The microtubule poison vinorelbine kills cells independently of mitotic arrest and targets cells lacking the APC tumour suppressor more effectively

Colorectal cancers commonly carry truncation mutations in the adenomatous polyposis coli gene. APC protein plays a role in microtubule stability, and loss of APC directly decreases microtubule stability. This change is at least partly responsible for the mitotic and apoptotic defects in APC-deficient cells (Dikovskaya et al., 2007; Fodde et al., 2001; Kita et al., 2006; Kroboth et al., 2007). Nevertheless, APC deficiency does not render cells more sensitive to death induced by the microtubule poisons taxol and nocodazole, largely because these cells have a partially compromised mitotic checkpoint (Dikovskaya et al., 2007). To determine whether the increased sensitivity of microtubules in APC-deficient cells could be exploited nevertheless, we further explored the effect of microtubule poisons on cells and tissue with different APC status.

Microtubule poisons are commonly used as therapeutic agents for cancer (Ellis et al., 1999). Their mode of action relies on changing microtubule dynamics, which disrupts formation of mitotic spindles. Taxol stabilizes microtubules and prevents microtubule depolymerization, whereas nocodazole and vinca alkaloids, such as vinorelbine, prevent polymerization (Beswick et al., 2006; Verdier-Pinard et al., 1999). In each case, loss of dynamic instability results in lack of normal microtubule function, which can lead to cell death following prolonged mitotic arrest (Gascoigne and Taylor, 2008; Varetti and Musacchio, 2008). Therefore, these drugs kill highly proliferative cancerous cells (Jordan and Wilson, 2004). However, this might only partially explain the therapeutic effect of these agents (Letai, 2008).

Although both nocodazole and vinorelbine lead to loss of polymerized microtubules, their modes of action, efficacy and toxicity differ (Perez, 2009; Uppal et al., 2007). Nocodazole induces mitotic arrest by activating the spindle assembly checkpoint (SAC) (Beswick et al., 2006; Varetti and Musacchio, 2008). Given that nocodazole impairs microtubule–kinetochore attachment, cells arrest in metaphase owing to an active SAC. Most commonly, cell death follows prolonged mitosis (Gascoigne and Taylor, 2008; Huang et al., 2009) but premature mitotic exit into tetraploid G1 can occur as a result of inappropriately decreased cyclin B1 levels, even in the presence of an active SAC. Vinorelbine was thought to act similarly to nocodazole by arresting cells in mitosis (Gonzalez-Cid et al., 1997). Although both drugs cause depolymerization of microtubules, vinorelbine induces the formation of tubulin paracrystals (Jean-Decoster et al., 1999), and might also downregulate the anti-apoptotic protein BCL2 in a p53-dependent manner (Bourgarel-Rey et al., 2009). These effects might explain the efficacy of vinorelbine as a chemotherapeutic agent in some cancers (Conroy, 2002).

Colorectal cancers commonly carry truncation mutations in the APC gene (Kinzler et al., 1991). The connection between APC, colorectal cancers and microtubules, together with the paucity of chemotherapy available for this disease prompted us to explore the mechanisms of vinorelbine-induced cell death (Zumbrunn et al., 2001). Loss of APC directly decreases microtubule stability and this change is at least partly responsible for the mitotic defects in APC-deficient cells (Dikovskaya et al., 2007; Fodde et al., 2001; Kita et al., 2006; Kroboth et al., 2007).

We found that vinorelbine, unlike nocodazole, induces significant cell death during interphase, suggesting that its activity is independent of prolonged mitotic arrest. Consistent with this finding, vinorelbine-induced apoptosis was lower in cells depleted of p53. Importantly, vinorelbine-induced apoptosis was higher in APC-depleted cells, even in cells lacking p53. Cell death induced by vinorelbine was accompanied by an increase in BIM (also known as BCL2L11) protein at mitochondria. BIM inhibits BCL2, so the recruitment of BIM to mitochondria correlates well with the increased apoptosis induced by vinorelbine in APC-deficient cells. This might also explain how vinorelbine induces cell death during interphase without mitotic arrest (Puthalakath et al., 1999). Consistent with the idea that enhanced recruitment of BIM is involved in the sensitivity of APC-deficient cells to vinorelbine, depletion of BIM decreased the sensitivity of APC-deficient cells.

These findings suggest that vinorelbine could be a useful chemotherapeutic agent for the treatment of colorectal cancer.

The extremely common lack of fully functional APC in colorectal cancers makes it attractive to exploit selective defects of such cells for therapy. Unlike the tumour suppressor p53, which contributes to apoptosis during interphase and in response to prolonged activation of the SAC (Castedo et al., 2004; Chi et al., 2009), APC has only been shown to contribute to the latter, and cells depleted of APC are more resistant to apoptosis induced by prolonged SAC activation (Chen et al., 2003; Dikovskaya et al., 2007). To explore further the contribution of APC status to cell killing by microtubule poisons, we tested how APC-deficient cells responded to vinorelbine in therapeutically relevant doses (Degardin et al., 1994) (Fig. 1A). We found that cell death induced by vinorelbine was more pronounced in cells lacking APC, indicated by the increased number of cells containing active caspase-3 (aCasp3) after 4 hours of vinorelbine treatment at a range of concentrations (Fig. 1A). This rapid response to vinrorelbine suggested that death did not involve mitotic arrest.

Microtubule poisons are generally thought to kill cells as a result of prolonged mitotic arrest (Jordan and Kamath, 2007); however, APC-deficiency has previously been show to protect against this process (Dikovskaya et al., 2007). To determine whether vinorelbine induces cell death during interphase or mitosis, we arrested cells in G1 with thymidine and treated them with vinorelbine or nocodazole (Fig. 1C,D). We measured aCasp3 every hour for 4 hours. After 4 hours, 26% of cells treated with vinorelbine were aCasp3 positive, whereas nocodazole treatment caused no change in the percentage of aCasp3-positive cells (Fig. 1C). Cells did not enter mitosis during this short experiment, as indicated by the lack of cells with 4N DNA content (Fig. 1D). This suggested that vinorelbine, but not nocodazole, induces apoptosis during interphase in G1.

We used RNA interference (RNAi) to deplete APC and p53 and routinely achieved 80–90% reduction in protein levels as shown by immunoblotting (Fig. 1B, 3A, 4A). Similar depletion was achieved in all subsequent experiments.

The experiments described above only revealed snapshots in time of cell populations (Fig. 1) and could not completely differentiate between cells dying during mitosis or cells dying after a failed mitosis, such as those resulting in tetraploidy or aneuploidy (Dikovskaya et al., 2007). To determine more precisely at what point during the cell cycle cells in an asynchronous population died when treated with vinorelbine, we monitored cells using time-lapse movies (Fig. 2; supplementary material Fig. S1B,C). To identify apoptotic cells, we used Nucview, a fluorescent reporter for active caspase-3 (Cen et al., 2008). In parallel, we used mCherry-labeled lamin B1 to report on the integrity of the nuclear envelope as a marker for interphase cells (Liu et al., 2003). Given that p53 is an important player in inducing apoptosis and is commonly mutated in many cancers, including colorectal cancer, we also compared the contribution of p53 to these responses by depleting p53 with small interfering RNA (siRNA). First, we measured the cell cycle distribution at the beginning of each recording, which started 2 hours after drug treatment began, and found a similar cell cycle distribution in vinorelbine and nocodazole-treated cells (15–27% of cells in mitosis for vinorelbine, and 25–37% for nocodazole) (Fig. 2A). The proportion of mitotic cells was not significantly different between these populations (P=0.09) (Fig. 2A).

After treatment with vinorelbine, the cell cycle distribution of cells when they entered apoptosis did not differ from the cell cycle distribution of cells at the beginning of the movies, indicating that vinorelbine-induced apoptosis was independent of the cell cycle (Fig. 2A,B; supplementary material Movies 1–6). Nocodazole, like vinorelbine, depolymerizes microtubules but induces cell death primarily during mitosis (Beswick et al., 2006; Brito and Rieder, 2009) and not in interphase (Fig. 1C). Therefore, nocodazole acted as a control for apoptosis induced by prolonged SAC activation (Fig. 2B). Indeed, in our assay we confirmed that cells predominantly died during mitosis, and not in interphase, when treated with nocodazole. This was in stark contrast to results with vinorelbine treatment: when treated with nocodazole only 13–26% of cells entered apoptosis during interphase versus 70–82% for vinorelbine (P<0.0001) (Fig. 2B). Importantly, our recordings showed that, in nearly all cases, apoptosis that occurred during interphase did not follow mitosis. Thus, vinorelbine appeared to kill cells independently of the cell cycle because the likelihood of a cell entering apoptosis was independent of its cell cycle stage at the beginning of the recording (P=0.26) (Fig. 2A,B) and this was independent of APC or p53 status.

Because vinorelbine induced cell death during interphase, we predicted that this drug would kill more cells than nocodazole. This was indeed the case, as shown by the higher overall incidence of cell death after vinorelbine treatment (Fig. 2B). Over the same time period, vinorelbine induced cell death in at least 90% of all cells, whereas nocodazole induced cell death in only 54% of cells, at most, during the 24-hour time-lapse movies (Fig. 2B). Under the same conditions, there was little cell death in cells not exposed to microtubule poisons, regardless of APC or p53 status (data not shown).

Nocodazole and vinorelbine depolymerize microtubules, but only vinorelbine causes the formation of tubulin paracrystals (supplementary material Fig. S2), which have been well characterized (Gigant et al., 2005). This prompted us to test whether differences in responses by microtubule-associated proteins could explain vinorelbine-induced cell death. Specifically, we examined the previously reported microtubule-associated pro-apoptotic protein BIM. BIM inhibits the anti-apoptotic protein BCL2 and also has been reported to be associated with microtubules in a manner dependent on dynein light chain 8 (Puthalakath et al., 1999). Importantly, in response to apoptotic stimuli, BIM translocates to mitochondria, where it can locally inhibit BCL2 (Puthalakath et al., 1999).

The role of BIM as a mediator of apoptosis and its association with microtubules prompted us to investigate whether recruitment of BIM to mitochondria was an early event in vinorelbine-induced apoptosis, and whether APC or p53 status affected such response (Fig. 3). We were unable to detect BIM at microtubules by using immunofluorescence and BIM was not recruited to vinorelbine-induced tubulin paracrystals (supplementary material Fig. S3A). However, BIM protein was slightly more abundant at mitochondria after treatment with vinorelbine but not nocodazole (Fig. 3). Importantly, in each case, the accumulation of BIM at mitochondria was significantly greater in cells lacking APC regardless of p53 status (Fig. 3). Depleting p53 on its own did not have a measurable effect on the localization of BIM to mitochondria (Fig. 3). The increase in mitochondria-associated BIM protein in APC-deficient cells is consistent with the idea that the altered microtubule dynamics in APC-deficient cells changes the ability of BIM to be recruited to mitochondria (Kroboth et al., 2007) (Fig. 3A,B). An increase in total BIM protein could not explain these observations because total BIM protein was reduced in APC deficient cells (Fig. 4A). Our data suggest that the recruitment of BIM to mitochondria contributes to vinorelbine-induced cell death particularly in APC-deficient cells.

The overall ability of vinorelbine to induce cell death during interphase correlated well with the recruitment of BIM to mitochondria: cells lacking APC were killed more readily and recruited more BIM to mitochondria (Figs 1, 3). p53 did not affect this recruitment and protected cells instead, suggesting that p53 acts independently of BIM (Figs 3, 4). When treated with 100 μg/ml vinorelbine, more than 37.3% of APC-depleted cells were positive for aCasp3 after 4 hours, compared with 23.3% in control cells. This was slightly enhanced, although not significantly different (P=0.42), in cells depleted of both APC and p53 (39.0%) (Fig. 4C). Depleting p53 alone reduced the number of aCasp3-positive cells to 18.8% (Fig. 4C). Together, these results suggest that APC depletion has a dominant effect over p53 depletion in enhancing vinorelbine-induced cell death. Interestingly, compared with control cells, p53-negative cells showed a lower basal level of apoptosis in the absence of vinorelbine (P<0.03), whereas APC-, and p53 and APC double-negative cells did not (Fig. 1A,B).

Furthermore, BIM depletion dampened vinorelbine-induced apoptosis in APC-deficient cells, confirming that it contributes to apoptosis induced by vinorelbine (Fig. 4C). When BIM was depleted, there was no significant difference in the number of aCasp3-positive cells between either control or p53 deficient cells (17.0%) and cells deficient in both APC and p53 (20.4%) (P>0.05, Fig. 4C).

Together these data suggest that the amount of BIM at mitochondria in response to the combined effect of vinorelbine and APC depletion exceeds the threshold that causes sufficient BCL2 inhibition to induce death. In this context, it is important to note that the increased recruitment of BIM to mitochondria does not depend on changes in β-catenin, given that HCT116 colorectal tumour cells, which express mutant β-catenin that is not degraded in response to APC depletion, also exhibit increased sensitivity to vinorelbine and have more BIM at mitochondria when APC is depleted (supplementary material Fig. S4; data not shown).

A correlation between BIM recruitment to mitochondria and interphase cell death is supported by our finding that nocodazole, which kills cells consequent to failed mitosis (Fig. 2B), did not induce BIM recruitment to mitochondria in control and APC-deficient cells (Fig. 3B). This observation also showed that recruitment of BIM to mitochondria is not merely a consequence of microtubule depolymerization (Fig. 3B). Whether the tubulin paracrystals induced by vinorelbine or other changes in free tubulin heterodimers contribute to the recruitment of BIM to mitochondria will require future investigation.

To determine whether the enhanced sensitivity to vinorelbine in cultured cells lacking APC was also true for cells in tissue, we administered vinorelbine to APC^Min/+ mutant mice with adenoma (which lack full-length APC) and compared the response of the adenoma tissue with that of physiologically normal tissue (which retains the wild-type APC allele) (Fig. 5). To determine the relationship between mitotic arrest and vinorelbine-induced death in cells in situ, we counted the number of apoptotic cells in the intestine of mice at various times after they were injected with vinorelbine or taxol (Fig. 5A,B). Just as in cultured cells, vinorelbine induced cell death in intestinal epithelial cells within 30 minutes after treatment, whereas tissue from taxol-treated animals only contained increased numbers of apoptotic cells 6 hours after treatment (Fig. 5A,B). Furthermore, in taxol-treated tissue, the appearance of apoptotic cells only occurred after mitotically arrested cells had accumulated, confirming that killing by taxol, but not vinorelbine, follows mitotic arrest (Fig. 5A,B). Consistent with the increased response in APC-deficient cells, vinorelbine induced apoptosis over threefold more efficiently in adenoma than in normal tissue when corresponding regions in the same animal were compared (Fig. 5C,D). Consistent with a defect in mitotic arrest in APC-deficient cells, cells in normal tissue arrested more efficiently in mitosis than cells in adenoma (Fig. 5E). Importantly, the increased sensitivity of APC-deficient cells to vinorelbine translated into a significantly reduced adenoma size in APC^Min/+ mice (P=0.0001) (Fig. 5F).

These results strongly suggest that vinorelbine could be a useful treatment for colorectal cancers with APC mutations.

The effectiveness of currently available chemotherapy for advanced colorectal cancer is limited, making the identification of suitable targets for this common human cancer an important goal (Segal and Saltz, 2009). Because most colorectal cancer cells carry APC mutations, and in line with the concomitant decrease in microtubule stability, we considered the possibility that treatment of colorectal cancer with microtubule poisons is a potentially useful approach.

However, we had previously found that APC-deficient cells died less readily than cells with normal levels of APC, owing to impaired mitotic arrest, when treated with taxol and nocodazole (Dikovskaya et al., 2007). This encouraged us to investigate further how another microtubule poison vinorelbine, which is currently used as a chemotherapeutic drug, produces cell death. To the best of our knowledge, vinorelbine has not been considered for therapy in colorectal cancer. Here, we report the potential usefulness of vinorelbine as chemotherapeutic for APC-deficient and colorectal cancer cells.

We discovered that vinorelbine kills cells independently of the cell cycle, showing for the first time that vinorelbine-induced cell death is not simply a consequence of prolonged mitotic arrest (Figs 1, 2, 4, 5). This was true in tissue in mice treated with vinorelbine (Fig. 5) and in cells (Figs 1, 2, 4). Time-lapse movies showed that cells enter apoptosis as a result of vinorelbine treatment during the cell cycle stage in which they first encountered the drug (Fig. 2). Our analysis showed that interphase death did not follow mitotic slippage (Fig. 2; supplementary material Fig. S1). These findings demonstrate that vinorelbine can induce cell death during interphase and does not rely on mitotic arrest.

We chose to examine the contributions of two of the most prevalent colorectal cancer mutations, APC and p53, to this process (Figs 2, 3 and 4). APC mutations are early events in tumorigenesis, whereas mutations in p53 tend to occur later. Both have also been implicated in death following mitotic arrest (Chi et al., 2009; Dikovskaya et al., 2007). Our results showed that loss of APC (depletion or mutation) renders cells more sensitive to vinorelbine-induced apoptosis (Fig. 1A; supplementary material Fig. S4A), whereas loss of p53 was slightly protective, suggesting that APC acts either downstream or independently of p53, given that p53 loss had no effect on the phenotype resulting from APC loss (Figs 2,4). Our finding that two relatively well-differentiated colorectal tumour cells lines (DLD1 and Caco-2) are killed more effectively by vinorelbine compared with cells expressing wild-type APC further supports our idea that APC mutant cells could be potentially selectively targeted by this agent. By contrast, more heterogeneous, generally less-differentiated SW480 cells responded to vinorelbine in a similar manner to cells expressing wild-type APC (supplementary material Fig. S4A). This suggests that more advanced tumours might not retain sensitivity to vinorelbine. However, a more detailed analysis will be required to identify all the required parameters that govern and predict sensitivity to vinorelbine.

Some p53 and APC protein co-fractionated with mitochondria, suggesting that their anti-apoptotic function might involve mitochondrial processes (Brocardo et al., 2008; Marchenko et al., 2000). To examine the relationship between the changes induced by microtubule poisons and the mitochondrial recruitment of known apoptotic regulators, we chose to investigate BIM in this context (Figs 3, 4). BIM has been described as a microtubule-bound, pro-apoptotic factor that is recruited to mitochondria to induce cell death through the inhibition of the key anti-apoptotic protein BCL2 (Merino et al., 2009; Puthalakath et al., 1999).

Vinorelbine treatment on its own induced some recruitment of BIM to mitochondria in control cells, whereas nocodazole did not (Fig. 3B). Most importantly, this effect of vinorelbine was significantly increased in APC-depleted cells but not in cells lacking p53 (Fig. 3A,B). Overall cellular BIM protein levels were not affected by altered APC expression (Fig. 4A). This suggested that APC-deficiency might prime cells for apoptosis by increasing the amount of BIM at mitochondria as a response to microtubule stress. These and previously published data suggest that the reduced microtubule stability in APC mutant cells enhances recruitment of BIM to mitochondria, possibly because of a reduction in the association of BIM with microtubules. Alternatively, APC could affect the association of BIM with mitochondrial proteins. The idea that other APC interactions contribute to the effects we observe is supported by our finding that cells that express only truncated fragments of APC, which are not thought to contribute to microtubule stability, are also more efficiently killed by vinorelbine when the truncated APC fragment is removed (supplementary material Fig. S4B). For instance, DLD1 and SW480 cells are killed more readily when APC has been depleted (supplementary material Fig. S4B). In HCT116 and SW480 cells, the increased killing is accompanied by increased BIM at mitochondria (data not shown). These data suggest that domains in N-terminal regions of APC directly contribute to events at mitochondria that affect apoptosis, consistent with the previous findings that N-terminal fragments of APC localize to mitochondria in tumour cells (Brocardo et al., 2008). These data, together with the increased sensitivity of colorectal tumour cells to vinorelbine, also suggest that changes in β-catenin or Wnt signalling are not required for the sensitivity to vinorelbine.

Another observation was that vinorelbine induced an increase in the apparent molecular weight of BIM, consistent with its phosphorylation. This was reduced in APC- and/or p53-deficient cells (Fig. 4A), suggesting that this post-translational modification has a protective role. However, in the absence of additional detailed information about the role of post-translational modification for the function of BIM, this is purely speculative.

Our data show that vinorelbine kills cells by different mechanisms than that of other microtubule poisons. Vinorelbine induces tubulin paracrystals and acts on interphase microtubules, which might be linked to this phenomenon (Lobert et al., 1998; Verdier-Pinard et al., 1999; Jean-Decoster et al., 1999) (supplementary material Fig. S2). Therefore we examined the association of microtubule-binding proteins with tubulin paracrystals in vinorelbine-treated cells. The paracrystals colocalized with acetylated tubulin, MAP2, EB1 (also known as MAPRE1), Clip170 (also known as CLIP1) and APC, only a little with CLASP1, and not at all with BCL2 and BIM (supplementary material Figs S2, S3). This raises the possibility that paracrystals sequester a subset of microtubule-binding proteins, competing them away from other locations in the cell, to invoke cellular changes that could promote cell death. Another possibility is that different modifications of tubulin itself contribute to the difference between the actions of vinorelbine and other microtubule poisons (supplementary material Fig. S2). For instance, a cellular sensor for free tubulin might be activated in cells treated with nocodazole, but not vinorelbine, owing to the sequestration of tubulin into paracrystals by the latter but not the former. How cells sense free versus polymerized tubulin remains unclear. Vinca alkaloids induce transcription of β3 tubulin (Bachurski et al., 1994; Saussede-Aim et al., 2009), and the ability of β-tubulin to regulate the stability of its mRNA (Bachurski et al., 1994) might be affected by vinorelbine. This might, in turn, have consequences that contribute to cell death and other cellular responses.

Importantly, the increased sensitivity of APC mutant cells to vinorelbine was measurable in murine adenomas lacking APC and in colorectal tumour cells (Fig. 5; supplementary material Fig. S4B). This observation reveals vinorelbine as a potentially useful chemotherapeutic for colorectal cancer. The details of how changes in APC expression contribute to the sensitivity of cells to vinorelbine, and how loss of p53 attenuates this effect, will require further investigation. Nevertheless, an important conclusion from our data is that the genotype of cells for these two commonly mutated tumour suppressor proteins, APC and p53, could be crucial when deciding how to treat colorectal tumours and has important implications for personalized cancer therapy.

Human osteosarcoma (U2OS) cells were obtained from Cancer Research UK, cultured in DMEM supplemented with 10% FBS, 1% penicillin–streptomycin stock solution (Life Technologies) and nonessential amino acids at a dilution of 1:100 (Sigma-Aldrich). For siRNA-mediated protein depletion, U2OS cells were transfected using INTERFERin (Polyplus) according to the manufacturer's instructions with 10 nM siRNAs targeting human APC and/or p53 (Qiagen), or a non-targeted control (Dharmacon) as described previously (Dikovskaya et al., 2007). For live-cell fluorescence markers, U2OS cells were co-transfected with 0.2 μg DNA encoding mCherry-tagged lamin B1 along with 10 nM APC and/or p53 siRNAs (Dharmacon, Qiagen) using Fugene 6 (Roche) according to the manufacturer's instructions. NucView (Biotium) was added to medium to detect caspase-3 activation as described previously (Cen et al., 2008). To induce mitotic arrest and cell death, U2OS cells were treated with 1.5 μg/ml nocodazole or various concentrations of vinorelbine (see figure legends). To determine BCL2 localization upon vinorelbine treatment, U2OS cells were co-transfected with 0.2 μg DNA encoding mCherry–tubulin (a kind gift from Markus Posch, University of Dundee, Dundee, UK) and 0.4 μg DNA encoding GFP–BCL2 (Wang et al., 2001).

Cells were fixed in a warm solution of 1.8% PFA in PHEM buffer (60 mM Pipes pH 6.9, 4 mM MgSO₄, 25 mM Hepes and 10 mM K-EGTA) for 15 minutes at 37°C, washed in PBS, and blocked in blocking buffer (0.1% Triton X-100, 2% BSA, 5% donkey serum, 0.02% NaN₃ in PBS supplemented with 50 mM NH₄Cl, for fixation only) for 1 hour. Primary antibodies against the following proteins were diluted in blocking buffer as follows: supernatant from hybridoma cells producing anti-tubulin antibody (YL1/2) at 1:25; BIM (AbCam) at 1:100; Clip170 at 1:300 [kindly provided by Niels Galjart (Tanenbaum et al., 2008), pericentrin (AbCam) at 1:2500, INCENP and KAP3 (BD Transduction Laboratories) at 1:200, EB1 (Transduction Laboratories) at 1:50, MAP2 (Sigma-Aldrich) was at 1:300; and anti-M-APC polyclonal (Näthke et al., 1996) at 1:500. For CLASP1 staining, cells were fixed in −20°C methanol for 10 minutes and then processed as described above for PFA fixation with no NH₄Cl in the blocking buffer. The antibody against CLASP1 (kind gift from Fedor Severin) was diluted 1:300 in blocking buffer. Secondary antibodies raised in donkey were conjugated to Alexa Fluor 488, Alexa Fluor 568 or Alexa Fluor 647 (Invitrogen), and were diluted 1:1000 in blocking buffer. Cells were counterstained with DAPI at 1 μg/ml for 20 minutes. High-resolution images were collected with a Delta-Vision Restoration imaging system (Applied Precision, Issaquah, WA) based on an Olympus IX70 inverted microscope stand, using a 100× 1.4NA objective lens. Images were acquired at 0.25 μm Z-intervals and recorded with CoolSnap CCD camera (Roper Scientific). Images were deconvolved and three-dimensionally rendered using SoftWoRx software (Applied Precision).

Lamin B1 and Nucview time-lapse movies were collected using a Nikon Eclipse Ti-E inverted microscope, a 40× 1.4 NA objective lens and a Photometrics Cascade EM-CCD camera. To acquire a time-lapse movie, an image was collected every 15 minutes for 24 hours, with the first time-point occurring 2 hours after 50 μg/ml vinorelbine treatment. Cells were kept at 37°C and under 5% CO₂ throughout the recording. Movies and images were analysed using ImageJ (http://rsb.info.nih.gov/ij/).

Mitochondrial isolation was performed using the AbCam Cytochrome C Releasing Apoptosis Assay Kit according to the manufacturer's instructions. Total protein concentration was measured using the Bradford Assay (BioRad) according to the manufacturer's instructions, to ensure equal protein loading on SDS-PAGE gels. We used the following proteins (the antibody is given in brackets) as loading controls: HSP60 (CST) for mitochondrial fractions, and tubulin (YL1/2), actin (C4 clone; MP Biomedicals) and glyceraldehyde 3-phosphate dehydrogenase (GAPDH; AbCam) for cytoplasmic proteins. Immunoblots were imaged and quantified with the LICOR Odyssey imaging system. To quantify protein amounts, the signal intensity for a protein was normalized to the signal intensity of an appropriate loading control in the same sample. Reported protein levels are relative to the control treatment, which was used as a reference in each individual experiment.

For total protein lysis, cells were lysed in MEBC buffer (50 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM EGTA, 5 mM EDTA, 0.5% NP-40, and 40 mM 2-glycerophosphate) supplemented with 10 μg/ml each of leuptin, pepstatin A and chymotrypsin (Peptide Institute, Inc.). Lysate was collected using cell scrapers and was then centrifuged (20 minutes, 20,000 g, 4°C). Supernatant was run on NuPage 4–12% gradient SDS-PAGE gels (Invitrogen) using MOPS running buffer (Invitrogen). The separated proteins were transferred onto a nitrocellulose membrane (Protran BA 79) with a 0.1 μm pore size (Whatman). Antibodies were diluted in blocking buffer (TBS containing 5% nonfat milk powder, 1% donkey serum and 0.02% Triton X-100) as follows: anti-APC and ALI mouse monoclonal antibody [against the N terminus; Cancer Research UK (Efstathiou et al., 1998) or crude serum of rabbit polyclonal antibody raised against the N-terminus of APC (anti-N-APC antibody) (Midgley et al., 1997), 1:1000; anti-α-tubulin rat monoclonal antibody (YL1/2), 1:100; anti-actin antibody, 1:500; anti-GAPDH antibody (AbCam), 1:2000; anti-BIM (AbCam) mouse monoclonal antibody, 1:1000; anti-p53 DO-1 mouse monoclonal antibody, 1:1000; and anti-HSP60 (CST) antibody, 1:1000. Secondary anti-rabbit, -mouse or -rat IRDye-800/680-conjugated antibodies (Rockland and Invitrogen) were diluted 1:5000 in blocking buffer.

To detect aCasp3, cells were fixed and stained using the aCasp3 phycoerythrin (PE) monoclonal antibody apoptosis kit (BD Biosciences) according to the manufacturer's instructions. In brief, cells were collected and fixed in 1% PFA on ice for 20 minutes and then washed in Saponin solution. The cell profile was analyzed on a flow cytometer (FACSCalibur; Becton Dickinson) using CellQuest Pro software (BD Biosciences). Data analysis was performed using Flowjo software (Tree Star).

Statistical significance between sample means was calculated using the Student's t test (using Graphpad Prism V5.0b).

Wild type or APC^Min/+ mice were injected intravenously with 10 mg per kg of body weight vinorelbine or 10 mg per kg of body weight taxol intraperitoneally (Radulescu et al., 2010). Tissue was harvested at the indicated times for the timecourse comparison (Fig. 3). Tissue was prepared and apoptosis and mitotic figures scored as previously described (Sansom et al., 2004; Cole et al., 2010) for five tumours each in APC^Min/+ mouse or for 25 crypts in three wild-type controls. To measure tumour size in APC^Min/+ mice, animals (45–50 days old) were injected once every 2 weeks with 10 mg per kg of body weight vinorelbine or vehicle. Tissue was harvested at 90 days and prepared for H&E staining, and the size of the indicated number of tumours was measured as described previously (Radulescu et al., 2010). All animal experiments were performed under the UK Home Office Guidelines.

We are grateful to all those who supplied reagents and have indicated this in the text. We thank R. J. S. Steele (Molecular Surgery, University of Dundee) for critical reading of the manuscript and members of the Näthke lab for helpful discussions.