Aspirin blocks proliferation in colon cells by inducing a G1 arrest and apoptosis through activation of the checkpoint kinase ATM


Colorectal cancer (CRC) is the most common gastrointestinal malignancy. Most of the clinical data on CRC prevention have come from the use of aspirin. Besides inhibition of cyclooxygenases, aspirin has a diversity of molecular effects that counteract colon carcinogenesis. Aspirin restrains cell proliferation by inducing a G1 arrest in colorectal cells. To determine which cell cycle checkpoint pathways are involved in this response, colorectal cell lines wild-type or defective for p53 and p21Waf1/Cip1 were treated with aspirin or the anti-proliferative drug sulindac sulfide, then assayed for proliferative activity, for cell cycle progression and apoptosis, for the activation and phosphorylation of checkpoint components and for the transcriptional up-regulation of p21Waf1/Cip1 and Bax. Aspirin and sulindac sulfide induced a G1 arrest within 48 h. While all cell lines responded in a comparable way to sulindac sulfide, the aspirin-induced G1 arrest was dependent on p21Waf1/Cip1—as cells lacking the cyclin-dependent kinase inhibitor failed to show this arrest—and on ataxia-telangiectasia-mutated kinase (ATM)—as the inhibitor caffeine abrogated the checkpoint. Moreover, aspirin induced cell death mainly in cells expressing p53. Aspirin induced the phosphorylation of p53 at residue Ser15 within 8 h in a caffeine-dependent manner, and also caused the activation of checkpoint kinase 2 and the cleavage of caspase 7. Our results suggest that aspirin induces a G1 arrest and apoptosis by activating p53 and p21Waf1/Cip1 in an ATM-dependent way. By activating these checkpoint pathways, aspirin may restrain uncontrolled proliferation of colorectal cells, enhance their response to stresses such as DNA damage and promote entry of abnormal cells into apoptosis.



Cancer chemoprevention is understood as the use of natural, synthetic or chemical agents to reverse, suppress or prevent carcinogenic progression to invasive cancer. Findings from epidemiological and clinical studies as well as animal models of colon carcinogenesis imply that some non-steroidal anti-inflammatory drugs (NSAIDs), salicylate derivatives and inhibitors of the cyclooxygenase (COX)-2 may be effective against the development of colorectal cancer (CRC) (1). Pharmacologically, NSAIDs inhibit COXs in various cell types and tissues. COX-2 inhibition is believed to underlie the chemopreventive effect of NSAIDs possibly through reducing cell proliferation, inducing apoptosis or modulating angiogenesis. NSAIDs inhibit the proliferation rate, alter the cell cycle distribution and induce apoptosis in colon cancer cell lines. Among the most studied agents of the COX-2 inhibitor group of compounds are aspirin and sulindac and derivatives (2,3). Salicylates were shown to have a chemopreventive role in CRC through COX-independent mechanisms, such as inhibition of the transcription factor nuclear factor kappa B (NF-κB) (4). Similar lines of research suggested that aspirin alters the expression of genes involved in mismatch repair and cell cycle progression (5). All these pieces of experimental evidence point to the interference of aspirin with cell cycle progression, although the biochemical pathways involved are still obscure.

To maintain genome stability and monitor the structure of chromosomes, eukaryotic cells have evolved surveillance mechanisms called cell cycle checkpoints that block cell cycle progression at specific stages to allow the cell to recover from the impairment (6). Checkpoint pathways comprise damage sensors, signal transducers and effectors. Ataxia-telangiectasia-mutated kinase (ATM) and ATM- and Rad3-related kinase (ATR) are phosphatidylinositol-3-like kinases central to the DNA damage response, which can cause activation of checkpoints, DNA repair and apoptosis (7). ATM is the protein product of the gene mutated in the multisystem disorder ataxia-telangiectasia, which is characterized by neuronal degeneration, immunodeficiency, chromosomal instability and cancer predisposition (8). The ATM pathway responds to the presence of double-strand breaks (DSBs) and can be activated during all phases of the cell cycle. In the absence of a checkpoint mechanism, premature progression through the cell cycle can be either lethal to the cell or result in oncogenic transformation. A hallmark of the transformed state of tumor cells is incompetent checkpoint control, resulting in accumulation of mutations and genetic abnormalities (9). When checkpoint control is compromised, initiation of S phase, or the onset of mitosis, occurs despite cellular damage, and the ensuing genetic instability may lead to the emergence of a malignant clone. Cells in which checkpoint control is disrupted are susceptible to the accumulation of additional genotoxic damage (10).

The tumor suppressor protein p53 is the major known mediator of the checkpoint-induced arrest in the G1 phase of the cell cycle. A variety of cellular stresses including DNA damage, hypoxia, nucleotide depletion, viral infection and cytokine-activated signaling pathways transiently stabilize the p53 protein, cause it to accumulate in the nucleus and activate it as a transcription factor (11). P53 induces cell cycle arrest, through the cyclin-dependent kinase (CDK) inhibitor p21Waf1/Cip1 (12), preventing the replication of damaged DNA. Alternatively, p53 can promote apoptosis, mainly through Bax (13), which is important for eliminating defective cells. Mutations in the p53 gene occur in half of all human cancers, and regulation of the protein is defective in a variety of others (14). Stabilization of p53 in response to ionizing radiation and ultraviolet light is dependent on the ATM and ATR kinases, respectively (15). ATM is required for homologous recombination and cell cycle checkpoint activation after DNA damage (16). In addition, germline mutations of ATM lead to radiation hypersensitivity, growth retardation, immunodeficiency and greatly increased cancer risk (17). Upon activation, ATM undergoes Ser1981 auto-phosphorylation and dimer separation (18). As a protein kinase, ATM functions by phosphorylating and activating a number of DNA repair and checkpoint proteins including p53 and checkpoint kinase 2 (Chk2), and it has been established previously that ATM is required to activate a p53-dependent cell cycle G1 arrest upon DNA damage (19). ATM and ATR both directly phosphorylate p53 in vivo on Ser15 and Ser37 causing stabilization (20). The ATM and ATR substrate Chk2 phosphorylates p53 on Ser20, which regulates the binding to and degradation by Mdm2 and thereby stabilizes p53 (21).

Our present study is focused on understanding the mechanisms by which aspirin exerts its control over cell cycle progression and activates cellular checkpoints in colon cells. We show that aspirin induces a caffeine-sensitive G1 arrest, which is dependent on p21Waf1/Cip1 and ATM, and cell death in cells expressing p53. Moreover, aspirin leads to phosphorylation of p53, γ-H2AX and Chk2 in an ATM-dependent manner, indicating that it activates a cellular checkpoint pathway including these proteins.

Materials and methods

Cell lines

HT29 and HCT116 CRC cells were obtained from American Type Culture Collection. HCT116p53−/− cells (p53-null) and HCT116p21−/− (p21Waf1/Cip1-null) were generated in Dr Vogelstein's laboratory as described (22). Cells were grown as monolayers in Iscove's modified Dulbeco's medium (Gibco/Invitrogen, Vienna, Austria) containing 2 nM glutamine and 10% fetal bovine serum at 5% CO2.

Antibodies, plasmids and reagents

Aspirin (Sigma, Vienna, Austria) was dissolved in culture medium at 10 mM and pH adjusted to 7.2 with NaOH. Sulindac sulfide (Sigma) was dissolved in DMSO at 400 mM and sterile filtered. Caffeine (Sigma) was dissolved in H2O and used at 5 mM. Cells were treated at subconfluent densities. All experiments were performed in triplicate. Total cell lysates were obtained as described (23) and western blots were performed according to standard procedures (23). Antibodies used were as follows: monoclonal antibody anti-p53 DO7 (Calbiochem, Vienna, Austria); polyclonal antibody (pAb) anti-phopsho-p53 Ser15 (Cell Signaling, Frankfurt am Main, Germany); monoclonal antibody anti-p21Waf1/Cip1 (Cell Signaling); pAb for cleaved caspase 7 (Cell Signaling); monoclonal antibody anti-ATM (Abcam, Cambridge, UK) and pAb anti-phospho ATM Ser1981 (Upstate, Vienna, Austria); pAb anti-Chk2 and pAb anti-phospho Chk2 Thr68 (Cell Signaling) and pAb anti-H2AX and pAb anti-phospho γ-H2AX Ser139 (Abcam). When needed, HCT116p53−/− cells were transfected with 0.5–1 μg pcDNA-p53 coding for full-length p53 (24) or empty vector (Invitrogen). pGL3-p21-luciferase and pGL3-Bax-Luciferase constructs were a gift from M.Oren (Weizmann Institute of Science, Israel) and they are described elsewhere in the text. All transfections were performed with the Effectene reagent (Qiagen, Vienna, Austria) according to the manufacturer's instructions.

Cell proliferation assay

Cells were plated in 96-well microtiter plates at a density of 5000 cells per well in medium. Twenty-four hours later, cells were treated with aspirin (0–10 mM) or sulindac sulfide (0–400 μM); untreated cells served as controls. After treatment for 48 or 72 h, cell proliferation was assayed using methylthiazolyldiphenyl-tetrazolium bromide salt (MTT) as described previously (23).

Flow cytometry analysis

Cells treated with aspirin or sulindac sulfide for 48–72 h were harvested and the cell cycle distribution was analyzed as described previously (23).

Luciferase assay

Cells were seeded in a 6-well culture plate at a density of 105 cells per well and transfected with 0.5 μg of either pGL3-Basic-Vector (Promega, Mannheim, Germany; containing no promoter upstream of the Luciferase gene), pGL3-Control-Vector (Promega; containing the SV40 constitutive promoter upstream of the Luciferase gene) or pGL3-based plasmids upstream of the Luciferase coding sequence ([containing the promoters of the p21Waf1/Cip1 or the Bax genes] pGL3-p21-Luciferase and pGL3-Bax-Luciferase, respectively). More in details, pGL3-p21-Luciferase contains a 2.5 Kb stretch, corresponding to the promoter sequence of the human p21Waf1/Cip1 gene digested with HindIII and subcloned into the luciferase reporter vector pGL3-Basic-Vector and containing a tandem binding site for p53 (12); pGL3-Bax-Luciferase Bax luciferase containing the proximal promoter region of the human bax gene fragment -715/-317 from the Bax gene promoter subcloned into BglII–HindIII sites of the pGL3-Basic-Vector (25,26). Transfected cells were treated with aspirin (10 mM) for 24 h. Cells were harvested in Reagent Lysis Buffer (Promega) and the luciferase activity of the lysates was measured using the Luciferase Assay System kit (Promega) following the manufacturer's instructions. Briefly, 50 mg of whole-cell lysates were mixed with 100 ml of reconstituted Luciferase Assay Substrate, and the light emitted was measured with a Lumat LB 9507 (Berthold, Vienna, Austria) tube luminometer. Relative units of light were defined by the formula: relative units of light = specific units of light/(control vector units basic vector units).


Proliferation inhibition by MTT was compared between different cell lines using the Student's T-test at a representative concentration. A P value below 0.05 was considered significant.


Aspirin reduces the proliferation of human colon cancer cell lines

In order to investigate the proliferation inhibitory effect of aspirin on colon cells, HCT116 and HT29 were treated with aspirin (0–10 mM) for 72 h. These cell lines have been chosen on the basis of their different expression of COX-2, p53 and downstream genes (Figure S1, supplementary data are available at Carcinogenesis online). In fact, HCT116 do not express COX-2 (27), but bear a wild-type form of p53, whereas HT29 express COX-2 but have a mutated p53 (p53R273H). Cellular proliferation was assessed by MTT assay. Aspirin treatment induced a concentration-dependent reduction in the proliferation rate of both cell lines (Figure 1A). However, 72 h after treatment, the number of viable cells was higher in the HT29 cell line than in HCT116 (e.g. at 0.62 mM aspirin: 68.2 ± 4.2 versus 45.6 ± 1.6% of control, P < 0.0001 or at 2.5 mM: 35.6 ± 1.7 versus 23.6 ± 1.0; P < 0.0001). As HT29 cells express a mutant form of p53 (p53R273H), we wished to investigate whether the difference in response of the 2 cell lines was due to an active p53 pathway. The p53-null HCT116p53−/− is a cell line isogenetic to the wild-type clone HCT116 created by targeted homologous recombination [Figure S1 and (28)]. HCT116 cells and HCT116p53−/− were incubated in the presence of 0, 2.5, 5 and 10 mM aspirin for 72 h (Figure 1B). We found that HCT116p53−/− cells were more resistant to aspirin treatment than HCT116 (p53-wt) (e.g. at 2.5 mM aspirin 47.6 ± 2.1 versus 31.9 ± 2.4% of control; P < 0.0001); moreover, re-introduction of p53 into HCT116p53−/− cells via transient transfection (Figure 1D) increased their sensitivity to aspirin treatment to an extent similar to that of HCT116 cells (Figure 1C). All these results point to a role of p53 for aspirin-mediated changes in proliferation and/or cell death. The results shown in Figure 1A and B have been normalized to the respective untreated control to account for the differences in the growth rate of the various colon cell lines. The corresponding non-normalized graphs, as well as the one in Figure 2B described here below, are shown as supplementary data are available at Carcinogenesis online (Figure S2A–D).

Aspirin induces G1 arrest and cell death in colon cells

It has been reported previously that aspirin causes cell cycle arrest and induces apoptosis in a number of cell lines (5,29). In order to establish whether the reduction in proliferation of colon cells was due to aspirin-induced changes in cell cycle progression, HT29, HCT116 and HCT116p53−/− were cultured in the presence of 2.5, 5, or 10 mM aspirin for 48 h and the cell cycle distribution was analyzed by flow cytometry. Upon treatment with 5 mM aspirin (Figure 2A), HCT116 cells arrested in G1 almost completely (82% of the cell population) within 48 h, whereas HCT116p53−/− and HT29 cells showed a less complete arrest (62 and 71%, respectively). High aspirin concentrations (10 mM) affected cell survival in HCT116 and HT29 but less so in HCT116p53−/− cells; the latter cell line arrested in both G1 and G2 (Figure 2A). The observed increase in the G1 population parallels the changes in the MTT experiments with higher aspirin sensitivity displayed in HCT116 than in HT29 or HCT116p53−/− cells. The strongest effect on G1 arrest was observed at 5 mM aspirin. As a control, cells were treated with 100 μM sulindac sulfide (at its IC50 concentration, as assessed by MTT assay on HCT116 cells) for 48 h (Figure 2B), which has been described to reduce the proliferation rate of HT-29 cells by inducing G1 arrest (30). We found that all three cell lines responded to sulindac sulfide to a similar extent (G1 population: HCT116 65%, HCT116p53−/− 59% and HT29 68%). These results suggest that p53 only partially controls the aspirin-mediated (but not sulindac sulfide mediated) G1 arrest and at higher concentrations the aspirin-induced cell death. However, cells lacking functional p53 were still able to activate checkpoint responses, which led to cell cycle arrest more in G1 and G2 and, to a lower extent, to apoptosis.

Aspirin activates p53 by phosphorylation at residue Ser15

The tumor suppressor p53 is known to play a key role in cell cycle arrest as well as apoptosis in response to various stresses such as DNA-damaging agents and anticancer drugs. Phosphorylation of p53 on Ser15 by the phosphatidylinositol-3-like kinases ATM and ATR triggers post-translational modifications that contribute to p53 stabilization. Thus, we examined whether p53 protein levels and phosphorylation were altered by aspirin treatment in our system. HT29 and HCT116 cells were exposed to 0, 2.5, 5 or 10 mM aspirin for 8 h. Analysis of total cell lysates indicated a concentration-dependent increase in phosphorylation of p53 on Ser15 in both cell lines but not of total protein levels (Figure 3A, top panels).

ATM is activated upon aspirin treatment in HCT116p53−/−cells

One of the major phosphatidylinositol-3 kinases that target p53 N-terminal residue Ser15 for phosphorylation is ATM (31). HCT116 and HT29 cells were treated with 5 or 10 mM aspirin for 8 h, in the presence or absence of caffeine, a known inhibitor of the ATM and ATR pathways (32). Phosphorylation of p53 at Ser15 decreased in a caffeine-dependent manner in HCT116 cells (Figure 3A, bottom panels) and in HT29 cells (data not shown), suggesting a role for ATM or ATR in the phosphorylation of p53 at Ser15. As we did not see activated ATR (i.e. chromatin-bound ATR; data not shown), we investigated the activation status of the ATM kinase by assessing its auto-phosphorylation at Ser1981 upon exposure to aspirin. A time course experiment revealed that ATM Ser1981 was phosphorylated within 2–4 h, with a peak at 8 h in HCT116 cells (Figure 3B). The kinetic of ATM activation paralleled that of the phosphorylation of the histone γ-H2AX at Ser139 (Figure 3B), a site targeted by ATM (19).

The checkpoint kinase Chk2 is activated upon aspirin treatment

As our data point to an activation of ATM by aspirin, we investigated further ATM substrates such as Chk2. Chk2 kinase plays a central role in modulating the cellular response to DNA damage, resulting in cell cycle arrest, DNA repair or apoptosis depending on the severity of the DNA damage and the cellular context. Chk2 is activated by phosphorylation on residue Thr68 by the kinase ATM (33). We treated HCT116, HCT116p53−/− and HT29 cells with 10 mM aspirin for 2, 4, 8 and 24 h. Although the kinase total levels were similar in all cell lines, Chk2 phosphorylation at Thr68 was detectable in p53-null cells as early as 2 h, whereas it was a much later event in the p53-wt cells (Figure 3C). In this time course experiment, p53 phosphorylation at Ser15 was visible within 8 h of aspirin treatment in HCT116 and HT29 cells and correlated with ATM activation. As expected, no p53 signal was detected in the HCT116p53−/− cells. The timing of phosphorylation of Chk2 and p53 is consistent with the activation of the ATM pathway.

ATM inhibitor caffeine blocks the cellular responses induced by aspirin

In order to further confirm the role of the checkpoint kinase ATM in the aspirin-induced responses, we repeated the treatment of colon cells HCT116 in the presence of caffeine, known to abrogate cellular checkpoint responses dependent on ATM and ATR (7). HCT116 cells were exposed to 5 or 10 mM aspirin for 48 h and 5 mM caffeine was added either at the same time as aspirin or 24 h later (Figure 3D). Analysis of cell cycle progression revealed that caffeine was able to prevent the activation of the aspirin-dependent G1 arrest and apoptosis; moreover, this effect was time dependent, as exposure to caffeine for 24 h only partially reversed both responses. Taken together, all these results indicate that the ATM checkpoint kinase mediates the effect of aspirin inhibition of proliferation in colon cells.

Aspirin-induced activation of caspase 7 is an early event

The cleavage and activation of caspase 7 is a well-established marker for the onset of apoptosis (34). When we tested whether the caspase cascade was activated in colorectal cell lines, we found that aspirin increased cleavage of caspase 7 (Figure 3C). Activation of p53 is known to mediate apoptosis, which has been observed upon treatment with aspirin or NSAIDs (5). Caspase 7 cleavage was more pronounced in the p53-wt cells and happened at an earlier time point (2 h). In the p53-mutant HT29 cells, the signal of caspase 7 cleavage was delayed but strong after 24 h, consistent with the pronounced cell death described earlier (Figure 2A), whereas in the HCT116p53−/− cells, caspase 7 cleavage was much weaker (also consistent with the lower extent of cell death shown in Figure 2A). These data suggest that aspirin induces apoptosis mostly in cells expressing either wild-type or mutant p53. However, a p53-independent mechanism must be responsible for the cleavage of caspase 7 (although weak) and cell death in p53-null cells.

Expression of the CDK inhibitor p21Waf1/Cip1 is essential for the aspirin-induced G1 arrest in colorectal cells

It is well established that p53 mediates cell cycle arrest in the G1 phase through transcriptional activation of p21Waf1/Cip1, an inhibitor of the CDKs 2, 3, 4 and 6. In HCT116 cells, an increase in p21Waf1/Cip1 levels was observed after 16 h treatment with aspirin (Figure 3C). However, the levels of p21Waf1/Cip1 also increased in p53-null and mutant cells, which correlated with increased phosphorylation levels of Chk2 (Figures 3C), indicating the presence of a p53-independent mechanism for p21Waf1/Cip1 up-regulation in these cells. A regulation of p21 levels by aspirin has been reported previously in HT29 cells, which express the transcriptionally inactive p53R273H mutant (35). In order to further investigate the role of p21Waf1/Cip1, we extended our study to HCT116p21−/− cells, in which the expression of the p21Waf1/Cip1 gene has been disrupted [Figure S1, supplementary data are available at Carcinogenesis online; (22)]. HCT116p21−/− cells were even more resistant to aspirin than HCT116p53−/− cells (e.g. at 2.5 mM aspirin 76.8 ± 10.3 versus 47.6 ± 2.1% of control; P < 0.0001; Figure 4A). Moreover, 48 h of treatment with aspirin failed to induce a G1 arrest (Figure 4C), which was clearly visible in the parental cell line HCT116 (Figure 4B). Instead, even low concentrations of aspirin (above 1.25 mM) induced cell death, suggesting that this apoptotic response is not dependent on p21Waf1/Cip1 but also that the disruption of p21Waf1/Cip1 could somehow sensitize the cell to apoptotic stimuli (36). In order to quantify the effect of aspirin on p21Waf1/Cip1 expression, we transfected HCT116, HCT116p53−/−, HCT116p21−/− and HT29 cells with a luciferase reporter plasmid bearing the p21Waf1/Cip1 minimal promoter upstream of the Luciferase gene. Upon aspirin treatment, all cell lines showed an increased expression of p21Waf1/Cip1-driven luciferase activity (Figure 4D), consistent with our western blot findings (Figure 3C). In an analogous experiment with a luciferase reporter plasmid containing the Bax promoter, only HCT116p53−/− failed to induce Bax-driven luciferase expression (Figure 4D). Furthermore, when p53 was reintroduced into the HCT116p53−/− cells via transient transfection, they re-gained the ability to up-regulate the expression of Bax upon treatment with aspirin (Figure 3D), suggesting that the induction of Bax, the major mediator of apoptosis upon DNA damage, is dependent on p53 expression (37).

Aspirin-induced G1 arrest is reversible

Activation and phosphorylation of the checkpoint kinases ATM and Chk2, as well as p53 phosphorylation at Ser15, are normally associated with cellular responses to DNA damage and/or DSBs. Moreover, the appearance of foci of γ-H2AX phosphorylation is characteristic of DNA damage and repair. In order to verify that aspirin did not cause permanent damage to cells, we exposed HCT116 cells to 5 or 10 mM aspirin for 48 h followed by culturing these cells under normal medium conditions for further 48 h (Figure 5). The G1 arrest and other cellular changes (which might mediate a G1/S checkpoint response) were fully reversible, as the profiles of the cell samples were very similar to those of untreated cells. In addition, after treatment with 10 mM aspirin, the remaining viable cells seemed to regain a normal proliferation profile. This suggests that aspirin does not produce permanent cell damage.


There is growing experimental and clinical evidence indicating that aspirin has cancer-preventive activity, in particular, in the colon (1). A possible mechanism for the antitumor properties of aspirin has been ascribed to its direct inhibition of COX-2 in colorectal cells (38). However, the anti-proliferative activity of aspirin does not correlate exclusively with its COX-2 inhibitory activity, as aspirin can induce apoptosis in cells lacking COX-2 expression [see Figure S1, supplementary data are available at Carcinogenesis online and (39)]. Although we have not tested the inhibition of prostaglandin production in our system, we do not believe that the observed effects may depend on the regulation of the prostaglandin pathway alone, as the latter is likely to be similarly regulated in the nearly isogenic cell lines HCT116, HCT116p53−/− and HCT116p21−/−, all negative for COX-2 (Figure S1, supplementary data are available at Carcinogenesis online), whereas these cell lines show distinctively different responses to aspirin.

Predisposition to cancer and radiosensitivity observed in Ataxia Telengectasia patients has been linked to chromosomal instability, abnormalities in genetic recombination and defective signaling to programmed cell death and several cell cycle checkpoints activated by DNA damage. ATM mutations (generally null alleles that truncate or destabilize the protein) are involved in the development of sporadic human cancers such as leukemia (40); loss of heterozygosity at the ATM locus was also found in about 30% of colorectal carcinomas (41). These observations predicted that the ATM gene encodes a protein, which plays a crucial role in sensing DNA damage and transducing signals that promote apoptosis (42). Furthermore, it was reported that ATM might have a role in activating defence mechanisms against oxidative stress (43). In a recently published study, the mechanisms by which the aspirin-related compound nitric oxide (NO)-releasing aspirin induced apoptosis in the human B-lymphoblastoid TK6 cell line were investigated. The authors observed that treatment with NO-aspirin led to DNA damage, histone H2AX phosphorylation on Ser 139 and ATM phosphorylation on Ser 1981, effects that were dependent on the NO-moiety of the compound. In fact, these responses were attenuated by the reactive oxygen species scavenger N-acetyl-L-cystein, suggesting that the DNA damage induced by NO-aspirin is caused by oxidative stress. In our system, aspirin did not affect the production of reactive oxygen species by activated polymorphonuclear neutrophils (Supplementary Figure S4, supplementary data are available at Carcinogenesis online), as measured by a lucigenin-based method, indicating that oxidative stress is not the mediator of the observed response to aspirin. We reported the activation of the ATM-dependent checkpoint independently of oxidative stress and of aspirin anti-inflammatory properties. Our evidence for ATM being involved in the checkpoint response mediated by aspirin includes auto-phosphorylation of ATM and of its targets γ-H2AX, p53 and Chk2 (Figure 3A–C) and the ability of the ATM inhibitor caffeine to revert the cell cycle arrest and apoptosis (Figure 3D).

Oncogenes and tumor suppressor genes play essential roles in colorectal carcinogenesis. In cancer cells defective for p53, the tumor suppressor is no longer able to control cell proliferation and is inefficient in preventing the emergence of genetically unstable cell populations. The most common changes of p53 in cancer tissues are missense mutations and loss of heterozygosity, leading to a ‘loss-of-function’ phenotype. Such alterations are found in many human cancers, including colon carcinomas (incidence of 70–75% of cases) (44). Phosphorylation at different residues in p53 has been shown to occur after cells were exposed to DNA-damaging agents. Upon cellular stress, p53 is phosphorylated by kinases targeting its C-terminal and N-terminal domains, which results in the activation of its transcriptional activity (24,45). Serine 15 is a known target of ATM and becomes phosphorylated after DNA damage (46). In our study, p53 phosphorylation at Ser15 was increased following aspirin treatment and was sensitive to the ATM and ATR inhibitor caffeine (Figure 3A). Caffeine was also able to reverse the cellular responses of G1 arrest and apoptosis induced by aspirin in a time-dependent manner (Figure 3D). Moreover, we detected ATM auto-phosphorylation, as well as phosphorylation of γ-H2AX and Chk2 at ATM-targeted sites (Figure 3B and C). Phosphorylation of γ-H2AX by ATM facilitates the formation of foci to which proteins known as mediators of the DNA damage checkpoint, such as MDC1 and BRCA1, are recruited (19). The assembly of these factors near a DSB coincides with checkpoint activation and recruitment of repair proteins. Therefore, the presence of γ-H2AX phosphorylation suggests that aspirin generates signaling resembling that of DSB-induced checkpoint. However, the exact nature of the molecular signal that activates this checkpoint remains to be determined.

Our results show that aspirin induces a G1 cell cycle arrest. An aspirin-dependent G1 arrest had been observed before in several tumor cell lines (5,29). A recent study (47) investigated the effects of aspirin on the NF-κB signaling in HCT116-derived cell lines, with regards to NF-κB translocation and induction of apoptosis. The authors found no effect of either p53 or hMLH1 in the inhibition of the NF-κB pathway by aspirin. A close analysis of some of their findings, though, also shows differences in the cell viability and response to apoptosis between HCT116 and HCT116p53−/− cells, which match our results. We therefore suggest that p53 has only a minor role in the cell cycle response to aspirin, as cells not expressing p53 still undergo a good G1 arrest. It has also been reported that the tumor suppressor p53 may up-regulate COX-2 expression and that COX-2 in turn inhibits p53-dependent transcription (48). However, as we showed, and others reported (49) that HCT116 cells do not express COX-2, this is unlikely to be the molecular mechanism behind the response to aspirin.

In our study, we pinpoint the CDK inhibitor p21Waf1/Cip1 as the major mediator for the aspirin-dependent G1/S checkpoint. A recent microarray-based study by Hardwick et al. (35) reported p21Waf1/Cip1 to be up-regulated, at both gene and protein levels, in HT29 cells treated with 5 mM aspirin. Consistently with these findings, p21Waf1/Cip1 were also raised in our system in all cells tested, including HCT116p53−/− and HT29 cells (Figure 3C; Figure S1, supplementary data are available at Carcinogenesis online). In HT29 cells, p53 is transcriptionally inactive, bearing a mutation (R273H) within its DNA-binding domain, whereas HCT116p53−/− do not express p53; therefore, the increase in the p21Waf1/Cip1 expression upon aspirin treatment is likely to be p53 independent.

The checkpoint kinase Chk2 has also been described to induce p21Waf1/Cip1-dependent senescence (50). Some other p53-independent mechanisms of increasing p21Waf1/Cip1 expression have been reported (51). Among these, the tumor suppressor kinase LKB1, which is often mutated in sporadic cancers (such as CRC), has been recently linked to p21Waf1/Cip1 expression (52). Germline mutations of LKB1 lead to Peutz–Jeghers syndrome, which is characterized by gastrointestinal polyps and cancer in different organs and to loss of LKB1 kinase activity (53). It would be interesting to investigate whether aspirin has any effect on the activation of the LKB1 kinase pathway and if it is useful in patients with Peutz–Jeghers syndrome.

Interestingly, we observed cell death (Figure 2A), caspase 7 cleavage (Figure 3C) and Bax-driven luciferase expression (Figure 4D) not only in wild-type but also in cells expressing p53R273H (but not in p53-null cells), suggesting that the mutant p53 is able to induce Bax expression, activation of the caspase cleavage cascade and thereby apoptosis. HT29 cells have been described previously as being sensitive to aspirin- or other drug-induced apoptosis (54). It was shown that elevated levels of p53R273H, although considered to be transcriptionally inactive, rendered transformed human cells susceptible to apoptosis through Bax (55). Our luciferase reporter assay indicates that p53R273H is able to bind to the Bax promoter, which might suggest that aspirin maintains chemopreventive activities in tumors in which mutations of p53 have occurred. We could speculate that this mechanism helps preventing clonal expansion of p53-mutated cells upon aspirin treatment.

Upon treatment with 10 mM aspirin, HCT116p53−/− cells accumulate in G1 and G2, indicating that both checkpoints have been activated (Figure 2A, middle panels). In a previous study with p53-deficient cells, a bypass of the G1/S checkpoint and an increase in G2/M arrest was described in response to DNA damage (56). Chk2 kinase, which mediates the cellular responses to DNA damage downstream of ATM (33), is known to induce arrest in the G2/M phase of the cell cycle by inhibiting the dual phosphatase Cdc25C and thus blocking activation of the CDC2–cyclin B complex and entry into mitosis (32). In HCT116p53−/− cells, a particularly strong phosphorylation of Chk2 was also observed (Figure 3C). These data lead us to hypothesize that, at high concentrations of aspirin, Chk2-dependent G2/M checkpoint pathways are activated in p53-deficient cells along a weak apoptotic response.

Our results taken together suggest a novel ATM-dependent mechanism of action for aspirin in the chemoprevention of colon cancer. Aspirin is mainly operating through the p21Waf1/Cip1 pathway for its anti-proliferative activity in a p53-independent manner. However, cells not expressing p53 are more resistant to aspirin-induced cell death suggesting that p53 has an important role in the activation of the apoptotic response in colon cells. Downstream effects of the ATM activation induce cleavage of caspase 7 and up-regulation of p21Waf1/Cip1 and Bax; the expression of p21Waf1/Cip1 upon aspirin treatment is transcriptionally regulated in the absence of p53 and may be dependent upon additional factors (Figure 6). It has been suggested that ATM is recruited to the site of DNA damage. ATM can directly bind to and phosphorylate proteins involved in DNA repair, such as c-Abl, Brca1, Nbs1 and replication protein A (57). We speculate that, by activating the G1/S checkpoint, aspirin increases the opportunity of colorectal cells to repair DNA damage before replication or to induce apoptosis both of which may contribute to maintaining the integrity of genomic DNA. Sensor protein complexes scan the DNA for abnormalities and translate these stimuli into activating signals for downstream target proteins, such as the ATM kinase. However, despite the numerous studies on molecular components of checkpoints, both the identities of these sensors and their mechanisms of action are still unclear. It seems possible that aspirin interferes with such sensors upstream of ATM. The identification of additional aspirin targets (besides COX-2) will help in the understanding of its antitumor effects and in the design of novel chemopreventive agents.

Aspirin is rapidly hydrolyzed to salicylate and other salicylic metabolites both in vivo (58) and in vitro (59). The clinical relevance of our data is strengthened by the fact that these concentrations of aspirin correspond to salicylate levels measured in the plasma of human patients, as well as to the therapeutic concentrations used in the treatment of arthritis (60). Comparisons between concentrations used on cell cultures and plasma levels, however, are somewhat artificial, because of the inability to accurately mimic in vivo metabolism and tissue concentration of aspirin in epithelial or tumor cells. In addition, the duration of drug exposure in cell culture is minimal when compared with long-term use for the reason of chemoprevention. In our study, the effects of aspirin on apoptosis and cell cycle were greatest when the cells were treated with a single dose of 5–10 mM aspirin for 48 h. A dose-dependent response in G1 arrest was already visible at lower concentrations in HCT116 cells (1.25–2.5 mM). Therefore, although the concentrations of aspirin used in our study may seem rather high, they are within the range that has been used in culture previously (1–20 mM) (4,5,61), and the effects observed may accurately reflect a biomechanism of repetitive dosing for long-term chemoprevention.

Supplementary material

Supplementary figures S1–S4 can be found at


We would like to thank Prof. Ted Hupp (University of Edinburgh, UK) for the kind gift of the HCT116p53−/− cells, HCT116p21−/− cells and the Luciferase plasmids, pGL3-p21-Luciferase and pGL3-Bax-Luciferase, Ms Cornelia Lichtenberger (Medical University of Vienna, Vienna) for her valuable help with the flow cytometry experiments and Dr James Hutchins (Research Institute for Molecular Pathology, Vienna) and Prof. Thurnher (Medical University of Vienna, Vienna) for the COX-1 and COX-2 antibodies. This work was supported by the Austrian Science Fund grants M-874-B14 (MGL) and P18270 (CG).