Numerous studies have indicated that exposure to nonsteroidal anti-inflammatory drugs is associated with a lowered risk of colorectal cancer. However, analyses of the effect of aspirin upon tumorigenesis in ApcMin/+ mice have yielded contrasting results. We show that adult dietary exposure to aspirin does not suppress intestinal tumorigenesis in ApcMin/+ mice, but that continual exposure from the point of conception does. To test whether this regime could suppress the phenotype of murine models of hereditary nonpolyposis colorectal cancer, Msh2-deficient mice were exposed to aspirin. This did not modify the mutator phenotype of Msh2−/− mice, but weakly extended survival. Finally, we analyzed (ApcMin/+, Msh2−/−) mice and found that lifetime aspirin exposure significantly delayed the onset of both intestinal and mammary neoplasia. Thus embryonic and perinatal exposure to aspirin suppresses neoplasia specifically associated with the loss of Apc function, opening a potential window of opportunity for nonsteroidal anti-inflammatory drug intervention.
Substantial epidemiological evidence shows that treatment with NSAIDS3 reduces the risk of developing colorectal cancer. The most commonly used NSAID has been aspirin, which has been reported to reduce the risk of colon cancer by up to 40%. Clinical studies using the NSAID sulindac have also reported reduced polyp counts in FAP patients (1, 2). The precise mechanism of NSAID action remains unclear, although the suppression of Cox2 is thought to be of pivotal importance. Significantly, Cox2 is overexpressed in ∼85% of human colorectal adenocarcinomas and adenomas from the ApcMin mouse (3). The definitive study highlighting the relevance of Cox2 to tumorigenesis showed that a Cox2-deficient background markedly suppressed intestinal neoplasia in mice carrying the Apcδ716 allele (4). The biological activities of aspirin and sulindac are not restricted to suppression of the Cox2 pathway. Aspirin, sulindac, and even the selective Cox2-inhibitor celecoxib (5) have been shown to induce apoptosis in a Cox2-independent fashion. Both aspirin and sulindac down-regulate β-catenin- and β-catenin/TCF4-mediated transcription (6, 7). NSAIDS also act on the nuclear factor κB signaling pathway (8), and recent data suggest this may be important for their antitumor activities (8). They have also been shown to specifically reduce the survival of genetically unstable (MSI+) MMR-deficient colorectal cancer cell lines (9), raising the possibility that aspirin may also suppress malignancy in hereditary nonpolyposis colorectal cancer families characterized by mutations in the MMR genes. Mice constitutively heterozygous for the ApcMin allele have been used to determine the ability of the NSAIDs to suppress intestinal malignancy. However with respect to aspirin, these studies have produced contrasting results. Virtually every study [e.g., Beazer-Barclay et al., (10) has shown that sulindac causes a reduction in the number of the spontaneous malignancies, apart from Oshima et al. (4) who use relatively low levels of sulindac]. However, of the studies that have investigated spontaneous intestinal malignancy after aspirin treatment, only two have shown suppression of malignancy in the ApcMin mouse. Two other studies failed to show suppression in either the ApcMin mouse or the Apc1638N mouse (6, 11, 12, 13). The basis for these discrepancies may lie within differences in the aspirin regime used. Shoemaker et al. (14) and Reitmair et al. (15) have argued that the majority of adenomas are fixed either in utero or perinatally just after birth.
Here we directly test the effect of increasing the period of aspirin exposure to include the entire period from the point of conception onwards. Furthermore, we have investigated whether this regime can modify the development of MMR-associated neoplasia by investigating the course of neoplasia in cohorts of Msh2−/− and (ApcMin/+, Msh2−/−) mice. We find that prolonged aspirin exposure dramatically enhances the suppression of Apc-associated neoplasia both within the intestine and the mammary gland but only weakly influences the phenotype of MMR deficiency.
Materials and Methods
Administration of Dietary Aspirin at Weaning.
C57BL/6 mice wild-type and heterozygous for the ApcMin mutation were placed on diets containing either 0, 200, or 400 mg/kg of aspirin (Harlan/Tekad). These levels of exposure are comparable with the highest doses used in previous studies (6, 11, 12, 13). Mice were monitored everyday for signs of disease, generally manifesting itself as anemia, loss of weight, and a hunched appearance.
Two experiments were performed: one where mice were killed at 150 days of age and one where mice were killed where they showed signs of disease.
Permanent Administration of Aspirin.
Matings segregating for progeny which were ApcMin/+, Msh2−/− or (ApcMin/+, Msh2−/−) were placed on aspirin diets containing either 0 or 400 mg/kg of aspirin before conception and throughout pregnancy and lactation. Progeny were weaned onto appropriate diets. Mice were killed when they showed signs of disease.
Tissues were removed, fixed in 10% neutral buffered formalin, embedded in paraffin, sectioned at 10 um, and stained with H&E before microscopic analysis. Scoring of intestinal lesion was achieved by removing the entire intestine at necropsy, flushing with PBS, and mounting en face. These preparations were then fixed in methacarn (4:2:1, methanol:chloroform:glacial acetic acid). Lesions were then scored macroscopically. Intestine was then wound into a “swiss” roll, which was subsequently embedded in paraffin and then sectioned as above.
Mutation Frequency at the Dlb-1b Locus.
The Dlb-1 assay was performed as described previously (16). For this assay, experimental cohorts were derived by backcrossing the Msh2 mutants to two different C57Bl/6 strains, one of which was homozygous for the Dlb-1 a allele and one of which was homozygous for the Dlb-1 b allele. Mice were subsequently intercrossed from these two lines to generate mice heterozygous at the Dlb-1 locus and segregating for all possible Msh2 genotypes.
Results and Discussion
Aspirin Does Not Suppress Intestinal Neoplasia when Administered at Weaning.
Wild-type and ApcMin mice were placed on aspirin-containing diets (either 200 mg/kg or 400 mg/kg) at weaning. (Fig. 1). Mice exposed to the aspirin diet gained weight normally when compared with mice on the control diet, and no ulceration or intestinal pathology such as perforation of the intestine was observed in any of the wild-type mice treated with aspirin. Two independent analyses were performed.
First, cohorts of mice either wild-type or heterozygous for the ApcMin allele were killed at 150 days of age, and the adenoma burden was assessed on whole mount preparations of the entire small intestine. No adenomas were seen in wild-type mice. Adenoma burden was as follows: for mock treated ApcMin mice (n = 5), 9.8 (±6.3); for mice treated with 200 mg/kg (n = 7), 9.7 (±5.88); and for mice treated with 400 mg/kg of aspirin (n = 5), 8.2 (±6.05). There was no difference in tumor burden at either 200 mg/kg (P = 1.00) or at 400 mg/kg (P = 0.75; Mann Whitney) compared with mock-treated mice, indicating that aspirin exposure was not modifying the ApcMin phenotype.
Second, cohorts of mice were permitted to age until they developed obvious symptoms of intestinal neoplasia, usually bleeding from the anus or anemia scored through whitening of the feet. Fig. 1 A shows a Kaplan-Meier plot reflecting survival over a 400-day period. Exposure to aspirin at either 200 mg/kg or 400 mg/kg did not alter the survival profile, again indicating that exposure to aspirin did not modify the ApcMin phenotype.
Aspirin Suppresses Intestinal Neoplasia when Administered in Utero.
Because Shoemaker et al. (14) and Reitmair et al. (15) have argued that the majority of adenomas are fixed before 6 days of age, we investigated the effect of aspirin exposure throughout embryogenesis and weaning. Cohorts of wild-type and ApcMin heterozygotes were derived from parents placed on aspirin-containing diets before conception. Dietary exposure to aspirin was maintained throughout and beyond weaning. We first wished to establish whether in utero exposure resulted in increased embryonic lethality of ApcMin heterozygotes, as has been reported for the Cox1- and Cox2-inhibitor piroxicam (17). Analysis of offspring showed this not to be the case, because there was no reduction in the number of ApcMin heterozygotes in progeny from aspirin-exposed parents (P = 0.2; χ2 test).
We next determined the survival profiles of each cohort (Fig. 1,A), which showed a significant increase in survival in ApcMin heterozygotes exposed to aspirin from conception onwards (P = 0.0004; log-rank test). This effect was sufficient to completely prevent the development of symptoms associated with intestinal neoplasia in 5 of 16 ApcMin at 500 days of age. Analysis of tumor burden and distribution in those mice that developed intestinal tumors showed no obvious differences in either small (Fig. 1 B) or large intestine (data not shown).
A number of different possibilities may underlie the discrepancies between the published studies examining the effect of aspirin on ApcMin-associated tumorigenesis. First, that it is a consequence of differences in the level of aspirin exposure—a possibility which seems unlikely given the similar levels used in each experiment. Second, that it is attributable to differences in the genetic background of the mice used, either at known modifiers of the ApcMin phenotype, such as Mom-1, or at as-yet unidentified modifiers (18). In this respect, it is notable that our study differs from previous studies by using mice homozygous for the relatively resistant C57/Bl6 Mom-1 allele. Third, that it is a consequence of different husbandry regimes; for example, mice housed in a sterile facility may respond very differently to those in a nonsterile facility.
Comparison of the data presented here with published experiments does not permit us to distinguish between these possibilities, but it does serve to underline the relatively fragile nature of aspirin-mediated suppression within this model. We show here that one of the key modifiers of the effectiveness of aspirin is the time point of exposure. By increasing the period of dietary exposure to include in utero and perinatal exposure, we have dramatically enhanced the ability of aspirin to modify the ApcMin phenotype. Within our experimental design, this enhancement is particularly notable given the failure to observe any effect in the ApcMin cohort treated postweaning.
These findings are consistent with the presence of a developmental “window” for adenoma formation either in utero or shortly after birth. In support of this, Shoemaker et al. (14) showed that perinatal exposure to chemical carcinogens specifically enhances intestinal tumor development in the adult. One possible explanation for this phenomenon is the huge expansion in the number of intestinal crypts which occurs 3 weeks after birth. This process of crypt fission inevitably expands the number of any mutation-bearing crypts present within the intestine and may therefore clonally expand any crypts bearing mutations at the Apc locus.
The presence of a very early window for neoplastic development raises serious questions about our understanding of Apc-mediated tumorigenesis. Loss of Apc function is considered to be the key initiating event in intestinal neoplasia and is strongly associated with both dysplasia and dysregulation of β-catenin [e.g., Kongkanuntn et al. (19)]. However these phenomena simply are not observed in mice aged <4 weeks, indicating either that loss of Apc alone is insufficient to lead to the up-regulation of β-catenin and development of dysplasia, or that loss of Apc (and thereby dysregulation of β-catenin) occurs as a secondary event. In this regard we have shown previously that a proportion of dysplasias occurring in a mismatch repair-deficient background are not characterized by increased β-catenin levels (19).
Aspirin clearly acts to suppress these early events either by deleting or suppressing the phenotype of those cells which carry an increased predisposition to adenoma formation. This seems unlikely to be through modulation of β-catenin levels, as has been suggested by Mahmood et al. (6), because, as stated above, elevated levels of β-catenin are not observed in young animals. A second possibility is that aspirin may suppress neoplasia through a Cox2-mediated pathway. Cox2 up-regulation is observed in 80–85% of human colorectal carcinomas, in 50% of colorectal adenomas, and within tumors arising in the ApcMin model (3). However, it seems unlikely that the primary effect of in utero aspirin exposure is mediated through Cox2 suppression, as Cox2 overexpression is associated with the later stages of tumorigenesis by modulating angiogenesis and levels of apoptosis within tumors (3).
Whatever the mechanism of action of aspirin, these studies demonstrate that the initiation of adenoma formation occurs very early in the ApcMin mouse, and that aspirin exposure at this time point can efficiently suppresses this process. Understanding the factors determining the predisposition to adenoma formation during embryonic development and perinatally might lead to greater insights into the molecular mechanisms of cancer in humans.
Aspirin Weakly Suppresses Neoplasia though not Mutation in Msh2−/− Mice.
Having established an effective protocol for aspirin exposure in the murine model of FAP, we determined whether this approach could modulate the phenotype of the murine model of hereditary nonpolyposis colorectal cancer. A significant subset of human intestinal tumors are characterized by mutations in the MMR pathway, and all murine models of MMR deficiency show increased predisposition to neoplasia.
Cohorts of Msh2−/− mice were either exposed to aspirin or were fed control diet from conception, and Kaplan-Meier survival curves were generated (Fig. 2 A). All 18 animals on the control diet were killed after the development of lymphoma, although 2 animals also had coexistent intestinal malignancy. All Msh2−/− mice exposed to aspirin also developed lymphoma; however, there was a small increase in survival compared with controls (P = 0.05; log-rank test). This slight shift in survival may reflect weak suppression of either intestinal malignancy or lymphomagenesis. This latter possibility is consistent with the one report that showed that aspirin exposure can result in the deletion of MSI-unstable cells in culture (9).
These results prompted us to determine directly whether aspirin exposure can suppress in vivo mutation in a mismatch repair-deficient background. Mutation frequency was scored at the Dlb-1b locus in Msh2 mutant mice continually exposed to either control or aspirin-containing diets. Using this assay, we have shown previously that Msh2-deficient mice have a mutator phenotype at the Dlb-1b locus (16, 20). Msh2-deficient mice were analyzed at 4 months of age after exposure to either control or aspirin-containing diet. No difference in mutation frequency was observed at this time point, demonstrating that this regime of aspirin exposure does not modify the mutator phenotype of Msh2-null epithelium (Fig. 2 B).
Aspirin Suppresses Intestinal and Mammary Neoplasia in ApcMin/+ Msh2−/− Mice.
In human colorectal cancer MMR-deficient tumors differ in a number of respects to MMR-proficient tumors. For example, MMR-deficient tumors express lower levels of Cox2 (21) as well as exhibiting very different patterns of mutation and genomic instability. These differences invoke different mechanisms for MMR-driven neoplasia, raising the possibility that aspirin exposure may have an additional effect in a MMR-deficient background, for example through the deletion of cells showing MSI.
The development of lymphoma in Msh2−/− mice largely precludes an analysis of the effect of aspirin upon Msh2-dependent intestinal neoplasia (22). However, Msh2 deficiency has been shown to greatly accelerate intestinal neoplasia in the ApcMin mouse (15), so permitting an analysis of the effect of aspirin in this context. In addition, mammary akanthoma occur in ApcMin/+ mice (23), and this is greatly accelerated in (ApcMin/+, Msh2−/−) mice, permitting a study of the effect of aspirin on the development of this lesion (see below). Cohorts of (ApcMin/+, Msh2−/−) mice were generated and exposed to either control diet or aspirin-containing diet from conception onwards, and Kaplan-Meier survival curves were generated (Fig. 3 A). Survival was markedly enhanced after aspirin exposure (log-rank test; P = 0.0001), although it was still reduced by comparison to untreated ApcMin heterozygotes. Aspirin exposure did not alter either tumor distribution or burden scored at the point of death (data not shown).
In the cohorts maintained on the control diet, a single mammary akanthoma was observed in a total of 15 ApcMin/+ mice, whereas none were detected in 18 Msh2−/− mice. Development of this tumor type was enhanced in ApcMin/+ Msh2−/− mice (11 tumors in 41 mice). In the cohorts exposed to aspirin, no mammary akanthomas were observed in 10 ApcMin/+ mice, whereas a single akanthoma was observed in 21 Msh2−/− mice. In the (ApcMin/+, Msh2−/−) cohort, five mammary akanthomas were observed in 20 mice. All of these mice also had intestinal neoplasia. Although this represents a similar prevalence to that observed in the untreated (ApcMin/+, Msh2−/−) cohort, the onset of akanthoma was significantly retarded in the aspirin-treated mice (Fig. 3 B). These results therefore show aspirin exposure significantly delays the onset of mammary tumorigenesis in the (ApcMin/+, Msh2−/−) background, consistent with several recent studies which have shown that both mutation of Apc and overexpression of Cox2 can lead to mammary neoplasia (24), and, furthermore, that selective Cox2 inhibitors can suppress rodent mammary tumorigenesis (25).
Overall, the median life span of mock-exposed ApcMin mice was 175 days compared with 330 days for aspirin-treated mice, an increase of 89% or 155 days. The comparable figures for the (ApcMin/+, Msh2−/−) cohort are 98 days (mock-treated) and 127 days (exposed), an increase of 29% or 29 days. This comparison suggests that the effect of aspirin was no greater in the (ApcMin/+, Msh2−/−) background. Taken together with the weak effect of aspirin on survival of Msh2−/− mice and the failure to observe an influence of aspirin upon mutation rate, these data argue that aspirin specifically suppresses tumorigenesis in an Apc-dependent manner. This conclusion contrasts the observation that in vitro aspirin can delete cells characterized by MSI (9). This contradiction may be explained by the surprising fact that adenomas developing in (ApcMin/+Msh2−/−) mice do not show microsatellite instability (15), which was independently confirmed here (data not shown). This demonstrates that gross microsatellite instability is not required to drive adenoma formation in this background, possibly because of the high selective pressure for adenoma formation consequent upon the ApcMin mutation. Thus, somewhat paradoxically, cells within the (ApcMin/+Msh2−/−) mice do not exhibit gross levels of MSI and therefore are not available as targets for aspirin-mediated deletion.
In summary, we have shown that dietary aspirin exposure can suppress tumorigenesis in the murine intestine and the mammary gland. This effect seems to be specifically associated with a loss of function of Apc and seems to only weakly modify the MMR phenotype. Critically, in the intestine, suppression only becomes apparent if exposure covers the period between conception and weaning. Obviously prophylactic treatment with aspirin of FAP patients during this window would be inappropriate because of the well-known deleterious side effects of aspirin. This, therefore, identifies a potential window of opportunity for the chemoprevention of intestinal neoplasia. The challenge will now be to establish whether NSAIDS (other than aspirin) can be therapeutically effective within this window.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the Cancer Research Campaign.
The abbreviations used are: NSAID, nonsteroidal anti-inflammatory drug; MSI, Microsatellite Instability; MMR, Mismatch Repair; FAP, familial adenomatous polyposis; Cox2, cyclooxygenase 2; Dlb-1b, dolichos biflorus.
We thank Andrea Leitch for microsatellite analysis, Mark Bishop for genotyping, Nathan Hill for maintenance of animal stocks, and Derek Scarborough for histology.