Abstract
Elimination of both alleles of the gene that encodes the cyclin kinase inhibitor p21WAF1/cip1 increases the frequency and size of intestinal tumors in Apc1638+/− mice that inherit a mutant allele of the Apc gene, and intermediate effects are seen if a single p21 allele is inactivated. The increased tumor formation is associated with altered cell maturation in the intestinal mucosa of the p21-deficient mice—increased cell proliferation, and decreased apoptosis, and goblet cell differentiation—that is also a function of p21 gene dosage. Moreover, a Western-style diet that mimics principal risk factors for colon cancer (high fat and phosphate, low calcium and vitamin D) accelerates tumor formation in Apc1638+/− mice, and the loss of a single or both p21 alleles is additive with the tumor-promoting effects of this diet, resulting in more and larger tumors, and a highly significant decrease in survival time. Thus, p21 normally suppresses Apc-initiated tumor formation and is haplo-insufficient in this regard. This is consistent with recent reports that Apc initiates tumor formation by up-regulating c-myc expression through alteredβ -catenin-Tcf signaling and that c-myc then up-regulates cdk4, whose activity is inhibited by p21. Decreased expression of p21 is also a marker of poor prognosis in patients, and the data presented suggest that dietary alterations in patients undergoing treatment for colon cancer might be highly effective in improving outcome.
INTRODUCTION
The p21WAF1/cip1 gene, a downstream effector of p53 (1) , is an inhibitor of cyclin-dependent kinase activity (2) and is therefore an important regulator of the cell cycle and potentially, of apoptosis and cell differentiation. In the intestinal tract, p21 is expressed as cells exit the proliferative compartment, and loss of both expression and topological regulation is detected early in colon tumor formation (3 , 4) . Absence of p21 is linked to inability of colon tumor cells to arrest in the G1 phase of the cell cycle (5 , 6) , and the cell cycle arrest and apoptosis of colon tumor cells stimulated by the short-chain fatty acid butyrate, the nonsteroidal anti-inflammatory drug sulindac, and radiation are all linked to induction of p21 (6, 7, 8, 9, 10, 11) .
Despite this evidence for an important role of p21 in the regulation of intestinal cell maturation and tumor formation, the targeted inactivation of the p21 gene in mice does not result in an obvious phenotype in the intestinal tract or other organs, although embryonic fibroblasts derived from such mice are defective in G1 checkpoint arrest (12) . However, it is possible that the loss of p21 is important in the formation of tumors that are initiated by other genetic events. In particular, loss of the wild-type APC gene initiates the development of almost all human colon cancers (13) , and mice that inherit an inactivated Apc allele develop intestinal tumors, principally in the small intestine, when they spontaneously lose or inactivate the remaining wild-type Apc allele (14, 15, 16, 17, 18) . Moreover, because Apc up-regulates c-myc expression through defectiveβ -catenin-Tcf signaling (19) and c-myc then up-regulates cdk4 (20) , a prediction is that p21, an inhibitor of cdk4 activity, should have important effects on the initiation of tumors by Apc.
Therefore, to determine whether p21 could alter intestinal tumorigenesis initiated by loss of Apc, we generated mice that inherited a mutant Apc allele and that were also either heterozygous or homozygous for loss of p21 and found that loss of p21 enhanced tumor formation in a dosage-dependent manner. Further, this increased tumorigenesis was associated with striking effects on cell maturation in the intestinal mucosa. Finally, because a Western-style diet has been shown to act on later stages of tumor promotion in enhancing Apc-initiated tumor formation, we investigated whether the loss of p21 was additive with a Western-style diet. Additive effects on tumor number and size were indeed seen, resulting in a highly significant decrease in life span for the mice. Thus, the combination of p21 loss and a Western-style diet has a profound impact on survival of mice with Apc-initiated tumors, which mimics the poorer prognosis for colon cancer patients whose tumors show decreased p21 expression (21) .
MATERIALS AND METHODS
The Apc1638 and p21 mouse models and methods for genotyping have been reported (12 , 16, 17, 18) . Apc1638+/− mice were mated with p21−/− mice to produce Apc1638+/−,p21+/− offspring (F1). F1 mice were mated to produce desired genotypes: Apc1638+/−,p21+/+, +/−, or −/−. At weaning (approximately 3–4 weeks), littermates were randomized to genetic/dietary groups and fed ad libitum either AIN76A or a Western-style diet that is formulated on the basis of nutrient density to mimic major risk factors for colon cancer in the Western diet: high in fat and phosphate and low in calcium and vitamin D (22 , 23) . Diets were from Teklad (Madison, WI).
Mice were weighed weekly and maintained on diet for 36 weeks or until they exhibited significant weight loss or other signs of extensive tumor formation. Mice were killed by CO2 overdose and cervical dislocation and rapidly dissected for evaluation of tumors and fixation of tissues, as described previously (16 , 18) . Proliferation and apoptosis were evaluated by staining for proliferating cell nuclear antigen (Zymed, South San Francisco, CA) or TUNEL 3 assay (Trevigen, Gaithersburg, MD), as described previously (24) . Goblet cells were detected by staining for mucins with Alcian blue or by immunohistochemical detection of mucins with MCM antibody (Ref. 25 ; a generous gift of A. Einerhand, The University of Amsterdam, The Netherlands), with detection by immunoperoxidase, using an ABC kit (Vector Laboratories, Burlingame, CA).
RESULTS
Apc1638+/− mice were mated with p21−/− mice to produce Apc+/−,p21+/− offspring (F1). The F1 mice were then mated to produce mice that were Apc+/−,p21+/+, +/−, or− /− (F2). By 36 weeks on a defined AIN76A diet, the F2 Apc+/−,p21+/+ mice developed intestinal tumors in 60% of the animals, at a frequency of 1.7 tumors per mouse (Fig. 1) ⇓ . This tumor incidence and frequency were identical to that of Apc1638+/− mice on a homogeneous B6 background (18) , thus eliminating the possibility that unlinked loci from the p21 mice had a significant effect on the Apc-initiated intestinal tumor formation. However, littermates that were Apc+/− and either p21+/− or− /− had a significantly higher tumor incidence of 95 and 100%, respectively. In addition, the tumor frequency per mouse was increased by 23% in the Apc+/−,p21+/− mice and by >118% in the Apc+/−, p21−/− mice. The effect on tumor size was also striking: tumors in Apc+/−,p21+/− mice were 26% larger than the tumors in Apc+/−,p21+/+ mice at 36 weeks, and in the Apc+/−,p21−/− mice, the tumors were 74% larger.
The incidence, frequency and size of gastrointestinal tumors in Apc+/−, p21+/+, +/−, or −/− mice fed AIN-76A diet for 36 weeks (∗, P < 0.05, ∗∗, P < 0.01 comparison with Apc+/−, p21+/+ mice).
In vitro, the short-chain fatty acid butyrate, a physiological regulator of cell maturation in the intestinal tract (7 , 24) , elevates p21 expression in association with stimulation of cell cycle arrest, differentiation markers, and apoptosis (7) . Thus, p21 elevation in cultured intestinal cells mimics the in vivo association of elevated p21 expression with the major pathways of intestinal cell maturation seen after cells exit the proliferative compartment (3 , 4) . We, therefore, investigated how the targeted inactivation of p21 in vivo is associated with alterations in cell maturation pathways in the intestine. In the duodenum, the principal site of tumor formation in these mice regardless of genotype, absence of a single p21 allele in the Apc+/− mice increased proliferation by 25%, and in the absence of both alleles, proliferation increased by 38%. We also detected decreased apoptosis, measured by TUNEL, of 47% in the Apc+/−,p21+/− mice and 68% in the Apc+/−,p21−/− mice (Fig. 2) ⇓ . Figure 2 ⇓ also shows that there is a major effect of the loss of a single or both p21 alleles on the ratio of proliferating to apoptotic cells in the duodenum. An important observation was that the increased proliferation and decreased apoptosis were not seen in the mucosa of the proximal colon of the Apc+/−,p21−/− mice compared with Apc+/− littermates wild type for p21 (data not shown). Because tumors formed only in the small intestine in these Apc+/− mice, even in the absence of p21, the changes in proliferation and apoptosis were tightly associated with the risk for tumor development in the tissue.
Proliferation and apoptotic index in the duodenal epithelia of Apc+/−,p21+/+,+ /−, or −/− mice fed AIN-76A diet for 36 weeks (∗, P < 0.05, ∗∗, P < 0.01 comparison with Apc+/−,p21+/+ mice).
We also investigated the effects of p21 on intestinal cell differentiation. We focused on the goblet or secretory cell lineage, because this lineage is often lost very early in aberrant crypt foci in humans at risk for development of colon cancer (26, 27, 28) , in mice treated with chemical carcinogens (27) , and in Apc1638+/− mice (27) . Figure 3 ⇓ illustrates a comparison of Apc+/−,p21−/− mice with the Apc+/−,p21+/+ mice, showing a substantial decrease in goblet cells in the duodenal mucosa that could be detected by Alcian blue staining (Fig. 3, A and B) ⇓ or by immunohistochemistry with MCM antibody, which recognizes mature mucin (Fig. 3, C and D) ⇓ . Frequencies of Alcian blue- and MCM-positive cells were determined for six mice in each genetic group. Table 1 ⇓ shows that the decrease in the number of these cells was approximately 25% in the p21+/− heterozygotes and 39% in the p21−/− homozygotes (P < 0.05 and P < 0.001, respectively, compared with p21 wild-type mice). Similar to the results for proliferation and apoptosis, no effect of the loss of p21 on goblet cell frequency was detected in the large intestine, where tumors do not form (not shown).
Identification of goblet cells in the duodenal epithelia of Apc+/−,p21+/+ or −/− mice fed AIN-76A diet for 36 weeks. Detection of goblet cells was by Alcian blue staining (A, B) or immunohistochemistry with an antibody (MCM; C, D) that recognizes mature murine mucin.
Alcian blue and MCM staining in the duodenum of Apc+/−, p21+/+, +/−, or −/− mice fed AIN-76A diet
The effects of p21 inactivation in the flat mucosa of the intestine at the anatomical site of tumor formation suggests that loss of p21 plays a role early in the development of tumors initiated by the Apc mutation. A Western-style diet that mimics the human dietary intake of high fat and phosphate and low calcium and vitamin D, which are risk factors for colon cancer, also enhances tumor formation in the Apc1638+/− mouse, but this has been reported to be attributable to later effects on tumor promotion (29 , 30) .
If the p21 and a Western-style diet indeed act at different times during Apc-initiated tumor formation, we would hypothesize that the effects of the two would be additive and independent. To test this, mice that were Apc+/−,p21+/+, +/−, or− /− were fed the Western-style diet. We found a striking effect on intestinal tumor formation first reflected in the survival of the animals. Apc1638+/− mice that were wild type for p21 and fed the defined AIN-76A diet developed intestinal tumors (i.e., Fig. 1 ⇓ ) but survived to 36 weeks of age (Fig. 4) ⇓ . Littermates that were Apc+/− and either p21+/− or −/− fed AIN-76A diet died somewhat earlier, coincident with the increase in tumor number and growth, and this was statistically significant for the Apc+/−,p21−/− group compared with Apc+/−,p21+/+ (P < 0.004). However, when fed the Western-style diet, all of the animals showed decreased survival compared with the same genetic groups fed AIN-76A (Fig. 4) ⇓ . For each genetic group, the effects of the Western-style diet in reducing life span were highly significant (P < 0.02, 0.003, and 0.015 for Apc+/−,p21+/+, +/−, and− /−, respectively). Most dramatic was the effect of a combination of absence of both alleles of p21 and the Western-style diet. Fewer than 29% of these mice survived to 36 weeks, whereas for the wild-type p21 mice on the standard or Western diet, the survival rates were 100 and 75%, respectively. The difference in survival for Apc+/−,p21+/+ mice fed AIN-76A compared with the Apc+/−,p21−/− mice fed the Western-style diet was significant at the P < 0.0001 level.
Survival of Apc+/−,p21+/+,+ /−, or −/− mice fed AIN-76A or Western-style diet.
Because many of the animals on the Western-style diet died before 36 weeks of age or had to be killed early because of their tumors, the number and size of tumors for this experiment and a comparison to the mice fed AIN76A is presented only for those mice that survived to 36 weeks of age. Table 2 ⇓ illustrates that for this subset, the frequency and size of the tumors was much greater for each p21 genotype fed the Western-style diet compared with those fed AIN76A. The effects of the Western-style diet appeared to be additive with the effects of the p21 mutation. For each p21 genetic group (wild type, heterozygotes, and homozygotes), tumor number and size were approximately double for the Western-style diet groups compared with the AIN76A diet groups (Table 2) ⇓ . As a result, the combination of the Western-style diet and the absence of both p21 alleles increased the number of tumors per mouse by 4.4-fold (P < 0.004) compared with p21 wild-type animals fed AIN-76A, and the size of the tumors increased by 3-fold (P < 0.001). This significant enhancement of Apc+/−-initiated tumor formation by a combination of inactivation of p21 and a Western-style diet led to the highly significant decrease in life span for the mice shown in Figure 4 ⇓ .
The frequency and size of gastrointestinal tumors in Apc+/−, p21+/+, +/−, or −/− mice fed AIN-76A diet or Western-style diet
DISCUSSION
We have demonstrated that the p21 gene normally suppresses tumor formation initiated by inactivation of the Apc gene. Although the largest effects were seen when both alleles of p21 were absent, absence of a single allele had intermediate effects, and these effects in the p21+/− mice were significant for tumor incidence in mice fed AIN-76A, for tumor frequency and size in mice fed a Western-style diet, and for proliferation, apoptosis, and goblet cell differentiation in the duodenal mucosa. In this regard, related molecules, such as p53, which is a regulator of p21 expression, and p27kip1, like p21 an inhibitor of cyclin-dependent kinases, have also been reported to have a tumor-enhancing effect upon loss of a single allele (31 , 32) .
A Western-style high-risk diet increased tumor formation in the Apc1638+/− mice (30) , and a high-fat diet also increases tumor formation in Min mice (33) . In the present study, the combination of loss of p21 and the consumption of the Western-style diet were additive on tumor formation and together resulted in much more frequent and larger tumors. The result of this was a highly significant decrease in life span of the mice.
The fact that p21 inactivation and a Western-style diet were additive on Apc-initiated tumor formation and that p21 inactivation had pronounced effects on cell maturation in the duodenal mucosa, whereas a Western-style diet has been reported to affect tumor formation in later stages (30) , suggests that the absence of p21 and the Western-style diet acted at different stages and pathways of tumor development and were independent. A role for p21 early in Apc-initiated tumor development is also consistent with reports that Apc inactivation up-regulates c-myc through its effects in altering β-catenin-Tcf signaling (19) and that c-myc in turn up-regulates cdk4 (20) , whose activity is inhibited by p21. Thus, our data support the suggestion that this pathway is important in the initiation of tumor formation by mutations in Apc.
These data have important implications for understanding prognosis of human colon cancer. Although p21 is not frequently lost during the development and progression of human colon tumors, it is down-regulated in expression (3 , 4) , and low expression of p21 in the tumors is an independent prognostic factor that is linked to poorer survival in colorectal cancer patients (21) . Our data on increased tumor formation with loss of p21 in mice are consistent with these clinical data, although it should be pointed out that in the mouse genetic model, p21 is also missing from the surrounding stromal cells, and there are as yet no data suggesting that p21 down-regulation is a characteristic of nonepithelial cells in tumors. It is potentially important that low p53 expression was not a prognostic marker in the clinical studies (21) , a fact that is perhaps consistent with a role for p21 in tumor suppression independent of p53 (34) , such as its p53-independent role in pathways of cell cycle arrest, differentiation, and/or apoptosis of leukemia cells (35 , 36) , muscle cells, (37) , or colonic cells treated with butyrate (7) or sulindac (9) . Moreover, the fact that mouse life span is most significantly reduced by a combination of a tumor-promoting diet and a genetic modifier of Apc-initiated tumor formation suggests that dietary alterations in patients undergoing treatment for colon cancer might be effective in improving either disease-free and/or overall survival, especially in an adjuvant setting when patients have been surgically cured of disease and the primary goal is prevention of recurrence or metastasis.
We have found gene dosage-dependent effects of the absence of p21 on increasing cell proliferation and decreasing apoptosis and on reducing the number of mature goblet cells in the mucosa of Apc+/− mice that are specifically linked to the site of tumor formation (duodenum versus proximal colon). We believe the effects on proliferation in the mucosa are more important than the effects on apoptosis, because, as we have previously reported, the number of apoptotic cells is very low in the mucosa, and elimination of apoptosis by genetic elimination of short-chain fatty acid metabolism is ineffective in producing colon tumors (24) . However, the percentage change in apoptotic cells with loss of p21 is larger than the percentage change in proliferating cells, and the ratio between proliferating and apoptotic cells both in the mucosa and in the initiated tumor may be crucial elements in determining how tumors form and respond to dietary factors that modulate tumor formation.
The loss of p21 expression may be linked to the decrease in goblet cells attributable to the continued proliferation of cells that prevents their differentiation. Alternatively, p21 could have a more specific role in regulating lineages of differentiation in the intestinal mucosa than that which is presently understood. However, regardless of the mechanism by which goblet cells are depleted by the loss of p21, it is particularly significant that loss of this lineage and mucin secretion is a characteristic of early, preneoplastic aberrant crypt foci in patients who are at risk for developing colon cancer, in rodents treated with colon-specific carcinogens, and in the Apc1638+/− mouse and thus may be an important factor in the promotion of these early lesions (24, 25, 26) .
There are clearly profound interactions between diet and genetics in the development and progression of colorectal cancer (30 , 33 , 38 , 39) . Our finding that the absence of p21 and a Western-style diet can significantly increase tumor formation in combination to a greater extent than either does alone again demonstrates the importance of considering both dietary and genetic factors in tumor formation, in chemoprevention, and in therapy.
Acknowledgments
We would like to thank P. Leder and colleagues for supplying the p21−/− mice.
Footnotes
-
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.
-
↵1 Supported in part by NIH Grants CA75246 and PO CA13330.
-
↵2 To whom requests for reprints should be addressed, at Department of Oncology, Albert Einstein Cancer Center, Montefiore Hospital, 111 East 210th Street, Bronx, NY 10467. Phone: (718) 920-4663; Fax: (718) 882-4464; E-mail: augen{at}aecom.yu.edu
-
↵3 The abbreviations used are: TUNEL, terminal deoxynucleotidyl transferase-mediated nick end labeling; MCM, murine colonic mucin.
- Received June 30, 2000.
- Accepted November 13, 2000.
- ©2001 American Association for Cancer Research.
References
Volume 61, Issue 2
