Understanding Phenotypic Variation in Rodent Models with Germline Apc Mutations

Tumor suppressor adenomatous polyposis coli (APC) is critical for maintaining cellular homeostasis in the intestine (1, 2). APC is a large (2,843 amino acids), multidomain protein that has been implicated in many cellular functions including cellular proliferation, differentiation, cytoskeleton regulation, migration, and apoptosis (3). Mechanistically, APC is best known for its ability to antagonize Wnt signaling by targeting the oncoprotein β-catenin for proteasomal degradation (4).

Acquiring a somatic APC mutation is an early, if not initiating, event in the great majority of colorectal tumors (5). Inheriting a germline APC mutation results in the development of hundreds to thousands of colonic polyps, a condition termed familial adenomatous polyposis (FAP). These precancerous polyps are thought to initiate following a somatic mutation in the wild-type APC allele (6, 7). To avoid the progression of these polyps into invasive carcinoma, prophylactic colon removal is recommended for FAP patients (8). There are no reports of humans with germline mutation of both APC alleles, consistent with early developmental lethality associated with complete loss of APC function (9–11). Germline and somatic APC mutations typically result in premature APC protein truncation and group between codons 1250 and 1464, a region termed the “mutation cluster region” (MCR; ref. 12).

A meta-analysis of genotype–phenotype correlation in patients with FAP showed that germline mutations in the MCR result in the most severe intestinal polyposis phenotype, with up to 5,000 polyps (13). Mutations on either side of the MCR are associated with an intermediate intestinal polyposis phenotype, whereas mutations that result in a truncation in APC after amino acid (a.a.) 1595 or before a.a. 157 are associated with an attenuated phenotype (AFAP), characterized by development of only a few polyps (13). Complete deletion of APC has been reported only rarely and results in an intermediate phenotype (14, 15).

More than two thirds of patients with FAP also have extra-colonic manifestations (13). Chronic hypertrophy of retinal pigment epithelium (CHRPE) is the most frequent phenotype, associated with APC truncation between a.a. 311 and 1446. Desmoid tumors, on the other hand, are associated with APC truncations 3′ to the MCR, after a.a. 1400. Duodenal and gastric tumors have been associated with APC mutations in 2 different regions, downstream of codon 1395 and between codons 564 and 1465 (13). It is important to note that these genotype–phenotype correlations are not rigid or complete, suggesting roles for other genetic and environmental factors in tumor development (13, 16).

For the past two decades, rodent models have been valuable for analysis of APC functions in intestinal homeostasis and tumor suppression (17, 18). APC is well conserved between human and rodent, with 92% similarity at the amino acid level (9, 19). Furthermore, some rodent models with germline Apc mutations that result in Apc protein truncation develop intestinal polyposis similar to that seen in patients with FAP (18). A brief summary of all published rodent models with germline Apc mutations appears in Tables 1 to 3, with a schematic provided in Figure 1.

Characterization of the many available Apc mouse and rat models has aided in discovery of various pathways important in colon carcinogenesis. Apc rodent models were also useful for elucidating the effect of various environmental and genetic factors on intestinal tumorigenesis and for testing potential chemoprevention and therapeutic agents. The many positive contributions of Apc mouse models have been reviewed previously (20, 21). As with most experimental systems, studies of the Apc models have also led to unanswered questions, particularly regarding phenotypic variation among the different models. Here, we review some of these variations, provide potential explanations, and pose challenges for future investigation.

As shown in Table 1 to 3, the average number of polyps varies greatly between different mouse models with germline Apc mutations. In addition, the number of polyps also varies in the same Apc mouse model maintained in different laboratories (17). These variations in intestinal polyp number in different models likely stem from the nature of the Apc mutations as well as environmental and genetic factors (17, 18). We propose that the number of intestinal tumors that develop in different Apc models and in the same model analyzed by different laboratories is influenced by one or more of the following factors.

In both patients with FAP and rodent models with germline Apc mutations, loss or inactivation of the wild-type APC/Apc allele is required for polyp formation (22, 23). The mechanism by which the second wild-type Apc allele is lost appears to depend on the Apc mouse model (24). Because this second Apc “hit” is essential for polyp initiation (10, 22, 25), the rate at which the second “hit” occurs will directly affect the number of intestinal polyps. Increasing the expected rate of these second “hits” through introduction of genomic instability, X-ray exposure, or injection with a mutagen significantly increases the number of polyps in Apc^Min/+ and Apc^1638N mice (26–30). It has been suggested that certain Apc mutations might lead to chromosomal instability, which could affect the rate of wild-type Apc loss (31).

Apc^1638N/+ mice develop relatively few intestinal polyps and the second Apc “hit” is usually inactivation of the wild-type Apc allele, predicted to be a rare event (24). On the other hand, Apc^Min/+ mice, in which the wild-type Apc allele is lost by means of a more frequent LOH event, develop considerably more polyps (24). Loss of the wild-type Apc allele in both Apc^Min/+ and Apc^1322T/+ mice, however, is reported to occur via LOH, yet these two mouse models have widely different polyp numbers (32). Although the rate and underlying mechanism of wild-type Apc allele loss might contribute to intestinal polyp numbers in Apc mouse models, it is unlikely that these are sole defining parameters.

Polyps must reach a certain size to be detectable. If two polyps are initiated at the same time, a more rapidly growing polyp should be detectable earlier than a slower growing polyp. The most recognized function of Apc is to antagonize the Wnt signaling pathway through inhibition of the activity of β-catenin as a transcription cofactor (4). As Wnt signaling can drive cellular proliferation, we might expect that different Apc mutations would lead to different levels of Wnt signal activation and different corresponding changes in cellular proliferation. In patients with FAP, mutations in the MCR are associated with the most severe intestinal phenotypes, whereas mutations outside the MCR lead to reduced polyp multiplicity (13). Notably, APC mutations 5′ and 3′ to the MCR result in higher and lower activation of Wnt signaling, respectively (33). This observation has led to the proposal that submaximal upregulation of Wnt signaling promotes more polyp growth than higher or lower elevation of Wnt signaling, the “just right” hypothesis (34, 35).

Wnt signaling has been assessed in many Apc mouse models. Some models have high polyp multiplicity and show elevated Wnt signaling in these polyps (Apc^Min/+, Apc^Δ716/+, Apc^1322T/+, and Apc^Δe1-15/+; refs. 10, 34, 35). Wnt signaling is also elevated in the few polyps that develop in Apc^NeoR/+ and Apc^NeoF/+ mice (36, 37). Apc^mNLS/mNLS mice have elevated Wnt signaling in intestinal epithelial cells (38, 39). Apc^1572T/1572T embryonic stem cells also have elevated Wnt signaling (38, 39). Neither Apc^mNLS/mNLS nor Apc^1572T/+ mice develop intestinal polyps (38, 39).

The “just right” hypothesis is supported by reports of increased polyp multiplicity in Apc^1322T/+ and Apc^Δe1-15/+ mice relative to Apc^Min/+ mice (34, 35). Compared with Apc^Min, Apc^1322T protein retains one 20-a.a. repeat that can bind to β-catenin and decrease Wnt signaling (34, 35). The Apc^Δe1-15 allele results in complete deletion of Apc and polyps in Apc^Δe1-15/+ mice also display less Wnt signaling than polyps in Apc^Min/+ mice (34). However, the “just right” hypothesis does not readily explain why Apc^Δ716/+ mice show higher activation of Wnt signaling and more polyps than Apc^Min/+ mice (40). In addition, several groups have reported that although loss of both Apc alleles is required to activate Wnt signaling (as assessed by nuclear translocation of β-catenin), this Apc loss is not sufficient for full Wnt signal activation (11, 41, 42). To establish the extent to which Wnt signaling and polyp growth contribute to phenotypic variation, Wnt signaling activities and proliferation rates must be directly compared in different Apc mouse models.

Another explanation of variation in polyp number among different Apc mouse models is negative selection of particular Apc genotypes. This negative selection could contribute to the “just right” hypothesis. Support for negative selection contributing to polyp phenotypes is provided by the observation that addition of Cdx2 or BubR1 mutations to Apc^Δ716/+ or Apc^Min/+ mice, respectively, results in reduced polyp multiplicity and increased apoptotic indices in the small intestines, despite the increased proliferation index in these cells (43, 44). Similarly, induction of a conditional Apc mutation in hematopoietic stem cells results in upregulation of Wnt signaling and increased stem cell proliferation with increased apoptosis and eventual exhaustion of the stem cell population (45). If this phenotype holds true for intestinal tissues, the “just right” hypothesis might explain the increased stem cell number in polyps from Apc^1322T/+ mice relative to those from Apc^Min/+, despite lower Wnt signaling in polyps from the former model relative to those from Apc^Min/+ mice.

It is possible that the effect of Apc genotypes on enterocyte differentiation contributes to differences in intestinal polyp number. For instance, compared with Apc^Min/+ mice, Apc^1322T/+ mice have a higher proportion of Paneth cells and cells that express stem cell markers (Lgr5, Bmi1, Msi1, and CD44), not only in adenomas but also in apparently normal intestinal epithelial cells (35). Cell fates that result from different Apc genotypes might alter tumor initiation or growth. Again, Wnt signaling is one of several factors proposed to affect differentiation.

It is well established that genetic and environmental factors affect intestinal polyp multiplicity in Apc mouse models. Polyp multiplicity in Apc^Min/+ mice varies greatly between laboratories (20–100/mouse; refs. 17, 18). This inconsistency might result from variations in diet, emergence of genetic modifiers, and even from different methods of polyp detection. A genetic modifier is a genetic locus that modifies the effect produced by a nonallelic locus. Modifier genes are present in different mouse strains and can even emerge in what is considered a congenic strain (46). Several modifier loci have been found to affect intestinal polyposis in Apc^Min/+ mice and are named modifier of min (Mom; reviewed in ref. 18). Some modifiers are single genes, others are thought to represent contiguous genes and some remain less well defined (47). The modifiers appear to function as recessive, dominant, or semidominant loci (17). The first identified modifier gene, Mom-1 (Pla2g2a), works in a cell nonautonomous manner, possibly by reducing inflammatory response in the gut (48–50). The Mom-2 (Atp5a1) allele is on the same chromosome as Apc (chromosome 18) and appears to inhibit loss of the wild-type Apc allele (48, 51). The mechanisms of action of other modifiers such as Mom-3, Mom-7, Mom-12, and Mom-13 are not understood (52–54).

Although identified in Apc^Min/+ mice, Mom genes likely also affect phenotypes of other Apc mouse models. For instance, the C3H/HeJ mouse strain carries at least one Mom allele that is absent from the C57BL/6 strain Mom-1 (48). Both Apc^Min/+ and Apc^Δ242/+ mice show reduced polyp multiplicity in the first generation mixed C57BL/6: C3H/HeJ mice compared with C57BL/6 mice (55). At present, there appears to be no direct examination of the effect of specific modifiers of Min on different Apc mouse models.

Environmental factors, such as intestinal flora, might also contribute to phenotypic variation (56). While intestinal flora appear to increase the number of polyps in Apc^Min/+ mice (57), Apc^Δ14/+ mice raised in pathogen-free conditions showed significant increases in intestinal polyp number (58).

Diet is another major environmental factor that clearly impacts the mouse phenotype (59–61). Although typically defined, the concentration of various vitamins, fiber, and total fat varies greatly between laboratory mouse diets. In our own experience, switching the mouse diet had a dramatic effect on polyp multiplicity in our Apc^Min/+ mouse colony. We found that the polyp burden per mouse significantly increased from 45.9 ± 4.5 in 10 Apc^Min/+ mice on Lab diet 5001 (Purina) to 81 ± 9.3 in 25 age-matched Apc^Min/+ mice on Harlan 2018 diet (P = 0.0006). Notably, the new diet (Harlan 2018) has a 24% increase in fat and decreased fiber, vitamin D, and folic acid by 42%, 67%, and 44%, respectively. Unfortunately, these interlaboratory variables such as diet confound direct comparison of the phenotypes of Apc mouse models studied in different laboratories.

APC interaction with cytoskeletal components, including actin filaments and microtubules, is thought to affect cell adhesion and migration (62, 63). Decreased cellular adhesion and migration in cells with APC mutations is expected to contribute to tumor formation (64). APC interacts with cytoskeletal proteins through its C-terminal region, which is absent in Apc from most mouse models (Fig. 1). Adding the C-terminal Apc region to Apc^1322T (as in Apc^ΔSAMP mice) did not change the phenotype (65). However, it is possible that cytoskeletal alterations affect later stages of tumor progression such as invasion and metastasis, which do not occur in most Apc mouse models (66). Currently, evidence supporting a direct role of the Apc C-terminus in intestinal phenotype variation among different Apc mouse models is lacking.

Apc rodent models have been generated using three different technologies: chemical mutagenesis screen, insertion of an antibiotic resistance gene, and Cre-lox–induced DNA excision. The Apc^Min/+ mouse, PIRC rat, and KAD rat were generated by chemical mutagenesis which resulted in a single basepair change in the Apc gene (9, 67, 68). Many other models, such as Apc¹³⁰⁹, Apc^1638N, and Apc^1638T, were generated through insertion of an antibiotic resistance gene into the Apc gene, thus introducing a nonsense mutation (69–71). In Apc^neoF and Apc^neoR alleles, the antibiotic resistance gene disrupts an enhancer sequence in intron 13 (36, 37). Other mouse models with Apc truncation including Apc^1322T/+ and Apc^Δe1–15 were generated using Cre-lox–mediated deletion of specific Apc regions. The later technology allowed removal of most exogenous DNA sequences originating from the targeting vector including the antibiotic resistance gene. The Apc^mNLS model contains mutations “knocked into” the Apc gene, with the antibiotic resistance gene subsequently removed by Cre-lox–mediated deletion (39).

The Apc^1638N/+ and Apc^1638T/+ models, which differ only by orientation of the inserted neomycin resistance gene, provide clear evidence for the contribution of extraneous DNA to phenotypic variation (69). Apc^1638N/+ mice express so little truncated Apc protein that they might be considered virtually null (69, 72); yet the described phenotype of Apc^1638N/+ mice is not similar to that of the Apc^Δe1-15 model, which has a complete deletion of the Apc gene (34, 72). The neomycin resistance gene clearly affects the phenotypes of these mice and if inserted in reverse orientation might affect not only Apc expression but also expression of genes upstream of Apc. It is possible that the 6-fold difference in intestinal polyp number between Apc^1322T/+ and Apc^1309/+ mice, which differs by only 13 amino acids, stems from the different technology used in their generation: Cre-lox–mediated deletion in Apc^1322T/+ versus insertion of an antibiotic resistance cassette in Apc^1309/+. However, other genetic and environmental factors may contribute to the variation between these two mouse models as well (32, 70). A final illustration of the challenges in generation of Apc mouse models is the Apc^Δ474/+ mice, which have a duplication of Apc exons 7 to 10. This feature complicates dissection of the contribution of exon duplication to the phenotype (73).

When analyzing the phenotypes of different Apc mouse models, another consideration is the level of expression of the mutant allele. Apc is a large multidomain protein. Truncations of Apc in most patients with FAP and rodent models leave N-terminal domains intact (Fig. 1). Although normal expression levels of truncated Apc protein have been verified in Apc^Δ716, Apc^Min/+, Apc^1322T, and Apc^1638T mice, this is not universally the case (32, 69, 74). In Apc^580D, Apc^Δ14, Apc^Δ474, and Apc^Δ242 models, the truncating mutation occurs before the final exon (15), and thus there is the possibility of a nonsense-mediated RNA decay. Truncated Apc was not detected in intestinal polyps from Apc^Δ580/+ mice or embryonic stem cells from Apc^Δ15/+ mice (75, 76), which suggests that these alleles might also be virtually null. A related consideration is the effect of the introduced mutation (and possibly the antibiotic selection cassette) on Apc folding. Although most of Apc is thought to be natively unfolded (77), the effects of mutations on inherently folded domains of Apc, and the consequences of potential folding defects in relation to phenotype, are not understood.

Tumors in most Apc mouse models occur mainly in the small intestine, whereas germline mutations of APC in humans result in tumors predominantly in the large intestine (21, 78). The PIRC Apc rat model has tumors in both small and large intestines (9, 13, 79). A pig model with germline Apc mutations was recently reported to develop polyps in the colon (80). In addition to this interspecies variation, mouse models with different germline Apc mutations show different distributions of intestinal polyps. Analysis of Apc^Min/+ mice with different genetic backgrounds has led to the hypothesis that polyp distribution is somehow linked to the mechanism by which the wild-type Apc allele is lost (24). Haigis and colleagues showed that in a B6 background, Apc^Min/+ mice develop polyps mainly in the distal half of the small intestine, and loss of the wild-type Apc allele occurs by means of LOH. In an AKR background, Apc^Min/+ mice develop polyps predominantly at the ileocecal junction, and inactivation of the wild-type Apc allele is achieved through allelic silencing. In the B6 background, Apc^Min/+ mice with additional mutations that inactivate the mismatch repair gene Mlh develop polyps all over the small intestine, and loss of the wild-type Apc allele is achieved through a point mutation. Apc^1638N/+ mice develop polyps in a similar distribution and appear to retain the wild-type Apc allele (24).

Mechanistically, two models have been proposed to explain the connection between polyp distribution and loss of the wild-type Apc allele. In the first model, the molecular machinery in different intestinal regions determines the mechanism of the second Apc “hit” and hence the distribution of polyps. This model is supported by the finding that mice in which the wild-type Apc allele is inactivated by the same mechanism (e.g., Apc^Min/+/Mlh^−/− and Apc^1638N/+) have similar polyp distributions (24). However, the finding that both Apc^1322T/+ and Apc^Min/+ mice lose the wild-type Apc allele through LOH, yet have different polyp distributions, does not support this model. A second model proposes that polyp growth is dictated by the Apc status but also by the particular environment of the different intestinal regions, independent of the mechanism of the second Apc mutation. Supporting this hypothesis, Apc^Δ716/+ mice with an additional mutation of Cdx2 exhibit more colonic and fewer small intestinal polyps. Yet, loss of the wild-type Apc allele occurs via LOH regardless of Cdx2 status (44). Similarly, a colonic shift of polyps has been described in Apc^Min/+ mice with an additional BubR1 mutation, although the mechanism of loss of the wild-type Apc allele in these mice was not reported (43). Mutation of both Cdx2 and BubR1 increases chromosomal instability and changes the proliferation and apoptotic indices in intestines of Apc^Δ714/+ and Apc^Min/+ mice, respectively (43, 44). Further support for the second model comes from Apc^Min/+ mice in a 129/Sv background, where additional mutations that inactivate Smad3 result in increased colonic tumors; yet, in both cases, loss of the wild-type Apc allele is achieved through LOH (81). Finally, PPARγ agonists increase colonic but not small intestinal tumors in Apc^Min/+ mice (82, 83). PPARγ is expressed in higher quantities in the colon and cecum relative to the small intestine, which might account for this differential effect (83).

An expansion of the “just right” hypothesis has been proposed to explain the variation in polyp distribution among patients with FAP and Apc^Min/+ and Apc^1322T/+ mice. The basal level of Wnt signaling is not the same in different intestinal regions. It was proposed that changes in Wnt signaling that result from specific Apc mutations cause optimal Wnt signaling for polyp growth only in certain intestinal regions. On the other hand, in other intestinal regions, these same Apc mutations will result in a higher or lower Wnt signaling level than what is optimal for tumor growth (84).

Perhaps some of these mechanisms can be clarified by studying Apc^Min-FCCC mice, which were generated by mating C57Bl/6J Apc^Min/+ males with Apc^+/+ females from an independent colony of C57Bl/6 mice maintained at Fox Chase Cancer Center (Philadelphia, PA). Apc^Min-FCCC/+ mice develop more colon polyps than do Apc^Min/+ mice, but the molecular basis behind this polyp shift has not been determined (85). Further clarification of the underlying mechanisms that control polyp distribution might also be achieved through careful analysis of Apc^Δ14/+ and Apc^580D/+ mice, which carry similar mutations (truncating the Apc protein at amino acid 580) but appear to have different polyp distributions. Apc^Δ14/+ mice develop more colonic polyps than do Apc^Min/+ mice. Apc^580D/+ mice develop a similar number of colonic polyps as Apc^Min/+ mice, although direct comparison of Apc^580D/+ and either Apc^Δ14/+ or Apc^Min/+ mice has not been reported (75, 86).

Although best known for its role to suppress colorectal tumorigenesis, APC mutations have been seen in other tumors including breast and liver carcinomas (4). In addition, both patients with FAP and rodent models with germline Apc mutations develop extraintestinal phenotypes (see Tables 1–3). As with the intestinal phenotype, the underlying mechanism for variation in extraintestinal phenotypes between patients with FAP and Apc rodent models as well as among different Apc rodent models is not completely understood. Patients with FAP have increased susceptibility to hepatic, pancreatic, thyroid, and brain tumors. They also develop desmoid tumors, dental anomalies, and congenital hypertrophy of retinal pigment epithelium. It is important to note that the penetrance of these extraintestinal phenotypes is variable in patients with FAP (16, 87). The basis behind this variation is not completely understood, although it seems to correlate with the APC germline as well as the acquired somatic mutations. (16, 33).

Apc rodent models also develop some of these extraintestinal manifestations; for example, Apc^1638N/+ mice develop desmoid tumors (72) and PIRC rats show mandibular osteoma (9). Other phenotypes described in patients with FAP have not been reported for Apc rodent models. The short life span of most Apc rodent models could prevent the full expression of some of these phenotypes. On the other hand, Apc rodent models manifest some other extraintestinal phenotypes that have not been described in patients with FAP (Tables 1–3). For example, many mouse models with germline Apc mutations develop mammary tumors. Although APC mutations and promoter methylation have been found in up to 70% of sporadic human breast cancers, patients with FAP do not appear at an increased risk for breast tumors (88–90). In addition, adenoacanthoma is a common type of mammary tumor that develops in Apc mouse models but it has not been reported in humans (91). Other extraintestinal phenotypes described in Apc rodent models include splenomegaly, abnormal hematopoiesis, changes in the serum lipid profile, gonadal changes, cutaneous cysts, and thyroid abnormalities. Differences in physiology, life span, and genetic content between human, mouse, and rat could be underlying causes.

Among different Apc mouse models, some extraintestinal phenotypes, such as anemia and splenomegaly, seem to correlate with the severity of intestinal polyposis. In contrast, mammary gland tumors in Apc mouse models appear to correlate with the severity of polyposis in only a few cases, such as in the Apc^Min/+ and Apc^Δ474/+ models. Very few Apc^Min/+ mice develop mammary tumors, whereas Apc^Δ474/+ mice develop mammary tumors at a rate that is almost double that seen in Apc^Min/+ mice (73, 91). In contrast, there are no reports of mammary tumor development in Apc mouse models with the most severe intestinal polyposis (Apc^Δ714, Apc^1322T, and Apc^ΔSAMP; refs. 32, 40, 65). Perhaps mice with severe polyposis die too early, before mammary tumors have a chance to develop. Apc^1572T/+ mice, which develop no intestinal polyps, have a fully penetrant mammary tumor phenotype in females. K14-cre-Apc^CKO/+ mice are a conditional model in which the Apc^Δ580 allele is expressed only in ectoderm-derived tissues including the mammary gland (75, 92). Mammary tumors from these mice have mutations in the wild-type Apc allele that cluster around codon 1530 consistent with the requirement of an optimal level of Wnt signaling for mammary tumorigenesis (38). It is likely that some of the genetic and environmental factors previously described also account for the variability in extraintestinal phenotypes among different Apc rodent models.

APC research has benefitted greatly from different rodent models with germline Apc mutations. However, genotype–phenotype correlation of these different models is confounded by many genetic and environmental factors. Use of standardized genetic backgrounds and environmental conditions in different laboratories should enable reliable genotype–phenotype analysis of these animals. This standardization will also shed light on the role of different Apc mutations in tumorigenesis. When possible, a direct comparative analysis of different models in the same laboratory will illuminate the contribution of many factors described in this review to phenotypic variation in rodent models with germline Apc mutations.

No potential conflicts of interest were disclosed.

The contents are solely the responsibility of the authors and do not necessarily represent the official views of NIH.

Conception and design: M. Zeineldin, K.L. Neufeld

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Zeineldin

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.L. Neufeld, M. Zeineldin

Writing, review, and/or revision of the manuscript: M. Zeineldin, K.L. Neufeld

This work was supported by grants from the National Cancer Institute (RO1 CA109220), the National Center for Research Resources (P20 RR016475), and the National Institute of General Medical Sciences (P20 GM103418) from the NIH.