Rapid Colorectal Adenoma Formation Initiated by Conditional Targeting of the Apc Gene

See allHide authors and affiliations

Science  03 Oct 1997:
Vol. 278, Issue 5335, pp. 120-123
DOI: 10.1126/science.278.5335.120


Familial adenomatous polyposis coli (FAP) is a disease characterized by the development of multiple colorectal adenomas, and affected individuals carry germline mutations in the APCgene. With the use of a conditional gene targeting system, a mouse model of FAP was created that circumvents the embryonic lethality of Apc deficiency and directs Apc inactivation specifically to the colorectal epithelium. loxP sites were inserted into the introns around Apc exon 14, and the resultant mutant allele (Apc 580S) was introduced into the mouse germline. Mice homozygous for Apc 580S were normal; however, upon infection of the colorectal region with an adenovirus encoding the Cre recombinase, the mice developed adenomas within 4 weeks. The adenomas showed deletion of Apc exon 14, indicating that the loss of Apc function was caused by Cre-loxP–mediated recombination.

Mutations in theAPC gene are responsible for FAP (1), a disease characterized by the development of thousands of colorectal adenomas (1). FAP patients have germline APC mutations, and their tumors show inactivation of the wild-type allele (2). Inactivation of both APC alleles also occurs frequently in sporadic colorectal adenomas (3). These results suggest that APC is a tumor suppressor gene, although the mechanism by which APC mutation leads to transformation of colorectal epithelial cells is largely unknown.

Several mouse lines carrying Apc mutations in their germline have been established as experimental models for FAP (4). For example, the Min mouse, which has a germline mutation in Apc, develops more than 100 intestinal adenomas, a subset of which show inactivation of the second Apc allele (4, 5). However, most of the tumors in the Min mouse develop in the small intestine rather than the colon (5), and it is difficult to study the early events in tumorigenesis in this model.

To study the initiation stage of colon adenoma formation, we have developed an Apc mutant mouse in which inactivation ofApc function is achieved by conditional targeting. This model is based on the Cre-loxP recombination system (6). We first introduced a pair of loxP sites into introns 13 and 14 of the Apc gene by targeted mutagenesis (Fig. 1A) (7). The targeting vector was introduced into embryonic stem (ES) cells by electroporation, and two homologous recombinant ES clones, cl 66 and 76, were obtained (Fig. 1B) (8). The results of Southern (DNA) blot analyses confirmed that these clones carry a mutantApc allele that is produced by homologous recombination with the targeting vector (Fig. 1C) (9). Two independent mutant mouse lines were established from the clones by blastocyst injection and subsequent breeding of chimeras (7).

Figure 1

Establishment of a mutant mouse line carrying a conditionally targeted Apc allele. (A) Structure of the conditionally targeted allele ofApc (Apc 580S) and the mutantApc allele resulting from the deletion mediated by Cre-loxP recombination (deleted allele,Apc 580D). The targeting vector was constructed by inserting one loxP site into intron 13 and the other, with a PGKneo cassette (Neo), into intron 14, resulting in exon 14 (E14) flanked by a pair of loxP sites (open triangles) (7). Positions of PCR primers used for the detection of each allele are indicated (P1 to P5). Positions of probes used for Southern blot analysis with Xba I (X) or Sac I (S) are also shown (probes A to C). (B) PCR analysis of the ES clones (8). DNA of J1 ES cells was used as a negative control (N) and DNA of plasmid pApc3.0 as a positive control (P) (8). Positive signals are evident for clones 66 and 76 (lanes 1 and 5), but not for the negative clones (lanes 2 to 4). (C) Southern blot analysis of ES clones (9). ES cell DNA was digested with Xba I or Sac I and hybridized with three different probes derived from the Apc gene (probes A and C) or a neo cassette (probe B). The results shown for cl 66 (lane 1) and J1 ES cells (lane 2) are representative. (D) Genotypes of F2 offspring obtained by double heterozygous breeding of Apc 580S F1 mice (11). Tail DNAs were analyzed by PCR using primers P3 and P4. The wild-type allele (W) generates a band of 226 bp, and Apc 580S (S) generates a band of 314 bp. The results shown for five offspring—one homozygous mutant (S/S), three heterozygotes (S/W), and one wild-type mouse (W/W)—are representative. (E) Analysis of Apc expression inApc 580S mice by Northern blotting (12). RNA extracted from the brains or colons ofApc 580S homozygotes (S/S) or wild-type mice (W/W) were analyzed. Autoradiograms obtained with two different exposures are shown. The amount of RNA applied was quantified by probing with β-actin cDNA. (F) Immunohistochemical analysis of Apc protein expression in the colon ofApc 580S homozygotes (S/S) or wild-type mice (W/W) (13). Scale bars, 30 μm.

Recombination mediated by the Cre recombinase deletes a region ofApc encompassing exon 14 and induces a frameshift mutation at codon 580. We therefore named this silent mutant alleleApc 580S (Fig. 1A). No tumors were observed inApc 580S heterozygotes after 1 year. Moreover, although homozygous Apc Minmutation typically causes embryonic lethality (10), 66 of the 253 F2 offspring obtained by double heterozygous breeding of F1 mice were identified as homozygous mutants ofApc 580S without any apparent phenotype (Fig. 1D) (11). This percentage of homozygotes (26.1%) is close to that expected from Mendelian inheritance. Northern (RNA) blot analysis revealed that Apc expression in the intestinal tract of homozygous mutants was ∼30% of that in wild-type mice (Fig.1E) (12). This was the only phenotypic alteration observed in Apc 580S mice, and an equivalent reduction in the amount of Apc mRNA was not detected in the brain (Fig.1E). We also analyzed Apc protein expression in the colon ofApc 580S homozygotes by immunohistochemistry and found the staining pattern to be indistinguishable from that of wild-type mice (Fig. 1F) (13). TheApc 580S homozygotes developed normally and had no detectable tumors over a 1-year period. We therefore concluded that the insertion of the loxP sites and the PGKneo cassette into the Apc introns did not impairApc functions in the mice and thatApc 580S could serve as a conditional mutant.

We next established a system that allowed controlled expression of Cre recombinase in the intestinal epithelium ofApc 580S mice. Recombinant adenoviruses are efficient agents for gene delivery; they have broad tissue specificity, and the expression of the delivered gene is transient, because the virus does not integrate into the host genome (14). This feature is important because constitutive expression of the Cre recombinase after recombination may in itself modify the cellular phenotype.

The Cre recombinase gene, driven by the SRα promoter, was introduced into an adenovirus vector from which the E1A and E1B genes had been deleted, and the virus (called AxSRαCre) was propagated in 293 cells, which endogenously express the E1A gene product (15). The function of the virally expressed recombinase was initially analyzed in cultured cells. β-geo 42 cells, established from NIH 3T3 cells by gene trapping, carry a copy of a recombinant provirus containing splice acceptor sequences (SA) and a fusion gene (β-geo) consisting of the neomycin resistance gene (neo) and β-galactosidase gene (lacZ) as the gene-trap machinery, flanked by a pair of loxP sites located in 5′ and 3′ long terminal repeats (LTRs) (Fig. 2A) (16). This cell line was infected with AxSRαCre (17), and the results of Southern blot analysis confirmed the deletion of the provirus from the genome of the infected β-geo 42 cells (Fig. 2B). DNA fragments specific for the expected deletion (4.3 and 6.2 kb) were detected in the AxSRαCre-infected cells when the multiplicity of infection (MOI) was ≥1.25, and almost all cells infected at an MOI of 30 had the expected deletion (Fig. 2B). Examination of lacZ expression (18) in the infected cells revealed that 100% of the cells had lostlacZ expression within 48 hours after infection at a MOI of ≥30 (Fig. 2C). These results indicate that the virus induced Cre-loxP–mediated recombination immediately after infection.

Figure 2

Analysis of AxSRαCre, a recombinant adenovirus expressing Cre recombinase. (A) Structure of the gene trap machinery inserted into the genome of β-geo 42 cells. A provirus, consisting of SA and a β-geo gene flanked by a pair of loxP sites located in LTRs, was integrated in the cellular DNA and expressed by trapping the expression of the endogenous gene (16). Therefore, these cells were blue when stained with 5-bromo-4-chloro-3-indolyl-β-d-galactosidase (X-gal). The region flanked by a pair of loxP sites is deleted as circular DNA from the genomic DNA by the expression of Cre recombinase. This recombination event can be detected by Southern blot analysis using a probe containing the loxP sequence. (B) Southern blot analysis of β-geo 42 cells infected with AxSRαCre. β-geo 42 cells were infected with AxSRαCre at the indicated MOIs (17). Genomic DNA was extracted from cells after incubation for 3 days, digested by Hind III, and analyzed by Southern blotting. (C) Time course of AxSRαCre-mediated DNA deletion, assessed by X-gal staining (18). Samples of β-geo 42 cells were infected with AxSRαCre at the indicated MOIs and were processed for X-gal staining after incubation for 24 or 48 hours. More than 1000 cells in each sample were analyzed, and the number of blue cells was counted.

Before introduction of AxSRαCre intoApc 580S mice, we infected cl 66 cells carrying the Apc 580S allele to confirm that Cre expression would induce the expected deletion (17). Of 20 subclones isolated after infection, 10 had the expected deletion of Apc exon 14, as assessed by Southern blotting and sequence analysis of genomic polymerase chain reaction (PCR) products (19). We next generated a mutant mouse line by injecting one of these clones into blastocysts. Reverse transcriptase–mediated PCR analysis of RNA extracted from intestine and brain of the mutant mouse carrying this deleted allele, Apc 580D, confirmed the deletion of the region encoded by exon 14 from theApc 580D transcripts, and resulted in a frame-shift mutation (19). AllApc 580Dheterozygotes developed multiple intestinal tumors, and Apc 580Dhomozygotes died during embryogenesis (19). These results indicate thatApc function is greatly impaired by the Cre-loxP–mediated deletion.

We next attempted to induce Apc mutation by expressing Cre recombinase in colorectal epithelial cells ofApc 580S homozygotes. For these experiments, we used both AxCANCre (20), an adenovirus that encodes the Cre enzyme with an artificial nuclear localization signal, and AxSRαCre. The viruses were injected into the colorectum ofApc 580S homozygotes through the anus (21), and tumor formation was analyzed by necropsy after 3 months. Homozygous mutants infected with AxCANLacZ (20), a recombinant adenovirus expressing the lacZ gene, did not develop any tumors. Colorectal adenomas developed in 80% of theApc 580S homozygotes infected with AxCANCre (7.1 tumors per mouse on average) and in 60% of those infected with AxSRαCre (2.2 tumors per mouse on average) (Table1).

Table 1

Colorectal tumor incidence in mice infected by rectal infusion with recombinant adenoviruses encoding the Cre recombinase. Tumor incidence is expressed as number of mice with tumors/number infected. ND, not done.

View this table:

To analyze earlier stages of adenoma formation, we next administered AxCANCre to Apc 580S homozygotes, heterozygotes, and wild-type mice, and analyzed them 4 weeks after infection. Colorectal tumors were observed in 20 of 24Apc 580S homozygotes (Table 1) (Fig.3A), but were not seen in any of the 14 heterozygotes nor in any of the 13 wild-type mice. All of the tumors were diagnosed as adenomas by histopathology (Fig. 3B), and immunohistochemical analysis (13) showed that intact Apc protein was not expressed in the adenoma cells (Fig. 3C). Scrape-PCR analysis of genomic DNA from these adenomas confirmed that all adenomas had lost the Apc 580S allele and contained the Apc 580D allele (Fig.4). This finding suggested that Cre-loxP–mediated inactivation of both Apcalleles was responsible for adenoma development.

Figure 3

Analysis of the colorectal region of mice infected with recombinant adenoviruses. (A) Tumors in the colorectal region of an Apc 580Shomozygote. The mouse was infected with AxCANCre (20) and killed after 4 weeks. Four tumors are evident. Scale bar, 3 μm. (B) Histological analysis of one of the tumors shown in (A) by hematoxylin and eosin staining. The typical histopathological appearance of an adenoma is indicated by arrows. Scale bar, 40 μm. (C) Immunohistochemical analysis of an adenoma shown in (B) with antisera to APC (13). The Apc protein is not expressed in the cells of the adenoma (arrows) but is present in normal epithelium (arrowhead). Scale bar, 80 μm. (D) A wild-type mouse was infected with AxCANLacZ, which encodes the lacZdriven by the CAGGS promoter, by injection through the anus. The mouse was killed after 48 hours and the colorectal region was stained for X-gal (18). Scale bar, 3 μm. (E) Histopathological analysis of the colorectal region of anApc 580S homozygote at 10 months after infection with AxCANCre. The invasion of submucosal vessels by adenocarcinoma cells is indicated by an arrow. Scale bar, 240 μm.

Figure 4

Molecular analysis of the mutantApc allele of adenomas. Genomic DNA was extracted from six adenomas (T1 to T6) and adjacent normal tissues (N1 and N2) by scraping paraffin sections of the colorectal region of two differentApc 580S homozygotes infected with AxCANCre; analysis was by PCR using primers P3, P4, and P5 (23). In all adenomas, DNA fragments of 258 bp, specific for Apc 580D (D), are evident with complete disappearance of those of 314 bp, specific forApc 580S (S).

The average number of adenomas of theApc 580S homozygotes treated with AxCANCre was 6.7, all occurring within ∼3 cm of the anal ring. This localization may be a result of our infection procedure. When we infected the colorectal region of wild-type mice (21) with AxCANLacZ, we found that essentially the same region was stained blue (Fig. 3D). About 10 to 20% of the cells displayed enzyme activity, indicating efficient expression of the gene transferred by AxCANLacZ. On this basis, we infer that Apc inactivation by AxCANCre infection may have occurred in a similar percentage of cells. The fact that relatively few adenomas were generated suggests that another event may be necessary for adenoma formation.

When we followed up homozygous mutants that were allowed to live after infection, five of six mice survived more than 1 year, possibly because of the regional localization of adenoma formation. By pathological analysis at necropsy, about 50% of their tumors showed invasion of the submucosal layer by tumor cells (Fig. 3E) and were diagnosed as adenocarcinomas. The conditional gene targeting system described here may allow chronological analysis of adenoma formation in a more precise manner than that allowed by other mouse models.


View Abstract

Stay Connected to Science

Navigate This Article