Tumorigenesis in Mlh1 and Mlh1/Apc1638N Mutant Mice

INTRODUCTION

CRC 5 has emerged as a suitable model to identify genes involved in the onset and progression of cancer. Two types of familial CRC have been studied extensively: (a) FAP is inherited as an autosomal dominant disorder, and individuals with this disorder are normal at birth and progressively develop a large number of intestinal and duodenal tumors. It has been shown that germ-line mutations in the tumor suppressor gene APC are responsible for FAP (1, 2, 3) . These results suggested that APC gene function is important to maintain the balance of normal proliferative potential and loss of intestinal epithelial cells, and APC is, therefore, termed a “gatekeeper” gene (4) . The observation that APC gene mutations are detectable in virtually all cases of sporadic CRC suggested that the mechanism of initiation of CRC involved mutation in the APC gene. (b) HNPCC is also inherited in an autosomal dominant fashion, and HNPCC individuals show a predisposition to colonic and other tumors (5, 6, 7, 8) . An important feature of tumors in HNPCC patients is that they have a characteristic RER⁺ phenotype (5, 6, 7, 8, 9) , which is manifested in the instability of DNA sequences containing short repeat sequences. Germ-line mutations in each of several genes implicated in repair of mismatched nucleotides in DNA have been shown to be responsible for HNPCC. Among the most prominent of these are MSH2 and MLH1, which are homologues of bacterial DNA MMR genes MutS and MutL, respectively. Because DNA MMR is an activity that is intrinsic to all replicating cells, it is important to understand how mutations in DNA MMR genes can cause a cancer predisposition. Knowledge of the mechanism of action of the MMR complex suggests that mutations in the repair genes cause cancer indirectly by increasing the frequency of oncogenic mutations in cells that lack MMR activity.

To understand the role of APC and the MMR genes in the onset and progression of colon cancer and to understand the normal role of these genes in development, differentiation, and cell/tissue homeostasis, we developed a series of mouse lines, each of which carries a mutation in a different gene. The first among these mutations was designated Apc1638N. In this model, we used gene targeting to introduce a neomycin phosphotransferase expression cassette at the position corresponding to codon 1638 of the mouse Apc gene. Mice that are heterozygous for Apc1638N are normal at birth but progressively develop aberrant crypt foci, colonic polyps, and tumors of the small intestine, including the duodenum. The mouse model is akin to human patients with attenuated FAP (10 , 11) .

A second line of mice was developed by targeting the mouse Mlh1 gene. Gene targeting allowed us to delete a portion of the 5′ end of the Mlh1 gene in mouse embryonic stem cells. In an initial report, we showed that the mutation we introduced results in the failure to produce any detectable MLH1 protein. Mice that are homozygous for the Mlh1 mutation are viable and their cells are defective in MMR and exhibit microsatellite instability; mice of both sexes are sterile (12) .

Here, we examined the susceptibility of Mlh1 mutant mice to tumor development. We present data showing that Mlh1 +/− and −/− mice have increased morbidity and develop lymphomas and tumors of the GI tract and a number of other organs. To assess the role of the Apc gene in tumorigenesis in Mlh1 mutant mice, we mated the Mlh1 mutant mice to Apc1638N +/− mice. Here, we show that the GI tumors, indeed, result from inactivation of the Apc gene.

MATERIALS AND METHODS

Mice.

The animals used in this study were generated by crossing Mlh1 heterozygote males and females of the F₁ and F₂ generation to generate homozygous, heterozygous, and wild-type mice. The animals were of a mixed genetic background (C57BL/6 × 129/Ola). To generate Mlh1/Apc1638N mutant mice, we crossed Mlh1 heterozygote F₁ animals with Apc1638N heterozygote animals (in the C57BL/6 genetic background). The Mlh1/Apc1638N double heterozygote offspring were interbred to generate mice homozygous for Mlh1 and heterozygous for Apc1638N.

Genotyping.

The animals were genotyped by PCR of tail DNA. For Mlh1 genotyping, the following primers were used: primer A, TGTCAATAGGCTGCCCTAGG; primer B, TGGAAGGATTGGAGCTGCTG; and primer C, TTTTCAGTGCAGCCTATGCTC. The reaction was performed in a 50-μl reaction mixture containing 100 ng of DNA, 10 ng/μl primer A and 5 ng/μl primers B and C, 1.3 mm MgCl₂, 0.2 mm each dNTP, and 0.5 units of Taq polymerase. Cycling conditions were 5 min at 94°C (1 cycle); 1 min at 94°C, 1 min at 56°C, and 30 s at 72°C (35 cycles); and 5 min at 72°C (1 cycle). The presence of the wild-type allele is indicated by a 350-bp PCR fragment (primers A and C) and the mutant allele by a 450-bp PCR fragment (primers A and B). For Apc1638N genotyping, the following primers were used: primer A, TGCCAGCACAGAATAGGCTG; primer B, TGGAAGGATTGGAGCTGCTG; and primer C, GTTGTCATCCAGGTCTGGTG. The reaction was performed in a 50-μl reaction mixture containing 100 ng of DNA, 4 ng/μl primers A and C and 8 ng/μl primer B, 1.2 mm MgCl₂, 0.2 mm each dNTP, and 0.5 units of Taq polymerase. Cycling conditions were 5 min at 94°C (1 cycle); 1 min at 94°C, 45 s at 58°C, and 30 s at 72°C (28 cycles); and 5 min at 72°C (1 cycle). The presence of the wild-type allele is indicated by a 300-bp PCR fragment (primers A and C) and the mutant allele by a 400-bp PCR fragment (primers A and B).

Analysis of Tumors.

After sacrifice, the GI tract was opened and examined under a dissecting microscope for tumors. In our analysis, duodenum is from the pylorus, the junction of the stomach and the duodenum, through the suspensory ligament of Treitz (∼7 cm in length). The remaining part of the small intestine is divided in half: the upper part is the jejenum and the lower part is the ileum (each region is ∼14 cm in length). Tumors, if found, the GI tract, and other organs, including the lungs, heart, liver, kidneys, and spleen, were removed and fixed in 10% neutral buffered formalin. Representative tissues from the tumors and organs were taken for processing and paraffin embedding, and others were frozen for subsequent DNA analysis. All tissue sections were prepared for H&E stain. The GI tumor tissues were studied for APC protein expression by immunohistochemistry, and the nature of the lymphomas was ascertained by immunotyping. The antibodies used in this study were: C-20, an antibody directed against the COOH terminus of the APC protein (Santa Cruz Biotechnology, Santa Cruz, CA); B-lymphocyte CD45R/B220, an antibody against a B cell-specific CD45 antigen (PharMingen, San Diego, CA); and an antibody against a T lymphocyte-specific antigen CD3 (Vector Laboratories, Burlingame, CA). Avidin-biotin-peroxidase technique was used for immunohistochemical stain. Deparaffinized slides were incubated with primary antibody overnight at 4°C, followed by addition of biotinylated goat antirabbit IgG and avidin-biotin-peroxidase complex. Color was developed in 3,3′-diaminobenzidine.

Microsatellite Instability Analysis.

DNA was extracted from tumor tissue and tails and subjected to PCR. End-labeled primer pairs were used to amplify sequences containing dinucleotide repeats D1Mit36, D7Mit91, D10Mit2, D14Mit15, and D18Mit15 (13) , and two others, JH104 and U12235, were used to amplify sequences containing mononucleotide repeats (primer sequences were as follows: JH104F, 5′-AGGTGATTGTAACGGGCATC-3′, and JH104R, 5′-TATCCTCTCAGTGGTGAGTG-3′; U12235F, 5′-GCTCATCTTCGTTCCCTGTC-3′, and U12235R, 5′-CATTCGGTGGAAAGCTCTGA-3′). Amplified PCR products were separated on a denaturing polyacrylamide gel and autoradiographed for analysis.

Analysis of Apc Mutations.

Protein truncation mutations within the segment of Apc corresponding to codons 677–1233 were identified by a modification of the PTT described previously (14 , 15) . To amplify specifically the wild-type allele of Apc, PCR amplification was first performed using the forward primer 5′-TACAGCACTTGAAATCTCACAG (nucleotides 1991–2012 of mouse Apc; GenBank accession no. M88127) and a wild-type allele-specific reverse primer, 5′-GTTGTCATCCAGGTCTGGTG (nucleotides 5123–5142). Genomic DNA (150–200 ng) from frozen tumor samples was amplified in 20-μl reactions containing 10 mm Tris-HCl (pH 8.6), 50 mm KCl, 0.1% Triton X-100, 1 mm MgCl₂, 0.2 mm dNTP, 0.2 μm primers, and 0.05 units/μl of Taq2000 (Stratagene). Cycling conditions were as described previously (15) , except that the annealing temperature was 60°C, and a total of 25 cycles were used. This reaction gave a 3150-bp product spanning Apc codons 664–1714. One-μl aliquots of these reaction products were reamplified in 50-μl reactions containing the forward primer 5′-CGGGATCCTAATACGA-CTCACTATAGGGAGACCACCATGGATGCATGTGGAACTTTGTGG-3′ and the reverse primer 5′-CGATCGAAGCTTGGACGCAGCTGATTCT-3′, each at 0.3 μm. Reaction conditions were as above, except that a total of 20 cycles were used and the annealing temperature was 57°C. The purified final PCR products were then used as templates in PTT assays performed with the TNT Quick coupled reticulocyte lysate system (Promega, Madison, WI) according to the manufacturer‘s protocol, and [³⁵S]methionine-labeled polypeptides were analyzed by 12% SDS-PAGE and fluorography. For characterization of the tumor-specific mutations, the PCR products were ligated into an SP64 plasmid vector. Individual clones were screened by PTT to identify those carrying mutations, and their DNA sequence was determined.

RESULTS

Mhl1 Mutant Mice Develop a Spectrum of Tumors.

The initial Mlh1 mutation was generated in E14-1 ES cells. The chimeric mice were mated with C57B1/6J (B6) mice. The F1 mice were continuously back-crossed to B6 mice. The analyses of Mlh1 mice were conducted on mice in the second and third back-cross generations. The initial Apc1638N mice were generated in E14-1 ES cells, which were derived from the 129/Ola strain of mice. We developed B6 mice that are congenic for the Apc1638N mutation by 10 or more generations of repeated back-crossing to B6.

Mlh1 heterozygotes were inter-crossed to generate Mlh1 +/+, Mlh1 +/−, and Mlh1 −/− mice. Each of the crosses yielded offspring of all three classes in the expected ratios. We have previously shown that the Mlh1 gene targeting resulted in a null mutant (12) . These results suggest that a reduction or complete absence of MLH1 protein is not detrimental to the normal development of mice.

Mice of the different Mlh1 genotypes were monitored for survival. The results of this analysis are shown in Fig. 1 ⇓ . Both heterozygous and homozygous Mlh1 mutant mice had reduced longevity compared to their wild-type littermates. Fifty % of the +/− mice died by the age of 18 months, whereas it took only 7 months for 50% of the −/− mice to die.

Several of the mice with Mlh1 mutation were sacrificed when they were moribund to assess whether they had developed tumors and, if so, of what origin. The results of this analysis are presented in Table 1 ⇓ .

Seven of 22 (32%) of Mlh1 +/− mice developed tumors, whereas a much larger proportion (13 of 18; 72%) of Mlh1 −/− mice developed tumors. Because some mice died prior to examination, it is possible that we underestimated the tumor frequency. The fact that the +/− mice were moribund at an average age of 9.8 months whereas the −/− mice were moribund earlier (7.0 months) shows that the Mlh1 −/− mice develop tumors much faster than their heterozygous littermates. In both cases, tumors outside the GI tract were as frequent as those within the GI tract. In contrast, 1 of 20 wild-type mice developed a single non-Hodgkin‘s lymphoma at 15 months of age.

We examined the distribution and nature of the tumors in the Mlh1 mutant mice. We were able to detect tumors in the duodenum, jejunum, and colon in the +/− mice, whereas the −/− mice had tumors in all parts of the GI tract tested. These included the stomach, duodenum, jejunum, ileum, and colon. Histological examination of the seven GI tumors indicated that the +/− mice had an equal proportion of adenocarcinomas and early invasive carcinomas, whereas 20 tumors in −/− mice contained 13% carcinomas, 37% early invasive carcinomas, and 50% adenomas.

Tumors outside the GI tract in Mlh1 mutant mice are of different types (Fig. 2) ⇓ . In +/− mice, we examined five tumors: one non-Hodgkin‘s lymphoma, one T-cell lymphoblastic lymphoma, one skin sweat gland carcinoma, two cervical squamous cell carcinomas, and one lung bronchio-alveolar carcinoma. In −/− mice, there was a preponderance of lymphomas (eight of nine) and one (one of nine) skin sweat gland carcinoma. Interestingly, the lymphomas were of either B- or T-cell origin. Of the eight lymphomas, seven were non-Hodgkin‘s lymphomas (five T-cell type and two B-cell type). The remaining lymphoma was classified as Hodgkin‘s disease.

The Combination of Mlh1 and Apc Gene Mutations Decreases Longevity and Increases Tumorigenesis.

We have shown previously that mice with a mutation in the Apc gene (Apc1638N) develop tumors late in the first year of their life and that each mouse develops relatively few (between one and three) tumors. All of these early tumors are in the GI tract. To assess the role of Apc in the GI tumors of Mlh1 mutant mice, we brought the Apc1638N mutation into mice carrying the Mlh1 mutation. Apc1638N heterozygotes were mated with Mlh1 heterozygotes. The double heterozygotes were identified by genotyping and mated again with Mlh1 heterozygotes. This mating provided mice that were wild-type or heterozygous for Apc1638N and that were either heterozygous or homozygous for Mlh1. These two classes of mice were monitored for survival (Fig. 3) ⇓ , and several mice were sacrificed when moribund and examined for the presence of tumors throughout the body. The results of this analysis are presented in Table 2 ⇓ , and a comparison of the tumor incidence with single mutant mice is presented in Fig. 4 ⇓ . Representative GI tumors are shown in Fig. 5 ⇓ .

The presence of the mutant Apc allele resulted in an increased tumor incidence and, in one combination, a significantly lower survival. Only 32% of Mlh1 heterozygotes developed tumors at an average age of 9.8 months whereas 85% of comparably aged double heterozygotes had tumors at an average age of 9.4 months. Perhaps what is remarkable is that Mlh1 −/−, Apc 1638N +/− mice were positive for fecal blood at 2–4 weeks of age and survived for only an average of 3.3 months, at which time all of the mice had tumors. An even more remarkable observation was made when we compared the average number of tumors per mouse in the GI tract. Results of this analysis are presented in Table 3 ⇓ . In Mlh1 heterozygotes, the GI tumor incidence increased 7-fold when the Apc1638N mutation was also present. In Mlh1 −/− mice, the addition of the Apc gene mutation resulted in a 40-fold increase in the number of GI tumors.

Distribution of Tumors in the GI Tract of Mlh1/Apc1638N Mice.

In Mlh1 −/− mice carrying Apc1638N, the increase in tumor number in the stomach and colon was a modest 4–5-fold, whereas the tumor numbers in duodenum, jejunum, and ileum had increased 25–100-fold. Among the GI tumors, we detected adenocarcinomas, early invasive carcinomas, and adenomas. For example, in Apc1638N+/−, Mlh1 −/− mice, of the 64 tumors examined, 22 (34%) were microadenomas, 23 (36%) were adenomas, 8 (13%) were early invasive carcinomas, and 11 (17%) were carcinomas.

The dramatic increase in GI tumor incidence in Mlh1/Apc1638N double mutants was not reflected in the incidence of non-GI tumors. In Mlh1 +/−, Apc +/+ mice, 5 of 22 (23%) had tumors outside the GI tract. Similarly, 3 of 13 (23%) of the Mlh1 +/−, Apc +/− had extra-GI tumors. Of these three, two were non-Hodgkin‘s lymphoma, and one was a skin carcinoma with lung metastasis (Fig. 2) ⇓ . In the Mlh1 homozygotes, 8 of 18 (44%) had extra-GI tumors whereas only 2 of 22 (9%) had similar tumors in the presence of the Apc gene mutation. These results suggest that the pathway for the onset and progression of GI tumors in Mlh1 mutant mice involves the Apc gene but that Apc does not seem to be important in the generation of lymphomas.

Further Evidence for the Involvement of APC in Mlh1 Mutant Mouse Tumors.

Tumors in HNPCC patients characteristically exhibit a RER⁺ phenotype that is manifested by microsatellite instability. If the Mlh1 mutant mice are to serve as true models for human HNPCC, the tumors we have observed must also show a similar RER⁺ phenotype. We examined this feature by isolating DNA from eight lymphoid tumors (three from Mlh1 +/− and five from Mlh1 −/−) and examining the DNA for the status of several microsatellites. We examined a total of seven markers, of which two amplify mononucleotide repeats while the rest amplify dinucleotide repeats. Using this assay, we observed that 43% of the reactions in DNA from Mlh1 +/− and 45% of reactions from Mlh1 −/− tumors showed abnormal size bands, consistent with a RER⁺ phenotype.

In GI tumors in FAP patients as well as sporadic colorectal tumors, inactivating mutations in APC have been observed (16 , 17) . Examination of intestinal tumors in mice with two different mutations of the Apc gene (Min/+, Apc1638N /+) revealed that a most consistent change is loss of the wild-type copy of the Apc gene through nondisjunction (11) . If the presence of Mlh1 mutation increases the incidence of mutations in the Apc gene leading to the onset and progression of cancer, we expect (a) loss of Apc gene expression in a majority of the tumors and (b) presence of point mutations in the unmodified copy of the Apc gene rather than allelic loss. We tested both of these predictions.

When APC protein expression was assayed in tumors from Apc1638N +/−, Mlh1 −/− mice by immunohistochemical methods and visual inspection, we observed that variable number of cells in the different tumor types lost APC expression (Fig. 5C) ⇓ .

To confirm the absence of Apc gene expression at the molecular level and to understand the mechanism of loss of Apc gene function, we tested the Apc gene for mutations. We initially used a PTT to assess for chain termination mutations in the Apc gene (14 , 18) . The reaction conditions we used allowed us to examine the status of the nonmodified Apc allele. We concentrated our efforts in the region corresponding to codons 664–1714 because mutations in this region account for a majority of mutations in human colonic tumors. DNA from five tumor samples was used as a template as described in “Materials and Methods.” The protein products were fractionated on a polyacrylamide gel, and bands were visualized by autoradiography. Results of a typical test are shown in Fig. 6A ⇓ . Each lane contained a band of M_r ∼75,000 corresponding to the full-length product together with smaller size bands common to all samples which may be products of degradation or premature termination. However, some lanes (e.g., T2 and T4) showed a novel product smaller than the wild-type product, suggesting that they result from a mutation in the Apc gene. To confirm this observation, the PCR products that yielded the smaller size polypeptides were cloned into a plasmid vector and individual clones assayed by the PTT test. Results of this analysis are shown in Fig. 6B ⇓ . In each case, clones were recovered which gave a product the same size as that seen in analysis of the corresponding tumor DNA sample (e.g., compare Fig. 6, B ⇓ , Lane 2, with A, Lane T2).

We sequenced the inserts from clones that give the M_r 23,000 product in T2 and the clone that gave the M_r 33,000 truncated product from T4. As shown in Table 4 ⇓ , the clone from T2 contained a C→T transition, resulting in R854Stop. The mutated base was within a stretch of GA dinucleotide repeats. In the second case, the mutation was also in a dinucleotide (CA) repeat, 929ΔCA that resulted in a frameshift leading to a premature chain termination codon. Because loss of Mlh1 gene expression leads to failure to correct single nucleotide as well as insertion/deletion mismatches, our results clearly show that the basis for accelerated tumor formation in these cases is loss of wild-type APC function resulting from the loss of MMR activity.

DISCUSSION

We developed two mouse models to study the mechanism of tumor formation in HNPCC. Mice with mutations in Mlh1, alone or in combination with mutations in Apc, constitute the two models. The majority of human patients with HNPCC are heterozygous for either MSH2 or MLH1 genes. In this report, we show that mice that are heterozygous or homozygous for a null mutation in the Mlh1 gene have a predisposition to cancer. There are differences between the +/− and −/− mice in terms of tumor incidence. Only 32% of +/− mice developed easily identifiable tumors at a mean age of 9.8 months, whereas 72% of −/− mice developed tumors at a mean age of 7.0 months. We have shown previously that cells from Mlh1 heterozygotes are capable of repairing mismatched DNA, whereas cells from −/− mice are not. These results indicate that the loss of MMR activity is probably a prerequisite for initiation of cancer in Mlh1 +/− mice. This notion was confirmed by our observations that tumors from Mlh1 +/− or Mlh1−/− mice show similarly high levels of microsatellite instability. It is of interest to note that older Msh2 +/− mice (19) , in a different genetic background than our Mlh1 mutant mice, developed 40% higher level of tumors compared to wild-type with no effect on survival.

The spectrum of tumors observed in the Mlh1 mutant mice is somewhat different from that seen in HNPCC. Although a large number of tissue types are involved in the tumors of HNPCC patients, lymphoid tumors are not very frequent. However, in the Mlh1 mutant mice, lymphoid tumors predominate the non-GI tumor category. It is also of interest that we have detected both T- and B-cell lymphomas. The presence of a large number of lymphoid tumors is not limited to Mlh1 mutant mice. Msh2 and Msh6 mutant mice also develop lymphomas, although the lymphomas in Msh2 mutant mice are exclusively of T-cell origin (19, 20, 21) , whereas the Msh6 mutant mice develop B- as well as T-cell lymphomas (22) . T-cell lymphomas were also detected in mice with p53 mutations (23) . These results might reflect some intrinsic differences between humans and mice in terms of lymphoma susceptibility. Prolla et al. (24) also observed a tumor susceptibility phenotype in an independently derived Mlh1 mutant homozygote.

Because APC plays an important gatekeeper role in colorectal tumorigenesis, we examined whether the cause of GI tumors in the Mlh1 mutant mice is through loss of APC function. A number of different experimental results support this view. When Mlh1 and Apc double mutants were generated by breeding, we observed that the time required for mortality, the number of tumors per mouse, and the stage of tumors were all affected. If Apc and Mlh1 acted in different pathways, the time of tumor onset would not be expected to change, and the tumor number would be additive. If inactivation of the Apc gene is an important prerequisite for tumor development, the time of onset and the number of tumors would be expected to change. This was, indeed, what we observed. In Mlh1 −/− mice, a mutated Apc gene resulted in a very significant (40-fold) increase in tumor numbers. These results clearly show that Apc gene mutation is an important early step in the onset of GI cancer in Mlh1 mutant mice.

Our results also showed that the mechanism of GI tumor development and lymphoid tumor development are distinct. When the Apc mutation was crossed into the Mlh1 mutant background, the lymphoid tumor incidence was unaffected. This observation, together with the fact that Apc mutant mice do not develop lymphoid tumors, suggests that the Apc gene does not play a role in lymphomagenesis in these mice. Alternatively, the double mutant mice could succumb to the GI tumors before the lymphoid tumors have an opportunity to develop.

Loss of Apc gene function is an important and early event in the development of GI tumors in mice and humans (25) . Tumors in the Mlh1/Apc1638N mice also follow the same pattern. We have shown the loss of Apc gene function by histochemical and molecular biological methods. However, the mechanism of loss of Apc gene function is different in Apc1638N and Mlh1/Apc1638N mice. We have previously shown (11) that, in Apc1638N mice, loss of Apc gene function is mediated through loss of the wild-type copy of Apc gene, apparently through chromosomal nondisjunction. In contrast, the tumors in the Mlh1/Apc1638N mice show mutations in the wild-type copy of the Apc gene. This mechanism is consistent with the known function of Mlh1 in DNA MMR.

The Apc1638N mice we developed are a good model for the attenuated form of FAP. Compared to Min/+ mice, Apc1638N mice develop fewer tumors and at a later time; these tumors progress into adenocarcinomas. As a result, the mice live longer, providing an opportunity to examine the role of environmental factors, such as diet, on the onset and progression of CRC. The combination of Mlh1 and Apc1638N mutations results in a large number of tumors at an early age. Although, in this respect, these mice are similar to Min/+ (26) , they have important differences. These mice are a model of HNPCC because of the MMR phenotype associated with them, whereas the Min/+ are not. Another significant difference is that the Min/+ mice do not develop carcinomas. It has been suggested that the number of tumors and their time of onset would preclude them from becoming carcinomas. However, the Mlh1/Apc1638N mice have a very high incidence of carcinomas. Therefore, these mice are uniquely suited for examining the development of carcinomas. Mice that are double mutant for Min and Pms2, a close relative of Mlh1, have been reported recently (27) . Although these mice show increased numbers of adenomas, they do not seem to progress to carcinomas. This difference also makes the Mlh1/Apc1638N mice excellent models for the study of the full spectrum of progression of the GI tumors.