An Msh2 Conditional Knockout Mouse for Studying Intestinal Cancer and Testing Anti-cancer Agents

Abstract

Background & Aims

Mutations in the DNA mismatch repair (MMR) gene MSH2 cause Lynch Syndromes I & II, and sporadic colorectal cancers (CRCs). Msh2^null mice predominantly develop lymphoma and do not accurately recapitulate the CRC phenotype.

Methods

We generated and examined mice with a conditional Msh2 disruption (Msh2^LoxP), permitting tissue-specific gene inactivation. ECMsh2^LoxP/LoxP mice carried an EIIa-Cre transgene and VCMsh2^LoxP/LoxP mice carried a Villin-Cre transgene. We combined the VCMsh2^LoxP allele with either Msh2^Δ7null (VCMsh2^LoxP/null) or Msh2^G674D mutations (VCMsh2^LoxP/G674D) to create allelic phase mutants. These mice were given cisplatin, or 5-fluorouracil/leucovorin and oxaliplatin (FOLFOX) and their tumors were measured by magnetic resonance imaging.

Results

Embryonic fibroblasts from ECMsh2^LoxP/LoxP mice do not express MSH2 and are MMR-deficient. Reverse transcription, PCR, and immunohistochemistry from VCMsh2^LoxP/LoxP mice demonstrated specific loss of Msh2 mRNA and protein from epithelial cells of the intestinal tract. Microsatellite instability (MSI) was observed in all VCMsh2 strains and limited to the intestinal mucosa. Resulting adenomas and adenocarcinomas had somatic Apc truncation mutations. VCMsh2^LoxP/LoxP mice did not develop lymphoma. Comparison of allelic phase tumors revealed significant differences in multiplicity and size. When treated with cisplatin or FOLFOX, tumor size was reduced in VCMsh2^LoxP/G674D but not VCMsh2^LoxP/null tumors. The apoptotic response to FOLFOX was partially sustained in the intestinal mucosa of VCMsh2^LoxP/G674D animals.

Conclusion

Msh2^LoxP/LoxP mice in combination with appropriate Cre recombinase transgenes have excellent potential for preclinical modeling of Lynch Syndrome, MMR deficient tumors of other tissue types, and use in drug development.

Background & Aims

Approximately 150,000 new cases of colorectal cancer (CRC) are diagnosed per year in the United States. More than 50,000 patients die from it yearly. Generally classified into familial predisposition syndromes and sporadic cancers, several critical genes involved in both have been identified. Familial adenomatous polyposis (FAP) is caused by mutations in the APC gene. Lynch Syndromes I & II are caused by mutations in the Mismatch Repair (MMR) genes. MSH2 was found to be one of the most commonly mutated MMR genes ^1-3. Msh2 is necessary for repair of base-base as well as insertional deletion mismatches and its absence results in increased mutation levels. Mice lacking MSH2 have a tumor disposition phenotype.

To develop mouse models for Lynch Syndrome three Msh2^null knockout mouse lines have been generated, two by targeted disruption of Msh2 exon 12 ^{4, 5} and one by disruption of exon 7 ⁶. Homozygous mutant mice of all three Msh2^null knockouts are MMR-deficient and display a highly increased predisposition to lymphoma. A proportion of older animals also develop intestinal neoplasms that are associated with Apc inactivation ⁷. However, the predominance of the lymphoma phenotype has limited the use of these animals as preclinical models.

We report a novel conditional knockout mouse model for the tissue-specific inactivation of Msh2 (Msh2^LoxP). In this model, MMR can be inactivated by Cre-LoxP-mediated inactivation of Msh2 in different tissues by the expression of various Cre-recombinase transgenes. To constitutively inactivate MMR similar to Msh2^null knockout mice we mated Msh2^LoxP mice with EIIa-Cre recombinase transgenic mice (termed ECMsh2^LoxP). To specifically inactivate MMR in the intestinal mucosa we combined the Msh2^LoxP allele with the Villin-Cre transgene (VCMsh2^LoxP). ECMsh2^LoxP/LoxP mice display complete MMR deficiency and have a cancer phenotype similar to Msh2^null knockout mice. In contrast, in VCMsh2^LoxP/LoxP mice MMR deficiency is limited to the intestinal epithelium and the mice develop exclusively intestinal neoplasms. These data show that Msh2^LoxP mice in combination with specific Cre recombinase transgenes allow the tissue-specific inactivation of MMR and the development of suitable mouse models for Lynch Syndrome.

We also demonstrate that it is possible to study allelic effects of different Msh2 mutations on intestinal tumorigenesis in VCMsh2^LoxP mice by combining the Msh2^LoxP allele with either a Lynch Syndrome related missense mutation (Msh2^G674D) or an Msh2^Δ7null mutation (Msh2^null). Tumors from these allelic phase mutants have also been tested for their response to two chemotherapeutic regimens, cisplatin and FOLFOX, and their growth recorded by Magnetic Resonance Imaging (MRI). Although some tumors in VCMsh2^LoxP/null mice were responsive to the two drugs, the majority were resistant to both chemotherapies. In contrast, almost all VCMsh2^LoxP/G674D tumors were found to generally respond well to cisplatin and FOLFOX. The differences in responsiveness of tumors correlated with the absence of a significant DNA damage response in VCMsh2^LoxP/null mice, and partial retention of this response in VCMsh2^LoxP/G674D mice.

Methods

Generation of Msh2^LoxP Mice

The targeting vector for the Msh2^LoxP mouse was made by recombinogenic methods ^{8, 9}. An Msh2 genomic fragment spanning exon 10 through intron 18 was PCR amplified from BAC clone 183K13 (RP-22 library) and subcloned. A LoxP site was introduced into Msh2 intron 12-13 followed by introduction of a LoxP-FRT-PGKneo^r-FRT selection cassette into Msh2 intron 11-12. The vector was linearized and transfected into WW6 ES cells ¹⁰. Male chimeric mice were generated and bred to C57BL/6J females to generate Msh2neo^{LoxP-FRT neo/+} F1 offspring. The PGKneo^r cassette was subsequently deleted in vivo by crossing Msh2neo^{LoxP-FRT neo/+} heterozygotes to FLP- deleter mice ¹¹. Offspring from these crosses were genotyped by PCR, Southern Blot and sequence analyses (data not shown), to confirm the integrity of the Msh2^LoxP allele. All procedures were in accordance with Institutional Animal Care and Use Committee Protocols.

Generation of Msh2^LoxP Cre Recombinase Transgenic Mouse Lines

Msh2^LoxP/+ mice were crossed with EIIa-Cre recombinase transgenic animals to generate ECMsh2^{LoxP/+ 12}. Heterozygotes were intercrossed to generate ECMsh2^LoxP/LoxP mice.

Msh2^Lox/p/+ mice were mated with B6;D2-Tg(Vil-Cre) to create VCMsh2^LoxP/+ mice, then intercrossed to create VCMsh2^LoxP/LoxP mice ¹³. VCMsh2^LoxP/+ mice were also mated to animals carrying the Msh2^Δ7 knockout allele (termed Msh2^null) ⁶ and the Msh2^G674D knock-in allele. Offspring with one floxed Msh2 allele and one mutant allele, VCMsh2^LoxP/null or VCMsh2^LoxP/G674D respectively were obtained.

PCR Genotyping Msh2^LoxP mice

Tail DNA was isolated using the DNAeasy kit (Qiagen, Valencia, CA) from ten day old mice. PCR primers used for genotyping were 184F (TACTGATGCGGGTTGAAGG), 184R (AACCAGAGCCTCAACTAGC), and 165R (GGCAAACTCCTCAAATCACG). Cycling conditions will be given upon request.

MMR analysis in ECMsh2^LoxP/LoxP MEF cell lines

Cytosolic extracts were prepared from MEF cells as described in Thomas et al. ¹⁴. A heteroduplex G-G substrate was prepared and DNA repair reactions were performed as previously described ^{14, 15}.

Western Blotting and Immunohistochemistry

MEF cell extracts were separated by SDS-PAGE and blotted onto Polyvinylidene fluoride membranes and probed with rabbit anti mouse MSH2 polyclonal antibody (MSH2 N-20:sc494, Santa Cruz, CA), an Msh6 monoclonal antibody (BD Biosciences, Franklin, USA) or a GADPH monoclonal antibody (Ambion for GADPH, Austin, TX).

For immunohistochemical analysis (IHC), monoclonal antibodies directed against Msh2 (N-20:sc494, Santa Cruz) and Apc (GTX15270, GeneTex, Inc.) were used.

Generation of Kaplan-Meier Survival Plots

Prism 3.0 software (Graphpad) was used to calculate percent survival of animals.

Histopathologic analysis

Mice were euthanized and the GI tract was removed, opened longitudinally, and fixed in 10% neutral-buffered formalin or Bouins solution. The number of tumors and their location was recorded under a dissecting microscope. For histological analysis, tumors were embedded in paraffin, sectioned to 5 μm and stained with hematoxylin and eosin. Relative tumor size was measured using a Vernier Caliper with fine adjustment.

MSI Analysis

Genomic DNAs from tail, spleen and flat mucosa were subjected to PCR amplification using a dilution assay as previously described ¹⁶. We screened for instability using a dinucleotide repeat marker, D17Mit123 ¹⁷. In intestinal tumors two dinulceotide markers were studied in undiluted DNA, D7Mit91 and TG27. PCR products were separated on denaturing 6% polyacrylamide gels and autoradiographed for analysis.

Apc Truncation Mutations

The analysis of truncation mutations to the Apc gene was performed as described earlier ¹⁸.

Drug Treatment

VCMsh2^LoxP/null and VCMsh2^LoxP/G674D animals were divided into 3 groups. The first group received an intraperitoneal injection (i.p.) with cisplatin (20 mg/kg body weight), five times every second day with a total dose of 100 mg/kg body weight. The second group received an i.p. injection with FOLFOX (5-fluorouracil/leucovorin, five sequential days (20 mg and 10 mg/kg body weight respectively); Oxaliplatin (1 mg/kg body weight) was injected i.p. once. The third group was a control group injected i.p. with Phosphate Buffered Saline (PBS) five times every second day. All drugs were purchased from Sigma-Aldrich Corp. (St. Louis, MO).

Magnetic Resonance Imaging (MRI) & 3-D Reconstitution of Images

Tumor sizes were measured before and after treatment by in vivo MRI imaging. Animals were positioned in a 40 mm “bird cage” MRI coil in a 9.4 T GE Omega vertical bore imaging system. A 51.2 mm field of view with a 256 × 256 pixel image matrix was used. Image slices were 1 mm thick with no gap between slices. Series of routine spinecho images along all three planes were acquired to reconstruct 3D images of the mouse. Typical parameters for GI tract studies were used with echo time of 30 ms, repetition time of 400 ms, and signal averaging 4 scans. Typically 24 images were acquired along each plane. Image data was analyzed using MATLAB based software. 3D reconstructions of the digestive tract were created using Amira 3.1 software.

DNA Damage Response

TUNEL assays were conducted on intestinal and spleen tissue from VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice, (Promega DeadEnd (TM) Fluorometric TUNEL System).

Results

The Generation of Msh2^LoxP Mice, ECMsh2^LoxP/LoxP Cell Lines, and MMR Measurement

We have generated a conditional knockout mouse line for Msh2 by flanking exon 12, encoding a portion of the essential ATPase domain of MSH2, with LoxP sites (Fig. 1A, B, & C). To constitutively delete exon 12, Msh2^LoxP/+ were mated with EIIa-Cre recombinase mice. In these mice, the Msh2^LoxP allele was transmitted in a normal Mendelian ratio, and ECMsh2^LoxP/LoxP mice obtained upon heterozygote intercrosses also developed normally. Western blot analysis of ECMsh2^LoxP/LoxP MEFs revealed that deletion of exon 12 resulted in the complete loss of MSH2 protein, and also reduced levels of MSH6 protein (Fig. 2D).

Figure 1.

Strategy for the production of Msh2 conditional knockout mutant mice, and PCR genotyping of offspring. (A) Gene targeting strategy. (B) PCR genotyping strategy. 1) pGEM markers, 2) Wildtype mouse, 3) Msh2^LoxP/LoxP, exon 12 deleted, 4) Msh2^LoxP/+. (C) Where Msh2 exon 12 has been deleted, primer pair 184F/184R amplifies no product, whereas primer pair 184F/165R amplifies a 340 bp product. Wildtype animals without the conditional allele amplify a 210 bp PCR product using primers 184F/184R.

Figure 2.

Molecular characterization of VCMsh2^LoxP/LoxP and ECMsh2^LoxP/LoxP mice. (A) Specific rearrangement of the Msh2 gene in the small and large intestines of VCMsh2^LoxP/LoxP animals, is absent in the kidney heart and spleen. Primer pairs A/C amplify a 340 bp product when exon 12 is deleted (intestine) and a 983 bp product when exon 12 is intact (kidney heart and spleen). (B) Measurement of MMR in ECMsh2^LoxP/LoxP MEFs. Cell lines EM2-1 and EM2-2 are MMR deficient and compare to the Exo1^-/- MMR deficient cell line, EM2 and Exo1^-/- cell lines complement each other. (C) MSI in VCMsh2^LoxP/LoxP mice (for MSI in C57B1/6J mice see 19). (D) Western blot analysis of EM2 MEF cell lines. (Δ12), MSH2 is absent from an EM2 cell line and shows reduced amounts of its complex partner, Msh6. WT, wildtype mouse embryonic fibroblast cell line. (E) IHC on VCMsh2^LoxP/LoxP small intestine (top) compared to wildtype intestine (bottom). (F) RT-PCR using Msh2 primers with GI tissues 1) wildtype, 2) VCMsh2^LoxP/LoxP.

To investigate the effect of Msh2 exon 12 deletion on MMR, cell extracts from two MEF lines (ECMsh2^LoxP/LoxP lines 1 and 2) were screened for their ability to repair a G-G mismatch from a 3′ nick. Neither extract repaired this single base mismatch. Repair deficiencies in both ECMsh2^LoxP/LoxP cell extracts were similar to repair deficiencies in an Exo1^Δ6/Δ6 (exon 6 deletion) extract, previously found defective for repair of single base mismatches ¹⁹. The MMR defect in both ECMsh2^LoxP/LoxP extracts was complemented by adding Exo1^Δ6/Δ6 extract, demonstrating that it was caused by the specific loss of MSH2. After complementation of the ECMsh2^LoxP/LoxP extracts by the Exo1^Δ6/Δ6 extract, a reduction in the blue/white plaque color ratios was found, demonstrating that repair activity in the complemented extracts is directed to the nicked strand, indicating strand-specific MMR (Fig.2B).

Inactivation of Msh2 by Villin-Cre in VCMsh2^LoxP Mice

To inactivate Msh2 in the intestinal mucosa, Msh2^LoxP/+ animals were intercrossed with Villin-Cre transgenic mice ¹³. Msh2 inactivation was confirmed by PCR (Fig. 2A) and RT-PCR (Fig. 2F). Genomic DNA isolated from kidney, heart and spleen displayed either minimal or no exon 12 deletion.

MSH2 expression was examined by immunohistochemistry (Fig. 2E). It was detectable in the cytoplasm and nucleus of epithelial cells in wild-type mice, but was absent in the intestinal mucosa of VCMsh2^LoxP/LoxP mice.

To study the effect of exon 12 deletion on MMR in vivo we analyzed MSI in different tissues of VCMsh2^LoxP/LoxP mice (Fig. 2C). At the D17Mit123 locus only 5 of 77 (6.5%) alleles were unstable in genomic tail DNA of VCMsh2^LoxP/LoxP mice and 6 of 82 (7.3%) were unstable in spleen genomic DNA which is comparable to the levels of MSI in wild-type mice for this marker¹⁹. However, epithelial cells in the intestinal mucosa of VCMsh2^LoxP/LoxP mice displayed a significant increase in MSI with 20 of 96 unstable alleles (20.8%) indicating that the MMR-deficiency is highly restricted to the intestinal epithelium in VCMsh2^LoxP/LoxP mice.

Survival and Tumor Development in ECMsh2^LoxP/LoxP and VCMsh2^LoxP/LoxP Mice

Both ECMsh2^LoxP/LoxP and VCMsh2^LoxP/LoxP mice were viable and fertile. However, a significant difference in survival was observed between the two mouse lines (p < 0.0001) (Fig. 3B). The median survival for VCMsh2^LoxP/LoxP animals was 12 months and all animals expired at 17 months. In contrast, ECMsh2^LoxP/LoxP mice had a reduced median survival of 6 months and all animals died by 11 months similar to Msh2^null/null knockout mice ^4-6.

Figure 3.

VCMsh2^LoxP/LoxP mice have decreased median survival. (A) An intestinal adenocarcinoma from a VCMsh2^LoxP/LoxP mouse. (B) Survival curves: wildtype, (green); VCMsh2^LoxP/LoxP, (red); ECMsh2^LoxP/LoxP, (blue, solid); Msh2^null (black, dashed). (C) An intestinal adenocarcinoma from a VCMsh2^LoxP/LoxP mouse stained with rabbit anti mouse E cadherin -24E10 (Cell Signaling Technology, Danvers, MA). The black square on the lower right shows tumor invasion of the muscularis and is enlarged in panel D). The smaller square on the lower right is enlarged in panel E) and shows multiple mitotic figures indicated by black arrows. F) IHC using antibody to Apc on a VCMsh2^LoxP/LoxP intestinal polyp (top), specific staining of Apc in macrophages of the intestinal epithelium (lower left), wildtype control (lower right).

The analysis of several moribund ECMsh2^LoxP/LoxP mice at 6 to 7 months of age revealed a high incidence of lymphoma (67% of animals) and a lower incidence of small intestinal tumors (33% of animals) indicating that the cancer predisposition phenotype in ECMsh2^LoxP/LoxP mice is comparable to that of Msh2^null/null knockout mice ^{4, 5} (Table 1).

Table 1. VCMsh2 ^LoxP/LoxP Tumor Incidence and Multiplicity.

Genotype	N	Age (mo) Mean ± SD	Sex (M:F)	Overall	Intestinal tumors n (%)		Intestinal Tumor Multiplicity (Mean ± SEM)
					Lymphoma	Intestine
VCMsh2 LoxP/LoxP	18	9.0 ± 1.1	1:0.8	16 (89)b	0 (0)	16 (89)b	1.61 ±0.30c
VCMsh2 LoxP/+	15	11.3 ± 1.8	1:1.5	0 (0)b	0 (0)	0 (0)b	0.00±0.00c
Msh2+/+ a	17	14.2 ± 2.7	1:04	l (6)b	0 (0)	l (6)b	0.06±0.00c

N, number of mice studied; n, number of mice with intestinal tumors.

^aMsh2 ^+/+ mice, VC positive and VC negative combined.

^bTumor incidence compared by Fisher exact probability: P<0.0001.

^cTumor multiplicity compared by Mann-Whitney or binomial calculation: P<0.0001.

In contrast, the cancer phenotype differed significantly between VCMsh2^LoxP/LoxP mice and Msh2^null/null knockout mice. A cohort of VCMsh2^LoxP/LoxP mice (n=18) was sacrificed at 9.0±1.1 months of age and analyzed for the presence of tumors. 89% of VCMsh2^LoxP/LoxP mice developed tumors in the small intestine at this age with a tumor multiplicity of 1.6±0.3 (Table 1). Histopathological analysis of 14 tumors showed that 50% were at the adenomas and the other 50% were highly invasive adenocarcinomas (Fig. 3A, C, D, & E). None of these animals developed lymphoma and only one in 150 VCMsh2^LoxP/LoxP mice developed lymphoma after 12 months of life. VCMsh2^LoxP/+ mice (n=15) at 11.3±1.8 months of age developed no tumors and 1 of 17 (6%) wild-type mice at 14.2±2.7 months of age developed an intestinal tumor (Table 1). We also analyzed the MSI status in the genomic DNA of the intestinal tumors in VCMsh2^LoxP/LoxP mice at two dinucleotide markers. We found that at the D7Mit123 marker 7 of 12 (58%) tumors tested were unstable, while at the TG marker ¹⁶ 11 of 12 (92%) tumors were unstable. All tumors showed MSI of at least one of the two markers tested.

Apc is Mutated in VCMsh2^LoxP/LoxP GI Tumors

Immunohistochemistry of intestinal tumors from VCMsh2^LoxP/LoxP mice were negative for Apc antibody staining (Fig. 3F).

Truncation mutations in VCMsh2^LoxP/LoxP GI tumors were found by IVTT analysis indicating somatic mutations to both Apc alleles may have occurred for tumor initiation (Table 2). Twenty out of thirty-one (64.5%) VCM2^LoxP/LoxP tumors screened were positive. Ten fragments were cloned and sequenced. Mutation types included mono and di-nucleotide deletions and C to T transitions leading to the formation of stop codons. These mutations are consistent with the types of mutations caused by loss of MMR function ¹⁸.

Table 2. Truncation mutations to Apc in tumors from VCMsh2 ^LoxP/LoxP.

Allelic Phasing of Intestinal Tumor Development in VCMsh2^LoxP Mice

The Msh2^LoxP allele provided an opportunity to examine the impact of different Msh2 mutations on intestinal tumor development. We generated VCMsh2^LoxP mice that carried the Msh2^LoxP allele in combination with either an Msh2^null (12) allele or the Msh2^G674D allele previously found in an HNPCC patient ²⁰ (Supplemental data). Deletion of exon 12 in the Msh2^LoxP allele by Villin-Cre expression leads to intestinal epithelial cells containing two Msh2^null alleles (that express no MSH2) in VCMsh2^LoxP/null mice or one Msh2^null and one Msh2^G674D allele (that only express MSH2^G674D) in VCMsh2^LoxP/G674D mice. The analysis of tumorigenesis in these mice revealed remarkable differences (Fig. 4 A & B). Mice of both lines displayed a strong predisposition to intestinal cancers. However, the tumor onset was delayed in VCMsh2^LoxP/G674D mice as compared to VCMsh2^LoxP/null mice. In VCMsh2^LoxP/null mice intestinal tumors were found beginning at 6 months of age and 50% of animals carried intestinal tumors at 10 months of age. In contrast, intestinal tumors were first detected in VCMsh2^LoxP/G674D mice at 10 months of age and 50% of animals carried tumors at 13 months of age (p<0.0001). In addition, the tumor number and size differed significantly between the VCMsh2^LoxP/null and VCMsh2^LoxP/G674D allelic phase mutants. While VCMsh2^LoxP/null mice developed 1.40±0.11 intestinal tumors, the VCMsh2^LoxP/G674D mice developed 3.43±0.42 tumors (p<0.0001). There was also a significant difference in tumor size. Tumors in VCMsh2^LoxP/null mice had an average diameter of 6.48±0.58 mm, while the tumors in VCMsh2^LoxP/G674D mice were significantly smaller with a size of 3.25±0.49 mm (P<0.003). In VCMsh2^LoxP/null mice 18% (3/17) of the intestinal tumors were adenomas and 82% (14/17) of tumors had progressed to adenocarcinoma. In VCMsh2^LoxP/G674D mice 37% (7/19) of intestinal tumors were adenomas, while 63% (12/19) were adenocarcinomas.

Figure 4.

Tumor size measurements by caliper and tumor number count, from the intestines of allelic phase mutants. (A) The average number of tumors between VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice varied significantly, as did the size (B). (C) Intestinal tumors for chemotherapy are visualized and measured by MRI. A six month old VCMsh2^LoxP mouse testing positive for occult blood, was subjected to MRI, successfully revealing one tumor. Two weeks later a second tumor was detected during a second MRI at the original location.

VCMsh2^LoxP/null and VCMsh2^LoxP/G674D Response to Cisplatin and FOLFOX

Allelic phase mutant mice were subjected to chemotherapy with either cisplatin or FOLFOX, and tumor response was analyzed by MRI (Fig. 4C). Tumors in both VCMsh2^LoxP/null and VC Msh2^LoxP/G674D mice continued growing after receiving Phosphate Buffered Saline (PBS) injections (i.p.). The two groups, however, differed in response to cisplatin or FOLFOX treatment. VCMsh2^LoxP/null tumors were predominantly resistant to both chemotherapies, and only a small number of tumors showed growth retardation. In contrast almost all tumors in VCMsh2^LoxP/G674D mice responded well to either cisplatin or FOLFOX treatment regimen (Fig. 5).

Figure 5.

Intestinal tumors from VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice, respond to chemotherapy. Tumors are measured in terms of relativity of MRI measurements. The tumor size at day 0 is 1, and the relative tumor growth or retardation is scored on the basis of percentage. Red lines indicate growth, green lines indicate retardation. The number of tumors for each treatment is as follows. VCMsh2^LoxP/null: PBS-7, cisplatin-8, FOLFOX-11. For VCMsh2^LoxP/G674D: PBS-7, cisplatin-9, FOLFOX-11.

To determine the molecular basis underlying the differences in intestinal tumorigenesis and drug response in VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice we performed in vivo analyses of MMR and the DNA damage response in the intestinal epithelium. While genomic DNA in the intestinal mucosa of wild type mice was previously shown to contain 4% unstable alleles at the dinucleotide marker D7Mit91²¹, genomic DNA in the mucosa of VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice displayed a significant increase in MSI at this marker that was comparable between both mouse lines (26/81 (32%) unstable alleles in VCMsh2^LoxP/null mice; 30/89 (31%) unstable alleles in VCMsh2^LoxP/G674D mice).

To determine whether VCMsh2^LoxP/G674D mice sustain the MMR-dependent DNA damage response function post chemotherapy, the number of apoptotic cells in intestinal crypts was analyzed 18 hours after FOLFOX injection (Fig. 6). These measurements showed a significant reduction in the apoptotic response in VCMsh2^LoxP/null mice as compared to Msh2^LoxP/LoxP (wild-type) mice (p<0.001). Although the apoptotic response in the mucosal epithelium of VCMsh2^LoxP/G674D mice was somewhat reduced in comparison to wild type mice, it was significantly higher than in VCMsh2^{LoxP/null mice} (p=0.0118). As expected the apoptotic response appeared normal in the spleen in all three groups of mice (p=0.2929). Overall, these studies indicate that although the intestinal epithelial cells in VCMsh2^LoxP/G674D mice display only a partial DNA damage response, it is still sufficient to significantly impact tumor growth after FOLFOX treatment.

Figure 6.

Apoptosis measured in intestinal mucosa of VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice, after 18 hours of FOLFOX treatment. (A) TUNEL staining (green fluorescence) in the intestinal mucosa of a. Msh2^LoxP/LoxP, b. Msh2^null, c. VCMsh2^LoxP/null, d. VCMsh2^LoxP/G674D mice with graphic analysis of the number of positive apoptotic cells per fifty crypts. (B) Apoptosis measured in the spleen of VCMsh2^LoxP/null and VCMsh2^LoxP/G674D mice, after 18 hours of FOLFOX treatment. A. TUNEL staining (green fluorescence) in the intestinal mucosa of a. Msh2^LoxP/LoxP, b. Msh2^null, c. VCMsh2^LoxP/null, d. VCMsh2^LoxP/G674D mice with graphic analysis of the number of positive apoptotic cells per fifty crypts.

Conclusions

We have generated a conditional Msh2^LoxP knockout allele in mice that permits better modeling of the intestinal cancer features of Lynch Syndrome than previous models. Although these earlier mouse models are prone to a variety of cancers, they predominantly develop aggressive lymphomas early in life which has limited their use as preclinical models ^4-6. Using the Msh2^LoxP allele in combination with the Villin-Cre transgene we show that the lymphoma phenotype of constitutional Msh2 knockout mice can be avoided and tumorigenesis restricted to the intestinal tract.

VCMsh2^LoxP/LoxP intestinal tissues display a high degree of MSI and tumors carry Apc mutations. A few documented Lynch Syndrome families exist with biallelic mutations in MMR genes ^22-24. These patients display severe reduction in life span and hematological malignancies, a phenotype that resembles neurofibromatosis, as well as CRCs. Since MSI can also be found in many solid tumors ²⁵, the Msh2^LoxP allele should be helpful in modeling these types of MSI positive cancers. We have found, for instance, that the Msh2^LoxP allele in combination with the Cre transgene under control of the human epithelial keratin 14 promoter (K14-cre) permits the generation of skin tumors (data not shown). The identification of tissue specific pathways altered in these tumors may be useful in identifying tissue specific cancer genes.

Similar to other mouse models of colorectal cancer, VCMsh2^LoxP/LoxP mice develop tumors predominantly in the small intestine, in contrast to Lynch Syndrome patients who typically develop tumors of the colon. Inactivation of an Apc^LoxP allele by Villin-cre also resulted mainly in small intestinal tumors. However, the tumor location could be shifted to the large intestine by colonic infection of Apc^LoxP/LoxP mice with Adenoviral-Cre (personal communication, Dr. Kenneth Hung). We are currently infecting the intestines of Msh2^LoxP/LoxP mice with Adenoviral-Cre to determine if this method may also be useful in permitting development of CRC without prior mutation to the Apc gene. Genetic approaches may also be useful in these experiments such as the introduction of retinoblastoma deficiency into Msh2^LoxP/LoxP mice, followed by infection with Adenoviral-Cre. While Rb deficiency by itself does not cause CRC ²⁶, it has been demonstrated to expand the compartment that tumors occur in to include the cecum and distal colon in Apc deficient mice ²⁷. We also combined the Msh2^LoxP allele with the Cdx2P-NLS-Cre transgene²⁸ and observed a shift of tumors to the large intestine. However, like Msh2^null mice, these animals also developed lymphomas (Kyeryoung Lee and Winfried Edelmann, unpublished observations).

Despite the tumor location in the small intestine, VCMsh2^LoxP mice have been highly useful in determining the effect of different Msh2 alleles on intestinal tumorigenesis. In this study, we introduced either an Msh2^null knockout allele that causes complete loss of MSH2, or the Msh2^G674D allele representing a Lynch Syndrome missense mutation into VCMsh2^LoxP mice. Interestingly, VCMsh2^LoxP/G674D mice developed a higher number of intestinal tumors. However, the tumors developed at a later age and their size was reduced compared to VCMsh2^LoxP/null mice. It is possible that the accelerated tumorigenesis and larger tumor size in VCMsh2^LoxP/null mice led to intestinal obstruction and early death, preventing higher tumor multiplicity later in life. These data indicate that Msh2 missense mutations can have distinct effects on intestinal tumorigenesis.

The Msh2^G674D mutation is located within the MSH2 ATPase domain at the identical amino acid residue as the Msh2^G674A mutation previously studied ²⁹. Neither mutation affect the DNA damage response function, and in vitro analysis using Msh2^G674D/G674D MEFs showed a normal apoptotic response to FOLFOX exposure as did MEFs from wildtype mice. Msh2^-/- MEFs displayed increased resistance (data not shown). Our in vivo analysis further showed that the Msh2^G674D mutation caused complete MMR deficiency in the intestinal mucosa of VCMsh2^LoxP/G674D mice but left the MMR-dependent DNA damage response function partially intact. These results support the idea that the DNA damage response function of MMR is important for the suppression of the early steps of intestinal tumorigenesis ³⁰.

Retention of the DNA damage response function in VCMsh2^G674D mice also had a significant impact on tumor response to chemotherapy. MRI based in vivo measurements of tumor growth showed that the intestinal tumors in VCMsh2^LoxP/G674D animals were sensitive to both cisplatin and FOLFOX, while tumors from VCMsh2^LoxP/null mice were generally resistant to both drugs. While cisplatin is commonly used to treat solid tumors ³¹ and sometimes GI tumors in combination treatment regimen (esophagus, stomach, anus, and sometimes small intestine), it is not routinely used to treat CRC. It is a platinum based drug that introduces GpG cross-links into DNA. These lesions are recognized by MutSα with high specificity; and MSH2-deficient cells display low level resistance to the drug ^{32, 33}. In vitro, MutSα binds to cross links, recruits MutLα, and interacts with the helicase domain of FANC-J (Fanconi Anemia protein) for cross link repair. The response of VCMsh2^LoxP/null and VCMsh2^LoxP/G674D tumors to cisplatin is consistent with the notion that complete loss of MSH2 diminishes cross link repair, and that less severe mutations might still be able to recruit repair factors for the removal of such lesions.

FOLFOX (5-fluoroacil (5-FU), leucovorin, and oxaliplatin) is commonly used for the treatment of late stage colorectal cancers. 5-FU-based adjuvant chemotherapy benefits patients with MSS tumors, but not patients with MSI positive tumors ³⁴. Our finding that VCMsh2^LoxP/null tumors are resistant to FOLFOX is consistent with these studies. There is significant in vitro evidence that MMR defective colon cancer cell lines are resistant to killing by 5-FU, and that re-expression of MMR genes in these cell lines reverses the phenotype ^{35, 36}. 5-FU acts to disrupt RNA synthesis and inactivate thymidylate synthase. It is also incorporated into DNA, and therefore could be processed by MMR. 5-FU substrates are recognized by MutSα, suggesting CRC chemosensitivity may in part be due to MMR protein recognition, leading to cell death either by futile repair cycles or signaling of cell cycle arrest and apoptosis ³⁷. VCMsh2^LoxP/G674D tumor sensitivity to FOLFOX implies that MutSα complexes interact with 5-FU adducts and mediate a cytotoxic response. It also raises the possibility that some tumors in human patients carrying MSH2 missense mutations that cause MSI without affecting DNA damage response will remain responsive to fluorouracil-based adjuvant chemotherapy.