The MutSβ complex is a modulator of p53-driven tumorigenesis through its functions in both DNA double strand break repair and mismatch repair

Abstract

Loss of the DNA mismatch repair protein MSH3 leads to the development of a variety of tumors in mice without significantly affecting survival rates, suggesting a modulating role for the MutSβ (MSH2-MSH3) complex in late onset tumorigenesis. To better study the role of MSH3 in tumor progression, we crossed Msh3^−/− mice onto a tumor predisposing p53-deficient background. Survival of Msh3/p53 mice was not reduced compared to single p53 mutant mice; however, the tumor spectrum changed significantly from lymphoma to sarcoma, indicating MSH3 as a potent modulator of p53-driven tumorigenesis. Interestingly, Msh3^−/− mouse embryonic fibroblasts displayed increased chromatid breaks and persistence of γH2AX foci following ionizing radiation, indicating a defect in DNA double strand break repair. Msh3/p53 tumors showed increased loss of heterozygosity, elevated genome-wide copy number variation, and a moderate microsatellite instability phenotype compared to Msh2/p53 tumors, revealing that MSH2-MSH3 suppresses tumorigenesis by maintaining chromosomal stability. Our results show that the MSH2-MSH3 complex is important for the suppression of late onset tumors due to its role in DNA double strand break repair as well as in DNA mismatch repair. Furthermore, they demonstrate that MSH2-MSH3 suppresses chromosomal instability and modulates the tumor spectrum in p53-deficient tumorigenesis, and possibly plays a role in other chromosomally unstable tumors as well.

Introduction

DNA mismatch repair (MMR) complexes function primarily in the detection and repair of mismatched bases that result from erroneous replication. Two heterodimeric MutS homolog (MSH) complexes, consisting of either MSH2-MSH6 (MutSα) or MSH2-MSH3 (MutSβ), are responsible for the recognition of these mismatched bases. MSH2-MSH6 binds to single base-base mismatches and small insertions/deletions (IDLs), whereas MSH2-MSH3 has a higher affinity for ≥2 base IDLs (1, 2). Germline mutations in MSH2 and MSH6 but not MSH3 are responsible for hereditary non-polyposis colorectal cancer/Lynch syndrome (HNPCC/LS); however, an (A)8 tract in the coding region of MSH3 was found to be frequently affected by frameshift mutations in MMR-deficient colorectal tumors, resulting in loss of MSH3 protein expression (3). In addition to its role in MMR, yeast studies revealed that during non-conservative homologous recombination, the MSH2-MSH3 complex is required to remove nonhomologous DNA ends during both the initiation of gene conversion and the resolution of single-strand annealing (SSA) intermediates that are initiated by a double strand break (DSB)(4, 5). SSA is a subset of the homologous recombination pathway for repairing spontaneous and induced DSBs that arise between repeated sequences occurring in cis. MSH2 and MSH3 recognize and stabilize the nonhomologous 3′ tails at the junction of double-stranded and single-stranded DNA to aid in either the recruitment or cleavage activity of Rad1-Rad10 (6). Although the mechanistic contribution of MSH3 in MMR and DSBR was thought to be overlapping, recent studies have shown that these two MSH3 activities are molecularly distinct (7). MSH2-MSH3 is also involved in triplet repeat (CAG·CTG) expansion diseases such as Huntington and myotonic dystrophy, and it has been shown in several mouse models that MSH2 and MSH3 are absolutely required to generate these expansions (8–12).

Previous observations showed that inactivation of Msh3 in mice leads to a moderate defect in the repair of insertion/deletion but not base-base mismatches (13), which might explain the absence of MSH3 mutations in HNPCC/LS families. Among other tumors, Msh3^−/− mice developed gastrointestinal tumors late in life, which corresponds with loss of MSH3 expression in late onset sporadic colorectal cancer (CRC) in humans (14). In human cancers, loss of MSH3 is associated with an MSI-low (MSI-L) phenotype at dinucleotide repeats and elevated microsatellite alterations at selected tetranucleotide repeats (EMAST) (15, 16). Loss or silencing of MSH3 also frequently occurs in a variety of other cancers, including non-small cell lung (17), ovarian, bladder (18), and breast (19) cancer. MSH3 polymorphisms were found to be associated with sporadic colorectal (14, 20, 21), prostate (22), and lung (23) cancer. These findings correlate with a recent report that unlike MutSα, the MutSβ complex is abundant in the majority of mouse tissues (24). Taken together, these studies indicate that MSH2-MSH3 might be important for tumor suppression in multiple tissues and especially in cases of late onset tumorigenesis.

Previous studies in mice showed that loss of Msh2 is sufficient to initiate and significantly accelerate tumorigenesis, especially in p53-deficient mice, and that tumorigenesis is associated with an MSI-high (MSI-H) phenotype (25). The MSH2-MSH6 complex is essential for the repair of base/base mismatches, and loss of either MSH2 or MSH6 therefore results in generation of a severe mutator phenotype and early onset cancers. In contrast, the absence of MSH3 mutations in early onset HNPCC/LS tumors and the late onset tumor phenotype in Msh3^−/− mice indicate that MSH3 is likely not involved in tumor initiation. To determine a possible role for MSH2-MSH3 in tumor progression, we crossed Msh3 null mice onto the tumor predisposing p53 null background to create cohorts of both Msh3^−/−p53^−/− and Msh3^−/−p53^+/− mice, compared their cancer predisposition phenotype to MMR-deficient Msh2/p53 mutant mice and analyzed the underlying molecular mechanisms. Although the effects of both Msh2- and Msh3-deficiency manifest from the earliest stages of life, tumor onset in Msh3^−/−p53^−/− mice occurred later in life in contrast to the early onset tumorigenesis in Msh2^−/−p53^−/− mice. We found that due to its role in DSBR rather than MMR, loss of Msh3 alters the tumor spectrum in p53 mutant mice by modulating the characteristics of the chromosomal instability (CIN) phenotype, resulting in elevated sarcomagenesis. We conclude that germline mutations in Msh3 can cause late onset tumorigenesis with various genomic signatures and modulate the tumor spectrum on cancer predisposing backgrounds.

Results

Altered cancer phenotype and elevated sarcomagenesis in Msh3/p53 mutant mice

To accelerate tumorigenesis in Msh3 null mice and study tumor progression, we intercrossed Msh3^+/− and p53^+/− knockout mice and generated two cohorts of mutant mice carrying homozygous mutations in Msh3 and either homozygous (Msh3^−/−p53^−/−) or heterozygous (Msh3^−/−p53^+/−) mutations in p53. To compare loss of Msh3 to the classic MMR phenotype, additional Msh2^−/−p53^−/− and Msh2^−/−p53^+/− cohorts were generated by intercrossing Msh2^+/− and p53^+/− mice.

Msh3^−/−p53^−/− double mutant mice showed reduced survival (median survival time 5 months, Figure 1a)when compared with Msh3^−/− alone (22 months)(13), but not when compared with p53^−/− single mutant mice (p = .31, Figure 1a). Msh3^−/−p53^+/− and p53^+/− mice also showed similar survival rates (p = .11), suggesting that loss of Msh3 does not contribute to the initiation or acceleration of tumorigenesis. In contrast, Msh2^−/−p53^−/− and Msh2^−/−p53^+/− mice showed significantly reduced survival compared to p53^−/− (2 vs. 6 months, p < .001) and p53^+/− (5 vs. 16 months, p< .001) mice respectively, confirming the strong tumorigenic MMR-deficient phenotype that has previously been described for Msh2^−/− (26–28), Msh2^−/−p53^−/− (25), and Msh2^−/−p53^+/− (29) mice.

Figure 1.

Loss of Msh3 modifies tumor phenotype of p53 null mice. (a) Tumor-free survival was significantly different for Msh2^−/−p53^−/− and p53^−/− mice (p< .001) and for Msh2^−/−p53^+/− and p53^+/− mice (p< .001). In contrast, survival was similar forMsh3^−/−p53^−/− and p53^−/− mice (p = .31) and for Msh3^−/−p53^+/− and p53^+/− mice (p = .11). Dashed lines, Msh2/p53 genotypes; solid lines, Msh3/p53 genotypes; dotted lines, single p53 genotypes. (b) Tumor type and incidence in mice with different Msh2, Msh3, and p53 genotypes. The percentage of sarcomas is increased in Msh3^−/−p53^−/− and Msh3^−/−p53^+/− mice compared to p53^−/− and p53^+/− mice, respectively (p = .04). Also, the carcinoma phenotype includes a small number of small intestinal tumors (Msh3^−/−p53^−/−, 1; Msh3^−/−p53^+/−, 2). (c) Microsatellite instability (MSI) analysis of tumors at the dinucleotide loci D7Mit91 and D17Mit123, determined by the number of unstable alleles divided by the total number of alleles scored. (d) Loss of heterozygosity (LOH) of the wild type p53 allele (^*) was demonstrated in Msh2^−/−p53^+/− or Msh3^−/−p53^+/− tumor samples using PCR (see also Table 1). Upper band, wild type allele (540 bp); lower band, p53 null allele (480 bp).

The tumor spectrum of p53^−/− mice is dominated by lymphomas, but also includes sarcomas and carcinomas (30). Surprisingly, loss of Msh3 induced a significant shift in the distribution of the various tumor types in p53 mutant mice: twice as many sarcomas were found in Msh3^−/−p53^−/− (32% vs. 17%, p = .04) and Msh3^−/−p53^+/− mice (53% vs. 30%, p = .04, Figure 1b) compared to tumors from p53^−/− and p53^+/− mice, respectively. In contrast, nearly 100% of MMR-deficient Msh2^−/−p53^−/− and Msh2^−/−p53^+/− mice succumbed to thymic lymphoma. Only a small number of Msh3/p53 mice developed gastrointestinal tumors. When tumors were analyzed for their MSI phenotype, which is the hallmark of MMR-deficiency, we found that most tumors from Msh2^−/−p53^−/− or Msh2^−/−p53^+/− mice showed MSI at the D7Mit91 and D17Mit123 dinucleotide repeats, compared to few tumors with MSI from Msh3^−/−p53 mutant mice(Figure 1c). Interestingly, none of the sarcoma samples included in the analysis (0/12) showed MSI. The MSI-L tumor phenotype indicates that defective MMR is not a main mechanism in Msh3^−/−p53-driven tumorigenesis.

LOH but not somatic mutation leads to loss of the wild type p53 allele in Msh3^−/−p53^+/− tumors

During tumorigenesis, MMR deficiency leads to the accumulation of somatic mutations in tumor suppressor genes and oncogenes. On the other hand, tumors with CIN usually display loss of heterozygosity (LOH) of tumor suppressor genes, which is often associated with defective DSBR pathways(31). Since loss of the tumor suppressor p53 is a critical event in tumorigenesis, we studied the genetic mechanisms by which the remaining p53 wild type allele was lost in Msh3^−/−p53^+/− and Msh2^−/−p53^+/− tumors. Consistent with the MMR-deficient phenotype, the majority of Msh2^−/−p53^+/− tumors (63%, 12/19, Table 1) showed mutation of p53. All p53 exons were sequenced, and mutations were found in exons 4, 5, 7, 8, 9, and 11 of Msh2^−/−p53^+/− tumors. In contrast, no p53 mutations were detected in the Msh3^−/−p53^+/− tumors (0/8); however, LOH at the wild type p53 allele occurred in almost all of the tumors (86%, 6/7) (Figure 1d and Table 1), suggesting a possible role for MSH3 in CIN. LOH of the wild type p53 allele was also found in some Msh2^−/−p53^+/− tumors (3/19, 16%), possibly reflecting loss of MSH2-MSH3 function in DSBR.

Table 1.

p53 mutation and LOH in p53^+/− tumors

Tumor sample	exon 4	exon 5	exon 7	exon 8	exon 9	exon 11.1	LOH
Msh2−/−p53+/−
29						del (A)7
47						del (A)7
90
66I	ins (T)6
73	Q112STOP
65							null
49
48					del (C)5
66II	NDa			del (A)5	ND	ND
63	ND			del (C)5	ND	ND
53	ND		del (G)6		ND	ND
88	ND	Q141STOP			ND	ND	wt
86	ND			del (A)5	ND	ND
20
36							wt
41	ND			del (C)5	ND	ND
40	ND		del (G)6		ND	ND
89							wt
14
Msh3−/−p53+/−
43	ND				ND	ND	wt
42	ND				ND	ND	wt
45	ND				ND	ND	wt
56	ND				ND	ND	ND
66	ND				ND	ND	wt
54	ND	ND			ND	ND	ND
36
64	ND	ND			ND	ND	ND
15	ND	ND	ND	ND	ND	ND	wt
19	ND	ND	ND	ND	ND	ND	wt

^aND, not determined

Msh3-deficient MEFs are defective in DSBR

In yeast, MSH2-MSH3 is involved in the processing of SSA recombination intermediates during certain forms of homologous recombination (5), and loss of this function might result in defective DSBR and contribute to the LOH phenotype in the tumors of Msh2^−/−p53^+/− and Msh3^−/−p53^+/− mice. To investigate if mouse MSH3 functions as part of the MSH2-MSH3 complex in the repair of DNA breaks, we first analyzed metaphases of untreated primary mouse embryonic fibroblast (MEF) cell lines. Msh3^−/− and Msh2^−/− MEFs both showed a significant increase in chromatid breaks (Figures 2a and 2b), further indicating a defective DSBR response. Since the number of breaks in Msh6^−/− cells was not significantly different from that in wild type cells (p = .21), loss of DSBR appears to be specifically due to the loss of MSH2-MSH3 complex function. Notably, all genotypes showed similar numbers of translocations.

Figure 2.

Loss of Msh3 induces a double strand break repair defect. (a) Metaphase spreads from Msh2^−/− (left) and Msh3^−/− (right) mouse embryonic fibroblasts (MEFs). Examples of chromatid breaks are magnified in insets. (b) Chromosomal aberrations were analyzed for >100 metaphase nuclei per genotype in MEF cell lines. Average numbers of chromatid breaks and translocations are shown. (c) Three categories of γH2AX foci accumulation were defined after staining with DAPI (blue) and anti-γH2AX antibody (red): 0–5, 6–10, and >10 foci per nucleus. (d) Deficient double strand break repair was defined by accumulation of γH2AX foci in Msh3^−/− and wild type MEFs 6h after irradiation with 1Gy.

To further analyze the observed DSBR defect, Msh3^−/− cells were subjected to a low dose of ionizing radiation (1Gy) and efficient resolution of DSBs was subsequently evaluated by counting γH2AX foci in the cell nuclei (Figure 2c). We did not observe a difference in the number of γH2AX foci one hour after irradiation (Figure S1), which suggests that both wild type and Msh3^−/− MEFs had a comparable accumulation of DSBs at early time-points and that γH2AX signaling is intact in Msh3^−/− cells. However, two hours after irradiation the majority of wild type cells (95%) showed background levels of 0–10 γH2AX foci per nucleus in contrast to only 45% of Msh3^−/− cells (Figure S1). Even 6 hours post-irradiation only 70% of Msh3^−/− cells were able to adequately resolve DSBs (p< .001, Figure 2d), and an increase of more than 10 γH2AX foci per nucleus could be seen in the remaining 30% of nuclei. Taken together, these data demonstrate a moderate DSBR defect in Msh3-deficient cells that is revealed by a delay in resolving DSBs.

Msh3 deletion is associated with chromosomal instability in p53-deficient tumors

To examine whether loss of MSH3 does lead to chromosomal instability during tumorigenesis, we used spectral karyotyping (SKY) to determine whether lymphomas of Msh3-deficient mice displayed any gross chromosomal abnormalities. In contrast to Msh2^−/−p53^−/− tumor cells, which showed close to normal ploidy (40 chromosomes/cell, Table 2), Msh3^−/−p53^−/− tumor cell karyotypes showed increased chromosomal instability characterized by significantly increased aneuploidy (55 chromosomes/cell, p< .001), translocations, deletions, duplications, and breaks (Figure 3). No specific recurring chromosomal aberrations were found; indicating that loss of MSH3 induces general chromosomal instability. However, the type of chromosomal changes between genotypes tended to be different: Msh3^−/−p53^−/− and Msh2^−/−p53^−/− tumor cells showed about two times more translocations compared to p53^−/− cells (1.2 and 0.9 vs. 0.5 per cell, Table 2), suggesting that the DSBR defect associated with loss of the MSH2-MSH3 complex that was observed earlier contributes to this type of chromosomal rearrangement. In contrast, a trend towards slightly higher numbers of deletions and duplications was observed in Msh3^−/−p53^−/− and p53^−/− cells compared to Msh2^−/−p53^−/− cells, reflecting the aneuploidy phenotype that is associated with the p53^−/− tumor background.

Table 2.

Chromosomal aberrations from SKY analysis

Tumor sample	# Cells	Average # aberrations per cell (± CI 95%)
		Chromosomes	Translocations	Deletions/Duplications
Msh3−/− p53−/−	23	55a (51–83)	1.2 (0.4–1.9)	1.0 (0.4–1.5)
Msh2−/− p53−/−	26	40 (39–41)	0.9 (0.5–1.3)	0.7 (0.2–1.2)
p53−/−	17	51 (51–54)	0.5 (0.1–0.8)	1.2 (0.4–1.9)

^aboth Msh3^−/−p53^−/− (p< .001) and p53^−/− (p = .001) tumors have significantly more chromosomes per cell compared with Msh2^−/−p53^−/− tumor cells

Figure 3.

MSH3 is involved in maintaining chromosomal stability. Examples of spectral karyotypes from lymphoma cells are shown for each genotype, indicating a close to normal karyotype and t(1,8) for Msh2^−/−p53^−/−, increased aneuploidy, del1, and t(1,7) for Msh3^−/−p53^−/−, and increased aneuploidy, 2del1 and del6 for p53^−/−. Red boxes indicate translocations and deletions.

To study the increase in CIN in Msh3/p53 tumors in more detail, we used the lymphoma samples that were used for SKY analysis in array comparative genomic hybridization (aCGH) to analyze genome-wide copy number variation (CNV). Since loss of Msh3 induces sarcomagenesis we also analyzed two groups of Msh3^−/−p53^+/− and p53^−/− sarcomas (Figure S2). Again, no increase in CNV was found to recur between samples at specific genomic regions, indicating general chromosomal instability. Compared with Msh2^−/−p53^−/− lymphoma samples, in which CNV was uncommon, this type of instability was significantly increased in Msh3^−/−p53^−/− and p53^−/− lymphomas (Figure 4). Log2 ratio distributions of gains and losses (Figure S3) and subsequent cluster analysis (Figure S4) revealed that both p53^−/− and Msh3^−/−p53^−/− lymphomas showed a significant increase in CNV (cluster 1) compared to Msh2^−/−p53^−/− tumors (cluster 3). Msh3^−/−p53^−/− and p53^−/− lymphoma samples clustered together (cluster 1) which indicates that the overall distribution of CNVs throughout the genome in Msh3^−/−p53^−/− tumors was similar to that of p53^−/− tumors and that loss of MSH3 does not induce a specific pattern of CNV distribution. Interestingly, the Msh3/p53 and p53 sarcoma samples (cluster 2) only showed a moderate increase in CNV compared to Msh3/p53 and p53 lymphoma samples. However, the CNV phenotype in Msh3/p53 sarcomas was significantly more severe than that found in Msh2/p53 lymphoma samples. Taken together, these results confirm that loss of MSH3, but not MSH2, is associated with an increase in overall chromosomal instability in tumor genomes and that the amount of CIN depends on the underlying predisposing genetic background and the tissue-specific context.

Figure 4.

Loss of Msh3 is associated with a high level of CNV. Chromosome copy number changes were analyzed as described before (48). To account for tumor heterogeneity, large segments with low-level copy number changes were considered as important as small segments with high-level changes, and no arbitrary log2 ratio was used. %, percentage of samples that showed gain (red) or loss (green). L, lymphoma; S, sarcoma.

Discussion

The MMR proteins MSH2, MSH6, MLH1 and PMS2 all play a major and well-described role in mismatch repair, and mutations in these genes have been found in patients with HNPCC/LS. Up until now, no mutations have been found in the other MMR proteins MSH3, MLH3 and EXO1 that could directly link them to the development of familial colorectal cancer. Although intestinal tumors developed in Msh3(13) and Mlh3(32) knockout mice late in life, these proteins appear to have a less pronounced role in MMR and tumor suppression. However, they have been implicated in other processes aside from MMR: apart from their roles in class switch recombination and somatic hypermutation (33–36), MLH3 is essential for processing DSBs at crossovers during meiosis (37), and EXO1 has been identified as a key mediator of DNA end resection during DSBR (38). Surprisingly, it has recently been found that MSH3 protein expression is higher in the majority of murine tissues when compared to MSH6, suggesting important and specific roles for MSH2-MSH3 in DNA repair and genomic instability (24). The similar phenotype of late onset tumors with mild MSI that appeared in Msh3^−/− and Exo1^−/− mice led us to hypothesize a role for MSH3 in tumor suppression via DSBR and maintenance of chromosomal stability. In this study, we showed that in mice loss of Msh3 leads to accumulation of unrepaired DSBs, and to a shift in tumor spectrum with increased CIN on a p53-driven tumor predisposing background.

As described before (25), mice with combined Msh2 and p53 ablation show independent segregation of the MSI phenotype compared to p53^−/− alone, which suggests that loss of Msh2 is dominant over the p53^−/− phenotype. Furthermore, their synergism in tumorigenesis suggests that they are not genetically epistatic. In contrast, Msh3^−/−p53^−/− mice show features that are more similar to the p53^−/− phenotype. The lack of synergism in tumorigenesis and the shift in tumor spectrum towards sarcomagenesis suggest that MSH3 might not be involved in tumor initiation but rather in tumor progression by modulating the p53-deficient phenotype. Interestingly, loss of p53 did not accelerate gastrointestinal tumorigenesis of Msh3^−/− mice since we only detected a small number of gastrointestinal carcinomas in Msh3/p53 mutant mice (Figure 1b). Since loss of MSH3 is associated with a variety of human cancer types, its effects might be evident on other predisposing backgrounds, implicating MSH3 as a general modulator of cancer phenotypes. The contribution of Msh3 deletion to tumorigenesis, however, depends on the genetic context and subsequent mechanism of tumorigenesis in different tissues, since Msh3^−/−Apc^1638N mice did not show a different tumor onset or phenotype compared to Apc^1638N mice (39).

Due to the strong aneuploidy phenotype caused by loss of p53(40, 41) it was not immediately clear whether loss of Msh3 contributed to the increase in CIN in Msh3^−/−p53^−/− tumors. However, although there were not enough tumor samples available for SKY analysis, we observed differences between tumors with combined Msh3/p53 deficiency and single p53 deficiency with a trend towards an increase in the average number of translocations per cell for Msh3^−/−p53^−/− tumors compared to p53^−/− alone, indicating defective DSBR caused by loss of the MSH2-MSH3 complex in these tumors. This defect in DSBR was also visible when we counted chromosomal aberrations in metaphase spreads, showing a significant increase in chromatid breaks in Msh3^−/− and Msh2^−/− MEFs. Interestingly, here the number of translocations was not significantly different between wild type, Msh3^−/− and Msh2^−/− MEFs, which might be due to technical difficulties with scoring translocations in regular metaphase spreads. Alternatively, this data suggests that loss of MSH2-MSH3-dependent DSBR results in an increase in the number of translocations during the clonal outgrowth of p53-deficient tumors and contributes to p53-driven tumorigenesis. The amount of chromosomal instability represented by CNV, however, is significantly increased in Msh3^−/−p53^+/− tumors (Figure 4 and Figure S2) compared to the CNV phenotype in p53^+/− tumors from our previous study (48) which indicates that loss of Msh3 is contributing to the chromosomal instability phenotype in Msh3^−/−p53^+/− tumors.

Our observations indicate a dual function for the MutSβ complex in genome maintenance and tumor suppression (Figure 5). Msh2-deficient mice display a strong MMR defect with accumulation of mutations and low chromosomal instability, indicating the essential role for MSH2 in MMR. On the other hand, a moderate DSBR defect is associated with the loss of either MSH2 or MSH3, and whereas the mild phenotypes with regard to CIN and tumorigenesis that are associated with the loss of MSH2-MSH3 dependent DSBR are difficult to observe in the dominant MMR tumor phenotype of Msh2-deficient mice, we were able to study them by deleting Msh3 on the appropriate tumor predisposing background. Because the DSBR defect is not as severe as the defect caused by loss of other canonical DSBR proteins, this further suggests a role for MSH3 in the repair of a subset of DSBs, possibly substrates of the SSA pathway (6, 7). Interestingly, loss of MSH3 was shown to be associated with accelerated tumor progression in MLH1-deficient colorectal cancers, and it was suggested that the effect of MSH3 loss on tumor progression might be related to another function of MSH3 unrelated to MMR, implicating its role in DNA DSBR by SSA (3). It remains unclear whether the role of MSH3 in DSBR is dependent on its recruitment via the MMR pathway, either during the process of homeologous recombination or when small IDLs or DNA mismatches (42) on opposite strands are in close proximity so as to mediate the formation of DSBs during the MMR process, or whether MSH2-MSH3 could instead be recruited directly to sites of DSBs. Regardless, MSH3 was shown to co-localize with γH2AX and DSB foci following genotoxic stress (43, 44), which suggests that it might be directly involved in a subset of DSBR events.

Figure 5.

Model representing the different functions of the Mutsβ (MSH2-MSH3) complex. Deletion of Msh2 induces a strong MMR defect and a moderate DSBR defect, leading to a dominant tumor phenotype with early onset tumors that are MSI-high and chromosomally stable. Deletion of Msh3 revealed a moderate DSBR defect next to the previously described moderate MMR defect, resulting in a modulated tumor phenotype of late onset tumors that were MSI-low and showed increased chromosomal instability.

Despite the moderate defect in DSBR Msh3 deficient mice showed a significant increase in sarcomagenesis, suggesting that loss of MSH3 targets sarcoma tumor suppressor genes and modulates p53-dependent tumorigenesis. When we analyzed lymphomas and sarcomas for genomic instability, Msh3^−/−p53^−/− and p53^−/− lymphomas showed the largest increase in CNV while Msh2^−/−p53^−/− mice displayed a low CNV phenotype. Interestingly, both Msh3/p53 and p53 sarcomas displayed significantly more CNV than Msh2/p53 lymphomas, but less CNV than Msh3/p53 and p53 lymphomas, and clustered together based on their moderate CNV profile. This indicates that loss of Msh3 causes a repair defect that contributes to genomic instability in a tissue-specific manner and promotes p53-driven sarcomagenesis. Since the sarcomas in our mouse cohorts are microsatellite stable, the increase in sarcoma incidence in Msh3/p53 mutant mice is likely caused by loss of Msh3-dependent DSBR which contributes to the moderate CIN that is associated with sarcomagenesis. However, loss of Msh3-dependent MMR might also play a role in p53-dependent sarcomagenesis by resulting in a low-level mutator phenotype that might be difficult to detect (2, 42).

In conclusion, our data indicate for the first time that the MSH2-MSH3 protein complex is involved in tumorigenesis through maintenance of chromosomal stability. In contrast to the MutSα complex (MSH2-MSH6), which has a dominant role in genome maintenance by MMR, MSH2-MSH3 functions both in MMR, indicated by an MSI-L phenotype, and DSBR, as has been recently shown in yeast (7). Loss of MSH3 can therefore contribute to tumorigenesis in two ways: by a mild MMR defect leading to MSI-L and low-level mutation accumulation, and by a DSBR defect that leads to a moderate increase in CIN. Since most sporadic colorectal cancers (CRCs) are CIN (85%) (45), the MSH2-MSH3 complex might play an important role in suppression of late-onset sporadic cancers that are MSI-L. Future studies should therefore include the analysis of Msh3 loss on multiple tumorigenic backgrounds to assess a more specific role for MSH3 in tumorigenesis.

Materials and methods

Animals, tumors and survival

Msh2^−/−, Msh3^−/− and Msh6^−/− knockout mice were previously generated in our lab (13, 28, 46). p53^−/− knockout mice (47) were purchased from Jackson Laboratories. All mice were on a congenic C57BL/6 background. Mice were intercrossed as described and observed until they became morbid or moribund. Tumors were removed and fixed in 10% buffered formalin or frozen at −150°C. After paraffin embedding of fixed tissue, sections were stained with hematoxylin and eosin. Statistical analysis of tumor incidence was completed with IBM SPSS Statistics version 20.0 (IBM Corp., Armonk, NY). The Kaplan-Meier method was used to compare curves for survival, with significance evaluated by two-sided log rank statistics. A Chi squared test was used to assess the influence of genotype on the tumor spectrum.

Microsatellite instability, LOH, and mutational analysis of p53

For MSI analysis, the D7Mit91 and D17Mit123 microsatellite loci were amplified from normal and tumor DNA and analyzed as previously described (39). Tumor DNA was also serially diluted and analyzed for loss of the wild type p53 allele as previously described (48). To analyze tumors for p53 mutations, primers amplifying all exons of p53 (including both transcript variants of exon 11) were used as described (49), followed by DNA sequencing.

Chromatid break and double strand break analysis

Primary mouse embryonic fibroblasts (MEFs, p4–6) were treated overnight with 0.1 μg/ml colcemid (Invitrogen, Carlsbad, CA) and subsequently swelled, fixed, and dropped onto glass slides to prepare metaphase spreads. DNA was stained using mounting medium with DAPI (Vector Laboratories, Inc., Burlingame, CA), and at least 100 metaphases in 2–3 different cell lines were analyzed per genotype.

For DSB analysis, MEFs were cultured on poly-L-lysine coated slides (Sigma, St. Louis, MO), irradiated with 1 Gy, and incubated at 37°C for 30 min – 6h. Cells were fixed with cold 1:1 methanol:acetone for 5 min, and washed twice with cold PBS for 5 min. After blocking with 10% FBS in PBS, cells were incubated overnight with a mouse monoclonal γH2AX antibody (Upstate, Billerica, MA; 1:500). Cells were again washed twice in cold PBS for 5 min, and incubated for 1h at room temperature with an Alexa Fluor 488-labeled goat anti-mouse antibody (Invitrogen). DNA was stained using mounting medium with DAPI. The experiment was repeated three times for two different cell lines per genotype, and a minimum of 100 cells was analyzed each time.

Metaphases and foci were counted using an Olympus BX61 microscope (Olympus America Inc., Center Valley, PA) with Sensicam QE cooled CCD camera (PCO, Kelheim, Germany). Significant differences between cell lines in the number of breaks or γH2AX foci were calculated using an independent two sample t test of equal variance.

Spectral karyotyping (SKY)

Thymic lymphomas were dissociated, filtered, and resuspended as described (50). To prepare metaphase spreads, cells were treated with colcemid for 3–6 hours, swelled, fixed, and dropped onto clean glass slides inside a humidity chamber. SKY was performed as previously described (51). Slides were hybridized with the combinatorially labeled whole chromosome painting probes (Applied Spectral Imaging, Inc., Carlsbad, CA) and metaphase images were captured using the Applied Spectral Imaging interferometer on an epifluorescence microscope (Zeiss). SKY karyotypes were then analyzed with SkyView^™ version 1.62 (Applied Spectral Imaging, Inc.). Metaphases were captured and analyzed using the nomenclature approved by the International Committee on Standardized Genetic Nomenclature for Mice (http://www.informatics.jax.org).

Array comparative genomic hybridization (aCGH)

Genomic DNA was isolated from frozen tumor tissue and matched normal tail tissue using the Qiagen DNeasy kit, and hybridized to NimbleGen Mouse CGH 3×720K Whole-Genome Tiling Arrays (Roche NimbleGen, Inc., Madison, WI) in four different experiments. To represent the wide variation of mouse sarcoma samples but also to homogenize between genotypes, groups were composed of one fibrohistiocytic, one hemangio- and one spindle cell sarcoma each.

Plotting summaries of copy number changes among all samples for each genotype was done as described before (48). In order to compare copy number changes between different sets of samples, distributions of the (log2) ratios for the segmented data were plotted for each sample. The tails of this distribution (weight) were used to indicate the amount of chromosomal instability: That is, for each distribution, the proportion of data points >3 s.d. of the distribution with the lowest variance was computed on the left and right tails, i.e. relative copy number increase (gain) or decrease (loss) according to the aCGH data. Samples were subsequently clustered using the k-means clustering algorithm. Consistency of the clusters obtained was evaluated by the silhouette method(52).