A Dominantly Acting Murine Allele of Mcm4 Causes Chromosomal Abnormalities and Promotes Tumorigenesis

Here we report the isolation of a murine model for heritable T cell lymphoblastic leukemia/lymphoma (T-ALL) called Spontaneous dominant leukemia (Sdl). Sdl heterozygous mice develop disease with a short latency and high penetrance, while mice homozygous for the mutation die early during embryonic development. Sdl mice exhibit an increase in the frequency of micronucleated reticulocytes, and T-ALLs from Sdl mice harbor small amplifications and deletions, including activating deletions at the Notch1 locus. Using exome sequencing it was determined that Sdl mice harbor a spontaneously acquired mutation in Mcm4 (Mcm4D573H). MCM4 is part of the heterohexameric complex of MCM2–7 that is important for licensing of DNA origins prior to S phase and also serves as the core of the replicative helicase that unwinds DNA at replication forks. Previous studies in murine models have discovered that genetic reductions of MCM complex levels promote tumor formation by causing genomic instability. However, Sdl mice possess normal levels of Mcms, and there is no evidence for loss-of-heterozygosity at the Mcm4 locus in Sdl leukemias. Studies in Saccharomyces cerevisiae indicate that the Sdl mutation produces a biologically inactive helicase. Together, these data support a model in which chromosomal abnormalities in Sdl mice result from the ability of MCM4D573H to incorporate into MCM complexes and render them inactive. Our studies indicate that dominantly acting alleles of MCMs can be compatible with viability but have dramatic oncogenic consequences by causing chromosomal abnormalities.


Introduction
Mouse models have been invaluable tools for studying human cancer. Many mouse models used for this purpose are reverse genetic, in that they involve genetically modified mice engineered to have lost a specific tumor suppressor gene (tsg) or to overexpress a specific proto-oncogene. More rarely, spontaneous or mutagen induced mouse models that result in tumor formation have been used to study tumorigenesis. Given the contribution of mouse models to understanding tumorigenesis, when a spontaneous mouse mutant that developed T-ALL arose in our colony, we pursued studies to both characterize the disease in these mice and to identify the causal mutation. The mutation was spontaneous and the phenotype dominant, so we named the mutant Spontaneous dominant leukemia (Sdl). We have identified a mutation in the Mcm4 gene as the likely causative genetic lesion in these mice.
MCM4 is part of the MCM2-7 heterohexameric complex that is involved in licensing origins of DNA replication prior to S phase. The MCM complex has ATPase activity and serves as the core of the replicative helicase that unwinds duplex DNA and drives progression of the replication fork [1]. Improper fork progression can lead to stalled forks, the potential for incomplete DNA replication and even fork collapse which may lead to double strand break (DSB) formation [2]. Therefore, the MCM proteins play important roles in maintaining genomic integrity, however their roles in tumorigenesis are just beginning to be elucidated.
Previous studies of murine Mcm genes have involved hypomorphic or gene-trap null alleles. Gene-trap alleles are heterozygous viable and homozygous lethal [3,4]. Mice harboring hypomorphic alleles of Mcm2 (Mcm2 IRES-CreERT2 ) [5] or Mcm4 (Mcm4 chaos3 ) [3] show reductions in MCM protein levels and develop tumors only in the homozygous state. Mcm4 chaos3 was discovered in a screen for mutations that cause increased micronucleus formation in reticulocytes and therefore promote chromosome instability. Mcm4 chaos3 results from a Phe345Ile substitution in MCM4, which is a residue that is involved in the interaction of MCM4 with MCM6 in the heterohexameric complex [3]. In Mcm4 chaos3/chaos3 mouse embryonic fibroblasts (MEFs), total and chromatin bound levels of MCM4 and other MCM proteins are reduced compared to wild-type [3,6]. This leads to a loss of backup origins that normally fire during replicative stress which is hypothesized to be the mechanism by which low levels of MCM proteins promote genomic instability [6,7]. Mcm4 chaos3/chaos3 mice develop tumors with long latency. Although the tumor spectrum varies with genetic background, Mcm4 chaos3/chaos3 mice have not been reported to develop T-ALL [3,6]. We have accumulated evidence that the early-onset T-ALL phenotype in Sdl mice results from a novel allele of Mcm4 (Mcm4 D573H ) that in the heterozygous state promotes chromosomal abnormalities that cause highly penetrant tumor formation.

Sdl causes primarily early onset T-ALL
The Sdl mutation arose in our colony in the germline of a breeder on the C57Bl/6 genetic background. We therefore pursued a recombination mapping strategy by utilizing out-crosses and backcrosses to the FVB/N and 129S1/SvImJ genetic backgrounds. A whole genome scan using simple sequence length polymorphisms (SSLPs) was performed and it was determined that mice of backcross generations that inherited C57Bl/6 markers at D16MIT131 and D16MIT4 on proximal Chr 16 rapidly became moribund ( Figure 1A) indicating linkage to this chromosomal location. Therefore, Sdl carriers can be identified by the presence of C57Bl/6 markers at these two SSLPs. Phenotypically, 94.2% (180 of 191) of moribund Sdl mice necropsied had signs of hematologic malignancy including mediastinal masses, splenomegaly and/or lymphadenopathy. Histologically, neoplastic cells filled hematopoietic tissues ( Figure 1B) and infiltration of neoplastic lymphocytes into non-hematopoietic organs was frequently observed ( Figure 1C). Leukemic cells are also found in the blood ( Figure 1C) and bone marrow ( Figure S1A). Sdlinduced disease was transplantable as tumors ( Figure S1B) developed with an average latency of 29 days in four of four immunocompromised nude/nude mice that received cells isolated from mediastinal masses from moribund Sdl mice.
Flow cytometry was used to determine the phenotype of hematologic tumors from four Sdl mice. Three mice developed disease early in life that was phenotypically T-ALL ( Figure 1D-1F and Table S1). The fourth mouse developed leukemia/lymphoma late in life, which expressed few lineage markers ( Figure 1G and Table S1). Southern analysis of early-onset leukemias/lymphomas from Sdl mice detects rearrangements of the T cell receptor (TCR) b locus [8] ( Figure S1). The majority of Sdl leukemias/lymphomas express TdT ( Figure S1) that, together with the surface phenotypes, indicates that most Sdl mice develop T-ALL with an immature phenotype.
Inter-crosses of Sdl heterozygotes were performed to determine the phenotype of Sdl homozygotes. No Sdl/Sdl mice were present at weaning, so embryos from timed pregnancies of Sdl inter-crosses were examined. No Sdl/Sdl embryos were detected even as early as 8.5 dpc (n = 69; 20 wild-type, 49 Sdl/+, 0 Sdl/Sdl; p,0.0001 Chi square test), indicating that Sdl is homozygous lethal early during embryonic development. Therefore, all carrier mice utilized for the experiments described here are Sdl/+.

Sdl causes subtle defects during thymic development
To further characterize the molecular basis of leukemogenesis in Sdl mice, microarray analysis was performed to detect mRNA expression differences between wild-type thymuses and thymuses from pre-leukemic Sdl carriers. To detect differentially expressed genes, the false discovery rate was controlled at 5%. Specifically, transcripts with a posterior probability of differential expression .95% and a q-value ,0.05 were considered to be significantly differentially expressed (see Methods). No transcripts were found to be significantly differentially expressed between wild-type and pre-leukemic Sdl carrier thymuses. Tests for common function did identify Gene Ontology (GO) sets enriched for differential expression between wild-type and carrier thymuses (Table 1), indicating some molecular differences between wild-type and carrier thymuses. However, a similar analysis of the Kyoto Encyclopedia of Genes and Genomes (KEGG) failed to detect any differences between wild-type and carriers.
To determine if Sdl impacts T cell development, flow cytometry was performed to characterize T cell developmental stages in Sdl carriers. Thymocytes were analyzed from Sdl carrier (n = 4) and wild-type siblings at 3.5 weeks of age (n = 4). Analysis of more mature thymocyte populations (CD4, CD8 and CD4/8 double positive) revealed a trend toward decreased levels of CD4+ cells in Sdl mice, however this did not reach statistical significance (p,0.064 Table 2). Lineage markers as well as CD44 and CD25 were then utilized to further analyze more immature double negative (DN) populations. There was a statistically significant decrease in the percentage of DN cells at the DN1 stage of development in Sdl mice (Table 2). Although no statistically significant differences in other DN cell populations were observed, flow cytometry profiles from individual mice revealed inter-animal differences, particularly in the DN3 population, in Sdl mice ( Figure  S2). Taken together, these data indicate that Sdl does cause subtle defects in thymocyte development; with some mice more severely affected than others. However, it is unlikely that Sdl causes T-ALL by directly promoting a block in thymocyte differentiation.

Sdl mice harbor a mutation in Mcm4 that causes chromosomal abnormalities
To identify the affected gene in Sdl mice, the chromosomal location of the Sdl mutation was further narrowed utilizing single nucleotide polymorphic (SNP) markers to analyze mice with recombination events in proximal Chr 16. Using this approach, the Sdl mutation was mapped to a 1.4 Mb candidate region ( Figure 2A) that contains 30.5 kb of annotated protein-coding sequence. No differences in expression levels of genes in the interval were detected by quantitative RT-PCR (qRT-PCR) analysis comparing 21-day-old Sdl carrier thymuses to 21-day-old control thymuses ( Figure S3). Exon capture followed by resequencing was performed on Sdl genomic DNA on the 129S1/ SvImJ congenic background. To ensure complete coverage in the Sdl interval, PCR amplification followed by Sanger sequencing was used to further examine exon and splice site sequences with fewer than 106 coverage following exon capture ( Figure S4 and Table S2). After eliminating known SNPs between C57Bl/6 and 129S1/SvImJ (Table S3), only one non-synonymous sequence difference in the Sdl interval was identified. This difference is a G to C missense mutation that causes a D573H substitution in Mcm4 (Mcm4 D573H ). This residue is conserved not only in MCM4 but also across all MCM2-7 subunits in eukaryotes ( Figure S5). This nucleotide change was present in all confirmed leukemic Sdl mice examined and was not detected in FVB/N, 129S1/SvImJ, or mice of the C57Bl/6 genetic background that were present in the colony at the time that Sdl arose ( Figure 2B). A cross of Sdl/+ to Mcm4 chaos3/+ did not produce any Sdl/Mcm4 chaos3 viable pups at p1 (n = 26; 7 wild-type, 10 Sdl/+, 9 Mcm4 chaos3/+ , and 0 Sdl/ Mcm4 chaos3 ; Chi square p value 0.0246), indicating non-complementation.
To further investigate if Sdl mice harbor phenotypes indicative of replicative stress, chromosomal aberrations were examined in both reticulocytes and mouse embryonic fibroblasts (MEFs) isolated from Sdl mice. Sdl mice harbor an ,18-fold increase in spontaneous micronucleated reticulocytes compared to noncarrier siblings ( Figure 2C and 2D). This is similar to the ,20fold increase reported for Mcm4 chaos3/chaos3 mice studied on a different strain background [3]. MEFs from Sdl carriers and noncarrier siblings were analyzed cytogenetically for chromosome breaks in the presence and absence of the DNA replication inhibitor aphidicolin (APH) ( Figure 2E-2F). More chromosome breaks were found in APH-treated Sdl MEFs compared to wildtype (p,0.02). Together, these observations indicate that Sdl causes chromosomal aberrations and increased sensitivity to exogenous replication stress, a phenotype that is consistent with MCM dysfunction [3,5]. Therefore, all evidence suggests that the Sdl phenotype is caused by Mcm4 D573H .

Mcm4 D573H acts in a dominant manner to promote tumorigenesis
Previously studied hypomorphic or gene trap null alleles of Mcms have indicated that minimum thresholds of MCM levels are needed for normal development and for tumor suppression in adults; and reductions in protein levels of other members of the MCM2-7 complex have been detected in Mcm2 and Mcm4 hypomorphic mice [3][4][5][6]. Therefore qRT-PCR and Western analyses were utilized to examine Mcm levels in 21-day-old wildtype thymuses and 21-day-old Sdl carrier thymuses. No reductions in mRNA levels for Mcm2-7 ( Figure 3A) or total or chromatin bound protein levels for MCM2 and MCM4 ( Figure 3B) were detected. MCM4 total and chromatin bound levels were also not reduced in Sdl MEFs compared to wild-type MEFs ( Figure S6). Therefore Mcm4 D573H does not promote tumorigenesis by simply causing a reduction in transcript or protein levels of Mcm4 or other Mcms.
Although genetically it acts dominantly, Mcm4 D573H could actually promote tumor formation in a recessive manner if lossof-heterozygosity (LOH) or epigenetic silencing at the Mcm4 locus occurs during tumor formation. To address these possibilities, RT-PCR followed by re-sequencing was used to examine if both wildtype and mutant Mcm4 alleles are expressed at the mRNA level in Sdl tumors. Peak heights of Sanger sequencing traces indicated that both alleles are expressed at similar levels in both Sdl tumors and in 21-day-old thymuses from Sdl carrier mice ( Figure 3C). As stromal cells are present in bulk tumors, Mcm4 allele expression was also examined in cell lines that were established from Sdl T-  Table S1. Mice depicted in D-F succumbed to early onset-disease that is phenotypically T-ALL. Within these animals, there is evidence of both CD4/8 double positive (DP) disease as well as CD8 single positive (SP) disease. The mouse in G became moribund with late-onset disease (354 days of age) and the tumor cells do not express most T cell antigens (see also ALLs. Both alleles were expressed at similar levels as they are in thymuses from pre-leukemic Sdl carrier mice ( Figure 3C). Therefore, tumorigenesis in Sdl mice does not require LOH, and the Mcm4 D573H allele acts dominantly to cause T-ALL.

The Sdl mutation generates a biologically non-functional helicase
To determine the impact of the Sdl mutation on MCM function, complementation studies in Saccharomyces cerevisiae were performed. These studies utilized a haploid strain that harbors a deletion of the chromosomal mcm4 locus in which viability is maintained by a URA3-mcm4 plasmid [9]. This strain was transformed with a TRP1 plasmid harboring mcm4 with the Sdl mutation engineered into the analogous yeast residue (mcm4 D632H , hereafter referred to as mcm4 Sdl ). Cloning into the TRP1 vector added a HA/10XHis tag, which has been shown to not compromise Mcm4 protein function in complementation tests [9] and allowed verification of Mcm4 Sdl protein expression by Western blotting (not shown). If the mcm4 Sdl allele expressed by the TRP1 plasmid complements the mcm4 genomic deficiency, then growth on -TRP+5-Fluoroorotic Acid (FOA) (restrictive conditions) will occur due to the ability to lose the wild-type mcm4 copy on the URA3 plasmid. TRP1-mcm4 wild-type and empty TRP1 vectors served as positive and negative controls, respectively. For each TRP1 vector, multiple individual colonies were analyzed for growth under restrictive conditions (example shown in Figure 4A). As expected, no empty TRP1 vector colonies (n = 37) grew while all TRP1-mcm4 colonies (n = 37) grew. Surprising, mcm4 Sdl showed an intermediate phenotype as 10 of 38 colonies grew. To further examine this phenomenon, mcm4 Sdl colonies were examined for the presence of mcm4 Sdl sequences ( Figure 4B). As expected, freshly isolated mcm4 Sdl colonies grown under permissive conditions (2URA 2TRP) harbored both wild-type mcm4 and mcm4 Sdl sequences due to the presence of both URA3-mcm4 and TRP1-mcm4 Sdl plasmids. However, mcm4 Sdl colonies that grew under restrictive conditions (2TRP +FOA) only harbored mcm4 wild-type sequences, indicating that a reversion or gene conversion involving mcm4 sequences on the TRP1-mcm4 Sdl plasmid had occurred. These results indicate that mcm4 Sdl generates a biologically non-functional helicase as it cannot complement a mcm4 genomic deletion. As Mcm4 D573H does not reduce MCM protein levels in mice, it is hypothesized that it causes chromosomal abnormalities and promotes tumorigenesis by stably incorporating into MCM heterohexamers and interfering with their normal function.

Sdl leukemias are characterized by intragenic deletions at the Notch1 locus
To further characterize the molecular basis of leukemogenesis in Sdl mice, microarray analysis was performed to compare expression in overt thymic tumors from Sdl mice to wild-type thymus. Utilizing the same criteria described above for analysis of pre-leukemic Sdl carriers, 3627 genes were found to be differentially expressed, of which 745 had $2 fold change in expression levels. The 20 significantly differentially expressed genes with the largest fold changes of increased and decreased expression in Sdl leukemias compared to wild-type thymuses are outlined in Table 3.
Microarray data indicated that the Notch1 pathway is activated in Sdl tumors as Notch1 itself and several Notch1 target genes including Myc, Hes1, Dtx1, Adam19, Hey, Heyl and Il2ra [10][11][12][13] were transcriptionally up-regulated in Sdl tumors compared to normal thymus (Table 3 and microarray data available at GEO). Expression levels of the Notch1 targets Hes1 and Myc were investigated by qRT-PCR ( Figure 5A), which revealed that they were expressed at equivalent levels in wild-type and carrier thymuses, but were approximately 2 fold up-regulated in tumors. As NOTCH1 activating mutations are present in .50% of human T-ALL [14]; exons 26, 27 and 34 which are the common sites of Notch1 mutational activation in murine leukemias [15] were sequenced from Sdl tumors. Only one point mutation in exon 26 was detected out of 13 tumors sequenced. Therefore, activation of Notch1 by point mutation is not a common mechanism in this model.
To further investigate the mechanism of Notch1 activation in Sdl T-ALL, Notch1 transcript levels were investigated by qRT-PCR  with primer pairs spanning several exon-exon boundaries ( Figure 5B). Five of five Sdl tumors examined showed higher levels of expression of 39 exons than 59 exons, and in four of five tumors 39 exons were expressed at higher levels than in normal thymus. These results were consistent with the presence of intragenic deletions removing 59 regions of the Notch1 locus that have been recently reported in murine leukemias. These deletions result in truncated or chimeric transcripts that produce NOTCH1 proteins that are constitutively active [16,17]. Two types of intragenic Notch1 deletions have been reported in murine T-ALLs. Both types of transcripts were shown to be translated beginning at M1727 in exon 28, produce intracellular NOTCH1 (ICN1) and activate a Notch1 reporter [16]. Type 1 were more common, had specific break points that occur immediately adjacent to sequences similar to RAG-signal sequences (RSSs) and had features consistent with being driven by RAG activity. Type 2 deletions were more rare (3 of 10 cell lines examined, two of which were sub-clones of the same tumor) and did not have evidence of RSSlike sequences at their breakpoints [16]. Type 1 deletions break at specific chromosomal locations, so genomic PCR can be used to detect them. No such deletions were detected in Sdl tumors. Type 2 deletions have varying breakpoints as they are not limited to RSS-like sequences, so they are difficult to detect via genomic PCR. However, RT-PCR can be used to detect the resulting abnormal chimeric transcripts. Such transcripts were detected in 12 of 15 tumors examined ( Figure 5C). Sequencing of the primary RT-PCR product from three separate tumors revealed splicing from exon 1 to exon 28. To attempt to clone the breakpoints in Notch1 in Sdl tumors, genomic PCR on a separate cohort of tumors was performed with various forward primers spanning exon 1 The Sdl mutation must map within regions harboring C57Bl/6 alleles in leukemic mice, but must be excluded from regions harboring C57Bl/6 alleles in non-leukemic mice. Analyzing two non-leukemic mice indicates that the Sdl mutation likely lies distal to 14.56 and proximal to 15.91. The two nonleukemic mice were bred and non-carrier status was verified. B) Sanger sequencing traces demonstrating the G to C substitution present in all confirmed Sdl carriers but absent from all wild-type strains examined. Arrow indicates dual G/C peak, asterisk indicates wild-type G peak. C) Sdl mice harbor increased numbers of spontaneous micronuclei. Example flow cytometry plots for wild-type (WT) and Sdl carrier (C) mice for micronuclei detection. Micronucleated normochromatic erythrocytes are propidium iodide (PI) positive but CD71 negative (lower right quadrant) and are expressed as a percentage of total erythrocytes. D) Quantification of micronuclei for sex and aged-matched wild-type (solid bar, WT) (n = 8) and Sdl carriers (striped bar, C) (n = 5 through intron 2 in combination with an exon 27 or exon 28 reverse primer. In 3 of 13 tumors, products were cloned and it was verified that the breakpoints do not possess evidence of RSS-like sequences. Two of three breakpoints had 2-4 bp microhomology ( Figure 5D). Therefore, it is hypothesized that the tumor spectrum of Sdl mice is, at least in part, due to the propensity to develop type 2 non-RAG driven deletions at the Notch1 locus.
To determine if Sdl T-ALLs harbor additional genomic aberrations, array CGH was performed on genomic DNA isolated from Sdl thymic tumors compared to DNA isolated from a noncarrier mouse. Although whole chromosome gains and losses were not detected, many small deletions and amplifications averaging 110 kb in size were present in tumors ( Figure 6). Therefore, the Sdl mutation promotes focal copy number changes and not aneuploidy.

Discussion
We have been studying a novel spontaneous mouse cancer model, Sdl, in which an early-onset T-ALL phenotype is inherited in a dominant manner. We have accumulated evidence that Mcm2 GT/+ mice have been reported to develop tumors, but only after one year of age and with approximately 75% penetrance [4]. Therefore, it has been proposed that a threshold level of MCM proteins (between 35 and 50% of normal for MCM2) is required for sufficient origin licensing to maintain genomic stability and prevent tumor formation [4]. We did not detect a reduction in total or chromatin bound MCM levels in Sdl mice, suggesting that the Mcm4 D573H allele acts in a different mechanism to cause tumorigenesis. However, it is also possible that Mcm4 D573H mice harbor small reductions in MCM levels that are beyond the detection limits of Western analysis. Although detailed aging studies were not presented, Mcm4 GT/+ mice were reported to be apparently normal [3]. However, a thorough study of the tumor phenotype of Mcm4 GT/+ mice will be required to determine the threshold levels of active MCM4 that are required to maintain genomic stability and prevent tumorigenesis.
The tumor spectrum and latency in Sdl is also very different from that observed in Mcm4 chaos3/chaos3 mice [3,6]. The reasons underlying these differences remain to be elucidated. As Mcm4 chaos3/chaos3 and Sdl have been studied on different strain backgrounds, genetic modifiers could contribute to the observed phenotypic differences. Alternatively, the recessively acting hypomorphic Mcm4 chaos3 and the dominantly acting Mcm4 D573H may have different consequences on origin licensing and DNA replication. In addition, a recent report found that MCM 2, 3, 5 and 7 regulate HIF1 activity and this function is likely independent from their function in the heterohexamer. A similar activity was not detected for MCM4 or 6 [18]. Therefore, the reduction in levels of total MCMs seen in Mcm4 chaos3/chaos3 mice could also influence HIF1 activity and have phenotypic consequences. A study of tumor and DNA replication phenotypes for both alleles on the same genetic background will be required to address the reasons for phenotypic differences between the two alleles.
Analysis of T cell differentiation in Sdl carriers revealed subtle defects, with some animals being more severely affected than others. One potential interpretation is that T cell differentiation is mostly normal in Sdl mice until genomic mutations due to replicative stress start to accumulate. In support of this, microarray data failed to detect any transcripts that are significantly differentially expressed in Sdl carrier thymuses compared to wild-type thymuses. Tests for common function did identify differences in genes with common gene functions including protein localization or targeting to mitochondria, chemokine binding or receptor activity and endothelial cell proliferation.
In contrast, many expression differences were detected between wild-type thymuses and Sdl leukemias. Many of the mostly profoundly down-regulated genes in Sdl leukemias are genes such as Prss16 and Tbata that are expressed in thymic epithelial cells [19,20]. This observation likely results from a lower ratio of T cells to thymic epithelial cells in normal thymus than in thymic lymphoma. Although non-T lineage cells are the minority of cells in the developing thymus, they nevertheless impacted our ability to identify genes that are down-regulated during T-ALL formation in Sdl mice. The transcripts with the greatest fold up-regulation in Sdl leukemias compared to normal thymus include genes with unknown function, metabolic genes, genes expressed during T cell activation and Notch1 target genes. RT-PCR in Sdl tumors demonstrated the presence of an aberrant Notch1 transcript splicing from exon 1 to exon 28 in 12 of 15 Sdl leukemias. Genomic PCR on a separate cohort of Sdl T-ALLs was able to clone genomic breakpoints in the Notch1 locus in 3 of 13 tumors. These breakpoints occurred in introns 2 and 27, introns 1 and 27, and introns 2 and 26. It is possible that the exon 1 to 28 splice is favored even when deletions leave more internal exons intact, or that our RT-PCR conditions failed to robustly amplify transcripts containing other aberrant splice variants. Alternatively, the genomic re-arrangements present at the Notch 1 locus may be more complex than can be detected by our genomic PCR. Nevertheless, the detected Notch1 transcript and lack of RSS-like sequences at the cloned breakpoints are both consistent with the presence of type 2 deletions at the Notch1 locus in Sdl T-ALLs. The vast majority of murine T-ALLs previously examined have harbored type 1 RAG-mediated deletions, while type 2 deletions were more rare. A predisposition to T-ALL has also been observed for Mcm2 hypomorphic mice [5,7] and array CHG detected deletions at the Notch1 locus in 4 of 8 of T-ALLs in Mcm2 mice [21]. One possibility to explain the tumor spectrum in Sdl mice and Mcm2 hypomorphic mice is that the integrity of the murine Notch1 locus is sensitive to replicative dysfunction in developing T cells and that replicative stress promotes the formation of type 2 All mcm4 wild-type colonies grew under restrictive conditions and empty vector colonies do not, as expected. A fraction of colonies expressing mcm4 Sdl mutation grew under restrictive conditions. C) A restriction fragment polymorphism was utilized to distinguish mcm4 Sdl from wild-type (WT) mcm4 sequences in the yeast strains described above. One mcm4 wild-type (W) colony and 10 mcm4 Sdl (S) colonies are shown. All freshly isolated mcm4 Sdl colonies grown under permissive conditions (2URA 2TRP) harbor both mcm4 Sdl and mcm4 wild-type sequences due to the presence of TRP1-mcm4 Sdl and URA3-mcm4 plasmids. All mcm4 Sdl colonies that grew under restrictive conditions lost mcm4 Sdl sequences, indicating that growth occurred due to a reversion or gene conversion event involving mcm4 Sdl sequences on the TRP1 plasmid and not due to the ability of mcm4 Sdl to complement the mcm4 genomic deletion. doi:10.1371/journal.pgen.1003034.g004 deletions at the Notch1 locus. As the majority of T cell development is completed by young adulthood, Mcm4 chaos3/chaos3 mice may not experience sufficient replicative stress to cause Notch1 deletions in developing thymocytes, which would allow them survive longer to develop other late-onset tumor types.
Previous array CGH studies of Notch1-driven mouse T-ALLs failed to detect tumor-specific chromosomal aberrations, indicating that chromosomal instability is not a general characteristic of mouse T-ALL [22]. In contrast, array CGH data of Sdl tumors did detect small amplifications and deletions but not whole chromosome gains and losses. This data is consistent with previous observations that an improved growth phenotype found in mcm4 Chaos3/Chaos3 diploid yeast is due to mutations in a few genes and not due to aneuploidy [23]. In addition, recent array CGH experiments on T-ALLs from Mcm2 hypomorphic mice also detected small genomic aberrations [21]. However, aberrations in T-ALLs in Mcm2 mice were primarily deletions, while both amplifications and deletions were found in Sdl T-ALLs. It is possible that functional differences between MCM helicase activity in Sdl and Mcm2 hypomorphic mice could explain this difference. However, it is also possible that strain specific modifiers can impact the types of aberrations generated by replicative dysfunction or selected for during tumorigenesis. Nevertheless, studies in yeast, Mcm2 hypomorphic mice and Sdl mice all support a model that replicative stress can contribute to tumorigenesis by generating smaller chromosomal aberrations and not by causing aneuploidy.
The residue impacted by the observed Mcm4 mutation in Sdl mice is part of the Walker B box, one of the structural motifs in MCM4 that is an integral part of the ATPase active site formed between MCM4 and MCM7 in the heterohexameric complex [24]. Engineering the D to H mutation into the analogous residue in yeast mcm4 failed to complement a mcm4 genomic deletion. Previous studies in yeast where the analogous D residue was mutated to A or T did complement a mcm4 deletion allele [9]. Mutation of the D and the adjacent E residue to N and Q, respectively, (DE.NQ) did however fail to complement [25]. As the D residue in the Walker B box is believed to be important for coordinating the Mg 2+ ion involved in ATP hydrolysis [26], the substitution of a positively charged H residue could result in a greater impact on MCM4 function than would mutation to an A or T ( Figure S5). Given the observation that total and chromatin bound MCM levels are not different in Sdl carrier and wild-type mice, this supports a model in which MCM4 D573H containing helicases are stable, yet functionally inactive.
The role of MCM proteins in promoting genomic instability during human cancer initiation and progression remains unclear. Immunohistochemistry detects MCM protein expression in many human tumor samples, as would be expected for rapidly dividing cells [27]. Knockdown of MCM 2, 3 or 7 in medulloblastoma cell lines caused inhibition of anchorage-dependent and independent growth; while their over-expression promoted cell migration, invasion and increased anchorage-independent growth [28]. MCM7 over-expression in epithelial progenitor cells sensitized mice to carcinogen-induced skin tumors but did not itself drive tumors by 1 year of age [29]. Over-expression of MCM7 alone in the prostatic epithelium did not promote phenotypes. However, over-expression of MCM7 along with a PTEN-targeting micro-RNA cluster encoded within the MCM7 human locus did initiate prostate tumorigenesis [30]. Although mutations in genes involved in DNA damage checkpoints and DNA damage repair are known to contribute to sporadic and hereditary tumorigenesis, it is unclear if genetic changes in the actual components of the replication machinery such as MCM proteins contribute to tumorigenesis in humans. A few point mutations in MCM subunits have been detected in human tumors [31]. However, the functional consequences of these mutations are currently unknown. Given clinical use and preclinical development of compounds that impact replication as cancer chemotherapies, it will be important to elucidate how MCMs contribute to tumor initiation and progression. Although previous studies have uncovered a tumor suppressive activity for Mcms, our studies of the Sdl model indicate that dominantly acting Mcms alleles can be compatible with viability but cause chromosomal abnormalities and highly penetrant tumor formation. Therefore, Mcm mutations with different functional consequences on MCM levels and activity have the potential to act as driver mutations during tumorigenesis.

Ethics statement
Mouse experiments were performed according to the institutional guidelines for animal care under the approval of the IACUC of the University of Minnesota and the University of Wisconsin.

Animals
Sdl arose in the germline of a Rosa-SB11 mouse maintained on the C57Bl/6 background [32]. Wild-type mice were purchased from Jackson Labs or Charles River. Non-carrier sibling mice were utilized as controls. Mcm4 chaos3 mice on the FVB/N genetic background were generously provided by Naoko Shima.

DNA capture and exon re-sequencing
The SureSelect XT Mouse All Exon Kit (Agilent Technologies) was used to capture exonic sequences from genomic DNA purified from a tail clip of a Sdl carrier. Sequencing was performed on the Illumina Genome Analyzer 2 platform as paired-end 76-bp reads. One lane of sequence was generated. Reads were aligned to the MM9 reference genome with BWA v0.5.5 [33]. The GATK (v1.0.4771) [34] was then used to do local realignment of the reads around all indel sites called in the mouse genomes project [35]. The base qualities of the BAM file were recalibrated with the GATK v1.0.4771 by masking all SNP and indel positions called in the mouse genomes project. SNP calling was carried out using SAMtools mpileup/bcftools (v0.1.16) [36]. For SAMtools mpileup, the following options were used: -d 500 -C50 -m3 -F0.002 -aug. The raw sequence data is available under ERA accession number ERP000474.

Array CGH
DNA was purified from thymic tumors from three Sdl mice on the 129S1/SvImJ genetic background. Tail clip DNA from a non-carrier also on the 129S1/SvImJ genetic background was utilized as reference DNA. Hybridizations were performed by WiCell research institute according to manufacturer's recommendations to the mouse CGH 36720 K Whole-Genome Tiling Arrays (NimbleGen). Nim-bleScan, CGH Fusion (RBS v1.0) (Infoquant) software was utilized for analysis and data visualization. Gains and losses were called with the following parameters: average log-ratio threshold of 0.2, a minimum aberration length of 5 probes and maximum p-value of 0.001.

Micronucleus assay
The micronucleus assay was performed essentially as described [37]. 5 week-old females on the FVB/N genetic background were analyzed.

Western analysis
Whole cell protein extracts were purified using RIPA buffer and protease inhibitors (Pierce). Cell fractionation was performed using the Qproteome Nuclear Protein Kit (Qiagen). Proteins were run on a 10% SDS-polyacrylamide gel and transferred to a PVDF membrane. Membranes were incubated with primary antibodies, followed by incubation with the secondary antibody and visualized using Western Lighting Plus-ECL (PerkinElmer).

RT-PCR and quantitative RT-PCR (qPCR)
RNA was isolated by Trizol (Invitrogen) and further purified using the RNeasy Mini Kit (Qiagen). All wild-type and Sdl carrier thymuses used for RNA extraction were from 21-day-old mice. Wild-type thymuses were isolated from non-carrier littermates. First strand cDNA was generated using a 20:80 mix of polyT:random decamer primers (Ambion Retroscript). Real-time PCR was completed using the LightCycler (Roche Applied Science). Briefly, LightCycler FastStart DNA MasterPLUS SYBR Green I kit (Roche Applied Science) was used with 5 ml cDNA from the reverse transcriptase reaction. Relative mRNA expression levels were calculated as described previously [38]. Go taq (Promega) was used for non-quantitative PCRs. t-tests or ANOVA followed by Tukey post hoc analysis were performed using Prism software.

Southern analysis
DNA was purified from thymic tumors, and TCR b rearrangement was detected as described [8].

Microarray analysis
RNA was isolated from female mice by Trizol (Invitrogen) extraction followed by further purification using the RNeasy kit (Qiagen). RNA from three animals for each group (21-day-old wild-type thymus, 21-day-old Sdl carrier thymus and Sdl overt thymic tumors) was pooled per array. Four RNA pools were analyzed per group. Amplified sense-strand cDNA was generated using the Ambion WT expression kit (Applied Biosciences), fragmented and labeled with the GeneChip WT Terminal Labeling Kit (Affymetrix) and hybridized to GeneChip Mouse Exon 1.0 ST Arrays (Affymetrix) by the University of Wisconsin Biotechnology Center Gene Expression Center. The data were normalized using rma [39] as implemented in the xps system, available at Bioconductor (www.bioconductor.org). Summaries at the transcript level using probes given a core level ranking were analyzed, as this annotation level is most conservative [40]. Antigenomic probes were used for background correction. EBarrays, an empirical Bayes hierarchical modeling approach, was used to identify differentially expressed (DE) genes [41,42]. With EBarrays, a gene was considered significantly DE if the posterior probability of DE under the log-normal Normal moderated variance (PP.LNNMV.DE) model exceeded 0.95, as this threshold controls the posterior expected false discovery rate (FDR) at 5% [42]. We also considered results from gene specific ttests with q-values calculated from the t-test p-values [43]. A list of genes with q-values ,0.05 has expected FDR of 5%. To help ensure that results were robust to the statistical method used, we define a gene to be significantly DE if its PP.LNNMV.DE.0.95 and its q-value ,0.05. Tests for enrichment of common function among sets of differentially-expressed genes were carried out using data from the Gene Ontology (GO) annotations and the Kyoto Encyclopedia of Genes and Genomes (KEGG). The R package allez was used to perform tests of enrichment for each GO category and KEGG pathway [44]. In general, the interpretation of p-values resulting from enrichment tests is not straightforward due to the many dependent hypotheses tested. Furthermore, the enrichment test tends to result in small p-values when groups with few transcripts are considered. The statistical methods underlying allez adjust for these factors, allowing increased power and sensitivity for identifying sets that are biologically meaningful.

Yeast studies
The mcm4 deletion strain and TRP1 plasmids were described previously [9] and were generously provided by Anthony Schwacha. The Sdl mutation in the equivalent residue (D632H) was generated by site-directed mutagenesis utilizing PfuUltra Hotstart taq (Agilent) followed by DpnI digestion and verified by sequencing. Cloning into the TRP1 plasmid added a HA/H10X tag that allowed Mcm4 D632H protein expression to be verified by Western blot. The engineered Sdl mutation creates a restriction fragment polymorphism that can be detected by PCR amplification followed by BssSI digestion.

MEF isolation and analysis
MEFs were isolated from 12.5-15.5 dpc embryos on the 129S1/ SvImJ genetic background. All MEFs were utilized at p3 or lower. For chromosome break analysis, MEFs were treated with aphidicolin or DMSO as vehicle only control for 24 hours. Colcemid was added for the final five hours prior to harvest. Three separate cultures were analyzed per genotype. Harvested cells were pelleted and resuspended in 0.075 M KCl solution and incubated at room temperature for 20 minutes, pelleted, and re-suspended in 3:1 (vol/ vol) methanol/acetic acid for 10 minutes twice. Cells were spread on a slide and stained with Giemsa stain (Sigma GS500) in a 1:20 dilution in water. Slides were washed briefly with 16 PBS and observed.

Histopathology
Tissues were fixed in 10% formalin, embedded in paraffin, sectioned and stained with hematoxylin and eosin by the University of Wisconsin RARC comparative pathology laboratory. Analysis was performed by the UWCCC Experimental Pathology Core.

Flow cytometry
Cryopreserved tumor cell samples were briefly thawed at 37uC and then incubated for 15 min at 37uC in DMEM containing 20% FBS, 10 U/ml Heparin (Sigma), and 0.25 mg/ml DNAse 1 (Roche). Cells were pelleted and re-suspended in HBSS without Ca/Mg, 2% FBS, 2.5% cell dissociation buffer (GIBCO, Invitrogen), 100 U/ml Penicillin G, and 100 mg/ml streptomycin. Tumor cell suspensions were pre-incubated with antibodies to FccRII/III (BD PharMingen) for 3 min to prevent nonspecific binding of labeled antibodies to the cell surface. They were then stained with monoclonal antibodies conjugated with biotin, phycoerythrin (PE), or TRI-COLOR (TC) for 20 min at 4uC. After washing, strepavidin-allophycocyanin (APC) was added for 20 min at 4uC to provide fluorescent signal for biotinylated antibodies. Staining was assessed with a FACSCalibur flow cytometer (Becton Dickinson) in four-color mode using CellQuest Pro and FlowJo software for analysis. For thymocyte analysis, dissociated cells were incubated with anti-CD4 FITC and anti-CD8 APC to examine CD4, CD8 and CD4/8 double positive T cells. To examine double negative (DN) populations, dissociated thymocytes were stained with a lineage cocktail (anti Gr-1-Biotin, cdTCR-Biotin, B220-Biotin, NK1.1-Biotin, TER-119-Biotin, Streptavidin FITC, CD4-FITC, Mac-1 FITC, CD8-FITC) along with CD25-PE, CD44-AlexaFluor700 and DAPI. DN thymocytes were defined as the FITC negative gate. At least 10,000 DN cells per animal were analyzed for CD25 and CD44 expression. DN1 cells were defined as Lin2 CD252 CD44+, DN2 cells as Lin2 CD25+ CD44+, DN3 cells as Lin2 CD25+ CD442 and DN4 cells as Lin2 CD252 CD442. All animals utilized for analysis of thymocyte developmental stages were female progeny of one Sdl mouse on the FVB/N genetic background.
Details of exon capture methods, antibodies utilized and primer sequences for genotyping and RT-PCR are provided as Text S1.  Figure S2 Inter-animal differences in T cell developmental defects in Sdl carriers. CD44 and CD25 staining of Lin-thymocytes was used to examine DN stages of T cell development in four non-carrier (wildtype, WT) and four Sdl carrier (C) thymuses. Differences in the percent of cells at the DN1 (CD252 CD44+) stage of development were statistically significantly decreased in Sdl carriers compared to wild-type (See Table 2), with animals C3 and C4 more severely affected than animals C1 and C2. Animals C3 and C4 also appear to have a defect at the DN3 (CD25+ CD442) to DN4 (CD252 CD442) transition. (TIF) Figure S3 qRT-PCR detects no differences in expression levels of genes in the Sdl interval between wild-type (solid bars) and carrier thymuses (striped bars). Wild-type thymus is normalized to 1.  [45] with the equivalent D residue highlighted (B) was utilized to model the impact of the D to H Sdl mutation (C) and the D to A mutation (D) that was previously shown in complementation tests to not impact the biologic activity of S. cerevisiae mcm4 [9]. The D to H substitution is predicted to have a greater impact on protein structure than the D to A substitution. Structures were visualized using PyMol [46].  Summary of exon and splice site sequences with fewer than 106 coverage by exon-capture re-sequencing that were examined by PCR amplification followed by Sanger sequencing. (XLSX)

Table S3
Non-synonymous changes identified by exon capture on proximal chromosome 16. * The change in Top3b falls outside of the Sdl interval (see Figure 2A) and was not confirmed by traditional Sanger sequencing in genomic DNA from the mouse utilized for exon capture or other carrier Sdl mice. Visual inspection indicates that the base lies adjacent to a low-complexity polyG tract and that the alternate call is likely due to read misalignment. (XLSX) Text S1 Supplemental methods. (DOC)