A unique mutator phenotype reveals complementary oncogenic lesions leading to acute leukemia

DNA replicative stress, typically in response to oncogene-induced hyperproliferation, has been linked to malignant transformation (1–3). Although the precise mechanisms by which replicative stress might result in cancer remain unknown, it is thought that chronic replicative stress leads to replication fork stalling, collapse, and subsequent DNA double strand breaks (DSBs) at or near the site of replication fork collapse (2). Inefficient repair of these DNA DSBs leads to genomic instability reflected by indels and structural variations involving genes important for malignant transformation (2).

Minichromosome maintenance complex component 2 (Mcm2) is a core component of the DNA replication-licensing complex required for replication initiation during S phase. A multi-subunit complex containing Mcm2–7 is responsible for the initial unwinding of DNA (4, 5). We previously generated a mouse strain containing a CreERT2 cassette flanked by an IRES knocked into the 3′ UTR of Mcm2, creating an Mcm2^{IRES−CreERT2} allele (henceforth referred to as Mcm2^cre) (6). Unexpectedly, this knockin allele resulted in diminished expression of Mcm2 protein. Mcm2^cre/cre mouse embryo fibroblasts expressed approximately one-third the amount of Mcm2 protein compared with wild-type (WT) controls. Remarkably, almost all Mcm2^cre/cre mice develop a lethal precursor T cell lymphoblastic leukemia/lymphoma (pre-T LBL) within 4 months of age. Previous studies identified abnormal karyotypes and increased levels of chromosome breaks in cultured cells with reduced expression of Mcm proteins (7, 8), and array comparative genomic hybridization demonstrated that pre-T LBL in Mcm2^cre/cre mice had numerous small (average < 0.5 Mbp) genomic deletions, including several genes known to be relevant for human pre-T LBL, such as E2a (Tcf3) and Pten (9), leading to the hypothesis that Mcm2 deficiency led to deletions of important tumor suppressor genes, resulting in malignant transformation. These findings led us to view Mcm2^cre/cre cells as a potential tool for inducing mutation, analogous to ionizing radiation (10), ethylnitrosourea (11), or retroviral insertion (12). In this view, dysregulation of Mcm2 protein is not directly oncogenic; instead, Mcm2 dysregulation leads to widespread genomic deletions, some of which are oncogenic. Cells that undergo a combination of mutations that dysregulate several complementary, collaborative pathways are selected in vivo because of a fitness advantage and emerge as a malignancy.

A NUP98-HOXD13 (NHD13) fusion gene has been identified in patients with myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) (13). Expression of an NHD13 fusion in the hematopoietic compartment of mice led to overexpression of Hoxa cluster genes and resulted in a highly penetrant MDS phenotype (14). Approximately 80% of NHD13 mice develop acute leukemia, most commonly AML, with the remainder dying from complications of MDS, such as severe anemia or infection, without signs of leukemic transformation (14). Leukemic transformation was frequently associated with spontaneous mutations of Nras, Kras, or Cbl, suggesting that these mutations collaborated with the NHD13 fusion and were biologically selected in vivo (15).

In addition to targeted resequencing of candidate genes, retroviral insertional mutagenesis (RIM) identified genes whose overexpression would collaborate with an NHD13 transgene to induce AML (16). Although RIM can identify gene inactivation events as well as gene activation events, tumor suppressor genes have typically not been identified through RIM screens, perhaps because 2 alleles need to be targeted for complete inactivation of many tumor suppressor genes, as opposed to a single allele for gene activation events (17). We hypothesized that the unique “deleter” phenotype found in Mcm2^cre/cre mice could be used to identify tumor suppressor genes in the context of AML, by crossing the NHD13 transgene onto an Mcm2^cre/cre background.

Mcm2^cre/cre NHD13⁺ mice do not develop AML. To determine whether Mcm2-deficient mice (Mcm2^cre/cre) can be used to identify tumor suppressor genes important for the development of AML, we generated Mcm2^cre/cre NHD13⁺ mice. Similar to prior studies, Mcm2 protein expression in thymus from 1-month-old Mcm2^cre/cre mice was only 28% that of WT mice; addition of the NHD13 transgene had no effect on the Mcm2 protein level (Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.131434DS1). Mice were euthanized when they presented with signs of leukemia, including weight loss, kyphosis, lethargy, visible lymphadenopathy, and dyspnea. Mcm2^cre/cre mice had a markedly decreased survival compared with mice that were heterozygous for this allele (Figure 1A), and adding the NHD13 transgene to the Mcm2^cre/cre mice (Mcm2^cre/cre NHD13⁺) led to a modest but significant decrease in survival. Previous reports had demonstrated that Mcm2^cre/wt mice were not predisposed to malignancy (6). To determine whether they had a subtle predisposition toward myeloid malignancy that could be uncovered by the NHD13 transgene, we compared the median survival of Mcm2^cre/wt NHD13⁺ to Mcm2^wt/wt NHD13⁺; the median survival was not significantly different (259 vs. 324 days) (Figure 1A) and was similar to our previous reports with NHD13⁺ mice (14, 18). In summary, these results showed that the NHD13 transgene did not accelerate the onset of disease in heterozygous Mcm2^cre/wt mice and caused a modest disease acceleration in Mcm2^cre/cre mice.

Figure 1

Mcm2^cre/cre NHD13⁺ mice do not develop AML. (A) Survival curve for all 6 genotypes, compared by log-rank test. *P < 0.05; ***P < 0.001. (B) Representative flow cytometry profile of malignant thymocytes from Mcm2^cre/cre NHD13^– and Mcm2^cre/cre NHD13⁺ mice with pre-T LBL. (C) Immunophenotype summary of pre-T LBL samples from thymus of Mcm2^cre/cre NHD13^– (n = 33) and Mcm2^cre/cre NHD13⁺ (n = 45) mice. DN, double negative.

Necropsy of both Mcm2^cre/cre NHD13⁺ and Mcm2^cre/cre NHD13^– mice revealed markedly enlarged thymus, splenomegaly, and hepatomegaly. Complete blood counts (CBCs) from the Mcm2^cre/cre NHD13^– cohort typically showed leukocytosis but were otherwise normal, whereas CBCs from the Mcm2^cre/cre NHD13⁺ cohort showed both leukocytosis and leukopenia, anemia, and thrombocytopenia, consistent with prior studies on NHD13 mice (ref. 14 and Supplemental Table 1). Leukemic subtype determined by flow cytometry demonstrated all but 1 of the Mcm2^cre/cre mice (with or without the NHD13 transgene) had infiltration of thymus, bone marrow, and spleen with malignant thymocytes that stained for CD4, CD8, or both. The diagnosis of pre-T LBL was further supported by clonal VDJ or DJ rearrangements of the Tcrb gene (Supplemental Table 2) and infiltration of parenchymal organs, such as the liver, kidney, and lung, with CD3⁺ lymphoblasts (Supplemental Figure 2).

The stage of thymocyte differentiation was influenced by the presence of the NHD13 transgene. Whereas most Mcm2^cre/cre NHD13⁺ mice displayed a CD4⁺CD8⁺ (DP) immunophenotype, a wide spectrum of pre-T LBL immunophenotypes were identified in the thymus of Mcm2^cre/cre NHD13^– mice, including DP, CD4⁺CD8^het, CD8⁺CD4^het, CD4⁺CD8^–, and CD4^–CD8^– (Figure 1, B and C). Notably, none of the 46 Mcm2^cre/cre NHD13⁺ mice developed an AML. The Mcm2^cre/wt NHD13⁺ and Mcm2^wt/wt NHD13⁺ mice developed disease with a similar immunophenotype, primarily AML, and less commonly pre-T LBL or B cell precursor acute lymphoblastic leukemia (BCP-ALL) (Supplemental Table 1). In summary, these results demonstrated that Mcm2^cre/cre NHD13⁺ mice did not develop AML, nor did Mcm2^cre/wt NHD13⁺ mice show accelerated myeloid leukemic transformation compared with Mcm2^wt/wt NHD13⁺ transgenic mice, suggesting that this strategy did not uncover AML tumor suppressor genes.

Recurrent deletions in pre-T LBL. Based on prior studies (9), we suspected that the pre-T LBLs identified in Mcm2^cre/cre NHD13⁺ mice were initiated, at least in part, by about 0.5 Mb genomic deletions. In addition, although there was no acceleration of AML in Mcm2^cre/wt NHD13⁺ mice compared to Mcm2^wt/wt NHD13⁺ mice, we wished to determine whether Mcm2^cre/wt NHD13⁺ mice had recurrent genomic deletions. Copy number alteration (CNA) analysis for pre-T LBL and AML samples was determined by sparse whole-genome sequencing (WGS; see Methods), an approach that can accurately identify copy number losses (deletions) and gains at a resolution of roughly 125 kb and map alteration breakpoints to a resolution of approximately 25 kb (19). As anticipated, we detected 4–32 small (100–1000 kb) deletions and gains in pre-T LBL from Mcm2^cre/cre samples (Figure 2A and Supplemental Figure 3, A–C), with focal deletions more common than gains (Supplemental Figure 3D).

Figure 2

Mcm2^cre/cre mice show recurrent deletions. (A) Whole-genome view of copy number alteration (CNA) analysis for germline, pre-T LBL (Mcm2^cre/cre NHD13^–, n = 11 mice; and Mcm2^cre/cre NHD13⁺, n = 18 mice), and AML (Mcm2^cre/wt NHD13⁺, n = 8); samples were more than 70% tumor tissue (thymus or BM) based on flow cytometry. Mouse chromosomes 1–19 and X are indicated. (B and C) Zoomed-in view of (B) Pten (chromosome 19) and (C) Bcl7a (chromosome 5) regions. Red indicates gain; blue indicates loss. Copy number loss is proportional to color. Darker blue is consistent with homozygous loss; lighter blue suggests heterozygous loss. Asterisk indicates the gains or deletions present in germline samples compared with reference C57BL/6 genome (mm9).

Inspection of Figure 2A shows recurrent interstitial deletions of chromosomes 5, 6, 7, 9, 10, 12, 14, 17, and 19; Mcm2^cre/cre NHD13⁺ pre-T LBL showed a deletion pattern similar to Mcm2^cre/cre NHD13^– pre-T LBL. We saw no recurrent acquired copy number alterations (CNAs) in the Mcm2^cre/wt NHD13⁺ AML samples, indicating that the heterozygous Mcm2^cre allele did not predispose to recurrent deletions, even in the presence of an NHD13-sensitizing transgene. A higher magnification view (Figure 2, B and C) demonstrates that (a) deletions can be more (e.g., Pten) or less (e.g., Bcl7a) tightly clustered, (b) deletions can be homozygous or heterozygous, and (c) no deletions are identical between different samples.

There were 15 common deleted regions (defined as deleted or gained in at least 15% of samples) in the 29 pre-T LBL primary tumors (11 Mcm2^cre/cre NHD13^– and 18 Mcm2^cre/cre NHD13⁺) (Supplemental Table 3); 9 regions were deleted in more than 60% of the samples. These included expected regions, such as Tcra and Tcrb, as well as previously reported regions that encompass known tumor suppressor genes, such as Pten (20), Tcf3 (21), Cdkn1a (22), Bcl11b (ref. 23 Supplemental Figure 4), and Zfp36l2 (24), and genes involved in normal T cell differentiation, including Tcf12, Bcl7a, and Bcl7c. A single region from chromosome 8 was recurrently amplified, specifically in Mcm2^cre/cre NHD13⁺ mice; there was no obvious candidate oncogene in this region.

To validate the small (125–1000 kb) deletions detected by sparse WGS, we performed conventional WGS, with about 28× coverage, for 3 Mcm2^cre/cre NHD13^– pre-T LBLs, using nonmalignant tail DNA as a germline control. WGS identified 93 genomic deletions in 3 Mcm2^cre/cre NHD13^– pre-T LBL (Supplemental Table 4). Eighty-five percent (57/67) of the CNAs identified by sparse WGS were verified (Supplemental Figure 5A). An additional 36 deletions were identified by WGS; most of these were smaller than 125 kb (Supplemental Table 4), consistent with the anticipated resolution of the sparse WGS method. When only deletions greater than 125 kb were considered, samples 2739 and 2854 were 100% concordant, whereas sample 2883 was only 68% concordant. One potential reason for the poor correlation in sample 2883 is the observation that this sample had substantial contamination with WT DP thymocytes (flow cytometry showed 60.5 % CD4⁺CD8^het and 35.9% WT CD4⁺CD8⁺ thymocytes) (Supplemental Figure 5, B–D).

Fifteen of the deleted regions involved TCR α, β, γ, or δ. Because these regions undergo programmed deletions in the course of normal development because of VDJ recombination, we focused our analysis on the remaining 78 deletions that were linked to the Mcm2 deficiency. We identified 11 homozygous deletions in the 3 samples; these homozygous deletions included Pten, Dnmt3a, Tcf3, and Bcl7a (Supplemental Table 5). An additional 9 deletions were identified that involved adjacent, but not overlapping, regions (Supplemental Table 5). Sample 2754 had 1 large Ikzf1 deletion, encompassing the entire gene, and a focal 9177-bp deletion, which deleted exons 4 and 5. Reverse transcriptase PCR (RT-PCR) revealed several splice forms of Ikzf1 not seen in WT BM (Supplemental Figure 5, E–G), all missing exons 4 and 5. The interstitial deletion of Ikzf1 and aberrant splice forms was reminiscent of IKZF1 deletions and aberrant isoforms associated with human lymphoid leukemia (25). Detailed analysis of the sequence at the deletion junction of the 78 deleted regions in pre-T LBL samples revealed that 51 of the 78 junctions had 1–5 bp of microhomology (Supplemental Table 4). Twenty-five junctions contained nontemplated insertions at the breakpoint junctions; some of these were quite extensive, 1–50 nucleotides in length.

We next used whole-exome sequencing (WES) to determine whether small indels or single nucleotide variants (SNVs) contributed to the development of pre-T LBL. WES results revealed 104 Tier 1 mutations in the 29 pre-T LBLs, or an average of 3 per tumor. Only 6 of the 29 samples had mutations in genes known or suspected to be involved in pre-T LBL, 3 Notch1, 2 Tp53, and 1 Ezh2 mutation (refs. 24, 26 and Supplemental Table 6). All of these mutations were confirmed (Supplemental Figure 6) by Sanger sequencing of tumor and normal tissue.

Ongoing deletions in Mcm2^cre/cre cell lines. We established immortal, cytokine-independent cell lines from 10 Mcm2^cre/cre pre-T LBL samples. Because the cell lines were uncontaminated by any normal tissue, we used genomic DNA from the cell lines to validate loss of DNA and protein expression (Supplemental Figure 7, A–C). The anticipated homozygous losses for Zfp36L2, Tcf3, and Pten were confirmed (Supplemental Figure 7A), as well as loss of Pten protein (Supplemental Figure 7C). However, sample 2773 remained positive in the PCR assay but did not produce Pten protein. A diagram of the deletions (Supplemental Figure 7B) demonstrates that although the deletions from 2773 were not overlapping, each allele had deleted substantial portions of the Pten coding sequence. Mcm2 protein expression was similar in the Mcm2-deficient cell lines and primary samples (Supplemental Figure 7D). To determine (a) whether the cell line that emerged in tissue culture represented the predominant clone in vivo and (b) whether the cell lines continued to undergo genomic deletions, we compared CNAs in primary tumor, early-passage (TCE, 1–1.5 months in culture) and late-passage (TCL, 2–6 months in culture) pre-T LBL cell lines. In all cases, the global pattern of deletions was similar to the primary tumor, indicating that the cell line that emerged in tissue culture was representative of the primary tumor (Figure 3). Deleted regions were scored for each trio and summarized in Table 1. A total of 135 deletions were identified; 90 were stable and present in all 3 samples, suggesting that these were truncal lesions. Twenty-two deletions were acquired at the early-passage stage; these could have been acquired during tissue culture or could have been present as a minor clone in the primary tumor. Only 10 new deletions were acquired at the late-passage stage, compared with 90 that were present initially and persistent.

Figure 3

Stable and ongoing CNAs in Mcm2^cre/cre pre-T LBL. CNAs of trios from mice 2875, 2854, 2880, 2730, 2773, 2696, 2795, 2869, 2883, and 2641 primary pre-T LBL tumor (PTT), and pre-T LBL cell line at early passage (1–1.5 months; TCE) and late passage (2–6 months; TCL). Color code as in Figure 2A.

Table 1

Summary of deletion patterns in Mcm2^cre/cre pre-T LBL cell lines

To gain further insight into the frequency of ongoing deletions in the Mcm2^cre/cre cell lines, we obtained single-cell clones by plating late-passage 2696 cell line at a limiting dilution. Five of 96 wells expanded and were analyzed for CNAs. Supplemental Figure 7E shows new deletions present in the single-cell clones, an average of 5 new deletions per clone compared with the late-passage cell line. These deletions involved Rpl5, Cntnap2, and Ccdc genes, which were not recurrently deleted in other Mcm2^cre/cre pre-T LBL samples; taken together, these findings suggest that although the deleter phenotype had become less pronounced during in vitro culture, ongoing interstitial deletions could be identified in the Mcm2^cre/cre cell lines.

Large Clone Capture sequencing identifies frequent Notch1 interstitial deletions. Mutations involving the extracellular heterodimerization (HD) domain and/or the C-terminal proline, glutamic acid, serine, and threonine (PEST) domain of NOTCH1 are found in more than 50% of both human and murine pre-T LBLs (26, 27). In addition, a recurrent deletion of Notch1 exon 1 and the adjacent 5′ regulatory sequences has been identified in mice but not humans (28, 29). This deletion leads to use of alternate Notch1 transcript initiation upstream of exon 27 and translation initiation at M1727 within exon 28, resulting in production of a truncated Notch1 protein that lacks the extracellular ligand-binding domain, retains the transmembrane domain and sensitivity to γ-secretase inhibitors, and is functionally similar to a NOTCH1 HD mutation (28). Given these findings, we were puzzled that only 3 Notch1 mutations were identified by WES. Because the WGS revealed an interstitial deletion in 1 of 3 samples, we hypothesized that smaller (5–125 kb) interstitial deletions not identified in the sparse sequence CNA assay might be common events in these pre-T LBLs. “Large Clone Capture” sequencing (LCC-Seq) is a technique that allows for the custom capture and next-generation sequencing of specific genomic regions of interest, similar in concept to WES (30). To search for Notch1 deletions, as well as identify precise breakpoints involved in deletions, we obtained bacterial artificial chromosome (BAC) clones that covered regions of chromosomes 2 (Notch1, Nup214-Abl1, and Ptpn1), 7 (Bcl7c), 10 (Tcf3), and 19 (Pten).

LCC-Seq identified 28 deletions involving Notch1 in 60 independent samples (Supplemental Table 7). Those deletions led to loss of Notch1 exons 2, 1–2, 1–4, 1–19, 1–23, 1–24, 1–27, 3–27, 16–26, or 16–27 (Supplemental Table 7 and Figure 4A). Aberrant Notch1 mRNA splice junctions were detected in samples that had deletions that retained exon 1 (Figure 4B and Supplemental Figure 8). Importantly, all the aberrant splice forms (loss of 3–27, 16–27) retained exon 28, which contains an alternate translation initiation site at Notch1 M1727 (28). Western blot and RT-PCR indicated that samples with Notch1 interstitial deletion produced an ICN protein and high Hes1 expression (Figure 4, C and D).

Figure 4

Interstitial deletions lead to activation of Notch1. (A) Summary of Notch1 deletions. Notch1 (ENSMUSG00000026923) exons are indicated in blue. Functional domains (EGF-like ligand-binding domain, HD, and intracellular domain of Notch1 [ICN]) as shown. Deleted exons of Notch1, frameshift mutations, missense mutations, and in-frame insertion are indicated. (B) Aberrant Notch1 mRNA splice forms in 2880, 2963, 2748, 2973, 2008, and 2897 thymus and 2773 TCL. RT-PCR for WT (WT thy1), 2880, 2963, 2008, and 2897 thymus and 2773 TCL was performed by using primers located in exons 1 and 29 of Notch1 to detect the junction of exon 2 and exon 28. RT-PCR for WT (WT thy2), 2748, and 2973 thymus was performed by using primers located in exons 13 and 29 of Notch1 to detect the junction of exon 15 and exon 28. The lanes were grouped from different gels as indicated by the vertical lines. (C) ICN expression in pre-T LBL primary tumor and pre-T LBL cell line. The lanes were grouped from different gels as indicated by the vertical lines. (D) RT-PCR analysis of Notch1 target Hes1 expression in WT thymus and pre-T LBL with Notch1 mutations. Error bars represent standard deviation of 3 technical replicates (n = 3).

To determine whether sustained high-level Notch1 signaling was required for vigorous growth of the Mcm2^cre/cre TCL, we treated Notch1-deleted cell lines with Compound E, a γ-secretase inhibitor that prevents cleavage of the membrane-bound ICN and resultant transport of the transcriptionally active ICN to the nucleus (26). Compound E did not affect growth of 2795 TCL, a cell line that was WT for Notch1 (Supplemental Figure 9A) but significantly inhibited the growth and ICN expression of Mcm2^cre/cre pre-T LBL cell lines with Notch1 mutations (Supplemental Figure 9, B–F). These results indicated that Mcm2^cre/cre TCLs require ongoing Notch1 signaling for active proliferation in vitro.

Mcm2^cre/cre NHD13⁺ mice develop BCP-ALL. In addition to the pre-T LBL detected in the thymus, 12 of 46 Mcm2^cre/cre NHD13⁺ mice had an unexpected B cell expansion in the BM and spleen (Figure 5A and Supplemental Table 1). None of the Mcm2^cre/cre NHD13^– mice showed this phenotype, in which a predominant CD19⁺B220^–/dim or CD19^–B220^dim population was identified in the BM and spleen, while the thymus was infiltrated with CD4⁺CD8⁺ pre-T LBLs (Figure 5A). Histological analysis showed infiltration of the BM, spleen, and kidney with clusters of B220⁺ cells; conversely, the thymus did not stain for B220 but stained for CD3 (Figure 5B), and BM was replaced with lymphoblasts (Figure 5C). The diagnosis of BCP-ALL was further supported by the presence of clonal Igh gene rearrangements (Figure 5D and Supplemental Table 8). In addition to BCP-ALL in Mcm2^cre/cre NHD13⁺ mice, 10% of Mcm2^cre/wt NHD13⁺ mice and 11% of Mcm2^wt/wt NHD13⁺ mice developed BCP-ALL (Figure 5E and Supplemental Table 8). The ratio of AML to BCP-ALL in Mcm2^cre/wt NHD13⁺ and Mcm2^wt/wt NHD13⁺ mice was 7:1 and 7:1, respectively, similar to previously published findings for NHD13 mice (14). However, the ratio of AML to BCP-ALL in the Mcm2^cre/cre NHD13⁺ mice was 0 because no Mcm2^cre/cre NHD13⁺ mice developed AML. The skewed proportion of BCP-ALL in Mcm2^cre/cre NHD13⁺ mice compared with Mcm2^wt/wt NHD13⁺ indicated that the Mcm2 hypomorph dramatically accelerated the development of BCP-ALL, but not AML, in the NHD13⁺ background (Table 2).

Figure 5

Mcm2^cre/cre NHD13⁺ mice develop BCP-ALL. (A) Flow cytometry plots of BM and spleen (SP) stained with CD19 and B220 and thymocytes (thy) stained with CD4 and CD8 from Mcm2^cre/cre NHD13⁺ mice with BCP-ALL. WT BM, SP, and Thy are controls. (B) Hematoxylin and eosin (H&E) and B220 and CD3 IHC of infiltrated BM, thymus, and kidney of mouse 2725 with concurrent BCP-ALL and pre-T LBL. Scale bar: 200 μm. (C) May-Grünwald-Giemsa–stained BM lymphoblasts from a mouse with BCP-ALL (number 2811). Scale bar: 50 μm. (D) Igh gene rearrangement assay for 2697 BM; clonal fragments indicated by a red asterisk were sequenced. (E) Frequency of leukemia subtypes by genotype (Mcm2^cre/cre NHD13^– [n = 33], Mcm2^cre/cre NHD13⁺ [n = 46], Mcm2^wt/wt NHD13⁺ [n = 18], and Mcm2^cre/wt NHD13⁺ [n = 41]).

Table 2

Total number of AMLs and BCP-ALLs in mice with indicated genotype

To gain insight into the leukemic transformation of Mcm2^cre/cre NHD13⁺ B cell precursors, we assessed the differentiation of B and T cells from Mcm2^cre/cre NHD13⁺ mice at 1 month of age, before overt leukemic transformation. Flow cytometry showed a modest decrease in B220⁺CD19⁺ precursors in Mcm2^wt/wt NHD13⁺ and Mcm2^cre/cre NHD13^– compared with WT mice; however, there was a marked decrease in B220⁺CD19⁺ precursors in Mcm2^cre/cre NHD13⁺ BM, suggesting impaired B cell maturation. Subfractionation of B cell precursors revealed a 7-fold decrease (21% vs. 3%) of B220⁺CD43^– pre-B cells in the Mcm2^cre/cre NHD13⁺ BM, while the proportion of pro-B cells was similar in all 4 genotypes (Supplemental Figure 10, A–C), suggesting a block in differentiation at the pro-B to pre-B stage of maturation. In contrast, we detected no clear abnormalities of T cell differentiation at 1 month of age in any thymocyte or splenocyte of the genotypes studied (Supplemental Figure 10, D and E).

Recurrent CNAs in BCP-ALL. Similar to pre-T LBL samples, we identified recurrent acquired CNAs in Mcm2^cre/cre NHD13⁺ BCP-ALL (Figure 6A and Supplemental Figure 11A). We also assessed CNAs in 4 Mcm2^cre/wt NHD13⁺ samples; as opposed to the pattern of 125–1000 kb deletions seen in Mcm2^cre/cre malignancies (both pre-T LBL and BCP-ALL), there were few 125–1000 kb deletions in the Mcm2^cre/wt NHD13⁺, consistent with prior observations that mice heterozygous for an Mcm2^cre allele are not prone to the deleter phenotype. Comparison of CNAs from pre-T LBL and BCP-ALL from the same mouse showed no similarities between the pairs (Supplemental Figure 11, B–D), indicating that the T and B cell malignancies did not arise from a common precursor.

Figure 6

CNA analysis of BCP-ALL. (A) CNA analysis of germline and BCP-ALL (Mcm2^cre/cre NHD13⁺ [n = 7] and Mcm2^cre/wt NHD13⁺ [n = 4] mice); samples were more than 60% tumor tissue (BM) based on flow cytometry. The IDs of Mcm2^cre/cre NHD13⁺ mice are indicated on the right side. Color code as in Figure 2A. (B and C) CNAs for Pax5 region (B) and Cebpb/Ptpn1 region (C). Color code as in Figure 2A. (D) PCR amplification of Ptpn1-deleted region. Faint signals in 2703 and 2725 could be due to haploinsufficiency or contamination with nonmalignant cells. (E) Ptpn1 mRNA expression, samples, and copy number of Ptpn1 (CNAs) are indicated. Error bars represent standard deviation of 3 technical replicates (n = 3). mRNA expression is markedly decreased in samples with 2 copies lost.

Five of 7 Mcm2^cre/cre NHD13⁺ BCP-ALLs showed deletion involving Pax5, one of the most frequently mutated genes in human BCP-ALL (Figure 6B and Supplemental Table 9). Moreover, the same 5 samples showed deletion (4 of 5 homozygous) of a region that encompassed Cebpb and Ptpn1 (Figure 6, C and D, and Supplemental Table 9). However, Cebpb is not highly expressed in normal B cell precursors, whereas Ptpn1 is highly expressed in B cell precursors. Figure 6E shows marked downregulation of Ptpn1 in 4 samples with putative homozygous deletions.

Recurrent Nup214-Abl1 fusion gene in Mcm2^cre/cre NHD13⁺ BCP-ALL. Four Mcm2^cre/cre NHD13⁺ BCP-ALLs showed copy number gain (estimated 1–2 copies) of a region bounded by Abl1 and Nup214 (Figure 7A and Supplemental Table 9). Given that NUP214-ABL1 fusions have been identified in patients with pre-T LBL (31) and BCP-ALL (32), we hypothesized that Nup214-Abl1 gains could result in a Nup214-Abl1 fusion. This putative fusion could be produced by an episome, as shown in human pre-T LBL (31), or by a tandem duplication (Figure 7B). We designed primers to amplify a fusion between Nup214 exon 23, 29, 31, 32, or 34 and Abl1 exon 2 based on the fusions in human pre-T LBL (31). A fusion PCR product was identified in all 4 samples that showed Nup214-Abl1 copy number gain (samples 2842, 2811, 2725, and 2703). In addition, a Nup214-Abl1 fusion was identified in 1 sample (sample 2905) that did not show an amplified region and in 2 of 3 samples that were not assayed for CNAs (Figure 7C). Nup214-Abl1 fusion genes were not identified in pre-T LBL, AML, or Mcm2^cre/wt NHD13⁺ BCP-ALL (Figure 7, C and D). Nucleotide sequencing demonstrated a Nup214 exon 32 to Abl1 exon 2 fusion in 6 samples; 1 sample (sample 2697) had a Nup214 exon 31 to Abl1 exon 2 fusion (Table 3). Crkl, a direct target of the Abl1 kinase (31), was phosphorylated in splenocytes that expressed the Nup214-Abl1 fusion (Figure 7E), indicating that Nup214-Abl1 acts as an activated Abl1 kinase. Imatinib, an inhibitor of Abl1 kinase (33), inhibited the growth of splenocytes, which expressed a Nup214-Abl1 fusion (Figure 7F).

Figure 7

Recurrent Nup214-Abl1 fusion gene detected in BCP-ALL. (A) CNAs for the Abl1- and Nup214-containing region of chromosome 2. Color code as in Figure 2A. (B) Schematic for putative Nup214-Abl1 fusion gene produced via episome or tandem duplication. (C) RT-PCR detection of Nup214-Abl1 fusion mRNA using Nup214 exon 29 forward primer and Abl1 exon 2 reverse primer. Pre-T LBL samples are from mice with genotype Mcm2^cre/cre NHD13⁺ and Mcm2^cre/cre NHD13^–, BCP-ALL samples are from mice with genotype Mcm2^cre/cre NHD13^–, and AML samples are from mice with genotype Mcm2^cre/wt NHD13⁺. B, bone marrow; T, thymus; LN, lymph node; SP, spleen. (D) RT-PCR does not show Nup214-Abl1 fusion mRNA in BCP-ALL from Mcm2^cre/wt NHD13⁺ mice; 2842 is a positive control. (E) Phosphorylation of Crkl in the spleen of mice (WT, 2842, 2703, 2725, 2905, 2811, 2888, and 2799) with Nup214-Abl1 fusion; v-Abl transformed B cell line is a positive control. (F) Imatinib treatment inhibited the growth of 2725 splenocytes that expressed a Nup214-Abl1 fusion. Growth curves of 2725 splenocytes treated with imatinib for 2 days. Cell number was counted by trypan blue exclusion. Error bars represent standard deviation of 3 technical replicates (n = 3).

Table 3

Sequence of Nup214-Abl1 fusion junctions

We used WES to identify Tier 1 (coding) mutations in Mcm2^cre/cre NHD13⁺ mice and found no recurrent mutations in the 11 samples assayed (Supplemental Table 10). In contrast, all 3 Mcm2^cre/wt NHD13⁺ samples showed at least 1 Tier 1 mutation in genes known to be relevant for human BCP-ALL (Sh2b3 [ref. 34], Jak1, Trp53, or Flt3 [ref. 35]) (Supplemental Table 10 and Supplemental Figure 12).

Nucleotide-level resolution of deletion breakpoints. As seen in Figure 2, the genomic regions most susceptible to the 10–1000 kb deletions triggered by the Mcm2 hypomorph are not random but occur within a limited number of regions. To determine whether the breakpoint junctions are focal and follow any clear rules or patterns (such as VDJ recombination), we used LCC-Seq to capture and analyze the breakpoints of several selected regions (Pten, Notch1, Tcf3, Bcl7c, Ptpn1, and Nup214-Abl1) at a nucleotide level. In total, we identified 245 deletions in 91 samples (Supplemental Table 11). Identical deletions found in primary tumor and cell lines were considered a single independent event. Detailed analysis of the sequence at the deletion junction of 245 independent deleted regions revealed that 155 of the 245 junctions had 1–7 bp of microhomology (Supplemental Table 11). Seventy junctions contained nontemplated insertions at the breakpoint junctions; some of these were quite extensive, 1–54 nucleotides in length. Given the precedent set by programmed VDJ rearrangement, in which signal sequences can occur at a distance (up to 38 bp) from the breakpoint junction, we analyzed nucleotide sequences in a 200-bp window (100-bp 5′ and 100-bp 3′) flanking both the upstream and downstream breakpoints, in both the WGS data set (Supplemental Table 4) and the LCC-Seq data set (Supplemental Table 11). Comparison of 156 windows from the WGS to 7332 randomly selected 200-bp windows revealed an increased likelihood of mononucleotide repeats in the Mcm2 tumor breakpoints identified by WGS (P = 0.009456; Supplemental Figure 13A and Supplemental Tables 12 and 13). Moreover, comparison of 490 windows from the 245 independent breakpoints identified by LCC-Seq with 7530 randomly selected 200-bp windows from chromosomes 2, 7, 10, and 19 (corresponding to the LCC-Seq chromosomes) also showed an increased frequency of mononucleotide repeats in the Mcm2 tumor breakpoints (P = 0.009671; Supplemental Figure 13B and Supplemental Tables 14 and 15). These findings are consistent with the observation that upon replication stress, DNA breaks are frequently generated at mononucleotide repeats (36).

Replication stress has been linked to many forms of cancer (1, 3). It has been speculated that replicative stress leads to DNA DSBs, with subsequent repair by non-homologous end joining (NHEJ), resulting in indel mutations and structural variations, including large interstitial deletions. It is thought that these mutations, such as deletions of important tumor suppressor genes, ultimately cause malignancy. These frequent DNA DSBs caused by increased replicative stress lead to activation of Chk1; Chk1 inhibitors have recently entered the clinic (37) as an approach to target unique vulnerabilities in the cancer cell. In this study, we show that mutation of a single gene (Mcm2) leads to a deleter phenotype and that resultant combinations of deletions involving several critical genes act in concert and are selected in vivo as T or B lymphoid malignancies.

Hypomorphs of several members of the Mcm2–7 DNA replicative helicase complex show malignancy or stem cell defects (9, 38–40), which is thought to be associated with replicative stress; in some cases, these malignancies were linked to recurrent interstitial deletions (7, 9, 41). We reasoned that tumor suppressor genes that are important for transformation of MDS to AML could be identified by crossing the Mcm2 mice with a deleter phenotype to NHD13 mice. However, all Mcm2^cre/cre NHD13⁺ developed T and/or B cell malignancies, but not AML, within 5 months of age. We also predicted that Mcm2^cre/wt mice, which express approximately 30%–40% less Mcm2 protein than WT mice, might have a subtle predisposition to malignancy that can be revealed by a cooperating oncogenic event (such as the NHD13 transgene). However, although Mcm2^cre/wt NHD13⁺ developed AML, the incidence, age of onset, and immunophenotype was similar to Mcm2^wt/wt NHD13⁺ (Figure 1A), suggesting that a single copy of the Mcm2^cre allele did not predispose to AML. Furthermore, CNA showed that the Mcm2^cre/wt NHD13⁺ AML samples did not have recurrent deletions, reinforcing the idea that Mcm2^cre/wt mice are not predisposed to malignancy. Although it was previously shown that Mcm2^cre/wt mice are not prone to recurrent deletions (9), it remains possible that Mcm2^cre/wt mice are prone to more infrequent deletions that are not evident because they are polyclonal and not discernable in a polyclonal population of cells. However, the observation that clonal Mcm2^cre/wt NHD13⁺ AML samples did not show deletions argues strongly against that possibility. Given that all Mcm2^cre/cre NHD13⁺ mice died of pre-T LBL or BCP-ALL by 5 months of age, it remains possible that these mice did not develop AML because they died before AML had a chance to develop.

There were 4–32 deletions per tumor. Nucleotide sequence analysis revealed that there was an increased likelihood of mononucleotide repeats at or near the deletion breakpoints, consistent with the hypothesis that polymerase pausing or stalling at sites of mononucleotide repeats leads to DNA DSBs (42, 43); improper repair of 2 such breaks via NHEJ leads to the interstitial deletions. Given that studies of human pre-T LBL have suggested that there may be 2–4 “driver” mutations per tumor (24), we considered that many of the deletions were likely to be passenger deletions. Recurrent deletions may still be passenger deletions in this model, selected by virtue of susceptibility to deletion. Some regions had numerous homozygous deletions, centered over a single gene, such as Pten. Alternatively, there were regions where the individual deletions were spread over a larger 2–3 Mb region, with no clear common deleted region, and few homozygous deletions, such as the region encompassing Bcl7a. We suspect the former are more likely to be driver deletions and the latter more likely to be passenger deletions. Using these criteria, the leading candidates for driver deletions were Pten and Tcf3. These candidates are supported by experimental studies that show homozygous Pten and Tcf3 deletions strongly predispose mice to pre-T LBL (21, 44). Moreover, PTEN deletions/mutations are quite common in human pre-T LBL (16%, ref. 20). In addition, although TCF3 deletions are uncommon in human pre-T LBL, enforced expression of TAL1 by a large variety of genomic rearrangements is quite common (24); in this context, it is important to note that enforced expression of TAL1 leads to a functional inactivation of TCF3 (45). Of note was a significantly increased likelihood of Dnmt3a mutations, especially homozygous Dnmt3a mutations (3/11 vs. 0/18; P = 0.0096), in Mcm2^cre/cre NHD13^– mice compared with the Mcm2^cre/cre NHD13⁺ mice. A plausible explanation for this observation is that NUP98 fusion proteins have been shown to increase stem cell self-renewal (46), as has Dnmt3a deletion (47). Thus a homozygous Dnmt3a deletion that promotes stem cell self-renewal may be redundant in the context of an NHD13⁺ cell that also leads to increased self-renewal potential.

Notch1 deletions represent a special case in terms of thymocyte transformation. In contrast to Pten and Tcf3, in which homozygous deletion of the genes predisposes to pre-T LBL, Notch1 mutations are most commonly monoallelic point mutations of the HD or PEST domain in both humans and mice (26, 27). Mutations in the HD lead to loss of requirement for extracellular Notch1 ligand binding, and PEST domain mutations result in decreased degradation of the transcriptionally active ICN, resulting in sustained activation of Notch1 target genes (26). A recurrent interstitial deletion of Notch1 exons 1–2 leads to translation initiation at an internal methionine residue (M1727), production of a protein that lacks the extracellular domain, and ligand-independent activation of Notch1 (28). In this study, we show recurrent Notch1 genomic deletions removed the Notch1 extracellular domain and retained the transcriptionally active ICN.

Consistent with the notion that driver mutations were produced by recurrent interstitial deletions in this model, WES studies showed relatively few SNVs or indels. The only recurrent Tier 1 mutations identified in 29 samples were 4 Notch1, 1 Ezh2, and 2 Tp53 mutations. These results stand in contrast to our prior WES studies of murine leukemia and lymphoma on a WT Mcm2 background, in which 60%–100% of malignancies had acquired Tier 1 mutations in genes known to be mutated in hematological malignancy (48, 49).

Serial analysis of the Mcm2^cre/cre pre-T LBL cell lines suggests that the Notch1 mutations are relatively late events, consistent with sequencing studies of human pre-T LBL, in which almost half the Notch1 mutations identified were thought to be subclonal (24). An overall scheme for cooperative lesions that result in transformation in this model is outlined in Supplemental Figure 14A. Increased stem cell self-renewal is conferred by either the NHD13 transgene or Dnmt3a inactivation. Tcf3 deletion results in a block to thymocyte differentiation, and Pten deletion leads to hyperproliferation. Finally, Notch1 mutations lead to ligand-independent growth; the cells no longer require Notch1 ligands supplied by thymic epithelial cells and are able to metastasize as well as expand in vitro.

CNA analysis of pre-T LBL cell lines indicated that the overall frequency of new deletions seemed to decrease with time (Table 1). The decreased frequency of new deletions could reflect accommodation to the Mcm2 protein deficiency; alternatively, it is conceivable that there are relatively few exquisitely sensitive regions susceptible to deletion in Mcm2 hypomorph pre-T LBL, and these are deleted relatively early. The ongoing Mcm2 deficiency suggests that these cell lines could be useful for the study of Mcm2 biology or evolutionary pressures in cancer. For instance, CNA analysis of cell lines selected for resistance to clinically relevant chemotherapy agents (for instance, vinca alkaloids or alkylating agents), could be used to identify genes and pathways important for this acquired resistance.

Approximately 26% of Mcm2^cre/cre NHD13⁺ developed concurrent BCP-ALL and pre-T LBL. CNA analysis of BCP-ALL was in general simpler than the pre-T LBL, with fewer deletions per sample. Pax5 and Ptpn1 were recurrently deleted, and these deletions were invariably accompanied by Nup214-Abl1 fusion similar to that seen in a subset of human pre-T LBL and BCP-ALL (31, 32). The Nup214-Abl1 fusions were functional, as shown by Crkl phosphorylation and sensitivity to imatinib. WES showed no recurrent Tier 1 SNVs, consistent with the notion that driver mutations in Mcm2 hypomorph mice were primarily CNAs. Taken together, the complementary pathways involved in this model match those predicted for human BCP-ALL (50). In this context, the collaborative pathways and genes are: (a) increased stem cell self-renewal conferred by the NHD13⁺ transgene, (b) impaired B cell differentiation caused by Pax5 deletion, and (c) hyperproliferation conferred by a Nup214-Abl1 fusion gene (Supplemental Figure 14B). Although the role of Ptpn1 deletion is unclear, it is interesting to note Ptpn1 deletion accelerates lymphoma onset in Trp53-deficient mice (51), conceivably by enforcing active kinase signaling (52), and deletions of the closely related PTPN2 are common in pre-T LBL patients with NUP214-ABL1 fusion. In both cases, the PTPN deletion is thought to be oncogenic through enforcing hyperactive kinase signaling (52, 53).

Taken together, these findings demonstrate that replicative stress can result in cancer through the generation of chromosomal rearrangements, most commonly interstitial deletions of about 50–1000 kb. The genetic lesions selected in vivo (Dnmt3a, Pten, Tcf3, and Notch1 for pre-T LBL and Pax5 and Abl1 for BCP-ALL) are frequent events in the corresponding human lymphoid leukemias, reinforcing the idea that collaborative pathways leading to lymphoid leukemias are similar between mice and humans. Combinations of CNAs involving tumor suppressor genes or oncogenes are selected because of fitness advantage and provide an in vivo mammalian model for evolution and selection within a time frame of months.

Mice and genotyping

Mcm2^cre/cre and NHD13-transgenic mice were generated as previously reported; congenic Mcm2 on a C57BL/6 background mice were generated as described previously (6, 9, 14). The genotyping of NHD13 and Mcm2 mice was performed as previously described (6, 9, 14) with primers listed in Supplemental Table 16. CBCs were performed on peripheral blood using a HEMAVET Multispecies Hematology Analyzer (CDC Technologies). Diagnosis of hematological malignancy was based on previously published consensus guidelines (54).

Flow cytometry, IHC, and immunoblots

Flow cytometry was performed as described previously (29) with the following antibodies: from eBioscience: Mac1 (CD11b)-PE (catalog 12-0112-82), CD4-PE (catalog 12-0043-82), Gr1-FITC (catalog 11-5931-85), Ter119-FITC (catalog 11-5921-85), c-Kit (CD117)-FITC (catalog 11-1171-82), CD43-FITC (catalog 11-0431-82), CD44-FITC (catalog 11-0441-82), CD25-PE (catalog 12-0251-82), CD4-APC (catalog 17-0042-82), and CD8-APC-780 (catalog 47-0081-82); and from BD Biosciences: CD8-FITC (catalog 553031), CD71-PE (catalog 553267), Sca-1-PE (catalog 553108), B220 (CD45R)-FITC (catalog 553088), and CD19-PE (catalog 553786). IHC and immunoblotting were performed as previously described (29). Formalin-fixed, paraffin-embedded tissue sections were stained with H&E, myeloperoxidase (A0398; Dako), CD3 (MCA1477; Bio-Rad), and B220 (553086; BD Biosciences). Stained sections were scanned and imaged as previously reported (29). For immunoblots, proteins were separated by 7.5% SDS-PAGE (Bio-Rad Laboratories) and transferred to nitrocellulose membranes (Thermo Fisher Scientific). Primary antibodies used were anti-Mcm2 (610701, Trans-duction Laboratories), anti-PTEN (D4.3) XP (9188, Cell Signaling Technology), anti–phospho-Crkl (Tyr207) (181, Cell Signaling Technology), anti–cleaved Notch1 (Val1744) (D3B8) (4147, Cell Signaling Technology), anti–β-actin (A5316, MilliporeSigma), and anti–α-tubulin (2125S, Cell Signaling Technology). After application of appropriate secondary antibodies conjugated to HRP (goat anti-mouse, 31430; and goat anti-rabbit, 31460, both from Thermo Fisher Scientific), signals were visualized using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific) and Amersham Hyperfilm ECL (GE Healthcare Ltd.).

Cell culture and cell lines

Pre-T LBL (2875, 2854, 2880, 2730, 2773, 2696, 2795, 2869, 2883, 2641, 2739, and 2973 TCL) cell lines were established from single-cell suspensions prepared from thymus (1 × 10⁶ cells) of sick mice and maintained in IMDM supplemented with 15% FBS, 100 mmol/L l-glutamine, and 100 μg/mL penicillin/streptomycin (Invitrogen, Thermo Fisher Scientific) without supplemental cytokines. Splenocytes from mouse 2725 that expressed a Nup214-Abl1 fusion were maintained in RPMI 1640 medium (Gibco, Thermo Fisher Scientific) supplemented with 10% heat-inactivated FBS, 1 mM l-glutamine, 100 U/mL streptomycin, 100 μg/mL penicillin, 50 μM 2-ME (all from Gibco, Thermo Fisher Scientific), and 10 ng/mL IL-7 (217-17; Peprotech).

Assessment of Compound E and imatinib mesylate treatment

Compound E (ALX-270-415, Enzo Life Science) was dissolved in DMSO and evaluated at a final concentration of 1 μM. Pre-T LBL cell lines (5 × 10⁴ /mL) were seeded in IMDM supplemented with 15% FBS, 100 mmol/L l-glutamine, and 100 μg/mL penicillin/streptomycin. The cells were treated with 1 μM Compound E or vehicle (DMSO) only for 4 days, and cell number was determined by trypan blue exclusion daily (TC20 Automated Cell Counter, Bio-Rad Laboratories, Inc). Imatinib mesylate (S1026, Selleckchem) was dissolved in DMSO and evaluated at 1, 5, and 10 μM. Splenocytes from mouse 2725 (5 × 10⁴ /mL) were seeded in RPMI 1640 medium supplemented with 10% heat-inactivated FBS, 1 mM l-glutamine, 100 U/mL streptomycin, 100 μg/mL penicillin, 50 μM 2-ME, and 10 ng/mL IL-7. The cells were treated with imatinib mesylate or vehicle only for 4 days, and cell number was determined by trypan blue exclusion daily.

PCR and Sanger sequencing

Genomic DNA was extracted with the DNeasy Blood & Tissue (QIAGEN) kit according to the manufacturer’s recommended protocol. PCR was performed using HiFi Taq polymerase mix (10790-020; Invitrogen, Thermo Fisher Scientific) and primers (Invitrogen, Thermo Fisher Scientific) as listed in Supplemental Table 16. Clonal Igh and Tcrb DJ or VDJ segments were identified using previously described PCR-based assays (29) with primers listed in Supplemental Table 16. RNA was extracted using TRIzol (Invitrogen, Thermo Fisher Scientific) and the manufacturer’s recommended protocol. cDNA was synthesized by reverse transcriptase using 1 μg RNA with SuperScript III enzyme and reagents (Invitrogen, Thermo Fisher Scientific). cDNA splice sites for Notch1, Ikzf1, and Nup214-Abl1 were identified by PCR amplification of cDNA and confirmed by Sanger sequencing. Selected mutations identified by WES or LCC-Seq were PCR amplified, purified, and confirmed by Sanger sequencing. Real-time quantitative PCR was performed with a Ptpn1 TaqMan primer probe (Mm00448427_m1, Thermo Fisher Scientific) and Hes1 TaqMan primer probe (Mm01342805_m1, Thermo Fisher Scientific) sets and ABI Fast Universal PCR Master Mix on the ABI Fast7500 system (Applied Biosystems, Thermo Fisher Scientific). Samples were normalized to endogenous 18S rRNA with TaqMan Ribosomal RNA Control Reagents. Sanger sequencing was performed at the National Cancer Institute Sequencing Minicore facility.

Sparse WGS for CNAs

Preparation of genomic libraries for sparse WGS was performed as previously described (19). In brief, 1 μg of genomic DNA was sonicated using the Covaris instrument followed by end repair and A-tail addition. TruSeq dual index adapters were subsequently ligated to DNA molecules, with ligated products enriched via PCR amplification. Indexed libraries were pooled and sequenced in multiplex fashion while targeting about 4 million reads per sample. Data analysis was performed as described previously (55) with the exception that new higher resolution bin boundaries were constructed to facilitate the analysis. Specifically, mouse genome build mm9 was divided into 120,000 bins, while accounting for unique mappability. This allowed analysis of the CNAs at a segment resolution of about 125 kb (5 consecutive bins for segmentation) and breakpoint resolution of about 25 kb (average width of each bin).

WES, WGS, and LCC-Seq and sequencing analysis

Library preparation. Sequencing libraries were prepared through tagmentation using the Nextera DNA Library Kit (Illumina, Inc.) according to the manufacturer’s instructions with the following modifications. PCR primers in the kit were replaced with i5 primer (5′-AATGATACGGCGACCACCGAGATCTACACNNNNNNNNTCGTCGGCAGCGTC-3′) and i7 primer (5′-CAAGCAGAAGACGGCATACGAGATNNNNNNNNGTCTCGTGGGCTCGG-3′), at a final concentration of 50 nM, where the Ns represent standard Nextera index sequences. Exome and Targeted Genomic CaptureExome capture was carried out using the SureSelect XT Mouse All Exon, 49.6 Mb Kit (Agilent Technologies, Inc.) following the manufacturer’s protocol. BAC clones tiling the genomic regions of interest (Supplemental Table 17) were obtained (BACPAC Resources Center), and DNA was prepared using the QIAGEN Large-Construct Kit. BACs were pooled and capture bait prepared using biotinylated random hexamers and Klenow enzyme in a random primed extension reaction. Bait coupled to streptavidin magnetic beads was hybridized to library pools for 24 hours followed by stringency washes and PCR using primers homologous to the outermost adapter sequences of the sequencing libraries (5′-AATGATACGGCGACCACC-3′ and 5′-CAAGCAGAAGACGGCATA-3′). Sequencing Captured libraries were quantified by quantitative PCR (KAPA Biosystems) and sequenced on a HiSeq 2500 System (Illumina, Inc.).

Sequencing analysis. Data processing and variant calling procedure followed the Best Practices workflow recommended by the Broad Institute (56, 57). Briefly, the raw sequencing reads were mapped to mouse genome build 10 (mm10) by the Burrows-Wheeler Aligner (58) followed by local realignment using the GATK suite (59) from the Broad Institute, and duplicated reads were marked by the Picard tools (http://broadinstitute.github.io/picard).

Somatic variant calling was performed by comparison of tumor to WT samples using the MuTect2 (60) somatic variant caller in the GATK suite. SnpEff (61) variant annotation and effect prediction tool and dbSNP 137 (National Center for Biotechnology Information) were used to annotate and predict effects of the variants. LUMPY (62) was used for structural variant discovery in the WGS and LCC-Seq.

These were the filtering criteria for the somatic variants in the WES data: variants were first filtered with the GATK recommended filtering criteria (https://software.broadinstitute.org/gatk/documentation/article.php?id=3225 under “Filtering Recommendations”) and then filtered with the following additional filters: (a) minimum fraction of altered reads in a tumor is 0.3; (b) minimum number of altered reads in a tumor is 2; (c) minimum log_fisher is 0.2; (d) impact effect is “High” or “Moderate” for Tier 1 indels and missense mutations; and (e) exclude SNPs reported in dbSNP build 137 or previously identified as germline variants in the NIH C57BL/6 colony.

These were the filtering criteria for the structure variants in the WGS data: (a) ALT = DEL; (b) SVLEN ≤ 1000; (c) normal AO = 0; and (d) manual check if deletions were present in the BAM files.

These were the filtering criteria for the structural variants (SVs) in the LCC-Seq data: (a) only keep SVs overlapping the bait regions; (b) maximum number of altered reads in the WT sample is 4; (c) minimum fraction of altered reads in a tumor is 0.1; and (d) manual check if deletions were present in the BAM files. Primary next-generation sequencing data are available at the National Center for Biotechnology Information’s Sequence Read Archive, accession numbers PRJNA565491, PRJNA565494, PRJNA589636, and PRJNA565492.

Analysis of repeat sequences associated with tumor breakpoints

The breakpoints mapped in recurrent deletions using WGS and LCC-Seq are extended 100 bp on either side — 100 bp into the deletion and 100 bp into the region intact in the tumor. Each deletion is associated with 2 breakpoints. The DNA sequence of this 200-bp window was obtained using bedtools getfasta command (63). Tandem Repeat Finder was used to locate the mononucleotide repeat sequences around each breakpoint (64).

For each data set (WGS and LCC-Seq), a random data set of a 200-bp window was generated using bedtools shuffle-chrom command to maintain a similar chromosome distribution of the random data set as the tumor deletion data set (63). The DNA sequence and repeat sequence were mapped as explained earlier.

For the tumor deletion and random data set, the enrichment of mononucleotide repeats was evaluated using Fisher’s Exact Test for Counts Data using R. The tumor deletion data set showed a significant enrichment of mononucleotide repeats with respect to the random data set.

Statistics

Data are displayed as mean ± standard deviation. Significance values were calculated using Student’s t test (unpaired, 1-tailed distribution) in Microsoft Excel. The number of independent experiments, statistical tests, and P values are indicated in figure legends. Survival was analyzed using GraphPad Prism 7.01 software and log-rank (Mantel-Cox) test. P < 0.05 was considered significant.

Study approval

All studies involving mice were approved by the National Cancer Institute (Bethesda) Intramural Animal Care and Use Committee and were performed according to protocols approved by the National Cancer Institute (Bethesda) Intramural Animal Care and Use Committee.