Use of chromosome engineering to model a segmental deletion of chromosome band 7q22 found in myeloid malignancies

Loss of chromosome 7 and deletion of a segment of the long arm (monosomy 7 and del(7q)) are recurring cytogenetic abnormalities in de novo and therapy-induced myeloid malignancies that are associated with advanced age, antecedent myelodysplastic syndrome (MDS), and resistance to current treatments.¹  Based on precedents in other cancers, it is likely that loss of one or more 7q tumor suppressor genes (TSGs) contributes to leukemogenesis. To facilitate the identification of candidate myeloid TSGs, Le Beau et al delineated 2 commonly deleted segments (CDSs) in patients with myeloid disorders characterized by a del(7q), a proximal interval within band q22 that accounts for most cases, and a second CDS in bands q32-34.²  Using an ordered set of yeast artificial chromosome clones as probes, these investigators then performed fluorescence in situ hybridization (FISH) experiments to further characterize leukemias with deletion breakpoints within 7q22 and implicated an approximately 2.5-Mb CDS as harboring a myeloid TSG. We and others have extensively characterized this CDS, identified and cloned multiple genes from the interval, analyzed leukemia samples for mutations in these candidate TSGs, and performed Taqman real-time quantitative reverse-transcriptase polymerase chain reaction (RT-PCR) assays to measure expression levels in normal and leukemic human bone marrows.^3-6  These studies did not uncover biallelic inactivation or epigenetic silencing of any candidate TSGs located within this CDS.^3-5  Thus, it was hypothesized that inactivation of a single allele (haploinsufficiency) of one or more TSGs located within the 2.5-Mb CDS might contribute to leukemogenesis.^4,5 

Recent technical advances such as high-throughput sequencing platforms and RNA interference (RNAi) provide powerful new tools for approaching the challenging problem of identifying haploinsufficient TSGs. Indeed, elegant RNAi-based studies by Ebert et al provided strong support for RPS14 as a haploinsufficient disease gene in the 5q− syndrome subtype of MDS.⁷  The use of “chromosome engineering” to delete large DNA segments in the mouse is another potent technique that can be harnessed to interrogate a region suspected of harboring a haploinsufficient TSG in vivo.^8,9  The essence of this strategy is to embed 2 complementary, but nonfunctional, fragments of a hypoxanthine phosphoribosyl transferase (HPRT) minigene cassette within loxP-containing targeting vectors that flank the region of interest. Cre recombination between the chromosomal loxP sites then joins the 2 HPRT fragments together and creates a functional HPRT minigene that can be used as a selectable genetic marker to identify embryonic stem (ES) clones with the desired recombination event. Advantages of chromosome engineering include the following: (1) it provides a viable, function-based alternative to traditional positional cloning strategies; (2) it represents an unbiased approach to identify segments that contain more than one TSG whose combined loss is required for the desired phenotype; and (3) it generates murine models of human disease that can be used to test new therapies and to study the biology of any embedded TSGs. Work that identified CHD5 as a TSG within 1p36.3 illustrates the potential of this strategy for cloning human cancer genes.¹⁰  Interestingly, many cancers with 1p36 loss show large deletions, and a recent study implicating K1F1Bβ as a second 1p36 TSG¹¹  raises the possibility that haploinsufficiency of several tumor suppressor genes contributes to some human cancers.

We developed mice harboring loxP sites that flank a 2-Mb DNA segment of mouse chromosome band 5A3 that is syntenic to the human 7q22 CDS found in myeloid malignancies and bred this strain with Mx1-Cre mice to induce the desired deletion in the hematopoietic compartment. Here we show that heterozygous Mx1-Cre, 5A3^fl/+ and homozygous Mx1-Cre, 5A3^fl/fl compound mutant mice excise the 5A3 interval in a small percentage of bone marrow cells that includes some hematopoietic stem cells (HSCs). These mice display normal steady-state hematopoiesis and do not spontaneously develop MDS or acute myeloid leukemia (AML). Intercross experiments showed that coinheritance of the excised 5A3^fl/+ segment does not cooperate with either oncogenic Kras expression or Nf1 inactivation in leukemogenesis. Similarly, induction of the Mx1-Cre 5A3^fl/+ allele did not affect the latency and penetrance of leukemia in mice that were injected with the MOL4070LTR retrovirus.¹²  In addition, loss of the 5A3 interval did not alter the response of MOL4070LTR-induced AMLs to antileukemia drugs. Together, these comprehensive studies suggest that the 7q22/5A3 interval is unlikely to harbor a haploinsufficient myeloid TSG. Our data demonstrate the feasibility of deleting large DNA segments in the HSC compartment in vivo and as a late event in leukemic cell populations, which may prove useful for creating strains of mice to investigate other chromosomal deletions found in human hematologic malignancies.

A 6.4-kb fragment containing part of introns 14 to 15, exon 15, and a portion of the 3′ untranslated region of the Srpk2 gene was cloned in the NotI and SpeI sites in the pG12-WT-NP-IRES-GFP vector (kind gift from Dr Allan Bradley, Wellcome Trust Sanger Institute, Cambridge, United Kingdom) and linearized with Eco47III. E14 ES cells (129P2 strain) were electroporated with 25 μg of linearized Srpk2 vector, and the clones were selected in puromycin (2 μg/mL) for 8 to 10 days. Correct gene targeting event was assessed by digesting ES cell DNA with HindIII and hybridizing Southern blots with a 1.3-kb probe containing sequences downstream of the genomic DNA used to construct the targeting vector. Three positive clones were expanded for a second round of targeting with a vector containing a 7.8-kb fragment of Lrrc17 genomic DNA that includes parts of intron 1, intron 2, and exon 2 in the pL13-NP backbone (kind gift from Dr Allan Bradley). Singly targeted ES cells were electroporated with 25 μg of linearized vector, and clones were selected in G418 (250 μg/mL) for 8 to 10 days. The desired gene targeting event was assessed by digesting the DNA with BamHI, followed by Southern blotting and hybridization with a 0.7-kb probe containing sequences downstream of the genomic DNA used to construct the targeting vector. Southern analysis was performed using Church-Gilbert solution at 60°C, and a final wash in 0.2× saline-sodium citrate solution, 0.1% sodium dodecyl sulfate.

Leukocyte DNA was isolated using the GFX genomic blood purification kit (Amersham), and tissue DNA was isolated as described.¹³  Genotyping to assess inheritance of the Mx1-Cre transgene, presence of the correct Lrrc17 integration flanking the targeted 5A3 interval, and somatic rearrangement to create the desired deletion was performed by PCR. The primers and assay conditions are given in supplemental Methods (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Total RNA was isolated from mouse bone marrow and frozen leukemia cells using the QIAGEN RNeasy Mini kit and treated with TURBO DNA-free kit (Ambion) according to the manufacturer's protocol. Total RNA (350 ng) was reverse transcribed using the High Capacity RNA-to-cDNA Master Mix according to the manufacturer's protocol (Applied Biosystems). The relative quantification of gene expression was performed by quantitative real-time PCR using the TaqMan Gene Expression Assays (Applied Biosystems). The primers, probes, and assay conditions are given in supplemental Methods.

Polyinosine-polycytidine (pIpC; Sigma or Amersham) was resuspended in phosphate-buffered saline at 10 mg/mL or 1 mg/mL. Mice received either a single dose of 100 μL of 10 mg/mL pIpC at 3 days or a single dose of 250 μL of 1 mg/mL pIpC at 3 weeks. For a high-dose regimen, mice received a dose of 250 μL of 1 mg/mL pIpC every other day 5 times. The breeding strategies for mouse strains are described in supplemental Methods. All mice involved in the research were housed in a specific pathogen–free facility at the University of California San Francisco, and all animal experiments were conducted under protocols approved by the University of California, San Francisco Institutional Animal Care and Use Committee.

Hematopoietic cell collection and pathologic and flow cytometric analysis were carried out as previously described.¹⁴ 

A MOL4070LTR viral stock was generated as described elsewhere.¹² 5A3^+/fl and Mx1-Cre mice were bred to generate Mx1-Cre, 5A3^+/fl pups on a F1 strain background, which were inoculated with 100 μL of MOL4070LTR virus intraperitoneally and monitored for signs of disease. MOL4070LTR integration sites were identified by performing PCR amplification (supplemental Methods) followed by either direct 454 sequencing or by cloning the products using a Zero Blunt cloning kit (Invitrogen), transforming TOP10 competent cells, and sequencing DNA minipreps isolated from bacterial cultures.

For serial transplantation of recombined bone marrow, adult recipient mice were lethally irradiated with 9.5 Gy from a cesium source. AML transplant recipients were irradiated with 4.5 Gy. After irradiation, the cells were injected into the retro-orbital sinus of recipient mice.

Mice that were injected with AMLs were treated with intraperitoneal cytosine arabinoside (Ara-C; 100 mg/kg once daily for 5 days) and doxorubicin (3 mg/kg once daily for 3 days). The drugs were administered together.

The human 7q22 CDS defined by Le Beau et al²  is flanked by the SRPK2 and FBXL13 genes, includes 14 genes, and does not contain any microRNA or noncoding RNAs (Figure 1A). This interval corresponds to a single region of orthologous synteny on mouse chromosome band 5A3 (Figure 1A). We isolated λ phage clones containing Srpk2 and FbxL13 and used these to construct targeting vectors (Figure 1B). Thirteen doubly targeted ES cell clones that carry loxP insertions in FbxL13 and Srpk2 were generated through sequential rounds of gene targeting (supplemental Figure 1A-B), and these clones were electroporated with a vector encoding Cre recombinase and grown in media containing neomycin or puromycin. Four clones were sensitive to both drugs, indicating that targeting had occurred in the correct relative orientation in cis (ie, on the same chromosome 5 homolog). The desired Cre-mediated recombination event was confirmed by Southern blot analysis (Figure 1C and supplemental Figure 1C). We injected doubly targeted (unrecombined) ES clones into blastocysts to generate 5A3^+/fl mice. Germline transmission of the latent mutant allele was screened by coat color, and confirmed by Southern blot and PCR-based genotyping (Figure 1D). Heterozygous 5A3^+/fl and homozygous 5A3^fl/fl mice were born at the expected ratio, and were grossly normal and fertile. To induce excision of the 5A3 interval, we transduced bone marrow cells and mouse embryonic fibroblasts with Cre-expressing retroviral vectors in vitro. However, we observed Cre-mediated cellular toxicity, leading to markedly reduced cell proliferation and changes in cellular morphology.^15,16  Attempts to circumvent Cre-mediated cellular toxicity by transduction using adenoviral or self-excising retroviral vectors encoding Cre recombinase never achieved detectable levels of recombination (data not shown).

The interferon-inducible Mx1-Cre transgene is broadly expressed in hematopoietic cells and has been used extensively to develop mouse models of hematologic malignancies.^14,17-20  We mated 5A3^fl/+ and Mx1-Cre mice, and administered a single dose of pIpC at 3 days of life to induce Cre recombinase expression. Compound Mx1-Cre, 5A3^fl/+ and Mx1-Cre, 5A3^fl/fl mutant mice appeared normal. PCR analysis of genomic DNA isolated from blood leukocytes confirmed the desired recombination in mice that were tested between 8 weeks and 2 years of age (Figure 2A). Cohorts of Mx1-Cre, 5A3^fl/+ (n = 35), Mx1-Cre, 5A3^fl/fl (n = 25), Mx1-Cre–negative 5A3^fl/+ (n = 20), 5A3^fl/fl (n = 22), and wild-type (wt; n = 20) littermates were aged for 2 years to characterize any disease phenotypes. These groups of mice showed no significant differences in overall survival, causes of death, or blood counts (Figure 2B and data not shown). Southern blotting of bone marrow DNA was performed to assess the efficiency of Mx1-Cre–mediated recombination. Although the level of recombination was low in most Mx1-Cre, 5A3^fl/+ and Mx1-Cre, 5A3^fl/fl mice (center and right lanes in Figure 2A), 3 (16%) of 19 Mx1-Cre, 5A3^fl/+ mice showed high levels of recombined cells at death (see left lane in Figure 2A). Each of these Mx1-Cre, 5A3^fl/+ mice was visibly sick, and pathologic analysis demonstrated centroblastic-type diffuse large B-cell lymphoma, histiocytic sarcoma, and possible infection in one animal each at 586 days, 755 days, and 568 days of age, respectively.

To determine whether increasing the percentage of recombined cells might modify the phenotype, we administered additional doses of pIpC (250 μg every other day for 5 doses) to 8-week-old Mx1-Cre, 5A3^fl/+ and Mx1-Cre, 5A3^fl/fl mice. Southern blotting demonstrated higher levels of recombination in bone marrow and spleen 4 days after the last pIpC dose in comparison with mice that received a single dose of pIpC after birth (Figure 2C). We isolated bone marrow RNA 1 day after the final dose of pIpC and performed RT-PCR to assess the expression of 5 genes located within the 5A3 region. Although this analysis was complicated by the effects of sequential injection of pIpC on gene expression in hematopoietic cells and by incomplete excision of the 5A3 interval, the overall expression of all 5 genes after pIpC treatment was lower in Mx1-Cre, 5A3^fl/+ mice (Figure 2D). However, this difference achieved statistical significance only for Mll5 (Figure 2D). We aged 5 Mx1-Cre, 5A3^fl/+ and 3 Mx1-Cre, 5A3^fl/fl mice that received high-dose pIpC for 21 to 65 weeks. All remained free of disease and had normal blood counts at death.

The observation that recombined cells persist in the blood of Mx1-Cre, 5A3^fl/+ and Mx1-Cre, 5A3^fl/fl mice 2 years after a single injection of pIpC suggests that the 5A3 interval is excised in HSCs. Consistent with this idea, PCR analysis of DNA extracted from c-kit⁺ Lin⁻ Sca-1⁺ (KLS) bone marrow cells from Mx1-Cre, 5A3^fl/fl mice that were treated with high-dose pIpC demonstrated recombination in this population (data not shown). To directly assess the long-term repopulating activity of recombined cells, we transplanted bone marrow from Mx1-Cre, 5A3^fl/fl mice into lethally irradiated wt recipients 4 days after high-dose pIpC treatment. These cells efficiently established donor-derived hematopoiesis that persisted for more than 4 months, which was characterized by the presence of 5A3-excised hematopoietic cells as detected by Southern blot analysis of the bone marrow (Figure 2E). Injecting bone marrow cells from these mice into secondary recipients resulted in donor-derived hematopoiesis with the 5A3 deletion again detected by Southern blotting. The intensity of the recombined fragment in comparison with the unrecombined fragment, which was a reflection of the proportion of 5A3-excised cells in the bone marrow, was relatively stable over serial transplantation (Figure 2E). Together, these functional data demonstrating that 5A3-excised cells possess long-term repopulating potential and contribute to hematopoiesis in secondary recipients infer that some HSCs contain this deletion.

Activating a latent Kras^G12D oncogene in Mx1-Cre, LSL-Kras^G12D mice induces an aggressive myeloproliferative disorder (MPD) that models juvenile and chronic myelomonocytic leukemias.^18,21  This MPD is fully penetrant and results in death 70 to 90 days after a single injection of pIpC. Somatic inactivation of Nf1 in Mx1Cre, Nf1^fl/fl mice causes a less aggressive monocytic MPD than Mx1-Cre, LSL-Kras^G12D mice.¹⁴  Based on data showing that monosomy 7/del(7q) coexists with NRAS, KRAS, and NF1 mutations in patients with de novo and therapy-induced myeloid malignancies,^22-26  we generated Mx1-Cre, LSL-Kras^G12D, 5A3^fl/+ and Mx1-Cre, Nf1^fl/fl, 5A3^fl/+ compound mutant mice to determine whether the 5A3 deletion alters the phenotype or latency of MPD. As shown in Figure 3A, all Mx1-Cre, LSL-Kras^G12D mice developed MPD with anemia as previously described after receiving a single dose of pIpC at weaning.^18,21  The survival of Mx1-Cre, LSL-Kras^G12D, 5A3^fl/+ mice was similar to Mx1-Cre, LSL-Kras^G12D littermates (Figure 3B). Southern blot analysis revealed the rearranged 5A3 allele in bone marrow cells from compound mutant mice with MPD (Figure 3C). To determine whether the 5A3 interval was deleted in the cells that have activated the latent Kras^G12D allele, we assayed colony-forming unit granulocyte-macrophage (CFU-GM) growth in methylcellulose cultures in the absence of cytokines. Bone marrow cells from Mx1-Cre, LSL-Kras^G12D mice demonstrate growth factor–independent CFU-GM growth only after Kras^G12D expression is induced by Cre-mediated excision of the inhibitory LSL cassette^18,21  (and data not shown). As expected, Mx1-Cre, LSL-Kras^G12D, 5A3^fl/+ progenitors also formed “spontaneous” CFU-GM colonies in the absence of cytokines, and PCR analysis of DNA from individual colonies revealed recombination of both the 5A3 interval and LSL cassette in approximately 30% of them (Figure 3D).

Because the aggressive MPD in Mx1-Cre, LSL-Kras^G12D mice could have obscured modest effects of the 5A3 deletion, we generated Mx1Cre, Nf1^fl/fl 5A3^fl/+ mice. As shown in Figure 3E, the survival of Mx1-Cre, Nf1^fl/fl, 5A3^fl/+ mice and of control Mx1-Cre, Nf1^fl/fl littermates that received a single dose of pIpC at 3 days was similar. In addition, Mx1-Cre, Nf1^fl/fl, 5A3^fl/+ and Mx1-Cre, Nf1^fl/fl mice developed a similar MPD (data not shown).¹⁴  Importantly, transformation to leukemia did not occur, even given the more indolent clinical course. We confirmed that cells with 5A3 deletion were present in the blood leukocytes of moribund Mx1-Cre, 5A3^fl/+, Nf1^fl/fl mice (Figure 3F). Together, our studies of Kras and Nf1 mutant mice infer that 5A3 deletion neither confers a selective growth advantage nor alters the phenotype in mice that develop MPD due to hyperactive Ras signaling.

Cytogenetic, molecular, and clinical studies of myeloid malignancies with monosomy 7/del(7q) strongly support a multistep model of leukemogenesis.^27,28  We used the MOL4070LTR retrovirus¹²  to perform a forward genetic screen to assess the effects of the 5A3 deletion on the latency, penetrance, and phenotype of leukemia. Mx1-Cre, 5A3^fl/+ and littermate control mice were injected with MOL4070LTR at 3 days of age, and this was followed by a single dose of pIpC at weaning. This design allowed us to investigate whether secondary loss of the 5A3 region confers a selective growth advantage and to screen any resulting leukemias for “second hit” integrations within the nontargeted 5A3 homolog. Mice were killed when they became moribund and hematopoietic tissues were collected for pathologic examination and molecular analysis. Although MOL4070LTR efficiently induced acute leukemia, the survival of Mx1-Cre, 5A3^fl/+ mice was similar to that of control littermates (Figure 4A). Southern blot analysis of bone marrow from Mx1-Cre, 5A3^fl/+ mice with leukemia revealed complete recombination of the 5A3 interval in 7 (23%) of 30 leukemias, partial recombination in 4 (13%), and no recombination in 19 (63%; Figure 4B). Serial transplantation confirmed that 5A3-deleted leukemias could efficiently induce disease in sublethally irradiated secondary recipients (Figure 4B). Within the Mx1-Cre, 5A3^fl/+ cohort, mice that developed leukemia with the loss of the 5A3 interval showed identical survival to mice in which the 5A3 deletion was not detected in the leukemic clone (data not shown). The 7 leukemias with complete loss of the 5A3 interval included 3 cases of acute myelomonocytic leukemia, 2 cases of AML without maturation, 1 case of MPD-like myeloid leukemia, and 1 case of T-cell acute lymphoblastic leukemia. This spectrum of disease is not significantly different from control littermates (data not shown).

We cloned retroviral integration sites from the 7 mice that developed 5A3-deleted leukemias (supplemental Table 1). Importantly, none of these insertions mapped within the nontargeted 5A3 homolog. Integrations in or near the Sox4 gene, which is a frequent common integration site in murine retroviral insertional mutagenesis screens,²⁹  was the only common integration site detected in 5A3-deleted leukemias and occurred in 4 independent 5A3-deleted AMLs (supplemental Table 2). By contrast, we did not detect Sox4 integrations in any of the 16 unrecombined 5A3 leukemias (data not shown). However, quantitative RT-PCR analysis of Sox4 expression did not reveal consistent changes in the 5A3-deleted AMLs (supplemental Table 2). In summary, retroviral insertional mutagenesis demonstrates that the 5A3 deletion occurs in cells that can initiate acute leukemia in vivo. However, loss of the 5A3 interval had no significant effect on either the time to death or the phenotype of MOL4070LTR-initiated acute leukemias.

We also performed quantitative RT-PCR to compare the expression levels of 5 genes within the targeted 5A3 segment in AMLs that had either retained or lost the interval (Figure 4C). These studies revealed marked variability in gene expression in individual AMLs of both 5A3 genotypes, which could be due to the effects of cooperating mutations or the heterogeneous phenotypes of these leukemias.

Clinical and epidemiologic data support the idea that monosomy 7 and del(7q) often occur as a cooperating event during leukemia progression.^{22,25,27,30,31}  The MOL4070LTR-induced Mx1-Cre, 5A3^fl/+ AMLs in which the 5A3 allele was not recombined provided an opportunity to test the effects of inducing the 5A3 deletion in established leukemias. To test the system, we selected a Mx1-Cre, 5A3^fl/+ leukemia with partial recombination (AML no. 18) and another in which the 5A3 interval is unrecombined (AML no. 102; Figure 4B center and right). We transplanted both leukemias into sublethally irradiated recipients and administered high-dose pIpC treatment beginning 10 days after transplantation (Figure 5A). As shown in Figure 5B, this protocol resulted in partial excision of the 5A3 region in both AMLs. However, serial transplantation of both leukemias, which are highly aggressive in secondary and tertiary recipients, did not result in outgrowth of recombined cells (data not shown).

Monosomy 7 and del(7q) are commonly observed in therapy-related leukemias and in other myeloid malignancies that are resistant to conventional chemotherapeutic regimens. To address whether loss of the 5A3 interval might modulate drug sensitivity in vivo, we identified clones of AMLs nos. 18 and 102 with approximately 50% recombined cells, transplanted them into sublethally irradiated mice, and treated these recipients with a regimen of doxorubicin and cytosine arabinoside that mimics AML induction therapy (Figure 5A).³²  All of the recipients with AML no. 18 transplant died 16 to 30 days later of progressive leukemia irrespective of treatment (Figure 5D). Southern blotting showed that 5A3 recombined cells were not selected during chemotherapy treatment, suggesting that this leukemia is resistant to doxorubicin and cytosine arabinoside regardless of the status of the 5A3 interval (Figure 5C). Treatment with doxorubicin and cytosine arabinoside in mice that received a transplant of AML no. 102 resulted in death due to severe leukopenia. Mice treated with the vehicle died of progressive leukemia in a similar time frame between 12 and 26 days (Figure 5E). These studies demonstrate that it is feasible to induce the 5A3 deletion as a late event in established AML clones and to administer and test the effects of therapeutic agents in leukemias that contain a mixture of unrecombined and recombined cells.

Integrated cytogenetic and FISH studies of 81 well-characterized myeloid malignancies with 7q deletions provided a starting point for the chromosome engineering approach and subsequent analyses reported in this paper.²  In that report, 7q22 was involved in approximately 80% of cases. Because almost all of the 7q22 deletions found in individual cases are relatively large, the proximal boundary of the 2.5-Mb CDS was defined by performing FISH on leukemias with deletions arising in 7q22 that extended toward the 7q telomere and the distal boundary was mapped by analyzing other cases with more centromeric interstitial deletions that terminated within 7q22.²  To model loss of this interval in mice, we generated a “latent” chromosomal deletion by flanking a segment of orthologous synteny on chromosome 5A3 with loxP sites and used the Mx1-Cre transgene to excise this interval in vivo. This approach accurately models the pathogenesis of human hematologic cancers in which a small population of transformed cells achieves clonal dominance in vivo. We also used retroviral mutagenesis to establish leukemias without the desired deletion and then tested the impact of removing the 5A3 interval as a late event. Our data demonstrate that it is feasible to use this approach to create large somatic deletions in hematopoietic cells in vivo. Although inefficient, administering a single dose of pIpC to neonatal mice resulted in a population of 5A3-deleted hematopoietic cells that was detectable for 2 years and persisted after serial transplantation. Moreover, multiple transplantable MOL4070LTR-induced AMLs contained the 5A3 deletion. These data provide strong evidence that the 5A3 deletion occurs in the HSC compartment and in leukemia-initiating cells.

Mx1-Cre, 5A3^fl/+ mice that were injected with pIpC did not spontaneously develop hematologic cancers, and the targeted 5A3 chromosome did not alter the phenotype or course of MPD in Mx1-Cre, LSL-Kras^G12D or Mx1-Cre, Nf1^fl/fl mice. Furthermore, retroviral mutagenesis caused acute leukemia with similar latency, penetrance, and phenotypic features in control and Mx1-Cre, 5A3^fl/+ mice. Finally, we found that the 5A3 deletion neither provides a clonal growth advantage in established AMLs nor alters responses to antileukemia drugs. The most straightforward interpretation of these extensive data is that the human 7q22/murine 5A3 CDS does not harbor a myeloid TSG. This idea is compatible with previous studies of myeloid malignancies with monosomy 7/del(7q), which revealed substantial heterogeneity and implicated the 7q22 CDS that we investigated as well as centromeric or telomeric DNA segments as potentially containing myeloid TSGs.^33-37 SRPK2 and EPO bracket a minimal 4.6-Mb interval that is lost in some individual leukemias.³⁸  This minimal DNA segment includes the 7q22/5A3 CDS that we targeted and an additional approximately 2 Mb of DNA located between FBXL13 and EPO. Interestingly, whereas the region of orthologous synteny in mouse 5A3 diverges beyond Fbxl13, most of the murine Fbxl13-Epo interval is located in a single segment of chromosome band 5G2 that could be targeted by chromosome engineering.

An alternative explanation for the absence of hematologic disease in Mx1-Cre, 5A3^fl/+ and Mx1-Cre, 5A3^fl/fl mice is that the TSG has a weak effect, but cooperates with the loss of other chromosome 7 genes in leukemogenesis. MLL5 is an intriguing candidate TSG within the 7q22 CDS.³  Three groups recently deleted the murine homolog and showed that homozygous Mll5 inactivation reduces the numbers of phenotypic long-term repopulating cells in the bone marrow and markedly impairs HSC function under competitive repopulating conditions.^39-41  If MLL5 plays a similar role in human HSC homeostasis, biallelic inactivation might reduce the fitness of potential leukemia-initiating cells. The observation that MLL3, a related gene located at chromosome band 7q34, is mutated in some human cancers raises the possibility that haploinsufficiency for MLL5 and MLL3 may interact in leukemogenesis.^42-44  A microdeletion located on chromosome band 7q21.3 that was recently implicated as containing a haploinsufficient myeloid TSG⁴⁵  might also cooperate with deletion of the 7q22 CDS. The idea that multiple 7q genes might contribute to leukemogenesis is consistent with recent studies of myeloid malignancies with 5q deletions, which suggest that haploinsufficiency for CTNNA1, APC, EGR1, RPS14, miR-145, and miR-146a cooperate to deregulate hematopoietic growth.^7,46-49 

Other potential explanations for the absence of myeloid disease in Mx1-Cre, 5A3^fl/+ mice include the following: (1) radiation exposure, treatment with an alklyating agent, or transducing bone marrow cells with myeloid oncogenes might have uncovered a role of the 5A3 interval in leukemogenesis; (2) leukemic outgrowth could require efficient excision of the 5A3 interval in nonhematopoietic cells within the bone marrow microenvironment; and/or (3) a species-specific requirement for a 7q22 myeloid TSG that is confined to human hematopoietic cells.

Despite any impact of the 5A3 deletion on the incidence or phenotype of malignant hematologic disease, our data demonstrate that chromosome engineering can be used to create latent chromosomal deletions that can be induced in vivo in HSCs and leukemia-initiating cells. This allows investigators to bypass haplolethal developmental effects and to investigate chromosomal loss as a late event in leukemogenesis. However, excision of the 2-Mb targeted interval was inefficient in Mx1-Cre, 5A3^fl mice, and our efforts to delete this segment by transducing hematopoietic and mouse embryonic fibroblasts with a Cre-expressing retroviral vector were unsuccessful. It is uncertain whether the low rates of somatic recombination that we observed will hold for other genomic segments or whether this potential problem can be overcome by modifying the targeting constructs so that recombination leads to expression of a robust reporter. As we pursued the studies reported in this paper, it became obvious that the low rate of somatic recombination precluded ascertaining either subtle effects on hematopoietic growth and differentiation or the potential role of nonhematopoietic cells within the marrow microenvironment in promoting leukemic growth. After many attempts, we succeeded in generating mice with the 5A3 deletion in the germline. Homozygous deletion of the 5A3 interval is embryonically lethal, whereas mice with heterozygous deletion of the 5A3 interval are grossly normal and have blood counts similar to their wild-type littermates at age 3 months (data not shown). These preliminary observations are consistent with the results we obtained with the Mx1-Cre, 5A3^fl mouse model, and additional studies are in progress to fully address the role of this interval in hematopoietic growth control and leukemogenesis.

We thank the members of the Shannon and Killeen laboratories for helpful discussions and gratefully acknowledge the assistance of the University of California, San Francisco Comprehensive Cancer Center's Mouse Pathology and Transgenic and Targeted Mutagenesis Core Facilities. We thank David Tuveson and Tyler Jacks for LSL-Kras^G12D mice and Luis Parada for Nf1^flox mice.

This study was supported by National Institutes of Health grants P01CA40046 (M.M.L.B, K.S.), U01CA84221 (M.M.L.B, S.C.K, N.K., K.S.), and R37CA72614 (K.S.); and by SCOR (Specialized Center of Research; M.M.L.B., S.C.K., K.S.), Scholar (N.K., S.C.K.), and Fellowship awards from the Leukemia & Lymphoma Society of America (J.C.Y.W.).

National Institutes of Health

Contribution: J.C.Y.W., Y.Z., M.M.L.B., K.H.L., M.T.T., E.F., K.W., and P.A. conducted experiments and analyzed data; L.W. provided essential reagents; S.C.K. performed pathologic analysis; K.A. analyzed MOL4070LTR integration site data; J.C.Y.W., M.M.L.B., N.K., and K.S. designed research and analyzed data; and J.C.Y.W., N.K., and K.S. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for Y.Z. is Unit of Hematopoietic Stem Cell and Transgenic Animal Models, Institut Pasteur of Shanghai, Chinese Institute of Sciences, Shanghai, China.

Correspondence: Kevin Shannon, Helen Diller Family Cancer Research Building, 1450 3rd St, Rm 240, San Francisco, CA 94158-9001; e-mail: shannonk@peds.ucsf.edu; or Nigel Killeen, Department of Microbiology and Immunology; University of California, 513 Parnassus Ave, San Francisco, CA 94143-0414; e-mail: nigel.killeen@ucsf.edu.