Mechanistic and Preclinical Insights from Mouse Models of Hematologic Cancer Characterized by Hyperactive Ras

Abstract

RAS genes are mutated in 5%–40% of a spectrum of myeloid and lymphoid cancers with NRAS affected 2–3 times more often than KRAS. Genomic analysis indicates that RAS mutations generally occur as secondary events in leukemogenesis, but are integral to the disease phenotype. The tractable nature of the hematopoietic system has facilitated generating accurate mouse models of hematologic malignancies characterized by hyperactive Ras signaling. These strains provide robust platforms for addressing how oncogenic Ras expression perturbs proliferation, differentiation, and self-renewal programs in stem and progenitor cell populations, for testing potential therapies, and for investigating mechanisms of drug response and resistance. This review summarizes recent insights from key studies in mouse models of hematologic cancer that are broadly relevant for understanding Ras biology and for ongoing efforts to implement rational therapeutic strategies for cancers with oncogenic RAS mutations.

Somatic RAS gene mutations were first identified in leukemia in the 1980s (reviewed in Bos 1989). Subsequent studies encompassing thousands of hematologic cancers have defined the overall incidence of NRAS, KRAS, and HRAS mutations in different blood cancers, the frequency of individual amino acid substitutions, the likely order in which these mutations are acquired during leukemogenesis, and the status of RAS mutations detected at diagnosis, during disease remission, and after relapse (reviewed in Ward et al. 2012). The NF1 tumor suppressor gene, which encodes a GTPase activating protein (GAP) that is a core component of the Ras/GAP molecular switch (reviewed in Cichowski and Jacks 2001), is also recurrently mutated in a number of hematologic cancers (Ward et al. 2012). We first summarize key findings from studies of human patients that have informed our current understanding of the role of these mutations in blood cancers and refer readers to a recent review for additional background information.

The overall prevalence of RAS/NF1 mutations in hematologic malignancies ranges from ∼5% in myelodysplastic syndrome (MDS) to ∼20% in acute myeloid leukemia (AML) to ∼40% in juvenile myelomonocytic leukemia (JMML). In contrast to most solid cancers, NRAS mutations predominate over KRAS mutations by a ratio of ∼3:1 in hematologic malignancies (Ward et al. 2012). Deep sequencing of human tumors indicates that RAS and NF1 mutations do not initiate most hematologic cancers, but typically comprise secondary events that cooperate with antecedent mutations in genes encoding transcription factors and proteins that regulate epigenetic programs in hematopoietic stem and progenitor cells (HSPCs) (Jan et al. 2012; Shlush et al. 2014; Lindsley et al. 2015). JMML, an aggressive cancer of young children, is a notable exception to this general rule (Jankowska et al. 2011; Sakaguchi et al. 2013; Stieglitz et al. 2015). Interestingly, RAS (and FLT3) mutations identified at diagnosis are invariably undetectable in blood and bone marrow samples from patients with AML and acute lymphoblastic leukemia (ALL) analyzed during remission (Lindsley et al. 2015). This contrasts with mutations in genes such as DMNT3A, TET2, and IDH1/2, which frequently persist (Chou et al. 2012; Corces-Zimmerman et al. 2014; Klco et al. 2015). Furthermore, RAS mutations that are present at diagnosis do not invariably reappear at disease relapse, underscoring the impact of treatment-induced selective pressure on the dominant clone and highlighting the subsequent genetic evolution that occurs in response to it. Finally, and importantly, when relapse follows a period of remission in patients with acute leukemia, molecular analysis frequently reveals outgrowth of a rare clone with intrinsic drug resistance that was already present at diagnosis (reviewed in Jan and Majeti 2013).

The broad subject of mouse cancer models driven by oncogenic Ras genes is reviewed elsewhere in the literature (Drosten et al. 2017; Jacks 2017). Here we focus on mouse models of hematologic malignancies characterized by RAS and NF1 mutations, summarize key insights from these systems, and discuss the advantages and potential liabilities of different experimental approaches. Importantly, as no mechanism-based treatments exist for the ∼25% of human cancers with KRAS, NRAS, or NF1 mutations, we emphasize the use of these models as platforms for investigating mechanisms of drug response and resistance that may inform efforts to target oncogenic Ras signaling in both hematologic and nonhematologic cancers.

TRANSGENIC AND RETROVIRAL TRANSDUCTION/TRANSPLANTATION MODELS

Oncogenic KRAS, NRAS, and HRAS mutations are distributed nonrandomly across different cancers. The prevalence of NRAS and, to a lesser extent, KRAS mutations in human myeloid malignancies prompted efforts to develop mouse models by expressing different mutant Ras alleles under the control of various promoters. Transgenic animals in which the mouse mammary tumor virus (MMTV) long terminal repeat (LTR) was used to drive oncogenic Nras expression developed T- and B-cell lymphoblastic lymphomas and mammary carcinomas (Mangues et al. 1996). Using this same promoter to drive transgenic expression of Hras caused B-cell lymphoblastic lymphomas at low frequency. While disappointing at the time because RAS gene mutations are more frequent in myeloid malignances and were not thought to contribute to lymphoid cancers, RAS and NF1 mutations were later identified in certain aggressive types of ALL such as early T-cell precursor ALL (ETP-ALL) and hypodiploid B lineage ALL (Zhang et al. 2012; Holmfeldt et al. 2013).

A seminal study showed that retroviral transduction of murine bone marrow with the BCR-ABL transgene, a potent upstream activator of Ras and of the Raf/MEK/ERK effector pathway (Sawyers 1999), followed by transplantation into irradiated recipient mice, caused myeloproliferative neoplasms (MPNs) that closely resembled human chronic myeloid leukemia (CML) (Daley et al. 1990). Following this paradigm, a number of groups pursued a similar strategy with RAS oncogenes (Fig. 1), which was facilitated by development of the murine stem cell virus (MSCV) vector. In one study, ∼60% of irradiated recipient mice injected with bone marrow cells that had been transduced with a retrovirus expressing an oncogenic Nras allele developed a spectrum of myeloid malignancies (MacKenzie et al. 1999). However, the phenotypic variability, incomplete penetrance, and prolonged latency combined with both impaired in vitro proliferation and a high rate of apoptosis in Nras-infected cells provided an early indication that levels of oncogenic Ras expression strongly modulate cellular phenotypes. Consistent with this idea, positioning the Nras oncogene downstream of an internal ribosomal entry site (IRES) in the MSCV retroviral backbone enhanced induction of MPNs and AML, which was likely caused by a lower level of expression (Parikh et al. 2007). A direct comparison of the tumorigenic capacity of oncogenic Hras, Nras, or Kras was performed by expressing each oncogene downstream of an IRES in the MSCV vector that also contained a green fluorescent protein (GFP) selectable marker (MSCV-IRES-GFP [MIG]). In this study, transduced (GFP-positive) bone marrow HSPCs expressing oncogenic Nras caused myeloid malignancies that resembled human chronic myelomonocytic leukemia (CMML) at low viral titers and AML at higher titers. HSPCs expressing oncogenic Kras uniformly caused a CMML-like disease in recipient mice, the latency of which correlated with viral titer. Finally, expressing oncogenic Hras exclusively produced an invasive AML-like disease and no effect of viral titer was observed (Parikh et al. 2007). The robust ability of oncogenic Hras to induce myeloid malignancies despite the very low frequency of HRAS mutations in patient samples (Ward et al. 2012), coupled with the observed effects of viral titer, suggests that the superphysiological levels of Ras expressed in these systems strongly influence disease phenotypes. Importantly, the hematologic cancers that emerge from retroviral transduction/transplantation experiments invariably exhibit clonal retroviral integrations (MacKenzie et al. 1999; Parikh et al. 2006, 2007), arguing that a specific level of oncogenic Ras protein expression is strongly selected for, or that misregulation of genes near the integration site or other cooperating somatic mutations are required to induce leukemia in vivo.

Figure 1.

Retroviral transduction/transplantation to generate acute leukemias. Bone marrow or fetal liver cells are isolated from donor mice, transduced with retroviral vectors, sorted to obtain a pure population, and then transplanted into irradiated recipient mice. These animals develop acute leukemias that can be harvested and analyzed. GFP, Green fluorescent protein.

Several recent studies have deployed modified retroviral transduction/transplantation or transgenic models to examine Ras membrane trafficking, the roles of different effector pathways in leukemogenesis, the ability of oncogenic Ras expression to cooperate with other mutations to generate myeloid malignancies, and to perform preclinical testing of various therapeutic strategies. As described in detail in Philips (2017), Ras processing was initially pursued as a therapeutic target in the 1980s and is an area of renewed interest. Retroviral transduction of a palmitoylation-defective form of oncogenic Nras followed by transplantation of sorted GFP-positive HSPCs abrogated myeloid disease and biochemical activation of downstream effectors in recipient mice, suggesting a requirement for Nras palmitoylation and subsequent plasma membrane localization to induce myeloid malignancies (Cuiffo and Ren 2010). However, a subsequent study showed that expressing Nras^G12D from the MSCV promoter has dominant negative effects on myeloid progenitor colony growth and Ras-regulated activation of effector pathways ex vivo (Xu et al. 2012). These data, in turn, raise the possibility that nonphysiologic levels of oncogenic N-Ras^G12D expression impaired the growth of transduced HSPCs before and after transplantation. Another recent application of the MIG vector transduction/transplantation system assessed the ability of “second-site” amino acid substitutions that impair the ability of K-Ras^G12D to bind Raf or PI3 kinase (PI3K) to initiate leukemia in vivo (Shieh et al. 2013). HSPCs transduced with these mutant Kras^G12D oncogenes efficiently generated aggressive T-ALLs in recipient mice, and these leukemias restored downstream effector activation by either acquiring “third-site” mutations within the Kras transgenes or by silencing phosphatase and tensin homolog (PTEN) expression (Shieh et al. 2013). Inducible transgenic expression of oncogenic Nras under control of the Vav promoter caused mast cell leukemia that was reversed by administering doxycycline to repress expression from a tetracycline response element upstream of Nras. This inducible allele also cooperated with an Mll-Af9 fusion gene to produce AMLs that required continuous oncogenic Nras expression to promote self-renewal of the leukemic population (Kim et al. 2009; Sachs et al. 2014). Similarly, retroviral transduction was used to coexpress oncogenic Nras with AML1/ETO or MLL fusion genes under control of the MSCV promoter in fetal liver cells, which were then transplanted into irradiated recipient mice to generate AMLs. These leukemias were subsequently used in preclinical studies to uncover differential effects of standard induction chemotherapy regimens resulting from a difference in the ability of each fusion protein to activate a p53 response (Zuber et al. 2009). This transduction/transplantation model was used more recently in combination with a tetracycline-regulated RNA interference (RNAi) system (Zuber et al. 2011a) to identify BRD4 as a therapeutic target in AML (Zuber et al. 2011b), and to uncover potential mechanisms of resistance to small molecule inhibitors of this epigenetic modifier (Rathert et al. 2015).

In summary, retroviral transduction/transplantation and transgenic models have helped elucidate important aspects of Ras biology and have also been used to assess drug response and resistance. As described below, the development of “knockin” mouse strains has provided investigators in the field with more robust and reliable experimental models that overcome the potential confounding effects of nonphysiologic levels of oncogenic Ras expression that also vary across different leukemias because of differences in the genomic location of the integrated transgene.

IN VIVO EFFECTS OF NF1 INACTIVATION AND OF ENDOGENOUS KRAS^G12D/NRAS^G12D EXPRESSION IN THE HEMATOPOIETIC COMPARTMENT

The World Health Organization classifies JMML and CMML as MPN/MDS “overlap” disorders. These aggressive hematologic cancers share overlapping clinical and biologic features, including excessive proliferation of cells in the monocytic lineage, anemia, splenomegaly, progression to AML in a subset of patients, and resistance to cytotoxic chemotherapy. One important difference between CMML and JMML is that mutations in genes that broadly regulate epigenetic programs appear to initiate CMML, but not JMML (Jankowska et al. 2011; Sakaguchi et al. 2013; Stieglitz et al. 2015). This observation, in turn, suggests that fetal HSPCs are uniquely susceptible to RAS-induced transformation in vivo. Indeed, JMML likely represents the clearest example of a human cancer driven by aberrant Ras signaling as ∼90% of JMML genomes harbor driver mutations in one of five genes involved in Ras signaling (NF1, NRAS, KRAS, PTPN11, CBL), but very few additional somatic alterations (Sakaguchi et al. 2013; Stieglitz et al. 2015; reviewed in Chang et al. 2014).

Given these human data, it was exciting when investigators showed that expressing Cre recombinase in the blood and bone marrow of mice harboring conditional mutant alleles of Kras, Nras, or Nf1 induced myeloid malignancies that recapitulated many aspects of JMML and CMML. A key advance in the field was the development of a conditional Lox-Stop-Lox (LSL) Kras^G12D “knockin” allele that allowed investigators to express oncogenic Kras from its endogenous locus following Cre-mediated excision of the inhibitory LSL cassette (Jackson et al. 2001). Although embryonic expression of oncogenic Kras with ubiquitous or germline promoters was lethal, instilling an Adeno-Cre vector into the lungs of LSL-Kras^G12D mice generated multifocal adenocarcinomas that correlated with the viral titer (Jackson et al. 2001). To examine the effects of oncogenic Kras expression in the hematopoietic compartment, two groups intercrossed LSL-Kras^G12D and Mx1-Cre mice (Braun et al. 2004; Chan et al. 2004). In this system, the interferon-inducible Mx1 promoter drives Cre recombinase expression following in vivo administration of polyI-polyC (pI-pC), a synthetic double-stranded RNA that induces endogenous interferon expression (Kühn et al. 1995). Mice expressing Kras^G12D under control of the Mx1 promoter developed fully penetrant and aggressive MPNs characterized by leukocytosis, splenomegaly, and bone marrow myeloid hyperplasia. Myeloid progenitor cells from these mice form colonies in methylcellulose media in the absence of cytokine growth factors and also display hypersensitivity to granulocyte-macrophage colony-stimulating factor (GM-CSF), a cellular hallmark of JMML (Emanuel et al. 1991). Taken together, these studies established a method for conditional expression of oncogenic Ras alleles from their endogenous loci in HSPCs, showed the ability of endogenous Kras^G12D to perturb hematopoietic growth at the single-cell level, and generated a new and robust model of JMML and CMML. An unexpected finding was that endogenous Kras^G12D expression in primary hematopoietic cells resulted in a modest increase in the levels of activated Ras-GTP, but low basal ERK and Akt phosphorylation (Braun et al. 2004). This observation provided one of the first hints that hyperactive Ras signaling induces potent negative biochemical feedback in primary cells, which was also observed in a subsequent analysis of defined bone marrow cell populations isolated from Mx1-Cre; LSL-Kras^G12D/+ mice (Van Meter et al. 2007).

To compare the effects of endogenous Kras^G12D versus Nras^G12D expression in colonic epithelium, a conditional LSL-Nras^G12D allele was expressed using the Fabpl-Cre promoter. Whereas mice expressing Kras^G12D developed colonic hyperplasia, Nras^G12D expression failed to induce proliferation of colon cells but instead conferred resistance to apoptosis (Haigis et al. 2008). Similarly, inducing expression of Nras^G12D in the hematopoietic compartment of Mx1-Cre; LSL-Nras^G12D/+ mice had a different outcome than expressing Kras^G12D in the same strain background. Although Mx1-Cre; LSL-Kras^G12D/+ mice invariably died from progressive MPNs by 4 months of age, Mx1-Cre; LSL-Nras^G12D/+ mice exhibited indolent myeloid disease that developed with greatly increased latency. Approximately 80% of these mice ultimately succumbed to a variety of hematologic malignancies by 1 year of age, which included MPNs, an MDS-like disorder, lymphoproliferative disease, and histiocytic sarcoma. Although bone marrow cells from Mx1-Cre; LSL-Nras^G12D/+ mice showed cytokine-independent growth and GM-CSF hypersensitivity in methylcellulose cultures, these phenotypes were much less pronounced than in Kras mutant progenitors. Total Ras protein expression and Ras-GTP levels were also significantly higher in myeloid lineage cells expressing Kras^G12D as compared to those expressing Nras^G12D (Li et al. 2011). These differences in K-Ras^G12D and N-Ras^G12D protein expression further suggested that Kras/Nras oncogene dosage might contribute to hematologic phenotypes. This hypothesis was tested directly by generating Mx1-Cre mice expressing one or two Nras^G12D alleles and showing that the homozygous Nras^G12D/G12D genotype caused an acute MPN that more closely resembled the aggressive myeloid disease induced by Kras^G12D expression. In addition, transplanting bone marrow cells isolated from heterozygous Nras^G12D mutant mice into lethally irradiated recipients resulted in predominantly CMML and rarely T-ALL, whereas transplantation of Nras^G12D/G12D cells exclusively caused T-ALL (Wang et al. 2010). Analyzing hemizygous Mx1-Cre; Nras^G12D/- mice confirmed that gene dosage, and not tumor suppressor activity of wild-type (WT) Nras, was responsible for the more aggressive biologic behavior of homozygous mutant Nras^G12D/G12D hematopoietic cells (Xu et al. 2013). Finally, heterozygous Nras^Q61L expression driven by Mx1-Cre caused an intermediate phenotype of MPNs in mice that was more aggressive and penetrant than heterozygous Nras^G12D expression but less severe than homozygous Nras^G12D expression (Kong et al. 2016). Because N-Ras^Q61L is more activated biochemically than N-Ras^G12D, these studies provide additional evidence that the degree and duration of oncogenic Ras output modulates cellular phenotypes. In this context, it is striking that Nras^G12D, which has less potent effects on the proliferation and differentiation of primary hematopoietic cells as compared to either Nras^Q61L or Kras^G12D, is also the most common mutation found in human blood cancers. Together, these data support the existence of substantial positive and negative pressure that selects for the outgrowth of HSPCs with specific levels of aberrant Ras pathway activation in vivo.

The consequences of disrupting Nf1 were first examined using gene targeting to generate mice expressing a null allele (Brannan et al. 1994; Jacks et al. 1994). Homozygous Nf1 mutant embryos die in utero from cardiac defects. Interestingly, however, Nf1-deficient fetal liver cells displayed GM-CSF hypersensitivity in methylcellulose cultures and caused JMML-like MPNs upon transplantation into lethally irradiated recipients (Bollag et al. 1996; Largaespada et al. 1996). Heterozygous Nf1 mutant mice are phenotypically normal over the first year of life and subsequently develop pheochromocytomas and JMML-like MPNs with incomplete penetrance (Jacks et al. 1994). These data are consistent with the tumor spectrum observed in persons with neurofibromatosis type 1 (NF1), a familial cancer syndrome caused by germline NF1 mutations. The tumor predispositions and other phenotypic features of NF1 disease, such as the role of germline and somatic NF1 mutations in tumorigenesis, and the role of NF1 GAP activity in regulating Ras signaling are discussed in detail in Cichowski (2017). Heterozygous Nf1 inactivation also cooperated with whole-body and focal radiation and the alkylating agent cyclophosphamide to induce breast cancers, myeloid malignancies, sarcomas, and pheochromocytomas (Chao et al. 2005; Nakamura et al. 2011).

To circumvent the embryonic lethality resulting from homozygous Nf1 inactivation and to create a mouse model of human NF1 disease, a conditional mutant Nf1 allele was generated (Zhu et al. 2001). Driving Cre recombinase expression from the Synapsin I promoter showed that tissue-restricted Nf1 inactivation caused abnormal development of the cerebral cortex and increased astrocyte proliferation (Zhu et al. 2001). Mx1-Cre; Nf1^flox/flox mice developed fully penetrant MPNs with a median survival of 7.5 months—an intermediate hematologic phenotype between those observed in Mx1-Cre; LSL-Nras^G12D/+ and Mx1-Cre; LSL-Kras^G12D/+ mice. This myeloid disease models aspects of JMML, including splenomegaly, bone marrow infiltration by myeloid cells, and GM-CSF hypersensitivity. Transplanting Nf1-deficient bone marrow caused MPNs in lethally but not sublethally irradiated recipient mice (Le et al. 2004). Mouse models exhibiting conditional Nf1 inactivation in the hematopoietic compartment have facilitated comparing the consequences of direct versus indirect activation of Ras signaling, and have also provided an additional model of MPNs for performing preclinical studies.

BIOLOGIC AND PRECLINICAL STUDIES IN KRAS, NRAS, AND NF1 MUTANT MICE

In addition to generating models that recapitulate the characteristics of hematologic malignancies driven by hyperactive Ras, conditional expression of Kras^G12D or Nras^G12D or ablation of Nf1 has elucidated fundamental aspects of Ras biology and tumorigenesis. The relative fitness of Nf1 mutant and WT HSPCs was assessed by performing repopulation experiments (Fig. 2) in which defined numbers of sorted Nf1-deficient bone marrow cells and WT “competitors” were injected into irradiated recipient mice (Zhang et al. 2001). In these studies, Nf1 mutant HSPCs outcompeted WT cells, but failed to induce MPNs unless a high ratio of mutant-to-WT cells was injected (Zhang et al. 2001). Similarly, Nras^G12D expression caused increased proliferation of highly purified hematopoietic stem cells (HSCs) and enhanced their competitive fitness but did not induce MPNs in recipients when coinjected with competitor cells. These data have important implications for modeling leukemogenesis, as they indicate that WT cells can suppress the tumorigenic potential of primary cells with oncogenic alterations that activate Ras signaling. In addition, endogenous Nras^G12D expression unexpectedly increased the frequency of cell divisions in a subpopulation of HSCs, but reduced this parameter in others. Enhanced self-renewal potential and competitive fitness was restricted to the infrequently dividing HSC population and was shown to be STAT5-dependent (Li et al. 2013). Oncogenic Kras expression increased HSC proliferation and conferred a competitive advantage over WT cells under repopulating conditions (Sabnis et al. 2009). In contrast to Nf1^–/– and Nras^G12D-expressing HSPC, Kras^G12D-expressing HSCs efficiently induced T-ALL under competitive repopulation conditions. These aggressive cancers were characterized by secondary mutations in Notch1, demonstrating the acquisition of cooperating mutations during progression to acute leukemia (Kindler et al. 2008; Sabnis et al. 2009). These studies show how genetically accurate mouse models are robust platforms for identifying cancer-initiating populations and for addressing fundamental questions regarding how endogenous oncogenic Ras expression perturbs stem cell fates. They also underscore the relative potency of oncogenic Kras versus oncogenic Nras or mutant Nf1 in leukemia initiation, albeit under the somewhat artificial experimental condition of competitive repopulation in irradiated congenic recipient mice.

Figure 2.

Competitive repopulation to assess fitness of hematopoietic stem cells (HSCs). Bone marrow cells are isolated from wild-type (WT) or mutant donor mice of different genetic backgrounds, then HSCs are sorted and mixed at a fixed ratio. These competitor cells are transplanted into irradiated recipient mice, then bone marrow is harvested and analyzed using fluorescence-activated cell sorting (FACS) to determine the percentage of mutant versus WT donor cells in the resulting population.

The observed GM-CSF hypersensitivity of Nf1-deficient progenitors in methylcellulose cultures raised the question of whether abnormal responses to exogenous stimuli are integral to the development of MPNs in vivo. To address this, fetal liver cells homozygous for mutations in both Nf1 and Gmcsf were transplanted into WT or Gmcsf-deficient recipients (Birnbaum et al. 2000). Remarkably, GM-CSF expression in either donor cells or the host bone marrow microenvironment was sufficient to induce MPNs, but Gmcsf inactivation in both contexts severely prolonged disease latency. Furthermore, transplanting doubly mutant bone marrow cells from WT recipients with established MPNs into secondary Gmcsf mutant recipients caused disease regression, which was reversed by exogenous GM-CSF administration (Birnbaum et al. 2000). In a related study, a mutation in the β common subunit of the GM-CSF receptor also attenuated the development of MPNs in Nf1 mutant mice (Kim et al. 2009). These data indicate that Nf1 inactivation cooperates with upstream inputs from activated growth factor receptors to drive aberrant proliferation in vivo, and are consistent with biochemical studies of primary Kras and Nf1 mutant cells, showing that Ras effector pathways are not fully activated under basal conditions but respond robustly to growth factor stimulation (Diaz-Flores et al. 2013).

Two recent studies examined additional requirements for the development of MPNs in Nf1 and Nras mutant mice. In the first, genetic disruption of Erk1/2 abrogated myeloid disease in Mx1-Cre; Nf1^flox/flox mice, supporting an essential role of aberrant Raf/MEK/ERK signaling in induction of MPNs (Staser et al. 2013). More recently, genetic ablation of Zdhhc9, which encodes a palmitoyltransferase, impaired the ability of Nras^G12D to induce MPNs and T-ALL (Liu et al. 2016). In addition to uncovering genes required for the development of Ras-driven diseases in hematopoietic cells, these studies credentialed the Raf/MEK/ERK pathway as a therapeutic target in MPNs characterized by hyperactive Ras signaling and support inhibition of the palmitoylation/depalmitoylation cycle as a potentially effective therapeutic strategy for NRAS mutant hematologic cancers.

Nf1, Nras, and Kras mutant mice with MPNs were also harnessed to directly test small molecules with the potential to inhibit hyperactive Ras signaling. The initial preclinical trial administered L744,832, a potent and selective farnesyltransferase inhibitor (Kohl et al. 1995), to mice transplanted with homozygous Nf1-deficient fetal liver cells (Mahgoub et al. 1999). Although this compound reduced Raf/MEK/ERK pathway activation and abrogated GM-CSF colony growth ex vivo, it had no effect on MPNs in vivo. Consistent with these data, L744,832 inhibited H-Ras prenylation in recipient mice but a substantial fraction of total Ras was processed normally (Mahgoub et al. 1999), which was almost certainly caused by “bypass” N-Ras and K-Ras prenylation by gerenylgerenyl transferase. This study accurately predicted the disappointing efficacy of farnesyltransferase inhibitors in the clinic.

Directly inhibiting the oncogenic Ras/GAP switch continues to pose great challenges, and this has stimulated the development of numerous small molecule inhibitors of various Ras effectors. MEK is a particularly promising therapeutic target in many cancers, and the Food and Drug Administration has approved two allosteric inhibitors (trametinib and cobimetinib) for advanced melanoma. Preclinical studies evaluating efficacy of the “first-generation” MEK inhibitor CI-1040 in Mx1-Cre; Nf1^flox/flox mice with MPNs showed no beneficial effects (Lauchle et al. 2009). By contrast, Mx1-Cre; LSL-Kras^G12D mice that received PD0325901 (PD901), a “second-generation” MEK inhibitor with optimized pharmacologic properties (Brown et al. 2007), had remarkable hematologic improvement and greatly enhanced survival (Lyubynska et al. 2011). This study also uncovered an ability of PD901 to block GM-CSF hypersensitivity in oncogenic Kras mutant bone marrow cells in vitro. However, Kras^G12D hematopoietic cells persisted after treatment, indicating that MEK inhibition rebalanced growth and differentiation in vivo. Consistent with this observation, some Kras mutant mice progressed to T-ALL despite continuous treatment with PD901, which supports differential dependence on the Raf/MEK/ERK pathway in myeloid versus lymphoid cells. To determine whether the discordant results of these preclinical trials reflected differential sensitivity of Nf1 and Kras mutant hematopoietic cells to MEK inhibition or indicated that sustained Raf/MEK/ERK pathway inhibition is essential for therapeutic efficacy, PD901 was administered to Mx1-Cre; Nf1^flox/flox mice with MPNs and induced profound reductions in leukocytosis and splenomegaly (Chang et al. 2013). As in mice with Kras mutant MPNs, MEK inhibition failed to eradicate Nf1 mutant HSPCs but rather exerted antiproliferative and prodifferentiation effects in the myeloid and erythroid lineages.

Preclinical trials in mouse models of MPNs and other NF1-associated tumors (Jessen et al. 2013) informed a recent phase I clinical trial of the MEK inhibitor selumetinib, which showed tumor regression in most children with plexiform neurofibromas (Dombi et al. 2016). A national clinical trial of MEK inhibition in patients with relapsed/refractory JMML is also expected to open later this year. While data from Kras and Nf1 mutant mice support the idea that MEK inhibition will induce hematologic improvement without eradicating mutant HSPCs, certain clinical observations suggest that this approach has curative potential. Specifically, infants with developmental disorders of the Noonan syndrome (NS) spectrum sometimes develop a transient MPN that is clinically indistinguishable from JMML, but resolves spontaneously (Kratz et al. 2005; Lauchle et al. 2006; Loh 2011). The causative germline PTPN11 and KRAS mutations in these patients typically encode mutant proteins that are less activated biochemically than the strong somatic gain-of-function alleles detected in JMML (Keilhack et al. 2005; Schubbert et al. 2006, 2007). In these cases, modestly activated SHP-2 and K-Ras mutant proteins likely fail to confer a durable growth advantage. By contrast, stronger gain-of-function somatic mutations such as K-Ras^G12D and SHP-2^E76K appear to be capable of fully transforming fetal hematopoietic cells, leading to JMML. If this idea is correct, it raises the provocative possibility that pharmacologic inhibition could reduce Raf/MEK/ERK signaling below a critical threshold level and thereby cure some JMML patients by allowing the normal developmental program to reassert itself (Fig. 3). The design of the forthcoming clinical trial includes correlative pharmacokinetic and molecular studies to directly address any underlying mechanisms of response.

Figure 3.

Potential therapeutic responses to MEK inhibition in juvenile myelomonocytic leukemia (JMML). Normal hematopoiesis is dynamically regulated during development with maturation from a fetal-like stem/progenitor to adult populations. The observation of a self-limited JMML-like myeloproliferative neoplasm (MPN) in infants with disorders of the Noonan syndrome spectrum suggests that a threshold of aberrant Ras/Raf/MEK/ERK signaling is required to fully transform fetal stem/progenitor cells. Based on this idea and preclinical data from Kras and Nf1 mutant mice with MPNs, possible beneficial outcomes of sustained pharmacologic MEK inhibition in JMML include (1) clinical improvement (reduction in leukocyte counts, improvement of anemia, less splenomegaly) with persistence of the malignant clone; or (2) eradication of the JMML clone caused by restoration of normal developmental processes regulating hematopoietic cell fates (dotted line). Clinical trials with molecular assessment of mutant allele burden in patient specimens are required to investigate these potential mechanisms of response. HSC, Hematopoietic stem cell.

Mx1-Cre; Kras^G12D mice with MPNs were also used to test the efficacy of GDC-0941, a pan-PI3K inhibitor (Akutagawa et al. 2016). Treatment resulted in disease remission similar to that observed with PD901. However, biochemical studies of bone marrow cells showed reduced Raf/MEK/ERK pathway activation in response to GDC-0941, an observation that is consistent with a previous study showing that PI3K functions both upstream and downstream of K-Ras^G12D in myeloid lineage cells stimulated with cytokine growth factors (Diaz-Flores et al. 2013). To more specifically inhibit PI3K/Akt signaling, cohorts of Kras and Nf1 mutant mice were treated with the Akt inhibitor MK-2206 and preclinical efficacy was observed in both models (Akutagawa et al. 2016). These data highlight the utility of preclinical studies in mouse models of oncongenic Ras-driven MPNs to elucidate potential cross talk between different Ras effector pathways.

MODELS OF ACUTE LEUKEMIA CHARACTERIZED BY NF1 INACTIVATION OR ENDOGENOUS KRAS^G12D/NRAS^G12D EXPRESSION

Whereas Nf1 inactivation and endogenous Kras^G12D or Nras^G12D expression in hematopoietic cells generated robust and tractable models of CMML and JMML, these mice do not develop acute leukemia. This observation and genomic analysis of human leukemias infer a requirement of cooperating mutations for progression to AML or ALL. Insertional mutagenesis is an unbiased forward genetic strategy for generating diverse primary leukemias. The first successful example of this method utilized the BXH-2 mouse strain that expresses a B-ecotropic murine leukemia virus (MuLV) to induce AML with long latency (Jenkins et al. 1982; Bedigian et al. 1984). Heterozygous Nf1 inactivation both increased the incidence of myeloid leukemia and accelerated disease onset in BXH-2 mice (Largaespada et al. 1996). Interestingly, these Nf1 mutant AMLs frequently exhibited loss of heterozygosity at the Nf1 locus, consistent with data from JMML patients with NF1 showing loss of the normal allele (Shannon et al. 1994). Over a decade later, retroviral insertional mutagenesis (RIM) was used to generate AMLs with unique cooperating genetic lesions in Mx1-Cre; Nf1^flox/flox mice (Lauchle et al. 2009). This experimental approach involved simultaneously injecting newborn mice with pI-pC to induce Nf1 inactivation and with the MOL4070LTR retrovirus to generate integrations throughout the genome (Fig. 4A). In this system, Nf1 cooperated with MOL4070LTR to induce AMLs that were heterogeneous with respect to morphology, myeloid surface markers, and retroviral insertions. Myb was identified as a common insertion site in multiple independent Nf1 mutant AMLs, providing evidence for cooperation between these two genetic loci in myeloid leukemogenesis (Lauchle et al. 2009).

Figure 4.

Conditional Ras activation in the hematopoietic compartment. (A) Newborn mice are injected with MOL4070LTR to generate diverse genomic integrations and with polyI-polyC (pI-pC) to inactivate Nf1 in hematopoietic cells, resulting in predominantly acute myeloid leukemia (AML) with less frequent emergence of T-cell acute lymphoblastic leukemia (T-ALL). (B) Retroviral mutagenesis is performed in newborn mice, then pI-pC is administered at weaning to express Nras^G12D in hematopoietic cells and generate AML. (C) Mice are injected with MOL4070LTR shortly after birth, then with pI-pC at weaning to activate Kras^G12D. When these animals develop a lethal myeloproliferative disorder (MPD), bone marrow is transplanted into irradiated recipient mice to generate both AML and T-ALL.

RIM was also performed in Mx1-Cre; Nras^G12D/+ mice using a modified approach (Fig. 4B) in which mice were injected with MOL4070LTR shortly after birth and pI-pC was administered at weaning (Li et al. 2011). This experimental system more closely recapitulates the pathogenesis of human AML in which RAS mutations generally comprise secondary events that cooperate with antecedent oncogenic translocations or with mutations in genes encoding hematopoietic transcription factors or proteins that broadly regulate epigenetic programs (Jan et al. 2012; Shlush et al. 2014; Lindsley et al. 2015). The Nras mutant AMLs generated in this study shared phenotypic features with the M4/M5 subtype of human AML, frequently exhibited loss of the WT Nras allele, and showed heterogeneous activation of Ras effector pathways. These AMLs were transplantable into sublethally irradiated recipient mice, and showed a clonal pattern of retroviral insertions that was stable upon secondary transplantation. The most frequent common insertion site was upstream of a gene encoding the zinc-finger transcription factor Evi-1 and resulted in a substantial increase in transcript expression specifically in leukemias harboring this integration (Li et al. 2011). Interestingly, NRAS mutations in human AML are frequently associated with both M4/M5 morphology and with EVI1 translocations (Bowen et al. 2005; Bacher et al. 2006).

In contrast to Nf1 and Nras mutant mice, Mx1-Cre; Kras^G12D/+ mice that were injected with MOL4070LTR uniformly died from MPNs. However, transplanting their bone marrow into sublethally irradiated recipients unexpectedly generated T-ALLs in ∼80% and AMLs in the remaining ∼20% of mice (Fig. 4C), which emerged with reduced latency compared to virus-injected WT animals. Primary T-ALLs were characterized by thymic masses, leukocytosis, extensive tissue infiltration, and an immature double-positive stage of lymphocyte development. Secondary transplant recipients also developed T-ALL with more profound bone marrow infiltration and variable thymic involvement (Dail et al. 2010). A common insertion site in Ikzf1, which encodes the lymphoid cell-specific transcription factor and known ALL tumor suppressor Ikaros (Rebollo and Schmitt 2003; Mullighan et al. 2007), was identified in Kras-driven T-ALLs and integrations in this gene appeared to result in proteins with dominant negative activity. Downstream Ras effector pathway activation was highly variable in Kras mutant and WT leukemias. Strikingly, somatic Notch1 mutations were identified in all Mx1-Cre; Kras^G12D/+ T-ALLs examined (Dail et al. 2010). This unanticipated observation demonstrates that cooperating events can arise from somatic alterations that are independent of retroviral insertions in RIM models of acute leukemia. AMLs generated by transplanting Mx1-Cre; Kras^G12D/+ bone marrow into sublethally irradiated recipients also contained clonal retroviral integrations within genes such as Myb and Evi5 (Burgess et al. 2017).

Taken together, data from RIM screens performed with the same virus in mice expressing mutant Nf1, Kras, and Nras in hematopoietic cells highlight the extent to which specific Ras pathway alterations can modulate the spectrum of acute leukemias that develop. This finding is consistent with the variable latency and penetrance of MPNs observed across the three genotypes described above. Importantly, these panels of genetically diverse transplantable leukemias were subsequently deployed to perform controlled preclinical trials of signal transduction inhibitors.

PRECLINICAL TRIALS AND STUDIES OF DRUG RESISTANCE IN MOUSE MODELS OF ACUTE LEUKEMIA

Acute leukemias generated by RIM recapitulate the genetic heterogeneity seen in advanced human cancers, and transplantation into recipient mice is an attractive preclinical platform (Fig. 5) for assessing therapeutic responses and elucidating mechanisms of drug resistance because (1) primary cancers are treated in immunocompetent mice; (2) retroviral integration patterns can be used to track the emergence of drug resistant clones; and (3) relapsed leukemia cells can be retransplanted to verify intrinsic resistance, test alternative therapies, and validate candidate resistance mechanisms. To examine the sensitivity of myeloid malignancies characterized by Nf1 inactivation to MEK inhibition, studies were initiated to assess the growth of bone marrow cells from WT mice, Nf1 mutant mice with MPNs, and Nf1-deficient AMLs generated by RIM in methylcellulose medium containing GM-CSF and various doses of CI-1040 (Lauchle et al. 2009). In this in vitro system, AML cells unexpectedly exhibited markedly enhanced sensitivity to MEK inhibition compared to either WT cells or those harvested from mice with MPNs. Transplanting Nf1-deficient AMLs into recipient mice and treating them in vivo revealed impressive efficacy of this compound and of PD901, which was characterized by a rapid clearance of circulating leukemic blasts and a greater than threefold increase in median survival compared to vehicle-treated animals. However, all of these mice eventually relapsed despite sustained treatment. AMLs that relapsed after a prolonged response showed intrinsic drug resistance upon retransplantation and retreatment and also exhibited clonal evolution by the criterion of one or more novel retroviral integrations detected on Southern blots. Cloning retroviral insertions from resistant AMLs identified RasGrp1 overexpression and disruption of one allele of Mapk14 as candidate resistance mechanisms that were validated functionally (Lauchle et al. 2009).

Figure 5.

Preclinical studies using acute leukemias generated by retroviral insertional mutagenesis (RIM). Mice expressing mutant Nf1, Nras, or Kras under the control of Mx1-Cre are subjected to RIM to generate large panels of cryopreserved acute leukemias. These cells can be serially transplanted into recipient mice, exposed to various drugs or combinations in vivo, and then harvested at relapse and analyzed.

Recipient mice transplanted with Nras^G12D AMLs showed a modest but significant survival benefit in response to treatment with PD901 or trametinib. In contrast, these AMLs did not respond to the pan-PI3K inhibitor GDC-0941 as a single agent, and combining this drug with PD901 also had limited effects. As was observed in Nf1-deficient AMLs, all of the recipient mice given MEK and/or PI3K inhibitors relapsed on treatment. However, none of these relapsed AMLs harbored novel retroviral integrations and all showed a similar response to MEK inhibition upon retransplantation. Further analysis revealed that MEK inhibition induces neither differentiation nor apoptosis of leukemia cells, but rather exerts its antileukemic effect by reducing proliferation of Nras^G12D AMLs (Burgess et al. 2014). Although this study showed modest efficacy of potent small-molecule MEK inhibitors, the lack of clonal evolution or acquired resistance in relapsed AMLs was an important limitation. Recently, an unbiased screen identified the creatine kinase pathway as a novel therapeutic vulnerability in AML cell lines (Fenouille et al. 2017). This finding was validated by transplanting primary Nras mutant AMLs generated by RIM, treating them with a chemical inhibitor of creatine biosynthesis, and showing that leukemia cells with an Evi1 insertion were much more sensitive to the drug (Fenouille et al. 2017).

Treatment with PD901 also extended the survival of mice transplanted with Kras^G12D AMLs with one leukemia showing an exceptional response, harboring a novel integration at relapse, and demonstrating both clonal evolution and intrinsic drug resistance upon retransplantation/retreatment. Extensive analysis of the parental AML and resistant subclone revealed a uniparental disomy event at the Kras locus that was followed by loss of the WT allele in the parental leukemia. This genetic configuration conferred a competitive advantage upon these primary AML cells, but also rendered them hypersensitive to MEK inhibition. Interestingly, the resistant AML exhibited trisomy 6 and a Kras genotype of two mutant and one WT allele. These data showed a role for serial genetic chances at the Kras locus in modulating Raf/MEK/ERK pathway dependence, which was subsequently also observed in a panel of human colorectal cancer cell lines. Furthermore, allelic imbalance at the KRAS locus was identified in over half of a large panel of advanced human cancers with diverse histologies harboring oncogenic KRAS mutations (Burgess et al. 2017).

Preclinical studies in Kras^G12D T-ALLs uncovered a pattern of dependencies on the downstream Ras effector pathways that was distinct from AML. Whereas these leukemias showed a modest response to PD901 and were refractory to GDC-0941, treatment with both inhibitors markedly extended survival (Dail et al. 2010, 2014). As in the AML models, all recipient mice ultimately died from drug-resistant leukemia, which exhibited remarkable clonal heterogeneity with numerous relapsed samples harboring different novel retroviral integrations. T-ALLs that relapsed after a prolonged response to treatment unexpectedly showed loss of retroviral integrations and somatic mutations that activated Notch1 and a corresponding elevation in PI3K pathway activation (Dail et al. 2014). These data have important implications for the treatment of T-ALL, as they suggest that the rational strategy of combining PI3K and NOTCH1 inhibitors might drive clonal outgrowth of drug-resistant cells.

CONCLUDING REMARKS

The tractable nature of the hematopoietic compartment and the availability of robust methodologies for isolating and functionally analyzing HSPCs in vivo have facilitated examining the effects of Ras pathway activation on HSPC biology and behavior. Extensive characterization of mouse models of MPNs driven by mutant Nf1, Kras, or Nras has revealed important differences at the level of individual HSPCs and, more globally, with respect to in vivo disease phenotypes and the extent to which downstream Ras effector pathways are activated following oncogene expression. Taken together, these studies indicate that heterozygous Kras^G12D is a more activating gain-of-function mutation than homozygous Nf1 inactivation, which is, in turn, more activating than a heterozygous Nras^G12D mutation. It will be interesting to perform similar comparisons in other tissue lineages. In addition to demonstrating differential effects of individual mutant alleles, these studies also underscore the importance of posttranslational modifications of Ras in tumorigenesis. Consistent with these findings, a recent genome-wide CRISPR-mediated screen to identify synthetic lethal interactions found that genes encoding proteins involved in Ras processing and components of the Raf/MEK/ERK pathway were essential specifically in human AML cell lines harboring oncogenic RAS mutations (Wang et al. 2017). Another article in this collection focuses on the topic of uncovering and validating candidate synthetic lethal interactions with oncogenic RAS (Aguirre and Hahn 2017).

Developing predictive preclinical models for testing new agents and drug combinations remains a fundamental challenge for advancing cancer medicine. In this regard, the ability to transplant human and mouse leukemias into cohorts of congenic recipient mice is a major advantage. Collections of patient-derived xenografts (PDXs) are also valuable for preclinical evaluation because they are derived from human cancers and can reflect the heterogeneity of a given patient population. However, the use of immunocompromised mice for engraftment precludes examination of any potential effects of the microenvironment on response and resistance to therapeutic agents, and the reduced genetic complexity of PDX models that results from the barrier of xenotransplantation (Klco et al. 2014) has potential limitations for characterizing relapse mechanisms. Whereas acute leukemias generated by RIM also recapitulate the genetic heterogeneity found in human tumors, they can be transplanted into congenic immunocompetent mice and interact with host cells in hematopoietic tissues. Although these models are dependent upon genetic and biochemical similarities between mouse and human cells, recent work has shown that insights from preclinical studies in RIM-induced AMLs can generate mechanistic hypotheses that are relevant to human cancers (Burgess et al. 2017). Mouse models of hematologic cancers generated by RIM can also be used to uncover mechanisms of resistance to conventional chemotherapeutic agents and to identify novel compounds that might overcome resistance to these drugs. Ideally, performing comprehensive in vivo preclinical drug testing that utilizes the RIM-driven leukemias described in this review in combination with banks of well-characterized human PDX models (Townsend et al. 2016) will identify promising therapeutic approaches for hematologic and solid cancers with oncogenic RAS mutations. The tractable experimental nature of the hematopoietic system makes it a particularly appealing model for addressing this fundamental problem.