Dominant Role of Oncogene Dosage and Absence of Tumor Suppressor Activity in Nras-Driven Hematopoietic Transformation

RAS genes encode ubiquitously expressed proteins (N-Ras, H-Ras, K-Ras4A, and K-Ras4B) that cycle between active GTP-bound and inactive GDP-bound conformations (Ras-GTP and Ras-GDP; ref. 1). Ras-GTP levels are regulated by the competing activities of guanine nucleotide exchange factors and GTPase-activating proteins (GAP), which enhance intrinsic Ras GTPase activity. Proteins encoded by RAS oncogenes, which accumulate in the GTP-bound state due to defective intrinsic GTP hydrolysis and resistance to GAPs, are exceedingly difficult targets for anticancer drug discovery due to their structural and biochemical properties (1).

Despite compelling evidence that oncogenic Ras proteins have dominant gain-of-function actions in cellular transformation, genetic studies in mice surprisingly have suggested that wild-type (WT) Ras exerts tumor suppressor activity in some cancers with oncogenic Ras mutations (2–5). However, mechanistic data regarding how normal Ras might antagonize oncogenic signaling are lacking.

Endogenous expression of Nras^G12D induces a myeloproliferative neoplasm in Mx1-Cre;Nras^G12D/+ mice that faithfully models human chronic and juvenile myelomonocytic leukemia (CMML and JMML; refs. 4, 6, 7). Hematologic disease is greatly accelerated in homozygous Nras^G12D-mutant mice (Mx1-Cre; Nras^G12D/G12D; refs. 8, 9). We deployed a conditional Nras-mutant allele to assess the relative contributions of NrasG12D oncogene dosage and tumor suppression by WT Nras in myeloid transformation. We find that elevated Nras^G12D expression drives myeloid transformation in vivo and strongly modulates cell growth, Ras signaling, and response to a targeted inhibitor in vitro. Consistent with these data, somatic uniparental disomy underlies loss of the WT allele in primary acute myeloid leukemia (AML) cells with Nras^G12D mutations, resulting in normal-to-increased Nras expression. Finally, NRAS expression is significantly elevated in human cancer cell lines with NRAS mutations, whereas KRAS expression is reduced, with a reciprocal pattern seen in cell lines with KRAS mutations.

We generated a Cre-dependent conditional Nras allele (Nras^2lox2) and conducted intercrosses to produce hemizygous (Mx1-Cre;LSL-Nras^G12D/2lox2), heterozygous (Mx1-Cre;LSL-NrasG12D^/+), and homozygous (Mx1-Cre;LSL-Nras^G12D/G12D) littermates on a C57Bl/6 strain background (Supplementary Fig. S1A). Use of this conditional Nras^2lox2 allele avoids potential confounding consequences of eliminating WT Nras expression throughout development and allowed us to simultaneously activate Nras^G12D expression and inactivate WT Nras in the hematopoietic compartment after birth (4). Efficient recombination of both conditional Nras alleles with loss of expression was observed 2 weeks later (Supplementary Fig. S1B and S1C). Western blot analysis confirmed that N-Ras protein levels are reduced in the bone marrow of hemizygous Nras-mutant mice (Fig. 1A), which we hereafter refer to as Nras^G12D/−.

Consistent with recent reports (8, 9), approximately 20% of Mx1-Cre;Nras^G12D/G12D mice died prematurely from T lineage acute lymphoblastic leukemia (Supplementary Fig. S2A). Surviving animals of all three Nras genotypes were euthanized at 6 months of age. All Nras^G12D/G12D mice had overt myeloproliferative neoplasm, which was characterized by leukocytosis with elevated blood neutrophil counts, splenomegaly, and anemia (Fig. 1B and Supplementary Fig. S2B). In contrast, hematologic parameters were normal in age-matched Mx1-Cre;Nras^G12D/+ and Mx1-Cre;Nras^G12D/− mice (Fig. 1B and Supplementary Fig. S2B). Flow cytometric analysis revealed increased numbers of immature monocytic (Mac-1⁺, Gr-1^lo) cells in the hematopoietic tissues of Nras^G12D/G12D mice, which is also observed in Nf1- and Kras-mutant mice with myeloproliferative neoplasm (ref. 7; Fig. 1C). This population was not expanded in hemizygous or heterozygous Nras-mutant mice.

We grew colony-forming unit granulocyte macrophage (CFU-GM) progenitors to assess the cell-intrinsic effects of Nras^G12D gene dosage and the consequences of inactivating WT Nras. Somatic NRAS mutations are highly prevalent in JMML and CMML, and CFU-GM progenitors from patients with these aggressive cancers are hypersensitive to granulocyte macrophage colony-stimulating factor (GM-CSF; ref. 7). Similarly, Nras^G12D/G12D bone marrow cells show cytokine-independent CFU-GM growth and pronounced GM-CSF hypersensitivity (Fig. 1D; refs. 8, 9). In striking contrast, CFU-GM from Mx1-Cre;Nras^G12D/− and Mx1-Cre;Nras^G12D/+ mice displayed normal cytokine responses (Fig. 1D). Phospho-flow cytometric analysis of Lin⁻/cKit⁺/CD105⁻/CD34⁺ bone marrow cells, which are highly enriched for myeloid progenitors (Supplementary Fig. S2C), revealed elevated basal levels of phosphorylated extracellular signal-regulated kinase (pERK) and enhanced responsiveness to GM-CSF in Nras^G12D/G12D cells compared with heterozygous and hemizygous Nras^G12D-mutant cells (Fig. 1E). In summary, phenotypic, functional, and biochemical analyses of age and strain-matched mice indicate that WT Nras lacks tumor suppressor activity in the hematopoietic lineage.

To determine whether exogenous WT Nras expression might antagonize the abnormal growth of Nras^G12D/G12D cells, we infected Mx1-Cre;Nras^G12D/G12D bone marrow with murine stem cell virus (MSCV) vectors encoding N-terminal GFP fused to WT N-Ras (N-Ras^WT), WT K-Ras (K-Ras^WT), or dominant negative N-Ras (N-Ras^N17; ref. 9). After sorting to isolate GFP-positive (GFP⁺) cells, CFU-GM colony growth was assayed in the presence of GM-CSF. Mx1-Cre;Nras^G12D/G12D bone marrow formed significantly more CFU-GM colonies than heterozygous or hemizygous Nras cells (Fig. 2A). Expressing N-Ras^WT or K-Ras^WT had no effect on CFU-GM colony growth from Nras^G12D/G12D bone marrow, whereas dominant negative N-Ras^N17 reduced growth by more than twofold (Fig. 2A).

To further investigate whether WT N-Ras interferes with oncogenic N-Ras^G12D–induced myeloid transformation, we infected WT fetal liver cells with MSCV vectors expressing N-terminal mCherry-tagged N-Ras^G12D in combination with GFP, GFP-tagged N-Ras^WT, or GFP-tagged K-Ras^WT. Flow cytometry and Western blotting revealed an equivalent increase in Ras protein levels in cotransduced cells (Fig. 2B and C). Importantly, neither WT Ras isoform suppressed the aberrant pattern of cytokine-independent CFU-GM progenitor growth induced by exogenous N-Ras^G12D expression (Fig. 2D).

We next assessed the functional and biochemical consequences of varying Nras oncogene dosage by infecting WT fetal hematopoietic cells with viruses encoding GFP-tagged N-Ras^G12D and sorting for different levels of GFP expression (Fig. 3A). As expected, increasing GFP intensity correlated with higher N-Ras protein levels (Fig. 3B). Progenitors expressing the highest levels of N-Ras^G12D showed cytokine-independent CFU-GM colony growth with a threshold level of expression required for myeloid transformation (Fig. 3B and C).

We also interrogated Ras signaling in transduced GFP⁺ bone marrow–derived macrophages that were first deprived of cytokines and serum and then stimulated with GM-CSF (10). Under these conditions, AKT, ERK, and S6 were not phosphorylated in starved macrophages infected with the empty pMIG vector, and these cells responded robustly to GM-CSF stimulation (Fig. 3D). Macrophages expressing low levels of N-Ras^G12D showed a modest increase in basal pAKT, pERK, and pS6, which were augmented by cytokine stimulation. In contrast, starved cells expressing the highest levels of N-Ras^G12D exhibited a further increase in basal levels of all three phospho proteins, but were unresponsive to GM-CSF (Fig. 3D). Thus, Akt and Erk activation are uncoupled from cytokine stimulation in primary myeloid cells expressing high levels of oncogenic N-Ras.

Treatment with the potent and selective MAP-ERK kinase (MEK) inhibitor PD0325901 restores a normal pattern of hematopoiesis in Kras- and Nf1-mutant mice with myeloproliferative neoplasm (11, 12). We therefore asked whether endogenous Nras^G12D gene dosage influences the sensitivity of bone marrow–derived macrophages to MEK inhibition. Macrophages grown directly from the bone marrows of Mx1-Cre;Nras^G12D/+ and Mx1-Cre;Nras^G12D/G12D mice expanded to a similar extent in the presence of a saturating concentration macrophage colony-stimulating factor (M-CSF; Fig. 3E). Remarkably, low concentrations of PD0325901 selectively reduced the growth of homozygous Nras-mutant macrophages (Fig. 3E) despite similar basal levels of ERK activation and sensitivity to inhibition by MEK inhibitor treatment (Supplementary Fig. S3A and S3B).

Nras^G12D expression cooperated with the MOL4070LTR retrovirus to efficiently induce AML in Mx1-Cre;Nras^G12D/+ mice that recapitulates morphologic and genetic features of human Nras-mutant AMLs (4). Somatic loss of the WT Nras allele occurs in many of these leukemias (Supplementary Fig. S4A; ref. 4). Importantly, however, real-time quantitative PCR analysis showed that Nras expression is normal or increased in AML blasts (Fig. 4A). Interestingly, Kras expression was reduced to levels below that in WT bone marrow (Supplementary Fig. S4B). Consistent with these data, Western blot analysis revealed elevated N-Ras protein levels in AML blasts, including Nras^G12D leukemias with somatic loss of the normal Nras allele (Fig. 4B). We conducted FISH to investigate the mechanism underlying somatic Nras inactivation in leukemias with loss of constitutional heterozygosity (LOH) and also used TaqMan PCR to assess Nras copy number. Both analyses supported somatic uniparental disomy (UPD) with duplication of the oncogenic Nras^G12D allele as the genetic basis underlying loss of the WT Nras allele in leukemias with LOH (Fig. 4C and D). Together, these studies indicate that genetic and transcriptional mechanisms converge to augment oncogenic N-Ras^G12D expression in AML.

To address the broad relevance of increased RAS oncogene expression in human cancer, we queried NRAS/KRAS mutational status and corresponding expression data across 957 cancer cell lines from different tissues (13). Cancer cell lines with NRAS mutations showed a highly significant increase in NRAS expression (Fig. 4E). Similarly, lines with KRAS mutations expressed more KRAS transcript on average. Furthermore, cancer cell lines with NRAS mutations showed a significant reduction in KRAS expression, whereas those with KRAS mutations downregulated NRAS (Fig. 4E). Similar trends were observed when the hematopoietic cell lines in this large collection were analyzed separately (Supplementary Fig. S5A). Gene expression data from a well-annotated collection of human AMLs revealed elevated NRAS expression compared with normal CD34⁺ progenitors in leukemias with and without Nras mutations (Supplementary Fig. S5B).

LOH is common in human cancer and classically represents a genetic “second hit” that results in homozygous inactivation of tumor suppressor genes (TSG) that negatively regulate cell growth through diverse mechanisms. Given this, the finding of frequent somatic oncogenic Ras mutations and loss of the corresponding WT allele in mouse cancers induced by chemical carcinogenesis raised the possibility that normal Ras proteins might also restrain malignant growth (2). Indeed, subsequent experiments showing that germline inactivation of one Ras allele greatly increased the incidence and biologic aggressiveness of genotoxin-induced skin and lung carcinoma supported this idea (5, 14). Though provocative, it has proven difficult to reconcile these genetic data with the dominant transforming ability of oncogenic Ras proteins and their biochemical properties. Although normal Ras genes may function to suppress tumorigenesis in some cell lineages, loss of the normal Ras allele may reflect selective pressure for cancers to increase oncogene dosage as they evolve. This idea is compatible with data showing that spontaneous transformation in primary Hras-mutant fibroblasts is associated with amplification of the mutant allele (15) and with frequent copy number gains, mutant allele–specific imbalance, and overexpression of oncogenic KRAS in human cancer cell lines (16). Interestingly, oncogenic Hras amplification is an early event in murine skin carcinogenesis models and may be associated with somatic UPD (2, 17).

Somatic NRAS mutations are common in hematologic malignancies, and Mx1-Cre;Nras^G12D mice provide a tractable and genetically accurate system for interrogating the putative TSG activity of WT Nras in early (myeloproliferative neoplasm) and late (AML) stage cancers that do not rely on chemical carcinogenesis. Our extensive genetic and functional analysis indicates that WT Nras lacks tumor suppressor activity in vitro and in vivo, and identifies increased oncogene expression as the major “driver” of aberrant growth. Consistent with these data, oncogenic NRAS mutations with UPD have been reported in human AML (18). A JMML with an NRAS mutation that acquired UPD after evolution to AML suggests a role of increased oncogenic NRAS dosage in disease progression (19). The pattern of Ras gene expression in murine Nras-mutant leukemias and in human cancer cell lines with oncogenic RAS mutations also supports the existence of selective pressure to amplify oncogenic signaling by increasing oncoprotein levels while simultaneously downregulating normal Ras.

The dominant role of mutant Ras amplification is consistent with data showing that expressing oncogenic K-Ras^G12D and N-Ras^G12D from their endogenous genetic loci results in remarkably little activation of canonical effectors in primary cells, which likely reflects the inhibitory effects of potent cellular feedback responses (20, 21). The dramatic effects of modulating N-Ras^G12D protein levels on basal activation of Ras effectors and on cytokine responses further suggest that cancer cells must titrate an “optimal” level of pathway activation to overcome negative feedback inhibition without evoking cell-cycle arrest or senescence.

The idea that normal Ras proteins might antagonize the transforming properties of their oncogenic counterparts suggested potential therapeutic strategies. Although this possibility remains viable in tissues such as skin and lung where normal RAS genes are likely to function as tumor suppressors, our data strongly argue that restoring or enhancing normal Ras expression will be ineffective in leukemias with oncogenic NRAS mutations. They also support revisiting this general question in other cancers, particularly given the broad pattern of elevated NRAS or KRAS expression in human cancer cell lines with mutations in each gene. Furthermore, as cancer cells amplify oncogenic signaling to optimize growth, they may become more dependent upon (“addicted to”) these activated pathways, which might be exploited therapeutically. The enhanced dependence of Nras^G12D/G12D macrophages on MEK supports this idea. Similarly, somatic UPD is common in AML with oncogenic FLT3 mutations and is associated with resistance to conventional anticancer agents (22). However, these leukemias are more sensitive to FLT3 inhibitors than AMLs without UPD (23). Thus, selective pressure favoring the outgrowth of clones with increased oncogenic RAS gene expression may also render them more susceptible to inhibitors of critical effector pathways.

Nras^2lox2 mice were generated by inserting loxP sites onto each side of exon 2 of the endogenous Nras locus in V26.2 C57BL/6 embryonic stem cells (24). A Frt-flanked Neo resistance cassette was also inserted into intron 1 (Supplementary Fig. S1A). After germline transmission of the targeted allele, the Neo resistance cassette was removed by crossing to an Flp deleter strain (Jackson Laboratory). Mx1-Cre;Nras^G12D mice were intercrossed with Nras^2lox2 mice to generate Mx1-Cre;Nras^G12D/G12D, Mx1-Cre;Nras^G12D/+, and Mx1-Cre;Nras^G12D/2lox2 mice. All mice received a single intraperitoneal injection of poly-I/poly-C (250 μg) at 3 weeks of age to activate Mx1-Cre expression (4). Mice were euthanized at 6 months of age to assess disease. Pathologic examinations were conducted as previously described (4).

Hematopoietic progenitor assays and flow cytometric analyses were conducted as previously described (4, 10).

Nras^WT, Kras^WT, Nras^N17, and Nras^G12D alleles containing either N-terminal GFP or N-terminal mCherry markers were cloned into the MSCV with expression driven by the internal ribosomal entry site. Retrovirally transduced E14.5 fetal liver cells from C57Bl/6 mice were sorted to isolate GFP⁺ cells, which were plated in methylcellulose medium to assess CFU-GM growth as described previously (4).

Biochemical analyses were conducted on cultured macrophages that were differentiated from transduced GFP⁺ fetal liver cells in 50 ng/mL M-CSF as described previously (10). Quantitative effects of PD0325901 on macrophage ERK signaling were determined by imaging and quantifying blots on an Odyssey imager

A labeled BAC probe containing the mouse Nras gene (RP23- 280E21; 150 kb) was labeled with 5-(3-aminoallyl)-dUTP by nick translation, followed by chemical labeling with amine-reactive Alexa Fluor 488 using the Ares DNA labeling kit. An 8-kb genomic probe containing the Nras gene was labeled with biotin-dUTP by nick translation and detected with streptavidin conjugated with Alexa Fluor 568. FISH was conducted as described previously (25). Cells were counterstained with 4, 6 diamidino-2-phenylindole-dihydrochloride. A minimum of 200 interphase nuclei and 10 metaphase cells were scored for each sample.

RNA purification and quantitative PCR analyses of primary AML cells that were generated by infecting Nras^G12D/+ mice with the MOL4070LTR retrovirus were conducted as previously described (4).

Genomic DNA from Nras-mutant AMLs was purified using a Qiagen RNEasy Mini kit. A total of 10 ng of DNA was used as a template for quantitative PCR experiments. The sequence for murine transferrin receptor (mTfrc) was used to normalize total amounts of DNA. The premixed probe and primer assay mixture used to quantify total amounts of genomic Nras were purchased from Applied Biosystems.

A total of 8 × 10⁵ bone marrow-derived macrophages from Mx1-Cre;Nras^G12D/G12D or Mx1-Cre;Nras^G12D/+ mice were plated in triplicate in 12-well plates in the presence of 10 ng/mL M-CSF and varying doses of PD0325901. Total numbers of viable cells were counted using the Beckman Coulter Vi-Cell XR at day 5.

NRAS and KRAS expression in cancer cell lines was extracted from gene-centric RMA-normalized mRNA expression data downloaded from Cancer Cell Line Encyclopedia (13). KRAS and NRAS mutational status was determined by hybrid capture and Oncomap assays. P values were calculated by two-tailed Student t test assuming unequal variance. To further examine NRAS and KRAS expression in human leukemia, we conducted microarray-based gene expression profiling data of pediatric AML and normal CD34⁺ samples generated using Affymetrix U133A microarrays (Affymetrix) according to the manufacturer's instructions, with data processed using MAS5 normalization and log2 transformation (26). This cohort comprised 108 samples including CD34⁺ bone marrow cells (n = 4), and included 74 without an NRAS mutation and 30 with a mutation. Primary data are available through http://www.ncbi.nlm.nih.gov/geo/ accession numbers GSE43176 (AML samples) and GSE33315 (CD34⁺ bone marrow).

No experiments were carried out on any cell lines to generate these data. All mouse experiments were approved by the Institutional Animal Care and Use Committee of the University of California, San Francisco.

No potential conflicts of interest were disclosed.

Conception and design: J. Xu, K. Shannon

Development of methodology: J. Xu, T. Jacks

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J. Xu, K.M. Haigis, A.J. Firestone, Q. Li, J. Downing, M.M. Le Beau

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Xu, A.J. Firestone, M.E. McNerney, Q. Li, S.-C. Chen, J. Nakitandwe, J. Downing, M.M. Le Beau, K. Shannon

Writing, review, and/or revision of the manuscript: J. Xu, J. Nakitandwe, J. Downing, M.M. Le Beau, K. Shannon

Study supervision: M.M. Le Beau, K. Shannon

Performed the FISH analysis of the samples: E. Davis

The authors thank J. Leopold (Pfizer, Inc.) for PD0325901 and A. Balmain and F. McCormick for suggestions. The authors also thank T. Huang and E. Hwang for assistance in critical experiments.

This work was supported by NIH grants R37CA72614, P01CA40046, and K08CA134649, a Specialized Center of Research award from the Leukemia and Lymphoma Society (LLS 7019-04), the American Lebanese Syrian Associated Charities (ALSAC) of St. Jude Children's Research Hospital, an American Cancer Society Fellowship (to J. Xu), and a Damon Runyon Cancer Research Foundation Fellowship (DRG-2149-13; to A.J. Firestone).