c-Met inhibition in a HOXA9/Meis1 model of CN-AML



Background: Hematopoiesis is a paradigm for developmental processes, hierarchically organized, with stem cells at its origin. Hematopoietic stem cells (HSCs) replenish progenitor and precursor cells of multiple lineages, which normally differentiate into short-lived mature circulating cells. Hematopoiesis has provided insight into the molecular basis of tissue homeostasis and malignancy. Malignant hematopoiesis, in particular acute myeloid leukemia (AML), results from impaired development or differentiation of HSCs and progenitors. Co-overexpression of HOX and TALE genes, particularly the HOXA cluster and MEIS1, is associated with AML. Clinically relevant models of AML are required to advance drug development for an aging patient cohort. Results: Molecular analysis identified altered gene, microRNA, and protein expression in HOXA9/Meis1 leukemic bone marrow compared to normal controls. A candidate drug screen identified the c-Met inhibitor SU11274 for further analysis. Altered cell cycle status, apoptosis, differentiation, and impaired colony formation were shown for SU11274 in AML cell lines and primary leukemic bone marrow. Conclusions: The clonal HOXA9/Meis1 AML model is amenable to drug screening analysis. The data presented indicate that human AML cells respond in a similar manner to the HOXA9/Meis1 cells, indicating pre-clinical relevance of the mouse model. Developmental Dynamics 243:172–181, 2014. © 2013 Wiley Periodicals, Inc.


Constant blood cell production is maintained by self-renewing hematopoietic stem cells (HSCs) (reviewed by Orkin and Zon, 2008). As hematopoietic cells proceed through the hierarchy, they lose their stem-cell capacity, and gradually their lineage potential until they become distinct mature functional cells. Hematopoiesis is a highly organised system, tightly regulated by complex cell-niche (Wilson et al., 2004; Robb, 2007; Felli et al., 2010) and molecular interactions. The Class I homeobox (HOX) transcription factors that regulate normal development (Duboule and Dolle, 1989; Tschopp et al., 2009) are highly expressed in self-renewing HSC and progenitor (HSPC) pools (Pineault, et al., 2002; Dickson et al., 2009; Lebert et al., 2010) where malignant hematopoiesis originates (Lapidot et al., 1994; Bonnet and Dick, 1997; Reya et al., 2001; Hope et al., 2004). HOX and TALE (three-amino-acid-loop-extension) co-factors are particularly associated with leukemia (Argiropoulos and Humphries, 2007) and for acute myeloid leukemia (AML) HOXA9 was reported as the top independent prognostic indicator of outcome (Golub et al., 1999).

AML patients are traditionally stratified into three risk groups: favourable, intermediate, and unfavourable based on cytogenetics (karyotype) and the presence or absence of chromosomal anomalies, e.g., translocations. A large proportion of AML patients with intermediate risk do not present with gross chromosomal anomalies and are subclassified as cytogenetically normal AML (CN-AML). Increased HOXA9 and MEIS1 expression is particularly associated with CN-AML (Bullinger et al., 2004; Grubach et al., 2008). Research into the potential for HOX and TALE genes to act as biomarkers of AML is currently ongoing (Dickson et al., 2013; Li et al., 2012) and correlations with response to chemotherapy have previously been reported (Nakamura et al., 1996; Afonja et al., 2000; Grubach et al., 2008; Zangenberg et al., 2009). Progress has also been made in identifying the mutational landscape of CN-AML (reviewed by Martinelli et al., 2012), which may lead to better therapies for these patients.

Standard-of-care induction chemotherapy is often too severe for elderly patients and not subtype specific. It is imperative that better tolerated effective treatments and relevant pre-clinical models are developed. Over-expression of individual HOXA genes alone or in combination with Meis1 results in leukemia (Thorsteinsdottir et al., 2002; Bach et al., 2010). Such retroviral transduction/transplantation mouse models recapitulate many facets of human malignancies and provide essential tools to interrogate molecular mechanisms of disease progression. Restriction of oncogene expression to values commonly seen in patient samples is necessary for the generation of more clinically relevant models.

We have developed a mouse model (A9M) that over-expresses both HOXA9 and Meis1 genes to levels reflective of CN-AML patients. Comparative analysis of patient microarray data (MILE) and gene expression data from the A9M model using connectivity mapping identified common potential therapies, indicating that the A9M model recapitulates CN-AML at the level of drug interaction (Ramsey et al., 2013). The focus of this investigation is to determine whether leukemia initiation by co-overexpression of HOXA9 and Meis1 results in differential gene, microRNA (miR), and protein expression that may help identify additional pathways for therapeutic testing. A selection of candidate drugs was used in a functional screen to examine anti-leukemia-specific activity and the cMet inhibitor SU11274 was selected as an exemplar for further studies extended to human AML cells.


HOXA9/Meis1 Model of Cytogenetically Normal Acute Myeloid Leukemia

Retroviral transduction of HOXA9-ires-MEIS1 (A9M) into fetal liver HSPCs has previously been used to develop aggressive leukemia in mice (Wilhelm et al., 2011; Ramsey et al., 2013). Use of the retroviral system allowed for the generation of a clonal murine leukemia (A9M) via the overexpression of HOXA9 and Meis1. Primary bone marrow cells from the A9M leukemia were subsequently used in downstream applications including: gene, microRNA, and protein expression arrays; a candidate drug screen of colony formation ability; cellular assays including cell cycle and apoptosis (Fig. 1A). In addition, microarray data from normal bone marrow and AML patients (MILE Study GSE13204) were initially mined to identify enrichment of HOX and TALE expression in CN-AML patients by hierarchical clustering and human AML cell lines used to examine drug sensitivity (Fig. 1B). Data from the mouse gene expression arrays and AML microarrays were further used for connectivity mapping (sscMap) analysis.

Figure 1.

Experimental design and strategy. A: Schematic representation of the generation and analysis of the A9M mouse model. B: Analysis of patient data and subsequent experimental procedures of leukemia cells and cell lines. *Data from the mouse gene expression arrays and AML microarrays were further used for connectivity mapping (sscMap) analysis.

Differential Gene Expression in A9M Leukemia

Q-PCR profile arrays were employed as a high throughput method to assess pathways associated with the A9M leukemic phenotype (stem cell pluripotency, immune and transcription factor). A total of 860 genes were analysed from cDNA obtained from total bone marrow of leukemic mice (A9M) or normal bone marrow (NBM) (Fig. 2A and data not shown). A significant change in gene expression was defined as a change of endogenous control-corrected Ct value greater or equal to two (ΔcCT ≥ 2). Almost half (47%) of the genes examined were differentially expressed in the A9M leukemia compared to NBM controls, with 45% of genes downregulated compared to an upregulation of only 2% (Fig. 2A). A similar pattern was observed in the pluripotent stem cell array wherein 21% of genes were expressed lower in A9M leukemia compared to NBM controls and only 4% of the genes investigated were expressed higher (data not shown). The genes that were expressed higher included cluster differentiation markers (CD28, CD34), cell cycle regulators (Nes, Sycp3), and epigenetic modifiers (Xist) (for full list see Ramsey et al., 2013). The high percentage of down-regulated genes across all the arrays suggested that, when constitutively over-expressed, Hoxa9 and Meis1 function to repress the transcription of a wide variety of genes. Alternatively, HoxA9 and Meis1 may act upon negative regulators of gene expression, such as microRNAs (miRs).

Figure 2.

Differential gene and miR expression in A9M leukemias. A: Scatter plot depicting the combined change in expression (ΔCt) of immune and pluripotent stem cell associated genes between NBM and A9M BM profiled using TaqMan® low-density array platforms. The majority of genes show decreased expression. Values represent the mean of four biological replicates from duplicate assays. B: Scatter plot depicting altered expression of miR for A9M BM compared to NBM analysed using TaqMan® Megaplex™ miR arrays. Of the 48 mIRs that showed differential expression, 71% were upregulated compared to 29% downregulated. miR IDs are plotted against ΔCt corrected to endogenous control MammU6. Mean values from four biological replicates analysed in duplicate are plotted. C: Waterfall plot depicting the 48 differentially expressed miRs. Fold change (calculated as 2ΔΔCT) is corrected to Mamm U6 RNA. Seven of the miRs were identified as having a potential role in hematopoiesis and were further validated by individual assays (*).D: A comparative histogram plot of changes in CT values (A9M v NBM) from the individual assay or Megaplex™ Array values for seven hematopoietic-related miRs. E: A cross-reference table of candidate miRs (n=6) with genes identified as downregulated in the A9M leukemia. The association was identified in silico by the miRWalk target prediction database.

Altered microRNA Expression in A9M Leukemia

miRs represent a class of non-coding RNAs with the capacity to negatively regulate gene expression by controlling the transition between mRNA and protein formation. Given that the A9M phenotype was associated with decreased expression of a panel of genes, primary bone marrow samples from leukemic (A9M) and normal (NBM) mice were analysed for changes in miR expression (Fig. 2B). When examined by ΔcCt, 12% (48/380) of miRs showed differential changes in expression and, interestingly, 71% (34/48) of these displayed upregulation in A9M samples compared to NBM controls with a range of 4-fold for miR-181c and 361-fold for miR-329 (Fig. 2C). miR-10 and miR-196 are known to be embedded within the Hox network (Yekta et al., 2008) and, interestingly, miR-196b demonstrated reduced expression (10-fold) in the A9M model. A subset of the candidate miRs (7/48), validated by individual assays (Fig. 2D) and reported to be involved in hematopoietic or leukemic development (Vasilatou et al., 2010; Narducci et al., 2013) was cross-referenced in silico against the A9M regulated gene signature, using the miR/Gene target prediction software miRWalk: http://www.umm.uni-heidelberg.de/apps/zmf/mirwalk/). Significant correlation (P < 0.01) was found between the altered miRs (except miR29b) and deregulated gene signature profiles with P values in some cases observed at <0.0006 (miR-221 targeting Gbx2). miR-221/222 had the highest number of target genes within the validated list with almost half of the correlations (9/20) predicted to be affected by this pair of miRs (Fig. 2E). Upregulation of miR-221/222 and miR-329 is associated with angiogenesis and these data indicate that future strategies to target knockdown of particular mIRs may be therapeutic in subtypes of AML.

A9M Leukemia Demonstrates Activation of Angiogenesis-Related Pathways

To further investigate the effect of HOXA9/Meis1 overexpression at the protein level, an angiogenesis proteome profiling array was examined. Increased expression of multiple analytes of angiogenesis-related pathways was observed following incubation of the capture antibodies with A9M protein lysates compared to NBM controls (Fig. 3A). The three reference positive controls (R1+, R2+, and R3+) were used to align the arrays and demonstrate equivalent loading. Exposed X-ray films were scanned and pixel densities obtained to quantify the level of protein expression over a wide dynamic range. Significant increased expression of 22/53 angiogenesis-related proteins was demonstrated in A9M cells compared to NBM controls (Fig. 3B). Levels of osteopontin, a known HOXA9 target, increased (over 6.5-fold) in the leukemia model, indicating congruence with other approaches. Of particular interest, hepatocyte growth factor (HGF), the ligand for the c-Met receptor that has been reported to be secreted from human leukemia cells (Gohda et al., 1995), was also significantly increased (> 10-fold) in the A9M model.

Figure 3.

Differential expression of angiogenic analytes in A9M leukemia. A: Representative scanned image (from n=3) of a mouse angiogenesis antibody array depicting differential expression of a number of analytes in A9M compared to NBM protein lysates. Reference positive controls (R1+, R2+, and R3+) confirm comparable loading and allow for template overlay to identify the proteins involved. Any protein/detection antibody complex present is bound by its cognate immobilized capture antibody on the membrane and visualised using chemiluminescence. B: Bar graph showing statistically significant changes in optical density levels in total cell lysates examined in duplicate. Graph shows mean values ± S.E.M (n=3 independent experiments; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001).

Candidate Drug Screen in A9M Cells

Combinatorial sscMap analysis of gene signatures from both the A9M model and patient samples (Ramsey et al., 2013) identified classes of drugs including antibiotics, anti-inflammatory agents, and histone deacetylase inhibitors, which may be therapeutic in the CN-AML subtype (Fig. 4A). To compliment this list, a search of the literature identified potential inhibitors of HOX-TALE or leukemic stem cell activity. A functional screen approach was taken, using methylcellulose culture of normal and A9M primary cells, to examine perturbation of colony-forming ability as the readout. In the primary screen, a change of the methylcellulose media color from red/orange to yellow due to pH was used to select for high cell number/metabolic activity (Fig. 4B). In the secondary screen, colonies were visualized by microscopy and images captured for analysis. The c-Met inhibitor SU11274 is presented as an exemplar of this approach (Fig. 4C). Reduced colony formation in the A9M cells was shown with several of the candidate drugs, some of which were further investigated in liquid culture. Reduced viability and increased apoptosis (caspase activation) was demonstrated for several of the candidates (Fig. 4D).

Figure 4.

Candidate screen and SU11274 inhibition of A9M colony formation. A: Venn diagram demonstrating overlap of 3/5 of the top drug connections identified by sscMap for A9M and MILE 13/CN-AML patient samples. B: Representative digital image of 96-well plate format screen of candidate drug therapies for A9M and NBM cells. Yellow wells indicate high metabolic activity resulting in exhaustion of media components resulting in an acidic pH, orange wells are indicative of low metabolic activity due to impaired cell/colony growth. C: Representative images of colony formation in NBM and A9M samples following 24-hr exposure to DMSO (0.01%), 1 μM or 10 μM SU11274. Colonies were stained with 1 mg/ml p-iodonitrotetrazolium violet for 16 hr and images captured at 4X using an inverted microscope (CKX41) with attached digital camera (E620) both Olympus, Essex, UK. D:) A table summarizing the activity of a set of candidate drugs and small molecule inhibitors, with known mechanism of actions on A9M colony formation, viability and apoptosis/caspase activation (-, not examined).

SU11274 Affects A9M and Human AML Cell Dynamics

The c-Met inhibitor, SU11274, was further assessed over a time course in human AML cell lines in liquid culture. A time and dose response was demonstrated for both OCI-AML3 and U937 cells with a significant reduction in cell number, as determined by a reduction in [ATP] (Fig. 5A) corresponding to a significant increase in caspase activity (Fig. 5B). Interestingly, the U937 cell line appeared to be more sensitive to the treatment than OCI-AML3 cells, which may reflect the more differentiated status of the cells. The increased caspase activity, indicative of apoptosis, was supported by cell cycle analysis and the increased number of sub G0/G1 events following SU11274-treatment. In contrast, OCI-AML3 cells showed minimal effects to low dose SU11274 (2 μM); however, the higher dose treatment (10 μM) resulted in G2/M accumulation of the cells (Fig. 5C). U937 and OCI-AML3 cells were further investigated by morphological analysis and 10 μM SU11274 treatment resulted in increased differentiation with concomitant loss of blast cells (Fig. 5D) consistent with the loss of A9M clonogenic ability previously shown.

Figure 5.

c-Met inhibition alters cell viability, apoptosis, cell cycle dynamics and differentiation of human AML cells. A: Bar graphs showing decreased viability in OCI-AML3 and U937 cells following incubation with SU11274 at the indicated dosage compared to 0.01% DMSO vehicle control for 24–72 hr. Viability is measured as relative fluorescence units based on [ATP] via CellTiter-Glo®. Mean values ± S.E.M. are plotted. **P ≤ 0.01, ***P ≤ 0.001. B: Bar graphs showing increased apoptosis in OCI-AML3 and U937 following incubation with SU11274 at the indicated dosage compared to 0.01% DMSO vehicle control for 24–72 hr. Apoptosis related caspase3/7 activity is measured as relative luminescence units using Caspase-Glo® 3/7. Mean values ± S.E.M. are plotted. **P ≤ 0.01, ***P ≤ 0.001. C: Representative flow cytometry histogram plots of DNA content (Propidium Iodide staining) for OCI-AML3 and U937 cells following SU11274 treatment at the indicated dosage for 72 hr compared to 0.01% DMSO controls. D: Representative morphology images of OCI-AML3 or U937 cells treated with 10 μM SU11274 or 0.01% DMSO. DMSO-treated cells (left panels) demonstrate a high number of blast cells (identified by high nuclear:cytoplasm) are identified by the arrows. SU11274-treated cells (right panels) demonstrate a high number of differentiated cells (low nuclear:cytoplasm and more defined nuclear shape, including band neutrophils) are identified by arrows (scale bar = 20 μm).


The ultimate aim of AML therapies is to specifically target leukemia-repopulating cells while sparing non-malignant HSCs. The developmental role of HOX genes has been clear for some time; additionally, the collaboration of the oncogenes HOXA9 and MEIS1 in CN-AML is well defined (Lawrence et al., 1999; Rozovskaia et al., 2001; Morgado et al., 2007; Ramsey et al., 2013). The restricted levels of HOXA9 and Meis1 gene expression in A9M bone marrow are similar to those seen in equivalent patient samples. We have previously shown overlap between the A9M model and CN-AML patient data, at the level of drug interaction (Ramsey et al., 2013) using sscMap analysis. Herein, we investigate further molecular mechanisms of the A9M-induced leukemia and identify a potential role for miR regulation of gene expression and angiogenesis-related pathways in disease progression.

miRs are a class of small noncoding RNAs (∼22 nucleotides) that silence target genes by complimentary base-pairing, usually in the 3′-untranslated region, which results in translational repression or target degradation. The complimentary overall increase in miR and decrease in gene expression in the A9M model is indicative of the high order of regulation and complexity within the HOX-TALE network and leukemia. Whether the increase in miR expression is an inhibitory cellular response to or a direct result of the oncogenic insult requires further study. However, it is thought-provoking that HOX and/or TALE transcription factors may directly regulate expression of a subset of miRs, which in turn regulates key pathways, resulting in the malignant phenotype. Indeed, miR-221/222 has been shown to be directly regulated by a HOX:TALE dimer (HOXB7/PBX2) in malignant melanoma (Errico et al., 2013), and reduced expression of miR-196b, also seen in the A9M model, results in enhanced expression of HOXA9 in chronic myeloid leukemia (Liu et al., 2013).

Increasing evidence suggests particular roles for miRs in cancer, and miR-150, which was downregulated in the A9M model, is reported as the most significantly and consistently downregulated miR in AML (reviewed by He et al., 2013). Of the A9M-associated upregulated miRs, miR-329 was the most highly expressed; along with miR-221/222, miR-329 is associated with angiogenesis, one of the hallmarks of cancer. A direct link between angiogenesis and hematopoietic growth factors is well established (Bikfalvi and Han, 1994) and several anti-angiogenic agents, including Sorafenib, are in clinical trials for AML (Ravandi et al., 2013, 2010). A protein array screen of the A9M leukemia demonstrated significant increased expression of angiogenesis-related analytes with known hematopoietic activity, e.g., VEGF and FGF2 (Pick et al., 2007) and IGFBPs, previously shown to be regulated by HOX proteins (Allander et al., 1995).

The aberrant upregulation of HGF/c-Met in A9M primary cells was of particular interest as a limited pathway that may be targeted by small molecule inhibitors. HGF binding to the c-Met receptor activates the PI3K and MAPK pathways (Eder et al., 2009). Abnormal activation of these pathways either by overexpression of the ligand (HGF) or overexpression or mutation of the receptor (c-Met) is involved in various cancers causing increased cell proliferation, survival, and metastasis (Peruzzi and Bottaro, 2006; Ma et al., 2005). SU11274 binds to the ATP-binding site of c-Met, inhibiting its function and preventing activation of downstream pathways (Berthou et al., 2004). SU11724 has been shown to reduce cell viability in vitro in a variety of cancers including glioblastoma (Huang et al., 2007), lung cancer (Ma et al., 2007), and melanoma (Kenessey et al., 2010). For these reasons, SU11274 was included with a group of candidate drugs, selected based on sscMap analysis from A9M leukemia and CN-AML patient samples.

We previously showed that Entinostat, an epigenetic modifying agent (Histone Deacetylase inhibitor), resulted in loss of leukemia maintenance in the A9M model (Ramsey et al., 2013). To extend these findings, six other epigenetic modifiers—DZnep, JQ1, Vorinostat, Romidepsin, Trichostatin-A, and Panobinostat—were included in the screen. In addition, pro/anti-inflammatories (Dinoprostone/PGE2), ionophores (Salinomycin, Nigericin), DNA-damaging agent (Etoposide), and an insecticide (Abamectin) were screened for anti-leukemic activity within the A9M model. The functional screen based on colony-forming ability identified SU11274 as a candidate drug for A9M-induced leukemia as compared to known effectors (TRAIL) and standard-of-care therapies (Cytarabine and Daunorubicin). Due to the drug overlap between A9M leukemia and CN-AML patients, identified by sscMap, the anti-leukemic effect of SU11274 was further investigated in human AML cell lines. Aberrant expression of HGF provides a survival mechanism in AML cell lines (Kentsis et al., 2012). C-Met inhibition resulted in decreased cell number, increased apoptosis, altered cell cycle dynamics, and marked differentiation in both cell lines. The OCI-AML3 cells appeared more resistant than the U937 cells to SU11274. Further investigation utilizing SU11274 or second-generation c-Met inhibitors alone or in combination with standard-of-care therapies may improve its efficacy in the treatment of CN-AML.

Together the data support the use of a clonal HOXA9/Meis1 mouse model of CN-AML for molecular analysis and pre-clinical drug screening. Although angiogenesis is a hallmark of cancer, it has mostly been studied in the solid tumor setting. The co-dependence of leukemia and endothelium within defined niches is currently of particular interest. With relevant preclinical models and access to second- or third-generation drugs, we can begin to develop appropriate combination treatments and identify accessible molecular markers for translational studies. The drug responses observed herein support the targeting of molecular mechanisms and pathways in the leukemia setting for the application of novel therapies.



Congenic donor CD45.1+ (C57Bl6/Pep3b) or (C57Bl/6-Ly5.1) and recipient CD45.2+ (C57Bl/6J) mice were bred and maintained in a specific pathogen-free (SPF) animal facility (BRU-QUB). Animal handling followed the guidelines of the UK Animals (Scientific Procedures) Act 1986. Experimental procedures were approved by the Ethical Review Committee (ERC) for Animal Research, Queen's University Belfast.

Bone Marrow Isolation

Mice were sacrificed by increasing CO2 concentration over a period of 5 min in accordance with Scientific Procedures Act 1986. Autopsies were performed for removal of femur and tibia bones. Bones were flushed with DMEM supplemented with 2% FCS. Cells were centrifuged at 800 rpm for 10 min at 4°C and re-suspended in 1 ml of DMEM plus 2% FCS. Counts were performed using Trypan Blue exclusion method (Invitrogen, Paisley, UK).

Gene Expression Profiling

Total BM was collected by flushing femurs and tibias of control or A9M mice, total RNA isolated, and cDNAs generated as previously described (Thompson et al., 2003). For stem cell pluripotency (part number 4385363) immune response (part number 4367786) profiling, cDNA samples were prepared with TaqMan™ Universal PCR mix and (200 ng/100 μl) loaded into each of eight ports present on the predesigned 384 well cards of the TaqMan™ Gene Expression Array. QPCR was performed as per the manufacturer's protocol and analyzed using the 7900 HT Sequence Detection System, all ABI (Applied Biosystems, Foster City, CA). 18S rRNA was used as the endogenous control calibrator for subsequent analysis. For patient samples, expression profiles were generated from human genome expression arrays (HG-U133A or HG-U133 Plus 2.0: Affymetrix, Santa Clara, CA) for MILE Class13 CN-AML plus other abnormalities not 11q23 (n=315) and MILE Class 18 according to the reported classification (NCBI Gene Expression Omnibus Accession number: GSE13204).

Low-Density microRNA Taqman Array

For miRNA cDNA synthesis, RNA was reverse transcribed using the Megaplex™ primer pool (Applied Biosystems Incorporation). This system allows profiling of hundreds of miR targets selected from the Sanger miR Base v10 using stem-looped RT primer pools. Each array was matched to these pools allowing for 375 miRs and 6 controls in pool A, and 210 miRs and 6 controls in pool B. miR expression profiles were acquired using the low-density miRNA Taqman array per the manufacturer's instructions or individual assays. Each array was prepared with 1 μg total RNA requiring no pre-amplification during PCR cycling. Cycling conditions were as follows; 40 cycles of 16°C for 2 min, 40°C for 1 min, 50°C for 30 sec, followed by 85°C for 5 min. All PCR reactions were performed on the Applied Biosystems Veriti thermal cycler. Each array was analysed on the 7900 system using TLDA default thermal-cycling conditions and Mamm-U6 used as the endogenous control (Applied Biosystems).

Angiogenesis Profiler Antibody Array

To study protein expression in primary mouse BM samples, mouse angiogenesis arrays were used according to the manufacturers' instructions (R&D Systems, Abingdon, UK). Briefly, collected BM cells were solubilised in lysis buffer, then mixed for 30 min at 4°C. The lysates were centrifuged at 14,000g for 5 min prior to quantification. Membranes were blocked for 1 hr at room temperature and lysates (200 μg each) were hybridized to the array prior to incubation in ECL chemiluminescent substrate (Merck Millipore, Billerica, MA). Membranes were exposed to X-ray film and analysed for optical density using a bioimaging system encompassing the AC1 Autodarkroom (Ultra-Violet Products, Cambridge, UK), an integrated Hamamatsu IEEE 1394 digital camera (Hamamatsu City, Japan), and the associated Labworks Version 4.5 software for Windows (UVP, Upland, CA), which allows for pixel density quantification over a wide dynamic range.

Methylcellulose Screen

Fresh primary bone marrow cells were collected from A9M leukemic or age-matched normal mice and allowed to recover in expansion media for 24 hr. Cells were routinely treated with 1 nM, 100 nM, 1 μM, and 10 μM of drug in the primary screen with the exception of panobinostat (7 nm and 25 nM), DZNep (200 nM, 500 nM, and 1 μM), JQI and Romidepsin (1 nM, 10 nM, and 100 nM), Vorinostat (10 nM, 100 nM, and 1 μM), Trichostatin-A (10 nM, 100 nM, and 300 nM). Cells were treated for 24 hr in liquid culture, then transferred to MethoCult® GF M3434 media (Stem Cell Technologies, France). Methylcellulose cultures were maintained in a humidified incubator at 37°C with 5% CO2 for up to 10 days when colony formation, colony counts, and photo images were captured and analyzed. Colony staining with 1 mg/ml p-iodonitrotetrazolium violet (INT, Sigma-Aldrich, Gillingham, UK) for 16 hr enabled visualization of colonies and determination of metabolic activity. Each screen was performed in triplicate in 96 well plates (1 × 104 cells per 100 μl).

Treatment of AML Cell Lines

The well-established human AML cell lines, OCI-AML3 or U937, were cultured for up to 72 hr in RPMI 10% FCS (Life Technologies, Paisley, UK) containing SU11274 at the stated concentration or DMSO (vehicle control 0.01%). Cell viability was quantified by CellTiter-Glo® and caspase activity by Caspase-Glo® 3/7 (Promega, Madison, WI) in accordance with the manufacturer's instructions.

Flow Cytometry and Cell Cycle Analysis

Flow cytometry analyses of leukemic cell populations were performed as described (Thorsteinsdottir et al., 2002). DNA content was measured with an LSR II flow cytometer. Sub G0/G1 populations, indicative of fragmented DNA and cell damage, were quantified using FACS Diva™ (BD Biosciences, Franklin Lakes, NJ) or FlowJo software (TreeStar Inc, Ashland, OR).

Morphology Analysis

AML cells were collected on glass slides at 500 rpm for 5 min using a cytospin III cytocentrifuge and associated funnel apparatus (Shandon, Pittsburgh, PA). Gross cell morphology was assessed using modified Wright's Stain and a Hematek 1000 Slide Stainer (HemTechnologies, Lebanon, NJ).

Statistical Analysis

Two-way ANOVA or Student t-tests were performed by Graph-Pad Prism software (GraphPad software, Version 5.0, La Jolla, CA) or SPSS software package (IBM, Portsmouth, UK). For all graphs, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.


We gratefully acknowledge Nadine Mayotte, Melanie Frechette (both IRIC, Montréal), and Dr. Shu-Dong Zhang (Bioinformatics, QUB) for technical support, in addition to the staff within the Biological Resource Unit, Bioinformatics, and Flow Cytometry Cores, Queen's University Belfast. A.T. was a recipient of The American Cancer Society for Beginning Investigator Fellowship from the UICC and supported by Leukemia and Lymphoma Research UK (grant numbers 09035 and 07016) and the Northern Ireland Leukemia Research Fund (NILRF). J.B. was supported by the Leukemia and Lymphoma Society of Canada and Canadian Cancer Society Research Institute, number 20399. N.M. and L.K. were supported by Leukemia and Lymphoma Research UK and J.R. was supported by the Department for Education and Learning Northern Ireland.