SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

MicroRNAs are short ribonucleic acids (RNAs) that play an important role in many aspects of cellular biology such as differentiation and apoptosis, due to their role in the regulation of gene expression. Using microRNA microarrays, we characterized the microRNA gene expression of 27 patients with acute myeloid leukemia (AML) with normal cytogenetics, focusing on the microRNAs differentially expressed between the M1 and M5 French–American–British (FAB) subtypes. An accurate delineation of these two AML entities was observed based on the expression of 12 microRNAs. We hypothesized that these microRNAs may potentially be involved in the differentiation block of M1 blasts and consequently monocytic differentiation. Using publically available mRNA data and microRNA target prediction software, we identified several key myeloid factors that may be targeted by our candidate microRNAs. The expression changes of the candidate microRNAs during monocytic differentiation of AML cell lines treated with Vitamin D and phorbol 12-myristate 13-acetate were examined. All six candidate microRNAs were significantly down-regulated over the time course by quantitative reverse transcriptase polymerase chain reaction suggesting a link between these microRNAs and monocytic differentiation. To further characterize these microRNAs, we confirmed by luciferase assays that these microRNA target several key myeloid factors such as MAFB, IRF8, and KLF4 identifying a possible mechanism for the control of differentiation by these microRNAs. Am. J. Hematol., 2011. © 2010 Wiley-Liss, Inc.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Acute myeloid leukemia (AML) is a rapidly fatal malignancy which occurs as a consequence of the progressive accrual of genetic aberrations in myeloid progenitor cells [1, 2]. These alterations lead to a block in progenitor cell differentiation and an increase in cell proliferation [1, 2]. AML is morphologically, molecularly, and prognostically heterogeneous [1]. Cytogenetic status at AML diagnosis provides the foundation of the WHO Classification system and remains one of the most robust factors predicting patient outcome [3]. Recurrent chromosomal translations in AML such as t(15;17), inv(16), and t(8;21) are associated with a favorable outcome whereas other abnormalities such as chromosome 5 or 7 monosomies are linked to a poor outcome [4]. Patients with cytogenetically normal (CN-AML) constitute the largest subgroup of AML (∼40%) [5], possessing an intermediate prognosis. Increasingly, recurrent mutations within critical gene sequences, such as FLT3 and NPM1, are being identified and may assist in refining prognosis, particularly in the CN-AML subgroup.

A hallmark of AML is a block in differentiation whereby AML blasts display a block in their development and maturation at a specific stage. It has been proposed that the disruption of differentiation observed in AML blasts provides one of the “hits” contributing to the etiology of the malignancy in addition to uncontrolled cell cycling, the so called “two hit hypothesis” [6]. Even in AMLs that contain a routinely detectable abnormality it appears that further genetic insults or hits are required to achieve the full malignant state, as a single insult is frequently observed to be unable to induce leukemic transformation [6].

Genome wide profiling of AML using microarray technology provides a means to further characterize the cellular biology of AML [7]. Indeed, the potential is present for expression profiling to identify genes which may provide the remaining hit to induce transformation. Gene expression profiling has shown utility in the diagnosis, prognosis, and biology of AML [8]. Although the majority of previously published reports have focused on mRNA expression, the advent of microRNA expression profiling has been accompanied by an increasing number of reports indicating that microRNA expression profiles may be similarly informative [8, 9].

MicroRNAs are short noncoding RNAs that have been shown to be involved in a variety of biological processes, including development and oncogenesis, via the post-transcriptional regulation of protein coding genes [10, 11]. MicroRNAs regulate the protein levels of several hundred target mRNAs, suggesting that the abnormal expression of individual microRNAs may have profound effects on numerous pathways.

This study aims to further characterize the potential involvement of microRNAs in CN-AML cellular biology. As described, many cases of CN-AML possess only one or no identifiable molecular abnormalities and therefore are likely to require other molecular mechanisms for transformation to have occurred. We hypothesized that dysregulation of microRNAs provides one of the required hits for transformation. This study particularly focuses on their potential role in the block of differentiation. We used the FAB M1 and M5 subtypes (AML without maturation and acute monoblastic or monocytic leukemia, respectively) as a model to study myelomonopoiesis. By matching the mRNA and microRNA gene expression profiles and their interplay in M1 and M5 patients, we have identified several microRNAs that may play a role in myeloid differentiation and provide evidence that these microRNAs target key myeloid differentiation factors.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Patient samples

CN-AML (n = 27) and normal (n = 8) samples were obtained after informed consent from the Department of Hematology, St Vincent's Hospital, Sydney, Australia, or from the Australian Leukemia and Lymphoma Group Tissue Bank. Bone marrow aspirates at time of diagnosis were separated by density centrifugation and the mononuclear cells were isolated and cryopreserved. A diagnostic bone marrow blast percentage of 50% was used a threshold for inclusion in this study with enrichment of blasts by density centrifugation with Ficoll.

RNA extraction

Cells were pelleted, washed once with phosphate buffered saline (PBS), and then lysed with 1 ml of Trizol (Invitrogen, Carlsbad, CA). Total RNA was isolated according to the manufacturer's guidelines. Total RNA integrity and purity was assessed using the Eukaryote Total RNA Nano chip on a Bioanalyzer (Agilent, Santa Clara, CA). Only RNA with a RNA Integrity Number >6.5 was used in the arrays.

Exiqon microRNA microarrays

MicroRNA microarrays were performed by the Adelaide Microarray Centre Australia. The microRNA microarrays consisted of 1,488 antisense microRNA oligonucleotide probes (miRCURY LNA microRNA probe set, Cat # 208010 V8.1, Exiqon, Vedbaek, Denmark) printed in duplicate onto epoxide coated microarray slides (Corning Life Sciences, Lowell, MA). For detection on the array, 5 μg of total RNA was labeled by the ligation of a fluorescently modified RNA dimer [12]. Two-sample (dual color) competitive hybridizations were performed using Cy3 and Cy5 labeled sample pairs. Experimental samples were run in duplicate and a dye swap was performed to control for labeling efficiency. For the reference channel the total RNA from normal patient samples were pooled in equal quantities.

Hybridization was performed for 16 hours at 60°C under LifterSlips (Erie Scientific, Portsmouth, NH) in 1× Exiqon hybridization buffer in a total volume of 25 μL. Slides were placed in Corning hybridization chambers and protected from light for the 16-hour incubation. Slides were washed using dilutions of the Exiqon Wash Buffer kit as recommended by the manufacturer. Slides were scanned at 10 μm resolution with a Genepix 4000B Scanner (Molecular Devices, Sunnyvale, CA).

Differentially expressed microRNAs

The raw intensity values were extracted using the image analysis software SPOT. This was followed by data pre-processing using the Bioconductor package limma [13]. Each array was normalized using the global loess method after subtraction of background intensities. Postnormalization, the log-2 fold-change values were obtained with respect to the reference (patients with non-leukemia).

Next, we obtained the differentially expressed (DE) microRNAs under two conditions—AML vs. normal patients and M1 vs. M5 patients with AML. For the first condition, we tested the null hypothesis that there was no difference in the average microRNA expression of M1 and M5 patients with AML. For the second condition, we tested the null hypothesis that the difference in log-2 fold change for M1 and M5 patients with AML was 0. The p-values obtained for the two null hypotheses were adjusted for multiple comparisons using the Benjamini and Hochberg method [14].

Because the Exiqon arrays have duplicate probes for each microRNA, a microRNA was said to be DE if one or both the probes had adjusted P value less than 0.05.

DE mRNAs

We obtained the mRNA microarray data for M1 and M5 patients with AML with normal karyotype from the previously published mRNA microarray data (GSE 1159) [15]. This single-color Affymetrix data was preprocessed using the Bioconductor package affy [16] with RMA background correction [17], quantile normalization [18], and summarization of gene expressions using the median polish algorithm. A total of 67 patients were considered for the analysis.

We tested the null hypothesis that there was no difference in the average mRNA expression of M1 and M5 patients with AML and obtained the P values. These P values were adjusted for multiple comparisons using the Benjamini and Hochberg method.

Because each mRNA has one or more probesets associated with it, an mRNA was said to be DE if one or more of the associated probesets had the adjusted P < 0.05. We visually assessed the microarray expression data by clustering the microRNA and mRNA samples based on DE microRNAs and mRNAs, respectively. These clusters are depicted in this report using heatmaps.

To discover the microRNAs that potentially regulate the mRNA expression, we performed a non-parametric gene-set test. First, for each microRNA that was statistically significant at one or more timepoints, we obtained putative target-mRNAs using three target-prediction databases —TargetScan [19], Pictar [20] and miRanda [21]— with similar prediction accuracy [22]. We focused on only those putative target-mRNAs that were statistically significant. Second, we tested the null hypothesis that the average log-2 fold change value for the target-mRNAs is the same as that for the remaining mRNAs. The null hypothesis was tested using the Wilcoxon test and the P values were adjusted for multiple comparisons using the Bonferroni-Holmes method. If a microRNA, previously identified as DE, had adjusted P < 0.05, it was said to be enriched and a potential regulator of mRNA expression. For each enriched microRNA, we obtained the list of mRNAs that had log-2 fold change value (median value over all probesets) opposite to that for the microRNA (median value over the duplicate probes).

Induction of monocytic differentiation of AML cells

The human AML cell lines HL60 (M2, [23]) and NB4 (M3, kindly provided by A/Prof Richard Lock, Children's Cancer Institute Australia) were used for these experiments and were maintained in Rosewell's Park Memorial Institute (Gibco, Carlsbad, CA) media supplemented with 10% fetal bovine serum, 2 mM glutamine, 100U/mL penicillin, and 100 μg/mL streptomycin (all from Gibco). Macrophage differentiation was induced by a 24 hour pulse of 100 nM 1,23-dihydroxy-vitaminD3 followed by 72 hour incubation with 20 nM phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich, St Louis, MO). Cells were collected at 0, 48, and 96 hours for total RNA extraction and cell characterization. A t test was performed to determine whether the difference in microRNA expression between the cells treated with vitamin D or phorbol myristate acetate and the untreated cells was statistically significant at timepoint 48 hrs or 96 hrs.

Cell lineage characterization

Cell lineage-specific marker expression was analyzed by flow cytometry in untreated and treated cells at various timepoints during differentiation. Flow cytometry analysis was performed using the following antibodies: flurorescein isothiiocyanate-conjugated CD14, phycoerythrin-conjugated CD15, allophycocyanin-conjugated CD11b, and flurorescein isothiiocyanate-conjugated CD71 together with their respective isotype controls (all from BD Pharmigen, San Jose, CA). 5 × 105 cells were incubated with 10 μL of a specific antibody or antibody combination and were incubated in the dark for 20 min at room temperature. Cells were washed twice with PBS containing 0.5% bovine serum albumin and 0.1% Sodium Azide and resuspended in 400 μL 0.5% paraformaldehyde. Cells were analyzed immediately by LSRII cell analyser (Becton Dickinson, Mountain View, CA), using FACSDiva software (Becton Dickinson) with a minimum acquisition of 30,000 events.

Morphological analysis was performed on cytospin preparations of treated and untreated time matched cells stained with Wright staining solutions. Cells were visualized using a Nikon Eclipse 50i microscope and photographed with a Nikon Digital Sight DLS2 camera.

Real time RT-PCR analysis

RNA was extracted from patient blasts using Trizol (Invitrogen) according to the manufacturer's protocol. One microgram of total RNA was used in the cDNA synthesis reaction using the nCode microRNA first strand synthesis system (Invitrogen). Real time polymerase chain reaction (PCR) was performed using the Platinum Sybr Green taq (Invitrogen) and the Rotorgene RG-3000 thermocycler (Qiagen, Hilden, Germany). A t test was performed to determine whether the difference in microRNA expression between M1 and M5 samples was statistically significant.

Luciferase assays

The 3′ untranslated region of each gene (incorporating the predicted miR binding site) was amplified from genomic DNA using primers 5′ modified with appropriate restriction enzyme consensus sites, cloned into pGEM T-easy (Promega, Madison, WI) and then directionally subcloned into pMIR-REPORT (Ambion, Foster City, CA) immediately downstream of the firefly luciferase open reading frame. The integrity of the construct was confirmed by automated fluorescent sequencing. 24 hours before transfection, Hela cells were plated in antibiotic free media at a density of 1 × 105 cells/mL in a 96-well format. Cells were co-transfected with 9 ng pMIR 3′ untranslated construct and 0.8 ng pRL-CMV (Invitrogen) renilla luciferase vector, alongside 30 nM pre-miR microRNA Precursor (Ambion) using Lipofectamine 2000 (Invitrogen). After 24 hours, cells were washed with PBS and luminescence was analyzed by the Dual Luciferase Assay (Promega) on a Fluostar Opima (BMG Labtech, Offenburg, Germany) platereader, with each result normalized to the renilla transfection control. Relative luciferase values for cells transfected by the relevant pre-miR microRNA precursor were compared to transfection with that of a nontargeting pre-microRNA negative control. Each transfection was performed in triplicate and the assay was independently repeated at least seven times. A t test was performed to determine whether the changes in luciferase values were statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Analysis of microRNA array data

In this study, the microRNA profiles of patient samples were analyzed, with a mean blast percentage of 74% with a similar number of patients in each of the represented FAB subgrouping (Table I). The microRNA expression profile of each patient was then determined and used in an unsupervised cluster analysis (Fig. 1). When compared with a pooled normal bone marrow sample (n = 8), 69 microRNAs were found to be DE between the patients with AML and the pooled normal bone marrow control. In concordance with previous studies, patients did not cluster based on their FAB subtype [15, 24].

thumbnail image

Figure 1. Unsupervised cluster analysis of newly diagnosed normal cytogenetics patients with AML based on microRNA expression. Patients do not group according to FAB classification suggesting that microRNA expression is unable to predict FAB classification when all normal cytogenetic patients with AML are considered; FAB classification indicated by pink (M1), yellow (M2), red (M4), or green (M5). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Table I. Clinical Parameters of 27 Newly Diagnosed Cytogenetically Normal Patients with AML
Sex
 Male15
 Female12
FAB subtype
 M16
 M28
 M46
 M57
Age
 Median48
 Range23–83
Blasts %
 Mean74.3
 Range52–93

The FAB classification system relies primarily of the morphology of blasts in the bone marrow at time of diagnosis [25]. The morphology of the blasts may indicate at which step of differentiation the malignant progenitor cells are halted, and we hypothesized that by comparing the gene expression of two FAB subtypes we may identify microRNAs that contribute to that differentiation block. In an effort to identify such microRNAs, the microRNA expression profiles of M1 and M5 subtypes were compared. The M1 subtype of AML describes leukemic blasts as being without maturation, while the M5 subtype contains blasts which have a monocytic morphology. The DE microRNAs, based on a comparison of M1 and M5 subtypes, are listed in Table II. If one or both the probes corresponding to a microRNA were DE, then the microRNA was considered to be DE. Therefore, the P values in Table II correspond to probes that were used to identify the DE microRNAs. A heatmap of M1 and M5 patients, based on the average expression of 12 DE microRNAs, accurately delineated the M1 and M5 subtypes (Fig. 2).

thumbnail image

Figure 2. Heatmap of patients with AML with a M1 or M5 FAB classification. Expression of 12 microRNAs (blue indicating a higher level of expression) is able to accuracy divide patients into FAB groupings; FAB classification indicated by pink (M1) or green (M5). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

Table II. Microarray Results Indicating Differential Expression of MicroRNAs Between M1 and M5 FAB Subtypes with a >1.5-Fold Change
microRNAFold changeP value
microRNAs overexpressed in M1
 Hsa-miR-130a3.9010.0002
 Hsa-miR-135b1.6270.0485
 Hsa-miR-146a6.140<0.0001
 Hsa-miR-146b4.0020.0002
 Hsa-miR-181a6.123<0.0001
 Hsa-miR-181a*3.9510.0003
 Hsa-miR-181b3.476<0.0001
 Hsa-miR-181d3.018<0.0001
 Hsa-miR-6633.7510.0300
microRNAs overexpressed in M5
 Hsa-miR-193a1.9040.0222
 Hsa-miR-212.9060.0300
 Hsa-miR-3701.8420.0150

Of note was the high representation of the miR-181 family of microRNAs, with miR-181a, −181b, −181d and 181a* over-expressed in the M1 subtypes when compared to the M5 subtype.

Identification of mRNA targets

To determine possible mRNA/microRNA interactions, we analyzed a publically available mRNA expression dataset of patients with CN-AML, according to the M1/M5 status [15]. A set of mRNAs that were DE between the M1 and M5 subtypes was identified. Although the delineation between M1 and M5 subtypes is not as accurate as that seen with the microRNA expression profiles, there is a clear separation of the majority of M1 and M5 patients (Fig. 3).

thumbnail image

Figure 3. Heatmap of mRNA expression data [15] of patients with AML with a M1 or M5 FAB classification. Although not as accurate as the microRNA data in Figure 2, the M1 and M5 subtypes do tend to group together based on mRNA expression profiles; FAB classification indicated by pink (M1) or green (M5). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

From this set of DE genes, a subset of genes plausibly associated with monocytic differentiation was identified. Similarly, those genes that characterize the undifferentiated blast cells of M1 were identified (Table III). Several key stem cell markers such as CD34, CD7, Kit, and Notch 1 were overexpressed in the M1 blasts as expected for a more stem cell-like population. Similarly, monocytic markers such as CD14, KLF4, SPI1, CEBPB and MAFB were over-expressed in the monocyte differentiated blasts of M5 patients.

Table III. Differential Expression of Key Phenotypic Markers for M1 and M5 Identified Using from the Valk Dataset [15]
 Entrez IDFold changeGene symbolGene name
  1. An increase in the expression of blast or stem cell markers was observed in the M1 subtype consistent differentiation block of M1 blasts, with a concordant increase in the expression of monocytic markers in the M5 subtype consistent with maturation of these blasts down the monocytic lineage.

MONOCYTIC MARKERS9935−11.23MAFBv-maf musculoaponeurotic fibrosarcoma oncogene homolog B (avian)
929−11.18CD14CD14 molecule
4332−5.02MNDAmyeloid cell nuclear differentiation antigen
604−4.29BCL6B-cell CLL/lymphoma 6
9314−4.16KLF4Kruppel-like factor 4 (gut)
7099−3.67TLR4toll-like receptor 4
2896−3.22GRNgranulin
1051−3.06CEBPBCCAAT/enhancer binding protein (C/EBP), beta
7100−3.05TLR5toll-like receptor 5
1052−2.46CEBPDCCAAT/enhancer binding protein (C/EBP), delta
6036−2.15RNASE2ribonuclease, RNase A family, 2 (liver, eosinophil-derived neurotoxin)
3726−2.15JUNBjun B proto-oncogene
6256−2.09RXRAretinoid X receptor, alpha
1959−2.04EGR2early growth response 2 (Krox-20 homolog, Drosophila)
671−1.87BPIbactericidal/permeability-increasing protein
2353−1.86FOSv-fos FBJ murine osteosarcoma viral oncogene homolog
3394−1.82IRF8interferon regulatory factor 8
6688−1.74SPI1spleen focus forming virus (SFFV) proviral integration oncogene spi1
566−1.53AZU1azurocidin 1
3398−1.52ID2inhibitor of DNA binding 2, dominant negative helix-loop-helix protein
BLAST OR STEM CELL MARKERS74901.56WT1Wilms tumor 1
26241.69GATA2GATA binding protein 2
514661.74EVLEnah/Vasp-like
796821.75MLF1IPMLF1 interacting protein
38151.77KITv-kit Hardy-Zuckerman 4 feline sarcoma viral oncogene homolog
6131.79BCRbreakpoint cluster region
533351.82BCL11AB-cell CLL/lymphoma 11A (zinc finger protein)
46131.83MYCNv-myc myelocytomatosis viral related oncogene, neuroblastoma derived (avian)
48511.83NOTCH1Notch homolog 1, translocation-associated (Drosophila)
43521.84MPLmyeloproliferative leukemia virus oncogene
20781.97ERGv-ets erythroblastosis virus E26 oncogene homolog (avian)
43302.63MN1meningioma (disrupted in balanced translocation) 1
9472.83CD34CD34 molecule
9242.97CD7CD7 molecule
798703.64BAALCbrain and acute leukemia, cytoplasmic

We focused on those microRNAs that were over-expressed in the M1 patients compared with M5 patients as it is possible that an increased level of these microRNAs in M1 repress expression of mRNAs that normally allow progression down the differentiation pathway. We then selected those microRNAs that are bioinformatically predicted to target genes that have been reported to be involved in myelomonocytic differentiation (miR-130, 135b, 146a, 181a, 181b, and 181d). Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) of microRNAs confirmed the array data with a significantly higher level of expression of these microRNAs in M1 patients compared to M5 patients (Fig. 4).

thumbnail image

Figure 4. qRT-PCR confirmation of DE microRNAs. Consistent with the microRNA results all six candidate microRNAs are significantly decreased in M5 patients. *P < 0.05, **P < 0.01, ***P < 0.001.

Download figure to PowerPoint

We next identified mRNAs potentially targeted by this set of set of microRNAs using the TargetScan, Pictar and miRBase target prediction databases. These predicted mRNA targets were then correlated with the Valk dataset to identify genes whose expression was appropriately inversely correlated with the identified microRNAs of interest. This generated a list of putative microRNA/mRNA pairs that could potentially contribute to differentiation block. The level of expression of mRNAs identified by this correlation of predictive bioinformatic data and mRNA expression data was then confirmed in our cohort of patients by qRT-PCR as depicted in Fig. 5, which shows mRNAs DE between M1 and M5 patients.

thumbnail image

Figure 5. qRT-PCR expression analysis of key monocytic genes identified using the Valk dataset ([15] and Table III). The increase in expression observed in the microarray data was confirmed in our cohort of patient was a significant increase in expression of monocytic genes in M5 patients. *P < 0.05, **P < 0.01.

Download figure to PowerPoint

Monocytic differentiation of AML cells

To investigate their potential role in monocytic differentiation, the expression level of the candidate microRNAs was assessed in the well established AML cell line models (HL60 and NB4) of monocytic differentiation. We confirmed in vitro differentiation by assessing the morphology and cell surface markers of cells before and after treatment (Fig. 6).

thumbnail image

Figure 6. A, Flow cytometry analysis of HL60 and NB4 cell lines treated with vitamin D and PMA showing an increase in the monocytic markers CD14 and CD11b. B, Morphology of HL60 and NB4 cell lines before and after treatment with the characteristic mature monocytic/macrophage appearance seen post treatment with vitamin D and PMA. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Download figure to PowerPoint

The expression levels of the candidate microRNAs over the time course of monocytic differentiation was assayed by qRT-PCR, with six candidate microRNAs showing a significant decrease in expression as the cells differentiated in response to Vitamin D and PMA (Fig. 7). This result suggests that the expression level of these microRNAs is inversely proportional to the maturation of the cell line towards a monocytic lineage. This observation fits the hypothesis that these microRNAs are targeting key myeloid transcription factors thereby inhibiting their role in monocytic differentiation.

thumbnail image

Figure 7. Expression of candidate microRNA levels in AML cell lines treated with vitamin D and PMA. For all six candidate microRNAs analyzed, a significant decrease in expression was observed at one or both time points during monocytic differentiation. *P < 0.05, **P < 0.01.

Download figure to PowerPoint

To investigate the specificity of the downregulation of the candidate microRNAs during monocytic differentiation, we assessed their change in expression during granulocytic differentiation of the HL60 and NB4 cell lines in response to 1 μM all-trans-retinoic acid (ATRA), using miR-223 which is known to be upregulated as a positive control (Fig. 8). Four of the six microRNAs (miR-181a and 181b were downregulated) were unchanged or increased in contrast to what was observed during monocytic differentiation.

thumbnail image

Figure 8. Expression of candidate microRNA levels during granulocytic differentiation of cell lines with ATRA. Both miR-181a and miR-181b were down-regulated and miR-130a was upregulated at 96 hours post-ATRA treatment. All other candidate microRNAs were unchanged after ATRA treatment, with miR-223 acting as a positive control. *P < 0.05, **P < 0.01.

Download figure to PowerPoint

Confirmation of the targeting of a given mRNA by a candidate microRNA was assessed by Luciferase assay. The candidate microRNAs were found to target key monocytic and myelocytic transcription factors and proteins (Fig. 9).

thumbnail image

Figure 9. Luciferase assays to determine the suppression of predicted target mRNAs by candidate microRNAs. A significant down-regulation of luciferase activity was observed indicating that several key monocytic markers are true targets of the candidate microRNAs. A scrambled microRNA species acted as a negative control. *P < 0.05, **P < 0.01, ***P < 0.001.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Gene expression profiling of AML has provided significant insights into the underlying molecular and biochemical abnormalities that drive the malignant process. In one of the largest studies of gene expression profiling in AML, Valk et al. [15] showed that patients clustered into groups based on mRNA gene expression profiles associated with gross chromosomal rearrangements, genetic mutations, and oncogene expression. Since this study mRNA expression profiling of patient with AML blasts has identified gene signatures associated with prognosis [26], response to chemotherapy [27–29] and epigenetically silenced genes [30]. With the introduction of microRNA microarrays, microRNA expression profiles have been associated with prognosis, cytogenetics, and recurrent gene mutations [31–33]. In this study, we aimed to further investigate whether microRNA gene expression profiles of specific AML subtypes can identify microRNAs contributing to the pathogenesis of AML by analyzing microRNA expression patterns according to morphological subtype.

An unsupervised cluster analysis of the microRNA expression profiles of our cohort of normal karyotype AMLs samples did not identify a discernable cluster pattern. This can be compared with Jongen-Lavrencic's analysis of 215 AML samples, which included a representation of the major cytogenetic ((including t(8;21), inv [16], t(8;21), and t(11q23)) and genetic (including NPM1 and CEBPA mutated) subtypes [9]. In this analysis, there was a clear separation into a complex cluster pattern by microRNA expression that in part reflected the underlying genetic defect. It was apparent from this study, however, that the clustering patterns were only party explained by the known genetic defect, illustrated by the fact that NPM1 mutant AML was represented in as many as six different clusters. No other group to date has presented an unsupervised cluster analysis of AML samples restricted to the normal karyotype grouping, however in view of Jongen-Lavencic and colleagues findings [9], it can be expected that with a large enough sample size that these clusters would emerge. What was clear, however, even in the much larger dataset of this study (compared with our 27 samples), was that there was no natural clustering that could be explained by the FAB morphological subtype.

Seeing that there was no such natural clustering was observed, we compared the microRNA expression profiles of M1 (AML with without maturation) and M5 (acute monoblastic and monocytic leukemia) FAB subtypes. On this basis we identified 12 candidate microRNAs that accurately separated M1 and M5 AML (Fig. 2). Of these, miR-146a, miR-181a, miR-146b, miR-181a*, miR130a, miR-663, miR-181b, miR-181d, and mir-135b were more highly expressed in M1 samples, whereas mir-21, miR-193a, and miR-370 were overexpressed in M5 samples. This informative set of microRNAs was partially foreshadowed in a study by Debernadi et al. [34], which established by qRT-PCR that miR-181a was more highly expressed in M1/M2 blasts compared with M4/M5 blasts. In this study, miR-181a was individually selected on a candidate microRNA basis owing to previous observations that this microRNA was involved in hematopoietic differentiation [35] rather than by more extensive microRNA expression profiling. This study also demonstrated that 163 mRNAs were negatively correlated with miR-181a levels as determined by U133A/B oligonucleotide array (Affymetrix), of which 15 had bioinformatically predicted microRNA binding sites.

In the same manner, we wished to identify plausible microRNA;mRNA pairs on the extended microRNA expression signature (compared with the study by Debernadi et al.) of 12 microRNAs separating M1 and M5 patient samples in our cohort. To this end, we used the microRNA target prediction databases to identify putative target mRNAs of the 9 microRNAs overexpressed in M1 blasts. We also reanalyzed a previously reported mRNA gene profiling dataset [15] to identify mRNAs that were DE between the M1 and M5 subtypes (Table III). Several well recognized stem cell markers were over-expressed in the M1 blasts such as CD34, KIT, and NOTCH1 and conversely monocytic markers including CD14, MAFB, SPI1 (PU.1), and KLF4 were over-expressed in M5 blasts. The bioinformatically predicted mRNA targets of the former analysis were then compared with the reanalysed Valk mRNA expression data [15], to identify mRNAs that were both bioinformatically predicted targets of these microRNAs and that were actually downregulated in AML samples. We found that six microRNAs had target mRNAs relevant to hematopoiesis, which satisfied this criterion. The strategy used in this study was able to identify microRNAs that have previously been shown to influence apoptosis, stem cell biology and hematopoiesis. In addition, to Debernadi's findings [34] miR-181a has been shown to modulate radiosensitivity [36] and apoptosis [37], whereas miR-181b influences multidrug resistance in cancer cell lines via the targeting of BCl-2 [38] and is down regulated in APL blasts after chemotherapy [39]. MiR-126 and −130a play a role in erythroid differentiation [40], miR-126 and −135 in megakaryocytic differentiation [41, 42] and miR-181b in granulopoiesis [43].

MicroRNAs over-expressed in the M1 subtype may target and suppress important differentiation factors and hold these blasts at a less differentiated stage of maturation. To further explore whether the six microRNAs that we identified are linked to this process, we performed further studies in an in vitro model system of monocytic differentiation. First, we examined the expression level of these microRNAs during monocytic differentiation of NB4 and HL60 cells treated with Vitamin D and PMA. All six candidate microRNAs exhibited a significant decrease in expression over the differentiation time-course (Fig. 7) that was consistent between cell lines. This data suggests that the microRNA expression of these six microRNAs is linked to the level of monocytic differentiation of the cells.

To determine the specificity of our candidate microRNAs in monocytic differentiation, we assessed their expression levels during granulocytic differentiation (Fig. 8). MicroRNAs 181a and 181b showed a decrease in expression in response to ATRA treatment, though not to the same magnitude as seen during monocytic differentiation. MicroRNA 181a is predicted to block differentiation of lymphoid progenitors [44], and to be down-regulated during murine myeloid differentiation [35], with our results confirming this down regulation during myeloid differentiation. MicroRNA 223 was used as a positive control due to its previously described increase in cell lines undergoing ATRA induced granulocytic differentiation [45]. Taken as a whole, we feel that our candidate microRNAs, with the exception of miR-181a and 181b, are specifically implicated in the control of monocytic differentiation.

Our candidate microRNAs were confirmed to target several key myeloid genes by reporter assay using a luciferase system (Fig. 9). KLF4, which we demonstrated to be a target for miR-130a and −135b, has previously been shown to function in a monocyte-restricted fashion to promote monocyte differentiation and furthermore knockdown of KLF4 inhibits phorbol ester induced monocytic differentiation [46]. These data suggest that the increased level of miR-130a and −135b in M1 patient blasts would decrease the level of KLF4 and contribute to the halt in the maturation of these blasts, stalling them at an earlier progenitor stage. The candidate microRNAs may further contribute to this maturation block via the targeting of the myeloid factors MAFB [47, 48], HOXA10 [49], IRF8 [50], and MCL1 [51]. Each of these was confirmed to be targeted by either individual or multiple microRNAs from our list of 6 candidate microRNAs as described above (Fig. 9). It is also possible to infer the presence of gene regulatory networks under microRNA control. For example, the HOXA10/MAFB regulatory network is involved in monocyte maturation [52]. Our reporter analysis has shown HOXA10 and MAFB to be repressed by miR-130a and miR-135b, respectively.

This report describes for the first time a set of microRNAs that are DE between M1 and M5 patients. These microRNAs may contribute to the maturation block observed in M1 blasts. The candidate microRNAs were found to decrease when AML cell lines are differentiated into monocytes and have been shown to target key myelomonocytic differentiation factors. It remains to be shown how these microRNAs may have come to be dysregulated. This report adds to the literature dealing with the microRNA control of hematopoiesis and gives credibility to a possible role for microRNA dysregulation in either providing a previously unrecognized second hit in AML or contributing to the pathogenic pathway of known AML genetic abnormalities.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank the Arrow Bone Marrow Transplant Foundation (M.L. and A.B.), Australian National Health and Medical Research Council (A.B. and D.A.), St Vincent's Hospital Clinic (D.M. and M.L.), and Sydney Medical Research Foundation (D.M.) for funding aspects of this project and the Australian Leukemia and Lymphoma Group Tissue Bank for contribution of patient samples.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References