Consistent Deregulation of Gene Expression between Human and Murine MLL Rearrangement Leukemias

Important biological and pathologic properties are often conserved across species. Although several mouse leukemia models have been well established, the genes deregulated in both human and murine leukemia cells have not been studied systematically. We performed a serial analysis of gene expression in both human and murine MLL-ELL or MLL-ENL leukemia cells and identified 88 genes that seemed to be significantly deregulated in both types of leukemia cells, including 57 genes not reported previously as being deregulated in MLL-associated leukemias. These changes were validated by quantitative PCR. The most up-regulated genes include several HOX genes (e.g., HOX A5, HOXA9, and HOXA10) and MEIS1, which are the typical hallmark of MLL rearrangement leukemia. The most down-regulated genes include LTF, LCN2, MMP9, S100A8, S100A9, PADI4, TGFBI, and CYBB. Notably, the up-regulated genes are enriched in gene ontology terms, such as gene expression and transcription, whereas the down-regulated genes are enriched in signal transduction and apoptosis. We showed that the CpG islands of the down-regulated genes are hypermethylated. We also showed that seven individual microRNAs (miRNA) from the mir-17-92 cluster, which are overexpressed in human MLL rearrangement leukemias, are also consistently overexpressed in mouse MLL rearrangement leukemia cells. Nineteen possible targets of these miRNAs were identified, and two of them (i.e., APP and RASSF2) were confirmed further by luciferase reporter and mutagenesis assays. The identification and validation of consistent changes of gene expression in human and murine MLL rearrangement leukemias provide important insights into the genetic base for MLL-associated leukemogenesis. [Cancer Res 2009;69(3):OF1109–16]

Chromosome translocations are among the most common genetic abnormalities in human leukemia. The MLL (mixed lineage leukemia) gene was identified as a common target of chromosomal translocations associated with human acute leukemias (1, 2). MLL is located on human chromosome 11 band q23 and on mouse chromosome 9. More than 50 different loci are rearranged in11q23 leukemias involving MLL in either acute myelogenous leukemia (AML) or acute lymphoblastic leukemia (ALL; ref. 3). MLL rearrangements are associated with a poor prognosis (4). MLL-ELL and MLL-ENL that result from t(11;19)(q23;p13.1) and t(11;19)(q23;p13.3), respectively (1, 5), are two common examples of these rearrangements. These two fusions are frequently involved in human AML, whereas MLL-ENL is also involved in human ALL. The translocations result in an in-frame fusion of the NH₂ terminus of the MLL gene and the COOH terminus of each partner gene. Retroviral-mediated gene transfer of MLL-ENL and MLL-ELL transforms primary myeloid progenitor cells and cause acute myeloid leukemia in mice (6, 7).

Gene expression profiles differ between distinct subtypes of leukemia and provide specific markers for clinical diagnosis. It is commonly observed that important biological/pathologic properties are often conserved across species (8, 9). Model organisms have contributed substantially to our understanding of the etiology of human disease and the development of new treatment methodologies (10). However, although genetically engineered mouse leukemia models have been well established (6, 7, 11, 12), there are few systematic studies to identify and study the genes that exhibit similar abnormal expression patterns in both human and murine leukemia cells. To perform an interspecies gene expression comparative study in leukemia, we used the serial analysis of gene expression (SAGE) technique (13) to compare gene expression between MLL-ELL or MLL-ENL myeloid leukemia progenitor cells and normal myeloid progenitor cells in both humans and mice. Herein, we report the identification and validation of differentially expressed genes commonly present in both human and murine MLL-ELL or MLL-ENL leukemias.

Patient samples. The patient samples were obtained at the time of diagnosis and with informed consent at University of Chicago and were stored in liquid nitrogen until use.

SAGE assay and data analysis. Cell purification, total RNA isolation, cDNA synthesis, and SAGE were carried out according to our established procedures (13–15). SAGE tag sequences were extracted with SAGE 2000 software. Tag counts were converted to counts per 100,000, and the expression data were cross-linked to UniGene clusters by extracting the 3′-most NlaIII SAGE tag for each transcript in each UniGene clusters. Only tags that matched to a single gene cluster were taken into account. All SAGE tags mapped to the same gene were then combined, and the sum of their counts in a given sample represented the expression level/signal of that gene in that sample (16, 17). χ² test was used to identify differentially expressed genes, as Man and colleagues (18) showed that χ² test has the best power and robustness in SAGE data analysis.

Functional classification and annotation of the candidate genes. We used the Database for Annotation, Visualization, and Integrated Discovery (DAVID; version 2008)⁶

for functional classification and annotation of candidate genes. DAVID provides a comprehensive set of functional annotation tools for investigators to understand the biological meaning behind large list of genes (19). A gene cluster means a set of genes belonging to the same functional category, as defined by DAVID.

Quantitative real-time PCR assays of protein coding or microRNA genes and data analysis. We performed quantitative real-time PCR (qPCR) to determine expression of protein coding genes with SYBR green dye using kits from Qiagene and microRNAs (miRNA) using TaqMan qPCR kits from Applied Biosystems. PGK1 or glyceraldehyde-3-phosphate dehydrogenase was used as endogenous controls for protein coding genes, whereas U6 RNA was used as an endogenous control for miRNAs. PCR reactions and data analyses were performed, as described previously (20, 21).

Methylation-specific PCR. The methylation status of the promoter region was determined by the methylation-specific PCR (MSP) method (22). The primers were designed to anneal specifically to methylated and unmethylated CpG dinucleotides in promoter regions of genes using the Primer3 program. Genomic DNA was isolated using QIAamp DNA mini kit (Qiagen). One microgram of DNA was used for bisulfite modification using the CpGenome DNA modification kit (Chemicon) according to the manufacturer's instructions. The bisulfite-converted DNA was amplified in a total volume of 20 μL using GeneAmp Gold buffer containing 4 mmol/L MgCl₂, 0.5 μmol/L of each primer, 0.2 mmol/L deoxynucleotide triphosphate, 5 μg bovine serum albumin, and 1.25 units of AmpliTaq Gold DNA polymerase (Roche). Hot start PCR was performed for 30 cycles, which consists of denaturation at 95°C for 30 s, annealing at 60°C for 30 s, and extension at 72°C for 45 s, followed by a final 7-min extension for all primer sets. The products were separated on 10% polyacrylamide gels.

Cell culture and 5-aza-2′-deoxycytidine (Decitabine) treatment. Human leukemia cell lines were grown at 37°C under 5% CO₂ in RPMI 1640 (Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen), 1% penicillin/streptomycin, and 1% HEPES. The cells were plated at 2 × 10⁶ per flask in 4 mL of medium. After 24 h, cells were exposed to 5-aza-2′-deoxycytidine (5-Aza-CdR; Fluka) at 1 μmol/L for 96 h. In parallel, untreated cells were used as a control. After 48 h of continuous exposure to 5-Aza-CdR, the medium was changed. Cells were harvested after an additional 48 h of incubation. Then, total RNA and genomic DNA were isolated from the treated and untreated cell lines for further qPCR and MSP assay, respectively.

Luciferase reporter and mutagenesis assay. MiR-17 expression plasmid or its control plasmid (i.e., MSCVpuro) was cotransfected into HEK293T cells with a single report plasmid (pMIR-Report plasmid; Ambion) containing either wild-type or mutated 3′ untranslated region (UTR) of an individual predicted target gene. Luciferase was measured 42 h after transfection. The firefly luciferase activity was then normalized to β-galactosidase activity. Experiments were repeated thrice, independently.

Eighty-eight genes were identified by SAGE analysis to be significantly abnormally expressed in both human and murine MLL-ELL and/or MLL-ENL leukemia. We used a modified SAGE technique (13) to compare gene expression profiles between MLL-ELL or MLL-ENL myeloid leukemia progenitor cells and normal myeloid progenitor cells (CD15+ or Gr-1+) in both humans and mice. We analyzed four patient samples (two with each fusion) and two retrovirally induced mouse leukemias containing either MLL-ELL or MLL-ENL fusions and a leukemic mouse cell line with an MLL-ELL fusion (Table 1). SAGE tags (484,303) were identified from the nine samples (total SAGE tag counts of each sample varies from 40,000 to 100,000, with 53,811 tags per sample on average), yielding 103,899 unique tags in human and 60,993 in mouse samples (see Table 1).

We identified 88 candidate genes that seemed to be deregulated in both human and murine MLL-ELL and/or MLL-ENL leukemias. Of them, 6 and 26 genes are up-regulated and down-regulated, respectively, in both types of leukemia; 3 and 13 genes are up-regulated and down-regulated, respectively, in MLL-ELL leukemias alone; 23 and 17 genes are up-regulated and down-regulated, respectively, in MLL-ENL leukemias alone (see Fig. 1A and Supplementary Table S1). All the significant genes have at least 3-fold difference in expression (tag counts per 100,000) between each leukemia sample and the normal control and with a χ² test P value of <0.05. As this criterion is stringent, it therefore limits the total number of the candidate genes identified.

Notably, the four most up-regulated genes (each has at least a 30-fold change in expression) in both types of human and mouse leukemia samples are three homeobox (HOX) genes (i.e., HOXA5, HOXA9, and HOXA10) and one HOX cofactor, MEIS1 (see Fig. 1A) The eight most down-regulated genes (each has at least a 30-fold change in expression on average) in both types of human and mouse leukemia samples include LTF, LCN2, MMP9, S100A8, S100A9, PADI4, TGFBI, and CYBB (see Supplementary Table S1). Note that the fold change of some up-regulated or down-regulated genes may be overestimated because of the limited number of SAGE tags examined (40,000–100,000 tags per sample).

Although 56 genes were identified to be significantly deregulated in only one type of leukemias, many of them exhibited a similar deregulation pattern in both types of leukemias (see Supplementary Table S1), albeit they were significant only in one according to our stringent criterion. If more samples were analyzed and/or a less stringent criterion was used, many of these genes would seem to be deregulated in both MLL-ELL and MLL-ENL leukemias.

Thirty-one of the 88 genes identified by SAGE have been previously reported to be deregulated in MLL rearrangement leukemia. Much effort has been expended previously to identify candidate genes that might be involved in MLL rearrangement leukemia using different methods, such as microarray and qPCR. To obtain a comprehensive list of candidate genes involved in MLL rearrangement leukemia, we searched 35 relevant publications (see Supplementary Table S2A) and collected a total of 1,647 genes that were reported to be significantly deregulated in MLL rearrangement leukemia. The top two most frequently reported genes are MEIS1 and HOXA9, having been reported as up-regulated genes by 16 and 15 different publications, respectively. Among the 1,647 genes, 370 were reported by at least two different studies. Of them, only 10 genes (all were reported by only two publications) have contradictory information regarding whether they are up-regulated or down-regulated in MLL rearrangement leukemia. The remaining 360 genes were reported consistently by various different studies, with 213 up-regulated and 147 down-regulated in MLL rearrangement leukemia compared with other subtypes of acute leukemias and/or normal controls. We found that 31 of the 88 genes (35%) identified from our SAGE analysis were reported by previous studies (see Supplementary Table S2B).

Validation of 81 genes by a large-scale qPCR assay. To validate the deregulation patterns of the genes identified from our SAGE analysis and those from the previous studies, we performed a large-scale qPCR validation for 81 genes in 20 samples. The 81 genes (see Supplementary Table S3) include 43 identified from our SAGE analysis and 38 reported only by others. Among the 43 genes identified by SAGE, 19 were also reported by others. The 20 samples include 12 human and 8 mouse samples. The human samples include three MLL-ELL and three MLL-ENL leukemia samples, in addition to six human normal control samples. The mouse samples include two retrovirally induced mouse leukemia samples containing either MLL-ELL or MLL-ENL fusion, two leukemia cell lines with MLL-ELL fusions, and one leukemia cell line with an MLL-ENL fusion, in addition to three murine normal control samples.

We found that 60 of the 81 tested genes (74%) seemed to be significantly deregulated in leukemia samples (see Fig. 1B). Of these, 50 candidate genes were deregulated in human and mouse MLL-ELL leukemia (29 up-regulated and 21 down-regulated) and 48 were deregulated in human and mouse MLL-ENL leukemia (32 up-regulated and 16 down-regulated), whereas 55 genes seemed to be significantly deregulated (34 up-regulated and 21 down-regulated) when combining the MLL-ELL and MLL-ENL leukemia samples together. As shown in Supplementary Table S3, 36 of the 43 SAGE candidate genes (84%) were confirmed to be significantly deregulated in leukemia samples in a similar mode (i.e., up-regulated or down-regulated) to SAGE data, whereas only 31 of the 57 genes (54%) reported previously were confirmed here in a similar mode to previous report(s). Notably, MYB and SMARCA4 were detected by SAGE as being up-regulated but were reported by others as being down-regulated in MLL rearrangement leukemia; the opposite is true for PTPRC, S100A8, ANXA1, and NCF1. Our qPCR result of these six genes is consistent with our SAGE analysis rather than previous reports (see Supplementary Table S3). In addition, four genes reported only by others, including PROM1, ADCY9, LST1, and TCF4, were detected by qPCR as being deregulated in MLL rearrangement leukemia but in a different mode from previous reports.

Functional classification and annotation of the candidate genes identified by SAGE. To gain further insight into the biological importance of these 88 candidate genes, we analyzed the functional categories of known genes in terms of gene ontology (GO) using the DAVID tools (19). One gene (i.e., MAGT1; down-regulated in MLL-ENL leukemia) is not in the DAVID gene database; therefore, only 87 genes were analyzed. We first performed Gene Functional Classification for the 87 genes and found that there were four gene clusters (only the clusters having at least three members were considered). Gene cluster 1 contains S100A8, S100A9, and S100A11 (enrichment score, 1.98); all of them were down-regulated in leukemia. Gene cluster 2 contains CTSS, CTSH, and MAN2B1 (enrichment score, 1.18); the first two were down-regulated, whereas the last one was up-regulated in leukemia. Gene cluster 3 contains ARHGEF1, CENTB2, and SMAP1L (enrichment score, 0.86); all of them were down-regulated in leukemia. Remarkably, gene cluster 4 contains eight genes, including HOXA5, HOXA9, HOXA10, MEIS1, MYB, SOX4, KLF4, and MLL5 (enrichment score, 0.75); the first six were up-regulated whereas the last two were down-regulated in leukemias.

We also conducted a gene enrichment and functional annotation analysis on the 87 genes with the functional annotation tools. As shown in Supplementary Table S4, among the 288 GO terms analyzed, our SAGE-detected candidate genes are specially enriched in 40 GO terms compared with the whole set of genes (30, 000) in the human genome, with at least 2-fold higher enrichment (P < 0.05) in each GO term, such as cell differentiation, cell development, defense response, apoptosis, and signal transduction. Remarkably, we found that our candidate genes enriched in some GO terms are either preferentially up-regulated (e.g., those in gene expression) or down-regulated (e.g., those in apoptosis or signal transduction) in our leukemia samples.

Therefore, we further compared the percentage of up-regulated genes with that of down-regulated genes enriched in each GO term. We found that, in 125 GO terms, the percentage of enriched up-regulated gene in the whole 32 up-regulated genes is at least 2-fold higher or lower than that of enriched down-regulated genes in the whole 55 down-regulated genes (see Supplementary Table S5). As exemplified in Fig. 2, up-regulated genes have a much higher percentage of enrichment in GO terms, such as gene expression, transcription, and regulation of transcription compared with down-regulated genes; opposite is true in GO terms, such as apoptosis, cell death, extracellular space, immune response, response to stress, signal transduction, and cell communication (see Supplementary Table s6 for the detailed gene lists). For example, HOXA5, HOXA9, HOXA10, MEIS1, MYB, SMARCA2, SMARCA4, RPUSD1, CDK6, RPL7A, RRAGC, and SOX4 are up-regulated genes that are involved in gene expression, whereas MCL1, LITAF, STK17B, FCER1G, TNFRSF1B, PTPRC, MMP9, ANXA1, and CYFIP2 are down-regulated genes that are involved in cell death and apoptosis (see Supplementary Tables S5 and S6).

Down-regulation of some candidate genes is likely associated with DNA methylation. As shown in Fig. 1A, the number of down-regulated genes is much greater than that of up-regulated genes (56 versus 32). The epigenetic silencing of tumor suppressor genes is a common event during carcinogenesis and often involves aberrant DNA methylation. To investigate whether DNA methylation is involved in the down-regulation of candidate genes in the MLL rearrangement leukemias, we performed MSP assay (22) to detect the status of CpG islands of three significantly down-regulated genes, including LTF, G0S2, and TGFBI. As shown in Fig. 3A, the CpG islands of all three tested genes are hypermethylated in three human ALL cell lines bearing MLL-ENL fusion (i.e., KOPN1, KOCL33, and KOCL51). After the treatment of 5-Aza-CdR on these three cell lines, the CpG islands of the three genes were partially demethylated (see Fig. 3B), and the expression of the three genes was dramatically up-regulated/restored with a >10-fold change (Fig. 3C). Indeed, a similar up-regulation of these three genes was also observed in five other MLL rearrangement leukemia cells lines, including both ALL (i.e., KOCL45, KOCL69, and SEM) and AML (KOCL48 and MV4-11) cell lines, after the demethylating treatment (see Fig. 3C). Therefore, the down-regulation of these three candidate genes in MLL rearrangement leukemia cells is likely associated with DNA hypermethylation.

Putative miRNA regulators of the 88 candidate genes detected by SAGE. One of the most exciting findings in life sciences in recent years is the discovery of an abundant class of small (∼22 nucleotides) noncoding RNAs, namely miRNAs. Recent studies showed that miRNAs play important regulatory roles in diverse biological processes and may function as oncogenes or tumor suppressors (23–25). Using the four currently available major prediction programs, including TargetScanS (26), Miranda (27), PicTar (28), and MAMI,⁷

we found that all the 88 candidate genes have putative miRNA regulation sites in their 3′ UTRs (see Supplementary Table S7).

A miRNA cluster is abnormally overexpressed in both human and murine leukemia samples with MLL-ELL or MLL-ENL fusion. We recently performed a large-scale, genome-wide miRNA expression profiling assay and observed that all the seven individual miRNAs from a unique polycistronic miRNA cluster, namely mir-17-92, residing in the C13orf25 gene locus at 13q31, including miR-17-5p, miR-17-3p, miR-18a, miR-19a, miR-20a, miR-19b, and miR-92, were significantly up-regulated in MLL rearrangement leukemias relative to normal controls or other leukemias (20). The mir-17-92 is a well-known oncogene (29). To validate whether this miRNA cluster is also abnormally overexpressed in murine leukemia cells with MLL-ELL or MLL-ENL fusion, we performed a qPCR to evaluate the expression levels of all seven individual miRNAs in 33 human and 7 mouse samples. The 33 human samples include 18 MLL rearrangement and 9 non–MLL rearrangement leukemia samples and 6 normal controls. The seven mouse samples include two MLL-ELL and two MLL-ENL leukemia samples and three Gr-1+ normal control samples. As shown in Fig. 4A, as in human samples, all seven individual miRNAs are overexpressed in mouse leukemia cells, although the level is not as high as in corresponding human leukemias.

Candidate targets of the miRNAs in the mir-17-92 cluster. It is believed that the central role of miRNAs is to regulate the expression of their target genes (23–25); thus, it is important to identify their targets. Because miRNAs usually function as negative regulators, it is expected that they will exhibit an inverse correlation with the expression of their targets. Recent findings indicate that animal miRNAs can not only repress protein synthesis but also induce mRNA degradation of a large portion of targets (30, 31). To identify targets of the miRNAs in the mir-17-92 cluster, we searched for candidate genes that were significantly down-regulated in MLL rearrangement leukemias as being reported by at least two different studies (including all the previous 35 studies and our SAGE analysis) and/or having been confirmed by our qPCR assay. We obtained a collection of 167 such candidate genes and found that 19 of them were predicted as putative targets of the miRNAs in the mir-17-92 cluster by at least two of the four prediction programs described above (see Supplementary Table S8). Of these, three genes (i.e., APP, RASSF2, and SH3BP5) have already been confirmed by our qPCR assay to be significantly down-regulated in both MLL-ELL and MLL-ENL samples (see Fig. 1B).

We further performed a luciferase reporter and mutagenesis assay to validate APP and RASSF2. A significantly negative effect (P < 0.01; paired t test) on luciferase activity was observed in the presence of miR-17 on 3′ UTR of APP or RASSF2, and such repression was lost when the predicted miRNA binding site in the 3′ UTR of APP or RASSF2 (in the reporter plasmid) was mutated (Fig. 4B), indicating that both APP and RASSF2 are direct targets of miR-17.

The power of comparative genomic analysis relies on the assumption that important biological properties are often conserved across species (9). Cross-species sequence comparison has been widely used to infer gene function, but it is becoming apparent that sequence similarity is not always proportional to functional similarity (32, 33). To determine the function of a gene precisely, therefore, we need to investigate not only its sequence characteristics but also its expression characteristics (9). The expression pattern of a gene can thus serve as a sensitive indicator of its function. In the present work, we conducted a genome-wide gene expression profiling assay on both human and mouse leukemia cells bearing MLL fusions using the SAGE method. We obtained 484,303 total SAGE tags for the nine samples and a total of 103,899 unique SAGE tags in the human and 60,993 in the mouse samples. We identified 88 genes that seemed to be significantly deregulated (32 up-regulated and 56 down-regulated) in both human and murine MLL-ELL and/or MLL-ENL leukemia (see Fig. 1A).

In a comparison with previous studies of human MLL rearrangement leukemias that used different methods, such as cDNA microarray, oligonucleotide microarray, and qPCR, we found that 31 of the 88 genes identified by SAGE have been reported previously but 57 are previously undescribed candidate genes that seem to be deregulated in MLL rearrangement leukemias. We further performed a large-scale qPCR assay to validate the deregulation of 81 candidate genes, including 43 genes identified from our SAGE analysis and 38 genes reported only by the previous studies in 20 samples, including 12 human and 8 mouse samples. Eighty-four percent (36 of 43) of the candidate genes detected by SAGE were confirmed, as were 54% (31 of 57) of the genes identified by previous studies. A lower confirmation/validation rate of the genes identified by previous studies relative to genes identified by our SAGE analysis is not surprising, because most of the previous studies compared MLL rearrangement leukemias with other acute leukemias, not with normal controls, and our qPCR focused on MLL-ELL and MLL-ENL samples that might not be included in a large number of previous studies. The fact that only 31 of the 88 genes detected by SAGE were reported previously and 84% of the tested SAGE candidate genes were confirmed by qPCR indicates that, although MLL rearrangement leukemias have been intensively studied previously, there are still many potentially important genes that have not been identified; thus, our study, using a different strategy/method, could provide additional important information to extend our understanding of the complex genetic alterations in MLL fusion-induced leukemogenesis.

Remarkably, HOXA5, HOXA9, HOXA10, and the HOX cofactor MEIS1 are the four most up-regulated genes observed in both types (i.e., MLL-ELL and MLL-ENL) of human and murine leukemia samples (Fig. 1A), which were further confirmed by our large-scale qPCR assay (Fig. 1B). HOX genes and MEIS1 are the best studied downstream targets of MLL and MLL fusion proteins, and their aberrant overexpression has been consistently recognized as the hallmark of MLL rearrangement leukemias (34–41). Interestingly, in a gene functional classification analysis using the DAVID tools (19), we found that HOXA5, HOXA9, HOXA10, and MEIS1 were clustered with four other candidate genes, including MYB, SOX4, KLF4, and MLL5; the first six were up-regulated whereas the last two were down-regulated in MLL-associated leukemias. Notably, all of these eight genes are regulators of gene expression and/or cell cycle; the first six genes (41–44) are positive regulators, whereas KLF4 (45) and MLL5 (46) are negative regulators. Thus, their up-regulation or down-regulation would promote gene expression and cell proliferation, which in turn would contribute to the development of leukemia.

LTF, LCN2, MMP9, S100A8, S100A9, PADI4, TGFBI, and CYBB are the eight most down-regulated genes (each has at least 30-fold change in expression) in both types of human and murine leukemia samples (Fig. 1A) and were also confirmed by our qPCR assay (Fig. 1B). Interestingly, five (i.e., LTF, MMP9, S100A8, S100A9, and TGFBI) of these eight genes are enriched in a GO term, namely extracellular space, and the other two genes (i.e., TNFAIP2 and LYZ) enriched in this term are also down-regulated in MLL-associated leukemias. In addition, LTF, S100A8, S100A9, and CYBB are involved in both response to stimulus and defense response, along with TGFBI involved in the former (see Supplementary Tables S5 and S6).

As shown in Fig. 2, up-regulated genes have a much higher percentage of enrichment in GO terms related to gene expression and transcription, whereas down-regulated genes are more enriched in GO terms related to apoptosis, signal transduction, and response. Thus, the up-regulation of genes responsible for gene expression and transcription but the down-regulation of genes responsible for apoptosis, signal transduction, and response can promote cell proliferation and inhibit apoptosis and, thereby, contribute to the development of leukemia.

It is well known that MLL fusion proteins can promote expression of many target genes, such as HOX genes and MEIS1 (3, 4). However, the mechanism underlying the down-regulation of a large number of genes in MLL rearrangement leukemias is unclear. In the present study, we showed that down-regulation of many genes in the leukemias might be associated with DNA methylation (see Fig. 3). Nonetheless, the genetic or epigenetic changes that caused the DNA methylation remain to be explored. Interestingly, we found that BMI1 was consistently overexpressed in human and murine MLL-associated leukemia cells relative to the normal controls (see Fig. 1B). BMI1 is a member of polycomb group proteins, which act as negative regulators of transcription (47). Thus, it is important to conduct systematic studies in the future to determine whether down-regulation of many genes in MLL-associated leukemias is related to the deregulation of BMI1 and/or other transcription repressor(s).

We found that all of the 88 candidate genes have putative miRNA regulators (see Supplementary Table S7). Interestingly, 19 putative targets of mir-17-92 seemed to be down-regulated in MLL rearrangement leukemias, as reported previously or detected by our SAGE assay; the down-regulation of APP, RASSF2, and SH3BP5 and the up-regulation of the mir-17-92 cluster have been confirmed by our qPCR. We have confirmed both APP and RASSF2 as direct targets of miR-17 through luciferase reporter and mutagenesis assay (see Fig. 4B). Interestingly, RASSF2 encodes a protein that contains a Ras association domain and is a novel tumor-suppressor gene, which has been observed to regulate Ras signaling and play a pivotal role in the early stages of colorectal tumorigenesis (48). Therefore, miRNAs may play an important role in the development of MLL rearrangement leukemia. In addition, recent studies also showed that some miRNAs, e.g., the miR-290 cluster, may control de novo DNA methylation through regulating Rbl2-dependent Dnmt expression (49, 50). Thus, besides the direct regulation of the expression of target genes, some miRNAs may also contribute to the DNA methylation of many genes and, thereby, regulate expression of these genes indirectly. Therefore, a combined study of miRNA and mRNA expression profiles would provide a more complete understanding of the complexity of molecular networks in diseases, such as MLL rearrangement leukemia.

In all, our study showed that the deregulation patterns of many protein-coding (e.g., HOX genes and MEIS1) and noncoding (e.g., individual miRNAs within the mir-17-92 cluster) genes observed in human leukemia cells are also conserved in mouse leukemia cells, highlighting the importance of these genes in leukemogenesis. The identification and validation of consistent changes of gene expression in human and murine MLL rearrangement leukemias contribute important information leading to the understanding of the complex functional pathways underlying MLL-associated leukemia, which in turn may lead to more effective therapy for this type of leukemia.

No potential conflicts of interest were disclosed.

Grant support: NIH grant CA127277 (J. Chen), Cancer Research Foundation Young Investigator Award (J. Chen), G. Harold and Leila Y. Mathers Charitable Foundation (J. Chen), Leukemia and Lymphoma Society Translational Research Grant (J.D. Rowley), Spastic Paralysis Foundation of Illinois, Eastern Iowa Branch of Kiwanis International (J.D. Rowley), NIH grants RO1CA104449 and P01CA105049 (M.J. Thirman), Leukemia and Lymphoma Society Scholar Award (M.J. Thirman), Leukemia and Lymphoma Society Specialized Center of Research (M.J. Thirman), family of Robert A. Chapski (M.J. Thirman), and NIH grant HG002600 (S.M. Wang).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. Y. Sato and K. Sugita for kindly providing KOPN1, KOCL33, KOCL45, KOCL48, KOCL51, and KOCL69 cell lines.