Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Molecular Targets for Therapy

The DNA methyltransferase inhibitors azacitidine, decitabine and zebularine exert differential effects on cancer gene expression in acute myeloid leukemia cells


The three DNA methyltransferase (DNMT)-inhibiting cytosine nucleoside analogues, azacitidine, decitabine and zebularine, which are currently studied as nonintensive therapy for myelodysplastic syndromes and acute myeloid leukemia (AML), differ in structure and metabolism, suggesting that they may have differential molecular activity. We investigated cellular and molecular effects of the three substances relative to cytarabine in Kasumi-1 AML blasts. Under in vitro conditions mimicking those used in clinical trials, the DNMT inhibitors inhibited proliferation and triggered apoptosis but did not induce myeloid differentiation. The DNMT inhibitors showed no interference with cell-cycle progression whereas cytarabine treatment resulted in an S-phase arrest. Quantitative methylation analysis of hypermethylated gene promoters and of genome-wide LINE1 fragments using bisulfite sequencing and MassARRAY suggested that the hypomethylating potency of decitabine was stronger than that of azacitidine; zebularine showed no hypomethylating activity. In a comparative gene expression analysis, we found that the effects of each DNMT inhibitor on gene transcription were surprisingly different, involving several genes relevant to leukemogenesis. In addition, the gene methylation and expression analyses suggested that the effects of DNMT-inhibiting cytosine nucleoside analogues on the cellular transcriptome may, in part, be unrelated to direct promoter DNA hypomethylation, as previously shown by others.


The therapeutic activity of three cytosine nucleoside analogues, 5-azacytidine (azacitidine, 5AC), 5-aza-2′-deoxycytidine (decitabine, DAC) and pyrimidin-2-one β-ribofuranoside (zebularine, Zeb), in acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS) has been the focus of numerous recent studies.1, 2, 3, 4 When applied at low doses, the compounds reduce genomic DNA methylation as a consequence of their irreversible binding to DNA methyltransferases (DNMT) after incorporation into newly synthesized DNA.5 Promoter DNA methylation is a major epigenetic mechanism of gene regulation and usually associated with gene silencing.6 Interference with aberrant DNA methylation, leading to the reactivation of silenced tumor suppressor genes, accounts, at least in part, for the DNMT inhibitors’ antitumoral effects.7 At higher concentrations, the hypomethylating activity of DNMT-inhibiting cytosine nucleoside analogues is overlaid by a nonspecific cytostatic effect comparable to that of the non-DNMT-inhibiting cytosine analogue cytarabine (AraC). Although clear differences exist between the DNMT-inhibiting cytosine nucleoside analogues with respect to cellular metabolism, enzyme specificity and nucleic acid incorporation, the agents are often viewed as interchangeable. Systematic comparisons of 5AC, DAC and Zeb in myeloid leukemia have yet to be reported. We hypothesized that the differences in pharmacokinetic properties might translate into diverse molecular activity of the DNMT inhibitors. In the study reported here, we used the Kasumi-1 AML cell line, which has been previously described as a very suitable model for studying comparative effects of DAC and AraC,8 to compare the in vitro activity of 5AC, DAC and Zeb and to investigate possible differences between the DNMT inhibitors and low-dose AraC. We examined how 5AC, DAC and Zeb at equivalent and low toxic concentration affect the proliferation, cell cycle, apoptosis and differentiation of Kasumi-1 cells. The DNA-hypomethylating capacity of each substance was quantified at fifteen independent loci and at genome-wide LINE1 sequences. Changes in gene expression after treatment with each DNMT inhibitor or AraC were studied globally and for selected single genes. We report that under equivalent and clinically relevant treatment schedules the three DNMT-inhibiting cytosine nucleoside analogues left reproducible and remarkably distinct ‘footprints’ in the gene expression profile of the cells, providing evidence that these drugs are not identical in function, and that their activity is clearly different from that of AraC.

Materials and methods

Cell cultures

The Kasumi-1 cell line9 was obtained by courtesy of Olaf Heidenreich (Newcastle University, UK). To confirm the identity of the cells, the presence of an AML1-ETO rearrangement was demonstrated by reverse transcriptase (RT)–PCR as described elsewhere.10 Cells were maintained in RPMI 1640 medium supplemented with 10% fetal bovine serum, penicillin/streptomycin and L-glutamine. AraC, 5AC, DAC and Zeb (Sigma-Aldrich, Seelze, Germany) were dissolved in phosphate-buffered saline. Cultures were treated for 72 h with three additions of substance at 24-h intervals. Cells were counted microscopically after standard Trypan blue staining.

Flow cytometry

For the evaluation of cytotoxic effects, cells were kept unstained and analyzed for forward- and side-angle light scatter on a FACScan cytometer (BD Biosciences, Heidelberg, Germany) using CellQuest software. In two-dimensional dot plots (Figure 1a), two distinct populations are seen, which represent normal Kasumi-1 cells (region R1 in Figure 1a) and pyknosis (region R2) as a result of spontaneous apoptosis or toxicity. Detection of early and late apoptotic cells was performed using dual staining with Annexin V-FITC and propidium iodide (both from BD Biosciences). For cell-cycle analysis, cells were fixed in cold 70% ethanol, washed in phosphate–citrate buffer, treated with ribonuclease and stained with propidium iodide. Myeloid and monocytic differentiation was assessed using CD11b-PE and CD14-PE antibodies (BD Biosciences); peripheral blood leukocytes from healthy donors were used as positive control.

Figure 1

(a) Cytotoxic effect of AraC, 5AC, DAC and Zeb on Kasumi-1 cells. Treatment-induced changes in cell size and granularity after treatment were assessed by measuring forward- and side-angle light scatter in a flow cytometer. The flow diagrams show one representative experiment (three replicates). (b) Inhibitory effect on proliferation. A total of 4 × 106 cells were seeded in each culture. The cell count after 72 h of treatment is shown. Error bars in both panels indicate one standard deviation (three replicate experiments).

Gene expression profiling

Cultured cells were used to isolate total RNA with the Trizol reagent (Invitrogen, Karlsruhe, Germany), generate cDNA and prepare biotin-labeled cRNA hybridization solutions using Affymetrix reagents and protocols (Affymetrix, Santa Clara, CA, USA). The solutions were hybridized to HG-U133A oligonucleotide microarrays, which after staining with phycoerythrin-conjugated streptavidin were read using a laser confocal scanner (Affymetrix). Quality control included assessment of RNA integrity on a Bioanalyzer 2100 (Agilent, Böblingen, Germany) and a number of parameters provided by the microarray data acquisition software (present call rate, scaling factor and 3′:5′ ratio of GAPDH and ACTB). Quality control results were impeccable for all profiles obtained.

Gene expression data processing

Fluorescence intensities and detection calls were determined from the scanner image data using Affymetrix GeneChip Operating software. Expression values for 22 215 probe sets on the U133A microarray were obtained and normalized between arrays to a standard target value of 500. A probe set was excluded from further analysis if it was flagged as ‘absent’ on all arrays, or if its signal intensity exceeded 100 in less than 2 of 10 arrays. A total of 12 092 probe sets passed this initial filtering step, representing 7906 individual genes. Using dChip software, the 90% confidence interval of fold-change in intensity between untreated cells and each experimental condition was calculated for each probe set.11 One-way analysis of variance (ANOVA) was performed using the algorithm built into dChip software.12

Bisulfite sequencing

Genomic DNA (1–2 μg) was chemically modified with sodium bisulfite using the EZ Methylation kit (Zymo Research, Orange, CA, USA). Bisulfite-treated DNA was amplified in PCR with tumor protein 73 (TP73)-specific primers IndexTermTAGGAGAAGTGGGTGGTAAGTTTT and IndexTermAACTCCCTACTATCCCCCAAA. Primers for cyclin-dependent kinase inhibitor 2B (CDKN2B) were published previously.13 PCR products were gel-purified using GFX columns (GE Healthcare, Freiburg, Germany) and ligated into the pCR 2.1 vector (Invitrogen) according to the manufacturer's instructions. Six to nine clones per sample were sequenced using BigDye terminator chemistry (Applied Biosystems, Darmstadt, Germany) and capillary electrophoresis. Analysis of the obtained sequences confirmed a complete bisulfite reaction in all samples.


Quantitative DNA methylation analysis at single cytosine-phosphate-guanine (CpG) units was performed by MassARRAY as previously described.14 Briefly, 1 μg of genomic DNA was treated with sodium bisulfite, PCR-amplified, in vitro transcribed, cleaved by RNase A and subjected to matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Primer sequences for PCR amplicons are available on request. Methylation standards (0, 20, 40, 60, 80 and 100% methylated genomic DNA) and correction algorithms based on the R statistical computing environment were used for data normalization.

Quantitative reverse transcriptase-polymerase chain reaction

Total RNA was reverse transcribed using oligo-dT primers and Superscript II (Invitrogen). Quantitative real-time PCR was performed on Mastercycler ep realplex (Eppendorf, Hamburg, Germany) using the ABsolute QPCR SYBR Green reaction mix (Thermo Scientific, Hamburg, Germany) and CDKN1C primers IndexTermCTGACCAGCTGCACTCGGGGATTTC/GCCGCCGGTTGCTGCTACATGA, HOXA9 primers IndexTermAAAACAATGCTGAGAATGAGAGCG/TGGTGTTTTGTATAGGGGCAC, or HGF primers IndexTermCAGCATGTCCTCCTGCATCT/GCACATTGGTCTGCAGTATTCAC. For each reaction, an endogenous ACTB control reaction (primers IndexTermCCAAGGCCAACCGCGAGAAGATGAC/AGGGTACATGGTGGTGCCGCCAGAC) was run in parallel. All reactions were set up in triplicate. Standard dilution curves were generated using RNA from untreated Kasumi-1 cells. Quantification of transcript expression was performed according to the relative standard curve method.


Effects of four cytosine nucleoside analogues on AML cell proliferation, apoptosis, cell cycle and differentiation

As a first step, we carefully established equivalent and low toxic culture conditions for the three structurally related DNMT inhibitors (5AC, DAC and Zeb), and the non-DNMT-inhibiting cytosine derivative AraC. Kasumi-1 cells were treated for 72 h with substance freshly added at 24-h intervals. Low drug concentrations, which allow the cells to cycle actively, were tested as the drugs are S-phase specific and therefore require active nucleic acid synthesis.15 The substantial cellular toxicity associated with higher concentrations8 was undesirable because of secondary effects related to apoptosis that would interfere with the primary gene expression changes induced by the drugs. Furthermore, high-dose schedules of DNMT inhibitors are no longer developed clinically. To define equivalent and low toxic concentrations of the four substances, toxic effects were analyzed by flow cytometry. Figure 1a illustrates that the exposure to AraC at 10 nM, 5AC at 500 nM, DAC at 50 nM and Zeb at 50 μM led to comparable, mild cytotoxicity in Kasumi-1 cultures after 72 h, affecting 10–20% of cultured cells. The proliferative capacity of Kasumi-1 cells under these conditions decreased by approximately 25% (Figure 1b), with the exception of 5AC where the inhibition of proliferation was less pronounced. These concentrations, which were used for all subsequent experiments, are similar to the estimated plasma concentrations achieved in current therapeutic regimens,16 supporting the clinical relevance of the in vitro studies reported here.

We next examined the effects of 5AC, DAC, Zeb and AraC on cell-cycle phases, induction of apoptosis and myeloid differentiation. Exposure to AraC led to a pronounced arrest in S phase even at concentrations as low as 10 nM (Figure 2a). S-phase arrest is a known property of AraC in AML.17 By contrast, treatment with any of the three DNMT inhibitors did not cause relevant changes in cell-cycle progression. The assessment of early and late apoptosis by two-color stain with Annexin V-FITC and propidium iodide (Figure 2b) confirmed for all four drugs their known activity as inducers of apoptosis. However, in concordance with the degree of cytotoxicity (Figure 1a), the proportion of apoptotic cells was modest at the low concentrations used in these experiments, ranging from 3 to 10% above the spontaneous rate of apoptosis in Kasumi-1 cultures (Figure 2b). Finally, no substantial induction of granulocytic or monocytic differentiation was observed after low-dose treatment with any of the cytosine nucleoside analogues (Figure 2c and d).

Figure 2

Effects of AraC, 5AC, DAC and Zeb on cell cycle, rate of apoptosis and cell cycle of Kasumi-1 cells. (a) Cell-cycle distribution as measured by cellular DNA content based on propidium iodide incorporation. Error bars indicate one standard deviation (three replicate experiments). (b) Induction of apoptosis. Flow cytometry analysis of Annexin V positivity (early apoptosis) and propidium iodide incorporation (late apoptosis) after treatment is shown. (c, d) Induction of differentiation. The shaded histogram shows the level of granulocytic (c) and monocytic (d) antigen expression as measured by CD11b or CD14 positivity in flow cytometry. The black curve represents healthy peripheral blood leukocytes used as positive control.

Effects of four cytosine nucleoside analogues on DNA methylation

To assess the DNA-hypomethylating activity of 5AC, DAC and Zeb at the equivalent and minimally cytotoxic concentrations determined above, the methylation status of two gene promoters (TP73 and CDKN2B) in Kasumi-1 cells was quantified by bisulfite sequencing (Figure 3a). The average methylation of the TP73 promoter fragment was 69% in untreated cells. As expected, there was no reduction in TP73 methylation after treatment with AraC. DAC decreased TP73 methylation to 57% (one-sided t-test P=0.02), but no change in methylation was observed after the equivalent dose and schedule of 5AC or Zeb. The CDKN2B promoter fragment was 97% methylated in untreated cells, in line with previous data.8 Methylation was unchanged after AraC treatment, but 5AC led to marginal hypomethylation (average 92%, one-sided t-test P=0.03) and DAC induced modest hypomethylation (average 86%, one-sided t-test P=0.11). Zeb had no hypomethylating effect. Of note, the marginal extent of CDKN2B hypomethylation under the low-dose 5AC schedule used here is in good agreement with clinical in vivo observations.18 We next extended the methylation analysis to the promoters of 14 additional genes and to LINE1 sequences, which are unbiased for any particular locus and so provide a global measurement of DNA methylation. MassARRAY technology was used to quantify DNA methylation at the CDKN1C, CDKN2B, CEBPA, CXCR4, DLK1, ID1, ID2, ID3, MPO, NFAT2CIP, PPAP2A, PRG2, RAB13, SMAD7 and TP73 loci and at genomic LINE1 sequences (Figure 3b; Supplementary Figure 1). These genes, most of which are of special interest in myeloid leukemia, were chosen because of reported methylation-dependent regulation (CDKN2B, CEBPA, MPO, TP73), or because they were subject to transcriptional regulation in our experiments (CDKN1C, CXCR4, DLK1, ID1, ID2, ID3, NFAT2CIP, PPAP2A, PRG2, RAB13, SMAD7; Supplementary Table 1). Consistent with the results obtained from bisulfite sequencing, MassARRAY confirmed dense methylation at the TP73 and CDKN2B promoters (Figure 3b). The upstream CEBPA promoter showed 77% methylation in Kasumi-1 cells, as reported previously.19 In addition, the analysis identified the RAB13 and DLK1 genes as targets of DNA hypermethylation in Kasumi-1 cells. By contrast, the CXCR4, MPO, PRG2, ID1, ID2, ID3, NFAT2CIP, PPAP2A and SMAD7 promoter sequences showed no or little methylation (Supplementary Figure 1). At all hypermethylated promoters examined, the MassARRAY data demonstrated a clear hypomethylating effect of DAC, and to a lesser degree of 5AC. Unexpectedly, Zeb at equitoxic concentration had no influence on DNA methylation at any of the loci examined. This is, however, in agreement with bisulfite sequencing and with a previous report where Zeb at low toxic concentration caused a 25% reduction in clonogenicity of HL-60 cells but did not decrease DNA methylation.20 The analysis of LINE1 sequences confirmed on a global genomic level that treatment with DAC resulted in more pronounced hypomethylation than 5AC treatment; again, Zeb had no discernible effect (Figure 3b, bottom). Taken together, the data indicate that DAC is a more potent DNA-hypomethylating agent in Kasumi-1 cells than 5AC at equitoxic concentration, and both agents are more effective than Zeb.

Figure 3

(a) Hypomethylation of the tumor protein 73 (TP73) and cyclin-dependent kinase inhibitor 2B (CDKN2B) gene promoters in Kasumi-1 cells after treatment with AraC, 5AC, DAC or Zeb. Cytosine-phosphate-guanine (CpG) dinucleotide methylation was assessed by sequencing multiple cloned alleles obtained from PCR on bisulfite-treated genomic DNA. Each horizontal line represents an individual allele. Filled circles represent methylated CpG sites; open circles, unmethylated CpG sites; gray circles, no data. The position of each cytosine nucleotide relative to the transcription start site is indicated at the top. (b) MassARRAY analysis to quantify DNA methylation of TP73, CDKN2B, CEBPA promoter sequences and global LINE1 sequences in untreated Kasumi-1 blasts and in cells treated with AraC, 5AC, DAC or Zeb. Cultures were performed in sequential triplicates. Rows represent DNA from individual cultures. Columns represent a single CpG site or a combination of CpG sites. Color coding reflects the degree of methylation (blue, 100%; light green, 0%; gray, no data). Average methylation for each amplicon is displayed as bar chart.

Distinct changes in global gene expression induced by azacitidine, decitabine and zebularine

To further address the functional diversity of 5AC, DAC and Zeb, we used high-density oligonucleotide microarrays to perform global gene expression profiling of Kasumi-1 myeloblasts after 72-h treatment with each of the three DNMT inhibitors, AraC or saline only. To exclude passage-specific effects and to ensure the reproducibility of the changes observed, the cell cultures were repeated under completely identical conditions several passages later and the resultant RNAs were processed on separate microarrays. After exclusion of probe sets without detectable signal under any treatment condition (corresponding to unexpressed genes), we used the data of 12 092 transcripts for further analysis. These transcripts represent 7906 individual genes. An unsupervised variance analysis was performed to estimate the similarities or disparities in global gene expression after treatment with the three DNMT inhibitors and AraC. To this end, the data set of 12 092 transcripts and 10 arrays was submitted to one-way ANOVA with a P-value threshold set at 0.001. This procedure identified 306 transcripts (estimated false discovery rate=0.04), which were then subjected to a hierarchical cluster algorithm. A color-coded expression ‘heat map’ (Figure 4a) demonstrated that gene expression profiles obtained after treatment with AraC, 5AC, DAC and Zeb were reproducible across the two independent experiments carried out with cells from different passages. More importantly, the dendrogram showed that the expression changes induced by each substance were clearly distinct (Figure 4a). To confirm the differences in a second analytical approach, filter criteria were applied for the robust determination of differentially expressed transcripts after treatment with each cytosine analogue. We considered a transcript as differentially expressed if (1) the lower bound of the 90% confidence interval of fold-change between untreated and treated cells was greater than 1.5, and (2) the difference between the mean expression levels in untreated and treated cells was greater than 100. The combination of an absolute difference criterion and a confidence bound of fold-change minimized the rate of false positives generated by technical variation. Thus, we found that 63 transcripts were regulated by treatment with 5AC (0.52% of detectable transcripts), 136 by DAC (1.12%) and 24 by Zeb (0.20%) (Supplementary Table 1). AraC altered the expression level of 57 transcripts (0.47%). The seemingly low numbers reflect the fact that the filtering criteria were set to rather stringent thresholds to avoid artifacts, at the possible cost of losing subtle changes in the expression data set.

Figure 4

(a) Two-dimensional hierarchical cluster analysis of differentially expressed transcripts. A total of 306 probe sets (columns) were identified using one-way analysis of variance (ANOVA) according to drug treatment (rows; two replicate cultures for each condition are denoted #1 and #2) at a 0.001 P-value threshold (expected number of false positives, 12). The hierarchical cluster dendrogram was calculated using Pearson's correlation as a distance metric and the Centroid linkage method, as implemented in dChip software. Signal intensities of each probe set were normalized across samples and color-coded according to standard deviations above (red) or below (blue) the mean. (b) Venn diagrams of differentially expressed transcripts in Kasumi-1 cells after equivalent low-dose treatment with 5AC, DAC or Zeb. The number in each area of the diagrams indicates probe set targets that were differentially expressed between treated and untreated cells (left diagram, upregulated; right diagram, downregulated). Transcripts that showed a response also to treatment with AraC were excluded to account for cytotoxic effects unrelated to DNA methyltransferase inhibition.

We next used the expression data set from cytarabine-treated cells as control for cytotoxicity unrelated to interference with DNA methylation. The assumption that AraC induces changes in gene expression via mechanisms other than DNA methylation is in agreement with the methylation analysis above. Of the 57 transcripts regulated after exposure to AraC, 7 overlapped with those responding to treatment with 5AC, 11 overlapped with DAC and 4 with Zeb. Thus, after exclusion of transcripts that also responded to AraC, we noted 24, 85 and 9 transcripts that were exclusively upregulated by 5AC, DAC and Zeb (Figure 4b); and 17, 24 and 6 transcripts that were exclusively downregulated. Importantly, no transcript changed in opposite direction between treatments with different DNMT inhibitors. This supports that the data reflect specific treatment effects rather than random variation. It was interesting to note that no gene was affected by all three DNMT inhibitors (Figure 4b; Supplementary Table 1). Together, the results indicate that the effect of each DNMT inhibitor on the cellular transcriptome is distinct and that the particular effects of each substance are stronger than the overlap between any two substances.

It was of interest to determine whether the DNMT-inhibiting cytosine nucleosides had a preferential effect on genes that possess an upstream CpG island. We examined 134 probe sets with differential expression after treatment with any DNMT inhibitor but not AraC, corresponding to 115 known genes. Of these, 75 had a promoter CpG island, 6 had an intragenic CpG-rich region and 34 had no CpG island. In comparison, 41 known genes were regulated by AraC but not by any DNMT inhibitor. Of these, 29 genes had a CpG island whereas 12 did not. Thus, the proportion of CpG island-containing genes among all regulated genes was similar for the DNMT inhibitors and AraC, suggesting that gene expression changes induced by 5AC, DAC or Zeb need not exclusively be linked to CpG island methylation.

Two interesting observations emerged from a Gene Ontology analysis of the genes regulated in our experiments. Of 85 genes upregulated by DAC, eight had a function in cell adhesion (RAB13, CD4, SELL, TGFB1I1, PSTPIP1, LPXN, COL14A1 and DCHS1) compared to 2.4% genome-wide (P<0.001). This overrepresentation was not noted in genes downregulated by DAC or in genes regulated by any of the other substances. Of 28 genes upregulated by AraC, six were functionally linked to immune response (TNFSF10, FCGR2A, LY75, CD8A, FCGR2C and CXCR4) compared to 2.7% genome-wide (P<0.001). Although preliminary, these observations provide a potential hint to functional diversity of the cytosine nucleoside analogues.

Differential activity of DNMT inhibitors at the single-gene level

We reviewed the reported function of transcripts of which the array data indicated differential expression after treatment with DNMT inhibitors, but not after AraC. We focused on three genes involved in myeloid development or leukemogenesis (CDKN1C, HOXA9 and HGF). The mRNA regulation of all three genes was verified using quantitative reverse transcriptase-polymerase chain reaction (Q-RT-PCR; Figure 5a, top and center panels). The CDKN1C gene, which encodes the cell-cycle regulator p57KIP2 and has a canonical 5′-CpG island, was markedly upregulated following treatment with 5AC and DAC, but not AraC or Zeb. This is in line with previous publications reporting CDKN1C gene induction after treatment with DAC.20, 21 In parallel, we noted a small but significant decrease in CDKN1C promoter DNA methylation after DAC treatment (Figure 5a, left). The HOXA9 homeobox gene, which also has a 5′-CpG island, is an activated oncogene in a subset of AML22 and MLL-rearranged leukemias,23 but is specifically repressed in AML1-ETO-positive AML.24 In our experiments, HOXA9 was upregulated by treatment with DAC but showed no change after exposing the cells to any of the other substances (Figure 5a, center). The HGF gene encodes the hepatocyte growth factor/scatter factor cytokine that is involved in proliferation and survival pathways in leukemia.25 HGF showed fourfold induction after 5AC whereas it was upregulated only twofold by all other substances (Figure 5a, right). This effect was reproducible across multiple independent probe sets on the array (Supplementary Table 1). However, the HGF gene is not associated with a CpG-rich DNA region. Further genes of interest that were subject to differential regulation in response to the four cytosine nucleoside analogues include three members of the inhibitor of DNA-binding protein family (ID1, ID2 and ID3), which were previously identified as retinoic-acid-responsive genes in acute promyelocytic leukemia;26 a Ras-associated small GTPase (RAB13); a member of the epidermal growth factor-like family (DLK1), which was previously implicated in myelodysplastic hematopoiesis27 and LIMS1, which interferes with the ERK1/2 pathway and appears to regulate Bim-mediated apoptosis.28 Of these, ID1, ID2, ID3 and DLK1 have standard 5′-CpG islands, RAB13 has a short CpG-rich fragment in its 5′ region, and LIMS1 has no CpG island. The methylation analysis of these six genes led to two interesting observations. First, DAC and 5AC (but not AraC) selectively enhanced the expression of ID1/ID2/ID3, but these genes were unmethylated in untreated Kasumi-1 cells and after treatment with any drug (Figure 5b). Second, whereas both 5AC and DAC hypomethylated RAB13 with comparable efficiency, RAB13 expression was increased after DAC but not after 5AC treatment (Figure 5b). Such disparity was also noted in DLK1 (Figure 5b). Together, the data demonstrate that the three DNMT inhibitors, under otherwise identical conditions, exert differential effects on the expression of several genes relevant to the pathogenesis of leukemia. Importantly, the diversity may, in part, be unrelated to differences in DNA-hypomethylating properties of the drugs.

Figure 5

(a) Verification of drug-specific effects on expression and DNA methylation of the CDKN1C, HOXA9 and HGF genes. The top panel illustrates the expression level of each gene in Kasumi-1 cells as measured by U133A oligonucleotide array. The middle panel depicts relative expression levels as determined by quantitative real-time PCR, using RNA from untreated cells as calibrator to generate standard curves. ACTB was used as endogenous control gene to standardize RNA input. The bottom panel shows changes in DNA methylation at the CDKN1C promoter (HOXA9 or HGF, not done). Error bars represent one standard deviation. (b) Expression and DNA methylation of the ID1, ID2, ID3, RAB13, DLK1 and LIMS1 genes in untreated Kasumi-1 cells and after treatment with AraC, 5AC, DAC or Zeb, as measured by U133A oligonucleotide array (probe sets 208937_s_at, 213931_at, 207826_s_at, 202252_at, 209560_s_at and 212687_at). The LIMS1 gene, which has no cytosine-phosphate-guanine (CpG) island, was not tested for methylation. Error bars represent one standard deviation.


Epigenetic alterations in leukemia, as opposed to genetic lesions, are pharmacologically reversible and therefore constitute an attractive target for novel therapeutic approaches. We characterized the cellular effects of three different DNMT-inhibiting cytosine nucleoside analogues, 5AC, DAC and Zeb, on Kasumi-1 AML blasts and examined the global patterns of gene expression after exposure to each substance under equivalent schedules. The effects were compared to those of a non-DNMT-inhibiting analogue, AraC. The study, which is the first benchmark of three hypomethylating agents using genome-wide RNA expression technology, demonstrated that each substance produces distinct patterns of gene induction and repression. Several other groups have used a similar approach, that is, global expression profiling after exposure of tumor cells to DNMT-inhibiting substances, with the goal of pharmacologically unmasking methylated genes to identify novel targets for epigenetic regulation.29, 30 Although this was not the primary aim of the present study, the CDKN1C and HOXA9 genes emerged from our experiments as potential candidates for epigenetic repression in Kasumi-1 AML cells.

Some important differences in structure and cellular pharmacology between the DNMT-inhibiting cytosine nucleoside analogues may account for the diversity observed. 5AC and DAC have a nitrogen in place of a carbon at position 5 of the pyrimidine ring but Zeb does not. 5AC is a ribonucleoside that is incorporated mainly into RNA; DAC and Zeb are incorporated only into DNA. Before incorporation can occur, the substances require phosphorylation to render the respective nucleotide forms: DAC is converted by deoxycytidine kinase whereas 5AC and Zeb are substrates of uridine-cytidine kinase.31, 32 The inhibitory activity of Zeb is not specific for DNMTs; Zeb is also a strong inhibitor of cytidine deaminase.33 In addition, the kinetics of activation/inactivation and the intracellular half-life of metabolites differ between each cytosine analogue. In this study, morphologic cytotoxicity was used to carefully define equivalent concentrations of the four analogues before performing global expression studies. We wish to point out that the titrations were not aimed at equally effective reduction of global or gene-specific DNA methylation, because with the high degree of RNA incorporation that 5AC undergoes, such an approach would have entailed excessive cytotoxicity for this particular drug. Besides, one of the four drugs studied (AraC) does not have DNA-hypomethylating activity.

DNMT-inhibiting cytosine nucleoside analogues are capable of reactivating epigenetically silenced genes by irreversibly binding DNMT and thus depleting intracellular enzyme activity. Hence, exposure of leukemia cell lines with extensive CpG island hypermethylation to these drugs would be expected to result in increased expression of the majority of responsive genes, whereas few genes would be downregulated. As the set of genes susceptible to methylation-dependent regulation in a given cell line is invariant, the profile of regulated genes would be expected to be similar after treatment with any of the DNMT inhibitors, but different from the pattern of regulation after treatment with AraC. We observed to the contrary that the transcriptional changes after treatment with each DNMT inhibitor, although clearly reproducible in an independent repeat experiment, showed remarkably little overlap. Thus, the expression profiling data reported here lend further support to the notion that the widespread perception of 5AC, DAC and Zeb as interchangeable DNMT inhibitors should be critically reappraised.34, 35, 36 A considerable number of genes were expressed at lower, not higher, levels after DNA-hypomethylating treatment, in line with previous observations.37 Several genes (CXCR4, ID1, ID2, ID3, NFAT2CIP, PPAP2A, PRG2, SMAD7) were upregulated even though they lacked relevant promoter methylation. Conversely, dense methylation and partial hypomethylation by 5AC or DAC were observed at the TP73 and CDKN2B loci, but neither gene was transcriptionally reactivated. Both 5AC and DAC efficiently hypomethylated the RAB13 and DLK1 loci, but only DAC induced the expression of these genes. It appears that gene reactivation after treatment with DNA-hypomethylating drugs is not necessarily a consequence of DNA methylation changes directly at gene promoters, as discussed by others.30, 34, 38, 39 For example, it was recently demonstrated that epigenetic reactivation of a specific microRNA in myeloid leukemia cell lines was responsible for decreased C/EBP-α levels after DNMT inhibitor treatment.19

Some interesting considerations for the use of DNMT inhibitors in clinical practice arise from the observation that 5AC, DAC, Zeb and AraC have different transcriptional regulatory potential in leukemic cells. First, there has been some debate whether 5AC or DAC at low doses constitutes a novel therapeutic principle in MDS/AML, or whether both are merely an equivalent of low-dose AraC. The results reported here provide support for the view that the therapeutic activity of the DNMT-inhibiting cytosine nucleoside analogues is different from that of AraC. This is consistent with the clinical observation that MDS/AML patients with monosomy 7 respond remarkably well to 5AC or DAC but not to AraC.18, 40, 41 Second, it is conceivable that some patients with MDS or AML refractory to one DNMT inhibitor might still respond to one of the others, as was indeed reported on the basis of a small series of cases.42 Third, a promising prospect would be the use of DNMT-inhibiting cytosine analogues in combination, with the goal of reciprocally complementing their activity. In a previous report on preclinical testing of DNMT inhibitor combinations, the authors note that the utility of Zeb may not lie so much in its own DNMT-inhibiting properties but rather in its potential to enhance the antileukemic activity of DAC by slowing the intracellular degradation of the latter.20 This is in good agreement with our results; we also observed no DNA-hypomethylating effect of low-dose Zeb.

In conclusion, we provide evidence that the cytosine nucleoside analogues have mechanisms of action that are not identical with each other, and different from cytarabine. The cellular basis for their antineoplastic activity is probably not limited to DNA hypomethylation.


  1. 1

    Lübbert M, Minden M . Decitabine in acute myeloid leukemia. Semin Hematol 2005; 42: S38–S42.

    Article  Google Scholar 

  2. 2

    Silverman LR, McKenzie DR, Peterson BL, Holland JF, Backstrom JT, Beach CL et al. Further analysis of trials with azacitidine in patients with myelodysplastic syndrome: studies 8421, 8921, and 9221 by the Cancer and Leukemia Group B. J Clin Oncol 2006; 24: 3895–3903.

    CAS  Article  Google Scholar 

  3. 3

    Scott SA, Lakshimikuttysamma A, Sheridan DP, Sanche SE, Geyer CR, DeCoteau JF . Zebularine inhibits human acute myeloid leukemia cell growth in vitro in association with p15INK4B demethylation and reexpression. Exp Hematol 2007; 35: 263–273.

    CAS  Article  Google Scholar 

  4. 4

    Kuendgen A, Lübbert M . Current status of epigenetic treatment in myelodysplastic syndromes. Ann Hematol 2008; 87: 601–611.

    CAS  Article  Google Scholar 

  5. 5

    Hurd PJ, Whitmarsh AJ, Baldwin GS, Kelly SM, Waltho JP, Price NC et al. Mechanism-based inhibition of C5-cytosine DNA methyltransferases by 2-H pyrimidinone. J Mol Biol 1999; 286: 389–401.

    CAS  Article  Google Scholar 

  6. 6

    Jones PA, Baylin SB . The fundamental role of epigenetic events in cancer. Nat Rev Genet 2002; 3: 415–428.

    CAS  Article  Google Scholar 

  7. 7

    Yoo CB, Jones PA . Epigenetic therapy of cancer: past, present and future. Nat Rev Drug Discov 2006; 5: 37–50.

    CAS  Article  Google Scholar 

  8. 8

    Berg T, Guo Y, Abdelkarim M, Fliegauf M, Lübbert M . Differential p15/INK4B reactivation and growth inhibition induced by 5-aza-2′-deoxycytidine in AML1/ETO-positive and negative myeloid cells. Leuk Res 2007; 31: 497–506.

    CAS  Article  Google Scholar 

  9. 9

    Asou H, Tashiro S, Hamamoto K, Otsuji A, Kita K, Kamada N . Establishment of a human acute myeloid leukemia cell line (Kasumi-1) with 8;21 chromosome translocation. Blood 1991; 77: 2031–2036.

    CAS  Google Scholar 

  10. 10

    Kozu T, Miyoshi H, Shimizu K, Maseki N, Kaneko Y, Asou H et al. Junctions of the AML1/MTG8(ETO) fusion are constant in t(8;21) acute myeloid leukemia detected by reverse transcription polymerase chain reaction. Blood 1993; 82: 1270–1276.

    CAS  PubMed  Google Scholar 

  11. 11

    Li C, Hung WW . Model-based analysis of oligonucleotide arrays: model validation, design issues and standard error application. Genome Biol 2001; 2: RESEARCH0032.

    CAS  PubMed  PubMed Central  Google Scholar 

  12. 12

    Li C, Wong WH . Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc Natl Acad Sci USA 2001; 98: 31–36.

    CAS  Article  Google Scholar 

  13. 13

    Nguyen C, Liang G, Nguyen TT, Tsao-Wei D, Groshen S, Lübbert M et al. Susceptibility of nonpromoter CpG islands to de novo methylation in normal and neoplastic cells. J Natl Cancer Inst 2001; 93: 1465–1472.

    CAS  Article  Google Scholar 

  14. 14

    Ehrich M, Nelson MR, Stanssens P, Zabeau M, Liloglou T, Xinarianos G et al. Quantitative high-throughput analysis of DNA methylation patterns by base-specific cleavage and mass spectrometry. Proc Natl Acad Sci USA 2005; 102: 15785–15790.

    CAS  Article  Google Scholar 

  15. 15

    Momparler RL, Samson J, Momparler LF, Rivard GE . Cell cycle effects and cellular pharmacology of 5-aza-2′-deoxycytidine. Cancer Chemother Pharmacol 1984; 13: 191–194.

    CAS  Article  Google Scholar 

  16. 16

    Momparler RL . Epigenetic therapy of cancer with 5-aza-2′-deoxycytidine (decitabine). Semin Oncol 2005; 32: 443–451.

    CAS  Article  Google Scholar 

  17. 17

    Sampath D, Cortes J, Estrov Z, Du M, Shi Z, Andreeff M et al. Pharmacodynamics of cytarabine alone and in combination with 7-hydroxystaurosporine (UCN-01) in AML blasts in vitro and during a clinical trial. Blood 2006; 107: 2517–2524.

    CAS  Article  Google Scholar 

  18. 18

    Raj K, John A, Ho A, Chronis C, Khan S, Samuel J et al. CDKN2B methylation status and isolated chromosome 7 abnormalities predict responses to treatment with 5-azacytidine. Leukemia 2007; 21: 1937–1944.

    CAS  Article  Google Scholar 

  19. 19

    Hackanson B, Bennett KL, Brena RM, Jiang J, Claus R, Chen SS et al. Epigenetic modification of CCAAT/enhancer binding protein alpha expression in acute myeloid leukemia. Cancer Res 2008; 68: 3142–3151.

    CAS  Article  Google Scholar 

  20. 20

    Lemaire M, Momparler LF, Bernstein ML, Marquez VE, Momparler RL . Enhancement of antineoplastic action of 5-aza-2′-deoxycytidine by zebularine on L1210 leukemia. Anticancer Drugs 2005; 16: 301–308.

    CAS  Article  Google Scholar 

  21. 21

    Sato N, Matsubayashi H, Abe T, Fukushima N, Goggins M . Epigenetic down-regulation of CDKN1C/p57KIP2 in pancreatic ductal neoplasms identified by gene expression profiling. Clin Cancer Res 2005; 11: 4681–4688.

    CAS  Article  Google Scholar 

  22. 22

    Nakamura T, Largaespada DA, Lee MP, Johnson LA, Ohyashiki K, Toyama K et al. Fusion of the nucleoporin gene NUP98 to HOXA9 by the chromosome translocation t(7;11)(p15;p15) in human myeloid leukaemia. Nat Genet 1996; 12: 154–158.

    CAS  Article  Google Scholar 

  23. 23

    Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML, Minden MD et al. MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia. Nat Genet 2002; 30: 41–47.

    CAS  Article  Google Scholar 

  24. 24

    Lasa A, Carnicer MJ, Aventin A, Estivill C, Brunet S, Sierra J et al. MEIS 1 expression is downregulated through promoter hypermethylation in AML1-ETO acute myeloid leukemias. Leukemia 2004; 18: 1231–1237.

    CAS  Article  Google Scholar 

  25. 25

    Accordi B, Pillozzi S, Dell’Orto MC, Cazzaniga G, Arcangeli A, Kronnie GT et al. Hepatocyte growth factor receptor c-MET is associated with FAS and when activated enhances drug-induced apoptosis in pediatric B acute lymphoblastic leukemia with TEL-AML1 translocation. J Biol Chem 2007; 282: 29384–29393.

    CAS  Article  Google Scholar 

  26. 26

    Nigten J, Breems-de Ridder MC, Erpelinck-Verschueren CA, Nikoloski G, van der Reijden BA, van Wageningen S et al. ID1 and ID2 are retinoic acid responsive genes and induce a G0/G1 accumulation in acute promyelocytic leukemia cells. Leukemia 2005; 19: 799–805.

    CAS  Article  Google Scholar 

  27. 27

    Sakajiri S, O’kelly J, Yin D, Miller CW, Hofmann WK, Oshimi K et al. Dlk1 in normal and abnormal hematopoiesis. Leukemia 2005; 19: 1404–1410.

    CAS  Article  Google Scholar 

  28. 28

    Chen K, Tu Y, Zhang Y, Blair HC, Zhang L, Wu C . PINCH-1 regulates the ERK–Bim pathway and contributes to apoptosis resistance in cancer cells. J Biol Chem 2008; 283: 2508–2517.

    CAS  Article  Google Scholar 

  29. 29

    Suzuki H, Gabrielson E, Chen W, Anbazhagan R, van Engeland M, Weijenberg MP et al. A genomic screen for genes upregulated by demethylation and histone deacetylase inhibition in human colorectal cancer. Nat Genet 2002; 31: 141–149.

    CAS  Article  Google Scholar 

  30. 30

    Schmelz K, Sattler N, Wagner M, Lübbert M, Dörken B, Tamm I . Induction of gene expression by 5-Aza-2′-deoxycytidine in acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) but not epithelial cells by DNA-methylation-dependent and -independent mechanisms. Leukemia 2005; 19: 103–111.

    CAS  Article  Google Scholar 

  31. 31

    Lee T, Karon M, Momparler RL . Kinetic studies on phosphorylation of 5-azacytidine with the purified uridine-cytidine kinase from calf thymus. Cancer Res 1974; 34: 2482–2488.

    CAS  PubMed  Google Scholar 

  32. 32

    Ben Kasus T, Ben Zvi Z, Marquez VE, Kelley JA, Agbaria R . Metabolic activation of zebularine, a novel DNA methylation inhibitor, in human bladder carcinoma cells. Biochem Pharmacol 2005; 70: 121–133.

    CAS  Article  Google Scholar 

  33. 33

    Cheng JC, Weisenberger DJ, Gonzales FA, Liang G, Xu GL, Hu YG et al. Continuous zebularine treatment effectively sustains demethylation in human bladder cancer cells. Mol Cell Biol 2004; 24: 1270–1278.

    CAS  Article  Google Scholar 

  34. 34

    Liang G, Gonzales FA, Jones PA, Orntoft TF, Thykjaer T . Analysis of gene induction in human fibroblasts and bladder cancer cells exposed to the methylation inhibitor 5-aza-2′-deoxycytidine. Cancer Res 2002; 62: 961–966.

    CAS  PubMed  PubMed Central  Google Scholar 

  35. 35

    Nguyen CT, Weisenberger DJ, Velicescu M, Gonzales FA, Lin JC, Liang G et al. Histone H3-lysine 9 methylation is associated with aberrant gene silencing in cancer cells and is rapidly reversed by 5-aza-2′-deoxycytidine. Cancer Res 2002; 62: 6456–6461.

    CAS  PubMed  PubMed Central  Google Scholar 

  36. 36

    Stresemann C, Brueckner B, Musch T, Stopper H, Lyko F . Functional diversity of DNA methyltransferase inhibitors in human cancer cell lines. Cancer Res 2006; 66: 2794–2800.

    CAS  Article  Google Scholar 

  37. 37

    Gius D, Cui H, Bradbury CM, Cook J, Smart DK, Zhao S et al. Distinct effects on gene expression of chemical and genetic manipulation of the cancer epigenome revealed by a multimodality approach. Cancer Cell 2004; 6: 361–371.

    CAS  Article  Google Scholar 

  38. 38

    Karpf AR, Lasek AW, Ririe TO, Hanks AN, Grossman D, Jones DA . Limited gene activation in tumor and normal epithelial cells treated with the DNA methyltransferase inhibitor 5-aza-2′-deoxycytidine. Mol Pharmacol 2004; 65: 18–27.

    CAS  Article  Google Scholar 

  39. 39

    Cheng JC, Yoo CB, Weisenberger DJ, Chuang J, Wozniak C, Liang G et al. Preferential response of cancer cells to zebularine. Cancer Cell 2004; 6: 151–158.

    CAS  Article  Google Scholar 

  40. 40

    Rüter B, Wijermans P, Claus R, Kunzmann R, Lübbert M . Preferential cytogenetic response to continuous intravenous low-dose decitabine (DAC) administration in myelodysplastic syndrome with monosomy 7. Blood 2007; 110: 1080–1082.

    Article  Google Scholar 

  41. 41

    Burnett AK, Milligan D, Prentice AG, Goldstone AH, McMullin MF, Hills RK et al. A comparison of low-dose cytarabine and hydroxyurea with or without all-trans retinoic acid for acute myeloid leukemia and high-risk myelodysplastic syndrome in patients not considered fit for intensive treatment. Cancer 2007; 109: 1114–1124.

    CAS  Article  Google Scholar 

  42. 42

    Borthakur G, Ahdab SE, Ravandi F, Faderl S, Ferrajoli A, Newman B et al. Activity of decitabine in patients with myelodysplastic syndrome previously treated with azacitidine. Leuk Lymphoma 2008; 49: 690–695.

    CAS  Article  Google Scholar 

Download references


Grant support: Kind Philipp Foundation T237/15893/2006 (to CF); José Carreras Leukemia Foundation DJCLS R06/42f (to ML); Research Commission of the Faculty of Medicine of the University of Freiburg (to CF and ML); Deutsche Forschungsgemeinschaft (to RC). We thank Dr A Heinzmann for providing the Affymetrix facility and J Heinze for excellent technical assistance with capillary sequencing. We sincerely thank Dr Peter A Jones, Dr Richard Momparler and Dr James Downing for critically reading the paper and providing helpful comments.

Author information



Corresponding author

Correspondence to C Flotho.

Additional information

Supplementary Information accompanies the paper on the Leukemia website (

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Flotho, C., Claus, R., Batz, C. et al. The DNA methyltransferase inhibitors azacitidine, decitabine and zebularine exert differential effects on cancer gene expression in acute myeloid leukemia cells. Leukemia 23, 1019–1028 (2009).

Download citation


  • myeloid
  • DNA methylation
  • epigenetic therapy
  • 5-azacytidine
  • 5-aza-2′-deoxycytidine

Further reading


Quick links