Azacytidine Inhibits RNA Methylation at DNMT2 Target Sites in Human Cancer Cell Lines Free

The pharmacologic inhibition of DNA methylation plays an important role in the treatment of human cancers (1, 2). 5-Azacytidine (azacytidine) and the closely related compound 2′-deoxy-5-azacytidine (decitabine) are cytosine analogues that have been established as potent inhibitors of DNA methylation, both in preclinical models and in cancer patients (3). Both drugs have shown significant clinical benefits in a variety of clinical trials (4, 5) and have received Food and Drug Administration approval for the treatment of myelodysplastic syndrome, a preleukemic bone marrow disorder. It has also been shown in various studies that azacytidine and decitabine induce DNA demethylation in myelodysplastic syndrome patients (6–9), but DNA methylation biomarkers for the predicition of clinical patient responses remain to be established (3).

The human genome contains four DNA methyltransferase genes (10). DNMT1 is the major biochemical activity in protein extracts from various tissues and has been designated as the maintenance methyltransferase for the postreplicative maintenance of DNA methylation patterns. DNMT3A and DNMT3B are two closely related enzymes that have been termed de novo methyltransferases and that are generally considered to play a major role in the establishment of methylation patterns. It has been shown that DNMT1, DNMT3A, and DNMT3B can be effectively depleted by azanucleosides (11). The mechanism of inhibition is widely assumed to depend on the incorporation of the fraudulent azacytosine base into DNA, followed by covalent trapping and subsequent degradation of the DNA methyltransferase enzymes. As a consequence, DNA will become passively demethylated when cells continue to replicate in the presence of reduced levels of DNA methyltransferase.

Despite the progress that has been made in characterizing the DNA demethylation responses of azacytidine and decitabine, numerous studies have indicated that the mode of action of these drugs is not limited to DNA methylation inhibition (3). In this context, it is particularly noteworthy that the effects of the ribonucleoside azacytidine on the RNA methyltransferase activity of the DNMT2 enzyme have not yet been investigated. DNMT2 contains all the catalytic signature motifs of conventional (cytosine-5) DNA methyltransferases, but has comparably low DNA methyltransferase activity (12). More recently, it was shown that DNMT2 methylates tRNA^Asp (13), which suggested that the substrate specificity of the enzyme might be different from other DNMTs. We have now analyzed the molecular effects of azacytidine in human cancer cell lines and found that the drug inhibits RNA methylation at DNMT2 target sites.

Cell culture. HCT116 cells were cultured in McCoy's 5a medium supplemented with 10% FCS. HL-60, K562, and ML-1 cells were cultured in RPMI 1640 supplemented with 5% l-glutamine, 10% FCS, 100 U/mL penicillin G sodium, and 100 μg/mL streptomycin sulfate. Cells were treated with azacytidine (Sigma) or decitabine (Calbiochem), as indicated.

Dnmt2 expression analysis. RNA was extracted using the RNeasy RNA Isolation kit (Qiagen) and on-column DNA digestion was performed using the RNase-Free DNase Set (Qiagen). For quantitative reverse transcription-PCR (RT-PCR), 5 μg of RNA were reverse transcribed using Superscript III reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Each cDNA sample was analyzed in triplicate using the QuantiFast SYBR Green PCR kit (Qiagen). QuantiTect Primer Assays (Qiagen) were used for DNMT1 (Hs_DNMT1_1_SG, QT00034335) and DNMT2 (Hs_TRDMT1_1_SG, QT00049896). The housekeeping gene GAPDH was amplified using the primers GAPDH_fwd (ATTCCACCCATGGCAAATTC) and GAPDH_rev (TCTCGCTCCTGGAAGATGGT). Using glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a reference, relative quantification was performed using the E-method of the LightCycler 480 software (Roche). Primer specific efficiencies were calculated from serial cDNA dilutions using the second derivative maximum method.

For protein expression analysis, 50 μg of protein extracts were separated on a 10% SDS-polyacrylamide gel, followed by immunoblotting on nitrocellulose and visualization using an anti-DNMT2 antibody (Santa Cruz SC-20702, 1:1000). For RNAi experiments, knockdown of DNMT2 was performed using On-Target plus SMART pool siRNA (Dharmacon). On-Target plus siControl nontargeting siRNA (Dharmacon) was used as a control. Cells were transfected with 100 nmol/L siRNA using DharmaFect 4 (Dharmacon) and subsequently cultured for 72 h, according to the manufacturer's instructions.

tRNA bisulfite sequencing. RNA bisulfite sequencing of tRNA^Asp was performed as described previously (14), with the following modifications: tRNA^Asp was reverse transcribed using either the reverse transcription stem-loop primer 5′-CTCAACTGGTGTCGTGGAGTCGGCAATTCAGTTGAGTGGCTCCCCA-3′ (tRNA sequence indicated in italics) or random hexamers. cDNA was PCR amplified using the bisulfite primers 5′-TAGTATAGTGGTGAGTATT-3′ (forward) and 5′-CGGCAATTCAGTTGAG-3′ (reverse for stem loop primed) or 5′-CTCCCCATCAAAAAATTA-3′ (reverse for hexamer primed). 28S rRNA methylation was analyzed using the reverse transcription primer 5′-ATCTAAACCCAACTCACA-3′ and the bisulfite primers 5′-TTTTAAGTAGGAGGTGTT-3′ (forward) and 5′-ATCTAAACCCAACTCACA-3′ (reverse). For 454 sequencing, bisulfite sequence–specific PCR amplicons were gel purified and reamplified using primer pairs containing 454 linker sequences.

Metabolic activity assay. Cells were seeded into multiwell plates. Twenty-four hours after seeding, cells were treated with various concentrations of azacytidine and decitabine, respectively, for 24 h. Metabolic activity was determined by the CellTiter-Blue assay (Promega) for HCT116 cells and by the CellTiter-Glo assay (Promega) for HL-60, K562, and ML-1 cells. Assays were performed according to the manufacturer's instructions. Standard curves for dose response analysis were plotted in SigmaPlot, version 10.0 (Systat Software) using the four-parameter logistic function. Functions were plotted only when metabolic activity was reduced sufficiently to meet the “s” shape criterion for the concentration range used in this assay.

DNA methylation analysis. Genomic DNA was purified using the DNeasy Blood and Tissue kit (Qiagen). Global methylation levels were determined by capillary electrophoresis, as described previously (15).

The cytidine analogue azacytidine has found increasing use for the clinical treatment of myeloid leukemias (2). Although it is generally assumed that azacytidine acts through covalent trapping of DNMT enzymes and inhibition of DNA methylation, the mode of action of this drug is not clearly understood (3). Two previous studies have suggested that azacytidine also affects tRNA methylation (16, 17), but these observations have never been followed up in detail. However, these findings, together with the observation that DNMT2 catalyzes the methylation of tRNA^Asp (8), raised the possibility that DNMT2 might be inhibited by azacytidine.

The relevance of DNMT2 for tumorigenesis has not been investigated yet. DNMT2 mutant mice do not display obvious mutant phenotypes (8) and have not been tested for their tumor susceptibility. Because mouse DNMT2 has been shown to be predominantly expressed in testis (18), we needed to establish DNMT2 expression in human cancer cell lines. As an initial step, we used quantitative RT-PCR to determine DNMT2 mRNA expression in HCT116 colon carcinoma and MCF-7 breast cancer cells. The results showed that DNMT2 mRNA is expressed in both cell lines, but at lower levels than DNA methyltransferase 1 (Fig. 1A). We also analyzed DNMT2 protein levels by Western analysis of protein extracts from HCT116 cells using DNMT2-specific antibodies. This revealed a polypeptide of ∼40 kDa that could be effectively depleted after transfection of the cells with siRNAs directed against DNMT2, thus confirming the specificity of the antibody (Fig. 1B). DNMT2 protein expression levels were subsequently analyzed in a panel of human cancer cell lines representing different tumor entities and different tissues of origin. The results showed variable DNMT2 protein expression in these cell lines (Fig. 1C), with comparably high levels in K562 and MCF7 cells and very low levels in A549 and HepG2 cells.

A previous study has established cytosine 38 (C38) of tRNA^Asp (Fig. 2A) as the primary substrate of DNMT2 in various model systems (13). To determine the tRNA^Asp methylation pattern in HCT116 cells, we used RNA bisulfite sequencing (14). Sequencing of 11 independent clones from control HCT116 cells showed that tRNA^Asp was methylated at C38, C48, and C49 (Fig. 2B). When cells were transfected with control siRNAs, this methylation pattern was not significantly affected (Fig. 2B). Sequencing of 14 independent clones from HCT116 cells that had been transfected with siRNAs directed against DNMT2 showed a complete loss of methylation at C38 (Fig. 2C), consistent with the notion that C38 is methylated by DNMT2 (13). Interestingly, the results also indicated reduced levels of methylation at C48 (Fig. 2C), a residue that has been considered to be methylated by the TRM4 family of tRNA methyltransferases (19). This finding raises the possibility that DNMT2 can also methylate C48 in tRNA^Asp, or that the methylation at this position is indirectly influenced by the methylation at C38. In contrast, C49 methylation, which is catalysed by the TRM4 family of proteins (19), was not affected in cells lacking DNMT2 (Fig. 2C). Our results thus confirm a role of DNMT2 in the site-specific methylation of tRNA^Asp in human cancer cells.

In subsequent experiments, we used the same methodology to test for effects of azacytidine on tRNA^Asp methylation. When HCT116 cells were treated with 1 μmol/L azacytidine for 1 day, methylation of C38 was considerably reduced, whereas methylation of C49 seemed unchanged (Fig. 2D). Of note, parallel treatment with 1 μmol/L decitabine did not cause loss of methylation in tRNA^Asp (Fig. 2D), and the difference in C38 methylation between the azacytidine- and decitabine-treated samples was statistically significant (P = 0.038, Fisher's Exact test). This indicates that the ribonucleoside azacytidine is more effective in inhibiting C38 methylation of tRNA^Asp than the deoxyribonucleoside decitabine.

Azacytidine and decitabine are widely considered to have similar modes of action. However, differences in the cellular effects of these drugs have also been shown (20). tRNA methylation is generally assumed to regulate tRNA functionality and thereby influences the metabolic activity of cells. The selective inhibition of tRNA methylation by azacytidine thus raised the possibility that azacytidine and decitabine might also show differential effects on the metabolic activity of human cancer cell lines. We therefore treated HCT116 cells with various concentrations of azacytidine or decitabine for 24 hours and determined the metabolic activity of the cells. The results showed that azacytidine caused a distinct drug response at low micromolar concentrations, whereas decitabine showed a similar effect only at 50-fold higher concentrations (Fig. 3). We also performed metabolic activity assays in a set of three cell lines (HL-60, K562, and ML-1) representing myeloid leukemias, the most frequent indication for the clinical use of azacytidine. All three cell lines showed a clear response to azacytidine, and a substantially weaker response to decitabine (Fig. 3). These results are consistent with the differential tRNA methylation inhibition activities of azacytidine and decitabine and suggest that inhibition of RNA methylation might be an important aspect in the cellular response to azacytidine.

Having shown that azacytidine affects the metabolic activity of myeloid leukemia cells, we also sought to confirm azacytidine-mediated demethylation of tRNA^Asp in these cell lines. Bisulfite sequencing of tRNA^Asp from control K562 cells showed a methylation pattern that was similar to control HCT116 cells, with methylation at C38, C48, and C49 (Fig. 4A). Treatment with 1 μmol/L azacytidine for 3 days caused significant (P < 0.001, Fisher's Exact test) demethylation at C38 and also at C48 (Fig. 4A). To confirm this observation in an additional myeloid leukemia cell line and to analyze the specificity of azacytidine-induced demethylation in greater detail, we used 454 technology to sequence tRNA^Asp and 28S rRNA in HL-60 cells. This approach allowed us to analyze four methylation sites modified by three distinct RNA methyltransferases (Fig. 4B). The evaluation of several hundred sequence reads revealed that tRNA^Asp from HL-60 cells was methylated at C38, C48, and C49 (Fig. 4C), in agreement with the methylation pattern observed in other cell lines. Treatment with 500 nmol/L azacytidine for 3 days caused substantial demethylation of C38. Similar effects were also observed at C48, which is consistent with our results obtained in K562 cells and after siRNA-mediated knockdown of DNMT2 in HCT116 cells. In contrast, C49 methylation, which is mediated by the TRM4 family of proteins (19), was not affected by azacytidine (Fig. 4C). Similarly, C4417 of 28S rRNA, which is methylated by a member of the p120/SUN family of RNA methyltransferases (19), did not show any azacytidine-dependent methylation changes (Fig. 4C). This demethylation pattern is consistent with a selective inhibition of DNMT2 activity.

To further characterize azacytidine-mediated loss of tRNA methylation, we also compared tRNA demethylation and DNA demethylation dose responses. To this end, HL-60 cells were treated for 3 days with various concentrations of azacytidine and C38 methylation levels of tRNA^Asp were determined by the analysis of 100 to 600 sequence reads per data point. In parallel analyses, genomic cytosine methylation levels were determined by capillary electrophoresis. The results confirmed a concentration-dependent demethylation of C38 from 54% in untreated control cells to 21% in cells treated with 500 nmol/L azacytidine (Fig. 5A). At higher drug concentrations, C38 methylation levels were similar to control levels (Fig. 5A). It is possible that higher drug concentrations cause a general inhibition of the RNA metabolism and a reduction in the synthesis of new and unmodified tRNA molecules. In contrast to the clearly defined concentration optimum for tRNA^Asp demethylation at 500 nmol/L, significant DNA demethylation could only be observed at the highest test concentration (1 μmol/L; Fig. 5A), thus indicating that tRNA^Asp is more sensitive to azacytidine-induced demethylation than DNA.

We also analyzed how repeated 3-day treatment cycles with low concentrations (200 nmol/L) of azacytidine affect tRNA^Asp methylation. C38 demethylation was readily detectable after the first treatment cycle (Fig. 5B). This effect was lost after the third treatment cycle, when methylation levels were similar to control levels (Fig. 5B). When we used capillary electrophoresis to determine genomic DNA methylation levels under the same conditions, we did not observe any significant DNA demethylation, even after three treatment cycles (Fig. 5B). These findings further illustrate the sensitivity of tRNA^Asp methylation to azacytidine and establish (de)methylation of tRNA^Asp as a novel molecular biomarker for azacytidine responses in human cancer cell lines. The reasons for the inability of azacytidine to demethylate tRNA^Asp after repeated drug incubation are presently unclear and will have to be investigated in future studies. It is possible that repeated doses of low azacytidine concentrations cause a general inhibition of the RNA metabolism, comparable with the effects observed after a single dose with high drug concentrations.

The role of the DNA methyltransferases DNMT1, DNMT3A, and DNMT3B in tumorigenesis has been extensively analyzed (21). In contrast, there are no published studies on the role of the fourth member of this family, DNMT2, in human cancers. DNMT2 has been shown to have a prominent tRNA methyltransferase activity in various model systems (13), which separates this enzyme from other DNMT proteins. In addition, this finding also raises the possibility that altered DNMT2 activity might contribute to tumorigenesis through novel pathways that are related to RNA methylation. Although the precise function of DNMT2-mediated RNA methylation remains to be established, tRNA methylation has been implied in the regulation of tRNA folding and stability (22, 23). It is conceivable that misfolded and/or less stable tRNAs might impact the rate of protein synthesis in cancer cells. Our recent analysis of tRNA methylation patterns in DNMT2 mutant Drosophila melanogaster has uncovered additional enzyme substrates,¹

Azacytidine has found widespread laboratory use as an inhibitor of DNA methylation and our results show that the drug can also inhibit cytosine methylation in tRNA^Asp with considerable specificity. Our bisulfite sequencing results revealed efficient demethylation of tRNA^Asp at C38 and at C48, thus establishing two sites as biomarkers for monitoring DNMT2 function during or after azacytidine treatment. Other methylated cytosines, such as C49 of tRNA^Asp, and C4417 of 28S rRNA, which are methylated by other RNA methyltransferases, were only weakly affected by azacytidine. The use of a DNA methyltransferase-like mechanism for the methylation of tRNA substrates (24) might provide an explanation for the particular sensitivity of DNMT2 to azacytidine. Because our study only analyzed a limited number of RNA methylation sites, we cannot exclude the possibility that other RNA methyltransferases are also inhibited by azacytidine. The detection of such effects might also have been obscured by general drug effects on RNA metabolism, thus resulting in a reduced representation of newly synthesized RNA molecules. In this scenario, the bisulfite sequencing approach would only allow us to recover RNA molecules that had been synthesized and methylated before drug treatment.

Regardless of the precise mechanism of tRNA methylation inhibition, our results may prove to be useful for the clinical development of azacytidine. Over the past few years, azacytidine, and its deoxynucleoside variant, decitabine, have been developed as archetypal drugs for epigenetic cancer therapy (25). Both drugs have shown to be similarly effective in inducing DNA demethylation in cancer cell lines and in patients (6–9, 20), which has fostered the notion that the mode(s) of action of both drugs are highly similar. Our results indicate that azacytidine is more effective than decitabine in inhibiting tRNA methylation, and thus provide an important example for a cellular pathway that is selectively affected by one of the drugs. In addition, our results also suggest that azacytidine-induced tRNA demethylation can be detected with higher sensitivity than azacytidine-induced DNA demethylation. It will be important to characterize the effects of azacytidine on RNA methylation patterns in patients undergoing azacytidine therapy and to evaluate RNA methylation markers for their potential as molecular markers predicting clinical responses.

No potential conflicts of interest were disclosed.

Grant support: Deutsche Forschungsgemeinschaft (Priority Programme Epigenetics; M. Schaefer and F. Lyko).

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 Mark Helm and Bodo Brueckner for valuable discussions, Stephan Wolf for 454 sequencing, and Anne Arens and Berit Haldemann for bioinformatics support.