DNA methylation-independent loss of RARA gene expression in acute myeloid leukemia

Despite progress in the understanding of the molecular abnormalities in acute myeloid leukemia (AML), overall survival rates remain low, reinforcing the need for more effective therapies.¹  Among the subtypes of AML, acute promyelocytic leukemia (APL), which in most patients is associated with the translocation between the retinoic acid receptor (RAR) α (RARA) and PML genes, leading to expression of the PML-RARα fusion oncoprotein (other much less common rearrangements involving RARA have been described² ), responds uniquely to differentiation therapy with all-trans retinoic acid (ATRA).³  Given that PML-RARα acts to inhibit the positive effects of physiological ATRA on regulation of gene expression and myelomonocytic differentiation, these dramatic therapeutic effects seemed at first paradoxical. However, progress in understanding the molecular mechanisms through which RARs and other nuclear receptors regulate gene expression, which involves ligand-mediated exchange of corepressor for coactivator⁴  and postactivation receptor degradation,⁵  provided a platform for understanding how administration of pharmacologic levels of ATRA can restore RARα signaling and differentiation in APL.⁶  Nevertheless, the molecular basis for the general lack of response of non-APL AML to ATRA remains poorly understood.

We have previously demonstrated that expression of the RARα2 isoform increases with differentiation of murine hematopoietic progenitors along the myelomonocytic lineage and that AML cell lines, including ATRA-resistant APL cells, do not express this isoform effectively.^7,8  This lack of RARα2 expression and ATRA response may, at least in part, be a reflection of a general impairment of ATRA signaling, which may involve genetic factors and/or epigenetic mechanisms. To examine this hypothesis, we have evaluated in parallel expression of the RARA gene and DNA methylation of its promoters in normal hematopoietic stem/progenitor cells and primary AML blasts.

Primary cells and cell lines were maintained under standard conditions as previously described.⁷  Approval was obtained from the Multi-Center Research Ethics Committee (MREC) for Wales (United Kingdom); Erasmus Medical Center (Rotterdam, the Netherlands); and Montefiore Medical Center (Bronx, NY) institutional review boards for these studies based on strictly anonymous use of archived samples or informed consent obtained in accordance with the Declaration of Helsinki. Following approval from local Research Ethics Committees, AML bone marrows were taken with consent from anonymized patients at diagnosis. Umbilical cord blood samples from anonymized healthy full-term infants were obtained at the time of delivery, and peripheral blood samples from healthy adults were obtained by venipuncture. Mononuclear cells were separated by Ficoll (Biochrom, Berlin, Germany), and cells expressing CD11b, CD33, CD34, or CD133 were isolated with the corresponding magnetic beads (Miltenyi Biotec, Auburn, CA).

Given that ATRA signaling via RARA is required for optimal myelomonocytic differentiation of murine progenitor cells^8,–10  and is disrupted by the PML-RARα fusion protein in APL,⁶  we sought to examine whether diminished RARA gene expression may be a general pathologic feature of AML. We compared by real-time reverse transcription–polymerase chain reaction (RT-PCR) the basal expression levels of the 2 major RARα isoforms in primary AML cells with AML cell lines and normal cord blood (CB)–derived human stem/progenitor cells (CD33⁺ or CD133⁺) and found that in primary non-APL AML and APL cells, the RARα1 copy numbers were 22% and 3.8%, respectively, of the value obtained for CD33⁺ CB cells (Figure 1A left panel). This contrasted with the value for AML cell lines, which was 62% of that for CD33⁺ CB cells. The diminution of RARα2 expression levels was even greater in primary non-APL AML (2.1%) and APL (1.9%) cells when compared with CD33⁺ CB cells (Figure 1A right panel). These values correlate with the data from AML cell lines, which showed a mean RARα2 copy number of 1.1% of that for CD33⁺ CB cells. Similar differences in RARα1 and α2 expression were observed relative to CD133⁺ CB cells; the mean values for RARα1 and α2 mRNA copy numbers in CD133⁺ CB cells were 1182 and 2418, respectively (Figure 1A). Although it is difficult to define what a good “normal” control is for AML cells, in this study we have used normal CB CD33⁺ and CD133⁺ progenitor/stem cells because AML blasts are usually CD33⁺ and most AML blasts are generally considered to derive from a primitive progenitor compartments. Additional analysis of the expression of 2 known ATRA target genes,¹¹ CEBPE¹²  and CYP26A1,¹³  revealed that their mRNA levels are also diminished in patients with AML relative to those found in normal CD133⁺ cells, suggesting a causal relationship for deregulation of RARA and these downstream target genes (Figure 1B).

Epigenetic modifications, in particular aberrant promoter hypermethylation, are frequently associated with deregulation of gene expression in many tumor types.¹⁴  We therefore examined whether DNA methylation could account, at least in part, for the diminished RARα1 and/or RARα2 expression in AML. Sequencing of bisulphite-modified genomic DNA from AML cell lines revealed that while the RARα1 promoter was never methylated (Figure 2Ai), the CpG islands present in the RARα2 promoter and 5′ untranslated regions (5′ UTRs) were always hypermethylated (Figure 2Aii).

In order to investigate the mechanisms responsible for loss of RARA expression in primary AML cells, we examined methylation of the RARα1 promoter (Figure 2Ai) and RARα2 (Figure 2Aii) promoter and 5′ UTRs in APL and non-APL AML patient samples. Surprisingly, in contrast to the promoter methylation observed in AML cell lines, none of the AML (including APL) patient samples displayed methylation of CpGs in either the RARα1 promoter or RARα2 promoter/5′ UTR. We also analyzed expression of the human tumor antigen PRAME (PReferential Antigen MElanoma), which has been reported to repress ATRA signaling in cell lines and is frequently overexpressed in a variety of human cancers.¹⁵  Normal CD33⁺ CB cells did not express PRAME, and while the K562 cell line was positive, only 2 of 8 APL and 1 of 7 non-APL AML samples examined expressed PRAME (Figure S1, available on the Blood website; see the Supplemental Materials link at the top of the online article); thus, PRAME could not be responsible for the repression of ATRA signaling in patients with AML. It is possible that DNA methylation-independent epigenetic events, such as loss of positively acting and/or acquisition of negatively acting histone modifications may play a role in the down-regulation of RARα2 expression in AML. Consistent with this notion, in AML patient samples containing sufficient numbers of cells for chromatin immunoprecipitation analysis we found that, relative to normal CD33⁺ cells, the RARα2 promoter was associated with diminished levels of histone H3 acetylation and dimethylation of lysine 4, which are markers of transcriptional competence^19,20  (Figure 2B).

An important aspect of our findings is that RARA expression is strongly diminished in both APL and non-APL AML. These results are consistent with a model where, in APL, PML-RARα makes a specific contribution to this phenotype, while in non-APL AML subtypes, similar effects on the RARA gene expression result from PML-RARα–independent mechanisms. For example, a recent report indicates that AML1/ETO can also interfere with ATRA signaling.²¹  It is also worth noting that similar levels of RARα2 expression are observed in AML cell lines with or without PML-RARα (Figure 1A) and the RARα2 promoter is methylated in all the examined AML cell lines, indicating that methylation and silencing of the RARα2 promoter can occur in a PML-RARα–independent manner. Although PML-RARα has been shown to recruit DNA methyltransferases²²  and MBD1,²³  as well as other chromatin-modifying activities, our results demonstrate that in fresh APL blasts, DNA methylation is not associated with the loss of RARα2 expression.

A lack of consistent RARα2 promoter methylation in AML patient samples was recently reported by Chim et al.²⁴  This study did not, however, compare RARα2 promoter (or RARα2 5′ UTR) methylation status with expression of the RARα2 isoform. The authors did report the RARα2 promoter to be methylated in 40% of patients with APL, but samples were analyzed by methylation-specific PCR without any indication of the extent of promoter methylation; this technique can provide false-positive results. Methylation of the RARβ2 promoter (a RARα2 paralog) in AML has also been studied by a number of different investigators, but the results of these studies have been somewhat contradictory in nature.^{21,22,25,–27}  Moreover, RARB is not readily expressed in normal hematopoietic cells or required for proper hematopoiesis,^9,28  and RARβ2 promoter methylation has no effect on patient survival.²⁵  In fact, our data indicate that, in contrast to normal cells, RARβ2 is expressed in a fraction of AML samples (Figure S2). Therefore, RARA represents a more physiologically relevant target for the investigation of impaired ATRA signaling in AML. In this respect, it may be interesting to examine whether the level of RARA expression correlates with prognosis in these hematopoietic neoplasms.

Taken together, our data indicate that diminished expression of the RARA locus is characteristic of AML and may be relevant to its pathogenesis. Although it appears that demethylating agents such as decitabine do not act directly on RARs in patients with AML, this class of drugs is therapeutically active in AML and myelodysplastic syndrome (MDS), and remains a target for research into combination therapy.²⁹  The results of this study emphasize the need for a better understanding of the mechanisms that block ATRA signaling and gene expression in AML as well as for research into combinatorial therapies using epigenetic drugs that, by targeting multiple processes, may be most effective in reactivating ATRA-sensitive gene expression and differentiation of AML cells.

The authors would like to acknowledge support from the Leukemia Research Fund of Great Britain, the Samuel Waxman Cancer Research Foundation, and the Kay Kendall Leukemia Fund.

Contribution: A.G., A.B., and A.Z. designed the research; A.G., A.B., K.P., and M.B.-C. performed the research; D.-C.Z, D.G., R.G., M.v.L, S.W., T.E., and G.H. contributed vital new reagents or analytical tools; A.G. and A.B. collected data; A.G., A.B., K.P., and A.Z. analyzed the data; and A.G., K.P., and A.Z. wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Arthur Zelent, Section of Haemato-Oncology, Institute of Cancer Research, Brookes Lawley Building, 15 Cotswold Road, Sutton, Surrey SM2 5NG, United Kingdom; e-mail: arthur.zelent@icr.ac.uk.