Epigenomics, microRNAs and leukemias

Recent advances in the molecular therapy of cancers are mostly due to improved classifications of tumor subtypes within a cancer entity. A classical example is acute myeloid leukemia (AML), a heterogeneous disease, initially diagnosed based on microscopy, resulting in the French–American–British classification. Further research, especially through the pioneering work of Dr Janet Rowley, led to the identification of recurring chromosomal abnormalities, thus improving risk assessment and providing a prognosis-oriented classification. These discoveries were accompanied by emerging sequencing technologies, adding more complexity to our understanding of leukemias by identifying common molecular aberrations, such as Fms-like tyrosine kinase 3 internal tandem duplication (FLT3-ITD) [1] and nucleophosmin (NPM1) mutations [2]. Accumulation of information about a disease entity is, in principle, a good thing, but creates a challenge to define homogeneous, biologically relevant and mutually exclusive entities as seen in the new WHO classification [3]. However, progress in AML therapy was mainly achieved through the intensification of therapy and the improvement of supportive care. This growing dilemma reflects the constantly growing complexity of cancer cells, whose pathophysiological degeneration affects all genetic regulatory layers, as shown for the epigenetic machinery, as well as post-transcriptional regulators such as microRNAs (miRNAs) [4]. Both research fields received a lot of attention and were quickly considered as important contributors to cancer development. In particular, the recent hype about miRNAs, tiny noncoding RNAs approximately 21–25 nucleotides in length, with regulatory functions in development and cell metabolism, generated high expectations for a better understanding of malignant cells. miRNA profiling has become increasingly simplified, and affordable for many researchers. This has resulted in a torrent of miRNA profiling papers covering many aspects of biology, including hibernation of ground squirrels [5]. Since Carlo Croce‘s group demonstrated the first direct link between miRNAs and leukemias in 2002 [6], deregulated miRNA expression could be demonstrated in almost every profiled cancer. This ‘profiling frenzy‘ led to the general assumption that miRNAs are important for almost every aspect of biology, and particularly cancer development. But, how true is this?

Considering that of all miRNA-related publications, almost 50% are review articles (including my own), research findings tend to be repeated. This raises questions about what is really known and whether we actually understand the role of miRNAs in non-malignant and malignant cells. Owing to the relatively easy access to bone marrow cells and their well-characterized hierarchy, hematopoiesis has proven to be a model system to study development and malignant transformation. Several in vitro differentiation studies highlighted a regulating role for miRNAs during lineage differentiation [7]. Despite the high expectations, in most cases single miRNAs seemed to have a fine-tuning role, regulating terminal differentiation. Only genetic depletion of several miRNAs, such as miR-17–92 [8], a cluster of six miRNAs, demonstrated, besides a postnatal lethality, a severe block in early B-cell differentiation. In contrast, loss of miR-223 [9] led to a rather mild phenotype with impaired granulocyte function and blocked terminal granulocytic differentiation.

With regards to hematological neoplasias, several miRNA expression signatures for various leukemias and leukemia subtypes have been described [10,11], but no miRNA or miRNA cluster has been shown to cause leukemia in vivo by its overexpression or downregulation. This might be mostly owing to multiple targets of a single miRNA, leading to a ‘dispersion effect‘ by knocking down multiple genes, but with low efficiency. These findings underscore the importance of an orchestrated expression of multiple miRNAs during lineage differentiation and leukemogenesis. But what is regulating the regulators?

So far, only fragmented knowledge about the regulation of miRNAs is available. Their expression is driven by multiple mechanisms and mainly depends on genomic location, transcriptional and post-transcriptional events. Besides post-transcriptional events, such as inefficient processing of miRNAs through RNASEN (Drosha) [12] or DICER1 (Dicer) [13], the location in the genome appears to be an important factor for miRNA deregulation in cancer, as many miRNAs are located in cancer-associated genomic regions [14]. However, the primary regulatory mechanism is probably the transcriptional control of miRNAs genes. In general, miRNAs can be hosted within introns as well as exons of a specific gene, or reside in intergenic regions. It has been shown that miRNAs are predominantly transcribed by RNA polymerase II [15]. As approximately half of the human promoter regions contain CpG-rich regions [16], it is not surprising that the expression of miRNAs is affected by promoter hypermethylation or global hypomethylation. In addition, miRNAs have been shown to directly control the expression of epigenetic effectors such as DNA methyl-transferases (DNMTs) and histone deacetylases (HDACs), as well as polycomb genes [17]. An excellent example is miR-223 [9], a miRNA that is exclusively expressed in myeloid cells and therefore has received a lot of scientific attention. miR-223 is thought to have a dual function, as genetic loss and retroviral overexpression showed opposite phenotypes by either impairing [9] or enhancing myeloid differentiation [18]. In order to balance its function, expression of miR-223 is controlled by several regulatory mechanisms, including an autoregulatory feedback loop [19], as well as epigenetic modifications [18]. Within this regulatory circuit, AML-ETO binds to the miR-223 promoter and recruits an epigenetic silencing complex consisting of DNMTs, HDAC and methyl-CpG-binding proteins, inhibiting pri-miR-223 transcription through CpG methylation. This, in turn, sustains the AML-specific block in myeloid differentiation, maintaining the characteristic AML phenotype. Indeed, several studies demonstrated that epigenetic factors control carcinogenesis not only by affecting the transcription of coding genes, but also of noncoding genes, such as miRNAs. Most of these recent studies were performed on solid tumor cell lines by treating them with DNMT-inhibiting agents [20] as well as HDAC inhibitors [21], followed by miRNA expression profiling or exploiting miRNA expression in DNMT1- and DNMT3b-knockout cell lines [22]. However, both approaches revealed increased expression levels for only 5–8% of all miRNAs, with a handful of miRNAs, such as miR-127 [20], miR-124a [22], let-7a-3 [23], miR-34b/c [24] and miR-9 [24]. Given the growing number of newly discovered miRNAs, the actual fraction of epigenetically regulated miRNAs could be significantly higher. As regulators of the epigenetic machinery, miRNAs add a powerful twist to the indirect control of gene expression on a broad scale. Specifically, re-expression of members of the miR-29 family within a lung cancer cell line led to disruption of de novo DNA methylation through post-transcriptional targeting of DNMT3a and DNMT3b [25]. This resulted in the re-expression of tumor suppressor genes, leading to apoptosis and, thus, impaired lung cancer growth in vivo. Other components of the epigenetic machinery also undergo posttranscriptional regulation through miRNAs, including HDACs through miR-1 [26], miR-140 [27] and miR-449a; as well as EZH2, a member of the polycomb complex, through miR-101 [28]. Depending on the transcriptional context of a cell, this could turn a miRNA into a tumor suppressor or oncogene. The fact that miRNAs interact on a bidirectional level with the epigenetic machinery layers even more complexity on top. For example, treatment with the DNMT inhibitor, 5-azacitidine, has been established as effective therapy in myelodysplastic syndrome (MDS) [29], despite the fact that chromatin remodeling drugs act on a global, nonspecific scale. It is therefore not possible to predict the direct and indirect effect of these epidrugs on the expression of most genes, including miRNAs. In turn, it has to be assumed that, most probably, miRNAs are already undocumented targets of epigenetic drugs in MDS [29], as exemplified for miR-145 and miR-146 in 5q-syndromes [30]. Other levels of regulation, namely the modulation of miRNA processing through epigenetic control of RNASEN and DICER1, could also be relevant and have been barely touched upon so far. Although DICER1 has been implicated in various biological processes, including centromeric silencing [31], no direct or indirect connection of epigenetic regulation of DICER1 or RNASEN has been reported yet. Understanding these complicated and rather unpredictable regulatory layers will refine our knowledge about hematological neoplasias, but increase the difficulty to create a rather homogeneous classification of complex diseases such as AML. However, as long as relevant treatment options emerge, the practical application rules over the theoretical knowledge.

Acknowledgments

I thank A Rouhi and VPS Rawat for critically reading the manuscript, as well as RK Humphries for his continuous support and guideship over the last years.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.