Long noncoding RNAs as regulators of pediatric acute myeloid leukemia

Acute myeloid leukemia (AML) accounts for approximately 20% of acute leukemias in children [1]. Although the overall survival of children with AML has significantly increased as a result of intensified therapy, hematopoietic stem cell transplantation, and improved supportive care over the past decades, around 25% of all patients still cannot be cured [2] — highlighting the urgent need to transfer discoveries about the molecular features of pediatric AML into new therapeutic approaches. Among the recent scientific developments in this field, comprehensive studies have revealed that the molecular landscape of childhood AML is shaped not only by oncogenic mutations and cytogenetic alterations but also by global changes in DNA methylation and gene expression affecting both protein-coding genes and noncoding RNAs [3, 4]. Noncoding RNAs in particular are emerging as important regulators of hematopoiesis and leukemogenesis and represent a largely understudied space in the search for new therapeutic strategies.

Long noncoding RNAs (lncRNAs) — defined as transcripts longer than 200 nucleotides that lack open reading frames — represent the largest group of noncoding RNAs and constitute two-thirds of the human transcriptome [5]. Different structural domains enable their interaction with RNA, DNA, and proteins and thereby allow the regulation of every stage of gene expression. Apart from their versatile roles in gene regulation on every possible transcriptional and posttranscriptional level, lncRNAs can directly interact with signaling pathways and contribute to the function of organelles such as exosomes or mitochondria.

Based on the mechanistic interaction of lncRNAs with other molecules, four different archetypes of lncRNA functions — namely signal, decoy, guide, and scaffold — have been defined in a seminal work by Wang and Chang in 2011 [6]. As the first archetype, signal lncRNAs, which are under precise transcriptional control, act as a molecular signal reflecting a specific developmental stage, cellular background, or a response to stimuli [6‐9]. LncRNAs belonging to the second archetype, decoy, bind and titrate away regulatory proteins or RNAs, thereby repressing transcription or translation of a target gene [10, 11]. Guide lncRNAs, which represent the third mechanistic archetype, direct regulatory protein complexes, chromatin modifiers, or transcription factors to their target site, resulting in either transcriptional activation or repression of the respective genomic locus [12, 13]. The fourth archetype, scaffold, describes lncRNAs as a structural platform at which different bound components of protein complexes or ribonucleoprotein complexes are assembled or can interact with each other [14, 15]. Even though increasing evidence now points toward complex lncRNA mechanisms that represent rather coexisting or overlapping features of the four classical archetypes, these four mechanistic subtypes still illustrate the wide range of possible modes of action of lncRNAs.

Nuclear-localized lncRNAs can exert independent regulatory effects on neighboring genes (function in cis) as well as on distant genes (function in trans), while cytoplasmic lncRNA mechanisms include competitive miRNA binding and interaction with translational proteins [16]. The most commonly demonstrated function of cis-acting nuclear lncRNAs, which are operating at their own site of transcription, is regulation of gene expression and chromatin modification [17]. The participation of cis-acting lncRNAs in transcriptional processes is supported by the high abundance of lncRNA genes in the proximity of regulatory elements of the human genome such as enhancers and promotors [18]. Protein-coding genes that are involved in transcriptional regulation, e.g., genes encoding transcription factors or chromatin modifiers, show a higher enrichment of closeby lncRNA genes than protein-coding genes of other functional categories, further indicating an essential contribution of cis-acting lncRNAs to the regulation of gene expression [19]. LncRNAs can act in cis to either activate or repress the expression of nearby genes through a variety of mechanisms. For instance, lncRNAs can activate gene expression in cis by recruiting proteins that establish spatial interactions such as chromatin loops, thereby enabling closer contact of an enhancer to the respective protein-coding gene [20, 21]. Other lncRNAs have been shown to activate gene expression of their target genes in cis in an transcript-independent manner by the process of their own transcription and splicing through recruiting cofactors, accumulation of transcriptional proteins, and establishing of activating chromatin marks [22, 23]. Conversely, cis-acting lncRNAs can also recruit chromatin modifiers that repress transcriptional activity at their genomic locus, such as the polycomb repressive complex 2 [24, 25]. Another lncRNA mechanism that results in decreased expression of the neighboring target gene is transcriptional interference: The transcription of a lncRNA can interfere with the transcription of the adjacent gene by impeding recruitment of necessary proteins such as transcription factors and chromatin remodeling proteins or by increasing nucleosome density, thereby preventing transcription factor access [26, 27]. For trans-acting lncRNAs, diverse functions in the modulation of distant gene expression have been demonstrated, with most of the studied examples exhibiting mechanisms that have been also described in the context of cis-acting lncRNAs. For instance, lncRNAs can facilitate transcriptional activation of distant target genes by the initiation of chromatin loop formation [28]. Similarly, several lncRNAs have been shown to repress transcription of target genes in trans through recruitment of chromatin-modifying complexes [7, 9, 29]. Another mechanism that has been described for cis-acting lncRNAs as well as for trans-acting lncRNAs is the formation of RNA-DNA hybrids, the so-called R-loops, that are recognized by transcription factors or chromatin modifiers and thereby lead to the activation or repression of transcription of the target gene [30].

Given their versatile cellular and molecular functions, it is no surprise that lncRNAs are involved in many essential physiological processes such as genomic imprinting and differentiation, as well as in the pathogenesis of diseases such as cancer, neurodegenerative disorders, and metabolomic diseases [31‐33].

Hematopoietic differentiation is a tightly regulated, hierarchically ordered process coordinated by the expression of specific gene programs. Numerous lncRNAs have been characterized in the context of hematopoiesis including lncRNAs that are involved in hematopoietic fate decision and lncRNAs whose deregulation contributes to the malignant transformation of hematopoietic progenitor cells [34]. Recent studies identified unique stage- and lineage-specific lncRNA signatures in distinct blood cell populations indicating an important contribution of lncRNAs to the homeostasis and regulation of hematopoiesis [3, 35, 36]. Here, we review a selection of well-characterized lncRNAs that are involved at different levels of hematopoietic differentiation.

Fetal lncRNA H19 is one of the best-characterized lncRNAs in embryonic development and tumorigenesis. Physiologically downregulated after birth, H19 is expressed in almost every type of human cancer [37, 38]. During embryonic development, the lncRNA facilitates the transition from endothelial cells to hematopoietic stem cells (HSCs), whereas in adult hematopoiesis, it is essential for maintaining HSC quiescence, thereby regulating the long-term homeostasis of HSCs [39, 40]. In addition, H19 is overexpressed in AML and correlates with poor prognosis. In vitro knockdown of H19 leads to decreased proliferation and increased apoptosis in AML cell lines — further supporting its potential oncogenic effect in AML [41]. Another example of a lncRNA implicated in HSC homeostasis is LncHSC-2, a nuclear lncRNA, which is expressed in HSCs and hematopoietic progenitors [42]. LncHSC-2 regulates long-term self-renewal and lymphoid differentiation of HSCs by binding to Tcf3, a transcription factor that is essential for HSC proliferation and differentiation into myeloid-lymphoid progenitor cells [42].

In addition to these and other mechanistically studied examples of lncRNAs involved in HSC maintenance and maturation, several comprehensive transcriptomic studies identified hundreds of lncRNAs enriched in HSCs that are co-expressed with lineage-specific transcription factors, indicating that lncRNAs represent another important layer of the complex regulatory network that tunes hematopoietic differentiation [35, 36, 42]. Accordingly, lncRNAs are specifically enriched and functionally relevant not only in the context of HSCs but also in hematopoietic progenitor cell populations and mature blood cell populations. For instance, lncRNA HOTAIRM1 is highly expressed during granulocytic differentiation and contributes to the modulation of target genes in cis and in trans that are essential for proper myelopoiesis [43].

LINC00173 is another example of a lncRNA that is essentially involved in myeloid differentiation. We identified LINC00173 to be specifically expressed in mature granulocytes [3]. Upon knockdown in hematopoietic stem and progenitor cells (HSPCs), granulocytic differentiation and phagocytic capacity are impaired, whereas the erythroid lineage remained unaffected. Further analyses revealed a direct interaction between LINC00173 and PRC2, as well as differential H3K27 trimethylation at the promoter regions of genes involved in stemness, megakaryopoiesis, and erythropoiesis [3].

During erythroid differentiation, lncRNA EPS is enriched only in erythroid progenitor cells and promotes terminal erythrocytic differentiation by repressing pro-apoptotic pathways [44]. Within the lymphoid lineage, numerous lncRNAs that regulate differentiation and contribute to the immune response have been described. An example is lncRNA NeST, which is specifically expressed in CD4+ T-helper 1 cells and regulates the transcription of inflammatory genes through recruiting histone methyltransferase complexes [45, 46].

Dysregulation of the hematopoietic system results in uncontrolled proliferation of immature progenitor cells and in a block of proper differentiation, ultimately leading to the development of leukemia. Several lncRNAs have been shown to contribute to leukemogenesis [34]. In addition, lncRNAs may also serve as biomarkers or predictive factors in this disease [34]. It has been demonstrated that specific lncRNA expression profiles can be utilized to distinguish between different known molecular and cytogenetic AML subgroups and may serve as independent predictors of clinical outcome [47]. Using several examples, we illustrate different functions that have been described for individual lncRNAs in AML. We also briefly discuss lncRNA loci that may have functional consequences in AML cells independent from the encoded transcripts.

The lncRNA HOTAIR is highly expressed in AML and serves as a predictor for poor clinical outcome [48]. In vitro studies in primary AML blasts suggest an oncogenic function of HOTAIR, where it supposedly acts as a decoy for the tumor-suppressive microRNA miR-193a [48]. An alternative mode of action has also been described, where HOTAIR exerts its oncogenic effect in AML through EZH2-mediated epigenetic silencing of the tumor suppressor gene p15 [49].

LncRNA ANRIL, which is upregulated in both AML and ALL, acts as an oncogenic lncRNA by epigenetic silencing of its antisense tumor suppressor gene p15 [14, 50] As for other oncogenic lncRNAs, recent studies indicate a correlation of ANRIL expression to poor survival in patients with AML [51].

In contrast to the previous examples, lncRNA IRAIN is downregulated in AML cell lines and patients with high-risk AML, indicating that lncRNAs might not only act as oncogenes but as tumor suppressors in AML, too [52]. This lncRNA is transcribed antisense from the insulin-like growth factor type I receptor (IGF1R) locus, which is known to promote proliferation of AML cells through the PI3K/Akt signaling pathway [53, 54]. Mechanistically, IRAIN is involved in the formation of an intrachromosomal chromatin loop connecting the IGF1R promoter to a putative enhancer element [52]. However, the functional implications of this mechanism have yet to be elucidated. Clinical data further support the suggested tumor-suppressive function of IRAIN in AML, demonstrating a correlation between low IRAIN expression and poor prognosis in non-M3 acute myeloid leukemia patients [55].

While the majority of lncRNAs have yet to undergo in-depth characterization, this selection of examples provides a glimpse into the diversity of lncRNA functions in the context of hematopoiesis and AML. Even for these better-studied examples, it should be noted that mechanistic details remain elusive, due in part to the extensive experimental labor required to discern between RNA-dependent and -independent effects originating from lncRNA loci [56]. As a case in point, our group recently described MYNRL15 — a pan-myeloid leukemia dependency locus involved in genome topology, whose lncRNA product is dispensable for its dependency phenotype [57]. We found CTCF-enriched lncRNA loci (C-LNCs) like MYNRL15 to be enriched for leukemia vulnerabilities and provide a catalog (www.​C-LNC.​org) in hopes of facilitating the functional classification of lncRNAs and the discovery of new oncogenic vulnerabilities [57].

In contrast to many other malignant diseases, AML occurs in all age groups, but children account for only a small proportion of all patients with AML. The molecular landscape of pediatric AML differs significantly from the molecular profile of adult AML. Chromosomal aberrations are more common in children than in adults with AML [4, 58]. In addition, the genes that are frequently mutated in adult AML (NPM1, DNMT3A, IDH1, IDH2, RUNX1, TP53) are less often affected in children. Other genes, such as FLT3 or GATA2, differ in terms of the exact location and frequency of the mutations between pediatric and adult AML [4]. Given these biological and clinical differences, it is essential that we refrain from simply transferring new findings from adult cell lines, mouse models, and clinical cohorts of adult AML patients to the pediatric setting. Rather, investigations that focus specifically on pediatric AML are needed, to refine current risk stratification criteria and to develop novel therapeutic strategies for children with AML. While most current examples of lncRNAs with roles in AML have first been described in adult contexts, there is now an increasing number of studies characterizing lncRNAs in pediatric AML.

Our group described, for the first time, subtype-specific lncRNA signatures for six major cytogenetic subgroups of pediatric AML, namely, Down syndrome (DS) and non-DS acute megakaryoblastic leukemia (AMKL), inv[16], t[8;21], and AML with KMT2A rearrangement (t[9;11] and t[10;11]) [3]. In the transcriptional landscape of normal and malignant hematopoiesis, most DS- and non-DS-AMKL samples, and KMT2A-r samples, cluster in close proximity to HSCs. Their lncRNA expression profiles are characterized by the absence of myeloid expression programs. In contrast, all other pediatric AML samples clustered in proximity to normal myeloid progenitor cells. We further uncovered a core lncRNA stem cell signature that is shared between HSCs and AML blasts of all different pediatric AML subgroups. High expression of this core lncRNA stemness program is significantly correlated with poor survival in a cohort of adult AML patients [3].

Other studies have focused on the in-depth characterization of individual lncRNAs in pediatric AML. Here, we will summarize lncRNAs that have been implicated in pediatric AML (Table 1) and will exemplarily discuss several individual lncRNAs, for which molecular functions have been studied in the context of pediatric AML. Luo et al. found that the lncRNA HOTTIP is overexpressed in NPM1-mutated and KMT2A-r AML cases and predicts poor outcome [59]. Mechanistically, they showed that HOTTIP alters the three-dimensional structure of the nearby HOXA locus and binds to posterior HOXA sites as well as other genes critically involved in hematopoiesis and leukemogenesis, resulting in the activation of an AML-specific transcriptional program. Of note, HOTTIP expression is sufficient to initiate leukemic transformation of HSCs in mice. Knockout of HOTTIP perturbs leukemic proliferation and prolongs the survival in AML mouse models, suggesting a novel therapeutic option for the treatment of pediatric AML [59].

A recent elaborate study has identified lncRNA CDK6-AS1 as a novel regulator in pediatric AML [60]. In a pediatric patient cohort, CDK6-AS1 was significantly overexpressed and associated with higher minimal residual disease after induction therapy. High CDK6-AS1 levels contributed to an immature phenotype in healthy HSCs and primary AML blasts, whereas silencing of the lncRNA led to increased hematopoietic differentiation of HSCs and to a rescue of the pathogenic undifferentiated state of AML blasts. Mechanistically, the authors could show that CDK6-AS1 regulates expression of its neighboring gene CDK6 by sharing a bidirectional promotor, and that the common CDK6-AS1/CDK6 axis downregulates RUNX1 signaling, which is essential for early hematopoietic differentiation. In addition, CDK6-AS1 activates mitochondrial biogenesis in healthy HSCs as well as in pediatric AML blasts. Interestingly, mitochondrial targeting, using Tigecycline, sensitizes AML blasts with high CDK6-AS1 expression to chemotherapy, supporting the concept of a mitochondrial vulnerability in these blasts. Overall, these findings identified CDK6-AS1 as an important regulator of early hematopoietic differentiation and leukemogenesis of pediatric AML and uncovered therapeutics targeting mitochondrial biogenesis as a novel treatment strategy in pediatric AML [60].

UCA1 is an additional example of an oncogenic lncRNA in adult and pediatric AML. UCA1 has been shown to be upregulated by CEBPα-p30, the CEBPα isoform that results from CEBPA mutations recurrently found in AML patients [86]. In pediatric and adult AML cell lines, UCA1 is upregulated, and knockdown of the lncRNA impairs leukemic viability, migration, and invasion through binding of various microRNAs, such as miR-126, miR-204, miR96-5p, and miR296-3p [61‐64]. Furthermore, UCA1 contributes to chemoresistance in pediatric AML by tethering miR-125a [65].

Other examples of relevant lncRNAs in childhood AML are MONC and MIR100HG, which are host genes for the homologous miRNA clusters miR-99a~125b-2 and miR-100~125b-1, respectively. These miRNA clusters are known to promote the progression of AMKL [87, 88]. Both lncRNA host genes are highly expressed in AMKL cells compared to cell lines of other pediatric AML subtypes. Lentiviral overexpression of MONC alters hematopoietic differentiation, independently of the expression of the oncogenic miRNA clusters [66].

LncRNA MEG3 is downregulated in adult and pediatric AML, supposedly by epigenetic modifications of its genomic locus. Mechanistically, MEG3 has been shown to activate p53 expression and DNMT3A, thereby inhibiting leukemogenesis [67]. Hypermethylation of the MEG3 promoter is associated with poor prognosis in adult AML patients [68]. In pediatric AML patients, higher expression of MEG3 correlates with better survival [69].

These are only a few examples of lncRNAs that have been characterized in the context of pediatric AML and that show correlation to clinically relevant subgroups and/or prognosis of the patients. In addition, a more general lncRNA scoring system based on the expression of 14 lncRNAs has been proposed to predict overall survival in children with AML [89].

In a recently published study from our group, the antisense lncRNA HOXA10-AS was identified as an essential regulator of hematopoiesis and as a novel oncogenic lncRNA in the context of pediatric AML with KMT2A rearrangements (KMT2A-r AML) [70]. Along the genome, HOXA10-AS is located at the posterior end of the HOXA cluster — one of the four highly conserved HOX gene clusters. Tightly controlled spatiotemporal expression of the different HOX genes is crucial for hematopoietic differentiation, and dysregulation of HOX genes such as HOXA9 and HOXA7 by KMT2A fusion proteins is responsible for leukemic transformation in KMT2A-r AML [90]. The HOX gene clusters harbor numerous lncRNAs that are expressed in the same specific pattern as their protein-coding neighbors during differentiation, indicating their putative biological importance [9]. Previous studies revealed that HOX lncRNAs are capable of regulating the expression of neighboring or distant protein-coding HOX genes, as well as of independent effects on other signaling pathways [9, 12, 91]. Although a handful of HOX lncRNAs have undergone further characterization, the role of the vast majority of HOX lncRNAs in AML remains unknown.

HOXA10-AS is transcribed from the antisense strand relative to the protein-coding gene HOXA10 and microRNA mir-196b, both of which are involved in hematopoiesis and in the pathogenesis of KMT2A-r AML. In our study, we confirmed that HOXA10-AS is overexpressed in KMT2A-r AML as well (Fig. 1A). KMT2A rearrangements are the most frequent cytogenetic aberrations in pediatric AML and predominantly affect infants [58, 92]. Gain-of-function experiments in cell lines and primary blasts showed increased leukemic growth of KMT2A-r AML cells upon HOXA10-AS overexpression. In complementary loss-of-function assays using shRNA-mediated knockdown, CRISPR-Cas9-induced excision, and LNA-GapmeRs, we further demonstrated that the maintenance of KMT2A-r AML cells depends on high HOXA10-AS expression (Fig. 1B). During normal hematopoiesis, HOXA10-AS is specifically expressed in HSCs and strongly downregulated during hematopoietic differentiation, whereas the neighboring genes HOXA10 and mir-196b remain highly expressed in early myeloid progenitor cells (Fig. 1C). This strict stem cell-specific expression of HOXA10-AS suggests an independent regulatory circuit and cellular function separate from that of the nearby genes. Hematopoietic differentiation assays upon lentiviral overexpression of HOXA10-AS in HSPCs revealed impaired monocytic differentiation in HOXA10-AS overexpressing cells (Fig. 1D). The observations that ectopic expression of HOXA10-AS impairs monocytic differentiation and that HOXA10-AS is overexpressed in KMT2A-r AML are consistent with the fact that KMT2A-r AML predominantly manifests as a monoblastic leukemia (AML FAB M5) [58]. Regarding the mechanistic characterization of HOXA10-AS, we found that its effects in hematopoiesis and leukemogenesis were independent of changes in expression of the neighboring oncogenes, arguing against a regulatory role for HOXA10-AS on the HOXA cluster in cis. This was supported by subcellular localization studies, which showed that the HOXA10-AS is mainly located in the cytoplasm. Indeed, microarray-based gene expression analysis uncovered a possible trans mechanism for HOXA10-AS involving the upregulation of NF-κB target genes in HOXA10-AS expressing early monocytic progenitors and KMT2A-r AML (Fig. 1 A and D). Finally, we provided a proof of principle of how HOXA10-AS could be leveraged towards clinical implementation, by demonstrating HOXA10-AS as a prognostic marker in AML and potential therapeutic target in pediatric KMT2A-r AML [70].