Molecular Medicine Israel

Circular RNAs drive oncogenic chromosomal translocations within the MLL recombinome in leukemia

Summary

The first step of oncogenesis is the acquisition of a repertoire of genetic mutations to initiate and sustain the malignancy. An important example of this initiation phase in acute leukemias is the formation of a potent oncogene by chromosomal translocations between the mixed lineage leukemia (MLL) gene and one of 100 translocation partners, known as the MLL recombinome. Here, we show that circular RNAs (circRNAs)—a family of covalently closed, alternatively spliced RNA molecules—are enriched within the MLL recombinome and can bind DNA, forming circRNA:DNA hybrids (circR loops) at their cognate loci. These circR loops promote transcriptional pausing, proteasome inhibition, chromatin re-organization, and DNA breakage. Importantly, overexpressing circRNAs in mouse leukemia xenograft models results in co-localization of genomic loci, de novo generation of clinically relevant chromosomal translocations mimicking the MLL recombinome, and hastening of disease onset. Our findings provide fundamental insight into the acquisition of chromosomal translocations by endogenous RNA carcinogens in leukemia.

Introduction

RNA-DNA hybrids are commonplace throughout the mammalian genome, including those formed during DNA replication (11 bp hybrid from Okazaki fragments), and transcription (8 bp hybrid within the RNA polymerase active site).1 However, longer tracts of such three-stranded nucleic acid structures, known as R loops, form when RNA base pairs with its cognate genomic DNA locus, displacing a loop of single-stranded DNA (ssDNA). The exposed ssDNA tracts are sites of genome instability and are susceptible to mutagenesis, which can manifest as double-strand DNA breaks (DSBs) from base excision repair.2 This R loop-mediated genome instability can manifest advantageously, including their indispensable role in immunoglobulin class switch recombination (CSR) in activated B cells.3 CSR requires the mutagenic enzyme, activation-induced cytidine deaminase (AID),4,5 which belongs to the apolipoprotein B mRNA editing enzyme, catalytic polypeptide-like (APOBEC) family of proteins. These APOBEC proteins deaminate deoxycytidine in ssDNA to deoxyuridine, a process that can cause DSBs. Since APOBEC proteins are expressed beyond B cells alone,6,7 there is potential for widespread, R loop-mediated DNA mutagenesis. Conceivably, aberrant R loop formation could stimulate oncogenic driver mutations, critical in the initiation of cancer.8

Acute leukemias carry the lowest mutational burden of any cancer type, on average <0.5 mutations per megabase.9 Conserved de novo chromosomal translocations between the H3K4 histone methyltransferase mixed lineage leukemia (MLL, also called KMT2A) gene and one of >100 known partner genes are found in about 5%–10% of all acute leukemia patients, but >70% of infant leukemias.10,11 The spectrum of translocation partner genes, called the MLL recombinome, varies between infant, pediatric, and adult cohorts and between acute lymphoblastic leukemia (ALL) and acute myeloid leukemia (AML). However, common translocation partner genes include MLLT1-ENLMLLT2-AFF1-AF4, and MLLT3-AF9, which account for approximately 50% of AML and 88% of ALL translocations.11,12 Rearrangements of the MLL gene also occur frequently in therapy-related acute leukemia, accounting for 15% of AMLs and are associated with particular anti-cancer treatments such as topoisomerase inhibitors, which promote DNA breaks.10 In the background of a paucity of additional gene mutations, this family of translocations can efficiently transform hematopoietic cells into leukemic stem cells13 and initiate aggressive forms of ALL and AML that are associated with particularly poor outcomes with 5-year survival rates of 5%–27%.14,15

Most MLL translocations map to an 8.3 kb breakpoint cluster region (bcr) comprising introns 5–12 of MLL and common sites within their partner gene introns.12,16 Some chromosome features have been suggested to play a role in delineating the MLL bcr, including transcriptionally active Alu elements in intron 9, and a DNase I hypersensitive site and internal RNA polymerase II (RNAPII)-driven promoter in intron 11.11,17,18 Despite these observations, the mechanisms underlying vulnerability of the bcr to translocation events and how it fuses with conserved sites in partner genes remain to be fully elucidated. Understanding the fundamental basis of these potent MLL rearrangements by focusing on their genomic landscape is critical for the prevention of leukemia.11

It is clear that R loops can form co-transcriptionally (cis R loops), but may also target loci of similar sequence composition in trans.1 Immunoprecipitation with the R loop-specific S9.6 monoclonal antibody coupled to high-throughput sequencing of the resident DNA (DNA:RNA immunoprecipitation sequencing, or DRIP-seq) or the bound RNA (DRIPc-seq) has shown that cis R loops comprise the vast majority of R loops genome wide.19 Defects in termination and elongation factors are thought to stall the RNA polymerase, increasing the time in which nascent RNAs are proximal to cognate DNA sequence and prolonging the unwound state of the DNA and promoting recruitment of AID through SPT5.20

The Fanconi anemia (FA) pathway is able to resolve R loops and prevent genome instability,21 with the mutation of the FA genes common in cancer, particularly AML.22 In addition, key roles in DNA repair are played by the ubiquitin-proteasome system including through the activation of the FA pathway23 and through recruitment of both 19S and 20S proteasome subunits to sites of DSBs.24 Defects in topoisomerases, or chemotherapeutic intervention with topoisomerase inhibitors (e.g., etoposide) can prevent relaxation of the negative supercoils ultimately leading to leukemogenic chromosomal translocations for, as yet, unidentified reasons.10,25 Furthermore, deficiencies in splicing, or non-canonical splicing, may unmask the RNA, making it more accessible to hybridize with DNA.26 Therefore, it is apparent that R loops can be augmented by non-canonical splicing and demarcate sites of genome instability. Through impairment of DNA repair, including but not limited to FA pathway inactivation and/or proteasome inhibition, a hallmark of cancer, it is conceivable that R loops could yield oncogenes by driving locus-specific mutagenesis.

It has recently become evident that another class of RNA, circular RNAs (circRNAs), are abundant and ubiquitous across eukaryotes but have been overlooked because they were difficult to detect by traditional methods.27,28,29 CircRNAs are covalently closed circles of single-stranded RNA that arise from non-canonical, back-splicing of pre-mRNA. CircRNAs can be formed from RNA-binding proteins such as Quaking, which can be misregulated in cancer.30,31 Consequently, circRNAs have been identified as cancer biomarkers and function by promoting hallmarks of cancer.32,33,34,35 One report identified circRNAs arising following the oncogenic fusion of MLL to known partner genes, and these fusion circRNAs (f-circRNAs) contributed to cellular transformation in leukemia.36 Critically, there has been no evidence for the capacity of circRNAs to drive oncogenic mutations.

By investigating the functional consequence of circRNA:DNA hybrids (circR loops) in human cancer circRNAs are shown to demarcate the sites of chromosomal translocations and drive oncogenic gene fusions via endogenous RNA-directed DNA damage (ER3D).

Section snippets

circR loops are prevalent across the human genome

To assess the contribution of circRNAs to R loops (Figure 1A) in human cells genome wide, we performed DRIP-seq on HEK293T cells. In accordance with previous studies in mammalian cells,38 we identified 20,467 discrete R loops (Figure 1B, red bars; Table S1). We assessed which of these peaks overlap with circRNAs at their cognate locus in HEK293T cells by cross-referencing R loops with matched, nuclear-fractionated circRNA sequencing (circRNA-seq) (Figure 1B, blue bars). To facilitate

Discussion

CircRNAs are increasingly shown to play functional roles in eukaryotic cells, including known hallmarks of cancer, mediated via their interactions with microRNAs, proteins, and more recently mRNA74 and DNA.40 Here, we show that circRNAs, which are formed co-transcriptionally,75 can form circRNA:DNA hybrids (circR loops) and cause DNA breaks resulting in chromosomal translocations. R loops formed during transcription have been shown to be dynamically resolved with an average half-life of…

Sign up for our Newsletter