Deregulation of the non-coding genome in leukemia.

Methodological advances that allow deeper characterization of non-coding elements in the genome have started to reveal the full spectrum of deregulation in cancer. We generated an inducible cell model to track transcriptional changes after induction of a well-known leukemia-inducing fusion gene, ETV6-RUNX1. Our data revealed widespread transcriptional alterations outside coding elements in the genome. This adds to the growing list of various alterations in the non-coding genome in cancer and pinpoints their role in diseased cellular state.

Approximately 80 % of the genome is transcribed into RNA species in at least some cell type or at some stage of development. Non-coding regulatory (non-housekeeping) RNAs are currently defined by their size, genomic location or presumptive function. Enhancer RNAs (eRNA), which have a length span from 0.1 to 10 kb, mainly fall into the category of long non-coding RNAs (lncRNAs) although they are better defined by their transcriptional regulatory function. Larger clusters of enhancers with multiple transcription factor (TF) binding sites and open chromatin marks are termed super-enhancers and they define cell identity. Locations of enhancer elements are often deduced from certain histone marks (H3K4me1, H3K27ac), transcription factor binding profiles (p300), or open chromatin states (eg. DNAse- and ATAC-seq). The development of global nascent RNA sequencing techniques, such as global run-on sequencing (GRO-seq), has revealed that transcription of eRNAs is highly correlated with marks such as H3K27ac (for review see ref. 6) and to transcription at nearby gene promoters, and is considered the most reliable mark of an active enhancer. The functions of eRNAs are yet unclear: they can be passive byproducts of transcription or function actively in recruitment of transcription factors (reviewed in ref. 10), like in the case of Yin-Yang (YY)1.

Misregulation of ncRNAs is common in cancer although recurrent structural variations have been challenging to find. For example, in a study with whole-genome sequencing of 150 tumor/normal pairs of chronic lymphocytic leukemia, only one recurrent non-coding mutation cluster was found at a potential regulatory element. However, this may also reflect the lacking annotations. We recently analyzed whole genome sequencing data from precursor B-cell acute lymphoblastic leukemia (pre-B-ALL) in the context of chromatin architecture and found that the topologically associated domains with the highest number of breakpoints contained unannotated ncRNAs. Functional studies manipulating lncRNA production in leukemia have shown diverse roles in cancer-related pathways. In addition, functional studies on enhancers have highlighted their overall role in cancer, as reviewed in ref. 17. In leukemia, somatic mutation of a non-coding element generated a MYB binding site upstream of oncogenic TAL1 locus, and a deletion of the mutated (but not wild type allele) super-enhancer in a T-ALL cell line decreased expression of TAL1 and impaired cell survival. Altered transcription at enhancers may also result from structural or quantitative changes in both enhancer elements and their regulating proteins. Duplication of NOTCH1-driven MYC enhancer was observed in T-ALL and its relevance demonstrated in a mouse knockout model. Moreover, aberrations in chromatin structure and especially in insulator regions induce abnormal gene expression, as exemplified by activation of TAL1 due to a deletion of upstream insulator element. Misregulated transcription during delicate differentiation processes in haematopoietic precursors may also cause cancer by predisposing to secondary mutations. Convergent transcription and RNA polymerase II stalling strongly correlate with structural variation clusters and seem to provide vulnerable regions for RAG and AID mediated double strand breaks in lymphoma and leukemia. Although ncRNA expression profiles using microarray or RNA-seq have been published (eg. refs. 22-26), many nascent transcripts have remained unnoticed because of rapid degradation of several ncRNA species. New methods to address this challenge have emerged, such as GRO-seq, PRO-seq or TT-seq that enable monitoring various nascent transcripts and engaged RNA polymerase II in leukemia.

We addressed this issue in the ETV6-RUNX1 (E/R, TEL-AML1) fusion positive leukemia, which represents 25 % of pediatric acute lymphoblastic leukemias, and causes alterations in gene expression that predispose to leukemia. With the help of an inducible E/R cell model and GRO-seq, we explored dynamics of gene expression and the activity of their regulatory elements simultaneously, exposing the transcriptional circuitry downstream of the E/R fusion (Fig. 1). We analyzed enhancers based on eRNA correlation with GRO-seq signal change at differentially expressed genes (transcript-centric approach). Secondly, we generated an enhancer-centric approach that directly applied the statistical framework on eRNA levels to identify significantly regulated enhancers (enhancer annotation was based on H3K27ac and RUNX1 ChIP-seq data) and correlated these changes to that of nearby transcripts. We found at least one similarly altered putative enhancer element within +/−400 kb for almost all the deregulated coding transcripts using transcript-centric approach. E/R regulated approximately 20% of transcribed regions with RUNX1 ChIP peaks, and 5% of CD19/20 (B-cell)-related enhancers. Interestingly, CD19/20 specific super-enhancers were mostly downregulated, implying a way for E/R to arrest cell differentiation.

Figure 1.

(A) A schematic representation of the ETV6-RUNX1 (E/R, TEL-AML1) fusion protein resulting from a recurrent t(12;21) translocation in pediatric pre-B acute lymphoblastic leukemia. ETV6-RUNX1 includes the pointed (PNT) domain of ETS variant 6 (ETV6) but lacks the ETS domain that is involved in DNA binding of the normal TF protein. The 480 aa long RUNX1 variant 1 (AML-1c, NP_001745) is illustrated with the point mutation R201Q in the Runt domain which impedes its DNA binding capability (this was used to generate E/Rmut in ref. 30). ID = Runx inhibitory domain. (B) GRO-seq signal (nascent RNA transcription) is shown for E/R-negative as red and E/R-positive samples as blue tracks at an example genomic region. Signals above and below the axis indicate plus and minus strands, respectively. RUNX1 ChIP peaks in SEM cells (GSE42075, ref. 42) and an enhancer marker H3K4me1 ChIP-seq in B-cells (GM12878, ref. 2) are shown and coincide with the GRO-seq signal. Three enhancer regions that are downregulated by E/R via RUNX1-mediated binding are highlighted. Nalm6-E/R = 24h expression of E/R in a pre-B-ALL cell line; REH = E/R-positive cell line; pre-B-ALL other = E/R-negative patient; pre-B-ALL E/R+ = E/R-positive patient.

It has been proposed that any transcription may possess regulatory activity. A recent study showed that half of the studied transcribed gene loci (12 lncRNA and 6 mRNA) regulated a nearby gene in cis independently of whether the locus was a coding or non-coding one. As the non-coding genome is only weakly conserved, most non-coding regions may function in a way which is not dependent on the sequence of transcript itself but rather the sequence of its promoter or its location in the genome. In the case of E/R leukemia, we classified 57 deregulated novel lncRNAs (over 5 kb long) as either potential eRNAs or lncRNAs based on the GRO-seq signal. One fourth of the novel and 3 of 7 annotated transcripts were concordantly differentially expressed in RNA-seq data with 8 E/R-positive and 9 other subtype pre-B-ALL patients. For example, KCNQ1OT1, which acts in epigenetic regulation, was upregulated in our E/R cell model GRO-seq and the patient RNA-seq data. Signal changes at ZEB1 and ZEB1-AS1 serve as an example of a simultaneous downregulation of gene and its promoter-associated RNA, with ZEB also being linked to cancer and late B cell differentiation. Functional roles of the novel transcripts in E/R leukemia remains to be explored in future. Nascent RNA profiles of diagnostic patient samples of distinct ALL subtypes will give further insights into the derailed transcriptional network downstream of the oncogenic TF fusions.

Already, thousands of regulatory lncRNA transcripts and hundreds of thousands of enhancer regions have been found. It is now known that ncRNAs are widely specific to a certain cell type and developmental stage. For example, most lncRNAs that are expressed at various stages of mouse B cell development are not expressed in a closely related T-cell lineage. A recent study noted that distal regulatory elements varied across distinct haematopoietic lineages so that they are better discriminators of cell identity than mRNA levels. This was also reflected in our work, where we noticed that sample separation based on quantification of global eRNA transcription was equally good as that based on quantification of transcription at protein coding regions. We can assume that the increasing knowledge of the interplay between various elements of genome and their transcriptional products will significantly contribute to our understanding of the diverse types of leukemia and cancer in near future.