Chimeric RNAs and Their Implications in Cancer

Abstract

Chimeric RNAs have been believed to be solely produced by gene fusions resulting from chromosomal rearrangement, thus unique features of cancer. Detected chimeric RNAs have also been viewed as surrogates for the presence of gene fusions. However, more and more research has demonstrated that chimeric RNAs in general are not a hallmark of cancer, but rather widely present in non-cancerous cells and tissues. At the same time, they may be produced by other mechanisms other than chromosomal rearrangement. The field of non-canonical chimeric RNAs is still in its infancy, with many challenges ahead, including the lack of a unified terminology. However, we believe that these non-canonical chimeric RNAs will have significant impacts in cancer detection and treatment.

Introduction

The term “chimeric RNAs” is not foreign to the field of cancer research. Starting from the discovery of BCR-ABL fusion resulting from “Philadelphia Chromosome”[1], chimeric RNAs are known to be products of gene fusions and considered ideal biomarkers for cancer. With modern technologies, they are being uncovered at an unprecedented rate. Several large databases including Mitelman [2], ChimerDB [3], ChiTaRs [4], FusionDB [5], dbCRID [6], TICdb [7], ConjoinG [8], FusionCancer [9] and HYBRIDdb [10], have collected thousands or more chimeric RNAs (Table 1). However, as next generation sequencing becoming popular and additional chimeric RNAs being discovered, increasing evidence is mounting toward a realization that goes against two traditional dogmas: chimeric RNAs are strictly the result of gene fusions, and that chimeric RNAs are unique to cancer.

Table 1.

Different databases hosting chimeric RNAs/gene fusions.

Database	Tumor	Non-tumor	Total entry	Web link
Mitleman	Yes	No	10861	https://ulib.iupui.edu/databases/mitelman-database-chromosome-aberrations-and-gene-fusions-cancer
ChimerDB3.0	Yes	No	33316	http://203.255.191.229:8080/chimerdbv31/mindex.cdb
ChiTaRs-3.1	Yes	No	42000	http://chitars.md.biu.ac.il/
dbCRID	Yes	Yes	2643	http://dbcrid.biolead.org
TICdb	Yes	No	28171	http://www.unav.es/genetica/TICdb/
ConjoinGene	Yes	Yes	800	https://metasystems.riken.jp/conjoing/index
FusionCancer	Yes	No	11839	http://donglab.ecnu.edu.cn/databases/FusionCancer/index.html
HYBRIDdb	Yes	Yes	46492	http://www.primate.or.kr/hybriddb/
FusionDB				Web link no longer active

We now know that chimeric RNAs can be made at the RNA level by intergenic splicing, and that they are not specific to cancer. Despite these new and exciting discoveries, the field of non-traditional chimeric RNA is still in its infancy. In addition, the discoveries of these new chimeric RNAs bring confusion to some current nomenclatures, and raise alarms for some traditional practices. In this communication, we try to clarify some of the terminology. We also reviewed the current knowledge about the mechanisms of chimeric RNA formation, and their categorizations, as well as discussed their implications in cancer diagnosis and treatment (Fig. 1).

Figure 1. Chimeric RNA centric view.

Chimeric RNAs can be generated by gene fusion, trans-splicing, and cis-splicing between adjacent genes (cis-SAGe). They can function as chimeric proteins, long non-coding chimeric RNAs (lnccRNA), or affect parental gene expression. Traditionally, they have been thought to be specific to cancer. However, increasing numbers are being found in normal physiology.

Terminology

Firstly, there is a need to clarify some terminology. Although tens of thousands of chimeric RNAs are deposited into various databases, the way in which they are defined is unclear, partially due to the ill-defined “gene” and unclear mechanism of generation for the chimeric RNAs [11**]. Various names have been used, including: transcription mediated fusions [12], gene fusions [13], conjoined genes [8,14*], complex genes [15], co-transcribed genes [16], spanning genes [17], hybrid genes [10], and tandem chimerism [18]. We prefer the definition of “gene” as a nucleotide sequence in a DNA molecule that acts as a functional unit for the production of a protein, a structural RNA, or a catalytic or regulatory RNA molecule [19]. We then define “chimeric RNA” as a fusion transcript composed of exons, or fragments of exons from different genes.

We propose to differentiate “chimeric RNA” from the traditional term, “gene fusion”. Even though in the past, labs including our own had been using the two terms interchangeably, the term “gene fusion” tends to leave the impression of a fusion event happening at the DNA (gene) level. “Chimeric RNA” on the other hand can be a general term to describe fusion transcripts regardless of their mechanism of formation. In this regard, we consider the terms “chimeric RNA”, “chimeric transcript”, “fusion RNA”, and “fusion transcript” to be the same. Additionally, when it comes to referencing chimeric RNAs that are composed of exons from two adjacent genes transcribing in the same direction, many terms have also been used, including: transcription induced chimeras [12], tandem RNA chimeras [18,20], conjoined genes [8], and read-through fusions [21**,22]. However, these terms do not differentiate cis-splicing (precursor RNA transcribing through gene boundary and splicing together exons of neighboring genes) from trans-splicing events. In fact, trans-splicing may have a higher incidence rate with genes in close proximity to each other. To distinguish between the two, we prefer to use the term “cis-splicing of adjacent genes (cis-SAGe)” [23] or “gene read-through” for the first mechanism. Since “gene read-through” was originally used to describe protein translation processes that skip a stop codon, we propose to use the term cis-SAGe to avoid confusion. Recently Yuan et al., proposed to restrict the term “chimeric RNA” to those that are formed by separate RNA transcripts [11**]. By their definition, fusion RNAs from gene fusion or cis-SAGe should not be called “chimeric RNAs”. Even though this restriction holds some merit, it is not practical, and will cause more confusion, as it requires additional mechanistic insights, that are currently missing for the vast majority of fusion RNAs. Our definition is simple, as it relies only on the genome annotation, and requires no additional knowledge regarding how the fusion RNAs are formed.

Generating Mechanism

Given the above definition, chimeric RNAs are known to be produced hitherto by three mechanisms (Fig. 2) [24]. The classic and well-studied mechanism is through gene fusions. Chromosomal rearrangements that include insertions, deletions, and translocations can result in cytogenetically distinct gene loci juxtaposed together. In the past, due to technical limitations, most gene fusions were discovered in haematological and childhood malignant tumors, including BCR-ABL resulting from t(9;22) in chronic myelogenous leukemia (CML) [1,25] and in acute lymphoblastic leukemia (ALL) [26] or acute myelogenous leukemia (AML) [27] ;TEL-AML1 (ALL with t(12 ; 21) [28], AML1-ETO (M2 AML with t(8 ; 21)) [29]; PAX3-FOXO1 (t(2;13) in alveolar rhabdomyosarcoma [30]. With modern high throughput technologies, recurrent gene fusions are uncovered in common epithelial cancers, such as TMPRSS2-ERG in prostate cancers [31], and BCAM-AKT2 in serous ovarian cancer [32**].

Figure 2. Three known generating mechanisms for chimeric RNAs.

Exons are depicted as blocks, and introns are indicated by lines. (A) Chromosomal rearrangement including translocation, deletion, and inversion. Shown here is a case of translocation. Gene fragments from different genomic loci are juxtaposed together. (B) RNA trans-splicing. Two separate pre-mRNA transcripts are spliced together. (C) cis-splicing between adjacent genes. The transcription machinery reads through two neighboring genes, and the exons from the two genes are spliced together.

The other two mechanisms sometimes are grouped together as “intergenic splicing”. One is trans-splicing, in which exons from two separate RNA transcripts are spliced together. Trans-splicing is well documented in lower eukaryotes [33,34], while it was first observed in vitro using mammalian cell extracts [35–38], and then in vivo in higher eukaryotes, including humans [39–46**,47*]. The molecular mechanism of trans-splicing in vertebrates is still elusive. It likely involves multiple factors, including transcriptional and splicing machinery, in conjunction with some sequence specificity, and three-dimensional proximity (detail discussion see [48**]).

Another intergenic splicing mechanism is cis-SAGe, which involves same strand neighbor genes. Traditionally, such chimeric RNAs containing exons of neighboring genes have been considered rare in mammalian cells, but recent studies incorporating systematic in silico analysis and paired-end RNA-Seq have identified many potential cis-SAGe chimeric RNAs [49,50]. In a recent study involving the analysis of both prostate cancer and non-cancerous samples, over 300 chimeric RNA events were observed, of which 30%, were characterized as cis-SAGe chimeras [21]. Another study classified 76% of their candidates as deriving from adjacent genes [51*]. Intuitively, cis-SAGe requires an active transcription from the 5′ parental gene, a readthrough transcript across gene boundaries, and alternative splicing between exons of the two genes. Factors involved in any of the three steps are likely to contribute to the regulation of cis-SAGe. Take SLC45A3-ELK4 e1e2 form for example. It is a recurrent cis-SAGe fusion in prostate cancer. Insulator binding factor CTCF negatively correlates with the expression of the fusion RNA, and silencing CTCF resulted induction of the fusion [52*]. In addition, CTCF silencing combined with RNA-Seq indeed revealed additional cis-SAGe fusions [21]. However, it is not the only factor. Even though CTCF binding to the insulators can regulate the fusion expression in different culture conditions for the same cell line [52*], among different cell lines the expression of SLC45A3-ELK4 correlates with the expression of the parental gene SLC45A3. In addition, forced RNA polII pausing by flavopiridol can induce the fusion RNA expression (unpublished). Recently, Vilborg et al. reported upregulation of ‘downstream of gene’-containing transcripts (DoG) under osmotic stress [53**]. It is thus likely that these same stresses may also enhance the level of cis-SAGe fusions.

Yuan et al. discussed three options for categorizing cis-SAGe fusions: a variant of upstream gene, dubbed as gene A; or a variant of downstream gene, dubbed gene B; or a totally new gene, gene C [11**]. Among the three, they favor the last option. We disagree with such classification, and propose to still dub them as A-B. Many of the parental genes involved in cis-SAGe are genuine genes that have well annotated TSS, and polyadenylation and termination sites. Their expression patterns are not identical, and they may have distinctive functions. In the case of SLC45A3-ELK4, SLC45A3 encodes a solute carrier, which is expressed on cell membrane. It is almost exclusively expressed in prostate tissue. ELK4 encodes an ETS family transcription factor, which is a nuclear protein. It is expressed in many tissues and cell types. Grouping the fusion into either of the parental gene variants, or calling it a new gene is not accurate, and will cause unnecessary confusion.

Categories

Chimeric RNAs could be classified into different categories by various criteria (Fig. 3). They can be grouped according to their generating mechanism as outlined above: gene fusion products, trans-splicing and cis-SAGe [48]. However, chimeric RNAs can also be sorted by other criteria. For instance, chimeras could be classified according to the chromosomal locations of their parental genes: parental genes located on different chromosomes (INTERCHR), neighboring genes transcribing the same strand (INTRACHR-SS-0GAP), and parental genes on the different strands of the same chromosome or with gaps in between (INTRACHR-OTHER) [21**]. Chimeras can be divided by the junction position relative to the exons of the parental genes, which include: both sides being known exon ends (E/E), one side being exon end, the other not (E/M or M/E); or both sides falling into the middle of exons (M/M) [46**]. Chimeras can also be grouped by their protein coding potentials: those in which the chimera’s coding sequence is in-frame with both parental genes (in-frame), those in which the reading frame of the downstream parental gene is different from the upstream parental genes (frame-shift), and those whose junction site falls outside of the open reading frame of either parental genes, or one or both parental genes are non-coding [46]. One note of caution is that the protein coding potential is only a prediction, which may not mirror how individual chimeric RNAs truly function. One example is the SLC45A3-ELK4 e1e2 fusion. Even though it encodes the ELK4 protein, it functions as a long non-coding chimeric RNA (lnccRNA) [54**]. All the above mentioned and even more categorizations may be used according to the topics researchers are investigating.

Figure 3. Example of the landscape of chimeric RNAs.

The data is from our analysis of over 7000 RNA-Seq samples covering 53 different tissues. (A) Chimeric RNAs can be grouped by different criteria. They can be grouped according to the location of parental genes, the fusion junction site relative to exons, and protein coding reading frames. (B) Chimeric RNAs may have different tissue distributions.

Chimeric RNAs in Cancer Diagnosis and Treatment

Gene fusions and their fusion products (RNA and protein) have had major impacts in cancer diagnosis and treatment. The oncogenic fusion BCR-ABL1 resulted from the t(9,22) translocation loci, named the ‘ Philadelphia chromosome’, was the target of an effective cancer drug known as, Gleevec (Imantinib) [25,55]. Recent advances in high-throughput sequencing and computational capabilities, which provide single nuclear-base resolution and long single strand coverage, have enabled the systematic discovery of novel chimeric RNAs in solid cancers [21**,49,56**,57–59]. EML4-ALK was detected in non-small cell lung cancer, which was treated with ALK inhibitors to improve prognosis [60,61]. In prostate cancer, fusions with promoters of certain inducible, or highly expressed genes were identified and characterized. The most well-know fusion is TMPRSS2-ERG, harbored by the majority of prostate cancers. In this fusion, the androgen-regulated TMPRSS2 promoter fuses to the coding region of ERG [62]. This fusion drives a unique transcriptional program, inducing DNA damage, invasion and metastasis [63*,64–67]. Wang et al., recently identified a series of peptides that interact specifically with the DNA binding domain of the ERG part of this fusion [63*]. This peptide inhibited ERG-mediated transcription, with reduced cell invasion, proliferation and tumor growth.

These textbook successes have led to great enthusiasm for using chimeric RNAs as biomarkers and drug targets, and at the same time three assumptions: 1, chimeric RNAs are solely produced by gene fusions at chromosomal level; 2, detecting chimeric RNA is equivalent of detecting gene fusions, and; 3, chimeric RNAs are unique to cancer cells.

However, increasing amounts of research has found that chimeric RNAs are also present in non-cancerous tissues and cells [44,45,68,69]. Some of them may also play important roles in normal physiology [46,70*]. In a recent survey of 117 RNA-Seq datasets covering 30 different non-cancerous tissues and cells, 291 recurrent chimeric RNAs were found [46]. Our ongoing research with over 7,000 RNA-Seq datasets of 44 tissues have also yield a large number of chimeric RNAs, of which many were identified in more than ten samples (unpublished). Even though the vast majority have not been experimentally validated, let alone functionally studied, their presence has already posed a challenge to the traditional view that fusion RNAs are cancer-specific features. In the field of cancer researchers, one common practice is still associate fusion RNAs discovered by RNA-Seq in certain cancer cell lines and tissues being equivalent to gene fusions, as well as to immediately assign them the title of cancer biomarkers. This practice has resulted in an explosion in the number of entries deposit into the Mitelman Database of Chromosome Aberrations and Gene Fusions in Cancer (https://cgap.nci.nih.gov/Chromosomes/Mitelman) [2]. As of August 4^th of 2017, 10,841 “ gene fusions” are included. However, many were not validated to be product of gene fusion or being cancer specific. We would likely to raise the alarm against this practice. Chimeric RNAs are not a phenomenon unique to cancer, and they are not all the results of gene fusions. Therefore, extensive validations involving sufficient numbers of non-cancer control samples need to be conducted before assigning a chimeric RNA being a cancer marker [70*]. Similarly, more evidence is needed in order to claim that a chimeric RNA is a product of gene fusion.

On the other hand, chimeric RNAs generated by non-traditional gene fusions add another layer of transcriptional regulation, which can go awry in cancer. Therefore, they represent a new repertoire for cancer biomarkers and/or therapeutic targets. For instance, the CYCLIN D1-TROP2 chimeric RNA was found in multiple types of cancer, with no involvement of chromosomal rearrangement [71]. YPEL5-PPP1CB, was found in 28% of CLL samples with no evidence for a genomic fusion between YPEL5 and PPP1CB [72]. The SLC45A3-ELK4 e1e2 form is an example of cis-SAGe fusion in the absence of corresponding fusions at the DNA level. Its level correlates with prostate cancer development, and yet the parental genes do not share that correlation [23,58]. In addition, Wen et al. identified seven fusions in acute myeloid leukemia with normal karyotyping [73]. These examples, and others, support that the abnormality of intergenically spliced chimeric RNAs may play an underappreciated role in cancer biology. Whole exome or even whole genome sequencing will miss the discovery of this pool of potential biomarkers and/or drug targets that are produced abnormally at the RNA level.

Conclusions and Future Perspectives

We aim to clarify some discrepancies in nomenclature within the field of chimeric RNAs. Gene fusions and their corresponding chimeric RNAs have had major impacts on cancer diagnosis and treatment. We predicted that the non-traditional chimeric RNAs produced by trans-splicing, cis-SAGe, and possibly other unknown mechanisms will contribute to the discovery of new biomarkers and potential drugs. At the same time, we want to sound the alarm for some of the misconceptions and incorrect practices.

In the future, new sequencing technologies, especially full-length sequencing, and development of new software will facilitate the discovery of additional chimeric RNAs. One of the bottlenecks is the lack of a high throughput approach to study the functions of chimeric RNAs. Mechanisms of formation for the non-traditional chimeric RNAs, as well as their relationships with gene fusions are also burning questions that are begging for answers.