Gene fusions and chimeric RNAs, and their implications in cancer

Abstract

Gene fusions are appreciated as ideal cancer biomarkers and therapeutic targets. Chimeric RNAs are traditionally thought to be products of gene fusions, and thus, also cancer-specific. Recent research has demonstrated that chimeric RNAs can be generated by intergenic splicing in the absence of gene fusion, and such chimeric RNAs are also found in normal physiology. These new findings challenge the traditional theory of chimeric RNAs exclusivity to cancer, and complicates use of chimeric RNAs in cancer detection. Here, we provide an overview of gene fusions and chimeric RNAs, and emphasize their differences. We note that gene fusions are able to generate chimeric RNAs in accordance with the central dogma of biology, and that chimeric RNAs may also be able to influence the generation of the gene fusions per the “horse before the cart” hypothesis. We further expand upon the “horse before the cart” hypothesis, summarizing current evidence in support of the theory and exploring its potential impact on the field.

Introduction

Gene fusions are hybrid genes resulting from the fusion of two previously separate genes. Gene fusions are formed by chromosomal rearrangement including translocation, inversion, deletion or tandem duplication. Interestingly, gene fusions are characteristic cytogenetic signatures of many cancer types, and have been successfully used as diagnostic tools. Often, gene fusions give rise to gene fusion transcripts and chimeric protein products, which have been used as targets for treatment. The well known examples are Gleevec (Imantinib) targeting BCR-ABL1 gene fusion and crizotinib targeting EML4-ALK gene fusion.1, 2

While the term “gene fusion” refers to DNA-level fusion events, “chimeric RNA” refers to any transcript composed of exons from different parental genes, including gene fusion transcripts. Contrary to popular belief, transcription of gene fusions is not the only means of chimeric RNA generation. Chimeric RNAs can also arise from trans-splicing of two separate precursor mRNAs and alternative splicing of a readthrough transcript, otherwise known as cis-splicing of adjacent genes (cis-SAGe).³

Interestingly, several studies indicate the presence of chimeric RNAs which exactly mimic common gene fusion transcripts in the absence of the corresponding gene fusion. Such observations have spurred “the cart before the horse” hypothesis, in which chimeric RNA can be generated first by trans-splicing and then guide genome rearrangement to form the corresponding gene fusion.⁴

Within this manuscript, we discuss known generation mechanisms for gene fusions as well as their relevance to cancer. Further, we present existing evidence in support of “the cart before the horse hypothesis” and discuss the implications and significance of these findings on our understanding of oncogenesis.

Mechanisms of chimeric RNAs generation

Transcription of gene fusions is a classic, well-studied mechanism of generating chimeric RNA. Rearrangements such as translocation, inversion, deletion, and tandem duplication have all been shown to produce gene fusions then transcribe to corresponding chimeric RNAs. The first renowned gene fusion, BCR-ABL, was discovered in human chronic myelogenous leukemia (CML). It is the result of translocation between the q arms of chromosomes 9 and 22 t(9; 22),⁵ and its chimeric transcript encodes for a fusion protein which is an altered, constitutively active ABL1 kinase. Another gene fusion, DNAJB1-PRKACA is the result of chromosome segmental deletion. It is now recognized as a biomarker of fibrolamellar hepatocellular carcinoma (FL-HCC), and introducing this deletion via CRISPR/CAS9 to adult mouse liver can successfully generate the Dnajb1–Prkaca gene fusion and effectively induce tumors resembling fibrolamellar hepatocellular carcinoma.6, 7 Additionally, the FGFR3-TACC3 gene fusion in human glioblastoma arises via tandem duplication and insertion.⁸ The fusion transcript encoding a protein, which has constitutive kinase activity and induces mitotic and chromosome separation defects then triggers aneuploidy.⁸ Subsequent studies reported this gene fusion in many other cancers, indicating that FGFR3-TACC3 is a recurring gene fusion in cancer.⁹ FGFR3-TACC3 is a potent oncogene that promotes the phosphorylation of PIN4, inducing mitochondrial respiration and promoting tumor growth.¹⁰

Another mechanism for chimeric RNA generation is cis-SAGe, in which two neighboring genes are transcribed into one precursor RNA by transcriptional readthrough, followed by RNA splicing between exons from the two neighboring genes.³ The cis-SAGe product SLC45A3-ELK4 was discovered by two independent groups11, 12 and has the potential to be a biomarker in prostate cancer.3, 13 Knockdown of SLC45A3-ELK4 in cancer cells causes reduction in cell proliferation.¹⁴ Another example of a cis-SAGe chimeric RNA is DUS4L-BCAP29, which was discovered in gastric and prostate cancer and plays a tumor-promoting role in gastric cancer.15, 16 However, our group found that DUS4L-BCAP29 also exists in a variety of normal tissues, and its growth promoting effect is not only found in cancer but also in normal physiology.¹⁷

A third category of chimeric RNAs arise from trans-splicing, in which exons from different RNA transcripts are spliced together.¹⁸ While there is limited direct evidence of chimeric RNAs arising from two separate precursor mRNAs, many chimeric RNAs have been detected with concurrent evidence of no corresponding genomic rearrangement and no readthrough transcription between its parental genes. The best example of a trans-spliced chimeric RNA is JAZF1-JJAZ1 (SUZ12), as described by Li et al in 2008. RNA trans-splicing assay was used to prove that the mechanism of this chimeric RNA in normal tissue is trans-splicing between precursor messenger RNAs for JAZF1 and JJAZ1.¹⁹ Other examples of trans-spliced chimeras are TMEM79-SMG5 and CYCLIN D1-TROP2. While there is no direct evidence of trans-splicing for either transcript, these chimeric RNAs have been found in cells without evidence of corresponding genomic rearrangement. Further, the orientation of the parental genes is not conducive to readthrough transcription. The TMEM79 gene and SMG5 are located on opposite strands in opposing orientations within the 1q22 locus, whereas CYCLIN D1 and TROP2 are located on entirely different chromosomes. Interestingly, TMEM79-SMG5 is highly differentially expressed in human prostate cancer samples,²⁰ and CYCLIN D1-TROP2 is able to transform naïve, primary cells in vitro and induce aggressive tumor growth in vivo in cooperation with activated RAS.²¹

Detection of gene fusions and chimeric RNAs

The first gene fusion was discovered by chromosome banding techniques. Additional techniques that can be used to detect gene fusions include Fluorescence In Situ Hybridization (FISH), Southern blotting, Comparative Genome Hybridization (CGH), PCR, and whole genome sequencing, which are based on chromosomal rearrangement. For the gene fusions that result in the formation of chimeric RNAs, RNA based assays can be used as surrogates. However, chimeric RNAs generated by trans-splicing or cis-SAGe cannot be detected by the DNA based assays, as they are produced in the absence of chromosomal rearrangement. Technologies used to detecting these chimeric RNAs are RT-PCR, Northern blotting, RNAse protection assay, and RNA sequencing. The development of microarray technology and Next-generation sequencing has lead to the discovery of a large number of gene fusions and chimeric RNAs. Thousands of chimeric RNAs and gene fusions are now deposited into several databases including Mitelman,²² FusionGDB,²³ ChimerDB,²⁴ FusionCancer²⁵ and ChiTaRs.²⁶

The cart: chimeric RNA which precede gene fusion

Remarkably, several chimeric RNAs found in healthy cells exactly mimic oncogenic fusion transcripts in the absence of the corresponding rearrangement. Trans-spliced JAZF1-JJAZ1 is a premier example of this phenomenon. The JAZF1-JJAZ1 gene fusion is generated by the t(7; 17) (p15; q21) translocation and is found in approximately 50% of human endometrial stromal sarcomas (ESSs).27, 28 The translocation unites the first 3 exons of JAZF1 to the last 15 exons of JJAZ1, and gives rise to a chimeric transcript, which is translated into chimeric protein. Forced overexpression of JAZF1-JJAZ1 in HEK 293 cells confers resistance to apoptosis and promotes cell proliferation when cooperating with suppression of the unrearranged JJAZ1 allele.²⁸ Li et al detected the identical JAZF1-JJAZ1 chimeric RNA and protein in normal endometrial stromal cells and proved absence of translocation between chromosome 7 and chromosome 17 in normal endometrial stromal cells.¹⁹ This study provided new insight into the relationship between chimeric RNA and gene fusion: the chimeric RNA normally generated by trans-splicing in developing tissue could potentially lead to genome rearrangement, thus generating the corresponding gene fusion through an unknown mechanism.⁴

Another “cart” is chimeric PAX3-FOXO1. The PAX3-FOXO1 gene fusion arises from a t(2; 13) (q35; q14) translocation exclusively expressed in alveolar rhabdomyosarcoma (ARMS). This rearrangement joins the DNA binding domain of PAX3 to the transactivation domain of FOXO1, creating a new transcription factor.²⁹ Yuan et al found that the chimeric PAX3-FOXO1 transcript identical to the gene fusion transcript in ARMS was transiently expressed in pluripotent cells differentiating into skeletal muscle without the t(2; 13) (q35; q14) translocation. Forced overexpression of PAX3-FOXO1 led to continuous expression of MYOD and MYOG which were also overexpressed in rhabdomyosarcoma cells,³⁰ whereas silencing the fusion led to failure of MYOD and MYOG expression. These findings support the idea that transiently expressed PAX3-FOXO1 by trans-splicing plays an important role in myogenesis. The generation of the PAX3-FOXO1 gene fusion results in constitutive expression of the chimeric transcript and resulting chimeric transcription factor, which establishes super enhancers directly to drive itself, MYOD1, and MYCN, and indirectly to drive MYOG, thus arresting cells in a premature myogenesis stage.³¹ Of note, PAX3-FOXO1 expression is not enough to cause transformation alone,32, 33 but promotes tumorigenesis in conjunction with inactivation of CDKN2A and overexpression of TERT and MYCN.³⁴

These examples each support the claim that chimeric RNAs serving a developmental or cell/tissue specific role in normal cells without chromosomal rearrangement may potentially mediate gene fusion in the same-lineage cancer cells.

RNA-mediated genome rearrangement

RNA-mediating genome rearrangement is not a new concept in biology. In fact, RNA-induced genome rearrangement is a common feature of ciliates due to nuclear dimorphism.35, 36 In the ciliate Oxytricha, scientists have shown that maternal RNA templates can guide DNA assembly, and disruption of these specific RNAs disables the corresponding gene assembly. Further, injection of synthetic RNA templates into Oxytricia can mediate targeted genome rearrangement.³⁷

RNA-mediated genetic change is not exclusively to ciliates. Shen et al used RNA-containing oligos as templates to repair a Double-Strand Break (DSB) in human cells and introduce base changes in genomic DNA.,³⁸ demonstrating that RNA sequences can have a direct role in DNA genetic modification and remodeling.

A recently published study offers some direct evidence that chimeric RNAs may facilitate gene fusion. Sachin et al showed that forced expression of a chimeric RNA can lead to genome rearrangement, resulting in generation of the corresponding gene fusion in mammalian cells.³⁹ The authors used the TMPRSS2-ERG and TMPRSS2-ETS fusions, common to prostate cancer, as their models for the study. These gene fusions are particularly interesting, as their parental genes are separated by considerable genomic distance, but move into close three-dimensional proximity in response to androgen stimulation. The TMPRSS2-ERG fusion was induced only in samples treated with dihydrotestosterone (DHT, a metabolite of testosterone) and expressing chimeric RNA templates spanning the canonical junction site. The authors proposed a potential mechanism in which the chimeric RNA template forms an imperfect stem with the sense TMPRSS2 and ERG or EVT1 genomic sequence, forming a three-way junction in a sequence-specific manner that brings the canonical breakpoints for the fusion into proximity. Surprisingly, the fusion was preferentially induced when expressing the antisense TMPRSS2-ERG template, potentially indicating that active transcription can prevent this junction from forming on the antisense DNA strand. In support of this hypothesis, RNA polymerase-II inhibition via α-amanitin successfully allowed for gene fusion generation by sense chimeric RNAs³⁹ (Fig. 1a). Whether these structures can form from endogenous chimeric RNA awaits further investigation.

Figure 1.

Schematic diagram of mechanisms of chimeric RNA mediated gene fusion. a, chimeric RNA invades chromosomal DNA of 2 genes and make a transient RNA/DNA hybrid, DNA break/repair mechanisms finally generated the corresponding gene fusion. b, RNA-poise model. Transcript of one gene (RNA 1) preinstalled on gene 2's genomic region, allowing for the possibility of trans-splicing (RNA Targeting). Spatial proximity of 2 genes could bring the 2 transcripts near each other, thus allowing for the possibility of trans-splicing (RNA Confinement). The RNA Confinement model also could enhance the possibility of genome rearrangement thus generate fusion gene.

Another study into RNA-DNA interactions proposed the RNA-Poise model as an explanation for the cart-before-the-horse phenomenon. This model emphasizes the importance of three-dimensional proximity in generating chimeric transcripts without rearrangement. In the first subtype of the model, or the RNA Targeting model, the transcripts of gene 1 preinstall onto gene 2's genomic sequence, thus allowing for the spatial proximity of transcripts of the two genes, and trans-splicing between transcripts. The second subtype of the RNA-Poise model, the RNA Confinement model, dictates that spatial proximity brings the nascent transcripts of gene 1 close to the genomic sequence of gene 2. In both cases, proximity of the genomic regions increases the chance of translocation (Fig. 1b). Additionally, this study also observed a singular lung cancer sample expressing the EML4-ALK chimeric RNA without harboring the corresponding EML4-ALK gene fusion. This is especially important, as the authors have potentially captured both the “before” and “after” states within the same study.⁴⁰

Chimeric RNAs and gene fusions as cancer diagnostic biomarkers and treatment targets

The tumor-specificity of gene fusions makes them ideal biomarkers for cancer. BCR-ABL1, has been widely utilized as a biomarker and prognosis factor in acute lymphoblastic leukemia (ALL) patients,⁴¹ and its specific inhibitor Gleevec (Imantinib) is used as an effective cancer drug targeting the BCR-ABL gene fusion for CML and ALL patients.¹ The TMPRSS2-ERG gene fusion is regarded as an early event in prostate cancer and is positively correlated with Gleason score, which is used to help evaluate the prognosis of men with prostate cancer.⁴² A series of peptides inhibiting ERG-mediated transcription have recently been identified by Wang et al and can reduce cell invasion, proliferation, and tumor growth.⁴³ ROS1 rearrangement and EML4-ALK account for 4% of non–small-cell lung cancers (NSCLC) carcinogenesis and is effectively targeted by crizotinib.2, 44, 45 The EVT6-NTRK3 fusion is found in 92% of human secretory breast carcinomas, and thus it is defined as a diagnostic biomarker.⁴⁶ The PAX-FOXO1 fusion is found in 80% of aRMS patients,47, 48 and has been recognized as a superior biomarker of poor event-free survival.49, 50

However, even though plenty of chimeric RNAs have been discovered in cancer and reported as biomarkers, not all are exclusively found alongside gene fusions, and not all are truly exclusive to cancer. Examples provided earlier in this manuscript such as JJAZ1-JAZF1 and PAX3-FOXO1 argue against exclusivity of these transcripts in cancer, and even several leukemia related fusion transcripts can be detected in healthy individuals.51, 52, 53 Further, many chimeric RNAs have been found in normal samples, 13 of which have been found to overlap with existing annotations of supposed cancer-specific chimeras.⁵⁴ Taken together, gene fusions and chimeric RNAs have major impacts on cancer diagnosis and treatment; however, detection of either is not necessarily indicative of cancer. Thus, it is important to thoroughly validate the candidate marker before projecting its use to translational applications.

Conclusions and future perspectives

In this review, we provide an overview of similarities and differences between gene fusions and chimeric RNAs. We establish known origins of gene fusions as well as chimeric RNAs, and note that each may be able to influence the generation of the other. Gene fusions and the generation of their downstream chimeric products are generally well-studied; however, the inverse “cart before the horse” hypothesis is gaining traction as a possible means for existing chimeric RNAs to influence DNA-level changes.

Both gene fusions and chimeric RNAs have strong associations with cancer. They have been successfully utilized to improve patient diagnosis, treatment, and prognosis, and their value in these avenues cannot be understated. As gene fusion databases expand with the development of RNA sequencing technology, we urge additional caution in validation of these markers, as much evidence has arisen to indicate that patterns of fusion events are not sweeping indications of cancer.

Improvement in sequencing technologies has accelerated the pace at which study into chimeric RNAs has progressed. We expect that with the advent of new sequencing technologies such as low-cost whole genome sequencing and full-length sequencing, this pattern will continue. Ongoing research into this field has the potential to elucidate new mechanisms for oncogenesis which could have significant translational consequences.

Conflict of interest

Authors declare no conflict of interest.