Recent progress in circular RNAs in human cancers

Highlights

• CircRNAs are no longer regarded as “non-coding” RNA, since they can be translated to protein driven by m⁶A modification.

• For most circRNAs rarely contain miRNA target sites, miRNA sponge might be not the predominant mode of circRNA function.

• CircRNAs were found to be dysregulated in cancer cell lines, tumor tissues, and even plasma samples from patients.

• CircRNAs were deemed to be promising biomarkers for the early diagnosis and prognosis prediction of cancer.

Abstract

Circular RNAs (circRNAs) are a large class of endogenous RNAs, formed by exon skipping or back-splicing events as covalently closed loops, which are expressed abundantly in mammalian cells. Although their biological functions remain largely unknown, recent studies show that circRNAs have three main functions in mammalian cells. First, circRNAs can regulate transcription and RNA splicing. Second, circRNAs function as microRNA (miRNA) sponges. Third, they can be translated into protein driven by N⁶-methyladenosine modification. Taking advantage of RNA sequencing (RNA-seq) technology, the expressions of circRNAs were found to be dysregulated in all types of cancer cell lines, tumor tissues, and even plasma samples from patients, which correlated with certain clinical characteristics, suggesting the potential roles of circRNAs in tumor progression. Considering their conserved sequences and stable structures, circRNAs were deemed to be promising biomarkers for the early diagnosis and prognosis prediction of cancer. In this review, we describe briefly the formation and properties of circRNAs, and focus mainly on recent progress in research into their function, regulation, and clinical relevance in different cancers.

Introduction

Circular RNAs (circRNAs) are a large class of endogenous RNAs that are formed by exon skipping or back-splicing events; however, they attracted little attention until their function in post-transcriptional regulation of gene expression was discovered. Potato spindle tuber viroid (PSTVd) was the first identified circRNA in 1976 when researchers studying potato spindle tuber disease observed that the viroid could infect plants and cause death. Different from viruses, the viroid lacks a protein envelope and the genome is a closed, single-stranded RNA molecule [1]. In 1979, Hsu and Coca-Prados observed the presence of a circular form of RNA in the cytoplasm of several eukaryotic cells using electron microscopy [2].

In the early 1990s, circRNAs in higher eukaryotes were discovered. The first clue to the mechanism of endogenous circRNA generation emerged from studies of the transcripts of the tumor suppressor gene DCC. Very low levels of transcripts of the DCC gene with exons joined accurately at consensus splice sites were found in normal and neoplastic cells, primarily in the nonpolyadenylated component of cytoplasmic RNA. These results demonstrated that the splicing process does not always pair sequential exons in the order predicted from their positions in the genome, which was called “exon scrambling” [3]. Cocquerelle et al. identified a scrambled transcript of the human c-ets-1 gene, which is non-polyadenylated and is expressed at much lower levels than the normal transcript [4]. They further determined the structure of these transcripts as circular RNA molecules containing only exons in the genomic order [5]. In adult mouse testis, circular transcripts of the testis-determining gene sex-determining region Y (Sry) were detected. These circular RNAs, which represent the most abundant transcript in the testis, are located in the cytoplasm, but are not bound substantially to polysomes [6]. Later, atypical RNA molecules containing an incomplete exon tandem repetition, or having exons with a different order compared with the corresponding genomic DNA, were identified from the Drosophila melanogaster muscleblind (mbl) locus. Considering its lack of polyadenylation and downstream splicing events, its small size, and polyacrylamide gel electrophoresis (PAGE) behavior, the non-canonical transcript mblE2E2′ was deemed to be the first identified circular RNA in invertebrates [7].

Recently, taking advantage of RNA-seq technology and bioinformatic tools, more circRNAs have been discovered and characterized. In 2012, Salzman et al. developed an algorithm to detect scrambled exons in RNA-seq datasets of five bone marrow samples from pediatric acute lymphoblastic leukemia (ALL). They identified more than 1232 genes with evidence of exon scrambling, which were further validated by reverse transcription polymerase chain reaction (RT-PCR). Intriguingly, all examples of exon scrambling were also detected in peripheral blood collected from the same ALL patients and H9 ES cells, providing strong support for the view that a circular RNA isoform resulting from a non-canonical mode of RNA splicing is actually a general feature of the gene expression program in diverse human cells [8]. This group further developed a new bioinformatic approach and investigated circRNA expression using published RNA-Seq data from Drosophila brains and a series of cancer and non-cancer cell lines. They showed that circRNA expression was conserved evolutionarily across model organisms. In addition, the expression profiles, the ratio of circular to linear transcripts, and the pattern of splice isoforms of circRNAs of individual genes were cell-type specific. They also estimated that circular RNAs might account for about 1% of poly(A) RNA in humans [9]. Jeck et al. identified over 25,000 distinct RNAs containing non-linear exons from 14.4% of actively transcribed genes in human fibroblasts. Surprisingly, the abundance of certain circRNAs was 10-fold more than that of associated linear mRNAs. Bioinformatic analysis revealed that these circularized exons were always flanked by long introns that contained complementary ALU repeats. Moreover, they found that circRNAs could be degraded by siRNAs, suggesting their potential role as competing endogenous RNAs [10], [11]. Combining ribominus sequencing data for HEK293 cells and human leukocyte data, Memczak et al. identified 1950 kinds of human circRNAs from at least two independent junction-spanning reads. They also identified 1903 mouse circRNAs (81 of these mapped to human circRNAs) and 724 circRNAs from various C. elegans developmental stages. Further studies focused on a human circRNA, which is antisense to the cerebellar degeneration-related protein 1 transcript (CDR1as), and identified the regulatory potential of circRNA as an miRNA antagonist [12]. Thereafter, they sequenced RNA in human peripheral whole blood and detected thousands of circRNAs reproducibly. Hundreds of circRNAs were expressed at much higher levels than the corresponding linear mRNAs, which were not accessible by classical mRNA specific assays, demonstrating the potential of circRNAs as biomarkers for human disease in an easily accessible body fluid [13].