Circular RNAs: a new class of biomarkers as a rising interest in laboratory medicine

Introduction

Over the last decade, whole-genome sequencing studies have revealed that a significant part of the human genome encodes transcripts that are not translated into proteins. These non-coding RNAs comprise a large group of different small (<200 nucleotides) to long (>200 nucleotides) RNA molecules such as microRNAs (miRNAs), piwi-interacting RNAs and large intergenic non-coding RNAs [1]. A new class of RNAs, named circular RNAs (circRNAs) in reference to their structure, was characterized and validated within the last 6 years after Salzman et al. [2] discovered in 2012 that circRNAs are ubiquitously expressed and highly conserved. CircRNAs are single-stranded covalently closed RNA molecules lacking 5′-3′ polarization and the characteristic poly(A) tail of linear RNA [3].

Covalently closed RNA rings were first discovered in viroids in the 1970s using electron microscopy [4]. Until the early 2010s, they were regarded as transcriptional trash due to mis-splicing, similar to non-coding RNAs [5]. However, the development of high-throughput sequencing technologies and the considerable advances in bioinformatics and systems biology in recent years has enabled researchers to identify human circRNAs as stable and abundant splicing products [6]. Similar to non-coding RNAs, circRNAs function as critical gene expression regulators in most cellular processes in both healthy and diseased tissues. The continuous increase of circRNA-related publications in clinical medicine highlights the potential development of circRNAs as promising biomarkers and therapeutic targets (Figure 1), suggesting that there are new diagnostic opportunities within laboratory medicine as well as analytical challenges in translating novel molecular insights into the potential for non-invasive biomarkers.

Figure 1:

Annual publications on circRNA indexed in the PubMed database.

The search was performed for the period from January 2012 to December 2017. The publications are classified based on whether they are related to basic research, cancer or other diseases.

This review should, therefore, draw the attention of clinical chemists and laboratory physicians to this rapidly developing field. The aim of this short overview is to focus on recent progress in circRNA research and to make the scientific community of laboratory medicine aware of this and future perspectives. Some particular issues facing circRNA research are shown through the example of renal cell carcinoma using our own data [7].

CircRNA characteristics and biogenesis

A major characteristic of circRNAs is how they are formed, which occurs through a back-splice junction that joins the downstream 3′-end of an RNA sequence with the upstream 5′-end during splicing [3]. CircRNAs result from an alternative splicing process in addition to the regular canonical splicing mechanism (Figure 2). The number of circRNAs differs between cell types. Approximately 500–17,000 circRNAs have been found in 15 distinct cancerous and non-cancerous cell lines, totaling ~47,000 circRNAs [6], [10]. Approximately 80%–90% of circRNAs in clinically relevant human tissues are derived from exonic gene sequences [11]. This relevance is also shown in the example of renal cell carcinoma (Figure 3A). Exonic circRNAs are usually composed of 1–5 exons [8], [12]. However, intronic, antisense and intergenic genomic regions can also serve as sources for circRNAs (Figures 2 and 3A). CircRNAs can also originate from a single host gene and can contain the same back-splice junction, forming multiple circRNA isoforms [10]. In the case of renal cell carcinoma, ~53% and 93% of host genes show more than one but less than five circRNAs, respectively (Figure 3B). However, more than 15 circRNAs can be derived from one host gene.

Figure 2:

Overview of circRNA biogenesis.

(A) CircRNAs are produced by back-splicing circularization; canonical splicing produces linear mRNAs. CircRNAs are preferably produced from internal exons with long flanking introns, and the number of exons being circularized varies. Introns can be included in the final circRNA. (B) Among the several cis-elements regulating circRNA production, repetitive or non-repetitive complementary sequences in flanking introns control circRNA production. Pairing of complementary sequences of two flanking introns promotes circRNA production, whereas pairing of complementary sequences within an intron promotes canonical splicing. (C) Trans-factors, such as RNA-binding proteins, can also promote or inhibit circRNA production. MBL and QKI, for example, bind to flanking introns to promote circRNA production, whereas adenosine-to-inosine editing by ADAR1 prevents pairing of flanking introns to inhibit circRNA production. (Adapted from [8], [9].)

Figure 3:

CircRNA expression is shown on the example of tissue samples from clear cell renal cell carcinoma.

(A) Genomic origin of circRNAs found in microarray analysis of clear cell renal cell carcinoma. The data originated from a microarray performed with seven paired malignant and adjacent non-malignant tissue samples from specimens of clear renal cell carcinomas after nephrectomy. (B) Number of circRNAs expressed per host gene in the same microarray. (Adapted according to the presentation of data at the 9th AuF Symposium, Freiburg, November 16–18, 2017; [7].)

Several possible mechanisms have been suggested for the biogenesis of circRNAs, which is regulated by cis- and trans-acting factors that control splicing [9]. Cis-regulatory elements include canonical splice sites bracketing the exons, exon length and position and the length of the flanking introns as well as reverse complementary sequences within introns. The majority of mammalian circRNAs are processed from internal exons with long flanking introns containing reverse complementary sequences [6], [13]. These complementary sequences can be repetitive, such as ALU repeats in humans, b1 repeats in mice and other repeats in drosophila or non-repetitive sequences as short as 30–40 nucleotides [6], [13], [14], [15], [16]. The pairing of these sequences strongly promotes circRNA production and is competitive. Pairing within one intron favors canonical splicing, whereas pairing of two flanking exons promotes circRNA production. The existence of several complementary regions results in the production of alternative competing circRNAs. However, not all circRNAs require intron pairing, especially in non-mammalian organisms. In Schizosaccharomyces pombe, for example, circRNAs are produced from an exon-containing lariat precursor [17]. Lariats are noose-like intronic RNA structures that have a 2′-5′ connection creating a circle and tail formed by the 3′ region. In lariat-driven back-splicing of exonic circRNAs, the downstream 3′- and the upstream 5′- ends of two exons are connected through spliceosome-regulated exclusion of intronic lariats.

In addition, RNA-binding proteins (RBPs), such as the muscleblind-like splicing regulator 1 (MBL) [18], the KH domain containing RNA binding factor (QKI) [19] or the RNA specific adenosine deaminase (ADAR1) [15], [20], can act as trans-factors during circRNA biogenesis. RBPs can facilitate, control or inhibit intron pairing independent of complementary sequences within the exons.

Biogenesis of circRNA can occur post- or co-transcriptionally. Host gene-independent expression has been described, for example, during cardiac differentiation, indicating that their biogenesis is at least partially independent of each other [21]. Although circRNA levels are generally lower than linear RNA levels and the kinetics of back-splicing is slower than the kinetics of canonical splicing, some circRNAs are highly abundant in specific cell types and tissues or are dependent on cellular and developmental states, suggesting the functional relevance of these molecules [6], [10], [12].

Biological functions of circRNAs

The striking expression differences of circRNAs are an indication of the possible functional relevance of these molecules. Recent investigations have shown that circRNAs can interfere with several molecular processes or are essentially attributed therein. Although circRNA research is at a very early stage, it is becoming apparent that their effects are based on their miRNA and protein binding properties as well as on their potential ability to modulate transcription and translation. Therefore, circRNAs can influence both nuclear and cytoplasmic processes.

CircRNAs can act as miRNA sponges, inhibit the cellular effect of miRNAs by binding them and thus competing with miRNA:mRNA binding [6], [12], [22], [23], [24]; therefore, the inhibited mRNA targets can be expressed again. The first discovered miRNA-sponging circRNA is derived from the gene CDR1 (cerebellar degeneration related protein 1) and regulates miRNA-7 and miR-671 in the brain [12], [22], [24]. CircHIPK3 (homeodomain interacting protein kinase 3), another miRNA sponge, has been functionally associated with cancer progression [6], [23]. CircHIPK3 is highly expressed in cancer tissue, has multiple miR-124 binding sites and controls proliferation in cancer cell lines [6], [23]. The miRNA-binding capacity of circRNAs has also been correlated with their ability to act as transporters of miRNAs from the nucleus into the cytoplasm and finally into the extracellular compartment through exosomes [25]. CircNet (http://circnet.mbc.nctu.edu.tw) and CircInteractome (https://circinteractome.nia.nih.gov) are useful web tools to screen for probable miRNA binding sites of a candidate circRNA [26], [27]. Tissue-specific data selection is also possible. These databases comprise sequencing data from numerous RNA samples of several studies that enable the user to find the miRNA binding sites of a given circRNA (Supplemental Table S1). The above-mentioned circRNAs showed possible effects on cellular processes by sponging specific miRNAs. Even more circRNAs are likely able to sponge miRNAs [27]. However, the exact functions of most circRNAs have yet to be proven experimentally and require further exploration.

CircRNA biogenesis has also been shown to compete with canonical splicing, creating an mRNA trap function [28]. This feature has been most clearly shown for mbl/ MBLN1, which is circularized in flies and humans [18]. Mbl has multiple binding sites in the flanking introns of the circularized exon and in the circRNA itself. When expressed, MBL protein has been shown to promote production of the circRNA by binding to the flanking introns. In turn, circMbl sponges up excess protein to regulate MBL protein levels [18].

Data show that intron-containing circRNAs, which are predominantly localized in the nucleus, enhance the transcription of their parental genes by interacting with RNA polymerase II, likely in association with the small nuclear U1 snRNA [29]. Another function of circRNAs related to transcription is that they bind to the DNA of their host gene, a phenomenon demonstrated in plants but not in eukaryotes until now [30]. This DNA-circRNA hybrid can also impair polymerase II activity.

With regard to the potential translation ability of circRNAs, certain circRNAs were recently shown to be associated with translating ribosomes [31]. The authors provided strong evidence for a 5′-cap-independent translation of circRNAs based on results of different experiments, including the detection of circRNA-encoded proteins and specific peptides. Using mass spectrometry, other authors also identified peptides that are unique to circRNA-encoded proteins [32].

Analytical challenges in circRNA research

Challenges exist in the determination of circRNAs, specifically regarding their sensitivity and specificity. The reasons for these challenges include the following: (a) the particularity of back-splice junctions in circRNAs; (b) the fact that the generation of circRNA and linear RNA from the same parent genes results in matching sequence portions, as exons are shared between the circRNA and the annotated mRNA; and (c) the frequent low abundance of circRNAs in comparison to the abundance of linear transcripts. Problems in the determination of circRNAs are comparable, if not more complicated, than the determination of other RNA forms. Therefore, a multistep strategy is advisable and has already proven successful for the reliable validation of miRNAs or piwi-interacting RNAs [33]. Such an approach should consist of three steps as follows: (1) the discovery and identification of circRNAs, (2) the validation of those circRNAs of interest using other molecular biology-based methods and (3) their reliable quantification.

Deregulated circRNA expression profiles in diseases, particularly in cancer

The detection of differentially expressed circRNAs in disease is a field of growing clinical and scientific interest. Although the function of most circRNA molecules in relation to physiology and disease has not been investigated to date, their differential expression in healthy and diseased tissues makes them promising biomarker candidates. For this purpose, deregulated circRNA candidates are identified by analyzing disease-specific circRNA expression profiles. As such, circRNAs could be used for the prevention, diagnosis, prognosis, treatment and/or follow-up of a pathological process.

Valid reports outlining the potential use of differential expression of circRNAs as a clinical tool are currently limited, although in 2017 a number of exploratory studies addressed this question. A few selected examples are given in Table 1. Comparison of circRNA expression in several human tissues from the same donors revealed high tissue specificity of circRNA and deregulation in diseased donors [11]. Memczak and colleagues [61] presented the first evidence that circRNAs could act as circulating biomarkers in whole blood, where s pecific circRNAs were detected in a higher abundance than their linear counterparts (Table 1, study no. 1). Because circRNAs consist of covalently closed rings without poly(A) tails, they show a superior stability than mRNAs, which are quickly degraded in body fluids. This feature makes circRNAs very interesting as non-invasive biomarkers. Further examples of deregulated circRNA expression, both in tissue and peripheral blood samples, exist for Alzheimer’s disease [71], rheumatoid arthritis [54], active pulmonary tuberculosis [72] and diabetes mellitus type 2 [55] (Table 1, study nos. 2–5).

Most studies to date have addressed deregulated circRNA expression in cancer (Figure 1). A comparative study between seven cancer types (bladder urothelial carcinoma, breast cancer, colorectal cancer, gastric cancer, hepatocellular carcinoma, clear cell renal carcinoma and prostate cancer) showed an overlap of 19,071 circRNAs and confirmed the general feasibility of circRNA detection in several cancer tissues [23]. Two current reviews and a cancer-specific circRNA database provide more detailed information on the role of circRNAs in different cancers [41], [73], [74]. For example, microarray data of lung adenocarcinoma samples detected higher expression of hsa_circ_0013958 in cancer tissue than in adjacent normal tissue in the same patients (Table 1, no. 6, [52]). Validation experiments with RT-qPCR in 49 tissue pairs supported the microarray results. Furthermore, differential expression was confirmed in plasma samples of these patients and corresponding cancer cell lines. For hepatitis B-related hepatocellular carcinoma, microarray data showed a significant up-regulation of hsa_circRNA_10338 in cancer tissue compared to expression in adjacent normal tissue (Table 1, study no. 7 [53]).

However, all of the examples show that these clinically oriented studies do not go beyond the concept of pilot studies and are mainly of a descriptive nature. Both analytical and clinical requirements for studies recommended in the guidelines of “Standards for Reporting of Diagnostic Accuracy Studies” and “Reporting Recommendations for Tumor Marker Prognostic Studies” including the multivariate evaluation of data have yet to be exploited [75], [76].

Future directions for laboratory medicine: circRNAs as promising biomarkers

To date, the biological functions and clinical significance of circRNAs in disease development and physiological processes remains unclear. However, circRNAs are expressed highly specific under physiological and pathological condition [6], [10], [20]. Furthermore, experimental data prove that circRNA expression is independent from the expression of linear counterparts originating from the same gene locus [21]. These facts underline the possible regulatory function of circRNAs, leading to their potential role as disease-indicating biomarkers with special interests for clinical laboratory scientists. However, as comprehensive functional studies of circRNAs are still missing to demonstrate their biological function, it cannot be excluded that the majority of circRNAs are possibly functionless transcriptional byproducts only tolerated by the cells. Therefore, future efforts in human circRNA research will be focused on unraveling which pathways are affected by specific circRNAs, either independently or as part of a network. In the case of cancer, for example, research will focus on how circRNAs act as tumor-suppressive or oncogenic factors [77], [78].

This tissue-based biological background is strongly connected with the challenge to utilize circRNAs for the same purpose as other nucleic acids as tissue-equivalent biomarkers because circRNAs could be released from diseased tissues or organs into easily accessible biofluids such as blood or urine [33]. Although the turnover of circRNAs and the mechanisms of their release into the extracellular compartment are unknown [79], numerous circRNAs have been found in the cell-free body fluids of healthy individuals and specific cancer patients [42], [56], [80], [81], [82]. RNA-Seq studies and RT-qPCR measurements have shown that tumor-specific circRNAs are enriched in serum-derived exosomes/extracellular vesicles but not in exosome-depleted serum [42], [82]. The presence of circRNAs in these vesicles and their specific molecular structure of covalently closed rings has been suggested as the reason for their high stability in body fluids because they are protected from RNases present in body fluids [42]. Moreover, blood cells, especially thrombocytes, contain a large number of circRNAs [61], [83]. They can be differentially released depending on the blood collection method (serum or plasma). The presence of these other sources of circRNAs could substantially interfere with the diagnostic specificity of circRNA testing.

To avoid these complications and potential errors, basic and applied researchers, in collaboration with clinicians, need to work together to translate circRNA research findings for the potential use of these circRNAs as biomarkers. The introduction of other nucleic acid-based assays into clinical practice has proven that this will be a multiphase process (Figure 4) [33]. Clinical chemists and laboratory physicians, with their specific knowledge of and experience with body fluids, should play a significant role in both the successful development of circRNAs assays and in their clinical validation. Laboratory scientists are the link between the first two (basic science) and the last two (clinical) phases of this multiphase process (Figure 4). In this process, after the discovery and experimental validation of a circRNA as mentioned above, clinical validation must be performed using robust RT-qPCR-based assays. All variables in the preanalytical phase (sample collection, processing and storage conditions), the analytical phase (RNA isolation protocols and quantification principles) and the postanalytical phase (data evaluation and normalization strategies) that could influence the determination of circRNAs must be considered and documented with standard operating procedures [76]. More detailed information is given in Supplementary Table S2. Special attention should be paid, based on experience with other nucleic acid-based assays, to the possible interfering effects of preanalytical factors on circRNA measurements in serum/plasma and urine samples [33]. Although cancer studies with serum-based measurements of circRNAs have been published [42], [81], [82], there has been no information until now on how the preparation of the sample can affect the measurements. For example, there are many miRNAs compiled in a catalogue that cannot be accurately measured in serum/plasma or urine samples containing free hemoglobin undetectable by visual inspection [84], [85]. Due to this interference, an additional quality criterion was implemented to eliminate inappropriate samples [84]. A similar approach could be established for circRNAs.

Figure 4:

Developmental phases of circRNA assays for clinical practice. (Adapted from [33].)

In the two clinical validation phases based on retrospective and prospective studies performed at single or at multiple centers (Figure 4), the laboratory scientist has the main task of supervising the quality of the measurements. Based on the analytical and biological variation of data, the laboratory scientist then contributes substantially to the practice-oriented evaluation of the data [86].

Conclusions

CircRNAs are an interesting new class of RNAs and could potentially be used as biomarkers in the future. In contrast to early reports that circRNAs are random splicing byproducts, whole genome studies have shown that circRNAs are stable and that their biogenesis is tightly regulated. The expression of circRNA is highly tissue- and cell-state specific, and exploratory studies have proven the general feasibility of the concept to use circRNAs as biomarkers. Their gene regulatory function, especially as a regulator of miRNAs, also makes them putative therapeutic targets. However, several limitations must be addressed before circRNAs can be evaluated in clinical studies, including detection biases and the absence of a consensus regarding validation protocols and nomenclature. More studies are needed to understand the deregulated expression, abundance, stability and function of circRNAs in specific diseases and disease stages and to confirm if this new class of RNAs is of clinical importance. For laboratory scientists, circRNA research and its translation into clinical practice is a great challenge but also a very exciting new task.