Tip of the iceberg: roles of circRNAs in hematological malignancies

Abstract

Circular RNAs (circRNAs) are a new class of covalently closed RNA molecules whose 3’- and 5’-ends are linked by a back-splicing event. Emerging evidence has shown that circRNAs play a vital role in the occurrence and development of many diseases and are promising biomarkers and therapeutic targets. However, knowledge of circRNAs in hematological malignancies is limited. In this review, the biogenesis, categories, characteristics, and functions of circRNAs are summarized, especially the roles of circRNAs in hematopoiesis and hematological malignancies.

Introduction

CircRNAs, first identified in 1976 in viruses to have a single-strand covalently closed structure, were considered viroids due to a lack of knowledge about them [1]. Subsequently, researchers discovered several circRNAs in different eukaryotic cells and speculated that they were generated by base-pairing between the two ends of RNA molecules [2]. In recent years, abundant evidence has shown that circRNAs are richly and widely expressed in eukaryotes and participate in most biological processes, including carcinogenesis. For example, serving as a sponge for miR-424-5p, circLARP4 suppresses cell proliferation and invasion of gastric cancer by regulating the expression of LAST1 [3]. CircCSNK1G3 can promote prostate cancer cell proliferation by acting as a stabilizer for miR-181b/d [4]. Ci-ANKRD52 has been suggested to regulate its parental gene by influencing the elongation of Pol II [5]. Some circRNAs, such as circZNF609 [6] and circ-MBL [7], function as templates for protein translation. A growing number of studies have shown that circRNAs are also involved in hematopoiesis and hematological malignancies. Although Bonizzato et al. [8] summarized the roles of circRNAs in hematopoiesis and hematological malignancies, they focused on the biogenesis and general functions of circRNAs due to the lack of research on hematological diseases at that time. Liu et al. [9] also summarized the roles of ncRNAs (noncoding RNAs) in acute myeloid leukemia, but they focused on acute myeloid leukemia and noncoding RNAs. This article provides a comprehensive outlook on circRNAs from their biological features such as biogenesis, categories, characteristics and functions to their roles in hematopoiesis and hematological malignancies. Although circRNAs have been shown to play a variety of roles in hematological diseases, our understanding of circRNAs may be just the tip of the iceberg.

Biogenesis and categories of circRNAs

The biogenesis of circRNAs can occur during and after transcription by a back-splicing process [10]. According to their different origins, four types of circRNAs have been found, namely, circRNA from pre-mRNA, tricRNA (tRNA intronic circRNA) [11-13] from pre-tRNA, f-circRNA (fusion-circRNA) [14] from gene fusions, and rt-circRNA [15,16] from transcription read-through. In addition, circRNAs originating from pre-mRNAs can be divided into four subcategories according to their composition: exonic circRNA (ecircRNA), circular intronic RNA (ciRNA), exon-intron circRNA (EIciRNA), and intergenic circular RNA (intergenic circRNA). The classification of circRNAs is shown in Table 1.

Table 1.

Classification of circRNAs

CircRNA	Origin	Components	Location	Potential function
ecircRNA	pre-mRNA	exon	cytoplasm	biomarker, miRNA sponge, translating, etc.
ciRNA	pre-mRNA	intron	nucleus	regulating splicing and transcription
EIciRNA	pre-mRNA	exon and intron	nucleus	regulating splicing and transcription
intergenic circRNA	pre-mRNA	intergenic sequence	-	-
tricRNA	intron-containing pre-tRNA	intron from pre-tRNA	-	-
f-circRNA	fusion gene	two or more genes from different strands	-	oncogene
rt-circRNA	read-through transcript	two or more adjacent gene from the same strand	-	-

EcircRNAs and EIciRNAs may have the same biogenetic mechanisms [17]. Three mechanisms have been proposed for their biogenesis: direct back-splicing or intron pairing-driven circularization [18] (Figure 1A), exon-skipping or lariat-driven circularization [18] (Figure 1B), and RNA-binding protein-driven circularization [19] (Figure 1C). CiRNA biogenesis relies on a motif consisting of a 7 nt GU-rich component near the 5’ splicing site and an 11 nt AC-rich element near the branchpoint site (Figure 1D) [5]. An intergenic circRNA derived from chromosome 5 was detected by CIRI (CircRNA Identification, a new circRNA identification tool). It contains two intronic circRNA fragments, which are spliced at the flanking GT-AG splicing signals to form a complete circRNA (Figure 1E) [20]. The intron-containing pre-tRNAs are cleaved into exon and intron termini by specific motifs and enzymes, and then the intron termini can be circularized to form tricRNA [11,13] (Figure 1F). F-circRNAs are generated from translocated or rearranged chromosomes, which results in the connection of two otherwise isolated genes. Subsequently, the flanking intron complementary sequences of the junction point allow the formation of a f-circRNA (Figure 1G) [14]. Rt-circRNAs are a novel class of circRNAs that contain several exons stemming from different genes [15,16]. In some cases, such as the pre-mRNA 3’ processing was suppressed, RNA Pol II can extend to the next gene, which produces read-through transcripts containing two or more adjacent genes [15], after which rt-circRNA can be formed by back-splicing when the intronic repeats are present in the 5’ end of the upstream gene and the 3’ end of the downstream gene (Figure 1J) [15,16].

Figure 1.

Mechanisms of circRNA biogenesis: A. Direct back-splicing or intron pairing-driven circularization; B. Exon-skipping or lariat-driven circularization; C. RBP-driven circularization; D. Biogenesis of ciRNA; E. Biogenesis of intergenic circRNA; F. Biogenesis of tricRNA; G. Biogenesis of f-circRNA; H. Canonical splicing of a fusion gene; I. Canonical splicing of pre-mRNA; J. Biogenesis of rt-circRNA.

Characteristics of circRNAs

To date, circRNAs have been confirmed to have at least the following characteristics. (1) Stability: the circRNA closed-loop structure without a 5’ cap and 3’ tail grants them resistance to RNase R or RNA exonuclease [21]. CircRNAs have a half-life of approximately 24 h, while mRNAs have a half-life of approximately 4-6 h at 4°C [22]. (2) Specificity: circRNAs have cell-type-specific and tissue-specific expression patterns [23]. (3) Abundance: circRNAs are extremely abundant, some with circular to linear ratios greater than 1 [24]. (4) Conservation: circRNAs are evolutionarily conserved in different species [25]. (5) Distribution: circRNAs can be detected in multiple tissues, such as peripheral blood [19], bone marrow [26-30] and exosomes [31].

Functions of circRNAs

Functions of the vast majority of circRNAs remain unclear, but increasing studies have revealed that circRNAs play a crucial role in most biological processes by regulating miRNAs or alternative splicing patterns, interacting with RBPs (RNA-binding proteins), producing proteins, etc. The potential functions of circRNAs are shown in Figure 2.

Figure 2.

Potential functions of circRNAs: A. Serve as miRNA sponges or traps; B. Interact with RBPs; C. Regulate splicing and transcription; D. Act as protein scaffolds; E. Translate proteins; F. CircRNA-derived pseudogenes; G. Transcription.

Serving as miRNA sponges, stabilizers, or traps

CircRNAs can act as miRNA sponges to sequester miRNAs, stabilize miRNAs, or trap miRNAs for storage or transport, thereby regulating the expression of downstream genes. The most typical case is CDR1as (CDR1 antisense), which contains over 70 miR-7 binding sites and a number of almost perfectly complementary miR-671 binding sites. CDR1as can act as a sponge or stabilizer to bind miR-7, leading to the downregulation of miR-7 and thereby restoring the expression of miR-7 target genes [32,33]. Additionally, by binding to miR-671, CDR1as directs its own Argonaute 2 (AGO2)-mediated cleavage, thereby releasing its miR-7 cargo [33,34].

Interacting with RNA-binding proteins

CircRNAs interact with RBPs to form circRNPs, which store, sequester, translocate, and sort proteins and further regulate gene expression [35]. For example, the MBL protein and circ-MBL are generated from the same gene locus, and circ-MBL can interact with the MBL protein to regulate the balance between MBL protein and circ-MBL [36]. Circ-Foxo3 interacts with both p21 and CDK2 (cycle-dependent kinase 2) to form a circ-Foxo3-p21-CDK2 ternary complex, which inhibits the cell cycle by inhibiting the binding of CDK2 and cyclin E and enhancing the binding of CDK2 and p21 [37].

Regulating splicing and transcription

The majority of ciRNAs and EIciRNAs are retained in the nucleus and can regulate gene expression. The ciRNA ci-ANKRD52 can promote the transcription of its parental gene by affecting Pol II elongation [5]. Circ-EIF3J and circ-PAIP2 are EIciRNAs that facilitate the expression of their parental genes in HeLa and HEK 293 cells by binding to the U1 small nuclear ribonucleic proteins (snRNPs) and interacting with RNA Pol II at the promoter regions of their parental genes [17]. Moreover, the biogenesis of circRNAs may be a way to regulate the expression of parental genes. For most circRNAs produced from exons of protein-coding genes, on the one hand, the more an exon is cyclized, the less it appears in the mRNA [38,39], and on the other hand, there is competition for the usage of 5’ and 3’ splice sites between the biogenesis of circRNAs and linear mRNAs [39].

Acting as protein scaffolds

CircRNAs can modulate protein-protein interactions by acting as dynamic scaffolds [35]. Circ-Foxo3 has both p53 and double mouse minute 2 (MDM2) biding sites, and the p53-MDM2 co-IP was affected by circ-Foxo3 overexpression or knockout, so circ-Foxo3 may serve as a scaffold to mediate the formation of the p53-MDM2 complex [40]. These results suggest that circRNAs with both enzyme and substrate binding sites may serve as scaffolds to close the distance between proteins and facilitate protein reactions.

Translating proteins

Although many circRNAs contain the canonical AUG initiation codon of their host gene, initially they were thought to be unable to be translated into proteins because they lack a 5’ cap, which is the machinery and factors necessary for the formation of translation initiation complexes. However, recently, some circRNAs have been found to be translated, which is driven by N⁶-methyladenosine (m⁶A) and ribosome entry site (IRES). The consensus m⁶A motif near the translation start site can drive the protein translation from circRNAs by recruiting the initiation factor eIF4G2 and the m⁶A reader YTHDF3 from the cytosol into the nucleus and binding to them [41]. The m⁶A-driven translation of circRNAs can be inhibited by the m⁶A demethylase FTO and promoted by the adenosine methyltransferase METTL3/14. Additionally, IRES can recruit and bind ribosomes to initiate translation in a cap-independent manner under stress conditions [42]. CircZNF609 [6] and circ-MBL [7] have been found to contain IRES that can bind to polysomes, and polypeptides translated from these molecules have also been verified.

CircRNA-derived pseudogenes

A classic approach for pseudogene production occurs when an mRNA is reverse transcribed into cDNA and the cDNA is inserted into the genome. Pseudogenes produced in this way maintain the same exon sequence as the parental linear mRNA. However, Dong et al. [43] discovered some circRNA-derived pseudogenes in both mice and human genomes, which had an exon-exon linkage in reverse order of their parental genes and might be generated by the same biogenesis mechanism as mRNA-derived pseudogenes.

CircRNAs in hematopoiesis

Hematopoiesis is a strictly regulated process in which hematopoietic stem cells differentiate into blood cells with specific functions and morphologies, and it involves transcription factors [44], miRNA [45], lncRNA [46], TNF [47] and other chemical factors [48]. CircRNAs have been found to be widely expressed in hematopoietic cells and mature blood cells, with expression that can be altered upon differentiation and can be cell-type specific, rendering such circRNAs as potential hematopoietic regulators. For example, circ-BACH1 (exons 2-4) is preferentially expressed in HSCs (hematopoietic stem cells) and MPPs (multipotent progenitors), circ-FNDC3B in NK cells, circ-MYBL1 and circ-SLFN12L in T cells and NK cells, circ-AKT3 and circ CCDC91 in lymphoid cells, and circ-BACH1 (exons 3-4) in monocytes [24]. Surprisingly, circRNAs are more enriched in enucleated cells, platelets, and red blood cells than in other hematopoietic cells [24,31,49]. Alhasan et al. [49] attributed this phenomenon to the fact that mRNAs are more easily degraded than circRNAs and proposed it as a signature of transcriptome degradation. However, platelets are translationally competent and exhibit active splicing upon platelet activation [50-52], and they share only a part of circRNAs with megakaryocytes [24], which leads us to speculate that circRNAs are involved in biological processes in these cells and are not just byproducts of transcriptome degradation. Preusser et al. confirmed this speculation and revealed that platelets could selectively release circRNAs into microvesicles and exosomes as regulatory signals for certain biological processes such as hemostasis, inflammation, and cancer metastasis [31]. Zhang et al. also found that circRNAs may participate in macrophage differentiation and polarization [53]. Additionally, some circRNAs have been discovered to protect bone marrow stromal cells from total body irradiation [54,55]. Cia-cGAS (circular RNA antagonist for cGAS; cGAS: cyclic GMP-AMP synthase) is richly expressed in long-term HSCs (LT-HSCs) [56], and elevated cia-cGAS can bind to cGAS in the nucleus to inhibit cGAS-mediated production of type I IFNs [57-59], thus protecting LT-HSCs from type I IFN-driven exhaustion [60,61]. Broadly and specifically expressed in hematopoietic cells, circRNAs are not the byproducts of transcriptome degradation but are potential biomarkers for hematopoietic cells and regulators of their differentiation, maturation, and functions. However, the mechanisms of how circRNAs regulate hematopoietic cell differentiation, maturation, and functions have yet to be elucidated. The circRNAs involved in hematopoietic cell differentiation, maturation and functions are shown in Figure 3.

Figure 3.

CircRNAs specifically expressed in hematopoietic stem cells, mature blood cells and circRNAs involved in hematopoietic cell differentiation, maturation, and functions. BM: bone marrow, PB: peripheral blood, HSC: hematopoietic stem cell, MPP: multipotent progenitor, CMP: common myeloid progenitor, CLP: common lymphoid progenitor, MkEP: megakaryocyte-erythrocyte progenitor, GMP: granulocyte-monocyte progenitor, TNKP: T lymphocyte-NK cell progenitor, BLP: B lymphocyte progenitor, MkP: megakaryocyte progenitor, EP: erythrocyte progenitor, EoP: eosinophil progenitor, BP: basophil progenitor, NP: neutrophil granulocyte progenitor, MP: monocyte progenitor, TLP: T lymphocyte progenitor, NKP: natural killer cell progenitor, PLT: platelet, RBC: red blood cell, Eo: basophil, Ba: basophil, Ma: mastocyte, Ne: neutrophil granulocyte, Mono: monocyte, DC: dendritic cell, T: T lymphocyte, NK: natural killer cell, B: B lymphocyte, Macro: macrocyte, MSC: mesenchymal stem cell, IFN: interferon, cGAS: cyclic GMP-AMP synthase, cia-GAS: circular RNA antagonist for cGAS.

CircRNAs in hematological malignancies

CircRNAs in acute myeloid leukemia

A large number of circRNAs specifically expressed in acute myeloid leukemia have been identified with the development of next-generation sequencing technology. Li et al. discovered that 464 circRNAs were differentially expressed with a 2-fold change cut-off in cytogenetically normal AML (CN-AML) compared with healthy controls. Among them, a dramatically downregulated circRNA, hsa_circ_0004277, can differentiate AML patients from the healthy individuals with an area under the receiver operating characteristic (AUC) curve of 0.957 in a cohort of 115 samples. Downregulation of this circRNA was restored after complete remission and was again downregulated at recurrence [26]. Similarly, Chen et al. performed circRNA microarray analysis on five pairs of newly diagnosed AML and iron deficiency anemia (IDA) cases, finding that 282 circRNAs were upregulated and 416 circRNAs were downregulated in AML. Next, the authors validated these findngs in a cohort of 87 AML and 45 IDA patients by qRT-PCR, confirming that circ-ANAPC7 was dramatically upregulated [27]. Likewise, Lei et al. tested samples from three pairs of AML and IDA patients for RNA-seq, but detailed information on the circRNA expression profiles was not provided. Only the circ_0009910 was suggested to be upregulated in AML in a group of 70 AML patients and 70 IDA controls, as determined by qRT-PCR. In addition, increased expression of circ_0009910 was found to be correlated with high-risk and a shorter overall survival rate [30]. In another study, circRNA microarray and genome-wide microarray analysis were performed simultaneously for the first time in 4 matched AML patients with and without extramedullary infiltration (EMI) as well as healthy volunteers. The results showed that 253 circRNAs and 663 genes were upregulated, and 259 circRNAs and 838 genes were downregulated in EMI AML patients compared with non-EMI AML patients. One of the upregulated circRNAs in AML with EMI, hsa_circRNA_0004520, was validated by performing RT-PCR on four pairs of EMI and non-EMI AML samples [28]. In a study by Li et al., 246 upregulated circRNAs and 262 downregulated circRNAs were identified by RNA-seq to be dynamically expressed during ATRA-induced differentiation. Some dysregulated circRNAs, including circ-HIPK2, were selected for validation by qRT-PCR, and the verification results were consistent with the RNA-seq results, indicating that some circRNAs showed dynamic regulation during ATRA-induced differentiation and may play a role in ATRA-induced differentiation [62]. Similarly, by screening the circRNA expression profiles of three matched THP-1 and doxorubicin-resistant THP-1 (THP-1/ADM) cell lines, Shang et al. revealed a total of 49 circRNAs to be significantly aberrantly expressed in the THP-1/ADM cell lines. Among them, 35 circRNAs were upregulated and 14 circRNAs were downregulated. One of the aberrant circRNAs, circ-PAN3, was found to be markedly upregulated in refractory/recurrent AML by examining its expression levels in 20 refractory/recurrent AML and 22 chemo-sensitive patients [63]. In several other papers, researchers have found circRNAs of interest by searching online databases or the literature. One of the circRNAs, circ-VIM, was discovered to be significantly upregulated in de novo AML and could discriminate AML (AUC = 0.741, 95% CI: 0.657-0.825), non-APL (acute promyelocytic leukemia) AML (AUC = 0.740, 95% CI: 0.652-0.825) and CN-AML (AUC = 0.749, 95% CI: 0.645-0.852) patients from controls in a 155 sample cohort. Moreover, circ-VIM was an independent prognostic factor because patients with elevated circ-VIM had shorter leukemia-free survival and overall survival [64]. CircRNA-DLEU2 was confirmed to be upregulated in a cohort of 20 CN-AML patients and 20 healthy controls [29]. The hsa_circ_100290, extracted from the GSE94591 dataset of the GEO database (http://www.ncbi.nlm.nih.gov/geo/) [65], was confirmed to be markedly upregulated in AML patients compared with the healthy controls in a cohort of 60 AML and 60 healthy samples [66]. CircMYBL2 was also identified from the GSE94591 dataset and was validated to be upregulated in FTL3-ITD (FMS-like tyrosine-3-internal tandem duplication) AML (an AML with poor prognosis) compared with non-FLT3-ITD AML patients by qPCR [67]. The expression levels of some circNPM1 variants (including hsa_circ_0075001) in AML samples are higher than in normal samples, and subtypes M0 and M1 share a higher expression of hsa_circ_0075001 than M2, M4 and M5 [68]. F-circRNAs have also been found in AML with aberrant chromosomal translocation; for example, f-circPR has been found in APL patients and APL cell lines with the PML/RARα and f-circM9 has been found in AML patients who carry the MLL/AF9 [14,69]. These results suggest that circRNAs are abundant and specific in AML, illustrating that circRNAs may serve as promising potential diagnostic biomarkers, prognostic biomarkers or monitoring indicators for AML.

In addition to acting as potential biomarkers, circRNAs can also regulate cell proliferation and apoptosis by sponging miRNAs or can participate in other modes of action involved in initiation, progression and drug resistance in AML. For instance, circRNA-DLEU2 has been demonstrated to promote PRKACB expression by inhibiting the expression of miR-496, thereby promoting the proliferation of AML cells in vitro and the formation of AML tumors in vivo [29]. Circ_0009910 has been observed to improve the growth of AML cells by inhibiting miR-20a-5p in vitro and in vivo [30]. In another study, knockdown of hsa_circ_100290 significantly arrested the cell cycle in G1 phase, promoted apoptosis, and increased the expression levels of apoptosis-related proteins, such as cyclin D1, CDK4, BCl-2, and cleaved caspase-3. Furthermore, bioinformatics and luciferase assays confirmed that miR-203 bound both hsa_circ_100290 and Rab10, and hsa_circ_100290 upregulated Rab10 expression by sponging miR-203, therby promoting the proliferation of AML cells and inhibiting apoptosis [66]. Circ-HIPK2, derived from a gene closely associated with the initiation and progression of AML as a transcriptional coactivators in the nucleus [70,71], has been shown to be needed for ATRA-induced differentiation of APL cells, but its overexpression alone could not induce differentiation of APL cells. MiR-124-3p can be targeted by both circ-HIPK2 and CEBPA and can suppress the expression of CEBPA, a critical hematopoietic transcription factor [72]. Overall, circ-HIPK2 might promote the ATRA-induced differentiation of APL cells by serving as a sponge for miR-124-3p to restore the expression of CEBPA [62]. CircPAN3, whose host gene PAN3 is located in amplified chromosome regions in AML patients with a complex karyotype [73], may employ the circPAN3-miR-153-5p/miR-183-5p-XIAP axis to mediate drug resistance by acting as a miR-153-5p sponge [63]. CirMYBL2 has been found to recruit and bind with the RNA-binding protein PTBP1, facilitating FLT3/FLT3-ITD translation by facilitating the interaction of PTBP1 and FLT3/FLT3-ITD, and ultimately promoting FLT3-ITD AML progression. In contrast, knockdown of circMYBL2 can suppress FLT3-ITD AML progression and impair the drug resistance of FLT3-ITD-positive AML patients [67]. The functions of the above circRNAs have been experimentally verified. The roles of the following circRNAs in AML have only been predicted by bioinformatic analysis or literature review. Li et al. used bioinformatics to predict the target miRNAs of the hsa_circ_0004277 and the target genes of the above miRNAs. Furthermore, GO (http://www.geneontology.org) [74] and KEGG (https://www.kegg.jp/) [75] analyses have shown that the target genes of hsa_circ_0004277 constitute signatures involved in various biological processes and signaling pathways, such as the regulation of cellular processes and cell junctions, as well as the MAPK, PI3K-Akt and HIF-1 signaling pathways [26]. Hsa_circ_0075001 has been observed to be inversely correlated with the expression of some genes that are known targets of miR-181 and are involved in the TLR signaling pathway [68]. Furthermore, miR-181 is commonly downregulated in CN-AML [76], and the TLR signaling pathway has been suggested to regulate the differentiation of leukemic stem cells [77,78], so hsa_circ_0075001 may function via a miR-181 family/TLR signaling pathway axis. Circ-ANAPC was predicted by web-based tools to contain miRNA response elements (MREs) for the hsa-miR-181 family. A KEGG pathway enrichment analysis showed that the hsa-miR-181 family was strictly related to the cancer-related pathway; therefore, circ-ANAPC7 was speculated to be involved in AML tumorigenesis by sponging the miR-181 family [27]. In another study, Lv et al. searched out 17 “kernel” circRNAs that were possibly closely related to EMI and further constructed a ceRNA (competing endogenous RNA) network of circRNA/miRNA/mRNA by bioinformatics analysis to confirm that circRNAs had a bearing on EMI via ceRNA networks [28]. A new class of circRNAs, f-circRNAs have also been shown to promote cellular transformation as pro-oncogenic RNAs. Although they alone were not sufficient to trigger tumorigenesis, they could promote the development of leukemia and confer resistance to therapy when combined with other oncogenic stimuli [14]. The circRNAs validated in hematological malignancies are shown in Table 2.

Table 2.

The circRNAs with potential biomarker value and therapeutic targets in hematological malignancies

Disease	CircRNA ID	Gene symbol	Regulation	Biomarker		Role	Potential mechanism	Ref
				Diagnosis	Prognosis
AML	circ_0004277	WDR37	↓	+	-		miR-138-5p/SH3GL2?	[26]
	circ-ANAPC7	ANAPC7	↑	+	-		miR-181 family	[27]
	circ_0009910	MFN2	↑	+	+	oncogene	miR-20a-5p	[30]
	circ_0004520	VAV2	↑	+	-		miRNA?	[28]
	circ_0100181 (circPAN3)	PAN3	↑	-	-	chemoresistance	miR-153-5p,miR-183-5p/XIAP	[63]
	circ-HIPK2		↑	+	-	facilitate ATRA-inducing differentiation	miR-124a/CEBPA	[62]
	circ-VIM	VIM	↑	+	+			[64]
	circ_0000488 (circRNA-DLEU2)	DLEU2	↑	+	-	oncogene	miR-496/PRKACB	[29]
	circ_100290	SLC30A7	↑	+	-	oncogene	miR203/Rab10	[66]
	circMYLB2	MYLB2	↑	+	+	oncogene	PTBP1/FLT3	[67]
	circ_0075001	NPM1	↑	-	-	oncogene?	miR-181/TLR signaling pathway?	[68]
	circPR	PML-RARα	-	+	-	oncogene, chemoresistance		[14]
	circMF9	MLL/AF9	-	+	-	oncogene, chemoresistance		[14]
ALL	circ-PVT1	PVT1	↑	+	-	oncogene	miR-let7/c-MYC, miR-125/Bcl-2?	[86]
CML	circ_100053		↑	+	+	chemoresistance		[82]
	circ_0080145	TNS3	↑	+	-	oncogene	miR-129b	[83]
	circBA9.3	BCR-ABL	↑	+	+	oncogene, chemoresistance	upregulate protein c-ABL1 and BCR-ABL1	[81]
CLL	circ_000070 (circ-CBFB1)	CBFB	↑	+	+	oncogene	miR-607/FZD/Wntβ signaling pathway	[90]
	circ_0132266	MTO1	↓	-	-	tumor suppressor	miR-337-3p/PML	[91]
lymphoma	circ_101303 (circ_LAMP1)	LAMP1	↑	+	-	oncogene	miR-615-5p/DDR2	[93]
MM	circ_000190		↓	+	+	tumor suppressor	miR-767-5p/MAPK	[97]

“↑” Upregulation; “↓” Downregulation; “+” Yes; “-” No; “blank” Unknown; “?” Not experimentally validated.

In conclusion, circRNAs may be associated with the initiation, progression, and drug resistance of AML through multiple action modes, indicating that they are the potential therapeutic targets.

CircRNAs in chronic myeloid leukemia

Chronic myeloid leukemia (CML) is a clonal disease of hematopoietic stem cells, characterized by the aberrant chromosome translocation of t (9;22), resulting in fusion of the BCR-ABL1 gene and leading to sustained tyrosine kinase activity [79]. In the management of CML, the detection of BCR-ABL mRNA by qRT-PCR is the “gold standard” for the evaluation of therapeutic effect and prediction of recurrence [80].

Recently, Pan et al. discovered a fusion circRNA, circBA9.3, that can promote cell proliferation and endow leukemic cells with resistance to TKIs (trosine kinase inhibitors, including imatinib, nilotinib, and dasatinib) by upregulating c-ABL1 and BCR-ABL1 protein expression levels instead of the BCR-ABL1 oncogene [81]. CircBA9.3 is generated from the fusion gene BCR-ABL1 of CML patients, which is markedly overexpressed in TKI-resistant CML patients compared with TKI-susceptible CML patients. Another circRNA, hsa_circ_100053, was shown to be remarkably increased in CML, which was verified by qRT-PCR of 151 CML and 100 healthy samples [82]. The expression level of hsa_circ_100053 has been related to the relative risk of Sokal (P = 0.037) and the mutation status of BCR/ABL (P = 0.008). Moreover, the high expression of hsa_circ_100053 was not only an independent prognostic factor for CML progression (P = 0.037), indicating a shorter overall survival, but also conferred drug resistance to CML patients and cell lines. In another study, Liu et al. identified 361 differentially expressed circRNAs between CML and control samples with a fold change threshold of >2, including 201 upregulated and 160 downregulated circRNAs [83]. Among them, the upregulation of hsa_circ_0080145 in CML patients and cell lines was validated by qRT-PCR. Subsequently, the finding that hsa_circ_0080145 could promote CML proliferation by serving as a sponge of miR-29b and could play an oncogenic role in the progression of CML was confirmed through dual luciferase assay and a series of loss-of-function and rescue experiments.

These studies have demonstrated that circRNAs are differentially expressed in CML, are potential diagnostic or prognostic biomarkers and even participate in the initiation, progression and drug resistance of CML. Most importantly, the result that circBA9.3 promotes cell proliferation and endows CML with TKI resistance by regulating c-ABL1 and BCR-ABL1 protein expression rather than BCR-ABL1 oncogene expression challenges the use of BCR-ABL1 mRNA detection in traditional CML patient management.

CircRNAs in lymphoid malignancies

Lymphoid malignancies are a group of clonal tumors that arise at different stages of B/T-cell development. Here we mainly focused on acute lymphoblastic leukemia (ALL), chronic lymphoblastic leukemia (CLL), lymphoma, and multiple myeloma (MM).

Acute lymphoblastic leukemia (ALL) is a heterogeneous group of lymphoid malignancies caused by the monoclonal proliferation and expansion of B/T precursor lymphoid cells in the bone marrow, blood and other organs [84]. With treatment, pediatric ALL has a cure rate of approximately 90% [85]. However, treatment of adults is unsatisfactory and full of challenges. Elucidating the pathogenesis of ALL is the primary task of research on ALL. Hu et al. showed that circ-PVT1 is produced from the long noncoding RNA PVT1 locus on chromosome 8q23, is mainly located in the cytosol and may play an oncogenic role in ALL [86]. Circ-PVT1 was strikingly elevated in ALL cell lines and ALL patient samples. In-depth functional experiments have demonstrated that circ-PVT1 knockdown can inhibit cell proliferation and induce apoptosis, accompanied by decreased levels of the c-MYC and Bcl-2 proteins. Circ-PVT1 can act as a sponge for miRNA let-7 [87] and miRNA-125 [88], and in turn, these microRNAs target c-MYC and Bcl-2, respectively. Therefore, knockdown of circPVT1 can arrest cell proliferation and promote apoptosis possibly by elevating miRNA let-7 and miRNA-125 levels to suppress the expression of c-MYC and Bcl-2 in ALL cell lines. However, this hypothesis has not been confirmed by further experiments.

Chronic lymphoblastic leukemia (CLL), the type of leukemia with the highest incidence in western countries, is a hematopoietic malignancy with clonal proliferation and accumulation of mature CD5-positive B-cells [89]. Although patient survival has been improved by introducing fludarabine, an anti-20 monoclonal antibody, and Bruton’s tyrosine kinase, the pathogenesis of CLL still needs further study because of its recurrence. Recently, a study provided new insight into the pathogenesis of CLL from the perspective of ncRNA, indicating that circ-CBFB generated from chromosome 1 was dramatically upregulated in blood samples from CLL patients [90]. Circ-CBFB was suggested to distinguish CLL patients from healthy controls with an AUC of 0.80 (95% CI: 0.69-0.90). High circ-CBFB expression was an independent predictor of prognosis and was associated with shorter survival time. Furthermore, circ-CBFB was shown to be an oncogenic stimulus for CLL through the circ-CBFB/miR-607/FZD/Wnt/β-catenin signaling pathway axis by serving as a miRNA sponge. In another study, Wu et al. revealed that circ_0132266 was dramatically downregulated in CLL patients compared with healthy controls and exerted a tumor-suppressive role in CLL by the miR-337-3p/PML axis, playing a role as a miRNA sponge [91].

Lymphoma is a heterogeneous lymphoid malignancy. Its clinical manifestations, therapeutic responses and prognosis vary with the histological type, clinical factors and molecular characteristics [92]. Deng et al. discovered that circ-LAMP1 was dramatically overexpressed in T-cell lymphoblastic lymphoma tissues compared with normal infantile thymus and may play an oncogenic role in T-cell lymphoblastic lymphoma because knockdown of circ-LAMP1 inhibited cell viability and dramatically increased the rate of apoptosis, and vice versa [93]. In-depth mechanistic studies have revealed that circ-LAMP1 promotes DDR2 expression and regulates cell proliferation and apoptosis in T-cell lymphoblastic lymphoma by sponging miR-615-5p. Deng et al. also stated that it was challenging to detect RNA in frozen samples. Dahl et al. encountered the same problem, and they presented a more accurate and sensitive circRNA quantification method, which they named NanoString assay [94]. The authors employed RNA-seq to screen the circRNA expression profiles of four different mantle cell lymphoma (MCL) cell lines and one multiple myeloma (MM) cell line and they discovered 813, 816, 741, 279 and 619 unique circRNAs, respectively. Then, they verified the structures and abnormal expression of some of the circRNAs by Sanger sequencing analysis, qRT-PCR and northern blotting analysis and found that the circRNA expression profiles could distinguish different B-cell malignancies by Hierarchical Cluster Analysis [95], which indicated that circRNAs might act as potential diagnostic and differential diagnostic biomarkers. Importantly, this study showed that the NanoString technology is superior to RNA-seq and qRT-PCR in sensitivity, specificity, and quantitative accuracy and is a better method for detecting circRNAs because it dose not require RT (reverse transcription) and PCR amplification, which is a major source of assay artifacts. NanoString technology is also reliable, even in low-quality RNA samples such as formalin-fixed and paraffin-embedded tissues.

Multiple myeloma (MM) is a highly heterogeneous disease characterized by the accumulation of clonal plasma in the bone marrow [96]. In the latest study, circ_0000190 was shown to be downregulated not only in MM cell lines but also in the bone marrow and plasm of MM patients compared with healthy controls [97]. International staging system stage I and Durie-Salmon stage I patients had higher circ_0000190 levels than stage II and III patients, and higher expression of circ_0000190 was positively correlated with a longer progression-free period and overall survival time, indicating that higher expression levels of circ_0000190 predicted excellent clinical outcome. In-depth research demonstrated that circ_0000190 inhibited miR-767-5p and then upregulated MAPK levels, ultimately arresting the cell cycle and inhibiting MM progression. Additionally, circ_0000190 was shown to inhibit tumor progression in an MM mouse model.

These studies revealed that circRNAs are broadly and specifically expressed in various subtypes of lymphoid malignancies as they are in other diseases, indicating that circRNAs can also be diagnostic biomarkers for lymphoid malignancies. Additionally, some of circRNAs have been associated with prognosis, some with disease occurrence and progression, and some may have antitumor effects.

Conclusions

In this review, the biogenesis, categories, functions, and characteristics of circRNAs were summarized, with emphasis on the roles of circRNAs in the hematological system. CircRNAs are widely and specifically expressed in hematopoietic stem cells, mature blood cells and various subtypes of hematological malignancies and can be detected in blood and bone marrow, making them promising potential diagnostic biomarkers. The expression of some circRNAs is associated with the current cancer risk stratification system, progression-free survival time, overall survival time, extramedullary infiltration and drug resistance, suggesting that they may be potential prognostic biomarkers and monitoring indicators. Some circRNAs are dynamically expressed during hematopoietic differentiation, some can protect mesenchymal stem cells from total body irradiation and some can maintain the self-renewal and differentiation ability of stem cells by sponging miRNAs or acting with other molecules, which suggests that circRNAs may be potential hematopoietic cell biomarkers and hematopoietic regulators. Additionally, some circRNAs have been confirmed to act with miRNAs or RBPs to regulate the expression of downstream genes and eventually participate in the occurrence and development of hematological malignancies as well as in drug resistance by influencing the cell cycle, signaling pathways and posttranscriptional regulation, which indicates that circRNAs can be potential therapeutic targets.

Although we have gradually unveiled the mystery of circRNAs, our understanding of circRNAs may be just the tip of the iceberg and still far from circRNA clinical application. Here, we also include some suggestions for future research on circRNAs in hematological malignancies. (1) It is advisable to provide the names of circRNAs cataloged in the circBase database (http://www.circbase.org/) [98], which is the most complete database for circRNAs annotation at present. The naming of circRNAs is currently very confusing, which makes it difficult for researchers to communicate with each other. (2) Although many unique circRNAs in hematologic malignancies have been identified by RNA-seq or circRNA microarrays, only a small portion of these circRNAs have been verified in a large number of samples, and multiple circRNAs remain to be investigated. (3) Many circRNAs have been identified to be differentially expressed in different stages of hematopoietic cells, but how they regulate differentiation, maturation and functions are still not clear, so the mechanism of circRNAs as hematopoietic regulators needs to be addressed in future studies. (4) Increasing circRNAs have been confirmed as diagnostic markers, but their sensitivity and specificity have yet to be verified. Combining several circRNAs or combining circRNAs with traditional diagnostic biomarkers may improve the diagnostic value. (5) Most studies on circRNAs in hematological malignancies have focused on the role of circRNAs as miRNA sponges. On the one hand, the exact mechanisms remain to be further studied, and on the other hand, other action modes of circRNAs are also worth exploring. (6) Some circRNAs have been confirmed to be sensitive to RNase, and it can be challenging to extract circRNAs from some low-quality specimens. More sensitive and accurate circRNA detection methods need to be developed. In general, research on circRNAs is still in its infancy, but with the development of the next-generation high-throughput sequencing technology, bioinformatic analysis and the use of online databases, we can expect more circRNAs and their action modes to be elucidated and applied to prevent and treat hematological malignancies.

Disclosure of conflict of interest

None.