RNA-binding proteins (RBPs) are involved in almost all layers of post-transcriptional regulation (RNA metabolism). RBPs establish highly dynamic interactions with both coding and noncoding RNAsm as well as multiple proteins to achieve these complicated metabolic processes. Although RBPs bind to various types of RNAs including messenger RNA (mRNA), ribosomal RNA (rRNA), small nuclear RNA (snRNA), small nucleolar RNA (snoRNA), and transfer RNA (tRNA), more than one-half of RBPs bind to mRNA to regulate its fate ^[1][14]. In this section, we review the regulatory mechanisms of mRNA and noncoding RNA metabolism by several RBPs in the context of normal and malignant cells.

Alternative splicing (AS) plays an important role in gene regulation, through which pre-mRNA transcripts are processed to produce multiple mRNA variants with different stability and protein-coding potential. AS perturbations are frequently observed across cancers. RBP is one of the molecular determinants of AS, and the disturbance of RBP network activity has a causal relationship with cancer development ^[2][3][15,16]. Functionally aberrant RBPs induce splice isomer conversions, which are widely involved in the regulation of cancer phenotypes including proliferation, apoptosis, cell-cycle progression, invasion and metastasis, angiogenesis, abnormal energy metabolism, and immune escape ^[4][17]. The main types of abnormal splicing in tumors are constituent splicing, exon skipping or inclusion, substitution of the 5′ or 3′ splice site, intron retention, and mutually exclusive exon ^[5][18]. Classical splicing regulators include serine–arginine (SR) proteins (such as SRSF2) and heterogeneous nuclear ribonucleoproteins (hnRNPs), which bind to exon or intron regulatory elements to promote or prevent the recognition of the 5′ splicing site by the spliceosomal U1 small nuclear ribonucleoprotein (snRNP). These proteins can also promote or block SF1, U2AF2, U2AF1, or U2 snRNP, thereby influencing the selection of splice sites and thus altering splicing decisions ^[4][17]. The aberrant expression of SR and hnRNP proteins in cancers suggests that dysregulation of these two types of splicing factors plays an important role in tumor progression. In addition, RNA splicing is frequently dysregulated in a variety of cancers, and hotspot mutations affecting key splicing factors, SF3B1, SRSF2, and U2AF1 are commonly enriched across cancers ^[6][7][19,20], strongly suggesting that aberrant RNA splicing is a new class of hallmark that contributes to the initiation and/or maintenance of cancers. These mutations in genes encoding splicing factors are commonly identified in a variety of hematologic malignancies ^[6][7][8][19,20,21], as well as in solid tumors such as breast cancers, lung cancers, and pancreatic cancers ^{[9][10][11][12][13]}[22,23,24,25,26]. The pathogenic roles of recurrent mutations affecting these splicing factors ^{[14][15][16][17]}[27,28,29,30] and the therapeutic strategies against cancers bearing these mutations ^[14][18][19][12,27,31] have been extensively studied and reviewed elsewhere ^[20][21][22][32,33,34].

Polyadenylation is a key process for the generation of mature RNA transcripts. Selective polyadenylation occurs within the 3′ UTR of mRNA and produces 3′ UTR of varying length by 3′-terminal cleavage and polyadenylation (CPA). Each transcript of mature mRNA contains a polyadenylate tail that determines its stability. RBPs involved in polyadenylation include U1 small nuclear ribonucleic protein (U1 snRNP), cleavage and polyadenylation-specific factor 1 (CPSF1), embryonic lethal abnormal vision (ELAV) L1/human antigen R (HuR), and poly(A) RNA-binding protein (examples include the cytoplasmic polyadenylate element-binding protein family, CPEB1–4), and the ZFP36 ring finger protein (ZFP36/TTP) ^[23][35]. CPEB family proteins recruit the translational inhibitory factors or cytoplasmic polyadenylation factors and regulate the length of the poly(A) tail. The C-terminal region of CPEB family proteins contains two RRM and two zinc finger-like motifs, as well as a variable N-terminal region ^[24][36]. For example, TP53 mRNA contains CPE in its 3′-UTR, which promotes polyadenylation. Burns et al. showed that CPEB1 enhanced TP53 mRNA polyadenylation and translation with the cytoplasmic poly(A) polymerase GLD-4 ^[25][37]. The researcheuthors demonstrated that TP53 mRNA has a short poly(A) tail and a reduced translational efficiency, leading to a decrease in p53 protein expression ^[26][38].

RNA stability is associated with its nucleotide sequence, modification, 5′ m⁷G cap, and 3′ poly (A) tail ^[27][39]. These determinants for RNA stability regulate the mRNA decay and translation initiation ^[28][40]. More specifically, mRNA degradation is mainly regulated by the following two mechanisms: (1) one mechanism starts with deadenylation of the 3′ poly (A) tail, which is followed by 5′ cap removal and 5′-to-3′ decay; (2) the other mechanism begins after hydrolysis of the 3′ poly (A) tail and 3′-to-5′ decay ^[28][40].

Several RBPs such as the mRNA decapping enzyme scavenger (DCPS), CUGBP Elav-like family member 2 (CELF2), insulin-like growth factor 2 mRNA protein (IGF2BP) family, HuR, QKI-5, RBMS3, and TARBP2 play important roles in cancer biology ^[29][6]. For example, Yamauchi et al. performed a genome-wide CRISPR/Cas9 screen in murine AML models with the oncogenic fusions CALM/AF10 and MLL/AF9 and identified DCPS as a promising target for AML ^[30][31][41,42]. The 5′ end of eukaryotic mRNA is characterized by a distinctive “cap”, which consists of an N7 methylated guanine (m⁷GpppN). The 5′ cap is important for promoting splicing of the first intron, exporting mRNA to the cytoplasm, and allowing translation of the mRNA. Removal of the cap (“decapping”) results in silencing of mRNA expression. The decapping enzyme DCPS is characterized as a pyrophosphatase that hydrolyzes the 5′ m⁷Gppp and m⁷Gpp cap structure generated following 3′-to-5′ and 5′-to-3′ decay. Messenger RNAs containing an AU-rich element (ARE) in the 3′ UTR are rapidly degraded in the cytoplasm. ARE-mediated decay is initiated by deadenylation, which is followed by the 3′-to-5′ decay through a complex of exonucleases (termed exosome). In this process, DCPS hydrolyzes the remaining cap ^[32][43]. On the other hand, the 5′-to-3′ decay pathway is initiated by cleavage of the 5′ cap structure to release the m⁷Gpp and monophosphorylated RNA, followed by the DCPS-mediated hydrolysis of m⁷Gpp to release m⁷Gp ^[33][44]. Therefore, DCPS plays an essential role in the final step of removing the residual cap in both directions of RNA decay. Intriguingly, DCPS was shown to be dispensable for normal hematopoiesis, which was supported by the observation that clinical parameters including blood cell counts in persons with germline biallelic loss-of-function mutations in DCPS were not significantly affected. In summary, DCPS is potentially an AML-selective vulnerability for which development of a targeted therapy is expected.

Another layer of regulation on the stability, as well as the translation, of mRNA is achieved by RBP-mediated control on intracellular localization ^[34][45]. Subcellular localization of mRNAs involves several steps and requires the coordinated involvement of multiple protein factors. Cis-motifs and postcode elements in mRNA 3′ UTR are the important factors for RBPs to coordinately involve this regulatory process ^[35][46]. One of the best examples is IGF2BP1 (also known as IMP1/ZBP1), which is a member of the conserved VICKZ family of RBP. IGF2BP1 controls cell adherence and polarity in breast cancer by physically binding to a subset of mRNAs that encode important mediators such as E-cadherin, β-actin, α-actinin, and ARP-16 for these cellular properties ^[36][47].

The translation process consists of the following three steps: initiation, extension, and termination. Controls on these processes are crucial for cancer development and progression both globally and in specific mRNAs which promote cancer biology, such as cell survival, transformation, metastasis, and stemness. Multiple factors including 43S ribosome initiation complex, cap-dependent mRNA translation, and cap-independent mRNA translation are involved in carcinogenesis. Dysregulations in these translational processes in malignant disorders are beyond the scope of this review and are excellently summarized elsewhere ^[37][38][39][48,49,50].

No-coding RNAs (ncRNAs) are commonly expressed RNAs in human cells that lack protein-coding ability. In the research results of FANTOM and ENCODE, two large genome projects, 80% of the human genome has transcriptional activity, while only 2% of the human genome codes for proteins. The noncoding regions of the human genome (98%) are primarily responsible for regulating gene expression ^[40][41][51,52]. RBPs also bind to noncoding RNAs such as microRNAs (miRNA), transfer RNAs (tRNA), siRNA, telomerase RNA, small nucleolar RNA (snoRNA), and splicing small nucleolar RNA (snRNA), which regulates multiple molecular processes including RNA splicing and modification, protein localization, and chromosomal remodeling ^[42][53]. The interactions between noncoding RNAs and RBPs are increasingly recognized as one of the basic mechanisms of gene regulation and plays a crucial role in cancer ^[43][54]. RBPs are key regulators of miRNA biogenesis and maturation. They promote or inhibit miRNA processing mainly through their effects on typical proteins such as Drosha and Dicer. Altered functions in such RBPs result in the impairment of miRNA processing, which in turn affects expression of cancer-associated genes ^[44][55].

The RBP LIN28 is one of the four factors sufficient to reprogram human somatic cells into induced pluripotent stem cells, which upregulates or inhibits the maturation of different members of the let-7 microRNA family in many cancer cells ^[45][56]. LIN28A binds to the terminal ring of precursor let-7 (pre-let-7) and enforces terminal uridine transferase (TUTase) ZCCHC11, which can acidify pre-let-7 polyuridine, thereby blocking miRNA biogenesis and tumor suppressor function. For LIN28B, the exact mechanism that causes let-7 inhibition remains controversial. Functionally, the reduction in let-7 miRNA leads to the overexpression of its oncogenic targets such as MYC, RAS, HMGA2, and BLIMP1 ^[46][57]. In addition to let-7, other miRNAs (miR-9, -107, -143, -200C, -370, and -638) containing the same tetranucleotide sequence motif (GGAG) as pre-let-7 undergo the same process. Most of these miRNAs have been identified as tumor suppressors, suggesting that LIN28 may promote cancer metastasis by inhibiting multiple metastasis-associated miRNAs ^[29][6].

CircRNAs are covalently closed RNA molecules, usually derived from precursor mRNA (pre-mRNA) through reverse splicing events, where the downstream 5′ splicing donor is reversely spliced to the upstream 3′ splicing receptor ^[47][58]. Recent evidence suggests that abnormal circRNA expression exists in almost all cancer types and plays an indispensable role in cancer pathogenesis as oncogenes or tumor suppressors ^[48][59]. CircRNA can act as a protein scaffold or antagonist that interacts with RBPs. For example, circACC1 acts as a protein scaffold to enhance the enzyme activity of the AMP-activated protein kinase (AMPK) holoenzyme by directly binding AMPK β and γ subunits. ^[49][60]. CircRHOBTB3 also inhibits metastasis in colorectal cancer by interacting with HuR, which in turn degrades HuR to reduce the expression level of the downstream target PTBP1 ^[50][61].