RNA-Binding Proteins in Cancer: Functional and Therapeutic Perspectives

Simple Summary RNA-binding proteins (RBPs) play central roles in regulating posttranscriptional expression of genes. Many of them are known to be deregulated in a wide variety of cancers. Dysregulated RBPs influence the expression levels of target RNAs related to cancer phenotypes, such as proliferation, apoptosis, angiogenesis, senescence, and EMT/invasion/metastasis. Thus, understanding the molecular functions of RBPs and their roles in cancer-related phenotypes can lead to improved therapeutic strategies. Abstract RNA-binding proteins (RBPs) crucially regulate gene expression through post-transcriptional regulation, such as by modulating microRNA (miRNA) processing and the alternative splicing, alternative polyadenylation, subcellular localization, stability, and translation of RNAs. More than 1500 RBPs have been identified to date, and many of them are known to be deregulated in cancer. Alterations in the expression and localization of RBPs can influence the expression levels of oncogenes, tumor-suppressor genes, and genome stability-related genes. RBP-mediated gene regulation can lead to diverse cancer-related cellular phenotypes, such as proliferation, apoptosis, angiogenesis, senescence, and epithelial-mesenchymal transition (EMT)/invasion/metastasis. This regulation can also be associated with cancer prognosis. Thus, RBPs can be potential targets for the development of therapeutics for the cancer treatment. In this review, we describe the molecular functions of RBPs, their roles in cancer-related cellular phenotypes, and various approaches that may be used to target RBPs for cancer treatment.


Introduction
RNA-binding proteins (RBPs) play central roles in regulating gene expression at the post-transcriptional level. Recent advances in screening techniques have led to the identification of more than 1500 RBPs, which account for about 7.5% of all protein-coding genes in the human genome. RBPs can interact with proteins and various classes of RNAs (mRNAs, ncRNAs, tRNAs, snRNAs, snoRNAs, and others) to form ribonucleoprotein (RNP) complexes [1]. RBPs recruit various factors and enzymes, and form different complexes in diverse combinations to modulate the fates and/or functions of target RNAs [2,3]. Conventional RBPs have various impacts on RNA metabolism, such as by regulating microRNA (miRNA) processing and the alternative splicing, alternative polyadenylation, subcellular localization, stability, and translation of RNAs. More than half of the RBPs harbor one or more RNA-binding domains (RBDs), such as the RNA-recognition motif (RRM), K-homology (KH) target mRNAs, which include BMP6, PDCD4, and TIMP4, and induces EMT [21]. KHSRP functions as either a tumorigenic or tumor-suppressive protein by regulating miRNA biogenesis [16][17][18][19][20][21].
Heteronuclear ribonucleoprotein A1 (hnRNP A1) harbors two RRMs and has been implicated in miRNA processing [22]. hnRNP A1 acts as an auxiliary factor for the processing of miRNA precursors, such as pri-miR-18a. It recognizes two UAG motifs of pri-miR-18a (one in the TL and one in the proximal stem region), forms a 1:1 complex with this miRNA, and increases the efficiency of Drosha cleavage by causing relaxation at the stem of the pri-miRNA [22]. In lung cancer, cervical cancer, prostate cancer, gastric cancer, and mesothelioma miR-18a downregulates target mRNAs, such as IRF2, PTEN, WNK2, SOX6, STK4, and PIAS3, to induce cancer progression and development. In other cancers, in contrast, miR-18a negatively regulates CDC42 and SREBP1 to suppress cancer cell proliferation, invasion, and metastasis [23]. Conversely, hnRNP A1 negatively regulates the processing of let-7a by binding to the conserved terminal loop of pri-let-7a-1 to block the interaction of KHSRP, which increases let-7a biogenesis [24]. hnRNP A1 and KHSRP play antagonistic roles in the processing of let-7a precursors [24].
SMAD, which is a signal transducer of TGF-β and BMP, positively regulates the maturation of miR-21 and miR-199a, which contain the RNA Smad-binding element (R-SBE) [25]. SMAD directly binds to this consensus sequence and recruits RNA helicase p68 (also called DDX5), which is a component of the Drosha microprocessing complex. The interactions of SMAD with pri-miR-21 and pri-miR-199a increase the expression levels of miR-21 and miR-199a by mediating their efficient cleavage by the Drosha complex [25]. Increased miR-21 negatively regulates the tumor suppressor, PDCD4, by binding to its 3 -UTR, and this promotes tumor invasion, intravasation, and metastasis [26].

Alternative Splicing
Alternative splicing is a critical mechanism through which pre-mRNA transcripts are processed to generate multiple mRNA variants with different stabilities and protein-coding potentials. Different splice variants of mRNAs contribute to protein diversity in cancer [27]. Alternative splicing events are controlled by the core spliceosome, which is a large molecular complex composed of small nuclear proteins (snRNPs), polypeptides, and RBPs. Alternative splicing is regulated by cis-acting elements (i.e., enhancers or silencers located in exons or introns) and trans-acting splicing factors, which promote or suppress exon splicing, respectively. RBPs, such as SRSF, hnRNPs, PTB, ESRP, and QKI, bind to cis-regulatory elements and form trans-acting splicing factors to regulate (negatively or positively) the splicing reaction [28]. Compared to normal cells, aberrant patterns of alternative splicing are seen in cancer cells. These alterations can lead to modulation of oncogenes or tumor-suppressive genes, and thereby help drive cancer development and progression [28,29].
Serine/Arginine-rich (SR) family proteins, which represent the best-characterized splicing regulators, have one or two RRMs and a C-terminal arginine (R) and serine (S) amino acid sequence (RS domain); these functional regions participate in RNA recognition and protein-protein interactions to recruit the spliceosome [30]. SR proteins contribute to multiple steps in splicing events, and they are involved in both constitutive and alternative splicing [31]. SRSF1 (also known as SF2/ASF) is known to alter the splicing of the protooncogene Ron, the tumor suppressor BIN1, the kinases MNK2 and S6K1, and the proapoptotic regulators Bcl-x, Mcl-1, CASP2, CASP9, and ICAD [32][33][34][35]. SRSF1 is modulated its alternative splicing activity through the interaction with the lncRNA, MALAT1. MALAT1 influences the distribution of SRSF1 at nucleus and nuclear speckles, thereby modulating the levels of phosphorylated SRSF1 [36]. In hepatocellular carcinoma (HCC), MALAT1-upregulated SRSF1 enhances the isoforms of the antiapoptotic gene BIM and the oncogenes S6K1 and TEAD1, leading to cancer cell proliferation, survival, and tumorigenesis [37]. Moreover, in glioblastoma, SRSF1 physically associates with circRNA SMARCAS5 to regulate the alternative splicing of VEGF, thereby affecting angiogenesis [38]. SRSF3 (also known as SRp20) is the smallest member of the SR protein family; it is involved in the alternative splicing of FoxM1 to generate the FoxM1a, b, and c isoforms. Conversely, SRSF3 overexpression increases the FoxM1b and 1c isoforms and results in upregulation of the FoxM1 targets, PLK1 and CDC25B [39]. In colon cancer cells, SRSF3 regulates the alternative splicing of HIPK2. Research revealed that SRSF3 depletion promotes skipping of 81 5 nucleotides (27 amino acids) from exon 8 of HIPK2. This generates HIPK2 ∆e8, which has no binding site for SIAH1 ubiquitin ligases, and is thus protected from proteasomal digestion [40]. Transcriptomic profiling of SRSF3-mediated altered splicing showed that SRSF3 binds to the CA(G/C/A)CC(C/A) sequence of the exonic splicing enhancer that is found in target genes, such as those encoding ETV1 and NDE1 [41]. In contrast to the above-described situation in colon cancer cells, SRSF3 is reduced or mislocalized in hepatocellular carcinoma. SRSF3 affects the alternative splicing of genes related to metabolism (HNF1a, ERN1, HMGCS1, DHCR7, and SCAP) or EMT (LIFR, Epb4.1l5, Myo1b, CTNND1, GIT2, and SLK) [42,43]. In colorectal cancer (CRC), the SRSP (splicing regulatory small protein) encoded by the lncRNA Loc90024 interacts with SRSF3 and promotes its binding to SP4. SRSF3-SRSP mediates the alternative splicing of SP4 and generates cancerous SP4 isoforms, leading to tumorigenesis and metastasis [44]. SRSF6 (formerly SRp55) is upregulated in a subset of tumors and is critical for tumor growth, initiation, and maintenance [45]. SRSF6 has been shown to regulate the splicing of one oncogene (INSR) and tumor suppressor genes (DLG1 and MKNK2), modulating their oncogenic and tumor-suppressive isoforms [45]. SPF45, also known as RBM17, is a component of the spliceosome and associates with the lncRNA Saf. In erythroleukemia and cervical carcinoma cells, SPF45-Saf binds to the Fas pre-mRNA and modulates alternative splicing of Fas; this increases soluble Fas, which eventually interacts with Fas ligand to protect cells from apoptosis [46]. hnRNP A/B family members, such as hnRNP A1 and hnRNP A2/B1, regulate alternative splicing by acting as antagonists of the SR proteins [47]. hnRNP A1 has two RBP domains and one arginine-glycine-rich (RGG) domain, and recognizes the UAGGG(A/U) sequence. hnRNP A2/B1 has a structure similar to that of hnRNP A1, and binds to (UUAGGG) N [47]. hnRNP A1 and hnRNP A2/B1 regulate the alternative splicing of the glycolytic enzyme, PKM2, and increase the PKM2/PKM1 ratio in cancer cells [48,49]. hnRNP A1 is overexpressed in hepatocellular carcinoma (HCC); this increases the level of CD44v6, which promotes invasiveness in HCC and is associated with a poor prognosis for HCC patients [50]. In glioblastoma, hnRNP A2/B1 functions as a proto-oncogene. This hnRNP affects alternative splicing and increases the expression of oncogenic isoforms of genes related to tumor suppressors (BIN1 and WWOX), anti-apoptotic proteins (CFLAR and CASP9), and a proto-oncogene (MST1R) [51]. hnRNP M controls isoform switching from CD44 variable (CD44v) to CD44 standard (CD44s) [52]. Upregulated hnRNP H generates a constitutively active isoform of the tyrosine kinase receptor, MST1R [53].
Polypyrimidine tract-binding protein (PTB, also known as hnRNP I) plays a role as a splicing repressor and preferentially interacts with pyrimidine-rich sequences and modulates cancer-relevant alternative splicing events, leading to exon skipping or inclusion [54]. In glioblastoma, PTB binds to the ISS element in the intron upstream of the FGFR-1α exon; it mediates skipping of the α exon to produce the FGFR-1β isoform [55]. The switch from FGFR-1α to FGFR-1β contributes to the malignant progression of astrocytic tumors [56]. USP5 is subjected to PTB-mediated alternative splicing. High-level expression of PTB generates USP5 isoform 2 and inhibits the production of USP5 isoform 1 [57]. In addition, PTB and hnRNP A1/A2 cooperate to regulate an alternative splicing event of PKM2; they bind to the ISS element of PKM and thereby increase the level of PKM2 [48,49].
RBM3, which is known as a proto-oncogene, is a member of the glycine-rich RBP family [61]. In HCC, upregulated RBM3 induces biogenesis of SCD-circRNA, leading to cell proliferation [62].
Src-associated in mitosis of 68 kDa (SAM68) is a member of the signal transduction and activation of RNA (STAR) protein family [63]. SAM68 is primarily localized in the nucleus and regulates alternative splicing. SAM68 is a target of ERK and its serine-threonine is phosphorylated in response to RAS/ERK signaling. Phosphorylated SAM68 induces the inclusion of the v5 exon of CD44 in a protein modification-dependent fashion [64]. In prostate cancer, the phosphorylation of SAM68 by signal transduction pathways increases its binding affinity to the proximal region of Cyclin D1 intron 4. SAM68 enhances alternative splicing of Cyclin D1 and inhibits the recruitment of U1-70K to its target genes, generating Cyclin D1b [65]. Meanwhile, SAM68 phosphorylated through Src-like kinases alters the splicing of Bcl-x, which encodes two splicing variants, anti-apoptotic Bcl-xL and pro-apoptotic Bcl-xS [66]. Importantly, SAM68 is involved in cancer-relevant splicing in a protein modification-dependent manner [64][65][66].
Quaking (QKI), a member of the STAR family, is involved in the biogenesis of circRNAs. During EMT, QKI binds to the intron of pre-mRNAs and promotes the formation of circRNAs, such as SMARCAS5, POLE2, OXNAD1, SHPRH, SMAD2, ATXN2, DOCK1, and GNB1. These EMT-associated circRNAs may regulate the migration, invasion, and metastasis of cancer cells [67].

Alternative Polyadenylation
Alternative polyadenylation is a critical process to generate mature RNA transcripts. Alternative polyadenylation occurs within the 3 -UTR of mRNAs and generates different length 3 -UTRs by 3 -end cleavage and polyadenylation (CPA). CPA is carried out by multimeric protein complexes, such as the cleavage and polyadenylation specificity factors (CPSFs), the cleavage stimulation factors (CSTF), and the mammalian cleavage factor complexes and II (CFIm and CFIIm). The 3 -UTR of an mRNA is critical for its maturation, stability, nucleo-cytoplasmic localization, and translation [68]. RBPs can also regulate the CPA of target mRNAs by either recruiting or competing with the polyadenylation machinery proteins [68].
The cytoplasmic polyadenylation element binding proteins (CPEBs), CPEB1-4, regulate the poly(A) tail length of mRNAs with a cytoplasmic polyadenylation element (CPE) by recruiting the translational repression or cytoplasmic polyadenylation factors. The CPEB-family proteins have C-terminal regions containing two RRMs and two zinc-finger-like motifs, and a variable N-terminal region [69]. CPEB1 can alter the gene expression profile by shortening or lengthening the 3 -UTRs of target mRNAs. The loss of CPEB1 results in poly(A) tail lengthening and increased translation of the mRNA for MMP9 in breast cancer cells [70]. CPEB1 and CPEB2 mediate the polyadenylation of the HIF-1αmRNA by binding to its 3 -UTR. Increased HIF-1αpolyadenylation affects its protein levels [71]. CPEB2 and CPEB3 are less well studied in terms of their roles in cancer, and future work is needed. In pancreatic ductal adenocarcinoma (PDA), overexpression of CPEB4 promotes poly(A) elongation and translational activation for the mRNA encoding tPA, which is a key regulator of PDA malignancy. tPA with a longer 3 -UTR has a higher translation efficiency, leading to high-level expression in PDA [72]. CPEB4 is highly expressed in the early stage of melanoma progression; it can control the polyadenylation of the melanoma drivers, MITF and RAB7A, through binding to their 3 -UTRs. Depletion of CPEB4 decreases the protein levels of MIFT and RAB7A through the shortening of their poly (A) tails [73]. CPEB1 and CPEB4 regulate the length and polyadenylation of the VEGF 3 -UTR in the nucleus and cytoplasm, respectively. The VEGF 3 -UTR harbors several CPE elements and alternative polyadenylation sites (PASs). In the nucleus, CPEB1 promotes alternative 3 -UTR processing of VEGF and increases the level of VEGF mRNAs with shorter 3 -UTRs that lack translation inhibitory elements, such as AU-rich elements (AREs) and miRNA-binding sites, and thus are less prone to degradation. CPEB1 also upregulates CPEB4 expression through alternative processing of the pre-mRNA. Activated CPEB4 contributes to positively regulating cytoplasmic VEGF expression through polyadenylation of its mRNAs [74].

RNA Localization
The subcellular localization of an mRNA or lncRNA is an important factor in the regulation of its stability and translation, and cancer-related RBPs often bind to RNAs to coordinate where they are localized and translated [75,76].
Among the CPEB family members, CPEB1 regulates the localization of ZO-1 mRNA, which encodes a critical tight junction component. The apical co-localization of CPEB1 with the ZO-1 mRNA is consistent with its role in RNA localization. When CPEB1 is depleted, the ZO-1 mRNA is randomly distributed and central cavity formation is disrupted, resulting in loss of epithelial cell polarity [77]. This impairment of cell polarity is linked to the metastatic potential, and CPEB1 depletion leads to alterations in EMT-related genes [70]. IGF2BP1 (also called IMP1/ZBP1) controls the subcellular localizations of the mRNAs encoding β-actin, E-cadherin, α-actinin, and Arp-16 (a component of the Arp 2/3 complex) [78,79]. IGF2BP1 binds to the zip-code of the β-actin mRNA via its COOH-terminal KH domains and localizes the β-actin mRNA to the cell periphery [78]. In breast carcinoma cells, IGF2BP1 promotes the localization of E-cadherin, α-actinin, and Arp-16 at cell-cell contacts. IGF2BP1 also regulates the localizations of mRNAs related to cell motility and focal adhesions to impact cancer invasion and metastasis [78,79]. IGF2BP2 (also called IMP2) binds to oxidative phosphorylation (OXPHOS)-related mRNAs, such as NDUFS3 and COX7b, and facilitates their localization to mitochondrial polysomes. The delivery of these mRNAs to mitochondria induces their translation and insertion into mitochondrial membranes, contributing to the assembly of complexes I and IV [80].
HuR and GRSF1 regulate the shuttling of the lncRNA RMRP from the nucleus to mitochondria. HuR, which is predominantly localized in the nucleus, promotes the nuclear export of lncRNA RMRP in a CRM1-dependent manner. After lncRNA RMRP is transported into mitochondria, GRSF1 is involved in the retention of lncRNA RMRP [81].
Another RBP, hnRNP K, can directly associate with the lncRNA MALAT1 and regulates its nuclear retention. hnRNP K binds to a sequence derived from Alu repeats and short interspersed nuclear element (SINEB1), and increases the distribution of MALAT1 at the nucleus [82,83].

RNA Stability
The stability of an RNA is determined by its 5 -terminal 7-methylguanosine (m 7 G) cap and 3 poly(A) tail. These two stability determinants protect mRNAs from decay and promote translational initiation [84]. mRNA decay is mediated through several pathways, including deadenylation of the poly(A) tail, removal of the 5 -terminal m 7 G cap (decapping), 5 →3 exonucleolytic decay, and exosome-mediated 3 →5 degradation. mRNAs targeted for decay are shuttled to cytoplasmic foci, such as processing bodies (P-bodies) or stress granules [84]. RBPs, such as AUF1, HuR, TTP, IGF2BP family proteins, and Wig1, can stabilize or destabilize specific target mRNAs and lncRNAs in cancer [85].
AU-rich element RNA-binding protein 1 (AUF1; also known as heterogeneous nuclear ribonucleoprotein D, or hnRNP D) comprises four proteins: p37 AUF1 , p40 AUF1 , p42 AUF1 , and p45 AUF1 . AUF1 binds to the AREs of mRNAs and primarily promotes mRNA decay via ARE-mediated decay (AMD) [86]. In NSCLC cells, p45 AUF1 is increased by the chemotherapeutic agent, prostaglandin A 2 , and this induction increases Cyclin D1 mRNA turnover by binding the Cyclin D1 3 -UTR [87]. AUF1 also negatively affects the stability of p21 and Cyclin D1 by directly binding to their 3 -UTRs. Knockdown of AUF1 was found to increase the stabilities and expression levels of p21 and Cyclin D1 [88]. AUF1 promotes the destabilization and decay of the p16 mRNA by binding to a stem-loop structure localized within its 3 -UTR [89]. AUF1 and HuR share the Cyclin D1, p21, and p16 mRNAs as common targets. They co-occupy the p21 and Cyclin D1 mRNAs at non-overlapping sites or competitively bind these mRNAs at common sites [88]. For decay of the p16 mRNA, meanwhile, AUF1 can function as a co-factor of HuR; the co-factors interdependently interact with a secondary structure within the p16 3 -UTR to destabilize the p16 mRNA [89]. Other cell cycle-regulatory proteins, such as p27 and pRB, are targets of AUF1, which negatively regulates their mRNA stabilities [90]. AUF1 also mediates the destabilization of apoptosis regulators (Bax, Bcl-2, Gadd45a, and CASP2), metastasis regulators (MMP9 and FGF9), inflammatory factors (GM-CSF, IL-6, and NOS), and DNA repair, replication, and replication regulators (Fos, TYMS, and JunD) [91]. Moreover, AUF1 interacts with and destabilizes the cancer-associated lncRNA NEAT1 [92].
HuR, a member of the Hu family of RNA-binding proteins, governs the stability of target mRNAs by binding to AREs in their 3 -UTRs. HuR-mediated mRNA stabilization depends on the subcellular localization of HuR, which is translocated into the cytoplasm in response to different stresses and stimuli, including DNA damage, oxidative stress, and chemical compound treatment. The increased cytoplasmic HuR binds to the 3 -UTRs of mRNAs related to oncogenes, cell cycle regulators, inflammation, and apoptosis. For example, HuR stabilizes the mRNAs for Cyclins (A, B1, D1, and E), IL-8, MMP9, HIF-1α, VEGF, SIRT1, and Snail, thereby elevating their proteins levels. In contrast, mRNA levels of c-Myc, Wnt5a, and p27 are downregulated by HuR [93]. Moreover, HuR cooperates with RBPs or miRNAs to regulate mRNA stability. In CRC, HuR interferes with the binding of miR-16 to the COX-2 mRNA, thereby increasing COX-2 expression [94]. HuR is involved in stabilizing various lncRNAs, such as NEAT1, lncRNA-HGBC, and OIP5-AS1 [95][96][97]. In ovarian cancer, HuR stabilizes NEAT1 by inhibiting its binding with miR-124-3p, and upregulation of NEAT1 in ovarian cancer induces cancer cell proliferation and invasion [95]. In gallbladder cancer, lncRNA-HGBC stability is increased by HuR. Stabilized lncRNA-HGCB functions as an miRNA sponge to prevent miR-502-3p from binding to the target gene SET. This leads to activation of the SET-AKT pathway, which supports cancer cell proliferation and metastasis [96]. HuR also interacts with and stabilizes the lncRNA OIP5-AS1. This interaction prevents the binding of HuR to target mRNAs, such as Cyclin A2, Cyclin D1, SIRT1, VHL, TP53, and WEE1, and thereby inhibits the proliferation of cervical carcinoma cells [97]. In contrast, HuR associates with the lncRNAs LincRNA-p21 and HOTAIR to facilitate let-7/Ago2-mediated decay of those lncRNAs [98,99].
Tristetraprolin (TTP) is a cysteine-cysteine-cysteine-histidine (CCCH) zinc-finger protein that recognizes AREs in the 3 -UTRs of target mRNAs. The binding of TTP to the 3 -UTR of an mRNA promotes its decay by shortening its poly (A) tail [100]. In the cytoplasm, TTP associates with the Ccr4-Not1 complex and recruits Caf1 deadenylase, leading to hydrolysis of the poly(A) tail of target mRNAs [101]. TTP also mediates decapping of target mRNAs by interacting with decapping complex, which includes Hedls, mRNA-decapping enzyme (Dcp) 1, Dcp2, enhancer of mRNA-decapping protein 3 (Edc3), and the RNA helicase Rck/p54 [102]. TTP mainly functions as a tumor suppressor by targeting mRNAs encoding proteins related to the cell cycle (Cyclin B1 and D1), cell death and proliferation (Bcl-2 and cIAP), angiogenesis (VEGF), and EMT (Snail, Twist1, ZEB1, SOX9, MACC1, MMP2, MMP9, and IL-6) [103]. Recent studies revealed that PD-L1 is a novel target of TTP in several types of cancer cell. In gastric cancer cells, overexpressed TTP negatively regulates the mRNA stability of PD-L1 and thereby contributes to the induction of anti-tumor immunity [104]. IGF2BP 1, 2, and 3 (also known as IMP, CRD-BP, VICKZ, ZBP, Vg1RBP/Vera, or KOC) represent a conserved RNA-binding protein family whose members have two N-terminal RRMs and four C-terminal KH domains. The IGF2BP family proteins are primarily localized in the cytoplasm; they promote target mRNA stability by interfering with endonuclease cleavage or miRNA binding [105][106][107][108][109][110]. IGF2BP1 associates with the coding region determinant (CRD) and 3 -UTR of the β-TrCP1and c-Myc mRNAs, and shields these mRNAs from endonucleolytic cleavage to prevent their degradation [105][106][107]. IGF2BP1 stabilizes the CD44 mRNA through binding to its 3 -UTR, and thereby contributes to cellular adhesion and invasion during cancer development and formation [107]. IGF2BP1 also binds to and stabilizes the β-TrCP1 mRNA by interfering with its miR-183-dependent interaction with Ago2 [108]. In contrast, IGF2BP1 mediates the destabilization of the liver cancer-associated lncRNA HULC by recruiting the Ccr4-Not1 complex in HCC [111]. IGF2BP2 shields the RAF1 mRNA from miR-195-mediated degradation [109]. IGF2BP3 protects HMGA2 from let-7a-directed mRNA decay [110]. Recent studies revealed that IGF2BPs are readers of N 6 -methyladenosine (m 6 A, methylation of adenosine at position 6); they recognize the consensus GG(m 6 A)C sequence and promote target mRNA stability [112]. In cancer cells, IGF2BP increases the stability and storage of the Myc mRNA [112].
Wig1 (also called ZMAT3) is a p53-target gene and can interact with double-stranded RNAs through its zinc-finger domains [113]. Wig1 regulates the stabilities of the p53, p21, and ACOT7 mRNAs by interacting with other proteins, such as hnRNP A2/B1 and Ago2 [114][115][116][117]. Wig1 binds to the ARE in the p53 3 -UTR and thereby protects the mRNA from deadenylation by hnRNP A2/B1 [114,115]. Wig1 mediates miRNA-mediated mRNA decay through an association with Ago2, which is a fundamental component of RNA-induced silencing complex (RISC). Wig1 preferentially binds to a stem-loop structure in the 3 -UTRs of the p21 and ACOT7 mRNAs; this facilitates the recruitment of Ago2, increasing the ability of RISC to access the p21 and ACOT7 mRNAs. The Wig1-and Ago2-containing RISC thus triggers the miRNA-mediated decay of the p21 and ACOT7 mRNAs [116,117].

Translational Regulation
Translational control is a critical means by which gene expression is tuned and regulated in cancer. RBPs, which form the dynamic component of the ribonucleoprotein (RNP) complex, are involved in diverse steps of translation, such as initiation, elongation, and termination. Most of the involved RBPs bind to the 5 -or 3 -UTR with different RBP-binding capacities, leading to different translation efficiencies [118].
Eukaryotic translation initiation factor 4E (eIF4E) is a component of the eIF4F translation initiation complex. eIF4E interacts with eIF4A (an RNA helicase) and eIF4G (a scaffolding molecule), and binds to the 5 -terminal m 7 G cap of mRNAs. After mRNAs are unwound, ribosomes are recruited to mRNAs and translation is initiated. eIF4E is overexpressed in cancer and exerts oncogenic potential by regulating mRNAs related to proliferation (c-Myc, CDK2, and Cyclin D1), metastasis (MMP9 and heparanase), loss of apoptosis (Mcl-1, Bcl-2, and survivin), and angiogenesis (VEGF and FGF2) [119].
HuR regulates the translation of target mRNAs containing ARE in their 3 -UTRs. HuR enhances the translation of the transcripts encoding ProTα, p53, and MSI1 by binding to their 3 -UTRs in an miRNA-independent fashion [120,121]. HuR increases the cytoplasmic abundance and translation of the ProTαmRNA by binding to its 3 -UTR, and an increased level of cytoplasmic ProTαmRNA has been correlated with anti-apoptotic effects in cancer cells [88]. An interaction between HuR and the p53 3 -UTR has been identified in ultraviolet C (UVC)-treated colon cancer cells. After UVC irradiation, the association of HuR with the p53 3 -UTR is increased, as is the translation of p53 [120]. HuR positively regulates the MSI1 mRNA and its translation in glioblastoma [121]. HuR and PTB (also known as hnRNP I) associate with the HIF-1α5 -and 3 -UTRs, respectively. These RBPs cooperatively increase HIF-1αtranslation in hypoxia [122]. In addition, HuR can negatively regulate mRNA translation in an miRNA-dependent manner. For example, HuR recruits miR-19 to the RhoB 3 -UTR, and thereby represses RhoB translation [123].
Musashi (MSI) proteins (MSI1 and MSI2) are members of the class A/B heterogeneous nuclear ribonucleoprotein (hnRNP) family. All MSI proteins have two RRMs (RRM1 and RRM2) for interacting with RNA. MSI1 and MSI2 can positively or negatively regulate mRNA translation in cancer. In breast and brain cancers, MSI1 binds to and represses the mRNA of NF-YA, which is a transcription factor that directs the expression of proteasome subunits. The decreased NF-YA protein level leads to downregulation of 26S proteasome subunit expressions [124]. MSI2 induces the translation of the HOXA9, Myc, and IKZF2 mRNAs, but represses that of the p27 mRNA [125,126]. In pancreatic adenocarcinoma, MSI1 and MSI2 positively regulate the translation of BRD4, c-Met, and HMGA2 [127]. Meanwhile, both MSI proteins suppress the translation of p21 and NUMB [126,127].
RBPs can contribute to translational control by recognizing structural RNA elements, such as the internal ribosome entry site (IRES) and TGF-β-activated-translation (BAT) element [128][129][130][131]. The IRES is a structural RNA element in the mRNA 5 -UTR that promote selective mRNA translation in a cap-independent manner. For example, HuR binds to the IRES of the XIAP mRNA and enhances its translation by increasing its recruitment into the polysome. The HuR-mediated induction of XIAP contributes to cytoprotection against apoptosis-inducing agents [128]. The BAT element is a stem-loop region with an asymmetrical bulge; it is located in the mRNA 3 -UTR. hnRNP E1 forms an RNP complex with eukaryotic elongation factor-1 A1 (eEF1A1) by binding to a 33-nucleotide BAT element in the 3 -UTR of the mRNAs encoding Dab2 and ILEI. hnRNP E1 blocks the release of eEF1A1 from the ribosomal A site, which inhibits the translational elongation of Dab2 and ILEI. This translational silencing regulates EMT in tumorigenesis and metastatic progression [129,130]. PDCD4 acts as a translational repressor by inhibiting the helicase activity of eIF4A. However, in eIF4A-independent manner, PDCD4 mediates translational repression by binding to a secondary structure in the coding region of target mRNAs. PDCD4 recognizes a secondary structure in the coding region of the A-Myb proto-oncogene, and inhibits its translational elongation [131]. Wig1 also interacts with a secondary structure of the ACOT7 mRNA and represses its translation. Wig1 bound to the ACOT7 mRNA forms a complex with RISC components (Ago2 and GW182) and the translation initiation factor, eIF5B, to inhibit translation of the ACOT7 mRNA [117].
Herein, we describe the mechanistic roles of RBPs in cancer. Schematic diagrams of representative RBPs are shown in Figure 1.

Implications of RBPs in Cancer Phenotypes
Cancer has common features, including abnormal cell growth and division. There are many causes of cancer development, including genetic and environmental factors [132,133]. Among the various causes, cancer development can be triggered by genetic alterations of RBPs [3]. RBPs can regulate the expression levels of proto-oncogenes or tumor suppressors, and that many types of RBPs can participate in cancer development [7]. Aberrant expression of these RBPs can affect every step of tumorigenesis, including proliferation, apoptosis, angiogenesis, senescence, and EMT/invasion/metastasis (Figure 2).

Proliferation
Excessive and unregulated cell proliferation can lead to the development of a tumor cell population that gradually becomes malignant. Therefore, it is very important that a potential therapy be able to restrict and inhibit the uncontrolled growth of cancer cells. Recent research has established that RBPs can be regarded as potential drivers of cancer progression, and most of the relevant RBPs modulate tumor cell proliferation [134]. IGF2BP1 and 3 have been considered as oncogenes due to their targeting of some cancer-related mRNAs [112]. IGF2BP1 promotes proliferation via regulating their mRNA targets, such as c-Myc, Ki67 and β-TrCP1 [105,135]. IGF2BP3 also regulates target mRNAs, including cell cycle regulators (TRIM25, Cyclin D1, D3, and G1), eIF4E-BP2, and Myc; these targets are downregulated by miRNAs or ribonucleases, which bind to their 3 -UTRs, leading to accelerated cell proliferation [136][137][138][139].
CPEB1 regulates translation by shortening the 3 -UTR of a target mRNA, and these changes in translation affect cell proliferation [140]. CPEB4 is overexpressed in pancreatic ductal adenocarcinoma and melanoma [72,73]. The cancer-related upregulation of CPEB4 can affect the expression levels of its target mRNAs, which include tPA, MITF, and RAB27A, thereby altering tumor growth [72,73].
hnRNP A1, hnRNP A2, and hnRNP I (also known as PTBP1) are highly expressed in glioblastoma and regulate the alternative splicing of PKM to promote tumor cell proliferation [48,49]. hnRNP I is upregulated in liver cancer and increases tumor growth through a pathway that requires miR-194/hnRNP I/Cyclin D3 signal transduction [141]. hnRNP D, also known as AUF1, binds AREs within the 3 -UTR of the Myc mRNA to compete with TIAR-1 and affect cell proliferation [142]. hnRNP F is upregulated in bladder cancer, where it interacts with the TPX2 protein to alter the expression levels of cell cycle-related genes, such as Cyclin D1 and p21, and thereby affect cell proliferation [143].
HuR contributes to tumor proliferation by modulating mRNAs of cell cycle-related genes, such as those for Cyclin A, Cyclin B1, and Cyclin E1 [144,145].
The La-related protein (also known as LARP3) activates the IRES of the Cyclin D1 mRNA to increase the Cyclin D1 protein level, and this promotes cell proliferation [146]. LARP1 is ubiquitously expressed and modulates mTORC1-regulated proliferation by effectively regulating the translation of mRNAs having a 5 terminal oligopyrimidine (TOP) [147].
RBM3, 5, 6, and 10 are generally mutated in various types of cancer, and are regarded as putative regulators of tumor progression [62,148]. These four RBMs can promote cell growth in liver and lung cancer by interacting with SCD-circRNA [62] or modulating alternative splicing of their target mRNAs, such as the Notch pathway regulator, NUMB [148].
LIN28A and LIN28B, which negatively regulate members of the let-7 miRNA family, promote cell proliferation in cancers derived from various origins. LIN28A and LIN28B regulate proliferation by directly or indirectly binding to and affecting the expression levels of cancer growth-related genes, including HER2 and HMGA1, in both let-7-dependent and -independent manners in breast cancer [149].
SAM68 binds intron 4 of Cyclin D1 and triggers its alternative splicing by interfering with the binding of splicing-related factors. The alternatively spliced Cyclin D1b variant has higher oncogenic potential than Cyclin D1a [65]. The binding affinity of SAM68 to a target mRNA is enhanced by its acetylation, which occurs through inhibition of the deacetylase, HDAC, or activation of the acetyltransferase, CBP, and plays a role in tumor proliferation [150].
QKI exhibits little or no expression in colon cancers due to irregular hypermethylation. Induction of QKI expression promotes the accumulation of p27 and membrane localization of β-catenin, leading to decreases in proliferation and tumorigenesis [151].
MSI1 downregulation reduces cell proliferation by triggering cell cycle arrest at G0/G1 phase. MSI1 also binds to the 3 -UTRs of p21 and p27, and reduces their expression [126]. In addition, loss of MSI1/2 affects the growth of pancreatic cancer from patient-derived tumors in mouse, reducing the expression of proto-oncogenes such as c-Met, Fos, and Fyn [127].

Apoptosis
Cancer cells not only have the ability to continually divide and grow; they are also able to avoid cell death. Apoptosis is an essential process by which normal cells decide to die under a harmful condition. However, cancer cells ubiquitously escape apoptosis to promote tumor progression. Various RBPs participate in this escape by regulating target apoptosis-related mRNAs, such as those for p53, Bcl, Fas, PARP, Caspases, and so on [152].
HuR regulates the stabilities of the mRNAs for MSI1, SIRT1, Bcl-2, Mcl-1, and ProTα to affect apoptosis [121,153,154]. HuR modulates the expression of PDCD4 through a competitive interaction with the RBP, T cell intracellular antigen-1 (TIA-1) [155]. TIA-1 promotes apoptosis by enhancing the alternative splicing of the Fas mRNA or stabilizing PDCD4 [155,156]. eIF4E is involved in regulating the expression levels of Bcl-2 and Bcl-xL, and thereby affects cell survival [157]. LARP family proteins contribute to suppressing cell growth by promoting apoptosis through regulation of Bax, Bcl-2, Bik, Mdm2, and XIAP [158,159]. CELF1 is upregulated in oral squamous cancer cell (OSCC), and its depletion in OSCC cells induced apoptosis by regulating the pro-apoptotic genes Bad, Bax, and JunD [160]. Enhanced expression of dead-end 1 (DND1) is detected in breast cancer cells and is correlated with prolonged survival in patients. Mechanistically, DND1 promotes apoptosis by enhancing the mRNA expression of BIM via a competitive interaction with miR-221 [161]. Reduction of MSI1 diminishes cancer cell proliferation and activates Caspase 3-mediated apoptosis in colon adenocarcinoma xenografted mouse models [162]. In breast cancer, KIN17 is associated with apoptosis. Depletion of KIN17 was associated with reduced cell proliferation and increased activity of Caspase 3/7, which promotes apoptosis [163]. The RBM family member, RBM10, is associated with apoptosis through its ability to regulate the levels of p53 and cleaved PARP [164].

Angiogenesis
Angiogenesis is the physiological process through which new blood vessels and capillaries are formed from preexisted vessels. Cancer proliferation, survival, and metastasis require oxygen and nutrients; thus, cancer needs access to blood vessels. Early observations revealed that rapidly developing tumors are highly vascularized compared to dormant tumors, suggesting that angiogenesis is required for cancer progression. Angiogenesis is promoted by angiogenic activators, such as VEGF, FGF2, and TNF-α [165]. RBPs can modulate the expression of angiogenic factors to play important roles in cancer progression.
HuR expression correlates with the grade of brain tumor [166]. High-level HuR expression is observed in rapidly growing and high-grade brain tumors, such as glioblastoma and medulloblastoma, whereas only weak HuR expression is detected in low-grade brain tumors. In perinecrotic regions of glioblastoma, significant HuR expression is detected and is linked to the expression of angiogenic growth factors. HuR has high binding affinities for the 3 -UTRs of angiogenic factors (VEGF, COX-2, IL-8), an immunomodulating factor (IL-6), TGF-β, and TNF-α, and thus plays important roles in regulating tumor proliferation and angiogenesis in the central nervous system [166]. Alternative splicing of TIA-1 in exon 5 generates a shortened TIA-1 (sTIA-1) isoform. Compared to normal tissues, sTIA-1 is elevated in colorectal cancer and is positively correlated with the tumor stage. VEGF-A is subject to sTIA-1-mediated alternative splicing. Depletion of sTIA-1 or overexpression of full-length TIA-1 (flTIA-1) increases the generation of the anti-angiogenic isoform, VEGF-A165b, leading to decreased tumor formation and vascularization in colon cancer xenografted mice, compared to those overexpressing sTIA-1 [167]. eIF4E affects angiogenesis by regulating the translation of its target mRNAs, VEGF, Cyclin D1, and FGF2 [168,169]. TTP functions as a tumor suppressor in colon cancer. In this disease, TTP is downregulated, whereas VEGF is highly expressed. In a human colon cancer xenograft mouse model, overexpression of TTP enhances degradation of the VEGF mRNA, resulting in the inhibition of tumor growth and angiogenesis [170]. LARP6 expression is elevated in breast cancers; such cancers show especially aggressive phenotypes due to the upregulations of MMP9 and VEGF [171].

Senescence
Cellular senescence is a multi-pronged biological process that leads to permanent cell-cycle arrest. Numerous intrinsic and extrinsic triggers, such as telomere length, genetic alterations, Reactive oxygen species (ROS) production, and chemo-or radio-therapy, can induce cellular senescence in both normal and cancer cells. Senescent cells usually share common characteristics, such as activation of the p53/p21 stress response and/or RB/p16 tumor suppressor pathways, positive staining of senescence-associated β-galactosidase (SA-β-gal), and the senescence-associated secretory phenotype (SASP) [172,173]. RBPs contribute to numerous alterations of gene expression during senescence, and can modulate tumor cell senescence.
HuR modulates cellular senescence by destabilizing the TIN2 mRNA and reducing its translation. Depletion of HuR leads to the mitochondrial localization of TIN2, and the resulting increase in the ROS level induces cellular senescence [174]. Wig1 decreases the p21 mRNA level through binding to the p21 3 -UTR stem-loop structure near the miRNA-recognition site and facilitating p21 mRNA decay through the recruitment of Ago2 to the target site of p21 mRNA. Knockdown of Wig1 increases p21 expression and results in premature senescence. Moreover, the mRNA expression level of Wig1 affects p21 mRNA level in a human lung cancer xenografted mouse model [116].
Fragile X-related protein 1 (FXR1) is upregulated in oral squamous cancer cells. Its depletion induces cellular senescence by inactivating the phosphatidylinositol 3-kinase/Akt signaling pathway and triggering the expressions of senescence-related genes, such as p53, PTEN, p21, and p27. Overexpressed FXR1 regulates the noncoding RNA, TERC, to avoid cellular senescence in oral squamous cancer cells [175]. AUF1 knockout mice exhibit increased cellular senescence in various normal and tumor tissues due to enhanced levels of senescence-related genes, such as p16, p19, and p21, and inactivated transcription of mTERT [176].
TTP controls the levels of p53 and hTERT by promoting the rapid mRNA decay of cellular ubiquitin ligase E6-associated protein (E6-AP). The TTP protein, which is downregulated in cervical cancer compared to normal cervix, functions as a tumor suppressor by inducing senescence through the regulation of p53 protein stability [177].
SRSF1 is an oncogenic splicing factor that is upregulated in various cancers. It interacts with the RPL-MDM2 complex and increases p53 protein expression and activity by blocking the proteasome-mediated degradation of p53, leading to oncogene-induced senescence [178].

EMT, Invasion, and Metastasis
EMT occurs in normal tissues during embryonic development, tissue regeneration, and wound healing, as well as in tumors undergoing progression and metastatic expansion. Epithelial cells can transform to a mesenchymal cell phenotype via several biological processes. When EMT occurs, cell-to-cell adhesion and cell polarity is inhibited; this facilitates invasion and metastasis in cancer. Transcriptional and post-transcriptional alterations of gene expression regulate EMT and the downstream alterations in cell morphology and function. RBPs can manage these processes by modulating the alternative splicing, stability, translation, and polyadenylation of mRNAs during EMT [179].
ESRP1 and/or ESRP2 further contribute to EMT by regulating the alternative splicing of FGFR2, Exo70, and CD44. Gain-and loss-of function studies showed that ESRP1 and 2 are required for the cell to modulate the expression of epithelial FGFR2 exon IIIb or the mesenchymal FGFR2 exon IIIc variant in various normal and cancer tissues [58]. This conversion of FGFR2 splicing yields a less aggressive phenotype and suppresses metastasis in pancreatic cancer [180]. ESRP1 also regulates the alternative splicing of Exo70, a component of the exocyst complex that tethers secretory Golgi vesicles to the plasma membrane prior to fusion. The epithelial Exo70 variant can affect the expression levels of EMT-related genes, such as those encoding Snail and ZEB2, inducing the epithelial phenotype and inhibiting tumor metastasis in breast cancers; in contrast, the mesenchymal Exo70 variant cooperates with the Arp2/3 complex to enhance tumor invasion by inducing actin polymerization [181]. In breast cancer, reduction of ESRP1 changes the variant expression from CD44v to CD44s and inhibits metastasis to the lung. Mechanistically, the ESRP1-mediated alternative splicing of the CD44 mRNA stimulates the cystine transporter (x-CT), which is involved in cysteine uptake and the accumulation of reductive glutathione (GSH); this decreases ROS and thereby prevents cancer cell metastasis [60]. hnRNP M can enhance breast cancer metastasis by stimulating the expression of mesenchymal specific CD44v via a competitive interaction with ESRP1 [52].
hnRNP family members play various roles in EMT and cancer metastasis. A loss-of-function study showed that hnRNP A2/B1 expression is specifically associated with intermediate and mesenchymal cell lines (A549 and H1703). The reduced expression level of hnRNP A2/B1 is related to an increase in E-cadherin and decreases in Twist1 and Snail1 (an inhibitor of E-cadherin), indicating that hnRNP A2/B1 may be correlated with lung metastasis [182]. hnRNP A2 activates invasion by modulating splicing of TP53INP2 mRNA [183]. hnRNP E1 can regulate TGFβ-mediated EMT mechanisms by binding within the 5 or 3 -UTR of its target mRNAs, such as those for Dab2, ILEI, or phosphatase of regenerating liver (PRL)-3, which will lead to activation or repression of these targets [130,184]. hnRNP E1 also inhibits the translation of Inhibin βA by binding to its target mRNA. The translational level of Inhibin βA is elevated after the knockdown of hnRNP E1, and this increased Inhibin βA expression promotes metastasis [185]. The expression of hnRNP R may also influence metastasis and tumor aggressiveness by increasing the mRNA stability of its target, Cyclin B1 and CENPF [186].
KHSRP is significantly expressed in highly metastatic non-small-cell lung carcinoma cells, and its depletion decreases tumorigenic phenotypes, including growth and metastasis, both in vitro and in vivo. KHSRP interacts with hnRNP C and then stimulates the IFN-α/JAK/STAT pathway to promote lung metastasis. Sustained KHSRP expression inhibits TGFβ-mediated EMT by activating miR192-5p to reduce EMT-related factors. In cooperation with hnRNP A1, KHSRP can facilitate alternative splicing to produce the epithelial-specific and EMT-related variants of ENAH, CD44, and FGFR2 [19].
CPEB1 has an inverse correlation with metastasis of breast cancer. Reduction of CPEB1 in TGFβ-mediated EMT results in metastasis via MMP9 mRNA polyadenylation and translation [70]. eIF4E is highly upregulated in colorectal cancer patients with metastatic tissues in liver. Reduction of eIF4E in colorectal cancer cells suppresses the invasion and migration by modulating the expression of MMPs and VEGF [187].
IGF2BP1 is specifically decreased in metastatic breast cancer due to its highly methylated promoter; it affects the migration and proliferation of metastatic cells [188]. IGF2BP2 and 3, contribute to the progression of metastatic breast cancer by modulating the formation of invadopodia, which are actin-rich protrusions of metastatic tumors that act to degrade the extracellular matrix (ECM). In triple-negative breast cancer, IGF2BP2 and 3 cooperate to promote cell migration and invasion by destabilizing the PR mRNA through recruitment of the CNOT1 deadenylase complex [189].
The La protein can promote the translation of the laminin B1 (LamB1) mRNA by binding within its IRES, and thereby contributes to invasion and metastasis in hepatocellular carcinoma [190]. LARP7 is expressed at a low level in breast cancer and interacts with the transcription elongation factors, 7 SK snRNP and P-TEFb. Depletion of LARP7 can promote the escape of P-TEFb from the 7 SK snRNP-harboring complex; the released P-TEFb can encourage transcription elongation by phosphorylating the two negative elongation factors and promoting the activity of PolII, which leads to EMT and enhances invasion and metastasis [191].
The overexpression of LIN28A and/or LIN28B is related to enhanced tumorigenesis and invasive progression in a murine model of invasive colorectal cancer, in association the colonic stem cell markers, LGR5 and PROM1 [192].
SAM68-knockout mice have uterine defects and SAM68-heterozygous mice generated by mating with MMTV-PyMT transgenic mice, in which rapid tumor growth is induced by expression of an oncogene, exhibit delayed tumorigenesis, decreased metastasis, and induction of tyrosine kinase-related signaling [193]. SAM68 also cooperates with scaffold/matrix-associated region-binding protein 1 (SMAR1); phosphorylation of SMAR1 by ERK-1/2 kinase or downregulation of SMAR1 itself lead to increased acetylation of SAM68, which in turn affects the alternative splicing of the CD44 mRNA and thereby modulates cancer metastasis [194].
RBM47 inhibits tumor progression and metastasis by increasing the DKK1 [195]. SRSF1 generates the ∆Ron variant, and enhances EMT and cell motility in breast and colon cancers [32].

RBPs as Therapeutic Targets in Cancer
As described in detail above, RBPs can affect the overall progression of various cancer types, including aspects such as cancer cell proliferation, survival, angiogenesis, metastasis, and senescence, by regulating the expression, localization, and post-translational modification of target mRNAs (Table 1). Accordingly, researchers are constantly seeking to develop drugs that can target RBPs to treat cancer patients [196]. Here, we will review a number of candidate therapeutics, including small-molecule drugs, inhibitors, therapeutic small peptides, and antisense oligonucleotides (ASOs), that are currently being developed and applied in clinical trials (Table 2).   [197] 4Ei-1 Converts to 7Bn-GMP. Antagonizes eIF4E cap binding and induces eIF4E proteasomal degradation, inhibiting translation initiation Breast, lung, mesothelioma [198] 4EGI-1, 4E1RCat, 4E2RCat, Ouabain, Perillyl alcohol, Rapamycin, Torin1, AZD8055 Interferes with association of eIF4E and eIF4G, inhibiting translation initiation Lung, melanoma, lymphoma, colon, liver, breast, prostate [199][200][201][202] MnK inhibitors Inhibits phosphorylation of eIF4E Melanoma, lymphoma, colon, lung [203] Therapeutic peptides Targets RRM1 and 2 of HuR, and inhibits RNA-binding activities of HuR Pancreas, colon, melanoma, brain, breast [209][210][211][212] CMLD-2 Downregulates the expression of HuR Competitively interacts with HuR and inhibits its interaction with target mRNAs Lung, thyroid [213,214] siRNA Enhances the efficiency with chemo-biologic combinatorial drug delivery system Lung [215,216] MSI Small molecule inhibitor Small-molecule drugs can target RBP functions in various human diseases, including cancer, and are already being clinically tested and reported for their anti-cancer effects. For example, eukaryotic translation initiator factor 4F (eIF4F) is an important protein complex that recruits the small ribosomal subunit (40S) to the 5 cap of mRNAs during cap-dependent translation initiation; this complex comprises the three proteins, eIF4E, eIF4A, and eIF4G [119]. eIF4E is a 5 cap-dependent translation initiation factor that is reportedly overexpressed in multiple cancers, where it is associated with aggressive phenotypes. Ribavirin is an antiviral guanosine analog that is able to physically imitate the 5 -terminal m 7 G cap-structure, and thereby interacts with the 5 cap of the eIF4E mRNA. This blocks the transport and translation of eIF4E-regulated oncogenes, such as Cyclin D1, to reduce tumorigenesis in vitro and in vivo [197]. N-7 benzyl guanosine monophosphate tryptamine phosphoramidate prodrug (4Ei-1) prevents eIF4E cap binding and triggers proteasomal degradation of eIF4E, thereby chemosensitizing breast and lung cancer to gemcitabine [198]. 4EGI-1, 4E1RCat, and 4E2RCat disrupt the cooperation of eIF4E with eIF4G and eIF4A to inhibit cap-dependent translation and promote pro-apoptotic effects in vitro and in vivo [199,200]. Ouabain, perillyl alcohol, and mTOR inhibitors (e.g., rapamycin, Torin1, and AZD8055) also interrupt the interactions of eIF4E and eIF4G in various cancer cells [201,202].
When MnK inhibitors were applied to in vitro and in vivo models of melanoma, lymphoma, colon cancer, and lung cancer, blocking the phosphorylation of eIF4E was found to inhibit cell proliferation and metastasis [203].
HuR shows aberrant expression and localization in various types of cancer. MS-444, okicenone, dehydromutactin, DHTS, and AZA-9 are nanomolar inhibitors of HuR that block its RNA-binding activities by targeting the RRM1 and RRM2 of HuR. In the presence of these molecules, HuR does not bind the AREs of target RNAs, such as IL-2, IL-1β, TNF-α, and COX-2, and c-fos; this reduces their HuR-mediated mRNA stabilization. These nanomolar inhibitors of HuR show anti-cancer effects in in vitro and in vivo models of colorectal cancer [209][210][211][212]. CMLD-2, another small-molecule inhibitor, inactivates HuR-mediated RNA stabilization. In non-small-cell lung carcinoma and thyroid cancer cells, CMLD-2 treatment decreases the mRNA expression of HuR or competitively binds to HuR, and thereby downregulates target mRNAs, such as Bax, Bcl-xL, and Mad2. CMLD-2 treatment therefore increases apoptosis and decreases tumor aggressiveness [213,214].
High-throughput analysis to identify potential inhibitors of MSI proteins was employed for cancer therapy. From among 30,000 candidate compounds, oleic acid was found to bind the RRM1 motif in MSI and inhibit its interaction with target mRNAs by promoting conformational changes. MSI can also modulate Stearoyl-CoA desaturase-1 which is required for fatty acid synthesis, eventually affecting cell division [217]. In another screening assay used to search for an inhibitor of MSI, (−)-gossypol, a natural compound known to have an anti-tumor effects in various cancers, was found to cooperate with MSI and inhibit tumor growth by enhancing apoptosis and autophagy in a mouse xenograft model [218]. Ro 08-2750 (Ro) is a novel small molecule that selectively inhibits MSI2 activity. It binds to RRM1 of MSI2 to disrupt its interaction with RNA. The Ro-mediated inactivation of MSI2 reduces the expression of target mRNAs, including those for TGF-βR1, c-Myc, Smad3, and HOXA9. In addition, Ro can inhibit leukemogenesis in both in vitro and in vivo models [219].
Overexpression of LIN28 in cancer inhibits the maturation of let-7 to promote tumor growth. Using a protein/RNA FRET assay, researchers identified compound 1632 as an inhibitor of LIN28. Compound 1632 inhibits the interaction of LIN28 with pre-let-7 and induces the maturation of let-7, resulting in disruption of tumorigenesis in vitro and in vivo in xenograft mice [221]. High-throughput screening identified a number of other LIN28 inhibitors, including compound 1, KCB3602, LI71, and TPEN. Compound 1, KCB3602, and LI71 interact with the CSD of LIN28 to block its interaction with let-7. TPEN destabilizes the ZKD of LIN28. These inhibitors disrupt LIN28-mediated oligouridylation to restore the level of let-7 in cancer and embryonic stem cells [222].
Therapeutic peptides comprise 55 or fewer amino acids; compared to antibodies, they are easier to synthesize and exhibit higher cell penetration and lower immunogenicity [223]. In ovarian cancer, 4EBP-based therapeutic peptides can bind eIF4E to prevent its cap-dependent translation, thereby disrupting tumor growth. This eIF4E-targeting peptide was further merged with an analogue of gonadotropin-releasing hormone (GnRH), which is expressed in the majority of ovarian cancer patients and is known for its anti-cancer effects. The GnRH-4EBP fusion peptide was reported to inhibit tumor growth without inducing any cytotoxicity in a xenograft model of epithelial ovarian cancer [204]. Synthetic peptides have also been used to interrupt the RBM38-eIF4E interaction, with the goal of suppressing translation of the p53 mRNA. Using this strategy, p53 levels were upregulated and tumor growth was inhibited in vitro and in vivo [205].
ASOs and small-interfering RNAs (siRNAs) are commonly utilized to regulate gene expression and have been applied to many diseases as nucleic acid drugs. Both oligonucleotides can bind their target RNAs by Watson-Crick base pairing to modulate splicing, target miRNAs, and inhibit translation [224]. Several FDA-approved or clinically tested drugs take advantage of these characteristics to engage in gene-specific silencing for the treatment of human diseases, including neurological diseases and viral infections [225]. Although no such drug has yet been approved for cancer treatment, many researchers are conducting relevant clinical and nonclinical studies, especially in the context of targeting RBPs. For example, ISIS 183750, which is an ASO drug against eIF4E, disrupts the proliferation of colorectal cancer cells and shows a synergetic effect with the chemical drug, irinotecan. Combination therapy with ISIS 183750 and irinotecan has been tested in phase I/II clinical trials among patients with advanced solid tumors and irinotecan-refractory colorectal cancer [206]. Another ASO, LY2275796, advanced to preclinical studies in mice, where it dose-dependently reduced eIF4E protein and tumors. In humans, a phase I clinical trial has been completed in patients with advanced tumors, and combination therapies with chemical drugs are currently in phase II trials [207]. siRNAs against eIF4E can inhibit tumor growth and stimulate the cytotoxic effects of cisplatin in human breast cancer in vitro and in vivo [208]. HuR can also be targeted with siRNAs, and a chemo-biologic combinatorial drug delivery system can be used to actively target the desired cells or tissues. A siRNA against HuR was loaded into a folic acid (FA)-conjugated polyamidoamine dendrimer (Den)-based nanoparticle, and this formulation was found to effectively decrease HuR expression and cell proliferation in lung cancer cells. In combination with cis-diamine platinum (CDDP), this nanoparticle exhibited synergistically improved anti-tumor effects with reduced cytotoxicity [215]. Similarly, when a siRNA against HuR was loaded to a transferrin receptor-targeted liposomal nanoparticle delivery system, HuR expression was disrupted and tumor growth was inhibited in a mouse lung cancer model [216]. Additionally, ASOs and siRNAs against MSI could inhibit the tumor growth of pancreatic and ovarian cancers in vitro and in vivo [127,220].
RBPs also have potential as a diagnostic and prognostic markers for different types of cancers due to their abnormal expression and functions for mRNA regulation. For example, IGF2BP3 is mostly upregulated and related with invasive phenotypes or poor prognosis in a variety of cancers [226,227]. In addition, cytoplasmic HuR expression is connected with more aggressive phenotypes and poor survival rates in ductal breast cancer from patients [228]. Upregulation of hnRNP C in gastric cancer induces chemoresistance, which is negatively associated with overall survival in patients [229]. Furthermore, research is ongoing to discover RBPs for using a diagnostic and prognostic markers that show cancer-specific expression using advanced bioinformatic tools such as microarray or newly analyzing RNA-seq data based on the Cancer Genome Atlas (TCGA) data [230,231].

Conclusions
Considerable evidence indicates that dysregulation of RBPs occurs in various human cancer types and affects every step of cancer development. RBPs can modulate gene expression at the post-transcriptional and translational levels to induce or reduce cancer-related genes. A number of RBPs have been found to simultaneously regulate numerous genes related to cancer development, resulting in diverse changes in cancer progression and suggesting targets for new therapeutic approaches. With the development of deep-running technologies, such as single-cell analysis and crosslinking and immunoprecipitation (CLIP), many new RBPs and their partners have been discovered and subjected to functional studies. Based on the results of these studies, researchers have attempted to target RBPs and/or their partners in clinical and non-clinical studies using siRNAs, ASOs, and small molecules. Some of these strategies have been found to prevent cancer development and growth, thereby yielding a therapeutic effect. However, there are still some limitations that must be overcome. More research should be conducted to elucidate the detailed functions of the discovered RBPs and to develop specific methods for targeting them in cancer treatment without affecting neighboring normal cells. In addition, it would be useful to improve the target specificity of drug delivery systems with the goal of further enhancing the effects of the developed drugs [232]. In conclusion, we need to develop a more comprehensive knowledge of RBPs in cancer, as they hold great promise as therapeutic targets for the treatment of cancer in the near future.

Conflicts of Interest:
The authors declare no conflict of interest.