Emerging role of tumor-related functional peptides encoded by lncRNA and circRNA

Open reading frames (ORFs) are nucleic acid sequences starting with ATG (or AUG in RNA) and continuing in three-base sets to a stop codon [69]. The length of sORFs is usually < 300 nt. Calculations [70] and ribosome analyses [43] have disclosed that thousands of unannotated ORFs are translated in various species [14, 16, 18, 71‐74]. Longer ORFs are the most likely to be encoded [75‐77]. Regulatory elements (IRES, m6A-modified conserved sites, and so on) upstream in the open reading frame mediate translation [78, 79]. The positional relationship between the ORF and the cyclization site is significant in circRNAs. In general, an ORF spanning the splicing site is the distinctive feature of circRNA-encoded peptides [80]. Websites and software used to predict ORFs are listed in Table 1.

An IRES is an RNA regulatory element that recruits ribosomes, implements ribosomal assembly and reading frame protein translation, and initiates protein translation independent of the 5' cap structure and direct translation [96‐98]. An earlier study found that a circRNA constructed in vitro with an IRES recruited ribosomes and underwent translation [61]. An IRES region has been found in a wide range of viral RNAs [99, 100]. The first was detected in the small RNA virus 5' untranslated region (5′-UTR). IRES have also observed in certain eukaryotic mRNAs. About 10% of mRNAs use IRES in the 5'-UTR to recruit ribosomes [101]. Moreover, the UTR of circ-ZNF609 may serve as an IRES facilitating circ-ZNF609 translation in a shear-dependent manner [78]. IRES may be difficult to detect in higher eukaryotes as these organisms have highly complex genomes and cellular regulatory networks. IRES appear mainly in the 5'-UTRs upstream of the ORFs they control. However, there are exceptions. Certain IRES may be seen between the ORFs while others reside within them [102‐104]. IRES sequences in cells are generally less active and efficient than those in viruses. Nevertheless, the former have good characteristics and are reliable [105, 106]. Endogenous ncRNAs with IRES may translate long polypeptide chains on a continuous ORF [107, 108]. The selective regulation of IRES-mediated translation participates in physiological and pathological processes such as cell growth, proliferation, differentiation, stress response, and apoptosis [98, 109, 110]. Websites currently used to predict IRES are listed in Table 2.

The m6A modification is very common in the mRNAs and ncRNAs of higher organisms [115, 116]. The m6A modification regulates mammalian gene expression [117], as well as RNA stability, localization, shearing, and translation at the post-transcriptional level. It was recently discovered that m6A has various effects on translation [118‐121]. Abnormalities in its regulatory mechanism are associated with tumorigenesis [122, 123]. Using ribosome profiling, computational prediction, and mass spectrometry, m6A-driven endogenous ncRNA translation has been found to be widespread [62, 124]. Numerous translatable endogenous circular RNAs probably contain m6A sites. To examine the ability of m6A to drive circRNA translation, an m6A-modified circRNA was constructed in vitro. The m6A reading protein YTHDF3 was tightly bound to the translation initiation factor eIF4G2. The latter promoted circRNA translation in cells [62]. Commonly used tools for screening m6A motifs are listed in Table 3.

Conserved sequences indicate potential functions and/or play important roles in cell development and regulation [137‐139]. As a rule, coding region sequences are highly conserved. Evolutionary conservation (including ncRNA sequences, sORFs, and small peptide amino acid sequences) may serve as a predictor in the analysis of the coding functions of ncRNAs [71, 75, 140‐142]. For conservation analysis, nucleotide- and protein-protein BLAST [23], UCSC [143, 144], and other websites may be consulted. Alternatively, software such as MegAlign [145], MEGA [146], and Clustal [147, 148] can be used.

Most of the current research on ncRNA-encoded peptides is based on data analyses performed by ribosome display technology [43, 141]. The evolution of high-throughput sequencing has yielded four detection methods for translation omics analysis [149]. These include polysome profiling, ribosome immunoprecipitation/ribosome affinity purification, ribosome profiling (also known as ribo-seq), and ribosome-nascent chain complex (RNC)-seq.

Polysome profiling separates polyribosomes by sucrose density gradient centrifugation as ribosomes have high sedimentation coefficients. The rate of sedimentation during gradient centrifugation increases with the number of ribosomes bound to the mRNA. Thus, mRNAs bound to different numbers of ribosomes may be separated in solution by centrifugation. mRNAs and their active translation ORFs in the separated components are then analyzed, and the output is used to evaluate the ncRNA coding potential [78]. However, the RNA recovery for translation is low in this method and may not suffice to meet the sample size requirement for full spectrum analysis [150, 151]. Ribosome profiling is a comprehensive quantitative method to sequence the mRNA segments in ribosomes [152]. It uses low-concentration RNase to digest the RNC, degrades the mRNA fragments without ribosome coverage, and sequences and analyzes RNA fragments ~22-30 bp long. These are known as ribosome footprints or ribosome-protected fragments. The ribosome distribution and density on each transcript can be determined, as well as the starting codon, ORF location, translation pause area, and other information [153‐156]. The ribosomal characteristics of hundreds of ORFs in annotated non-coding genes, as well as new peptides, may also be identified from ribo-seq data [141, 157‐161]. During translation, ribosomes bind and move along the mRNA chain and gradually synthesize a protein polypeptide chain based on the codon triplet information in the mRNA template. During this process, an RNC is formed. Ribosomes and tandem mRNA precipitates may be separated by sucrose density gradient centrifugation and the mRNA further purified and separated for high-throughput sequencing, known as RNC-seq [162]. Using this method, ncRNA binding to ribosomes may also be analyzed. ncRNA can be translated into proteins in RNCs [79]. In ribosome immunoprecipitation/ribosome affinity purification, specific fusion marker proteins are used to bind ribosomal large subunits, and antibodies against these markers isolate the polymers. The mRNAs and ncRNAs are then isolated for microarray or sequencing analysis [163, 164].

Proteomics can be used to discover and directly detect micropeptides encoded by ncRNAs, which in turn provides the most intuitive evidence that ncRNAs can encode small peptides. Among them, biological mass spectrometry is a common identification and analysis method for these micropeptides. Zhang et al. used immunoprecipitation combined with liquid chromatography tandem mass spectrometry (LC-MS/MS) to characterize the unique amino acid sequences encoded by circ-FBXW7, circ-SHPRH, and circPINT. The distinctive peptides identified in the mass spectrometry results also matched the ORF prediction results [67, 68, 79]. Commonly used software or databases that can be used for protein sequence alignment and peptide search of mass spectrometry data include UniProt [165] and Mascot daemon [166].

As more attention has been given to ncRNA-encoded proteins, several experimental identification methods have emerged to detect these proteins. To verify predicted reading frame expression, FLAG-labeled expression vectors constructed in vitro are imported into cells. Western blots identify distinct bands at the expected molecular weight, indicating that the artificially constructed ncRNA with the FLAG label was translated. CRISPR has also been used to knock FLAG labels into endogenous ncRNA coding regions and detect endogenous protein expression. Sucrose density gradient centrifugation, puromycin treatment, and other techniques determine the extent to which the target ncRNA recruits and binds the ribosomes in the translation machinery. Dual luciferase and other reporting assays elucidate IRES activity and predict ncRNA encoding ability [78, 167]. Overexpression and mutation experiments demonstrate the functions of each regulatory sequence and site. Use of vectors with manipulation of translational elements, such as mutated forms of the predicted IRES, m6A modification sites, or ATG start codons, may confirm whether the translation occurs as normal and the phenotype is consistent. Endogenous translation products may be identified by western blot or with specific antibodies such as those designed for unique amino acid sequences across the circRNA splicing site [67, 68] or common amino acid sequences encoded by lncRNA- and circRNA-derived transcripts [168]. In this way, the translational functions of the endogenous circRNAs may be verified, and the overexpression and knockdown of the translation products can be simulated. The proteins and polypeptides in the samples are isolated and determined by LC-MS/MS.

The circular form of the SNF2 histone linker PHD RING helicase (SHPRH) gene encodes the protein SHPRH-146aa. Circ-SHPRH and SHPRH-146aa are highly expressed in normal human brain tissue and downregulated in glioblastoma. Cyclization in circ-SHPRH results in the tandem stop codon UGAUGA. The entire circ-SHPRH is translated into a 146-aa protein by starting and stopping translation with overlapping genetic codes. An antibody against the unique amino acid sequence generated by the ORF spanning the splicing site and identification of the SHPRH-146aa amino acid sequence by LC-MS/MS confirmed that circ-SHPRH was translated into SHPRH-146aa. The latter participates in the development of central nervous system cancer through regulation of protein ubiquitination pathways. SHPRH-146aa overexpression in U251 and U373 glioblastoma cells reduces their malignancy and tumorigenicity in vitro and in vivo. SHPRH-146aa protects full-length SHPRH from degradation by ubiquitin proteases. It also stabilizes SHPRH as an E3 ligase by ubiquitinating proliferating cell nuclear antigen. In this manner, it inhibits cell proliferation and tumorigenicity [68, 169] (Fig. 1a).

Circ-AKT3 is formed by the cyclization of the third to seventh exons of AKT3. It is 524-nt long and localized mainly to the cytoplasm. When it is driven by an active IRES, circ-AKT3 encodes a 174-aa protein, AKT3-174aa, via the overlapping start-stop codon UAAUGA. AKT3-174aa has the same amino acid sequence as residues 62–232 of AKT3. Compared with normal brain tissue, AKT3-174aa is downregulated in glioblastoma tissue. AKT3-174aa, but not circ-AKT3, acts as a tumor suppressor. AKT3-174aa overexpression inhibits glioblastoma cell proliferation, radiation resistance, and tumorigenicity. The PI3K/Akt pathway plays central roles in various oncogenic signaling pathways promoting glioblastoma development and progression [170, 171]. After PI3K activation, Akt is recruited to the membrane via the PH-domain and is fully activated after Thr308 and Ser473 are sequentially phosphorylated. PDK1 directly phosphorylates Akt at Thr308. This initial step is the most critical in Akt activation. The amino acid sequence of AKT3-174aa is partially identical to that of AKT3. Thus, AKT3-174aa competitively interacts with activated PDK1, inhibits Akt phosphorylation at Thr308, and negatively regulates the PI3K/Akt signaling pathway [172] (Fig. 1b).

Zhang et al. identified certain circRNAs by performing circRNA transcriptome and RNC-RNA sequencing and bioinformatics integration analysis on normal human astrocytes and U251 glioblastoma cells. The second exon of the lncRNA LINC-PINT formed the circular molecule circPINT by self-cyclization. The latter contained an sORF and a natural IRES encoding an 87-aa polypeptide translated from endogenous circPINT exon 2 rather than linear LINC-PINT, termed PINT87aa. It is localized mainly to the nucleus, directly interacts with PAF1, regulates the PAF1/POLII complex, inhibits the transcriptional elongation of the downstream oncogenes cpeb1, sox-2, c-Myc, cyclin D1, and others, and inhibits the proliferation and tumorigenesis of glioblastoma and other cancer cell types [79] (Fig. 1c).

Circ-FBXW7 may have an ORF spanning the splice site. It is highly conserved among different species and encodes a 185-aa protein driven by an IRES independently of the 5' cap translational machinery. Circ-FBXW7 was able to be translated in human cells using a construct harboring a FLAG sequence before the ORF stop codon. The circ-FBXW7 IRES-mut vector, which has a mutation in the IRES sequence, was transfected into U251 and U373 cells. However, cells transfected with the vector formed a circular RNA similar to circ-FBXW7. Therefore, it is FBXW7-185aa rather than circ-FBXW7 that induces cell cycle arrest and hinders glioma cell proliferation. FBXW7a is the most abundant isoform of FBXW7. It uses c-Myc as a tumorigenesis regulator for ubiquitination-induced degradation. The de-ubiquitinating enzyme USP28 stabilizes c-Myc by binding it via interaction with the N-terminus of FBXW7a. The protein FBXW7-185aa translated by circ-FBXW7 has a relatively higher affinity for USP28. It functions as bait and competitively inhibits USP28 from binding FBXW7a. In this way, it perturbs c-Myc stabilization induced by USP28 and shortens its half-life. FBXW7-185aa is a synergistic parental gene that encodes FBXW7a, stabilizes c-Myc, inhibits tumor cell proliferation and malignant phenotypes, and impedes malignant glioma progression [67] (Fig. 1d).

Human papillomaviruses produce the 472-nt oncogene called CircE7 containing the entire E7 ORF. CircE7 is modified by m6A and is localized mainly to the cytoplasm. It is closely associated with polyribosomes and may be translated to the E7 oncoprotein. The latter process is upregulated by cellular stressors such as heat shock. E7 translation was increased two- to four-fold under 42 °C heat shock. CircE7 knockdown in CaSki cervical cancer cells reduces E7 protein levels, inhibits cancer cell proliferation and colony formation, and suppresses tumor growth and malignancy. CircE7 is essential for E7 protein expression and transformation in CaSki cervical cancer cells both in vitro and in transplanted tumors [124] (Fig. 1e).

The protein UBAP1-AST6 is translated from a lncRNA, is localized mainly to the nucleus, and is expressed in A549 lung cancer cells. UBAP1-AST6 promotes cancer, and its overexpression significantly induces cancer cell proliferation and colony formation [173] (Fig. 1f).

CircPPP1R12A is highly expressed in colon cancer tissues and serves as a prognostic marker of survival. Patients with increased circPPP1R12A have comparatively poorer overall survival. CircPPP1R12A contains a short 216-nt ORF encoding the conserved 73-aa peptide circPPP1R12A-73aa. Silencing CircPPP1R12A markedly inhibits colon cancer cell proliferation, migration, and invasion. Construction of the FLAG-circPPP1R12A overexpression vector with an initiation codon mutation ATG/ACG confirmed that it is circPPP1R12A-73aa rather than circPPP1R12A that plays key roles in colon cancer cell proliferation, invasion, and metastasis. The YAP1-specific inhibitor peptide 17 dramatically attenuated colon cancer cell proliferation, migration, and invasion promoted by circPPP1R12A-73aa overexpression. Induction of colon cancer growth and metastasis by circPPP1R12A-73aa was validated in vitro and in vivo by activating the Hippo-YAP signaling pathway [174] (Fig. 2a).

The lncRNA HOXB-AS3 is a tumor suppressor that is substantially downregulated in highly metastatic and primary colorectal cancer (CRC) tissues. HOXB-AS3 binds ribosomes and encodes a highly conserved 53-aa peptide called HOXB-AS3. It is endogenous, naturally occurring, and widely expressed in various tumor tissues. HOXB-AS3 inhibits cancer cell proliferation, invasion, and metastasis and suppresses tumor growth. Colon cancer patients with low HOXB-AS3 levels generally have poor prognoses. HOXB-AS3 competitively binds arginine in the RGG motif of hnRNP A1. In this manner, it blocks hnRNP A1 binding to the pyruvate kinase M (PKM) EI9 sequence, antagonizes hnRNP A1-mediated PKM splicing regulation, and inhibits PKM 2 subtype formation and miR-18a production. HOXB-AS3 downregulates PKM2 but upregulates PKM1. PKM2 is a key regulator of aerobic glycolysis and increases lactic acid production. Therefore, HOXB-AS3 inhibits aerobic glycolysis in CRC cells. The loss of HOXB-AS3 is a key oncogenic event in CRC metabolic reprogramming [175] (Fig. 2b).

CircLgr4 is highly expressed in advanced CRC and is associated with poor prognosis. LGR4 is also highly expressed in colorectal tumors and activates Wnt/β-catenin signaling via ubiquitination and FZD receptor stabilization. Thus, it drives colorectal stem cell self-renewal and invasion. CircLgr4 encodes the circLgr4-peptide, which interacts with LGR4 to activate the LGR4-Wnt signaling pathway. CircLgr4 drives colorectal stem cell self-renewal and invasion in a manner dependent on LGR4. The circLgr4-peptide-Lgr4 axis may be used in targeted CRC therapy [176] (Fig. 2c).

Circβ-catenin is derived from CTTNB1, which encodes β-catenin, a major regulator of the Wnt pathway in liver cancer. Circβ-catenin is upregulated in hepatocarcinoma tissues and is localized mainly to the cytoplasm. It has an ORF and an active IRES encoding the 370-aa β-catenin isomer β-catenin-370aa. Circβ-catenin knockdown inhibits hepatoma cell growth and migration in vitro and in vivo, impedes tumorigenesis and metastasis, and suppresses the Wnt/β-catenin pathway. Construction of the circβ-catenin expression vector with an initiation codon mutation disclosed that its functionality could be attributed to its protein-coding ability rather than its non-coding property. Circβ-catenin knockdown had no effect on the CTTNB1 mRNA level but significantly reduced the β-catenin protein level. β-catenin stability is closely linked to its phosphorylation state. After β-catenin is phosphorylated by GSK3β, it is ubiquitinated by the ubiquitin ligase β-TrCP and degraded by the proteasome. β-catenin-370aa, encoded by circβ-catenin, interacts with GSK3β and acts as a bait to block it from binding the full-length β-catenin protein. In this manner, it represses GSK3β-induced β-catenin degradation. In liver cancer, β-catenin-370aa stabilizes β-catenin by reducing its ubiquitination, activating the Wnt/β-catenin pathway, and promoting tumor growth [168] (Fig. 2d).

LINC01420 is a lncRNA that is highly expressed in nasopharyngeal carcinoma. The overall survival rate is low in patients with nasopharyngeal carcinoma presenting with elevated LINC01420 expression. LINC01420 knockdown significantly inhibits nasopharyngeal carcinoma cell invasion [177]. The sORF of LINC01420/LOC550643 encodes a highly sequence-conserved microprotein named nobody. It interacts with an mRNA capping protein, directly binds EDC4, removes the 5' cap from mRNA, promotes 5'-to-3' decay, and regulates the degradation of normal and aberrant transcripts. Nobody is localized mainly to P-bodies. Its level decreases with increasing P-body number. The latter perturbs the homeostasis of endogenous cellular nonsense-mediated decay substrates. Nevertheless, the effects of this process on tumor growth, development, and metabolism are unclear [178] (Fig. 2e).

LINC00961 is substantially downregulated in human non-small cell lung cancer (NSCLC). Low tissue LINC00961 levels are associated with clinical stage, lymph node metastasis, and shorter survival time in NSCLC patients [179, 180]. LINC00961 may also inhibit tumor progression in oral squamous and renal cell carcinoma, glioma, and other cancers [181‐184]. Matsumoto et al. reported that LINC00961 is translatable. Its encoded small peptide SPAR is localized to the late lysosome and interacts with lysosomal V-ATPase. SPAR functions upstream of Rags and the Ragulator complex and at the v-ATPase level. It induces interactions of the v-ATPase-Ragulator-Rags supercomplex. SPAR impedes lysosomal mTORC1 reuptake, inhibits mTORC1 activation by amino acid stimulation, and affects muscle regeneration [185‐187]. Circ-ZNF609 is formed from the cyclization of the second exon of ZNF609. It is upregulated in nasopharyngeal carcinoma, renal and breast cancer, and other cancers. Circ-ZNF609 knockdown dramatically inhibits cancer cell proliferation, invasion, and metastasis [188‐190]. Bozzoni et al. reported that circ-ZNF609 was strongly expressed in muscle cells, highly conserved evolutionarily, and contained a 753-nt ORF. Its UTR had IRES-like activity and encoded a protein in a splicing-dependent manner. This peptide regulated myoblast proliferation [78]. MiPEP-200a and miPEP-200b, encoded by primary miRNAs (miR-200a and miR-200b), can inhibit the migration of prostate cancer cells by regulating the epithelial to mesenchymal transition of tumor cells [191].

CircRNAs, lncRNAs, and the small peptides they encode may regulate tumorigenesis. Moreover, certain ncRNAs in various species encode proteins that regulate various biological and disease processes in vivo. For example, peptides 11–32 aa long encoded by sORFs from polished rice control epidermal differentiation in Drosophila by modifying the transcription factor Shavenbaby [14]. Myomodulin (MLN) is a highly conserved micropeptide encoded by a 138-nt ORF in a lncRNA. MLN structurally and functionally resembles phospholipids and phosphatidylcholine and inhibits SERCA in a similar manner. In this way, MLN regulates muscle motility [15]. A muscle-specific lncRNA encodes a small 34-aa peptide, DWORF. It increases the activity of SERCA pumps which, in turn, enhance cardiac contractility during a heart attack [192]. In the Drosophila heart, a lncRNA (pncr003:2L) encodes two peptides ≤ 30-aa long that regulate calcium transport and affect muscle contraction [23]. Pauli et al. found that the short, conserved polypeptide Toddler encoded by a lncRNA in zebrafish promotes cell movement during gastrulation by activating APJ/apelin receptor signaling [193]. Pri-miR171b from alfalfa and pri-miR165a from Arabidopsis produce peptides that promote the accumulation of mature miRNAs and downregulate target genes regulating root development [194]. Kadener et al. reported that circMbl-encoded proteins are enriched in synaptosomes and modulated by starvation and FOXO [94]. Abou-Haidar et al. found a covalently closed, 220-nt circular RNA in a viroid. The translated protein was rich in basic amino acids, expressed only in RYMV-infected rice plants, and bound homologous (scRYMV) and heterologous [potato virus X] RNA [195].