Small open reading frame-encoded microproteins in cancer: identification, biological functions and clinical significance

Abstract

The human genome harbors approximately twenty thousand protein-coding genes, and a significant portion of life science research focuses on elucidating their functions and the underlying mechanisms. Recent studies have revealed that small open reading frame (sORF), originating from non-coding RNAs or the 5’ leader sequences of messenger RNAs, can be translated into small peptides called microproteins through cap-dependent or cap-independent mechanisms. These microproteins interact with diverse molecular partners to modulate gene expression at multiple regulatory levels, thereby playing critical roles in various biological processes. Notably, sORF-encoded microproteins exhibit aberrant expression patterns in cancer and are implicated in tumor initiation and progression, expanding our understanding of cancer biology. In this review, we introduce the translational mechanisms and identification methods of microproteins, summarize their dysregulation in cancer and their biological functions in regulating gene expression, and emphasize their roles in driving hallmark events of cancer. Furthermore, we discuss their clinical significance as diagnostic and prognostic biomarkers, as well as therapeutic targets.

Introduction

The majority of human genome sequences can be transcribed into RNA molecules. However, only approximately 3% of them are capable of encoding proteins, indicating most transcripts belong to non-coding RNAs (ncRNAs) [1]. Compelling evidence demonstrate that ncRNAs function as dynamic regulators to play crucial roles in different cellular processes, and their dysregulation is associated with disease pathogenesis [2, 3]. With the advancement in bioinformatics analysis and experimental technologies, an increasing number of ncRNAs with small open reading frame (sORF) have been discovered to encode small peptides called microproteins, typically fewer than 100 amino acids [4]. Besides ncRNAs, including long non-coding RNA (lncRNA) [5], circular RNA (circRNA) [6] and primary microRNA (pri-miRNA) [7], the 5’ leader sequences of messenger RNAs (mRNAs) [8] have also been identified as the sources of microprotein.

The sORF-encoded microproteins are conserved among species and exhibit tissue- or development-specific expression patterns. Growing evidence indicate that sORF-encoded microproteins play important biological functions under physiological conditions, such as signal transduction and intercellular communication during embryonic development [9, 10], regulation of autoimmune responses and control of muscle performance [11, 12]. Notably, aberrant expression of sORF-encoded microproteins has been observed in human diseases, such as cardiovascular diseases, muscular dystrophy and cancer [13,14,15,16,17,18]. In cancer progression, dysregulated microproteins may play oncogenic or tumor suppressive roles, highlighting their potential to serve as diagnostic and prognostic biomarkers and promising therapeutic targets for cancer management [19, 20].

In this review, we briefly introduce the translational mechanisms and identification methods of sORF-encoded microproteins, summarize how microproteins are dysregulated in cancer and regulate gene expression, emphasize their biological functions during tumorigenesis, and finally highlight their clinical significance.

Translational mechanisms of sORF-encoded microprotein

Cap-dependent translation

Eukaryotic mRNAs are linear RNA molecules usually with a 7-methylguanosine (m7G) cap at the 5′ end and a poly(A) tail at the 3’ end. Typically, mRNAs are translated into proteins in a cap-dependent fashion. In this process, the small ribosomal subunit (40 S) first associates with the eIF4F complex, which includes the translation initiation factor eIF4E, to recognize the 5′ cap, followed by ribosome scanning along the mRNA to locate the first start codon AUG [21, 22]. Upon recognizing the start codon, the 40 S subunit recruits the large ribosomal subunit (60 S) to form the 80 S translating ribosome for polypeptide synthesis [23].

Most lncRNAs have a 5′ cap structure and a 3′ poly(A) tail, referred to as ‘mRNA-like’ ncRNA [24,25,26], indicating that the sORF-encoded microproteins from lncRNA are translated in the cap-dependent manner (Fig. 1A). Employing the RiboTaper algorithm [27], it was found that although 96% of these mRNA-like ncRNAs can interact with the translational machinery, about 1% have the potential to produce proteins [24].

Fig. 1

Translational mechanisms of microproteins. (A) Some lncRNAs with a 5′ cap structure and a 3′ polyA tail are translated into microproteins in a cap-dependent manner; (B) The 5’UTR of mRNAs is translated by leaky scanning; (C) IRES-mediated transaltion; (D) m⁶A modification-mediated transaltion

Leaky scanning serves as an alternative translation mechanism similar to the traditional cap-dependent translation. In leaky scanning, the 40S ribosomal subunit recognize the start codon of upstream open reading frame (uORF) located within the 5’-untranslated region (UTR) of mRNA in a cap-dependent manner, where it recruits the 60 S ribosomal subunit and translation initiation factors to trigger the translation of the uORF into a microprotein (Fig. 1B) [28]. Alternatively, the 40 S subunit can bypass upstream start codons and continue to scan downstream until it reaches the main ORF (mORF), resulting in the initiation of mORF translation [29, 30].

The proteins from these two translation modes may have distinct localizations. For example, the mORF of SLC35A4 mRNA encodes a 324-amino-acid protein localized to the Golgi apparatus, while the uORF of the same mRNA is translated into a 103-amino-acid microprotein localized to the mitochondrial inner membrane, known as SLC35A4-MP [31, 32]. Typically, the translation of uORF can be initiated by either AUG or non-AUG codons. For example, Na et al. showed that the translation of uORF within LAMA3 mRNA begins at non-AUG codon AGG, producing the alt-LAMA3 protein [33].

IRES-dependent translation

Due to the lack of 5′ cap structure and 3′ poly(A) tail, circRNAs and certain lncRNAs are translated through a cap-independent mechanism. Internal ribosome entry sites (IRESs), typically located upstream of their corresponding ORF, are highly structured cis-acting RNA elements that enable the recruitment of ribosomes to or near start codon, thereby facilitating protein translation through a cap-independent pathway [34]. Increasing studies have reported that circRNAs and certain lncRNAs are translated via IRES-dependent translation (Fig. 1C). For example, Xiao’s group reported that circPDHK1 had an ORF with an AUG start codon and an IRES sequence with ~ 150 nucleotides in length, and experimentally validated that circPDHK1 promoted tumor growth and metastasis in clear cell renal cell carcinoma (ccRCC) through encoding a functional peptide termed PDHK1-241aa [35]. Yu et al. proved that DNA damage facilitated the association of ribosome with the IRES region of lncRNA CTBP1-DT, thus enabling the translation into the novel DDUP protein with 186 amino acid residues [36]. Noteworthy, the activities of IRESs often require assistance from other factors known as IRES-transacting factors (ITAFs). In melanoma cells, heterogeneous nuclear ribonucleoprotein A1 (hnRNP-A1) has been reported to promote the IRES-dependent translation of the lncRNA meloe to produce the microprotein MELOE-1, a melanoma-specific neoantigen [37].

m⁶A-dependent translation

N6-methyladenosine (m⁶A) is a type of RNA modification in which the N6 position of adenosine (A) within RNA molecules is methylated [38, 39]. m⁶A modification is dynamically regulated by methyltransferases called writers and demethylases termed as erasers, and this modification plays important roles in protein translation of sORF (Fig. 1D). For example, Zheng et al. found that overexpression of methyltransferases METTL14 and METTL16 significantly increased the m⁶A levels of circMIB2 and promoted the expression of its encoded microprotein MIB2-134aa [40]. In contrast, knockdown of the demethylase ALKBH5 dramatically increased LINC00278 m⁶A level and LINC00278-encoded 21-amino-acid microprotein [41].

m⁶A modification may be specifically recognized and bound by certain RNA-binding proteins called readers, such as IGF2BP1 and YTHDF1/2/3, to recruit translation machinery to facilitate the translation of sORFs [16, 42,43,44]. For example, Zeng et al. reported that YTHDF3 bound to the m⁶A sites of circ-YAP and recruited eIF4G2 translation initiation complex, thus facilitating ribosome assembly to produce the protein YAP-220aa [44]. Besides, YTHDF2 is reported to recognize the m⁶A modification of circMET and regulate the expression of circMET-encoded microprotein [16].

Interestingly, the lncRNA HNF4A-AS1 produces a small 51-amino acid peptide (sPEP1) using a mechanism different from either IRES- or m⁶A-mediated translation. Song et al. revealed that miR-409-5p interacted with HNF4A-AS1 to facilitate sPEP1 translation through recruiting the translation initiation factor eIF3G [45]. However, it remains unclear whether it operates with the cap-dependent translation.

Identification methods of sORF-encoded microproteins

Bioinformatics tools for microprotein prediction

Bioinformatics tools are commonly employed to predict microproteins by analyzing genomic or transcriptomic data from databases. These tools identify microproteins either by analyzing ORF or by coordinating analysis of m⁶A modifications or IRES elements with the presence of ORF (Fig. 2A).

Fig. 2

Identification methods for microproteins. (A) Prediction of microproteins from transcriptomic and genomic data by bioinformatics tools; (B) Microprotein identification by ribo-seq; (C) Discovery of microproteins by MS; (D) Experimental validation of microproteins using exogenous expression and endogenous detection

Currently, most tools predict ORFs by sequence alignments or alignment-free methods, as exemplified by ORF-FINDER, OrfPredictor, CPAT and GeneMarkS-T (Table 1). Among them, ORF-FINDER is the earliest developed tool for ORF searching in cDNA by sequence alignments [46]. Subsequently, Min et al. developed OrfPredictor to predict protein-coding regions in EST-derived sequences. OrfPredictor utilizes BLASTX to guide the identification of coding regions with a hit and predicts coding region ab initio for sequences without a hit [47]. With the advent of next-generation sequencing, the prediction of ORF in bulk transcriptomic data requires faster and more accurate analytical tools, boosting the development of machine learning-based ORF prediction techniques. Notably, Wang et al. introduced a supervised machine learning software called CPAT (Coding-Potential Assessment Tool), which utilizes a logistic regression model based on four sequence features: ORF length, ORF coverage, Fickett score and hexamer usage bias, thus enabling accurate identification of coding and noncoding transcripts from a pool of candidates [51]. Unlike CPAT, Borodovsky’s team developed an unsupervised machine learning tool, GeneMarkS-T, which makes manually curated preparation of training sets unnecessary and is robust with respect to the presence of transcripts assembly errors [48]. Besides, sORF Finder and MiPepid are specifically designed to predict sORF (smaller than 300 base pairs). sORF Finder evaluates sORF coding potential based on the nucleotide composition bias between coding and non-coding sequences [52], while MiPepid, a machine-learning tool, demonstrates superior performance in evaluating the coding potential of the test set, achieving an overall accuracy of 0.96, compared to sORF Finder that attains an overall accuracy of 0.87 [53].

Table 1 Tools for prediction of ORF, IRES and m⁶A sites

As mentioned above, IRES sites located upstream of the ORF in lncRNA and circRNA could recruit ribosomes to initiate the translation of downstream ORF. IRESite, IRESPred and IRESfinder are currently the most commonly used tools for IRES prediction. IRESite provides a collection of experimentally validated IRES sequences and related information, allowing users to query whether the RNA sequence of interest contains known IRES elements [55]. However, it cannot predict novel IRES elements for the subjected sequences. IRESPred predicts IRES elements via machine learning model, which was built based on sequence and secondary structure characteristics of UTRs and the probabilities of interactions between UTRs and small subunit ribosomal proteins [58]. However, IRESPred has some limitations, such as the use of unvalidated non-IRES sequences as negative samples, a small training dataset, and the inability to handle large-scale data. Subsequently, Song’s team developed IRESfinder, which leverages 19 carefully selected k-mer features to construct a logistic regression model, achieving 80% precision and 73% accuracy in independent testing [56]. IRESfinder is also employed in CircBank database for IRES prediction within circRNA [63].

Various m⁶A site prediction tools have been reported, including SRAMP [59], MethyRNA [61], M6AMRFS [60] and M6APred-EL [62]. These tools primarily predict m⁶A sites based on RNA sequence features (i.e., nucleotide patterns, m⁶A consensus motifs), RNA secondary structures and translation-related signals. These tools have been widely applied to the studies of microprotein-encoding transcripts. For instance, Duan et al. used SRAMP to show that circMAP3K4 has a high potential for m⁶A modifications, which may be involved in circMAP3K4 translation [43].

Despite the development of these bioinformatics tools for microprotein prediction, the results may include false positives. To address this challenge, techniques such as ribosome profiling sequencing (Ribo-Seq), which provides direct evidence of translation activity, and mass spectrometry (MS), which directly proves the presence of microproteins, can be combined with bioinformatics predictions to improve the accuracy and reliability of microprotein identification.

Ribo-Seq

Ribo-Seq is a high-throughput sequencing-based technology to study global translation activity within cells [64]. The experimental workflow includes cell lysis to release ribosomes and their bound mRNAs, followed by nuclease digestion of free mRNAs unprotected by ribosomes. Ribosomes and their protected mRNA fragments are then isolated using ultracentrifugation or affinity purification. The ribosome-protected fragments, typically 28–32 nucleotides in length, are extracted to construct a cDNA sequencing library for high-throughput sequencing. Subsequent data analysis involves mapping the short fragments to the reference genome to identify translation initiation and termination sites, translation efficiency and potential coding regions. This technique is widely used in exploring the translational potential of sORFs (Fig. 2B) [65].

To analyze Ribo-Seq results, various software tools have been developed to explore the translational potential of ORFs (Table 2). RiboTaper, RibORF and RiboCode detect actively translated regions by utilizing the triplet periodicity of ribosome-protected fragments (RPFs) [27, 67, 69]. These tools have proven instrumental in microprotein research. For example, van Heesch et al. employed bioinformatics tools including RiboTaper to identify 1,090 microproteins encoded by lncRNAs and circRNAs in human heart [80], while Jackson et al. used RibORF and RiboCode to identify 224 microproteins in mouse bone marrow-derived macrophages [81]. However, the output of these methods could be affected by low-quality or low-coverage data. To address these issues, the computational method called PRICE was developed, which employs the Expectation-Maximization (EM) algorithm to model experimental noise and accurately infer codon activity probabilities [70]. Additionally, tools like RiboToolkit and GWIPS-Viz enable online analysis of Ribo-Seq data [77, 78].

Table 2 Tools for ORF prediction in ribo-seq data

MS-based approaches

MS-based proteomics can provide direct and robust evidence to identify sORF-encoded microproteins by detecting specific peptide fragments of microproteins [82]. The workflow for MS-based identification of sORF-encoded microproteins includes microprotein enrichment, MS sample preparation, MS data acquisition and raw data analysis (Fig. 2C).

Due to the low abundance of microproteins, different methods are developed to enrich microproteins from complex biological samples. Based on the properties of microproteins (e.g., size, hydrophobicity and charge), diverse enrichment methods can be employed, such as gel separation, molecular weight cutoff filters (10 or 30 kDa), precipitation with organic solvents (e.g., acetonitrile, methanol acetic acid), C8 solid-phase extraction (C8-SPE), and size-exclusion chromatography (SEC) [83,84,85]. The differences in enrichment methods leads to biases in microprotein identification. For instance, Zhang et al. found that C8-SPE and molecular weight cutoff methods had only a 7.58% overlap in identified microproteins, but combination of both methods successfully identified 762 novel microproteins across multiple tissues and cell lines [86]. Moreover, fractionation can effectively reduce sample complexity and improve microprotein identification [85, 87]. For example, Yang et al. used the SEC fractionation to enrich mouse microproteins, achieving a 1.4-fold increase in microproteins compared to unfractionated method [85].

Because of their small molecular weight, microproteins could be subjected to MS without enzymatic digestion into peptides. For example, Wang et al. identified 241 microproteins from Hep3B cells using undigested protein samples [88]. Despite this, enzymatic digestion is usually employed for the identification of microproteins. For enzymatic digestion, trypsin is commonly used to cleave peptide chains at the carboxyl side of lysine (K) and arginine (R) residues. However, microproteins may have few arginine and lysine residues, resulting in compromised sequence coverage detected by MS. Therefore, a multi-enzyme digestion strategy combining trypsin with other proteases, such as Glu-C, Lys-C or Asp-N, can be employed. Kaulich et al. demonstrated that digestion of multiple proteases in LC-MS/MS analysis improved the sequence coverage and the number of identified peptides for microproteins [89].

Data-Dependent Acquisition (DDA) is a commonly used MS data acquisition mode that selects precursor ions for data collection based on signal intensity, which has successfully identified thousands of microproteins in different species [86, 87, 90]. However, DDA tends to favor the detection of precursor ions with strong signals, often overlooking low-abundance molecules. Another acquisition mode, Data-Independent Acquisition (DIA), fragments ions across all mass ranges, capturing of both low- and high-abundance molecules and thereby enhancing its suitability for microprotein detection. For example, Martinez et al. successfully identified 85 microproteins in mouse adipocytes using DIA [91]. Because the spectra generated by DIA are complex to analyze [92], directDIA was thus developed to simplify the workflow and improve sensitivity and accuracy [93].

Currently, some publicly available databases have been developed for the analysis of MS raw data to identify microproteins, such as SmProt [94], sORFs.org [95] and OpenProt [96] (Table 3). Data in sORFs.org are derived from ribosome profiling, while SmProt and OpenProt integrate additional data sources like genomic and MS data. These databases provide detailed microprotein information, such as their sequences, genomic locations and start codons. Notably, SmProt and sORFs.org include both AUG and non-AUG initiated microproteins, while OpenProt only has AUG-initiated ones. Given the overlap between these databases is relatively low [88], combination of multiple databases may improve the identification of microproteins.

Table 3 Biological functions of lncRNA-encoded microproteins in cancer

In MS data analysis, some spectra may fail to match peptides in the database. For unmatched spectra, de novo sequencing can be employed to identify microproteins [101], as exemplified by the pNovo software, which conducted de novo peptide sequencing and mapped 1,682 peptides to 2,544 sORFs randomly distributed across human chromosomes [88]. In addition, Pan et al. combined database-dependent analysis with de novo sequencing to successfully identify 1074 microproteins in mouse tissues [102].

Experimental validation of microproteins

Microproteins predicted or detected as above are subsequently validated using two experimental strategies: exogenous expression and endogenous detection. For exogenous expression, sORF is fused with fluorescent protein (e.g., GFP, mCherry) or epitope tag (e.g., FLAG, HA, Myc) to construct its expression plasmid, which is then introduced into cells to express the microprotein of interest [5, 103]. Specifically, epitope tag or fluorescent protein lacking start codon is cloned into the C-terminus of target sORF, and antibodies against the fluorescent protein or epitope tag are then used to detect the expression of microprotein by immunoblotting and immunofluorescence. For instance, Zhang et al. demonstrated that four lncRNA with sORF may encode microproteins by ectopic expression of microproteins with FLAG and GFP tags [85].

The endogenous microproteins can be detected through the specific antibodies developed against these microproteins. For instance, Li et al. validated the expression of microprotein MIAC using customized monoclonal antibodies [104]. However, due to the small molecular weight and low antigenicity of certain microproteins, it is challenging to develop specific and effective antibodies. In this case, CRISPR/Cas9-mediated gene editing can be employed to insert fluorescent or epitope tags into the target DNA sequence, enabling the detection of microprotein expression and localization by the corresponding tag antibodies. For example, Na et al. generated Cas9-directed knock-in HEK293T cell lines with a 3xGFP11-FLAG-HA tag appended to the 3’ end of corresponding ORF and validated the endogenous expression and subcellular localization of tagged microprotein, shedding light on their potential biological functions [33].

Dysregulation of sORF-encoded microproteins in cancer

Accumulating evidence has demonstrated that the expression of sORF-encoded microproteins is widely dysregulated in a variety of malignancies. These alternations are often associated with disruptions in the regulatory mechanisms governing RNA abundance, translation efficiency, or protein stability, as outlined in Tables 4 and 5.

Table 4 Representative databases for microprotein research

Table 5 Biological functions of circRNA-encoded microproteins in cancer

Dysregulation of specific transcription factors can lead to abnormal expression of microprotein-coding transcripts. For example, the transcription factor GATA3 has been shown to suppress the transcription of LINC00887 by binding to two responsive elements within its promoter. In ccRCC, reduced GATA3 expression results in the upregulation of LINC00887 and its encoded microprotein, ACLY-BP [106]. Conversely, in hepatocellular carcinoma (HCC), the TGF-β-activated transcription factor SMAD3 promoted the transcription of LINC02551, leading to a marked upregulation of its encoded microprotein, JunBP [122].

Unlike linear RNAs, circRNAs are generated by the back-splicing of primary transcripts, a process regulated by cis-acting elements (e.g., Alu repeats) in the flanking introns and diverse trans-acting factors. Dysregulation of these factors probably contributes to the aberrant biogenesis of circRNAs encoding microproteins. For instance, the DExH-box helicase 9 (DHX9) specifically binds to Alu repeats, thereby inhibiting circularization by disrupting the pairing of these repeats [184]. In intrahepatic cholangiocarcinoma (ICC), DHX9 expression is decreased upon IL-6 stimulation, leading to elevated levels of the circRNA cGGNBP2 and its encoded microprotein, cGGNBP2-184aa [154]. In addition to helicase like DHX9, splicing factors also exert indispensable roles in governing circRNA biogenesis. Song et al. demonstrated the splicing factor SLU7, which was significantly downregulated in triple-negative breast cancer (TNBC), bound to Alu elements within primary transcripts of circCAPG to inhibit its circularization, ultimately resulting in reduced expression of its encoded microprotein, CAPG-171aa [151]. However, how these trans-acting factors selectively regulate the biogenesis of certain microprotein-coding circRNAs remains poorly understood and warrant further investigation.

Cellular RNAs undergo extensive structural and chemical modifications, many of which are essential for their biogenesis and function regulation. Consequently, dysregulation of RNA modifications in cancer leads to aberrant expression of transcripts encoding microproteins. METTL14 plays a pivotal role in rewiring RNA behavior by introducing m⁶A modifications on target transcripts. In HCC, METTL14-mediated m⁶A modification on circSTX6 is found to suppress its expression, leading to the significant downregulation of the circSTX6-encoded microprotein, circSTX6-144aa. This observation is further supported by the negative correlation between METTL14 expression and the levels of circSTX6 and circSTX6-144aa in HCC tissues [167]. In contrast, in breast cancer (BC), METTL14 is shown to upregulate the expression of lncRNA LY6E-DT and its encoded microprotein MRP. Notably, the knockdown of IGF2BP1, a well-characterized m⁶A “reader” protein, reversed METTL14-induced upregulation of LY6E-DT, suggesting that METTL14-mediated m⁶A modifications promote LY6E-DT expression in an IGF2BP1-dependent manner [129]. These findings highlight the dual roles of METTL14 in regulating target RNA abundance, likely mediated by the recruitment of distinct m⁶A reader proteins to the modified transcripts. Furthermore, it would be of interest to investigate how other types of modifications contribute to the regulation of microprotein-coding RNAs.

During translation, eukaryotic initiation factor 3 (eIF3), a multiprotein complex composed of 13 distinct subunits (eIF3a-m), plays a critical role in both cap-dependent and cap-independent translation initiation. Notably, eIF3J has been shown to exert an inhibitory effect on the translation of a subset of circRNAs by impeding the binding of eIF3a and eIF3b to these circRNAs [185]. However, in HER2-positive BC, the transcriptional repression of eIF3J by NRF2 results in the enhanced translation of circ-β-TrCP peptide that confers trastuzumab resistance [183].

Certain E3 ubiquitin ligases facilitate the transfer of ubiquitin from ubiquitin carrier proteins to target microproteins, thereby promoting their degradation and reducing their abundance. For example, the microprotein circMAP3K4-455aa, encoded by circMAP3K4, is highly expressed in HCC. Its stability is regulated by the E3 ubiquitin ligase MIB1, which shortens its half-life through ubiquitination [43].

Biological function of sORF-encoded microproteins in cancer

Regulating gene transcription

Emerging studies have shown that certain sORF-encoded microproteins can regulate gene expression through directly interacting with transcription factors or their associated binding partners (Fig. 3A). For example, Xiang et al. reported that the microprotein PINT87aa, encoded by LINC-PINT, bound to the DNA-binding domain of the transcription factor FOXM1, effectively inhibiting its transcription activity in HCC cells [177]. Similarly, the microprotein CORO1C-47aa, encoded by hsa_circ_0000437, is shown to interact directly with the PAS-B domain of ARNT to prevent ARNT from binding to the transcription factor TACC3, ultimately suppressing VEGF transcription in endometrium tumor [171].

Fig. 3

Examples of the biological functions of microproteins. (A) Microproteins bind to transcription factors to regulate gene transcription; (B) Microproteins interact with splicing factors to regulate RNA splicing; (C) Microproteins interact with RNA-binding proteins to regulate the stability of target mRNAs; (D) Microproteins recruit translational factors to regulate protein translation; (E) Microproteins influence the stability of target proteins

Regulating RNA splicing

During cancer-associated transcriptome reprogramming, sORF-encoded microproteins have been shown to precisely regulate alternative splicing by interacting with key splicing factors, including members of the heterogeneous nuclear ribonucleoprotein (hnRNP) family, and the serine/arginine-rich splicing factor (SRSF) family (Fig. 3B). For instance, the microprotein PRDM16-DT, encoded by LINC00982, competitively interacts with hnRNP A2B1 to prevent the binding of hnRNP A2B1 to exon 9 of CHEK2 transcript, leading to the formation of the long isoform (L-CHEK2) while simultaneously suppressing the production of the short CHEK2 variant [138]. Similarly, another study demonstrated that the microprotein SRSP, encoded by LOC90024, enhanced the binding of SRSF3 to exon 3 of the Sp4 transcript. This enhanced interaction facilitates the selective inclusion of exon 3 to produce the oncogenic long isoform L-Sp4 protein while inhibiting the expression of the non-oncogenic short isoform S-Sp4 [143]. These findings highlight the ability of microproteins to modulate splicing patterns, thereby contributing to cancer progression.

Regulating mRNA stability

Certain RBPs, such as IGF2BP1 and HNRNPC, can specifically bind to the 3’-untranslated regions (3’-UTRs) of target mRNAs to regulate RNA stability [186,187,188]. Emerging evidence indicates that certain sORFs-encoded microproteins interact with these RBPs to influence the stability and fate of target mRNAs (Fig. 3C). For example, the microprotein RBRP, encoded by LINC00266-1, binds to IGF2BP1 and enhances its ability to recognize m⁶A modifications on c-Myc mRNA, leading to the increased stability and translation efficiency of c-Myc mRNA during tumorigenesis [140]. Similarly, the microprotein MRP, derived from the lncRNA LY6E-DT, interacts with HNRNPC to strengthen the binding of HNRNPC to epidermal growth factor receptor (EGFR) mRNA, enhancing the stability of EGFR mRNA and the expression of EGFR protein in BC [129].

Regulating protein translation

sORF-encoded microproteins can function as scaffold to modulate the assembly and function of translation-related complexes, therefore controlling specific translational programs in cancer (Fig. 3D). For example, the microprotein APPLE, encoded by lncRNA ASH1L-AS1, facilitates the interaction between PABPC1 (poly(A)-binding protein cytoplasmic 1) and eIF4G (eukaryotic initiation factor 4G), consequently promoting mRNA circularization and the assembly of the eIF4F initiation complex to support a specific pro-cancer translation program [108].

Regulating post-translational modification

Post-translational modifications (PTMs), including ubiquitination and phosphorylation, are critical regulatory mechanisms that govern protein abundance and function. The ubiquitination modification is a dynamic process governed by the bidirectional regulation of E3 ubiquitin ligases and deubiquitinases (e.g., USPs). sORF-encoded microproteins have been shown to interact with these enzymes, either facilitating or impeding their ability to recognize and modify specific substrates. For instance, the microprotein circZKSaa, derived from the circZKSCAN1, interacts with the E3 ubiquitin ligase FBXW7 to promote the ubiquitination and subsequent degradation of mTOR, thereby suppressing mTOR signaling [169]. In contrast, the microprotein N1DARP (Notch1 degradation-associated regulatory protein), encoded by LINC00261, disrupts the interaction between Notch1 intracellular domain (N1ICD) and the ubiquitin-specific peptidase 10 (USP10), leading to the polyubiquitination of N1ICD via K11 and K48 linkages and the inhibition of both canonical and non-canonical Notch1 signaling pathways [133]. Additionally, microproteins also regulate protein stability through phosphorylation modification. For example, the microprotein HEATR5B-881aa, encoded by circHEATR5B, directly interacts with the Jumonji C domain-containing protein (JMJD5) and reduces its stability by promoting phosphorylation at the S361 site [174]. However, the precise molecular mechanisms underlying this process remain to be fully elucidated. Given the importance of other PTMs, such as acetylation and methylation, in modulating protein stability and function, it would be interesting to explore whether and how microproteins influence protein homeostasis by regulating these modifications.

The roles of sORF-encoded microproteins in cancer

Emerging evidence has highlighted the oncogenic or tumor-suppressive roles of sORF-encoded microproteins in the onset and progression of various cancers. These microproteins contribute to cancer biology through distinct mechanisms, including: (i) the modulation of proliferative signaling pathways, (ii) the evasion of programmed cell death, (iii) the regulation of angiogenesis, (iv) the control of metastatic potential, and (v) the reprogramming of cellular metabolism.

Modulating proliferative signaling

c-Myc is a well-known oncogenic transcription factor that potently initiates and sustains tumor growth programs. Its expression is frequently upregulated in cancer. Recent studies reveal the involvement of sORF-encoded microproteins in regulating c-Myc abundance, thereby maintaining proliferative signaling for cancer growth. For instance, the tumor-associated peptide RBRP, encoded by LINC00266-1, interacts to IGFBP1 to enhance the interaction between IGFBP1 and c-Myc mRNA, increasing mRNA stability and driving colorectal cancer (CRC) progression [140]. Conversely, FBXW7-185aa, a microprotein encoded by the circular isoform of the E3 ligase FBXW7α transcript, shortens the half-life of c-Myc protein by accelerating FBXW7α-mediated c-Myc degradation [173]. Furthermore, a recent study identified a secretory 114-amino-acid microprotein, MPEP, encoded by an ORF within 5‘-UTR of MYC mRNA. MPEP functions as an agonistic ligand for the TRKB receptor tyrosine kinase to promote glioblastoma stem cell growth independently of MYC protein function [189]. These findings provide novel insights into biological significance of noncanonical ORFs and their encoded microproteins in cancer progression.

The MAPK/ERK signaling pathway, a critical pro-proliferative cascade frequently dysregulated in various cancers, can be modulated by multiple microproteins. In esophageal squamous cell carcinoma (ESCC), the microprotein circUBE4B-173aa has been identified as a direct interactor of MAPK1, enhancing its phosphorylation and thereby promoting the MAPK/ERK-mediated cell proliferation [168]. In addition, the MAPK gene encodes a microprotein, MAPK1-109aa, which is derived from its circular transcript of MAPK1. In gastric cancer (GC), MAPK1-109aa exerts a tumor suppressive effect by competitively binding to MEK1, the upstream kinase of MAPK1. This interaction inhibits the activation of MAPK1 and its downstream pro-proliferative signals [175].

Resisting cell death

Programmed cell death is an essential process for organisms to maintain internal homeostasis, whereas cancer cells have evolved diverse mechanisms to evade cell death programs including apoptosis and ferroptosis. A growing body of evidence highlights the important roles of cancer-associated microproteins in cell death (Fig. 4).

Fig. 4

Regulation of apoptosis and ferroptosis by microproteins

Apoptosis is a highly controlled process of cell death to eliminate damaged or abnormal cells in a caspase-dependent or caspase-independent ways. The expression of caspases, key executioners of apoptosis, can be regulated by specific microproteins. For example, the microprotein YY1BM, encoded by m⁶A-modified lncRNA LINC00278, has been shown to inhibit apoptosis in ESCC cells by downregulating caspase-3 expression. Mechanistically, YY1BM hinders the interaction between the transcription factor Yin Yang 1 (YY1) and the androgen receptor (AR), thereby suppressing the transcription of eukaryotic elongation factor 2 kinase (eEF2K). The downregulation of eEF2K relieves its inhibitory phosphorylation of eukaryotic elongation factor 2 (eEF2), ultimately promoting the translation and expression of caspase-3 in ESCC cells [41]. In caspase-independent apoptosis, apoptosis-inducing factor (AIF), a mitochondrial protein, undergoes N-terminal cleavage to form a soluble fragment upon apoptotic stimuli. Then the soluble AIF translocates to the nucleus to induce cell death [190]. Notably, the microprotein circMAP3K4-455aa, encoded by m⁶A-modified circMAP3K4 and upregulated in HCC, has been demonstrated to confer resistance to cisplatin-induced apoptosis by directly interacting with AIF to prevent its N-terminal cleavage and subsequent nuclear translocation, implicating microproteins in cisplatin resistance [43].

Mitophagy is a selective form of autophagy that removes damaged mitochondria to prevent mitochondrial outer membrane permeabilization. Wang et al. revealed that C-IGF1R, a microprotein encoded by circRNA cIGF1R, suppressed mitophagy by interacting with mitochondrial membrane protein VDAC1 to inhibit its ubiquitination, thus promoting apoptosis in EGFR-TKIs-resistant non-small cell lung cancer (NSCLC) cells [155]. Hence, these findings indicate that microproteins act as a molecular switch to regulate the transition from a drug-tolerant persister state to the apoptotic process.

Distinct from apoptosis, ferroptosis is an iron-dependent cell death process characterized by uncontrolled lipid peroxidation. Nuclear receptor coactivator 4 (NCOA4) is a selective cargo receptor mediating the autophagic degradation of ferritin, a process essential for iron homeostasis. Recently, Wang et al. revealed that the microprotein circFOXP1-231aa promoted ferroptosis in ICC cells by interacting with deubiquitinating protease OTUD4 to enhance the stability and expression of NCOA4 [160]. On the contrary, the transcription factor NRF2 acts as a protective regulator against ferroptosis by targeting components of the ferroptosis cascades. Employing machine learning, Jiang et al. predicted that microprotein encoded by LINC02381 modulated ferroptosis in glioblastoma through regulating NRF2 signaling pathway [147], providing a new strategy for experimental design to validate microprotein functions. However, the broader implications of microproteins in other forms of cell death, such as necrosis and pyroptosis, remains poorly understood.

Regulating angiogenesis

The rapid growth of tumors requires a heavy supply of both oxygen and nutrients, which necessitates the formation of new blood vessels from pre-existing vasculatures via angiogenesis, a process predominately regulated by growth factors including members of VEGF family [191]. Studies have shown the regulation of VEGF expression by microproteins. For example, the microprotein XBP1SBM, encoded by the lncRNA MLLT4-AS1, promotes VEGF transcription by enhancing the nuclear localization of transcription factor XBP1s, thereby driving angiogenesis in TNBC [146]. Conversely, other microproteins, such as ASRPS peptide encoded by LINC00908 and CORO1C-47aa encoded by hsa_circ_0000437, act as negative regulators of VEGF transcription. Specifically, ASRPS directly binds to the coiled-coil domain (CCD) of STAT3 and downregulates STAT3 phosphorylation and subsequent VEGF transcription [110]. Similarly, CORO1C-47aa competitively binds to ARNT and prevents its interaction with the transcription factor TACC3, thereby inhibiting TACC3-mediated VEGF activation [171]. Collectively, these findings underscore the diverse and context-dependent roles of microproteins in the regulation of tumor angiogenesis.

Regulating metastasis

The TGF-β/Smad signaling pathway is widely characterized as a pivotal pro-metastatic signaling cascade, primarily through enhancing epithelial-mesenchymal transition (EMT), a critical early step enabling primary tumors to gain invasive and metastatic capabilities. Recent studies reveal the roles of TGF-β/Smad-regulated microproteins in EMT and tumor metastasis. In GC, activation of the TGF-β/Smad pathway upregulates the expression of the circular RNA transcript of E-cadherin (circ-E-Cad) and its encoded microprotein C-E-Cad [152]. Elevated levels of C-E-Cad subsequently enhance the expression of transcription factors Snail and Slug, leading to the downregulation of E-cadherin and the upregulation of N-cadherin and vimentin, thereby driving EMT and promoting tumor metastasis (Fig. 5, Left). Conversely, TGF-β/SMAD signaling has been reported to downregulate the expression of LINC00665-encoded microprotein CIP2A-BP in TNBC metastasis. CIP2A-BP directly binds with CIP2A to replace PP2A’s B56γ subunit, thus releasing PP2A activity, which inhibits PI3K/AKT/NFκB pathway, resulting in decreased levels of MMP-2, MMP-9, and Snail and impairment of EMT process. Clinically, downregulation of CIP2A-BP in TNBC patients is associated with metastasis and poor survival in TNBC patients [115].

Fig. 5

Regulation of cancer metastasis by microproteins. Activation of Wnt/β-catenin and IL6/STAT3 pathways significantly promotes metastasis. TGF-β/Smad signaling pathways regulates microprotein expression, then markedly increases the expression of pro-EMT transcription factors such as Snail

The Wnt/β-catenin signaling represents another critical pathway to drive tumor metastasis. The stability of cytoplasmic β-catenin is tightly controlled by the APC/AXIN/GSK-3β complex and serves as a critical regulatory switch in Wnt activation. Within this context, microproteins derived from components of this pathway have emerged as key regulators of Wnt-mediated metastasis. The microprotein AXIN1-295aa, encoded by a circular transcript of AXIN1, is highly expressed in GC and promotes Wnt/β-catenin signaling and metastasis [150]. Mechanistically, AXIN1-295aa competitively interacts with APC to inhibit GSK3β-mediated degradation of β-catenin, allowing cytoplasmic β-catenin to accumulate and translocate into the nucleus, where it drives the transcriptional activation of genes associated with GC metastasis (Fig. 5, Middle). Similarly, the microprotein circβ-catenin-370aa, encoded by circular β-catenin, promotes Wnt/β-catenin signaling by competitively binding to GSK3β and protected β-catenin from GSK3β-induced degradation [170]. Additionally, the abundance of GSK3β itself can be regulated by microproteins. In TNBC, the microprotein EIF6-224aa interacts with MYH9 to prevent its degradation. This stabilization enhances MYH9-mediated destruction of GSK3β via the ubiquitin-proteasome pathway, thereby amplifying β-catenin signaling and facilitating TNBC metastasis [172].

The pro-inflammatory cytokine IL-6 initiates a signaling cascade by binds to its cognate receptor (IL-6R) to activate JAK kinase, thereby leading to the phosphorylation and nuclear translocation of STAT3 [192]. In ICC, IL-6 has been shown to upregulate the circRNA GGNBP2 and its encoded microprotein, cGGNBP2-184aa [154]. This microprotein directly interacts with STAT3 to facilitate STAT3^Tyr705 phosphorylation. Thus, IL-6/cGGNBP2-184aa/STAT3 forms a positive feedback loop to sustain constitutive activation of IL-6/STAT3 signaling and promote ICC metastasis (Fig. 5, Right), underscoring the complexity of cancer metastatic networks.

Reprogramming cellular metabolisms

Metabolic reprogramming represents a hallmark of cancer, characterized by the dynamic adjustments of metabolic pathways for nutrients (e.g., glucose, fatty acid) in cancer cells to meet their demands for rapid proliferation, enhanced survival and invasive capabilities [193]. Several sORF-encoded microproteins have been identified as effective regulators of cancer metabolic reprogramming (Fig. 6).

Fig. 6

Regulation of metabolic reprogramming by microproteins

The “Warburg effect” is a defined feature of metabolic reprogramming in cancer, characterized by the preferential reliance of cancer cells on glycolysis for energy production, even in the presence of sufficient oxygen [194]. Glycolysis, a multi-step enzymatic process involving the breakdown of glucose, is catalyzed by a series of enzymes, such as hexokinase 2 (HK2), phosphoglycerate kinase 1 (PGK1), pyruvate kinase M (PKM) and lactate dehydrogenase A (LDHA). The abundance of these enzymes can be regulated by microproteins through both transcriptional and post-transcriptional mechanisms. For example, Fang et al. revealed that the transcription factor Myeloid Zinc Finger 1 (MZF1) promoted the transcription of HK2 and PGK1, thereby promoting aerobic glycolysis in neuroblastoma, whereas a 21-amino acid microprotein derived from the 5‘-UTR of MZF1 mRNA (termed MZF1-uPEP) functioned as a negative regulator of the MZF1/HK2/PGK1 signaling axis. Specifically, MZF1-uPEP interacts with YY1 to inhibit its transactivation activity, leading to the downregulation of MZF1 and its downstream glycolytic targets [195]. LDHA catalyzes the conversion of pyruvate to lactate and is frequently upregulated in various cancers. In glioblastoma, the microprotein P4-135aa, encoded by the pseudogene MAPK6P4, promotes the translocation of KLF15 into nucleus, where KLF15 directly binds to the promoter region of LDHA and enhanced its transcription [196]. Similarly, the microprotein circMRCKα-227aa, encoded by circMRCKα, enhances LDHA transcription and glycolysis in HCC. Mechanistically, circMRCKα-227aa enhances USP22-mediated deubiquitinating and upregulation of HIF-1α, driving HIF-1α-induced LDHA transcription [163]. Alternative splicing of PKM pre-mRNA determines cellular metabolic phenotypes. The inclusion of exon9 generates the PKM1 isoform, which favors oxidative phosphorylation, while the exclusion of exon 9, mediated by the splicing factor hnRNP A1, produce the PKM2 isoform, which promotes aerobic glycolysis. In CRC, the microprotein HOXB-AS3, encoded by the lncRNA HOXB-AS3, competitively binds to hnRNP A1, and this interaction antagonized the hnRNP A1-mediated production of PKM2 and subsequent aerobic glycolysis [121]. These findings provide novel mechanistic insights in the complex regulatory networks governing glucose metabolic reprogramming in cancer.

The abundance of fatty acids is elaborately controlled by the dynamic equilibrium between their synthesis and oxidation, with the disruptions in this metabolic balance being a recurrent feature in cancer [197]. In ESCC, the activity of key anabolic enzymes, including fatty acid synthase (FASN) and stearoyl-CoA desaturase (SCD), is significantly elevated to boost fatty acid synthesis. Conversely, the microprotein pep-KDM4A-AS1, encoded by lncRNA KDM4A-AS1, has been shown to suppress fatty acid anabolism in ESCC by inhibiting the expression of FASN and SCD [137]. During fatty acid catabolism, carnitine palmitoyl transferase I (CPT1A) facilitates the transport of long-chain fatty acids into the mitochondria for β-oxidation [198]. Zhu et al. demonstrated that the microprotein pep-AKR1C2, encoded by exosomal lncAKR1C2, promoted CPT1A expression through reducing YAP phosphorylation and subsequently enhancing YAP-induced transcriptional activation of CPT1A, leading to an accelerated fatty acid oxidation and increased ATP production in GC [135]. However, the role of sORF-encoded microproteins in regulating other enzymes involved in fatty acid metabolism remains to be further elucidated.

Mitochondria are pivotal in metabolic reprogramming, primarily through their role in generating ATP via oxidative phosphorylation (OXPHOS), a fundamental process that can be assessed by oxygen consumption rate (OCR), membrane potential, and ATP production rate [199]. Several sORF-encoded microproteins have been identified to interact with mitochondrial proteins to regulate their functions. For example, the mitochondria-located microprotein SMIM26, encoded by LINC00493, interacts with acylglycerol kinase (AGK) and the glutathione transporter regulator SLC25A11. This interaction traps AGK within mitochondria and subsequently inhibits AGK-mediated AKT phosphorylation [141]. Thus, SMIM26 exerts a tumor suppressive role in ccRCC, and its low expression correlates with poor prognosis for ccRCC patients. Similarly, the microprotein miPEP133, derived from the primary miR-34a transcript, functions as an anti-cancer agent in mitochondria. It interacts with mitochondrial heat shock protein 70 kDa (HSPA9) to prevent the binding of HSPA9 to its partner proteins, leading to a decrease in mitochondrial membrane potential across multiple cancer types [7]. In contrast, the microprotein ASAP, encoded by LINC00467, promotes CRC progression by elevating ATP synthase activity. Specifically, ASAP interacts with ATP synthase subunits α and γ (ATP5A and ATP5C) to promote the assembly of ATP synthase, thereby increasing mitochondrial OCR and ATP production [109]. Hence, the interplay between microprotein and mitochondria highlights the complexity of energy metabolism in cancer cells.

Clinical significance of sORF-encoded microproteins in cancer

The expression profiles of sORF-encoded microproteins are closely associated with cancer progression and clinical outcomes, highlighting their potential as diagnostic and prognostic biomarkers. Furthermore, the important functions of microproteins during cancer pathogenesis underscore their therapeutic potential for cancer treatments.

sORF-encoded microproteins as potential diagnostic and prognostic biomarkers

Compared to normal tissues, sORF-encoded microproteins exhibit distinct expression patterns in cancer samples, making them ideal candidates as diagnostic biomarkers. For example, the microprotein MRP is found to be significantly upregulated in BC tissues. Its expression levels effectively distinguish patients with lymph node metastasis from those without, achieving an area under the curve (AUC) value of 0.7112, indicating its potential in predicting lymph node metastasis in BC [129]. Notably, microproteins encoded by ncRNAs are detectable not only in tumor tissues but also in various body fluids, highlighting their potential as non-invasive biomarkers. In a study by Pei et al., the levels of microproteins in the serum of NSCLC patients and healthy donors were compared, revealing that the microprotein ATMLP, encoded by lncRNA AFAP1-AS, is significantly upregulated in NSCLC patients. Importantly, ATMLP exhibits superior diagnostic efficiency (AUC = 0.852) compared to the conventional biomarker CEA (AUC = 0.746), underscoring its potential for early NSCLC detection [111]. Nevertheless, the diagnostic applicability of microproteins in other body fluids, such as saliva and urine, remains underexplored.

sORF-encoded microproteins also hold promise as prognostic markers for cancer progression and patient outcomes. In glioblastoma, high levels of microproteins such as MET404 and C-E-Cad are associated with poor overall survival (OS), suggesting their potential as prognostic indicators for glioblastoma [16, 153]. Similarly, high expression of other microproteins including ASAP [109], RBRP [140] and SRSP [143] is correlated with poor OS of CRC patients. Notably, RBRP and SRSP are further identified as independent prognostic factors associated with advanced clinical stages and higher histological grades [140, 143]. Conversely, high levels of microproteins such as ASRPS encoded by LINC00908 in TNBC and circATG4B-222aa encoded by circATG4B in CRC are correlated with improved OS [110, 156]. The prognostic significance of sORF-encoded microproteins has also been explored in other malignancies, including pancreatic ductal adenocarcinoma (PDAC) [17], HCC [167], ovarian cancer (OV) [36] and acute myeloid leukemia (AML) [108].

Although microproteins have emerged as promising biomarkers for cancer diagnosis and prognosis, their translation into clinical practice still faces several challenging issues, such as the identification of robust, reproducible and cancer type-specific biomarkers, along with the establishment of standardized detection protocols. Additionally, it is essential to develop more efficient, specific, and reliable analytical techniques for microprotein detection in clinical samples.

sORF-encoded microproteins as therapeutic targets

Given that sORF-encoded microproteins exhibit either oncogenic or tumor-suppressive functions, inhibiting oncogenic microproteins or restoring/enhancing the function of tumor-suppressive microproteins could be promising strategies for cancer therapy.

To achieve effective anti-cancer effects, many researches focus on restoring and enhancing the function of microproteins with the tumor suppressive role. For example, Zhai et al. found that LINC00261-endoced microprotein N1DARP inhibited tumor growth through regulating the USP10-Notch1 signaling axis [133]. Based on this discovery, the researchers developed SAH-mAH2-5, a cell-penetrating peptide designed to mimic the helical structure of N1DARP while exhibiting enhanced physicochemical stability. Moreover, SAH-mAH2-5 demonstrates potent anti-tumor activity against Notch1-activated PDAC cells, with minimal off-target and systemic toxicity [133]. Similarly, Dong et al. identified a conserved 9-amino acid peptide, ENSEP3, encoded by NEAT1, suppressed ESCC proliferation through inhibiting RAF expression and its downstream MAPK pathway. In patient-derived xenograft (PDX) models, synthetic ENSEP3 peptides specifically inhibits MAPK pathway activation, leading to significant suppression of ESCC tumor growth [116]. In addition, delivering plasmids encoding microproteins using nanocarriers can enhance the tumor-suppressive efficacy of microproteins. For example, Lei et al. identified that the circPDE5A-encoded microprotein PDE5A-500aa exerted the tumor-suppressive functions, and the delivery of its expression plasmid using a reduction-responsive nanoplatform (Meo-PEG-S–S-PLGA) effectively suppressed ESCC growth and metastasis both in vitro and in vivo [176].

Increasing studies have demonstrated that targeting oncogenic microproteins by shRNA or CRISPR/Cas9 system can effectively suppress tumor growth. For instance, Song et al. demonstrated that lentivirus-mediated delivery of shRNA targeting the oncogenic microprotein sPEP1, administered via intravenous injection, remarkably inhibited tumor growth [45]. Ge et al. reported that intratumoral injection of CRISPR/Cas9 vectors targeting the microprotein ASAP significantly inhibited CRC growth in PDX mouse models [109]. Similarly, an AAV-mediated Cas9/sgRNA delivery system is successfully applied for in vivo knockdown of the oncogenic HCP5-132aa microprotein, yielding substantial tumor growth inhibition in PDX models [119].

In addition to genetic modification approaches, targeting oncogenic microproteins with specific antibodies has also emerged as a promising strategy in cancer treatment. For example, Gao et al. reported that antibodies targeting the circ-E-Cad-encoded microprotein C-E-Cad effectively reduced STAT3 phosphorylation and inhibited the proliferation of glioma stem cell (GSC) [153]. Given that C-E-Cad activates EGFR signaling through its interaction with the CR2 domain of EGFR, the combination administration of C-E-Cad-targeting antibodies and EGFR -targeting antibodies dramatically suppressed tumor growth and improved survival rates in GSC-xenograft models. Similarly, the combination of MET404Ab, an antibody targeting circMET-encoded peptide MET404, with Onartuzumab, an FDA-approved MET antibody, demonstrated significant efficacy in inhibiting glioblastoma progression [16]. Owing to their small molecular weight and inherent immunogenic properties, microproteins may serve as potential antigens, enabling immune system to generate antibodies against tumors in vaccine therapy. For instance, Zeng et al. identified the RNF10 uPeptide, derived from the 5’-UTR of RNF10 mRNA, as an immunogenic antigen in a CT26 murine tumor model, where it was specifically recognized by CD8⁺ T cells to confer significant anti-tumor activity in mice. Notably, HLA-A2-restricted cytotoxic T lymphocytes (CTLs) isolated from pancreatic cancer patients recognizes the RNF10 uPeptide epitope (RLFGQQQRA) and then lysed HLA-A2⁺ pancreatic cancer cells expressing the RNF10 uPeptide [200]. Similarly, Kikuchi et al. identified the PVT1 peptide, encoded by lncRNA PVT1, as a novel tumor-specific antigen in CRC. The PVT1 peptide is presented by HLA class I molecules and recognized by CD8⁺ tumor-infiltrating lymphocytes (TILs) and peripheral blood mononuclear cells (PBMCs) from CRC patients, highlighting its potential application in CRC vaccine development [201].

Summary and perspectives

Advances in high-throughput sequencing and MS technologies have revealed a vast repertoire of previously undetected microproteins, significantly contributing to the complexity and diversity of proteomes. Many of these microproteins are encoded by sORFs located within ncRNAs or the UTRs of mRNAs. Despite substantial progress in this field, the identification of microproteins remain challenging due to their short length and low abundance. Thus, a concerted effort is required to optimize approaches such as MS and Ribo-seq to enhance detection sensitivity and specificity. Furthermore, the integration of multi-omics data holds promise for enabling a comprehensive characterization of the microprotein landscape. However, findings derived from such high-throughput screening approaches necessitate rigorous downstream validation to rule out false-positive results. Therefore, the development of highly specific and efficient antibodies against the microprotein of interest is essential to support experimental research and facilitate clinical validation.

Dysregulation of sORF-encoded microproteins has been increasingly implicated in various cancers. Although some microproteins have been demonstrated to play oncogenic or tumor-suppressive functions during tumor progression, the precise molecular mechanisms underlying their roles remain incompletely elucidated. Thus, it is essential to identify novel microproteins involved in tumorigenesis and to investigate their potential as therapeutic targets in cancer. Although multiple strategies have been explored to target microproteins for cancer therapy, proteolysis-targeting chimeras (PROTACs) has not yet been applied to oncogenic microproteins. Given the unique advantages of PROTACs, such as their ability to overcome drug resistance [202], developing highly specific and effective PROTAC degraders that selectively target cancer-associated microproteins while minimizing off-target effects represents a promising direction for future research.