An RNA-binding protein (RBP) is a protein that directly interacts with RNA, either transiently or as part of a stable molecular assembly. Such proteins help determine how RNA is produced, processed, transported, translated, stored, and degraded. They also support the structural and catalytic functions of RNA in complexes such as the ribosome. Together, RNA and its associated proteins form ribonucleoprotein particles, commonly abbreviated RNPs.
RNA-binding proteins operate at nearly every stage of gene expression. Some recognize short nucleotide sequences, others bind particular RNA shapes, and many combine several recognition mechanisms. Their activities connect information encoded in DNA with the changing requirements of the cell. The category therefore includes both specialized regulatory factors and components of fundamental cellular machinery.
The boundaries of the RNA-binding proteome depend partly on how binding is defined and measured. A protein may bind RNA as its principal biological activity, as one of several established functions, or only under particular conditions. Distinguishing direct contact from indirect association, and biochemical binding from a demonstrated physiological function, is central to understanding this diverse group.
Scope, discovery, and evolutionary diversity
RNA-binding proteins are a functional category rather than a single evolutionary family. They include structural proteins, RNA-processing factors, regulatory proteins, and enzymes whose substrates are RNA. Some remain associated with an RNA for much of its lifetime; others bind briefly during a processing reaction or regulatory transition. A ribosomal protein, a splicing regulator, and an RNA-degrading nuclease can therefore all qualify as RBPs despite having very different structures and activities.
Early understanding developed through studies of ribosomes, viral particles, and RNA-containing material in the cell nucleus. During the 1960s, biochemical work established that newly synthesized nuclear RNA associates with proteins in heterogeneous nuclear ribonucleoprotein particles, or hnRNPs. Subsequent purification and molecular characterization revealed families of proteins involved in packaging and processing nascent transcripts. The discovery of intervening sequences in eukaryotic genes in 1977 made RNA processing a central problem and helped drive the identification of proteins involved in RNA splicing.
The expansion from individual factors to whole RNA-binding proteomes depended on experimental advances. In 2012, studies led by Alfredo Castello and by Alexander Baltz independently combined ultraviolet crosslinking, purification of polyadenylated RNA, and protein identification by mass spectrometry. These approaches recovered established RBPs alongside numerous proteins not previously recognized as RNA-associated. A widely used 2014 census by Stefanie Gerstberger, Markus Hafner, and Thomas Tuschl catalogued 1,542 human RBPs using domain information and experimental evidence. This was an operational catalogue, not a final count: later methods sampled additional RNA classes and cellular conditions.
RNA-binding activities occur throughout cellular life. In bacteria, proteins such as Hfq and ProQ organize regulatory RNA interactions. Archaea possess RNA-binding proteins associated with translation, RNA maturation, and RNA-guided modification. In eukaryotes, nuclear compartmentalization and extensive transcript processing support large repertoires of proteins involved in splicing, export, and cytoplasmic regulation. Conserved molecular architectures coexist with lineage-specific expansions, particularly among proteins regulating development and cell differentiation.
Not all newly detected RNA contacts imply a newly discovered regulatory function. Abundant metabolic enzymes may bind RNA reproducibly, but the consequences must be established separately. A convincing functional assignment usually combines direct binding evidence with changes in RNA behavior after perturbation, supported by experiments that distinguish RNA binding from the protein’s other activities. This distinction prevents the experimental RNA-binding proteome from being treated as a list of equally well-characterized RNA regulators.
Molecular recognition and RNA-binding domains
RNA presents several kinds of molecular information: nucleotide sequence, local secondary structure, three-dimensional shape, chemical modifications, and the arrangement of exposed and buried surfaces. RBPs recognize these features through combinations of hydrogen bonds, electrostatic interactions, aromatic stacking, and shape complementarity. Positively charged residues can interact with the negatively charged phosphate backbone, but charge attraction alone rarely explains biological specificity. Recognition of particular bases and geometries is usually necessary to discriminate among the many RNAs present in a cell.
The RNA recognition motif (RRM) is one of the most widespread RNA-binding domains in eukaryotes. Typically about 80–90 amino acids long, it contains a four-stranded antiparallel beta sheet backed by two alpha helices. Conserved aromatic residues often contact exposed RNA bases on the beta-sheet surface. Nevertheless, different RRMs use additional loops, terminal extensions, or alternative surfaces, so their shared architecture does not imply a single binding mode. Proteins containing several RRMs can recognize longer sequences or combine contacts with separated RNA segments.
The K homology domain, usually about 70 amino acids long, commonly recognizes a short stretch of single-stranded RNA within a cleft. A conserved GXXG loop, where X denotes a variable residue, contributes to its characteristic binding surface. Other proteins use zinc-binding modules, Pumilio repeats, or double-stranded RNA-binding domains. Pumilio proteins illustrate a modular recognition principle: an array of repeats contacts successive bases, allowing sequence preference to be distributed along an extended protein surface. By contrast, a double-stranded RNA-binding domain typically recognizes the geometry of an A-form RNA duplex through contacts involving grooves and the sugar–phosphate backbone.
For a simple equilibrium interaction between protein and RNA ,
A lower dissociation constant, , indicates tighter binding under the specified conditions. Cellular occupancy, however, also depends on protein concentration, competing RNAs, accessibility, and assembly with other factors. Tight binding measured with an isolated RNA fragment does not necessarily mean that the same site is occupied inside a cell.
Many RBPs contain several binding modules and intrinsically disordered regions. These regions lack a single stable structure in isolation but can provide flexible contacts, regulatory modification sites, and interaction surfaces for partner proteins. RNA helicases add a dynamic component: by coupling nucleotide-dependent conformational changes to RNA or RNP remodeling, they alter which sequences and structures are accessible. Consequently, specificity is often an emergent property of an entire molecular assembly rather than a short sequence motif recognized by one isolated domain.
Nuclear RNA processing and assembly of export-competent transcripts
In eukaryotic cells, RBPs begin associating with many transcripts while RNA polymerase is still synthesizing them. Nascent RNA is therefore not simply released as an unstructured chain. Its folding and protein occupancy develop together, influenced by transcriptional speed, local sequence, and recruitment of processing factors. The resulting assemblies coordinate processing events and help distinguish properly matured transcripts from incomplete or defective products.
The spliceosome provides a prominent example of RNA–protein cooperation. This large, dynamic assembly removes introns from precursor messenger RNA and joins the surrounding exons. Small nuclear RNAs help recognize splice sites and form the catalytic center, while numerous proteins organize assembly, stabilize intermediates, and promote rearrangements. U1 small nuclear RNP participates in recognition of the 5′ splice site, whereas U2 small nuclear RNP engages the branch-point region. ATP-dependent remodeling factors help establish and rearrange these interactions before the chemical steps of splicing.
Regulatory RBPs influence alternative splicing, allowing a gene to generate different RNA isoforms. Serine/arginine-rich proteins frequently promote splice-site recognition, while members of the hnRNP families can enhance or repress particular events. Their effects depend strongly on binding position and surrounding factors; neither family is exclusively activating or repressing. Tissue-enriched regulators provide additional specificity. NOVA proteins preferentially recognize YCAY motifs, where Y is a pyrimidine, and participate in neuronal splicing programs. RBFOX proteins recognize UGCAUG-related sequences and regulate many tissue-dependent exons.
Formation of the messenger RNA 3′ end also depends on coordinated RNA recognition. During cleavage and polyadenylation, protein complexes recognize upstream and downstream sequence elements, cleave the precursor, and support addition of a poly(A) tail. WDR33 and CPSF30 contribute to recognition of the common AAUAAA signal, while CstF64 recognizes downstream U- or GU-rich regions. Selection among alternative processing sites changes the length of untranslated regions or, in some cases, the coding potential of the resulting transcript. These changes can alter which cytoplasmic regulators subsequently bind.
Mature messenger RNPs acquire export factors, including the NXF1–NXT1 receptor system, which supports passage through nuclear pores. Processing and export are coupled through adaptor complexes rather than occurring as entirely independent stages. The exon junction complex is deposited approximately 20–24 nucleotides upstream of many exon–exon junctions and can influence later RNA localization, translation, and surveillance. Meanwhile, nuclear retention and degradation pathways limit the escape of improperly processed RNA. Nuclear RBPs thus help establish both the sequence composition of a transcript and its molecular history, which can affect its fate after export.
Translation, RNA turnover, and quality control
Cytoplasmic RBPs regulate how efficiently an mRNA produces protein and how long that mRNA remains available. During translation initiation, eIF4E recognizes the 5′ cap of many eukaryotic mRNAs, while eIF4G acts as a scaffold linking several initiation factors. Cytoplasmic poly(A)-binding protein binds the poly(A) tail and interacts with initiation machinery. These connections can promote productive initiation and help coordinate events at opposite ends of the transcript, although mRNAs do not all adopt a single fixed circular structure.
Sequence-specific RBPs frequently act through untranslated regions. ELAVL1, also called HuR, binds many U-rich or AU-rich sequences and can influence RNA stability and translation in a context-dependent manner. ZFP36, or tristetraprolin, recognizes AU-rich elements in selected transcripts and recruits factors promoting their decay. Such regulatory sequences occur in mRNAs encoding cytokines, transcriptional regulators, and other proteins whose production must change rapidly. The same RNA can contain binding sites for competing or cooperating factors, making its output dependent on the composition of the surrounding regulatory network.
Iron metabolism provides a particularly well-defined regulatory mechanism. Under iron-limited conditions, iron-regulatory proteins bind iron-responsive elements in selected transcripts. Binding to an element in the 5′ untranslated region of ferritin mRNA inhibits translation initiation, reducing production of an iron-storage protein. Binding to multiple elements in the 3′ untranslated region of transferrin receptor 1 mRNA protects that transcript from degradation. When iron is abundant, IRP1 can acquire an iron–sulfur cluster and function as cytosolic aconitase, while IRP2 abundance is regulated through protein degradation. Similar RNA-binding logic therefore coordinates two different outputs according to cellular iron status.
Small regulatory RNAs act through protein partners rather than functioning as free guides. Argonaute proteins bind microRNAs and related small RNAs, presenting their sequences for target recognition. In animals, many microRNA–Argonaute complexes repress protein production and recruit machinery that shortens poly(A) tails and promotes mRNA decay. For appropriate guide–target combinations, catalytically competent Argonaute proteins can cleave the target RNA directly. Guide loading, pairing geometry, and associated proteins determine which outcome occurs.
RNA quality-control systems also depend on RBPs. Nonsense-mediated decay involves factors including UPF1 and can reduce transcripts containing premature termination codons, although it also regulates some normal transcripts. Other pathways respond to stalled ribosomes or defective RNA ends. Decay commonly proceeds through deadenylation, removal of the 5′ cap, and exonucleolytic degradation, but the order and relative contribution of these steps vary. RBPs connect recognition of a particular RNA state to the enzymes that execute its removal.
RNA localization, transport, and cytoplasmic assemblies
The location of an RNA can be as important as its abundance. Localizing an mRNA allows protein synthesis near a particular membrane, organelle, or cellular subregion, reducing the need to transport the completed protein. This is especially important in polarized cells and in neurons, whose axons and dendrites can extend far from the cell body. RBPs recognize localization elements in transcripts and connect them to transport machinery, anchoring sites, and translational regulators.
A well-studied example is beta-actin mRNA localization. IGF2BP1, historically called zipcode-binding protein 1 in this context, recognizes a localization element in the transcript’s 3′ untranslated region. Association with this protein helps couple transport to repression of premature translation. In suitable cellular contexts, phosphorylation of the regulator promotes relief of repression near the destination. The system illustrates how one RBP can integrate sequence recognition, movement, and local protein production rather than performing only one isolated function.
During transport, mRNAs commonly assemble into particles containing several proteins and sometimes multiple RNA molecules. These particles interact with motors moving along the cytoskeleton, including microtubules and actin filaments. Their composition can change during movement or upon arrival. Localization may also arise through diffusion followed by selective capture, or through preferential RNA stabilization at a destination. Consequently, the presence of an RNA-containing particle does not by itself demonstrate directed motor-driven transport.
Stress granules are cytoplasmic assemblies enriched in untranslated messenger RNPs and proteins such as G3BP1 and G3BP2. They often form when translation initiation is inhibited, including during some forms of oxidative or thermal stress. Their composition and physical properties depend on the stress, cell type, and time after induction. Processing bodies are distinct assemblies enriched in factors involved in translational repression, decapping, and RNA turnover. The two structures can exchange components and sometimes associate spatially, but they are not interchangeable, and neither is required for every instance of the process with which it is associated.
Many RNA-rich assemblies are studied as biomolecular condensates, in which multivalent molecular interactions concentrate selected components without a surrounding membrane. RNA can promote assembly by bridging proteins, but at other concentrations it can inhibit condensation by saturating available binding sites. Some assemblies show properties consistent with liquid–liquid phase separation; others are strongly influenced by scaffolds, active remodeling, or gel-like networks. A rounded appearance or recovery after photobleaching is therefore insufficient on its own to establish a phase-separation mechanism. The functional question is how assembly changes RNA access, reaction rates, or molecular availability under physiological conditions.
Bacterial, organellar, and viral RNA-binding systems
Bacteria possess extensive post-transcriptional regulation despite lacking a nucleus and the typical eukaryotic machinery for pre-mRNA processing. Hfq, first characterized as a host factor for bacteriophage Qβ RNA replication, forms a ring-shaped assembly that binds regulatory small RNAs and their target transcripts. Different surfaces recognize different RNA features, helping bring complementary RNAs together and influencing their stability. Pairing can block ribosome access, expose a translation-initiation region, or alter susceptibility to degradation, depending on the target.
ProQ represents another bacterial RNA-binding system with a distinct structural basis. In several bacterial species, it associates with structured RNAs and contributes to their stability or regulatory activity. CsrA illustrates direct protein-mediated regulation: it binds GGA-containing RNA motifs, often in exposed loops, and can control translation by occupying sites near ribosome-binding regions. In well-characterized Csr systems, regulatory RNAs containing multiple CsrA-binding sites sequester the protein, changing the fraction available to bind messenger RNAs. This creates a regulatory circuit based on competition for a limiting RBP.
Organelles have additional requirements arising from their specialized genomes and gene-expression systems. In the mitochondrion and chloroplast, nucleus-encoded RBPs contribute to RNA maturation, stability, editing, and translation. Pentatricopeptide repeat proteins are particularly abundant in plants. Their repeated helical modules recognize RNA in a largely modular fashion, and different family members specify particular transcripts or processing events. In plant organelles, some participate in site-specific cytidine-to-uridine editing through complexes that include, or are associated with, catalytic editing components.
RNA-guided defense systems also rely on RNA-binding proteins. In CRISPR systems, guide RNAs associate with Cas proteins and direct recognition of complementary nucleic acids. Depending on the system, the target may be DNA or RNA. The RNA is not merely a passive structural component: its sequence supplies targeting information, while the protein provides recognition checkpoints and catalytic activities. Related principles appear in eukaryotic RNA interference, although the protein families and evolutionary histories differ.
Viruses encode proteins that bind RNA to package genomes, regulate translation, assemble replication complexes, and influence host responses. Influenza A viral genomic RNA, for example, is organized into RNPs containing nucleoprotein and the viral polymerase complex. Host RBPs can either support viral replication or restrict it. Proteins involved in innate immunity, including RIG-I and MDA5, recognize particular features of viral or otherwise unusual RNA and initiate signaling. These systems demonstrate that RNA binding can function not only in gene expression but also in identifying molecular features associated with infection.
Regulation of RBP activity and combinatorial networks
An RBP’s effect depends on more than the presence of its preferred RNA sequence. Protein abundance, subcellular localization, accessibility of the binding site, competing factors, and the state of the RNA-processing machinery all influence the outcome. A site recognized strongly in an isolated biochemical assay may remain unoccupied in a cell because it is buried in RNA structure or covered by another protein. Conversely, cooperative assembly can stabilize interactions that would be weak on their own.
Post-translational modifications provide rapid control over these variables. Phosphorylation can change interaction surfaces or nuclear transport, while arginine methylation can alter interactions involving arginine-rich regions. Ubiquitin-dependent pathways regulate the abundance of some RBPs. The consequences are protein- and site-specific: adding a phosphate group does not uniformly activate or inhibit RNA binding. In splicing factors, modification can influence both interactions with other proteins and the balance between different nuclear compartments.
RNA chemistry creates another regulatory layer. N6-methyladenosine, abbreviated m6A, can alter local RNA behavior and recruit proteins containing YTH domains. Some proteins recognize the modified base through a defined binding pocket; others respond indirectly because modification changes RNA structure or competition among binding partners. The outcome can involve processing, stability, localization, or translation. It depends on the transcript and cellular setting rather than following a universal rule that a particular modification always stabilizes or destabilizes RNA.
Individual RBPs often bind many transcripts, while individual transcripts can bind many RBPs. This many-to-many organization produces combinatorial regulation. Two factors may compete for overlapping sequences, cooperate through direct protein contacts, or act sequentially as an RNA moves between cellular compartments. Collections of functionally related transcripts can share regulatory factors even when their genes occupy different genomic locations. This organization is sometimes described as a post-transcriptional regulon, emphasizing coordinated control at the RNA level.
Feedback is common in these networks. Some RBPs regulate processing of their own transcripts, limiting further protein production when they become abundant. Members of the serine/arginine-rich splicing-factor family can promote inclusion of sequences that direct their mRNAs toward nonsense-mediated decay. TDP-43 also participates in autoregulatory control of its RNA. Such mechanisms help maintain dosage, but their operation depends on the integrity of the surrounding processing and degradation systems. Network analysis must therefore distinguish a direct binding-dependent effect from secondary changes caused by altered cell state, stress, or the misregulation of another regulatory protein.
Experimental identification and mechanistic analysis
No single experiment establishes every aspect of an RBP’s function. Studies typically separate four questions: whether a protein directly contacts RNA, which RNAs it contacts, where on those RNAs it binds, and what consequence the interaction has. Different techniques answer different subsets of these questions. Agreement among biochemical, structural, and cellular evidence is especially important because RNA–protein interactions can be transient, condition-dependent, or disrupted during sample preparation.
Purified-component assays establish binding under controlled conditions. Electrophoretic mobility-shift assays detect changes in RNA migration during gel electrophoresis, while fluorescence anisotropy and calorimetric methods can quantify affinity or thermodynamic properties. Competition experiments test sequence preference, and systematic substitutions identify important nucleotides or amino acids. These assays can distinguish direct recognition from association mediated by another protein, but their simplified conditions may omit essential partners or RNA structures present in cells.
Immunoprecipitation-based methods recover RNAs associated with a selected protein. Conventional RNA immunoprecipitation can identify candidate targets, but without appropriate crosslinking it may retain indirect associations or allow interactions to change after lysis. Crosslinking and immunoprecipitation (CLIP) methods use ultraviolet irradiation to create covalent links at some direct RNA–protein contacts before purification. Variants including HITS-CLIP, iCLIP, PAR-CLIP, and eCLIP differ in labeling, library construction, and treatment of crosslink-associated sequence information.
CLIP data require careful interpretation. Ultraviolet crosslinking is not equally efficient for all nucleotide–amino-acid contacts, and highly abundant RNAs can generate substantial background. Fragmentation, antibody performance, and library preparation introduce additional biases. Input controls, biological replication, and analysis of enrichment are therefore important. A binding peak identifies evidence of contact, not necessarily a regulatory effect. Conversely, failure to detect a peak does not prove that binding never occurs.
Proteome-wide methods reverse the experimental perspective by identifying proteins associated with an RNA population. Oligo(dT)-based interactome capture enriches proteins crosslinked to polyadenylated RNAs, whereas other approaches recover broader RNA classes or isolate RNA–protein material through phase partitioning. Proteomics then identifies the recovered proteins. Structural methods—including X-ray crystallography, nuclear magnetic resonance, and cryo-electron microscopy—reveal contacts and conformational arrangements at complementary scales.
Functional analysis combines these measurements with depletion, targeted mutation, or genome editing. RNA sequencing measures transcript abundance and isoform changes, while ribosome profiling measures the distribution of ribosomes on transcripts. Rescue with a wild-type protein and comparison with carefully characterized binding-defective variants can test whether RNA recognition is required. A variant must also be checked for altered folding, abundance, or localization; otherwise, loss of function cannot be attributed specifically to loss of RNA binding.
Human disease and biotechnology
Human genetic and pathological studies show that disruption of RNA-binding proteins can affect many tissues, with particularly prominent examples in the nervous system. The relevant mechanisms include altered RNA recognition, defective localization, loss of normal processing activity, and abnormal protein assembly. These mechanisms can coexist. Visible protein accumulation is not by itself sufficient to explain every molecular defect, and disease-associated changes must be distinguished from normal, reversible RNP assembly.
TDP-43, encoded by TARDBP, and FUS are nuclear RBPs associated with subsets of amyotrophic lateral sclerosis and frontotemporal degeneration. TDP-43 pathology is present in most ALS cases, although important genetic subtypes show different pathology. Loss of normal nuclear TDP-43 activity can permit inclusion of normally suppressed cryptic exons. Well-characterized examples involve STMN2 and UNC13A, genes relevant to neuronal function. These findings connect a protein’s nuclear RNA-processing activity with specific changes in disease-associated transcripts without reducing the entire disease process to one target.
Fragile X syndrome usually results from expansion of a CGG repeat in the FMR1 gene, followed by silencing and loss of fragile X messenger ribonucleoprotein, FMRP. FMRP binds many neuronal RNAs and participates in control of their translation and localization. Spinal muscular atrophy most commonly results from loss or pathogenic alteration of SMN1. The SMN protein is central to assembly of small nuclear RNPs, illustrating how defects in RNP biogenesis can be clinically important even when the affected protein’s role is not simply recognition of a short messenger RNA sequence.
Altered RBP expression or activity also occurs in cancer. Mutations affecting the splicing factors SF3B1, SRSF2, and U2AF1 are recurrent in particular hematological malignancies and some solid tumors. These alterations change splice-site selection through mechanisms that differ among the affected proteins. Their existence does not imply that all splicing disruption has the same biological outcome, or that every change in an RBP observed in a tumor is a causal driver. Mechanistic studies distinguish recurrent mutations, downstream RNA alterations, and effects on cellular behavior.
In biotechnology, RNA-binding modules are used to visualize, isolate, and manipulate selected RNAs. The bacteriophage MS2 coat protein–RNA hairpin interaction is widely used to label engineered transcripts in living cells. Programmable Pumilio-derived proteins can be designed to recognize selected sequences, and RNA-targeting CRISPR systems provide another route to sequence-directed manipulation. Engineered RBPs can also recruit processing or regulatory activities to chosen transcripts. Their performance depends on target accessibility, off-target binding, expression level, and whether the added protein or RNA tag alters the behavior being measured.