CLIP-Seq (crosslinking and immunoprecipitation followed by sequencing) is a family of methods that freeze RNA-protein contacts in living cells and then read out exactly where on the transcriptome those contacts occurred. The “crosslinking” step is what makes the whole approach work: it welds an RNA-binding protein to the RNA it was touching at that instant, so the complex survives the harsh purification steps that follow. Over the past fifteen years, researchers have developed several distinct crosslinking strategies, each with different trade-offs in efficiency, resolution, and practicality, and the choice of method shapes nearly every downstream result.
Why Crosslinking Is the Defining Step
RNA-binding proteins hold onto their targets through weak, transient interactions. Without crosslinking, those contacts fall apart the moment you lyse a cell. UV light solves this by creating a covalent bond between a nucleotide base and an amino acid that happens to be sitting right on top of it. The reaction occurs at essentially zero distance, triggered by free-radical chemistry between the two molecules when they absorb UV photons.1Nature Communications. Nucleotide-amino acid π-stacking interactions initiate photo cross-linking in RNA-protein complexes Uridines and guanosines are the most reactive bases, though most amino acids can participate on the protein side. The bond is irreversible under normal conditions, which means the RNA fragment stays attached to the protein through immunoprecipitation, gel electrophoresis, and enzymatic digestion of everything else.
After crosslinking, the unprotected RNA is chewed away by RNases, leaving only the short stretch that was shielded by the protein. That fragment is then converted into a sequencing library, and the resulting reads tell you where on the genome the protein was bound. The entire protocol, from UV exposure to sequencing data, was first described as HITS-CLIP (high-throughput sequencing of RNA isolated by CLIP), and every variant since then modifies one or more of these core steps to improve resolution, efficiency, or ease of use.
Standard UV Crosslinking at 254 nm
The original approach uses short-wavelength ultraviolet light (254 nm, in the UV-C range) to crosslink natural, unmodified RNA directly to whichever protein is in contact. No special reagents need to be added to the cells beforehand, which is a major practical advantage: you can apply 254 nm crosslinking to any cell type or organism without worrying about metabolic labeling. The protocol covalently links transiently interacting RNA-protein complexes, then removes unprotected RNA by RNase digestion and detects the resulting complexes by gel analysis.2Europe PMC. RNA-protein UV-crosslinking Assay
The downside is that 254 nm crosslinking is relatively inefficient. Only a small fraction of protein-RNA contacts actually form a covalent bond, which means you need a lot of starting material. It also lacks specificity in a particular sense: because the UV energy is high enough to trigger crosslinks at many nucleotide-amino acid interfaces, including some that may be only fleetingly close, 254 nm exposure can generate a higher proportion of nonspecific hits. One proteomic comparison found that 254 nm UV identified more candidate RNA-binding proteins than longer wavelengths but a smaller fraction of them were annotated as known RNA-binding proteins, suggesting that the increased sensitivity came at the cost of specificity.3Molecular Cell. Proteomic Mapping of RNA-Binding Regions in Vivo
PAR-CLIP and Photoactivatable Nucleosides
PAR-CLIP (photoactivatable ribonucleoside-enhanced crosslinking and immunoprecipitation) takes a different approach. Instead of hitting natural RNA with 254 nm light, you feed cells a modified nucleoside, most commonly 4-thiouridine (4SU), which gets incorporated into newly made RNA. Then you crosslink with 365 nm UV-A light, a much gentler wavelength that specifically excites the thio group on 4SU rather than native bases.
This switch has two consequences. First, crosslinking efficiency goes up compared to conventional 254 nm CLIP. Second, and arguably more useful, the crosslinked 4SU creates a characteristic mutation during reverse transcription: a T-to-C change in the resulting cDNA. If you use the other common photoactivatable nucleoside, 6-thioguanosine, you instead see G-to-A transitions. These mutations pinpoint the exact crosslink site with single-nucleotide resolution and, critically, let you computationally separate real signal from the background of co-purifying RNA fragments that were never actually bound by the protein.4PubMed Central. PAR-CLIP: A Method for Transcriptome-Wide Identification of RNA Binding Protein Interaction Sites
The trade-off is that PAR-CLIP requires metabolic labeling, which limits it to cells that actively take up nucleosides in culture. Applying it to intact tissues is harder, though not impossible, as we will see below. The RNase digestion step also demands careful optimization: the type and concentration of RNase, the incubation time, and the temperature all determine how small the RNA fragments get. Fragments need to land in the 20 to 40 nucleotide range to map uniquely to the genome. Too little digestion leaves fragments too large for precise mapping; too much digestion destroys them before they can be sequenced.5PubMed Central. Optimization of PAR-CLIP for transcriptome-wide identification of binding sites of RNA-binding proteins
iCLIP and Nucleotide-Resolution Mapping
Individual-nucleotide resolution CLIP (iCLIP) tackles a problem that the original HITS-CLIP largely ignored. When reverse transcriptase copies an RNA fragment that still has a crosslinked amino acid residue stuck to it, the enzyme frequently stalls and falls off at the crosslink site instead of reading through. In standard CLIP, those truncated cDNAs are lost because the library preparation requires a full-length copy. iCLIP rescues them through an intramolecular cDNA circularization step that captures truncated cDNAs and turns the truncation point itself into a precise marker of where the protein was bound.6PubMed Central. iCLIP: protein-RNA interactions at nucleotide resolution
A direct comparison of CLIP and iCLIP for the Nova protein illustrated how this difference matters in practice. Standard CLIP data showed that cDNA deletions (another signature of crosslinks) had a preference for TTT motifs, while iCLIP’s truncation-based approach was more likely to identify clusters of YCAY motifs, which are the established primary binding sites for Nova.7PubMed Central. Analysis of CLIP and iCLIP methods for nucleotide-resolution studies of protein-RNA interactions In other words, the choice of method influenced which binding motif you would conclude the protein preferred, which is a reminder that the crosslinking and library preparation steps are not neutral observers of biology.
eCLIP and the Problem of Background Signal
Enhanced CLIP (eCLIP) focuses less on the crosslinking chemistry itself (it still uses 254 nm UV-C) and more on controlling for the noise that can contaminate any CLIP experiment. The key innovation is a paired “size-matched input” control. After cell lysis and RNA fragmentation, a small aliquot of the pre-immunoprecipitation sample is set aside and processed through the exact same library preparation steps as the immunoprecipitated sample, including the membrane size-selection step. This input control captures nonspecific background signal from the same size range, as well as biases introduced during ligation, reverse transcription, PCR, and gel migration.8PubMed Central. Robust transcriptome-wide discovery of RNA binding protein binding sites with enhanced CLIP (eCLIP)
Why does this matter so much? Work on PAR-CLIP showed that covalently crosslinked background binding is common, reproducible, and apparently universal across labs. Without correcting for it, many apparent “targets” of an RNA-binding protein are actually just abundant cellular RNAs that stick to everything. One study demonstrated that applying background correction to PAR-CLIP data for the protein Caprin1 revealed a previously unrecognized RNA recognition element that was invisible without the correction.9PubMed Central. Advancing the functional utility of PAR-CLIP by quantifying background binding to mRNAs and lncRNAs Computational tools like PureCLIP have also been developed to address this, using statistical models that explicitly account for nonspecific background and sequence biases when calling crosslink sites from single-nucleotide CLIP data.10PubMed Central. PureCLIP: capturing target-specific protein-RNA interaction footprints from single-nucleotide CLIP-seq data
How the Choice of Reverse Transcriptase Changes Your Data
One underappreciated variable in any CLIP experiment is the reverse transcriptase (RT) enzyme used to convert the crosslinked RNA fragments into cDNA. Different RTs behave very differently when they encounter a crosslinked nucleotide. Some stop cold at the crosslink (producing truncated cDNAs that iCLIP-style methods exploit). Others read through and introduce a deletion or mutation at that position. Still others do a mix of both, and the ratio depends on buffer conditions as well as the enzyme itself.
A systematic comparison found that the proportion of read-through reads was highly variable across different RTs, ranging from about 2% to 24% for the same protein. AffinityScript produced the lowest proportion of read-through reads, while three other tested enzymes generated more variable and generally higher proportions, in the 8 to 25% range.11Nucleic Acids Research. Monitored eCLIP: high accuracy mapping of RNA-protein interactions Separately, work on eCLIP showed that AffinityScript and TGIRT essentially never deleted the crosslinked base, while other enzymes showed variable deletion rates, meaning that the common computational approach of looking for deletions to find crosslink sites simply does not work with every RT.12PubMed Central. Variation in single-nucleotide sensitivity of eCLIP derived from reverse transcription conditions
Among RTs used in related RNA structure probing assays, the pattern holds: some enzymes like AMV and SuperScript III mostly register modifications as stops, while others like TGIRT-II are biased toward producing mutations that read through.13PubMed Central. Interpreting reverse transcriptase termination and mutation events for greater insight into the chemical probing of RNA The practical upshot is that your choice of RT determines whether you should analyze your data by looking for truncations, deletions, or mutations, and comparing datasets generated with different enzymes without accounting for this can lead to contradictory conclusions.
Chemical Crosslinking as a Complement to UV
UV crosslinking, whether at 254 nm or 365 nm, captures only contacts where a nucleotide base and an amino acid are in very close proximity and properly oriented. Chemical crosslinkers can reach contacts that UV misses. A recent study compared UV crosslinking to chemical crosslinking with agents like mechlorethamine (nitrogen mustard), diepoxybutane, and formaldehyde, analyzing the crosslinked sites by mass spectrometry. Chemical crosslinking increased crosslinked protein yields in living E. coli cells roughly four-fold compared to UV and generated different types of crosslink species, expanding the structural information accessible from a single experiment. Together, the UV and chemical approaches produced a comprehensive inventory of nucleic acid-protein crosslink sites at amino acid resolution for over 1,500 proteins.14Oxford Academic. Chemical crosslinking extends and complements UV crosslinking in analysis of RNA/DNA nucleic acid–protein interaction sites by mass spectrometry
Chemical crosslinkers are not yet routinely used in CLIP-Seq library preparation workflows, but their ability to capture a different subset of contacts makes them a natural complement for studies aiming at a complete picture of how a protein interfaces with RNA.
Comparing UV Wavelengths Side by Side
Not all UV is equal, and the wavelength you choose sets the balance between how many crosslinks you get and how trustworthy they are. The same proteomic study that flagged 254 nm specificity issues also tested 312 nm (UV-B) and 365 nm (UV-A, the PAR-CLIP wavelength). UV-B at 312 nm offered the best compromise between sensitivity and specificity: it crosslinked enough material to detect many RNA-binding proteins, and a high proportion of those candidates were already known to bind RNA. UV-A at 365 nm was the most accurate, meaning a very high fraction of identified proteins were true RNA binders, but at a considerable loss in sensitivity: it simply did not crosslink enough material to detect weaker or less abundant interactions.3Molecular Cell. Proteomic Mapping of RNA-Binding Regions in Vivo
This three-way comparison reveals a theme that runs through all CLIP-Seq method development: every gain in one dimension (resolution, specificity, ease of use) tends to come with a cost somewhere else (sensitivity, required input, computational complexity).
Faster Protocols and Non-Radioactive Detection
A persistent headache in classical CLIP is the use of radioactive isotopes to visualize RNA-protein complexes on membranes. Autoradiography works well but is slow, tightly regulated, and impractical in labs without isotope infrastructure. Infrared CLIP (irCLIP) replaces the radioactive label with an infrared-dye-conjugated, biotinylated adaptor. This adaptor ligates with the same efficiency as a standard one but reduces the time required for visualization from overnight autoradiography to minutes on an infrared imager, a speedup of more than ten- to a hundred-fold. The attomole sensitivity of the infrared adaptor also enables rapid quality-control checks at critical steps by simple dot blotting.15PubMed Central. irCLIP platform for efficient characterization of protein—RNA interactions
Building on irCLIP, a streamlined variant called quick-irCLIP combines elements of iCLIP and irCLIP into a protocol that can be completed in less than three days while still achieving single-nucleotide resolution.16MethodsX. Quick-irCLIP: Interrogating protein-RNA interactions using a rapid and simple cross-linking and immunoprecipitation technique These speed improvements matter because CLIP experiments are notoriously finicky, with many failure points. Being able to see your RNA-protein complexes quickly and iterate on conditions within a few days instead of a week or more makes optimization far less painful.
Taking CLIP Into Tissues
Most CLIP experiments are done on cells grown in a dish, where UV light can penetrate a monolayer easily. Intact tissues are a different story. The short wavelength of 254 nm UV barely penetrates past the surface of a tissue, which means crosslinking efficiency drops dramatically in anything thicker than a single cell layer.
One solution is to bring PAR-CLIP’s 365 nm crosslinking strategy into animals. In a study called viP-CLIP (in vivo PAR-CLIP), researchers gave mice repeated intraperitoneal injections of 4-thiouridine, which peaked in serum within 30 minutes and remained detectable in the blood for about two hours after a single dose. After repeated dosing over 15 hours, organs were harvested, flash-frozen, ground into powder, and then crosslinked with 365 nm UV light. Comparing this approach to standard 254 nm crosslinking, the 365 nm strategy improved RNA recovery from liver tissue several-fold, reaching levels comparable to what you would get from cells in culture. The 4SU dosing protocol showed no signs of liver toxicity, as measured by standard serum markers.17Nature Communications. In vivo PAR-CLIP (viP-CLIP) of liver TIAL1 unveils targets regulating cholesterol synthesis and secretion
This kind of tissue-level CLIP data has been applied to study RNA-binding proteins in liver, brain, and other organs where the biology cannot be faithfully recapitulated in cell culture.
Biological Discoveries Driven by CLIP-Seq
The different crosslinking methods have been applied across a wide range of biological questions, but the mapping of microRNA targeting stands out as one of the most impactful. MicroRNAs silence genes by guiding the Argonaute protein (Ago2) to complementary sites on messenger RNAs. CLIP-Seq of Ago2 directly reveals where on the transcriptome these silencing events happen, which is far more informative than computational prediction alone.
HITS-CLIP applied to human brain tissue uncovered roughly 7,000 stringent Ago2 binding sites that were highly enriched for conserved sequences corresponding to abundant brain microRNAs.18PubMed Central. Transcriptome-wide discovery of microRNA binding sites in human brain PAR-CLIP, meanwhile, was used in cell culture to map the binding sites and regulatory consequences for several intensely studied RNA-binding proteins and microRNA complexes, including PUM2, QKI, and all four human Argonaute proteins.19PubMed Central. Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP Specialized peak-calling tools have been benchmarked specifically on Ago2-CLIP data, with one method called miRBShunter validated on both HITS-CLIP and PAR-CLIP datasets in human stem cells to improve identification of microRNA-target interactions.20PubMed Central. From benchmarking HITS-CLIP peak detection programs to a new method for identification of miRNA-binding sites from Ago2-CLIP data
Beyond Crosslinking: Antibody-Free and Single-Cell Approaches
A newer generation of methods sidesteps both UV crosslinking and immunoprecipitation entirely. STAMP (surveying targets by APOBEC-mediated profiling) fuses an RNA-binding protein to a cytidine deaminase enzyme. When the fusion protein binds RNA in a living cell, the deaminase edits nearby cytidines to uridines, leaving a permanent mark in the RNA sequence that can be read out by standard RNA-Seq. Because no crosslinking or immunoprecipitation is needed, STAMP can work in single cells. When tested with the RBFOX2 protein, single-cell STAMP revealed consistent editing signal near known RBFOX2 binding sites across individual cells, demonstrating that the approach can define binding sites at single-cell resolution.21PubMed Central. Robust single-cell discovery of RNA targets of RNA binding proteins and ribosomes
A related strategy called TRIBE uses a fusion of an RNA-binding protein to the editing domain of ADAR, an enzyme that converts adenosine to inosine. This has been used to profile RNAs localized to stress granules, the phase-separated cytoplasmic structures that form under cellular stress. Because TRIBE works through enzymatic editing rather than crosslinking, it can interrogate compartment-specific RNA-protein interactions in settings where UV light cannot reach or where cell numbers are too low for conventional CLIP.22PubMed Central. Identification of the stress granule transcriptome via RNA-editing in single cells and in vivo
Long-Read Sequencing Meets CLIP
All the methods described so far produce short sequencing reads, typically under 150 nucleotides, which means you see each binding site in isolation. You cannot tell from a standard CLIP library whether two binding sites on the same transcript were occupied simultaneously on the same RNA molecule. A method called dirCLIP addresses this by combining UV crosslinking and immunoprecipitation with amplification-free direct nanopore long-read sequencing. In direct RNA mode, the nanopore detects perturbations in electrical current caused by amino acid adducts still covalently attached to the RNA. In direct cDNA mode, binding sites appear as mutations. Benchmarking dirCLIP with the SRSF3 protein showed over 75% concordance with short-read data while additionally revealing isoform-selective binding and enabling detection of co-occurring binding sites on single RNA molecules.23bioRxiv. dirCLIP profiles variant-specific RNA-protein interactions via nanopore long-read sequencing
This capacity to see multiple binding events on the same full-length transcript opens up questions that short-read CLIP simply cannot address, such as whether a splicing factor binds cooperatively to multiple introns of a pre-mRNA or whether binding at one site on a transcript excludes binding at another.
Computational Peak Calling and What It Means for Interpretation
Regardless of which crosslinking method you use, the sequencing reads have to be turned into a list of binding sites through a step called peak calling. This is where bioinformatics removes the nonspecific signal and identifies regions of the transcriptome with statistically enriched read coverage. Both protocol-specific peak callers (designed for the mutation signatures of PAR-CLIP or the truncation patterns of iCLIP) and more generic tools are available.24PubMed. Computational analysis of CLIP-seq data The choice of peak caller can affect the final binding-site list as much as the choice of crosslinking method, which is why benchmarking studies that compare multiple peak callers on the same dataset are important for the field. The existence of matched input controls, as in eCLIP, substantially improves peak calling because the algorithm has a direct measure of what the background looks like rather than relying on statistical assumptions.
Picking a Method for a New Project
With so many CLIP variants available, a practical question for any researcher starting a new project is which to choose. The decision usually comes down to a handful of considerations:
- Cell culture vs. tissue: Standard 254 nm CLIP and eCLIP work well for cultured cells. For intact tissues, PAR-CLIP with in vivo 4SU labeling or antibody-free approaches like STAMP and TRIBE are more practical.
- Resolution needs: If single-nucleotide resolution is essential, iCLIP, eCLIP with truncation analysis, or PAR-CLIP’s characteristic mutations are the strongest options.
- Input material: irCLIP and quick-irCLIP are designed for situations where starting material is limited, and their non-radioactive visualization makes optimization faster.
- Single-cell questions: STAMP and TRIBE are currently the only approaches that work at the single-cell level, though they require genetic construction of fusion proteins and do not use crosslinking.
- Isoform-level information: dirCLIP with nanopore sequencing is the emerging option for seeing binding events in the context of full-length transcripts.
No single method dominates across all these dimensions, and some of the most informative studies use two complementary approaches on the same biological system. The field is still actively evolving, with each new variant trying to solve a specific bottleneck rather than replace everything that came before.