RNA–Protein Interactions: an Intro to CLIP-seq

The question CLIP answers

An RNA-binding protein (RBP) recognizes specific RNAs and acts on them. CLIP-seq, short for cross-linking and immunoprecipitation followed by sequencing, answers: for a given RBP, which RNA positions does it physically bind, across the entire transcriptome? The output is a map of binding sites (peaks), which you then turn into a binding motif and a list of regulated transcripts.

Why crosslinking is the whole trick

You could just pull the protein down with an antibody and sequence whatever RNA comes along, which is the older method called RIP. The problem is that RNAs re-associate and stick non-specifically after you break the cell open, so RIP is noisy. CLIP fixes this with a covalent crosslink made before lysis: a pulse of UV light forms a permanent covalent bond between the protein and the exact RNA nucleotides it is touching at that instant (a "zero-distance" crosslink). Because the bond is covalent, the RNA stays attached through harsh, stringent washes that strip away everything not directly bound. That is what gives CLIP its specificity.

The core workflow

1. UV crosslink living cells, freezing each RBP onto the RNA it is bound to.
2. Lyse and lightly digest with RNase to trim the RNA down to a footprint around the protein.
3. Immunoprecipitate the RBP with a specific antibody, capturing the protein plus its crosslinked RNA.
4. Stringent wash to remove non-crosslinked, background RNA.
5. Recover the RNA, build a sequencing library (with adapters and unique molecular identifiers), and sequence.
6. Map and call peaks: enriched pile-ups mark binding sites.

The CLIP family: HITS-CLIP, PAR-CLIP, iCLIP, eCLIP

You will meet four names. They differ mainly in how they pin down the exact crosslink position and how they control for background.

Method	Key idea	How it pinpoints the site
HITS-CLIP	The original transcriptome-wide CLIP (also just "CLIP-seq").	Peak of reads over background.
PAR-CLIP	Cells are fed 4-thiouridine (4SU), which incorporates into RNA and crosslinks efficiently under 365 nm UV.	Reverse transcription reads the crosslinked 4SU as a characteristic T→C transition, a single-nucleotide signature of the contact point.
iCLIP	Exploits that reverse transcriptase usually truncates at the crosslinked nucleotide (the leftover peptide blocks it).	The read's truncation site marks the crosslink at single-nucleotide resolution.
eCLIP	ENCODE's robust, efficient version with a size-matched input control.	Truncation site, with enrichment measured against the paired input.

The trend across these is toward single-nucleotide resolution and proper controls. PAR-CLIP's T→C transitions and iCLIP/eCLIP's truncation sites both locate the contact far more precisely than a broad read pile-up.

From reads to binding sites: the analysis

A typical eCLIP-style pipeline looks like this. You do not need to memorize the tools, but you should recognize the shape:

# 1. Trim adapters and remove PCR duplicates using the UMIs
# 2. Align the footprint reads to the genome (e.g. STAR)
# 3. Collapse duplicates by UMI so each original molecule counts once
# 4. Call peaks and test enrichment AGAINST the size-matched input
$ clipper --species hg38 --bam IP.bam --outfile peaks.bed
# 5. Keep peaks reproducible across replicates, then find the motif

The end products are a BED file of high-confidence binding sites, an enriched sequence motif (the protein's preferred sequence or structure), and the set of transcripts it regulates, which you can then cross with expression data.

Check your understanding

Next in Track 4

Epitranscriptomics 101: m6A and RNA modifications →