3' Cleavage and Polyadenylation

Transcription by RNA Polymerase II does not end because the polymerase reads a DNA sequence meaning "stop." Instead, Pol II transcribes straight through the true 3′ end of the mature mRNA into "downstream" genomic sequence, sometimes for hundreds to thousands of nucleotides. Embedded in this nascent transcript is a cis-acting element called the poly(A) signal (PAS), most commonly the hexamer AAUAAA (or a close variant such as AUUAAA), positioned roughly 10–30 nucleotides upstream of the actual cleavage site. A multi-subunit protein assembly recognizes this signal together with flanking sequence elements, and endonucleolytically cleaves the RNA at the poly(A) site (usually after a CA dinucleotide). The polymerase, meanwhile, keeps transcribing downstream RNA that is now doomed. Immediately after cleavage, a template-independent enzyme, poly(A) polymerase (PAP), adds a run of roughly 200–250 adenosine residues to the new 3′-OH end, generating the poly(A) tail. Two features distinguish this process from ordinary enzymology and make it a favorite teaching example of co-transcriptional coupling. First, the entire process happens while transcription is still occurring — the cleavage/polyadenylation machinery rides along on the phosphorylated CTD (carboxy-terminal domain) of the large Pol II subunit, using the CTD as a co-transcriptional recruitment platform, a fact you should already recognize from the general logic of CTD-coupled RNA processing. Second, the act of cleavage is causally linked to transcription termination itself: once the RNA is cut, the polymerase — still transcribing the downstream, now-unprotected RNA — becomes a target for a 5′→3′ exonuclease that degrades that downstream RNA and eventually catches up to and evicts Pol II from the DNA template, a mechanism called the "torpedo" model. Together, 3′ cleavage, polyadenylation, and Pol II termination form one interlocking co-transcriptional module, executed by more than a dozen polypeptides working as a single machine, whose choices (which poly(A) site, how long the tail) shape mRNA export, stability, localization, and translation — themes picked up in the study of mRNA Export and Stability.

1. The cis-acting code. Three sequence elements, read as RNA, define a poly(A) site:

• The AAUAAA hexamer (or variants AUUAAA, AAUACA, etc., collectively covering >90% of human sites), located 10–30 nt upstream of the cleavage site. This is the single most important determinant; point mutations in it abolish cleavage almost completely.
• A U-rich or GU-rich downstream sequence element (DSE), located roughly 0–20 nt downstream of the cut site, which is bound cooperatively with the hexamer.
• Auxiliary upstream sequence elements (USEs), often U-rich, that enhance efficiency at many genes but are not universally required.
• The actual cleavage site itself usually lies at a CA dinucleotide, generating a new 3′-OH on the upstream fragment ending typically ...NAAUAAA...CA| (cut here).

2. Assembly of the 3′ processing machinery. At least four multi-subunit complexes and several accessory factors assemble co-transcriptionally on this RNA code:

CPSF-73 is the catalytic endonuclease — a metallo-β-lactamase-fold enzyme, homologous to the enzyme that cleaves histone mRNAs and to RNase J-family nucleases in bacteria — but it only cuts efficiently when correctly positioned by the surrounding complex reading the hexamer and DSE.

3. Endonucleolytic cleavage. Once CPSF and CstF are correctly positioned by the sequence code (with CFIm and CFIIm often stabilizing weak sites), CPSF-73 cleaves the phosphodiester backbone, typically at a CA dinucleotide 10–30 nt downstream of AAUAAA. This generates two RNA fragments:

• The upstream fragment, ending in a free 3′-OH — this becomes the mature mRNA 3′ end and is the substrate for polyadenylation.
• The downstream fragment, which retains the 5′-monophosphate created by the cut, remains attached to the still-transcribing polymerase, and is now uncapped and unprotected.

4. Poly(A) tail synthesis. PAP, recruited by CPSF (specifically CPSF-160/Fip1), adds AMP residues one at a time from ATP to the new 3′-OH, releasing pyrophosphate — a template-independent polymerization reaction chemically related to but mechanistically distinct from Pol II transcription. In the nucleus, PABPN1 binds the growing tail and both stimulates PAP processivity and acts as a "molecular ruler" that helps terminate tail growth at a characteristic length (~200–250 nt in mammals, shorter in yeast, ~70 nt). The finished mRNA — capped 5′ end plus new poly(A) 3′ end — is now competent for nuclear export, a step foreshadowed here and developed under mRNA export biology.

5. Coupling to Pol II termination — the torpedo mechanism. After cleavage, Pol II is still transcribing the downstream genomic DNA, producing the doomed downstream RNA fragment that lacks a 5′ cap. The 5′→3′ exoribonuclease Xrn2 (Rat1 in yeast) recognizes this uncapped 5′-monophosphate end and degrades the downstream RNA co-transcriptionally, chasing the elongating polymerase. When Xrn2 physically catches Pol II, it destabilizes the elongation complex — likely by removing RNA that helps hold Pol II on the template and/or by direct interaction with termination factors — causing Pol II to release from DNA. This is the "torpedo" model of termination. It works together with an "allosteric" component: passage through certain downstream sequences (often GC-poor, poly(A)-signal-like "sink" elements) slows Pol II elongation rate, giving the exonuclease time to catch up — so the modern consensus view fuses the once-competing allosteric and torpedo models into a single kinetic race.

A brief note on why the "chase" makes chemical sense. RNA polymerase elongation complexes are extraordinarily stable — stable enough to transcribe genes hundreds of kilobases long without falling off. Simply reaching a downstream sequence cannot, by itself, easily explain abrupt release. The torpedo model explains the puzzle: it isn't the DNA sequence that ejects Pol II, but a race between an exonuclease degrading the trailing RNA and Pol II's own forward movement, converted into termination once the enzyme physically collides with (or otherwise structurally destabilizes) the transcribing polymerase, often assisted by CTD-bound termination factors like Pcf11.

1. In vitro reconstitution proved cleavage and polyadenylation are separable, sequence-directed biochemical reactions (Moore & Sharp, 1985). Before the mechanism was understood, it was unclear whether 3′ end formation required cleavage at all, or whether poly(A) polymerase might simply extend the primary transcript at whatever 3′ end RNA Pol II happened to stop at. Using nuclear extracts from HeLa cells, Michael Moore and Phillip Sharp showed that synthetic pre-mRNA substrates containing the SV40 late poly(A) signal were specifically cleaved at a defined site even in the absence of ATP-dependent polyadenylation, generating a precise upstream fragment with a 3′-OH. When ATP was present, that same upstream fragment was then extended into a poly(A) tail. Critically, mutating the AAUAAA hexamer abolished both cleavage and polyadenylation, while the downstream fragment, once released, was rapidly degraded in the extract — foreshadowing the torpedo idea decades later. This established that (a) cleavage precedes and is mechanistically distinct from polyadenylation, (b) both reactions require the same upstream cis element, and (c) the whole process can be reconstituted biochemically outside cells, opening the door to purifying the responsible factors (CPSF, CstF, CFI, CFII, PAP were subsequently identified by biochemical fractionation of exactly these extracts).

2. UV crosslinking and immunoprecipitation identified which proteins contact which RNA elements. Using radiolabeled RNA substrates crosslinked by UV irradiation to nuclear extract proteins, followed by immunoprecipitation with specific antibodies, researchers (including Tom Wilusz, James Manley, and colleagues through the 1980s–90s) showed that CstF-64 crosslinks specifically to the downstream U/GU-rich element, while CPSF subunits crosslink to the AAUAAA hexamer — directly demonstrating the division of labor predicted by mutagenesis studies. Later structural work (cryo-EM structures of human CPSF core bound to RNA, ~2018–2019, from the Doudna, Tong, and Conti labs) visualized WDR33 and CPSF-30 reading out the AAUAAA sequence directly through base-specific contacts, converting decades of biochemical inference into an atomic-resolution picture.

3. Xrn2 knockdown/genetic experiments proved the torpedo model of termination (Kim, West, Proudfoot; Luo & Bentley; ~2004–2006). If Xrn2-mediated degradation of the downstream RNA is what triggers Pol II release, then removing Xrn2 activity should specifically impair termination without blocking cleavage. Torben Heick Jensen, Nick Proudfoot, and Steven West studied Pol II occupancy downstream of poly(A) sites using chromatin immunoprecipitation (ChIP) in human cells depleted of Xrn2 by RNAi, and in yeast strains carrying a catalytically dead allele of the ortholog Rat1 (the temperature-sensitive rat1-1 mutant, exploited earlier by Proudfoot's group in a landmark 2004 Nature paper by Kim, Krogan, Proudfoot, and colleagues). The result: with functional Xrn2/Rat1, Pol II ChIP signal fell off sharply within a few hundred base pairs past the poly(A) site, exactly as expected for prompt termination. With Xrn2/Rat1 inactivated, Pol II continued transcribing for kilobases beyond the normal termination point — a dramatic transcriptional readthrough phenotype — even though cleavage at the poly(A) site itself still occurred roughly normally. This is the key control: cleavage happened, but termination specifically failed, showing that it is the exonucleolytic degradation of the downstream RNA (not cleavage per se) that drives polymerase release. Tethering experiments artificially recruiting Rat1/Xrn2 to elongation complexes further showed this was sufficient to promote premature termination, closing the causal loop.

4. Depletion/mutation of AAUAAA in vivo confirmed the hexamer is necessary in living cells, not just extracts. Site-directed mutagenesis of AAUAAA in expressed reporter genes (classic experiments using SV40 and adenovirus late transcripts, and later genome-wide CRISPR-based poly(A)-site perturbation studies) consistently shows loss of correct 3′ processing, readthrough transcription, and use of cryptic downstream sites — confirming that the same sequence code identified biochemically operates as the actual determinant of gene expression in cells.

Example 1 — Predicting the effect of a poly(A) signal point mutation on a reporter transcript. A student engineers a plasmid expressing a reporter mRNA with the canonical hexamer `AAUAAA` located 22 nt upstream of a CA cleavage site, followed 15 nt further downstream by a `UGUGUGU` DSE. She introduces a single point mutation changing the hexamer to `AAGAAA`. Reasoning through the mechanism:

• WDR33/CPSF-30 recognize AAUAAA via direct, largely sequence-specific base contacts; `AAGAAA` disrupts this recognition (it is not one of the efficient near-cognate variants like AUUAAA).
• CPSF can no longer be stably positioned near the CA site even though CstF-64 can still bind the intact downstream U/GU element — but CstF alone binds only weakly and cannot recruit CPSF-73 efficiently without hexamer-dependent anchoring.
• Predicted outcome: greatly reduced cleavage efficiency at this site.
• Consequence for the polymerase: without efficient cleavage, the downstream fragment is not generated with an unprotected 5′ end, Xrn2 is not efficiently loaded, and transcriptional readthrough occurs — Pol II continues past the normal termination point until it reaches a downstream cryptic or fortuitous poly(A) signal (common in genomic readthrough experiments).
• Net prediction: longer, aberrant transcript, reduced levels of correctly terminated/polyadenylated mature mRNA, reduced protein output — matching the actual phenotype seen in classical AAUAAA mutagenesis experiments.

Example 2 — Alternative polyadenylation (APA) and 3′ UTR length: a quantitative reasoning problem. A gene has two poly(A) sites: a proximal site 300 nt after the stop codon (weak hexamer AUUAAA, weak DSE) and a distal site 4,000 nt after the stop codon (canonical AAUAAA, strong DSE, plus a strong CFIm-binding UGUA element). In actively proliferating cells (e.g., embryonic or cancer cells), CFIm25 (NUDT21) levels are often reduced. Reasoning:

• CFIm25 normally promotes usage of poly(A) sites that carry UGUA upstream elements — often the distal, stronger site in many genes — by helping loop the 3′ end processing complex onto weak or distant hexamers.
• If CFIm25 is depleted (a phenomenon documented in several cancers and in proliferative cell states), the relative advantage of the distal site is reduced.
• Prediction: cells shift toward preferential use of the proximal poly(A) site, producing transcripts with shorter 3′ UTRs.
• Downstream consequence: shorter 3′ UTRs often lose microRNA target sites and destabilizing AU-rich elements that would otherwise reside in the longer 3′ UTR — predicting increased mRNA stability and higher protein output from the shortened isoform.
• This exact logic (CFIm25 loss → widespread 3′ UTR shortening → oncogene stabilization) was demonstrated genome-wide using 3′-seq (a sequencing method that specifically maps poly(A) site usage) in cells depleted of CFIm25, linking a basic 3′-end-processing factor directly to a cancer-relevant regulatory switch.

Nick Proudfoot has spent much of his career asking a question that sounds almost too simple: how does RNA Polymerase II know when to stop? Trained in molecular biology in the UK during the era when RNA processing was just being resolved from transcription itself, Proudfoot became convinced that termination could not be understood in isolation from 3′ end cleavage — a view that was for years a minority position against models emphasizing purely sequence-triggered polymerase pausing. His laboratory's work on the yeast rat1-1 temperature-sensitive exonuclease mutant, and later on human Xrn2, provided the decisive genetic evidence for the torpedo model, converting what had been a plausible but untested idea (first floated in the 1990s by Lykke-Andersen, and others building on earlier proposals) into an experimentally established mechanism. Proudfoot's group also did foundational work mapping the sequence elements of the poly(A) signal itself in the 1980s, meaning he contributed to both halves of the puzzle this article covers — the cis-code and the termination mechanism it ultimately triggers. His broader research program has continued to explore how transcription termination integrates with chromatin, R-loop formation, and genome stability.

Human disease and genetic diagnosis. Point mutations that destroy or create poly(A) signals cause defined human genetic diseases — for example, certain α-thalassemia mutations disrupt the poly(A) signal of the α-globin gene, causing transcriptional readthrough, reduced correctly processed mRNA, and reduced globin protein. Clinical variant interpretation pipelines now routinely flag mutations in the AAUAAA hexamer of disease genes as likely pathogenic for exactly this mechanistic reason.

Cancer biology and alternative polyadenylation. Genome-wide 3′-seq studies show widespread 3′ UTR shortening in proliferating and cancer cells (via CFIm25/NUDT21 downregulation and other mechanisms), which removes microRNA target sites and stabilizes oncogene transcripts — making APA patterns an active area of cancer biomarker and therapeutic-target research.

Antiviral and antiparasitic drug development. Because CPSF-73's endonuclease active site is structurally related to, but distinguishable from, bacterial and some pathogen nucleases, and because viral infection often hijacks or disables host 3′ processing (e.g., influenza NS1 protein binds CPSF-30 to block host mRNA 3′ end formation, shutting down host gene expression while sparing viral mRNAs, which are capped by a distinct "cap-snatching" mechanism), this pathway is directly relevant to understanding viral pathogenesis and potential antiviral targets.

Biotechnology and synthetic biology. Engineered poly(A) signals of defined strength are routinely used in expression vectors, CRISPR guide constructs, and mRNA vaccine design (the mRNA vaccines for SARS-CoV-2 rely on carefully engineered, efficient 3′ UTRs and poly(A) tails, whose length and design measurably affect mRNA stability and translational output in vivo) — turning a basic mechanistic insight directly into a tool used in a global vaccination campaign.

Agricultural and model-organism research. Comparative studies of poly(A) signal usage across plant and animal genomes inform genome annotation pipelines (correctly identifying the 3′ boundary of a gene depends on correctly predicting or experimentally mapping its poly(A) site), which matters for crop genome engineering and functional annotation of newly sequenced species.

Ask the tutor: How does alternative polyadenylation interact with alternative splicing to generate protein diversity from a single gene?