Eukaryotic RNA Processing: Capping, Splicing, Polyadenylation

A newly made human transcript is a molecular liability. Fresh from RNA polymerase II, it has a naked 5′ triphosphate. This screams "foreign RNA" to innate immune sensors. Its coding sequence is interrupted by intron junk. This junk averages ten times the length of the exons it separates. Its 3′ end just stops. No signal tells the cell where the message ends. Before this pre-mRNA can leave the nucleus and be translated, it must be trusted. Three chemical operations must convert it into a stable, correctly bounded, intron-free messenger. These are: capping of the 5′ end, splicing out of introns, and cleavage/polyadenylation of the 3′ end. These three processes matter for a key reason. They are where eukaryotic gene expression earns its extra layer of information processing over bacteria. A bacterial mRNA is translated essentially as transcribed. A human pre-mRNA, by contrast, is a substrate for combinatorial exon selection. The same primary transcript can yield structurally and functionally distinct proteins. This depends on which exons are joined. This phenomenon is called alternative splicing. It touches the overwhelming majority of human multi-exon genes. Capping and polyadenylation are not mere end-decorations. They are checkpoints that couple RNA quality to transcriptional output and export competence. The cap and poly(A) tail together license translation. They do this by circularizing the mRNA through protein bridges. The big picture is that RNA processing is transcription-coupled chemistry. The C-terminal domain (CTD) of RNA polymerase II's largest subunit acts as a moving scaffold. This scaffold is phosphorylation-coded. It recruits capping enzymes, splicing factors, and cleavage/polyadenylation machinery. These are recruited in the correct temporal order as the polymerase transcribes. Splicing itself is not protein catalysis at its chemical core. It is RNA catalysis instead. It involves two sequential SN2 phosphoryl-transfer (transesterification) reactions. These are run by a ribonucleoprotein machine, the spliceosome. Its catalytic center is built from snRNA. This echoes the ribosome's RNA-based peptidyl transferase center. It also reinforces the RNA World hypothesis. This article works through the chemistry of each step in mechanistic detail. It covers the guanylyltransferase/methyltransferase cascade that builds the cap-0 and cap-1 structures. It covers the two-step transesterification mechanism of splicing with its branch-point lariat intermediate. It covers the sequence logic of 5′ splice site, branch point, and 3′ splice site recognition. It also covers the endonucleolytic cleavage/non-templated polyadenylation reaction that defines the mRNA 3′ end. Together these connect back to Promoters, Initiation, and Termination of Transcription. They also connect forward to The Genetic Code: Structure and Properties and Regulation of Gene Expression: Chemical Logic.

Key takeaways

The 5′ cap is a 7-methylguanosine joined to the transcript by an unusual 5′–5′ triphosphate bridge, added co-transcriptionally by a three-enzyme cascade (triphosphatase, guanylyltransferase, methyltransferase).
Splicing removes introns via two sequential SN2 transesterifications — no ATP is consumed in the chemical step itself, only in spliceosome assembly/remodeling — passing through a branched lariat intermediate.
The spliceosome is a ribozyme-based machine (U1, U2, U4, U5, U6 snRNPs) whose catalytic core is built from U2/U6 snRNA base-pairing, coordinating two catalytic Mg²⁺ ions exactly like group II self-splicing introns.
Three cis-elements — the 5′ splice site (GU), the branch point adenosine (A), and the 3′ splice site (AG) — are recognized with degenerate consensus sequences, explaining alternative splicing and the need for auxiliary enhancers/silencers.
3′ end formation is endonucleolytic cleavage downstream of an AAUAAA poly(A) signal, followed by template-independent addition of ~200 (nuclear)/shorter (cytoplasmic-stabilized) adenosines by poly(A) polymerase.
Cap and poly(A) tail are read by eIF4E and PABP respectively, which bridge to form a closed-loop mRNA that is the actual translationally competent unit.
Key idea: all three processing steps are transcription-coupled reactions organized in real time on the phosphorylated CTD "tracks" of RNA polymerase II, converting a fragile primary transcript into a stable, correctly spliced, translationally licensed mRNA.

Where this lives today

Splicing chemistry is now a direct drug target. Nusinersen (Spinraza) and risdiplam are approved therapies for spinal muscular atrophy that work by shifting the alternative splicing of the paralogous gene SMN2 — an antisense oligonucleotide (nusinersen) or a small molecule (risdiplam) biases inclusion of exon 7, restoring functional SMN protein levels from a gene that would otherwise splice it out. Antisense oligonucleotide "splice-switching" drugs of exactly this design class are now approved or in trials for Duchenne muscular dystrophy (exon-skipping drugs like eteplirsen), certain forms of familial dysautonomia, and various cancers with splicing-factor mutations (e.g., recurrent SF3B1 mutations in myelodysplastic syndrome and chronic lymphocytic leukemia, which perturb branch-point recognition and are themselves a target for small-molecule spliceosome modulators such as H3B-8800). Cap chemistry is central to mRNA vaccine and therapeutic manufacturing: both the Pfizer-BioNTech and Moderna COVID-19 vaccines require efficient, correctly structured cap-1 addition on in-vitro-transcribed mRNA, either enzymatically (using recombinant capping enzymes and 2′-O-methyltransferase) or co-transcriptionally via synthetic cap analogs (such as CleanCap reagents that incorporate a cap-1 structure directly during T7 transcription), because uncapped or cap-0 RNA is a strong trigger of innate immune sensors (RIG-I) and interferon responses that would destroy vaccine efficacy and induce inflammatory toxicity. Poly(A) biology underlies alternative polyadenylation (APA) as a widespread regulatory mechanism: many genes have multiple competing poly(A) sites, and shifting which one is used (often shortening 3′UTRs in rapidly proliferating cells, including many cancers) changes which microRNA target sites and RNA-binding-protein sites are retained in the 3′UTR, altering mRNA stability and translation without changing the encoded protein — a layer of regulation now studied genome-wide with 3′-end sequencing methods and increasingly implicated in tumor biology. Splicing and capping machinery are also front-line antiviral and diagnostic targets: many viruses (influenza via "cap-snatching," where viral polymerase steals host caps to prime its own mRNA synthesis) hijack host capping and splicing machinery, and spliceosome inhibitors (pladienolide B derivatives, targeting SF3B1) are in active oncology clinical development, exploiting cancer cells' heightened dependence on correct splicing fidelity under oncogenic stress.

:::faq[Why does alternative splicing matter for something like The Genetic Code: Structure and Properties?] The genetic code specifies how a linear mRNA sequence is translated into amino acids, but which linear mRNA sequence actually reaches the ribosome is determined upstream by which exons are included or excluded during splicing. A single gene can therefore produce multiple distinct mature mRNAs, each read by the same fixed genetic code but yielding structurally different proteins — different combinations of protein domains, regulatory motifs, or even premature stop codons that trigger nonsense-mediated decay. Understanding the code's triplet logic and understanding splicing's exon-selection logic are complementary: one explains how a given sequence is decoded, the other explains which sequence exists to decode in the first place.

:::faq[How do capping, splicing, and polyadenylation relate to Regulation of Gene Expression: Chemical Logic?] All three processes are additional layers of information flow beyond simple "on/off" transcriptional control, and each is itself regulated: alternative splicing choices respond to tissue-specific SR-protein and hnRNP concentrations, alternative polyadenylation site choice changes 3′UTR-dependent stability and localization, and capping efficiency/quality affects nuclear export and translational competence. Because RNA processing is coupled to the phosphorylation state of RNA polymerase II's CTD, transcriptional regulatory signals (which control CTD phosphorylation dynamics) propagate directly into processing outcomes — meaning the chemical logic of gene regulation extends well past the promoter into every step examined here.

Ask the tutor: How does alternative splicing generate different protein isoforms from the same gene, and how do SR proteins and hnRNPs tip the balance between competing splice sites?

The physical picture — capping, catalysis, and cleavage

Transcription by RNA polymerase II produces a pre-mRNA with a 5′-triphosphate end. As soon as roughly 20–30 nucleotides have emerged from the polymerase — while the CTD of Rpb1 is phosphorylated on Ser5 of its heptad repeat (Tyr1-Ser2-Pro3-Thr4-Ser5-Pro6-Ser7)ₙ — three enzymatic activities act in sequence. RNA triphosphatase removes the γ-phosphate from the 5′ end, leaving a diphosphate. Guanylyltransferase then attacks this diphosphate with GTP, forming the signature 5′–5′ triphosphate linkage (a phosphoamide-like linkage chemically unlike any normal phosphodiester in the chain) and releasing pyrophosphate. Finally, guanine-N7 methyltransferase uses S-adenosylmethionine (SAM) to methylate N7 of the newly added guanine, producing the cap-0 structure, m⁷GpppN. In higher eukaryotes a further 2′-O-methylation of the first (and sometimes second) transcribed nucleotide's ribose produces cap-1 (and cap-2), which helps the cell's own mRNA evade cytoplasmic RNA sensors (RIG-I, MDA5) that are tuned to detect uncapped or improperly capped viral RNA. Splicing removes introns from the pre-mRNA by a mechanism chemically identical in logic to two consecutive nucleophilic substitutions at phosphorus. The spliceosome — composed of five small nuclear ribonucleoproteins (U1, U2, U4, U5, U6 snRNPs) plus over a hundred accessory proteins — assembles de novo on each intron, recognizes three short degenerate sequence elements (5′ splice site, branch point, 3′ splice site), and catalyzes: (1) attack of the branch-point adenosine's 2′-OH on the 5′ splice site phosphate, generating a lariat intermediate with an unusual 2′–5′ phosphodiester branch, and (2) attack of the freed 5′ exon's 3′-OH on the 3′ splice site phosphate, joining the exons and releasing the intron as a lariat. Both steps are transesterifications — the total number of phosphodiester bonds is conserved, no water is consumed, and no external energy source is chemically required, though ATP-dependent RNA helicases remodel the spliceosome's RNA-RNA and RNA-protein contacts between and within these steps. 3′-end formation is mechanistically distinct: it is an endonucleolytic cleavage of the nascent transcript roughly 10–30 nucleotides downstream of the hexanucleotide signal AAUAAA (recognized by the cleavage and polyadenylation specificity factor, CPSF), followed by template-independent, distributive addition of adenosine monophosphates by poly(A) polymerase using ATP as the nucleotide donor, releasing pyrophosphate at each step — mechanistically a phosphoryl transfer chemically related to (but not templated like) normal RNA synthesis.

Figure

The 5′ cap: an inverted 5′–5′ triphosphate bridge with N7-methylguanine

Figure

Spliceosome two-step transesterification: branch-point attack then exon ligation

The core ideas

The chemistry of splicing is best summarized as two in-line nucleophilic substitutions at phosphorus, each proceeding with inversion of configuration at the electrophilic phosphate — the diagnostic signature of a direct, single-step SN2 mechanism rather than a double-displacement (retention) pathway.

\text{Step 1: } \; \text{Branch-A(2′-OH)} + \text{p}_{5'SS}\text{-Exon2} \;\longrightarrow\; \text{Branch-A(2′-5′-p)-Exon1(free 3′-OH)} + \text{Exon2(free 5′)}

where: the branch-point adenosine's 2′-OH is the nucleophile, the phosphate at the exon1/intron boundary (5′ splice site) is the electrophile, and exon 1 is released bearing a free 3′-OH while the intron becomes covalently looped onto itself.

\text{Step 2: } \; \text{Exon1(3′-OH)} + \text{p}_{3'SS}\text{-Exon2} \;\longrightarrow\; \text{Exon1–Exon2 (spliced mRNA)} + \text{intron lariat}

where: exon 1's newly liberated 3′-OH is now the nucleophile, the phosphate at the intron/exon2 boundary (3′ splice site) is the electrophile, and the products are the ligated mRNA plus the excised lariat intron (later linearized by a debranching enzyme and degraded). The overall stoichiometry of splicing is bond-neutral:

\Delta G^{\circ\prime}_{\text{splicing (chemical step)}} \approx 0 \text{ kcal/mol}

where: each transesterification breaks one phosphodiester bond and forms another (a phosphate ester for a phosphate ester), so no net high-energy bond is consumed in the chemical step itself — the reaction is thermoneutral to first approximation, and the large energetic cost of splicing is paid instead in ATP-dependent conformational proofreading and remodeling by DEAD/DEAH-box helicases (Prp5, Prp28, Brr2, Prp16, Prp22, Prp43) during spliceosome assembly and catalytic activation. Splice site strength can be quantified as a log-odds match to the consensus, an information-content view of recognition:

S = \sum_{i} \log_2\!\left(\frac{f_i(b)}{p_0(b)}\right)

where: $f_i(b)$ is the observed frequency of base $b$ at position $i$ of real splice sites, $p_0(b)$ is the background base frequency (~0.25 for each base), and $S$ (in bits) measures how strongly a given sequence resembles the consensus 5′ splice site (MAG│GURAGU, where │ marks the exon–intron boundary) or 3′ splice site (Y-rich tract–NCAG│). For the cap methylation step, the relevant relationship is the group-transfer thermodynamics of SAM as methyl donor:

\text{SAM} + \text{GpppN-RNA} \longrightarrow \text{SAH} + \text{m}^7\text{GpppN-RNA}

where: SAM (S-adenosylmethionine) donates a methyl group to N7 of the terminal guanine via nucleophilic attack of the guanine N7 lone pair on the SAM methyl carbon (SN2, with the sulfonium center as leaving-group-activator), producing S-adenosylhomocysteine (SAH) and the methylated cap; this reaction is strongly exergonic (~ -10 to -12 kcal/mol under physiological SAM/SAH ratios) because the sulfonium ion in SAM is an exceptional leaving group, driving methyl transfer essentially to completion. For 3′-end formation, poly(A) addition is a repeated, template-independent phosphoryl transfer:

(\text{RNA})_n\text{-3′-OH} + \text{ATP} \longrightarrow (\text{RNA})_{n+1}\text{-3′-OH-AMP} + \text{PP}_i

where: poly(A) polymerase uses the same two-metal-ion mechanism as other nucleotidyl transferases (two Mg²⁺ ions positioned by conserved aspartates activate the 3′-OH nucleophile and stabilize the pentacovalent transition state and pyrophosphate leaving group), and pyrophosphate hydrolysis by cellular pyrophosphatase pulls the reaction forward, exactly as it does in DNA and RNA polymerization.

Derivation AMechanism of the first transesterification: branch-point attack at the 5′ splice site

Goal. Show explicitly, step by step, how the spliceosome catalyzes attack of the branch-point adenosine 2′-OH on the phosphodiester at the exon1–intron junction, and why the reaction is a clean in-line SN2 substitution at phosphorus rather than a two-step addition–elimination through a stable pentacovalent intermediate.

Step 1 — Substrate positioning. After U1 snRNP base-pairs with the 5′ splice site (5′ end of intron, consensus GURAGU) via Watson–Crick pairing to the U1 snRNA 5′ end, and U2 snRNP base-pairs with the branch-point sequence (consensus YNYURAC, where the branch A is bulged out of the U2 snRNA duplex), the U4/U6•U5 tri-snRNP joins. ATP-dependent remodeling by the helicase Brr2 unwinds U4 from U6, allowing U6 to displace U1 at the 5′ splice site and to base-pair extensively with U2 snRNA. This U2/U6 helix, together with U6's internal stem-loop, folds into the catalytic core that positions two catalytic Mg²⁺ ions (metal ion 1 and metal ion 2) in a fold structurally and mechanistically homologous to self-splicing group II introns.

Step 2 — Activation of the nucleophile. The branch-point adenosine, bulged out of its Watson–Crick pairing with U2 snRNA, presents its ribose 2′-OH into the active site. Catalytic metal ion 1 (Mg²⁺) coordinates the 2′-oxygen, lowering its pKa and stabilizing developing negative charge as a proton is lost (to solvent or to a general base), generating a reactive 2′-alkoxide nucleophile. This is the same "metal-assisted deprotonation" strategy nucleases and polymerases use to generate reactive oxyanions without requiring extremes of pH.

Step 3 — In-line nucleophilic attack. The activated 2′-O⁻ attacks the phosphorus atom of the phosphodiester bond linking the last nucleotide of exon 1 to the first nucleotide of the intron (the 5′ splice site scissile phosphate), approaching from the face opposite the 3′-oxygen leaving group (in-line, 180° O–P–O geometry). This produces a trigonal bipyramidal transition state with partial bonds to both the incoming 2′-oxygen nucleophile and the outgoing 3′-oxygen leaving group, both apical, and the three non-bridging substituents equatorial.

Step 4 — Leaving-group stabilization and bond resolution. Catalytic metal ion 2 (Mg²⁺) coordinates the developing negative charge on the 5′-oxygen leaving group of the intron (the exon 1 3′-oxygen becomes the leaving group in this bond — precisely, the phosphate being attacked was originally the 5′-phosphate of the first intron nucleotide, so exon 1's 3′-oxygen leaves as the new bond to the intron's 2′-branch forms). As the transition state collapses, the P–O bond to the old 3′-exon linkage breaks heterolytically, and a new phosphodiester bond forms between the branch-point 2′-oxygen and the intron's 5′-phosphate.

Step 5 — Product distribution. The result is a branched "lariat" intermediate: the intron is now covalently attached to itself through an unusual 2′–5′ phosphodiester bond at the branch adenosine (which is now trisubstituted — 3′–5′ to its upstream neighbor within the intron, 2′–5′ to the intron's own 5′ end), while exon 1 is released bearing a free 3′-OH, poised as the nucleophile for the second transesterification. Because the mechanism proceeds through a single pentacoordinate transition state rather than a stable phosphorane intermediate, stereochemistry at phosphorus is inverted — the diagnostic experimental signature (using phosphorothioate-substituted, chirally pure substrates) that confirmed this is a direct, single in-line displacement, just as in the second step of splicing and in the analogous self-splicing group II intron reaction. The net conservation of phosphodiester bonds means:

\Delta n_{\text{P–O bonds broken}} = \Delta n_{\text{P–O bonds formed}} = 1

confirming no external phosphoanhydride bond energy is required for this chemical step; the spliceosome's role is entirely to organize substrate geometry, activate the nucleophile via metal coordination, and stabilize the transition state and leaving group — true catalysis by proximity, orientation, and electrostatics, not by driving an otherwise unfavorable reaction with an energy source.

Go deeper · Derivation B · 2 minEndonucleolytic cleavage and template-independent polyadenylation at the 3′ end

Derivation BEndonucleolytic cleavage and template-independent polyadenylation at the 3′ end

Goal. Derive how the poly(A) signal is converted into a defined, stable 3′ end, distinguishing the two chemically separate steps — site-specific endonucleolytic cleavage, and distributive, non-templated nucleotidyl transfer — and show why the poly(A) tail length is kinetically, not thermodynamically, determined.

Step 1 — Recognition of the cis-elements. As RNA polymerase II transcribes through the poly(A) signal region, the hexanucleotide AAUAAA (10–30 nt upstream of the eventual cleavage site) is bound by CPSF (cleavage and polyadenylation specificity factor, specifically its CPSF-30/WDR33/Fip1 subcomplex reading the hexamer directly). A U/GU-rich downstream sequence element (DSE) is bound cooperatively by CstF (cleavage stimulation factor) via its CstF-64 subunit. These two multiprotein complexes, bridged by additional cleavage factors (CFIm, CFIIm), loop the RNA so that the actual cleavage site — typically a CA dinucleotide — is positioned in the active site of the endonuclease CPSF-73, a metallo-β-lactamase-fold enzyme.

Step 2 — Endonucleolytic hydrolysis. Unlike the transesterifications of splicing, 3′-end cleavage is a genuine hydrolysis: CPSF-73 uses two Zn²⁺ ions to activate a water molecule as nucleophile, which attacks the scissile phosphate in-line with the 3′-oxygen leaving group of the upstream cleavage product, releasing a new mRNA 3′-OH end and a downstream fragment bearing a 5′-phosphate. The downstream fragment (still attached to the elongating polymerase) is rapidly degraded 5′→3′ by the exonuclease Xrn2, which is proposed to also help terminate transcription by "torpedoing" and dislodging the polymerase — mechanistically linking 3′-end processing directly to transcription termination.

Step 3 — Recruitment of poly(A) polymerase. The free 3′-OH generated by cleavage is immediately bound by poly(A) polymerase (PAP), tethered to the CPSF complex, together with the nuclear poly(A)-binding protein (PABPN1 in mammals). PAP requires no template; it selects ATP specifically (over other NTPs, based on active-site shape complementarity to adenine) and catalyzes nucleotidyl transfer using the identical two-metal-ion mechanism used by all template-dependent and template-independent polymerases.

Step 4 — Two-metal-ion nucleotidyl transfer, iterated. For each adenosine added: metal ion A coordinates and deprotonates the RNA 3′-OH, generating a 3′-alkoxide nucleophile; this alkoxide attacks the α-phosphate of incoming ATP in-line, displacing pyrophosphate (stabilized and charge-shielded by metal ion B) as the leaving group, through a pentacovalent transition state exactly analogous to the phosphotransfer step of DNA/RNA polymerases and to Step 3 of the splicing mechanism above. Because there is no template to specify a stop position, iteration continues stochastically.

Step 5 — Kinetic control of tail length. Initially PAP alone works slowly and distributively (dissociating after each addition). Binding of PABPN1 to the growing tail dramatically stimulates PAP's processivity, but PABPN1 molecules occupy roughly 25–27 nucleotides of poly(A) each; once approximately 200–250 adenosines have been added (about 10 PABPN1 footprints), the complex reaches a size at which further stimulation saturates and polyadenylation slows sharply. This gives:

L_{\text{poly(A)}}^{\text{nuclear}} \approx 200\text{–}250 \text{ nt}

where: $L_{\text{poly(A)}}^{\text{nuclear}}$ is set by a kinetic/steric "molecular ruler" (PABPN1 occupancy translating into a processivity cutoff), not by a thermodynamic equilibrium — the reaction is essentially irreversible under cellular conditions (driven by pyrophosphate hydrolysis) and would run indefinitely without this counting mechanism. In the cytoplasm, tails are subsequently shortened by deadenylases (e.g., the CCR4-NOT and PAN2-PAN3 complexes) as part of normal mRNA turnover and translational control, so the functional cytoplasmic tail length observed at any moment is typically shorter and is itself a regulatory readout of mRNA age and stability.

Worked examples

Example 1 — Foundational: energetics of the SN2 phosphoryl transfer in splicing step 1. Consider the first transesterification in isolation. The bond broken is a 3′–5′ phosphodiester (exon1–intron junction); the bond formed is a 2′–5′ phosphodiester (branch A to intron 5′ end). Both are phosphodiester linkages of very similar intrinsic bond energy (phosphodiester hydrolysis free energies are typically quoted around -5 to -6 kcal/mol per bond relative to full hydrolysis to free phosphate and two free hydroxyls, but here we are not hydrolyzing — we are transesterifying, exchanging one phosphodiester for another).

\Delta G^{\circ\prime}_{\text{step 1}} = \Delta G^{\circ\prime}_{\text{bond formed (2′–5′ phosphodiester)}} - \Delta G^{\circ\prime}_{\text{bond broken (3′–5′ phosphodiester)}}

Using representative literature values of approximately -5.5 kcal/mol (3′–5′ phosphodiester, "normal" linkage) and approximately -5.0 kcal/mol (2′–5′ phosphodiester, slightly less stable due to steric strain from the branched ribose), we get:

\Delta G^{\circ\prime}_{\text{step 1}} \approx (-5.0) - (-5.5) = +0.5 \text{ kcal/mol}

This small, near-thermoneutral value (a few tenths of a kcal/mol, well within the precision of such estimates) confirms the qualitative claim above: splicing chemistry itself is not strongly driven either forward or backward by bond energetics; it is directionality (irreversibility in the cell) that comes from downstream steps — release of the lariat, ATP-dependent disassembly of the spliceosome by Prp43, and rapid debranching/degradation of the excised intron — that pull the overall pathway forward via Le Chatelier–type product removal, not from the chemical step's intrinsic thermodynamics. Limiting case: if the branch-point 2′–5′ linkage were exactly as stable as a normal 3′–5′ linkage ( $\Delta G^{\circ\prime}_{\text{formed}} = \Delta G^{\circ\prime}_{\text{broken}}$ ), the chemical step would be perfectly thermoneutral ( $\Delta G^{\circ\prime}=0$ ), and this is in fact the accepted approximation used in most biochemistry texts — the small nonzero value here simply reflects the modest steric strain of the branched lariat structure, and both signs of small deviation are chemically reasonable.

Example 2 — Applied: scoring a real human 5′ splice site against consensus and predicting the effect of a point mutation. The consensus 5′ splice site (positions -3 to +6 relative to the exon–intron boundary, intron in capitals) is approximately:

\text{C A G}\,|\,\text{G U A A G U}

with position-specific frequencies (approximate, from large splice-site databases) at the invariant GU (positions +1, +2) essentially 100% conserved, and substantial degeneracy elsewhere (e.g., position +5 is A in ~ 84% of sites, G in most of the remainder). Take a real exon–intron boundary with sequence C A G | G U G A G U (differing from consensus only at position +4, G instead of A). Using approximate position frequencies $f$ (A at +4) ≈ 0.71, $f$ (G at +4) ≈ 0.10 (real human 5′SS position weight matrices), and background $p_0 = 0.25$ :

S_{+4,\text{consensus }A} = \log_2\!\left(\frac{0.71}{0.25}\right) = \log_2(2.84) \approx 1.51 \text{ bits}

S_{+4,\text{observed }G} = \log_2\!\left(\frac{0.10}{0.25}\right) = \log_2(0.40) \approx -1.32 \text{ bits}

The mutant site loses:

\Delta S = S_{+4,\text{observed }G} - S_{+4,\text{consensus }A} = -1.32 - 1.51 = -2.83 \text{ bits}

relative to the fully consensus sequence at that one position. Summed across all positions this gives the total splice-site strength score used by tools such as MaxEntScan; a drop of nearly 3 bits at a single position is large enough, in a real weak splice site, to shift U1 snRNP binding equilibrium enough to cause exon skipping or intron retention — precisely the class of mechanism underlying many disease-causing splicing mutations (for example, in SMN2 in spinal muscular atrophy, where a single synonymous-looking exonic change disrupts an exonic splicing enhancer and weakens exon 7 inclusion). Limiting case: if the mutant nucleotide matched consensus exactly, $\Delta S = 0$ and no change in splice-site recognition would be predicted from this position alone — the calculation correctly reduces to "no effect" when there is, in fact, no sequence change, confirming the log-odds formula behaves sensibly at its trivial input.

Example 3 — Exam/competition level: reconstructing an intron's boundaries and lariat branch point from partial sequencing data, then computing the free 5′-exon length released after step 1. A pre-mRNA (numbering nucleotides 1–300 for convenience) is transcribed from a gene with one intron. Mapping data show: the exon1/intron boundary is at position 100 (so exon 1 = nt 1–100, and the intron begins at nt 101 with the sequence starting GUAAGU, consistent with consensus); a candidate branch-point adenosine is found at position 250, within a distinctive sequence context UACUAAC — the classic yeast branch-point consensus (in mammals the consensus is more degenerate, UNYURAC, but this exact sequence is textbook yeast); and the intron/exon2 boundary (3′ splice site, ending …YAG) is at position 280, so exon 2 begins at nt 281.

(a) Verify the branch point is positioned correctly. In both yeast and humans, the branch-point adenosine must lie a short, defined distance upstream of the 3′ splice site — typically 18–40 nucleotides in mammals (yeast is more rigidly ~ 18–24 nt due to a near-invariant branch consensus). Here:

d_{\text{branch to 3′SS}} = 280 - 250 = 30 \text{ nt}

This falls within the biologically observed window, so U2AF (the U2 auxiliary factor, which recognizes the polypyrimidine tract and 3′ AG in the region between the branch point and the 3′ splice site) has enough room to bind and hand off to U2 snRNP — confirming the assignment is mechanistically plausible, not merely a sequence match.

(b) Compute the size of the lariat intron after both catalytic steps. The full intron spans nt 101–280, a total length of:

L_{\text{intron}} = 280 - 101 + 1 = 180 \text{ nt}

After step 1 (branch attack at nt 101's phosphate), exon 1 (nt 1–100, length 100 nt) is released with a free 3′-OH, while the intron (nt 101–280, 180 nt) remains circularized onto itself via nt 250's new 2′–5′ bond — forming the lariat plus still-attached exon 2 (a "lariat-exon2" intermediate, total 180 + 20 nt of exon2 already present depending on total transcript length; here exon 2 alone is nt 281–300, length 20 nt, so the intermediate is 180 + 20 = 200 nt).

(c) Compute the mature mRNA length after step 2. Step 2 ligates exon 1 (100 nt) directly to exon 2 (20 nt), releasing the intron as a free lariat (180 nt, still branched at nt 250, now numbered internally 1–180 relative to its own sequence with the branch nucleotide at internal position 150). The mature spliced mRNA length is:

L_{\text{mRNA}} = L_{\text{exon1}} + L_{\text{exon2}} = 100 + 20 = 120 \text{ nt}

(d) Sanity check by conservation of total nucleotides. Before splicing, total pre-mRNA length was 300 nt (100 exon1 + 180 intron + 20 exon2). After splicing: mature mRNA (120 nt) + excised lariat intron (180 nt) = 300 nt exactly, confirming no nucleotides are lost or gained in the chemical steps — consistent with the transesterification mechanism's bond-neutral stoichiometry derived above. Limiting case: if the branch point were placed impossibly close to the 3′ splice site (say $d = 3$ nt, far below the ~18 nt minimum), U2AF and the second-step machinery would have no room to assemble; real spliceosomes reject such geometries, and this failure mode is exactly what is observed experimentally when branch points are engineered too close to the 3′SS — a useful consistency check that the numerical window used here (18–40 nt) is not arbitrary but mechanically load-bearing.

Limiting cases, scope & common traps

Situation	What actually happens	Common misconception
Intron with a weak/degenerate 5′ splice site	Reduced U1 binding, competition from alternative splice sites → exon skipping or intron retention (alternative splicing)	Assuming splicing is all-or-none and every intron is removed identically every time
No branch-point adenosine available (mutated to another base)	Nearby cryptic adenosines can sometimes substitute at reduced efficiency; splicing efficiency drops, aberrant transcripts increase	Assuming the reaction has an absolute, immutable requirement for adenine only — the 2′-OH is essential, but weak use of near-cognate branch points does occur
Capping enzyme fails to act (e.g., experimentally blocked)	Transcript is degraded rapidly by 5′→3′ exonucleases (Xrn1-family) and fails nuclear export/translation	Thinking the cap is merely a "protective cap" analogous to a plastic bottle cap with no active recognition role — it is actively read by CBC and eIF4E
Poly(A) signal (AAUAAA) mutated	Read-through transcription to a downstream cryptic poly(A) site, or transcriptional termination failure entirely	Assuming polyadenylation and transcription termination are fully independent processes — they are mechanistically coupled via Xrn2 "torpedo" termination
Self-splicing group I/II introns (no spliceosome)	Chemically analogous transesterification (group II) or direct GTP/guanosine-initiated transesterification (group I), but catalyzed by the intron RNA itself with no protein machinery required	Assuming all splicing requires the spliceosome — group I and II introns are autocatalytic ribozymes, historically important evidence for RNA catalysis
Very short introns (<70–80 nt in humans)	Splicing efficiency drops sharply below a minimum intron length because spliceosomal components sterically cannot assemble across too short a span	Assuming intron length is irrelevant to splicing as long as consensus sequences are present

Common trapMisconception: the poly(A) tail and the 5′ cap are just "stability tags" with no active reading

Both structures are directly bound by dedicated proteins that couple them mechanistically to translation and quality control: the cap is bound by the nuclear cap-binding complex (CBC) and, in the cytoplasm, by eIF4E; the poly(A) tail is bound by poly(A)-binding protein (PABP, cytoplasmic PABPC1 in humans). eIF4E and PABP interact indirectly through eIF4G, physically circularizing the mRNA into a closed loop. This "closed-loop" architecture is why cap and poly(A) tail synergize far more than additively in stimulating translation — removing either one collapses the loop, and this is a real biochemical event, not a metaphorical "protection," with measurable stimulatory effects on ribosome recruitment.

The chemist behind it

The discovery that genes in eukaryotes are interrupted by introns — arguably the single most surprising molecular biology finding of the late 1970s — came from independent electron-microscopy experiments by Phillip Sharp's group and Richard Roberts's group, both examining adenovirus mRNA hybridized back to its genomic DNA template. Instead of a simple linear duplex, they observed characteristic "R-loop" structures with unpaired single-stranded DNA loops bulging out of the RNA–DNA hybrid — visual, unambiguous proof that mature mRNA is shorter than, and non-colinear with, the gene that encodes it. This single image reoriented the entire field: genes were not simply continuous coding sequences, and a new class of machinery had to exist to stitch coding exons together after transcription.

Origins · 1 minPhillip Allen Sharp and how this was discovered

Phillip Allen Sharp

1944– · Molecular biologist, MIT; co-discoverer of RNA splicing and introns, 1993 Nobel Prize in Physiology or Medicine shared with Richard J. Roberts

Sharp's electron-microscopy R-loop experiments on adenovirus-2 major late transcripts in 1977 showed that the mature mRNA hybridized to four separate, non-contiguous regions of the viral genome, with intervening genomic DNA looping out unpaired — the first direct visualization of what would be named introns and exons (terms coined shortly after by Walter Gilbert). Sharp's subsequent work with cell-free nuclear extracts helped establish that splicing occurs via a lariat intermediate and defined the two-step transesterification mechanism, and his laboratory contributed foundational work characterizing the snRNP components of the spliceosome. He later co-founded Biogen and helped pioneer RNA interference therapeutics research at the Koch Institute, linking his splicing discoveries to a broader career in RNA-based biotechnology.

History

- 1977 — Sharp and Roberts independently discover split genes and introns in adenovirus, via R-loop electron microscopy.

- 1978 — Gilbert coins the terms "exon" and "intron."

- 1981–1984 — Steitz and colleagues show U1 snRNA base-pairs directly with the 5′ splice site; branch-point lariat intermediate identified biochemically by Sharp, Green, and colleagues using in vitro splicing extracts.

- 1982 — Cech discovers self-splicing group I introns in Tetrahymena rRNA, the first demonstration that RNA alone can catalyze a splicing reaction (shared 1989 Nobel Prize with Altman for ribozyme chemistry).

- 1993 — Sharp and Roberts awarded the Nobel Prize in Physiology or Medicine for the discovery of split genes.

- 2009 onward — cryo-EM structures of the fully assembled, catalytically active spliceosome (Shi and colleagues) reveal the U2/U6 snRNA catalytic core and two-metal-ion active site directly, confirming decades of biochemical prediction.

Why does splicing need to remove introns at all — why not just transcribe exons directly?

Eukaryotic genes are physically interrupted by intron sequences within the genomic DNA itself, so RNA polymerase II necessarily transcribes introns along with exons as it moves through the gene; there is no way to "skip" them at the DNA level during transcription. Introns likely persist evolutionarily because they enable alternative splicing (dramatically expanding proteome diversity from a fixed number of genes), provide substrates for regulatory small RNAs, and in some models reflect ancient exon-shuffling that assembled modern protein domains from separately encoded ancestral pieces. Whatever the evolutionary origin, once introns exist in the genome, a dedicated RNA-based excision system — the spliceosome — became necessary to convert the primary transcript into a translatable open reading frame.

Is the spliceosome a ribozyme, and how do we know?

Yes, in essence: the catalytic core of the activated spliceosome is built from base-paired U2 and U6 snRNA, which together fold to position two catalytic Mg²⁺ ions in geometry essentially superimposable on the catalytic core of self-splicing group II introns — RNA molecules known to catalyze this same two-step transesterification entirely without protein. Cryo-EM structures of catalytically active spliceosomes directly visualize this RNA-based active site, and biochemical experiments (metal-rescue experiments using phosphorothioate substitutions, which restore activity only when the correct metal-coordinating oxygen is replaced by sulfur alongside a metal ion substitution) confirm the snRNA, not the many associated proteins, performs the actual chemistry. The proteins are essential for assembly, fidelity, and ATP-dependent remodeling, but the catalytic center is RNA.

What happens if the poly(A) signal is deleted or destroyed by mutation?

Cleavage and polyadenylation specificity factor (CPSF) cannot assemble at the correct site, so the normal cleavage/polyadenylation event fails there. RNA polymerase II typically continues transcribing downstream, often for a considerable distance, until it encounters a fortuitous cryptic poly(A) signal further along or a genuine terminator; because 3′-end cleavage is mechanistically coupled to Xrn2-mediated "torpedo" termination, failure to cleave can also delay or prevent normal transcription termination, causing extended, often unstable, read-through transcripts. This is a well-documented consequence of some disease-associated 3′UTR mutations and is also exploited experimentally to map novel/alternative 3′ ends.

How does the cell distinguish its own capped mRNA from viral RNA lacking a proper cap?

Cytoplasmic pattern-recognition receptors such as RIG-I and MDA5 are tuned to bind RNA 5′ ends bearing an uncapped triphosphate or an unmethylated (cap-0, lacking 2′-O-methylation) cap structure, triggering a type I interferon antiviral response. Host mRNAs are efficiently converted to cap-1 (with 2′-O-methylation of the first transcribed nucleotide by the cap-specific 2′-O-methyltransferase CMTR1), which is a poor RIG-I ligand, effectively marking the RNA as "self." Many viruses have evolved their own capping or cap-snatching (influenza) or 2′-O-methylation enzymes specifically to evade this surveillance, and synthetic mRNA therapeutics/vaccines must replicate cap-1 chemistry closely to avoid triggering unwanted innate immune activation.

Can splicing occur without a spliceosome at all?

Yes — group I and group II self-splicing introns, found in some organellar, bacterial, and lower-eukaryotic rRNA/mRNA genes, catalyze their own excision without any protein machinery. Group II introns use essentially the identical two-step transesterification/lariat mechanism as the spliceosome (strong evidence that the spliceosome's RNA core is evolutionarily derived from, or convergent with, group II intron chemistry), while group I introns use a chemically related but distinct mechanism initiated by an external guanosine nucleophile rather than an internal branch-point adenosine. These ribozymes were pivotal evidence, alongside the ribosome's peptidyl transferase center, for the RNA World hypothesis — that RNA once served as both genetic material and catalyst before proteins and DNA became specialized for those roles.