Alternative Splicing

How splicing works, in chemistry

Each intron is bounded by GU at its 5' end and AG at its 3' end, with a branch-point adenosine ~20–50 nt upstream of the 3' splice site. The spliceosome assembles step-wise:

1. E complex. U1 snRNP base-pairs with the 5' splice site; U2AF binds the polypyrimidine tract and 3' AG.
2. A complex. U2 snRNP base-pairs with the branch point, bulging out the branch A.
3. B complex. The U4/U6.U5 tri-snRNP joins; the spliceosome is now assembled but catalytically inactive.
4. B activation. U1 leaves the 5' splice site; U6 takes over and base-pairs with U2 to form the catalytic core. U4 dissociates.
5. First transesterification. The 2'-OH of the branch-point A attacks the 5' splice site phosphodiester bond, freeing the upstream exon and forming a lariat intron-exon intermediate.
6. Second transesterification. The 3'-OH of the freed upstream exon attacks the 3' splice site, joining the two exons and releasing the lariat intron, which is then debranched and degraded.

Both steps are simple Sn2-like nucleophilic substitutions — the same chemistry as group II self-splicing introns and a likely ancestor of the spliceosome.

Splicing decision diagram

  5' exon ── GU ─────── intron ─────── A ── (Py)n ── AG ── 3' exon
              │                        │              │
              ▼                        ▼              ▼
            U1 binds              U2 binds         U2AF / 3' SS

            ▼ (U4/U6.U5 joins, U1+U4 leave, U6+U2 form catalytic core)

            branch-A 2'-OH attacks 5' SS  →  lariat intermediate
                                            │
            3'-OH of upstream exon attacks 3' SS  →  joined exons + lariat

Five canonical alternative splicing patterns

Pattern	What changes	Frequency in humans	Functional effect	Classic example	Aberrant example
Exon skipping (cassette)	An internal exon is included or excluded	~40% of alternative events	Adds or removes a protein domain	Tropomyosin TPM1 muscle isoforms	SMN2 exon 7 skipping → SMA
Intron retention	An intron is kept in the mature mRNA	~5% in mammals; majority in plants	Often introduces premature stop → NMD	Plant flowering FLC regulation	NF1 intron retention in tumors
Alternative 5' splice site	Different 5' GU chosen	~8%	Shortens or extends exon	BCL2L1 Bcl-x(L) vs Bcl-x(S)	LMNA progerin in progeria
Alternative 3' splice site	Different 3' AG chosen	~18%	Shortens or extends exon	FN1 fibronectin EDB region	Cancer-specific MDM4 splicing
Mutually exclusive exons	Exactly one of two (or more) exons included	~5%, but most striking	Switches functional domain	FGFR2 IIIb (epithelial) vs IIIc (mesenchymal)	—
Alternative promoters / poly-A	Different first or last exon	Common; often combined	Tissue-specific UTRs and translation	DRD2 dopamine receptor isoforms	—

One transcript can use several patterns simultaneously: Drosophila Dscam combines four mutually-exclusive cassettes; mammalian troponin T uses cassette and alternative-3' events.

Real-world significance

• Tissue identity. Brain has the highest splicing complexity. Microexons (3–27 nt) regulated by SRRM4 are nearly brain-exclusive and misregulated in autism (Irimia, Cell 2014).
• Cancer. Mutations in splicing factors (SF3B1, SRSF2, U2AF1) are found in 60–80% of myelodysplastic syndromes. They reprogram exon usage genome-wide, generating tumor-specific neoantigens for immunotherapy.
• SMA treatment. Nusinersen (Spinraza, 2016) and risdiplam (Evrysdi, 2020) restored SMN2 exon 7 inclusion and turned a fatal infant disease into chronic-care. Splicing's first proof-of-concept as a drug target.
• Duchenne muscular dystrophy. Eteplirsen (Exondys 51, 2016), golodirsen, viltolarsen, casimersen — antisense oligos that force skipping of mutated dystrophin exons to restore an in-frame (if shorter) protein.
• Sex determination. Drosophila sex is controlled by a splicing cascade: Sex-lethal → transformer → doublesex. Each step is alternative-3'-splice-site choice.
• Adaptive immunity. Membrane vs secreted IgM are produced by alternative splicing of the same heavy-chain transcript — the very RNA that pushed Sharp and Roberts toward split genes in 1977.

Variants and special cases

• Trans-splicing. Splice sites on different RNA molecules joined. Standard in trypanosomes; rare in mammals (JAZF1-JJAZ1).
• Recursive splicing. Long introns removed in successive segments — important for dystrophin.
• Back-splicing → circRNAs. A downstream 5' splice site joins an upstream 3' splice site, producing a circular RNA. CDR1as sponges miR-7 in brain.
• Self-splicing introns. Group I and group II introns splice without protein (Cech, Altman, Nobel 1989).
• Minor (U12) spliceosome. Splices ~700 human introns with non-canonical boundaries; U4atac mutations cause MOPDI.

Pitfalls and clinical traps

• RNA-seq quantification. Short reads cannot resolve all isoforms. Tools like Salmon, Kallisto, and rMATS estimate isoform abundance with assumptions; long-read sequencing (Iso-Seq, Nanopore) is the rigorous answer.
• Reference-annotation bias. If an isoform is missing from your annotation (Ensembl, GENCODE), its reads get misassigned. Always inspect novel splice junctions.
• Cryptic splice sites. Mutations can create new GT or AG dinucleotides that the spliceosome uses. Many disease alleles act this way (Tay-Sachs HEXA, β-thalassemia HBB) — in silico splice predictors (SpliceAI) catch most.
• Progeria. A single C1824T LMNA point mutation activates a cryptic 5' splice site, producing the truncated progerin protein. The classical example of splicing pathology from a non-splice-site SNV.
• Splicing-NMD coupling. Many isoforms exist transiently and are degraded by NMD; their detection depends on inhibiting NMD (cycloheximide, UPF1 knockdown). Steady-state RNA-seq under-counts these regulated targets.
• SMA carrier testing. SMN1 and SMN2 are 99% identical; confusing them in PCR-based diagnostics has led to misdiagnoses. MLPA or droplet-digital PCR is the standard.