The Spliceosome and Alternative Splicing
Lariat Intron Excision, Small Nuclear Ribonucleoproteins, and the DSCAM Isoform Explosion
Richard Roberts and Phillip Sharp independently hybridized mature adenovirus-2 messenger RNA to the single-stranded viral DNA that encoded it in 1977. Under the electron microscope, three giant loops of uncopied single-stranded DNA bulged out from the hybrid duplex. That image shattered the bacterial rule that one gene is a continuous, uninterrupted stretch of DNA.
In the human genome, protein-coding exons make up only 1.5 percent of all DNA bases, whereas non-coding introns inside genes span roughly 25 percent. A typical human gene is split across eight or nine introns. The human dystrophin gene spans $2.4\text{ million}$ base pairs—taking 16 hours to transcribe—only to have 99.4 percent of its sequence spliced out.
Unlike the ribosome, which stays intact between jobs, the spliceosome assembles from scratch on every single intron and disassembles once the cut is sealed. Five small nuclear ribonucleoproteins (U1, U2, U4, U5, and U6 snRNPs) and over 150 proteins converge on the pre-mRNA. Eight ATP-burning DExD/H-box RNA helicases remodel the complex step by step.
The two chemical cuts that remove an intron require zero external energy from ATP because the total number of phosphodiester bonds never changes. First, the $2'\text{-OH}$ group of an intronic branch-point adenosine attacks the $5'$ GU splice site to form a $2'\text{-5'}$ looped lariat. The freed upstream exon's $3'\text{-OH}$ then attacks the $3'$ AG site to join the exons.
Between 2015 and 2017, cryo-EM structures solved at $3.6\text{ \AA}$ by Yigong Shi, Kiyoshi Nagai, and Reinhard Lührmann proved Thomas Cech's 1986 evolutionary hypothesis. The spliceosome is a protein-directed ribozyme. Inside its catalytic core, U6 and U2 snRNAs fold into a triple helix that coordinates two $\text{Mg}^{2+}$ ions identically to self-splicing Group II introns.
When the Human Genome Project found only 20,000 protein-coding genes—barely more than a millimeter-long nematode worm—alternative splicing explained the discrepancy. Deep RNA sequencing by Christopher Burge and Benjamin Blencowe in 2008 showed that 95 percent of multi-exon human genes are spliced in multiple ways. Those combinations generate over 100,000 distinct protein isoforms across human tissues.
The extreme limit of alternative splicing occurs in the fruit fly (Drosophila melanogaster) axon-guidance gene Dscam1, discovered by Lawrence Zipursky in 2000. Its pre-mRNA contains 95 mutually exclusive alternative exons arranged in four clusters of 12, 48, 33, and 2 cassettes. Choosing one cassette from each cluster allows a single gene to produce 38,016 distinct receptor proteins.
Why does a fruit fly having only 14,000 genes need 38,016 versions of one receptor? Every developing fly neuron expresses a random subset of 15 to 50 Dscam1 isoforms on its membrane. Because each Dscam1 protein binds exclusively to an identical isoform on contact to trigger repulsion, sister dendrites repel one another while mingling freely with neighboring neurons.
In the vertebrate inner ear, hair cells along the basilar membrane of the cochlea tune their electrical resonance to sound frequencies spanning 20 to 20,000 hertz. Much of that continuous frequency gradient is built by alternative splicing of a single gene (Slo). The spliceosome generates over 500 splice variants of the BK potassium channel with graded opening speeds.
How does the vertebrate spliceosome avoid getting lost inside a 50,000-nucleotide intron while searching for a tiny 140-nucleotide exon? Susan Berget proved in 1995 that the machinery uses exon definition. U2AF at the upstream $3'$ splice site and U1 snRNP at the downstream $5'$ site first bridge across the compact 140-base exon via SR proteins before pairing across the intron.
Human geneticists estimate that 15 to 30 percent of all disease-causing single-base mutations damage health by corrupting pre-mRNA splicing rather than mutating an amino acid directly. In Hutchinson-Gilford progeria, a silent $\text{C}\to\text{T}$ substitution in codon 608 of the LMNA gene leaves the glycine unchanged. Yet it activates a cryptic splice site that deletes 50 amino acids from lamin A.
Adrian Krainer and colleagues exploited splicing regulation to develop nusinersen (Spinraza), approved in 2016 as the first treatment for infantile spinal muscular atrophy. Injected into the cerebrospinal fluid, the 18-nucleotide antisense oligonucleotide masks an intronic splicing silencer on SMN2 pre-mRNA. That block coaxes the spliceosome to retain exon 7 and restore full-length motor-neuron protein.