Genes are written with interruptions. Before the cell can use one, it has to cut out the filler and stitch the useful pieces together. A canonical splice site is the standard signal that marks where those cuts go: a short letter pattern, GT at the start of the filler and AG at its end, found at about 99 in 100 of these boundaries. So when a genetic report flags a change at a canonical splice site, it is pointing at one of the most tightly protected positions in a gene, and that is why such changes are taken seriously.
Why Genes Need Cutting in the First Place
Most human genes are split into exons, the parts that end up in the final instructions, and introns, the filler stretches between them. The gene is first copied into a draft message that still contains every intron. The cell must then remove each intron and join the exons in order, a step called splicing. That leaves each intron with two boundaries the cell has to find precisely, its start and its end.
Precision matters because protein instructions are read three letters at a time. A cut shifted by even one or two letters scrambles everything after it, like misaligning every word after a typo. A scrambled message usually produces a useless protein or gets destroyed.
A splice site isn’t a gap or a scissor-shaped object. It’s a short run of letters in the sequence that the cutting machinery recognizes as a boundary marker.
The GT and AG Letters That Make a Site Canonical
An intron’s start is called the donor site and its end the acceptor site, terms you may see on a genetic report. In the canonical pattern, the donor begins with the letters GT and the acceptor ends with AG.1PubMed Central. Natural selection affects frequencies of AG and GT dinucleotides at the 5′ and 3′ ends of exons In mammals, about 98.7% of introns use this GT-AG pair.2PubMed Central. Analysis of canonical and non-canonical splice sites in mammalian genomes
The letters around them matter too. A typical donor reads CAG/GTAAGT and a typical acceptor reads NYAG/G, with the slash marking the cut.3PubMed Central. Splicing mutations in human genetic disorders: examples, detection, and confirmation At the donor, the slash falls between the end of the exon and the start of the intron. Letters to its left belong to the exon that is kept, and letters to its right belong to the intron that is removed. So GT is not part of the kept message; it’s the first two letters of the filler that gets discarded. At the acceptor it works the other way: AG is the last two letters of the intron, just before the slash.
In a pattern like NYAG/G, N stands for any of the four letters and Y stands for a C or T. The pattern describes typical sequences rather than one exact string. Only the GT and AG pair is close to fixed. The surrounding letters vary between genes, and that variation affects how strongly each site is recognized.
You may also see GU instead of GT. In DNA the donor reads GT, but in the RNA copy the T becomes a U. Sources and reports write GT-AG or GU-AG, and both describe the same canonical site.
Positions on a report are counted from the cut point. The +1 and +2 positions are the G and T just inside the intron’s start, and -1 and -2 are the A and G at the intron’s very end. A report entry like c.123+1G>A means the first letter of the intron changed from G to A.
The cutting machinery is built partly from small RNA molecules that pair with these letters.4PubMed Central. Extended base pair complementarity between U1 snRNA and the 5′ splice site does not inhibit splicing in higher eukaryotes, but rather increases 5′ splice site recognition One of them, U1, recognizes the donor by pairing with it, which starts the whole process.5PubMed Central. Extended base pair complementarity between U1 snRNA and the 5′ splice site does not inhibit splicing in higher eukaryotes, but rather increases 5′ splice site recognition For the acceptor, the machinery finds a landing spot inside the intron, the branch point, then picks the first AG after it.6PubMed. Scanning from an independently specified branch point defines the 3′ splice site of mammalian introns
What Happens When a Canonical Splice Site Is Changed
The most classic disease-causing changes hit the +1 and +2 letters of the donor or the -1 and -2 of the acceptor.7PubMed Central. Splicing mutations in human genetic disorders: examples, detection, and confirmation Such changes usually cause exon skipping, where a whole exon is left out of the message.8PubMed Central. Splicing mutations in human genetic disorders: examples, detection, and confirmation9PubMed Central. Splicing mutations in human genetic disorders: examples, detection, and confirmation The change can instead lead the cell to use a hidden backup cutting spot nearby.10PubMed Central. Splicing mutations in human genetic disorders: examples, detection, and confirmation Splice site changes are estimated to account for about 15% of point mutations that cause inherited disease.
Backup sites can win even against a healthy site. In patients with one inherited immune disorder, a backup donor site out-competed the normal one even though both looked equally strong.11PubMed Central. Activation of cryptic splice sites in three patients with chronic granulomatous disease Whether a backup site takes over depends on one being nearby and on the strength of neighbouring sites.12PubMed. Extensive in silico analysis of NF1 splicing defects uncovers determinants for splicing outcome upon 5′ splice-site disruption
When an exon is skipped, the message is shorter, and the protein built from it may be missing a whole segment that had a job to do. If the skipped exon’s length is a multiple of three letters, the rest of the message still reads in the right frame and the protein may be shortened but otherwise intact. If it isn’t, everything after the gap is misread, and the protein is usually nonfunctional or the message is destroyed.
When a backup cutting spot is used, the cut lands a little to one side of the normal boundary, so the exon ends up slightly longer or shorter than intended. A shift of a few letters usually changes the reading frame, while a shift of exactly three or a multiple of three may leave it intact. The result can also be a mixture, with some messages made correctly and others not, which can mean a milder condition than a complete loss.
For a variant on your report, labs often treat +1/+2 and -1/-2 changes as likely to disrupt splicing. The final call also depends on the gene and on known disease links. A change near a splice site is not automatically harmful, though. Only the core canonical letters carry the strongest expectation, and nearby positions are less predictable and often need more evidence. It’s reasonable to ask your genetic counselor whether RNA testing has confirmed the effect.
Non-Canonical Splice Sites and Why They Exist
About 0.56% of mammalian introns start with GC instead of GT and still end in AG.13PubMed Central. Analysis of canonical and non-canonical splice sites in mammalian genomes In protein-making human genes, GC-AG introns run about 0.83%,14PubMed Central. GC-AG Introns Features in Long Non-coding and Protein-Coding Genes Suggest Their Role in Gene Expression Regulation but in human genes that do not make proteins they reach about 3%.15PubMed Central. GC-AG Introns Features in Long Non-coding and Protein-Coding Genes Suggest Their Role in Gene Expression Regulation A GC donor pairs less tightly with the recognizing component than a GT donor, so on its own it would be a weaker signal. These sites compensate with surrounding letters in the exon and intron that pair especially well with the machinery.16PubMed Central. Human GC-AG alternative intron isoforms with weak donor sites show enhanced consensus at acceptor exon positions Donor sites also differ in how well they pair with U1, which is a large part of their strength. When two donors compete, the stronger one usually wins, though as the backup-site example above shows, other factors can tip the balance.
A rare group of introns uses AT-AC and is removed by a separate, minor spliceosome.17PubMed. A novel spliceosome containing U11, U12, and U5 snRNPs excises a minor class (AT-AC) intron in vitro18PubMed. Minor spliceosome and disease That is a distinct cutting machine with its own small RNA parts, built to recognize a different family of introns. It also processes some introns that begin with GT and end with AG, so the letters alone don’t always tell you which machinery will do the cutting. The remaining 0.73% of introns are split among many small, rare patterns.19PubMed Central. Analysis of canonical and non-canonical splice sites in mammalian genomes
On a report, a variant at a GC or AT-AC site isn’t automatically a canonical change in the usual sense, so labs may judge it with different evidence. If a variant turns a GC donor into a GT donor, the site may become stronger rather than broken, so the direction of the change matters.
How Labs Predict Splicing Effects
Labs commonly use a tool called SpliceAI to predict splicing effects directly from DNA letters.20eLife. OpenSpliceAI provides an efficient modular implementation of SpliceAI enabling easy retraining across nonhuman species Such tools correctly classify roughly 89% to 98% of variants in sets dominated by canonical-site changes.21PubMed Central. Analyzing the performance of deep learning splice prediction algorithms For changes deep inside introns, though, standard cutoffs miss most harmful variants.22PubMed Central. Analyzing the performance of deep learning splice prediction algorithms In cancer, deep intronic changes often create new cut sites at existing GT or AG letters.23PubMed Central. Comprehensive characterisation of intronic mis-splicing mutations in human cancers
Tools like SpliceAI give a score between 0 and 1 for each type of effect, such as losing or gaining a donor or acceptor, and a higher score means a bigger expected change. A score is a probability-style estimate, not a measurement, so it’s a reason to look further rather than a final answer. Labs choose a cutoff for what counts as a likely effect: a stricter one misses more real cases, and a looser one flags more false alarms.
Lab tests can measure splicing for thousands of variants at once, and RNA testing of a patient’s cells can confirm a single one.24bioRxiv. Massively parallel assay of human splice variants reveals cis-regulatory drivers of disease-associated and cell type-specific splicing regulation25PubMed Central. Massively parallel reporter assays discover de novo exonic splicing mutants in paralogs of Autism genes RNA testing usually takes cells from blood and reads the message made from the gene to see which exons are actually joined. If the gene isn’t active in blood, the lab may need special methods or a different tissue. A mix of normal and abnormal messages in the result can mean the variant has a partial effect, though normal messages may also come from the other, unchanged copy of the gene.
Changes inside exons count too, because splicing also relies on helper signals within them. Gaining a signal that suppresses use of an exon is a strong cause of skipping, while losing a helper signal alone predicts poorly.
When you’re given a splice prediction, a few questions help. Was it confirmed by RNA testing? Is the variant at a canonical position or deep in an intron? Which tool and cutoff were used? Has the same variant been seen in others with the condition?
Can Splicing Problems Be Treated
In spinal muscular atrophy, a key exon, exon 7, is often left out of the message.26PubMed Central. Antisense Oligonucleotides for the Treatment of Spinal Muscular Atrophy Short designed molecules can block signals that cause it to be skipped, so the exon is kept and a full-length message is made.27PubMed Central. Antisense Oligonucleotides for the Treatment of Spinal Muscular Atrophy This approach strongly improved disease in animal models and has moved into patient treatment.28PubMed Central. Antisense Oligonucleotides for the Treatment of Spinal Muscular Atrophy
The drug is a short strand built to match a specific stretch of the message, so it sticks to that spot like a matching puzzle piece. By covering a signal that tells the cell to skip an exon, it lets the cutting machinery treat the exon as usable. The effect is on the message, not the DNA, so the underlying genetic change stays the same and treatment generally needs to be repeated.
Such drugs are designed for one specific gene and change, and they exist for only a few conditions so far. Knowing a disease is caused by splicing helps doctors judge whether such approaches apply. The first question to ask is whether the condition is known to be caused by a splicing problem at all. The second is whether your exact variant has been studied, because a drug designed for one change does not automatically work on another in the same gene. A specialist can tell you about trials for your specific variant.