Splicing changes which sequence reaches the ribosome without necessarily changing the underlying DNA.
By the end: Distinguish exons, coding sequence, introns, and untranslated regions, and reason about exon skipping.
Separate retained sequence from translated sequence
A typical nuclear protein-coding pre-mRNA receives a 5′ cap and, after cleavage, a 3′ poly-A tail. Splicing removes introns and joins retained exon sequences. These events help govern processing, export, stability, and translation. They need not occur as isolated stages after the entire transcript has finished.
An exon is a region retained in a particular mature transcript. Exons are not necessarily entirely protein-coding. Mature mRNA usually includes untranslated regions, or UTRs, before and after its coding sequence. A stop codon ends translation; it does not define the end of the RNA molecule.
Retention and translation mark different parts of an RNA
Processing acts on an RNA molecule that already contains sequence copied from DNA. Removing an intron changes the covalent connections in that RNA; it does not remove the corresponding DNA interval from the chromosome. The retained pieces become adjacent in the mature message. This new adjacency is an exon junction. A mature RNA can therefore place two bases next to each other even though many genomic bases separate their corresponding positions. Explaining this difference requires tracking which molecule is being discussed. A shorter RNA sequence is not evidence that the cell's genomic copy has undergone the same deletion.
Exon boundaries describe retention in a particular transcript. Coding boundaries describe what a ribosome translates in a specified event. Draw a mature message with an untranslated region before AUG, a coding interval beginning at AUG, an in-frame stop codon, and another untranslated region afterward. An exon can extend across either translation boundary, or it can lie wholly in an untranslated region. Thus there are two overlapping partitions of the same RNA. One partition follows how the message was assembled; the other follows how it will be read. They need not divide the molecule at the same positions.
One mature RNA has exon boundaries and translation boundaries
The colored interval marks the stipulated main coding region, including its stop. Its boundaries cut through retained exon blocks rather than defining those blocks. All drawn widths are schematic; untranslated regions can be exonic.
The 5′ cap is a chemical modification of the RNA end, involving a modified guanine nucleotide linked in a distinctive way. It is not a conventional three-base codon placed before the coding sequence. Proteins that recognize the cap help connect end chemistry to protection and translation initiation. Likewise, the poly-A tail is an added stretch of adenine-containing RNA nucleotides after cleavage at the 3′ end of many nuclear protein-coding transcripts. Its presence is not explained by ordinary copying of an equally long genomic run of T bases. Template copying and subsequent addition are different steps.
A typical message's 3′ untranslated region separates its stop codon from the cleavage and tail-addition boundary. Stopping translation therefore leaves untranslated RNA downstream. That region can contain sites recognized by molecules affecting message stability or use. The 5′ untranslated region can also affect how translation begins. Calling these regions untranslated states what happens during the specified protein-synthesis event, not whether their nucleotides have biological effects. A change in an untranslated exon could leave the main peptide sequence unchanged while changing how many peptide molecules the cell produces. The sequence consequence and the regulatory consequence are separate questions.
Processing does not have to wait for an entire long pre-mRNA to accumulate. The RNA emerges progressively, so some end modification and splicing can occur while downstream sequence is still being transcribed. This ordering makes physical sense: the early 5′ end is available before the last 3′ portion exists. A drawing that places full-length pre-mRNA above full-length mature RNA is a comparison of sequence organization, not proof that cells always complete one isolated stage before beginning the next. Overlapping steps still have dependencies. A segment cannot be joined to a downstream exon before the relevant RNA has become available.
The cap and tail are end features, not peptide residues
A cap symbol precedes a retained RNA body, and a short drawn A tail follows it. The cap linkage and all lengths are schematic. The tail represents post-transcriptional addition; it is not shown as a copied genomic T run. Neither feature is a stop amino acid.
The model below intentionally removes most of this organization to isolate one decision about retained coding sequence. It has no untranslated regions, cap chemistry, or explicit tail. Their absence in the display must not become a claim that real mature messages lack them. Conversely, adding an imaginary cap symbol would not supply the recognition machinery or demonstrate export. The useful deduction is narrower: selecting retained segments determines a joined RNA sequence. Whether that product matures, persists, reaches the cytoplasm, and supports translation requires additional molecular steps that the short bookkeeping example holds outside its boundary.
Compare alternative products
Alternative splicing can join different exon combinations from the same locus. Our invented message retains E1 = AUG and E3 = UUUUAA, while E2 = GCU is optional. Including E2 gives AUG GCU UUU UAA; skipping it gives AUG UUU UAA. The DNA need not change.
This example uses coding blocks with lengths divisible by three, so skipping E2 preserves the downstream reading frame. Real exon boundaries need not coincide with codon boundaries. Removing four coding nucleotides usually changes downstream grouping unless another change compensates. Some splice products are degraded instead of producing stable proteins.
Joining a three-base block preserves downstream grouping
Splicing requires recognition of appropriate boundaries and formation of a new backbone connection between retained sequences. In the common nuclear mechanism, a spliceosome, a complex containing RNAs and proteins, carries out the reactions. The removed intron forms a looped intermediate often called a lariat. A structural drawing of that loop explains how one RNA region can leave while the flanking exons become connected. It does not predict the correct boundaries from an arbitrary sequence. Recognition involves more information than the three short coding blocks supplied here, and our model has no rules for evaluating real splice sites.
For the constructed message, keep the order of bases fixed within each block. E1 contributes AUG, E2 contributes GCU, and E3 contributes UUUUAA. Joining all three produces twelve bases. Joining E1 directly to E3 produces nine. Both products start at the stipulated first base, so their triplets are AUG GCU UUU UAA and AUG UUU UAA respectively. The run of U bases makes exact counting worthwhile: E3 contains four U bases followed by two A bases. Its first three U bases form UUU, and its fourth U starts the stop triplet UAA.
Removal changes RNA connections without changing DNA
An intron is drawn as a curved RNA interval between two retained segments. After its removal, the retained segments are joined. The isolated loop suggests a lariat, but omits reaction chemistry, spliceosome structure, and actual boundary sequences. Lengths are schematic.
Triplets are groups of three RNA nucleotides read together during translation. Removing exactly three coding bases moves every downstream base three positions earlier. A downstream base that began a triplet therefore still begins a triplet; the same is true for its second and third positions. In this example the start of E3 moves from position seven to position four. Both positions begin triplets when counting from base one. The unchanged grouping is the reason UUU and UAA retain their meanings. It follows from arithmetic and a fixed starting point, not from a general property that exon skipping always preserves a protein.
At the residue level, the complete-inclusion product specifies methionine, alanine, and phenylalanine before termination. The shortened product specifies methionine and phenylalanine before termination. A residue is an amino acid incorporated into a peptide chain; it is not a nucleotide. The stop triplet contributes no residue. Thus removing three nucleotides reduces the peptide by one residue in this particular arrangement. Both chains still have an amino end, labeled N, and a carboxyl end, labeled C. Those chemical ends should not be relabeled 5′ and 3′, which describe nucleic-acid strands instead of peptides.
A four-base removal gives a contrasting arithmetic result without requiring a detailed new gene example. Four contains one full triplet plus one extra base. After the deletion, downstream bases are displaced by one triplet position relative to the old grouping. The later triplets will generally change until another event restores the frame or translation terminates. Length alone cannot tell which new stop will appear, because that depends on the remaining letters. Likewise, removal of three bases preserves downstream grouping but may alter a codon across a junction when the boundaries do not align with complete codons as they do here.
Removing GCU leaves the later UUU and UAA triplets intact
Every cell is one nucleotide. The complete-inclusion and skipped RNAs retain the same E3 letters and downstream grouping. The peptides have three and two residues respectively, with N and C chemical ends; UAA contributes no residue.
Preserved grouping is therefore a local sequence conclusion, not a verdict on protein function. Losing alanine might matter greatly at one position and little at another. The short model contains no fold, binding partner, localization signal, or activity assay with which to decide. It also assumes both RNA products are available for reading. Some real processing outcomes produce messages that are removed by quality-control pathways instead. Those possibilities do not invalidate the triplet calculation. They show why a correct calculation of a possible peptide must remain distinct from evidence that a stable, functional protein accumulates in a cell.
Work through the mechanism
Join retained exon sequence
E1 AUGintronE2 GCUintronE3 UUUUAA
Include all three
RNA read 5′ to 3′. Offset 0.
Base 1AUGMet
Base 4GCUAla
Base 7UUUPhe
Base 10UAAStop
Peptide: Met · Ala · Phe
An in-frame stop was reached.
Selected mature product
RNA read 5′ to 3′. Offset 0.
Base 1AUGMet
Base 4GCUAla
Base 7UUUPhe
Base 10UAAStop
Peptide: Met · Ala · Phe
An in-frame stop was reached.
Constructed coding blocks, not a real gene. Dashed intron links omit sequence and length; no splice-site prediction. Choosing a product leaves the full-inclusion reference visible.
Product
Joined sequence
Peptide in standard code
Include E2
AUG GCU UUU UAA
Met–Ala–Phe
Skip E2
AUG UUU UAA
Met–Phe
Stop is a signal, not a fourth amino acid.
Compare the selected message with the unchanged inclusion reference
The processing display always keeps the complete-inclusion product visible. Its other panel shows the selected product. Start with inclusion selected and compare the panels base by base. Both contain the same twelve-base RNA and three-residue peptide. Choose skipping next. Only the selected panel changes to the nine-base RNA and two-residue peptide. This arrangement separates the change in product from a change in the reference. It also prevents a common bookkeeping error: mentally subtracting the middle block twice because the source drawing and the selected product are both visible on the page.
Track identities rather than merely comparing lengths. In the inclusion product, bases one through three come from E1, four through six from E2, and seven through twelve from E3. In the skipped product, bases one through three still come from E1, but four through nine now come from E3. E3's bases are retained in order, not reverse-complemented. Splicing joins RNA segments in this stated order; it is not another template-copying event. A correct nine-base output that contains a reversed E3 would still be a wrong molecular account, even though its length happened to match the expected answer.
E3 changes position but retains its own base order
Position numbers count mature-RNA bases from the stipulated start. E3 starts at base seven with inclusion and base four with skipping. Both are first positions of triplets in frame zero. These exact mature lengths do not determine the omitted intron lengths.
The source drawing labels omitted intervals as introns, but those dashed links do not encode a number of nucleotides. Consequently, the mature lengths can be calculated exactly while the full pre-mRNA length cannot. Twelve retained bases plus two unknown intron lengths gives an unknown total. Skipping E2 removes its three bases from the selected product in addition to leaving the introns out. It does not prove that the removed material existed as one contiguous three-base intron in the source RNA. The final adjacency and the physical route by which it formed are related but different descriptions.
Now compare what the shared E3 sequence can establish. Detecting E3 would be compatible with either inclusion or skipping, since both products contain it. Detecting the E1-to-E3 junction would distinguish the skipped product in this invented two-product system. Detecting E1-to-E2 would instead support inclusion. This reasoning depends on the set of possible products: in a more complex gene, another transcript might contain one of those same junctions. A sequence feature is discriminating only relative to the alternatives under consideration. The model makes those alternatives explicit rather than supplying a universal rule about any three-exon gene.
The controls select product identity, not the proportion of a cellular RNA population. A screen showing one skipped product does not mean every RNA in a tissue has skipped E2, or that inclusion and skipping occur equally often. To see the difference, imagine ten RNA molecules. All ten could include E2, all ten could skip it, or the population could be mixed. Each molecule still follows one of the displayed sequence patterns. The display can explain each pattern while remaining silent about their frequencies. Frequency requires measurements or an additional population model, neither of which is provided by the selection button.
A shared segment and a product-specific junction answer different questions
Sequence feature
Include E2
Skip E2
E3: UUUUAA
Present
Present
E2: GCU
Present
Absent
E1|E2: AUG|GCU
Present
Absent
E1|E3: AUG|UUUUAA
Absent
Present
The E3 sequence occurs in both products. Only skipping creates the direct E1|E3 adjacency in this two-product model. The comparison states sequence presence, not detection sensitivity or relative cellular abundance.
Changing the selection also leaves the underlying DNA unspecified and unchanged within the example. The switch does not introduce a genomic deletion or edit the letters in an exon. It changes which supplied blocks are retained in the selected RNA. Real regulatory molecules can influence such choices, and DNA variants can sometimes alter them, but the display does not decide which cause is responsible. Its complete output is a joined sequence and its standard-code peptide under stipulated initiation. Keeping that contract narrow makes the model useful for checking retention and frame without pretending to predict expression in an actual organism.
Connect regulation to evidence
Splice-site recognition depends on sequence and associated machinery. A variant outside the translated sequence can alter splicing. Measuring total RNA shared by all products may miss a change in their proportions. Junction-specific measurements can distinguish which exons were joined.
A predicted splice product is a hypothesis until supported in the relevant cells. Different tissues can use different products. Our display omits actual splice-site motifs, intron lengths, UTRs, cap chemistry, and quality-control pathways. It is a bookkeeping model of retained sequence, not a splice predictor.
Equal total RNA can conceal different retained products
Consider two constructed RNA populations with the same total of one hundred messages containing E1 and E3. Population A contains eighty inclusion messages and twenty skipped messages. Population B contains twenty inclusion messages and eighty skipped messages. A measurement recognizing a shared E3 region could report the same total in both populations under equal per-message detection. Yet a measurement recognizing E2 would differ fourfold, eighty versus twenty. The contrast comes entirely from product composition, not from a change in total message count. This arithmetic assumes the measurements distinguish their stated targets and does not describe experimental signal calibration.
The same counts show why a junction measurement can answer a sharper question than total RNA. The E1-to-E3 junction appears twenty times in A and eighty times in B under the two-product assumption. Its increase supports more skipped product in the sampled population. It does not identify why skipping increased. A change in processing choice, different survival of the products, or a change in the types of cells contributing RNA could all affect the observed mixture. Product-specific abundance still combines formation and persistence. Measuring a more specific target improves the question without automatically supplying its causal answer.
One hundred messages can have different product mixtures
Each symbol represents ten messages. Circles labeled I are inclusion messages, and triangles labeled S are skipped messages. Both populations total one hundred; E2-containing counts change from eighty to twenty while skipped-junction counts change from twenty to eighty.
Length provides another partial view. The model's inclusion product has twelve bases, and its skipped product has nine. If a size measurement could resolve that difference, it could distinguish these two products by length. But two different nine-base RNAs can have the same apparent size, and limited resolution can merge nearby lengths. Therefore, agreement in size cannot prove agreement in nucleotide order or exon composition. Sequence evidence across a junction contains different information from a size measurement. Neither should be described as a complete molecular identity test without stating what interval or feature it actually resolves.
A genomic comparison and an RNA comparison can also disagree for informative reasons. Suppose two cell types have the same measured DNA sequence across a locus but different mature exon junctions. Alternative processing is compatible with those observations because the assembly decision occurs on RNA. The DNA evidence alone does not show that every regulatory condition is the same; the cells may contain different RNA-binding proteins or other factors. Conversely, a DNA difference near a boundary is only a candidate explanation for an RNA difference. Demonstrating an association between them does not establish that the particular change caused the processing outcome.
Any claim about a resulting protein must account for further steps. A detected mature RNA might be rare in the cytoplasm, translated inefficiently, or removed before many peptides accumulate. A peptide that is made may have altered stability or activity. The inclusion and skipped sequences here support a limited standard-code comparison, with initiation fixed at base one and termination at UAA. They do not supply rates for any of those later events. Even equal numbers of the two messages would not guarantee equal numbers of their peptide products. Translation and protein removal introduce additional quantities beyond retained RNA composition.
Equal length does not establish sequence identity
Both separately listed RNAs have nine nucleotides, but the middle triplet differs. A measurement that resolves only length cannot distinguish them by base identity. These are sequence specimens, not a prediction of their expression or function.
A bounded conclusion can still explain a real distinction. If the total shared-region signal is unchanged while the skipped junction becomes more abundant, the evidence supports changed product composition under the stated assay assumptions. If a corresponding cDNA crosses E1 directly into E3, it records that joined RNA sequence rather than proving a matching genomic deletion. If a predicted peptide lacks alanine in this model, the standard code explains the sequence difference but not its functional importance. These conclusions connect retention, measurement, and translation while preserving the missing links that would need separate evidence in a cellular claim.
Exercises
A four-base coding exon
A mature message loses a four-base segment within its coding region. Does the downstream reading frame necessarily stay the same?
Hint
Compare the removed length with a multiple of three.
Worked answer
Removed coding bases
Length modulo 3
Downstream frame
3
0
Preserved
4
1
Shifted
This assumes no compensating insertion or alternative initiation.
No. Four is not divisible by three, so downstream codon grouping shifts in this simple deletion model. The new peptide and stop position depend on the actual sequence. A frame shift does not specify a universal phenotype.
An exon with no amino acids
A retained exon lies entirely before the translation start site. Is calling it an exon a mistake?
Hint
Retention in RNA and translation are separate definitions.
Worked answer
Property
This exon
Retained in mature RNA
Yes
Translated into the main peptide
No
Potential regulatory role
Possible, requires evidence
Noncoding does not mean functionless.
No. It can form part of the 5′ UTR. It remains in mature RNA but does not encode amino acids in that translation event. Its sequence may still affect stability, localization, or translation.
Sources and further reading
Selected textbook passages support the mechanisms below. The prose and diagrams are original, and examples are constructed rather than measured. These references are not a full literature review or independent specialist validation. Optional links are not runtime dependencies.