Alternative splicing

Alternative splicing, or alternative RNA splicing, or differential splicing, is an alternative splicing process during gene expression that allows a single gene to produce different splice variants. For example, some exons of a gene may be included within or excluded from the final RNA product of the gene.^[1] This means the exons are joined in different combinations, leading to different splice variants. In the case of protein-coding genes, the proteins translated from these splice variants may contain differences in their amino acid sequence and in their biological functions (see Figure).

Biologically relevant alternative splicing occurs as a normal phenomenon in eukaryotes, where it increases the number of proteins that can be encoded by the genome.^[1] In humans, it is widely believed that ~95% of multi-exonic genes are alternatively spliced to produce functional alternative products from the same gene^[2] but many scientists believe that most of the observed splice variants are due to splicing errors and the actual number of biologically relevant alternatively spliced genes is much lower.^[3][4]

Alternative splicing was first observed in 1977.^[5][6] The adenovirus produces five primary transcripts early in its infectious cycle, prior to viral DNA replication, and an additional one later, after DNA replication begins. The early primary transcripts continue to be produced after DNA replication begins. The additional primary transcript produced late in infection is large and comes from 5/6 of the 32kb adenovirus genome. This is much larger than any of the individual adenovirus mRNAs present in infected cells. Researchers found that the primary RNA transcript produced by adenovirus type 2 in the late phase was spliced in many different ways, resulting in mRNAs encoding different viral proteins. In addition, the primary transcript contained multiple polyadenylation sites, giving different 3' ends for the processed mRNAs.^[7][8][9]

In 1981, the first example of alternative splicing in a transcript from a normal, endogenous gene was characterized.^[7] The gene encoding the thyroid hormone calcitonin was found to be alternatively spliced in mammalian cells. The primary transcript from this gene contains 6 exons; the calcitonin mRNA contains exons 1–4, and terminates after a polyadenylation site in exon 4. Another mRNA is produced from this pre-mRNA by skipping exon 4, and includes exons 1–3, 5, and 6. It encodes a protein known as CGRP (calcitonin gene related peptide).^[10][11] Examples of alternative splicing in immunoglobin gene transcripts in mammals were also observed in the early 1980s.^[7][12]

Since then, many other examples of biologically relevant alternative splicing have been found in eukaryotes.^[1] The "record-holder" for alternative splicing is a D. melanogaster gene called Dscam, which could potentially have 38,016 splice variants.^[13]

In 2021, it was discovered that the genome of adenovirus type 2, the adenovirus in which alternative splicing was first identified, was able to produce a much greater variety of splice variants than previously thought.^[14] By using next generation sequencing technology, researchers were able to update the human adenovirus type 2 transcriptome and document the presence of 904 splice variants produced by the virus through a complex pattern of alternative splicing. Very few of these splice variants have been shown to be functional, a point that the authors raise in their paper.

"An outstanding question is what roles the menagerie of novel RNAs play or whether they are spurious molecules generated by an overloaded splicing machinery."^[14]

Five basic modes of alternative splicing are generally recognized.^{[1][2][16][15]}

• Exon skipping or cassette exon: in this case, an exon may be spliced out of the primary transcript or retained. This is the most common mode in mammalian pre-mRNAs.^[15]
• Mutually exclusive exons: One of two exons is retained in mRNAs after splicing, but not both.
• Alternative donor site: An alternative 5' splice junction (donor site) is used, changing the 3' boundary of the upstream exon.
• Alternative acceptor site: An alternative 3' splice junction (acceptor site) is used, changing the 5' boundary of the downstream exon.
• Intron retention: A sequence may be spliced out as an intron or simply retained. This is distinguished from exon skipping because the retained sequence is not flanked by introns. If the retained intron is in the coding region, the intron must encode amino acids in frame with the neighboring exons, or a stop codon or a shift in the reading frame will cause the protein to be non-functional. This is the rarest mode in mammals but the most common in plants.^[15][17]

In addition to these primary modes of alternative splicing, there are two other main mechanisms by which different mRNAs may be generated from the same gene; multiple promoters and multiple polyadenylation sites. Use of multiple promoters is properly described as a transcriptional regulation mechanism rather than alternative splicing; by starting transcription at different points, transcripts with different 5'-most exons can be generated. At the other end, multiple polyadenylation sites provide different 3' end points for the transcript. Both of these mechanisms are found in combination with alternative splicing and provide additional variety in mRNAs derived from a gene.^[1][16]

These modes describe basic splicing mechanisms, but may be inadequate to describe complex splicing events. For instance, the figure to the right shows 3 spliceforms from the mouse hyaluronidase 3 gene. Comparing the exonic structure shown in the first line (green) with the one in the second line (yellow) shows intron retention, whereas the comparison between the second and the third spliceform (yellow vs. blue) exhibits exon skipping. A model nomenclature to uniquely designate all possible splicing patterns has recently been proposed.^[15]

Genuine alternative splicing occurs in both protein-coding genes and non-coding genes to produce multiple products (proteins or non-coding RNAs). External information is needed in order to decide which product is made, given a DNA sequence and the initial transcript. Since the methods of regulation are inherited, this provides novel ways for mutations to affect gene expression.^[33]

Alternative splicing may provide evolutionary flexibility. A single point mutation may cause a given exon to be occasionally excluded or included from a transcript during splicing, allowing production of a new protein isoform without loss of the original protein.^[1] Studies have identified intrinsically disordered regions (see Intrinsically unstructured proteins) as enriched in the non-constitutive exons^[34] suggesting that protein isoforms may display functional diversity due to the alteration of functional modules within these regions. Such functional diversity achieved by isoforms is reflected by their expression patterns and can be predicted by machine learning approaches.^[35][36] Comparative studies indicate that alternative splicing preceded multicellularity in evolution, and suggest that this mechanism might have been co-opted to assist in the development of multicellular organisms.^[37]

Research based on the Human Genome Project and other genome sequencing has shown that humans have only about 30% more genes than the roundworm Caenorhabditis elegans, and only about twice as many as the fly Drosophila melanogaster. This finding led to speculation that the perceived greater complexity of humans, or vertebrates generally, might be due to higher rates of alternative splicing in humans than are found in invertebrates.^[38][39] However, a study on samples of 100,000 expressed sequence tags (EST) each from human, mouse, rat, cow, fly (D. melanogaster), worm (C. elegans), and the plant Arabidopsis thaliana found no large differences in frequency of alternatively spliced genes among humans and any of the other animals tested.^[40] Another study, however, proposed that these results were an artifact of the different numbers of ESTs available for the various organisms. When they compared alternative splicing frequencies in random subsets of genes from each organism, the authors concluded that vertebrates do have higher rates of alternative splicing than invertebrates.^[41]

Changes in the RNA processing machinery may lead to mis-splicing of multiple transcripts, while single-nucleotide alterations in splice sites or cis-acting splicing regulatory sites may lead to differences in splicing of a single gene, and thus in the mRNA produced from a mutant gene's transcripts. A study in 2005 involving probabilistic analyses indicated that greater than 60% of human disease-causing mutations affect splicing rather than directly affecting coding sequences.^[42] A more recent study indicates that one-third of all hereditary diseases are likely to have a splicing component.^[20] Regardless of exact percentage, a number of splicing-related diseases do exist.^[43] As described below, a prominent example of splicing-related diseases is cancer.

Abnormally spliced mRNAs are also found in a high proportion of cancerous cells.^[44][45][46] Combined RNA-Seq and proteomics analyses have revealed striking differential expression of splice isoforms of key proteins in important cancer pathways.^[47] It is not always clear whether such aberrant patterns of splicing contribute to the cancerous growth, or are merely consequence of cellular abnormalities associated with cancer. For certain types of cancer, like in colorectal and prostate, the number of splicing errors per cancer has been shown to vary greatly between individual cancers, a phenomenon referred to as transcriptome instability.^[48][49] Transcriptome instability has further been shown to correlate grealty with reduced expression level of splicing factor genes. Mutation of DNMT3A has been demonstrated to contribute to hematologic malignancies, and that DNMT3A-mutated cell lines exhibit transcriptome instability as compared to their isogenic wildtype counterparts.^[50]

In fact, there is actually a reduction of alternative splicing in cancerous cells compared to normal ones, and the types of splicing differ; for instance, cancerous cells show higher levels of intron retention than normal cells, but lower levels of exon skipping.^[51] Some of the differences in splicing in cancerous cells may be due to the high frequency of somatic mutations in splicing factor genes,^[46] and some may result from changes in phosphorylation of trans-acting splicing factors.^[33] Others may be produced by changes in the relative amounts of splicing factors produced; for instance, breast cancer cells have been shown to have increased levels of the splicing factor SF2/ASF.^[52] One study found that a relatively small percentage (383 out of over 26000) of alternative splicing variants were significantly higher in frequency in tumor cells than normal cells, suggesting that there is a limited set of genes which, when mis-spliced, contribute to tumor development.^[53] It is believed however that the deleterious effects of mis-spliced transcripts are usually safeguarded and eliminated by a cellular posttranscriptional quality control mechanism termed nonsense-mediated mRNA decay [NMD].^[54]

One example of a specific splicing variant associated with cancers is in one of the human DNMT genes. Three DNMT genes encode enzymes that add methyl groups to DNA, a modification that often has regulatory effects. Several abnormally spliced DNMT3B mRNAs are found in tumors and cancer cell lines. In two separate studies, expression of two of these abnormally spliced mRNAs in mammalian cells caused changes in the DNA methylation patterns in those cells. Cells with one of the abnormal mRNAs also grew twice as fast as control cells, indicating a direct contribution to tumor development by this product.^[33]

Another example is the Ron (MST1R) proto-oncogene. An important property of cancerous cells is their ability to move and invade normal tissue. Production of an abnormally spliced transcript of Ron has been found to be associated with increased levels of the SF2/ASF in breast cancer cells. The abnormal isoform of the Ron protein encoded by this mRNA leads to cell motility.^[52]

Overexpression of a truncated splice variant of the FOSB gene – ΔFosB – in a specific population of neurons in the nucleus accumbens has been identified as the causal mechanism involved in the induction and maintenance of an addiction to drugs and natural rewards.^{[55][56][57][58]}

Recent provocative studies point to a key function of chromatin structure and histone modifications in alternative splicing regulation. These insights suggest that epigenetic regulation determines not only what parts of the genome are expressed but also how they are spliced.^[59]

Transcriptome-wide analysis of alternative splicing is typically performed by high-throughput RNA-sequencing. Most commonly, by short-read sequencing, such as by Illumina instrumentation. But even more informative, by long-read sequencing, such as by Nanopore or PacBio instrumentation. Transcriptome-wide analyses can for example be used to measure the amount of deviating alternative splicing, such as in a cancer cohort.^[60]

Deep sequencing technologies have been used to conduct genome-wide analyses of both unprocessed and processed mRNAs; thus providing insights into alternative splicing. For example, results from use of deep sequencing indicate that, in humans, an estimated 95% of transcripts from multiexon genes undergo alternative splicing, with a number of pre-mRNA transcripts spliced in a tissue-specific manner.^[2] Functional genomics and computational approaches based on multiple instance learning have also been developed to integrate RNA-seq data to predict functions for alternatively spliced isoforms.^[36] Deep sequencing has also aided in the in vivo detection of the transient lariats that are released during splicing, the determination of branch site sequences, and the large-scale mapping of branchpoints in human pre-mRNA transcripts.^[61]

More historically, alternatively spliced transcripts have been found by comparing EST sequences, but this requires sequencing of very large numbers of ESTs. Most EST libraries come from a very limited number of tissues, so tissue-specific splice variants are likely to be missed in any case. High-throughput approaches to investigate splicing have, however, been developed, such as: DNA microarray-based analyses, RNA-binding assays, and deep sequencing. These methods can be used to screen for polymorphisms or mutations in or around splicing elements that affect protein binding. When combined with splicing assays, including in vivo reporter gene assays, the functional effects of polymorphisms or mutations on the splicing of pre-mRNA transcripts can then be analyzed.^[20][23][62]

In microarray analysis, arrays of DNA fragments representing individual exons (e.g. Affymetrix exon microarray) or exon/exon boundaries (e.g. arrays from ExonHit or Jivan) have been used. The array is then probed with labeled cDNA from tissues of interest. The probe cDNAs bind to DNA from the exons that are included in mRNAs in their tissue of origin, or to DNA from the boundary where two exons have been joined. This can reveal the presence of particular alternatively spliced mRNAs.^[63]

CLIP (Cross-linking and immunoprecipitation) uses UV radiation to link proteins to RNA molecules in a tissue during splicing. A trans-acting splicing regulatory protein of interest is then precipitated using specific antibodies. When the RNA attached to that protein is isolated and cloned, it reveals the target sequences for that protein.^[64] Another method for identifying RNA-binding proteins and mapping their binding to pre-mRNA transcripts is "Microarray Evaluation of Genomic Aptamers by shift (MEGAshift)".net^[65] This method involves an adaptation of the "Systematic Evolution of Ligands by Exponential Enrichment (SELEX)" method^[66] together with a microarray-based readout. Use of the MEGAshift method has provided insights into the regulation of alternative splicing by allowing for the identification of sequences in pre-mRNA transcripts surrounding alternatively spliced exons that mediate binding to different splicing factors, such as ASF/SF2 and PTB.^[67] This approach has also been used to aid in determining the relationship between RNA secondary structure and the binding of splicing factors.^[22]

Use of reporter assays makes it possible to find the splicing proteins involved in a specific alternative splicing event by constructing reporter genes that will express one of two different fluorescent proteins depending on the splicing reaction that occurs. This method has been used to isolate mutants affecting splicing and thus to identify novel splicing regulatory proteins inactivated in those mutants.^[64]

Recent advancements in protein structure prediction have facilitated the development of new tools for genome annotation and alternative splicing anlaysis. For instance, isoform.io, a platform guided by protein structure predictions, has evaluated hundreds of thousands of isoforms of human protein-coding genes assembled from numerous RNA sequencing experiments across a variety of human tissues. This comprehensive analysis has led to the identification of numerous isoforms with more confidently predicted structure and potentially superior function compared to canonical isoforms in the latest human gene database. By integrating structural predictions with expression and evolutionary evidence, this approach has demonstrated the potential of protein structure prediction as a tool for refining the annotation of the human genome.^[68]

There is a collection of alternative splicing databases.^[69][70][71] These databases are useful for finding genes having pre-mRNAs undergoing alternative splicing and alternative splicing events or to study the functional impact of alternative splicing.

• AspicDB database
• Intronerator database
• ProSAS database

• Exitron
• Polyadenylation § Alternative polyadenylation
• Trans-splicing