Regulated splicing of large exons is linked to phase-separation of vertebrate transcription factors.

Although large exons cannot be readily recognized by the spliceosome, many are evolutionarily conserved and constitutively spliced for inclusion in the processed transcript. Furthermore, whether large exons may be enriched in a certain subset of proteins, or mediate specific functions, has remained unclear. Here, we identify a set of nearly 3,000 SRSF3-dependent large constitutive exons (S3-LCEs) in human and mouse cells. These exons are enriched for cytidine-rich sequence motifs, which bind and recruit the splicing factors hnRNP K and SRSF3. We find that hnRNP K suppresses S3-LCE splicing, an effect that is mitigated by SRSF3 to thus achieve constitutive splicing of S3-LCEs. S3-LCEs are enriched in genes for components of transcription machineries, including mediator and BAF complexes, and frequently contain intrinsically disordered regions (IDRs). In a subset of analyzed S3-LCE-containing transcription factors, SRSF3 depletion leads to deletion of the IDRs due to S3-LCE exon skipping, thereby disrupting phase-separated assemblies of these factors. Cytidine enrichment in large exons introduces proline/serine codon bias in intrinsically disordered regions and appears to have been evolutionarily acquired in vertebrates. We propose that layered splicing regulation by hnRNP K and SRSF3 ensures proper phase-separation of these S3-LCE-containing transcription factors in vertebrates.

Extremely low-frequency electromagnetic field induces acetylation of heat shock proteins and enhances protein folding.

The pervasive weak electromagnetic fields (EMF) inundate the industrialized society, but the biological effects of EMF as weak as 10 µT have been scarcely analyzed. Heat shock proteins (HSPs) are molecular chaperones that mediate a sequential stress response. HSP70 and HSP90 provide cells under undesirable situations with either assisting covalent folding of proteins or degrading improperly folded proteins in an ATP-dependent manner. Here we examined the effect of extremely low-frequency (ELF)-EMF on AML12 and HEK293 cells. Although the protein expression levels of HSP70 and HSP90 were reduced after an exposure to ELF-EMF for 3 h, acetylations of HSP70 and HSP90 were increased, which was followed by an enhanced binding affinities of HSP70 and HSP90 for HSP70/HSP90-organizing protein (HOP/STIP1). After 3 h exposure to ELF-EMF, the amount of mitochondria was reduced but the ATP level and the maximal mitochondrial oxygen consumption were increased, which was followed by the reduced protein aggregates and the increased cell viability. Thus, ELF-EMF exposure for 3 h activated acetylation of HSPs to enhance protein folding, which was returned to the basal level at 12 h. The proteostatic effects of ELF-EMF will be able to be applied to treat pathological states in humans.

Position-specific binding of FUS to nascent RNA regulates mRNA length.

More than half of all human genes produce prematurely terminated polyadenylated short mRNAs. However, the underlying mechanisms remain largely elusive. CLIP-seq (cross-linking immunoprecipitation [CLIP] combined with deep sequencing) of FUS (fused in sarcoma) in neuronal cells showed that FUS is frequently clustered around an alternative polyadenylation (APA) site of nascent RNA. ChIP-seq (chromatin immunoprecipitation [ChIP] combined with deep sequencing) of RNA polymerase II (RNAP II) demonstrated that FUS stalls RNAP II and prematurely terminates transcription. When an APA site is located upstream of an FUS cluster, FUS enhances polyadenylation by recruiting CPSF160 and up-regulates the alternative short transcript. In contrast, when an APA site is located downstream from an FUS cluster, polyadenylation is not activated, and the RNAP II-suppressing effect of FUS leads to down-regulation of the alternative short transcript. CAGE-seq (cap analysis of gene expression [CAGE] combined with deep sequencing) and PolyA-seq (a strand-specific and quantitative method for high-throughput sequencing of 3' ends of polyadenylated transcripts) revealed that position-specific regulation of mRNA lengths by FUS is operational in two-thirds of transcripts in neuronal cells, with enrichment in genes involved in synaptic activities.

tRIP-seq reveals repression of premature polyadenylation by co-transcriptional FUS-U1 snRNP assembly.

RNA processing occurs co-transcriptionally through the dynamic recruitment of RNA processing factors to RNA polymerase II (RNAPII). However, transcriptome-wide identification of protein-RNA interactions specifically assembled on transcribing RNAPII is challenging. Here, we develop the targeted RNA immunoprecipitation sequencing (tRIP-seq) method that detects protein-RNA interaction sites in thousands of cells. The high sensitivity of tRIP-seq enables identification of protein-RNA interactions at functional subcellular levels. Application of tRIP-seq to the FUS-RNA complex in the RNAPII machinery reveals that FUS binds upstream of alternative polyadenylation (APA) sites of nascent RNA bound to RNAPII, which retards RNAPII and suppresses the recognition of the polyadenylation signal by CPSF. Further tRIP-seq analyses demonstrate that the repression of APA is achieved by a complex composed of FUS and U1 snRNP on RNAPII, but not by either one alone. Moreover, our analysis reveals that FUS mutations in familial amyotrophic lateral sclerosis (ALS) that impair the FUS-U1 snRNP interaction aberrantly activate the APA sites. tRIP-seq provides new insights into the regulatory mechanism of co-transcriptional RNA processing by RNA processing factors.

Molecular Hydrogen Enhances Proliferation of Cancer Cells That Exhibit Potent Mitochondrial Unfolded Protein Response.

Molecular hydrogen ameliorates pathological states in a variety of human diseases, animal models, and cell models, but the effects of hydrogen on cancer have been rarely reported. In addition, the molecular mechanisms underlying the effects of hydrogen remain mostly unelucidated. We found that hydrogen enhances proliferation of four out of seven human cancer cell lines (the responders). The proliferation-promoting effects were not correlated with basal levels of cellular reactive oxygen species. Expression profiling of the seven cells showed that the responders have higher gene expression of mitochondrial electron transport chain (ETC) molecules than the non-responders. In addition, the responders have higher mitochondrial mass, higher mitochondrial superoxide, higher mitochondrial membrane potential, and higher mitochondrial spare respiratory capacity than the non-responders. In the responders, hydrogen provoked mitochondrial unfolded protein response (mtUPR). Suppression of cell proliferation by rotenone, an inhibitor of mitochondrial ETC complex I, was rescued by hydrogen in the responders. Hydrogen triggers mtUPR and induces cell proliferation in cancer cells that have high basal and spare mitochondrial ETC activities.

Competitive regulation of alternative splicing and alternative polyadenylation by hnRNP H and CstF64 determines acetylcholinesterase isoforms.

Acetylcholinesterase (AChE), encoded by the ACHE gene, hydrolyzes the neurotransmitter acetylcholine to terminate synaptic transmission. Alternative splicing close to the 3΄ end generates three distinct isoforms of AChET, AChEH and AChER. We found that hnRNP H binds to two specific G-runs in exon 5a of human ACHE and activates the distal alternative 3΄ splice site (ss) between exons 5a and 5b to generate AChET. Specific effect of hnRNP H was corroborated by siRNA-mediated knockdown and artificial tethering of hnRNP H. Furthermore, hnRNP H competes for binding of CstF64 to the overlapping binding sites in exon 5a, and suppresses the selection of a cryptic polyadenylation site (PAS), which additionally ensures transcription of the distal 3΄ ss required for the generation of AChET. Expression levels of hnRNP H were positively correlated with the proportions of the AChET isoform in three different cell lines. HnRNP H thus critically generates AChET by enhancing the distal 3΄ ss and by suppressing the cryptic PAS. Global analysis of CLIP-seq and RNA-seq also revealed that hnRNP H competitively regulates alternative 3΄ ss and alternative PAS in other genes. We propose that hnRNP H is an essential factor that competitively regulates alternative splicing and alternative polyadenylation.

Splicing regulation and dysregulation of cholinergic genes expressed at the neuromuscular junction.

We humans have evolved by acquiring diversity of alternative RNA metabolisms including alternative means of splicing and transcribing non-coding genes, and not by acquiring new coding genes. Tissue-specific and developmental stage-specific alternative RNA splicing is achieved by tightly regulated spatiotemporal regulation of expressions and activations of RNA-binding proteins that recognize their cognate splicing cis-elements on nascent RNA transcripts. Genes expressed at the neuromuscular junction are also alternatively spliced. In addition, germline mutations provoke aberrant splicing by compromising binding of RNA-binding proteins, and cause congenital myasthenic syndromes (CMS). We present physiological splicing mechanisms of genes for agrin (AGRN), acetylcholinesterase (ACHE), MuSK (MUSK), acetylcholine receptor (AChR) α1 subunit (CHRNA1), and collagen Q (COLQ) in human, and their aberration in diseases. Splicing isoforms of AChET , AChEH , and AChER are generated by hnRNP H/F. Skipping of MUSK exon 10 makes a Wnt-insensitive MuSK isoform, which is unique to human. Skipping of exon 10 is achieved by coordinated binding of hnRNP C, YB-1, and hnRNP L to exon 10. Exon P3A of CHRNA1 is alternatively included to generate a non-functional AChR α1 subunit in human. Molecular dissection of splicing mutations in patients with CMS reveals that exon P3A is alternatively skipped by hnRNP H, polypyrimidine tract-binding protein 1, and hnRNP L. Similarly, analysis of an exonic mutation in COLQ exon 16 in a CMS patient discloses that constitutive splicing of exon 16 requires binding of serine arginine-rich splicing factor 1. Intronic and exonic splicing mutations in CMS enable us to dissect molecular mechanisms underlying alternative and constitutive splicing of genes expressed at the neuromuscular junction. This is an article for the special issue XVth International Symposium on Cholinergic Mechanisms.

IntSplice: prediction of the splicing consequences of intronic single-nucleotide variations in the human genome.

Precise spatiotemporal regulation of splicing is mediated by splicing cis-elements on pre-mRNA. Single-nucleotide variations (SNVs) affecting intronic cis-elements possibly compromise splicing, but no efficient tool has been available to identify them. Following an effect-size analysis of each intronic nucleotide on annotated alternative splicing, we extracted 105 parameters that could affect the strength of the splicing signals. However, we could not generate reliable support vector regression models to predict the percent-splice-in (PSI) scores for normal human tissues. Next, we generated support vector machine (SVM) models using 110 parameters to directly differentiate pathogenic SNVs in the Human Gene Mutation Database and normal SNVs in the dbSNP database, and we obtained models with a sensitivity of 0.800±0.041 (mean and s.d.) and a specificity of 0.849±0.021. Our IntSplice models were more discriminating than SVM models that we generated with Shapiro-Senapathy score and MaxEntScan::score3ss. We applied IntSplice to a naturally occurring and nine artificial intronic mutations in RAPSN causing congenital myasthenic syndrome. IntSplice correctly predicted the splicing consequences for nine of the ten mutants. We created a web service program, IntSplice (http://www.med.nagoya-u.ac.jp/neurogenetics/IntSplice) to predict splicing-affecting SNVs at intronic positions from -50 to -3.

H-InvDB in 2013: an omics study platform for human functional gene and transcript discovery.

H-InvDB (http://www.h-invitational.jp/) is a comprehensive human gene database started in 2004. In the latest version, H-InvDB 8.0, a total of 244 709 human complementary DNA was mapped onto the hg19 reference genome and 43 829 gene loci, including nonprotein-coding ones, were identified. Of these loci, 35 631 were identified as potential protein-coding genes, and 22 898 of these were identical to known genes. In our analysis, 19 309 annotated genes were specific to H-InvDB and not found in RefSeq and Ensembl. In fact, 233 genes of the 19 309 turned out to have protein functions in this version of H-InvDB; they were annotated as unknown protein functions in the previous version. Furthermore, 11 genes were identified as known Mendelian disorder genes. It is advantageous that many biologically functional genes are hidden in the H-InvDB unique genes. As large-scale proteomic projects have been conducted to elucidate the functions of all human proteins, we have enhanced the proteomic information with an advanced protein view and new subdatabase of protein complexes (Protein Complex Database with quality index). We propose that H-InvDB is an important resource for finding novel candidate targets for medical care and drug development.

Muscle-specific lack of GFPT1 in knock-in mice triggers ER stress to alleviate misfolded proteins.

Pathogenic variants in GFPT1, encoding a key enzyme to synthesize UDP-N-acetylglucosamine (UDP-GlcNAc), cause congenital myasthenic syndrome (CMS). We made a knock-in (KI) mouse model carrying a frameshift variant in Gfpt1 exon 9 simulating a CMS patient. As Gfpt1 exon 9 is exclusively included in striated muscles, Gfpt1-KI mice were deficient for Gfpt1 only in skeletal muscles. In Gfpt1-KI mice, (i) UDP-HexNAc, CMP-NeuAc, and protein O-GlcNAcylations were reduced in skeletal muscles; (ii) aged Gfpt1-KI mice showed poor exercise performance and abnormal neuromuscular junction structures; and (iii) markers for unfolded protein response (UPR) were elevated in skeletal muscles. Denervation-mediated enhancement of ER stress in Gfpt1-KI mice facilitated protein folding, ubiquitin-proteasome degradation, and apoptosis, whereas autophagy was not induced and protein aggregates were markedly increased. Lack of autophagy was accounted for by enhanced degradation of FoxO1 by increased Xbp1-s/u proteins. Similarly, in Gfpt1-silenced C2C12 myotubes, ER stress exacerbated protein aggregates and activated apoptosis, but autophagy was attenuated. In both skeletal muscles in Gfpt1-KI mice and Gfpt1-silenced C2C12 myotubes, maladaptive UPR failed to eliminate protein aggregates and provoked apoptosis.

Calcitriol ameliorates motor deficits and prolongs survival of Chrne-deficient mouse, a model for congenital myasthenic syndrome, by inducing Rspo2.

Signal transduction at the neuromuscular junction (NMJ) is compromised in a diverse array of diseases including congenital myasthenic syndromes (CMS). Germline mutations in CHRNE encoding the acetylcholine receptor (AChR) ε subunit are the most common cause of CMS. An active form of vitamin D, calcitriol, binds to vitamin D receptor (VDR) and regulates gene expressions. We found that calcitriol enhanced MuSK phosphorylation, AChR clustering, and myotube twitching in co-cultured C2C12 myotubes and NSC34 motor neurons. RNA-seq analysis of co-cultured cells showed that calcitriol increased the expressions of Rspo2, Rapsn, and Dusp6. ChIP-seq of VDR revealed that VDR binds to a region approximately 15 â€‹kbp upstream to Rspo2. Biallelic deletion of the VDR-binding site of Rspo2 by CRISPR/Cas9 in C2C12 myoblasts/myotubes nullified the calcitriol-mediated induction of Rspo2 expression and MuSK phosphorylation. We generated Chrne knockout (Chrne KO) mouse by CRISPR/Cas9. Intraperitoneal administration of calcitriol markedly increased the number of AChR clusters, as well as the area, the intensity, and the number of synaptophysin-positive synaptic vesicles, in Chrne KO mice. In addition, calcitriol ameliorated motor deficits and prolonged survival of Chrne KO mice. In the skeletal muscle, calcitriol increased the gene expressions of Rspo2, Rapsn, and Dusp6. We propose that calcitriol is a potential therapeutic agent for CMS and other diseases with defective neuromuscular signal transmission.

Full-length transcriptome-based H-InvDB throws a new light on chromosome-centric proteomics.

H-Invitational Database (H-InvDB; http://hinv.jp/ ) is an integrated database of all human genes and transcripts that started in an international collaborative research project for establishing a functional annotation database of human full-length cDNAs. Because H-InvDB contains an abundance of information for human transcripts, including not only well-characterized protein-coding transcripts but also those without experimental evidence at the protein level, this will be a useful information resource for identifying novel and uncharacterized human proteins (so-called missing proteins). By extending predicted protein data in H-InvDB, we developed the H-Inv Extended Protein Database (H-EPD; http://hinv.jp/hinv/h-epd/ ). From now on, we plan to carry out a database-driven proteome research that makes full use of H-EPD to promote discoveries in the current and future C-HPP. Furthermore, we will push forward with the integration of genome, transcriptome, and proteome databases using a unique tool for connecting distributed databases and would like to develop a knowledge discovery system by incorporating data mining tools.

FexSplice: A LightGBM-Based Model for Predicting the Splicing Effect of a Single Nucleotide Variant Affecting the First Nucleotide G of an Exon.

Single nucleotide variants (SNVs) affecting the first nucleotide G of an exon (Fex-SNVs) identified in various diseases are mostly recognized as missense or nonsense variants. Their effect on pre-mRNA splicing has been seldom analyzed, and no curated database is available. We previously reported that Fex-SNVs affect splicing when the length of the polypyrimidine tract is short or degenerate. However, we cannot readily predict the splicing effects of Fex-SNVs. We here scrutinized the available literature and identified 106 splicing-affecting Fex-SNVs based on experimental evidence. We similarly identified 106 neutral Fex-SNVs in the dbSNP database with a global minor allele frequency (MAF) of more than 0.01 and less than 0.50. We extracted 115 features representing the strength of splicing cis-elements and developed machine-learning models with support vector machine, random forest, and gradient boosting to discriminate splicing-affecting and neutral Fex-SNVs. Gradient boosting-based LightGBM outperformed the other two models, and the length and nucleotide compositions of the polypyrimidine tract played critical roles in the discrimination. Recursive feature elimination showed that the LightGBM model using 15 features achieved the best performance with an accuracy of 0.80 ± 0.12 (mean and SD), a Matthews Correlation Coefficient (MCC) of 0.57 ± 0.15, an area under the curve of the receiver operating characteristics curve (AUROC) of 0.86 ± 0.08, and an area under the curve of the precision-recall curve (AUPRC) of 0.87 ± 0.09 using a 10-fold cross-validation. We developed a web service program, named FexSplice that accepts a genomic coordinate either on GRCh37/hg19 or GRCh38/hg38 and returns a predicted probability of aberrant splicing of A, C, and T variants.

Dysregulation of Aldh1a2 underlies motor neuron degeneration in spinal muscular atrophy.

Lower motor neuron degeneration is the pathological hallmark of spinal muscular atrophy (SMA), a hereditary motor neuron disease caused by loss of the SMN1 gene and the resulting deficiency of ubiquitously expressed SMN protein. The molecular mechanisms underlying motor neuron degeneration, however, remain elusive. To clarify the cell-autonomous defect in developmental processes, we here performed transcriptome analyses of isolated embryonic motor neurons of SMA model mice to explore mechanisms of dysregulation of cell-type-specific gene expression. Of 12 identified genes that were differentially expressed between the SMA and control motor neurons, we focused on Aldh1a2, an essential gene for lower motor neuron development. In primary spinal motor neuron cultures, knockdown of Aldh1a2 led to the formation of axonal spheroids and neurodegeneration, reminiscent of the histopathological changes observed in human and animal cellular models. Conversely, Aldh1a2 rescued these pathological features in spinal motor neurons derived from SMA mouse embryos. Our findings suggest that developmental defects due to Aldh1a2 dysregulation enhances lower motor neuron vulnerability in SMA.

Splicing regulation of GFPT1 muscle-specific isoform and its roles in glucose metabolisms and neuromuscular junction.

Glutamine:fructose-6-phosphate transaminase 1 (GFPT1) is the rate-limiting enzyme of the hexosamine biosynthetic pathway (HBP). A 54-bp exon 9 of GFPT1 is specifically included in skeletal and cardiac muscles to generate a long isoform of GFPT1 (GFPT1-L). We showed that SRSF1 and Rbfox1/2 cooperatively enhance, and hnRNP H/F suppresses, the inclusion of human GFPT1 exon 9 by modulating recruitment of U1 snRNP. Knockout (KO) of GFPT1-L in skeletal muscle markedly increased the amounts of GFPT1 and UDP-HexNAc, which subsequently suppressed the glycolytic pathway. Aged KO mice showed impaired insulin-mediated glucose uptake, as well as muscle weakness and fatigue likely due to abnormal formation and maintenance of the neuromuscular junction. Taken together, GFPT1-L is likely to be acquired in evolution in mammalian striated muscles to attenuate the HBP for efficient glycolytic energy production, insulin-mediated glucose uptake, and the formation and maintenance of the neuromuscular junction.

H-DBAS: human-transcriptome database for alternative splicing: update 2010.

H-DBAS (http://h-invitational.jp/h-dbas/) is a specialized database for human alternative splicing (AS) based on H-Invitational full-length cDNAs. In this update, for better annotations of AS events, we correlated RNA-Seq tag information to the AS exons and splice junctions. We generated a total of 148,376,598 RNA-Seq tags from RNAs extracted from cytoplasmic, nuclear and polysome fractions. Analysis of the RNA-Seq tags allowed us to identify 90,900 exons that are very likely to be used for protein synthesis. On the other hand, 254 AS junctions of human RefSeq transcripts are unique to nuclear RNA and may not have any translational consequences. We also present a new comparative genomics viewer so that users can empirically understand the evolutionary turnover of AS. With the unique experimental data closely connected with intensively curated cDNA information, H-DBAS provides a unique platform for the analysis of complex AS.

H-InvDB in 2009: extended database and data mining resources for human genes and transcripts.

We report the extended database and data mining resources newly released in the H-Invitational Database (H-InvDB; http://www.h-invitational.jp/). H-InvDB is a comprehensive annotation resource of human genes and transcripts, and consists of two main views and six sub-databases. The latest release of H-InvDB (release 6.2) provides the annotation for 219,765 human transcripts in 43,159 human gene clusters based on human full-length cDNAs and mRNAs. H-InvDB now provides several new annotation features, such as mapping of microarray probes, new gene models, relation to known ncRNAs and information from the Glycogene database. H-InvDB also provides useful data mining resources-'Navigation search', 'H-InvDB Enrichment Analysis Tool (HEAT)' and web service APIs. 'Navigation search' is an extended search system that enables complicated searches by combining 16 different search options. HEAT is a data mining tool for automatically identifying features specific to a given human gene set. HEAT searches for H-InvDB annotations that are significantly enriched in a user-defined gene set, as compared with the entire H-InvDB representative transcripts. H-InvDB now has web service APIs of SOAP and REST to allow the use of H-InvDB data in programs, providing the users extended data accessibility.

Extremely low-frequency pulses of faint magnetic field induce mitophagy to rejuvenate mitochondria.

Humans are frequently exposed to time-varying and static weak magnetic fields (WMF). However, the effects of faint magnetic fields, weaker than the geomagnetic field, have been scarcely reported. Here we show that extremely low-frequency (ELF)-WMF, comprised of serial pulses of 10 µT intensity at 1-8 Hz, which is three or more times weaker than the geomagnetic field, reduces mitochondrial mass to 70% and the mitochondrial electron transport chain (ETC) complex II activity to 88%. Chemical inhibition of electron flux through the mitochondrial ETC complex II nullifies the effect of ELF-WMF. Suppression of ETC complex II subsequently induces mitophagy by translocating parkin and PINK1 to the mitochondria and by recruiting LC3-II. Thereafter, mitophagy induces PGC-1α-mediated mitochondrial biogenesis to rejuvenate mitochondria. The lack of PINK1 negates the effect of ELF-WMF. Thus, ELF-WMF may be applicable for the treatment of human diseases that exhibit compromised mitochondrial homeostasis, such as Parkinson's disease.

IntSplice2: Prediction of the Splicing Effects of Intronic Single-Nucleotide Variants Using LightGBM Modeling.

Prediction of the effect of a single-nucleotide variant (SNV) in an intronic region on aberrant pre-mRNA splicing is challenging except for an SNV affecting the canonical GU/AG splice sites (ss). To predict pathogenicity of SNVs at intronic positions -50 (Int-50) to -3 (Int-3) close to the 3' ss, we developed light gradient boosting machine (LightGBM)-based IntSplice2 models using pathogenic SNVs in the human gene mutation database (HGMD) and ClinVar and common SNVs in dbSNP with 0.01 ≤ minor allelic frequency (MAF) < 0.50. The LightGBM models were generated using features representing splicing cis-elements. The average recall/sensitivity and specificity of IntSplice2 by fivefold cross-validation (CV) of the training dataset were 0.764 and 0.884, respectively. The recall/sensitivity of IntSplice2 was lower than the average recall/sensitivity of 0.800 of IntSplice that we previously made with support vector machine (SVM) modeling for the same intronic positions. In contrast, the specificity of IntSplice2 was higher than the average specificity of 0.849 of IntSplice. For benchmarking (BM) of IntSplice2 with IntSplice, we made a test dataset that was not used to train IntSplice. After excluding the test dataset from the training dataset, we generated IntSplice2-BM and compared it with IntSplice using the test dataset. IntSplice2-BM was superior to IntSplice in all of the seven statistical measures of accuracy, precision, recall/sensitivity, specificity, F1 score, negative predictive value (NPV), and matthews correlation coefficient (MCC). We made the IntSplice2 web service at https://www.med.nagoya-u.ac.jp/neurogenetics/IntSplice2.

A comprehensive survey of human polymorphisms at conserved splice dinucleotides and its evolutionary relationship with alternative splicing.

BACKGROUND: Alternative splicing (AS) is a key molecular process that endows biological functions with diversity and complexity. Generally, functional redundancy leads to the generation of new functions through relaxation of selective pressure in evolution, as exemplified by duplicated genes. It is also known that alternatively spliced exons (ASEs) are subject to relaxed selective pressure. Within consensus sequences at the splice junctions, the most conserved sites are dinucleotides at both ends of introns (splice dinucleotides). However, a small number of single nucleotide polymorphisms (SNPs) occur at splice dinucleotides. An intriguing question relating to the evolution of AS diversity is whether mutations at splice dinucleotides are maintained as polymorphisms and produce diversity in splice patterns within the human population. We therefore surveyed validated SNPs in the database dbSNP located at splice dinucleotides of all human genes that are defined by the H-Invitational Database. RESULTS: We found 212 validated SNPs at splice dinucleotides (sdSNPs); these were confirmed to be consistent with the GT-AG rule at either allele. Moreover, 53 of them were observed to neighbor ASEs (AE dinucleotides). No significant differences were observed between sdSNPs at AE dinucleotides and those at constitutive exons (CE dinucleotides) in SNP properties including average heterozygosity, SNP density, ratio of predicted alleles consistent with the GT-AG rule, and scores of splice sites formed with the predicted allele. We also found that the proportion of non-conserved exons was higher for exons with sdSNPs than for other exons. CONCLUSIONS: sdSNPs are found at CE dinucleotides in addition to those at AE dinucleotides, suggesting two possibilities. First, sdSNPs at CE dinucleotides may be robust against sdSNPs because of unknown mechanisms. Second, similar to sdSNPs at AE dinucleotides, those at CE dinucleotides cause differences in AS patterns because of the arbitrariness in the classification of exons into alternative and constitutive type that varies according to the dataset. Taking into account the absence of differences in sdSNP properties between those at AE and CE dinucleotides, the increased proportion of non-conserved exons found in exons flanked by sdSNPs suggests the hypothesis that sdSNPs are maintained at the splice dinucleotides of newly generated exons at which negative selection pressure is relaxed.