Quantitative evaluation of all hexamers as exonic splicing elements

Splicing analysis of STAT3 tandem donor suggests non-canonical binding registers for U1 and U6 snRNAs

Tandem donor splice sites (5'ss) are unique regions with at least two GU dinucleotides serving as splicing cleavage sites. The Δ3 tandem 5'ss are a specific subclass of 5'ss separated by 3 nucleotides which can affect protein function by inserting/deleting a single amino acid. One 5'ss is typically preferred, yet factors governing particular 5'ss choice are not fully understood. A highly conserved exon 21 of the STAT3 gene was chosen as a model to study Δ3 tandem 5'ss splicing mechanisms. Based on multiple lines of experimental evidence, endogenous U1 snRNA most likely binds only to the upstream 5'ss. However, the downstream 5'ss is used preferentially, and the splice site choice is not dependent on the exact U1 snRNA binding position. Downstream 5'ss usage was sensitive to exact nucleotide composition and dependent on the presence of downstream regulatory region. The downstream 5'ss usage could be best explained by two novel interactions with endogenous U6 snRNA. U6 snRNA enables the downstream 5'ss usage in STAT3 exon 21 by two mechanisms: (i) binding in a novel non-canonical register and (ii) establishing extended Watson-Crick base pairing with the downstream regulatory region. This study suggests that U6:5'ss interaction is more flexible than previously thought.

The debranching enzyme Dbr1 regulates lariat turnover and intron splicing

The majority of genic transcription is intronic. Introns are removed by splicing as branched lariat RNAs which require rapid recycling. The branch site is recognized during splicing catalysis and later debranched by Dbr1 in the rate-limiting step of lariat turnover. Through generation of a viable DBR1 knockout cell line, we find the predominantly nuclear Dbr1 enzyme to encode the sole debranching activity in human cells. Dbr1 preferentially debranches substrates that contain canonical U2 binding motifs, suggesting that branchsites discovered through sequencing do not necessarily represent those favored by the spliceosome. We find that Dbr1 also exhibits specificity for particular 5' splice site sequences. We identify Dbr1 interactors through co-immunoprecipitation mass spectrometry. We present a mechanistic model for Dbr1 recruitment to the branchpoint through the intron-binding protein AQR. In addition to a 20-fold increase in lariats, Dbr1 depletion increases exon skipping. Using ADAR fusions to timestamp lariats, we demonstrate a defect in spliceosome recycling. In the absence of Dbr1, spliceosomal components remain associated with the lariat for a longer period of time. As splicing is co-transcriptional, slower recycling increases the likelihood that downstream exons will be available for exon skipping.

Copper-binding proteins and exonic splicing enhancers and silencers

Eukaryotic DNA codes not only for proteins but contains a wealth of information required for accurate splicing of messenger RNA precursors and inclusion of constitutively or alternatively spliced exons in mature transcripts. This "auxiliary" splicing code has been characterized as exonic splicing enhancers and silencers (ESE and ESS). The exact interplay between protein and splicing codes is, however, poorly understood. Here, we show that exons encoding copper-coordinating amino acids in human cuproproteins lack ESEs and/or have an excess of ESSs, yet RNA sequencing and expressed sequence tags data show that they are more efficiently included in mature transcripts by the splicing machinery than average exons. Their largely constitutive inclusion in messenger RNA is facilitated by stronger splice sites, including polypyrimidine tracts, consistent with an important role of the surrounding intron architecture in ensuring high expression of metal-binding residues during evolution. ESE/ESS profiles of codons and entire exons that code for copper-coordinating residues were very similar to those encoding residues that coordinate zinc but markedly different from those that coordinate calcium. Together, these results reveal how the traditional and auxiliary splicing motifs responded to constraints of metal coordination in proteins.

Exonic splicing code and coordination of divalent metals in proteins

Exonic sequences contain both protein-coding and RNA splicing information but the interplay of the protein and splicing code is complex and poorly understood. Here, we have studied traditional and auxiliary splicing codes of human exons that encode residues coordinating two essential divalent metals at the opposite ends of the Irving-Williams series, a universal order of relative stabilities of metal-organic complexes. We show that exons encoding Zn2+-coordinating amino acids are supported much less by the auxiliary splicing motifs than exons coordinating Ca2+. The handicap of the former is compensated by stronger splice sites and uridine-richer polypyrimidine tracts, except for position -3 relative to 3' splice junctions. However, both Ca2+ and Zn2+ exons exhibit close-to-constitutive splicing in multiple tissues, consistent with their critical importance for metalloprotein function and a relatively small fraction of expendable, alternatively spliced exons. These results indicate that constraints imposed by metal coordination spheres on RNA splicing have been efficiently overcome by the plasticity of exon-intron architecture to ensure adequate metalloprotein expression.

Autonomous transposons tune their sequences to ensure somatic suppression

Transposable elements (TEs) are a major constituent of human genes, occupying approximately half of the intronic space. During pre-messenger RNA synthesis, intronic TEs are transcribed along with their host genes but rarely contribute to the final mRNA product because they are spliced out together with the intron and rapidly degraded. Paradoxically, TEs are an abundant source of RNA-processing signals through which they can create new introns1, and also functional2 or non-functional chimeric transcripts3. The rarity of these events implies the existence of a resilient splicing code that is able to suppress TE exonization without compromising host pre-mRNA processing. Here we show that SAFB proteins protect genome integrity by preventing retrotransposition of L1 elements while maintaining splicing integrity, via prevention of the exonization of previously integrated TEs. This unique dual role is possible because of L1's conserved adenosine-rich coding sequences that are bound by SAFB proteins. The suppressive activity of SAFB extends to tissue-specific, giant protein-coding cassette exons, nested genes and Tigger DNA transposons. Moreover, SAFB also suppresses LTR/ERV elements in species in which they are still active, such as mice and flies. A significant subset of splicing events suppressed by SAFB in somatic cells are activated in the testis, coinciding with low SAFB expression in postmeiotic spermatids. Reminiscent of the division of labour between innate and adaptive immune systems that fight external pathogens, our results uncover SAFB proteins as an RNA-based, pattern-guided, non-adaptive defence system against TEs in the soma, complementing the RNA-based, adaptive Piwi-interacting RNA pathway of the germline.

An intronic RNA element modulates Factor VIII exon-16 splicing

Pathogenic variants in the human Factor VIII (F8) gene cause Hemophilia A (HA). Here, we investigated the impact of 97 HA-causing single-nucleotide variants on the splicing of 11 exons from F8. For the majority of F8 exons, splicing was insensitive to the presence of HA-causing variants. However, splicing of several exons, including exon-16, was impacted by variants predicted to alter exonic splicing regulatory sequences. Using exon-16 as a model, we investigated the structure-function relationship of HA-causing variants on splicing. Intriguingly, RNA chemical probing analyses revealed a three-way junction structure at the 3'-end of intron-15 (TWJ-3-15) capable of sequestering the polypyrimidine tract. We discovered antisense oligonucleotides (ASOs) targeting TWJ-3-15 partially rescue splicing-deficient exon-16 variants by increasing accessibility of the polypyrimidine tract. The apical stem loop region of TWJ-3-15 also contains two hnRNPA1-dependent intronic splicing silencers (ISSs). ASOs blocking these ISSs also partially rescued splicing. When used in combination, ASOs targeting both the ISSs and the region sequestering the polypyrimidine tract, fully rescue pre-mRNA splicing of multiple HA-linked variants of exon-16. Together, our data reveal a putative RNA structure that sensitizes F8 exon-16 to aberrant splicing.

Benchmarking splice variant prediction algorithms using massively parallel splicing assays

BACKGROUND

Variants that disrupt mRNA splicing account for a sizable fraction of the pathogenic burden in many genetic disorders, but identifying splice-disruptive variants (SDVs) beyond the essential splice site dinucleotides remains difficult. Computational predictors are often discordant, compounding the challenge of variant interpretation. Because they are primarily validated using clinical variant sets heavily biased to known canonical splice site mutations, it remains unclear how well their performance generalizes.

RESULTS

We benchmark eight widely used splicing effect prediction algorithms, leveraging massively parallel splicing assays (MPSAs) as a source of experimentally determined ground-truth. MPSAs simultaneously assay many variants to nominate candidate SDVs. We compare experimentally measured splicing outcomes with bioinformatic predictions for 3,616 variants in five genes. Algorithms' concordance with MPSA measurements, and with each other, is lower for exonic than intronic variants, underscoring the difficulty of identifying missense or synonymous SDVs. Deep learning-based predictors trained on gene model annotations achieve the best overall performance at distinguishing disruptive and neutral variants, and controlling for overall call rate genome-wide, SpliceAI and Pangolin have superior sensitivity. Finally, our results highlight two practical considerations when scoring variants genome-wide: finding an optimal score cutoff, and the substantial variability introduced by differences in gene model annotation, and we suggest strategies for optimal splice effect prediction in the face of these issues.

CONCLUSION

SpliceAI and Pangolin show the best overall performance among predictors tested, however, improvements in splice effect prediction are still needed especially within exons.

The human genome contains over a million autonomous exons

Mammalian mRNA and lncRNA exons are often small compared to introns. The exon definition model predicts that exons splice autonomously, dependent on proximal exon sequence features, explaining their delineation within large introns. This model has not been examined on a genome-wide scale, however, leaving open the question of how often mRNA and lncRNA exons are autonomous. It is also unknown how frequently such exons can arise by chance. Here, we directly assayed large fragments (500-1000 bp) of the human genome by exon trapping, which detects exons spliced into a heterologous transgene, here designed with a large intron context. We define the trapped exons as "autonomous." We obtained ∼1.25 million trapped exons, including most known mRNA and well-annotated lncRNA internal exons, demonstrating that human exons are predominantly autonomous. mRNA exons are trapped with the highest efficiency. Nearly a million of the trapped exons are unannotated, most located in intergenic regions and antisense to mRNA, with depletion from the forward strand of introns. These exons are not conserved, suggesting they are nonfunctional and arose from random mutations. They are nonetheless highly enriched with known splicing promoting sequence features that delineate known exons. Novel autonomous exons are more numerous than annotated lncRNA exons, and computational models also indicate they will occur with similar frequency in any randomly generated sequence. These results show that most human coding exons splice autonomously, and provide an explanation for the existence of many unconserved lncRNAs, as well as a new annotation and inclusion levels of spliceable loci in the human genome.

Efficient exon skipping by base-editor-mediated abrogation of exonic splicing enhancers

Duchenne muscular dystrophy (DMD) is a severe genetic disease caused by the loss of the dystrophin protein. Exon skipping is a promising strategy to treat DMD by restoring truncated dystrophin. Here, we demonstrate that base editors (e.g., targeted AID-mediated mutagenesis [TAM]) are able to efficiently induce exon skipping by disrupting functional redundant exonic splicing enhancers (ESEs). By developing an unbiased and high-throughput screening to interrogate exonic sequences, we successfully identify novel ESEs in DMD exons 51 and 53. TAM-CBE (cytidine base editor) induces near-complete skipping of the respective exons by targeting these ESEs in patients' induced pluripotent stem cell (iPSC)-derived cardiomyocytes. Combined with strategies to disrupt splice sites, we identify suitable single guide RNAs (sgRNAs) with TAM-CBE to efficiently skip most DMD hotspot exons without substantial double-stranded breaks. Our study thus expands the repertoire of potential targets for CBE-mediated exon skipping in treating DMD and other RNA mis-splicing diseases.

PDIVAS: Pathogenicity predictor for Deep-Intronic Variants causing Aberrant Splicing

BACKGROUND

Deep-intronic variants that alter RNA splicing were ineffectively evaluated in the search for the cause of genetic diseases. Determination of such pathogenic variants from a vast number of deep-intronic variants (approximately 1,500,000 variants per individual) represents a technical challenge to researchers. Thus, we developed a Pathogenicity predictor for Deep-Intronic Variants causing Aberrant Splicing (PDIVAS) to easily detect pathogenic deep-intronic variants.

RESULTS

PDIVAS was trained on an ensemble machine-learning algorithm to classify pathogenic and benign variants in a curated dataset. The dataset consists of manually curated pathogenic splice-altering variants (SAVs) and commonly observed benign variants within deep introns. Splicing features and a splicing constraint metric were used to maximize the predictive sensitivity and specificity, respectively. PDIVAS showed an average precision of 0.92 and a maximum MCC of 0.88 in classifying these variants, which were the best of the previous predictors. When PDIVAS was applied to genome sequencing analysis on a threshold with 95% sensitivity for reported pathogenic SAVs, an average of 27 pathogenic candidates were extracted per individual. Furthermore, the causative variants in simulated patient genomes were more efficiently prioritized than the previous predictors.

CONCLUSION

Incorporating PDIVAS into variant interpretation pipelines will enable efficient detection of disease-causing deep-intronic SAVs and contribute to improving the diagnostic yield. PDIVAS is publicly available at https://github.com/shiro-kur/PDIVAS .

Deciphering RNA splicing logic with interpretable machine learning

Machine learning methods, particularly neural networks trained on large datasets, are transforming how scientists approach scientific discovery and experimental design. However, current state-of-the-art neural networks are limited by their uninterpretability: Despite their excellent accuracy, they cannot describe how they arrived at their predictions. Here, using an "interpretable-by-design" approach, we present a neural network model that provides insights into RNA splicing, a fundamental process in the transfer of genomic information into functional biochemical products. Although we designed our model to emphasize interpretability, its predictive accuracy is on par with state-of-the-art models. To demonstrate the model's interpretability, we introduce a visualization that, for any given exon, allows us to trace and quantify the entire decision process from input sequence to output splicing prediction. Importantly, the model revealed uncharacterized components of the splicing logic, which we experimentally validated. This study highlights how interpretable machine learning can advance scientific discovery.

Alu RNA fold links splicing with signal recognition particle proteins

Transcriptomic diversity in primates was considerably expanded by exonizations of intronic Alu elements. To better understand their cellular mechanisms we have used structure-based mutagenesis coupled with functional and proteomic assays to study the impact of successive primate mutations and their combinations on inclusion of a sense-oriented AluJ exon in the human F8 gene. We show that the splicing outcome was better predicted by consecutive RNA conformation changes than by computationally derived splicing regulatory motifs. We also demonstrate an involvement of SRP9/14 (signal recognition particle) heterodimer in splicing regulation of Alu-derived exons. Nucleotide substitutions that accumulated during primate evolution relaxed the conserved left-arm AluJ structure including helix H1 and reduced the capacity of SRP9/14 to stabilize the closed Alu conformation. RNA secondary structure-constrained mutations that promoted open Y-shaped conformations of the Alu made the Alu exon inclusion reliant on DHX9. Finally, we identified additional SRP9/14 sensitive Alu exons and predicted their functional roles in the cell. Together, these results provide unique insights into architectural elements required for sense Alu exonization, identify conserved pre-mRNA structures involved in exon selection and point to a possible chaperone activity of SRP9/14 outside the mammalian signal recognition particle.

Large-scale functional screen identifies genetic variants with splicing effects in modern and archaic humans

Humans coexisted and interbred with other hominins which later became extinct. These archaic hominins are known to us only through fossil records and for two cases, genome sequences. Here, we engineer Neanderthal and Denisovan sequences into thousands of artificial genes to reconstruct the pre-mRNA processing patterns of these extinct populations. Of the 5,169 alleles tested in this massively parallel splicing reporter assay (MaPSy), we report 962 exonic splicing mutations that correspond to differences in exon recognition between extant and extinct hominins. Using MaPSy splicing variants, predicted splicing variants, and splicing quantitative trait loci, we show that splice-disrupting variants experienced greater purifying selection in anatomically modern humans than that in Neanderthals. Adaptively introgressed variants were enriched for moderate-effect splicing variants, consistent with positive selection for alternative spliced alleles following introgression. As particularly compelling examples, we characterized a unique tissue-specific alternative splicing variant at the adaptively introgressed innate immunity gene TLR1, as well as a unique Neanderthal introgressed alternative splicing variant in the gene HSPG2 that encodes perlecan. We further identified potentially pathogenic splicing variants found only in Neanderthals and Denisovans in genes related to sperm maturation and immunity. Finally, we found splicing variants that may contribute to variation among modern humans in total bilirubin, balding, hemoglobin levels, and lung capacity. Our findings provide unique insights into natural selection acting on splicing in human evolution and demonstrate how functional assays can be used to identify candidate causal variants underlying differences in gene regulation and phenotype.

In silico methods for predicting functional synonymous variants

Single nucleotide variants (SNVs) contribute to human genomic diversity. Synonymous SNVs are previously considered to be "silent," but mounting evidence has revealed that these variants can cause RNA and protein changes and are implicated in over 85 human diseases and cancers. Recent improvements in computational platforms have led to the development of numerous machine-learning tools, which can be used to advance synonymous SNV research. In this review, we discuss tools that should be used to investigate synonymous variants. We provide supportive examples from seminal studies that demonstrate how these tools have driven new discoveries of functional synonymous SNVs.

Possible involvement of silent mutations in cancer pathogenesis and evolution

Recent studies have shown that some silent mutations can be harmful to various processes. In this study, we performed a comprehensive in silico analysis to elucidate the effects of silent mutations on cancer pathogenesis using exome sequencing data derived from the Cancer Genome Atlas. We focused on the codon optimality scores of silent mutations, which were defined as the difference between the optimality of synonymous codons, calculated using the codon usage table. The relationship between cancer evolution and silent mutations showed that the codon optimality score of the mutations that occurred later in carcinogenesis was significantly higher than of those that occurred earlier. In addition, mutations with higher scores were enriched in genes involved in the cell cycle and cell division, while those with lower scores were enriched in genes involved in apoptosis and cellular senescence. Our results demonstrate that some silent mutations can be involved in cancer pathogenesis.

Benchmarking splice variant prediction algorithms using massively parallel splicing assays

Background

Variants that disrupt mRNA splicing account for a sizable fraction of the pathogenic burden in many genetic disorders, but identifying splice-disruptive variants (SDVs) beyond the essential splice site dinucleotides remains difficult. Computational predictors are often discordant, compounding the challenge of variant interpretation. Because they are primarily validated using clinical variant sets heavily biased to known canonical splice site mutations, it remains unclear how well their performance generalizes.

Results

We benchmarked eight widely used splicing effect prediction algorithms, leveraging massively parallel splicing assays (MPSAs) as a source of experimentally determined ground-truth. MPSAs simultaneously assay many variants to nominate candidate SDVs. We compared experimentally measured splicing outcomes with bioinformatic predictions for 3,616 variants in five genes. Algorithms' concordance with MPSA measurements, and with each other, was lower for exonic than intronic variants, underscoring the difficulty of identifying missense or synonymous SDVs. Deep learning-based predictors trained on gene model annotations achieved the best overall performance at distinguishing disruptive and neutral variants. Controlling for overall call rate genome-wide, SpliceAI and Pangolin also showed superior overall sensitivity for identifying SDVs. Finally, our results highlight two practical considerations when scoring variants genome-wide: finding an optimal score cutoff, and the substantial variability introduced by differences in gene model annotation, and we suggest strategies for optimal splice effect prediction in the face of these issues.

Conclusion

SpliceAI and Pangolin showed the best overall performance among predictors tested, however, improvements in splice effect prediction are still needed especially within exons.

Aberrant splicing prediction across human tissues

Aberrant splicing is a major cause of genetic disorders but its direct detection in transcriptomes is limited to clinically accessible tissues such as skin or body fluids. While DNA-based machine learning models can prioritize rare variants for affecting splicing, their performance in predicting tissue-specific aberrant splicing remains unassessed. Here we generated an aberrant splicing benchmark dataset, spanning over 8.8 million rare variants in 49 human tissues from the Genotype-Tissue Expression (GTEx) dataset. At 20% recall, state-of-the-art DNA-based models achieve maximum 12% precision. By mapping and quantifying tissue-specific splice site usage transcriptome-wide and modeling isoform competition, we increased precision by threefold at the same recall. Integrating RNA-sequencing data of clinically accessible tissues into our model, AbSplice, brought precision to 60%. These results, replicated in two independent cohorts, substantially contribute to noncoding loss-of-function variant identification and to genetic diagnostics design and analytics.

Rescue of blood coagulation Factor VIII exon-16 mis-splicing by antisense oligonucleotides

The human Factor VIII ( F8 ) protein is essential for the blood coagulation cascade and specific F8 mutations cause the rare bleeding disorder Hemophilia A (HA). Here, we investigated the impact of HA-causing single-nucleotide mutations on F8 pre-mRNA splicing. We found that 14/97 (∼14.4%) coding sequence mutations tested in our study induced exon skipping. Splicing patterns of 4/11 (∼36.4%) F8 exons tested were especially sensitive to the presence of common disease-causing mutations. RNA-chemical probing analyses revealed a three-way junction structure at the 3' end of intron 15 (TWJ-3-15). TWJ-3-15 sequesters the polypyrimidine tract, a key determinant of 3' splice site strength. Using exon-16 of the F8 gene as a model, we designed specific antisense oligonucleotides (ASOs) that target TWJ-3-15 and identified three that promote the splicing of F8 exon-16. Interaction of TWJ-3-15 with ASOs increases accessibility of the polypyrimidine tract and inhibits the binding of hnRNPA1-dependent splicing silencing factors. Moreover, ASOs targeting TWJ-3-15 rescue diverse splicing-sensitive HA-causing mutations, most of which are distal to the 3' splice site being impacted. The TWJ-3-15 structure and its effect on mRNA splicing provide a model for HA etiology in patients harboring specific F8 mutations and provide a framework for precision RNA-based HA therapies.

Regulation of pre-mRNA splicing: roles in physiology and disease, and therapeutic prospects

The removal of introns from mRNA precursors and its regulation by alternative splicing are key for eukaryotic gene expression and cellular function, as evidenced by the numerous pathologies induced or modified by splicing alterations. Major recent advances have been made in understanding the structures and functions of the splicing machinery, in the description and classification of physiological and pathological isoforms and in the development of the first therapies for genetic diseases based on modulation of splicing. Here, we review this progress and discuss important remaining challenges, including predicting splice sites from genomic sequences, understanding the variety of molecular mechanisms and logic of splicing regulation, and harnessing this knowledge for probing gene function and disease aetiology and for the design of novel therapeutic approaches.

Identification and in silico characterization of CSRP3 synonymous variants in dilated cardiomyopathy

BACKGROUND

Synonymous variations have always been ignored while studying the underlying genetic mechanisms for most of the human diseases. However, recent studies have suggested that these silent changes in the genome can alter the protein expression and folding.

METHODS AND RESULTS

CSRP3, which is a well-known candidate gene associated with dilated cardiomyopathy (DCM) and hypertrophic cardiomyopathy (HCM), was screened for 100 idiopathic DCM cases and 100 controls. Three synonymous variations were identified viz., c.96G > A, p.K32=; c.336G > A, p.A112=; c.354G > A, p.E118=. A comprehensive in silico analysis was performed using various web based widely accepted tools, Mfold, Codon Usage, HSF3.1 and RNA22. Mfold predicted structural changes in all the variants except c.96 G > A (p.K32=), however it predicted changes in the stability of mRNA due to all the synonymous variants. Codon bias was observed as evident by the Relative Synonymous Codon Usage and Log Ratio of Codon Usage Frequencies. The Human Splicing Finder also predicted remarkable changes in the regulatory elements in the variants c.336G > A and c.354 G > A. The miRNA target prediction using varied modes available in RNA22 revealed that 70.6% of the target sites of miRNAs in CSRP3 were altered due to variant c.336G > A while 29.41% sites were completely lost.

CONCLUSION

Findings of the present study suggest that synonymous variants revealed striking deviations in the structural conformation of mRNA, stability of mRNA, relative synonymous codon usage, splicing and miRNA binding sites from the wild type suggesting their possible role in the pathogenesis of DCM, either by destabilizing the mRNA structure, or codon usage bias or else altering the cis-acting regulatory elements during splicing.

Computational prediction of human deep intronic variation

BACKGROUND

The adoption of whole-genome sequencing in genetic screens has facilitated the detection of genetic variation in the intronic regions of genes, far from annotated splice sites. However, selecting an appropriate computational tool to discriminate functionally relevant genetic variants from those with no effect is challenging, particularly for deep intronic regions where independent benchmarks are scarce.

RESULTS

In this study, we have provided an overview of the computational methods available and the extent to which they can be used to analyze deep intronic variation. We leveraged diverse datasets to extensively evaluate tool performance across different intronic regions, distinguishing between variants that are expected to disrupt splicing through different molecular mechanisms. Notably, we compared the performance of SpliceAI, a widely used sequence-based deep learning model, with that of more recent methods that extend its original implementation. We observed considerable differences in tool performance depending on the region considered, with variants generating cryptic splice sites being better predicted than those that potentially affect splicing regulatory elements. Finally, we devised a novel quantitative assessment of tool interpretability and found that tools providing mechanistic explanations of their predictions are often correct with respect to the ground - information, but the use of these tools results in decreased predictive power when compared to black box methods.

CONCLUSIONS

Our findings translate into practical recommendations for tool usage and provide a reference framework for applying prediction tools in deep intronic regions, enabling more informed decision-making by practitioners.

SPiP: Splicing Prediction Pipeline, a machine learning tool for massive detection of exonic and intronic variant effects on mRNA splicing

Modeling splicing is essential for tackling the challenge of variant interpretation as each nucleotide variation can be pathogenic by affecting pre-mRNA splicing via disruption/creation of splicing motifs such as 5'/3' splice sites, branch sites, or splicing regulatory elements. Unfortunately, most in silico tools focus on a specific type of splicing motif, which is why we developed the Splicing Prediction Pipeline (SPiP) to perform, in one single bioinformatic analysis based on a machine learning approach, a comprehensive assessment of the variant effect on different splicing motifs. We gathered a curated set of 4616 variants scattered all along the sequence of 227 genes, with their corresponding splicing studies. The Bayesian analysis provided us with the number of control variants, that is, variants without impact on splicing, to mimic the deluge of variants from high-throughput sequencing data. Results show that SPiP can deal with the diversity of splicing alterations, with 83.13% sensitivity and 99% specificity to detect spliceogenic variants. Overall performance as measured by area under the receiving operator curve was 0.986, better than SpliceAI and SQUIRLS (0.965 and 0.766) for the same data set. SPiP lends itself to a unique suite for comprehensive prediction of spliceogenicity in the genomic medicine era. SPiP is available at: https://sourceforge.net/projects/splicing-prediction-pipeline/.

Familial history and prevalence of BRCA1, BRCA2 and TP53 pathogenic variants in HBOC Brazilian patients from a public healthcare service

Several studies have demonstrated the cost-effectiveness of genetic testing for surveillance and treatment of carriers of germline pathogenic variants associated with hereditary breast/ovarian cancer syndrome (HBOC). In Brazil, seventy percent of the population is assisted by the public Unified Health System (SUS), where genetic testing is still unavailable. And few studies were performed regarding the prevalence of HBOC pathogenic variants in this context. Here, we estimated the prevalence of germline pathogenic variants in BRCA1, BRCA2 and TP53 genes in Brazilian patients suspected of HBOC and referred to public healthcare service. Predictive power of risk prediction models for detecting mutation carriers was also evaluated. We found that 41 out of 257 tested patients (15.9%) were carriers of pathogenic variants in the analyzed genes. Most frequent pathogenic variant was the founder Brazilian mutation TP53 c.1010G > A (p.Arg337His), adding to the accumulated evidence that supports inclusion of TP53 in routine testing of Brazilian HBOC patients. Surprisingly, BRCA1 c.5266dupC (p.Gln1756fs), a frequently reported pathogenic variant in Brazilian HBOC patients, was not observed. Regarding the use of predictive models, we found that familial history of cancer might be used to improve selection or prioritization of patients for genetic testing, especially in a context of limited resources.

Mechanism and modeling of human disease-associated near-exon intronic variants that perturb RNA splicing

It is estimated that 10%-30% of disease-associated genetic variants affect splicing. Splicing variants may generate deleteriously altered gene product and are potential therapeutic targets. However, systematic diagnosis or prediction of splicing variants is yet to be established, especially for the near-exon intronic splice region. The major challenge lies in the redundant and ill-defined branch sites and other splicing motifs therein. Here, we carried out unbiased massively parallel splicing assays on 5,307 disease-associated variants that overlapped with branch sites and collected 5,884 variants across the 5' splice region. We found that strong splice sites and exonic features preserve splicing from intronic sequence variation. Whereas the splice-altering mechanism of the 3' intronic variants is complex, that of the 5' is mainly splice-site destruction. Statistical learning combined with these molecular features allows precise prediction of altered splicing from an intronic variant. This statistical model provides the identity and ranking of biological features that determine splicing, which serves as transferable knowledge and out-performs the benchmarking predictive tool. Moreover, we demonstrated that intronic splicing variants may associate with disease risks in the human population. Our study elucidates the mechanism of splicing response of intronic variants, which classify disease-associated splicing variants for the promise of precision medicine.

Higher-order epistasis and phenotypic prediction

One core goal of genetics is to systematically understand the mapping between the DNA sequence of an organism (genotype) and its measurable characteristics (phenotype). Understanding this mapping is often challenging because of interactions between mutations, where the result of combining several different mutations can be very different than the sum of their individual effects. Here we provide a statistical framework for modeling complex genetic interactions of this type. The key idea is to ask how fast the effects of mutations change when introducing the same mutation in increasingly distant genetic backgrounds. We then propose a model for phenotypic prediction that takes into account this tendency for the effects of mutations to be more similar in nearby genetic backgrounds.

Contemporary high-throughput mutagenesis experiments are providing an increasingly detailed view of the complex patterns of genetic interaction that occur between multiple mutations within a single protein or regulatory element. By simultaneously measuring the effects of thousands of combinations of mutations, these experiments have revealed that the genotype–phenotype relationship typically reflects not only genetic interactions between pairs of sites but also higher-order interactions among larger numbers of sites. However, modeling and understanding these higher-order interactions remains challenging. Here we present a method for reconstructing sequence-to-function mappings from partially observed data that can accommodate all orders of genetic interaction. The main idea is to make predictions for unobserved genotypes that match the type and extent of epistasis found in the observed data. This information on the type and extent of epistasis can be extracted by considering how phenotypic correlations change as a function of mutational distance, which is equivalent to estimating the fraction of phenotypic variance due to each order of genetic interaction (additive, pairwise, three-way, etc.). Using these estimated variance components, we then define an empirical Bayes prior that in expectation matches the observed pattern of epistasis and reconstruct the genotype–phenotype mapping by conducting Gaussian process regression under this prior. To demonstrate the power of this approach, we present an application to the antibody-binding domain GB1 and also provide a detailed exploration of a dataset consisting of high-throughput measurements for the splicing efficiency of human pre-mRNA 5′ splice sites, for which we also validate our model predictions via additional low-throughput experiments.

Performance evaluation of differential splicing analysis methods and splicing analytics platform construction

A proportion of previously defined benign variants or variants of uncertain significance in humans, which are challenging to identify, may induce an abnormal splicing process. An increasing number of methods have been developed to predict splicing variants, but their performance has not been completely evaluated using independent benchmarks. Here, we manually sourced ∼50 000 positive/negative splicing variants from > 8000 studies and selected the independent splicing variants to evaluate the performance of prediction methods. These methods showed different performances in recognizing splicing variants in donor and acceptor regions, reminiscent of different weight coefficient applications to predict novel splicing variants. Of these methods, 66.67% exhibited higher specificities than sensitivities, suggesting that more moderate cut-off values are necessary to distinguish splicing variants. Moreover, the high correlation and consistent prediction ratio validated the feasibility of integration of the splicing prediction method in identifying splicing variants. We developed a splicing analytics platform called SPCards, which curates splicing variants from publications and predicts splicing scores of variants in genomes. SPCards also offers variant-level and gene-level annotation information, including allele frequency, non-synonymous prediction and comprehensive functional information. SPCards is suitable for high-throughput genetic identification of splicing variants, particularly those located in non-canonical splicing regions.

Modeling splicing outcome by combining 5'ss strength and splicing regulatory elements

Correct pre-mRNA processing in higher eukaryotes vastly depends on splice site recognition. Beyond conserved 5'ss and 3'ss motifs, splicing regulatory elements (SREs) play a pivotal role in this recognition process. Here, we present in silico designed sequences with arbitrary a priori prescribed splicing regulatory HEXplorer properties that can be concatenated to arbitrary length without changing their regulatory properties. We experimentally validated in silico predictions in a massively parallel splicing reporter assay on more than 3000 sequences and exemplarily identified some SRE binding proteins. Aiming at a unified 'functional splice site strength' encompassing both U1 snRNA complementarity and impact from neighboring SREs, we developed a novel RNA-seq based 5'ss usage landscape, mapping the competition of pairs of high confidence 5'ss and neighboring exonic GT sites along HBond and HEXplorer score coordinate axes on human fibroblast and endothelium transcriptome datasets. These RNA-seq data served as basis for a logistic 5'ss usage prediction model, which greatly improved discrimination between strong but unused exonic GT sites and annotated highly used 5'ss. Our 5'ss usage landscape offers a unified view on 5'ss and SRE neighborhood impact on splice site recognition, and may contribute to improved mutation assessment in human genetics.

Protocol to identify host-viral protein interactions between coagulation-related proteins and their genetic variants with SARS-CoV-2 proteins

Here, we describe a bioinformatics pipeline that evaluates the interactions between coagulation-related proteins and genetic variants with SARS-CoV-2 proteins. This pipeline searches for host proteins that may bind to viral protein and identifies and scores the protein genetic variants to predict the disease pathogenesis in specific subpopulations. Additionally, it is able to find structurally similar motifs and identify potential binding sites within the host-viral protein complexes to unveil viral impact on regulated biological processes and/or host-protein impact on viral invasion or reproduction. For complete details on the use and execution of this protocol, please refer to Holcomb et al. (2021).

Deep Molecular Characterization of Milder Spinal Muscular Atrophy Patients Carrying the c.859G>C Variant in SMN2

Spinal muscular atrophy (SMA) is a severe neuromuscular disorder caused by biallelic loss or pathogenic variants in the SMN1 gene. Copy number and modifier intragenic variants in SMN2, an almost identical paralog gene of SMN1, are known to influence the amount of complete SMN proteins. Therefore, SMN2 is considered the main phenotypic modifier of SMA, although genotype−phenotype correlation is not absolute. We present eleven unrelated SMA patients with milder phenotypes carrying the c.859G>C-positive modifier variant in SMN2. All were studied by a specific NGS method to allow a deep characterization of the entire SMN region. Analysis of two homozygous cases for the variant allowed us to identify a specific haplotype, Smn2-859C.1, in association with c.859G>C. Two other cases with the c.859G>C variant in their two SMN2 copies showed a second haplotype, Smn2-859C.2, in cis with Smn2-859C.1, assembling a more complex allele. We also identified a previously unreported variant in intron 2a exclusively linked to the Smn2-859C.1 haplotype (c.154-1141G>A), further suggesting that this region has been ancestrally conserved. The deep molecular characterization of SMN2 in our cohort highlights the importance of testing c.859G>C, as well as accurately assessing the SMN2 region in SMA patients to gain insight into the complex genotype−phenotype correlations and improve prognostic outcomes.

Selection of Olduvai Domains during Evolution: A Role for Primate-Specific Splicing Super-Enhancer and RNA Guanine Quadruplex in Bipartite NBPF Exons

Olduvai protein domains (also known as DUF1220 or NBPF) have undergone the greatest human-specific increase in the copy number of any coding region in the genome. Their repeat number was strongly associated with the evolutionary expansion of brain volumes, neuron counts and cognitive abilities, as well as with disorders of the autistic spectrum. Nevertheless, the domain function and cellular mechanisms underlying the positive selection of Olduvai DNA sequences in higher primates remain obscure. Here, I show that the inclusion of Olduvai exon doublets in mature transcripts is facilitated by a potent splicing enhancer that was created through duplication within the first exon. The enhancer is the strongest among the NBPF transcripts and further promotes the already high splicing activity of the unexpanded first exons of the two-exon domains, safeguarding the expanded Olduvai exon doublets in the mature transcriptome. The duplication also creates a predicted RNA guanine quadruplex that may regulate the access to spliceosomal components of the super-enhancer and influence the splicing of adjacent exons. Thus, positive Olduvai selection during primate evolution is likely to result from a combination of multiple targets in gene expression pathways, including RNA splicing.

Exonic splicing code and protein binding sites for calcium

Auxilliary splicing sequences in exons, known as enhancers (ESEs) and silencers (ESSs), have been subject to strong selection pressures at the RNA and protein level. The protein component of this splicing code is substantial, recently estimated at ∼50% of the total information within ESEs, but remains poorly understood. The ESE/ESS profiles were previously associated with the Irving-Williams (I-W) stability series for divalent metals, suggesting that the ESE/ESS evolution was shaped by metal binding sites. Here, we have examined splicing activities of exonic sequences that encode protein binding sites for Ca2+, a weak binder in the I-W affinity order. We found that predicted exon inclusion levels for the EF-hand motifs and for Ca2+-binding residues in nonEF-hand proteins were higher than for average exons. For canonical EF-hands, the increase was centred on the EF-hand chelation loop and, in particular, on Ca2+-coordinating residues, with a 1>12>3∼5>9 hierarchy in the 12-codon loop consensus and usage bias at codons 1 and 12. The same hierarchy but a lower increase was observed for noncanonical EF-hands, except for S100 proteins. EF-hand loops preferentially accumulated exon splits in two clusters, one located in their N-terminal halves and the other around codon 12. Using splicing assays and published crosslinking and immunoprecipitation data, we identify candidate trans-acting factors that preferentially bind conserved GA-rich motifs encoding negatively charged amino acids in the loops. Together, these data provide evidence for the high capacity of codons for Ca2+-coordinating residues to be retained in mature transcripts, facilitating their exon-level expansion during eukaryotic evolution.

Deep learning modeling m6A deposition reveals the importance of downstream cis-element sequences

The N⁶-methyladenosine (m⁶A) modification is deposited to nascent transcripts on chromatin, but its site-specificity mechanism is mostly unknown. Here we model the m⁶A deposition to pre-mRNA by iM6A (intelligent m⁶A), a deep learning method, demonstrating that the site-specific m⁶A methylation is primarily determined by the flanking nucleotide sequences. iM6A accurately models the m⁶A deposition (AUROC = 0.99) and uncovers surprisingly that the cis-elements regulating the m⁶A deposition preferentially reside within the 50 nt downstream of the m⁶A sites. The m⁶A enhancers mostly include part of the RRACH motif and the m⁶A silencers generally contain CG/GT/CT motifs. Our finding is supported by both independent experimental validations and evolutionary conservation. Moreover, our work provides evidences that mutations resulting in synonymous codons can affect the m⁶A deposition and the TGA stop codon favors m⁶A deposition nearby. Our iM6A deep learning modeling enables fast paced biological discovery which would be cost-prohibitive and unpractical with traditional experimental approaches, and uncovers a key cis-regulatory mechanism for m⁶A site-specific deposition.

Transgene-design: a web application for the design of mammalian transgenes

Summary

Transgene-design is a web application to help design transgenes for use in mammalian studies. It is predicated on the recent discovery that human intronless transgenes and native retrogenes can be expressed very effectively if the GC content at exonic synonymous sites is high. In addition, as exonic splice enhancers resident in intron containing genes may have different utility in intronless genes, these can be reduced or increased in density. Input can be a native gene or a commercially ‘optimised’ gene. The option to leave in the first intron and to protect or avoid other motifs is also permitted.

Availability and implementation

Transgene-design is based on a ruby for rails platform. The application is available at https://transgene-design.bath.ac.uk. The code is available under GNU General Public License from GitHub (https://github.com/smuehlh/transgenes).

Supplementary information

Supplementary data are available at Bioinformatics online.

Analysis of Pathogenic Pseudoexons Reveals Novel Mechanisms Driving Cryptic Splicing

Understanding pre-mRNA splicing is crucial to accurately diagnosing and treating genetic diseases. However, mutations that alter splicing can exert highly diverse effects. Of all the known types of splicing mutations, perhaps the rarest and most difficult to predict are those that activate pseudoexons, sometimes also called cryptic exons. Unlike other splicing mutations that either destroy or redirect existing splice events, pseudoexon mutations appear to create entirely new exons within introns. Since exon definition in vertebrates requires coordinated arrangements of numerous RNA motifs, one might expect that pseudoexons would only arise when rearrangements of intronic DNA create novel exons by chance. Surprisingly, although such mutations do occur, a far more common cause of pseudoexons is deep-intronic single nucleotide variants, raising the question of why these latent exon-like tracts near the mutation sites have not already been purged from the genome by the evolutionary advantage of more efficient splicing. Possible answers may lie in deep intronic splicing processes such as recursive splicing or poison exon splicing. Because these processes utilize intronic motifs that benignly engage with the spliceosome, the regions involved may be more susceptible to exonization than other intronic regions would be. We speculated that a comprehensive study of reported pseudoexons might detect alignments with known deep intronic splice sites and could also permit the characterisation of novel pseudoexon categories. In this report, we present and analyse a catalogue of over 400 published pseudoexon splice events. In addition to confirming prior observations of the most common pseudoexon mutation types, the size of this catalogue also enabled us to suggest new categories for some of the rarer types of pseudoexon mutation. By comparing our catalogue against published datasets of non-canonical splice events, we also found that 15.7% of pseudoexons exhibit some splicing activity at one or both of their splice sites in non-mutant cells. Importantly, this included seven examples of experimentally confirmed recursive splice sites, confirming for the first time a long-suspected link between these two splicing phenomena. These findings have the potential to improve the fidelity of genetic diagnostics and reveal new targets for splice-modulating therapies.

Systematic characterization of short intronic splicing-regulatory elements in SMN2 pre-mRNA

Intronic splicing enhancers and silencers (ISEs and ISSs) are two groups of splicing-regulatory elements (SREs) that play critical roles in determining splice-site selection, particularly for alternatively spliced introns or exons. SREs are often short motifs; their mutation or dysregulation of their cognate proteins frequently causes aberrant splicing and results in disease. To date, however, knowledge about SRE sequences and how they regulate splicing remains limited. Here, using an SMN2 minigene, we generated a complete pentamer-sequence library that comprises all possible combinations of 5 nucleotides in intron 7, at a fixed site downstream of the 5′ splice site. We systematically analyzed the effects of all 1023 mutant pentamers on exon 7 splicing, in comparison to the wild-type minigene, in HEK293 cells. Our data show that the majority of pentamers significantly affect exon 7 splicing: 584 of them are stimulatory and 230 are inhibitory. To identify actual SREs, we utilized a motif set enrichment analysis (MSEA), from which we identified groups of stimulatory and inhibitory SRE motifs. We experimentally validated several strong SREs in SMN1/2 and other minigene settings. Our results provide a valuable resource for understanding how short RNA sequences regulate splicing. Many novel SREs can be explored further to elucidate their mechanism of action.

Machine Learning for Designing Next-Generation mRNA Therapeutics

Over just the last 2 years, mRNA therapeutics and vaccines have undergone a rapid transition from an intriguing concept to real-world impact. However, whereas some aspects of mRNA therapeutics, such as the use of chemical modifications to increase stability and reduce immunogenicity, have been extensively optimized for over two decades, other aspects, particularly the selection and design of the noncoding leader and trailer sequences which control translation efficiency and stability, have received comparably less attention. In practice, such 5' and 3' untranslated regions (UTRs) are often borrowed from highly expressed human genes with few or no modifications, as in the case for the Pfizer/BioNTech Covid vaccine. Focusing on the 5'UTR, we here argue that model-driven design is a promising alternative that provides unprecedented control over 5'UTR function. We review recent work that combines synthetic biology with machine learning to build quantitative models that relate ribosome loading, and thus translation efficiency, to the 5'UTR sequence. We first introduce an experimental approach that uses polysome profiling and high-throughput sequencing to quantify ribosome loading for hundreds of thousands of 5'UTRs in parallel. We apply this approach to measure ribosome loading in synthetic RNA libraries with a random sequence inserted into the 5'UTR. We then review Optimus 5-Prime, a convolutional neural network model trained on the experimental data. We highlight that very accurate models of biological regulation can be learned from synthetic data sets with degenerate 5'UTRs. We validate model predictions not only on held-out data sets from our random library but also on a large library of over 30 000 human 5'UTR fragments and using translation reporter data collected independently by other groups. Both the experiment and model are compatible with commonly used chemically modified nucleosides, in particular, pseudouridine (Ψ) and 1-methyl-pseudouridine (m¹Ψ). We find that, in general, 5'UTRs have very similar impacts when combined with different protein-coding sequences and even in the context of different chemical modifications. We demonstrate that Optimus 5-Prime can be combined with design algorithms to generate de novo sequences with precisely defined translation efficiencies. We emphasize recent developments in design algorithms that rely on activation maximization and generative modeling to improve both the fitness and diversity of designed sequences. Compared with prior approaches such as genetic algorithms, we show that these approaches are not only faster but also less likely to get stuck in local sequence optima. Finally, we discuss how the approach reviewed here can be generalized to other gene regions and applications.

Massively parallel reporter assays discover de novo exonic splicing mutants in paralogs of Autism genes

To determine the contribution of defective splicing in Autism Spectrum Disorders (ASD), the most common neurodevelopmental disorder, a high throughput Massively Parallel Splicing Assay (MaPSY) was employed and identified 42 exonic splicing mutants out of 725 coding de novo variants discovered in the sequencing of ASD families. A redesign of the minigene constructs in MaPSY revealed that upstream exons with strong 5' splice sites increase the magnitude of skipping phenotypes observed in downstream exons. Select hits were validated by RT-PCR and amplicon sequencing in patient cell lines. Exonic splicing mutants were enriched in probands relative to unaffected siblings -especially synonymous variants (7.5% vs 3.5%, respectively). Of the 26 genes disrupted by exonic splicing mutations, 6 were in known ASD genes and 3 were in paralogs of known ASD genes. Of particular interest was a synonymous variant in TNRC6C - an ASD gene paralog with interactions with other ASD genes. Clinical records of 3 ASD patients with TNRC6C variant revealed respiratory issues consistent with phenotypes observed in TNRC6 depleted mice. Overall, this study highlights the need for splicing analysis in determining variant pathogenicity, especially as it relates to ASD.

Pseudoexon activation in disease by non-splice site deep intronic sequence variation - wild type pseudoexons constitute high-risk sites in the human genome

Accuracy of pre-messenger RNA (pre-mRNA) splicing is crucial for normal gene expression. Complex regulation supports the spliceosomal distinction between authentic exons and the many seemingly functional splice sites delimiting pseudoexons. Pseudoexons are nonfunctional intronic sequences that can be activated for aberrant inclusion in mRNA, which may cause disease. Pseudoexon activation is very challenging to predict, in particular when activation occurs by sequence variants that alter the splicing regulatory environment without directly affecting splice sites. As pseudoexon inclusion often evades detection due to activation of nonsense-mediated mRNA decay, and because conventional diagnostic procedures miss deep intronic sequence variation, pseudoexon activation is a heavily underreported disease mechanism. Pseudoexon characteristics have mainly been studied based on in silico predicted sequences. Moreover, because recognition of sequence variants that create or strengthen splice sites is possible by comparison with well-established consensus sequences, this type of pseudoexon activation is by far the most frequently reported. Here we review all known human disease-associated pseudoexons that carry functional splice sites and are activated by deep intronic sequence variants located outside splice site sequences. We delineate common characteristics that make this type of wild type pseudoexons distinct high-risk sites in the human genome.

Inherited retinal diseases: Linking genes, disease-causing variants, and relevant therapeutic modalities

Inherited retinal diseases (IRDs) are a clinically complex and heterogenous group of visual impairment phenotypes caused by pathogenic variants in at least 277 nuclear and mitochondrial genes, affecting different retinal regions, and depleting the vision of affected individuals. Genes that cause IRDs when mutated are unique by possessing differing genotype-phenotype correlations, varying inheritance patterns, hypomorphic alleles, and modifier genes thus complicating genetic interpretation. Next-generation sequencing has greatly advanced the identification of novel IRD-related genes and pathogenic variants in the last decade. For this review, we performed an in-depth literature search which allowed for compilation of the Global Retinal Inherited Disease (GRID) dataset containing 4,798 discrete variants and 17,299 alleles published in 31 papers, showing a wide range of frequencies and complexities among the 194 genes reported in GRID, with 65% of pathogenic variants being unique to a single individual. A better understanding of IRD-related gene distribution, gene complexity, and variant types allow for improved genetic testing and therapies. Current genetic therapeutic methods are also quite diverse and rely on variant identification, and range from whole gene replacement to single nucleotide editing at the DNA or RNA levels. IRDs and their suitable therapies thus require a range of effective disease modelling in human cells, granting insight into disease mechanisms and testing of possible treatments. This review summarizes genetic and therapeutic modalities of IRDs, provides new analyses of IRD-related genes (GRID and complexity scores), and provides information to match genetic-based therapies such as gene-specific and variant-specific therapies to the appropriate individuals.

The Impact of Pro-Inflammatory Cytokines on Alternative Splicing Patterns in Human Islets

Alternative splicing (AS) within the β cell has been proposed as one potential pathway that may exacerbate autoimmunity and unveil novel immunogenic epitopes in type 1 diabetes (T1D). We employed a computational strategy to prioritize pathogenic splicing events in human islets treated with IL-1β + IFN-γ as an ex vivo model of T1D and coupled this analysis with a k-mer based approach to predict RNA binding proteins involved in AS. In total, 969 AS events were identified in cytokine-treated islets, with the majority (44.8%) involving a skipped exon. ExonImpact identified 129 events predicted to impact protein structure. AS occurred with high frequency in MHC Class II-related mRNAs, and targeted qPCR validated reduced inclusion of Exon5 in the MHC Class II gene HLA-DMB. Single molecule RNA FISH confirmed increased HLA-DMB splicing in pancreatic sections from human donors with established T1D and autoantibody positivity. Serine and Arginine Rich Splicing Factor 2 was implicated in 37.2% of potentially pathogenic events, including Exon5 exclusion in HLA-DMB. Together, these data suggest that dynamic control of AS plays a role in the β cell response to inflammatory signals during T1D evolution.

Linking genome variants to disease: scalable approaches to test the functional impact of human mutations

The application of genomics to medicine has accelerated the discovery of mutations underlying disease and has enhanced our knowledge of the molecular underpinnings of diverse pathologies. As the amount of human genetic material queried via sequencing has grown exponentially in recent years, so too has the number of rare variants observed. Despite progress, our ability to distinguish which rare variants have clinical significance remains limited. Over the last decade, however, powerful experimental approaches have emerged to characterize variant effects orders of magnitude faster than before. Fueled by improved DNA synthesis and sequencing and, more recently, by CRISPR/Cas9 genome editing, multiplex functional assays provide a means of generating variant effect data in wide-ranging experimental systems. Here, I review recent applications of multiplex assays that link human variants to disease phenotypes and I describe emerging strategies that will enhance their clinical utility in coming years.

Evidence in disease and non-disease contexts that nonsense mutations cause altered splicing via motif disruption

Transcripts containing premature termination codons (PTCs) can be subject to nonsense-associated alternative splicing (NAS). Two models have been evoked to explain this, scanning and splice motif disruption. The latter postulates that exonic cis motifs, such as exonic splice enhancers (ESEs), are disrupted by nonsense mutations. We employ genome-wide transcriptomic and k-mer enrichment methods to scrutinize this model. First, we show that ESEs are prone to disruptive nonsense mutations owing to their purine richness and paucity of TGA, TAA and TAG. The motif model correctly predicts that NAS rates should be low (we estimate 5–30%) and approximately in line with estimates for the rate at which random point mutations disrupt splicing (8–20%). Further, we find that, as expected, NAS-associated PTCs are predictable from nucleotide-based machine learning approaches to predict splice disruption and, at least for pathogenic variants, are enriched in ESEs. Finally, we find that both in and out of frame mutations to TAA, TGA or TAG are associated with exon skipping. While a higher relative frequency of such skip-inducing mutations in-frame than out of frame lends some credence to the scanning model, these results reinforce the importance of considering splice motif modulation to understand the etiology of PTC-associated disease.

What's Wrong in a Jump? Prediction and Validation of Splice Site Variants

Alternative splicing (AS) is a crucial process to enhance gene expression driving organism development. Interestingly, more than 95% of human genes undergo AS, producing multiple protein isoforms from the same transcript. Any alteration (e.g., nucleotide substitutions, insertions, and deletions) involving consensus splicing regulatory sequences in a specific gene may result in the production of aberrant and not properly working proteins. In this review, we introduce the key steps of splicing mechanism and describe all different types of genomic variants affecting this process (splicing variants in acceptor/donor sites or branch point or polypyrimidine tract, exonic, and deep intronic changes). Then, we provide an updated approach to improve splice variants detection. First, we review the main computational tools, including the recent Machine Learning-based algorithms, for the prediction of splice site variants, in order to characterize how a genomic variant interferes with splicing process. Next, we report the experimental methods to validate the predictive analyses are defined, distinguishing between methods testing RNA (transcriptomics analysis) or proteins (proteomics experiments). For both prediction and validation steps, benefits and weaknesses of each tool/procedure are accurately reported, as well as suggestions on which approaches are more suitable in diagnostic rather than in clinical research.

Interpretable prioritization of splice variants in diagnostic next-generation sequencing

A critical challenge in genetic diagnostics is the computational assessment of candidate splice variants, specifically the interpretation of nucleotide changes located outside of the highly conserved dinucleotide sequences at the 5' and 3' ends of introns. To address this gap, we developed the Super Quick Information-content Random-forest Learning of Splice variants (SQUIRLS) algorithm. SQUIRLS generates a small set of interpretable features for machine learning by calculating the information-content of wild-type and variant sequences of canonical and cryptic splice sites, assessing changes in candidate splicing regulatory sequences, and incorporating characteristics of the sequence such as exon length, disruptions of the AG exclusion zone, and conservation. We curated a comprehensive collection of disease-associated splice-altering variants at positions outside of the highly conserved AG/GT dinucleotides at the termini of introns. SQUIRLS trains two random-forest classifiers for the donor and for the acceptor and combines their outputs by logistic regression to yield a final score. We show that SQUIRLS transcends previous state-of-the-art accuracy in classifying splice variants as assessed by rank analysis in simulated exomes, and is significantly faster than competing methods. SQUIRLS provides tabular output files for incorporation into diagnostic pipelines for exome and genome analysis, as well as visualizations that contextualize predicted effects of variants on splicing to make it easier to interpret splice variants in diagnostic settings.

High-throughput splicing assays identify missense and silent splice-disruptive POU1F1 variants underlying pituitary hormone deficiency

Pituitary hormone deficiency occurs in ∼1:4,000 live births. Approximately 3% of the cases are due to mutations in the alpha isoform of POU1F1, a pituitary-specific transcriptional activator. We found four separate heterozygous missense variants in unrelated individuals with hypopituitarism that were predicted to affect a minor isoform, POU1F1 beta, which can act as a transcriptional repressor. These variants retain repressor activity, but they shift splicing to favor the expression of the beta isoform, resulting in dominant-negative loss of function. Using a high-throughput splicing reporter assay, we tested 1,070 single-nucleotide variants in POU1F1. We identified 96 splice-disruptive variants, including 14 synonymous variants. In separate cohorts, we found two additional synonymous variants nominated by this screen that co-segregate with hypopituitarism. This study underlines the importance of evaluating the impact of variants on splicing and provides a catalog for interpretation of variants of unknown significance in POU1F1.

Intermixing the OPN1LW and OPN1MW Genes Disrupts the Exonic Splicing Code Causing an Array of Vision Disorders

Light absorption by photopigment molecules expressed in the photoreceptors in the retina is the first step in seeing. Two types of photoreceptors in the human retina are responsible for image formation: rods, and cones. Except at very low light levels when rods are active, all vision is based on cones. Cones mediate high acuity vision and color vision. Furthermore, they are critically important in the visual feedback mechanism that regulates refractive development of the eye during childhood. The human retina contains a mosaic of three cone types, short-wavelength (S), long-wavelength (L), and middle-wavelength (M) sensitive; however, the vast majority (~94%) are L and M cones. The OPN1LW and OPN1MW genes, located on the X-chromosome at Xq28, encode the protein component of the light-sensitive photopigments expressed in the L and M cones. Diverse haplotypes of exon 3 of the OPN1LW and OPN1MW genes arose thru unequal recombination mechanisms that have intermixed the genes. A subset of the haplotypes causes exon 3- skipping during pre-messenger RNA splicing and are associated with vision disorders. Here, we review the mechanism by which splicing defects in these genes cause vision disorders.

Let It Go: HIV-1 cis-Acting Repressive Sequences

After human immunodeficiency virus type 1 (HIV-1) was identified in the early 1980s, intensive work began to understand the molecular basis of HIV-1 gene expression. Subgenomic HIV-1 RNA regions, spread throughout the viral genome, were described to have a negative impact on the nuclear export of some viral transcripts. Those studies revealed an intrinsic RNA code as a new form of nuclear export regulation. Since such regulatory regions were later also identified in other viruses, as well as in cellular genes, it can be assumed that, during evolution, viruses took advantage of them to achieve more sophisticated replication mechanisms. Here, we review HIV-1 cis-acting repressive sequences that have been identified, and we discuss their possible underlying mechanisms and importance. Additionally, we show how current bioinformatic tools might allow more predictive approaches to identify and investigate them.

An artificial neural network approach integrating plasma proteomics and genetic data identifies PLXNA4 as a new susceptibility locus for pulmonary embolism

Venous thromboembolism is the third common cardiovascular disease and is composed of two entities, deep vein thrombosis (DVT) and its potential fatal form, pulmonary embolism (PE). While PE is observed in ~ 40% of patients with documented DVT, there is limited biomarkers that can help identifying patients at high PE risk. To fill this need, we implemented a two hidden-layers artificial neural networks (ANN) on 376 antibodies and 19 biological traits measured in the plasma of 1388 DVT patients, with or without PE, of the MARTHA study. We used the LIME algorithm to obtain a linear approximate of the resulting ANN prediction model. As MARTHA patients were typed for genotyping DNA arrays, a genome wide association study (GWAS) was conducted on the LIME estimate. Detected single nucleotide polymorphisms (SNPs) were tested for association with PE risk in MARTHA. Main findings were replicated in the EOVT study composed of 143 PE patients and 196 DVT only patients. The derived ANN model for PE achieved an accuracy of 0.89 and 0.79 in our training and testing sets, respectively. A GWAS on the LIME approximate identified a strong statistical association peak (rs1424597: p = 5.3 × 10^-7) at the PLXNA4 locus. Homozygote carriers for the rs1424597-A allele were then more frequently observed in PE than in DVT patients from the MARTHA (2% vs. 0.4%, p = 0.005) and the EOVT (3% vs. 0%, p = 0.013) studies. In a sample of 112 COVID-19 patients known to have endotheliopathy leading to acute lung injury and an increased risk of PE, decreased PLXNA4 levels were associated (p = 0.025) with worsened respiratory function. Using an original integrated proteomics and genetics strategy, we identified PLXNA4 as a new susceptibility gene for PE whose exact role now needs to be further elucidated.

Role of Splicing Regulatory Elements and In Silico Tools Usage in the Identification of Deep Intronic Splicing Variants in Hereditary Breast/Ovarian Cancer Genes

Simple Summary

There is a significant percentage of hereditary breast and ovarian cancer (HBOC) cases that remain undiagnosed, because no pathogenic variant is detected through massively parallel sequencing of coding exons and exon-intron boundaries of high-moderate susceptibility risk genes. Deep intronic regions may contain variants affecting RNA splicing, leading ultimately to disease, and hence they may explain several cases where the genetic cause of HBOC is unknown. This study aims to characterize intronic regions to identify a landscape of “exonizable” zones and test the efficiency of two in silico tools to detect deep intronic variants affecting the mRNA splicing process.

Abstract

The contribution of deep intronic splice-altering variants to hereditary breast and ovarian cancer (HBOC) is unknown. Current computational in silico tools to predict spliceogenic variants leading to pseudoexons have limited efficiency. We assessed the performance of the SpliceAI tool combined with ESRseq scores to identify spliceogenic deep intronic variants by affecting cryptic sites or splicing regulatory elements (SREs) using literature and experimental datasets. Our results with 233 published deep intronic variants showed that SpliceAI, with a 0.05 threshold, predicts spliceogenic deep intronic variants affecting cryptic splice sites, but is less effective in detecting those affecting SREs. Next, we characterized the SRE profiles using ESRseq, showing that pseudoexons are significantly enriched in SRE-enhancers compared to adjacent intronic regions. Although the combination of SpliceAI with ESRseq scores (considering ∆ESRseq and SRE landscape) showed higher sensitivity, the global performance did not improve because of the higher number of false positives. The combination of both tools was tested in a tumor RNA dataset with 207 intronic variants disrupting splicing, showing a sensitivity of 86%. Following the pipeline, five spliceogenic deep intronic variants were experimentally identified from 33 variants in HBOC genes. Overall, our results provide a framework to detect deep intronic variants disrupting splicing.

Missense RHD single nucleotide variants induce weakened D antigen expression by altering splicing and/or protein expression

BACKGROUND

Although D variant phenotype is known to be due to genetic defects, including rare missense single nucleotide variants (SNVs), within the RHD gene, few studies have addressed the molecular and cellular mechanisms driving this altered expression. We and others showed previously that splicing is commonly disrupted by SNVs in constitutive splice sites and their vicinity. We thus sought to investigate whether rare missense SNVs located in "deep" exonic regions could also impair this mechanism.

STUDY DESIGN AND METHODS

Forty-six missense SNVs reported within exons 6 and 7 were first selected from the Human RhesusBase. Their respective effect on splicing was assessed by using an in vitro assay. An RhD-negative cell model was further generated by using the CRISPR-Cas9 approach. RhD-mutated proteins were overexpressed in the newly created model, and cell membrane expression of the D antigen was measured by flow cytometry.

RESULTS

Minigene splicing assay showed that 14 of 46 (30.4%) missense SNVs alter splicing. Very interestingly, further investigation of two missense SNVs, which both affect codon 338 and confer a weak D phenotype, showed various mechanisms: c.1012C>G (p.Leu338Val) disrupts splicing only, while c.1013T>C (p.Leu338Pro) alters only the protein structure, in agreement with in silico prediction tools and 3D protein structure visualization.

CONCLUSION

Our functional data set suggests that missense SNVs damage quantitatively D antigen expression by, at least, two different mechanisms (splicing alteration and protein destabilization) that may act independently. These data thereby contribute to extend the current knowledge of the molecular mechanisms governing weakened D expression.