Abstract: Accurate and systematic transcriptome-wide detection of 5-methylcytosine (m5C) has proved challenging, and there are conflicting views about the prevalence of this modification in mRNAs. Here we report an experimental and computational framework that robustly identified mRNA m5C sites and determined sequence motifs and structural features associated with the modification using a set of high-confidence sites. We developed a quantitative atlas of RNA m5C sites in human and mouse tissues based on our framework. In a given tissue, we typically identified several hundred exonic m5C sites. About 62–70% of the sites had low methylation levels ( 40% methylation). Cross-species analysis revealed that species, rather than tissue type, was the primary determinant of methylation levels, indicating strong cis-directed regulation of RNA methylation. Combined, these data provide a valuable resource for identifying the regulation and functions of RNA methylation. A quantitative atlas of RNA m5C sites in human and mouse tissues based on a new discovery pipeline allows the identification of sequence motifs and structural features associated with the modification and provides a resource for future studies.

Abstract: Predicting the effects of genetic variants on splicing is highly relevant for human genetics. We describe the framework MMSplice (modular modeling of splicing) with which we built the winning model of the CAGI5 exon skipping prediction challenge. The MMSplice modules are neural networks scoring exon, intron, and splice sites, trained on distinct large-scale genomics datasets. These modules are combined to predict effects of variants on exon skipping, splice site choice, splicing efficiency, and pathogenicity, with matched or higher performance than state-of-the-art. Our models, available in the repository Kipoi, apply to variants including indels directly from VCF files.

Abstract: Mutations in the dystrophin gene cause Duchenne muscular dystrophy (DMD), which is characterized by lethal degeneration of cardiac and skeletal muscles. Mutations that delete exon 44 of the dystrophin gene represent one of the most common causes of DMD and can be corrected in ~12% of patients by editing surrounding exons, which restores the dystrophin open reading frame. Here, we present a simple and efficient strategy for correction of exon 44 deletion mutations by CRISPR-Cas9 gene editing in cardiomyocytes obtained from patient-derived induced pluripotent stem cells and in a new mouse model harboring the same deletion mutation. Using AAV9 encoding Cas9 and single guide RNAs, we also demonstrate the importance of the dosages of these gene editing components for optimal gene correction in vivo. Our findings represent a significant step toward possible clinical application of gene editing for correction of DMD.

Abstract: Spliceosome mutations are common in myelodysplastic syndromes (MDS) and acute myeloid leukaemia (AML), but the oncogenic changes due to these mutations have not been identified. Here a global analysis of exon usage in AML samples revealed distinct molecular subsets containing alternative spliced isoforms of inflammatory and immune genes. Interleukin-1 receptor-associated kinase 4 (IRAK4) was the dominant alternatively spliced isoform in MDS and AML and is characterized by a longer isoform that retains exon 4, which encodes IRAK4-long (IRAK4-L), a protein that assembles with the myddosome, results in maximal activation of nuclear factor kappa-light-chain-enhancer of B cells (NF-κB) and is essential for leukaemic cell function. Expression of IRAK4-L is mediated by mutant U2 small nuclear RNA auxiliary factor 1 (U2AF1) and is associated with oncogenic signalling in MDS and AML. Inhibition of IRAK4-L abrogates leukaemic growth, particularly in AML cells with higher expression of the IRAK4-L isoform. Collectively, mutations in U2AF1 induce expression of therapeutically targetable 'active' IRAK4 isoforms and provide a genetic link to activation of chronic innate immune signalling in MDS and AML.

Abstract: Synonymous mutations have been viewed as silent mutations, since they only affect the DNA and mRNA, but not the amino acid sequence of the resulting protein. Nonetheless, recent studies suggest their significant impact on splicing, RNA stability, RNA folding, translation or co-translational protein folding. Hence, we compile 659194 synonymous mutations found in human cancer and characterize their properties. We provide the user-friendly, comprehensive resource for synonymous mutations in cancer, SynMICdb ( http://SynMICdb.dkfz.de ), which also contains orthogonal information about gene annotation, recurrence, mutation loads, cancer association, conservation, alternative events, impact on mRNA structure and a SynMICdb score. Notably, synonymous and missense mutations are depleted at the 5'-end of the coding sequence as well as at the ends of internal exons independent of mutational signatures. For patient-derived synonymous mutations in the oncogene KRAS, we indicate that single point mutations can have a relevant impact on expression as well as on mRNA secondary structure. Synonymous mutations do not alter amino acid sequence but may exert oncogenic effects in other ways. Here, the authors present a catalogue of synonymous mutations in cancer and characterise their properties.

Abstract: Gene knock outs (KOs) are efficiently engineered through CRISPR-Cas9-induced frameshift mutations. While the efficiency of DNA editing is readily verified by DNA sequencing, a systematic understanding of the efficiency of protein elimination has been lacking. Here we devised an experimental strategy combining RNA sequencing and triple-stage mass spectrometry to characterize 193 genetically verified deletions targeting 136 distinct genes generated by CRISPR-induced frameshifts in HAP1 cells. We observed residual protein expression for about one third of the quantified targets, at variable levels from low to original, and identified two causal mechanisms, translation reinitiation leading to N-terminally truncated target proteins or skipping of the edited exon leading to protein isoforms with internal sequence deletions. Detailed analysis of three truncated targets, BRD4, DNMT1 and NGLY1, revealed partial preservation of protein function. Our results imply that systematic characterization of residual protein expression or function in CRISPR-Cas9-generated KO lines is necessary for phenotype interpretation.

Abstract: The introduction of insertion-deletions (INDELs) by non-homologous end-joining (NHEJ) pathway underlies the mechanistic basis of CRISPR-Cas9-directed genome editing. Selective gene ablation using CRISPR-Cas9 is achieved by installation of a premature termination codon (PTC) from a frameshift-inducing INDEL that elicits nonsense-mediated decay (NMD) of the mutant mRNA. Here, by examining the mRNA and protein products of CRISPR targeted genes in a cell line panel with presumed gene knockouts, we detect the production of foreign mRNAs or proteins in ~50% of the cell lines. We demonstrate that these aberrant protein products stem from the introduction of INDELs that promote internal ribosomal entry, convert pseudo-mRNAs (alternatively spliced mRNAs with a PTC) into protein encoding molecules, or induce exon skipping by disruption of exon splicing enhancers (ESEs). Our results reveal challenges to manipulating gene expression outcomes using INDEL-based mutagenesis and strategies useful in mitigating their impact on intended genome-editing outcomes.

Abstract: Interest in extrachromosomal circular DNA (eccDNA) molecules has increased recently because of their widespread presence in normal cells across every species ranging from yeast to humans, their increased levels in cancer cells and their overlap with oncogenic and drug-resistant genes. However, the majority of eccDNA (microDNA) in mammalian tissues and cell lines are too small to carry protein coding genes. We have tested functional capabilities of microDNA by creating artificial microDNA molecules mimicking known microDNA sequences and have discovered that they express functional small regulatory RNA including microRNA and novel si-like RNA. MicroDNA are transcribed in vitro and in vivo independent of a canonical promoter sequence. MicroDNA that carry miRNA genes form transcripts that are processed by the endogenous RNA-interference pathway into mature miRNA molecules, which repress a luciferase reporter gene as well as endogenous mRNA targets of the miRNA. Further, microDNA that contain sequences of exons repress the endogenous gene from which the microDNA were derived through the formation of novel si-like RNA. We also show that endogenous microDNA associate with RNA polymerases subunits, POLR2H and POLR3F. Together, these results suggest that microDNA may modulate gene expression through the production of both known and novel regulatory small RNA.

Abstract: The molecular mechanisms of exon definition and back-splicing are fundamental unanswered questions in pre-mRNA splicing. Here we report cryo-electron microscopy structures of the yeast spliceosomal E complex assembled on introns, providing a view of the earliest event in the splicing cycle that commits pre-mRNAs to splicing. The E complex architecture suggests that the same spliceosome can assemble across an exon, and that it either remodels to span an intron for canonical linear splicing (typically on short exons) or catalyses back-splicing to generate circular RNA (on long exons). The model is supported by our experiments, which show that an E complex assembled on the middle exon of yeast EFM5 or HMRA1 can be chased into circular RNA when the exon is sufficiently long. This simple model unifies intron definition, exon definition, and back-splicing through the same spliceosome in all eukaryotes and should inspire experiments in many other systems to understand the mechanism and regulation of these processes.

Abstract: A well-known limit of genome browsers is that the large amount of genome and gene data is not organized in the form of a searchable database, hampering full management of numerical data and free calculations. Due to the continuous increase of data deposited in genomic repositories, their content revision and analysis is recommended. Using GeneBase, a software with a graphical interface able to import and elaborate National Center for Biotechnology Information (NCBI) Gene database entries, we provide tabulated spreadsheets updated to 2019 about human nuclear protein-coding gene data set ready to be used for any type of analysis about genes, transcripts and gene organization. Comparison with previous reports reveals substantial change in the number of known nuclear protein-coding genes (now 19,116), the protein-coding non-redundant transcriptome space [now 59,281,518 base pair (bp), 10.1% increase], the number of exons (now 562,164, 36.2% increase) due to a relevant increase of the RNA isoforms recorded. Other parameters such as gene, exon or intron mean and extreme length appear to have reached a stability that is unlikely to be substantially modified by human genome data updates, at least regarding protein-coding genes. Finally, we confirm that there are no human introns shorter than 30 bp.

Abstract: Precursor messenger RNA (pre-mRNA) splicing is executed by the spliceosome. In the past 3 years, cryoelectron microscopy (cryo-EM) structures have been elucidated for a majority of the yeast spliceosomal complexes and for a few human spliceosomes. During the splicing reaction, the dynamic spliceosome has an immobile core of about 20 protein and RNA components, which are organized around a conserved splicing active site. The divalent metal ions, coordinated by U6 small nuclear RNA (snRNA), catalyze the branching reaction and exon ligation. The spliceosome also contains a mobile but compositionally stable group of about 13 proteins and a portion of U2 snRNA, which facilitate substrate delivery into the splicing active site. The spliceosomal transitions are driven by the RNA-dependent ATPase/helicases, resulting in the recruitment and dissociation of specific splicing factors that enable the reaction. In summary, the spliceosome is a protein-directed metalloribozyme.

Abstract: Among RNA viruses, the order Nidovirales stands out for including viruses with the largest RNA genomes currently known. Nidoviruses employ a complex RNA-synthesizing machinery comprising a variety of non-structural proteins (nsps). One of the postulated drivers of the expansion of nidovirus genomes is the presence of a proofreading 3'-to-5' exoribonuclease (ExoN) belonging to the DEDDh family. ExoN may enhance the fidelity of RNA synthesis by correcting nucleotide incorporation errors made by the RNA-dependent RNA polymerase. Here, we review our current understanding of ExoN evolution, structure, and function. Most experimental data are derived from studies of the ExoN domain of coronaviruses (CoVs), which were triggered by the bioinformatics-based identification of ExoN in the genome of severe acute respiratory syndrome coronavirus (SARS-CoV) and its relatives in 2003. Although convincing data supporting the proofreading hypothesis have been obtained, from biochemical assays and studies with CoV mutants lacking ExoN functionality, the features of ExoN from most other nidovirus families remain to be characterized. Remarkably, viable ExoN knockout mutants were obtained only for two CoVs, mouse hepatitis virus (MHV) and SARS-CoV, whose RNA synthesis and replication kinetics were mildly affected by the lack of ExoN function. In several other CoV species, ExoN inactivation was not tolerated, and knockout mutants could not be rescued when launched using a reverse genetics system. This suggests that ExoN is also critical for primary viral RNA synthesis, a property that poorly matches the profile of an enzyme that would merely boost long-term replication fidelity. In CoVs, ExoN resides in a bifunctional replicase subunit (nsp14) whose C-terminal part has (N7-guanine)-methyltransferase activity. The crystal structure of SARS-CoV nsp14 has shed light on the interplay between these two domains, and on nsp14's interactions with nsp10, a co-factor that strongly enhances ExoN activity in vitro assays. Further elucidation of the structure-function relationships of ExoN and its interactions with other (viral and/or host) members of the CoV replication machinery will be key to understanding the enzyme's role in viral RNA synthesis and pathogenesis, and may contribute to the design of new approaches to combat emerging nidoviruses.

Abstract: Huntington's disease (HD) is a currently incurable and, ultimately, fatal neurodegenerative disorder caused by a CAG trinucleotide repeat expansion within exon 1 of the huntingtin (HTT) gene, which results in the production of a mutant protein that forms inclusions and selectively destroys neurons in the striatum and other adjacent structures. The RNA-guided Cas9 endonuclease from CRISPR-Cas9 systems is a versatile technology for inducing DNA double-strand breaks that can stimulate the introduction of frameshift-inducing mutations and permanently disable mutant gene function. Here, we show that the Cas9 nuclease from Staphylococcus aureus, a small Cas9 ortholog that can be packaged alongside a single guide RNA into a single adeno-associated virus (AAV) vector, can be used to disrupt the expression of the mutant HTT gene in the R6/2 mouse model of HD following its in vivo delivery to the striatum. Specifically, we found that CRISPR-Cas9-mediated disruption of the mutant HTT gene resulted in a ∼50% decrease in neuronal inclusions and significantly improved lifespan and certain motor deficits. These results thus illustrate the potential for CRISPR-Cas9 technology to treat HD and other autosomal dominant neurodegenerative disorders caused by a trinucleotide repeat expansion via in vivo genome editing.

Abstract: Splicing modifiers promoting SMN2 exon 7 inclusion have the potential to treat spinal muscular atrophy, the leading genetic cause of infantile death. These small molecules are SMN2 exon 7 selective and act during the early stages of spliceosome assembly. Here, we show at atomic resolution how the drug selectively promotes the recognition of the weak 5' splice site of SMN2 exon 7 by U1 snRNP. The solution structure of the RNA duplex formed following 5' splice site recognition in the presence of the splicing modifier revealed that the drug specifically stabilizes a bulged adenine at this exon-intron junction and converts the weak 5' splice site of SMN2 exon 7 into a stronger one. The small molecule acts as a specific splicing enhancer cooperatively with the splicing regulatory network. Our investigations uncovered a novel concept for gene-specific alternative splicing correction that we coined 5' splice site bulge repair.

Abstract: Mutations in the β-catenin gene (CTNNB1) have been implicated in the pathogenesis of some cancers. The recent development of cancer genome databases has facilitated comprehensive and focused analyses on the mutation status of cancer-related genes. We have used these databases to analyze the CTNNB1 mutations assembled from different tumor types. High incidences of CTNNB1 mutations were detected in endometrial, liver, and colorectal cancers. This finding agrees with the oncogenic role of aberrantly activated β-catenin in epithelial cells. Elevated frequencies of missense mutations were found in the exon 3 of CTNNB1, which is responsible for encoding the regulatory amino acids at the N-terminal region of the protein. In the case of metastatic colorectal cancers, inframe deletions were revealed in the region spanning exon 3. Thus, exon 3 of CTNNB1 can be considered to be a mutation hotspot in these cancers. Since the N-terminal region of the β-catenin protein forms a flexible structure, many questions arise regarding the structural and functional impacts of hotspot mutations. Clinical identification of hotspot mutations could provide the mechanistic basis for an oncogenic role of mutant β-catenin proteins in cancer cells. Furthermore, a systematic understanding of tumor-driving hotspot mutations could open new avenues for precision oncology.

Abstract: Myotonic dystrophy (dystrophia myotonica, DM) is a multi-systemic disease caused by expanded CTG or CCTG microsatellite repeats. Characterized by symptoms in muscle, heart and central nervous system, among others, it is one of the most variable diseases known. A major pathogenic event in DM is the sequestration of muscleblind-like proteins by CUG or CCUG repeat-containing RNAs transcribed from expanded repeats, and differences in the extent of MBNL sequestration dependent on repeat length and expression level may account for some portion of the variability. However, many other cellular pathways are reported to be perturbed in DM, and the severity of specific disease symptoms varies among individuals. To help understand this variability and facilitate research into DM, we generated 120 RNASeq transcriptomes from skeletal and heart muscle derived from healthy and DM1 biopsies and autopsies. A limited number of DM2 and Duchenne muscular dystrophy samples were also sequenced. We analyzed splicing and gene expression, identified tissue-specific changes in RNA processing and uncovered transcriptome changes strongly correlating with muscle strength. We created a web resource at http://DMseq.org that hosts raw and processed transcriptome data and provides a lightweight, responsive interface that enables browsing of processed data across the genome.

Abstract: Pre-mRNA-splicing and adenosine to inosine (A-to-I) RNA-editing occur mostly cotranscriptionally. During A-to-I editing, a genomically encoded adenosine is deaminated to inosine by adenosine deaminases acting on RNA (ADARs). Editing-competent stems are frequently formed between exons and introns. Consistently, studies using reporter assays have shown that splicing efficiency can affect editing levels. Here, we use Nascent-seq and identify ∼90,000 novel A-to-I editing events in the mouse brain transcriptome. Most novel sites are located in intronic regions. Unlike previously assumed, we show that both ADAR (ADAR1) and ADARB1 (ADAR2) can edit repeat elements and regular transcripts to the same extent. We find that inhibition of splicing primarily increases editing levels at hundreds of sites, suggesting that reduced splicing efficiency extends the exposure of intronic and exonic sequences to ADAR enzymes. Lack of splicing factors NOVA1 or NOVA2 changes global editing levels, demonstrating that alternative splicing factors can modulate RNA editing. Finally, we show that intron retention rates correlate with editing levels across different brain tissues. We therefore demonstrate that splicing efficiency is a major factor controlling tissue-specific differences in editing levels.

Abstract: During exon ligation, the Saccharomyces cerevisiae spliceosome recognizes the 3′-splice site (3′SS) of precursor messenger RNA (pre-mRNA) through non–Watson-Crick pairing with the 5′SS and the branch adenosine, in a conformation stabilized by Prp18 and Prp8. Here we present the 3.3-angstrom cryo–electron microscopy structure of a human postcatalytic spliceosome just after exon ligation. The 3′SS docks at the active site through conserved RNA interactions in the absence of Prp18. Unexpectedly, the metazoan-specific FAM32A directly bridges the 5′-exon and intron 3′SS of pre-mRNA and promotes exon ligation, as shown by functional assays. CACTIN, SDE2, and NKAP—factors implicated in alternative splicing—further stabilize the catalytic conformation of the spliceosome during exon ligation. Together these four proteins act as exon ligation factors. Our study reveals how the human spliceosome has co-opted additional proteins to modulate a conserved RNA-based mechanism for 3′SS selection and to potentially fine-tune alternative splicing at the exon ligation stage.

Abstract: Apoptosis plays a vital role in cell homeostasis during development and disease. Bcl-x, a member of the Bcl-2 family of proteins, is a mitochondrial transmembrane protein that functions to regulate the intrinsic apoptosis pathway. An alternative splicing (AS) event in exon 2 of Bcl-x results in two isoforms of Bcl-x with antagonistic effects on cell survival: Bcl-xL (long isoform), which is anti-apoptotic, and Bcl-xS (short isoform), which is pro-apoptotic. Bcl-xL is the most abundant Bcl-x protein and functions to inhibit apoptosis by a number of different mechanisms including inhibition of Bax. In contrast, Bcl-xS can directly bind to and inhibit the anti-apoptotic Bcl-xL and Bcl-2 proteins, resulting in the release of the pro-apoptotic Bak. There are multiple splice factors and signaling pathways that influence the Bcl-xL/Bcl-xS splicing ratio, including serine/arginine-rich (SR) proteins, heterogeneous nuclear ribonucleoproteins (hnRNPs), transcription factors, and cytokines. Dysregulation of the AS of Bcl-x has been implicated in cancer and diabetes. In cancer, the upregulation of Bcl-xL expression in tumor cells can result in resistance to chemotherapeutic agents. On the other hand, dysregulation of Bcl-x AS to promote Bcl-xS expression has been shown to be detrimental to pancreatic β-cells in diabetes, resulting in β-cell apoptosis. Therefore, manipulation of the splice factor, transcription factor, and signaling pathways that modulate this splicing event is fast emerging as a therapeutic avenue in the treatment of cancer and diabetes.

Abstract: Circular RNAs (circRNAs) are a new class of RNAs, and many studies have identified thousands of circRNAs in tumor cells. Fibronectin type III domain-containing protein 3B (FNDC3B) circular RNA (circFNDC3B, circBase ID: hsa_circ_0006156) circularizes with exons 5 and 6. Gibson Assembly DNA technology was used to construct a circFNDC3B expression vector without a splice site and restriction enzyme site. We showed that circFNDC3B increased migration and invasion in gastric cancer (GC). Ectopic expression of circFNDC3B reduced the level of E-cadherin protein to promote the epithelial-mesenchymal transition in GC. RNA immunoprecipitation assays and RNA pull-down assays confirmed that circFNC3B increased CD44 expression, which was associated with cell adhesion, via the formation of a ternary complex of circFNDC3B-IGF2BP3-CD44 mRNA. These results indicated that circFNDC3B was associated with the degree of malignancy to highlight the specific characteristics of cell invasion.

Abstract: Low frequency coding variants in TREM2 are associated with Alzheimer disease (AD) risk and cerebrospinal fluid (CSF) TREM2 protein levels are different between AD cases and controls. Similarly, TREM2 risk variant carriers also exhibit differential CSF TREM2 levels. TREM2 has three different alternative transcripts, but most of the functional studies only model the longest transcript. No studies have analyzed TREM2 expression levels or alternative splicing in brains from AD and cognitively normal individuals. We wanted to determine whether there was differential expression of TREM2 in sporadic-AD cases versus AD-TREM2 carriers vs sex- and aged-matched normal controls; and if this differential expression was due to a particular TREM2 transcript. We analyzed RNA-Seq data from parietal lobe brain tissue from AD cases with TREM2 variants (n = 33), AD cases (n = 195) and healthy controls (n = 118), from three independent datasets using Kallisto and the R package tximport to determine the read count for each transcript and quantified transcript abundance as transcripts per million. The three TREM2 transcripts were expressed in brain cortex in the three datasets. We demonstrate for the first time that the transcript that lacks the transmembrane domain and encodes a soluble form of TREM2 (sTREM2) has an expression level around 60% of the canonical transcript, suggesting that around 25% of the sTREM2 protein levels could be explained by this transcript. We did not observe a difference in the overall TREM2 expression level between cases and controls. However, the isoform which lacks the 5′ exon, but includes the transmembrane domain, was significantly lower in TREM2- p.R62H carriers than in AD cases (p = 0.007). Using bulk RNA-Seq data from three different cohorts, we were able to quantify the expression level of the three TREM2 transcripts, demonstrating: (1) all three transcripts of them are highly expressed in the human cortex, (2) that up to 25% of the sTREM2 may be due to the expression of a specific isoform and not TREM2 cleavage; and (3) that TREM2 risk variants do not affect expression levels, suggesting that the effect of the TREM2 variants on CSF levels occurs at post-transcriptional level.

Abstract: Baratela-Scott syndrome (BSS) is a rare, autosomal-recessive disorder characterized by short stature, facial dysmorphisms, developmental delay, and skeletal dysplasia caused by pathogenic variants in XYLT1. We report clinical and molecular investigation of 10 families (12 individuals) with BSS. Standard sequencing methods identified biallelic pathogenic variants in XYLT1 in only two families. Of the remaining cohort, two probands had no variants and six probands had only a single variant, including four with a heterozygous 3.1 Mb 16p13 deletion encompassing XYLT1 and two with a heterozygous truncating variant. Bisulfite sequencing revealed aberrant hypermethylation in exon 1 of XYLT1, always in trans with the sequence variant or deletion when present; both alleles were methylated in those with no identified variant. Expression of the methylated XYLT1 allele was severely reduced in fibroblasts from two probands. Southern blot studies combined with repeat expansion analysis of genome sequence data showed that the hypermethylation is associated with expansion of a GGC repeat in the XYLT1 promoter region that is not present in the reference genome, confirming that BSS is a trinucleotide repeat expansion disorder. The hypermethylated allele accounts for 50% of disease alleles in our cohort and is not present in 130 control subjects. Our study highlights the importance of investigating non-sequence-based alterations, including epigenetic changes, to identify the missing heritability in genetic disorders.

Abstract: Transforming growth factor-β (TGF-β) is major inducer of epithelial-to-mesenchymal transition (EMT), which associates with cancer cell metastasis and resistance to chemotherapy and targeted drugs, through both transcriptional and non-transcriptional mechanisms. We previously reported that, in cancer cells, heightened mitogenic signaling allows TGF-β-activated Smad3 to interact with poly(RC) binding protein 1 (PCBP1) and together they regulate many alternative splicing events that favors expression of protein isoforms essential for EMT, cytoskeletal rearrangement, and adherens junction signaling. Here we show that the exclusion of TGF-β-activated kinase 1 (TAK1) variable exon 12 requires another RNA-binding protein, Fox-1 homolog 2 (Rbfox2), which binds intronic sequences in front of exon 12 independently of the Smad3-PCBP1 complex. Functionally, exon 12-excluded TAK1∆E12 and full-length TAK1FL are distinct. The short isoform TAK1∆E12 is constitutively active and supports TGF-β-induced EMT and nuclear factor kappa B (NF-κB) signaling, whereas the full-length isoform TAK1FL promotes TGF-β-induced apoptosis. These observations offer a harmonious explanation for how a single TAK1 kinase can mediate the opposing responses of cell survival and apoptosis in response to TGF-β. They also reveal a propensity of the alternatively spliced TAK1 isoform TAK1∆E12 to cause drug resistance due to its activity in supporting EMT and NF-κB survival signaling.