Alternative splicing in complex disease – Does it play a causal role?
Alternative splicing of mRNA molecules is an important feature of virtually every complex disease. Why? Complex diseases are associated with alterations in many genes. In addition, genetic variations in the human genome can exert numerous effects on gene expression including changes in RNA splicing. RNA alternative splicing is the means whereby a gene produces messenger RNAs (mRNA) containing different collections of exons, that when translated into protein, alter the protein’s function. Significant associations between RNA splicing events and disease phenotypes have been documented in many complex diseases including those involving the brain, such as alcohol use disorder (AUD). Hence, RNA alternative splicing is an important quantitative molecular trait (i.e., endophenotype) for complex disease. Nevertheless, whether RNA splicing is a causal factor or merely a result induced by the disease itself, is not known. So, how do we tell if the association between RNA splicing and disease stems from genetic variation, or is simply an effect of the disease?
Challenge
This question is one that geneticists would love to answer but it is very difficult to address. The major challenge in determining the contribution of a splicing event to the risk of disease is that splicing changes contributing to a disease are always confounded by the splicing changes induced by the disease. The key to addressing this conundrum is to dissect the genetic component of the splicing outcome from cellular environmental effects using computational prediction models.
Mendelian randomization-based approach for RNA splicing analysis in alcohol use disorder
In theory, there are multiple ways to build prediction models to map genotypes to splicing outcomes. However, our aim was to be able to directly impute splicing outcomes from individual genotypes. The benefits of this approach are threefold: (1) achieving greater precision and specificity; (2) enabling direct experimental verification; and (3) utilizing individual genotype information. To this end, we adapted PrediXcan, a genotype-based prediction method for gene expression, into a novel platform to predict RNA splicing. Using splicing models built upon data from the CommonMind Consortium (CMC), we further designed a Mendelian randomization (MR)-based approach and applied it to AUD datasets from COGA and OZ-ALC to identify causal RNA alternative splicing events. Next, we further replicated our results in large GWAS via Generalized Summary-based MR. We also devised a bioinformatics approach to predict the downstream genes regulated by these events. Our discoveries differ in nature from previous reports, which mainly focused on splicing associations with AUD status. Our study reported, for the first time, how splicing events in specific genes may lead to alcohol addiction. AUD is a prevalent psychiatric disorder that is a major preventable cause of death affecting over a hundred million people around the world (29.5 million in the United States alone). The methodological framework that we established will facilitate translational studies in public health by prioritizing new causal genes whose splicing is genetically regulated. Our approach is also applicable to RNA splicing studies of other complex diseases, e.g., cannabis, opioid and nicotine use disorders, Parkinson’s disease, Alzheimer’s disease, or dementia.
Alternative splicing of genes contributes to AUD risk
We identified six novel splicing events using our MR-based framework. One skipped exon in ELOVL7 is an outstanding contributor to AUD risk. Interestingly, this exon is in the 5’ untranslated region of the gene which is not involved in coding protein sequence. Typically, such an event would not likely be appreciated in genetic studies because it has no direct links to protein function. Nonetheless, we found that the change of the genetic propensity for the inclusion of this exon correlates with differential expression of ELOVL7 itself, as well as more than 200 other genes. Therefore, as a noncoding exon, this putative causal splicing event may be involved in post-transcriptional or translational regulation. Of note, ELOVL7 was initially identified as downregulated in prefrontal cortex in individuals with alcohol dependence. However, their finding could be interpreted as the change in ELOVL7 expression either contributes to alcoholism or just results from alcohol exposure. Now our finding indicates that the alternative splicing of ELOVL7 clearly contributes to the risk of AUD. In this way, our study revitalized this gene from the perspective of RNA splicing. Thereby, our study provides evidence of causality for findings from previous GWAS and differential gene expression analyses; thus, shedding new insights into the molecular mechanisms of AUD. We also found other putative causal splicing events including skipped exons in DRC1, LINC00665, NSUN4, SRRM2 and TBC1D5; and most of these exons are also non-coding. Their host genes are relevant to alcohol-related diseases, neurological disorders, or cancers. Similarly, their roles, including the causality in AUD, have not been reported anywhere before.
What pathways do the findings highlight? – Neuro, immune, and neuroimmune
Because assessing the functions of splicing events in the brain is difficult, we devised a rigorous bioinformatics approach to predict the downstream genes regulated by the splicing events. We found that each of the splicing events mediates differential expression for a number of downstream genes, varying from several to several hundreds. Although each event regulates a distinctive set of downstream genes, we found significant consensus at the level of biological pathways. These pathways include neural pathways such as gliogenesis, neurogenesis, and nervous system development; as well as immune pathways such as multiple inflammatory, cytokine and antiviral response cascades. Most significantly, neuroimmune pathways were commonly enriched, specifically, TNF, NF-κB, IL6-JAK-STAT3, IL2-STAT5, NOD-like receptor (NLR) signaling pathways, as well as T cell activation and differentiation. Our results not only provide additional evidence that immune activation or neural activities are consequential effects of problematic alcohol use, but also provide new evidence that these pathways contribute to the development of AUD. Thus, our findings strongly emphasize that the multifaceted molecular functions of neuroimmune pathways in alcohol dependence are much more profound than previously recognized.
Another interesting finding was that complex regulatory pathways such as the epithelial-mesenchymal transition (EMT) were also enriched by the downstream genes of the AUD-causal splicing events. EMT underlies many crucial bioprocesses including neural tube formation and cancer metastasis. It is unclear why EMT was impacted by the AUD-causal splicing events; however, LINC00665 is known to regulate EMT in cancer, and alcohol stimulates EMT in cancer cells leading to progression. Thus, our finding clearly indicates a role of splicing in the development of alcohol-related cancers that is not likely to be a mere bystander effect.
Conclusion
We are still trying to understand the molecular mechanisms that underlie alcohol addiction. But one thing is for sure, RNA splicing plays a profound role in the development of alcoholism. It will be important to study the mechanisms and functions of the RNA-level regulatory events in the brain with advanced experimental techniques and systems, as well as to conduct further functional studies of neuroimmune molecular mechanisms in complex disease. For now, we hope that our data highlight the impact of RNA splicing in the genetics of AUD and provide a foundation for future studies of RNA splicing in complex disease.
Genetics of RNA alternative splicing in alcohol use disorder
