Identifying changes to cellular physiology that drive multiple sclerosis risk
Final Report Abstract
Multiple sclerosis (MS) is a chronic autoimmune inflammatory disorder of the central nervous system. The risk for developing MS is heritable, and over 200 genetic risk factors across the entire genome have been identified. The molecular and cellular mechanisms driving MS risk, however, are yet to be understood and the translation of these genetic findings into an understanding of underlying pathophysiology is challenging. In order to influence a complex trait, a genetic variant must also act on underlying molecular and cellular traits. These traits are therefore by definition also genetic traits and are altered by the same genetic variant. One approach to understand genotype-phenotype associations, therefore, is to exploit pleiotropic effects where one genetic variant is associated with multiple traits, to identify molecular and cellular traits underlying a genetic association with a complex trait. Such cellular and molecular traits - especially gene expression levels - have been systematically mapped across many cell types, usually as quantitative traits in cohorts of healthy donors. The aim of this project was the identification of shared genetic effects between cellular immune traits, gene expression traits and MS risk using JLIM, a recently developed robust method to determine whether two different traits are driven by the same underlying genetic effect at a single genetic locus. Using cellular immune phenotypes from healthy individuals and gene expression data from three different immune cell types - T cells, monocytes and neutrophils - we could successfully identify a small number of cellular immune phenotypes and gene expression traits that share a genetic effect with MS risk. As the next logical step, we directly assessed for causality between these identified traits using a Mendelian randomization approach. These analyses could however not provide direct evidence for causality between these gene expression and cellular immune traits as well as MS risk, which might be explained by a lack of power due to the relatively small size of the available data sets. We therefore expanded the scope of this project to assess on a broader scale whether the identification of shared genetic effects between traits is meaningful in the context of identifying causal relationships between traits. To identify causally linked traits directly via causality testing, without prior information, we are obliged to test a variety of outcomes across different cell types and conditions experimentally, which is not scalable in many cases. One alternative approach is to exploit shared associations in an unbiased fashion across traits to generate mechanistic hypotheses, which can then be tested in a more focused way. To test this approach, using JLIM, we first identified pairs of gene expression and immune phenotype traits that share genetic association signals. We then used polygenic risk scores to demonstrate that trait pairs with shared genetic associations are more likely to share a broader genetic correlation, as we would expect for causally linked traits. Lastly, we directly assessed causality between these traits by using two different Mendelian randomization approaches and could show that evidence for causality is higher in trait combinations with evidence for a shared genetic effect. In summary, we demonstrate that we can find genetic associations shared between immune cell traits and gene expression traits and show that traits sharing an association signal tend to be causally related, rather than subject to horizontal pleiotropy, where a genetic effect acts on two otherwise independent traits. These results show that shared association testing is a valuable tool for translating genetic associations to complex traits into testable molecular, cellular and physiological mechanistic hypotheses.
Publications
- Shared associations identify causal relationships between gene expression and immune cell phenotypes. Communications Biology, Vol. 4. 2021: 279.
Gasperi C., Chun S., Sunyaev S.R., Cotsapas C.
(See online at https://doi.org/10.1038/s42003-021-01823-w)
