PROJECT SUMMARY/ABSTRACT
The human X chromosome has long been hypothesized to play a significant role in the etiology of sex-biased
diseases and traits, particularly autoimmune diseases like lupus. Despite its importance, the X chromosome is
largely understudied in genetic association and functional genomics studies. Such an omission is largely due to
its unique biology. The hemizygosity of XY males necessitates XX females to achieve dosage compensation of
X-linked genes by the means of X-chromosome inactivation (XCI). Thus, in females most genes are expressed
only from the active X (Xa) and remain silent on the inactive X (Xi). However, up to 10% of genes consistently
escape XCI in healthy females and are transcribed from both Xa and Xi. And 15-30% of genes variably escape
XCI in a subset of females or tissues. We hypothesize such inter- and intra- individual heterogeneity is genetically
influenced and disease relevant. However, the genetic architecture of X-linked genes and XCI escape remains
poorly understood. We recently showed, for the first time, that XCI escape has significant heritability and variable
XCI escape genes have a significantly increased enrichment of heritability in female-biased traits when
compared to sex-balanced traits. Although promising, the annotations for XCI states were inferred only from
lymphoblast cell lines, which could differ across human tissue/cell types. Recent works have shown disease
heritability is enriched in regions surrounding genes specific to disease relevant tissues. Thus, matching each
trait to relevant tissue/cell type specific XCI states in heritability analysis could pinpoint the disease and trait
relevant tissue/cell types in which escape from XCI plays a significant role (Aim 1). To do so, we first propose
an empirical bayes method to infer XCI escape states in an individual sample invariant of XCI mosaicism or
presence of transcribed heterozygous SNPs in population scale bulk RNA-seq data. Unlike previous efforts, our
method maximizes the samples and X-linked genes assayed to construct the most comprehensive XCI escape
landscape across human tissue/cell types to date. Such a complete map will enable robust heritability estimation.
Next, to understand the genetic influence on variable XCI escape, we propose a two-step method that accurately
models XCI mosaicism and genetic regulation of Xa/Xi by jointly modeling male and female samples to detect
associations with Xa and Xi expression levels (Xa-/Xi- QTL) (Aim 2). Our method offers substantial advancement
and improves power to detect Xa-/Xi- QTLs compared to other approaches that 1) assume genetic regulation on
Xa and Xi are largely similar or 2) attribute total expression of X-linked genes to expression from Xa and Xi. We
will apply our approach across tissue/cell types and integrate the identified Xa-/Xi- QTL with existence genome
wide association studies to identify tissue/cell type specific Xa and Xi gene and trait associations. We will apply
our methods to some of the largest datasets for a variety of traits including lupus, diabetes, and addiction. Overall,
our proposed methods will allow comprehensive assessment of the role the X chromosome and XCI escape
plays in the etiology of diseases and traits, particularly those that are sex-biased like autoimmune diseases.
