Identifying the genetic determinants of transcription factor activity

Abstract

Analysis of parallel genotyping and expression profiling data has shown that mRNA expression levels are highly heritable. Currently, only a tiny fraction of this genetic variance can be mechanistically accounted for. The influence of trans-acting polymorphisms on gene expression traits is often mediated by transcription factors (TFs). We present a method that exploits prior knowledge about the in vitro DNA-binding specificity of a TF in order to map the loci ('aQTLs') whose inheritance modulates its protein-level regulatory activity. Genome-wide regression of differential mRNA expression on predicted promoter affinity is used to estimate segregant-specific TF activity, which is subsequently mapped as a quantitative phenotype. In budding yeast, our method identifies six times as many locus-TF associations and more than twice as many trans-acting loci as all existing methods combined. Application to mouse data from an F2 intercross identified an aQTL on chromosome VII modulating the activity of Zscan4 in liver cells. Our method has greatly improved statistical power over existing methods, is mechanism based, strictly causal, computationally efficient, and generally applicable.

Figure 1

Overview of our transcription-factor-centric approach to detecting trans-acting sequence variation. (A) We construct a matrix containing the promoter-binding affinity for each combination of the upstream non-coding sequence of a particular gene and the position-specific affinity matrix (PSAM) of a particular transcription factor (TF). (B) The promoter-binding affinity matrix is interpreted as a regulatory connectivity matrix and used to infer a matrix containing the regulatory activity of each TF in each segregant. For each segregant independently, multivariate genome-wide linear regression of segregant-specific differential mRNA expression on the matrix of promoter affinity for all TFs is performed. The coefficients from this linear fit represent (differential) protein-level TF activities. (C) For each TF independently, we treat the inferred activity as a quantitative phenotype and use genetic linkage analysis across all segregants to identify loci that genetically modulate TF activity. Whenever TF activity is statistically associated with genotype at a particular genetic marker, this shows as a high log-odds (LOD) score indicating the presence of a TF activity quantitative trait locus, or ‘aQTL’.

Figure 2

Inferred differences in TF activity between the BY and RM parental strains. Shown are the t-values corresponding to the regression coefficients in a multivariate linear model that predicts genome-wide differential mRNA expression from predicted binding affinity of upstream promoter regions.

Figure 3

Overview of the trans-acting genetic modulators of TF activity mapped using our method. All transcription factors that have at least one significant aQTL region at a 5% FDR are shown. Transcription factors are sorted according to the chromosomal position of their maximum LOD score. Putative causal gene assignments are indicated in green (local aQTL: TF encoded by gene in aQTL) or red (protein–protein interaction identified between TF and gene in aQTL).

Figure 4

(A) Inferred activity of Stb5p in parental strains and segregants. The first and second columns show the activity of Stb5 in six replicates of a BY-reference comparison and six replicates of a RM-reference comparison. The third and fourth columns show the activity of Stb5p for segregants that inherited the BY and RM allele, respectively, at the STB5 locus. (B) LOD score profile for the activity of Stb5p. An asterisk denotes the STB5 locus.

Figure 5

(A) Activity of Fkh1p and Fkh2p across all segregants. The activity of Fkh1p is negatively correlated with that of Fkh2p. The yellow dots correspond to segregants carrying the BY allele at the CDC28 locus, the green dots to those carrying the RM allele. (B) Schematic diagram illustrating the antagonistic modulation of Fkh1p and Fkh2p by Cdc28p. Although the transcriptional targets of Fkh1p are more highly expressed in segregants carrying the BY allele at the CDC28 locus, the opposite is true for the targets of Fkh1p.

Figure 6

(A) Inferred activity of Zscan4p across all F2 mouse population. Each column shows the activity of Zscan4 in homozygous C57BL/6J (BB), heterozygous (BD), and homozygous DBA/2J (DD) mice at aQTL positions, respectively. (B) LOD score profile for Zscan4p.