Genetic variation can affect molecular circuits that control mRNA levels1-4
. In particular, genetic variants that affect the absolute abundance of transcripts are called expression quantitative trait loci (eQTLs), whereas variants that affect changes in transcript abundance after perturbation by external stimuli are referred to as responsiveness expression quantitative trait loci (reQTLs)4
. Recent studies demonstrated the utility of testing reQTLs in the context of a single stimulus (e.g.
, ionizing radiation) in yeast and mammals1, 4-10
; some reQTLs could be detected in response to stimulus, but not using baseline mRNA abundance levels.
Although most molecular circuits can respond to multiple distinct stimuli, it remains unknown how the same reQTL would behave in response to different stimuli in the same cell type. The reQTL associated with a given gene could either affect that gene's responsiveness to all the stimuli to which the gene responds (stimulus non-specific reQTLs; , left), or may influence gene expression in response to some but not all of the stimuli to which the gene responds (stimulus-specific reQTLs, , right,). Earlier studies monitored reQTLs only during a single stimulus and hence could not distinguish these possibilities.
Determining the stimulus-specificity of a reQTL can, in principle, allow us to position it within complex circuits. For example, consider a circuit with three input signals, two target genes, and the network connecting them (). Determining a trans-reQTL in just one stimulus (e.g., s2, ), would position it upstream of its target gene(s) (e.g., x1, ), but cannot predict on which of the multiple paths from stimulus to target it resides (e.g., both pathways 2 and 3 are possible positions, ). However, if one determines a trans-reQTL across a sufficiently diverse set of stimuli (e.g., s1, s2, and s3, ), it can be positioned on a particular pathway in the circuit (e.g., pathway 2, ). A similar rationale applies to cis-acting variants (). If the genetic architecture cannot be explained in the context of the known circuitry, we can use the inconsistency to refine the circuit with new putative connections.
Here, we systematically study these questions by developing a framework, involving several new computational methods, for multi-stimulus reQTL analysis. We apply this approach to study the transcriptional responsiveness of bone-marrow-derived primary dendritic cells (DCs) to stimulation with different pathogen components, across a panel of recombinant inbred (RI) BXD mice, generated by crossing the parental C57BL/6J (B6) and DBA/2J (D2) strains11
. We find that stimulation-specific reQTLs are common, both in cis
and in trans
, and that many trans
-reQTLs likely function the same way in response to different stimuli. By comparing the responsiveness of genes and their underlying reQTLs across multiple stimuli, we refine known molecular circuits and infer the position of reQTLs within them.