association between smoking and atherosclerosis was mostly reflected (or mediated) by its effect on the expression of these four genes. The fact that PTGDS and SASH1 expression remained associated with carotid plaques after adjustment on smoking status may indicate a broader implication of these genes in atherosclerosis than the sole effect induced by smoking. However it is also possible that the expression of these two genes more faithfully reflects tobacco consumption than the dichotomous variable used to define smoking. This illustrates the dual aspect of the transcriptome which may be viewed either as an element in a causal chain or as reflecting ongoing processes with no implied causation. Because genetics may help to dissect causal pathways, we examined whether the best cis eSNPs associated with expression of the smoking-related genes were also associated with carotid atherosclerosis, but no such association was detected (Table 7).
