Our quantitative protein-level measurements show that, in any given growth condition, responsiveness to paralog deletion is restricted to a small number of genes. Responsiveness occurs at both the transcriptional and posttranscriptional level. With almost no exceptions, such paralog responsiveness occurs only when the genes are synthetic lethal, namely, when they have an overlapping biochemical function that is critical for growth in the tested conditions. Removing or adding the need of a function, either by supplying its end product or by shifting to conditions in which its product is not required, specifically determines whether or not a given gene will respond to deletion of its paralog.
The mechanisms underlying need-based responsiveness are most likely complex. In principle, responsiveness of a gene to deletion of its paralog could reflect either a direct response to the absence of the paralogous protein (similar to supply control), or an indirect response to the absence of its function (similar to demand control 
) (Figure S8A
. A simple mathematical model of a metabolic pathway exemplify that indirect responsiveness should depend on the presence of the product of the pathway in the environment (Figure S8B and S8C
; Text S2
). Indeed, we found that for the amino acid biosynthetic genes, the addition of the amino acid end product eliminates paralog responsiveness (), suggesting that responsiveness is not due to the absence of the paralogous protein but rather to the absence of its function. Such paralog responsiveness may therefore reflect a simple end-product regulation of genes. This supports the demand strategies previous identified in glycolysis 
. Indeed, feedback regulation often occurs in the first committed step of a pathway, and these metabolic branching points are known to be enriched for duplicated genes 
This logical argument is based on the notion that addition of the end product of a pathway supplements its biosynthetic function. The argument, therefore, does not apply to conditions that instead of supplying the end product simply remove the need of the function. For example, yeast cells need to accumulate glycerol only in osmotic stress; removing the osmotic stress relieves the need for the glycerol biosynthetic pathway not by externally supplying its end product, glycerol, but rather by generating conditions in which this end product is not needed. This is in contrast to the case of the amino acid biosynthetic genes; we therefore cannot conclude from our data that the mechanism underlying responsiveness of Hxk1, Rhr2, and Gpd2 is indirect. Indeed, the responsiveness of Hxk1 may be mediated by direct regulation of its paralog; nuclear Hxk2 is involved in repression of HXK1
and expression of its own gene, HXK2 
. In agreement with these observations, we find that either the absence of glucose or the absence of HXK2
results in Hxk1 up-regulation (). These differences in the underlying mechanisms of responsiveness underscore the breadth of its functional roles and suggest that in some cases, responsiveness to paralog deletion could even depend on the presence of other (nonparalogous) genes 
Genetic redundancy is a salient feature of living organisms. It has long been discussed under what circumstances genetic redundancy is evolutionary stable 
and how redundancy can contribute to genetic robustness 
. Interestingly, we uncovered a set of genes that are not up-regulated under a specific condition unless their paralogs are deleted. This and other cases of need-based responsiveness of genes to the absence of their paralogs could play an adaptive role in the compensation of functions that are compromised by genetic, environmental, or stochastic perturbations.