Gene co-regulation is highly conserved in the evolution of eukaryotes and prokaryotes

Abstract

Differences between species have been suggested to largely reside in the network of connections among the genes. Nevertheless, the rate at which these connections evolve has not been properly quantified. Here, we measure the extent to which co-regulation between pairs of genes is conserved over large phylogenetic distances; between two eukaryotes Caenorhabditis elegans and Saccharomyces cerevisiae, and between two prokaryotes Escherichia coli and Bacillus subtilis. We first construct a reliable set of co-regulated genes by combining various functional genomics data from yeast, and subsequently determine conservation of co-regulation in worm from the distribution of co-expression values. For B.subtilis and E.coli, we use known operons and regulons. We find that between 76 and 80% of the co-regulatory connections are conserved between orthologous pairs of genes, which is very high compared with previous estimates and expectations regarding network evolution. We show that in the case of gene duplication after speciation, one of the two inparalogous genes tends to retain its original co-regulatory relationship, while the other loses this link and is presumably free for differentiation or sub-functionalization. The high level of co-regulation conservation implies that reliably predicted functional relationships from functional genomics data in one species can be transferred with high accuracy to another species when that species also harbours the associated genes.

Figure 1

Measuring conservation of co-regulation between E.coli and B.subtilis. This figure illustrates the various contexts in E.coli that the genes from an operon in B.subtilis can occur in. The various possibilities should be taken into account when measuring conservation of co-regulation between the two species: genes can be absent from E.coli (the blue gene), genes can still be together in an operon (the red and green gene); genes can be in different operons but still share the same transcription factor (the red/green genes versus the yellow gene) or a gene can be in another operon and be regulated by a different transcription factor (the purple gene versus the others).

Figure 2

A scatter plot of co-expression correlations from microarray experiments and correlation in transcription factor binding profiles from ChIP experiments. Correlations between expression profiles are computed in the normal fashion (see Methods). Correlations between the ChIP binding profiles are used here to allow a quantitative evaluation of the relation between both data sets, which would be conceptually complicated if the ChIP-on-chip data were treated by the presence/absence of transcription factor binding. The correlation coefficient is 0.034. This is a very weak correlation, but it nevertheless is very significant with P < 10⁻¹⁶ when tested against the normal distribution.

Figure 3

Possible outcomes of co-regulation evolution after gene duplication. This figure illustrates the dilemma with respect to inparalogs, which occurs when studying co-regulation. We delineate three different situations: complete conservation, partial conservation and complete loss. We argue (see text) that gene co-regulation is conserved, not only in the case of complete conservation but also in the case of partial conservation. Ce denotes C.elegans and Sc denotes S.cerevisiae.

Figure 4

Distribution of rs of C.elegans orthologous pairs. We plot the r of the C.elegans orthologous pairs of reliably co-regulated yeast pairs to measure the degree of conservation. The distribution shows a very clear tendency to high co-expression correlation coefficients. The amount of conservation can be measured by comparing this distribution to the distribution of rs among all C.elegans genes.