Coordinated evolution of co-expressed gene clusters in the Drosophila transcriptome

Abstract

Background: Co-expression of genes that physically cluster together is a common characteristic of eukaryotic transcriptomes. This organization of transcriptomes suggests that coordinated evolution of gene expression for clustered genes may also be common. Clusters where expression evolution of each gene is not independent of their neighbors are important units for understanding transcriptome evolution.

Results: We used a common microarray platform to measure gene expression in seven closely related species in the Drosophila melanogaster subgroup, accounting for confounding effects of sequence divergence. To summarize the correlation structure among genes in a chromosomal region, we analyzed the fraction of variation along the first principal component of the correlation matrix. We analyzed the correlation for blocks of consecutive genes to assess patterns of correlation that may be manifest at different scales of coordinated expression. We find that expression of physically clustered genes does evolve in a coordinated manner in many locations throughout the genome. Our analysis shows that relatively few of these clusters are near heterochromatin regions and that these clusters tend to be over-dispersed relative to the rest of the genome. This suggests that these clusters are not the byproduct of local gene clustering. We also analyzed the pattern of co-expression among neighboring genes within a single Drosophila species: D. simulans. For the co-expression clusters identified within this species, we find an under-representation of genes displaying a signature of recurrent adaptive amino acid evolution consistent with previous findings. However, clusters displaying co-evolution of expression among species are enriched for adaptively evolving genes. This finding points to a tie between adaptive sequence evolution and evolution of the transcriptome.

Conclusion: Our results demonstrate that co-evolution of expression in gene clusters is relatively common among species in the D. melanogaster subgroup. We consider the possibility that local regulation of expression in gene clusters may drive the connection between adaptive sequence and coordinated gene expression evolution.

Figure 1

Relationships among the seven species in the D. melanogaster subgroup that were analyzed. Times of speciation events follow estimates from [26].

Figure 2

Sliding window heat map of p-values resulting from the clustering analysis across species projected onto the D. melanogaster genome. Approximate position on chromosomes is plotted along the x-axis and window size on the y-axis. Centromere proximal regions are indicated by yellow shading on the chromosome. The spectrum runs from highly significant p-values (red) to highly non-significant p-values (dark blue).

Figure 3

Heat map of p-values resulting from the sliding window analysis within D. simulans projected onto the D. melanogaster genome. Color coding follows Figure 2.

Figure 4

Effects of probe masking on the estimated expression level. Each circle reflects the slope of the regression divided by mean squared error for a single probe set for a regression using all 48 samples: A) Effect of masking random probes for probe sets where all 14 probes are perfect matches for both D. melanogaster and D. yakuba, B) Effect of masking in probe sets where probes masked have diverged between D. melanogaster and D. yakuba (all other probes in a set are perfect matches for both species), C) Effect of masking for all probe sets included in the analysis.