Statistical tests for natural selection on regulatory regions based on the strength of transcription factor binding sites

Abstract

Background: Although cis-regulatory changes play an important role in evolution, it remains difficult to establish the contribution of natural selection to regulatory differences between species. For protein coding regions, powerful tests of natural selection have been developed based on comparisons of synonymous and non-synonymous substitutions, and analogous tests for regulatory regions would be of great utility.

Results: Here, tests for natural selection on regulatory regions are proposed based on nucleotide substitutions that occur in characterized transcription factor binding sites (an important type functional element within regulatory regions). In the absence of selection, these substitutions will tend to reduce the strength of existing binding sites. On the other hand, purifying selection will act to preserve the binding sites in regulatory regions, while positive selection can act to create or destroy binding sites, as well as change their strength. Using standard models of binding site strength and molecular evolution in the absence of selection, this intuition can be used to develop statistical tests for natural selection. Application of these tests to two well-characterized regulatory regions in Drosophila provides evidence for purifying selection.

Conclusion: This demonstrates that it is possible to develop tests for selection on regulatory regions based on the specific functional constrains on these sequences.

Figure 1

Changes in binding site strength in the absence of selection. a) shows time courses for the strength of a transcription factor binding site (S) in a simulation of evolution without selection. The strength of a real binding site (dotted trace) usually decreases from the strength expected for real binding sites (E[S|motif]) to that expected under background residue frequencies (E[S|bg]). An example of a binding site created from background sequence in the absence of selection (solid trace) is also shown. b) shows the probability (p) of observing a change in score of size ΔS given than a sequence is a binding site (dotted trace) or a background sequence (solid trace).

Figure 2

Distribution of the proposed statistic under the null hypothesis. In a simulation of molecular evolution under the null hypothesis (see text for details) the statistic proposed shows good agreement with the expected standard normal behavior (dotted trace) either using the mean and variance of ΔS computed exactly (unfilled bars) or in the long-time limit (filled bars). Inset is a comparison of the P-value as computed under the standard normal assumption and the number of times that value of statistic or greater was observed in 1000 simulations, using either the exact (Xs) or long-time limiting (squares) values for the mean and variance of ΔS.

Figure 3

Dependence of the distribution of ΔS on evolutionary parameters. a) shows the probability distribution of ΔS as evolutionary distance varies (coloured solid traces) for the Bcd matrix under an HKY model with transition-transversion rate ratio set to 2. The distribution rapidly converges to the long-time limit distribution (dotted trace). b) shows the probability distribution of ΔS as the transition-transversion rate ratio (kappa) varies (coloured solid traces) for the Bcd matrix under an HKY model with evolutionary distance set equal to 0.3 substitutions per site. Once again, the distribution converges to the long-time limit (dotted trace). Inset in both is the convergence of the mean of ΔS (squares) to the long-time limit (dotted trace). Distributions are for real binding sites evolving in the absence of selection.

Figure 4

Characterized Bcd sites in the hb anterior activator. a) schematic of the Bcd sites in the hb anterior activator. b) alignments of the binding sites. Bold residues indicate the substitutions that were included in the analysis. Numbers below the alignments indicate the relative positions in the Bcd binding motif. Abbreviations are D. - Drosophila, vir - virilis, pse - pseudoobscura, ana - ananassae, ere - erecta, yak - yakuba, sim - simulans, sec - sechelia, mel - melanogaster