• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Sep 13, 2011; 108(37): 15276–15281.
Published online Aug 29, 2011. doi:  10.1073/pnas.1105814108
PMCID: PMC3174679
Evolution

Two types of cis-trans compensation in the evolution of transcriptional regulation

Abstract

Because distant species often share similar macromolecules, regulatory mutations are often considered responsible for much of their biological differences. Recently, a large portion of regulatory changes has been attributed to cis-regulatory mutations. Here, we examined an alternative possibility that the putative contribution of cis-regulatory changes was, in fact, caused by compensatory action of cis- and trans-regulatory elements. First, we show by stochastic simulations that compensatory cis-trans evolution maintains the binding affinity of a transcription factor at a constant level, thereby spuriously exaggerating the contribution of cis-regulatory mutations to gene expression divergence. This exaggeration was not observed when changes in the binding affinity were compensated by variable transcription factor concentration. Second, using reciprocal introgressions of Drosophila, we show that relative expression of heterozygous alleles from two distinct species often varied significantly between different species backgrounds, indicating the possible action of cis-trans compensation. Taken together, we propose that cis-trans hybrid incompatibilities are accumulating much faster than generally considered.

Keywords: compensatory evolution, trans-regulatory mutation, epistasis, cis-trans interaction, reproductive isolation

A first step to understanding the evolution of gene expression is to decompose expression variation into cis- and trans-regulatory components (14). This decomposition can be done by combining an allele-specific expression assay using F1 hybrids with expression assays in pure species backgrounds (1, 4). From these assays, two kinds of expression differences are determined for a gene: (i) difference between two parental strains and (ii) difference between two alleles in an F1 hybrid. When the difference between parental strains is entirely due to trans-regulatory changes, we would not expect any difference between the two alleles in the hybrid condition. However, when the difference between parental strains is entirely due to cis-regulatory changes, this difference should be reproduced as an allelic difference in the hybrid. A recent application of this method identified a greater contribution of cis-regulatory changes in inter- than intraspecific comparisons, suggesting that cis-regulatory mutations are fixed more preferentially during evolution than trans-regulatory mutations (5).

In the above decomposition, allelic differences in F1 hybrids are entirely attributed to cis-regulatory variation under the assumption that trans-regulatory factors should affect both alleles equally. However, some types of cis-trans interactions can violate this assumption by differently affecting the two alleles. For example, by adapting to an evolving cis-regulatory element of the same species, a transcription factor may bind less efficiently with the cis-regulatory element from the distinct species. In the extreme case where the binding between species does not occur at all, an allelic difference in a hybrid would be similar to the parental difference as if there were only cis-regulatory variations. Thus, when incompatibilities between cis- and trans-regulatory elements reduce heterospecific binding, the contribution of cis-regulatory changes to gene expression divergence would be spuriously exaggerated.

Here, by specifically discriminating two distinct types of cis-trans compensation, we show, based on stochastic simulations, that compensatory regulatory evolution under stabilizing selection could reduce heterospecific binding and therefore explain the hypothetical increase in the relative contribution of cis-regulatory changes without invoking preferential fixation of cis- over trans-regulatory mutations (5). We also show by allele-specific expression assays that the relative expression of heterospecific alleles from closely related Drosophila species often varies depending on the genetic background. Combining the simulation and experiential observations, we hypothesize that cis-trans hybrid incompatibilities are evolving much faster than previously thought.

Stochastic Simulations

One-Activator, One-Target System.

In the simplest case with a pair of a transcriptional activator and its target gene, our simulations keep track of evolutionary changes in the concentration of the trans-acting activator ([T]) and its dissociation constant to the cis-regulatory element of the target gene (K) (Simulation Methods). These two properties jointly regulate the expression level of the target gene; more precisely, it is expressed as a monotonically increasing function of the quotient [T]/K. The gene expression level, in turn, affects the fitness of an individual. The dissociation constant K is also described as a function of two variables (a and c), each variable characterizing the binding properties of the activator and the cis-regulatory element, respectively. Defined in this way, two distinct types of compensatory cis-trans changes are expected under stabilizing selection: (i) K-[T] compensation, which refers to correlated evolution of K and [T] in such a way that an increase (or decrease) in the transcriptional activator abundance cancels out the decreased (increased) binding affinity as measured by the inverse of the dissociation constant (and vice versa), and (ii) ac compensation, which refers to coordinated changes in a and c that maintain the binding affinity at a constant level.

Following the previous study (4), we studied the linear regression of the relative allelic expression in F1 hybrids on the relative expression difference between parental strains. When both types of compensatory changes were allowed in the simulations (Fig. S1), we consistently observed a gradual reduction in heterospecific binding (Fig. 1) and an asymptotic increase in the linear regression over time (Fig. 2, black line). These dynamical changes are reflected in the scatter plot of the relative allelic expression in hybrids against the relative expression between strains (4), where we observed a continual increase in the number of genes for which the relative allelic expression in hybrids did not exceed the parental difference (referred to as cis + trans genes in ref. 6); this increase was accompanied by a concomitant decrease in the number of genes for which the allelic difference in hybrids was greater than or in the opposite direction from the parental difference (cis × trans genes in ref. 6) (Fig. S2A).

Fig. 1.
Temporal changes in the binding affinity of the transcriptional activator to its cis-regulatory element. Coevolutionary dynamics of the transcriptional activator concentration [T] (as a function of f) (Simulation Methods) and its dissociation constant ...
Fig. 2.
Temporal changes in the linear regression of the relative allelic expression in F1 hybrids on the relative expression difference between parental strains. We computed the linear regression of the relative allelic expression in F1 hybrids on the relative ...

By contrast, when a was held constant so that ac compensation was prohibited, no increasing tendency of the linear regression was observed (Fig. 2, gray line), despite the action of K-[T] compensation. In this case, the expression level as a function of [T]/K was even larger in the hybrids (for which K is expected to be uncorrelated with [T]) than in the parental strains (for which K is expected to be positively correlated with [T] under K-[T] compensation). We also observed many cis × trans genes but an almost complete absence of the cis + trans genes (Fig. S2B).

One-Activator, Two-Target System.

Many transcription factors regulate the expression of multiple genes pleiotropically. This pleiotropic effect imposes strong selective constraints on both [T] and a, which may retard the accumulation of compensatory cis-trans changes. For example, when a transcriptional activator affects multiple targets simultaneously, mutational changes in [T] or a that optimize the expression level of a target gene may deviate the expression levels of other targets away from their optima. Because of this potentially deleterious side effect, trans-regulatory mutations have been considered to play only a limited role in the evolution of transcriptional regulation.

To see how constraints on trans-regulatory changes affect the expression divergence, we conducted additional simulations with two target genes (denoted G1 and G2) that are simultaneously affected by a single activator and independently subjected to stabilizing selection. More specifically, we focused on the expression divergence of gene G1 while changing the strength of selection on the level of G2 expression (s2) (Simulation Methods). Again, we observed a gradual reduction in heterospecific binding and an asymptotic increase in the linear regression over time, even when the trans-regulatory changes may produce deleterious pleiotropic effects (s2 > 0) (Fig. 3). Because a larger value of s2 implies a greater selective constraint on the trans-regulatory changes, we expect that the above results with s2 > 10 apply also when more than two target genes are involved in the regulatory system. Accordingly, a gradual accumulation of compensatory cis-trans changes would still be possible even in the presence of deleterious pleiotropic effects of trans-regulatory changes as long as the number of genes regulated by a single activator is not too large.

Fig. 3.
Temporal changes in the linear regression when a transcriptional activator pleiotropically affects two target genes. We computed the linear regression of the relative allelic expression in F1 hybrids on the relative expression difference between parental ...

Expression Assays

Seeing that a specific form of compensatory cis-trans regulation (here referred to as ac compensation) could, in theory, account for the putative increase in the contribution of cis-regulatory mutations during evolution, we now turn to an empirical issue of whether such cis-trans regulation does indeed occur in nature. When trans-regulatory changes equally affect two alleles in a hybrid, allelic differences in heterozygous individuals can entirely be attributed to cis-regulatory variation, and the relative expression of the two alleles should be the same, regardless of the genetic background (7). By contrast, when ac compensation is present and trans-regulatory changes do not affect the two alleles equally, heterospecific binding would be reduced by the accumulation of incompatibilities between cis- and trans-regulatory elements. As a consequence, the relative expression of heterospecific alleles should vary depending on the genetic background. For example, if expression of an allele should be higher in its conspecific background than in a heterospecific background, the expression ratio of the two alleles may then be greater than unity in one parental background but smaller in the other background.

To test these predictions, we constructed two sets of reciprocal introgression lines between Drosophila species. In the first set, we introgressed a second chromosomal region encompassing the centromere of D. mauritiana into D. simulans and the homologous region of D. simulans into D. mauritiana; in a similar manner, a right arm region of the second chromosome was reciprocally introgressed between D. mauritiana and D. sechellia in the second set. All three species diverged more recently than the species pair used in ref. 5. We then measured the expression ratio of heterospecific alleles in the two different species backgrounds. We chose six genes for the first set of introgressions and a single gene for the second set of introgressions based on available data on expression levels in these species and their hybrids (Materials and Methods for Expression Assays).

The expression ratios differed significantly between the two backgrounds in three of seven genes studied (two genes from the first set and one gene from the second set of introgressions) (Fig. 4). For example, the D. mauritiana allele of vulcan (vlc) gene was expressed lower than the D. simulans allele in the heterospecific (simulans) background, but it was expressed higher in the conspecific (mauritiana) background. Consistent with our hypothesis that ac compensation reduces binding affinities of transcriptional activators to their heterospecific cis-regulatory elements, the relative expression level of the D. mauritiana allele was higher in its conspecific than heterospecific backgrounds in all three genes that showed statistically significant differences.

Fig. 4.
Allele-specific expression assay using Drosophila introgression lines. Relative expression of heterospecific alleles was studied for six genes [CG15219, vulcan (vlc), Actin 42A (Act42A), CG7856, Trap1, and phtf] in a set of reciprocal introgressions into ...

Discussion

The conjecture that the linear regression coefficient represents the relative contribution of cis-regulatory changes to the expression divergence (4, 5) implicitly assumes that the total divergence between parental strains can be exclusively decomposed into cis- and trans-acting components. This decomposition may become inappropriate if trans-regulatory changes differently affect two alleles in an F1 hybrid (8, 9). In fact, our simulation analyses highlight the possibility that the higher regression in inter- than intraspecific comparisons (5) was caused by a reduction in heterospecific binding under ac compensation and not by preferential fixation of cis-regulatory mutations. Similar trends were observed, although in a less evident manner, when the transcriptional activator had pleiotropic effects on multiple targets. This result suggests that the evolutionary role of trans-regulatory mutations would not be much restricted by deleterious side effects of pleiotropy (although the mutational effect of transcription factors may be less pleiotropic than usually considered) (10).

In the simulations, we have explicitly discriminated two types of compensatory cis-trans regulation, namely ac vs. K-[T] compensation. These two types of compensation have generally been studied separately (6, 11M14), and their differential contributions to the evolution of transcriptional regulation have not been anticipated. Moreover, despite a lack of formal proof, it is generally thought that ac compensation, unlike K-[T] compensation (6), evolves too slowly to be seen in a closely related species pair such as D. melanogaster and D. simulans studied in ref. 5; these two species diverged ~1.2 Mya (15). We tackled this conjecture by measuring relative expression of heterospecific alleles in reciprocal introgression lines of three Drosophila sibling species. These species diverged only for ~0.4 Myr (15) but still gave evidence for a possible action of cis-trans compensation, as exemplified here by reduced heterospecific binding.

A clear example of such ac compensation was recently reported in a yeast adaptor protein-1 transcription factor (16). In our own case, there is a putative cis-regulatory change in the odorant-binding protein 57e (Obp57e) gene of D. sechellia, namely a 4-bp insertion ~70 bp upstream of the translation start site (17). When introduced into a D. melanogaster background, a construct containing an upstream sequence of D. sechellia failed to drive reporter expression in any parts of the fly body, but when the 4-bp insertion was removed from the construct, the normal expression was successfully restored (17), a result suggestive of an involvement of this insertion in ac compensation.

For most of the introgression lines used in this study, introgressed segments contain more than 1,000 genes (Materials and Methods for Expression Assays). In addition, the expression of genes outside the introgressed regions may have also been affected (18). Therefore, the introgressions do not represent pure species backgrounds. However, these complexities do not affect our conclusion. The important point here is that, because two alleles are not regulated equally in a hybrid, allelic differences in heterozygotes vary depending on the genetic background and therefore do not adequately quantify the cis-regulatory effects. We found such background dependency in three of seven genes examined (Fig. 4), although this is potentially an overestimation because of the biased choice of genes. For a correct decomposition of expression divergence into cis- and trans-regulatory components, the relative allelic expression in F1 hybrids must be insensitive to the genetic background. We show here that this is not always the case.

Recently, based on a genome-wide survey of mRNA abundance in two species of Drosophila and their hybrids, a relatively large contribution of trans-regulatory changes to expression divergence was reported (19). However, this analysis also suffers from the same methodological problem of ignoring trans-regulatory changes that differently affect two alleles in a hybrid (4, 5), and we emphasize that the extent of cis-regulatory divergence is still overestimated.

In contrast to our interspecific comparisons, background dependency of the relative allelic expression was not observed within species (7, 20). This finding suggests that natural selection is promoting compensatory cis-trans evolution. This is what was found in the simulations, where compensatory cis-trans changes gradually accumulated over time and eventually led to reduced heterospecific binding (Fig. 1). Combining our simulation and experimental results, we here propose that, as a result of pervasive ac compensation, cis-trans hybrid incompatibilities are accumulating on a much shorter timescale than considered previously (11, 12).

The rapid evolution of ac compensation has several implications. First, like other compensatory changes (6, 21), it can potentially lead to a rapid development of reproductive isolation. Second, what many think of as adaptive evolution (2224) may, in fact, represent compensation for genetic alterations rather than adaptive response to changing environments (25). This notion is corroborated by the finding that pathogenic missense mutations in humans and D. melanogaster are frequently fixed in other organisms (compensated pathogenic deviations) (21, 26). Third, cis-trans hybrid incompatibilities may be used to identify transcription factors that interact with genes of interest (27).

Simulation Methods

Target Gene Expression.

The expression level of a target gene is determined by the probability that an RNA polymerase (RNAP) complex binds to its promoter through interaction with a bound transcriptional activator (28). Despite several factors that would affect this probability, we restrict our attention, for the sake of simplicity, to evolutionary changes in the concentration of an activator ([T]) and its dissociation constant to the cis-binding element (K), keeping all other variables constant. The concentration [T] has a logistic form

equation image

where [T]max is the maximum possible level of concentration achieved when f = ∞, τ designates the concentration achieved when f = 0, and f represents an evolving variable that is the direct target of mutational changes (see below). The dissociation constant K is determined by a difference between the two numbers, each assigned to the activator and cis-regulatory element (a and c, respectively) (13):

equation image

where κ designates the smallest dissociation constant achieved when a = c. We assume that the binding affinity of a transcriptional activator to its cis-regulatory element can be quantified by the inverse of the dissociation constant (1/K) (Fig. 1). In the simulations, we set [T]max = κ and τ = κ (2/15). Hence, assuming that the binding probability of the activator to the cis-regulatory element is given by An external file that holds a picture, illustration, etc.
Object name is pnas.1105814108i1.jpg (29), the upper bound of this probability would be 0.5 (achieved when f = ∞ and a = c), whereas the lower bound would be 0 (achieved when f = −∞ or ac = ± ∞); it becomes 2/17 at a fitness optimum given by f = a = c = 0.

In each strain, the expression level is kept near the optimum under so-called Gaussian stabilizing selection, meaning that the fitness of an individual with an expression level Z is given by w(Z) = exp[−s (Z − θ)2]. Here, θ denotes the optimum level of gene expression, whereas s is a constant measuring the strength of selection; larger (smaller) s implies stronger (weaker) stabilizing selection on the target gene expression level.

In the present formulation, the RNAP promoter binding probability for each allele in the parental homozygous strain i (zP(i)) is given by

equation image

Here, [Pol] and KPol are the concentration and dissociation constant of the RNAP complex, respectively; ω is a cooperativity factor between the bound activator and the RNAP complex (28). Assuming a diploid autosomal locus and additivity of the two alleles, the expression level of an individual is obtained by Z = 2zP(i).

We also assume that the allele-specific concentration of the activator [Ti] is halved in hybrids. Hence, the corresponding probability for allele i in the hybrid between strains i and j (zH(i)) is given by

equation image

Here, Kik is a dissociation constant of the activator from strain k to the cis-binding element in strain i, where k [set membership] {i, j}. The expression level of a hybrid individual is again obtained by adding the probabilities for the two alleles: Z = zH(i) + zH(j). Throughout the simulations, we set [Pol]/KPol = 1/50 and ω = 20 (28).

Mutation.

Each mutation affects only one of the evolving properties f, a, and c. Mutations that increase or decrease f by an amount df occur at rate vf per generation. Mutation effect and rate for a are ±da and va, respectively, whereas the effect and rate for c are ±dc and vc, respectively. Because of the difference in the mutation target size, we assume vfva, vc.

Simulating Evolutionary Dynamics.

At the fitness optimum (Z = θ), any mutational change is necessarily deleterious, and its fixation entails a degradation of population fitness. Subsequent to this substitution event, the next mutation may bring back the population to the fitness peak by compensating for the deleterious effect of the first mutation; otherwise, population fitness may decay even more substantially by fixing mutations that shift the population farther away from the optimum. Previous work on compensatory evolution has found that, when selection against deleterious intermediates is weak and linkage between interacting loci is loose, evolution proceeds by sequentially accumulating independent substitutions in a stepwise manner (30). To simulate the process of evolution by successive fixation of mutant alleles in an otherwise monomorphic (homogeneous) background, the present analysis makes a simplifying assumption that only a single locus at most can be polymorphic at any time. The stochastic effect of random genetic drift is incorporated into the simulations by assuming a finite panmictic population of constant size n = 100,000. Then, the fixation probability of a new mutation at the polymorphic locus is computed based on its fitness effect conditional on other loci being monomorphic (31). If fixed, the mutation moves the population to a new state, from which the fixation probability of the next mutation is computed.

By repeating this substitution process, the evolutionary divergence in 10,000 pairs of orthologous genes is simulated based on the relative probabilities of possible substitution events that may potentially occur in the next step. The time interval between consecutive substitution events is also determined based on the total rate of substitution at each step. Starting from a fitness optimum given by f = a = c = 0, each round of simulations is initiated by a burn-in period of 107/U generations (where U [equivalent] vf + va + vc designates the total mutation rate), after which a single panmictic population starts to diverge by splitting into two isolated populations that evolve independently thereafter.

Regressing the Relative Allelic Expression in F1 Hybrids on the Expression Difference Between Parental Strains.

We follow ref. 4 and study the simulated linear regression of the (log2-transformed) relative allelic expression (log2[zH(i)/zH(j)]) in F1 hybrids on the (log2-transformed) relative expression difference (log2[zP(i)/zP(j)]) between parental strains i and j. We study two cases in particular. In the first case, a is a constant (va = 0), and only c mutates; consequently, ac compensation does not occur. In the second case, both a and c are allowed to mutate, and therefore ac compensation may occur together with K-[T] compensation.

Pleiotropy of an Activator with Two Target Genes.

We also consider a system with two genes (denoted G1 and G2) that are pleiotropically affected by a single activator. As in the above model with a single target, the expression of gene G1 is jointly regulated by the concentration of the transcriptional activator [T] and its dissociation constant to the cis-binding element of G1, which is given by κ exp[(ac1)2]. Similarly, the level of G2 expression is regulated by three variables, [T], a, and c2, in which the dissociation constant to the cis-binding element of G2 is κ exp[(ac2)2]. Together, these properties determine the expression levels of the two target genes (denoted Z1 and Z2), which, in turn, influence the fitness of an individual in a multiplicative manner: w(Z1, Z2) = exp[−s1 (Z1 − θ1)2s2 (Z2 − θ2)2]. Here, si quantifies the strength of stabilizing selection on the level of Gi expression as the deviation of Zi from its optimum θi (i [set membership] {1, 2}). Note that, by setting s2 = 0, we recover the model without pleiotropy, with the fitness function given by w(Z1) = exp[−s1 (Z1 − θ1)2]. When s1, s2 > 0, mutational changes in [T] or a may have detrimental pleiotropic effects, because mutations that minimize |Z1 − θ1| may inflate the deviation of Z2 from θ2. Although only two target genes are explicitly modeled in the present analysis, we expect that, as long as we focus on the regulation of G1 expression, situations with a larger number of simultaneously regulated targets can be studied by assigning a larger value of s2 in the simulations.

Materials and Methods for Expression Assays

Introgression Lines.

Using D. simulans white line (w501; stock number 14021–0251.011) and D. mauritiana P{lacW} insertion line V1 (P{lacW}42B; stock number 14021–0241.65) (32), each obtained from the University of California San Diego Drosophila Species Stock Center, we constructed reciprocal introgression lines, one carrying a small D. mauritiana segment surrounding the P{lacW} insert in an otherwise D. simulans genetic background and the other containing a D. simulans segment around the same insert in a D. mauritiana background. Because the P{lacW} acts in a semidominant fashion, heterozygous and homozygous flies can be readily discriminated. Experiments were initiated by crossing w501 females to V1 males. Then, hybrid females heterozygous for the P{lacW} insert were repeatedly backcrossed for 15 generations to w501 males to obtain a V1 introgression line and also to V1 males to obtain a w501 introgression line. Three replicate sublines were constructed for each introgression. Because of a low recombination rate, the introgressed segments encompass the centromere and commonly span a cytological region 36E3-37B8;42C7-43A4, which contains 1,564–1,777 genes.

Similarly, using D. sechellia white line (33) and D. mauritiana insertion line DEE1 (P{lacW}57A; stock number 14021–0241.115) (32), we constructed another independent set of introgression lines (DEE1 and sechellia introgressions) by 10 generations of backcrossing. In all sublines of these introgressions, introgressed segments cover a region 54E7-55C2;57B16-57E9 in which there are 535–720 genes.

All flies were kept in 12/12-h light/dark cycle at 22 °C.

Crosses, RNA and DNA Extraction, and cDNA Synthesis.

For each subline of V1 and w501 introgressions, we set up three replicate vials with 10 heterozygous introgression males and 10 pure-species females (w501 females for the V1 introgression males and V1 females for w501 introgression males). Flies were then transferred to fresh vials every 2 d (or 3 d for one case). Ten heterozygous males eclosed for 2 h (Zeitgeber time; ZT23 to ZT1) were homogenized at ZT9 on the same day. RNA was extracted from two-thirds of the whole fly lysate (RNeasy Mini kit; Qiagen), whereas DNA was extracted from the remaining one-third of the lysate (GenElute Mammalian Genomic DNA Miniprep Kit; Sigma). Total RNA was treated with DNase I two times, one time during and one time after the extraction (RNase-Free DNase set; Qiagen and DNase I, Amplification Grade; Invitrogen). PCR experiments ensured that there was practically no DNA contamination. oligo(dT)-primed cDNA was synthesized from 1 μg total RNA (SuperScript III First-Strand Synthesis System for RT-PCR; Invitrogen). The experiments included three or four independent RNA samples extracted on different days per subline, yielding a total of 10 replicates per introgression line. Because these expression assays were conducted for RNA from the whole body, they may not be sensitive enough to detect tissue-specific expression differences.

For sublines of DEE1 and sechellia introgressions, crosses were also made as described above. We collected a group of 20 virgin heterozygous males from each subline and dissected all their legs at ZT0 (19–25 h after the eclosion). We then extracted total RNA from these 120 legs (RNeasy Micro kit; Qiagen), whereas DNA was extracted from remaining body parts. One or two groups of 20 flies were dissected in a single day. We constructed four or five replicates per subline, each on a different day, to yield a total of 13 replicates per introgression line. A one-half aliquot was used for cDNA synthesis.

Measurements and Statistics of Allele-Specific Expression.

We used pyrosequencing to measure relative expression of heterospecific alleles in virgin males heterozygous for the introgression. For the assays of V1 and w501 introgressions, we chose six genes [CG15219, vulcan (vlc), Actin 42A (Act42A), CG7856, Trap1, and phtf] present on both introgressions. These genes were suggested to be underexpressed in D. mauritianaD. simulans hybrids, but their statistical significance was only marginal (34). Moreover, only one (CG15219) of six genes was listed as underexpressed in another genome-wide survey (35). Hence, whether they are really underexpressed in hybrids remains a question. We chose Obp57e gene for an independent expression assay using DEE1 and sechellia introgressions. This gene is differentially expressed in D. sechellia and D. simulans (17), but no data are available on its expression in D. mauritiana or hybrids. Because of this potential bias in selecting the genes, we may have overestimated the background dependency of the expression ratio of heterospecific alleles.

We determined DNA sequences of the chosen genes for each of the two parental lines. Based on the sequence data, we designed PCR and pyrosequencing primers using Primer3 software (36) and Primo SNP 3.4 (http://www.changbioscience.com/primo/primosnp.html), which is summarized in Table S1.

For each gene, all samples were measured on one plate. To correct for possible experimental bias and error, an allelic ratio in cDNA was divided by the ratio in genomic DNA from the same biological sample and then log2-transformed before statistical analyses. Because the analysis of variance detected no subline effect in any tests, we tested the significance of differences between the two backgrounds, without distinguishing sublines, using the t test (or Welch's approximate t test in one case where the variances differed significantly between the V1 and w501 introgressions) (37). Three genes showed statistically significant differences in allelic expression ratio between reciprocal introgressions at a false discovery rate of less than 5% (vlc: t = −4.21, df = 17, P < 0.001, two-tailed; phtf: t = −2.75, df = 17, P < 0.014, two-tailed; Obp57e: t = −2.61, df = 19, P < 0.018, two-tailed). For these three genes, second assays were carried out to confirm the results by using independently synthesized cDNA for vlc and phtf and additional cDNA samples for Obp57e. Because of the inequalities in sample size and also the error variance between the two assays, we performed the t tests separately on the second assays, combined P values from these two assays, and then evaluated the differences between the two backgrounds using the χ2 test (37).

Supplementary Material

Supporting Information:

Acknowledgments

We thank K. Fujimoto for suggestions, K. Fujikawa and S. Tamura for experimental assistance, A. Takahashi for technical advice, and N. Saitou for provision of equipment. K.R.T. was supported by an NIG Research Fellowship. This study was funded by NIG Cooperative Research Programs 2010-A72 (to K.R.T.), 2011-A16 (to K.R.T.), 2010-B5 (to T.M.), and 2011-B4 (to T.M.) and Grants-in-Aid for Scientific Research 20570100 (to T.T.-S.-K.) and 23570123 (to T.T.-S.-K.).

Footnotes

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. G.G. is a guest editor invited by the Editorial Board.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1105814108/-/DCSupplemental.

References

1. Cowles CR, Hirschhorn JN, Altshuler D, Lander ES. Detection of regulatory variation in mouse genes. Nat Genet. 2002;32:432–437. [PubMed]
2. Yan H, Yuan W, Velculescu VE, Vogelstein B, Kinzler KW. Allelic variation in human gene expression. Science. 2002;297:1143. [PubMed]
3. Yvert G, et al. Trans-acting regulatory variation in Saccharomyces cerevisiae and the role of transcription factors. Nat Genet. 2003;35:57–64. [PubMed]
4. Wittkopp PJ, Haerum BK, Clark AG. Evolutionary changes in cis and trans gene regulation. Nature. 2004;430:85–88. [PubMed]
5. Wittkopp PJ, Haerum BK, Clark AG. Regulatory changes underlying expression differences within and between Drosophila species. Nat Genet. 2008;40:346–350. [PubMed]
6. Landry CR, et al. Compensatory cis-trans evolution and the dysregulation of gene expression in interspecific hybrids of Drosophila. Genetics. 2005;171:1813–1822. [PMC free article] [PubMed]
7. Wittkopp PJ, Haerum BK, Clark AG. Independent effects of cis- and trans-regulatory variation on gene expression in Drosophila melanogaster. Genetics. 2008;178:1831–1835. [PMC free article] [PubMed]
8. Wang H-Y, et al. Complex genetic interactions underlying expression differences between Drosophila races: Analysis of chromosome substitutions. Proc Natl Acad Sci USA. 2008;105:6362–6367. [PMC free article] [PubMed]
9. Zhang X, Borevitz JO. Global analysis of allele-specific expression in Arabidopsis thaliana. Genetics. 2009;182:943–954. [PMC free article] [PubMed]
10. Wagner GP, Lynch VJ. The gene regulatory logic of transcription factor evolution. Trends Ecol Evol. 2008;23:377–385. [PubMed]
11. Dover GA, Flavell RB. Molecular coevolution: DNA divergence and the maintenance of function. Cell. 1984;38:622–623. [PubMed]
12. Shaw PJ, Wratten NS, McGregor AP, Dover GA. Coevolution in bicoid-dependent promoters and the inception of regulatory incompatibilities among species of higher Diptera. Evol Dev. 2002;4:265–277. [PubMed]
13. Johnson NA, Porter AH. Rapid speciation via parallel, directional selection on regulatory genetic pathways. J Theor Biol. 2000;205:527–542. [PubMed]
14. Crocker J, Tamori Y, Erives A. Evolution acts on enhancer organization to fine-tune gradient threshold readouts. PLoS Biol. 2008;6:e263. [PMC free article] [PubMed]
15. Cutter AD. Divergence times in Caenorhabditis and Drosophila inferred from direct estimates of the neutral mutation rate. Mol Biol Evol. 2008;25:778–786. [PubMed]
16. Kuo D, et al. Coevolution within a transcriptional network by compensatory trans and cis mutations. Genome Res. 2010;20:1672–1678. [PMC free article] [PubMed]
17. Matsuo T, Sugaya S, Yasukawa J, Aigaki T, Fuyama Y. Odorant-binding proteins OBP57d and OBP57e affect taste perception and host-plant preference in Drosophila sechellia. PLoS Biol. 2007;5:e118. [PMC free article] [PubMed]
18. Lu X, et al. Genome-wide misexpression of X-linked versus autosomal genes associated with hybrid male sterility. Genome Res. 2010;20:1097–1102. [PMC free article] [PubMed]
19. McManus CJ, et al. Regulatory divergence in Drosophila revealed by mRNA-seq. Genome Res. 2010;20:816–825. [PMC free article] [PubMed]
20. Main BJ, et al. Allele-specific expression assays using Solexa. BMC Genomics. 2009;10:422. [PMC free article] [PubMed]
21. Kondrashov AS, Sunyaev S, Kondrashov FA. Dobzhansky-Muller incompatibilities in protein evolution. Proc Natl Acad Sci USA. 2002;99:14878–14883. [PMC free article] [PubMed]
22. Smith NGC, Eyre-Walker A. Adaptive protein evolution in Drosophila. Nature. 2002;415:1022–1024. [PubMed]
23. Fay JC, Wyckoff GJ, Wu C-I. Testing the neutral theory of molecular evolution with genomic data from Drosophila. Nature. 2002;415:1024–1026. [PubMed]
24. Andolfatto P. Adaptive evolution of non-coding DNA in Drosophila. Nature. 2005;437:1149–1152. [PubMed]
25. Hughes AL. Near neutrality: Leading edge of the neutral theory of molecular evolution. Ann N Y Acad Sci. 2008;1133:162–179. [PMC free article] [PubMed]
26. Kulathinal RJ, Bettencourt BR, Hartl DL. Compensated deleterious mutations in insect genomes. Science. 2004;306:1553–1554. [PubMed]
27. Michalak P, Noor MAF. Association of misexpression with sterility in hybrids of Drosophila simulans and D. mauritiana. J Mol Evol. 2004;59:277–282. [PubMed]
28. Buchler NE, Gerland U, Hwa T. On schemes of combinatorial transcription logic. Proc Natl Acad Sci USA. 2003;100:5136–5141. [PMC free article] [PubMed]
29. von Hippel PH, Berg OG. On the specificity of DNA-protein interactions. Proc Natl Acad Sci USA. 1986;83:1608–1612. [PMC free article] [PubMed]
30. Innan H, Stephan W. Selection intensity against deleterious mutations in RNA secondary structures and rate of compensatory nucleotide substitutions. Genetics. 2001;159:389–399. [PMC free article] [PubMed]
31. Kimura M. On the probability of fixation of mutant genes in a population. Genetics. 1962;47:713–719. [PMC free article] [PubMed]
32. True JR, Weir BS, Laurie CC. A genome-wide survey of hybrid incompatibility factors by the introgression of marked segments of Drosophila mauritiana chromosomes into Drosophila simulans. Genetics. 1996;142:819–837. [PMC free article] [PubMed]
33. Higa I, Fuyama Y. A new mutant of D. sechellia. Drosophila Information Service. 1993;72:198.
34. Moehring AJ, Teeter KC, Noor MAF. Genome-wide patterns of expression in Drosophila pure species and hybrid males. II. Examination of multiple-species hybridizations, platforms, and life cycle stages. Mol Biol Evol. 2007;24:137–145. [PubMed]
35. Michalak P, Noor MAF. Genome-wide patterns of expression in Drosophila pure species and hybrid males. Mol Biol Evol. 2003;20:1070–1076. [PubMed]
36. Rozen S, Skaletsky HJ. In: Bioinformatics Methods and Protocols. Misener S, Krawetz SA, editors. Totowa, NJ: Humana Press; 2000. pp. 365–386.
37. Sokal RR, Rohlf FJ. Biometry: The Principles and Practice of Statistics in Biological Research. New York: Freeman; 1995.

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...