Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2015 Sep 25.
Published in final edited form as: Cell. 2014 Sep 25;159(1):58–68. doi: 10.1016/j.cell.2014.09.003

Evolution of DNA specificity in a transcription factor family produced a new gene regulatory module

Alesia N McKeown 1,*, Jamie T Bridgham 1,*, Dave W Anderson 1, Michael N Murphy 3, Eric A Ortlund 3, Joseph W Thornton 2
PMCID: PMC4447315  NIHMSID: NIHMS629207  PMID: 25259920

SUMMARY

Complex gene regulatory networks require transcription factors (TFs) to bind distinct DNA sequences. To understand how novel TF specificity evolves, we combined phylogenetic, biochemical, and biophysical approaches to interrogate how DNA recognition diversified in the steroid hormone receptor (SR) family. After duplication of the ancestral SR, three mutations in one copy radically weakened binding to the ancestral estrogen response element (ERE) and improved binding to a new set of DNA sequences (steroid response elements, SREs). They did so by establishing unfavorable interactions with ERE and abolishing unfavorable interactions with SRE; also required were numerous permissive substitutions, which nonspecifically improved cooperativity and affinity of DNA binding. Our findings indicate that negative determinants of binding play key roles in TFs’ DNA selectivity and—with our prior work on the evolution of SR ligand specificity during the same interval—show how a specific new gene regulatory module evolved.

Transcription factor specificity and the evolution of gene regulatory networks

Development, homeostasis, and other complex biological functions depend upon the coordinated expression of networks of genes. Thousands of transcription factors (TFs) in eukaryotes play key regulatory roles in these networks, because their distinct affinities for DNA binding sites, for other proteins, and for small molecules allow them to specifically regulate the expression of unique sets of target genes in response to various hormones, kinases, and other upstream molecular stimuli. Most studies of the evolution of gene regulation have focused on how changes in cis-regulatory DNA can bring a new target gene under the influence of an existing TF (Carroll, 2008; Wray, 2007) or on changes in protein-protein interactions among TFs (Brayer et al., 2011; Lynch et al., 2011; Baker et al., 2012). Although TF specificity for DNA can and does evolve (Baker et al., 2011; Sayou et al., 2014), little is known concerning the molecular mechanisms and evolutionary dynamics by which such changes occur. In turn, it remains unclear how distinct gene regulatory modules – defined as a transcription factor, the molecular stimuli that regulate it, and the DNA target sequences it recognizes – emerge during evolution. If TFs are constrained by selection to conserve essential ancestral functions (Stern and Orgogozo, 2009), how can new regulatory modules ever arise? Do specific modules evolve by partitioning the activities of an ancestral TF that is promiscuous in its interactions with DNA targets and molecular stimuli (Sayou et al., 2014), or by acquiring entirely new interactions (Teichmann and Babu, 2004)? What is the genetic architecture of evolutionary transitions in TF specificity, and what kinds of biophysical mechanisms mediate these changes? Answering these questions requires dissecting evolutionary transitions in TFs’ capacity to interact specifically with DNA and molecular stimuli. Ancestral protein reconstruction, combined with detailed studies of protein function and biochemistry, has the potential to accomplish this goal (Harms and Thornton, 2010).

The knowledge gap concerning transcription factor evolution mirrors uncertainty about the physical mechanisms that determine TFs’ specificity for their DNA targets. DNA recognition is usually thought to be determined by favorable interactions—especially hydrogen bonds but also van der Waals interactions—between a protein and its preferred DNA sequences (Garvie and Wolberger, 2001; Rohs et al., 2010). Supporting this view, structural studies have established that positive interactions are typically present in high-affinity complexes of protein and DNA. Specificity, however, is determined by the distribution of affinities across DNA sequences, and it is unclear whether positive interactions sufficiently explain TFs’ capacity to discriminate among targets. In principle, negative interactions that reduce affinity to non-target binding sites—such as steric clashes or the presence of unpaired polar atoms in a protein-DNA complex—could also contribute to specificity (von Hippel and Berg, 1986). Evaluating the role of negative interactions in determining specificity, however, requires analyzing not only high-affinity TF/DNA complexes but also poorly bound ones, which are vast in number and difficult to crystallize. We reasoned that by focusing on a major evolutionary transition in DNA specificity during the history of a family of related TFs, we could gain direct insight into the genetic and biophysical factors that cause differences in DNA recognition (Harms and Thornton, 2013).

Steroid receptors coordinate distinct gene regulatory modules

Steroid hormone receptors (SRs), a family of ligand-activated transcription factors, are a model for the evolution of TF specificity. SRs initiate the cascade of classic transcriptional responses to sex and adrenal steroid hormones in vertebrate physiology, reproduction, development, and behavior (Bentley, 1998). These proteins contain a conserved DNA-binding domain (DBD), which directly binds to DNA sequences in the vicinity of the target genes they regulate; they also contain a conserved ligand-binding domain (LBD), which binds hormonal ligands and then attracts coregulatory proteins, leading to ligand-regulated changes in gene expression (Bain et al., 2007; Beato and Sanchez-Pacheco, 1996; Kumar and Chambon, 1988). Additional poorly conserved N-terminal and hinge domains mediate other SR activities. All SRs bind as dimers to inverted palindromic DNA sequences consisting of two six-nucleotide half-sites separated by a variable three-nucleotide spacer (Fig. 1A, (Welboren et al., 2009; So et al., 2007; Lundback et al., 1993; Umesono and Evans, 1989; Beato et al., 1989)).

Fig. 1. Evolution of novel specificity occurred via a discrete shift between AncSR1 and AncSR2.

Fig. 1

Open in a new tab

(A) Architecture of SR response elements. All SRs bind to an inverted palindrome of two half-sites (gray arrows) separated by variable bases (n). x, sites at which ERE and SREs differ. (B) SR phylogeny comprises two major clades, which have non-overlapping specificity for ligands (stars) and REs (boxes). Preferred half-sites for each clade are shown; bases that differ are underlined. Ancestral and extant receptors are colored by RE specificity (purple, ERE; green, SREs; blue, extended monomeric ERE). Orange box, evolution of specificity for SREs; number of substitutions on this branch and the total number of DBD residues are indicated. Nodal support is marked by the approximate likelihood ratio statistic: unlabeled, aLRS 1 to 10; •, aLRS 10 to 100; ••, aLRS>100. Scale bar is in substitutions per site. (C) AncSR1 specifically activates reporter gene expression driven by ERE (purple bar), with no activation from SRE1 (light green) or SRE2 (dark green); AncSR2’s specificity is distinct. Bar height indicates fold-activation relative to vector-only control. (D) Ancestral binding affinities reflect distinct specificities for ERE vs. SREs. Bars heights indicate the macroscopic affinity (KA,mac) of binding to palindromic DNA response elements, measured using fluorescence polarization. Colors as in panel C. (E-G) The components of macroscopic binding affinity—affinity for a half-site (K1) and cooperativity of binding (ω)—by AncSR1 and AncSR2, were estimated by measuring KA,mac on a full palindromic RE and K1 on a half-site, then globally fitting the data to a model containing both parameters. Error bars show SEM of three experimental replicates. See Fig. S1; Tables S1-S3.

There are two phylogenetic classes of SRs in vertebrates, which have distinct specificities for both DNA and hormonal ligands: the two SR classes therefore mediate distinct regulatory modules (Fig. 1B). One class, the estrogen receptors (ERs), are activated by steroid hormones with aromatized A-rings (Eick et al., 2012) and bind preferentially to estrogen response elements (ERE, a palindrome of AGGTCA) (Welboren et al., 2009). The other class contains the receptors for the non-aromatized steroid hormones, including androgens, progestagens, glucocorticoids, and mineralocorticoids (AR, PR, GR, and MR; (Eick et al., 2012); this class of SR preferentially binds to steroid response elements (SREs), including palindromes of AGAACA (SRE1) or AGGACA (SRE2) (So et al., 2007; Chusacultanachai et al., 1999). The two classes’ DNA specificities are distinct: ERs bind poorly to and do not activate SREs, whereas members of the AR/PR/GR/MR group bind poorly to and do not activate ERE (Zilliacus et al., 1992). Although SRs can and do bind variants of these classic sequences (Welboren et al., 2009; So et al., 2007), the classical ERE and SRE sequences are physiologically relevant and have been the subject of extensive biochemical and structural analysis (Beato et al., 1989; Luisi et al., 1991; Zilliacus et al., 1992; Lundback et al., 1993; Schwabe et al., 1993).

Understanding the evolution of a TF-mediated regulatory module requires understanding the origin of the TF’s interactions with both upstream stimuli and DNA targets. We recently reported on the mechanisms by which the two classes of SRs evolved their distinct specificities for aromatized or nonaromatized hormones (Eick et al., 2012; Harms et al., 2013). Here we use ancestral protein reconstruction (Harms and Thornton, 2013; Thornton, 2004; Harms and Thornton, 2010) to identify the genetic, biochemical, and biophysical mechanisms for the evolution of the distinct DNA specificity in the two classes of SRs. The results, together with previous findings on the evolution of SR ligand specificities, allow us to provide a detailed historical and mechanistic account for the evolution of a new regulatory module.

RESULTS

A discrete evolutionary transition in DNA specificity

To characterize the evolutionary trajectory of DNA recognition in the SRs, we first used ancestral protein reconstruction to infer the DBDs of the ancestral protein from which all SRs descend (AncSR1) and of the ancestor of all ARs, PRs, GRs, and MRs (AncSR2, Fig. 1B). Both proteins predate the evolutionary emergence of vertebrates, more than 450 million years ago (Eick et al., 2012). We used maximum likelihood phylogenetics to infer the best-fit evolutionary model and phylogenetic tree for 213 SRs and related nuclear receptors from a wide variety of animal taxa using sequences of both the DBD and LBD (Fig. S1). We then inferred the maximum likelihood amino acid sequences of the DBD and the posterior probability distribution of amino acids at each sequence sites at the phylogenetic nodes corresponding to AncSR1 and AncSR2 (Fig. S1A-B). The vast majority of sites in the two sequences were reconstructed with little or no uncertainty; only 3 sites in AncSR2 and 12 in AncSR1 were reconstructed ambiguously, defined as having an alternate state with posterior probability >0.20 (Table S1).

The distinct specificities of extant SRs could have evolved by partitioning the activities of a promiscuous ancestor among descendants or by a discrete switch from ancestral to derived forms of specificity. To distinguish among these possibilities, we synthesized coding sequences for the inferred ancestral DBDs and characterized their functions and physical properties. We focused on the capacity to bind ERE, SRE1, and SRE2, because these classical REs differ only at two bases in the half-site and are completely distinct in their responses to the two classes of SR (Zilliacus et al., 1992). Using a dual luciferase reporter assay in cultured cells (Fig. 1C), we found that AncSR1 had DNA specificity like that of extant ERs, driving strong activation from ERE but exhibiting no expression above background from SREs. AncSR2, in contrast, specifically activated from both SREs but did not activate from ERE. These results are consistent with the strong sequence similarity between AncSR1 and extant ERs and between AncSR2 and the vertebrate ARs, PRs, GRs, and MRs (Fig. 1B) and are further corroborated by the pattern of RE specificities across extant members of the SR family tree: because all known descendants of AncSR2 recognize SREs and all other family members and close outgroups bind ERE-like sequences, the most parsimonious expectation by far is SRE-specificity by AncSR2 and ERE-specificity by AncSR1 (Eick and Thornton, 2011), the most parsimonious expectation for AncSR1 is ERE-specificity.

Robustness to uncertainty

To determine whether the inferred functions of AncSR1 and AncSR2 are robust to uncertainty about the ancestral sequences, we synthesized reconstructions of each ancestor that contain every plausible alternate residue. These sequences represent the far edge of the “cloud” of plausible estimates of the true ancestral sequence and are different from the ML sequences at more residues than the expected number of errors in each ML reconstruction (Table S1). These alternative reconstructions therefore provide a conservative test of the robustness of inferences about the ancestral proteins’ functions.

We synthesized and assayed these alternate reconstructions and found that the DNA specificities of the alternate reconstructions were nearly identical to those of the ML ancestors (Fig. S2A). Moreover, the sequences of extant SRs indicate that none of the plausible alternative residues in AncSR1 or AncSR2 are sufficient to change DNA specificity (Table S2).

Taken together, these data indicate that the ancestral SR was ERE-specific, and recognition of SREs emerged via a discrete change in specificity during the interval between AncSR1 and AncSR2 (Fig. 1B). This transition involved a complete loss of activation from the ancestrally preferred ERE and a wholesale gain of novel activation on SREs.

Thermodynamic basis for evolution of new DNA specificity

We next sought to understand the biochemical basis for this ancient change in DNA recognition by expressing and purifying ancestral proteins and characterizing their thermodynamics of binding to DNA. We used fluorescence polarization to determine the macroscropic binding affinity (KA,mac) of each ancestral DBD for labeled DNA probes containing palindromic ERE or SREs. The relative affinities followed those in the activation assays, with AncSR1 showing strongly preferential binding to ERE and AncSR2 preferentially binding SREs (Fig. 1D, Table S3). Both bound much more weakly to their non-target REs, with affinity apparently too low to activate reporter transcription. These data indicate that the evolutionary transition in the DBD’s DNA specificity was due primarily to changes in DNA-binding affinity for the two classes of binding sites (see (Bain et al., 2012).

The macroscopic affinity of an SR dimer for a palindromic DNA sequence is determined by two components: the half-site binding affinity (K1) of each monomer for its half-site and the binding cooperativity (ω) between half-sites, defined as the fold excess of the macroscopic affinity beyond that expected if each monomer binds independently (Fig. 1E, (Hard et al., 1990). To estimate these parameters, we performed fluorescence polarization binding experiments with both half-site and palindromic DNA constructs and globally fit the parameters of a two-monomer cooperative binding model to these data.

We found that AncSR1 binds ERE with high half-site affinity and low cooperativity. In contrast, AncSR2 displays much lower half-site affinity but greater cooperativity (Fig. 1F-G, Table S3). AncSR2’s novel RE specificity therefore evolved through a trade-off in the energetic mechanisms of binding: the protein’s direct interactions with DNA became weaker as its specificity changed, but this effect was offset by an increase in cooperativity of binding. As a result, the derived DBD retained macroscopic DNA binding affinity for its favored targets similar to that of its ancestor, but for a new family of DNA sequences. These ancient changes in binding energetics persist to the present: human ERs, like AncSR1, bind DNA with high half-site affinity and low cooperativity, whereas human GR, like AncSR2, displays considerable cooperativity but lower half-site affinity (Alroy and Freedman, 1992; Hard et al., 1990).

Atomic structures of ancestral DBDs

To identify the causes of these evolutionary changes in DNA binding and recognition, we determined the crystal structures of AncSR1-DBD bound to ERE and of AncSR2-DBD bound to SRE1 at 1.5 and 2.7 Å, respectively (Fig. 2, Table S4). Although their sequences are only 54% identical, AncSR1 and AncSR2 have very similar conformations (RMSD for protein backbone atoms = 0.82 Å). Each monomer buries a recognition helix (RH) in the DNA major groove of one half-site and makes additional contacts to the DNA backbone; the monomers contact each other via a dimerization surface composed of an extended loop coordinated by a zinc atom (Schwabe and Rhodes, 1991; Schwabe et al., 1993; Luisi et al., 1991).

Fig. 2. Structures of ancestral proteins give insight into the molecular determinants of specificity.

Fig. 2

Open in a new tab

(A) X-ray crystal structures of AncSR1 bound to ERE (left); AncSR2 bound to SRE1 (right). Cartoon shows protein dimers; surface shows DNA. Black arrow, beginning of unresolved C-terminal tail. Dotted line, unresolved AncSR1 loop near dimerization interface. Cyan spheres, sites of permissive substitutions. Grey spheres, zinc atoms. (B) Enlarged view of recognition helix in the DNA major groove (black box in A). Sticks, side chains of RH residues making polar contacts with DNA. Dotted lines, hydrogen bonds and salt bridges from protein to DNA. (C) Buried solvent-inaccessible surfaces in Å2 at the protein-DNA and protein-protein interfaces in the crystal structures for each protein chain. Parentheses, calculations when residues unresolved in the AncSR1 crystal structure are excluded. See Table S4.

Despite these general similarities, there are several differences between the AncSR1 and AncSR2 structures. First, AncSR1’s RH makes more hydrogen bonds to DNA than AncSR2 does (Fig. 2B). Second, the loop that connects the RH to the dimerization surface is disordered in AncSR1 but adopts a resolved structure in AncSR2. Third, AncSR1 buries ~60% more of its surface area at the DNA interface than AncSR2 does, but AncSR2 buries ~40% more surface in its dimerization interface than AncSR1 (Fig. 2C). These differences are consistent with AncSR1’s greater affinity for DNA half-sites and AncSR2’s greater cooperativity of dimeric binding.

Recognition helix substitutions are necessary but not sufficient for evolution of the derived function

We next sought to identify the evolutionary genetic changes that caused specificity to change between AncSR1 and AncSR2. We focused first on the recognition helix, because it makes the only direct contacts to bases in the DNA half-site. There are ten residues in the RH, but only three changed between AncSR1 and AncSR2—e25G, g26S, and a29V (Fig. 3A, with lower and upper cases denoting ancestral and derived states, respectively). All three residues are strictly conserved in the AncSR1-like state in all ERs and the AncSR2-like state in all AR, PR, GR, and MRs (Fig. S3A). This region is also known to play an important role in the specificity of extant SRs (Alroy and Freedman, 1992; Zilliacus et al., 1992).

Fig. 3. Genetic basis for evolution of new DNA specificity.

Fig. 3

Open in a new tab

(A) AncSR1 and AncSR2 sequences. Substitutions between AncSR1 and AncSR2 are shown. Dots, conserved sites. ˆ, recognition helix (RH) and *, permissive substitutions. Grey box, RH. (B) Effect of RH and 11 permissive (11P) substitutions in luciferase reporter assays. Lower and upper case letters denote ancestral and derived states, respectively. Fold activation over vector-only control is shown, with SEM of three replicates. (C) RH substitutions shift half-site affinity among REs, and permissive substitutions non-specifically increase half-site affinity and cooperativity. The corners of the square represent genotypes of AncSR1, with or without RH and 11P substitutions. At each corner, circle color shows RE preference; numbers are the ratio of the KAmac for binding to SRE1 (upper) or SRE2 (lower) versus ERE. Along each edge, vertical bar graphs show the effect of RH or permissive substitutions on the energy of association for the dimeric complex (grey background); contributions of effects on half-site binding (beige) and cooperativity (cyan) are shown. Bar color shows effects on binding to ERE (purple), SRE1 and SRE2 (light and dark green, respectively). Graphs in the square’s center show the effect of 11P and RH combined. Mean ± SEM of three experimental replicates is shown. See Figs. S2-S4; Tables S3 and S5.

To test the hypothesis that these three substitutions were the main determinants of the evolutionary change in DNA specificity, we first reversed them to their ancestral state in AncSR2 (generating AncSR2+rh). As predicted, these changes are sufficient to restore the ancestral preference for ERE over SREs in a luciferase assay (Fig. 3B). They do so by restoring the DBD’s capacity to activate transcription from ERE while dramatically decreasing SRE activation.

We also determined the crystal structure of AncSR2+rh on ERE at 2.2 Å and found that reversing these three substitutions largely restores the ancestral protein-DNA interface (Fig. S2B-C). The interactions of AncSR2+rh with ERE-specific nucleotides are almost identical to those made by AncSR1. Only a few minor differences are apparent in non-specific interactions to the DNA backbone and to nucleotides outside of the half-sites, presumably because of differences in crystallization conditions or protein sequence outside the RH. Taken together, these data indicate that the RH substitutions were the primary determinants of the evolutionary change in half-site specificity from ERE to SREs.

To determine whether the RH substitutions were also sufficient causes of the shift in specificity, we introduced the derived RH states into AncSR1 (Fig. 3B). Surprisingly, activation was entirely abolished on all REs tested (Fig. 3B). This result is robust to uncertainty about the ancestral sequence: introducing the RH substitutions – which are inferred unambiguously – into the reconstruction of AncSR1 containing all plausible alternative amino acids caused the same effect (Fig. S2A). The lack of activity is not due to differences in protein expression between AncSR1 and AncSR1+RH (Fig. S2D), implying that the RH substitutions strongly compromise DBD function when introduced into AncSR1, rather than depleting protein in the cell. The derived RH states, however, are conserved in AncSR2 and all its descendants, all of which activate transcription. These data indicate that additional epistatic substitutions, which permitted the DBD to tolerate the RH substitutions must have also occurred during the AncSR1/AncSR2 interval.

Permissive substitutions outside the DNA interface were required for the evolution of new specificity

To identify these permissive substitutions, we divided the 35 other substitutions that occurred during the AncSR1/AncSR2 interval into 8 groups based on contiguity in the linear sequence and tertiary structure (Fig. S3A). We tested the hypotheses that each group contained permissive substitutions by reverting it to the ancestral state in AncSR2: reversing a permissive substitution in the context of the derived RH should compromise function. We found that just three groups, containing a total of 16 amino acid replacements, significantly reduced activation when reversed, indicating that the derived states at these sites are necessary for full DBD function and therefore contribute to the permissive effect (Fig. S3B, Table S5).

Using a series of forward and reverse genetic experiments testing the effects of the individual mutations within these groups, we ruled out a role for several substitutions and narrowed the set of permissive changes to 11 historical substitutions (11P) distributed among the three structural groups (Fig. S4A-C, Table S5). When the derived residues at these sites are introduced into the nonfunctional AncSR1+RH, they rescue activation and recapitulate the evolution of the derived DNA specificity (Fig. 3 A-B). Their permissive effect is robust to uncertainty about the precise sequence of AncSR1 (Fig. S2A). All three groups are necessary for the full permissive effect (Fig. S4D, Table S5).

These substitutions are permissive in that they are required for the protein to tolerate the derived RH, but when introduced into AncSR1 they have no effect on specificity; rather, they enhance activation non-specifically on ERE and SREs alike (Fig. 3B). Taken together, these data indicate that a large number of permissive mutations, which did not themselves affect specificity, were required for the specificity-switching substitutions to be tolerated.

The effect of these ancient permissive mutations persists to the present. We found that introducing the derived RH states from the human GR into human ERα results in a nonfunctional DBD, just as it did in AncSR1, consistent with the fact that the lineage leading to ERs branches from the rest of the SR phylogeny before AncSR2’s permissive mutations occurred (Fig. S2E). Adding the 11P into the nonfunctional ERα+RH protein, however, rescued activation and yielded a DBD with preference for SREs. Conversely, the ancestral RH states can be introduced into human GR, where they dramatically increase activation on ERE, just as they do in AncSR2 (Fig. S2E; Alroy and Freedman, 1992; Zilliacus et al., 1991). Taken together, these results indicate that the ancient RH and permissive substitutions provide a sufficient genetic explanation for the evolution of the distinct DNA specificities of the two major classes of extant SRs.

Evolution of specificity by negative protein-DNA interactions

Having identified the genetic changes that caused the evolution of AncSR2’s new specificity, we sought to understand the biophysical mechanisms by which they did so. We first measured the effect of the RH substitutions on the energetics of sequence-specific DNA binding. We found that they improve the DBD’s macroscopic binding preference for SREs by a factor of 30,000; this effect is caused by a 2,000-fold reduction in affinity for ERE and a 15-fold increase in SRE affinity (Fig. 3C, Table S3). These effects are entirely attributable to changes in half-site binding affinity, as the RH substitutions do not affect cooperativity (Fig. 3C).

To understand the atom-level mechanisms for the effects of the RH mutations, we compared crystal structures of the ancestral DBDs containing the ancestral or derived RH amino acids in complex with both ERE and SRE1; we also performed molecular dynamics (MD) simulations of AncSR1, AncSR1+RH, and AncSR2, each bound to ERE, SRE1 and SRE2. In principle, the evolutionary change in DNA specificity could have been caused by changes in positive interactions – hydrogen bonds or van der Waals attractions between protein and DNA atoms – or in negative interactions, such as electrostatic or steric clashes. If the change in specificity were solely due to changes in positive interactions, then the RH substitutions would reduce favorable interactions with ERE and increase favorable interactions with SREs.

Contrary to this prediction, we found that the RH substitutions primarily change negative interactions between the DBD and DNA binding sites, relieving clashes with SRE and establishing new ones with ERE. The ancestral RH does form more hydrogen bonds on ERE than on SREs, and the RH substitutions reduce the number of hydrogen bonds to ERE (Fig. 4A, S5E); these observations are consistent with the view that positive interactions are the primary determinants of specificity. By removing hydrogen bond acceptors, however, these substitutions also establish negative polar interactions, leaving polar groups on ERE-specific bases unpaired and leading to penetration of transient solvent molecules into the protein-DNA interface (Fig. S5A-D). The effect of these negative interactions is expected to be much stronger than the loss of the positive interactions: eliminating a protein-DNA hydrogen bond would reduce binding affinity only slightly, because the same number of total hydrogen bonds would form whether or not the protein and DNA are bound to each other or free in solvent. In contrast, leaving an unpaired polar atom at the protein-DNA interface results in more hydrogen bonds in the unbound than the bound state, leading to a much larger difference in energy between the bound and unbound states and a much more dramatic reduction in affinity (von Hippel and Berg, 1986).

Fig. 4. Recognition helix substitutions change DNA specificity by altering negative interactions.

Fig. 4

Open in a new tab

(A) In MD simulations, RH substitutions reduce hydrogen bonds to ERE but do not increase hydrogen bonds to SREs. Bars show mean number of direct hydrogen bonds from all 10 RH residues to DNA (Purple, ERE; light green, SRE1; dark green, SRE2), each sampled across three MD trajectories, with SEM. (B) RH substitutions reduce packing efficiency at the protein-DNA interface on ERE, but do not improve packing on SREs. Bars show the mean number of atoms in the 10 RH residues within 4.5 Å of a DNA atom. (C) Ancestral residue glu25 (sticks) shifts position due to steric clashes with T-4 and T-3 of SRE1. A representative sample frame from MD trajectories is shown for AncSR1 with ERE (purple) or SRE1 (green). DNA is shown as surface, with atoms in the variable bases -4 and -3 shown as lines; methyls of T-4 and T-3 are spheres. (D-F) Repositioning of glu25 by SREs causes Lys28 to shift, reducing hydrogen bonds to DNA. (D) The average position of these residues in MD trajectories of AncSR1 with various REs is shown when all atoms in the protein-DNA complex are aligned. Distance of lys28 from hydrogen bond acceptor G2 on ERE is shown in black. (E) Displacement of glu25 and lys28 of AncSR1 on SREs relative to their position on ERE. The mean positions of all atoms in each MD trajectory were calculated, the DNA atoms in these “mean structures” were aligned in pairs: bars shows the average distances from the atoms in complexes with SRE1 (dark green) or SRE2 (light green) to the corresponding atom in ERE were calculated. Purple bars, distances between pairs of atoms from independent ERE trajectories. Displacement toward the center of the palindrome was scored as positive, away as negative. Each bar shows the distance averaged across atoms in a residue and three pairs of trajectories with SEM. (F) Lys28 forms fewer hydrogen bonds to DNA on SREs than on ERE. Points show the mean number of hydrogen bonds formed by each RH residue to different REs, with SEM for three MD trajectories. (G,H) Effect of introducing e25G and other RH substitutions on half-site binding affinity (G) and transcriptional activation (H). See Figs. S6-S7, and Table S3. (I) Summary of mechanisms by which ancestral RH excludes SREs. Ancestral glu25 and conserved residue Lys28 form hydrogen bonds (black dotted lines) with ERE bases. These side chains would sterically clash with methyl groups of SRE1 and SRE2, so they are repositioned and are unable to form hydrogen bonds to DNA, leaving unpaired donors (blue) and acceptors (red) at the DNA-RH interface. The RH substitutions resolve the steric clash and remove the unfulfilled donor on e25, increasing SRE affinity. See Figs. S5-S6.

The improvement in SRE binding also cannot be explained by an increase in SRE-specific positive interactions. The RH substitutions do not increase the total number of hydrogen bonds on SRE1 and actually reduce the number of hydrogen bonds on SRE2 (Fig. 4A). They do so by eliminating or weakening hydrogen bonds formed by the ancestral protein to SREs without forming enough new hydrogen bonds to compensate. Although the derived RH does establish one novel hydrogen bond from derived residue Ser26 to the DNA backbone, this interaction actually forms more frequently on ERE than on SREs (Fig. S5E). Overall, AncSR1+RH (like AncSR2) forms equal numbers of hydrogen bonds with ERE and SREs, indicating that hydrogen bonding does not explain the evolution of preference for SREs. As for van der Waals interactions, the RH substitutions reduce the efficiency of packing on ERE, but they do not improve packing on SREs (Fig. 4B). Taken together, these results indicate that changes in positive interactions hydrogen bonds and van der Waals forces—do not explain AncSR2’s increase in affinity or its preference for SREs.

If new SRE-specific positive interactions do not explain the increase in affinity for SREs caused by the RH substitutions, what mechanisms do mediate this effect? We found that the RH substitutions improve SRE affinity by relieving SRE-specific steric and electrostatic clashes with the ancestral RH. Crystal structures and MD simulations both show that the long sidechain of glu25 sterically clashes with T-4 and T-3 of SREs; these bases contain large methyl groups that protrude into the DNA major groove of SREs, but are absent from the corresponding bases in ERE (Fig. 4C, Fig. S6A-E). As a result of this clash, glu25 is forced to move away from the major groove of SREs and, in turn, to displace the conserved residue Lys28, which in high-affinity complexes forms hydrogen bonds to DNA bases that do not vary among REs (Fig. 4D-E). As a result, Lys28 forms fewer hydrogen bonds on SREs compared to ERE (Fig. 4F). Additionally, by pushing the negatively charged glu25 away from the bases in the center of the major groove, the SRE-protein interface is left with numerous unpaired hydrogen bond donors and acceptors, leading to water penetration into the interface with SREs (Fig. S6F-H). The RH substitutions ameliorate this clash by replacing glu25 with the much smaller Gly, thus relieving the negative effect of the glu on SRE binding.

To test the hypothesis that removing glu25 improves SRE recognition by relieving negative interactions, we used site-directed mutagenesis to introduce e25G alone into AncSR1 containing the permissive mutations. We found, as predicted, that SRE affinity and activation were enhanced, despite the fact that Gly25 makes no apparent favorable interactions with SREs (Fig. 4G-H).

The other two RH substitutions preferentially reduce recognition of ERE, apparently by establishing additional ERE-specific negative interactions. When g26S and a29V are added to e25G, yielding the derived RH genotype, they reduce affinity and activation on all REs, but do so much more severely on ERE than SREs (Fig. 4G-H). The mechanism for this effect is not obvious in the structures or simulations (Fig. S6I-J), but it does not involve eliminating hydrogen bonds or van der Waals interactions with ERE: neither ancestral amino acid forms hydrogen bonds to ERE (Fig. 4F), and they do not pack more efficiently against ERE than the derived amino acids do (Fig. S6K).

Taken together, these data indicate that differences in sequence-specific positive interactions do not explain the switch in specificity caused by the RH substitutions. Rather, negative interactions that interfered with SRE binding in the ancestral state were lost, and new negative interactions that impair binding to ERE were gained (Fig. 4I). The result was to transform the DBD’s ancestral ERE-preference into AncSR2’s derived SRE-preference. A secondary effect was to reduce affinity for the preferred DNA sequence and thus to require permissive substitutions for activation to be maintained.

Permissive substitutions non-specifically improve affinity for both the derived and ancestral REs

Permissive substitutions are often thought to act by increasing thermodynamic stability, allowing the protein to tolerate mutations that confer new functions but compromise stability (Bershtein et al., 2006; Gong et al., 2013). Using reversible chemical denaturation, however, we found that the 11P substitutions do not increase stability, and the RH substitutions do not decrease stability (Fig. 5A-B).

Fig. 5. Permissive substitutions do not improve protein stability or dimerization in the absence of DNA.

Fig. 5

Open in a new tab

(A) Crystal structure of AncSR2 bound to SRE1. Sites of permissive substitutions are shown as Cα spheres; red, cyan, and orange indicate clustered groups of sites. Only one residue in the C-terminal group is shown). (B) Permissive substitutions (11P) do not increase protein stability. ΔGH2O, calculated Gibbs free energy of chemically induced unfolding; m, slope of the unfolding transition; CM, denaturant concentration at which 50% of protein is folded. (C,D) Permissive substitutions do not increase protein dimerization in the absence of DNA, measured by analytical ultracentrifugation. Distribution (C) and best-fit values (D) of sedimentation velocity coefficients (S20,w) for AncSR1 (left) or AncSR1+11P (right) at 0.5 mM. The fraction of the total signal under the dominant peak (% total), the estimated molecular weight of that peak (MW) and the expected molecular weight of the monomeric protein (MWtheo) show that AncSR1 and AncSR2 are both predominantly monomeric. RMSD, root mean square deviation of the data from the model; f/f0, total shape asymmetry. Signal at higher MW peaks may reflect aggregation due to high protein concentration.

Because the RH substitutions radically reduce affinity for ERE and only weakly increase affinity for SREs – yielding a low-affinity receptor for both kinds of element – we hypothesized that the permissive substitutions might offset these effects by increasing affinity in a non-sequence specific manner. As predicted, introducing 11P into the ancestral background increases macroscopic binding affinity by increasing both cooperativity and half-site affinity on all REs (Fig. 3C), indicating a tradeoff in the energetics of binding between the permissive and specificity-switching substitutions during evolution.

The crystal structures suggest that the permissive substitutions cause these effects by enhancing nonspecific protein-protein interactions at the dimerization interface and non-specific interactions with the DNA backbone and minor groove. Two of the permissive substitutions (v39H and v42L) may facilitate dimer formation, because they are located on the loop that links the RH to the dimerization surface (Fig. 5A). In AncSR1, as in human ERα the loop is unresolved, but it is fully resolved in complexes containing the derived state at these residues, including AncSR2, AncSR2+rh, and the human GR (Luisi et al., 1991). Using analytical ultracentrifugation, we found that the permissive substitutions do not measurably increase DBD dimerization in solution (Fig. 5C-D). We therefore propose that v39H and v42L contribute to cooperativity by stabilizing the dimerization interface in a DNA-dependent manner. Consistent with this view, this loop has been shown in extant SRs to undergo functionally relevant conformational changes when DNA is bound (Wikstrom et al., 1999; Berglund et al., 1997; Watson et al., 2013; Meijsing et al., 2009). The remaining permissive substitutions may enhance non-specific DNA binding because they are involved in contacts to the DNA backbone or other base-nonspecific interactions. Substitution w22L is adjacent to several backbone-contacting residues (Fig. 5A), and the other permissive substitutions are in the C-terminal tail; although unresolved in our ancestral crystal structures, this region binds directly to the DNA backbone or minor groove just outside the core RE in other nuclear receptors (Helsen et al., 2012; Roemer et al., 2006; Meijsing et al., 2009).

Taken together, our findings indicate that numerous permissive substitutions, which increased nonspecific affinity, were necessary for the affinity-reducing effects of the RH mutations to be tolerated. The evolving DBD therefore traversed sequence space extensively without changing its specificity, reaching regions relatively distant from AncSR1, before the transition to a new function via the RH substitutions could be completed. Selection for the derived specificity could not have driven this exploration; either neutral chance processes (such as drift and linkage) or selection for functions unrelated to specificity must therefore have played crucial roles in the evolution of AncSR2’s DNA recognition mechanism.

DISCUSSION

Evolution of a new gene regulatory module

These results, together with our previous work on the evolution of the ancestral ligand binding domain, elucidate the mechanisms by which the distinct regulatory modules mediated by the two classes of extant SRs evolved from an ancestral module mediated by a single TF. We recently reported that AncSR1’s LBD also had ER-like functions, responding specifically to estrogens; after duplication of AncSR1, AncSR2 lost estrogen sensitivity entirely and gained activation by nonaromatized steroids (Eick et al., 2012; Harms et al., 2013); during this period, androgens and progestagens were already produced as intermediates in the synthesis of estrogens (Eick and Thornton, 2011). Our present findings therefore establish that during the interval after the duplication of AncSR1, both AncSR2’s LBD and DBD both evolved entirely new specificities for upstream stimuli and downstream DNA targets (Fig. 6A). The other protein lineage produced by this duplication, which led to the present-day estrogen receptors, maintained the specificity of the ancestral signaling module essentially unchanged for hundreds of millions of years.

Fig. 6. Evolution of a new regulatory module.

Fig. 6

Open in a new tab

(A) After duplication of AncSR1, the ancestral specificity for estrogens (purple stars) and ERE (purple box) was maintained to the present in the ER lineage. In the lineage leading to AncSR2, ancestral specificity for both DNA and hormone was lost, and novel sensitivity evolved for SREs (green box) and nonaromatized steroids (green star). A new set of target genes (light grey) was thus activated in response to different stimuli. Green hashes mark the branch on which these events occurred. (B-D) Other potential evoutionary trajectories for evolving new functions would interfere with the ancestral signaling network. (B) Evolution of new specificity for DNA or ligand would cause activation of old targets by new stimuli, or activation of new targets in response to ancestral stimuli. (C-D) Evolution of promiscuity in one or both domains would cause similar effects. (E) The shift in specificity from ERE (purple helices) to SREs (green helices) in AncSR2 involved losing favorable interactions (orange arrows) to ERE, losing unfavorabl negative interactions (red bars) to SRE, and gaining unfavorable interactions to ERE. Offsetting the loss of positive interactions in the DNA major groove, AncSR2 evolved favorable non-specific DNA contacts (blue arrows) and protein-protein interactions (white arrows in dimer interface) that increased cooperativity.

By evolving distinctly new specificities in both domains after gene duplication, a new regulatory module was established without interfering with the functional specificity of the ancestral module. If one domain of AncSR2 had retained the ancestral specificity while the other evolved new interactions, the information conveyed by the ancestral signaling system would have been compromised by noise: ancestral targets would have been activated by additional stimuli, or the ancestral stimuli would have activated additional targets (Fig. 6B). A similar effect would have ensued if the DBD and/or LBD became promiscuous (Fig. 6C-D). Because the new specificities for hormone and DNA evolved during the same phylogenetic interval, we cannot determine which appeared first. It is possible that a promiscuous DBD arose as an evolutionary intermediate during the transition between the distinct RE-specificities of AncSR1 and AncSR2. If it did, however, it did so transiently, was abolished relatively rapidly, and left no promiscuous descendants that persist in present-day species. Thus, the distinct AncSR2-mediated signaling module arose by establishing new functional connections and, just as importantly, by actively erasing the ancestral connections.

In both domains, just a few key mutations – three in the DBD and two in the LBD (Harms et al., 2013) – changed the protein’s binding preferences by many orders of magnitude. These substitutions dramatically impaired interactions with the ancestral partner and, to a lesser extent, improved binding of the ancestral TF to the derived partner. In both domains, the biophysical mechanisms for this transition involved changes in negative determinants of specificity: the key mutations introduced unfavorable steric or electrostatic clashes with estrogens or ERE and removed clashes that in the ancestral state impaired binding to nonaromatized steroids and SREs (Harms et al., 2013). These data indicate that negative determinants of specificity – mechanisms that actively prevent binding to “non-target” partners – played key roles in the evolution of the new AncSR2-mediated regulatory module (Fig. 6E).

Negative determinants of specificity: mutational constraints on TF evolution

AncSR2’s new DNA specificity was conferred by a complex set of changes: three RH-mediated mutations that changed exclusionary interactions and a large number of permissive mutations that offset the affinity-reducing effects of the specificity-switching mutations. Why did evolution not utilize a simpler mechanism to cause the shift in specificity, such as gains and losses of positive interactions? We propose that differences in the abundance of mutational opportunities to establish negative vs. positive mechanisms of specificity determined the evolutionary trajectory by which AncSR2’s new mode of DNA recognition evolved.

As a protein evolves, it drifts through a “neutral network” of neighboring genotypes with similar functional outputs; it may cross into a network that encodes different functions, if one is accessible by mutation and compatible with selective constraints (Smith, 1970; Wagner, 2008). Biophysical considerations suggest that there may be few mutational opportunities to increase affinity in a sequence-specific fashion. Establishing a new sequence-specific positive interaction in the complex, heterogeneous interface with DNA would require introducing a side chain of fairly precise length, angle, volume, polarity, and charge to interact favorably with a feature of DNA that is unique to the target sequence, all without disrupting other aspects of the protein-DNA complex. In contrast, the requirements to establish a negative interaction via a steric or electrostatic clash are likely to be considerably less precise, as are those to abolish a hydrogen bond and thereby leave unpaired polar atoms in an interface. Thus, just as the integrated architecture of protein folds makes mutations that stabilize proteins more rare than those that destabilize them (Bloom et al., 2006), the biophysical architecture of protein-DNA interactions should make mutations that shift specificity by establishing new sequence-specific positive interactions much more rare than those that do so by reducing affinity for non-target sequences.

Evolutionary trajectories that utilize predominantly negative mechanisms to achieve specificity – like those during the evolution of AncSR2’s DBD and LBD – should therefore be more likely to be realized than those that change specificity by establishing new, sequence-specific positive interactions. Consistent with this view, directed evolution experiments that select for specific binding to a new DNA target typically reduce affinity (Rockah-Shmuel and Tawfik, 2012). Further, studies that select for binding without selecting for specificity usually increase affinity in a non-specific fashion (Cohen et al., 2004), indicating that increased affinity often evolves because of non-specific positive interactions, but specificity is realized largely through sequence-specific negative interactions.

Although they are more numerous, mutations that shift specificity by negative, exclusionary interactions would be eliminated by natural selection if they were to reduce affinity to a level below that required for target gene activation, as the RH substitutions do if introduced directly into AncSR1. The historical permissive mutations, by increasing cooperativity and nonspecific affinity, moved the evolving AncSR2 into a region of its neutral network in which the historical specificity-inducing mutations could be tolerated. This evolutionary dynamic is similar to that observed for permissive mutations that increase protein stability and therefore allow destabilizing mutations that confer new functions to be tolerated (Bloom et al., 2006). In the present case, however, the critical parameter is the binding affinity of a protein-DNA complex, rather than the stability of the protein fold. Because macroscopic binding affinity is determined by both half-site affinity and cooperativity, permissive mutations that enhance either parameter – or both, as is the case for the evolution of the SR DBD—could facilitate the evolution of new TF specificity and the rewiring of transcriptional circuits (Tuch et al., 2008; Li and Johnson, 2010).

Because of the limitations imposed by mutational opportunities and purifying selection, AncSR2 evolved distinct, high-affinity DNA binding using a mechanism that is not the simplest or most elegant form imaginable for a TF-DNA complex. But it was the mechanism that happened to be available, given AncSR2’s chance wanderings through sequence space and the constraints imposed by the physical architecture of SR proteins, DNA, and the interaction between them. That ancient, awkward mechanism persists to the present.

EXPERIMENTAL PROCEDURES

Ancestral sequences and posterior probability distributions for AncSR1 and AncSR2 DBDs were inferred using maximum-likelihood phylogenetics from an alignment of 213 peptide sequences of extant steroid and related receptors, the maximum likelihood gene family phylogeny, and the best-fit evolutionary model (JTT+G) (see Eick et al., 2012). Complementary DNAs coding for these peptides were synthesized and subcloned and expressed as fusion constructs with the NFκB-activation domain in CV-1 cell line. Activation was measured using a dual luciferase assay in which firefly luciferase expression was driven by four copies of ERE or SRE. Variant proteins were generated using Quikchange mutagenesis and verified by sequencing. To measure the energetics of binding, tagged DBDs were expressed in E. coli and purified by affinity chromatography; we measured the change in fluorescence polarization of 6-FAM labeled double-stranded DNA oligos as protein concentration increased. Oligos containing a single half-site or a full palindromic element were assayed, and the data were globally fit to a two-site model with a cooperativity parameter to determine the half-site affinity and the cooperativity coefficient (the fold-increase in the KA of dimeric binding compared to the expected value if the monomers bind independently (Hard et al., 1990)). To measure protein stability we used circular dichroism to measure the reversible loss of secondary structure in increasing guanidinium chloride. Protein dimerization was assayed by sedimentation velocity analytical centrifugation. For crystallography, purified DBDs were crystallized in complex with palindromic DNA oligos and diffracted at the Advanced Photon Source; structures were determined using molecular replacement. Atomic coordinates were deposited as AncSR1:ERE (PDB 4OLN, 1.5 Å), AncSR2:SRE1 (4OOR, 2.7 Å), AncSR2+rh:ERE (4OND, 2.2 Å), and AncSR2+rh:SRE1, (4OV7, 2.4 Å). Molecular interactions were characterized with molecular dynamics simulations using Gromacs, TIP3P waters and AMBER FF03 parameters for protein and DNA. For each condition, three replicate 50 ns simulations were run, starting from crystal structures of ancestral proteins; historical mutations were introduced and energy minimized before MD simulation. For details, see Extended Experimental Procedures in Supplemental Information.

Supplementary Material

1
2
3
4
5
6
7

Acknowledgments

We thank Geeta Eick for phylogenetic analysis and Mike Harms for extensive advice. We thank Vincent Lynch, Pete von Hippel, members of the Thornton Lab, and Will Hudson for comments on the manuscript. The University of Oregon ACISS cluster provided computing resources. This work was supported by NIH R01-GM104397 (JWT), a Howard Hughes Medical Institute Early Career Scientist Award (JWT), NIH training grant 5-T32-GM-7759-33 (ANM), and AHA award 11PRE7510085 (DWA).

Footnotes

AUTHOR CONTRIBUTIONS

ANM, JTB, and JWT conceived the project. All authors designed the experiments and analyzed data. JTB performed the functional characterizations of ancestral proteins and their variants and identified key historical substitutions; ANM performed the biochemical and biophysical characterizations of ancestral proteins and their variants; DWA performed the molecular dynamics simulations; MNM and EAO performed X-ray crystallography and preliminary biophysical characterizations. ANM and JWT wrote the paper, with contributions from all authors.

SUPPLEMENTAL INFORMATION

Supplemental Information includes Extended Experimental Procedures, 6 figures, and 6 tables.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Alroy I, Freedman LP. DNA binding analysis of glucocorticoid receptor specificity mutants. Nucleic Acids Res. 1992;20:1045–1052. doi: 10.1093/nar/20.5.1045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bain DL, Heneghan AF, Connaghan-Jones KD, Miura MT. Nuclear receptor structure: implications for function. Annu Rev Physiol. 2007;69:201–220. doi: 10.1146/annurev.physiol.69.031905.160308. [DOI] [PubMed] [Google Scholar]
  3. Bain DL, Yang Q, Connaghan KD, Robblee JP, Miura MT, Degala GD, Lambert JR, Maluf NK. Glucocorticoid receptor-DNA interactions: binding energetics are the primary determinant of sequence-specific transcriptional activity. J Mol Biol. 2012;422:18–32. doi: 10.1016/j.jmb.2012.06.005. [DOI] [PubMed] [Google Scholar]
  4. Baker CR, Booth LN, Sorrells TR, Johnson AD. Protein modularity, cooperative binding, and hybrid regulatory states underlie transcriptional network diversification. Cell. 2012;151:80–95. doi: 10.1016/j.cell.2012.08.018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Baker CR, Tuch BB, Johnson AD. Extensive DNA-binding specificity divergence of a conserved transcription regulator. Proc Natl Acad Sci USA. 2011;108:7493–7498. doi: 10.1073/pnas.1019177108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Beato M, Chalepakis G, Schauer M, Slater EP. DNA regulatory elements for steroid hormones. J Steroid Biochem. 1989;32:737–747. doi: 10.1016/0022-4731(89)90521-9. [DOI] [PubMed] [Google Scholar]
  7. Beato M, Sanchez-Pacheco A. Interaction of steroid hormone receptors with the transcription initiation complex. Endocr Rev. 1996;17:587–609. doi: 10.1210/edrv-17-6-587. [DOI] [PubMed] [Google Scholar]
  8. Bentley PJ. Comparative Vertebrate Endocrinology. Cambridge, UK: Cambridge University Press; 1998. [Google Scholar]
  9. Berglund H, Wolf-Watz M, Lundback T, van den Berg S, Hard T. Structure and dynamics of the glucocorticoid receptor DNA-binding domain: comparison of wild type and a mutant with altered specificity. Biochemistry. 1997;36:11188–11197. doi: 10.1021/bi970343i. [DOI] [PubMed] [Google Scholar]
  10. Bershtein S, Segal M, Bekerman R, Tokuriki N, Tawfik DS. Robustness-epistasis link shapes the fitness landscape of a randomly drifting protein. Nature. 2006;444:929–932. doi: 10.1038/nature05385. [DOI] [PubMed] [Google Scholar]
  11. Bloom JD, Labthavikul ST, Otey CR, Arnold FH. Protein stability promotes evolvability. Proc Natl Acad Sci U S A. 2006;103:5869–5874. doi: 10.1073/pnas.0510098103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Brayer KJ, Lynch VJ, Wagner GP. Evolution of a derived protein-protein interaction between HoxA11 and Foxo1a in mammals caused by changes in intramolecular regulation. Proc Natl Acad Sci USA. 2011;108:E414–20. doi: 10.1073/pnas.1100990108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Carroll SB. Evo-devo and an expanding evolutionary synthesis: a genetic theory of morphological evolution. Cell. 2008;134:25–36. doi: 10.1016/j.cell.2008.06.030. [DOI] [PubMed] [Google Scholar]
  14. Chusacultanachai S, Glenn KA, Rodriguez AO, Read EK, Gardner JF, Katzenellenbogen BS, Shapiro DJ. Analysis of estrogen response element binding by genetically selected steroid receptor DNA binding domain mutants exhibiting altered specificity and enhanced affinity. J Biol Chem. 1999;274:23591–23598. doi: 10.1074/jbc.274.33.23591. [DOI] [PubMed] [Google Scholar]
  15. Cohen HM, Tawfik DS, Griffiths AD. Altering the sequence specificity of HaeIII methyltransferase by directed evolution using in vitro compartmentalization. Protein Eng Des Sel. 2004;17:3–11. doi: 10.1093/protein/gzh001. [DOI] [PubMed] [Google Scholar]
  16. Eick GN, Colucci JK, Harms MJ, Ortlund EA, Thornton JW. Evolution of minimal specificity and promiscuity in steroid hormone receptors. PLoS Genet. 2012;8:e1003072. doi: 10.1371/journal.pgen.1003072. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Eick GN, Thornton JW. Evolution of steroid receptors from an estrogen-sensitive ancestral receptor. Mol Cell Endocrinol. 2011;334:31–38. doi: 10.1016/j.mce.2010.09.003. [DOI] [PubMed] [Google Scholar]
  18. Garvie CW, Wolberger C. Recognition of specific DNA sequences. Mol Cell. 2001;8:937–946. doi: 10.1016/s1097-2765(01)00392-6. [DOI] [PubMed] [Google Scholar]
  19. Gong LI, Suchard MA, Bloom JD. Stability-mediated epistasis constrains the evolution of an influenza protein. Elife. 2013;2:e00631. doi: 10.7554/eLife.00631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Hard T, Dahlman K, Carlstedt-Duke J, Gustafsson JA, Rigler R. Cooperativity and specificity in the interactions between DNA and the glucocorticoid receptor DNA-binding domain. Biochemistry. 1990;29:5358–5364. doi: 10.1021/bi00474a022. [DOI] [PubMed] [Google Scholar]
  21. Harms MJ, Eick GN, Goswami D, Colucci JK, Griffin PR, Ortlund EA, Thornton JW. Biophysical mechanisms for large-effect mutations in the evolution of steroid hormone receptors. Proc Natl Acad Sci USA. 2013;110:11475–11480. doi: 10.1073/pnas.1303930110. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Harms MJ, Thornton JW. Analyzing protein structure and function using ancestral gene reconstruction. Curr Opin Struct Biol. 2010;20:360–366. doi: 10.1016/j.sbi.2010.03.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Harms MJ, Thornton JW. Evolutionary biochemistry: revealing the historical and physical causes of protein properties. Nat Rev Genet. 2013;14:559–571. doi: 10.1038/nrg3540. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Helsen C, Kerkhofs S, Clinckemalie L, Spans L, Laurent M, Boonen S, Vanderschueren D, Claessens F. Structural basis for nuclear hormone receptor DNA binding. Mol Cell Endocrinol. 2012;348:411–417. doi: 10.1016/j.mce.2011.07.025. [DOI] [PubMed] [Google Scholar]
  25. Kumar V, Chambon P. The estrogen receptor binds tightly to its responsive element as a ligand-induced homodimer. Cell. 1988;55:145–156. doi: 10.1016/0092-8674(88)90017-7. [DOI] [PubMed] [Google Scholar]
  26. Li H, Johnson AD. Evolution of transcription networks—lessons from yeasts. Curr Biol. 2010;20:R746–53. doi: 10.1016/j.cub.2010.06.056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Luisi BF, Xu WX, Otwinowski Z, Freedman LP, Yamamoto KR, Sigler PB. Crystallographic analysis of the interaction of the glucocorticoid receptor with DNA. Nature. 1991;352:497–505. doi: 10.1038/352497a0. [DOI] [PubMed] [Google Scholar]
  28. Lundback T, Cairns C, Gustafsson JA, Carlstedt-Duke J, Hard T. Thermodynamics of the glucocorticoid receptor-DNA interaction: binding of wild-type GR DBD to different response elements. Biochemistry. 1993;32:5074–5082. doi: 10.1021/bi00070a015. [DOI] [PubMed] [Google Scholar]
  29. Lynch VJ, May G, Wagner GP. Regulatory evolution through divergence of a phosphoswitch in the transcription factor CEBPB. Nature. 2011;480:383–386. doi: 10.1038/nature10595. [DOI] [PubMed] [Google Scholar]
  30. Meijsing SH, Pufall MA, So AY, Bates DL, Chen L, Yamamoto KR. DNA binding site sequence directs glucocorticoid receptor structure and activity. Science. 2009;324:407–410. doi: 10.1126/science.1164265. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Nelson CC, Hendy SC, Shukin RJ, Cheng H, Bruchovsky N, Koop BF, Rennie PS. Determinants of DNA sequence specificity of the androgen, progesterone, and glucocorticoid receptors: evidence for differential steroid receptor response elements. Mol Endocrinol. 1999;13:2090–2107. doi: 10.1210/mend.13.12.0396. [DOI] [PubMed] [Google Scholar]
  32. Rockah-Shmuel L, Tawfik DS. Evolutionary transitions to new DNA methyltransferases through target site expansion and shrinkage. Nucleic Acids Res. 2012;40:11627–11637. doi: 10.1093/nar/gks944. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Roemer SC, Donham DC, Sherman L, Pon VH, Edwards DP, Churchill ME. Structure of the progesterone receptor-deoxyribonucleic acid complex: novel interactions required for binding to half-site response elements. Mol Endocrinol. 2006;20:3042–3052. doi: 10.1210/me.2005-0511. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Rohs R, Jin X, West SM, Joshi R, Honig B, Mann RS. Origins of specificity in protein-DNA recognition. Annu Rev Biochem. 2010;79:233–269. doi: 10.1146/annurev-biochem-060408-091030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sayou C, Monniaux M, Nanao MH, Moyroud E, Brockington SF, Thevenon E, Chahtane H, Warthmann N, Melkonian M, Zhang Y, Wong GK, Weigel D, Parcy F, Dumas R. A promiscuous intermediate underlies the evolution of LEAFY DNA binding specificity. Science. 2014;343:645–648. doi: 10.1126/science.1248229. [DOI] [PubMed] [Google Scholar]
  36. Schwabe JW, Chapman L, Finch JT, Rhodes D. The crystal structure of the estrogen receptor DNA-binding domain bound to DNA: how receptors discriminate between their response elements. Cell. 1993;75:567–578. doi: 10.1016/0092-8674(93)90390-c. [DOI] [PubMed] [Google Scholar]
  37. Schwabe JW, Rhodes D. Beyond zinc fingers: steroid hormone receptors have a novel structural motif for DNA recognition. Trends Biochem Sci. 1991;16:291–296. doi: 10.1016/0968-0004(91)90121-b. [DOI] [PubMed] [Google Scholar]
  38. Smith JM. Natural selection and the concept of a protein space. Nature. 1970;225:563–564. doi: 10.1038/225563a0. [DOI] [PubMed] [Google Scholar]
  39. So AY, Chaivorapol C, Bolton EC, Li H, Yamamoto KR. Determinants of cell- and gene-specific transcriptional regulation by the glucocorticoid receptor. PLoS Genet. 2007;3:e94. doi: 10.1371/journal.pgen.0030094. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Stern DL, Orgogozo V. Is genetic evolution predictable? Science. 2009;323:746–751. doi: 10.1126/science.1158997. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Teichmann SA, Babu MM. Gene regulatory network growth by duplication. Nat Genet. 2004;36:492–496. doi: 10.1038/ng1340. [DOI] [PubMed] [Google Scholar]
  42. Thornton JW. Resurrecting ancient genes: experimental analysis of extinct molecules. Nat Rev Genet. 2004;5:366–375. doi: 10.1038/nrg1324. [DOI] [PubMed] [Google Scholar]
  43. Tuch BB, Li H, Johnson AD. Evolution of eukaryotic transcription circuits. Science. 2008;319:1797–1799. doi: 10.1126/science.1152398. [DOI] [PubMed] [Google Scholar]
  44. Umesono K, Evans RM. Determinants of target gene specificity for steroid/thyroid hormone receptors. Cell. 1989;57:1139–1146. doi: 10.1016/0092-8674(89)90051-2. [DOI] [PubMed] [Google Scholar]
  45. von Hippel PH, Berg OG. On the specificity of DNA-protein interactions. Proc Natl Acad Sci U S A. 1986;83:1608–1612. doi: 10.1073/pnas.83.6.1608. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Wagner A. Neutralism and selectionism: a network-based reconciliation. Nat Rev Genet. 2008;9:965–974. doi: 10.1038/nrg2473. [DOI] [PubMed] [Google Scholar]
  47. Watson LC, Kuchenbecker KM, Schiller BJ, Gross JD, Pufall MA, Yamamoto KR. The glucocorticoid receptor dimer interface allosterically transmits sequence-specific DNA signals. Nat Struct Mol Biol. 2013;20:876–883. doi: 10.1038/nsmb.2595. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Welboren WJ, Sweep FC, Span PN, Stunnenberg HG. Genomic actions of estrogen receptor alpha: what are the targets and how are they regulated? Endocr Relat Cancer. 2009;16:1073–1089. doi: 10.1677/ERC-09-0086. [DOI] [PubMed] [Google Scholar]
  49. Wikstrom A, Berglund H, Hambraeus C, van den Berg S, Hard T. Conformational dynamics and molecular recognition: backbone dynamics of the estrogen receptor DNA-binding domain. J Mol Biol. 1999;289:963–979. doi: 10.1006/jmbi.1999.2806. [DOI] [PubMed] [Google Scholar]
  50. Wray GA. The evolutionary significance of cis-regulatory mutations. Nat Rev Genet. 2007;8:206–216. doi: 10.1038/nrg2063. [DOI] [PubMed] [Google Scholar]
  51. Zilliacus J, Dahlman-Wright K, Wright A, Gustafsson JA, Carlstedt-Duke J. DNA binding specificity of mutant glucocorticoid receptor DNA-binding domains. J Biol Chem. 1991;266:3101–3106. [PubMed] [Google Scholar]
  52. Zilliacus J, Wright AP, Norinder U, Gustafsson JA, Carlstedt-Duke J. Determinants for DNA-binding site recognition by the glucocorticoid receptor. J Biol Chem. 1992;267:24941–24947. [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS629207-supplement-1.pdf (834.8KB, pdf)
2
NIHMS629207-supplement-2.pdf (2.3MB, pdf)
3
NIHMS629207-supplement-3.pdf (475.4KB, pdf)
4
NIHMS629207-supplement-4.pdf (467.6KB, pdf)
5
NIHMS629207-supplement-5.pdf (1.1MB, pdf)
6
NIHMS629207-supplement-6.pdf (5MB, pdf)
7
NIHMS629207-supplement-7.docx (183.9KB, docx)

RESOURCES