Transition state for protein–DNA recognition

Abstract

We describe the formation of protein–DNA contacts in the two-state route for DNA sequence recognition by a transcriptional regulator. Surprisingly, direct sequence readout establishes in the transition state and constitutes the bottleneck of complex formation. Although a few nonspecific ionic interactions are formed at this early stage, they mainly play a stabilizing role in the final consolidated complex. The interface is fairly plastic in the transition state, likely because of a high level of hydration. The overall picture of this two-state route largely agrees with a smooth energy landscape for binding that speeds up DNA recognition. This “direct” two-state route differs from the parallel multistep pathway described for this system, which involves nonspecific contacts and at least two intermediate species that must involve substantial conformational rearrangement in either or both macromolecules.

Protein recognition of specific double-stranded DNA sequences (sequence readout) is a primary event in gene expression and genome replication. Sequence recognition is thought to start by the diffusion of the protein to the DNA site along a mixture of 1D and 3D searches (1). Some proteins diffuse mainly in 1D over long distances (2), whereas others favor a 3D pathway (3). Once at the target site, proteins without an enzymatic activity may bind to DNA along a single kinetic route (4–9) or parallel routes (10, 11). Each route may either be two-state (4–6, 10) or include populated kinetic intermediates (7–11). In the final consolidated complex, sequence recognition is mediated by specific contacts between protein residues and DNA bases (direct sequence readout) and by the conformational energetics of noncontacted bases (indirect sequence readout) (12–16).

The development of “native” sequence readout along protein–DNA binding pathways is not well understood because of the limited number of accessible experimental models. Binding experiments that compare the kinetics of recognition of cognate and noncognate sequences and of wild-type sites and point mutants showed that some proteins do not display any level of sequence discrimination in the transition state for binding (4, 5, 17, 18). Other proteins do discriminate between sequences in the transition state (6, 19, 20), sometimes as much as in the final complex (19, 20). The detailed energetics of sequence readout and unspecific protein–DNA backbone interactions may also be analyzed by the effect that mutations exert on the binding kinetics (5, 6, 20, 21). Unfortunately, the global interpretation of protein–DNA binding reactions requires an extensive and systematic mutagenesis of the interface.

Sequence recognition in protein–DNA binding is often associated with conformational changes in either or both reagents. Changes in the protein range from subtle rearrangements (9, 14, 22–24) to local folding reactions (25, 26). The energy landscape theory predicts that protein flexibility speeds up DNA binding (27, 28). The DNA site can be slightly (9, 14, 22, 23) or strongly bent (8) in the final complex. Bending and binding may occur in a stepwise (8) or concerted manner (7).

The complex between the C-terminal domain of the human papillomavirus type 16 E2 master regulator (E2C) and a cognate double-stranded DNA oligonucleotide is a model system for specific protein–DNA binding. On binding of HPV16 E2C to DNA, the conformational changes are relatively large for the DNA and small for the protein (9, 14, 23, 24, 29, 30). However, the DNA recognition elements of E2C, helix1, and the β2–β3 loop, show substantial flexibility in the unbound form (24, 31). The energetic contributions of the individual E2C residues to site-specific binding, and the overall binding thermodynamics are well characterized (13, 15, 24). Direct and indirect readout were shown to be decoupled (14, 15). In stopped-flow experiments where E2C recognizes a short target oligonucleotide, binding takes place along two parallel routes, one of them two-state and the other with at least two populated intermediates (10). The multistate route could be a model for a stepwise transition from nonspecific to specific binding after 1D diffusion (26, 32) and the two-state route for binding after 3D diffusion. Here, we analyze the transition state ensemble of the two-state route for E2C–DNA binding through an extensive mutagenesis of amino acids and bases at the complex interface and stopped-flow kinetic measurements, testing for direct sequence readout, interface packing, and DNA bending in the transition state.

Results

Structure and Thermodynamics of the E2C–DNA Interface.

Papillomavirus E2C domains form 8-stranded homodimeric β-barrels that expose four α-helices (9, 14, 22–24). The E2C target DNA site consists of two highly conserved four-base half-sites connected by a four-base linker (14, 33). On complexation, specific contacts are formed mainly between the two symmetrical DNA-binding helices and consecutive major grooves of the DNA half-sites. A short loop between the second and third β strands forms nonspecific contacts with the DNA backbone (9, 14, 22–24). The 3D structure of the HPV16 E2C–DNA complex at atomic resolution has not been determined to date. We have modeled the HPV16 E2C–DNA interface based on the structures of the complexes of HPV18, HPV6, and BPV1 E2C domains with DNA (9, 14, 22, 23) and on the NMR backbone structure of the HPV16 E2C–DNA complex (24) (see supporting information (SI) Text, Figs. S1–S6, and Tables S1 and S2). Residues N294, K297, C298, Y301, and R302 in helix1 form contacts to DNA bases. N294, K297, Y301, and R302 also contact the DNA backbone. Residues T295, R300, K304, and K305 in helix1; V324, K325, and K327 in the β2–β3 loop; and T316, K349, and T353 outside of the main recognition elements contact the DNA backbone only. Residues N294, K297, C298, R300, Y301, and R302 are >94% conserved in alpha papillomaviruses and are the main determinants of E2C sequence specificity (33). Fig. 1A shows a surface representation of the DNA-bound conformation of the HPV16 E2C domain (24). The recognition helix is at the bottom left of the image and the β2–β3 loop at the top. We have recently used 20-point mutants to characterize the contribution of 12 of the 15 recognition residues to the equilibrium binding of HPV16 E2C to one of its cognate target DNA sites (15) (Table 1, Figs. 1A and 2A, black bars). We will use these equilibrium measurements as a starting point to interpret our kinetic results.

Fig. 1.

Effect of point mutations on the stability of the E2C–DNA complex (A) (15) and the transition state for binding (B), represented on the surface of the DNA-bound conformation of the HPV16 E2C homodimer (24). Residues in helix1 are grouped at the bottom left part of the plot, residues in the β2–β3 loop are grouped at the top of the plot. Residues contacting the DNA bases are underlined. Mutations to alanine are shown, except for K297R, R300M, and R302M. Residues are colored according to ΔΔG, in a scale that goes from −2 (green) to 0 (yellow) to 2 kcal/mol or more (red). The equilibrium data derived from kinetic measurements were used whenever available (Table 1). Uncharacterized residues are shown in gray.

Table 1.

Thermodynamics and kinetics of E2C–DNA binding for E2C point mutants

E2C variant	ΔΔGeq from equilibrium*	kON/108, M−1s−1	kOFF, s−1	ΔΔGeq from kinetics*†	ΔΔGon*‡	φbinding§
Wt	–	8.7	0.32	–	–	–
Helix1 residues with direct contact to bases
N294A	1.75	¶	¶	‖	‖	‖
K297A	1.89	5.8	0.41	0.39**	‖	‖
K297R	3.33	0.54	5.20	3.31	1.65	0.50
C298S	1.25	7.9	0.12	−0.53**	‖	‖
C298A	1.22	1.4	0.37	1.17	1.09	0.93
C298G	1.28	1.9	0.42	1.07	0.90	0.85
R300M	1.33	††	0.46	‖	‖	‖
Y301F	0.16	8.9	0.48	0.23	−0.01	‡‡
Y301A	1.00	2.1	0.49	1.10	0.85	0.77
R302M	0.89	2.1	0.69	1.31	0.85	0.65
Helix1 residues (contacts with the DNA backbone only)
T295A	−0.82	8.5	0.05	−1.09	0.01	−0.01
K304A	1.38	4.3	3.50	1.84	0.42	0.23
K305A	0.90	6.8	0.72	0.63	0.15	0.23
β2–β3 loop residues (contacts with the DNA backbone only)
V324A	0.17	9.1	0.24	−0.20	−0.03	‡‡
K325A	0.63	4.5	0.46	0.61	0.39	0.64
K325R	−0.08	13	0.19	−0.55	−0.24	‡‡
K327A	0.75	3.5	0.74	1.04	0.54	0.52
K327R	0.09	8.3	0.27	−0.07	0.03	‡‡
Mutations probing flexibility of the E2C–DNA interface
Y301N	2.04	3.6	6.79	2.34	0.52	0.22
R302Q	1.08	0.59	0.14	1.11	1.60	1.44

The average error in the determination of the rate constants is 10%, which propagates to an average error of 0.1 kcal/mol in the ΔΔG-values and of <0.2 for φ_binding.

*In kcal/mol, data from ref. 15.

^†Calculated as RT·(ln(k_ON^wt/k_ON^mut) − ln(k_OFF^wt/k_OFF^mut)).

^‡Calculated as RT·ln(k_ON^wt/k_ON^mut).

^§ΔΔG_on/ΔΔG_eq from kinetics.

^¶Not measured. This mutant could not be purified in sufficient amounts because of its high tendency to aggregate.

^‖Not determined.

**ΔΔG_eq from equilibrium and kinetic measurements do not agree.

^††Binding rate was independent of concentration.

^‡‡ΔΔG_eq too small to calculate φ_binding.

Fig. 2.

Mutational analysis of the two-state pathway for E2C–DNA binding. (A) Effect of point mutations on the stability of the E2C–DNA complex (black bars) and the transition state for binding (red bars). (B) Side-chain interactions in the transition state for binding of E2C to its target DNA, as measured by φ_binding. ΔΔG_eq for mutations Y301F, V324A, K325R, and K327R is too small to calculate φ_binding.

Binding Kinetics of E2C Point Mutants.

We measured the kinetics for binding and dissociation of the E2C–DNA complex for 19 of the 20 mutants of our dataset to characterize the transition state ensemble of the two-state route for E2C–DNA binding (see Methods and SI Text) (Table 1). The average error in the determination of the rate constants is 10%. We were able to extract the rate constants for association (k_ON) and dissociation (k_OFF) corresponding to the two-state route for 16 of the 19 mutants (Table 1 and SI Text). For these 16 variants there is an excellent agreement between the changes in binding free energy measured in equilibrium (15) and those calculated from k_ON and k_OFF (Table 1 and SI Text), considering the complexity of the reaction. The association kinetics for the R300M variant is concentration-independent (data not shown), and the ΔΔG_eq calculated from k_ON and k_OFF do not agree for the K297A and C298S variants. This suggests that the two-state route for binding is not kinetically significant in these three mutant complexes and that these three side chains play a role in directing binding of wild-type E2C along this route. We conclude that the overall mechanism for E2C–DNA binding is fairly robust to mutation.

Direct Readout in the Transition State for E2C–DNA Binding.

The free-energy barrier that separates the unbound reagents from the consolidated complex determines the rate constant for formation of the E2C–DNA complex along the two-state route. The effect of mutations on this free energy barrier, ΔΔG_on, can be extracted from the association rate constants for the wild-type (k_ON^wt) and mutant complexes (k_ON^mut):

The results are shown in Table 1, Fig. 1B, the red bars in Fig. 2A and the SI Text. The average error in ΔΔG_on is 0.1 kcal/mol. Six mutations probe four side chains in E2C helix1 that contact the DNA bases directly and thus are responsible for direct sequence readout at equilibrium. Mutation Y301F does not have a measurable effect on k_ON, indicating that the hydroxyl group of Y301 does not stabilize the transition state relative to the unbound state. All other mutations probing direct contacts to DNA bases decrease k_ON at least 4-fold. We have also analyzed the kinetics of E2C binding to two DNA oligonucleotides without specific amino acid-base contacts (Table 2, EBNA and iset) (10). Loss of all specific contacts to the DNA bases has a ΔΔG_ON of 4.3–4.9 kcal/mol (Table 2) in good agreement with the added effect of mutations K297R, C298A, Y301A, and R302M at 4.4 kcal/mol. This suggests that the observed reduction in k_ON for these mutants are mainly due to the loss of specific contacts with the DNA bases and not to loss of unspecific contacts from K297, Y301, and R302 to the DNA backbone. We observe that the distribution of the energetic perturbation to the transition state for binding is similar to that attributed to “direct readout” of the DNA sequence at equilibrium.

Table 2.

Thermodynamics and kinetics of E2C–DNA binding for different DNA sites

DNA site	Perturbation	ΔΔGeq from equilibrium, kcal/mol	ΔΔGeq from kinetics, kcal/mol*	ΔΔGon†	φbinding‡
EBNA§	All bases	5.48	6.05	4.90	0.81
iset¶	All bases	6.41	6.89	4.27	0.62
BOV‖	Linker bases	1.91	1.41	0.08	0.06

Data from refs. 10 and 13. Rate constants were assigned to the two-state route to achieve an agreement with equilibrium measurements. In all three cases, this could be achieved with only one combination of rate constants. The sequence of the wild-type oligonucleotide is TCAACCGATTTCGGTTAC.

*Calculated as RT·(ln(k_ON^wt/k_ON^mut) − ln(k_OFF^wt/k_OFF^mut)).

^†Calculated as RT·ln(k_ON^wt/k_ON^mut).

^‡ΔΔG_on/ΔΔG_eq from kinetics.

^§Sequence GGGTAGCATATGCTACCC.

^¶Sequence TACTTGACAGGTCCATGT.

^‖Sequence CCGACCGACGTCGGTCGG.

Eight mutations probe six side chains in helix1 and the β2–β3 loop that contact the DNA backbone only. These mutations decrease k_ON at most 2.5-fold. The added effect of mutations of these residues to alanine is 1.5 kcal/mol, a small figure compared to 4.4 kcal/mol for mutations of residues contacting the DNA bases. This suggests that contacts of E2C side chains to the DNA backbone stabilize the transition state modestly relative to the unbound reagents, in contrast to the effect on the final complex (15).

Altogether, we observe that most E2C side chains contacting the DNA stabilize the transition state for binding relative to the unbound reagents. However, the contribution of side chains contacting the DNA bases is clearly larger than that of side chains that only contact the DNA backbone.

φ-Value Analysis of Side Chain Interactions in the Transition State for E2C–DNA Binding.

In the preceding section, we examined the role of E2C side chains in stabilizing the transition state for binding relative to the unbound reagents. We will now study the E2C–DNA interactions in the transition state, that is, to which degree the transition state for binding resembles the equilibrium complex as opposed to the unbound reagents. To this end, we make use of φ-value analysis (34, 35). φ_binding measures to which degree the energetic contribution of the interactions made by the mutated side chain in the transition state (ΔΔG_on) resembles the energetic contribution of the interactions made by the mutated side chain in the final complex (ΔΔG_eq).

This is a normalized measure that goes, in principle, from 0 (if the energetics of the mutated side chain in the transition state ensemble resemble the unbound reagents) to 1 (if the energetics of the mutated side chain in the transition state ensemble resemble the final complex) (34, 35).

The results for the E2C–DNA complex are shown in Fig. 2B, Table 1, Fig. 3A, and the SI Text. The average error in φ_binding is <0.2. Mutations probing side chains in helix1 that make direct contacts to the DNA bases have φ_binding ranging between 0.5 and 0.9. This is in agreement with the averaged φ_binding of 0.6–0.8 for specific side chain–base interactions measured by using the iset and EBNA oligonucleotides (Table 2). Mutations probing side chains in helix1 that make contacts to the DNA backbone only have φ_binding ranging between 0 and 0.25, whereas mutations probing side chains in the β2–β3 loop that make contact only to the DNA backbone have φ_binding of 0.5 and 0.6. The φ_binding analysis of the two-state route for formation of the E2C–DNA complex suggests that the energetics of protein–DNA interactions in the transition state is not uniform. The energetics of the direct contacts between side chains and DNA bases are similar to those observed in the final complex.

Fig. 3.

Rate-equilibrium free-energy relationship (34) of E2C point mutants. (A) Side-chain interactions in the transition state for binding of E2C to its target DNA, as measured by φ_binding, represented on the surface of the DNA-bound conformation of the HPV16 E2C homodimer (24). (B) Effect of point mutations on the stability of the E2C–DNA complex and the transition state for binding. We highlight the mutations introducing new polar groups in the E2C–DNA interface, Y301N and R302Q (red filled circles), and the reference mutations Y301A and R302M (black filled circles).

Introduction of New Polar Side Chains in the E2C–DNA.

All φ_binding measured for the E2C–DNA complex is smaller than 1, indicating that the interface is not fully consolidated in the transition state for binding along the two-state route. We have probed the dynamic nature of the interface in the transition state ensemble with two additional mutations of side chains making contacts with the DNA bases, Y301N and R302Q (Table 1 and Fig. 3B). The side chains of residues Y301 and R302 interact with the DNA in the transition state, as indicated by φ_binding close to 0.7 for the Y301A and R302M mutations. Compared with variant R302M, mutation R302Q introduces a new polar amide group into the interface. This change destabilizes the transition state relative to the unbound reagents by 0.75 kcal/mol and stabilizes the final complex by 0.2 kcal/mol. We interpret that the new glutamine side chain forms transient destabilizing interactions in the transition state that disappear on progressing to the final complex.

Compared with variant Y301A, mutation Y301N introduces a new polar amide group into the interface. This change destabilizes the final complex relative to the unbound reagents by 1.3 kcal/mol, while stabilizing the transition state slightly by 0.3 kcal/mol. We interpret that the newly introduced asparagine side chain forms destabilizing interactions that are better tolerated in the transition state than in the final complex. From the behavior of the Y301N and R302Q mutations, we infer that the E2C–DNA interface in the transition state and in the final complex differ in their ability to accommodate new polar groups.

Indirect Readout Does Not Stabilize the Transition State for E2C–DNA Binding.

HPV16 E2C barely changes its average structure on DNA binding in solution (24), whereas the DNA bends toward the minor groove by ≈40°, regardless of binding affinity (14, 29). The four linker bases bend on binding, whereas the two conserved half-sites do not significantly change conformation (9, 14, 22, 23, 30). Because the linker bases are not in direct contact with the protein in the E2C–DNA complex, but their composition does affect the binding energetics (23, 30), the energetic contribution of DNA deformation is called “indirect readout” (12–15).

We have determined the contribution of indirect readout in the transition state for binding of E2C to a DNA site with a mutated linker sequence (ACGT for ATTT, “BOV” entry in Table 2) (10). The free BOV oligonucleotide adopts a nearly straight conformation, whereas a sequence similar to the wild-type oligonucleotide is prebent toward the minor groove in the absence of E2C, although further bending takes place on binding (30). The mutated linker destabilizes the final complex by 1.4 kcal/mol relative to the wild-type linker, as expected for a reduced bendability. However, it destabilizes the transition state by only 0.1 kcal/mol. This indicates that E2C discriminates between linker sequences in the final complex but not in the transition state for binding.

Discussion

The transition state ensemble for binding of E2C to its target DNA along the two-state route has a bipartite structure according to φ_binding-values (Figs. 2B and 3A). A uniformly structured interface region includes specific contacts between helix1 and DNA bases and between the β2–β3 loop and the DNA backbone. Residues in this region of the protein form interactions that stabilize the transition state, speeding up formation of the HPV16 E2C–DNA complex mainly by direct sequence readout (Figs. 1B and 2A). Neutral residues C298 and Y301 mediate specific interactions in the transition state, indicating that direct sequence readout in the transition state cannot be explained only in terms of electrostatic steering. The effects of mutations on k_ON are fairly homogeneous within this region, without kinetic “hot spots,” similar to equilibrium measurements on this complex (15). Our analysis shows that the interactions between helix1 and the DNA backbone are absent or highly distorted in the transition state. This region contributes mainly to the kinetic stability of the consolidated complex by lowering the k_OFF.

It has been suggested that plasticity of the binding partners speeds up specific protein–DNA recognition (27, 28). Interestingly, E2C helix1 and the β2–β3 loop are highly dynamic in the unbound protein (31). The E2C target DNA is also flexible in solution (36). This poses the question of how much of the inherent flexibility of the binding partners is lost in the transition state ensemble for E2C–DNA binding. The change in k_ON induced by a mutation is generally smaller than the change in the equilibrium constant for binding (Figs. 1 and 2A), indicating that E2C–DNA intermolecular interactions are not fully consolidated in the transition state compared with the final complex. Along the same line, the target DNA is not bent into its final conformation in the transition state (Table 2) and the results for the Y301N and R302Q mutations point at differences in interface packing between the transition state and the final complex (Fig. 3B). We infer that both the E2C recognition elements and the target DNA retain some conformational plasticity in the transition state for binding. Conformational flexibility in the transition state may favor the high degree of development of intermolecular interactions in the transition state indicated by the φ_binding-values (37) and, consequently, a fast direct readout of the base sequence.

Equilibrium binding of E2C to DNA is associated with a negative change in heat capacity (38), pointing at a net decrease in solvation and vibrational freedom on binding (39). In the two-state route, the full change in heat capacity takes place only after the transition state (10), compatible with a highly solvated transition state with many degrees of vibrational freedom. We propose that interfacial water molecules mediate an ensemble of many different interactions in the transition state ensemble that underlies the observed flexibility (16). On conversion into the final complex water molecules are expelled from the interface (10).

We have presented clear evidence that direct sequence readout takes place in the transition state ensemble of the two-state route for E2C binding to DNA. That is, the E2 protein binds much faster to its target sequences than to any other site. From an energy landscape perspective, we interpret that the two-state binding route of wild-type E2C has a smooth energy landscape dominated by the native sequence-specific interactions present in the final complex, similar to many protein folding and association reactions (16). The residues responsible for transition state sequence readout and R300, which guides E2C along the two-state route, are >94% conserved in alpha papillomaviruses (33). Moreover, mutations Y301N and R302Q introduce nonnative interactions, indicating that side chains not present in nature can introduce some degree of “frustration” in the binding landscape (16). This suggests that kinetic sequence readout and the funneled energy landscape of the two-state route is a selected property of the E2C domain that may play a role in vivo. Experiments on other protein–DNA complexes by using mutated or nonspecific oligonucleotides or partial mutagenesis of the protein suggested that some of them also discriminate between sequences in the transition state (6, 19–21), but others do not (4, 5, 17, 18). These behaviors may be due to different biological requirements for fast binding and kinetic stability of each complex. Our results contribute to the search for a consensus view of sequence readout in the kinetics of protein binding to double- and single-stranded nucleic acids (40, 41), equivalent to recent progress on the understanding of protein–protein association (35, 42).

Methods

The association and dissociation kinetics of the HPV16-E2C wild-type and mutant complexes were measured as described previously (10). Our buffer conditions were 20 mM bis-Tris·HCl, pH 7.0, 1 mM DTT, 10 mM MgCl₂ and 50 mM NaCl at 298 K, as for the equilibrium binding experiments with the mutants (15). See SI Text for details. All calculations were performed by using ProFit (Quantumsoft). The structure representations were prepared with Pymol (43).