Direct substitution and assisted dissociation pathways for turning off transcription by a MerR-family metalloregulator

Abstract

Metalloregulators regulate transcription in response to metal ions. Many studies have provided insights into how transcription is activated upon metal binding by MerR-family metalloregulators. In contrast, how transcription is turned off after activation is unclear. Turning off transcription promptly is important, however, as the cells would not want to continue expressing metal resistance genes and thus waste energy after metal stress is relieved. Using single-molecule FRET measurements we studied the dynamic interactions of the copper efflux regulator (CueR), a Cu⁺-responsive MerR-family metalloregulator, with DNA. Besides quantifying its DNA binding and unbinding kinetics, we discovered that CueR spontaneously flips its binding orientation at the recognition site. CueR also has two different binding modes, corresponding to interactions with specific and nonspecific DNA sequences, which would facilitate recognition localization. Most strikingly, a CueR molecule coming from solution can directly substitute for a DNA-bound CueR or assist the dissociation of the incumbent CueR, both of which are unique examples for any DNA-binding protein. The kinetics of the direct protein substitution and assisted dissociation reactions indicate that these two unique processes can provide efficient pathways to replace a DNA-bound holo-CueR with apo-CueR, thus turning off transcription promptly and facilely.

Bacteria often dwell in environments with high concentrations of metals. Some of these metals are essential, but many are toxic. Even the essential metals, for example iron and copper, can become detrimental above a certain concentration inside cells. Many biological processes are thus present to regulate and maintain intracellular metal homeostasis (1–9). One of them is through metalloregulators, which respond to metal ions and regulate the transcription of genes that protect the bacteria from metal-induced stress (5–7, 10). The MerR-family metalloregulators respond to many metal ions with high selectivity and sensitivity, such as Hg²⁺ and Cu²⁺ (5, 11, 12).

All MerR-family metalloregulators are homodimeric proteins. They regulate transcription via a DNA distortion mechanism (5, 13–16). They recognize specific dyad-symmetric DNA sequences within a promoter, and both their apo and holo forms bind DNA tightly. In the absence of metal, the metalloregulator bends the DNA; in this configuration RNA polymerase (RNAp) cannot interact with both -10 and -35 sequences properly and transcription is repressed. Upon binding metal, the metalloregulator changes its conformation and further unwinds the DNA slightly to allow proper RNAp interactions with the -10 and -35 sequences; transcription is then activated.

Although the mechanisms of transcription activation by MerR-family metalloregulators are well-studied (5, 13–16), little is yet known about how transcription activation is reversed. Turning off transcription promptly is important, however, as the cells would not want to continue expressing metal resistance genes and thus waste energy after metal stress is relieved. Metal dissociation to convert a holo-metalloregulator to its apo form would be the simplest way to turn off transcription, but is unlikely as the metal is bound tightly (often by cysteine ligands) and metal−cysteine bond dissociation is slow (17). For example, CueR, the Cu⁺-responsive MerR-family metalloregulator in Escherichia coli, has a Cu⁺ binding affinity of approximately 10^-21 M (18). Although thiol ligand exchange can possibly facilitate Cu⁺ removal from the binding site as observed for copper chaperones (19), no evidence exists that CueR can undergo similarly facile ligand exchange reactions. Then a holo-metalloregulator has to be replaced somehow by its apo-protein to turn off transcription. Here the simplest scenario would be for the holo-protein to unbind from DNA, followed by the binding of an apo-protein. What then are the timescales of the protein unbinding and the subsequent binding? Are there any alternative, and more efficient, pathways to turn off transcription?

Here we use single-molecule FRET (smFRET) measurements (20, 21) to address the above questions. We focus on CueR, which regulates the transcription of CopA, a membrane transporter that pumps Cu⁺ out of the cytoplasm, and CueO, a periplasmic multicopper oxidase that facilitates copper removal from the cell (6, 22). By examining CueR–DNA interactions at the single-molecule level, we found that apo-CueR could turn off transcription via a direct protein substitution pathway and an assisted protein dissociation pathway; both pathways are unprecedented and more efficient than the generic unbinding-plus-binding pathway. Moreover, CueR can spontaneously flip while staying bound on DNA and it has two different binding modes, both of which are advantageous for its regulatory function.

Results

Experimental Strategy.

We used smFRET to probe CueR interactions with a 25-bp dsDNA (SI Appendix, Fig. S1A). One end of this DNA had a FRET donor, Cy3, and the other end on a different strand had a biotin for surface immobilization, ensuring that only dsDNA was immobilized and fluorescently monitored in the experiments. The homodimeric CueR was labeled with a single FRET acceptor, Cy5, and supplied in a flowing solution. Upon CueR binding to DNA, FRET occurs, changing the Cy3 and Cy5 fluorescence intensities, which report the CueR–DNA interactions.

We used two types of DNA. One contains the dyad-symmetric sequence recognized by CueR and is referred to as the (specific) DNA; this 25-bp sequence also spans the entire footprint of CueR at the promoter region (23). The other does not contain the CueR-recognition sequence and is referred to as the nonspecific DNA. We made two CueR variants to place the Cy5 label (SI Appendix, Fig. S2A). In one variant, Cy5 is attached to C129 within the metal-binding domain of one monomer (i.e., CueR_Cy5-C129). The other has the Cy5 attached to E96C located at the α-helix dimerization domain (i.e., CueR_Cy5-E96C). In vitro transcription assay confirmed that both variants were as active as the wild-type CueR in activating transcription controlled by the copA promoter (SI Appendix, Fig. S1B and C).

Two Protein Binding Orientations on DNA.

Fig. 1A shows the E_FRET trajectory for a surface-immobilized DNA molecule interacting with holo-CueR_Cy5-C129. Three E_FRET states are clear at E₀ ∼ 0.07, E₁ ∼ 0.25, and E₂ ∼ 0.92. Correspondingly, the histogram of the E_FRET trajectories shows three peaks centered at E₀, E₁, and E₂ (Fig. 1F); the relative peak areas reflect the relative stabilities of the chemical species associated with the respective E_FRET states.

Fig. 1.

Dynamic CueR–DNA interactions. (A) Single-molecule E_FRET trajectory of an immobilized Cy3-DNA interacting with holo-CueR_Cy5-C129 (2 nM). The cartoons on the right show CueR_Cy5-C129 in two binding orientations. The FRET donor (green sphere) and acceptor (red sphere) are drawn on the DNA and protein at their approximate locations. The gray line is original data; the red line is after nonlinear filtering. (B) Same as A, but with 2 nM holo-CueR_Cy5-E96C. (C) Same as A, but using the nonspecific DNA and 4 nM holo-CueR_Cy5-C129. Here, τ₀ is denoted as τ_off, and τ₁ and τ₂ are denoted together as τ_on. (D) Same as A, but with a mixture of holo-CueR_Cy5-C129 and holo-CueR_Cy5-E96C of 5 nM each. (E) Same as A, but with a mixture of apo-CueR_Cy5-C129 and holo-CueR_Cy5-E96C of 5 nM each. The blue arrows denote the transitions from the holo-protein bound states to the apo-protein bound states, and the black arrows denote the reverse transitions. (F) Histogram of E_FRET trajectories of holo-CueR_Cy5-C129–DNA interactions at [holo-CueR_Cy5-C129] = 2 nM. Compiled from > 500 E_FRET trajectories. Solid lines are fits of Voigt functions centered at about 0.07, 0.25, and 0.92, with percentage peak areas of 90.5 ± 0.1%, 4.9 ± 0.3%, and 4.6 ± 0.3%, respectively. Bin size, 0.01. (G) Same as F, but with 2 nM holo-CueR_Cy5-E69C. The three peaks center at about 0.07, 0.42, and 0.62, with percentage peak areas of 89.5 ± 0.1%, 5.5 ± 0.3%, and 5.0 ± 0.3%, respectively. (H) [CueR_Cy5-C129] dependence of the (E₂ + E₁)/E₀ peak area ratios obtained from data such as in F. The solid lines are fits with Eqs. 8 (holo) and 9 (apo) (see also SI Appendix, section S15).

The E₀ state has the lowest E_FRET value, which indicates the farthest FRET donor–acceptor distance; it corresponds to a free DNA without a bound protein (i.e., the unbound state). Its assignment was verified by a control experiment where no protein was present. The dominance of the E₀ peak in the E_FRET histogram comes from the low concentrations of the protein (≤ 10 nM), where the DNA stays mostly at the unbound state.

The E₁ and E₂ states, whose larger values indicate FRET donor–acceptor proximities, can be assigned to the two orientations of holo-CueR_Cy5-C129 bound to DNA. CueR is a homodimer; the Cy5 label at one monomer makes the protein asymmetric. The distance between the Cy5 on the protein and the Cy3 on DNA is closer in one binding orientation than in the other, resulting in two different E_FRET values (i.e., E₂ and E₁) (Fig. 1A, Right, cartoons). Consistently, the E₁ and E₂ peaks have about the same area in the E_FRET histogram (Fig. 1F), as the two protein binding orientations are chemically equivalent except for the labels. Moreover, with increasing holo-CueR_Cy5-C129 concentration (i.e., [holo-CueR_Cy5-C129]), the E₁ and E₂ peak areas in the E_FRET histogram increase relative to that of E₀ (Fig. 1H), consistent with that increasing protein concentration should lead to higher populations of protein–DNA complexes. Similarly, using apo-CueR_Cy5-C129, three E_FRET states were observed, corresponding to the unbound and the two protein-bound states with opposite binding orientations (SI Appendix, Fig. S4A and B).

To confirm the assignments of the E₁ and E₂ states, we further studied CueR_Cy5-E69C (SI Appendix, Section S3). The Cy3–Cy5 distance in its E₂ binding orientation should be longer than that for CueR_Cy5-C129, whereas that in its E₁ orientation should be shorter (Fig. 1B, Right, cartoons). Indeed, for CueR_Cy5-E96C–DNA interactions, E₂ decreased to approximately 0.62 and E₁ increased to approximately 0.42 (Fig. 1B and G).

When CueR_Cy5-C129 interacted with the nonspecific DNA, three E_FRET states were still observed (Fig. 1C). Therefore, CueR can always bind in two orientations regardless of the DNA sequence. The E₁ and E₂ values here are slightly shifted compared with those in interactions with the specific DNA (Fig. 1A), likely because the protein can slide more freely along the DNA in the absence of the recognition sequence.

Protein Flipping on DNA.

In the E_FRET trajectories of both holo- and apo-CueR–DNA interactions (Fig. 1A and SI Appendix, Fig. S4A), the E₀↔E₁ and E₀↔E₂ transitions correspond to the reversible protein–DNA binding and unbinding. Interestingly, direct E₁↔E₂ transitions also occur. These E₁↔E₂ transitions could be due to (i) protein flipping, i.e., the protein flips its binding orientation spontaneously without detaching from DNA completely, or (ii) protein unbinding followed by a rapid rebinding in a different orientation during which the unbound E₀ state was not detected because of limited experimental time resolution (approximately 50 ms). The second scenario is unlikely, however, because these E₁↔E₂ transitions still occur at protein concentrations as low as 0.5 nM, where protein rebinding is slow and the unbound state should be readily observed. Therefore, the direct E₁↔E₂ transitions indicate that a DNA-bound CueR can spontaneously flip its orientation without complete detachment from DNA.

In contrast, no direct E₁↔E₂ transitions were observed for CueR_Cy5-C129 interacting with the nonspecific DNA (Fig. 1C). This contrast indicates that protein flipping occurs only when CueR binds to the specific DNA sequence, in which CueR distorts the DNA structure as shown by past structural studies on MerR-family regulators (5, 13–15).

Single-Step Protein Binding Kinetics.

In the E_FRET trajectories (e.g., Fig. 1A), τ₀ is the microscopic dwell time on the unbound E₀ state; its statistical properties, such as its average and distribution, contain the information about the protein binding kinetics. Here, , where denotes averaging, represents the single-molecule rate of CueR binding to DNA. Expectedly, increases with increasing [holo- or apo-CueR] (Fig. 2A). Moreover, the distribution of τ₀ follows a single-exponential decay in the presence of either holo- or apo-CueR (Fig. 2B). This distribution indicates that CueR binding to DNA follows simple single-step binding kinetics, i.e., containing only one rate-limiting step (see later for the derived probability density function of τ₀, which quantitatively describes the distribution of τ₀).

Fig. 2.

Single-step protein binding kinetics. (A) [CueR] dependence of for CueR_Cy5-C129–DNA interactions. Solid lines are fits with Eq. 5 (see also SI Appendix, section S12). (B) Distributions of τ₀ for holo- and apo-CueR_Cy5-C129 interactions with DNA, both at 0.5 nM protein concentration. Data compiled from about 500 (holo) and 600 (apo) E_FRET trajectories. Bin size, 2 s. Both distributions were normalized to the first point for comparison. Solid lines are fits with a single exponential function Nγ exp(-γτ); γ(holo) = 0.038 ± 0.002 s^-1, γ(apo) = 0.031 ± 0.004 s^-1, and N is a scaling parameter.

Two Different Binding Modes of Protein.

τ₁ and τ₂ are the microscopic dwell times on the E₁ and E₂ states, the two protein-bound states of opposite binding orientations (Fig. 1A and B). They contain the information about the kinetics of protein unbinding as well as what is happening kinetically during the protein-bound states. Interestingly, the distributions of τ₁ and τ₂ for holo-CueR_Cy5-C129–DNA interactions both follow a double-exponential decay (Fig. 3A and SI Appendix, Fig. S8A). This double-exponential distribution indicates that there are two major kinetic species within each protein-binding orientation, i.e., holo-CueR has two different binding modes on DNA. The distributions of τ₁ and τ₂ are identical within experimental error (SI Appendix, Fig. S8A), consistent with the E₁ and E₂ states being equivalent. Similar double-exponential distributions of τ₁ and τ₂ were also observed for apo-CueR_Cy5-C129–DNA interactions (Fig. 3A and SI Appendix, Fig. S4C).

Fig. 3.

Two different binding modes of CueR on DNA. (A) Distributions of τ₂ for holo- and apo-CueR_Cy5-C129 interactions with specific DNA, both at 2 nM protein concentration. Data compiled from about 500 (holo) and 700 (apo) E_FRET trajectories. Bin size, 0.15 s. The two distributions were normalized to the first data point for comparison. Solid lines are fits with a sum of two exponentials: N[Aγ₁ exp(-γ₁τ) + (1 - A)γ₂ exp(-γ₂τ)]. For holo: γ₁ = 0.55 ± 0.01 s^-1, γ₂ = 6.4 ± 0.2 s^-1, and A = 0.68 ± 0.01; for apo: γ₁ = 0.35 ± 0.02 s^-1, γ₂ = 1.6 ± 0.06 s^-1, and A = 0.57 ± 0.02. (B) Distributions of τ_2→0 and τ_2→1 at [holo-CueR_Cy5-C129] = 2 nM. Solid lines are global fits with a sum of two exponentials using the same parameters for the holo data in A. Bin size, 0.15 s. (C) Distribution of τ_on for holo-CueR_Cy5-C129 interactions with the nonspecific DNA at [holo-CueR_Cy5-C129] = 7 nM. To increase statistics, τ₁ and τ₂ from about 225 E_FRET trajectories were combined here as τ_on. Bin size, 0.05 s. Solid line is a fit with Nγ exp(-γτ); γ = 12.8 ± 0.6 s^-1.

In contrast, when the nonspecific DNA was used, the distribution of the dwell times on the protein-bound states follows a single-exponential decay (Fig. 3C), indicating that only one major kinetic species is present within each protein-binding orientation. Therefore, the two different binding modes of CueR on DNA are present only when CueR recognizes the specific DNA sequence.

Past studies have shown that MerR-family regulators distort DNA structure upon binding to the specific dyad-symmetric sequence (5, 13, 14). We can thus attribute one of the two kinetic species within each binding orientation to a complex in which CueR distorts the DNA structure. For the other kinetic species, we could attribute it to a CueR–DNA complex in which the CueR binds DNA in a way as if the DNA is nonspecific; this attribution is reasonable because CueR does bind nonspecific DNA and the binding mode here must be different from that of the specific DNA binding mode. No structural information is yet available about CueR, or any MerR-family regulators, in complex with a nonspecific DNA. Just for illustrative purposes, we drew the cartoon of this complex having the DNA structure undistorted (Fig. 1C, Right).

Moreover, τ₂ can be divided into two subtypes: τ_2→1 and τ_2→0, depending on whether an E₂ → E₁ or E₂ → E₀ transition concludes a τ₂ period in the E_FRET trajectories. For holo-CueR_Cy5-C129–DNA interactions, τ_2→1 and τ_2→0 follow the same double-exponential distribution with identical exponents (Fig. 3B), indicating that upon leaving the E₂ state, transitions to the E₁ or E₀ state must start from the same kinetic species of the two binding modes within the E₂ state. Similar behaviors and conclusions also apply by analyzing τ_1→2 versus τ_1→0 (SI Appendix, Fig. S8D), as well as for apo-CueR_Cy5-C129–DNA interactions (SI Appendix, Fig. S4E and F).

Direct Protein Substitution and Assisted Protein Dissociation.

For the microscopic dwell time τ₂ (or the equivalent τ₁, SI Appendix, Fig. S5), its average represents the lifetime of the protein-bound state E₂. And, represents the rate of leaving the state E₂. Surprisingly, increases linearly with increasing [holo- or apo-CueR] (Fig. 4A). This dependence indicates that a protein molecule coming from the solution must disrupt the protein–DNA complex, thus shortening its lifetime.

Fig. 4.

Direct protein substitution and assisted dissociation. (A) [CueR] dependence of for CueR_Cy5-C129–DNA interactions. Solid lines are fits with Eqs. 6 (holo) and 7 (apo). (B) The [holo-CueR] dependence of for holo-CueR_Cy5-C129 interactions with the nonspecific DNA; the τ₁ and τ₂ are combined here and denoted as τ_on. Solid line is a fit with a horizontal line at 5.9 ± 0.1 s^-1. Each data point in A and B is an average of data from about 250 E_FRET trajectories. (C) Dependence of N_2→1/N_2→0 on [holo-CueR_Cy5-C129]. Solid line is a fit with Eq. 3. (D) Dependence of N_2→1/N_2→0 on [apo-CueR_Cy5-C129]. Solid line is a fit with Eq. 4. (Inset) Same data but the y-value is the inverse, N_2→0/N_2→1.

Leaving the E₂ state can lead to either state E₁, the other protein-bound state with the opposite protein orientation, or state E₀, the unbound state. The former would involve the incoming protein to replace the incumbent protein on DNA and form a new complex with an opposite binding orientation (50% probability), namely a direct protein substitution. The latter would involve the incoming protein to carry away the incumbent protein, namely an assisted protein dissociation.

To differentiate these two possibilities, we examined the relative transition probabilities to the E₁ and E₀ states after the protein–DNA complex leaves the E₂ state. This relative transition probabilities can be quantified by the ratio N_2→1/N_2→0, where N_2→1 is the number of E₂ → E₁ transitions and N_2→0 is that of E₂ → E₀ transitions observed in the E_FRET trajectories. If the direct protein substitution dominates, N_2→1/N_2→0 would increase with increasing protein concentration because more E₂ → E₁ transitions are expected at higher protein concentrations. If the assisted protein dissociation dominates, N_2→1/N_2→0 would decrease with increasing protein concentration.

Strikingly, both behaviors were observed, but it depends on the metallation state of CueR. For holo-CueR–DNA interactions, N_2→1/N_2→0 increases with increasing protein concentration, whereas for apo-CueR, the opposite was observed (Fig. 4C and D). Therefore, both the direct protein substitution and the assisted protein dissociation operate when a CueR molecule interacts with a CueR–DNA complex. To our knowledge, these are unique examples of direct protein substitution and assisted protein dissociation known to any DNA-binding proteins.

To verify the direct protein substitution process by holo-CueR, we performed an experiment using a mixture of holo-CueR_Cy5-C129 and holo-CueR_Cy5-E96C. Direct transitions were indeed observed in the E_FRET trajectories between the states of holo-CueR_Cy5-C129–DNA complexes (E₁ ∼ 0.25, E₂ ∼ 0.92) and those of holo-CueR_Cy5-E96C–DNA complexes (E₁ ∼ 0.42, E₂ ∼ 0.62), reporting the exchange of proteins on DNA (Fig. 1D and SI Appendix, Fig. S6).

Note that the direct protein substitution can lead to E₁↔E₂ transitions in the E_FRET trajectories. These protein-substitution-caused E₁↔E₂ transitions do not preclude the presence of spontaneous protein flipping, however, because the E₁↔E₂ transitions still occur at very low protein concentrations, where protein substitution is negligible.

As a control, and in contrast, when the nonspecific DNA was used, the lifetime of the CueR–DNA complex is independent of the protein concentration, indicating no occurrence of either the direct protein substitution or the assisted protein dissociation (Fig. 4B). Therefore, these two processes only occur to a CueR–DNA complex in which CueR recognizes the specific sequence and distorts the DNA structure.

Interaction Mechanism and Quantitative Kinetics.

Combining the information determined above, we can formulate a minimal kinetic model of CueR–DNA interactions that contains the following processes (Fig. 5): (i) CueR binds to DNA reversibly (i.e., k₁ and k_-1) with single-step binding kinetics; its two binding orientations are distinguishable by asymmetrically labeling the protein; and the DNA that contains the recognition sequence can have a distorted structure in the complex (i.e., I₁ and I₂). (ii) CueR can flip its orientation spontaneously on DNA (i.e., k₄); this flipping occurs only when CueR recognizes the specific sequence, where the DNA structure is distorted. (iii) In each binding orientation, CueR interacts with DNA in two different binding modes, which interconvert (i.e., k₃ and k_-3) and are observed only when CueR recognizes the specific DNA sequence. Besides the form in which DNA structure is distorted, we propose that the other form is similar to that when CueR binds to a nonspecific DNA (i.e., and ). (iv) The CueR–DNA complex (I₁ or I₂) can undergo direct substitution, where a holo-CueR molecule from the solution replaces a bound holo-CueR protein directly (i.e., k_2a), or undergo assisted dissociation, where an apo-CueR molecule from the solution helps carry away a bound apo-CueR molecule (i.e., k_2b). For the direction substitution, the replacing protein has a 50% probability of ending up with the opposite orientation to the incumbent one. Either the direct substitution or the assisted dissociation occurs only when CueR recognizes the specific DNA sequence. And, all the transitions that depart from the E₂ (or E₁) state start from the same species I₂ (or I₁). Control experiments using vesicle trapping of a CueR–DNA pair further showed that the monomer-dimer equilibrium of CueR (if it exists) and the quartz surface effect were insignificant in considering the CueR–DNA interaction kinetics (SI Appendix, section S8). By removing the artificial differentiation of the two orientations by the FRET labels, this mechanistic model simplifies to SI Appendix, Fig. S12, in which only three distinct species exist: the free DNA, and the two CueR–DNA complexes of the two different binding modes.

Fig. 5.

Mechanistic model of CueR–DNA interactions. Here, k’s are the kinetic constants and described in the text. P denotes protein.

Using the kinetic mechanism in Fig. 5 we can derive the probability density functions, f, for the dwell times τ₀, τ₁, and τ₂ (SI Appendix, sections S11 and S12); these f’s quantitatively describe the normalized distributions of the following dwell times:

[1]

[2]

Here [P] stands for protein concentration; M, N, C, and D are all functions of [P] and the k’s defined in Fig. 5. Here, f₀(τ) is a single-exponential distribution function, as observed for the distribution of τ₀ (Fig. 2B), and reflecting the single-step binding kinetics. The equivalent f₂(τ) and f₁(τ) are double-exponential distribution functions, consistent with experimental observations (Fig. 3A and SI Appendix, Fig. S4D) and reflecting the two kinetic species in both the E₂ and E₁ states associated with CueR’s two different binding modes on DNA. Note that for f₂(τ) and f₁(τ), either k_2a or k_2b is close to zero, depending on whether apo-CueR or holo-CueR is concerned.

Similarly, the probability density functions for τ_2→0 and τ_2→1 are predicted to also follow double-exponential distributions with identical exponents (SI Appendix, Eq. S7 and S8), as observed in Fig. 3B. From them we can derive N_2→1/N_2→0, the ratio between the numbers of E₂ → E₁ and E₂ → E₀ transitions in the E_FRET trajectories:

[3]

[4]

Eq. 3 predicts that N_2→1/N_2→0 will increase linearly with increasing [P] when k_2b = 0, consistent with Fig. 4C and supporting that the direct substitution process dominates for holo-CueR–DNA interactions (i.e., k_2a≫k_2b ≈ 0). On the other hand, Eq. 4 predicts that N_2→1/N_2→0 will decrease with increasing [P] when k_2a = 0, consistent with Fig. 4D and supporting that the assisted protein dissociation dominates for apo-CueR–DNA interactions (i.e., k_2b≫k_2a ≈ 0).

From the probability density functions it follows:

[5]

[6]

[7]

where K_3D = k₃/k_-3. Eq. 5 gives that , which represents the protein binding rate, equals k₁[P], as expected and observed (Fig. 2A). Regarding and , Eq. 6 applies to holo-CueR–DNA interactions, and Eq. 7 to apo-CueR–DNA interactions; both are linear functions of [P], as observed (Fig. 4A and SI Appendix, Fig. S5). This kinetic model also gives the ratio of peak areas in the E_FRET histograms (Fig. 1H; SI Appendix, section S15):

[8]

[9]

Using the above equations to fit data, we determined the kinetic parameters for both holo-CueR and apo-CueR interactions with the specific DNA (Table 1 and SI Appendix, section S16). Strikingly, for holo-CueR, its rate constant for direct protein substitution (k_2a ∼ 134 × 10⁶ M^-1 s^-1) is about 15 times larger than its binding rate constant (k₁ ∼ 9 × 10⁶ M^-1 s^-1), whereas for apo-CueR, its rate constant for assisted dissociation (k_2b ∼ 60 × 10⁶ M^-1 s^-1) is 10 times larger than its binding rate constant (k₁ ∼ 6 × 10⁶ M^-1 s^-1). These larger rate constants indicate that the presence of a CueR on DNA facilitates another CueR molecule in finding the recognition sequence, leading to either direct protein substitution or assisted protein dissociation, both of which have functional advantages in transcription regulation (see Discussion).

Table 1.

Kinetic parameters for CueR interactions with specific DNA

Processes	Kinetic parameters	Holo	Apo
Binding	k1( × 106 M-1 s-1)	9 ± 4	6 ± 1
Unbinding	k-1(s-1)	1.1 ± 0.3	0.4 ± 0.1
Dissociation constant	KD( = k-1/k1)(nM)	122 ± 63	67 ± 20
Interconversion	k3(s-1)	2.0 ± 0.2	0.4 ± 0.1
	k-3(s-1)	1.4 ± 0.1	0.54 ± 0.07
	K3D = k3/k-3	1.4 ± 0.2	0.74 ± 0.2
Substitution	k2a( × 106 M-1 s-1)	134 ± 20	N/A
Assisted dissociation	k2b( × 106 M-1 s-1)	N/A	54 ± 22
Flipping	k4(s-1)	0.13 ± 0.05	0.21 ± 0.03

N/A, not applicable.

Discussion

Using smFRET we have quantified the dynamic interactions of CueR with both specific and nonspecific DNA. We found that (i) both holo- and apo-CueR can flip their binding orientations spontaneously on DNA, a phenomenon that has also been observed for other DNA binding proteins (24); (ii) both holo- and apo-CueR have two different binding modes on DNA: one in which they recognize the specific sequence and distort the DNA structure and the other likely mimicking their interactions with nonspecific DNA; and (iii) when bound to DNA, holo-CueR can undergo direct substitution, whereas apo-CueR can undergo assisted dissociation, both by another protein molecule from the surrounding solution. Similar behaviors were also observed when a longer DNA sequence containing the entire promoter region was used (SI Appendix, section S19). To our knowledge, both the direct substitution and the assisted dissociation are unique examples known for any transcription factors as well as for DNA-binding proteins in general. Moreover, all these processes (i.e., flipping, two different binding modes, direct substitution, and assisted dissociation) occur only when CueR is interacting with a DNA that contains the recognition sequence.

All above features of CueR can provide advantages for its regulatory function. Being able to flip spontaneously on DNA indicates that the bound CueR is highly dynamic. This dynamic nature, especially for holo-CueR, may facilitate transcription initiation, which involves large structural rearrangements of associated proteins and DNA (16). The two different binding modes, in which CueR interacts with DNA, specifically or nonspecifically, are beneficial for CueR in its searching for the recognition sequence in the large bacterial chromosome [approximately 4.6 million base pairs for E. coli (25)]. Being able to interact with DNA nonspecifically would help CueR sliding along the chromosome; and upon locating the recognition sequence, the CueR–DNA complex can interconvert to the specific binding mode, thereby distorting the DNA structure to regulate transcription. This sliding along DNA via nonspecific interactions has long been realized to be advantageous for reducing the dimensionality of site search for DNA-binding proteins (26).

Both the direct protein substitution and the assisted dissociation can provide efficient pathways for CueR to turn off transcription after activation. As Cu⁺ dissociation from holo-CueR is unlikely due to its tight binding (18), holo-CueR, which is the activator, needs to be replaced by apo-CueR, which is the repressor, to turn off transcription. A generic way would be for holo-CueR to unbind from DNA first, followed by the binding of an apo-CueR, which would be the dominant form of the protein inside a cell after the activation of copper resistance genes. From the rate constant k_-1 (1.1 s^-1; Table 1), holo-CueR unbinding takes about 0.9 s (Fig. 6, step i). The timescale of the subsequent apo-CueR binding depends on the intracellular protein concentration. Depending on growth conditions, an E. coli cell has about 320–400 copies of CueR (27), corresponding to [P] ≈ 355–440 nM with a cell volume of approximately 1.5 fL (18). With the rate constant k₁ ∼ 6 × 10⁶ M^-1 s^-1 (Table 1), apo-CueR binding to DNA takes about 0.4–0.5 s (= 1/(k₁[P]); Fig. 6, step ii). Therefore, this generic pathway takes a total of approximately 1.4 s to turn off transcription.

Fig. 6.

Pathways for turning off transcription by CueR. The timescales are denoted for relevant kinetic steps, including (i) unbinding, (ii) binding, (iii) direct substitution, and (iv) assisted dissociation.

Alternatively, the bound holo-CueR can be assisted to dissociate or be directly substituted by an apo-CueR (Fig. 6, steps iii and iv). It is reasonable to assume that apo-protein can either assist a holo-protein to unbind or substitute a holo-protein, as apo- and holo-proteins do not differ significantly in most of their kinetic parameters (Table 1). The direct substitution of holo-CueR by an apo protein was indeed observed when we used a mixture of apo-CueR_Cy5-C129 and holo-CueR_Cy5-E96C (Fig. 1E, and SI Appendix, section S21). Using the rate constant k_2b (54 × 10⁶ M^-1 s^-1), the assisted dissociation takes about 0.05 s (= 1/k_2b[P]; Fig. 6, step iv), about 18 times faster than the spontaneous unbinding (about 0.9 s; Fig. 6, step i). Taking into account the subsequent apo-protein binding, the assisted dissociation would accelerate the unbinding-plus-binding pathway by approximately two times for turning off transcription. On the other hand, using the rate constant k_2a (134 × 10⁶ M^-1 s^-1; Table 1), the direct substitution takes about 0.02 s (= 1/k_2a[P]) to reach the apo-protein bound, transcription-repressed state (Fig. 6, step iii), which is about 70 times faster than the unbinding-plus-binding pathway (Fig. 6, step i and ii). Therefore, the direct substitution of holo-CueR by apo-CueR would be the most efficient pathway for turning off transcription. It is worth noting that cellular conditions are not exactly the same as in our experiments and kinetic constants may thus differ; nevertheless, our results show that both the direct protein substitution and the assisted protein dissociation are possible pathways for turning off transcription.

Past studies have shown that for the prototype Hg²⁺-responsive metalloregulator MerR, a protein called MerD might mediate the unbinding of holo-MerR from DNA for turning off transcription (28). No evidence has yet been found, however, for a MerD homologue for CueR or other MerR-family regulators. As all known MerR-family metalloregulators share the DNA distortion mechanism for turning on transcription (5, 13, 14), it is likely that many of them share a common mechanism for turning off transcription. The direct substitution pathway and assisted dissociation pathway may thus be common mechanisms for MerR-family metalloregulators to turn off transcription efficiently after transcription activation.

Materials and Methods

Materials and methods are described in SI Appendix, section S1.