Flexible Target Recognition of the Intrinsically Disordered DNA-Binding Domain of CytR Monitored by Single-Molecule Fluorescence Spectroscopy

Abstract

The intrinsically disordered DNA-binding domain of cytidine repressor (CytR-DBD) folds in the presence of target DNA and regulates the expression of multiple genes in E. coli. To explore the conformational rearrangements in the unbound state and the target recognition mechanisms of CytR-DBD, we carried out single-molecule Förster resonance energy transfer (smFRET) measurements. The smFRET data of CytR-DBD in the absence of DNA show one major and one minor population assignable to an expanded unfolded state and a compact folded state, respectively. The population of the folded state increases and decreases upon titration with salt and denaturant, respectively, in an apparent two-state manner. The peak FRET efficiencies of both the unfolded and folded states change continuously with denaturant concentration, demonstrating the intrinsic flexibility of the DNA-binding domain and the deviation from a strict two-state transition. Remarkably, the CytR-DBD exhibits a compact structure when bound to both the specific and nonspecific DNA; however, the peak FRET efficiencies of the two structures are slightly but consistently different. The observed conformational heterogeneity highlights the potential structural changes required for CytR to bind variably spaced operator sequences.

Introduction

Intrinsically disordered proteins (IDPs) and intrinsically disordered regions (IDRs) are abundant in eukaryotic and prokaryotic proteomes and play vital cellular roles.¹⁻³ Many IDRs are characterized by low sequence complexity with a high proportion of polar and charged residues that perform their functions without forming fixed tertiary structures. On the other hand, another class of IDRs possess high-complexity sequences reminiscent of foldable proteins and fold to specific structures under favorable conditions. Such IDRs are prevalent in transcription regulators,^4,5 whose conformational landscapes are characterized by a complex interplay between folding and binding, leading to distinct folding–binding mechanisms, such as conformational selection,⁶ induced folding,⁷ synergistic binding,⁸ fly casting,⁹ specific binding without folding,¹⁰ conformational switching,¹¹ and conformational funneling.¹² These mechanisms are related to the heterogeneity and the dynamics of IDRs over varied time scales,^13,14 which can be examined by various methods, including nuclear magnetic resonance (NMR) spectroscopy, small-angle X-ray scattering, molecular dynamics simulation, and especially single-molecule Förster resonance energy transfer (smFRET) spectroscopy.¹⁴⁻¹⁷

The DNA-binding domain of the cytidine repressor protein (CytR-DBD) is an example of such an IDR that is disordered in the absence of DNA but helps the protein to bind site specifically to its operators upon concomitant folding.^18,19 However, the large heterogeneity inherent to the domain confounds the relation between dynamics and cellular function. CytR belongs to the lactose repressor (LacR) family and regulates the expression of transport proteins and enzymes involved in ribonucleotide salvage and catabolism in bacteria.^20,21

CytR operators are flanked by two binding sequences for cAMP receptor protein (CRP), enabling CytR and CRP to control gene expression via multiple pathways.²³⁻²⁶ For example, the binding of one CRP dimer to one of the CRP binding sites causes the activation of its promoter by recruiting RNA polymerase. The subsequent binding of the CytR dimer to the operator represses the activation through the quaternary interaction between CytR and CRP.^27,28 The binding of cytidine to CytR reduces the affinity of CytR to the operator in the presence of CRP.²⁵

A monomer unit of CytR consists of an intrinsically disordered DNA-binding domain (residues 1–68) including the hinge region and a folded core dimerization domain (69–341) involved in cytidine binding and the recognition of CRP (Figure 1A).^29,30 While a crystal structure of the full-length CytR or its DNA-binding domain is not available, a solution NMR study demonstrated that CytR-DBD is largely unstructured in the absence of DNA but forms a canonical helix–turn–helix motif (Figure 1B; composed of helices I, II, and III) upon the association to a short specific DNA.¹⁸ The N-terminal residues before helix I and the C-terminal residues after helix III are disordered even in the DNA-bound form. In the case of three other structurally well-known LacR family members (LacR,³¹ PurR,³² and FruR³³), helices I to III form a stable helix–turn–helix structure without DNA. Thus, the folding of the disordered N-terminal domain upon binding to the operator DNA is unique to CytR and is assumed to trigger a conformational change in the C-terminal domain, leading to the control of gene expression.^34,35

Figure 1.

(A) CytR consists of an intrinsically disordered DNA-binding domain (DBD) including the hinge region and a folded core dimerization domain. The amino acid sequence of DBD is shown with positively charged residues colored blue and negatively charged residues colored red. Orange boxes indicate the first three helices, and the blue shaded box indicates the putative region that might form the helix IV suggested from the multiple sequence alignment of the hinge regions of PurR, LacR, and CytR.²² The current sample corresponds to residues 2–65 flanked by two cysteines that are introduced for the labeling with fluorophores. (B) Solution NMR structure of CytR-DBD in the presence of cognate DNA (PDB ID: 2L8N). The terminal regions before helix I and after helix III including helix IV are unstructured in the NMR model.

CytR preferentially binds to the operators consisting of two specific octamer repeats arranged in direct or inverted orientations separated by a spacer with varying lengths from 0 to 11 bp.^36,37 The affinity of CytR-DBD to the target DNA is low (K_d ≈ 10 μM) and is only marginally stronger than that to the nonspecific DNA.¹⁹ In addition, CytR-DBD is highly frustrated electrostatically, binds promiscuously to multiple sites, and binds with the cognate DNA in a heterogeneous manner as exemplified by a negative cooperativity in the binding isotherm.³⁸⁻⁴¹ These observations suggest that the binding of CytR-DBD to DNA might be rather nonspecific and flexible. However, CytR-DBD needs to differentiate the specific and nonspecific DNA sequences to enable the C-terminal domain to control the gene expression via interactions with CRP. For many DNA-binding proteins, the differences in structure on binding specific and nonspecific sequences are subtle.⁴²⁻⁴⁴ The structural malleability and complex folding–binding modes found in IDPs operating as conformational rheostats may require the small or zero thermodynamic barriers, i.e., downhill-like protein (un)folding or reorganization.^7,19 Thus, understanding how the folding of CytR-DBD fine-tunes its conformational heterogeneity to balance binding specificity and affinity toward its target sequences is still an open question.

The large-scale structural reordering of CytR-DBD has been investigated by various spectroscopic and theoretical methods. CytR-DBD, while being predominantly disordered, still contains a significant secondary structure, the relative extents of which change gradually depending on solution conditions.^19,39,45,41 The compact structure reminiscent of the DNA bound form populates as a minor state even in the absence of DNA,^40,45 whose relative population increases upon an increase in the solution ionic strength or crowder density.^19,46 The heat capacity profile of the disordered CytR-DBD at the low ionic strengths displays a gradual and broad transition; however, at high salt concentrations, an apparent two-state transition with an excess heat capacity peak is observed.¹⁹

In this work, we have explored the unfolding and the target recognition mechanisms of CytR-DBD using a variant of smFRET spectroscopy, the alternating laser excitation (ALEX) measurements.⁴⁷ Trends in the smFRET data of CytR-DBD in solutions containing different concentrations of salt, denaturant, and the specific and nonspecific DNAs were examined. We find that CytR-DBD populates two distinct substates at equilibrium in slow exchange: one minor population which is compact folded-like and another mostly unfolded. While the unfolded population changes its structure gradually, the interconversion between the compact and extended conformations occurs in an apparent two-state manner. In addition, the binding of CytR-DBD to both specific and nonspecific DNA sequences is demonstrated to cause compaction of the domain with discernible structural differences.

Materials and Methods

Protein Expression and Purification

To study CytR-DBD by smFRET measurements, a double cysteine variant of CytR-DBD including the hinge region sequence was constructed, which has one cysteine inserted between Met1 and Lys2 and another cysteine at the C-terminus (E66C). Since the first methionine is excised in E. coli, the sample corresponds to residues 2–65 flanked by two cysteines enabling the labeling with fluorescent probes as we explain below. The protein expression and purification protocol have been reported previously.⁴¹ The gene for the double cysteine mutant of CytR-DBD was cloned into the pTXB1 vector containing chitin-binding domain (IMPACT, New England Biolabs) and transformed into E. coli, BL21(DE3). The cells were grown at 37 °C until the optical density of the culture reached ∼1.2 and were further grown for 2 h after adding 2 mM isopropyl β-d-1-thiogalactopyranoside (IPTG). The cells were harvested, lysed, and centrifuged, and the supernatant was passed through a chitin resin column (New England BioLabs) and washed with 200 mM sodium phosphate buffer at pH 8.0. The cleavage reaction was initiated by equilibrating the column in the same buffer with 50 mM β-mercaptoethanol and incubated for 12–14 h. The cleaved protein was eluted from the column using 50 mM phosphate buffer at pH 8.0, passed through a cation exchange column (BioRad, MiniPrep HighS), and finally purified through a size exclusion column (HiLoad 16/600 Superdex 75pg, GE Healthcare) equilibrated with 150 mM ammonium acetate buffer at pH 8.0. The fractions containing purified CytR-DBD were lyophilized. If necessary, the buffers were supplemented with 1 mM DTT.

Fluorophore Labeling of CytR-DBD for smFRET Measurements

The labeling protocol was designed according to the method described by Gois et al.⁴⁸ [Yoshida et al., manuscript in preparation]. The lyophilized protein was dissolved in 20 mM ammonium acetate buffer at pH 7 to obtain a protein concentration of approximately 150 μM. To reduce possible disulfide bonds, a 10-fold excess of tris(2-carboxyethyl)phosphine (TCEP) was added and the mixture was incubated for 2 h. To protect the N-terminal cysteine, 20 equiv of 2-formylphenylboronic acid in DMF at 10 mM was added, and the mixture was incubated for 1 h. The C-terminal cysteine was labeled with Alexa Fluor 488 (Alexa488) C₅ maleimide (Invitrogen) by adding its DMSO solution at 16 mM at a molar ratio of fluorophore to protein of 1:1 for 4 h at room temperature. To unblock the N-terminal cysteine, a hundred molar excess of O-benzylhydroxylamine hydrochloride in DMF at 10 mM concentration was added and the mixture was incubated for 1 h. Three times molar excess of Alexa Fluor 647 (Alexa647) C₂ maleimide (Invitrogen) in DMSO at 20 mM concentration was added, and the sample was kept overnight at 4 °C. The unreacted fluorophores were removed from the solution by a PD-10 desalting column (GE Healthcare) equilibrated with 50 mM sodium phosphate buffer at pH 8 and then concentrated by ultrafiltration (Amicon Ultra 4, Merck Millipore). The properly labeled protein was separated from the other singly and doubly labeled components by using cation exchange chromatography on an ÄKTA purifier system (GE healthcare) equipped with a TSKgel BioAssist S column (TOSOH Bioscience). The collected colored fractions were examined by MALDI-TOF MS, and fractions containing the properly double-labeled protein were collected and buffer exchanged to 20 mM sodium phosphate buffer at pH 7. The concentrated sample was then snap-frozen and stored at −80 °C. The ion-exchange chromatogram and mass spectroscopic data of the separated samples are presented in the Supporting Information.

Apparatus for the ALEX Measurements

All single-molecule fluorescence experiments were performed in a home-built confocal microscope. The scheme explaining the optical setup of the system is presented in the Supporting Information. The system is equipped with two diode lasers, LDB-160-488 (Tama Electric) whose lasing wavelength is 486 nm (for the excitation of donor fluorophore) and LDB-160-639 (Tama Electric) whose lasing wavelength is 638 nm (for the excitation of the acceptor fluorophore). After the ellipsoidal beam of the donor laser was reshaped using an anamorphic prism pair (PS883-A, Thorlabs), the laser lights were combined by a dichroic mirror (Di02-R514, Semrock) and were introduced into a polarization-maintaining single-mode fiber. The output from the fiber was collimated by an achromatic collimator (C80FC-A, Thorlabs), passed through an iris (SM2D25, Thorlabs), and reflected by a wedge plate. The wedge plate enabled the introduction of the excitation beam into the light path leading to the objective without using a dichroic mirror. The excitation beam was reflected vertically, introduced into a water immersion objective (NA 1.2, CFI PlanApo 60XC WI, Nikon) in the inverted setup, and focused into the sample solution (∼500 μL) placed on the glass-based dish (thickness of 0.15–0.18 mm, Iwaki). Fluorescence photons from the sample molecules were collected by the same objective, passed through the wedge plate and a dual bandpass filter (ZET488/640m, Chroma Technology), and were separated into donor and acceptor fluorescence by a dichroic mirror (593 nm cut-on wavelength, #67-083, Edmund Optics). The separated donor and acceptor fluorescence photons were passed through bandpass filters (for donor fluorescence, FBH520-40, Thorlabs, and for acceptor fluorescence, FF01-676/29, Semrock) and were imaged with two achromatic lenses (AC254-200-A-ML, Thorlabs) onto multimode optical fibers with a core diameter of 100 μm that served as a confocal pinhole. The fibers were coupled to two single-photon avalanche diodes (SPAD, SPCM AQRH-14-FC, Excelitas). Signals from SPADs were recorded as photon arrival times by using a counter (PCIe-6612, National Instruments) operated by home-built LabVIEW software (National Instruments).

For all the ALEX measurements, the excitation lasers were modulated with square waves at a frequency of 50 kHz and a duty ratio of 0.5. The phase of the square wave for the acceptor excitation was inverted with that for the donor excitation. The excitation intensities were ∼60 μW for the donor and ∼100 μW for the acceptor at the back aperture of the objective. The instruments were controlled by the homemade software programed with LabVIEW (National Instruments).

ALEX Measurements and Data Analysis

A double-labeled CytR-DBD stock solution of 3–4 μM in 50 mM sodium phosphate at pH 7 was diluted to a final concentration of 200–300 pM in the single-molecule fluorescence buffer (SMF buffer) containing 34 mM sodium phosphate and 47 mM NaCl adjusted at pH 6.0. The ionic strength of the SMF buffer was 100 mM. The ionic strength experiments were performed by adjusting the NaCl concentration in the SMF buffer. For the DNA-binding experiments, the duplex DNAs purchased from IDT (Integrated DNA Technologies) were dissolved in the stock buffer containing 10 mM Tris-HCl and 1 mM EDTA at pH 7.0 at 100 μM concentration. The stock solutions of double-labeled CytR-DBD and DNA were diluted in the SMF buffer at an appropriate ratio immediately before the ALEX measurements. All measurements were performed at ∼23 °C. To prevent charge-driven surface adhesion of the positively charged proteins, the surfaces of the glass-based dishes were coated with MPC polymer. Briefly, glass-based dishes were first cleaned with a solution containing 10% H₂O₂ and 10% NH₃ for 10 min, washed using distilled water, and dried by flowing compressed gas. Next, 20 μL/cm² of 0.1% MPC polymer (Lipidure-CM5206; NOF) solution was placed on the glass-based dish and dried at room temperature. ALEX-smFRET measurements were conducted by collecting fluorescence signals typically for 30–60 min to detect a sufficient number of bursts.

The arrival times of all the detected photons by the donor and acceptor channels were stored on the open-source photon-HDF5 file format⁴⁹ and then analyzed using the FRETBursts software.⁵⁰ First, the local background rate of each photon type (total of four types, both donor and acceptor photons during donor and acceptor excitations) was calculated in windows of 30 s. Typical background rates for the donor and acceptor photons during the donor excitation and acceptor photons during the acceptor excitation were ∼0.8, ∼0.3, and ∼0.8 kcps, respectively. Second, the fluorescence bursts were searched for each of the channels by selecting the time regions where the local count rate calculated using arrival times of 10 consecutive photons exceeded the count rate threshold which was set to equal to 6 times the local background rate. Third, stronger bursts were selected from the searched bursts by using the burst size threshold of 40 for the sum of donor and acceptor photons obtained during the donor excitation period. This procedure eliminates the acceptor-only species efficiently. Fourth, another round of burst selection was performed by setting the burst size threshold of 40 for the acceptor photons collected during the acceptor excitation period to eliminate the donor-only bursts efficiently. The selected bursts were recorded into the data matrix describing the duration and the photon numbers of each burst, f_Dex^D_em, f_Dex, f_Aex^A_em, and f_Aex, where f_X^Y denotes the photon numbers of a burst detected in the Y-channel with the excitation of fluorophore X.

The raw burst data were corrected in multiple steps to build an accurate FRET efficiency histogram considering three correction parameters: γ accounting for the differences in fluorescence quantum yields and photon detection efficiencies of donor and acceptor; α and δ accounting for the donor leakage into the acceptor channel (Lk = αf_Dex^D_em) and acceptor direct excitation by the donor excitation laser (Dir = δf_Aex), respectively. From the photon counts in each channel, apparent FRET efficiency (E_app) and stoichiometry (S_app) for each burst were calculated as eqs 1 and 2:

1

2

where F^FRET was characterized as eq 3:

3

The corrected FRET efficiency E was calculated based on eq 4 by using the previously reported method:⁵¹

4

where ϕ_A and ϕ_D are fluorescence quantum yields and η_A and η_D are the efficiencies of fluorescence detection for the acceptor and donor, respectively. The fluorescence quantum yields provided by the manufacturer of the fluorophores were used for the calculation without correcting for possible changes in different buffer solutions. The detection efficiencies were estimated numerically by calculating eq 5:

5

where the subscript i stands for donor (d) or acceptor (a), T_i=a,d(λ) is the transmittance of the optical filter on the acceptor or donor detection path at wavelength λ, q(λ) is the quantum efficiency of the detector at wavelength λ, and f_i=a,d(λ) is the normalized fluorescence spectrum of the acceptor or donor fluorophore. The γ value estimated for the current system and the FRET pair Alexa488/Alexa647 is 0.48. The typical values of α and δ were 0.008 and 0.004, respectively.

To estimate the peak FRET efficiencies and populations of the folded and unfolded states, the FRET histograms were fitted with a combination of two Gaussian distributions by setting the position of the FRET peaks and width as variable parameters, and the area under the fitted histogram curves for each subpopulation was determined by numerical integration to calculate the folded and unfolded populations in each histogram. This analysis is the simplest approximation of FRET efficiency distributions for the evaluation of peak FRET efficiency and population and ignores the asymmetric deformation of FRET efficiency distributions due to a large deviation of γ from 1.⁵¹ Details of the fitting procedures and the fitted parameters are described in the Supporting Information.

Results

Structural Heterogeneity of CytR-DBD in the Absence of DNA

Our sample, CytR-DBD, corresponds to the residues from 2 to 65 of CytR with additional two cysteines introduced at the N- and C-termini. In the presence of the cognate DNA sequence, residues 12–18, 23–30, and 38–51 form helices I, II, and III, respectively, resulting in a helix–turn–helix structure. Helix II is assumed to contact the major groove of the target DNA, while both the terminal regions before helix I and after helix III are unstructured.¹⁸ CytR-DBD, in the absence of DNA, samples a heterogeneous ensemble of conformations in equilibrium as exemplified by the appearance of multiple resonances for residues and by the rapid exchange of amide protons to deuterium.¹⁸ Interestingly, helices I and III exhibit partial helical structures even in the DNA-free form.¹⁸

To explore the extent of residual structure in the absence of DNA, we carried out smFRET measurement on the double-labeled CytR-DBD at pH 6.0, 0.1 M ionic strength condition (see Materials and Methods). Figure 2A plots a two-dimensional histogram of the observed bursts as a function of apparent FRET efficiency (E_app) and apparent fluorophore stoichiometry (S_app). Bursts centered near E_app = 0 and S_app = 1 and those scattered along the S_app = 0 axis, corresponding to donor-only and acceptor-only bursts, respectively (not shown), were removed from the 2-D plot by deselecting bursts having no acceptor photons during acceptor excitation and bursts having no donor photons during donor excitation (see Materials and Methods). After removing the donor-only and acceptor-only bursts, we can identify two populations of bursts: a major population centered at around E_app = 0.3 and S_app = 0.2 and a minor population at around E_app = 0.7 and S_app = 0.2. In some experiments, an additional residual population at a transfer efficiency close to zero was observed (even after the elimination of donor-only signals) due to a large signal from donor-only molecules. Considering that the power of the laser used for acceptor excitation was slightly higher than that for the donor, the two populations showing S_app centered at ∼0.2 can be assigned to the molecules containing both the donor and acceptor fluorophores. A standard FRET efficiency (E) histogram obtained by using all the burst data remaining in the two-dimensional histogram (Figure 2A) after γ correction is shown in Figure 2B. CytR-DBD at low ionic strength conditions and pH 6.0 displays a bimodal behavior, with the major population displaying a peak FRET efficiency at E = 0.47 (implying a relatively expanded structure) and a minor population having a peak FRET efficiency at E = 0.82 (indicating a collapsed structure). This observation clearly demonstrates that the DNA-free CytR-DBD samples a heterogeneous ensemble of conformations in its native ensemble.

Figure 2.

(A) Two-dimensional (2-D) histogram of the apparent FRET (E_app) and stoichiometry (S_app) of the doubly labeled CytR-DBD at pH 6.0, 0.1 M ionic strength. (B) FRET efficiencies (E) extracted from the 2-D histogram shown in panel A. Green and blue curves are the fits with two Gaussian distributions for unfolded and folded populations, respectively.

Tuning the Subpopulations by Modulating the Solution Conditions

The major and minor populations detected in the smFRET efficiency histograms can be assigned to the mostly unfolded state and the compact folded state, respectively. To confirm the assignment and to further investigate the properties of the two populations, we studied the effect of perturbations on the FRET efficiency distributions. First, we examined the effect of changing the ionic strength, since CytR-DBD in the absence of DNA is known to populate a native-like compact structure under high ionic strength conditions.¹⁹ Second, we examined the effect of denaturants on the peak FRET efficiency positions. For each condition, we obtained two-dimensional E_app–S_app histograms (not shown) similar to those presented in Figure 2A and converted them into standard FRET efficiency histograms as shown in Figure 2B.

ALEX-smFRET measurements were conducted under different salt concentrations to alter the ionic strength of the solutions (Figure 3A). It is apparent that the amplitude of the high FRET efficiency peak increases upon an increase in ionic strength, in agreement with earlier ensemble experiments.¹⁹ The high concentration of ions shields the electrostatic repulsion of the positively charged residues in CytR-DBD, enabling the formation of a compact folded structure in the absence of DNA. With an increase of ionic strength from 100 mM to 1 M, the compact folded state is preferentially stabilized, as evident from the more than 3-fold rise of its relative population compared to the unfolded state, which concomitantly decreases in population (Figure 3B).

Figure 3.

(A) Dependence of FRET efficiency (E) on ionic strength of the buffer solution. The green and blue curves are the fits with two Gaussian distributions. (B) Bar plots represent the integrated population distribution of the unfolded and folded state at different ionic strengths, following the color code in panel A. Error ranges were estimated from fits as explained in the Supporting Information.

To further confirm the assignment of the two populations, we examined the effect of denaturant by adding urea (Figure 4A). As the concentration of urea increases, the relative population of the compact form decreases, with a concomitant increase in the expanded state amplitude (Figure 4B). The observation is similar to the standard unfolding transition of two-state proteins and is consistent with the interpretation that the compact and expanded conformations detected in the FRET histograms correspond to the folded-like and unfolded conformations, respectively. In addition to the changes in the populations, the peak FRET efficiency of the expanded conformation shifts gradually to low efficiencies with increasing urea concentration (Figure 4C). The shift demonstrates an expansion of the unfolded state at higher concentrations of urea. Additionally, the peak FRET efficiency of the folded state moves to lower values on increasing urea concentrations, suggesting structural changes in the native ensemble (Figure 4C).

Figure 4.

(A) Dependence of FRET efficiency (E) on the concentration of urea. The data shown in the top panel of Figure 3A is represented as the 0 M urea condition. The green and blue curves represent the fits with two Gaussian distributions for unfolded and folded subpopulations, respectively. (B) Bar plots showing the percentage population of the unfolded and folded states at different urea concentrations. Error ranges were estimated from fits as explained in the Supporting Information. (C) Peak FRET efficiency values as a function of urea concentration. Note that there is a consistent decrease in the peak FRET efficiency for both the unfolded (green circles) and folded (blue triangles) states. Error bars for the condition in the absence of urea correspond to the standard deviation from 7 independent experiments as explained in the Supporting Information.

A comparison of the current results with circular dichroism (CD) measurements suggests that the unfolded state in the absence of urea possesses a fractional content of secondary structures. The CD spectrum of nonlabeled CytR-DBD at 100 mM ionic strength and 298 K in the absence of urea possesses a significant ellipticity at 222 nm (−7000 deg·cm²·dmol^–1).⁴⁶ This value corresponds to ∼44% of the total signal change between the fully folded state (−14200 deg·cm²·dmol^–1, detected at 2.5 M ionic strength)¹⁹ and the fully unfolded state (−1390 deg·cm²·dmol^–1, detected in the presence of 6 M urea).⁴¹ However, the folded population estimated by smFRET at 100 mM ionic strength and ∼298 K is only ∼20%. Accordingly, the CD signal cannot be explained without assuming that the unfolded population in the absence of urea possesses a fractional amount of secondary structures. Furthermore, the gradual decrease of the CD signal upon the addition of urea demonstrates that the residual secondary structure is gradually destabilized along with urea-induced expansion of the unfolded structure.⁴¹

Structural Transitions on Binding to Specific and Nonspecific DNA Sequences

To understand the target recognition mechanism of CytR, we examined the structure of CytR-DBD bound to different DNA sequences. As an example of the specific DNA, we used the left half site of the uridine phosphorylase (udp) operator (14bp) sequence (Specific 1), on which the previous NMR study was carried out (Table 1). We chose the left half-site since its 8 bp recognition site (colored red in Table 1) was a better match to the consensus recognition motif defined by SELEX than the right half-site.²⁹ To further examine the flexibility in the target recognition of CytR, we chose other specific DNAs, Specific 2 and Specific 3, having a slightly longer length (42bp). In Specific 2, the 8 bp recognition site of the udp operator was altered while keeping the central 4 bp (TGCA) sequence intact (colored red). The recognition site was flanked by mostly randomized sequences. While we initially thought the extended bases did not contain any binding sites, there was an additional sequence at the 3′ terminus that was reported to be bound by CytR (gray highlighted), as CytR binds to multiple specific sites at different promoters.²⁶ In Specific 3, the central recognition sequence in Specific 2 was altered to a nonspecific sequence but the 3′ sequence was still present (gray highlighted). Finally, we used the (AC)₇ sequence as the representative for a fully nonspecific DNA.

Table 1. DNA Sequences Examined.

^aThe 8 bp recognition site of the udp operator is colored red in Specific 1. The central 4 bp of the recognition site kept in Specific 2 is colored red. The additional binding motif of CytR is shaded gray in Specific 2 and Specific 3.

To monitor potential structural transitions in CytR-DBD upon binding to specific DNA, we conducted the ALEX experiments with an unlabeled 14bp udp operator, Specific 1 (Figure 5). Upon increasing the DNA concentration, the relative abundance of the folded fraction increases monotonically. At 3 μM DNA, a near complete shift of the occupancies from the expanded fraction to the folded fraction occurred. Here, the peak FRET efficiency of the folded component is observed at E = 0.78, which is slightly lower than the value of 0.82 extracted by the fitting of the FRET distributions in the absence of DNA (Figure 2B). This suggests that there might be a structural difference between the folded structure in the absence and presence of the specific DNA sequence.

Figure 5.

FRET histograms of labeled CytR-DBD at increasing concentrations of unlabeled 14 bp udp (Specific 1, concentrations indicated). The green and blue curves are the fits with two Gaussian distributions.

To further examine the structural changes, we performed the same measurements in the presence of other DNA sequences. The high FRET efficiency population of CytR-DBD increases in the presence of the varied DNA sequences, in agreement with similar results from ensemble binding isotherms.¹⁹ The peak FRET efficiencies of the folded population from binding to Specific 2 and Specific 3 are 0.73 and 0.78, respectively. Note that these numbers are similar to that observed for Specific 1 and lower than the values observed in the absence of DNA (Figure 6A,B). Finally, to test the binding of a nonspecific DNA, we examined 14 bp (AC)₇ that was previously reported to bind to CytR-DBD.¹⁹ Though we find that the affinity toward this 14 bp sequence is lower than the other DNAs examined (smaller relative amplitude of the high FRET peak at the same DNA concentration), an increase in the population of the folded state is detected (Figure 6A). Thus, these data provide direct and first evidence that even a random DNA sequence having no similarity with the consensus sequence can induce large structural compaction of the protein.

Figure 6.

(A) FRET histograms of CytR-DBD in the absence of DNA (first panel), presence of 3 μM 14 bp Specific 1 (second panel), 8 μM 42 bp Specific 2 (third panel), 8 μM 42 bp Specific 3 (fourth panel), and 8 μM 14 bp (AC)₇ (fifth panel). (B) Peak FRET efficiency values of the folded and unfolded states for different DNA sequences following the color code in panel A. Error bars for the condition without DNA correspond to the standard deviation from 7 independent experiments as explained in the Supporting Information.

Discussion

Structural Flexibility of the Expanded Unfolded State and the Compact Folded State

In the absence of DNA and at low ionic strength, we identify two populations having peak FRET efficiencies of 0.82 and 0.47 and assign them to the compact folded state and the expanded unfolded state, respectively. The population of the folded state is estimated to be ∼20% at ∼297 K. The folded and unfolded population increases and decreases, respectively (Figure 3B), at higher concentrations of salt without exhibiting additional FRET efficiency peaks. The two populations exhibit the opposite trend with increasing concentrations of urea (Figure 4B). The addition of specific or nonspecific DNA causes a similar increase in the folded state population. Thus, the current results are a direct experimental confirmation that the native ensemble of CytR-DBD, though mostly disordered, samples a compact folded state. In addition, our results demonstrate that the folded and unfolded states convert in an apparent two-state manner (discussed below) and that the peak FRET efficiencies of both states change with the increasing concentration of urea (Figure 4C).

The unfolded state of CytR exhibits random coil dimensions but with a fractional amount of residual secondary structure. The unfolded ensemble shows a peak FRET efficiency of 0.47 in the absence of urea. If we assume the simple Gaussian chain model for the unfolded state and the Förster distance (R₀) of 55 Å for the Alexa488–Alexa647 pair,¹⁷ we obtain a mean end-to-end distance (R_ee) of ∼66 Å, which can be converted to a radius of gyration (R_g) of ∼26 Å. If we assume the self-avoiding walk model,⁵² we obtain R_ee of ∼64 Å, which can be converted to an R_g of ∼25 Å. These values are consistent with the reported R_g of 26 Å of hydrodynamic radius measurements based on dynamic light scattering.⁴¹ The peak FRET efficiency for the unfolded state in 3 M urea is 0.33, which corresponds to R_ee of ∼81 Å and R_g of ∼33 Å in the Gauss chain model, and R_ee of ∼77 Å and R_g of ∼30 Å in the self-avoiding walk model. The increase in unfolded state dimensions at higher urea concentrations is also commonly observed in the unfolded state of many proteins, which could be attributed to the reduction of the hydrophobic effect (as water is a poor solvent).^17,53−59 In addition, the changes in chain dimensions are typically associated with the gradual loss of residual secondary structure content. We reported recently that molecular crowding induced by crowding agents like polyethylene glycol (PEG) causes gradual compaction of the unfolded state of CytR-DBD.⁴⁶ Taken together, the unfolded state of CytR-DBD is sensitive to changes in solvent conditions, which proves that the balance between various intraprotein and protein–water interactions determines the dimension of the CytR-DBD unfolded state.

Structural flexibility is also detected in the folded state of CytR-DBD. In the absence of urea, the folded state exhibits a peak FRET efficiency of 0.82, which shifts to 0.77 at 2 M urea and 0.78 at high salt concentrations. Under the assumption of a fixed distance between the two fluorophores, the efficiency of ∼0.8 corresponds to a fluorophore–fluorophore distance of ∼44 Å. The two terminal residues of the folded three-helix bundle of CytR-DBD are Met12 and Glu50, whose C_α’s are separated by only 10 Å.¹⁸ The unstructured residues flanking the folded domain should make the effective distance between the fluorophores longer. The lowering of the peak FRET efficiency in urea could therefore be ascribed to an increase in the effective length of the unstructured domain or partial unfolding of the structured domain. A similar increase in the effective length of the unstructured domains might also explain the lowering of FRET efficiency in the folded state. These results demonstrate the structural flexibility of the folded state of CytR-DBD, indicating that the free energy landscape for the folded state is indeed very malleable.

Deviation from a Strict Two-State Transition

The consistent observation of the folded and unfolded populations at different concentrations of salt, denaturant, and specific and nonspecific DNA shows an apparent two-state transition between the compact folded state and the expanded unfolded state. Each of the two populations might further be a combination of partially structured substates; however, the current results cannot resolve the possible substates likely because of their fast exchange within the burst period of a few milliseconds. Two FRET efficiency distributions, one obtained by selecting bursts with durations shorter than 6 ms and the other longer than 6 ms, were indistinguishable from each other, suggesting that the exchange of the two states occurs in the slower time scale (Figure S4). The two-state-like transition is consistent with some of the past investigations. The kinetic response of CytR-DBD measured by stopped-flow experiments after the rapid dilution of urea is characterized by a single kinetic phase highlighting an exchange between two signal-competent states with a rate constant of the order of 30–200 s^–1 between 283 and 306 K.⁴⁵ At around 295 K, the kinetic constant is ∼80 s^–1, consistent with the exchange slower than the typical duration of the burst data. The change in fluorescence signal is sensitive only to the formation of a long-range interaction between the ends of helices I and III, similar to that reported via smFRET in this study. The semiquantitative statistical modeling of the conformational landscape of CytR points to thermodynamic barriers of the order of ∼2–5 kJ mol^–1 between different states suggestive of marginal barriers, if any, in one-dimensional free energy profiles.⁴¹ Thus, CytR-DBD appears to exchange populations in an apparent two-state manner to the extent detectable from FRET efficiency measuring only the end-to-end distances.

However, there are several distinct observations that point to the deviations from a strict two-state behavior in the structural transition of CytR-DBD. First, the estimated folded population of 20% from smFRET is on the higher side as analysis of populations via the statistical mechanical model point to a folded population which is less than 10% even at 288 K.⁴⁵ The folded population at 297 K (conditions of the current study) is accordingly expected to be significantly lower. It is therefore likely that the “folded ensemble” as identified by the high FRET efficiency peak includes additional substates that are not discernible via FRET efficiency. Second, the native ensemble as identified by FRET consistently moves to lower values on the addition of urea, highlighting structural changes that cannot be readily explained by the strict two-state model. Third, all-atom simulations based on a replica exchange Monte Carlo method with implicit solvents predict a rather flat landscape for the substates of CytR-DBD with different contents of native structures,⁴¹ supporting that the unfolded population might also be an ensemble of rapidly fluctuating substates. Fourth, the salt-induced folding transition when monitored by stopped-flow fluorescence kinetics did not result in a chevron-like rate dependence, which is a necessary (but not a sufficient) condition for two-state folding.^45,60 Effectively, it is possible that FRET that maps the distance between the N- and C-terminal regions of the protein (which are close to each other in the native structure) is insensitive to structural variations in the rest of the structure (say, in the helical regions) manifesting as an apparent two-state transition when monitored along this axis. A similar observation was reported for the unfolding of a B domain of protein A (BdpA), a three-helix bundle protein, in which the gradual structural transition of the native and unfolded structure was observed.^51,61 A more nuanced answer to the question of the folding mechanism in CytR-DBD can be potentially obtained by measuring FRET between donor and acceptor positioned at different regions of the protein structure.

The absolute heat capacity profile of CytR-DBD under low salt conditions displays a broad transition with no excess heat capacity peak.⁴¹ A variable-barrier model analysis of the thermogram demonstrated that the thermal transition occurs as a barrier-less process rather than a barrier-limited transition.⁴⁶ The thermally unfolded state of CytR-DBD is highly collapsed⁴¹ and is distinct from the unfolded ensemble at room temperature. Hence, we propose that the thermal transition of CytR-DBD occurs as a downhill-type transition from the expanded unfolded state at room temperature to a collapsed unfolded state at higher temperatures and under low-salt conditions. However, the heat capacity measurements at the higher salt concentrations detected a large shift from no excess heat capacity peak to a profile with a distinct peak.¹⁹ It is important to note that the presence of an excess heat capacity peak is not a sufficient condition for the two-state folding, as even downhill folders exhibit excess heat capacity peaks.^62,63 However, the appearance of the peak could imply a switch from a barrier-less transition at low salt concentrations to an apparent two-state transition at high salt concentrations (>0.6 M) upon thermal perturbation. Though experimentally challenging, temperature-dependent smFRET could provide a direct confirmation of this intriguing possibility.

Implication for Regulation Mechanisms

A key to understanding the regulation mechanism of CytR, which should differentiate the target and nontarget DNA sequences, is the observed small but consistent lowering of the peak FRET efficiency bound to the specific DNA (∼0.78) compared to that at ∼0.82 for the folded state in the absence of DNA and in the presence of the nonspecific (AC)₇. NMR experiments highlight a 36% reduction in the NOE signals when CytR-DBD is bound to a nonspecific DNA compared to that in the presence of specific DNA.⁶⁴ Furthermore, while the amide protons located in helices I and III are protected from deuterium exchange in the presence of specific DNA, all amide protons exchange rapidly in the presence of nonspecific DNA.⁶⁴ Taken together, the NMR results suggest that CytR-DBD is only loosely bound to the noncognate DNA, sampling conformations within the time scale of NMR measurements. Combining this with our smFRET results, it appears that the compact overall structure in terms of the distance between the N and C termini is maintained in the CytR-DBD while still undergoing structural changes. We thus propose that the observed flexibility of CytR-DBD may be functionally significant and that the small but distinct structural differences, arising from the hinge region flexibility, might trigger the regulation of CytR.

The detection of the single FRET peak for the component bound to the noncognate DNA, where multiple conformations were suggested to exchange in the time scale of NMR measurements,⁶⁴ might be consistent with the apparent two-state transitions detected by the salt-induced folding and urea-induced unfolding processes. As discussed above, the current smFRET data detect only the end-to-end distance of the labeled CytR-DBD and cannot resolve possible heterogeneity in the other regions of the protein. Furthermore, these observations provide structural clues for the negative cooperativity (Hill coefficient, n_H < 1) in the binding of CytR-DBD to cognate DNA; that is, the structural flexibility observed could manifest as a range of binding affinities and therefore manifest as broad binding isotherms.⁴¹

While CytR belongs to the LacR family of proteins, the regulation mechanisms of the family members are slightly different from each other. LacR folds to a different final conformation when bound to different operator sequences,^65,66 and the hinge helix folds only upon binding to the specific DNA and requires protein–protein interaction between the two related hinge helices.^31,33 In the cases of purine repressor (PurR) and fructose repressor (FruR), the hinge helix does not form without their dimerization and complexation with the DNA operator.^33,32 Whereas PurR-DBD alone is unable to show specific DNA binding, LacR-DBD alone is sufficient for the specificity.³² While LacR and PurR cause a significant bend or kink of the bound DNA, CytR alone does not bend DNA significantly^30,67,68 and requires the association of CRP to alter the DNA structure. These differences might be ascribed to the difference in the unstructured regions flanking the helix–turn–helix domain of LacR family members. In the N-terminus, the first helix for other members of LacR starts from the 3rd–4th residue, whereas it starts from the 12th residue in CytR. In the hinge region of DBD, CytR has the helix breakers Pro57 and Gly59, whereas the other family members have a conserved helix maker Alanine.²² Further investigations with a CytR construct not containing the disordered hinge region would give us important clues regarding the structural rigidity and their roles. Comparison of the gene regulation mechanisms with other family members is necessary to understand the significance of the diversity of the family proteins including CytR.

Conclusions

The folding mechanism of CytR-DBD was investigated by using smFRET spectroscopy. The capability of single-molecule spectroscopy to differentiate between the expanded unfolded state and the compact folded state allowed us to reveal transitions between the two populations upon the changes in the various solution conditions. The small but distinct difference in the peak FRET efficiencies for CytR-DBD bound to the specific and nonspecific DNAs might explain the initial triggering event causing the tertiary structural change of CytR. The binding of a DNA-binding domain of a transcription factor to DNA alters the dynamical properties of the bound protein in a DNA-sequence-dependent manner, which might contribute to the formation of a correct regulatory complex.

Glossary

Abbreviations

CytR-DBD

DNA-binding domain of cytidine repressor

smFRET

single-molecule Förster resonance energy transfer

IDPs

intrinsically disordered proteins

IDRs

intrinsically disordered regions

LacR

lactose repressor

CRP

cAMP receptor protein

ALEX

alternating laser excitation

R_g

radius of gyration

BdpA

B domain of protein A

PurR

purine repressor

FruR

fructose repressor

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jpcb.2c02791.

• Supporting text explaining the fitting and error estimation procedures used for the analysis of the FRET efficiency distributions; tables listing the fitting parameters obtained by the analysis of the FRET efficiency distributions; figures showing ion-exchange chromatogram and mass spectroscopic data of the labeled samples and the optical setup for ALEX-smFRET measurement (PDF)

Author Contributions

S. Mitra, A.N.N., and S.T. designed the experiments. H.O. designed and constructed the ALEX measurement system. S. Mitra and H.O. performed the single-molecule experiments. D.R. expressed and purified the sample. S. Mitra, H.O., T.K., and S. Mizukami labeled the sample. S. Mitra, H.O., A.N.N., and S.T. analyzed the data and wrote the paper.

The authors declare no competing financial interest.

Special Issue

Published as part of The Journal of Physical Chemistry virtual special issue “Hiro-o Hamaguchi Festschrift”.