Residual Structure in the Denatured State of the Fast Folding UBA(1) Domain from the Human DNA Excision Repair Protein HHR23A

Abstract

The structure of the first ubiquitin-associated domain from HHR23A, UBA(1), was determined by X-ray crystallography at 1.60 Å resolution, and its stability, folding kinetics and residual structure under denaturing conditions have been investigated. The concentration dependence of thermal denaturation and size-exclusion chromatography indicate that UBA(1) is monomeric. Guanidine hydrochloride (GdnHCl) denaturation experiments reveal that the unfolding free energy, ΔG_u°′(H₂O), of UBA(1) is 2.4 kcal mol⁻¹. Stopped-flow folding kinetics indicates sub-millisecond folding with only proline isomerization phases detectable at 25 °C. The full folding kinetics are observable at 4 °C, yielding a folding rate constant, k_f, in the absence of denaturant of 13,000 s⁻¹ and a Tanford β-value of 0.80, consistent with a compact transition state. Evaluation of secondary structure via circular dichroism shows that residual helical structure in the denatured state is replaced by polyproline II structure as GdnHCl concentration increases. Analysis of NMR secondary chemical shifts for backbone ¹⁵NH, ¹³CO, and ¹³Cα atoms between 4 and 7 M GdnHCl shows three islands of residual helical secondary structure that align in sequence with the three native-state helices. Extrapolation of the NMR data to 0 M GdnHCl demonstrates that helical structure would populate to 17 – 33% in the denatured state under folding conditions. Comparison with NMR data for a peptide corresponding to helix 1 indicates that this helix is stabilized by transient tertiary interactions in the denatured state of UBA(1). The high helical content in the denatured state, which is enhanced by transient tertiary interactions, suggests a diffusion-collision folding mechanism.

Graphical Abstract

INTRODUCTION

There is ongoing interest in how residual structure in the denatured states of proteins seeds the folding mechanism and biases primary structure toward the native-state fold.¹ In addition to residual secondary structure, there is good evidence for long-range interactions within the denatured state from a wide range of methods.^2–4 NMR studies used to determine residual structure at residue-level resolution have either been carried out at ~8 M urea^5–11 or under less harsh acid denaturing conditions.^{6,7,9,12–14} Occasionally, it has been possible to study denatured states of protein in aqueous solution under folding conditions^9,15–17 or by using peptide fragments of proteins.^18,19 In general, the data show that high concentrations of urea eliminate much of the residual structure evident under acid unfolding conditions and thus denatured states in the presence of denaturants have been considered poor models for the denatured state that pertains under folding conditions.

Thermodynamic studies^20–22 and Förster resonance energy transfer (FRET) investigations^23,24 indicate that there is gradual loss of residual structure as denaturant concentration increases, suggesting that experiments at lower concentrations of denaturants could provide insight into residual structure in unfolded proteins. Results from FRET studies have been brought into question by small-angle X-ray scattering (SAXS) measurements, which indicate little to no effect of denaturant concentration on the structure of the denatured state and attribute the observed effects to the large hydrophobic dyes used for FRET studies.²⁵ However, recent studies with small fluorophores show significant compaction of the denatured state at low denaturant concentration and attribute the SAXS results to the bias of SAXS toward conformers with a larger radius of gyration, R_g.²⁶

There are relatively few studies at residue-level resolution on the denaturant dependence of residual structure in denatured proteins.^27,28 Here, we investigate residual structure in the denatured state of a small three-helix bundle domain in 4 – 7 M guanidine hydrochloride (GdnHCl) using NMR secondary shifts. This three-helix bundle, the internal ubiquitin-associated domain, UBA(1), from the human homolog of Saccharomyces cerevisiae Rad23, HHR23A, DNA excision repair protein is an ~45 residue domain. Its structure has been determined previously by NMR.²⁹ The structure of the C-terminal UBA domain from HHR23A, UBA(2), has also been solved by NMR spectroscopy³⁰ and is very similar (backbone RMSD of 1.6 Å) to that of UBA(1), despite only 20% sequence indentity.²⁹ The folding and stability of UBA(2)^31,32 and the properties of its denatured state have been studied,³³ but the folding and stability of UBA(1) remain uncharacterized. Proteins containing UBA domains are involved in cellular processes such as DNA excision repair,^29,30 NF-κB and kinase signaling pathways^34,35 and targeting of ubiquitinated proteins to the proteasome.^35–37 The UBA domains of HHR23A interact with its ubiquitin-like (UBL) domain trapping HHR23A in a closed conformation that may be optimized for its role in nucleotide excision repair.³⁸ The UBL domain also binds to the S5a subunit of the proteosome disrupting binding to the UBA domains and causing HHR23A to adopt an open conformation. The UBA domains of HHR23A are known to bind to polyubiqitin chains. The open conformation of HHR23A is thought to promote inhibition of the proteasome.³⁸

Three-helix bundle domains are well-studied and include some of the fastest folding proteins.^39–44 Many of these fast folding proteins demonstrate two-state behavior, although there is evidence of deviations from two-state behavior.^45,46 The transition states (TS) of some three-helix bundles have been characterized. Their folding mechanisms have been shown to range from diffusion-collision to nucleation condensation,^42,47–50 even within the same fold.⁵¹ The degree of residual helical structure in the denatured state of the engrailed homeodomain family is believed to control the nature of the folding mechanism.⁵¹ For the villin headpiece subdomain, there is considerable evidence of residual structure in the denatured state,^18,19 however, the connection to its folding dynamics is less clear.⁵⁰

The UBA domain has a distinct topology (SCOP: UBA-type fold, up-and-down bundle with a clockwise topology) relative to other well-studied three-helix bundles. In the current work, we present analysis of a 1.60 Å X-ray structure of UBA(1), characterization of the thermodynamics and kinetics of UBA(1) folding and then use UBA(1) as a model system to show that the degree of residual α-helical secondary structure in the denatured state increases monotonically as the concentration of GdnHCl decreases. We also present initial evidence of long-range interactions in the denatured state of UBA(1).

EXPERIMENTAL PROCEDURES

Expression of the UBA Domains and the H1 Peptide.

pGEX-2T plasmids with UBA(1) and UBA(2) coding sequences were supplied by Juli Feigon at UCLA.²⁹ The plasmids contain a glutathione S-transferase (GST) tag for efficient purification of the UBA domains and a human thrombin cleavage site. To produce the pGEX-2T(TEV) plasmid for expression of UBA(1), the thrombin cleavage site (LVPRGS) was converted to a Tobacco Etch Virus (TEV) protease cleavage site (ENLYFQGS) by two steps of site-directed mutagenesis. Both protease sites yield the same UBA(1) construct after proteolytic removal of GST. Both expression plasmids were used for UBA(1) expression and purification. UBA(1) truncated after the first α-helix (H1 Peptide) was prepared by replacing the Arg177 codon (CGA) with a stop codon (TGA) in the pGEX-2T(TEV) plasmid. A C344A variant of UBA(2) was produced to prevent dimerization through formation of an intermolecular disulfide bond.³² Site-directed mutagenesis was performed using the Agilent QuikChange Lightning Site-Directed Mutagenesis Kit with mutagenic primers from Invitrogen (Table S1). All mutations were confirmed by DNA sequencing (Eurofins Genomics).

Phage T1-resistant BL21(DE3) Escherichia coli cells (New England BioLabs, Inc.) were transformed with plasmids coding for UBA(1), C344A UBA(2), or the H1 Peptide. 10 mL overnight LB medium cultures were used to inoculate 1 L of LB medium in 2.8 L Fernbach flasks. All media contained 100 μg/mL ampicillin. The 1 L cultures were grown at 37 °C until an OD₅₅₀ of 0.6 was reached. Then, expression was induced with 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) at 30 °C for 3 hours. The cells were harvested by centrifugation at 5,000 rpm for 10 minutes (Lynx 6000 centrifuge, F12-6x500 LEX rotor, Thermo Scientific) and frozen at −80 °C or lysed immediately.

Purification of the UBA Domains and the H1 Peptide.

Cells were lysed with Bugbuster detergent (EMD Millipore) in the presence of benzonase nuclease (EMD Millipore). For the H1 Peptide, 5 mM β-mercaptoethanol (β-ME) was added to detergent prior to lysis to prevent oxidation of the disordered peptide (see the matrix-assisted laser desorption/ionization time-of-flight, MALDI-TOF, mass spectrometry section). Cellular debris was pelleted by centrifugation at 11,600 rpm for 20 minutes at 4 °C (Lynx 6000 centrifuge, F14-6x250y rotor, Thermo Scientific). Purification was by GST-affinity chromatography (EMD Millipore).

To remove GST from fusion proteins with thrombin cleavage sites the Thrombin Cleavage Capture kit (EMD Millipore) was used. The released GST was separated from UBA(1) or UBA(2) by GST-affinity chromatography.

To cleave off GST from fusion protein with a TEV protease cleavage site, 30 μg of TEV protease was added per mg of fusion protein. The solution was incubated overnight at 4 °C with shaking at 75 rpm, concentrated by ultrafiltration and then loaded onto a Superdex Peptide 10/300 GL high performance size-exclusion column (GE Healthcare) connected to an AKTA FPLC (GE Healthcare) to separate GST from UBA(1) or the H1 Peptide.

The purities of UBA(1), C26A UBA(2) and the H1 Peptide were confirmed by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and MALDI-TOF mass spectrometry. The concentrations of purified UBA(1), UBA(2) and the H1 Peptide were determined from absorbance at 280 nm using extinction coefficients estimated by the method of Pace et al.⁵² (UBA(1), ε₂₈₀ = 5960 M⁻¹ cm⁻¹; UBA(2), ε₂₈₀ = 1490 M⁻¹ cm⁻¹; H1 Peptide, ε₂₈₀ = 2980 M⁻¹ cm⁻¹).

Preparation of ¹³C-/¹⁵N-labeled UBA(1) and the H1 Peptide.

¹³C-/¹⁵N-labeled UBA(1) and H1 Peptide were expressed as described above except that M9 minimal medium containing 4 g/L of U-¹³C₆-D-glucose and 1 g/L of ¹⁵NH₄Cl (Cambridge Isotope Laboratories) was used for 1 L cultures. After addition of IPTG, the induction period was 5 h at 32 °C for UBA(1) and 16 h at 16 °C for the H1 Peptide. Protein was isolated from cell lysate as described above.

MALDI-TOF Mass Spectrometry.

A Bruker microflex MALDI-TOF mass spectrometer calibrated with Bruker standards (cytochrome c, insulin, myoglobin, and ubiquitin) was used to acquire mass spectra. MALDI-TOF mass spectra of UBA(1) showed two distinct peaks, one at 5,634.95 m/z corresponding to the molar mass of monomeric UBA(1) (theoretical MW 5,635.3) and another peak at 11,269.22 m/z corresponding to an apparent dimer (Figure S1). For the H1 Peptide, in addition to a monomolecular peak at 2,526.5 m/z (theoretical MW 2529.8), the initial preparation showed a peak at 2542.7 m/z attributable to monooxidation of a methionine residue. Inclusion of 5 mM β-ME during lysis and purification eliminated methionine oxidation.

Size-exclusion Chromatography.

A Superdex Peptide 10/300 GL high performance column (GE Healthcare) was used to determine if UBA(1) dimerizes in the micromolar concentration range. The column was equilibrated with 50 mM sodium phosphate, pH 6.5, 100 mM NaCl (pH 6.5 phosphate buffer) prior to loading samples. UBA(1) in pH 6.5 phosphate buffer was loaded onto the size-exclusion column. Equine cytochrome c (Sigma) dissolved in pH 6.5 phosphate buffer was used as a molecular weight standard. To assess whether UBA(1) and UBA(2) from HHR23A bind to each other, 50 μM UBA(1) and 50 μM C26A UBA(2) in pH 6.5 phosphate buffer were mixed for variable incubation times prior to size-exclusion chromatography.

Concentration Dependent Thermal Unfolding.

UBA(1) was diluted to approximately 5, 20, 50, 100 and 200 μM in pH 6.5 phosphate buffer. The quartz cuvette pathlength used at each UBA(1) concentration (in brackets) was: 1 cm (5 μM), 0.4 cm (20 and 50 μM) and 0.1 cm (100 and 200 μM). Circular dichroism (CD) data were acquired using an Applied Photophysics Chirascan CD Spectrophotometer with a Peltier temperature controller. For temperature ramping experiments, the temperature first was equilibrated to 6 °C and then increased at a rate of 1 °C per minute. Data were collected at 222 nm and 250 nm in 2 °C increments until 90 °C was reached. A magnetic stir bar was used to provide for homogenous heating. A thermocouple was placed inside the cuvette to determine the actual temperature. For the 0.1 cm cuvette, where stirring was not possible, the thermocouple was placed adjacent to the cuvette. At each UBA(1) concentration, pre- and post-thermal melt CD spectra from 200 to 250 nm were obtained to assess the recovery of α-helical secondary structure upon cooling. The reversibility of UBA(1) thermal unfolding was close to 100% at all concentrations up to 100 μM. At 200 μM, the reversibility of the thermal unfolding was not complete.

Ellipticity of buffer alone at 222 nm was use as a background correction for ellipticity at 222 nm. θ₂₂₂ was plotted versus temperature in Kelvin and eq 1 was fit to the data.⁵³

$${\text{θ}}_{222}(\text{T})=\frac{({\text{θ}}_{\text{N}}+m_{\text{N}}\text{T})+({\text{θ}}_{\text{D}}+m_{\text{D}}\text{T})\times \text{exp}\left[-\left\{\mathrm{\Delta }{\text{H}}_{\text{m}}\left(1-\frac{\text{T}}{{\text{T}}_{\text{m}}}\right)-\mathrm{\Delta }{\text{C}}_{\text{p}}\left[({\text{T}}_{\text{m}}-\text{T})+\text{Tln}\left(\frac{\text{T}}{{\text{T}}_{\text{m}}}\right)\right]\right\}/\text{RT}\right]}{1+\text{exp}\left[-\left\{\mathrm{\Delta }{\text{H}}_{\text{m}}\left(1-\frac{\text{T}}{{\text{T}}_{\text{m}}}\right)-\mathrm{\Delta }{\text{C}}_{\text{p}}\left[({\text{T}}_{\text{m}}-\text{T})+\text{Tln}\left(\frac{\text{T}}{{\text{T}}_{\text{m}}}\right)\right]\right\}/\text{RT}\right]}$$ (1)

In eq 1, θ_N and m_N, and θ_D and m_D, are the intercept and the slope of the native- and denatured-state baselines respectively, T_m is the midpoint temperature for thermal unfolding, ΔH_m is the enthalpy of unfolding at T_m and ΔC_p is the heat capacity increment for thermal unfolding. Given the close correlation between the number of residues in a protein and the magnitude of ΔC_p,⁵⁴ we set ΔC_p to a value previously obtained for the UBA(2) domain of HHR23A, 0.45 kcal mol⁻¹ K⁻¹,³² to fit eq 1 to the θ₂₂₂ versus T data. All other parameters in eq 1 were allowed to vary freely. Reported parameters are the mean and standard deviation of three independent trials.

Global Unfolding by GdnHCl Denaturation.

GdnHCl denaturation was carried out in 20 mM MES, 40 mM NaCl, pH 6.5 (CD buffer). An approximately 6 M GdnHCl solution containing 5 μM UBA(1) and CD buffer was prepared, and its exact GdnHCl concentration was determined from the refractive index of the solution.⁵⁵ This solution was titrated into a 5 μM solution of UBA(1) in CD buffer using a Hamilton Microlab 500 Titrator coupled to an Applied Photophysics Chirascan CD Spectrophotometer. Temperature was controlled at 25 °C with a Peltier device. Ellipticity was measured at 222 nm using 250 nm as background, θ_222corr = θ₂₂₂ - θ₂₅₀. Plots of θ_222corr versus GdnHCl concentration were fit to eq 2, which assumes two-state unfolding and a linear dependence of the free energy of unfolding, ΔG_u, on GdnHCl concentration.^56–58

$${\text{θ}}_{222\text{corr}}=\frac{\left[({\theta }_{\text{N}}+m_{\text{N}}[\text{GdnHCl}])+({\theta }_{\text{D}}+m_{\text{D}}[\text{GdnHCl}])\cdot \text{exp}\left\{\frac{m[\text{GdnHCl}]-\mathrm{\Delta }{\text{G}}_{\text{u}}^{°\prime }({\text{H}}_2\text{O})}{\text{RT}}\right\}\right]}{1+\text{exp}\left[\frac{m[\text{GdnHCl}]-\mathrm{\Delta }{\text{G}}_{\text{u}}^{°\prime }({\text{H}}_2\text{O})}{\text{RT}}\right]}$$ (2)

In eq 2, θ_N and m_N are the intercept and slope of the native state baseline, θ_D and m_D are the intercept and slope of the denatured state baseline, m is rate of change of ΔG_u with respect to GdnHCl concentration and ΔG_u°′(H₂O) is the free energy of unfolding extrapolated to 0 M GdnHCl at pH 6.5. For GdnHCl denaturation at 4 °C, tyrosine fluorescence data were collected simultaneously with CD data. Excitation was at 280 nm with emission measured after passing through a 305 nm cutoff filter and an emission monochromator set to 315 nm. An equation of the form of eq 2 was fit to plots of fluorescence at 315 nm versus GdnHCl concentration. Reported parameters are the mean and standard deviation of at least three independent trials.

Stopped-Flow Refolding and Unfolding Kinetics.

Purified 200 μM UBA(1) in CD buffer or in CD buffer containing concentrated GdnHCl was mixed 1:10 with CD buffer containing variable concentrations of GdnHCl using an Applied Photophysics SX20 Stopped-flow Spectrophotometer. Folding and unfolding were monitored using tyrosine fluorescence (excitation at 280 nm). Fluorescence was collected at 90° to the direction of excitation using a PM tube with a 295 nm cut-off filter in front of it. The final GdnHCl concentration after mixing was verified using a Fisher Scientific refractometer. For mixing at 25 °C, only a proline isomerization phase was observable. UBA(1) in CD buffer with the maximum GdnHCl concentration was mixed with CD buffer containing the maximum GdnHCl concentration of GdnHCl, and native UBA(1) in CD buffer was mixed with CD buffer to obtain denatured and native baselines, respectively, for fluorescence at 25 °C. Fluorescence amplitude was plotted versus GdnHCl to demonstrate the amplitude loss that occurred within the dead time of the stopped-flow (2 ms at 25 °C as measured by reduction of 2,6-dichlorophenolindophenol⁵⁹).

Experiments were also carried out at 4 °C to capture the entire folding process. The dead-time for folding was evaluated using the software of the Applied Photophysics SX20 Stopped-flow Spectrophotometer at a range of final GdnHCl concentrations (1.45 to 6.09 M) at 4 °C using a drive ram fitted with a transducer. The dead-time was found to be 1.62 ± 0.06 ms. Therefore, 1.6 ms was added to all time points before a single exponential function was fit to the data. Eq 3 was fit to chevron plots of the natural log of the observed rate constants, ln(k_obs), versus GdnHCl concentration to extract the rate constants for folding and unfolding in the absence of denaturant, k_f(H₂O) and k_u(H₂O), respectively.

$$\text{ln}(k_{\text{obs}})=ln((k_{\text{f}}({\text{H}}_2\text{O})\times exp(-m_{k_{\text{f}}}\times [\text{GdnHCl}]))+(k_{\text{u}}({\text{H}}_2\text{O})\times exp(m_{k_{\text{u}}}\times [\text{GdnHCl}])))$$ (3)

In eq 3, m_kf is the slope of the folding branch of the Chevron plot and m_ku is the slope of the unfolding branch of the Chevron plot. The equilibrium ΔG_u°′(kin), m-value and the Tanford β-values, β_T, are calculated with eqs 4, 5 and 6, respectively.⁶⁰

$$\text{Δ}{\mathrm{G}}_{\mathrm{u}}{}^{°}\prime (\mathrm{kin})=\mathrm{RT}ln(k_{\mathrm{f}}/k_{\mathrm{u}})$$ (4)

$$m=\mathrm{RT}(m_{k_{\mathrm{f}}}+m_{k_{\mathrm{u}}})$$ (5)

$${\mathrm{\beta }}_{\mathrm{T}}=m_{k_{\mathrm{f}}}/(m_{k_{\mathrm{f}}}+m_{k_{\mathrm{u}}}).$$ (6)

CD Spectra of UBA(1) and the H1 Peptide as a Function of GdnHCl Concentration.

Samples of 50 μM UBA(1) and the H1 Peptide were prepared in 50 mM MES, 100 mM NaCl, pH 6.5 and contained concentrations of GdnHCl ranging from 0 M to 7 M. CD spectra from 200 to 250 nm were measured with a Chirascan CD spectrophotometer at 25 °C and a pathlength of 1.0 mm, a step size of 1 nm, a bandwidth of 1 nm and a 5 s acquisition time per data point. Spectral scans were repeated in triplicate for GdnHCl concentrations from 0 M to 3 M and in quintuplicate for GdnHCl concentrations from 4 M to 7 M to obtain standard deviations at each wavelength. Spectra were also baseline-corrected with respect to 250 nm after determining standard deviations.

NMR Chemical Shift Analysis of UBA(1) and the H1 Peptide.

Samples of labeled UBA(1) (0.39 – 0.86 mM) and the H1 Peptide (0.18 – 0.24 mM) were prepared in pH 6.5 phosphate buffer, and GdnHCl concentrations ranging from 4 M to 7 M. All samples contained D₂O, NaN₃, and DSS (4,4-dimethyl-4-silapentane-1-sulfonate) at final concentrations of 10%, 0.2%, and 50 μM, respectively. For backbone assignments of UBA(1) in the denatured state, HNCO,^61,62 C(CO)NH,⁶³ HNCACB,⁶⁴ HN(CA)CO⁶⁵ and CBCA(CO)NH⁶⁶ spectra were acquired at 4 and 6 M GdnHCl using standard VNMR pulse sequences. Chemical shifts for ¹⁵NH, ¹³CO, and ¹³Cα were obtained from ¹H-¹⁵N HSQC,⁶⁷ HNCO, and HNCA^61,68 spectra, respectively, using the methods of Poulsen and co-workers.⁶⁹ To increase digital resolution in the indirect dimensions, narrow spectral widths were used such that a number of signals were aliased in the ¹⁵N dimension, and signals from glycine ¹³Cα were aliased in the ¹³C dimension of the HNCA spectra. Tables S2 to S6 detail the spectral widths used, number of increments, resolution, and other NMR experimental parameters for both UBA(1) and the H1 Peptide. All data were acquired at 25 °C with a 600 MHz VNMRS spectrometer equipped with a triple-resonance probe. Experiments with UBA(1) were repeated in triplicate with individually prepared samples to obtain standard deviations for chemical shifts. Experiments with the H1 Peptide were not repeated, and the error was approximated by the median standard deviation value obtained for UBA(1) for all residues and all GdnHCl concentrations for the relevant nucleus.

All spectral chemical shifts were referenced indirectly to DSS using established gyromagnetic ratios. Spectra were processed using NMRPipe with cosine bell squared apodization and zero-filling to the next power of 2 for 2D spectra and the second-next power of 2 for 3D spectra. NMR data were analyzed using CCPNmr Analysis software.⁷⁰ Crosspeaks were picked using the built-in peak-picking utilities. In the case of overlap, crosspeaks were deconvoluted using the MATLAB Curve Fitting Toolbox or LSQCURVEFIT commands and fit to a multi-dimensional multi-peak Lorentzian lineshape (Figure S2). The error of chemical shifts associated with overlapping crosspeaks is estimated as either the standard deviation of the triplicated deconvoluted shift values or the average 95% confidence interval from nonlinear fitting to Lorentzian lineshapes, whichever is larger. The chemical shifts are reported in Supporting Information, Tables S7 (BRMB 51260) and S8 (BRMB 51257).

Crystallization and Structure Determination of UBA(1).

UBA(1) was purified as described above and concentrated to 20 mg/mL in 50 mM HEPES, 150 mM NaCl, pH 8.0. Crystals that diffracted to 1.60 Å resolution were obtained at 20 °C with vapor diffusion from a sitting drop containing a 1:1 mix of the protein solution and a reservoir solution of 0.1 M Tris pH 8.5, 1.5 M ammonium sulfate, 12% (v/v) glycerol. X-ray diffraction data were collected at the Stanford Synchrotron Radiation Lightsource beamline 9-2 with a DECTRIS PILATUS 6M detector. The data were indexed, integrated and scaled in the P4₃ space group using XDS⁷¹ and Aimless.⁷² The structure was solved by molecular replacement using PHASER, integrated into the PHENIX software suite.⁷³ Model building was accomplished in PHENIX, and the structure was refined through iterative cycles of manual adjustment in Coot⁷⁴ and refinement of atomic positions, real space, occupancy, and thermal parameters in PHENIX,⁷³ yielding R_work = 0.162 and R_free = 0.205. Data collection and refinement statistics are provided in Table S9. Coordinates have been deposited at the Protein Data Bank (www.rcsb.org) under PDB code: 6W2H. All molecular images were produced with PyMol.⁷⁵

RESULTS

Novel Insights from a High-Resolution Crystal Structure of UBA(1).

Although an NMR structure of UBA(1) is available,²⁹ we also solved the structure of UBA(1) by X-ray crystallography (Table S9). This new 1.60 Å resolution structure allows us to better define the hydrogen bonding patterns in the turns between the helices. Figure 1 shows an overlay of the X-ray structure with the NMR structure. The two structures are similar with an all-atom RMSD of 1.917 Å and a backbone RMSD of 1.329 Å for residues 160 – 201. There are small changes in the turn conformations and the orientation of the helices, but the overall topology is the same.

Figure 1.

Overlay of NMR²⁹ (gray, PDB ID: 1IFY) and X-ray (blue, PDB ID: 6W2H) structures of UBA(1). The structures were superimposed using backbone atoms of residues 160–201. In the X-ray structure the helix ranges are; helix 1, residues 159 – 172; helix 2: residues 176 – 187; and helix 3, residues 190 – 201.

In a small helical bundle, the structures of turns and the helix capping motifs that direct the conformations of the turns are expected to be important to the topology and stability of the fold. Helix-capping motifs have been extensively analyzed.^76,77 The common nomenclature for the entry into a turn at the C-terminus of a helix is …C4-C3-C2-C1-Ccap-C′-C′′-C′′′… and for the exit from a turn at the N-terminus of a helix is …N′′′-N′′-N′-Ncap-N1-N2-N3-N4… For many turns between helices, the C′/C′′/C′′′ and N′′′/N′′/N′ designations will overlap. At the beginning of a turn the amide NH of the Ccap residue is hydrogen bonded to the C=O of C4 and bridges the helix and the turn. Similarly, the C=O of the Ncap residue is hydrogen bonded to the amide NH of N4 providing a bridge between the turn and the helix that follows.

Although sequence conservation of UBA domains is low, the first turn of UBA domains shows a high preference for the sequence MGY.²⁹ Figure 2A shows the hydrogen bonding pattern in turn 1. The MGY sequence is bracketed by the Ccap residue of helix 1, Ser172, and the Ncap of helix 2, Glu176. Serine is less common at the Ccap position.^76,77 The amide NH of Ser172 forms a bifurcated hydrogen bond that completes the last helical hydrogen bond to the carbonyl of Thr168 and helps to cap the carbonyl of Glu169 (C3). The Ser172 side chain also caps the carbonyl of Glu169, a less common mode of C-capping.⁷⁷ In the common Schellman C-capping motif with Gly at the C′ position, the amide NHs at C′ and C′′ cap the carbonyls of C2 and C3, respectively. However, the amide NH of Gly174 (C′′) instead caps the carbonyl of Met171 (C1), while C′/C2 capping is still observed between the amide NH of Met173 (C′) and the carbonyl of Ile170 (C2). The carbonyls of Ser172, Met173 and Gly174 are exposed to solvent with hydrogen bonds to H₂O or polar groups of adjacent molecules in the crystal lattice.

Figure 2.

Hydrogen bonds (yellow dashed lines) around (A) turn 1 and (B) turn 2 of UBA(1) as observed in the X-ray structure (PDB ID: 6W2H). Waters are shown as red spheres.

The N-capping of helix 2 does not have the classic reciprocal side chain motifs of the capping box^76,78 and big box^76,79 motifs between the amide NHs and side chains of N3 and Ncap or N′ residues, respectively. Instead, the carbonyl of Glu176 at the N-cap position forms a bifurcated H-bond to the N3 (Arg 179) and N4 (Val180) amide NHs (Figure 2A). H-bonds from the side chains of Glu176 and Glu178 to their own amide NHs are not denoted with dashed lines in Figure 2A because the electron density that supports the position of these side chains in the structural model is weak. The amide NH and CO of Tyr175 and the amide NH of Arg177 lack polar groups within hydrogen bonding distance, but are either solvent exposed or are just outside of H-bonding distance of a polar group.

Turn 1 also contains two interesting long-range contacts, a salt bridge between Arg177 and Glu164 in helix 1, and a water-mediated polar contact between the side chain of Tyr175 and the carbonyl of Leu198 at the C-terminus of helix 3. Thus, turn 1 is mediated by two less common capping motifs, has many stabilizing polar contacts and two long-range contacts that may help to establish the topology of UBA(1).

Sequence conservation is much poorer in turn 2.²⁹ However, the Ncap residue for helix 3, position 190, shows a strong preference for Asn. There are only two residues between the Ccap residue of helix 2, Ser187, and the Ncap residue of helix 3, Asn190. The carbonyl of Ser187 is hydrogen bonded to the amide NH of Asn190 forming a β-turn between the Ccap of helix 2 and the Ncap of helix 3 (Figure 2B). The ϕ, ψ angles of Tyr188 (58.5°, 38.0°) and Asn189 (68.7°, 16.2°) are similar to those of a type III′ β-turn.⁸⁰ The turn is further stabilized by a hydrogen bond between the side chain of Ser187 and the carbonyl of Asn190. In the NMR structure the ϕ, ψ angles of Tyr188 and Asn189 are similar to those of our X-ray structure, however, the main chain H-bond between Ser187 and Asn190 is not present.²⁹

A β-turn following a helix can combine with a Schellman motif.⁷⁷ The C-capping pattern of helix 2 by turn 2 resembles that of a non-Gly Schellman motif⁷⁶ with capping of C3 (Leu184) by the amide NH of C′′ (Asn189) and of C2 (Arg185) by the amide NH of C′ (Tyr188) (Figure 2B). However, the amide NH of Tyr188 forms an H-bond to the C3 carbonyl, as well, and the hydrophobic interaction between C3 (Leu184) and C′′/C′′′ (Asn189/Asn190) common in Schellman motifs is absent.^76,81 The carbonyl of C1 (Ala186) is capped by H-bonds to two waters, one of which mediates a long-range contact to Tyr197 of helix 3.

The Ncap of helix 3, Asn190, is overrepresented at the Ncap position as is Pro191 at the N1 position.^76,77 The side chain of Asn190 caps both the amide NHs of the N2 (His192) and N3 (Arg193) (Fig 2B). Thus, turn 2 of UBA(1) like turn 1 is also supported by a strong network of hydrogen bonds and makes long-range contacts to an adjacent helix.

Oligomeric State of UBA(1).

The observation of dimers in the MALDI-TOF mass spectrum of UBA(1) (Figure S1) led us to investigate the oligomeric state of UBA(1). To ascertain whether or not dimerization occurs, we followed the concentration dependence of thermal unfolding of UBA(1). If dimerization is occurring, a shift in the midpoint temperature, T_m, for thermal unfolding would be expected.³⁵ Parameters from a fit of eq 1 to the thermal unfolding data (Figure S3), acquired for UBA(1) concentrations ranging from about 4 μM to about 100 μM, are given in Table 1. Over this concentration range, we observe no significant increase in T_m, indicating that UBA(1) does not change oligomerization state between 4 and 100 μM. Size-exclusion chromatography of UBA(1) loaded at a concentration of 100 μM shows that it elutes well after equine cytochrome c (Figure S4), consistent with UBA(1) being monomeric at this concentration. Finally, evaluation of intermolecular contacts in the UBA(1) crystal structure by PISA⁸² also indicates that there are no biologically relevant interfaces that could lead to dimerization. The largest interface observed is 505 Ų. These results are consistent with the NMR study of UBA(1), which also found no evidence for dimerization of UBA(1).²⁹ Thus, all data indicate that UBA(1) does not form homodimers, suggesting that the dimer peak observed in MALDI-TOF mass spectra is an artifact of the ionization process.

Table 1.

Thermal Unfolding Parameters for UBA(1) as a Function of Concentration at pH 6.5.^a

UBA(1) concentration (μM)	Tm (K)	ΔHm (kcal mol−1)	ΔGu°′(25 °C) (kcal mol−1)	ΔGu°′(4 °C) (kcal mol−1)
3.8	339 ± 1	35 ± 5	3.1 ± 1.1	3.7 ± 1.3
20.9	339.67 ± 0.07	31.6 ± 0.4	2.7 ± 0.2	3.0 ± 0.3
56.5	340.0 ± 0.4	31.1 ± 0.8	2.6 ± 0.4	3.0 ± 0.5
99.3	340 ± 1	32 ± 2	2.7 ± 0.9	3.1 ± 1.0

^aΔG_u°′(25 °C) and ΔG_u°′(4 °C) were extrapolated from T_m, using the Gibbs-Helmholtz equation, the fitted ΔH_m values, and the ΔC_p for UBA(2) given in Experimental Procedures.

Previous data indicated that the yeast homolog of HHR23A, Rad23, dimerizes through its UBA domains.⁸³ No data were provided to show whether one or both UBA domains were required. Given that UBA(1) does not dimerize, a possible explanation for the observation that Rad23 requires UBA domains for homodimerization could be that UBA(1) binds to UBA(2).²⁹ However, when a mixture 50 μM UBA(1) and 50 μM UBA(2) is analyzed by size-exclusion chromatography, we see no evidence that these two UBA domains interact (Figure S4). NMR studies also demonstrate that the UBA domains of HHR23A do not interact with each other and furthermore show that HHR23A is likely a monomer, unlike Rad23.³⁸ Thus, our results are consistent with previous studies which suggest that the oligomerization states of HHR23A and Rad23 differ.

Thermodynamics and Kinetics of GdnHCl Denaturation of UBA(1) at 25 °C.

The thermodynamic stability of UBA(1) was characterized by GdnHCl denaturation at 25 °C (Figure 3). The data were fit to eq 2 yielding ΔG_u°′(H₂O) = 2.39 ± 0.05 kcal mol⁻¹, m = 1.16 ± 0.02 kcal mol⁻¹ M⁻¹ and a midpoint GdnHCl concentration, C_m, of 2.06 ± 0.06 M. The ΔG_u°′(H₂O) is similar to the extrapolated value of ΔG_u°’(25 °C) ≈ 2.7 kcal mol⁻¹ obtained from thermal unfolding data (Table 1). The stability obtained for UBA(2) from HHR23A by GdnHCl denaturation (ΔG_u°′(H₂O) = 1.34 ± 0.06 kcal mol⁻¹; m = 1.12 ± 0.01 kcal mol⁻¹ M⁻¹ and C_m = 1.19 ± 0.01 M)³¹ is lower than that of UBA(1).

Figure 3.

Equilibrium GdnHCl denaturation of UBA(1) at 25 °C in 20 mM MES, 40 mM NaCl, pH 6.5 monitored by CD at 222 nm with 250 nm as background, θ_222corr.

UBA(1) contains four tyrosine residues. When UBA(1) unfolds in GdnHCl, there is a significant decrease in fluorescence (Figure S5). This decrease in fluorescence provides a convenient means to monitor the folding kinetics of UBA(1) as a function of GdnHCl concentration. When UBA(1) in buffer or 5 M GdnHCl is mixed with a range of GdnHCl concentrations at 25 °C, only the final fluorescence is observable (Figure S6A). A plot of final fluorescence versus GdnHCl (Figure S6B) is consistent with the equilibrium unfolding data shown in Figure 3. Thus, at 25 °C UBA(1) is a fast folding protein that folds within the dead time of our stopped-flow instrument. On a 50 s time scale a low-amplitude phase attributable to proline isomerization is observable (Figure S7). The rate constants for this phase are near 0.1 s⁻¹ and are relatively invariant as a function of GdnHCl concentration (Table S10). For refolding, the amplitude for this phase is maximal at low GdnHCl concentration and for unfolding the amplitude decreases as GdnHCl concentration increases above C_m, as is typical for proline isomerization phases (Figure S8 and Table S10).⁶⁰

Thermodynamics and Kinetics of GdnHCl Denaturation of UBA(1) at 4 °C.

To slow the folding and unfolding of UBA(1) so that we could determine the rate constants, k_obs, for the main folding and unfolding phases, we carried out stopped-flow experiments at 4 °C. A chevron plot showing the GdnHCl concentration dependence of ln(k_obs) is shown in Figure 4. The folding arm (closed triangles) depends more steeply on GdnHCl concentration than the unfolding arm (open triangles) of the chevron plot. The fit of eq 3 to the data yields k_f = 1.3 ± 0.2 x 10⁴ s⁻¹, k_u = 50 ± 6 s⁻¹, m_kf = 1.76 ± 0.09 M⁻¹ and m_ku = 0.43 ± 0.03 M⁻¹. Using eqs 4 and 5, we obtain ΔG_u°′(kin) = 3.1 ± 0.1 kcal mol⁻¹ and m = 1.21 ± 0.05 kcal mol⁻¹ M⁻¹. ΔG_u°′(kin) obtained at 4 °C is within error of ΔG_u°’(4 °C) extrapolated from the thermal unfolding data (Table 1). Eq 6 yields a Tanford β-value, β_T, of 0.80 ± 0.05 indicating a compact TS for the folding of UBA(1).

Figure 4.

Chevron plot for the folding (filled triangles) and unfolding (open triangles) of UBA(1) as a function of GdnHCl concentration. Data were collected at 4 °C in 20 mM MES, 40 mM NaCl at pH 6.5.

Because of the long extrapolation involved in obtaining ΔG_u°’(4 °C), we also investigated the GdnHCl unfolding of UBA(1) at 4 °C by CD and tyrosine fluorescence (Figure 5). The fit of eq 2 to the GdnHCl denaturation data monitored by CD at 4 °C yields ΔG_u°′(H₂O) = 3.1 ± 0.1 kcal mol⁻¹, m = 1.20 ± 0.03 kcal mol⁻¹ M⁻¹ and C_m = 2.62 ± 0.05 M. The fit of eq 2 to the GdnHCl denaturation data monitored by tyrosine fluorescence at 4 °C yields ΔG_u°′(H₂O) = 3.08 ± 0.09 kcal mol⁻¹, m = 1.19 ± 0.02 kcal mol⁻¹ M⁻¹ and C_m = 2.59 ± 0.06 M. The consistency of the thermodynamic parameters obtained with two different spectroscopic probes indicates two-state folding. The ΔG_u°′(kin) and m-value obtained from the kinetic folding data at 4 °C are essentially identical to the thermodynamic parameters obtained from equilibrium GdnHCl denaturation at 4 °C, indicating that the kinetics of UBA(1) folding are two-state.

Figure 5.

Equilibrium GdnHCl denaturation of UBA(1) at 4 °C in 20 mM MES, 40 mM NaCl, pH 6.5 monitored by tyrosine fluorescence at 315 nm, F₃₁₅, and circular dichroism at 222 nm with 250 nm as background, θ_222corr. CD and fluorescence data were collected simultaneously on the same sample using an Applied Photophysics Chirascan CD spectrometer. Solid curves are fits to eq 2. The open data points were not included in the fit to the tyrosine fluorescence data.

Secondary Chemical Shifts of Denatured UBA(1) and the H1 Peptide.

NMR spectroscopy is a powerful tool to study protein conformations in the denatured state.^2–4,84 Secondary chemical shifts (SCSs) in particular report on phi and psi angles, providing information on the degree of residual secondary structure formation for a particular denaturing condition. The magnitude of SCSs corresponds to the population of α-helix or β-strand due to conformational averaging during data acquisition. The sign of the SCS depends on the nucleus for each type of secondary structure. For both ¹³Cα and ¹³CO nuclei, positive and negative SCSs indicate α-helix and β-strand formation, respectively, but for ¹⁵NH nuclei, positive and negative SCSs indicate β-strand and α-helix formation, respectively. For all nuclei, SCS values of random coil conformations are expected to be near zero.

Both UBA(1) and the H1 Peptide produce high quality 2D and 3D NMR spectra in denaturing concentrations of GdnHCl. Representative ¹H-¹⁵N HSQC spectra with assignments are shown in Figure S9. Under denaturing conditions, chemical shifts show significant dependence on GdnHCl concentration for both UBA(1) and the H1 Peptide (Figure S10).

To quantify the residual structure in the denatured state of UBA(1) and the H1 Peptide, secondary chemical shifts (SCS) as a function of GdnHCl were determined. Figure 6 shows the ¹³Cα secondary shifts obtained at 4, 5 and 6 M GdnHCl using 7 M GdnHCl as a random coil reference state for UBA(1). Random coil chemical shifts are sequence dependent.⁸⁵ Using the chemical shifts of UBA(1) under strongly denaturing conditions as a reference state intrinsically accounts for this sequence dependence.^28,86 The data reveal three islands of positive SCS indicating persistent helical structure in the denatured state. These three envelopes are separated by short segments of near zero or negative magnitude SCSs at all reported concentrations of GdnHCl. Thus, the conformational distribution of the interhelical regions does not differ strongly from that of 7 M GdnHCl, no matter the strength of the denaturing condition. By contrast, the SCSs of the three regions of residual helical structure slowly dissipate as the strength of the denaturing condition increases. The gray panels in Figure 6 correspond to the positions of the three helices of UBA(1) in the native-state structure. The three islands of residual helical structure approximately localize to the segments of the primary structure that are helical in the native state of the protein (Figure 1). Similar patterns of residual helical structure are observed for the ¹³CO and ¹⁵NH SCSs (Figures S11 and S12). However, as noted previously,²⁸ the helical envelopes of the ¹⁵NH SCSs are shifted ~4 residues from their positions in the ¹³CO and ¹³Cα SCSs because the first four NHs of a helix are not involved in helical hydrogen bonds indicating that NH---O=C hydrogen bonds contribute strongly to the ¹⁵NH SCS (Figure S12).

Figure 6.

¹³Cα secondary shifts, SCS(Cα), using chemical shifts at 7 M GdnHCl as a random coil reference state at 25 °C for (A) the H1 Peptide and (B) UBA(1). Open symbols connected with dashed lines are data points where the chemical shift was extracted by deconvoluting overlapping crosspeaks. Gray panels indicate the positions of the helices in the native structure of UBA(1) (PDB ID: 6W2H).

The data in Figure 6 show that there is significant residual structure in all three helices of UBA(1) with the relative order of residual helical structure being helix 1 ≈ helix 2 > helix 3. Residual structure under denaturing conditions would be expected to mirror the intrinsic stability of the individual helical segments. However, the program AGADIR,⁸⁷ which predicts helical stability in the absence of tertiary interactions, indicates that the central helix of UBA(1) has a considerably higher intrinsic helical propensity than the first and third helices (Figure S13). This difference between the predictions of AGADIR and the data in Figure 6 might suggest that there are stabilizing interactions between the helices of UBA(1) under denaturing conditions.

To test this possibility, we truncated UBA(1) by putting a stop codon in place of Arg177 (H1 Peptide). Thus, the H1 Peptide contains the disordered N-terminal extension, helix 1 and the first turn of UBA(1) ending with the N-cap residue of helix 2. The SCS data of the H1 Peptide also reveal native-like helix formation between residues 160 and 173 (Figure 6). The magnitudes of the helical SCSs for the H1 Peptide are smaller than for helix 1 in full length UBA(1) at all GdnHCl concentrations, suggesting that helix 1 is stabilized by tertiary interactions under denaturing conditions. However, the magnitude of this effect is much smaller than indicated by AGADIR. We further analyze these differences in the stability of helix 1 induced by tertiary interactions in the denatured state of UBA(1) in the Discussion section.

In comparing the UBA(1) NMR data to the H1 Peptide NMR data, we also noted some differences in the behavior of crosspeak linewidths. At 5, 6, and 7 M GdnHCl the linewidths for both UBA(1) and the H1 Peptide were uniformly small (Tables S4 and S6), as expected for a denatured protein. At 4 M GdnHCl, we observe significant increases in ¹⁵NH and ¹³CO linewidths for some residues in UBA(1) (Figure S14), while the linewidths stay small for the H1 Peptide (Figure S15). Residues with increased linewidths are mostly within helical regions, although not all residues within helical regions have increased linewidths. Sometimes, an increased linewidth coincides with the appearance of a shoulder on a crosspeak (see Val181 and Leu199 crosspeaks in Figure S10B), which may indicate that some residues experience more than one environment in the denatured state at 4 M GdnHCl.

CD Spectra of UBA(1) and the H1 Peptide as a Function of GdnHCl Concentration.

To gain further insight into the nature of the residual secondary structure of UBA(1) and the H1 Peptide under denaturing conditions, we acquired CD spectra of both as a function of GdnHCl concentration (Figure 7). At low concentrations of GdnHCl (Figure 7B, inset), UBA(1) shows the characteristic signature of α-helical content with a negative minimum in ellipticity at 222 nm. For the H1 Peptide, ellipticity at 222 nm is a shoulder consistent with the smaller helical content expected for a short peptide (Figure 7A). At 5 M GdnHCl, 222 nm is no longer a local minimum for UBA(1). Instead, a positive maximum near 230 nm begins to appear that is characteristic of a polyproline II (PP_II) helix.⁸⁸ The elliptical contribution of PP_II helix at 222 nm is relatively small,⁸⁹ so the presence of significant negative ellipticity at 222 nm indicates the presence of residual α-helix structure in addition to the PP_II helix. For the H1 Peptide, the positive ellipticity near 230 nm is more pronounced than for UBA(1) as GdnHCl concentration increases above 5 M (Figure S16). Previous work has shown that PP_II helices in peptides and proteins are stabilized by increased concentrations of chaotropic denaturants.^89–91 In the Discussion, we estimate the relative amounts of α-helix and PP_II structure for UBA(1) and the H1 Peptide.

Figure 7.

CD spectra, mean residue ellipticity, [θ]_MRW, versus wavelength for (A) the H1 Peptide and (B) UBA(1) as a function of GdnHCl concentration at 25 °C. GdnHCl concentrations are 0 M (light brown), 1 M (red), 2 M (maroon), 3 M (purple), 4 M (blue), 5 M (teal), 6 M (dark green) and 7 M (light green). In (B), the main panel is shown on the same scale as for the H1 Peptide so that the spectra from 4 to 7 M GdnHCl can be seen better. The inset in (B) is scaled to show the full spectra of UBA(1) at all GdnHCl concentrations.

DISCUSSION

Thermodynamic and Kinetic Properties of the Folding of UBA(1).

Our results confirm previous data²⁹ showing that UBA(1) from HHR23A is a monomer and not a dimer, unlike some other UBA domains.^35,92 UBA(1) has a modest stability of ~2.4 kcal mol⁻¹ at 25 °C, which is nevertheless ~1 kcal mol⁻¹ higher than the stability of UBA(2) from HHR23A.³¹ Like many small helical domains, UBA(1) is a fast folder at 25 °C.^40,41 The main folding phase is complete within the ~2 ms dead time of our stopped-flow instrument at 25 °C, with only a slow proline isomerization phase, likely due to Pro191 at the N-terminus of helix 3, being observable. At 4 °C, the main folding phase is measureable yielding k_f ≈ 13,000 s⁻¹ corresponding to a folding time, τ_f, of ~ 77 μs. The relative contact order (RCO) of UBA(1) is 0.0813 (%RCO = 8.13).⁹³ The relationship between lnk_f and %RCO [lnk_f(predicted) = 16.94 – 0.76×%RCO]⁹⁴ yields k_f(predicted) ≈ 47,000 s⁻¹ (τ_f ≈ 21 μs). Given the deadtime of our stopped flow and the fact that we only observe proline isomerization phases at 25 °C (Figures S6 and S7), 20 μs is a reasonable upper limit for τ_f at 25 °C. At the bottom of the chevron plot, k_obs would be on the edge of observable with our stopped flow (using m_kf from our kinetic data at 4 °C, the minimum k_obs would be ~1850 s⁻¹). The fastest folding proteins have τ_f in the range of 0.5 – 10 μs at 20 – 25 °C.^{43,44,47–49,51,95,96} Based on the upper limit of ~20 μs for τ_f of UBA(1) at 25 °C, the folding time of UBA(1) is near this range.

The magnitude of k_f has been observed to decrease as the TS for folding becomes more compact (i.e. as β_T increases).⁹³ However, contrary to this expectation, the TS for UBA(1) folding is compact based on the β_T of ~0.8. Other fast folding three-helix bundle domains also have β_T values >0.7: B-domain of protein A, β_T ≈ 0.8;^42,43,48 protein GA, β_T ≈ 0.7 (mechanical unfolding);⁹⁶ and engrailed homeodomain, β_T ≈ 0.8.⁵¹ For the three-helix-bundle villin headpiece domain, the lnk_u vs 1/T dependence is linear, whereas that of lnk_f vs 1/T is curved indicating that most of the heat capacity change occurs before the TS in the folding direction, consistent with a TS of native-like compactness.⁵⁰ At least for small three-helix bundles, a compact TS may promote fast folding.

Residual α-helical Structure in the Denatured State of UBA(1) Versus the H1 Peptide as a Function of GdnHCl Concentration.

Based on the equilibrium unfolding data in Figure 3, UBA(1) is 97.8% unfolded in 4 M GdnHCl and 99.7% unfolded at 5 M GdnHCl. Thus, NMR and CD data acquired at 4 M and above can be used to evaluate residual structure in the denatured state of UBA(1). By 7 M GdnHCl, UBA(1) is >99.99% unfolded and thus represents a good approximation of the random coil reference state for UBA(1). McConnell’s treatment of the effects of exchange on NMR linewidths⁹⁷ produces linewidths consistent with those observed in Figure S14 at 4 M GdnHCl. All residues with linewidths >10 Hz have large differences in ¹⁵NH chemical shift of between 4 and 5 ppm for the denatured state relative to the native state of the UBA(1) domain.⁹⁸ If we extrapolate the upper limit for τ_f of ~20 μs at 25 °C for UBA(1) to 4 M GdnHCl using m_kf obtained at 4 °C, a 2.2% population of the native state would produce the linewidth increases observed for 4 M GdnHCl relative to 5 M GdnHCl for these resonances. Thus, the increased linewidths for some resonances at 4 M GdnHCl can be attributed to exchange with the native state.

To quantify residual structure in the denatured state of UBA(1) at 4, 5 and 6 M GdnHCl relative to 7 M GdnHCl as a random coil reference state, we have used the ¹³Cα secondary chemical shifts for each of the 20 amino acids in the RefDB database (Table 2).⁹⁹ We have chosen to use ¹³Cα chemical shifts for quantification of residual structure in the denatured state because the sensitivity of random coil chemical shifts to GdnHCl concentration, as reported for nine amino acids at pH 5,¹⁰⁰ is small for ¹³Cα (−27 ± 21 ppb/M) compared to ¹³CO (−86 ± 16 ppb/M) and ¹⁵NH (116 ± 27 ppb/M). The relative insensitivity of random coil ¹³Cα chemical shifts to GdnHCl concentration is also evident from the behavior of the ¹³Cα chemical shifts of the STLV sequence at the N-terminus of UBA(1), which is disordered in the native state (Figure 1). The SCS of these residues change very little for ¹³Cα (Figure 6) compared to the SCS values for ¹³CO and ¹⁵NH (Figures S11 and S12) as GdnHCl concentration increases. The GdnHCl dependence of random coil chemical shifts reported at pH 5 for Thr (−53 ppb/M), Leu (−33 ppb/M) and Val (−42 ppb/M) in GGXGG peptides¹⁰⁰ are higher than what we observe (Figure 6) for these residues at the N-terminus of UBA(1) with respective values of −2.8 ppb/M, 2.9 ppb/M, and −14.6 ppb/M for UBA(1) and −3.9 ppb/M, −4.0 ppb/M and −19.3 ppb/M for the H1 Peptide. These abated solvent dependencies are close to our estimated error in determining chemical shift values and may reflect the higher pH used here or possibly the higher solvent accessibility expected for backbone atoms within GGXGG model peptides versus the backbone atoms of a denatured protein.

Table 2.

Percent Residual α-Helix in the Denatured State of UBA(1) and the H1 Peptide Calculated from ¹³Cα SCS at 25 °C.^a

[GdnHCl]	UBA(1)			H1 Peptide
	Helix 1	Helix 2	Helix 3	Helix 1
4.0 M	6.8%	5.9%	3.3%	5.1%
5.0 M	4.1%	3.9%	2.1%	2.9%
6.0 M	2.0%	1.9%	1.1%	1.3%

^a13Cα SCS referenced to 7 M GdnHCl shown in Figure 6 were converted to % helix using RefDB SCSs and then averaged over each helical region. The range of each helix is based on the X-ray structure of UBA(1) as outlined in Figure 6. The error is estimated to be ±0.2% based on propagation of the error in the chemical shifts from the triplicate data acquired for UBA(1) at each concentration of GdnHCl.

Using the ¹³Cα SCSs from refDB,⁹⁹ we observe significant residual helical structure in all three helices of UBA(1) at 4, 5 and 6 M GdnHCl (Table 2). The residual helical structure in helices 1 and 2 of UBA(1) is 6 – 7% in 4 M GdnHCl, about double that of helix 3. As GdnHCl concentration increases to 5 and 6 M, helical content drops, with the amount of residual helical structure in helices 1 and 2 remaining about double that of helix 3 at all GdnHCl concentrations. The residual structure in helix 1 of UBA(1) is 1 – 2% greater than in the H1 Peptide at all GdnHCl concentrations. This observation is consistent with transient long-range interactions between helix 1 and the C-terminal half of UBA(1) in the denatured state enhancing the degree of residual helix formation in helix 1. Persistent tertiary interactions in protein denatured states have been detected previously by paramagnetic relaxation enhancement,^{8,9,14,101,102} residual dipolar coupling^10,103,104 and mutation-induced changes in secondary chemical shifts.⁶⁹

To verify the magnitude of residual structure determined by ¹³Cα SCSs, we have also evaluated residual α-helix and PP_II structure using the CD data shown in Figure 7 (Table S11). Analysis of the CD data in Figure 7 suggests that UBA(1) globally contains up to 9.2% α-helix on the cusp of folding at 4 M GdnHCl, which reduces to 4.2% α-helix under strong denaturing conditions at 7 M GdnHCl (Table S11). At high concentrations of GdnHCl (≥ 4 M), the helical content of UBA(1) is substantially higher than what is expected from the two-state unfolding model used to analyze global unfolding of UBA(1), which notably predicts less than a half-percent helical content at concentrations of 5 M GdnHCl and above. Additionally, a steady increase in PP_II content is observed as GdnHCl increases as has been previously observed in small proteins and peptides denatured with chaotropic agents.^89–91,105 It has been postulated that the concurrent decrease of α-helix and increase of PP_II helix as GdnHCl concentration increases may be mediated through the perturbation of the equilibrium between the two secondary structures.¹⁰⁶ The H1 Peptide behaves similarly to UBA(1), but has distinctly less α-helix and more PP_II relative to UBA(1) (Figure S16, Table S11) under highly denaturing conditions.

Because we used 7 M GdnHCl as a reference state for evaluation of α-helix content by NMR in the denatured state (Table 2), the α-helical content derived from CD data must be referenced to 7 M GdnHCl to allow direct comparison of helical content determined by the two methods. Table 3 compares the results of these estimates of % helix as a function of GdnHCl from CD and NMR data. Overall, the agreement between CD and NMR data with regard to residual α-helical structure in the denatured state of UBA(1) and the H1 Peptide is very good. Some of the small discrepancies between the % helix values derived from NMR and CD data may result from overlap of the PP_II signal with the α-helical signal in CD spectra, which is not accounted for in the equation we used to evaluate % helix from CD data (Table S11). This issue is complicated by the sequence-dependent shift in λ_max away from the canonical 230 nm for PP_II toward 222 nm depending on the number of proline residues in the sequence.^89,107,108 In addition, there is disagreement over whether or not PP_II causes significant secondary chemical shifts, which makes it difficult to determine the effect of the presence of PP_II on estimating residual α-helical content under denaturing conditions.^109,110 These issues aside, the good agreement between CD and NMR estimates of residual α-helical structure relative to the 7 M GdnHCl reference state indicates that the estimates of residual α-helical structure for each of the individual helices in Table 2 are reasonable. The good correspondence also indicates that the SCS compiled in RefDB provide reliable values for 100% helical content for a given residue, despite the significant standard deviations on many SCS values.⁹⁹

Table 3.

Percent Residual α-Helix for UBA(1) and the H1 Peptide Derived from CD and NMR data at 25 °C with 7 M GdnHCl as a Random Coil Reference State.

	% Helix
	UBA(1)		H1 Peptide
[GdnHCl]	CDa	NMRb	CDa	NMRb
4.0 M	4.0 ± 0.3	4.3 ± 0.1	2.4 ± 0.4	3.4 ± 0.2
5.0 M	1.6 ± 0.3	2.7 ± 0.1	1.5 ± 0.8	1.9 ± 0.2
6.0 M	0.3 ± 0.3	1.4 ± 0.1	0.6 ± 0.6	0.8 ± 0.2

^aDifference values (X M – 7 M) are calculated with the values for % helix in Table S11. Reported errors are from standard propagation of error.

^bFractional helicity was calculated as Cα SCS divided by RefDB statistical SCS⁹⁹ averaged over all residues and converted into a percent. Error is propagated from the standard deviation of the chemical shifts from three independent data sets for UBA(1). For estimating error in the % helix of the H1 Peptide, the error in chemical shifts was assumed to be the same as for UBA(1).

Application of Helix-Coil Theory to the GdnHCl Dependence of Residual Helical Structure in the Denatured State of UBA(1) and the H1 Peptide.

The NMR data in Figure 6 show that residual helical structure is limited to the residues within the boundaries of the three helices in the native-state structure of UBA(1). Therefore, in applying helix-coil theory to the CD data for the H1 Peptide and the denatured state NMR data for the H1 Peptide and UBA(1), we have assumed that helical structure is confined to the residues that are helical in the native state X-ray structure. For the CD data of the H1 Peptide, we assume that helical structure is restricted to the 14 residues that correspond to Thr159 to Ser172, and the fractional helicity of this segment calculated accordingly (see Supporting Methods). We used a simplified Zimm-Bragg helix-coil model^111,112 that assumes a linear dependence of the free energy of helix propagation on GdnHCl concentration and a uniform helix propagation parameter for all residues (Supporting Methods).^113,114 Figure 8A shows two fits of the Zimm-Bragg model to a plot of f_helix from CD data versus GdnHCl concentration for the H1 Peptide. One fit includes all data points while the other omits the 0 M GdnHCl data point. When the 0 M data point is removed, the correlation coefficient increases significantly, suggesting that the 0 M data point deviates from the helix parameters required to fit the remaining data. Charge screening of the negatively-charged H1 Peptide by GdnH⁺ could be the cause of this deviation. GdnH⁺ may shield the electrostatic repulsion of the 3 glutamate side chains within the helical region of the H1 Peptide stabilizing the helical conformation at low GdnHCl concentrations. For charged helical proteins and peptides,^115–117 charge screening has been observed to promote helix stabilization at GdnHCl concentrations up to about 2 M.^115,118 Although the fit without the 0 M data point is better, the helix propagation parameter at 0 M GdnHCl, s_o, and the denaturant m-values from the two fits are in close agreement (Table 4). The m-value obtained is similar to that obtained for urea denaturation of alanine-based peptides (0.023 kcal mol-res⁻¹ M⁻¹),¹¹³ but less than that obtained for GdnHCl denaturation of alanine-based peptides (0.043 kcal mol-res⁻¹ M⁻¹).¹¹⁴ However, about half of the magnitude of the GdnHCl m-value is due to Cl⁻,¹¹⁴ which may depend on specific side chain electrostatic interactions.¹¹⁹ The s_o value is somewhat less than that for alanine-based peptides (s_o = 1.34).¹¹³ Thus, the H1 Peptide has strong but somewhat lower helical propensity than alanine-based peptides.

Figure 8.

(A) f_helix derived from CD (open inverted triangles) and NMR data (closed triangles) for the H1 Peptide plotted versus GdnHCl concentration. The lines (dashed, all CD-derived data points, R² = 0.988; dotted, all CD-derived data points except 0 M GdnHCl, R² = 0.998; solid, NMR data points) are fits of the Zimm-Bragg helix-coil model to the data. (B) f_helix derived from NMR data for helices 1, 2 and 3 of UBA(1) and for the H1 Peptide plotted versus GdnHCl concentration. Solid curves are fits of the Zimm-Bragg helix-coil model to the data.

Table 4.

Zimm-Bragg Helix-Coil Parameters for Denatured State Residual Helical Structure in UBA(1) and the H1 Peptide.

Protein	Helix	n a	s o	m,b kcal mol-res−1 M−1
	CD-derived Parametersc,d
H1 Peptide	1	14	1.145 ± 0.003	0.0202 ± 0.0005
			(1.131 ± 0.005)	(0.018 ± 0.001)
	NMR-derived Parametersd
H1 Peptide	1	14	1.180 ± 0.001	0.0224 ± 0.0001
UBA(1)	1	14	1.23 ± 0.01	0.024 ± 0.001
UBA(1)	2	12	1.27 ± 0.02	0.023 ± 0.002
UBA(1)	3	12	1.16 ± 0.02	0.024 ± 0.002

^an is the number of residues in the helix and was used as the helix length in the Zimm-Bragg model.

^bFor the m-value, mol-res stands for moles of amino acid residues.

^cThe unbracketed values are from the fit excluding the 0 M GdnHCl data point. The values in brackets are from the fit with all data points.

^dStandard errors reported for s_o and m were determined from the goodness-of-fit of the model to experimental data.

To allow direct comparison between residual helical structure measured by CD and that detected by NMR SCS data, f_helix derived from NMR data with 7 M GdnHCl as a random coil reference state was corrected for the residual helical structure detected by CD methods at 7 M GdnHCl (Supporting Methods). For the H1 Peptide, the corrected f_helix values derived from NMR SCS data are very similar to those obtained from CD data. The fit of the Zimm-Bragg model to the data is also very similar (Figure 8A), and s_o obtained from the fit is nearly identical to that obtained from CD-derived data (Table 4).

The f_helix data in the denatured state of helices 1, 2 and 3 for UBA(1) derived from NMR SCS, corrected in a similar manner (Supporting Methods), are shown in Figure 8B. The magnitude of f_helix for helices 1 and 2 is similar and about twice that of helix 3 across the range of 4 to 7 M GdnHCl. Notably, f_helix for helix 1 in the context of UBA(1) is much greater than that of helix 1 within the H1 Peptide in 4 to 7 M GdnHCl (Figure 8B). Strictly speaking, helix-coil theory developed for isolated helices cannot be applied to the full UBA(1) protein if there are helix-helix interactions in the denatured state. However, helix-helix interactions are expected to be transient and thereby the perturbation to helix-coil theory will be small. Thus, to a first approximation, helix coil theory provides a reasonable way to evaluate the stabilization of helix 1 caused by the coupled equilibrium of docking the isolated helix onto adjacent transient helices in the denatured state of UBA(1). The apparent s_o value for helix 1 in the denatured state of UBA(1) yields a measure of the magnitude of the stabilizing effect of transient interactions with the rest of the protein. We have previously analyzed the effects of stabilization of helical structure in short alanine-based peptides by metal binding using helix-coil theory.¹²⁰ The primary effect is an apparent increase in the helix-coil propagation parameter lending support for use of this approach here.

The Zimm-Bragg helix-coil model was fit to the NMR-derived f_helix values allowing extrapolation of residual structure observed under denaturing conditions to that expected under folding conditions. The values of the apparent s_o values and m from the fits are given in Table 4. The model predicts 33 ± 1%, 28 ± 2% and 17 ± 2% residual helical structure for helices 1, 2 and 3, respectively, in the denatured state of UBA(1) under folding conditions in 0 M GdnHCl (Figure 8B). For helix 1 in the H1 Peptide, the NMR-derived data extrapolates to 25.9 ± 0.2% helix in 0 M GdnHCl, which compares well with the helical content of 21.7 ± 0.2% obtained from the fit to the CD data for 1 to 7 M GdnHCl (Figure 8A). The 25 – 50% increase in denatured state residual helical structure in helix 1 when the complete sequence of UBA(1) is present is consistent with stabilization of helical structure by transient long-range interactions in the denatured state of UBA(1).

Plots of f_helix versus sequence position for UBA(1) and the H1 Peptide at 4 M, 5 M and 6 M GdnHCl using 7 M GdnHCl as a random coil reference state show the pattern expected from helix-coil theory indicating that its use to analyze our data is reasonable (Figure S17). In particular, at all GdnHCl concentrations, f_helix is largest near the center of each helix and smallest at the ends of each helix. The f_helix determined by the fit of the Zimm-Bragg model to the data follows the experimental trend well at each GdnHCl concentration indicating the model, despite its simplicity, provides a reasonable representation of the data.

As noted above, at 4 M GdnHCl, the population of the native state is estimated to be 2.2%. It is possible that exchange with the native state could affect our evaluation of f_helix for the denatured UBA(1) in 4 M GdnHCl. For ¹³Cα resonances, using τ_f = 20 μs and McConnell’s treatment of the effects of exchange on NMR linewidths and chemical shifts,^{97 13}Ca resonances will be in intermediate exchange, which reduces the effect of exchange on chemical shift relative to what would be observed in the fast exchange regime. We estimate that exchange with the native state could contribute approximately 0.007 to the apparent f_helix for helix 1 at 4 M GdnHCl. If we reduce f_helix by this amount at 4 M GdnHCl for helix 1, we obtain m = 0.022 ± 0.002, and an apparent s_o of 1.21 ± 0.02 and helical content at 0 M GdnHCl of 30 ± 3 %. All parameters are within error of those obtained without this correction (Table 4). The correction reduces the increase in the helicity of helix 1 of UBA(1) under folding condition relative to the H1 peptide to 12 – 38 %.

The increase in residual structure in the denatured state as GdnHCl concentration decreases could affect the values of ΔG_u°′(H₂O) obtained from GdnHCl denaturation data because the denatured state will be less stable in the transition region than at 0 M GdnHCl. Using the fits to helix-coil theory in Figure 8, we estimate the denatured state will be stabilized by about 0.3 kcal/mol between C_m (~2,0 M GdnHCl) and 0 M GdnHCl. The increase in residual structure as GdnHCl concentration decreases should result in a progressive decrease in the m-value. The gradual decrease in the m-value as GdnHCl concentration decreases in the transition region, may partially compensate for the small overestimate in ΔG_u°′(H₂O) expected to result from linear extrapolation from the unfolding midpoint. Similar effects are expected for thermal denaturation data as residual helical structure will be less near T_m (340 K) than at 298 K. Despite these effects, similar stabilities at both 4 °C and 25 °C are obtained for UBA(1) from thermal and GdnHCl denaturation methods.

Implications for the Folding Mechanism of UBA(1).

Analysis of our CD and NMR data using the Zimm-Bragg helix-coil model indicates that UBA(1) has significant helical structure in the range of 17 – 33% in helices 1, 2 and 3 in the denatured state under folding conditions. The helical content of helix 1 may be somewhat overestimated because of electrostatic repulsion at lower salt concentration, however, the resulting decrease in helical content is expected to be small (Figure 8A). Positive and negative residues are more evenly distributed in helices 2 and 3. Under low salt folding conditions, the i, i+3 E176/R179 and Arg193/Glu196 interaction in helices 2 and 3, respectively, should be stabilizing in the denatured state under folding conditions.^121,122 For these helices, residual helical structure under folding conditions may be underestimated. The high helical content in the denatured state of UBA(1) would be expected to promote a diffusion-collision mechanism for folding, where helices form first then collide and stabilize each other.¹²³ This mechanism has been demonstrated for several helical bundles containing helices with high intrinsic helical propensity.^{45,51,124,125}

The availability of residue resolution data on residual structure in the denatured state of both UBA(1) and the H1 Peptide provides the opportunity to determine which parts of helix 1 are stabilized the most under denaturing conditions for UBA(1) versus the H1 Peptide. In the diffusion-collision model of folding, the most stabilized region of helix 1 should correspond to contact points between helix 1 and other helices of UBA(1). Figure 9A shows a plot of the difference in f_helix, Δf_helix, between UBA(1) and the H1 Peptide at 4 M, 5 M and 6 M GdnHCl for helix 1. Positive values of Δf_helix are observed for each residue of helix 1 at all concentrations of GdnHCl, corresponding to a relative average increase of 35 ± 5% at 6 M GdnHCl to 56 ± 22% at 4 M GdnHCl in the magnitude of f_helix across the length of the helix. At 4 M GdnHCl, increases in Δf_helix are biased toward the C-terminus of helix 1. In UBA(1), the most buried residues in helix 1 reside at the C-terminal end of the helix and include residues Leu167 and Ile170 with solvent-accessible surface areas of < 1 Ų and 10 Ų, respectively, as determined by PISA.⁸² Notably, Leu167 exhibits the largest Δf_helix value at 4 M GdnHCl (relative increase in f_helix of 80 ± 20%) and is thus most perturbed by truncation of UBA(1). In the native state of UBA(1), the Leu167 side chain makes numerous contacts, including intrahelical side-chain packing with i ± 3 and i ± 4 residues Tyr163, Glu164, and Met171, and interhelical side-chain interactions with helix 2 residues Arg177, Val180, Val181, and Leu184 (Figures 9B and 9C). No direct contacts are made between Leu167 and the residues of helix 3. Instead, nearby residues Met166 and Ile170 directly interact with helix 3 through N-capping residue Pro191 and helix 3 residue Val195, respectively. It is therefore possible that helix 1 C-terminal residues Met166 – Ile170, in particular Leu167, make transient, native-like hydrophobic interactions with the side chains of Val and Leu residues at the N-terminus of helix 2, thereby stabilizing local helical structure in the latter half of helix 1. Interestingly, Leu167 also exhibits the highest Δf_helix at 5 M GdnHCl while other residues, such as Met166 and Ile170, no longer have as prominent Δf_helix values. This observation suggests that residual tertiary interactions involving helix 2 also exist at 5 M GdnHCl and are largely mediated by Leu167. Val and Leu residues are prominent in the contact matrix of UBA(1). ILV clusters involving branched, aliphatic side-chains have been suggested to be cores of stability in partially-unfolded states¹²⁶ and may be a major contributor to initiation of tertiary folding in UBA(1). In our recent study using His-heme loop formation to evaluate non-random structure in the denatured state of UBA(1),¹²⁷ we showed that UBA(1) has a large ILV cluster capable of stabilizing residual structure near both turns. The first turn of UBA(1) is one of the few sequence regions of UBA domains that is conserved across evolution.²⁹ It is, therefore, possible that this conserved turn sequence promotes transient interactions between helix 1 and helix 2 that stabilize residual structure in helix 1 in UBA(1) relative to the H1 Peptide. However, our current data only provide indirect support for this possibility.

Figure 9.

(A) Per-residue Δf_helix values for the N-terminal section of UBA(1) relative to H1 Peptide (f_helix,UBA(1) – f_{helix,H1 Peptide}) at 4 M (red), 5 M (green) and 6 M GdnHCl (blue). (B) Residues surrounding Leu167 (transparent green spheres) from helix 1 in the hydrophobic core of UBA(1). (C) C_α-C_α distance heat map for UBA(1). The close contacts between Leu167 and Arg177 and Val 180/Val181 are highlighted with a red square and a red rectangle, respectively.

Based on the native state structure of UBA(1), we have emphasized hydrophobic interactions as a possible cause of the stabilization of helix 1 through transient interactions with helix 2. However, molecular dynamics simulations of an 8-residue alanine peptide,¹²⁸ show that a nearby helical template increases the helical content of the (Ala)₈ peptide by reducing the effective dielectric near the template and thereby strengthening electrostatic interactions, which stabilize the hydrogen bonds of the (Ala)₈ helix. Thus, stabilization of helix 1 of UBA(1) in the denatured state could, in part, result from a similar effect when it is near helix 2 in a helical conformation.

Analysis by Chemical Shift Perturbation Leads to Similar Conclusions.

As an alternative approach, we can look at the helix 1 residues from the perspective of chemical shift perturbations (CSPs), wherein we define Δδ_CSP = δ_UBA(1) – δ_H1Pep. This approach allows us to extend the analysis to 7 M GdnHCl and provides a crosscheck on the conclusions from the Δf_helix analysis based on SCS referenced to 7 M GdnHCl in Figure 9. Δf_helix values derived from CSPs, Δf_helix,CSP, are obtained by dividing Δδ_CSP by the magnitude of the α-helix secondary shift for the amino acid in RefDB.⁹⁹ This analysis diminishes the magnitude of Δf_helix for helix 1 in UBA(1) versus the H1 Peptide in 4 M GdnHCl (Figure S18). However, qualitatively, the conclusions are the same. The C-terminal end of helix 1 is more stable in UBA(1) compared to the H1 Peptide. The increase in Δf_helix,CSP is largest for Leu167 in 4 M GdnHCl, but is noticeably diminished for Glu164 and Ser172 compared to Δf_helix values derived from SCS (Figure 9). The reason for this discrepancy can be observed in the Δf_helix,CSP values at 7 M GdnHCl (Figure S18). Here, Δf_helix,CSP values are negative indicating an overall difference in residual structure between the two 7 M GdnHCl reference states for UBA(1) and the H1 Peptide, which we use to determine Δf_helix values. Given the current state of understanding of the SCS of PP_II structure,^109,110 a detailed analysis of the differences in the reference states for UBA(1) versus the H1 Peptide is not possible beyond what CD data provide (Table S11). Analysis of SCSs using either Δf_helix or Δf_helix,CSP lead to the same conclusions regarding the observed increase in residual structure in helix 1 for UBA(1) relative to the H1 Peptide.

CONCLUSIONS

We have characterized the kinetics and thermodynamics of folding of the small three-helix bundle, UBA(1), and the residual structure in its denatured state. UBA(1) has modest stability, but folds on the μs time scale even at 4 °C. NMR data obtained in 4 to 7 M GdnHCl show that all three helices retain significant residual structure in the denatured state that progressively increases as the denaturing conditions are weakened. Extrapolating the f_helix for residual structure in each helix using helix-coil theory shows that the population of helical structure in the helices of UBA(1) likely ranges from 17 to 33 percent at 0 M GdnHCl. Thus, under folding conditions, the strong structural bias in the denatured state, which may be promoted by hydrophobic interactions, will play an important role in directing formation of the native state of UBA(1). The degree of residual helical structure in UBA(1) suggests that it likely folds by a diffusion-collision mechanism,^42,123,124 although further kinetic studies would be necessary to directly demonstrate this folding mechanism.

ABBREVIATIONS

β-ME

β-mercaptoethanol

CD

Circular dichroism

CSP

chemical shift perturbation

ΔG_u°′(H₂O)

standard free energy of unfolding in the absence of denaturant at pH 6.5

f_helix,i

fraction helix for residue i

FRET

Förster resonance energy transfer

GdnHCl

guanidine hydrochloride

GST

glutathione S-transferase

H1 Peptide

UBA(1) with a stop codon in place of Arg177, containing only the first helix

HHR23A

human homolog of the Saccharomyces cerevisiae Rad23 DNA excision repair protein

IPTG

isopropyl β-d-1-thiogalactopyranoside

MALDI-TOF

matrix-assisted laser desorption/ionization time-of-flight

PP_II

polyproline II secondary structure

SAXS

small-angle X-ray scattering

SCS

secondary chemical shift

T_m

melting temperature

TS

transition state

UBA(1)

HHR23A ubiquitin-associated domain 1