Helical Propensity Affects the Conformational Properties of the Denatured State of Cytochrome c′

Abstract

Changing the helical propensity of a polypeptide sequence might be expected to affect the conformational properties of the denatured state of a protein. To test this hypothesis, alanines at positions 83 and 87 near the center of helix 3 of cytochrome c′ from Rhodopseudomonas palustris were mutated to serine to decrease the stability of this helix. A set of 13 single histidine variants in the A83S/A87S background were prepared to permit assessment of the conformational properties of the denatured state using histidine-loop formation in 3 M guanidine hydrochloride. The data are compared with previous histidine-heme loop formation data for wild-type cytochrome c′. As expected, destabilization of helix 3 decreases the global stabilities of the histidine variants in the A83S/A87S background relative to the wild-type background. Loop stability versus loop size data yields a scaling exponent of 2.1 ± 0.2, similar to the value of 2.3 ± 0.2 obtained for wild-type cytochrome c′. However, the stabilities of all histidine-heme loops, which contain the helix 3 sequence segment, are increased in the A83S/A87S background compared to the wild-type background. Rate constants for histidine-heme loop breakage are similar for the wild-type and A83S/A87S variants. However, for histidine-heme loops that contain the helix 3 sequence segment, the rate constants for loop formation increase in the A83S/A87S background compared to the wild-type background. Thus, residual helical structure appears to stiffen the polypeptide chain slowing loop formation in the denatured state. The implications of these results for protein folding mechanisms are discussed.

Introduction

There has been considerable debate over whether formation of secondary structure promotes or inhibits efficient protein folding (1, 2, 3). Initial collapse of a protein from a highly disordered structure has been shown to promote formation of secondary structure (4, 5), suggesting that collapse to a nativelike topology may precede secondary structure formation. Furthermore, the work of Fersht (6) supports a nucleation-condensation model for protein folding where a diffuse nucleus of sequentially distant residues establishes the topology of a protein at the transition state. However, there is also considerable experimental evidence in support of the diffusion-collision model (7, 8, 9, 10). This model posits that flickering elements of secondary structure, which anneal to form tertiary contacts at the transition state, are essential for efficient folding (11). A related model, the zipping and assembly model (12, 13), is based on the observation that folding rates are proportional to the contact order of the native state (14). This model, which assumes short-range interactions (helical turns, β-turns) form first, followed by the assembly of the tertiary structure by agglomeration of these elements, successfully predicts protein folding ϕ-values (12, 13, 15, 16). The success of these various contrasting models suggests that for different proteins, tertiary interactions or formation of secondary structure may provide equally good solutions for efficient protein folding (17).

The ϕ-value analysis studies, a primary method for assessing structure in the transition state for protein folding, typically assume that the denatured state is highly disordered with minimal residual structure (6). Even though many studies are consistent with denatured proteins having the properties of a random coil (18, 19), there is substantial evidence that denatured states of proteins retain residual structure, which can either be nativelike or nonnative in nature (20, 21). Amino acid substitutions can affect the thermodynamics of the denatured state (22, 23, 24, 25, 26, 27, 28, 29, 30), which could affect the interpretation of ϕ-values (31). Thus, a thorough understanding of the conformational properties of the denatured states of proteins is important for understanding the process of protein folding.

There are numerous examples of residual helical structure in denatured proteins because helix content is readily evaluated with chemical shifts determined by NMR spectroscopy (32, 33, 34, 35). It is also evident that hydrophobic clusters can persist (20, 21, 36, 37, 38, 39, 40), even under highly denaturing conditions (41). For acid-denatured proteins (35, 42) and in cases where the unfolded state can be studied in the absence of denaturing agents (43, 44, 45), significantly more residual structure is evident. There is also evidence that long-range tertiary interactions exist in disordered and denatured states of proteins (46, 47, 48, 49, 50, 51, 52), which are responsive to amino acid substitutions (53, 54, 55). However, some proteins show less evidence of long-range interactions in the denatured state (56). Recent work on the N-terminal domain of ribosomal protein L9 (43, 57) shows that long-range tertiary interactions in the denatured state are not dependent on the presence of significant residual secondary structure. This observation raises the question of how the conformational properties of the denatured state are affected by residual secondary structure. In particular, does the presence of residual secondary structure favor or disfavor contacts between distant parts of the protein primary structure in the denatured state? And does it matter if the points of contact contain the element of residual secondary structure?

To address these questions, we have extended our work on loop formation in the denatured state of Rhodopseudomonas palustris cytochrome c′ (Cytc′) (58, 59). Cytc′ is a four-helix bundle protein containing four long ∼20 residue helices (60) (Fig. 1). Cytc′ also contains a c-type heme and thus is readily adapted to study the conformational properties of the denatured state using His-heme loop formation methods (22, 58, 59, 61, 62, 63, 64, 65, 66, 67). Cytc′ is believed to have a compact denatured state and thus tertiary interactions are likely (68). Helix 3 of Cytc′ is predicted by Agadir to have considerably higher intrinsic helical propensity than the other helices (see below). To study the effect of helical propensity on the conformational properties of the denatured state, a variant of Cytc′, which strongly reduces the intrinsic helical propensity of the helix 3 sequence segment (residues 75–95), has been prepared. The same set of single histidine variants used in the previous studies on His-heme loop formation in the denatured state of Cytc′ were prepared. Our results show that, if the helix 3 sequence segment is contained within the denatured state loop, the His-heme loop forms more readily in 3 M guanidine hydrochloride (GdnHCl) when the helical propensities of residues in helix 3 are reduced. For His-heme loops that do not contain the helix 3 sequence segment, the effects are minimal.

Figure 1.

Structure of R. palustris Cytc′ (PDB: 1A7V) (60) showing the positions of histidine and serine substitutions. The heme is shown as a stick model colored by element. Substitution sites are indicated with spheres at the positions of the β-carbons. Amino acid substitutions are indicated with standard notation next to each sphere. To see this figure in color, go online.

Materials and Methods

Preparation of Cytc′ variants by mutagenesis

All primers were designed using the QuikChange Primer Design tool (Agilent Technologies, Santa Clara, CA; see Table S1). Alanine to serine substitutions were introduced first at positions 83 and 87 using the pETcp vector (69) containing the pseudo wild-type (pWT, Q1A substitution) gene for Cytc′. The Q1A substitution is present to avoid cyclization of the N-terminal glutamine to pyroglutamate (69). The single histidine substitutions were then introduced using the pETcp vector (69) carrying the gene for pWT Cytc′ with the A83S/A87S substitutions (H3_αLo Cytc′). Site-directed mutagenesis was performed using the QuikChange II PCR-based mutagenesis kit (Agilent Technologies). Mutations were confirmed by dideoxy sequencing at the Genomics Core Facility at the University of Montana.

Expression and purification of Cytc′

Competent BL21-DE3 Escherichia coli cells (EdgeBio, Gaithersburg, MD) were transformed with both pETcp (ampicillin resistance) and pEC86 (70) (chloramphenicol resistance). pEC86 provides the genes for the E. coli cytochrome c maturation operon, allowing for covalent heme attachment. A full plate of transformed colonies was resuspended in 3 mL of sterile LB medium and used to inoculate 1 L cultures of Terrific Broth (12 g/L tryptone, 24 g/L yeast extract, 4 mL glycerol, 2.3 g/L KH₂PO₄, and 12.5 g/L K₂HPO₄) containing 100 μg/mL ampicillin and 34 μg/mL chloramphenicol. These cultures were grown overnight at 30 or 37°C with shaking at 165–200 rpm. Cell pellets were harvested using a Sorvall GS-3 rotor in a Sorvall RC-5C+ floor model centrifuge (Thermo Fisher Scientific, Waltham, MA) running at 5000 rpm for 10 min. The cell pellet was stored at −80°C.

Cell were lysed by osmotic shock, as previously described after three freeze/thaw cycles (59). Protein purification was as described previously (58, 59). Briefly, the crude protein extract was diluted to 5 mM ionic strength and adjusted to pH 5.0 using glacial acetic acid. This solution was loaded onto a CM Sepharose (GE Healthcare Life Sciences, Marlborough, MA) column equilibrated with CM Low buffer (5 mM NaOAc, pH 5.0). Cytc′ was eluted using a 200 mL linear salt gradient from 0 to 500 mM NaCl using CM High buffer (5 mM NaOAc and 500 mM NaCl, pH 5.0). Protein solution eluted from the CM Sepharose column was exchanged into HPLC Buffer A (10 mM sodium phosphate, pH 6.0) by ultrafiltration. Final purification was by HPLC using a ProteinPak SP-8HR cation exchange column (Waters, Milford, MA). The gradient used was as follows: 0–10 min 0% B, 10–11 min increase linearly to 2% B, 11–31 min linear increase to 12% B, 31–45.9 min increase linearly to 15% B, 45.9–46 min increase linearly to 100% B, 46–58 min hold at 100% B, 58–58.1 min decrease to 0% B, and 58.1–68 min hold at 0% B (flow rate: 2 mL/min; HPLC buffer B: 10 mM sodium phosphate, 500 mM NaCl, pH 6.0). A MALDI-TOF mass spectrum was collected to check purity and identity of each variant. This gradient consistently produced protein with one peak and at the expected mass. A DU 800 Spectrophotometer (Beckman Coulter, Brea, CA) was used to check the protein concentration. As the spectrum is pH dependent, the concentration was measured in 100 mM sodium phosphate buffer at pH 7.0 using ε₃₉₈ = 85,000 M⁻¹ cm⁻¹ (71).

Global thermodynamic stability measurements

To measure global stability, GdnHCl denaturation monitored by circular dichroism (CD) spectroscopy was employed. Protein was exchanged into CD Buffer (20 mM MES, 40 mM NaCl, pH 6.5) using size exclusion chromatography (Sephadex G25; Sigma-Aldrich, St. Louis MO). GdnHCl titrations were carried out as described previously using a Microlab 500 titrator (Hamilton Robotics, Reno, NV) coupled to a Chirascan CD spectrometer (Applied Photophysics, Leatherhead, Surrey, United Kingdom) (72, 73). Ellipticity was monitored at 222 and 250 nm for 25 s. The 250 nm signal was subtracted as a background (θ_220corr = θ₂₂₂–θ₂₅₀). All data were acquired at 25°C.

Plots of θ_220corr versus GdnHCl concentration were fitted to a six-parameter equation that evaluates the slope and intercept of the native and denatured state baselines and assumes a linear free energy relationship for the dependence of the free energy of unfolding, ΔG_u, on GdnHCl concentration (74). ΔG_u°′(H₂O), the free energy of unfolding in the absence of denaturant and m, the rate of change of the free energy of unfolding, ΔG_u, with respect to GdnHCl concentration were obtained from these fits. The midpoint for unfolding, C_m, is evaluated as the ratio of ΔG_u°′(H₂O) to m.

Denatured state loop formation

His-heme loop formation for Cytc′ variants was measured in the denatured state (3.0 M GdnHCl, 5 mM Na₂HPO₄ and 15 mM NaCl) by pH titration, as previously described (63). Loop equilibria as a function of pH were monitored with the heme Soret band (350–450 nm). Titrations were carried out at room temperature (22 ± 1°C).

Corrected absorbance at 398 nm, A_398corr (= A₃₉₈–A₄₅₀), was plotted versus pH and fit to a modified Henderson-Hasselbalch equation,

$$A_{398}=\frac{A_{\text{LS}}+A_{\text{HS}}{10}^{n_p\left(\text{p}{K}_{\text{a}}\left(\text{obs}\right)-\text{pH}\right)}}{1+{10}^{n_p\left(\text{p}{K}_{\text{a}}\left(\text{obs}\right)-\text{pH}\right)}},$$ (1)

to extract the apparent pK_a of loop formation, pK_a(obs), and the number of protons, n_p, associated with loop formation. In Eq. 1, A_LS is the corrected absorbance at 398 nm at high pH where the His-heme loop is formed and the heme is low spin and A_HS is the corrected absorbance at 398 nm at low pH where the His-heme loop is broken and the heme is high spin.

Kinetics of His-heme loop breakage

To monitor the kinetics of His-heme loop breakage in the denatured state, stopped-flow mixing experiments were performed at 25 ± 0.1°C in 3.0 M GdnHCl using a SX20 stopped-flow spectrophotometer (Applied Photophysics). The dead time of the instrument was 0.7 ms as determined by the reduction of 2,6-dichlorophenolindophenol with ascorbate (75). The experiments involved 1:1 mixing of 6 μM protein in 10 mM MOPS, pH 6.8, and 3.0 M GdnHCl with a 100 mM sodium citrate buffer, pH 3.0 or 3.5, in 3.0 M GdnHCl to achieve the desired ending pH of 3.1 or 3.6 at 3 μM protein concentration. Loop breakage was monitored at 398 nm. The final pH of each experiment was measured from the reaction products (stopped flow effluent) immediately after the reaction. Protein was exchanged into 10 mM MOPS, pH 6.8, using a Sephadex G25 size exclusion chromatography column (Sigma-Aldrich) equilibrated with the MOPS buffer. The initial 6 μM protein sample in 3.0 M GdnHCl was prepared by diluting protein obtained from the Sephadex G25 column with 10 mM MOPS, pH 6.8 and ∼6 M GdnHCl (exact concentration obtained by refractive index measurements (76)) buffered with 10 mM MOPS, pH 6.8. Seven to fifteen stopped-flow runs were collected for each variant and the data were fit to a single exponential equation.

Fractional helicity prediction

Agadir (http://agadir.crg.es/) (77) was used to evaluate the relative tendency of regions of the primary sequences of pWT and H3_αLo Cytc′ to form helical structure. The Agadir calculations were done at 22°C, pH 7 and an ionic strength of 0.1. The per-residue fractional helix content of the most stable part of helix 3 (residues 80–91) was modeled using a homopolymer version of the simplified (2 × 2 matrix) version of the Zimm-Bragg model (78). The nucleation constant, σ, was set to 0.003 (79) and the propagation parameter, s, was adjusted to match the average fractional helix content of this segment of helix 3 (15.9%, s = 1.139) predicted by Agadir. To estimate the effect of GdnHCl on helix content, ln(s) was adjusted to 3 M using m = 0.0434 kcal mol⁻¹ M⁻¹ (79), yielding s = 0.913 in 3 M GdnHCl. This value for s was used to estimate residual helix content on a per-residue basis in the denatured state in 3 M GdnHCl for the central stable segment of helix 3 using the simplified Zimm-Bragg model.

Results

Design of Cytc′ variants

The Agadir algorithm predicts that the third long helix of Cytc′ is intrinsically more stable than the other three helices (Fig. S1 A). At 22°C, the temperature used for denatured state loop formation measurements, Agadir predicted an average percent helix of 11.2% for the helix 3 sequence segment (positions 75–95), peaking at >20% helix at position 83. The Agadir algorithm predicts that the other three helices of Cytc′ have far lower tendencies to form helix: helix 1 (residues 4–26), average percent helix of 2.1%; helix 2 (residues 34–54), average percent helix of 1.3%; and helix 4 (residues 99–120), average percent helix of 1.3% (Fig. S1 A). Therefore, to examine the effect of the preference of the primary sequence for helical structure on the conformational properties of the denatured state, substitutions to destabilize helix 3 were designed. In the segment of helix 3 running from residues 83 to 92, 7 of 10 residues are alanines, strongly favoring helical secondary structure (80). Ala⁸³ and Ala⁸⁷ are separated by one turn of helix near the center of the helix and thus substitution at these sites will maximally affect the intrinsic stability of helix 3 (81). Solvent exposure calculations with GETAREA 1.1 (http://curie.utmb.edu/getarea.html) (82), show that Ala⁸³ and Ala⁸⁷ are 78% and 86% solvent exposed, respectively. Thus, these residues can be used to reduce the intrinsic preference of the helix 3 sequence segment for helical secondary structure with minimal perturbation to tertiary interactions in the native state of Cytc′ (Fig. 1). Ala→Ser substitutions are used to substantially reduce helical propensity at these sites (80), while only modestly changing side-chain sterics. The Agadir algorithm predicts that the A83S/A87S substitutions will decrease the average helical content of the helix 3 sequence segment to 2.7% at 22°C, with helical content <5% for all residues in this sequence segment (Fig. S1 A). Thus, replacing alanines 83 and 87 of Cytc′ with serine will allow evaluation of the effect of the intrinsic helix-forming propensity of the primary structure, on the conformational properties of the denatured state.

For denatured state His-heme loop formation measurements in 3 M GdnHCl at 22°C, the fractional helicities are expected to be much lower than those predicted by Agadir in water at 22°C. The central segment of helix 3 (residues 80–91) is particularly stable at 22°C (average helix content of 15.9%). The alanine content of residues 80–91 is 50%. Thus, the helical properties of this sequence segment were evaluated with a homopolymer version of the simplified Zimm-Bragg model (Fig. S1 B). Adjusting the propagation parameter obtained in water, using the GdnHCl m-value reported for alanine-based peptides (79), indicates that the residual helical content of this sequence segment decreases to 4.6% in 3 M GdnHCl, with residues near the center of this sequence segment having fractional helix content >6% (Fig. S1 B). This predicted residual helical content is similar to that reported for denatured proteins in this range of GdnHCl concentration using NMR secondary shifts (32). A similar estimate of the effect of 3 M GdnHCl on the helix content of residues 80–91 with the A83S/A87S substitution indicates that the helical content drops from 3.2% in water to 1.3% in 3 M GdnHCl, with no individual residue exceeding 1.6% helical content.

To assess conformational properties in the denatured state, His-heme loop formation is used. The same set of single histidine variants used in our previous studies on the conformational properties of the denatured state of Cytc′ (58, 59) has been prepared in the A83S/A87S (H3_αLo) background to allow direct comparison with the conformational properties of the denatured state of pWT Cytc′. Sequence positions with high solvent accessibility were chosen for substitution with histidine (59) to minimize the effect of these substitutions on stabilizing tertiary interactions in the native state. This set of 13 variants (Fig. 1) allows measurement of His-heme loop formation for loop sizes ranging from 10 to 111 residues.

Global stability of Cytc′ variants

GdnHCl denaturation was used to evaluate the effect of destabilizing helix 3 with the A83S/A87S substitutions (H3_αLo variant) on the global stability of Cytc′. Fig. 2 shows that GdnHCl denaturation of H3_αLo Cytc′ occurs at lower GdnHCl concentration, consistent with the A83S/A87S substitutions significantly decreasing global stability, ΔG_u°′(H₂O), relative to pWT Cytc′ (Table 1). For single histidine variants in the H3_αLo background as His-heme loop size in the denatured state increases, ΔG_u°′(H₂O) tends to be closer to the thermodynamic properties of H3_αLo Cytc′. However, variants with shorter His-heme loops tend to be less stable. For example, the GdnHCl denaturation curve for H3_αLo K13H Cytc′ (loop size = 101) nearly overlays that of H3_αLo Cytc′, whereas the GdnHCl denaturation curve for H3_αLo E73H Cytc′ (loop size of 41 residues) is shifted toward lower GdnHCl concentration (Fig. 2).

Figure 2.

Representative GdnHCl denaturation data monitored by circular dichroism at 222 nm for single histidine variants of Cytc′ in the H3_αLo background relative to H3_αLo Cytc′. Previously published data for pWT Cytc′ are shown for comparison. Data were acquired at pH 6.5 and 25°C. To see this figure in color, go online.

Table 1.

Thermodynamic Parameters for GdnHCl Unfolding of Single Histidine Cytc′ Variants in the H3_αLo Background at pH 6.5 and 25°C

Single Histidine Variant	Loop Size	ΔGu°′(H2O) (kcal/mol)	m-Value (kcal/(mol∗M))	Cm (M)
A104H	10	5.25 ± 0.18 (6.08 ± 0.26)	5.81 ± 0.11 (5.59 ± 0.12)	0.90 ± 0.01 (1.09 ± 0.05)
K97H	17	5.88 ± 0.16 (6.80 ± 0.09)	5.52 ± 0.25 (5.28 ± 0.07)	1.07 ± 0.06 (1.29 ± 0.01)
A91H	23	5.97 ± 0.04 (7.07 ±0.09)	5.59 ± 0.05 (5.75 ± 0.20)	1.07 ± 0.01 (1.23 ± 0.03)
K84H	30	5.78 ± 0.24 (5.89 ± 0.34)	5.80 ± 0.30 (5.32 ± 0.32)	1.00 ± 0.01 (1.11 ± 0.01)
E73H	41	5.27 ± 0.09 (5.76 ± 0.13)	5.39 ± 0.07 (5.08 ± 0.14)	0.98 ± 0.01 (1.13 ± 0.05)
A66H	48	6.82 ± 0.11 (6.79 ± 0.09)	5.46 ± 0.06 (4.88 ± 0.06)	1.19 ± 0.01 (1.39 ± 0.04)
D58H	56	5.91 ± 0.35 (6.72 ± 0.09)	4.97 ± 0.33 (4.80 ± 0.11)	1.18 ± 0.01 (1.40 ± 0.02)
K49H	65	6.15 ± 0.08 (6.75 ± 0.10)	5.05 ± 0.04 (4.66 ± 0.07)	1.22 ± 0.01 (1.45 ± 0.01)
K39H	75	6.32 ± 0.09 (6.91 ± 0.15)	5.18 ± 0.09 (4.70 ± 0.10)	1.22 ± 0.02 (1.47 ± 0.01)
K31H	83	6.60 ± 0.14 (7.41 ± 0.29)	5.27 ± 0.08 (5.04 ± 0.20)	1.25 ± 0.02 (1.47 ± 0.03)
K20H	94	5.96 ± 0.34 (5.89 ± 0.13)	4.91 ± 0.29 (4.17 ± 0.09)	1.22 ± 0.01 (1.41 ± 0.01)
K13H	101	7.41 ± 0.42 (8.17 ± 0.11)	5.01 ± 0.29 (4.91 ± 0.07)	1.472 ± 0.003 (1.66 ± 0.05)
D3H	111	7.12 ± 0.30 (7.41 ± 0.17)	5.69 ± 0.25 (5.08 ± 0.15)	1.253 ± 0.003 (1.46 ± 0.05)
H3αLo	—	6.78 ± 0.19	4.56 ± 0.20	1.49 ± 0.02
pWT	—	8.81 ± 0.22	5.14 ± 0.13	1.71 ± 0.04

Parameters for single histidine variants in the pWT background are given in brackets and are from Rao et al. (59).

The midpoint for GdnHCl denaturation, C_m, is lower for single histidine variants in the H3_αLo background as compared to the pWT background, and in general, C_m decreases as the loop size becomes smaller (Fig. S2; Table 1). The change in C_m, ΔC_m, for the H3_αLo background versus the pWT background is 0.20 ± 0.04 M (Fig. 3). For single histidine substitutions in or near helix 3 (E73H, K84H, A91H), the shift to lower C_m appears to be smaller. The largest deviation from the ΔC_m average occurs for the K84H substitution in the center of helix 3, which yields ΔC_m of only 0.11 ± 0.01 M.

Figure 3.

Changes in C_m, ΔC_m, for Cytc′ single histidine variants in the pWT versus the H3_αLo backgrounds. Histidine-heme loop sizes are given in parentheses. To see this figure in color, go online.

The global stability follows the same broad trend; ΔG_u°′(H₂O) decreases as loop size decreases (Fig. S3; Table 1). Most variants in the H3_αLo background show a decrease in the free energy of unfolding, ΔG_u°′(H₂O), compared to the pWT background, which varies from 0.5 to 2 kcal/mol. However, the D3H, K20H, A66H, and K84H variants have similar ΔG_u°′(H₂O) in both backgrounds (Table 1).

Denatured state loop formation equilibria

Cytc′ has heme covalently bound though a CXXCH heme attachment motif, which permits measurement of His-heme loop formation in the denatured state. Each single histidine variant places a histidine at a specific sequence distance from the closest point of heme attachment (Cys¹¹³ in Cytc′), allowing determination of the stability of His-heme loops as a function of loop size. The ionizable side chain of histidine allows loop stability to be measured with a simple pH titration experiment under denaturing conditions using absorbance in the heme Soret region (Fig. 4 A; Fig. S4). The spin state change from low spin (when histidine is bound to the heme near neutral pH) to high spin (when water is bound at low pH) causes a shift in the Soret band from ∼408 to ∼396 nm as pH is lowered from approximately pH 8 to approximately pH 3. An isosbestic point is initially observed at ∼404 nm, indicating that loop formation is two-state. Below pH 3.2, the isosbestic point shifts to ∼399 nm. The shift in isosbestic point likely results from protonation of His¹¹⁷ as the His¹¹⁷-heme bond breaks (see Fig. 4 A). Titration data are fit to Eq. 1 (Materials and Methods) yielding an apparent pK_a, pK_a(obs), as a measure of loop stability. All variants are completely unfolded in 3.0 M GdnHCl (Fig. 2; Table 1), the conditions used to monitor denatured state loop formation.

Figure 4.

Denatured state loop formation with Cytc′ variants. (A) Given here is a schematic representation of His–heme loop formation in the denatured state of Cytc′. (B) Given here are representative denatured state pH titration curves for histidine-heme loop formation at 22 ± 1°C in 3.0 M GdnHCl. Solid curves are fits to Eq. 1 in Materials and Methods. To see this figure in color, go online.

The primary structures of pWT and H3_αLo Cytc′ contain only the histidine of the heme attachment site (His¹¹⁷, Fig. 4 A). Loop formation is still observed due to lysine ligation (83). The pK_a(obs) is near 7.3 for both pWT and H3_αLo Cytc′ (Fig. 4 B; Table S2), providing an upper limit for the range over which pK_a(obs) can be measured. Single histidine variants in the H3_αLo background show titration curves shifted toward lower pH (Fig. 4 B). As with global stability, the variants which form the longest His-heme loops yield pK_a(obs) most similar to H3_αLo Cytc′ (Fig. 4 B, H3_αLo D3H, loop size = 110). As loop size decreases, loop stability increases (i.e., pK_a(obs) decreases; see Fig. 4 B, H3_αLo A66H, loop size = 49; H3_αLo E73H, loop size = 41).

Equilibrium loop formation can be broken down into a two-step process: histidine deprotonation followed by binding of the deprotonated histidine to the Fe³⁺ of the heme. Thus, pK_a(obs) is the sum of the intrinsic pK_a for histidine ionization, pK_a(HisH⁺), and the pK for His-heme loop formation with the deprotonated histidine, pK_loop(His):

$$\text{p}{K}_{\text{a}}\left(\text{obs}\right)=\text{p}{K}_{\text{a}}\left({\text{HisH}}^{+}\right)+\text{p}{K}_{\text{loop}}\left(\text{His}\right).$$ (2)

Dar et al. (58) report an average pK_a(HisH⁺) of 6.78 ± 0.15 for Cytc′ variants in the pWT background for GdnHCl concentrations from 3.0 to 6.0 M, which allows calculation of pK_loop(His) (Table S2). In general, pK_loop(His) becomes more negative (His-heme loop more stable) as loop size decreases Fig. 5 A.

Figure 5.

(A) Shown here is the loop stability, pK_loop(His), versus loop size, n (plotted logarithmically), for single histidine variants in the H3_αLo versus the pWT background of Cytc′. pWT data are from Rao et al. (59) and were converted to pK_loop(His) using pK_a(HisH⁺) = 6.78 from the more recent work of Dar et al. (58). Data obtained for His-heme loop formation with polyalanine sequences also are shown for comparison (61). The solid curves are fits to Eq. 6. The D3H data points were not included in the fits to Eq. 6 because of the higher uncertainty in pK_loop(His) for these variants. (B) Shown here is the plot of ΔpK_loop(His) versus loop size. The dashed line is the average value of ΔpK_loop(His). To see this figure in color, go online.

For variants with the largest His-heme loop sizes, biphasic loop formation was observed (Fig. S5). The same variants show biphasic denatured state loop formation in both the pWT (59) and the H3_αLo backgrounds. For these variants, loop formation curves are broadened and the data do not fit well to Eq. 1. For these variants, the loop is large enough that the affinity of histidine for the heme is insufficient to complete the spin state transition. Thus, a lysine binds to the heme at higher pH and finishes the spin state transition leading to the observed biphasic behavior. Data for these variants were fit to (63)

$$A_{398}=\frac{A_{\text{HS}}+A_{\text{LS}}\left(\left(\frac{{10}^{-\text{p}{K}_{\text{loop}}\left(\text{His}\right)}}{1+{10}^{\text{p}{K}_{\text{a}}\left({\text{HisH}}^{+}\right)-\text{pH}}}\right)+\left(\frac{{10}^{-\text{p}{K}_{\text{loop}}\left(\text{Lys}\right)}}{1+{10}^{\text{p}{K}_{\text{a}}\left({\text{LysH}}^{+}\right)-\text{pH}}}\right)\right)}{1+\left(\frac{{10}^{-\text{p}{K}_{\text{loop}}\left(\text{His}\right)}}{1+{10}^{\text{p}{K}_{\text{a}}\left({\text{HisH}}^{+}\right)-\text{pH}}}\right)+\left(\frac{{10}^{-\text{p}{K}_{\text{loop}}\left(\text{Lys}\right)}}{1+{10}^{\text{p}{K}_{\text{a}}\left({\text{LysH}}^{+}\right)-\text{pH}}}\right)},$$ (3)

and parameters from the fit are listed in Table S3. To fit data to this equation, the intrinsic pK_a of lysine, pK_a(LysH⁺), was set to 10.5. pK_loop(Lys) in Eq. 3 is the pK for binding of a deprotonated lysine to the heme. All other parameters in Eq. 3 are as defined for Eqs. 1 and 2. The pK_a(HisH⁺) values obtained from these fits range from 6.7 to 6.8 (Table S3), consistent with the average value of 6.78 ± 0.15 derived from Cytc′ variants in the pWT background with biphasic loop formation equilibria (58).

A plot of pK_loop(His) versus the log of loop size, n (Fig. 5 A), shows that pK_loop(His) is linearly dependent on log(n) for single histidine variants in both the pWT and H3_αLo backgrounds. For shorter loops, up to and including the helix 3 sequence segment (residues 75–95), the magnitude of pK_loop(His) is similar in both backgrounds. However, with longer loops, for which this sequence segment is fully included in the loop, pK_loop(His) is uniformly more negative (loop formation more favorable) in the H3_αLo background than in the pWT background. A plot of ΔpK_loop(His) versus loop size (Fig. 5 B) shows that the largest increases in denatured state loop stability (most negative ΔpK_loop(His)) are observed for loops that include the helix 3 sequence segment, but with the histidine still close to the C-terminal end of this sequence (E73H and A66H). However, ΔpK_loop(His) is negative for all loops that include the helix 3 sequence segment.

His-heme loop formation data from polyalanine inserts at the N-terminus of iso-1-cytochrome c acquired in 3 M GdnHCl (61) are plotted alongside the data for the single histidine variants in the pWT and H3_αLo backgrounds of Cytc′ (Fig. 5 A). The polyalanine data show a tight linear correlation, whereas the single histidine variants in both the pWT and H3_αLo backgrounds of Cytc′ show sizeable scatter about a linear correlation.

Kinetics of denatured state loop breakage and formation

To gain further insight into the effects of loop length and residual structure in 3.0 M GdnHCl, the kinetics of histidine-heme loop breakage of each variant were measured by pH-jump stopped-flow experiments in the denatured state. Previous work has shown that the kinetics of His-heme loop formation are consistent with a model involving rapid histidine deprotonation followed by binding of histidine to the heme (62). In this model, the observed rate constant, k_obs, has the pH dependence given by

$$k_{\text{obs}}=k_b+k_f\left(\frac{K_a\left({\text{HisH}}^{+}\right)}{K_a\left({\text{HisH}}^{+}\right)+\left[{\text{H}}^{+}\right]}\right),$$ (4)

where k_b and k_f are the rate constants for loop breakage and formation, respectively; and K_a(HisH⁺) is the acid dissociation constant for histidine. Therefore, if pH ≪ pK_a(HisH⁺), then k_obs is approximately equal to k_b. Downward pH jumps from pH 6.8 to pH 3.6 and pH 3.1 yield very similar values for k_obs (Table S4), indicating that this limit has been reached. Plots of k_b versus loop size, n, using k_obs at pH 3.1 for single histidine variants in the H3_αLo versus the pWT background of Cytc′, show that k_b is unaffected by decreasing the intrinsic stability of the helix 3 sequence segment (Fig. 6 A). The magnitude of k_b values for His-heme loops formed in the denatured state of Cytc′ is considerably more variable than for His-heme loops containing polyalanine segments (Fig. 6 A).

Figure 6.

Histidine-heme loop formation kinetics in the denatured state for the H3_αLo background versus the pWT (58) background. (A) Shown here is the His-heme loop breakage rate constant, k_b, as a function of loop size, n, derived from k_obs obtained with a pH jump from pH 6.8 to pH 3.1. Data were obtained at 25°C in 3.0 M GdnHCl. Data for His-heme loop breakage with polyalanine sequences are shown for comparison (61). (B) Given here is a plot of Log(k_f) versus loop size, n (plotted logarithmically). Loop formation rate constants, k_f, were derived as described in the text. Values of k_f are tabulated in Table S5. The solid curves are fits to a linear equation. These fits do not include the D3H data point for either the H3_αLo background or the pWT background. To see this figure in color, go online.

Because His-heme loop formation is two-state, rate constants for loop formation, k_f, can be calculated using pK_loop(His) and k_b (k_f = k_bK_loop(His)). A plot of Log(k_f) versus Log(n) is linear (Fig. 6 B) as observed for plots of pK_loop(His) versus Log(n), with k_f decreasing with increasing loop size. For loops that do not include all of the helix 3 sequence segment (A104H, K97H, A91H, and K84H), k_f is similar in the pWT and H3_αLo backgrounds of Cytc′. However, for the remaining loops, which contain the helix 3 sequence segment, k_f is larger in the H3_αLo background compared to the pWT background of Cytc′.

Discussion

Effect of destabilizing helix 3 of Cytc′ on global stability

Host-guest experiments with proteins and peptides indicate that an Ala→Ser substitution should lead to a decrease in helix stability of 0.4 kcal/mol in protein systems (84) to 0.8 kcal/mol in alanine-based peptides (80). Because both Ala⁸³ and Ala⁸⁷ are at solvent-exposed positions, the primary effect of these substitutions should be on the stability of the helix 3 secondary structure. Thus, the double substitution in helix 3 might be expected to lead to a decrease in protein stability in the range of 0.8–1.6 kcal/mol. The decrease in stability of H3_αLo Cytc′ relative to pWT Cytc′ is 2.0 ± 0.3 kcal/mol (Table 1), near the upper limit of this range. Most of the single histidine variants have a decrease in stability of 0.6–1.1 kcal/mol, near the lower limit of the expected range. Thus, for the most part, the effect of destabilizing helix 3 on global stability is as expected. A few of the single histidine variants (K84H, A66H, K20H, and D3H) show no change in stability in the H3_αLo versus the pWT background. The change in C_m, ΔC_m, for the H3_αLo versus the pWT background, is very consistent throughout all the single histidine variants (Fig. 3), except for substitutions in or near helix 3. The variation in the changes in ΔG_u°′(H₂O) results largely from variability in the m-value, which reflects the cooperativity of the GdnHCl unfolding transition. The larger-than-expected decrease in ΔG_u°′(H₂O) for H3_αLo Cytc′ relative to pWT Cytc′ may reflect the longer extrapolation from the unfolding transition region involved in evaluating ΔG_u°′(H₂O) for this variant than for the single histidine variants. This longer extrapolation is coupled to a significant decrease in the m-value of H3_αLo Cytc′ relative to pWT Cytc′ (Fig. S6). The single histidine variants (K84H, A66H, K20H, and D3H), which have minimal effects on ΔG_u°′(H₂O) in the H3_αLo versus the pWT background, all have significant increases in the m-value for the H3_αLo versus the pWT background (Fig. S6). The cause of the changes in the m-value in the H3_αLo versus the pWT background could reflect changes in denatured state compactness (85) or the population of intermediates during unfolding (86, 87).

Studies on proteins with engineered disulfide bonds often show a decrease in denaturant m-values relative to the wild-type protein lacking disulfides, indicating that formation of a loop in the denatured state leads to a more compact denatured state (88, 89, 90). Effects of His-heme loop formation on m-values are likely more complex because, unlike with disulfide bonds, the loop is not formed in the native state. The m-values observed for variants with short denatured state His-heme loops are actually larger than the m-value for H3_αLo (or pWT) Cytc′ (Fig. S6 A; Table 1), for which the Lys-heme loop is not strongly populated at pH 7. The same phenomenon is observed for single histidine variants of iso-1-cytochrome c relative to the variant with no histidine that can form a denatured state His-heme loop (91). Short loops are the most stable loops (Table S2), and thus the driving force for loop formation may make the unfolding transition more cooperative. Variants with shorter His-heme loops also sequester little of the chain within the loop in the denatured state. Thus, the compaction of the denatured state due to loop formation also will be minimal. As observed for single histidine variants of iso-1-cytochtome c (91), the m-value decreases as loop size increases and more of the protein sequence is constrained within the loop. This observation appears to be consistent with loop formation, causing a more compact denatured state as observed for proteins with engineered disulfide bonds (88, 89, 90). The error bars on Δm are large. However, in general Δm is positive (Fig. S6 B, larger for the H3_αLo variants) for most pairs of single histidine variants. His-heme loop formation is also more favorable for the H3_αLo variants (Fig. 5 B). Thus, more favorable denatured state loop formation may increase the cooperativity of unfolding in the H3_αLo versus the pWT background.

The plot of ΔC_m versus sequence position (Fig. 3) shows an interesting deviation around helix 3 (residues 75–95). In particular, the ΔC_m for the K84H variant is half of the average. Agadir predicts that adding the K84H substitution to pWT Cytc′ decreases the average % helix from 11.2 to 4.2% at 22°C. However, Agadir predicts that adding the K84H substitution to H3_αLo only decreases the average % helix from 2.7 to 1.9% at 22°C. Thus, the destabilizing effect on helix 3 of the K84H substitution in the H3_αLo versus the pWT background is small because the helix is already strongly destabilized.

Effect of destabilizing helix 3 of Cytc′ on denatured state loop stability

For a random coil polymer, the dependence of loop stability on loop size is assumed to be due only to loop entropy, ΔS_loop. Loop entropy for a random coil polymer increases with the natural log of the number of monomers in the loop, n, as given by the Jacobson-Stockmayer equation (92):

$$\Delta S_{\text{loop}}=-{\nu }_{3}R\text{ln}\left(n\right)+R\text{ln}\left({\left(3/2\mathit{\pi }{C}_n{\ell }^2\right)}^{{\nu }_3}{V}_i\right),$$ (5)

where ν₃ is the scaling exponent, R is the gas constant, C_n is Flory’s characteristic ratio, ℓ is the distance between monomers, and V_i is the volume within which the ends of the chain must be constrained for the loop to form. The more familiar scaling exponent, ν, for the dependence of the radius of gyration, R_g, on the number of residues in a denatured protein is near 0.6 (random coil with excluded volume) (19). For loop formation, the chain ends can explore the entire volume with radius, R_g. Thus, ν₃ is expected to be 1.8 (threefold larger) for a random coil with excluded volume (92, 93). Simulations indicate that ν₃ likely has values from 2.1 to 2.4 for a random coil with excluded volume (94, 95). For His-heme loop formation, ν₃ is readily determined from a plot of loop stability versus log of the loop size (61),

$$\text{p}{K}_{\text{loop}}\left(\text{His}\right)=\text{p}{K}_{\text{loop}}{\left(\text{His}\right)}_{\text{ref}}+{\nu }_3\text{log}\left(n\right),$$ (6)

where pK_loop(His)_ref corresponds to pK_loop(His) for n = 1. The fits of pK_loop(His) versus log(n) data to Eq. 6 for the pWT and H3_αLo variants (Fig. 5 A, solid curves) yield ν₃ = 2.3 ± 0.2 and ν₃ = 2.1 ± 0.2, respectively. Thus, the macroscopic behavior for both data sets is that of a random coil with excluded volume. A plot of pK_loop(His) versus log(n) for His-heme loop formation with polyalanine segments engineered into the N-terminus of yeast iso-1-cytochrome c (61) shows little scatter about the fit to Eq. 6 (Fig. 5 A). For His-heme loop formation with Cytc′ in either the pWT or the H3_αLo background, there is considerable scatter about the fit to Eq. 6, indicating considerable deviation from random coil behavior at the level of local sequence. Furthermore, the pattern of scatter of the data points about the fit to Eq. 6 is identical in the pWT versus the H3_αLo backgrounds of Cytc′ (Fig. 5 A). This observation indicates that the pattern of scatter is not random and that reducing the intrinsic helical propensity of the helix 3 sequence segment does not affect deviations from random coil behavior at the level of local sequence. It is known from simulations that local residual structure, even at the level of full secondary structure, is compatible with random coil scaling exponents (96).

Although the local behavior of the denatured state of Cytc′ is unaffected by decreasing the intrinsic helical propensity of the helix 3 sequence segment, there are evident long-range effects on pK_loop(His) (Fig. 5; Table S2). His-heme loops, which include the helix 3 sequence segment, are more stable (ΔpK_loop(His) is negative; Fig. 5 B). Thus, decreasing the helical propensity of helix 3 appears to decrease the stiffness of this sequence segment of Cytc′ under denaturing conditions favoring loop formation. NMR studies in concentrated denaturants show that the small levels of residual secondary structure present under these conditions can cause measurable chain stiffness as determined by transverse relaxation experiments (32, 97). The effect on ΔpK_loop(His) appears to be more pronounced for loop formation with histidines that are closer to the C-terminal end of the helix 3 sequence segment (A66H and E73H; see Fig. 5 B). Thus, residual helical structure in the denatured state appears to create a local excluded volume effect similar to the local excluded volume effect of the heme when denatured state His-heme loop formation requires the chain to wrap around the heme (64). Studies on loop formation with (GlySer)_n polypeptides suggest that hydrogen bonding of Ser to the peptide backbone stiffens the polypeptide chain (98). Thus, the A83S/A87S substitutions could stiffen rather than increase the flexibility of the polypeptide chain. The results in Fig. 5 suggest that the disruption of residual helical hydrogen bonding in the denatured state by the A83S/A87S substitutions decreases chain stiffness more than Ser hydrogen bonding to the backbone increases chain stiffness.

Effect of destabilizing helix 3 of Cytc′ on the kinetics of denatured state loop formation

For denatured state His-heme loop formation with polyalanine inserts engineered into the N-terminus of yeast iso-1-cytochrome c, loop breakage rate constants, k_b, are relatively invariant as a function of loop size (61) (Fig. 6 A). For Cytc′, k_b varies over a much wider range as a function of loop size (Fig. 6 A), indicative of local nonrandom behavior in the denatured state. As with the pattern of scatter about the dependence of pK_loop(His) on log(n) (Fig. 5 A), the pattern of variation in k_b is unaffected by reducing the intrinsic helical propensity of the helix 3 sequence segment. Thus, reducing the intrinsic helical propensity of the helix 3 sequence segment does not affect the local nonrandom behavior of the denatured state of Cytc′.

The lack of any significant effect of the A83S/A87S double substitution on k_b indicates that the long-range effects of reducing the intrinsic helical propensity of the helix 3 sequence segment on equilibrium loop formation (Fig. 5) must result from the rate constant for denatured state loop formation, k_f. The plot of log(k_f) versus log(n) (Fig. 6 B) shows that k_f is similar in the pWT versus the H3_αLo background for His-heme loops that do not contain all of the helix 3 sequence segment. However, for longer loops that contain all of this sequence segment, k_f is larger in the H3_αLo background than in the pWT background (Fig. 6 B; Table S5). Thus, reducing the propensity to form residual helical structure in the denatured state enhances the flexibility of the polypeptide chain lowering the average end-to-end His to heme distance. As with ΔpK_loop(His) (Fig. 5 B), the change in k_f for the H3_αLo versus the pWT background, Δk_f, is largest for histidines closest to the C-terminal end of the helix 3 sequence segment (E73H, A66H; see Table S5). Thus, residual helical structure limits contact formation in the denatured state most for residues near the residual structure.

Implications of residual structure in the denatured state for protein folding mechanisms

Numerous studies on protein denatured states demonstrate that long-range nonrandom contacts occur in the denatured states of proteins (43, 46, 47, 48, 49, 50, 53, 54, 55, 57). Folding mechanisms that alternatively rely on flickering elements of secondary structure (7, 8, 9, 10, 11) versus diffuse long-range interactions (6) appear equally suited to efficient folding. It has been observed that a continuum may exist between these divergent mechanisms (17). Therefore, it is of interest to consider whether the properties of the denatured state may play a role in modulating the balance between the importance of secondary structure and long-range contacts in a folding mechanism. Meng et al. (57) have recently observed that long-range contacts appear to persist even when little residual secondary structure remains in the denatured state. This observation seems counterintuitive given the cooperative stabilization expected from the interaction of two preformed elements of secondary structure (99). Our results show that decreasing the propensity to form residual structure in the denatured state makes the polypeptide chain more flexible and thus more able to form long-range contacts in the denatured state. The expected decrease in the rate of protein folding due to the stiffening of the protein backbone by residual helical secondary structure has been noted previously (3). Thus, loss of residual secondary structure by increasing the probability of long-range contacts will tip the balance toward a folding mechanism with a diffuse nucleus of tertiary interactions. Under folding conditions where residual secondary structure is more populated in the denatured state (43, 44, 45), the effects observed here should be even more pronounced. In summary, our results show that the intrinsic properties of the denatured state will naturally allow a folding mechanism to move along a continuum between a mechanism dominated by secondary structure formation and one dominated by tertiary contacts.

Author Contributions

T.A.D. both performed research and wrote the article. B.E.B. designed research and wrote the article.

Editor: Daniel Raleigh.