Effect of an Imposed Contact on Secondary Structure in the Denatured State of Yeast Iso-1-cytochrome c

ABSTRACT: There is considerable evidence that long-range interactions stabilize residual protein structure under denaturing conditions. However, evaluation of the effect of a specific contact on structure in the denatured state has been difficult. Iso-1- cytochrome c variants with a Lys54 → His mutation form a particularly stable His−heme loop in the denatured state, suggestive of loop-induced residual structure. We have used multidimensional nuclear magnetic resonance methods to assign 1H and 15N backbone amide and 13C backbone and side chain chemical shifts in the denatured state of iso-1-cytochrome c carrying the Lys54 → His mutation in 3 and 6 M guanidine hydrochloride and at both pH 6.4, where the His54−heme loop is formed, and pH 3.6, where the His54−heme loop is broken. Using the secondary structure propensity score, with the 6 M guanidine hydrochloride chemical shift data as a random coil reference state for data collected in 3 M guanidine hydrochloride, we found residual helical structure in the denatured state for the 60s helix and the C-terminal helix, but not in the N-terminal helix in the presence or absence of the His54−heme loop. Non-native helical structure is observed in two regions that form Ω- loops in the native state. There is more residual helical structure in the C-terminal helix at pH 6.4 when the loop is formed. Loop formation also appears to stabilize helical structure near His54, consistent with induction of helical structure observed when His− heme bonds form in heme−peptide model systems. The results are discussed in the context of the folding mechanism of
cytochrome c.

The progression of events to a fully folded protein from a disordered state has been an area of considerable interest,and much progress has been made toward understanding this process.1,2 The earliest events in protein folding likely involve the formation of simple loops. Considerable effort has focused on understanding the factors that control both the kinetics3−9and the thermodynamics10−18 of loop formation in disorderedpeptides and denatured proteins. While the bulk properties of protein denatured states are consistent with random coil behavior,19,20 there is considerable evidence that significant residual structure can exist in the denatured states of proteins.21−25 Numerous nuclear magnetic resonance (NMR)studies,26−54 studies of the thermodynamics of electrostaticinteractions in the denatured state,55−62 Förster resonance energy transfer (FRET) studies,63−65 and some small-angle X- ray scattering (SAXS)66−69 data support this conclusion. NMR studies of denatured proteins have been particularly fruitful.These studies have shown evidence for both native and non- native secondary structure in protein denatured states.22−26,30,35−42,54,70−73 NMR transverse relaxation experi- ments are sensitive to local chain stiffness, providing evidence of hydrophobic clusters and residual secondary structure in protein denatured states as well as more flexible regions that may act as molecular hinges early in folding.38−42,52,71,72,74−78Paramagnetic relaxation enhancement NMR experiments have further demonstrated long-range interactions in the denatured states of many proteins.43,44,46,48,49,51,53,79−81 In a few cases, mutational analysis has demonstrated the importance of long- range contacts in stabilizing structure in the denatured state.76,77,82 However, it has been more difficult to study the effect of a specific well-defined contact on residual structure in the denatured state of a protein.Two methods have been used extensively to impose defined contacts in the denatured state of a protein: engineered or naturally occurring disulfide bridges83−92 and histidine−heme loops.

In the case of disulfide bonds, the denaturant m- value for global unfolding is often decreased relative to that inthe absence of the disulfide bond,84,87,91 indicating that the disulfide bond induces a more compact denatured state.93 Disulfide bonds have been used to probe folding pathways and can lead to remarkable acceleration of folding rates.87−89,92 However, the effect on folding rates appears to depend on whether the disulfide bond encompasses a part of the protein that forms structure in the transition state.86 In some cases,changes in the folding mechanism occur.89,91 For a Cys85− Cys102 disulfide bond engineered into barnase, a significant dependence of disulfide loop stability in the denatured state on guanidine hydrochloride (GdnHCl) concentration was ob- served (m-value of ∼0.4 kcal mol−1 M−1).83 NMR data in this study were inconclusive with regard to the nature of the residual structure induced by disulfide loop formation. However, molecular dynamics simulations indicated the disulfide bond stabilized a nativelike hairpin loop in the denatured state.83 Notably, hydrophobic clusters observed in the denatured state of wild type barnase appeared to be disrupted by the Cys85−Cys102 disulfide bond.For denatured state His−heme loop formation, GdnHCl m-values observed for the stability of the loops are typically near0.2 kcal mol−1 M−1 for yeast iso-1-cytochrome c (iso-1-Cytc) and the four-helix bundle cytochrome c′.11,16 These small m- values indicate that loop formation in the denatured state buriesonly a small amount of hydrophobic surface area for most single-histidine variants of these proteins. However, in the case of a K54H variant of iso-1-Cytc, a relatively large GdnHCl m- value of 0.45 kcal mol−1 M−1 was observed,16 indicating burial of a more significant amount of hydrophobic surface area upon His54−heme loop formation in the denatured state similar tothat buried when the Cys85−Cys102 disulfide bond of barnaseforms.83

In addition, kinetic studies showed an unexpectedly slow rate of breakage of the very stable His54−heme loop, also consistent with stabilization of the closed loop by residual structure in the denatured state.15 While the His54−heme loop is a non-native contact, it presents an advantageous model system for characterizing the effects of a specific well-defined long-range interaction on a protein denatured state.To investigate residual structure produced by the His54− heme loop formed by the K54H variant of iso-1-Cytc at residue level resolution, we assigned chemical shifts in this variantunder denaturing conditions in the presence and absence of the His54−heme loop. We analyzed the observed chemical shifts relative to random coil shifts using the secondary structure propensity (SSP) score developed to assess residual structure in disordered proteins.94 We observe significant residual secon- dary structure in the denatured state in 3 M GdnHCl in the presence and absence of the His54−heme loop. As with barnase,83 loop formation changes the distribution of residual structure.Preparation of Iso-1-Cyc Variants. Preparation of variants started with the pRbs_BTR1 vector95 carrying the coding sequence for a TM_T(−5)Q/N52I variant. TM stands for triple mutant and is shorthand for a variant with H26N, H33N, and H39Q mutations, which eliminate His−heme misligation in the denatured state of iso-1-Cytc.18,96 In addition to the T(−5)Q and N52I mutations, this variant also contains a C102S mutation to prevent disulfide dimerization during physical studies. The pRbs_BTR1 vector is a derivative of the pBTR1 vector,97,98 which carries the iso-1-Cytc and heme lyase genes (CYC1 and CYC3, respectively) from Saccharomyces cerevisiae, allowing covalent attachment of heme to the CLQCH heme attachment sequence of iso-1-Cytc in the cytoplasm of Escherichia coli.

The sequences of oligonucleotides used for mutagenesis are listed in Table S1. Vector DNA for the expression of the TM_T(−5)P/N52I variant was prepared by the QuikChange protocol (Agilent Technologies) using the oligonucleotides Q(−5)P and Q(−5)P-r. In a similar manner,the vector for expression of the TM_T(−5)P/N52I/K54H variant was prepared by the QuikChange protocol using double-stranded pRbs_BTR1 DNA carrying the coding sequence for the TM_T(−5)P/N52I variant and oligonucleo- tides K54H(I52) and K54H(I52)-r. The entire coding regions of the TM_T(−5)P/N52I and TM_T(−5)P/N52I/K54H variants of iso-1-Cytc were verified by dideoxy sequencing (Murdock DNA Sequencing Facility, University of Montana). Expression of Isotopically Labeled Iso-1-Cytc. Com- petent BL21-DE3 E. coli cells (EdgeBio) were transformed with the pRbs_BTR1 vector carrying the coding sequence of the TM_T(−5)P/N52I/K54H variant and plated on LB agar plates containing 100 μg/mL ampicillin. Colonies on a plate were suspended in 3 mL of sterile LB medium, and the suspension was used to inoculate 1 L of minimal medium containing 100 μg/mL ampicillin. The 1 L of minimal medium consisted of the following components: 10 mL of 0.1 g/mL 15NH4Cl, 1 mL of 1 M MgCl2, 100 μL of 1 M CaCl2, 1 mg of thiamine, 5 mg of FeSO4·7H2O, 1 mg of biotin, 4 g of [13C]glucose, 16.8 mg of δ-aminolevulinic acid, and 100 mL of a 10 × M9 salt solution at pH 7.4. The water, the M9 salt solution, and the MgCl2 solution were each autoclaved. The rest of the components were filter sterilized. The culture in minimal medium was grown at 37 °C while being shaken at 125 rpm for 24 h. Cells were harvested using a Sorvall GS-3 rotor in a Sorvall RC-5C Plus floor model centrifuge running at 5000 rpm for 10 min. The supernatant was discarded, and theresulting cell pellet was stored at −80 °C.Expression of Iso-1-Cytc in Rich Media. Iso-1-Cytc variants were also expressed in Terrific Broth that contained 100 μg/mL ampicillin. The rest of the growth procedure was the same as for the preparation of isotopically labeled protein in minimal medium.Lysis and Crude Purification of Iso-1-Cytc. The cell pellet was removed from the −80 °C freezer, allowed to thaw in a 4 °C cold box, and then returned to the −80 °C freezer. Two more rounds of freezing and thawing were performed.

On thefinal thaw, the cells were resuspended in lysis buffer [100 mM Tris (pH 8.0), 500 mM NaCl, 1 mM EDTA, 2 mM PMSF, anda few crystals of DNase I and RNase A]. Lysis buffer was used at a level of 2 mL/g of cells to be lysed. The cell slurry was run through a French press (SLM Aminco) 8−10 times to break the cells. This cell slurry was centrifuged twice, using a GSA rotor in a Sorvall RC-5C Plus floor model centrifuge at 10000 rpm for 30 min, keeping the supernatant. Undesired proteins were precipitated by bringing the solution to 50% saturation with ammonium sulfate (31.4 g of ammonium sulfate/100 mL of lysate) and stirring for 4−5 h or overnight at 4 °C. The solution was centrifuged at 10000 rpm for 30 min, and the pellet was discarded. The supernatant was transferred to 3500 molecular weight cutoff (MWCO) dialysis tubing and placed in12.5 mM sodium phosphate (pH 7.2), 1 mM Na2EDTA, 2 mM β-mercaptoethanol (β-ME) dialysis buffer and stirred gently overnight at 4 °C. The dialysis tubing was transferred to a second bottle of dialysis buffer, and the contents were stirred gently for 24 h. The iso-1-Cytc solution was removed from the dialysis bags and centrifuged at 10000 rpm for 30 min to remove any precipitate.Purification of Iso-1-Cytc by CM-Sepharose. Approx- imately 100 mL of CM-Sepharose resin was equilibrated in CM buffer A [50 mM sodium phosphate (pH 7.2), 1 mM EDTA, and 2 mM β-ME], and the dialyzed iso-1-Cytc solution was added to the resin. After being stirred at 4 °C for 1 h, the resinwas allowed to settle, and the buffer was decanted off. The resin, now pink with iso-1-Cytc protein, was resuspended in a small amount of CM buffer A and packed into a column. Approximately 300−400 mL of CM buffer A was run through the column as a wash. A 200 mL linear 0 to 0.8 M NaCl gradient was used to elute the protein from the CM-Sepharose by mixing CM buffer A and CM buffer B (CM buffer A with 0.8nm, θ222, diagnostic of α-helix, and 250 nm, θ250, used as background, with 25 s of signal averaging at each wavelength. All data were acquired at 25 °C. θ222corr (=θ222 − θ250) was plotted against GdnHCl concentration and fit to eq 1θ222corr =⎡ m[GdnHCl] − ΔG◦′(H O) ⎤benchtop Jouan CR4i centrifuge. Purification of Iso-1-Cytc by HPLC.

Unlabeled protein was reduced with a small amount of sodium dithionite prior to HPLC, whereas 13C- and 15N-labeled protein was oxidized with 5 mg of potassium ferricyanide/mg of crude protein prior to HPLC. Protein was loaded onto a Bio-Rad Uno S6 cation exchange column and eluted with a gradient formed by mixing HPLC buffer A and HPLC buffer B [50 mM sodium phosphate and 1 M NaCl (pH 7.0)]. The following gradient was used: 0% B from 0 to 7 min, linear increase to 30% B from 7 to 34 min, held at 30% B from 34 to 40 min, linear increase to 100% B from 40 to 43 min, held at 100% B from 43 to 50 min, lineardecrease to 0% B from 50 to 53 min, and held at 0% B from 53to 63 min, at a flow rate of 3 mL/min. The largest peak was collected and exchanged back to HPLC buffer A and concentrated using the Millipore centrifugal filters mentioned above. The protein concentration was determined spectroscopi- cally using a Beckman Coulter DU 800 spectrophotometer at isosbestic points of 339, 526.5, and 541.75 nm of the iso-1-Cytc spectrum as well as with the oxidized state extinction coefficient at 550 nm.99,100For the 13C- and 15N-labeled protein, a mass spectrum was obtained with a Bruker microflex matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometer to check the efficiency of labeling. The observed peak near 13252 Da is close to the expected mass of 13252.1 Da for uniformly 13C- and 15N-labeled protein.Global Thermodynamic Stability Measurements. To measure global stability, GdnHCl denaturation was employed and monitored by circular dichroism (CD) spectroscopy. Protein in HPLC buffer A was oxidized with ferricyanide and then separated from ferricyanide and exchanged into CD buffer [20 mM Tris, 40 mM NaCl, and 1.0 mM NaEDTA (pH 7.0)] by passing the protein solution through a Sephadex G-25 column equilibrated with CD buffer. The concentration of the eluted protein was determined as described above. A volume of 850 μL of 4 μM native protein in CD buffer was prepared and placed in a 10 mm × 4 mm quartz Suprasil cell with space for a stir bar (Hellma part no. 119004F-10-40). A stock solution of 4 μM protein in approximately 6 M GdnHCl buffered with CDbuffer was prepared from a 7.5−8.0 M GdnHCl solution buffered with CD buffer and the eluted protein stock in CD buffer.

GdnHCl titration was performed by titrating the 6 Mwhere θN and θD are the ellipticities of the native and denatured protein, respectively, at 0 M GdnHCl, mD is the denaturant dependence of the ellipticity of the denatured state, m is the GdnHCl concentration dependence of ΔGu, the free energy ofunfolding, and ΔGu°′(H2O) is the free energy of unfoldingextrapolated to 0 M GdnHCl. Reported parameters are the average and standard deviation of three or four separate titrations.Denatured State Loop Formation. His−heme loop stability was measured in the denatured state (3 M GdnHCl, 5 mM Na2HPO4, and 15 mM NaCl) by pH titration. Loopequilibria were monitored at the heme Soret band (398 nm). Titrations were performed at room temperature (22 ± 1 °C). A 3× protein 3× buffer stock solution (9 μM protein, 15 mM Na2HPO4, and 45 mM NaCl) was made. The refractive index of an approximately 6 M GdnHCl solution was measured using water as a background to determine the exact GdnHCl concentration.101A 3 mL sample that contained 3 μM protein, 1× buffer, and 3 M GdnHCl was prepared by combining appropriate amounts of 3× protein 3× buffer stock, a 6 M GdnHCl solution, and water. The pH was measured with a Denver Instruments UB-10 pH meter; 60 μL of the sample were then transferred into a Beckman Coulter microcell cuvette, and an absorbance spectrum was acquired from 350 to 450 nm using a Beckman Coulter DU 800 spectrophotometer. To change the pH without changing the protein or GdnHCl concentration of the sample, 3× protein 3× buffer stock (20 μL), ∼6 MGdnHCl (∼30 μL), and HCl (∼10 μL) in appropriate ratios,for a total volume of 60 μL, were added back to the sample and thoroughly mixed by pipetting. The volume of HCl added was constant, but the concentration of the acid was varied such that the pH of the sample decreased by ∼0.2 pH unit with each addition. After each pH adjustment, the pH and the absorption spectrum of the sample were measured as described above. At least three pH titrations were completed for each variant. The absorbance at 450 nm was used as a baseline and subtracted from the absorbance values at 350−450 nm.

The adjustedabsorbance at 398 nm, A398corr (=A398 − A450), was plottedversus pH and fit to a modified Henderson−Hasselbalch equation (eq 2) to determine the apparent pKa of loop formation, pKa(obs), and the number of protons, np, linked tothe process.Hamilton Microlab 500 titrator interfaced with an AppliedPhotophysics Chirascan CD spectrometer. A small stir bar was present in the cuvette and would stir at approximately 200 rpmfor 100 s after the addition of a new denaturant stock to make certain mixing and equilibration were complete. The sample was allowed to settle without being stirred for 10 s followed by the acquisition of data. The ellipticity was monitored at 222where ALS is A398corr when the loop is formed and the heme is in a low-spin state and AHS is A398corr when the loop is broken and the heme is in a high-spin state, with water occupying the sixth coordination site.Preparation of Iso-1-Cytc Samples for NMR Spectros-a Denver Instruments UB-10 pH meter. The procedure for changing the pH of the sample was very similar to the procedure described above for denatured state pH titrations. For example, to change the pH in the 3 M GdnHCl sample without changing the protein or GdnHCl concentration, a total of 12 μL was added to the sample: 3× protein 3× buffer stock(4 μL), ∼6 M GdnHCl solution (∼6 μL), D2O (1.2 μL), and HCl (∼0.8 μL); then the sample was thoroughly mixed by pipetting. The volume of HCl added was constant, but theconcentration of the acid was varied as needed to reach the desired pH value.NMR Experiments. Chemical shifts of TM_T(−5)P/ N52I/K54H iso-1-Cytc in 3 M GdnHCl at pH 6.4 and 3.6and in 6 M GdnHCl at pH 6.4 and 3.6 were assigned using two- dimensional (2D) 1H−15N HSQC102 and three-dimensional (3D) HNCACB,103 CBCA(CO)NH,104 C(CO)NH,105HNCO,106,107 and HN(CA)CO108 spectra, which were acquired under all four solution conditions (Table S6).

The experiments were performed at 25 °C using a 600 MHz Varian NMR System spectrometer equipped with a triple-resonance probe or a 13C-enhanced salt-tolerant cold probe. Standard BioPack (VnmrJ 2.2C software, Agilent) sensitivity-enhanced pulse sequences that utilize gradients for coherence selection were employed. 1H chemical shifts were referenced relative to internal DSS (4,4-dimethyl-4-silapentane-1-sulfonic acid), and 13C and 15N chemical shifts were referenced indirectly.109,110 The data were processed with NMRPipe.111 A squared cosine- bell window function and zero filling were applied in the 1H and 13C dimensions, and linear prediction to double the number of data points, a squared cosine-bell window function, and zero filling were applied in the 15N dimension prior to the respective Fourier transforms.Chemical shift assignments were made with the aid of Sparky (http://www.cgl.ucsf.edu/home/ sparky/)112 and CcpNmr Analysis version 2.2.2.113Evaluation of Secondary Structure in the DenaturedState. The 13Cα and 13Cβ chemical shifts were used to estimate the secondary structure propensities in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc. The chemical shifts were re-referenced, and the SSP score was calculated using eq 3 as implemented in SSP version 1.02.94random coil and secondary shifts and their standard deviations that are provided with the SSP program come from the RefDB database.114 A weighted average over a five-residue window is used to minimize the effects of chemical shifts that are poor measures of secondary structure, such as the chemical shifts of glycine residues. As the score reflects the fraction of secondary structure at a specific position, fully formed secondary structure would have an absolute value of 1. Regions of α and β structure yield positive and negative SSP scores, respectively.

RESULTS
Design of TM_T(−5)P/N52I/K54H Iso-1-Cytc. The goal of this work is to determine the effect that formation of a simple loop has on the presence of ordered structure in an unfolded protein. In previous work, we have shown that a histidine at position 54 of iso-1-Cytc forms a particularly stable histidine−heme loop in the denatured state.15−18,96 Thedenaturant dependence of the stability of the His54−hemeloop in the denatured state is large compared to those of other histidine−heme loops, suggesting induction of structure upon loop formation in the denatured state.16,83 Thus, the His54− heme loop is well-suited as a model system for the effects of loop formation on residual structure in the denatured state.Variants of iso-1-Cytc used to monitor histidine−heme loop formation in the denatured state contain a single histidine besides His18 from the CXXCH heme attachment motif. Thehistidine used for loop formation is engineered into a surface accessible site (Figure 1). When iso-1-Cytc unfolds, the weak native state bond between Met80 and the heme iron breaks (Figure 1). The engineered histidine forms a strong bond with the heme iron, causing the formation of the denatured state histidine−heme loop. Thus, particularly stable histidine−heme loops can decrease the global stability of iso-1-Cytc by preferentially stabilizing the denatured state of iso-1-Cytc.115 The His54−heme loop strongly stabilizes the denatured state ofiso-1-Cytc.16−18 Thus, K54H variants of iso-1-Cytc are onlymarginally stable.17,18 To ameliorate this problem, we use mutation N52I, which is a global stabilizer of yeast iso-1-Cytc (see Figure 1).116 Previous work shows that this mutation stabilizes the K54H variant by 2 kcal/mol117 and should favor higher yields of isotopically labeled protein for NMR experiments.As in our previous work, the iso-1-Cytc variants used here are built on the TM variant (H26N, H33N, and H39Q), in whichnative host, S. cerevisiae, the N-terminal amino group can beacetylated in vivo as a post-translational modification, preventing competition between the N-terminal amino group and histidine for heme binding in the denatured state.17 However, post-translational acetylation of the N-terminal amino group is not straightforward when producing isotopically labeled protein in E. coli.

The N-terminal α-amino group can be converted into an α-keto group via a transamination reaction,118,119 but the yield is too low for use with isotopically labeled protein samples. Conversion of the terminal amino group to proline provides another strategy for reducing the level of interference from the N-terminus of the protein. The pKa of 10.47 ± 0.06 (ionic strength μ = 0.1) for the secondary amine of proline120,121 is 1.5 units higher than the pKa of 8.97± 0.03 (μ = 0.1) for the α-amino group of threonine,121 thewild type residue at the N-terminus of iso-1-Cytc. Thus, N- terminal proline is expected to interact more weakly with the heme in the denatured state and interfere less with His−heme loop formation than a primary N-terminal α-amino group. TM_T(−5)P/N52I and TM_T(−5)P/N52I/K54H iso-1-Cytc variants were prepared for the work described here (see Figure 1).Stability of Iso-1-Cytc Variants. GdnHCl denaturation data for the TM_T(−5)P/N52I and TM_T(−5)P/N52I/ K54H variants of iso-1-Cytc are shown in Figure 2. The TM_T(−5)P/N52I/K54H variant unfolds at a GdnHCl concentration considerably lower that that of the TM_T(−5)P/N52I variant. At low concentrations of GdnHCl, we observe an increase in the negative ellipticity of iso-1-Cytc followed by a leveling off. This behavior is commonly observed for iso-1-Cytc.122 Previous work from the laboratories of Tsong and Goto has shown that this effect is due to specific ion binding by Cl−.123,124 We typically exclude these data points from our data analysis and fit the data only after the native baseline has leveled off.125−128 Fits of the data in Figure 2 to a linear free energy relationship (eq 1, Experimental Procedures)aCm is the GdnHCl concentration at the midpoint of the unfolding titration. bParameters are taken from ref 117. The AcTMI52 and AcH54I52 variants are identical to the TM_T(−5)P/N52I and TM_T(−5)P/N52I/K54H variants, respectively, except that insteadof proline at position −5, they contain T(−5)S and K(−2)L mutationsso that the N-terminus is acetylated when expressed in S. cerevisiae.indicate that the K54H mutation destabilizes iso-1-Cytc by ∼3 kcal/mol (Table 1).

Comparison with GdnHCl denaturation data for the equivalent variants expressed in yeast with N- terminal acetylation, AcTMI52 and AcH54I52,117 shows that the E. coli-expressed variants with proline at the N-terminus have stabilities similar to those of their yeast-expressed counterparts.In Figure 2, it evident that the cooperative unfolding of the TM_T(−5)P/N52I/K54H variant is complete by 1.5 M GdnHCl. To be conservative, NMR studies were performed at 3 and 6 M GdnHCl to ensure that only residual structure in fully denatured TM_T(−5)P/N52I/K54H iso-1-Cytc is being monitored and not in partially folded forms of the protein.Loop Formation in the Denatured State. Met80 is a weak ligand for the heme of Cytc when the iron of the heme is in the ferric state.129,130 When Cytc unfolds, Met80 (see Figure 1) is replaced by histidines or lysines from the Cytc sequence,131−133 which have an affinity for ferric heme higher than that of methionine. For yeast iso-1-Cytc, we have engineered the protein so that a single histidine in the sequence can form a loop of defined size in the denatured state.16,17 When the pH is decreased, the histidine becomes protonated, and the loop is broken. Deprotonation of His54 and coupled His−heme loop formation can be described by an apparent pKa, pKa(obs) (Figure 3). The process is readily monitored at micromolar concentrations of iso-1-Cytc due to the shift in the intense Soret absorption band of the heme caused by the spin state change that results from replacing the strong field ligand, histidine, with the weak field ligand, water.with His54 binding in the denatured state of the TM_T(−5)P/ N52I/K54H variant. Furthermore, the pKa(obs) for binding of lysine to the heme of iso-1-Cytc in the denatured state shifts from 7.5 in the absence of a histidine to 9.0 in the presence of His54.133 Thus, the proline secondary amino group should not displace His54 from the heme until the pH is significantly above 7 in denaturing solutions of GdnHCl. We note that the number of protons, np, linked to loop formation is ∼1 in all cases (Table 2), as expected, because a histidine (or proline secondary amino group) must be deprotonated for the denatured state His−heme loop to form (Figure 3).Studies of the TM_T(−5)P/N52I/K54H Variant in theDenatured State by NMR Spectroscopy.

The GdnHCl unfolding titration data for the TM_T(−5)P/N52I/K54H variant in Figure 2 show that it is fully in the denatured state above 1.5 M GdnHCl. The NMR studies described here were performed at 3 and 6 M GdnHCl. Thus, the data report on theproperties of the fully denatured state of the TM_T(−5)P/ N52I/K54H variant at two different concentrations of GdnHCl.From the denatured state pH titration data in Figure 4, it is evident that in 3 M GdnHCl the His54−heme loop of the TM_T(−5)P/N52I/K54H variant is essentially completely formed at pH 6.4 and almost completely broken at pH 3.6. Thus, NMR studies in the denatured state at pH 6.4 and 3.6can provide for a comparison of secondary structure in the denatured state in the presence and absence of the His54− heme loop. It is important to evaluate quantitatively the degree to which the Pro(−5) secondary amino group competes with His54 for heme binding at pH 6.4 in the denatured state. The pKa(obs) can be broken down into two terms, pKa(LH+), which describes ionization of the ligand, and pKFeL, which describes the affinity of the fully deprotonated ligand [His54 or Pro(−5) secondary amino group] for ferric heme, as given by eq 4pKa(obs) = pKa(LH+) + pKFeL (4)For the pKa(LH+) of the secondary amino group of proline, pKa values for the N-terminal amino group in short peptides provide an appropriate estimate. A pKa value of 8.80 is reported for L-Pro-L-Leu-Gly-CONH2 (μ = 0.1; 25 °C).121 Using this value for pKa(LH+), along with the pKa(obs) in Table 2 for TM_T(−5)P/N52I, yields a pKFeL of −2.13 for N-terminalproline secondary amino group−heme binding in the denatured state in 3 M GdnHCl. Similarly, pKFeL = −2.04 for His54−heme binding in the denatured state in 3 M GdnHCl using the pKa(obs) for TM_T(−5)P/N52I/K54H in Table 2and a pKa(LH+) of 6.78, an average value for histidineionization in denaturing concentrations of GdnHCl extracted from His−heme loop formation data at eight different sequence positions from 3 to 6 M GdnHCl.11 The pH dependence of the affinity of each ligand relative to ferric heme with water bound at low pH (Figure 3) is given by eq 5.130,133We collected NMR data and assigned 1H and 15N backbone amide and 13C backbone and side chain chemical shifts in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc at pH6.4 to report on residual secondary structure with the His54−heme loop in place and at pH 3.6 to report on residual secondary structure in the absence of this loop. Figure 5 showsWe can use eq 5 to evaluate the partition function for heme− ligand binding and the population of each heme−ligand pair133 as a function of pH. For the denatured state of TM_T(−5)P/ N52I/K54H in 3 M GdnHCl, we find that ∼93% of TM_T(−5)P/N52I/K54H has water bound to the heme at pH 3.6, with the remainder having His54 as the ligand. At pH6.4 in 3 M GdnHCl, ∼96% of TM_T(−5)P/N52I/K54H has the His54−heme loop, whereas only ∼3% forms the Pro(−5)− heme loop, with the remainder having water bound to the heme.

Thus, interference from Pro(−5)−heme loop formation in the denatured state is minimal at pH 6.4 in 3 M GdnHCl. The pKa(obs) of AcH54I52 is 5.65 in 6 M GdnHCl,16 which leads to a pKFeL of −1.13 Thus, for the denatured state of TM_T(−5)P/N52I/K54H in 6 M GdnHCl, we estimate that∼99% of TM_T(−5)P/N52I/K54H has water bound to the heme at pH 3.6, with the remainder having His54 as the ligand.At pH 6.4 in 6 M GdnHCl, ∼72% of TM_T(−5)P/N52I/ K54H has the His54−heme loop, whereas only ∼10% forms the Pro(−5)−heme loop, leaving ∼18% of the protein still having water bound to the heme. The strong GdnHCl dependence of the His54−heme loop lowers the proportion of the protein in the His54−heme conformer at pH 6.4, but it still remains the predominant form in solution.2D 1H−15N HSQC spectra for the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc at pH 6.4 and 3.6 in both 3 and 6 M GdnHCl. The low chemical shift dispersion under all four denaturing conditions is typical for unfoldedproteins. No partially unfolded forms are expected at 3 or 6 M GdnHCl based on the GdnHCl denaturation data in Figure 2. The crosspeaks in the 1H−15N HSQC spectra are indeed sharp under all conditions, as expected for fully unfolded proteins. Another potential concern is aggregation. However, the sharp signals that we observe in the NMR spectra and the small-angle X-ray scattering (SAXS) data reported by the laboratories of Doniach and Bilsel68,69 indicate that cytochrome c at millimolar concentrations does not aggregate in concentrated GdnHCl solutions. Also, the radius of gyration, Rg, of 30.3 ± 0.1 Å68observed in 3−5 M GdnHCl with SAXS measurements is in the range expected for an unfolded monomeric protein of the size of cytochrome c.134At pH 6.4, where the His54−heme loop is formed, backbone amide chemical shifts were assigned for 85 of 103 non-proline residues in 3 M GdnHCl and for 76 of 103 non-prolineresidues in 6 M GdnHCl (Tables S2 and S4). Not surprisingly, assignments could not be made for residues located near the paramagnetic low-spin ferric heme. In 3 M GdnHCl, backbone amide chemical shifts of K11−E21, which are adjacent to theCXXCH heme attachment site (C14−H18), could not beassigned. In 6 M GdnHCl, the unassigned region adjacent to the heme attachment site is more extensive, with backbone amide chemical shifts missing for L9−N26. The signals from amides whose chemical shifts could not be assigned in 6 M GdnHCl have low intensities in the HSQC spectrum in 3 M GdnHCl.

The somewhat lower protein concentration in the 6 M GdnHCl samples is likely the primary reason assignments could not be made for residues 9, 10, and 22−26. The other region where assignments could not be made was near His54, which is bound to the heme at pH 6.4. In 3 M GdnHCl, backbone amide chemical shifts were not assigned for D50 and I53 to N56. In 6 M GdnHCl, the region of incomplete assignments is again larger, with backbone amide chemical shifts missing for A51−L58.At pH 3.6, where the polypeptide chain of TM_T(−5)P/N52I/K54H iso-1-Cytc is unconstrained, backbone amide chemical shifts were assigned for 83 of 103 non-proline residues in both 3 and 6 M GdnHCl (Tables S3 and S5). Backbone amide chemical shift assignments could not be made for residues T8−N26 adjacent to the heme attachment site in 3M GdnHCl and for L9−V28 in 6 M GdnHCl. The region withmissing assignments near the heme attachment site is larger at pH 3.6 than at pH 6.4, possibly because of the paramagnetic ferric heme being high spin at low pH.Interestingly, a signal from the backbone amide of F36 was not identified in the 1H−15N HSQC spectra under any of the four denatured state conditions used for our NMR experiments. Signals from the neighboring residues, including I35 and G37, were present in the HSQC spectra, although they are weak at pH 3.6 (Figure 5). The correlations between the backbone amide of G37 and the 13C′, 13Cα, and 13Cβ chemical shifts of F36 could be observed in the 3D triple-resonance experiments under all conditions (Tables S2−S5). F36 is not near either the heme attachment site or His54.

DISCUSSION
SSP Score Analysis of the Denatured State of Iso-1- Cytc. To assess residual secondary structure in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc, we used the SSP score developed by Forman-Kay and co-workers with the re- referencing algorithm for 13Cα and 13Cβ chemical shifts.94 The re-referencing shifts are near −0.2 for all four experimental conditions. These shifts are somewhat large, given that all spectra were referenced to internal DSS,94 and may reflect perturbations caused by the paramagnetic heme. The SSP scores with the His54−heme loop formed at pH 6.4 in thedenatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc at 3and 6 M GdnHCl are shown in Figure 6A. Both positive (α- helix) and negative (β-sheet) SSP scores are observed. In general, the scores are more positive (or less negative) at 3 M GdnHCl than at 6 M GdnHCl. The profile of SSP scores is similar in the denatured state of TM_T(−5)P/N52I/K54Hiso-1-Cytc in 3 and 6 M GdnHCl at pH 3.6 when the loop is broken (Figure 6B). There is evidence of residual helical structure corresponding to two of the three native α-helices of iso-1-Cytc, the 60s helix and the C-terminal helix, in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc bothwith and without the His54−heme loop formed. Themagnitudes of the SSP scores, however, are large for a fully unfolded protein at high denaturant concentrations72 and are more similar to the magnitudes of SSP scores observed for acid- denatured proteins,45,70−72 which typically retain more residual structure.Correcting the SSP Score for the Paramagnetic Heme. A potential concern is the effect of the paramagnetic heme on chemical shifts for residues constrained to be near the heme in the denatured state either by the CLQCH heme attachment site or by the His54−heme bond at pH 6.4. To correct for this possibility, we used the SSP scores in 6 M GdnHCl as a reference state for the SSP scores for the denatured state in 3 M GdnHCl. Chemical shifts in 6 M GdnHCl have been used previously as a reference state to quantify residual secondarystructure in the denatured state of apoflavodoxin under milder (3.4 M GdnHCl) denaturing conditions.135 Using chemical shifts in 6 M GdnHCl also inherently corrects for the sequence dependence of random coil chemical shifts.

Because of the effects of GdnHCl on the stability of the His54−heme loop, this subtraction is not perfect at either pH but should still eliminate the majority of the effects of the paramagnetic heme.M GdnHCl as a reference state with the His54−heme loop formed at pH 6.4 and in its absence at pH 3.6. The ΔSSP scores in general fall between −0.1 and 0.1. Most ΔSSP scores are positive, indicative of residual helical structure on the order of 5−10% helix in the denatured state of TM_T(−5)P/N52I/ K54H iso-1-Cytc in 3 M GdnHCl at pH 3.6 and 6.4.Common Features of Residual Secondary Structure in the Denatured State in the Presence and Absence of the His54−Heme Loop. In general, the regions of residualhelical structure in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl, in the presence (pH 6.4) and absence (pH 3.6) of the His54−heme loop, correspond to portions of iso-1-Cytc that are helical in the native state (Figure 7). However, there are also regions of non-native helical structure that are present in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl, in thepresence and absence of the His54−heme loop. Regions ofnon-native secondary structure are common in denatured proteins.22,26,30,35−42,70−73There are three long helices in the native state of iso-1-Cytc. The presence or absence of residual structure in the denatured state of these sequence regions of TM_T(−5)P/N52I/K54H iso-1-Cytc is discussed first. The N-terminal helix has no consistent stretches of significant residual helical structure in the region between residues 2 and 7 that can be detected with a ΔSSP score (Figure 7). The large ΔSSP scores at positions 6 and 7 (positive at pH 6.4 and negative at pH 3.6) may have considerable error because the NMR signals from these residues are very weak and thus the accuracy of the chemical shifts is likely lower. The lack of residual helical structure in this portion of the N-terminal helix in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl is consistent with the lower intrinsic helical propensity predicted by Agadir137 for residues 2−7 of the N-terminal helix (FigureS1). The 60s helix is populated to approximately 5−8% in boththe open and closed loop forms of the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl (Figure 7).

There is a region of non-native helix preceding the 60s helix (approximately residues 55−61), which is populated to a similar level. A peptide fragment corresponding to the C-terminal helix of horse cytochrome c has been shown to have significant helical content.138 Consistent with this observation, there is also significant residual helical structure in the C- terminal helix, which is populated up to 5−7% (Figure 7) in thepresence and absence of the His54−heme loop in thedenatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl (Figure 7).Residual helical structure is also observed in the denatured state in 3 M GdnHCl outside of the sequence regions corresponding to the three long helices in the native state of TM_T(−5)P/N52I/K54H iso-1-Cytc. Starting at Pro76 in themiddle of Ω-loop D (residues 70−85), there is non-nativeresidual helical structure on the order of ∼5% in the presence and absence of the His54−heme loop (Figure 7). There is also residual helical structure near residue 50, which begins near residue 45 and partially overlaps the short native 50s helix, which is formed by residues 49−55 in the native state of iso-1- Cytc. There is essentially no residual helical content in the C-terminal half of the native state 50s helix in the absence of loop formation (Figure 7, pH 3.6; chemical shift data are not available for this region in the presence of the His54−heme loop at pH 6.4). The low helical propensity of histidine139 relative to that of the wild type lysine residue may be a factor in this case.The absence of an observable signal from the backbone amide of F36 under any of the denaturing conditions tested is intriguing. It could be indicative of line broadening due to intermediate exchange through participation in a transient hydrophobic cluster or perhaps transient interactions with the heme. At pH 6.4, I35 and G37 give strong amide NH crosspeaks (Figure 5). At pH 3.6, the I35 and G37 crosspeaks are weaker (Figure 5; at 6 M GdnHCl, G37 is visible but I35 is below the contour level; at 3 M GdnHCl, both crosspeaks are below the contour level). Thus, it appears that peak broadening is more pronounced when the loop is broken in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc and thepolypeptide chain is less constrained.

Roder’s lab has shownon the basis of NMR experiments that F36 and the preceding hydrophobic residue contribute significantly to nativelike structure that is present in an unfolding intermediate and also involves the N- and C-terminal helices and some residues in the 60s and 70s helices.140 Our data indicate that F36 also may promote structure in the fully denatured state even in the presence of high denaturant concentrations. Studies of heme peptide models show that Phe and Trp can make stabilizing contacts with the heme.141−143 In general, aromatic residues appear to be important for nucleating residual structure in the denatured state.10,11,13,22,27,29,76,77Effect of the Imposed His54−Heme Loop on Residual Structure in the Denatured State. The effect of the His54− heme loop on residual secondary structure in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl is relatively small (Figure 7). Significant changes occur in the C- terminal helix. At pH 6.4 (loop formed), there is residual helicalstructure over much of the length of the C-terminal helix, whereas the C-terminal end of the helix has no significant residual structure at pH 3.6. This effect could be due to protonation of Glu103, which is expected to have a pKa near4.35 in concentrated GdnHCl solutions,144 leading to a loss of stabilizing salt bridge interactions with Lys99 and Lys100 (i, i + 3 and i, i + 4 helical contacts). The differences in residual helical structure in the presence (pH 6.4) and absence (pH 3.6) of the His54−heme loop at the C-terminal end of the C-terminal helixapparent in Figure 7 are not predicted well by Agadir (Figure S1). The constraint imposed by loop formation may cause an increase in the number of transient long-range interactions between the C-terminal helix and structure induced by the His−heme loop in the denatured state of TM_T(−5)P/N52I/ K54H iso-1-Cytc in 3 M GdnHCl, which stabilizes the C- terminal helix (see below).The only other segment of TM_T(−5)P/N52I/K54H iso-1- Cytc that shows a significant enhancement of helical contentupon loop formation is the region between residues 42 and 49, which immediately precedes His54. Given the proximity of these residues to the His54−heme contact, some caution should be exercised in interpreting these ΔSSP scores.

However, in 3 M GdnHCl at pH 6.4 in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc, the intensity of the backbone amide crosspeaks for the sequence preceding His54shows periodicity suggestive of helical structure. His54 and I53 are not observed. I52 has a weak crosspeak. The A51 crosspeak is barely above the lower threshold of the spectrum shown in Figure 5, and D50 is not observed. In a helical structure, I52 would be farthest from the heme and D50 would be expected to be closest to the heme surface. Studies of heme peptide model systems show that when a His-containing peptide binds to heme there is induction of α-helix in the pep- tide.141−143,145−147 Furthermore, the helical structure inducedby His54−heme binding in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl mayprovide a docking site for the C-terminal helix, leading to theobserved increase in helical content for the C-terminal helix. In 6 M GdnHCl at pH 6.4, the amide NH crosspeaks are not observed from A51 to L58. The lower protein concentration in the 6 M GdnHCl experiments is likely a factor. However, the crosspeak for D50 is clearly visible, albeit weak (Figure 5), suggesting that the helical structure present in 3 M GdnHCl has been disrupted in 6 M GdnHCl.Implications of Residual Structure in the Denatured State for the Folding Mechanism of Cytochrome c. Kinetic hydrogen−deuterium exchange (HX) data for horse Cytc are consistent with the N- and C-terminal helices coming together to form the earliest folding intermediate, followed by the 60s helix and the Ω-loop running from residue 19 to 36.148,149 Our chemical shift data in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl showthat significant residual helical structure exists in both the 60sand C-terminal helices. Thus, these helices appear to have flickering structure that predisposes them to be involved in the early steps of the folding mechanism of cytochrome c. The portion of the N-terminal helix, which we can observe, does not contain significant residual helical structure. The helical contents of isolated peptides corresponding to the N-terminal (3% helix), 60s (7% helix), and C-terminal (27% helix) helices of horse Cytc show that only the C-terminal helix is intrinsically stable.138 In a compact molten globule at pD 2.2 and 1.5 M NaCl (A state), the protection factors of the N-terminal, 60s, and C-terminal helices are similar.150 The same remains true at pD 2.2 and 0.02 M NaCl, where the acid state of horse Cytc is expanded to a pre-molten globule state, although the protection factors for the N-terminal, 60s, and C-terminal helices are 2 orders of magnitude lower in the pre-molten globule state.151,152 For the pre-molten globule form of horse Cytc, the protection factors suggest that the equilibrium constant for helix formation is near 10 for each of these helices, much higher than expected from the helical content of these helices inisolation. The similar protection factors suggest formation of modestly stable submolecular folding units via long-range interactions between these helices.

In 3 M GdnHCl, the helical content decreases a further 2 orders of magnitude (5− 10% helical content) in the 60s and C-terminal helices (Figure 7). However, the similar helical content despite the significantly different intrinsic stabilities of isolated peptides corresponding to these helices138 suggests the possibility of transient interactions between these helices even under the strong denaturing conditions of 3 M GdnHCl. The low helical contentat the Gly83-Gly84 segment (Figure 7) may act as a flexible molecular hinge that facilitates these long-range interactions in the denatured state as suggested for Gly-containing sequencesegments in work from the laboratories of Wright and Hosur.40−42,52Our data also suggest that Phe36, which is part of the loop that folds in synch with the 60s helix, may be involved in a transient hydrophobic cluster in the denatured state involving the heme. Data from Roder’s lab also indicate that Phe36 is involved in an equilibrium folding intermediate.140 The equilibrium HX data for both the molten globule and pre- molten globule states of horse Cytc show that the amide NH group of Phe36 is protected significantly from exchange.150,151 Thus, Phe36 also may play an important role in establishing nativelike topological constraints, through a transient inter- action with the heme (see Figure 1) in the denatured state, providing for efficient folding.

CONCLUSION
The ΔSSP scores observed in the denatured state of TM_T(−5)P/N52I/K54H iso-1-Cytc in 3 M GdnHCl at pH 6.4, when an imposed loop between His54 and the heme is present, and at pH 3.6, when the loop is absent, indicate that modest amounts of residual helical content (5−8%) remain in the 60s and C-terminal helices. Significant equilibrium HX protection factors also are observed for these helices in the molten globule and pre-molten globule states of horse Cytc. The portion of the N-terminal helix that can be detected in our data appears to be largely devoid of residual helical structure in the presence and absence of the His54−heme loop. The similar residual helical content in the 60s and C-terminal helices in 3 M GdnHCl is consistent with data for both the pre-molten globule and molten globule states of horse Cytc.150,151 Collectively, these observations suggest that the early formation of structure in these helices during folding is rooted in intrinsic topological biases that persist in the denatured state. In the presence of the His54−heme loop, there is an increase in the residual helical structure content that is evident at the C-terminus of the C- terminal helix. There may also be some enhancement of helical content for residues 42−49 preceding the His54−heme bond and in the 60s helix following it. Overall, the observed effects of His54−heme loop formation are weak but similar in magnitude to those of the long-range changes in helical structure induced by mutations in the denatured state of acyl-coenzyme A binding protein.