Influence of the Interdomain Interface on Structural and Redox Properties of Multiheme Proteins

Abstract

Multiheme proteins are important in energy conversion and biogeochemical cycles of nitrogen and sulfur. A diheme cytochrome c₄ (c₄) was used as a model to elucidate roles of the interdomain interface on properties of iron centers in its hemes A and B. Isolated monoheme domains c₄-A and c₄-B, together with the full-length diheme c₄ and its Met-to-His ligand variants, were characterized by a variety of spectroscopic and stability measurements. In both isolated domains the heme iron is Met/His-ligated at pH 5.0, as in the full-length c₄, but becomes His/His-ligated in c₄-B at higher pH. Intradomain contacts in c₄-A are minimally affected by the separation of c₄-A and c₄-B domains, and isolated c₄-A is folded. In contrast, the isolated c₄-B is partially unfolded and the interface with c₄-A guides folding of this domain. The c₄-A and c₄-B domains have propensity to interact even without the polypeptide linker. Thermodynamic cycles have revealed properties of monomeric folded isolated domains, suggesting that ferrous (Fe^II) but not ferric (Fe^III) c₄-A and c₄-B are stabilized by the interface. This study illustrates effects of the interface on tuning structural and redox properties of multiheme proteins and enriches our understanding of redox-dependent complexation.

Graphical Abstract

INTRODUCTION

Heme proteins are involved in many biological functions, including catalysis, electron transfer (ET), and signaling.¹ Structure and folding of these proteins are influenced by the presence of the heme cofactor and the polypeptide packing around it in turn tunes the reactivity of this group.^2-4 While there is a growing understanding of the effects of the polypeptide-derived first and second coordination sphere on redox properties of the heme iron,^4-7 more distant effects remain less understood. Yet, studies of protein association suggest that effects emanating from the protein surface could greatly influence potentials and reactivity of these metal centers.^8-11 Further, the presence of the nearby hemes or other redox centers and their oxidation state affect redox properties of multiheme proteins.^12-14

Proteins having two or more hemes are common, particularly in bacteria associated with low-oxygen or anaerobic environments.^15-16 These proteins are important for energy conversion, enabling long-range ET processes, including those that connect inside and outside of cells, and catalyzing key reactions in biogeochemical cycles of nitrogen and sulfur.^16-18 While some multiheme proteins have no distinct interdomain boundaries,^19-20 others have domains separated in distinct modules, with clearly defined interfaces.^21-22 The interface contacts influence the coordination environment and solvation of the hemes,^12-13 and redox-dependent changes in interface contacts serve to activate catalysis.¹² Despite the functional importance of interdomain interface in heme proteins, systematic studies of their effects are scarce due to the inherent complexity of multidomain proteins and difficulty in data interpretation of composite effects.

Diheme proteins could be considered as building blocks of multiheme proteins and thus serve as models for exploring the effects of the interdomain interfaces on redox properties of the heme iron centers. Herein, we investigate a diheme protein cytochrome c₄ (c₄) from Pseudomonas stutzeri (Ps), to decipher the role of the interdomain interface on the structure, stability, and reduction potentials of its two domains. The two domains c₄-A and c₄-B (Figure 1) are tethered by a flexible linker and, as suggested by the crystal structure,²² related by a pseudo-two-fold axis passing between the inner heme propionates (HPs) of the two hemes. The two domains have high structural similarity with a backbone root-mean-square deviation (RMSD) of 1.11 Å, while sharing relatively low sequence identity (34.5%). The heme iron centers in c₄-A and c₄-B are both Met/His-ligated. The interdomain contacts involve hydrogen bonds and hydrophobic interactions and appear to be highly symmetric (Figure S1).

Figure 1.

(A) Structure of Ps c₄ (PDB ID: 1M70)³³ showing the two heme domains related by a pseudo-two-fold symmetry and the Met/His ligation at the heme iron centers. The c₄-A and c₄-B domains correspond to amino acids 1-92 (green) and 93-190 (magenta), respectively. (B) Schematic representation of the studied protein constructs for the full-length c₄, its two isolated domains (c₄-A and c₄-B), and two M-to-H variants (M66H and M167H). Ligands to the heme iron are shown.

Characterization of the full-length c₄ protein and its two isolated domains c₄-A and c₄-B by a variety of spectroscopic techniques and stability measurements has revealed that the interface has distinct effects on properties of ferric (Fe^III) and ferrous (Fe^II) proteins. The split of the full-length c₄ into two isolated domains exposed sites involved in forming the interdomain interface, increasing the tendency of the domains for oligomerization and causing pH-dependent misligation of the heme iron in c₄-B. Oligomerization of the isolated domains complicates studies of the interdomain interface, but by taking advantage of alterations in redox-dependent stabilization upon replacement of the heme iron axial ligand in monomeric Met-to-His variants of c₄, a full thermodynamic analysis of interface effects became possible. The study illustrates how combining the two monoheme domains in a diheme protein structure modulates redox properties, folding, and recognition of the domains and provides a framework for analyses of elusive interface effects in other metalloprotein systems.

MATERIALS AND METHODS

General.

All chemicals were purchased from Fisher Scientific Inc. and VWR International, unless noted otherwise. Buffers were prepared using reagent-grade chemicals. Water was purified to a resistivity of 18.2 MΩ cm using Barnstead E-Pure Ultrapure Water Purification System. All experiments were done at room temperature unless stated otherwise.

Data analyses were performed using MATLAB_R2019a (MathWorks).²³ Images of protein structures were prepared with UCSF Chimera (version 1.31.1).²⁴

Construction of Expression Plasmids.

The plasmid carrying the gene for Ps ATCC14405 c₄ was previously described.²⁵ The genes encoding the amino acids 1-92 and 93-190 were amplified and subcloned into the pET22b(+) vector through BamHI and XhoI sites to prepare constructs 22b-Ps c₄-A and 22b-Ps c₄-B for expression of the two isolated domains. The extra nucleotides between the pelB leader and the sequence corresponding to the full-length Ps c₄ or isolated domains were removed by mutagenesis using a Quikchange kit (Agilent Technologies, Inc.) to yield the final plasmids 22b-Ps c₄ΔN, 22b-Ps c₄-AΔN, and 22b-Ps c₄-BΔN. Variants M66H and M167H were created by mutagenesis using the 22b-Ps c₄ΔN plasmid as a template. Plasmid DNA was extracted and purified with the Miniprep Kit (Qiagen), and the sequence was verified at the Molecular Biology and Proteomics Core Facility (Dartmouth College).

Expression and Purification of Proteins.

The proteins expressed using the 22b-Ps c₄ΔN, 22b-Ps c₄-AΔN, and 22b-Ps c₄-BΔN plasmids are referred to in this study as c₄, c₄-A, and c₄-B, respectively. All protein variants were expressed and purified according to the published procedure for Ps c₄²⁵ with the following modifications. The transformed cells were plated onto a LB agar plate containing 100 mg/L ampicillin and 68 mg/L chloramphenicol and incubated for about 20 hours at 37 °C. The large-scale cultures in TB media containing 100 mg/L ampicillin and 34 mg/L chloramphenicol were incubated for 16-20 hours at 30 °C and 180 rpm. The harvested cells were resuspended in a solution containing 30% sucrose (w/v), 100 mM Tris at pH 8.0, and 1 mM EDTA followed by osmotic shock to extract the periplasmic proteins. The supernatant from osmotic shock was loaded onto a Q Sepharose Fast Flow batch column (about 60 mL resin). The eluted protein was precipitated by the addition of (NH₄)₂SO₄ powder (30 g per 100 mL solution). The resuspended protein was then loaded onto a HiTrap SP HP 5 mL cation exchange column pre-equilibrated with a 10 mM acetate buffer at pH 5.5. The flow-through was collected and exchanged into a 10 mM Tris-HCl buffer at pH 7.5. Then, the protein solution was further purified using a HiTrap Q HP 5 mL anion exchange column.

Spectroscopic Measurements.

Electronic absorption spectra were collected on an Agilent 8453 diode-array spectrophotometer. Circular dichroism (CD) spectra in the far-UV range were recorded on a J815 CD spectropolarimeter equipped with a variable temperature Peltier cell device (JASCO). Extinction coefficients of c₄ proteins were determined using a pyridine hemochrome assay.²⁶ pH titrations were performed and analyzed as previously described.²⁷ The pH titrations of ferric c₄, c₄-A, and c₄-B in the unfolded state were performed in the presence of 4 M guanidine hydrochloride (GuHCl).

EPR spectra were recorded on a Bruker EMX 300 X-band EPR spectrometer (Bruker Biosciences Corp.) at 10 K and the experimental parameters were adapted from previous studies of cytochrome (cyt) c proteins.²⁸ Before measurements, proteins were treated with excess of potassium fern cyanide to yield ferric proteins, repurified, and concentrated to about 500 μM. All samples were prepared in a 100 mM sodium phosphate buffer at pH 7.4 or a 100 mM sodium acetate buffer at pH 5.0.

¹H NMR spectra were recorded on a 500 MHz Bruker NMR spectrometer (Bruker Biosciences). Ferrous samples were prepared in a 50 mM sodium phosphate buffer at pH 7.4, or a 50 mM sodium acetate buffer at pH 5.0 containing 10% (v/v) D₂O under anaerobic conditions in a nitrogen-filled glove box (COY Laboratory Products), and 5 mM sodium dithionite was kept in the solution to ensure the heme iron stayed reduced during the measurements. Data collection and analyses were done as previously described.²⁸

Resonance Raman (RR) spectra were recorded at room temperature and at 110 K using a 407 and 568 nm laser excitation from a krypton laser (Innova 302C, Coherent) on solutions with protein concentrations 150 μM (c₄-A and c₄-B) and 75 μM (c₄) either in a 50 mM sodium phosphate buffer pH 7.5 or a 50 mM sodium acetate buffer pH 5.0. The ferrous proteins were prepared by addition of 1 mM sodium dithionite to samples of ferric proteins. For measurements at room temperature, scattered photons were collected using a 90° geometry on Raman capillaries; for low temperature measurements, samples in NMR tubes were maintained at 110 K inside a liquid nitrogen cold-finger and scattered photons were collected using a backscattering geometry. Raman photons were focused on the entrance slit of a McPherson 2061/207 spectrograph equipped with a liquid N₂-cooled CCD camera (LN1100PB, Princeton Instruments) with attenuation of the Rayleigh scattering from a long-pass filter (razorEdge, Semrock). Frequency calibrations were performed using indene and aspirin and are accurate within ± 1 cm⁻¹. Preventing photoreduction of the ferric proteins with the 407 nm excitation required laser powers below 1 mW and rapid reciprocal translation of the Raman capillaries across the laser beam. Electronic absorption spectra collected directly from the Raman capillaries were unchanged after acquisition of the RR data, confirming the integrity of the samples.

Determination of Oligomeric States.

Oligomeric states of ferric and ferrous proteins were examined by size exclusion chromatography (SEC). Superdex 200 Increase 10/300 GL (Cytiva product 28990944) was pre-equilibrated with a 50 mM sodium acetate buffer at pH 5.0 at a flow rate of 1 mL/min. Small volumes (about 150 μL) of proteins at 50-500 μM were injected onto the column and absorbance signals at 280 and 410 nm were monitored. The column was calibrated with the following standards: apoferritin (443 kDa), β-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and horse heart cyt c (12.4 kDa). The column void volume was determined by the elution volume of blue dextran.

Spectroelectrochemistry.

Electrochemical experiments were carried out in a buffer containing 50 mM Tris and 100 mM NaCl at pH 7.0 as previously described.²⁷ The final concentration of the protein was 25-50 μM. 100 μM sodium dithionite was included in the samples to make ferrous proteins at the start of the titration. The following mediators (50 μM each) were added into the sample solution to facilitate electrical interaction between protein and electrode: p-benzoquinone (293 mV), TMPD (260 mV), 1,2-naphthoquinone (143 mV), and phenazine methosulfate (80 mV).

Chemical Denaturation.

The protein global stability at 20 °C was examined by chemical denaturation using GuHCl and monitoring the ellipticity at 222 nm of equimolar protein samples on a JASCO-J815 CD spectropolarimeter. A series of GuHCl solutions were prepared in a 100 mM sodium acetate buffer at pH 5.0 and concentrations of the denaturant were determined by refractive index measurements²⁹ using an AO Scientific Instrument ABBE Mark II Digital Refractometer.

For GuHCl titrations of ferric proteins, the denaturation process was initiated by adding the protein into GuHCl solutions and equilibrating samples for at least 15 min. Each protein sample in the series was then transferred into a 1-mm cuvette for the CD measurements. The final protein concentrations were about 10 μM for the full-length wild type (WT), M66H, and M167H. The final protein concentrations were about 20 μM for c₄-A and c₄-B.

For GuHCl titrations of ferrous proteins, the samples were prepared in a nitrogen-filled glove box (COY Laboratory Products) and put in a 4-mm quartz anaerobic cuvette; 0.5 mM sodium dithionite was included to keep the protein reduced. The final protein concentrations were 2.5-3.0 μM for the full-length WT, M66H, and M167H. The final protein concentrations were 5-6 μM for c₄-A and c₄-B. Changes in ellipticity at 222 nm were plotted against GuHCl concentrations and fit to eqs 1 and 2 for one- and two-transition curves,³⁰ respectively. Data analyses followed the previously described method of fitting the transition region of unfolding curves,^31-32 with some modifications. The four parameters, $m_{\mathrm{f}}$, $b_{\mathrm{f}}$, $m_{\mathrm{u}}$, and $b_{\mathrm{u}}$, representing features of the pre- and post-unfolding processes (approximated as linear dependencies) were first determined. These four parameters were then fixed at the values derived from the first round of fitting and the data were fit to eq 1 or eq 2, yielding the final values of ${[\mathrm{D}]}_{1∕2}$, $m_{\mathrm{D}}$, and ${\Delta G}_{\mathrm{u}}({[\mathrm{D}]}_{1∕2}\times m_{\mathrm{D}})$. This iterative procedure, focusing on the linear regions at first, helps in fitting dependences to equations with many parameters but underestimates errors.³¹

$$f([\mathrm{D}])=\frac{m_{\mathrm{f}}[\mathrm{D}]+b_{\mathrm{f}}+(m_{\mathrm{u}}[\mathrm{D}]+b_{\mathrm{u}})\times \text{exp}\left[\frac{m_{\mathrm{D}}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2})}{RT}\right]}{1+\text{exp}\left[\frac{m_{\mathrm{D}}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2})}{RT}\right]}$$ eq 1

$$\begin{gathered} f([\mathrm{D}])=\frac{m_{\mathrm{f}1}[\mathrm{D}]+b_{\mathrm{f}1}+(m_{\mathrm{u}1}[\mathrm{D}]+b_{\mathrm{u}1})\times \text{exp}\left[\frac{m_{\mathrm{D}1}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2,1})}{RT}\right]}{1+\text{exp}\left[\frac{m_{\mathrm{D}1}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2,1})}{RT}\right]} \\ +\frac{(m_{\mathrm{u}2}[\mathrm{D}]+b_{\mathrm{u}2})\times \text{exp}\left[\frac{m_{\mathrm{D}2}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2,2})}{RT}\right]}{1+\text{exp}\left[\frac{m_{\mathrm{D}2}([\mathrm{D}]-{[\mathrm{D}]}_{1∕2,2})}{RT}\right]} \end{gathered}$$ eq 2

Molecular Dynamics Simulations.

Structural models of the full-length c₄ and the two isolated domains c₄-A and c₄-B were constructed based on the crystal structure of ferric c₄ from Ps ATCC11607 strain (PDB ID: 1M70).³³ There are four chains (A, B, C, and D) in the crystal structure, in which chain A exhibits notable differences in residues 161-165 compared to other three chains, presumably due to crystal packing. Chains B, C, and D are all highly similar, with alpha carbon (Cα) RMSD values of 0.57±0.03 Å, and chain B was selected as an initial structure for the molecular dynamics (MD) simulations. Not Just a Molecular Dynamics (NAMD)³⁴ was used for simulations, and structural analysis was performed by Visual Molecular Dynamics (VMD),³⁵ as previously described³⁶ with the following modifications. The solvation of each initial structure was achieved by adding a water box of TIP3 model, which extended 10 Å from the protein surface. The system was first minimized for 5,000 steps with conjugate gradients using CHARMM22 force field parameters³⁷ for geometry optimization and further equilibrated for 50,000,000 steps (2 fs per step, 100 ns in total) at 300 K and 1 atm using the Langevin piston method. The trajectory data were saved every 10,000 steps (20 ps) and the snapshots (1000 frames) from the trajectory data of 80-100 ns representing the states in equilibrium were used for the following analyses. The root mean-square fluctuations (RMSF) of residues were evaluated in VMD using positions of Cα. The contact frequencies with a distance cutoff of 10 Å were calculated with MD Analysis toolkit³⁸ in Python 3 based on the residue-to-residue distances defined by the closest distance between any two non-hydrogen atoms. The heat maps of contact frequencies were then generated in MATLAB. Solvent-accessible surface areas (SASA) for the heme over the trajectories were calculated in VMD using a probe radius of 1.4 Å.

Structural Analyses.

To investigate similarity of monoheme c₄-A and c₄-B domains to other heme proteins, structure-based similarity searches against PDB90 subset database were performed in the DALI server.³⁹ The output structures with a Z score (an optimized similarity score defined as the sum of equivalent residue-wise Cα-Cα distances between two proteins) above 4 were manually examined. The resulting structures are all of the heme c-containing proteins. After excluding the structures from multiheme proteins and the redundant structures for the same protein or mutants, those of proteins with a single c-type heme were used for structural comparisons.

Structural models of c₄-A and c₄-B homodimers were generated using AlphaFold2-multimer Colab⁴⁰ (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb) with two copies of residues for c₄-A and c₄-B as input sequences, respectively. No template information is used and complex prediction "AlphaFold-multimer-v2" is applied. The interdomain interface in the full-length c₄ and the dimerization interface in c₄-A and c₄-B homodimers were analyzed in InterProSurf⁴¹ web server. Residues that establish hydrophobic contacts at the interface were determined using LigPlus+.⁴² Electrostatic potentials of the protein surfaces were calculated with PDB2PQR and APBS tools⁴³ in Chimera and mapped on the protein structures.

RESULTS

Preparation of c₄ Variants.

The red color of cell pellets indicated successful expression of all the variants; the yield of pure proteins after the final step of the HiTrap Q column purification was 5-10 mg per L of the culture medium. The purity of the proteins was confirmed by SDS-PAGE gel (Figure S2A). Heme staining of the SDS-PAGE gel showed the presence of hemes in all the variants (Figure S2B).

Secondary Structure.

Secondary structure of the proteins was probed by far-UV CD spectroscopy. Both the full-length c₄ and the two isolated domains c₄-A and c₄-B showed features characteristic of an α-helical structure. CD measurements for ferric c₄ and c₄-A at pH 7.4 yielded α-helical contents similar to those calculated based on the crystal structure of the protein,²² while the α-helical content of ferric c₄-B calculated from CD spectra was lower than that from the crystal structure (Figure 2A and Table S1). In contrast, CD measurements for all three ferrous proteins (c₄, c₄-A, and c₄-B) revealed α-helical contents fully consistent with the crystal structure (Figure 2B and Table S1). These findings suggest disruption of the secondary structure of c₄-B in the ferric but not the ferrous state. Interestingly, the α-helical content of ferric c₄-B increases at lower pH (Figure 2A).

Figure 2.

Far-UV CD spectra of (A) ferric and (B) ferrous c₄ (black), c₄-A (green), and c₄-B (magenta) variants at pH 7.4. The spectrum of ferric c₄-B at pH 5.0 is also shown (magenta-dash) and reveals changes in the secondary structures of this variant at lower pH. The mean residue molar ellipticity (y-axis) accounts for the difference in the number of residues.

Heme Ligation.

Ferric State.

The electronic absorption spectra of ferric proteins at pH 7.4 are indicative of six-coordinate low-spin (6c LS) heme ligation in the intact full-length c₄ and the two isolated domains c₄-A and c₄-B (Figure 3A, Table S2). The presence of a characteristic charge-transfer (CT) band at ~700 nm⁴⁴ further supports Met ligation to the ferric heme iron for c₄ and c₄-A. In contrast, this band is not observed for c₄-B at pH 7.4, implying that the heme iron ligand in this variant may not be Met. The EPR spectrum of ferric c₄ at pH 7.4 (Figure 3B) exhibits two sets of signals corresponding to Heme A and Heme B, as previously described.^25,45 The EPR spectrum of c₄-A is very similar to that for Heme A in the full-length c₄, suggesting a similar environment of this heme in both proteins. The EPR spectrum of c₄-B reveals typical rhombic features of a 6c LS ferric heme iron that deviate slightly from those of the full-length c₄.⁴⁶

Figure 3.

(A) Electronic absorption spectra at room temperature, (B) EPR spectra at 10 K, (C) spectral changes with pH for ferric c₄-B, and (D-F) RR spectra at room temperature of ferric c₄ (black), c₄-A (green), and c₄-B (magenta) variants at pH 7.4. The spectra of ferric c₄-B at pH 5.0 are also shown (magenta-dash).

The RR spectra of ferric c₄ and c₄-A are nearly identical at pH 7.4. The spectra are easily assigned by comparison to RR results with cyt c⁴⁷ and the prior RR characterization of Ps c₄.⁴⁸ The spectra obtained with Soret (407 nm) excitation show ν₄, ν₃, and ν₁₀ modes at 1370, 1503, and 1638 cm⁻¹, respectively; these frequencies are consistent with a 6c LS heme configuration (Figure 3D). These modes, however, are upshifted by 2-3 cm⁻¹ in c₄-B and match frequencies previously assigned to the His-misligated forms of cyt c.⁴⁹ Comparison of the RR spectra obtained with Soret (407 nm) and Q-band (568 nm) excitations allows assignment of ν₁₁ and ν₁₉ modes at 1559 and 1587 cm⁻¹ in ferric c₄. Again, while these frequencies are unchanged in c₄-A, they are upshifted by 7 and 2 cm⁻¹ in c₄-B (Figure 3E). In addition to these changes in porphyrin macrocycle vibrations, the ν(C_a-S) and δ(C_βC_aC_b) modes observed with c₄-B are downshifted by 7 cm⁻¹ relative to full-length c₄ and c₄-A (Figure 3F), indicating conformational rearrangements of the thioether linkages.

The RR spectra of ferric c₄ and c₄-A at pH 5.0 are very similar to those at pH 7.5 (Figure S3), suggesting no pH-dependent changes in the heme environment. In contrast, the RR spectra of ferric c₄-B exhibit a pH dependence, with downshifts of all porphyrin macrocycle modes at pH 5.0 to produce a set of frequencies that are more similar to those in the full-length c₄. The C_a-S stretch and C_β-C_a-C_b deformation modes of c₄-B are also affected by pH, upshifting toward frequencies seen in the full-length c₄ at lower pH (Figure 3D, 3E, and 3F). The pH dependence of ferric c₄-B is also apparent in the electronic absorption spectra as the ~700 nm CT band, which is absent at pH 7.4, is clearly visible at pH 5.0 (Figure 3A).

Titrations monitoring electronic absorption spectra of ferric c₄, c₄-A, and c₄-B as a function of pH were carried out (Figure 3C and Table S3). Upon decrease in pH (<4.0), the Soret band of the isolated c₄-A and c₄-B domains blueshifts to ~395 nm and increases in intensity, accompanied by appearance of a band at ~620 nm, characteristic of the high-spin (HS) heme.⁴⁴ The titration of ferric full-length c₄ has been previously reported.^50-51 The Soret band of c₄ also blueshifts to ~395 nm and a new broad feature at ~370 nm appears. The CT band at 705 nm decreases and the band at ~620 nm increases in intensity. The pK_a values for the LS-to-HS heme transition are 3.3, 2.5, and 3.0 (Table S3) for c₄, c₄-A, and c₄-B, respectively. A similar transition has been well-characterized during acid unfolding of cyt c and has been assigned to the replacement of the LS axial ligand by water.⁴⁴ In the case of c₄-B, a second transition with pK_a,2 = 5.5±0.2 is readily apparent (Figure 3C) during which the heme remains LS. Taken together, these spectroscopic results suggest that, while His ligation in ferric c₄-B dominates at near neutral pH, the Met ligation, as in the native full-length c₄ protein, is mostly regained at pH 5.0 (78±6% Met-ligated based on pK_a,2). The proteins c₄ and c₄-A are Met-ligated.

Ferrous State.

Electronic absorption spectra of ferrous c₄, c₄-A, and c₄-B are consistent with features of the 6c LS ferrous heme (Figure 4A and Table S2).⁴⁴ At pH 7.4, two ¹H NMR signals at −3.20 and −3.59 ppm in spectra of the full-length c₄ are assigned to the methyl protons of the axial ligands Met66 and Met167 (Figure 4B). For c₄-A, there are two signals for the methyl protons of Met66 at pH 7.4, suggesting differences in its environment, but only one signal at pH 5.0 (Figure S4A). Varying concentrations of ferrous c₄-A between 25 and 400 μM introduces distinct changes in the spectrum of this variant (Figure S4B). For cyt c₄-B, there is only one signal for the methyl protons of Met167 at both pH 7.4 and 5.0, at high as well as low protein concentrations, and its position is very close to that of the corresponding signal in c₄.

Figure 4.

(A) Electronic absorption, (B) ¹H NMR, and (C, D) RR spectra of ferrous c₄ (black), c₄-A (green), and c₄-B (magenta) variants at pH 7.4.

The RR spectra of ferrous c₄, c₄-A, and c₄-B are remarkably similar (Figure 4C and 4D). Changing the pH of the samples from 7.5 to 5.0 has no effect on the RR spectra of ferrous variants (Figure S3). Our RR analysis supports the 6c LS ferrous heme iron configuration in all three variants. Specifically, in-plane skeletal modes ν₄, ν₃, ν₂, and ν₁₀ are observed at 1360, 1492, 1592 and 1622 cm⁻¹, respectively, for c₄ and c₄-A. For c₄-B, the ν₁₀ mode is at the same frequency and the ν₄, ν₃, and ν₂ modes upshift by only 1 cm⁻¹. Other in-plane skeletal modes ν₂₁, ν₂₀, ν₂₉, and ν₁₁ (at 1310, 1382, 1395, and 1545 cm⁻¹, respectively), are the same in all three variants.

Out-of-plane porphyrin vibrations, such as pyrrole folding modes γ₂₁ at 564 and 548 cm⁻¹, pyrrole swivel modes γ₁₂ and γ₂₂ at 518 and ~450 cm⁻¹, also show nearly identical frequencies and intensities in all the proteins, supporting a conserved level of out-of-plane porphyrin deformation. Subtle distinction between c₄-A and c₄-B is apparent, for example, the γ₂₂ mode is observed at 445 and 451 cm⁻¹ for ferrous isolated c₄-A and c₄-B domains, respectively, and these features accumulate to produce an unresolved broad band at 448 cm⁻¹ in the full-length c₄. The thioether ν(Ca-S) is at 688 cm⁻¹ in all the variants. The C-C-C bending modes from the peripheral groups are also similar, with the only small deviation of a lower enhancement for one of the two thioether deformation modes for c₄-B.

Isolated Domains Oligomerize in a Redox-Dependent Manner.

Oligomeric states of the protein variants were determined by SEC (Figures 5 and S4C). With 150 μL injections of ~200 μM proteins, the full-length c₄ is a monomer in both ferric and ferrous states. Ferric c₄-A and c₄-B domains are a dimer and a monomer at pH 5.0, respectively. Both isolated domains further dimerize at pH 5.0 upon reduction of the heme iron, forming a tetramer and a dimer. While ferrous c₄-A is a tetramer at pH 5.0, it appears to exist primarily as a dimer at pH 7.4 (Figure S4C).

Figure 5.

(A) SEC profiles of ferric (solid) and ferrous (dash) c₄ (black), c₄-A (green), and c₄-B (magenta) variants at pH 5.0 from an injection of 150 μL of ~200 μM proteins and (B) oligomeric states derived from analyses of SEC data. The elution volumes of blue dextran and standards are indicated in dashed gray vertical lines and the molecular weight values of the standards are showed on the top in kDa. From left to right: blue dextran, apoferritin (443 kDa), β-amylase (200 kDa), alcohol dehydrogenase (150 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and horse heart cyt c (12.4 kDa). The shoulder peak in panel A (labeled with *) represents a small population of ferric c₄-A formed during the elution from a SEC column in the presence of air, evidenced by the color and electronic absorption spectrum of this elution fraction.

Based on the results of SEC experiments (Figure S4C) and concentration dependence of the NMR spectrum of ferrous c₄-A (Figure S4B), we hypothesize that the two sets of Met66 signals (Figures 4B) arise from a dimer and tetramer in slow exchange (~45 s⁻¹) on the NMR timescale. Intensities of these signals provide a strategy to estimate the dissociation constant K_d for the tetramer-dimer equilibrium, yielding K_d~200 μM for ferrous c₄-A at pH 7.4. Since the domain is a tetramer at 25 μM at pH 5.0, we could only set the upper boundary (K_d<25 μM) at this pH. A similar approach to quantify dimer-monomer equilibrium is not applicable to ferrous c₄-B because only one set of Met167 methyl signals was detected (Figure 4B). The finding of the concentration independence of signals in NMR spectra, however, suggests the upper limit K_d<25 μM for c₄-B.

Interestingly, adding ferrous c₄-B to ferrous c₄-A at pH 7.4 modifies the NMR spectra of the two isolated domains, yielding a separate set of methionine methyl signals at the same positions as in the full-length c₄ (Figure S4D). These signals grow in intensity at higher concentrations of c₄-B and the original signals of c₄-A homotetramer and c₄-B homodimer decrease in intensity, suggesting that the c₄-A/c₄-B heterodimer is formed.

Reduction Potentials Decrease in the Isolated Domains.

Spectroelectrochemistry measurements reveal that reduction potentials of the heme iron in the isolated monoheme c₄-A and c₄-B domains decrease compared to those in the full-length c₄ (Table 1 and Figure S5). This observation could be qualitatively explained by the loss of the interdomain interface and, accordingly, more exposure of the heme groups when the two domains are separated. However, the quantitative interpretation needed to understand the origin of these effects is complicated by the oligomerization of c₄-A and c₄-B and changes in the folded state of c₄-B upon the polypeptide truncation. Our subsequent studies resolved these complications with measurements of redox-dependent protein stability and introduction of two additional c₄ variants.

Table 1.

Reduction Potentials of the Heme Iron in Diheme Ps c₄, its Monoheme Isolated Domains, and the Diheme Met-to-His Variants

	Measureda			Calculatedb
Variants	Oligomeric states	$E_{\mathrm{m}}$ (mV vs SHE)		$E_{\mathrm{m}}$ (mV vs SHE)
	(FeIII ⇌ FeII)	Heme A	Heme B
c 4	Monomer ⇌ Monomer	258 ± 4 (Ref25)	364 ± 6 (Ref25)	N/A
c4-A	Dimer ⇌ Tetramerc	151 ± 3	N/A	150±121d
c4-B	Monomer ⇌ Dimer	N/A	268 ± 5	75±25e
M66H	Monomer ⇌ Monomer	20±3	355±3	N/A
M167H	Monomer ⇌ Monomer	268±3	−32±4	N/A

^aMeasured at pH 7.0.

^bCalculated $E_{\mathrm{m}}$ for the monomeric isolated fragments from analyses of thermodynamic cycles.

^cDepends on the concentration of c₄-A, see Figure S4B.

^dFor the f domain.

^eFor the puf domain. The $E_{\mathrm{m}}$ for the f c₄-B domain is estimated to be 247 mV.

The spectroelectrochemistry titration curve for the mixture of c₄-A and c₄-B at equimolar concentrations of proteins shifts to higher potential values than those in isolated c₄-A and c₄-B (Figure S5A), consistent with formation of the c₄-A/c₄-B complex. The protein mixture under these conditions consists of both bound and unbound species (Figure S4D) and apparent potentials are only about 20 mV lower than those in the full-length c₄.

Global Stability of the Full-Length c₄ and the Isolated c₄-A and c₄-B Domains.

When unfolding of a protein involves only folded and unfolded states, a two-state transition model is widely used to interpret denaturation data, providing three thermodynamic parameters ([GuHCl]_1/2, $m_{\mathrm{D}}$, and ${\Delta G}_{\mathrm{u}}$).^31-32 The parameters have the following meaning: [GuHCl]_1/2 is the denaturant concentration at which a half of the protein population is unfolded; $m_{\mathrm{D}}$ value correlates with the change in SASA of the protein upon unfolding; and ${\Delta G}_{\mathrm{u}}$ is the free energy of the unfolding in zero denaturant and equals to ${[\text{GuHCl}]}_{1∕2}\times m_{\mathrm{D}}$. This model has been widely applied to folding studies of monoheme cyt proteins.^52-53

Ferric c₄-A and c₄-B display single-transition profiles but two unfolding transitions are apparent for ferric c₄ (Figure 6A). Comparison of the results for the isolated domains to those for the full-length protein (Table 2) allowed us to assign the first (lower [GuHCl]_1/2) and second (higher [GuHCl]_1/2) transitions associated with c₄ to unfolding of the c₄-B and c₄-A domains, respectively. This sequence differs from the one previously suggested based on the spectroscopic features of the two hemes rather than the polypeptide structural markers we employ here.⁵¹ The ${\Delta G}_{\mathrm{u}}$ difference of 3.0 kcal mol⁻¹ (Tables 2 and 3) for c₄-B in the isolated domain and in the full-length c₄ presumably arises from the contribution of the interdomain interface and any distinct intradomain contacts for the two forms.

Figure 6.

GuHCl denaturation curves at pH 5.0 for (A) ferric and (B) ferrous c₄ (black), c₄-A (green), and c₄-B (magenta) showing fractions of the unfolded state (χ_u) from CD ellipticity measurements at 222 nm. (C) Scheme depicting the interface breakup and unfolding of c₄. The interface is annotated as Int and shown in black-dash line. The folded and unfolded domains are annotated as A_f (or B_f) and A_u (or B_u), respectively. Only folded and unfolded species, without intermediate species, are shown in this simplified scheme. Our further analyses derive other states in the unfolding process; these are shown in Figures S10 and S11.

Table 2.

Thermodynamic Parameters Describing Unfolding of Diheme Ps c₄, its Monoheme Isolated Domains, and the Diheme Met-to-His Variants by GuHCl Denaturation at pH 5.0

		Ferric			Ferrous
Variants	Heme	[GuHCl]1/2, M	$m_{\mathrm{D}}$, kcal mol−1 M−1	${\Delta G}_{\mathrm{u}}$, kcal mol−1	[GuHCl]1/2, M	$m_{\mathrm{D}}$, kcal mol−1 M−1	${\Delta G}_{\mathrm{u}}$, kcal mol−1
c 4	A	1.68 ± 0.04	2.63 ± 0.35	4.42 ± 0.60	3.30 ± 0.42	2.20 ± 0.19	7.24 ± 0.66
	B	0.87 ± 0.02	3.91 ± 0.35	3.37 ± 0.31
c4-A	A	2.05 ± 0.13	2.71 ± 0.12	5.55 ± 0.26	4.55 ± 0.31	1.72 ± 0.08	7.83 ± 0.38
c4-B	B	0.14 ± 0.02	2.82 ± 0.16	0.40 ± 0.06	2.50 ± 0.35	2.24 ± 0.21	5.58 ± 0.57
M66H	A	1.17 ± 0.04	2.29 ± 0.29	2.68 ± 0.35	1.17 ± 0.06	3.82 ± 1.30	4.47 ± 1.53
	B	0.35 ± 0.01	3.81 ± 0.25	1.33 ± 0.10	2.19 ± 0.03	2.50 ± 0.31	5.46 ± 0.69
M167H	A	1.69 ± 0.02	2.35 ± 0.18	3.98 ± 0.30	4.01 ± 0.07	2.65 ± 0.69	10.64 ± 2.76
	B	0.59 ± 0.01	3.79 ± 0.28	2.22 ± 0.17	1.43 ± 0.03	3.79 ± 0.56	5.44 ± 0.81

Table 3.

Effects of the Interdomain Interface on the Monoheme c₄-A and c₄-B Domains

Changea	Heme A	Heme B
${\Delta E}_{\mathrm{m}}$ (mV)	119±121	289±26
${\Delta \Delta G}_{\text{Int}}^{2-3}$ (kcal mol−1)b	−2.7±2.8	−6.6±0.6
${\Delta G}_{\text{Int}}^3$ (kcal mol−1)c	0	−3.0
${\Delta G}_{\text{Int}}^2$ (kcal mol−1)d	−2.7	−9.6

^aChange in the parameter with introduction of the interface: parameter in the full-length c₄ minus parameter in the isolated c₄-A (or c₄-B) domain.

^b${\Delta G}_{\text{Int}}^2-{\Delta G}_{\text{Int}}^3$.

^c,dFree energy corresponding to the interface-introduced stabilization of ferric and ferrous domains, respectively.

All three ferrous variants displayed single-transition profiles (Figure 6B). The similar $m_{\mathrm{D}}$ values of ferrous c₄, c₄-A, and c₄-B exclude the scenario of c₄ unfolding as a single unit, because in such a case, the $m_{\mathrm{D}}$ value for the full-length protein should have been the sum of $m_{\mathrm{D}}$ values for the two isolated domains.⁵⁴ Therefore, the two domains in ferrous c₄ either unfold cooperatively or, if the difference in the stability parameters of c₄-A and c₄-B is small, independently. Because of the composite transition in c₄ and oligomerization of c₄-A and c₄-B, parameters in Table 2 for ferrous proteins are just apparent values.

Intrinsic Stability and Reduction Potentials of Isolated Domains.

A simple thermodynamic cycle (Figure S6A) connecting four states (ferric-folded (Fe^III─X_f, F), ferric-unfolded (Fe^III─X_u, U), ferrous-folded (Fe^II─X_f, F), ferrous-unfolded (Fe^II─X_u, U)) for a monoheme protein X has been used in the past to illustrate the relationship between unfolding free energies $({\Delta G}_{\mathrm{u}})$ and reduction potentials $(E_{\mathrm{m}})$ for monoheme cyt c proteins.⁵² The reduction potential of the unfolded cyt c could be considered to be similar to that of the heme-containing peptide N-acetyl microperoxidase-8 AcMP8 (or its ligand adducts) and depends on the nature of the distal ligand.^55-56 A similar formalism could be extended to c₄. Without specifying yet intermediate steps, unfolding of the full-length protein can be envisioned as a combination of breakup of interfacial contacts and unfolding of the folded c₄-A and c₄-B domains (Figure 6C).

Electronic absorption spectra were used to monitor changes in the heme environment with pH for ferric c₄, c₄-A, and c₄-B in the presence of 4 M GuHCl, the condition under which all three variants are unfolded (Figure S7 and Table S3). Misligation by nonnative His residues is well documented in unfolded cyt c.^49,57-60 The phenomenon is pH-dependent as at lower pH His gets protonated and no longer coordinates to the heme iron. Besides the two native His ligands (His18 and His123), there are two additional His residues (His142 and 190) in c₄ and both are found in the polypeptide region corresponding to c₄-B. Based on this information and the pK_a range from other unfolded heme iron misligates,⁶¹ we assign the pK_a of 4.8 for unfolded c₄-B to the equilibrium between His-misligated and water-coordinated ferric heme iron species. The pK_a of 6.6 for unfolded c₄-A is assigned to the equilibrium between N-terminus-ligated and water-coordinated ferric heme iron species.⁶² There are two pK_a values observed for the unfolded ferric c₄: 6.4 and 4.9. By analogy with isolated domains, we attribute them to the equilibrium between N-terminus-ligated and water-coordinated ferric heme iron species in the c₄-A domain and the equilibrium between His-misligated and water-coordinated ferric heme iron species in the c₄-B domain, respectively. Analyses of spectroscopic data in Figure S7 suggest that in the full-length unfolded c₄ at pH 5.0, c₄-A is water-ligated and c₄-B is a mixture of the water-ligated (40%) and His-misligated (60%) species.

Electronic absorption spectra of unfolded ferrous c₄, c₄-A, and c₄-B at pH 5.0 all exhibit features characteristic of a 5c heme iron, as in AcMP8.^63-64 The information on potentials of the heme iron in water- and His-ligated AcMP8 exists^64-65 and now that we know the ligation state of Hemes A and B in unfolded c₄ variants, we could apply the AcMP8 data to set estimates of potentials of these hemes at −140 mV and −150 mV, respectively.

Having potential estimates for the hemes in the unfolded proteins, we then extended the original thermodynamic cycle in Figure S6A to include oligomerization of folded proteins. The cycle in Figure S8A outlined in dash-orange led to the $E_{\mathrm{m}}$ value of 75 mV for the monomer c₄-B domain (Figure S8B and Table 1). The oligomerization of ferric c₄-A and a single unfolding transition in the ferrous full-length c₄ have prevented application of this thermodynamic cycle to c₄-A. The derived potential of 75 mV for the isolated monomeric c₄-B domain is higher that the value of −10 mV⁶⁴ for the Met-ligated AcMP8, suggesting at least partial encapsulation of the heme by the surrounding polypeptide. The potential of 75 mV for c₄-B, however, is about 175 mV lower than that of the molten-globular form of cyt c,⁶⁶ suggesting the higher extent of the heme exposure in the former case. We therefore refer to ferric c₄-B in its monomeric form as partially unfolded (puf).

Characterization and Stability of M66H and M167H c₄ Variants.

The formalism in Figure 6C includes folded species in the absence of interface. In order to examine their properties and circumvent complications of oligomerization of protein fragments, we decided to evaluate full-length variants with altered relative stabilities of the two domains. Two variants of the full-length WT c₄, M66H and M167H, with the Met residue corresponding to the axial ligand at Heme A or Heme B replaced by a His residue, were prepared. Studies of cyt c have established that the replacement of the Met ligand to His leads to a decrease in the reduction potential of the heme iron and alters the stability of the protein.^36,67-68

As with other variants, the oligomeric states, secondary structure, heme ligation, reduction potentials, and redox-dependent stability of the two Met-to-His variants were characterized. SEC profiles suggest that the two Met-to-His c₄ variants are monomers (Figure S9A). The α-helical contents of M66H and M167H are very similar to that of WT c₄ (Figure S9B). In the ¹H NMR spectra of ferrous variants (Figure S9C), signals corresponding to the methyl protons of the remaining Met ligand are at the same positions as that in WT c₄, suggesting that the heme of the unmutated domain maintains its native environment. Further, the reduction potential (Table 1 and Figure S5B) for the heme iron of the unmutated domain changes very little compared to the values in WT c₄ (by 10±5 and 9±7 mV for Heme A and Heme B, respectively). In contrast, the potential of the other heme, with the native Met ligand mutated to His, decreases dramatically. These results suggest that the Met-to-His mutations introduce minimal structural perturbations but change the ligation at one of the two hemes.

The unfolding profiles for the M66H and M167H in both ferric and ferrous states display two separate transitions (Figure 7). Studies of M80H cyt c have revealed that the stability of the His-ligated protein decreases only slightly in the ferric state, but drops dramatically in the ferrous state compared to the stability of the Met-ligated WT.³⁶ With this knowledge, we assign the first unfolding transition to be of the c₄-B domain in both ferric variants. Assuming then that the His-ligation does not change much the stability of ferric c₄-A and c₄-B, the unfolding order of the two domains in M66H and M167H is the same as that in ferric WT c₄. In the case of ferrous variants, the domain with the His-ligated heme is presumed to have a lower stability and the first unfolding transition is the one that corresponds to that domain. In ferrous M66H, c₄-A unfolds before c₄-B because ferrous c₄-A is destabilized by the M66H mutation. In ferrous M167H, c₄-B unfolds before c₄-A. The sequential unfolding equilibria for M66H and M167H are depicted in Figures S10 and S11, respectively.

Figure 7.

GuHCl denaturation curves at pH 5.0 for (A) ferric and (B) ferrous c₄ (black), M66H (magenta), and M167H (green) showing changes in CD ellipticity at 222 nm.

We have designed M66H to decrease the stability of c₄-A in our attempt to generate the folded c₄-B species in the absence of the interface. However, the following factors have worked against this strategy. Because ferric c₄-B domain unfolds before the c₄-A domain, the species Au^III-Bf^III is not populated during unfolding and the unfolding energy for the ferric folded c₄-B domain $({\Delta G}_{\text{Bf}}^3)$ cannot be obtained. Ferrous c₄-A is destabilized by introduction of the His ligand and unfolds before the Met-ligated c₄-B. But, because c₄-B partially unfolds upon the removal of the interdomain interface, A_u^II-B_puf^II rather than A_u^II-B_f^II, is populated during the unfolding and the experimental value of 5.5 kcal mol⁻¹ does not report on the unfolding energy of ferrous folded c₄-B $({\Delta G}_{\text{Bf}}^2)$. Interestingly, the ${\Delta G}_{\mathrm{u}}$ value of 5.5 kcal mol⁻¹ for the second unfolding transition in ferrous M66H (Figure S10) is close to the ${\Delta G}_{\mathrm{u}}$ value of 5.6 kcal mol⁻¹ for ferrous c₄-B from the analyses of redox-dependent stability and dimerization of the isolated c₄-B (Figure S8). This finding suggests that c₄-B has similar properties if the N-terminal polypeptide corresponding to c₄-A is unfolded or missing.

For M167H, as intended, thermodynamic analyses have revealed the properties of the folded “isolated” monomeric c₄-A. The four species involved in the second unfolding transition (dash-orange square: A_f^III-B_u^III, A_u^III-B_u^III, A_f^II-B_u^II, and A_u^II-B_u^II) were connected. Since the interdomain contacts no longer exist, it was assumed that the A_f-B_u species presents a combination of redox properties of the “isolated” monomeric c₄-A and unfolded c₄-B. Within this cycle, the $E_{\mathrm{m}}$ for unfolded c₄-B $(E_{\mathrm{m}},{\mathrm{B}}_{\mathrm{u}})$ cancels out, and the $E_{\mathrm{m}}$ for the folded monomeric c₄-A $(E_{\mathrm{m}},{\mathrm{A}}_{\mathrm{f}})$ is calculated to be 150 mV (Table 1).

MD Simulations.

MD models of c₄, c₄-A and c₄-B are shown in Figures 8A-C. Comparison of residue RMSF values suggests that the split of c₄ into two isolated domains greatly increase the flexibility of c₄-B but does not perturb much c₄-A (Figure S12A). Five regions (Figure 8D), in loops (Region 1 (19-28), Region 2 (52-60), Region 3 (90-97), Region 4 (125-136), and Region 5 (156-166)), experience notable changes when the two domains become isolated (Figure S13). All the regions, except for Region 2, become more flexible in the isolated domains. In the full-length c₄, Region 1 of c₄-A and Region 4 of c₄-B form contacts with Helix 2 of the other domain (residues 139-151 and 36-53, respectively), which add constraints to Regions 1 and 4. Interactions between Region 2 of c₄-A and Region 5 of c₄-B seem to decrease conformational fluctuations for Region 5 but increase them for Region 2. Region 3 in c₄ within the hinge loop is in close proximity to Region 1 in c₄-A and Helices 2 and 3 in c₄-B. Upon the split of the full-length protein, these interdomain interactions are no longer possible.

Figure 8.

MD structural models of (A) c₄-A, and (B) c₄-B isolated domains, and (C) full-length c₄. The crystal structure of Ps c₄ (PDB ID: 1M70)³³ is shown in tan for comparison. (D) Five regions (Regions 1-5), showing dramatic changes in the RMSF values in isolated domains compared to those in the full-length c₄, are highlighted in green (Regions 1 and 2 in c₄-A) and magenta (Regions 4 and 5 in c₄-B). Regions 3 corresponds to the hinge loop in the full-length protein and its residues 90-92 and 93-97 are colored in green and magenta, respectively.

Among the five regions, Region 4 and most of Region 3 (93-97), both in c₄-B, display particularly high RMSF values in the isolated c₄-B domain compared to those of the corresponding regions in the full-length c₄. Intradomain contacts maps (Figure S14) reveal that upon the removal of the neighboring domain some of the contacts between Helices 1 and 3 in c₄-B are lost or decrease in contact frequency, while the corresponding interhelical contacts in c₄-A are largely unperturbed. Our inspection of interactions between Helices 1 and 3 has identified a salt-bridge interaction that may be responsible for the difference of the two domains. In c₄-A, Lys10 of Helix 1 forms salt bridges with Asp77 and Asp80; in contrast, the interactions of the corresponding Lys111 in c₄-B only involve Asp178 (analogous to Asp77) (Figure S15). While Lys10 is positioned right in the middle of Helix 1, Lys111 is closer to the neighboring turn.

DISCUSSION

Structure of Diheme Ps c₄ and Comparison with Monoheme Proteins.

We first wondered about the principal differences of c₄-A and c₄-B in comparison to structurally similar Met/His-ligated c-type monoheme proteins (Table S4). All the identified monoheme proteins have three α–helices as a core for class I⁴⁴ cytochromes (Figure S16), well-aligned with the corresponding helices of c₄-A and c₄-B, and can be organized into two groups based on the protein oligomeric state in solution: homodimers (Group I) and monomers (Group II).

Crystal structures of proteins in Group I have interfaces between the two monomers closely mimicking the interdomain interface in c₄, through a two-fold axis centered at the middle of the hydrogen bond between the two inner HPs of the hemes (Figure S16A). For homodimers in Group I, the calculated SASA values of the heme group are 40-60 Ų and reduction potentials of the heme iron are 250-280 mV.^69-71 The SASA and reduction potential values of c₄-A and c₄-B domains in the intact full-length c₄ (Table S4) are similar. When, however, the second monoheme molecule for each structure in Group I is removed during analysis, the SASA values increase to 100-140 Ų, similarly to what happens in c₄-A and c₄-B in the absence of the other domain.

Monomeric proteins in Group II have lower structural similarity with c₄-A and c₄-B (Table S4). In particular, there are differences at the heme edge around the two HPs. The majority of these monomeric proteins have additional structural elements, such as an α–helix, β–sheet, or loop that surround the HPs region (Figure S16C). Their heme SASA values are about 50 Ų and reduction potential values are in the range 230-350 mV.^72-75 There is one particular structure within Group II, of Sinorhizobium meliloti (Sm) SorU (Figure S16B),⁷⁶ with a more exposed heme group (the heme SASA value is 147 Ų), similar to that in c₄-A and c₄-B, when the neighboring domain is not included. The reduction potential of Sm SorU is 110 mV,⁷⁷ consistent with the high solvent accessibility of the heme group. Taken together, these structural analyses suggest that, to reduce heme exposure and maintain high reduction potential, monoheme proteins arrange their polypeptides to enclose the heme edge. Otherwise, the proteins (or individual domains) tend to dimerize and form an interface centered at the HPs that encapsulates the heme groups.

The Interface Provides Structural Stabilization to the c₄-B Domain and Prevents Heme Misligation.

Our characterization of c₄-B has revealed perturbations in the secondary structure, heme exposure, and ligation to the heme iron in this domain in the absence of interactions with c₄-A, leading us to conclude that the isolated c₄-B is partially-unfolded. In the full-length protein, the more stable c₄-A likely guides folding of the neighboring c₄-B: interdomain contacts formed with c₄-A decrease the conformational flexibility of c₄-B.

Misligation is only observed in c₄-B and not in c₄-A, in accord with the sequence location of the His residues. While the N-terminal amino group could also serve as a ligand to the heme iron in unfolded proteins,⁶² it does not appear to do so in c₄-A or the full-length c₄ in the absence of a denaturant. These findings suggest that structural contacts in c₄-A prevent this coordination. We cannot fully exclude the possibility of the N-terminal amino group contributing to the formation of some of the misligated species in c₄-B but note that previous studies of unfolded proteins suggest that His residues, when available, are the main causes of misligation.⁴⁹

Interestingly, interactions with c₄-A not only increase the stability of c₄-B (Table 2) and decrease its conformational fluctuations (Figure S12) but also engage His142 in forming a hydrogen bond to the backbone of Ser24 at the interdomain interface. In addition, His190 is hydrogen-bonded to Gly188 in folded c₄-B. The His misligation is thus prevented in the full-length protein by structural constraints, which is important for functional integrity of this and other electron carriers under varying pH conditions.

The Interface Prevents Nonnative Dimerization of the c₄-A Domain.

The native intradomain contacts are minimally perturbed in c₄-A (Figure S14) upon separation of the two domains. Because there are no additional His residues in c₄-A, heme misligation is avoided and the heme iron is Met-ligated at both pH 5.0 and 7.4. The CD spectrum and denaturation curves suggest that the secondary structures are maintained, and the stability of c₄-A is unaltered upon removal of the neighboring c₄-B. We thus refer to the isolated c₄-A as folded (f).

While ferric c₄-B is a monomer at pH 5.0, ferric c₄-A is a dimer. Previous studies suggest that isolated domains of multidomain proteins may dimerize (or oligomerize) using the native-like interface, even without the enhancement of high local domain concentrations.⁷⁸ Based on the symmetrical nature of the domain interface in c₄, one would expect both c₄-A and c₄-B to dimerize in their ferric state, but it is not the observation here. Therefore, we considered if perhaps other, nonnative, interface sites are being used when ferric c₄-A dimerizes.

We have employed AlphaFold2-multimer Colab⁴⁰ to identify possible sites on the surface in c₄-A that could promote dimerization of this domain and be distinct from the native interface in the full-length c₄. Five models (Figure S17B) were generated and only one of them uses a part of the native interface for dimerization (Figure S17A). The nonnative sites in these models are not identical, but all encompass the area around the heme thioether next to the outer HP. This area is clearly apart from the native interface centered at the inner HP. The nonnative interface is primarily hydrophobic with apolar fraction of ~80%, which is higher than that for the native interface in the full-length c₄ (55%).²⁵ The total interface area for the c₄-A dimer models, ranging between 580-1460 Ų, is smaller than the value (1680 Ų)²⁵ in the full-length c₄. Dimerization through these nonnative interface sites does not exclude the simultaneous use of the native interface (Figure S17), therefore, under conditions favoring native-like contacts, it may be still possible to engage the native interface for oligomerization.

However, the intact full-length c₄ does not dimerize, suggesting that some distinct properties of the truncated protein favor the use of the nonnative interface. It is yet unknown how the interface may prevent nonnative dimerization, but, based on our MD simulations, we hypothesize that more flexible Region 2 in the full-length c₄ lowers the affinity for dimerization at nonnative sites. Testing this hypothesis and pinpointing the exact location of the nonnative interface will require experimental verification but a recent computational analysis highlights the importance of protein flexibility in modulating binding.⁷⁹

AphaFold2 was also employed to predict configurations of the c₄-B dimer (Figure S17B). The nonnative interface is again hydrophobic, with apolar fraction of ~75%, and shows a small interface area, ranging between 780-1340 Ų. As with c₄-A, all five models have nonnative dimerization sites involving the area around the heme thioether next to the outer HP, which mainly includes Region 4 (125-136) (Figure 8D). In contrast to Region 2 in c₄-A, Region 4 in c₄ becomes more dynamic upon separation of the two domains (Figure S12A). The enhanced flexibility and less compact structure of c₄-B may decrease the propensity of dimerization of this domain at nonnative sites.

Redox-Dependent Interface Stabilization.

Four (three positive and one negative) charged residues from both c₄-A and c₄-B are found at the interdomain interface of the intact full-length c₄ (Figure S1). The inventory of residue contacts (Table S5) suggests that, while charges of sidechains Lys42, Asp46, Arg61, Lys148, and Arg158 may be fully or partially balanced through electrostatic interactions with other residues or HPs, Lys31, Asp131, and Lys137 should be considered fully charged and thus their inclusion at the interface will have an unfavorable contribution from the side chain desolvation that may or may not be canceled by favorable interdomain interactions from these residues. In addition, the positive charges of Lys31 and Lys137 brought by the interface into the vicinity of Heme B and Heme A, respectively, are expected to be destabilizing for the ferric heme with the net charge of +1 compared to the ferrous heme with the net charge of zero.² These analyses have led us to hypothesize that the stabilization induced by the native interface may be minimal for ferric c₄-A and c₄-B but will increase for their ferrous counterparts.

Our estimates from analyses of ¹H NMR spectra of ferrous c₄-A suggest K_d<25 μM. If both the native-like and nonnative interfaces are employed in the ferrous c₄-A tetramer, then affinity of which interface does this K_d represent? The nonnative interface is mainly hydrophobic, but the native interface includes hydrogen-bonding and electrostatic interactions. One could then expect that the native interface will be affected by changes in charges at or near the interface, but the nonnative interface will not. Reduction of the heme iron results in the net charge of heme group changing from +1 to zero. The decrease in pH from 7.4 to 5.0 may lead to protonation of acidic residues, which would alleviate repulsion of charge at the interface. Our observations that both reduction of the heme iron and decrease in pH increase the propensity for dimerization of c₄-A dimers (forming the tetrameric species) and c₄-B monomers (forming the dimeric species) suggest that our K_d estimates (<25 μM) for ferrous c₄-A relate to the native interface.

Because of the symmetric nature of the native interface in c₄, one should expect similar values of K_d for c₄-A and c₄-B if the native interface is employed in their dimerization. Indeed, our K_d~3 μM for c₄-B from the thermodynamic cycle in Figure S8 is similar to K_d<25 μM from c₄-B from NMR experiments. Dimerization driven by the reduction of the heme iron does not take place in the intact full-length c₄, when the interface (which is by definition native) is formed by contacts between c₄-A and c₄-B, suggesting competition between hetero- and homodimerization for the same interaction site. The polypeptide linker favors the interactions of the two domains by bringing them in proximity to each other. But even without the linker, the two isolated domains form a heterodimer upon mixing, with properties similar to that of the intact c₄. It is logical to assume then that the native interface is involved in this complex as well. Importantly, the populations of homodimer and homotetramer decrease in the presence of the other domain. These multiple pieces of evidence suggest that the native-like interface, similar to the one in the full-length c₄, is employed for dimerization of the isolated domains upon reduction of the heme iron.

We have justified above that the stabilization induced by the native interface is minimal for ferric folded domains. Accordingly, we assume that for the folded c₄-A domain, the interface stabilization $({\Delta G}_{\text{Int}}^3)$ is equal to zero. The isolated c₄-B in solution, however, is partially unfolded and its interactions with c₄-A help with the buildup of intradomain contacts. From thermodynamic analyses of the M167H variant (Figure S11), the reduction potential for the folded c₄-A in the absence of the interdomain interface is 150 mV. Comparison of this value with that of c₄-A in WT (258±4) or M167H (268±3 mV) suggests that the interface stabilizes ferrous c₄-A by 2.7 kcal mol⁻¹ (120 mV) (Figure 9). Since c₄-A is already folded in the absence of the interface, the stabilization effects derive exclusively from interfacial contacts (Figure 6C). Because of the symmetric nature of the interdomain interface, we can assume that, if both isolated domains were already folded, the extent of stabilization from the interfacial contacts for c₄-A and c₄-B should be similar and we use 2.7 kcal mol⁻¹ determined from c₄-A for both domains. Then we can separate the ${\Delta G}_{\text{Int}}^2$ value of −9.6 kcal mol⁻¹ in Table 3 into stabilization effects of interface contacts (2.7 kcal mol⁻¹) and folding (6.9 kcal mol⁻¹) and derive the intrinsic reduction potential of isolated folded c₄-B (247 mV) (Figure 9).

Figure 9.

Energy diagrams for ferric (Fe^III, blue) and ferrous (Fe^II, orange) (A) c₄-A and (B) c₄-B domains illustrating stabilization effects of the interdomain interface. The free energy of ferric puf c₄-B was set to zero to serve as a reference for all other levels.

The above thermodynamic analyses help to understand the unfolding equilibria in WT c₄. The thermodynamic cycle of WT c₄ (Figure S18) involving the four species (A_f^III-Int-B_f^III, A_u^III-B_u^III, A_f^II-Int-B_f^II, and Au^II-B_u^II) leads to ${\Delta G}_{\text{Af}}^2+{\Delta G}_{\text{Bf}}^2$ of 23.4 kcal mol⁻¹. The value of ${\Delta G}_{\text{Af}}^2$ is calculated to be 11.1 kcal mol⁻¹ by following the orange-dash cycle, subsequently allowing an estimation on ${\Delta G}_{\text{Bf}}^2$ of 12.3 kcal mol⁻¹. The finding of similar unfolding free energies for ferrous c₄-A and c₄-B in the absence of the interface rationalizes our observation of a composite unfolding transition in ferrous WT c₄ (Table 2 and Figure 7). Further, the observation of a smaller experimental value (${\Delta G}_{\text{exp}}^2=7.2$ kcal mol⁻¹, Table 2) for ferrous c₄ than the values of ${\Delta G}_{\text{Af}}^2$ (11.1 kcal mol⁻¹) and ${\Delta G}_{\text{Bf}}^2$ (12.3 kcal mol⁻¹) supports cooperative, rather than independent unfolding of the two domains in ferrous c₄.

CONCLUSIONS

Our study of the full-length c₄ and its two isolated domains c₄-A and c₄-B has revealed that the interface affects properties of the constituent fragments in a number of ways. Monoheme proteins arrange their polypeptides to enclose the heme edge; if this is not possible, the proteins dimerize and form an interface centered at HPs. Both c₄-A and c₄-B dimerize in this way and ferrous c₄-A even forms a tetrameric species using both native and nonnative interface sites. In the full-length c₄, with c₄-A and c₄-B in the same polypeptide, the formation of oligomers is avoided. Further, the interdomain interface also promotes the formation of intradomain contacts and prevents misligation in c₄-B.

The interface in c₄ has distinct effects on properties of ferric and ferrous domains and selectively stabilizes ferrous species. The redox-dependent stabilization results in the 120 mV upshift of the intrinsic reduction potential of each of the folded domains. The redox-dependent stabilization from the intraprotein interface we observe here is similar to the redox-dependent interprotein interactions in unidirectional electron flow.⁸⁰ We have quantified such stabilization and anticipate that similar approaches could be broadly adapted for other redox metalloproteins to decipher mechanisms of redox regulation in both intra- and interprotein systems.

SYNOPSIS.

A series of cytochrome c₄ variants were characterized by spectroscopic techniques and stability measurements to probe effects of the interdomain interface on tuning structural and redox properties of multiheme proteins. The study illustrates how combining the two monoheme domains in a diheme protein modulates ligation and redox properties of the heme iron centers as well as folding and recognition of the domains, enriching understanding of multiredox assemblies and redox-dependent complexation.