The structure of the diheme cytochrome c₄ from Neisseria gonorrhoeae reveals multiple contributors to tuning reduction potentials

Abstract

Cytochrome c₄ (c₄) is a diheme protein implicated as an electron donor to cbb₃ oxidases in multiple pathogenic bacteria. Despite its prevalence, understanding of how specific structural features of c₄ optimize its function is lacking. The human pathogen Neisseria gonorrhoeae (Ng) thrives in low oxygen environments owing to the activity of its cbb₃ oxidase. Herein, we report characterization of Ng c₄. Spectroelectrochemistry experiments of the wild-type (WT) protein have shown that the two Met/His-ligated hemes differ in potentials by ~100 mV, and studies of the two His/His-ligated variants provided unambiguous assignment of heme A from the N-terminal domain of the protein as the high-potential heme. The crystal structure of the WT protein at 2.45 Å resolution has revealed that the two hemes differ in their solvent accessibility. In particular, interactions made by residues His57 and Ser59 in Loop1 near the axial ligand Met63 contribute to the tight enclosure of heme A, working together with the surface charge, to raise the reduction potential of the heme iron in this domain. The structure reveals a prominent positively-charged patch, which encompasses surfaces of both domains. In contrast to prior findings with c₄ from Pseudomonas stutzeri, the interdomain interface of Ng c₄ contributes minimally to the values of the heme iron potentials in the two domains. Analyses of the heme solvent accessibility, interface properties, and surface charges offer insights into the interplay of these structural elements in tuning redox properties of c₄ and other multiheme proteins.

Introduction

Neisseria gonorrhoeae (Ng) is a human pathogen of growing concern owing to the impact of Ng infections on human health [1, 2] and increasing spread of antibiotic-resistant Ng strains [2, 3]. The ability of Ng to efficiently transition between denitrification and aerobic respiration enables this pathogen to thrive in a variety of environments [4]. The sole cytochrome (cyt) c oxidase in Ng is a cbb₃-type cytochrome that catalyzes oxygen reduction and proton pumping at its active site under aerobic conditions [5, 6], therefore this enzyme and the proteins it interacts with may serve as powerful antibiotic targets. The diheme protein cytochrome c₄ (c₄) has been suggested as an electron donor to cbb₃ oxidase in multiple bacteria, acting alone or together with another heme protein, cytochrome c₅ (c₅), to transfer electrons from the bc₁ complex to cbb₃ oxidase [7–10]. In Neisseria gonorrhoeae (Ng), mutations in genes for c₄ and c₅ have revealed increased growth inhibition of the bacteria when placed in highly aerated environments [9]. A strain having deletions of both c₄ and c₅ was not possible to isolate under aerobic conditions, and this result has been interpreted to suggest that these two proteins are involved in electron transfer to cbb₃ oxidase [9].

Genomic analysis showed widespread presence of c₄ proteins within proteobacteria [8]. Solution properties of multiple c₄ proteins have been reported [8, 11–17], but at this point, structural information is only available for c₄ proteins from Pseudomonas stutzeri (Ps) [18], Pseudomonas aeruginosa (Pa) [19], and Acidithiobacillus ferrooxidans (Af) [20]. Ps c₄ is the most well-characterized and several of its variants have been examined [21].

In Ps c₄, the two hemes differ in their reduction potentials by ~100 mV, which is thought to be governed by the distinct surface charges of the two heme domains [18]. A similar relationship between potentials and surface charges exists for c₄ proteins from several other species [8, 12, 19]. For some c₄ proteins in which the two heme groups are equipotential, such as ones from Pa [19] and Pseudoalteromonas haloplanktis (Ph) [15], no distinct charge patches are present. At the same time, the two hemes in both Pa and Ph c₄ differ in their solvent accessible surface area (SASA) [19]. The interdomain interface has been suggested to also influence potentials of c₄ proteins [19, 20], but its effects are complex. Recent efforts in our group have uncovered that the interdomain interface in Ps c₄ plays a role in folding of the c₄-B (C-terminal) domain, prevents homodimerization of the c₄-A (N-terminal) domain, and increases potentials of both folded c₄-A and c₄-B by 120 mV by stabilizing the ferrous forms of the domains [21].

Herein, we report the solution properties and crystal structure of Ng c₄. The two hemes of Ng c₄ greatly differ in their solvent exposure. The structure reveals a prominent positively-charged patch, which encompasses surfaces of both c₄-A and c₄-B domains. We correlate these structural features with reduction potentials of the heme centers, determine the contribution of the interdomain interface to potentials of the isolated domains, and suggest implications for reactions of c₄ with its putative redox partners.

Materials and Methods

General.

Water was purified to a resistivity of 18.2 MΩ•cm using a Barnstead E-Pure Ultrapure Water Purification System. Reagent-grade chemicals were used to prepare all the solutions. Data fitting was done with MATLAB R2023a, and NMR data were processed with MestReNova 14.1.1 [22]. Sequence analysis was performed using SnapGene 7.0.1 [23]. Protein structures were visualized using UCSF Chimera 1.16 [24] or ChimeraX 1.6.1 [25].

Plasmid Construction and Protein Expression.

The gene encoding Ng c₄ (accession number AE004969) was synthesized by Genscript and subcloned into the pET-22b(+) vector using the NcoI and XhoI sites. The pelB signal sequence was included to target the protein to the periplasm. Site-directed Met-to-His mutations were performed using a QuikChange kit (Agilent). The prepared plasmids were sequenced at Dartmouth College Genomics and Molecular Biology Shared Resources.

Each of the constructed plasmids encoding c₄ was co-transformed with the plasmid pEC86 [26] into E. coli BL21 (DE3) cells (Agilent). The latter plasmid contains genes needed for maturation of c-type cytochromes [26]. The cells were plated on Luria Broth (LB)-agar plates containing 100 μg/mL ampicillin and 68 μg/mL chloramphenicol.

Colonies that appeared overnight were used to inoculate 6 mL Terrific Broth (TB) medium containing the same antibiotics, and the samples were left to shake at 220 rpm and 37 °C for 8 hours. 1 mL of this starter culture was transferred to 1 L TB media containing 100 μg/mL ampicillin and 68 μg/mL chloramphenicol. The large (1 L) cultures were grown for 18 hours at 220 rpm and 37 °C (WT and M166H) or 30 °C (M63H). For M63H, 100 μM isopropyl β-D-1-thiogalactopyranoside (IPTG) was added after 18 hours and cultures were left to shake for an additional 4 hours at 30 °C. Cells were harvested by centrifugation at 4,000 rpm for 25 minutes at 4 °C using a Beckman J6 centrifuge.

Protein Purification.

Harvest and osmotic shock of the cells were performed as previously described [19]. The supernatant from the osmotic shock was dialyzed overnight against a 10 mM sodium phosphate buffer at pH 6.8 (Buffer A) and subsequently applied onto a Fast Flow cation-exchange batch column (SP Sepharose resin, GE Healthcare) equilibrated with the same buffer. The crude product was eluted with Buffer A containing 50 mM NaCl (WT and M166H) or 100 mM NaCl (M63H), concentrated to 50 mL using a stirred cell concentrator (Amicon), and dialyzed overnight against Buffer A to remove salt. The dialyzed sample was purified on a HiTrap SP HP 5 mL cation-exchange column (GE Healthcare) equilibrated with Buffer A and eluted with a gradient of Buffer B (Buffer A containing 300 mM NaCl). The eluted fractions were then dialyzed against Buffer A, loaded onto a Mono S 1 mL cation-exchange column (GE Healthcare) equilibrated with Buffer A and eluted with a gradient of Buffer B. The purity of eluted fractions was verified by SDS-PAGE gels with Coomassie staining. Heme staining was used to assess the presence of the covalently attached heme groups in the prepared proteins. The oligomeric state of the proteins was evaluated by size exclusion chromatography (SEC) using a Superdex 200 Increase 10/300 GL column (Cytiva product 28990944). The column was equilibrated with a 10 mM Tris buffer at pH 7.5 containing 100 mM NaCl and 150 μL of ~10 μM proteins was loaded. The column was calibrated with the same standards as in our previous work [21].

Sequence Alignments.

The Basic Local Alignment Search Tool (BLAST) using the Ng c₄ protein sequence as the subject sequence was used to create alignments with proteins from other Neisseria species. The top 100 hits on a search of the standard database were manually evaluated to filter duplicates. In addition, the Ng c₄ protein sequence was used as a template to identify homologs in other bacteria with BLAST; Neisseria bacteria were excluded from the search. The obtained 1,000 hits were filtered by excluding those with the number of residues of >270 or <150. Then the top 500 sequences were used in multiple sequence alignments by MAFFT [27]. Sequence logo images were created based on the MAFFT output using online WebLogo [28] tool (https://weblogo.berkeley.edu/logo.cgi. The protein sequences of c₅ from Ng (GenBank: EFF40252.1) and Pa (GenBank: AAG08685.1) were used as templates in BLAST to find homologs of diheme and monoheme c₅ respectively.

Spectroscopic Measurements.

Electronic absorption spectra were collected using an Agilent 8453 diode-array or Shimadzu UV-1201 spectrophotometers. Protein concentrations were quantified using the absorbance of the Soret band with extinction coefficients determined in hemochrome assays [29]. Ferric (Fe(III)) and ferrous (Fe(II)) forms of the proteins were obtained by oxidation with potassium ferricyanide (K₃Fe(CN)₆, Acros Organics) and reduction with sodium dithionite (Na₂S₂O₄, Fisher Chemicals), respectively. The excess of redox reagents was removed using a PD-10 column (GE Healthcare).

Circular dichroism (CD) spectra were collected on a JASCO-J815 CD spectrometer in the far-UV (200-250) region; a quartz cuvette with path-length of 1 mm was used in these measurements. Spectra were recorded at a scanning speed of 100 nm/min, and five spectra were averaged. Experiments were done with ferric proteins; samples were prepared at concentrations of 15 μM in a 50 mM sodium phosphate buffer at pH 6.8 by treatment with 10 mM K₃Fe(CN)₆, and excess of the oxidizing reagent was removed using a PD-10 column (GE Healthcare).

¹H NMR spectra were collected on a 500 MHz Bruker NMR spectrometer. Samples of ferrous variants at concentrations of 0.5 mM were prepared in a degassed 50 mM sodium phosphate buffer at pH 6.8; other sample preparation details were as in our previous work [19].

NMR Redox Titration.

¹H NMR spectra were used to evaluate changes during redox titration of WT Ng c₄. A ferrous protein sample was prepared in a glovebox (Coy Laboratory Products) under a nitrogen atmosphere by reduction with 10 mM Na₂S₂O₄ (Fisher Chemicals), and the excess of the reductant was removed using a PD-10 column (GE Healthcare) equilibrated with a 10 mM sodium phosphate buffer at pH 6.8. Before the measurements, 1 mM sodium trimethylsilylpropanesulfonate (DSS) (Sigma-Aldrich) and 10% D₂O were introduced. The oxidant K₃Fe(CN)₆ was gradually added to the protein sample and changes in the intensity of the ¹H signals from methyl protons of the Met ligands were monitored at each addition.

Measurements of Reduction Potentials.

Spectroelectrochemical titrations of Ng c₄ variants were performed in both oxidative and reductive directions, as previously described [19]. The potentials were varied by a WaveNow USB potentiostat and the electronic absorption spectra in the range from 500 to 600 nm were collected. The protein sample was allowed to equilibrate for at least 15 minutes; the necessary timing was determined by monitoring the current with the Aftermath software (Pine Research Instrumentation).

The percentage of ferrous hemes (% Fe(II) hemes) was calculated from absorbance signals at 550 nm using Eq. 1, where A_550,sample is the absorbance of the baseline-corrected protein sample at a given potential, A_{550,Fe(II)Fe(II)} is the absorbance of the baseline-corrected Fe(II)Fe(II) sample, and A_{550,Fe(III)Fe(III)} is the absorbance of the baseline-corrected Fe(III)Fe(III) sample. Baseline correction was performed as previously described [30]. The dependencies of the % Fe(II) hemes on the applied potential E were fit to Eq. 2 to determine reduction potentials E_m1 and E_m2 of the two hemes.

$$\%\mathrm{Fe}(\mathrm{II})\mathrm{hemes}=\frac{A_{550,\mathrm{sample}}-A_{550,\mathit{Fe}(\mathrm{III})\mathrm{Fe}(\mathrm{III})}}{A_{550,\mathit{Fe}(\mathrm{II})\mathrm{Fe}(\mathrm{II})}-A_{550,\mathrm{Fe}(\mathrm{III})\mathrm{Fe}(\mathrm{III})}}$$ (1)

$$\%\mathrm{Fe}(\mathrm{II})\mathrm{hemes}=\frac{a_1}{1+\exp \left[\frac{(E-E_{m1})\times n_1\times F}{\mathrm{RT}}\right]}-\frac{a_2}{1+\exp [\frac{(E-E_{m2})\times n_2\times F}{\mathrm{RT}}]}$$ (2)

Crystallization and Structure Determination of Ng c₄.

Ferric Ng WT c₄ protein was exchanged into a 10 mM Tris buffer at pH 7.5 containing 100 mM NaCl by SEC, which confirmed that the protein exists as a monomer in solution at concentrations used for crystallization. Before crystallization, the electronic absorption spectrum of the protein was recorded, to confirm that the heme irons are in their ferric state. Crystallization conditions were initially assessed at 16 mg/mL Ng c₄ with a Wizard JCSG+ (Rigaku) screen of 0.2 μl protein + 0.2 μl well solution and 0.2 μl protein + 0.1 μl well solution drops in a 96-well tray using sitting drop vapor diffusion. Large, ~300 μm, crystals appeared after about 100 days under the condition of 1.6 M MgSO₄ and 0.1 M MES at pH 6.5. Crystallization conditions were then further optimized in a 24-well tray using hanging drop vapor diffusion. Crystals for data collection were produced from 21 mg/mL ferric Ng c₄ in drops containing 1 μl protein + 1 μl well solution (2.2 M MgSO₄ and 0.1 M MES at pH 6.5). Prior to flash freezing in liquid nitrogen, crystals were cryoprotected with 20% glycerol.

Diffraction data were collected at a wavelength of 0.92 Å at the beamline at the National Synchrotron Light Source II (NSLS II), Brookhaven National Laboratory. The data were indexed and integrated using XDS and scaled using aimless (CCP4) [31]. Phasing was accomplished by molecular replacement using Phaser-MR in PHENIX [32, 33], with the structure of Pa c₄ (PDB ID: 6Q2U) [19] as a search model. The initial model was constructed using AutoBuild in PHENIX, followed by iterative cycles of manual building in COOT and refinement in PHENIX. [34, 35] Four distance constraints, between the heme iron and the S atom of the Met ligand in each heme of the two protein chains in the asymmetric unit (ASU), were generated using eLBOW in PHENIX [36] and included in the refinements. The final structure was deposited in the Protein Data Bank under ID 8UF3.

Molecular Dynamics (MD) Simulations.

Protein MD models were constructed based on the crystal structures of c₄ proteins from Ps (PDB: 1M70 [18], chain B) and Ng (PDB: 8UF3). NAMD 2.13 [37, 38] was employed for simulations, and structural analysis was performed by Visual Molecular Dynamics (VMD) 1.9.3 [39]. Two independent simulations for Ng c₄ were carried out using the previously described settings for Ps c₄ [21]. The analyses on the MD models, including the root mean-square fluctuations (RMSF) of residues and SASA of hemes, were performed as previously described [21].

Structural Analyses.

Structures of c₄ were compared and Root Mean Square Deviations (RMSD) were calculated in ChimeraX-1.6.1 [25]. The interdomain interfaces were analyzed in InterProSurf [40]. Hydrophobic and hydrogen-bonding contacts at the interface were generated using LigPlot+ [41]. Electrostatic potentials of the protein surfaces were calculated with PDB2PQR and APBS tools [42] and color-mapped on the protein structures in Chimera. Structures in the Protein Data Bank were searched to identify monoheme c-type cytochromes having a similar polypeptide fold with the two domains of Ps c₄ in the DALI server [43] using a 90% sequence identity threshold (PDB90) and manually examined as previously described [21]. Structural models of diprotein complexes of Ng c₄ with c₁, c₅ or the CcoP subunit of cbb₃ oxidase were generated using AlphaFold2-Colab [44] with the following sequences from GenBank: EFF40949.1 (c₁), EFF40252.1 (c₅), and EFF40211.1 (CcoP subunit of cbb₃ oxidase). The top five ranked models were used for analyses.

Results

Characterization of Ng c₄ Variants in Solution.

The SDS-PAGE gels of recombinant Ng WT, M63H, and M166H c₄ variants confirmed the purity of the prepared protein samples (Fig. S1A–B). The SEC profiles suggested that all three variants exist as monomers in solution (Fig. S1C). CD spectra of the variants (Fig. S2) have revealed signatures at 208 and 222 nm characteristic of an α-helical structure. The spectra and calculations of the α-helical content (Table S1) suggest no major changes in the protein secondary structure upon the Met-to-His mutations.

The λ_max of the Soret band of WT c₄ is at 410 nm and 416 in the ferric and ferrous proteins, respectively (Fig. 1A–B and Table S2). Total protein concentrations from the Bradford assay, coupled with the absorption values for the heme group from the hemochrome assay, have confirmed the presence of the two heme groups in the protein ([Heme]/[c₄]=1.9±0.1). While the position of the Soret band is the same in the ferric Ng c₄ and Ps c₄, there is a 1-nm difference for the band λ_max in the ferrous proteins [21]. The Soret bands of ferric Met-to-His variants are blue-shifted relative to that of WT (Fig. 1A and Table S2); such shifts are in accord with the change in the axial ligand to the heme iron from Met to His and are similar to the trend observed with Ps c₄ variants [21]. Two lower energy charge-transfer bands, at 620 nm and 698 nm, are apparent in the electronic absorption spectrum of ferric WT (Fig. 1C and Table S3). The spectra of the His-ligated variants have revealed that heme B is the one responsible for the 620 nm spectral feature, characteristic of the high-spin species [45].

Fig. 1.

Electronic absorption spectra of (A) ferric and (B) ferrous Ng c₄ variants at pH 6.8. (C) The charge-transfer band near 700 nm of ferric variants.

The ¹H NMR spectra (Fig. 2A) of the ferrous WT protein exhibit two well-defined upfield signals at −3.11 and −3.51 ppm, which we attribute to the ɛ-CH₃ protons of the axial Met ligands of hemes A and B. Similar signals in ¹H NMR spectra of ferrous Ps c₄ associated with the ɛ-CH₃ protons of the axial Met residues are found at −3.20 and −3.59 ppm [21]. As in Ps c₄, the signals from the remaining Met axial ligand in ¹H NMR spectra of ferrous His variants of Ng c₄ are at the same positions as those from two Met ligands in WT, suggesting that the heme environment of the domain with the native Met/His ligation of the heme iron is unperturbed by the Met-to-His mutation of the ligand to the heme iron in the neighboring domain.

Fig. 2.

(A) Upfield ¹H NMR spectra of ferrous Ng c₄ variants at pH 6.8. (B) Titration of ferrous WT Ng c₄ with potassium ferricyanide.

Reduction Potentials and Redox Cooperativity.

Spectroelectrochemistry titrations have required fits of the data to two apparent reduction potentials (Eq. 2), 314 ± 5 mV and 217 ± 8 mV, for the two hemes in WT (Table 1). The titration curves of M63H and M166H each show signatures of two differently ligated hemes, with reduction potentials more than 230 mV apart (Fig. 3). Values for the Met/His-ligated heme centers in M63H and M166H are the same as apparent values of Met/His-ligated hemes in WT (Table 1). These results allow for straightforward assignment of the higher- (314 ± 5 mV) and lower- (217 ± 8 mV) potential values as those corresponding to heme A and heme B, respectively.

Table 1.

Reduction potentials at pH 7.0 of heme iron centers in Ng and Ps c₄ proteins from spectroelectrochemistry

Species	Variant	Em (mV vs SHE)
		Heme A	Heme B
Ng	WT	314 ± 5	217 ± 8
	M63H	−50 ± 6	210 ± 19
	M166H	325 ± 6	−28 ± 7
Ps a	WT	258 ± 4	364 ± 6
	M66H	20 ± 3	355 ± 3
	M167H	268 ± 3	−32 ± 4

^aFrom ref. [21].

Fig. 3.

Spectroelectrochemical titrations of Ng c₄ variants at pH 7.0.

A redox titration followed by ¹H NMR has corroborated these assignments. Gradual additions of ferricyanide to ferrous WT have first led to the disappearance of the methyl signal from Met166 followed by that from Met63, suggesting that the heme iron ligated by Met166 is more readily oxidized and thus the potential of heme B is lower than that of heme A (Fig. 2B). Since the potentials of the Met/His-ligated heme A and heme B are the same in WT and the mutants, the redox state of one heme does not modify the potential of its neighbor, meaning there is no (or minimal) redox cooperativity between the heme centers and the measured apparent potentials correspond to intrinsic values.

Crystal Structure of Ng c₄.

The crystals diffracted to 2.45 Å in the space group I 2 3, with unit-cell parameters a = b = c =168.05 Å and α = β = γ = 90° (Table 2). Two protein molecules are present in each ASU (Fig. S3A). The densities for most of the residues 3-189 (Fig. S3B–C) are well defined, except for the sidechains of Glu7, Gln86, Lys95, Glu175, and Lys179; these five sidechains were not built in the final model. The electron densities for the two covalently-bound heme groups of Ng c₄ are readily apparent (Fig. S3D). The structure of Ng c₄ (Fig. 4) is comprised of ten helices, and in general, the polypeptide backbone exhibits high structural similarity with Ps c₄ (RMSD = 1.2 Å), despite sequence similarity of 40.0% (Fig. 5). The two heme domains c₄-A and c₄-B are related by a pseudo two-fold symmetry, with the symmetry axis crossing the O—O hydrogen bond (2.43 Å, Fig. S4 and Table S4) between the two inner heme propionates (HPs). The relative positions of the two hemes are similar with those in Ps c₄ (Fig. S4 and Table S4).

Table 2.

Data collection and refinement statistics for Ng c₄^a

Data collection
Space group	I 2 3
Cell dimensions
a, b, c (Å)	168.05, 168.05, 168.05
α, β, γ (°)	90, 90, 90
Resolution (Å)	48.51–2.45 (2.55–2.45)
Rmeas	0.312 (3.074)
Rpim	0.068 (0.654)
I/σ(I)	7.8 (1.3)
CC1/2	0.995 (0.536)
Completeness (%)	100.0 (100.0)
Redundancy	21.2 (22.0)
Refinement
Resolution (Å)	39.61–2.45
No. reflections: work (free)	29063 (2875)
Rwork / Rfree	0.2037/0.2429
No. atoms
Protein	2719
Heme	172
Sulfate	15
Water	52
B factors (Å2) (all)
Protein	51.76
Heme	42.32
Sulfate	87.53
Water	49.86
R.m.s. deviations
Bond lengths (Å)	0.009
Bond angles (°)	1.053

^aValues in parentheses are for the highest resolution shell.

Fig. 4.

Crystal structures of c₄ from Ng (blue PDB ID: 8UF3, this work) and Ps (gray, PDB ID: IM70 [18]).

Fig. 5.

Sequence alignments of Ng c₄ with c₄ proteins from Pa, Ps, and Af. The CXXCH motif and the axial Met ligand are highlighted in orange and blue for the N-terminal and C-terminal heme domains, respectively. The boundary for the two domains is indicated by a dash line (red). The sequence identity with Ng c₄ is given at the end of the sequences. The three loops (Loop1, Loop2, and Loop3) showing variations among these c₄ species are highlighted in yellow. The insert I₁ in Ps c₄ [19] is outlined in green.

There are notable differences in the structures of Ng and Ps c₄ proteins in three loops (Fig. 5): Loop1 (residues 52-59), Loop2 (residues 123-137), and Loop3 (residues 159-163). Loop1 is located right before the helix containing the Met63 ligand to the heme iron in c₄-A. The length of this loop varies in c₄ proteins from different species [19]; the number of residues is eight for Ng c₄ and fourteen for Ps c₄. The sequence of three residues Thr-His-Gly in Loop1 is not present in either of the three structurally characterized c₄ proteins but is found in c₄ from other Neisseria species (Fig. S5). Among twelve known c₄ sequences from different Neisseria bacteria, the length of Loop1 remains the same. Thr56 and Gly58 are fully conserved, ten out of twelve sequences have His as residue 57, and the other two have Thr (Fig. S5A). WebLogo image generated for 500 c₄ sequences from bacteria other than Neisseria, with sequence identities from 45% to 100%, show the same length of Loop1 as in Ng c₄ but significant variations in residues 56 and 57 (Fig. S5B).

The c₄-A and c₄-B domains of Ng c₄ share only 34% sequence identity but structurally are very similar (RMSD = 1.3 Å). When comparing different c₄ proteins in our previous study [19], we have defined the insert I₁ in c₄-A within Loop1 as a variable-length polypeptide fragment (Fig. 5). For consistency, we keep the same notation in this report. When comparing the c₄-A and c₄-B domains of Ng c₄, three other inserts I₂ (residues 105-112), I₃ (residues 58-60), and I₄ (residues 125-130) are apparent (Fig. S6). The inserts I₂ and I₃ are also present in Ps c₄ [19], but the insert I₄ is only present in Ng c₄ because there are five additional residues in its Loop2 (Fig. 5). The conformation of I₂ in Ng c₄ is similar to that of I₂ in Ps c₄, but the conformations of I₃ are different in the two proteins. Because I₃ does not involve contacts with the neighboring chains in ASU of either Ng or Ps c₄, the difference may relate to differences in conformational dynamics of this protein region in the two proteins. The insert I₃ in Ng c₄ is shorter than that in Ps c₄ (Fig. S6), and MD simulations (Fig. S7) suggest that the polypeptide fragment corresponding to I₃ is less flexible in Ng c₄ compared to that in Ps c₄.

The c₄-A and c₄-B domains of Ng c₄ differ in the number of charged residues (Table S5). The net charges are +4 and +1 at pH 7.0 for c₄-A and c₄-B, respectively. The surface of Ng c₄ is mainly positively-charged, but there is also a small negatively-charged patch consisting of residues 175-177 on the surface of the c₄-B domain (Fig. 6). The positively-charged surface spans both domains.

Fig. 6.

Representation of the electrostatic surface of Ng c₄. The electrostatic potentials are calculated using APBS in Chimera and the protein surface is colored by its electrostatic potential in units of kT/e from −5 (red) to +5 (blue).

Heme Environment.

Hemes of both c₄-A and c₄-B are Met/His-ligated. The Fe-Met (S6 atom) and Fe-His (Sδ2 atom) and Fe-His (Nε2 atom) distances are within the range of those in other monoheme and diheme c-type cytochromes [18–20, 46]. The SASA values for the two hemes, from analyses of the crystal structure as well as snapshots from MD simulations, suggest that in Ng c₄ heme A is more enclosed than heme B (Table S5 and Fig. S8). The enclosure of heme A (Fig. S9) is largely due to the sidechain of His57 forming a hydrogen bond with the sidechain of Gln152 (Fig. 7B and Fig. S9), while its backbone oxygen atom interacting with the backbone nitrogen of Ala60.

Fig. 7.

Hydrogen-bonding networks of the inner (A) and outer (B) HPs in Ng and Ps c4 proteins. The distance between the oxygen atom of the inner HP and a water molecule is shown. In both c₄-A and c₄-B, the oxygen atom of the outer HP is next to an Arg residue (55 and 161, respectively). In Ps c₄, the outer HPs form extensive interactions with residues (Gln and Lys) from the other domain. In Ng c₄, only the hydrogen bond between Gln152 (analog of Gln149 in Ps c₄) and HP of heme A is conserved. Two new intradomain hydrogen bonds exist (backbone N of Ser59 and heme A; sidechain O of Tyr156 and heme B) in Ng c₄. The conserved contacts between Ng and Ps c₄ are indicated by *.

The oxygen atoms of the inner HPs of the two hemes are within hydrogen-bonding distance. Each inner HP (Fig. 7A) interacts with a Tyr and an Arg; these interactions are conserved in multiple c₄ proteins [18–20]. In Ps c₄, each inner HP hydrogen-bonds to a water molecule with an O—O distance of 2.9-3.0 Å. In Ng c₄, the interaction of the inner HP of heme A with a water molecule is readily apparent (2.7 Å), but the electron density for the water molecule (3.2 Å) next to the inner HP of heme B is relatively weak and this water molecule was not included in the deposited structure.

The intradomain contacts of outer HPs with Arg55 and Arg161 in Ng c₄ are similar to the corresponding contacts in the Ps protein. A new intradomain contact of outer HP of heme A with the backbone nitrogen of Ser59 is present in Ng c₄ (Fig. 7B). This interaction between Ser59 and HP may displace the outer HP of heme A away from Gln152 in the Ng protein. We observe one hydrogen bond to this outer HP, in contrast to two hydrogen bonds between the outer HP of heme A and the analogous Gln149 in Ps c₄. The outer HP of heme B forms a hydrogen bond with the sidechain OH of Tyr156 and with Arg161. A similar set of interactions (Tyr50-Arg55-outer HP) is formed near heme A in c₄ from Af [20], but not in the proteins from Ps or Pa.

Interdomain Interface.

A total area of the interdomain interface in Ng c₄ is 1592±30 Ų, which is somewhat smaller than 1703±58 Ų in Ps c₄, but the areas obtained from MD models are similar (Table S5). The fraction of interface area that is apolar is greater in the Ng protein (Table S5). Apolar interactions at the interdomain interface of Ng c₄ include contacts found in other structurally characterized c₄ proteins as well as distinct ones (Fig. S10). His145 packs against Ala30 in Ng c₄ compared to His142 that packs against Pro25 and Lys31 in Ps c₄ and Arg34 in Ng c₄ (which corresponds to Lys31 in Ps c₄) packs against Gly143. Also, there are several additional interdomain contacts (His57—Ala155/Arg161, Pro87—Arg140, His39—Met128/Pro129) not found in Ps c₄.

The unique polar contacts in Ng c₄ are His39 (Nε2)—Ala126 (O) and His57 (Nε2)— Gln152 (Oε1) (Fig. 8 and Fig. S10). Because of the difference in sequence of the two proteins (Fig. 5), the interfacial contacts Arg38—Asp131 and Lys42/Asp46—Asn160 found in Ps c₄ are lost in Ng c₄. In comparison to Ps c₄ [19], there are fewer charged residues at the interdomain interface of Ng c₄. Contacts of the outer HPs to residues in the neighboring domain (Fig. 7B) contribute to the interactions at the interdomain interface. The outer HP of heme A in Ng c₄ forms a hydrogen bond to Gln152 of c₄-B; this contact is similar to the bifurcated hydrogen bond to Gln149 in Ps c₄. Another interdomain contact of the outer HP of heme A is with Lys148 in Ps c₄, but there is no contact with the corresponding Glu151 in Ng c₄. The outer HP of heme B interacts with Lys42 and Gln43 in Ps c₄, but not with the corresponding residues (His45 and Gln46) in the Ng protein.

Fig. 8.

Interacting residues at the interface for the two domains of Ng c₄ as determined by LigPlot+. Hydrogen bonds and hydrophobic interactions are depicted as dashed lines and eyelashes, respectively.

Correlations of Reduction Potentials with Protein Structure.

Structures available in the Protein Data Bank have been analyzed to identify Met/His-ligated monoheme C-type cytochromes having a similar polypeptide fold and similar SASA values of the heme group to those of the two domains of the previously characterized Ps c₄ [21]. The dependence of potentials on surface charge for the combined set of four monoheme proteins and two protein domains is described by a positive linear relationship with a slope of 14.5 mV per charge (Fig. S11). This slope value is in accord with previous estimates of surface charge contributions to heme iron potentials from both theoretical (12 mV) [45] and experimental (6-23 mV) studies [45, 47].

We postulated next that when the heme SASA of a heme protein is in a similar range, the above correlation could help to estimate reduction potentials of monoheme proteins as well as intrinsic potentials of individual domains of diheme proteins in the absence of the interface. The SASA value of heme B in Ng c₄ is similar to those of the proteins in Fig. S11. The interface contribution to its experimental potential of 217 ± 8 mV in the diheme Ng c₄ could be estimated by positioning this value on the correlation plot and accounting for the net charge of +1 of this domain. The data point for Ng c₄-B is very close to the correlation line developed for the set of proteins that do not have an interface or have the interface contribution subtracted (Fig. S11). This analysis suggests that the interface does not appear to alter much the intrinsic potential of heme B in the case of Ng c₄, in contrast with the increase by 120 mV in the case of Ps c₄ [21].

Because the interdomain interface in Ng c₄ is generally symmetrical (Fig. 8 and Fig. S10), both heme A and heme B are expected to be minimally affected by the interface formation in the Ng protein. Heme A of Ng c₄ has a smaller SASA compared to other hemes in the correlation in Fig. S11, thus applying the correlation in Fig. S11 is not suitable for this heme and required a different approach. Tezcan et al. explored the correlation between the reduction potential and the heme solvent exposure based on an analysis of eleven hemes from structurally characterized c-type cytochromes [48]. We have revisited this earlier reported dependence of the potential of heme iron on heme exposure, adding more protein entries and focusing exclusively on hemes from monoheme and diheme Met/His-ligated c-type cytochromes (Fig. S12A). Both hemes A and B of Ng c₄ follow the correlation.

We wondered next if accounting for the net charge contribution to the potential could strengthen the correlation between the potential and SASA of the heme edge. However, instead, the overall correlation (Fig. S12B) became weaker, although the negative relationship between the two parameters remained. These findings suggest that, while SASA and surface charge both contribute to tuning potentials, their effects may not be straightforward to separate when both properties change or there are other confounding effects. Nevertheless, one can conclude that the positive surface charge and lower heme exposure in c₄-A are factors that contribute to the higher reduction potential of heme A relative to that of heme B in Ng c₄.

Insights on Recognition of Redox Partners from AlphaFold2 models.

Given a prominent positively-charged surface in Ng c₄, we wanted to examine if this area may be employed for interactions with the previously suggested redox partners of c₄. Models for the diprotein complexes of c₄ with c₁, CcoP, and c₅, all from Ng, were generated with AlphaFold2 (Fig. S13–15). The Ng complex of c₄ with c₁ (Fig. S13) is similar to the corresponding Pa complex [7]. In the complexes from both bacteria, c₄-B is the domain that interacts with c₁ and the surfaces of the two proteins complement each other electrostatically. Apolar interactions between the two proteins appear to be of importance as well: apolar area represents 68% of the 2382 Ų total interface area.

The two heme domains CcoP-A and CcoP-B at the N-terminus of the triheme CcoP from Ng (CcoP-AB) are similar in protein sequence to those of the diheme CcoP from Ps and Pa [7, 49]. In the Ng complex of c₄ with CcoP-AB (Fig. S14), the c₄-B domain again is the one interacting with CcoP and residues in both CcoP-A and CcoP-B are involved in the complex formation. The interaction area in c₄-B is similar to that employed for interactions with c₁ but the interprotein interface is smaller (1543 Ų) and apolar contacts are more prominent (72% of the total area is apolar). The electrostatic complementarity of the two protein surfaces is evident in this complex as well.

For Ng c₅, a polypeptide fragment of 47 residues linking the two heme domains could not be defined by AlphaFold2. At the moment, structures of only monoheme c₅ proteins are known and just from two bacteria, Azotobacter vinelandii and Shewanella violacea [50, 51]. There are monoheme c₅ proteins from many other bacteria, including Ps and Pa [52]. In contrast, Ng, other Neisseria bacteria, as well as Klebsiella pneumoniae, Pseudogulbenkiania sp., Leeia aquatica have c₅ proteins that are diheme. The AlphaFold2-generated models of the Ng complex of c₄ with c₅ have yielded structures with either c₄-A or c₄-B domains or both the domains interacting with c₅ (Fig. S15). However, without concrete structural data on diheme c₅ proteins, the information content of these models is likely minimal.

Discussion

The Ng c₄ crystal structure is an important addition to the currently very limited set of structurally characterized c₄ proteins as it illustrates distinct structural features that govern redox behavior of these proteins. The particularly tight encapsulation of one of the hemes, the minimal contribution of the interface to the heme potentials, and the prominent positively-charged surface area encompassing both domains are the key differences that set Ng c₄ apart from the other c₄ proteins whose structures we know. Whether these features are a norm or an exception will depend on the continuing characterization of this protein family, which is critical for wholistic mapping of bacterial respiration networks.

The lower heme exposure of the c₄-A domain seems to be connected to the lid formed by three residues Thr-His-Gly, conserved in Neisseria species, but not in other bacteria. The SASA of heme A in Ng c₄ of 27.5±7.7 Ų is not only lower than that of heme B in this protein but is the lowest among heme SASAs of other c₄ proteins [19], comparable to that of iso-1 cytochrome c from Saccharomyces cerevisiae [53] and c₅ from Azotobacter vinelandii [51]. The tighter heme encapsulation is expected to raise the reduction potential of the iron center, which is critical for function of electron carriers at later parts of biological electron-transfer chains.

The considerable scatter in the correlation plots in Fig. S12 suggests that additional factors, besides the heme exposure and the protein surface charge, may contribute to tuning reduction potentials in the series. Charges from residues right next to the heme group can dramatically modulate reduction potentials of the heme iron [54, 55]. Well-known examples are a carboxylate hydrogen-bonding to the His ligand in peroxidases, which lowers the reduction potential [56, 57], as well as the buried Arg residue found in many cytochromes, which destabilizes the ferric heme iron and stabilizes the negative charge of the HPs [15]. Such an effect may be a factor influencing the stark deviation of C6 from Phaeodactylum tricornutuma in Fig. S12; Arg44 interacts with Asp37 and Glu84 in this protein, an interaction not seen in other c₆ proteins [58], which may help to delocalize the negative charges of these two carboxylates, increasing the potential of the heme to a greater degree than anticipated from SASA and surface charge predictions alone. Redox-dependent structural changes are also known to influence the electrostatic environment and hence potentials of heme proteins [45, 55, 59, 60]. Interestingly, the scatter in the correlation plot greatly increases after correcting for surface charge in the region of small SASAs, while the correction seems to strengthen the correlation in the region of larger SASAs. The proteins associated with the particularly high deviation from the correlation line in Fig. S12B have highly-charged surface patches near their heme groups, such as mitochondrial cytochromes or photosynthetic c₆ proteins [45, 46, 61]. Analyses of their potentials may require a comprehensive way of treating electrostatic contributions that accounts for charge screening within their patches and any residue-specific local charge or dipolar effects.

Studies of protein complexes provide precedents for tuning reduction potentials of metal centers by complex formation, with implications for regulation of electron transfer [20, 62]. In Ps c₄, the size of the interdomain interface and its effect in upshifting the heme iron potentials by 120 mV is comparable to that of known metalloprotein complexes. In contrast, despite a very similar size, the effect of the interface on potentials is minimal in Ng c₄. There are fewer charged residues at the interdomain interface of Ng c₄, which are four Arg residues at positions 34, 55, 140, and 161 (Fig. 8). Four positively-charged residues (Lys31, Arg61, Lys137, and Arg158) are found at the corresponding positions in Ps c₄, but, in addition, there are Arg38, Lys42, Asp46, Asp131, and Lys148 at the interface. The charges of Arg55 and Arg161 in Ng c₄ (or Arg61 and Arg158 in Ps c₄) may have been balanced by inner and outer HPs of the heme from the same domain. The inclusion to the interface of the charged Arg34 and Arg140 (or corresponding Lys31 and Lys137 as well as Arg38 in Ps c₄) is expected to destabilize the ferric forms of the domains by bringing these positively-charged groups next to the heme iron with the net charge of +1 [63, 64]. In contrast, the negatively-charged Asp131 at the interface of Ps c₄ should stabilize the ferric forms of the two domains in this protein [65, 66]. Because in Ps c₄ Lys42 forms an intradomain salt bridge with Asp46, the charges of these two residues may be complemented within c₄-A, so the effect of these two residues may be minimal upon the interface formation. However, formation of the interdomain interface in Ps c₄ brings two Lys residue (Lys42 in c₄-A and Lys148 in c₄-B) to contact the outer HPs of the neighboring domains (Fig. B), and, based on previous studies [64, 67–70], such contacts are likely to raise the potential of the heme iron [21]. Importantly, analogous Lys-HP contacts are not there in Ng c₄, because of the difference in the protein sequence (Fig. 5). While these arguments frame general expectations, the effects of charges on potentials are difficult to predict quantitatively as changes in solvation and contacts with other residues affect stability of proteins upon interface formation [71], which is relevant to both ferric and ferrous forms of c₄.

Formation of transient complexes is important for efficiency of electron transfer in redox chains [72, 73]. Long-range electrostatic interactions often guide the formation of an encounter complex, which subsequently rearranges to a reactive configuration. The advantage of balancing more persistent hydrophobic interactions with electrostatic complementarity allows for fast dissociation of such complexes [72]. Electrostatic interactions between surfaces of a donor and an acceptor could allow for recognition and binding of redox partners. In contrast to the bipolar distribution of surface charges in Ps c₄, the surface of Ng c₄ is predominantly positively-charged (Fig. 6). Because there are no large distinct negatively-charged patches in Ng c₄, it is unlikely that binding of distinct redox partners to this protein is differentiated simply by charge complementarity. Hydrophobic contacts and reduction potentials, different for heme A and heme B, could then be of importance in controlling electron transfer to and from this protein.

Redox partners of c₄ from Ng have not been explicitly identified, but based on the studies of other bacteria, the following possibilities could be considered. In the structure of the Pa supercomplex, c₄ has been found to interact with the c₁ subunit of complex III (cyt bc₁) and c₅ [7]. Studies of gene deletion strains and cbb₃ activities in Pa and Vc have suggested that c₄ is an electron donor to cbb₃ oxidase with its periplasmic fragment of CcoP subunit as the docking site of c₄ [8, 74] [52]. The AlphaFold2 models of c₄ complexes in Ng (Fig. S13–14) suggest that charge complementarity may play in recognition of c₄ with c₁ or CcoP, but apolar contacts seem to dominate at the interfaces of these two complexes. It is not clear, however, if the prominence of apolar contacts at these interfaces is a bias of AlphaFold2 toward structurally characterized complexes, many of which are persistent [75]. To properly define the nature of the redox complexes and their role in the Ng aerobic respiration, characterization of structural and binding properties of transient complexes is required.

Despite their ambiguity, the models in Fig. S13–14, stimulate an interesting discussion point. The c₄-B domain is the domain interacting with both c₁ and CcoP. The 217±8 mV potential of heme B in c₄ sets the respective upper and lower limits for potentials of these redox partners. The known reduction potentials of these proteins from other bacteria are consistent with these limits. The potentials of hemes in CcoP are 269 and 344 mV in Ps and Vc, respectively [76]. Bacterial bc₁ complexes have the c₁ subunit with the potential in the range from 160 to 335 mV [77]. If the lower-potential heme domain in Ng and also in Ps c₄ [19] is employed in electron transfer in the aerobic pathway, what does the higher-potential heme domain then do? Its role may be in regulation of the aerobic respiratory complex as an electron sink, to deal with an excess of reducing equivalents when concentrations of oxygen are low, or to interact with other protein components such as c₅ as suggested in the Pa supercomplex [7], or act as an electron carrier in denitrification. A role of c₄ in denitrification of Ng has been previously proposed [78]. Based on the structural and potentials data for Ng c₄ in this report, the protein components and interactions in this pathway are worth examining further.

Conclusions

The crystal structure of Ng c₄ has revealed a large positively-charged patch, encompassing surfaces of both heme domains, and a tight enclosure of heme A. The former is likely used in recognition of redox partners, including several that have been implicated in other bacteria, and the latter may control directionality of electron transfer. Analyses in this study illustrate the role of a number of contributors to tuning reduction potentials of hemes in multiheme proteins. The distinct structural and redox properties of c₄ from different bacteria point to versatile strategies for recognition of redox partners in respiration.

Abbreviations

Af

Acidithiobacillus ferrooxidans

ASU

Asymmetric unit

c ₄

Cytochrome c₄

c ₅

Cytochrome c₅

HP

Heme propionate

MD

Molecular Dynami

MES

2-(N-morpholino)ethanesulfonic acid

Ng

Neisseria gonorrhoeae

Pa

Pseudomonas aeruginosa

Ph

Pseudoalteromonas haloplanktis

Ps

Pseudomonas stutzeri

RMSD

Root Mean Square Deviation

RMSF

Root Mean Square Fluctuations

SASA

Solvent accessible surface area

SEC

Size exclusion chromatography

Tris

Tris(hydroxymethyl)aminomethane