Redox sensor properties of human cytoglobin allosterically regulate heme pocket reactivity

Abstract

Cytoglobin is a conserved hemoprotein ubiquitously expressed in mammalian tissues, which conducts electron transfer reactions with proposed signaling functions in nitric oxide (NO) and lipid metabolism. Cytoglobin has an E7 distal histidine (His81), which unlike related globins such as myoglobin and hemoglobin, is in equilibrium between a bound, hexacoordinate state and an unbound, pentacoordinate state. The His81 binding equilibrium appears to be allosterically modulated by the presence of an intramolecular disulfide between two cysteines (Cys38 and Cys83). The formation of this disulfide bridge regulates nitrite reductase activity and lipid binding. Herein, we attempt to clarify the effects of defined thiol oxidation states on small molecule binding of cytoglobin heme, using cyanide binding to probe the ferric state. Cyanide binding kinetics to wild-type cytoglobin reveal at least two kinetically distinct subpopulations, depending on thiol oxidation states. Experiments with covalent thiol modification by NEM, glutathione, and amino acid substitutions (C38S, C83S and H81A), indicate that subpopulations ranging from fully reduced thiols, single thiol oxidation, and intramolecular disulfide formation determine heme binding properties by modulating the histidine-heme affinity and ligand binding. The redox modulation of ligand binding is sensitive to physiological levels of hydrogen peroxide, with a functional midpoint redox potential for the native cytoglobin intramolecular disulfide bond of −189 ± 4 mV, a value within the boundaries of intracellular redox potentials. These results support the hypothesis that Cys38 and Cys83 on cytoglobin serve as sensitive redox sensors that modulate the cytoglobin distal heme pocket reactivity and ligand binding.

Graphical Abstract

Introduction

Vertebrate hemoproteins such as globins and cytochromes are highly conserved and play critical functions in oxygen transport, electron transfer reactions, nitric oxide (NO) metabolism, and monooxygenation of fatty acids, drugs, and steroids. The mammalian hemoproteins all contain a redox-active, iron-containing heme prosthetic group bound to the protein via a proximal ligand, often a histidine or a cysteine. Depending on the specific protein, the distal pocket may contain an amino acid that binds this crucial position of heme iron; when bound in such a manner, exogenous ligands cannot bind until the amino acid dissociates. For proteins like hemoglobin, its distal E7 histidine cannot bind the iron, which allows for small molecules binding, while simultaneously regulating such ligands’ binding affinities [1]. However, there are two classes of mammalian hemoprotein in which the iron is coordinatively saturated. The first class consists of cytochromes (excluding the P450 superfamily), such as cytochrome c, which function predominately as electron transfer proteins. Having a bound distal amino acid tunes the oxidation/reduction potential, but importantly also limits exogenous ligand binding and prevents ‘off-target’ chemistry.

The second class of coordinatively-saturated mammalian hemoprotein includes the relatively recently discovered hemoproteins neuroglobin (Ngb), cytoglobin (Cygb), and likely androglobin [2], though unlike cytochromes, these proteins fluctuate between hexa- and pentacoordinate. While Ngb and Cygb may share the same α-helical globin fold as hemoglobin and myoglobin, their distal histidine is in an equilibrium between bound and unbound conformations in both the ferric and ferrous (deoxy) forms [3–6]. The hexacoordinate globins appear to be more ancient than the pentacoordinate globins, with Cygb being the closest hexacoordinate relative of the pentacoordinate oxygen carrying globins [7]. In fact, two forms of Cygb are found in zebrafish: one hexacoordinate and akin to mammalian Cygb, and one that is mostly pentacoordinate, supporting the phylogenetic evidence [8]. However, the biological functions of Ngb and Cygb have not been fully elucidated. Ngb is predominately found in vertebrate neurons, while Cygb is ubiquitously expressed in diverse mammalian cells; however, they both have cellular levels considerably lower than that of myoglobin or hemoglobin, suggesting a limited role in oxygen storage [9,10]. Proposed functions of both Ngb and Cygb include reactive oxygen species (ROS) detoxification, signaling reactions, and NO metabolism (NO dioxygenation and nitrite reduction) [11–17]; Cygb specific functions may also include lipid peroxidation and the regulation of apoptotic responses via reactive oxygen and nitrogen species sensing during vascular injury and remodeling [18–20].

Tetrameric hemoglobin heme pockets are allosterically regulated via the protein’s quaternary structure. Though monomeric, Ngb and Cygb exhibit a form of heme pocket modification via surface cysteine residues, which may be thought of as a form of redox allostery [21]. Such regulation is akin to what is observed in ASK1 [22], FoxO paralogs [23,24], Keap1 [25], and numerous other proteins [26]. Ngb was first shown to be redox regulated through the formation of an intramolecular disulfide bond, which results in a more open pocket via movement of the E-helix containing the distal histidine, effectively shifting the equilibrium to favor a more open, pentacoordinate state [27]. Indeed, we have previously shown the utility of this equilibrium shift in Ngb using nitrite reduction, which requires reaction with an ‘open’ distal pocket found in pentacoordinate heme [13]. The presence of an oxidized, intramolecular disulfide bond in human Ngb doubles the rate constant for nitrite reduction activity [13]. Similarly, this effect has been demonstrated in Cygb in which the nitrite reductase activity of the intramolecular form is approximately 50-fold higher than the reduced form of Cygb [28]. In fact, this form of Cygb has the largest observed rate constant for a mammalian nitrite reductase, hinting at a crucial role for these cysteines in modulating ligand electron transfer reactions. Further study of the redox sensor properties of these cysteines in human Cygb is required.

While a crystal structure of Cygb with the intramolecular disulfide bond intact has not been determined, various computational models have been developed [18,29]. Using these models, the formation of the intramolecular disulfide bond triggers rearrangements of the secondary structure. In particular, in the disulfide bond structure, the E-helix moves away from the heme, facilitating His81 (E7) dissociation from the heme iron and hence altering the distal histidine binding equilibrium (K_His, Figure 1).

Figure 1. Structural effects of disulfide bond formation in Cytoglobin.

The crystal structure of wild-type Cygb without the intramolecular disulfide bond (top left) and a model for the structure with the intramolecular disulfide bond (top right) are shown [18]. The formation of the disulfide bond between Cys38 and Cys83 displaces the E helix containing the distal histidine away from the heme prosthetic group, thus greatly decreasing ${\text{K}}_{\text{His}}\left(k_{{\text{His}}_{\text{on}}}/k_{{\text{His}}_{\text{off}}}\right)$.

It is proposed here and by others that modulation of Cygb via intramolecular disulfide reduction and oxidation regulates heme pocket reactivity [6,28,30–32]. However, the assessment of such reactivity requires defined redox states, and the molecular mechanism for this effect has not been well characterized. A major limitation of prior studies is the nature of testing for redox sensitive cysteine oxidation states in reactions that involve ferrous Cygb such as nitrite reduction, oxygen affinity [33], and CO binding. These assessments create artifacts as both cysteine oxidation and ligand binding to a ferrous heme requires competing redox conditions: the oxidation of cysteine residues and the reduction of the ferric heme for maintained nitrite reduction, CO binding activity, and oxygen affinity. Put simply, experiments that oxidize the cysteines may also oxidize the ferrous heme to ferric heme, thus blocking ligand binding, as well as vice versa, thus precluding the study of the oxidized cysteines. Furthermore, reactions rates can be complicated based on the formation of both intramolecular and intermolecular/mixed disulfide bonds with other thiol containing species.

To overcome these limitations, we take advantage of the binding of cyanide to the stable ferric heme species as a tool for probing thiol-redox dependent regulation of exogenous ligand binding. This allows for a more precise study of the discrete oxidation states of the cysteine residues. In these studies, we establish the ferric ligand binding rates associated with reduced cysteines, intramolecular disulfide formation, and single/double C38S and C83S mutants. We further use this methodology to test the sensitivity of this redox modulation to physiological levels of hydrogen peroxide and in vivo redox potentials. These results support the hypothesis that Cys38 and Cys83 on cytoglobin serve as sensitive redox sensors that modulate cytoglobin heme pocket reactivity and ligand binding.

Materials and Methods

Reagents

All reagents were obtained from Sigma (St. Louis, MO) and used as received unless stated otherwise. Buffers were prepared with 0.5 mM EDTA where necessary to eliminate confounding effects of free metal ions on oxidation of thiols. All pH measurements were performed with a Fisherbrand Accumet pH meter equipped with a Hamilton MiniTrode. All experiments were performed in 100 mM phosphate buffer, pH 7.4, unless otherwise stated.

Expression and purification of recombinant globins

Molecular biology procedures were performed using standard techniques. Human wild-type Cygb (UniProtKB Q8WWM9, wtCygb) was expressed in E. coli cells carrying the pET28-HsaCygb plasmid as described [8,15]. Cygb mutations were incorporated in the wild-type plasmid using the Quikchange kit (Stratagene) with adequate primers for each mutation. Expression and purification of wild-type Cygb and Cygb mutants was carried out in E. coli SoluBL21 cells (Genlantis, San Diego, CA) carrying the wild-type or mutant Cygb plasmid as previously described [8,15].

Cyanide-binding kinetics to Cygb and mutants

Cygb (wild-type or mutants) was mixed with dissolved excess potassium ferricyanide (K₃[Fe(CN)₆]) in 100 mM phosphate buffer, pH 7.1 to oxidize any remaining ferrous protein with bound diatomic species (e.g., CO, NO, O₂). Unlike other diatomic species, the ferric globin preferentially binds cyanide. The ferric Cygb was then passed through a Sephadex G-25 gravity column loaded polyacrylamide beads (Bio-Rad, 10DG desalting column) and equilibrated with pH 7.1, 100 mM phosphate buffer to remove excess ferricyanide. The concentration is then determined (ε = 11 mM⁻¹cm⁻¹ at 530 nm for hexacoordinate ferric Cygb or 500 nm for pentacoordinate ferric H81A-type proteins) as reported previously [15] using UV-Vis spectroscopy and diluted to 10 μM in the pH 7.1 buffer. Single spectra data and slower, single phase kinetics were recorded either using a Cary 50 spectrophotometer (Agilent) or an Agilent HP8453 diode array spectrophotometer equipped with a Peltier thermostatting device (HP89090). All solutions were prepared using air-tight Hamilton and in a nitrogen atmosphere glovebox (Coy Laboratories, Grass Lake, MI) for anaerobic conditions.

Pseudo first order kinetics were performed using a thermostatted (37.0°C ± 0.2) SX-20 stopped-flow instrument fitted with a direct mount photodiode array (Applied Photophysics, Ltd.). A 100 mM stock solution of potassium cyanide (KCN, Sigma Aldrich) was prepared in a sealed container fresh daily in 100 mM phosphate buffer, pH 11.5, as to avoid loss as gaseous HCN (pK_a = 9.2). The KCN stock was diluted to 250 μM – 40 mM in 4 mL aliquots of 1 mM, pH 11.5 phosphate buffer. 50/50 mixing of this buffer in the stopped-flow instrument with the 100 mM, pH 7.1 phosphate buffer containing the Cygb results in a pH of 7.4 as confirmed after experiments. Note that all concentrations injected in the stopped flow are half the injected amount, accounting for equal mixing. Kinetic fittings were made using Prism GraphPad 8.3. The best fit was chosen (i.e., single, double, triple exponentials) according to the best residuals of each fit at three wavelengths: the Soret decrease (407–416 nm) the Soret increase (425 nm) and the Q-band increase (548 nm). Each fit was made out to approximately eight half-lives. Fits take the form of Equation 1 at a given wavelength with appropriate background spectra.

$$\Delta \text{Abs}=A_1\cdot e^{-\left(k_{obs}1\right)\cdot t}+\dots A_n\cdot e^{-\left(k_{obs}n\right)\cdot t}$$ Equation 1

where A₁ (the pre-exponential term) represents the relative amount of a given phase in multi-phase fittings, k_obs1 is the corresponding observed rate constant determined over time, t. Under such conditions, A₁ is converted into % of the fastest phase via Equation 2.

$$A_1=\Delta {\text{Abs}}_0\cdot \left(\%A_{\mathit{\text{fast}}}\cdot 0.01\right)$$ Equation 2

Plots of the generated rate constants (k_obs) were plotted against KCN concentration, and the slopes (second order rate constants) were determined. Each point of k_obs at the various KCN concentrations is an average of 3–10 experiments. All plots of k_obs vs concentration KCN exhibited near zero intercepts, indicating very limited cyanide off-reactions.

NEM treatment

wtCygb cysteines were redox inactivated using N-ethylmaleimide (NEM). Globins were first reduced using 20-fold excess dithiothreitol (DTT, Roche) on ice for 1 hour, followed by incubation with 100-fold excess NEM for 1 hour. The Cygb heme was then re-oxidized with 100-fold excess ferricyanide, and the protein passed through a Sephadex G25 column to isolate and the concentration determined.

GSH/GSSG reduction potential determination

As it was determined that ferric wtCygb binding to cyanide fits well to a double exponential (or single where appropriate) with a fast phase representing an intramolecular disulfide and the slow phase representing all other forms of Cygb at concentrations of KCN under 3 mM, a constant amount of KCN (750 μM) was used for these experiments. Kinetics were performed as described above on the stopped-flow but at pH 7.0 and in the presence of 2.5 mM total concentration of reduced glutathione (GSH) and oxidized glutathione (GSSG). Different ratios of GSH/GSSG were used to simulate changes in cellular redox environment (GSH:GSSG – 100:0, 50:50, 25:75, 12:88, 6:94, 2:98, 0.1:99.9, 0:100), and were added to Cygb for 30 minutes on ice before running the assay. A double exponential kinetic analysis was performed on the time course of each result; the determined observed rate constants from each measurement should be unaffected by changes in GSH/GSSG amounts. However, differences in the amount of the formed intramolecular disulfide (the fast-kinetic phase) were determined by the relative percentage of Afast to A_slow. Using the standard reduction potential of the glutathione couple of −240 mV, the ratios of glutathione were converted to reduction potential and plotted against the %A_fast. From the fit of the data to the Nernst equation, the functional midpoint potential representing formation of the intramolecular disulfide was determined. These experiments were repeated under anaerobic conditions as otherwise described above.

Probing the effect of hydrogen peroxide on Cygb cysteines

Cyanide binding kinetics under hydrogen peroxide (H₂O₂) conditions were completed in similar manner as the GSH/GSSG experiments, with the same target parameters: constant k_obs’s and changes in %A_fast. Under low oxygen conditions, atmosphere regulated using a glovebox, wtCygb cysteines were fully reduced with 20-fold excess of TCEP (tris(2-carboxyethyl)phosphine, Oakwood Chemical) on ice for 1 h, which is bulky enough to prevent reduction of the ferric heme (iron reduction was not observed), unlike with DTT (complete iron reduction was observed within 30 min). The excess TCEP was then removed by a G25 Sephadex column equilibrated with buffer. This TCEP-treated Cygb (4–10 μM) was then exposed to concentrations of H₂O₂ (30% solution, diluted to 0–100 μM) at 37 °C. After 3 minutes of exposure, stopped flow mixing experiments with 750 μM KCN were performed. All experiments were repeated under room air conditions (except for initial anaerobic TCEP reduction of cysteines).

Assessment of ferryl species formation

4–8 μM of TCEP-treated Cygb was also exposed to 50 or 100 μM peroxide under hypoxic or normoxic conditions and observed for formation of a ferryl (broad absorbance centered ~595 nm). These spectra were recorded every 6 seconds over 3 minutes to observe both ferryl formation and subsequent decay. In order to establish conclusively ferryl formation, 11 μM TCEP-treated Cygb was treated with 22 μM sodium sulfide (Na₂S, 2 equivalents) was added together in a cuvette in 500 μL 100 buffer under the different oxygen levels. Upon mixing, an immediate slight redshift of the Q-bands was observed, owing to HS⁻ binding. 50 μM H₂O₂ was then added (5 μL addition) to induce ‘sulfcytoglobin’ formation, as exhibited by increases in absorbance at 585 and 600 nm.

Determination of cysteine availability after reaction with H₂O₂

A stock solution of 4-DPS (3 mM) was prepared in phosphate buffer mixed with 10% ethanol to help dissolution. 4–5 μM of TCEP-treated Cygb was added to 1 total mL in a quartz cuvette under normoxic conditions and was exposed 4 μL (50 μM) 4,4’-dipyridyl disulfide (4-DPS) for 5 minutes. The absorbance of this read at 324 nm represents the total number of available reduced thiols [34]. The experiment was repeated, except this time exposed to 50 μM H₂O₂ for 3 minutes before addition of 4-DPS for 5 minutes, representing the available amount reduced thiols after reaction with peroxide. Control spectra of wtCygb, wtCygb + H₂O₂, 4-DPS alone, and 4-DPS + H₂O₂ were taken at the matching concentrations.

Results

Ferric wild-type Cygb cyanide binding kinetics

Cyanide has been widely studied as a ligand for the ferric form of heme proteins [35,36]. The use of cyanide binding to ferric Cygb avoids some of the challenges with heme oxidation in other reactions that have been probed such as CO binding or nitrite reduction, which require small molecule binding to ferrous heme. Reactions of cyanide with ferric Cygb allows for independent oxidation of thiols without competing effects on the heme.

The full spectral changes of a typical reaction of ferric wtCygb under pseudo first-order conditions with excess potassium cyanide at pH 7.4 and 37°C are shown in Figure 2a. Using stopped flow techniques, reactions with ferric wtCygb and cyanide are deemed complete within 10 seconds, as shown in the time course for three select wavelengths (Figure 2b). Fitting of the traces to a single exponential function shows significant deviations consistent with a complex process, whereas a fit to a double exponential function appears more accurate based on the fit residuals (Figure 2c). The second order rate constants were thus determined by fitting the data to a double-exponential equation for KCN concentrations ranging from 250 μM to 3 mM (Figure 2d). It should be noted that a triple-exponential fit did perform better for the wild-type protein, suggesting the presence of a third protein subpopulation (Table 1). However, there was no substantial improvement in the fit from the use of a three-exponential equation over a double-exponential equation in the rest of experimental conditions studied (with kinetics followed for 1 minute on the stopped flow). Longer observation using standard UV-Vis spectroscopy did show this slowest phase at these concentrations, but the fast phase was lost to the dead time (data not shown). A more detailed analysis and discussion of the possible significance of this third phase is included in the supplementary section.

Figure 2. Recombinant wild-type human cytoglobin pseudo-first order kinetics with potassium cyanide (KCN).

Example of spectral changes (a) and temporal changes (b) over the course of 10 seconds with 8.5 μM ferric Cygb and 750 μM KCN. Traces in b represent the fit to a double exponential equation. Residuals of single and double exponentials of data in b are shown in c (the residual for 414 nm is shown). The determined pseudo-first order rate constants vs KCN concentration (d) allow for the calculation of second order rate constants from the slopes (dotted lines). Gray squares in d represent the percentage of the fast phase at each KCN concentration. Combined experiments in d used either 4, 7, 8.5, or 10 μM Cygb, and all experiments were conducted in 100 mM phosphate buffer adjusted to pH 7.4 at 37 °C. Error bars represent standard deviations of four total experiments monitoring three separate wavelengths.

Table 1.

Second order rate constants of wild-type Cygb and select mutants reacting with cyanide at pH 7.4, 37°C, 100 mM phosphate buffer. Values are shown as average ± SD of the mean, with n ≥ 3. Values in parenthesis indicate the relative amplitude (%) of each phase.

Cygb	kfast (M−1s−1)	kmiddle (M−1s−1)	kslow (M−1s−1)
wild-type (double exp. fit)	5200 ± 400 (70)	600 ± 300 (30)	--
wild-type (three-exp. fit)	5800 ± 500 (55)	400 ± 200 (27)	30 ± 5 (18)
Cygb-NEM	--	--	22 ± 4 (100)
C38S/C83S	--	--	28 ± 3 (100)
wt + 5 mM GSH	--	--	21 ± 2 (100)
C38S	--	510 ± 70 (17)	23 ± 1 (83)
C83S	--	540 ± 50 (18)	24 ± 1 (82)
TCEP-treated Cygb + 1 eq 4-DPS	4000 ± 500 (75)	300 ± 100 (25)	--
H81A	28000 ± 1000 (100)	--	--
H81A/C38S/C83S	30500 ± 800 (100)	--	--
H81A-NEM	30000 ± 1000 (100)	--	--

Despite pseudo-first order conditions, the residuals from the single-exponential fit indicate an apparently more complex, biphasic reaction (Figure 2c, Table 1). The clean isosbestic points observed (Figure 2a), argue against sequential reactions. These data are consistent with two optically identical reactants resulting in two optically identical products at two different rates and suggest different heme reactivities. This difference can be due to Cygb subpopulations with different surface cysteine oxidation states, as previously described for nitrite reduction by Ngb and Cygb [13,28]. As the spectra are dominated by the strong absorptivity coefficients of the heme moiety, small differences in absorbance due to the surface cysteines do not affect isosbestic points, thus, the simultaneous presence of a subpopulation with intramolecular disulfides and a reduced subpopulation would be consistent with these observations. It is worth noting that our expressed and purified recombinant wtCygb results in about 80% of the fast phase dominating the cyanide binding kinetics, consistent with a predominant intramolecular disulfide species formed during protein purification as previously described [18]. Finally, the surface cysteines, Cys38 and Cys83, are the only cysteines present in human Cygb.

Effects of substitution of the distal histidine on cyanide binding rates

Computational models indicate that the intramolecular disulfide can apply a mechanical tension to the E-helix, weakening the distal histidine (His81) binding to the heme iron. This change results in a more “open” distal heme pocket and thus a smaller K_His value [30,37]. Distal histidine dissociation is required for ligand binding in the ferrous form [27]. To test the hypothesis of bound distal histidine His81 blocking ferric small molecule binding, which has been previously implied [31], we performed several experiments. First, we mutated the His81 to a considerably smaller and non-coordinating alanine (H81A Cygb). The ferric H81A absorbance spectrum is similar to ferric hemoglobin, which also has an available coordination site in the distal pocket (Figure 3a). Pseudo-first order kinetics were performed to generate a single second order rate constant, which is 6-fold larger than the fastest wild-type phase and about 1000-fold larger than the slowest phase (Figure 3c, Table 1). As there is no competing histidine-mediated pre-equilibrium (K_His) and presumably enough space in the distal pocket for cyanide to bind in the preferred 180° configuration, the binding rate is much faster (Figure 3b). Additionally, the spectral changes are well-fitted by a single exponential equation, suggesting that there are no Cygb subpopulations that modulate ligand binding rates. The absence of subpopulations was confirmed in additional experiments where the thiols were i) completely reduced by DTT and alkylated by reaction with NEM or ii) removed by mutation of the cysteines (H81A-NEM and H81A/C38S/C83S, respectively). The respective rate constants of all these H81A Cygbs are similar (blue lines, Figure 3c, Table 1). These data suggest that distal histidine to heme affinities (K_His) mediate the observed different binding affinities of cyanide with wtCygb.

Figure 3. Reaction of cyanide with recombinant H81A mutated Cygb with and without reactive cysteines.

Example of spectral changes (a) and temporal changes (b) over the course of 2 seconds with 6 μM H81A and 250 μM KCN. Traces in b indicate fitting to a single exponential. The determined pseudo-first order rate constants vs KCN concentration (c) yields the second order rate constant (blue) and shows a much faster and less complex rate than the wtCygb (green circles and black squares). Combined experiments in c used either 6 or 10 μM H81A (blue circles), 5 or 9 μM H81A-NEM (blue squares) where the two cysteines were first reduced with DTT followed by alkylation with NEM, or 5 or 7 μM H81A/C38S/C83S (blue diamonds). All experiments were conducted in 100 mM phosphate buffer adjusted to pH 7.4 at 37 °C. Error bars represent standard deviation of four total experiments monitoring three separate wavelengths.

Effect of the redox state of the Cys38 and Cys83 on cyanide binding rates

To evaluate the effect of the redox state of cysteines 38 and 83 on the reaction rates of ferric wtCygb with cyanide (Figure 2), the two cysteines were either reduced with DTT and alkylated with NEM (Cygb-NEM) or mutated to serine residues (C38S/C83S). Consistent with nitrite reduction in ferrous Cygb, we hypothesized that the faster phase is due to the intramolecular disulfide subpopulation of Cygb. We then anticipate that modified Cygb species that are unable to form this intramolecular disulfide would lack this fast phase. In agreement with this hypothesis, the reaction of cyanide with Cygb-NEM and C38S/C83S was considerably slower than with wild-type. Moreover, traces fit well to single exponential equations (Figure 4a). The plots of k_obs versus cyanide concentration for derivation of the second order rate constant of Cygb-NEM and C38S/C83S are shown in Figure 4b (and Table 1). The rate constants (22 and 28 M⁻¹s⁻¹ respectively) are an order of magnitude lower than either of the two rate constants derived from wtCygb (5200 and 600 M⁻¹s⁻¹, Figure 2d, Table 1). Although such phase is not clearly observed for the wild-type Cygb at times collected on our stopped flow instrument, a three-exponential fit of the wild-type experimental data at higher concentrations of KCN (≥4 mM) does reveal a slow phase consistent with the Cygb-NEM and C38S/C83S results (Table 1, Supplementary Figure 1) and each second order rate constant separated by roughly an order of magnitude. Reduction of the thiols using TCEP followed by oxidation with a single equivalent of 4,4’-dipyridyl disulfide (4-DPS) regenerates the wild type data yielding the fastest and ‘middle’ rates (Supplementary Figure 2). Our data suggest that the subpopulation corresponding to this ‘middle’ phase may arise from the oxidation status and interaction of either Cys38 or Cys83 alone (Supplementary Figures 3 & 4), though there may be other possible interpretations. A more detailed discussion is included in the supplementary material. For simplicity, we focus on the double exponential treatment of the data, which seems to adequately explain most of the observed phenomena. The data are consistent with a ‘redox allostery’ of single and double oxidation of the thiol groups Cys38 and Cys83 to a sulfenic acid and/or a disulfide respectively, modulating K_His, with these larger terms being eliminated by reduction and/or C38S/C83S mutations.

Figure 4. Reaction of Cygb species without redox sensitive cysteines with cyanide.

Reduction or removal of Cygb cysteines result in a single, slow phase upon reaction with KCN (a); the example shown is with 1 mM KCN with 4 μM wtCygb treated first with DTT then excess NEM. The determined second order rate constants (b) of the mutation of the two cysteines (C38/C83) to serines (purple squares), the reduced and NEM-protected wtCygb (yellow squares) are close in value. Combined experiments in b used either 4 or 7 μM Cygb, and all experiments were conducted in 100 mM phosphate buffer adjusted to pH 7.4 at 37 °C. Error bars represent standard deviation of three total experiments monitoring three separate wavelengths.

Redox potential for the formation of the Cygb intramolecular disulfide bond

The redox environment of a cell is related to local concentrations of GSH and GSSG [38]. In order for the Cygb cysteines to function as a redox sensor, it is essential that they are responsive to redox potential fluctuations within the biological redox window [39,40]. Given the distinct kinetic differences, using cyanide as a probe is ideal for exploring the effect of the cysteines on the distal heme pocket of Cygb under biologically relevant redox conditions. However, the effect of mixed disulfide formation on the ferric Cygb (i.e., glutathione bound Cygb) needed to be established first. Given that tension of the intramolecular disulfide is needed to trigger the shift in the E-helix resulting in the increased the rate of ligand binding, we work under the assumption that an oxidized mixed disulfide would not cause the necessary strain and thus would behave like a reduced protein or the C38S/C83S mutant (Scheme 1).

Scheme 1.

Various forms of Cygb used in the ferric cyanide binding kinetics. The reduced form of the protein (with and without NEM protection), the C38S/C83S mutant, and each form of glutathione mixed disulfide appear to be kinetically equivalent and represent slow cyanide binding. The intramolecular form (top) is kinetically distinct, representing a fast form of cyanide binding.

Addition of excess GSH (2.5 mM) to wtCygb (7 μM) results in complete conversion of all intramolecular disulfides to mixed glutathione disulfides while also leaving the reduced cysteine subpopulation intact as reduced thiols alone should be unreactive toward each other. The resulting single, slow phase is analogous to those observed for Cygb-NEM and C38S/C83S (Figure 5a), supporting the hypothesis that fully reduced or oxidized, but glutathionylated, C38 and C83 lower heme ligand binding affinity (Scheme 1).

Figure 5. Using cyanide as a proxy for ligand reactions with cytoglobin with variable intramolecular disulfide formation states due to redox environment shifting with reduced and oxidized glutathione.

The reaction of 7 μM wtCygb with cyanide in the presence of excess (2.5 mM) GSH yields observed rates similar to the C38S/C83S mutant, resulting in a single slow rate constant (a). The observed rates values for the fast phase of the wtCygb reaction is shown for comparison (green circles) along with the C38S/C83S mutant (purple squares). Temporal changes of 5 or 7 μM wtCygb and 750 μM KCN at different glutathione ratios are shown in b, right, with a total glutathione concentration of 2.5 mM. At 50% of each glutathione (sky blue), the data fit well to a single exponential. Each other GSSG/GSH ratio fit to a double exponential, with the faster term increasing its contribution to the two phases as the environment became more oxidizing (b, left). The relative percentage of the fast term (%A_fast, intramolecular disulfide) was plotted against the reduction potentials, derived from the glutathione ratios via the Nernst equation (c, purple). The observed pseudo-first order rate constants of the faster and slower phases remain constant, as expected (c, red circles and black squares, righthand axes). The functional midpoint potential for intramolecular disulfide formation is −189 ± 4 mV. Error bars represent standard deviation of three total experiments monitoring three separate wavelengths.

Since we can discern between the more closed and open distal pocket of the ferric Cygb using cyanide as depicted in Figure 5b, and glutathionylation results in a slow rate constant for cyanide binding commensurate with cysteine-free Cygb, we can determine the reduction potential for the disulfide bond with this technique. By modulating the GSH/GSSG ratio at a constant total concentration (2.5 mM) and keeping cyanide concentration constant, we fit the observed traces to double exponential equations to determine the relative amount of each subpopulation. If our hypothesis is correct, the reaction at different GSH/GSSG ratios will be fit to a double-exponential equation with constant observed rates but variable amplitudes for each phase. The intramolecular disulfide subpopulation is represented by the amount of fast phase and the reduced/mixed disulfide subpopulation is represented by the amount of slow phase present in the term. The more oxidizing the conditions, the greater the amount of fast phase as seen in Figure 5b. These ratios can be used to calculate the functional oxidation/reduction potential for the formation of the intramolecular disulfide bond (E’) using the standard reduction potential of the GSSG/2GSH couple (-240 mV) in the Nernst equation (Equation 3) [41].

$$E^{\prime }=E^{°\prime }-\frac{RT}{nF}\text{ln}\frac{{[\text{reductant}]}^x}{{[\text{oxidant}]}^y}$$ Equation 3

As the cyanide concentration is unchanged in each reaction with different GSH/GSSG ratios, the observed large and small rate constants should remain uniform as just the redox environment changes; these constants do remain reasonably close. These rate constants are plotted on the right axis of Figure 5c. Thus, changes to the overall rate are dictated by the relative amount of each phase of the reaction, not a change in rate constant, validating this method. It is important to note that these data are consistent under hypoxic and normoxic conditions. A plot of the percentage of the fast phase of the reaction versus the reduction potential of the samples calculated from the known GSSG/GSH ratio yields a functional midpoint redox potential for the wild-type Cygb intramolecular disulfide bond of −189 ± 4 mV (Figure 5c, left axis). This value is well within the boundaries of reported cellular redox potentials [38].

Effect of hydrogen peroxide on cyanide binding rates

Using a similar approach to the reduction potential experiments, we tested the effect of H₂O₂ on wtCygb where the thiols had been fully pre-reduced with TCEP under low oxygen conditions (herein referred as TCEP-treated Cygb). H₂O₂ can generate sulfenic acids with reduced, redox-sensitive cysteines; a vicinal thiol may then react forming an intramolecular disulfide (Scheme 2).

Scheme 2.

Generalized scheme depicting formation of transitory sulfenic acid leading to C38-C83 disulfide formation in the presence of the vicinal thiol. The initial sulfenic acid may form readily on either C38 or C83.

The study of the thiol state using nitrite reductase activity or CO binding is complicated by side reactions of the active ferrous form with H₂O₂ or oxygen. As cyanide binding reactions involve the ferric species, we can probe the effect of H₂O₂ and oxygen on the cysteines without such limitations. These experiments are otherwise analogous to the GSH/GSSG experiments above: only the redox conditions vary, and the cyanide concentration is held constant. Thus, the observed rate constants of the two phases should remain constant no matter the peroxide concentration, and only the relative amplitudes of the fast and slow phases are expected to change. Under low hypoxic (0.5 – 1% O₂) conditions, TCEP-treated Cygb (16 μM) with mild peroxide treatment (0–48 μM) results in a H₂O₂ concentration dependent increase of the fast phase, indicating an increase in intramolecular disulfide subpopulation (Figure 6a). This Cygb pre-treated with peroxide (up to 80 μM) reacts with cyanide to generate spectra indistinguishable to normal ferric wtCygb binding cyanide (Supplementary Figure 5 vs. Figure 2a). When these H₂O₂ experiments were conducted in the same manner but under aerated conditions, the observed formation of the intramolecular disulfide was surprisingly entirely blunted (Figure 6b). These data may suggest overoxidation of one or both cysteines, as generation of a sulfinic or sulfonic acid would preclude intramolecular bond formation. It is worth noting that the kinetics of the reaction of peroxide with the cysteine to yield the disulfide cannot be measured using this method, but these reactions were all measured after 3 minutes of incubation with H₂O₂ at 37°C. Either way, these results strongly suggest a major change in the distal pocket under mild oxidizing conditions (low hydrogen peroxide) with and without oxygen.

Figure 6. Effect of hydrogen peroxide on the intramolecular disulfide formation.

In the presence of low to no oxygen (a), the increase of the contribution of the faster reaction (% fast phase, blue circles), indicating intramolecular disulfide formation, is acute, maxing out 40 μM H₂O₂ (~2 equivalents to wtCygb), followed by a plateau at higher concentrations. The effect is abolished in the presence of room air (b, open black circles). The pseudo first order rate constants (a and b, red squares and green diamonds, right axes) associated with cyanide binding remain relatively unchanged and within error of k_obs determined at this KCN concentration in earlier wtCygb experiments. Peroxide incubation times are ~3 minutes at 37°C. Error bars represent standard deviation of three total experiments monitoring three separate wavelengths.

The question remains if the reduced cysteines are modified upon exposure to peroxide and air. According to our results, it seems likely that one or both of the cysteines under normoxic/hydrogen peroxide are unavailable, and are not oxidizing to an intramolecular disulfide. In order to discern a difference between the relative available number of reduced thiols under normoxic conditions with and without hydrogen peroxide, we used 4-DPS under excess conditions, which reacts with reduced thiols generating a weakly absorbing mixed disulfide and an equivalent of 4-mercaptopyridine, which has a distinct absorbance at 324 nm [34]. Using TCEP-treated Cygb, 50 μM of H₂O₂ was added for 3 minutes or not added in the control, and followed by a 5-minute exposure to 50 μM 4-DPS. From these spectra, the spectra of wtCygb alone and of 4-DPS alone were subtracted to generate difference spectra (Supplementary Figure 6). Peroxide treatment resulted in roughly half the absorbance at 324 nm as wtCygb untreated with peroxide, suggesting that roughly half of the thiols are unavailable for intramolecular disulfide formation. It is worth noting that addition of H₂O₂ to 4-DPS over 10 minutes did not result in spectral changes from the 4-DPS alone spectrum. Thus, the changes observed are indeed due to Cygb thiol availability and not due to side reactions of 4-DPS with hydrogen peroxide.

Given the use of a hemoprotein and H₂O₂, the potential role of ferryl formation was considered. WtCygb can form a ferryl species with a distinct spectrum, though this was previously observed at much higher concentrations of peroxide (5 mM) than those used here [37]. Under aerobic conditions and the higher concentrations of peroxide used for the experiments above (50 – 100 μM), this ferryl spectrum is visible and maximized within 20 seconds, though decays again after about 40 seconds of peroxide exposure (Supplementary Figure 7a). However, the ferryl is not observed under hypoxic conditions in the same time frame (Supplementary Figure 7b) and even out to 3 minutes. To confirm if the ferryl is formed at all under these conditions or is simply being consumed before it can be spectroscopically observed, two equivalents of sodium sulfide (Na₂S) was added before addition of H₂O₂. The ferric spectrum for ‘sulfcytoglobin’ is clear under aerobic conditions, and some sulfcytoglobin also appears to form under low oxygen conditions, though in considerably less amount (Supplementary Figure 7c and 7d). Such a difference suggests a difference in steady state ferryl formation between normoxic and hypoxic conditions.

Discussion

Nitrite reduction and CO binding have been previously used to assess the open probability of the distal heme pocket in both Ngb and Cygb. We have previously reported a single second order rate constant for human wtCygb nitrite reduction (0.4 M⁻¹s⁻¹) [8], whereas Reeder and Ukeri have determined different rates for the protein when the cysteines are reduced (0.63 M⁻¹s⁻¹) versus when an intramolecular disulfide bond is present (32.3 M⁻¹s⁻¹) [28]. Our initial result is consistent with a Cygb with fully reduced thiols. No faster subpopulation was observed as these experiments were performed using strong reducing agents to maintain a ferrous heme species. The reducing conditions necessary to maintain a reduced heme complicate the study of the allosteric effects of cysteine oxidation on heme pocket ligand binding. A quantitative assessment of the effects of oxidation of single thiols and intramolecular disulfide formation on heme ligand binding under more physiological, redox signaling conditions has not been performed. As Cygb can maintain distal histidine ligation in both the ferrous and ferric states [42], we have used cyanide binding to the ferric heme to explore the unique effect of redox sensitive cysteines on the ligand accessibility to the distal pocket of human Cygb. In this work we have (1) determined the kinetics of cyanide binding to wtCygb and key mutants, as well as generated critical second order rate constants under biologically relevant conditions; (2) used the observed rates of cyanide binding to Cygb to determine the approximate functional midpoint reduction potential of formation of the intramolecular disulfide; and (3) used this same methodology to explore the dynamic nature of these thiols under the effect of hydrogen peroxide under aerobic and hypoxic conditions.

In studies evaluating cyanide binding to ferric Cygb, a triple-exponential equation did provide the somewhat better fit of the experimental data of cyanide binding to wtCygb as seen in Table 1, especially at high concentrations of cyanide (Supplementary Figure 1), with each second order rate constant separated by an order of magnitude. Our results agree with previous work for Cygb and Ngb, which has found that for the ferrous globin the intramolecular disulfide triggers a more open pocket and thus faster-reacting phase, while all other forms (reduced cysteines and dimers) have a more closed pocket [13,28,32] and that ligand affinity hinges on the presence of the distal histidine (Figure 3). Though not critical to the results determined under more dynamic redox conditions throughout this work, we have also shown that three different binding rates exist in the ferric state, seemingly corresponding to three different pre-equilibria states of the distal histidine. While this has been indicated previously [31], little has been done to reconcile this apparent ‘middle’ phase. From our single cysteine mutant experiments, the observed phase appears to be triggered by a singular Cygb surface thiol through oxidation (Supplementary Figure 4). However, this oxidation is not necessarily to a disulfide given the relevant concentration range used (Supplementary Figure 3) [6], and mixed disulfide formation leads to only slow phase (Figure 5a); therefore, it may possibly be a sulfenic acid. This phase may not be observable in the ferrous state with nitrite reduction due to the relative magnitude of the second order rate constants; there is not as much separation between the fastest and slowest observed nitrite reduction rate constants and thus a middle constant may be obscured if too close in value to one or the other.

Since we are able to monitor the effects on the distal pocket using the ferric form of the heme, we avoid use of strong reductants such as dithionite to accurately determined the functional midpoint potential of intramolecular disulfide formation using cyanide binding kinetics. We approximate this potential to be −189 ± 4 mV under these conditions using the Nernst equation and the GSH/GSSG standard reduction potential (Figure 5c). This value is well within established cellular redox potentials [38], as nearly all cells and tissues show potentials in the range of −260 to −150 mV [40]. Moreover, this value for human Cygb is close to the value found for the human Ngb disulfide bond of −194 ± 3 mV [13]. Generally speaking, as more rapidly proliferating cells are more reducing by up to 60 mV more than differentiated cells, and apoptotic cells are more oxidizing up to 60 mV more than growth arrested cells, these data indicate the Cygb distal pocket would be relatively closed (hexacoordinate, reduced cysteines) when under a proliferative state and more open (pentacoordinate, intramolecular disulfide) in confluent and differentiating cells as well as pre-apoptotic/apoptotic cells. As others have noted, overexpression of Cygb results in inhibition of cellular proliferation, specifically halting G1 to S-phase progression [43,44].

Cygb has been shown to be induced under hypoxic and oxidative stress conditions [45]. HIF-1-α binding sites are found in the Cygb promoter region [46]. Further, Cygb appears to be a critical protein responsive to oxidative stress in promoting viability in muscle progenitor cells, though the mechanism is unknown [47]. Cygb is also suggested to be a tight regulator of programmed cell death in vascular smooth muscle cells via protection from NO-dependent toxicity [19,20]. As such protection is mediated through NO dioxygenation [48], a more open pocket from an intramolecular disulfide would enhance oxygen binding and subsequent nitrate generation. Thus, oxidative stress triggered by pre-apoptotic conditions would ‘switch on’ the efficiency of NO dioxygenation given the midpoint potential of intramolecular disulfide formation. It is worth noting that lipid metabolism may be responsible for some or all of these listed effects – and is affected by intramolecular disulfide formation [18] – though not enough is yet known.

We further used the cyanide/ferric Cygb technique to assess the sensitivity of the Cygb cysteines to hydrogen peroxide, as some cysteines readily generate a transient sulfenic acid which in turn reacts with a vicinal thiol to generate a disulfide. The results in Figure 6 indicate that Cygb is highly peroxide sensitive, but only under conditions of little to no oxygen. Given that the maximum is reached at around 40 μM H₂O₂ or about ~2 equivalents of Cygb in the experiment at about 1% oxygen, Cygb appears to be a physiologically sensitive to peroxide under hypoxic conditions. Since Cygb is found in low micromolar concentrations in cells, these data indicate that low micromolar fluxes of H₂O₂ stimulate formation of the intramolecular disulfide and opening of the pocket.

Assuming such reactivity likely occurs at the cysteine closest to the heme, the sensitivity of the Cys83 apparent rapid generation of a sulfenic acid could be facilitated via a transient peroxide-bound ferric heme (compound 0 type complex). Lewis acidic metals do catalyze sulfenic acid formation as shown in proteins like Keap1, yielding an intramolecular disulfide or mixed disulfide [49]. Further, cysteines with adjacent basic amino acids also have lower pK_as, resulting in kinetically favored formation of sulfenic acids [50,51]; the Cys83-Arg84 combination is conserved across mammalian Cygbs [52]. An oxidation by a compound I type ferryl species cannot be ruled out in spite of its low detection under hypoxic conditions (Supplementary Figure 7), as this ferryl may be kinetically depleted and not reach an observable steady state. These postulations will be explored further elsewhere.

The blunting of fast phase signal with the cyanide probe under normoxic conditions of peroxide treatment is more puzzling. As the ferric heme does not have significant oxygen affinity, the addition of oxygen to a ferric hemoprotein is unlikely to have an effect on the distal pocket directly; i.e., it does not affect cyanide binding, nor should it affect peroxide in the distal pocket and subsequent sulfenic acid formation. It seems unlikely that the apparent insensitivity of the thiols to peroxide under normoxic conditions results in cysteines that are not oxidized, as they are oxidized to the disulfide under hypoxic conditions. A possible explanation is that oxygen contributes to overoxidation of the cysteines, though the mechanism is unclear, preventing intramolecular disulfide formation and thus keeping the distal pocket relatively closed. While such hyperoxidation is typically triggered by excess hydrogen peroxide [50], some protein cysteines are overoxidized readily by room air, such as Cys111 in human superoxide dismutase 1 [53,54]. Further, the loss of about half of the available reduced cysteines from hypoxic to normoxic conditions with hydrogen peroxide is consistent with an overoxidation event of a single Cygb cysteine (Supplementary Figure 6). However, alternative, yet to be determined mechanisms may still fit well with these data and we are currently exploring other possibilities. In any case, it is apparent that under normoxic conditions, formation of the intramolecular disulfide does not readily occur with H₂O₂ and the distal pocket remains relatively closed, indicating that the redox sensor status of these thiols can be further modulated by oxygen levels. Given the upregulation of Cygb under hypoxic conditions [45], the sensitivity of intramolecular disulfide bond fits with previous hypotheses that Cygb is a nitrite reductase, as low oxygen concentrations are necessary for efficient nitrite reduction [15,28,55], as well as postulations that Cygb serves as a lipid peroxidase and is involved in ROS/RNS scavenging [48]. Disulfide and thiol reducing proteins such as thioredoxin, glutaredoxin, and protein disulfide isomerases generally have been implicated in the cysteine redox switching of Ngb [56]. Logically, given Cygb’s apparent reduction potential and redox sensitivity, Cygb may also be additionally regulated by such disulfide controlling proteins, though this is beyond the scope of this paper and will be explored in later work. However, these thiols appear to be sensitive enough to trigger a redox switch to changes in the cellular redox environment on their own.

It is worthwhile to note that cellular redox conditions are dynamic. There should be a cellular ‘sweet spot’ that triggers intramolecular disulfide formation given the above discussed redox conditions are sometimes at odds (hypoxia is often associated with more reducing conditions). Additionally, a more apoptotic cell, increased mitochondrial ROS leak under hypoxia [57], or subcompartmentalization would also overcome this juxtaposition and trigger the allosteric opening of the reactive distal pocket of Cygb, allowing for changes in the activity of Cygb via redox switching depending on the cellular conditions.

Conclusions

Redox sensing proteins are crucial to oxidative stress response, and hemoproteins are often involved in such a response via ROS signaling, NO metabolism, and lipid metabolism, among others. Coupled with the ubiquity of Cygb in mammalian cells and its hypoxic induction, the results presented here regarding the redox sensitivity of these allosteric cysteines directly affecting the distal heme pocket, indicate that Cygb could be categorized as a switchable oxidative stress-mediating protein. Though the manner in which Cygb mediates oxidative stress is not formally known – and likely depends on cell-type/circumstance – it likely includes NO metabolism and lipid peroxidation. Continued studies of Cygb function in different thiol oxidation states under cellular oxidative stress conditions are merited.

Highlights.

• Ubiquitously found mammalian hemoprotein cytoglobin is a redox sensor protein

• We measured cyanide binding kinetics to probe redox allostery of Cygb cysteines

• The midpoint potential of Cygb disulfide bond is in the biological redox range

• Mild oxidizing conditions trigger disulfide bond formation and enhance reactivity

Abbreviations

4-DPS

4,4’-dipyridyl disulfide

ASK1

Apoptosis signal-regulating kinase 1

Cygb

cytoglobin

DTT

dithiothreitol

FoxO

Forkhead box protein O

GSH

reduced glutathione

GSSG

oxidized glutathione

HIF-1-α

hypoxia-inducible factor 1-alpha

KEAP1

Kelch-like ECH-associated protein 1

NEM

N-ethylmaleimide

Ngb

neuroglobin

NO

nitric oxide

ROS

reactive oxygen species

TCEP

tris(2-carboxyethyl)phosphine