Engineering a highly selective, hemoprotein-based scavenger as a carbon monoxide poisoning antidote with no hypertensive effect

Abstract

Carbon monoxide (CO) poisoning causes 50,000–100,000 emergency department visits and ~1,500 deaths in the United States annually. Current treatments are limited to supplemental and/or hyperbaric oxygen to accelerate CO elimination. Even with oxygen therapy, nearly half of CO poisoning survivors suffer long-term cardiac and neurocognitive deficits related to slow CO clearance, highlighting a need for point of care antidotal therapies. Given the natural interaction between CO and ferrous heme, we hypothesized that the hemoprotein RcoM, a transcriptional regulator of microbial CO metabolism, would make an ideal platform for CO-selective scavenging from endogenous hemoproteins. We engineered an RcoM truncate (RcoM-HBD-CCC) that exhibits high CO affinity (K_a,CO = 2.8×10¹⁰ M⁻¹), remarkable selectivity for CO over oxygen (K_a,O2 = 1.4×10⁵ M⁻¹; K_a,CO/K_a,O2 = 1.9×10⁵), thermal stability (T_m = 72°C), slow autoxidation rate (k_ox = 1.1 h⁻¹). In a murine model of acute CO poisoning, infused RcoM-HBD-CCC accelerated CO clearance from hemoglobin in red blood cells and was rapidly excreted in urine. Moreover, infused RcoM-HBD-CCC elicited minimal hypertension in mice compared to infused hemoglobin, attributed to a comparatively limited reactivity toward nitric oxide (NO) via dioxygenation (k_NOD(RcoM) = 6–8×10⁶ M⁻¹s⁻¹ vs k_NOD(Hb) = 6–8×10⁷ M⁻¹s⁻¹). These data suggest that RcoM-HBD-CCC is a safe, selective, and efficacious CO scavenger. Additionally, by limiting hypertension RcoM-HBD-CCC improves end-organ adverse effects compared with hemoglobin-based therapeutics.

Introduction

Carbon monoxide (CO) poisoning results in approximately 50,000–100,000 hospital visits in the U.S. each year, and between 1,500–2,000 people die from CO exposure annually (1, 2). Even with oxygen supplementation, the current standard of care treatment for CO poisoning, survivors often suffer increased long-term mortality (3), as well as long-term cardiac and neurocognitive deficits (4–10). When inhaled in excess, CO, which is impossible to detect as a tasteless, odorless, and colorless gas, attenuates aerobic respiration due to decreased oxygen (O₂) delivery and utilization (1, 2). CO binds the iron-containing cofactor heme with high affinity in hemoproteins; CO binds the blood oxygen-carrier hemoglobin (Hb) to form carbonylhemoglobin (HbCO, historically called ‘carboxyhemoglobin’) with an affinity 200 to 400-fold greater than oxygen, thereby diminishing oxygen delivery to tissues (11, 12). CO binding to hemoglobin stabilizes the allosteric transition from low-affinity tense state hemoglobin (Hb-T) to high-affinity relaxed state hemoglobin (Hb-R), increasing the binding affinity of CO and other diatomic ligands and further limiting oxygen binding and delivery (13). In oxygen-starved tissues, CO specifically inhibits the hemoprotein cytochrome c oxidase in the mitochondrial electron transport chain, uncoupling respiration and attenuating cellular energy production (14–17). Diminished mitochondrial respiration causes damage in tissues with high energy demands, such as cardiac and neuronal tissues, contributing to long-term sequelae in survivors of CO poisoning (1, 2). We note that the above adverse effects are the result of excessive inhalation of CO, but that CO is also produced endogenously as a byproduct of heme catabolism such that up to 2% of circulating Hb is bound to CO. This endogenous CO acts as an important signal in physiological and pathophysiological contexts (18).

To date, no antidotal therapy for CO poisoning exists, and treatments are limited to inhalation of 100% normobaric or hyperbaric oxygen. These treatments enhance CO clearance by increasing the partial pressure of oxygen and thereby increasing the rate of CO exchange in red blood cell (RBC) hemoglobin in the lungs. With 100% normombaric oxygen, the half-life of HbCO is decreased from 320 minutes to around 74 minutes (1). Hyperbaric oxygen administration can further accelerate the CO elimination; however, there is often a several hour delay between diagnosis of CO poisoning, patient transport, and treatment, reducing the effectiveness of this strategy (7, 19, 20). The latency between diagnosis and treatment, as well as persistence of long-term sequelae in patients both with and without supplemental oxygen inhalation treatment, highlight the need for point-of-care antidotes to treat CO poisoning.

Our laboratory has specialized in a promising therapeutic strategy to treat acute CO poisoning that relies on administration of high-affinity CO scavenging agents. These scavenging agents leverage the strong interaction between CO and the Fe(II) (ferrous) heme cofactor bound to a protein or small-molecule scaffold, and tight CO binding to these agents sequesters CO away from endogenous hemoprotein sites (21–24). We first demonstrated this concept with a high CO affinity recombinant hemoprotein variant of human neuroglobin (Ngb-H64Q-CCC) to treat acute CO poisoning in vivo (22, 25). Infused Ngb-H64Q-CCC, which exhibits 500-fold higher CO affinity than hemoglobin, depletes the fraction of HbCO in RBCs, reverses hemodynamic collapse, restores mitochondrial function, and improves survival in an otherwise lethal murine model of acute CO poisoning. Infusion of other hemoproteins that have similar CO affinities to RBC hemoglobin, including alkylated hemoglobin and myoglobin (Mb), also improve hemodynamic outcomes and survival in murine models of acute CO poisoning, though these proteins likely function by both providing modest CO scavenging and enhancing tissue oxygen delivery (23). In addition to these hemoproteins, several small molecule-based scavengers have shown promise as CO sequestration agents in recent years (21, 24, 26, 27). Recently, Mao et al showed that treatment with the water-soluble synthetic heme mimic, hemoCD, demonstrated efficacy against CO poisoning in mice (24). Taken together, these proof-of-concept studies demonstrate the feasibility of intravenous CO scavenging to treat acute CO poisoning; however, questions remain regarding these molecules’ optimal therapeutic window.

Generally, hemoprotein-based scavengers must 1) possess a high CO binding affinity and kinetic parameters to outcompete physiological hemoprotein targets (e.g., hemoglobin, K_d,CO = 1.7 nM for Hb-R) (12); 2) exhibit selectivity for CO over oxygen, as high scavenger oxygen affinity would inhibit CO scavenging; 3) be stable to thermal and chemical degradation to prevent heme release and subsequent adverse reactivity in physiological environments (28); and 4) exhibit redox stability, i.e., slow autoxidation from Fe(II)-O₂ to Fe(III) heme, as only the Fe(II) heme readily binds CO. Additionally, as shown historically in the development of hemoglobin-based blood substitutes, intravenous addition of hemoproteins exhibit nitric oxide (NO) scavenging reactions, triggering hypertension and heme oxidation leading to release of pro-inflammatory ferric heme. Such off-target reactions lead to acute renal failure and increased mortality in pre-clinical and clinical trials (29, 30). To circumvent these off-target effects, we seek to diversify the potential pool of CO scavenging therapeutic candidates through engineering of non-globin-based hemoproteins. By fine-tuning the amino acid composition of the hemoprotein scaffold, we can optimize critical properties for CO scavenging (ligand binding parameters, redox stability, and pharmacokinetic profile) while mitigating off-target effects observed through NO reactivity.

A hemoprotein called the regulator of CO metabolism (RcoM) protein, a bacterial, non-globin transcription factor that activates aerobic CO metabolism in the presence of CO (31, 32) may serve as an ideal platform to engineer a safe, efficacious CO scavenger. Originally isolated from the soil bacterium Paraburkholderia xenovorans, RcoM utilizes heme to sense low environmental concentrations of CO (0.1 to 25 nM) in aerobic environments, suggesting that RcoM exhibits high CO affinity, selectivity for CO over oxygen, and heme redox stability (33). The native protein is a homodimer comprised two domains: (1) an N-terminal, sensory PAS (Per-Arnt-Sim) domain, where heme and CO binding occurs, and (2) a C-terminal, DNA-binding LytTR domain (34). Studies of the RcoM-2 paralog from P. xenovorans (PxRcoM-2) report nanomolar CO binding affinity; however, a truncate bearing only the N-terminal heme-binding domain (HBD) exhibits 16-fold higher CO binding affinity (K_d,CO = 0.25 nM for PxRcoM-2-HBD) (35). Importantly, oxygen binding to full-length RcoM has not been observed to date, suggesting exquisite CO selectivity. Ferrous heme in RcoM is six-coordinate; the protein binds heme via a proximal histidine (His77) and a distal methionine (Met104), the latter of which is readily replaced by CO (31, 36, 37). This Met104 residue, which saturates the heme coordination environment, may impart additional heme stability and ligand selectivity without compromising CO binding affinity.

Given the inherent characteristics of RcoM and its truncate, we hypothesize that we can devise a safe, efficacious CO scavenging hemoprotein using the RcoM platform, but several important ligand binding and stability criteria require further optimization. Herein, we present the design and characterization of a novel hemoprotein-based scaffold based on the PxRcoM-1 paralog as a promising CO scavenging therapeutic. We demonstrate the efficacy of this scavenger, RcoM-HBD-CCC, in a murine model of severe CO poisoning, where RcoM administration accelerates CO clearance. We further demonstrate that this scavenger elicits neither hypertension nor organ-specific toxicity upon intravenous infusion in mice due to limited reactivity with NO.

Results

Design and recombinant expression of heme-loaded RcoM-HBD-CCC

We introduced two modifications to the native PxRcoM-1 protein sequence to improve CO scavenging properties. First, we removed the C-terminal, DNA-binding LytTR domain (residues 154 to 266), yielding a truncate bearing the N-terminal, heme-binding PAS domain (Figure S1a). This C-terminal domain is not involved in direct CO binding. Second, we introduced three Cys-to-Ser amino acid substitutions (C94S, C127S, and C130S) to the HBD truncate to eliminate the potential for intermolecular disulfide bond formation at high protein concentrations, as informed by previous studies (22). To facilitate isolation of high-purity recombinant protein, we appended a C-terminal 6xHis affinity tag. Initial expression of this RcoM-HBD-CCC construct, cloned into pET28a and expressed in soluBL21 E. coli, resulted in high protein yields (~44–49 mg purified holoprotein/L culture) but low heme loading (16–20% based on a 1:1 ratio of protein monomer to heme cofactor, Table S1). To increase the amount of heme-loaded protein relative to total isolated protein, RcoM-HBD-CCC was co-expressed with E. coli ferrochelatase (EcFeCH), which catalyzes the final step in heme biosynthesis. Co-expression of RcoM-HBD-CCC with EcFeCH dramatically improved heme loading (102–136% heme loading, see Materials and Methods), although overall protein yields diminished (4–15 mg purified holoprotein/L culture) due to EcFeCH co-expression.

Spectroscopic properties and stability of the RcoM-HBD-CCC heme

The RcoM-HBD-CCC truncate bears a six-coordinate heme in both Fe(III) and Fe(II) oxidation states (Figure 1A and 1B, Table S2). Spectroscopic data suggest that Fe(III) heme from the native RcoM-1 homolog from P. xenovorans is axially coordinated by a charged thiolate (from Cys94) and neutral imidazole (from His74)(31, 36). The Fe(III) heme of RcoM-HBD-CCC is also six-coordinate, as evidenced by the positioning of the Q-band (visible) absorbance features at 537 nm and 565 nm, despite the fact that the native coordinating Cys ligand (Cys94) is replaced by a non-coordinating Ser residue. It is possible that a different protein-derived ligand occupies the sixth axial position, giving rise to the low-spin, Fe(III) spectral features. Similar observations have been reported after removing the distal ligand in some cytoglobin and neuroglobin mutants (38–41). RcoM-HBD-CCC holoprotein exhibits very high thermal stability, as measured by loss of Fe(III) heme signal, with a melting temperature value, T_m, of 72 °C in phosphate-buffered saline (PBS, 10 mM, pH 7.4) (Figure 1C). Chemical reduction of the RcoM-HBD-CCC heme with sodium dithionite gives rise to an UV-Vis absorbance feature with two sharp bands at 531 nm and 562 nm, consistent with formation of a low-spin, six-coordinate heme center. Native RcoM undergoes a redox-mediated ligand switch in which the charged thiolate of Cys94 is replaced by a neutral thioether, Met104 (37). Ferrous heme coordination by Met104 is maintained in RcoM-HBD-CCC, as substitution of Met104 with a non-coordinating Leu residue (M104L) broadens the Q-bands, consistent with the formation of a five-coordinate, high-spin species (Figure S2). We note here that though the Fe(II) heme in RcoM-HBD-CCC is weakly bound by the distal Met104, for clarity we will use the term “unliganded Fe(II) heme” to describe RcoM-HBD-CCC species without an exogenous diatomic ligand bound (i.e., CO, NO, and oxygen).

Figure 1.

RcoM-HBD-CCC spectroscopic and stability properties All measurements carried out in phosphate buffered saline (PBS, 10 mM, pH 7.4). (A) Heme coordination environment for RcoM-HBD-CCC. (B) Electronic absorption (UV-Vis) spectra for RcoM-HBD-CCC heme species. (C) Thermal unfolding of RcoM-HBD-CCC bearing Fe(III) heme. Inset: Decrease in absorbance at 413 nm, corresponding to dissociation of Fe(III) heme from the protein. The melting curve was fit using the Santoro-Bolen equation (red line). (D) Comparison of purified RcoM-HBD-CCC spectra before and after chemical oxidation with excess potassium ferricyanide. Top: Reduction of heme using sodium dithionite prior to chemical oxidation yields an admixture of unliganded Fe(II) and Fe(II)-CO species. Bottom: Reduction of heme using sodium dithionite after chemical oxidation yields pure unliganded Fe(II) species. (E) Autoxidation kinetics for Fe(II)-O₂ RcoM-HBD-CCC heme under aerobic conditions at 37 °C. Inset: Kinetic traces were fit to single exponential decay functions to determine the observed rate of autoxidation, k_autox = 1.1 h⁻¹ (red lines).

The ligand binding properties of RcoM-HBD-CCC differ from those of the native full-length RcoM sensor protein. We observe a majority fraction of RcoM-HBD-CCC bearing Fe(II)-CO heme (55–85% Fe(II)-CO heme; 45–15% Fe(III) heme) after isolation under typical aerobic expression conditions from E. coli without addition of CO to culture media (Figure 1D), an observation consistent with prior studies of RcoM truncates and full-length variants bearing the M104L substitution (31, 36, 37, 42). As with other recombinant CO scavengers, such as Ngb-H64Q-CCC (22, 40, 43), isolation of CO-bound protein during heterologous production is indicative of high CO binding affinity. Potassium ferricyanide oxidizes Fe(II)-CO RcoM-HBD-CCC, yielding Fe(III) heme (Figure 1D) and releasing heme-bound CO. This re-oxidized protein undergoes facile chemical reduction with sodium dithionite to generate the homogeneous Fe(II) (ferrous) Met104-bound species. This ferrous species also readily binds nitric oxide (NO) to yield spectroscopic features consistent with a low-spin, six-coordinate Fe(II)-NO species (Table S2). Additionally, Fe(II) RcoM-HBD-CCC forms a stable Fe(II)-O₂ adduct not previously described for full-length or other truncated RcoM variants (Figure 1B). While the peak maxima of these Fe(II)-O₂ features (Q-bands at 540 nm and 573 nm) are very similar to those of the Fe(II)-CO species, the band shapes and relative intensities are distinct. The RcoM-HBD-CCC Fe(II)-O₂ adduct is relatively stable towards decomposition into Fe(III) heme and superoxide with an apparent autoxidation rate, k_autox, of 1.1 h⁻¹ (t_1/2 = 38 min) at 37 °C (Figure 1E), faster than hemoglobin or myoglobin but much slower than rates reported for cytoglobin or neuroglobin (40, 44). This rate slows to 0.15 h⁻¹ (t_1/2 = 277 min) at 22 °C (Figure S3).

CO binding parameters for RcoM-HBD-CCC

To quantify the CO affinity constant (K_a,CO) of RcoM-HBD-CCC, we first sought to measure kinetic rate constants for CO binding (k_on,CO) and dissociation (k_off,CO). We can then compute K_a,CO as a ratio of binding association and dissociation rate constants (K_a,CO = k_on,CO/k_off,CO. We initially attempted to quantify k_on,CO using flash photolysis-rebinding studies; however, we observed minimal CO escape from the heme pocket after photolysis, consistent with complete geminate recombination of CO and Fe(II) heme (vide infra). Given that CO escape from the heme pocket is needed to quantify k_on,CO, this result precluded the use of flash photolysis-rebinding. Instead, we followed CO binding kinetics for RcoM-HBD-CCC using stopped-flow electronic absorption (UV-Visible) spectroscopy, observing CO binding to RcoM-HBD-CCC on the millisecond timescale (Figure 2A). CO binding kinetics followed a single exponential function under pseudo-first order conditions (Figure 2A, right); changes in observed rates varied linearly as a function of CO concentration, giving rise to the second-order rate constant (k_on,CO) of 4.67×10⁴ M⁻¹s⁻¹ (Figure 2D). In a prior stopped-flow study investigating CO binding kinetics to Fe(II) heme in a HBD truncate of the PxRcoM-2 ortholog, the observed binding rate constant approached a limiting value at CO concentrations of 0.5 mM in PBS, presumably due to the limitations of Met104 dissociation from heme (42). We observed no such limiting rate of CO binding; however, we note that we were unable to achieve a concentration of CO greater than 0.3 mM in solution using our stopped-flow setup.

Figure 2.

Determination of kinetic and thermodynamic parameters for CO binding to RcoM-HBD-CCC. (A) (Left) Spectral changes in the visible (Q-band) region upon stopped-flow rapid mixing of Fe(II) RcoM-HBD-CCC (10 μM, solid black line) and CO-saturated PBS at 25 °C and 297 μM CO, to yield Fe(II)-CO RcoM-HBD-CCC (solid blue line). (Right) Corresponding kinetic traces following the loss of signal from Fe(II) RcoM-HBD-CCC (absorbance at 564 nm, grey line) and formation of Fe(II)-CO RcoM-HBD-CCC (absorbance at 578 nm, red line) at 25 °C. Black dashed lines depict best fits to a single exponential function. (B) Plot of observed pseudo first-order rate constants for CO binding (k_obs) to Fe(II) RcoM-HBD-CCC as a function of CO concentration at 25 °C. Each data point represents the average observed rate constant for four separate mixing events. A simple linear regression yields k_on,CO = 4.67×10⁴ M⁻¹s⁻¹. (C) Kinetics of CO dissociation from Fe(II)-CO RcoM HBD-CCC (6.6 μM) at 25 °C, as measured by replacement with NO over 8 days. Reference data for Fe(II)-CO (blue dashed line) and Fe(II)-NO (black dashed line) RcoM-HBD-CCC species are superimposed over CO displacement data, normalized to highest peak intensity for each respective species. (D) Corresponding kinetic trace of CO dissociation, as measured by change in Soret absorbance at 423 nm as CO is replaced by NO at Fe(II) heme (green circles). The data were fit to a single-exponential decay function (black line) to determine k_on,CO = 1.67×10⁻⁶ s⁻¹. (E) Transient electronic absorption spectra for Fe(II)-CO RcoM-HBD-CCC at different delay times (t) following CO dissociation by flash photolysis. Soret features are plotted as difference spectra where ΔOD(t)=Abs(t)-Abs(t₀) and t₀ is the initial Fe(II)-CO spectrum. (F) CO geminate rebinding kinetics measured at 437 nm following flash photolysis (blue line). The black line represents the curve of best fit to a single exponential function (t_½= 195 ps; τ=281 ps).

CO dissociation from the RcoM-HBD-CCC heme is extremely slow. We measured CO dissociation kinetics under an atmosphere of anaerobic NO using UV-Vis spectroscopy by following CO replacement with NO (Figure 2C). We observed a slow isosbestic shift as pre-formed Fe(II)-CO protein is slowly converted to the Fe(II)-NO species. By fitting the change in absorption for the Fe(II)-CO Soret peak at 423 nm to a single exponential function, we obtained a first-order dissociative rate constant (k_off,CO) of 1.67×10⁻⁶ s⁻¹ (t_1/2 = 115 h, Figure 2D).

Taking the ratio of experimentally determined rate constants for CO binding and dissociation, we obtain a value of K_a,CO = 2.80×10¹⁰ M⁻¹ for RcoM-HBD-CCC, a value nearly 50-fold higher than that of Hb-R (Table 1). We also note that the observed CO dissociation rate constant for RcoM-HBD-CCC is the lowest for any characterized hemoprotein (35), and it is slow dissociation of CO from heme that primarily drives the high CO binding affinity for this engineered CO scavenger (Table 1).

Table 1.

Summary of kinetic and thermodynamic binding parameters for oxygen and CO for select hemoproteins.

protein	O2 kon (M−1s−1)	O2 koff (s−1)	O2 KA (M−1)	CO kon (M−1s−1)	CO koff (s−1)	CO KA (M−1)	M (KCO/KO2)
RcoM-HBD-CCC	(8.2×105)*	5.9	1.4×105	4.7×104	1.7×10−6	2.7 to 2.8×1010	1.9×105
Hb R1	5.0×107	15	3.3×106	6.0×106	1.0×10−2	6.0×108	180
Hb T1	4.5×106	1.9×103	2.4×103	8.3×104	9.0×10−2	9.2×105	390
StHb2	2.4×107	15	1.6×106	5.1×106	1.3×10−2	3.9×108	242
Ngb-H64Q-CCC3	7.2×108	18	3.9×107	1.6×108	4.2×10−4	3.8×1011	9.7×103
hemoCD4	4.7×107	1.3×103	3.6×104	1.3×107	2.5×10−4	5.2×1010	1.4×106

^1.Values from Cooper, C.E. Biochim. Biophys. Acta Bioenerg. 1999

^2.Values from Xu, Q. et al. J. Clin. Invest. Insight 2022

^3.Values from Azarov, I. et al. Sci. Transl. Med. 2016

^4.Values from Kano, K. et al. Inorg. Chem. 2006

^*calculated from K_A,O2 and k_off,o2

Slow dissociation of CO from the RcoM-HBD-CCC heme is likely driven by fast and complete geminate recombination. Ultrafast transient electronic absorption spectroscopy was employed to quantify geminate CO rebinding kinetics for Fe(II)-CO RcoM-HBD-CCC. Following a 200 femtosecond pulse at 570 nm, we observe a difference absorbance spectrum characteristic of a transition from 6-coordinate, low-spin Fe(II) to 5-coordinate high-spin Fe(II) heme (Figure 2E), consistent with complete CO photolysis. Within 1 ns, the amplitude of this difference spectrum rapidly diminished, and the spectrum returned to that of Fe(II)-CO RcoM-HBD-CCC. The loss of 5-coordinate, high-spin Fe(II) signal followed a single exponential decay (t_1/2=195 ps; t=281 ps) and approached a signal plateau that was 3.5% that of the initial amplitude (Figure 2F), demonstrating rapid and nearly complete geminate recombination of CO.

Oxygen binding parameters for RcoM-HBD-CCC

In addition to possessing a CO binding affinity (K_d) in the picomolar range, RcoM-HBD-CCC exhibits remarkable selectivity for CO over oxygen. A low autoxidation rate allowed for direct quantification of the oxygen binding affinity (K_a,O2) for Fe(II) RcoM-HBD-CCC using standard tonometry methods and UV-Vis spectroscopy to monitor oxygen binding (Figure S3). Using the Hill equation to fit the oxygen binding curve, we obtain an association binding constant K_a,O2 = 1.4×10⁵ M⁻¹ (P₅₀ = 3.98 mmHg) for RcoM-HBD-CCC. This value is five orders of magnitude lower than that of CO, giving rise to a selectivity constant (M-value) of 1.9×10⁵ (Table 1). We used stopped-flow UV-Vis spectroscopy to quantify oxygen dissociation from Fe(II) RcoM-HBD-CCC heme in the presence of excess CO, obtaining a k_off,O2 of 5.9 s⁻¹ (Figure S3). With the experimental values for K_a,O2 and k_off,O2, we computed the oxygen association rate constant (k_on,O2) for Fe(II) RcoM-HBD-CCC, obtaining a value of 8.2×10⁵ M⁻¹s⁻¹. Table 1 compares these values to R- and T-state hemoglobin, as well as these parameters for other published CO scavengers.

In vitro CO scavenging by RcoM-HBD-CCC

RcoM-HBD-CCC effectively sequesters CO from RBC-encapsulated HbCO under aerobic and anaerobic conditions in vitro. To model CO scavenging in circulation, we incubated CO-saturated murine RBCs (>90% HbCO) with Fe(II)-O₂ RcoM-HBD-CCC at equimolar final concentrations of 100 μM heme under aerobic conditions at 37 °C. Using UV-Vis spectroscopy, we followed CO transfer from encapsulated Hb to extracellular RcoM and used spectral deconvolution to quantify the fraction of CO-bound and CO-free hemoproteins in RBC and extracellular compartments (Figure 3). We observed rapid loss of 0.72 ± 0.03 equivalents of CO from RBC Hb in approximately five minutes, with a HbCO decay half-life of 27 ± 4 s. A slightly higher than expected fraction of Fe(II)-CO RcoM-HBD-CCC from CO transfer from Hb was observed (0.86 ± 0.04 equivalents), likely due to a small amount of excess CO dissolved in solution during the preparation of CO-saturated RBCs. CO transfer is slightly slowed at 25 °C under otherwise identical conditions (HbCO decay half-life of 38 ± 2 s), attributed to the decreased kinetics of ligand exchange (Figure S4A). Under these same conditions, when one equivalent of Fe(II)-O₂ RcoM-HBD-CCC is incubated with a five-fold excess of RBC-encapsulated HbCO, nearly complete transfer of one equivalent of CO from Hb to RcoM is observed at an accelerated rate, with a HbCO decay half-life of 15 ± 2 s (Figure S4B). Under anaerobic conditions and 25 °C, we observe complete, rapid transfer of one equivalent of CO from RBC-encapsulated deoxyHb to Fe(II) RcoM-HBD-CCC (Figure S4C). Taken together, these data demonstrate that RcoM-HBD-CCC sequesters the majority of CO from HbCO within minutes in RBCs under both aerobic and anaerobic conditions.

Figure 3.

RcoM-HBD-CCC rapidly scavenges CO from RBCs ex vivo under aerobic conditions. CO-saturated murine RBCs (>90% HbCO, 13 μM heme) were incubated with RcoM-HBD-CCC (>99% Fe(II)-O₂, 13 μM heme) at 37 °C for 10 min total. Representative time courses following spectral features for lysed RBCs (A) and cell-free RcoM-HBD-CCC (B). (C) Combined kinetic data summarizing CO transfer from RBC-encapsulated HbCO (red circles) to extracellular RcoM-HBD-CCC (blue squares). Fractions of CO-bound species were determined by spectral deconvolution. Each data point represents the average value from three technical replicates, and error bars denote ± 1 SD. Dashed lines represent best fits to single-exponential functions (k_obs = 0.027 ± 0.003 s⁻¹).

In vivo CO scavenging by RcoM-HBD-CCC in a severe, nonlethal CO poisoning model

RcoM-HBD-CCC accelerates CO clearance from circulation in mice challenged with acute, severe CO poisoning. Using a previously established model of severe, non-lethal CO poisoning (23), we assessed hemodynamic outcomes and scavenger-dependent changes in the fraction of CO-bound hemoglobin in RBCs ($f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$) in mice. Tracheally-intubated, anaesthetized animals under mechanical ventilation were exposed to a 3% CO gas mixture balanced with air for 1.5 min resulting in HbCO levels of ~76% in these animals. Three minutes after CO exposure, PBS vehicle or Fe(II)-O₂ RcoM-HBD-CCC (scavenger) in PBS was administered intravenously via a jugular vein catheter at scavenger protein dosages of 255 mg/kg (15 μmol¹/kg), 510 mg/kg (30 μmol/kg), or 1020 mg/kg (60 μmol/kg) and a dosing volume of 10 mL/kg body weight, or a total scavenger volume of ~300 mL (Figure 4A). All RcoM-HBD-CCC samples used for in vivo testing contained <0.1% of inactive Fe(III) heme by spectral deconvolution prior to infusion (Figure S5). Following CO inhalation, mean arterial pressure (MAP) dropped from 80–85 mmHg to ~50 mmHg (Figure 4B, Table S3. Infusion occurs with a transient increase in blood pressure to values around 100 mmHg in both vehicle and scavenger treatment groups, followed by a stabilization to blood pressure values at or slightly above baseline within 10 min of CO exposure.

Figure 4.

RcoM-HBD-CCC accelerates CO clearance in a murine model of severe CO poisoning. (A) (left) Experimental scheme for severe CO poisoning mouse model. Red arrows denote 15 uL blood draws. Body temperature was monitored and maintained at 37 °C throughout the course of the experiment. (right) RcoM dosage and animal numbers. (B) Changes in mean arterial pressure (MAP) as a function of time in the severe, nonlethal CO poisoning model. The average starting blood pressure across all treatment groups is depicted by the vertical dashed line at 83.6 mmHg. (C) (left) Comparison of CO clearance, as measured by the difference in $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ before and after infusion, $\mathrm{\Delta }{f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$(T2-T1), as a function of scavenger dose. Statistical significance between vehicle and treatment groups was assessed using ordinary one-way ANOVA (*, p<0.05; ***, p<0.001; ****, p<0.0001). (right) Changes in the fraction of circulating HbCO ($f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$) after CO exposure and delivery of RcoM-HBD-CCC. Values for $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ were quantified by spectroscopic analysis of lysed RBCs isolated from whole blood drawn at four time points throughout the experiment. All $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ values were adjusted to a starting value of 0.76, the average initial $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ at T1 across all treatment groups. (D) Representative spectroscopic data for plasma samples (diluted 40-fold in PBS) taken at different time points following RcoM-HBD-CCC infusion (T2 to T4) in the severe CO poisoning model. Spectra are compared before (solid lines) and after chemical reduction using sodium dithionite (dashed lines) to highlight the degree of CO binding. (E) Assessment of RcoM-HBD-CCC speciation and concentration in plasma. Values for f_CO-RcoM (top) and scavenger concentration (bottom) were quantified by spectroscopic analysis of plasma samples isolated from whole blood. (F) Representative images of urine samples collected 45 min after acute CO exposure and subsequent infusion with scavenger. (G) Spectroscopic analysis of urine samples reveal that RcoM-HBD-CCC scavenger eliminated was >90% CO-bound. For hemodynamic data, error bars denote ± SEM; for all other data, error bars denote ± SD.

CO clearance was assessed by monitoring changes in $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ in lysed RBCs from samples drawn just before scavenger infusion (T1), after infusion (T2), 22.5 min (T3), and 44.5 min (T4) after CO exposure via a carotid arterial catheter (Figure 5C). Values for $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ were computed as follows:

$$f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}=\frac{{\left[\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}\right]}_{RBC}}{{\left[\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{H}\mathrm{b}\right]}_{RBC}}$$ (1)

where ${\left[\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}\right]}_{\mathrm{R}\mathrm{B}\mathrm{C}}$ and ${\left[\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{H}\mathrm{b}\right]}_{\mathrm{R}\mathrm{B}\mathrm{C}}$ are quantified from spectral deconvolution of lysed RBCs. Additionally, RBCs are lysed in the presence of 20 μM sodium dithionite, which eliminates Fe(III) heme sites and scavenges heme-bound oxygen. As a result, only deoxyHb (i.e., Fe(II) unliganded Hb) and HbCO species remain in the lysate, and reference data (molar absorptivity values at 1 nm increments from 450–700 nm) for these two spectroscopically distinct species are used to quantify their absolute concentrations in the admixture. The average starting value for $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ in all treatment groups was 0.760 ± 0.018. In control animals infused with PBS vehicle, a decrease in $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ is observed over time, consistent with both physiological clearance due to gas exchange in the lungs and loss of CO to peripheral tissues. The change in $f_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ values before and after infusion, $\mathrm{\Delta }{f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ (T2-T1), shows a significant, dose-dependent acceleration of CO clearance with RcoM-HBD-CCC infusion for all scavenger groups (${\mathrm{\Delta }f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}=-0.104\pm 0.008,-0.132\pm 0.006,-0.167\pm 0.024$, respective for increasing doses) compared to the control group ($\mathrm{\Delta }{f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}=-0.079\pm 0.011$, Figure 4B). No significant differences in $\mathrm{\Delta }{f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$ were observed at subsequent time points between control and scavenger groups (Figure S6).

Figure 5.

Intravenous infusion of RcoM-HBD-CCC does not elicit hypertension in mice with or without CO exposure. (A) Mean arterial pressure (MAP) as a function of time in the severe, nonlethal CO poisoning model with intravenous infusion of PBS vehicle (n=4, black), StHb at a dose of 800 mg/kg (50 μmol/kg heme; n=3, red), or RcoM-HBD-CCC at a dose of 1020 mg/kg protein (60 μmol/kg heme; n=4, blue). Average starting blood pressure across all treatment groups was 83.6 mmHg (black dashed line). Black arrows denote ten-minute time intervals following CO exposure. (B) Highlighted values for change in MAP (ΔMAP) following the initiation of CO exposure. Each data point represents the difference in blood pressure between the starting value at t=0 and ten-minute time intervals following CO exposure for each replicate. (C) MAP as a function of time in healthy animals after intravenous infusion of PBS vehicle (n=3, black), StHb at a dose of 480 mg/kg protein (30 μmol/kg heme; n=3, red), or RcoM-HBD-CCC at a dose of 510 mg/kg protein (30 μmol/kg heme; n=4, blue). Average starting blood pressure across all treatment groups was 86.0 mmHg (black dashed line). Note that the y-axis range in panel C differs from that in panel A. (D) Highlighted ΔMAP values as in panel B. Statistical significance between vehicle and treatment groups was assessed using one-way ANOVA with multiple comparisons between means in each treatment group (*, p<0.05). (E) Reaction scheme for the estimation of the NO dioxygenation rate constant for RcoM-HBD-CCC (k_NOD,RcoM) through a competition reaction with StHb. (F) Final amount of NO scavenging due to NO dioxygenation by StHb (red bar, 27.4 ± 0.8 nmol) and RcoM-HBD-CCC (blue bar, 2.8 ± 0.1 nmol), as determined by spectral deconvolution (n=4 technical replicates). (G) Correlation between apparent rate of NO dioxygenation (k_NOD) and ΔMAP recorded t = 10 min (3.5 min after complete infusion) in non-CO-poisoned mice. Points represent the mean ΔMAP across biological replicates in each treatment group. Values for k_NOD were derived from the literature for Hb (7 × 10⁷ M⁻¹s⁻¹) and Mb (3.4 × 10⁷ M⁻¹s⁻¹; refs 52 and 53) and from competition kinetics experiments for RcoM-HBD-CCC (7 × 10⁶ M⁻¹s⁻¹) and Ngb-H64Q-CCC (3 × 10⁷ M⁻¹s⁻¹). Note that RcoM-HBD-CCC and StHb (30 μmol/kg heme) were administered at a higher dose than equine myoglobin (Mb) and Ngb-H64Q-CCC (20 μmol/kg heme). For hemodynamic data, error bars denote ± SEM; for all other data, error bars denote ± SD.

RcoM-HBD-CCC is rapidly saturated with CO in circulation. As with RBC Hb speciation in vitro, we followed the speciation of intravenous RcoM-HBD-CCC scavenger in plasma, isolated from blood samples used to measure RBC HbCO described above. Spectroscopic data for diluted plasma samples are consistent with rapid CO binding by the scavenger at all doses (Figure 4D). We observed minimal changes in spectral features for plasma samples upon addition of sodium dithionite, which eliminates all oxygen, including any bound to RcoM-HBD-CCC. The lack of spectral changes further confirm a high fraction of Fe(II)-CO RcoM-HBD-CCC vs oxygen bound, ferric or unliganded RcoM-HBD-CCC, as the spectral features of the former are unaffected by the presence of the dithionite reducing agent. Using spectral deconvolution, we determined the fraction of CO-bound RcoM-HBD-CCC (f_CO-RcoM) to be greater than 90% in all scavenger treatment groups at all time points (Figure 4E, top). Total scavenger concentrations in plasma directly correlate with dosing; elimination half-lives for each group occur between 30–40 min, though we note that we did not follow to complete elimination from plasma here (Figure 4E, bottom).

RcoM-HBD-CCC is rapidly excreted in the urine. As with plasma scavenger concentrations, the concentration of RcoM-HBD-CCC eliminated in the urine increases linearly with the dose administered in this model (Figure 4F and 4G). A urine sample was isolated from each animal ~45 min following exposure to CO. Spectral features for RcoM-HBD-CCC recovered from urine were consistent with those observed in plasma samples. Moreover, spectral deconvolution of urine samples after treatment with sodium dithionite reveal greater than 90% CO-bound RcoM-HBD-CCC in all treatment groups, indicating RcoM-HBD-CCC maintains CO ligation through facile renal clearance and excretion in urine.

RcoM-HBD-CCC NO scavenging kinetics mitigate vasoconstriction in vivo

Unlike hemoglobin and modified hemoglobins (23), infusion of Fe(II)-O₂ RcoM-HBD-CCC does not elicit an increase in blood pressure in the context of severe, nonlethal CO poisoning. The hemodynamic parameters for animals treated with RcoM-HBD-CCC (1020 mg/kg) compared to animals treated with a similar dose of cell-free, 2,3-diphosphoglycerate stripped human hemoglobin (StHb, 800 mg/kg, 50 μmol/kg heme) under analogous CO poisoning conditions are consistent with this observation (Figure 5A,B). For all animals in all treatment groups, a transient decrease in MAP to a minimum value of 40–50 mmHg was observed 3–4 min after the start of CO exposure (Figure 5A, Table S3), consistent with what was observed in Figure 4B. Likewise, an increase in MAP was observed just prior to infusion, likely to due to a combination of adrenergic response and introduction of fluid (45–47). During the infusion period, animals in all groups (including PBS vehicle) exhibit a rapid, transient increase in MAP to a maximum value of ~100 mmHg, followed by a decrease to pressure values just above baseline over a ~4 min period following infusion. In animals administered RcoM-HBD-CCC or PBS vehicle, MAPs stabilized just above basal values for the duration of the experiment. In contrast, MAPs increase by ~15 mmHg in animals administered StHb between t=8 min and t=20 min, and this hypertensive effect persists for the duration of the experiment.

To eliminate CO exposure as a confounding factor, we infused StHb or RcoM-HBD-CCC in healthy mice administered medical air without CO (Figure 5C,D). Non-CO-poisoned animals in all treatment groups exhibited stable blood pressure values of 84–86 mmHg prior to infusion (Table S4). After infusion of RcoM-HBD-CCC or PBS vehicle, animals exhibited a small, transient decrease in blood pressure (ΔMAP= −5.5 ± 1.3 mmHg for RcoM-HBD-CCC and −2.1 ± 4.5 mmHg for PBS at t=10 min), followed by a return to near-basal MAP values by t=40 min. In contrast, healthy animals administered StHb exhibited an immediate, significant increase in blood pressure after infusion (ΔMAP= +8.0 ± 4.3 mmHg at t=10 min). That StHb induced hypertension is not surprising given the well-documented vasoconstrictive effects of cell-free oxyhemoglobin due to rapid NO consumption via NO dioxygenation (48–50). However, we were surprised to find that RcoM-HBD-CCC, which stably binds oxygen and may therefore also scavenge NO via NO dioxygenation, did not elicit hypertension in vivo.

It is possible that RcoM-HBD-CCC-induced hypertension is mitigated by rapid clearance from plasma. The RcoM-HBD-CCC monomer (17 kDa) is nearly one-quarter the size of the Hb tetramer (65 kDa), and the RcoM-HBD-CCC elimination half-life (t_1/2 = 30–40 min) is correspondingly shorter than that estimated for Hb (t_1/2 ~ 2 h) (23). We therefore assessed the vasoactivity of two hemoproteins comparable in size (and presumably elimination half-life) to RcoM-HBD-CCC: myoglobin (Mb; 17 kDa) and Ngb-H64Q-CCC (17 kDa). Infusion of either oxyferrous Mb or oxyferrous Ngb-H64Q-CCC at a dose of 340 mg/kg (20 μmol/kg heme) elicited hypertension within one minute of infusion in healthy (non-CO-poisoned) mice (Figure S7). This hypertensive effect was sustained for more than 10 min (Figure S7). These results suggest that clearance rate is not the primary factor contributing to the apparent lack of vasoactivity for RcoM-HBD-CCC.

Having eliminated rapid plasma clearance as a factor, we hypothesized that slow NO dioxygenation could account for the lack of hypertension elicited by RcoM-HBD-CCC. To test this hypothesis, we determined the rate constant for NO dioxygenation (k_NOD) for Fe(II)-O₂ RcoM-HBD-CCC. NO dioxygenation kinetics for oxyferrous hemoproteins are exceedingly rapid (k_NOD,Hb = 6 to 8 ×10⁷ M⁻¹s⁻¹) and difficult to measure, even with a stopped-flow instrument (51–53). To circumvent this limitation, we carried out a competition experiment in which equimolar amounts of Fe(II)-O₂ RcoM-HBD-CCC, Fe(II)-O₂ StHb, and NO were mixed in a sealed cuvette (Figure 5E). As one equivalent of NO reacts irreversibly with one equivalent of Fe(II)-O₂ hemoprotein to make a ferric hemoprotein, the final fraction of Fe(III) heme for each protein directly correlates to the amount of NO scavenged. Spectroscopic changes were rapid and reached final equilibrium within 5 minutes of mixing (Figure S8). Spectral deconvolution of the final reaction mixture was facilitated by spectroscopic differences between the two Fe(III) species (Figure S8). The final distribution of Fe(III) species was 90.2 ± 0.6% Hb and 9.8 ± 0.6% RcoM-HBD-CCC (Figure 5F, Table S5). This final distribution reflects the relative rates between RcoM-HBD-CCC (k_NOD,RcoM) and StHb (k_NOD,Hb), given that NO dioxygenation is irreversible and fast for both proteins. We therefore estimate an upper limit of k_NOD,RcoM of 6–8 ×10⁶ M⁻¹s⁻¹ from known values of k_NOD,Hb (51–53). Using the same competition methodology for Ngb-H64Q-CCC, we estimated k_NOD,Ngb = 2.6–3.4 × 10⁷ M⁻¹s⁻¹ (Figure S8), a value 4–5-fold higher than that of RcoM-HBD-CCC and comparable to reported rate constants for Mb (k_NOD,Mb = 3.4 × 10⁷ M⁻¹s⁻¹) (52). Consistent with prior studies, our combined hemodynamic and kinetics data suggest a strong correlation the value for k_NOD and the degree of hypertension induced upon hemoprotein infusion (Figure 5G). We conclude that the relatively sluggish k_NOD value observed for RcoM-HBD-CCC minimizes the degree of NO scavenging that occurs upon infusion of the protein, resulting essentially no hypertension at the doses examined.

RcoM-HBD-CCC safety and toxicity studies in healthy mice

Healthy (non-CO poisoned) mice tolerated intravenous infusion of RcoM-HBD-CCC without noticeable organ injury. Fe(II)-O₂ RcoM-HBD-CCC (170 mg/kg) was administered via tail vein catheter to a separate cohort of healthy, non-CO poisoned mice. All mice administered RcoM-HBD-CCC (N=3) and PBS vehicle (N=6) survived to the end of the 48-hour monitoring period, and no significant differences in animal behavior were observed. No accumulation of RcoM-HBD-CCC in tissue (lung, liver, spleen, or kidney) was observed after the 48-hour observation period by Western blot against anti-6xHis antibodies, nor were plasma biomarkers for liver or kidney function elevated compared to animals administered PBS vehicle, demonstrating minimal organ-specific toxicity (Figure S8).

Discussion

In this study, we engineered a variant of the microbial CO-sensing transcription factor, RcoM, with ideal properties for a pharmaceutical CO antidote. We hypothesized that RcoM, which utilizes heme to sense low environmental concentrations of CO in aerobic environments, would act as an ideal platform to design a selective, high-affinity CO scavenger. Consistent with this hypothesis, we identified a truncated variant of RcoM, RcoM-HBD-CCC, that exhibits extremely high affinity for CO (K_d,CO = 37 pM), but only modest affinity for O₂ (K_d,O2 = 7.1 μM). In contrast to previously reported hemoprotein-based CO scavengers, which are globin-based (22, 23), the 17 kDa RcoM-HBD-CCC likely adopts a tertiary structure consistent with a PAS domain. The ~100 amino acid PAS domain, found in all kingdoms of life, contains a highly conserved structural motif comprised of a five-stranded β-sheet with intervening α-helices that together form a ligand binding pocket that can be specifically modified to accommodate a wide variety of ligands (53). In the case of RcoM, this binding pocket bears a heme (31, 32). Like native RcoM, heme irreversibly binds to RcoM-HBD-CCC under physiological conditions, as evidenced by the protein’s high thermal stability (T_m = 72 °C). Unlike globin-based CO scavengers, the heme iron in RcoM-HBD-CCC remains coordinatively saturated in both ferric and ferrous oxidation states (Figure 1A), likely contributing to the stability, relatively slow autoxidation, diatomic ligand selectivity, and lack of organ-specific toxicity for RcoM-HBD-CCC.

The exceedingly slow dissociation of CO from Fe(II) heme drives high CO binding affinity and selectivity for RcoM-HBD-CCC. Our experimental data suggest a first-order rate constant of 1.67×10⁻⁶ s⁻¹ for CO dissociation from Fe(II) RcoM-HBD-CCC. This value for RcoM-HBD-CCC is consistent with a dissociation rate constant previously measured for the heme-binding domain for the PxRcoM-2 ortholog (k_CO,off = 3.5×10⁻⁶ s⁻¹) (35). As with PxRcoM-2, we find that slow dissociation for RcoM-HBD-CCC is likely attributed to fast (ps) and almost complete geminate recombination (Figure 2E). For most hemoproteins, geminate recombination is readily observed for O₂ and NO, but to a far lesser extent for CO (55). Fast geminate recombination with CO has been characterized in a handful of hemoproteins, including the other microbial CO-sensing transcription factor, CooA (56–58), the PAS-containing microbial oxygen sensor, DosP (59), truncated bacterial hemoglobin (60), and Met80 variants of the electron transfer protein cytochrome c (61). Extensive studies of these proteins using ultrafast ligand rebinding spectroscopy, high-resolution structural data, mutagenesis, and molecular dynamics simulations suggest that the distal heme pocket positionally restricts CO in an orientation perpendicular to the heme plane (55). It is possible that a similar effect is operative in RcoM-HBD-CCC (35, 42), though no high-resolution structural data exist to date to confirm this hypothesis. While the underlying structural features remain uncharacterized, we speculate that RcoM likely evolved to maximize sensitivity towards CO while preventing off-target activation by oxygen in aerobic niches of CO-oxidizing bacteria. From a therapeutic standpoint, this selectivity improves the CO scavenging activity, while also limiting heme-iron oxidation, protein oxidative or thermal denaturation, and NO scavenging, especially in the presence of CO.

The kinetic ligand binding parameters for RcoM-HBD-CCC enable oxygen delivery and fast, irreversible CO scavenging in the context of acute CO poisoning. Given that oxygen escapes the heme pocket five times faster than CO, we reasoned that RcoM-HBD-CCC could be administered bearing Fe(II)-O₂ heme in the context of CO poisoning. At or below oxygen tensions of ~20 mmHg, there will be rapid net release of oxygen from RcoM-HBD-CCC (t_1/2 for oxygen dissociation is 0.1 s), allowing for stoichiometric delivery of one equivalent of oxygen for every RcoM-HBD-CCC heme. After oxygen release, unliganded RcoM Fe(II) heme sites are free to irreversibly bind and sequester CO in circulation or provide further oxygen transport and delivery depending on the relative concentrations of CO and oxygen in plasma. Consistent with these kinetic parameters, mixing cell-free Fe(II)-O₂ RcoM-HBD-CCC with RBC-encapsulated HbCO in a 1:1 ratio in vitro resulted in ~75% transfer of CO from Hb to RcoM within 5 minutes of mixing (Figure 3). This transfer improves to 100% under strictly anaerobic conditions (where unliganded Fe(II) RcoM-HBD-CCC is used) and suggests that RcoM-HBD-CCC should act as an effective CO scavenging agent in vivo during acute CO poisoning. Importantly, these data validate our prior theoretical analysis of CO scavenging kinetics, which suggested that, for a sufficiently selective scavenger, the rate of CO sequestration is primarily dictated by the rate of CO dissociation from HbCO (23).

RcoM-HBD-CCC accelerated CO clearance in a severe, murine model of acute, inhaled CO poisoning. Upon RcoM-HBD-CCC infusion, we observed a rapid, dose-dependent decrease in HbCO levels ($\mathrm{\Delta }{f}_{\mathrm{H}\mathrm{b}\mathrm{C}\mathrm{O}}$) following CO exposure. Even at the smallest dose of RcoM-HBD-CCC administered (255 mg/kg), we observe a quantifiable difference in HbCO levels vs PBS vehicle immediately following infusion. While no significant differences in HbCO levels were observed at subsequent time points between control and scavenger groups, we note that mice exhibit an estimated 20-fold faster rate of CO clearance compared to humans (t_1/2HbCO = 15 min vs mice t_1/2HbCO = 320, respectively) (22, 23). As a result, we anticipate differences in HbCO clearance between treated and control patient groups to be more pronounced over longer time periods in humans. Consistent with in vitro kinetics, which suggest rapid, irreversible CO binding by RcoM-HBD-CCC, the scavenger was >90% CO-bound in plasma immediately following infusion and throughout the course of the experiment (Figure 4). Roughly one hour after infusion, Fe(II)-CO RcoM-HBD-CCC was identified at high concentration in the urine. These results are consistent with rapid binding, sequestration, and elimination of CO by RcoM-HBD-CCC.

Remarkably, intravenous infusion of RcoM-HBD-CCC bearing Fe(II)-O₂ heme did not lead to an increase in MAP in healthy animals nor in animals exposed to CO. In the presence of CO, the heme-protein scavengers do not usually cause significant hypertensive effect, due to the large excess of CO versus NO. no matter the heme protein used (22, 23), the incidental observation of a lack of hypertensive effect in the absence of CO was surprising given the well-established hypertensive effects of cell-free, oxyferrous globin proteins employed as hemoglobin-based oxygen carriers (HBOCs) (48–50). Mechanistic investigations revealed that HBOC-derived hypertension is primarily driven by depletion of tonic NO levels through fast and irreversible NO dioxygenation, and there is a strong correlation between NO dioxygenation rate (k_NOD) and the extent of hypertension induced upon intravenous infusion (52, 53, 63, 64). For example, in rats infused at the same dosage, recombinant human hemoglobin (rHb0.1) with k_NOD = 5.8 × 10⁷ M⁻¹s⁻¹ gives rise to an increase in MAP of 28.4 ± 5.7 mmHg, while a Hb variant (rHb4) with 30-fold lower NO dioxygenation rate (k_NOD = 2 × 10⁶ M⁻¹s⁻¹) only gives rise to MAP increase of 7.6 ± 1.0 mmHg, a four-fold lower effect compared to that of rHb0.1 (62). Our hemodynamic data for three globin proteins (StHb, Mb, and Ngb-H64Q-CCC) administered to non-CO-poisoned mice recapitulate this trend (Figure 5G). Considering this correlation, we hypothesized that the lack of RcoM-HBD-CCC vasoactivity may derive from a slow NO dioxygenation rate. In support of this hypothesis, we estimated an upper-limit value for k_NOD,RcoM (6–8 × 10⁶ M⁻¹s⁻¹) that was ~10% that of k_NOD,_StHb. Thus, even under conditions where infused RcoM-HBD-CCC is not immediately saturated by CO (e.g., a misdiagnosis of CO poisoning), slow NO dioxygenation would limit off-target hypertension. Notably, the discovery of an intravenously infused oxyferrous hemoprotein that does not trigger hypertension represents a significant advancement in the field of hemoprotein therapeutics, and we are directing future work towards the development of RcoM variants as potential non-globin-based, oxygen-carrying blood substitutes.

RcoM-HBD-CCC administered at a screening dose was well-tolerated by healthy (non-CO poisoned) animals. Organ-specific blood chemistry markers for liver and kidney damage were nominal after 48 hours, and these results were comparable to those observed in preliminary safety studies of StHb (23). Western blotting of organ tissues using a 6xHis specific antibody indicates minimal accumulation of scavenger in lung, liver, spleen, and kidney after 48 hours. These results represent a significant improvement in safety and tolerability over Ngb-H64Q-CCC, which was well-tolerated in CO-poisoned animals, but whose safety was not investigated in healthy, non-CO poisoned animals (22). Administration of Ngb-H64Q-CCC bearing Fe(III) heme caused kidney damage, suggesting a mechanism involving adverse heme reactivity, such as heme release or heme-facilitated reactive oxygen species generation (64–66). We note here that off-target hypertensive effects of Ngb-H64Q-CCC were not quantified in our prior studies as all reported hemodynamic data for Ngb-H64Q-CCC infusion were recorded in the context of severe CO poisoning (30–95% HbCO). Under these conditions, CO replaces O₂ at all Ngb-H64Q-CCC heme sites within 1.2 minutes and thereby prevents oxidative depletion of tonic NO, which does not react with Fe(II)-CO heme (22). In our current study, we find that infusion of oxyferrous Ngb-H64Q-CCC does elicit hypertension in non-CO-poisoned mice (Figure S7). We further observe a strong correlation between hypertension and the apparent rate of NO dioxygenation (Figure 5G). Taken together, our safety and tolerability data for RcoM-HBD-CCC suggest that this protein precludes off-target oxidative reactivity regardless of CO binding status. RcoM-HBD-CCC infusion did not elicit hypertension, and the scavenger was safely excreted into the urine within hours of intravenous infusion in the absence of CO. Further dose range finding studies to assess the complete therapeutic window for RcoM-HBD-CCC are currently in progress.

In conclusion, we have successfully engineered a high-affinity, CO scavenging hemoprotein, RcoM-HBD-CCC, based on the heme-binding domain of the bacterial CO sensor, RcoM. Ligand binding kinetics and reactivity data reveal remarkable selectivity for CO over oxygen, as well as limited reactivity towards NO. These biochemical properties are consistent with a distal heme pocket that evolved to facilitate exclusive interactions between CO and heme through hydrophobicity, restricted solvent access, and/or positional restriction of heme-bound ligands. Intravenous infusion of RcoM-HBD-CCC during acute inhaled CO poisoning accelerated clearance of CO from circulating RBC Hb, and CO-bound scavenger was rapidly excreted in urine. The biochemical selectivity of RcoM-HBD-CCC translated to minimal off-target reactivity in heathy animals and in the context of CO poisoning. In contrast to globin-based hemoproteins, RcoM-HBD-CCC infusion did not induce hypertension, nor did RcoM induce organ-specific toxicity. These data demonstrate that RcoM-HBD-CCC may act as a safe and efficacious treatment for acute CO poisoning. This study also underscores the importance of heme pocket structure and amino acid composition in dictating small molecule reactivity.

Materials and Methods

Statistical analyses were performed with GraphPad Prism 9.0 using methods described in the figure legends. Detailed materials and procedures for the design and cloning of expression constructs, heterologous production of recombinant proteins, spectroscopic analysis, determination of rate constants and binding affinities, ultrafast kinetics experiments, in vitro modeling of CO scavenging, and animal studies can be found in the SI Appendix. In vivo models of acute CO poisoning were adapted from prior reports (22, 23). All animal studies were performed using protocols approved by the Institutional Animal Care and Use Committee at the University of Pittsburgh and in accordance with National Institutes of Health guidelines.

Significance Statement.

Carbon monoxide (CO) poisoning remains the most common non-drug related poisoning worldwide. Despite a high incidence of mortality and long-term sequelae for survivors, no specific antidotal therapy is available to treat CO poisoning. This study employs a novel hemoprotein engineering strategy to design a high affinity CO sequestration agent with previously unattainable ligand selectivity that improves CO clearance and limits hypertension in a murine model of acute CO poisoning.