2017 MILITARY SUPPLEMENT: CURRENT CHALLENGES IN THE DEVELOPMENT OF ACELLULAR HEMOGLOBIN OXYGEN CARRIERS BY PROTEIN ENGINEERING

Abstract

This article reviews the key biochemical mechanisms that govern O₂ transport, NO scavenging, and oxidative degradation of acellular hemoglobin and how these ideas have been used to try to develop strategies to engineer safer and more effective hemoglobin-based oxygen carriers (HBOCs). Significant toxicities due to acellular hemoglobin (Hb) have been observed after the administration of HBOCs or after the lysis of red cells, and include: (a) rapid clearance and kidney damage due to dissociation into dimers, haptoglobin binding, and macrophage activation; (b) early O₂ release leading to decreased tissue perfusion in capillary beds; (c) interference with endothelial and smooth muscle signaling due to nitric oxide (NO) scavenging; (d) autooxidization of heme iron followed by production of reactive oxygen species; and (e) iron overload symptoms due to hemin loss, globin denaturation, iron accumulation, and further inflammation. Protein engineering can be used to mitigate some of these side effects but requires an in-depth mechanistic understanding of the biochemical and biophysical features of hemoglobin that regulate quaternary structure, O₂ affinity, NO dioxygenation, and resistance to oxidation, hemin loss, and unfolding.

INTRODUCTION

The objective of this article is to review the biochemical strategies that have been used to try to engineer safer and more effective acellular hemoglobin-based oxygen carriers (HBOCs). In the preceding article on the toxicity of hemoglobin-based oxygen carriers (HBOCs), Alayash (1) reviewed the history, development, and side effects of some key commercial products, with an emphasis on oxidative reactions of the heme group. Our focus is on the fundamental biochemical mechanisms that govern clearance, O₂ transport, NO scavenging, and denaturation of acellular hemoglobin, all of which determine the physiological activity and toxicities of HBOC products. These processes are described in molecular detail to allow the development of strategies to inhibit the reactions that lead to serious side effects.

The use of hemoglobin-based transfusion alternatives to donated blood has been explored since the 1930s (2–4). The development of these blood substitutes has matured along with a greater understanding of the chemistry and physiology of erythrocytes and hemoglobin. Initial tests focused on replacing whole blood in animal models with acellular hemoglobin-based oxygen carriers (HBOCs). Identification of the different physiological roles of whole blood informed the requirements for distinct transfusing agents. For example, saline solutions were identified early on as substitutes for the volume and isotonic properties of blood that are important for maintaining heart pumping. When the role of hemoglobin in oxygen transport was fully established, it became apparent that acellular hemoglobin could be used for efficient oxygen transport during transfusion (3).

The intrinsic value of acellular blood substitutes is that these products can potentially eliminate clinically significant problems associated with using donated blood (5, 6). First, acellular hemoglobin solutions do not contain antigens and require no blood typing. Second, production of blood substitutes can be done in vitro under sterile laboratory conditions, reducing or eliminating the risk of transmitting disease or infection. Third, the shelf life of most acellular HBOCs is dramatically longer than that of donated blood (in some cases years versus one month), which is crucial for production, storage, and transport to the scene of trauma. Fourth and perhaps most important, large supplies of HBOCs can be maintained and stockpiled, eliminating shortages and transfers between institutions during large-scale emergencies.

Unfortunately, there are significant problems associated with the use of acellular hemoglobin products (7–12). Initial studies with acellular hemoglobin began in the early 20th century. In these experiments, hemoglobin was purified from cell lysates and injected directly into patients (2, 4). Purified acellular hemoglobin was able to provide oxygen delivery without antigenic reactions, but not without significant side effects. Injection of unmodified acellular hemoglobin into animals and humans resulted in the rapid appearance of high levels of hemoglobin in urine, hypertension, and renal failure (2–4).

DISSOCIATION OF ACELLULAR HEMOGLOBIN INTO DIMERS

A summary of the causes of plasma hemoglobin toxicity is given in Fig. 1. A list of some of the acellular hemoglobin based products, most of which have undergone human trials, is given in Table 1 along with some of their properties. In general, it was difficult to obtain blood pressure effects, autooxidation rates, and hemin loss kinetic parameters under comparable conditions for all these products. As result care should be taken when making comparisons and the primary literature in the footnotes should be examined.

Fig. 1.

Possible causes of acellular hemoglobin toxicity

Table 1. Summary of O₂ binding, blood pressure effects, autooxidation, and hemin loss properties of HBOC products that have been used in human trials.

Except for the P₅₀ values, most of these parameters are under reported, particularly the rates of autooxidation and hemin loss. Thus, the values for the last four columns may be approximate. The companies should be consulted for exact values. ΔMAP represents the increase in mean arterial pressure in 10% topload or 50% isovolemic exchange experiments with rats or guinea pigs. The rate constants for autooxidation and hemin loss were normally measured at 37 °C, near physiological pH. NR indicates not reported.

Hemoglobin Product Name (other names)	Modification	Company	Oxygen P50 mmHg	ΔMAP mmHg (Top load)	kautox initial h−1	k-H (β) h−1	k-H (α) h−1
Stroma-free HbA Tetramer (Dimer)	None		8a	≥25b	~0.001c (~0.03)	1.5 (15)c	0.3 (0.6)c
HemAssist@ (DCLHb, ααHb)	Human HbA chemically crosslinked tetramer	Baxter International	32d,e	~30f	NR	NR	NR
Polyheme@ (SFH-P)	Glutaraldehyde crosslinked HbA with covalently attached pyridoxal-phosphate	Northfield Laboratories	20 – 22d, 26–30e	30f	0.007g	8.1g	1.2g
Hemopure@ (HBOC-201)	Glutaraldehyde crosslinked bovine Hb	Biopure	32–38d, 40e	15h	NR	4.3i	0.4i
Hemospan@ (MP4)	Pegylated human HbA	Sangart	10e	≤5j	0.021g	40g	1.3g
Optro1.1@	Crosslinked Recombinant HbA with β N108K mutation	Somatogen	33k	≥20k	NR	52g	1.5g
Sanguinate@	Pegylated bovine Hb	Prolong	12l	≤10m	NR	NR	NR
rHb2.0 (rHb3011)	Crosslinked Recombinant HbA with (βV67W-αH58Q/L28W)	Baxter Hemoglobin Therapeutics	31–33n	≤5n	0.41o	90o	90o

^aMaillet et al. 2008 (38)

^bMcCarthy et al. 2001 (84)

^cThe initial k_autox value for native HbA was determined in this report. The values of k_-hemin for tetrameric (and dimeric) HbA were taken from Hargrove et al.1997 (77)

^dChen et al. (85)

^eNatanson et al. (86)

^fGulati et al.1995 (87)

^gVandegriff et al. 2006 (88)

^hKatz et al. 2010 (89)

ⁱUnpublished data from Olson, J.S. and Light, W. R., 2008.

^jVandegriff et al.2003 (90)

^kRaat et al. 2005 (47)

^lAbuchowski 2017 (91)

^mAbuchowski 2010 (92)

ⁿHermann et al. 2007 (93)

^oVarnado et al. 2013(5)

Bunn et al. (13, 14) identified the dissociation of tetrameric hemoglobin into dimers as the probable cause of renal toxicity and rapid clearance. Hemolysis due to red cell aging, chronic inflammation, and exposure to chemical oxidants shows similar toxicities (15, 16). When human adult hemoglobin is diluted into plasma, the tetramers dissociate into dimers, which can pass through the glomerular filters. Alternatively, the dimers can rapidly bind to haptoglobin to form complexes that are specifically taken up and degraded by macrophages through a pathway involving the CD163 receptor and stimulation of heme oxygenase activity. Schaer, Alayash, and their colleagues have also shown that CD163 can also take up hemoglobin alone (albeit at a much higher concentration), and that these interactions may reduce oxidative damage by acellular hemoglobin, although this effect has not been shown directly (17–19).

To prevent dissociation, the first generation of acellular hemoglobins products used chemical crosslinking methods to polymerize hemoglobin and prevent dimerization. Although some of these polymerized and crosslinked hemoglobins led to a reduction in side effects, there were still issues with hypertension and reduced cardiac output (see (20) or reviews (8–10, 12, 21) and Table 1). One of the initial roadblocks for optimization of the chemically crosslinked hemoglobins was the variability of unmodified, non-crosslinked hemoglobin tetramers in the preparations. This problem was solved by Biopure, Inc. (Hemopure^@) and Northfield Laboratories (Polyheme^@) in the late 1980s by reproducibly removing virtually all non-crosslinked Hb tetramers from their polymerized samples ((22, 23) and Table 1).

The second generation of recombinant hemoglobin oxygen carriers used more precise chemical reactions or genetic manipulation to produce cross-linked tetrameric HbA variants. The first simple hemoglobin tetramer product contained α chains crosslinked with bis-(3,5-dibromosalycil) fumarate between two α Lys99 side chains (HemAssist^@, Baxter, Table 1) (24). Several years later, recombinant DNA technology was used to fuse two α globin genes with a glycine codon linker to create a di-α subunit that associated with β subunits to create a crosslinked recombinant hemoglobin tetramer, called rHb0.1 (25). However, tests with these tetrameric HBOCs still revealed significant vasoconstrictive effects and renal toxicity (Table 1 and (8, 21, 26)). As with the polymerized HBOCs, the circulation half-lives of these simple crosslinked tetramers were significantly increased, and no hemoglobin was excreted in urine. Marden, Ho, and their coworkers constructed and evaluated a recombinant hemoglobin that spontaneously becomes an octamer by formation of two inter-tetramer disulfide bonds between Cys side chains introduced at the β Gly83 and α Asn78 positions (27).

Genetic crosslinking of the hemoglobin genes and their expression in E. coli allowed the use of rational protein engineering approaches to determine the underlying causes of the side effects associated with acellular HBOCs (28, 29). Previously, Varnado et al. (5) have reviewed the methodology used to express and purify recombinant HBOCs in bacteria, the problems associated with this technology, and various attempts to mitigate the side effects of acellular Hb. In this paper, we review the mechanisms and strategies for mitigating the key processes that are associated with plasma hemoglobin toxicity (numbered 1–4 in Figure 1).

OXYGEN BINDING AND TRANSPORT

Optimization of ligand binding parameters is critical to mimic the oxygen transport properties of intact red blood cells. One of the first differences noted between extracellular and intracellular hemoglobin was the increased O₂ affinity of the cell-free protein due to the loss of intracellular 2,3-bis-phosphoglycerate, which decreases oxygen affinity inside erythrocytes. Chemical and genetic crosslinking also often reduces P₅₀ (the partial pressure of O₂ when hemoglobin is half saturated). The efficiency of oxygen transport is governed primarily by the release of O₂ from hemoglobin and diffusion into actively respiring tissues (i.e. heart muscle, brain tissue, etc.) The diffusion gradient at 50% unloading is proportional to the P₅₀ of the HBOC, and as a result, the efficiency of transport increases for hemoglobins with lower O₂ affinity, as long as the P₅₀ is low enough for complete uptake in the lungs (29–31).

In the case of red cells flowing through capillaries, there is additional diffusion resistance due to both unstirred plasma layers around the red cell and cell free layers along the capillary surface. These resistances do not occur when the hemoglobin is present in the plasma phase, and as a result, the efficiency of O₂ transport increases 2–3 fold for acellular HBOCs (29, 32–35). A consequence of this enhanced delivery is that the dosage of acellular hemoglobin can be 2–3 times lower than that for whole blood (i.e. ~ 4–7 g/dl of acellular Hb versus 10–14 g/dl Hb in whole blood). However, Winslow and colleagues have argued that this difference leads to premature O₂ unloading in arterioles and less delivery to capillary beds (21, 36). To avoid this problem, he and his colleagues proposed that acellular HBOCs should have a higher oxygen affinity than what is needed for efficient delivery by red cells and argued that Hemospan^@ or MP4 was a more efficient HBOC because of its lower P₅₀ (Table 1). Regardless of whether this idea is correct, it is relatively straightforward to engineer the oxygen affinity of Hb-based blood substitutes.

The basic biochemical mechanisms that govern O₂ affinity in human hemoglobin are well understood as outlined in Fig. 2 and as described by Olson and coworkers (37–39). First, iron reactivity is governed by the ease of movement of the Fe(II)-proximal His(F8) complex into the plane of the porphyrin ring, which in turn is regulated by the R (high affinity) to T (low affinity) quaternary conformational transition. Allosteric effectors that bind and stabilize the T quaternary conformation can raise P₅₀. This strategy was employed to raise the P₅₀ of three initial products: Polyheme^@ by covalently attaching pyridoxal phosphate (a 2,3-bisphoshoglycerate analogue) (Table 1 and (23)); Hemopure^@ by using bovine Hb, which binds Cl⁻ very tightly in the T state (22); and HemAssist^@ by chemically crosslinking HbA in the deoxy or T state conformation (24). This approach has also been taken with genetically crosslinked rHb, using naturally occurring mutations that favor the low affinity T state (i.e. Hb Presbyterian (β N108K) in Optro1.1^@ (25)). O₂ affinity can also be adjusted by distal pocket mutations which alter the strength of hydrogen bonding to bound O₂ from the polar E7 side chain (39) or enhance steric hindrance by the apolar B10, E11, or G8 amino acids that line the back of the distal pocket (40) (see Fig. 2, where the amino acid positions are given as residue number from the N-terminal position along one of the 8 helices labeled A through G). Many examples of these types of allosteric and active site substitutions occur in nature and serve to guide protein-engineering efforts (37–39). One of the more interesting studies involved examining substitutions in recombinant woolly mammoth hemoglobin that allowed the protein to maintain efficient oxygen transport during exposure to very low temperatures at the animal’s extremities (41, 42). Thus, it is relatively straightforward to chemically or genetically engineer HBOCs with altered O₂ affinities and rate constants.

Fig. 2. Factors governing O₂ binding in human adult hemoglobin (HbA).

The crystal structure of the oxygenated form of the tetramer and the active site of an α subunit in adult human oxyhemoglobin (PDB access code 2DN1).

NO SCAVENGING AND HYPERTENSION

The most serious side effects of acellular hemoglobins involve oxidative processes that lead to methemoglobin formation and include autooxidation and the reaction of nitric oxide (NO) with oxyhemoglobin (HbO₂). For example, increases in systemic blood pressure due to acellular hemoglobin (Hb) have been observed in hemolytic diseases, transfusions with old blood, and administration of extracellular HBOCs (Table 1, column 5 (8, 15, 16). In the early 1990s, there was considerable discussion of what role extracellular hemoglobin played in these hypertensive events. However, currently almost all workers in the field agree that scavenging of nitric oxide (NO) by hemoglobin inhibits vasorelaxation, which is mediated by the NO complex of soluble guanylyl cyclase (sGC) in blood vessel walls. The key reaction involves dioxygenation of NO by O₂ bound to reduced hemoglobin and produces nitrate (NO₃⁻) and methemoglobin (26, 43). As shown in Fig. 3, NO enters the active site of an oxyhemoglobin subunit (HbO₂) and quickly reacts with the bound O₂ to generate transient intermediates that rapidly dissociate into oxidized hemoglobin (metHb) and nitrate. The overall reaction is very fast (μs to ms), irreversible, and its speed is determined by the rate of NO diffusion into the active site of the protein. Some direct NO binding to deoxyhemoglobin does occur, but it is very minor because most of the plasma and red cell hemoglobin is oxygenated in arteriole blood. In addition, the NO that is bound reversibly to deoxyHb will slowly dissociate and then be dioxygenated by surrounding HbO₂ molecules. Thus, NO scavenging is due exclusively to irreversible dioxygenation, producing NO₃⁻ and metHb.

Fig. 3. Four key reactions involved in NO scavenging and oxidative degradation of human hemoglobin.

The reactions of H₂O₂ are also shown, and these processes are described in detail by Alayash and coworkers (paper that is part of this series).

There are three strategies for inhibiting NO scavenging. The first strategy is to encapsulate the hemoglobin in a cell or vesicle, which prevents the HbO₂ from getting up to and into endothelial cells (44, 45). To be consumed NO has to diffuse through cell free plasma layers adjacent to the blood vessel walls and through unstirred layers adjacent to the cell surface. In contrast acellular hemoglobin flows up to and into the blood vessel walls where it can directly interfere with NO signaling between endothelial and smooth muscle cells. A second strategy for reducing NO scavenging is to inhibit acellular Hb extravasation into the endothelium by increasing the size of the hemoglobin molecule or coating its surface with polyethylene glycol polymers (46). Various groups have shown that increasing the size of the HBOC reduces the extent of the hypertensive side effect (see Fig. 7 in Olson et al. (26)), and PEGylation of acellular Hb by itself also inhibits elevation of mean arterial blood pressure (36). These effects explain why the polymerized HBOCs listed in Table 1 (i.e. Polyheme^@ and Hemopure^@) have lower hypertensive side effects (increases in mean arterial blood pressure, ΔMAP, in Table 1) than the simple crosslinked tetramers (Hemassist (DCLHb or ααHb) and Optro(rHb1.1)) and why the pegylated HBOCs (Hemospan^@ or MP4 –pegylated human Hb) and Sanguinate^@ (pegylated bovine Hb) also have lower blood pressure effects.

The third strategy for decreasing NO scavenging is to reduce the bimolecular rate of NO dioxygenation by genetic engineering. In the late 1990s, the group at Somatogen (later Baxter Hemoglobin Therapeutics (28)) introduced large aromatic amino acids (Phe, Trp) at the B10, E11, and G8 helical positions in the α and β subunits of recombinant hemoglobin (rHb) to decrease the rate of NO entry by preventing its capture in the back of the distal portion of the heme pocket (see Fig. 2). They created a library of rHbs with bimolecular rates of NOD dioxygenation (k′_NOD values) that decreased from ~ 80 μM⁻¹s⁻¹ to ~ 4 μM⁻¹s⁻¹, and observed corresponding 10 to 20-fold decreases in the mean arterial pressure (MAP) effect in experiments using an isovolemic exchange transfusion protocol in rats (28). These results were later extended to total peripheral resistance (TPR) measurements (26, 29). The observed increases in MAP or total peripheral resistance (TPR) only correlated with the rate of NO dioxygenation and not with the oxygen affinity of the genetically crosslinked HBOC variants (26, 29). These studies led to the development of a product called rHb2.0 by Baxter Hemoglobin Therapeutics, Inc., which contained α Leu(B10)Trp/His(E7)Gln and β Val(E11)Trp mutations. rHb2.0 had little or no hypertensive side effect in transfusion experiments with animal models (Table 1 and (47)). However, as described in Varnado et al. (5), this product oxidized and denatured relatively quickly and was not successful. This failure points out the need to consider all the key properties at the same time when developing engineering strategies, i.e. the rate of NO dioxygenation needs to be reduced without decreasing resistance to oxidative degradation. The rapid reaction of HbO₂ with free NO also suggests that efforts to mitigate the hypertensive side effect of HBOCs by simply adding or attaching NO releasing agents are unlikely to be successful because the heme bound O₂ will rapidly react with the free gas, oxidizing the hemoglobin and generating nitrate.

AUTOOXIDATION

Many workers in the HBOC field have argued that oxidative degradation of acellular hemoglobin is the major cause of the observed side effects, particularly those associated with endothelial damage and inflammation caused by oxidative stress from reactive oxygen species (8, 9). Similar pathological effects are associated with transfusions using old blood and with hemolytic diseases. Spontaneous autooxidation of oxyhemoglobin generates both methemoglobin and superoxide, which rapidly dismutates to hydrogen peroxide (H₂O₂), a very reactive oxygen species. As shown in Fig. 3, the resultant H₂O₂ can in turn react with either reduced or oxidized hemoglobin to produce even more reactive protein radicals that damage both hemoglobin itself and nearby proteins and membrane components (48).

In red blood cells, reductases quickly reverse the oxidation of intracellular hemoglobin, and superoxide dismutase and catalase efficiently remove O_2•⁻ and H₂O₂ (Fig. 1). However, no superoxide or catalase is present in plasma, and as a result, these reactive oxygen species lead to cyclic redox reactions that can produce free radicals, which cause significant damage to hemoglobin itself, plasma proteins, blood vessels, and surrounding tissues. H₂O₂ can react with either reduced deoxyHb (FeII) or oxidized metHb (FeIII) to produce highly reactive ferryl (FeIV) species and protein radicals. These ferryl forms can be reduced back to ferric forms by cyclic pseudoperoxidase activities, which sustain the generation of radicals and ROS (48, 49). These cyclic redox reactions appear to account in part for oxidative damage to the kidneys, where the acellular hemoglobin tends to accumulate.

In recent reviews, Alayash (1, 10, 50, 51) has described the reactions of acellular Hb with H₂O₂, and Cooper and his colleagues at the University of Essex have described various approaches to mitigate these processes including mutations in β chains to enhance reduction of radicals (52–55). In this paper, we have focused in much greater detail on the initial autooxidation reaction, which is the key rate-determining step in the oxidative degradation of hemoglobin. Ironically, most workers in the hemoglobin field avoid studying autooxidation because of the slowness of the reaction, the instability of the reaction products, and the complexity of the mechanism (missing numbers in Table 1). However, in our view inhibition of this reaction, even if only 2 to 3-fold, would generate a much more stable, effective, and commercially viable product. Thus, we have presented below a detailed molecular description of this process in the hope that the proposed mechanisms will lead to engineering strategies to reduce the rate of autooxidation of acellular hemoglobin products.

Many of the important phenomenological features of autooxidation of oxyhemoglobin and oxymyoglobin were established in the 1930s using van Slyke Apparatus manometric methods to control O₂ pressure and measure O₂ binding capacity (56–58). However, attempts at detailed biochemical interpretations were not made until many years later (59–61). Early studies by Brooks on hemoglobin (57) and George and Stratmann (62) on myoglobin showed that the rate of autooxidation, k_autox, has a “bell” shaped dependence on the concentration of dioxygen, [O₂], with a maximum k_autox at the P₅₀ of the globin sample (i.e. [O₂]_free at 50% saturation) and then a smaller, limiting value of k_autox equal to 0.1 to 0.02 h⁻¹ at very high [O₂] (Fig. 4). They also showed that k_autox dramatically increases with decreasing pH in the range from 9 to 5.

Fig. 4. Dependence of the rate of autooxidation of bovine oxyhemoglobin on [O₂] in 0.6 M phosphate at pH 5.7, 30° C.

Data were taken from Brooks (57) and plotted versus [O₂] in μM instead of P_O2 in mmHg. The dashed line in the main figure is an empirical fit to the combined mechanism shown in Fig. 5, using the equations derived in Brantley et al. (61) for sperm whale myoglobin. The dependence of the bimolecular rate of oxidation and the unimolecular rate of oxidation on [O₂] are shown in the inset as gray dashed and solid lines.

Roughly 60 years later, Brantley et al. (61) examined the factors that governed autooxidation of mammalian myoglobin and proposed the complex scheme shown in Fig. 5 based on work with a large library of recombinant mutants, previous studies by Caughey and coworkers (59, 63) on the effects of cyanide and azide on the rates of autooxidation of globins, and theoretical considerations of direct superoxide dissociation, originally discussed by Weiss in 1964 (64). The outer valence electron in ferrous iron can be partially transferred to bound dioxygen to produce a complex that resembles low spin ferric iron and bound superoxide (O₂•⁻). If protonated, the bound O₂•⁻ can dissociate as the neutral hydrosuperoxide and then rapidly migrate out of the distal pocket at rates similar to those of the neutral diatomic gases (~ 10⁷s⁻¹ (65)). This unimolecular process is shown as Mechanism 1 on the right hand side of Fig. 5. Under physiological conditions, the rate of this process will be slow because the extent of protonation of bound O₂ is very small and the probability of thermal dissociation of HO₂• is also very low. This mechanism predicts a hyperbolic dependence of k_autox on [O₂], which is identical to that for simple reversible O₂ binding, and a limiting value at high [O₂] equal to the thermal rate of HO₂• dissociation, i.e. a value of ~ 0.04 h⁻¹ for the results obtained by Brooks (57) in Fig. 4.

Fig. 5. Mechanisms involved in the autooxidation of HbO₂.

This scheme was taken from Brantley et al. (61) where the mechanism of autooxidation of myoglobin was examined quantitatively.

The mechanism in Fig. 5 also explains the strong inverse dependence of k_autox on pH, with the rates increasing several-fold for each decrease in pH unit from ~ pH 8 to pH 6 (56, 57, 60, 61). The pKa for superoxide in solution is roughly 4.9 so that at neutral pH the extent of protonation is very small for the free superoxide anion and perhaps even lower when it is bound to iron (61). This scheme also explains the key role of the distal histidine in inhibiting autooxidation by forming a strong hydrogen bond with the Fe-O₂ complex. Hydrogen bonding to His(E7) blocks protonation of the bound ligand by lowering the effective pKa even further. When His(E7) is replaced with apolar amino acids, the rates of autooxidation of mammalian Hbs and Mbs increase markedly, by factors of 10 to 1000, depending on the size of the amino acid side chain (61).

The unimolecular, HO₂• dissociation mechanism cannot explain the marked 3 to 4-fold increase in k_autox as [O₂] decreases into the region of the hemoglobin’s P₅₀ value (Fig. 4). In 1982, Wallace et al. (59) proposed a mechanism similar to the bimolecular process shown in the middle of Fig. 5. This second mechanism involves transient formation of a 6-coordinate ferrous complex in which an external ligand or nucleophile binds on the distal side of the heme group and pulls the iron atom into the plane of the porphyrin ring. Under these conditions, outer sphere electron transfer to an acceptor can occur without a large quantum mechanical barrier and forms a stable Fe(III) complex with the proximal His(F8) and the nucleophile (59). In the case of autooxidation, the acceptor is dissolved O₂ which reacts with the edge of the porphyrin ring to produce O₂•⁻ directly. In this mechanism, no autooxidation can occur until some deoxyhemoglobin is formed to react with either an added nucleophile like azide or a solvent water molecule. For this mechanism, k_autox is zero at high [O₂] when no deoxyHb is present and at very low [O₂] when no electron acceptor is present. The maximum rate will occur at ~ 50% saturation. This mechanism explains why the observed rate of autooxidation shows a “bell-shaped” dependence on [O₂] and why the observed rate at low [O₂] increases when strong ferric ligands such as azide, cyanide, and imidazole are added.

Hargrove and coworkers (66) have also pointed out that this mechanism explains why autooxidation of naturally occurring hexacoordinate plant and microbial hemoglobins is so fast. In these cases, the distal histidine is binding to the iron atom in the ferrous state, inhibiting reversible O₂ binding and, at the same time, speeding up electron transfer to solvent O₂ and other electron acceptors. In myoglobins and hemoglobins, the structure of the distal pocket evolved to keep the distal histidine away from the iron atom in the native tertiary structure to prevent rapid autooxidation by this bimolecular mechanism, but still close enough to form a strong hydrogen bond to prevent both simple dissociation of the bound O₂ and protonation to facilitate the dissociative mechanism.

As shown in Fig. 4 and discussed in Brantley et al. (61) both mechanisms are occurring in hemoglobins and myoglobins. If the P₅₀ of the hemoglobin is low, then autooxidation occurs primarily by the dissociative process in air or at higher [O₂]. However, if the P₅₀ is high, the rate of autooxidation will have a greater contribution from the bimolecular process. A detailed discussion of the effects of mutagenesis on these processes in Mb is given in Brantley et al. (61), and we are currently working on trying to apply those ideas to studies with recombinant Hbs.

Autooxidation of hemoglobin has three additional complications compared to monomeric myoglobin, all of which are physiologically relevant. First, the rate of autooxidation, k_autox, of human adult oxyhemoglobin depends dramatically on total protein concentration. This observation was first made by Zhang and Rifkind (67) and then quantitated more thoroughly by Mollan et al. (68). An example of the dramatic increase in k_autox at low protein concentration is shown in Fig. 6 using data that we have recently collected. The observed k_autox value becomes 30 to 40-fold higher at very low concentrations where human oxyhemoglobin dissociates into dimers. Mollan et al. (68) analyzed this dependence quantitatively, using a tetramer to dimer dissociation constant K_4,2, ≈ 1 μM, and obtained fitted limiting values of 0.01 h⁻¹ for tetramers (upper gray line, Fig. 6) and 0.4 h⁻¹ for dimers (lower gray line, Fig. 6).

Fig. 6. Effect of quaternary structure on autooxidation of human hemoglobin in 0.1 M Na phosphate, pH 7.0, 37° C.

rHb0.1 is genetically crosslinked human hemoglobin containing a glycine linker between the C terminus of one α chain and the N-terminus of the other α chain. The points represent triplicate measurements and the fraction of HbO₂ remaining was obtained from deconvoluting the observed spectra into components representing HbO₂, metHb, hemichromes, and light scattering due to precipitation (83) The upper and lower gray lines represent theoretical time courses for completely tetrameric and dimeric HbA, respectively, based on the analysis of Mollan et al. (68).

The second complication is that the autooxidation of Hb tetramers could potentially be cooperative, with oxidation of the first few subunits facilitating the oxidation of the remaining ones. In the absence of ligands, equilibrium redox curves for human HbA do show cooperativity (69), and the time course we observe for autooxidation of almost fully tetrameric human HbA at high protein concentrations shows some acceleration (Fig. 6). However, no one has yet attempted to analyze the extent of the increase in rate in terms of an allosteric model, and this acceleration is lost at lower protein concentrations (≤ 100 μM heme) where most experiments are done and significant levels of non-cooperative HbO₂ dimers are present. However, the important observation is that the initial rate of HbA autooxidation is small and on the order of 0.001 h⁻¹ for tetramers at 37 °C, pH 7 and underestimates the speed of the overall reaction (Table 1 and Fig. 6).

The third potential complication is differences between the autooxidation rates of the α and β subunits. Shikama and coworkers (70) have argued that the time course for HbA autooxidation is biphasic and that the α subunits autooxidize more rapidly. In unpublished work, we have looked for differences between the autooxidation rates of the subunits by reacting partially oxidized samples of HbA with azide and measuring the relative fractions of oxidized β versus α subunits. Because met-β subunits react ~ 5 times more rapidly with azide than met-α, the fraction of subunits that react rapidly with azide provides the ratio of met-β/met-α. This approach was used by MacQuarrie and Gibson (69) to show that β subunits have a more positive reduction potential than α subunits, implying that the α subunits might more readily oxidize. However, under the conditions shown in Fig. 6, the ratio met-β/met-α was unchanged throughout the autooxidation time course for HbO₂ at 100 μM. In addition, if the subunits were significantly different, the overall time course should have been biphasic and not accelerating under these conditions. At lower protein concentrations, the overall time course does become exponential and sometimes biphasic, and we do see evidence that the α subunits autooxidize more rapidly (≤ 2.5-fold), but the differences are small.

The key issue for this review is whether or not the mechanisms and results shown in Figs. 4–6 can be used as a framework to engineer HBOCs with greater resistance to autooxidation. Most HBOC preparations contain crosslinked hemoglobin to prevent dissociation into dimers, which was done initially to prevent clearance by the kidneys. Thus, many of these stabilized tetramers or Hb polymers actually do show smaller k_autox values than HbA at low protein concentrations because dissociation into rapidly oxidizing dimers is prevented (i.e. 10 μM rHb0.1 in Fig. 6 compared with 10 μM HbA and see values of k_autox for the HBOCs in Table 1). However, at higher more physiological globin concentrations, all of these crosslinked hemoglobin preparations show larger rates of autooxidation than native HbA tetramers due to the effects of the mutations or chemical modifications used for crosslinking (i.e. compare 100 μM rHb0.1 with 100μM HbA, the upper gray curve in Fig. 6 and initial k_autox values in Table 1).

Both mechanisms shown in Fig. 5 require loss of hydrogen bonding from the distal histidine to facilitate either O₂ dissociation into deoxyHb or to allow protonation of bound O₂ and dissociation of HO₂•⁻. Autooxidation can be slowed dramatically by inserting more hydrogen bond donating amino acids in the active site as is seen in some invertebrate and bacterial hemoglobins. However, these proteins have such high O₂ affinities that they are not capable of O₂ transport (71, 72).

A better strategy is to try to limit solvent accessibility to the distal pocket because water is required, either to combine transiently with deoxyHb to facilitate the outer sphere bimolecular process or to act as a proton donor to the bound O₂ in the unimolecular mechanism. We were successful in reducing the rate of autooxidation of Mb roughly 5-fold by inserting a Phe for Leu(B10), which limited water penetration into the active site (73). However, this same strategy has not worked for human HbA. So far no one has been able to construct a mutant or chemically modified Hb with a lower rate of autooxidation than native tetrameric human HbA (i.e. see (74, 75)).

In our view, hemoglobin autooxidation is governed by transient water penetration into the distal pocket, which is dampened when more rigid quaternary states occur. This idea would explain the dramatic ~ 100-fold decrease in k_autox observed in going from isolated subunits (~ 1–2 h⁻¹ at 37° C) to dimers (~ 0.4 h⁻¹) to tetramers (≥ 0.01 h⁻¹). Thus, mutations that dampen fluctuations in the tertiary structures of tetrameric subunits need to be considered but are not obvious from static crystal structures. Choices of regions to mutate or modify will require both computational analyses and screening of large libraries of variants with changes in regions at interfaces and turns between helical regions (i.e. CD, EF, and FG corners). Finally the results in Figs. 4 and 6 also indicate that the autooxidation characteristics of potential HBOCs require studies as a function of O₂ concentration and pH. Maximum rates of autooxidation in vivo will occur in capillaries adjacent to actively respiring muscle and neuronal tissues where [O₂] approaches the P₅₀ of the product and the blood pH is lowered.

HEMIN DISSOCIATION

The third key process in the oxidative degradation of hemoglobin is hemin dissociation (Figs. 1 and 3). This process leads to irreversible precipitation of apoglobin, incorporation of the redox active hemin into nearby serum proteins and endothelial membranes, and self-aggregation of the hemin to form insoluble particulates. These three reactions prevent reformation of HbO₂ from metHb even when reducing agents are added. MetHb can be reduced back to HbO₂ or deoxyHb if the hemin is still bound by adding ascorbate, glutathione, or other naturally occurring reducing agents. In red cells, autooxidation is reversed by methemoglobin reductases using NADPH, and catalase and superoxide dismutase to prevent an accumulation of H₂O₂ (Fig. 1). Thus, another key engineering goal in the HBOC field is to reduce the rate of hemin dissociation.

As shown in Fig. 7B, dissociation of heme requires hydrolysis of the Fe-His(F8) bond and then movement of the Fe-protoporphyrin IX out of the protein. The strength of the Fe-His(F8) bond depends on the oxidation state of the iron atom and the presence or absence of bound ligands. Heme does not dissociate readily when the iron is reduced and is completely inhibited in the presence of bound ligands, particularly CO. The ferrous forms of myoglobin and hemoglobin have equilibrium heme dissociation constants ≤ 10⁻¹⁶ M (76). In contrast, the equilibrium constants for hemin dissociation from the aquo-ferric forms of these proteins are one hundred to a million-fold higher and on the order of 10⁻¹⁴ to 10⁻¹⁰ M (76, 77). In effect, autooxidation has to occur before the heme group can dissociate.

Fig. 7.

A. Time courses for hemin dissociation from various quaternary states of HbA and HbA dimers bound to haptoglobin (Hp) 1.1. The time courses for isolated α and β subunits, metHbA dimers, metHbA, and crosslinked met-rHb0.1were computed using the rate constants determined by Hargrove et al. (77). The hemin loss assay was performed in 0.1 M potassium phosphate, pH7.0, 37° C in 20% sucrose as an osmolyte to inhibit precipitation of apohemoglobin at the end of the reaction (77). The time course for hemin dissociation from the Hp:HbA dimer complex was computed from the rate constant reported by Mollan et al. (68). B. Active site of human 7agr; subunits showing how penetration of water can facilitate both autooxidation and hemin dissociation.

Twenty-five years ago Hargrove et al. (78) developed an assay to measure the rate of hemin dissociation from metMb and metHb. They constructed a double mutant of sperm whale myoglobin in which the distal histidine (His(E7)) was replaced with a Tyr (H64Y), which chelates to the ferric iron atom and gives the holo metMb a green color. The distal valine (Val(E11)) was replaced with a Phe (V68F) to enhance the stability of the apoprotein. When excess H64Y/V68F apoMb is mixed with metHb samples, the color of the solution turns from brown to green as the apoMb reagent scavenges hemin from the hemoglobin sample. The rate-limiting step is hemin dissociation, which occurs with rate constants, k_-H, on the order of 40 to 0.3 h⁻¹, and half-times of 1 to 2 minutes to 2 to 3 hours, depending on pH and protein concentration.

As with autooxidation, the rate of hemin dissociation from HbA depends strongly on oligomeric structure and thus globin concentration (77). Comparisons of time courses for hemin dissociation from isolated subunits, dimers, tetramers, and the dimer:haptoglobin complex are shown in Fig. 7A using data taken primarily from Mollan et al. (68). Monomeric subunits are highly unstable in the ferric form and lose hemin completely in 10 to 20 minutes; dimers lose all of their hemin to the scavenger in an hour; and tetramers are significantly more stable. The fast and slow phases for hemin loss from dimers and tetramers represent hemin dissociation from the β and α subunits respectively (77, 78). The most remarkable result in Fig. 7 is the dramatic stabilization of bound hemin by the binding of HbA dimers to haptoglobin (68). Virtually no loss of hemin was observed even when the apoMb scavenging agent was added in great excess. Baek et al. (15) observed a similar result using apohemopexin as a scavenging agent and suggested that a key role of haptoglobin is to take up metHb dimers and prevent hemin loss until the complex can be processed by macrophages.

By analogy with the results for autooxidation, our interpretation of the effects of quaternary structure is that the formation of αβ subunit interfaces or the Hp:HbA dimer complex dampens fluctuations in the globin tertiary structure, which markedly reduces water penetration into the proximal portion of the heme pocket and inhibits hydration of the Fe-His(F8) bond. Using a large library of recombinant variants, we were able to identify more specific interactions that inhibit hemin dissociation from myoglobin and most of which involve favorable electrostatic interactions with the heme propionates (79). However, in the case of hemoglobin these interactions either don’t exist or are weaker, accounting in part for the 30 to 100-fold higher rates of hemin dissociation from metHb versus from metMb. Introducing more favorable electrostatic interactions in the β subunits of hemoglobin does appear to inhibit hemin loss from HbA dimers but the effects are small (≤ 2-fold) and not observed in Hb tetramers. As in the case of autooxidation, we have not been able to engineer rationally single point mutants with rates of hemin loss that are smaller than those for native HbA tetramers. This result is emphasized in the last two columns of Table 1, where all the HBOCs, even when polymerized, have higher rates of hemin dissociation than tetrameric HbA.

Again, we feel that the key to reducing both autooxidation and hemin loss is to understand the dynamics of water penetration into the heme pocket and then to engineer α and β subunits with reduced rates and extents of solvent movement into protein. The Hp:HbA dimer complex serves as a remarkable example of slowing hemin loss. However, a simple static comparison of the crystal structures of the Hp:dimer complex and tetrameric Hb shows almost no differences in the positions of the α and β polypeptide chains with respect to their heme groups. In both cases, the heme propionates are still exposed to solvent (80). Thus, the dramatic differences shown in Fig. 7A must be due to differences in the fluctuations of the structures that allow water to transiently penetrate and hydrolyze the Fe-His(F8) bond. The dilemma is that hemin loss is slow (as is autooxidation) so the amount of water inside the pocket at anytime is very small and undetectable by conventional NMR and crystallographic methods. Molecular dynamics simulations are needed to identify regions of the structure that enhance or inhibit fluctuations. Then these regions need to be mutated to try to inhibit water penetration.

APOGLOBIN STABILITY AND EXPRESSION

The final process in the degradation of hemoglobin is the unfolding and precipitation of the apoprotein (Reaction 4 in Figs. 1 and 3). In general, apohemoglobin is difficult to work with at temperatures above 10° C and rapidly precipitates under physiological conditions at 37°C. Almost all previous studies of apoHb denaturation involved irreversible studies of thermal or acid denaturation, dehydration, or solublization in sodium dodecyl sulfate. Recently, we examined for the first time reversible unfolding of apoHb at pH 7, 10° C using GdnHCl as a denaturant, which prevents the unfolded subunits from precipitating (81). Results for apoHbA are summarized in Fig. 8. The process involves two major steps. Loss of hemin generates an apo-dimer (D state), which retains most (~ 70–80%) of the helical structure of the holoprotein tetramer. The first unfolding step for this apodimer involves melting of the heme pocket with loss of most of the B, C, D, E, and F helices but retention of the helices (A, G, and H) that are involved in the α₁β₁ interface. The second step involves dissociation of the dimer interface and formation of almost completely unfolded monomers. The molten globule intermediate, I_D, is a dimer, which can still bind hemin to form hemichrome spectral species that can refold into native holoprotein. As indicated in Fig. 8, mutations that stabilize the heme pocket inhibit the first phase of apoglobin unfolding, whereas mutations that stabilize the α₁β₁ dimer interface inhibit the second phase of unfolding. The most dramatic of the latter mutations have genetically crosslinked α subunits, i.e. rHb0.1 as described in Fig. 7. In this case, complete denaturation requires disruption of two dimer interfaces to generate two unfolded β subunits and one di-α subunit (81). Thus, crosslinking does inhibit equilibrium unfolding of hemoglobin as would be expected. This mechanism also helps explain why, under most conditions, human fetal hemoglobin is more resistant to denaturation than adult hemoglobin because the α₁γ₁ interface in HbF is stronger than the α₁β₁ interface in HbA (81).

Fig. 8. ApoHb unfolding involves two major steps.

These data for apoHbA were taken from Samuel et al. (81) and represent denaturation by GdnHCl in 0.2 M potassium phosphate buffer, pH 7, at 10° C in the presence of 1 mM dithiothreitol. The first phase involves unfolding of the heme pockets in both subunits to form a dimeric molten globule state and the second phase involves dissociation of the dimer intermediate into unfolded monomers.

Because our measurements involve reversible unfolding, the observed mechanism also applies to the folding and assembly of holohemoglobin as shown in Fig. 9, which represents a general scheme for the biosynthesis of hemoglobin either in pre-erythroid cells or in bacteria. The arguments in favor of this scheme for hemoglobin are given in Samuel et al. (81) and Varnado et al. (5), and similar ideas have been proposed for the biosynthesis of myoglobin (82). The scheme in Fig. 9 serves to indicate which factors are crucial for efficient production of holoHb and which are important for preventing denaturation.

Fig. 9.

Mechanisms for the biosynthesis, folding, and assembly of human hemoglobin tetramers.

Assuming heme availability, holoprotein expression yield is governed by native apoglobin stability, which, in the case of hemoglobin, includes both the strength of the dimer interface and the extent of folding of the heme pocket. In a study using myoglobin as a simple model system, we have shown that the highest expression yields are observed for variants in which distal histidine and valine are replaced with large apolar amino acids, i.e. Leu or Phe, which greatly stabilize the folded heme pocket and facilitate the uptake of heme (82). We have observed that qualitatively similar increases in production yields in E. coli occur for His(E7)Leu/Val(E11)Phe mutations in hemoglobin. During biosynthesis the rate of holoprotein formation is governed by the fraction of native apoprotein (native D state in Fig. 9), which is capable of reacting with heme, versus the fraction of unfolded chains (U_M states in Fig. 9), which irreversibly self-aggregate. In the case of Mb, we have shown quantitatively that there is strong, direct correlation between holoprotein yield and the overall stability of the folded native apoglobin state obtained from fits to GdnHCl-induced unfolding curves, which are similar to those in Fig. 8. This correlation holds for expression in cell-free systems, E. coli, and in mammalian muscle (82), and we assume the same situation occurs for human hemoglobin.

Thus, the expression yields of holo-myoglobins and hemoglobins depend strongly on the absolute value of the equilibrium constant for formation of the folded native apoglobin state, starting from unfolded chains. This parameter determines the fraction of newly expressed polypeptide chains that are capable of rapidly binding heme versus the fraction that remains unfolded and irreversibly aggregates. In contrast, the yield of holoprotein is little affected by hemin affinity or resistance of the oxygenated complex to autooxidation (82).

However, mutations that enhance expression yields do not always yield holoproteins that are functional or resistant to oxidative degradation. Replacing the distal histidine with Leu enhances the stability of the apoprotein, but markedly decreases O₂ affinity (P₅₀ ≥ 100 μM), enhances autooxidation (k_autox ≈ 0.5 h⁻¹), and speeds up hemin dissociation (k_-H ≥ 20 h⁻¹) in human α subunits (83). The latter detrimental properties are due to loss of hydrogen bonding capacity of the E7 amino acid side chain. His(E7) stabilizes bound O₂ and inhibits its protonation, which decreases P₅₀ and k_autox, respectively (Fig. 5). The His(E7) side chain also hydrogen bonds to water coordinated to the Fe(III) atom in metHb, inhibiting dissociation of the hemin group. These results provide a good example of how functionality can compromise globin stability and expression and that the evolution of a successful phenotype involves a fine balance of protein properties. It also points out the need to consider simultaneously all the key functional and stability properties when developing protein engineering strategies.

FUTURE DIRECTIONS

The four key characteristics of HBOC products are efficient oxygen delivery, low rates of NO scavenging, resistance to oxidative degradation, and high production yields. O₂ affinity is easily engineered with chemical modifications or site-specific mutations based on the well-established mechanisms shown in Fig. 2 (29, 72). Slowing extravasation of the HBOC into the endothelium by increasing its size or by PEGylation reduces the extent of NO scavenging and the hypertensive side effect (26, 29). The bimolecular rate constant for NO dioxygenation can also be reduced more directly by distal pocket mutations that inhibit ligand capture (26, 28, 29).

The key to limiting the oxidative degradation of HBOCs is to inhibit autooxidation. Heme loss and globin denaturation will not occur under physiological conditions if the protein is kept in a reduced holoprotein state. However, once the iron atom is oxidized, hemin loss can be rapid if acceptors (albumin, apohemopexin, membranes, receptors, and lipoproteins) are present, and most of the resultant apoglobins are relatively unstable at 37° C and begin to irreversibly precipitate. Redox reactions with the H₂O₂ produced by autooxidation can also accelerate degradation (1). Although myoglobins have been engineered with rates of autooxidation and hemin loss that are lower than those for the native or wild-type protein, we have been unable to construct Hb site-specific mutants with rates that are lower than those for native HbA tetramers.

Both autooxidation and hemin loss require interactions of the heme group or bound O₂ with water in the heme pocket (Figs. 6 and 7), which in turn requires fluctuations in the tertiary structure that allow solvent penetration into the protein interior. In our view, there are two keys to developing a successful strategy to reduce autooxidation and hemin loss rates. First, computational methods need to be developed and applied systematically to examine where fluctuations and solvent penetration into the heme pocket occurs. The goals would be to identify what regions facilitate fluctuations and which are dampened by dimer and tetramer formation or binding to haptoglobin, and then to examine whether or not these regions are more rigid in myoglobins. Our initial guesses involve regions at the CD, EF, and FG corner, and multiple mutations are probably needed to change these more global motions. Second, high throughput methods need to be developed to screen large libraries of hemoglobin mutants for enhanced resistance to oxidative degradation. In our study of Mb expression, we developed a cell-free translation system for expressing, partially purifying, and characterizing the spectral properties of libraries of myoglobin variants. This assay could easily be adapted to measure rates of autooxidation or hemin loss for hemoglobin in multiple well plates. Similar approaches can be taken to enhance apoglobin stability where again single point mutations normally have only small effects unless they are in crucial areas of the distal pocket or subunit interfaces, but even then multiple substitutions are required to have a large effect.