Blood Substitutes

1. Field of Invention

This invention relates to novel hemoglobin compositions useful as substitutes for red blood cells, and to methods of preparing same. It further relates to the preparation, using recombinant DNA technology, of mutant hemoglobins for use in such compositions.

2. Information Disclosure Statement

It is not always practical to transfuse a patient with donated blood. In these situations, use of a red blood cell substitute is necessary. The product must effectively transport O.sub.2, just as do red blood cells. ("Plasma expanders", such as dextran and albumin, do not transport oxygen.) The two types of substitutes that have been studied most extensively are hemoglobin solutions and fluorocarbon emulsions.

Hemoglobin (Hgb) is the oxygen-carrying component of blood. Hemoglobin circulates through the bloodstream inside small enucleate cells called erythrocytes (red blood cells). Hemoglobin is a protein constructed from four associated polypeptide chains, and bearing prosthetic groups known as hemes. The erythrocyte helps maintain hemoglobin in its reduced, functional form. The heme iron atom is liable to oxidation, but may be reduced again by one of two enzyme systems within the erythrocyte, the cytochrome b.sub.5 and glutathione reduction systems.

Hemoglobin exhibits cooperative binding of oxygen by the four subunits of the hemoglobin molecule (two alpha-globins and two beta-globins in the case of Hgb A), and this cooperativity greatly facilitates efficient oxygen transport. Cooperativity, achieved by the so-called heme-heme interaction, allows hemoglobin to vary its affinity for oxygen. Hemoglobin reversibly binds up to four moles of oxygen per mole of Hgb. At high oxygen concentration, such as that found in the lungs, the oxygen affinity is high and hemoglobin is almost saturated with oxygen. At low oxygen concentration, such as that found in actively respiring tissue, the oxygen affinity is lowered and oxygen is unloaded.

Oxygen-carrying compounds are frequently compared by means of a device known as an oxygen dissociation curve. This curve is obtained when, for a given oxygen carrier, oxygen saturation is graphed against the partial pressure of oxygen. The percentage of saturation increases with partial pressure according to a sigmoid relationship. The P.sub.50 is the partial pressure at which the oxygen-carrying solution is half saturated with oxygen. It is thus a measure of oxygen-binding affinity; the higher the P.sub.50, the more loosely the oxygen is held.

When the oxygen dissociation curve of a oxygen-carrying solution is such that the P.sub.50 is less than that for whole blood, it is said to be "left-shifted."

The oxygen affinity of hemoglobin is lowered by the presence of 2,3-diphosphoglycerate (2,3-DPG), chloride ions and hydrogen ions. Respiring tissue releases carbon dioxide into the blood and lowers its pH (i.e. increases the hydrogen ion concentration), thereby causing oxygen to dissociate from hemoglobin and allowing it to diffuse into individual cells.

The ability of hemoglobin to alter its oxygen affinity, increasing the efficiency of oxygen transport around the body, is dependent on the presence of the metabolite 2,3-DPG. Inside the erythrocyte 2,3-DPG is present at a concentration nearly as great as that of hemoglobin itself. In the absence of 2,3-DPG "conventional" hemoglobin binds oxygen very tightly and would release little oxygen to respiring tissue.

Aging erythrocytes release small amounts of free hemoglobin into the blood plasma where it is rapidly bound by the scavenging protein haptoglobin. The hemoglobin-haptoglobin complex is removed from the blood and degraded by the spleen and liver.

It is clear from the above considerations that free native hemoglobin A, injected directly into the bloodstream, would not support efficient oxygen transport about the body. The essential allosteric regulator 2,3-DPG is not present in sufficient concentration in the plasma to allow hemoglobin to release much oxygen at venous oxygen tension, and free hemoglobin would be rapidly inactivated as an oxygen carrier by auto-oxidation of the heme iron.

Nonetheless, solutions of conventional hemoglobin have been used as RBC substitutes. The classic method of preparing hemoglobin solutions employs outdated blood. The red cells are lysed and cellular debris is removed, leaving what is hopefully "stromal-free hemoglobin" (SFH).

Several basic problems have been observed with this approach. The solution must be freed of any toxic components of the red cell membrane without resorting to cumbersome and tedious procedures which would discourage large-scale production. DeVenuto, "Appraisal of Hemoglobin Solution as a Blood Substitute", Surgery, Gynecology and Obstetrics, 149: 417-436 (1979).

Second, as expected, such solutions are "left-shifted" (lower P.sub.50) as compared to whole blood. Gould, et al., "The Development of Polymerized Pyridoxylated Hemoglobin Solution as a Red Cell Substitute", Ann. Emerg. Med. 15: 1416-1419 (Dec. 1986).

Third, SFH has a half-life in the circulatory system of only about 2-4 hours. This is because oxyHgb partially dissociates into a dimer that is small enough to be filtered by the kidney.

Finally, SFH has a high colloid osmotic pressure (COD). Thus, administration of SFH in a dose that would have the same oxygen-carrying capacity as a unit of packed red blood cells is inadvisable, since the high osmotic pressure (60 mm Hg) would cause a massive influx of water from the cells into the bloodstream, thus dehydrating the patient's tissues. This consideration limits the dose of SFH to about 6-8 gm Hgb/dl.

In an effort to restore the desired P.sub.50, researchers added 2,3-DPG to the hemoglobin solution. Unfortunately, 2,3-DPG was rapidly eliminated from the circulation. Scientists then turned to other organic phosphates, particularly pyridoxal phosphate. Like 2,3-DPG, these compounds stabilized the "T state" of the Hgb by forming a salt bridge between the N-termini of the two beta chains. The pyridoxylated hemoglobin had a P.sub.50 of 20-22 torr, as compared to 10 torr for SFH and 28 torr for whole blood. While this is an improvement over SFH, the pyridoxylated Hgb remains "high affinity" relative to whole blood.

Hemoglobin has been chemically modified (by intramolecular or intermolecular crosslinking) to increase intravascular retention and reduce osmotic pressure. Unfortunately, this polymerization also causes a "left shift" of the molecule's oxygen dissociation curve. Thus, for polymerized-pyridoxylated Hgb, the P.sub.50 is about 18 torr.

For chemical modifications of hemoglobin, See Iwashita, U.S. Pat. Nos. 4,412,989 and 4,301,144 (with polyalkylene glycol), Iwasaki, U.S. Pat. No. 4,670,417 (with polyalkylene oxide), (with a polysaccharide); Nicolau, U.S. Pat. Nos. 4,321,259 and 4,473,563 (with inositol phosphate); Wong, U.S. Pat. Nos. 4,710,488 and 4,650,786 (with inositol phosphate and a polysaccharide); Bonhard, U.S. Pat. No. 4,336,248 (with other proteins or gelatin derivatives); Walder, U.S. Pat. Nos. 4,598,064 and 4,600,531 (intramolecularly crosslinked hemoglobin) and Ajisaka, U.S. Pat. No. 4,377,512 (with inulin).

The human alpha- and beta-globin genes have both been cloned and sequenced. Liebhaber, et al., P.N.A.S. (U.S.A.) 77: 7054-58 (1980)(alpha-globin genomic DNA); Marotta, et al., J. Biol. Chem., 252: 5040-53 (1977)(beta-globin cDNA).

Nagai and Thorgerson (Nature, 309: 810-812, 1984) expressed in E. coli a hybrid protein consisting of the 31 amino-terminal residues of the lambda cII protein, an Ile-Glu-Gly-Arg linker, and the complete human beta globin chain. They cleaved the hybrid at the single arginine with blood coagulation factor Xa, thus liberating the beta-globin chain.

Later, Nagai, et al., P.N.A.S. (U.S.A.), 82: 7252-55 (1985) took the rDNA-derived human beta globin, naturally derived human alpha globin, and a source of heme and succeeded in producing active human hemoglobin. Additionally, they produced two semi-artificial analogues of the naturally occurring mutant hemoglobins Hb Nympheas and Hb Daphne by site-directed mutagenesis of the cloned beta-globin gene, expression of the modified gene, and combination of the rDNA-derived beta chain with naturally occurring alpha chain and a source of heme. Like the naturally occurring mutants, these semiartificial analogues exhibited increased oxygen affinity as compared to "normal" hemoglobin. In subsequent studies, the structural basis for this change in oxygen binding was established. Luisi and Nagai, Nature, 320: 555-56 (1986); and cp. Nagai, et al., Nature, 329: 858-860 (Oct. 1987)(similar production of mutant hemoglobins with replacements of Val(67beta)Ell).

Surprisingly, the expression of the human alpha globin gene in heterologous cells has not been reported. Indeed, merely substituting the alpha globin gene for the beta globin gene in the E. coli expression system of Nagai, et al. results in negligible expression of the lambda cII protein/alpha-globin fusion. It appears that there is something about the native alpha globin mRNA sequence which interferes with efficient translation in a bacterial host.

We have discovered that the disadvantages of hemoglobin solutions as blood substitutes are overcome if a mutant hemoglobin species is selected which, in a typical cell-free blood substitute solution, would impart to the solution a P.sub.50 comparable to that of non-mutant hemoglobin in RBC-bound state. Naturally occurring hemoglobin mutants which, in the erythrocyte environment, would impart to the erythrocyte a P.sub.50 higher than the normal value for whole blood (28 torr) are of particular interest, both in their own right and for what they teach about the structural foundations of oxygen affinity. It is expected that many such "right-shifted" species, outside the erythrocyte environment (and thus the right-shifting influence of 2,3-DPG), will assume a P.sub.50 comparable to or greater than that of the normal P.sub.50 for whole blood.

For the purposes of this invention, the term "conventional hemoglobin A" refers to the species of Hgb A whose alpha and beta chains are of the amino acid sequence given in FIG. 1. This is the species which is most frequently found in human erythrocytes and which imparts to such erythrocytes a P.sub.50 of about 28 torr. A "hemoglobin A mutant" is defined as any species of hemoglobin A whose alpha or beta chain is of an amino acid sequence different from that set forth in FIG. 1. A "low affinity" hemoglobin A mutant is one which has a P.sub.50 at least about 10% greater than that of "conventional hemoglobin A" in the same environment. It is particularly desirable that its P.sub.50 be at least that twice of conventional (wild type) hemoglobin A in the absence of 2,3-DPG. A "recombinant" hemoglobin is one composed of an alpha and beta globin at least one of which is obtained by expression of a globin gene carried by a recombinant DNA molecule, whether the hemoglobin is a conventional hemoglobin or a mutant species.

A large number of naturally occurring low affinity Hgb A mutants are known. (See Table I). The mutations may appear in either the alpha or the beta chains of the molecule (or both, of course). Thus, Hgb Hazebrouck is a beta mutant (38(C4):thr.fwdarw.pro) whose P.sub.50 is 36 (in whole blood), declining to 27-29 torr in vitro.

Clearly, one cannot depend on nature to provide an adequate supply of these low affinity mutants. Consequently, the mutant polypeptide chain will usually be prepared artificially, either by direct polypeptide synthesis, or, more preferably, by in vivo expression of the corresponding mutant gene in a suitable host cell. This gene may be obtained directly from the genome of the mutant erythrocyte precursor (a mature erythrocyte does not contain DNA), as a complementary DNA (cDNA) transcribed from the messenger RNA of the mutant erythrocyte precursor, by direct polynucleotide synthesis, or, preferably, by in vitro mutagenesis of the gene encoding conventional hemoglobin.

If one of the chains is identical to that of "conventional" hemoglobin, it may be obtained either naturally or synthetically. In addition, to create a functional hemoglobin molecule, it is necessary to provide the prosthetic heme groups and to couple the alpha and beta chains.

It is also within the contemplation of this invention to prepare and use non-naturally occurring low affinity mutants by appropriate modification and expression of the alpha or beta globin gene and subsequent assembly of a recombinant hemoglobin. Methods for selecting candidate sequences and evaluating their suitability for use in a blood substitute product are described herein.

We also have discovered that human alpha-globin may be obtained from a bacterial host by (1) constructing a fused gene which comprises not only the alpha-globin gene but also at least a portion of the beta globin gene, separated by spacer DNA encoding a selective protease cleavage site; (2) expressing the fused gene in the form of a fusion protein; and (3) cleaving the fusion protein at the aforementioned cleavage site in order to liberate the human alpha-globin.

As a result of this discovery, it is possible to prepare entirely artificial human hemoglobin, that is, hemoglobin in which both the alpha and beta globin chains are formed in cells other than human erythrocytes. Such fully artificial hemoglobin is ideal for use in blood substitute solutions. Of course, semi-artificial hemoglobins (one chain obtained from non-erythrocyte source) may still be used.

When extracted natural hemoglobin is used as a blood substitute, one must be concerned with the toxicity of red blood cell membrane components which might contaminate the product. It is known that erythrocyte stroma can cause dyspnea, bronchospasm, hypotension, arrythmia, disseminated intravascular coagulation, activation of complement, and renal, myocardial and hepatic changes associated with ischemia and acute inflammation. See Feola, Surgery, Gynecology & Obstetrics, 166: 211-222 (March 1988); MacDonald, et al., F.A.S.E.B. J., 2(6) Abstr. 8217 (1988); Stone, et al., Surgery, Gynecology and Obstetrics, 149: 874-876 (1979); Rabiner, et al, J. Exp. Med., 126: 1127-42 (1967). While purified preparations of natural hemoglobin are known (so called "stroma-free hemoglobin"), Feola comments, "a truly pure hemoglobin solution has not been produced."

Another concern with natural hemoglobin is contamination with infectious agents communicated by blood. Bove, Progr. Hematol., 14: 123-145 (1986) reported that hepatitis viruses, cytomegalovirus, Epstein-Barr virus, serum parvoviruses, syphilis, malaria, filariasis, trypanosomiasis, babesiosis, numerous pathogenic bacteria, and AIDS are all transmitted by blood transfusions. AIDS has even been transmitted by blood transfusions screened as negative for HIV antibody. Ward, et al., New Engl. J. Med., 318: 473-78 (1988).

The alpha-globin of the present invention may be identical in sequence to natural normal human alpha-globin or to a naturally occurring hemoglobin mutant, or it may be an alpha-globin mutant which is not known in nature.

Mutant hemoglobins, whether of reduced or enhanced oxygen affinity, may be of value for altering O.sub.2 concentrations in cell cultures, or for extracting O.sub.2 from fluids.

The appended claims are hereby incorporated by reference into this specification as a statement of the preferred embodiments.

FIG. 1 shows the amino acid sequence of the (a) alpha and (b) beta chains of conventional human hemoglobin A, and the nucleotide sequences encoding these chains.

FIG. 2 shows selected DNA sequences and enzyme restriction maps of (a) M13 mpll FX, (b) pLcIIFX beta, and pLcII beta, and (c) pLCIIFX-beta-FX-alpha and (d) mpIIFXalpha. Note that pLCIIFX alpha lacks the codon encoding beta His 2 in FIG. 2(b).

FIG. 3 shows equilibrium oxygen binding curves for artificial conventional human hemoglobin (open boxes) and mutant human hemoglobin having the structure of Hgb Kansas (.alpha..sub.2 .beta..sub.2 Asn.sup.102 .fwdarw.Thr)(shaded boxes). 0.1 M HEPES, pH 7.4, 0.1 M NaCl, 25.degree. C.

One aspect of this invention is the cloning and expression of an exogenous alpha globin gene in a suitable host. The host may be a procaryotic cell (such as a E. coli) or a eucaryotic cell (such as a yeast cell or a mammalian cell). The alpha globin gene will usually code on expression for a polypeptide corresponding to a naturally occurring human alpha globin, normal or abnormal, but may correspond to a nonhuman alpha globin, or indeed be a non-naturally occurring analogue of known alpha globins. Preferably, a mutant alpha globin is prepared and assembled into a low affinity hemoglobin mutant for use in a blood substitute composition.

The alpha globin gene is expressed as part of a fused gene which also codes on expression for at least a portion of the beta globin gene. In a preferred embodiment, the alpha and beta globin sequences are separated by spacer DNA encoding a selective protease cleavage site, in particular, a site susceptible to specific cleavage by Factor Xa.

Preferably, the aforementioned fused gene includes a subsequence which codes on expression for the 20 amino acid N-terminal of beta globin.

As previously noted, the alpha globin of the present invention is advantageously combined with rDNA-derived beta globin and a source of heme to obtain fully artificial (derived entirely from sources other than blood) hemoglobin. Such hemoglobins, and especially low oxygen affinity hemoglobin mutants produced through use of selectively modified alpha and/or beta globin genes, are of value as blood substitutes. Semi-artificial hemoglobins, in which the only chain obtained from a non-erythrocyte source is a mutant sequence and the mutation imparts reduced oxygen affinity to the molecule, are also encompassed by the present invention and may also be used as blood substitutes. Unless otherwise indicated, the term "artificial" embraces both wholly and semi-artificial forms.

In order to appreciate possible strategies for designing low affinity alpha or beta mutants of hemoglobin for use as blood substitutes, it is necessary to understand the structure of the hemoglobin molecule.

The structure of conventional hemoglobin is well known. We herewith incorporate by reference the entire text of Bunn and Forget, eds., Hemoglobin: Molecular. Genetic and Clinical Aspects (W. B. Saunders Co., Philadelphia, PA: 1986) and of Fermi and Perutz "Hemoglobin and Myoglobin," in Phillips and Richards, Atlas of Molecular Structures in Biology (Clarendon Press: 1981).

The primary structure of a polypeptide is defined by its amino acid sequence and by identification of any modifications of the side chains of the individual amino acids.

About 92% of the normal adult human hemolysate is Hgb A (designated alpha2 beta2, because it comprises two alpha and two beta chains). The alpha chain consists of 141 amino acids (See FIG. 1). The iron atom of the heme (ferroprotoporphyrin IX) group is bound covalently to the imidazole of his 87 (the "proximal histidine"). The beta chain is 146 residues long (see FIG. 1) and heme is bound to it at His 92.

Other recognized hemoglobin species are Hgb A.sub.2, Hgb A.sub.1a, Hgb A.sub.1b, and Hgb A.sub.1c, as well as the rare species Hgb F, Hgb F.sub.1, Hgb Gower-1, Hgb Gower-2, Hgb Portland, Hgb H, and Hgb Bart. They are distinguished from Hgb A by a different selection of polypeptide chains.

Segments of polypeptide chains may be stabilized by folding into one of two common conformations, the alpha helix and the beta pleated sheet. In its native state, about 75% of the hemoglobin molecule is alpha-helical. Alpha-helical segments are separated by segments wherein the chain is less constrained. It is conventional to identify the alpha-helical segments of the each chain by letters, e.g., the proximal histidine of the alpha chain is F8 (residue 8 of helix F). The non-helical segments are identified by letter pairs, indicating which helical segments they connect. Thus, nonhelical segment BC lies between helix B and helix C. In comparing two variants of a particular hemoglobin chain, it may be enlightening to attempt to align the helical segments when seeking to find structural homologies. For the amino acid sequence and helical residue notation for conventional human hemoglobin A.sub.o alpha and beta chains, see Table 4.

The tertiary structure of the hemoglobin molecule refers to the steric relationships of amino acid residues that are far apart in the linear sequence, while quaternary structure refers to the way in which the subunits (chains) are packed together. The tertiary and quaternary structure of the hemoglobin molecule have been discerned by X-ray diffraction analysis of hemoglobin crystals, which allows one to calculate the three-dimensional positions of the very atoms of the molecule.

In its unoxygenated ("deoxy", or "T" for "tense") form, the subunits of hemoglobin (alpha1, alpha2, beta1, and beta2) form a tetrahedron having a twofold axis of symmetry. The axis runs down a water-filled "central cavity". The subunits interact with one another by means of Van der Waals forces and hydrogen bonds (and, in the case of deoxyhemoglobin, by "salt bridges"). The alpha1beta1 and alpha2beta2 interfaces remain relatively fixed during oxygenation. In contrast, there is considerable flux at the alpha1beta2 (and alpha2beta1) interface. In its oxygenated ("oxy", or "R" for "relaxed" form), the intersubunit distances are increased.

The deoxy conformation is stabilized by numerous interactions, including, for example, the hydrogen bond between Tyr42alpha and Asp99beta. In the oxygenated form, this bond is broken and a new one formed between Asp94alpha and Asn102beta.

Various different approaches to modification of hemoglobin may be adopted. In each case, a candidate mutant is selected which, on the basis of the available evidence, is believed to be likely to have a lower affinity for oxygen than conventional hemoglobin.

In making this selection, it is possible to consider not only the known effects of various mutations of human hemoglobin, but also the oxygen binding capacity of known forms of animal hemoglobins, and of related compounds such as carboxyhemoglobin, methemoglobin, myoglobin, etc.

Thanks to our success in overcoming the problems of expressing the alpha globin gene in a heterologous system, it is now possible to conveniently prepare alpha globin mutants.

Several low oxygen affinity hemoglobin alpha chain mutants are already known. Of these, Hb Titusville (alpha94Asp=>Asn), Hb Setif (alpha94Asp.fwdarw.Tyr), Hb Torino (alpha43Phe.fwdarw.Val), Hb Hirosaki (alpha43Phe.fwdarw.Leu) and Hb Moabit (alpha86Leu.fwdarw.Arg) are of special interest.

Alpha globins are more readily oxidized than beta globins because the His(F8) to O.sub.2 bond on the alpha chain is slightly stronger than on beta, so that an electron is more readily lost by the oxygen. Alpha globins could be modified to make them less oxidizable, for example, by the change alpha63His.fwdarw.Gln or Val.

Chloride ion binds to the alpha chain by bridging between the N-terminal NH.sub.3.sup.+ and the hydroxyl of alpha131Ser. The effect of chloride binding is to increase the P.sub.50 slightly. It is believed that by changing alpha131 to Glu, Asp or Asn one could achieve the same effect without resorting to chloride. Alternatively, the pK.sub.a of the N-terminal could be increased. The natural human N-terminal is valine, with a pK.sub.a of9.72. This could be replaced with Ile (9.76), Pro (10.60) or Thr (10.43).

Beta globin mutants, and additional alpha globin mutants which are likely to impart reduced oxygen affinity, are discussed below.

It is not possible to raise the plasma concentration of 2,3-DPG sufficiently to maximize the oxygen carrying efficiency of free conventional hemoglobin in the blood. This problem can be overcome by stabilizing the T state with additional salt-bridges or hydrogen bonds introduced by protein engineering. The cross-linking of hemoglobin may itself stabilize the T structure to some extent.

Hydrogen bonds and ionic salt bridges are the predominant stabilizing forces on the surface of proteins. Hydrogen bonds are weak non-ionic bonds formed between electronegative atoms (eg. oxygen, nitrogen, sulfur) and protons that are covalently attached to other electronegative atoms. Individually, hydrogen bonds are weak (eg .about.1 kcal/mol), but in a protein there are hundreds to thousands of hydrogen bonds that collectively amount to a large stabilizing force. An example of an hydrogen bond that is important to hemoglobin structure is the hydrogen bond formed between alpha-asp.sup.94 and beta-asn.sup.102 in the oxystate. When either of these residues is mutated to a residue that can no longer form this hydrogen bond the oxy state is destabilized and the molecule has a much lower O.sub.2 affinity. Hg Kansas (beta thr.sup.102), Hg Beth Israel (beta ser.sup.102), Hg Richmond (beta lys.sup.102), Hg St. Mande (beta tyr.sup.102), Hg Titusville (alpha asn.sup.94) and Hg Setif (alpha tyr.sup.102) are all examples of the importance of this hydrogen bond. Other likely non-natural mutants that will achieve the same effect are beta asp.sup.102, beta glu.sup. 102, beta arg.sup.102, beta his.sup.102, beta gly.sup.102 and beta cys.sup.102 ; alpha gln.sup.94, alpha thr.sup.94, alpha ser.sup.94, alpha lys.sup.94, alpha gly.sup.94 and alpha arg.sup.94.

Ionic interactions are salt forms formed between juxtaposed residues of opposite charge. These interactions are of much greater strength mole for mole than hydrogen bonds. An example of attraction of unlike charges would be the interaction between a lys and asp residue; at physiologic pH both of these residues are charged (positive and negative, respectively). Repulsion of two juxtaposed positive charges or two juxtaposed negative charges may also occur; interactions such as these are destabilizing.

Stabilization of the deoxy state of Hg by 2,3-DPG is an example of ionic interaction. The 2,3-DPG molecule is highly charged at neutral pH (5 negative charges) and interacts with eight residues in the 2,3DPG pocket that are positively charged. It is felt that by engineering more positive charge into this binding pocket that 2,3-DPG would bind more tightly to the engineered Hgb than to HgbA. Another example is the alpha.sub.1 /beta.sub.2 interface where alpha asp.sup.94 hydrogen bonds to beta asn.sup.102. Replacing beta asn.sup.102 with a negatively charged group such as asp or glu will interfere with oxy state stabilization by charge repulsion with the like charged alpha asp.sup.94.

Thus, certain amino acid residue changes can facilitate the formation of the desired hydrogen bonds and salt bridges.

The T conformation can also be stabilized by the substitution of cysteine residues for other residues. Cysteine residues may be cross-linked by disulfide bonds. Examination of the published x-ray data of methemoglobin suggests that the alpha.sub.1 /beta.sub.1 interface would be a logical place to put the disulfide. Conveniently, there are two alpha ala residues, G17 and G18 (ala is sterically similar to cys) adjacent to beta G14 cys. Hence, at first glance it is thought that alpha G17 or G18 would be likely spots for cys residues to be added. Further strategy about where to place disulfides can be guided by Thornton, J. M. J. Mol. Biol. 151, 261-287, 1981. Oxidation of cysteines to disulfide (Cystines) bonds can be carried out by treatment with O.sub.2 or catalyzed by thioredoxin (Pigiet, V Am. Biotech. Lab 6, 48-52, 1988).

Heme (haem) is the prosthetic group of hemoglobin, myoglobin, catalase, peroxidase, and cytochrome b. The heme is inserted in a cleft between the E and F helices. The heme iron is linked covalently to the imidazole nitrogen of the "proximal" F8 histidine. The "distal" Ell valine appears to guard the access of oxygen to the heme pocket.

Val-Ell and His-E7 are highly conserved residues which are in Van der Waals contact with the oxygen molecule liganded to the heme iron atoms of hemoglobin; by replacing these residues the intrinsic oxygen affinity of hemoglobin can be altered. Val-Ell has been replaced with Ile, Leu, Ala and Met. The oxygen affinity of the Ala-Ellbeta mutant was higher than that of HbA; that of the Ile-Ellbeta mutant was lower. X-ray crystallographic study of the latter mutant showed that the delta-methyl group of the Ile side chain must be pushed to one side if oxygen is to bind to the iron atom.

Another alteration that we have made is beta His63.fwdarw.Phe. This mutant has an extraordinarily low oxygen affinity (See Table 3).

It has therefore been shown that the oxygen affinity of hemoglobin can be altered at will by replacing residues close to the oxygen binding site. By adjusting oxygen affinity in this way the efficiency of oxygen transport can be maximized in the absence of allosteric effectors such as 2,3-DPG.

The following residues of human deoxyhemoglobin are, on a nearest atom-to-nearest atom basis, within 4 angstroms of the heme moiety: Alpha B13(D)Met, C7(E)Tyr, CE1(D)Phe, CE3(E)His, CE4(D)Phe, E7(D)His, E10(D)Lys, E1I(D)Val, E14(D)Ala, F4(P)Leu, F7(P)Leu, F8(P)His, FG3(P)Leu, FG5(P)Val, G4(P)Asn, G5(P)Phe, G8(D)Leu, H15(P)Val and H19(P)Leu; and Beta B13(D)Leu, C7(E)Phe, CD1(D)Phe, CD3(E)His, CD4(D)Phe, E7(D)His, E10(D)Lys, E11(D)Val, E14(D)Ala, F4(P)Leu, F7(P)Leu, F8(P)His, FG3(P)Leu, FG5(P)Val, G4(P)Asn, G5(P)Phe, G8(D)Leu, H15(P)Val, and H19(P)Leu. See Fermi, et al., J. Mol. Biol., 175: 159-174 (1984). (In the above list, "P" denotes "proximal", "D", "distal", and "E", "edge-on".)

These residues are therefore candidates for modification. Consideration should also be given to pairs contacted through bound water molecules. See Ladner, et al., Mol. Biol., 114: 385-414 (1977).

Mutations of beta residues 42 (CDI), 45 (CD4) and 70 (E14) are of particular interest. Other beta residues of interest include 43 (CE1), 46 (CE4), 58 (E7), 61 (E10) and 62 (E11). Alpha residues of interest include 43 (CE1), 46 (CE4), 58 (E7), 61 (E10) and 62 (E11).

In general, mutations around the heme-O.sub.2 binding site that interfere with O.sub.2 binding are also desirable because of their low-affinity O.sub.2 binding nature. Replacing residues that are adjacent to the face of heme that binds O.sub.2 can result in lower affinity. A naturally occurring mutant Hg Bristol (beta.sup.67 val.fwdarw.asp) has been described with low affinity. Other mutants that are desirable are the beta ile.sup.67 described herein, beta asp.sup.67 and beta glu.sup.67. Other residues are also in the vicinity of the O.sub.2 binding site. Histidine E7 (beta his.sup.63) can be replaced with phe which results in very low O.sub.2 affinity. The other likely residue for mutation is beta phe.sup.42 ; replacement with trp is likely to result in low O.sub.2 affinity. The corresponding residues of the alpha chain may be altered instead of, or in addition to, these preferred mutations of the beta chain.

The oxygen affinity and cooperativity of hemoglobin depend on the relative stabilities of the T (low affinity) and R (high affinity) quaternary states. These two states are in equilibrium with each other but one state may be favored by mutations at the alpha.sub.1 beta.sub.2 contact or in the central cavity. There are many naturally occurring mutations at these sites, and careful study of these should be of value in designing a hemoglobin molecule with the desired properties.

Thus alpha 1 residues 37(C2)Pro, 38(C3)Thr, 40(C5)Lys, 41(C6)Thr, 42(C7)Tyr, 44(CD2)Pro, 88(F9)Ala, 91(FG3)Leu, 92(FG4)Arg, 94(G1)Asp, 95(G2)Pro, 96(G3)Val, 97(G4)Asn, 140(HC2)Tyr, and 141(HC3)Arg are all known to lie within 4 angstroms of at least one residue of the beta2 chain of human deoxyhemoglobin. Similarly, beta2 residues 146(HC3)His, 145(HC2)Tyr, 105(G7)Leu, 102(G4)Asn, 101(G3)Glu, 100(G2)Pro, 99(G1)Asp, 98(FG5)Val, 97(FG4)His, 43(CD2)Glu, 41(C7)Phe, 40(C6)Arg, 37(C3)Trp, 36(C2)Pro, 35(C1)Tyr, and 34(B16)Val lie on the other side of the alpha1beta2 interface of human deoxyhemoglobin.

Mutation of beta102(G4)Asn is particularly preferred. Hg Kansas is a known low affinity mutant in which this residue is altered to Thr. As previously mentioned, the oxy state is stabilized by a hydrogen bond between beta Asn102 and alpha Asp94. The preferred mutation would disrupt this hydrogen bond.

Another favored mutation at this site is to beta Asp102. The negative charge-negative charge repulsion between this group and the alpha Asp94 would further destabilize the oxy state.

Still other preferred mutations at the beta102 locus would be to Ser (Hb Beth Israel), Lys (Hb Richmond) and Tyr (Hb St. Mande).

Inside the erythrocyte the cytochrome b.sub.5 and glutathione reduction systems serve to maintain hemoglobin in the active ferrous form. Free hemoglobin in the bloodstream would rapidly be at least partially oxidized to the ferric state since there is no such means of keeping it in the reduced form outside the red cell. It is possible to stabilize the ferrous form by replacing the Val-Ell residue with a large aliphatic amino acid such as Ile or Leu. A large side chain at this position prevents electron donors from reaching the iron atom and so slows the rate of autoxidation. The His63.fwdarw.Phe mutant is also of interest.

If hemoglobin is cross-linked both intermolecularly to form higher molecular weight aggregates and intramolecularly to prevent dissociation into dimmers then it will neither react with haptoglobin nor pass through the glomerular membrane of the kidney. If the oxygen binding properties of haemoglobin are to be unaffected it is important that the cross-linking does not prevent conformational change of the protein. This is because the heme-heme interaction arises from a reversible transition between tow quaternary structures: the T structure with low oxygen affinity. These two structures with high oxygen affinity. These two structures differ in the contact between the alpha.sub.1 and beta.sub.2 subunits. Therefore this contact should be allowed to undergo the conformational changes accompanying oxygen association and dissociation.

Site-directed mutagenesis can be used to replace certain surface residues of haemoglobin with cysteinyl residues. Protein engineering techniques similar to those used to introduce new disulfide bridges to lysozyme, subtilisin and dihydrofolate reductase can be used. Hemoglobin molecules carrying such --SH groups can be cross-linked by disulfide bridges or via a bifunctional thiol reagent. It should also be noted that there is a natural mutant (Hb Ranier) in which His 143 beta is replaced by Cys and the newly introduced cysteine forms a disulfide bridge with Cys 93 beta in vivo. This mutant is more stable than native Hb.

Table 2 is a list of candidate, non-naturally occurring hemoglobins which are expected to exhibit a lower affinity for oxygen than does conventional hemoglobin.

After determining the amino acid sequence changes which distinguish the desired hemoglobin from conventional hemoglobin, it is necessary to design the expression vector. The most convenient starting point is a nucleotide sequence which codes upon expression for conventional hemoglobin. This sequence then may be modified by site-specific mutagenesis.

Techniques of site-specific mutagenesis are well known, and this invention is not limited to any particular technique. The two principal techniques are the gapped duplex A.A., Kruse, K. B., Brown, J.L. BioTechniques 6, 338-339, 1988) and M-13 (Zoller, M. J. and Smith, M. Meth. Enz. 100, 468-500, 1987) methods.

Alternatively, a subsequence with the desired mutation may be synthesized and then ligated to other subsequences to form the desired molecule.

The gene must be placed under the control of a promoter. Either a constitutive or an inducible promoter may be used; the advantages and disadvantages of each are well known in the molecular biology art. A promoter must be chosen which is functional in the host. Thus, a bacterial promoter would be chosen if the expression vector is to be introduced into a bacterial host, a yeast promoter, if a yeast host, and a mammalian promoter, if a mammalian cell host. It should be understood that the promoter of a viral gene of a virus which infects the chosen host cell may also be used. The present invention does not depend upon the choice of either the promoter or the host. However, it is desirable to choose a host so that the subsequent purification of the mutant hemoglobin is not unduly complicated.

For the same reason, it is preferable, but not required, that the mutant hemoglobin be expressed as a moiety of a fusion protein. Attempts to express the alpha chain other than as a fusion protein were largely unsuccessful. However, the entire alpha chain sequence could be expressed as part of a fusion protein including a portion of the beta chain, and separated from the latter by a spacer providing a selective cleavage site. The hemoglobin is obtained merely by placing the secreted fusion protein in an appropriate cleavage environment. Numerous fusion protein/cleavage system combinations are now known in the art.

After expressing, recovering and purifying the mutant hemoglobin, its P.sub.50 is measured in accordance with the protocol set forth in the Reference Example. Preferably, its P.sub.50 is at least 10% higher than that of conventional hemoglobin A when measured in the same environment.

While it is not practical to combine stroma-free conventional hemoglobin with sufficient 2,3-DPG to restore its P.sub.50 to intra-erythrocyte levels, it may be possible to combine a modestly right-shifted mutant hemoglobin with a small amount of 2,3-DPG or functionally similar organic phosphate (e.g., pyridoxal phosphate or ATP-dialdehyde) so as to mimic the oxygen-carrying capacity of whole blood. The half life of the organic phosphate may be improved by encapsulating the organic phosphate and the mutant hemoglobin in a liposome to obtain a "simulated erythrocyte," or by covalently attaching the organic phosphate group to the hemoglobin.

Our preferred method of measuring the P.sub.50 of purified hemogoblin solutions for the purpose of the appended claims is as follows.

Hemoglobin-oxygen equilibrium data are measured using a thin film technique (Imai, K. Meth. Enz. 76, 438-449, 1981). A sample of hemoglobin (0.6 mM) in buffer (50 mM Bis-Tris or 100 mM HEPES), pH 7.4, 0.1M NaCl, is applied to the cell and then equilibrated at 25.degree. C. The hemoglobin is saturated with O.sub.2 by a stream of air, or air/O.sub.2 if the hemoglobin has low O.sub.2 affinity or if the local barometric pressure prevents ambient oxygen from achieving a high enough partial pressure to saturate the molecule. Deoxygenation is achieved by closing the cell to O.sub.2 flow and flushing the cell with N.sub.2 (>99.98% pure). The oxygen equilibrium curve is obtained by plotting the change in absorbance at 560 nm against the pressure of O.sub.2 in the cell. Percent saturation is determined by measuring the A.sup.560 at a given partial pressure (i) divided by the A.sup.560 of the beginning, totally saturated Hgb solution Hgb solution [A.sup.560 (i)/A.sup.560 (100%).times.100 =% saturation]. The P.sub.50 is defined as the partial pressure (i) of O.sub.2 required to cause 50% saturation of O.sub.2 binding sites.

P.sub.50 may also be measured under other conditions, but it should be noted that many environmental factors affect hemoglobin's oxygen affinity. The effect of Ph, CO.sub.2, inorganic anions, organic phosphates and temperature on P.sub.50 are discussed in Bunn and Forget, HEMOGLOBIN:MOLECULAR, GENETIC AND CLINICAL ASPECTS 37-47, 95-98 (W. B. Saunders Co.: 1986).

Since many determinations of whole blood oxygen binding curves are made under standard physiologic conditions (37.degree. C., pH=7.4, PCO.sub.2 =40 mm Hg), it may be necessary to adjust literature figures. In this context, it should be noted that a 10.degree. C. increase results in nearly a two-fold increase in P.sub.50, while the dependence of P.sub.50, while the dependence of P.sub.50 on pH is approximately given as delta log P.sub.50 /delta pH=-0.5.

Comparing P.sub.50 values of purified Hb preparation to P.sub.50 values of whole blood can be problematic. Whole blood, or isolated RBC's, contain many components that naturally modulate the shape of the hemoglobin-oxygen binding curve. The RBC encapsulates Hgb in the presence of a high concentration of the effector molecule 2,3-DPG; a molecule that causes Hgb to have a markedly lower affinity for O.sub.2. Other intra-erythrocyte components also affect the shape of the binding curve: ATP, Cl.sup.-, CO.sub.2, H.sup.+, orthophosphate, methemoglobin and carboxyhemoglobin. These substances are not normally present in purified HgB solutions and thus, the P.sub.50 value of purified Hgb is lower than that found in whole blood. One very important modulator of Hgb-oxygen affinity is Cl.sup.- ion. Cl.sup.- ion is found outside the erythrocyte in the blood serum at a physiologic concentration of approximately 0.15M. Since Cl.sup.- causes a lower O.sub.2 affinity, a Hgb solution with a P.sub.50 measured in vitro may well have much lower O.sub.2 affinity if infused into the blood stream. Another problem with measuring O.sub.2 binding of whole blood is that RBCs are quite fragile and in the process of manipulating the erythrocyte into the instrument used to measure the O.sub.2 binding it is inevitable that at least a small percentage of the RBCs will lyse. Lysed RBCs leak Hgb into the surrounding media away from 2,3-DPG; hence, since free Hgb has a higher affinity than intraerythrocyte Hgb, lysed RBCs will have a higher O.sub.2 affinity and can cause a falsely low P.sub.50 value for whole blood P.sub.50 determinations. It is widely accepted that under physiologic conditions whole blood has a P.sub.50 value of 26-28 mmHg. When Hgb is isolated from whole blood, however, the measured P.sub.50 is on the order of 1-10 mmHg depending on the investigators experimental conditions. For these reasons it is most accurate to measure Hgb-oxygen equilibria with purified Hgb molecules under strict conditions of buffer, pH and salt concentration. Unfortunately, there are no accepted "standards" for all investigators to measure Hgb oxygen binding in in vitro systems.

Still, as many mutant hemoglobins are first identified in patient's whole blood, one would like to be able to compare the relative affinities of native and mutant Hgb for O.sub.2, between whole blood and purified Hgb preparations. An example of this is Hgb Chico (beta lys.sup.66 .fwdarw.thr) (Table 1). If one examined only the P.sub.50 value of the purified mutant Hgb (10.1 mmHg) one would note that Hgb has a P.sub.50 value less than that for normal whole blood (27.2 mmHg). Still, when that hemoglobin is measured in RBCs under physiologic conditions it is apparent that it does have a higher P.sub.50 than normal whole blood (38 mmHg). One cannot predict the degree that the P.sub.50 value will change going from whole blood Chico to a purified Hgb Chico if it were infused into the bloodstream as a blood substitute. One can conclude, however, that the P.sub.50 will be higher than it is in pure form, and that by reacting the mutant Hgb with organic phosphates that P.sub.50 will be even higher.

Note also that whole blood oxygen binding curves are customarily determined under standard physiologic conditions (37.degree. C., pH 7.4, PCO.sub.2 =40 mmHg) and red cell 2,3-PDG varies with age, sex and condition.

M13 mpll FX encodes a sequence (Ile-Glu-Gly-Arg) including the recognition site for factor X.sub.a. This M13 derivative can be used to join any coding sequence to the factor X.sub.a recognition sequence. See Nagai, EP Appl 161,937 (CELLTECH LTD.) However, this invention is not limited to the use of either M13 or of the factor Xa cleavage system.

All DNA manipulations were carried out essentially as described by Maniatis et al (`Molecular Cloning` Cold Spring Harbour, N.Y., 1982). A temperature-sensitive lysogenic strain MZ-1 (galK.sub.am 8attL BamN.sub.7 N.sub.53 cI857 Hl, his-, ilv-, bio-, N.sup.+, a gift from Dr. K. McKenney and available on request from Medical Research Council) was used as a host strain for plasmids containing lambda P.sub.L promoter and transformation was carried out by the method of Remaut et al (Gene 15, 81-93 (1981)). Other promoters and host strains could have been employed.

T4 DNA ligase was prepared from strain NM989 (Murray et al, J Molec Biol 132, 493-505 (1979) and Tait et al, J Biol Chem 255, 813-815 (1980)). Restriction enzymes were purchased from New England BioLabs.

Two oligonucleotides dTACCCTCGATGGATC and dCATCGAGGGTAGGCC were synthesized by a phosphotriester method on a controlled pore glass support (Sproat et al, Tetrahedron Lett, 24, 5771-5774 (1983)) and purified by HPLC (Gait et al, Nucleic Acids Research 10, 6243-6254 (1982)). These oligonucleotides encode the peptide sequence (gly)-ser-ile-glu-gly-arg in a BamH1-Stu1 linker. The two oligonucleotides were allowed to anneal after phosphorylation with T4 polynucleotide kinase (P-L, Biochemicals) and r[gamma.sup.32 P]ATP (3000 Ci/m mol, Amersham) and ligated to form concatamers. The DNA was then digested with Bam HI and cloned into the dephosphorylated Bam HI site of M13 mpll (Vieira et al, Gene 19, 259-268 (1982)) to yield M13 mpll FX, as shown in FIG. 1a, which forms blue plaques in the presence of isopropyl-beta-D-thiogalacto-pyranoside and 5-bromo-4-chloro-3-indolyl-beta-d-galactoside (Sigma).

40 micro g of cloned human alpha-globin cDNA were digested with the restriction enzymes Nco I and Apa I. The single-stranded ends of the excised alpha-globin fragment were trimmed by incubation at 0.degree. C. for 10 minutes with 200 units of mung bean nuclease (P-L Biochemicals) in 30 mM sodium acetate pH 4.6, 50 mM sodium chloride, 1 mM zinc chloride, 5% glycerol. The alpha-globin sequence was then cloned into the Stu I site of M13 mpll FX described above (Nagai & Thogersen, Nature 309, 810-812). The DNA sequences of several clones were determined (Sanger et al Proc. Natl. Acad. Sci. USA, 74, 5463-5467 (1977)) and a clone in which the first valine codon of the alpha-globin gene was joined to the DNA sequence encoding the factor X.sub.a recognition site (Ile-Glu-Gly-Arg) was named mpll FX alpha-globin.

Plasmids pLcIIFX beta and pLcII beta direct efficient production of a hybrid protein consisting of the 31 amino-terminal residues of the lambda CII protein and the complete human beta-globin, with and without the factor X.sub.a cleavage site, respectively.

The Eco-Hind III fragment containing the multi-restriction sites was cut out from M13 mp10 (Vieira eta al, supra) and ligated to Eco RI-Hind III cut pLc245 (Remaut et al, supra) to form pLmp10. The 319 bp Alu I fragment containing the nutR, t.sub.R1 sites and a part of the cII gene was cut out from pKG1805 (McKenney, K PhD Dissertation, The Johns Hopkins University (1982)) and cloned into the Sma I site of M13 mp10 in the same orientation with respect to the beta-galactosidase alpha-peptide gene. The Eco RI-Hind III fragment containing the lambda DNA sequence was then cut out and cloned into the Eco RI-Hind III site of pLmp10 to yield pLcII.

A complete human beta-globin cDNA sequence was reconstructed by joining restriction fragments prepared from an incomplete cDNA clone (pJW102) (Wilson et al, Nucleic Acids Research 5, 563-581 (1978)) and a genomic DNA clone (Lawson et al, Cell 21, 647-651 (1980)) and cloned into the Sma I-Hind III site in M13 mp9. M13 mp9 beta cDNA thus obtained was opened at the Nco I site which is located at the initiation codon and treated with Klenow DNA polymerase (Boehringer Mannheim) in the presence of 100 micro M 4dNTP to obtain flush ends. The beta-globin cDNA sequence was then cut out with Hind III and inserted into the Bam HI (filled-in)-Hind III site of pLcII so that the beta-globin gene was fused to the lambda cII gene in phase via a small linker DNA derived from M13 mp10.

In order to construct pLcIIFX beta, M13 mp9 beta cDNA was opened with Nco I and 40 micro g of DNA was treated with 200 units of mung bean nuclease (P-L Biochemicals) in 30 mM Na-acetate pH 4.6, 50 mM NaCl, 1 mM ZnCl.sub.2, 5% glycerol at 0.degree. C. for 10 min to remove the 5' protruding end. The beta-globin cDNA sequence was cut out with Hind III and cloned into the Stu I-Hind III cut M13 mpll FX. The DNA sequence was determined by the dideoxy chain termination method (Sanger et al, PNAS 74, 5463-5467 (1977)) to ensure that the first valine codon of the beta-globin gene was preceded by the DNA sequence coding for Ile-Glu-Gly-Arg. Then, the Bam HI fragment containing a part of the beta-globin sequence was cut and cloned into Bam HI digested pLcII beta to form pLcIIFX beta, as shown in FIG. 1b.

M13 mpll FX beta-globin DNA was prepared in single-stranded form and a BglII site was introduced into the beta-globin sequence using a mutagenic oligodeoxynucleotide, dACCAACTTCAGATCTGTTACCTTG, called KN83, to obtain mpll cII FX beta FX. The replicative form of this mutant clone was digested with SacI and Hind III, and the resulting cII FX beta FX fragment was cloned into Sac I/Hind III cut pLmpII to form pLcII FX beta FX. This recombinant plasmid was digested with BglII and the 5' terminal phosphate groups of the linear DNA were removed with calf intestinal alkaline phosphatase. The replicative form of mpll FX alpha-globin was digested with BamHl and the FX alpha-globin containing fragment was ligated with the linearised pLcII FX beta FX to form pLcII FX beta FX alpha. This plasmid encodes a fusion protein consisting of the 31 amino-terminal residues of the lambda phage cII protein, the tetrapeptide Ile-Glu-Gly-Arg, the 20 amino-terminal residues of human beta-globin, the tetrapeptide Ile-Glu-Gly-Arg and human alpha-globin at the carboxyl end. Transcription of the fusion protein gene is initiated at the lambda P.sub.L promoter and is regulated by lambda repressor.

A defective lambda phage lysogenic strain of E.coli QY13 (a gift from S. Brenner, and available on request from Medical Research Council) harboring pLcIIFX beta FX alpha-globin or pLcIIFX beta-globin was grown at 30.degree. C. in 2xTY medium (16 g tryptone, 10 g yeast extract and 5 g sodium chloride-liter) in the presence of ampicillin (25 micro g/ml). When the optical density (600 nm) reached 1.5-1.6, the temperature was quickly raised and maintained at 42.degree. C. for 15 min, followed by further incubation at 37.degree. C. for 3-4 hr. The cells were harvested and frozen in liquid nitrogen.

The cells (100g) were thawed and suspended in 80 ml of 50 mM Tris-HCI (pH 8.0)/25% sucrose (wt/vol)/1 mM EDTA and lysed by addition of lysozyme (200 mg). Then, MgCl.sub.2, MnCl.sub.2 and DNase I were added to final concentration of 10 mM, 1 mM and 10 micro g/ml, respectively. After 30 min incubation 200 ml of 0.2 M NaCl/1% deoxycholic acid/1.6 % Nonidet P-40 (vol/vol)/20 mM Tris-HCl(pH7.5)/2 mM EDTA were added to the lysate, which was then centrifuged at 5000.times.g for 10 min. Then the pellet was suspended in 0.5 % Triton X-100/1 mM EDTA and centrifuged. This procedure was repeated until a tight pellet was obtained. The protein pellet was finally dissolved in 8 M urea/ 25 mM Tris-HOAc (pH5.0)/1 mM EDTA/1 mM dithiothreitol in the case of cIIFX beta-globin fusion protein. In the case of cIIFX beta-FX-alpha-globin fusion protein, the pellet was first dissolved in 6M guanidine hydrochloride/25 mM Tris-HOAc (pH 5.0)/1 mM EDTA/1 mM dithiothreitol.

The fusion protein solution was then applied to a 4.times.10 cm CM-Sepharose (Pharmacia) column equilibrated with the same buffer. The fusion protein was eluted with a linear gradient formed with 500 ml of 8M urea/25 mM Tris-HOAc pH 5.0/1 mM EDTA/1 mm dithiothreitol and 500 ml of the same buffer with 0.2M NaCl. The fusion protein was further purified on a 5.times.60 cm Sephacryl S-200 column equilibrated with 5M guanidine-HCl/50 mM Tris-HCl/1 mM EDTA/1 mM dithiothreitol to remove any trace of impurities. The combined fraction was extensively dialyzed against 50 mM Tris-HCI (pH 8.0)/0.5 M urea/1 mM CaCl.sub.2.

The cIIFX beta FX alpha-globin or cIIFX beta-globin fusion protein was incubated at 0.degree. C., with blood coagulation factor X.sub.a that had been activated with Russell's viper venom immobilized on cyanogen bromide-activated Sepharose-6B. 100 micro 1 aliquots were removed after intervals of 2, 5, 15, 30, 60 and 120 minutes. 100 ul of protein sample buffer (Laemmli, 1970) and 1 ul of 100 mM DTT were added to each aliquot, which was then boiled before being applied to an SDS polyacrylamide gel. Factor X.sub.a cutting of the cIIFX beta FX alpha fusion protein gives rise to a number of polypeptide products. This is due to the presence of two recognition sites within the protein. Complete digestion releases three polypeptides, a cII protein fragment and a beta-globin fragment, both with the tetrapeptide Ile-Glu-Gly-Arg at the carboxyl terminus, and the desired alpha-globin. Partial digestion of the fusion protein gives two other products.

Factor Xa cleavage of cIIFX-beta yields two products, a cII fragment and the desired beta globin.

25 mg of hemin-Cl was dissolved in 2.5 ml of 0.1 N KOH and 30 diluted with 20 ml of water and 2.5 ml of 1 M KCN. The native beta chain was diluted in 20 mM K phosphate buffer pH 5.7, 1 mM EDTA, 1 mM dithiothreitol (DTT) and bubbled with CO. Alpha globin produced in E.coli was dissolved in 8 M urea/50 mM Tris-CI pH 8/1 mM EDTA/1 mM DTT at the concentration of 5 mg/ml and incubated at room temperature for 3 hours. The alpha-globin solution was added dropwise to 20 volumes of 30 mM K phosphate buffer pH 5.7, 1 mM EDTA, 1 mM DTT with gentle stirring. The hemin-dicyanide solution (1.2 equivalent to the alpha-globin) was added dropwise to the alpha-globin solution and the beta chain was added in slight excess. The semi-artificial Hb was dialyzed overnight against 0.1 M K phosphate pH 7.6 1 mM EDTA, 1 mM KCN.

Beta-globin (100 mg) was dissolved in 8 M urea, 50 mM Tris-CI pH 8.0, 1 mM DTT, 1 mM EDTA at the concentration of 5 mg/ml and incubated at room temperature for hr. The beta-globin solution was added dropwise to 16 volumes of alpha chain solution (either isolated from Hgb A, or produced by recombinant means) (3.2 mg/ml) in 10 mM Tris-CI pH 8.0. The hemindicyanide solution (1.2 equivalents to beta-globin) was added dropwise with gentle stirring. The semi-artificial Hb was dialyzed against 2 changes of 0.1 M K phosphate pH 7.4 1 mM EDTA, 1 mM KCN.

The lyophilized recombinant alpha and beta globins were dissolved in 8M urea/50 mM Tris-Cl, pH 8.0/1 mM EDTA/1 mM DTET, diluted to a concentration of 5 mg/ml and incubated at room temperature for 3-4 hours. The alpha globin was then diluted to 0.3 gm/ml with chilled 20 mM K.sub.2 HPO.sub.4, pH 5.7/1 mM EDTA/1 mM DTT. Hemin (25 mg) was dissolved in 2.4 mg 0.1 M KOH, diluted with an equal volume of 1 M KCN; this solution was then made 0.1 mg.ml in hemin and 20 mM K.sub.2 HPO.sub.4, pH 6.7 with stock phosphate buffer. Hemin from this solution was added to a 2.8 molar excess to chilled alpha-globin; and equal molar amount of beta-globin was added and the solution was dialyzed at 420 C. overnight against 0.1 M K.sub.2 HPO.sub.4, pH 7.6/1 mM EDTA/1 mM KCN.

The artificial Hb was concentrated by ultra-filtration using diaflo PM-10 membrane (Amicon) and transferred into a 200 ml screw-top test tube with a rubber septum. The hemoglobin solution was deoxygenated by evacuation and flushing with N.sub.2, and then the solution was saturated with CO. 100 mM Sodium dithionite solution was prepared anaerobically in a 20 ml screw-top test tube with rubber septum. 4.5 equivalents of dithionite were added to the Hb solution with a syringe, and the mixture incubated on ice for 15 min.

The Hb solution was gel-filtered against 10 mM Na phosphate buffer pH 6.0 on a 4.times.40 cm Sephadex G-25 (fine) column. The Hb was then applied to a 2.times.10 cm CM-52 (Whatman) column equilibrated with the same buffer and the chromatography was developed with a linear gradient of 500 ml 10 mM Na phosphate buffer pH 6.0 and 500 ml of 70 mM sodium phosphate buffer pH 6.9. CO was removed from Hb by photolysis under a stream of oxygen. This Hb shows native oxygen binding properties.

A wholly artificial hemoglobin may be prepared by the combination of alpha-globin and beta-globin both produced in E.coli, or any other host of a non-erythroid nature, with a source of heme.

A synthetic oligonucleotide of sequence dGGAGCCTGAAAGTCTCAGGA was designed from published mRNA sequence information [Bunn & Forget, eds., Hemoglobin: Molecular, Genetic and Clinical Aspects, W. B. Saunders Co., Philadelphia, PA, 169-222 (1986)]and synthesized on a controlled glass support. The oligonucleotide was gel purified [Lloyd et al., BioTechniques 4, 8-10 (1986)]and used to prime the site-specific mutagenesis of M13 mp10 cIIFXbeta -globin by the methods of Zoller and Smith [Methods in Enzymology 100, Academic Press, New York, 468-500 (1983)].

The mutagenic oligonucleotide was complementary to the beta-chain structural gene sequence flanking and centered upon the wild-type codon for Asn.sup.102. At this triplet specific base substitutions were designed into the oligonucleotide to specify ACT-Thr.sup.102, the amino acid sequence alteration which is characteristic of the Kansas beta-globin mutant [Bonaventura & Riggs, J. Biol. Chem. 243, 980-991 (1968)]. The particular Thr codon employed in this substitution, while differing from that found in the original Kansas isolate (Bonaventura and Riggs, supra), is preferentially utilized in highly expressed E. coli genes [Grantham et al., Nucleic Acids Res 9, r43-r74 (1981)].

The products of the in vitro mutagenesis reaction were transformed into competent E. coli .sup.M Z-1 [galK.sub.am 8attL BAmN.sub.7 N.sub.53 cI857 H1, his.sup.-, ilv.sup.-, bio.sup.- N.sup.+, a gift of Dr. K. McKenney and available on request from the Medical Research Council), by standard procedures of CaCl.sub.2 -shock [Maniatis et al., Molecular Cloning, Cold Spring Harbor Laboratory, New York, 250-251 (1982); Nagai & Thogersen, Methods in Enzymology, supra].

Transformants harboring the desired mutant M13 bacteriophage constructs were then identified by differential plaque-hybridization screenings at high stringency using gamma [.sup.32 P]-end-labeled oligonucleotides as probes.

The reactants used to prepare each of the phosphorylated hybridization probes were 300 pM (2 ug) of oligonucleotide, 100 pM (0.7 mCi) gamma-[.sup.32 P]-ATP (specific activity approximately 6000 Ci/mM), and 15 units T4 polynucleotide kinase in a total reaction mixture of 50 ul. After 37o C incubation for 2 h, the end-labeled oligomers were purified away from orthophosphate and unincorporated precursor nucleotide using reverse phase C-18 sep-paks (Waters Associates Milford, MA). This latter procedure involved loading the phosphorylation reaction mixture onto the C-18 cartridge in an aqueous salt solution, eluting orthophosphate and unincorporated ATP with water followed by 10% methanol, and then eluting the purified oligomer with 60% methanol. The probes employed comparatively in the differential hybridization analyses were both the mutagenic oligonucleotide and another 20-mer (dGGAGCCTGAAGTTCTCAGGA) which is perfectly complementary to the wild-type beta-chain DNA sequence in the same coding region.

After identification and plaque purification (Zoller & Smith, supra) of several of the desired MI3 phage constructs, one of the resultant Thr.sup.102 mutants, termed M13 mp10 cIIFX beta-globin (Thr.sup.102), was further verified by DNA sequence analysis [Sanger et al., Proc. Natl. Acad. Sci. USA 74, 5463-5467 (1977)]to contain the desired mutation at codon-102, and only that particular amino acid sequence alteration in the beta-chain structural gene coding sequence.

A large scale preparation of M13 mp10 cIIFX-globin(Thr.sup.102) RF DNA was conducted as follows [Recinos, Ph.D. Dissertation, Vanderbilt University (1987)]. The host E. coli was grown at 37.degree. C. overnight in M9 minimal medium (Maniatis et al., supra) plus 2 ug/ml thiamine. 0.3 ml of this cell culture was then diluted (1:50) into 14.7 ml 2X YT medium and growth at 37.degree. C. was continued for an additional 2 h. The latter culture was again diluted (1:10) into a final volume of 150 ml 2X YT, and this cell solution was inoculated with the plaque purified mutant M13 phage construct at a multiplicity of infection of approximately one. This phage infection was then shaken vigorously at 37.degree. C. for 14 h, and cells for RF preparation were harvested by centrifugation (5000.times.g, 10 min, 4.degree. C.). The mutant phage supernatant was stored at -20.degree. C. for use in scaled-up versions of the protocols (Zoller & Smith, supra) for phage purification and for single-stranded template preparation.

Double-stranded RF DNA was purified from the cell pellets as follows. Pellets were frozen in an alcohol dry-ice bath for 10 min, thawed at 20.degree. C. and completely resuspended on ice in 10 ml 25% sucrose, 50 mM Tris-HCl (pH 8.0). Lysozyme was added to a final concentration of 4 mg/ml, and incubation was continued on ice for 5 min. EDTA was then added to a final concentration of 80 mM, and again after 5 min on ice, an equal volume of 0.5% Triton X-100, 50 mM Tris-HCl (pH 8.0), 62.5 mM EDTA was added. This solution was kept on ice for 15 min more, and then 5 M NaCl was added to a final concentration of 1 M. The last solution was loaded into Beckman Type 70 Ti rotor bottles, and after a further incubation on ice for 3 h, was centrifuged at 40,000 rpm for 75 min at 15.degree. C. RF DNA was decanted with the supernatant and precipitated at -20.degree. C. for 20 min with the addition of an equal volume of isopropanol. DNA precipitates were pelleted and resuspended in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA, 100mM NaCl, and treated with RNase (final conc. 100 ug/ml) for 2 h at 37.degree. C. This solution was phenol and chloroform extracted (one time each), and the DNA was ethanol precipitated and resuspended in 30 ml 10 mM Tris-HCl (pH 8.0), 1 mM EDTA. The DNAs were twice banded by CsCl-ethidium bromide density gradient equilibrium centrifugation. Form I DNA bands were extracted 4.times. with CsCl-saturated isopropanol to remove ethidium bromide, and DNA and CsCl were ethanol precipitated. CsCl was removed from DNA pellets by resuspension in and dialysis against 10 mM Tris-HCl (pH 8.0), 0.2 mM EDTA. A final ethanol precipitation and resuspension in 0.5 ml Tris-HCl (pH 8.0) yielded 150 ug of purified M13 mp10 cIIFXbeta-globin(Thr.sup.102) RF DNA for use in subcloning the mutant beta-globin structural gene into the beta-globin expression construct.

The mutated beta-chain sequence was moved into the beta-chain expression vector, pLcIIFXbeta-globin (nic.sup.-), by the following procedures. Mutant clone RF (50 ug) was restriction enzyme digested with Sac I and Hind III, and the resultant cIIFX -globin (Thr-102) fragment was isolated from a 1% preparative agarose gel (Maniatis et al., supra) and gel-purified by a phenol freeze-thaw procedure [Benson, BioTechniques 2, 77-78 (1984)]. Approximately 200 ug of expression vector DNA was isolated and purified from E. coli QY13 transformant cell pellets by methods nearly identical to those described above for the Rf preparation. This plasmid DNA (20 ug) was similarly restricted with Sac I and Hind III and further treated with bacterial alkaline phosphatase (Bethesda Research Laboratories, Gaithersburg, MD, as directed) to remove the 5' phosphates from the vector DNA, thereby preventing recircularization without insert DNA in subsequent ligation reactions.

The purified mutant insert DNA fragment was then ligated back into the expression construct at moderate insert end-concentration, replacing the wild-type protein coding sequences previously therein. Ligation reaction conditions (modified from procedure of New England Biolabs, Inc., Beverly, MA) were: 11 ug vector DNA and 2.1 ug purified insert DNA in 500 mM Tris-HCl (pH 7.8), 100 mM MgCl.sub.2, 6 mM ATP, 2 mM dithiothreitol; total reaction volume 125 ul. Incubation for the ligation was at 16.degree. C. for 10 h. The final ligation reaction mixture was used to transform competent E. coli QY13 with selection for ampicillin resistance. Transformants harboring the desired plasmid construct for the expression of mutant beta-chain were identified by differential colony hybridization screenings [Grunstein & Hogness, Proc. Natl. Acad. Sci. USA 72, 3961-3965 (1975), with modifications] using the end-labeled mutagenic and wild-type oligonucleotides described above as probes. The correct plasmid construct was further verified by restriction analysis and by its expression of a protein which is chromatographed differentially from the wild-type cIIFX beta-globin fusion product by HPLC. The mutant beta-globin was produced, purified and combined with alpha-globin as described for native beta-globin.

Mutation of the val.sup.67 codon was introduced into the beta-globin cDNA sequence in MI3 mp10 cIIFXbeta-globin using the mutagenic primer (dGCACCGAGGATTTTCTTGCC) as above. The mutant beta-globin was produced, purified and combined with alpha globin as described for native beta-globin to obtain a mutant hemoglobin.

Mutation of the his.sup.63 codon was introduced into the beta-globin cDNA sequence in M13 mp10 cIIFXbeta-globin using the mutagenic primer (dTTCTTGCCGAAAGCCTTCA) as above. The mutant beta-globin was produced, purified and combined with alpha globin as described for native beta-globin to obtain a mutant hemoglobin.

Oxygen equilibrium studies for Hgb (beta phe.sup.63) Hgb (beta ile.sup.67) were performed in 0.05 M bis-Tris pH 7.4, 0.1 M NaCl, at 25.degree. C. using the automated recording apparatus of K. Imai (Meth. Enz. 76, 438-449, 1981 and for Hgb (beta thr.sup.102) in 0.1 M HEPES pH 7.4, 0.1 M NaCl using a thin layer optical cell (Gill, S. J. Meth. Enz. 76, 427-438, 1981). Results are shown in Table 3.

The purified hemoglobin is incorporated into a physiologically acceptable blood substitute solution. A preferred solution includes the following components:

Hgb (gm/1): 60-120

Sodium (mEq/1): 135-145

Potassium (mEq/1): 3.5-4.5

Chloride (mEq/1): 90-110

Preferably, the solution has a pH of 7.3-7.5, an osmolality of 280-310, and an oncotic pressure of 20-30 mm Hg. Osmolality is controlled by concentration of hemoglobin and of the electrolytes, as well as by the optional ingredient glucose (preferably 0-30 gm/l). The oncotic pressure is controlled by the concentration of the hemoglobin and by its degree of crosslinking. Agents, such as albumin (0-70 gm/l), dextran (0-100 gm/l) and polyethylene glycol (0-25 gm/l) may be added to increase oncotic pressure. Moreover, to reduce the degree of methemoglobin formation, anti-oxidant or free radical scavengers, such as mannitol (0-20 gm/l), glutathione (0-4 gm/l), ascorbic acid (0-0.3 gm/l) and vitamin E (0-100 IU/1) may be provided.

If a low oxygen affinity mutant hemoglobin is employed, it may be desirable or necessary to adjust the P.sub.50 of the solution to the preferred level by suitable choice of electrolytes, pH and other characteristics of the composition. Preferably, the final solution has a P.sub.50 of 24-32 torr under standard physiological conditions.

1) Hemoglobin 1987, 11, 241-308.

2) Ohba, Y.; Miyaji, T.; Matsuoka, M.; Yokoyama, M.; Numakura, H.; Nagata, K.; Takebe, Y.; Izumu, Y.; Shibata, S. Biochemi. Biophvs. Acta 1975, 405, 155-160.

3) Beretta, A.; Prato, V.; Gallo, E.; Lehmann, H. Nature 1968, 217, 1016-1018.

4) Knuth, A.; Pribilla, W.; Marti, H. R.; Winterhalter, K. H. Acta Haematol 1979, 61, 121-124.

5) Schneider, R. G.; Atkins, R. J.; Hosty, T. S.; Tomlin, G.; Casey, R.; Lehmann, H.; Lorkin, P. A.; Nagei, K. Biochem Biophys, Acta 1975, 400, 365-373.

6) Moo-Penn, W. F.; Bechtel, K. C.; Schmidt, R. M.; Johnson, M.H.; Jue, D. L.; Schmidt, D. E.; Dunlap, W. M.; Opella, S. J.; Boneventura, J.; Boneventura, C. Biochemistry 1977, 16. 4872-4879.

7) Moo-Penn, W. F.; McPhedran, P.; Bobrow, S.; Johnson, M. H.; Jue, D. L.; Olsen, K. W. Amer. J. Hematol 1981, 11, 137-145.

8) Idelson, L. I.; Didkowsky, N. A.; Casey, R.; Lorkin, P. A.; Lehmann, H. Nature 1974, 249, 768-770.

9) Gacon, G.; Belkhodja, O.; Wajcman, H.; Labie, D. Febs Lett 1977, 82, 243-246.

10) Blouquit.; Delanoe,-Garin, J.; Lacombe, C.; Arous, N.; Cayre, Y.; Peduzzi, J.; Braconnier, F.; Galacteros, F.; Febs Lett. 1984, 172, 155-158.

11) Dacie, J. V.; Shinton, N. K.; Gaffney, P. J.; Carrell, R. W.; Lehmann, H. Nature 1967, 216 663-665.

12) Keeling, M. M.; Ogden, L. L.; Wrightstone, R. N.; Wilson, J. B.; Reynolds, C. A.; Kitchens, J. L.; Huisman, T. H. J. Clin. Invest. 1971, 50, 2395-2402.

13) Bratu, V.; Larkin, P. A.; Lehmann, H.; Predescu, C. Biochem. Biophvs. Acta. 1971, 251, 1-6.

14) Ogata, K.; Ho, T.; Okazaki, T.: Dan, K.; Nomura, T.; Nozawa, Y.; Kajita, A. Hemoglobin 1986, 10, 469-481.

15) Yeager, A. M.; Zinkham, W. H.; Jue, D. L.; Winslow, R. M.; Johnson, M. H.; McGuffey, J. E.; Moo-Penn, W. F. Ped. Res. 1983, 17, 503-507.

16) Charache, S.; Brimhall, B.; Milner, P.; Cobb, L. J Clin. Invest. 1973, 52. 2858-2864.

17) Marinucci, M.; Giuliani, A.; Maffi, D.; Massa, A.; Giampolo, A.; Mavilio, F.; Zannotti, M.; Tantori, L. Biochem. Biophys. Acta. 1981, 668, 209-215.

18) Garel, M. C.; Hasson, W.; Coquelet, M. T.; Goosens, M.; Rosa, J.; Arous, N. Biochem. Biophys. Acta. 1976, 420, 97-104.

19) Shih, D. T.; Jones, R. T.; Shih, M. F. C.; Jones, M. B.; Koler, R. D.; Hemoglobin 1987, 11, 453-464.

20) Steadman, J. H.; Yates, A.; Huehns, E. R.; Brit., J. Haematol 1970, 18, 435-446.

21) Stamotoyannopoulos, G.; Parer, J. T.; Finch, C. New Eng. J. Med. 1969, 281, 915-919.

22) Anderson, N. L.; Perutz, M. F.; Stamatoyannopoulos, G. Nature New Biol. 1973, 243, 275-276.

23) Jones, R. T.; Brimhall, B.; Pootrakul, S.; Gray, G. J. Mol. Evol. 1976, 9, 37-44.

24) Konotey-Ahulu, F. I. D.; Gallo, E.; Lehmann, H.; Ringelhann, B. J. Med. Genet. 1968, 5, 107-111.

25) Schneider, R. G.; Hosty, T. S.; Tomlin, G.; Atkins, R.; Brimhall, B.; Jones, R. T. Biochem Genet. 1975, 13, 411-415.

26) Sugihara, J.; Imamura, T.; Nagafuchi, S.; Boneventura, J.; Boneventura, C.; Cashon, R. J. Clin. Invest. 1985, 76, 1169-1173.

27) Tatsis, B.; Sofroniadou, K.; Stergiopoulas, C. I. Blood 1976, 47, 827-832.

28) Merault, G.; Keclard, L.; Saint-Martin, C.; Jasmin, K.; Campier, A.; Delanoe Garin, J.; Arous, N.; Fortune, R.; Theodore, M.; Seytor, S.; Rosa, J.; Blouquit, Y.; Galacteros, F. Febs Lett. 1985, 184, 10-13.

29) Imar, K.; Morimoto, H.; Kotani, M.; Shibata, S.; Miyaji, T.1 Matsutomo, K. Biochem. Biophys. Acta. 1970, 200, 197-202.

30) Ahern, E.; Ahern, V.; Hilton, T.; Serjeant, G. D.; Serjeant, B. E.; Seakins, M.; Lang, A.; Middleton, A.; Lehmann, H. Febs Lett. 1976, 69, 99-102.

31) Boneventura, J.; Riggs, A.; J. Biol. Chem. 1968, 243, 980-991.

32) Nagel, R. L.; Lynfield, J.; Johnson, J.; Landeau, L.; Bookchin, R. M.; Harris, M. B. N. Eng. J. Med. 1976, 295, 125-130.

33) Arous, N.; Braconnier, F.; Thillet, J.; Blouquit, Y.; Galacteros, F.; Chevrier, M.; Bordahandy, C.; Rosa, J. Febs Lett. 1981, 126, 114-116.

34) Efremov, G. D.; Huisman, T. H. J.; Smith, L. L.; Wilson, J. B.; Kitchens, J. L.; Wrightston, R. N.; Adams, H. R.; J. Biol. Chem. 1969, 244, 6105-6116.

35) Turner, J. W.; Jones, R. T.; Brimhall, B.; DuVal, M. C.; Koler, R. D. Biochem. Genet. 1976, 14, 577-585.

36) Imamura, T.; Fujita, S.; Ohta, Y.; Hanada, M.; Yanase, T. J. Clin. Invest. 1969, 48, 2341-2348.

37) Moo-Penn, W. F.; Wolff, J. A.; Simon, G.; Vacek, M.; Jue, D. L.; Johnson, M. H. Febs Lett. 1978, 92, 53-56.

38) King, M. A. R.; Willshire, B. G.; Lehmann, H.; Marimoto, H. Br. J. Haem. 1972, 22, 125-134.

39) Ranney, H. M.; Jacobs, A. S.; Nagel, R. L. Nature 1967, 213, 876-878.

40) Minnich, V.; Hill, R. J.; Khuri, P. D.; Anderson M. E. Blood 1965, 25, 830-838.

41) Ohba, Y.; Miyaji, T.; Murakami, M.; Kadowaki, S.; Fujita, T.; Oimoni, M.; Hatanaka, H.; Ishikawa, K.; Baba, S.; Hitaka, K,; Imai, K. Hemoglobin 1986, 10, 109-126.