The Nicotinamide Adenine Dinucleotides as Allosteric Effectors of Human Hemoglobin*

The oxygen binding properties of human hemoglobin are appreciably altered by the nicotinamide dinucleotides NADH, NADP+, and NADPH, These cofactors are important in the control of many metabolic pathways and in providing reductive potential for a number of enzymatic reactions, including in vivo reduction of methemoglobin. Specific binding of these cofactors to hemoglobin and their potential for acting as allosteric modifiers of hemoglobin function have not been previously recognized. Detailed oxygen binding studies utilizing a thin-layer method suggest that the nicotin- amide dinucleotides bind with high affinity to the deox-yhemoglobin tetramer at the /3 chain anion-binding site and stabilize the low affinity “T-state” conformation. Stripped Hb A 20 is half-saturated at P60 is to 3.8,7.1, and respectively. The Bohr Hb A in 0.05 M HEPES buffer sensitive to these effectors and is raised from 0.25 to about 0.65

The Nicotinamide Adenine Dinucleotides as Allosteric Effectors of Human Hemoglobin* (Received for publication, April 21, 1986) Robert Cashon$, Celia Bonaventura, and Joseph Bonaventuras The oxygen binding properties of human hemoglobin are appreciably altered by the nicotinamide dinucleotides NADH, NADP+, and NADPH, These cofactors are important in the control of many metabolic pathways and in providing reductive potential for a number of enzymatic reactions, including in vivo reduction of methemoglobin. Specific binding of these cofactors to hemoglobin and their potential for acting as allosteric modifiers of hemoglobin function have not been previously recognized. Detailed oxygen binding studies utilizing a thin-layer method suggest that the nicotinamide dinucleotides bind with high affinity to the deoxyhemoglobin tetramer at the / 3 chain anion-binding site and stabilize the low affinity "T-state" conformation. Stripped Hb A in 0.05 M N-2-hydroxyethylpiperazine-N"2-ethanesulfonic acid (HEPES) buffer, pH 6.5, at 20 "C is half-saturated at a p 0 2 of 1.6 mm Hg. In the presence of 0.5 m M NADH, NADP+, or NADPH, the P60 is raised to 3.8,7.1, and 12.5 mm Hg, respectively.
The Bohr factor for stripped Hb A in 0.05 M HEPES buffer is sensitive to these effectors and is raised from 0.25 to about 0.65 by the addition of NADPH. The data suggest that routine use of these effectors in studies of human hemoglobin variants or the allosteric mechanism of Hb A be considered carefully. The relatively low intraerythrocytic levels of the nicotinamide dinucleotides in relation to hemoglobin dictate that these cofactors cannot significantly affect in vivo oxygen delivery. However, the converse is theoretically possible. The binding of the cofactors to hemoglobin and the preferential binding of their reduced forms may affect cofactor-dependent metabolic processes in red blood cells.
In early studies of hemoglobin function, it was not understood why hemolysates prepared by water lysis had oxygen affinities significantly higher than that of the intact red cell. In fact, dialyzed hemolysates had oxygen affinities nearly as high as those of myoglobin ( Fig. 1 in Benesch and Benesch, * This work was supported by Grant ESO 1908 from the National Institutes of Health and Grant N00014-83-K-0016 from the Office of Naval Research. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ Supported by Blood Resources Training Grant 5-T32-HL-07057 from the National Institutes of Health.
3 To whom correspondence should be addressed. B Partially supported by the National Science Foundation/Conselho Nacional de Ciencia e Technologia Cooperative Program. 1969). Although 2,3-diphosphogIycerate was reported to be present in high concentrations in human red blood cells as early as 1941 (Rapoport and Guest, 1941), it was not until 1967 that Chanutin and Curnish (1965a, 1965b and Benesch (1967a, 1967b) discovered that 2,3diphosphoglycerate was an important allosteric effector of hemoglobin function. It was apparent why earlier investigators were puzzled by the experiments done with lysed red cells: dilution of the hemoglobin caused the ratio of 2,3diphosphoglycerate to hemoglobin to decrease and dialysis led to still further decreases in 2,3-diphosphoglycerate. The physiological significance of the allosteric interaction of 2,3-&phosphoglycerate with human Hb A is that its preferential binding with the deoxy form of the molecule leads to a decrease in oxygen affinity, thereby facilitating oxygen delivery to the respiring tissues surrounding the capillary beds. In subsequent years, a wide variety of anionic effectors of human and non-human tetrameric hemoglobin function have been discovered, as reviewed elsewhere (Bonaventura and Bonaventura, 1980). The best known of these anionic effectors are organic and inorganic phosphates, chloride, and lactate. Carbon dioxide also lowers hemoglobin's oxygen affinity, not only by carbamino formation at the amino-terminal residues of the a and chains, but also by its action, in the form of bicarbonate, as an anionic effector of hemoglobin function (see Antonini and Brunori, 1971). To this list should be added another class of metabolically important compounds which are found within the red cell: the nicotinamide dinucleotides NADPH, NADP', and NADH. As will be shown in this paper, these compounds act as heterotropic allosteric effectors of hemoglobin function and bind to the human hemoglobin tetramer with high affinity.
In order to investigate whether the nicotinamide dinucleotides exert their effect on oxygen binding to Hb A by interacting with the /3 chain anion-binding site, we made use of a human hemoglobin variant whose amino acid substitution is in the positively charged cluster of /3 chain residues in the major anion-binding site. This variant, hemoglobin Providence-Asp, has aspartate residues at the 0-82 positions of the two @ chains in the human hemoglobin tetramer instead of lysine residues (Moo-Penn et al., 1976) and has been shown to have a marked decrease in affinity for 2,3-diphosphogtycerate .
Knowledge of the binding affinity of the nicotinamide dinucleotides and their in vivo molar ratios to hemoglobin suggest that a different type of molecular control may occur in connection with these compounds. In contrast to 2,3diphosphoglycerate and similar effectors that act to alter the oxygen transport function, the preferential binding of the reduced forms of the nicotinamide dinucleotides by hemoglobin allows for possible control or modulation of metabolism.

MATERIALS AND METHODS
Hb A, Hb Providence-Asn, and Hb Providence-Asp were prepared and stripped of anionic cofactors as previously described . Samples were stored immersed in liquid nitrogen until use. 8-NADPH, &NADP', 0-NADH, and 8-NAD' were obtained from Sigma. Some comparative studies were carried out using a-NADH and a-NADPH, also purchased from Sigma. Unless otherwise stated, all references to nicotinamide dinucleotide cofactors in this paper refer to the 0 forms. by the spectrophotometric method of Riggs and Wolbach (1956) or Routine oxygen equilibrium measurements were performed either using a modified Hemoscan oxygen dissociation analyzer in a stepwise mode. When used in parallel experiments, good agreement was obtained between the two methods. Extended oxygen binding curves were determined using a thin-layer binding cell (Dolman and Gill, 1978) obtained from the laboratory of Dr. Stanley Gill at the University of Colorado at Boulder. These precisely determined curves were analyzed as described by Johnson (1984) in conjunction with a nonlinear least squares curve fitting program (Johnson et al., 1981). These procedures allow estimation of the parameters associated with the oxygen binding model of Monod et al., (1965). Experiments were conducted in 0.05 M HEPES' buffer with pH adjusted with NaOH unless otherwise specified. By this means, it was possible to evaluate the effector binding over a wide pH range without chloride present.

Nicotinamide Dinucleotides as Heterotropic Effectors of
Hemoglobin Function-A thin-layer cell as described by Dolman and Gill (1978) was used to obtain precise oxygen binding curves for analysis of cofactor binding to hemoglobin (see "Materials and Methods"). Fig. 1 shows representative Hill plots of oxygen binding to stripped Hb A and Hb A in the presence of NADPH and in the presence of 2,3-diphosphoglycerate. It is evident from the figure that NADPH acts in an analogous fashion to 2,3-diphosphoglycerate in altering the oxygen binding properties of hemoglobin. Alterations in the shape of the ligand binding curves are reflected in the slopes of the Hill plots at half-saturation, with n50 values of 2.9, 2.0, and 3.2 for stripped Hb A, Hb A with NADPH, and Hb A with 2,3-diphosphoglycerate, respectively. The depressed value of n50 in the presence of NADPH is due to a more asymmetrical binding curve rather than to decreased cooperativity since Hill coefficients as high as those seen with 2,3-diphosphoglycerate are observed at higher levels of oxygen saturation. Like 2,3-diphosphoglycerate, the nicotinamide dinucleotides exert the major part of their effect on the Tstate of the hemoglobin tetramer while leaving the R-state virtually unaffected. The P50 for the T state, calculated from the linear extension of the binding curves at low saturation where the Hill plots have slopes of unity, is 0.7 in the completely stripped condition and increases to 1.18 and 1.84 in the presence of NADPH and 2,3-diphosphoglycerate, respectively. At high (saturating) concentrations of NADPH, the shift in oxygen affinity at half-saturation is somewhat less than that seen with saturating concentrations of 2,3-diphosphoglycerate (log P m at pH 7.3 of 0.8 and 1.2, respectively). These results were independent of the buffer used in that equivalent data were obtained when Hb A was in 0.05 M Tris, a condition that results in a lowered oxygen affinity due to the C1added as HCl required for pH adjustment. Studies of the a and / 3 forms of the reduced nicotinamide dinucleotide cofactors showed that for both NADPH and NADH there are no significant differences between the effects of the a and / 3 isomers on the P~o values for human hemoglobin.
Throughout the pH range studied, the reduced nicotinamide dinucleotides are more effective at lowering hemoglobin's oxygen affinity than are the oxidized forms of the cofactors. This phenomenon is linked to the greater affinity of the reduced forms for deoxyhemoglobin, as illustrated by the titrations shown in Fig. 2. The titrations were analyzed using an iterative curve-fitting technique in conjunction with the equation of Szabo and Karplus (1976) to estimate the relative affinities of these compounds for Hb A. As described by these authors, the dependence of log Pm on cofactor concentration may be modeled by equations that take into consideration the binding to both oxy and deoxy states, with the former of importance only at relatively high cofactor concentration (above 1 mM in the case of 2,3-diphosphoglycerate binding to Hb A). Table I shows the binding constants calculated by this procedure. Binding constants calculated for 2,3-diphosphoglycerate are presented for comparison, based on data obtained under identical experimental conditions. Three important points are apparent from these estimates.
1) The three nicotinamide dinucleotide cofactors studied are bound preferentially to the deoxy-Hb A tetramer (Table  I). This is analogous to the stabilization of the unliganded Tstate as seen with 2,3-diphosphoglycerate.
2) The reduced forms, NADPH and NADH, bind more tightly than do their respective oxidized forms. There is little or no effect of oxidized NAD on oxygen binding by Hb A.
3) For the conditions of this study, where the use of HEPES buffer avoids the necessity of having chloride present as a competitive anion, it is possible to directly compare the calculated binding affinities of 2,3-diphosphoglycerate and the nicotinamide dinucleotide cofactors. As shown in Table I, deoxy-Hb A binds these forms (except NAD+) with high affinity. A striking difference between the effectors is the stronger binding of 2,3-diphosphoglycerate to oxy-Hb A.
The upper trace shown in Fig. 2 is a NADPH titration carried out with Hb A in the presence of 1.5 mM inositol hexaphosphate. Under these conditions, the oxygen affinity is initially low, due to the strong allosteric influence of inositol hexaphosphate. It remains so until the NADPH concentration is 5 mM or higher where the cofactor binding is significantly competitive with that of inositol hexaphosphate. The titration suggests a common binding site for the two classes of effectors and supports the fact that NADPH binds to the hemoglobin tetramer with high affinity. Fig. 3 shows the pH dependence of Hb A in the presence of the four nicotinamide dinucleotides studied. It is notable that not only is the affinity for oxygen lowered by NADPH, NADP+, and NADH, but also that the magnitude of the Bohr effect is increased from 0.25 in stripped Hb A to approximately 0.65 in the presence of 0.5 mM NADPH, NADP+, or NADH. This indicates a linkage between proton binding and cofactor binding that is reminiscent of that seen with the binding of other organic phosphates to Hb A. Table I1 compares the effects of the nicotinamide dinucleotide cofactors on oxygen binding by Hb A and Hb Providence-Asp at pH 6.5 and 7.3. It is relevant that Hb Providence-Asp at pH 6.5 shows little response to 2,3-diphosphoglycerate and is essentially unaffected by 2,3-diphosphoglycerate at pH 7.3 . The same pattern is seen for the nicotinamide dinucleotide effectors, as is evident from the data of Table 11. The amino acid substitution of Hb Provi-

Effect of nicotinamide dinucleotide cofactors on oxygen binding
Oxygen binding measurements were made in 50 mM HEPES buffer at 20 "C at a hemoglobin concentration of 60 pM. Alog P , refers to the difference in log PW values in the presence and absence of the specified cofactor.

Cofactor-induced Asymmetry in Equilibrium Oxygen Bind-
ing Curves-Evidence of cofactor-induced asymmetry is sometimes evident in equilibrium oxygen binding studies. Fig.   4 is a plot of log P , uersus log Pso for Hb A and Hb Providence-Asp with varying cofactor concentrations. In a symmetrical oxygen binding curve with no differential association of cofactor between the different chains, one would expect to find P,,, = Pm and a plot of log P, versus log Ps0 to have a slope of approximately 1. It is apparent from Fig. 4 that for Hb A in the presence of low levels of 2,3-diphosphoglycerate (left half of curves), a large amount of asymmetry is introduced by the cofactor. This is not the case as the ratio of 2,3-diphosphoglycerate to hemoglobin rises or in the presence of NADPH, NADP+, or NADH. In these latter cases, the points all fall on a line with slope approximately 1. This same line contains all points for the two Hb Providence variants (in the presence of 2,3-diphosphoglycerate) which (as discussed earlier) have a greatly reduced association with 2,3-diphosphoglycerate due to the amino acid substitution in the anion-binding site. Heme concentration was 60 p~ for each experiment. Each point represents a different cofactor concentration (increasing from lower left to upper right. 0, Hb A + 2,3-diphosphoglycerate (DPG), pH 6.2; A, Hb Providence-Asn + 2,3-diphosphoglycerate, pH 6.2; 0, Hb Providence-Asp + 2,3-diphosphoglycerate, pH 6.2; *, H b A + NADPH, pH 7.3; X Hb A + NADP+, pH 7.3.

DISCUSSION
Stereochemistry of Hb A Nicotinamide Dinucleotide Interactions-It is clear from hemoglobin oxygen binding studies done in the presence of the nicotinamide dinucleotides NADPH, NADP+, and NADH that these cofactors act as heterotropic allosteric effectors of hemoglobin function. These effects appear to be exerted by interaction of the cofactors with the /3 chain anion-binding site, a situation analogous to that seen with 2,3-diphosphoglycerate, inositol hexaphosphate, and ATP. In particular, we have shown that both LY and /3 forms of these cofactors bind with high affinity to Hb A, with binding constants like those of the abovementioned organic phosphates. Greatly decreased effects on the oxygen affinity of Hb Providence-Asp suggest that /3-82 lysine is at least a contributor to the binding site of the nicotinamide dinucleotides. The competition between NADPH and inositol hexaphosphate effects (Fig. 2) also supports the hypothesis that the nicotinamide dinucleotides alter oxygen affinity by interactions with residues that contribute to the / 3 chain anion-binding site. Fig. 5 presents a schematic representation of the open form of NAD+ (a conformation it assumes upon binding to lactate dehydrogenase) and indicates the regions of the molecule that appear to be of importance in its interactions with Hb A. The two central phosphates are reasonable candidates in light of their charge complementarity to the positively charged residues of the , f 3 chain anion-binding site. The enhanced Bohr effects seen in the presence of these effectors also closely resembles that seen with 2,3-diphosphoglycerate, where Bohr protons are clearly derived from phosphate groups (Benesch and Benesch, 1969;de Bruin et al., 1974). The cofactors are markedly larger than 2,3-diphosphoglycerate, however, and the studies of Arnone (1972) indicated that the binding site is rather "tailor-made" for an anion the size of 2,3-diphosphoglycerate. Although we cannot predict what conformation the nicotinamide dinucleotide cofactors assume upon binding to Hb A, the available data dictate considerable specificity in their binding interactions. Both the 2"phosphate of NADPH and NADP+ and the charge state of the nicotinamide ring are

NAD+.
The numbered sites represent features which data indicate are important in binding of the nicotinamide dinucleotide cofactors to hemoglobin. Site 1, presence (NADP) or absence (NAD) of the 2'phosphate on the ribose of the AMP moiety; site 2, redox state of the nicotinamide ring; site 3, two phosphate groups of the pyrophosphate linkage.
important in the interaction with the binding site. A comparison of the size of 2,3-diphosphoglycerate with those of the extended forms of NADPH, NADP+, and NADH reveals a significant difference in both size and the distribution of charged groups which are thought to take part in binding. Although for many NADP+-and NAD+-dependent enzymes the cofactor binds in its extended conformation (for discussion, see Fita and Rossman, 1985), in the case of catalase the bound NADP+ has been shown to be folded into a righthanded helix (Fita and Rossman, 1985). It is possible that a similar folding process could bring the charged groups of the nicotinamide dinucleotides into a conformation which facilitates its binding to hemoglobin. It is likewise possible that at least some of the charged groups on the nicotinamide dinucleotide cofactors interact with groups adjacent to, but not actually a part of, the /3 chain anion-binding site as defined by the studies of Arnone (1972). The calculated binding constants suggest that the conformational differences between deoxy and oxy forms of Hb A more strictly exclude the nicotinamide dinucleotide interactions with the oxy form than occurs for either 2,3-diphosphoglycerate or inositol hexaphosphate. Further clarification of the stereospecificity of these interactions will require x-ray structural analysis.
Effect of Nicotinamide Dinucleotides on I n Vivo Oxygen Transport-The discovery that the reduced and oxidized nicotinamide dinucleotides NADPH, NADP', and NADH can act as heterotropic allosteric effectors of hemoglobin function suggests the possibility that these interactions may have physiological importance. If such a physiological role does exist, it may be related more to the metabolism of the red blood cell than to hemoglobin's ability to bind oxygen at the air-lung interface and deliver it to respiring tissues. Although the affinity of NADPH binding to hemoglobin (micromolar range) is "high," the concentration of the nicotinamide dinucleotides in erythrocytes, compared to the hemoglobin concentration, dictates that the quantitative role of these molecules in lowering the oxygen affinity of hemoglobin can only be very small. Hemoglobin tetramer is present at a concentration of approximately 5 mM in the erythrocyte (Wintrobe, 19671, and 2,3-diphosphoglycerate is present in approximately equimolar amounts (Guest, 1942). The nicotinamide dinucleotides, however, are present at molar ratios of less than 1:50 compared to hemoglobin tetramer (Gross et aL, 1966). This is clearly not consistent with an intracellular role for the nicotinamide dinucleotide cofactors as major modulators of hemoglobin oxygen affinity.
Possible Metabolic Effect of Hemoglobin-Nicotinamide Dinucleotide Interactions-Another possibility to be considered is that physiologically important interactions would occur if binding of the nicotinamide dinucleotide cofactors to hemoglobin reduces their effective concentrations within erythrocytes. These cofactors are important in control of many metabolic pathways and provide the reducing potential for a number of enzymatic reactions. Among the latter are enzymatic reactions responsible for in uiuo reduction of methemoglobin. Moreover, our binding data (Table I) suggest that the modulation of intraerythrocytic levels of nicotinamide dinucleotide cofactors could potentially operate on two levels. First, we find a stronger interaction between hemoglobin and the reduced forms of the cofactors, suggesting that intracellular binding could affect the redox potential within the cell by lowering the effective concentrations of NADH and NADPH compared to the oxidized forms. Second, our data show that the nicotinamide dinucleotide cofactors bind almost exclusively to the deoxy form of the Hb A tetramer. This suggests that the binding of the cofactors to hemoglobin in uiuo would be greatest in the relatively deoxygenated blood leaving the capillary beds in the tissues and much less in the oxygenated blood leaving the lungs. Thus, one might predict that in oxygenated blood, the absolute levels of free nicotinamide dinucleotide cofactors would be high with respect to deoxygenated blood. One would also predict relatively low values for the ratio of oxidized to reduced cofactor forms free in oxygenated blood (i.e. oxygenated blood would be reduced with respect to deoxygenated blood to the extent that the redox potential is affected by shifts in the nicotinamide cofactor ratios). These two sets of intracellular conditions might then be expected to favor different subsets of enzymatic activity as the red cells course through the bloodstream. Inherent in this possible scenario of events is the realization that physiological differences may be attributable to mutations in the human hemoglobin tetramer or evolutionary differences between species that affect the p chain anionbinding site and its ability to bind 2,3-diphosphoglycerate and the nicotinamide dinucleotide cofactors. Significant metabolic consequences with regard to NAD-and NADP-linked enzymatic processes may result that are not classically considered to be affected by functional changes in the hemoglobin tetramer.
Evidence exists that a significant portion of nicotinamide dinucleotide cofactors within human red blood cells are bound within the protein fraction. Studies by Jacobasch et al. (1974) based on measured total NAD levels in the cell and the NAD+/ NADH ratio calculated on the basis of the measuredpyruvatel lactate ratio suggest that a large proportion of intracellular NADH is bound and is not part of the free pool. Recent studies (Kirkman, et al., 1986) strongly suggest that the availability of intracellular NADP+ to glucose-6-phosphate dehydrogenase is reduced by the binding of the cofactor to other proteins within the cell. Although the data cited did not implicate hemoglobin as a major effector of this intracellular NADP+ binding, it should be noted that the experiments reported were carried out under oxy conditions where our data imply that the binding interactions would be minimal or nonexistent.
The presence within the erythrocytes of a large number of other anions that can bind to hemoglobin could reduce the influence of these interactions. Since we suggest that the nicotinamide dinucleotides interact with the /3 chain anionbinding site, the dominant competing anion is 2,3-diphosphoglycerate. Red cells contain a relatively large amount of this effector, and it binds to hemoglobin with high affinity. Calculations* based on the relevant affinity constants and the reported concentrations of hemoglobin and its effectors within erythrocytes suggest that alterations of red cell metabolism may indeed result from the differential binding of the reduced forms of the nicotinamide dinucleotide cofactors. Specifically, at the estimated value of KN for NADPH and at a 1:1 ratio of Hb A to 2,3-diphosphoglycerate, 93% of the reduced dinucleotide could exist in the form of a hemoglobin-dinucleotide complex in deoxygenated solutions. The possibility exists, therefore, that hemoglobin may serve as a "sink" for a significant quantity of the intraerythrocytic nicotinamide dinucleotides.
The results presented here demonstrate in vitro interactions between hemoglobin and the nicotinamide nucleotides. The in vivo condition is undoubtedly complex, but these interactions are of potential significance in any of the redox interactions carried out by the circulating erythrocytes. The in vitro results reported here serve to clarify yet another class of molecules that can act as modulators of oxygen binding by hemoglobin and will hopefully also stimulate studies of the physiological consequences of these interactions.