Characterization of the ionizable groups interacting with anionic allosteric effectors of human hemoglobin.

Benzenehexacarboxylate binds preferentially to the deoxy form of human hemoglobin and lowers its oxygen affinity. Saturating amounts of this anionic cofactor produce changes in the oxygen affinity of hemoglobin A which are intermediate between those produced by 2,3-diphosphoglycerate and inositol hexaphosphate. Between pH 6 and 9 the interaction of deoxyhemoglobin A with benzenehexacarboxylate results in the absorption of protons. The proton absorption may be ascribed to three pairs of groups having pK values near 5.59, 6.78, and 7.66 where the pK values are shifted by 1.26 pH units upon interaction with benzenehexacarboxylate. In deoxyhemoglobin Deer Lodge (p2His + Arg) only two pairs of groups appear to be responsible for the proton absorption. Their apparent pK values are 5.55 and 7.18, respectively, with a ApK = 1.49 upon the interaction with benzenehexacarboxylate. These results are compared with the crystallographic models of Arnone (Arnone, A. (1972) Nature 237, 146-149) and Arnone and Perutz (Arnone, A., and Perutz, M. F. (1972) Nature 249,34-36). It is proposed that in hemoglobin A the groups with pK = 5.59 are the p143 histidines: those with pK = 6.78 are the 81 valines and those with pK = 7.66 are the j32 histidines. In hemoglobin Deer Lodge, the groups with pK = 7.18 are inferred to be the fil valines. Substitution of an aspartyl residue for a lysine in deoxyhemoglobin ProvidenceAsp@SZLys -+ Asp) greatly reduces the interaction with the effector above pH 7.5 as anticipated by the mentioned models of Arnone and Arnone and Perutz. The liganded form of hemoglobin A differs from its deoxy derivative in the interaction with benzenehexacarboxylate. In the liganded derivative the NH,, terminus of the p chain and the p2 histidine do not appear to experience pK changes upon addition of benzenehexacarboxylate. Thus the liganded derivatives of hemoglobins A and Deer Lodge (PtHis

+ Arg) show very similar proton absorption. These findings suggest that the site at which anionic effecters are bound differs in liganded and unliganded hemoglobin.
The difference could arise from conformational changes which make different groups available for interaction with the effector. The CD spectrum in the Soret region of the carbon monoxide derivative of hemoglobin A is sensitive to benzenehexacarboxylate and inositol hexaphosphate while the same derivatives of hemoglobins Deer Lodge and Providence-Asp are not. In contrast, comparable CD spectra of the deoxy derivatives of the hemoglobins are all similar and do not show sensitivity to either benzenehexacarboxylate or inositol hexaphosphate.
Crystallographic models of Arnone (1) and Arnone and Perutz (2) indicate that the interaction of human deoxyhemoglobin with polyanion effecters is regulated by the electrostatic interaction of the negative charges of the effector with a cluster of 8 positive groups in the (u~*&~ tetrameric molecule. Some of these groups lose their positive charge when the pH is increased from pH 7 to 9. Studies on the interaction of human hemoglobin with BPC' (3) showed that the carboxyl groups of the effector do not undergo changes in ionization state in this pH range. Consequently, measurement of the protons absorbed upon the interaction of deoxyhemoglobin and polycarboxylates was seen as a possible tool for studying the characteristics of the ionizable groups of hemoglobin which interact with anionic effecters. It has previously been shown that polycarboxylate effecters modulate hemoglobin oxygen affinity in a manner analogous to 2,3-diphosphoglycerate and other polyphosphate effecters (4). In this study we used BHC rather than BPC because it has a higher affinity for hemoglobin, making it possible for liganded hemoglobin to be approached (3,4). This paper reports a study of the ionization characteristics of side chains which in liganded and unliganded hemoglobin interact with polyanions. We used BHC as the effector, and Hb A plus two mutant hemoglobins as model proteins. The '  Both these procedures require the input of initial estimates for the unknown parameters. These are then refined by the algorithms so as to minimize the sum of square residuals.
In all cases the results obtained were independent from the initial estimates.
Fittings between simulated curves and experimental determinations were subjected to the following considerations.
The number of groups involved in the interaction cannot be less than the maximum value of H,,, experimentally obtained.
In addition, the models of Arnone (1) and Arnone and Perutz (2)  The hemoglobin species here considered are those protonated in i = 0, 1, 2, ... n of the n groups that form salt bridges with the effector, and ionize in the pH range investigated.
Also as previously shown (3) s q P,Ri where P, is the affinity constant of hemoglobin for the effector when all of the positive groups considered are unprotonated and R is the ratio of the ionization constant of each group in the presence of the effector to that in the absence of the effector. Taking oi and p, = 1 a, as the fraction of unprotonated and protonated form of each ith group, respectively, the distribution of the various molecular species considered is given by the terms of the product When each term of the expansion is multiplied by the proper P, their sum gives P as in the expression: This is the equation used for simulating the pH dependence of the affinity constant of hemoglobin for the effector.

Interaction of Deoxyhemoglobin
A with BHC-The pH dependence of oxygen affinity of stripped hemoglobin A in the presence and absence of DPG, BHC, or IHP is shown in Fig.  1. The effect of BHC is intermediate to that of DPG and IHP. The value of n1,2, in Hill plots of the oxygen binding, remained between 2.5 and 2.8 in all cases. Representative Hill plots at neutral pH are shown in Fig. 2.
In Fig. 3 the titration of deoxyhemoglobin A with BHC was followed potentiometrically, monitoring the amount of HCl necessary to restore the original pH of the solution after each addition of BHC. The titration was performed near pH 7.0 where the reaction was essentially stoichiometric.
The sharp break of the titration curve indicates a stoichiometric ratio of 1 mol of effector/mol of tetrameric hemoglobin.
The interaction of deoxyhemoglobin with BHC above pH 7.4 is nonstoichiometric.
This makes it possible to produce a plot of the log Y/(1 -Y) against the logarithm of the free 15. with the effector with a common pK shift. The pK values assigned to the groups before the interaction with BHC were 5.59, 6.78, and 7.66, with a ApK of 1.26. Table  I lists these parameters and the 95% confidence limits obtained for them.
We will call the groups with pK = 5.59 the acid groups, and those with pK = 6.78 and pK = 7.66 the neutral groups.
In Fig. 6    , with the parameters listed in Table I, and a value of PO = 3000 M-I, AF" = -4645 cal. The upper line represents a simulation of pH dependence of the affinity of deoxyhemoglobin Deer Lodge for BHC as obtained from the parameters listed in Table I and a value of P, corresponding to a AF" = -9300 cal.

Ionizable
Groups and Hemoglobin's Allosteric Effecters Interaction of Oxyhemoglobin A with BHC-In Fig. 3 a stoichiometric titration of oxyhemoglobin with BHC is shown. Also in this case, as in deoxyhemoglobin A, the stoichiometry of the reaction proved to be 1:l. The pH dependence of H,,, for the interaction of oxyhemoglobin with BHC is shown by the open circles in Fig. 7. It is substantially different from that of deoxyhemoglobin A. Above pH 7, the low affinity of liganded hemoglobin for BHC made it impossible to objectively determine the end point of the titration. Above pH 7.5 practically no pH change was observed when BHC was added to the protein solutions. The data presented in Fig. 6 are raw values, therefore not corrected for the number of protons liberated by BHC. Since the data obtained covered only a limited pH range, no numerical simulation was conducted.
Interaction of BHC with Deoxyhemoglobin Deer Lodge-The reaction of deoxyhemoglobin Deer Lodge with BHC was studied in the same way as for hemoglobin A. In this case the stoichiometry was also 1:l. The pH dependence of H,,, is represented in Fig. 5. It appears that in the entire pH range investigated the absorption of protons was less than in the case of deoxyhemoglobin A. The interpolating line is the best fit obtained with Equation 1, assuming that two pairs of identical groups interact with the effector with a common pK shift. The assigned pK values were 5.55 and 7.18, with a ApK = 1.49. Table I lists these parameters and the 95% confidence limits obtained for them.
The affinity of the effector for the protein appeared to be higher than in the case of deoxyhemoglobin A. The reaction was essentially stoichiometric at all pH values investigated.

Znteraction of Oxyhemoglobin
Deer Lodge with BHC -The stoichiometry of the reaction was measured at pH 6.5 as previously described. Also in this case the titration showed a stoichiometry of 1:l. The pH dependence of H,,, for the reaction is presented in Fig. 7 together with the data obtained for normal hemoglobin. This parameter appeared to be very similar to that obtained for oxyhemoglobin A. Also in this case the interaction with BHC at pH near and above 7.5 produced a negligible or totally absent absorption of protons.
Interaction of Deoxyhemoglobin Providence-Asp with BHC -The number of determinations performed with this hemoglobin were very limited due to the small amount of protein available. Only three determinations were possible (pH 7.0, 7.5, and 8.0). The interaction at pH 7.0 gave an absorption of protons comparable to that of normal hemoglobin (1.96 eq/tetramer), and the reaction was essentially stoichiometric. At pH 7.5 the absorption of protons was less than in normal hemoglobin (1.50 eq/tetramer). At pH 8.0 the titration with BHC failed to give a distinct plateau at the end of the titration and H,,, could not be measured. This was due to the low affinity of the mutant deoxyhemoglobin for BHC (probably below 1O-3 M-I) at this pH.
Measurements of Circular Dichroism -The circular dichroism of carbonmonoxyhemoglobin A in the Soret region is sensitive to the presence of 10m3 M BHC and IHP as shown in Fig. 8. The same is not true for the CD spectra of the carbonmonoxy derivatives of hemoglobins Deer Lodge and Providence-Asp, which otherwise were not distinguishable from that of normal hemoglobin. Addition of sodium dithionite (0.1 mg/ml) to the various solutions did not modify these results, nor did the method of hemoglobin preparation. This experiment was repeated several times to ensure that the effect observed with Hb A was not related to the presence of methemoglobin or to a particular preparation of the protein. deoxy derivatives of the three hemoglobins investigated were very similar to each other and were not sensitive to the addition of effecters.

DISCUSSION
Groups That Znteract with BHC in Deoxyhemoglobin-In the attempt to determine the ionization characteristics of the groups in deoxyhemoglobin that interact with BHC, rather than searching for a unique set of parameters in the various simulations, we searched for the best numerical solution possible within the framework of existing models for the binding of effecters to deoxyhemoglobin.
The assumptions derived from the models of Arnone (1) and Arnone and Perutz (2) not only produced a very good correspondence between experimental and simulated data, but the fitting parameters are very close to data obtained from entirely different approaches. According to the crystallographic models there are Ionizable Groups and Hemoglobin's Allosteric Effecters three pairs of groups in deoxy hemoglobin A that produce proton absorption in our experiments. We have assigned pK values near 5.59, 6.78 and 7.66 to these groups. Results obtained by Garner et al. (19) would assign a pK of 6.85 to the /31 valines. Results of Fung et al. (20) assign a pK of about 7.5 to the 02 histidines.
We regard these values as virtually identical to those assigned to the neutral groups in our simulations.
This suggests that the acid groups with pK = 5.59 are the p143 histidines. This value might appear somewhat low for a histidyl residue, but it is within the reported variation range of the pK values of these side chains in globular proteins (21,22). The proximity of lysyl residues at p&32 and at p144 to the p143 histidines might explain the relatively low apparent pK of these histidyl residues.
The apparent pK values so obtained for the ionization of the acid and neutral groups include electrostatic factors which we have not been able to evaluate quantitatively and the averaging factor introduced by assuming the same pK shift for all of the interacting residues. Indeed the pK shift is related to the distance between the opposite charges, and probably it is not identical in all of the salt bridges which are formed.
It must be stressed that these apparent values are regulated by the protein conformation. Thus variations in these values may reflect structural modifications at least in the area where the binding of effecters occurs.
In hemoglobin Deer Lodge the apparent pK of the @l valines increases from 6.78 to 7.18 and the ApK also increases from 1.25 to 1.49. This indicates that the average length of the salt bridges is shorter in this hemoglobin than in hemoglobin A, suggesting at least a slight rearrangement of the groups in the binding site. This rearrangement, in turn, might increase the distance between the positive charges of residues pl and /32, thereby allowing a higher apparent pK for the pl valines.
Affinity of BHC for Deoxyhemoglobin A and Deer Lodge -In Fig. 6 the pH dependence of the overall affinity of deoxyhemoglobin A for BHC was simulated using a value of P,, = 3000 Mm'. This value of P,, corresponds to a @ of -4645 cal which is consistent with two salt bridges formed between BHC and the 682 lysines at high pH range. As can be seen the simulation perfectly interpolates the experimental points. This confirms the assumption implied in Equation 7 that the binding of BHC is primarily regulated by the salt bridges formed between hemoglobin and the effector.
The additional simulation presented in Fig. 6 was obtained for hemoglobin Deer Lodge using a value of P,, which gives a aF" of -9300 cal. This would correspond to the formation of four salt bridges at high pH. The additional salt bridges at high pH in hemoglobin Deer Lodge are those produced by the arginines in position /32. It can be readily seen that the resulting affinity constant for the interaction of deoxyhemoglobin Deer Lodge and BHC is distinctly higher than that of deoxyhemoglobin A at all pH values. This is consistent both with the stoichiometric behavior shown in the binding of BHC by hemoglobin Deer Lodge in all of the experiments performed, and with the participation of the p2 arginines in the binding.
Interaction of BHC with Hemoglobin Providence-Asp -The behavior of the interaction of hemoglobin Providence-Asp with BHC is drastically different from that of normal deoxyhemoglobin. It is improbable that this difference is produced only by the treatment with sodium dithionite used during the preparation of the mutant hemoglobin. Control experiments indicate that such treatment does not significantly alter the functional behavior of normal hemoglobin.
The crystallographic model of Arnone (1) and Arnone and Perutz (2) suggests that the presence of the aspartate in position p82 decreases the affinity of hemoglobin Providence-Asp for BHC not only because one salt bridge becomes impossible, but also because of active repulsion between the negative charge of the aspartyl carboxyl group and the negative charges of the effector. It appears that the interaction of deoxyhemoglobin Providence Asp with BHC is accompanied by absorption of protons from acid groups, while the absorption of protons from neutral groups could not be measured because of the low affinity of this hemoglobin for BHC above pH 7.5. This suggests that the interaction of BHC with the /31 and /32 residues is greatly reduced by the presence of aspartate at the /382 position.  (23) and Brygier et al. (24) in the sense that above pH 7.5 proton absorption by the liganded form becomes immeasurably small. The absorption of protons decreases rapidly with increasing pH, changing from about 2.3 protons/mol at pH 7.0 to nearly none with an increase of less than 1 pH unit. This suggests that there may be a cooperative protonation of the side chains which participate in the interaction. Similar cooperativity has been reported for the interaction of BHC with liganded p chains (10). Additionally, it is clear that the absorption of protons by the neutral groups in deoxyhemoglobin is absent in liganded hemoglobin. De Bruin et al. (23) and Brygier et al. (24) suggested that these groups have lower pK values in liganded hemoglobin than in deoxyhemoglobin.
Our data on hemoglobin Deer Lodge are relevant to this hypothesis. Comparison of the interaction of the liganded derivatives of hemoglobin A and Deer Lodge with BHC reveals no differences. The substitution of arginine for histidine at the p2 position might be expected to produce a lower absorption of protons and a higher affinity for the effector in liganded hemoglobin Deer Lodge (as for its unliganded derivative) if the p2 residues were involved in the interaction.
It seems more probable that the binding sites are different in liganded and unliganded hemoglobin, as proposed by Kilmartin (251, or that the binding sites are topologically similar (between the adjacent p subunits), but the different conformations of the protein make different groups available to the interaction. The latter has been proposed by Salahuddin and Bucci in connection with experiments on isolated /3 chains (10).
Optical Activity Measurements -It is relevant to stress that in the Soret region of the spectrum the optical activity of the unliganded derivatives of the three hemoglobins investigated were indistinguishable.
This supports the idea that the potentiometric interaction of deoxyhemoglobins Deer Lodge and Providence-Asp with BHC is different from that of normal hemoglobin because of changes in the electrostatic environments of the mutant hemoglobins, not because of special protein conformations produced by the mutations.
It is important to note that the CD spectra of the unliganded derivatives are not sensitive to either BHC or IHP. This suggests that the small conformational change detected by Arnone (1) and Arnone and Perutz (21, in the deoxyhemoglobin A crystals, in the presence of polyphosphate, did not affect the conformation of the heme pocket. Additionally, this supports the hypothesis that the modification of the oxygen affinity produced by anionic effecters is due primarily to their preferential binding to deoxyhemoglobin (26). The CD spectra of the carbonmonoxy derivative of hemoglobin A is sensitive to both BHC and IHP. This indicates modifications in subunits tertiary structure, and is consistent with observations of Lindstrom and Ho (271,Adams and Schuster (28), and Perutz et al. (29). We do not ascribe these results to a major shift in the equilibrium between R and T conformations since the proton absorption of liganded Hb A in the presence of BHC is distinctly different from that of the unliganded form. The conformational change produced in liganded hemoglobin by the addition of effecters is consistent with the possible cooperative protonation of the protein groups involved in the interaction.
This was discussed in regard to the interaction of BHC with the isolated p subunit of hemoglobin A (10).
It is surprising that the CD spectra of the carbonmonoxy derivatives of hemoglobins Deer Lodge and Providence Asp are nearly insensitive to both BHC and IHP. The interaction was not tested potentiometrically for hemoglobin Providence Asp, while in the case of hemoglobin Deer Lodge the interaction produced an absorption of protons similar to that obtained with normal liganded hemoglobin.
We do not have a simple explanation for this phenomenon.
Functional fmplicatians -In the alkaline range, near pH 8, the affinity of hemoglobin Deer Lodge for BHC was higher than that of normal hemoglobin. In the same pH range, the affinity of hemoglobin Providence-Asp for BHC was much lower than that of normal hemoglobin.
From these observations we would predict that, in this pH range, the effects of polyanions would be enhanced in hemoglobin Deer Lodge, and greatly reduced in hemoglobin Providence-Asp. These predictions, based on the potentiometric study of the interaction of BHC with these deoxyhemoglobins, are consistent with functional studies conducted with IHP on hemoglobins Deer Lodge and Providence-Asp (6, 7).
These studies show the utility of abnormal human hemoglobins as probes of hemoglobin's anion binding site. From the results obtained, it appears that the model proposed by Arnone (1) and Arnone and Perutz (2) for the interaction of deoxyhemoglobin with polyphosphates is also valid for the interaction of deoxyhemoglobin with BHC.