Stability of the heme-globin linkage in alphabeta dimers and isolated chains of human hemoglobin. A study of the heme transfer reaction from the immobilized proteins to albumin.

The stability of the heme-globin linkage in alphabeta dimers and in the isolated chains of human hemoglobin has been probed by studying the transfer of heme from the proteins immobilized onto CNBr-activated Sepharose 4B to human albumin. The kinetic and equilibrium features of the reaction have been measured spectrophotometrically given the stability of the heme donors and the ease with which heme donor and acceptor can be separated. Isolated alpha and beta chains transfer heme to albumin at similar rates (1 6 x 10(-2) s-1 at pH 9.0 and 20 degrees C) in the ferrous CO-bound and in the ferric state. In alpha beta dimers the heme-globin linkage is strengthened considerably, albeit to a different extent in the ferrous CO-bound and ferric met-aquo derivatives. Only in the latter heme is lost at a measurable rate, 0.065 +/- 0.011 x 10(-2) s-1 for alpha heme and 2.8 +/- 0.6 x 10(-2) s-1 for beta heme at pH 9.0 and 20 degrees C, which is very close to the rate measured with soluble met-aquo-hemoglobin at micromolar concentrations. These results indicate that in human hemoglobin the heme-globin linkage in the alpha chains is stabilized by interactions between unlike chains at the alpha1 beta1 interface, whereas heme binding to the beta chains is stabilized by interactions at the alpha1beta2 interface. These long range factors have to be taken into account in addition to the local factors at the heme pocket when evaluating the effect of point mutation and chemical modification.

The polypeptide chain-heme equilibrium is an important element in the description of all hemoproteins. However, a direct measure of this parameter is difficult due to the high stability of the holoprotein under physiological conditions (1). The heme-apoprotein linkage, therefore, is usually assessed by studying the heme transfer reaction from a donor to a heme acceptor protein. When the heme acceptor has a very high affinity for heme, like apomyoglobin, the kinetics of the heme transfer process is followed. This is governed by the rate of heme release from the donor protein as the rate of heme association is very high (2,3). When the heme donor and acceptor proteins have similar affinities for heme, the heme is partitioned between the two proteins and hence both kinetic and equilibrium features of the heme transfer reaction can be exploited (4,5).
In the case of human hemoglobin (HbA) there are relatively few quantitative estimates of the stability of the heme-globin linkage despite its relevance in the study of mutant and chemically modified hemoglobins (2)(3)(4)(5)(6). The interaction between heme and globin is known to be affected by a number of parameters. Of major importance are the redox state of the protein and pH (2,4,5). Thus, heme is released only from oxidized, met-aquo-hemoglobin, but not from the reduced protein and more so upon departure from neutral pH values. In met-aquohemoglobin, the rate of heme release from ␤ chains is considerably faster than from ␣ chains. Furthermore, the kinetics of heme dissociation from ␣␤ dimers is much greater than from ␣ 2 ␤ 2 tetramers (5,6). In the isolated ␣ and ␤ chains to our knowledge the heme-globin affinity has never been measured, presumably due to the very marked tendency of the respective globins to precipitate. As a matter of fact in all studies of heme dissociation from hemoglobin or from myoglobin the formation of free, precipitable globin upon heme depletion of the ferric protein is a common problem (3,5,6) that has been alleviated in part by adding sucrose to the incubation medium (2).
We have addressed the problem of globin precipitation in a different manner and propose the use of ␣␤ dimers immobilized on Sepharose 4B (7-10) as heme donor. This material offers a number of advantages: it is stable after heme depletion, it does not undergo changes in state of association, it can be lyophilized, and, most importantly, it can be separated easily from the acceptor protein. Therefore formation of the heme-acceptor complex can be followed spectrophotometrically in a facile way under any experimental condition. As heme acceptor we have used human albumin which is endowed with two high affinity heme binding sites (11) and whose affinity for heme does not change in the pH range 5-10 (12). Transfer of heme from hemoglobin to albumin to form methemalbumin is known to occur in the blood of patients with plasma hemoglobinemia and "in vitro" when albumin is mixed with ferric hemoglobin (4,5). The immobilized ␣␤ dimers behave like the soluble protein in that they do not transfer heme to albumin when in the oxy or CO form, but only when in the oxidized state, an indication that the immobilization process does not produce significant alterations in the heme environment. The direct comparison of the rate of heme loss from soluble dimers has confirmed this contention.
The behavior of isolated ␣ and ␤ chains both in solution and after immobilization on Sepharose 4B has also been studied. Quite unexpectedly, soluble ␣ and ␤ chains transfer heme readily to albumin even when oxygenated or in the CO-bound form, a finding which points to a major change in the heme pocket induced by the assembly of unlike chains into a heterodimer.

MATERIALS AND METHODS
Human serum albumin, crystallized and lyophilized, was a commercial product of Sigma and was used without further purification. Albumin concentration was determined spectrophotometrically at 280 nm, using the molar absorbance E 280 nm 1 cm ϭ 9.8 ϫ 10 Ϫ5 . Human hemoglobin was prepared according to Berger et al. (13) and was immobilized covalently in the oxygenated state on CNBr-activated Sepharose 4B (Pharmacia Biotech Inc., Uppsala, Sweden). The coupling reaction was carried out in 0.1 M sodium bicarbonate, pH 8.3, in the presence of ethanolamine in 10:1 molar excess with respect to heme. The suspension was stirred at room temperature for 1 h and washed thereafter on a filter funnel as described in Ref. 9. This procedure leads to immobilization of hemoglobin as ␣␤ dimers which are coupled to Sepharose via either chain and maintain the capacity to interact in a specific and reversible manner with soluble dimers (7)(8)(9)(10). The concentration of immobilized oxyhemoglobin was typically around 8 mg/ml of packed resin. It was determined on every preparation using a 1-mm light path cell and a Cary 219 spectrophotometer; the effect of turbidity was minimized by the use of protein-free gel in the reference cell (9). Immobilized oxy-␣␤ dimers (Fig. 1A) were oxidized by addition of two to three equivalents of potassium ferricyanide at neutral pH; excess ferricyanide and the ferrocyanide produced during the oxidation reaction were removed by washing extensively the gel on a filter funnel. The immobilized met-␣␤ dimers were then lyophilized in the presence of 10% sucrose. The optical absorption spectrum of the immobilized met-␣␤ dimers equilibrated with buffers of pH 6.5-9.0 ( Fig. 1, B-D Isolated ␣ and ␤ chains were prepared according to Geraci et al. (14). Heme transfer from the soluble chains to albumin was followed in polyacrylamide gel electrophoresis experiments carried out according to Davis (15). The chains were mixed with albumin, and the mixture was incubated at 4 or 20°C and then subjected to electrophoresis. The gels were stained for heme with benzidine and for protein with Coomassie Blue. The immobilization reaction was carried out on the oxygenated derivative in 0.1 M phosphate buffer at pH 7.4 and yielded a concentration of immobilized chains between 1 and 5 mg/ml of packed resin. Oxidation of the immobilized chains was performed as described for hemoglobin.
Heme Transfer Experiments with Immobilized ␣␤ Dimers-A known amount of lyophilized, immobilized met-␣␤ hemoglobin dimers (ranging from 25 to 400 mg) was mixed in a test tube with 2.5 ml of buffer containing albumin at a known concentration and the test tube was placed in a water bath thermostatted at 20°C. At established times, a filter sampler (Porex Medical, Fairburn, GA) was gently pressed into the test tube to separate the solid phase from the supernatant, whose absorbance was measured between 450 and 650 nm on a Cary 219 spectrophotometer. Formation of methemalbumin can thus be followed without interference from immobilized hemoglobin. Methemalbumin concentration was determined using a millimolar specific absorbance (heme basis) of 9.5 at 530 nm; this value was obtained with the pyridine hemochromogen method (1). The buffers used in the heme transfer experiments were: 0.1 M Tris-HCl plus 0.1 M NaCl at pH 9.0 or 7.5; 0.05 M bis-Tris 1 -HCl plus 0.1 M NaCl at pH 7.0 or 6.5; 0.1 M phosphate at pH 7.18.
Analysis of the Heme Transfer Experiments with Immobilized ␣␤ Dimers-Heme transfer from immobilized met-␣␤ dimers to albumin can be described by the minimal reaction scheme, where ␣␤ is the immobilized dimer, A stands for albumin and H for heme. In view of the different behavior of ␣ and ␤ chains, the kinetic analysis was carried out by approximating the overall reaction to the following equilibria, where ␣ and ␤ are the hemoglobin chains in the immobilized ␣␤ dimer. An iterative diagonalization of the rate constants matrix relative to the processes described by the equilibria (Equations 1-3) was used to fit the time course of methemalbumin formation. The second order processes were linearized by assuming dH/dt ϭ 0 (2). In the fitting procedure: (i) the rate constants k A and k ϪA were fixed at the values determined in independent experiments, i.e. at 5 ϫ 10 4 M Ϫ1 s Ϫ1 and 3.2 ϫ 10 Ϫ4 s Ϫ1 ; (ii) the concentrations of heme-depleted ␣ and ␤ chains as well as the dissociation rate constants k Ϫ␣ and k Ϫ␤ were allowed to float; (iii) the amount of heme associated with the globin chains and with albumin was fixed to the value determined experimentally at the end of the reaction (see Fig. 8), thereby fixing the values of the rates for heme binding, k ␣ and k ␤ , at each iteration. Heme Transfer Experiments with Soluble Hemoglobin-In these experiments human hemoglobin after oxidation with an excess of potassium ferricyanide was equilibrated with 0.1 M Tris-HCl plus 0.1 M NaCl and 0.3 M sucrose, pH 9.0, on a Sephadex G25 column. Hemoglobin solutions at 5 ϫ 10 Ϫ6 M (in heme) were mixed with the same volume of 5 ϫ 10 Ϫ6 M albumin in the same buffer and the heme transfer reaction was followed at 410 nm in a Hewlett-Packard 8542A spectrophotometer for about 180 min at 20°C. The kinetics of heme transfer was analyzed in terms of three first order processes corresponding to: 1) heme release from the ␤ chains in ␣␤ dimers, 2) heme release from ␤ chains in ␣ 2 ␤ 2 tetramers, 3) a slow process which comprises heme release from the ␣ chains and heme release from methemalbumin. In the fitting procedure the relative amplitudes of processes 1 and 2 were fixed at the amounts of dimers and tetramers present at equilibrium on the basis of the dimer-tetramer association constant measured in parallel ultracentrifugation experiments. The amplitude of the slow phase and the three rate constants were allowed to float.
Reaction of Albumin with Soluble Hemin-Titration experiments were carried out at a constant albumin concentration of 1.3 M in 0.1 M Tris-HCl plus 0.1 M NaCl, pH 9.0, by adding small volumes of a hemin solution in the same buffer at 20°C. Given the low concentrations involved a 10 cm cell was used. Formation of the methemalbumin complex was followed at 403 nm. At any given point in the titration the contribution of free hemin to the observed absorbance was subtracted. The data were analyzed according to a simple two sites Adair equation, where Y is the fractional saturation of albumin with heme, [H] is the concentration of free hemin, and K 1 and K 2 are the affinity constants of the two albumin binding sites. The kinetic parameters of the reaction were measured at 20°C on an Applied Photophysics (Applied Photophysics Ltd., Leatherhead, United Kingdom) stopped flow apparatus (dead time 3 ms) upon mixing hemin with albumin in 0.1 M Tris-HCl plus 0.1 M NaCl, pH 9.0. All reactant concentrations are after mixing unless otherwise stated.
All the algorithms used to fit the experimental data were elaborated with the software package Matlab (The Math Works Inc., Natick, MA).
Sedimentation velocity experiments were performed using a Beckman model XL-A analytical ultracentrifuge at 40,000 rpm and 20°C over the concentration range 0.0043-0.5 g/dl (0.27-3.1 ϫ 10 Ϫ5 M heme). The gradient of protein concentration in the cells was determined by absorption scans along the centrifugation radius at a single wavelength (410, 540, or 630 nm) with a step resolution of 0.001 cm. Sedimentation coefficients were evaluated with the software provided by Beckman and were reduced to s 20,w according to standard procedures. The weight fraction of tetramers, ␣, at any given concentration c, was calculated on the basis of the measured value of s 20,w (a weight average property) and of s 4 c Ϫ s 2 c , where s 4 c and s 2 c correspond to the sedimentation velocity of tetramers and dimers, respectively, at concentration c. In turn, s 4 c ϭ s 4 0 (1 Ϫ 0.07c) and s 2 c ϭ s 2 0 (1 Ϫ 0.07c) with c expressed in g/dl (16). The value of s 4 0 , the sedimentation coefficient for the tetramer at zero protein concentration, namely 4.7 S, was used to calculate s 2 0 , the corresponding value for the dimer, assuming that sedimentation coefficient is proportional to (M r ) 2/3 (16,17). The curves for K 2,4 , the dimertetramer association constant, were calculated from the mass law expression according to Ref. 17.
Sedimentation equilibrium experiments were performed using a Beckman model XL-A analytical ultracentrifuge at 20,000 rpm and 10°C in 0.1 M Tris-HCl plus 0.1 M NaCl, pH 9.0, over the concentration range 0.75-5.0 ϫ 10 Ϫ6 M heme. The data (20 averages/scan) were analyzed with the software "Multi" for self-associating systems provided by Beckman.

RESULTS
Heme Binding to Albumin-The reaction between heme and albumin has been reinvestigated prior to performing the heme transfer experiments. The titration curve of albumin with hemin (Fig. 2) was analyzed according to Equation 4 and points to the presence of two high affinity binding sites for hemin, characterized by K 1 ϭ 7.8 ϫ 10 7 M Ϫ1 and K 2 ϭ 2.3 ϫ 10 8 M Ϫ1 in good agreement with the values reported in Ref. 11.
The kinetics of the reaction has been investigated by mixing in a stopped flow apparatus 3.6 M hemin with albumin at concentrations varying between about 20 and 250 M. Hemin was used at a low concentration to minimize its tendency to polymerize. The time course is monophasic and has been analyzed as a first order reaction. The pseudo first order rate constant (Fig. 3) depends linearly on albumin concentration up to approximately 25 M; at higher protein concentrations a constant value of 1.3 s Ϫ1 is reached. This value, which relates to a rate-limiting monomolecular step, is close to the rate of hemin depolymerization (18). From the data in Fig. 3 an association rate constant for hemin binding to albumin of 5 (Ϯ0.9) ϫ 10 4 M Ϫ1 s Ϫ1 can be obtained. This value in combination with the average affinity constant determined independently (see above) yields an average rate of heme dissociation from albumin of 3.2 (Ϯ0.11) ϫ 10 Ϫ4 s Ϫ1 .
Heme Transfer Experiments from ␣ and ␤ Chains to Albumin-Either type of chain releases heme to albumin at pH 9.0 not only when oxidized, but also when in the CO-liganded state. The shift of a substantial amount of heme from CO-bound ␣ and ␤ chains to albumin is clearly shown by the electrophoresis patterns of Fig. 4. Under similar experimental conditions, there is no heme transfer from CO-liganded HbA to albumin as reported by Bunn and Jandl (4).
Immobilized CO-liganded ␣ and ␤ chains exposed to albumin at pH 9.0 and 20°C behave similarly to the soluble ones and give rise readily to a significant amount of methemalbumin (Fig. 5A). It is of interest that methemalbumin is formed indicating that CO is lost upon or before binding. An equilibrium distribution is attained after about 180 min; at equilibrium, when albumin is in 3-fold molar excess (in terms of heme binding sites), about 50% of the heme is in the form of methemalbumin. Under similar experimental conditions immobilized oxidized chains loose all their heme (Fig. 5A).
The heme transfer process displays two clearly separated kinetic phases (Fig. 6). The time course was fitted using the scheme and the approximations presented under "Materials and Methods" by considering the contribution of only one chain. The rate of the slow phase was taken as 3.2 ϫ 10 Ϫ4 s Ϫ1 , which corresponds to the rate of heme dissociation from albumin estimated from the measurements on the heme-albumin system. The rate of the fast phase, which can be assigned to heme release from the ␣ or ␤ chains is 4.5-7.5 ϫ 10 Ϫ2 s Ϫ1 and does not depend within experimental error on the chain type and on the state of heme oxidation (Table I).
In a further set of experiments the observation that methemalbumin is formed upon transfer of heme from the CO-bound chains was exploited as it enables measurement of heme transfer from soluble chains to albumin. Upon mixing ␤ chains at 7  (2), a biphasic reaction is observed (Fig. 7). The fast rate, 0.2 ϫ 10 Ϫ2 s Ϫ1 , is roughly 5-fold slower than that measured with immobilized ␤ chains.
Heme Transfer Experiments from Immobilized ␣␤ Dimers to Albumin-It was first established that immobilized ␣␤ dimers, like soluble hemoglobin, do not transfer heme to human albumin when in the oxy, CO, or CN-met form, but dissociate heme rapidly in the ferric, met-aquo form (Fig. 5B). This finding is of importance as it shows that the immobilization and lyophilization steps do not produce significant alterations in the heme environment.
Thereafter heme transfer experiments were carried out by mixing different amounts of immobilized met-aquo ␣␤ dimers with a constant concentration of albumin at pH 9.0 and incubating the mixture at 20°C. Under these experimental conditions an equilibrium distribution is reached within 24 h. The heme transferred from the immobilized dimers to albumin at equilibrium is around 30% when heme is in excess over albumin and approaches 100% at the lowest heme/albumin ratio tested (Fig. 8); conversely, the saturation of albumin with heme increases with increase in heme/albumin molar ratio (data not shown).
The time course of methemalbumin formation at different heme/albumin molar ratios is given in Fig. 9A. It can be described by a fast and a slow process whose apparent rate differs by approximately 50-fold, both rates increase slightly upon increasing pH from 6.5 to 9.0 (see Table I). The amplitude of the  fast process predominates at the higher heme/albumin molar ratios (Table II). Bunn and Jandl (4), when studying the exchange of hemes between hemoglobins A and F, observed that the process is biphasic and proposed that the fast rate reflects the dissociation of heme from non-␣ chains and the slower one dissociation from the ␣ chains. This assignment has been confirmed recently by Hargrove et al. (2) with the aid of mutant hybrid methemoglobins and valence hybrids in which one subunit is oxidized.
The data in Fig. 9 can be interpreted in the same way: when hemoglobin is in excess, rapid dissociation of heme from the ␤ chains predominates as indicated by the greater amplitude of the fast process; when albumin is in excess, the contribution of heme dissociation from the ␣ chains becomes relevant and the amplitudes of the fast and slow process are approximately equal. Under all conditions, as the reaction proceeds, there is an accumulation of the heme-albumin complex, and thus, a significant contribution of the rate of hemin dissociation from albumin, in particular to the slow phase. This interpretation was substantiated by the analysis of the heme transfer reaction carried out as outlined under "Materials and Methods." Both rate constants pertaining to albumin, k A and k ϪA , 5 ϫ 10 4 M Ϫ1 s Ϫ1 and 3.2 ϫ 10 Ϫ4 s Ϫ1 , respectively, were calculated using the data of Figs. 2 and 3. Likewise, the equilibrium partitioning of hemin between albumin and the immobilized ␣␤ dimers was fixed at the value determined at the end of the experiment reported in Fig. 8. The global fit of the whole set of data yields rate constants of 6.5 ϫ 10 Ϫ4 and 2.8 ϫ 10 Ϫ2 s Ϫ1 , respectively, for the slow and fast phase; the fitted time courses are shown in Fig. 9.
Heme transfer experiments similar to those just described were carried out also at pH 7.5 and 6.5. The amount of heme transferred to albumin at equilibrium increases at any given  pH with decrease in the heme/albumin molar ratio; at any given ratio of albumin to heme it decreases with decrease in pH. The rate constants which describe the time course of heme exchange are affected only slightly by pH (Table I). Representative fits at pH 7.5 are included in Fig. 9. Last a set of heme transfer experiments from soluble hemoglobin was performed at pH 9.0 and 20°C. These measurements were designed to compare under the same conditions of pH and temperature the rate of heme loss from immobilized dimers with that from soluble dimers. To this end it appeared necessary to determine the dimer-tetramer association constant of methemoglobin by ultracentrifugation. To our knowledge there are no reports in the literature on this equilibrium. Under the conditions used for the heme transfer experiments, namely Tris buffer I ϭ 0.1 M ϩ 0.1 M NaCl at pH 9.0 and 20°C, the dimer-tetramer association constant was found to lie between 3.2 ϫ 10 4 and 8.0 ϫ 10 4 M Ϫ1 (dimer basis) by sedimentation velocity (Fig. 10). Preliminary sedimentation equilibrium experiments carried out at 10°C yielded 5 ϫ 10 4 M Ϫ1 . In view of the small temperature dependence of the dimerization reaction at alkaline pH values (⌬H°4 ,2 13.2 Ϯ 2 kcal/mol; Ref. 19), the K 2,4 value corrected to 20°C, 9 ϫ 10 4 M Ϫ1 , is close to the upper limit obtained by sedimentation velocity. Thus, when the concentration of soluble met-hemoglobin is 2.5 ϫ 10 Ϫ6 M (heme), the fraction of dimers is close to 90%. The fit to the heme transfer data at this hemoglobin concentration, carried out as described under "Materials and Methods," yields rates of 1.1 ϫ 10 Ϫ2 s Ϫ1 and 0.2 ϫ 10 Ϫ2 s Ϫ1 for heme loss from ␤ chains in dimers and tetramers, respectively, and 3.8 ϫ 10 Ϫ4 s Ϫ1 for the slow process (Fig. 10). The rate of heme release from ␤ chains in soluble dimers is in very good agreement with that from immobilized ones (Table I). Likewise the rate of the slow process is fully consistent with the values obtained for heme loss from chains in the dimer (Table I) and from methemalbumin.
Sedimentation velocity experiments were also carried out in 0.15 M phosphate buffer, pH 7.0 and 20°C, in order to assess the amount of tetramers present in solutions of met-hemoglo-bin under conditions similar to those used by Hargrove et al. (2). Dissociation into dimers is less pronounced than in Tris buffer at pH 9.0 (K 2,4 ϭ 1.5 ϫ 10 5 M Ϫ1 ). This finding is not unexpected since dissociation into dimers of oxyhemoglobin, which entails cleavage of the same ␣ 1 ␤ 2 interface, is likewise enhanced at alkaline pH values (20). DISCUSSION The experiments presented here show that the stability of the heme-globin linkage in ␣␤ dimers and isolated chains of human hemoglobin can be probed by studying the heme transfer reaction from the immobilized proteins to albumin. Soluble and Sepharose-bound heme-proteins transfer heme to albumin in a similar fashion. From a qualitative viewpoint, only those derivatives which transfer heme to albumin in solution do so when immobilized, and conversely, those derivatives which do not release heme to albumin in solution do not release heme when bound to Sepharose (Fig. 5, A and B). At a quantitative level the direct comparison carried out with both ␣␤ dimers and isolated chains indicates that the immobilization step does not alter the heme environment significantly while providing material that is not easily denatured after heme loss. Thus, the rates of heme transfer from the immobilized proteins are increased only 3-to 6-fold relative to the soluble ones (Table I). Immobilized heme donors have other advantages; for example they do not undergo changes in state of association, which may complicate analysis of the heme transfer reaction in solution, and can be separated easily from the heme acceptor. The latter property enables one to monitor the relevant spectral changes easily even when the spectral properties of heme donor and acceptor are very similar. In brief, immobilized heme donors provide a solution to several problems that have limited the study of the heme transfer reaction in solution (2,5,6).
In the present work human albumin has been used as the heme acceptor, because its affinity for heme is comparable with that of human hemoglobin (11). This feature permits determination not only of the kinetic, but also of the equilibrium aspects of the heme transfer reaction. In turn, knowledge of the amount of heme partitioned at transfer equilibrium between the hemoprotein and albumin provides a useful constraint in the fit of the time courses to the reaction scheme.
The most interesting finding of this study is that the stability of the heme-globin linkage changes dramatically upon formation of ␣␤ dimers. This change is especially striking in the ferrous CO-bound state: at pH 9.0 and 20°C the isolated chains transfer heme significantly to albumin over a time scale of a few hours, whereas no heme release takes place from ␣␤ dimers (Figs. 4 and 5), which behave like the ␣ 2 ␤ 2 tetramer (4). The assembly of unlike chains, therefore, brings about a major rearrangement in the heme pocket of the ferrous protein which results in considerable strengthening of the heme-globin interaction, such that in practice heme dissociation cannot be measured.
Heme oxidation enhances heme release in human hemoglobin (2,4,5) and does so also in the ␣␤ dimers (Fig. 5B), but has a very small effect on the heme transfer properties of the isolated chains (Table I). This difference in behavior can be ascribed to the different nature of the ferric forms of the proteins. The isolated chains give rise to six-coordinate low spin hemichromes, in which the iron-proximal histidine bond is effectively covalent, while HbA and its immobilized dimers form high spin met-aquo derivatives in which this bond is weakened. It is of interest that isolated ␣ and ␤ chains loose their heme at approximately the same rate (1-6 ϫ 10 Ϫ2 s Ϫ1 ), whereas in the ␣␤ dimer heme dissociation from the ␣ chains is some 50 times slower than from the ␤ chains (e.g. 0.065 ϫ 10 Ϫ2 s Ϫ1 versus 2.8 ϫ 10 Ϫ2 s Ϫ1 at pH 9.0 and 20°C). In turn release