The Association Reaction between Hemoglobin and Carbon Monoxide as Studied by the Isolation of the Intermediates

Transcription

1 THE JOURNAL OF BIOLOGICAL CHEMISTRY by The American Society for Biochemistry and Molecular Biology, Inc. Val. 267, No. 13, Issue of May 5, pp , 1992 Printed in U. S. A. The Association Reaction between Hemoglobin and Carbon Monoxide as Studied by the Isolation of the Intermediates IMPLICATIONS ON THE MECHANISM OF COOPERATIVITY* (Received for publication, June 7, 1991) Michele PerrellaSB, Norman DavidsV, and Luigi Rossi-Bernardi From the SDipartimento di Scienze e Tecnologie Bwmediche, Via Celorin, 2, Milano, Italy and the IIPennsylvania State University, University Park, Pennsylvania The concentrations of the intermediates in the asso- inequivalent. They found that the free energy of dimer assemciation reaction between human hemoglobin CO and at bly into tetramers of species 11, 12, 21, and 22 is different 20 OC, ph 7, under conditions of negligible dissociation from the energy of species 01 and the energy of the remaining of the ligand, were measured by cryogenic techniques. species. According to Ackers and Smith (5), the discovery The monoligated species were predominant at all val- that intermediates in different states of ligation (ie. the ues of overall ligand bound studied. The analysis of the monoligated and some diligated species) share the same value experimental data assuming a scheme of four consec- of this thermodynamic property and that intermediates in the utive reactions indicated that the binding rates insame state of ligation, such as the diligated species,have creased in a continuous fashion. A significant accelerdifferent energies, is in contrast with the predictions of the ation after the binding of the second molecule of ligand occurred in the presence 0.1 of M KCl, but not with the mechanism developed by Monod, Wyman and Changeaux to addition of an excess of inositol hexaphosphate, indidescribe cooperativity (6). Furthermore, Perrella et al. (7) cating that major functional, and possibly structural, have shown that species 21 and 22, both with one a and one transitions occur at the diligated state. p subunit in the ligated state, have different energies. Species Differences in the concentrations of the intermedi- 21 is likely to have the same energy as 11 and 12, and species ates in the same state of ligation were observed under 22 has the same energy as 23 and 24. This indicates that the all conditions. The analyses of the data on the basis of observed energy difference between species 21 and species 22, schemes of multiple pathways of reaction indicated 23, and 24 is related to the different configurations of the that the 0 subunits reacted about 1.5 times faster than ligated subunits within the tetramer and not to the structural the a subunits in the first ligation reaction. After the differences between the a and (3 subunits. To describe coopaddition of inositol hexaphosphate, the a subunits re- erativity in this, as in other models of ligation (8), Ackers and acted about 1.5 times faster than the 0 subunits in the first ligation step, but the overall rate of the first CO binding step was unchanged. The configuration of the four subunits, two a and two p, in the hemoglobin tetramer confers to the molecule a 2-fold axis of symmetry, which is mantained in the transition from the unligated, T, to the ligated, R, state (1). The subunits are structurally inequivalent, a crucial feature for cooperativity in hemoglobin; for it is well known that the tetramer of hemoglobin H, composed of four (3 subunits, is uncooperative (2). Scheme 1 shows a topographic representation of the 10 ligation states and the various pathways leading from deoxyhemoglobin, 01, to fully ligated hemoglobin, 41. The scheme also indicates the different intersubunit contacts. The a& and a& contacts remain unchanged in the transition from T to R, the alp2, azb1, and the weak alaz contacts depend on the state of ligation (3). Smith and Ackers (4) were the first to demonstrate in a model system, where ligation is mimicked by cyanide bound to the ferric heme, that the pathways from 01 to 41 can be * This work was supported by the Consiglio Nazionale delle Ricerche, Rome. 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. 5 To whom correspondence and reprint requests should be addressed. a744 Smith (5) formulated a combinational global switch mechanism, which uses the concept of induced-fit from the sequential mechanism of cooperativity developed by Koshland et al. (9). Carbon monoxide is the closest approximation to the physiological ligand of hemoglobin. Methods are now available for trapping the intermediates in the reaction between hemoglobin and CO (10-12), which provides the opportunity to verify whether the discoveries of Ackers and collaborators pertain only to artificial models of ligation or can be confirmed in a system where the ligand reacts reversibly with the iron protein in the ferrous state, in the absence and presence of modulators. Perrella et al. (12) have measured the concentrations of the intermediates in the equilibrium reaction between hemoglobin and CO and used the relationship between the concentrations and the free energy of assembly of dimers into tetramers for the intermediates (5) to calculate the values of the energies compatible with the observed concentrations. They calculated an energy value intermediate between those of Hb and HbCO for the monoligated species, as was also found in the deoxy/ cyanomet system, but the energetics of the two systems differed in regard to the diligated species. The concentrations of the diligated species calculated assuming the distribution of energy values of the ligation model of Ackers and collaborators were much higher (e.g. >60% of the total at 50% CO saturation) than the values observed for these species (about l%), some of which were below the limit of detection by the method. The energies of the diligated species, calculated to be com- M. Perrella and L. Rossi-Bernardi, manuscript in preparation.

2 01 Intermediates in Association Reaction between Hemoglobin and Carbon Monoxide 8745 ;J 4' species are minor and point to the diligated state as the step where a major functional, and possibly structural, transition occurs. A difference in the concentrations of the intermediates in the same state of ligation was found under all conditions. However, because of the experimental error, such a difference was significant mainly for the monoligated intermediates. In 0.1 M KC1, 10 mm phosphate, ph 7, the heterogeneity was explained by assuming that the /3 subunits reacted faster than the a subunits, at least in the first binding step. After addition of IHP, the order of reactivity toward CO of the subunits was reversed, but the overall rate of binding the first CO molecule was unchanged. The implications of these findings for the mechanism of hemoglobin cooperativity, particularly in regard to the question addressed above, are discussed. SCHEME 1. Topography of the ligation states in hemoglobin (4). The ten ij species are identified by the state of ligation i (i = 1-4) and an arbitrary index j 6 = 1-4). The configuration of the 01 and /3 subunits in a tetramer are identified in species 01. The lines joining the subunits indicate the intersubunit contacts. Arrows indicate the pathways in the on reactions of ligand L. The indexes which identify a rate constant kk,i refer to the reactant subunit, product subunit, and product ligation state, respectively. patible with this situation, have to be equal or close to the energy of HbCO, implying that the diligated CO-intermediates are similar in function and that their quaternary structures are similar to that of HbCO. These calculations indicate that there is no apparent contradiction between the functional behavior of hemoglobin in its interaction with CO, under the conditions of the equilibrium study of Perrella et al. (12), and the predictions of the Monod-Wyman-Changeaux model (6, 14). Are the mechanisms of cooperativity in the models of ligation studied by Ackers and collaborators and in the CO-hemoglobin interaction qualitatively different or are they different quantitative expressions of the same mechanism? In this paper, we report on the isolation of the intermediates in the association reaction between hemoglobin and carbon monoxide. The experimental approach consisted in mixing rapidly hemoglobin solutions at various concentrations with a buffer containing a constant substoichiometric amount of carbon monoxide. The time of reaction, about 100 ms, allowed quantitative binding of the ligand, but was not enough for an appreciable dissociation of the bound ligand. The intermediates were then trapped by cryogenic chemical quenching of the protein solutions, and the products of the chemical quenching were separated by cryogenic IEF' (10-12). This experimental approach does not provide a direct measurement of the rates of reaction, comparable with the classical kinetic stopped-flow experiments (15). A kinetic simulation of the observed distributions of intermediates allows the determination of the ratios of the rates along a reaction pathway. The results of this study in the absence and presence of IHP indicate that kinetic differences among the four diligated The abbreviations used are: IEF, isoelectric focusing; IHP, inositol hexaphosphate; BIS-TRIS, 2-[bis(2-hydroxyethyl)amino]-2-(hydroxymethyl)-propane-1,3-diol; Hb, deoxyhemoglobin; HbCO, carbonmonoxyhemoglobin; Hb+, methemoglobin; Hb+CN-, cyanomethemoglobin; SSE, sum of square errors. The symbol (ox) following a CO-hemoglobin intermediate indicates that the unligated hemes are oxidized. The rate constants throughout the paper, indicated as k instead of the classical 1' symbol (15), are "intrinsic," i.e. apparent constants divided by a statistical factor, unless otherwise specified. MATERIALS AND METHOD Preparation of HbAo from Adult Red Blood Cell Lysate-Oxyhemoglobin (4 g) was gel-filtered in 5 mm sodium phosphate, 0.5 mm EDTA, ph 6.8, and loaded onto a CM-cellulose column (8 X 30 cm) that was equilibrated with the same buffer. Elution was carried out with 7.5 mm sodium phosphate, 0.5 mm EDTA, ph The ph values of the buffers were measured at 20 "C, hut the column was equilibrated at 5 "C. This procedure separated all minor hemoglobins from the Hb& fraction. The HbAo fraction contained -1.7% Hb'. The protein was stabilized against further methemoglobin formation by the addition of catalase and superoxide dismutase (Boehringer Mannheim) and was stored in liquid nitrogen. The Hb' concentration was determined by the cyanomethemoglobin method (16). Trapping the Intermediates in the Association Reaction-The methodology for trapping the intermediates between hemoglobin and CO at equilibrium (10-12) was modified as follows. Deoxyhemoglobin, 1.2 to 6 mm (heme concentration), was prepared in a tonometer thermostatted at 20 "C. The CO solvent was prepared by deoxygenating 0.1 M KC1 in 10 mm phosphate, ph 7.0, in a glass bottle that was then filled with pure CO at a positive pressure of 2-5 cm Hg. The bottle was kept in a water bath at 20 "C. The buffer and the hemoglobin solution were drawn into the thermostatted syringes of a continuous flow apparatus (17) directly from the bottle and tonometer. This procedure was adopted because the design of the continuous-flow apparatus did not allow separate filling of the syringes. The concentration of carbon monoxide in the syringe after loading was not always reproducible, depending on the pressure of CO in the bottle, difference between room temperature and the temperature of the thermostatted syringe during the loading procedure, etc. About 2 ml of reactants were used to clear the reaction tube from oxygen and measure the ph and the hemoglobin concentration after mixing. A sample of reactants, ml, was injected anaerobically, ms after mixing, into a reactor at -30 "C that contained the quenching medium and was mixed by a Radiometer TTA-80 stirrer (Radiometer, Copenhagen). The quenching medium was 1 ml of thoroughly deoxygenated 20 mm sodium phosphate buffer, ph 7.5 at 20 "C, diluted with ethylene glycol to give a water/glycol ratio of about 1 to 1 after quenching and containing 5 to 10 pl of a deoxygenated 0.9 M solution of ferricyanide in water. The ph of the hydroorganic buffer, -8.5 at -30 "C (18), was increased to about 10 by the addition of an aliquot of 0.5 M Tris in 50% (v/v) ethylene glycol to a final concentration of 20 mm, 10 s after quenching. Separation of the Partially Oxidized Intermediates-The separation and identification of the quenched intermediates by low temperature isoelectric focusing have been described elsewhere (19). The separation was carried out on gel tubes (4 X 90 mm) thermostatted at -25 "C and loaded with p1 of the quenched sample. After h of focusing, the distance between Hb' and HbCO was mm. Only nine components were isolated since two intermediates, 21(0x) and 22(0x), have identical isoelectric points. The quantitation of the separated components was carried out by two different procedures, densitometry of color slides of the gels, and chemical assay of the components eluted from the gel (12). To assay the protein, the gels were extruded and sliced. The slices from four tubes containing the same components were pooled and suspended in 2 ml of 20 mm NaOH for 24 h. Pyridine (0.5 ml) was added, and, after 2-3 h, the

3 8746 Intermediates in Association Reaction between absorbance at 418 nm was read in a cuvette containing 20 mg of dithionite. The baseline was also determined by this method. A blank run in parallel with any separation of the intermediates was made by quenching a sample of Hb and a sample of HbCO in a ratio corresponding to the CO saturation of the experiment. The amount of protein loaded per gel tube was approximately similar to the amount of Hb plus HbCO that was present in the experiment. After focusing, the blank gel and the experimental gel were sliced similarly. The blank allowed correction for the product of hybridization of Hb+ and HbCO, traces of the products of incomplete oxidation of Hb and partial oxidation of HbCO, a smear of denatured protein, and turbidity of debris from the gel matrix. Controls-Controls were carried out to validate the procedure and, in particular, to check that (a) the unligated subunits were quantitatively oxidized by ferricyanide at low temperature, (b) no significant oxidation of the ligated subunits occurred during cryogenic quenching and removal of the excess oxidant by electrofocusing, (c) dimer exchange reactions did not alter the concentrations of the intermediates during cryogenic quenching and electrofocusing. These controls are summarized in previously published work (10-12) and are described in detail elsewhere.' An additional control of the chemical quenching was carried out by measuring spectrophotometrically the amount of CO reacted with hemoglobin. The syringes of the continuous-flow apparatus were reloaded after the quenching experiment, and a sample of reactants was injected into a Thurnberg cuvette containing 40 mm borate buffer, ph 9 (4 ml) and -15 mg of dithionite. The absorbance of the solution was read at 540, 555, and 568 nm before and after pure CO was bubbled into the cuvette. The precision was 1-2% CO saturation. The spectrophotometric value of CO saturation was then compared with the value calculated from the concentrations of the intermediates. The two values agreed in general within 5-8% CO saturation. Discrepancies greater than 2% CO saturation were mainly due to fluctuations in the CO concentration of the buffer which yielded different amounts of HbCO in the two separate mixing experiments required for such comparative measurements. Analysis of Data-A modification of the Finite Element Method developed by Davids and Berger (20) for kinetic problems was used. Briefly, the approach consists of a simulation of the kinetics to determine the current error and to provide feedback to calculate the amount of correction needed to iterate again, until by convergence, the best fit is obtained. In the simulation, the data were the experimental values of the concentrations of the intermediates and the parameters that were varied to fit these concentrations were the apparent rate constants. Such rate constants are calculated relative to a reference rate which may be chosen as unity. Thus, the simulation provides information only on the ratios of the rates unless the reference rate is truly representative of the absolute rate under the same experimental conditions. The value of 6.0 p"' s" measured by Gibson (21) in 50 mm phosphate, ph 7, and at 20 "C, for the binding rate ofco to triligated hemoglobin was assumed as the reference rate, unless otherwise specified. The structure of the computational loop in Fortran 77 was as follows (the symbols should be self-explanatory). First ligation step: and so forth for the successive steps. Time intervals, DT, of 0.01 and ms yielded parameter values differing by less than 2%. The initial values of the CO concentrations used in the simulations were calculated using the values of the overall CO saturation determined from the experimental concentrations of the intermediates, and the protein concentrations of the mixed reactants that were discarded to waste before quenching. Convergence was obtained readily using starting values of the parameters 2-10 times the values at convergence. In the analyses of complex schemes of reactions, convergence was obtained, but occasionally the analyses of some sets of data yielded negative values for some of the parameters since the program did not contain constraints in this regard. The procedure was to fit each set of concentrations of intermediates at a particular value of overall CO saturation and to average the values of the rate constants obtained from each fit. This procedure Hemoglobin and Carbon Monoxide was adopted to check for deviations due to systematic errors. On the other hand, a global fit through the data might give only a marginal advantage over this approach in the analyses of multiple reaction pathways given the limited number of data, the error, and the complexity of the problem. Assuming values of the rate constants from the literature (21-23), simulations were carried out to test whether (a) the concentration of free CO dropped to a negligible value in the time allowed for the reaction between hemoglobin and CO, under the conditions of protein concentration of this work, and (b) the off reaction ofco had a significant effect under these conditions. Table I shows the concentrations of intermediates relative to an experiment carried out at a value of overall CO saturation less than 25%. The experiment was carried out in duplicate. The concentrations of the intermediates obtained from two separate quenchings were corrected using two blanks differing (-30%) for the total amount of protein loaded onto the gels. Table I1 shows the results of the analyses of the blank-corrected data of Table I and the raw data (not shown in Table I) assuming a scheme of four consecutive reactions. The data shown in Table 111 were obtained similarly. RESULTS Distributions of Intermediates in 0.1 M KC1, 10 mm Phosphate, ph 7.0 The concentrations of the intermediates at different values of overall CO saturation are shown in Fig. 1. The functional heterogeneity among the subunits, ignored in Fig. 1, is shown in Fig. 2. Simple Scheme of Four Consecutive Reactions To compare the values of the rate constants with those obtained by Gibson (21), the data were analyzed on the basis of a simple scheme of four consecutive reactions assuming no functional heterogeneity among intermediates in the same state of ligation. The results of the analysis of each set of concentrations of intermediates at the different overall CO saturation values of Fig. 1 are shown in Table 111. All the data TABLE I Concentrations (%) of intermediates ij (see Scheme I) from a duplicated experiment at overall CO saturation S = 0.18 Columns l(a and b) and 2(a and b) show the data corrected using different blanks (see text). Conditions were: 20 "C, 10 mm phosphate, 0.1 M KCl, ph 7. GI la lb 2a 2b Average f f f 0.35 BO CO + B(1, 1) (@-subunit liganded) f 0.65 BO + CO + B(1, 2) (a-subunit liganded) f 0.95 D = KON(1, J, 1)*BO*CO*DT (KOFF term omitted) f 0.44 CO = CO - D f 0.36 BO = BO - D f 0.55 B(1, J) = B(1, J) + D S TABLE I1 Rate constants for a scheme of four consecutive reactions These reactions were calculated by the Finite Element Method (see text) using the blank-corrected values of the concentrations of the intermediates reported in Table I and assuming k, = 6.0 PM-' s" (21). Values of the rate constants calculated from the raw data are shown in parentheses. la lb 2a 2b Average kl (0.161) kz (0.354) f kg (2.01)

4 Intermediates in Association Reaction between Hemoglobin and Carbon Monoxide TABLE I11 Rate constants kl, k2, k3 calculated by the Finite Element Method assuming a scheme of four consecutive reactions and k4 = 6.0 pm' s-l (21) Conditions were: 20 "C, 10 mm phosphate, 0.1 M KCl, ph C Wbl" WR CS,: IC0lC P k, k, k3 SSE x 10' mm % mm S ( ) (-C 0.077) (& 0.325) Hb concentration at zero time of reaction. * Overall CO saturation at quenching time. M, spectrophotometric determination; C, value calculated from the concentrations of the intermediates and used for data analysis. Concentration of CO in the solvent at zero time of reaction, calculated using the (S)C value and the protein concentration after mixing. Time of reaction before quenching. fi FIG. 1. Fractional concentrations fi (i = 0-4) of the intermediates in different states of ligation i versus the overall CO saturation S, calculated from the fi values, at 20 "C, ph 7.00 f 0.05, in 10 mm phosphate containing 0.1 M KCl. Subunit heterogeneity is considered in Fig. 2. 0, Hb; 0, HbCO; 0, (see Scheme 1); X, ; A, The full lines were calculated assuming a reaction scheme of four consecutive steps with rates k, = 0.137, k, = 0.318, ks = 1.62, and k, = 6.0 p~"s-'. fiil%) 4 c 1 FIG. 2. Fractional concentrations fij of the intermediates ij (see Scheme 1) versus the overall CO saturation S, calculated from the fij values. Other conditions are as in Fig. 1. A: 0, 11; 0, 12; B: A, ; 0, 23; 0, 24; C 0, 31; 0, 32. The full lines were calculated assuming the rate constants averaged from the analyses of the data according to Scheme D (see text). shown in Table I11 refer to experiments run in duplicate and analyzed as described under "Materials and Methods." The solid lines in Fig. 1 represent the distributions of intermediates calculated for kl = (k 0.032), k2 = (k 0.077), k3 = (k 0.325), in FM-' s" units, as obtained by averaging the values reported in Table 111. Schemes of Multiple Pathways of Reaction The analysis of the data according to Scheme 1 could not be carried out because the number of variable parameters (16 rate constants) exceeds the number (8) of independent values of the concentrations of the intermediates. In addition, the concentrations of intermediates 21 and 22 were assumed equal since the corresponding components were not resolved by IEF. Such an assumption is commented on later. Four partial reaction schemes containing 5 to 8 variable parameters were explored. In each scheme, the values of the rate constants for the binding of CO to triply ligated hemoglobin, kl14 and kzl4, were set equal to 3.0 p ~ " s-l and the rate constants for CO binding to the CY and 0 subunits of Hb, klll and k121, were considered variable parameters. The schemes differed in the assumptions made on the functional heterogeneity of the subunits in the second and third ligation reactions. Scheme A (Five Parameters)-The rate constants for the binding of CO to the a subunits of the monoligated species were set equal to each other ( ki3, = kz12 = kzz2) and considered a variable parameter. This same assumption was made for the binding to the p subunits (kl12 = klz2 = kz4j. The rate constants in the third binding reaction were all set equal and their value was the fifth variable parameter. Scheme B (Six Parameters)-The same assumptions of Scheme A were made for the second binding reaction. The rate constants for the binding of CO to the a subunits of the diligated species were set equal to each other (kl13 = k213 = k423), and their value was considered a variable parameter. A similar assumption was made for the binding to the 0 subunits (km = km = k313). Scheme C (Seven Parameters)-In the second binding step, the rate constants for CO binding to the CY subunit to yield species 23, k132, and to the p subunit to yield species 24, kzc2,

5 8748 Intermediates in Association Reaction between Hemoglobin and Carbon Monoxide were considered variable. The rate constants for the binding of CO to the a subunit of species 12 to yield species 21 and 22 were set equal to each other, kz12 = kzz2, and were considered variable. Similarly, the rate constants for the binding of CO to subunit of species 11 to yield species 21 and 22 were set equal to each other, kllz = kiz2 and were considered variable. For the third ligation step, the assumptions made were as in Scheme A. Scheme D (Eight Parameters)-The assumptions made for the second ligation step were as in Scheme C, and those for the third ligation step were as in Scheme B. Table IV reports the values of the ratios k121/klll obtained from the analyses of the data according to Schemes A-D. The values of the rate constants for CO binding to and (Y subunits in the second and third ligation steps were not significantly different due to the large error. As explained above (see Materials and Methods ), a global fit to the data was not carried out. Instead, a separate analysis was carried out for each set of concentration data corresponding to the various values of overall CO bound. The rate constants from each set were averaged and used to calculate the SSE value relative to that set of data. The sum of the SSE values relative to all the sets of data, ESSE, obtained for Schemes A-D is also shown in Table IV to provide an approximate comparison among the various schemes. Distributions of Intermediates in 0.1 M KC1 Containing IHP, PH 7 The data obtained in the presence of 2 and 10 mm IHP were similar and are shown in Figs. 3 and 4. The analysis of the data yielded for the four-step scheme kz/kl = 1.73, k3/kz = 1.70, k4/k3 = The values of the rate constants depend on the value assigned to k4 in the presence of IHP. The analysis of the data in the absence of IHP, Scheme C (see above) gave: klll/k121 = 1.47 ( ). The kinetic heterogeneity of the a subunits in the successive ligation TABLE IV Ratios of the rate constants for CO binding to the and a subunits of deoxyhemoglobin at 20 C in 10 mmphosphate, 0.1 M KCl, ph 7, according to multiple pathway schemes (see text) Scheme k,m Jk, 1 I ZSSE x 10 fi A 1.62 f B C D f f as 0.4 Q2 a2 0.. FIG. 3. Fractional concentrations fi (i = 0-4) of the intermediates in different states of ligation i versus the overall CO saturation s, calculated from the fi values, at 20 C, ph 7.00 f 0.06, in 10 mm phosphate buffer containing 0.1 M KC1 and 2 or 10 mm IHP (four data points each). Subunit heterogeneity is considered in Fig. 4.0, Hb; 0, HbCO; 0, (see Scheme 1); X, ; A, The lines were calculated assuming a reaction scheme of four consecutive steps with rates k, = (f 0.031), k2 = ( ), ks = ( ), and k, = 1.4 pm sk1 (29). fij(%) 5 O-C P A 10 Q S FIG. 4. Fractional concentrations fij of intermediates ij (see Scheme 1) versus overall CO saturation S calculated from the fij values. Other conditions are as in Fig. 3. A: 0, 11; 0, 12; B: A, ; 0, 23; 0, 24; C 0, 31; 0, 32. The lines were calculated using the values of the rate constants averagedfrom the results of the analyses of the data according to Scheme C (see text). reactions was undetermined due to the large error. The averaged rates obtained from the analysis according to thi scheme were used to calculate the curves in Fig. 4. DISCUSSION Accuracy and Errors-The accuracy in the determination of the concentrations of the intermediates was checked by controls that indicated that the unligated subunits were specifically and quantitatively oxidized by ferricyanide, and no redistribution of ligated subunits occurred by dimer-exchange reactions during the quenching and electrofocusing procedures. In addition, two independent methods used for the determination of the concentrations of the components separated by IEF (i.e. densitometry and chemical assay of the protein; see Materials and Methods ) gave similar and reproducible results when applied to the analysis of various samples of partially oxidized HbCO. The chemical assay method was preferred because it allowed a more accurate estimation of the baseline. In the range of overall CO saturations, 25% < S > 75%, the error in the concentrations of the intermediates was higher than that observed at intermediate saturation values because of the paucity of some species. The uncertainty in the blank contributed greatly to magnify such an error. To illustrate a representative case among the experiments whose data are reported in Table 111, Tables I and I1 give details on the concentrations of intermediates and their analyses according to the scheme of four consecutive reactions. Columns 1 (a and b) and 2 (a and b) in Table I show the concentrations relative to two different quenchings yielding S = 18%, corrected using two different blanks (see Materials and Methods ). The error in the concentrations is 40% for the monoligated species, about 10% for the sum of intermediates 21 and 22, and varies from 20% to 100% for other minor species. The error in the calculated rate constants, shown in Table 111, is 10-30%. It is not surprising that the error in the rates is less than the error in the concentrations of the intermediates since the rate constants are calculated by a kinetic simulation involving the entire kinetic pathway. In addition, the largest errors often regard the concentrations of species that are minor with respect to the other species in the same state of ligation, e.g. in Table I the error in the concentration of species 23 is loo%, but the concentration of this intermediate is only 10% of the total concentration of the diligated molecules. A comparison of the values of the rate constants calculated from the blank corrected data and those obtained from the raw data, shown in parentheses in Table 11, illustrates how critical is the choice of the baseline under the conditions of low CO saturation. The differences, in percent, between the values of the rates from the blank-corrected data of Table 11

6 Intermediates in Association Reaction between (average) and the averaged values from the experiments at various CO saturations reported in Table I11 are about 40%, that is twice the error reported in Table 111. Such differences relative to the rates obtained from the raw data are the same as the errors reported in Table 111. At intermediate CO saturation values, the error was 5-10% for the monoligated species and 10-50% for other minor species. In the presence of IHP, the concentrations of the minor species were significantly increased, and the precision of the concentration measurements was greater. The rate constants shown in Table I11 do not appear to be affected by systematic errors except perhaps those due to the baseline problem discussed above. The error of the averaged rate constants relative to the various CO saturations is about 20%. The differences between the overall CO saturations measured spectrophotometrically and those calculated from the concentrations of the intermediates were usually in the range of 1-2% CO saturation, the same as the precision of the spectrophotometric measurements. However, in some cases, discrepancies up to 5-8% CO saturation were observed, as shown in Table 111. Such gross discrepancies were caused by occasional fluctuations in the concentration ofco in the buffer due to technical problems (see "Materials and Methods"). The value of overall CO saturation used for the calculation of the rate constants was that obtained from the concentrations of the intermediates, and its error was also in the range 1-2%, as shown in Table I. In conclusion, the errors shown in Table I11 (-f20%) appear to be a reasonable estimate of the errors in the determination of the rate constants for a scheme of four consecutive reactions. Clearly, the large error in the concentration of some intermediates has more dramatic effects on the determination of the rate constants according to schemes of multiple pathways of reaction. Since only the concentrations of the monoligated intermediates were significantly different, the various models explored in this work indicate that only klpl and kill were significantly different. Distributions of the Intermediates in 0.1 M KC1-The concentrations of the intermediates in the association reaction were significantly larger than the concentrations found at equilibrium under similar conditions and by the same cryogenic technique (11, 12). The monoligated species were predominant at all values of overall CO binding, as shown in Fig. 1. The ratios kz/kl = 2.32, k3/k2 = 5.09, k4/k3 = 3.7, and k4/kl = 44, calculated for a sequential scheme of four reactions, indicate that the rates increase monotonically with CO binding. The largest change after ligation of two CO molecules suggests that significant structural changes must occur at the diligated state. Fig. 5 shows the distribution of intermediates calculated assuming values of the rate constants kl = 0.11 k 0.01, kp = 1.00 f 0.45, k3 = 0.11 f 0.01, and k4 = 6.0 p ~-l s-l, FIG. 5. Fractional concentrations f, (i = 0-4) of the intermediates in different state of ligation i uersus the overall CO saturation S calculated from the fi values, assuming a reaction scheme of four consecutive steps with rates kl = 0.11, k, = 1.00, k~ = 0.11, and k, = s" from Ref. 21. Hemoglobin and Carbon Monoxide 8749 as calculated by Gibson (21) from stopped-flow data at 20 "C, ph 7, in 50 mm phosphate, indicating a non-monotonic increase in the rate constants. The high concentration of diligated species predicted by this simulation, which is far outside the range of error in our experimental determination of the intermediates, is not due to the error in the value of kp. If kl = k2 = k3 = 0.11, the calculated concentration of diligated molecules at 50% CO saturation is still greater than 35% of the total. The explanation is that the low value of k3 represents a "bottle-neck" in the chain of reactions. In preliminary experiments carried out in 50 mm phosphate, ph 7, the concentrations of the intermediates were similar to those found under the conditions of this work. This suggests that the discrepancy between the experimental distributions of intermediates and those predicted on the basis of Gibson's rate constants was not due to differences in the experimental conditions. Another simulation assumed that due to systematic errors in the cryogenic technique the concentrations of the diligated species were underestimated (50%) and those of the triligated species were overestimated (50%). The concentrations of Hb, HbCO, and monoligated species were recalculated to give the same experimental value of overall CO saturation, but were not found to be significantly different. The values of the rate constants that fitted the new distributions of intermediates were kl = f 0.016, kp = f 0.035, k3 = f p ~ " s-', showing again a monotonic increase in the rates. Our work is in substantial agreement with several aspects of the classical stopped-flow studies of CO binding to hemoglobin of Roughton and Gibson (15, 21, 24), such as the prediction of a larger population of intermediates under dynamic conditions than at equilibrium and the value of the rate constant for the binding of the first CO molecule ( pc"' s-'). However, our work does not confirm the indication from the stopped-flow experiments of a non-monotonic increase in the rate constants which requires the building-up of a high concentration of diligated species in the course of CO binding (24) and also under the conditions of our study, as shown in Fig. 5. Since the controls exclude a significant error in the cryogenic technique and the principle underlying the study of the association reaction between hemoglobin and CO in our work and in the stopped-flow studies is similar, we tentatively suggest that the spectrophotometric stopped-flow measurements are liable to misinterpretation because of nonlinearity effects due to the different optical responses of the a subunits, as reported by Gray and Gibson (25), and of the various intermediates, as suggested by Coletta and Geraci (26). Fig. 2 shows that the concentrations of the intermediates in the same state of ligation were different. The concentrations of species 21 and 22 could not be determined separately. However, the concentration of species 21 plus 22 was found to be equal, within error, to the sum of the concentrations of species 23 and 24. This suggests that, although differences in the rates of CO binding to the a subunits are possible, because of the different tertiary structures of the subunits, the different topological configurations of the subunits have only minor effects on the kinetic properties of species 21 and 22. The results of the analyses of the data in Fig. 2 according to simplified schemes of multiple pathways of reaction that are summarized in Table IV indicated that subunits react 1.5 to 2.1 times faster than the a subunits in the first ligation step. A slight difference in the rates of CO binding to

7 8750 Intermediates in Association Reaction between Hemoglobin ana! Carbon Monoxide the two subunits in the successive reactions is possible, but it could not be quantitated because of the experimental error. Similarly, the various schemes in Table IV fitted the data in a comparable way. Sharma (27,28) has studied the rates of CO binding to the CO intermediates 21, 23, and 24. A straightforward comparison with our results is not possible because complex kinetic schemes, involving also rates of dimer exchange and conformational reactions, were used by Sharma to interpret the data. However, our results agree qualitatively with Sharma's estimation that in the absence of phosphate subunits react faster than the a subunits in both the T and R conformations. Distributions of the Intermediates in 0.1 M KC1 Containing IHP-The increase in the concentrations of all the intermediates in the presence of IHP with respect to the absence of the phosphate is shown in Fig. 3. Ratios of the rate constants k2/kl = 1.7, k3/k2 = 1.7, k4/k3 = 3.7, and k4/kl = 14 were calculated assuming a scheme of four consecutive reactions. A comparison of these values with those obtained from the data in the absence of IHP clearly indicates that the organic phosphate makes the binding process less cooperative (decrease in k4/kl) and affects most significantly the kinetic properties of the diligated species (decrease in k3/k2). Gray and Gibson (29) reported values of k4 = 7.4 and 1.4 p"' s" for the binding of CO to triligated hemoglobin in 0.1 M NaCl buffered at ph 7 by BIS-TRIS in the absence and in the presence of IHP, respectively. If these values are taken as reference, the rate constants calculated from our data for the binding of the first CO molecule in the absence and in the presence of IHP are and kt"' s-', respectively. This suggests that the decrease in the value of k2/k1 from 2.3 to 1.7 on addition of IHP is due mainly to a decrease in the rate of binding of the second CO molecule. This conclusion should hold even if the reference rate constants do not apply strictly to our experimental conditions, as long as the ratio of the rates in the absence and presence of organic phosphate is the same. Fig. 4 shows that the concentrations of the a-ligated inter- mediates were greater than the concentrations of intermediates in the same state of ligation. However, the sum of the concentrations of species 23 and 24 was again equal, within error, to the concentration of species 21 plus 22. In the first binding step, the a subunits reacted about 1.5 times faster than subunits. The heterogeneity in the binding reactions of the subunits after the first step could not be estimated because of the experimental error. These observations suggest that: (a) IHP induces a change in the relative rates of binding to the subunits of the first CO molecule, but not in the overall rate of binding to deoxyhemoglobin, (b) IHP does not induce kinetic heterogeneity in species 21 and 22, which is consistent with the evidence discussed above that these species are kinetically similar and with the indication from the equilibrium study (12) that these species are structurally similar, and (c) IHP modifies the course of the CO binding reaction through a modulation of the kinetic properties of all the intermediates, but is most effective on the diligated species, confirming the crucial role of the diligated state in the functional and structural transitions in this reaction. Implications on the Mechanism of Cooperatiuity-The experimental distributions of intermediates in the binding of CO to hemoglobin and the phenomenological description of the data reported in this paper focus on two important aspects of this reaction. Firstly, the rates increased in a continuous fashion, with a significant acceleration after the binding of two molecules of CO, which was not observed in the presence of IHP. Within the experimental error, the four diligated species were similar in their kinetic behavior, both in the absence and presence of organic phosphate. In this regard, the results of this study under dynamic conditions are consistent with the results of the equilibrium study (12) and could be analyzed as well as the CO kinetics of Roughton, Gibson, and collaborators by the approach of Hopfield et al. (30) to show that they are consistent with the Monod-Wyman-Changeaux model of allostery (6). Secondly, the functional heterogeneity of the a subunits, significantly in the monoligated species, was observed to be dependent on the nature of the modulator. In 0.1 M KCl, subunits reacted about 1.5 times faster than the a subunits with the first CO molecule. On addition of IHP, the a subunits reacted about 1.5 times faster than subunits, but the overall rate of the first CO binding step was unchanged, within error. These findings are consistent with the observations of several investigators. Sawicki and Gibson (31) found that flash photolysis of CO from HbCO in the presence of IHP yields a quickly reacting hemoglobin species different from that obtained in the absence of phosphate, indicating that IHP binding yields an altered quaternary structure. Ho and collaborators (32, 33), using the NMR technique, found that O2 and CO binding is random at equilibrium in buffers containing chloride ions. On addition of 2,3-diphosphoglycerate, they observed an increased binding of O2 to the a subunits and, after addition of IHP, exclusive binding of O2 and preferential binding of CO to the a subunits. Perutz (34), in his stereochemical mechanism of the T to R transition in hemoglobin, has proposed that the tertiary structural changes induced by ligand binding to the hemes are coupled to the quaternary stability of the protein through the rupture of eight salt bridges. Tertiary structural changes have been demonstrated in crystals of deoxyhemoglobin reacted with CO at the a subunits (35). Functional studies on hemoglobin (36) suggest that intermediates in the R quaternary structure may have subunit tertiary structures different from the tertiary structures of the protein in the fully ligated state. This observation is consistent with our finding of a ratio for k4/k3 = 3.7, indicating significant differences in the intrinsic rates of CO binding between the di- and triligated intermediates and the results of our equilibrium studies (12). In fact, these showed that the di- and triligated CO intermediates have free energies of dimer assembly into tetramers similar to that of HbCO, suggesting that the quaternary structures of these intermediates are of the R type. All these studies clearly indicate that the elegant simplicity of the Monod-Wyman-Changeaux model of allostery (6) cannot account for the variety of the heterotropic interactions and their complex interplay and stress the importance of tertiary structural modifications within the two quaternary structures of hemoglobin that could represent additional intermediate thermodynamic states observed by Ackers and collaborators in their study of the artificial models of ligation (4,8). In conclusion, to answer the question posed in the introduction to this paper, the different energetics of the deoxy/ cyanomethemoglobin system and of the hemoglobin-co interactions could be the expression of quantitative differences between similar mechanisms in systems which differ in coop- erativity, as also pointed out by Ackers (37). The highly cooperative nature of the hemoglobin-co interactions, indicated by the high value of the Hill's coefficient, nh = 3.4 (121,

8 Intermediates in Association Reaction be tween Hemoglobin and Carbon Monoxide 8751 requires a low concentration of intermediates (38), which was indeed observed (12). By contrast, the complex between an oxidized subunit and cyanide reduces the cooperativity of the remaining deoxy sites, as proved by the low value of nh ( ) observed in the O2 binding studies to the cyanomet species 23 and 24 (39) and to the cyanomet intermediates of Scheme 1 that are prepared from hemoglobin stabilized against tetramer dissociation by intramolecular cross-linking, which retains much of the cooperativity of native hemoglobin (40). Acknowledgments-We are grateful to Marilena Ripamonti (C.N.R., Milano) for the preparation of hemoglobin Ao and to Stuart Davids for his collaboration in the modification of the Finite Element Method of data analysis. Kim Vandegriff, Robert Berger, and Giuseppe Geraci helped us in improving this paper with many useful discussions. A preliminary report on part of this work is found in Ref. 13. REFERENCES 1. Perutz, M. (1989) Q. Reo. Biophys. 22, Benesch, R., Benesch, R. E., and Enoki, Y. (1968) Proc. Natl. Acad. Sci. U. S. A. 61, Baldwin, J. M., and Chothia, C. (1979) J. Mol. Biol. 129, Smith, F. R., and Ackers, G. K. (1985) Proc. Natl. Acad. Sci. U. S. A. 82, Ackers, G. K., and Smith, F. R. (1987) Annu. Reu. Biophys. Biophys. Chem. 16, Monod, J., Wyman, J., and Changeaux, J. P. (1965) J. Mol. Biol. 12, Perrella, M., Benazzi, L., Shea, M., and Ackers, G.K. (1990) Biophys. Chem. 35, Smith, F. R., Gingrich, D., Hoffman, B. M., and Ackers, G. K. (1987) Proc. Natl. Acad. Sci. U. S. A. 84, Koshland, D. E., Nemethy, G., and Filmer, D. (1966) Biochemistry 5, Perrella, M., Benazzi, L., Cremonesi, L., Vesely, S., Viggiano, G., and Rossi-Bernardi, L. (1983) J. Biol. Chem. 258, Perrella, M., Sabbioneda, L., Samaja, M., and Rossi-Bernardi, L. (1986) J. Bid. Chem. 261, Perrella, M., Colosimo, A., Benazzi, L., Ripamonti, M., and Rossi- Bernardi, L. (1990) Biophys. Chem. 37, Perrella, M., Colosimo, A., Benazzi, L., Samaja, M., and Rossi- Bernardi, L. (1988) Symposium on Oxygen Binding Heme Proteins. Structure, Dynamics, Function and Genetics, Asilomar Conference Grounds, Pacific Grove, CA 14. Ackers, G. K., and Johnson, M. L. (1981) J. Mol. Biol. 147, Gibson, Q. H. (1959) Progr. Biophys. Biophys. Chem. 9, Evelyn, K. A., and Malloy, H. T. (1938) J. Biol. Chem. 126, Perrella, M., Benazzi, L., Cremonesi, L., Vesely, S., Viggiano, G., and Berger, R. L. (1983) J. Biochem. Biophys. Methods 7, Douzou, P. (1977) Cryobiochemsitry: An Introduction, Academic Press, New York 19. Perrella, M., Cremonesi, L., Benazzi, L., and Rossi-Bernardi, L. (1981) J. Biol. Chem. 256, Davids, N., and Berger, R. L. (1964) Commun. Am. Comput. Mach. ASSOC. 7, Gibson, Q. H. (1973) J. Biol. Chem. 248, Sharma, V. S., Schmidt, M. R., and Ranney, H. M. (1976) J. Biol. Chem. 251, Samaja, M., Rovida, E., Niggeler, M., Perrella, M., and Rossi- Bernardi, L. (1987) J. Biol. Chem. 262, Gibson, Q. H., and Roughton, F. J. W. (1957) Proc. R. SOC. Lond. B Biol. Sci. 146, Gray, R. D., and Gibson, Q. H. (1971) J. Biol. Chem. 246, Coletta, M., and Geraci, G. (1991) Eur. J. Biochem. 196, Sharma, V. S. (1988) J. Biol. Chem. 263, Sharma, V. S. (1989) J. Biol. Chem. 264, Gray, R. D., and Gibson, Q. H. (1971) J. Biol. Chem. 246, Hopfield, J. J., Shulman, R.G., and Ogawa, S. (1971) J. Mol. Biol. 61, Sawicki, C. A., and Gibson, Q. H. (1976) J. Biol. Chem. 251, Johnson, M. E., and Ho, C. (1974) Biochemistry 13, Viggiano, G., and Ho, C. (1979) Proc. Natl. Acad. Sci. U. S. A. 76, Perutz, M. F. (1970) Nature 228, Liddington, R., Derewenda, Z., Dodson, G., and Harris, D. (1988) Nature 331, Lee, A.W., Karplus, M., Poyart, C., and Bursaux, E. (1988) Biochemistry 27, Ackers, G. K. (1970) Biophys. Chem. 37, Gill, S. J., Richey, B., Bishop, G., and Wyman, J. (1985) Biophys. Chem. 21, Brunori, M., Amiconi, G., Antonini, E., and Wyman, J. (1970) J. Mol. Biol. 49, Miura, S., Ikeda-Saito, M., Yonetani, T., and Ho, C. (1987) Biochemistry 26,