Allosteric Properties of Carbamylated Hemoglobins*

The spectra, ligand-binding properties, and confor- mational change of carbamylated human hemoglobin A, specifically modified at its a-NHz groups, have been compared with those of hemoglobin A. In the Soret region, the spectra of all deoxyhemoglobins are iden- tical. The Soret bands of carbonmonoxy azc/3z and a2fiZc are shifted in opposite directions relative to hemoglobin A, azC/l2 having its Soret band at the longer wave- length. Carbamylation of either chain slows oxygen dissociation from the a chain in the R-state, although the effect is larger for azCflz and aZcf12C than a+". Carbamylation of either chain also slowed oxygen bmd-ing to the a chain while the rate for the B chain is not changed significantly, indicating that the chains influence each other in the R-state. "he rates of oxygen binding after full and partial laser photolysis of oxygen-saturated az'/lz appear identical. The rate of oxygen rebinding for a&", after partial (10%) photolysis is the same as that for a2c/#2, but following full photolysis, biphasic rebinding was observed due to the appearance of T-state molecules. Carbon monoxide binding studies under photostationary conditions suggested an L value for a2cflzc smaller than that for a2pZc and hemoglobin A. The R to T transition of the deoxy form of azC& is about 10 times slower than hemoglobin A. Carbamylation of the p chain does not slow the transition, indicating that carbamylation of the a chain affects the allosteric equi- librium

' The abbreviations used are: azc/32, a&', and (~~~/ h~, the hemoglobin specifically modified at the NH2 termini of the (Y chains, p chains, and all four subunits, respectively; Hb, hemoglobin; PMB, p-mercuribenzoate. of allosteric behavior in the absence of powerful effectors, and sometimes tend to dissociate to dimers more than Hb A. The carbamylated hemoglobins, however, may be expected to show a full range of allosteric properties, analogous to, but different from, those of Hb A.

RESULTS AND DISCUSSION
The carbamylated hemoglobins were prepared substantially as described in Refs. 1-3 with a number of minor modifications set out in detail in Appendix 1.' As a preliminary to the kinetic work, spectra were obtained for the deoxy and carboxy derivatives using a digital spectrophotometer (8). As described in detail in Appendix 2, the spectra of the deoxy derivatives were identical with those from hemoglobin A, but the carbamylated CO derivatives show shifts of the Soret band. The band is moved in opposite directions relative to Hb A for the two singly carbamylated derivatives. For azC,& the shift is about

nm toward longer wavelengths, while for a & '
it is about 213 as great and to shorter wavelengths. The a2',82' is almost identical with the azCB2. When the PMB derivatives were compared with PMB-treated Hb A, PMB-a' was closely similar to PMB-a; the PMB-pC gave a more complex spectrum suggesting sharpening of the Soret. There is a correlation between the position of the Soret band and affinity for ligand with the higher affinity derivatives having their Soret band at longer wavelengths.
Kinetics of Ligand Reactions-Data for the dissociation of oxygen from fully liganded hernoglobin were collected to represent a dissociation reaction from the R-state. The reactions are relatively slow, and as dimers and tetramers are alike, so far as is known, it is legitimate to use quite dilute solutions. The experiments also give estimates of the relative rates of oxygen and CO binding to the R-state without requiring the collection of additional data. The results, set out in detail in Appendix 3, may be summarized by saying that carbamylation of either chain slowed oxygen dissociation from the a chain though the effect was larger for a2& than for azfiZc. Doubly carbamylated hemoglobin resembled (YZ'PZ quite closely, and the control experiments with Hb A agreed well with the data of Olson et al. (lo), although the conditions were slightly different.
The ratio of the rate of 0 2 binding to the rate of CO binding showed a similar pattern. Carbamylation of either chain slows the rate of oxygen binding to the a chain relative to that of CO, but carbamylation of the 01 chain has a larger effect. It is doubtful if the rate for the / 3 chain is altered. * Portions of this paper (including Appendices 1-7 and additional references) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. These results were unexpected because there is little evidence that the chains influence each other in the R-state.
There is, in fact, not much difference between dimers and tetramers in their ligand reactions, and only quite moderate differences between separated chains and chains present as tetramers. Olson et al. (10) found only small differences between t~ chains free and in combination, and thought that polymerization might be responsible for the larger differences seen with p chains. As already mentioned, dimers account for a considerable part of the heme in the dilute solutions used in these experiments, and it is possible that there are interactions between the chains in them. It is unlikely that there are large differences between dimers and tetramers, as this would have shown up as heterogeneity of the time course of replacement of oxygen by carbon monoxide.
Oxygen Binding-All of the experiments were performed by laser photolysis which, if applied to a solution with a low initial oxygen saturation may give information about binding and dissociation of T-state hemoglobin. The amplitude of the excursion after photolysis is a good approximate measure of the initial saturation, and if the experiment is carried out using a tonometer the lower part of the dissociation curve may be obtained. If photolysis begins with a solution near saturation with oxygen, partial photolysis will usually give data on combination with the R-state, and comparison between full and partial photolysis may permit some conclusions about the rate of the R to T conformation change following photolysis.
Examination of azc/32 at low initial saturations gave a pattern of reaction rates which differs significantly from that of Hb A.
As shown in Appendix 5, Fig. 1, A and B, there is a slow rate of about 350/s which accounts for about one-half of the observed reaction at high saturations, and much more at the lowest ones. The rate increases with the initial level of saturation of the hemoglobin before photolysis. The second component has a mean rate of about 1500/s and also increases with increase in initial saturation. Not only is the affinity for oxygen very high, with p s of the order of 1 ~L M but the initial part of the equilibrium curve is much steeper still. As discussed in Appendix 5, this is Likely to be due to a small degree of inequivalence in the amounts of the chain preparations used, and is not considered further. As a start toward interpretation of the results, an approximate calculation of the value of L required to give the high affinity observed (halfsaturation at about 1 p~ 0 2 ) gives a value of only about 500, on the assumption that the properties of the R-and T-states are not changed. This is of the order of reciprocal c (the ratio of KT to KR). In consequence, the dissociation velocity from T will not contribute fully to the observed rates, most of which must be attributed to the combination reaction. The rate of 10 X 106/M/s is similar to the faster of the rates reported by Sawicki and Gibson (11) for Hb A. It would then be reasonable to assign the faster rate to the R-state. This would account for its rapid increase in amplitude and rate with increasing initial saturation. As the rates of dissociation of oxygen from the R-state have been measured to give an average of about 1O/s almost all of the observed rate must arise from the combination reaction, and the implication is that the rate of the R to T conversion is low for this derivative. If it were the same as for Hb A (12) virtually no R-state deoxyhemoglobin would survive the first 0.1 ms after the flash and the time courses at different saturations would be as much alike for (~2~f l 2 as for Hb A. Detailed evidence on this point is presented later. With these assumptions, the R-state rate becomes 60 X 106/M/s, which is the same as for Hb A (11). There are so many difficulties associated with any attempt at more detailed analysis that none is presented. The results appear compatible with the idea that carbamylation affects allosteric behavior rather than the ligand reactions of the hemes and the value of L required to fit in with this is at least lo00 times smaller than reasonable estimates for Hb A ( 7 , 8 ) .
The pattern of a&' reactions with oxygen is quite different. To begin with, the affinity for oxygen is considerably lower in the given experimental conditions, although still about twice that of Hb A in terms of the pressure of oxygen required for half-saturation. The time course of oxygen rebinding after f u l l photolysis of partially saturated hemoglobin closely resembles that of Hb A, showing at least two components, and three when data are collected over a wide range of initial saturations. The effect of changing initial saturation on the amplitudes of the components is the opposite of that for azcBz. The slow component, which is substantially slower, increases in amplitude with initial saturation, while the rapid one decreases. The rate of the rapid component does not change with saturation, which is consistent with all the hemoglobin reaching the same state (T) soon after the flash, and with attribution of the observed rate primarily to the dissociation reaction. It is suggested that the rapid reaction corresponds to the T-state reaction, as with Hb A, and that the slower component arises from the redistribution of ligand from a kinetically determined family of states to an equilibrium one in which cooperativity favors species with large numbers of ligand molecules over those with few. As discussed in Ref. 12, such a redistribution is a slow process, as dissociation is a necessary preliminary. Approximate calculations to give a rough idea of the value of L, with the same assumptions as used for m e / & , suggest about 4 X lo5, which is about 1000 times greater than the estimate for t~2 p 2~.
The discussion so far has centered on the behavior of deoxy-Hb. The results from photolysis experiments with saturated (air equilibrated) Hb differ from each other in a way consistent with the idea that the R-state is more readily populated in u Z c p~ than in a&', and that deoxy a2' pz' switches from R to T less quickly than a & ' . With t~2 ' & the rate and time course of oxygen binding after full and partial photolysis appear identical. With the course of rebinding after partial (10%) photolysis is much the same as that for aZcP2, but foliowing full photolysis a notably biphasic reaction is seen. This may be explained as due to the appearance of T-state within the time of the experiment with a2p2' but not with an'/?2. The approximate value of L is 1.5 X lo5 under the standard conditions used.
The properties of the final derivative azcp2' may be dealt with quite briefly. It is intermediate between a2'p2 and aZpZc, perhaps favoring a 2 ' P 2 over a2P2'. Its affinity is high, and the kinetics does not suggest rapid R to T conversion of the deoxy form. The details of these experiments are set out in full in Appendix 5, together with a figure showing the equilibrium data.
The results so far presented in the various appendices allow the interpretation that the main effect of carbamylation is upon the allosteric equilibrium, but the argument is circular in that the conclusion is used in calculating the allosteric parameters. The actual numbers obtained for L also depend on the applicability of the model to the data of Ref. 11. An obvious extension of the experiments is to use other ligands. The results for the kinetics of the carbon monoxide reactions suggest it as a suitable candidate because there is no effect of carbamylation on the accessible rates. Unfortunately, its high affinity makes it difficult to study the equilibria with hemoglobins, and there is no easy method to measure the rate of CO dissociation from the T-state. These drawbacks can be largely overcome by measuring the equilibria in a strong light. The quantum yield is high and the dark dissociation reactions slow. It is easy, therefore, to decrease the apparent a f i t y greatly, so making measurement easier, and substituting a known rate of photo dissociation for the poorly defined dark reactions. The procedure, suggested by Brunori et al. (13), has been used by Torkelson and Gibson (14) with menhaden hemoglobin to simplify the application,of the allosteric model. The residual difficulty remains that the partition between the R-and T-states depends upon L and c, and this is true in the light as well as in the dark. Equilibrium measurements made in a strong light must still depend, to some extent, on the values used for the dark dissociation reactions of carbon monoxide.
A complication in using human hemoglobin is its tendency to dissociate to dimers, and it is probably this difficulty which prevented Brunori et al. (13) from obtaining useful results with the dilute (2 to 4 ~L M in heme) solutions they used. This was not a problem for menhaden Hb (14) because many fish hemoglobins dissociate much less than mammalian hemoglobins (15). The difficulty was met by including suitable terms in the equilibrium equation, and also by using flash photolysis as the method for determining the fractional saturation. If the concentration of carbon monoxide is low enough so that recombination after the flash is not competitive with the R to T transition, the course of carbon monoxide recombination will show two phases, a faster one due to binding to dimers, and a slower due to binding to T-state tetramer. (Although true under most circumstances, there will be only one phase when the carbon monoxide is small and the concentration of T-state hemes is 30 times that of heme in dimers.) The amount of dimer may thus be measured and allowed for in interpreting the results. In principle, the appearance of dimers as a function of saturation of a hemoglobin solution is directly calculable from the model. In practice, it was found easy to represent the course of the appearance of dimer with increasing saturation (Appendix 6, Fig. l), and the flash method does indeed seem to measure this quantity. Although interesting in themselves, the results failed to contribute toward the specific matter of the allosteric parameters for the carbamylated hemoglobins. It was found in numerical trials that the value of L required to reproduce a given affinity (measured aspm) was very sensitive to the light dissociation rate specified to the model. This is so great as to make comparisons between experiments performed on different days of somewhat doubtf u l value. With this restriction, the results given in Appendix 6 do seem to show, however, that az'P2 and a2'P2' have smaller values of L than a & ' and Hb A. The next step in seeking to define differences in allosteric behavior between the carbamylated hemoglobins was to examine the kinetics of carbon monoxide binding over a wide range of ligand concentrations using the method of flash photolysis. The significant tetramer to dimer dissociation of the liganded derivatives gives an assurance that at least a substantial number of the tetramer molecules must be in the R-state. If ligand is removed from them quickly enough, the immediate product is deoxyhemoglobin in the R-state. These molecules may either combine at once with ligand, or they may change in conformation to the T-state, and, in doing so decrease their rate of combination with CO by some 30-fold. It is therefore easy to distinguish the molecules which bind in the R-state from those which bind in the T-state by observing the progress of ligand rebinding, and quantitative estimates of the rate of the R to T transition can be made by analyzing the progress curves.
Unlike the equilibrium experiments described earlier, the new results were unambiguous. Carbamylation of the a chain, alone or together with the p chain, gives a derivative for which the rate of the R to T transition of the deoxy form is some 10 times slower than the same change in Hb A. The rate of the R to T transition for (~$2' appears to be close to that of Hb A. These results show almost beyond doubt that carbamylation of the a chain affects the allosteric equilibrium in deoxyhemoglobin, and do so without requiring detailed calculations of values for L. Further, the behavior of the individual carbamylated hemoglobins agrees with the indications given by the combination of equilibrium and kinetic experiments described in earlier paragraphs. An additional method of observation of the R to T transition in deoxyhemoglobin has also been applied to experiments with carbon monoxide as ligand. This depends on the old observation that R-state deoxyhemoglobin has an absorption spectrum different from that of T-state deoxy. If, then, observations are made at the carboxy-deoxy isobestic point, the allosteric transition following flash photolysis may be observed directly rather than inferred from changes in ligand-binding rates. In hemoglobin A the change is so rapid that laser methods are required to follow it (16), but with the two a chain-carbamylated derivatives the change can readily be observed using apparatus based on conventional photographic flash tubes.
These kinetic experiments suffer from two weaknesses, fist, only the rate of the transition from R to T is accessible. The reverse reaction is much too slow to influence the observed kinetics. Second, simply interpretable results are available for the RO to TO change only. Although partially liganded intermediates can be prepared, the absorbance changes associated with their conformation switch are usually too small to follow with the precision required for useful results. A detailed description of the procedures and results is given in Appendix 7 and its associated figures.
General Discussion-The main conclusion to emerge from this preliminary study of the kinetics of ligand binding to the carbamylated hemoglobins is that, although there is some effect on the kinetics of individual ligand reactions for derivatives which may reasonably be regarded as in similar allosteric states, these effects are small as compared with those mediated through changes in the allosteric parameter L. The changes in intrinsic rate constants follow a definite pattern with significant changes associated with a chain carbamylation, and, where chain identification has been possible, it is the a chain which is affected (see Appendix 3, Table I). It seems to be true, also, that oxygen reactions are more affected than reactions of carbon monoxide. In comparison, the effects of fi chain carbamylation are small, but with a tendency to be opposite in sign to those of a chain carbamylation. In at least some of the properties studied, P chain carbamylation tends to cancel the effect of a chain carbamylation so that the doubly carbamylated compound is intermediate between the singly carbamylated forms.
The finding that the spectra of the CO compounds are shifted by carbamylation is unexpected. In this case too, the effects of a and P chain carbamylation are in opposite directions. It is difficult to relate the spectral~changes to allosteric effects, since all of the CO forms must be in the R-state. They are associated with the tetramer, however, since the effects on the isolated a chains were small by comparison with those in the tetramer. There is a suggestion that the R-state in the carbamylated hemoglobins is not identical with that of hemoglobin A under similar conditions. AU our work has been carried out in solutions buffered with phosphate, and it is, at present, only an attractive speculation that the spectral shifts are mediated by the inhibition of chloride binding reported by O'Donnell et al. (5) to follow a chain carbamylation. It is a speculation which leads to experimentally verifiable predictions for Hb A as well as for the carbamylated hemoglobins.
Turning to the allosteric effect of carbamylation, it is again evident that the a chain dominates, and that the allosteric equilibrium is indeed shifted in az'pz as compared with Hb A or with a&'. As with the intrinsic rates, the effects of a chain carbamylation are partly reversed by / 3 chain carbamylation.
Although the functional effects are large, it has not been possible to obtain sound estimates of the actual sue of L. It was particularly disappointing that the studies of carbon monoxide binding in the photostationary state failed in this respect, especially when the experiments were arranged so that L was the only parameter to be determined. Although numbers are not available, an interpretation of the results in terms of the two-state model was easy and natural, and allowed rationalization of the data for oxygen binding at low saturations, although the data themselves differed widely between the derivatives. The actual oxygen affinities observed under our conditions are hard to compare with previous results although we have found a larger difference between azcpZ and Hb A than O'Donnell et al. ( 5 ) . This may be explained in terms of their work on chloride and proton binding. Their experiments were performed in phosphate-free media chiefly at pH 7.6. We used 0.1 M KPi at pH 7. If, as seems likely, a chain carbamylation strongly favors the Rstate, by preventing phosphate from binding at a site near to Val la, the apparent difference between az' pz and Hb A will be greater the more the solution conditions favor the T-state for Hb A. The effect on p50 in our experiment (Appendix 3, Fig. 5 ) is only 3 times greater than that reported in Fig. 2 of Ref. 5 and there is no necessary conflict between their result and ours.
The rate of the R to T transition is affected by carbamylation in the direction to be expected on the basis of the equilibrium experiments, and the effects are considerable, with a 20-fold span between aZ'flz and Comparison of the traces for a&" with published data of Sawicki and Gibson (16) for Hb A, suggests that azp2c was little affected by carbamylation. Again, az' &' is intermediate between azcpz and In summary, the two-state model is able to rationalize a a2pzc.
complex series of results to a reasonable approximation with a minimum of assumptions, and the results are entirely consistent with the recent structural work of O'Donnell et al. (5).
Although it is axiomatic that kinetic data are incapable of proving the truth of a mechanistic hypothesis, the natural and easy description of the results so far obtained at least justify making further extension of observations to examine the predictions which may be made using the model together with