The effect of 2,3-diphosphoglycerate on the tetramer-dimer equilibrium of liganded hemoglobin.

Abstract The effect of 2,3-diphospho-d-glycerate (P2-glycerate) on the kinetic behavior of deoxyhemoglobin generated by the rapid dissociation of gaseous ligand (either by flash photolysis of carbon monoxide hemoglobin (HbCO) or by deoxygenation of oxyhemoglobin (HbO2) in the presence of sodium dithionite) is consistent with a stabilization of the tetrameric state of liganded hemoglobin by the organic phosphate. The fraction (α) of rapidly reacting hemoglobin produced by pulsed laser photolysis of phosphate-free HbCO in 0.05 m 2,2'-bis(hydroxymethyl)-2,2',2''-nitriloethanol-0.1 m NaCl, pH 7.0, was independent of CO concentration below about 500 µm, but increased systematically with dilution of the hemoprotein. The apparent tetramer-dimer dissociation constant (Kl4,2) calculated from the dependence of α on [HbCO] was 3.6 ± 1.0 µm in the absence of phosphates and decreased to 1.4 ± 0.3 µm when 1 mm P2-glycerate was added. These values of Kl4,2 are similar to estimates for liganded hemoglobin obtained in sedimentation experiments conducted at pH 7.0 in 0.1 m phosphate (Edelstein, S. J., Rehmer, M. J., Olson, J. S., and Gibson, Q. H. (1970) J. Biol. Chem. 245, 4372–4381) and in 0.1 m Tris-0.09 m NaCl (Kellett, G. L. (1971) J. Mol. Biol. 59, 401–424). The magnitude of the Soret absorbance drift accompanying deoxygenation of dilute solutions of HbO2 in the presence of sodium dithionite was decreased by added 1 mm P2-glycerate, and the second order rate constant characterizing the drift phase was increased at 20° from 0.54 ± 0.09 µm-1 s-1 to 1.57 ± 0.18 µm-1 s-1 by the organic phosphate. Since the drift has been shown to result from the formation of deoxy tetramers from deoxy αβ dimers (Kellett, G. L., and Gutfreund, H. (1970) Nature 227, 921–926), this result is also consistent with the proposed phosphate inhibition of dimer formation by liganded hemoglobin.

The Effect of 2,3=Diphosphoglycerate on the Tetramer-Dimer Equilibrium of Liganded Hemoglobin* (Received for publication, September 26, 1973) ROBERT D. GRAY From the Department of Bioclremistrg, ll~liversity of Louisville School of dletlicine, Health Sciences Center, Louisville, Keducky /,0201 SUMMARY The effect of 2,3-diphospho-D-glycerate (Pg-glycerate) on the kinetic behavior of deoxyhemoglobin generated by the rapid dissociation of gaseous ligand (either by flash photolysis of carbon monoxide hemoglobin (HbCO) or by deoxygenation of oxyhemoglobin (HbO,) in the presence of sodium dithionite) is consistent with a stabilization of the tetrameric state of liganded hemoglobin by the organic phosphate. The fraction (a) of rapidly reacting hemoglobin produced by pulsed laser photolysis of phosphate-free HbCO in 0.05 M 2,2'-bis(hydroxymethyl)-2,2',2"-nitriloethanol-O.l M NaCl, pH 7.0, was independent of CO concentration below about 500 pM, but increased systematically with dilution of the hemoprotein.
The apparent tetramer-dimer dissociation constant (Kt,2) calculated from the dependence of (Y on [HbCO]  The magnitude of the Soret absorbance drift accompanying deoxygenation of dilute solutions of HbOL in the presence of sodium dithionite was decreased by added 1 mM P2-glycerate, and the second order rate constant characterizing the drift phase was increased at 20" from 0.54 f 0.09 PM-' s-l to 1.57 f 0.18 PM-' s-l by the organic phosphate. Since the drift has been shown to result from the formation of deoxy tetramers from deoxy & dimers (KELLETT, G. L., AND GLTT-FREUND, H. (1970) Nature 227,[921][922][923][924][925][926], this result is also consistent with the proposed phosphate inhibition of dimer formation by liganded hemoglobin.
In 1934 Roughton (1) reported that deoxyHb' produced by dissociation of oxygen from HbOz in the presence of sodium * This work was supported by National Science Foundation Grant (;B-32283. 1 The abbreviations used are: deoxyHb, deoxyhemoglobin; dithionite maintains under certain conditions an anomalously high reaction rate with CO for a brief time. Since then a number of so-called "rapidly reacting" species of hemoglobin have been described, all of which are characterized by an increased ligand affinity and a slightly decreased Sorct absorption whose masimutn is redshifted apprositnatcly 2 nm (2). In each case the presence of the high affinity material can presutnably be traced to the transient or pertnanent existence of deoxyHb in the ligandbound confortnation.
For example, Ivhen the amount of CO removed from HbCO by flash photolysis is restricted to about 554, the rate of CO recombination at pH 7 is 3 ~1r-l s-l compared to an initial rate of about 0.4 ~~1-l s-l when total photolysis is achieved (3). These two rate extremes have been ascribed to CO binding to the intermediate species Hb,(CO)$ and to Hbr, respectively, and thus probably reflect structural differences (3, 4) between the ligand-bound and ligand-free conformations of the protein.
h second form of rapidly reacting hemoglobin was observed by Gibson (5) when Hash photolysis of sheep HbCO was carried out at low temperature and alkaline pI-I. The proportion of rapid recombinant increased in this instance with increased CO concentration, but was independent of IIbCO concentration.
On this basis Gibson suggested that this alkaline form of hemoglobin, designated IIb*, represents a transient form of deoxyHb which exists in the high affinity, ligand-bound conformation for a short period after photolysis.
h third form of quickly reacting hemoglobin appears at low hemoglobin concentration (6, 7). Recent experiments correlating sedimentation and kinetic data have shown that the rapidly reacting component produced by dilution at neutral pH in the concentration range 0.1 to 100 PM is the liganded C@ dimer (8).
The effect of removing endogenous phosphate compounds (9-11) on the kinetics of hemoglobin recombination with CO after Nash photolysis has been mentioned by several investigators. Gibson and Parkhurst (12), Gray (13), and MacQuarrie and Gibson (14) observed relatively large amounts of rapid hemoglobin when stripped HbCO was photolyzed.
It is obviously important to elucidate the structural basis for this increased proportion of rapidly reacting material which appears with stripped HbCO, both in order to interpret the kinetics and also to appreciate fully the role of organic phosphates in hemoglobin-l&and equilibria. The results of the kinetic experiments reported here (subject, of course, to the inherent inability of kinetic data alone to define unambiguously chemical structures) suggest that stripped, liganded hemoglobin dissociates to rapidly reacting dimers to a greater degree than when Pz-glycerate or inorganic phosphate is present.2 The following reaction scheme may facilitate presentation and discussion of the results. In the scheme, A$,2 and ky,2 are the rate constants for formation of dimers from liganded and unliganded tetramers, respectively; ki,q and ky,, are the corresponding rate constants for the dimer-dimer association reaction. The pseudo-first order rate constant for CO binding to qB dimers is k, and that for the slower binding of CO to tetramers is k,. A CO concentration is used so that insignificant adjustment of the tetramerdimer equilibrium occurs during the cyclic displacement of bound CO by the flash and the reassociation dark reaction.
where F is the fractional amount of the rapid species present at time (t) = 0 and kf and k, represent the apparent first order rate constants for the reaction of the rapid and slow species with CO. F, kf, and k, were treated as adjustable paramet'ers and the best value of each obtained by minimizing the error in the normalized absorbance changes using the normal equations for F, kf, and k,. The standard deviation of the parameters was determined from the error matrix.
In all cases at least a 5-fold excess of CO was present in order to assure the approximate pseudo-first order requirement for CO and Hb combination. Computations were carried out using programs written for the Nova 1200 computer. The experimental protocol for the stopped flow deoxygenation experiments depended on which reaction was of interest. Data collection was initiated 3 ms after flow-stopping switch for the deoxygenation reaction.
When the kinetics of the slow absorbance changes subsequent to deoxygenation was to be studied, data 3 The intensity of the measuring light beam did not significantly alter the results; see Table I. collection started 100 ms after flow stopping.
A reference voltage was collected 30 or 60 s later. This procedure had the advantage that by omitting observation of the initial deoxygenation (AA N 1) reaction, a greater sensitivity could be used to observe the slow reaction (AA N 0.1). Estimation of the actual absorbance change in the slow phase was difficult because: (a) at 429 nm the slow changes amounted usually to less than 10% of the total AA; (5) the slow second order reaction followed a first order reaction which made it impossible to establish accurately the "zero" time for the second order reaction; (c) occasional artifacts resulting from dithionite were experienced.
The requirements adopted for using a particular data set were: (a) the series must show linear l/AA versus time plots over 90% of the total AA at 429 nm; (b) no detectable difference in the AA measured whether the final reference voltage was collected 30 or 60 s after the last data point. An estimate of AA due to the slow second order reaction was obtained by extrapolating plots of the reciprocal of the absorbance change versus time to zero time.

RESULTS
Laser Photolysis of NbCO-The kinetics of recombination of stripped human Hb subsequent to photolysis in the presence and absence of Pz-glycerate is illustrated in Fig. 1. The kinetics was measured at 437 nm, a wavelength isosbestic for both the rapidly and slowly reacting species. In the absence of phosphates the rapid species contributed 32.5 + 0.4y0 to the observed absorbance change; when the sample was supplemented with 623 PM Pz-glycerate, the percentage of rapid material decreased to 25.3 f 0.6%. This analysis can, at best, be only approximate since it is well known that the second order rate constant for CO binding to Hb depends upon the fractional saturation with ligand (26). However, it does serve to point out the effect of P2-glycerate.
In experiments shown in Table I, 0.1 M phosphate also depressed the proportion of rapidly reacting hemoglobin to about the same extent. Fig. 2 shows that the fraction of rapid material depends on the wavelength of the observing light.
The lines are calculated from the data in spectrum of the isolated (Y and p chains and the slow species has the HbCO-deoxyHb difference spectrum of cooperative hemoglobin tetramers.
The experimental points of Fig. 2 show that the spectral characteristics of the rapidly and slowly reacting species compare favorably with that expected if the unliganded rapidly reacting species has the chain like HbCO-dcoxyHb difference spectrum characteristic of noncooperative hemoglobin derivatives.
The relationship between HbCO concentration and the fraction of rapid reaction, (Y, is shown in Fig. 3 for stripped hemoglobin (open circles) and for hemoglobin in the presence of PZglycerate (@filled circles). A value of Kt, 2 was calculated for each data point in Fig. 3 using the expression or*[HbCO]/(l -ar) ; average values for Kt,2 of 3.6 f 1.0 PM for stripped and 1.4 f 0.3 PM for hemoglobin with Pz-glycerate were obtained.
The agreement between the behavior observed and that predicted on the basis of the dimerization scheme suggests that the rapidly reacting hemoglobin observed subsequent to flash photolysis of stripped HbCO can be accounted for entirely by assuming it to be the ap dimer.  An approximate equilibrium constant for binding of PZ-glycerate to stripped HbCO in 0.05 M bis-tris-0.1 M NaCl, pH 7.0, can be obtained by observing the fraction of photolytically produced rapid hemoglobin as a function of Pz-glycerate concentration (Fig. 4). The concentration of Pz-glycerate required to give half the maximal decrease in rapid hemoglobin is in the range 100 to 300 PM. This relatively high concentration reveals that the interaction is weak compared to phosphate binding by deoxy Hb (28). The difference in the two curves of Fig. 4 was not investigated further but may result from the use of different hemoglobin preparations.
The effect of CO concentration on the fraction of rapidly reacting stripped hemoglobin was also tested. Increasing CO from an excess of 5-fold to nearly 30.fold had only a small effect on cr (Table II).
This experiment indicates that a first order conformational change which converts rapidly to slowly reacting hemoglobin after the photolytic flash does not become rate-limiting in the absence of phosphates in the range of CO concentrations tested.
Deoxygenation of HbOz- Gibson and Roughton (discussed in Ref. 3) noted that slow absorbance changes accompany the rapid deoxygenation of HbOz by sodium dithionite under certain carefully controlled conditions.
The magnitude of this so-called "drift phase" depends on the wavelength of observation and also on the concentration of HbOz (28,29). Kellett and Gutfreund (29) correlated the size of the drift phenomenon with the fraction of oxygenated crp dimers (determined by sedimentation equilibrium (30)) in 0.01 M Tris-0.09 M NaCI-1 mM EDTA, pH 7.0. Their results suggested that the absorbance changes following the deoxygenation process were the result of the relatively slow (ky,, = 0.43 FM-' s-l) association of the unliganded dimers produced by deoxygenation of liganded hemoglobin in dilute solution.
The effect of phosphates on the mag-

Dependence of fraction of rapidly reacting stripped hemoglobin on [CO]
Photolysis of 14.7 pM stripped HbCO in 0.05 M bis-tris-0.1 M NaCl, pH 7.0. The temperature was 21 f 2". The rate constants are the apparent first order constants estimated by least squares analysis of the average of at least four successive experiments as described in the text.  sequently, this author repeated the deoxygenation experiments using stripped HbOz in the presence and absence of Pz-glycerate. Figs. 5 and 6 show the wavelength dependence of the deoxygenation reaction of stripped HbOz by dithionite. Fig. 5 illustrates that the apparent rate of deoxygenation of stripped HbOz progressively decreases toward the end of the reaction when observed at 429 nm. In contrast, no such deceleration is found when the reaction is followed at 437 nm. This result shows the absence of dithionite-dependent artifacts within the time range of these experiments (-60 s). Fig. 6 shows that the kinetic difference spectrum of the slow phase after HbOz deoxygenation exhibim the characteristics expected of conversion of a noncooperative derivative to a cooperative form of unliganded hemoglobin (2).
When the magnitude of the absorbance change due to the slow phase is plotted as a function of the initial concentration of HbOt, the results of Fig. 7 are obtained.
In Fig. 7 the lines corresponding to stripped and Pz-glycerate-supplemented solutions of hemoglobin arc the theoretical ones based on A~gr$~"' = 12 (determined by the method of Kellett and Gutfreund (29) in 1 M NaI) and 1<t,2 = 3.6 PM for stripped and 1.4 pM for phosphate-liganded hemoglobin.
The results of the deoxygenation nitude or the kinetics of the drift phase was not reported.
Con-experiments are thus consistent with the interpretation of the 3 to 17.6 PM) for stripped and P2-glycerate-supplemented hemoglobin at 20", respectively. Note also that values of the intercepts are inversely related to the initial HbOz concentration and are decreased in magnitude by the dissociating agent NaI and increased by Pi-glycerate. DISCUSSION The liganded, or R allosteric conformation of hemoglobin (31, 32) binds CO at a rate about 80 times that of the 2' or deox) conformation.
Of the several mechanisms discussed in the introductory section of this paper in which phosphate compounds might influence the transition between the high and low affinity forms subsequent to ligancl removal, the data suggest that the primary mechanism is a stabilization of the liganded tetrameric structure relative to the dimer in the presence of phosphates.
The dependence of (Y on [HbCO] (Fig. 3) follows the relationship predicted on the assumption that dilution of the protein solution results in dissociation of slowly reacting tetramers to rapidly reacting dimers. The value of k't,Z calculated from (Y and [HbCO] Table 1 show that 1 rnM I'*-glycerate added to stripped HbCO decreased the K ,",Z to values close to those observed earlier (8) in phosphate buffers.
The relatively high concentration of I'*-glycerate required to achieve half the maximum phosphate effect (about 250 pM, Fig.  4), taken with the fact that P2-glycerate equilibrates rapidly with deoxyHb (18,33,34) also is consistent with the importance of an interaction between the phosphate compound and liganded, rather than ligand-free hemoglobin.
The association constant of Pr-glycerate and HbCO must be about 4 x lo3 M-I (assuming one binding site per tetramer? as judged from the data in Fig. 4. This value is consistent with equilibrium dialysis data of Caldwell and Nagel (35).
The relative independence of (Y with respect to [CO] eliminates the possibility that phosphates increase the rate of a first order R --t T transformation.
Although the data of Fig. 8 show that the rate of deoxydimer association is increased by Pz-glycerate, the increase is not so large that a significant fraction of dimeric deoxyHb could associate to form slowly reacting tetramers prior to CO binding.
For example in 1 rnhf Pz-glycerate, the first half-life for dimer association would be approrimately 60 ms when [HbCO] = 35 pM, whereas the half-time for CO binding to the rapidly reacting dimer was less than 10 ms.
The effect of P2-glycerate on the drift phase following dithionite-induced deoxygenation of gtripped HbOs also supports the conclusion that liganded hemoglobin is stabilized by phosphate compounds.
The reduction in the extent of the drift phase by Pz-glycerate (Figs. 7 and 8), the dependence on [HbOz], and the apparent second order kinetics all are best interpreted by as-  (16, 37) is about 5 PM (38). Thus phosphate compounds may provide a stabilization free energy of about 3.6 Cal per mole at 20" for tetrameric deosyHb compared to the dimeric species.
The kinetic basis of the Pz-glycerate stabilization of deoxyHb tetramers must primarily be an effect of the phosphate on the dissociation rate constant for dimer formation.
*issuming the validity of the above estimation of KY,? for stripped deosyHb, the dissociation rate constant ky,2 is approximately 1000 X 1OV s-l. III contrast, the dissociation rate constant in the presence of phosphates would be only 5 x 1OF 0. Thus the rate of association of tleoxy hemoglobin dimers to form tetramers is enhanced by a factor of 2.9, whereas the rate of dissociation to form dimers is retarded by about 200.fold by P2-glycerate.
Lindstrom and Ho (39), based on an anion-induced pert)urbation of the XXR spectra of IIbCO and HbOz and on the known locus of PZmglycerate binding in deosyHb (40, 41), suggest that the binding site of organic polyphosphates is between the NHz termini of the p chains as in the deosygenatecl derivative. Thus it would appear that Pz-glycerate might serve as an ionic cross-linking agent which contributes to the forces holding together a pair of c@ dimers. The cross-link would contribute about 560 cal per mole to the stability of liganded tetrameric hemoglobin compared to the dimeric state at 20".
It is not easy to see how inorganic phosphate could serve in a similar cross-linking fashion, as is suggested by its ability to depress the apparent extent of dimerization (Table I). It is perhaps possible that two phosphate anions, bound between the NHlmterminal group of each p chain and the corresponding t-ammonium group of lysine p82 (as suggested by &none for deoxyHb (40)) could interact, perhaps by means of hydrogen bonding between the -0-H of 1 phosphate and an osygen atom of the second. Alternatively, the change in tertiary structure shown in the study of Lindstrom and Ho (39) might) lead also to an increased stability of the quaternary structure in the presence of anions. This explanation seems unlikely.
The effects of Cl-and phosphate observed in the NMR experiments are apparently very similar, whereas Cl-enhances dimerization of liganded hemoglobin (8,30,36), and inorganic phosphate apparently inhibits it'. Elucidation of the inorganic phosphate effect will require further experimentation. I am further