The Binding of Folyl- and Antifolylpolyglutamates to Hemoglobin*

A binding method that detects only the strongest binding site for a ligand on a protein has been used to show that folates and folate analogs, conjugated with poly-y-glutamates, are bound to hemoglobin. When the concentration of hemoglobin is much larger than that of the polyglutamate, as is the case in the red cell, the fraction bound is a direct function of the hemoglobin concentration and is independent of the total polyglu- tamate concentration. Binding to deoxyhemoglobin tetramers is competitive with 2,3-diphosphoglycerate. In oxyhemoglobin the folyl and methotrexate polyglutamates are bound preferentially by free a@ dimers, but removal of the pteridine moiety leads to tetramer binding even in oxyhemoglobin. Changes in the length of the polyglutamate side chain and alterations of the pteridine structure such as reduction and/or methyla- tion have a much larger effect on the constant for binding to deoxyhemoglobin tetramers than on that for oxyhemoglobin dimers. The implications of these results for the storage of pteroylpolyglutamates in the erythrocyte and their release from the red cell under the influence of the degree of oxygenation and varia- tions in the 2,3-diphosphoglycerate pH 7.5. The reaction was started by the addition of 1.7 units of phosphoglycerate mutase (Sigma). The increase in absorbance at 240 nm was linear for the first 3 min and proportional to the DPG concentration from 2 to 8 X lo-' M. Binding Measurements-All binding studies were carried out by ultrafiltration as previously described (Benesch et al., 1983) at room temperature with varying hemoglobin concentrations. 0.05 M bistris buffer, pH 7.3, was used throughout, except for the DPG binding experiments, which were conducted in 0.05 M Tris at the same pH, since Tris has no significant absorbance at 240 nm. Hemoglobin-Oxygen Equilibria-The effect of PteGlu, on the oxygen equilibrium curve of hemoglobin was measured at 20 "C in 0.05 M bistris buffer, pH 7.3, 0.1 M Cl-, 0.1 mM EDTA, using previously described methods (Benesch et al., 1965,1973). The concentration of the polyglutamate was kept constant at 2.5 X lo-, M and that of the hemoglobin was varied from 1.6 X 1O"j M to 5 X M. For the lowest hemoglobin concentration the absorbances at 415, 420, and 430 nm were corrected for the absorbance of PteGlu,.

A binding method that detects only the strongest binding site for a ligand on a protein has been used to show that folates and folate analogs, conjugated with poly-y-glutamates, are bound to hemoglobin. When the concentration of hemoglobin is much larger than that of the polyglutamate, as is the case in the red cell, the fraction bound is a direct function of the hemoglobin concentration and is independent of the total polyglutamate concentration. Binding to deoxyhemoglobin tetramers is competitive with 2,3-diphosphoglycerate. In oxyhemoglobin the folyl and methotrexate polyglutamates are bound preferentially by free a@ dimers, but removal of the pteridine moiety leads to tetramer binding even in oxyhemoglobin. Changes in the length of the polyglutamate side chain and alterations of the pteridine structure such as reduction and/or methylation have a much larger effect on the constant for binding to deoxyhemoglobin tetramers than on that for oxyhemoglobin dimers. The implications of these results for the storage of pteroylpolyglutamates in the erythrocyte and their release from the red cell under the influence of the degree of oxygenation and variations in the 2,3-diphosphoglycerate level are discussed.
Folic acid as well as analogs such as aminopterin or methotrexate, used in cancer chemotherapy, are conjugated in vivo with glutamic acid to form poly-y-glutamates of varying chain length (Baugh et d., 1976;Galivan, 1979;Covey, 1980;Kisliuk, 1981;McGuire and Bertino, 1981;Jolivet et al, 1982). These highly charged compounds are, of course, bound electrostatically in a nonspecific fashion to many proteins (Zamierowski and Wagner, 1974;Waxman et aL, 1977;Holm et al., 1980;Wagner, 1982) including hemoglobin (Hansen et aZ., 1981). However, we have shown previously (Benesch et aL, 1983) that a more specific binding of pteroylheptaglutamate (PteG1u71) to hemoglobin can be demonstrated under conditions where the concentration of the folate is of the same order as that in the red cell, i.e. about 1 PM, and hemoglobin is in large excess. In this way nonspecific binding is eliminated and only the strongest binding site is occupied. The binding constant is then equal to the hemoglobin concentration at which half the total folate is bound. In the case of deoxy-Hb it was found that 1 mol of PteG1u.l is bound per tetramer over a 10-fold range of hemoglobin concentration. This was not so with oxy-Hb where it was shown that free a@ dimers instead of tetramers are responsible for the binding of PteGlu7 (Benesch et d., 1983).
In this paper a variety of folyl-and antifolylpolyglutamates has been investigated to determine the effect of structural alterations on the binding energy. These include variations in the length of the polyglutamate side chain, as well as reduction, substitution, and removal of the pteridine portion of the molecule. It will also be shown that 2,3-&phosphoglycerate, the main intraerythrocytic regulator of oxygen transport, inhibits polyglutamate binding to deoxy-Hb competitively. The conclusions from these experiments agree well with the binding site for PteGlu7 on deoxy-Hb that has been identified crystallographically, The methods used will illustrate a general strategy for treating the binding of a small molecule in very low concentration to a protein present in large excess. This is, of course, particularly relevant for those biological situations where proteins occur in high molar concentration, such as hemoglobin in the red cell (-5 mM) or serum albumin in plasma (-0.5 mM).

MATERIALS AND METHODS
Hemolysates of normal human blood were prepared as described previously (Benesch et al., 1972). The major hemoglobin component (HbA) was isolated by chromatography on DEAE-Sepharose-Fast Flow (Pharmacia) at 4 "C using a linear gradient of 0.05 M Tris buffer, pH 8.3-7.3. Hb Kansas (B'MA"Th') was isolated essentially as described by Bonaventura and Riggs (1968) except that the gradient was flattened by using 4 liters instead of 1 liter of each buffer. After 40 h a good separation between Hb Kansas and HbA was obtained. Hemolysates and purified hemoglobin were stored in liquid Nz. Hemoglobin concentration was measured spectrophotometrically at 540 nm after conversion to methemoglobin cyanide ( 6 = 4.4 X lo4 M" cm").
Assay Methods-The very low concentrations of the polyglutamate conjugates which were essential for these measurements required sensitive assays, accurate to about 0.01 nmol.
All the 14C-labeled compounds were assayed in 5 ml of Aquasol-2 in a Packard Tri-Carb Liquid Scintillation Counter.
p-AB-Glu7 was determined fluorimetrically after reaction with fluorescarnine (Furness and Loewen, 1981). 0.2 ml of the sample was mixed with 0.1 ml of freshly prepared fluorescamine (10 mg/50 ml of acetone). After 5 min at room temperature, 1.5 ml of HzO was added and the fluorescence was read at 500 nm using an excitation wavelength of 395 nm.
The assay of methotrexate and MTX-GlQ was based on the stoichiometric inhibition of dihydrofolate reductase as described by Peterson et al. (1975). Various amounts of methotrexate were incubated at 37 "C for 5 min with 0.01 unit/2 ml of bovine liver dihydrofolate reductase (Sigma) and 100 p~ NADPH (Sigma) in 0.6 M KC1 and 0.6 M sodium acetate buffer, pH 6.0. The enzyme reaction was started by the addition of 40 pl of 3.3 mM dihydrofolate (Sigma)/2 ml. The decrease in the initial reaction rate was linear with methotrexate or MTX-Glu, concentration from 5 to 25 pmol/2 ml. DPG concentrations >lo4 M were determined by phosphate analysis as described by Ames and Dubin (1960). For the low concentration range (down to 2 X M) the assay was based on the catalytic effect of DPG on the reaction: 2-phosphoglycerate $3-phosphoglycerate, using a modification of the method of Towne et al. (1957). The reaction mixtures at 30 "C contained the sample, 40 units of enolase (Sigma), 1.6 pmol of 3-phosphoglyceric acid (Sigma) and 20 pmol of MgSO, in 3 ml of 0.07 M Tris buffer, pH 7.5. The reaction was started by the addition of 1.7 units of phosphoglycerate mutase (Sigma). The increase in absorbance at 240 nm was linear for the first 3 min and proportional to the DPG concentration from 2 to 8 X lo-' M.
Binding Measurements-All binding studies were carried out by ultrafiltration as previously described (Benesch et al., 1983) at room temperature with varying hemoglobin concentrations. 0.05 M bistris buffer, pH 7.3, was used throughout, except for the DPG binding experiments, which were conducted in 0.05 M Tris at the same pH, since Tris has no significant absorbance at 240 nm.
Hemoglobin-Oxygen Equilibria-The effect of PteGlu, on the oxygen equilibrium curve of hemoglobin was measured at 20 "C in 0.05 M bistris buffer, pH 7.3, 0.1 M Cl-, 0.1 mM EDTA, using previously described methods (Benesch et al., 1965(Benesch et al., ,1973. The concentration of the polyglutamate was kept constant at 2.5 X lo-, M and that of the hemoglobin was varied from 1.6 X 1O"j M to 5 X M. For the lowest hemoglobin concentration the absorbances at 415, 420, and 430 nm were corrected for the absorbance of PteGlu,.

RESULTS
The binding of small molecules to proteins is usually carried out by titrating the protein with increasing amounts of the ligand and measuring the moles bound per mole of protein as a function of the free ligand concentration. Binding constants for a site or group of sites can then be obtained, provided the moles bound per mole of protein approach a plateau value (Klotz, 1982). We found earlier that with an excess of pteroylpolyglutamates over hemoglobin no plateau is reached (Benesch et al., 1982). For this reason (as well as for the sake of biological reality) we have reversed the usual concentration ratio of protein to ligand, using a large excess of hemoglobin. Under these conditions only the tightest binding site will be occupied and its binding constant is given by: -PteGlu,] which reduces to since [Hb-PteGlu,] is negligible compared to the total hemoglobin concentration. This formulation was first used successfully to demonstrate specific binding of PteGlu7 to deoxy-Hb tetramers and oxy-Hb dimers (Benesch et al., 1983). The validity of Equation 2 is confirmed by two predictions: 1) The ratio of bound to free ligand should be a function only of the hemoglobin concen-tration and independent of the total ligand concentration when hemoglobin is in large excess. This is borne out by the data in Table I. 2) If a ligand binds with a 1:l stoichiometry even when it is present in excess over the protein, as DPG does to deoxy-Hb (Benesch et al., 1968), then the two ways of defining and measuring the binding constant should be equivalent, i.e. the molar hemoglobin concentration at which half the total DPG is bound should be numerically equal to the free DPG concentration at which 0.5 mol is bound per mole of hemoglobin. The binding curve of DPG to deoxy-Hb was therefore measured (a) using M hemoglobin and varying the DPG concentrations from 3 X to 2 X M and (b) at a constant DPG concentration of M and hemoglobin concentrations from 7 X lop6 to 1.2 X low4 M. Fig. 1 shows that the two curves are indeed identical.

The Affinity of Different Folylpolyglutamates for Hemoglobin
Binding to Deoxyhemoglobin-The effect of chain length (and therefore the number of negative charges) illustrates the contribution of electrostatic interactions to the binding energy (Table 11, column 1). It is clear that the strength of binding to deoxy-Hb increases with the number of glutamates and that the effect of two additional ones is about the same, i.e. 2-fold, both for the pteroyl and the 5-methyltetrahydropteroyl compounds.     The binding curves of four different heptaglutamates are shown in the form of Hill plots in Fig. 2. It is noteworthy that they are all parallel with a slope of one, as required by Equation 2. Surprisingly, the introduction of a methyl group at N5 and especially at Nlo weakens the binding more than the removal of the entire pteridine moiety. The 4-amino group of methotrexate is unprotonated at neutral pH (Poe, 1977) and is therefore unlikely to contribute significantly to the weaker binding of this compound.
The effect of DPG on polyglutamate binding by deoxy-Hb was investigated with molar ratios of DPG to hemoglobin close to unity as is the case in the red cell. As before, the concentration of polyglutamates was 1 X M, so that both DPG and hemoglobin were in great excess. From the data in Table 111 it is clear that the fraction of polyglutamate bound is decreased progressively by increasing the DPG concentration at each hemoglobin concentration. Furthermore, if DPG and the polyglutamates compete for the same site on the protein, then only those hemoglobin molecules which are not complexed with DPG should be "active" in binding polyglutamates. The calculated values in Table I11 al. (1967), who showed that glycogen cannot bind to the tetramer form of phosphorylase a and that the dimer-glycogen complex is therefore the active form of this enzyme.
Since the tetramer/dimer dissociation constant of oxy-Hb is only about loT6 M (Edelstein et al., 1970;Chu and Ackers, 1980), the concentration of dimers in oxy-Hb is much less than that of tetramers in deoxy-Hb at the protein concentrations used for this work. Therefore, the fraction of polyglutamate bound is always less for oxy-Hb than for deoxy-Hb at the same hemoglobin concentration and, conversely, the hemoglobin concentration for equal binding is much larger for oxy-than for deoxy-Hb (Table 11). It is also evident that the effect of the polyglutamate chain length on the binding constant is noticeably smaller in the case of the oxy-Hb dimers ( Table 11). The selective affinity for oxy-Hb dimers was observed with all the polyglutamates included in this study with one notable exception. This is shown in Fig. 3 where the Hill plots based on Equation 3 are linear with a slope of one except for the paminobenzoylheptaglutamate. This compound evidently binds to tetramers even in oxy-%, since the binding curve fits Equation 2 rather than 3 (Fig. 3).
Just as is the case for the glutamate chain length, methylation of the pteridine ring and its state of oxidation has much less effect on the affinity for oxy-Hb dimers than for deoxy-Hb tetramers. The numerical values of the binding constants of the four heptaglutamates to oxy-and deoxy-Hb are listed in Table IV.
The binding of folylpolyglutamate to oxy-Hb dimers is borne out by two other lines of evidence.  low oxygen affinity and a 50-100-fold increase in the dissociation of the liganded form into cup dimers (Atha and Riggs, 1976). Fig. 4 shows that COHb Kansas does indeed bind more PteGlu-, than HbA over the whole range of hemoglobin concentrations. The binding curve of the mutant hemoglobin fits Equation 3 very well (Fig. 4, inset) using the published value for the tetramer-dimer dissociation constant (K4,2 = 5.4 X M). The resulting binding constant of Hb Kansas dimers for PteGlu7 is 6.6 X lov5 M compared to 2.5 X M for HbA. Evidently the substitution which loosens the contact between the cup dimers also reduces the affinity of PteGlu-, for these dimers.

TABLE IV Binding constants of heptaglutamates to hemoglobin
(b) Another consequence of folylpolyglutamate binding to oxy-Hb dimers is that the allosteric shift in the oxygen affinity should vary with the hemoglobin concentration. In the case of PteGlu., the affinities for deoxy-Hb tetramers and oxy-Hb dimers happen to be equal (Table IV). Therefore &0 should be unaffected by this polyglutamate at a hemoglobin concentration at which the molar concentrations of oxy-Hb dimers and deoxy-Hb tetramers are equal, i.e. 7.5 X loF7 M or 0.005 gfl00 ml hemoglobin. Above this concentration the P5o"in the presence of a constant concentration of PteGlu7-should increase progressively with increasing hemoglobin concentration since the proportion of oxy-Hb dimers decreases with the total hemoglobin concentration. The results in Fig. 5 confirm the expectation that, in contrast to other compounds like DPG (Benesch et aL, 1969), the allosteric effect of PteGlu, increases with increasing hemoglobin concentration.

DISCUSSION
The method developed here is unusual because binding is measured with a large excess of protein over ligand. The binding constant is then defined as the protein concentration for half-saturation of the small molecule. This has several notable advantages: 1) It permits the recognition of the tightest binding site even in the case of weakly bound ligands. 2) In a dissociating protein system the affinity of the ligand for either the monomer or the polymer can be measured easily without disturbing the equilibrium.
3) The simplicity of the equation under these conditions avoids the ambiguities associated with measurements of multiple binding sites in the presence of excess ligand (Klotz, 1982). Therefore the association constants for the tightest binding site are easily determined.
As far as the location of this site on deoxy-Hb is concerned, the demonstration that DPG competes with the polyglutamates suggests that these compounds are bound at the same site, i.e. at the entrance to the central cavity between the , ! 3 chains (Arnone, 1972). Crystallographic evidence obtained by Arthur Arnone2 has shown that the pteroyl group of PteGlu? is buried deep inside the central cavity of deoxy-Hb where it interacts with phenylalanine (~3 6 .
The glutamate residues can be seen at the entrance to the central cavity near the basic residues of the DPG binding site, although, of course, they give rise to a much weaker positive density in the difference map. This evidence accounts very well for the binding of all the pteroylpolyglutamates to the tetramer in deoxy-Hb but not in oxy-Hb where the entrance to the central cavity is considerably smaller (Perutz, 1965). It also explains why the p-aminobenzoylheptaglutamate, which lacks the bulky pteridine residue, is the only one of these compounds which is able to penetrate the oxy-Hb tetramer (Fig. 3).
The oxy-Hb dimer binding site for pteroylpolyglutamates is obviously inaccessible in the undissociated tetramer. It is therefore likely that it involves the same region by which these dimers are linked within the tetramer. This is reminiscent of the interaction between oxy-Hb dimers and haptoglobin (Benesch et al., 1976;Wejman et al., 1984). Other ligands have been shown to have an affinity for oxy-Hb dimers, for example Zn2+ (Gray, 1980;Gray and Dean, 1982) and inositol hexaphosphate (Hensley et ul., 1975;White, 1976) and even for deoxy-Hb dimers (Weiderman and Olson, 1975).
Erythrocytes contain relatively high concentrations of folylpoly-y-glutamates but their function there is not known. The major forms are 5-CH3H4 folates with 4-7 glutamate residues (Shin et Perry and Chanarin, 1977). The chemotherapeutic folate analog, methotrexate, is similarly conjugated with glutamic acid, and accumulates in the red cell in high concentrations (daCosta and Iqbal, 1981;Kamen et al., 1981Kamen et al., , 1984. In view of the results reported here, hemoglobin binding probably plays an important role in storing folates and antifolates in the red cell. By analogy with results on other cells (Rosenblatt et al., 1978;Balinska et al., 1981;Galivan and Balinska, 1983;Poser and Sirotnak, 1983), free pteroylpolyglutamates are likely to pass through red cell membranes, albeit more slowly than the monoglutamate forms. The rate of efflux would, of course, only be proportional to the concentration of polyglutamates not bound to hemoglobin, and it is therefore very relevant that Cooper and Peyman (1982) found a significant efflux of intact pteroylpentaglutamate from red cells with a low hemoglobin concentration. Since we have also shown that the proportion bound is directly influenced by DPG concentration (Table 111) and oxygen pressure (Table   IV), these two variables would participate in regulating the rate of efflux from the erythrocyte.
Using the dissociation constants in Table IV, the fraction of free 5-CH3H4PteGlu7 or MTX-Glu6 can be calculated at red cell hemoglobin concentrations and various oxygen pressures and DPG levels. Thus, at a red cell hemoglobin concentration of about 5 mM and a normal venous pOz of 40 mm where hemoglobin is about 75% oxygenated and with normal DPG levels of about 0.8 mol of DPG/mol of hemoglobin, only 27% of the total 5-CH3H4-PteGlu7 apd 17% of the MTX-Glu6 would not be bound to hemoglobin. Under anoxic conditions, however, where the molar DPG/Hb ratio rises well above 1, much greater amounts of the polyglutamates will be released.
In conclusion, it seems probable that the binding of folates and antifolates may be just one example of a hitherto ne- glected role of hemoglobin. Other biologically important molecules such as vitamins, coenzymes, and peptides, as well as some drugs (Perutz and Poyart, 19831, which occur in the red cell in minute molar concentrations compared to that of hemoglobin might be similarly bound. Thus, hemoglobin could act as a storage reservoir from which these compounds are released under the influence of such regulatory factors as oxygen and DPG. In the erythrocyte the enormous hemoglobin concentration (about 30 g/lOO ml) would lead to significant binding even of substances which have a relatively low affinity for the protein.