Kinetics and Activation Parameters of the Reaction of Cyanide with Free Aquocobalamin and Aquocobalamin Bound to a Haptocorrin from Chicken Serum*

The kinetics of the reaction of cyanide with free aquocobalamin (HzOCbl) and with aquocobalamin bound to a vitamin Blz-binding protein (haptocorrin) from chicken serum (HC-HzOCbl) have been investi-gated as a function of temperature and pH. The mech- anism of replacement of HzO from protein-bound and free HzOCbl is apparently the same and involves attack of both CN- (kl) and HCN (kz). The reactions with HC-HzOCbl are somewhat slower than those with free HZOCbl, kl being 22-fold smaller and kz 7-fold smaller at 25 “C. The relatively small effect of the protein on the rate constants supports the view that the metal in HC-HzOCbl is readily accessible to solvent. Activation parameters suggest that the transition states for ligand substitution are stabilized by nucleophilic participation (at least by CN-) and that ligand substitution on the protein-bound cobalamin proceeds through more ordered, concerted transition states. The latter effect suggests that the Co-0 bond of HZOCbl is strengthened upon binding to the haptocorrin. of 1.0 ml. Cyanide solutions were prepared fresh daily and their pH adjusted to match that of the reaction solutions. pH was measured using a Radiometer PHM 84 pH meter and a Radiometer Type B combined glass electrode with electrode, samples, standards, and rinse water incubated at the measurement temperature. Kinetic traces were monitored for at least five half-times and analyzed by plotting ln(A, - A,) uersus t using standard linear regression analysis to obtain a pseudo-first-order rate constant, kp". The relevant second-order rate constants, kf?, were obtained from the slope of plots of kYb" versus cyanide concentration in which the latter varied by a factor of 20-200, also by linear regression analysis. A similar technique was used to study the reaction of free HZOCbl (-3.2 X M) with cyanide, but reactions were monitored at 560 nm, the isosbestic point for conversion of CNCbl to dicyanocobalamin (40). The pK, for proton dissociation from the aquo ligand of HC- HzOCbl was determined spectrophotometrically at 540 nm by adjust-ing the pH of a continuously stirred 4.0 ml sample of 3.1 X lo-' M HC-HzOCbl in 50 mM buffer (equal concentrations of MES, bicine, and CAPS) and KC1 (to ionic strength 1.0 M), immersed in a ther- mostatted bath at the appropriate temperature, with 4 M NaOH. A typical titration from pH 5.3 to 11.3 entailed addition of 35.8 p1 of base (a volume change of <1%) and caused a variation in ionic strength from 0.97 to 1.01 M. An analogous procedure was used for free H,OCbl, but the absorbance change was monitored at 351.5 nm.

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. duced by a conformational change in the protein upon binding substrate, few specific hypotheses of how such steric activation might arise have been advanced. One such hypothesis, the "mechanochemical trigger" mechanism is quite attractive on several grounds. It envisages an enzyme-mediated compression of the long (2.24 A) axial Co-N(Bz) bond (19) leading directly to homolytic fission of the Co-C (Ado) bond. This would provide rationalizations for the presence of a pendent axial ligand in this unique pentadentate chelate as well as its steric bulk, and for the relatively tight binding of AdoCbl to most AdoCbl-requiring enzymes. Model systems also exist for the labilization of sterically strained alkylcobalamins upon decreasing the distance between the benzimidazole of the nucleotide loop and the cobalt ion (20)(21)(22)(23)(24)(25)(26)(27)(28). For example the slow homolysis of the base-off form of neopentylcobalamin in acid solution under aerobic conditions (tlh = 28 days) is increased some 500-fold when the reaction is conducted at neutral pH, even though the benzimidazole nucleotide appears to be only about 30% coordinated (26).
Derivatives of vitamin BIZ are also very tightly bound by a wide range of proteins including the gastric intrinsic factors, the serum transcobalamins, and the haptocorrins (also known as R-binders or cobalophilins) obtained from diverse sources (see, for example, Refs. 29 and 30). A variety of cobalamins with different fi axial ligands (CN-, HzO, OH-, C1-, CH3, and Ado) have similar affinities for such proteins (31-34). However, since alteration of the nucleotide loop significantly decreases this affinity (35)(36)(37) it is generally believed that the cobalamins bind to such proteins with the fi face pointing toward the solvent. Although the binding of many Blz derivatives to such proteins has been studied, little work has been done to determine the effects of the protein moiety on the chemical and physical properties of bound cobalamins.
We are currently involved in studies of the chemistry of cobalamins bound to such B12-binding proteins as models for the interaction of AdoCbl with AdoCbl-requiring enzymes. Recently using a 70-kDa glycoprotein haptocorrin from chicken serum, we have found that the 31P NMR resonances of protein-bound cobalamins are shifted some 0.7-1.0 ppm downfield from the positions of the free cobalamin resonances (38). This is apparently due to changes in phosphodiester conformation which may involve steric compression of the axial Co-N(Bz) bond (38).
In the current work we have attempted to use cyanide as a probe of the chemistry of protein-bound cobalamins. Cyanide has been shown to be a useful probe ligand for the metal center of metalloproteins. For instance, the kinetics of the reaction of cyanide with microperoxidase-8 (MP-8), a heme octapeptide obtained by proteolysis of cytochrome c which retains His-18 as proximal (axial) ligand, have recently been studied (39). Comparison of the results with the kinetics of the reaction of cyanide with hemoproteins containing a single His ligand, such as myoglobin, hemoglobin, various peroxidases, and cytochrome c oxidase, has shown how the protein moiety controls and modifies the ligand binding properties of the metallo prosthetic group (39). Since the kinetics and mechanism of the reaction of free HzOCbl with cyanide in aqueous solution at 25 "C have been reported (40), we have extended this approach to the complex of H20Cbl with the chicken serum haptocorrin (HC-H2OCb1) and have determined the kinetics of its reaction with cyanide as a function of pH and temperature. The results of these investigations are the subject of this paper.
The kinetics of cyanide addition to HC-HzOCbl to produce HC-CNCbl were followed by the absorbance change at 364 nm (close to the ?-band of the product (38)) using 1.0-cm pathlength semi-micro quartz cuvettes on a Cary 219 recording spectrophotometer fitted with a thermostatted cell block. The temperature was monitored in the cuvettes with a thermistor device (Yellowsprings Instruments) calibrated against NBS calibrated thermometers. Pseudo-first-order conditions were maintained with HC-H20Cbl -1.2 X lo6 M and total cyanide at least 10-fold that concentration. pH was maintained with 0.1 M appropriate buffer (acetate, phosphate, bicine, or CAPS), and ionic strength was maintained at 1.0 M with KCl. Reactions were initiated by addition of 100 pl of temperature-equilibrated HC-HzOCbl stock solution to temperature-equilibrated cuvettes containing buffer, KCl, and cyanide for a final volume of 1.0 ml. Cyanide solutions were prepared fresh daily and their pH adjusted to match that of the reaction solutions. pH was measured using a Radiometer PHM 84 pH meter and a Radiometer Type B combined glass electrode with electrode, samples, standards, and rinse water incubated at the measurement temperature.
Kinetic traces were monitored for at least five half-times and analyzed by plotting ln(A, -A,) uersus t using standard linear regression analysis to obtain a pseudo-first-order rate constant, kp". The relevant second-order rate constants, kf?, were obtained from the slope of plots of kYb" versus cyanide concentration in which the latter varied by a factor of 20-200, also by linear regression analysis. A similar technique was used to study the reaction of free HZOCbl (-3.2 X M) with cyanide, but reactions were monitored at 560 nm, the isosbestic point for conversion of CNCbl to dicyanocobalamin (40).
The pK, for proton dissociation from the aquo ligand of HC-HzOCbl was determined spectrophotometrically at 540 nm by adjusting the pH of a continuously stirred 4.0 ml sample of 3.1 X lo-' M HC-HzOCbl in 50 mM buffer (equal concentrations of MES, bicine, and CAPS) and KC1 (to ionic strength 1.0 M), immersed in a thermostatted bath at the appropriate temperature, with 4 M NaOH. A typical titration from pH 5.3 to 11.3 entailed addition of 35.8 p1 of base (a volume change of <1%) and caused a variation in ionic strength from 0.97 to 1.01 M. An analogous procedure was used for free H,OCbl, but the absorbance change was monitored at 351.5 nm.

RESULTS
The variation with temperature in the pKa of HCN, HzOCbl, and HC-HZOCbl, and the enthalpy and entropy of ionization, obtained from plots of In K. versus 1/T (not shown), are listed in Table I. The spectral changes associated with the pKa values of H20Cbl and HC-H20Cbl were found to be completely reversible in the pH range 5.3 to 11.5. Consequently, the haptocorrin appears to be stable in alkaline solution at least up to pH 11.5.
Two determinations of k?p at 25 "C, for the reaction of free

HCN, HZOCbl, and HC-H20Cbl
Ionic strength 1.0 M (KCl). The reaction was found to proceed too rapidly at 45 "C for reliable determination of second-order rate constants by conventional spectroscopy. Therefore, the variation of k?? with pH for the reaction of free H20Cbl with cyanide was studied at 5" and 15 "C. The results are shown in Fig. 1. In agreement with a previous report (40), no evidence of saturation kinetics was found at high cyanide concentration (up to 0.5 M at pH 3.99, 5 "C).

AH
The mechanism of reaction of cyanide with H20Cbl is well established (40  iterative, nonlinear least squares program using either a simplex minimization algorithm or a Newton-Raphson procedure using Marquardt's algorithm. The results, which were nearly identical using either minimization routine, are summarized in Table 11. Alternatively (40), the second-order rate constants may be expressed in terms of addition of CN-to H20Cbl by dividing kyp by the fraction of cyanide present as CNand the fraction of cobalamin present as the aquo species (Equation 12). Data    (12) reactant are shown in Fig. 2 and governed by the rate law of Equation 13, which converges to k1 at high pH and to kzcrK,/ KHCN at low pH. This is an instructive way to view the data in that the unusual pH rate profiles of Fig. 2 clearly show a pHindependent region at high pH (representing rate-limiting addition of CN-to H20Cbl), an acid-catalyzed region at intermediate pH (representing rate-limiting addition of HCN to form the N-bound species), and a pH-independent region at low pH (due to rate-limiting isomerization of the N-bound species to the C-bound species, Equation 8).
Values of kyp for reaction of HC-H20Cbl with cyanide were determined over the pH range 3.5 to 11.2 at 5,25, and 45 "C.
As with free H20Cbl, no evidence was found for saturation kinetics up to cyanide concentrations as high as 0.5 M. The  Tables I and 11. pH rate profiles obtained are shown in Fig. 3 and the values of the kinetic parameters obtained by curve-fitting to Equation 11 are given in Table 11. pH rate profiles for kc,, (Equations 12 and 13), i.e. corrected for ionization of the starting materials, are shown in Fig. 4.

TABLE I11 Activation parameters for addition of C N -and HCN to HZOCbl and HC-HzOCbl kl(CN-) k2(HCN) Complex
Activation parameters for addition of CN-(Le. kl, Equation  1) and HCN (i.e. kp, Equation 6) to free H,OCbl and HC-HzOCbl were determined from the temperature-dependence of k, and kz via plots of In(kh/kBT) versus 1/T (not shown) where h and kB are Planck's constant and Boltzman's constant, respectively. The values obtained are listed in Table   111.

DISCUSSION
The enthalpy and entropy of ionization of the aquo ligand of free H20Cbl has not been reported previously. As can be seen from the temperature dependence of the equilibrium constant for the self-ionization of water ( A H = 13.3 kcal mol", A S = -19.4 cal mol" OK-' , calculated from the data in Ref. 45), ionization of water entails a positive enthalpy change but a negative entropy change. The latter value suggests that the products of ionization cause substantial ordering of solvent as would be expected. Coordination to the metal center causes a substantial reduction in the enthalpy of ionization and a smaller reduction in the magnitude of the entropy change. Both effects can be reasonably attributed to stabilization and charge delocalization of the conjugate base (OH-) by coordination.
On binding of HzOCbl to the haptocorrin, the pK, is raised from 8.10 to 8.29 at 25 "C. This is a consequence of somewhat higher enthalpy of ionization; the entropies of ionization of free and protein-bound H20Cbl are not significantly different. Binding of ferric porphyrins to various proteins can increase the pK. of Fe(II1)-coordinated water, decrease it, or leave it relatively unchanged. For instance, the pK, of aquo-MP-8 is 8.90 (46) while the pK, values of human metHb (47), horse metMb (48), and sperm whale metMb (47) are 8.05,8.93, and  8.99, respectively. Whether these differences are due to enthalpic or entropic factors is not yet known, and there appears to be at present no means of predicting the effect of the protein on the ionization of a metalloaquo center. However, in the case of HC-H20Cbl the small effect of the protein on the pK, implies that the microenvironment of the coordinated water is not significantly different from that of free HzOCbl in bulk water, i.e. the upper axial ligand position of the cobalamin is not buried in a protein region of significantly different dielectric than solvent.
The temperature dependence of the pK, of HCN has been reported (49) but at a low ionic strength ((0.01 M) from which the values AH = 11.4 kcal mol" and A S = -6.7 cal mol" OK-' can be calculated with pK. = 9.21 at 25 "C. Our results show that HCN is a stronger acid at higher ionic strength as would be anticipated for a neutral Bronsted acid.
The good agreement obtained between the theoretical curves and the experimental data in Figs. 3 and 4, as well as the similarity of the reactant ionization-corrected pH rate Cyanide Reaction Kinetics profiles for free HzOCbl (Fig. 2) and the protein-bound species (Fig. 4) suggest that HC-HzOCbl and free HzOCbl react with cyanide by the same mechanism. However, at 25 "C, the reaction of CN-is 10.8 times slower and that of HCN 2.3 times slower with HC-HzOCbl than with free H20Cbl. When compared with the effect which some hemoproteins have on such reactions, this effect is modest. The second-order rate constants for reaction of CN-and HCN with ferric MP-8 are 6.0 X lo6 M-' s" and 4.8 X lo3 M-' s-', respectively (39). In contrast the rate of reaction of metMb with HCN is insignificant compared with its rate of reaction with CN-(50), and the rate constants for reaction of CN-with metMb and metHb are of the order of 2-5 X 10' M" s-' (51, 52), i.e. the protein slows down the reaction by about 20,000-fold. This has been ascribed both to poor ligand accessibility to the metal and to steric crowding of the heme distal side (39). We may conclude, in agreement with previous suggestions about cobalaminbinding proteins (35, 36), that the cobalamin-binding site of the haptocorrin from chicken serum leaves the cobalamin relatively open to the solvent. This, of course, explains the success of affinity chromatography procedures in which the cobalamin moiety is immobilized by attachment to the upper axial ligand position (38).
The present results, together with those of Reenstra and Jencks (40) show that both CN-and HCN (presumably reacting through N) act as nucleophiles toward the Co(II1) center of HzOCbl. A similar effect has been shown with the ferric heme center of MP-8 (39). Although we were unable to observe the postulated N-bound intermediate cyanide complex, precedents do exist, as has been pointed out (40). We may further add that Betterton (53) observed that when HCN reacts with diaquocobinamide, isosbestic points are not immediately established but shift with time in a subsequent irreversible reaction, indicating formation of an intermediate. In addition the transient formation of what is probably an Nbound cyanide species on reaction of hemin with cyanide has been observed spectroscopically (54, 55). There would, therefore, appear to be little doubt about the validity of the proposed reaction scheme.
It is generally agreed that the mechanism of ligand substitution reactions of six-coordinate cob(II1)alamins in aqueous solution is a predominately dissociative (i.e. Id) process (40, 56, 57). Hence, the rate of substitution of Hz0 in HzOCbl by small anions such as I-, SCN-, and CN-is relatively insensitive to the nature of the incoming ligand (58). However, substitution rates have been found to vary by nearly two orders of magnitude for a larger series of ligands in which equilibrium constants vary by 11 orders of magnitude (59). This observation prompted Reenstra and Jencks (40) to suggest that there is some stabilization of the transition state by the incoming ligand (i.e. nucleophilic participation) at least for some ligands. On the other hand some of the variability in ligand substitution rates may be due to other factors. If the incoming ligand is capable of hydrogen bonding to the acetamide side chains directed upwards toward the p face of cobalamin, rate constants can decrease significantly. For example, the second-order rate constant for substitution of Hz0 by SCN-is 7.1 x lo3 M" s-' (58), but 21.5 and 1.05 M-' s-' for NH20H and CH3NHz, respectively (60). It thus appears that factors such as ligand charge, and steric, hydrophobic-, and hydrogen-bonding effects between the ligand and the corrin ring side chains play a role in the variability of ligand substitution rates.
Nonetheless, our results on the activation parameters for substitution of HzOCbl by CN-and HCN provide support for the importance of nucleophilic participation in the ligand substitution transition state. The transition state for substitution by CN-is enthalpically stabilized by 6 kcal relative to that for HCN substitution (Table 111). As CN-must be expected to be a far stronger nucleophile than HCN (CN-is more basic toward the proton than HCN by over 11 orders of magnitude (61)), this observation provides strong support for nucleophilic stabilization. Entropies of activation, however, are much more difficult to interpret due to the importance of differential solvation of ground and transition states. The positive entropy of activation for HCN substitution suggests that the entropy gain due to loss of the leaving ligand is largely uncompensated by entropy loss due to bonding of the incoming ligand. This is consistent with a strictly Id mechanism with little or no nucleophilic participation and a symmetric transition state. However, the value of AS* for HCN substitution is very similar to that expected for a simple gas phase dissociation reaction (62) suggesting that there is little change in solvation of the departing ligand in the transition state. This makes it difficult to rationalize the negative entropy of activation for CN-substitution, since the entropy loss due to participation of the incoming-ligand in the transition state cannot be expected to exceed the entropy gain due to dissociation of the leaving ligand. Nonetheless, the substantial decrease in ASf for CN-substitution relative to HCN substitution is consistent with a very much more ordered transition state in which significant nucleophilic participation has made the mechanism somewhat more concerted (40).
On binding to the protein, the reaction rates change considerably (Table 11) and at 25 "C CN-and HCN react with virtually the same rate constants with HC-HzOCbl. The activation parameters show that this is due to (i) an increase in AHz and an even more unfavorable ASf for replacement of HzO by CN-and (ii) a slightly smaller AHH+ but significantly less favorable AS# for replacement of HzO by HCN. The transition states for replacement of HzO by CN-and HCN are thus both more ordered in protein-bound than in free HzOCbl. This suggests that there is less Co-0 bond breaking in both transition states and that both displacement reactions have become somewhat more concerted (40). This would be the anticipated result if binding of HzOCbl to the protein leads to a strengthening of the Co-0 bond. Excellent structural evidence has been obtained for a mutual dependence of Co-C and trans Co-L bonds in organocobaloximes (63) and related cobalamin models (64). Thus, shortening of the axial Co-N bond leads to a strengthening of the trans Co-C bond in such organocobalt species (63). A similar effect has recently been shown to operate in organocobalamins (65). An analogous effect may be responsible for the decreased reaction rate of HC-H20Cbl with cyanide. Recent 31P NMR measurements (38) of cobalamins bound to the haptocorrin from chicken serum show that the nucleotide loop phosphodiester resonance is shifted downfield as much as 1 ppm upon binding of cobalamins to the protein. This effect has been definitely attributed to a change of phosphodiester conformation upon binding. Earlier NMR studies with free cobalamins (66-69) have shown that nucleotide loop conformation is dependent on the axial Co-N(Bz) bond length. Thus, a possible explanation for the more ordered, and somewhat concerted ligand substitution transition states for protein-bound H20Cbl is steric compression of the axial Co-N(Bz) bond leading to a strengthened Co-0 bond via the anticipated electronic trans effect. Further studies to clarify this question are in progress.