The Mechanism of Water Proton Nuclear Magnetic Resonance Relaxation in the Presence of Mammalian and Aplysia Metmyoglobin Fluoride

Abstract A mechanism of proton exchange leading to water proton relaxation is proposed for horse heart metmyoglobin fluoride which is mediated by H2O molecules and involves the distal histidine. The mechanism is based on a comparison of water proton NMR relaxation (T1-1 (H)) in the presence of horse and Aplysia metmyoglobin fluoride; Aplysia metmyoglobin fluoride was chosen for comparison because it lacks the distal histidine. Proton relaxation for horse heart metmyoglobin fluoride has been observed over a pH range of 5.0 to 11.5 in H2O and 95% D2O solutions while simultaneously monitoring the optical absorbance at 542 nm. For horse heart metmyoglobin fluoride, proton relaxation shows three pK values, at pH 6.2, 8, and one at a higher pH, which varies with the fluoride ion concentration due to the equilibrium metmyoglobin fluoride + OH- ⇋ MbOH + F-. Optical spectroscopy shows two pK values, one at pH 6.2 and one above 9. The latter pK has a large deuterium isotope effect in the optical spectrum, but none for proton relaxation. Theorell and Ehrenberg ((1951) Acta Chem. Scand. 5, 823) reported changes in the magnetic susceptibility of metmyoglobin fluoride with pK values of 6.3 and 8.11; however, the magnetic susceptibility increases when T1-1 (H) decreases and vice versa. We therefore conclude that the proton-iron distance changes twice during the titration.

A mechanism of proton exchange leading to water proton relaxation is proposed for horse heart metmyoglobin fluoride which is mediated by Hz0 molecules and involves the distal histidine. The mechanism is based on a comparison of water proton NMR relaxation (TIP1 (H)) in the presence of horse and Apl ysia metmyoglobin fluoride ; API ysia metmyoglobin fluoride was chosen for comparison because it lacks the distal histidine. Proton relaxation for horse heart metmyoglobin fluoride has been observed over a pH range of 5.0 to 11.5 in H20 and 95% DzO solutions while simultaneously monitoring the optical absorbance at 542 nm. For horse heart metmyoglobin fluoride, proton relaxation shows three pK values, at pH 6.2,8, and one at a higher pH, which varies with the fluoride ion concentration due to the equilibrium metmyoglobin fluoride + OH-ti MbOH + F-. Optical spectroscopy shows two pK values, one at pH 6.2 and one above 9. The latter pK has a large deuterium isotope effect in the optical spectrum, but none for proton relaxation. Theorell and Ehrenberg ((1951) Acla Chem. Stand. 5, 823) reported changes in the magnetic susceptibility of metmyoglobin fluoride with pK values of 6.3 and 8.11; however, the magnetic susceptibility increases when TIP1 (H) decreases and vice versa. We therefore conclude that the proton-iron distance changes twice during the titration.
In the presence of sufficient fluoride to convert metmyoglobin to metmyoglobin fluoride, the water proton relaxation rate, T,-' (H), is enhanced by approximately a factor of 236 over Z'-l (H) of the aquo complex (l). ' We will discuss three types of results in this paper. First, the effect of using I)20 as a solvent on the titration behavior of the proton relaxation and the ab- sorbance at 542 nm. The NMR experiment is sensitive only to protons and therefore can potentially measure the behavior of the protonated ligand, while the optical spectrum measures the behavior of the predominant species, which in 95y0 DzO is Mb2 with a deuterated ligand.
Thus by comparing the NMR and optical pK values in D20 it should be possible to learn something about the groups which are titrated.
The second aspect is the comparison of the values of Tl+ (H) for MbF as a function of pH with the magnetic susceptibility data of Theorell and Ehrenberg (2). From these data we can draw conclusions about the relative iron-proton distance during the course of the titration. Third, we will propose a mechanism for proton exchange in MbF solutions based on a comparison of the TIP1 (H) behavior of Aplysia Mb, which lacks a distal histidine, with that of horse Mb.

EXPERIMENTAL PROCEDURE
Commercial horse heart myoglobin preparations were used throughout.
Measurements were made on Seravac Lots 22L, 50, 52C, and 57A. Lyophilized material was dissolved in buffer and centrifuged.
Myoglobin was prepared from the buccal muscle of Aplysia calijornicans as described by Rossi-Fanelli and Antonini (3). trans-1,2-Diaminocyclohexane tetraacetic acid (CDTA) was added to the buffer used for grinding to a concentration of lO+ M in an attempt to avoid heavy metal contamination.
The crude protein was then concentrated and run in 0.02 2-amino-2-methyl-1,3-propanediol at pH 8.7 on a 25-cm column containing 5 cm of DEAE-Sephadex on the top and 20 cm of Sephadex G-50 (fine). Titrations were performed by making a high pH preparation and a low pH preparation and mixing the two to obtain intermediate point,s. This procedure avoids changes in Mb, F-, and ionic strength due to dilution or addition of reagents. Samples in D20 were prepared by dissolving lyophilized material directly in DtO phosphate buffer. All experiments were carried out in 0.1 M phosphate buffer. Background relaxation (T1-l (H) due to the intrinsic relaxation rate of protons in water and to the contribution of the globin or protein part) was determined by the addition of sufficient cyanide to saturate the heme site. The value of T1-l (H) in the presence of MbCN was then measured and subtracted from the value of T,-l(H) obtained under the same conditions in the absence of cyanide. That this is a valid procedure has been shown by demonstrating that !Z'l-1 due to MbCN is equal to that due to diamagnetic Mb02 or MbCO (4). Using this approach it is possible to separate the paramagnetic contribution, data was observed. Temperature was controlled at 25 f 0.2" unless otherwise specified.

RESULTS
Titration Behavior in DzO and HzO-The pH dependence of TI-1 (H) and the optical density at 542 nm are shown in Fig. 1. Results are shown for data taken in I-I20 and 95% DzO. The NMR data have three pK values, at pH 6.2, pH 8, and above pH 9. When the optical data arc taken at 542 nm no optical transition is detected at pH 8; however, pK values are observed at 6.1 and above 9. This is in agreement with the data of Theorell and Ehrenberg (2). The data in Fig. 1 were plotted in terms of pH meter readings and the scale for DzO was then shifted 0.4 pH unit on the horizontal axis3 (6). The graphs were then shifted vertically until the Tlm1 (H) values of MbCN solutions of the samples used for the titrations in D,O and Hz0 occurred at the same point.
This procedure compensates experimentally for the theoretically anticipated reduction in the background relasation observed in D,O (7). The resulting overlap of Tl-' (H) for MbF in DzO and Hz0 represents a true equality of T$ in the two solvents.
Note that the optical density scale is not shifted vertically.
Fluorine NMR Relaxation--'gF NMR relaxation was observed for commercial myoglobin samples. The values of molar relaxivity for fluorine, Tlm1 (F), showed great variation from one commercial lot to the next, while the molar relaxivity for protons remained constant from lot to lot. A titration with sodium azide at pH 5.3 while monitoring TIP1 (H), Tl-' (F), and absorba The commonly used correction of 0.4 unit is based on experimental observations in nonbiological systems; the very good fit which this value gives for the rather complex shape of the NMR data seems to justify its use in this case. The total absorbance density change was about 0.25 A units. No correction for proton or fluorine background was made.
ante at 542 nm, which is characteristic of the fluoride complex, showed no correlation between fluorine relaxation and TJ-' (H) or absorbance, while the latter two quantities were closed parallel, as shown in Fig. 2. A subsequent analysis for copper revealed concentrations ranging from 0.8 pg/lOO mg of Mb for Lot 22L to 3.8 pg/lOO mg of Mb for Lot 52C, which correlates with the relative magnitude of TIP1 (F) observed for each lot. The extreme sensitivity of Tl+ (F) to Cu(I1) is well known (5), and we attribute the measured relaxation to copper impurities. Removal of Cu(I1) led to Tl values of the order of background relaxation, so we may place an upper limit on F-exchange at the heme site of 1 x IO2 0.
Aplysia illetmyoglobin-The aquo and fluoride complexes of Aplysia metmyoglobin were studied as a function of temperature and pH. For both complexes, only slight decreases in Tl-' (H), less than 5% of the total relaxivity, were observed when cyanide or azide was used to saturate the heme site, as determined by visible spectroscopy. This is equivalent to a molar relaxivity of 0.1 & 0.05 x 10e3 M-' se1 for the aquo complex minus background.
This value may be compared to a value of 0.8 X 10e8 &f-l se1 for the difference between the acid aquo complex and the cyanide complex of horse heart metmyoglobin at 25". Likewise, the transitions from MbHOH to MbOH and from MbF to MbOH, detected by observation of the visible spectrum, were not accompanied by changes in Tl-' (H) as observed for mammalian myoglobins; nor did increasing the temperature to 60" result in an increase in Tl-1 for either complex.
In contrast, increasing the temperature to 40" nearly doubles the relaxivity of the mammalian acid aquo complex.
We therefore conclude that there is no relaxation attributable to protons in the first coordination sphere of the heme iron. It is possible that the where P,,, is the ratio of the number of proton sites in the complex to sites in solution.
As a system approaches the NMR fast exchange limit, rM becomes smaller and Tii increases until it reaches the limiting value of PMTi...
Since TM, the lifetime in the complex, usually decreases as the temperature increases, while TIM is constant or increases inversely with viscosity, we regard the absence of temperature dependence in our data as evidence of NMR fast exchange (9).
The formula used for calculating Tii (H), the relaxation rate in the complex, is (10,11) where peff is the effective magnetic moment calculated from xnr; the magnetic susceptibility, fi, yl, and /3 are physical constants in the appropriate units listed under Table I; r is the Fe(II1) to proton distance in angstroms; and 7c is the correlation time. If reasonable values of rc and xM (see Table I), and r (2.85 to 3.35 A for H-bonded Fe F-H complexes (1)) are substituted into Equation 2, then Ti.. calculated for MbF is equal to the value actually observed for T&.
If the system were in NMR slow exchange, the observed value would be less than the calculated value by the amount predicted from Equation 1. This is further evidence that the protons in MbF solutions are in NMR fast exchange4 (9). A similar conclusion has been drawn by Mildvan et al. (1). If changes in Tl-' (H) do not reflect changes in the kinetic rate of proton exchange with the paramagnetic center, then they reflect changes in either the mag-4 In fast exchange a nucleus requires more than one encounter with the paramagnetic center to "completely relax" and the measured relaxation time depends on the relaxation time in the complex and the ratio of sites available in the complex to sites available in the solution.
In slow exchange, relaxation is complete in one encounter and Tl-1 measures the rate-limiting step.
netic susceptibility of the complex or changes in the Fe(II1) to proton distance. T1-l is calculated from the experimental results by subtracting out the contribution of the solvent water and t,he diamagnetic contribution of the protein to the measured T,, as described under "Experimental Procedure." To evalua';e PM, when proton relaxation is measured in water, the appropriate value to use for the "sites in solution" is simply two times 55.5 M, the concentration of protons in pure water. For the number of sites in the complex, we assume there is one proton site per myoglobin fluoride molecule.
Due to the small size of the heme cavity, the only other likely possibility is that there may be two sites, in the case of "outer sphere" relaxation via a water molecule in the heme cavity.5 Outer sphere relaxation is also a fast exchange phenomenon and therefore one must distinguish between inner and outer sphere fast exchange by some means other than temperature dependence.
One can, however, eliminate the second mechanism by comparing MbF with MbHOH relaxation, since the products of the magnetic susceptibility (2) and the correlation time are about the same.6 The temperature dependence of MbHOH (1)7 indicates that protons in MbHOH are in NMR slow exchange, in contrast to protons in MbF solutions. It seems reasonable to assume that the same outer sphere mechanism would be available to MbHOH as is available to MbF. If the relaxation by MbF were completely outer sphere, we would expect MbHOH to have t'he greater relaxivity, since it would have both an outer sphere and a kinetically limited contribution. Since it is smaller, we conclude that the single site model is valid.
Titration Behavior in DzO and HzO-When the NMR data in DzO are compared to the data in HzO, the most st,riking feature is how little the pK values change, and how little change occurs in the magnitude of T1+ (H) after the background is compensated for. In the MbF portion of the titration, the lack of change in Tl-' (H) indicates that no substantial changes have occurred in the geometry or magnetic susceptibility of the heme site when Mb is dissolved in DzO. When the optical spectrum is observed in DzO, the pK of the low pH titration (6.2) does not change appreciably, but t.he pK of the high pH titration (above 9) is shifted to a much higher pH. Table II lists the NMR and optical pK values for two different fluoride concentrations.
Qualitatively, it is clear that the high pH titration represents a transition from MbF to MbOH.
Both Tim1 (H) and the opt.ical spectrum change from values characteristic of the fluoride complex to values characteristic of the hydroxide complex (1). Quantitatively, the value of pK,,,, the pH at which the conversion from MbF to MbOH is half-complete, also increases when the fluoride concentration is increased, as expected for a simple displacement of fluoride by hydroxide.
Calculated values of pK,,, are presented in Table II. We will now consider the following model of the events occurring at the heme site.
6 To account for the magnitude of the relaxivity such a hypothetical water site would have to be occupied at all times and the oxygen of the water, if it were present, should be visible by x-ray crystallography.
6 It can be seen from Equation 1 that these are the only variables that need to be considered.
J.L~~ is related to xrn by ,&r = 2.85 (x~T)~'~, and since it is squared in Equation 1, xm enters linearly. For MbHOH xmrc is (13.3 X 10e3) (2.77 X IO-lo) or 3.68 X lo-la, while for MbF x,,,T~ is (14.4 X 10m3) (2.17 X lo+') or 3.13 X lo+. The values used for 7r were obtained from EPR and NMR dispersion. The paramagnetic contribution to the mo1a.r relaxivity, T;;, at 25" is 0.8 X lo+ M-I s-1 for the mammalian acid aquo complex and is 2.1 X lo+ M-I 8-I for the-maximum value of the mammalian fluoride complex.
7 Unpublished observation.   where KJ+ is 1.26 X 10-g moles per liter corresponding to a pK of 8.9 for the acid-alkaline transition. b PI%, calculated for KF = 0.023 (14). c PI&, calculated for KF = 0.01 (15).l d Average of several measurements, the pK,b, (DZO) corresponds to the pH meter reading at the midpoint of the titration +0.4 pH unit as discussed in the text.
When the optical spectrum is observed in DzO, the high pH titration corresponds to the reaction The assumption of H bonding to the distal histidine is consistent with the x-ray structure (12). When !/'-I (H) is observed, the pK for the transition from the fluoride to the hydroxide complex is about 0.4 pH unit lower than the optical pK (see Table II). The simplest explanation for the difference is that the NMR titration involves only t,he direct participation of protons, since T1-l (H) detects only the protonated species. In other words, the pK detected by T1-l (H) in DzO at high pK is due to the reaction

MbFe(III)-F-H-his(7E) cf MbFe(III)-0-H-his(7E)
If the H-bonded protons did not participate in Z'-' (II) and relaxation was via an outer sphere mechanism, then we would expect the optical and NMR pK values to be shifted upward to the same pH in DzO instead of becoming separated as observed. This is further evidence that an outer sphere mechanism does ... HisE7 not contribute to relaxation by MbF and hence only a single proton site is responsible for paramagnetic relaxation. The separation of the opt.ical and NMR pK values in the case of the high pH titration may be contrasted to the effect of DzO on the ionization which occurs at pH 6.2. Since in this case the NMR and optical pK values are still 6.2 in DzO, we are observing the effect of a group whose ionization causes a structural or electronic change which alters the environment of the heme and thus indirectly changes both the optical and magnetic properties of the site. This lack of deuterium isotope effect may be related to the work of l'eisach and Hlumberg (13) who recently suggested that in some ferric heme proteins the proximal histidine, which is bonded through Nt to the heme iron and is H-bonded through N6 to Leu (4F), loses its N&-H bond, effectively becoming deprotonated.
If the ionization with a pK of 6.2 were due to the ionization of the proximal histidine, it would account for both the lack of a deuterium isotope effect and the failure of the optical and NivlR pK values to separate. The ionization of the proximal histidinc also results in a decrease of electron density at the iron (lo), which could account for the increase in KF which accompanies the pH 6.2 ionization (14,15). It is possible that the group with a pK of 8 is the N nitrogen of the distal histidine.
It is interesting to note that the observed pH dependence of the relaxation implies that the fluorine in metmyoglobin is protonated up to pH 10 and above, in contrast to HF which has a pK of 2.9. However, this is not surprising if one considers this proton as participat.ing in a complex somewhat analogous to a metal chelate complex, that is, Fe(III)-F-H-distal histidine. Then a conservative estimate (16) of the equilibrium constant for the dissociation of the proton from the complex is the product of KhisKHF; if one chooses values of pKHF = 2.9 and pKhi, = 6.5 and 7.5 this would result in a pK for the H-bonded proton between 9.5 and 10.5.

Dependence
of TIP1 (II) and xMM, and the Fe(IIl)-Proton Distance-The second aspect of the pH dependence of RlbF to be discussed is the origin of the two small changes in TIP' (H) at pH 6.2 and pH 8, and their relation to the magnetic susceptibility of myoglobin fluoride measured by Theorell and Ehrenberg (a), who also found two transitions at pH 6.03 and pH 8.11, in addition to the major one from i\bF and ;\IbOH.
Equation 2 shows that these transitions could be due to either changes in xrn, the magnetic succeptibility, or in r, the iroll-proton distance. However, where their susceptibilities increase we find that !i"?
(H) decreases, and vice versa, contrary to what would be cxpetted from Equation 2 if r. and r do not, change appreciably. It seems reasonable to assume that changes in rc large enough to account for the observed results (nearly 50';; of Tii in the pK 6.2 transition) do not occur. A value of 2 X lo-lo s \vas used in the calculation for rc; values of 2.2 x lo-i0 and 1.7 x lo-i0 s were measured by EPR and NMR dispersionl (4), respectively. This is the value of the electron relaxation time, Tie; it is the relevant quantity, since the tumbling time of the protein and the lifetime of a nucleus at the heme site are much longer. We therefore attribute the small changes in TIP1 (II) to changes in the iron-proton distance. These distances were calculated using Equation 2 and are presented in Table 1. The absolute values of I are of course affected by the choice of re, the accuracy of the measurements of TIP1 (H) and xm, as well as the approsimations involved in the theory from which T1.M is calculated.
The relative values of r are not as subject, to these uncertainties, however, and indicate the potential sensitivity of proton relasation to structural changes. The actual numbers obtained are in the range (2.85 to 3.35 A) determined by x-ray crystallography for H-bonded metal fluoride con~plcscs, rather than in the range for covalently bonded metal fluoride complexes (2.68 to 2.74 A) (1).
Mecimnism of Proton Exchange-There are two major justifications for the assertion that the proton exchange observed in MbF solutions is due to a single proton bonded to the fluorine. The first is the separation of the optical and NMR pK values observed in D,O, as discussed under the section on titration behavior in DzO and HzO. The second is the argument presented under the section on theory and preliminary observations based on a comparison of the magnitude of Tii in MbF and MbHOII solutions.
Since the proton relaxation due to MbF is in the NMR fast exchange region we can use Equation 1 and the data in Table I to place a lower limit of about 2 x 1Oj s-i on the eschangc rate. From the study of fluorine relaxation we can put an upper limit on the rate of fluorine eschange at 1 x lo2 s-i. Since the rate of proton exchange is so different from that of fluorine exchange, it is clear that exchange of HF does not occur; this conclusion is in agreement with that of Xlildvan el al. (1). At pH 7 and 8, the proton concentration is so low that to observe an exchange rate of 2 x lo5 s-i, a kinetic rate constant of 2 x lOi or greater would be required, exceeding even diffusion limited rate constants. It is thus improbable that we are observing simple proton exchange, therefore the species participating in exchange must be intact water molecules.
We can reconcile these facts by postulating that (a) the species entering the heme cavity is an intact water molecule which either ionizes before reaching the iron and recombines before leaving, or (b) that the water molecule exchanges its protons in a concerted reaction with the proton bound to fluorine.
A mechanism similar to Case b been proposed by llgenfritz and Schuster (17) for the transition from MbHOII to MbOH observed by T-jump.
Proton relaxation was also investigated in solutions of Aplysia metmyoglobin for both the acid and alkaline aquo complex and the fluoride complex. Aplysia myoglobin lacks the distal his-

2919
tidine and yet the visible spectrum of the complexes studied are quite similar to the spectra of mammalian myoglobins. No paramagnetic contribution Ivas observed for any of these complexes, even at high temperatures.
There are two possible explanations: (a) there are no protons to eschange, and (b) the exchange is too slow to be observed.
One might accept Case a for MbF; however, in view of the spectra1 similarities between flplysia and mammalian mgoglobins, it is an unlikely explanation for the aquo complexes.
Cast b suggests that the distal histidine plays a role in catalyzing proton exchange or ionization.
To summarize, we believe that proton relaxation in hIbF solutions is a cast where the primary contribution to relaxation may be attributed to a single site and yet, the exchange with the solution is mediated by intact water molecules, not protons or hydronium ions. Furthermore, the exchange between the site and the water molecules is catalyzed by the distal hist.idine.