Structural dynamics of myoglobin: ligand migration and binding in valine 68 mutants.

We have combined Fourier transform infrared/temperature derivative (FTIR-TDS) spectroscopy at cryogenic temperatures and flash photolysis at ambient temperature to examine the effects of polar and bulky amino acid replacements of the highly conserved distal valine 68 in sperm whale myoglobin. In FTIR-TDS experiments, the CO ligand can serve as an internal voltmeter that monitors the local electrostatic field not only at the active site but also at intermediate ligand docking sites. Mutations of residue 68 alter size, shape, and electric field of the distal pocket, especially in the vicinity of the primary docking site (state B). As a consequence, the infrared bands associated with the ligand at site B are shifted. The effect is most pronounced in mutants with large aromatic side chains. Polar side chains (threonine or serine) have only little effect on the peak frequencies. Ligands that migrate toward more remote sites C and D give rise to IR bands with altered frequencies. TDS experiments separate the photoproducts according to their recombination temperatures. The rates and extent of ligand migration among internal cavities at cryogenic temperatures can be used to interpret geminate and bimolecular O2 and CO recombination at room temperature. The kinetics of geminate recombination can be explained by steric arguments alone, whereas both the polarity and size of the position 68 side chain play major roles in regulating bimolecular ligand binding from the solvent.

Myoglobin (Mb) 1 has served as a model system for protein structure-dynamics-function studies for more than 40 years. This small globin reversibly binds diatomic ligands like oxygen (O 2 ), nitric oxide (NO), or carbon monoxide (CO). Almost 50 years ago, its three-dimensional average structure was resolved in atomic detail (1). The diffraction data revealed a polypeptide chain tightly wrapped around a heme group, leaving no direct pathway from the outside to the central heme iron where the ligands bind. Therefore, it became evident early on that protein motions are essential for Mb to perform its function.
Initially, ligand binding was assumed to be a simple one-step process (2). This view had to be refined when flash photolysis experiments showed non-exponential internal recombination and multistep kinetics (3). Over the years, a more detailed picture of ligand binding has emerged. In all current models, the ligand has to overcome a sequence of energy barriers to migrate from the solvent to the active site. For a long time, the underlying structural determinants of these barriers remained controversial. Interpretations of the nature of the kinetic intermediates included proximal structural transitions of the protein that affect the reactivity at the heme iron atom (4,5) and/or migration of the ligand to different docking sites (6,7). The initial models were, in the absence of direct structural information, merely speculative.
In recent years, a clear structural picture of the ligand binding reaction in Mb has begun to emerge (Fig. 1). It is based on an integrated experimental approach, combining x-ray crystallography of photoproducts, time-resolved spectroscopy and protein engineering. After photolysis of MbCO crystals at liquidhelium temperatures, photolyzed ligands were found to reside in the primary docking site B in close vicinity of the heme iron (8 -10). Two additional secondary photoproducts, C and D, were identified that are characterized by CO ligands residing in separate internal protein cavities (11)(12)(13). A total of four cavities spacious enough to accommodate a xenon atom, the socalled Xe cavities, are known to exist in the Mb structure (14). Recent time resolved x-ray diffraction data have given direct evidence of their role as transient ligand docking sites at physiological temperature (15). These open volumes appear to facilitate both ligand capture and escape and might play a role in regulating the reaction of NO with bound O 2 , which is an important, newly discovered physiological function of Mb (16 -19).
The new crystallographic evidence of secondary docking sites prompted us to re-examine the functional role of the highly conserved distal valine at position 68, which is located both adjacent to the ligand binding site and along the pathway to the xenon (Xe) 4 cavity, as shown in Fig. 1. This buried hydrophobic side chain has been the subject of extended ligand binding studies (20 -27). Egeberg et al. (24) measured association and dissociation rate coefficients of Val-68 mutants at room temperature by stopped flow rapid mixing and flash photolysis techniques. Their results indicated that Val-68 is part of the overall kinetic barrier to ligand binding. A monotonic decrease of the ligand association rate with increasing size of the Val-68 side chain suggested that residue 68 regulates ligand approach to and capture at the heme iron. In 1995, Quillin et al. (25) correlated the functional properties of V68A, V68L, V68I, and V68F mutants with changes observed in their crystal structures.
In V68F MbCO, the benzyl side chain points away from the ligand-binding site and occupies a region in the back of the distal pocket, partially filling the Xe4 cavity ( Fig. 1, right  panel). The bulky side chain appears to trap photolyzed ligands near the heme iron, enhancing both the rate and extent of CO, O 2 , and NO geminate rebinding. Molecular dynamics simulations suggest that the volume of residue 68 governs ligand escape to remote sites within the protein interior. The polarity of amino acid 68 also affects the bimolecular rate of ligand binding. The V68T mutation in pig Mb causes a ϳ10-fold decrease in kЈ O 2 , the overall rate coefficient for O 2 binding from solvent (26). This mutation also affects ligand migration into and out of the distal pocket, but its effects on ligand movement to secondary docking sites have not been examined carefully. In this study, the effects of polarity and side chain volume on the lifetimes and spectral characteristics of the B and C/D state intermediates at cryogenic temperatures have been examined more systematically and in great detail.
Determination of photoproducts by x-ray diffraction methods is only feasible if the states are populated to at least 20 -30%. Fourier transform infrared (FTIR) spectroscopy is a more sensitive technique that allows resolution of different species with fractional populations as low as 1%. Sensitivity to local structure can be combined with time resolution to study dynamics. Measurements carried out as a function of temperature, for instance with the TDS technique, yield information on the energy barriers governing internal migration and rebinding of the photodissociated ligand. The CO stretch bands of hemebound and photodissociated CO serve as powerful tools to probe the electrostatic fields and accessible space in the immediate vicinity of the CO molecule inside the protein matrix (28 -31). Wild-type (wt) MbCO shows a few discrete stretching bands for heme-bound CO. These A state bands have provided valuable insights into the structural details of the distal heme pocket (32,33). In wild-type MbCO, the taxonomic substates A 0 , A 1 , and A 3 associated with CO stretch ( C-O ) bands at ϳ1965, ϳ1945, and ϳ1933 cm Ϫ1 differ mainly in the position of the distal histidine, His-64 (34,35). Structural, vibrational, and CO rebinding properties of these substates and their corresponding photoproducts are summarized in Table I. At low pH, the imidazole side chain is protonated and swings toward the solvent to better solvate the positive charge (32). The resulting open conformation has been associated with A 0 , it is characterized by an apolar pocket. The closed conformation is related to A 1 and A 3 , in which the positive partial charge of the N ⑀ -H of the His-64 imidazole interacts electrostatically with the bound CO (34, 36 -39).
Photodissociated CO ligands that are trapped inside the protein matrix typically display multiple sharp infrared (IR) lines near 2130 cm Ϫ1 (Table I), providing clear evidence that structurally well-defined ligand docking sites exist (28 -30). In analogy to the A state bands, the individual stretch frequencies are affected by the local environment, mainly through Stark effects of the local electric field acting on the CO dipole (40 -43). Upon photolysis below ϳ10 K, the photodissociated ligand settles into the initial docking site B (30). Opposite orientations of the CO molecule with respect to the heme center have been associated with two IR bands, B 2 and B 1 , at 2119 and 2131 cm Ϫ1 (44). B 0 at ϳ2150 cm Ϫ1 represents CO ligands at site B in the A 3 conformation with a strong interaction with His-64 even in the unbound state (29). The corresponding band for the opposite orientation was suggested to be hidden underneath B 1 (29,43). After prolonged illumination, especially at higher temperatures, ligands are observed at more remote secondary sites with slower reaction pathways back to the iron atom (4,5,45). Based on previous investigations of native, L29W, and L29Y/ H64Q/T67R sperm whale Mbs, these additional photoproduct states are expected to be characterized by CO-stretching bands different from B 0, B 1 , and B 2 (40,42,43,46). 2 To differentiate between steric and electrostatic effects of the distal valine on ligand migration and recombination, mutations were selected (i) to alter the electrostatic potential adjacent to the binding site and/or the primary docking site and (ii) to change the open volume in the interior of the distal pocket. Replacement of the non-polar Val-68 by the isosteric threonine (V68T) provides an unequivocal test of the importance of polarity in determining the rate of ligand binding. The V68S mutant has a smaller and polar side chain and V68M a larger polar one. The V68F and V68W mutations introduce large aromatic side chains into the distal pocket and V68Y combines polarity with an increase in volume. To relate the low temper- The Leu-29 side chain is cut away to make the bound CO visible. Leu-69 and Ile-28 were removed to allow visualization of the Xe4 pocket. The labeled amino acid side chains are those closest to the bound ligand. The Val-68 side chain is marked in green. The other unlabeled apolar side chains are marked in yellow and include, starting from the top right and going clockwise around the bound CO, Phe-46, Phe-33, Leu-32, Ile-111, Leu-72, and Ala-71. The polar external residues are marked in cyan and include the Thr-67 (unlabeled upper left) and His-64 side chains. The spaces making up the B and C (Xe4) sites are labeled in wild-type MbCO. In the Phe-68 mutant, the Xe pocket is almost completely filled; the cavity comprising the B site is smaller due to rotation of the C␦ atom of Ile-107 about the C␥-C␤ bond; but there is no hindrance by the benzyl side chain immediately adjacent to the iron atom.
ature spectroscopic data to physiological ligand binding, room temperature flash photolysis and rapid mixing data were collected for both geminate and bimolecular CO and O 2 binding to the mutants. The observed kinetic traces are interpreted in terms of the results of the low temperature FTIR-TDS experiments.

MATERIALS AND METHODS
Sample Preparation-Plasmids containing the mutant Mb genes were transformed into Escherichia coli strain Tb1, expressed and purified as described previously (48). IR samples were prepared by dissolving the lyophilized protein at a concentration of ϳ15 mM in cryosolvent (75% glycerol, 25% potassium phosphate buffer (v/v), pH 8) and subsequent reduction with excess dithionite under a CO atmosphere. For the kinetic measurements, lyophilized protein was dissolved in 100 mM sodium phosphate buffer, pH 8 at a concentration of ϳ10 M and sealed in a 1 ϫ 1 ϫ 3 cm 3 cuvette. The solution was equilibrated with 1 atm of CO and reduced with excess sodium dithionite solution. MbO 2 samples were prepared by adding sodium dithionite to a ferric Mb solution followed by a rapid passage through a G25 Sephadex column to remove excess reducing agent (for flash photolysis) or by exchanging CO from reduced MbCO samples by equilibration with 1 atm O 2 under a bright light source (for the on-rate determinations).
Cryospectroscopy and Photolysis Setup-A few microliters of the protein solution were placed between two CaF 2 windows (diameter, 25.4 mm) separated by a 75-m thick Mylar washer. The windows were sandwiched inside a block of oxygen-free, high-conductivity copper mounted on the cold-finger of a closed-cycle helium refrigerator (model SRDK-205AW, Sumitomo, Tokyo, Japan). The sample temperature was measured with a silicon temperature sensor diode and regulated by a digital temperature controller (model 330, Lake Shore Cryotronics, Westerville, OH). A continuous wave, frequency-doubled Nd-YAG laser (model Forte 530 -300, Laser Quantum, Manchester, UK), emitting 300 milliwatt output power at 532 nm, was used to photolyze the sample. The laser beam was split and focused with lenses on the sample from both sides. The standard photolysis rate k L was determined to be ϳ20 s Ϫ1 at low temperature. A Fourier transform infrared (FTIR) spectrometer (IFS 66v/S, Bruker, Karlsruhe, Germany) equipped with an InSb detector was employed to collect transmission spectra in the mid-infrared between 1800 and 2400 cm Ϫ1 at a resolution of 2 cm Ϫ1 .
Temperature Derivative Spectroscopy-TDS is an experimental protocol designed to investigate thermally activated rate processes that are characterized by distributed enthalpy barriers (5,29,49). This twodimensional technique has proven to be particularly useful for detailed investigations of ligand recombination because it allows separation of populations not only by their spectral components (wavenumber, ) but also by their rebinding properties (recombination temperature, T).
Photolysis creates an initial non-equilibrium set of intermediate states within the sample. The subsequent TDS measurement records the relaxation of the sample back to equilibrium as the temperature is ramped up linearly with time in the dark at a rate ␤ ϭ 5 mK/s. During the TDS experiment, FTIR transmission spectra I(,T) are taken every 1 K. The integrated absorbance A of the spectral band taken at the lowest temperature represents the total photolyzed population, N. The change in integrated absorbance, ͐⌬Ad, is assumed to be proportional to the change in population of a state, ⌬N, during the acquisition of two successive spectra. Absorbance changes arise from two different rate processes, CO rebinding to the heme iron and CO migration to other locations in the protein. Distributed barrier heights and recombination rates are universal features of protein ensembles at low temperatures (3,50). Only small changes in the absorbance are expected at the lowest temperatures because merely a few molecules can surmount the barrier for recombination. At higher temperatures, ͐⌬Ad will increase because more molecules are able to rebind. At even higher temperatures, ͐⌬Ad will decrease again as the population of the photodissociated states gets depleted. Therefore, the TDS protocol ensures that the different rate processes are sorted according to their activation enthalpy barrier, which is, to a good approximation, linearly related to the temperature at which the reaction is observed.
For a quantitative modeling of the TDS data, the derivative of the population with respect to temperature, dN/dT, is approximated by calculating absorbance difference spectra ⌬A(,T) from transmission spectra I(,T) at successive temperatures. Two assumptions are implicit in this procedure. (i) The change in absorbance is proportional to the rebinding population, and (ii) there is no intrinsic temperature dependence in the measured spectra. TDS data are conveniently presented as contour plots of the absorbance change on a surface spanned by the wavenumber () and temperature (T) axes.
Geminate and Bimolecular Recombination at Room Temperature-Experiments were carried out with a home-built flash photolysis apparatus, using a 6-ns (full width at half maximum) pulse from a frequency doubled Nd:YAG laser (model Surelite, Continuum, Santa Clara, CA) for photodissociation. Ligand binding was monitored with light from a tungsten source, which was passed through a monochromator set at 436 nm. A photomultiplier tube (model R5600U, Hamamatsu Corp., Mid- b Whereas A 0 is associated with samples at low pH, a band at similar frequency in high pH samples corresponds to substate A 3 , which is most likely characterized by a neutral histidine inside the heme pocket, with N 2 protonated but pointing away from the bound ligand (32). c See Ref. (29).
dlesex, NJ) measured the changes in transmittance. The data were recorded with a digital storage oscilloscope from 10 ns to 50 s (model TDS 520, Tektronix, Wilsonville, OR) and a home-built logarithmic time-base digitizer (Wondertoy II) from 2 s to 100 s. Absorbance traces were scaled according to the absorbance of the Soret band. Subsequently, they were normalized with respect to the trace of wild-type MbCO. The fraction of ligands that escape into the solvent, N S (CO), was set to 0.96 for wild-type MbCO (51). MbO 2 traces were adjusted by an additional factor of 1.5 to account for the different extinction coefficients of MbCO and MbO 2 in the Soret band. In this normalization procedure, the quantum yields for both CO and O 2 are assumed to be unity (52,53). Previous studies using 80 fs excitation pulses have shown that ϳ70 -80% of photoexcited MbO 2 complexes decay back to the MbO 2 bound state within 2-10 ps (52,54,55).
In the simplest model, ligand recombination from the solvent can be described by a three-state kinetic scheme, with bound state A, photolyzed state B with the ligand trapped inside the protein and solvent state S (51). The apparent rate coefficient S that describes bimolecular recombination is given by Equations 1 and 2, with rate coefficient k BA for recombination from B to A, pocket occupation factor P B , which represents the equilibrium coefficient for finding CO at the geminate (B) site (if there were no recombination at the heme iron), and N S , the probability for ligands to escape from the protein after dissociation. The rate coefficients for transitions of ligands between states B and S are denoted as k BS and kЈ SB . This simple model does not explicitly take into account the existence of several internal docking sites. However, more complicated kinetic schemes can be simplified to an effective three-well model under certain conditions, as we have shown recently for ligand binding in wild-type Mb (43). Note also that Equation 1 yields the bimolecular rate coefficient kЈ CO  ). O 2 dissociation rate coefficients, k O 2 , were determined independently by analyzing stopped-flow rapid mixing time courses in which bound oxygen was displaced by carbon monoxide. CO association rate coefficients were obtained either by flash photolysis of MbCO samples equilibrated with 1 atm CO using the 300 ns dye laser or by mixing deoxyMb with various concentrations of CO in the stopped flow rapid mixing apparatus. CO dissociation rate coefficients were measured by mixing MbCO samples in 50 M CO with 1000 M NO. Overall equilibrium coefficients for O 2 and CO binding were calculated from the ratios of the association and dissociation rate coefficients. In these experiments, the reaction conditions were 20°C, 0.1 M potassium phosphate, 0.3 mM EDTA, pH 7.0. Time courses for O 2 and CO binding to the V68M and V68Y at room temperature were biphasic (see Fig. 10), making assignment of the overall equilibrium coefficients ambiguous.

Infrared Spectra of Bound and Photodissociated CO at 3 K-Each
MbCO sample was cooled to 3 K in the dark. Subsequently, a transmission spectrum I dark and, after 1-s illumination, a second transmission spectrum I light were taken. The difference spectrum was calculated as Ϫlog(I light /I dark ); it reflects the changes within the sample caused by laser illumination. The FTIR absorbance difference spectra in Fig. 2 give a first impression of the influence of specific amino acids at position 68 on the IR stretch bands of heme-bound and photodissociated CO at very low temperature. The effects of polar amino acid replacements (polar mutants) are shown in Fig. 2a, and the effects of aromatic amino acid substitutions (aromatic mutants) are shown in Fig. 2b. This grouping of the mutants is kept throughout this work. All peak frequencies are compiled in Table II.
In the spectral region of the bands of heme-bound CO, wild type, and most Val-68 mutant MbCO samples display a dominant A substate band at ϳ1945 cm Ϫ1 (A 1 ) and a smaller one at ϳ1930 cm Ϫ1 (A 3 ). A minor band at ϳ1970 cm Ϫ1 is also discernible. In V68Y, A 3 is the dominant substate. Mutant V68T displays a pronounced band at ϳ1965 cm Ϫ1 as well as a minor peak at ϳ1938 cm Ϫ1 . The A bands are shifted to lower frequencies in V68M (1918,1937, and 1960 cm Ϫ1 ).
In the same figure, we observe a number of bands between 2110 and 2160 cm Ϫ1 , associated with photoproducts B 0 , B 1 , and B 2 . The absolute values of these peak frequencies indicate a bond order of ϳ2.8 for the non-covalently bound CO. The polar mutations at position 68 produce relative small effects, Ϯ 3 cm Ϫ1 , on the positions of the C-O peaks of the B states as compared with wild type (Fig. 2a, Table II). The V68T mutation produces a small 3 cm Ϫ1 increase in the positions of the B 2 and B 1 peaks. A similar increase in C-O for the B 1 state of V68S is visible. In contrast, the phenylalanine, tyrosine, and tryptophan replacements all cause pronounced splitting of the photoproduct bands (Fig. 2b, Table II). The peak position of B 2 shifts from 2119 to ϳ2110 cm Ϫ1 and the B 1 peak to 2138 cm Ϫ1 . In V68Y, we note a shoulder on B 2 , an additional minor photoproduct band at ϳ2130 cm Ϫ1 and a strong B 0 band, due to the large population in A 3 . In most mutants, B 0 is at 2149 -2151 cm Ϫ1 ; however, in V68F MbCO, the C-O band for this substate is shifted by ϳ6 cm Ϫ1 to 2156 cm Ϫ1 .
Interconversion of B States and Geminate Recombination from Site B-Brief illumination for 1 s at 3 K cleaves the bond between the heme iron and the CO ligand. The appearance of two distinct bands, B 1 and B 2 , in the photoproduct spectra shows that photolyzed ligands are trapped at the initial docking site B within the distal pocket, presumably in two distinct orientations. To examine the thermally induced interconversion of the B states and geminate rebinding, TDS experiments were performed subsequent to 1-s illumination. The TDS maps monitoring the absorbance changes in the bands of hemebound CO (left columns in Figs. 3 and 4) show CO recombination from only one intermediate state (except for V68S and V68M) to each bound state conformation. Recombination to substate A 1 ( C-O ϳ1945 cm Ϫ1 ) dominates most of the contour plots and reaches a maximum at ϳ40 K (and 15 K for V68S). A 3 generally rebinds at significantly higher temperatures (ϳ70 K). Weak contours representing recombination to a third A substate with a high frequency C-O band are seen for V68S and V68M MbCO. Rebinding occurs at temperatures similar to the formation of A 1 .
The TDS maps of the photodissociated CO bands in Figs. 3 and 4 (right column) show several positive and negative peaks. If only rebinding occurred, the maps would display only negative contours (dashed lines) due to loss of absorbance in the photoproduct B states. At ϳ20 K, however, the population increase in the 2130 cm Ϫ1 band, B 1 (A 1 ), is concomitant with the decrease at 2118 cm Ϫ1 , B 2 (A 1 ), implying a population transfer from one photoproduct state to another in addition to net rebinding. Additional solid contours at ϳ2150 cm Ϫ1 , most pronounced in V68Y (Fig. 4f), indicate a population increase in B 0 . Above 70 K, B 1 is depleted, recombination in A 1 is complete, and rebinding from B 0 to A 3 occurs. The temperature of maximum transfer from B 2 to B 1 is lowest for V68S (ϳ12 K) and increases for the series wild type (ϳ15 Remarkably, the TDS plot of V68T MbCO shows only rebinding. We do not observe the transition from B 2 to B 1 in this mutant (Fig. 3f).
Ligand recombination in V68T occurs from both the 2122 cm Ϫ1 band at 20 K and from the 2133 cm Ϫ1 band at ϳ30 K. As in all other mutants, rebinding occurs in A 3 at temperatures above ϳ70 K.
Ligand Movement into Secondary C Sites-Prolonged illumination at elevated temperatures enables ligands to populate intermediate states with higher barriers to recombination (5,42,43,57,58). 2 These slowly rebinding intermediates are formed by ligand migration to remote sites in the protein interior, i.e. the Xe4 and Xe1 sites. To survey the effects of Val-68 replacements on ligand migration, the mutant MbCO samples were cooled slowly from 160 to 3 K under steady illumination. Subsequently, FTIR transmission spectra were collected and referenced against 3-K transmission spectra taken without prior illumination. The upper and lower panels in Fig. 5 display the stretch bands of photolyzed CO after brief (1 s at 3 K, panels a and b) and extended (slow cooling from 160 to 3 K, panels c and d) illumination, respectively. All spectra were normalized to equal areas of 1 OD cm Ϫ1 .
After extended illumination, V68S and V68T, like wild type, display a pronounced peak at ϳ2132 cm Ϫ1 that was not ob-  (20), 2130 (7) 2150 (27)  V68W 1928 (28) 1945 (71) 1967 (1) 2110 (56) 2138 (36) 2148 (8) a Bands positions were determined at 3 K, with an estimated experimental error of Ϯ0.5 cm Ϫ1 . b The assignment of a particular band to A 0 , A 1 and A 3 is based on the three-dimensional structure and not on the stretch frequency. c Photoproduct bands were assigned according to their stretch frequencies. served after short illumination. V68M shows a new, large peak at 2134 cm Ϫ1 , and less intense ones at 2126 and 2118 cm Ϫ1 . In all polar mutants, residual population is present in B 2 ; B 0 is no longer visible (Fig. 5c). In contrast to the results for the polar mutants, the continuous illumination spectra of V68F, V68Y, and V68W MbCO show marked persistence of the peaks assigned to the B 0 , B 1 , and B 2 substates (Fig. 5d). Compared with brief illumination, B 1 at ϳ2138 cm Ϫ1 has lost intensity. Instead, a band at 2130 cm Ϫ1 has emerged in V68F and V68W. In V68Y, this band was already present after brief illumination but has gained intensity.
The FTIR absorbance difference spectra measured after extended illumination suggest that photolyzed ligands indeed get trapped at intermediate sites different from the initial docking site B. Hence, FTIR-TDS experiments were performed subsequent to slow cooling under illumination to observe ligand recombination as a function of temperature. The A state TDS contour map of wild type in Fig. 6a shows recombination in A 1 at ϳ50, 80, and 120 K. The plot of V68S also depicts recombination in substate A 1 at ϳ40 K (and 15 K). Additional recombination in A 1 is apparent as a long tail stretching out to 120 K. In V68T, a second population rebinding at 90 K is clearly separated from the peak at 25 K, and a third subpopulation is discernible around 180 K. The contour plot of V68M shows rebinding to the main band at 1937 cm Ϫ1 at 25 and 160 K.
At most two subpopulations of ligands recombine in substate A 3 , at ϳ70 and ϳ160 K. For V68M, recombination in A 3 is confined to temperatures around 180 K. In V68S and V68M, contours around 1965 cm Ϫ1 are associated with recombination to an additional substate A x (32).
The TDS contour maps of the photoproducts (Fig. 6, right column) of wild-type and V68T MbCO show similar features, which differ significantly from those measured after 1-s illumination (Fig. 3, right column). They are characterized by exchange at ϳ20 K, strong negative contours around 2132 cm Ϫ1 that stretch over a significant portion of the temperature range, and a second, less distinct feature at 2119 cm Ϫ1 that merges with the dominant one at ϳ100 K. V68S MbCO displays a markedly different TDS map compared with the one after short illumination (Fig. 3d). The extensive exchange feature at 15 K is smaller, and recombination from the band at 2132 cm Ϫ1 extends up to 100 K. For V68M MbCO, the TDS map after prolonged illumination is very complex, showing multiple exchange processes. This suggests that many more conformational substates are present in this mutant than in wild type and the V68S and V68T mutants.
The TDS maps of the aromatic mutants after extended illumination are shown in Fig. 7. They resemble those measured after 1-s illumination (Fig. 4). In the A state maps, recombination in A 1 is visible at ϳ40 K and in A 3 at ϳ70 K. All three mutants show additional recombination features in the A 1 substate frequency range at ϳ90 K. In V68F and V68Y, some recombination within the A 3 conformation occurs also at ϳ160 K. The photoproduct maps of V68F/Y/W exhibit similar contours compared with those after short illumination, except for additional peaks at ϳ2130 cm Ϫ1 related to photoproducts at more remote sites. The pronounced B 2 to B 1 exchange occurs at ϳ20 K for all three mutants, and ligands rebind in A 1 from B 1 and in A 3 from B 0 .
The TDS data in Figs. 6 and 7 were integrated along the wavenumber axis and plotted in Fig. 8 for both the polar and aromatic mutants separately for the A 1 (Fig. 8, a and c) and A 3 frequency ranges (Fig. 8, b and d). The total area under each original difference spectrum was normalized to 1 OD cm Ϫ1 to allow comparisons of populations. Note that the relative amounts of each substate are plotted on a linear scale in Fig. 8, whereas the contours in the TDS maps are spaced logarithmically.
The differences in the temperature maxima for recombination between the polar and wild-type proteins versus the aromatic mutants are fairly dramatic. Mutants with smaller side chains at position 68 show extensive rebinding at high temperatures. In contrast, most of the rebinding within the A 1 conformation occurs in the aromatic amino acid mutants at low temperatures, ϳ40 K, even after prolonged illumination. The aromatic amino acid mutants also show relatively low temperature maxima (ϳ60 K) for rebinding in A 3 .
Geminate Recombination at Room Temperature-Nanosec- ond laser photolysis experiments were performed on mutant MbCO and MbO 2 samples to relate the low temperature FTIR-TDS data to physiological binding processes. The MbCO samples were equilibrated with 1 atm of CO; the MbO 2 samples were prepared in buffer equilibrated with air (0.2 atm O 2 ). Time courses for both geminate and bimolecular rebinding from solvent are shown in Fig. 9. Two parameters were extracted from these data: (i) the amplitude N S representing the fraction of Mb molecules from which ligands escape into the solvent after photodissociation, and (ii) S , the apparent rate coefficient for recombination from the solvent (Equation 1 and Table III). In most samples, geminate recombination of CO and O 2 extends out to ϳ100 ns. Generally, O 2 rebinds faster from the solvent than CO. The amplitudes N S (CO) and N S (O 2 ) vary significantly for each mutant. For instance, only 4% of CO rebinds geminately in wt, but ϳ50% geminate recombination occurs for O 2 . The pronounced differences between polar and aromatic mutants that were observed in the FTIR-TDS experiments at cryogenic temperatures are also apparent in the room temperature kinetic traces. The amplitudes of ligand escape, N S (CO), for the polar mutants, V68M, V68T, and V68S, are similar to the wild-type value. In contrast, there is extensive geminate recombination of CO in the aromatic Val-68 mutants, with a progressive decrease in ligand escape to the solvent with increasing side chain size; N S (CO) equals 0.61, 0.41, and 0.22 for V68F, V68Y, and V68W MbCO, respectively. The kinetics of geminate rebinding to many of the position 68 mutants have been described in detail by Scott et al. (56). Time courses for O 2 rebinding are shown in Fig. 9, c and d for comparison with the MbCO derivatives. N S (O 2 ) is considerably  Tables IV and V and Fig. 9 and 10).  A 1 (a, c) and A 3 (b, d) calculated from TDS data (Figs. 6 and 7) and plotted as a function of temperature. a and b, wild type, V68M, V68S, and V68T; c and d, wild-type, V68F, V68Y, and V68W. smaller in the V68M, V68S, and V68T mutants and below 10% in the aromatic mutants. In wild-type MbO 2 and the polar mutants, the fraction of nanosecond geminate recombination is between 40 and 20%, whereas more than 90% of the ligands rebind geminately in V68F and V68W (56). In addition, the aromatic mutants show markedly reduced rate coefficients for rebinding from the solvent, S (CO) and S (O 2 ), (Table III), as discussed below. Tables IV and V. Sample time courses for bimolecular ligand binding to wild-type and V68Y Mb are shown in Fig. 10. In these experiments, complete photolysis was achieved with an intense, ϳ0.5 s dye laser excitation pulse, and only rebinding from the solvent phase is observed. With two exceptions, the time courses for both O 2 and CO association are monophasic and can be described by a single exponential expression in which the observed rate depends linearly on ligand concentration. Mutants V68M and V68Y show markedly biphasic time courses for the binding of O 2 and less heterogeneous behavior for CO binding (Tables IV and V, Figs. 9 and 10). The kinetic data for these mutants were fitted with two exponentials to obtain estimates for the fast and slow bimolecular rate coefficients. These mutants also show biphasic time courses for O 2 dissociation, but not for CO dissociation, whereas all the other mutants showed monophasic behavior.

Effects on Overall Association and Dissociation Rate Coefficients and Equilibrium Binding-Summaries of the effects of the polar and aromatic substitutions on O 2 and CO binding at 20°C are given in
In general, the polar 68 mutations in sperm whale Mb cause significant decreases in the association rate and equilibrium coefficients for the O 2 binding and increase the dissociation rate coefficient (Table IV). The most pronounced effects are observed for the sperm whale V68T mutant, which shows 10and 20-fold decreases in kЈ O 2 and equilibrium constant K O 2 , respectively, and a 3-fold increase in k O 2 . In contrast, the V68S and V68T mutations have much less effect on CO association rate coefficients, causing only 20 -30% decreases in kЈ CO . However, the rate of CO dissociation, k CO , does increase 2-4-fold. The net result is a decrease in CO affinity. However, the change in K CO is smaller than that of K O 2 (Table IV). Similar effects on ligand binding were observed for pig V68T Mb by Smerdon et al. (26). In V68M Mb, the faster reacting component appears to have O 2 and CO affinities similar to those of wild type, but both the association and dissociation rate coefficients are 3-5-fold smaller. The slower reacting conformation shows extremely small association rate coefficients, which are similar to those observed for V68W Mb. Since there is no compensating decrease in the rate of ligand dissociation, this slowly reacting conformation has a very low affinity for both O 2 and CO.
The aromatic mutations at position 68 cause decreases in overall association and dissociation rate coefficients, particularly for O 2 binding, but in these cases, there is little effect on net O 2 or CO affinity. The lack of change in K O 2 and K CO is due to compensating decreases in both kЈ O 2 , kЈ CO , and k O 2 , k CO . As in the case of V68T, the effects of the V68F and V68Y mutations on kЈ CO or S (CO) are much smaller than those on kЈ O 2 or S (O 2 ) (Tables III and V). Only in the cases of the slow reacting V68Y component and the V68W mutant, do we see substantial decreases in the rates of CO binding compared with wild type. However, even in these cases, there is little decrease in K CO or K O 2 .

Effects of Polar and Aromatic Position 68 Mutations on
Bound CO-The IR-stretching frequencies of heme-bound CO are only weakly affected by apolar mutations at position 68 ( Figs. 1 and 2). Even when direct steric hindrance occurs, as in the case of Ile-68 MbCO that shows an ϳ10-fold reduction in affinity, there is little change in C-O (59). As shown in Fig. 1, the ␥ 2 -CH 3 of Val-68 makes contact with the O atom (red) of the bound ligand and appears to cause some tilting of the Fe-C-O complex. However, this hindrance has been shown experimentally and theoretically to have little effect on C-O (59,60).
In contrast, introduction of the isosteric and polar threonine, V68T, induces large perturbations in the spectrum of hemebound CO, despite only small changes in the local structure as compared with wild type (Fig. 1). The crystal structure of pig Mb V68T MbCO reveals that the OH side chain group of Thr-68 donates a hydrogen bond to the main chain carbonyl oxygen of His-64 MbCO (26,27). Consequently, the non-bonding electrons of the hydroxyl oxygen are oriented toward the CO oxygen, thus opposing backbonding and shifting C-O to higher values. As a result, the dominant A state band of mutant V68T appears at a frequency (1964 cm Ϫ1 ) that would normally be associated with an open A 0 -like bound conformation, where the distal histidine His-64 has moved out of the distal heme pocket (32,36). The rather high temperature for recombination in the minor substate (ϳ70 K) at 1937 cm Ϫ1 as well as the peak position of the corresponding photoproduct band, 2149 cm Ϫ1 , identify this conformation as A 3 . Here again, the band is most likely shifted to the comparably high frequency of 1937 cm Ϫ1 because of the hydrogen-bonding network. The same effect was observed in hemoglobin mutant ␤V67T (61,62).
In V68M MbCO, the IR-stretching bands of the bound ligand are shifted toward lower frequencies with respect to wild-type MbCO, which is indicative of more positive charge near the bound oxygen atom (59, 63-67). As no structural data are available for this mutant, we can only speculate that the loss of  (78). Rate coefficients for V68T were taken from unpublished data collected by Olson,Eich,and Dou (47) and re-measured for this work. Association rate coefficients were measured using either rapid mixing or 300 ns laser photolysis techniques. The dissociation rate coefficients were determined from the analysis of rapid mixing, ligand displacement reactions. The errors in the rates and equilibrium coefficients are ϷϮ20% and ϷϮ35%, respectively, based on completely independent experiments with different starting mutant or wt samples (78).
Wild-type 17  hindrance from the ␥ 2 -CH 3 group of Val-68 allows the N⑀-H atom of the distal histidine to move closer to the bound CO, thus increasing the population of A 3 -like states.
In general, the aromatic 68 mutants have an increased fraction of protein molecules in substate A 3 compared with wildtype MbCO. In V68Y MbCO, A 3 is the dominant conformation, and intense recombination from B 0 is observed in this mutant at ϳ70 K (Figs. 4f and 7f). Surprisingly, the spectra of V68F and V68W MbCO are more similar to those of wild-type MbCO than those of V68M or V68Y MbCO, at least at low temperatures. The A substate ratios are in general dependent on temperature (46,68), and in V68F, more of the A 3 -like band (ϳ36%) occurs at room temperatures (59) than at 3 K (18%). The x-ray structure of V68F MbCO shows that the aromatic side chain lines the wall of the distal pocket ( Fig. 1) and, despite of its large size, does not come into direct contact with the bound ligand (25). Assuming a similar orientation for the indole side chain in V68W, there is more open space adjacent to the bound ligand and near His-64 (Fig. 1, right panel). Presumably, this lack of hindrance facilitates stronger hydrogen-bonding interactions between the distal histidine and the bound ligand, enhancing the populations of the A 3 state and the B 0 photoproduct.
Effects on the Spectra of CO in the Primary Docking Site B-After 1-s illumination at 3 K, photolyzed ligands are observed to universally settle into the initial docking site B as judged by the photoproduct bands ( Figs. 1 and 2). This observation underscores the role of site B as the most easily accessible state after photodissociation. There is only a single CO location in the x-ray structure of the primary photoproduct B both at low temperature and at short times (ϳ1 ns) (8 -10, 69). The FTIR spectrum, however, shows typically two distinct infrared bands, B 1 and B 2 , for CO at the primary docking site. This doublet feature is not unique for the B site. A survey of photoproducts of many different Mb mutants revealed that CO spectra of the different photoproduct sites frequently exhibit two bands (40,43,46,70). They arise from Stark splitting of the IR band; opposite orientations of the CO dipole interact differently with the internal electric field at the respective docking site (41). The photoproduct maps in Fig. 3 show that the presence of polar amino acids (V68T, V68S, V68M) has only a minor effect on the internal electrostatic field in the B site, as compared with the large effects exerted by these residues on the bound ligand. In V68T, the Stark splitting is unchanged with respect to wild type, but both B 1 and B 2 are shifted to higher frequencies. This result indicates an increase in the bond order of photodissociated CO, presumably as a result of changes in the electrostatic field in the heme pocket due to the Thr-OH. Residues Ser-68 and Met-68 lead to a slight increase of the separation between the B 1 and B 2 bands, which could be due to a slightly changed location of the distal histidine or direct effects from the mutant side chains.
Aromatic residues at position 68 (V68F, V68Y, V68W), however, increase the Stark splitting between B 1 and B 2 significantly. Fig. 1 shows that photolyzed ligands at site B are located in direct vicinity to the aromatic side chain. We have previously proposed that the introduction of aromatic -electron systems into the intrinsically hydrophobic distal cavity alters the local electric field markedly. This conclusion was drawn from FTIR investigations on mutants L29W (11,42,46) and L29Y/H64Q/T67R (40), where an aromatic amino acid was introduced at position 29. Additionally, the splitting of the B state bands in the aromatic mutants may also be enhanced by interactions with the N⑀-H atom of the distal histidine. The flexibility of the imidazole side chain is increased by the loss of hindrance by ␥ 2 -CH 3 in the naturally occurring Val-68 side chain (Fig. 1, left panel). In the structure of V68F MbCO, we see that the distal histidine is located further inside the pocket and thus closer to B state photoproducts. This interpretation is supported by the increased equilibrium population of the A 3 substates for these mutants, which indicates a closer approach of the His-64 side chain to the heme iron and bound ligand.
The TDS maps after brief illumination (Figs. 3 and 4) indicate recombination exclusively from the primary docking site B. At 3 K, ligands are photodissociated with a large excess of kinetic energy and trapped in one or the other of two metastable states, B 1 and B 2 (44). At ϳ20 K, however, thermal energy is sufficient for the population to readjust toward minimal free energy. The individual temperature of ligand reorientation increases in the sequence V68S Ͻ wild type Ͻ V68M Ͻ V68F Ͻ V68W Ͻ V68Y (Figs. 6 and 7). We suggest that the volume of side chain 68 sterically governs CO rotation. In this context, it  2 and CO binding at pH 7, 293 K for V68F, V68Y, and V68W mutants of sperm whale Mb. The parameters for wt and V68F were taken from Springer et al. (78). The rate coefficients for V68W were taken from Scott et al. (56) and Dou et al. (79) and re-measured for this work. Parameters for V68Y were measured for this work. The errors in the rates and equilibrium coefficients are ϷϮ20% and ϷϮ35%, respectively.  2 and MbCO apply to the same conformation, one that allows more rapid ligand entry and exit. The opposite assignment would indicate conformations with markedly different O 2 and CO equilibrium coefficients. The former assignment seems more logical since neither the V68F nor V68W mutation caused a dramatic change in K O 2 or K CO .  Table V. is remarkable that rotation at site B appears to be almost non-existent in V68T, even though the valine and threonine side chains are isosteric. The A state TDS map of V68T MbCO after brief illumination indicates extensive recombination in the 10 -15 K temperature region. Thus, recombination from state B competes effectively with reorientation of B 2 to B 1 . Smerdon et al. (26) observed more rapid geminate recombination in pig V68T MbO 2 and suggested that, after photodissociation, the His-64 side chain may rotate around the C␤-C␥ bond to interact with the Thr-68-OH instead of swinging inward over the iron atom (see Fig. 1). In the former case, the iron atom would remain less hindered, whereas in the latter case the imidazole side chain would hinder bond formation. A similar but less dramatic phenomenon is seen in the V68S A state TDS map (Fig. 3c); recombination occurs in part directly from B 2 at 18 K and from B 1 at 42 K.
In the aromatic mutants, the prominent decrease of the temperatures of maximum rebinding in the A 1 and A 3 conformations implies that a factor common to both substates is responsible for the lower barriers to recombination. Our results and those of Carver et al. (23), Quillin et al. (25), and Scott et al. (56) demonstrate that large residues at position 68 trap photodissociated ligands closer to the iron atom and inhibit ligand movement to more remote sites, particularly the Xe4 cavity. This effect enhances the geminate recombination rate in both the A 1 and A 3 conformations. In addition, the loss of hindrance by the ␥ 2 -methyl group of Val-68 also facilitates bond formation with the iron atom, regardless of the exact position of the distal histidine. Similar results were obtained in sol-gels of Val-68 mutants of sperm whale myoglobin (71).
Structural Interpretation of the B States-Lim et al. (44) have associated the B 1 photoproduct with a CO molecule in the B cavity of an A 1 -type protein conformation, with the ligand oxygen atom pointing toward the heme iron. B 2 was suggested to represent the opposite orientation based on the rates of appearance of the bands. Our results suggest the opposite orientation, with the C atom being closest to the iron in the B 1 substate and away from it in the B 2 substate. Thus, at ϳ20 K, CO in state B 2 must first re-orient to the B 1 conformer before rebinding can occur, because only the C atom can act as a Lewis base and react with the iron.
The evidence in favor of our interpretation is 3-fold. First, the proposed orientations explain why recombination occurs mainly from state B 1 and slowly or not at all from B 2 at cryogenic temperatures. Second, in the B 2 orientation, interactions between the atom of the photolyzed ligand and the N⑀-H atom of the distal histidine would favor a lower frequency resonance structure (greater fraction of OϭC:). In the B 1 state, the N⑀-H atom of His-64 would interact with the C atom of the ligand and stabilize the charged, triple bond resonance structure (-:CϵOϩ) giving rise to a higher C-O . This interpretation provides a simple explanation of the Stark splitting of the B 2 and B 1 states. Third, in the A 3 conformation, the distal histidine is closer to the bound ligand than in A 1 molecules (35) and gives rise to the high frequency B 0 photoproduct, suggesting that the ligand in B 0 is oriented as in B 1 , with the carbon atom pointing toward the heme iron. Because the distal histidine in the A 3 conformation is closer to the heme iron, it sterically hinders geminate rebinding and increases the temperature required to observe this process to ϳ70 K. The proximity of the N⑀-H atom of His-64 to the iron atom in the B 0 substate also increases the strength of electrostatic interactions with the C atom of the dissociated ligand. As a result, the C-O of the ligand in B 0 is both predicted and observed to be significantly higher (ϳ2150 cm Ϫ1 ) than that of the other photoproducts. The band corresponding to the opposite orientation of CO in B 0 has not yet been found. It was proposed to be hidden underneath B 1 at ϳ 2130 cm Ϫ1 (29,43). However, it could also be that the close proximity of the His-64 side chain to the O atom of bound CO in the A 3 conformation sterically enhances translation without rotation of the ligands after photodissociation. Consequently, the photoproduct for the A 3 conformation may contain only the orientation with the C atom pointing back to the iron atom.
Additional evidence of our interpretation of ligand orientation in the B substates and the importance of interactions with the distal histidine is provided by TDS experiments on H64L and H64V Mb mutants (70). For these mutants, the splitting between B 1 and B 2 photoproducts is almost absent, and rebinding occurs at very low temperatures because there is no steric hindrance by the distal histidine. Similarly, at low pH, wildtype MbCO adopts the A 0 conformation where the histidine has swung out of the distal pocket. Under these conditions, rapid recombination is observed at low temperatures (5). At higher temperatures, where protein motions are activated, the imidazole side chain plays a dynamic role. After ligand escape from the distal pocket, the His-64 side chain rotates deeper into the distal pocket as in deoxyMb and may further occlude the active site. Indeed, time-resolved x-ray structures of photolyzed wildtype MbCO crystals with nanosecond time resolution show that at ambient temperature, His-64 rotates to a position above the heme iron within about 4 ns after photolysis and is close enough to the dissociated ligand to influence its vibrational properties (69). A summary of these interpretations of the photoproduct states is given in Table I.
Access to Secondary Docking Sites-The FTIR spectra in Fig.  5 have indicated a markedly different effect of extended illumination on wild-type and V68(M/S/T) on the one hand and V68(F/Y/W) on the other hand. The A state TDS maps upon slow cooling (Figs. 6 and 7), and the temperature-dependent integrated absorbances differences calculated from the TDS maps (Fig. 8) show that rebinding in wild-type MbCO occurs at ϳ40, 80, and ϳ120 K in A 1 molecules and at ϳ70 and ϳ160 K in A 3 molecules. In accordance with the TDS plots after 1-s illumination at 3 K, we associate the contours peaking at the lowest temperatures with recombination from the initial docking site B.
For the V68F, V68Y, and V68W mutants, most of the ligands rebind from site B because the large aromatic side chains almost completely suppress ligand escape from the initial site B into more remote sites. Except for the appearance of a small additional band at ϳ2130 cm Ϫ1 , the TDS maps generated after 1-s illumination at 3 K are identical to those generated after prolonged illumination. From Fig. 1, it is easily seen why movement from B to C is possible in photolyzed wild-type MbCO, whereas in V68F, ligands are essentially confined at site B; the Xe4 site has been almost completely filled. Escape to Xe1 is also suppressed by the voluminous side chain at position 68, as initially suggested by Elber and Karplus (72). These authors considered residue 68 to be part of the gate that controls ligand migration toward the proximal side. In their studies on O 2 binding to Mb mutants V68W and V68F, Scott et al. (56) also found only minute geminate recombination from secondary sites.
In V68S, there is a very large subpopulation recombining at ϳ50 K. The subpopulation recombining in A 3 at 160 K is even larger than the one rebinding at ϳ70 K, indicating that migration away from the primary site B is unhindered. A comparison of the integrated A 1 absorbances after 1-s and extended illumination (Fig. 8a) shows that the subpopulation rebinding at ϳ50 K does not originate from site B, but rather from a more remote site. The comparatively low temperature for this process implies a lower barrier to return from secondary sites. Hence, a smaller side chain at position 68 not only enhances ligand escape to, but also ligand return from remote sites. Again, this can be readily interpreted from the structures in Fig. 1 since the side chain of residue 68 forms part of the gate to movement into the Xe4 cavity.
The photoproduct maps of wild-type, V68M, and V68T MbCO after prolonged illumination are governed by dominant negative contours at 2132 cm Ϫ1 that extend up to 150 K. These features indicate significant rebinding from remote C or D sites. Although the A state TDS maps reflect recombination of two and three subpopulations in this temperature range, an unambiguous characterization of individual secondary sites based on the photoproduct spectra is not feasible.
Geminate and Bimolecular Rebinding to the Polar Mutants at Room Temperature-The FTIR-TDS experiments have provided valuable information about the occupancy of secondary sites at cryogenic temperatures. It is clear from these results and the structure of V68F in Fig. 1 that ligand migration toward remote intermediate sites is governed by the volume of the residue 68 side chain. This steric influence of residue 68 on ligand migration within the protein is also seen in the geminate rebinding traces recorded after flash photolysis at ambient temperature ( Fig. 9).
In wild-type and mutants V68M, V68S, and V68T, secondary ligand storage sites within the Xe cavities are accessible, as seen from the slow-cool experiments (Fig. 8). In V68S, ϳ90% of the CO escape, an observation already reported by Sugimoto et al. (73). The small Ser-68 side chain reduces the steric hindrance at the heme iron and thus enhances the geminate process (17,74). At the same time, the polarity of the serine side chain can stabilize distal pocket water in deoxyMb, an effect that would inhibit ligand entry and non-covalent binding in the B site. These two effects appear to cancel and, therefore, the rate coefficient S (CO) and bimolecular coefficient kЈ CO are similar to those of wild-type MbCO.
For the isosteric V68T sperm whale mutant, rate coefficients S (CO) and kЈ CO are also similar to those of wild-type Mb. Comparable results were observed for pig V68T MbCO by Smerdon et al. (26). These authors also showed that the dissociation rate, k CO , had increased 4-fold. The same effect was seen in b-V67T Hb (62) and demonstrates that the Fe-CO bond is destabilized by the polar Thr-68 side chain. The values of S and kЈ O 2 for O 2 binding to V68T Mb, however, are about an order of magnitude smaller than those for wild-type Mb (Tables  III and IV). In 1993, Smerdon et al. (26) explained the significant decrease in kЈ O 2 for pig V68T Mb in terms of a hydrogenbonding network between the distal histidine, a water molecule present in the deoxy protein and the hydroxyl side chain of Thr-68. Disruption of this network to expel the water causes a ϳ10-fold decrease in the rate and equilibrium coefficient for ligand entry into the mutant protein. The rate-limiting step for bimolecular O 2 binding is given by kЈ SB (for N S Ͻ Ͻ 1). Consequently, the decrease in kЈ SB causes O 2 binding to become limited by movement into the protein (Equation 1). In the case of CO binding, k BA is still much smaller than the rate of escape from the protein, k BS , and as a result, kЈ CO for the V68T mutant, like that for wild-type Mb, is given by k BA ⅐(kЈ SB /k BS ). For bimolecular CO association, the decrease in kЈ SB /k BS due to stabilization of distal pocket water is compensated by the increase in k BA , as inferred from the lower recombination temperatures in the TDS map for the V68T mutant, and there is little net change in kЈ CO or S (see Tables III and IV).
Bimolecular recombination in V68M is biphasic, showing a fast component with CO and O 2 rate coefficients that are 4 -8fold lower and a second slow component ϳ50-fold lower than those of the wild-type protein. Similar biphasic time courses were observed for O 2 and CO association with V68Y Mb (Fig.  10). In this case, the O 2 dissociation curves also showed a very rapid and a very slow component. The structural causes of this heterogeneity are still unknown. Clearly, there must be widely different conformations of the methionine and tyrosine side chains that do not interconvert on millisecond to second time scales since the heterogeneous behavior occurs even at low ligand concentrations and in the O 2 dissociation reaction. Similar conformational heterogeneity under physiological conditions has been reported for the single L29Y and H64M mutants. This points to a characteristic problem of these side chains that are large and can be fixed in different positions by polar interactions (21,75).
Geminate and Bimolecular Rebinding to the Aromatic Mutants at Room Temperature-The CO-rebinding traces of the aromatic mutants V68F, V68Y, and V68W (Fig. 9b) show clearly that the fraction of ligands escaping into the solvent, N S (CO), is significantly reduced, to 0.61, 0.41, and 0.22, respectively. The effect is even more pronounced for O 2 ligands (Table  III). Three general mechanisms have been suggested for enhancement of geminate recombination (17,74): (i) trapping of ligands adjacent to the iron by packing the back of the distal pocket, (ii) reduction or elimination of steric hindrance at the heme iron, and (iii) enhancement of the iron reactivity by proximal effects. The phenylalanine, tyrosine, and tryptophan insertions in V68F, V68Y, and V68W are examples of the first two mechanisms. The voluminous side chains sequester photolyzed ligands at site B close to the heme iron, preventing movement into remote docking sites. Similar sequestering effects are seen when Ile-107 is replaced with larger amino acids that fill the back of the distal pocket (56,76).
Lack of secondary substitution on the ␤ carbon of the position 68 side chain increases the accessibility of the iron atom to rebinding by internal ligands (see Fig. 1, right). As a result, the V68F, V68Y, and V68W mutants show small values of N S . As shown in Tables III and V, the decreased bimolecular association rate coefficients and S values for these mutants scale with lower N S values. Investigations of wild-type and single mutants with selectively blocked xenon cavities have shown, however, that for CO rebinding, a small N S is associated with occluded secondary sites and a fast rate coefficient S , as the probability of incoming CO ligands to reside at site B, the only site from where recombination can occur, is enhanced (43). Extremely low bimolecular rebinding rates were observed for mutant L29W, owing to an occluded primary docking site B (42,56,77). For the V68F, V68Y, and V68W mutants we suggest that, immediately after the flash, all photolyzed CO ligands are recovered at site B. While most of them rebind directly from B because of the roadblock at position 68, a small fraction escapes into the solvent. The resulting protein relaxes into the deoxy conformation. The distal histidine side chain swings over the iron atom and brings a water molecule into the space between N⑀-H of His-64 and the face of the aromatic ring of the position 68 side chain (25). As a result there is little "free space" to capture ligands in the distal pocket of the deoxy forms of the Phe-68, Tyr-68, and Trp-68 mutants. The decrease of the bimolecular recombination rate coefficient correlates with the size of residue 68. Additional support for the steric influence of residue 68 on the initial docking site is based on the photoproduct TDS maps (Figs. 3 and 4). They reveal that ligand reorientation at the primary docking site B is opposed by the presence of voluminous side chains at position 68.
In 2001, Scott et al. (56) have argued that the decreased open space in the distal pocket is responsible for the slower bimolecular rates in mutants with bulky side chains reaching into the distal cavity, including the V68F and V68W mutants. They compared the distal pocket with a baseball glove that catches incoming ligands. If the glove is already partially filled, incoming ligands are not as readily trapped. Here, we have shown clearly that aromatic side chains at position 68 decrease rotational motion of ligands in site B (Fig. 4), prevent movement into the more remote Xe cavities (Figs. 7 and 8), and enhance interactions with the distal histidine in the photodissociated B states. All of these effects lower the fraction of escape by facilitating internal recombination and at the same time decrease the rates of ligand entry into the protein by reducing the free volume available to capture ligands in the distal pocket. Moreover, these aromatic side chains compromise access to the primary docking site, which is the key intermediate for bond formation to the heme iron (42,43). CONCLUSION The IR spectra of heme-bound and photodissociated CO provide convincing evidence that the CO ligand serves as an internal voltmeter. The distal histidine and aromatic amino acids in the vicinity of the bound or dissociated ligand exert the most pronounced effect on the electrostatic field surrounding the CO and, concomitantly, on the CO stretching frequencies. Similarities as well as differences between the O 2 and CO time traces can be explained by the interplay of size and polarity of residue 68. Geminate recombination and ligand movement in the protein interior are regulated primarily by steric interactions with the position 68 side chain, which regulates access to the iron atom and governs the size and interconnection of the B and C site cavities. In contrast, both the polarity and size of the position 68 side chain play major roles in regulating bimolecular ligand binding from the solvent and thermal dissociation of the iron-ligand bond.