Thr-E11 regulates O2 affinity in Cerebratulus lacteus mini-hemoglobin.

The mini-hemoglobin from Cerebratulus lacteus (CerHb) belongs to a class of globins containing the polar Tyr-B10/Gln-E7 amino acid pair that normally causes low rates of O2 dissociation and ultra-high O2 affinity, which suggest O2 sensing or NO scavenging functions. CerHb, however, has high rates of O2 dissociation (kO2 = 200-600 s(-1)) and moderate O2 affinity (KO2) approximately 1 microm(-1)) as a result of a third polar amino acid in its active site, Thr-E11. When Thr-E11 is replaced by Val, kO2 decreases 1000-fold and KO2 increases 130-fold at pH 7.0, 20 degrees C. The mutation also shifts the stretching frequencies of both heme-bound and photodissociated CO, indicating marked changes of the electrostatic field at the active site. The crystal structure of Thr-E11 --> Val CerHbO2 at 1.70 A resolution is almost identical to that of the wild-type protein (root mean square deviation of 0.12 A). The dramatic functional and spectral effects of the Thr-E11 --> Val mutation are due exclusively to changes in the hydrogen bonding network in the active site. Replacing Thr-E11 with Val "frees" the Tyr-B10 hydroxyl group to rotate toward and donate a strong hydrogen bond to the heme-bound ligand, causing a selective increase in O2 affinity, a decrease of the rate coefficient for O2 dissociation, a 40 cm(-1) decrease in nuCO of heme-bound CO, and an increase in ligand migration toward more remote intermediate sites.

Globins are found in all kingdoms of living organisms. Their functions have been the subject of active debate. In addition to O 2 transport and storage (1)(2)(3), several novel functions have been proposed recently, including control of NO levels in microorganisms and nerve tissue, O 2 sensing, and dehaloperoxidase activity (4 -8). Nerve tissue Hbs are found in both vertebrates and invertebrates. Neuroglobin is a recently discovered member of the globin family, whose in vivo function is still unknown (9 -14). It is expressed in specific regions of vertebrate brains, displays low sequence identity to conventional Hbs or Mbs, 1 and is characterized by a bis-His-Fe hexacoordinate heme structure (11,12,15,16). In contrast, the nerve tissue Hbs found in mollusc, annelid, arthropod, nemertean, and nematode species (17,18) appear to store and/or transport O 2 to support brain and axon function during temporary hypoxia (18 -21).
The nerve tissue and body wall Hbs of the nemertean worm Cerebratulus lacteus (CerHb) 2 are the smallest naturally occurring Hbs, composed of only 109 amino acids. Analysis of the three-dimensional structure of nerve tissue CerHb has shown that the typical 3-over-3 globin fold is edited markedly (22). The N-terminal A-helix is deleted; the GH region is extended; and the C-terminal H-helix is shortened. Both sequence and fold comparisons suggest that CerHb is equally distant from all known globin tertiary structures, supporting its identification with a new superfamily, the mini-Hbs (22). CerHb contains a large elongated tunnel in its interior. Ligands may enter and exit CerHb through this apolar tunnel between the E-and H-helices, whose solvent access is made easier by the absence of the A-helix, the presence of small residues lining the cavity, and the shorter spans of some of the individual helices (22). Similar apolar cavities/tunnels are observed in a variety of heme-containing enzymes that react with gases or apolar ligands (23).
The ligand-binding site of both body wall and nerve CerHb is highly unusual since it contains three polar residues, Tyr-B10, Gln-E7, and Thr-E11. Normally, Hbs that contain the Tyr-B10 and Gln-E7 side chains show enhanced O 2 affinity and mark-edly reduced rates of dissociation due to strong hydrogen bonds between these polar amino acid side chains and the hemebound ligand (24,25). The Tyr-B10 hydroxyl group and Gln-E7 N-⑀ atom in wild-type nerve CerHb are very close (Յ3.0 Å) to the heme-bound O 2 atoms, indicating significant electrostatic interactions (22) . In addition, the imidazole ring of the proximal histidine adopts a staggered orientation with respect to the pyrrole N atoms in CerHbO 2 , a conformation that allows inplane movement of the iron atom with minimal steric hindrance. Leghemoglobins have similar staggered proximal geometries and show very high affinities for all ligands (26,27). In contrast, the affinity of CerHb for O 2 is moderate (P 50 Ϸ0.6 torr), and the rate coefficient for O 2 dissociation is unexpectedly large, 200 -600 s Ϫ1 (Table I).
To resolve this apparent discrepancy between structure and observed functional properties, we have surveyed the effects of point mutations of the three polar amino acids in the active site of nerve CerHb (Tyr-B10, Gln-E7, and Thr-E11) on O 2 binding. The largest effects were observed when Thr-E11 was changed to a small, apolar amino acid like Val. The consequences of this replacement on structure and function were examined in detail by determining the crystal structure of the Thr-E11 3 Val CerHbO 2 mutant, measuring overall rates of O 2 and CO binding to wild-type and mutant CerHbs, and examining the IR spectral properties of the corresponding heme-bound and photodissociated CerHbCO complexes.

Measurement of Overall Rates of Ligand Association and Dissociation at Ambient Temperature-Wild-type and mutant recombinant Cer-
Hbs were expressed and purified as described previously using a synthetic gene with codon usage optimized for expression in Escherichia coli (22). The native body wall CerHb was isolated from whole animals and purified as described (18). Most of the recombinant nerve CerHb and also the body wall CerHb sample were isolated in the reduced state and used directly. Samples containing significant amounts of ferric CerHb were reduced by dithionite and run down a G-25 column to remove excess reducing agent. All experiments were carried out in 0.1 M phosphate buffer, pH 7.0, 2.0 mM EDTA, 20°C. The sample solutions were drawn into a gas-tight syringe and inserted into a sealed cuvette (1 mm path length), previously equilibrated with the appropriate O 2 partial pressure. O 2 association time courses were measured after complete laser photolysis of CerHbO 2 samples, using a 300-ns excitation pulse from a Phase-R model 2100B dye laser (28). Simple exponential rebinding time courses were observed for most of the samples, and kЈ O 2 was determined from the slope of plots of k obs versus [O 2 ]. The association rate coefficient for NO binding to wild-type CerHb was measured in the same way. Cuvettes were filled with degassed buffer, equilibrated with 1, 1/2, and 1/4 atm NO (2,000, 1,000, and 500 M, respectively), and ϳ100 M deoxy-CerHb was added. Bimolecular recombination was measured at 436 nm using the 300-ns dye laser for photoexcitation. The value of kЈ NO was taken from the slope of k obs versus [NO]. O 2 dissociation rate coefficients were determined by analyzing time courses for ligand replacement in a stopped-flow spectrometer. In these experiments, 5-10 M CerHbO 2 samples in buffer containing various concentrations of free O 2 were mixed with buffer equilibrated with 1 atm CO. The O 2 dissociation rate coefficients were calculated from where r obs is the observed first order replacement rate coefficient, and kЈ CO and kЈ O 2 are the association rate coefficients for CO and O 2 binding, which were determined independently (29).
CO association time courses were measured either in stopped-flow rapid mixing experiments, in which the reduced deoxygenated protein was mixed with various concentrations of CO, or in laser photolysis experiments where rebinding was followed after complete photodissociation. Rate coefficients for CO dissociation were determined by analyzing stopped-flow time courses, in which the heme-bound CO was displaced with high concentrations of NO (28,29). Equilibrium coefficients for O 2 and CO binding were calculated from the ratio of the association and dissociation rate coefficients.
Crystallization and Data Collection-The Thr-E11 3 Val CerHbO 2 mutant was crystallized by vapor diffusion techniques (protein concen-tration 27 mg/ml) under conditions matching those of the wild-type protein (22). Elongated prismatic crystals (about 0.05 ϫ 0.05 ϫ 0.2 mm 3 ) grew within 1 week. The crystals were stored in 2.8 M ammonium sulfate, 50 mM sodium acetate, pH 6.0, and transferred to the same solution, supplemented with 15% glycerol (v/v), immediately before data collection at 100 K. Thr-E11 3 Val CerHbO 2 crystals are isomorphous with those of the wild-type protein (orthorhombic space group P2 1 2 1 2 1 , with unit cell constants a ϭ 42.7 Å, b ϭ 43.4 Å, c ϭ 59.7 Å, one molecule per asymmetric unit). A diffraction data set was collected at the ESRF synchrotron facility (beam line ID14 -2, Grenoble, France) up to 1.70 Å resolution and processed using DENZO, SCALEPACK (30), and programs from the CCP4 suite (31). Crystallographic refinement was performed using the programs REFMAC (32) and O (33) for model building/inspection. The Thr-E11 3 Val CerHbO 2 model was refined to a final R-factor of 16.7% and R-free of 19.5%. Weak electron density was observed only for residues Lys-B15 and Lys-E5. Atomic coordinates and structure factors for the Thr-E11 3 Val CerHbO 2 mutant have been deposited with the Protein Data Bank (34), with accession codes 1v07 and r1v07sf, respectively.
Room Temperature FTIR Spectroscopy-Approximately 20 l of 2-3 mM CerHb were equilibrated with either 1 atm of CO or N 2 in stoppered Eppendorf tubes. One l of a 200 mM dithionite solution in 100 mM phosphate buffer, pH 7.0, was added to reduce any oxidized iron and to remove molecular oxygen. The tube was vortexed and spun in a microcentrifuge to remove any precipitate. With an airtight syringe, equilibrated with N 2 gas, the protein sample was rapidly added to a CaF 2 BioCell IR cuvette (5 mm thickness ϫ 50 mm diameter, separated by a 40-m spacer; BioTools, Inc., Edmonton, Alberta, Canada) to obtain a uniform, bubble-free film. The windows of the cuvette were quickly sealed. The cuvette was placed in the sample chamber of a Nicolet Nexus 470 FTIR spectrometer (Nicolet Instrument Corp., Madison, WI), which was purged with N 2 gas 1 h prior to and during data collection. Spectra were recorded at 1 cm Ϫ1 resolution in the 1800 -2100 cm Ϫ1 region. Up to 128 interferograms were collected for both the CerHbCO and deoxy-CerHb control samples. The final FTIR spectra were corrected for buffer and protein background by computing Cer-HbCO minus deoxy-CerHb difference spectra.
Low Temperature FTIR Spectroscopy-For the IR experiments at cryogenic temperatures, the lyophilized protein was dissolved at a concentration of ϳ20 mM in cryosolvent (75% glycerol, 25% 1.0 M potassium phosphate buffer (v/v), final pH 7.0), stirred under a CO atmosphere, and reduced with sodium dithionite. The sample solutions were kept between two CaF 2 windows (diameter 25.4 mm), separated by a 75-m thick Mylar washer, and sandwiched inside a block of oxygen-free high conductivity copper. This assembly was mounted on the cold finger of a closed cycle helium refrigerator (model SRDK-205AW, Sumitomo, Tokyo, Japan). A digital temperature controller (model 330, Lake Shore Cryotronics, Westerville, OH) allowed adjusting the temperature between 3 and 320 K. Photolysis was achieved by a continuous wave, frequency-doubled Nd-YAG laser (model Forte 530 -300, Laser Quantum, Manchester, UK), emitting ϳ300 milliwatt output power at 532 nm. IR transmission spectra were collected between 1800 and 2400 cm Ϫ1 , at a resolution of 2 cm Ϫ1 , using an FTIR spectrometer equipped with an InSb detector (IFS 66v/S, Bruker, Karlsruhe, Germany).
Fourier Transform Infrared Temperature Derivative Spectroscopy-FTIR-TDS is an experimental protocol designed to measure thermally activated rate processes with distributed barriers (35). In the first step, the sample is photodissociated by a specific illumination protocol that selectively populates the desired intermediate state(s). Subsequently, FTIR transmission spectra are taken every kelvin while the temperature is increased linearly in time (typically at a rate 5 mK/s) over a certain temperature interval. In the TDS analysis, absorbance difference spectra are calculated for successive temperatures. Frequently, the change in the spectral area of an infrared band that occurs during acquisition of two successive spectra can be taken as proportional to the change in the population of CO molecules contributing to the band. Population changes arise from ligand rebinding and ligand diffusion among different docking sites. Both rate processes are governed by thermal activation. The temperature ramp protocol ensures that they occur sequentially with respect to the height of the activation enthalpy barrier. For a simple two-state reaction, the temperature axis can be converted to an enthalpy axis, with the barrier height approximately proportional to the ramp temperature. The TDS data are presented as contour plots of the absorbance change on a surface spanned by the wave number and temperature axes, with solid (dotted) lines indicating increasing (decreasing) absorbance.
Flash Photolysis at Cryogenic Temperature-For low temperature flash photolysis experiments, dilute samples with a protein concentra-tion of ϳ10 M were prepared as described above. The sample was loaded in a 10 ϫ 10 ϫ 2.5 mm 3 polymethyl methacrylate cuvette in thermal contact with a home-built copper sample holder in a closed cycle helium cryostat (model 22, CTI Cryogenics, Mansfield, MA), equipped with a Lake Shore Cryotronics (Westerville, OH) model 330 digital temperature controller. Samples were photolyzed by a saturating 6-ns (full width at half-maximum) pulse from a frequency-doubled, Q-switched, Nd-YAG laser (model Surelite II, Continuum, Santa Clara, CA). Photolysis-induced optical absorbance changes were monitored in the Soret region, using light from a tungsten source (model A 1010, PTI, Brunswick, NJ). The wavelength of the monitoring beam was adjusted to 436 nm by a monochromator. Intensities were measured with a photomultiplier tube (model R5600U, Hamamatsu Corp., Middlesex, NJ) and 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. At least three single transients were averaged at each temperature.

O 2 Binding to Wild-type and Distal Pocket Mutants of
CerHb-In order to examine how CerHb is able to achieve an Mb-like O 2 affinity, we replaced Tyr-B10, Gln-E7, and Thr-E11 with apolar residues of similar size and surveyed the overall O 2 binding parameters of recombinant nerve CerHb. Typical time courses for O 2 binding and dissociation are shown in Fig. 1. The computed association, dissociation, and equilibrium coefficients for the series of single mutants are given in Table I. Rate coefficients for O 2 binding to body wall CerHb isolated from intact worms were determined as controls for comparison with the values for the recombinant nerve proteins. Samples of native nerve CerHb were not available.
Although nerve and body wall CerHbs have O 2 affinities similar to mammalian Mbs, their association and dissociation rate coefficients are both 10-fold larger than those of muscle Mbs (Table I). The value of kЈ O 2 ϭ 240 M Ϫ1 s Ϫ1 for wild-type nerve CerHb indicates an easily accessible distal pocket and suggests that the rate-limiting step for O 2 binding is movement from the solvent into the protein interior. To confirm this conclusion, we measured the bimolecular rate of NO binding to reduced wild-type CerHb and observed kЈ NO 2 ϭ 230 M Ϫ1 s Ϫ1 . Scott et al. (36) have shown that kЈ NO is equal to the bimolecular rate of ligand entry into all mutant Mbs that have pentacoordinate deoxygenated heme groups. Thus, the equivalence of kЈ NO 2 and kЈ O 2 indicates that O 2 binding to wild-type CerHb is diffusion controlled.
The rate coefficients for O 2 dissociation from wild-type nerve and native body wall CerHbs are also remarkably large (k O 2 ϭ 200 -600 s Ϫ1 , Table I) compared with most other O 2 -binding proteins containing polar groups in their distal pockets. Mutation of Gln-E7 in CerHb to either His or Leu causes only small changes (Ϸ2-fold) in k O 2 , kЈ O 2 , and K O 2 ( Table I). Replacement of Tyr-B10 by either Phe or Ala results in an ϳ2-fold decrease in kЈ O 2 , 2-fold increase in k O 2 , and ϳ4-fold decrease in K O 2 ( Table  I). The effects of these E7 and B10 mutations are strikingly small compared with the effect of the isosteric Thr-E11 to 3 Val mutation, which causes an 8-fold decrease in kЈ O 2 (Fig. 1a) and a spectacular 1,000-fold decrease in the rate of O 2 dissociation ( Fig. 1, b and c, and Table I). This single mutation converts CerHb from a protein with a modest O 2 affinity (P 50 Ϸ 0.6 torr) typical of mammalian Mbs and Hbs with O 2 storage/transport functions to an ultra-high affinity Hb (P 50 Ϸ 0.003 torr).
Overall CO Binding to Wild-type and Thr-E11 3 Val CerHb at Ambient Temperatures-Overall rate coefficients for CO binding to wild-type neural, native body wall and Thr-E11 3 Val CerHb are listed in Table II and compared with the corresponding parameters for CO binding to wild-type sperm whale Mb and native Ascaris Hb, which has Tyr-B10, Gln-E7, and  Ile-E11 residues in the distal pocket. The CO association rate coefficients for CerHb are 100-fold higher than those for other animal Hbs and Mbs, indicating that the iron atom is highly reactive and accessible in the wild-type protein. The Thr-E11 to 3 Val mutation causes ϳ10and ϳ7-fold decreases in kЈ CO 2 and k CO , respectively (Table II). The net result is little change in CO affinity, which is in striking contrast to the ϳ130-fold increase in O 2 affinity produced by the same mutation ( Table I).
Determination of the Crystal Structure of Thr-E11 3 Val CerHbO 2 -In order to examine the biophysical mechanism behind the dramatic effect of the Thr-E11 hydroxyl group on ligand binding, we determined the three-dimensional structure of Thr-E11 3 Val CerHbO 2 using difference Fourier methods, and we compared it with that of the wild-type protein (22). The final model was refined to 1.70 Å resolution. The mutant structure contains 109 residues (plus an extra N-terminal Met residue), 92 water molecules, one bound O 2 molecule, and one sulfate anion (R-factor ϭ 16.7%, R-free ϭ 19.5%), with ideal stereochemical parameters (37) (Table III).
Optimal superposition of the Thr-E11 3 Val mutant and the wild-type CerHbO 2 structures yielded an r.m.s.d. value of 0.12 Å (calculated for 109 C-␣ atom pairs). Thus, the two molecules are virtually structurally indistinguishable, despite their remarkably different O 2 binding properties. There are no significant changes (Յ0.05 Å) in the Fe-O 2 and Fe-N-⑀ (His-F8) coordination bond distances. The staggered azimuthal orientation of the proximal His-F8 side chain relative to the heme pyrrole N atoms is completely conserved, presumably because the imidazole ring makes energetically equivalent van der Waals contacts with residue Phe-G5 and forms a hydrogen bond to the Leu-F5 carbonyl O atom in both the wild-type and mutant proteins.
The only significant structural differences between the wildtype and mutant proteins are movements of the Tyr-B10 and Val-E11 side chains away from each other and a slight "inward" movement of the heme-bound O 2 molecule in the Thr-E11 3 Val mutant (Fig. 2). In wild-type CerHbO 2 , the distance between the O-atom of the Tyr-B10 side chain and the O-␥1 atom of the Thr-E11 hydroxyl group is only 2.59 Å, implying that a strong hydrogen bond is "holding" the Tyr-B10 side chain near the Thr-E11 helical position (Table IV). The Thr-E11 3 Val mutation results in rotations of the Tyr-B10 side chain about the C␣-C␤ and C␤-C␥ bonds, which shift the O-atom of Tyr-B10 roughly ϳ0.7 Å away from the C-␥2 atom of 3 Val-E11 and toward the heme-bound ligand ( Fig. 2 and Table  IV). The close approach of the Tyr side chain to the heme-bound ligand causes a slight bending of the Fe-O 2 complex.
The N-⑀ atom of the Gln-E7 side chain in wild-type CerHb is also close to the heme-bound ligand (Table IV). There appears to be a moderately strong hydrogen bond with the O(2) ligand atom and a much weaker interaction with the O(1) atom, based on N-⑀ 3 O distances of 2.61 and 3.35 Å, respectively (Table  IV). In the Thr-E11 3 Val CerHbO 2 mutant, the heme-bound ligand moves away from the Gln-E7 side chain, increasing the N-⑀ 3 O (2) distance to 2.80 Å. However, the Tyr-B10 hydroxyl O-atom moves to within ϳ3 Å of the amide side chain and may be competing with the ligand for the N-⑀ proton.
In wild-type CerHbO 2 , the Thr-E11 O-␥1 atom is 3.12 Å away from the main chain O atom of the Gln-E7, the amino acid that is one turn toward the N-terminal of the E-helix (Table IV). This observation suggests strongly that the Thr-E11 hydroxyl group is donating a proton to the carbonyl O atom. In the Thr-E11 3 Val CerHbO 2 mutant, the apolar mutant side chain rotates in the opposite direction, downward and toward the heme, causing the distance between the Gln-E7 carbonyl O  atom and the C-␥2 atom to increase by ϳ0.3 Å (Table IV and Fig. 2). The hydrogen bond between the Thr-E11 hydroxyl group and the Gln-E7 carbonyl defines the direction of the electrostatic interaction with the Tyr-B10 side chain (see arrows in Fig. 2). The nonbonded electrons of the Thr-E11 O-␥1 atom are pointing toward the Tyr-B10 side chain and accept a proton from the phenolic O-atom. This hydrogen bond appears to be strong since the distance between the oxygen atoms is 2.59 Å (Table  IV). Donation of the phenolic proton to the Thr-E11 hydroxyl group directs the nonbonded electrons of the O-atom of Tyr-B10 toward the heme-bound ligand, which, in the case of hemebound O 2 , also has a partial negative charge. This electrostatic repulsion provides a mechanistic explanation for the much lower O 2 affinity and much higher O 2 dissociation rate coefficient in the wild-type protein as compared with the Thr-E11 3 Val mutant, where the phenolic proton of Tyr-B10 can be donated to the heme-bound ligand and stabilize it. At 298 K, the FTIR spectrum of wild-type CerHbCO at pH 7.0 shows a single peak of heme-bound CO at 1979 cm Ϫ1 , which is indicative of a negative charge close to the bound ligand (Fig.  3). This result supports our interpretation of the direction of the electrostatic interactions between the side chains of Tyr-B10 and Thr-E11 and the main chain carbonyl of Gln-E7 in wild-type CerHb. The nonbonded electrons of the Tyr-B10 hydroxyl group are pointed toward the heme-bound CO, stabilizing the Fe-Ϫ C'O ϩ tautomer. Replacing Thr-E11 with either Ala or Val causes a dramatic 40 cm Ϫ1 decrease in CO to 1930 -1940 cm Ϫ1 , indicating a reversal of the electric field (Fig.  3). This shift correlates with the movement of the Tyr-B10 side chain and rotation of the phenolic hydrogen toward the hemebound ligand, enhancing the fraction of FeϭCϭO tautomer.
The Tyr-B10 to Phe and Ala mutations cause a much smaller 8 -15 cm Ϫ1 decrease in CO , due to the loss of the negative charge of the nonbonded electrons of the phenolic O-atom. The Gln-E7 to Leu mutation results in an upshift to 1990 cm Ϫ1 . The Gln-E7 to His mutation has the opposite effect; it decreases CO to 1962 cm Ϫ1 . These observations indicate that, in wildtype CerHb, the hydrogen bond donated by Gln-E7 to the heme-bound ligand partially compensates for the electric field created by the negative partial charge of Tyr-B10. In the Gln-E7 3 Leu CerHbCO mutant, hydrogen bonding by the amide N-⑀ proton is absent, and CO shifts to higher frequency. In the His-E7 mutant, a stronger hydrogen bond to the hemebound CO occurs, shifting CO to lower frequencies.
FTIR Absorbance Difference Spectra at 3 K-Extensive studies on wild-type sperm whale Mb and a large number of single mutants have shown that ligands sample several transient docking sites on their way into and out of the protein (36,(43)(44)(45)(46)(47)(48)(49)(50). At cryogenic temperatures, the ligands can be trapped selectively at these intermediate sites. The rates and the extent of ligand migration among these docking sites at cryogenic temperatures provide clues about the physiological ligand binding process. To survey the existence of well defined transient docking sites in CerHb, FTIR absorbance difference spectra of wild-type and Thr-E11 3 Val CerHbCO were recorded after a 1-s illumination at 3 K to disrupt the bond between the heme iron and CO (Fig. 4). These spectra consist of a contribution in the spectral region of heme-bound CO that represents the fraction of iron-ligand complexes or A states that can be  and proximal sites (lower part, hosting His-F8). Short fragments of the Band E-helices are displayed as cyan ribbons. The heme is shown in red. The relevant residues are displayed with light gray bonds, using black labels for the wild-type CerHb structure. Structural modifications observed in the Thr-E11 3 Val CerHb mutant are highlighted by displaying residues Tyr-B10 and 3 Val E11 in orange. The dashed arrows indicate hydrogen bond donation in the wild-type protein (blue arrows) and in the Thr-E11 3 Val mutant (orange arrow). The O 2 molecule is displayed at the sixth coordination site of the heme iron as a purple diatomic species (partly covered by arrows). Lys-E10 is electrostatically coupled to both heme propionates and contributes to the shielding of the distal heme pocket. photolyzed (negative bands in Fig. 4). Wild-type CerHbCO shows a single A state band at ϳ1982 cm Ϫ1 , indicating a structurally homogeneous binding site with a negative partial charge near the CO oxygen, as was observed at room temperature (Fig. 3).
The A substate bands are connected with much weaker bands between 2100 and 2160 cm Ϫ1 representative of photolyzed CO. The stretch bands of the photoproducts show surprisingly narrow line shapes, indicative of homogeneous environments and minimal orientation disorders of the CO molecules in the intermediate sites. By analogy to the A state bands, the stretching frequencies, CO , of dissociated CO are also affected by the local environment, mainly through Stark effects of the local electric field acting on the CO dipole (43)(44)(45). In the photoproduct spectrum of wild-type CerHb, only one major CO band is observed at 2125 cm Ϫ1 , with a very minor (Յ5%) component visible at 2134 cm Ϫ1 . This result suggests that either there is only one major photoproduct with a single CO orientation or the Stark effects are too small to cause differences in the CO stretch frequency of the different CO conformers.
Cooling under continuous illumination enables photodissociated ligands to populate additional transient docking sites with higher recombination barriers (51). In sperm whale Mb, ligands have been recovered in internal hydrophobic pockets called xenon cavities after their ability to bind large noble gases (43-45, 49, 50, 52, 53). In CerHbCO, ligands can also migrate away from the initial docking site as is evident from the pronounced changes in the photoproduct spectrum after extended illumination (Fig. 4). The major photoproduct band is shifted from 2125 to 2122 cm Ϫ1 , and the minor band at 2134 cm Ϫ1 is significantly enhanced.
Replacement of Thr-E11 with Val gives rise to a dominant CO band at 1938 cm Ϫ1 and a minor band at 1955 cm Ϫ1 for heme-bound CO (Fig. 4). In the spectral region of photolyzed CO, two major bands are observed at 2123 and 2142 cm Ϫ1 , and two minor bands at intermediate positions (2131 and 2134 cm Ϫ1 ) appear as a broad feature. The existence of four peaks implies that ligands populate more than one intermediate site even after only brief illumination at 3 K. Extended illumination markedly enhances the absorbance of the bands centered at 2131 and 2134 cm Ϫ1 , but small fractions of the other two bands remain. Thus, the features in the 2131-2134 cm Ϫ1 region appear to represent ligands at sites that are further away from the iron atom or have a higher energy barrier against recombination. TDS on Wild-type CerHbCO-In TDS experiments, populations of the photodissociated ligands can be sorted according to the temperature at which they rebind to the iron atom. This temperature is a measure of the enthalpy barrier to geminate recombination. The method also helps to establish the number of intermediate docking sites and their order of interconversion (35).
In a first experiment, the absorbance changes in the bands of heme-bound CO were monitored as a function of temperature after wild-type CerHbCO was photolyzed at 3 K with a brief 1-s illumination (Fig. 5a). CO rebinding started immediately, even at 3 K, as seen from the prominent contours centered at ϳ1982 cm Ϫ1 . This result demonstrates unambiguously that the majority of the photodissociated ligands have very low enthalpy barriers against recombination at the iron. Less intense contours at ϳ1938 cm Ϫ1 arise from the 13 C 16 O isotope rebinding in the same heme-bound conformation.
The absorbance changes in the photoproduct bands are shown in Fig. 5b. The decrease in the major broad feature centered at 2125 cm Ϫ1 is associated with ligand rebinding to the heme iron from the initial docking site B and correlates directly with the absorbance increase at ϳ5 K in the dominant A state band (Fig. 5a). It can be resolved into two narrower bands with peaks at 2122 and 2126 cm Ϫ1 (Fig. 5b). A small solid contour at 2134 cm Ϫ1 and ϳ20 K indicates thermally induced ligand migration to an additional site, which is designated state C. Ligands rebind from this state at significantly higher temperatures, ϳ50 K, than from the initial B states. This process is seen as the peak in the photoproduct map at ϳ2134 cm Ϫ1 .
This pattern is quite different from that in sperm whale Mb (44), where recombination peaks at ϳ50 K for the A 1 ( CO ϭ 1945 cm Ϫ1 ) and at ϳ80 K for the A 3 conformer ( CO ϭ 1933 cm Ϫ1 ). Moreover, there are two spectroscopically well separated B state bands, B 1 at 2131 cm Ϫ1 and B 2 at 2119 cm Ϫ1 (Table II). Conversion of B 2 to B 1 occurs in the 10 -30 K temperature range. Rebinding in A 1 occurs exclusively from B 1 at 50 K. A second TDS experiment was performed after wild-type CerHbCO was slowly cooled from 160 to 3 K under constant illumination (Fig. 6, a and b). At 160 K, all heme-bound CO molecules are photolyzed. At these high temperatures, ligands may rebind but will be photolyzed again. Because the temperature is ramped down, rebinding will be slowed continuously. Ligands will eventually have insufficient thermal energy to overcome the barriers against recombination and become trapped in more remote sites, each of which will have a characteristic distribution of enthalpy barriers g(H BA ).
The resulting TDS maps of wild-type CerHb differ markedly from the ones generated after brief illumination at 3 K. Instead of rebinding readily at ϳ5 K, most of the dissociated CO molecules rebind at much higher temperature, ϳ80 K. Small subpopulations still show low barriers with recombination peaks at ϳ5 and ϳ40 K. In the photoproduct map (Fig. 6b), there are no features at low temperatures corresponding to rebinding from the initial docking site. Instead, absorbance decreases in the region associated with state C (2131-2134 cm Ϫ1 ) occur at ϳ40 K. Solid contours in this same temperature range imply either (i) that the photolyzed ligands do not recombine directly from intermediate state C, but migrate to yet another intermediate site D where they finally rebind at ϳ80 K, or (ii) ligand reorientation at site D. Recombination from state D is indicated by the feature centered at 2123 cm Ϫ1 and 80 K.
After prolonged illumination of sperm whale MbCO, many key features of the 1-s illumination photoproduct map are retained (44). "New" solid and dotted contours indicate ligand migration into and recombination from the Xe-4 and Xe-1 pockets (states C and D), but the extent of the migration toward more remote sites is much less than that seen for CerHb.
Temperature Derivative Spectroscopy on Thr-E11 3 Val Cer-HbCO-The A state contour maps of the Thr-E11 3 Val Cer-HbCO mutant after both short and prolonged illumination are much more complex than those of the wild-type protein (Figs. 5, c and d, and Fig. 6, c and d). The mutant shows a very broad band of heme-bound CO centered at 1938 cm Ϫ1 and minor features in the 1950 -1960 cm Ϫ1 region, indicating multiple A state conformations. After a 1-s illumination, rebinding to form the dominant substate is maximal at ϳ30 K. In the minor A states (1955 and 1968 cm Ϫ1 ), recombination occurs maximally at 70 and 5 K, respectively.
The spectral features in the photoproduct map of the Thr-E11 3 Val mutant are much more dispersed (2115-2145 cm Ϫ1 ) than those for wild-type CerHb. The large solid contours at 2134 cm Ϫ1 and ϳ20 K indicate significant thermally induced migration from the two initial photoproducts, with peaks at 2123 and 2145 cm Ϫ1 , to a new state, most likely the same state C seen in the wild-type protein (Fig. 5b). Rebinding from the secondary site C is seen as dotted contours centered at 2134 cm Ϫ1 and peaking at ϳ50 K (Fig. 5d).
The A state contour plot of the Thr-E11 3 Val CerHb mutant after extended illumination shows that the dominant A state subpopulation still forms maximally at ϳ40 K (Fig. 6c). Only a few ligands have migrated to more remote sites. Rebinding from these sites occurs at markedly higher temperatures (ϳ90 and ϳ150 K, Fig. 6c). The photoproduct map is now dominated by contours at 2134 cm Ϫ1 , representing rebinding from state C that occurs maximally at ϳ50 K (Fig. 6d). However, the exchange seen in the 1-s illumination experiment for the B to C state transition is absent (Fig. 5d and Fig. 6d).
Flash Photolysis at Cryogenic Temperatures-To investigate the enthalpy barriers against ligand rebinding and migration

FIG. 5. TDS contour plots of CO rebinding in wt CerHb and Thr-E11 3 Val
CerHb after 1-s illumination at 3 K. a and c, absorbance changes monitored in the IR bands of heme-bound CO. b and d, absorbance changes monitored in the IR bands of photolyzed CO. Solid (dotted) lines indicate an absorbance increase (decrease). Contours are spaced logarithmically. more quantitatively, laser photolysis experiments were performed at cryogenic temperatures, and absorbance changes for geminate recombination in wild-type and Thr-E11 3 Val CerHb were monitored at 436 nm (Fig. 7). The time courses between 40 and 180 K were fitted with a two-state model that invokes a static distribution of activation enthalpy barriers, g(H), between bound and photoproduct state. The fraction of heme groups that is still unligated at time t after the flash, N(t), is given by Equation 1, The measured absorbance change, ⌬A(t), is taken to be proportional to N(t). For a thermally activated barrier crossing, the dependence of the rebinding rate coefficient k(H,T) on temperature T is given by the transition state expression shown in Equation 2, with pre-exponential constant A, reference temperature T 0 set to 100 K, and the universal gas constant R. In the calculations, the ⌫ distribution (Equation 3) was chosen as a model function for g(H) (54,55).
For wild-type CerHbCO, global nonlinear least squares fit-ting of the experimental data yielded a pre-exponential A ϭ 10 7.9 s Ϫ1 , with the g(H) distribution peaking at 2.2 Ϯ 0.2 kJ/mol (␣ ϭ 1.15 mol/kJ; H min ϭ 0.5 kJ/mol, see Table II). The data, however, deviate from the calculated kinetic traces starting at 60 K. At times longer than 100 ms, recombination is slower than predicted. Based on the TDS data, we propose that a fraction of the photodissociated ligands migrates to state C during the initial recombination phase and that the slower geminate phase represents ligand recombination from this more remote site. By contrast, the experimental traces for the Thr-E11 3 Val CerHbCO mutant in the 60 -180 K range (Fig.  7b) can be fitted with a single g(H) distribution, with parameters A ϭ 10 8.5 s Ϫ1 and H peak ϭ 5.77 Ϯ 0.2 kJ/mol (␣ ϭ 0.82 mol/kJ; H min ϭ 0.5 kJ/mol, see Table II). Only the experimental data at 40 K and below deviate significantly from the fit. At these very low temperatures, recombination is faster than predicted, due to quantum-mechanical tunneling of the CO (55,56). The markedly increased enthalpic barrier in the Thr-E11 3 Val mutant compared with that in the wild-type CerHb implies that rebinding probably occurs exclusively from intermediate site C.

DISCUSSION
Electrostatic Regulation of O 2 Affinity-Our structural interpretation of how Thr-E11 decreases the O 2 affinity in wild-type CerHb is shown in Fig. 2. The Thr-E11 hydroxyl group donates its proton to the main chain carbonyl O atom of Gln-E7. A similar interaction is observed in the Val-E11 to Thr mutant of pig Mb (57), where the Thr-E11 side chain is rotated about the C␣-C␤ bond, upward and away from the heme, bringing the O-␥1 atom of Thr-E11 0.5 Å closer to the His-E7 carbonyl O atom. In fact, this hydrogen bonding pattern is seen for most Thr side chains present in ␣-helices, particularly when the face containing the hydroxyl group is located in an apolar environment (i.e. in membranes or facing the interior of a protein (58)). The nonbonded electrons of the Thr-E11 O-␥1 atom redirect the phenolic hydrogen of Tyr-B10 away from the bound ligand to form a strong hydrogen bond between the two amino acid side chains. This interaction "holds" Tyr-B10 next to the E-helix. Instead of donating a favorable hydrogen bond to the hemebound ligand, the nonbonded electron pairs of the Tyr-B10 Oatom point directly toward the ligand. The resulting electric field destabilizes the heme-bound O 2 by electrostatic repulsion, accounting for the large dissociation rate coefficient and moderate K O 2 value, despite the favorable proximal geometry and the hydrogen bonding interaction with the amide side chain of Gln-E7. As shown in Fig. 2 and Table IV, the Thr-E11 side chain in wild-type CerHbO 2 cannot interact directly with the heme-bound ligand (distance Ն4.3 Å) or strongly with the Gln-E7 side chain (distance Յ3.4 Å). These distances do not change in the Thr-E11 3 Val mutant. This geometry is significantly different from that in mammalian Mbs, where both the C-␥2 atom of the naturally occurring Val-E11 side chain and the O-␥1 atom in Thr-E11 mutants are within 3 Å of the heme-bound ligand and interact directly and unfavorably with it (57,59).
The dramatic functional effects caused by replacing Thr-E11 with Val confirm the proposed mechanism for regulating O 2 affinity. This isosteric mutation causes a 0.7-Å movement of the Tyr-B10 hydroxyl group away from the Val-E11 side chain and a 0.3-Å movement of the C-␥2 atom of Val-E11 away from the main chain carbonyl O atom of Gln-E7 ( Fig. 2 and Table  IV). Both movements reflect disruption of the hydrogen bonding network that is present in the wild-type CerHb due to the Thr-E11 hydroxyl group. In the Thr-E11 3 Val CerHbO 2 mutant, the Tyr-B10 group is "free" to donate a strong hydrogen bond to bound ligands. This favorable interaction is manifested The conclusion that the Thr-E11 to Val mutation causes a reversal of the electric field at the ligand-binding site is strongly supported by the FTIR spectra of the corresponding CerHbCO complexes (Figs. 3 and 4 and Table II). At both ambient and low temperatures, heme-bound CO in wild-type CerHb shows a narrow band centered at ϳ1980 cm Ϫ1 , a frequency that is only seen in model heme compounds and mutant sperm whale Mbs that have negative partial charges adjacent to the heme-bound ligand (38 -42). The Thr-E11 3 Val mutation causes a shift of the CO band to 1938 cm Ϫ1 , a frequency characteristic of model compounds and Mbs with strong hydrogen bonds between the heme-bound ligand and surrounding amino acid side chains, usually His-E7, Gln-E7, Tyr-B10, and/or Asn-E11 (38 -42).
Thr-E11 and Ligand Selectivity-The effects of the Thr-E11 to Val mutation on the functional properties of CerHb provide dramatic confirmation of the theory that distal electrostatic interactions, and not steric hindrance, govern ligand selectivity in heme proteins. Despite the change in the electrostatic field, the Thr-E11 to Val mutation causes little net change in the overall affinity for CO because the FeCO complex is inherently neutral (Table II). In contrast, K O 2 increases 130-fold due to the change from the unfavorable electrostatic repulsion between the bound ligand and the nonbonded electrons of the Tyr-B10 O-atom to the formation of a strong hydrogen bond between the phenolic hydroxyl group and the highly polar FeO 2 complex. Thus, the Thr-E11 to Val mutation changes CerHb from a protein that favors CO binding by a factor of ϳ400 to a mutant protein that has roughly equal affinity for CO and O 2 (M values in Table II). These marked changes occur with little or no alteration of protein structure and heme-ligand coordination geometry (Fig. 2). Similar selective enhancement of the O 2 affinity is seen when hydrogen bonding between Tyr-B10 and heme-bound O 2 occurs naturally (Table II,  Regulation of the Overall Rates of Ligand Binding-The formation of strong hydrogen bonds with heme-bound O 2 requires side chain atoms of adjacent polar amino acids to be within 2.5-3.0 Å of the sixth coordination position of the heme iron. This close proximity often reduces the rate coefficient for ligand association because access to the iron atom is hindered in the dissociated "deoxy" state by the polar side chains, which either move closer to the iron atom or bind solvent water molecules. Thus, in mammalian Mbs, mutation of the distal His-E7 to apolar amino acids results in ϳ5to 60-fold increases in the association rate coefficients for O 2 and CO binding, even though the O 2 affinity decreases markedly (60). In wild-type CerHb, the opposite situation occurs. The Tyr-B10 side chain is held next to the E-helix by hydrogen bonding to Thr-E11, preventing the aromatic side chain from swinging toward the heme iron, accounting in part for the large absolute values of kЈ O 2 and kЈ CO . When Thr-E11 is replaced by Val, however, the Tyr-B10 side chain moves freely toward the binding site and the phenolic hydroxyl group inhibits access to the heme iron atom in the unliganded state. This movement gives rise to the 10-fold decreases in both kЈ O 2 and kЈ CO 2 and the marked increase in the enthalpic barrier against geminate recombination (Tables I and II). The large absolute values of kЈ O 2 and kЈ CO 2 of wild-type CerHb are also the result of the staggered orientation of the proximal imidazole base, which reduces the barrier to in-plane movement of the iron atom, and the ease of ligand entry into the protein, presumably through the apolar interior tunnel.
Structural Interpretations of the Photoproduct States and Geminate Recombination-As shown in Figs. 3-7, altering the hydrogen bonding network between Tyr-B10, Thr-E11, and the carbonyl O atom of Gln-E7 has profound effects on the kinetic and spectral properties of photodissociated CO trapped in the protein. In wild-type Mb and most mutants that retain His-E7, the initial photoproduct generated by short illumination at 3 K exhibits two IR bands. Lim et al. (61) have investigated the nature of ligand motion within Mb by measuring femtosecond time-resolved infrared spectra of photolyzed CO. They observed two trajectories for the CO ligands into the initial docking site B, which can be distinguished kinetically. The trajectory that leads to the faster rebinding B state was assigned to CO molecules that translate directly into the docking site with the carbon atom still pointing toward the heme iron. The other trajectory involves rotation of the ligand molecule into the initial docking site with the O atom pointing back toward the iron, explaining why it cannot rebind without first rotating into the other reactive B state orientation. These two conformers are readily distinguished spectroscopically, due primarily to interactions with the N-⑀ proton of His-E7 that cause a Stark splitting into two photoproduct bands. This electrostatic interaction enhances the triple bond character of the B 1 substate with the C atom pointing toward the iron atom and His-E7, giving rise to the CO band at 2131 cm Ϫ1 . In contrast, interaction with His-E7 decreases the bond order of CO in the B 2 substate where the O atom is pointing back toward the iron, resulting in the CO band at 2119 cm Ϫ1 (43). When the distal His-E7 in Mb is replaced with Val or Leu, the splitting is markedly reduced or gone, and the barriers to recombination are reduced significantly.
In wild-type CerHb, the Tyr-B10 side chain is held away from the iron and the photodissociated ligands. As a result, there is little Stark splitting of the bands for the two CO orientations in the initial docking site B and little or no steric barrier against rebinding. In the TDS experiments after 1-s illumination, the T max value for rebinding to wild-type CerHb is very low, ϳ5 K compared with ϳ50 K for wild-type sperm whale Mb, and the peak value of the enthalpy barrier against geminate rebinding from site B is only ϳ2 kJ/mol compared with ϳ10 kJ/mol for wild-type sperm whale Mb (Table II). Prolonged illumination ("pumping") of CerHbCO is required to trap photodissociated CO molecules at more remote sites C and D because rebinding from the initial state B is so fast (Fig. 6b).
However, when Thr-E11 is replaced by Val, the Tyr-B10 side chain rotates back toward the iron, interacts with the photodissociated CO, causes marked splitting of the B state peaks (Fig. 4), and creates a large steric barrier to rebinding. As a consequence, the ligands that are initially trapped at site B will not rebind as readily to the heme iron, which allows them to escape to the more remote site C. This migration is indicated by the prominent positive peak in the photoproduct map of the mutant (Fig. 5d). Very similar features have been observed for Mb triple mutant YQR (Leu-B10 3 Tyr/His-64 3 Gln/ Thr-67 3 Arg) (62). The photoproduct map of Thr-E11 3 Val CerHb after prolonged illumination looks deceivingly simple compared with the one of the wild-type CerHb (Fig. 6, b versus  d). There is a single broad feature centered at ϳ50 K and 2134 cm Ϫ1 indicating CO rebinding exclusively from the C site. This assignment is supported by the flash photolysis experiments at cryogenic temperatures. However, the A state map indicates additional rebinding at higher temperatures (i.e. the peak at 150 K, 1938 cm Ϫ1 ) (Fig. 6c), presumably from ligands dispersed within the apolar channel.
Regardless of the exact structural interpretation, the prolonged illumination experiments show that ligands are trapped more readily at remote sites in both wild-type and mutant CerHb than in sperm whale Mb, despite the higher barrier against recombination in the mammalian protein. Thus, the remote sites in the apolar channel in CerHb almost certainly play a key role in facilitating ligand capture and entry into the distal pocket. However, more mutagenesis work is needed to define the exact role of the channel and to establish a detailed kinetic mechanism for ligand association in CerHb.
Conclusions-Although a mini-Hb, CerHb is part of a series of naturally occurring globins with Gln and Tyr at the E7 and B10 positions. These proteins include bacterial Hbs, flavohemoglobins, and invertebrate Hbs, with the most well studied examples being Vitreoscilla Hb (63-66), E. coli (HMP) flavohemoglobin (67)(68)(69)(70)(71)(72), and Ascaris Hb (73)(74)(75). The functions of these protein are still controversial and may involve O 2 sensing, NO dioxygenation, O 2 scavenging, and peroxidase activity. Most researchers no longer believe that these hemoglobins serve as O 2 storage proteins or transporters. All of them appear to have high O 2 affinities and low O 2 dissociation rate coefficients, when these parameters are measured directly (65,67,69,75). Ascaris Hb exhibits the most extreme behavior, showing a P 50 for O 2 binding of ϳ0.002 torr and k O 2 ϭ 0.004 s Ϫ1 (76,77), indicating that this protein almost certainly functions as an O 2 scavenger. In contrast, the nerve and body wall CerHbs have moderate Mb-like O 2 affinities, P 50 Ϸ0.6 torr, very large dissociation rate coefficients, k O 2 ϭ 200 -600 s Ϫ1 , and a clearly defined O 2 storage function to allow neuronal and muscle activity under hypoxic conditions (18).
Despite these marked differences in physiological function, the active sites of Ascaris and C. lacteus Hbs are structurally very similar. Both proteins show a staggered orientation of the proximal His-F8 side chain with respect to the pyrrole N atoms of the porphyrin ring. This proximal geometry is characteristic of a highly reactive iron atom. Short distances between the side chains of Gln-E7 and Tyr-B10 and the heme-bound ligand indicate strong electrostatic interactions (24,73). The key difference is the presence of the polar Thr-E11 side chain in the active site of CerHb instead of the apolar Ile-E11 side chain in Ascaris Hb. The mutagenesis results show that this single amino acid replacement is sufficient to explain the remarkably different physiological properties of the two Hbs. The single Thr-E11 3 Val point mutation can convert CerHb from an O 2 storage protein to an O 2 scavenger analogous to Ascaris Hb and vice versa.