A leucine residue "Gates" solvent but not O2 access to the binding pocket of phascolopsis gouldii hemerythrin.

A leucine residue, Leu-98, lines the O(2)-binding pocket in all known hemerythrins. Leu-98 in recombinant Phascolopsis gouldii hemerythrin, was mutated to several other residues of varying sizes (Ala, Val), polarities (Thr, Asp, Asn), and aromaticities (Phe, Tyr, Trp). UV-visible and resonance Raman spectra showed that the di-iron sites in these L98X Hrs are very similar to those in the wild type protein, and several of the L98X hemerythrins formed stable oxy adducts. Despite the apparently tight packing in the pocket, all of the L98X Hrs except for L98W, had second order O(2) association rate constants within a factor of 3 of the wild type value. Similarly, the O(2) dissociation rate constant was essentially unaffected by substitutions of larger (Phe) or smaller (Val, Thr) residues for Leu-98. L98Y Hr showed a 170-fold decrease in the O(2) dissociation rate constant and a large D(2)O effect on this rate, which are attributed to a hydrogen-bonding interaction between the Tyr-98 hydroxyl and the bound O(2). Significant increases in autoxidation rates were observed for all of the L98X Hrs other than X = Tyr. These increases in autoxidation rates are attributed to increased solvent access to the binding pocket caused by inefficient packing (Phe), smaller size (Val, Ala), or increased polarity (Thr, Asp, Asn) of the residue 98 side chain. A leucine at position 98 appears to have the optimal size, shape, and hydrophobicity for inhibition of solvent access. Thus, "gating" of small molecule access to the binding pocket of Hr by Leu-98 is not evident for O(2), but is evident for solvent.

A leucine residue, Leu-98, lines the O 2 -binding pocket in all known hemerythrins. Leu-98 in recombinant Phascolopsis gouldii hemerythrin, was mutated to several other residues of varying sizes (Ala, Val), polarities (Thr, Asp, Asn), and aromaticities (Phe, Tyr, Trp). UVvisible and resonance Raman spectra showed that the di-iron sites in these L98X Hrs are very similar to those in the wild type protein, and several of the L98X hemerythrins formed stable oxy adducts. Despite the apparently tight packing in the pocket, all of the L98X Hrs except for L98W, had second order O 2 association rate constants within a factor of 3 of the wild type value. Similarly, the O 2 dissociation rate constant was essentially unaffected by substitutions of larger (Phe) or smaller (Val, Thr) residues for Leu-98. L98Y Hr showed a 170-fold decrease in the O 2 dissociation rate constant and a large D 2 O effect on this rate, which are attributed to a hydrogen-bonding interaction between the Tyr-98 hydroxyl and the bound O 2 . Significant increases in autoxidation rates were observed for all of the L98X Hrs other than X ‫؍‬ Tyr. These increases in autoxidation rates are attributed to increased solvent access to the binding pocket caused by inefficient packing (Phe), smaller size (Val, Ala), or increased polarity (Thr, Asp, Asn) of the residue 98 side chain. A leucine at position 98 appears to have the optimal size, shape, and hydrophobicity for inhibition of solvent access. Thus, "gating" of small molecule access to the binding pocket of Hr by Leu-98 is not evident for O 2 , but is evident for solvent.
Hemerythrin (Hr) 1 is an oligomeric non-heme iron, O 2 -carrying protein found mainly in coelomic cells of a few marine invertebrate phyla (1). The octameric Hr from the sipunculid worm, Phascolopsis gouldii, is the most thoroughly characterized of this group. The eight essentially identical subunits exhibit no cooperativity in O 2 binding. Each 13.5-kDa subunit consists of a four-helix bundle protein backbone surrounding an oxo-/hydroxo-bridged di-iron site. Scheme 1 depicts the diiron site structure and also depicts a proposed structural mech-anism for reversible binding of O 2 to the site (1).
The structures depicted in Scheme 1 for the deoxy and oxy iron sites have been largely confirmed by x-ray crystallography and various spectroscopies (1). Several lines of evidence (summarized by Brunold and Solomon (Refs. 2 and 3)) show that the O 2 binding equilibrium of Hr is best formulated as an internal two-electron/one-proton transfer reaction. k on and k off are the O 2 association and dissociation rate constants, respectively. The bound O 2 is, thus, formally a hydroperoxo ligand with its proton hydrogen-bonded to the oxo bridge. Studies measuring O 2 affinities and association and dissociation rate constants have been published for several Hrs (4 -8), and have confirmed the second and first orders of these reactions, respectively. With one exception (9), the O 2 association/dissociation reactions of Hrs were reported to have monophasic kinetics at room temperature, implying that any intermediate species, such as that shown in Scheme 1, does not accumulate.
The O 2 -binding pocket of Hr is lined with a set of conserved, hydrophobic residues, whose sequential and spatial positions are indicated in Fig. 1. One or more side-chain atoms of residues Ile-28, Phe-55, Trp-97, Leu-98, and Ile-102 are within 4 Å of the coordinated O 2 atoms in oxyHr (1). L98C␦1 is 3.6 Å from atom O 2 of the bound dioxygen (cf. Scheme 1 and Fig. 1 for atom numbering). Based on this close distance, it was proposed that Leu-98 could limit ingress or egress of O 2 to or from the binding pocket, thereby functioning as a steric "gate" (11). However, this proposal has not heretofore been subjected to a direct experimental test. Based on laser flash photolysis studies of another oxyHr, opening of a gate was proposed to lead to escape of O 2 into solvent following its photodissociation from Fe2 (6). However, this putative "gate opening" occurred on the time scale of microseconds or less, which is much faster than the thermal O 2 dissociation rates of any known Hr (15-120 s Ϫ1 ; Ref. 12). Furthermore, in order to accurately model the exper-imental data, the gate was proposed to be permanently open in the deoxy form (13). Two mutations of the residue corresponding to Leu-98 in a myoHr, the monomeric counterpart to Hr, led to rapid autoxidation of the di-iron site, such that O 2 binding kinetics and equilibria could not be measured (14). Autoxidation, formulated as Reactions 2 and/or 3, is presumably initiated by solvent entry into the binding pocket of the oxy form (15,16

EXPERIMENTAL PROCEDURES
General Molecular Biology Procedures-Molecular biology procedures not described below followed those of Sambrook et al. (17) or Ausubel et al. (18). Restriction enzymes were obtained from either Roche Molecular Biochemicals or Promega, Inc. Oligonucleotides were purchased from Operon Technologies, Inc., Biopolymers, Inc., Integrated DNA Technologies, Inc., or the Great American Gene Co.. PCR products were purified from agarose electrophoresis gels using Wizard PCR Preps DNA purification kits (Promega). Plasmids were purified using Midi Prep plasmid purification kits (Qiagen, Inc.). Nucleotide sequencing was carried out in the Molecular Genetics Instrumentation Facility at the University of Georgia. Escherichia coli cultures were grown either in LB/amp or in LB/amp containing agar. Protein overexpression was monitored by Tricine SDS-polyacrylamide gel electrophoresis (19) on samples of E. coli cultures removed before and after induction (as described below), and on cell lysate fractions.
Cloning of the P. gouldii Hr Gene-Specimens of P. gouldii were obtained live from the Marine Biological Laboratory (Woods Hole, MA). Hemerythrocytes (the hemerythrin-containing coelomic cells) were isolated from the combined coelomic fluid of approximately a dozen specimens by low speed centrifugation and washing in artificial seawater. Total RNA was isolated from ϳ2 ml of packed hemerythrocytes using a total RNA isolation kit from Stratagene, Inc., and assuming that 1 ml of packed hemerythrocytes corresponded to 1 g of tissue. The precipitated RNA was stored at Ϫ20°C. The recombinant P. gouldii Hr gene was cloned from the total RNA by reverse transcription-PCR using a Perkin-Elmer Cetus GenAmp ® RNA PCR kit. The reverse transcription-PCR mix contained, in addition to the standard ingredients, ϳ1 g of the P. gouldii hemerythrocyte total RNA (dissolved in 2 l of 10 mM Tris and 1 mM EDTA, pH 8) and 0.25 M each of two degenerate primers, DMK2 (5Ј-GCTGCAGTAAGGAGGTTTAACATGGGNTTYCCNATHC CNGA-Y-3Ј) and DMK3 (5Ј-ATGCAAGCTTADATYTTNCCYTTRTAYTT-3Ј), in a total volume of 50 l. The degenerate portions of the nucleotide sequences of DMK2 and DMK3 were based on the published N-and C-terminal amino acid sequences, respectively, of P. gouldii Hr (1), and either PstI or HindIII restriction sites (underlined) were added to the 5Ј ends of the primers. The following thermal cycling sequence was used in the PCR: once for 2 min at 95°C, 35 times for (1 min at 95°C and 1 min at 42°C), once for 7 min at 60°C. The resulting PCR product was ligated into the PstI and HindIII restriction sites of pBluescript KS ϩ (Stratagene, Inc.). The resulting plasmid was sequenced to confirm insertion of the Hr gene, which was then isolated and purified from a PstI/HindIII digest of the plasmid. This purified, PstI/HindIII-digested Hr gene was used as template in a PCR with ϳ0.5 M primers DMK6 (5Ј-TATACATatgGGTTTCCCGATTCCG-3Ј) and DMK3. The nucleotide sequence of DMK6 duplicated that of the N-terminal end of the cloned P. gouldii Hr gene with the start codon indicated in lowercase, and an NdeI restriction site (underlined) incorporating the start codon was added to the 5Ј end. The thermal cycling sequence used for the PCR was: 30 times for (2 min at 95°C, 2 min at 60°C, and 3 min at 72°C), and once for 5 min at 72°C. The PCR product was ligated into the NdeI/HindIII restriction sites of pT7-7 (20). Nucleotide sequencing of the resulting pT7-7 derivative, designated pDK4 -1, confirmed the correctness of the inserted Hr gene sequence. The cloned Hr gene sequence in pDK4 -1 was deposited as GenBank accession no. AF220529.
Construction of L98X-mutated Hr Genes-Site-directed mutations in the Hr gene were obtained by the method of splicing by overlap extension (21), which requires two rounds of PCRs for each mutation. The first round consisted of two PCRs, both using BglII-digested pDK4 -1 as template. One first-round PCR used the oligonucleotide T7 (5Ј-TAAT-ACGACTCACTATAGGG-3Ј) and one oligonucleotide, L98X, with the sequence 5Ј-GTG GTT GAC NNN CCA GCT CTT-3Ј as primers. The nucleotides, NNN, consisted of one of the following sequences (5Ј 33Ј) for each L98X mutation: X ϭ F, GAA; X ϭ Y, GTA; X ϭ V, GAC; X ϭ T, AGT; X ϭ W, CCA; X ϭ D, GTC; X ϭ N, AAC; X ϭ A, GCA. The other first-round PCR used oligonucleotide T7R (5Ј-TCAGACCAAGTT-TACTCA-3Ј) and the corresponding reverse complement of the L98X oligonucleotide as primers. T7 and T7R duplicate nucleotide sequences in pT7-7 upstream of the NdeI restriction site and downstream of the HindIII restriction site, respectively (20). The second round of PCR used the two combined, purified PCR products from the first round as template with T7 and T7R as primers. For both rounds the PCR reaction mixtures contained ϳ5 ng of template DNA, ϳ0.08 M amounts of oligonucleotide primers, 0.25 M amounts of each of the dNTPs, and 0.5 unit of Taq polymerase in PCR reaction buffer (10 mM Tris-HCl, pH 8.9, 50 mM KCl, 2.5 mM MgCl 2 ) to a final volume of 50 l. The PCR temperature-cycling sequence was: 1 ϫ 5 min at 95°C, 30 ϫ [1 min at 94°C, 55°C for 1 min, and 2 min at 72°C]. The second-round PCR products were restriction-digested with NdeI and HindIII and ligated into the corresponding restriction sites of pT7-7. Those plasmids containing the correctly mutated L98X Hr gene sequences were used to transform E. coli BL21(DE3) (22).
Overexpression, Isolation, and Purification of Recombinant P. gouldii Hrs-Fifty-milliliter cultures of E. coli BL21(DE3) transformed with either pDK4 -1 or one of its L98X-mutated derivatives were grown overnight at 37°C in LB/amp with shaking at 250 rpm. Four of these 50-ml overnight cultures were used to inoculate four 1-liter batches of LB/amp, and these 1-liter cultures were incubated under the same conditions of temperature and shaking. When the OD 600 of the cultures reached ϳ1.0, either isopropyl-␤-D-thiogalactoside to 0.4 mM final concentration or 4.0 g of ␤-D-lactose was added to induce overexpression of the Hr gene. (Either isopropyl-␤-D-thiogalactoside or lactose gave comparable levels of Hr overexpression.) The induced cultures were incubated for another 2-4 h at the same temperature and shaking speed. All subsequent operations were performed at room temperature, except for centrifugations and pressure ultrafiltrations, which were at 4°C. The harvested, combined cells from four 1-liter cultures were resuspended in ϳ100 ml of 50 mM HEPES, 150 mM Na 2 SO 4 , pH 7.5 (buffer), and then lysed by sonication. The lysed cell suspension was centrifuged at 30,000 ϫ g for 30 min. The overexpressed Hr was found to be exclusively in the pellet, which was washed with buffer and then resuspended in ϳ12 ml of 6 M guanidine-HCl in buffer. To this suspension was added 200 l of ␤-mercaptoethanol, and the mixture was stirred vigorously for 5-24 h aerobically at room temperature to facilitate dissolution of the inclusion bodies. The green-brown suspension was then centrifuged at 30,000 ϫ g for 30 min. The supernatant (ϳ40 ml) was transferred to a 250-ml beaker with a magnetic stir bar. Ten volumes of buffer were added dropwise over 2-5 h, with stirring at room temperature. The resulting suspension was centrifuged at 30,000 ϫ g for 30 min, and the supernatant was discarded. The insoluble protein pellet was re-dissolved in guanidine-HCl/␤-mercaptoethanol as before; only several minutes of gentle stirring were required to re-dissolve the pellet at this stage. The solution (ϳ10 ml) was transferred to a 500-ml Schlenk-type flask fitted with a pressure-equalizing addition funnel, which contained 100 ml of buffer. This system was sealed, attached to a vacuum manifold, and alternately evacuated and flushed with argon while stirring until frothing of the protein solution was negligible. The addition funnel was briefly removed from the flask under a positive pressure of argon, and ϳ0.1 g of solid ferrous ammonium sulfate was rapidly added to the protein solution. The addition funnel was replaced, anaerobiosis was re-established, and the 100 ml of buffer was added dropwise at a constant rate over a period of ϳ10 h with gentle stirring. After complete addition of the buffer, the protein solution was centrifuged aerobically at 30,000 ϫ g for 30 min. The resulting pink/yellow solution, which contained soluble, iron-containing Hr, was concentrated by ultrafiltration (Amicon, YM30 membrane, 30-kDa cut-off) under argon pressure to ϳ5 ml. Two 50-to 5-ml dilution/reconcentration cycles with buffer were performed to remove residual reagents. To ensure homogeneity, the Hr was converted to the met form by addition of 5-10 small crystals of potassium ferricyanide. This solution was stirred overnight at 4°C. The ϳ5 ml of protein solution was then loaded onto a Superdex 200 16/60 gel filtration column (Amersham Pharmacia Biotech) pre-equilibrated with buffer. The Hr was eluted with buffer at a flow rate of 0.5 ml min Ϫ1 as a single band, and was collected and concentrated by ultrafiltration (Centricon 30, 30-kDa cut-off). Both wild type and L98X Hrs were similarly overexpressed, isolated, and purified. The purified Hrs were stored at Ϫ80°C. Amounts and purities of the recombinant Hrs were estimated by UV-visible spectrophotometry (⑀ 330 ϭ 6400 M Ϫ1 cm Ϫ1 /subunit) and A 280 /A 330 ratio (typically 4 -4.5) (23). SDS-polyacrylamide gel electrophoresis showed a single protein band for the recombinant Hrs after this purification procedure. Four liters of E. coli culture typically yielded 50 mg of purified soluble recombinant Hr.
Preparation of Deoxy-and OxyHrs-MetHr in buffer, prepared as described above, was used as the starting point. The deoxy forms of the recombinant Hrs were prepared by anaerobic dialysis of samples of metHr against ϳ5 mM sodium dithionite solutions in buffer for ϳ24 h at room temperature either in a Coy ® anaerobic chamber or in a Schlenktype flask attached to a vacuum manifold. The protein was then dialyzed three times against 100 volumes of deoxygenated buffer to remove excess dithionite. The resulting protein solution was centrifuged at 10,000 ϫ g for 5 min inside the Coy chamber in order to remove any precipitate. The oxy forms were prepared by briefly bubbling air through a freshly prepared solution of deoxyHr. Aliquots (ϳ1 ml) of the oxyHrs were frozen rapidly in liquid nitrogen and stored at Ϫ80°C.
Measurement of O 2 Affinities of Hrs-O 2 affinities for wild type and L98X Hrs were measured by spectrophotometric tonometry. The method has been previously described for native P. gouldii Hr (24). Solutions of the deoxyHr to be examined (1.0 -1.5 ml, 50 -100 M in subunits) were anaerobically transferred to a specially adapted 1-cm pathlength cell, which was then attached to the tonometer. The O 2 association constant was expressed as P 50 , the partial pressure of O 2 at which [oxyHr]/[Hr] total ϭ 0.5, as determined from exponential fits to plots of A 500 versus PO 2 .
Measurements of O 2 Association, O 2 Dissociation, and Autoxidation Kinetics of Hrs-Rates of O 2 dissociation from oxyHr were measured by stopped-flow spectrophotometry using a rapid scanning monochromator from Olis, Inc. A previously described O 2 -scavenging method was employed (4) by mixing the oxyHr solution (ϳ200 M in di-iron sites) with 100 mM sodium dithionite solution in buffer. Spectra were collected every millisecond following a dead time of 2 ms. Absorbance changes at 500 nm versus time were fit to a single exponential decay for Ͼ5 half-lives using non-linear least squares fitting to determine the firstorder dissociation rate constant, k off . Measurements of the temperature-dependence of k off between ϳ6°C and ϳ35°C, controlled to Ϯ0.01°C by a circulating bath, were used to determine enthalpies and entropies of activation from Eyring plots. Rates of O 2 association with deoxyHrs were measured by stopped-flow spectrophotometry in the same buffer and on the same instrument as for the dissociation rates. This method has also been described previously for native P. gouldii Hr (4). Solutions of the deoxyHrs, freshly prepared as described above, were used for these experiments. A small amount of the deoxyHr solution was exposed to air, and the resulting oxyHr (⑀ 500 ϭ 2, 200 M Ϫ1 cm Ϫ1 ) (25) was used to determine the concentration of the deoxyHr. Prior to transferring the deoxyHr, the loading syringe was filled with ϳ100 mM sodium dithionite and incubated for ϳ30 min in order to scavenge any adventitious O 2 . The loading syringe was then washed three times with dithionite-free anaerobic buffer. A gas-tight syringe containing the deoxyHr solution was rapidly transferred to the loading syringe on the stopped-flow spectrophotometer, and the protein solution was injected into the loading syringe. The deoxyHr solution remained anaerobic as judged by the lack of significant absorbance at 500 nm. Aerated buffer was loaded into the second stopped-flow syringe. The concentration of dissolved O 2 in the aerated buffer was measured with a Clark-type oxygen electrode (Yellow Springs Instruments, Inc.). The concentration of O 2 was 140 M after mixing, which was a 5-10-fold molar excess over Hr di-iron sites (12-30 M after mixing), and pseudo first order behavior was therefore, assumed. The increase in absorbance at 500 nm was monitored versus time, or, alternatively, absorption spectra from 350 nm to 600 nm were collected at 1-ms intervals. Data were plotted as absorbance at 500 nm versus time, and the resulting curves were least-squares fit to single exponentials over Ն3 half-lives from which the second order O 2 association rate constants, k on , were derived. Activation parameters were determined as described above. Autoxidation half-lives at 25°C were estimated by monitoring the distinctive changes between oxyHr and metHr absorption spectra (23) on a Shimadzu UV2101PC scanning spectrophotometer. For autoxidation half-times that were shorter than a few seconds, the rates were determined on the Olis stopped-flow spectrophotometer by mixing the deoxy form with O 2 solution as described above for the O 2 association rates.
Resonance Raman Spectra-Resonance Raman spectra were obtained on Hr samples prepared at the University of Georgia and shipped frozen in dry ice to the Oregon Graduate Institute. Spectra were obtained at ice temperature using a 90°scattering geometry on a McPherson 2061 spectrograph with an 1800-groove grating and a Princeton Instruments liquid N 2 cooled (LN100PB) CCD detector. The excitation sources were a Coherent Innova 300 Kr ϩ laser for 413.1 nm and a Spectra Physics Ar ϩ 164 laser for 514.5 nm and 488.0 nm. Raman samples in buffer were 1-2 mM in di-iron sites.

Isolation of the Recombinant P. gouldii Hrs-The P. gouldii
Hr gene was cloned from cDNA of hemerythrin-containing coelomic cells. The derived amino acid sequence was identical to that published for one of the minor amino acid variants of native P. gouldii Hr in which a threonine is substituted for a glycine at position 79 (26). Native P. gouldii Hr consists of a total of five amino acid substitution variants in varying proportions; none of these variants is known to affect any spectroscopic or functional properties of the di-iron site (27). Both the recombinant wild type and L98X P. gouldii Hrs were expressed in E. coli as insoluble apoproteins using the T7 polymerase/promoter overexpression system (20,22). Based on a previously described procedure used to reconstitute native P. gouldii Hr from the apoprotein (23, 28), a method was developed for resolubilizing the recombinant wild type P. gouldii apoHr from the inclusion bodies and for incorporation of iron such that a di-iron site with properties (see below) essentially indistinguishable from that in native P. gouldii Hr was obtained. The same method was used to obtain a series of soluble di-iron-containing recombinant L98X P. gouldii Hrs. All of these recombinant Hrs were octameric, as judged by gel filtration, as is native P. gouldii Hr (1).
Spectral Properties of the Recombinant P. gouldii Hrs-After resolubilization, iron incorporation, and oxidation with ferricyanide, all of the recombinant wild type, and L98Y, -F, -V, -W, -T, -D, -N, and -A P. gouldii Hrs contained di-iron sites whose absorption spectra (cf. Fig. 2) were very similar to that of native P. gouldii metHr at pH 7.5 (23). The absorption features at ϳ320 and ϳ360 nm, and shoulder at ϳ480 nm, are due to -oxo 3 Fe(III) LMCT transitions, and this set of features serves as a fingerprint for -oxo-di-iron(III) sites with supporting carboxylate bridges (29,30). The absorption spectrum of the recombinant L98Y metHr was found to have a weak, broad absorption between 500 and 700 nm, which was not present in the wild type or other L98X metHr spectra. This feature showed variable intensity in multiple preparations of L98Y metHr, and neither its intensity nor position was affected by pH between 6.0 and 9.0. The L98Y metHr spectrum shown in Fig. 1 contains a very minor contribution from this feature. Resonance Raman spectra, discussed below, indicate that this absorption is due to a portion of the protein in which a phenolate, presumably from Tyr-98, is ligated to Fe(III). Resonance Raman spectra of the recombinant wild type metHr (data not shown) exhibited the characteristic s (Fe-O-Fe) and as (Fe-O-Fe) stretching frequencies of the oxo-bridged di-iron(III) site at 508 and 758 cm Ϫ1 (weak), respectively. These frequencies mimic those of the native P. gouldii metHr (510 and ϳ753 cm Ϫ1 , respectively; Ref. 31), which indicates conservation of the bent Fe-O-Fe angle (ϳ125°) enforced by the two additional carboxylate bridges (32). An extensive and detailed analysis of the resonance Raman spectra of the recombinant metHrs can be found elsewhere (33).
Upon anaerobic reduction with sodium dithionite, and subsequent re-exposure to air, the wild type and L98V, -F, -Y, and -T P. gouldii Hrs formed oxy adducts that were sufficiently stable at room temperature to obtain their characteristic UVvisible absorption spectra (cf. Fig. 3). In addition to the peak at 330 nm and shoulder at ϳ360 nm (both due to oxo3 Fe(III) LMCT), all of these latter Hrs displayed a broad absorption centered near 500 nm, which is due to the hydroperoxo3Fe(III) LMCT transition (30). Either very small or negligible perturbations of the wild type oxyHr spectrum are evident in the spectra of the L98X oxyHrs. These spectra completely bleach in the deoxy forms, which thereby provides a convenient spectroscopic monitor of O 2 binding.
Laser excitation into the hydroperoxo3 Fe(III) LMCT transition of native P. gouldii oxyHr is known to enhance the (O-O) and (Fe-O 2 ) Raman-active stretching frequencies of the bound O 2 (31). The analogous Raman experiments on the recombinant oxyHrs produced the spectra shown in Fig. 4. The frequencies for the recombinant wild type oxyHr at 844 and 503 cm Ϫ1 are identical to those previously reported for (O-O) and (Fe-O 2 ), respectively, of native P. gouldii oxyHr (31). Small (ϩ2 cm Ϫ1 ) upshifts in these frequencies are observed in the L98F oxyHr. In L98Y oxyHr, the (O-O) frequency has shifted 3 cm Ϫ1 downward to 841 cm Ϫ1 . The (Fe-O 2 ) frequency region of the L98Y oxyHr Raman spectrum is complicated by contributions from a minor portion of the met form that invariably contaminates concentrated samples of oxyHrs. In the case of L98Y Hr, the met frequencies are apparently enhanced to a greater extent than for wild type or L98F met forms. Comparisons to resonance Raman spectra of L98Y metHr (data not shown; Ref. 33) revealed that the Raman features at 498 and 515 cm Ϫ1 are due mostly to s (Fe-O-Fe) of two different met forms, one of which has a tyrosinate, presumably Tyr-98, coordinated to Fe(III). The feature at 573 cm Ϫ1 in Fig. 4 is attributed to (Fe-O) of this coordinated tyrosinate in the met form (34). These comparisons also indicate some residual intensity from the (Fe-O 2 ) of L98Y oxyHr near 505 cm Ϫ1 . In any case both the resonance Raman and UV-visible absorption spectra indicate that, in those L98X Hrs that form a stable oxy adduct, the bound O 2 is a peroxo ligand resulting from the same internal redox reaction (Reaction 1) as for the wild type Hr. described O 2 -scavenging method (4), the O 2 dissociation rates of the recombinant P. gouldii oxyHrs could be conveniently monitored by stopped-flow spectrophotometry. Within a relatively narrow range of deoxyHr concentrations and excess O 2 (cf. "Experimental Procedures"), the O 2 association kinetics of the recombinant P. gouldii Hrs could also be reliably measured by stopped-flow spectrophotometry between 5 and 35°C. Autoxidation half-times (Reactions 2 and 3) were also estimated. In those cases where the autoxidation rates were on the stopped-flow time scale (L98A, -N, and -D, cf. Table I), the initial formation of the 500-nm feature characteristic of oxyHr was observed, followed by a decrease in absorbance toward the met form, which is the product of autoxidation (Reactions 2 and 3). L98W Hr was an exception, and is discussed separately. The O 2 association and dissociation kinetics were best fit to monophasic processes in all cases (except L98W). Even when the O 2 association reactions were monitored in the wavelengthscanning mode of the stopped-flow spectrophotometer with higher protein concentrations under non pseudo first order conditions, no intermediate chromophores were evident. The results of these O 2 affinity and kinetics measurements are collected in Table I. The O 2 affinity and rate constants for the recombinant wild type P. gouldii Hr closely parallel those previously reported for native P. gouldii Hr (4,24).
Given the relatively narrow temperature range, we were able to examine, the small temperature dependence of the O 2 association rates, and the large extrapolations used to obtain ⌬S ‡ from the Eyring plots, quantitative comparisons among the activation parameters of the various Hrs listed in Table I is not warranted. However, a few qualitative points are noteworthy.
Although the values of ⌬S ‡ on determined for the recombinant P. gouldii Hrs (Table I) are significantly more negative than previously reported for native P. gouldii Hr (4), they are not inconsistent with a second order associative reaction, such as occurs in O 2 -carrying proteins (cf., e.g., Coletta et al. (Ref. 35)). A ⌬S ‡ on of Ϫ11 cal/mol-K was reported for Tz myoHr (5). Previous kinetic and calorimetric studies on native P. gouldii Hr have shown the overall O 2 association reaction of native P. gouldii Hr to be exothermic by Ϫ9 to Ϫ12 kcal/mol (4,36). The calculated ⌬H ϭ ⌬H ‡ off Ϫ ⌬H ‡ on values for the O 2 association equilibria of the recombinant P. gouldii Hrs (cf. Table I) are also exothermic, but, once again, somewhat more so than reported for the native P. gouldii Hr.
Effects on O 2 Association-Perhaps the most striking aspect of the data in Table I is the similarities in k on among the wild type and several L98X Hrs. The rates for O 2 association are all within the range of 1-3 ϫ 10 6 M Ϫ1 s Ϫ1 (except for L98W). These high rates, and the low activation enthalpies, are incompatible with an O 2 association mechanism in which bond formation or ligand substitution occurs in the rate-limiting step. Rather, these parameters suggest an open coordination site for O 2 binding, as is indeed observed for Fe2 (cf. Scheme 1) in the x-ray crystal structure of deoxyHr (1). The O 2 association parameters listed in Table I are consistent with a diffusioncontrolled associative mechanism (6,(37)(38)(39). According to this mechanism, two effects could lower the second order rate constants below that expected for a purely diffusion-controlled encounter (ϳ10 9 M Ϫ1 s Ϫ1 ). First, the time scales required for structural fluctuations of the protein to create transient internal cavities are longer than that for collisional encounters, and second, favorable interactions of the small molecule exist with only a fraction of the protein surface and/or internal matrix. Perhaps due to modest steric inhibition by the larger side chain, L98F Hr shows an approximately 3-fold reduction in k on , which is reflected in its approximately 3-fold lower O 2 affinity, i.e. ϳ3-fold higher P 50 (cf . Table I). However, the Leu-98 mutants with aliphatic residues smaller than leucine did not show an increase in k on or O 2 affinity, which would be expected if steric restrictions at Leu-98 regulated O 2 association. Thus, a narrow range of O 2 association rate constants is observed for Hrs having a wide range of sizes and polarities at residue 98. We, therefore, conclude that Leu-98 does not limit the rate of O 2 access to the binding pocket of P. gouldii Hr. The fact that Hrs were ϳ1 mM in di-iron sites in 50 mM HEPES and 150 mM Na 2 SO 4 (pH 7.5). Spectra were acquired at ice temperature using 514-nm laser excitation. Other spectral conditions are given under "Experimental Procedures."

TABLE I Kinetic and thermodynamic parameters characterizing O 2 binding to recombinant wild type and L98X Pg Hrs
Conditions for all recombinant Pg Hr measurements: 50 mM HEPES and 200 mM Na 2 SO 4 (pH 7.5) and 25°C. Subscripts "on" and "off" indicate O 2 association and dissociation, respectively. Experimental uncertainties: k on , Ϯ10 -15%; k off , Ϯ4%; ⌬H ‡ on and ⌬S ‡ on , Ϯ20 -30%; ⌬H ‡ off , Ϯ4%; ⌬S ‡ off , Ϯ10 -30%; P 50 , Ϯ10 -30%.  Table  I are consistent with significant Fe-O 2 bond weakening or breakage during the step limiting O 2 dissociation. This conclusion had been reached previously for native Hrs and myoHr based on both temperature and pressure dependences of the dissociation rates (8,39). Since one C␦ atom of Leu-98 is reported to be 3.6 Å from a bound O 2 atom in the crystal structure of oxyHr (1), this residue could in principle sterically hinder expansion of the Fe2-O 2 coordination sphere. However, changing Leu-98 to either smaller (Thr, Val) or larger (Phe) residues did not affect k off (cf . Table I). Thus, Leu-98 does not appear to modulate the rate of O 2 dissociation from P. gouldii Hr.
The O 2 dissociation rate constant for L98Y Hr is 170 times slower than that of wild type Hr. The approximately 15-fold higher O 2 affinity (i.e. lower P 50 ) of L98Y Hr is qualitatively consistent with the rate constants. Such high O 2 affinities (Ͻ0.5 mmHg) are difficult to accurately measure by spectrophotometric tonometry. Since the similarly sized and shaped phenyl side chain in L98F oxyHr produced no detectable perturbation in k off , steric inhibition is unlikely to be responsible for the much slower dissociation rate in L98Y oxyHr. An alternative explanation is that the Tyr-98 hydroxyl forms a hydrogen-bond with O-1 (or possibly O-2) of the bound O 2 (cf. Scheme 1), thereby stabilizing the oxy adduct. As a test of the hydrogenbonding explanation, the O 2 dissociation rate of L98Y oxyHr was measured in buffered D 2 (31). This latter difference is also consistent with an altered hydrogen bonding pattern in L98Y oxyHr. Given the very slow autoxidation rate of L98Y oxyHr (even slower than for wild type), it is unlikely that solvent introduction into the binding pocket is responsible for these altered isotope effects.
Effects on Autoxidation-Wild type, L98V, L98T, L98N, L98D, and L98A Hrs autoxidize successively more rapidly, even though their O 2 association rates are all very similar (cf. Table I). This comparison indicates that both the size and hydrophobicity of the residue 98 side chain contribute barriers to solvent access in the binding pocket when bound O 2 is also present. The importance of side chain volume is illustrated by the fact that L98A Hr was the most rapidly autoxidizing of the series. The importance of hydrophobicity can be illustrated by the relative autoxidation rates of L98V and L98T oxyHrs. The Val and Thr side chains are approximately the same size and shape, but the less hydrophobic substitution mutant, L98T, autoxidizes 8 times faster. L98N and -D oxyHrs autoxidize ϳ4,000 times faster than does L98T, further emphasizing the importance of pocket polarity in enhancing autoxidation rates. Thus, Leu-98 plays a significant role in inhibiting autoxidation of P. gouldii Hr, due both to its size and its hydrophobicity. The very slow autoxidation rate of L98Y Hr, even slower than for wild type, is presumably related to the same stabilizing interactions giving rise to its very slow O 2 dissociation rate. The L98F mutation resulted in a 10-fold increase in autoxidation rate relative to wild type. Since this mutation did not affect k off or significantly perturb the oxyHr absorption and Raman spec-tra, the increased autoxidation rate of L98F cannot be due to a direct perturbation of the Fe2-O 2 adduct structure. Therefore, the increased autoxidation rate of L98F must be due to structural perturbations in the binding pocket without any concomitant stabilizing interactions, as occurs in L98Y Hr.
L98W Hr-The rate of O 2 association for L98W deoxyHr is ϳ50 times slower than for wild type Hr. The inset to Fig. 5 shows that the chromophoric species formed on the stoppedflow time scale for the L98W deoxyHr reaction with O 2 is different from that of wild type (or any of the other L98X Hrs). The absorption maximum of the first detectable species is at ϳ650 rather than 500 nm. This particular experiment was not performed under pseudo first order conditions in order to facilitate detection of this intermediate. No absorption feature having a maximum near 500 nm was observed subsequent to the traces shown in Fig. 5. Only the spectrum attributable to the met form developed. Autoxidation of L98W Hr is extremely rapid, with a half-time of 70 ms. The spectrum of the end product of the reaction between L98W deoxyHr and O 2 is shown in Fig. 4. Comparison with the spectra of Fig. 1 shows that this product has an oxo-bridged di-iron site typical of metHrs.

DISCUSSION
Although the structure and function of the di-iron site in Hr is well established, the role of conserved amino acid residues, other than those providing metal ligands, is less well characterized. The essentially concerted redox and proton transfer process that occurs upon O 2 binding to Hr (Reaction 1 and Scheme 1) is self-contained within the di-iron site (3), and the conserved pocket residues in Hr are all hydrophobic (cf. Fig. 1). Therefore, if these pocket residues influence the O 2 binding process, they must do so by relatively indirect means.
The UV-visible and resonance Raman data (Figs. 2-4) indicate that the essential features of the native oxyHr di-iron site are preserved in the L98X oxyHrs. Thus, any effects of these Each stopped-flow spectrum was collected over 1 ms, and succeeding spectra were acquired at 5-ms intervals from lowest to highest absorbance. Arrow indicates direction of absorbance changes with time. mutations on the kinetics or thermodynamics of O 2 binding cannot be due to significant alterations in the ground state of the oxy di-iron site. The "steric gate" function previously proposed for Leu-98 in Hr (14), is not supported by the kinetics of O 2 association and dissociation for a series of L98X Hrs. Despite the apparently tight packing in the pocket, substitutions of Leu-98 with larger, smaller, or more polar residues had little or no effect on the second order O 2 association rate constant. Similarly, larger and smaller residues could be substituted for Leu-98 with no effect on the O 2 dissociation rate constant. Thus, Leu-98 does not affect expansion of the Fe2-O 2 coordination sphere, assuming that is the rate-limiting step in O 2 dissociation (39). L98Y was the only mutation we tested that showed a large effect on k off . This mutation also gave a much larger D 2 O effect on k off than did wild type. Therefore, a hydrogen-bonding interaction that has been either introduced or enhanced relative to the wild type could be responsible for the much lower O 2 dissociation rate of L98Y Hr. The hydroxyl of Tyr-98 would not interact favorably with the other hydrophobic residues lining the binding pocket, thereby favoring its localization near the polar Fe2-O 2 moiety. The hypothetical structure shown in Scheme 2 is consistent with our data for L98Y oxyHr, assuming that the proton between O 2 and the oxo bridge is transferred during the rate-limiting step of O 2 dissociation (40).
The x-ray crystal structure of L98Y metHr has been solved, and the position of the Tyr-98 hydroxyl indicates that the structure shown in Scheme 2 is feasible. Details of the L98Y metHr crystal structure will be reported separately. 2 Substitutions of Leu-98 with successively smaller or more polar residues significantly increased the autoxidation rate of P. gouldii oxyHr. Due to rapid fluctuations of protein matrices in solution, it has been proposed that hydrophobicity rather than volume limits water occupancy in buried cavities within proteins (37). This proposal, in fact, rationalizes the hydrophobic nature of the O 2 -binding pocket in Hr. The 8-fold faster autoxidation of L98T versus L98V Hr suggests that pocket residue hydrophobicity does influence the rate-limiting step, which is presumably entry of water into the binding pocket of the oxy form. The much faster autoxidations of L98N and -D oxyHrs reinforce this conclusion. However, the size of the Leu-98 pocket residue in Hr also appears to have a large influence on autoxidation rates. The mutant with the smallest side chain we tested, L98A, autoxidizes ϳ6,000 times faster than does L98T, and the alanine side chain is, if anything, more hydrophobic than that of threonine (41). We have found that L98A metHr binds larger molecules, such as phenol, than can be accommodated in the wild type metHr-binding pocket. 3 We propose that the larger O 2 -binding pocket created in L98A Hr increases the probability of simultaneous pocket occupancy by O 2 and solvent, thereby increasing the autoxidation rate.
The autoxidation rate of L98Y oxyHr is significantly slower than that of wild type (cf. Table I). We propose that the hydrogen bond-stabilized oxy structure, such as shown in Scheme 2, inhibits solvent attack. Such a "locked" pocket structure may also dampen side-chain fluctuations that normally admit solvent. The 10-fold increase in autoxidation rate of L98F relative to wild type can be explained as a disruptive "wedge" effect of substituting a larger and aromatic side chain into the tightly packed native structure. Recall that L98F has an ϳ3-fold slower O 2 association rate, suggesting a small perturbation of the O 2 diffusion pathway by the larger side chain. However, as noted above, the L98F mutation does not affect the essential structural features of the native oxy site nor the O 2 dissociation rate. Therefore, we suggest that the enhanced autoxidation rate of L98F oxyHr is due to less efficient packing of the phenyl ring with other side chains at and near the O 2 -binding pocket, thereby expanding the average size of the pocket in the oxy form. Given the rate-limiting steps proposed above, this expansion would not be expected to affect the rate of O 2 dissociation, but could increase solvent occupancy in the pocket, and, therefore, the rate of autoxidation.
Although this disruption-in-packing explanation could also apply to L98W Hr, the behavior of this mutant was distinctive. The initial spectrophotometrically detectable reaction of O 2 with L98W deoxyHr was 15-50-fold slower than with any other deoxyHr we tested, and this slower reaction yielded a transient 650-nm absorbing species. Such an intermediate has not heretofore been reported in reactions of O 2 with Hrs or myoHrs. This altered reactivity is likely due to extreme steric crowding in the binding pocket of L98W Hr, thereby preventing O 2 from coordinating in its usual fashion to Fe2 (cf. Scheme 1). Two possible candidates for the 650-nm chromophore are a diferricperoxo species, in which the peroxo ligand bridges Fe1 and Fe2 (42,43), or a mixed-valent Fe II Fe III species resulting from "outer-sphere" one-electron oxidation of the di-iron site by O 2 (44). Although mixed-valent forms of Hr are well known (45), they have not been detected during autoxidation of native P. gouldii Hr (15). Whatever the nature of the 650-nm intermediate may be, it leads to much more rapid autoxidation than any other L98X Hr tested. Thus, this intermediate is unlikely to be on the pathway to the wild type O 2 adduct.
The increases in autoxidation rates observed for the L98X Hr mutants in the absence of large effects on O 2 association/dissociation kinetics suggest an important and specific role in inhibition of autoxidation for this residue. Our results suggest that the conserved residues lining the binding pocket kinetically inhibit autoxidation by efficient packing of their hydrophobic side chains around the Fe2-O 2 moiety. This packing minimizes pocket volume and, thereby, limits occupancy of the pocket by solvent. A leucine side chain at position 98 appears to have the optimal combination of size, shape and hydrophobicity for this role. This combination of properties may have led to its conservation in all known Hrs (46). Nevertheless, Leu-98 does not modulate the rate-limiting steps of O 2 entry into or exit from the binding pocket of P. gouldii Hr. Thus, "gating" by pocket residues in Hr is evident for solvent but not for O 2 , and the gate for solvent entry appears to consist of both hydrophobic and steric barriers. Our recent investigations of corresponding pocket residue-mutated myoHrs led us to similar conclusions. 4