Waterproofing the Heme Pocket

The ability of myoglobin to bind oxygen reversibly depends critically on retention of the heme prosthetic group. Globin side chains at the Leu89(F4), His97(FG3), Ile99(FG5), and Leu104(G5) positions on the proximal side of the heme pocket strongly influence heme affinity. The roles of these amino acids in preventing heme loss have been examined by determining high resolution structures of 14 different mutants at these positions using x-ray crystallography. Leu89 and His97 are important surface amino acids that interact either sterically or electrostatically with the edges of the porphyrin ring. Ile99 and Leu104 are located in the interior region of the proximal pocket beneath ring C of the heme prosthetic group. The apolar amino acids Leu89, Ile99, and Leu104 “waterproof” the heme pocket by forming a barrier to solvent penetration, minimizing the size of the proximal cavity, and maintaining a hydrophobic environment. Substitutions with smaller or polar side chains at these positions result in exposure of the heme to solvent, the appearance of crystallographically defined water molecules in or near the proximal pocket, and large increases in the rate of hemin loss. Thus, the naturally occurring amino acid side chains at these positions serve to prevent hydration of the His93-Fe(III) bond and are highly conserved in all known myoglobins and hemoglobins.

Three important functions of the globin portion of myoglobin are to sequester heme, enhance coordination with the proximal histidine, and inhibit oxidation of the iron atom. In myoglobin, the heme pocket is surrounded by four of the eight globin helices. The B, C, and E helices form the top and sides of the porphyrin binding site, and the F helix forms the bottom (as oriented in Fig. 1). Heme binding causes the globin to form a more compact structure with an approximately 20% increase in helicity (1). Three edges of the porphyrin are buried in the protein interior and stabilized by hydrophobic interactions with apolar side chains that line the heme pocket (2)(3)(4). The fourth edge contains the solvent-exposed heme propionates, which interact electrostatically with polar amino acid side chains located on the surface of the protein.
The affinity of free heme for binding a single imidazole is very weak, whereas the affinity for binding a second base is very high. Thus, hexacoordinate bis-imidazole complexes are much more stable than the pentacoordinate intermediate (5)(6)(7). Although a much weaker ligand, water competes effectively against the binding of the first imidazole base in aqueous solution, causing the apparent equilibrium dissociation constant to be Ն10 Ϫ2 -10 Ϫ4 M (7). Hemoglobins and myoglobins have evolved to stabilize the mono-imidazole heme complex to facilitate reversible O 2 binding to the sixth coordination position. This stabilization appears to be accomplished by removing water molecules from the vicinity of the proximal coordination site.
The side chains of the proximal amino acids, Leu 89 (F4), His 97 (FG3), Ile 99 (FG5), and Leu 104 (G5), are within 4 Å of the porphyrin ring. Leu 89 is located at the entrance of a hydrophobic cavity underneath pyrrole B. Xenon gas binds readily in this proximal pocket, which has been designated the Xe1 site (8). Photodissociated O 2 and CO access both the distal Xe4 site and this Xe1 proximal site on nanosecond to microsecond time scales, suggesting that these spaces may be functionally significant (9 -13, 34). His 97 forms a hydrogen bond with the heme-7-propionate and is a hydrophobic barrier to solvation of the proximal heme pocket on the other side of the heme group. Ile 99 and Leu 104 are located just beneath the pyrrole rings B and C, respectively. Leu 104 forms one of the internal sides of the Xe1 cavity. Hargrove et al. (3) suggested that both these amino acids may be important for keeping the internal region of the proximal pocket anhydrous and preventing disruption of the His 93 (F8)-Fe bond. To test these ideas, we have examined the effects of mutations at these four key positions on the crystal structure of recombinant sperm whale myoglobin.

MATERIALS AND METHODS
Preparation of Proteins-Recombinant wild-type and mutant sperm whale myoglobins were constructed, expressed, and purified as described by Springer and Sligar (15) and Carver et al. (16). Recombinant wild-type myoglobin differs from native myoglobin by the initiator methionine required for bacterial expression and an Asp 122 to Asn mutation that was a result of an error in the original sequence determination. These changes do not affect the function of the recombinant protein (17) but do cause differences in crystallization conditions (18). The protein samples were concentrated to about 1 mM in heme, frozen, and stored under liquid nitrogen until ready for use.
Measurement of Hemin Dissociation Rate Constants-Hemin dissociation rates were measured by monitoring the absorbance changes associated with the transfer of hemin from holoprotein to excess H64Y/ V68F apomyoglobin at pH 5.0 and 7.0, as described by Hargrove et al. (19). Briefly, the transfer of hemin from 6 M metmyoglobin to 30 M H64Y/V68F apomyoglobin was measured by following the absorbance decrease at 409 nm as "green" H64Y/V68F holo-metmyoglobin was formed. The experiments were carried out with 0.15 M buffer and 0.45 M sucrose at 37°C in either sodium acetate at pH 5.0 or potassium phosphate at pH 7. The values for the rate of hemin loss for most of the mutants were originally reported in Hargrove et al. (3); the remainder were measured for this report (20).
Preparation of Protein Crystals-Prior to crystallization, recombinant proteins were purified further by HPLC 1 (LDC/Milton Roy). Samples were dialyzed against 20 mM sodium phosphate, 1 mM EDTA, pH 6.5, at 4°C. The dialyzed samples were loaded onto a preparative weak cation exchange column (Custom LC) equilibrated with the dialysis buffer. The samples were eluted using a linear gradient of 0.0 -1.0 M sodium chloride in 20 mM sodium phosphate, 1 mM EDTA, pH 6.5. Fractions containing the recombinant protein were combined and dialyzed against 20 mM Tris-HCl, 1 mM EDTA, pH 9.0, at 4°C. A few of the myoglobins, particularly the Leu 89 mutants, were unstable because of increased rates of heme loss and subsequent precipitation, which made protein purification and crystallization difficult. To overcome this problem, the unstable samples were reduced with sodium dithionite prior to purification to strengthen the covalent bond anchoring the heme group to the globin. Excess reducing agent was removed by passing the sample through a Sepharose column equilibrated with "no salt" HPLC buffer. Crystals of the HPLC-purified recombinant proteins were produced using the hanging drop vapor diffusion method employing a pH and ammonium sulfate gradient, as described by Phillips et al. (18). Briefly, 10 l of 20 mg/ml HPLC-purified protein was mixed with an equal volume of 2.4 -3.0 M ammonium sulfate, 20 mM Tris-HCl, 1 mM EDTA, pH 8.0 -9.0, with well solutions having concentrations of 2.4 -3.0 M ammonium sulfate. During sample preparation those mutants which were reduced for HPLC purification reoxidized to the met form. Crystals grew in several weeks to sizes from 0.1 to 0.5 mm. The Ile 99 to Ser mutation failed to crystallize with these conditions. The other mutations for which heme loss rates were given and no structure is presented were not screened for crystallization (Leu 89 to Ala and Ser and His 97 to Ala).
X-ray Diffraction Data Collection and Refinement-X-ray diffraction data were collected on a Rigaku R-AXIS IIC imaging plate area detector with a Siemens rotating-anode source. Typically, 60 -90 frames were collected at a detector distance of 70 mm with a 30-min/frame exposure time and a oscillation range of 1°/frame. The data were processed using the XDS and XSCALE software packages (21).
In all cases, 10% of the reflections were set aside for a test set and the calculation of R free . The X-PLOR package (22) was used for the initial refinement of the model structures and for the generation of difference maps, using the Engh and Huber (23) parameter set for stereochemical restraints. The coordinates of wild-type metmyoglobin (18) were used as a starting model in the refinement of the mutant structures. An iterative procedure of manually fitting the model using the molecular graphics program CHAIN (24) followed by Powell minimization, solvent B factor, and occupancy refinement, and individual B factor refinement for all atoms was repeated until R free was unchanged, and no significant differences were observed in electron density maps. The mutated side chains were changed only after ͉2F o Ϫ F c ͉ and ͉F o Ϫ F c ͉ difference Fourier maps indicated clear positions for these altered side chains. Peaks in the difference Fourier maps were modeled as water molecules using the program PEAK 2 or OH2. 3 After the refinement converged in X-PLOR, the model was refined further using the conjugate gradient and full matrix least squares method in SHELXL (25). Estimated standard deviations of the atomic positions of the heme iron and the 24 atoms of the pyrrole rings and methionine bridges were calculated from refinement without the imposition of standard distances and planarity restraints, as described in Ref. 25.
Sequence Alignment-All myoglobin sequences were downloaded from the National Center of Biotechnology Information. About 460 sequences were obtained through the Entrez Protein search engine, which included all myoglobin related data from SwissProt, Protein Information Resource, Protein Data Bank, and Protein Research Foundation entries, etc. Incomplete myoglobin sequence fragments were immediately discarded. Most of the myoglobin sequences from the Protein Data Bank are from known myoglobins, and their mutants were also omitted in the sequence alignment. The first sequence alignment containing 203 sequences was performed using Clustal W (version 1.8) (26). Duplicate sequences with different entry names were recognized and sorted in a log file, and extra copies of duplicate entries were discarded. The final multiple alignment contains 99 myoglobin sequences, and the nomenclature for helical position was taken from Dickerson and Geis (e.g. F4 is the fourth amino acid along the F-helix in sperm whale myoglobin) (27). Similar alignments were used for the ␣ and ␤ subunit sequences of vertebrate hemoglobins.

O 2 Binding and Rates of Hemin
Loss-Although the focus of this work is a structural interpretation of the effects of proximal pocket mutations on hemin loss (Table I), it is important to note that some of these amino acid replacements do have significant effects on ligand binding. Table II presents a list of the O 2 binding properties of all 14 mutants whose structures have been determined by x-ray crystallography. There is a progres- 1 The abbreviation used is: HPLC, high pressure liquid chromatography.   (3). The values for wild type and the Leu 104 mutants were measured for this study (20). In these later measurements, more variable rates of hemin loss from the wildtype controls were observed, and the average value for wild type was higher than the value of 0.01 h Ϫ1 reported in Hargrove et al. (3). The variability in the wild type k ϪH values at pH 7 is due to the slowness of the reaction that requires collecting time courses for over 24 h. Much less variability is observed for those mutants with higher rates of hemin loss.
Myoglobin  (Table II). In all four cases, there is no obvious correlation with hemin loss rate constants. A detailed interpretation of the O 2 binding parameters for a more complete set of proximal mutants will be presented elsewhere and has been discussed by Liong (20), Dou (28), and Scott et al. (34). None of these ligand binding effects are large in comparison with the dramatic decreases (Ͼ10-fold) observed for the resistance of these mutants to hemin dissociation. Native and wild-type recombinant sperm whale myoglobin lose hemin at rates of 1.0 and 0.01-0.05 h Ϫ1 at pH 5 and 7, respectively. Table I lists rates of hemin loss from 20 different proximal pocket mutants of myoglobin. The pH dependence of hemin loss from myoglobin is due to protonation of both His 93 , which facilitates disruption of the proximal imidazole-Fe bond, and His 64 , which then rotates out into solvent causing loss of distal water coordination (29). As a result, hemin loss at pH 5 is rapid; data collection can be completed within 5 h; and the rate constants are well defined. At pH Ն 7, protonation of His 93 is greatly inhibited; coordinated water is stabilized by hydrogen bonding to the neutral form of the side chain of His 64 ; hemin dissociation is much slower; and the time courses must be recorded for over 24 h. During these long incubation times, apoprotein denaturation often occurs, making data analysis difficult. As a result, the small rate constants observed at pH 7 (k ϪH Յ ϳ0.1 h Ϫ1 ) have much greater uncertainty (almost a factor of 2) than those for the more unstable mutants and those obtained from measurements at lower pH values.
Most of the proximal pocket mutations cause Ն10-fold increases in the rate of hemin loss (3). Removal of the proximal histidine base, as in the H93G mutant (30), results in an extremely high rate of hemin dissociation, k ϪH ϭ 660 and 140 h Ϫ1 at pH 5 and 7, respectively (Table I). This result is clearly due to the lack of a covalent link anchoring the heme group to the globin. Many of the Leu 89 mutants also show dramatic increases in the rate of hemin loss, with the order being: The rates of hemin loss from the His 97 mutants show both a size and a charge dependence with the order of k ϪH being: H97D Ϸ H97E Ϸ H97A Ͼ H97Q Ͼ H97V Ͼ Ͼ H97F Ϸ native/ wild type. Interestingly, replacing either Leu 89 or His 97 with a phenylalanine side chain is relatively conservative with respect of the rate of hemin loss, particularly at pH 5 ( Table I).
Replacement of Ile 99 with serine results in a large 40-fold increase in the rate of hemin loss, whereas mutation to alanine and valine results in only moderate increases. Leu 104 mutations also exhibit increases in the rate of hemin loss. The correlation with size is weak; however, the L104A mutation does cause almost a 20-fold increase in k ϪH at both pH values (Table I). The lack of effect of the polar L104N replacement on hemin loss at pH 5 is even more difficult to interpret without a structure of the mutant. Placement of an amide side chain in the proximal pocket would be expected cause an influx of water molecules, which should enhance disruption of the His 93 -Fe bond.
X-ray Structures-Crystals of all but one of the 14 metmyoglobin mutants grew in the hexagonal P6 space group with one molecule/asymmetric unit. The His 97 (FG3) to Asp mutant crystallized in the orthorhombic P2 1 2 1 2 1 space group, also with one molecule per asymmetric unit. Statistics for data collection and refinement are listed in Table III. The limit of diffraction (d min ) was 2.2 Å or better for all of the crystals, and the quality of the diffraction data, as assessed by the internal agreement of symmetry-related reflections (R merge ), was acceptable. Refinement of the models converged with reasonable crystallographic R factors (R cryst and R free ). There is good agreement between the calculated and observed structure factors, with an average value of 15.8% for R cryst . The difference between R cryst and R free was 6.1% or less. The deviations from ideal stereochemistry in the refined structures were also within normal limits, as summarized in Table IV.
The displacement of the iron from the heme plane varied slightly among the 14 mutants examined (Table V). However, there appears to be no correlation between Fe 3ϩ displacement and the rate of hemin dissociation from the met forms of these mutants. Table V also lists the bond length between the proximal imidazole ring and the heme iron (His 93 N⑀2 -Fe), showing that this parameter is virtually the same for all 14 mutants. Furthermore, the deviations in heme planarity observed in most of these proximal pocket mutants are of the same magnitude as those observed in wild-type and distal pocket mutants of sperm whale myoglobin (31). Thus, changes in heme planarity, imidazole-Fe 3ϩ bond distance and Fe 3ϩ displacement do not appear to be the underlying causes of the large increases in the rate of hemin loss.
The mutations at positions 89, 97, 99, and 104 produce only minor changes in the overall structure of the globin moiety. The major alterations are localized in the region of the mutated side chain and include significant changes in local water structure (Figs. 2-3). These local structural modifications were examined closely and used to interpret the changes in k ϪH .
Leu 89 Mutants-A space-filling model of wild-type myoglobin shows that Leu 89 is packed against the heme forming a significant part of the outer covering of the proximal portion of the heme pocket. The His 93 (F8) is hidden in a hydrophobic cavity that is protected from the solvent molecules located on the surface of the protein. Replacing leucine at position 89 with glycine creates a clear pathway for the entry of solvent into the heme pocket (Fig. 3B). Electron density indicative of at least three internal water molecules is seen in the ͉F o Ϫ F c ͉ omit map for L89G metmyoglobin (Fig. 2). The three internal water molecules are observed inside the Xe1 pocket of this mutant and are connected to four other water molecules located on the protein exterior (Fig. 3B).
The hydration and increase of polarity in the heme pocket of Gly 89 metmyoglobin greatly facilitates disruption of the His 93 -Fe bond. As a result, the rate of hemin loss from L89G myoglobin is greater than or equal to that of the H93G mutant where there is no covalent attachment between the heme and the globin. A similar dramatic increase in the rate of hemin loss is observed for the L89S mutation, where the small hydroxymethyl side chain probably also allows an influx of water and increases the polarity of the heme pocket. The L89A mutation also causes a substantial increase in k ϪH , but the effect is significantly smaller than that seen for the L89G and L89S mutants. Mutation of Leu 89 to an amino acid with similar size and polarity, such as L89F, has a much smaller effect on the rate of hemin loss, particularly at pH 5. As shown in Fig. 3C, the aromatic side chain of Phe 89 metmyoglobin forms a hydrophobic barrier to solvation which is equivalent to that of the native leucine side chain. Trp 89 also provides a steric barrier to solvation (Fig. 3D). However, the L89W mutation causes a 17-fold increase in the rate of hemin loss at pH 5, which is most likely due both to steric hindrance between the large indole side chain and the rigid porphyrin ring and to the increased polarity of the indole side chain.
His 97 Mutants-The location of His 97 (FG3) is shown in Fig.  4A where the protein molecule has been rotated 90°to the right relative to its position in Fig. 1. Replacing His 97 with small (H97A and H97V) or negatively charged (H97D and H97E) side chains results in large increases in k ϪH . The crystal structures of H97V and H97D show that the heme-7-propionate is no longer interacting with this side chain and that a channel has been created that connects the solvent exterior to the proximal histidine (Fig. 4, B and E, respectively). In both mutants, hydration of the heme pocket is significantly smaller than that seen for the L89G mutation, which almost completely exposes the proximal histidine to water. The smaller channel and extent of hydration explain why the effects of H97V and H97D mutations on hemin loss are less than those observed for the L89G, A, and S replacements.
Replacing His 97 with glutamine was expected to have little effect on the metmyoglobin structure because the amide side chain can still form a hydrogen bond with the heme-7-propionate. However, as shown in Fig. 4D, the glutamine side chain is not large enough to completely seal the opening to the hydrophobic cavity, resulting in significant increases in k ϪH at both neutral and acid pH (Table I). Phenylalanine can replace His 97 with very little effect on the rate of hemin loss at either pH. (Table I). In the crystal structure of this mutant, the phenyl ring is located in the same position as the naturally occurring imidazole ring (Fig. 4C). Phe 97 also appears to provide some electrostatic stabilization through interaction between the positive edge of the phenyl multipole and the heme-7-propionate. Thus, the native His 97 side chain provides both a  steric barrier to solvation and electrostatic stabilization of one of the heme propionates.
Ile 99 Mutants-The side chain at the FG5 position is important in maintaining the exact position of the porphyrin ring in the heme pocket. Replacing the native isoleucine with a smaller side chain results in tilting of the B ring toward the proximal side of the heme pocket because of the loss of the C␦ atom of the sec-butyl side chain (Fig. 5). Hargrove et al. (3) suggested that an apolar side chain at position 99 is important for keeping the heme pocket anhydrous and observed that the I99S mutation results in a 90-fold increase in hemin loss at pH 5. Mutation to alanine results in a more moderate 20-fold increase. Other than changes in the tilt of porphyrin ring, the crystal structures of I99A and I99V do not show significant alteration of the protein or hydration of the heme pocket with well defined internal water molecules. Thus, the causes of the high rates of hemin dissociation from the I99A, I99V, and I99S mutants are more ambiguous and could be due to the increase in internal space and/or changes in the tilt of the heme which weaken the Fe-His 93 bond.
Leu 104 Mutants-Replacing Leu 104 (G5) with alanine and valine causes moderate to small increases in the rate of hemin loss (from 20-to 5-fold, respectively). Ordered water molecules are found in the proximal portion of the heme pocket in the L104A and L104N mutants (Fig. 6A) and almost certainly account for the increases in k ϪH for these proteins. The presence of these interior water molecules was unexpected. There is no obvious connection to the solvent phase because Leu 89 still blocks direct access to solvent, making visualization of the internal waters difficult (Fig. 6A).
There are two possible pathways for entry of the water molecules seen in the L104A and L104N mutants. Solvent could enter the heme pocket through the distal histidine gate and then migrate to the proximal Xe1 site as is seen for the apolar ligands, O 2 and CO (9 -13, 34). Alternatively, water molecules could diffuse into the protein via multiple transient channels created by thermal fluctuations in the protein matrix (32). There are two well defined surface water molecules within hydrogen bonding distance of Gln 105 (G6), and when Leu 104 is replaced with Asn, a chain of three water molecules is found in this area and extends from the exterior surface of the amide side chain out into the solvent phase along the G helix.
In general, water molecules are not seen in the proximal and distal cavities of myoglobin because these spaces are lined with hydrophobic amino acids. However, in the L104A mutation, the proximal cavity is probably large enough to sequester two internal water molecules that appear to hydrate the interior edge of the porphyrin ring and the Xe1 pocket. The net result is a ϳ20-fold increase in the rate of hemin dissociation.
The amide side chain in the L104N mutation is highly polar. Two internal water molecules are attached to this side chain in the crystal structure of the metmyoglobin mutant. Remarkably, this mutation causes little increase in the rate of hemin loss at either pH 5 or 7. The water molecules in the L104N mutant appear to be attached strongly to each other and to the amide side chain by strong hydrogen bonds with O-O and N-O distances of 2.6 to 3.1 Å (Fig. 6B). Thus, the hydroscopic nature of the Asn side chain appears to prevent these internal water molecules from solvating the heme group and disrupting the proximal histidine-iron bond. In contrast, the more mobile water molecules present in the L104A mutant do facilitate disruption of the Fe-His(F8) bond (Table I). DISCUSSION The four proximal pocket amino acids examined appear to "waterproof " the heme pocket. They maintain a favorable apolar environment for the heme group and prevent the hydration of the proximal imidazole base. Leu 89 (F4) seals the hydrophobic cavity underneath pyrrole B, preventing solvation of the Xe1 site in the proximal portion of the heme pocket. Decreasing the size or increasing the polarity of the side chain at position 89 results in dramatic increases in the rate of hemin loss. There appears to be strong selective pressure to seal off the heme pocket as judged by the high degree of conservation at the F4 amino acid. The frequency of leucine at this position is 87% in the 99 known myoglobin sequences, and the remaining major variation is conservative (7% Phe). Leu(F4) is even more conserved in vertebrate hemoglobins showing a frequency of 99% in both the ␣ (279 sequences) and ␤ (224 sequences) subunits.
His 97 (FG3) is part of the extensive hydrogen-bonding network found at the exterior portion of the proximal pocket of myoglobin. Disruption of the hydrogen-bonding network by mutation of His 97 to smaller, apolar, or negatively charged side chains has little effect on the ligand binding properties of the protein (10,20,34). However, loss of favorable electrostatic interactions with heme-7-propionate or opening of a channel to the proximal histidine does accelerate the rate of hemin loss up to 40-fold.
Both Ile 99 and Leu 104 are located in the interior region of the heme pocket beneath the heme prosthetic group and are important in maintaining the position of the porphyrin ring, which in turn influences the reactivity of the heme iron toward ligands (10,20,34). Mutations at both positions have significant effects on the rate of hemin dissociation. Substitution with an alanine at position 104 enlarges the proximal pocket enough to allow the entry of water molecules that promote hemin loss. In the L104N mutant, solvent also enters the proximal cavity, but the water molecules are sequestered by the Asn side chain. In effect, the amide group acts as a hydroscopic agent to keep the Fe-His 93 "dry," preventing a large increase in the rate of hemin dissociation.
All of these results show that the proximal environment of the heme group must be kept anhydrous and that direct exposure to solvent must be minimized. Tilton et al. (8) have shown that there are internal cavities at both the proximal and distal sides of the heme group and visualized them by adding Xe to crystals and determining the binding sites. The most stable Xe site(Xe1) is circumscribed by Leu 89 , His 93 , Leu 104 , and Phe 138 . In the native protein, this cavity is too small and hydrophobic to accommodate water molecules. A small fraction of the apolar ligands O 2 and CO do reach this location on ns to s time scales after laser photolysis or under high pressure (9 -13, 34). However, if the Xe1 cavity is opened to solvent (L89G), enlarged (L104A), or made more polar (L104N), water does enter and destabilize the proximal His 93 -Fe bond.
The idea that solvation of the heme pocket facilities hemin loss was first proposed to account for the instability of Hemoglobin Boras (␤ Leu(F4) 3 Arg), which exhibits abnormally high rates of hemin loss, formation of semihemoglobin, and precipitation (14,27,33). The structures presented here show directly and convincingly that Leu 89 does protect the proximal pocket from hydration and subsequent loss of the heme pros- thetic group. Clearly, there is strong selective pressure to prevent hydration of the heme pocket, as can be seen by the high degree of conservation of large apolar side chains at the Leu 89 (F4), Ile 99 (FG5), and Leu 104 (G5) positions in myoglobin. At the FG5 position, the frequencies are 83% Ile, 12% Val, 3% Leu, and 2% others in the 99 known myoglobin sequence; those at the G5 position are 78% Leu, 17% Phe, and 5% other. In contrast, there is much more variability at the His 97 (FG3) position. In most cases, large side chains are present at FG3 to seal off the proximal pocket, and when polar side chains are present, they interact with the heme propionates.
The results described in this report confirm experimentally the necessity of a hydrophobic pocket to bind and retain the heme prosthetic group in myoglobin and show directly by crystallography the presence of water molecules around the heme when this hydrophobic pocket is disrupted. Furthermore, the character and contributions of particular proximal side chains on heme binding and retention have been dissected.