Crystal Structures of the Ferrous Dioxygen Complex of Wild-type Cytochrome P450eryF and Its Mutants, A245S and A245T INVESTIGATION OF THE PROTON TRANSFER SYSTEM IN P450eryF*

Cytochrome P450eryF (CYP107A) from Saccaropolyspora ertherea catalyzes the hydroxylation of 6-deoxy-erythronolide B, one of the early steps in the biosynthesis of erythromycin. P450eryF has an alanine rather than the conserved threonine that participates in the activation of dioxygen (O 2 ) in most other P450s. The initial structure of P450eryF (Cupp-Vickery, J. R., Han, O., Hutchinson, C. R., and Poulos, T. L. (1996) Nat. Struct. Biol. 3, 632–637) suggests that the substrate 5-OH replaces the missing threonine OH group and holds a key active site water molecule in position to donate protons to the iron-linked dioxygen, a critical step for the mo-nooxygenase reaction. To probe the proton delivery system in P450eryF, we have solved crystal structures of ferrous wild-type and mutant (Fe 2 (cid:1) ) dioxygen-bound complexes. The catalytic water molecule that was pos-tulated to provide protons to dioxygen is absent, although the substrate 5-OH group donates a hydrogen bond to the iron-linked

utilizing dioxygen (O 2 ) and electrons derived from NAD(P)H. P450s play an important role in steroid hormone biosynthesis, drug metabolism, xenobiotic detoxification, and steroid biosynthesis (2). Some bacterial P450s participate in the biosynthesis of pharmacologically important molecules. For example, cytochrome P450eryF (P450eryF, CYP107A) hydroxylates 6-deoxyerythronolide B (6-DEB) to erythronolide B at the C-6 atom as one step in erythromycin A production in Saccaropolyspora ertherea (3,4). Electrons provided from NAD(P)H via cytochrome P450 reductase or an iron-sulfur protein are required to reduce the iron-linked O 2 molecule to the peroxide level, whereas protons are needed to protonate the distal O 2 oxygen atom to promote heterolysis of the O-O bond. Much of what we know regarding the proton shuttle system in P450s derives from studies with P450cam (5,6). Thr 252 and Asp 251 are critical residues in an H-bonded network that enables water protons to be delivered to the iron-linked O 2 molecule (6 -9). P450eryF is unusual in that alanine (Ala 245 ) replaces the threonine found in most other P450s. However, the substrate 5-OH group is located close to the oxygen binding site, suggesting that the substrate 5-OH replaces the missing threonine OH group (10). Another unique feature of P450eryF is that a water molecule is located between the substrate and Ala 245 and is in position to donate a proton to an iron-linked O 2 molecule. This proton supply mechanism is consistent with mutagenic (11) and computational studies (12)(13)(14). Nevertheless, an important lesson from P450cam is that the H-bonded network and water structure changes significantly in the O 2 complex (15). Therefore, developing detailed mechanistic schemes based on P450 structures in the Fe 3ϩ state may be of limited value. In addition to the possibility of similar structural changes in P450eryF, the "catalytic" water in P450eryF is only 3.6 Å from the ferric iron so this water must reposition in the O 2 complex. To further probe the potentially unique features of the P450eryF proton relay system, we solved the structure of the WT O 2 ⅐P450eryF at 1.7 Å. We also solved the structure of oxy-complexes of two mutants, A245T and A245S, that have significantly reduced activity (11). The crystal structures of the O 2 ⅐P450eryF complexes, together with the O 2 -P450cam structure, provide important insights on the differences and similarities in P450 O 2 activation mechanisms.

MATERIALS AND METHODS
Crystallization-WT and mutants P450eryF were expressed and purified as reported earlier (16). The natural substrate 6-DEB was from KOSAN Bioscience (Hayward, CA). Crystallization of P450eryF followed published procedures (16). 1 l of mother liquor consisting of 0.1 M Tris-HCl, pH 8.5, 0.3 M sodium acetate, and 30 -35% polyethylene glycol 4000 was overlaid over 1 l of a WT protein drop without 6-DEB in the sitting drop vapor diffusion crystallization tray (Hampton Research, Aliso Viejo, CA). Fine needle clusters that formed in a few days were crushed into microcrystals and used for microseeding. Microseed-* This work was supported by National Institutes of Health Grant GM33688 (to T. L. P.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The ing was carried out 24 h after setting up crystallization sitting drops consisting of 0.1 M Tris-HCl, pH 8.5, 0.28 M sodium acetate, 22% polyethylene glycol 4000, and the protein solution containing 0.3 mM 6-DEB. Single crystals formed within a few days. The ferric (Fe 3ϩ ) substrate-bound WT crystals belong to space group P2 1 2 1 2 1 with unit cell dimensions of a ϭ 53.9 Å, b ϭ 78.64 Å, and c ϭ 97.21Å (Table I). The mutant crystals were prepared by cross-microseeding using substratefree WT crystals.
Preparation of Ferrous (Fe 2ϩ ) Dioxygen-bound Single Crystals-The following manipulations with the exception of dioxygen-soaking were carried out at room temperature. Oxidized crystals were transferred into cryobuffer (mother liquor with 20% glycerol) and then reduced with 10 mM sodium dithionite in a glove box (COY, Grass Lake, MI) with the dioxygen concentration kept Ͻ15 ppm. A clear color change from dark brown to bright red was apparent within 2 min. The crystals were kept in the dithionite solution for another 20 min. After reduction, crystals were washed thoroughly in deoxygenated cryobuffer for 10 min in the glove box. Reduced crystals then were soaked in the oxygen-saturated cryobuffer under atmospheric condition at Ϫ10°C. The crystal was quickly scooped up in a nylon loop and flash-cooled in liquid nitrogen in preparation for data collection.
Data Collection and Processing-All of the data were collected at the Stanford synchrotron radiation laboratory (SSRL) beamline 9-2 using a Quantum-315 CCD detector (Area Detector Systems Corp., Poway, CA). The wavelength used was 0.92 Å, and the beam was attenuated to 50% by using an aluminum filter to reduce photo-induced reduction of the dioxygen complex. Crystals were maintained at 100 K in a cryostream. Data were indexed, integrated, and scaled with HKL2000 (17).
Model Building and Refinement-For the WT O 2 ⅐P450eryF complex, rigid body refinement using CNS (18) and the ferric WT P450eryF structure was used as an initial model (Protein Data Bank code 1JIO). After a single cycle of energy minimization and B-factor refinement, both 2F o Ϫ F c and F o Ϫ F c electron density maps showed the presence of dioxygen coordinated to the iron atom. The dioxygen molecule then was added to the model, and the refinement continued. Bond angles and distances for the dioxygen-iron interactions were weakly restrained. The Fe-O bond was restrained to 1.9 Å using a force constant of 4.0 kcal/mol/Å 2 , whereas the Fe-O-O bond angle was retrained to 130°with a force constant of 35 kcal/mol/rad 2 . Water molecules were added by using the water-pick program of CNS with the threshold of 5.0 . After another cycle of refinement, several water molecules were added manually around the substrate. The graphic program O (19) was used for the model building. The same procedure was used for the mutant-oxy structures. Cavity volume calculations were done using VOIDOO (20).

RESULTS AND DISCUSSION
Active Site Structure of WT O 2 ⅐P450eryF-The simulatedannealing omit maps for the O 2 -WT complex clearly show dioxygen coordinated to the heme iron (Fig. 1). The Fe-O bond and Fe-O-O angle were refined to 1.81 Å and 128°, respectively. These values are very close to those determined from the P450cam-oxy crystal structure, 1.81 Å and 131° (15). These Fe-O bond distances also agree well with the 1.80-Å distance determined from the early extended x-ray absorption fine structure work on oxy-P450cam (21).
The distal oxygen atom is sandwiched between and is nearly equidistance from the Ala 241 and Ala 245 C ␤ atoms. The Ala 241 where I i is the intensity of an observation and ͗I͘ in the mean value for that reflection and the summations are overall reflections.
b Values for the highest resolution shell are within parentheses.
where F o and F c are the observed and calculated structure factor amplitudes, respectively. 5% of the reflections were randomly picked for the calculation of R free , and the same set of the test reflections was maintained throughout the refinement.
carbonyl O atom is 3.3 Å from the distal O atom. The C-6 atom of the substrate, which is the hydroxylation site, is 3.2 Å from the proximal oxygen atom. The substrate C-5 OH forms a 2.7-Å H-bond with the distal oxygen atom. These steric and H-bond interactions force the dioxygen ligand to point toward the I-helix cleft between Ala 241 and Ala 245 . Fig. 2 shows a superposition of O 2 ⅐P450eryF and oxidized P450eryF. Although the substrate slightly rotates about C-13, O 2 binding does not change the substrate-heme distances significantly, probably because of restrictions imposed by the Hbond between 5-OH and O 2 . This is different than in P450cam where diatomic ligand binding, CO and O 2 , displaces the substrate by ϳ1 Å further from the heme (15,22,23).
From the crystallographic analysis of the ferric enzyme, water molecule 519, which hydrogen bonds to the carbonyl of Ala 241 and the substrate 5-OH group, was proposed to be the direct proton donor to the dioxygen ligand (10). This view was supported by the A245S and A245T mutant structures that show a much higher mobility of water 519 and exhibit 16 and Ͻ1% wild-type activities, respectively (11). As shown in Fig. 2, the distal oxygen atom of O 2 is only 0.7 Å from water 519 in the ferric enzyme. Thus, it might be anticipated that water 519 moves into a position for direct H-bond donation to the distal oxygen atom, similar to what is observed in P450cam (Fig. 3) (15). However, there is no evidence for the presence of water 519. Although there is a small cavity near the O 2 molecule, this cavity is much smaller than in P450cam (77 versus 206 Å 3 ) and is too small to accommodate a water molecule. Therefore, the water molecule previously assumed to be the proton donor has been expelled from the active site, leaving no water directly H-bonded to the distal oxygen as in P450cam.
In P450cam, there is a substantial change in the conformation of the I-helix and the H-bonded water network that provides a continuous H-bonded link among the distal dioxygen O atom, water molecules, and key active site groups such as Thr 252 (15). In sharp contrast, there is little change in the I-helix or local H-bonded network in P450eryF. The H-bond network found in ferric P450eryF involving Glu 360 , Ser 246 , Ala 241 carbonyl, and waters 53 and 63 (Fig. 1) remains intact in the O 2 -complex with the exception that Wat519 has been expelled. We next probed the functional relevance of this hydrogen bond network by solving the structures of mutant-O 2 complexes that have reduced enzyme activity.
Active Site Structures of A245S and A245T Mutant O 2 ⅐P450eryF-The A245S and A245T mutants have substantially decreased enzyme activities, 16 and Ͻ1%, respectively (11). The O 2 -complex structures were solved to see whether the mutations resulted in perturbations in the local H-bonded network that could be related to the loss in activity. Even though Ala 245 has been replaced by residues of quite a different size and polarity at a critical location in the active site, the overall ligand geometry remains unchanged (Figs. 4 and 5). The similar conformation suggests that the other groups interacting with the ligand, the substrate, and Ala 241 are the key determinants for the dioxygen ligand conformation. As in the WT structure, the mutants maintain the same substrate 5-OH-O 2 H-bond and the C ␤ atom of Ala 241 forms one wall of the O 2 cavity, both of which force the ligand to point toward the I-helix.
Although the activity is decreased by 84% by the A245S mutation, the active site structure including water is essentially unchanged with the exception that the OH group of Ser 245 points toward the I-helix cleft to make new H-bonds with water 63 and the carbonyl group of the Ala 241 (Fig. 4). A possible source of the decrease in activity in the A245S mutant is a disruption in the strict H-bond donor/acceptor relationship in the proton transfer system because of the new H-bonds formed with the mutant Ser 245 side chain. In the O 2 ⅐A245T complex (Fig. 5), the larger threonine side chain results in larger and more functionally relevant structural changes. The 5-OH group of the substrate makes van der Waals contact with the C␥ atom of Thr 245 , pushing Thr 245 toward the I-helix cleft (Fig. 5C). As a result, the ␥-OH of Thr 245 occupies a similar position as water 63 in the ferric state and thus displaces water 63 from the active site. The O␥ atom of Thr 245 is too far (4 Å) from the O␥ atom of Ser 246 to make an H-bond. Therefore, the H-bond network is indeed disrupted at this site in this mutant (Fig. 5A). Thr 245 also contacts the distal O 2 O atom (Fig. 5C). These interactions and especially the disruption of the water H-bonded network by the displacement of water 63 probably disrupts efficient proton transfer to dioxygen, which explains the extremely low activity of the A245T mutant.
Also of interest is a comparison between the A245T mutant and P450cam (Fig. 6). In P450cam-oxy complex, Thr 252 moves closer to the O 2 ligand and donates an H-bond to the distal O 2 O atom. This movement of Thr 252 away from the I-helix and the associated change in the I-helix conformation allow additional waters to move into the I-helix groove and thus complete the H-bonded network to the dioxygen. However, Thr 245 in P450eryF cannot adopt the Thr 252 -like conformation because of steric interactions with the 5-OH group of the substrate. Thus, Thr 245 does not form an H-bond with the ligand to assist with dioxygen activation and instead inhibits the dioxygen activation process. To accommodate the larger substrate and to make the substrate 5-OH group available for the dioxygen activation, it appears that P450eryF evolved to replace the threonine with alanine.
Ortiz de Montellano and co-workers (24) have reported that A245T has higher activity than WT in testosterone hydroxylation. A possible cause of higher activity is that testosterone does not clash with Thr 245 , placing the threonine side chain in a similar position as P450cam, and recovers water 63. These plausible structural changes would enable proton supply through the WT-like hydrogen bond network and Thr 245 -assisted O 2 activation.
Dioxygen Activation by P450eryF-The only direct H-bond donor to dioxygen in P450eryF is the substrate 5-OH. The 5-OH group is critical, because conversion of the 5-OH group to a carbonyl eliminates enzyme activity (11). It has generally been assumed that the 5-OH group replaces the I-helix threonine found in most other P450s, and as the O 2 -P450cam structure shows (15), Thr 252 is in position to donate an H-bond to the distal oxygen atom, similar to the substrate 5-OH in P450eryF. However, the similar roles of Thr 252 and the P450eryf substrate 5-OH group came into question when Kimata et al. (25) reported that the substitution of the O ␥ proton with methyl group (Thr to O-Me Thr) retains significant hydroxylation activity, suggesting that the O␥ proton is not directly involved in catalysis. Although both P450cam and P450eryF have similar solvent H-bonded networks involving the I-helix, P450cam but not P450eryF has a water directly H-bonded to the distal oxygen. Although water 519 appeared to be an ideal candidate for the direct proton donor to dioxygen based on the Fe 3ϩ structure, once dioxygen binds the cavity, housing water 519 is simply too small to accommodate a water molecule and hence water 519 is expelled from the active site. Although both P450cam and P450eryF deliver protons from bulk solvent to dioxygen via an ordered network of H-bonds, the precise nature of the proton delivery network is substantially different, primarily because P450eryF uses the substrate itself as a proton donor and, unlike P450cam, water appears not to be a direct proton donor to dioxygen.
A possible proton delivery scheme is outlined in Fig. 7. The crystal structure shows that water 63 is too far to H-bond to dioxygen but it might be possible to transfer a proton via the carbonyl group of the Ala 241 , which is 3.3 and 2.6 Å from the distal oxygen atom and water 53, respectively. However, it is important to note that the species that accepts protons is not the oxy-complex, Fe(III)-O-O Ϫ , but rather the peroxy complex that forms after the second electron transfer step, Fe(III)-O-O 2Ϫ . A water situated between water 63 and dioxygen would form a direct water-dioxygen H-bond as in P450cam. The one caveat in this view is that the space between dioxygen and water 63 is too small (ϳ77 Å 3 ), compared with ϳ206 Å 3 in P450cam, to accommodate water. Hence, if water does bind here, there must be some rearrangement in the active site once the peroxy intermediate forms. In the absence of any information supporting such a change in structure, we propose the mechanism shown in Fig. 7. In this scheme, one proton is delivered from water 63 via the Ala 245 carbonyl. The most probable other proton source is the substrate 5-OH, because the 5-OH is the only direct H-bond partner for O 2 and the 5-OH proton is essential for the catalytic activity. The proton abstraction from 5-OH by O 2 transiently forms a substrate alkoxide, which is stabilized by the H-bond with the leaving water molecule. The energetic feasibility of forming an alkoxide is supported by density functional calculations that show that the iron-linked peroxy dianion has a higher proton affinity than the substrate alkoxide (14). After O-O bond fission, a leaving water molecule completes an H-bonded network that provides a proton to the substrate alkoxide via Glu 360 . The Fe(V)-O intermediate or its electronic equivalent next hydroxylates the C-6 atom to complete the catalytic event. One final problem is the ultimate source of protons, which for both P450cam and P450eryF is assumed to be bulk solvent. In both structures, there is no continuous H-bonded connection between the active site and bulk solvent. However, when the concerted motion of side chains and water is taken into account, the proton delivery pathway in P450cam does connect to the molecular surface (26). Thus, dynamics may well play a critical role in enabling bulk solvent protons to be delivered to the internal H-bonded network.

CONCLUSIONS
The ferric P450eryF structure suggested that a well ordered active site water molecule H-bonded to a substrate OH group serves as a direct proton donor to the iron-linked dioxygen (10). Although ferric P450cam has no homologous water molecule, the O 2 -P450cam structure does have a water molecule H-bonded to the Thr 252 OH group capable of donating protons to dioxygen (15). Therefore, the previous picture of O 2 activation mechanisms for these two P450s shares common features: a catalytic water molecule H-bonded to a substrate or side chain OH group capable of donating protons to dioxygen. In forming the O 2 ⅐P450eryF complex, however, the water molecule that makes an H-bond to the substrate is expelled from the active site, leaving no water capable of serving as a direct proton donor to dioxygen. Instead the substrate 5-OH is the only direct H-bond donor to the ironlinked dioxygen and most probably serves as a direct proton donor. In addition, the hydrogen bond network, water 63-Glu 360 -Ser 246 -water 53-Ala 241 carbonyl in the I-helix cleft, is proposed as the proton transfer pathway. Thus, it appears that the details of the proton shuttle machinery differ substantially in these two P450s. These results illustrate the limited picture on the structural underpinnings of catalysis that can be derived from resting state structures while emphasizing the importance of structural studies on critical intermediates.