Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency

doi:10.1016/j.jmb.2007.08.015 J. Mol. Biol. (2007) 373, 633–651

Filling a Hole in Cytochrome P450 BM3 ImprovesSubstrate Binding and Catalytic Efficiency

Wei-Cheng Huang†, Andrew C. G. Westlake†, Jean-Didier MaréchalM. Gordon Joyce, Peter C. E. Moody and Gordon C. K. Roberts⁎

Henry Wellcome Laboratories ofStructural Biology, Departmentof Biochemistry, University ofLeicester, Leicester LE1 9HNUK

Received 1 May 2007;received in revised form9 July 2007;accepted 7 August 2007Available online21 August 2007

*Corresponding author. E-mail addr† W.-C.H. and A.C.G.W. contribu

work.Present addresses: J.-D. Maréchal,

Química, Unitat de Química Física, Ude Barcelona, Edifici C.n., 08193 BelM. Gordon Joyce, Structural ImmunLaboratory of Immunogenetics, NatAllergy and Infectious Diseases, NaHealth, 12441 Parklawn Drive, RockAbbreviations used: P450s, cytoch

P450BM3, cytochrome P450 BM3 fro(CYP102A1); DMSO, dimethyl sulforoot-mean-square deviation.

0022-2836/$ - see front matter © 2007 E

Cytochrome P450BM3 (CYP102A1) from Bacillus megaterium, a fatty acidhydroxylase, is a member of a very large superfamily of monooxygenaseenzymes. The available crystal structures of the enzyme show non-productive binding of substrates with their ω-end distant from the iron ina hydrophobic pocket at one side of the active site. We have constructed andcharacterised mutants in which this pocket is filled by large hydrophobicside-chains replacing alanine at position 82. The mutants having phenyla-lanine or tryptophan at this position have very much (∼800-fold) greateraffinity for substrate, with a greater conversion of the haem iron to the high-spin state, and similarly increased catalytic efficiency. The enzyme asisolated contains bound palmitate, reflecting this much higher affinity. Wehave determined the crystal structure of the haem domain of the Ala82Phemutant with bound palmitate; this shows that the substrate is bindingdifferently from the wild-type enzyme but still distant from the haem iron.Detailed analysis of the structure indicates that the tighter binding in themutant reflects a shift in the conformational equilibrium of the substrate-free enzyme towards the conformation seen in the substrate complex ratherthan differences in the enzyme–substrate interactions. On this basis, weoutline a sequence of events for the initial stages of the catalytic cycle. TheAla82Phe and Ala82Trpmutants are also verymuchmore effective catalystsof indole hydroxylation than the wild-type enzyme, suggesting that theywill be valuable starting points for the design of mutants to catalysesynthetically useful hydroxylation reactions.

© 2007 Elsevier Ltd. All rights reserved.

Keywords: cytochrome P450; substrate binding; crystal structure; indole;fatty acids
Edited by I. Wilson
ess: [email protected] equally to this

Departament deniversitat Autònomalaterra, Spain;ology Section,ional Institute oftional Institutes ofville, MD 20852, USA.romes P450;m B. megateriumxide; RMSD,

lsevier Ltd. All rights reserve

Introduction

Cytochromes P450 are a superfamily, currentlynumbering more than 6000, of haem-thiolate mono-oxygenase enzymes, found in almost all forms oflife, which catalyse the activation of molecularoxygen1 and the addition of an atom of oxygen totheir substrate.2 There is considerable sequencediversity within the superfamily; those memberswhose structures have been determined share acommon overall fold, while differing markedly intheir active site architecture, leading to very diversesubstrate specificity. Members of the superfamilyinclude both enzymes of high specificity involved inthe biosynthesis of, for example, steroids (inmammals, insects, plants, fungi and bacteria3–6),and polyketide antibiotics,4 and also “de-toxifying”

d.

mailto:[email protected]

http://dx.doi.org/10.1016/j.jmb.2007.08.015

634 Substrate Binding in Cytochrome P450 Mutants

enzymes, including the well-known mammaliandrug-metabolising P450s,3 which have a broadsubstrate specificity, allowing them to metabolise avery wide range of compounds.The bacterial cytochromes P450 CYP101 (P450cam)

from Pseudomonas putida and CYP102A1 (P450BM3)from Bacillus megaterium have been extensivelystudied structurally and mechanistically.7–9 Theyhave also been used as the starting point for theconstruction of mutants with altered specificity foruse in chemical synthesis, where the ability ofP450s to insert an oxygen atom into an unactivatedC–H bond has potentially valuable applications.P450BM3, identified as a fatty acid hydroxylase,10,11

is a 119 kDa polypeptide which contains a P450domain and a diflavin NADPH-cytochrome P450reductase domain11,12 similar to that in the mamma-lian drug-metabolising mono-oxygenase system.Unlike most P450s, therefore, which require addi-tional electron transfer proteins, P450BM3 is cataly-tically self-sufficient. Perhaps because of this,P450BM3 also has the highest catalytic activity ofany P450 mono-oxygenase identified to date.P450BM3 catalyses the hydroxylation of C12–C16

saturated fatty acids at the (ω-1), (ω-2) and (ω-3)positions.10,11 The crystal structure of the substrate-free form of the haem domain13 reveals a longhydrophobic active site channel extending from theprotein surface, where there are charged residuessuitable for binding a carboxylate group, to thehaem iron, a structure entirely consistent with theobserved fatty acid hydroxylase activity. The crystalcontains two molecules in the asymmetric unit, onewith an open substrate access channel and onewhere it is slightly more “closed”; the twomoleculeshave different intermolecular contacts in the crystal,and it seems likely that in solution there is a dynamicequilibrium between these and perhaps additionalconformations of the access channel.13 In the crystalstructures of the protein bound to the substratespalmitoleic acid14 and N-palmitoylglycine15 theprotein is seen to have undergone a significantconformational change involving a displacement ofthe F and G helices. Studies of the A264E mutanthave provided evidence to support the idea thatthese “substrate-free” and “substrate-bound” con-formations, which are quite distinct from the “open”and “closed” structures seen in the crystal structureof the free enzyme, co-exist in equilibrium insolution in the absence of substrate.16 However, innone of the available structures of substrate com-plexes is the fatty acid bound with the (ω-1), (ω-2)and/or (ω-3) carbon atoms positioned close to theiron in a position for hydroxylation. Instead, the “ωend” of the fatty acid becomes sequestered in ahydrophobic pocket between phenylalanine resi-dues 81 and 87, created by rotation of the aromaticring of phenylalanine 87 by ∼90° and a rearrange-ment of nearby side-chains (notably isoleucine 263and leucine 437), with the terminal methyl group ofthe substrate in contact with alanine 82. In thisposition, the ω to ω-6 carbon atoms of the fatty acidare all between 7.5 Å and 10 Å from the iron centre,

too distant for hydroxylation. (Binding of substratesremote from the haem has also been observed incrystal structures of a number of eukaryoticP450s.17–19) All the available crystal structures ofP450 BM3 thus represent inactive states of theenzyme–substrate complexes. This is in sharp con-trast to P450cam, P450eryF and P450epoK,20,23where the natural substrates are bound such thatthe sites of hydroxylation are positioned within 5 Åof the iron, and little or no rearrangement isobserved relative to the substrate-free form. NMRrelaxation experiments24 suggested that laurate and12-bromo-laurate bind to P450BM3 in solution atroom temperature in a similar way to that seen forpalmitate in the crystal structures. However, recentevidence from solid-state NMR for differences in theenvironment of some active site residues, includingPhe81 and Phe87, in the N-palmitoylglycine complexof P450BM3 between room temperature and −30 °C,led to the suggestion,25,26 supported by moleculardynamics simulations,27 that this may reflect anequilibrium between a low temperature bindingmode which is seen in the crystal, and a differentmode of binding at room temperature. It remains tobe established whether the mode of binding seen inthe fatty acid–P450BM3 complexes in the crystal ison the catalytic pathway, or whether it represents a“dead-end” complex not normally populated underphysiological conditions.The existence of a hydrophobic pocket within the

active site cavity, but relatively distant from theiron, also has implications for attempts to engineerthe enzyme's activity towards novel target sub-strates. It seems likely that small hydrophobicmolecules could preferentially bind in the pocketbetween phenylalanine residues 81 and 87, resultingin non-productive complexes. There have beenmany reports of changing the substrate specificityof P450BM3 by substitution of phenylalanine 87by a smaller residue such as glycine, alanine orvaline.7,28–34 Such substitutions would destroy thehydrophobic pocket, but would also increase theactive site volume, which may result both in re-duced coupling of NADPH consumption to productformation, due to less efficient exclusion of waterfrom the active site, and in decreased regiospecifi-city of hydroxylation due to increased mobility ofthe substrate within the active site.We have sought to investigate this issue by filling

the hydrophobic pocket rather than destroying it.This we have done by substitution of alanine 82 bythe larger hydrophobic residues isoleucine, pheny-lalanine and tryptophan. Examination of the struc-ture suggests that these substitutions should excludeprogressively larger volumes of the hydrophobicpocket and, in the case of the aromatic substitutions,perhaps fill it altogether. The mutants A82F andA82W are observed to bind fatty acids orders ofmagnitude more tightly compared to the wild-type,and to show substantially increased catalytic effi-ciency. In addition, these mutants are found to beeffective catalysts of the oxidation of indole, result-ing in formation of indigo, suggesting that they may

635Substrate Binding in Cytochrome P450 Mutants

exhibit generally improved activity towards smallmolecules.

Results and Discussion

As shown in Figure 1(a), in the crystal structuresof substrate complexes of P450 BM3 the fatty acidchain binds in a hydrophobic pocket betweenphenylalanine residues 81 and 87, with the terminalmethyl group in contact with Ala82. We haveexamined the substitution of Ala82 by a range ofother residues. As shown in Figure 1(b)–(d), thispredicts that substitution by isoleucine, phenylala-nine and tryptophan would be expected to fillprogressively more of the pocket; the phenylalanine

Figure 1. The active site of the N-palmitoyl-glycine compmodels of the mutants with isoleucine (b), phenylalanine (c) orchain at position 82 is shown in yellow, and the van der Waspace-filling representation of the substrateN-palmitoylglycinein the crystal structure. The modelled structures in (b)–(dhydrophobic pocket discussed in the text and overlapping wi

and tryptophan substitutions in particular wouldlead to a substantial predicted overlap with the endof the fatty acid chain, suggesting that they wouldprevent the binding of the fatty acid in the modeseen in the crystal structures. We therefore made andcharacterised the A82I, A82F and A82W mutants.

Fatty acid binding

Unlike the wild-type enzyme and the A82Imutant, the A82F and A82W mutants as purifiedwere in a predominantly high-spin state, suggestingtight binding of a substrate-like molecule, presum-ably an endogenous fatty acid from the Escherichiacoli host cells. This could not be removed byextensive dialysis or by gel filtration. However, treat-

lex of the wild-type enzyme ((a) PDB code 1JPZ) and oftryptophan (d) in place of alanine at position 82. The side-als surface of the substrate-binding pocket in magenta. Ais superimposed in blue, occupying the position observed

) show the larger side-chains at position 82 filling theth the bound fatty acid.


ment of the purified protein with a small excess ofNADPH, followed by buffer exchange using exten-sive ultrafiltration to remove reaction products,resulted in conversion to a predominantly low-spin form. Final preparations of the A82F and A82Wmutants contained 20% and 25% high spin haem,respectively.In line with these qualitative observations, marked

differences in fatty acid binding were observedbetween the wild-type enzyme and the A82I mutanton the one hand, and the A82F and A82W mutantson the other (Figure 2; Table 1). Wild-type and A82Ienzymes exhibited Kd values of approximately250 μM for laurate, with approximately 50% con-version to high-spin haem, whereas the A82F and

Figure 2. Fatty acid binding to wild-type and mutant enzfatty acid binding for (a) wild-type enzyme, (b) A82I, (c) A8210 μM enzyme without substrate (blue), and in the presenBottom: Concentration dependence of the optical absorption cand (f) 0.25 μM A82F enzyme. Note the difference in enzbetween (e) and (f).

A82W mutants bound this substrate 600 to 800-foldmore tightly, with Kd values of approximately0.4 μM and N90% conversion to high-spin. Weattempted to study the kinetics of binding oflaurate to the enzyme by stopped-flow methods,following the changes in the UV/visible absorp-tion. However, in line with an earlier report onthe wild-type enzyme,35 the observed rate was toofast (N800 s−1) for accurate determination bystopped-flow for either the wild-type or the A82Fmutant.In the case of palmitate, both wild-type and the

mutant enzymes bound the substrate tightly andwith substantial conversion to high spin haem(Table 1), but noticeable differences were again

yme. Top: Changes in UV/visible absorption spectra onF, (d) A82W mutants. In each case spectra are shown force of 2 mM laurate (orange) or 10 μM palmitate (red).hange on laurate binding to (e) 2.5 μM wild-type enzymeyme concentration and in substrate concentration scale

Table 1. Binding and kinetic parameters for fatty acid binding to wild-type and mutant P450BM3

Enzyme

Laurate Palmitate

Kd (μM)a High spin (%) KM (μM)b kcat (s−1)b Kd (μM)a High spin (%)

Wild-type 270(±14) 53 265(±19) 28(±1) 0.14(±0.01) 87A82I 240(±17) 50 320(±16) 45(±1) 0.27(±0.01) 85A82F 0.34(±0.03) 92 b20 26(±1) b0.1 100A82W 0.43(±0.04) 93 b20 43(±1) b0.1 100

a From optical titration experiments; see Materials and Methods.b From measurements of NADPH consumption; see Materials and Methods.


observed between wild-type and A82I on the onehand, and the A82F and A82Wmutants on the other.Palmitate binding to A82F and A82W was too tightto determine by this method (Kdb0.1 μM) and wasaccompanied by complete conversion to high-spinhaem; in comparison wild-type and A82I gavebetween 85% and 90% high-spin. The evidencethat the endogenous fatty acid bound to the A82Fenzyme as isolated is palmitate (see below) suggeststhat this substrate has a dissociation constant ordersof magnitude below 0.1 μM, since it cannot beremoved by dialysis. These data thus demonstratethat replacing Ala82 by larger, rigid side-chainsleads to a marked increase in the affinity of theenzyme for fatty acids and also in the degree ofconversion from low-spin to high-spin iron. Bycontrast, the substitution with the larger but moreflexible isoleucine side-chain has essentially noeffect.

Catalytic activity towards laurate

The rates of NADPH consumption were measuredfor the mutant and wild-type enzymes with laurateas substrate; the catalytic activity against palmitatecould not be estimated for either mutant or wild-type, since the rate was too fast to be measured inthe standard assay and the protein concentrationcould not be lowered any furtherwithout dissociationof the catalytically functional dimer.36 The tighterbinding of fatty acids to the A82F and A82Wmutantswas reflected in decreased KM values, while turn-over rates were similar for all the mutants (Table 1).Assuming that KM∼Kd for A82F and A82W as forwild-type and A82I, the combination of much tighterbindingwith unchanged turnover numbers leads to amuch greater catalytic efficiency (kcat/KM) of the twomutants with aromatic side-chains at position 82,estimated as 8×107 M−1s−1 (A82F) and 1×108 M−1s−1

(A82W) as compared to 1×105 M−1s−1 for the wild-type enzyme.The products formed from the actions of the wild-

type and mutant enzymes on laurate were identifiedby NMR spectroscopy; all of the mutants gave amixture of (ω-1), (ω-2) and (ω-3) products, withonly small differences in product distribution. A82Fand A82I appeared to favour (ω-2) hydroxylationcompared to wild-type, whilst A82W gave a slightlyhigher proportion of (ω-1) hydroxylated product(Figure 3).

The A82W mutant exhibited relatively poorcoupling of NADPH consumption to productformation, as demonstrated by the significantamount of unmetabolised laurate remaining in thereaction mixture (Figure 3). Comparison of inte-grated peak areas allowed an estimate of approxi-mately 45% coupling, whereas the other mutantsgave much higher coupling efficiencies, estimated tobe N90% in all cases. A82W was also notably lessstable than the other mutants, being inactivated afteran average of ∼1000 turnovers per enzyme mole-cule. In view of the relatively poor coupling, thismutant may be inactivated by generation of reactiveoxygen species such as peroxide. Addition ofcatalase to the reactions did indeed increase thelifetime of the protein, although it did not affordcomplete protection of activity, suggesting that thatA82W may be inactivated by active oxygen speciesgenerated at the iron before they can diffuse out ofthe active site.Both the A82F and the A82W mutants showed

significantly higher rates of NADPH consumptionin the absence of substrate than did the wild-type orA82I enzymes (wild-type, 0.14 s−1; A82I, 0.19 s−1;A82F, 2.6 s−1; A82W, 5.6 s−1); this is most likely toreflect the fact that, as noted above, these twomutants exist to a significant extent in the high-spinstate in the absence of substrate.

Crystal structure of the fatty acid complex ofP450BM3 A82F

In order to provide structural explanations for themarked differences in properties of the A82F andA82W mutants as compared to the wild-typeenzyme, we determined the crystal structure of thehaem domain of P450BM3 A82F. To date allcrystallographic studies of P450BM3 have involvedthe haem domain of the enzyme rather than theintact enzyme, in which the flexibility of the twodomains may inhibit crystallisation. We thereforeexpressed and purified the haem domain of theA82F mutant. As with the full-length protein, theA82F haem domain was isolated in a high-spinform, suggesting again that a substrate-like mole-cule was bound; this could not be removed either bygel filtration or extensive dialysis (nor, in theabsence of the reductase domain, by NADPH-supported catalytic turnover). Its identity wasestablished by extracting a protein sample with

Figure 3. 1H NMR spectra ofthe products of hydroxylation oflaurate by wild-type and mutantP450BM3. The methyl region of thespectra is shown for ((a)–(d)) reac-tion mixtures containing (a) A82W,(b) A82F, (c) A82I, and (d) wild-typeenzyme, and (e) a laurate standard.The methyl resonances of laurateand its hydroxylation products areidentified as (1) 11-hydroxylaurate(δ=1.071, M=2, J=6.24 Hz), (2) 10-hydroxylaurate (δ=0.808, M=3,J=7.17 Hz), (3) 9-hydroxylaurate(δ=0.803, M=3, J=7.53 Hz), and(4) laurate (δ=0.775, M=3, J=6.68 Hz). The triplet partially over-lapping the 11-hydroxylauratemethyl resonance (δ=1.096, M=3,J=7.15 Hz) is from an impurity inthe NADPH, and is also present incontrol reactions containing nosubstrate.


dichloromethane, and analysis of the resultantsolution by liquid chromatography–mass spectro-scopy. The dominant component was identified aspalmitic acid, with m/z=255 and an identicalretention time to an authentic standard; spikingthe sample with authentic standard confirmed theidentification. The presence of this fatty acid in theactive site of A82F is unsurprising; as describedabove, palmitate binds extremely tightly to the full-length protein, and it has been identified as one ofthe three major fatty acids in the E. coli host cells.37

Since the palmitate could not be removed, wecrystallised the substrate complex of the haemdomain using conditions similar to those reported.16

Crystals were obtained in the P212121 space group,and diffracted to 2.8 Å resolution; in contrast toprevious P450BM3 crystals, six chains were identi-fied in the asymmetric unit. The structure wassolved by molecular replacement, using the wild-typeN-palmitoylglycine-bound structure (PDB code1JPZ) as a search model. Strong electron density wasobserved in the active site indicating the presence ofbound fatty acid, and the structure was therefore

refined with palmitate modelled in this position; theextent of the electron density was entirely consistentwith the identification of the bound fatty acid aspalmitate. The final model shows well-definedelectron density for both the substrate and activesite residues, including the phenylalanine side-chainat position 82.As can be seen in Figure 4(a), the structure of the

A82F mutant is very similar to that of the wild-typehaem domain complexed with N-palmitoylglycine(PDB code 1JPZ), and also to the palmitoleic acidcomplex (PDB code 1FAG). The six chains in the unitcell all adopt essentially the same conformation,with pairwise root-mean-square deviation (RMSD)values between 0.49 Å and 0.83 Å; the differencesbetween them are confined mainly to loop regionsand surface side-chains, particularly to regionswhich are known to move on substrate binding,including the A helix, the β1 region and the F–Gloop. There are, however, some noticeable differ-ences in the binding of the substrate between thedifferent chains in the crystal; these are discussedbelow.


The replacement of the small aliphatic side-chainof A82 with the bulky aromatic side-chain ofphenylalanine in the mutant is readily accommo-dated with minimal changes to the protein structurecompared to the two reported substrate complexesof the wild-type enzyme (RMSDb0.74 Å; Figure4(a)). The ring of Phe82 fits between Phe81 andPhe87, and as predicted fills the hydrophobic pocketinto which the fatty acid terminus binds in the wild-type structures (Figure 4(b)). This changes the shapeof the substrate-binding channel and reduces itsoverall length, forcing the substrate to adopt adifferent binding position, with the carboxylategroup closer to the protein surface (Figure 4(c)).This apparently results in disruption of a hydrogenbond normally formed between the substratecarboxylate group and tyrosine 51. Perhaps as aresult of this, there are distinct differences in thebinding position of the substrate between the sixchains in the asymmetric unit (see SupplementaryData). This is in sharp contrast to the wild-typestructures, where the four chains in 1FAG and thetwo chains in 1JPZ each exhibit almost identicalsubstrate binding modes. In A82F, the greatestdifference in substrate binding position is betweenchains E and F, with a 2.5 Å shift in the location ofthe carboxylate group between these two; thecarboxylate is also between 2 Å and 4 Å closer tothe surface than is that of palmitoleic acid in thewild-type complex. The considerable conforma-tional flexibility of the side-chain of arginine 47appears to help compensate for the repositioningof the carboxylate group between the variousstructures.The “filling in” of the hydrophobic pocket in the

substrate binding channel by the A82F mutationresults in a significant displacement of the fatty acid(ω-1) to (ω-3) carbon atoms, which are the sites ofmetabolism, from their positions in the wild-typestructures. The terminal methyl group of thesubstrate now points towards the side-chain ofE267, some 4–6 Å (depending on which chain iscompared) away from its location adjacent to A82 inthe wild type structure (Figure 4(c)). The keyamino acid residues in the binding site adoptessentially identical positions to those in the N-palmitoylglycine complex of the wild-type enzyme(Figure 4(c)). Notwithstanding the fact that thebinding position of the substrate has changed,there is virtually no change in the regiospecificityof hydroxylation.In the structure of the complex of the wild-type

enzymewithN-palmitoyl-glycine (PDB code 1JPZ15),there is no water molecule coordinated to the iron(consistent with a high-spin complex), but there is anordered water molecule positioned close to the ironand hydrogen-bonded to A264 and T268. In thepalmitoleic acid complex (PDB code 1FAG14), thereare no water molecules in the PDB file, but asimilarly located water molecule is seen in theelectron density map for two of the four chains. Thiswater molecule was identified by Haines et al.15 aspotentially crucial to the mechanism of proton

transfer in P450BM3. However, in the crystalstructure of the palmitate complex of the A82Fmutant there is electron density corresponding tothis water molecule only in one of the six chains(chain D) (Figure 4(b) and Supplementary Data). It isnot clear why this should be so, since there is spacefor this water in the structure and the potentiallyhydrogen bonding residues are in the same place; itmay be that the altered position of the terminalmethyl group of the substrate makes this locationmore hydrophobic and thus water binding weaker.In any case, the absence of this water in the structureof the A82F mutant, which has a greater catalyticefficiency than the wild-type enzyme, raises ques-tions about the mechanistic importance of a tightlybound water molecule near the iron. There is nodoubt that it is essential for a water molecule to haveaccess to this region of the active site close to thehaem, and probably for it to be oriented byhydrogen-bonding to T268,38 in order to participatein essential proton transfers, but it may not benecessary for it to be tightly bound in the oxidisedform of the complex.All available crystal structures of substrate com-

plexes of P450BM3 clearly represent inactive states ofthe enzyme–substrate complex. Unlike in otherstructures such as P450cam where substrate isbound with the position of metabolism close to theiron centre,20,21,39 in the complexes of wild-type orthe A264E mutant of P450BM3 with palmitoleicacid14,16 or of wild-type enzyme with N-palmitoyl-glycine15 the fatty acid is bound distantly from theiron. Similarly in the structure of the palmitatecomplex of the A82F mutant the fatty acid bindstowards the “roof” of the active site pocket (Figure 4(b)and (c)), and the distances from the iron centre to the(ω-1)–(ω-3) carbon atoms, while somewhat shorterthan in thewild-type complexes, are still between 7Åand 9 Å, suggesting that, as in the wild-type, thecrystal structure obtained represents an unproduc-tive mode of substrate binding. For none of thestructures of P450BM3 is there unambiguous evi-dence that the crystals are catalytically active; wide-spread attempts to reduce the enzyme in the crystalhave met with failure due to crystal cracking.However, since very similar structures of theenzyme–substrate complex are obtained for differentsubstrates, for wild-type and mutant enzyme and indifferent space groups, it seems likely that thesestructures do represent relevant complexes. Thusit appears that, at some point in the catalytic cycleafter the initial substrate binding, both wild-typeand A82F P450BM3 must rearrange to conforma-tions where the fatty acid substrate binds closer tothe iron. Indeed previous work in our laboratoryusing paramagnetic relaxation experiments hassuggested that substrates may move by as muchas 6 Å on reduction of the iron centre in the wild-type protein.40 The very similar product distribu-tions in wild-type and A82F suggest that in this“rearranged” conformation the substrate has asimilar orientation in both wild-type and mutantcomplexes.


McDermott et al.25–27 have linked this requirementfor a movement of the substrate into a “catalytic”position to the equilibrium between low-spin andhigh-spin states of the complex, corresponding tosix-coordinate (with an axial water ligand) and five-

Figure 4 (legend

coordinate states of the haem iron. Noting the factthat the low-spin – high-spin equilibrium is clearlytemperature dependent in the wild-type enzyme,26they have suggested that the crystal structure of theN-palmitoyl-glycine complex, in which the substrate

on next page)

Figure 4. Structure of the palmitate complex of thehaem domain of the A82F mutant of P450BM3. (a) Overallstructure of the palmitate complex of the A82F mutant(cyan) compared to theN-palmitoylglycine complex of thewild-type enzyme (PDB code 1JPZ; green). The haemgroup is shown in red, the palmitate bound to A82F inbrown and the N-palmitoylglycine bound to wild-type inpurple; the mutated Phe82 residue is highlighted inyellow. The F and G helices are indicated. (b) Stereo viewof the active site in the crystal structure of the A82Fmutant. The side-chain of the mutated Phe82 residue ishighlighted in magenta. The bound palmitate is in darkgreen, the haem in red and Tyr51, Phe87 and Phe81 arein yellow. The blue mesh is the electron density mapcalculated using the omit procedure (two cycles) inSFCHECK.57 The omit maps for all six chains in theasymmetric unit are shown in Supplementary Data. (c)Comparison of the substrate-binding site in the substrate-free and substrate bound wild-type enzyme with that inthe substrate-bound A82F mutant. Relevant substratecontact residues are shown. The A82F mutant withpalmitate bound (2UWH) is in pink, the wild-type withN-palmitoylglycine bound (1JPZ) is in cyan and the wild-type without substrate (2BMH) is in green.


is bound in a “distantly bound” position, corre-sponds to a low-spin state. As discussed above, inthis complex there is a water molecule very close tothe iron;15 however, the distance and angle betweenthe water and the iron, along with the lack of anybonding electron density between them, are moreconsistent with the original assignment of thecrystal structure as a high spin complex. McDermottet al.26,27 suggest that at room temperature, wherethe complex is in the high-spin state and wherecatalytic activity is observed, the substrate hasmoved close to the iron. It is clear from the availablestructures that the aromatic ring of phenylalanine87 appears to form a “barrier” between the subs-trate in its “distantly bound” position and the haemiron, and they propose that rotation of this ring is akey feature of the room temperature high-spincomplex.26,27

In the light of this evidence for a temperaturedependence of the low-spin – high-spin equilibriumand perhaps of the mode of substrate binding, weexamined the UV/visible spectra of the palmitatecomplexes of the wild-type and mutant enzymes asa function of temperature between 4 °C and 30 °C(Figure 5). The wild-type enzyme and the A82I

mutant showed significant shifts to low-spin as thetemperature was decreased, reaching an estimated50% low-spin at 4 °C for wild-type and 70% low-spin for A82I; however, in marked contrast therewas no significant temperature dependence of thespectra for the A82F or A82W mutants, whichremained completely high-spin at all temperaturesexamined. Taken together with the absence of awater molecule near the haem iron in the crystalstructure, this strongly suggests that the palmitatecomplex of the A82F mutant is high-spin both in thecrystal and in solution, and hence that the move-ment of the substrate into the “catalytic” position isnot coupled to the spin state change; rather, acomplete spin state change can occur with thesubstrate in the distantly bound position.

Structural origin of the tighter fatty acid binding

Experimentally we observe that fatty acids bind∼1000-fold more tightly to A82F than to the wild-type enzyme. However, the crystal structure of thepalmitate complex offers no clear-cut explanationfor this. Indeed the position of the fatty acid appearsless ordered than in the wild-type, presumably dueto disruption of the hydrogen bond normallyformed between the carboxylate and tyrosine 51,and this would be expected to result in weakerbinding. Substitution of alanine 82 with the largerphenylalanine residue is likely to exclude watermore efficiently from the complex, resulting insomewhat tighter binding due to hydrophobiceffects, but it seems unlikely that this effect wouldbe sufficient to explain the magnitude of theobserved decrease in Kd; experiments using cam-phor derivatives designed to fill in “holes” in theprotein–substrate complex of P450cam resulted inmuch more modest increases in binding affinity.41

Since comparison of the substrate-bound struc-tures provides no clear-cut explanation for the tighterbinding to the A82F mutant, we need to considerin more detail the complete substrate bindingprocess, and in particular the conformational changewhich occurs on substrate binding to wild-typeP450BM3.14,15 Examination of the crystal structureof the substrate-free wild-type enzyme shows thatalanine 82 is surrounded closely by phenylalanineresidues 81 and 87 and isoleucine 263, such thatthere appears to be insufficient space to tolerate thesubstitution to phenylalanine (Figure 6). In thestructure of the substrate complexes of the wild-type enzyme, the alkyl chain of the fatty aciddisplaces the side-chains of isoleucine 263 andleucine 437, rotating them towards the F and Ghelices, and directly resulting in the repacking anddisplacement of this structural unit; the F helix isdisplaced about half a turn along its length, and byabout two-thirds of its width laterally, between thetwo conformations. As noted above, the structure ofthe palmitate complex of the A82F mutant is verysimilar to that of the substrate complexes of thewild-type enzyme, and this is specifically true of thisregion of the structure and of Ile263. Examination of

Figure 5. Temperature dependence of the optical absorption spectra of the palmitate complexes of wild-type andmutant P450BM3. Spectra are shown for (a) wild-type enzyme, and (b) A82I, (c) A82F, and (d) A82W mutants. In eachcase, spectra shown were recorded at 30 °C (red), 20 °C (orange), 10 °C (light blue), and 4 °C (dark blue).


these structures suggests that the bulky side-chain ofphenylalanine 82 itself would induce the sameconformational change in isoleucine 263 (andhence in the F and G helices) in the absence of thesubstrate (Figure 6). The structural evidence thusindicates that, in the absence of substrate, P450BM3A82F would adopt a conformation closely resem-bling the substrate-bound form (as observed for theA264E mutant16). The substantial increase in bind-ing affinity for fatty acids observed as a result of thesubstitution of Ala82 by phenylalanine would thusarise not from any change in interactions betweenthe substrate and the enzyme as a result of themutation, but rather from a shift in the conforma-tional equilibrium in the free enzyme. By contrast,examination of the structure indicates that substitu-tion of alanine 82 by isoleucine can be accommo-dated in the substrate-free conformation and wouldnot perturb the conformational equilibrium; thiswould be consistent with the observation that theA82I mutant resembles the wild-type in many of itsproperties.

The substrate binding process

Combining the information from the present workwith that already available from structural, spectro-scopic and simulation studies allows us to postulatea sequence of events for the early stages of thecatalytic cycle of P450BM3. First, we suppose thatthe free enzyme exists in an equilibrium betweentwo conformations, corresponding to the “substrate-

free” and “substrate-bound” conformations. Thesubstrate binds preferentially to the substrate-boundconformation; the data for the A82F and A82Wmutants suggest that this preference must be severalhundred-fold, although there is insufficient informa-tion available to provide a precise value. In thisinitial complex, the substrate is “distantly bound”,with the (ω-1) to (ω-3) carbon atoms 7 −9 Å from theiron. In spite of this, the spin state equilibrium isshifted over towards the high-spin state, almostcompletely in the case of palmitate binding. Thisspin state shift is attributed to displacement of thewater molecule from the sixth coordination positionof the iron. The substrate itself is clearly too far fromthe iron to displace the water directly; however, thebound substrate makes contact with the side-chainof Phe87, leading to a reorientation of this side-chain14,15 which would in turn displace the water.There is space for the displaced water to remainwithin the active site, where it is likely to play a rolein proton transfers, but from the structure of theA82F complex it appears that tight binding of thedisplaced water is not essential. If the low-spin –high-spin transition arises from the rotation ofPhe87, what is it that prevents this from occurringin the absence of substrate? In the substrate-freeconformation of the enzyme, Leu437 extends intothe active site and approaches Phe87 quite closely;its two methyl groups appear to lock the ring ofPhe87 in the “vertical” position, preventing it fromdisplacing water, and keeping the iron low-spin.Substrate binding and the conversion to the subs-

Figure 6. Displacement of Ile263 on substrate binding. (a) Wild-type N-palmitoylglycine-bound (PDB code 1JPZ); (b)wild-type substrate-free (PDB code 2BMH); (c) A82F palmitate bound; (d) A82F modelled into the wild-type substrate-free structure, demonstrating the clash between F82 and L263 in this conformation. Residue I263 is highlighted in darkblue and residue 82 in yellow. The other residues shown are F81 and F87 above and below (A/F)82, and residues on the Fhelix, F173, M177 and L181.


trate bound conformation displaces Leu437 signifi-cantly away from Phe87, such that the ring can nowrotate and displace the water ligand from the iron. Inthe A82F and A82W mutants, the substitution itselfleads to displacement of Ile263 and hence Leu437,and allows rotation of Phe87; as a result, for boththese mutants there is a significantly larger propor-tion of high-spin iron in the absence of substrate(20%–25%), together with a greater “leak rate” ofNADPH consumption in the absence of substrate.The next step in the catalytic cycle is the essential

movement of the substrate into a position closer tothe iron, appropriate for hydroxylation, and againPhe87 appears to play a key role. Examination ofthe structure suggests that this residue forms a“barrier” to this movement, and support for thiscomes from studies of isotope effects on the hydro-xylation of small substrates by wild-type P450BM3

and the F87A mutant.42 Molecular dynamics simula-tions provide evidence for the coupling of themovement of the substrate and the movement ofthe side-chain of Phe87.27 This central role for Phe87is consistent with a number of reports of substantialeffects of mutation of this residue on substratespecificity.28,31,32,34,43 Further work aimed at pro-viding support for this postulated sequence ofevents is in progress.

Oxidation of indole by A82 mutants

Unlike the wild-type enzyme or the A82I mutant,cells expressing P450BM3 A82F or P450BM3 A82Wwere observed to produce an insoluble blue dyeduring normal growth. This has been previouslyobservedwith certain human P450 isoforms, notablyCYP2A644,45 andwith theF87VmutantofP450BM3,46


and attributed to the formation of indigo byoxidation of indole either present in the growthmedium or endogenous to the E. coli host cells. Thesimple colorimetric nature of this phenomenon hasalso led to its use in screening libraries of randommutants to identify variants of P450BM3 andCYP2A6 with enhanced indole hydroxylationactivity.46,47 Wild-type P450BM3 has been reportedto produce detectable amounts of indigo only at highpH in a reaction driven by cumene hydroperoxiderather than NADPH.48 We therefore investigatedfurther the interaction of the purified A82 mutantenzymes with indole.The binding of indole to wild-type and mutant

P450BM3 gave type I spectral changes in the haemSorêt UV/visible absorbance, with a decrease inabsorbance at 420 nm and an increase at 390 nm,indicative of displacement of the axial water ligandfrom the haem iron and conversion to the high-spinstate (Figure 7). For the wild-type enzyme and theA82I mutant, plots of Δ(A390-A420) versus indoleconcentration demonstrated weak binding, with

estimated apparent dissociation constants of∼10 mM or greater; accurate values could not beobtained due to the limited solubility of indole. Bycontrast, the A82F and A82W mutants, which alsogave typical type I substrate-binding spectra atindole concentrations ≤10 mM, clearly bind indolemuch more tightly. Data in the concentration rangeup to 8 mM indole were best fitted by a two-sitebinding model, with apparent Kd values for the tightbinding site of ∼0.1 μM, similar to the concentrationof protein used in the experiment, and Kd values forthe second site ∼1000-fold higher (Table 2). Even theweaker site shows indole binding at least 35 to 65-fold tighter than that observed for wild-typeenzyme. The two sites contributed approximatelyequally to the final overall absorbance change, andthe magnitude of this change showed that the totalspin state changes on indole binding were substan-tial, approximately 80% for A82Wand 95% for A82F,indicating efficient displacement of the axial waterligand from the haem iron. On addition of highconcentrations (∼18 mM) of indole to the A82F or

Figure7. Indole binding towild-type and mutant enzyme. Changesin UV/visible absorption spectra ofwild-type enzyme (top) and A82Fenzyme (bottom) on indole binding.Spectra are shown for enzyme with-out substrate (blue), and in the pre-sence of 8 mM indole (orange) or18 mM indole (red).

Table 2. Indole binding and turnover by wild-type and mutant cytochrome P450BM3

Enzyme

Indole bindinga Indole turnoverb

Kd/Kd1 Kd2 KM (mM) nH ΔVmax (s−1)

Wild-type 11(±5) mM – 16(±5) 1.7(±0.3) 3.7±0.9A82I 13(±7) mM – 14(±4) 1.8(±0.4) 5.5±1.4A82F 0.1(±0.04) μM 317(±91) μM 3.4(±0.9) 1.2(±0.2) 20.8±2.6A82W 0.08(±0.05) μM 168(±62) μM 0.2(±0.1) 0.6(±0.1) 25.9±2.8

a From optical titration experiments; see Materials and Methods.b Frommeasurements of NADPH consumption, fit by the modified Hill equation as described in the Materials andMethods. Note that

ΔVmax is the increase in the rate of NADPH consumption on addition of saturating indole concentrations, not the rate of productformation; for the latter, see Table 3.


A82W mutants a further spectral change wasobserved (Figure 7), with a shift in the Sorêt peakto 424 nm to give a type II binding spectrum,typical of inhibitor complexes involving directnitrogen coordination to the haem iron, suggestingweak binding directly to the haem iron. Thesubstitution of alanine 82 by larger, rigid side-chains in the A82F and A82W mutants thus leads toa very marked increase in affinity of the enzyme forindole. Studies of indole turnover demonstrate thatthis is accompanied by a substantial increase incatalytic efficiency.Indole hydroxylation by P450 enzymes generally

gives rise to a complex mixture of soluble andinsoluble products45,46 (see Supplementary Data).Initial hydroxylation may take place at either the 2-position to give oxindole, or at the 3-position to givethe rather unstable indoxyl, which dimerisesthrough a non-enzymatic pathway to give theinsoluble dye indigo. In addition, indoxyl may befurther oxidised, most probably non-enzymatically,to isatin, which can then form a heterodimer withindoxyl to generate the insoluble dye indirubin. Thesoluble and insoluble products from the action ofwild-type and mutant P450BM3 on indole wereseparated by centrifugation. The soluble com-pounds were characterised by HPLC, while thewater-insoluble dyes were redissolved in dimethylsulfoxide (DMSO) and analysed by TLC, opticalspectroscopy and 1H-NMR; details of the analysesare given in Supplementary Data. The predominantsoluble product was oxindole, with trace quantities

Table 3. Product formation and NADPH consumption dcytochrome P450BM3

EnzymeIndigo

formeda (μM)Oxindole

formeda (μM)Coupling of p

NADPH co

Wild-type 2.5(±0.9) 7(±0.6)A82I 11(±2.8) 12(±0.4) 1A82F 41(±8.8) 34(±2.0) 3A82W 22(±2.3) 48(±1.1) 3

a Reaction mixtures contained 1 μM enzyme, 8 mM indole and 300were measured on completion of the reaction. Indigo formationredissolved in DMSO and oxindole formation by peak area integrationand Methods and Supplementary Data).

b Since one molecule of indigo is formed from two molecules of the100×(([oxindole formed]+2×[indigo formed])/[NADPH consumed])

c Calculated from the NADPH consumption rate and the percentag

of isatin, while the major water-insoluble productwas indigo, with only a small amount of indirubin.For wild-type enzyme and the A82I mutant, plots

of the rate of NADPH consumption versus indoleconcentration did not show complete saturation by18 mM indole, consistent with the low affinity forindole indicated by the optical titration experiments.The data were best fitted by the Hill equation(modified to account for the non-zero rate ofNADPH consumption in the absence of substrate),leading to estimated Hill coefficients of 1.7 in bothcases. The kinetic parameters for the wild-typeenzyme (Table 2) are in agreement with thosepublished.48 For the A82F and A82W mutants anincrease in NADPH consumption rate with increas-ing indole concentration was observed at concen-trations up to approximately 8 mM, above whichmarked inhibition occurred. This is consistent withthe optical binding studies, which demonstrateformation of a type II complex at these concentra-tions, typical of inhibitor binding. Estimating thekinetic parameters from the data at indole concen-trations below 8 mM, it is clear that the KM valuesare lower (by a factor of 4.7 for A82F and 81 forA82W) and the Vmax values higher, by a factor of 6,for A82F and A82W than for the wild-type; inaddition, for these mutants the Hill coefficient isnot significantly different from one (Table 2). Littleincrease in the NADPH consumption rate over thatobserved in the absence of substrate was observedat low indole concentrations, suggesting at most amodest contribution to the turnover rates by the

uring indole hydroxylation by wild-type and mutant

roduct formation tonsumption (%)b

Rate of NADPHconsumption (s−1)

Rate of productformation (s−1)c

4(±1.4) 1.1 0.031(±2.6) 1.5 0.199(±9) 16 6.01(±8) 23 5.7

μM NADPH in 50 mM potassium phosphate (pH 8.0); productswas estimated from UV/vis absorbance of insoluble productsof the HPLC chromatogram of the soluble products (see Materials

precursor product indoxyl the percentage coupling is defined as:.e coupling.


very tight indole binding site observed by opticaltitration.The coupling of product formation to NADPH

consumption in the hydroxylation of indole wasdetermined by measuring the formation of oxindoleand indigo in reaction mixtures containing 8 mMindole and 200 μM NADPH; by ignoring the smallamounts of isatin and indirubin formed, this willslightly underestimate the degree of coupling. Asshown in Table 3, very little product was generatedby wild-type enzyme or the A82I mutant, and thecouplingwas very poor. The substitution of Ala82 byphenylalanine or tryptophan, on the other hand, ledto marked increases in the coupling of productformation to NADPH consumption, and to a 170-fold increase in the rate of product formation. Takentogetherwith the≥25-fold increase in kcat/KMvaluesestimated from the measurements of NADPH turn-over, it is clear that the A82F andA82W substitutionslead to a substantial increase in the efficiency ofindole hydroxylation by the enzyme.

Conclusions

Cytochrome P450BM3 is well-known to catalysethe hydroxylation of C12–C16 saturated fatty acidsat the (ω-1), (ω-2) and (ω-3) positions.10,11 However,one of the continuing puzzles with regard to thisenzyme is the fact that in none of the availablestructures of substrate complexes is the fatty acidbound with these carbon atoms positioned closeenough to the iron for hydroxylation. Instead, the “ωend” of the fatty acid becomes sequestered in ahydrophobic pocket between phenylalanine resi-dues 81 and 87, with the carbon atoms which arehydroxylated all between 7.5 Å and 10 Å from theiron centre, clearly too distant for hydroxylation.This is in contrast to enzymes such as P450cam andP450eryF,20–22 where the natural substrates arebound in such a way that the sites of hydroxylationare positioned within 5 Å of the iron. In an attemptto understand the significance of this hydrophobicpocket in substrate binding, we have made a seriesof mutants at alanine 82 in which the pocket ispredicted to be filled by hydrophobic side-chains insuch a way as to prevent the binding of fatty acids inthe way seen in structures of the wild-type enzyme.The structure of the A82Fmutant in its complex withpalmitate showed that the pocket had indeed beenfilled by the side-chain of Phe82 and the mode ofbinding of the fatty acid had indeed changed.However, rather than finding that the ω end of thefatty acid was now binding closer to the iron, weobserved a shift in the position of the fatty acidtowards the surface of the protein so that the (ω-1),(ω-2) and (ω-3) carbons remained 7 Å −9 Å from theiron. This clearly indicates that this “distant”binding position does not result simply from theavailability of the hydrophobic pocket betweenPhe81 and Phe87.Nonetheless, the A82F and A82W mutants, in

which this pocket has been filled with large rigid

side-chains, showed a remarkable increase inaffinity for fatty acids, to the extent that endogenouspalmitate could not be removed by extensivedialysis, and in catalytic efficiency. It is unusual fora simple single-site substitution to lead to such amarked (∼800-fold) increase in substrate affinity,but detailed comparison of the mutant and wild-type complexes did not reveal any differences in theenzyme–substrate interactions which are likely toaccount for this. The most likely explanation is thatthe substitution of the large side-chains at position82 favours the substrate-bound conformation, seenin the complexes with palmitoleate and N-palmi-toyl-glycine and in the substrate-free A264E mutant,over the substrate-free conformation, due to inter-actions of Phe82 or Trp82 with Ile263, leading in turnto movements of the F and G helices. As discussedby Joyce et al.16 the free enzyme may exist in anequilibrium between these two conformations. Bychanging the position of this equilibrium in theabsence of substrate, the mutations will decrease theproportion of the binding energy used up in shiftingthe equilibrium on substrate binding, leading to anet increase in affinity.The observation of the formation of a blue

pigment by cells expressing the A82F or A82Wmutants of P450BM3 was the first indication thatthese mutants might be able efficiently to hydro-xylate indole, with the formation of indigo, and thiswas unambiguously confirmed by studies of theisolated enzymes. The data suggest that, whilst bothwild-type and A82I can bind indole, albeit weakly, itadopts a predominantly unproductive bindingmode in the active sites of these proteins, since theamount of product formed is so low. It is possiblethat a significant proportion of the indole bound tothese enzymes is located in the hydrophobic pocketbordered by Phe81, Phe87 and residue 82. In themutants in which this pocket is filled by a largerside-chain at position 82, the efficiency of hydro-xylation of indole is markedly increased, due both toincreased kcat/KM values and to increased couplingbetween NADPH consumption and product forma-tion. The evidence suggests that indole binding toP450BM3 is complicated, involving the binding ofseveral indole molecules in catalytically productive,non-productive and inhibitory positions; there isother evidence for the simultaneous binding of morethan one substrate molecule to P450BM3 and itsmutants.49,50 Notwithstanding these complications,the A82F and A82W mutants are clearly much moreeffective catalysts of indole hydroxylation than is thewild-type enzyme. There are two possible contribu-tions to this improved catalytic efficiency. First, thesubstitution will lead to the removal of the potentialnon-productive binding site in the hydrophobicpocket; simple docking calculations (Figure 8) doindeed suggest that the preferred binding site forindole in the wild-type enzyme is the hydrophobicpocket, while in the A82F mutant it prefers to bindmuch closer to the haem, with the 2 and 3-positionsclosest to the iron, in position for hydroxylation.Secondly, if, as suggested above, the mutation leads

Figure 8. Docking of indole into the active site of P450 BM3. Indole is shown as a space-filling model, docked into theactive site of (a) the wild-type enzyme (1JPZ) and (b) the A82F mutant (2UWH). Side-chains of F81 (yellow), F82/A82(pink) and F87 (green) are shown. Docking was performed by using GOLD3.1.1.62


to a shift in the conformational equilibrium of theenzyme, this will increase the affinity for indole andin addition, with its effect on the environment ofPhe87, will promote conversion to the high-spinstate and hence the rate of electron transfer.Whichever is the more important contribution, thepresent results suggest that these mutants are notonly very efficient fatty acid hydroxylases but alsoefficient hydroxylases of small hydrophobic mole-cules and hence may be useful biocatalysts inorganic synthesis51,52 and perhaps in the biodegra-dation of polycyclic aromatic hydrocarbons.28,53

Materials and Methods

Materials

The QuikChange XLmutagenesis kit was obtained fromStratagene, UK, and oligonucleotide primers from theProtein and Nucleic Acid Chemistry Laboratory, Univer-sity of Leicester, UK. Restriction enzymes were obtainedfrom New England Biolabs. Chromatography columnsand media from were obtained from Amersham Bios-ciences, Complete™ Protease inhibitors from Roche andindirubin from Biomol International LP, UK. All otherchemicals, of analytical grade or higher, were from SigmaAldrich UK Ltd.

Site-directed mutagenesis

Wild-type and mutant proteins were expressed usingthe plasmid pGLWBM3, encoding full-length P450BM3,which was a kind gift of Professor L.-L. Wong, Universityof Oxford, UK. The plasmid was modified to incorporate anumber of silent restriction sites to facilitate easyrecombination of mutants in regions of the gene corre-sponding to the various substrate recognition sites, and toallow simple excision of the region encoding the reductasedomain, thus providing a straightforward route to expres-sion of the haem domain of mutants of interest forstructural studies (full details are given in SupplementaryData). Mutagenesis was carried out using the Stratagene

QuikChange XL kit according to the manufacturer'sinstructions. Primers used were as follows (mismatchesare underlined):

A82I F: 5′-GCTTAAATTTGTACGTGATTTTATCGGA-GACGGGTTATTTACAAGC-3′A82I R: 5′-GCTTGTAAATAACCCGTCTCCGATAAAAT-CACGTACAAATTTAAGC-3′A82F F: 5′-GCTTAAATTTGTACGTGATTTTTTCGGA-GACGGGTTATTTACAAGC-3′A82F R: 5′-GCTTGTAAATAACCCGTCTCCGAAAAAAT-CACGTACAAATTTAAGC-3′A82W F: 5′-GCTTAAATTTGTACGTGATTTTTGGGGA-GACGGGTTATTTACAAGC-3′A82W R: 5′-GCTTGTAAATAACCCGTCTCCCCAAAAAT-CACGTACAAATTTAAGC-3′

All mutant genes were fully sequenced to confirm theabsence of any undesired mutations.

Protein expression and purification

Protein was expressed and purified using a modifica-tion of methods described;28 the full-length protein andthe haem domain were purified in essentially the sameway. Briefly, E. coli JM109 cells harbouring pGLWBM3were grown in 2×YT medium containing 100 μg/ml ofampicillin at 30 °C for 6–8 h to mid-log phase. The growthmedium was supplemented by the addition of 0.5 mM δ-aminolaevulinic acid, 25 mM ferric citrate and 1 ml/l of asaturated solution of riboflavin, and incubation wascontinued for 12–16 h. Cells were harvested by centrifuga-tion, resuspended in buffer A (50 mM potassiumphosphate, 1 mM EDTA, 1 mM benzamidine, 1 mMDTT (pH 7.5)) supplemented with Complete™ Proteaseinhibitors, and stored at −20 °C until processed. All of theA82 mutants expressed well in E. coli, producing levels ofpolypeptide similar to those seen for the wild-typeenzyme. Cells were lysed by sonication, and the solublefraction was loaded onto a DEAE-Sepharose Fast-Flowcolumn pre-equilibrated with buffer A. The protein waseluted using a gradient of 0–0.5 M KCl in buffer A, andfractions exhibiting the highest haem content pooled, andconcentrated and desalted by ultrafiltration. This was thenloaded onto a HiLoad Q-Sepharose 26/10 HP column pre-

Table 4. Statistics of data collection and refinement

Space group P212121Cell dimensions a=116.8 Å α=90°

b=147.0 Å β=90°c=183.4 Å γ=90°

Data range (Å) 2.80–73.52 (2.80–2.95)Number of measurements 440,872 (57,179)Number of independent reflections 76,791 (10,567)Mean ((I)/sd(I)) 10.6 (2.5)Rmerge (%) 0.185 (0.638)Data completeness (%) 98.0 (93.7)Number of non-hydrogen atoms 23,027Number of water molecules in the

asymmetric unit521

Number of reflections used in refinement 72,833Rwork 0.22Rfree 0.30ESU based on free R value (Å) 0.482RMSD in bond length (Å) 0.008RMSD in bond angle (°) 1.383Missing residues in chains A, B, C, D, E, F 459–472Analysis of Ramachandran plot (using

PROCHECK)63 (%)Most-favouredregions 86.5

Additional allowedregions 11.6

Generously allowedregions 1

Mean B factor (Å2) 22.8

Values in parentheses refer to the outer bin.


equilibrated with buffer A, and eluted with a gradient of0.2M–0.5 M KCl in buffer A. Again fractions with thehighest haem content were pooled, and concentrated anddesalted by ultrafiltration. Finally, to remove any boundsubstrate-like molecules from the full-length enzyme, theprotein was treated with five molar equivalents ofNADPH for 5 min at room temperature, and any productsremoved by extensive ultrafiltration prior to storage at−20 °C. Purification led to protein of N90% homogeneity asjudged by SDS–PAGE. All of the mutants exhibited typicalP450 spectra on reduction in the presence of carbonmonoxide, with absorbance maxima at 448 nm andminimal (b5%) formation of P420.

Optical spectroscopy

UV/visible spectroscopy was carried out using a Cary300 Bio UV/visible spectrophotometer equipped with aPeltier temperature control unit and Cary WinUV soft-ware. All experiments were conducted in buffer contain-ing 50 mM potassium phosphate (pH 8.0), at 30 °C unlessotherwise stated. Optical titrations were conducted bymeasuring the absorbance changes of the haem Sorêtband on addition of increasing quantities of ligand to aprotein solution containing 0.1–0.5 μM P450 in the samplecuvette, and to buffer solution in the reference cuvette.The data were corrected for any dilution of the proteinduring the course of the experiment, and the change inabsorbance difference between 390 nm and 420 nm,Δ(A390-A420), was plotted against substrate concentration,and dissociation constants were estimated using non-linear regression using Microcal Origin software, theprotein concentration being measured by CO-differencespectroscopy.54

Assay of catalytic activity

NADPH consumption assays were carried out usingreaction mixtures containing 0.1–0.5 μMP450 and varyingconcentrations of substrates as described. Reactions wereinitiated by the addition of 200–300 μM NADPH, and thedecrease in absorbance at 340 nm monitored. Initial rateswere calculated from the first 20 s of the reaction.Apparent KM values were estimated by plotting initialrate versus concentration and fitting with equation (1):

V ¼ V0 þ DVmaxSðKM þ SÞ , ð1Þ

amodification of the standardMichaelis–Menten equationwhich allows for a non-zero reaction rate at zero substrateconcentration due to uncoupling reactions.V0 is the rate ofNADPH consumption in the absence of substrate, andΔVmax the maximal increase in rate at saturating substrateconcentrations, such that the final maximal observed rateis equal to (Vo+ΔVmax). For substrates that showedevidence of cooperativity, a similarly modified version ofthe Hill equation was used to estimate maximal rates, thesubstrate concentration giving 50% of the maximal rate,S0.5, and the Hill coefficient, nH. In view of the sub-stoichiometric incorporation of haem in some of themutants, all catalytic rates were normalised to the haemcontent measured by CO-difference spectroscopy.54

Analysis of reaction products

The products of fatty acid hydroxylation were analysedby NMR using Bruker AMX500 and AMX600 instruments

as described.55 Reactions were carried out in 2H2O buffer,using 200 μM NADPH and 200 μM laurate such that theconcentration of dissolved oxygen (ca 250 μM) should notbe limiting, and were initiated by the addition of NADPH.The absorbance at 340 nm was monitored to ensure thereactions reached completion; NADPH consumption ratesin 2H2O were found to be approximately 90% of those inwater. Reaction mixtures were transferred to NMR tubesand the 1H-NMR spectra measured directly.The analysis of the products of indole hydroxylationwas

carried out separately for the soluble and insolubleproducts. HPLC analysis of the water-soluble products ofthe reactions of indole with P450BM3 was carried out asdescribed45 using an Agilent 1100 series instrumentequipped with diode-array UV/visible and fluorescencedetection, and a Zorbax Reverse Phase SB C-18 4.6 mm×250 mm column (5 μm particle size). All products wereidentified by demonstration of identical retention volumesto authentic standards, and quantitated by comparison ofUV/visible peak area integrations to known concentra-tions of authentic standards. The insoluble products wereanalysed by thin layer chromatography as described46 andproducts identified by comparison to authentic standardsof indigo and indirubin. In addition, the insoluble productsfrom a 1 ml reaction containing 1 μM P450BM3, 8 mMindole and 300 μM NADPH in 50 mM potassium phos-phate (pH 8) were redissolved in 600 μl of 2H6-DMSO andthe 1H-NMR spectra were recorded.

Crystallography

The A82F haem domain was crystallised by the sittingdrop method at 4 °C. Sitting drops were prepared byadding 2.5 μl of precipitant mixture to 2.5 μl of 6 mg/mlhaem domain; crystals were obtained by using a pre-cipitant mixture of 140 mM MgCl2, 25% (w/v) polyethy-lene glycol 2000MME and 100 mM Mes (pH 5.0). Crystalswere immersed in 10% polyethylene glycol 200 in mother


liquor as a cryo-protectant before being mounted on arayon loop and flash-cooled in liquid nitrogen. Diffractiondatawere collected at the European Synchrotron Radiationfacility (Grenoble, France) on ID14-EH3 using an ADSCQ4R CCD detector. The 180° data were collected at 100 Kwith 1° oscillations. Data were processed and scaled usingMOSFLM56 and SCALA.57 The auto-indexing routines inMOSFLM and examination of reflections along theprincipal axes established the space group as P212121with cell dimensions a=117.7 Å b=147.9 Å c=184.0 Å. Mo-lecular replacement calculations were performed withPHASER1.3,58,59 using a search model based on thesubstrate-bound wild-type structure (PDB code 1JPZ)edited to remove solvent and substrate molecules. Sixmolecules were found in the asymmetric unit, correspond-ing to a solvent content of 49%. Initial maps clearly showedpositive difference density corresponding to the phenolicside-chain of residue 82 and to bound substrate.REFMAC557,60 was used to conduct crystallographicrefinement and COOT61 was used for manual rebuildingand density interpretation. The data collection and finalrefinement parameters are given in Table 4.

Docking calculations

The protein-ligand docking was performed usingGOLD3.1.162 without any constraints. The templateprotein PDB files used were 1JPZ for the wild-typeenzyme and 2UWH for the A82F mutant. The ligandswere then docked into a sphere of radius 15 Å around theiron centre of haem.

Accession number

The refined coordinates and structure factors have beendeposited with the RCSB Protein Data Bank (PDB code2UWH).

Acknowledgements

This work was supported by the BBSRC (grantE20186) and by an ORSAS grant (to W.-C. H.). Weare grateful to Professors A.W. Munro and Paul M.Cullis for valuable discussions at the start of thisproject, to Professor Luet Wong for the gift of thepGLWBM3 plasmid, to Dr Fred Muskett for helpwith the NMR experiments and to Professor PeterFarmer and Dr Don Jones for the liquid chromato-graphy–mass spectroscopy analyses.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2007.08.015

References

1. Makris, T. M., Denisov, I., Schlichting, I. & Sligar, S. G.(2005). Activation of molecular oxygen by cytochrome

P450. In Cytochrome P450: Structure, Mechanism andBiochemistry (Ortiz de Montellano, P. R., ed), 3rd edit.,pp. 149–182, Kluwer Academic/Plenum Publishers,New York.

2. Ortiz de Montellano, P. R. & De Voss, J. J. (2005).Substrate oxidation by cytochrome P450 enzymes. InCytochrome P450: Structure,Mechanism and Biochemistry(Ortiz de Montellano, P. R., ed), 3rd edit., pp. 183–246,Kluwer Academic/Plenum Publishers, New York.

3. Guengerich, F. P. (2005). Human cytochrome P450enzymes. InCytochrome P450: Structure, Mechanism andBiochemistry (Ortiz de Montellano, P. R., ed), 3rd edit.,pp. 377–530, Kluwer Academic/Plenum, New York.

4. Kelly, S. L., Kelly, D. E., Jackson, C. L., Warrilow, A. G.S. & Lamb, D. C. (2005). The diversity and importanceof microbial cytochromes P450. In Cytochrome P450:Structure, Mechanism and Biochemistry (Ortiz de Mon-tellano, P. R., ed), 3rd edit., pp. 585–618, KluwerAcademic/Plenum Publishing, New York.

5. Nielsen, K. A. & Moller, B. L. (2005). CytochromeP450s in plants. In Cytochrome P450: Structure, Mechan-ism and Biochemistry (Ortiz de Montellano, P. R., ed),3rd edit., pp. 553–584, Kluwer Academic/PlenumPublishers, New York.

6. Warren, J. T., Petryk, A., Marques, G., Jarcho, M.,Parvy, J. P., Dauphin-Villemant, C. et al. (2002).Molecular and biochemical characterization of twoP450 enzymes in the ecdysteroidogenic pathway ofDrosophila melanogaster. Proc. Natl. Acad. Sci. USA, 99,11043–11048.

7. Munro, A. W., Leys, D. G., McLean, K. J., Marshall,K. R., Ost, T. W. B., Daff, S. et al. (2002). P450 BM3: thevery model of a modern flavocytochrome. TrendsBiochem. Sci. 27, 250–257.

8. Poulos, T. L. (2005). Intermediates in P450 catalysis.Phil. Trans. Roy. Soc. ser. A, 363, 793–806.

9. Pylypenko, O. & Schlichting, I. (2004). Structuralaspects of ligand binding to and electron transfer inbacterial and fungal P450s. Annu. Rev. Biochem. 73,991–1018.

10. Narhi, L. O. & Fulco, A. J. (1986). Characterization of acatalytically self-sufficient 119,000-dalton cytochromeP-450 monooxygenase induced by barbiturates inBacillus megaterium. J. Biol. Chem. 261, 7160–7169.

11. Fulco, A. J. (1991). P450BM-3 and other induciblebacterial P450 cytochromes: Biochemistry and regula-tion. Annu. Rev. Pharmacol. Toxicol. 31, 177–203.

12. Narhi, L. O. & Fulco, A. J. (1987). Identification andcharacterization of two functional domains in cyto-chromeP-450BM-3, a catalytically self-sufficientmono-oxygenase induced by barbiturates in Bacillus megate-rium. J. Biol. Chem. 262, 6683–6690.

13. Ravichandran, K. G., Boddupalli, S. S., Hasemann,C. A., Peterson, J. A. & Deisenhofer, J. (1993). Crystalstructure of hemoprotein domain of P450BM-3, aprototype for microsomal P450's. Science, 261,731–736.

14. Li, H. Y. & Poulos, T. L. (1997). The structure of thecytochrome P450BM-3 haem domain complexed withthe fatty acid substrate, palmitoleic acid.Nature Struct.Biol. 4, 140–146.

15. Haines,D.C., Tomchick,D.R.,Machius,M.&Peterson,J.A. (2001). Pivotal role of water in the mechanism ofP450BM-3. Biochemistry, 40, 13456–13465.

16. Joyce, M. G., Girvan, H. M., Munro, A. W. & Leys, D.(2004). A single mutation in cytochrome P450 BM3induces the conformational rearrangement seen uponsubstrate binding in the wild-type enzyme. J. Biol.Chem. 279, 23287–23293.


17. Ekroos, M. & Sjogren, T. (2006). Structural basis forligand promiscuity in cytochrome P450 3A4. Proc.Natl. Acad. Sci. USA, 103, 13682–13687.

18. Williams, P. A., Cosme, J., Vinkovic, D. M., Ward, A.,Angove, H. C., Day, P. J. et al. (2004). Crystal structuresof human cytochrome P450 3A4 bound to metyraponeand progesterone. Science, 305, 683–686.

19. Williams, P. A., Cosme, J., Ward, A., Angova, H. C.,Vinkovic, D. M. & Jhoti, H. (2003). Crystal structure ofhuman cytochrome P4502C9 with bound warfarin.Nature, 424, 464–468.

20. Li, H. Y., Narasimhulu, S., Havran, L.M.,Winkler, J. D.& Poulos, T. L. (1995). Crystal structure of cytochromeP450cam complexed with its catalytic product, 5-exo-hydroxycamphor. J. Amer. Chem. Soc. 117, 6297–6299.

21. Raag, R. & Poulos, T. L. (1989). Crystal structure of thecarbon monoxide-substrate-cytochrome P-450CAMternary complex. Biochemistry, 28, 7586–7592.

22. Cupp-Vickery, J. R. & Poulos, T. L. (1995). Structure ofcytochrome P450eryf involved in erythromycin bio-synthesis. Nature Struct. Biol. 2, 144–153.

23. Nagano, S., Li, H. Y., Shimizu, H., Nishida, C., Ogura,H., de Montellano, P. R. O. & Poulos, T. L. (2003).Crystal structures of epothilone D-bound, epothiloneB-bound, and substrate-free forms of cytochromeP450epoK. J. Biol. Chem. 278, 44886–44893.

24. Modi, S., Primrose, W. U., Boyle, J. M. B., Gibson, C. F.,Lian, L. Y. & Roberts, G. C. K. (1995). NMR studies ofsubstrate binding to cytochrome P450 BM3: compar-isons to cytochrome P450 cam. Biochemistry, 34,8982–8988.

25. Jovanovic, T. & McDermott, A. E. (2005). Observationof ligand binding to cytochrome P450 BM-3 by meansof solid-state NMR spectroscopy. J. Amer. Chem. Soc.127, 13816–13821.

26. Jovanovic, T., Farid, R., Friesner, R. A. & McDermott,A. E. (2005). Thermal equilibrium of high- and low-spin forms of cytochrome P450 BM-3: Repositioning ofthe substrate? J. Amer. Chem. Soc. 127, 13548–13552.

27. Ravindranathan, K. P., Gallicchio, E., Friesner, R. A.,McDermott, A. E. & Levy, R. M. (2006). Conforma-tional equilibrium of cytochrome P450 BM-3 com-plexed with N-palmitoylglycine: a replica exchangemolecular dynamics study. J. Amer. Chem. Soc. 128,5786–5791.

28. Carmichael, A. B. & Wong, L. L. (2001). Proteinengineering of Bacillus megaterium CYP102 - the oxida-tion of polycyclic aromatic hydrocarbons. Eur. J.Biochem. 268, 3117–3125.

29. Appel, D., Lutz-Wahl, S., Fischer, P., Schwaneberg, U.& Schmid, R. D. (2001). A P450 BM-3 mutanthydroxylates alkanes, cycloalkanes, arenes and het-eroarenes. J. Biotechnol. 88, 167–171.

30. Cirino, P. C. & Arnold, F. H. (2002). Regioselectivityand activity of cytochrome P450 BM-3 and mutantF87A in reactions driven by hydrogen peroxide. Adv.Synth. Catal. 344, 932–937.

31. Cowart, L. A., Falck, J. R. & Capdevila, J. H. (2001).Structural determinants of active site binding affinityand metabolism by cytochrome P450 BM-3. Arch.Biochem. Biophys. 387, 117–124.

32. Graham-Lorence, S., Truan, G., Peterson, J. A., Falck,J. R., Wei, S. Z., Helvig, C. & Capdevila, J. H. (1997).An active site substitution, F87V, converts cytochromeP450 BM-3 into a regio- and stereoselective (14S,15R)-arachidonic acid epoxygenase. J. Biol. Chem. 272,1127–1135.

33. Li, Q. S., Ogawa, J., Schmid, R. D. & Shimizu, S.(2001). Residue size at position 87 of cytochrome

P450BM-3 determines its stereoselectivity in propyl-benzene and 3-chlorostyrene oxidation. FEBS Letters,508, 249–252.

34. Sulistyaningdyah, W. T., Ogawa, J., Li, Q. S., Maeda,C., Yano, Y., Schmid, R. D. & Shimizu, S. (2005).Hydroxylation activity of P450BM-3 mutant F87Vtowards aromatic compounds and its application tothe synthesis of hydroquinone derivatives fromphenolic compounds. Appl. Microbiol. Biotechnol. 67,556–562.

35. Munro, A. W., Daff, S., Coggins, J. R., Lindsay, J. G. &Chapman, S. K. (1996). Probing electron transfer inflavocytochrome P-450 BM3 and its componentdomains. Eur. J. Biochem. 239, 403–409.

36. Neeli, R., Girvan, H. M., Lawrence, A., Warren, M. J.,Leys, D., Scrutton, N. S. & Munro, A. W. (2005). Thedimeric form of flavocytochrome P450 BM3 is cataly-tically functional as a fatty acid hydroxylase. FEBSLetters, 579, 5582–5588.

37. Magnuson, K., Jackowski, S., Rock, C. O. & Cronan,J. E., Jr (1993). Regulation of fatty acid biosynthesis inEscherichia coli. Microbiol. Rev. 57, 522–542.

38. Yeom, H., Sligar, S. G., Li, H. Y., Poulos, T. L. & Fulco,A. J. (1995). The role of Thr268 in oxygen activation ofcytochrome P450BM-3. Biochemistry, 34, 14733–14740.

39. Raag, R. & Poulos, T. L. (1991). Crystal structures ofcytochrome P-450CAM complexed with camphane,thiocamphor, and adamantane: factors controllingP-450 substrate hydroxylation. Biochemistry, 30,2674–2684.

40. Modi, S., Sutcliffe, M. J., Primrose, W. U., Lian, L. Y. &Roberts, G. C. K. (1996). The catalytic mechanism ofcytochrome P450 BM3 involves a 6 Åmovement of thebound substrate on reduction. Nature Struct. Biol. 3,414–417.

41. Helms, V., Deprez, E., Gill, E., Barret, C., Hoa, G. H. B.& Wade, R. C. (1996). Improved binding of cyto-chrome P450cam substrate analogues designed to fillextra space in the substrate binding pocket. Biochem-istry, 35, 1485–1499.

42. Rock, D. A., Boitano, A. E., Wahlstrom, J. L., Rock,D. A. & Jones, J. P. (2002). Use of kinetic isotopeeffects to delineate the role of phenylalanine 87 inP450(BM-3). Bioorg. Chem. 30, 107–118.

43. Noble, M. A., Miles, C. S., Chapman, S. K., Lysek,D. A., Mackay, A. C., Reid, G. A. et al. (1999). Rolesof key active-site residues in flavocytochrome P450BM3. Biochem. J. 339, 371–379.

44. Gillam, E. M. J., Aguinaldo, A. M. A., Notley, L. M.,Kim, D., Mundkowski, R. G., Volkov, A. A. et al.(1999). Formation of indigo by recombinant mamma-lian cytochrome P450. Biochem. Biophys. Res. Commun.265, 469–472.

45. Gillam, E. M. J., Notley, L. M., Cai, H. L., De Voss,J. J. & Guengerich, F. P. (2000). Oxidation of indoleby cytochrome P450 enzymes. Biochemistry, 39,13817–13824.

46. Li, Q. S., Schwaneberg, U., Fischer, P. & Schmid, R. D.(2000). Directed evolution of the fatty-acid hydro-xylase P450 BM-3 into an indole-hydroxylatingcatalyst. Chem. Eur. J. 6, 1531–1536.

47. Nakamura, K., Martin, M. V. & Guengerich, F. P.(2001). Random mutagenesis of human cytochromeP450 2A6 and screening with indole oxidationproducts. Arch. Biochem. Biophys. 395, 25–31.

48. Li, Q. S., Ogawa, J., Schmid, R. D. & Shimizu, S. (2005).Indole hydroxylation by bacterial cytochrome P450BM-3 and modulation of activity by cumene hydro-peroxide. Biosci. Biotechnol. Biochem. 69, 293–300.


49. van Vugt-Lussenburg, B. M. A., Damsten, M. C.,Maasdijk, D. M., Vermeulen, N. P. E. & Comman-deur, J. N. M. (2006). Heterotropic and homotropiccooperativity by a drug-metabolising mutant ofcytochrome P450 BM3. Biochem. Biophys. Res. Commun.346, 810–818.

50. Rock, D. A., Perkins, B. N. S., Wahlstrom, J. & Jones,J. P. (2003). A method for determining two substratesbinding in the same active site of cytochromeP450BM3: an explanation of high energy omegaproduct formation. Arch. Biochem. Biophys. 416, 9–16.

51. Bernhardt, R. (2006). Cytochromes P450 as versatilebiocatalysts. J. Biotechnol. 124, 128–145.

52. Urlacher, V. B., Makhsumkhanov, A. & Schmid, R. D.(2006). Biotransformation of beta-ionone by engi-neered cytochrome P450 BM-3. Appl. Microbiol. Bio-technol. 70, 53–59.

53. Li, Q. S., Ogawa, J., Schmid, R. D. & Shimizu, S. (2001).Engineering cytochrome P450BM-3 for oxidation ofpolycyclic aromatic hydrocarbons. Appl. Environ.Microbiol. 67, 5735–5739.

54. Omura, T. & Sato, R. (1964). The carbon monoxide-binding pigment of liver microsomes II. Solubiliza-tion, purification, and properties. J. Biol. Chem. 239,2379–2385.

55. Oliver, C. F., Modi, S., Sutcliffe, M. J., Primrose, W. U.,Lian, L. Y.&Roberts, G. C.K. (1997). A single mutationin cytochrome P450 BM3 changes substrate orienta-

tion in a catalytic intermediate and the regiospecifi-city of hydroxylation. Biochemistry, 36, 1567–1572.

56. Leslie, A. G. W. (1992). Recent changes to theMOSFLM package for processing film and imageplate data.

57. Collaborative Computational Project (1994). TheCCP4 suite: programs for protein crystallography.Acta Crystallog. sect. D, 50, 760–763.

58. Storoni, L. C., McCoy, A. J. & Read, R. J. (2004).Likelihood-enhanced fast rotation functions. ActaCrystallog. sect. D, 60, 432–438.

59. McCoy, A. J., Grosse-Kunstleve, R. W., Storoni, L. C. &Read, R. J. (2005). Likelihood-enhanced fast transla-tion functions. Acta Crystallog. sect. D, 61, 458–464.

60. Potterton, E., Briggs, P., Turkenburg, M. & Dodson, E.(2003). A graphical user interface to the CCP4program suite. Acta Cryst allog. sect. D, 59, 1131–1137.

61. Emsley, P. & Cowtan, K. (2004). COOT: model-building tools for molecular graphics. Acta Crystallog.sect. D, 60, 2126–2132.

62. Verdonk, M. L., Cole, J. C., Hartshorn, M. J., Murray,C. W. & Taylor, R. D. (2003). Improved protein-liganddocking using GOLD. Proteins: Struct. Funct. Genet. 52,609–623.

63. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein structures.J. Appl. Crystallog. 26, 283–291.

Filling a hole in cytochrome P450 BM3 improves substrate binding and catalytic efficiency

Documents