Structures of the Apo and FAD-Bound Forms of 2-Hydroxybiphenyl 3-monooxygenase (HbpA) Locate Activity Hotspots Identified by Using Directed Evolution

The FAD-dependent monooxygenase HbpA from Pseudomonas azelaica HBP1 catalyses the hydroxylation of 2-hydroxybiphenyl (2HBP) to 2,3-dihydroxybiphenyl (23DHBP). HbpA has been used extensively as a model for studying flavoprotein hydroxylases under process conditions, and has also been subjected to directed-evolution experiments that altered its catalytic properties. The structure of HbpA has been determined in its apo and FAD-complex forms to resolutions of 2.76 and 2.03 Å, respectively. Comparisons of the HbpA structure with those of homologues, in conjunction with a model of the reaction product in the active site, reveal His48 as the most likely acid/base residue to be involved in the hydroxylation mechanism. Mutation of His48 to Ala resulted in an inactive enzyme. The structures of HbpA also provide evidence that mutants achieved by directed evolution that altered activity are comparatively remote from the substrate-binding site.


Introduction
Microbial flavoprotein monooxygenases (FPMOs) are involved in ah ost of important biochemical processes in ar ange of organisms, with many roles in catabolism and natural-product biosynthesis. [1,2] Owing to their ability to selectively functionalise organic molecules of interest, they also have great potential as biocatalysts for preparative industrial reactions. [3,4] As urvey by van Berkela nd Fraaije in 2006 divided the FPMOs into six sub-classes A-F. [3] Subclass "A", which consists of the aromatic flavin-dependent hydroxylases, is an attractive group of enzymes from the perspective of preparative biocatalysis, as these enzymes catalyse the selective hydroxylation of aromatic substrates to phenols, the abioticp reparation of which requires harsh conditions and toxic reagents that are inconsistent with contemporary demands for sustainable chemical synthesis. The model enzyme from subclass "A" has been the NADPH-plus-FAD-dependent para-hydroxybenzoate hydroxylase (PHBH), [5] which catalysest he hydroxylation of the substrate to 3,4-dihydroxybenzoate (1,S cheme 1), and for which an X-ray structure [6] and mechanism [7] were first proposed in 1979.
PHBH hass erved as am odel for structural and mechanistic studies in the aromatic hydroxylasese ver since. [8] PHBH and other class "A" FPMOs are thought to catalysea romatic hy-droxylation reactions through am echanism of two distinct phases.I nt he first, the nicotinamide cofactor NAD(P)H is used to reduce FAD. This reduction is stimulated by the binding of the aromatic substrate, which acts as an effector.I nt he second half of the reaction, reduced FADH 2 reacts with molecular oxygen to produce aC 4a-(hydro)peroxyflavin speciest hat acts as the oxygenating agent in catalysis. Hydroxylation of the aromatic nucleus is thought to occur by electrophilic aromatic substitution, in which the hydroperoxide is the electrophile and the hydroxybenzoate aromatic ring is the nucleophile. [8] The nucleophilicity of the aromatic ring is increased by deprotonation of the phenolic hydroxy group by active-site residues. Extensive structurals tudies on PHBH [9][10][11] have also revealed that catalysis is characterised by am obile flavin that exists in The FAD-dependent monooxygenase HbpA from Pseudomonas azelaica HBP1 catalyses the hydroxylation of 2-hydroxybiphenyl (2HBP) to 2,3-dihydroxybiphenyl (23DHBP). HbpA has been used extensively as am odel for studying flavoprotein hydroxylases under process conditions, and has also been subjected to directed-evolution experiments that altered its catalytic properties. The structure of HbpA has been determined in its apo and FAD-complex formst or esolutions of 2.76 and 2.03 , re-spectively.C omparisons of the HbpA structure with those of homologues, in conjunction with amodel of the reaction product in the active site, reveal His48 as the most likely acid/base residue to be involved in the hydroxylation mechanism.M utation of His48 to Ala resulted in an inactive enzyme. The structures of HbpA also provide evidence that mutants achieved by directede volution that altereda ctivity are comparatively remote from the substrate-binding site. different conformations depending on the step in the catalytic cycle. For flavin reduction, the FADi sd isplaced away from the substrate binding site to the periphery of the enzyme ("out" position), where it can be reduced through interaction with NAD(P)H. In the presence of substrate, the flavin is located in the substrate binding site ("in" position). Reduced FADH 2 in the "in" positioni sa ble to participate in both the reaction with oxygen, to form the hydroperoxide, and then the aromatic hydroxylation.
In addition to PHBH, other structurally and mechanistically related class "A" FPMOs that have been characterised range from phenol hydroxylase (PHHY), which forms catechol (2; [12,13] Scheme 1), to 3-hydroxybenzoate 6-hydroxylase [14] and biosynthetic enzymes including RebC, which is involved in the productiono fr ebeccamycin, [15,16] and aklavinone-11-hydroxylase (RdmE), which hydroxylates the anthracycline precursor aklavinone (3). [17] In 1988, Kohler and co-workersr eported the isolation of the bacterium Pseudomonas azelaica HBP1, which was able to employ as sole carbon source the fungicide 2-hydroxybiphenyl (2HBP). [18] The breakdown of the biaryl core was initiated by the ortho-hydroxylation of the substrate to yield 2,3-dihydroxybiphenyl (23DHBP). This aromatic hydroxylation was later attributed to ac lass "A" FPMO, 2-hydroxybiphenyl 3-monooxygenase (HbpA), which was isolated and characterised. [19] Mechanistic studies confirmed that HbpA shared many of the catalytic characteristics of class "A" FPMOs such as PHBH. [20] The gene encoding HbpA was cloned and overexpressed in the heterologous host Escherichia coli [21] to facilitatei ts application to the gram-scaleh ydroxylation of ar ange of 2-substituted phenols [22] and it has also served as am odel for the activity of flavin-dependent hydroxylases under process conditions, including studies that investigated the stability of the purified enzyme in the presence of co-solvents such as methanol and decanol. [23] HbpA was also the subject of successful directedevolution experiments that yieldedm utantst hat weres uperior to the wild-type in their biotransformation of tert-butyl phenol [24] guaiacol and 2-sec-butylphenol, [25] displayed improvedcoupling of NADH oxidation to substrate hydroxylation, and also am utant capable of the hydroxylation of indole. [26] In an effort to shed light on the structural consequences of evolution experiments,a nd also to provide furtheri nformation on substrate recognition and mechanism in HbpA, preliminaryXray crystallographic studies on the enzyme were performed. [27] Although crystals were obtained, as olution of the structure was frustrated by the short lifetime of the crystals within the X-ray beam;2 0c rystals were required to provide as ingle dataset. In this report, using an ew genetic construct and exploiting advances in X-ray data collection time, we have used single crystals of HbpA to determine its structure in two forms: an FAD-bound form that allows characterisation of the active site, and an apo form that, although lacking flavin, gives extra information on the location of residues and structure of mobile loops that are absentfrom the FADc omplex. The structure of HbpA allows the beneficial mutation sites previously identified to be put into at ertiary structural context for the first time, and also provides ar obustp latform for furtherp rotein engineering of this useful enzyme.

Results and Discussion
Quality of the models Sub-cloning of the hbpA gene and its expression from the YSBLIC-3C vector resulted in ap rotein that is 21 residues longer than the native protein by virtue of as pacert hat leads from the N-terminal methionine to the histidine tag, and also incorporates aC 3p roteasec leavage site. The tag was not removed prior to crystallisation. Protein resulting from this construct gave crystals of both ad ifferent space group (P1r ather than C2) and cell dimensions from those obtained by Meyer and co-. [27] These crystalsh ad sufficient longevity within the Xray beam for ac omplete dataset to be collected by using one crystal in each case for the apo form of the enzymea nd the FADc omplex. The structure of the apo form was obtained first, after which it was determined that soaking with 1mm FAD was necessary to obtain structural data in which FADw as at full occupancy within each active site of HbpA. In the apo form, the amino acid backbone was complete from residues Ser4 to Arg228, Ala239 to Trp254 and Glu268 to Arg565. The FADc omplex was less complete in each subunit, with density of aq uality sufficient for building present from Ser4 to Gly194, Ser203 to Asp211, Asp222 to Arg228, then Gly237t oM et243; Trp250 to Trp254 then Glu268 to Arg585. In each case, the breaks in electron density corresponded to loop regionsb etween the seven b-strands of the substrate binding domain D2,asd escribed in detail below.
The closests tructuralh omologueso fH bpA, as revealed by analysisu sing the DALI server [31] are the flavin-dependent monooxygenase RebC, from Lechevalia aerocologines (PDB ID: 4EIP;2 8% sequence identity; Z-score 39.8;r msd 2.6 o ver 489 backboneC aatoms), [15,16] the aklavinone-11-hydroxylase (RdmE)f rom Streptomycesp urpurascens [17] (PDB ID:3 IHG;3 5% sequence identity; Z-score 39.2;r msd2 .2 o ver 474 backbone Ca atoms) and the phenol monooxygenase (PHHY) from Trichosporon cutaneum (Z-score 35.2;1 PN0/1FOH;2 2% sequence identity; rmsd 2.9 over 476 backboneC aatoms). [12,13] The FADb inding domain, D1,i saR ossman-type domain with ac entral mixed b-sheet of six strands, bordered by an additional sheet of four strands, on the side nearestt he enzyme periphery,a nd four substantial a-helices (a1, a6, a10 and a11) at the interface with the thioredoxin domain, D3.T he loop of amino acids 35-52 betweens trand b2a nd helix a2r uns behind, and provides close contacts with, the isoalloxazine ring of FAD. HbpA lacks the large additional loop region in D1 that is present in PHHY (formed by residues 170-210), which is also absenti nR dmE andw hich is thought to constitute a" lid" over the active site of PHHY. [12] The middle domain, D2,c onsists of al arge seven-stranded mixed b-sheet that provides the floor of the substrate binding site. Poor density in the loops connecting these strands accounts for the most of the chain breaks in both the apo and especially the FAD-complex structures listed above and is indicative of ah igher level of mobility in this region.T he sheet is adjacent to ah elix-turn-helix-turnhelix subdomainf ormed by residues 325-427 of domain D2, which incorporates helices a12, a13 and a14 ( Figure 1) and which constitutes one of the major structural differences between HbpA, RebC, RdmE and PHHY.I nP HHY, this subdomain is smaller,l acking the equivalent of a13, and in RdmE is almost absent,r eplaced by al oop of only 15 amino acids (RdmE 373-388). This large sub-domain in HbpA is almostc ertainly important for dimer stabilisation within the HbpA tetramer (vide infra). Domain D2 is connected by al ong loop from amino acids 428-471 to the thioredoxin domain D3,w hich,a lthough of unknown function in FPMOs, is involved in many reciprocal contactsw ith the helix-turn-helix-turn-helix region 325-427i n domain D2 of the neighbouring monomer.

Structure of the HbpA tetramer
Data on the tetrameric assembly observed in the crystal structure ( Figure 2), which were provided by PISA analysis, [32] appear to provide good evidence foraphysiological tetramer, as suggested previously: [19] the contact area between monomers A and C is 1243 2 ,w ith a DG value for this interaction of À8.5 kcal mol À1 ;t he values for the dimer formedb ym onomers A and D were 1037 2 and À12.7 kcal mol À1 .T here are ten hydrogen bonds and 13 salt bridges between the AC pair,a nd between the AD pair are 11 and eight, respectively.T he AC dimer is held together by reciprocal interactions between domain D3 andh elix a14 of domain D2,a nd also by hydrogen bonds between the Gln102 side chains in the short helices a5 that connectd omains 1a nd 2. The AD pair is held together by extensive reciprocal interactions between the N-terminal fourstranded b-sheets and helices a3a nd a6o fd omain D1.T here is al arge cavity at the centre of the tetramer into which the side chains of several hydrophilic amino acids of each monomer,including Glu75 and Tyr76, are projected.
The aromatic hydroxylases PHHY [12] andR dmE [17] are reported to exist in solutiona sah omodimer and am onomer,r espectively.The determinants of the AC homodimer interactions in PHHY (1FOH) are very different from those of the AC dimer found in HbpA;i nt he PHHY dimer,m uch more extensive reciprocal interactions exist between domains 2a nd 3, giving ac ontact area of 1900 2 ,a sar esult of ar elative closureo f the AC subunits relative to HbpA. Dimer formation appearst o be assisted in PHHY by al arge movement of the loop 170-210 (absent in HbpA) in domain 1i no nly one of the monomers that brings this loop into close contact with its dimer partner. Interactions between the more stable dimer pair AD of HbpA (as indicated by DG values) are actually governed by reciprocal interactions between the D1 domains, which make comparatively few interactions in the PHHY AC dimer,a sar esult of the extended loop. The structure of D1 appearst ob eq uite well conserved between HbpA and RdmE, so the reason for the failure of the latter to oligomerise in an equivalent way to the HbpA AD dimer is not clear,b ut the inability to form an HbpA AC-type dimer might be attributed to its lacko fasubstantial subdomainwithin D2.

Comparison of the apo and FADs tructures
Although lacking the FADc oenzyme, the apo structure of HbpA has proveduseful in providing information on secondary structurale lements and residues that are missing from the FAD-complex structure. The overall rmsd betweent he two structures is low,a t0 .85 o ver 538 backbone Ca atoms, as would be expected, but there is ap ronounced shift in the relative orientation of domain D2 away from domain D1 as ar esult of FADb inding;t his results in am ore open active site. In the apo structure, the backboneC aatoms of Arg242 and Met243 on strand b15 in D2 closest to FAD, for example, each move 2.5 a way from the binding site in as uperimposition of the two structures. The relative shift in domains is coincident with an absence of some electron density in regions in the FADc omplex compared to the apo structure. These include density for the loop 195-203 that connects domains D1 and D2 as wella sa dditional loops between the seven b-strands of domain D2,i ncluding region 245-249 between b15 and b16 and region 213-219b etween b13 and b14, which features the short helix a8i nt he apo structure. The reasons for the mobility of these loops on FADb inding are unclear;t hey correspond neither to substrate entry channels-as proposed, for example, for RdmE [17] -nor to the region of the protein that would move as ar esult of FADm oving to the "out" position, as also observed in RdmE.T he structural reorganisation upon FAD binding is also manifested in the orientation of the loop between residues 36 and4 9b etween b2a nd b3, which movest o accommodate the isoalloxazine ring of FAD.

FADa nd NADH binding
In the FADc omplex, the flavin is bound within HbpA in the "in" position described for other class "A" FPMOs. As ac onsequenceo ft he relative movement of domains D1 and D2,r esidue Trp293 on strand b17 has movedt os hield the isoalloxazine ring of FADf rom the solvent. The ADPm oiety of FADi s bound in ap ocket near the surface of D1.T he adenine ring is stacked against the side chain of Arg37, with the exocyclic NH 2 bondedt ot he backbone carbonyl of Tyr144. The adenine ring also makes hydrogen bonds with the peptidic NH groups of Arg37 and Tyr144. The ribose hydroxy group is H-bonded to the side chain of Asn36,a nd the diphosphate bonds to the peptidic NH groups of Ala17 and Asp313, and also to water molecules. The ribitol side chain interacts with the side chains of Asp313,A rg46 and Gln120. The tricyclic isoalloxazine ring of FADi sb ound deeper within the protein at the base of domain D1 and near the interface with D2,w ith the isoalloxazine ring stacked between loop 46-49 preceding strand b3, and loop 319-324b etween b21 and a10. The loop above the flavin provides the side chains of Ser47, which is directly above the central FADr ing, and His48, which is discussed in more detail below.T he loop beneath the flavin provides Pro320a nd Met321 (Figure3A), whicha lso help to form one wall of the substrate binding pocket, al arge hydrophobic cavity whose other side formed by the seven-stranded b-sheet of domain D2.
The sheet of domain D2 provides many hydrophobic side chains to the active site pocket, including Trp97, Met223 and Trp225.A mongt he hydrophobic side chains of the active site, there are also two hydrophilic residues, Arg242 and His48, which are situated to the front of the plane of the isoalloxazine ring ( Figure 3A). The imidazole ring of His48 is 5.5 f rom the C4a atom of FADa tw hich the hydroperoxide is formed. It is also 4.0 f rom aw ater molecule to the front and beneath the FADr ing that is well placed to mimict he hydroxy group of the hydroperoxide. Putative roles for His48 in the mechanism of HbpA are discussed in the context of am odel complex structure below.
The structure of HbpA also allows ac omparative analysis of the cofactor binding loop, which contains residues that assist in discriminating between the nicotinamide cofactors NADH and NADPH. In NADPH-dependentP HBH, specificity for the phosphorylated cofactor is thought to be determined largely by Arg33,T yr38 and Arg42, with replacement of Tyr38 by an acidic residue leading to as hift in activity toward NADH. [33] The equivalent residues in HbpA:A rg37, Ser40 and Ser42 clearly provide al ess positivelyc harged environment for the NADPH phosphate, but there are no close structurally homologous acidic residues that suggest that NADHs pecificity in HbpA is achieved through carboxylate-ribose hydroxy interactions.
Comparison of the HbpA active site with FPMO homologues PHHY and aklavinone-11-hydroxylase RdmE Determination of the structure of HbpA allows its active site to be compared with that of its closest structural and functional homologues in an effortt od istinguish the determinantso f mechanism and substrate specificity in the enzyme. Despite numerousa ttempts at both co-crystallisation and soakingw ith both 2HBP and 23DHBP,n ot ernary complex of HbpA was forthcoming. We therefore sought to model the product into www.chembiochem.org the active site of the enzyme by using PHHY and RdmE as the bases. Interestingly,i ns ubstrate complexes for both PHHY and RdmE, the nucleophilic carbonsofthe substrate to be hydroxylated superimpose very well( Figure 3B), at ad istance of approximately 5.3 a nd an angle of approximately 608 down from the plane of the FADi soalloxazine ring. This, in addition to the orientation of the polycyclic ring system in RdmE,h as provided anchoring positions for modelling the HbpA product 23DHBP into the substrate-binding pocket of HbpA.
It might be expectedt hat some conservation of active-site residues betweenP HHY and HbpA would be observed, given the formal similarity of their substrates;o ft he hydrophobic side chains in the active site, for example, Met80 (PHHY) is conserved as Met77 (HbpA). However,f or protic residues that interact directly with the phenol hydroxy group in PHHY,a nd have thusb een implicated in mechanism, [12] Asp54, which is thought to assist in phenolate stabilisation,i sr eplaced in HbpA by His48, and Tyr289, which is thought to act as the proton donor to the FADp eroxo-anion is replaced by Ala240. Given these differences, it is clear that, despite similarities in substrate structure, different mechanismso fs ubstrate orientation and intermediate stabilisation operate in HbpA.
Ac omparison of the active site of HbpA with that of RdmE, [17] is also informative in that, in common with HbpA, this enzymec atalyses the hydroxylation of ap olycyclic aromatic substrate. The recognition of al arger hydrophobic substrate is reflected in the hydrophobic make-up of the substrate binding channel in both enzymes. In the HbpA model product complex, the phenylr ing of the product is accommodated in ah ydrophobic pocket formedb yt he side chains of Pro320, Met321, Trp225 and Met223, similar to the binding mode of the polycyclic ring system of aklavinone in RdmE, although the phenylr ing is rotated9 0 8relative to the phenol ( Figure 3B). The phenol is in ah ydrophobic pocket formedb yI le49 and Trp97 near the pyrimidinedione ring of FAD.
As with PHHY, the aklavinone substrate is also thought to be activated by deprotonation in RdmE, which,i nt his case, is accomplished by Tyr224, which forms ah ydrogen bond with the C6 hydroxy group of aklavinone para to the C11a tom. If deprotonation of the C2 hydroxy group of the substrate is also to be as ignificant step in the HbpA mechanism, the only candidate residue for deprotonation is His48. However,i nt he product model complex of HbpA, His48 is hydrogen bonded to the new C3-hydroxy group of the product. In the HbpA-FADc omplex, the His48 NE2 atom is 4.0 f rom aw ater molecule that is itself 3.6 f rom the FADC 4a atom, and ideally placed to mimic the hydroxy group of ah ydroperoxide intermediate. It might be therefore that His48 is the protond onor to the nascent hydroperoxide, the role thought to be fulfilled in PHHY by Tyr289. Interestingly,H is48 forms ac leard yad with Asp117 in the second shell of residues around the actives ite. The only other hydrophilic residuet hat is close to the FADi s Arg242; the terminal guanidinium group is 8.2 f rom the FAD C4a, but also only 3.6 f rom the imidazole ring of the His48 side chain.

The His48Ala mutant of HbpAisi nactive for the hydroxylation of 2HBP
The location of His48,A sp 117a nd Arg242w ithin the active site prompted us to make site-directed mutantso fH bpA featuring alaninea tt hese positions in order to test the effect of these mutations on the ability of the enzyme to bind and to hydroxylate 2HBP.E ach of the mutant genes encoding His48Ala, Asp117Ala and Arg242Ala was expressed at similar levels in the soluble fraction as compared to the wild-type (WT). The apparent K m values (K mapp )f or 2HBP for each of the mutantsw ere then determined (Table 1) by UV spectrophotometry,m easuring the oxidation of NADH at 340 nm at increasing substrate concentration, as described by Meyer et al. [25] Each mutant displayed Michaelis-Menten-type kinetics with 2HBP,w itht he WT K mapp value of 3.3 mm being of as imilar order to that determined by Meyer et al. (2.6 mm). [25] The K mapp of His48Ala was approximately the same as for the WT,b ut Asp117Ala and Arg242Ala, displayed values approximately 1.7 times greater than WT in each case. Each mutant was clearly able to oxidise NADH. The maximum rate of NADHo xidation in these experiments,r ecorded in the presence of saturating concentrations of 2HBP (25 mm)w as 0.02 s À1 for the WT enzyme, but was reduced approximately fourfold for the His48Alaa nd Asp117Ala mutants, ande ightfold fort he Arg242Ala mutant.
However,t he rate of NADH oxidation as approximately the same for all variants in the absence of substrate;p erhaps this indicates that the ability of 2HBP to act as an effector is compromised in each of these mutants.
The rates of NADH oxidation are often not indicative of the catalytic performance of HbpA with respectt os ubstrate hydroxylation,h owever,o wing to the inefficiency of electrontransfer steps between cofactors ands ubstrate in the mechanism ("decoupling"). [25] The hydroxylation abilities of WT HbpA and mutant enzymes were therefore assayed by using the appearance of product on HPLC, as described in the Experimental Section. When WT HbpA was incubated with 0.5 mm 2HBP and NADH in threefold excess, the oxygenation to 23DHBP was complete after 2h.N oc onversiont op roduct was obtained for reactions containing either the His48Ala or Asp117Ala mutants. For the Arg242Ala mutants, approximately 15 %c onversion to the product was obtained after 2h.T he resultsa ppear to suggest that both His48 and Asp117 are essential for the oxygenation reaction to occur,b ut that Arg242 is not. The precise role of His48 as ap ossible proton donor awaits the outcome of other experiments,i ncluding the determination of an HbpA www.chembiochem.org structure in the presence of the substrate or product. The role of Asp117 might be either merely to orient the correct tautomer of the His48 imidazole ring for catalysis, as in ribonuclease A, [34] or to reduce the pK a of the histidine NE-2 proton, thereby assisting its functiona sacatalytic acid. Ther ole of Arg242i su nclear,b ut the higher K mapp observed for this mutant,c oupled with the mobility of the domain bearing this residue (as revealed by comparison with the apo structure), suggestst hat interaction between this residue and the substrate cannot be ruled out. Althought he precise role of His48 in catalysis remains unconfirmed, its identification as the major protic residue neart he C4a atom now permits more thorough investigation of its mechanistic role.

Location of activity hotspots in HbpA
In the absence of structure, HbpA was subjected to directedevolution experiments that resulted in mutants capable of hydroxylating 2-tert-butylphenol, [24,25] guaiacol [25] and indole. [26] The structure of HbpA allows the sites of mutations in these variants to be located for the first time. The variant Val368Ala/ Leu417Phe was reported to have eightfold improved catalytic efficiency towards guaiacol. Another variant, Ile244Val displayed highera ctivity towards 2-sec-butylphenol, guaiacol and 2-HBP.Am odel of HbpA constructed by using the structure of PHHY as at emplate, placed Val368 and Leu417 near the surface of the HbpA monomer,a nd within the substrate binding pocket. [23] Val368 in each subunit is indeed at the periphery of the HbpA tetramer in helix a11 within the flavin binding domain D1 (Figure 4A), and superimposes with Ile412 of PHHY, but is approximately 13 f rom the FAD, with the loop formed by residues 317-324b etweent hem.L eu417 is one of the first residues in the long helix a14 that helps to form the larger substrate binding domain D2 and is actually 27 d istantf rom the FADm olecule ( Figure 4B)L eu417 superimposes in the region of Met438 of PHHY,r ather than Val480, as first thought. [23] Here it makes stabilising hydrophobic interactions with Val381, Leu385 and Met421, which support the b-sheet underlying the substrate-binding pocket. Electron density for Ile244 is absenti nt he FAD-complex structure, but the adjacent residue,M et243, places Ile244 within the loop that separates strands b15 and b16 in the b-sheet underlying the substrate binding pocket (Figure 4C), and superimposing with residue Leu285 in PHHY,not with the catalytic residue Tyr289 of PHHY within the substrate binding pocket, as previously suggested. [23] As each of the three mutated residues is too distant from the FADb inding site to be implicated in catalysis directly,o ther roles in substrate transport or active-site access must be considered. Val368 is on helix a11,w hich corresponds to region 354-363 in RebC, which is proposed to act as the gateway for substrate andp roduct in that enzyme. In HbpA, helix a11 forms one side of ahydrophobic channel, the opening of which also includes the mobile loop 195-203, as revealed by the apo structure. The mobility in this region of the protein could mean that it serves as ag ateway in HbpA for entry of the aromatic substrate.

Conclusions
Flavoprotein aromatic hydroxylases are promising candidates for roles in both the preparative hydroxylation of aromatic substrates and in bioremediation processes. The determination of their structure is as ignificant advance in being able to interpret the results of in vitro evolution experiments and in informing new protein-engineering experiments. In addition to providing the first structural perspective on mechanism and specificity in HbpA, the structure also reveals new avenues for rational engineering experiments that could be targeted towards improving or altering its activity.

Experimental Section
HbpA gene cloning, expression and protein purification:T he gene encoding HbpA was provided by Bartlomiej To maszewski (Technical University of Dortmund, Germany). The gene was subcloned into the pET-YSBLIC-3C vector,i nw hich cloned genes become equipped with as equence encoding an N-terminal His 6 tag. The HbpA gene was amplified by PCR from the template plasmid by using the primers 5'-CCAGG GACCA GCAATG TCGA ATTCT GCAGA AACTG ATGTT CTTAT TGTGG-3' (forward) and 5'-GAGGA GAAGG CGCGT TACGC CCTCC CAAGG ATGCT CTTCA C-3' (reverse). Following agarose gel analysis of the PCR product, the relevant band was eluted from the gel by using aP CR Cleanup kit (Qiagen). The gene was then subcloned into the pET-YSBLIC-3C vector according to published techniques. [35] The recombinant plasmid was www.chembiochem.org used to transform E. coli XL1-Blue cells (Novagen), yielding colonies that, in turn, gave plasmids by standard miniprep procedures; these were sequenced to confirm the identity and sequence of the gene.
The recombinant plasmid containing the HbpA was used to transform E. coli BL21(DE3) cells by using kanamycin (30 mgmL À1 )a sa ntibiotic marker on lysogeny broth (LB) agar.S ingle colonies from an agar plate grown overnight were used to inoculate cultures of LB (5 mL), which were then grown overnight at 37 8Cw ith shaking at 180 rpm. The starter cultures served as inocula for 1L cultures of LB in which cells were grown until the optical density (OD 600 ) had reached av alue of 0.8. Expression of HbpA was then induced by the addition of isopropyl b-d-1-thiogalactopyranoside (IPTG, 1mm). The cultures were then incubated at 18 8Ci na no rbital shaker at 180 rpm for approximately 18 h. The cells were harvested by centrifugation for 15 min at 4225 g in aS orvall GS3 rotor in aSorvall RC5B Plus centrifuge.
Cell pellets were resuspended in Tris·HCl buffer (20 mL, 50 mm, pH 7.5) containing NaCl (300 mm;h enceforth referred to as "buffer A"), per litre of cell culture. The cell suspensions were then sonicated for 10 45 sb ursts at 4 8Cw ith 30 si ntervals. The soluble and insoluble fractions were separated by centrifugation for 30 min at 26 892 g in aS orvall SS34 rotor.T he crude cell lysate from 1Lcell culture was filtered and then loaded onto a5mL His-Trap FF crude column (GE Healthcare), which was washed with buffer A. The column was eluted with imidazole (linear gradient of 20-300 mm)o ver 20 column volumes at af low rate of 2.5 mL min À1 .F ractions containing HbpA, as determined by SDS-PAGE analysis were combined, and the volume was reduced to yield aprotein concentration of 10 mg mL À1 .The concentrated protein was ad eep yellow colour,s uggestive of bound oxidised FAD within the protein.
Protein crystallisation:C rystallisation conditions for HbpA were determined by using aM osquito robot (TTP Labtech, Melbourn, UK) in conjunction with commercially available screens. Trials were conducted in 96-well plates with sitting-drop format by using 300 nL drops (150 nL protein plus 150 nL precipitant solution). Positive hits were scaled up in 24-well Linbro dishes by using the hanging-drop method of crystallisation, with crystallisation drops containing protein solution (1 mL) and precipitant reservoir (1 mL). The best HbpA crystals were obtained in drops that contained PEG 3350 (18 %, w/v), KSCN (0.15 m)i nB is·Tris propane buffer (pH 5.5), with ap rotein concentration of 10 mg mL À1 .D espite the yellow colour of the protein solution and of the crystals, the occupancy of flavin was found to be poor in the active sites of the solved structure;t his led to the apo structure described above.I n order to obtain as tructure with bound FAD, it therefore proved necessary to soak the HbpA crystals in as olution of the mother liquor with added FAD( 1mm), as well as 20 %e thylene glycol (as cryoprotectant) for 5min prior to flash-cooling in liquid nitrogen for X-ray diffraction analysis. Crystals were tested for diffraction by using aR igaku Micromax-007HF X-ray generator fitted with Osmic multilayer optics and am ar345 imaging plate detector (Marresearch, Norderstedt, Germany). Those crystals that diffracted to greater than 3resolution were retained for full dataset collection at the synchrotron.
Data collection, structure solution, model building and refinement of HbpA:C omplete datasets for apo-HbpA and its FADc omplex were collected on beamlines I04 and I04-1, respectively,a tt he Diamond Light Source (Didcot, UK). Data were processed and integrated by using XDS [36] and scaled by using SCALA [37] within the Xia2 processing system. [38] Data collection statistics are given in Ta ble 2. The structure of HbpA was solved by using BALBES, [39] which selected am onomer of structure PDB ID 3IHG, the aklavinone-11-hydroxylase, [17] as as earch model. The solution contained four molecules in the asymmetric unit;t his represents one tetramer.T he structures were built and refined by using iterative cycles of Coot [40] and REFMAC [41] employing local NCS restraints in each case. For the FADc omplex of HbpA, the omit maps, after building and refinement of the protein backbone and side chains, revealed clear residual density at the active sites, which was successfully modelled as FAD. The final structures of apo-HbpA and the FAD complex had R cryst /R free values of 19.1/24.1 and 17.0/20.3, respectively.T he structures were validated by using PROCHECK. [42] Refinement statistics are presented in Ta ble 2. The Ramachandran plot for apo-HbpA showed 93.5 %o fr esidues to be situated in the most favoured regions, in addition 5.6 %w ere allowed and 0.9 % were outlier residues. For the FADc omplex, the respective values were 95.9, 3.4 and 0.7 %. Coordinates and structure factors for both apo-HbpA and the FADc omplex have been deposited in the Protein Data Bank with the accession codes 4CY6 and 4CY8, respectively.
Site-directed mutagenesis:T he mutants of HbpA, His48Ala, Asp117Ala and Arg242Ala were generated by using aC lontech In-Fusion kit, according to the manufacturer's instructions, from the primers listed in Ta ble 3. Mutants genes were expressed, and proteins were purified according to the methods described above for the wild-type.
Docking:A utomated docking was performed by using AutoDock Vina 1.1.2. [43] Am onomer structure of the FADc omplex of HbpA was prepared by using AutoDock utility scripts. Coordinates for the product 23DHBP (4)w ere prepared by using PRODRG. [44] The active site of HbpA was contained in ag rid of 30 3224 with 0.375 s pacing, centred around the catalytic centre (co-ordinates: À9.947, 34.123, À16.228) which was generated using AutoGrid in the AutoDock To ols interface. The dockings were performed by Vina, therefore the posed dockings were below 2 rmsd. The results generated by Vina were visualised in AutoDock To ols 1.5.6 where the ligand conformations were assessed both upon lowest Vina energy,b ut also according to criteria established by previous studies on mechanism and substrate selectivity in this enzyme class (vide supra).