RETUNING RIESKE-TYPE OXYGENASES TO EXPAND SUBSTRATE RANGE

Rieske-type oxygenases are promising biocatalysts for the destruction of persistent pollutants or for the synthesis of fine chemicals. In this work, we explored pathways through which Rieske-type oxygenases evolve to expand their substrate range. BphAE(p4), a variant biphenyl dioxygenase generated from Burkholderia xenovorans LB400 BphAE(LB400) by the double substitution T335A/F336M, and BphAE(RR41), obtained by changing Asn(338), Ile(341), and Leu(409) of BphAE(p4) to Gln(338), Val(341), and Phe(409), metabolize dibenzofuran two and three times faster than BphAE(LB400), respectively. Steady-state kinetic measurements of single- and multiple-substitution mutants of BphAE(LB400) showed that the single T335A and the double N338Q/L409F substitutions contribute significantly to enhanced catalytic activity toward dibenzofuran. Analysis of crystal structures showed that the T335A substitution relieves constraints on a segment lining the catalytic cavity, allowing a significant displacement in response to dibenzofuran binding. The combined N338Q/L409F substitutions alter substrate-induced conformational changes of protein groups involved in subunit assembly and in the chemical steps of the reaction. This suggests a responsive induced fit mechanism that retunes the alignment of protein atoms involved in the chemical steps of the reaction. These enzymes can thus expand their substrate range through mutations that alter the constraints or plasticity of the catalytic cavity to accommodate new substrates or that alter the induced fit mechanism required to achieve proper alignment of reaction-critical atoms or groups.

to Gln 338 Val 341 Phe 409 , metabolize dibenzofuran two and three times faster than BphAE LB400 , respectively. Steady-state kinetic measurements of single-and multiple-substitution mutants of BphAE LB400 showed the single Thr 335 Ala and the double Asn 338 Gln Leu 409 Phe substitutions contribute significantly to enhanced catalytic activity towards dibenzofuran. Analysis of crystal structures showed that the Thr 335 Ala substitution relieves constraints on a segment lining the catalytic cavity allowing a significant displacement in response to dibenzofuran binding. The combined Asn 338 Gln Leu 409 Phe substitutions alter substrate-induced conformational changes of protein groups involved in subunit assembly and in the chemical steps of the reaction. This suggests a responsive induced-fit mechanism that retunes the alignment of protein atoms involved in the chemical steps of the reaction. These enzymes can thus expand their substrate range through mutations that alter the constraints or plasticity of the catalytic cavity to accommodate new substrates or that alter the induced-fit mechanism required to achieve proper alignment of reaction-critical atoms or groups.
In addition, crystal structures of the oxygenase component of carbazole 1,9a-dioxygenase and of the binary complex of the oxygenase and ferredoxin components provided evidence that conformational changes are required to suitably align the Rieske clusters of the ferredoxin and oxygenase components (24). These observations are consistent with a mechanism whereby an induced-fit process is involved in ROs substrate binding and catalytic function. Understanding how conformational adjustments influence the turnover rate and how they can be modified to enhance activity toward new substrates will aid the development of new, better-performing catalysts.
Unlike biphenyl, the fused rings of dibenzofuran are locked in a co-planar conformation (Fig. 1), and this molecule is poorly oxygenated by BphAE LB400 (25,26). In previous reports (7, (7). The library was produced by changing Asn 338 and Ile 341 of BphAE p4 simultaneously by saturation mutagenesis (7), but an additional spontaneous mutation Leu 409 Phe occurred in this mutant. In this study, in order to gain more insight into the pathways through which ROs evolve to expand their substrate range, we identified the mutations in BphAE p4 and in BphAE RR41 that contribute most to their enhanced activity toward dibenzofuran and analyzed the crystal structures of BphAE RR41 and its dibenzofuran-bound form to evaluate the consequences of the mutations.
A previously described two-step site-directed mutagenesis protocol (7) was used to create a set of mutants representing the six mutants that can be produced by single and double substitutions of Asn 338 Gln, Ile 341 Val and Leu 309 Phe of BphAE p4 (Table 1). These bphAE mutants were cloned in pQE31 or in pET14b. DNA protocols were generally according to Sambrook et al. (29). DNA from each mutant was sequenced at the Genome Quebec DNA Sequencing Center (Montreal, Quebec, Canada). Biphenyl and dibenzofuran were of the highest purity grade available from AccuStandard (New Haven, CT). Purified enzyme preparations were also analyzed by HPLC gel filtration chromatography using a Waters Protein Pak 300 SW column (7.8 x 300 mm), as described previously (31). Monitoring enzyme activity with purified enzyme preparations-Reconstituted His-tagged purified BPDO preparations were used to monitor enzyme activity and metabolite production. In this case, the genes expressing each enzyme component were cloned into pET-14b (Novagen, Madison, WI) and expressed in E. coli C41(DE3). The components were produced as recombinant Histagged protein and purified by affinity chromatography on high performance Ni-Sepharose resin (GE Healthcare) (7). The concentration of each purified component was determined by spectrophotometry (31)(32)(33). Enzymatic reactions were performed as described previously (7) at 37 o C, in pH 6.0 MES 50 mM buffer, and in a volume of 400 l containing 1.2 nmol of each of the His-tagged enzyme component and 200 nmol NADH. Substrate depletion and metabolite production were analyzed and quantified by gas chromatographymass spectrometry (GC-MS) using previously published protocols (7). The steady-state kinetic parameters of all BphAEs were determined by recording oxygen consumption rates using a Clarke-type Hansatech model DW1 oxygraph (34) for concentrations of biphenyl and dibenzofuran varying between 5 and 150 M. Kinetic parameters reported in this investigation were obtained from analysis of at least two independently produced preparations tested in triplicate. Crystal structure analyses-Purification, crystallization, and preliminary X-ray diffraction properties of BphAE RR41 have been communicated elsewhere (35). The procedures to prepare crystals of BphAE RR41 and its dibenzofuran-bound form were identical to those described for BphAE p4 (17). The crystal structures were obtained and analysed using the same approaches and software used in studies of BphAE p4 (17). Crystal structures of BphAE RR41 and its dibenzofuran-bound form were compared to crystal structures of BphAE LB400 (RCSB Protein Data accession codes 2XR8) and its biphenyl-bound form (2XRX) as well as BphAE p4 (2XSO) and its 2,6-dichlorobiphenylbound form (2XSH). PDB accession codes-The coordinates have been deposited with the RCSB Protein Data Bank (http://deposit.rcsb.org/) under accession codes 2YFI for BphAE RR41 and 2YFJ for its complex with dibenzofuran. RR41 and their variants with dibenzofuran-Based on the sum of areas under GC-MS peaks of metabolites produced when 1.2 nmol enzyme was incubated for 2 min with 100 M dibenzofuran, BphAE p4 and BphAE RR41 produced, respectively, three and four times more metabolites than BphAE LB400 (Fig.  2) . Consistent with the single time-point measurements, the apparent k cat value for BphAE RR41 is approximately 1.5 times higher than that of BphAE p4 and 3 times higher than for BphAE LB400 based on the oxygen consumption rates recorded for variable concentrations of dibenzofuran ( Table 2). These results show the superior ability to metabolize dibenzofuran of BphAE RR41 compared to BphAE p4 and BphAE LB400 .

Steady-state kinetics of BphAE p4 , BphAE
In order to identify the mutations that contribute most to the enhanced activity of BphAE p4 and BphAE RR41 toward dibenzofuran, we assayed all mutants carrying single or multiple mutations at positions 335, 336, 338, 341 or 409 (Table 1). Based on the sum of areas under GC-MS peaks, the amounts of metabolites produced by the T335A mutant and by BphAE p4 (T335A/F336M) were similar and about three times higher than the amounts for BphAE LB400 and its F336M mutant (Fig. 2) . The superior ability of the T335A mutant to metabolize dibenzofuran was confirmed by steady-state kinetics (Table 2). Furthermore, the single F336M substitution lowered k cat and k cat /K m for the reaction with biphenyl. This identified the T335A substitution as responsible for the enhanced ability of BphAE p4 to metabolize dibenzofuran.
The apparent k cat values for the mutants T335A/F336M/I341V and T335A/F336M/I341V/L409F toward biphenyl and dibenzofuran were lower than for BphAE p4 . In addition, introducing the single Ile 341 Val or the double Ile 341 Val Leu 409 Phe mutations into BphAE p4 did not contribute to enhanced activity toward dibenzofuran. We did not determine steady-state kinetic parameters for variant T335A/F336M/L409F. However, the observation that recombinant E. coli cells producing this enzyme did not degrade dibenzofuran more efficiently than BphAE p4 (not shown) indicates that changing Leu 409 of BphAE p4 to Phe did not influence activity toward dibenzofuran. Variants T335A/F336M/N338Q and T335A/F336M/N338Q/I341V were poorly active toward biphenyl and dibenzofuran because they are not assembled correctly (see below). The steady-state kinetic parameters obtained with these two variants were too difficult to determine accurately and therefore are not reported here. On the other hand, the apparent k cat value for T335A/F336M/N338Q/L409F toward dibenzofuran was in the same range or even higher than for BphAE RR41 (T335A/F336M/N338Q/I341V/L409F) ( Table 2). Furthermore, replacing Asn 338 Leu 409 of BphAE p4 with Gln 338 Phe 409 created a double mutant exhibiting enhanced activity toward dibenzofuran.
The poor activity of variants T335A/F336M/N338Q and T335A/F336M/N338Q/I341V is the result of hexamer misassembly-Unlike BphAE RR41 and T335A/F336M/N338Q/L409F, variants T335A/F336M/N338Q and T335A/F336M/N338Q/I341V were poorly active, which suggests a detrimental effect of the Asn 338 Gln substitution on enzyme activity. When the proteins of IPTG-induced recombinant E. coli cells producing variant BphAEs were extracted under denaturing conditions and separated by SDS-PAGE, the intensities of bands corresponding to the  and the  subunits were similar for all strains (Fig. 3A). This shows the level of expression of BphAE is similar for all IPTG-induced E. coli clones expressing these variants. However, analysis of cell extracts prepared under non-denaturing conditions revealed significantly lower intensities of the bands corresponding to the  subunits for variant T335A/F336M/N338Q and T335A/F336M/N338Q/I341V compared to the other variants (Fig. 3B). This suggests misassembly or early dissociation of     hexamers resulting in loss of  subunits into the non-soluble protein fraction. HPLC gel filtration analysis of freshly purified His-tagged BphAE of variants T335A/F336M/N338Q and T335A/F336M/N338Q/I341V confirmed that less than 10% of each protein preparation exhibited the expected  3  3 association pattern (not shown). Therefore, the Asn 338 Gln substitution hinders hexamer assembly unless a concomitant Leu 409 Phe substitution is introduced. Nevertheless, the double Asn 338 Gln Leu 409 Phe substitution is beneficial for enhancing catalytic properties towards dibenzofuran.
Overall structure of BphAE RR41 -The overall structure of BphAE RR41 and of its dibenzofuran-bound form are very similar to the crystal structures of the BphAE p4 :2,6-dichlorobiphenyl complex (17), which contains triplets of  dimers associated into two hexamers in the asymmetric unit (chains ABCDEF and chains GHIJKL). For both structures, the final refined models contain residues Asn 18 to Phe 143 plus Phe 153 to Pro 459 of each  subunit, residues Phe 9 to Phe 188 of each  subunit and 1217 water molecules. The diffraction data and refined models are characterized in Table  3.
Superposition of all C  atoms for the  dimer of chains AB with chains CD-KL yielded rmsd values of 0.2-0.3 Å. The poorly ordered residues and protein segments were the same as observed for BphAE p4 (17), including segments comprising residues Ile 247 to Lys 263 and Glu 280 toVal 287 of the  subunit and 9-17 and 158-164 of the  subunit. In the BphAE RR41 :dibenzofuran complex structure, dibenzofuran could be identified clearly in (F o -F c ) difference Fourier maps in all active sites of the ABCDEF hexamer. Similar to BphAE LB400 :biphenyl, a water molecule lies between the Fe ++ and dibenzofuran at approximately 2 Å from the catalytic iron. Electron density maps of the active site residues of the BphAE RR41 and of its dibenzofuran-bound form are shown for dimer AB in Fig. 4.
When the BphAE RR41 :dibenzofuran BphAE LB400 :biphenyl structures are superposed, the positions of the carbons targeted for hydroxylation, C-4a and C-4 of dibenzofuran and C-2 and C-3 of biphenyl, are nearly the same in  dimers AB, CD and EF (not shown). In  dimers AB, CD and EF of BphAE RR41 :dibenzofuran, the substrate is in an orientation that would favor 4,4a angular attack. This is unexpected since biochemical data revealed dioxygenation of the lateral 1,2 and 3,4 carbons of dibenzofuran as by far most favored for BphAE RR41 (7). In the crystal structure, the furan ring's oxygen atom contacts the water ligand of the active site Fe ++ atom. Based on observations drawn from the naphthalene dioxygenase (NDO) crystal structure, Karlsson et al. (36) proposed a reaction cycle for ROs in which Fe ++ coordinates this water in states prior to dioxygen binding. When dioxygen binds it intercalates side-on between the iron and the substrate displacing the water. Thus, a subsequent adjustment in substrate position or orientation is possible such that the lateral attack would occur. In essence, the dibenzofuran-O-water interaction seen in the crystal form could produce a false inference about the points of attack because the structure reports on a state prior to dioxygen binding.
Structural analysis of the influence of residue 335 on catalytic properties toward dibenzofuran-The average active site cavity volume of the αβ dimers AB, CD, EF of BphAE RR41 was comparable to that of BphAE p4 (17) (1073 Å 3 as calculated using CASTp program (37)). The corresponding atoms of the reactive ring of dibenzofuran and of biphenyl interact with the same residues of the  subunits of BphAE RR41 and BphAE LB400 (Gln 226 , Phe 227 , Asp 230 , Met 231 , Leu 233 , Ala 234 , His 323 and Leu 333 ) and they are located at approximately the same distances (not shown). Therefore, neither the overall size of the cavity nor the constraints on the reactive ring are affected by the mutations introduced in BphAE RR41 . The residues lining the distal portion of the BphAE RR41 catalytic cavity are the same as those in BphAE LB400 and BphAE p4 (Phe 384 , Phe 378 , Val 287 , Ser 283 , Phe/Met 336 , Leu 333 , Gly 321 , Tyr 277 , His 239 , Ala 234 and Met 231 ) (Fig. 5). However, the overall shape of the cavity of BphAE RR41 :dibenzofuran differs significantly from that of BphAE LB400 :biphenyl, but is similar to that of BphAE p4 :2-chlorobiphenyl (Fig. 5). This is caused principally by the replacement of Phe 336 by Met combined with the conformational freedom of Gly 321 . The new catalytic properties of BphAE p4 toward dibenzofuran (and in part those of BphAE RR41 ) can thus be attributed to structural changes in the distal portion of the substrate binding pocket. As noted for BphAE p4, the Thr 335 Ala mutation relieves constraints on the Val 320 :Gln 322 segment allowing displacement of the Gly 321 carbonyl such that it moves away from the substrate (17). In this case, the removal of Gly 321 from dibenzofuran reduces the influence it exerts on the substrate's distal ring. This is significant because dibenzofuran is obligatory coplanar: an altered placement of the distal ring would influence the orientation of the proximal ring inside the catalytic pocket.
Structural analysis of the influence of residues 338 and 409 on enzyme stability-Based on the crystal structure of BphAE RR41 , as well as the structures of BphAE LB400 and BphAE p4 (17), Gln 338 and Phe 409 are too distant from each other to interact. In order to understand the effect of these two mutations we need to examine closely the overall structure of the  and  subunits and the contacts at , , and  interfaces.
The overall crystal structure of the  subunit of BphAE RR41 is very similar to that of other biphenyl dioxygenases (16,17) and naphthalene dioxygenase (2). It includes a long twisted sixstranded  sheet, with three helices on the inward side of the sheet and a loop made of residues 9-20 ( Fig. 6A) on its outward side. Many polar interactions are uniformly distributed between the -sheet residues of vicinal  subunits and between two of the helices and -sheet residues of the vicinal subunit. In addition, helix 3 and strand  are in contact with the  subunit. This suggests the  subunit plays a key role in subunit assembly.
As reported for other dioxygenases (2,16), the  subunit comprises two domains, the Rieske and catalytic domains. Unlike the  subunits, the crystal structures show unevenly distributed contacts between vicinal  subunits; these occur principally at the junction between the Rieske domain of one  subunit and the catalytic domain of its vicinal subunit (Fig 6A).
The Rieske domain is dominated by antiparallel  strands from which two hairpin structures protrude to form two fingers that hold the [2Fe-2S] center. The catalytic domain contains the catalytic Fe ++ , which lies against a eight-stranded antiparallel  sheet on one side and is surrounded on the other sides by helices and loops (Fig. 6B). Residues Arg 101 , His 102 , Arg 103 and Gly 104 of the Rieske domain form the tip of hairpin 1. This short segment (in blue on Fig. 6B and C) is embedded inside a matching trough of the vicinal  subunit; it faces helix 6 of the catalytic domain, comprised of residues Trp 220 to Ser 229 , and it also contacts a short segment ( comprised of residues Pro 408 , Phe 409 and Asn 410 and located at the edge of the catalytic domain. Agr 103 forms a polar contact with Glu 225 of helix 6. Arg 101 forms polar contacts with Pro 408 and Asn 410 , and Arg 104 forms a polar contact with Asn 410 (Fig. 6C). Therefore, the crystal structures show residue Phe 409 is located within a stretch of amino acids that appears to play an important role in subunit assembly and/or in maintaining the stability of the oligomeric structure. In BphAE RR41 , Phe 409 is approximately 4.6 Å from Phe 222 of helix 6 (Fig.  6C). Alignment of dimers AB, CD, EF, GH, IJ and KL of BphAE RR41 and its dibenzofuran-bound form with dimer AB, CD, EF, GH, IJ and KL of substrate-free and bound forms of BphAE LB400 or BphAE p4 shows Phe 409 of BphAE RR41 aligns very well with Leu 409 of BphAE LB400 or BphAE p4 (not shown). In the latter enzymes, however, Leu 409 is at an average distance of 5.4 Å from Phe 222 . This could explain why replacing Leu 409 of BphAE p4 with a larger side chain in Phe 409 suppresses the negative impact of the Asn 338 Gln mutation on hexamer assembly. Through its interaction with Phe 222 , Phe 409 seems to help stabilize subunit assembly by reinforcing the role played by segment Pro 408 -Asn 410 in holding the subunits together.
Structural analysis of the influence of residues 338 and 409 on catalytic properties-Prior studies identified mutations within a subsequence called region III that influenced the oxygenase's catalytic properties (12)(13)(14). This region includes a loop between strands 18 and 19 and a portion of strand 19. Residues 338 and 341 are both located on strand 19 (Fig. 6C). Superposition of the catalytic domains of BphAE RR41 and BphAE LB400 reveals minor variations in strand 19 that can be attributed to the longer side chain of Gln 338 in BphAE RR41 (not shown). Strand 19 faces helix 12, which interacts with the tip of hairpin-2 of the vicinal Rieske domain and includes Fe ligand Asp 388 . The tip of Rieske domain hairpin-2 includes Ser 121 , Tyr 122 and His 123 . These residues make polar contacts with Thr 237 and Thr 238 located at the junction between helices 7 and 8 on which are located Fe ligands His 233 and His 239 . In addition, Ser 121 and His 123 make polar contacts with Gln 226 and Asp 230 , two residues believed to be involved in the reaction mechanism (36,38) and found in the catalytic cavity at the level of the proximal ring of the substrate (17) (Fig. 6B). Furthermore, Tyr 122 has a polar contact with Trp 392 and Ser 121 with Asn 391 of helix 12. Gln 338 forms two polar contacts with Arg 340 and the latter forms two polar contacts with Glu 385 of helix 12. Arg 340 is also close enough from Glu 385 to form a salt bridge with this residue (Fig. 6C). Therefore, Arg 340 and Gln 338 are located such that their conformation can influence the distribution in space of helices 12, 8, 7 and 6, which are critical for catalytic activity and subunit assembly.
Analysis of the crystal structure did not suggest a clear-cut mechanism by which Gln 338 and Arg 340 exert these effects. However, the longer length of Gln 338 side chain compared to Asn 338 might disturb a key state not observed in the crystal or the internal dynamics of the protein. With respect to the latter possibility, it is clear from structural analysis that residues on helix 12 move considerably during substrate binding showing this protein segment is rather adaptable (Fig. 7). This movement is likely required to suitably align the reactive atoms for progression along the chemical reaction.
In ROs, the proximity between the Rieske cluster and the iron in the active site of the adjacent  subunit is consistent with a mechanism involving a transfer of electron across the interface between two subunits (2,16,17). This is corroborated by the fact that full activity requires that  and  subunits associate into a  3   configuration (31). Such a mechanism must demand precise alignment of the amino acids involved in electron transfer between the Rieske cluster and the mononuclear iron of each adjacent  subunits highlighting the importance of the protein atoms involved in the subunit interface.
Therefore, the crystal structure analysis is consistent with an induced-fit response required to reorganize the active site and facilitate the interplay between protein atoms critical for the reaction. The Asn 338 Gln substitution might disturb the conformation of helices 12 and 6 resulting in subunit instability or misassembly. Conversely, the double Gln 338 Phe 409 mutation may affect the conformational fluctuations of these helices in such a way that it enhances the roles of protein residues such as Asn 388 , Gln 226 and Asp 230 that are located on the helices and presumed to be involved in the chemical steps of the reaction (2,38).

DISCUSSION
In this study we examined the crystal structural of BphAE RR41 , an evolved RO that oxidizes dibenzofuran more efficiently than its BphAE LB400 and BphAE p4 parents. In spite of the limitations of crystal structure analyses, the study revealed two pathways through which ROs evolve to expand their substrate range.
Traditionally, enzyme engineering to alter the substrate range involves mutations at residues lining the catalytic pocket. This approach has been applied successfully in many circumstances (39)(40)(41)(42)(43)(44)(45). Reducing the size of a side chain or altering charge distributions can generate enzyme with new catalytic properties.
However, other studies have shown that several residues not in direct contact with the substrate can significantly change BPDO's catalytic properties toward biphenyl analogs (21,22,44). In this work we confirm the importance of the Thr 335 Ala mutation. In altering the plasticity of the catalytic cavity, this mutation allows the carbonyl of residue Gly 321 to move away from the substrate. In a previous work, we showed this movement was required to increase the space available to bind the bulky 2,6-dichlorobiphenyl in a productive orientation (17). Because dibenzofuran is obligatory co-planar, any misplacement of the distal ring would influence the orientation of the proximal ring inside the catalytic pocket. Therefore, consistent with an induced-fit mechanism, in BphAE p4 and BphAE RR41, the displacement of Gly 321 appears to be required to reduce the influence it exerts through atomic interactions on the substrate's distal ring.
In this work, we highlighted a second and more subtle route to changes in substrate range, which implies that in ROs, either one or both of the induced-fit or protein dynamic processes are involved to place the protein atoms involved in the reaction into proper relationships that facilitate catalysis. The reaction catalyzed by ROs is complex; it not only involves substrate binding and release of product, but also one dioxygen molecule is required in the reaction and electrons must be transferred from the ferredoxin component to the Rieske cluster of one  subunit and then to the catalytic iron of the vicinal  subunit. Furthermore, a recent report showed residues at the interface between the Rieske domain and the catalytic domain move during formation of the complex between the carbazole 1,9a-dioxygenase's oxygenase and ferredoxin components (24). This implies reaction-critical atoms from the Rieske domain must align properly with those of the vicinal catalytic domain and the reaction-critical atoms of the catalytic domain must align properly to work together during the catalytic process. Structural analysis shows residues located on secondary structures 6 and 12 are involved in subunit assembly and biochemical data suggest they are involved in the catalytic reaction (electron transfer and protonation) (2,38). The fact that these residues move during substrate binding is consistent with a substrate-induced retuning process required to suitably align the protein atoms involved in the chemical steps of the reaction. In such a context, by altering the interactions occurring between secondary structure elements surrounding the catalytic center, the Asn 338 Gln mutation generates a protein unable to stabilize the  3  3 assembly previously shown to be required for activity (31). However, the double Asn 338 Gln and Leu 409 Phe substitution generates an  subunit that supports a stable hexamer and where the retuning process is improved compared to its BphAE LB400 and BphAE p4 parents, resulting in a more efficient and faster catalytic reaction. ROs can thus be engineered to enhance their catalytic properties toward new substrates by altering the process involved in fine-tuning the interplay between of the reaction-critical atoms.
Many questions remain unanswered; crystal structure analysis did not determine a clear-cut mechanism by which the double Asn 338 Gln Leu 409 Phe substitution affects the enzyme structure and catalytic properties, and our data do not determine which of the enzymatic steps are accelerated during the reaction. Although the data do not provide any direct demonstration that the Asn 338 Gln and Leu 409 Phe substitutions either affect an induced-fit or protein dynamic mechanism involved in the catalytic reaction, it is clear from crystal structure analysis that these residues occupy strategic positions whereby they can interact with reaction-critical protein atoms/groups and affect oligomeric assembly. Furthermore, residues of helix 12 and 6, and especially Asp 388 and Gln 226 , which are postulated to play a key role in the catalytic reaction (36) moved significantly during substrate binding in all variants (BphAE LB400 , BphAE p4 and BphAE RR41 ).
Altogether, our analysis shows that evolving ROs to change their substrate specificity is a rather complex enterprise that does not involve exclusively mutations at key residues in direct contact with the substrate. It appears that some mutations affect key residues associated with necessary conformational changes that are more difficult to identify by a rational approach, but that are required to allow productive or improved interplay of reaction-critical atoms both inside and outside of the substrate binding pocket.

FOOTNOTES
This work was supported by the Natural Sciences and Engineering Research Council of Canada (NSERC) (Grant #RGPIN/39579-2007). X-ray diffraction data were collected at APS using Southeast Regional Collaborative Access Team (SER-CAT) 22-ID beamline; supporting institutions may be found at www.ser-cat.org/members.html. PK thanks the Department of Science and Technology, India for providing a young scientist grant.