Alteration of the Co-substrate Selectivity of Deacetoxycephalosporin C Synthase

Deacetoxycephalosporin C synthase is an iron(II) 2-oxoglutaratedependent oxygenase that catalyzes the oxidative ring-expansion of penicillin N to deacetoxycephalosporin C. The wild-type enzyme is only able to efficiently utilize 2-oxoglutarate and 2-oxoadipate as a 2-oxoacid co-substrate. Mutation of arginine 258, the side chain of which forms an electrostatic interaction with the 5-carboxylate of the 2-oxoglutarate co-substrate, to a glutamine residue reduced activity to about 5% of the wild-type enzyme with 2-oxoglutarate. However, other aliphatic 2-oxoacids, which were not co-substrates for the wild-type enzyme, were utilized by the R258Q mutant. These 2-oxoacids “rescued” catalytic activity to the level observed for the wild-type enzyme as judged by penicillin N and G conversion. These co-substrates underwent oxidative decarboxylation as observed for 2-oxoglutarate in the normal reaction with the wild-type enzyme. Crystal structures of the iron(II)- 2-oxo-3-methylbutanoate (1.5 Å), and iron(II)-2-oxo-4-methylpentanoate (1.6 Å) enzyme complexes were obtained, which reveal the molecular basis for this “chemical co-substrate rescue” and help to rationalize the co-substrate selectivity of 2-oxoglutaratedependent oxygenases.

DAOCS is a member of the iron(II) and 2-oxoglutarate-dependent oxygenase family, which catalyze a wide variety of oxidative reactions (6,7). DAOCS, DACS, and DAOC/DAC synthase belong to a subgroup of more closely related enzymes, which have significant primary sequence homology to one another (6). Also included in this group are two enzymes which do not use 2-oxoglutarate as a co-substrate, isopenicillin N synthase (IPNS), the iron-dependent oxidase responsible for formation of the penicillin nucleus, and 1-amino-1-carboxycyclopropane oxidase (ACCO) which catalyzes the last step during ethylene biosynthesis in plants.
Mechanistic understanding of iron(II), 2-oxoglutaratedependent oxygenases have been significantly advanced by the recently determined crystal structures of DAOCS (1,8,9) and clavaminic acid synthase (10). The DAOCS crystal structures revealed the presence of a number of arginine residues within the active site, with arginine 258 (part of a conserved RXS motif) being involved in co-substrate binding (1,8). The equivalent RXS motif residues in Aspergillus nidulans IPNS, arginine 281, and serine 283, bind the ␣-carboxylate group of the L-␦-(␣-aminoadipoyl)-L-cysteinyl-D-valine substrate (11). Both of these residues have been shown to be important for substrate binding in the C. acremonium isoenzyme by site-directed mutagenesis (12,13).
Here we report studies on the 2-oxoglutarate co-substratebinding function of arginine 258 in DAOCS by site-directed mutagenesis. Activity of the R258Q mutant was reduced in the presence of 2-oxoglutarate, but could be fully restored using aliphatic 2-oxoacids as alternative co-substrates. The process is called "chemical co-substrate rescue." The molecular basis for this phenomenon has been investigated by crystallographic analyses of the R258Q mutant complexed with iron(II) and alternative 2-oxoacids. The results have implications for clinically observed mutations to phytanoyl-CoA 2-hydroxylase that result in Refsum's disease. 2

EXPERIMENTAL PROCEDURES
Materials-All chemicals were obtained from the Sigma-Aldrich Chemical Co. or E. Merck and were at least analytical grade or higher. Oligonucleotides were synthesized by V. Cooper (Dyson Perrins Laboratory, Oxford) using Applied Biosystem DNA synthesizers (Models 380B and 394). Reagents were supplied by Roche Molecular Biochemicals (ATP); MBI (1 kilobase and 100 base pair DNA gel markers); Bio-Rad (mutagenesis reagents, competent cells, Bradford Reagent); New England Bio-Labs (enzymes for molecular biology); Promega (Wizard Plus miniprep DNA purification system, Wizard Plus SV miniprep DNA purification system); and Qiagen Ltd. (RNase A).
Site-directed Mutagenesis-The primer (5Ј-ACTGGAGGTCTGGCT-GCTGC-3Ј) was annealed to the DAOCS gene in the pET24a vector (1) by heating to 70°C and cooled over 30 min to 35°C, and mutagenesis performed by the method of Kunkel et al. (14,15). The presence of the R258Q mutation was confirmed by automated sequencing, using an ABI Prism sequencer.
Protein Expression, Purification, and Activity Measurements-Wildtype and R258Q mutant enzymes were purified as described (1), except that a 80 -320 mM NaCl gradient over 800 ml was used to elute the Q-Sepharose anion-exchange column. HPLC assays were carried out as described in Ref. 1 with 5 mM of the appropriate 2-oxoacid substituted for 2-oxoglutarate when assaying the R258Q mutant. The products resulting from 2-oxoacid oxidation were characterized by larger scale incubation (2 ml) with the required 2-oxoacid (10 mol) in the presence and absence of penicillin G (20 mol) (1). 2-Oxoglutarate conversion was assayed using [ 14 C 1 ]2-oxoglutarate as reported (16). Apparent kinetic parameters (Ϯ S.D.) were determined by HPLC assay (1) and the spectrophotometric method of Dubus et al. 3 At least six concentrations of the variable substrate was used (in triplicate) at a single defined concentration of the second substrate (Table II). Data from the HPLC (1) and spectrophotometric 3 assays were derived as described.
Crystallization, Data Collection, and Analysis of the R258Q Mutant-DAOCS was crystallized by vapor diffusion using the hanging drop method at 14°C, using protein (2 l) and precipitating solution (2 l). The precipitating solution contained 100 mM HEPES-NaOH, pH 7.0, 1.5-1.7 M ammonium sulfate, 3-6% (w/v) glycerol, and 5 mM 2-oxoglutarate, 2-oxo-4-methylpentanoate (␣-ketoisocaproate) or 2-oxo-3methylbutanoate (␣-ketoisovalerate). Crystals in the R3 space group resembling those of the wild-type enzyme (1) appeared after about 1 week. Prior to data collection crystals were soaked under anaerobic conditions in mother liquor supplemented with 5 mM iron(II) sulfate, followed by cryo-protectant containing 25% (v/v) glycerol. Data was collected using a Rigaku rotating anode generator fitted with 345 MAR Research image plate detector and at Beamline 9.6 of the SRS, Daresbury, United Kingdom.
Images were processed with the program MOSFLM (17) and data reduced with the CCP4 suite (18). Each data set was assessed for merohedral twinning based on the H-statistic (19). 4 The approximate twin fraction obtained was used in the program DETWIN (18), and a more accurate value for the twin fraction was obtained by minimization of the residual twinning.
The starting model was that of the 2-oxoglutarate complex (9) with all water and ligands removed and confirmed with the iron-2-oxoglutarate complex of DAOCS (8). Rigid body refinement, simulated annealing, and 12-19 rounds of energy minimization and B-factor refinement were carried out using the CNS suite (20,21). The iron and ligands were built after 5-6 rounds of refinement. Iron-ligand distances were determined using restrained refinement, with initial parameters based on those previously observed (1). Water molecules were added using the CNS procedure water_pick.inp (20). Visual inspection of maps was accomplished using O (22).

Biochemical and Kinetic Analysis of the R258Q Mutant-The
R258Q mutant was obtained via site-directed mutagenesis, and the enzyme purified to Ͼ99% homogeneity (as judged by SDSpolyacrylamide gel electrophoresis) (1). Analyses by ESI MS (observed 34,532 Ϯ 6 Da; calculated 34,525 Da) and DNA sequencing indicated that no undesired mutations were present, while CD analyses suggested there were no gross changes to the secondary structure of the enzyme.
Both the wild-type enzyme and the R258Q mutant were analyzed for their ability to catalyze the conversion of penicillin N and penicillin G to their respective products using discontinuous HPLC assays for cephem production ( Table I). The results demonstrated that wild-type DAOCS has, at most, only traces of activity with all 2-oxoacids tested other than 2-oxoglutarate and 2-oxoadipate. The R258Q DAOCS mutant was able to convert 2-oxoglutarate to succinate and carbon dioxide in the presence of either penicillin N and penicillin G. Two experiments were carried out (in duplicate) to investigate the coupling of penicillin and 2-oxoglutarate conversion by the R258Q mutant. The level of conversion of penicillin N (28%) and penicillin G (Ͻ Ͻ 5%) (relative to wild-type enzyme) was much lower than the levels of 2-oxoglutarate conversion (67 and 29%, respectively), suggesting this mutation causes significant uncoupling of 2-oxoglutarate and penicillin oxidation. This contrasts with the wild-type enzyme in which penicillin oxidation is tightly coupled (with a ratio of 1.06:1 for the native enzyme using 2-oxoglutarate and penicillin N (23), and 1.07:1 for the recombinant enzyme using 2-oxoglutarate and penicillin G (9)). This probably accounts (at 3   Activities are normalized to wild-type (WT) enzyme (24 g/assay) with penicillin N (92 nmol/min/ mg). The activity of wild-type enzyme with 2-oxoglutarate and penicillin G was 46 nmol/min/mg. Wild-type enzyme showed no significant activity with any other 2-oxoacid by HPLC assays except with 2-oxoadipate (64%) and pyruvate (5%) with penicillin N as substrate.
b Activities measured with the spectrophotometric assay. Activities are calculated at the optimum 2-oxoacid concentration and compared to wild-type enzyme at 2-oxoglutarate (100 M). Standard deviations for these data are estimated to be 10 -15% with both assays. No activity with either penicillin was observed for the wild-type or R275Q mutant in the presence of oxaloacetate, ketomalonate, phenylpyruvate, mercaptopyruvate, or indole-3-pyruvate. c ND, not determined. least in part) for the poor efficiency of penicillin oxidation by the R258Q mutant. The possibility of penicillin oxidation being uncoupled from 2-oxoglutarate oxidation should be taken into consideration when comparing the apparent kinetic parameters in Tables I, II, and Table III. The R258Q mutant was able to utilize several other 2-oxoacids as co-substrates when penicillin N was a substrate, most notably 2-oxo-3-methylbutanoate, 2-oxo-4-methylpentanoate, and 2-oxo-5-thiahexanoic acid. The alternative 2-oxoacids were also utilized when penicillin G was the prime substrate although the activity was somewhat lower. As with the reaction with penicillin N, the two best co-substrates were 2-oxo-3methylbutanoate and 2-oxo-4-methylpentanoate. The lower activity for penicillin G compared with penicillin N is probably due to poorer coupling of penicillin oxidation to co-substrate conversion for the former. The increased tolerance for alternative co-substrates by the single R258Q mutation to the wildtype enzyme is noteworthy. 1 H NMR and ESI MS analyses were used to characterize products of the incubation of 2-oxo-4-methylpentanoate and 2-oxo-3-methylbutanoate with the R258Q mutant in the presence and absence of penicillin G. No evidence for their modification, other than oxidative decarboxylation to give 3-methylbutanoate and 2-methylpropanoate, respectively, was obtained. This implies that 2-oxo-4-methylpentanoate and 2-oxo-3-methylbutanoate fulfill the same role as 2-oxoglutarate in the wild-type enzyme.
Apparent steady-state kinetic parameters for the R258Q mutant with the "unnatural" 2-oxoacids were determined at a single concentration of the second substrate. Analysis of the R258Q mutant showed that 2-oxo-4-methylpentanoate and 2-oxo-3-methylbutanoate are utilized with a efficiency similar (as judged by K cat /K m ) to that of 2-oxoglutarate with the wildtype enzyme (Tables II and III). Detailed kinetic analyses have revealed that substrate inhibition of the wild-type enzyme occurred at high concentrations of 2-oxoglutarate. 3 A similar although less pronounced phenomenon was observed for the R258Q mutant with 2-oxo-4-methylpentanoate and 2-oxo-3methylbutanoate as co-substrates with penicillin G as substrate. Due to the low levels of activity of the R258Q mutant with 2-oxoglutarate in the presence of penicillin G it was not possible to obtain accurate values for the kinetic parameters, but the apparent K m for 2-oxoglutarate was increased by Ͼ1000-fold compared with the wild-type enzyme. 3 Crystallographic Analysis of the R258Q Mutant-Crystal structures of the R258Q mutant complexes with iron(II) and 2-oxo-4-methylpentanoate and iron(II) and 2-oxo-3-methylbutanoate were determined to high resolution (1.5 and 1.6 Å, respectively) ( Table IV). The structures had similar overall structures to the iron(II)-2-oxoglutarate complex of the wildtype enzyme (root mean squared deviation C␣ differences ϭ 0.267 and 0.282 Å, respectively) (8). Both structures show that the 2-oxoacid moiety is ligated to the iron in the active site in a bidentate manner similar to that observed for 2-oxoglutarate in the wild-type enzyme (1,8). The protein ligands are provided with the side chains of His-183, Asp-185 and His-243, and, as before, the sixth ligand to the iron is water. This water is proposed to be displaced by binding of the "prime" substrate, allowing oxygen binding and initiation of the catalytic cycle (1,24). Most of the proteins side chains that bind to 2-oxoglutarate in the wild-type enzyme occupy similar positions to these structures. However, the side chain of glutamine 258 is now disordered. Binding of the side chains of 2-oxo-4-methylpentanoate and 2-oxo-3-methylbutanoate is apparently via hydrophobic interactions involving those residues which normally "bind" to the C-3 and C-4 methylene groups of 2-oxoglutarate (i.e. Met-180, Ile-192, Leu-204, Phe-225, Val-245, and Val-262) (Fig. 1). The result of this interaction is that the co-substrate side chains are slightly displaced from the positions occupied by the methylene groups of 2-oxoglutarate. DISCUSSION The R258Q mutant was chosen for analysis as it was hoped that the glutamine side chain would allow for an active site with an intact conformation, while allowing other 2-oxoacids with hydrophobic side chains to bind. The utilization of alternative co-substrates has been previously investigated for several oxygenases related to DAOCS. 2-Oxoadipate, which possesses an additional methylene group compared with 2-oxoglutarate, has been shown to be a co-substrate for clavaminic acid synthase (25), taurine dioxygenase (26), DAOC/ DAC synthase (27), and DAOCS from N. lactamdurans (28). ␥-Butyrobetaine hydroxylase is also able to utilize both 2-oxoadipate and oxaloacetate, the latter of which is one methylene group shorter than 2-oxoglutarate, albeit with greatly increased K m values (29). In contrast, 2,4-dichlorophenoxyacetate dioxygenase is reported to be able to utilize a number of alternative 2-oxoacids, albeit relatively inefficiently (30). Other than IPNS and ACCO, the only known non-heme-dependent oxygenase that does not utilize 2-oxoglutarate as its preferred "co-substrate" is 4-hydroxyphenylpyruvate dioxygenase. However, the structure of 4-hydroxyphenylpyruvate dioxygenase places it in the family of catechol dioxygenases, (31) and it carries out a four-electron oxidation and decarboxylation of the same 2-oxoacid, 4-hydroxyphenylpyruvate. It is related to the 2-oxoglutarate-dependent oxygenases by mechanism and can utilize 3-thienylpyruvate (32), ␣-ketoisocaproate (2-oxo-4methylpentanoate) (33), and 2-oxo-5-thiahexanoic acid (34) as (co)substrates. In contrast, ␣-ketoisovalerate (2-oxo-3-methylbutanoate) and ␣-keto-␤-methyl-n-valerate (2-oxo-3-methylpentanoate), derived from valine and isoleucine, respectively, were not converted to their corresponding products (32).
The results show that the side chain of arginine 258 is a major determinant of the 2-oxoacid co-substrate selectivity of DAOCS and, by implication, related oxygenases. The decrease in activity of the R258Q DAOCS mutant observed with 2-oxo- glutarate is probably due to the loss of a favorable ionic interaction between the 5-carboxyl group of 2-oxoglutarate and the guanidino group of arginine 258. Thus, it appears that 2-oxo-4-methylpentanoate and 2-oxo-3-methylbutanoate are utilized as efficient co-substrates by the R258Q mutant because they possesses hydrophobic side chains similar to that of 2-oxoglutarate but do not have the unfavorable polar interactions. As proposed for prolyl hydroxylase (35), the structural and biochemical results imply that DAOCS has three key subsites for 2-oxoglutarate binding, viz. the iron-ligation site, the RXS motif (Arg-258 and Ser-260) binding to the 5-carboxylate, and a hydrophobic region binding to the C3-and C4-methylene groups of 2-oxoglutarate. These results also demonstrate that the catalytic effectiveness of the 2-oxoacid co-substrate is dependent on which prime substrate is utilized (i.e. penicillin N versus penicillin G). A factor in determining catalytic efficiency is the level of coupling of 2-oxoacid oxidation to penicillin oxidation. Several recent studies on related enzymes have suggested that dioxygen binding to the iron-2-oxoglutarate enzyme complex is promoted by the presence of the prime substrate in the active site (10,24,36). Oxidative decarboxylation of the 2-oxoacid co-substrate is thought to generate a reactive ferryl species, which is responsible for oxidation of the prime substrate (37). Crystallographic evidence for the presence of this intermediate in the reaction of the iron(II)-dependent oxygenase, isopenicillin N synthase has been recently obtained (38).
The kinetic results further support the idea that cooperative binding of the 2-oxoacid and penicillin sites is required for efficient catalysis. The relatively large amount of uncoupled conversion of 2-oxoglutarate in the presence of penicillin G for the R258Q DAOCS mutant may be a reflection of an "editing" mechanism used by the enzyme to select for the correct or "correctly bound" prime substrate (1).
Studies on human prolyl-4-hydroxylase (39) have shown that mutation of His-501 lead to increases in the level of uncoupled conversion of 2-oxoglutarate by about 12-fold. This appears to be due to a relative reduction in the rate of prolyl-residue hydroxylation compared with that of 2-oxoglutarate conversion. The authors proposed that His-501 directed binding of the 2-oxoacid moiety of 2-oxoglutarate and accelerated the breakdown of the ferryl intermediate. In the light of the results presented in this paper it may be that His-501 is ligating to the 5-carboxylate of the 2-oxoglutarate co-substrate in addition to ligation by Lys-493, and that mutation led to a decrease in coupling of 2-oxoglutarate and prime substrate oxidation.
The metal-binding sites of non-heme oxygenases containing iron are closely related to those of hydrolytic enzymes containing either zinc or iron (40). Before the concentration of dioxygen reached present day levels it is possible that iron-dependent enzymes were involved in many non-redox reactions, e.g. hydrolytic reactions, reflecting the bioavailability of this metal (41). Upon increases in dioxygen levels it is possible that some of these enzymes evolved to use zinc, thereby avoiding deleterious oxidative damage, while others evolved into oxygenases/ oxidases. The increases in dioxygen levels probably also caused proliferation of the Kreb's cycle of which 2-oxoglutarate is an intermediate. The parallel increases in dioxygen and 2-oxoglutarate concentrations may have also influenced the evolution of selectivity for 2-oxoglutarate over other less abundant or ubiquitous 2-oxoacids such as 2-oxoadipate. 2-Oxoacids are common metabolites and readily form bidentate ligands with iron. The question therefore arises how and why the non-heme iron(II) and 2-oxoglutarate-dependent oxygenases appear to have evolved a (relatively strict) selectivity for 2-oxoglutarate. One reason may be that 2-oxoglutarate is a common primary metabolite, but other 2-oxoacids, e.g. pyruvate, are also ubiquitous. The proposed binding mode of 2-oxoglutarate at the active site of DAOCS rationalizes the utilization of 2-oxoglutarate and its close analogue 2-oxoadipate, and the deselection of 2-oxoacids with hydrophobic side chains. The results presented here and elsewhere (39) also demonstrate that incorrect binding of 2-oxoacids can increase the extent of prime substrate uncoupled oxidation. Furthermore, it seems that cooperativity between the 2-oxoacid and the prime substrate-binding sites is required for efficient catalysis. The exact mechanism for this cooperativity in the catalytic process during and subsequent to dioxygen binding requires further investigation. The evolution of the use of 2-oxoglutarate as a cosubstrate may reflect a need for precision (including ordered binding/release) during part of the catalytic process. This may be best achieved using a relatively rigid electrostatic interac-tion such as that formed between Arg-258 and the 5-carboxylate rather than a hydrophobic interaction or other electrostatic/hydrogen bonding interactions. Thus, the selection of 2-oxoglutarate over pyruvate may also reflect mechanistic considerations.
We have termed the restoration of activity by the use of a modified 2-oxoacid "chemical co-substrate rescue." The chemical rescue of mutant enzymes has been extensively investigated, although studies have focused on restoring the role of a mutated amino acid residue, e.g. by using organic amines (42,43), ions (44,45), or a modified residue involved in metal binding (46). The rescue of activity using modified (co)-substrates has been relatively underexplored, although ATP derivatives have been elegantly used to investigate the function of kinases (47). We have recently demonstrated chemical cosubstrate rescue of several clinically observed mutants of phytanoyl-CoA 2-hydroxylase using alternative 2-oxoacids. 2 The results in this paper give the first indications of the molecular and structural basis for this process, and suggest that many non-heme-dependent oxygenases control 2-oxoacid cosubstrate selectivity in a similar manner.