Terminally Truncated Isopenicillin N Synthase Generates a Dithioester Product: Evidence for a Thioaldehyde Intermediate during Catalysis and a New Mode of Reaction for Non‐Heme Iron Oxidases

Abstract Isopenicillin N synthase (IPNS) catalyses the four‐electron oxidation of a tripeptide, l‐δ‐(α‐aminoadipoyl)‐l‐cysteinyl‐d‐valine (ACV), to give isopenicillin N (IPN), the first‐formed β‐lactam in penicillin and cephalosporin biosynthesis. IPNS catalysis is dependent upon an iron(II) cofactor and oxygen as a co‐substrate. In the absence of substrate, the carbonyl oxygen of the side‐chain amide of the penultimate residue, Gln330, co‐ordinates to the active‐site metal iron. Substrate binding ablates the interaction between Gln330 and the metal, triggering rearrangement of seven C‐terminal residues, which move to take up a conformation that extends the final α‐helix and encloses ACV in the active site. Mutagenesis studies are reported, which probe the role of the C‐terminal and other aspects of the substrate binding pocket in IPNS. The hydrophobic nature of amino acid side‐chains around the ACV binding pocket is important in catalysis. Deletion of seven C‐terminal residues exposes the active site and leads to formation of a new type of thiol oxidation product. The isolated product is shown by LC‐MS and NMR analyses to be the ene‐thiol tautomer of a dithioester, made up from two molecules of ACV linked between the thiol sulfur of one tripeptide and the oxidised cysteinyl β‐carbon of the other. A mechanism for its formation is proposed, supported by an X‐ray crystal structure, which shows the substrate ACV bound at the active site, its cysteinyl β‐carbon exposed to attack by a second molecule of substrate, adjacent. Formation of this product constitutes a new mode of reaction for IPNS and non‐heme iron oxidases in general.

These and other studies, including ar ange of computational and spectroscopic investigations, [20][21][22][23][24] have led to ad etailed chemicalu nderstanding of IPNS catalysis, and of the roles played by some key active site residues beyond the iron-binding residues. [25] The side-chain amide carbonyl oxygen of Gln330,t he penultimate residue of the enzyme, ligates to the active site metal only in the absence of substrate, as observed in the structure of the IPNS:Mn II complex (Figure S1 a). [5] In the structure of the anaerobic IPNS:Fe II :ACV complex [6] and structures with substrate analogues, [13,[26][27][28][29] the side-chain of Gln330 is displaced from the metal to enable ligationo ft he substrate sulfur.M ovement of Gln330 enables formation of as alt bridge between thec arboxylate of Thr331 and Lys98, with the six Cterminalr esidues (Asn326-Thr331) taking up ap ositiont hat extends the final helix (a-10), thuse nclosing the substrate in the active site (Figure 1b and Figure S1). Gln330 is highly conserveda cross IPNS proteins from different species, but it is not essential for catalysis. [30] Changing Gln330 to Ala or Leu, or deleting af ew amino acids (two or six) from the C-terminal, leads to significant reductions in specifica ctivity (k cat ca. 10 %o f wildtype), butt he enzyme remains catalytically viable;d eleting eight residues ablates activity. [30] Reports concerning 2-oxoglutarate oxygenases, which are from the same structural subfamily as IPNS, have observed that their C-terminal region can exert important control over reactivity.F or instance, the C-terminal residues of gibberellin 20-oxidases are involved in regulation of specificity, [31] deletingt he last five residues of deacetoxycephalosporin synthase (DAOCS) yields an enzyme that converts the 2-oxoglutarate co-substratet os uccinate at the same rate as the wildtype butw hich does not couple this to oxidation of the primary substrate (penicillin N), [32] and the Cterminal region is important in catalysis by the hypoxia-sensing enzyme prolyl hydroxylase domain 2( PHD2). [33] To better characterise the hydrophobic binding pocket aroundt he ACV valine residue (Figure 1a)a nd to investigate the role of the dynamic C-terminus in protecting the IPNS active site after substrate binding, we constructed ar ange of IPNS variants, determined their catalytic competencies, and investigated the products formed when these modified proteins react with ACV.T he interactionb etween substrate ACV and binding pocket wasi nvestigated by changing Leu223,L eu231, Val272 and Pro283, whereas interactions betweent he protein C-terminal and the substrate were investigated by substituting ar esiduet hat contributes indirectly to substrate binding (Lys98), and by extending (*332Q) or shortening (I325*) the terminal residues. The choice and design of mutants is discussed furtheri nt he Supporting Information. Unexpectedly,d eletion at the C-terminuse nablesf ormation of an entirely new type of enzyme-catalysed thiol oxidation product. We report the characterisation of this enzymatically unprecedentedp roduct as the ene-thiol tautomer of ad ithioester, made up from two molecules of ACV linked between the thiol sulfur of one tripeptide and the oxidised cysteinyl b-carbon of the other.

Results and Discussion
Investigation of hydrophobic substrate binding pocket IPNS variants were constructed as shown in Tables S1 and S2 (see the Supporting Information) and assayedf or activity in terms of IPN production from ACV.The results reveal that alteration of the hydrophobic bindingp ocket (Figure 1) reduces Scheme1.IPNS-catalysed oxidation of:a)its naturals ubstrate ACV 1 to the bicyclic penam IPN 2 by means of af our electron(4e) oxidationi nw hich one molecule of O 2 is consumed per molecule of ACV,w hich loses four hydrogen atoms; b) the analogue ACG 3 to the shuntm etabolite 4 (a hydrated aldehyde) by means of at wo electron( 2e) oxidation; [13,19] and c) the depsipeptideanalogue ACOV 5 by means of another 4e process, in whicha ll the oxidizing potential of O 2 is directed at the cysteinyl b-carbon to form thiocarboxylic acid 6 (as characterised crystallographically). [15] the capacity of the enzymet of unction efficiently ( Figure 2), with some apparent variations for specific amino acids. For a more detailedd iscussiono ft hese substitutions and their effects, see the Supporting Information.
Product distributions with the "tailless" I325* mutant Previous work with I325* IPNS has shown it retains the ability to convert ACV 1 to IPN 2,b ut with only 10 %o fw ildtype activity. [30] Further examination of elutionp rofiles from HPLC analyses of ACV turnover by the I325* truncated variant revealed an ew "split" peak, due to an ew product designated henceforth as compound 7.T his new product eluted from the C18 column between IPN 2 and ACV thiol 1 (FigureS2);t he production of 7 increased with time at approximately the same rate as IPN 2,a nd was not observedi nc ontrolr eactions using heat inactivated enzyme. This new product wasg enerated in sufficientq uantities to allow isolation and characterisation. The split peak observed was not resolvable under all HPLC conditions screened and was collected as as ingle fraction. Subsequenta nalysis showed the fraction to contain an equilibrating mixtureo ft hio-keto/enol tautomers (see below).
[ACV] 2 (disulfide) was observed eluting later than ACV thiol 1, IPN 2 and the new compound 7,w hereas several other new speciesw ere also observed but were not produced in sufficient quantities to enable characterisation.
Largerq uantitieso f7 were obtainedb yr epeating the incubation and employing multiple chromatographic isolations, generating sufficient quantitieso fp ooled 7 for MS and NMR analyses. As econd HPLC purification step (using ah igherc oncentration of ammonium bicarbonate, see the Experimental Section) was required to separate 7 from traces of co-eluting ACV thiol 1.T ypically ACV 1 (100 mg) yielded only very small quantities of 7 (< 1mg) due to the relatively poor turnover of ACV by I325* IPNS. Some product loss was also incurred during isolation, by virtue of repeated HPLC purification steps and the apparent instability of 7.D uring these investigations, it was noted that 7 displayed ac haracteristica bsorption band at l max = 317 nm ( Figure S3 a, Supporting Information), as confirmedduring subsequent analyses (Supporting Information).
High resolution mass spectrometric analysis (positive ion mode)g ave am olecular ion for 7 at m/z 723.2697 (i.e., [  i.e.,[ MH] + with m/z 725.2850.) The mass of 7 did not change after reactionw ithe xcess sodium borohydride, indicating that the new compound does not contain ad isulfide SÀS. Quantitative reactiono f7 with Ellman's reagent (5,5'-dithiobis-(2-nitro- Figure 1. a) Key residues in the hydrophobic binding pocket aroundt he valinyls ide chain of ACV 1 at the active site of the IPNS:Fe II :ACV complex (PDB ID 1BK0). [6] The side-chains of Leu223( at "1 o'clock" relative to ACV valine), Leu231(11 o'clock),V al272 (9 o'clock)a nd Pro283(3o'clock) are shown. b) Structure of the wholeIPNS:Fe II :ACV complex, highlightingt he core double-stranded b-helix fold (green), and the C-terminal a-helical region,r esidues 313-331 (cyan). Note the position of the six C-terminalr esidues Asn326-Thr331 at the top of this helix,e nclosing ACV in the active site.Iron is in orange, key substrate-binding residues are magenta, and ACV 1 is yellow.See also FigureS1i nt he SupportingI nformation for comparisono ft his structure with the structure of IPNS without substrateb ound.  [34] using e = 13 600 m À1 cm À1 fort he TNB À anion and quantifyinga gainst adithiothreitolstandard, indicated the presence of asingle thiol group in compound 7.
The reaction of 7 with dithiothreitol (DTT) led to the isolation of ACV thiol 1 (assigned by comparison of HPLC retention times and mass spectra with an authentic sample) and as pecies of m/z 514, consistentw ith as pecies ([MH] + )a rising from DTT and the desaturated half of 7 (i.e.,A CV-2 H, Figure S4). The absorption band at 317 nm in the originali solated species is maintained on reactionw ith DTT.I odoacetamide reacts only once with 7 to give ap roduct of mass 780 Da ([MH] + ), i.e. consistent with the addition of one equivalent of CH 2 CONH 2 in a single alkylation reaction, shifting the absorbance maximum to 297 nm. When both iodoacetamide and DTT are simultaneously added to 7,a ll speciesd escribed in this paragraph are observed ( Figures S4 and S5 andT able S3 in the Supporting Information).
These data, combined with furthera nalysis by NMR (see below),s uggested that 7 is an oxidised ACV dimer,i nw hich one ACV has been doubly oxidised at cysteine (i.e.,u ndergone af our-electron oxidation) andc ontains an a,b-desaturatedc ysteinyl residue (i.e.,i sp redominantly the ene-thiol tautomer of ad ithioester) and the second ACV is covalently linked through its thiol-derived sulfur to the cysteinyl b-carbono ft he first ACV ( Figure 3).
Analysis of the 500 MHz 1 HNMR spectrum of 7 ( Figure 4, Ta ble S4 in the Supporting Information), along with COSY and HMQC spectra (FiguresS5a nd S6 ai nt he Supporting Information), enabled identification of spin-coupled systemsc orresponding to two valinyl moieties and two aminoadipoyl components,b ut only one cysteine. The aliphatic regiono ft he 1D 13 CNMR spectrum (data not shown) supported this assignment;h owever,t he limited amount of materiala vailable meant that the signal-to-noise ratio of this spectrum was too low to reliably revealt he carbonyl resonances or other quaternary centres.F urther information was therefore soughti nt he HMBC long-range heteronuclear correlation spectrum (Figures S6 ba nd S7). Though most of the correlations observed were as expectedf or the parent ACV 1,t wo additional correlations are of particular note. The first was from the valine a-proton of the modified half of 7 (blue in Figure 3) to the carbonyl of the modified cysteine. This carbonyl resonance is seen at 166.8 ppm, upfield relative to the "intact" ACV cysteine carbonyl at 171.4 ppm;s uch an upfield shifti sc onsistent with this carbonyl being adjacent to the unsaturated ene-thiol functionality in 7.T he second noteworthy correlation is from both bprotons of the intact (i.e.,g reen) cysteine to ac arbon at 229.7 ppm, ar esonance that is consistentw ith the thiocarbonyl carbon of ad ithioester,i .e. the "keto" form of thio-enol 7. [35][36][37] This correlation is indicative of an ew -CH 2 -S-C-S linkage derived from the cysteine groupso ft wo ACV building blocks. The presence of such alinkageisconsistent with the molecular formula deduced above (C 28 H 46 N 6 O 12 S 2 )a nd the outcome of reactions with DTT and iodoacetamide, that is:a nA CV dimer containing four fewer hydrogens than two fully reduced ACV molecules, in which three hydrogens have been removed from the modified cysteinyl moiety (blue) and only one from the other cysteine (green).
The NMRd ata thus suggestt hat 7 exists in at hio-keto/enol equilibrium with the thiocarbonyl form 8 (Figure 5a), consistent with the split peak seen in the HPLC elutionp rofile. This meanst hat the a-proton of the modified cysteinei se xchanged for deuterium in the thio-ketof orm under the solvent conditions used for NMR analysis. This observation, combined with its remotenessf rom non-exchangeable hydrogens, prevents direct observation of the a-carbon of the modified cysteine residue in the HMQC, HMBC and direct 13 CNMR spectra.
Investigatingt he literature for previously reported exemplars of the core a-amido-dithioester or ene-thiol tautomers een in the new compound 7/8 (the region highlighted in red in Figure 5a)r eturnsalimited number of related structures (Figure 5b). In ab iological context, ene-thiols are intermediates in the biosynthesis of the dithiopyrrolone core of antibiotics such as holomycin, [38] and thioenols/thioaldehydes are intermediates in catalysis by cysteine-derivative decarboxylases that are in- is an oxidised dimer of two molecules of ACV 1:one ACV (green)isi ntact (but for its thiol SÀH), the other (blue) has been doubly oxidised at cysteine (from thiol to thiocarboxylic acid oxidation state) andcontains an a,b-desaturated cysteinyl residue. Notethat the E/Z stereochemistryoft he alkene in 7 (and analogous products) is not definedb yt he NMR analyses andmay equilibrate through tautomerism. Stereochemistry at all other positionsisa ssumedt ob ea si nACV 1. volved in phosphopantetheine and lantibiotic biosynthesis. [39] Amino acid and dipeptide dithioester derivatives such as 9 have been used as probes to study mechanistic aspects of the cysteinep rotease papain, [40,41] whereas arylated cyclic compounds 10 have been explored as inhibitors of protein isoprenyl transferases. [42] Interestingly,e ne-thiols have been identified as potent inhibitors of di-Zn II ion-dependent metallo-blactamases (MBLs), through hydrolytic fragmentation of rhodanine derivative 11 (ML302)t og enerate the ene-thiol 12 and compound 13 (Figure5c). [43] Thus, although there is no direct evidencef or the biological production of an oxidase-catalysed ene-thiol product, it is ac onceivable mechanism for the productiono fm etallo-b-lactamase inhibitors in order to potentiate b-lactam antibiotic activity.

Furthercharacterisation of the I325* mutant
Furthere xperiments incubating the I325* variant with ACV 1 and the di-or tripeptide analogues N-acetyl-l-Cys-l-Val 14, Nacetyl-l-Cys-Gly 15, N-acetyl-l-Ser-l-Val 18,o rl-d-(a-aminoadipoyl)-l-cysteinyl-d-allo-isoleucine 19 ( Figure S9, Ta ble S5, Supporting Information) show that in the presence of ACV,t he truncated enzyme can form ene-thiol structures analogous to compound 7 only if the added ACV analogue contains ac ysteinyl moiety.T hus, when I325* was incubated with am ixture of ACV 1 ande ither N-acetyl-l-Cys-l-Val 14 or N-acetyl-l-Cys-Gly 15,i tr eactedt og ive the analogousc ysteine-oxidised ACV/ dipeptide conjugates 16 and 17 in addition to 7,w hereas with am ixture of ACV and N-acetyl-l-Ser-l-Val 18 it did not (Figure S9, Supporting Information). None of these dipeptides reacted without ACV being present (this is perhaps unsurprising, as none are substrates for wildtype IPNS) and corresponding experiments with l-cysteine itself did not give an analogousdimerised product. However, l-d-(a-aminoadipoyl)-l-cysteinyl-dallo-isoleucine 19 reactedt og ive ac ysteine-oxidised dimer 20, analogoust o7 ( Figure S9 d). Thus, it seems that I325* binds ACV 1 in am anner similar to the wildtype enzyme, to give an intermediate that can also react with relatedt hiols from solution.
To investigate the mechanism of formation of the ene-thiol products,acrystal structure for I325* in complex with iron(II) and ACV was determined (see PDB ID 2BJS, Figure 6, Figure S10 and Ta ble S6 in the SupportingI nformation). The protein chain in the I325* crystal structure adopts the same fold as the wildtype IPNS:Fe II :ACV complex (PDB ID 1BK0), the active site iron(II) is bound by the 2-His-1-carboxylate facial triad of protein-derived ligands( the side-chains of His214, Asp216 and His270). [9,10] The "primary" substrate molecule (ACV1)i sbound in as imilarm anner as for the wildtype IPNS: Fe II :ACV complex,t hat is, is held by as alt bridge from its aminoadipoyl carboxylate to the side-chain of Arg87, ligation of the cysteinyl thiolate to iron, and hydrogen bonding between the valinyl carboxylate and several protein side-chains (as seen Figure 5. a) NMRs pectroscopyexperiments indicatethat the ene-thiol 7 (of undetermined alkene stereochemistry) exists in equilibrium with thiocarbonyl tautomer 8;b)previously reported compounds 9 and 10 that include the a-amido-thio-ene-thiol motif seen in 7,ord ithioester 8;c)rhodanine derivative ML302 11 is apotent metallo-b-lactamase (MBL) inhibitor, acting via fragmentationt o12 and 13.S ee the main text for discussion of results with compounds 9-13. Figure 6. Stereoview representation showing atom positionsa nd electron densityf or the substrate (ACV1)att he active site of the I325*IPNS:Fe II :ACV complex,w ith a2 mF 0 -DF c (s A )e lectron density map shown in blue at 1.5 s for the substrate. The catalytic metal is shownn ear the top of the figure( orange sphere), proximate to the thiolate sulfura nd valinyl isopropyl groupo fACV1.( Notethat thisisopropylg roup likely has multiple conformations, indicating greater rotational freedom than is seen with the wildtype protein, in structure 1BK0.) Electron density due to the second ACV molecule (ACV2) occupies the position that is taken up by the tail of the intact enzyme (1BK0). Additional electrond ensity adjacent to the thiol sulfur of ACV2 is thought to be due to partial occupancyb ya nA CV-methane thiol disulfide (showni nthis image overlayingt he valinyli sopropyl group of ACV2), discussedi nm ore detail in the main text.A tom colours: carbon = green;nitrogen = blue;oxygen = red;s ulfur = yellow;i ron = dark orange. with the naturals ubstrate and analogues). [6,7,13,28] Several minor changes are evident relative to the wildtype protein, including minor alterations in the positions of the aminoadipoyl and valinyl side chains of the active site-bound substrate, and as mall rotation in the side-chain of Phe211r elative to the wildtypes tructure ( Figure S10 in the Supporting Information). The position of this side-chain has previously been observed to change upon turnover of tripeptide substrates to bicyclic penam products, and it has been proposed that-like the Cterminal helix-this benzyl group functions as as hieldt ot he active site. [7,44] The idea of aP he side-chain functioning to "gate" the active site of am etalloenzyme has also been posited to operate in bacterial persulfide dioxygenases (Fe II -dependent members of the metallo-b-lactamase superfamily,w hich oxidise glutathione persulfide to sulfite and glutathione), in which the side-chain of Phe184 has been observed to occupy alternative conformations in the presence and absence of glutathione. [45] Notably,i nt he I325* structure (Figure6)t he b-carbono ft he ACV1 cysteine is clearly exposed to the externals olution due to the absence of the C-terminal helix that would otherwise cover this region.F urthermore, density consistent with a second ACV molecule (ACV2)i so bserved near the primary, iron-bound ACV (ACV1). This ACV2 sits in an area of the structure close to the region that is occupied by residues Gly329 and Gln330 in structure1 BK0 of the intact enzyme;p resumably in the wildtype protein, these residues prevent entry of "ACV2" to the active site, so the cysteine-oxidisedp roduct 7 is not seen in wildtype incubations. However,i ti sp erhaps possible that binding of "ACV2"i nt his site may in part reflect the workings of as ubstrate capture mechanism. The valinyl and cysteinyl side-chains of ACV2 appear to be well-ordered, but electron density for the aminoadipoyl residue was not resolved beyondi ts Ca atom. This observation maps nicely to the results of I325* incubations with ACV analogues 14, 15 and 18, which showedt hat the aminoadipoyl moiety can be replaced with an acetyl group, but that ac ysteinyl residue linked to a second amino acid are required for formation of cysteine-oxidised dimeric products like 7 (see above and Figure S9, Supporting Information). Extra electron density is evidenta djacent to the cysteinyls ulfur of ACV2;t he size and shapeo ft his density suggesta nother sulfur atom (S')b ound in the mode of a disulfidel inkage. Af urther very small peak of density is apparent at ap osition that likely corresponds to another atom (C') covalently bound to the additional sulfur S'. We postulate that the electron density observed around ACV2 in the second ACV "bindings ite" of I325* IPNS corresponds to partial occupancy by ACV thiol 1,a nd partial occupancy by the disulfide of ACV and methane thiol, which is an impurity present in the fermentation-derived ACV 1. [46] Amechanism for formation of the oxidised product 7 The consensus mechanism for the IPNS-mediated transformation of ACV 1 to IPN 2 (Figure 7a) [6,7,24,25] begins with displacement of Gln330 from the active site of the apo-enzyme complex 21 as first ACV and then O 2 bind to iron to form the com-plexes 22 and then 23.O xygen binds end-on in the site opposite Asp216, [6] avoiding the bridged bindingm ode seen with other non-hemeiron enzymes and thusenabling oxidase activity (the primary reactionp athway for IPNS) in preference to oxygenase activity.O xygen sitsa djacent to the cysteinyl bcarbon of ACV and can then selectively abstractt he pro-S hydrogen from this positiont of orm at hioaldehyde/Fe-peroxide intermediate 24. This step is irreversible and, in the wildtype enzyme, commits ACV to the oxidative bicyclisation pathway. [47] The thioaldehyde/Fe-peroxide can then mediate blactam closure via 25 [20,21,48] to generate am onocyclic/ferryl (i.e.,F e IV -oxo) intermediate 26.D irect evidence for af erryl species has been reported for other members of the non-heme iron oxygenase family (i.e.,e nzymes closely related to IPNS), [49][50][51] and in model complexes that mimic the actives ite of non-heme iron(II) enzymes. [52] The iron(IV)-oxo species in 26 can then effect thiazolidine closure through hydrogen atom abstraction from the valinyl b-carbon to generate radical intermediate 27,w hich drives thiazolidine closure to afford the IPN complex 28.W ith I325*, we propose that the initial steps of the reaction mechanism proceed as normal, with ACV and O 2 bindinga nd abstraction of the cysteinyl pro-S b-H leading to a thioaldehyde/Fe-peroxide intermediate 24* ( Figure 7b). This speciesm ay continue the standard reaction cycle to form IPN 2,s ignificant quantities of which are generated even with I325*. However, an alternative pathway is availablei nt he absence of the C-terminal residues, which, as noted previously, would otherwise adopt ap ositiont hat extends the final helix (a-10) relative to the Mn:IPNS structure and enclose the substrate in the active site. [6] Without these residues shielding the reaction centre, the thioaldehyde in 24*i ss usceptible to attack by at hiolate (or thiol) nucleophile (ACV2) from outside the active site, leadingt oa ne nzyme-bound hemi-thioacetal/ Fe IV -oxo intermediate 29.I ns upport of the proposed mechanism, one of us has previously reported ah igh-yielding reaction in which at hiolate nucleophile attacks am etal-bound thioaldehyde (formed by hydride abstractioni namodel rhenium complex). [53] The iron(IV)-oxo species 29 incorporates two ACV moieties covalently linked by aC ÀSb ond. Reaction of the high valent iron(IV)-oxo intermediate to remove the second cysteinyl b-hydrogen from the active-site-bound ACV1 would afford the iron-bound thio-ester species 30.( This step is very similar to the mechanism believed to operate in the reaction of IPNS with the depsipeptide substrate analogue d-(l-a-aminoadipoyl)-l-cysteine d-a-hydroxyisovaleryl ester (ACOV), which leads ultimately to at hiocarboxylic acid product.) [15] In intermediate 30,t here is no longerat hiolate tether between substrate and iron. It seems probable that the absence of this link, combined with increased motilityi nducedb yt he pendant ACV2 moiety sitting outside the active site, would facilitate departure of the whole "bis-ACV" product 8 from the active site. Tautomerisation of the dithioester 8 in the reaction buffer would give the observed product 7.
It seems likely that the thioaldehydei ni ntermediate 24* would be susceptible to reaction with other extraneous nucleophiles in addition to the thiol(ate) of as econd ACV molecule. However any such products have not been observed in signifi-cant quantities nor successfully isolated during the current study.T hus, it is possible that the observed ACV2 binding mode reflects, at least in part, am echanism of capture of ACV by IPNS.

Conclusions
Substitution of key residues around the substrate binding pocket of IPNS demonstrate that the hydrophobic nature of this bindingp ocket is vital to the successful progression of IPNS catalysis to form bicyclic products, and its steric profile is highly optimised to the catalytically productivec onformation of the native substrate ACV 1.E venm inor steric changes to these residues, such as Leu to Ile or Val, are poorly tolerated, whereas reducing hydrophobicity by introducing Thr in place of Valo rL eu is highly deleterious. The geometricc onstraints imparted by Pro283a re also critical tot he catalytic efficiency of the IPNS activesite.
Removing the seven C-terminal residues to create the tailless I325* mutantp rovides important insight into the mechanism of IPNS catalysis, particularlyt he role played by these residues in shielding the reactive species generated at the enzyme active site during ACV turnover.I th as been observed previous-ly that, in the complex of wildtype IPNS with ACV,t hese seven residues "adopt ac onformation that extends the final helix (a-10) relative to the Mn:IPNS structure and encloses the substrate in the active site." [6] Thus, the C-terminal residues shield the activesite andprotectt he reactive speciespresent in intermediates 23-27 (Figure 7a)f rom alternative reaction paths. It follows that, in am utant enzyme lacking these residues, the active site is more exposed and open. The formation of 7/8 from reaction of the taillessI 325* mutant with ACV provides direct evidence that reactive intermediates generated in this more accessible active site can be intercepted by alternative reaction partnersa nd diverted from the reactionc ourse seen with the wildtype enzyme.H owever,t he intermolecularr eaction with as econd ACV that leads to 7/8 is presumablyd isfavoured relative to intramolecular reactiont oc lose the b-lactam ring (leadingi nt ime to IPN), whichi sw hy 7 is only formed as am inor by-product alongside significant quantities of the penam 2.
It is possible that binding interactions that bring ACV2 near to the actives ite reflect am echanism by which IPNS captures ACV and delivers it to the active site. This may in part helpt o explain the high efficiency of IPNS in the cellular context where it is involved in the large scale commercial production  [25]).T his scheme shows the "ligand donor" model for peroxidecleavage (24!26), in which the peroxide first reacts with the adjacent water molecule at iron, and the resulting iron-bound hydroxide then mediates deprotonation of the cysteinyl-valine amide through aproton shuttle. Recentc omputationals tudies [20,21,48] suggest this routei sm orelikelyt han alternative "substrate donor" model,i nwhich the peroxide species directly removes the NÀHp roton of the cysteinyl-valine amide.b)Proposed mechanism for the conversion of ACV 1 to oxidised product 7 by the I325* mutant,inwhich the putativet hioaldehyde intermediate of penicillins. Furthermore,I PNS has am ode of reactivity that is to date unique, and it is likely that this oxidasee volvedf rom the much more prevalent 2-oxoglutarate (2-OG) dependent oxygenases.C omparison of IPNS and 2-OG oxygenase structures reveals that ACV occupies the region of the actives ite normally involved in 2OG binding. Thus,i ti sp ossible that the ACV2 binding site in part reflects the position originally occupied by ACV (or ar elateds ubstrate) in a2 OG dependent progenitor of IPNS.
Finally,t he unusualr eaction observed with I325* mutant IPNS provides compelling evidencef or the proposed involvement of at hioaldehyde intermediate in the reactionc ycle of wildtypeI PNS. It has long been thought that such as pecies is involved in this reaction;h owever,d irectly characterisingi th as provedd ifficult. [15][16][17][18]24] The dithioester product 7/8 reported here is most plausibly formed through interception of at hioaldehydea tt he IPNS active site by as econd ACV thiol/thiolate (Figure 6b). Observation of this shunt product is therefore strong support, albeit indirect, for the involvement of at hioaldehydeintermediate in IPNS catalysis.
Conversion of the cysteine-containing tripeptide 1 to ad ithioester product 7/8 constitutes an ew mode of reactivity for IPNS and non-heme iron(II) oxidases/enzymes in general.T his reactionr equires two electrono xidation of the cysteinyl bcarbon twice, from the CH 2 carbon of at hiol to the thiocarboxylic acid oxidation state, combined with formation of as econd CÀSb ond at this same carbon centre. IPNS has previously been shown to mediate thiocarboxylate formation at this position in two different substrate analogues, [15,58] but direct enzymatic formation of ad ithioester is unprecedentedf or this enzyme and others in its class.

Production of mutante nzymes
Substitutions were introduced into the wildtype Aspergillus nidulans IPNS gene by the unique site elimination method (GE Healthcare, UK), and were confirmed by DNA sequencing. Primers for these mutations are given in Ta ble S1 (Supporting Information). Wildtype A. nidulans IPNS was produced as described previously. [5] In brief, about 50 mg of > 90 %p ure protein (by SDS-PAGE), was produced by anion exchange chromatography.F or crystallisation, further purification was performed by gel filtration on Sepharose S75 and ion exchange chromatography over MonoQ resin. [5] This yielded protein of > 95 %p urity by SDS-PAGE analysis. Electrospray ionisation mass spectrometry was used to confirm the presence of the predicted mass changes (Table S2, Supporting Information). ACV 1 was isolated and purified from the crude extract of af ermentation broth of Cephalosporin acremonium strain Ta keda N2, which is mutated in IPNS. [59][60][61] This strain is discussed in more detail in the Supporting Information. Incubation of the crude material with at wo molar excess of tris(carboxyethyl)phosphine at room temperature [62] was followed by HPLC on aH ypersil 250 10 mm C 18 column using ag radient of methanol in aqueous ammonium bicarbonate (10 mm)a nd UV detector.F ractions containing ACV 1 were freeze-dried to give aw hite solid, which was stored at À80 8Cuntil required.

Enzyme assays and isolation of products
Incubations with IPNS were typically conducted in af inal volume of 1mL, which contained ascorbic acid (1 mm), tris(carboxyethyl)phosphine (TCEP,1 m m), catalase (0.3 mg mL À1 ), iron(II) sulfate (10 mm), ACV 1 (40 mm), and IPNS (0.2 mm). Incubations were carried out at 28 8Cw ith stirring and aliquots removed at specified time points, mixed with an equal volume of methanol and flash-frozen in liquid nitrogen before HPLC analysis. Analytical HPLC was carried out on a2 50 4.6 mm C 18 column using ag radient of acetonitrile in aqueous potassium phosphate buffer (25 mm, pH 6.8) at af low rate of 1mLmin À1 . [30] Product quantities were determined by integration of peak areas in the chromatogram of the elution as measured through an absorbance detector at 214 nm. Identification of ACV 1 and IPN 2 was by comparison with authentic standards. Scaled-up incubations with the mutant contained ascorbic acid (1 mm), TCEP (1 mm), catalase (0.3 mg mL À1 ), ACV 1 (5 mm), iron(II) sulfate (100 mm), IPNS I325* mutant (10 mm), and ammonium bicarbonate (25 mm)i nafinal volume of 1mL. After incubation at 28 8C for 40 min, second aliquots of iron(II) sulfate and TCEP equal to the amounts of each already present were added and incubation was continued for af urther 40 min. Products were isolated by subjecting the whole assay mixture to HPLC separation, using ag radient of methanol in 25 mm aqueous ammonium bicarbonate on a250 10 mm C 18 Hypersil column. Fractions containing IPN 2 and the new product 7 were collected and freeze dried. The resultant white solids were stored at À80 8Cp rior to further analysis. If further purification was required, as econd HPLC separation was performed using the same gradient of methanol in 50 mm ammonium bicarbonate.
For NMR analyses, the new product 7 purified from multiple incubations was pooled. NMR spectra were acquired using aB ruker AVANCE DRX500 spectrometer equipped with a5mm 1 H{ 13 C, BB} TBI probe for 1 Ha nd 1 H-13 C2 Dd ata or on aB ruker AMX500 spectrometer equipped with a5mm BBO broadband observe probe for 1D 13 Cd ata. Samples were prepared in 3mmt ubes in D 2 O (140 mL) with 1 Ha nd 2D spectra collected at 283 Ks ot hat no 1 H resonances were masked by the solvent peak. 2D gradient, selected COSY,H MQC and HMBC spectra were collected with 2Kt 2 data points over 256 t 1 increments using 8, 128 and 512 transients per increment, respectively (total experiment times of 1, 12 and 52 h, respectively). The 1D 13 Cs pectrum was collected for 16 000 transients over a14hperiod.

Crystallography
Crystals of the IPNS mutant I325* were grown anaerobically,t ransferred into cryoprotectant solution and flash-frozen as previously described. [63,64] Data were collected at 100 Ku sing synchrotron radiation of wavelength 0.93487 at beamline ID14-3 of the European Synchrotron Radiation Facility (ESRF), Grenoble, France equipped with a1 65 mm MAR Research CCD detector.
Data were indexed and integrated with MOSFLM [65] and scaled using SCALA from the CCP4 suite of programs. [66] An initial structure was generated by rigid body refinement of protein atoms from the IPNS:Fe II :ACV model (PDB 1BK0) [6] against the new data. Further refinement was carried out using REFMAC [67] and Coot for model building. [68] The active site iron atom, substrate and water molecules were added in the course of refinement. Crystallographic coordinates and structure factors for the I325* IPNS mutant have been deposited in the Protein Data Bank (PDB) under the accession number 2BJS.