What unanticipated intermediates occur during isopenicillin N synthase catalysis?

The thioaldehyde sulfur unexpectedly leaves the iron

Isopenicillin N synthase (IPNS) is an Fe(II)-dependent oxidase that converts the tripeptide ACV into the bicyclic penicillin ring system, a four-electron oxidation that builds a β-lactam fused to a thiazolidine ring. In the starting anaerobic IPNS:Fe(II):ACV complex, ACV binds the iron through its cysteinyl sulfur. The paper asked what the transient precursors of the two ring-closing steps actually look like, because those structures had remained elusive.

The study exposed anaerobic IPNS:Fe(II):ACV microcrystals to O2 and followed the reaction by time-resolved serial femtosecond crystallography (tr-SFX) at X-ray free-electron lasers, paired with time-resolved X-ray emission spectroscopy (tr-XES). At the 3, 4 and 6 s timepoints, the refined models show the ACV sulfur moving from about 2.3 Å in the anaerobic IPNS:Fe(II):ACdV structure to about 4.2 Å from the iron. The maps fit an sp3 cysteinyl α-carbon, which the authors read as a thioaldehyde (TAL) that does not coordinate the metal rather than an enethiol(ate) intermediate.

The same timepoints show mixed populations of iron-bound unmodified ACV and iron-released TAL, with the TAL share rising from 65:35 ACV:TAL at 3 s to 50:50 at 4 s and 30:50 at 6 s. tr-XES at these times is consistent with reduction back to Fe(II). The sulfur release is therefore a measured structural feature of the reaction, not an inferred one.

Restored Fe–S coordination marks β-lactam closure

The 8 s and 10 s isomorphous difference maps show the Fe–S bond returning at roughly 2.4 Å. The authors modelled a monocyclic β-lactam intermediate (MCI) at 75% occupancy in both datasets, with residual density consistent with a partly occupied TAL. They compare this MCI with a structure obtained from the substrate analogue L-δ-(α-aminoadipoyl)-L-cysteinyl-D-(S-methyl)-cysteine, which yields a stable MCI after O2 exposure, and report that the two structures closely correlate.

Deuterium-labelled ACV (ACdV) was what made this stage observable. A 10 s PAL-XFEL structure showed partial isopenicillin N (IPN) formation alongside the MCI, and about 24 h of O2 exposure gave an IPNS:Fe:IPN complex with apparent full occupancy and clearly tetrahedral product geometry. Iron occupancy fell to roughly 50% after 24 h while the IPN product stayed bound, and diffraction quality and lattice order were maintained.

The authors describe the closure as a stereoelectronically favoured 4-exo-trig cyclization of the TAL that restores Fe–S coordination. In crystallo they saw no other products, which contrasts with solution studies where about 5% of a shunt product, a hydrated aldehyde, appeared with ACdV; the paper notes that low levels in crystals may fall below detection.

A three-water chain decides whether the lactam closes

A chain of three water molecules, Fe–W1–W2–W3, extends from the iron-bound water W1 away from the active site and is held by the backbone carbonyls of T221 and L250 and the side chain of N252. Normalized B-factor analysis of the SFX structures shows increased mobility of W2 in the 8 s and 10 s datasets, which the authors link to solvent rearrangement during catalysis.

Mutagenesis at N252 tested that link. N252A, N252Q and N252E all produced IPN as the sole observed product in solution, though the Q and E variants had substantially impaired catalytic efficiency; N252Q and N252E also displaced W2 through added steric bulk. The structurally conservative N252D variant behaved differently: in solution it showed a marked drop in catalytic efficiency and produced no observed IPN, instead giving a hydrated formyl-glycine product as its sole characterized product. All N252 variants still formed IPN in crystallo.

The authors use this split to argue that the water chain stabilizes intermediates, particularly around MCI formation and reaction before IPN. The contrast between crystallo and solution outcomes for N252D is a caution about how far any single environment reports on the other.

Removing the iron-bound water stalls catalysis at the thioaldehyde

The H270E variant replaces the iron-coordinating histidine with glutamate. Its Fe(II):ACV structure shows four-coordinate iron ligated by His214, Asp216, Glu270 and the ACV thiolate, with W1 pushed from about 2.1 Å in wild-type IPNS to about 3.8 Å, stabilized instead by Asp216 and Glu270 side-chain oxygens.

Adding O2 to this variant in solution produced a transient blue species (λmax 611 nm, half-life of several minutes). HPLC–MS and NMR identified S-alkylated enethiol E- and Z-diastereomers, matched against synthetic standards, as products of a TAL that had been released and then tautomerized or dimerized. In crystals exposed to O2, ACV sulfur left the iron: about 25% of ACV was rearranged after 30 min and about 75% by 3 h, while iron occupancy fell from 75% to 15%. No clear density for O2 or O2-derived species was seen.

The authors interpret this as support for a pathway in which the distal peroxide oxygen deprotonates Fe-bound W1, and the resulting hydroxide relays the valine amide NH proton to enable MCI formation. Without correctly positioned W1, the TAL is released instead of cyclizing. The paper states that peroxide deprotonation and ferryl formation were not directly visualized in the XFEL datasets, probably because they are transient or partly occupied, so this step rests on the H270E accumulation plus computation.

An Fe(IV) route gains indirect support from computation

Molecular dynamics of the Fe(III)–superoxo complex gave a stable trajectory with an average root mean square deviation of 1.78 Å. A representative structure was refined by quantum mechanics/molecular mechanics, yielding spin densities of Fe 4.03 and O2 −0.51 and an Fe–S distance of 2.42 Å, consistent with an Fe(III)–superoxo species and with the experimental Fe–S length.

Cysteinyl hydrogen abstraction from that state produced an IPNS:Fe(II):HO2:TAL intermediate in which the Fe–S bond lengthens to 3.12 Å and a TAL C=S double bond appears, matching the tr-SFX observation of sulfur release. The authors also report that the modelled route converges on an MCI linked through the cysteinyl sulfur to an Fe(IV)=O(OH2) species, which then abstracts the valine β-CH to close the thiazolidine ring.

The paper is explicit that the overall sequence, Fe(III)-superoxo:ACV, Fe(II)-hydroperoxide:TAL, Fe(IV)=O:MCI and Fe(II):IPN, is consistent with earlier solution-phase Mössbauer work that independently identified Fe(III)-superoxo and Fe(IV)=O intermediates. The new contribution is structural placement of the TAL and MCI stages, not a direct view of every step.

Helices away from the active site move during turnover

B-factor analysis of backbone atoms, normalized by Z-score so datasets could be compared, flagged the α3 and α10 helices as more mobile. α3 B-factors rose progressively to 10 s and returned to starting levels in the IPN complex, and two distinct α3 conformations were modelled at 6, 8 and 10 s, suggesting conformational heterogeneity. α10 stayed elevated in all O2-exposed structures, which the authors connect to the movement needed to release IPN, with lattice packing preventing it.

The authors note that the extent of conformational change in solution is likely more extensive than in crystals because of lattice constraints. That caveat matters for reading the B-factor trends as a picture of solution dynamics rather than a direct measurement of them.

Water positioning, not the ring chemistry alone, sets the outcome

The paper's practical weight sits in the comparison between variants that keep the water chain and the one that does not. N252A, N252Q and N252E still made IPN in solution, with the latter two much less efficient, and N252D made a hydrated aldehyde instead. H270E, which loses Fe-bound W1, accumulated released TAL products. Both point to the same priority: the position of active-site water, and the second-shell residue that holds it, gates whether the MCI closes to penicillin or the TAL escapes.

For readers weighing this against earlier mechanistic proposals, the decisive new observation is the non-coordinating TAL at 3–6 s together with the restored Fe–S bond at 8–10 s. Those two structural states, plus the variant behaviour, are what the Fe(IV)-mediated, water-dependent route rests on. The steps between them, peroxide cleavage and ferryl formation, remain inferred rather than visualized, so the mechanism should be treated as strongly supported at the TAL and MCI stages and indirect at the ferryl stage.

Sources used in this answer