Bioorganometallic mechanism of action, and inhibition, of IspH

Abstract

We have investigated the mechanism of action of Aquifex aeolicus IspH [E-4-hydroxy-3-methyl-but-2-enyl diphosphate (HMBPP) reductase], together with its inhibition, using a combination of site-directed mutagenesis (K _M,V _max), EPR and ¹H, ²H, ¹³C, ³¹P, and ⁵⁷Fe-electron-nuclear double resonance (ENDOR) spectroscopy. On addition of HMBPP to an (unreactive) E126A IspH mutant, a reaction intermediate forms that has a very similar EPR spectrum to those seen previously with the HMBPP “parent” molecules, ethylene and allyl alcohol, bound to a nitrogenase FeMo cofactor. The EPR spectrum is broadened on ⁵⁷Fe labeling and there is no evidence for the formation of allyl radicals. When combined with ENDOR spectroscopy, the results indicate formation of an organometallic species with HMBPP, a π/σ “metallacycle” or η ²-alkenyl complex. The complex is poised to interact with H⁺ from E126 (and H124) in reduced wt IspH, resulting in loss of water and formation of an η ¹-allyl complex. After reduction, this forms an η ³-allyl π-complex (i.e. containing an allyl anion) that on protonation (at C2 or C4) results in product formation. We find that alkyne diphosphates (such as propargyl diphosphate) are potent IspH inhibitors and likewise form metallacycle complexes, as evidenced by ¹H, ²H, and ¹³C ENDOR, where hyperfine couplings of approximately 6 MHz for ¹³C and 10 MHz for ¹H, are observed. Overall, the results are of broad general interest because they provide new insights into IspH catalysis and inhibition, involving organometallic species, and may be applicable to other Fe₄S₄-containing proteins, such as IspG.

Enzymes that catalyze the formation of isoprenoids are of interest as drug targets. There are two main pathways involved in the early steps in isoprenoid biosynthesis: The mevalonate pathway found in animals and in pathogens such as Staphylococcus aureus, Trypanosoma cruzi, and Leishmania spp. (the causative agents of staph infections, Chagas’ disease and the leishmaniases), and the nonmevalonate or Rohmer pathway found in most pathogenic bacteria, as well as in the malaria parasite, Plasmodium falciparum (1). Both pathways lead to formation of the C₅-isoprenoids isopentenyl diphosphate (IPP, 1) and dimethylallyl diphosphate (DMAPP, 2). In the later stages of isoprenoid biosynthesis, these C₅-compounds then form the farnesyl diphosphate (FPP) and geranylgeranyl diphosphate (GGPP) used in protein prenylation, sterol, and carotenoid biosynthesis. Understanding how the enzymes catalyzing these “downstream” events function has led to a better understanding of e.g. how FPP synthase (2) and GGPP synthase function, and can be inhibited (3); the discovery that bisphosphonates have potent antiparasitic activity (4); the clinical use of amiodarone (a squalene oxidase and oxidosqualene cyclase inhibitor) against Chagas’ disease (5; 6) and leishmaniasis (7); anticancer agents that inhibit both FPPS and GGPPS (8); as well as the discovery that cholesterol lowering agents (squalene synthase inhibitors) can function as antivirulence agents, against S. aureus (9). However, there have been few compounds discovered that block the nonmevalonate pathway, fosmidomycin being the notable exception (10). In this article, we focus on the last enzyme in the nonmevalonate pathway, IspH (LytB), with the goal of obtaining a better understanding of its mechanism of action, and inhibition.

The IspH (LytB) enzyme HMBPP (E-4-hydroxy-3-methyl-but-2-enyl diphosphate) reductase (EC 1.17.1.2) catalyzes the 2H⁺/2e^- reduction of HMBPP (3) to form an approximately 5∶1 mixture of IPP and DMAPP: The enzyme is essential for survival and is not found in humans, so is an attractive target for drug development (11). The structures of IspH from Aquifex aeolicus (12) and Escherichia coli (13) have recently been reported and indicate trefoil-like protein structures with a central Fe₃S₄ cluster (14), whereas EPR (15), Mössbauer (16, 17), reconstitution and catalytic activity (15, 17) measurements have all been interpreted as indicating that an Fe₄S₄ cluster is the catalytically active species. Ligand-free IspH has an “open” structure (12), whereas IspH cocrystallized with diphosphate has a “closed” structure (13) in which a serine-X-asparagine (SXN) loop is involved in hydrogen bonding with a PPi ligand. The mechanism of action of IspH is controversial and there have been many different proposals (13, 15, 18 –21) (Fig. S1). However, none of these models has yet been supported by any spectroscopic evidence, and none have led to the development of IspH inhibitors. Here, we report spectroscopic results that indicate the involvement in catalysis of π/σ metallacycle intermediates similar to those found for ethylene and allyl alcohol when bound to a nitrogenase FeMo cofactor (22 –24). Then, based on these results, we show that alkynes can inhibit IspH, forming once again, metallacyles or π/σ complexes.

Results and Discussion

The Role of Protein Residues.

We first investigated the role of protein residues in the IspH mechanism. In previous work, we noted that in addition to E126, His42, and His124 were also totally conserved residues, were located in the active site region, and were likely essential for catalytic activity, a conclusion now supported by mutagenesis results (13). However, the exact role of these residues was unclear. We thus determined the K _M and V _max values for three mutants: H42A, H124A, and E126A. In the case of the E126A mutant, activity was so low (V _max < 0.025 μmol min^-1 mg^-1) that K _M could not be measured. But with the H124A mutant, we found that although V _max was low (0.05 μmol min^-1 mg^-1 versus 1.16 μmol min^-1 mg^-1 for the wild-type enzyme), K _M was essentially unchanged (7 μM versus 5 μM, for the wild-type enzyme). This indicates that H124 is not a major contributor to substrate binding, but is essential for catalysis, suggesting that H124 may be involved in delivering H⁺ to E126 and the bound HMBPP. In the case of H42, however, we find in the H42A mutant that there is an increase in K _M (from 7–74 μM), indicating a role in substrate binding, consistent with the crystallographic observation that H42 hydrogen bonds to a bound diphosphate ligand (13). There is, nevertheless, also a 5-fold decrease in V _max, due perhaps to the possibility that several more distal residues could also be involved in proton transfer.

EPR of IspH with Bound Ligands.

As shown by Wolff et. al. (15), as-isolated, oxidized IspH (from E. coli) has low activity and exhibits an EPR spectrum characteristic of an S = 1/2 [Fe₃S₄]⁺ cluster. On reconstitution (with DTT, Fe³⁺ and S^2-) and under reducing (excess dithionite) conditions, a broad, S = 1/2 EPR spectrum characteristic of an [Fe₄S₄]⁺ cluster is obtained (15), essentially identical to that we find for reduced A. aeolicus IspH (Fig. 1 A). On addition of HMBPP, the spectrum sharpens and several components can be seen (Fig. 1 B). This suggests that HMBPP might bind to the cluster, or that HMBPP reacts and that the IPP/DMAPP products bind to the cluster, or that both occur. The observation that the IPP product can bind is supported by the observation that the spectrum of reduced, wild-type IspH plus IPP (Fig. 1 C) is similar to that of the HMBPP spectrum. Clearly, this apparent reactivity complicates any determination of the details of HMBPP binding, so we next investigated the binding of HMBPP to the E126A mutant of IspH. In our previous work (12) we proposed that E126 was a key catalytic residue, a suggestion now confirmed by mutagenesis results (13). If E126 is inactivated by mutagenesis (to Ala), this should block catalysis, and might enable observation of an early reaction intermediate. As shown in Fig. 1 D, the spectrum of the reduced E126A mutant + HMBPP is in fact now quite different to that seen with the wild-type enzyme + HMBPP, confirming this idea. The spectrum is broadened in an ⁵⁷Fe-labeled sample (Fig. S2C), indicating an [Fe₄S₄]⁺rather than an organic radical, origin. Moreover, the g-values (g = 2.124,1.999,1.958, by simulation) are, interestingly, quite similar to those seen previously with ethylene and allyl alcohol bound to the α-70^Ala mutant of a nitrogenase FeMo cofactor protein [ethylene: g = 2.123,1.978,1.949 (24); allyl alcohol: g = 2.123,1.998,1.986 (22)] with, on average, only a |0.01| difference between the IspH and nitrogenase g-values. In nitrogenase, ethylene and allyl alcohol have been shown [via ENDOR and/or density functional theory calculations (22 –24)] to bind as π/σ “metallacycles,” as shown e.g. in Fig. 1 E. And because HMBPP is simply a substituted alkene, one possible explanation of the E126A + HMBPP spectrum is that HMBPP binds to reduced IspH in a similar manner, that is, as a π or π/σ “metallacycle” or “η ²-alkenyl” complex, similar to that shown in Fig. 1 F (which is based on the nitrogenase/allyl alcohol structure and contains Mo and “X”). That such a complex could form with HMBPP is supported by the results of ligand docking calculations using Glide (25) (Fig. 1 G and H) in which it can be seen that HMBPP can bind with its diphosphate occupying the “PPi” site seen crystallographically, at the same time that its double bond interacts with the unique fourth Fe, added computationally here as described previously (12). Based on the observation that only reduced Fe₄S₄ clusters bind alkynes (26), it seemed likely that only the reduced IspH would bind HMBPP in this manner, and to investigate this binding question in more detail, we used ENDOR to probe the HMBPP- reduced IspH interaction in more detail.

Fig. 1.

9 GHz EPR of A. aeolicus IspH, and molecular models for ligand interactions. (A) Wild-type IspH. (B) Wild-type IspH + HMBPP. (C) Wild-type IspH + IPP. (D) E126A IspH + HMBPP. Spectra were all obtained on samples reduced with 20 equivalents sodium dithionite (see SI Text for more details), and are essentially identical to those obtained in samples prepared via photoreduction (e.g. Fig. S2A and B). Microwave power was 1 mW for A–C, and 0.1 mW for D. (E) Structure of nitrogenase Fe/Mo cofactor + allyl alcohol. (F) As (E) but edited to more clearly show the allyl alcohol binding site. Mo (Cyan); “X” (Blue). (G) Glide docking result for HMBPP docking to IspH active site region. (H) As (G) but after molecular mechanics optimization.

²H, ¹³C, ³¹P, and ⁵⁷Fe ENDOR of IspH + Alkene Ligands.

A test of this π/σ metallacycle complex formation hypothesis with reduced IspH is that there should be significant hyperfine couplings (A) with the ligand if it indeed bonds to the S = 1/2, [Fe₄S₄]⁺ cluster. We thus prepared the following series of ²H and ¹³C isotopically labeled HMBPPs (4–6), and investigated their ENDOR spectra when bound to IspH. On binding 4 to E126A IspH, we clearly see (Fig. 2 A, Bottom Trace) a low frequency feature in the Mims ENDOR (27) difference spectrum attributable to the methyl deuterons, characterized by A ∼ 0.4 MHz. At higher frequency, there is a peak centered at the ³¹P Larmor frequency, but with a very small (< 0.1 MHz) hyperfine coupling, in the ENDOR spectra of both 3 and 4 with IspH (Fig. 2 A, Inset in Black). Essentially the same results are found on binding of 5 (which contains a 4-²H₁ group), inset in blue above the 5-²H₃ spectra, with A again being approximately 0.4 MHz for ²H and < 0.1 MHz for ³¹P. With [u-¹³C] HMBPP, 6, there are two major features (having A = 1.70,0.80 MHz, Fig. 2 B), which we attribute to the olefinic carbons, ¹³C2,3. Signals from ¹³C1, 4 and 5 are not observed, due we believe to very small couplings. These EPR and ²H, ¹³C ENDOR results support the idea that HMBPP bound to reduced IspH forms a metallacycle (that is, a π or η ²-alkenyl σ-like complex), just as found with ethylene and allyl alcohol binding to the nitrogenase FeMo protein cofactor (22 –24).

Fig. 2.

ENDOR spectra of IspH with HMBPP. (A) E126A IspH + 4 (Black) or 5 (Blue). Spectra are difference spectra (²H-¹H); the ³¹P region is the unsubtracted ENDOR result. (B) E126A IspH + 6 minus E126A + 3, showing solely ¹³C ENDOR. (C) Unliganded wild-type [u-⁵⁷Fe]-IspH, with natural abundance ⁵⁷Fe control subtracted. (D) [u-⁵⁷Fe]-E126A IspH + HMBPP, ⁵⁷Fe ENDOR, Davies selective pulse. (E) As (D) but nonselective pulse. (F) As (D) but natural abundance ⁵⁷Fe-IspH. (G) Difference spectrum of (E) and (F).

⁵⁷Fe ENDOR spectra of [⁵⁷Fe]-IspH ± 3 also support the idea that 3 may bind directly to the Fe₄S₄ cluster. The ⁵⁷Fe Davies ENDOR spectrum (27) (π/2_mw = 12 ns, nonselective pulse) of IspH in the absence of any added ligands is broad (Fig. 2 C) and there are three major features present. On binding 3, a sharper ENDOR spectrum is seen. Using selective (π/2_mw = 48 ns) pulses in a Davies ENDOR spectrum, features attributable to ¹H as well as ⁵⁷Fe are observed (Fig. 2 D); with nonselective pulses (π/2_mw = 12 ns) only two major Fe features are apparent (Fig. 2 E) and these features (as expected) are absent in a control ([⁵⁶Fe] IspH + 3) spectrum (Fig. 2 F). The difference spectrum (2E-2F) is shown in Fig. 2 G. No additional small ⁵⁷Fe hyperfine couplings are observed, as confirmed by Mims ENDOR. These results show that there are major changes in the electronic structure of the reduced Fe₄S₄ cluster on HMBPP addition, and strongly support the idea that HMBPP binds chemically to the cluster. This in turn supports the idea that the hyperfine couplings seen in the ²H/¹³C ENDOR results are not solely due to a dipolar or through space interactions but rather, arise (at least in part) from through bond, Fermi contact couplings. So, the results of both EPR and ENDOR spectroscopy indicate π or π/σ complex formation with reduced IspH and HMBPP, leading to the catalytic mechanism proposal described in the following.

Catalytic Mechanism of IspH.

The models shown in Fig. 1 G and H indicate that when the HMBPP diphosphate docks to the “PPi” site and the double bond forms a π/σ alkenyl complex with the (reduced) Fe₄S₄ cluster, the CO₂H group of E126 is approximately 2.3 Å from the HMBPP 4-OH group. This then suggests the catalytic mechanism shown in Fig. 3 in which: (i) After initially binding to the oxidized cluster via oxygen (Fig. 3 A), on reduction a π (or π/σ) complex forms (Fig. 3 B and C), “activating” the substrate for the subsequent deoxygenation. The structure shown in Fig. 3 B is the π-complex, but the proposed reaction can arguably be better followed via the η ²-alkenyl/metallacycle, Fig. 3 C. (ii) This reduced complex is then protonated via E126 (and H124A), resulting in formation of water and an η ¹-allyl complex, Fig. 3 D. (iii) After a second reduction, the η ³-allyl complex forms (Fig. 3 E and F), with protonation at C2 or C4 then forming the two products, DMAPP and IPP (Fig. 3 G and H). One potential drawback of this mechanism might be that the E126 carboxyl is not acidic enough to protonate the 4-OH. However, H124 is close by (approximately 5 Å from the H124 N^δ to E126 carboxyl oxygen, PDB file #3F7T), and could be involved in increasing the acidity of E126, either directly or via an intervening H₂O. The source of H⁺ in the second part of reaction, the reductive cleavage, could be via E126 (in the case of DMAPP formation), or solvent (e.g. H₃O⁺ in a “water” site), for IPP (because there are no other obvious H⁺ donors apparent).

Fig. 3.

Proposed mechanism of action of IspH. (A) Oxidized Fe₄S₄ cluster plus HMBPP (12); (B) the Fe₄S₄ cluster has been reduced by one electron; the structure is generally similar to that found for allyl alcohol bound to nitrogenase; (C) π/σ complex (shown for simplicity as the η ²-alkenyl metallacycle) is protonated by E126/H124A; (D) After protonation, the 4-OH is removed as H₂O, forming an η ¹-allyl complex; (E) the Fe₄S₄ cluster is reduced by second e^-, and subsequently forms a η ³-allyl (π) complex, (F); (G) protonation at C4 results in DMAPP product; (H) IPP is formed by protonation at C2.

A simplified version of the mechanism, together with the locations of the reactants/intermediates/products in IspH is illustrated in Fig. 4. In Fig. 4 A, we show a docking pose for HMBPP bound to reduced IspH (PDB file #3F7T; chosen because it contains a PPi ligand and adopts the “closed” conformation) in which the diphosphate hydrogen bonds to S225, N227, and H42 while the side-chain fits the pocket near the Fe₄S₄ cluster. Note of course that, based on the spectroscopic and activity results, we show an Fe₄S₄ (not an Fe₃S₄) cluster. The double bond in HMBPP forms a π (or π/σ) complex with the reduced Fe₄S₄ cluster, based on the ENDOR results. As is well known in organometallic chemistry, such complexes are typically neither completely π nor completely σ, as discussed by Pelmenschikov et al. (23), but for convenience in drawing reaction mechanisms, is shown as a σ (η ²-alkenyl) complex in Figs. 3 and 4. On protonation of the 4-OH by E126 (possibly involving H124 and H₂O), the η ¹-allyl complex and water form, Fig. 4 B. After the second reduction, the η ³-complex forms (Fig. 4 C), which is then protonated to form the two products (Fig. 4 D). Although radical transition states or reactive intermediates may also be involved, the EPR and ENDOR spectra we do observe are unlikely to arise from allyl radicals because (i) allyl radicals would have 2 or 3 partially resolved peaks around the free electron g-value; (ii) they would have very large [approximately 50 MHz (28)] ¹³C hyperfine couplings; (iii) they would not broaden on ⁵⁷Fe-labeling; plus (iv), the E126A mutant is essentially unreactive. The model proposed is, attractive because it invokes well-established precedent for the formation of π-complexes between Fe₄S₄ clusters and unsaturated species (alkynes, cyclopentadienides) (26, 29); such complexes are seen with the “parents” ethylene and allyl alcohol bound to the nitrogenase FeMo cofactor; we observe hyperfine couplings with bound substrate; docking puts the double bond next to the unique Fe, and the 4-OH close to the putative H⁺ donor, E126; and there is no evidence for any allyl radical intermediates.

Fig. 4.

Protein-based structural models for IspH catalysis. (A) HMBPP docks to the reduced active site with diphosphate binding to S225, N227, and H42, double bond forms π/σ complex with Fe₄S₄ cluster; (B) HMBPP 4-OH is protonated by E126, loss of H₂O results in η ¹-allyl complex; (C) cluster is rereduced by one electron, forming an η ³-allyl (π) complex; (D) η ³-allyl complex is protonated at C4 forming DMAPP and oxidized cluster. Protonating the alternate η ³-allyl complex (or η ¹-allyl species) at C2 results in IPP product formation (as in Fig. 3). Numbering is for the A. aeolicus protein.

The Metallacycle Model Leads to Novel Inhibitors.

The binding of acetylene to simpler Fe₄S₄ clusters, forming π complexes, Fig. 5 A, has been observed in model systems such as [Fe₄S₄(SPh)₄]^2-/3- (26, 30), and diphenylacetylene (7) binds to [Fe₄S₄(SPh)₄]^2- where it can be cis-hydrogenated by NaBH₄ (31). This leads to the idea that acetylenes might bind to, and thus inhibit, IspH. We thus next investigated the inhibition of IspH by propargyl alcohol, 8: Inhibition was very low, K _i > 10 mM. On the other hand, propargyl diphosphate (9) was a more potent IspH inhibitor (Fig. 5 B) with a K _i = 0.97 μM. This indicates that acetylenic diphosphates bind to IspH, most likely with their diphosphates occupying the “PPi” site, while their alkyne groups bind to the fourth Fe, forming a π/σ metallacycle. A docking pose for 9 bound to IspH is illustrated in Fig. 5 C and clearly indicates that the alkyne is in close apposition with the fourth Fe, forming a metallacycle.

Fig. 5.

Inhibition of IspH by alkynes. (A) Proposed binding mode of acetylene to an Fe₄S₄ cluster (26, 30). (B) IspH inhibition by 9; IC₅₀ = 6.7 μM; K _i = 0.97 μM. (C) Glide docking pose for 9 bound to IspH. The E. coli IspH structure (PDB 3F7T) and numbering are shown. (D–F) EPR spectra of three alkynes interacting with IspH: (D) 8; (E) 9; (F) 10. Microwave power was 0.05 mW. All spectra show major g-value changes from those seen in the inhibitor-free protein, Fig. 1 A.

If such metallacycles form, it should be possible to investigate them spectroscopically. In particular, we expect to see major changes in the [Fe₄S₄]⁺ EPR spectra, plus, ENDOR should indicate sizeable hyperfine couplings. The EPR spectra of IspH (+8,9, and 10) are shown in Fig. 5 D–F and indicate large g-value changes (from Fig. 1 A) on ligand addition, suggesting an interaction with the Fe₄S₄ cluster. The Davies ¹H ENDOR spectrum (Fig. 6 A) of IspH + 8 shows an approximately 10 MHz hyperfine coupling feature, while in the Davies ENDOR spectrum of IspH + [3-²H₁]-8, the 10 MHz feature is greatly attenuated (Fig. 6 A, Inset) and is replaced at low field by a 1.5 MHz hyperfine coupling feature in the (²H - ¹H) difference Mims ENDOR spectrum (Fig. 6 B), attributable to the 3-deuteron. The large ¹H and ²H hyperfine couplings are similar to those seen in the nitrogenase system (22) and indicate formation of an organometallic species.

Fig. 6.

Alkyne inhibitor ENDOR results. (A) IspH + 8, Davies selective pulse, showing a large ¹H hyperfine coupling that is attenuated (Inset, Blue) in an 80% [3-²H₁]-8 labeled sample. (B) IspH + [3-²H₁]-8, (²H - ¹H) difference Mims ENDOR spectrum, showing a ²H ENDOR signal with 1.5 MHz coupling. (C) IspH + [u-¹³C₃]-8, difference Mims ENDOR spectrum showing solely ¹³C ENDOR. (D) IspH + [u-¹³C₃]-9 difference Mims ENDOR spectrum showing solely ¹³C ENDOR; the inset shows the ³¹P Mims ENDOR signals from IspH+ unlabeled 9. (E) [u-⁵⁷Fe]-IspH + 8, Davies selective pulse. (F) The same as (E), but with Davies nonselective pulse. (G) Natural abundance ⁵⁷Fe control, Davies nonselective pulse.

Additional evidence for formation of Fe-C bonds per se come from the results of ENDOR experiments using both [¹³C₃]-8 and [¹³C₃]-9 (Fig. 6 C and D), where we see strong ¹³C ENDOR peaks having quite large (6 MHz) hyperfine couplings. This, together with the docking result (Fig. 5 C), suggests that both 8 and 9 bind to the Fe₄S₄ cluster at the fourth Fe, forming a π/σ complex, as with acetylene binding to model Fe₄S₄ clusters (26, 30). Also present in the ENDOR spectrum of 9 bound to IspH is a feature due to ³¹P (Fig. 6 D Inset, hyperfine coupling = 0.3 MHz). We favor a sideways-on π-complex rather than the alternative end-on binding mode because (i) diphenylacetylene (7), which lacks a terminal H, binds to reduced Fe₄S₄ clusters (31) but as with acetylene (30), is still cis-reduced [by ²H (31)]; (ii) the shift in the C ≡ C vibrational Raman spectrum (26) of C₂H₂ on binding to reduced Fe₄S₄ clusters is relatively small (approximately 60 cm^-1), whereas shifts seen on acetylide formation (or in mononuclear complexes) are typically 2–3x larger (32); (iii), compound 10, which cannot bind end-on to IspH, still exhibits (Fig. 5 F) a rhombic EPR spectrum similar to that seen with the terminal alkynes, Fig. 5 D and E; and (iv), the observation of the ²H ENDOR signal from the terminal ²H in 8 is consistent with this binding mode: FeH would have a larger coupling, while SH would exchange away.

The ⁵⁷Fe ENDOR spectra of [u-⁵⁷Fe]-IspH + 8 (and a ⁵⁶Fe control) are shown is Fig. 6 E–G. The Davies selective pulse spectrum (π/2_mw = 48 ns, Fig. 6 E) has both ¹H and ⁵⁷Fe contributions. The ¹H signal is edited out by using nonselective pulses (π/2_mw = 12 ns), Fig. 6 F, and now we see two main features with hyperfine couplings of 28,35 MHz. No signal is seen with the ⁵⁶Fe control sample (Fig. 2 G) under the same conditions. Likewise, no signals are apparent in the low frequency region when using a Mims pulse sequence, ruling out the presence of very small ⁵⁷Fe hyperfine couplings. Overall, these ⁵⁷Fe ENDOR results are quite similar to those seen with HMBPP, Fig. 2, and are quite different to those seen in the absence of these ligands, supporting again a direct interaction with the cluster.

Conclusions

In summary, the mechanism we propose is unique in that it involves a novel, bioorganometallic mechanism for IspH catalysis, that is, the formation of transition state/reactive intermediates (π/σ complexes; η ³-allyl complexes) containing metal-carbon bonds, together with the involvement of at least two residues (H124, E126) that deliver H⁺ to the active site. The ability of olefins (as well as acetylenes and cyclopentadienides) to bind to [Fe₄S₄]⁺ (or [Fe₆Mo₂S₈(SPh)₉]^3- clusters) is well known in model systems, as well as in the nitrogenase system, and provides precedent for the structures proposed. Of particular importance is the observation that ethylene and allyl alcohol (HMBPP “minus” the Me and -CH₂OPP substituents), form π/σ complexes with Fe (in the nitrogenase FeMo cofactor), and similar metallacycles are likely to form (as reactive intermediates or transition states) when HMBPP binds to reduced IspH, with the 4-OH then being ideally poised to be protonated (by E126/H124). This π/σ bonding would not be present with an Fe₃S₄ cluster, which just presents 3S^2- to the olefin. When compared with previous proposals, the results we have described above differ in several important respects. First, organometallic species (i.e. containing Fe-C bonds) are involved. Second, we find no evidence for stable, carbon-based radicals. Third, there are specific roles for key catalytic residues (in particular, E126). Fourth, the mechanism requires the presence of a fourth, unique Fe. The likely involvement of an η ³ (π) complex (the allyl anion) is also attractive, because other anionic π-complexes, cyclopentadienides, are well known to form π-complexes with Fe₄S₄ clusters. These results are also of broader general interest because they have led to the discovery of the first μM inhibitor of IspH, and based again on EPR, ENDOR, and computational docking results, we propose that these types of inhibitors bind into the IspH active site with their diphosphates occupying the “PPi” site, while their alkyne groups form π/σ complexes with the unique fourth Fe. Overall, these results can be expected to lead to new types of inhibitors of IspH, as well as of other Fe₄S₄ cluster-containing proteins containing “unique” fourth Fe atoms, such as IspG, where again, organometallic species are likely to be involved in catalysis.

Methods

Samples for EPR spectroscopy were reduced either by adding Na₂S₂O₄ or by photoreduction in the presence of 5-deazaflavin. EPR spectra were collected at X-band using a Varian E-122 spectrometer together with an Air Products helium cryostat. Data acquisition parameters were typically field center = 3250 G; field sweep = 800 G; modulation = 100 kHz; modulation amplitude = 5 G; time constant = 32 ms; 60 s per scan; 8 s between each scan; and temperature = 15 K. Samples for ENDOR spectroscopy were reduced by adding Na₂S₂O₄. Pulsed ENDOR spectra were obtained on a Bruker ElexSys E-580-10 FT-EPR X-band EPR spectrometer using an ENI A 300RF amplifier and an Oxford Instruments CF935 cryostat at 15 K (8 K for unliganded wt [⁵⁷Fe]-IspH). Davies pulsed ENDOR experiments were carried out for ⁵⁷Fe samples and the IspH + 8 sample using a three pulse scheme (π _mw-T-π/2_mw-τ-π _mw-τ-echo, π _rf applied during T) (27) with the excitation field set to correspond to g ₂. ⁵⁷Fe Davies ENDOR spectra were collected with either π/2_mw = 12 ns (nonselective) or π/2_mw = 48 ns (selective) pulse excitation. The selective pulse scheme was also used for acquiring ¹H ENDOR spectra of IspH + 8. For other samples, Mims pulsed ENDOR with a three pulse scheme (π/2_mw-τ-π/2_mw-T-π/2_mw-τ-echo, π/2_mw = 16 ns, and π _RF applied during T) (27) was used, with the excitation field set to correspond to g ₂. τ-averaging (32 spectra at 8 ns step) was used to reduce the blind spots that arise from the τ-dependent oscillations.

Further details of site-directed mutagenesis, protein purification and reconstitution, enzymatic assays, computational aspects, and compound synthesis are reported in SI Text, Supplementary Methods.