Radical SAM catalysis via an organometallic intermediate with an Fe–[5′-C]-deoxyadenosyl bond

See allHide authors and affiliations

Science  13 May 2016:
Vol. 352, Issue 6287, pp. 822-825
DOI: 10.1126/science.aaf5327

Catching a radical in action

Many enzymes catalyze reactions through the production of radical intermediates. Radical SAM enzymes, the largest superfamily of enzymes in nature, do this by using an iron-sulfur cluster to cleave S-adenosylmethionine and produce a radical intermediate. Using freeze quenching, Horitani et al. were able to trap a previously unseen radical intermediate from bacterial pyruvate formate-lyase activating enzyme. Spectroscopy revealed that the intermediate consists of a short-lived covalent bond between the terminal carbon of 5′-deoxyadenosyl and the single iron atom of the iron-sulfur cluster. Not only does the observation of this radical expand our mechanistic understanding of radical SAM enzymes, but it expands the range of enzyme active sites or cofactors that function through an organometallic center.

Science, this issue p. 822


Radical S-adenosylmethionine (SAM) enzymes use a [4Fe-4S] cluster to cleave SAM to initiate diverse radical reactions. These reactions are thought to involve the 5′-deoxyadenosyl radical intermediate, which has not yet been detected. We used rapid freeze-quenching to trap a catalytically competent intermediate in the reaction catalyzed by the radical SAM enzyme pyruvate formate-lyase activating enzyme. Characterization of the intermediate by electron paramagnetic resonance and 13C, 57Fe electron nuclear double-resonance spectroscopies reveals that it contains an organometallic center in which the 5′ carbon of a SAM-derived deoxyadenosyl moiety forms a bond with the unique iron site of the [4Fe-4S] cluster. Discovery of this intermediate extends the list of enzymatic bioorganometallic centers to the radical SAM enzymes, the largest enzyme superfamily known, and reveals intriguing parallels to B12 radical enzymes.

The radical SAM superfamily is the largest known superfamily of enzymes, with more than 100,000 functional domains found throughout all kingdoms of life (1). These enzymes use a [4Fe-4S]+ cluster and S-adenosylmethionine (SAM) to catalyze a diverse array of finely tuned (2) radical reactions that are central to key pathways, such as heme and chlorophyll biosynthesis, vitamin biosynthesis, DNA repair, and metal cluster assembly (1). Spectroscopic studies of the radical SAM enzymes pyruvate formate-lyase activating enzyme (PFL-AE) and lysine 2,3-aminomutase (LAM) demonstrated that these enzymes bind SAM as a classical coordination complex in which the amino and carboxyl moieties of SAM chelate the unique iron site of a [4Fe-4S] cluster, as confirmed by x-ray structure determinations (36). One-electron reduction of SAM by the [4Fe-4S]+ cluster initiates SAM cleavage, generating the 5′-deoxyadenosyl radical intermediate (5′-dAdo•), which carries out subsequent radical chemistry. To date, neither the 5′-dAdo• radical nor any SAM- or 5′-dAdo-–derived intermediate has been detected for any radical SAM enzyme. However, the use of the enzymatically active SAM analog 3′,4′-anhydro-S-adenosylmethionine has allowed characterization of the allylically stabilized anhydroadenosyl radical, a functional analog of the 5′-dAdo• radical, generated within the active site of LAM during the LAM-catalyzed reaction (79).

Pyruvate formate-lyase activating enzyme is a radical SAM enzyme that catalyzes the formation of a catalytically essential glycyl (G•) radical on Gly734 (G734) of PFL, a central enzyme in anaerobic glucose metabolism (Fig. 1) (10). We used rapid freeze-quench (rfq) electron paramagnetic resonance (EPR) and electron nuclear double-resonance (ENDOR) spectroscopies to trap and characterize a catalytically competent intermediate species formed during the reaction catalyzed by PFL-AE. We hoped that this approach would allow us to capture the elusive 5′-dAdo• radical intermediate that is implicated as the central species in radical SAM mechanisms, but instead our work yielded a state with unanticipated properties.

Fig. 1 Activation of PFL by PFL-AE, with concomitant cleavage of SAM to methionine and 5′-deoxyadenosine.

When reduced PFL-AE is rapidly mixed with PFL and SAM, followed by rfq of the reaction at times ranging from 25 ms to 1.0 s after mixing, the characteristic [4Fe-4S]+ EPR signal of reduced PFL-AE/SAM (g values: g = 2.01, 1.88, 1.87) (11) is diminished in intensity, concomitant with the appearance of a new EPR signal, denoted Ω (g|| = 2.035, g = 2.004), whose intensity reaches a maximum at 500-ms quench times (fig. S1). At longer quench times, the Ω signal is lost in parallel with the appearance of the G• radical signal that is characteristic of activated PFL, which indicates that Ω is an intermediate in the formation of the G• radical.

To explore the reactivity of Ω, we cryo-annealed a frozen 500-ms quench sample for 1 or 3 min at progressively higher temperatures, up to 220 K. During cryo-annealing of the frozen solid, the Ω signal is lost in parallel with quantitative generation of the PFL G• g-2 radical signal, with its characteristic doublet from hyperfine coupling, A ~ 15 Gauss (G), to the α proton (Fig. 2A) (12). This annealing progression confirms that Ω is a catalytically competent reaction intermediate that can generate the G• radical product by H atom extraction from G734 of PFL and can even do so in the frozen solid. As a control, the rfq reaction was carried out using PFL Gly734→Ala734 (G734A), where the G• radical site was substituted with alanine. The rfq EPR spectrum at 12 K is identical to the one for the PFL wild type, which indicates that Ω is formed by cleavage of SAM in these reactions (Fig. 2B). However, the EPR spectrum at 40 K shows that Ω does not proceed to produce a G• radical, at a quench time ≥500 ms or during cryo-annealing.

Fig. 2 EPR spectra showing the formation of the PFL glycyl radical (G•) from Ω.

(A) (Top) EPR spectra of mixture of photoreduced PFL-AE and PFL/SAM freeze-quenched at ~77 K (500 ms), stored at 77 K, and then annealed at progressively higher temperatures for the indicated times (see supplementary materials). At 12 K, the G• radical spectrum is highly saturated and its amplitude diminished; at 40 K, the signal from the rapidly relaxing Ω is correspondingly diminished. The spectra here have had the residual intensities at both temperatures subtracted out (fig. S2), with one exception: Ω is completely lost after annealing at 220 K; the dashed curve shows the residual signal from the saturated G• radical. (Bottom) Populations of Ω and the G• radical, relative to the final (220 K) G• radical concentration taken as 100% (Sum), as derived from EPR spectra (see supplementary materials). (B) X-band EPR spectra for photoreduced PFL-AE freeze-quenched (77 K) 500 ms after mixing with PFL G734A/SAM, with spectra collected at 12 and 40 K. Conditions: microwave frequency = 9.23 GHz, microwave power = 1 mW, 100-kHz modulation amplitude = 8 G, and T, as indicated. The gain at a given T is fixed.

We conclude from these observations that (i) Ω is associated with PFL-AE and SAM or 5′-dAdo; (ii) Ω is a precursor in the formation of the G• radical of the substrate protein PFL; and (iii) in Ω, the target G734 has been positioned so precisely that H atom abstraction to generate the G• radical can take place without major conformational changes, which are quenched in the frozen solid.

The g values for Ω are not compatible with assignment to a “free” organic radical, such as the elusive 5′-dAdo• radical, which would exhibit a near-isotropic g ~ 2.0 signal. Moreover, the EPR spectrum for the primary-carbon 5′-dAdo• radical would necessarily exhibit well-resolved splittings arising from the hyperfine coupling between the odd electron in the 2p π orbital of C(5′) and the two equivalent protons of 1H2C(5′) (fig. S3), but these are absent in the spectrum of Ω. The g values are similar to those of compound ES of cytochrome c peroxidase, where they reflect a weak exchange coupling between an organic radical and an integer-spin metal-ion center (1315). In our work, the analogous situation would involve the 5′-dAdo• radical spin-coupled to an integer-spin [4Fe-4S]2+ cluster. However, for multiple reasons, this model is also unambiguously incapable of describing Ω, not least because the EPR spectrum for such a spin-coupled center would necessarily exhibit the same 1H hyperfine couplings to the 1H2C(5′) of the 5′-dAdo• radical that are required for the free radical, unchanged by the exchange interaction (1315) (fig. S3), but absent in the spectrum of Ω.

To determine whether Ω nonetheless involves the 5′-dAdo• radical in some fashion, we carried out rfq experiments, using 13C-labeled SAM and 35-GHz continuous wave (CW) and pulsed ENDOR spectroscopies to examine Ω. When Ω was formed with SAM—in which all adenosyl carbons, both of the base and ribose, are 13C ([adenosyl-13C10]-SAM)—CW ENDOR spectra collected at g = 2.004 disclosed a strong coupling to 13C (Fig. 3A, top). This signal can be simulated by a hyperfine interaction with a single 13C with a large isotropic coupling, aiso = 9.4 MHz, and an axial dipolar coupling, 2T ≳ 5:3 MHz,, depending on the extent to which the dipolar direction deviates from the g plane. A second, more weakly coupled 13C signal arising from 13C at other positions (A ~ 0.7 MHz) is revealed by the Mims pulse technique (Fig. 3A, inset).

Fig. 3 35-GHz ENDOR spectra at g for photoreduced PFL-AE freeze-quenched with PFL/SAM.

To first order, an ENDOR spectrum of a spin I = 1/2 nucleus (N) in a frozen solution comprises a superposition of signals from different orientations, each signal a doublet at frequencies ν± = |ν(N) ± A/2|, where ν(N) is the nuclear larmor frequency and A is the orientation-dependent hyperfine coupling (23). For 13C, A/2 << ν(13C), and it is convenient to plot spectra versus ν – ν(13C). For 57Fe, ν(57Fe) << A/2, and spectra are plotted versus ν. (A) 13C CW ENDOR for [adenosyl-13C10] SAM. The green dashed lines denotes the best-match simulation to the axial hyperfine tensor (see supplementary materials). Simulation parameters: aiso = 9.4 MHz, 2T = 5.3 MHz, and β = 90°. Conditions: microwave frequency = 35.39 GHz, microwave power = 1 mW, 100-kHz modulation amplitude = 1.3 G, rf sweep rate = 1 MHz/s, and T = 2 K. (Inset) Mims ENDOR spectrum. Conditions: microwave frequency = 35.20 GHz; MW pulse length, (π/2) = 50 ns; τ = 500 ns; and T = 2 K. (B) Mims ENDOR spectrum from [methyl-13C] SAM. Conditions: microwave frequency = 35.08 GHz; MW pulse length, (π/2) = 50 ns; τ = 500 ns; and T = 2 K. (C) 57Fe CW ENDOR for 57Fe-enriched Ω and photoreduced PFL-AE. (Top) CW ENDOR spectra for 57Fe-enriched (red) and natural-abundance (gray) rfq samples. (Bottom) Frequency sweep and randomly hopped stochastic CW ENDOR spectra (23) for 57Fe-enriched reduced PFL-AE. Conditions: microwave frequency = 35.45 GHz and 35.07 GHz for rfq and 57Fe-enriched reduced PFL-AE, respectively; microwave power = 1 mW; 100-kHz modulation amplitude = 1.3 G; rf sweep rate = 1 MHz/s; stochastic CW ENDOR cycle, rf-on = 3 ms, rf-off = 1 ms; sample collection time = 3 ms; and T = 2 K. See supplementary materials for details.

The strong isotropic coupling to one 13C of SAM reveals that this carbon is covalently integrated into the paramagnetic center of Ω. However, in addition to the absence of a large 1H2C(5′) hyperfine coupling (see above), the magnitude of aiso is ~10-fold too low (16, 17) to be attributed to the sought-after 5′-dAdo• primary-carbon radical, either free or weakly coupled to the cluster spin. As alternative possibilities, ω might be a precursor to SAM cleavage with enhanced bonding between the sulfonium sulfur and the unique Fe of the [4Fe-4S]1+ cluster, or a product of electron transfer in which the unpaired electron of Ω is on the sulfur of an intact SAM neutral radical. However, in either case we would expect that each of the three carbons bound to the sulfur would exhibit similarly strong hyperfine couplings. Instead, experiments using [methyl-13C]-SAM revealed that the coupling of the methyl 13C (A ~ 0.5 MHz) (Fig. 3B) is roughly 20 times weaker than that of the strongly coupled adenosyl 13C, ruling out these alternatives.

These considerations lead to the idea that Ω involves a SAM fragment associated with a spin S = 1/2 form of the [4Fe-4S] cluster. To test this idea, we used 57Fe-enriched PFL-AE to prepare Ω. ENDOR spectroscopy of this sample revealed an 57Fe signal centered at A/2 = ~17 MHz, essentially the same as the much-better-resolved signal for the 57Fe-labeled protein in its paramagnetic [4Fe-4S]+ state (Fig. 3C). This establishes that the electron spin giving rise to the Ω EPR signal resides on the iron-sulfur cluster, whereas the isotropic coupling to a single carbon of dAdo requires a carbon of the 5′-dAdo fragment to be covalently bound to the paramagnetic cluster. Although all carbons of the adenosyl moiety of SAM were 13C in this sample, the only reasonable interpretation is that the 5′ C of the 5′-dAdo• radical created by reductive cleavage of SAM has formed an Fe–C bond with the unique cluster iron and gives rise to the stronger coupling observed in the 13C ENDOR spectrum of Ω (Fig. 4). The weakly coupled carbon observed in the Mims ENDOR spectrum (Fig. 3B) would then be assigned to the 4′ C of dAdo.

Fig. 4 Model for bio-organometallic intermediate Ω.

Whether methionine remains coordinated to the unique iron site is not currently known. Blue, nitrogen; white, carbon; red, oxygen; yellow, sulfur; orange, iron.

In the commonly proposed mechanism of radical SAM enzymes, an electron from a [4Fe-4S]+ cluster is transferred to the sulfonium of SAM, thereby promoting S-[5′-C] bond cleavage to generate the 5′-dAdo• radical intermediate (1). If this is indeed the initial step in catalysis, then our current results would indicate that in PFL-AE, the 5′-dAdo• radical thus formed by SAM reduction adds to the [4Fe-4S]2+ cluster product of the electron transfer to generate Ω, an organometallic intermediate in which the S = 1/2 cluster, formally [4Fe-4S]3+, is covalently linked to 5′-dAdo through an Fe–C bond between [5′-C] and the unique iron of the cluster. As a plausible alternative mechanism, nucleophilic attack of the unique iron of the [4Fe-4S]+ cluster on the 5′ C of SAM initiates SAM cleavage, to release methionine and generate the [4Fe-4S]3+-adenosyl intermediate, Ω. In this case, the 5′-dAdo• radical is truly “never free” (9).

Regardless of the mechanism for its formation, the organometallic intermediate Ω proposed here for PFL-AE provides intriguing parallels to the adenosylcobalamin cofactor, which is used by B12 radical enzymes to initiate radical catalysis by homolytic cleavage of the Co(III)-[5′-C]-deoxyadenosyl bond to generate a 5′-dAdo• radical and Co(II) cobalamin (18, 19). Analogously, the H atom abstraction from G734 by Ω would begin with homolytic cleavage of the [4Fe-4S]3+ Fe-[5′-C]-adenosyl bond to generate [4Fe-4S]2+ and a 5′-dAdo• radical, which then abstracts a precisely positioned hydrogen atom from G734 of PFL.

The results reported here provide evidence that, contrary to expectation, cleavage of SAM by PFL-AE generates a catalytically competent bio-organometallic intermediate (Ω) in which the 5′-dAdo moiety derived from SAM is covalently bound through the 5′ C to the [4Fe-4S] cluster (Fig. 4). This intermediate generates the G• product radical on PFL during annealing of the activated protein-protein complex in the frozen solid at temperatures at or below 170 K (Fig. 2). This confirms that, in the ternary complex, the structures of both PFL and PFL-AE have rearranged so that the target G734 of PFL is proximate to the [4Fe-4S]/SAM radical-generating construct of PFL-AE. This proximity, which was predicted on the basis of structural studies of PFL-AE in complex with a heptamer peptide of PFL (5), requires that formation of the PFL-AE/PFL complex involves substantial conformational changes in PFL, whose G734 residue is normally buried 8 Å from the protein surface (20, 21).

If Ω is truly formed through nucleophilic attack of the unique Fe on the 5′ C of SAM as a means to cleave the S-[5′-C] bond of SAM and generate a radical initiator Fe-[5′-C] bond, then such an intermediate may well be common to many, or even all, radical-SAM enzymes. In particular, it is interesting to consider whether the formation of Ω might represent a mechanistic difference between those radical SAM enzymes that, like PFL-AE, use SAM as a cosubstrate and those that use SAM as a cofactor; further studies will be required to address this possibility. The mechanistic importance of such an intermediate may lie in providing a means to control and store the reactive 5′-dAdo• radical intermediate until the target hydrogen atom of the substrate is bound appropriately for abstraction. Given the complexity of the 170-kDa substrate PFL and the requirement for large protein conformational changes during its activation by PFL-AE (5, 21), the need for such storage and control seems plausible. In either case, the discovery of this organometallic radical initiator moiety in a radical-SAM enzyme provides an additional parallel to the B12 radical enzymes: Not only do both classes use a 5′-dAdo• radical in catalytic H atom abstraction, but both also employ a precursor complex with a direct metal-[5′-C] bond.

The work presented here reveals a catalytically competent [4Fe-4S]-cluster–bound 5′-deoxyadenosyl species in PFL-AE. This radical initiator Fe-[5′-C]-adenosyl complex expands the growing list of enzymatic bioorganometallic centers, whose first entry was coenzyme B12 (18, 19) but now includes the active-site metal clusters of hydrogenases (22), nitrogenase (23, 24), and a catalytic intermediate of IspH (25, 26). Our study shows that radical SAM enzymes, the largest known superfamily of enzymes, can also function through an organometallic center.


Materials and Methods

Supplementary Text

Figs. S1 to S3

References (2739)


Acknowledgments: This work was funded by the NIH (grant GM 111097 to B.M.H. and grant GM 54608 to J.B.B.). We thank B.-H. Huynh and C. Krebs for contributions in the early stages of this study.

Stay Connected to Science

Navigate This Article