An organometallic complex formed by an unconventional radical SAM enzyme

Abstract

Pyrococcus horikoshii Dph2 (PhDph2) is an unusual radical SAM enzyme involved in the first step of diphthamide biosynthesis. It catalyzes the reaction by cleaving S-adenosylmethionine (SAM) to generate a 3-amino-3-carboxy propyl (ACP) radical. To probe the reaction mechanism, we synthesized a SAM analog (SAM_CA), in which the ACP group of SAM is replaced with a 3-carboxyallyl group. SAM_CA is cleaved by PhDph2, yielding a paramagnetic (S = 1/2) species, which is assigned to a complex formed between the reaction product, α-sulfinyl-3-butenoic acid, and the [4Fe-4S] cluster. Electron-nuclear double resonance (ENDOR) measurements with ¹³C and ²H isotopically labeled SAM_CA support a π-complex between the C=C double bond of α-sulfinyl-3-butenoic acid, and the unique iron of the [4Fe-4S] cluster. This is the first example of a radical SAM related [4Fe-4S]⁺ cluster forming an organometallic complex with an alkene, shedding additional light into the mechanism of PhDph2 and expanding our current notions for the reactivity of [4Fe-4S] clusters in radical SAM enzymes.

Graphical Abstract

Diphthamide is a post-translationally modified histidine residue on archaeal and eukaryotic translation elongation factor 2 (EF2), a protein essential for ribosomal protein synthesis. Its biosynthesis involves at least seven proteins.^1,2 The first step involves the transfer of 3-amino-3-carboxypropyl (ACP) group derived from S-adenosyl-L-methionine (SAM) to the histidine residue of EF2. In vitro, the [4Fe-4S]-containing radical SAM (RS) enzyme PhDph2³ or the eukaryotic Dph1-Dph2 homodimer⁴ is sufficient for this step, with dithionite as the reducing agent. In contrast to classical RS enzymes that cleave the C_5′,Ade–S bond of SAM to form 5′-deoxyadenosyl radical (5′-dA•) (Scheme 1A)⁵, PhDph2 cleaves the C_γ,Met–S bond to form an ACP radical (Scheme 1B).³

Scheme 1.

PhDph2 is different from classical RS enzymes.

The formation of the ACP radical is supported by the PhDph2-catalyzed generation of 2-aminobutyric acid and homocysteine sulfinic acid in the absence of the substrate protein, PhEF2.³ However, the ACP radical has not been directly observed even when we carried out freeze quench-electron paramagnetic resonance (EPR) experiments (Figure S1). This is not surprising because the ACP radical is expected to be short-lived. Following the strategy developed by Frey and coworkers to generate a more stable allylic analog of the 5′-dA•^6.7, we synthesized a SAM analog, SAM_CA, in which the ACP group of SAM was replaced with a 3-carboxyallyl group (Scheme 2). If PhDph2 could accept SAM_CA as a substrate, a 3-carboxyallyl radical would be generated, which should be more stable and allow direct observation by EPR. As shown below, even with SAM_CA, we could not detect the radical intermediate. However, surprisingly, we found that a dithionite quenched radical product formed an organometallic complex with the Fe-S cluster, revealing an interesting reactivity of the [4Fe-4S] cluster in RS enzymes.

Scheme 2.

Structures of SAM_CA and aza-SAM_CA.

We initially examined whether SAM_CA could serve as a substrate of PhDph2. High-performance liquid chromatography (HPLC) showed that all the SAM_CA was converted to 5′-deoxy-5′-methylthioadenosine (MTA) within 20 min, whereas in the absence of dithionite or PhDph2, no SAM_CA was cleaved (Figure 1A). These results demonstrate that PhDph2 is able to catalyze the cleavage of the ‘correct’ C-S bond of SAM_CA. In the presence of the substrate protein PhEF2, the same reaction product was detected (Figure S2). However, PhEF2 was not modified by SAM_CA. This was likely because SAM_CA generated a more stable radical (see more experimental evidence for the stability below), which was not active enough to react with PhEF2.

Figure 1.

A). HPLC traces of the SAM_CA cleavage reaction (reaction time 20 min) catalyzed by PhDph2 at room temperature. B). X-band CW EPR spectra of PhDph2 in the absence and presence of SAM_CA. A spectral simulation for the S = 1/2 species is shown as green dotted line. T = 12 K.

We then prepared freeze quench samples and used EPR to detect radical signals. The dithionite-reduced PhDph2 in the absence of SAM_CA showed a typical [4Fe-4S]¹⁺ EPR signal at 12 K (Figure 1B),³ which was temperature-dependent and broadened beyond detection at 30 K (Figure S1). When PhDph2 in the presence of dithionite was allowed to react with SAM_CA for 30 s, the [4Fe-4S]¹⁺ signal was replaced by a nearly axial signal, with a split feature at g₁ (Figure 1B). The ratio of this g₁ splitting at X- and Q-band scaled with the ratio of microwave frequencies (Figure S3), showing that the splitting represents two different conformations and not a hyperfine interaction with a paramagnetic nucleus. Simulation of the spectra gave for the one conformer principal g-values g = [2.092, 2.010, 1.990], and for the second one g = [2.074, 2.010, 1.990]. Both signals exhibited similar relaxation properties and were hardly detectable above 70 K (Figure S1). The large g-anisotropy and the absence of proton hyperfine splittings suggest that it is not a free allyl radical.^6,8 In contrast, the signals of the new species formed in the reaction of the reduced PhDph2 with SAM_CA exhibited an “isotropic” g-value, g_iso = (g₁₁+g₂₂+g₃₃)/3 > g_e = 2.0023,⁹ which is reminiscent of that observed in high potential iron-sulfur proteins (HiPIP) and [4Fe-4S]³⁺ intermediates trapped with the [4Fe-4S] proteins IspG¹⁰ and IspH¹¹. With ⁵⁷Fe-enriched PhDph2, the EPR signal became broadened (Figure 1B), further supporting that the new species (termed CA) is associated with the [4Fe-4S] cluster.

When we monitored the reaction at different time points, the EPR signal of CA persisted for >30 min, even after all SAM_CA was consumed (Figure S4). This suggested that CA was a complex of the reaction product with the [4Fe-4S]¹⁺ cluster, the latter regenerated by dithionite reduction of the [4Fe-4S]²⁺ produced in the reaction. To test this hypothesis, protein-free aliquots of the reaction products were added to ‘fresh’ dithionite-reduced PhDph2. The same anisotropic EPR signal as that of CA was obtained, supporting that CA is a product complex (Figure S5).

To understand the structure of CA, we analyzed the reaction products using ¹H-NMR. Several new peaks were observed from the reaction containing PhDph2, SAM_CA, and dithionite (Figure S6), which were absent in control reactions without dithionite. Some of the new peaks were assigned to MTA. Crotonic acid and its isomer, 2-butenoic acid, which are the hydrogen abstraction products of the carboxyallyl radical, were not formed by comparison to the spectra of their standards. Three new peaks (a–c) were assigned to γ-sulfinylcrotonic acid (Figure S6) and the remaining three peaks (d, e and f) around 5.1 and 5.8 ppm were assigned to α-sulfinyl-3-butenoic acid (Figure S6); the remaining proton on α-sulfinyl-3-butenoic acid is localized at 3.4 ppm based on ¹H-¹H COSY (correlation spectroscopy) (Figure S14, S15). Using 1,1-²H₂-SAM_CA as the substrate, peaks c, e and f (Figure S7) disappeared, which supported the assignments. The assignments were confirmed by ¹H/¹³C HMBC (Heteronuclear Multiple Bond Correlation), HSQC (Heteronuclear Single Quantum Correlation) and ¹H-¹H COSY (Figure S8 – S15). The reaction products were also chemically modified to allow detection by LC-MS (Figure S16). Our data thus demonstrated the formation of γ-sulfinylcrotonic and α-sulfinyl-3-butenoic acid, which likely resulted from the recombination of the carboxyallyl radical with the dithionite-derived SO₂^−• radical, similar to previously reported reactions of radicals with dithionite ^3,12,13. The NMR signal intensities suggested roughly equal formation of the two products.

We then performed 35 GHz CW/pulse ¹³C, ¹H and ²H ENDOR with 1-¹³C, 2-¹³C, 3-¹³C, 1,1-²H₂ and 1,1,2,3-²H₄ labeled SAM_CA (see Scheme 2). Field-modulated 35 GHz CW ENDOR spectra obtained from reactions with 1-¹³C, 2-¹³C labeled SAM_CA both exhibit ¹³C doublets with broad component lines, split by the ¹³C hyperfine coupling, A ~ 5–8 MHz (Figure 2). An estimate of the magnitude of the hyperfine coupling tensors for these two ¹³C sites of SAM_CA (Table S1) was obtained through analysis of 2D field-frequency patterns of ENDOR spectra collected across the EPR envelopes of the SAM_CA isotopologs (Figures S17, S18). The analysis showed comparable isotropic hyperfine tensors, A_iso ~ 7 MHz for both 1-¹³C and 2-¹³C. For 3-¹³C labeled SAM_CA, a considerably smaller isotropic coupling, A_iso ~ 0.6 MHz, was estimated using Mims pulsed ENDOR (Figure 2A, Inset).

Figure 2.

(A) 35 GHz CW ¹³C ENDOR spectra for CA with 1-¹³C and 2-¹³C-labeled SAM_CA collected at g₂; the doublets are split by the hyperfine coupling, and centered at the Larmor frequency as described in SI. Inset: Mims ENDOR spectra of 3-¹³C SAM_CA. (B) ¹H Davies ENDOR difference spectra of CA generated by subtracting from the spectrum of ¹H-SAM_CA (natural abundance) the spectra for 1,1,2,3-²H-labeled SAM_CA (upper) and 1,1-²H-labeled SAM_CA (lower). Spectra were collected at g₂ and normalized by EPR echo height.

To further probe the structure of CA, ¹H and ²H pulse ENDOR measurements were performed with 1,1-²H₂, and 1,1,2,3-²H₄ labeled SAM_CA. Only the pulsed ENDOR techniques resolved individual features, but the lower sensitivity of pulsed vs CW ENDOR allowed spectra to be collected only at the intensity maximum of the EPR envelope, near g₂. The ¹H signals from the labeled sites were obtained by subtracting the spectra obtained using the natural-abundance SAM_CA (Figure 2B). After subtraction of the 1,1-²H₂ spectrum, a doublet of ¹H peaks assignable to the 1,1-¹H nuclei was obtained with coupling constant of A ~ 7 MHz; subtraction with the 1,1,2,3-²H₄ labeled SAM_CA showed an enhanced intensity for this doublet (Figure 2B), indicating that the 1,1-¹H and 2-¹H nuclei had comparable couplings (assigned based on 2-¹³C data). In addition, weak broad features corresponding to a doublet with A ~ 2 MHz were assigned to 3-¹H. Mims ²H ENDOR spectra (Figure S19 and S20) provided support for this assignment. The substantial isotropic couplings to 1-¹³C and 2-¹³C (A_iso ~ 7 MHz), the deviations of the g-values from g~2, and the ⁵⁷Fe broadening imply the spin is primarily localized on the cluster, but with close C-Fe interaction.

The similarity of 1-¹³C and 2-¹³C hyperfine couplings, as well as the similarity of the 1,1-²H and the 2-²H couplings, suggests that both 1-C and 2-C have similar direct interactions with the unique Fe of the [4Fe-4S] cluster. In contrast, the weaker 3-¹³C and 3-²H hyperfine couplings show that 3-C does not experience a significant direct interaction with the Fe. These results suggest that CA only consists of one of the two products, α-sulfinyl-3-butenoic acid, coordinated to the unique Fe site of the [4Fe-4S] cluster. If CA also involved a complex with γ-sulfinylcrotonic acid, 2-C and 3-C should have similar strong hyperfine couplings, and 1-C should have weaker coupling, contrary to our observations (Figure 2A). We further confirmed this by utilizing the instability of α-sulfinyl-3-butenoic acid. After incubating the protein-free aliquots of the reaction products at room temperature overnight, the peaks corresponding to α-sulfinyl-3-butenoic acid disappeared in the ¹H-NMR spectrum (Figure S21), while the signals from γ-sulfinylcrotonic acid remained. The instability of α-sulfinyl-3-butenoic acid is likely due to the acidity of the α-proton, which is adjacent to two strong electron-withdrawing groups.¹⁴ When the mixture containing only the γ-sulfinylcrotonic acid product was added to ‘fresh’ dithionite-reduced PhDph2, the signal associated with CA was not be detected by EPR (Figure S22). This result supports the notion that γ-sulfinylcrotonic acid does not form an EPR-active complex with the [4Fe-4S] cluster and that CA is best described as a complex formed between α-sulfinyl-3-butenoic acid and the [4Fe-4S] cluster.

We then probed whether the sulfinyl and carboxyl of α-sulfinyl-3-butenoic acid groups are required for CA formation. We synthesized a non-cleavable analog of SAM_CA, aza-SAM_CA (Scheme 2)¹⁵. The freeze quench-EPR experiment with aza-SAM_CA showed only a [4Fe-4S]⁺ signal (Figure S23) that was similar to that of the reduced cluster in PhDph2 in the absence of SAM_CA. Therefore, CA is not formed by binding of uncleaved SAM_CA to the [4Fe-4S] cluster. Addition of crotonic acid to PhDph2 also failed to generate the CA species. Thus, the sulfinic group in α-sulfinyl-3-butenoic acid is likely involved in the formation of CA. Similarly, adding allylsulfinic acid to PhDph2 did not generate the CA signal either. Thus, the carboxyl group in α-sulfinyl-3-butenoic acid is also necessary. The results suggest that both the sulfinic group and the carboxylate of α-sulfinyl-3-butenoic acid bind to the Fe-S cluster, facilitating the formation of a π-complex between that Fe and the C=C double bond of the product (Figure 3).

Figure 3.

Plausible structures of CA and formation mechanism.

CA can have two formal resonance structures (Figure 3, structures a and b). To further probe the oxidation state of CA, we attempted to reduce CA by low-temperature γ-irradiation (cryoreduction)¹⁶ (Figure S24). The EPR signal did not decrease upon γ-irradiation, as would be expected for a [4Fe-4S]³⁺ cluster. Similarly, extra dithionite could not reduce CA, which is inconsistent with the expectation for a [4Fe-4S]³⁺ cluster (Figure S25)¹⁷. In contrast, when we tried to oxidize CA with K₃[Fe(CN)₆], CA was oxidatively degraded to a [3Fe-4S]⁺ cluster (Figure S25), an outcome commonly observed in similar treatments of [4Fe-4S]⁺ clusters of RS enzymes with chemical oxidants. Thus, the Fe-S core in CA behaves like a [4Fe-4S]⁺ cluster. We also sought to use ⁵⁷Fe Mössbauer spectroscopy. However, owing to the rather modest yield of CA (50% of total Fe) and the presence of at least one additional EPR-silent Fe-containing product, the analysis of the spectra does not permit definitive conclusions about the electronic structure of CA (Figures S26, 27 and 28).

With the characterization of reaction products and the structure of CA, we proposed a plausible mechanism for the formation of CA (Figure 3). The [4Fe-4S]⁺ cluster in PhDph2 provides one electron to SAM_CA, cleaving the C_CA-S bond and generating a 3-carboxyallyl radical, MTA, and a [4Fe-4S]²⁺ cluster. The 3-carboxyallyl radical reacts with a dithionite-derived SO₂^−• radical, forming γ-sulfinylcrotonic acid and α-sulfinyl-3-butenoic acid. The sulfinic group, the carboxyl group, and the C=C double bond of α-sulfinyl-3-butenoic acid, coordinate the [4Fe-4S]⁺ cluster (reduced by extra dithionite) to generate the π-complex. The coordination of the sulfinic and carboxyl group helps to place the double bond in a favorable position for π interactions with the unique Fe. The two conformers revealed by EPR (Figure 1B) could be attributed to different binding orientations of the α-sulfinyl-3-butenoic acid to [4Fe-4S]⁺ cluster (the three chelating functional groups may exchange positions).

In summary, using SAM_CA, we have detected an EPR-active species, CA, formed in the enzymatic reaction of PhDph2. Various experiments demonstrate that CA is a complex formed between the [4Fe-4S]¹⁺ cluster and one of the reaction products, α-sulfinyl-3-butenoic acid, coordinated to the unique cluster Fe through the sulfinyl and carboxylate oxygens, with the C=C double bond forming an organometallic π-complex with the Fe. Although we could not accumulate the radical intermediate that we set out to detect with SAM_CA, formation of both α-sulfinyl-3-butenoic acid and γ-sulfinylcrotonic acids provides additional support for the radical mechanism of PhDph2 reaction, as a nucleophilic mechanism cannot account for their formation. Interestingly, although the EPR g tensors of CA shows similarity to those of [4Fe-4S]³⁺-like clusters observed in two enzymes involved in the non-mevalonate pathway, IspG and IspH, the cluster in CA behaves like a [4Fe-4S]⁺ cluster.

The present case can be described as the reaction of the unique Fe of the [4Fe-4S]⁺ cluster with the alkene product to form a stable organometallic complex. Such an alkene complex has reported for the nitrogenase FeMo-cofactor and the ethylenic product of alkyne reduction¹⁸, and for mutants of IspH and its substrate HMBPP.^11,19 However, similar organometallic complex had not been demonstrated for RS enzymes until very recently, when the [4Fe-4S]²⁺ cluster of the RS enzyme, pyruvate formatelyase activating enzyme is shown to react with the 5′-dA• to form a highly reactive organometallic intermediate.²⁰ This study and our current result suggest that the unique iron of [4Fe-4S] clusters in RS enzymes can readily form a C-Fe complex. Such a property not only may help to explain the unusual chemistries catalyzed by the [4Fe-4S]-containing RS enzymes, but also may open up new avenues to tune the activity or develop new chemistry for this enzyme class or for other [4Fe-4S]-containing enzymes.