Radical S-adenosyl-l-methionine FeS cluster implicated as the sulfur donor during albomycin biosynthesis

Abstract

Carbon–sulfur bond-forming reactions in natural product biosynthesis largely involve Lewis acid/base chemistry with relatively few examples catalysed by radical S-adenosyl-l-methionine (SAM) enzymes. The latter have been limited to radical-mediated sulfur insertion into carbon–hydrogen bonds with the sulfur atom originating from a sacrificial auxiliary iron–sulfur cluster. Here we show that the radical SAM enzyme AbmM encoded in the albomycin biosynthetic gene cluster catalyses a sulfur-for-oxygen swapping reaction, transforming the furanose ring of cytidine 5′-diphosphate to a thiofuranose moiety that is essential for the antibacterial activity of albomycin δ₂. Thus, in addition to its canonical function of mediating the reductive cleavage of SAM, the radical SAM catalytic cluster of AbmM appears to play a role in providing the sulfur introduced during the AbmM-catalysed reaction. These discoveries not only establish the origin of the thiofuranose core in albomycin δ₂ but, more importantly, also emphasize the functional diversity of radical SAM catalysis.

Albomycins (1–3) are produced by species of Streptomycetes and are characterized by a thioheptose dipeptide structural component known as SB-217452 (4), which is a potent seryl-tRNA synthetase inhibitor^1–6 (Fig. 1). SB-217452 is linked in albomycins to the ferrichrome-type siderophore 5, which is critical for the uptake of albomycins by pathogenic bacteria via their iron acquisition systems such that they operate as ‘Trojan horse’ antibiotics⁷. Albomycins are further distinguished by a thioether linkage that connects C1′ and C4′ of the furanose ring, rendering them structurally unique among primary and secondary nucleoside metabolites. Moreover, a synthetic analogue of 3 which contains an oxygen in place of the sulfur is biologically inactive, indicating that the thiofuranose ring is critical for the antibacterial activity of albomycins⁸. However, the mechanism by which sulfur is incorporated into the furanose ring in albomycins remains obscure.

Fig. 1 |. Proposed biosynthetic pathway of albomycins.

Albomycins are composed of two principal structural features including a siderophore (5) and the seryl-tRNA synthetase inhibitor SB-217452 (4), which has a unique thiofuranose moiety. Construction of the thiofuranose involves the formation of intermediate 6 but is otherwise unclear. P, phosphate group; PP, diphosphate group; PPP, triphosphate group; PLP, pyridoxal 5′-phosphate.

In this Article a radical-mediated, sulfur-for-oxygen swapping reaction is identified in the biosynthetic pathway of albomycin δ₂. The swapping of the bridging oxygen for sulfur is catalysed by the twitch radical S-adenosyl-l-methionine (SAM) enzyme AbmM encoded in the albomycin biosynthetic abm gene cluster and results in transformation of the furanose core of cytidine 5′-diphosphate (CDP) to a thiofuranose ring. Moreover, the incorporated sulfur appears to be associated with the radical SAM catalytic iron–sulfur (FeS) cluster rather than an auxiliary cluster. Thus, beyond its widely recognized role in mediating the reductive cleavage of SAM, the catalytic [Fe₄S₄] cluster in AbmM may also be responsible for donating the sulfur atom to be incorporated into the furanose core.

Results

Functional characterization of AbmM

Most of the proteins encoded in the abm gene cluster responsible for the biosynthesis of albomycin δ₂ (1) have already been characterized^9–11. In particular, AbmH has been shown to be a PLP-dependent transaldolase that catalyses coupling of the 5′-oxo-4′-thionucleoside 6 and l-threonine to form the (5′S,6′S)-α-amino-β-hydroxyacid-containing adduct 7 (ref. 10; Fig. 1). However, the oxyribose analogue of 6 is a poor substrate for AbmH¹⁰, implying that the incorporation of sulfur into the nucleoside occurs before AbmH-catalysed C5′–C6′ bond formation. Among the few genes that remain unassigned in the abm cluster, abmM is a likely candidate for the putative sulfur insertion activity. AbmM (UniProt ID I1ZAZ9) is annotated as a twitch radical SAM enzyme^12–14 and is predicted to harbour two [Fe₄S₄] clusters on the basis of sequence analysis^9,15. As albomycin δ₂ (1) is a cytidine-type nucleoside, AbmM is hypothesized to catalyse the conversion of cytidine or a cytidine-based ribonucleotide (for example, 9–11) to its thiosugar counterpart during the early stages of albomycin biosynthesis.

To investigate the possible involvement of AbmM in albomycin biosynthesis, the abmM gene in the producing strain Streptomyces sp. ATCC 700974 was deleted, and the resulting strain was found to produce neither albomycin δ₂ (1) nor SB-217452 (4). However, production could be restored when the abmM gene was reintroduced to the deletion mutant (Extended Data Fig. 1a–d). Thus, a functional abmM gene is essential for the in vivo production of albomycins. The N-His₆-small ubiquitin-like modifier (SUMO)-tagged AbmM was expressed in Escherichia coli, aerobically purified and treated with SUMO protease Ulp1 to yield AbmM with the native sequence (Extended Data Fig. 1e). Ultraviolet–visible (UV–vis) absorption spectroscopy of the as-isolated protein showed a broad band at 410 nm that decreased upon treatment with dithionite, consistent with one or more [Fe₄S₄]²⁺ clusters bound by AbmM (ref. 16; Extended Data Fig. 1f).

Subsequent iron/sulfide quantitation showed that the as-purified AbmM contains 7.0 ± 0.4 equiv. Fe and 4.7 ± 0.4 equiv. S per protein monomer. Repeating the iron/sulfur quantitation after the mutation of all three conserved cysteine residues in the CxxxCxxC radical SAM domain to alanines (AbmM-C56/C60/C63A, the ΔRS-cluster variant) yielded 5.9 ± 0.1 equiv. Fe and 3.8 ± 0.2 equiv. S. These results implied that AbmM has one auxiliary [Fe₄S₄] cluster (AUX cluster) in addition to the canonical radical SAM [Fe₄S₄] cluster (RS cluster) with at least partial decomposition of the latter during aerobic purification from E. coli.

To study the activity of AbmM, cytidine (8) and cytidine-based nucleotides (9–11) were initially tested as possible substrates for AbmM under anaerobic conditions without previous FeS cluster reconstitution of the enzyme, which appeared to exhibit sufficient activity for the purposes of substrate screening (Methods). When 0.4 mM CDP (10) was incubated with 34 μM AbmM, 0.6 mM SAM, 10.0 mM 1,4-dithiothreitol (DTT), 1.0 mM reduced nicotinamide adenine dinucleotide phosphate (NADPH), 1.0 mM methyl viologen (MV) and 2.0 mM sodium sulfide (Na₂S), which can serve as a potential sulfur donor, a new product was detected after 12 h, together with 5′-deoxyadenosine (5′-dA) in an approximate 1:1 ratio using HPLC (Fig. 2b). l-Methionine was also detected via liquid chromatography–mass spectrometry (LCMS) analysis (Extended Data Fig. 2a). It was noted that the reducing system comprising NADPH and MV could be substituted by sodium dithionite (DT) to detect the new product. Electrospray ionization mass spectrometry (ESI-MS) analysis of the new product revealed a mass increase of 32 Da from CDP, implying the insertion of one sulfur atom into CDP (Extended Data Fig. 2b). The new product was dephosphorylated using calf intestinal phosphatase (CIP) to facilitate further characterization. ESI-MS fragmentation analysis of the resulting nucleoside indicated that sulfur incorporation is at the ribose moiety instead of the cytosine nucleobase, the molecular mass of which remained unchanged (mass-to-charge ratio (m/z) 111.04) (Fig. 2c).

Fig. 2 |. AbmM-catalysed sulfur insertion into CDP.

a, Structural characterization of the AbmM product. b, HPLC analysis of the AbmM reaction with CDP (10). AU, absorbance units. c, ESI-MS spectrum of 14 (m/z 276) generated from the AbmM product (13) following dephosphorylation with CIP.

Whereas the dephosphorylated product 14 eluted as a single peak via HPLC analysis (Extended Data Fig. 2e), its NMR spectrum showed a 69:31 mixture of two isomers at pH 5–6 (Supplementary Fig. 3). Notably, no H-4′ peak was observed for either isomer (Supplementary Fig. 3), and the ¹³C resonance of C1′ for both isomers appears at 64.5 ppm, which is similar to that of 4′-thiocytidine (12; 65.5 ppm) but clearly distinct from that of cytidine (8; 96.2 ppm) (Fig. 2a and Supplementary Fig. 3). By contrast, the C4′ signal for both isomers is shifted downfield (94.0 or 92.1 ppm) compared with 8 (83.7 ppm) and 12 (52.1 ppm) (Fig. 2a). These results indicated that the AbmM product has a thiobridge between C1′ and C4′ and a hemithioketal moiety at C4′. Thus, the major and minor product species after dephosphorylation were assigned as (4′S)-14 and (4′R)-14, respectively, because through-space correlation between H-6 and H-5′ was only observed for the minor epimer via NOESY (nuclear Overhauser effect spectroscopy) (Supplementary Fig. 7). Off-diagonal exchange cross-peaks were also observed for 14 using EXSY (exchange spectroscopy), suggesting that the two stereoisomers are interconvertible (Supplementary Fig. 8b–d). Furthermore, when 14 was reacted with monobromobimane (mBBr), which forms adducts with free thiols¹⁷, a new product consistent with open-14-bimane was detected via LCMS (Extended Data Fig. 3a,b). This adduct could be reduced using NaBH₄ (or NaBD₄) to yield two isomeric species (15) with a mass increase of 2 Da (or 3 Da with NaBD₄). Collectively, these results support assignment of the AbmM product as 13 (Fig. 2a). AbmM was also found to accept cytidine 5′-triphosphate (CTP; 11) as a substrate, but not cytidine (8) or cytidine 5′-monophosphate (CMP; 9) (Extended Data Fig. 2d).

Isotope labelling experiments

A putative mechanism for the AbmM-catalysed reaction is shown in Fig. 3a. This hypothesis predicts hydrogen-atom abstraction from the C4′ position of CDP (10) by a 5′-deoxyadenosyl radical (5′-dA^•; 16) following the reductive cleavage of SAM^16,18. This is consistent with the observation of 95% deuterium incorporation into 5′-dA (17) and no deuterium retention in dephosphorylated 14 when [4′-²H]-CDP (98% D) was used as the substrate (Fig. 3b,c). Experiments with [3′-²H]-CDP and [5′,5′′-²H₂]-CDP led to no deuterium transfer to 5′-dA and retention of the deuterium label(s) in 14 (Extended Data Fig. 3c,d). The proposed mechanism is also consistent with the absence of observable ¹⁸O incorporation into 13 when the reaction was conducted in H₂¹⁸O, which implies that the bridging oxygen atom in CDP (10) is retained as the C4′-hydroxyl group in 13 (Extended Data Fig. 3e).

Fig. 3 |. Mechanistic investigation of AbmM catalysis.

a, Proposed mechanism of AbmM catalysis with CDP (10). b,c, ESI-MS analysis of 5′-dA (17) (b) and 14 (c) produced in the AbmM-catalysed reaction with [4′-²H]-CDP. The analysis was performed after workup with CIP.

Reconstitution by the FeS cluster scaffold proteins

The mechanism in Fig. 3a implies that the incorporated sulfur originates from free sulfide (19 → 20 → open-13). To test this hypothesis, the reaction using as-purified AbmM was repeated with Na₂³⁴S (>99% ³⁴S). Following overnight incubation, 14 was found to contain 90% ³⁴S; however, this value was reduced to 64% when the reaction was analysed after 10 min (Fig. 4a). Furthermore, 14 and 5′-dA (17) were still produced in reactions without Na₂S, although the presence of 2 mM Na₂S boosted the number of turnovers from one to between seven and eight per protein monomer following incubation for 12 h (Extended Data Fig. 2c). These results suggested that exogenous Na₂S is not the immediate sulfur donor in the AbmM-catalysed reaction, and an enzyme-bound donor such as an FeS cluster that can be replenished by free sulfide to facilitate multiple turnovers instead serves in this role.

Fig. 4 |. In vitro reactions of AbmM with various sulfur sources.

a, ESI-MS analysis of 14 produced during the AbmM-catalysed reaction using Na₂³⁴S after 10 min (top) and after 12 h (bottom). b, Formation of 14 (product) in the AbmM reaction with various external sulfur sources. The reaction mixtures contained 267 μM CDP, 600 μM SAM, 15 μM AbmM, 2 mM DT, 300 μM Na₂S or 150 μM FeS cluster carriers or scaffold proteins. The analysis was performed after the incubation mixture had been treated with CIP. Circles, incubation with AbmM alone; squares, with Na₂S; hexagons, with IscU; rhombuses, with SufU; triangles, with NfuA. [Product], product (14) concentration. c, Formation of 14 (product) in the AbmM reaction with various concentrations of NfuA. Circles, incubation with 20 μM NfuA; rhombuses, with 40 μM NfuA; triangles, with 80 μM NfuA; hexagons, with 160 μM NfuA. In b and c, the time courses were measured in triplicate, and the data points denote the mean ± s.d.

Whereas in vitro FeS cluster reconstitution in radical SAM proteins can be facilitated by the presence of Na₂S and FeCl₃ alone, in vivo this process typically involves multiple proteins^19–21. For example, the FeS cluster carrier protein NfuA and the scaffold protein IscU have been shown to facilitate reconstitution of the enzyme lipoyl synthase (LipA) to sustain its catalytic activity in E. coli²². Scaffold protein SufU is a homologue of IscU (34%/55% identity/similarity) encoded by the suf operon of the albomycin-producing strain Streptomyces sp. ATCC 700974. Therefore, to investigate a model where an [Fe₄S₄] cluster in AbmM acts as the sulfur donor and SufU serves to support AbmM catalysis, the sufU gene was heterologously expressed in E. coli, and partially purified SufU was reconstituted with FeCl₃ and Na₂S before desalting (Supplementary Fig. 1). A broad absorbance at 410 nm of the reconstituted SufU suggested the presence of an [Fe₄S₄]²⁺ cluster as reported for E. coli IscU²² (Supplementary Fig. 2). Incubation of 150 μM SufU with 15 μM reconstituted AbmM and 267 μM CDP for 90 min in the absence of Na₂S generated 86 ± 2 μM 14 after CIP treatment (Fig. 4b). By contrast, an equivalent concentration (300 μM) of Na₂S led to much less production of 14 (25 ± 3 μM) although it still increased the production of 14 compared with conditions without an added sulfur source (12 ± 2 μM). Furthermore, NfuA and IscU from E. coli could also be substitutes for SufU in the AbmM assay, with the former displaying a more pronounced effect in supporting AbmM turnover with up to 153 ± 10 μM 14 produced after a 90 min incubation (Fig. 4b). Moreover, total product generation after incubation for 90 min correlated with increasing NfuA concentration, with a slope of 0.9 ± 0.1 and an intercept of 23 ± 8 μM (Fig. 4c), which implies one additional micromole of turnover per micromole NfuA once the initially reconstituted AbmM has been consumed. These results are consistent with one of the two [Fe₄S₄] clusters in AbmM serving as the sulfur donor, and that this cluster can be reconstituted by SufU, NfuA or IscU to support multiple turnovers.

X-ray crystal structure of AUX-cluster-bound AbmM

Iron–sulfur clusters ([Fe₄S₄] or [Fe₂S₂]) are known to act as sulfur donors in C–S bond-forming reactions catalysed by radical SAM enzymes such as biotin synthase (BioB)^23,24 and LipA^25–27 in the biosynthesis of biotin and lipoic acid, respectively. In both cases, one bridging sulfur of the auxiliary cluster serves as the sulfur donor. In addition, one of the residues coordinating the auxiliary clusters of BioB and LipA is not cysteine but is instead an arginine residue in BioB or a serine in LipA. It has been suggested that this unique ligand may facilitate iron dissociation and thus sulfur transfer during lipoic acid biosynthesis²⁷. The reactions catalysed by the methylthiotransferases MiaB²⁸ and RimO²⁹ are two other related examples in which their auxiliary clusters also participate in sulfur transfer. For MiaB, the sulfur donor is the methylated sulfide component of its [Fe₃S₄] auxiliary cluster²⁸.

The structure of AbmM was determined at 2.5 Å resolution following crystallization under anaerobic conditions (Extended Data Fig. 4a,b and Extended Data Table 1). The asymmetric unit contains two molecules of AbmM. The structure solved for AbmM was found to bind only one auxiliary [Fe₄S₄] cluster (AUX cluster), which is coordinated by C267/C270/C276/C294 in the C-terminal domain as predicted (Extended Data Fig. 4c). By contrast, the catalytic [Fe₄S₄] cluster (RS cluster) was not observed, and the loop containing the conserved C56/C60/C63 motif was found to be disordered. Attempts to obtain a structure of AbmM with the RS cluster bound were unsuccessful. A comparison between the crystal structure of AbmM and its structure predicted using the neural network-based model AlphaFold2³⁰ (Extended Data Fig. 4d) revealed a root mean squared deviation value of 0.71 Å, which suggests that the latter is a reasonable model for AbmM. The disordered region in the predicted structure adopts a conformation that is suitable for binding an RS cluster and aligns well with that of other twitch radical SAM enzymes such as MoaA (refs. 31,32; Extended Data Fig. 4e). However, the AUX cluster in AbmM appears less likely to be the sulfur source, because it is coordinated by four cysteine residues, which may render this cluster more stable (Extended Data Fig. 4c). Moreover, the AUX cluster may be difficult to access as it is located ~19 Å away from the RS cluster where formation of the substrate radical occurs (for example, 18 or 19); however, a conformational change of AbmM upon substrate binding or RS cluster reconstitution cannot be excluded¹³ (Extended Data Fig. 4f). This distance is longer than those observed in BioB and LipA, for which the AUX cluster is positioned within 12–16 Å from the RS cluster^24,27. Hence, the RS cluster may be a more suitable sulfur donor during the AbmM-catalysed reaction, given its closer proximity to the predicted substrate radical.

Mössbauer spectroscopy

To study the roles of the two [Fe₄S₄] clusters during AbmM catalysis, they were characterized using a combination of ⁵⁷Fe Mössbauer spectroscopy and electron paramagnetic resonance (EPR) spectroscopy following expression in ⁵⁷Fe-enriched media. The preparation of an AbmM mutant that does not coordinate the auxiliary cluster (that is, ΔAUX cluster) proved futile due to insolubility; by contrast, however, the C56/C60/C63A (ΔRS cluster) mutant, which does not bind the RS cluster, could be obtained in soluble form. Before reduction, both reconstituted ⁵⁷Fe-labelled wild-type AbmM and ΔRS-cluster AbmM showed almost identical Mössbauer spectra (Extended Data Fig. 5a,b) exhibiting a major quadrupole doublet feature that is typical of [Fe₄S₄]²⁺ clusters, with an isomer shift (δ) of ~0.45 mm s⁻¹ and a quadrupole splitting (|ΔE_Q|) of 1.16 mm s⁻¹ (see Extended Data Tables 2 and 3 for simulation parameters). Upon treatment with DT, a mixture of [Fe₄S₄]²⁺ and [Fe₄S₄]⁺ clusters was found for both enzymes (Extended Data Fig. 5c,d). After incubating SAM and CDP (10) with reduced wild-type AbmM (reacted ⁵⁷Fe-AbmM), the formation of a high-spin (S = 2) ferrous species (Fe^II-A) was detected, which represented ~17% of the total iron in the sample (Fig. 5a,b and Extended Data Tables 2 and 3). The Mössbauer parameters of this species (δ = 0.80 mm s⁻¹, |ΔE_Q | = 3.30 mm s⁻¹) are consistent with ferrous ions coordinated by a thiolate-rich ligand environment. Thus, the formation of Fe^II-A may be attributed to the coordination of DTT in the buffer with iron released following degradation of one of the two clusters during AbmM catalysis.

Fig. 5 |. ⁵⁷Fe Mössbauer spectra of wild-type AbmM with selective ⁵⁷Fe labelling.

a–f, Mössbauer spectra of DT-treated ⁵⁷Fe-AbmM (a), reacted DT-treated ⁵⁷Fe-AbmM in the presence of 10 and SAM (b), DT-treated ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) (c), reacted DT-treated ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) in the presence of 10 and SAM and freeze-trapped after 15 min (d), DT-treated ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) (e) and reacted DT-treated ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) in the presence of 10 and SAM and freeze-trapped after 15 min (f). g, Difference spectrum of spectrum f minus spectrum e. The measurement conditions were 4.2 K and the applied external magnetic fields parallel to the γ-radiation as indicated in the figure. Throughout, black bars show the experimental data with the bar length indicating the statistical uncertainty of each data point, red solid lines represent the overall simulations, dashed coloured lines show the individual components of the spectroscopic simulations: blue, [Fe₄S₄]²⁺; green, [Fe₄S₄]⁺; orange, Fe^II-A; light green, Fe^II-B; cyan, Fe^II-C. The simulation parameters are listed in Extended Data Tables 2 and 3.

As the Mössbauer parameters of unreacted wild-type and ΔRS-cluster enzymes were indistinguishable (Extended Data Fig. 5), selective ⁵⁷Fe labelling experiments were performed to study the structural integrity of the RS and AUX clusters. To determine the degree to which the AUX cluster is susceptible to iron exchange, the ΔRS-cluster variant was overexpressed in natural-abundance Fe-enriched media and analysed via Mössbauer spectroscopy without previous reconstitution. A very weak quadrupole doublet signal typical of an [Fe₄S₄]²⁺ cluster (Extended Data Fig. 5e, red trace) was observed in the as-isolated enzyme before and after treatment with the chelating agent 1,10-phenanthroline (PT). This weak Mössbauer signal is due to the small amount (2.2%) of ⁵⁷Fe present at natural abundance. The Mössbauer spectrum was essentially unchanged after reconstitution with ⁵⁷Fe and treatment with PT, despite a small increase in the overall intensity (Extended Data Fig. 5e, green trace). A comparison of the iron content (Extended Data Fig. 5 caption) with changes in the Mössbauer signal intensity before and after ⁵⁷Fe reconstitution implied that less than 10% of the AUX cluster in the ΔRS-cluster variant was either exchanged or introduced into the enzyme from externally supplied iron during Fe reconstitution. Therefore, the majority of the AUX cluster does not appear to be susceptible to iron exchange, thereby facilitating the selective labelling of the AUX and RS clusters with ⁵⁶Fe versus ⁵⁷Fe.

Accordingly, wild-type AbmM was subsequently overexpressed in ⁵⁷Fe-enriched media and characterized without previous reconstitution. A Mössbauer quadrupole doublet typical of an [Fe₄S₄]²⁺ cluster (Extended Data Fig. 5f, orange trace) was observed and remained essentially unchanged upon incubation with PT (Extended Data Fig. 5f, blue trace), despite losing most of its activity (Supplementary Fig. 14). Consistent with the aforementioned controls using the ΔRS-cluster variant, these results indicated that the Mössbauer signal of ⁵⁷Fe-AbmM without reconstitution is mostly due to the AUX cluster and that iron extraction of the AUX cluster by PT is minimal. Therefore, wild-type AbmM was prepared containing a ⁵⁷Fe-enriched AUX cluster and a natural-abundance RS cluster (⁵⁷Fe-PT–⁵⁶Fe-AbmM, AUX-labelled) by overexpressing AbmM in ⁵⁷Fe-enriched media followed by treatment of the isolated protein with PT and then reconstitution with natural-abundance FeCl₃. The corresponding construct of ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled), which contains a natural-abundance AUX cluster and a ⁵⁷Fe-enriched RS cluster, was prepared in an analogous manner.

The ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) construct exhibited nearly identical Mössbauer spectra, indicative of a quadrupole doublet with parameters typical of an [Fe₄S₄]²⁺ cluster before and after incubation with DT. The latter spectrum showed less than 10% (relative to the total iron in the sample) [Fe₄S₄]⁺ cluster signal (Fig. 5c). In addition, the Mössbauer spectrum of reduced the ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) construct remained unchanged after incubation with SAM and CDP for 15 min before freeze-trapping (Fig. 5c,d and Extended Data Tables 2 and 3). These results indicated that the AUX cluster in AbmM cannot be easily reduced by DT and does not show detectable degradation during the enzymatic reaction. By contrast, spectra of the reduced ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) construct showed a greater mixture of ~29% [Fe₄S₄]²⁺ and 36% [Fe₄S₄]⁺ clusters (Fig. 5e and Extended Data Tables 2 and 3). Two additional mononuclear high-spin ferrous species (Fe^II-B and Fe^II-C; Extended Data Tables 2 and 3) accounted for an additional ~23% of the total iron in the sample, probably due to non-cluster-bound iron. Moreover, the [Fe₄S₄]⁺ species in reduced ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) was no longer discernible after incubation with SAM and CDP for 15 min before freeze-trapping, whereas the two mononuclear ferrous species increased to ~50% of the total iron (Fig. 5f and Extended Data Tables 2 and 3). These observations suggest that the AUX cluster remains unchanged during the enzymatic reaction, whereas the RS cluster is susceptible to both reduction by DT and degradation to mononuclear ferrous species after incubation with SAM and CDP. As shown in Extended Data Tables 2 and 3, the depletion of the reduced RS cluster (from 36% to undetectable) matches the increase in mononuclear ferrous ion (from 23% to 50%). The aforementioned Mössbauer results were further corroborated by EPR data (Supplementary Figs. 15 and 16), which showed that two slightly different S = ½ [Fe₄S₄]⁺ cluster signals—corresponding to the AUX-[Fe₄S₄]⁺ and the RS-[Fe₄S₄]⁺ clusters—were observed via EPR (g = [2.06, 1.92, 1.87] for AUX-[Fe₄S₄]⁺ and g = [2.02, 1.91, 1.90] for RS-[Fe₄S₄]⁺)²¹. The S = ½ [Fe₄S₄]⁺ signal from the RS-[Fe₄S₄]⁺ cluster was almost depleted upon the reaction of AbmM with CDP in the presence of SAM and DT.

Labelling experiments with ³⁴S

Labelling experiments with ³⁴S were also performed to narrow down the immediate sulfur donor during AbmM catalysis. Controls examining the susceptibility of the AUX cluster to sulfide exchange involved treatment of the as-isolated ΔRS-cluster variant with PT before the reconstitution with FeCl₃ and with Na₂³⁴S before a final round of PT treatment to remove possible FeS aggregates. When the resulting protein was denatured and worked up with mBBr, only 5% S-(bimane)₂ was found to contain ³⁴S (Extended Data Fig. 6a). This outcome indicates that the AUX cluster does not undergo extensive sulfide exchange during the FeS reconstitution protocol, which is consistent with the aforementioned Mössbauer results regarding AUX cluster stability and resistance to Fe exchange. Therefore, a solution of as-isolated wild-type AbmM expressed in the presence of natural-abundance sulfide was treated with PT and reconstituted with FeCl₃ and Na₂³⁴S to selectively enrich the RS cluster with ³⁴S (³²S-PT–³⁴S-AbmM, RS-labelled). Indeed, denaturation and derivatization of the enriched enzyme with mBBr demonstrated that approximately 45% of the resulting S-(bimane)₂ contained ³⁴S (Extended Data Fig. 6b), consistent with selective enrichment of only the RS cluster and not the AUX cluster. When the selectively enriched wild-type AbmM (³²S-PT–³⁴S-AbmM, RS-labelled) was subsequently incubated with 10 in the absence of additional sulfide, ESI-MS analysis of the product indicated the incorporation of 95% ³⁴S into the product (Extended Data Fig. 6c). These results are consistent with a mechanistic model where the RS cluster rather than the AUX cluster serves as the sulfide donor during AbmM catalysis. The substantial incorporation of ³⁴S (64%) into 14 after a 10 min incubation with Na₂³⁴S and reductants (DT or NADPH/MV) in the presence of DTT suggests that the sulfur atoms in the RS cluster can exchange with exogenous sulfide before sulfur insertion into CDP (Fig. 4a). The apparently rapid exchange is consistent with a previous report that the RS cluster in the DT-treated twitch radical SAM enzyme BtrN undergoes rapid exchange with exogenous Fe, whereas the AUX cluster remains inert³³.

Discussion

In this Article we have demonstrated that AbmM is a twitch radical SAM enzyme that catalyses the oxidation of CDP/CTP to the 4′-hydroxy-4′-thionucleotide 13. Our mechanistic studies indicate that this reaction proceeds via hydrogen-atom abstraction from C4′, which subsequently induces cleavage of the C1′–O bond and the addition of sulfur to C1′. Subsequent cyclization generates the sulfur-bridged ribose, effectively swapping oxygen for sulfur in the furanose ring. This chemistry differs from that of other radical SAM-catalysed thiolation reactions, which typically involve sulfur insertion into C–H bonds^22,24,27. Incorporation of sulfur leads to the deactivation of AbmM with minimal reactivation in the presence of Na₂S, which suggests that free sulfide is not an immediate sulfur donor and must first be more tightly integrated into the structure of AbmM before it can participate in the reaction. Moreover, the observation of improved catalytic turnover with external FeS cluster carrier proteins implies that the sulfur donor is a component of an FeS cluster bound by AbmM. However, the AUX cluster in AbmM remains stable during turnover. This stands in contrast to the RS cluster, which both undergoes decomposition during turnover—observable via Mössbauer spectroscopy—and appears to be the source of the incorporated sulfur according to selective ³⁴S labelling of the RS- and AUX-[Fe₄S₄] clusters in AbmM.

These findings indicate a model in which the RS cluster serves in some manner as the immediate sulfur donor (Fig. 6). This chemistry would distinguish AbmM catalysis from that of BioB^23,24 and LipA^22,25–27, which instead rely on an auxiliary [Fe₂S₂] or [Fe₄S₄] cluster as the sulfur donor. Unlike the radical SAM enzymes BioB and LipA, AbmM is a twitch radical SAM enzyme, for which the AUX-[Fe₄S₄] cluster is typically situated quite distant from the site of initial substrate radical generation and is stabilized via by four cysteine ligands, features which appear to be recapitulated in the partial crystal structure of AbmM. Thus, the AUX cluster may not be a suitable donor in the active sites of twitch radical SAM enzymes, in contrast to BioB and LipA. Nevertheless, substituting any one of the four cysteine residues coordinating the AUX cluster with alanine resulted in protein insolubility, suggesting that the AUX cluster plays a structural role. Furthermore, AUX clusters in twitch radical SAM enzymes are typically difficult to reduce, as exemplified by BtrN³⁴; however, EPR analysis indicated that the AUX cluster in AbmM can be partially reduced by DT, raising the possibility that it could participate in redox processes during catalysis, a hypothesis that requires further investigation.

Fig. 6 |. Proposed mechanism of AbmM.

In this mechanism, the RS cluster in AbmM serves both to mediate the reductive cleavage of SAM and to donate the incorporated sulfur.

Exactly how the RS cluster participates in the radical thiolation step likewise remains to be fully clarified. In particular, the reduction products of SAM (that is, 5′-dA and methionine) would presumably block access to the RS cluster by the substrate radical. This may require an unanticipated disposition of the substrate radical relative to the RS-[Fe₄S₄] cluster, an FeS cluster with an additional sulfur ligand or decomposition of the RS cluster to liberate free sulfide before the thiolation step, possibly induced by methionine dissociation. In any case, entry into the albomycin δ₂ biosynthetic pathway involves a novel radical-mediated swapping of the bridging furanose oxygen for sulfur and may have broad implications in the biosynthesis of sulfur-containing natural products^15,35–37.

Methods

Materials and general methods

Streptomyces sp. strain ATCC 700974 was obtained from the American Type Culture Collection. E. coli DH5α, from Bethesda Research Laboratories, was used for routine cloning procedures³⁸. The protein overexpression host E. coli BL21 Star (DE3) was obtained from Invitrogen. Vector pET28b(+) for protein overexpression was purchased from Novagen. Vector pYH7 used for the gene inactivation experiments was a gift from Y. Sun at Wuhan University³⁹. Methylation-deficient E. coli ET12567/pUZ8002 was used for the intergeneric conjugation⁴⁰. Plasmid pIB139 is a Streptomyces–E. coli shuttle vector that can site-specifically integrate into Streptomyces chromosomes⁴¹. The vector pSUMO was kindly provided by S. Booker at The Pennsylvania State University.

All chemicals and reagents were purchased from Sigma-Aldrich or Fisher Scientific and were used without further purification unless otherwise specified. Standard genetic manipulations of E. coli were performed as described by Sambrook and Russell³⁸. DNA sequencing was performed at the core facility of the Institute of Cellular and Molecular Biology, the University of Texas at Austin. Tetrahydrofuran was distilled from sodium/benzophenone, and dichloromethane was distilled from calcium hydride under an argon atmosphere. Oligonucleotide primers were prepared by Integrated DNA Technologies. Kits for DNA gel extraction and spin minipreps are products of Qiagen. The PureLink Genomic DNA Mini Kit used was acquired from Invitrogen. Enzymes and molecular weight standards used in the cloning experiments were obtained from New England Biolabs. KAPA HiFi DNA polymerase was purchased from KAPA Biosystems. Reagents for SDS–PAGE were ordered from Bio-Rad. Sterile syringe filters were bought from Fisher Scientific. Amicon YM-10 ultrafiltration membranes are products of Millipore. Silica gel column chromatography was carried out using SiliaFlash P60 (230–400 mesh, SiliCycle). NaBD₄ (99% D), elemental ³⁴S (99.79% ³⁴S) and [¹⁸O]water (97% ¹⁸O) were obtained from Cambridge Isotope Laboratories. d-[3-²H]Ribose (99% D) and d-[4-²H]ribose (99% D) were purchased from Omicron Biochemicals.

DNA concentrations were measured using a NanoDrop ND-1000 UV-vis instrument from Thermo Fisher Scientific. HPLC was performed using a Beckman System Gold 125 Solvent Module with a 166 detector or an Agilent Technologies 1260 Infinity HPLC System with a diode array multiple wavelength UV–vis absorbance detector (G1315D, Agilent Technologies) using a C18 reversed-phase column (Microsorb 100–5 C18 4.6 × 250 mm, Agilent Technologies) unless otherwise noted. LC-ESI-TOFMS analysis was performed using an Agilent Technologies HPLC system equipped with a pump (G1311C), an auto sampler (G1329B) and a time-of-flight mass spectrometer (G6230B) with an electrospray ionization source. LCMS separations were performed using a Poroshell 120 EC-C18 column (2.7 μm, 4.6 × 100 mm) along with an Eclipse plus C18 guard column (1.8 μm, 2.1 × 5 mm) at a flow rate of 0.4 ml min⁻¹ using 0.1% formic acid in H₂O (solvent A) and acetonitrile (solvent B). The elution was conducted using the following gradient programme unless otherwise specified: 0–10 min, 0–30% B; 10–12 min, 30–100% B; 12–13 min, 100–0% B; 13–18 min, 0% B. The obtained LCMS data were analysed using MassHunter software (Agilent Technologies). NMR spectra were recorded using a Varian DirectDrive 600 MHz spectrometer, a Bruker Avance III HD 500 MHz NMR spectrometer equipped with a CryoProbe Prodigy or a Varian DirectDrive 400 MHz spectrometer at the Nuclear Magnetic Resonance Facility at the University of Texas at Austin. MestRe Nova⁴² was used to analyse the NMR spectra. Deuterated solvents were used as internal standards in the NMR spectra unless stated otherwise. Chemical shifts are reported as parts per million (ppm) relative to those of CDCl₃, 7.26 ppm for ¹H NMR and 77.16 ppm for ¹³C NMR, respectively. Anaerobic work was performed in an anaerobic glovebox (Coy Laboratory Products) under an atmosphere of >98% N₂, ~2% H₂ and <25 ppm O₂.

Cloning of sufU and abmM

The sufU gene was amplified via polymerase chain reaction (PCR) from the genomic DNA of Streptomyces sp. ATCC 700974 using primers sufU-F and sufU-R shown in Supplementary Table 3. The PCR-amplified gene fragment was purified using a Gel Extraction Kit from Qiagen, digested with NdeI and XhoI restriction enxymes, and ligated into pET28b(+) vector, which had also been digested using the same restriction enzymes. The resulting plasmid sufU/pET28b(+) was sequenced using the T7 or T7-terminal universal primers. The abmM gene was PCR-amplified from the genomic DNA of Streptomyces sp. ATCC 700974 using primers abmM-F and abmM-R listed in Supplementary Table 3. The PCR-amplified gene fragment was purified using the Gel Extraction Kit from Qiagen and inserted into the pSUMO vector after digestion with BsaI and XhoI restriction enzymes. The procedures were based on the cloning method using SLiCE (Seamless Ligation Cloning Extract)⁴³. The resulting plasmid abmM/pSUMO was sequenced using the T7 or T7-terminal universal primers. The plasmids abmM/pET28b(+) and abmM-ΔRS/pET28b(+) used for expression of N-His₆-AbmM and N-His₆-AbmM-ΔRS were constructed using primers abmM-pET28b(+)-F and abmM-pET28b(+)-R listed in Supplementary Table 3.

Expression and purification of SufU

The plasmid sufU/pET28b(+) was used to transform the E. coli BL21 Star (DE3) strain for protein expression. An overnight culture of E. coli BL21 Star (DE3)-sufU/pET28b(+) grown in the LB medium (10 ml) containing kanamycin (50 μg ml⁻¹) was used to inoculate one litre of the same growth medium. The culture was incubated at 37 °C with shaking (200 revolutions per min (r.p.m.)) until the absorbance at 600 nm (A₆₀₀) reached ~0.6. Protein expression was then induced by the addition of isopropyl β-d-1-thiogalactopyranoside to a final concentration of 0.1 mM, and the cells were allowed to grow at 18 °C and 120 r.p.m. for an additional 20 h. The cells were collected by centrifugation at 4,000 × g for 10 min and stored at −80 °C until lysis. All purification steps were carried out at 4 °C. The thawed cells were re-suspended in 50 mM HEPES buffer (pH 7.5) containing 10% (v/v) glycerol, 10 mM imidazole and 300 mM NaCl. The cells were disrupted by sonication using ten pulses of 10 s each with a 20 s cooling pause between each pulse. The resulting lysate was centrifuged at 20,000 × g for 20 min, and the supernatant was mixed with Ni-nitrilotriacetic acid (NTA) resin. Bound protein was eluted using 50 mM HEPES buffer (pH 7.5) buffer containing 10% (v/v) glycerol, 250 mM imidazole and 300 mM NaCl. The collected protein solution was dialysed against 50 mM HEPES buffer (pH 7.5) (3 × 1 l) containing 300 mM NaCl and 10% (v/v) glycerol. The protein solution was then flash-frozen in liquid nitrogen and stored at −80 °C until use (Supplementary Fig. 1).

Expression and purification of AbmM

N-His₆-SUMO-AbmM was also expressed and purified on the basis of the same procedure described for SufU. After purification by Ni-NTA resin and dialysis against 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol, the SUMO-tagged AbmM was treated with SUMO protease Ulp1_403–621 (prepared as previously described^44,45, 0.87 μM) in the presence of DTT (1 mM) at 4 °C for 6 h. The protein solution was loaded onto a column filled with Ni-NTA resin that had been pre-equilibrated with 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol. The desired native AbmM appeared as a brown-coloured band on the column and was eluted with 50 mM HEPES buffer (pH 7.5) containing 10% (v/v) glycerol, 10 mM imidazole and 300 mM NaCl. The collected protein solution was dialysed against 50 mM HEPES buffer (pH 7.5) (3 × 1 l) containing 300 mM NaCl and 10% glycerol. Iron and sulfur bound to AbmM were quantified on the basis of the reported methods^46,47. The protein solution was concentrated to approximately 170 μM, as determined using the NanoDrop instrument, flash-frozen in liquid nitrogen and stored at −80 °C until use (Extended Data Fig. 1e).

Typical AbmM assay conditions and analysis

CDP (10, 0.2 or 0.4 mM) was incubated with as-purified AbmM (34 μM), Na₂S (2 mM), SAM (0.6 mM), NADPH (1 mM), MV (1 mM) and DTT (10 mM) in HEPES buffer (40 mM, pH 7.5) at room temperature under an anaerobic atmosphere. Assays were conducted using AbmM without iron and sulfur reconstitution unless otherwise noted. The AbmM reaction was also performed using DT (2 mM) instead of NADPH and MV. The reaction volume was typically 50–100 μl. After the incubation, the reaction mixture was removed from the glovebox and filtered through a YM-10 membrane filter. The filtrate was then analysed using LCMS or HPLC with UV absorbance detection. For the LCMS analysis, the filtrate (10 μl) was diluted in H₂O (400 μl), and a 10 μl volume of the diluted sample was subjected to LCMS analysis using the method described above. For the HPLC-UV analysis, the filtrate (100 μl) was injected onto an anion exchange column (Dionex CarboPac PA1, 4 × 250 mm) and the column was eluted using H₂O as mobile phase A and 1 M ammonium acetate in H₂O as mobile phase B at a flow rate of 1 ml min⁻¹ with the following gradient programme: 0–33 min, 0–90% B; 33–36 min, 90% B; 36–37 min, 90–0% B; 37–42 min, 0% B. Elution was monitored by setting the UV detector at 271 nm (Fig. 2). Analysis of the AbmM reaction was also performed after the incubation mixture was treated with CIP (0.05 U μl⁻¹) for 1 h. The mixture was filtered using YM-10 centrifugal filtration to remove protein and analysed by LCMS as described above. Methods for the isolation and characterization of 14 are described in Supplementary Information.

Expression and purification of AbmM and the ΔRS-cluster variant for Mössbauer and ³⁴S cluster labelling experiments

The plasmid abmM/pET28b(+) and the plasmid pDB1282⁴⁸ were used to co-transform the E. coli BL21 Star (DE3) strain for protein expression. An overnight culture of E. coli BL21 Star (DE3)-abmM/pET28b(+)-pDB1282 grown in the LB medium (5 ml) containing kanamycin (50 μg ml⁻¹) and ampicillin (100 μg ml⁻¹) was used to inoculate of M9 medium (1 l) with the identical antibiotics. The culture was incubated at 37 °C with shaking (200 r.p.m.) until A₆₀₀ reached ~0.3. l-Arabinose (2 g) was added to the culture, which was then incubated further until A₆₀₀ reached ~0.6. The culture was cooled on ice for 1 h. Protein expression was then induced by the addition of isopropyl β-d-1-thiogalactopyranoside and FeCl₃ to a final concentration of 0.1 mM and 50 μM, respectively. The cells were allowed to grow at 18 °C and 120 r.p.m. for an additional 20 h. The purification was carried out in the same way as described for NfuA. The N-His₆-AbmM-ΔRS cluster was expressed and purified using the same procedure as described for N-His₆-AbmM. For ⁵⁷Fe-enriched N-His₆-AbmM and the N-His₆-AbmM-ΔRS cluster, regular FeCl₃ was replaced with ⁵⁷FeCl₃.

Reconstitution of N-His₆-AbmM and the ΔRS-cluster variant for Mössbauer analysis

All of the procedures are performed anaerobically. A solution of 100 μM as-isolated ⁵⁷Fe-enriched N-His₆-AbmM in 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol in a total volume of 3 ml was chilled on ice packs and equilibrated with the anaerobic atmosphere for 1 h. DTT was added to a final concentration of 5 mM. After incubation for 1 h, a solution of ⁵⁷FeCl₃ was added over 20 min to a final concentration of 0.5 mM. This was followed by the addition of a solution of Na₂S over 20 min to a final concentration of 0.5 mM. The resulting mixture was incubated for 3 h before desalting with an Econo-Pac 10DG Desalting Prepacked Gravity Flow Column (PD-10 column, Bio-Rad) using 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol as the elution buffer. The brown-coloured protein eluate was concentrated using an Amicon Ultra-0.5 centrifugal filter unit with a molecular weight cut-off of 10 kDa (Millipore). Reconstitution of the ⁵⁷Fe-enriched ΔRS-cluster variant was carried out in the same way as described above.

Selective ⁵⁷Fe labelling experiments on the AbmM-ΔRS cluster

All of the procedures were conducted anaerobically. The buffer system was 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol. A solution of 135 μM as-isolated ΔRS-cluster variant was incubated with 2 mM PT and 2 mM DT on ice packs overnight. After desalting through a PD-10 column, the protein eluate was obtained as a ⁵⁶Fe-PT-AbmM-ΔRS cluster. A solution of 105 μM ⁵⁶Fe-PT-AbmM-ΔRS cluster was reconstituted with 5 mM DTT, 0.84 mM ⁵⁷FeCl₃ and 0.84 mM Na₂S. After desalting through the PD-10 column, the obtained protein eluate was concentrated as the ⁵⁶Fe-PT–⁵⁷Fe-AbmM-ΔRS cluster. Another round of PT treatment was conducted on the ⁵⁶Fe-PT–⁵⁷Fe-AbmM-ΔRS cluster to remove excess FeS aggregates, where a solution of 178 μM ⁵⁶Fe-PT–⁵⁷Fe-AbmM-ΔRS cluster was incubated with 3 mM PT and 3 mM DT on ice packs overnight. After desalting through the PD-10 column, the obtained protein eluate was concentrated as the ⁵⁶Fe-PT–⁵⁷Fe-PT-AbmM-ΔRS cluster. The iron content analysis is shown in caption of Extended Data Fig. 5.

Selective ⁵⁷Fe labelling experiments on wild-type AbmM

All of the procedures were conducted anaerobically. The buffer system was 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol. A solution of 118 μM as-isolated ⁵⁷Fe-enriched N-His₆-AbmM was incubated with 2 mM PT and 2 mM DT on ice packs overnight. After desalting through the PD-10 column, the protein eluate was obtained as ⁵⁷Fe-PT-AbmM. A solution of 76 μM ⁵⁷Fe-PT-AbmM was reconstituted with 5 mM DTT, 0.3 mM regular FeCl₃ and 0.3 mM Na₂S. After desalting through the PD-10 column, the obtained protein eluate was concentrated as ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled). A solution of 112 μM as-isolated N-His₆-AbmM was incubated with 2 mM PT and 2 mM DT on ice packs overnight. After desalting through the PD-10 column, the protein eluate was obtained as ⁵⁶Fe-PT-AbmM. A solution of 77 μM ⁵⁶Fe-PT-AbmM was reconstituted with 5 mM DTT, 0.31 mM ⁵⁷FeCl₃ and 0.31 mM Na₂S. After desalting through the PD-10 column, the obtained protein eluate was concentrated as ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled).

Preparation of Mössbauer samples

All of the procedures were conducted anaerobically. Each sample was in a total volume of 300 μl, in 50 mM HEPES buffer (pH 7.5) containing 300 mM NaCl and 10% (v/v) glycerol. For the 0.3 mM DT-treated ΔRS-cluster variant (Extended Data Fig. 5c), a solution of 0.3 mM reconstituted N-His₆-tagged AbmM-ΔRS cluster was incubated with 10 mM DTT and 2 mM DT for 10 min. For 0.2 mM DT-treated AbmM (Extended Data Fig. 5d), a solution of 0.2 mM reconstituted N-His₆-tagged AbmM was incubated with 10 mM DTT and 2 mM DT for 10 min. For Fig. 5a, the sample is the same as that in Extended Data Fig. 5d. For Fig. 5b, a solution of 0.2 mM reconstituted N-His₆-tagged AbmM was incubated with 10 mM DTT, 2 mM DT, 0.3 mM SAM and 0.2 mM CDP for 10 min. For Fig. 5c, a solution of 0.15 mM ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) was incubated with 4 mM DT for 15 min. For Fig. 5d, a solution of 0.15 mM ⁵⁷Fe-PT–⁵⁶Fe-AbmM (AUX-labelled) was incubated with 1 mM CDP, 1.5 mM SAM and 4 mM DT for 15 min. For Fig. 5e, a solution of 0.15 mM ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) was incubated with 4 mM DT for 15 min. For Fig. 5f, a solution of 0.15 mM ⁵⁶Fe-PT–⁵⁷Fe-AbmM (RS-labelled) was incubated with 1 mM CDP, 1.5 mM SAM and 4 mM DT for 15 min.

Mössbauer spectroscopy

Mössbauer spectra were recorded using constant acceleration Mössbauer spectrometers equipped with a Janis SVT-400 variable-temperature cryostat (weak-field). Isomer shift (δ) values are reported with respect to the centroid of the spectrum of α-iron metal at room temperature. Simulations of Mössbauer spectra were carried out using the program WMOSS (SEE Co.) and SpinCount (M. P. Hendrich, Carnegie Mellon University)⁴⁹. For simulations of diamagnetic species, the spin Hamiltonian formalism was invoked according to equation (1) shown below, in which the first term describes the interaction between the electric field gradient and the nuclear quadrupole moment, and the last term represents the nuclear Zeeman splitting of ⁵⁷Fe nuclei. All symbols have their usual meanings. For the simulation of paramagnetic species arising from the [Fe₄S₄]⁺ clusters, the additional spin Hamiltonian terms were considered according to equation (2) listed below, where the first term describes the electronic Zeeman interactions and the second term describes the ⁵⁷Fe electron–nuclear hyperfine interactions⁵⁰.

$$H_{\text{n}}=\sum _i\left\{\frac{\text{e}Q{V}_{zz,i}}{4I_i(2I_i-1)}\left[3I_{z,i}^2-I_i(I_i+1)+{\eta }_i(I_{x,i}^2-I_{y,i}^2)\right]-g_{\text{n}}{\beta }_{\text{n}}{\mathbf{I}}_i\cdot \mathbf{B}\right\},$$ (1)

where H_n represents nuclear spin Hamiltonian; e is the elementary charge; Q is the nuclear quadrupole moment; I_i is the nuclear spin quantum number of the ith nucleus (for ⁵⁷Fe nuclei, I = 3/2); V_xx,i, V_yy,i and V_zz,i are the principal components of the electric field gradient tensor at the ith nucleus, V_i, following the convention |V_xx,i| ≤ |V_yy,i| ≤ |V_zz,i|; η_i is the asymmetry parameter of V_i, defined as ${\eta }_i=\frac{V_{xx,i}-V_{yy,i}}{V_{zz,i}};{\mathbf{I}}_i$ is the total nuclear spin operator of the ith nucleus, with Cartesian components I_x,i, I_y,i and I_z,i defined in the same principal axis frame; g_n is the nuclear g factor; β_n is the nuclear magneton; and B is the external magnetic field vector.

$$H_{\text{e}}=\beta \mathbf{B}\cdot g\cdot \mathbf{S}+\mathbf{S}\cdot A_i\cdot {\mathbf{I}}_i$$ (2)

where H_e represents electron spin Hamiltonian, β is the Bohr magneton, B is the external magnetic field vector, g is the electron g tensor, S is the total electron spin operator, A_i is the hyperfine coupling tensor between the electron spin and the nuclear spin for the ith nucleus, and I_i is the total nuclear spin operator of the ith nucleus.

Extended Data

Extended Data Fig. 1 |. Gene deletion of abmM and analysis of purified AbmM.

(a) LCMS analysis of the culture of the ΔabmM deletion mutant showing extracted ion chromatogram (EIC) traces corresponding to the [M + H]⁺ signal from albomycin δ₂ (1) (m/z = 1046.3). (b) LCMS analysis of the culture of the ΔabmM deletion mutant showing EIC traces corresponding to the [M + H]⁺ signal from SB-217452 (4) (m/z = 477.1). (c) LCMS analysis of the culture of the ΔabmM strain complemented with abmM showing EIC traces corresponding to the [M + H]⁺ signal from albomycin δ₂ (1) (m/z = 1046.3). (d) LCMS analysis of the culture of the ΔabmM strain complemented with abmM showing EIC traces corresponding to the [M + H]⁺ signal from SB-217452 (4) (m/z = 477.1). (e) SDS-PAGE of AbmM purification. Lane 1: molecular weight markers; Lane 2: purified N-His₆-SUMO-AbmM (48.2 kDa); Lane 3, digestion of N-His₆-SUMO-AbmM; Lane 4, purified AbmM without the tag (36.1 kDa). This experiment was repeated independently three times with similar results. (f) UV-vis absorption spectra of AbmM. Solid line, 15 μM as-isolated AbmM; dotted line, 15 μM as-isolated AbmM after treatment with 200 μM sodium dithionite.

Extended Data Fig. 2 |. Studies of the AbmM reaction.

(a) Detection of l-methionine by LCMS. EIC traces corresponding to the [M + H]⁺ signal of l-methionine (m/z = 150.0583) are shown. A small amount of l-methionine was detected from the reaction mixture without AbmM or dithionite likely due to nonenzymatic degradation of SAM. (b) LCMS analysis of the AbmM reaction mixture. The [M – H]⁻ signal at m/z 402.0130 arises from CDP (10, calc for C₉H₁₄N₃O₁₁P₂⁻, 402.0109). The [M – H]⁻ signal at m/z 433.9849 is consistent with the mass of 13 (calc. for C₉H₁₄N₃O₁₁P₂S⁻, 433.9830). (c) LCMS analysis of the AbmM reaction carried out under different conditions. The analysis was conducted after the incubation mixture was treated with CIP. Three EIC traces (m/z = 276.06, 244.09, and 252.2, corresponding to the [M + H]⁺ signals from 14, 8, and 5′-deoxyadenosine, respectively) are overlaid. (d) LCMS analysis of the AbmM reaction with different substrates such as cytidine, CMP, CDP and CTP (8-11). Prior to the analysis, the incubation mixture was treated with CIP. Three EIC traces (m/z = 276.06, 244.09, and 252.2, corresponding to the [M + H]⁺ signals from 14, 8, and 5′-deoxyadenosine, respectively) are overlaid. Other tested compounds, cytidine 2′:3′-cyclic monophosphate (2′:3′-cyclic CMP), cytidine 3′:5′-cyclic monophosphate (3′:5′-cyclic CMP), 2′-deoxycytidine (2′-dC), 2′-deoxycytidine 5′-monophosphate (2′-dCMP), 2′-deoxycytidine 5′-diphosphate (2′-dCDP), deoxycytidine 5′-triphosphate (2′-dCTP), adenosine 5′-diphosphate (ADP), adenosine 5′-triphosphate (ATP), guanosine 5′-diphosphate (GDP), guanosine 5′-triphosphate (GTP), thymidine 5′-diphosphate (TDP), thymidine 5′-triphosphate (TTP), uridine 5′-diphosphate (UDP), and uridine 5′-triphosphate (UTP), were not accepted by AbmM as judged by no formation of 5′-deoxyadenosine (data not shown). (e) HPLC analysis of purified 14.

Extended Data Fig. 3 |. AbmM product and deuterium labelling experiments.

(a) and (b) LCMS analysis of the AbmM reaction mixture treated with CIP/mBBr or CIP/mBBr/NaBH₄. EIC traces corresponding to the [M + H]⁺ signal from each species are overlaid (m/z 466.14 for 14-bimane and m/z 468.16 for 15). In (b): Bottom trace: ESI-MS analysis of 14-bimane generated from the AbmM product upon treatment with CIP/mBBr (calc for C₁₉H₂₄N₅O₇S⁺, 466.1391). Middle trace: ESI-MS analysis of 15 generated from the AbmM product upon treatment with CIP/mBBr/NaBH₄ (calc for C₁₉H₂₆N₅O₇S⁺, 468.1547). Top trace: ESI-MS analysis of [4′-²H]-15 generated from the AbmM product upon treatment with CIP/mBBr/NaBD₄ (calc for C₁₉H₂₅DN₅O₇S⁺, 469.1610). (c) ESI-MS analysis of 5′-dA and (d) 14 generated in the AbmM reaction using deuterated CDPs as substrates. The analysis was conducted after the incubation mixture was treated with CIP (5′-dA, calc for C₁₀H₁₄N₅O₃⁺, 252.1091; 14, calc for C₉H₁₄N₃O₅S⁺, 276.0649; [3′-²H]-14, calc for C₉H₁₃DN₃O₅S⁺, 277.0711; [5′,5′-²H₂]-14, calc for C₉H₁₂D₂N₃O₅S⁺, 278.0774). (e) ESI-MS analysis of 14 (calc for C₉H₁₄N₃O₅S⁺, 276.0649) generated in the AbmM reaction in H₂¹⁸O (52% ¹⁸O). The analysis was conducted after the incubation mixture was treated with CIP.

Extended Data Fig. 4 |. Structural analysis of AbmM.

(a) Overall dimeric structure of AbmM shown in purple. (b) Monomer of AbmM. (c) C-terminal AUX cluster. (d) Overlay of the crystal and predicted structures of AbmM. The predicted structure is shown in grey. (e) The RS cluster binding loops of AbmM (crystal and predicted structures) and MoaA (shown in blue, PDB 2FB3). (f) Distance between the docked RS cluster and AUX cluster in AbmM.

Extended Data Fig. 5 |. ⁵⁷Fe Mössbauer spectra of selected AbmM samples.

(a) 0.3 mM reconstituted ΔRS-cluster variant; (b) 0.2 mM reconstituted AbmM; (c) 0.3 mM DT-treated ΔRS-cluster variant; (d) 0.2 mM DT-treated AbmM; (e) 0.3 mM ⁵⁶Fe-PT-AbmM-ΔRS cluster (red) and 0.3 mM ⁵⁶Fe-PT-⁵⁷Fe-PT-AbmM-ΔRS cluster (green); (f) 0.15 mM as-isolated ⁵⁷Fe-AbmM (orange), 0.15 mM ⁵⁷Fe-PT-AbmM (blue), and 0.15 mM ⁵⁷Fe-PT-⁵⁶Fe-AbmM (AUX-labelled) (dark green). The measurement conditions are indicated with external magnetic fields parallel to the γ-radiation. Bars show experimental data with statistical uncertainty of the data (the length of the vertical bars), red solid lines represent overall simulations. Simulation parameters are listed in Extended Data Table 2. The sample of ⁵⁷Fe-PT-AbmM was prepared by overexpressing AbmM in media enriched with ⁵⁷Fe and subsequently treated with PT before spectroscopic characterization. The sample of ⁵⁶Fe-PT-⁵⁷Fe-PT-AbmM-ΔRS cluster was prepared by overexpressing AbmM ΔRS-cluster variant in media containing natural abundance Fe, followed by treatment of the isolated protein with PT and then reconstitution with ⁵⁷FeCl₃. The resulting sample was further treated with PT. Other combinations such as ⁵⁶Fe-PT-AbmM-ΔRS cluster were prepared in an analogous manner. The Fe content analysis yielded (Fe/protein monomer) 2.4 for as-isolated AbmM-ΔRS cluster, 2.4 for ⁵⁶Fe-PT-AbmM-ΔRS cluster, 6.0 for ⁵⁶Fe-PT-⁵⁷Fe-AbmM-ΔRS cluster, and 2.7 for ⁵⁶Fe-PT-⁵⁷Fe-PT-AbmM-ΔRS cluster.

Extended Data Fig. 6 |. ³⁴S labelling experiments.

(a) ESI-MS spectrum of S-(bimane)₂ obtained by reacting mBBr with the denatured ³⁴S-reconstituted ΔRS-cluster AbmM variant after another round of PT treatment. (b) ESI-MS spectrum of S-(bimane)₂ obtained by reacting mBBr with the denatured ³⁴S-reconstituted wildtype AbmM (³²S-PT-³⁴S-AbmM, RS-labelled). (c) ESI-MS spectrum of 14 generated from the single-turnover reaction of 10 with ³⁴S-reconstituted wildtype AbmM (³²S-PT-³⁴S-AbmM, RS-labelled). The ³⁴S enrichment was determined after correcting for natural abundance of ³⁴S.

Extended Data Table 1 |.

Data collection and refinement statistics

	AbmM
Data collection a
Space group	P 21 21 21
Cell dimensions
a, b, c (Å)	70.79, 81.03, 131.84
β (°)	90.0, 90.0, 90.0
Resolution (Å)	48.24–2.5 (2.6–2.5)b
Rmerge (%)	8.4 (84.6)b
//σ/	14.2 (2.1)b
Completeness (%)	99.9 (99.9)b
CC1/2	0.998 (0.790)
Redundancy	6.7 (7.0)b
Refinement
Resolution (Å)	48.2–2.5
No. reflections	26864
Rwork/Rfree	22.9/27.9
No. atoms
Protein	4545
Ligand/ion	56
Water	65
B-factors
Protein	54.0
Ligand/ion	62.7
Water	46.0
R.m.s. deviations
Bond lengths (Å)	0.002
Bond angles (°)	0.479
Data access
PDB	8WM2

^aData were collected from one crystal.

^bValues in parentheses are for highest-resolution shell.

Extended Data Table 2 |.

Mössbauer parameters of ⁵⁷Fe sites employed in the simulations

Site	S	δ (mm/s)	|ΔEQ| (mm/s)	η	A/gnβn (T)	Note
I	0	0.45(2)b	1.16(2)b,c	-	-	[Fe4S4]2+
II	½	0.50a	+1.32a	0.8a	[−23.1, −23.7, −20.2]a	[Fe4S4]+, the FeII/FeIII pair
III	½	0.58a	+1.89a	0.32a	[+19.3, +9.8, +6.3]a	[Fe4S4]+, the FeII/FeII pair
IV	2	0.80(1)b	3.30(2)b,c	-	-	Mononuclear FeII_A
V	2	1.15(1)b	2.19(2)b,c	-	-	Mononuclear FeII_B
VI	2	1.20(1)b	3.12(2)b,c	-	-	Mononuclear FeII_C

^aThe spectral simulation parameters for the [Fe₄S₄]⁺ cluster are adapted from ref. 51.

^bThe values included in the parentheses indicate the uncertainty of the simulation parameters.

^cThe signs of |ΔE_Q| are not determined.

Extended Data Table 3 |.

Iron species contents in Mössbauer simulations

Sample	[Fe4S4]2+	[Fe4S4]+	FeII_A	FeII_B	FeII_C
Reconstituted 57Fe-ΔRS-cluster variant	90%a	-	-	-	-
Reconstituted 57Fe AbmM	90%a	-	-	-	-
DT-treated 57Fe-ARS-cluster variants	64%a	36%b	-	-	-
DT-treated 57Fe-AbmM	64%a	30%b	≤ 5%	-	-
Reacted 57Fe-AbmM	65%a	18%b	17%	-	-
DT-treated 57Fe-PT-56Fe-AbmM (AUX-labelled)	90%	-	-	-	-
Reacted 57Fe-PT-56Fe-AbmM (AUX-labelled)	93%	-	-	-	-
DT-treated 56Fe-PT-57Fe-AbmM (RS-labelled)	29%	36%b	-	12%	11%
Reacted 56Fe-PT-57Fe-AbmM (RS-labelled)	34%	-	-	23%	27%

^aTwo doublets within uncertainty or asymmetric doublet are employed.

^bSite II and III are fixed to 1:1 ratio. The best simulation requires the use of the slightly different parameters (δ = 0.75 mm/s, |ΔE_Q | = 3.16 mm/s) from the listed parameters for Fe^II_A site in Extended Data Table 2.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41929-025-01367-w.

Data availability

The data that support the findings of this study are available within the article and its Supplementary Information. The coordinates and the structure factor amplitudes for AbmM have been deposited at the Protein Data Bank with accession code 8WM2. Source data are provided with this paper.