A new role for radical SAM enzymes in the biosynthesis of thio(seleno)oxazole RiPP natural products

Abstract

Ribosomally synthesized post-translationally modified peptides (RiPPs) are ubiquitous and represent a structurally diverse class of natural products. The ribosomally encoded precursor polypeptides are often extensively modified post-translationally by enzymes that are encoded by co-clustered genes. Radical S-adenosyl-l-methionine (SAM) enzymes catalyze numerous chemically challenging transformations. In RiPP biosynthetic pathways, these transformations include the formation of C-H, C-C, C-S, and C-O linkages. In this manuscript, we show that the Geobacter lovleyi sbtM gene encodes a radical SAM protein, SbtM, which catalyzes the cyclization of a Cys/SeCys residue in a minimal peptide substrate. Biochemical studies of this transformation support a mechanism involving H-atom abstraction at the C-3 of the substrate Cys to initiate the chemistry. Several possible cyclization products were considered. The collective biochemical, spectroscopic, mass spectral, and computational observations point to a thiooxazole as the product of the SbtM catalyzed modification. To our knowledge, this is the first example of a radical SAM enzyme that catalyzes transformation involving a SeCys containing peptide, and represents a new paradigm for formation of oxazole-containing RiPP natural products.

Graphical Abstract

Ribosomally synthesized post-translationally modified peptides (RiPPs) are a structurally diverse class of natural products. These peptides are synthesized on the ribosome and are subsequently post-translationally modified to build-in functional and biological diversity^1,2. Over the last dozen years, radical S-adenosyl-l-methionine (SAM) enzymes have emerged as being involved in many chemically challenging reactions in pathways that produce RiPP natural products³. Members of the radical SAM superfamily are generally identified on the basis of a conserved CX₃CX₂C motif, whose thiolates are responsible for the coordination of the three iron corners of a catalytic [4Fe-4S] metallocluster. The fourth iron of this cluster coordinates the α-amino and α-carboxylate moieties of SAM, binding the cofactor to the active site^4–6. Radical SAM enzymes require a reductant, such as dithionite (dT), to reduce the [4Fe-4S] cluster from a catalytically inactive +2 state to a catalytically active +1 state. The reduced [4Fe-4S] cluster cleaves SAM, generating a 5’-deoxyadenosyl radical (dAdo•), which initiates the transformation of the substrate, often by H-atom abstraction^7,8 (Fig. 1A). Although SAM is sometimes reformed during the catalytic cycle by some, in many radical SAM enzymes, it is used stoichiometrically⁹.

Figure 1. Radical SAM Enzymes and RiPP maturation.

(A) The reductive cleavage of SAM to generate a 5’-deoxyadenosyl radical (dAdo•) is catalyzed by a [4Fe-4S] cluster. The sulfur and iron ions are denoted in yellow and red, respectively. The dAdo• initiates the catalytic cycle and either the cofactor is reformed at the end of the cycle, or used it stoichiometrically. (B) Examples of RiPP natural products with C-S, C-C, and C-O bonds are installed by radical SAM enzymes. (C) The gene cluster from Geobacter lovleyi contains a gene coding for a radical SAM enzyme (sbtM), which is co-clustered with one coding for a precursor peptide (sbtMa)²⁰. The DNA sequences of the SbtMa peptide contain one or more SeCys (U) coding amber codons. The minimal sequence used in this study is shown with U or C at the site modified by SbtM shown in red.

Many radical SAM proteins that function as RiPP maturases install chemically distinct crosslinks in their respective natural products². The most common of these reactions are thioether crosslinks of varied regiochemistry. The first thioether crosslinking enzyme to be identified, AlbA, is a sactipeptide (sulfur to α-carbon thioether) synthase that catalyzes the insertion of three thioether crosslinks in the precursor peptide of the antibiotic, subtilosin A¹⁰ (Fig. 1B). More recently, RiPP maturase proteins have been identified, which instead of forming a C-S bond between a cysteine residue and a Cα, catalyze a crosslink to either the Cβ¹¹ or Cγ¹² positions of the acceptor amino acid, thereby expanding the regiochemical outcomes.

The roles of radical SAM maturases in RiPP biosynthetic pathways, however, are not limited to the formation of C-S bonds. For example, PqqE catalyzes the formation of a C-C bond between the Cγ of a glutamate and the C3 of a tyrosine in the precursor to the bacterial redox cofactor pyrroloquinoline quinone (PQQ), PqqA¹³. The number of C-C bond forming RiPP maturases has a large repiotre¹⁴, including StrB, an enzyme that catalyzes a crosslink between a lysine and tryptophan residue in the biogenesis of streptide, a peptide involved in quorum sensing¹⁵ (Fig. 1B). Moreover, the recent discovery that TqqB catalyzes the formation of an ether crosslink between a glutamine and threonine residue has broadened the scope of radical-mediated chemistry in RiPPs to also include C-O bond formation¹⁶. Additional examples of radical-mediated transformations in RiPP pathways also include protein splicing to remove tyramine within a precursor protein to form an α-keto-β-amino amide¹⁷, with epimerization occurring at the α-carbon of various amino acids^18,19. Therefore, radical SAM RiPP maturases catalyze the formation of C-H, C-C, C-O, and C-S bonds in their respective substrates (Fig. 1B).

The radical SAM peptide maturase genes are often co-clustered with an open reading frame (orf) encoding the putative peptide substrate, as well as those coding for any other tailoring proteins that are required to synthesize the full natural product. Several years ago, Haft and Basu surveyed bacterial genomes, and identified potential RiPP gene clusters that encode at least one radical SAM enzyme and a putative peptide substrate²⁰ (Fig. 1C). One of these clusters, initially called the SCIFF proteins (six cysteines in forty five), has since been shown to catalyze a C-S crosslink between a Cys thiol and a carbon of a Thr residue²¹. The SCIFF family are part of a larger group of ranthipeptides that catalyze thioether crosslinks to positions other than the alpha carbon of an amino acid².

One of the clusters identified in this study was predicted to encode a selenocysteine (SeCys) containing peptide natural product with the necessary selenocysteine insertion elements (SECIS) also present. As with other bacterial SeCys-containing proteins, the SECIS elements are comprised of an amber UGA stop codon, with a predicted stem-loop structure located immediately downstream of the peptide orf^22,23. SeCys is a rare amino acid, and to our knowledge, there are no known examples of a post-translationally modified SeCys containing RiPP natural products. On the basis of sequence alignment, the proximal maturase (sbtM) is predicted to be a member of a subgroup of radical SAM enzymes, which has a cysteine rich C-terminal extension that coordinate two auxiliary [4Fe-4S] clusters (Fig. S1). These additions to the radical SAM core are termed SPASM domains, so-called because they were initially identified in the biosynthesis of subtilosin, PQQ, anaerobic sulfatase, and mycofactocin biosynthesis. While SPASM domains are prevalent in radical SAM enzymes that modify peptides, the exact function of these auxiliary clusters remains unknown. However, where available, data suggests that they are required for overall activity and not for the cleavage of SAM^21,24–30.

The study that identified the putative SeCys RiPP cluster postulated that these peptides could be of different lengths, some with potentially two SeCys modification sites²⁰. In general, these putative peptides have a relatively well-conserved leader sequence, which is followed by a core region that contains the SeCys residue(s). Unlike the N-terminus, however, the C-terminal regions are hypervariable in sequence. While this study clearly identified a group of novel peptide-maturase pairs, it did not provide any insight as to the nature of the modification(s) that are installed by the radical SAM enzyme. In the intervening time since the paper²⁰ was published, NCBI now identifies homologs of this radical SAM maturase as SbtM and the peptide a selenobacterocin. Interestingly, some of the peptides identified in the original study have since been updated to indicate the presence of at least two SeCys residues. To our knowledge, these peptides have not been detected in nature, and so the identity of the modification, the length of the peptide, or the SeCys content are all unknown. The sequence similarity network (SSN) of SbtM (Fig. 2) shows that it co-clusters in the network with other proteins of unknown function, that are often annotated as selenobiotic biosynthesis enzymes. Members of Geobacter and Desulfobacteria make up most of the closest neighbors in the cluster.

Figure 2. SSN of peptide modifying radical SAM enzymes.

The amino acid sequences of several characterized SPASM domain-containing proteins were individually subjected to BLAST analyses and the resulting sequences analyzed by the suite of programs at EFI (https://efi.igb.illinois.edu). Detailed procedures are in the Supporting Information³¹.

In this manuscript, we characterize the reaction between the radical SAM enzyme, SbtM, and a 53 amino acid peptide substrate reported by Haft and Basu²⁰. The peptide contains the first SeCys site, as well as additional amino acids beyond the SeCys that are conserved in other members of the family, but dispenses with the hypervariable region. Our data show that the radical SAM enzyme is a new member of the RiPP family that introduces a thiooxazole moiety in the peptide containing Cys. The modification also occurs when SbtM is incubated with the SeCys containing peptide, which is presumably the biologically encoded amino acid. Site directed mutagenesis studies show that while SeCys and Cys are tolerated at the position, Ala and Ser are not transformed. The cyclization is accomplished by the removal of two hydrogen atoms from the β-carbon of Cys, one proton from the Cys amide nitrogen, and a fourth proton from the α-carbon of the peptide. The radical SAM enzyme is introducing a C-O linkage to form an oxazole functional group in the peptide, expanding the repertoire of modifications installed by radical SAM RiPP maturases.

EXPERIMENTAL PROCEDURES

Cloning and Expression of SbtM.

The sbtM gene (WP_083768621) was codon optimized and cloned into the pET28a(+)-TEV plasmid by Genscript using the restriction sites of NdeI and XhoI. The codon optimized gene sequence of sbtM from Geobacter lovleyi is shown in Fig. S2 and the protein sequence is shown in Fig. S3.

The plasmids, sbtM-pET28a(+)-TEV and pDB1282, were co-transformed into Escherichia coli BL21 (DE3) T1 resistant cells (NEB C2527). The plasmid pDB1282 contains the isc operon from Azotobacter vinelandii under arabinose control. The operon is often included in the expression of radical SAM enzymes because it assists in the assembly of iron-sulfur clusters in heterologously expressed proteins^32,33. The transformation mixture was plated onto plates containing LB agar supplemented with 34 μg/mL kanamycin, 100 μg/mL ampicillin, and placed at 37 °C overnight. A single colony was selected from the plate and used to inoculate an overnight culture (0.1 L) of LB containing 34 μg/mL kanamycin and 100 μg/mL ampicillin grown at 37 °C with shaking at 175 rpm. An aliquot (0.01 L) of the overnight culture was used to inoculate six 2.8 L Fernbach flasks containing 1 L each of LB supplemented with 34 μg/mL kanamycin and 100 μg/mL ampicillin. The cultures were grown at 37 °C and 175 rpm to an OD_{600 nm} of ~0.5. At this point, the cultures were supplemented with 0.05 mM iron (III) chloride hexahydrate and 0.05 mM cysteine and expression of the isc operon from the pDB1282 was induced with the addition of 0.05% (w/v) arabinose. The temperature and shaking were then reduced to 18 °C and 120 rpm, respectively. At an OD_{600 nm} ~0.7, expression from the sbtM-pET28a(+)-TEV plasmid was induced with the addition of 0.1 mM IPTG. The cultures were grown overnight (~18 h) and the cells were harvested by centrifugation at 5,000 xg. The cell pastes were flash frozen in liquid nitrogen and stored at −80 °C until use.

Purification of SbtM.

The purification of SbtM was performed in a Coy Laboratories anaerobic chamber maintained with 97% N₂/ 3% H₂ atmosphere. The cell paste (20 g) was placed in a metal beaker and resuspended in ~0.2 L of 0.05 M KPi (pH 7.4) buffer containing 0.5 M KCl, 0.05 M imidazole, 100 μg/mL lysozyme, and 1 mM PMSF. The cells were lysed by a Branson digital sonifier operated at 50% amplitude for a total of 12 min (15 sec on/ 45 sec off) while stirring on ice. The suspension was centrifuged at 18,000 xg for 35 min at 4 °C. The clarified lysate was then loaded onto a 5 mL nickel sulfate charged HisTrap HP column (GE Healthcare) that had been equilibrated with loading buffer containing 0.05 M KPi (pH 7.4), 0.5 M KCl, and 0.05 M imidazole. The column was washed with 5 column volumes (CV) of loading buffer and the enzyme was eluted with a linear gradient over 8 CV to 0.5 M imidazole in the loading buffer. All fractions containing SbtM were identified via color and/or SDS-PAGE gel. The pooled fractions were desalted on a BioGel P6 DG desalting gel 100–200 mesh (wet) (Bio-Rad) into 0.02 M Tris•HCl (pH 8.0) containing 5 mM DTT.

The resulting protein was further purified by anion exchange chromatography as follows. The desalted fractions from the affinity step were loaded onto a Q-Sepharose FF column (11 × 3 cm) (GE Healthcare) which had been equilibrated with buffer containing 0.02 M Tris•HCl (pH 8.0) and 5 mM DTT. SbtM was eluted with a 0.2 L linear gradient to 0.5 M KCl in the loading buffer. Fractions containing the desired protein were identified by color and/or SDS-PAGE gel and desalted into buffer containing 0.05 M PIPES•NaOH (pH 7.4), 0.15 M KCl, and 5 mM DTT. The protein was then flash frozen with liquid nitrogen and stored at −80 °C.

The concentration of the desalted SbtM was determined by the Bradford method, using BSA as a standard. The protein was reconstituted by mixing with 15–25 molar equivalents of FeCl₃ hexahydrate (dissolved in saturated sodium bicarbonate) and Na₂S nonahydrate (in water). The iron (III) chloride hexahydrate was added dropwise to the protein followed by the sodium sulfide, which was dissolved in water. The reconstitution mixture was stirred for 4 h at room temperature. The resulting mixture was clarified by centrifugation at 16,000 xg for 5 min to remove any debris and desalted on a BioGel P6 DG desalting gel 100–200 mesh (wet) (Bio-Rad) into buffer containing 0.05 M PIPES•NaOH (pH 7.4), 0.15 M KCl, and 5 mM DTT. The protein was concentrated to ~1 mL with an Amicon concentrator under N₂ with a YM-10 membrane (Millipore).

The reconstituted enzyme was further purified by a Sephacryl 16/60 S-200 column equilibrated with buffer containing 0.05 M PIPES•NaOH (pH 7.4), 0.15 M KCl, and 5 mM DTT. The protein was eluted isocratically at 1 mL/min and fractions containing SbtM were identified by color and/or SDS-PAGE gel. The pooled fractions were concentrated to ~0.5 mL, aliquots were flash frozen in liquid nitrogen, and stored at −80 °C. SbtM was quantified by a Bradford assay with BSA as a standard.

Amino acid analysis and Iron concentration determination.

To obtain a correction factor to correlate the Bradford assays to the actual protein concentration, five independent preparations of SbtM were subjected to amino acid analysis (by the Molecular Structure Facility at the University of California-Davis). For the analysis, the protein samples were hydrolyzed under vacuum in solution containing 6 N HCl and 1 % phenol at 110 °C. The samples were resuspended with 40 nmol/mL norleucine as an internal standard. The samples were analyzed by a Hitachi 8800 analyzer that had been calibrated with amino acid standards for protein hydrolysate on the Na-based Hitachi 8800 (Sigma, A-9906). These standards were verified by the National Institute of Standards and Technology (NIST) standard reference material 2389a. The protein samples were injected onto a Concise ion-exchange column (AminoSep Beckman Style Na+, part #AAA-99–6312) with a secondary reaction with ninhydrin for detection that uses Pickering Na buffers. The correction factor for the Bradford assays was thus determined to be 0.61. Iron content of the reconstituted protein was determined though ICP-MS on two separate enzyme preparations at the Center for Water, Ecosystems, and Climate Science in the Department of Geology and Geophysics at the University of Utah. For the analysis, the enzyme preparations were diluted to a concentration of 2–10 μM trace metals grade nitric acid prior to submission. The determination of iron in soluble protein solutions was performed with a triple quadruple inductively coupled plasma mass spectrometer (ICPMS, Agilent 8900, Santa Clara, California, USA) at the ICPMS Laboratories, Dept. of Geology and Geophysics, University of Utah. Protein solutions were diluted 1:40 with 2.4% HNO₃ and 10 ng In/mL was added as internal standard. An external calibration curve was prepared from 1,000 mg/L single elemental standard (Inorganic Ventures, Christiansburg, VA, USA). Concentrations of Fe in the six calibration solutions were 0, 8.3, 20.7, 66.2, 165.5 and 331.1 ng Fe/mL respectively, and all contained 10 ng In/mL. Diluted samples, blanks and the calibration solutions were run in the ICPMS using a double-pass quartz spray chamber; PTFE nebulizer and dual-syringe introduction system (Teledyne, AVX71000), platinum cones and sapphire injector in a quartz platinum-shielded torch. Fe and In were detected at masses of 56 and 115, respectively, with a flow of 8 mL He/min in the collision cell. Limit of determination (LoD) was calculated as three times the standard deviation of the blanks, multiplied by the 40-fold dilution factor used for samples. The instrument is located in a filtered air positive pressure lab and sample handling and dilutions were performed in laminar flow benches and using calibrated pipettors (Eppendorf Reference, Hamburg, Germany). Certified Reference Material CRM 1643f (National Institute of Standards and Technology, Gaithersburg, MD, USA) was diluted 1:20 and run together with the samples and external calibration curve as quality control for the calibration. The measure value for Fe in CRM 1643f agreed with the certified value within 10%.

Synthesis of unlabeled, Cys35SeCys, [¹³C₂, ¹⁵N]-Gly34, [¹³C₃, ¹⁵N]-Cys35, [2,2 D₂]-Gly34, [3,3 D₂]-Cys35 SbtMa peptides.

All SbtMa peptides and variants were synthesized on a PS3 Peptide Synthesizer (Protein Technologies Inc.) at a 0.025 mmol scale using solid phase peptide synthesis (SPPS) methodology. The isotopically enriched Fmoc-[¹³C₂, ¹⁵N]-Gly, Fmoc-[¹³C₃, ¹⁵N]-Cys S-trityl, Fmoc-[2,2 D₂]-Gly, Fmoc-[3,3 D₂]-Cys S-trityl were purchased from Cambridge Isotope Laboratories. The Fmoc-SeCys(Mob)-OH was purchased from Sigma-Aldrich. The syntheses were carried out essentially as described by Bruender et al²¹, with the following modifications. After precipitating the peptide in ice-cold diethyl ether, the solution was poured over a Büchner-funnel and vacuumed to collect the peptide and continued to dry for ~1 h prior to being resuspended in ~40 mL of water. The suspension was then placed in a sonicator bath for 30 min to aid in dissolution of the peptide, flash frozen in liquid nitrogen, and lyophilized. The cleavage and deprotection of SbtMa peptide containing SeCys at position 35 followed the procedure described by SteMarie et al³⁴, except that the cleavage buffer contained 2% 2,2’-dithiobis-5-nitropyridine (DTNP) and did not include thioanisole. The resulting Sec-5-Npys adduct was deprotected with 5 equivalents of ascorbate (pH 4.5) at 25 °C for 4 h. and stirred for 3 h in 10 mM DTT to reduce diselenides and HPLC-purified on a Phenomenex Jupiter C12 preparative column (21.2 mm × 250 mm, 4 μm particle size, 90 Å pore size) with Buffer A as 0.1% TFA (HPLC Grade) in nanopure water and Buffer B as 0.1% TFA (HPLC Grade) in acetonitrile (HPLC Grade). The separation was run at 5 mL/min with a linear gradient of Buffer A from 88%−60% for 95 min. The column was washed with 100% Buffer B from 96–125 min and returned to 88% Buffer A from 125–130 min, followed by and reequilibration at 88% Buffer A from 130–155 min. Fractions containing the desired peptide were identified via LC-MS using one of two instrumental set-ups (Ultimate 3000 HPLC with a diode array detector interfaced to a LTQ OrbiTrap XL mass spectrometer, or a Vanquish UHPLC with a diode array detector connected to a Q-Exactive mass spectrometer) fitted with a Hypersil GOLD C4 column (4.6 × 250 mm, 5 μm particle size) for separation at 0.2 mL/min. For the LC-MS program, Buffer A was LC-MS Optima water (Fisher)/ 0.1% (v/v) LC-MS Optima TFA (Fisher) and Buffer B was LC-MS Optima acetonitrile (Fisher)/ 0.1% (v/v) LC-MS Optima TFA (Fisher). The 15 min separation consisted of washing the column with 100% A for 3 min, followed by a linear gradient to 100% B from 3 to 6 min, followed by washing with 100% B from 6–8 min. The column was returned to 100% A over 3 min and equilibrated in the starting conditions for an additional 3 min. The MS detectors were operated in positive ion mode and the FT analyzer settings are as follows: 100,000 resolution for the LTQ OrbiTrap and 70,000 resolution for the Q Exactive, 1 microscan, and 200 ms maximum injections time. All data were analyzed on Xcalibur software (ThermoFisher).

Enzymatic reaction of SbtMa with SbtM.

All modification reactions were conducted at room temperature in a Coy Laboratories anaerobic chamber with 97% N₂/ 3% H₂ atmosphere. The reactions contained 0.05 M PIPES•NaOH (pH 7.4), 10 mM DTT, 10 mM dT, 2 mM SAM (enzymatically synthesized and purified as previously described³⁵), ~0.8 mg/mL SbtMa, and 0.015–0.03 mM SbtM, in a total volume of 0.1 mL. Control reactions were also conducted in the absence of dT, SAM, and SbtM, respectively. The reactions were initiated with the addition of the peptide substrate and quenched after at least 4 h with the addition of 30 μL of 30% (w/v) TCA (ACS grade). The samples were centrifuged at 10,000 rpm in a microcentrifuge to remove the precipitated enzyme.

Alkylation of SbtMa.

Where alkylation by iodoacetamide was desired, Tris•HCl (pH 8.0) was added to a final concentration of 0.1 M to the quenched peptide reactions followed by iodoacetamide to a final concentration of 30 mM. The mixtures were incubated for 1 h in the dark and the excess iodoacetamide was quenched by the addition of DTT to 30 mM final concentration.

U/HPLC-MS analysis of enzymatic reactions.

The enzymatic reactions were analyzed using either an Ultimate 3000 HPLC with a diode array detector connected to a LTQ OrbiTrap XL mass spectrometer, or a Vanquish UHPLC with a diode array detector connected to a Q-Exactive mass spectrometer. The LTQ OrbiTrap XL or Q-Exactive was operated in positive ion mode and the FT analyzer was set to 100,000 resolution, 1 microscan, and 200 ms maximum injections time. The data were analyzed using Xcalibur software. For the analysis, an aliquot (98 μL) was injected onto a Hypersil GOLD C4 column (4.6 × 250 mm, 5 μm particle size) (Thermo Fisher) pre-equilibrated in 0.1% (v/v) LC-MS Optima TFA (Fisher) in LC-MS Optima water (Fisher). The separation was carried out at 0.2 mL/min with Buffer A 0.1% (v/v) TFA in water and Buffer B Optima grade acetonitrile with 0.1% (v/v) TFA. The separation utilized the following program: washing with 0% B from 0–3 min; linear gradient to 99% B from 3–28 min; washing with 100% B from 28–34.9 min; equilibration at 0% B from 35–40 min.

CID fragmentation of unmodified and modified SbtMa.

To obtain sufficient material for fragmentation by direct infusion, the reactions described above were scaled up 5-fold to 0.5 mL and the SbtMa was increased to ~2 mg/mL. Following TCA treatment and centrifugation, the reaction mixtures were desalted by C4 ZipTips (Millipore) following the manufacturer’s protocols. The analyzer was first tuned to the mass of the SbtMa. For the analyses, the +4 charge state envelope corresponding to the modified and unmodified SbtMa was isolated in the CID cell using an isolation width of 1.0 m/z and 0.1 ms activation time and fragmented with a collision energy of 20%. The analysis of fragmentation products used mMass software.

NMR of unmodified and modified SbtMa.

The modification reactions were scaled up to 1 mL and contained 0.05 M PIPES•NaOH (pH 7.4), 10 mM DTT, 10 mM DT, 2 mM SAM, ~0.1 mg/mL SbtMa, and 7–13 μM SbtM. To obtain sufficient material, ~60 individual (1 mL) reactions were combined after overnight incubation. The reactions were pooled and lyophilized and resuspended in ~5 mL of water, filtered through a sterile syringe filter with a 0.2 μm polyethersulfone membrane (VWR), and purified via a Phenomenex Jupiter C12 prep column following the procedure detailed above for SPPS peptides. The fractions containing modified peptide were identified by LC-MS analysis of the fractions, pooled, and lyophilized. These modified peptides were used for the ¹³C experiments. The lyophilized modified peptide was then resuspended in 50 mM Tris•HCl (pH 8.0) and reacted with 20–30 mM iodoacetamide for 1 h in the dark. The reaction was quenched by the addition of 30 mM DTT and purified on the Phenomenex Jupiter C12 prep column with the same procedure detailed for SPPS peptides. These modified peptides were used for the ¹⁵N NMR experiments. The unmodified SbtMa samples were prepared by resuspending lyophilized SbtMa in 50 mM Tris•HCl (pH 8.0) and following the same iodoacetamide procedure and purification as for the modified samples. The lyophilized peptide was dissolved in 50 mM KPi (pH 6.0) 10% D₂O for ¹⁵NHSQC and HNCACB experiments. For the ¹³C experiments, lyophilized unmodified peptide was resuspended in 50 mM KPi (pH 6.0), 2 mM DTT, 100% D₂O. The NMR spectra were acquired on an Agilent DirectDrive 500 MHz NMR spectrometer at the D.M. Grant NMR Center at the University of Utah. The 2D-NMR data were analyzed by VNMRJ software and the 1D-NMR data and HMBC were analyzed by Mestrenova software.

FTIR of unmodified and modified SbtMa.

Lyophilized SbtMa peptide was mixed with dry KBr using an agate mortar and pestle and pressed into a pellet. FTIR measurements were run on a Bruker 80v interferometer with a DLaTGS detector, 4 cm⁻¹ resolution, and 128 scans. Measurements with modified or unmodified SbtMa were carried out under vacuum at 4.32 mbar and illuminated with a globar MIR source of non-polarized light. Transmission spectra were processed using an auto spline function implemented in the OPUS software.

DFT on unmodified and modified truncated SbtMa.

Density functional theory calculation were performed with Gaussian16 Version A.03³⁷. Truncated peptide models CH₃-GlyCys-CH₃ were built in Gaussview and truncated with a methyl group at the N- and C- terminal sides of the GlyCys fragment. Both unmodified and modified truncated peptides were optimized using the B3LYP functional^38–41 with the 6–311G basis set for all atoms. Normal vibrational modes were extracted from analytical frequency calculations and scaled by 0.966 due to the tendency of the B3LYP to overestimate covalency⁴².

RESULTS

Protein expression and purification.

The gene encoding the putative radical SAM maturase, sbtM, was codon optimized and cloned into pET28a(+)-TEV plasmid for heterologous expression in E. coli. To ensure that protein was replete with the requisite [4Fe-4S] cofactors, the cells were co-transformed with pDB1282. This plasmid contains the isc operon from A. vinelandii, which previous studies have shown aid in iron-sulfur cluster assembly^32,33. The enzyme was purified via His₆ affinity chromatography followed by a Q-Sepharose anion exchange under anaerobic conditions. Sequence analysis (Fig. S1) suggests the protein is likely to have three [4Fe-4S] clusters, therefore the maturase was reconstituted with 15–25-fold molar excess iron and sulfide. The reconstitution step was followed by gel-filtration and the purified protein is judged to be >95% pure based on SDS-PAGE analysis (Fig. S4). To permit accurate quantification of cluster stoichiometry, amino acid analysis was carried out revealing a correction factor of 0.61 for Bradford assays with BSA as standard. The preparations of protein reconstituted in the presence of 25-fold excess iron contain 13.6 ± 1.9 mol, whereas those reconstituted in the presence of lower concentrations of iron were less replete. This is consistent with the presence of at least two [4Fe-4S] clusters per polypeptide chain. The putative minimal substrate SbtMa (see Fig. 1) was prepared by standard SPPS methods, as described in the Experimental Procedures. The majority of the experiments presented in this study utilize SbtMa containing Cys at position 35 because of its ease of synthesis and stability; however, a SeCys containing analog was also prepared by SPPS and utilized as discussed below.

SbtM modifies SbtMa.

Preliminary studies demonstrate that when SbtM is combined with Cys-containing SbtMa (see Fig. 1), a new product is formed in a manner that requires the presence of SAM, dT, and the maturase (Fig. 3). Samples were analyzed by LC-MS employing high resolution mass analyzer, after quenching of the reaction mixtures with TCA (see Fig. S5 for full spectra). The regions in the MS spectra corresponding to the +4 charge state are shown in Fig. 3A. In the absence of the enzyme, the substrate displays a monoisotopic peak at m/z of 1273.90, which is within 3.1 ppm of the expected value for the unmodified peptide. However, when SAM, dT, and SbtM are included, the envelope shifts the monoisotopic peak to m/z of 1272.89. This overall mass change of 4.03 amu is consistent with a modification that potentially involves the loss of 4 protons.

Figure 3. SbtM modifies SbtMa.

The panels show the +4 charge state of the mass spectra of the SbtMa treated ± SAM, ± dT, or ± SbtM. The orange and blue lines mark the positions of the monoisotopic peaks for the modified and unmodified peptides, respectively. The spectra in (A) correspond to reactions containing Cys-SbtMa. The calculated m/z for unmodified and modified SbtMa occur at 1273.8976 and 1272.8898 amu, respectively, and the measured monoisotopic features are within 3.1 ppm of the theoretical. In (B), the spectra corresponding to carbamidomethylated peptide are shown. The calculated m/z for the monoisotopic peak before and after modification occur at 1288.1530 and 1287.1452 amu, respectively. The measured values are within 2.9 ppm of the theoretical. In (C), spectra obtained with SeCys-SbtMa after carbamidomethylated are shown. The calculated m/z for the monoisotopic peak before and after modification occur at 1300.1391 and 1299.1313 amu, respectively. The measured values are within 0.8 ppm of the theoretical. In (D) the UV-visible spectra of the modified Cys-SbtMa (blue), modified SeCys-SbtMa (pink) and unmodified Cys-SbtMa (black) peptides are compared. The modified peptides exhibits a feature at 304 nm and 313 nm (λ_max), respectively, suggesting the presence of a conjugated unsaturation. The spectra of the SeCys and Cys containing peptides shown in (D) are normalized to the λ_max of the respective peptide. The spectrum of the unlabeled peptide was normalized at 304 nm, whereas the peptide from a control reaction that did not include SbtM does not. By contrast, the Se-Cys containing peptide shows a red shift to 313 nm. This data support the formation of a conjugated unsaturation in the peptide.

To determine if the modification involved the thiol of the Cys residue, the reaction mixtures were treated with iodoacetamide to alkylate the thiol prior to being analyzed by LC-MS. The monoisotopic peak for the unmodified peptide under these conditions is at m/z of 1288.16 (Fig. 3B), which is within 2.9 ppm of the expected value for a peptide containing a single carbamidomethylated Cys residue. Interestingly, the SbtM modified peptide was alkylated as well, in addition to exhibiting a monoisotopic mass shift of of 4.03 amu relative to the unmodified peptide. These observations suggest that the modified peptide retains a reactive sulfur, although the structure of the modified moiety remains unknown.

To further probe the nature of the modification we also examined the UV-visible spectra of the peptide eluting from the HPLC column in experiments where a diode-array detector was placed in-line with the MS instrument. Fig. 3D compares the spectral data for modified and unmodified peptides. To our surprise, the Cys-containing modified peptide exhibits λ_max=304

We also synthesized the SbtMa peptide containing SeCys at position 35 instead of Cys (see Fig. 1), to determine if the modification reaction occurs with SeCys, which is presumed to be the naturally occurring amino acid. Due to the difficulty of reducing diselenides that form in the SPPS workup, the experiments involving alkylation by iodoacetamide were completed after the reaction. Unlike the Cys containing peptides, Se gives rise to a complex pattern because of the presence of numerous naturally occurring and abundant stable isotopes (⁷⁶Se, 9.23%; ⁷⁷Se, 7.6%; ⁷⁸Se, 23.69%, ⁸⁰Se, 49.80; and ⁸²Se, 8.82). In these experiments, the monoisotopic peak of the +4 charge state of the peptide occurs at m/z of 1300.14, which is within 0.8 ppm of the expected value for a singly carbamidomethylated SeCys residue (Fig. 3C). The observed +4 charge state envelope for the singly carbamidomethylated peptide (Fig. 3C) is identical to the simulated envelope (Fig. S6A). After incubation with dT, SAM, and SbtM, however, the envelope shifts to place the monoisotopic peak at m/z of 1299.13 (Fig. 3C). The simulated envelope of the +4 charge state for a peptide that has undergone a loss of 4.03 amu is consistent with the observed data (Fig. S6B).

SbtM catalyzes modification of the GC motif in SbtMa.

We next investigated the location of the modification in the SbtMa via collision-induced dissociation (CID) of the modified SbtMa by MS-MS fragmentation. The reaction mixture and control reaction missing SbtM were introduced into the mass spectrometer by direct infusion after quenching with TCA and removal of salts. For the analysis, the envelope corresponding to the +4 charge state was isolated and fragmented in the CID cell of the instrument. The fragmentation data were analyzed by mMass and all of the b and y ions that were identified are listed in Table S1 and Table S2 for the modified peptide and control peptide, respectively. The y ions proved highly informative as close examination of the data reveals that both samples exhibit species with identical mass up to y18. All species from y20 (+2 charge state) onwards display a 2 amu difference between the modified and unmodified samples, corresponding to a 4 amu difference. The region in the fragmentation spectra that highlights this key difference is shown in Fig. 4A. These data show that the 4 amu loss observed in the presence of SbtM can be traced to residues Gly34 and Cys35.

Figure 4. SbtM modifies SbtMa at position 35.

(A) The CID fragmentation patterns of the modified and unmodified SbtMa peptides show identical y18 fragments. By contrast, the y20 fragments show a 4 amu difference. Note that the y18 fragment is in the +1 charge state whereas the y20 fragment is in the +2 state. (B) The [3,3 D₂]-Cys35 SbtMa peptide undergoes a mass change of 6 amu in the presence of SbtM. By contrast, the [2,2 D₂]-Gly34 undergoes a 4 amu change upon modification. The measured monoisotopic features with both variants are within 5 ppm of the expected values. Note that in both cases, the envelope still shows some residual substrate.

To probe the significance of this observation, peptides containing Ala or Ser in the place of Cys35 were synthesized and incubated with SbtM. Unlike the Cys and SeCys containing counterparts, the Ala and Ser containing SbtMa peptides exhibited identical mass spectral features after incubation with SbtM, under conditions where parallel experiments with Cys-containing peptide showed a 4.03 amu loss (Fig. S7). These observations are significant because they suggest that while the modification maintains the reactivity of the thiol/selenol residue, the identity of the residue at position 35 of the peptide (Cys or SeCys) is critical for the enzymatic transformation.

SbtM removes two hydrogens from Cys35.

To probe the nature of the modification introduced by SbtM isotopologs of the SbtMa peptide were synthesized in which Gly34 or Cys35 were replaced with [2,2 D₂]-Gly or [3,3 D₂]-Cys, respectively. SbtM was incubated with each of the isotopologs and the resulting products were analyzed as described above for unlabeled substrate. The monoisotopic peak of the +4 charge state envelope for the [2,2 D₂]-Gly34 SbtMa peptide is at m/z of 1274.40 amu, which is within 1.7 ppm of the expected value (Fig. 4B). Incubation with SbtM leads to a 4.03 amu shift in the mass of peptide, which is identical to that observed with the unlabeled substrate. In contrast, the monoisotopic peak for the +4 charge state of the SbtMa peptide substituted with [3,3 D₂]-Cys35 (at 1274.40 amu) shifts to a m/z of 1272.89 amu. This corresponds to an overall mass loss of 6.04 amu, indicating the loss of two protons during the modification as well as two deuterium atoms at the β-position of Cys35 (Fig. 4B). These data clearly implicate the Cys35 as the site of modification.

SbtM catalyzes H-atom abstraction from Cys35.

Radical SAM enzymes generally initiate catalysis by H-atom abstraction from their respective substrate. Once we had determined the Gly-Cys region of the peptide is the site of chemistry, we examined H-atom transfer between these residues and dAdo• with the deuterated isotopologs of the substrate. Intriguingly, when SbtM is incubated with a peptide containing [3,3 D₂]-Cys35, the dAdo obtained from the reaction is enriched in singly deuterated dAdo (Fig. S8). By contrast, dAdo from [2,2 D₂]-Gly34 peptide does not show any enrichment beyond that expected from natural abundance of the isotope. The enrichment of singly deuterated dAdo in the reactions containing the labeled Cys is consistent with H-atom transfer between the β-carbon of Cys35 and dAdo•. Collectively, these data show that at least two of the hydrogens that are lost upon conversion of the substrate to the product are derived from the β-carbon of Cys35.

NMR characterization of structure.

To elucidate the structure of the modified peptide, SbtMa containing [¹³C₂, ¹⁵N]-Gly34 and/or [¹³C₃, ¹⁵N]-Cys35 was synthesized by SPPS purified by HPLC and lyophilized. The resulting peptides were subjected to modification by SbtM, and followed by treatment with iodoacetamide. The HPLC-purified modified and carbamidomethylated peptides were resuspended in buffer containing 10% D₂O/H₂O and analyzed by NMR.

We examined the fate of the amide protons of the Gly34 and Cys35 through a ¹⁵N-HSQC experiment, which correlates amide protons to the amide nitrogen. The full NMR spectra are shown in Fig. S9. In a control experiment, unmodified SbtMa isotopolog containing [¹³C₂, ¹⁵N]-Gly34 and [¹³C₃, ¹⁵N]-Cys35, the HN-Gly at (8.26, 110.4 ppm) and HN-Cys (8.26, 118.9 ppm) correlations are clearly observed (Fig. 5A). After incubation with SbtM, we observe correlations corresponding to the HN-Gly and HN-Cys from unreacted SbtMa, as well as a single new correlation at (8.46, 110.3 ppm) assigned to Gly. Significantly, we do not observe any modified peak for the amide of Cys35, indicating that a structural change has occurred leading to the loss of the NH proton (Fig. 5B).

Figure 5. NMR analysis of the SbtM-modified peptide.

The ¹⁵N-HSQC spectra of unmodified (A) and modified (B) SbtMa which shows the loss of the Cys amide proton. The HNCACB of unmodified (C) and modified (D) SbtMa confirm the loss of the Cys amide proton. (E) Shows correlations in the unmodified peptide. (F) Shows the ¹³C spectra of the unmodified and modified SbtMa. The labels for the resonances in the unmodified peptide are shown in black and those in the modified ones are shown in blue.

To obtain additional structural information, a HNCACB experiment was carried out to correlate the amide proton to the adjacent carbon centers. With unmodified peptide containing [¹³C₂, ¹⁵N]-Gly34 and [¹³C₃, ¹⁵N]-Cys35, the previously assigned amide proton of the Cys and Gly at 8.26 ppm (Fig. 5A) both correlate to the α-carbon of Gly (45.4 ppm), through the peptide carbonyl. In addition, the amide proton of the Cys residue shows a correlation to the α- and β-carbons of the Cys residue at 55.75 and 36.06 ppm, respectively (Fig. 5C). We note that the HNCys-βCys has a negative intensity (denoted in red) as is expected of β-carbons in HNCACB experiments.

Upon modification by SbtM, the HNCACB spectrum obtained with peptide containing [¹³C₂, ¹⁵N]-Gly34, [¹³C₃, ¹⁵N]-Cys35 exhibits a single new correlation between the amide proton of Gly at 8.47 ppm and the α-carbon of Gly at 39.2 ppm (Fig. 5D). The amide proton resonance at 8.47 ppm is similar to that observed in the ¹⁵N-HSQC of the modified product (Fig. 5B). Significantly, all the correlations to the amide proton of the Cys are absent in the modified peptide, consistent with the loss of the Cys amide proton, which was also observed in the ¹⁵N-HSQC spectrum. We note that an identical set of two-dimensional NMR measurements were also made with SbtM modified peptide labeled with only [¹³C₂, ¹⁵N]-Gly34, that had not undergone reaction with iodoacetamide. The only difference between these data and the spectra presented in Fig. S10 and S11 were small changes in the chemical shifts. These data are included in the Supporting Information. The loss of the NH (Fig. 5B) and β-protons both in the HNCACB (Fig. 5D) and the deuterium labeling data (Fig. 4B) strongly support the notion that the modification catalyzed by SbtM involves cyclization of the side chain in some fashion that also imparts significant aromatic character to the product, as suggested by the UV-visible spectrum of the modified peptide (Fig. 3D). Therefore, we also compared the ¹³C-NMR spectra of unmodified and SbtM modified ¹³C and ¹⁵N labeled SbtMa peptide (Fig. 5F). In addition to the resonances from the ¹³C labeled atoms in the peptide, we also observe two additional features at 76 and 29 ppm for DTT. The spectrum of the modified peptide (see Fig. S12) exhibits some of the same features as the SbtM-modified peptide, as would be expected from unmodified peptide that remains in the sample. We also observe several sharp resonances < 100 ppm, presumably from small molecules that are co-purified in the course of the workup. However, we observe 3 new resonances at 118, 120 and 165 ppm, all of which are in the aromatic region. This observation will be examined further below, but are consistent with formation of a conjugated system involving the Cys sidechain and amide.

FTIR of the modified peptide.

Further investigation towards the nature of the modification was carried out using FTIR spectroscopy. The spectra of the unmodified and modified SbtMa peptide were measured under vacuum in KBr pellets (Fig. 6). The unmodified SbtMa exhibits a characteristic v_S-H stretch at 2555 cm⁻¹, which is attributed to a thiol^43–46. This peak is not able to be resolved in the modified peptide, though an additional feature at ~ 2180 cm⁻¹ is observed. The modification also introduces several new, sharp features at ~1550 cm⁻¹.

Figure 6. Analysis of the SbtMa by FTIR.

The experimental FT-IR spectra show the unmodified (black) and modified (red) SbtMa peptide. Asterisks (*) represent new features present in the modified peptide. Scaled features highlighting the S-H stretching region are shown in the inset.

Additionally, the modified peptide also exhibits significant changes in the amide II region. The secondary structures of the unmodified or modified peptides are not known. However, changes in the amide I and amide II regions are known to reflect changes in the structure of the peptide backbone. Changes in the amide II region are proposed to result from changes to the coupling of the backbone C-N stretches along the peptide after cyclization of the Cys residue⁴⁸. Overall these changes are indicative of the modification disrupting the peptide backbone, similar to the effect of proline residues in short peptides and proteins^43,49–51. However, since the three-dimensional structure of the peptides are not known, further structural interpretation of the changes in the amide I and amide II stretches is not possible.

Nature of the modified SbtMa.

In considering the structure of the modified SbtMa, one needs to take into account all of the MS, UV-visible, NMR, and FT-IR observations. SbtM catalyzes the loss of 4 amu from the SbtMa peptide. The modification requires the presence of the Cys/SeCys residue at position 35. Two of the hydrogen atoms lost in this reaction are from the β-position of the Cys/SeCys at position 35 and the third form the corresponding amide nitrogen. The product appears to retain an iodoacetamide-sensitive group even after the modification by SbtM. The location from which the fourth hydrogen that is lost is not known. We know, on the basis of the H-transfer data, that at least one (and possibly both) β-hydrogens of the Cys/SeCys are transferred to dAdo• forming dAdo in two successive H-atom transfer events. Finally, the Fe-content suggests the presence of multiple [4Fe-4S] clusters, one of which, by analogy to other radical SAM enzymes, we posit is involved in the activation of SAM to generate the dAdo•, while at least one more may be involved in the chemical transformation.

By analogy to anaerobic sulfatase maturing enzyme (AnSME)^52–54 we propose that SbtM catalyzes H-atom transfer from β-carbon of Cys followed by oxidation of the resulting species by a second cluster to form a thioformylglycine intermediate (Fig. 7). This intermediate could have several fates. It could undergo cyclization with the carbonyl moiety of the Gly-Cys amide bond to form a 5-membered ring. Alternatively, it can form a 4-membered ring in an analogous manner, with the oxygen from the carbonyl of Cys/SeCys, or, a 3-membered ring by cyclizing with the amide of the Gly-Cys peptide bond. Each of these could subsequently undergo a second H-atom transfer from the β-carbon of the Cys/SeCys, which upon oxidation would yield a thione product. Both the 5-and 3-membered ring products can undergo tautomerization to present a thiol moiety. However, the tautomerization to the aromatic thiooxazole (5-membered) is more favorable than the tautomerized aziridine thione (3-membered) which would be antiaromatic.

Figure 7.

Possible products of SbtMa peptide. The peptide is proposed to initially undergo H-atom transfer to dAdo to form a radical intermediate that is oxidized to thioformylglycine. This thioformylglycine can be trapped in one of three manifolds to form either a 3-, 4- or 5-membered intermediates. Evidence for each is discussed in the text.

All of the initial products proposed in Fig 7 are isobaric and could not be distinguished on the basis of MS. However, we can eliminate the 4-membered ring from consideration because it would still maintain the Cys amide, which is inconsistent with the absence of an N-H resonance in the HSQC data (Fig. 5B) and because optimization by DFT leads to a ring opened product. DFT calculations, using a truncated dipeptide model (see Experimental Procedures) suggest that the 5-membered ring containing product is more stable than one containing a 3-membered ring (see Table. S3). Indeed, upon aromatization, the thiooxazole structure is 26 kcal/mol more stable than the aziridine thione. While the inability to identify a resonance corresponding to the α-proton of the Cys strongly points to the aromatization of the oxazoline to the oxazole, we wondered if a similar process could also occur with the aziridine structure. DFT calculations suggest that the analogous tautomerization of this structure is highly unfavorable relative to the aziridine thione (by 42.4 kcal/mol), likely because of the fact that the resulting structure is antiaromatic.

Both the aziridine thione and the thiooxazole products would be reactive towards iodoacetamide. While the antiaromatic structure would not be favored, resonance delocalization from the amide could lead to activation of the thione for carbamidomethylation. However, a key difference between the aziridinethione and the thiooxazole structure is that they will have differential reactivity towards mild reducing agents. One would predict that the aziridine thione would undergo reduction by a hydride transfer reagent, whereas the thiooxazole would not⁵⁵. Therefore, we reduced SbtM-modified Cys-SbtMa peptide with either NaBH₄ or NaBD₄ (Fig. S14). The data clearly show that the envelope of the charge states of the product is identical to those of reactions that were not quenched with either NaBH₄ or NaBD₄. This observation is most consistent with the assignment of the modified amino acid as a thiooxazole.

NMR spectroscopic observations lend further support for the thiooxazole assignment. Perhaps the most crucial observation is the fact that we see three resonances in the ¹³C NMR (Fig. 5F) that are similar to those that have been reported in other oxazole structures^56,57. These could be assigned to the three carbons of the oxazole (see Fig. 7). To obtain additional insights into the ¹³C NMR data, particularly the 165 ppm resonance, a set of ¹³C-HSQC spectra were obtained with unmodified and SbtM-modified peptide (Fig. 8). The chemical shift of the α-proton of Gly shifts from 3.9 to 4.2 ppm in the modified peptide. In the ¹³C-HMBC spectrum of the modified peptide, these protons are correlated with the 165 ppm ¹³C resonance, directly support the connectivity of the Gly-Cys, and the assignment of the 165 ppm feature to what was the carbonyl carbon of the unmodified peptide. Since the modification reaction was not complete, we also observe a correlation between the carbonyl carbon of the Gly and the α-protons of the unmodified peptide (at 3.8, 172 ppm).

Figure 8. ¹³C NMR characterization.

¹³C HSQC of unmodified (A) and SbtM modified (B) SbtMa peptide. The HMBC spectrum of modified SbtMa peptide zoomed in on the aromatic region (C) and the unmodified (D) and modified (E) structures with the pink dots corresponding to the ¹³C HSQC correlations and the green arrow corresponding the HMBC correlation.

The UV-visible features attributed to the product further support the assignment. The λ_max=304 nm feature in the UV-visible data and its red shift to 313 nm in the Se-Cys containing SbtM-modified peptide are further suggestive of a conjugated system (see Fig. 3D). While unsubstituted oxazoles have UV-visible features that are generally in the UV region, substitutions that extend conjugation of the system lead to shifting of the spectral feature⁵⁸.

In addition to the significant changes in the amide I and II bands, which we have already discussed above, two additional differences between the modified and unmodified FT-IR spectra are noteworthy. The modified peptide shows a new band at 2150 cm⁻¹, which is low for an S-H stretch. This feature was probed via DFT analytical frequency calculations of optimized unmodified and modified peptide models. Of the potential modified products evaluated, only the calculated spectrum of the thiooxazole species displayed an intense vibration at 1985 cm⁻¹ (Fig. S15) originating in the S-H stretching (Table S4). This predicted feature is somewhat lower than normal S-H stretches because the optimized structure reveals strong intramolecular hydrogen bonding which shifts this band to lower energy. The only other dipeptide model that displays a feature in this region is the 1H-azirine-2-thiol (see Table S3), which DFT calculations show to be 68.4 kcal/mol higher in energy than the thiooxazole. Literature precedents suggest that oxazoles would have strong features in the IR in the region of 1500–1600 cm^{−1 59}. The spectra of the modified SbtMa have features in this region of the FT-IR spectrum. While not diagnostic, these features are more consistent with a thiooxazole product.

In summary, we have methodically considered all reasonable structures that would account for the biochemical, chemical, mass spectrometric, NMR, UV-visible, FT-IR and computational observations. The aromatic thiooxazole is the structure that is most consistent with all of these observations.

DISCUSSION

The transformations that are catalyzed by members of the radical SAM superfamily often have no parallels in polar reaction spaces. It is perhaps not surprising that these enzymes are often found in the biosynthetic pathways of natural products because of their ability to catalyze challenging reactions and introduce rare functional groups that bestow selective advantage(s) to the producing organisms. SbtM is a radical SAM enzyme that was first identified bioinformatically in gene clusters that were presumed to be involved in the biogenesis of SeCys-containing RiPP natural products. The biological role of the mature SbtMa remains to be established.

A working mechanism for SbtM is shown in Fig. 9. Biochemical studies show that SbtM could support formation of at least two [4Fe-4S] clusters, one of which is likely coordinated by the conserved radical SAM cluster binding motif⁴. By analogy to other SPASM domain containing radical SAM enzymes^21,24–30, one of the remaining auxiliary clusters is likely proximal to the substrate in the active site and is involved in the chemistry, whereas the last cluster presumably plays an electron relay role. We propose that in the first step, H-atom transfer occurs from the β-carbon of Cys/SeCys at position 35 to the dAdo• to generate a substrate radical intermediate. This intermediate is then oxidized by the auxiliary cluster to form a thioformylglycine. This thioformylglycine intermediate likely engages the carbonyl oxygen and undergoes cyclization with concomitant deprotonation of the amide nitrogen. The resultant O,S-acetal is oxidized to a thione moiety by a H-atom transfer event likely from the β-carbon of the Cys residue at position 35, followed by a second oxidation. This thione intermediate undergoes tautomerization to form a stable aromatic thiooxazole. As discussed in the results, multiple peptide products are possible; however, we favor the thiooxazole on the basis of the observations reported in this manuscript. While oxazole moieties have been observed in natural products, the mechanism by which SbtM catalyzes the overall transformation is radically different. Oxazoles, such as those in microcin B17 are formed in two steps that involve activation of the peptide carbonyl oxygen followed by crosslinking to the Ser residue. In the second step, a flavoenzyme dehydrogenates the ring to form the oxazole.^60–63.

Figure 9. Mechanistic proposals for SbtM.

SbtM catalyzes the oxidative cyclization of Cys/SeCys residue in the SbtMa. In this mechanism, two successive H-atom transfer events are proposed, each of which is followed by oxidation via the auxiliary cluster. Note that the auxiliary cluster is presumably re-oxidized after each event to maintain an oxidized state.

The parallels between SbtM and the cyclodehydratase/dehydrogenase systems are striking. Both function by activating the substrate to present an electrophilic carbon center. However, in SbtM’s case, it activates the β-carbon of a Cys/SeCys residue by oxidation, whereas, the cyclodehydratase activates the amide carbonyl carbon. In SbtM, the oxidation to the oxazole is predicted to also be carried out by the SbtM using a second equivalent of SAM. Therefore, SbtM dispenses with the two-step pathway for oxazole formation by carrying out cyclization and oxidation in the same active site using H-atom activation. The auxiliary cluster of SbtM serves as a waystation for the reducing equivalents that are extracted from the substrate and are presumably shuttled away from the enzyme through the second auxiliary cluster.

In the nearly 20 years since the radical SAM superfamily was described, the repertoire of reactions catalyzed by these enzymes has expanded to include nearly every type of chemical reactivity. Members of this ancient superfamily have been discovered in biological niches that range from metabolism to biosynthesis of natural products that confer selective advantage to the host. The physiological role, and in fact, the identity of the natural product being produced by SbtM is not known. However, the reaction uncovered here is significant in several respects. The observation provides a parallel pathway for oxazole formation in biology that relies on an ancient and abundant iron-sulfur cofactor. The reaction catalyzed by SbtM is potentially the first characterized modified SeCys containing RiPP natural product. The 53 amino acid peptide characterized here should be considered a minimal substrate for SbtM. Future studies with longer substrates, particularly those that contain multiple potential modification sites, would be important in understanding the scope and role of the modification in nature. Finally, the mechanistic paradigm uncovered in this manuscript extends the repertoire of reactivity in the radical SAM superfamily.

ABBREVIATIONS

BSA

bovine serum albumin

CID

collision-induced dissociation

CV

column volume

dAdo•

5’-deoxyadenosine radical

dAdo

5’ deoxyadenosine

DFT

density functional theory

dT

dithionite

DTT

dithiothreitol

DNTP

dithiobis-5-nitropyridine

Fmoc

flurenylmethyloxycarbonyl

FTIR

Fourier-transform infrared spectroscopy

Gnd•HCl

guanidinium hydrochloride

HPLC

high performance liquid chromatography

HSQC

heteronuclear single quantum coherence

ICP-MS

Inductively coupled plasma-mass spectrometry

IPTG

isopropyl β-D-1-thiogalactopyanoside

KPi

potassium phosphate

LB

Lennox broth

Mob

methoxybenzyl

NMR

nuclear magnetic resonance

5-Npys

2-thio(5-nitropyridyl)

orf

open reading frame

PIPES

1,4-piperazinediethanesulfonic acid

PMSF

phenylmethanesulfonyl fluoride

Q-Sepharose

Quaternary Ammonium Sepharose

RiPP

ribosomally synthesized post-translationally modified peptide

SAM

S-adenosyl-l-methionine

SDS-PAGE

sodium dodecyl sulfate polyacrylamide gel electrophoresis

SECIS

selenocysteine insertion sequence

SPPS

solid phase peptide synthesis

TCA

trichloroacetic acid

TCEP

tris(2-carboxyethyl) phosphine

TEV

tobacco etch virus

Tris

tris(hydroxymethyl)aminomethane