The B₁₂-Independent Glycerol Dehydratase Activating Enzyme from Clostridium butyricum Cleaves SAM to Produce 5’-Deoxyadenosine and not 5’-Deoxy-5’-(methylthio)adenosine

Abstract

Glycerol dehydratase activating enzyme (GD-AE) is a radical S-adenosyl-l-methionine (SAM) enzyme that installs a catalytically essential amino acid backbone radical onto glycerol dehydratase in bacteria under anaerobic conditions. Although GD-AE is closely homologous to other radical SAM activases that have been shown to cleave the S-C(5’) bond of SAM to produce 5’-deoxyadenosine (5’-dAdoH) and methionine, GD-AE from Clostridium butyricum has been reported to instead cleave the S-C(γ) bond of SAM to yield 5’-deoxy-5’-(methylthio)adenosine (MTA). Here we re-investigate the SAM cleavage reaction catalyzed by GD-AE and show that it produces the widely observed 5’-dAdoH, and not the less conventional product MTA.

Graphical Abstract

The radical S-adenosyl-l-methionine (SAM) superfamily of enzymes is large and diverse, catalyzing a wide range of reactions throughout all kingdoms of life.^1–4 These reactions are thought to be initiated by a set of common steps that include coordination of SAM to the unique iron of a [4Fe-4S] cluster,^5–7 followed by reductive cleavage to generate the 5’-deoxyadenosyl (5’-dAdo•) radical intermediate responsible for abstracting a hydrogen atom from substrate.^{8, 9} The 5’-dAdo• is the same central intermediate involved in radical reactions catalyzed by B₁₂ radical enzymes; despite its important role in biological radical reactions, however, the 5’-dAdo• eluded direct observation and characterization until recently.^10–12 Interestingly, both radical SAM enzymes and B₁₂ radical enzymes harbor this reactive primary carbon radical as an organometallic species: the Co-C bond containing cofactor adenosylcobalamin in B₁₂ enzymes and the Fe-C bond containing Ω intermediate in radical SAM enzymes.^{13, 14} In each case, homolytic metal-carbon bond cleavage generates the key 5’-dAdo• radical intermediate,⁹ which has been shown to be exquisitely controlled in the radical SAM enzymes.¹⁵ Upon liberation from Ω, the 5’-dAdo• radical abstracts H• from native or alternative substrates to generate the initial substrate radical.^{8, 9, 16}

The radical SAM enzyme glycerol dehydratase activating enzyme (GD-AE) from Clostridium butyricum catalyzes the activation of the B₁₂-independent glycerol dehydratase (GD)¹⁷ by hydrogen atom abstraction from a conserved glycine residue (G763) to generate the active enzyme;¹⁸ GD-AE is thus a member of a subclass of radical SAM enzymes that function to activate glycyl radical enzymes.¹⁹ Unlike these other glycyl radical enzyme activating enzymes (GRE-AEs) however, GD-AE has been reported to produce 5’-deoxy-5’-(methylthio)adenosine (MTA), rather than 5’-deoxyadenosine (5’-dAdoH), as the adenosyl-containing product of SAM cleavage.²⁰ While photoinduced reductive cleavage of SAM can result in enzyme-dependent differences in regioselectivity,^{12, 21, 22} radical SAM enzyme catalysis typically involves S-C(5’) bond cleavage to generate 5’-dAdoH and methionine. In contrast, MTA would be formed if the S-C(γ) bond were reductively cleaved, with the other product of that cleavage being a 3-amino-3-carboxypropyl radical (ACP•) (Figure 1).¹ Thus the reported observation of MTA and not 5’-dAdoH during GD-AE catalysis suggested that an alternate S-C bond is cleaved during catalysis by this enzyme, and that ACP• rather than 5’-dAdo• abstracts the H-atom from GD to generate the active enzyme.²⁰

Figure 1:

Reductive cleavage pathways in radical SAM enzymes. SAM chelates the unique iron of a [4Fe-4S]⁺ cluster in the active site (center). Reductive cleavage of SAM generally involves homolytic S-C5’ bond cleavage to generate methionine and an intermediate 5’-deoxyadenosyl radical (5’-dAdo•, right). GD-AE has been reported¹⁹ to instead homolytically cleave the S-Cγ bond to generate methylthioadenosine (MTA) and an intermediate aminocarboxypropyl radical (ACP•, left).

A SAM-derived ACP• intermediate is also implicated in the diphthamide biosynthetic enzyme Dph2 from Pyrococcus horikoshii.²³ While Dph2 catalyzes radical-based chemistry using SAM and an iron-sulfur cluster, sequence and structural comparisons place it outside of the canonical radical SAM superfamily;^{1, 24} further, the ACP• generated does not carry out H-atom abstraction but rather adds to substrate and is subsequently incorporated into the product diphthamide.²³ During the Dph2-catalyzed reaction, the ACP• binds to the cluster at the unique iron to form an organometallic intermediate that is similar to the Ω organometallic intermediate observed for canonical radical SAM enzymes.^25–27 As of this writing, GD-AE remains an anomaly as the only canonical radical SAM enzyme reported to catalyze reductive cleavage of SAM at a location other than the S-C(5’) bond.

As part of our efforts to understand the factors that control regioselectivity and mechanism in radical SAM enzymes, we have examined the reaction catalyzed by GD-AE. In the course of these studies, we found that contrary to the prior report,²⁰ GD-AE in fact cleaves SAM at the S-C(5’) bond to generate methionine and 5’-dAdoH as products. While small amounts of MTA were observed in our assays, control experiments showed that the MTA was a contaminant that did not increase in an enzyme-dependent manner. Here we describe these results and their implications in regards to the consensus SAM cleavage mechanism that occurs within all characterized members of the radical SAM superfamily.

Materials and Methods

GD-AE Cloning and Protein Expression

The DNA sequence of Clostridium butyricum GD-AE was optimized using the DNAworks application,²⁸ synthesized using a commercial vendor (Integrated DNA Technologies, Coralville, IA, USA), and cloned into a modified pET42 vector (EMD Millipore, Billerica, MA, USA), which appends an N-terminal 10X-histidine-SUMO tag (the SUMO domain corresponds to the yeast Smt3 gene). The DNA sequence encoding the 10X-histidine-SUMO tag was a gracious gift from Dr. Christopher Bahl. The resulting plasmid was used to transform BL21-CodonPlus(DE3)-RIL cells (Agilent Technologies, Santa Clara, CA, USA) along with pDB1282, a modified pAra13 plasmid carrying the genes for 6 ISC operon genes from Azotobacter vinelandii (iscA, iscS, iscU, hscA, hscB, and fdx).²⁹ The pDB1282 plasmid was a gracious gift from Dr. John Peters. Cells were grown and harvested using a modified version of a previously published method.³⁰ Six 2.8 L Fernbach flasks, each with 1.5 L of LB medium (with 100 μg/mL ampicillin and 50 μg/mL kanamycin) were each inoculated with 50 mL of saturated overnight culture and grown at 37°C and 230 RPM. At a 600 nm absorbance of 0.3, a sterile filtered solution of 20% arabinose in water was added to a final concentration of 0.2% and the incubation temperature was decreased to 25°C. At a 600 nm absorbance of 0.6, a sterile filtered solution of 1 M IPTG (isopropyl β-D-thiogalactopyranoside) in water was added to a final concentration of 0.1 mM, and solutions of 300 mM ferrous ammonium sulfate and 300 mM L-cysteine, each dissolved in water, were added to final concentrations of 200 μM each. After an additional 4 hours of shaking at 25°C, ferrous ammonium sulfate and L-cysteine were again added to final concentrations of 200 μM each. At this point, 1 drop of Antifoam-SE was added to each flask, which were then sealed and incubated at 4°C for 16 hours while being actively sparged with nitrogen gas.

Alternatively, the optimized GD-AE coding sequence was inserted into pET-45b expression vector (GenScript) between the Acc651 and BamHI restriction sites, in-frame with a hexahistidine tag. The GD-AE/pET-45b was used to transform chemically competent Rosetta-pLysS(DE3) E. coli cells. Following transformation, a single colony was transferred into 50 ml LB (10 g/L Tryptone, 10 g/L KCl, 5 g/L Yeast Extract) liquid culture supplemented with 100 μg/ml Ampicillin and 33 μg/ml Chloramphenicol. The seed culture was used to inoculate 6 L TB media (24 g/L Yeast Extract, 12 g/L Tryptone, 89 mM KPi pH 7.0, 0.5% glycerol). The cultures were incubated at 37°C with 180 rpm shaking speed until the optical density of the cultures reached OD₆₀₀=0.6–0.8, then the cultures were transferred to 4°C for 3 hours. The protein expression was induced by the addition of 0.5 mM IPTG, furthermore the media were supplemented with 600 μM FAS and 600 μM L-cysteine, the cultures were shaken typically for 16–18 hours at 16°C. The cells were harvested by centrifugation (5000 rpm 10 min 14°C), the cell paste was frozen in liquid nitrogen and kept at −80°C until further use.

GD-AE Protein Purification

All steps were completed in an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI, USA) housed in a 4°C room.

For purification of GD-AE expressed with a 10xHis-SUMO tag, cell pellet from 27 L of culture was resuspended in 150 mL of lysis buffer (50 mM Tris, pH 7.5, 300 mM KCl, 1 mM phenylmethylsulfonyl fluoride, 250 μg/mL lysozyme, 25 μg/mL deoxyribonuclease I, 25 μg/mL ribonuclease A, 1% TritonX-100, 100 μM dithiothreitol (DTT)) and incubated with stirring for 1 hour. The lysis solution was clarified by centrifugation for 60 minutes at 38,000 g and the supernatant was flowed over a 5 mL nickel-NTA sepharose column, which was pre-equilibrated with loading buffer (50 mM Tris, pH 7.5, 300 mM KCl). The nickel column was next washed with 50 mL of imidazole wash buffer (50 mM Tris, pH 7.5, 300 mM KCl, 30 mM imidazole), prior to elution with 50 mL of elution buffer (50 mM Tris, pH 7.5, 300 mM KCl, 400 mM imidazole). The brown eluate was immediately added to a 90 mL Sephadex G25 desalting column pre-equilibrated with storage buffer (50 mM Tris, pH 7.5, 300 mM KCl, 10% glycerol). The desalted protein was immediately snap frozen in liquid nitrogen. The next day the protein was thawed and DTT added to a final concentration of 100 μM. Next, a 1/10 molar equivalent of purified SUMO protease was added to the protein solution, which was then incubated for 3 hours at 4°C. The DNA vector encoding the SUMO protease (corresponding to the yeast Ulp1 gene) was a gracious gift from Dr. Christopher Bahl at the University of Washington. Following the 3 hour incubation, the cleavage reaction was flowed over a fresh 5 mL nickel-NTA sepharose column pre-equilibrated with storage buffer. The brown flow-through was aliquoted and snap frozen. Protein concentration was quantified using a modified version of the method of Bradford³¹ and iron content was determined using a SpectrAA 220 Fast Sequential Atomic Absorption Spectrometer (Varian, Palo Alto, CA, USA).

For purification of the 6xHis-tagged protein lacking the SUMO tag, the frozen cell paste was transferred into the Coy chamber where it was resuspended in lysis buffer (75 mM Tris-HCl pH 9.0, 300 mM KCl, 20% glycerol, 25 mg/ml MgCl₂, 0.2 mg/ml DNAse, 0.2 mg/ml RNAse, 0.2 mg/ml lysozyme and 1% Triton X-100. The cells were lysed for 20–30 minutes at room temperature. The chemical-enzymatic lysis was followed with mechanical cell disruption. The partially lysed cell suspension was transferred to a 4°C Coy chamber were the suspension was sonicated (40% amplitude, 5 sec on, 10 sec off, 5 minutes in total). The sonicated suspension was centrifuged for 45 minutes with 15000 rpm at 14°C. The resulting supernatant was loaded onto Ni-NTA columns which were equilibrated with 75 mM Tris-HCl pH 9.0, 300 mM KCl, 20% glycerol. The column was washed extensively with equilibration buffer until the UV absorption reached nearly zero, then the column was washed with 75 mM Tris-HCl pH 9.0, 300 mM KCl, 20% glycerol, 100 mM imidazole to eluate the non-specifically binding proteins. The 100 mM imidazole wash was followed by the elution of GD-AE with 75 mM Tris-HCl pH 9.0, 300 mM KCl, 20% glycerol, 500 mM imidazole. The excess imidazole was removed using 10 kDa Amicon filters. The purified protein was aliquoted, flash frozen in liquid nitrogen and stored at −80°C until further use.

Glycerol Dehydratase Cloning and Protein Expression

The codon optimized Glycerol dehydratase (GD) gene (generated as described above for GD-AE) from Clostridium butyricum was inserted into a pET-45b vector (Genscript) between the Acc65I and BamHI restriction sites. The GD/pET-45b was used to transform chemically competent Rosetta-pLysS(DE3) E. coli cells. Following transformation, a single colony was transferred into 30 ml LB (10 g/L Tryptone, 10 g/L KCl, 5 g/L Yeast Extract) liquid culture supplemented with 100 μg/ml Ampicillin and 33 μg/ml Chloramphenicol and grown overnight (37°C, 180 rpm shaking). The seed culture was used to inoculate 2 L TB media (24 g/L Yeast Extract, 12 g/L Tryptone, 89 mM KP_i pH 7.0, 0.5% glycerol). The cultures were incubated at 37°C with 180 rpm shaking speed until the optical density of the cultures reached OD₆₀₀=0.5–0.6, then the cultures were placed in an ice-water bath for 1 hour. Protein expression was induced by the addition of 0.5 mM IPTG. Cells were then shaken at 18°C, 180 rpm for 21 hours. Two hours after induction, 1 mL of Antifoam-SE was added to each flask. The cells were harvested by centrifugation (6000 rpm, 15 min, 4°C) and the cell paste was frozen in liquid nitrogen and kept at −80°C until further use.

Glycerol Dehydratase Protein Purification

For purification of the 6xHis-tagged GD, the frozen cell paste was transferred into the Coy chamber where it was resuspended in lysis buffer (50 mM HEPES, pH 7.5, 100 mM KCl, 10% glycerol, 1 mM PMSF, 1.15 mg/ml MgCl₂, 0.01 mg/ml DNAse, 0.45 mg/ml lysozyme, and 10 mg/mL Triton X-100). The cells were lysed for 1 hour at room temperature and then sonicated (40% amplitude, 5 seconds on, 15 seconds off, 5 minutes total time). The sonicated suspension was centrifuged for 45 minutes at 15000 rpm and 4°C. The resulting supernatant was loaded onto a Ni-NTA column that was equilibrated with 10 column volumes of Buffer A (50 mM HEPES, pH 7.5, 100 mM KCl, 10% glycerol) and 10 mM imidazole. The column was then washed with Buffer A with increasing amounts of imidazole in a step-wise manner starting with 50 mM, then 100 mM, 150 mM, and finally 500 mM to elute any remaining protein. Each fraction was collected separately and subsequently flash frozen in liquid nitrogen and stored at −20°C overnight. Each fraction was then thawed and analyzed by SDS-PAGE for which fractions contained the GD. Each fraction contained the GD but with decreasing amounts of contaminating proteins (Figure S3). Fractions eluted with 150 mM and 500 mM imidazole were pooled and de-salted using a 50 kDa MWCO Amicon spin filter by diluting with Buffer A 10x and centrifuging at 6000 rpm for a total of 3 times. Lastly, the concentrated and de-salted GD was centrifuged in an Eppendorf tube at 13,000 rpm, 10 minutes to remove any precipitated protein. The GD was then aliquoted, flash frozen in liquid nitrogen, and stored at −80°C until further use. GD was quantified by its absorbance at 280 nm using the molar extinction coefficient ε_280nm = 76,210 M⁻¹ cm⁻¹, calculated using the Protparam tool (https://web.expasy.org/protparam/) assuming all cystines are reduced.

Cloning and Purification of CsdA

The DNA sequence of Escherichia coli cysteine desulfurase (ecCsdA) was codon optimized and synthesized using a commercial vendor (Genscript) and cloned into a pET-19b vector using the NdeI and BamHI restriction sites. The resulting plasmid was transformed into a BL21(DE3)pLysS cell line. Six 2.8 L Fernbach flasks, each with 1.0 L of LB medium (supplemented with 25 mM KPi, pH = 7.6, 5% glucose, 100 μg/mL ampicillin, and 33 μg/mL chloramphenicol) were each inoculated with 30 mL of saturated overnight culture and grown at 37°C and 180 RPM. Once the OD₆₀₀ reached 0.3, the cells were placed in a 4°C fridge for 1 hour. The cells were then induced with IPTG (0.5 mM final concentration) and allowed to grow for 3 hours at 37°C and 180 RPM shaking. Cells were harvested by centrifugation (6000 RPM, 4°C, 20 minutes) and the cell paste was frozen in liquid nitrogen and stored at −80°C until further use.

For purification of ecCsdA with a 6xHis Tag, the cell paste was brought into an anaerobic COY chamber and resuspended in lysis buffer (75 mM HEPES, pH = 8.0, 300 mM KCl, 5% glycerol, 0.2 mg/ mL DNAse, RNAse, and lysozyme, 25 mg/ mL MgCl₂, and 1% Triton X-100) at a ratio of 2 mL/ g cell paste and incubated at room temperature for 40 minutes. The cell lysate was then moved to a 4°C anaerobic COY chamber for sonication (40% amplitude, 5 sec on, 10 sec off, 5 minutes in total). The lysate was clarified by centrifugation for 45 minutes at 15,000 rpm, 4°C and the supernatant was loaded onto a 5 mL nickel-NTA Sepharose column, which was pre-equilibrated with 5 column volumes of equilibration buffer (75 mM HEPES, pH 8.0, 300 mM KCl, 5% glycerol). The nickel column was washed with 100 mL of imidazole wash buffer (75 mM HEPES, pH 8.0, 300 mM KCl, 5% glycerol, 100 mM imidazole), prior to elution with 50 mL of elution buffer (75 mM HEPES, pH 8.0, 300 mM KCl, 5% glycerol, 400 mM imidazole). The eluate was concentrated to 4 mL using a 30 kDa Amicon spin filter, then de-salted by passing over a NAP25 column equilibrated in equilibration buffer. The de-salted ecCsdA was aliquoted, frozen in liquid nitrogen, and stored at −80°C.

Quantification of Protein, Iron, and Sulfide

Purified GD-AE protein was quantified following the procedure by Barr et al.³² with the minor modification of using 10 kDa MWCO Amicon spin filters to buffer exchange and concentrate the denatured protein. To buffer exchange, the protein was concentrated to 200 μL, diluted to 500 μL with 6 M guanidinium HCl, and concentrated again to 200 μL, repeating for a total of 7 times. The concentrated protein solution was used for downstream analyses. This procedure provided a correction factor of 0.851 ± 0.004.

Iron was quantified following the procedure by Németh et al.³³ with modifications. An iron AA standard (Bicca, 1000 ppm in 3% HCl) was used to generate a calibration curve by first diluting to 0.5 nmol/ μL in water. Following the addition of perchloric acid and water, each solution was allowed to sit at room temperature for 30 minutes. The solutions were then centrifuged at 13,000 rpm for 15 minutes at room temperature. After the addition of bathophenanthroline disulphonic disodium salt, ammonium acetate, and ascorbic acid, the solutions reacted for 20 minutes. The UV-Visible spectrum of each sample was measured with MQ water as a blank. The absorbance at 535 nm was plotted against nmol of iron to generate a calibration curve that was used to calculate the amount of iron in each protein sample.

Acid-labile sulfide was quantified following the procedure by Broderick et al.³⁴ with minor modifications. Each sample and standard were brought to a volume of 100 μL using filtered (0.2 μm) deionized water instead of pH 8 water. Also, following the addition of 1% ZnOAc and 12% NaOH, each sample was allowed to sit for at least 15 minutes before the addition of N,N-dimethyl-p-phenylenediamine (0.1% in 5 M HCl) and FeCl₃ (23 mM in 1.2 M HCl).

Enzymatic Reconstitution of Iron-Sulfur Clusters in GD-AE

The purified GD-AE was brought into an anaerobic COY chamber to thaw, along with E. coli cysteine desulfurase (ecCsdA). Ferrous ammonium sulfate hexahydrate (FAS), L-cysteine hydrochloride monohydrate, and dithiothreitol (DTT) were brought into the COY chamber as powders and brought up to 100 mM stock concentrations in Buffer A (80 mM Tris HCl, pH = 9.0, 150 mM KCl, 20% glycerol). The purified GD-AE was diluted to 100 μM with Buffer A in a scintillation vial with a small stir bar. FAS, L-cysteine, and DTT were added to the vial to final concentrations of 1 mM. Then ecCsdA was added to the vial to a final concentration of 2 μM to initiate the reconstitution. The reaction mixture was stirred gently for 2.5 hours, after which the contents were centrifuged at 6000 rpm for 10 minutes to remove any precipitates. The supernatant was then de-salted by passing over a 90 mL column packed with Sephadex G-25 resin equilibrated with 2 column volumes of Buffer A. The eluate was collected and concentrated down to approximately 1 mL with a 30 MWCO Amicon spin filter by centrifuging at 6000 rpm for 3 minutes, 3 times, with gentle agitation after each centrifugation. The reconstituted GD-AE was then aliquoted and frozen in liquid nitrogen, and stored at −80°C.

Electron Paramagnetic Resonance Spectroscopy

Electron paramagnetic resonance (EPR) spectroscopy followed general procedures described previously.³⁵ Continuous wave EPR X-band EPR spectroscopy was performed using a Bruker EMX EPR spectrometer equipped with a Bruker/ColdEdge (Sumitomo Cyrogenics) 10 K waveguide cryogen free system with an Oxford MercuryiTC controller unit and helium Stinger recirculating unit (Sumitomo Cyrogenics, ColdEdge Technologies). For analysis of iron-sulfur clusters, GD-AE was left non-reduced, reduced with 0.60 mM Na-DT for 10 minutes, or reduced with 0.60 mM Na-DT and incubated with 0.90 mM SAM in the glove box. Each sample was then transferred to EPR tubes, capped, and flash frozen in liquid nitrogen outside the glove box. Each sample was measured at varying temperatures and powers as indicated in the figure captions. Metal-based signals were recorded with a microwave frequency of 9.38 GHz, microwave power of 1 mW, receiver gain of 10000, modulation frequency of 100 kHz, modulation amplitude of 10 G, and time constant of 168.84 ms, with a scanning range from 3000 to 4000 G. The glycyl radical generated on GD by GD-AE was recorded at 75 K using a microwave frequency of 9.38 GHz, microwave power of 212 μW, 2 G modulation amplitude, and 100 kHz modulation frequency. The scan width is 300 G centered at 3350 G. EPR simulations were performed with Easyspin³⁶ within the Matlab R2020b software suite (Mathworks inc.). All spectra were cavity subtracted and baseline corrected.

Enzyme Activity Assays with GD Peptide Substrate

All enzyme activity assays were carried out in a Unilab glove box (O₂ ≤ 1 ppm; MBraun, Stratham, NH, USA) or an anaerobic chamber (O₂ ≤ 20 ppm; Coy Laboratory Products, Grass Lake, MI, USA) at 24 – 31°C. SAM was prepared enzymatically as previously described.⁶ The GD 7-mer peptide (RVAGYSA) corresponding to amino acid residues 760 – 766 and containing the site of H-atom abstraction by GD-AE (G763) was purchased from GenScript. As-isolated GD-AE (8.0 or 16 μM) was combined with SAM (1.0 – 2.1 mM), DTT (5.0 mM), and the 7-mer target peptide from glycerol dehydratase (3.43 mM) in assay buffer composed of 50 mM Tris-Cl pH 7.5, 300 mM KCl, 10% v/v glycerol. The assay was initiated by the addition of dithionite (2.0 – 2.1 mM). At indicated time points (0.5, 7, 15, 30, and 60 minutes), reaction aliquots were removed and quenched by addition of 1/3 volume of 1 M sodium acetate, pH 4.5 and agitation by vigorously pipetting air into the solution. Quenched aliquots were stored frozen at −20°C for 16 to 24 hours and were then centrifuged at 20,000 g for 10 minutes. Each supernatant was injected onto a Phenomenex Kinetex 5μ PFP 100Å column equilibrated in 98% solvent A (water with 0.1% acetic acid) and 2% solvent B (acetonitrile with 0.1% acetic acid) on a Series 1100 HPLC (Agilent Technologies, Santa Clara, CA, USA). The column was washed with 2% solvent B at 1 mL/minute for 3.5 minutes, after which a linear gradient was begun, reaching 60% solvent B at 11 minutes. The 60% solvent B was maintained for 3 minutes prior to re-equilibrating the column with 2% solvent B. SAM eluted in the void volume, while 5’-dAdoH eluted with a retention time near 7.8 minutes and MTA eluted near 9.0 minutes. SAM, 5’-dAdoH, and MTA were detected by their absorbance at 254 nm and dAdo and MTA were identified by comparison of their retention times to those of 5’-dAdoH and MTA standards that were run separately, by standard addition of 5’-dAdoH and MTA to assay samples, and by mass spectrometry. Quantification was carried out by integrating peaks in the ChemStation software, and standard curves were fit using linear regression in Excel. MTA and 5’-dAdoH were quantified by comparison to standard curves of their respective standard solutions. To generate the chromatograms in Figure 5, the raw chromatographic data was processed in Excel as follows: a blank run was subtracted from each data set and the SAM standard was baseline corrected (as it was collected in a different run than the other samples and blank run).

Figure 5.

Kinetics of 5’-dAdoH and MTA production by GD-AE. A. 5’-dAdoH production as a function of time. B. MTA production as a function of time. Assays are: control containing SAM (2.0 mM) in buffer (green); GD peptide (3.4 mM) and SAM (2.0 mM) in buffer (purple); GD-AE (16 μM) and SAM (2.0 mM) in buffer (blue); complete reaction containing GD-AE (16 μM), SAM (2.0 mM), and GD peptide (3.4 mM) (orange). All assays contained 2.0 mM dithionite. Data shown are the averages of at least 5 independent assays per condition.

Mass spectrometry of 5’-dAdoH

An HPLC run like that depicted in Figure 5G was fractionated and the fraction corresponding to the 5’-dAdoH peak was dried, resuspended in 0.1% formic acid in water, and run on a Waters Acquity BEH C18 reverse phase (100 × 2.1) mm 1.7 μm column on an Agilent Infinity 1290 UPLC coupled to an Agilent 6538 Q-TOF mass spectrometer. The mobile phases were A: 0.1% formic acid in water, B: 0.1% formic acid in acetonitrile with a flow rate of 0.130 μL/min. The mass spectrometer was run in ESI mode, ion mode positive, scanning from 50–1000 m/z with a scan rate of 1 spectra/sec, Vcap of 3500 V, Nebulizer at 55 psig, Fragmentor voltage at 100 V, octapole RF at 500 V. The 5’-dAdoH had an observed mass of 252.1 g/mol, in close agreement with the calculated mass of 251.2 g/mol and with the observed mass of 252.1 g/mol of a pure 5’-dAdoH standard also analyzed by LC-MS.

Activation of GD by GD-AE

In the glove box, 22.3 μM GD and 37.7 μM GD-AE were exchanged into assay buffer (50 mM HEPES, pH = 7.5, 100 mM KCl) using a 10 kDa MWCO Amicon spin filter to remove any glycerol from the storage buffers. They were then transferred to a clean Eppendorf tube where 889 μM Na-DT was added. GD-AE was allowed to reduce for 8 minutes before the addition of 905 μM SAM to initiate the reaction. The reaction mixture was transferred to an EPR tube and quenched by flash freezing in liquid nitrogen after 30 minutes. The sample was then analyzed by CW X-band EPR spectroscopy for glycyl radical detection. Following analysis by EPR spectroscopy, the sample was thawed anaerobically. A 20 μL aliquot of the sample was removed and quenched with ½ volume of 1 M sodium acetate, pH = 4.5 and centrifuged at 14,000 rpm for 15 minutes to remove precipitated protein. The supernatant was transferred to a clean HPLC vial and injected onto a Shimadzu LC-2060C liquid chromatograph equipped with a reversed phase column (Phenomenex Kinetex F5, 2.6 μm, 100 Å, 100 × 4.6 mm). The HPLC method was the same as used for the enzyme activity assay except a washing step was included in which after the 3 minutes at 60% B, the concentration of B was increased to 90%, held for 2 minutes, and then decreased to 2% B and re-equilibrated for the next run. Detection was performed at 254 nm. 5′-dAdoH eluted at ~6 minutes and MTA at ~6.8 minutes. Analytes were verified by the method of standard addition and quantified by comparison to standard curves of authentic standards. Each HPLC trace was baseline corrected by subtracting a blank run.

Generation of GD-AE Homology Model

The amino acid sequence of Clostridium butyricum glycerol dehydratase activating enzyme (GD-AE) was aligned to the amino acid sequence of Escherichia coli pyruvate formate-lyase activating enzyme (PFL-AE) using the Muscle algorithm in Geneious and a BLOSUM matrix (Geneious version 6.1.7, http://www.geneious.com).³⁷ The resulting alignment was then manually edited. A model was generated wherein the structure of Clostridium acidiurici 2[4Fe-4S] ferredoxin (PDB ID 2fdn)³⁸ was inserted between residues 46 and 47 of the structure of PFL-AE bound to the PFL peptide (PDB ID 3cb8).³⁹ The orientation of the ferredoxin domain relative to PFL-AE was chosen by aligning the endpoints of the ferredoxin domain to the insertion site on PFL-AE using PyMOL (The PyMOL Molecular Graphics System, Version 1.3 Schrödinger, LLC.). Based on the edited sequence alignment, the PFL-AE-ferredoxin fusion model was modified to the sequence of GD-AE using Foldit.⁴⁰ The sequence of the PFL peptide was modified to match the GD peptide sequence and this peptide was present throughout all subsequent modeling steps. The ferredoxin and radical SAM domains of the resulting model were then separated into two distinct chains and the rigid body relationship between the domains, as well as all torsion angles were minimized using the FastRelax application in RosettaScripts.^{41, 42} Intra-molecular distance constraints were applied to prevent excessive backbone change within each domain, but no constraints were applied between the domains. The model in which the ferredoxin domain most closely associated with the radical SAM domain was recombined into a single chain and used for further modeling. The connection points between the ferredoxin domain and the radical SAM domain were remodeled using the Rosetta Remodel application.⁴³ Insertions and deletions in the GD-AE sequence relative to the PFL-AE sequence were also corrected at this point using Rosetta Remodel. The resulting model with the best core packing according to the RosettaHoles algorithm⁴² was used in the following steps. All torsion angles within the model were subjected to a second round of minimization using the FastRelax application in RosettaScripts and the resulting model with the best total score was used in the final step.

To generate the final GD-AE homology model, SAM and [4Fe-4S] cluster molecules were added to the model and the positions of the SAM and nearby protein side chains (Glu 158, Met 179, Arg 218, Leu 251, and Tyr 256) were modified in Coot⁴⁴ to alleviate clashes between SAM and the protein. GD-AE sequence covariance constraints from the Gremlin server⁴⁵ were used to validate inter-residue contacts throughout the homology model as well as the orientation of the ferredoxin domain relative to the radical SAM domain, although there were few pairs of co-varying residues between the ferredoxin and radical SAM domains.

Results

Purification and Characterization of GD-AE

GD-AE was expressed in E. coli as a 10X-His-SUMO fusion for the purposes of purification; following purification, the 10X-His-SUMO tag was removed prior to further experiments (Figure S1).⁴⁶ In order to optimize iron-sulfur cluster incorporation during protein expression, this GD-AE/10x-His-SUMO fusion was co-expressed with the ISC operon from Azotobacter vinelandii.²⁹ The resulting purified GD-AE contained ~6 irons per protein monomer, less than the 12 irons per monomer expected given the one CX₃CX₂C radical SAM motif and two CX₂CX₂CX_NC ferredoxin motifs present in the GD-AE sequence. Expression and purification of the GD-AE/6x-His fusion (lacking the SUMO tag, Figure S2) also provided protein with less than the expected 12 irons per protein. This protein was enzymatically reconstituted using CsdA, cysteine, and iron to generate protein containing 10.59 ± 0.18 irons and 12.03 ± 0.53 sulfides per protein. Given that the ratio of iron to sulfide is close to 1:1, it is likely that all of the iron bound to GD-AE is present in iron-sulfur clusters.

Samples of reconstituted GD-AE were examined by EPR spectroscopy. The reconstituted protein shows a small signal characteristic of a [3Fe-4S]⁺ cluster (Figure 2, top). Upon reduction of the reconstituted GD-AE with sodium dithionite (DT), multiple overlapping axial or near-axial EPR signals appear that are consistent with S = ½ [4Fe-4S]¹⁺ clusters (Figure 2). These multiple overlapping EPR signals are consistent with the fact that GD-AE has three cysteine motifs, with the radical SAM motif expected to bind a site-differentiated [4Fe-4S] cluster and the two ferredoxin motifs each expected to bind a ferredoxin-type [4Fe-4S] cluster. Temperature-dependence of the EPR signals (Figures 3 and S4) shows a rapid loss of signal intensity above 12 K, with essentially no signal observed at 40 K; this behavior is consistent with the fast-relaxing properties typically associated with [4Fe-4S]⁺ clusters. Examination of the power-dependence of these EPR signals (Figure S5) reveals behavior that is also fully consistent with their assignment to [4Fe-4S]⁺ clusters.

Figure 2. EPR spectroscopy of GD-AE.

EPR samples were prepared using 6x-His-GD-AE protein (57 μM, 10.59 ± 0.18 iron atoms per protein). Top, GD-AE as-reconstituted. Middle, GD-AE reduced with dithionite. Bottom, GD-AE reduced with dithionite followed by addition of SAM. Sample components, when present, were at the following concentrations: SAM (0.90 mM) and dithionite (0.60 mM). EPR parameters: 12 K, 9.38 GHz, 1 mW, 100 kHz modulation frequency, 10 G modulation amplitude.

Figure 3. Temperature-dependent EPR of reduced GD-AE.

EPR spectra of reduced GD-AE (57 μM, 10.59 ± 0.18 Fe/protein) in the absence (left) or presence (right) of SAM were recorded at the temperatures indicated. The loss of signal intensity at temperatures above 12 K is indicative of the presence of rapidly-relaxing [4Fe-4S]⁺ clusters. Sample components, when present, were at the following concentrations: SAM (0.90 mM) and dithionite (0.60 mM). EPR parameters: 12 K, 9.38 GHz, 1 mW, 100 kHz modulation frequency, 10 G modulation amplitude.

Addition of SAM to the DT-reduced enzyme perturbs the EPR signals, while also causing a decrease in the total spin intensity (Figure 2). The loss of some signal intensity upon addition of SAM to the reduced GD-AE is consistent with nonproductive reductive cleavage of SAM at the radical SAM [4Fe-4S]⁺ cluster, which would cause oxidation to the EPR-silent [4Fe-4S]²⁺ state. The ferredoxin clusters and any radical SAM cluster that does not get oxidized will continue to contribute to the observed EPR signal for this sample, however the EPR signal of the radical SAM cluster is expected to be perturbed by the binding of SAM. The observed spectrum therefore remains complex as a result of multiple EPR-active clusters. The temperature relaxation (Figures 3 and S4) and the power saturation (Figure S4) behavior of the reduced GD-AE with SAM added are consistent with the presence of [4Fe-4S]⁺ clusters.

Assays of GD-AE for SAM Cleavage

GD-AE was assayed for the formation of SAM-derived 5’-dAdoH and MTA using an HPLC-based method. For these assays, a heptamer peptide (RVAGYSA) corresponding to the glycine loop of GD¹⁸ and containing the target G763 was used in place of the full GD substrate. A heptamer peptide was previously shown to substitute for the full glycyl radical enzyme substrate when assaying the glycyl radical activating enzyme PFL-AE.⁴⁷ Our results in Figure 4 demonstrate that the same is true for GD-AE: the corresponding GD heptamer peptide used herein acts as a substrate in stimulating reductive SAM cleavage by GD-AE. The complete reaction (Figure 4G) shows the production of a significant amount of 5’-dAdoH when GD-AE, SAM, 7-mer peptide, and dithionite are incubated for 60 minutes. A small peak corresponding to MTA can also be seen for this reaction, however its peak height is similar to the MTA peaks observed for reactions lacking GD peptide (4F), lacking GD-AE (4E), and lacking both GD-AE and GD peptide (4D). We note that the SAM used in the assays (prepared in-house using published protocols)³⁵ does contain a small amount of MTA contamination, a degradation product of SAM (Figure 4C). The presence of a comparably small MTA peak for all these samples (including SAM alone), in contrast to the dramatically increased 5’-dAdoH peak for the complete reaction relative to all others, indicates that the GD-AE catalyzed reaction involves cleavage of the S-C(5’) bond of SAM, contrary to a previously published report.²⁰ This GD-AE catalyzed reductive homolytic cleavage of the S-C(5’) bond of SAM is presumably accompanied by 5’-dAdo• mediated abstraction of H• from the target glycyl residue of the peptide substrate to generate 5’-dAdoH and Gly•; the resulting glycyl radical is not observed, likely due to rapid solvent quenching of the exposed peptide backbone radical.

Figure 4.

HPLC based endpoint assays of GD-AE. Chromatograms A – C are standards, while chromatograms D – G are 120 minute reactions as indicated. A. 1 μM 5’-dAdoH. B. 1 μM MTA. C. 5 mM SAM. D. Control reaction containing SAM (2.0 mM) and dithionite (2.0 mM) but lacking GD-AE and GD peptide. E. Control reaction containing GD peptide (3.4 mM), SAM (2.0 mM), and dithionite (2.0 mM), but lacking GD-AE. F. Reaction of GD-AE (16 μM), SAM (2.0 mM), and dithionite (2.0 mM), in the absence of GD peptide, to probe uncoupled SAM cleavage. G. Complete reaction containing GD-AE (16 μM), GD peptide (3.4 mM), SAM (2.0 mM), and reductant (2.0 mM). The 5’dAdoH peak from HPLC was collected and analyzed by LC-MS, confirming the identity of the 5’-dAdoH.

We also evaluated the rates of change of MTA and 5’-dAdoH concentrations over a 2-hour assay, and these results further support our conclusion that GD-AE catalyzes a conventional SAM reductive cleavage in which the S-C(5’) bond is broken (Figure 5). We observed that GD-AE in the absence of the GD peptide catalyzes a slow nonproductive cleavage of SAM (0.0098 ± 0.002 moles of dAdo/mole of GD-AE/minute), consistent with the uncoupled SAM cleavage observed for many other radical SAM enzymes.¹ In the presence of the GD peptide, the rate of 5’-dAdoH production by GD-AE increases to 0.16 ± 0.02 moles of 5’-dAdoH/mole of GD-AE/minute, consistent with observations for other radical SAM enzymes,¹ where SAM cleavage rates are significantly increased in the presence of substrate (Figure 5A). In contrast, although a small increase in MTA is observed over the course of the 2-hour assay, the same increase is observed in all samples, regardless of the presence or absence of GD peptide and GD-AE (Figure 5B).

Activation of GD by GD-AE

In order to examine the reactivity of GD-AE with its native substrate GD, the GD-AE sample used for EPR spectroscopic characterization (Figure 2) was thawed and mixed with purified GD (final concentrations were 38 μM GD-AE and 22 μM GD). This mixture was exchanged into assay buffer, then SAM and dithionite were added, and the sample was frozen for EPR spectroscopy after 30 min. The EPR spectrum recorded at 75 K (Figure 6) shows the presence of a doublet signal as expected for a glycyl radical, demonstrating the activation of the GD by GD-AE. This sample was then analyzed for production of 5’-dAdoH and MTA, to determine whether turnover of the native substrate GD provided the same SAM cleavage products as observed during turnover of the GD peptide. As can be seen in Figure 7, HPLC reveals the presence of 5’-dAdoH in the GD/GD-AE/SAM EPR sample; this 5’-dAdoH peak is absent in an identical sample prepared without GD-AE. MTA, a common product of SAM degradation, is also observed in both samples but is unchanged in the sample containing GD-AE. Together, these results substantiate that GD-AE cleaves the S-C5’ bond of SAM to form 5’-dAdoH during the activation of GD, as it does when using the 7-mer GD peptide as a model substrate.

Figure 6. EPR spectroscopy of activated GD.

EPR sample (black) contained 38 μM 6x-His-GD-AE, 22 μM GD, 0.9 mM SAM, and 0.89 mM DT. The signal was simulated (red) using g = 2.0045, 2.0035, 2.0025 and A = 40 MHz. EPR parameters: 75 K, 9.38 GHz, 0.21 mW, 100 kHz modulation frequency, 2 G modulation amplitude.

Figure 7. HPLC Chromatogram of GD Activation Reaction.

The HPLC chromatogram of the full reaction containing GD, GD-AE, SAM, and DT (top) shows a 5’-dAdoH peak that is absent in the sample lacking GD-AE. MTA, a common decomposition product of SAM, is observed with equal intensity in both samples, indicating it is not enzymatically formed by GD-AE.

Discussion

Radical SAM enzymes use a site-differentiated [4Fe-4S] cluster to initiate the regioselective reductive cleavage of SAM to generate an organic radical intermediate. In these enzymes, SAM coordinates the unique iron through its amino and carboxylate moieties,⁷ with the sulfonium sulfur sitting sufficiently close to the cluster to exhibit orbital overlap.⁶ In most of the radical SAM enzymes characterized to date, it is the S-C(5’) bond of SAM that is specifically cleaved, giving rise to methionine and a 5’-deoxyadenosyl radical intermediate, the latter of which converts to 5’-dAdoH upon H-atom abstraction from substrate.¹ Recent insights into the origins of regioselectivity in the reductive cleavage of SAM in radical SAM enzymes point to the Jahn-Teller effect associated with the one electron reduced sulfonium center of SAM coupled to active-site forces as the key determinants of S-C bond cleavage regioselectivity.¹²

The previous observation of alternate SAM cleavage products for GD-AE was surprising given the significant sequence homology to other well-characterized glycyl radical enzyme activating enzymes,¹⁹ notably PFL-AE,⁴⁸ 4HPD-AE⁴⁹ and aRNR-AE,⁵⁰ which have been shown to produce the standard 5’-dAdoH product of reductive SAM cleavage. Because there is no X-ray crystal structure of GD-AE available, we generated a homology model to provide structural insight. The homology model (Figure 8) utilized PFL-AE (PDB ID 3cb8)^{39, 51} to model the radical SAM domain of GD-AE, which contains the active site and catalytic iron-sulfur cluster, and a ferredoxin (PDB ID 2fdn)³⁸ as a template for the ferredoxin domain insertion of GD-AE.

Figure 8.

Homology model of GD-AE. A. Model of GD-AE (green cartoon) bound to its substrate peptide from GD (cyan cartoon). The crystal structures of the templates used to generate the homology models are superimposed (E. coli PFL-AE, PDB ID 3cb8, magenta cartoon, for the radical SAM domain and Clostridium acidiurici 2[4Fe-4S] ferredoxin, PDB ID 2fdn, yellow cartoon, for the N-terminal ferredoxin domain insertion). The [4Fe-4S] clusters, ligating cysteines, and SAM molecule are shown as sticks. B. Close up view of the active site of the GD-AE homology model (green sticks) superimposed onto the structure of PFL-AE (magenta sticks) bound to SAM (purple sticks).

The homology model supports a close structural homology between the radical SAM domain of GD-AE and PFL-AE. The CX₃CXϕC [4Fe-4S] cluster-binding motif in GD-AE bears a tryptophan rather than tyrosine at the ϕ position, a substitution occasionally observed in radical SAM enzymes.^{52, 53} In addition, GD-AE has tyrosine and histidine residues in place of the two histidine residues in PFL-AE that pack onto the face of the adenine ring. Together, these differences could modestly perturb the position of the adenine ring of SAM in GD-AE relative to that in PFL-AE. However, GD-AE retains other key active site residues, including Asp181 and Lys183 in the same structural positions as the Asp129 and Lys131 in PFL-AE, which interact with the ribose hydroxyl groups and methionine carboxylate of SAM. Overall, the homology model is consistent with an active site architecture and a SAM binding configuration in GD-AE that is similar to that observed in PFL-AE (Figure 8).

In order to probe the active-site iron-sulfur cluster and catalytic activity of GD-AE, we expressed and purified the protein and carried out EPR spectroscopic studies. The reconstituted protein exhibits a small signal attributed to [3Fe-4S]⁺ clusters, a common observation for radical SAM enzymes due to the labile unique iron of the [4Fe-4S] cluster.¹ Upon reduction with DT, the [3Fe-4S]⁺ signal disappears and signals characteristic of [4Fe-4S]⁺ clusters appear (Figure 2). Addition of SAM leads to perturbations of this signal and some loss of signal intensity (Figure 2); this is consistent with the binding of SAM to the unique iron of the radical SAM [4Fe-4S] cluster, in addition to some nonproductive SAM cleavage accompanied by cluster oxidation to the EPR-silent [4Fe-4S]²⁺ state. These spectroscopic observations are all in line with those reported for other radical SAM enzymes.

With the homology model and spectroscopic properties pointing to nothing unusual about the active-site of GD-AE, we decided to re-investigate the reported SAM cleavage activity.²⁰ As shown in Figures 4 and 5, we found that GD-AE, like other characterized radical SAM enzymes, catalyzes nonproductive SAM cleavage (SAM cleavage in the absence of substrate) at a low rate to produce 5’-dAdoH. The rate of GD-AE catalyzed SAM cleavage is significantly increased in the presence of the 7-mer peptide substrate corresponding to the glycyl radical loop of GD. MTA was also detected in these assays, however it was present in the SAM stock used in the assay and did not increase in the presence of GD-AE. In fact, as shown in Figure 5, the small increase in MTA over time was observed in all samples, regardless of the presence or absence of GD-AE. In contrast, substantial 5’-dAdoH was formed only in the complete reaction containing both GD-AE and the substrate GD peptide; in this case, the 5’-dAdoH increased over time at a rate consistent with the range of rates observed for other radical SAM enzymes.¹

In order to probe the identity of the SAM cleavage products generated by GD-AE when acting on the native substrate GD, we combined GD-AE, GD, SAM, and reductant and then used EPR spectroscopy to look for generation of the glycyl radical on GD; Figure 6 clearly shows a glycyl radical, with its characteristic doublet EPR signal, is generated on GD. This sample was then subjected to HPLC analysis, revealing that 5’-dAdoH, but not MTA, was generated during GD-AE catalyzed activation of GD (Figure 7). MTA, a common SAM degradation product, is observed in the control samples lacking GD-AE in Figures 4 and 7; we have observed that the amount of MTA in a solution of SAM is generally small but can increase with time as the SAM sits at ambient temperature. The prior observation of MTA during the GD-AE catalyzed reaction,²⁰ therefore, may have resulted simply from a small amount of MTA in their SAM stock solution. Note that this prior study provided only a single 15 s time point and did not demonstrate the time-dependent formation of MTA,²⁰ as we do here for the formation of 5’-dAdoH.

Together, our results demonstrate that GD-AE catalyzes a reaction employing the typical S-C(5’) bond cleavage observed for other radical SAM enzymes, as opposed to the S-C(γ) bond cleavage previously reported.²⁰ Thus as of this writing, all characterized canonical radical SAM enzymes initiate radical catalysis via the same initial S-C(5′) bond cleavage to generate the 5’-deoxyadenosyl radical intermediate. While Dph2 performs radical SAM-like chemistry via cleavage of the S-C(γ) bond of SAM,²³ it is not a member of the canonical superfamily given the absence of the key conserved sequence and structural features.^{23, 52} Radical SAM superfamily members therefore, though remarkably diverse in the chemistry they catalyze, appear to have in common initial mechanistic steps, including the generation of the 5′-deoxyadenosyl radical intermediate from S-C(5′) bond cleavage of SAM.

Synopsis:

The radical S-adenosyl-l-methionine (SAM) enzyme glycerol dehydratase activating enzyme (GD-AE) binds multiple [4Fe-4S] clusters and catalyzes the formation of the catalytically essential glycyl radical of glycerol dehydratase (GD). We show here that contrary to a prior report, GD-AE catalyzes the reductive cleavage of the SAM S-C5’, rather than the S-Cγ, bond.

ABBREVIATIONS

SAM

S-adenosyl-L-methionine

5’-dAdo

5’-deoxyadenosine

ACP

3-amino-3-carboxypropyl

GD-AE

B12-Independent glycerol dehydratase activating enzyme

GD

glycerol dehydratase

MTA

5’-Deoxy-5’-(methylthio)adenosine

Dph2

2-(3-amino-3-carboxypropyl)histidine synthase

4HPD-AE

4-hydroxyphenylacetate decarboxylase activating enzyme

RSEs

radical SAM enzymes

PFL-AE

pyruvate formate-lyase activating enzyme

aRNR-AE

anaerobic ribonucleotide reductase activating enzyme

ISC

Iron Sulfur Cluster

PMSF

phenylmethylsulfonyl fluoride

DTT

dithiothreitol

DT

sodium dithionite

EPR

electron paramagnetic resonance