Total synthesis of (+)-12-epi-fischerindole U isonitrile and its activity as a substrate of the halogenase WelO5

Abstract

Mononuclear non-heme Fe(II) and 2-oxoglutarate-dependent enzymes (Fe(II)/2OG) constitute an enzyme superfamily that oxidizes substrates by activating molecular oxygen (O₂) via the highly oxidized iron(IV)-oxo (ferryl) intermediate. Despite the high similarity of their active sites, the enzymes of this superfamily perform a diverse array of reactions such as hydroxylation, halogenation, epoxidation, desaturation etc. WelO5, a member of this superfamily, regio- and stereospecifically monochlorinates a free-standing substrate, 12-epi-fischerindole U isonitrile (FIUI), at its aliphatic C(13) to give 12-epi-fischerindole G (FIG). Researchers have proposed that a dynamic reconfiguration of the WelO5 active site governs the selective halogenation of FIUI. Unfortunately, the requirements of spectroscopic and structural methods for large amounts of fully assembled samples of protein–substrate complex impede the investigation of such reconfiguration. In this work, we describe a synthesis of (+)-FIUI that makes use of the Baran group’s strategy for coupling of indoles with carbonyl compounds. We apply EPR and LC–MS methods to demonstrate the ability of WelO5 to bind the synthetic (+)-FIUI and subsequently convert it to FIG.

Graphical abstract:

Introduction

Fe(II) and 2-oxoglutarate-dependent enzymes (Fe(II)/2OG) are involved in myriad biological processes such as biosynthesis of natural products, transcription, DNA repair, mRNA demethylation and many others.^{1, 2} They are structurally characterized by a distorted double-stranded β-helix (known as cupin or jelly-roll fold) and a conserved chelating motif to bind the Fe(II) at the active site.³ This conserved motif, termed the facial triad,⁴ is typically composed of a pair of His residues and a carboxylate ligand (Asp or Glu). The majority of Fe(II)/2OG enzymes are hydroxylases; i.e., they install an OH group on the primary substrate. However, researchers have recently recognized that a large number of functionally divergent subclasses eschew hydroxylation in favor of other transformations.^{2, 4, 5, 6, 7} In halogenases, such as SyrB2, WelO5, and BesD, the Gly residue replaces the carboxylate ligand in the facial triad to accommodate coordination of halide (Cl, Br) to the metallocenter.^{5, 6, 8} The initial stages of the reaction are similar between hydroxylases and halogenases. These enzymes undergo the reaction of 2OG and O₂ in the presence of the primary substrate to form a highly reactive iron(IV)-oxo (ferryl) intermediate, with a release of CO₂.⁹ Consequent hydrogen atom transfer (HAT) from the target substrate site to the metallocofactor results in formation of the X-Fe(III)-OH (X=Cl, Br, N₃) intermediate and a substrate radical (•C). In hydroxylases, this radical then abstracts OH from Fe(III) (oxygen rebound mechanism).^10,11 In halogenases, however, the rebound is avoided in favor of halide transfer. The exact mechanism of such oxygen rebound avoidance is yet to be resolved. Early studies of SyrB2 suggested that halogenases direct reactivity towards halogenation by positioning the target carbon closer to the halide than the ferryl oxo group.^11–13 Unlike SyrB2, which operates on its substrate only when the latter is appended to a partner protein SyrB1, WelO5 operates on a stand-alone substrate, providing a convenient platform for investigating the halogenation reaction and hydroxylation avoidance.^{6, 14} The enzyme acts only on 12-epi-fischerindole U isonitrile (FIUI, 1a) (Figure 1), halogenating it at the aliphatic C(13) and in the equatorial orientation to give 12-epi-fischerindole G (1b) exclusively.⁶ Biologically, WelO5 is a apart of a gene cluster responsible for biosynthesis of welwitindolinones valuable for their antibacterials and antifungal properties.¹⁵

Figure 1.

Some 12-epi-fischerindoles.

Intriguingly, the crystallographic structure of the WelO5 active site shows the target C(13)–H bond of 1a positioned directly above the oxygen binding site of Fe(II) cofactor (“inline” position, trans to the C-terminal His ligand).¹⁴ Such substrate–cofactor deposition seems to be incompatible with the halide transfer, and it should result in a selective hydroxyl group rebound due to the group’s proximity to the intermediate •C(13). Therefore, researchers have proposed that in the process of installing the haloferryl intermediate, a reorganization of the cofactor takes place, resulting in an Fe(IV)=O bond orientation (“offline”) that is orthogonal to that of hydroxylases.^14,16 However, researchers have yet to conclusively demonstrate such reorganization. Spectroscopically, one can evaluate the substrate–cofactor disposition by replacing the ferryl intermediate with its faithful mimic, a vanadyl group (V^IV≡O). This cofactor exhibits a complex EPR spectrum owing to the hyperfine interaction of the unpaired electron (S = 1/2) with ⁵¹V (I = 7/2) nucleus. The anisotropy of the interaction affords precise determination of the cofactor coordination and is sensitive to structural perturbations due to the presence of primary and cosubstrates.¹⁷

Unfortunately, the need for relatively large sample amounts impedes such investigations. Whereas heterologous protein expression and isolation from E. coli hosts are well-established and optimized for high yields, obtaining the primary substrate 1a in sufficient quantities remains challenging. In the past, researchers obtained 1a by isolation from the native organism, Hapalosiphon welwitschii UTEX B1830,⁶ which was then successfully utilized to obtain crystal structures of the {Fe-NO}⁷ adduct of WelO5.¹⁴ Such methodology, while possible, is extremely laborious, has relatively low yield, and precludes any modifications of the substrates, e.g., by site-specific isotope labeling. Therefore, in this work, we have developed a synthetic route to 1a. Furthermore, we show that our synthetic 1a can successfully bind to the WelO5-vanadyl complex and be chlorinated by WelO5·Fe(II)·2OG in the presence of oxygen.

Results and discussions

The Baran group has been the leader in developing synthetic methods that lead to fischerindoles,^18–22 although neither they nor any others have reported the synthesis of FIUI specifically. We decided to use the Baran group’s syntheses of ent-12-epi-fischerindole G (ent-1b) and ent-12-epi-fischerindole U isothiocyanate (ent-1c) as models for the synthesis of 1a.^19–21 Coincidentally, the Maji group very recently reported the synthesis of 11-epi-1a,²³ but not 1a itself. The Ang Li group has also recently disclosed some very nice approaches to some fischerindoles and related compounds, though not to 1a.²⁴

We began our synthesis of 1a by converting (S)-carvone into cyclohexanone 2 using the three-step procedure developed by the Zografos group for converting (R)-carvone into ent-2 (Scheme 1).²⁵ Later, the Rawal group prepared 2 itself by a different procedure.²⁶ We then subjected 2 to Baran’s indole–ketone coupling protocol to give 3. The Baran group had previously prepared ent-3 from (R)-carvone by a different sequence of the same steps to begin their synthesis of ent-1c.¹⁹ Our synthetic 3 had the opposite optical rotation to Baran’s reported ent-3, reflecting the opposing stereochemistry.

Scheme 1.

Synthesis of α-indolyl cyclohexanone 3.

The Baran group had next added CF₃SO₃H to ent-3 to promote an intramolecular Friedel–Crafts alkylation between C(2) of the indole and C(15) of the isopropenyl group, reportedly affording tetracyclic ketone ent-4 in 31% yield along with 59% recovered starting material (Scheme 2).¹⁹ We were unable to obtain more than about a 20% yield of 4 under these conditions, even after isolating the unreacted 3 and resubjecting it to the reaction conditions, and the 4 that we did obtain was not clean. Scale-up also failed to yield larger quantities of 4. Furthermore, the Baran group reported that reductive amination of ketone ent-4 with ammonium acetate provided axial primary amine ent-5, but, in our trials, reductive amination of 4 did not provide any material that we could identify as 5.

Scheme 2.

Baran group’s reported route from ent-3 to ent-5.

The reasons for these discrepancies remain unknown, but we established an alternative synthetic route by heeding an aside in the Fukuyama group’s report on the synthesis of hapalindole G.¹⁸ We first reduced ketone 3 with NaBH₄, obtaining the alcohol in approximately 9:1 dr as measured by ¹H NMR (Scheme 3). The ¹H NMR spectrum of the major diastereomer, 6, indicated that the OH group was equatorial; specifically, H(11) (δ 2.96: t, 10.9 Hz) showed a large coupling to H(10), typical of a fixed 1,2-diaxial orientation, as well as a large coupling to the H of the OH group. The intramolecular Friedel–Crafts alkylation of 6 promoted by 2.5 N HCl in EtOH (1:1) then joined C(2) and C(15) much more cleanly and in better yield than that of 3 did, affording tetracyclic alcohol 7.¹⁸ As expected, the ¹H NMR spectrum of 7 indicated that the OH group remained equatorial, as indicated by the large coupling of H(11) (δ 3.45: dd, 9.0 Hz, 10.8 Hz) to H(10) as well as to the H of the OH group. We found that 6 and 7 began to decompose within hours, so we quickly converted 7 to tetracyclic mesylate 8 with Ms₂O without first purifying 6 or 7, obtaining 8 in 34% yield over the three steps. The large coupling constant between H(11) (δ 4.50: d, 11.0 Hz) and H(10) in the ¹H NMR spectrum of 8 indicated that the OMs group in 8 remained equatorial, as expected.

Scheme 3.

Sidestepping the problematic Friedel–Crafts reaction of 3 and reductive amination of 4.

We completed the synthesis of 1a (Scheme 4) by applying to 8 the procedures that the Baran group had used to prepare 12-epi-fischerindole G.²⁰ We began by displacing the mesylate group of 8 with lithium azide in DMF to provide axial azide 9 in 29% yield. The ¹H NMR spectrum of 9 clearly indicated that the substitution at C(11) had inverted its configuration, with only very small, unresolved coupling between the now equatorial H(11) (δ 4.18; broad singlet) and H(10). We then used sodium amalgam to reduce axial azide 9 to axial primary amine 5, the enantiomer of which the Baran group had reported synthesizing by reductive amination of ketone ent-4 with ammonium acetate.¹⁹ After formylating the amino group of 5 to provide formamide 10,^{19, 27} we treated 10 with diphosgene²⁸ to provide 1a in 18% yield. The ¹H NMR resonance of H(11) in 1a (δ 4.34; broad singlet) confirmed the retained axial orientation of the isonitrile at C(11). The ¹H NMR spectrum of 1a also featured a resonance at δ 3.16 with a striking multiplicity of 1:2:2:2:2:2:1 that we attributed to the axial H(10) and its three three-bond couplings to axial H(15) (J_d = 11.1 Hz), equatorial H(11) (J_d = 3.7 Hz), and axial isonitrile N (J_t(1:1:1) = 3.7 Hz also), consistent with the reported ¹H NMR spectrum of 1a that had been isolated from its natural source.²⁹ The report describing the isolation of 1a from nature did not report the specific rotation of natural 1a, so we we were unable to compare the specific rotation of our synthetic 1a to that of natural 1a to establish that they were homomers, but because we had started our synthesis of 1a with (S)-carvone, we were confident that we had made the natural enantiomer. Our confidence turned to certainty when WelO5 accepted our synthetic 1a as a substrate, as described next.

Scheme 4.

Completion of the synthesis of 1a.

To verify that the synthetic 1a can indeed be modified by WelO5, we first investigated the ability of 1a to bind to WelO5. Past research on Fe(II)/2OG dependent enzymes universally demonstrated the sensitivity of the vanadyl EPR spectra to the presence of substrates.^{13, 17, 30} Hence, this effect can be used to interrogate substrate binding abilities in WelO5. As expected, distinct changes in the spectral features of WelO5·VO·Cl⁻·succinate complex were observed with the addition of 1a (Figure 2). Furthermore, we observed a near-complete conversion of the spectra, most evident in the low-field components (see insets in Figure 2), indicating a near-quantitative incorporation of 1a into the WelO5·V(IV)O·succinate·Cl⁻ complex. We were able to simulate both spectra using distinct sets of spin Hamiltonian parameters. We also observe narrowing of spectral features upon binding of 1a, suggesting a decrease in the heterogeneity of the metallocofactor environment. We provide the simulations and the simulation parameters in the Supporting Information (Figure S40).

Figure 2.

X-band CW-EPR spectra of WelO5·VO·Cl⁻·succinate complex in the presence of substrate 1a (red) and without it (blue). Insets highlight spectral features in the low- and high-field regions. Spectra were accumulated at 35 K with a modulation amplitude of 0.5 mT and a microwave frequency of 9.434 GHz.

Next, we performed an enzymatic activity assay using heterologously expressed and purified WelO5 from E. coli BL21(DE3). Incubation of WelO5 aerobically at 30 °C for 60 minutes in the presence of Fe(II), 2OG, and NaCl rapidly converted 1a to two distinct products, with the major product being the expected chlorinated product 1b (exact mass calculated based on chemical formula: 339.1623) and the minor product being the hydroxylated variant 1d (321.1961). The substrate and products were analyzed by high-resolution LC–MS by monitoring their absorbance at 280 nm and observed masses (Figure 3A). LC–MS clearly separates compounds 1b, 1a, and hydroxylated minor product 1d, as expected from the differences in their polarities. Furthermore, analysis of the exact mass and isotope abundance by high-resolution MS permitted the assignment of these peaks to the substrate and products (Figure 3B). The isotopic mass spectra signatures of 1a (m/z 305.2009, 100%), (m/z 306.2046, 23%), (m/z 307.2082, 3%) along with the converted 1b (339.1621, 100%), (m/z 341.1593, 32%), (m/z 340.1650, 23%), (m/z 342.1626, 7%) and 1d (m/z 321.1913, 100%), (m/z 322.1994, 23%), (m/z 323.2025, 3%) (Figure 3B) matched well with the values calculated from natural isotope abundances. The retention times for 1a, 1b, and 1d were 14.7, 15.6, and 6.3 minutes, respectively. We estimated the 1b:1d product ratio to be 21:1 based on their absorbance at 280 nm, consistent with previous reports.^{6, 31} These results show unequivocally that our synthetic 1a, prepared from (S)-carvone, is in the natural enantiomeric series.

Figure 3.

In vitro characterization of WelO5 with synthetic 1a. (A) LC and (B) MS data shown for the in vitro assay conducted with WelO5 (80 μM), 1a (200 μM), 2-OG (2 mM), NaCl (10 mM), ammonium iron(II) sulfate hexahydrate (500 μM) incubated at 30 °C for 60 min. We obtained LC–MS data with an Agilent Zorbax Extend-C18 column (4.6 mm × 50 mm, 1.8 μm particle size) at 35 °C that was equilibrated in 50% solvent A (water with 0.1% formic acid) and 50% solvent B (1:1 MeOH:ACN with 0.1% formic acid). Throughout each injection, we applied a gradient that increased solvent B to 80% over 30 min. We observed the isotopic mass spectrum signatures of 1a (m/z 305.2009, 100%; m/z 306.2046, 23%; m/z 307.2082, 3%), 1b (339.1621, 100%; m/z 341.1593, 32%; m/z 340.1650, 23%; m/z 342.1626, 7%), and hydroxylated product 1d (m/z 321.1913, 100%; m/z 322.1994, 23%; m/z 323.2025, 3%).

To summarize, we have successfully synthesized the natural enantiomer of FIUI and showed that WelO5 recognized it as its substrate. To our knowledge, ours is the first synthesis of FIUI, although of course our synthesis derives heavily from syntheses of other fischerindoles. In the future, we plan to use synthetic 1a and its deuterated congeners to gain more insight into the mechanism by which WelO5 halogenates FIUI instead of hydroxylating it.

Experimental section

General methods

We purchased all chemical reagents from commercial suppliers and used them without further purification. We carried out all reactions under a nitrogen atmosphere. We monitored reactions by Sorbtech Silica G TLC plates (UV 254), using UV light and P-anisaldehyde stain as a visualizing agent. We recorded the ¹H and ¹³C NMR spectra on a 400 MHz Bruker Avance NEO spectrometer. We confirmed structural assignments with additional information from gNOESY, gCOSY, and gHSQC experiments. We reported chemical shifts (δ) in ppm, and we denoted multiplicities as s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), dd (doublet of doublet), dt (doublet of triplet), dq (doublet of quartet). We reported coupling constants (J) in Hz. We collected GC–MS data on an Agilent Technologies 6890N GC system equipped with a split injection port and with 5973 mass selective detector (MSD), operated using Chemstation software. We acquired high resolution mass spectra (HRMS) with an Agilent 6230 TOF LC/ MS system and AB SCIEX Triple TOF 5600 in electrospray ionization mode (ESI+). We recorded IR spectra on a Nicolet 6700 and a Nicolet iS10 FT-IR spectrophotometer. We obtained optical rotations on a Jasco DIP-370 digital polarimeter. Melting points were determined using a Thermo 1001D Mel-Temp Electrothermal Melting Point Apparatus.

(2S,3S,6S)-2-(3-Indolyl)-2-isopropenyl-6-vinyl-6-methylcyclohexanone (3):

We added n-BuLi (2.5 M solution in hexanes, 10.1 mL, 25.3 mmol) to a solution of hexamethyldisilazane (6.06 mL, 28.94 mmol) in THF (7.23 mL, 1 M) at −78 °C under an atmosphere of N₂. We allowed this solution to warm to 0 °C, waited 30 min, and cooled it back to −78 °C. In the meanwhile, we prepared a second solution containing cyclohexanone^{25, 26} 2 (1.29 g, 7.23 mmol) and indole (1.69 g, 14.46 mmol) in THF (7.23 mL, 1 M) under an atmosphere of N₂. We slowly added the second solution to the first and then allowed the mixture to stir at −78 °C for another 30 min. We then added solid copper (II) 2-ethylhexanoate (3.79 g, 10.84 mmol) to the reaction mixture all at once, allowing the reaction mixture to stir at −78 °C for another 10 min before allowing it to warm to room temperature. We quenched the reaction mixture by adding 1 N aq. HCl (100 mL), and we extracted the solution with EtOAc (3 × 100 mL). We washed the combined organic layers with 1 N aq. NaOH (3 × 50 mL), water, and brine, dried them over Na₂SO₄, and evaporated the solvent. Flash chromatography of the residue on silica gel (20% EtOAc in hexanes) gave indole-substituted ketone 3 (1.09 g, 3.72 mmol, 51%) as a white solid, mp 176–178 °C, [α]²⁰_D = −97.7° (c = 0.0105, CH₂Cl₂), whose NMR spectra matched those of the reported compound.^{19 1}H NMR (400 MHz, CDCl₃): δ 8.07 (broad s, 1H), 7.32 (d, 7.8 Hz, 1H), 7.25 (d, 8.5 Hz, 1H), 7.11 (t, 7.5 Hz, 1H), 7.04 (t, 7.5 Hz, 1H), 6.82 (s, 1H), 6.15 (dd, 17.7 Hz, 10.7 Hz, 1H), 5.33 (d, 10.7 Hz, 1H), 5.18 (d, 17.7 Hz, 1H), 4.60 (s, 1H), 4.52 (s, 1H), 4.27 (d, 12.5 Hz, 1H), 2.91 (dt, J_t =12.4 Hz, J_d = 3.4 Hz, 1H), 2.28 – 2.08 (m, 2H), 1.91 – 1.73 (m, 2H), 1.52 (s, 3H), 1.21 (s, 3H). ¹³C{¹H} NMR (101 MHz, CDCl₃): δ 211.0, 146.8, 143.0, 136.1, 127.6, 123.3, 121.5, 119.2, 118.8, 116.1, 112.0, 111.6, 111.3, 53.4, 52.3, 48.8, 39.2, 28.7, 25.0, 18.4. IR (neat): 3349, 1702 cm⁻¹. GC–MS (EI): 98% pure; MS (EI) m/z: M⁺ Calcd for C₂₀H₂₃NO 293.2; Found 293.1. HRMS (ESI) m/z: [M+ H]⁺ Calcd for C₂₀H₂₄NO ⁺ 294.1852; Found 294.1849.

(1S,2S,5S,6S)-6-(3-Indolyl)-5-isopropenyl-2-vinyl-2-methylcyclohexanol (6):

We dissolved ketone 3 (250 mg, 0.85 mmol) in MeOH (8.5 mL, 0.1 M) under N₂, then cooled it to 0 °C. We added sodium borohydride (64.5 mg, 1.70 mmol) to the reaction mixture and allowed it to stir for 30 min. We then allowed the mixture to warm to room temperature and allowed it to stir for another 15 min. We then added sat. aq. NH₄Cl solution (20 mL) and extracted the aqueous layer with EtOAc (3 × 10 mL). We combined the organic layers, washed with them with water and brine, dried them over Na₂SO₄, and evaporated the solvent to give crude alcohol 6 (250 mg). We used this impure material directly in the next step. ¹H NMR (400 MHz, CDCl₃): δ 8.09 (broad s, 1H), 7.70 (d, 7.9 Hz, 1H), 7.26 (d, 8.0 Hz, 1H), 7.15 (t, 7.6 Hz, 1H), 7.08 (t, 7.5 Hz, 1H), 6.88 (s, 1H), 6.37 (dd, 17.8 Hz, 11.1 Hz, 1H), 5.28 (d, 11.1 Hz, 1H), 5.22 (d, 17.9 Hz, 1H), 4.50 (s, 1H), 4.41 (s, 1H), 3.64 (s, 1H), 2.96 (t, 11.1 Hz, 1H), 2.72 (t, 12.4 Hz, 1H), 1.98 – 1.89 (m, 1H), 1.87 – 1.72 (m, 1H), 1.64 – 1.49 (m, 3H), 1.48 (s, 3H), 1.20 (s, 3H). ¹³C{¹H} NMR (101 MHz, CDCl₃): δ 147.8, 145.1 (broad), 140.8, 136.6 (broad), 123.0 (broad), 121.9, 119.7 (broad), 119.3, 114.6 (broad), 114.3, 111.4, 111.1, 81.3 (broad), 50.4 (broad), 43.2 (broad), 41.7, 36.7, 28.0, 27.8, 19.1. GC–MS (EI): 95% pure; MS (EI) m/z: M⁺ Calcd for C₂₀H₂₅NO 295.2; Found 295.2.

(6aS,9S,10S,10aS)-9-Vinyl-6,6,9-trimethyl-5,6,6a,7,8,9,10,10a-octahydroindeno[2,1-b]indol-10-ol (7):

We dissolved the crude tricyclic alcohol 6 (250 mg, 0.85 mmol) in EtOH (8.5 mL, 0.1 M), added 2.5 N aq. HCl (8.5 mL, 0.1 M), and allowed the solution to reflux for 1 h. We allowed the reaction mixture to cool to room temperature, neutralized it with 1 N aq. NaOH, and extracted with CH₂Cl₂ (3 × 10 mL). We combined the organic layers, washed them with water and brine, dried the organic layers over Na₂SO₄, and evaporated the solvent to give crude tetracyclic alcohol 7 (250 mg). We used this impure material directly in the next step. ¹H NMR (400 MHz, CDCl₃): δ 7.84 – 7.74 (m, 2H), 7.32 – 7.26 (m, 1H), 7.12 – 7.03 (m, 2H), 6.26 (dd, 17.7 Hz, 11.0 Hz, 1H) 5.29 (dd, 10.9 Hz, 1.7 Hz, 1H), 5.24 (dd, 17.7 Hz, 1.6 Hz, 1H), 3.45 (t, 9.8 Hz, 1H), 2.81 (tt, 10.7 Hz, 1.4 Hz, 1H), 2.03-1.94 (m, 3H), 1.65 (ddt, J_d = 16.0 Hz, 1J_d = 3.1 Hz, J_t =3.0 Hz, 1H), 1.58 – 1.42 (m, 2H), 1.35 (s, 3H), 1.22 (s, 3H), 1.05 (s, 3H).¹³C{¹H} NMR (101 MHz, CDCl₃): δ 151.9, 140.5, 139.6, 124.5, 120.6, 120.5, 119.8, 117.9, 115.7, 111.3, 81.0, 62.0, 46.8, 43.4, 40.7, 38.7, 25.7, 25.3, 20.8, 20.6. C₂₀H₂₅NO.

(6aS,9S,10S,10aS)-9-Vinyl-6,6,9-trimethyl-5,6,6a,7,8,9,10,10a-octahydroindeno[2,1-b]indol-10-yl methanesulfonate (8):

We dissolved the crude tetracyclic alcohol 7 (250 mg) in pyridine (8.5 mL, 0.1 M) and added methanesulfonic anhydride (296 mg, 1.70 mmol). We heated the reaction mixture to 65 °C and allowed it to stir for 30 min. After allowing it to cool back to room temperature, we diluted the mixture with Et₂O (50 mL) and washed the organic layer with 1 N aq. HCl (3 × 20 mL), water, and brine. We dried the solution over Na₂SO₄ and evaporated the solvent. Purification by flash chromatography on silica gel (20% EtOAc in hexanes) gave mesylate 8 (108 mg, 0.28 mmol, 34% after 3 steps) as a brown solid, mp 147–149 °C, [α]²⁰_D = +77° (c = 0.0051, CH₂Cl₂), ¹H NMR (400 MHz, CDCl₃): δ 7.89 (s, 1H), 7.76 – 7.68 (m, 1H), 7.36 – 7.27 (m, 1H), 7.10 (dt, J_d = 5.0 Hz, J_t = 2.1 Hz, 2H), 6.19 (dd, 17.8 Hz, 11.2 Hz, 1H), 5.30 – 5.07 (m, 2H), 4.49 (d, 10.9 Hz, 1H), 3.25 (t, 10.6 Hz, 1H), 3.05 (s, 3H), 2.28 – 2.11 (m, 2H), 1.67 – 1.53 (m, 3H), 1.41 (s, 3H), 1.31 (s, 3H), 0.92 (s, 3H).¹³C{¹H} NMR (101 MHz, CDCl₃): δ 152.8, 140.5, 140.2, 124.8, 121.2, 120.4, 119.8, 115.6, 114.3, 111.6, 92.8, 61.8, 45.9, 44.7, 40.6, 38.5, 37.4, 29.1, 25.6, 21.7, 21.3. IR (neat): 3392, 1171 cm⁻¹. HRMS (ESI) m/z: [M+ H]⁺ Calcd for C₂₁H₂₈NO₃S⁺ 374.1784; Found 374.1782.

(6aS,9S,10R,10aS)-10-Azido-9-vinyl-6,6,9-trimethyl-5,6,6a,7,8,9,10,10a-octahydroindeno[2,1-b]indole (9):

We dissolved the mesylate 8 (100 mg, 0.26 mmol) in DMF (8.66 mL, 0.03 M), added 20 % aq. LiN₃ solution (445 μL, 9.1 mmol), and heated the mixture to 100 °C for 24 h. After allowing the reaction mixture to cool to room temperature, we diluted it with Et₂O (30 mL), washed it with water and brine, dried it over Na₂SO₄, and concentrated it in vacuo. Purification by flash chromatography on silica gel (15% EtOAc:hexanes) gave azide 9 (25 mg, 0.078 mmol, 29%) as a pale yellow waxy solid, [α]²⁰_D = +86.9° (c = 0.014, CH₂Cl₂). ¹H NMR (400 MHz, CDCl₃): δ 7.84 (s, 1H), 7.48 (t, J = 4.6 Hz, 1H), 7.30 (dd, J = 6.1, 3.4 Hz, 1H), 7.23 – 6.94 (m, 2H), 5.94 (dd, 17.5 Hz, 10.9 Hz, 1H), 5.24 – 5.11 (m, 2H), 4.18 (s, 1H), 3.20 (dd, 10.6 Hz, 2.6 Hz, 1H), 2.33 (dq, J_d = 10.8 Hz, J_q = 4.2 Hz, 1H), 1.78 (d, 9.5 Hz, 1H), 1.64 – 1.47 (m, 3H), 1.35 (s, 3H), 1.19 (s, 3H), 1.00 (s, 3H). ¹³C{¹H} NMR (101 MHz, CDCl₃) δ 152.2, 144.6, 139.6, 124.3, 120.6, 119.9, 118.4, 116.0, 113.4, 111.7, 70.7, 54.7, 43.4, 42.6, 40.2, 32.3, 27.9, 25.2, 20.9, 20.7. IR (neat): 2106 cm^–1. GC: 90% pure; MS (EI) m/z: M⁺ Calcd for C₂₀H₂₄N₄ 320.2; Found 320.1.

(6aS,9S,10R,10aS)-9-Vinyl-6,6,9-trimethyl-5,6,6a,7,8,9,10,10a-octahydroindeno[2,1-b]indol-10-amine (5):

We dissolved azide 9 (75 mg, 0.234 mmol) in EtOH (2.3 mL, 0.1 M) and added sodium/mercury amalgam (523 mg, 2.34 mmol), then allowed the reaction to reflux for 3 h. After the reaction mixture cooled to room temperature, we diluted it with EtOAc (30 mL), and we then washed this solution with water and brine. We dried the solution over Na₂SO₄ and evaporated the solvent in vacuo to give crude 5. Purification by flash chromatography on silica gel (15% MeOH:CH₂Cl₂) gave less but still somewhat impure primary amine 5 (44 mg, 0.14 mmol, 64%) as a pale yellow waxy solid, [α]²⁰_D = +18.4° (c = 0.003, CH₂Cl₂). We used this impure material directly in the next step. ¹H NMR (400 MHz, CDCl₃) δ 7.95 (s, 1H), 7.47 (dd, J = 6.1, 3.4 Hz, 1H), 7.34 – 7.28 (m, 1H), 7.11 – 7.05 (m, 2H), 5.98 (dd, J = 17.7, 10.9 Hz, 1H), 5.19 – 5.04 (m, 2H), 3.64 (t, J = 2.4 Hz, 1H), 3.19 (d, J = 10.9 Hz, 1H), 2.35 (td, J = 11.3, 6.4 Hz, 1H), 1.75 – 1.50 (m, 5H), 1.33 (s, 3H), 1.24 – 1.17 (m, 1H), 1.09 (s, 3H), 1.01 (s, 3H). ¹³C{¹H} NMR (101 MHz, CDCl₃) δ 153.1, 146.8, 139.6, 124.3, 120.5, 119.8, 118.2, 116.2, 112.4, 111.7, 65.9, 56.2, 53.7, 44.2, 41.5, 40.4, 31.5, 27.2, 25.3, 21.2, 20.6. IR (neat): 3400 cm⁻¹. GC–MS (EI): 96% pure; MS (EI) m/z: M⁺ Calcd for C₂₀H₂₆N₂ 294.2; Found 294.2. HRMS (ESI) m/z: [M+ H]⁺ Calcd for C₂₀H₂₇N₂⁺ 295.2169; Found 295.2169.

(6aS,9S,10R,10aS)-10-Formamido-9-vinyl-6,6,9-trimethyl-5,6,6a,7,8,9,10,10a-octahydroindeno[2,1-b]indole (10):

We sequentially added amine 5 (44 mg, 0.149 mmol), formic acid (6.18 μL, 0.163 mmol), 2-chloro-4,6-dimethoxy-1,3,5-triazine (31.3 mg, 0.178 mmol), DMAP (0.5 mg, 0.004 mmol), and N-methylmorpholine (19.7 μL, 0.178 mmol) to CH₂Cl₂ (1.49 mL, 0.1 M), and we allowed the mixture to stir at room temperature for 2 h. We diluted the reaction mixture with CH₂Cl₂ (30 mL), washed it with 1 N HCl (2 × 20 mL), saturated aq. NaHCO₃ (20 mL), water, and brine, and dried it over Na₂SO₄. We then evaporated the solution in vacuo to give crude 10. Purification by flash chromatography on silica gel (60% EtOAc:hexanes) gave 10 (35 mg, 0.108 mmol, 73%). Its NMR spectra showed a mixture of diastereomers. IR (neat): 3294, 1671 cm⁻¹. GC–MS (EI): 94% pure; MS (EI) m/z: M⁺ Calcd for C₂₁H₂₆N₂O 322.2; Found 322.2.

12-epi-Fischerindole U isonitrile (1a):

We cooled a solution of crude formamide 10 (12 mg, 37 μmol) in CH₂Cl₂ (0.3 mL, 0.1 M) under an atmosphere of N₂ to 0 °C and added N-methylmorpholine (10.2 μL, 92 μmol) and diphosgene (4.5 μL, 37 μmol), allowing the reaction mixture to stir for 2 h. We then quenched the reaction by adding sat. aq. NaHCO₃ solution and extracting with Et₂O (3 × 2 mL). We combined the organic layers, washed them with brine solution, dried the organic layer over Na₂SO₄, and concentrated the solution in vacuo. Purification by flash chromatography on silica gel (20% EtOAc:hexanes) gave 1a (4.3 mg, 0.014 mmol, 18%) as a white solid, [α]²⁰_D = +26.7° (c = 0.0014, CH₂Cl₂). ¹H NMR (400 MHz, CDCl₃): δ 7.84 (s, 1H), 7.51 – 7.41 (m, 1H), 7.36 – 7.28 (m, 1H), 7.15 – 7.04 (m, 1H), 5.86 (ddd, 17.7 Hz, 10.9 Hz, 0.9 Hz, 1H), 5.23 (d, 3.3 Hz, 1H), 5.19 (dd, 2.8 Hz, 0.7 Hz, 1H), 4.34 (broad s, 1H), 3.16 (dd of 1:1:1 triplet, J_d = 11.1 Hz, 3.7 Hz, J_t(1:1:1) = 3.7 Hz, 1H), 2.36 (dt, J_t = 11.1 Hz, J_d = 4.1 Hz, 1H), 1.90 (ddt, J_d = 13.3 Hz, 3.5 Hz, J_t = 1.6 Hz, 1H), 1.79 – 1.54 (m, 3H), 1.39 (s, 3H), 1.27 (s, 3H), 1.03 (s, 3H).¹³C{¹H} NMR (101 MHz, CDCl₃): δ 157.4 (1:1:1 multiplet, 5.0 Hz), 152.2, 142.8, 139.6, 124.0, 120.8, 119.9, 118.2, 114.8, 114.6, 111.7, 62.6 (1:1:1 multiplet, 4.5 Hz), 55.0, 42.2, 40.9, 40.2, 31.9, 28.1, 25.2, 20.9, 20.6. IR (neat): 2137 cm^–1. GC–MS (EI): 95% pure by GC; MS shows M⁺ = 304.2 amu (calcd for C₂₁H₂₄N₂: 304.2 amu). MS (ESI) m/z: [M + H]⁺ Calcd for C₂₁H₂₅N₂: 305.2018; Found: 305.2016.

Overexpression and purification of WelO5

The Boal group generously provided WelO5 plasmid (pQTEV::welO5). The plasmid backbone contained an N-terminal His₆-tag followed by a TEV (Tobacco Etch Virus) protease cleavage site and Ampicillin as bacterial resistance. Following the transformation of the plasmid into BL21(DE3) competent cells (New England Biolabs^®), we overexpressed cells in Lysogeny Broth (LB) media at 180 RPM, 37 °C until an optical density of 0.6 was reached. After growth, we cooled the cells down on ice, and we induced protein expression with 1 mM (final) isopropyl β-D-thiogalactopyranoside (IPTG). We continued protein expression overnight at 18 °C with shaking at 120 RPM. We harvested cells by centrifuging the cell-rich LB media at 18,000 RPM for 15 minutes. Liquid N₂ flash-froze the cell pellet, which we then stored at −80 °C.

We purified proteins via immobilized metal affinity chromatography using Ni-Penta resin (Macherey-Nagel). We used the purification protocol from the Liu group⁶ with slight modification; imidazole concentration was reduced to 5 mM in the lysis buffer (100 mM HEPES, 300 mM NaCl, 0.1% Tween-20, 10 mM β-mercaptoethanol, pH 8). Increasing the imidazole concentration in the elution buffer to 250 mM eluted protein from the column. We removed incorporated metal ions and the His₆-tag simultaneously in the initial round of dialysis. We added TEV protease to the eluted protein and dialyzed the mixture against 100 mM HEPES, 300 mM NaCl, and 10 mM EDTA, pH 8, at 4 °C using a 6–8 kDa dialysis membrane. We removed His₆-tagged TEV protease, along with the cleaved His₆-tag from WelO5, by running a second Ni-Penta affinity column. We recovered cleaved WelO5 from the column with lysis buffer as the eluent. An Amicon 10K MWCO centrifugal filter (Millipore-Sigma) further concentrated eluting fractions containing protein. We removed EDTA with two more rounds of dialysis against 100 mM HEPES, 300 mM NaCl. We analyzed WelO5 for homogeneity and purity by a 10% SDS-PAGE (Figure S53).

WelO5 In vitro assay

In-vitro assay of WelO5 using the synthesized 1a (stock dissolved in DMSO) was carried out on a 100 μL scale according to the previously published protocol.³² In short, we combined WelO5 (80 μM), 1a (200 μM final concentration dissolved in MeOH), 2OG (2 mM), NaCl (10 mM), and ammonium iron(II) sulfate hexahydrate (500 μM) in 50 mM Tris pH 8.0 and incubated at 30 °C for 1 h. We terminated the reaction by extracting with ethyl acetate and analyzed the extract by LC–MS.

LC–MS Analysis

We injected the sample into a Thermo Scientific Vanquish UHPLC system, equipped with a diode array detector (DAD), which was coupled to a Thermo Scientific Q Exactive HF-X mass spectrometer featuring an H-ESI ion source. We chromatographically separated the analytes on an Agilent Zorbax Extend-C18 column (4.6 mm × ID 50 mm length, 1.8 μm particle size) at 35 °C that was equilibrated in 50% solvent A (water with 0.1% formic acid) and 50% solvent B (1:1 MeOH:ACN with 0.1% formic acid). Throughout a single injection, we applied a gradient to increase solvent B to 80% over 30 min. We detected analytes at 280 nm via the DAD, and simultaneously, detected in positive ion mode using a full-scan method with an H-ESI capillary temperature of 320 °C, a resolution of 120,000, an AGC target of 1×10⁶, and a scan range set to m/z 150–450.

CW EPR Spectroscopy

Prior to sample preparation, we washed solutions of isolated WelO5 with 1 mM ethylenediaminetetraacetic acid (EDTA) in 100 mM HEPES buffer by three dilute-concentrate cycles using Amicon^® Ultra Centrifugal Filter, 10 kDa MWCO concentrator. We obtained vanadyl-incorporated samples of WelO5 by incubating WelO5 with vanadyl sulfate (VOSO₄) at 1:1 molar ratio. We prepared the EPR samples in 250 μL aliquots with the following final concentrations of constituents: WelO5 (1 mM), vanadyl sulfate (1.1 mM), succinate (10 mM), NaCl (50 mM). We prepared samples in 100 mM HEPES buffer at pH 8 without additional NaCl. For the sample with substrate, we added synthetic 1a to the sample solution at a final concentration of 5.8 mM. We removed the EDTA by subsequent two dilute-concentrate cycles with 100 mM HEPES buffer.

We analyzed the EPR spectra using a Magnettech MS-5000X spectrometer equipped with an Oxford ESR900 variable-temperature helium-flow cryostat via a custom framework. A LakeShore 335 Cryogenic Controller controlled the temperature. The microwave frequency was 9.434 GHz, microwave power was set to 100 μW, modulation amplitude was set to 0.5 mT, and the scan time was 300 s with 4096 points.

We used Kazan Viewer, a home-written suite of utilities in MATLAB (MATLAB r2017a, The Mathworks Inc.), to process data and simulate spectra.³³ We employed the “pepper” utility from the EasySpin software package to perform the simulation shown in the Supporting Information.

Safety statement

The synthesis of 9 uses the inorganic azide salt, lithium azide (19597-69-4). Metal azides are very toxic, particularly when dissolved in a polar aprotic solvent such as DMF (68-12-2). Synthetic chemists should handle them with extra caution.

Supplementary Material

IR, NMR, and mass spectra and GC and LC traces of compounds prepared during this study (38 pages), scan of SDS-PAGE gel showing expression of WelO5 (1 page), EPR spectra of WelO5·VO·Cl–·succinate including simulations (1 page).

Data Availability

The data underlying this study are available in the published article and its Supporting Information.