Enzymatic spiroketal formation via oxidative rearrangement of pentangular polyketides

The structural complexity and bioactivity of natural products often depend on enzymatic redox tailoring steps. This is exemplified by the generation of the bisbenzannulated [5,6]-spiroketal pharmacophore in the bacterial rubromycin family of aromatic polyketides, which exhibit a wide array of bioactivities such as the inhibition of HIV reverse transcriptase or DNA helicase. Here we elucidate the complex flavoenzyme-driven formation of the rubromycin pharmacophore that is markedly distinct from conventional (bio)synthetic strategies for spiroketal formation. Accordingly, a polycyclic aromatic precursor undergoes extensive enzymatic oxidative rearrangement catalyzed by two flavoprotein monooxygenases and a flavoprotein oxidase that ultimately results in a drastic distortion of the carbon skeleton. The one-pot in vitro reconstitution of the key enzymatic steps as well as the comprehensive characterization of reactive intermediates allow to unravel the intricate underlying reactions, during which four carbon-carbon bonds are broken and two CO2 become eliminated. This work provides detailed insight into perplexing redox tailoring enzymology that sets the stage for the (chemo)enzymatic production and bioengineering of bioactive spiroketal-containing polyketides.

Moreover, details on the biosynthesis of the rubromycins and the spiroketal moiety remain scarce. First insights originated from extensive gene-inactivation studies with the griseorhodin (grh) A biosynthetic gene cluster of Streptomyces sp. JP95 isolated from the marine tunicate Aplidium lenticulum, which was expressed in the heterologous producer S. albus sp. J1074 9,10 . Initial steps resemble typical type II polyketide pathways involving a minimal polyketide synthase (PKS) that likely utilizes an acetyl-CoA starter unit and 12 malonyl-CoA extender units to generate a highly reactive acyl-carrier protein (ACP)-bound poly-β-ketone chain. Following enzyme-catalyzed regioselective ketoreduction, cyclization, aromatization and ACP elimination, further tailoring reactions modify the polyketide backbone and lead to the advanced and highly oxidized intermediate collinone (3) (previously also isolated from a heterologous producer expressing parts of the rubromycin biosynthetic gene cluster 15 ), which may serve as a direct precursor for spiroketalization 10 . This would necessitate an extensive oxidative backbone rearrangement as well as the elimination of two C 1 units, which may be mediated by mechanistically versatile flavin-dependent enzymes [16][17][18][19][20][21][22] that often facilitate redox tailoring reactions in natural product biosynthesis ( Fig. 1) 16,19 .
Here, we report the full in vitro reconstitution of enzymatic spiroketal formation in the biosynthesis of rubromycintype polyketides. We elucidate the conversion of 3 into the 1 kb Fig. 1 Overview of the proposed biosynthesis of bacterial rubromycin-type polyketides and final pathway products. a Griseorhodin A biosynthetic gene cluster encoding, e.g., the minimal type II polyketide synthase (PKS), cyclases, and tailoring enzymes 9 . b Initial steps afford a reactive acyl-carrier protein (ACP)-bound poly-β-ketone, which is subsequently cyclized and modified to 3. Compounds 3 and 11 were previously identified in the course of gene deletion experiments (ΔgrhO5 and ΔgrhO6, respectively, encoding flavin-dependent tailoring enzymes investigated in this work) and assigned as putative advanced intermediates 10 . The conversion of 3 into 4 via 8 and 11 (dashed box) and additional intermediates was elucidated in this work. A ketoreductase (presumably GrhO10) then converts 4 into 13. c Examples of mature rubromycins likely formed from 13. [5,6]-spiroketal-containing 7,8-dideoxy-6-oxo-griseorhodin C (4) via various reactive intermediates by the concerted action of the flavoprotein monooxygenases GrhO5 and GrhO6, as well as the flavoprotein oxidase GrhO1 that are encoded by the grh gene cluster. This process is primarily mediated by the multifunctional monooxygenase GrhO5 that oxidatively rearranges the carbon backbone and ultimately forms a [6,6]-spiroketal and is assisted by GrhO1, before the ring-contracting GrhO6 generates the [5,6]spiroketal pharmacophore found in mature rubromycin polyketides (Fig. 1).

Results
Flavoprotein monooxygenase GrhO5 initiates spiroketal formation by rapid collinone reduction. S. albus sp. J1074 KR8 (ΔgrhO5) was previously reported to produce 3 as major shunt metabolite, which may thus represent the native substrate of GrhO5 10 . To investigate this, 3 was isolated from the S. albus mutant, while GrhO5 (fused with an N-terminal maltose binding protein tag) was obtained from the heterologous producer Escherichia coli BL21 DE3 (see Online Methods section for details on gene cloning as well as production and purification of enzymes and compounds). GrhO5 is predicted to function as flavoprotein monooxygenase based on the amino acid sequence 10 and is homologous to the NAD(P)H-and FAD-dependent class A flavoprotein monooxygenases with "glutathione reductase type" Rossmann fold 21 . Typically, these enzymes catalyze aromatic hydroxylation reactions via an electrophilic flavin-C4ahydroperoxide oxygenating species, while some members instead act as Baeyer-Villiger monooxygenases (BVMOs) that employ a nucleophilic flavin-C4a-peroxide anion 22,23 . The purified enzyme showed an intense yellow coloration indicative of a bound flavin cofactor that was further determined as flavin adenine dinucleotide (FAD; Supplementary Fig. 1). Under optimized assay conditions, GrhO5-dependent consumption of 3 could indeed be observed by UV-Vis spectroscopy in the presence of the electron donor NADPH (≈20% activity with NADH; see Supplementary  Fig. 2 for kinetics). To further investigate this and elucidate the reaction course, samples from enzyme reactions were quenched after different incubation times, the compounds extracted and then analyzed by reverse-phase high performance liquid chromatography (RP-HPLC).
First, GrhO5 catalyzed the rapid conversion of 3 into intermediate 5 ( Supplementary Fig. 3). Extracted 5 featured a distinct UV-Vis spectrum and intense yellow color, as compared to the purple-red 3. Liquid chromatography high-resolution mass spectrometry (LC-HRMS) indicated that 5 represents a reduced form of 3, which spontaneously reoxidized in the presence of O 2 , as shown by the color change and confirmed by RP-HPLC ( Supplementary Fig. 3). This was further supported by the nonenzymatic chemical reduction of 3 (using Ti(III) citrate or DTT), which also afforded 5 ( Supplementary Fig. 3a). Notably, compared to the much faster GrhO5-dependent 5 formation, NADPH (±free FAD) only reduced 3 at very low rates ( Fig. 2 and Supplementary Fig. 2c). To solve the structure of 5 and of other compounds described below, large scale enzymatic assays were conducted. Anaerobic conditions enabled the complete conversion of 3 into 5, which was afterwards extracted, purified via RP-HPLC, and lyophilized. NMR spectroscopy ( 1 H NMR, 13 C NMR, HSQC, HMBC, Supplementary Figs. 4-7) in a sealed, anaerobic tube then identified 5 as ring A-reduced dihydrocollinone featuring a naphthohydroquinone moiety (Fig. 3a).
Dihydrocollinone is oxygenolytically cleaved by GrhO5 and converted to secocollinone. Compound 5 was not further processed by GrhO5 under anaerobic conditions, suggesting a succeeding O 2 -dependent reaction probably in form of an oxygenation. Indeed, in presence of O 2 , 3 was first reduced to 5 and then further converted; however, significantly reduced amounts of putative intermediates/products were subsequently retrieved. We thus suspected that GrhO5 converts 5 into a labile and/or insoluble intermediate. To investigate this, chemical derivatization of the putative intermediate was attempted after quenching of the enzyme assay. Methylation of the reactive phenol groups by dimethyl sulfate (DMS) proved suitable and afforded a completely methylated and reduced derivative (6) of an enzymatically hydroxylated intermediate based on LC-HRMS and tandem MS (MS 2 ) analysis ( Supplementary Fig. 9). Comprehensive NMR spectroscopy (Supplementary Figs. 10-13) of 6 revealed a ruptured D-ring as well as a hydroxylated E-ring. We propose the native intermediate to be secocollinone 7, which is likely derivatized to 6 by non-enzymatic reduction (via NADPH), keto-enol tautomerization (that triggers water elimination at C18), and complete methylation of the free hydroxyl groups (Fig. 3a).
To confirm that 7 represents an intermediate, the time course for 6 formation was scrutinized in the GrhO5 assay. For that, samples were withdrawn at different time points, quenched, and methylated. HPLC analysis of the derivatized samples then showed the formation of significant amounts of 6 at 1 min, a maximum amount at 3 min, and only residual amounts after 4 min, consistent with the envisaged formation of 7 from 5 ( Fig. 3a and Supplementary  Fig. 9a). Based on these results, we propose the following reaction sequence. First, ring A of 3 is reduced to 5 (Fig. 3a, step I), before aromatic hydroxylation at C4 of 5 (step II) prompts the first C-C bond cleavage of ring D via retro-aldol condensation to afford 7 (step III). The flavin-mediated hydroxylation most likely proceeds via an electrophilic flavin-C4a-hydroperoxide (Fl C4aOOH ) oxygen transferring agent, as discussed below.
Secocollinone is rearranged to the [6,6]-spiroketal-containing dihydrolenticulone. Concurrent with consumption of 7, compound 8 emerged in the HPLC chromatograms, followed shortly by shunt product 9 and on-pathway intermediate 10 (Figs. 2 and  3ab). Notably, isolated 8 proved highly unstable and LC-HRMS analysis verified the spontaneous conversion into both 9 and 10 ( Supplementary Fig. 14). Moreover, compound 10 rapidly autooxidized to compound 11, analogous to the oxidation of 5 into 3, as verified by RP-HPLC and LC-HRMS ( Supplementary Fig. 15). Compounds 9 and 11 were only moderately stable in aqueous solution, thus explaining why prolonged incubation (>30 min) of 3 with GrhO5 did not yield significant amounts of enzymatic product. Surprisingly, 11 was identified as the previously reported lenticulone featuring a [6,6]-spiroketal 10 and 10 accordingly as the respective ring A-reduced dihydrolenticulone based on LC-HRMS/MS 2 analysis, UV-Vis spectra, NMR spectroscopy, and comparison with a standard acquired from S. albus sp. J1074 MP66 (ΔgrhO6) 10 (Fig. 3b and Supplementary Figs. [16][17][18][19]. In contrast, comprehensive NMR spectroscopy and MS 2 fragmentation of shunt product 9 revealed a non-spiro compound with opened C-ring (Fig. 3b To consolidate these results, we propose that 7 first undergoes retro-aldol cleavage of ring C generating an α-keto aldehyde motif at C5-C6 (Fig. 3a, step IV). Then, a series or reversible steps takes place including carbonyl-hydrations, tautomerization, and ring D re-formation via hemiacetal formation (steps V-VII), before isomerization affords a C5-C6 α-hydroxy acid (step VIII). Subsequently, the C6 hydroxyl group is oxidized to the respective ketone, likely mediated by the flavin cofactor of GrhO5 (step IX). Lastly, ring C is reinstalled via attack of the C18 hydroxyl group on the C6-ketone to afford 8 (step X). Although highly labile 8 could not be isolated for structural elucidation by NMR, isotope-labeling studies with 8 corroborate this proposal. Accordingly, up to two 18 O atoms derived from H 2 18 O were observed in 8 by LC-HRMS/MS 2 ( Supplementary Fig. 26), consistent with water-dependent α-hydroxy acid formation (part of steps V-VIII) and incorporation of 18 O via reversible carbonyl hydrate formation (step VI).
Next, we propose that spontaneous oxidation of ring B of 8 allows the third C-C bond cleavage via facile decarboxylation. This is enabled by an appropriately positioned C15-ketone in ring B that acts as electron sink, thereby producing on-pathway intermediate 10 after keto-enol tautomerization, before ring A autooxidation yields 11 (step XI, Fig. 3b). Interestingly, in addition to reduction of 3, GrhO5 was also capable of reducing 11 into 10 with NAD(P)H ( Supplementary Fig. 17). Alternatively, decarboxylation can occur (without prior oxidation) through spiroketalopening enabled by the C17-ketone of ring C to afford shunt product 9 after tautomerization and autooxidation of ring A (step XII). To further investigate this mechanistic proposal, both 9 and 11 were enzymatically produced in presence of either 18 (Fig. 3b). The UV-Vis data were furthermore fully consistent with the proposed redox states of these intermediates and clearly indicated a napthohydroquinone moiety with reduced rings A and B for 5, 8, and 10 ( Supplementary  Fig. 29a), while spectral features of 3, 9, and 11 suggested oxidized quinonic A rings ( Supplementary Fig. 29b).
The quinone-forming oxidase GrhO1 boosts dihydrolenticulone formation. Although GrhO5 proved sufficient for in vitro formation of 10, ≈50% of shunt product 9 was generated. This prompted us to consider additional enzymes that could mediate the oxidation of ring B of 8 and thus promote formation of 10/11. Previous studies with deletion mutants suggested that GrhO1 may be involved in spiroketal maturation 10 . GrhO1 also seemed a prime candidate because it is a member of the vanillyl alcohol oxidase/para-cresol methylhydroxylase (VAO/PCMH) flavoenzyme family that typically catalyze (non-oxygenative) two-electron oxidation reactions rather than hydroxylations, although some exceptions exist 18,24 . Hence, we supplemented our in vitro GrhO5 assays with heterologously produced polyhistidine-tagged GrhO1 that features a covalently bound flavin (Supplementary Fig. 30). Indeed, the presence of GrhO1 minimized production of 9 and boosted formation of 10 ( Fig. 2), whereas purified 9 (and its reduced form) could not be converted by GrhO1 (Supplementary Fig. 31). These results indicate that GrhO1 indeed catalyzes the selective oxidation of ring B of 8 (step XI, Fig. 3b) and assists GrhO5 in [6,6]-spiroketal formation. Next, we analyzed the GrhO5 and GrhO1-produced 11 by circular dichroism (CD) spectroscopy, revealing a racemic mixture. Presumably, enzymatically formed 11 is chiral at first but undergoes spontaneous racemization via rapid ring-opening and recylization. Consistent with that, the racemate could not be separated by chiral HPLC (Supplementary Fig. 32).
GrhO6 completes formation of the mature [5,6]-spiroketal moiety. GrhO6 is also predicted to be a member of the class A flavoprotein monooxygenases and has 31.4% amino acid (aa) identity (70% sequence coverage) with GrhO5. The ΔgrhO6 mutant strain was previously shown to accumulate 11, suggesting a possible role for GrhO6 in spiroketal formation 10 . Hence, polyhistidinetagged GrhO6 was heterologously produced and isolated, which contained a tightly, but non-covalently bound FAD cofactor similar to GrhO5 ( Supplementary Fig. 33). Indeed, upon complementing the GrhO5/GrhO1 assay with GrhO6, a stable main product 12 was formed in vitro (  [39][40][41][42]. Indeed, when 4 was dissolved in water and incubated at pH 2 or pH 7 with temperatures ranging from 4 to 50°C, it was non-enzymatically converted to 12 in aqueous phase at neutral pH (i.e. enzyme assaylike conditions), while remaining largely stable under acidic conditions ( Fig. 4 and Supplementary Fig. 43a). When this reaction was carried out in H 2  To confirm that 11 is a substrate for GrhO6, the compound was purified and used for separate in vitro assays with GrhO6. Similar to GrhO5, GrhO6 first reduced 11 into 10 before forming 12/4 in the presence of NADPH (≈40% activity with NADH), as quantified by UV-Vis spectroscopy (see Supplementary Fig. 2 for kinetics) and further analyzed by RP-HPLC, whereas 9 was not accepted as substrate ( Supplementary Fig. 49a, b). Moreover, the requirement for NAD(P)H suggested a monooxygenase functionality for GrhO6. This was further confirmed by 18 Table 6). GrhO6 likely employs a similar mechanism to GrhO5. Following NAD(P)H-dependent 10 formation, GrhO6 then likely catalyzes an aromatic hydroxylation at C2 of 10 (step XII, Fig. 3c), which dearomatizes ring B and affords a C3ketone in β-position to the C1-lactone moiety. This enables the subsequent decarboxylation of C1 via formation of a transient oxocarbenium that is quenched through intramolecular nucleophilic attack by the introduced C2-hydroxyl-group and thereby generates the [5,6]-spiroketal of 4 ( Fig. 3c). CD spectroscopy verified the chirality of compound 4 that features the same configuration at the C6a as 1 10,25 , suggesting that [5,6]-spiroketal formation is stereochemically controlled by GrhO6 (Supplementary Fig. 50).
Metabolomic and genomic analyses support a universal pathway for spiroketal formation in the rubromycin polyketide family. In contrast to 3, 4, 11, the intermediates 5, 7, 8, 9, 10, 12 were hitherto uncharacterized. To verify the authenticity of some key pathway intermediates, we extracted liquid cultures at various time points during growth of S. albus sp. J1074 (heterologous 1 producer) 10 , Actinoplanes ianthinogenes (native producer of purpuromycin 26 ), and S. puniceus (native producer of 1 and griseorhodin C) 25 . RP-HPLC and LC-HRMS analysis clearly confirmed the presence of 10 (that likewise underwent spontaneous oxidation to 11) and 12 in various cell free lysates, while 5, unstable 8, and shunt product 9 were not observed. Also, 7 could not be directly detected as a result of the poor interaction with chromatographic stationary phases. However, complex compound mixtures including insoluble compounds prone to polymerization were previously reported (but not further characterized) in the ΔgrhO1 mutant strain S. albus J1074 KR5 10 . Indeed, upon treatment of the cell free lysate of a culture of this strain with DMS, 6 could be obtained ( Supplementary Fig. 9b).
Furthermore, genomic data suggest that the biosynthetic gene clusters for production of spiroketal-containing 1 (S. sp. JP95), 2 (of S. collinus) 15 , hyaluromycin (of S. hyaluromycini MB-PO13) 27 , and of heliquinomycin (S. piniterrae) 28 harbor grhO5, grhO1, and grhO6 homologs (the genome of the Fig. 3 Proposed spiroketal formation in rubromycin polyketide biosynthesis. Redox steps catalyzed by GrhO5, GrhO1, and GrhO6 are highlighted with the respective enzyme symbols. Other steps may be catalyzed or occur spontaneously in the enzymes' active sites, see text for details. The rings A-E of compounds that undergo modification in the ensuing biosynthetic step are color-coded according to the final product 4 (bottom left); important carbons are numbered for each step. Note that the carbon numbering of intermediates up to 8 is according to compound 3, whereas different carbon numberings are used for 10/11 and the final product 4. Oxygen atoms derived from O 2 and H 2 O (based on isotope-labeling experiments) are highlighted in red and blue, respectively. Water-derived 18 O is most likely also incorporated from spontaneous and reversible keto hydration as shown in step VI (gray arrows). a Proposed GrhO5-mediated conversion of 3 into 8. For secocollinone detection and structural characterization by NMR, derivatization to 6 by dimethyl sulfate (DMS) was required (gray dashed box). b Formation of on-pathway intermediate 10 (black arrows) and shunt product 9 (gray arrows) from 8. GrhO1 boosts 10 formation, while 9 production is minimized (see Fig. 2). c Proposed GrhO6-catalyzed formation of 4 from 10/11. Presumably, 3 and 11 are largely skipped as intermediates in the reducing environment of the cell. Dashed arrows indicate autooxidation steps. Electron-dependent steps (indicated by 2[H]) require NADPH as preferred reductant for both GrhO5 and GrhO6, see text. The spiroketal-configuration of 11 (racemic) and 4 (chiral) were analyzed by CD spectroscopy, see text. All intermediates in boxes were characterized by NMR and/or HRMS. Note that the final product 4 spontaneously converts to 12 (see below). bacterial purpuromycin producer is not available yet). Compared to that, respective candidate genes were absent in the biosynthetic gene clusters of the non-spiroketal compounds benastatin A, fredericamycin A (featuring a carbaspirocycle), pradimicin A, lysolipin, and arixanthomycins ( Fig. 5 and Supplementary Fig. 51). To further verify a common pathway for spiroketal formation in the rubromycin polyketide family, we acquired heterologously produced polyhistidine-tagged RubLthe predicted functional homolog of GrhO5 encoded by the rubromycin biosynthetic gene clusterand showed that it catalyzed the virtual identical conversion of 3 into 10/11 ( Supplementary Fig. 52).

Discussion
Ever since their discovery, rubromycin-type polyketides have sparked the interest of synthetic chemists and biochemists because of their complex framework and potent bioactivities. While spiroketal groups are occasionally encountered in natural products, adjacent aromatic rings giving rise to distinct bisbenzannulated Fig. 4 Chemical preparation of 4 from 13 (isolated from S. albus KR7 10 ). Synthesized 4 converts into 12 in aqueous solution at neutral pH, as also observed in the enzyme assays (Fig. 2). Note that the structure of 12 is supported by NMR and IR data, as well as DFT calculations, but a C6-ketohydrate instead of the shown ring-opened compound 12 cannot be ruled out (see Supplementary Fig. 48). Enzymatic steps are indicated with dashed arrows, nonenzymatic chemical steps with bold arrows. A ketoreductase may deter water addition in vivo en route to 1. The stereochemical configuration of 13 and 4/12 is depicted as previously determined for 1 and verified in this work. DMP: Dess-Martin periodinane, DCM: dichloromethane. Same colors indicate high amino acid sequence similarities and predicted similar functions. The biosynthetic gene clusters for production of the spiroketal-containing compounds griseorhodin A (1, AF509565), heliquinomycin (NZ_SUMB01000006), hyaluromycin (NZ_BCFL01000018), and rubromycin (2, AF293355) are shown. These clusters harbor grhO5 (green), grhO1 (blue), and grhO6 (green) homologs. Corresponding homologs are connected with a shadow and the percentages of the amino acid sequence identities of the encoded proteins are shown. The grhO8 and grhO9 genes also show homology to grhO5 and grhO6 and encode FAD-dependent monooxygenases involved in early redox tailoring steps. The biosynthetic gene clusters of the non-spiroketal pentangular polyketides lack grhO1, grhO5, and grhO6 gene candidates necessary for spiroketal formation (see Supplementary Fig. 51 for full comparison). NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-21432-9 ARTICLE NATURE COMMUNICATIONS | (2021) 12:1431 | https://doi.org/10.1038/s41467-021-21432-9 | www.nature.com/naturecommunications spiroketals are exceedingly rare 30 . These spiroketals are remarkably robust owing to the stabilizing effect of the connected aryl functionalities through delocalization of the lone pairs of electrons from the oxygen atoms 30 . In this work, we provide detailed insights into the enzymology and biochemistry underlying the installment of the [5,6]-spiroketal pharmacophore into the backbone of rubromycin natural products. These redox tailoring processes are driven by the multifunctional spiroketalase GrhO5 (assisted by GrhO1) that produces key intermediate 10, which is subsequently transformed by GrhO6 into 4. Both GrhO5 and GrhO6 resemble characterized class A flavoprotein monooxygenases, e.g., GrhO5 is related to the aromatic hydroxylase RdmE (40.7% aa identity, 100% coverage) from aklavinone biosynthesis 31 and to XanO4 (42.3% aa identity, 98% coverage) catalyzing oxygenation and demethoxylation in xanthone biosynthesis 32 . Moreover, GrhO5 resembles the predicted aromatic hydroxylases GrhO8 (45.5% aa identity, 98% coverage) and GrhO9 (44% aa identity, 94% coverage) from 1 biosynthesis that presumably catalyze early tailoring steps en route to 3 10 . On the other hand, GrhO6 is more closely related to aromatic hydroxylases such as OxyS (43.9% aa identity, 94% coverage) from oxytetracycline biosynthesis 33 and verified BVMOs, e.g., MtmOIV (45% aa identity, 92% coverage) from mithramycin biosynthesis 34 (Supplementary Fig. 53).
Our results imply a reaction sequence for GrhO5 that combines versatile redox chemistry (quinone reduction, aromatic hydroxylation, and alcohol oxidation) with various non-redox steps (e.g., retro-aldol ring cleavage and spiroketalization). While flavoenzyme functionalities such as quinone reduction 35 , quinone reduction coupled to aromatic hydroxylation 36 , or alcohol oxidation 37 have been reported before, a single flavoenzyme that combines these steps and additionally drastically rearranges the carbon backbone is noteworthy. Notably, although initial substrate reduction is not common for flavoprotein monooxygenases 22 , it seems to be a more widely employed strategy in natural product biosynthesis for the activation of quinonic compounds prior to further redox transformations of the resulting hydroquinones or respective diketotautomers, e.g., in aflatoxin or actinorhodin polyketide biosynthesis 36,[38][39][40] . In the case of GrhO5, however, the first reduction of ring A is distant from the actual hydroxylation site and serves a further purpose: it is the requisite for the conversion of 8 into 10, as a quinonic ring A would deter ring B oxidation and thus lead to accumulation of shunt product 9 (Fig. 3b). The flavoprotein oxidase GrhO1 additionally counteracts 9 generation by catalyzing the selective oxidation of the hydroquinonic ring B to the p-quinone and thereby redirecting the metabolic flow to formation of 10 (Fig. 3b). Presumably, oxidized intermediates such as 3 and 10 might be largely skipped in the reductive environment of the cell and the quinonereducing flavoenzyme functionality of GrhO5 and GrhO6 would thus be primarily required for the salvaging of autooxidation products.
Finally, spiroketal maturation is completed by flavoprotein monooxygenase GrhO6 that converts the [6,6]-spiroketal moiety into the mature [5,6]-spiroketal. Based on the nucleophilicity of C2 of substrate 10, an aromatic hydroxylation appears likely (an alternative carbonate-producing Baeyer-Villiger oxidation of the C1-lactone 10 is conceivable, but chemically unfavorable 37 ). This hydroxylation then enables the final fourth C-C bond cleavage via decarboxylation, followed by recyclization to the [5,6]-spiroketal of 4 (Fig. 3c). Our combined data suggest a universal flavoenzyme-mediated pathway for spiroketal formation in the rubromycin polyketide family. Presumably, final tailoring steps on the molecular periphery of 4 then expedite the diversification of rubromycin polyketides. Notably, 4 readily formed ring-opened derivative 12 and it remains to be seen if this compound can be further converted or represents a dead-end product. Most likely, keto reduction is required to decrease ring-strain and prevent this non-enzymatic spiroketal-opening in vivo, consistent with the evident lack of mature rubromycins with C6-ketone functionalities.
The complex flavin enzymology underlying the formation of the rubromycin pharmacophore thus provides an astonishing solution for the generation of stable chroman spiroketals in nature. These complex redox transformations of a polycyclic, aromatic backbone starkly contrast with synthetic approaches (that rely on convergent syntheses in which the naphthazarin and isocoumarin moieties are separately generated and combined prior to spiroketal generation 12 ) as well as biosynthetic strategies and precursor molecules. Conventional enzyme-dependent routes rely, e.g., on a dihydroxyketone motif allowing hemiketal formation and dehydrative cyclization to afford the [6,6]-spiroketal of the pesticide avermectin 41 . Similarly, the proposed cascade-like [5,6]-spiroketal formation in the generation of the polyether antibiotics such as lasalocid A and monensin requires a hydroxy-ketone-epoxide motif and proceeds via initial hemiketal formation followed by epoxide-dependent spiroketalization 42 . Moreover, a radical mechanism has been proposed in griseofulvin biosynthesis 43 . Compared to these straightforward pathways, precursor 3 lacks evident structural requisites and functionalities for spiroketal formation and instead the pentangular precursor undergoes an unanticipated series of oxidative backbone rearrangements. Other noteworthy examples of flavoenzyme-driven oxidative rearrangements in type II polyketide biosynthesis are found, e.g., in chartreusin 44 , xanthone 32 , and rishirilide 45 biosynthesis. Interestingly, an alternative strategy to disrupt aromatic polyketide planarity is employed in enterocin biosynthesis 18,46,47 , where a single flavoprotein monooxygenase drastically rearranges a linear poly-β-ketone chain 18,24,48,49 rather than the cyclized backbone. Aside from flavoenzymes, other types of multifunctional monooxygenases (e.g., cytochrome P450s and non-heme iron α-ketoglutarate-dependent enzymes) have also been reported in natural product biosynthesis 50 .
Taken together, this study contributes to our understanding of the catalytic versatility of flavin-dependent enzymes in natural product biosynthesis and lays the foundations for future (chemo) enzymatic formation of novel rubromycin analogs and hybrid compounds. Our work illustrates a counterintuitive principle for the generation of structurally complex polyketide scaffolds and spiroketal pharmacophores.

Methods
Gene cloning, heterologous production, and protein purification Gene cloning. The grhO5, grhO1, and grhO6 genes were amplified from the cosmid pMP31 9 containing the grh gene cluster (all oligonucleotide primers from Sigma-Aldrich are shown in Supplementary Table 7). The grhO5 gene was inserted between the NdeI and PstI sites of the expression vector pMAL-c5x-His (New England Biolabs), the grhO1 gene between the NcoI and EcoRI sites of the expression vector pHis8 51 and the grhO6 gene between the EcoRI and PstI sites of the expression vector pETDuet-1 (Novagen, Merck KGaA). The rubL gene was synthesized (codon optimized) and purchased already cloned between the NdeI and XhoI sites of the expression vector pET-16b (BioCat).
Heterologous production and purification of GrhO5. The recombinant pMal-c5x-His-grhO5 plasmid was transformed into E. coli BL21 (DE3) (Thermo Fisher Scientific). LB-medium was inoculated with the transformed strain and supplemented with 100 μg/mL ampicillin at 37°C overnight under shaking at 180 rpm. Then, 1 L of TB-medium with 100 μg/mL ampicillin was inoculated with 10 mL of the starter culture and grown to an OD 600 of 0.5-0.6. The culture was induced with a final concentration of 250 μM IPTG and the cells were grown overnight at 18°C under shaking (130 rpm). The cells were retrieved by centrifugation at 4000 × g for 20 min at 4°C. Cell pellets were suspended in buffer A (20 mM Tris pH 7.4, 200 mM sodium chloride) with additional 1 mg/mL lysozyme and 1 mM phenylmethylsulfonyl fluoride and then stirred for 30 min at 4°C. Then, the cells were lysed by sonication and the cell fragments removed by centrifugation at 18,000 × g for 30 min at 4°C. The supernatant was filtered through a 0.45-μm syringe filter and then loaded onto a 5-mL MBPTrap HP Dextrin Sepharose TM column (GE Healthcare) connected to an FPLC system. Following a washing step with buffer A, bound mbp-tagged GrhO5 was eluted by a gradient ranging from 0% to 100% buffer B (20 mM Tris pH 7.4, 200 mM sodium chloride, 10 mM maltose) over 5 column volumes. GrhO5 was concentrated using a 30-kDa MW cut-off MACROSEP spin column (Pall Cooperation) and subsequently desalted with a 5-mL HiTrap desalting column (GE Healthcare). The identity and purity of recombinant GrhO5 was analyzed by SDS-PAGE.
Heterologous production and purification of GrhO1. The recombinant pHis8-grhO1 plasmid was transformed into E. coli Lemo21 (DE3). The production and purification was performed according to the procedure described above using kanamycin (50 μg/mL) and chloramphenicol (30 μg/mL) as antibiotics and 750 μM L-rhamnose for optimized expression levels. A 5-mL HisTrap TM FF crude column (GE Healthcare) was used for purification. Unbound protein was removed by washing with buffer A (50 mM Tris pH 7.1, 300 mM sodium chloride, 20 mM imidazole, 10% glycerol) and 0-6% buffer B (50 mM Tris pH 7.1, 300 mM sodium chloride, 500 mM imidazole, 10% glycerol) over five column volumes and bound his-tagged protein was then eluted with 100% buffer B. The protein was concentrated using a 10-kDa MW cut-off MACROSEP spin column (Pall Cooperation) and then further purified by size exclusion chromatography with a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) equilibrated with buffer A (50 mM Tris, pH 7.4, 300 mM sodium chloride) and connected to a FPLC system. Finally, 10% (v/v) glycerol was added to the fractions containing GrhO1. The identity and purity of the eluted protein was confirmed by SDS-PAGE.

Collinone (3) production and purification.
Spores of the S. albus grhO5 knock-out strain were inoculated into 50 mL of TSB medium supplemented with apramycin (50 μg/mL) and grown at 30°C, 200 rpm for 24 h. The resulting culture was used to inoculate 3 L of TSB medium distributed to 1 L flasks with stainless-steel springs. Each flask contained 300 mL of culture supplemented with apramycin (50 μg/mL) and was incubated at 30°C, 200 rpm for 4 days. The cells were harvested by centrifugation and the pellet was suspended in dH 2 O (300 mL). The pellet/water solution was acidified to pH 1 with 2 N HCl, EtOAc (500 mL) was added and the mixture then stirred overnight. The supernatant was also acidified to pH 1 with 2 N HCl and extracted with EtOAc (3 × 500 mL for 1 L culture). The combined extracts of the supernatant were purified by column chromatography [SiO 2 , CH 2 Cl 2 /MeOH = 9:1 (ν/ν)]. Fractions containing 3 (detected via thin layer chromatography (TLC)) were combined, dried, and further purified by semi-preparative HPLC. The extract of the pellet fraction was directly purified by semi-preparative HPLC. An average of 10 mg/L cell culture of 3 was obtained. The purity of the red solid 3 was confirmed by highresolution liquid chromatography mass spectrometry (LC-HRMS) and 1D NMR analysis. The acquired NMR data are in accordance with previously published data 2,15  Lenticulone (11) production in S. albus KR42. Spores of the S. albus KR42 strain were inoculated into 50 mL of LB medium supplemented with apramycin (50 μg/mL) and grown at 30°C, 200 rpm for 4 days. The cells were harvested by centrifugation and the supernatant was acidified to pH 3 with 2 N HCl and extracted with EtOAc (2 × 100 mL). The solvent of the combined extracts was removed under reduced pressure and the residue dissolved in CH 3 CN and EtOAc. After centrifugation for 10 min at 18,000 × g, samples were analyzed by HPLC using a NUCLEODUR 100-5 C18ec column (varioprep, 250 mm × 10 mm ID, 5 μM, Macherey-Nagel) coupled with a UNIVERSAL RP guard column (4 mm × 3 mm ID, Macherey-Nagel).
Photometric assays for kinetic parameters of GrhO5 and GrhO6. The kinetic parameters for GrhO5 and GrhO6 were determined photometrically at wavelengths of 556 for collinone (3) and and 536 nm for lenticulone (11) using a Shimadzu Photometer (UV-1650PC, Shimadzu, Duisburg, Germany). A mixture of 50 mM Tris pH 7.4, 1.5 mM NADPH or NADH and substrate (16,39,79, 392 μM collinone (3) or 20, 50, 100 μM lenticulone (11), both dissolved in CH 3 CN) was prepared. The reaction was started by addition of 1 μM GrhO5 or GrhO6 (or free FAD as control) that allowed the measurement of the linear decrease in absorption. The turnover rates were calculated based on extinction coefficients of 4850 L × mol −1 × cm −1 at 556 nm (3) and 1700 L × mol −1 × cm −1 at 536 nm (11) according to Beer-Lambert law. The slope of the absorption was read out by the software (UV Probe 2.43, Shimadzu).
Enzymatic assay with GrhO5, GrhO1, and GrhO6. The enzyme assay was performed identical to the standard GrhO5 assay with additional 10 μM of GrhO6 and/or 5 μM of GrhO1. Reactions were quenched after 8 min and analyzed as described above.
Enzymatic assay to isolate dihydrocollinone (5). The enzymatic assay was performed under anaerobic conditions using the above described assay composition. The reaction was started in an anaerobic chamber by adding solution A (50 mM Tris, 1 μM GrhO5, 1% (v/v) HPCD and collinone (0.25 mM, dissolved in CH 3 CN, final conc. of 2.5% (v/v)) to solution B (10% glycerol, 1.5 mM NADH and NADPH). After 1 min the reaction mixture was extracted [EtOAc/formic acid = 9:1 (ν/ν)] inside the anaerobic chamber. The mixture was centrifuged for 30 s and the layers then separated inside the anaerobic chamber. The next steps were performed under minimal exposure to O 2 . The solvent was removed under reduced pressure and the residue dissolved in CH 3 CN and EtOAc. After centrifugation for 10 min at 18,000 × g, compound 5 was purified by HPLC using a NUCLEODUR 100-5 C18ec column (varioprep, 250 mm × 10 mm ID, 5 μM, Macherey-Nagel) coupled with a UNIVERSAL RP guard column (4 mm × 3 mm ID, Macherey-Nagel). The yield of 5 was~12 mg of orange solid. The product was analysed by 1 H, 13  Enzymatic assays with compound 9. A solution of 50 mM Tris pH 7.4, 1.5 mM NADH/NADPH, 10 μM of either GrhO5, GrhO1 or GrhO6, and 1% (ν/ν) HPCD) was prepared. The reaction was started by addition of compound 9 (dissolved in CH 3 CN, final conc. of 2.5% (ν/ν) CH 3 CN in assay) before incubation for 5 min at 30°C while shaking at 750 rpm. The reaction was quenched by addition of EtOAc/ formic acid [9:1 (ν/ν)] and extracted by vigorous shaking. The mixture was centrifuged for 30 s and the layers then separated. The solvents were removed under reduced pressure and the compounds dissolved in CH 3 CN (20 μl) and EtOAc (20 μl). After centrifugation for 10 min at 18,000×g, samples were analyzed by HPLC.
Enzymatic assay to isolate compound 9 and compound characterization. A large scale enzyme assay (200 mL distributed to Eppendorf tubes each containing 500 μL) was performed using the same composition of constituents as described for the standard GrhO5 assay. The reaction was quenched after 15 min and compound 9 purified as described above for 5 but under aerobic conditions. The yield of 9 was~2 mg of red solid. The product was analysed by 1 H, 13 104.2, 110.7, 116.1, 118.4, 128.3, 128.3, 137.4,  141.7, 141.7, 147.3, 147.3, 151.2, 151.2, 161.5, 166.1, 166.1, 183.1, 190.0, 197.9; Most signals are unambiguously assigned, also due to HSQC and HMBC signals. Some signals of quaternary carbon atoms are extraordinarily strong because of the complete overlap of two carbons. The signals in italics show the carbon atoms with lacking HSQC and HMBC signals that are assigned due to chemical shifts observed for biosynthetically related molecules (see Supplementary Figs. 20-24); Notably, the proton attached to C13 could not be detected in DMSO-d 6 , possibly because of a proton exchange with the adjacent carboxyl group or interaction with metals that is often observed for such compounds. However, the measurement in 1,4-dioxane-d 8 revealed the proton attached to C13 and HRMS as well as extensive MS/MS fragmentation including 18  Standard assay with lenticulone (11) and GrhO6. A solution of 50 mM Tris pH 7.4, 1.5 mM NADH and/or NADPH, 10 μM GrhO6, and 1% (ν/ν) HPCD was prepared. The reaction was started by addition of lenticulone (dissolved in CH 3 CN, final conc. of 2.5% (ν/ν) CH 3 CN in assay) before incubation for 30 min at 30°C while shaking at 750 rpm. The reaction was quenched by addition of EtOAc/formic acid [9:1 (ν/ν)] and extracted by vigorous shaking. The mixture was centrifuged for 30 s and the layers then separated. The solvents were removed under reduced pressure and the compounds dissolved in CH 3 CN (20 μl) and EtOAc (20 μl). After centrifugation for 10 min at 18,000 × g, samples were analyzed by HPLC.
Enzymatic assay to isolate lenticulone (11) and compound characterization. The large scale enzyme assay (75 mL distributed to Eppendorf tubes each containing 1 mL) was performed using the same composition of constituents as described for the standard GrhO5 assay with additional 5 μM of GrhO1. The reaction was quenched after 10 min and 10/11 were isolated as described above for 5 but under aerobic conditions (collected compound 10 subsequently autooxidized to 11). The yield of 11 was~1.1 mg of red solid. The product was analysed by 1 H, 13 C NMR, UV/Vis, and ESI-HRMS. The acquired NMR data are in accordance with previously published data 10  Reduction of lenticulone (11). The enzymatic assay was performed under anaerobic conditions. The reaction was started in an anaerobic chamber by adding solution A (50 mM Tris pH 7.4, 1.5 mM NADH and/or NADPH, 5 μM GrhO5 or GrhO6 or 1 mM Ti(III)citrate and 1% (ν/ν) HPCD) to solution B (10% glycerol, NADH and NADPH). After 1 min the reaction mixture was extracted [EtOAc/formic acid = 9:1 (ν/ν)], the layers separated and the mixture analyzed on a photometer inside the anaerobic chamber. 18 O-labeling assays with GrhO5. The dioxygen in solution A (Tris pH 7.4, 50 mM, 10 μM GrhO5, and 1.5 mM NADH and NADPH) and solution B (0.5 mM compound 3, 1% (v/v) HPCD and 10% glycerol) was thoroughly removed by 10 cycles of treatment with vacuum and nitrogen gas. Solution B was added to solution A in an anaerobic chamber and the reaction was started by injection of about 97% 18 O 2 gas (Campro Scientific). The reaction was stirred at room temperature for 1 and 5 min and then extracted [EtOAc/formic acid = 9:1 (ν/ν)]. The solvent was removed under reduced pressure and the residue dissolved in CH 3 CN (20 μl) and EtOAc (20 μl) before LC-HRMS analysis. The assays with H 2 18 O were performed corresponding to typical GrhO5 assays with at least 55% of labeled water in the reaction mixture. Secocollinone (7) derivatization via methylation and compound characterization. S. albus KR5 spores were inoculated into 100 mL of TSB medium and grown overnight. The resulting seed culture was used to inoculate 9 L of TSB medium, which was distributed to thirty 1 L baffled flasks with stainless-steel springs. Flasks were incubated at 200 rpm for 4 days at 30°C. The cells were harvested by centrifugation and resuspended in H 2 O (500 mL) with trifluoroacetic acid (0.1% TFA; at pH 3). EtOAc (1 L) was added and the solution was stirred overnight at room temperature. After centrifugation, the layers were separated and the solvent of the organic layer was removed under reduced pressure. In total, 1.4 g crude extract was obtained from 9 L of culture. To a suspension of crude extract (1.4 g) of S. albus KR5 and anhydrous K 2 CO 3 (15 g, 108.5 mmol) in dry acetone (150 mL), dimethyl sulfate (DMS) (17 mL, 22.6 g, 179.3 mmol) was added dropwise. The mixture was stirred at room temperature for 1 h and then heated under reflux for 20 h at 40°C. Afterwards, the mixture was cooled to room temperature and diluted with H 2 O (300 mL). After stirring for additional 30 min, the mixture was extracted with EtOAc (500 mL). The layers were separated and the organic layer was washed with water. The solvent was removed under reduced pressure. The remaining extract was purified by column chromatography [SiO 2 , CH 2 Cl 2 /EtOAc = 1:1 (ν/ν)]. Fractions containing 6 (detected via TLC) were combined, dried, and further purified by semi-preparative HPLC. After a second purification step, the isolated yield was 4 mg of a yellow solid. The purity of 6 was confirmed by LC-HRMS and NMR analysis. TLC (CH 2 Cl 2 /EtOAc,  (13) and compound characterization. S. albus KR7 spores (50 μL) were inoculated into 50 mL LB medium and grown at 30°C, 200 rpm for 24 h. 5 mL of the starter culture was transferred into 10 flasks containing 300 mL LB medium and cultivation was continued at 30°C, 200 rpm for 96 h. This procedure was repeated four times resulting in a total culture volume of 12 L. After cell harvesting by centrifugation (60 min, 4000 × g, 4°C) the supernatant was separated from the pellet and the pellet was resupended in H 2 O (4 × 500 mL). Pellet suspensions were acidified with HCl (6 M, 20 mL) in 500 mL portions. EtOAc (2 × 500 mL) was added and the mixture was stirred overnight. The combined organic layers were concentrated under reduced pressure and the residue subjected to a short silica column [SiO2, CH2Cl2/MeOH/AcOH = 95:5:0.1]. Fractions containing compound 13 were combined. Solvents were removed under reduced pressure and further purified via semi-preparative HPLC, which yielded 120 mg of red solid from the pellet. The supernatant was acidified with conc. HCl (10 mL/L) and extracted with EtOAc (1 L culture with 2 × 500 mL). The combined organic layers were filtered and evaporated under reduced pressure yielding 2.60 g of crude extract from the supernatant. Further purification was performed by column chromatography [SiO2, CH 2 Cl 2 /MeOH/AcOH = 98:2:0.1 (ν/ν) (2 L), then CH 2 Cl 2 / MeOH/AcOH = 95:5:0.1 (ν/ν)] and provided 750 mg of crude product. The crude product was further purified via semi-preparative HPLC. Solvents were removed under reduced pressure yielding again 11.6 mg of red solid isolated from the supernatant. The overall yield was 131.6 mg of red solid compound 13. The product was analysed by 1 H, 13  .89 eq) was added and the reaction mixture was stirred at room temperature for 24 h. The solvent was removed under reduced pressure at room temperature and the residue dissolved in CH 3 CN (100 μL). After centrifugation (10 min, 18,000 × g, 4°C), 4 was isolated by semi-preparative HPLC. The collected compound 4 was directly frozen in liquid nitrogen and solvents were removed by subsequent lyophilization for at least 24 h yielding 15 mg of 4 (30.5 μmol, 75%) as dark purple solid. The product was analysed by NMR, UV/Vis and ESI-HRMS. The acquired NMR data are in accordance with previously published data 53 (750 rpm). The reactions were quenched by addition of EtOAc/formic acid [9:1 (ν/ν)] and extracted by vigorous shaking. The mixture was centrifuged for 30 s and the layers then separated. The solvents were removed under reduced pressure and the residues dissolved in CH 3 CN (40 μl). After centrifugation for 10 min at 18,000 × g, samples were analyzed by HPLC.
Isolation of compound 12 and compound characterization. A. ianthinogenes spores (60 μL) were inoculated into 60 mL GYM medium and grown at 30°C and 200 rpm for 48 h. 5 mL each of the preculture were then transferred into 10 flasks equipped with a stainless-steel spring containing 200 mL GYM medium and culture was continued at 30°C and 200 rpm for~96 h. The metabolite production was monitored by HPLC. If sufficient amounts of compound 12 were detected, the culture was harvested by centrifugation (30 min, 4000 × g, 4°C). The separated supernatant was acidified with HCl (6 M to pH = 1) before dichloromethane (700 mL) was added. The mixture was stirred vigorously for 1 h followed by centrifugation (10 min, 2000 × g, 4°C). The solvent of the combined extracts was removed under reduced pressure and the residue dissolved in CH 3 CN. After centrifugation for 10 min at 18,000 × g, samples were analyzed and compound 12 isolated by HPLC using a NUCLEODUR 100-5 C18ec column (varioprep, 250 mm × 10 mm ID, 5 μM, Macherey-Nagel) coupled with a UNIVERSAL RP guard column (4 mm × 3 mm ID, Macherey-Nagel). The purification via semi-preparative HPLC yielded 31 mg of dark purple solid. The product was analysed by 1 H, 13  The signals of 16 carbon atoms are unambiguously assigned, also due to HSQC and HMBC signals. The remaining nine carbon atoms are not detectable possibly due to interaction with metals that is often observed for such compounds. However, IR spectroscopy and DFT calculations ( Supplementary Fig. 48) as well as high-resolution MS/MS spectrometry and UV-Vis data ( Supplementary Fig. 34: UV-Vis spectrum is indicative of intact western napthoquinone moiety of which some NMR signals are missing; the MS/ MS fragmentation pattern resembles compound 4) strongly support this structure; UV/Vis (CH 3  High Performance Liquid Chromatography. Compounds from enzymatic assays were analyzed or prepared by RP-HPLC on a 1100 series chromatographic system (Agilent Technologies) with a VP NUCLEODUR 100-5 C18ec column (250 × 10 mm ID, 5 μM, Macherey-Nagel) and a UNIVERSAL RP guard column (4 × 3 mm ID, Macherey-Nagel). The mobile phases were water with 0.1% trifluoroacetic acid (TFA; solution A) and acetonitrile with 0.1% TFA (solution B). A gradient was run with 2% B from 0 to 1 min, 2-100% B from 1 to 15 min, 100-2% B from 20 to 21 min and 2% B from 21 to 24 min. As second purification step for NMR samples, a Kinetex® Biphenyl 100 Å column (250 × 10 mm ID, 5 μM, Phenomenex) was used with a UNIVERSAL RP guard column (4 × 3 mm ID, Macherey-Nagel). The mobile phases were water with 0.1% TFA or 0.1% formic acid (FA; solution A) and acetonitrile with 0.1% TFA or 0.1% FA (solution B). Isocratic methods were used for compound 6 (90% B) and compound 9 (60% B) at a flow rate of 3-5 mL/min at 30°C and absorbance was measured at 254 nm as well as 350 nm, 488 nm, and 500 nm. For analysis of the flavin cofactors, a NUCLEODUR Sphinx RP (150 × 3 mm ID, 5 μM, Macherey-Nagel) with a UNIVERSAL RP guard column (4 × 3 mm ID, Macherey-Nagel) was used. The mobile phases were water with 0.1% TFA (solution A) and acetonitrile with 0.1% TFA (solution B). The following gradient was used: 2% B for 0-1 min, 2-100% B from 1 to 15 min, 100% B from 15 to 20 min and 2% B from 20 to 21 min with a flow rate of 1 mL/min at 30°C. Absorbance was measured at wavelengths of 210 nm, 254 nm, 260 nm, and 280 nm.
Chromatographic and mass spectral conditions. In total, 5 μL of sample were injected and analyzed with an Agilent 6520 QTOF equipped with a dual ESI ion source. Chromatographic separation was performed with a flow rate of 0.75 mL/min, temperature controlled at 30°C in a column oven compartment with a NUCLEODUR Sphinx RP column (150 mm × 3 mm ID, 5 μM, Macherey-Nagel) coupled with a UNIVERSAL RP guard column (4 mm × 3 mm ID, Macherey-Nagel). The gradient was as follows: 0-1.5 min, 10% B; 1.5-22.5 min, 100% B; 22.5-30 min, 100% B; 30-30.1 min, 10% B; and 30.1-37.5 min, 10% B. Mobile phase A and mobile phase B consisted of water with 0.1% formic acid and acetonitrile with 0.1% formic acid, respectively. Reference mass correction was performed with constant infusion of purine and HP-0921 with 10 μL/min through the secondary nebulizer needle. Mass spectral analysis was performed in full-scan mode and in MS² mode. MS-Source parameters were as follows: Gas temperature, 250°C; drying gas flow, 12 L/min; nebulizer pressure, 30 psi; capillary voltage, 3000 V, fragmentor voltage, 150 V; skimmer voltage, 65 V; octopole voltage, 750 V. Collision energy was set to 36 V, yielding overall best fragmentation efficacy. Additional measurements of intermediates of enzyme assays were conducted with a Waters ACQUITY I-class UPLC in combination with a Waters HSS T3 (C18) column (2.1 mm × 100 mm, 1.8 μm particle size) coupled to a Waters Acquity photo diode array detector (Waters) and a Waters Synapt G2-Si HDMS electrospray ionization (ESI)/quadrupole time-of-flight (Q-TOF) system. The following software was used for MS and HPLC data analysis: Agilent MassHunter Workstation Acquisition B.04.00, Agilent OpenLAB ChemStation Edition Rev. C.01.07, Agilent MassHunter Workstation Qualitative Analysis B.08.01, Agilent OpenLAB ChemStation Edition Rev. C.01.07, and Waters MassLynx V4.1.
Flash chromatography. Silica gel chromatography was performed using silica gel from Macherey-Nagel (60, particle size 0.040-0.063 mm). The solvents for the mobile phase were chosen as described in the according sections.
IR spectroscopy and density functional theory calculations. Solid state IR spectra of compounds 4 and 12 were recorded on a Tensor 27 FTIR spectrometer (Bruker) equipped with an attenuated total reflection (ATR) cell (Pike instruments) at a resolution of 4 cm −1 . To facilitate the assignment of characteristic vibrational modes, 11 conformers of 4 and 27 conformers of 12 were modeled using the conformer search algorithm in SPARTAN (Wavefunction, Inc.). To identify the most abundant conformers for each compound, geometry optimization, energy calculation, and frequency calculation on all unique conformers were performed at the DFT level (B3LYP/6-31 G(d)) in Gaussian 16 Rev. C.01 (Gaussian, Inc.). A conformer was identified with 99% abundance for 4 and a conformer with 97% abundance for compound 12 (calculated with respect to ΔG) that were used for comparison to the corresponding experimental spectra. IR bands were constructed with Lorentzian band shapes around each calculated intensity with a band width of 6 cm −1 . Calculated frequencies were uniformly scaled by 0.97. Characteristic bands were assigned through visualization of vibrational modes in GaussView 6 (Gaussian Inc.). NMR spectroscopy. 1 H-and 13 C-NMR spectra of compounds 9 and 11 were recorded with an Avance III HD 400 instrument (Bruker) ( 1 H: 400 MHz, 13 C: 100 MHz). All other 1 H-and 13 C-NMR spectra were recorded with a FT-NMR DMX 600 instrument (Bruker) ( 1 H: 600 MHz, 13 C: 150 MHz). DMSO, CDCl 3 and 1,4-dioxane-d 8 were used as solvents. Chemical shifts refer to the δ-scale and are reported in ppm with a solvent resonance as an internal standard (CHCl 3 : δ H 7.26; CDCl 3 : δ C 77.16, DMSO: δ H 2.50, δ C 39.51, 1,4-dioxane-d 8 : δ H 3.53). Data was analyzed and processed using the Bruker TopSpin 4.0.7 software.