Traveling two diverging roads, cytochrome-P450 catalyzed demethylation and γ-lactone formation in bacterial gibberellin biosynthesis

Abstract

Biosynthesis of the gibberellin plant hormones evolved independently in plants and microbes, but the pathways proceed via similar transformations. The combined demethylation and γ-lactone ring forming transformation is of significant mechanistic interest, yet remains opaque. The relevant CYP112 from bacteria was probed via activity assays and ¹⁸O₂ labeling experiments. Notably, the ability of tert-butyl hydroperoxide to drive this transformation indicates use of the ferryl-oxo (Compound I) from the CYP catalytic cycle for this reaction. Together with the confirmed loss of C-20 as CO₂, this necessitates two catalytic cycles for carbon-carbon bond scission and γ-lactone formation. The ability of CYP112 to hydroxylate the δ-lactone form of GA₁₅ shown by the labeling studies is consistent with the implied use of a further oxygenated heterocycle in the final conversion of GA₂₄ to GA₉, with the partial labeling of GA₉ demonstrating that CYP112 partitions its reactants between two diverging mechanisms.

Graphical Abstract

Gibberellins are important phytohormones, requiring complex biosynthetic processes that independently evolved in bacteria and fungi as well as plants. The requisite coupled demethylation and γ-lactone ring formation, catalyzed by the bacterial cytochrome P450 CYP112, was mechanistically probed, providing insight into this complex transformation.

Gibberellins (GAs) are essential hormones in plants, which has led some plant-associated fungi and bacteria to produce GA to manipulate their host plants. The biosynthetic steps of the individual pathways are essentially identical in plants and bacteria, and differ from the fungal pathway only in the order of occurrence of one peripheral (C-3β) hydroxylation step. Yet on the enzyme level it is evident from the low sequence identity that GA biosynthesis evolved separately in plants, fungi and bacteria.^[1] One hallmark transformation in GA biosynthesis is the conversion of 20-carbon GAs to the 19-carbon GAs, involving the combined loss of a methyl group and formation of an intramolecular γ-lactone bridge. In plants this complex multi-step reaction is catalyzed by iron/α-ketoglutarate-dependent dioxygenases termed GA 20-oxidase (GA20ox), while it is catalyzed by cytochrome P450 (CYP) monooxygenases in fungi (CYP68B) and bacteria (CYP112).^[1a] Although fungi and bacteria both use CYPs, these fall into different families that share less than 15% sequence identity. Nonetheless the reactions catalyzed by the three individual enzymes appear to be identical, and involve the stepwise oxidation of C-20 from a methyl group in GA₁₂ in plants and bacteria, and GA₁₂ as well as the 3β-hydroxy derivative GA₁₄ in fungi, to the corresponding alcohols GA₁₅ or GA₃₇ and further oxidation to the aldehyde function, yielding GA₂₄ or GA₃₆, respectively. The final step is combined loss of the oxidized methyl group and formation of a γ-lactone bridge between carbon C-19 and C-10, producing GA₉ or GA₄, respectively (Scheme 1). While the fungal enzyme does not react with the intermediates GA₁₅/GA₃₇ and GA₂₄/GA₃₆, the bacterial and plant enzymes can use GA₁₅ and GA₂₄ as substrates to produce GA₉.^{[1a, 1b]} Plants and fungi also produce the C-20 carboxylic acids GA₂₅ or GA₁₃ as side products, since neither the plant GA20ox nor the fungal CYP68B further convert GA₂₅/GA₁₃ to GA₉/GA₄ (Scheme 1).^[2] For the fungal CYP68B very little is known about the catalytic mechanism, only that C-20 appears to be lost as CO₂.^[3] With one plant GA20ox catalyzing this multi-step reaction some artificial substrates have been tested, and it has been demonstrated that C-20 also is lost as CO₂.^[4] The combination of the apparently straightforward first two reactions with the complex last reaction, and the known differences between plants and fungi, makes the catalytic mechanism of CYP112 of significant interest, which was investigated here.

Scheme 1.

Reactions catalyzed by CYP112, CYP68B and GA20ox in GA biosynthesis from bacteria, fungi and plants, respectively. Sequential oxidation and elimination of C-20, while the plant and fungal enzyme release CO₂ upon elimination of C-20, this was unknown for the bacterial CYP112.

Previously, isoforms of CYP112 were characterized by addition of GA₁₂ to recombinant Escherichia coli cultures.^{[1b, 1d]} This however did not permit detailed investigation of the catalytic mechanism. Here the isoform from Erwinia tracheiphila (EtCYP112), chosen as it is in the same Enterobacteriaceae family as E. coli, was His-tagged, purified and found to exhibit full activity in a reconstituted system with purified ferredoxin reductase from E. tracheiphila (EtFdR) and spinach ferredoxin (Fd), converting GA₁₂ to GA₉ (Figure 1). Under optimal conditions (i.e., with a molar excess of NADPH), no intermediates were detected, even when turnover was not yet complete, indicating their retention in the active site. It was possible to observe the expected C-20 hydroxylated intermediate GA₁₅ and aldehyde intermediate GA₂₄ under NADPH-limited conditions. These assays also produced the C-20 carboxylate derivative GA₂₅ (Figures 1 and S1). However, while GA₁₅ and GA₂₄ were readily converted to GA₉, GA₂₅ is not (Figure S2).

Figure 1.

In vitro enzyme activity of EtCYP112. GC-MS chromatograms of assays with purified EtCYP112, spinach Fd and EtFdR under limiting NADPH concentrations (A), or NADPH in excess stopped after 30 min (B), with GA12 as substrate.

The form in which C-20 is released by EtCYP112 was examined by ¹⁸O₂ labeling, using the reconstituted enzymatic system with an excess of NADPH under an ¹⁸O₂ atmosphere, with GC-MS analysis of the head-space. In the absence of substrate only unlabeled CO₂ was found, while addition of GA₁₂ resulted in production of CO₂ with one or two ¹⁸O labels (Figure 2). The observed loss of label presumably originates from the interconversion of CO₂ with bicarbonate during the incubation period of the enzyme assay.^[5] Thus, EtCYP112 clearly releases C-20 as CO₂. In addition, the resulting GA₉ product was extracted, methylated and analyzed by GC-MS. Surprisingly, 44% had incorporated one ¹⁸O (Figures 3 and S2, and Table S1). This was found not only in the molecular ion, but also the fragments at m/z = 298/299 and 270, which retain the γ-lactone ring,^[6] wherein the ¹⁸O label must have been inserted.

Figure 2.

Figure CO₂ loss during γ-lactone formation. (A) GC-MS chromatograms of enzyme assays with purified EtCYP112, spinach Fd, EtFdR, NADPH in excess and either no substrate or GA₁₂ as the substrates under an ¹⁸O₂ atmosphere. (B) MS-spectra of the CO₂ peak.

Figure 3.

In vitro enzyme activity of EtCYP112 under an ¹⁸O₂ atmosphere. (A) GC-MS chromatograms of enzyme assays with purified EtCYP112, spinach Fd and EtFdR with an excess of NADPH and GA₁₂ or under limiting NADPH concentrations with GA₁₂, GA₁₅ (closed lactone) or GA₂₄ as the substrate under an ¹⁸O₂ atmosphere. (B) MS-spectra of GA₉ showing the molecular ion at m/z 330 or 332 and two larger fragments corresponding to the loss of the methylated acid at C-7.

¹⁸O incorporation during the transformation of GA₁₂ to GA₉ was further probed with limiting NADPH, enabling examination of the intermediates GA₁₅ and GA₂₄. While this reduced the amount of ¹⁸O incorporated into GA₉ (to 14%), a substantial proportion of the observed GA₂₄ accumulated two ¹⁸O (37%). The remainder of the GA₂₄ contained a single ¹⁸O. GA₁₅ is observed in the closed δ-lactone form, and was essentially fully labeled with a single ¹⁸O. This demonstrated that GA₁₂ is converted to GA₁₅ via hydroxylation rather than direct formation of the δ-lactone ring. Analogous labeling experiments were carried out with GA₁₅ (open lactone), GA₁₅ (closed δ-lactone) or GA₂₄ as substrate (Figure 3 and Table S1). GA₉ produced from these intermediates did not contain ¹⁸O. However, GA₂₄ produced from the closed (δ-lactone) form of GA₁₅ was fully labeled with ¹⁸O. Thus, EtCYP112 readily hydroxylates the δ-lactone form of GA₁₅ to the lactol form of GA₂₄. Given that GA₂₄ is observed in its aldehyde form,^[7] the incorporated ¹⁸O is retained by direct opening of the lactol. Starting from the open form of GA₁₅, 67% of GA₂₄ was labeled with ¹⁸O, indicating that EtCYP112 produces a geminal-diol, which upon dehydration loses ~50% of the label and forms the observed GA₂₄ aldehyde. The greater amount of labeled species observed here originates from both a slight bias towards removal of the unlabeled hydroxyl-group,^[8] but presumably largely from partial closure of the open GA₁₅ to its δ-lactone form in the assay buffer before conversion by EtCYP112.

To further investigate formation of the lactol form of GA₂₄, the C-7 and C-19 carboxylate groups were methylated in GA₁₂, GA₁₅, and GA₂₄. This was expected to block lactone formation in GA₁₅ and GA₉. However, C-19 methylation seems to be unstable in GA₁₅, and is quickly lost by conversion to the δ-lactone upon purification in organic solvents. The instability of the C-19 methyl ester in GA₁₅ is also reflected in assays with the enzymatic system. In particular, MeGA₁₂ was efficiently converted into GA₉ and, under an ¹⁸O₂ atmosphere, all GA₉ product contains one ¹⁸O label. MeGA₁₅ (closed δ-lactone) also is transformed into GA₉, but does not incorporate any ¹⁸O label, while MeGA₂₄ is not acted upon by EtCYP112 (Figures 4 and S3, and Table S1).

Figure 4.

In vitro enzyme activity of EtCYP112 with MeGA₁₂ and MeGA₁₅ under an ¹⁸O₂ atmosphere. (A) GC-MS chromatograms of enzyme assays with purified EtCYP112, spinach Fd and EtFdR under limiting NADPH concentrations with MeGA₁₂ or MeGA₁₅ (closed lactone) as the substrate under an ¹⁸O₂ atmosphere. (B) MS-spectra of GA₉ showing the molecular ion at m/z 330 or 332 and two larger fragments corresponding to the loss of the methylated acid at C-7.

It has been suggested that GA₂₄ can be converted into GA₉ in a single catalytic cycle via the atypical use of a ferric-peroxo activated group,^[2a] and similar carbon-carbon bond cleavage reactions catalyzed by CYP1A2 and CYP17A1 have been suggested to utilize superoxo or peroxo states of the heme-iron.^[9] In CYPs these early stages of the catalytic cycle can be bypassed by using H₂O₂ or tert-butyl hydroperoxide (tBuOOH) as the reducing agent instead of molecular oxygen and electrons from NADPH (Scheme S2).^[10] However, while the overall activity of EtCYP112 with H₂O₂ is significantly lower, limiting conversion of GA₁₂ to GA₁₅ and GA₂₄, it is able drive the conversion of GA₁₅ to GA₂₄ and GA₉, as well as GA₂₄ to GA₉. tBuOOH also can drive conversion of GA₂₄ to GA₉ (Figure S4).

The incorporation of ¹⁸O into a significant proportion of the GA₉ γ-lactone ring indicates that one of the oxygens incorporated by EtCYP112 is transferred to the C-19 carboxylate. This must stem from the initial hydroxylation of GA₁₂ to the open form of GA₁₅ that then undergoes dehydration to the δ-lactone ring form.^[7b] Indeed, the essentially complete labeling of the GA₁₅ produced from GA₁₂ demonstrates that this reaction invariably involves hydroxylation. The ability of EtCYP112 to catalyze hydroxylation of the δ-lactone form of GA₁₅ to the lactol form of GA₂₄ is demonstrated by both the complete ¹⁸O labeling of the GA₂₄ produced from the closed (lactone) form of GA₁₅, and the significant double-labeling of the GA₂₄ produced from GA₁₂. EtCYP112 can also hydroxylate the open form of GA₁₅ to the geminal-diol form of GA₂₄, as demonstrated by the partial labeling of the resulting GA₂₄.

The most interesting transformation is however the conversion of GA₂₄ into GA₉, requiring both carbon-carbon bond scission and γ-lactone ring formation. While MacMillan has proposed several reaction mechanisms,^[2a] these were focused on the plant dioxygenases. Nevertheless, CYP monooxygenases also use an iron co-factor to bind molecular oxygen, and both of these oxygenases generally utilize activated ferryl-oxo complexes (Compound I) for catalysis.^[10–11] Thus, the proposed mechanisms are adaptable to CYPs (Schemes S3 and S4). In addition, CYPs are known to catalyze reactions via ferryl-hydroxo (Compound II), ferric-superoxo, ferric-peroxo,^[9] or ferric-hydroperoxo species (Compound 0) as well (Scheme S2).^[12] However, the ability of EtCYP112 to use H₂O₂ or tBuOOH to drive this transformation precludes the use of either the superoxo or peroxo complexes, or Compound 0, respectively. Thus, EtCYP112 uses Compound I in the transformation of GA₂₄ to GA₉, which is consistent with the implied use of Compound I for the hydroxylation reactions shown here for the two preceding transformations, as well as the mechanism proposed for the similar reaction catalyzed by CYP19A1.^[13] However, release of C-20 as CO₂ rules out transformation of GA₂₄ to GA₉ via a single oxidation cycle, even with consideration of alternative reactivity with Compound II, despite the inability of the C-20 carboxylate GA₂₅ to serve as a substrate. It has been suggested that the plant GA20ox also uses two oxidation cycles, but proceeds via a covalent intermediate.^[4b] However, there is no evidence for such a mechanism with EtCYP112. On the other hand, the intermediacy of a C-20 geminal-diol cyclic anhydride equivalent may explain why the open C-20 carboxylate GA₂₅ is not a substrate. Consistent with a requirement for a cyclic anhydride intermediate is the inability of EtCYP112 to use MeGA₂₄ as a substrate, as it cannot form such a heterocycle.

The results presented here indicate that, similar to CYP170A1 in albaflavenone biosynthesis,^[14] EtCYP112 utilizes two diverging routes in transformation of GA₁₂ to GA₉. This is demonstrated by the partial labeling of the GA₉ produced from GA₁₂, and contrast with the complete labeling observed with MeGA₁₂, which is biased towards formation of the closed (δ-lactone) form of GA₁₅. Accordingly, rather than choose, EtCYP112 partitions the travel of its reactants between two mechanistic roads that diverge at GA₁₅ (Scheme 2). Initial hydroxylation to the open form of GA₁₅ can be followed by dehydration to the δ-lactone form, with subsequent hydroxylation of either producing the gem-diol or lactol forms of GA₂₄, respectively (Scheme S4). These distinct intermediates can then undergo separate conversion to GA₉ via two sequential oxidation reactions. This presumably involves formation of the C-20 gem-diol cyclic anhydride equivalent in both cases, reflecting either direct hydroxylation of δ-lactol, or cyclization of gem-diol GA₂₄ mediated by the ability of Compound II to abstract an additional hydrogen. The resulting anhydride gem-diol then undergoes coupled carbon-carbon bond scission and γ-lactone ring formation, with release of CO₂, again relying on abstraction of another hydrogen by Compound II. Altogether, despite their independent evolution, plants, fungi and bacteria catalyze the same series of reactions in the coupled demethylation and γ-lactone ring formation transformation step of gibberellin biosynthesis.^[3–4] The C-20 carboxylate equivalent GA₂₅ does not serve as a productive intermediate for any of the relevant oxygenases, and the C-20 gem-diol cyclic anhydride equivalent indicated here for the bacterial CYP112 has also been suggested for plants.^[2a] Thus, despite their independent evolutionary origins, these oxygenases seem to have converged on similar mechanistic routes, perhaps reflecting the chemical constraints of this complex transformation as well as biochemical constraints of the utilized iron-dependent oxygenases.

Scheme 2.

Proposed bifurcation and reconversion of pathways from GA₁₂ to GA₉ catalyzed by EtCYP112.