Unified biogenesis of ambiguine, fischerindole, hapalindole and welwitindolinone: Identification of a monogeranylated indolenine as a cryptic common biosynthetic intermediate by an unusual magnesium-dependent aromatic prenyltransferase

Abstract

Biochemical characterizations of aromatic prenyltransferase AmbP1 and its close homologs WelP1/FidP1 in hapalindole-type alkaloid biosynthetic pathways are reported. These enzymes mediate the magnesium-dependent selective formation of 3-geranyl 3-isocyanovinyl indolenine (2) from cis-indolyl vinyl isonitrile and GPP. The role of magnesium cofactor in AmbP1/WelP1/FidP1 catalysis is highly unusual for a microbial aromatic prenyltransferase, as it not only facilitates the formation of 2 but also prevents its rearrangement to an isomeric 2-geranyl 3-isocyanovinyl indole (3). The discovery of 2 as a cryptically conserved common biosynthetic intermediate to all hapalindole-type alkaloids, suggests an enzyme-mediated Cope rearrangment and aza-Prins-type cyclization cascade is required to transform 2 to a polycyclic hapalindole-like scaffold.

Hapalindole-type alkaloids are a large group of indole monoterpenoids with broad-spectrum antimicrobial and antitumor activities that are exclusively produced by stigonematalean cyanobacteria.¹ Their structural diversity and complexity constitutes a daunting puzzle that has provoked many intriguing biosynthetic proposals in the past three decades,² whereas the underlying genetic, biochemical and molecular basis remained unknown until very recently.³

By strategic identification and comparative analysis of ambiguine (amb),^3a welwitindolinone (wel)^3b and fischerindole (fid)^3c biosynthetic gene clusters in three distinct stigonematalean cyanobacteria that are able to generate all major structural phenotypes of hapalindole-type alkaloids, we recently provided compelling genetic and biochemical evidence that cis-indolyl vinyl isonitrile (Z)-1 and geranyl pyrophosphate (GPP) are the conserved common biosynthetic precursors to this family of natural products (Fig. 1).^3a–c We also demonstrated that the chlorine substitution in hapalindole-type alkaloids is introduced by a new family of 2-oxoglutarate-dependent nonheme iron halogenase via late stage aliphatic C-H bond functionalization in the context of welwitindolinone biogenesis.^3d

Figure 1.

Conserved intermediates identified in the early-stage biosynthesis of hapalindole-type alkaloids.

With the initial disclosure of amb gene cluster,^3a we proposed the tri- or tetracyclic scaffolds in hapalindoles and fischerindoles may arise from a single step enzymatic unification between (Z)-1 and GPP by aromatic prenyltransferase AmbP1 that could act as a terpene cyclase.^{3a, 4} This enzymatic transformation would selectively define the stereochemistry of four consecutive carbon centers (C15, C10, C11, C12). This proposal implicates the stereochemical diversity across these four carbon centers in haplindole-type alkaloids from different producers would be due to the alterable stereochemical configuration of their precursors (i.e. (Z)-1 vs. (E)-1 and GPP vs. neryl pyrophosphate (NPP)). However, this proposal was readily disfavored after we identified the wel and fid gene clusters,^3b–c which encode the identical set of genes for (Z)-1 and GPP assembly and do not have the capacity to biosynthesize (E)-1 and NPP. These observations led us to reformulate the role of AmbP1 and its close homologs WelP1 and FidP1 as a catalyst to transform (Z)-1 and GPP to a cryptic common intermediate,^3b which can be subsequently enzymatically tailored to give structurally diverged hapalindole-type molecules (Fig. 1).

Here, we report the biochemical characterization of AmbP1 and its homologs WelP1/FidP1 and the structural elucidation of their enzymatic products from (Z)-1 and GPP. AmbP1 is shown to be an unusual aromatic prenyltransferase that can forward-transfer a geranyl group selectively to the C-3 carbon of (Z)-1 to give a monogeranylated indolenine 2. This enzymatic transformation strictly depends on the presence of Mg²⁺, highly unusual for microbial aromatic prenyltransferase.⁵ The identification of 2 as a cryptic intermediate, common to all hapalindole-type alkaloid biosynthetic pathways, suggests an enzyme-mediated sequential Cope rearrangement and aza-Prins-type cyclization cascade has to be in place for the stereoselective generation of tricyclic hapalindole scaffold as observed in 12-epi-hapalindole C.

To assess the biochemical function of AmbP1 and its homologs WelP1 and FidP1 from the wel and fid gene clusters (ESI, Fig. S1), they were overexpressed in E. coli and isolated as N-terminal His₇-tagged fusion proteins (see ESI). Upon incubation of each protein with (Z)-1 and GPP with MgCl₂ (5 mM) in Tris buffer (50 mM, pH=8.0), the rapid generation of two isomeric products 2 and 3 (both with m/z=305 [M+H]⁺) were observed by LC-MS analysis (Fig. 2a). The UV absorption spectra (190–350 nm) of products 2 and 3 do not match those of tri- or tetracylic hapalindoles or fischerindoles (ESI, Fig. S2). Instead, the UV absorption spectrum of 3 resembles that of (Z)-1, suggesting the presence of an intact indolyl vinyl isonitrile group, whereas the spectrum of 2 lacks distinct feature at 250–350nm, suggesting a de-aromatization of the indole ring (Fig. 2a). The enzymatic products derived from WelP1 and FidP1 are perfectly identical to that from AmbP1, confirming they are functionally identical that corroborates with the bioinformatic prediction as the sequences of WelP1 and FidP1 are 96% and 100% identical to AmbP1 (ESI, Fig. S1). In addition, (E)-1, the diastereomer of (Z)-1, is not a substrate for AmbP1, supporting our previous conclusion that (Z)-1 and GPP are the common biogenetic precursors to all hapalindole-type alkaloids,^3b,f and ruling out the involvement of (E)-1 in their biogenesis as suggested by others.^{2a, 3e}

Figure 2.

Initial characterization of AmbP1 and its homologs WelP1/FidP1 revealed Mg²⁺-dependent divergent enzymatic generation of monogeranylated indolenine 2 and indole 3 from (Z)-1 and GPP. a) Comparative HPLC chromatographs showing (Z)-1, but not (E)-1, is the substrate for AmbP1 and its homologs WelP1/FidP1, which can be converted to two products 2 and 3 in the presence of GPP with MgCl₂ in Tris buffer (pH 8.0). b)-c) HPLC chromatographs showing the time course of AmbP1-mediated formation of 2 or 3 as the major enzymatic product from (Z)-1 and GPP in Tris buffer (pH 8.0) in the presence or absence of MgCl₂. d) Selective formation of 2 or 3 from (Z)-1 and GPP by AmbP1 with MgCl₂ at pH 9.0 or EDTA at pH 6.0.

As a mixture of 2 and 3 were generated in our initial assay, we proceeded with optimization to improve the selectivity. AmbP1 belongs to the ABBA aromatic prenyltransferase superfamily, of which a hallmark feature is that their enzymatic activities do not depend on the presence of divalent metal cations, such as Mg^2+.5 However, an HPLC-based time course analysis of AmbP1 enzymatic activity at pH 8.0 clearly shows the presence of MgCl₂ (5 mM) facilitates the generation of 2 and accelerates the turnover of (Z)-1, in comparison with the assays without MgCl₂ (Fig. 3b–c). This unexpected observation suggests Mg²⁺ plays an important role in AmbP1 catalysis and prompted us to systematically examine the effect of Mg²⁺ concentration and pH on the outcome of this enzymatic transformation. We found with increased Mg²⁺ concentration (up to 20 mM) and pH (up to 9.0), 2 was preferentially formed (ESI, Fig. S3a–b), whereas the exclusion of Mg²⁺ and other divalent metal ion by the addition of EDTA (10 mM) at a decreased pH (up to 6.0) facilitates the formation of 3 (ESI, Fig. S3b–c), albeit slows down the enzymatic turnover of (Z)-1. With the optimized conditions for selective formation of 2 and 3 identified (Fig. 2d), the isolation of sub-mg of 2 and 3 were achieved by scaling up the enzymatic reaction. A combination of 1-D ¹H and 2-D COSY, HSQC and HMBC NMR analyses (ESI, Fig. S4–11) allowed for the unambiguous structural assignment of 2 as 3-geranyl 3-isocyanovinyl indolenine (ESI, Table S1) and 3 as 2-geranyl 3-isocyanovinyl indole (ESI, Table S2). With the procurement of pure 2 and 3 as analytical standards, we were able to deduce the steady state kinetic parameters for AmbP-1 mediated enzymatic conversation of (Z)-1 to 2 and 3. At pH 8.0 with MgCl₂ (10 mM), K_m for (Z)-1 is 15.4±3.2 μM with a k_cat of 6.9±1.2 s⁻¹. The catalytic efficiency is in accordance with those of well-characterized microbial ABBA prenyltransferases.⁶ At pH 6.5 with EDTA (10 mM), the k_cat of AmbP1 drops nearly 8-fold to 0.82±0.19 s⁻¹, corroborating with our initial AmP1 time course analysis at different pH with or without Mg²⁺ (Fig. 2, ESI Fig. S3). Screening of other divalent metal ions (Ca²⁺, Zn²⁺, Mn²⁺) that can serve as alternatives for Mg²⁺ in ABBA prenyltransferases^5b,c readily revealed Mg²⁺ cannot be replaced, as all of them either reduced or abolished the enzymatic activity of AmbP1 and are ineffective in mediating the selective formation of 2 from (Z)-1 (Fig. 3a). As 3,3-di-alkyl substituted indolenines are well known to undergo facile 1,2-alkyl shift to generate thermodynamically more stable 2,3-disubstitued indole under acidic conditions,⁷ we hypothesize that 3 is the shunt product from 2. At pH 6.5, the spontaneous rearrangement of 2 to 3 occurs with an estimated t_1/2 of 6 h (ESI, Fig. S12). Intriguingly, the addition of AmbP1 at pH 6.5 in the presence of EDTA (10 mM) accelerates the rearrangement of 2 to 3 by at least 10-fold (Fig. 3b). This accelerated rearrangement however, was effectively suppressed to the basal level by the addition of MgCl₂ (20 mM) (Fig. 3c). This observation implicates Mg²⁺ likely interacts with the indole-containing substrate in the AmbP1 active site and the Mg²⁺-dependency of AmbP1 is mechanistically distinct from other Mg²⁺-dependent prenyltransferases where the role of Mg²⁺ is for activating pyrophosphate as a Lewis acid.⁸

Figure 3.

Conversion of (Z)-1 and GPP to 2 is the preferred enzymatic outcome of AmbP1 that depends on the presence of Mg²⁺. 1) Comparative HPLC chromatographs showing Mg²⁺ is the only divalent metal ion cofactor that allows for the selective formation of 2 by AmbP1. 2) HPLC chromatographs showing the time course of conversion of 2 to 3 is facilitated by the presence of AmbP1 at pH 6.5 with EDTA (10 mM). 3) HPLC chromatographs showing the time course of AmbP1-mediated conversion of 2 to 3 at pH 6.5 is readily suppressed by the presence of MgCl₂ (20 mM). 4) NMR analysis of AmbP1 provides biophysical evidence on its interaction with Mg²⁺. Superposition of the 2D ¹H-¹⁵N HSQC spectra of AmbP1 in the absence (black) and presence of Mg²⁺ (green) at ~1:10 molar ratio of protein to Mg²⁺. Selected resonances that exhibit noticeable chemical shift changes are encircled in magenta ovals. Note that residues marked with magenta asterisks also experience chemical shift perturbations; however, the bound resonances are too broad for detection.

Collectively, the functional role of Mg²⁺ in AmbP1 catalysis can be summarized as two-fold: facilitating the selective formation of 2 and suppressing its rearrangement to 3. To gain direct insight on the interaction of Mg²⁺ with AmbP1 protein, we carried out a NMR titration experiment. Chemical shift mapping of interactions by NMR titration experiments is a sensitive and powerful approach to identify binding sites in a protein at the residue level.⁹ Upon complex formation, the ligand modifies the contact surface by either inducing a conformational change and/or altering the electronic environment of the nuclei that reside at the interface in the complex. We used ¹⁵N-labeled AmbP1 protein and recorded changes in the 2D ¹H-¹⁵N HSQC spectrum in the absence (free) and in the presence of Mg²⁺ (Fig. 3d). A superposition of the free protein spectrum and the one in the presence of a 10 fold molar excess of Mg²⁺ identifies resonances affected by Mg²⁺ binding that are enclosed in magenta ovals or labelled with asterisks (Fig. 3d). Several resonances experience changes, either being shifted or broadened, clearly confirming that AmbP1 interacts with Mg²⁺.

Having disclosed AmbP1 as an unusual Mg²⁺-dependent aromatic prenyltransferase that can selectively geranylate C-3 position of (Z)-1 to give indolenine 2, we further examined its substrate scopes for prenyl donors and acceptors (Fig. 4). In addition to GPP, dimethylallyl pyrophosphate (DMAPP) can be also processed by AmbP1 when combined with (Z)-1 (Fig. 4a). Two enzymatic products 2a and 3a were detected, of which UV absorption spectra shared the same features of 2 and 3 (ESI, Fig. S13), suggesting they are structural homologs. While 3a is selectively formed at pH 6.0 with EDTA (10 mM) (Fig. 4a, trace 1), the formation of 2a was not selective at pH 9.0 with MgCl₂ (20 mM) (Fig. 4a, trace 2), different from what was observed with GPP as a substrate, suggesting Mg²⁺-dependent enzymatic property of AmbP1 is GPP-specific. Corroborating with this observation, when DMAPP and GPP were used in 1:1 ratio with (Z)-1, GPP is selectively processed to either give 1 or 2 (Fig. 4a, traces 3–4), with only trace amount of 2a and 3a generated. We also tested NPP (the diastereomer of GPP) and farnesyl pyrophosphate (FPP) as substrates for AmbP1, but they were not processed when combined with (Z)-1 (ESI, Fig. S14). These observations provide convincing evidence that GPP is the authentic natural substrate for AmbP1.

Figure 4.

Evaluation of alternative pyrophosphate and indole substrates confirms AmbP1 is a C-3 indole prenyltransferase and prefers GPP as the prenyl donor. a) Comparative HPLC chromatographs showing AmbP1 can accept DMAPP as a prenyl donor, but prefers GPP. b) Comparative HPLC chromatographs showing AmbP1 can also process 4 and 5 with GPP to give 6.

Alternative prenyl acceptors, including indole 4 and indole-3-carboxaldehyde 5, were also processed by AmbP1 when combined with GPP. In both cases, 3-geranylated indole 6 was isolated as the sole enzymatic products (Fig. 4b) (ESI, Fig. S15–18 and Table S3), either at pH 6.0 with EDTA (10 mM) or pH 9.0 with MgCl₂ (20 mM), providing additional evidence that the C-3 of indole serves as the default nucleophile for GPP with AmbP1. The selective generation of 6 from 5 by AmbP1 implicates an enzymatic sequence of C3-geranylation, followed by C3-deformylation, is likely operant. Interestingly, the enzymatic turnover of 4 or 5 to 6 appeared faster in the absence of Mg²⁺ (Fig. 4b, traces 1/3 versus 2/4), highlighting the unusual Mg²⁺–dependent feature of AmbP1 is highly specific to its native substrates (Z)-1 and GPP.

The discovery of 2 as the kinetically preferred enzymatic product of AmbP1 and its homologs WelP1/FidP1 from (Z)-1 and GPP, provides strong evidence that it is the conserved biosynthetic intermediates for all hapalindole-type alkaloids and raises the question how it can be transformed into the polycyclic hapalindole-like scaffold. Indolenine 2 contains a 1,5-diene group embedded between the proximal C=C bond of geranyl group and that of vinyl isontrile, a natural motif for a pericyclic reaction (Fig. 5). This [3,3]-sigmatropic Cope rearrangement will lead to the diastereoselective formation of an alternate 1,5-diene 2a, where the distal C=C bond in the geranyl group can serve as a nucleophile to trap the newly generated conjugated imine via an aza-Prins-type cyclization to give 12-epi-hapalindole C (7), which can be further enzymatically diversified to other hapalindole-type scaffolds (Fig. 5).

Figure 5.

Proposed enzymatic conversion of 2 to 7 via sequential Cope rearrangement and aza-Prins-type cyclization as a conserved biosynthetic step for the late-stage divergent generation of hapalindole-type alkaloids.

Recently, Li and coworkers have utilized 2a-type intermediates in the stereoselective preparation of hapalindole-type molecules,¹⁰ where they demonstrated 2a-type molecules are able to undergo aza-Prins-type cyclization to give 7-like molecules. The resulting substituents at C10 and C15 are stereoselectively placed in a trans-configuration as the reaction proceeds through an all-equatorial chair-like transition state, supporting the hypothesis outlined in Fig. 5. The rearrangement of 2 to 2a can be potentially facilitated by AmbP1 itself, as fungal ABBA-type indole prenyltransferases have been proposed to use Cope rearrangement as an alternative pathway for selective prenylation.¹¹ We extensively examined conditions by combining 2 and AmbP1 with various exogenous factors, such metal ions, heats and lights, but none of these combinations led to the formation of 7 or related hapalindole-type molecules (data not shown), suggesting additional enzyme(s) will be required for this process. During the late-stage preparation of this manuscript, a separate report emerged that described the cell-lysate fraction of the amb producer Fischerella ambigua UTEX1903 can facilitate the conversion of 2 to 12-epi-hapalindole U, a constitutional isomer of 7 with identical stereochemistry at C10–12 and C15,¹² which corroborates with the proposal outlined in Fig. 5. A previously annotated functionally unknown AmbU4 protein was cited as the main effector for the observed enzymatic rearrangement.¹² As both amb and related wel and fid gene clusters encode multiple copies of ambU homologs with variable protein sequences (50–60% sequence identify),^3a–c it remains to be elucidated on the true functional roles of these enzymes in directing the structural diversities in each hapalindole-type alkaloid producer originated from a single intermediate 2, characterized in this report.

In summary, this work featured the discovery of 2 as a cryptically conserved biosynthetic intermediate for all hapalindole-type molecules, assembled by a highly unusual Mg²⁺-dependent aromatic prenyltransferase AmbP1 (and its homologs WelP1 and FidP1). This disclosure implies a rare enzyme-mediated Cope rearrangement and aza-Prins cyclization cascade may be required as a key enzymatic step to diverge 2 into structurally distinct tri- or tetracyclic hapalindole/fischerindole scaffolds, and bring additional excitement into the investigation of hapalindole-type alkaloid biogenesis, which has proven to be a treasure trove of new enzymatic transformations.¹³