Identification and Characterization of A Welwitindolinone Alkaloid Biosynthetic Gene Cluster in Stigonematalean Cyanobacterium Hapalosiphon welwitschii

Welwitindolinones (Figure 1A) are a unique family of indole monoterpene alkaloids that were originally isolated from true-branching heterocystous filamentous cyanobacterium Hapalosiphon welwitschii with a broad range of biological activities.^[1] In particular, (N-methyl)welwitindolinones B and C (1-3) are dual functional antitumor agents that exhibit antimitotic activity with multidrug resistance (MDR) reversal properties.^[2] Except welwitindolinone A (4), all welwitindolinones isolated to date are 3,4-disubstituted oxindoles with a signature bicyclo[4.3.1]decane core motif,^{[1, 3]} which has captivated considerable interest from the synthetic chemistry community worldwide that resulted in several total chemical syntheses of this family of molecules.^[4]

Figure 1.

Structures of welwitindolinones (A) and 12-epi-hapalindoles/ fischerindoles (B) isolated from H. welwitschii W. & G .S. West and UTEX B1830 (C-12 stereocenters are highlighted) and their distinctions from hapalindoles G/U (C), the backbone structural motifs in ambiguines (C-12 stereocenters are highlighted). (D) Biosynthetic proposals on the stereodivergent generations of 5a/c, 6a/c and 8 from (E)- or (Z)-9 with 10, formulated by Moore and others.^{[1, 4d, 6]}

Welwitindolinones are believed to be biosynthetically related to other hapalindole-type molecules as their native producers also generate hapalindoles and fischerindoles (Figure 1B),^{[1, 3]} The biogenesis of 4, a putative biosynthetic intermediate for bicyclo[4.3.1]decane-containing welwitindolinones, was initially proposed by Moore that involves a cationic cyclization of an oxidized 12-epi-hapalindole E derivative.^[1] More recently, Baran and coworkers demonstrated that 4 could be readily obtained by a XeF₂-mediated oxidative ring contraction of 12-epi-fischerindole I (7),^[4c] which is implicated as the direct biosynthetic precursor to 4. However, no genes or proteins associated with welwitindolinone biosynthesis have been identified and characterized to date.

We recently initiated a collective effort in understanding the genetic and molecular basis for the biosynthesis of hapalindole-type natural products and disclosed the first biosynthetic gene cluster for ambiguine (amb) isonitriles.^[5] Initial studies on the amb pathway allowed for the conclusive demonstration that the biosynthesis of ambiguines does not involve the previously proposed β-ocimene.^[6] Identification of the amb pathway also sheds light on the roles of Rieske-type oxygenases in the late-stage structural diversification of ambiguines and implies the involvement of AmbP1, an aromatic prenyltransferase in the generation of core hapalindole scaffold from geranyl pyrophosphate (GPP) and 3-((Z)-2′-isocyanoethenyl) indole (10, vide infra).

Distinct from ambiguines with a shared hapalindole G or U motif,^[6-7] the hapalindoles and fischerindoles co-isolated with welwitindolinones possess different polycyclic ring frameworks, but share otherwise identical terpenoid backbone stereochemistry with hapalindoles G/U (8) (Figure 1C), except for an inverted C-12 quaternary stereocenter. This stereochemical discrepency at C-12 was suggested to be derived from a pair of diastereomeric monoterpene precursors, namely (E) and (Z)-β-ocimenes (9), according to the original biosynthetic proposal formulated by Moore and others (Figure 1D).^{[1, 4d, 6]} Since the study of the amb pathway has ruled out the involvement of (Z)-9 in the biogenesis of hapalindoles G/U, questions remain open on what is nature’s underlying principle to generate these structural diversities. To this end, we resorted to the identification of a biosynthetic gene cluster for welwitindolinones and related hapalindoles/fischerindoles to gain additional insights on the biogenesis of hapalindole-type natural products.

We chose H. welwitschii UTEX B1830, a xenic strain to investigate the biosynthesis of welwitindolinones as it is readily available in the public domain and has been reported to produce identical hapalindole-type molecules as H. welwitschii W. & G .S. West.^[1] We initially examined the metabolite profiles of H. welwitschii UTEX B1830 by combining HPLC with UV-spectral fingerprints and high resolution mass spectral (HRMS) analyses (Figure SI-1B) and ensured it generated the structural diversities as previously reported.^[1] We then extracted the genomic DNA of H. welwitschii UTEX B1830 and subjected it to de novo genome sequencing using a Roche 454 GS FLX+ system (SI Methods). The draft assembly of total reads resulted in nearly 10,000 contigs that total 15 Mbp, confirming the xenic status of H. welwitschii UTEX B1830. Using this pseudometagenomic data, we carried out nucleotide BLAST using genes in the amb pathway as bioinformatic leads. This effort led to the identification of 11 contigs, including a single 21-kbp contig that resembles the genetic sequence from ambC1 to ambC3 in the amb gene cluster, with the remaining contigs having an average size of 1-2 kbp and lacking homologous end-joining sequences. Subsequent gap repairings relied extensively on Sanger sequencing of carefully designed cross-contig PCR amplicons (SI Methods) in order to bypass highly sequence-repetitive regions (Figure SI-2) to successfully map out the sequence and directionality of the entire welwitindolinone (wel) biosynthetic gene cluster that spans 36 kbp long (Figure 2).

Figure 2.

Illustration of the welwitindolinone (wel) biosynthetic gene cluster from H. welwitschii UTEX B1830 and its comparison with the ambiguine (amb) biosynthetic gene cluster from F. ambigua UTEX1903. Wel gene functions are grouped based on their putative roles associated with welwitindolinone and 12-epi-hapalindole/fischerindole biosynthesis and their coded protein sequences were compared with those in the amb pathway for similarities (highlighted with symbols Δ/*/#). Full annotation of wel ORFs can be found in the supporting information (Table SI-1)

Functional annotation of 30 protein-coding open reading frames (ORFs) in the wel gene cluster revealed striking similarity to those in the amb pathway (Figure 2 & Table SI-1), providing an initial glimpse on two highly related biosynthetic machineries for the assembly of welwitindolinones, ambiguines and related hapalindoles. The presence of transposable elements (welS1/S2) in the wel cluster, similar to orf2/3 at the boundary of the amb cluster, highlights the mobile nature of these pathways, suggesting horizontal gene transfer (HGT) may be responsible for the wide occurrence of hapalindole-producing cyanobacteria in the Stigonematalean family. Except transposase-coding welS1-2 and response regulator-coding welR3, proteins encoded by 19 ORFs that span 27 kbp long from welD4 to welC1 share extremely high sequence identities (90-99%) to those in the amb gene cluster, implicating they are likely functionally identical to their amb homologues in the context of regulating and assembling key biosynthetic intermediates for welwitindolinone and ambiguine biogenesis (Figure 3A). To correlate with the bioinformatic predictions, we overexpressed WelP1 and WelP2 in E. coli, and demonstrated that WelP2 is a dedicated geranyl diphosphate synthase to provide GPP (SI Methods), whereas both WelP1 and WelP2 lack the ability to convert GPP to β-ocimene, identical to what was observed for AmbP1 and AmbP2 in ambiguine biosynthesis.^[5] To validate the functions of welI1-3, we cloned them as a single operon under a pBAD promoter (SI Methods) and examined its biosynthetic product in vivo. Overexpression of welI1-3 and ambI1-3 in E. coli both led to the robust production of 10 that matched the synthetic standard (Figure 3B) by HPLC analysis with no E-isomer of 10 observed. These experiments collectively demonstrated that welwitindolinone and ambiguine biosynthesis adopt identical pathways for assembling early biosynthetic intermediates GPP and 10.

Figure 3.

Prediction and characterization of wel genes involved in the assembly of intermediates GPP and 10 for welwitindolinone biosynthesis. (A) Predicted functions of WelD1-4, WelT1-5, WelI1-3 and WelP2. (B) In vivo characterization of WelI1-3 enzymatic product in E. coli and its comparison with that derived from AmbI1-3.

Upstream of welD4 in the wel cluster there are eight ORFs that show more notable differences, compared to those embedded in the amb pathway (Figure 2). In particular, welM is unique to the wel gene cluster and predicted to encode a SAM-dependent methyltransferase, suggesting it is likely responsible for the generation of N-methylwelwitindolinones (1b, 3b) from their non-methylated precursors (1a, 3a) (Figure 4A). To validate the function of WelM, we overexpressed and purified it from E. coli (SI Methods) and isolated its putative substrate 3a (Figure SI-3) from H. welwitschii UTEX B1830. Incubation of 3a with recombinant WelM and S-adenosylmethionine (SAM) rapidly generated a new product, of which the retention time over a C18 HPLC column matches that of authentic 3b (Figure 4B). Thorough characterizations of the enzymatic product derived from 3a and WelM by 1D/2D NMR and HRMS analysis confirmed its structural identity to be 3b (Figures SI-4/5/6). The kinetic profiles of WelM (K_m=2.43±0.18μM and k_cat=2.17±0.12s⁻¹ for 3a; SI Methods) are in line with other recently characterized amide N-methyltransferases,^[8] providing additional evidence that 3a is a natural substrate for WelM. We also examined the substrate promiscuity of WelM, motivated by the observations that several methyltransferases in small molecular natural product biosynthesis possess broad substrate tolerance. ^[9] Among a broad range of substrates tested, including non-substituted oxindole, 3-methyl oxindole and several readily isolatable hapalindole-type molecules (Figure SI-7A), only 4 that contains an oxindole backbone appended with a spirocyclobutane monoterpene unit turned out to be an acceptable substrate for WelM (Figure SI-7B). However, 4 is a very poor substrate with an apparent k_cat ca. 1,000-fold less than that for 3a. Overall, these experiments demonstrate that WelM is a highly efficient narrow-substrate N-methyltransferase that was likely evolved to specifically recognize 3,4-disubstituted oxindoles with a bicyclo[4.3.1]decane motif. This dedicated N-methylation is likely the final step for the welwitindolinone biosynthesis and allows for the generation of 3b with reduced cytotoxicity and enhanced MDR activity in comparison with its precursor 3a.^[2a] The dedicated activity of WelM constitutes an excellent example of how nature uses late-stage tailoring enzymes to diversify its product inventory with enhanced biological profiles.

Figure 4.

Prediction and characterization of WelM involved in a late stage N-methylation for welwitindolinone biosynthesis. (A) Proposed function of WelM in the generation of N-methylated welwitindolinones; (B) Characterization of WelM and its enzymatic product derived from 3a.

The remaining biosynthetic genes embedded in the wel cluster encode five nonheme iron (NHI)-dependent oxygenases, including four full length Rieske-type oxygenases (WelO1-O4) and a Fe(II)/α-ketoglutarate-dependent oxygenase (WelO5). While the number and diversity of wel oxygenases mirrors those in the amb pathway, their protein sequence identities are visibly lower (61-79%) (Table SI-1), in comparison with the rest of wel/amb biosynthetic enzymes (vide supra). Classification of fifteen welwitindolinones and 12-epi-hapalindoles/fischerindoles derived from the wel pathway based their oxidation states at the indole terpenoid cores (Figure 5A) clearly illustrates a need of five distinct 2e oxidation events to complete the full oxidative maturation of welwitindolinones. This analysis also re-asserts the proposed role of WelO5 as a candidate NHI-dependent halogenase for the stereoseletive generation of a C-Cl bond at the C-13 of hapalindole/fischerindole backbones as required for the generation of 5c/5d/6c from 5a/5b/6a respectively (Figure 5B). WelO1-4 are four highly homologous Rieske oxygenases (Figure SI-8), which are likely the main catalysts for the oxidative transformations of fischerindoles to welwitindolinones (Figure 5C), based on oxidation state analyses and a recently disclosed nonenzymatic oxidative transformation of 6c to 4.^[4c] In addition, distinct from Rieske oxygenases in the amb pathway, the iron-sulfur domain sequences in WelO1-4 are virtually identical (Figures SI-2/SI-8), implicating they likely originate from a common ancestor during HGT and are evolved specifically around the NHI catalytic domains.

Figure 5.

(A) Classification of welwitindolinones and 12-epi-hapalindoles/fischerindoles isolated from H. welwitschii based on the oxidation state of core indole-monoterpenoid scaffold. (B) Proposed role of WelO5 as a Fe(II)/α-ketoglutarate dependent halogenase in the generation of 5c/5d/6c from 5a/5b/6a. (C) Proposed roles of WelO1-4 (Rieske-type oxygenases) in the oxidative maturation of welwitindolinones from 12-epi-fischerindoles. A putative pathway to generate 3c from 6c by four sequential 2e-oxidations was shown.

Functional characterizations of WelI1-3/WelP2/WelM as well as bioinformatic correlations of WelO1-5 with the structural diversities of hapalindole-type molecules derived from H. welwitschii UTEX B1830 provided strong biochemical supports for linking the wel pathway to welwitindolinone biosynthesis. Comparative analysis of the wel and amb pathways collectively demonstrates the roles of GPP and 3-((Z)-2′-isocyanoethenyl) indole as common intermediates for the biosynthesis of welwitindolinones and ambiguines, and the critical roles of NHI-dependent oxygenases, particularly Rieske oxygenases as the key catalysts for late-stage oxidative structural diversifications.

The in vitro characterization of WelI1-3/WelP2 and inability of WelP1/P2 to generate β-ocimine, as shown in the amb pathway, solidified our conclusion that the stereodivergent generations of 12-epi-hapalindoles/fischerindoles from the wel pathway, in contrast to ambiguines with a conserved hapalindole G/U motif from the amb pathway, is not derived from the stereochemical differences of biosynthetic precursors (Figure 1D), as those proposed in the literature.^{[1, 4d, 6]}. Early studies on the amb pathway by us implicate a possible role of AmbP1 in directly fusing GPP and 10 to 8a through a cationic cascade.^[5] However, as the sequences of WelP1 and AmbP1 are virtually identical (Table SI-1), it seems unlikely that they will be the sole catalyst responsible for the stereodivergent generation of 5a, 6a and 8a. In light of the recent disclosures of oxygenase-mediated terpene-like cyclizations of geranylated aromatic polyketides,^[10] it is plausible that both WelP1 and AmbP1 will mono-geranylate 10 with GPP and subsequently rely on a dual functional oxygenase in the AmbO or WelO families. In addition, the wel pathway does not contain a dedicated gene for transferring a nucleophilic “sulfur” to isonitrile to generate isothiocyanate group. It is possible that a pathway-independent sulfur carrier protein (such as ThiS in thiamin biosynthesis) or an endogenous cystein desulfurase harboring a reactive thiocarboxylate or a cystein persulfide may serve as the sulfur donor,^[11] as recently proposed for 2-thiosugar biosynthesis.^[12] Efforts in delineating the remaining mysteries associated with hapalindole-type natural product biosynthesis are currently undergoing in our laboratory.

Experimental Section

Experimental details, including SI Methods, Figure SI-1 to SI-10, Table SI-1&2, are described in the supporting information. Nucleotide sequence of the wel cluster was deposited in Genbank with accession number KF811479.