Heterologous production of the C33–C45 polyketide fragment of anticancer apratoxins in a cyanobacterial host

Abstract

A polyketide synthase sub-cluster of cytotoxic apratoxin A was isolated from a Moorena bouillonii environmental DNA library and engineered with a thioesterase II domain for heterologous expression in the filamentous cyanobacterium Anabaena sp. PCC7120. Further engineering with a rhamnose-inducible promoter led to the production of (2R,3R,5R,7R)-3,7-dihydroxy-2,5,8,8-tetramethylnonanoic acid, a stereogenically rich chiral building block important to the efficient synthesis of apratoxin analogs, representing the first synthetic biology attempt to this type of polyketide fragment.

Graphical Abstract

Polyketides (PKs) and PK-containing natural products (e.g., polyketide-peptide and -terpenoid hybrids) are large families of secondary metabolites of terrestrial and marine plants, animals, fungi and bacteria. These compounds have intriguing structural complexity often with various stereoselective modifications and possess extremely rich functional diversity. Indeed, 1% of known PKs possess drug activity and tens of PK-related drugs (e.g., derivatives of erythromycin, epothilone and dolastatin 10) have been clinically used to treat diseases ranging from cancer and infection to hypocholesterolemia.¹ Continuous exploration of PKs and their hybrids would be expected to offer new therapeutic leads. However, the supply of these compounds for bioactivity studies remains challenging as total chemical synthesis, particularly PK (sub)structures, and isolation from biological sources have achieved limited successes, with some notable exceptions where multigram synthesis and process development have been developed.²

Depsipeptides are a class of PK-containing natural products with a rich diversity in structure and function. The generation of stereochemically rich PK fragments (e.g., unit A of cryptophycins) is often the most demanding task in depsipeptide synthesis.^{3,4 Once} PK substructures are available, solid phase peptide synthesis can be feasible to assemble natural and unnatural depsipeptides in drug research. Remarkably, nature elegantly synthesizes PKs directly from simple acyl-CoAs catalyzed by polyketide synthases (PKSs).

Apratoxin A (Fig. 1) is a cyclic PK-peptide hybrid produced by the marine filamentous cyanobacterium Moorena bouillonii,⁵ and it possesses potent anticancer activity by inhibiting cotranslational translocation early in the secretory pathway, directly targeting Sec61α.^{6, 7} This compound serves as the starting point for preclinical investigations and medicinal chemistry campaigns to improve its activities. Apratoxin A carries two PK fragments, and the longer one (C33-C45, APK), (2R,3R,5R,7R)-3,7-dihydroxy-2,5,8,8-tetramethylnonanoic acid, has one characteristic terminal t-butyl group, two hydroxy and two methyl groups in stereochemically defined orientation (Fig. 1). Despite significant efforts in developing different synthetic strategies, available routes to the APK remain lengthy and laborious,^{4, 8} limiting their scalability and applicability for the preparation of apratoxins and analogs. Indeed, once in hand, APK could be used to generate improved synthetic analogs such as apratoxin S4 and S9 (Fig. 1) with differences in the C27-C31 unit, lacking the Michael acceptor and differing in configurations at C30 to modulate activity and selectivity.⁹ An economically feasible simple route to APK would be a major advancement towards the development of this promising class of anticancer, antiviral and antiangiogenic agents acting through a novel mechanism of action.^{7, 9–11}

Fig. 1.

Structures of the natural product apratoxin A and synthetic analogs with improved properties.

Heterologous expression in surrogate hosts is a proven and useful strategy to produce microbial and plant natural products and analogs.¹² Escherichia coli, Streptomyces strains and yeast are common hosts in these studies. Recently, model cyanobacterial strains have received increased attention for the heterologous production of cyanobacterial natural products.¹³ For example, we have developed the unicellular cyanobacterium Synechocystis sp. PCC6803 to produce the mycosporine-like amino acid shinorine,¹⁴ while the filamentous cyanobacterium Anabaena sp. PCC7120 (hereafter referred to as Anabaena) has been used to produce multiple types of cyanobacterial natural products (e.g., lyngbyatoxin A,¹⁵ cryptomaldamide¹⁶ and columbamides¹⁷). Here we demonstrate the heterologous production of APK in Anabaena expressing an engineered PKS sub-cluster.

The biosynthesis of apratoxin A is deduced on the genetic basis of its 58-kb gene cluster that was previously identified through a combination of single-cell sequencing and genomic library screening.¹⁸ The cluster is comprised of 12 biosynthetic genes (aprA-L). The synthesis of APK presumably involves seven enzymes, AprA-G, whose corresponding genes span a 25-kb chromosomal region (Fig. 2). AprA is a loading PKS module for synthesizing the t-butyl terminus, functionally similar to the one in curacin biosynthesis.¹⁹ AprB is a bimodule PKS that elongates the intermediate by four carbons and stereospecifically introduces a -OH group. Subsequently, an HMG-CoA synthase (HCS)-like gene cassette (AprC-F), frequently associated with β-branching in PK biosynthesis, introduces the β-Me group (in pink, Fig. 2).²⁰ AprG then extends the intermediate chain with two carbons, introduces one α-Me group, and reduces the β-ketone to a chiral -OH group (Fig. 2). APK tethered on the thiolation (T) domain of AprG is then sequentially assembled with l-cysteine, malonyl-CoA, l-tyrosine, l-alanine, l-isoleucine, and l-proline by AprH-L.¹⁸ The elongated intermediates receive multiple in-line modifications (e.g., N-, O-, and C-methylations), and the macrolactonization via nucleophilic attack by the terminal -OH of APK generates the final structure of apratoxin A (Fig. 1).

Fig. 2.

The biosynthetic scheme of the long polyketide fragment (APK) of cytotoxic cyanobacterial depsipeptide apratoxin A. The isolated 25-kb APK sub-cluster carries 7 genes encoding three PKSs (in blue) and one complete set of HCS cassette (in maroon) for β-methylation. A type-II thioesterase domain (TEII) in the engineered sub-cluster releases the APK with a carboxylate terminus.

The genetic understanding of apratoxin biosynthesis sets the stage for the heterologous production of APK, which would be supported further by public access to the complete sequence information of the cluster. We therefore sought to identify the APK sub-cluster (aprA-G) by screening a metagenomic fosmid library prepared from the M. bouillonii sample VPG14–77 collected at Piti Bomb Holes in Guam (Fig. S1s). VPG14–77 produced apratoxin A as confirmed by high resolution MS (HRMS),¹H NMR and optical rotation analysis of isolated compound (Fig. S2, Table S1). The 100,000-clone library represented >80-fold coverage of the entire M. bouillonii genome (estimated to be <10 Mb) and was archived in seven pools. One pair of degenerate primers was designed on conserved regions of HCS enzymes for the biosynthesis of curacin A, jamaicamide A and mupirocin and used to screen the metagenomic library (Figs. S3–S4, Table S2). We successfully amplified the product of the expected size (447 bp) from the isolated metagenomic DNA (Fig. S5) and recovered three single HCS-positive clones from the library. Two clones (bAprat1 and bAprat8) are identical and are not part of the apratoxin cluster, but the clone bAprat14 harbors modular PKSs, HCS-cassette, and NRPS genes (GenBank ID:MG890637.1) with gene order and domain organization highly similar to the reported APK sub-cluster (aprA-G, Fig. 2), aprH and the ketosynthase (KS) domain of aprI (Table S3). Compared with the published one,¹⁸the APK sub-cluster reported here carries 112-amino acid longer AprB and AprG with one extra dehydratase (DH) domain. The DH domain lacks a conserved active site motif HXXXGXXXXP, likely being nonfunctional. These differences between the two APK sub-clusters indicate divergence in the clusters within cyanobacterial species collected from different geographic locations. Moreover, we previously isolated apratoxin E from another collection of Moorena sp.²¹ Compared with apratoxin A, the APK moiety of apratoxin E carries the C=C at C34–35 and lacks the C44 methyl, suggesting functional variations of AprG’s DH and methyltransferase (MT) domains. Of note, the configuration at C30 in apratoxin E is 30R as in apratoxin S9 (Fig. 1).^{9, 22} Interestingly, aprE-H of the cluster identified here are translationally coupled as the stop codon of preceding gene overlaps with the start codon of succeeding gene, fixing corresponding biosynthetic enzymes with equal molarity at the translational level.²³

We next sought to engineer the aprA-G sub-cluster for the heterologous production of APK in Anabaena (Fig. 2). We selected a replicative BAC vector pRL838 with both erythromycin and chloramphenicol resistance markers (Ery^R and Cm^R) for cloning the sub-cluster,²⁴ whose expression was controlled by a proven strong promoter Ptrc (Fig. 3). To release the APK from the AprG T domain, we chose the type-II thioesterase (TEII) domain of the erythromycin cluster (Fig. S6).²⁵ Ptrc and codon-optimized TEII were first cloned into pRL838. The expression vector pRL838-Apra (~34 kp) was then constructed by Gibson assembly using five fragments, including three (Apra I, 7,446 bp; Apra II, 7,127 bp; Apra III, 10,279 bp) amplified from bAprat14 and two backbone fragments (838-ptrc, 4,738 bp; 838-TEII, 4,959 bp) amplified from the above engineered pRL838 vector (Fig. 3, Fig. S7). The stop codon of aprG was eliminated in PCR amplification for functional co-transcription with TEII. The Cm^R was split between 838-ptrc and 838-TEII to facilitate the screening of clones carrying the properly assembled construct. Assembled constructs were further confirmed by digestion with three restriction enzymes (Fig. S8).

Fig. 3.

A schematic representation of the construction of engineered APK sub-cluster for the heterologous expression in Anabaena.

We then introduced pRL838-Apra into Anabaena utilizing triparental conjugation.²⁶ However, multiple attempts yielded the same observation: transformed cyanobacterial cells gradually lost viability during segregation and selection. As a control, we prepared a construct for expressing the biosynthetic gene cluster of mycosporine-like amino acid shinorine under the Ptrc promoter¹³ in Anabaena. Positive Anabaena transformants were successfully selected and produced shinorine in culture (Fig. S9). These results suggested the potential toxicity of expressed APK on the host cells. We therefore replaced the constitutive Ptrc promoter in pRL838-Apra with a Co²⁺ or rhamnose inducible promoter (PcoaT or Prha) (Fig. 3). The high efficient promoter switch was developed by combining in vitro Cas12a digestion²⁷ and HiFi assembly with in vivo Red-ET homologous recombination (detailed in supplementary material, Fig. S10A).²⁸ Engineered constructs were identified by PCR screening (Fig. S10B). The resultant constructs carrying PcoaT or Prha were named as pRL838-cApra and pRL838-rApra, respectively, and were conjugated into Anabaena. After multiple rounds of segregation and selection, positive transformants of both constructs, named Anabaena cApra and rApra, were confirmed by PCR screening (Fig. S11) and were then cultured in BG-11 medium. Of note, the growth of Anabaena wild type and rApra was similar but the cApra was significantly slower. After 21 days, the cultures were diluted to OD₇₃₀ of 0.1 and the production of APK was induced by 1 μM CoCl₂ or 5 mM rhamnose for 7 additional days. LC-HRMS analysis of the methanolic extract of wet biomass of Anabaena cApra and rApra identified a new peak (Fig. 4A), which was missing in those of the vector control and uninduced Anabaena rApra as well as the uninduced cApra (Fig. S12). The peak content showed an expected [M+Na]⁺ value of APK at 269.1718 (C₁₃H₂₆NaO₄⁺; calculated [M+Na]⁺:269.1723) in HRMS analysis (Fig. S13). Tandem HRMS analysis identified the ions of two fragments with loss of one and two water molecules and multiple additional ions of predicted fragments (Fig. 4B,Fig. S14). To further confirm the content identity of the new peak, we prepared the APK authentic standard from a previously synthesized intermediate⁹ in a 5-step route (Scheme S1). The structures of synthetic intermediates and the final product APK were confirmed by HRMS and¹H and ¹³C NMR analyses (Figs. S15–24). The new peak of the extracts of Anabaena cApra and rApra had the same retention time and MS/MS fragmentation pattern as the authentic standard (Fig. 4A–B,Fig. S25). These results indicated that APK was successfully produced in Anabaena expressing the engineered apratoxin PKS sub-cluster. This work also confirms the cluster identified in the current and previous studies¹⁸ responsible for the apratoxin biosynthesis. Compared with the PcoaT promoter, Prha demonstrated a better production of APK (Fig. 4A). Indeed, we observed upregulated transcription of AprA-G 24 h and 120 h after rhamnose induction (Fig. 4C). In addition, rhamnose at 5 mM gave rise to the highest yield (9.7 mg/L of culture), compared with 1 mM and 10 mM inducer (Fig. S26).

Fig. 4.

Heterologous production of APK in engineered Anabaena. A: The total-ion current (TIC) traces of the methanolic extracts of Anabaena cApra and rApra identified a new peak shaded in brown that showed the same retention time as the synthetic APK standard. This peak was missing in uninduced Anabaena rApra and the cells transformed with the pRL838-Ptrac vector. B: HRMS/MS spectrum of [M+H]⁺ of the new peak content. The m/z values of predicted APK fragments were labeled in red. C: RT-qPCR analysis of AprA-G transcription in Anabaena rApra after being induced by 5 mM rhamnose for 0, 24, and 120 h. The transcription level of RNA subunit of ribonuclease P (rnpB) was determined and used to normalize the signals of the biosynthetic genes in the same strain. Data represent mean ± standard deviation (n = 6).

In conclusion, the present work demonstrates the successful production of the C33–C45 PK fragment of the cyanobacterial depsipeptide apratoxin A using the filamentous cyanobacterium Anabaena as the host. The cyanobacterial platform can provide a controllable means to supply APK and designed analogs for the modular synthesis of apratoxins for preclinical investigations. The results also provided the first functional characterization of the apratoxin gene cluster. The successful heterologous expression of APK depended on the use of multiple synthetic biology approaches to redesign the sub-cluster and address the toxicity issue. These approaches can find broader applications in studying entire or partial known and cryptic cyanobacterial biosynthetic gene clusters. Finally, this work supports the development of cyanobacterial chassis for the environmentally friendly production of high-value chemicals.

Data Availability

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