Expanding Polycyclic Tetramate Macrolactam (PoTeM) Core Structure Diversity by Chemo‐Enzymatic Synthesis and Bioengineering

Abstract

Polycyclic tetramate macrolactams (PoTeMs) represent a growing class of bioactive natural products that are derived from a common tetramate polyene precursor, lysobacterene A, produced by an unusual bacterial iterative polyketide synthase (PKS)/non‐ribosomal peptide synthetase (NRPS). The structural and functional diversity of PoTeMs is biosynthetically elaborated from lysobacterene A by pathway‐specific cyclizing and modifying enzymes. This results in diverse core structure decoration and cyclization patterns. However, approaches to directly edit the PoTeM carbon skeleton do currently not exist. We thus set out to modify the PoTeM core structure by exchanging the natural l‐ornithine‐derived building block by l‐lysine, hence extending macrocycle size by an additional CH₂ group. We developed streamlined synthetic access to lysobacterene A and the corresponding extended analog and achieved cyclization of both precursors by the cognate PoTeM cyclases IkaBC in vitro. This chemo‐enzymatic approach corroborated the catalytic competence of IkaBC to produce a larger macrolactam yielding homo‐ikarugamycin. We thus engineered the adenylation domain active site of IkaA to directly accept l‐lysine, which upon co‐expression with IkaBC delivered a recombinant bacterial homo‐ikarugamycin producer. Our work establishes an entirely new PoTeM structural framework and sets the stage for the biotechnological diversification of the PoTeM natural product class in general.

Polycyclic tetramate macrolactams (PoTeMs) derive from ornithine‐based lysobacterene A. To expand the PoTeM core structure carbon skeleton by a CH₂‐unit, a synthetic access to lysobacterene A and an extended analog was developed, and cyclization by cognate ikarugamycin cyclases IkaBC was achieved in vitro. Building on these results, the adenylation domain of the NRPS IkaA was programmed to directly incorporate l‐lysine, giving biotechnological access to the first backbone‐engineered PoTeM, homo‐ikarugamycin.

Introduction

Polycyclic tetramate macrolactams (PoTeMs) are a growing class of natural products with complex structures and a broad range of biological activities. The first member ever discovered was the antibiotic and antiprotozoal ikarugamycin (1), which was isolated in 1972 from Streptomyces phaeochromogenes var. ikaruganensis. ^[1] The core structure of 1 and all other PoTeMs known to date is assembled by an unusual bacterial iterative type I PKS/NRPS (IkaA).[ ² , ³ ] The iPKS system produces two long‐chain polyene building blocks. The NRPS system utilizes an l‐ornithine and transfers the polyenes onto its two amino functions (Scheme 1). The thioesterase catalyzes tetramic acid formation with concomitant release of the central PoTeM biosynthetic intermediate lysobacterene A (2 a). Post‐PKS/NRPS processing is achieved by the oxidoreductase IkaB and the alcohol dehydrogenase IkaC, leading to a cyclization sequence that generates the characteristic 5–6–5‐carbocyclic pattern of 1 (Scheme 1).[ ³ , ⁴ ]

Scheme 1.

Biosynthesis and structural diversity of PoTeMs. All PoTeMs derive from a common precursor lysobacterene A (2 a), which is produced by an iPKS/NRPS (e.g., IkaA). Oxidoreductases (light blue) catalyze the formation of different carbocyclic patterns, leading to a broad variety of PoTeM core structures. Tailoring enzymes further diversify the PoTeM structural spectrum by introducing functional groups. PoTeM hydroxylases (PH, purple) can act on a broad variety of PoTeMs and introduce a hydroxy group at C‐3 (3–6). Monooxygenases like CftA and IkaD (green) add epoxides, ketones, and hydroxy groups, resulting in, e.g., 8 and 9. Additionally, the tetramic acid can be methylated by a methyl transferase (red) towards 10. Recently, also halogenated derivatives have been discovered, e.g., 12 (orange).

Since the discovery of 1, a large number of structurally diverse PoTeMs has been discovered from phylogenetically diverse organisms. While the core biosynthetic steps, leading to the production of intermediate 2 a, are identical in all biosyntheses, the large structural variety of PoTeMs derives from the utilization of different cyclizing oxidoreductases generating a variety of carbocyclic structural patterns. ^[5] Beyond the prototypical 5–6–5‐cyclization pattern of 1, further motives include the 5–5–6‐cyclization pattern present in heat‐stable antifungal factor (HSAF, 3) ^[2] and in frontalamides A/B (4/5), ^[6] the 5–5‐cyclization pattern of alteramide A (6), ^[7] and the 5–4–6‐cyclization pattern of compound D (7). ^[8] The structural diversity of PoTeMs is further expanded by tailoring enzymes that catalyze late‐stage decoration of the carbon skeleton. A widespread modification is hydroxylation at C‐3 (Scheme 1) by PoTeM hydroxylases.[ ⁹ , ¹⁰ ] Additionally, hydroxy functions and ketones (e.g., in clifednamide C, 8), ^[11] epoxides (e.g., in capsimycin B, 9), ^[12] and methyl groups (e.g., in lysobacteramide B, 10; via 5–5–6‐cyclized and hydroxylated 11) ^[13] can be introduced by tailoring enzymes (Scheme 1). Moreover, even halogenated congeners have been found recently (e.g., capsimycin D, 12). ^[12]

Interestingly, post‐PKS/NRPS PoTeM biosynthetic enzymes frequently show high promiscuity and can act on different PoTeM core structures. Over the last few years, new, artificial PoTeMs were generated capitalizing on this property by interchanging modifying enzymes,[ ⁹ , ¹⁴ , ¹⁵ ] or by combining cyclization enzymes from different PoTeM biosynthetic gene clusters.[ ¹⁶ , ¹⁷ ] However, all these engineering approaches so far focused on the deletion, addition, or exchange of cyclization or late‐stage modifying enzymes, while approaches targeting direct alterations of the carbon backbone of PoTeMs do not currently exist. Within this study, we thus aimed for the generation of a PoTeM analog with an edited carbon backbone. We focused on exchange of the amino acid building block from ornithine to l‐lysine, thereby extending the macrolactam by one CH₂‐group. To probe the promiscuity of the cyclization enzymes IkaBC, we established synthetic access to lysobacterene A (2 a) and its extended analog homo‐lysobacterene A (2 b) and show that both precursors are accepted as substrates, thereby for the first time generating access to a ring‐expanded, new‐to‐nature PoTeM analog, homo‐ikarugamycin (13). To enable future biotechnological PoTeM diversification, we furthermore reprogrammed the selectivity of the adenylation domain of IkaA to generate a recombinant bacterial homo‐ikarugamycin (13) production platform.

Results and Discussion

Chemical Synthesis of the Polyene Tetramate Intermediates 2 and in vitro Promiscuity Tests of IkaB and IkaC

To probe the required promiscuity of the cyclization enzymes IkaBC, we first established synthetic access to ornithine‐derived native precursor 2 a, and its extended analog, lysine‐derived precursor 2 b (cf. ESI, chapter 2.1). Retrosynthetic analysis of the desired polyenes 2 (Scheme 2) led to four building blocks, namely the core tetramic acid (green), the C₂ moiety at C‐3 (red) for attachment of the southern conjugated polyene chain (blue), and the C₁₂ side chain connected to the C ^δ/ϵ amine of the corresponding amino acid (yellow).

Scheme 2.

Retrosynthetic analysis of IkaA intermediate 2 with key reactions. Individual building blocks are depicted in green (tetramic acid), red (C₂ moiety for attachment of southern side chain), blue (polyene aldehyde), and yellow (C₁₂ polyene carboxylic acid). Both polyene chains (blue and yellow) can be accessed by Iridium‐catalyzed stepwise chain elongation.[ ¹⁸ , ¹⁹ ]

The central tetramic acids 14 were synthesized in a two‐step procedure, starting from the corresponding twofold N‐Boc‐protected amino acids l‐ornithine (15 a) or l‐lysine (15 b). EDC‐ and DMAP‐assisted acylation of Meldrum's acid with 15 a/15 b introduced the required C₂ moiety. Heating of the crude product in EtOAc then delivered the desired cyclic products in 78 % (14 a) and 76 % (14 b) yield (Scheme 3A). These intermediates were acylated in 3‐position with Bestmann's ylide in THF, yielding ylides 16 a (77 %) and 16 b (81 %).

Scheme 3.

Preparation of the central building blocks. A. Synthesis of ylides 16 and B. stepwise chain elongation towards conjugated olefin side chains (acid 25 and aldehyde 26).

As two different conjugated polyene chain lengths were necessary for the installation of the polyene side chains, we employed a methodology involving stepwise C₂ chain elongation of commercially available ethyl sorbate (17) by an Ir‐catalyzed reductive Horner–Wadsworth–Emmons (HWE) olefination established by Wunderlich et al. ^[19] [Ir(coe)₂Cl]‐catalyzed hydrosilylation with diethylsilane produced the corresponding diethylsilyl acetal 18 (Scheme 3B), which was directly used in a KOSiMe₃‐assisted HWE reaction with triethyl phosphonoacetate (19). This delivered (E,E,E)‐ethyl octatrienoate (20) in 54 % yield.

For the generation of the northern C₁₂ carboxylic acid chain (Scheme 2), this procedure was repeated twice (see Scheme 3B for individual molecule numbers), eventually yielding all‐trans ethyl dodecapentaenoate (24, see SI, chapter 2.1.3 for individual yields), which was saponified using KOH, liberating the corresponding acid 25 (61 %).

For the southern C₈ olefin, the intermediate silyl acetal of ethyl octatrienoate 21 was hydrolyzed with aqueous HCl in THF, resulting in formation of the all‐trans octatrienal 26 in 40 % yield (Scheme 3B).

With this aldehyde in hands, we engaged in KO ^t Bu‐mediated Wittig olefination of ylides 16, ^[20] thus attaching the southern olefin side chain in acceptable yields (Scheme 4A; 27 a: 41 %; 27 b: 51 %).

Scheme 4.

Fusing the individual compounds towards 2. A. Wittig‐olefination of ylides 16 and B. completion of the synthesis of IkaA intermediates 2.

To accomplish the synthesis of the desired IkaA intermediates 2, lastly both nitrogen functionalities were liberated using HCl in dioxane, before being directly used for peptide coupling with carboxylic acid 25 using PyBOP and NEt₃ (Scheme 4B; details in ESI, chapter 2.1.5). These reactions had to be performed at small scale (<5 mg 27), as workup and isolation of the notoriously unstable compounds 2 was difficult for larger amounts. HRMS data and UV spectra of the products 2 (isolated by preparative HPLC) thus obtained (cf. ESI, chapter 2.1.5) were in agreement with literature data (available for 2 a). ^[21]

In previous work from our laboratory, 1 was enzymatically synthesized in vitro from its building blocks acetyl‐CoA, malonyl‐CoA, and l‐ornithine using recombinant IkaA, IkaB, and IkaC. ^[22] This method was adapted to convert 2 a and 2 b to PoTeMs. IkaB and IkaC were expressed in E. coli BL21 pGro7 with an N‐terminal octahistidine tag and purified by affinity chromatography on Ni‐NTA beads (cf. ESI, chapter 3.4). Successful conversion of 2 a and 2 b to ikarugamycin (1) and homo‐ikarugamycin (13), respectively, was observed by HR‐MS (Figure S15). This chemo‐enzymatic investigation hence corroborated suitable substrate promiscuity of IkaBC that allowed the generation of the first PoTeM with an extended carbon framework, homo‐ikarugamycin (13).

Engineering Homo‐Ikarugamycin (13) Production in vivo

Having validated the catalytic ability of IkaB and IkaC to process elongated homo‐lysobacterene A (2 b), we aimed at engineering the PoTeM biosynthetic machinery to accept lysine to facilitate biotechnological production of homo‐ikarugamyin (13). Towards achieving this goal, we initially tested the promiscuity of the A domain of IkaA for lysine activation in vitro. Therefore, the NRPS of ikaA was cloned into a pET‐28a vector (Figure S4/S5), ^[2] expressed in E. coli BAP1, and purified by Ni‐NTA chromatography using the genetically attached C‐terminal His‐tag (cf. ESI, chapter 3.4). An HPLC‐MS/MS‐based multiplexed hydroxamate assay (HAMA; cf. ESI, chapter 3.5 for details, and chapter 2.2 for procedures of standard synthesis) ^[23] with l‐ornithine and l‐lysine as substrates revealed the expected high activity of wild‐type IkaA for l‐ornithine. However, l‐lysine was not activated under the tested conditions (Figure S16).

Consequently, the engineering of IkaA focused on adjusting the A domain selectivity. To keep the changes to IkaA as small as possible, only the active site of its encoded A domain was mutated, with the aim of enabling the activation of l‐lysine. To determine suitable mutations, the MIBiG database was searched for l‐lysine‐activating NRPS A domains. ^[24] 52 candidates were found (Figure S26) and their active site residues were determined with reference to the crystal structure of gramicidin synthetase 1 (PheA). ^[25] Several domains shared the same substrate‐conferring codes or had large similarities and were thus grouped (Table S5). This analysis resulted in 15 representative active‐site compositions containing two to eight point mutations when compared to the native IkaA A domain sequence (Figure 1). The section of the A domain, containing all active site residues (312 bp), was ordered for every mutant as synthetic, linear DNA from Twist Bioscience. The expression vector (pET‐28a‐ikaA‐NRPS) was amplified by PCR without the region to be replaced (Figure S6). The synthetic DNA fragments and the vector backbone were fused by sequence‐ and ligation‐independent cloning (SLIC; cf. ESI, chapter 1.2). ^[26] The mutated A domains were tested in vitro for their ability to activate l‐ornithine and l‐lysine again using the hydroxamate assay (HAMA). ^[23] All constructs lost their ability to activate l‐ornithine, and only one construct (mut6) was activating l‐lysine, albeit with a relatively low turnover (Figure S16). The construct mut6 contained two mutations, E281D and I308V. Both introduced mutations thus corresponded to changes of the respective amino acids to smaller homologs. This likely generated a larger active site substrate pocket, which provides sufficient space for the larger l‐lysine to bind.

Figure 1.

NRPS expression plasmid and substrate‐conferring codes of IkaA and 15 representative l‐lysine activating A domains. The residues were determined according to PheA and contained two to eight differences (grey) compared to IkaA. Mutations of lysine‐activating construct mut6 are labeled in green.

Subsequently, the amino acid changes in mut6 were introduced into the in vivo ikarugamycin expression vector (pSET152‐ermE*‐ikaABC; Figure S9/S10). To elucidate the influence of each of the mutations individually and combined, we generated IkaA single mutant I308V, single mutant E281D, and a mutant containing both changes to test PoTeM production in comparison to the wild‐type system in vivo.

For each single mutant strain, the native expression vector was amplified in two overlapping parts with a primer pair introducing the corresponding point mutation (Figure S11). The half‐plasmids were fused by Gibson assembly. ^[27] For the double mutant strain, the vector backbone was also amplified in two overlapping parts but without the region of the active site of the A domain. Together with the synthetic DNA of mut6, a Gibson assembly was performed to connect the three DNA parts (cf. ESI, chapter 3.3 for details).

The plasmids were conjugated into Streptomyces albus KO5 ^[28] as heterologous host and cultivated in ISP‐4 medium. The cells were harvested, extracted with methanol/acetone (1 : 1), the dried extracts re‐dissolved in MeOH, and analyzed using HPLC‐HRMS/MS with separation on C₁₈ reversed‐phase material (cf. ESI, chapter 3.6 and 3.7 for details). Interestingly, mutant I308V showed production levels of ikarugamycin (1) comparable to the wild‐type strain (m/z [M+H]⁺: 479.290, 15.3 min, Figure 2). Two additional peaks with a slightly increased retention time on the RP18 column and an m/z of 493.305 [M+H]⁺ were observed, which fits the expected mass change of +14 u compared to 1 for the desired product homo‐ikarugamycin (13). The compound with the highest retention time (16.6 min) was found to be N‐methyl ikarugamycin (28), corroborated by MS/MS fragmentation analysis (ESI chapter 3.7.1; Figures S17B and S19), a modification already described from native ikarugamycin‐producing strains. ^[29] To our delight, the compound with a retention time of 15.6 min was unambiguously identified as 13 by its MS/MS fragment pattern (Figures S17C and S20). Moreover, the significant change in polarity of 28 versus 1 (resulting in the retention time shift of +1.3 min on RP18) was used as further proof for this assignment, as N‐methylation has a high impact on the topological polar surface area (TPSA) value of the substances (see ESI, chapter 3.7.2). In contrast, the retention time of 13 versus 1 only increases by 0.3 min, which corresponds to the identical calculated TPSA values of these molecules. ^[13] To our knowledge, this is the first time that a PoTeM incorporating an amino acid building block different from l‐ornithine was observed. However, the amount of 13 was only around 0.5 % of that of 1. Interestingly, upon its identification using our recombinant mutant expression system, 13 was also present in barely detectable amounts in the native ikarugamycin producer, which contrasts our A domain activation experiments mentioned above.

Figure 2.

HRMS analysis of the cell extracts from Streptomyces expressing mutated ika. A. EIC for mass of ikarugamycin (1). B. EIC for the mass of homo‐ikarugamycin (13) with extended carbon skeleton (green). I) Wild‐type, II) I308V, III) double mutant, and IV) E281D. The intensities of chromatograms in B were scaled up by a factor of five.

In the double mutant, the amount of 1 was dramatically reduced, while the yield of 13 was eight times higher compared to the I308V single mutant (observed by HRMS; Figure 2). The E281D single mutant was even more efficient: the amount of 1 was decreased to practically undetectable levels and simultaneously, 13 was more than 36 times more abundant than in the I308V single mutant. Consequently, the selectivity of the enzyme was switched by a factor of approximately 10³ to an enzyme that almost exclusively activates l‐lysine. In summary, we thus generated for the first time a recombinant production system of a novel l‐lysine‐containing PoTeM structural framework.

Structural Analysis of the A Domain

The structure of the wild‐type A domain of IkaA was predicted using AlphaFold ^[30] and superimposed onto the structure of PheA (PDB code: 1AMU, Figure S3, cf. ESI, chapter 3.2 for details). ^[25] With phenylalanine bound to PheA as a reference, l‐ornithine was modeled into the active site of IkaA. The main interaction of this substrate with active‐site residues was formed by Glu281 and Asp342, which were found to have a distance of 2.2 Å and 3.2 Å from the δ‐amino group of l‐ornithine, and thus stabilize positioning of the substrate in the binding pocket (Figure 3A). In the double mutated enzyme mut6, Glu281 and the neighboring Ile308 were exchanged to their smaller homologs Asp and Val, respectively. Thus, the larger l‐lysine can fit into the active site and the ϵ‐amino group can form an interaction with Asp281 (2.4 Å, Figure 3C). However, due to the remaining Asp342, l‐ornithine can still bind to the active site leading to the formation of 1 (Figure 3B). As already observed experimentally, mutation of Ile308 does not increase the binding of l‐lysine, but instead, a mutation of Asp342 might be favorable.

Figure 3.

Active site of the IkaA structure predicted by AlphaFold and modeling of substrate binding. A. In the wild‐type IkaA, ionic interactions between Glu281 (2.2 Å) and Asp342 (3.2 Å) to the amino group of l‐ornithine are possible. B. In the double mutant mut6 (E281D, I308V), Asp342 can still interact with l‐ornithine (pink). C. Through mutation of Glu281 to Asp the active site pocket is extended, making space and allowing binding for the longer l‐Lys (purple) substrate. In addition, the interaction between the carboxylic acid group of Asp281 and the ϵ‐amino group of l‐Lys provides specificity.

Interestingly, the respective A domain encoded by the related HSAF gene cluster naturally contains a valine at the corresponding position of Ile308. ^[2] We also predicted the structure of the A domain of the HSAF NRPS module using AlphaFold. The architecture of its active site is nearly identical to IkaA‐I308V (Figure S25). However, the A domain was reported to be nearly as active for l‐lysine when compared to l‐ornithine in vitro, ^[2] but formation of l‐lysine‐derived HSAF was not reported.

Incorporation of further, even larger amino acids, particularly homo‐lysine, might become possible by further increasing the size of the A domain active site cavity, e.g., by mutations E281S and/or I308V/A.

Conclusions

Within this work, we developed a synthetic route to the common PoTeM biosynthetic precursor lysobacterene A (2 a) and the lysine‐derived extended analog 2 b, comprising tetramic acid formation, Ir‐catalyzed polyene synthesis, and fusion of the building blocks by Wittig and acylation reactions. Availability of 2 a/b enabled in vitro assays on the substrate promiscuity of the cyclization enzymes IkaBC, revealing their catalytic competence to form ikarugamycin (1) and the novel homo‐ikarugamycin (13) structural framework. This set the stage for the generation of a recombinant production platform for 13. We were able to dramatically switch the selectivity of the A domain of IkaA from l‐ornithine to l‐lysine by directed mutagenesis of a single residue (E281D). The corresponding mutant lost the ability to produce 1 and exclusively led to the production of 13 in vivo.

Both ways to the new lysine‐derived intermediate set the stage for the (bio‐)synthesis of further lysine‐derived PoTeMs: On the one hand, synthetic 2 b can be converted in vitro by a variety of related cyclizing enzymes. Alternatively, together with our plug‐and‐play system for a fast and easy PoTeM production, ^[31] cyclizing genes from other PoTeM gene clusters can be combined with the mutated ikaA to produce lysine‐derived PoTeMs in vivo. Both approaches will enable future structure–activity studies on these completely new natural product derivatives.

Supporting Information

The authors have cited additional references within the Supporting Information.[ ³² , ³³ , ³⁴ , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ³⁹ , ⁴⁰ ]

Conflict of Interests

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Schuler S., Einsiedler M., Evers J. K., Malay M., Uka V., Schneider S., Gulder T. A. M., Angew. Chem. Int. Ed. 2025, 64, e202420335. 10.1002/anie.202420335

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information or from the corresponding author upon request.