Genome‐Guided Discovery of Antimalarial 4‐Amino‐2,4‐Pentadienoate‐Containing Cyclolipodepsipeptides

ABSTRACT

4‐Amino‐2,4‐pentadienoate‐containing cyclolipodepsipeptides (APD‐CLDs) represent a structurally distinctive family of natural products known for their selective activity against hypoxic cancer cells. To explore the structural diversity of APD‐CLDs, we have identified and prioritized cryptic APD‐CLD biosynthetic gene clusters (BGCs) for compound discovery. Using a combination of genetic and chemical methods, we successfully activated three dormant BGCs, leading to the discovery of 12 new APD‐CLDs. These newly discovered metabolites significantly expanded the diversity of the APD‐CLD family, with chloromalamides and arabimalamides representing the first halogenated and glycosylated members, respectively. Unexpectedly, chloromalamides and arabimalamides exhibited potent antiplasmodial activity, with IC₅₀ values in the 25–161 nM range against drug‐sensitive and multidrug‐resistant Plasmodium falciparum strains. Phenotypic studies revealed arabimalamide B halted parasite development during the asexual blood stage life cycle, resulting in enlarged digestive vacuoles, dispersed hemozoin, and ultimately reduced reinvasion efficiency. These phenotypes are reminiscent of the effect of chloroquine and other 4‐aminoquinoline drugs, suggesting that arabimalamides may disrupt the parasite's heme detoxification mechanism. Biosynthetic studies identified key scaffold‐forming and modifying enzymes, including a rare membrane glycosyltransferase in arabimalamide biosynthesis. Together, these findings unveil APD‐CLDs as new antimalarial lead scaffolds and set the stage for structural diversification and optimization.

Halogenated and glycosylated 4‐amino‐2,4‐pentadienoate‐containing cyclolipodepsipeptides (APD‐CLDs) exhibit potent antiplasmodial activity (IC₅₀ = 25–161 nM) against drug‐sensitive and resistant Plasmodium falciparum strains.

1. Introduction

Natural products are a chemically diverse reservoir of bioactive compounds, many of which have served as the foundation for clinically important therapeutics. Among these, naturally occurring cyclodepsipeptides constitute a promising class of chemical scaffolds with diverse pharmacological activities [1, 2]. Defined by their characteristic cyclic structures composed of proteogenic or nonproteinogenic amino acids linked through amide or ester bonds, cyclodepsipeptides exhibit high target‐binding specificity, conformational rigidity, and resistance to proteolytic degradation. Owing to these properties, cyclodepsipeptides continue to garner considerable interest as lead compounds in the search for new therapeutic agents [3, 4, 5].

4‐Amino‐2,4‐pentadienoate‐containing cyclolipodepsipeptides (APD‐CLDs) represent a structurally distinctive family of cyclodepsipeptides. To date, a total of 21 APD‐CLDs have been isolated from actinobacteria. All APD‐CLDs contain the signature electrophilic APD moiety, a lipophilic alkyl chain, and two amino acid units (AA1 and AA2, Figure 1). Among the APD‐CLDs, members of the rakicidin [6, 7, 8, 9, 10, 11] and BE‐43547 [12, 13, 14] subfamilies exhibited particularly compelling bioactivity, including hypoxia‐selective cytotoxicity against solid tumors at nanomolar concentrations. This selective activity in hypoxic microenvironment, a hallmark of solid tumors, underscores their potential for the development of tumor‐specific anticancer therapies. Structure–activity relationship (SAR) studies of natural and synthetic APD‐CLDs have shown that the electrophilic APD warhead is essential for the anticancer activity [8, 15, 16, 17], likely by forming a covalent linkage with a nucleophilic residue in the binding pocket of protein targets [18]. These findings established the APD moiety as the key pharmacophore defining the biological function of this unique family of natural products.

FIGURE 1.

Structures of representative APD‐CLDs and the biosynthetic modular PKS/NRPS systems. Representatives of APD‐CLDs are shown with the APD moiety in blue. BE‐43547 APD‐CLDs are synthesized by PKS/NRPS hybrid systems that feature a different module arrangement than those of rakicidin, boholamide, vinylamycin, and microtermolide‐like APD‐CLDs. (Abbreviation of domain names: A, adenylation; AT, acyltransferase; C, condensation; CP, carrier protein; DH, dehydratase; KR, ketoreductase; KS, ketosynthase). The DH domain of PKS4 is designated as DH* as it is distinct from the canonical DH domain, as discussed in the text.

The biosynthesis of APD‐CLDs involves a modular enzymatic assembly line composed of four polyketide synthase (PKS) modules and three nonribosomal peptide synthetase (NRPS) modules [13]. A defining feature of this biosynthetic machinery is a characteristic bimodular NRPS–PKS unit (Figure 1). This bimodular unit was presumed to be responsible for synthesizing the electrophilic APD warhead, with the dehydratase (DH) domain hypothesized by Poulsen [13], Igarashi [19], and others to catalyze two successive dehydration reactions. The DH domain putatively eliminates the β‐hydroxyl group and hydroxyl group of the serine residue in the growing chain, yielding the dienoate functionality that contributes to the chemical reactivity of the APD moiety (Figure S1). Overall, the structural diversity observed across APD‐CLDs arises from: (i) variation in the alkyl side chain synthesized by the first three PKS modules, (ii) variation of the two amino acid residues (AA1 and AA2) incorporated by the last two NRPS modules, and (iii) positional rearrangement of AA1 and AA2 within the cyclopeptide scaffolds (Figure 1).

Malaria continues to pose a major global health burden, particularly in sub‐Saharan Africa and Southeast Asia [20]. Current malaria treatment strategies mainly rely on artemisinin‐based combination therapies, including commonly used combinations such as artemether–lumefantrine, artesunate–amodiaquine, and dihydroartemisinin–piperaquine [21]. However, the long‐term efficacy of these regimens is threatened by rapidly evolving parasite resistance and limited drug chemotypes. Resistant strains of Plasmodium falciparum to artemisinin and its derivatives, as well as to other standard antimalarial drugs, have been reported and are increasingly spreading around the world [22]. Therefore, there is an urgent need to replenish the antimalarial drug pipeline with leads featuring novel chemical scaffolds. This study unveiled unprecedented halogenated and glycosylated APD‐CLDs through genome mining and activation of silent biosynthetic pathways. The new APD‐CLDs exhibited potent antimalarial activity against P. falciparum, displaying sub‐micromolar IC₅₀ values with a profound impact on blood‐stage parasite growth and development. Our findings highlight the unexpected antimalarial potential of APD‐CLDs, positioning them as unique lead compounds for further preclinical development and investigation.

2. Results and Discussion

2.1. Genome Mining of APD‐CLD Biosynthetic Pathways

The modular enzymatic machinery involved in APD‐CLD biosynthesis includes a characteristic NRPS‐PKS bimodule responsible for APD synthesis. Within this bimodule, the NRPS module features a serine‐specific adenylation (A) domain, while the PKS module contains an atypical KR‐DH didomain (Figures 1 and S1). The KR–DH configuration, which contrasts with the more common DH–KR configuration found in modular PKSs, may be linked to noncanonical enzymatic functions and novel chemistry. To survey the prevalence of the KR‐DH configuration in natural product biosynthesis, we analyzed a dataset of 220,645 DH sequences retrieved from microbial genomes. Our analysis revealed that the majority of DH domains are associated with the DH‐KR configuration (n = 186,303), while only a minority (n = 10,327) are associated with the KR‐DH configuration (Figure S2). Among the 10,327 hits, only 2,354 (23%) were found in hybrid PKS/NRPS systems for producing APD‐CLDs and other putative secondary metabolites with a hybrid polyketide‐peptide biosynthetic origin.

The 2,354 PKS/NRPS systems were grouped into seven distinct clades (Clades I to VII) based on the divergence of their KR‐DH sequences (Figure 2). Clades I to VII differ in the composition of their PKS and NRPS modules. Clade I includes systems closely related to those involved in APD‐CLD biosynthesis. In contrast, PKS/NRPS systems from Clades II to VII do not seem to contain the APD‐synthesizing NRPS(Ser)‐PKS bimodule, as their A domains are predicted to recognize amino acids other than l‐serine. The secondary metabolites produced by the NRPS/PKS systems in Clades II to VII remain unknown. Considering their atypical KR‐DH configuration and the potential for the DH domain to catalyze noncanonical reactions, it will be of considerable interest to identify their secondary metabolite products in the future.

FIGURE 2.

Genome mining of PKS/NRPS systems containing the atypical KR‐DH didomain. A phylogenetic tree of KR‐DH didomains derived from the 2,354 PKS/NRPS systems that contain the KR‐DH didomain is shown on the left. The tree is composed of seven clades, corresponding to PKS/NRPS systems with distinct domain organization. Clade I is associated with APD‐CLD BGCs. On the right, a BGC similarity network generated by BiG‐SCAPE for the BGCs associated with Clade I is presented. The network contains 309 nodes, with each node representing a BGC. StrR‐containing BGCs are highlighted in orange, and the three targeted BGCs are labeled. BGC3 was predicted to produce an APD‐CLD with an AA2 residue distinct from those found in microtermolides and vinylamycin. The predicted products of the BGCs are shown, with amino acids represented by their single‐letter codes and “X” indicating unknown residues.

The BGCs associated with the KR‐DH didomains in Clade I were further analyzed using the Biosynthetic Gene Similarity Clustering and Prospecting Engine (BiG‐SCAPE) [23]. The 309 BGCs were classified into eight subgroups (Figure 2) based on the identity and positioning of the two amino acid units AA1 and AA2. The majority (n = 285) are predicted to produce APD‐CLDs that contain adjacent AA1 and AA2, as observed in rakicidins, microtermolides, and boholamides. Only a small subset (n = 24) is predicted to produce BE‐43547‐type APD‐CLDs with AA1 and AA2 separated by the APD moiety. Among the 285 BGCs, three were found to contain a flavin‐dependent halogenase gene, suggesting they may synthesize halogenated APD‐CLDs (e.g., BGC1, Figure 2). Another interesting BGC (BGC 2) contains a gene that encodes a membrane‐bound glycosyltransferase, which is rare in natural product biosynthesis [24, 25], hinting at the potential production of an unprecedented glycosylated APD‐CLD. The bioinformatic analysis together revealed cryptic BGCs with the capacity to generate structurally novel natural products, including new APD‐CLDs. Targeting the APD‐CLD BGCs that encode halogenase and glycosyltransferase enzymes may yield compounds with distinctive bioactivities, as halogenation and glycosylation are well known to modulate the biological and pharmacokinetic properties of natural products.

2.2. Production of APD‐CLDs Through BGC Activation

Based on the genome mining results, we targeted the putative APD‐CLD BGCs from three microbial strains for compound discovery (Figures 2 and 3A). Wenjunlia vitaminophila and Amycolatopsis aidingensis harbor BGC1 and BGC2, which are predicted to produce halogenated and glycosylated APD‐CLDs, respectively. Meanwhile, Streptomyces tasikensis P46 contains BGC3 (Figures 2 and 3A), which is predicted to produce a microtermolide or vinylamycin‐like APD‐CLD with a distinct AA2 residue. To optimize the fermentation conditions for metabolite production, we conducted a systematic OSMAC (One Strain–Many Compounds) screening [26], testing 30 different agar and liquid media that varied in carbon, nitrogen, and inorganic salt composition (Table S1). Despite these efforts, none of the three strains produced the expected APD‐CLDs, suggesting that the BGCs may be inactive under the fermentation conditions.

FIGURE 3.

Discovery of novel APD‐CLDs through the activation of silent BGCs. (A) Comparison of the cma, ama, and mct BGCs that were predicted to produce novel APD‐CLDs (also see Tables S2–S4 for gene annotation). (B) HPLC chromatograms show the production of APD‐CLD upon activation by the overexpression of a transcriptional activator and supplementation with chemical elicitors and biosynthetic precursors. (C) Chemical structures of the newly discovered APD‐CLDs.

About half of the APD‐CLD BGCs, including those from W. vitaminophila and S. tasikensis sp. P46, encode a pathway‐specific StrR‐type transcriptional activator [27, 28, 29] (Figure 2). These strR‐type genes are typically positioned 500–3,000 bp upstream of the first pks gene (Figure 3A). The putative StrR‐binding motif (GTTCGACTG(N)₁₁CAGTCGAAC) is highly conserved across APD‐CLD BGCs, suggesting a shared regulatory mechanism (Figure S3). To exploit the pathway‐specific StrR regulator for BGC activation, we constructed two gene overexpression plasmids using a pSET152‐based conjugative–integrative shuttle vector [30], with the StrR‐type genes cma1 and mct1 placed under the control of the kasOp* promoter. We subsequently generated the overexpression mutants W. vitaminophila::cma1 and S. tasikensis::mct1 with the overexpression plasmids.

When we applied the OSMAC strategy to W. vitaminophila::cma1, we observed the production of new metabolites (1–3) under several fermentation conditions, with the highest titers achieved on GYM agar (Figure 3B, left panel). LC–MS analysis confirmed the presence of amino acid residues and a chlorine atom in 1–3, suggesting the activation of the APD‐CLD pathway. From GYM agar plates, we also detected new metabolites in the culture medium of S. tasikensis::mct1, although the initial fermentation titers were very low (Figure 3B, middle panel). While screening chemical elicitors to enhance metabolite production, we found that supplementing the overexpression mutant with Cu²⁺ (0.2 mM) significantly boosted the fermentation titers (Figure 3B, middle panel), enabling us to isolate compounds 4–10 for structural elucidation.

We could not apply the regulator overexpression strategy to A. aidingensis, as the ama BGC does not encode a StrR‐type regulator. To explore alternative activation strategies, we tested elicitors including metal ions, antibiotics, and other organic compounds. The addition of CuSO₄ (0.1 mM) induced the production of metabolites with characteristic APD‐CLD absorption spectra, though fermentation titers remained low. Considering that the ama BGC contains a glycosyltransferase gene (ama4), we hypothesized that supplementing the culture medium with a sugar precursor might enhance the production of glycosylated products. Screening media supplemented with various sugar sources revealed that the use of M5 agar (Table S1) containing glucose resulted in robust production of new compounds 11 and 12 (Figure 3B, right panel). ¹H NMR analysis confirmed the presence of both the APD warhead and a sugar moiety in these metabolites (Table S9). Notably, the production of 11 and 12 was not observed with glucose alone, highlighting the critical role of Cu²⁺ in eliciting metabolite production.

A major challenge in genome‐guided natural product discovery is the activation of silent BGCs, for which broadly applicable methods remain limited [31]. Our previous studies have demonstrated that successful activation of silent BGCs typically requires iterative, trial‐and‐error strategies involving diverse genetic and chemical interventions [32, 33, 34]. In this study, we achieved activation of the APD‐CLD biosynthetic pathways through a combination of gene overexpression, chemical elicitation, and precursor supplementation. The regulatory gene strR was found in over half of the APD‐CLD BGCs. Its overexpression may represent a generally applicable strategy for accessing other cryptic APD‐CLDs. Our studies also suggest that activation of gene expression alone is often insufficient to drive the production of certain secondary metabolites, owing to the scarcity of biosynthetic precursors or cofactors. For example, the biosynthesis of arabimalamides requires both Cu²⁺ and glucose. While glucose likely serves as the biosynthetic precursor for the α‐d‐arabinofuranose moiety as described below, the role of Cu²⁺ remains unclear; it may function as a cofactor for copper‐dependent enzymes or a stimulus of metal‐responsive regulatory pathways.

2.3. Structural Features of the New APD‐CLDs

Following the successful activation of the APD‐CLD pathways, we conducted large‐scale fermentations to produce sufficient quantities of metabolites for structural elucidation. The W. vitaminophila::cma1 strain yielded several structurally related APD‐CLDs, from which three metabolites (1–3) were isolated and purified. The most abundant compound, designated chloromalamide A (1), was identified as a 7″‐chloro‐tryptophan‐containing lipo‐cyclopeptide (Table S7 and Figure S4). Key HMBC correlations established the connectivity between Gly, 7″‐Cl‐Trp, the APD moiety, and the alkyl chain (Figure S4), confirming the planar structure of 1 (Figure 3C, left panel). Chloromalamides B (2) and C (3) are closely related congeners, differing from 1 by minor structural variations in the alkyl chain, either replacing the terminal isopropyl group with an ethyl group (2) or removing a methylene unit (3). The defining feature of 1–3 is the 7″‐chloro‐tryptophan residues, making chloromalamides the only halogenated APD‐CLDs reported to date.

From the S. tasikensis::mct1 strain, we isolated seven APD‐CLDs (compounds 4–10) that share the same cyclopeptide core with microtermolide A and vinylamycin (Figure 3B, middle panel; Table S8 and Figure S4) [35, 36]. Designated microtermolides C–I, compounds 4–10 differ from microtermolide A primarily by variations in the length of their alkyl side chains. While our initial sequence analysis of the A domains has predicted an amino acid other than l‐valine at the AA2 position, we still observed l‐valine at this position. This finding highlights the limitations of current bioinformatic tools in accurately predicting the substrate specificity of NRPS adenylation domains.

The A. aidingensis strain produced two new APD‐CLDs, arabimalamides A (11) and B (12). These compounds are molecular isomers with slightly different alkyl chains. HMBC and COSY analyses confirmed the connectivity of the two amino acid residues (proline and tyrosine), the APD moiety, and the polyketide‐derived alkyl chain (Figure S4). The NMR spectra also revealed diagnostic signals consistent with the presence of a furanose sugar moiety, with an anomeric carbon at δ _C 106.6 ppm and an anomeric proton at δ _H 5.37 ppm, along with two oxygenated methine carbons at δ _C 85.1 and 82.4 ppm (Table S9). Comparison with literature data identified the sugar as α‐arabinofuranose [37, 38]. HMBC correlations between H‐7a″ of the sugar and C‐7″ of the tyrosine residue confirmed the site of glycosylation at the tyrosine hydroxyl group (Figure 3B, right panel, Figure S4). Apart from the unique α‐arabinofuranose moiety, the presence of proline (AA1) and tyrosine (AA2) distinguishes arabimalamides from other APD‐CLDs. The incorporation of proline at the AA1 position, rather than the typical glycine or alanine, may introduce greater conformational rigidity to the cyclic scaffold, potentially influencing its biological activity.

The absolute configurations of the APD‐CLDs were determined using a combination of experimental and bioinformatic methods. Marfey's analysis [39] of the hydrolysates of 1 and 12 revealed the presence of 7″‐chloro‐d‐Trp in 1, and l‐Pro and d‐Tyr in 12 (Figure S6A). The d‐configuration of Trp and Tyr agrees with the epimerase domain in the final NRPS module of the respective PKS/NRPS systems, as discussed later. The relative configuration of the 2‴‐ethyl, 3‴‐hydroxy, and 4‴‐methyl substituents in the chloromalamide alkyl chain was predicted as 2‴R*,3‴R*,4‴S* based on J‐based configuration analysis [40, 41] and the NOESY correlations (Figures S4–S5). For the 3‴‐hydroxy group of chloromalamides, stereoselectivity analysis of the ketoreductase (KR) domain [42] predicted its configuration as “R” (Figure S6B&C). The absolute configuration of the alkyl chain in chloromalamides was thus assigned as 2‴R,3‴R,4‴S. Given the similar ¹H, ¹³C chemical shift of H/C 3‴‐4‴, the ³ J _{H2 α ‴‐H3‴}, ³ J _{H3‴‐H4‴} coupling constants, and the NOESY correlations (Tables S7 and S9, Figure S5) in both chloromalamides and arabimalamides, we propose that the 3‴‐hydroxy and 4‴‐methyl substituents in arabimalamides share the same configuration as those in chloromalamides (Figure 3C). The absolute configuration of the α‐arabinofuranose in arabimalamides was determined to be “d” by comparing the optical rotation ([α]_D ²⁰ = ‐117.0° c 0.075) of the hydrolyzed arabinose with standard d‐arabinose ([α]_D ²⁰ = ‐118°).

2.4. Biosynthesis of Chloromalamides and Arabimalamides

The biosynthetic pathways from compounds 1–12 share a hybrid PKS/NRPS assembly line composed of four PKS (modules 1–3 and 5) and three NRPS modules (modules 4, 6, and 7) (Figures 4 and 5). Consistent with the previously proposed biosynthetic mechanism for APD‐CLDs [13, 19, 43], the biosynthesis is likely initiated by a fatty acyl‐AMP ligase (FAAL) domain, which activates and transfers a fatty acid starter unit to the first ACP domain. The relaxed substrate specificity of the FAAL domains likely accounts for the structural diversity of the alkyl chains observed in compounds 1–12. The APD moiety is likely synthesized by modules 4 and 5, with the DH domain of module 5 putatively catalyzing two successive dehydration reactions to yield the characteristic diene functionality. Phylogenetic relationship analysis and structural modelling (Figure S7) suggest the cyclization and concurrent release of the intermediate from the PKS/NRPS assembly line is most likely mediated by the terminal condensation domain of Cma9 (Cma9‐C_T).

FIGURE 4.

Proposed biosynthetic mechanism for chloromalamides (A–C). Cma10 is a standalone KR protein and Cma4 is a tryptophan halogenase. (FAAL, fatty acyl‐AMP ligase; ACP, acyl carrier protein; KS, ketoacyl synthase; AT, acyltransferase; DH, dehydratase; KR, ketoreductase; ER, enoyl reductase; A, adenylation; C, condensation; E, epimerization; T, thiolation).

FIGURE 5.

Proposed biosynthetic mechanism for arabimalamides. Ama10 is a standalone KR; and Ama4 is a membrane glycosyltransferase. (FAAL, fatty acyl‐AMP ligase; ACP, acyl carrier protein; KS, ketoacyl synthase; AT, acyltransferase; DH, dehydratase; KR, ketoreductase; ER, enoyl reductase; A, adenylation; C, condensation; E, epimerization; T, thiolation; PRPP: β‐d‐phosphoribose diphosphate; DPR: decaprenylphosphoryl‐β‐d‐ribose; DPX: decaprenylphosphoryl‐β‐d‐2‐ketoribose; DPA: decaprenylphosphoryl‐β‐d‐arabinofuranose).

For chloromalamides, the chlorine atom is likely introduced onto l‐tryptophan by the flavin‐dependent halogenase Cma4. Structural modeling with AlphaFold3 [44] indicated that the substrate‐binding pocket of Cma4 can accommodate l‐tryptophan but not chloromalamide (Figure S8), suggesting the chlorination is likely to occur before cyclopeptide assembly. This hypothesis is further supported by the high sequence identity (73%) between Cma4 and the tryptophan halogenase PyrH [45]. An additional feature of chloromalamides is the 1‐hydroxyethyl substituent in the alkyl chain, which is likely derived from ethylmalonyl‐CoA [46]. This precursor could be synthesized by the crotonyl‐CoA reductase Cma3 [46], followed by hydroxylation by the cytochrome P450 monooxygenase Cma11 (Figure 3A, Figure 4, Table S2). Furthermore, an epimerase domain in the final NRPS module is likely responsible for generating the d‐configuration of the chloro‐d‐Trp moiety [47].

As the first glycosylated APD‐CLDs, the arabimalamides contain an α‐d‐arabinofuranose moiety whose biosynthetic origin is of considerable interest. The only candidate glycosyltransferase is encoded by the ama4 gene from the ama BGC (Figure 5). The ama4 gene encodes a membrane‐bound glycosyltransferase that belongs to the GT39 family (Figure S9) [24, 25]. This is surprising because the biosynthesis of glycosylated natural products typically relies on soluble glycosyltransferases that employ nucleotide‐activated sugars as sugar donors [48]. Enzymes in the GT39 family are typically involved in glycosylating proteins using decaprenyl‐phosphorylated sugar donors rather than nucleotide‐activated sugars [24]. To our knowledge, only one other enzyme from this family has been reported to be involved in natural product biosynthesis, where it installs arabinofuranose onto the aglycone of primycin A1–A3 [49]. Hence, although its function remains to be validated by biochemical characterization, Ama4 could represent a rare example of a membrane‐bound glycosyltransferase involved in a secondary biosynthetic pathway, utilizing decaprenyl‐phosphoryl‐β‐d‐arabinofuranose (DPA) as the sugar donor (Figure 5).

The ama BGC does not appear to contain genes for d‐arabinofuranose biosynthesis. d‐arabinofuranose is commonly found in the cell walls of bacteria such as mycobacteria, where it is part of the arabinogalactan and lipoarabinomannan components [50]. We identified five genes (araf1–5, Table S3) in the A. aidingensis genome that share sequence homology with the d‐arabinofuranose‐synthesizing genes from mycobacteria. These genes could form the arabinofuranose biosynthetic pathway in A. aidingensis, supporting both cell wall biogenesis and arabimalamide biosynthesis. We propose an arabinofuranose‐biosynthesizing pathway composed of five arabinofuranose‐synthesizing enzymes (Figure 5). Araf1 encodes a ribose‐phosphate pyrophosphokinase [51] that transfers pyrophosphate from ATP to the C‐1 position of d‐ribose‐5‐phosphate, generating PRPP (β‐d‐phosphoribose diphosphate). Araf2 functions as a decaprenol‐1‐phosphate:5‐phosphoribosyltransferase [52], attaching a decaprenyl group to PRPP. Araf3 acts as a 5'‐phosphoribosyl‐monophospho‐decaprenol phosphatase [50], likely catalyzing C‐5 dephosphorylation to produce decaprenyl‐phospho‐β‐d‐ribose (DPR). In mycobacteria, the decaprenyl‐phosphoribose epimerase, composed of DprE1 and DprE2 subunits, converts DPR into DPA [53]. Given the high sequence homology, we propose that Araf4 (DprE1 homolog) oxidizes DPR to a keto intermediate (DPX), which is then reduced by Araf5 (DprE2 homolog) to generate the sugar donor DPA.

2.5. Chloromalamides and Arabimalamides Exhibited Antiplasmodial Activity

To evaluate whether the newly identified APD‐CLDs exhibit hypoxia‐selective anticancer activity, we performed cytotoxicity assays under both normoxic and hypoxic conditions. Representative compounds 1, 3, 8, and 12 showed moderate cytotoxicity against four different types of cancer cells, with IC₅₀ values ranging from 1.8 (for 3) to 36.0 µM (for 12) under normoxia (Table S10). These compounds exhibit moderately (<3.6‐fold) or negligibly enhanced cytotoxicity in these cells under hypoxia, in contrast to rakicidins and BE‐43547s, which display up to 79‐fold enhanced activity at hypoxia [6, 13, 14]. When assessing their antimicrobial properties, compounds 1, 3, 8, and 12 showed only moderate activity against Gram‐positive bacteria (Staphylococcus aureus and Streptococcus pneumoniae), and negligible activity against Gram‐negative bacteria (Escherichia coli and Pseudomonas aeruginosa), as well as the fungal pathogens Candida albicans and Aspergillus fumigatus (data not shown).

In contrast, when we evaluated the antimalarial activity of the new APD‐CLDs by performing intraerythrocytic growth inhibition assays with the malarial parasite strains P. falciparum 3D7 (a drug‐sensitive strain) and P. falciparum Dd2 (a multidrug‐resistant strain), we found compounds 1–3 and 12 displayed potent activity against the 3D7 strain, with nanomolar IC₅₀ values ranging from 24.7 to 126.0 nM (Figures 6A and B). These compounds retained nanomolar potency against the Dd2 strain, with less than a twofold reduction in IC₅₀. Microtermolide G (8), by contrast, showed no significant activity in the 10–800 nM range. We found that compounds 1–3 induced hemolysis of uninfected red blood cells (RBC), with hemolytic concentrations (HC₅₀) between 256.5 and 361.4 nM (Figure S10). In contrast, the hemolysis caused by 12 occurred at higher concentrations with an HC₅₀ value of 5.0 µM (20‐h assay, Figure S11). We evaluated the cytotoxicity of the representative compounds in three normal human cell lines (NCM460, HK‐2, and HaCaT, Table S11). From the assays, we concluded that arabimalamide B (12) has the most favorable selectivity index of 164 (IC₅₀ NCM460 /IC₅₀ P. falciparum Dd2) or 31.1 (HC₅₀/IC₅₀ P. falciparum Dd2) as an antimalarial lead compound among 1–12, based on the anti‐plasmodial activity (IC₅₀ = 160.6 nM against Dd2 strain) and cytotoxicity toward NCM460 human cell line (IC₅₀ = 26.3 µM) or RBC (HC₅₀ = 5.0 µM).

FIGURE 6.

Effects of APD‐CLDs on the blood‐stage survival and intraerythrocytic life cycle of P. falciparum. (A) SYBR‐green‐based response assay of compounds 1, 2, 3, 8, and 12, with the positive control dihydroartemisinin (DHA) and chloroquine (CQ). (B) IC₅₀ values derived from the dose–response assays. (C) Illustrating diagram of the 96‐h life cycle of P. falciparum (Top) and Giemsa‐stained images of ring‐synchronized P. falciparum treated with arabimalamide B (12) at 100 and 200 nM. The digestive vacuole of the parasite is marked by the dashed circles. The brown‐colored hemozoin is visible in the food vacuoles. The written time corresponds to the time taken after adding arabimalamide B (12) to the parasite culture. (D) Percentage parasitemia across 96 h after exposure to different concentrations of compound 12 at the ring stage. Negative control (DMSO) and positive controls (DHA and CQ) are also shown.

The IC₅₀ values of chloromalamides are comparable to those of established antimalarials such as chloroquine and some newly discovered antimalarial drug leads [54, 55, 56]. The presence of a chlorine atom in chloromalamides may contribute to their bioactivity, as halogenation is known to enhance drug–target interactions and improve pharmacokinetic properties [57]. In fact, approximately 25% of marketed drugs, including chloroquine, contain halogen substituents that confer certain advantages. A significant impediment to the therapeutic potential of chloromalamides is their hemolytic activity, with the hemolytic concentration ranging from 257 to 361 nM. To develop chloromalamides as viable antimalarial leads, structural modifications aimed at reducing hemolytic toxicity will be critical. Interestingly, microtermolides, the APD‐CLDs that differ from chloromalamides only at the AA1 (Gly vs. Ala) and AA2 (Val vs. Cl‐Trp) positions, induced hemolysis at much higher concentrations (Figure S11). This suggests that the hemolytic activity of APD‐CLDs is strongly influenced by the identity of the two amino acid residues, particularly AA2. We also observed that microtermolides C–F (4, 7, and 8) did not exhibit detectable hemolytic activity at concentrations up to 5.0 µM, whereas microtermolide I (10), which contains a longer alkyl chain, induced hemolysis with an HC₅₀ of 3.4 µM. These observations suggest that hemolytic activity increases with alkyl chain length. Hence, future optimization efforts should focus on modifying the AA2 residue and shortening the alkyl chain to mitigate membrane‐disruptive effects and reduce hemolytic liability while preserving or even improving antiplasmodial efficacy.

2.6. Arabimalamide B Induced Digestive Vacuole Enlargement and Hemozoin Dispersion in P. falciparum

Among the newly identified APD‐CLDs, the arabimalamides emerged as the most promising antimalarial leads with high selectivity indices. To further understand the impact of arabimalamide B (12) on the intraerythrocytic development of P. falciparum, we examined phenotypic changes induced by arabimalamide B during the parasite's asexual blood stage cycle. In the assays, the parasites were first synchronized to the ring stage. In the absence of arabimalamide B or positive control dihydroartemisinin (DHA) and chloroquine (CQ), the parasites developed into the trophozoite stage, the most metabolically active phase, characterized by hemozoin production within the digestive vacuole [58]. The parasites then progressed to the schizont stage, where they became segmented and multinucleated before releasing merozoites to invade new RBCs, thereby initiating a new intraerythrocytic development cycle (IDC) (Figure 6C). Introduction of arabimalamide B at the ring stage profoundly impacted the parasite development and survival, particularly from the second IDC onward. Most notably, arabimalamide B significantly altered the morphology of the digestive vacuole and the distribution of hemozoin during the trophozoite and schizont stages. Hemozoin, an insoluble brown crystalline pigment, is formed through the polymerization of toxic heme, liberated during hemoglobin digestion, as part of a critical heme detoxification mechanism for the parasite [59]. In the treated parasites, digestive vacuoles appeared markedly enlarged, and hemozoin was more diffusely distributed compared to controls.

In addition to the impact on the digestive vacuole, arabimalamide B also delayed or halted parasite development in a dosage‐dependent manner. At 100 nM, parasite progression was slowed, reaching only the schizont stage by 86–96 h after arabimalamide B was introduced, whereas the control parasites had already completed reinvasion and returned to the ring stage by 86 h (Figure 6C). At 200 nM, the parasites exhibited abnormal vacuole morphology and dispersed hemozoin early in the first IDC (23–43 h) and subsequently collapsed and died by 62 h, failing to transition into the next IDC. These findings suggest that arabimalamide B disrupts essential developmental processes in P. falciparum, potentially through interference with the digestive vacuole biogenesis and heme detoxification.

The phenotypic effect of arabimalamide B on P. falciparum resembles those induced by the hemin‐binding quinoline‐type antimalarials such as chloroquine, suggesting that arabimalamide B may inhibit hemozoin formation by interfering with the parasite's heme detoxification pathway [59, 60]. The disruption of digestive vacuole integrity and impaired hemozoin biogenesis likely compromise schizont maturation and reduce merozoite yield, thereby contributing to the observed reduction in reinvasion efficiency. Furthermore, arabimalamide B exhibited comparable antiplasmodial activity against both chloroquine‐sensitive and ‐resistant P. falciparum strains, suggesting a lack of cross‐resistance. This indicates that arabimalamide B could not be efficiently exported by the P. falciparum Chloroquine Resistance Transporter (PfCRT) [61], a protein located on the digestive vacuole membrane that mediates the efflux of chloroquine and other quinoline drugs, contributing to resistance. Given the distinct mode of action and activity against the resistant strain, arabimalamides and their structural derivatives hold potential as replacements for chloroquine or other quinoline‐based drugs in artemisinin‐based combination therapies for the treatment of drug‐resistant malaria.

We further quantified the antiplasmodial effect of arabimalamide B by measuring parasitemia, the percentage of RBCs infected by P. falciparum, across different concentrations (Figure 6D). Arabimalamide B demonstrated a strong effect on parasite load, as parasitemia dropped below 0.5% at 200 nM by the end of the 96 h time course. The reduction in parasitemia at 800 nM exceeded those observed for 200 nM DHA and 1 µM CQ, two clinically important antimalarials. From the parasitemia, we also calculated the parasitized erythrocyte multiplication rate (PEMR) [62], defined as the number of ring‐stage infected RBCs per 1000 RBCs after schizogony divided by the number of infected RBCs per 1000 before schizogony. Arabimalamide B inhibited PEMR in a dose‐dependent manner during the initial invasion phase (23–38 h posttreatment), as detailed in Table S12. Notably, 100 nM of arabimalamide B reduced the second invasion phase (72–86 h posttreatment) by approximately 80% compared to the control, while 200 nM was sufficient to eliminate P. falciparum during the second IDC. Given that the 23–38 h and 72–86 h windows correspond to critical periods of merozoite reinvasion into RBCs, these results suggest that arabimalamide B lowered the parasite's reinvasion efficiency, likely by interfering with pathways essential for host cell entry.

To further evaluate the potential of APD‐CLDs as schizonticidal lead compounds, it is essential to generate structural analogs for SAR studies and to identify their cellular targets. Synthetic routes for APD‐CLDs have been established, enabling the chemical synthesis of structural analogs [9, 10, 11, 13, 14, 15, 16, 17, 63, 64, 65, 66, 67]. Alternatively, with recent advances in NRPS and PKS engineering [68, 69, 70], it is now feasible to engineer NRPS and PKS assembly lines through domain or module swapping to produce novel APD‐CLD analogs. We envision that one or both of the final two NRPS modules can be replaced to create new APD‐CLDs containing different amino acids at the AA1 and AA2 positions. The substitution of AA1 and AA2, or the shortening of the alkyl chain, can potentially further reduce the cytotoxicity. The hydrophilic arabinofuranose moiety likely plays a significant role in improving the water solubility and binding specificity of arabimalamides toward their protein targets. The arabinofuranose moiety can also be replaced by other sugar moieties using promiscuous glycosyltransferases to modulate the protein‐drug interaction and pharmacokinetic properties. Meanwhile, the molecular targets and mechanisms of action of chloromalamides and arabimalamides still remain unknown. The observed impact on digestive vacuole morphology and hemozoin formation suggests that arabimalamides may interfere with proteins or pathways involved in digestive vacuole biogenesis or heme detoxification. Given the presence of the electrophilic APD warhead, these compounds are likely to act as irreversible inhibitors that form covalent complexes with their target proteins [18]. To elucidate the mechanism of action, target identification strategies such as the use of affinity‐tagged probes [71] or cellular thermal shift assay (CETSA) [72] will be valuable for probing parasite protein targets. The insights gained from the SAR and target identification studies will play a critical role in driving lead optimization through rational design.

3. Conclusion

We have expanded the APD‐CLD family through genome mining and activation of silent BGCs. Among the newly identified members, chloromalamides and arabimalamides represent the first halogenated and glycosylated APD‐CLDs, respectively, thereby broadening the structural diversity of this natural product class. Chloromalamides and arabimalamides demonstrated potent antiplasmodial activity against P. falciparum, including strains resistant to existing antimalarial drugs. Mechanistic studies revealed that these compounds disrupt hemozoin formation during the trophozoite and schizont stages, and significantly reduce reinvasion efficiency. These phenotypic effects are reminiscent of those observed with quinoline‐class antimalarials, yet arabimalamides retain efficacy against chloroquine‐resistant strains, highlighting their potential to overcome existing resistance mechanisms. These findings position the APD‐CLDs, particularly the arabimalamides and their future structural analogs, as lead compounds for the development of new antimalarial therapies. Continued efforts in structural optimization, target identification, and validation of in vivo efficacy and toxicity will reveal the full potential of the compounds as antimalarial leads.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: anie71875‐sup‐0001‐SuppMat.Docx.

Experimental methods, spectroscopy, and DNA data are provided in the supporting information. The authors have cited additional references within the Supporting Information [25, 27, 35, 36, 39, 40, 41, 42, 44, 45, 47].

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.