Integrated omics approach to unveil antifungal bacterial polyynes as acetyl-CoA acetyltransferase inhibitors

Bacterial polyynes are highly active natural products with a broad spectrum of antimicrobial activities. However, their detailed mechanism of action remains unclear. By integrating comparative genomics, transcriptomics, functional genetics, and metabolomics analysis, we identified a unique polyyne resistance gene, masL (encoding acetyl-CoA acetyltransferase), in the biosynthesis gene cluster of antifungal polyynes (massilin A 1, massilin B 2, collimonin C 3, and collimonin D 4) of Massilia sp. YMA4. Crystallographic analysis indicated that bacterial polyynes serve as covalent inhibitors of acetyl-CoA acetyltransferase. Moreover, we confirmed that the bacterial polyynes disrupted cell membrane integrity and inhibited the cell viability of Candida albicans by targeting ERG10, the homolog of MasL. Thus, this study demonstrated that acetyl-CoA acetyltransferase is a potential target for developing antifungal agents.

P olyynes or polyacetylenes contain a conformationally rigid rod-like architecture and an electron-rich consecutive acetylene moiety. Hundreds of polyynes have been discovered, out of which compounds have mainly been isolated from terrestrial plants and marine organisms 1,2 . In contrast to polyynes from plant sources, bacterial polyynes contain a distinct terminal alkyne with conjugated systems, which causes bacterial polyynes to be more unstable. This instability has discouraged surveys of bacterial polyynes using a bioactivity-guided isolation approach. Nevertheless, more than ten bacterial polyynes have been recorded in a few species. Notably, most polyynes have been reported to have a broad spectrum of antimicrobial effects. For instance, cepacins, isolated from Pseudomonas cepacian SC 11783 (taxonomically reclassified as a Burkholderia diffusa), was reported to have antibacterial activity against the majority Gram-negative bacteria and Staphylococci spp., and anti-oomycetal activity against Pythium ultimum [1][2][3][4] . Caryoynencin, isolated from Pseudomonas caryophylli (taxonomically reclassified as Burkholderia caryophylli), was reported to have antibacterial activity against Escherichia coli, Klebsiella pneumoniae, Staphylococci aureus, and Bacillus subtilis 5 . In addition, collimonins isolated from Collimonas fungivorans Ter331 6,7 and Sch 31828 isolated from Microbispora sp. SCC1438 8 were reported to have antifungal activity against Aspergillus niger and Candida spp., respectively.
Here, we used a multi-omic approach and genetic engineering to identify the bioactive polyynes of Massilia sp. YMA4, characterized their biosynthesis gene cluster (BGC) named massilin (mas) BGC, and delineated the antifungal mechanism of bacterial polyynes. Furthermore, heterologous co-expression of mas BGC further illustrated the conserved genes critical to the electron-rich consecutive alkyne moiety and an additional gene responsible for structure derivatives modification of the polyynes. Furthermore, a comparative genomic analysis uncovered the self-resistance gene and identified bacterial polyynes as antifungal agents that target the first enzyme of ergosterol biosynthesis, acetyl-CoA acetyltransferase. Finally, crystallographic analysis unveiled the detailed binding model of polyynes to the acetyl-CoA acetyltransferase.

Results and discussion
Transcriptomics analysis reveals polyynes as antifungal agents and their encoding BGC in Massilia sp. YMA4. An antagonism assay of Massilia sp. YMA4 against C. albicans revealed a distinct phenotype showing that the antifungal metabolites were produced in potato dextrose agar (PDA) medium but not in yeast malt agar (YMA) medium ( Supplementary Fig. 1). In addition, the antifungal metabolites were unstable in the extract and laborious to purify for bioassays using the classic bioactivityguided isolation approach, which includes purification with click chemistry, exclusion of oxygen and light, or careful optimization 7,12,16 . Therefore, to reveal the antifungal metabolites and their BGCs, we used a combined transcriptomics and metabolomics approach to explore the compounds produced in the two different media (PDA and YMA). First, the BGCs in the genome of Massilia sp. YMA4 were characterized via DeepBGC 17 with default settings and integrated with the criteria: DeepBGC score > 0.7 to characterize 13 BGCs in Massilia sp. YMA4 (Supplementary Data 1). Then, the differentially expressed genes (192 upregulated and 226 downregulated with P-value < 0.05 and |fold change | > 2) from transcriptomics analysis of PDA vs. YMA cultured cells were mapped into the 13 characterized BGCs. Among them, only one BGC was upregulated in PDA compared to YMA (Supplementary Data 1), and we named this gene cluster massilin (mas) BGC with 12 transcribed genes (masA to masL) (Fig. 1a).
To characterize the metabolites produced from mas BGC, we constructed mutant strains through insertion mutation at the masH (encoding fatty acid desaturase) gene locus in Massilia sp. YMA4 ( Supplementary Fig. 2). The result of the antagonism assay illustrated that YMA4::masH lost antifungal activity against C. albicans ( Supplementary Fig. 3), which demonstrated that mas BGC was crucial for the antifungal activity of Massilia sp. YMA4. Furthermore, a study of polyyne BGC (pgn BGC) in P. protegens had determined that the mutation of homolog (pgnH) of masH and masI in pgn BGC can effectively disrupt the production of protegencin (also known as protegenin A) 12 . We then conducted mass spectrometry-based untargeted metabolomics to analyze the differential features in wild type and biosynthesis-deficient mutant strain YMA4::masH. Via target isolation of differential features, we purified four polyynes (collimonin C/D 1, 2, and massilin A/B 3, 4) from ethyl acetate extract of Massilia sp. YMA4 (Fig. 1b, c). Their structures were elucidated using high- Among the four polyynes identified from Massilia sp. YMA4, collimonin C 1 and collimonin D 2 were initially isolated from C. fungivorans Ter331 7 . A new compound with an ene-triyne moiety was named massilin A 3 and identified as a racemate at C(6)OH of an unsaturated hexadecanoic acid. Another new ene-diyne-ene compound, massilin B 4, was presumed to be the precursor of collimonin C 1 or collimonin D 2. Notably, massilin B 4 was more chemically stable than other polyynes with a terminal alkyne.
Phylogenetic analysis of polyyne BGCs and mas heterologous co-expression revealed the core components for polyyne biosynthesis. The in-silico mining and comparative studies of polyynes/ terminal alkyne containing BGCs indicated that jamABC homologs (fatty acyl-AMP ligase (FAAL), fatty acid desaturase (FAD), and acyl carrier protein (ACP)) in jamaicamides (jam) BGC would be the minimal module for polyynes biosynthesis 18,19 . In addition, the gene cluster K in C. fungivorans Ter331 (collimonin BGC, which is abbreviated to col BGC) recruited not only FAAL, FAD, and ACP but also hydrolase/ thioesterase (H/TE) and rubredoxin (Rd) 6,7 . Therefore, we further explored the architecture of 56 bacterial polyynes BGCs with mas BGC as a reference to confirm the conserved composition of polyyne biosynthetic associated enzymes. It contains the FAAL-2x FAD-ACP-FAD-H/TE-Rd ( Supplementary Fig. 23).
To complement the mutagenesis results, we further established an E. coli heterologous co-expression system of mas genes to determine whether the genes participate in polyyne biosynthesis. The engineered E. coli mas-(heterologous co-expression with masD to masJ) produced a relatively nonpolar polyyne (Fig. 2a). This new ene-triyne polyyne was named massilin C 5 (Supplementary Table 5 and Supplementary Figs. [24][25][26][27][28]. The results demonstrated that masD to masJ are the core genes in the biosynthesis of the ene-triyne moiety. On the other hand, the expression of masB in Massilia sp. YMA4 was upregulated when cultured in PDA medium (Supplementary Data 1), like the core genes masD to masJ. Therefore, we proposed masB, a dioxygenase gene, as a candidate for the hydroxylation of polyynes since cytochrome P450s or dioxygenases can execute similar reactions 20,21 . The masB incorporated strain, mas + , produced collimonin C/D 1, 2 ( Fig. 2b and Supplementary Fig. 24), which suggested massilin C 5 is a biosynthetic intermediate and further hydroxylated by MasB to generate polyynes 1-4. Furthermore, the absence of masJ significantly decreased the accumulation of polyynes in mas expressed E. coli (mas-' and mas + ', Fig. 2), indicating the crucial role of rubredoxin in polyyne biosynthesis. As an electron transporter, rubredoxin can shuttle reducing molecules from NAD(P)H to membrane-bound alkane hydroxylases 22 , which is likely crucial to activation/regeneration of acetylenase/desaturase for the poly-dehydrogenation process in polyyne biosynthesis.
MasL serves as a polyyne direct target with a protective function. "Antibiotic resistome" refers to the notion that not only antibiotic resistance genes present in resistant pathogens but also those in the antibiotic producers often harbor resistance genes for self-protection 23,24 . Sometimes the resistance genes are located within the BGC of antibiotics 25 . To reveal the target protein of the bacterial polyynes of interest, we adopted the concept of the antibiotic resistome. We conducted phylogenetic analysis of bacterial polyyne BGCs and classified the BGCs into two chemotaxonomic groups: a palmitate-derived (C16) and a stearatederived (C18) group ( Fig. 3a and Supplementary Fig. 29). In the palmitate-derived group BGCs, cepacin (ccn) BGC branched out before the most recent common ancestor of mas/col BGCs.
Interestingly, the gene encoding the major facilitator superfamily (MFS) transporter, implicated in multidrug resistance and transports of small molecules and xenobiotics 26 , is preserved in col BGC but lost in mas BGC (Fig. 3b, Supplementary Table 6, and Supplementary Data 2). This implies that Massilia sp. YMA4 has an alternative protection mechanism for self-protection from polyynes. Notably, a conspicuous gene masL, an acetyl-CoA acetyltransferase gene, remains in mas BGC and ancestor ccn BGC but is absent in col BGC. We, therefore, proposed that masL is a potential self-resistance gene of polyynes.
To evaluate the protective effect of masL, we first heterologously expressed masL in polyyne-sensitive C. albicans. The expression of masL rescued fungal cell viability from polyyne inhibition under minimum inhibitory concentration ( Fig. 4 and Supplementary Table 7). Furthermore, an in vitro acetyl-CoA acetyltransferase (MasL) inhibition assay showed that polyynes (collimonin C/D 1, 2, and massilin A 3) inhibited the MasL enzyme activity (Table 1 and Supplementary Fig. 30). Since the drug target identification could be achieved by gene overexpression for gain-of-function of resistance 27 , we suggest masL serves as a self-resistance gene in the mas BGC, and the acetyl-CoA acetyltransferase it encodes is the target of collimonin C/D 1, 2 and massilin A 3. This targeting manner might explain why the growth of mas expressed E. coli is unaffected when producing polyynes without masL because E. coli has an acetyl-CoA acetyltransferase-independent route, the methylerythritol phosphate (also called deoxyxylulose 5-phosphate) pathway, to produce isoprenoid 28,29 . Moreover, ccn BGC divided into mas BGC and col BGC independently, and each contains a different self-protection mechanism, which implies a deletion event after horizontal gene transfer 30 .
Polyynes are covalent inhibitors of acetyl-CoA acetyltransferase. To investigate how bacterial polyynes target acetyl-CoA acetyltransferase, we solved the crystal structures of MasL in its apo and two holo (bound collimonin C 1 and collimonin D 2) forms at 1.78 Å, 1.66 Å, and 1.40 Å resolution, respectively ( Table 2). The asymmetric unit (space group P1 for apo MasL and MasL-collimonin D complex; P2 1 for MasL-collimonin C complex) of both structures contains a tetramer of the protein (Fig. 5, Table 2 and Supplementary Fig. 31 Fig. 32). The N-and C-domains form a typical five-layered fold (α-β-α-β-α) as observed in the structures of the other type II biosynthetic thiolases including Zoogloea ramigera PhaA 31 , Clostridium acetobutylicum CEA_G2880 32 , and Aspergillus fumigatus ERG10A 33 . The L-domain displays an α/β fold with a tetramerization loop associated with the C-domain.
Many high-resolution atomic structural models of the substrate-binding complex revealed the Claisen condensation reaction and binding model within the reaction pocket of acetyl-CoA acetyltransferases. In apo and holo forms of MasL, the substrate-binding pocket was located on the enzyme's surface facing the opposite dimer of the tetrameric assembly. The pocket was a tunnel shape of~10 Å depth with~6-8 Å diameter for the linear pantothenic moiety of coenzyme A (CoA) extending through the reactive center. The reactive center in MasL contained reactive cysteine residues Cys90 and nucleophilic activation residues His350 and Cys380. In the MasLcollimonin C complex, the conjugated polyyne tail extended into the MasL substrate binding site and formed a covalent bond between the terminal carbon (C16) of the alkyne moiety and the reactive cysteine sulfhydryl moiety of Cys90 via nucleophilic addition (Supplementary Figs. 33 and 34). Thus, the terminal alkyne was essential for polyyne activity through covalent modification on the reactive cysteine sulfhydryl group of acetyl-CoA acetyltransferases.
Although we did not obtain a good X-ray diffraction result surrounding the reactive Cys90 in the MasL-collimonin D complex, the nucleophilic addition mechanism of polyynes to the acetyl-CoA acetyltransferase was confirmed by mass spectrometry analysis (Supplementary Fig. 35 and Supplementary Data 3). The observation of collimonin C/D (+C 16 H 18 O 4 )-derived adducts of enzyme reactive cysteine residue provided convincing evidence of protein irreversible S-alkylation via nucleophilic addition to the conjugated terminal alkyne. Furthermore, the observation of massilin A (+C 16 H 18 O 3 )-derived adducts suggest that massilin A 3 would also covalently inhibit acetyl-CoA acetyltransferase activity. In the further comparative analysis of the polar interactions in the activation pocket of MasL in MasL-collimonin C and MasLcollimonin D complex, the C(7)OH moiety of collimonin C/D 1, 2, His158 of MasL, and a water molecule formed a strong polar interaction network, including a direct hydrogen bond (3.00-3.16 Å) and a water-mediated hydrogen bond between C(7) OH and His158 (Fig. 5). Furthermore, although there was no significant induced-fit within the pocket, the collimonin C/D 1, 2 caused the Arg135 on the tetramerization loop to swap and form a hydrogen bond interaction and salt bridge, respectively, across the two subunits within the binding site (Fig. 5). These polar interacting residues for the collimonin C/D 1, 2 (inhibitor) binding were similar to CoA (product) in other thiolase models 31 . Therefore, the acetyl-CoA acetyltransferase inhibitions by collimonin C/D 1, 2 are competitive binding models against substrate acetyl-CoA.
The stereochemistry of hydroxyl groups on polyynes affects their binding affinity toward acetyl-CoA acetyltransferase. The C6 chiral center of collimonin C/D 1, 2 is critical to forming different polar interactions. The C(6)OH of collimonin C 1 had  more flexibility concerning the spatial direction (with a dihedral angle to C(7)OH from 109˚to 170˚) and built a sophisticated polar interaction network with the amide of Pro249 in the pantetheine loop, Arg135, and multiple water molecules ( Fig. 5 and Supplementary Fig. 36). In contrast, the C(6)OH of collimonin D 2 did not form a polar interaction network at the contrasting face and showed fixed spatial direction (dihedral angle to C(7)OH from 32.9˚to 42.6˚) (Supplementary Fig. 36). The polar interaction network derived from C(6)OH of collimonin C 1 further affected the distal carboxyl group binding model. Furthermore, the reduced binding strength from the salt bridge (in MasLcollimonin D complex) to the hydrogen bond (in MasLcollimonin C complex) corresponded to their inhibition kinetic, that collimonin D 2 has a lower K I (42.84 μM) than collimonin C 1 (297.10 μM) ( Table 1). The above information suggests that the stereochemistry of the hydroxyl group on polyynes is vital for initial non-covalent complex affinity. This salt bridge/hydrogen bond formation will dominate the MasL binding affinity in dihydro-type polyynes 1 and 2.   Candida albicans ERG10 is the antifungal target of polyynes.

masD E F G H I masD E F G H I J D E F G H I masB
As the bacterial polyynes target acetyl-CoA acetyltransferase MasL, we further evaluated the inhibition effect of bacterial polyynes on ERG10, an acetyl-CoA acetyltransferase homolog in C. albicans. ERG10 is the first enzyme of the ergosterol biosynthesis pathway and is crucial for fungal cell membrane formation. The enzyme activity of ERG10 was inhibited by polyynes 1-3 ( Supplementary  Fig. 37), and correspondingly the overexpression of ERG10 in C. albicans also rescued the cell viability from their inhibition (Fig. 4).
Although we did not obtain the X-ray crystallography for the polyyne-ERG10 complex, the mass spectrometry analysis showed that polyynes 1-3 target the reactive Cys90 of ERG10 (Supplementary Fig. 38). A previous study reported that the homologous gene in Aspergillus fumigatus, erg10A is essential for survival and reduced expression leads to increased susceptibility to oxidative and cell wall-perturbing agents 33 . In the TEM analysis, we also observed that bacterial polyynes disrupted the cell membrane integrity in C. albicans (Supplementary Fig. 39). This result corresponded to the earlier finding that protegencin (or protegenin A) permeated algal cells and caused cell lysis 14 . Notably, massilin C 5, the analog of protegencin (or protegenin A), possessed antifungal activity against C. albicans (MIC 12.22 μM, Supplementary Fig. 22) and a comparable ERG10 inhibitory potency to polyynes 1-3 ( Supplementary  Fig. 37). However, neither the heterologous expression of MasL nor ERG10 in C. albicans was able to recover the cell viability under massilin C 5 treatment (Fig. 4). Together these results suggested that the polyynes without a hydroxy moiety would possess multiple antifungal targets.

Conclusion
The practical application of the antibiotic resistome concept with in-silico genome mining and comparative genetics analysis facilitated drug target identification. Our integrated approaches revealed that bacterial polyynes inhibit fungal growth by disrupting ergosterol biosynthesis. The crystallography analysis revealed that bacterial polyynes inhibit acetyl-CoA acetyltransferase via irreversible S-alkylation of the reactive residue with the functional terminal alkyne. In addition, the hydroxyl moiety stereochemistry affects the inhibitor affinity. We characterized the core biosynthesis module and modification genes for polyynes through genetic engineering. These findings will aid in the understanding of the structure-activity relationships of the polyynes and their production via microbial fermentation.   Massilia sp. YMA4 was identified for phylogenetic tree construction. The analysis was completed by using MEGA 10 with default parameters 35 . iTOL was used to present the results of phylogenetic analysis 36 . The RNA sequencing and transcriptome analysis were performed using Illumina MiSeq system (Illumina, USA) and CLC genomics workbench (version 11, CLC bio, Denmark), and the 92.89-98.11% reads were successfully mapped to the Massilia sp. YMA4 genome (GenBank assembly accession: GCA_003293715.1) for RPKM quantification (see Supplementary Methods and Supplementary Data 1 for details). The differential gene expression analysis was analyzed through CLC software with default pipeline and settings. Identification of differentially expressed genes (DEGs) with the expression (|Fold change (FC) | ≥ 2 with P-value < 0.05) was based on RPKM and analyzed using the Empirical analysis method 37 . Two biologic replicates of each condition were analyzed.

Methods
Construction of mas co-expression platform in Escherichia coli. Individual mas genes were cloned to the pET-28b(+) (Merck, Germany) using restriction enzyme digestion and ligation (see Supplementary Methods and Supplementary  Tables 8-10 for details). To obtain functional masH in the heterologous E. coli system, we recruited native and codon-optimized masH and the homolog (Bv4687) from the palmitate-derived (C16) species B. vietnamiensis LMG 10929 (see Supplementary Fig. 23, 29 and 40) into modular co-expression. The backbone of the constructed modules, including pET-22a(+), pACYCduet, and pRSFduet, were purchased from Merck (Germany). However, polyyne accumulations were observed only in combination with Bv4687. Thus, masH was replaced by Bv4687 in further co-expression assay. The artificial gene clusters were then subcloned to the destination vector and resulted in three modules for better flexibility of adjusting composition, including pET-22a(+)-masD-masE-masF-masG, pACYCduet-Bv4687 (masH homolog from B. vietnamiensis LMG 10929, Genbank accession number WP_011881350)-masI-masJ, and pRSFduet-masB. Two or three modules were co-transformed to E. coli strain C41 (Yeastern Biotech, Taiwan), and selected colonies were then inoculated in LB medium with suitable antibiotics. The expression was induced by adding isopropyl-β-D-thiogalactopyranoside (IPTG) and cultured for two days. The extractions were carried out by adding equal volumes of EtOAc and sonicated for 30 min. Finally, the extracts were concentrated and transferred to DMSO for further analysis.
Minimum inhibitory concentration determination and genetic rescue assay. The minimum inhibitory concentration (MIC) measurement was modified from R J Lambert's method 38 . Different concentrations of collimonin C 1, collimonin D 2, massilin A 3, and amphotericin B were prepared in YPD. The C. albicans cell viabilities were seeded with initial OD 600 0.05 and incubated at 37˚C. After 24 h incubation, the final OD 600 was recorded by Epoch 2 Microplate (BioTek Instruments, USA) for MIC calculation. The MIC of polyynes was built with cell viability (%) of different concentrations, fitting into the modified Gompertz function 38 .
For the genetic rescue assay, the ERG10 overexpression and masL heterologous expression strains (Supplementary Fig. 41) were seeded with OD 600 0.05 in YPD treated with MIC of each polyyne at 37˚C and supplied with 40 µg/mL doxycycline for gene expression. The C. albicans cell viabilities were recorded at 24 h. The statistical results were analyzed using GraphPad Prism (Version 8, GraphPad Software, USA) with multiple t-test analyses (FDR < 0.05).
Enzymatic inhibition assays. The enzymatic inhibition assay of acetyl-CoA acetyltransferase was conducted using a modified method 39 (see Supplementary Methods for details). The releasing CoA was monitored using a fluorescent probe. The fluorescence intensity of each experiment was obtained by subtracting the fluorescence intensity of the polyyne-free reaction. The measurement and calculation of polyynes inhibition kinetic refer to the previous covalent inhibitor model 40 . The progress curves used the hyperbolic regression model to calculate the kinetic inhibition parameters using GraphPad Prism (Version 8, GraphPad Software, USA).
Protein expression and crystallization. The recombinant MasL protein was purified from E. coli cells (see Supplementary Methods for details). For MasLcollimonins complex, 20 µM MasL was incubated with 100 µM collimonin C 1 or D 2 for further purification and crystallization. A freshly thawed aliquot of MasL, MasL-collimonin C complex, and MasL-collimonin D complex was concentrated to 20 mg/mL for an initial crystallization screening of ca. 500 conditions (Academia Sinica Protein Clinic, Academia Sinica). X-ray diffraction experiments were conducted by the National Synchrotron Radiation Research Center (Hsinchu, Taiwan) at the TLS beamline 15 A or the TPS beamline 05 A. The MasL, MasL-collimonin C, and MasL-collimonin D complex structures were solved by the molecular replacement method using the structure of thiolase from Clostridium acetobutylicum (pdb ID 4N44) as the search model. The molecular figures were produced with Maestro (Schrödinger Release 2021-1: Maestro, Schrödinger, USA).
Statistics and reproducibility. For Fig. 4, the cell viabilities were normalized to the mock treatment. The standard deviation was calculated based on three replicates and the two-tailed Student's t-test for statistical analysis. The experiments in this study, including mas co-expression in E. coli, minimum inhibitory concentration determination, genetic rescue assay, and enzymatic inhibition assay, were performed in three biological replicates (N = 3).