Analyzing the biosynthetic potential of antimicrobial‐producing actinobacteria originating from Indonesia

We investigated the biosynthetic potential of soil‐associated actinobacteria originating from Indonesia, identified as Streptomyces luridus and as Streptomyces luteosporeus . Antimicrobial assays indicated inhibitory activity by both strains against the pathogen Pseudomonas aeruginosa , with S. luteosporeus particularly inhibiting the growth of Bacillus subtilis . PCR‐amplification, cloning, and sequencing of ketosynthase (KS) domains of type I modular polyketide (PKS‐I) and adenylation (AD) domains of non‐ribosomal peptide synthetase (NRPS) indicated the diversity of KS and AD domains derived from both Indonesian Streptomyces . Further phylogenetic analysis showed that KS domains from the subclass cis‐AT PKS can be classified as being a part of a loading module or an extension module, along with their predicted substrate specificity. The results suggest that both strains are a potential source of novel biosynthetic pathways. This genetic anal‐ ysis approach can be used as a fast guide to obtain insight into natural product biosynthetic gene diversity in microorganisms.


Introduction
Phylum Actinobacteria have long been recognized as the major source of compounds representing 45% of all bioac tive microbial metabolites discovered so far (Bérdy 2005). Many actinobacterial natural products belong to polyke tides and nonribosomal peptides, two large classes of natural products with various potent biological activities, such as antitumor, antivirus, antibiotics, and immune suppressants (Cane and Walsh 1999). From the biosyn thetic point of view, the huge diversity of most polyke tide structures is generated from simple carboxylic acid monomers via a series of enzymatic reactions mediated by polyketide synthases (PKSs) (Risdian et al. 2019; Her tweck 2015. Risdian et al. (2019) stated that among three types of bacterial PKSs known to date (types I, II, and III), type I PKS (PKSI) in Streptomyces has at tracted increasing attention due to the involvement of this type in the biosynthesis of many macrocyclic polyketides (macrolides) with various bioactivities. Type I PKS system consists of modules, which act in a noniterative way. Each module harbors a set of cat alytic domains that are responsible for one cycle of polyke tide elongation and modification. A module minimally consists of three domains: an acyltransferase (AT) do main, an acyl carrier protein (ACP) domain, and a ketoacyl synthase (KS) domain (Hertweck 2009; Rawlings 2001. Each module receives an acyl chain from the upstream module onto KS domain and a suitable extender unit onto ACP through acyltransferase (AT) activity. ACP in this module serves as an anchor for the building block. Sub sequent KScatalyzed addition of the extender unit to the acyl chain results in a polyketide product that can be mod ified by optional domains such as ketoreductase (KR), de hydratase (DH), and enoylreductase (ER) (Watanabe et al. 2003; Weissman 2015. Due to the highly conserved re gions and intermediate specificity of KS domains in PKS systems, KS sequences are ideally used as a basis not only for PKS classification, but also for predicting structures of intermediates and even final biosynthetic products (Apari cio et al. 1996; Nguyen et al. 2008. A remarkable structural and functional diversity of nonribosomal peptides are biosynthetically built from proteinogenic and nonproteinogenic amino acids often along with other unusual acids catalyzed by nonribosomal peptide synthetases (NRPSs) (Fischbach and Walsh 2006). A typical NRPS module minimally consists of an adenyla tion (AD) domain, a thiolation (T) domain, and a conden sation (C) domain (Fischbach and Walsh 2006). The AD domain responsible for amino acid activation and incor poration is the bestcharacterized among NRPS domains; and therefore it is suitable to be used as a basis for detect ing NRPS systems and predicting the amino acid building blocks (Röttig et al. 2011).
Screening of genes encoding PKS and NRPS systems based on the detection of KS and AD domains have ex tensively been used to obtain preliminary insights into the biosynthetic potential of various bacterial groups, includ ing the genus Streptomyces (AyusoSacido and Genilloud 2005). The objectives of this work are to analyze the biosynthetic potential of two uncharacterized actinobacte rial strains by analyzing KS and AD domains of the PKSI and NRPS genes. We tested the antimicrobial activity, de termined the taxonomic identity of both strains, and pre dicted their possible environmental distribution based on 16S rRNA analysis. The outcome of these studies become an important basis for genome sequencing to reveal the promising biosynthetic pathways and for chemical analy sis to identify secondary metabolites for drug discovery.

Bacterial strains and cultivation
Two different bacterial strains, designated as ID 04 677 and SDS14LU, used in this work were obtained from Dr. Puspita Lisdiyanti at Research Center for Biotechnol ogy, Indonesian Institute of Sciences (LIPI). Both strains were preserved or deposited with the names A375 and A538 in Indonesian Culture Collection (INaCC) operated by LIPI. The strain ID 04 677 was derived from Pur wadadi, West Java, which was identified in this work based on 16S rRNA gene analysis using 27F and 1492R primers (Weisburg et al. 1991). The strain SDS14LU was derived from Mount Tambora Savana, Bima, West Nusa Tenggara, which was previously identified as Strep tomyces thioluteus (Fawzya et al. 2016 DH5α ATCC® 53868 and pGEMT Easy (Promega) were used for DNA cloning. All reagents and kits used were in high quality obtained from various manufacturers, such as Promega, Thermo Scientific, and Roche.

16S rRNA gene analyses
Genomic DNA was isolated from each strain according to the CTAB method (Piel et al. 2004) with a slight modifica tion and used as the template for the PCRamplification of nearly complete 16S rDNA. The PCR mixture contained 1x GoTaq Green Master Mix (Promega), 1 µM each 16S primer, and 0.5 µl DNA template in total volume of 50 µl. The PCR program was set up at 30 cycles, consist ing of a predenaturation at 95°C for 3 min, denatura tion 95°C for 30 seconds, annealing 55°C for 1 min, and elongation at 72°C for 1.5 min, and final elongation 72°C , 5 min. The target PCR product ( 1400 bp) was sep arated on the 1% agarose gel, purified from the gel, and subjected to DNA sequencing using the same primers em ployed in the PCR amplification. DNA sequencing was conducted using the Big Dye Terminator Cycle sequencing kit on ABI Prism 3700 DNA analyzer (Applied Biosys tems, USA). The resulting forward and reverse reads of each 16S rRNA gene sequence were assembled and sub jected to the Mega BLAST mode of NCBI (Morgulis et al. 2008) against the preformatted 16S microbial database downloaded from the updated BLAST databases (https://ft p.ncbi.nlm.nih.gov/blast/db/) in Geneious Prime sequence analysis software (ver. 10.2.3). The resulting 16S se quence database created from the maximum BLAST hits of 200 was aligned with the 16S rRNA gene sequences of both ID 04 677 and SDS14LU strains using MUSCLE (Edgar 2004). The alignment result was saved as MEG file and used to build a phylogenetic tree in Molecular Evolu tionary Genetic Analysis version 10.0.5 (MEGA X) (Ku mar et al. 2018) based on the NeighborJoining method (Saitou and Nei 1987). To investigate the microbial habit ability, the 16S rRNA gene sequences of both strains were submitted to MetaMetaDB, a database and analytic system for investigating microbial habitability (Yang and Iwasaki 2014). This MetaMetaDB analysis performed a BLAST search against 2,949,852 representative 16S rRNA se quences from 61 diverse environments (update dataset by November 6, 2014) (http://mmdb.aori.utokyo.ac.jp/).

Antimicrobial assays
Streptomyces strains were individually cultivated on ISP2 agar plates at 30°C for a week. The culture agar in each plate was cut into small pieces in square shape (approx imately 1 cm x 1 cm). The agar pieces were transferred into a beaker glass containing 100200 ml ethyl acetate and mixed well. The beaker glass was closed and let overnight at room temperature. The ethyl acetate extract was transferred into an empty beaker glass and evapo rated overnight in a fume hood. The dried extract was dissolved in methanol at the final concentration of 100 mg/ml. The extract dissolved in MeOH (50 µl) was ap plied onto a sterile paper disc in the clean bench, let it dry for 15 min, and placed on an agar plate containing a test organism. Four bacterial species as mentioned above were used as test organisms for antimicrobial assays. Each test bacterial strain was cultivated overnight in 5 ml Nutrient Broth (NB). The NB culture of each tested bacterial strain (100 µl) was mixed with 8 ml of semisolid (half recipe) MuellerHinton Agar (MHA), and subsequently poured evenly on the surface of HMA plate (full recipe). The pa per discs containing extracts were individually placed on the top of HMA plates containing test bacteria. The HMA plates with paper discs were incubated at 30°C for 24 h. Zone of inhibition around discs was measured.

Cloning into E. coli and sequencing
KSencoding fragments were PCRamplified from the bacterial cell fraction mentioned above using the primer set KSDPQQF (5' MGNGARGCNNWNSMNATGGAYCCNCARCANMG 3') and KSHGTGR (5' GGRTCNCCNARNSWNGTNCCNGTNCCRTG3') (Piel 2002). PCRamplification of regions coding for AD domain was carried out using the primer pair AGfor (5'GCSTACSYSATSTACACSTCSGG3') and AGrev (5'SASGTCVCCSGTSCGGTAS3') (AyusoSacido and Genilloud 2005). The PCR composition and condition are the same as described above, except the annealing temperature was set up at 60.1°C. The PCR products were separated on 1% agarose gel in the electrophoresis (Power Pac Basic, BioRad). Target fragments of ap proximately 700 bp were extracted from the gel using GeneJET Gel Extraction and DNA Cleanup Micro Kit (Thermo Scientific). The extracted DNA solution was concentrated to 7 µl by evaporator, followed with ligation with pGEMT Easy (Promega). The ligation product was transformed into electrocompetent cells of E. coli DH5α using the electroporator Micropulser (BioRad), added with 1 ml LB, and incubated at 37°C for 1 h in a shaking incubator. The transformed cell suspension was spread on LB plates containing 100 µg/ml ampicillin previously supplemented with 40 µl of 0.1 M isopropyl βD1thiogalactopyranoside (IPTG) and 40 µl of 5bromo4chloro3indolylβDgalactopyranoside (XGal). After overnight incubation at 37°C, the re sulting white colonies were picked up, transferred to LB agar plates, 96well microplates, and tubes containing 5ml LB, followed by overnight incubation. Plasmid constructs were individually recovered from cultures using GeneJet Plasmid Miniprep Kit (Thermo Scientific). The resulting plasmid samples were individually cut with EcoRI to check the presence of inserts, and subsequently digested with BamHI to know the restriction pattern. Recombinant plasmids with unique restriction patters were subjected to DNA sequencing using the primer T7 (5'GTAATACGACTCACTATAGGG3') that recognizes the cloning vector.

Bioinformatic analyses of KS and AD sequences
MIBiG (Minimum Information about a Biosynthetic Gene cluster) database (version 2.0) containing biosynthetic gene clusters of known function was initially downloaded in FASTA format from the Genomic Standards Con sortium website (https://mibig.secondarymetabolites.org/) (Kautsar et al. 2020). All of the DNA sequences encoding KS and AD domains isolated in this work were translated into amino acid sequences using the Webbased transla tion tool Expasy (http://www.expasy.org/tools/dna.html), which was subsequently subjected to BLASTp analysis against the MIBiG database operated in Geneious Prime (ver. 10.2.3). Of the resulting maximal BLAST hits of 50, several domain sequences were selected, downloaded, and aligned with the KS and AD sequences obtained in this work using Clustal Omega program (Sievers and Hig gins 2018). A phylogenetic tree was then constructed us ing the neighborjoining method (Saitou and Nei 1987) or the Maximum Likelihood method and JTT matrixbased model (Jones et al. 1992) in Mega X (Kumar et al. 2018). Resampling method with bootstrapping values inferred from 1000 replicates was set up on each tree branch to es timate the reliability of phylogenetic tree reconstruction. Based on this phylogenetic tree, we determined the affil iation or classification of KS sequences. The sequence data was also analyzed using some other programs, such as NaPDoS for comparison to a broad set of curated refer ence genes from wellcharacterized biosynthetic pathways (Ziemert et al. 2012) and NRPSpredictor2 (Röttig et al. 2011) for predicting the substrate specificity of adenyla tion domains. The Enzyme Function Initiative -Enzyme Similarity Tool (EFIEST) SSN analysis (Gerlt et al. 2015) of AD sequences was conducted to generate a protein se quence similarity network.

Taxonomic affiliation, antimicrobial activity, and environmental habitability
We evaluated the taxonomic affiliation and antimicrobial activity of two cultivated Streptomyces strains originated from Indonesia. One strain (ID 04 677) isolated from a soil habitat in Purwadadi appeared as small roughround colonies with whitecolored aerial mycelium on the ISP2 solid medium (Supplementary Figure 1a) (Labeda et al. 2012). Further phylogenetic analysis re vealed its close relationship with S. luridus, which was in the same clade as the following characterized Strep  (Saitou and Nei 1987) showing the predictive KS classification based on substrate specificity. (b) Cooccurrence network of AD domains based on substrate specificity: blue for Gly, pink for Orn, and orange for Thr. Edges (lines) indicate that pairwise alignments of nodes. This was conducted by EFI-EST SSN analysis (Gerlt et al. 2015(Gerlt et al. , 2011, which is visualized by Cytoscape (Shannon et al. 2003). Pairwise identity (%) between AD sequences obtained in this study and representaive ADs from other biosynthetic pathways. Note: IAD1 = ID677AD1, IAD2 = ID677AD2, IAD3 = ID677AD3.  Figure 1b). The outcome of this 16S rRNA analysis strongly suggested the taxo nomic identity of ID 04 677 as S. luridus, and therefore we named it as Streptomyces luridus ID 04 677.
Interestingly, the EtOAc extract prepared from this strain exhibited antimicrobial activity against P. aerugi nosa (Fig. 1C), a Gramnegative pathogenic bacterium known as the leading cause of morbidity and mortal ity in cystic fibrosis (CF) patients and nosocomial in fections (Moradali et al. 2017). To get insight into the habitability of this strain in diverse environmental FIGURE 2 Phylogenetic analysis of two ketosynthase sequences (SDS14KS4 and SDS14KS5) from Streptomyces luteosporeus SDS14LU, which involved a total of 17 KS sequences from cis-AT modules with the specificity of different substrates. The optimal tree was constructed in MEGA X (Kumar et al. 2018) using the Maximum Likelihood method and JTT matrix-based model (Saitou and Nei 1987). Representative types of cis-AT KS domains in loading modules (usually fused with a first extender module) with their starter units are indicated with blue bullets for acetyl-CoA, red bullets for propionyl-CoA, and a green bullet for methylbutyryl-CoA. Representative KS domains from cis-AT loading modules with the substrate specificity for and α-methylated intermadite. The extension units are shown with pink bullets for methylmalonyl-CoA and orange bullets for malonyl-CoA.

Secondary metabolite biosynthetic potential
Using the degenerate primers for detecting PKS (Piel 2002) genes, we PCRamplified and cloned KSencoding fragments into E. coli. Subsequent sequencing and analyz ing some clones with correctsized inserts indicated that four amplicons belonged to KS domains characterized by the presence of typical KS conserved motifs (e.g., MD PQQR) and two of the conserved catalytic triad (CHH) (Robbins et al. 2016) (Supplementary Figure 3). They re tain the active sites CSG/SSL and HGTGT, suggesting its ability to catalyze translocation of carbon chain and decar boxylative condensation (Robbins et al. 2016).
One of the two KS amplicons from S. luridus ID 04 677 (ID677KS1) exhibited high identity (73%74%) with those in cisAT PKS modules, exemplified by PieA6 in the biosynthesis of piericiden A1 (Engl et al. 2018) (Supple mentary Figure 4). Interestingly, the sequence alignment between ID677KS1 and other KS domains from the same cisAT PKS clade indicated highly conserved amino acid residues that could be responsible for its specificity de terminant, as has been shown for EryKS3 (erythromycin PKS), PikAIII (pikromycin PKS), MlsA2KS (mycolac tone PKS) (Murphy et al. 2016) (Supplementary Figure  4). Based on NaPDoS analysis (Ziemert et al. 2012), the amplicon ID677KS1 displayed the highest similarity (87% identity) with KS in cisAT PKS catalyzing lymphostin biosynthesis (Aotani et al. 1997) (Supplementery Table 1). The second KS (ID677KS4) from S. luridus ID 04 677 was closely related to the first KS in the extension mod ule McyD (Nishizawa et al. 1999) (Supplementary Table  1). This sequence analysis strongly suggests the classifi cation of both KS amplicons into cisAT PKS type.
Further phylogenetic analysis against several KS do mains from some representative biosynthetic pathways re trieved from MIBiG database (Kautsar et al. 2020) clas sified ID677KS1 into the extension module clade and ID677KS4 into the loading module type (Figure 1a). ID677KS1 was phylogenetically closer to PieA6KS2 (74% pairwise idenity), a cisAT KS domain that accepts an acyl intermediate containing αmethyl and βketo moi ety (Liu et al. 2012), suggesting that they may share a similar substrate specificity (Figure 1a). ID677KS4 phy logenetically belonged to the same clade as KS domains in the loading modules. It was closer to the KS from MtaB with the substrate specificity for methyl butyryl CoA (Silakowski et al. 1999).
BLAST search showed that two different KS se quences from the strain S. luteosporeus SDS14LU (SDS14KS4 and SDS14KS5) were similar with many KS domains from type I PKS system (Supplementary Figure  6). The first KS sequence (SDS14KS4) shared the identity of 41% to 53% with those in the cisAT modules, such as McyD (microcystin PKS) (Tillett et al. 2000). The other KS (SDS14KS5) showed the high identity of 7476% with those in the cisAT modules, such as PikAI_Q9ZGI51KSB (pikromycin PKS) (Xue et al. 1998) (Supplementary Table  3). This sequence analysis suggests that both SDS14KS4 and SDS14KS5 belong to cisAT PKS system. Phylogenetic analysis of both KS sequences against those from several characterized biosynthetic pathways di vided them into two distinct cisAT PKS clades, which were separated from the FAS clade (Figure 2). A close look at the multiple alignment suggests the presence of conserved residues in cisAT KS of loading modules that are different from the counterpart residues in cis AT KS of extension modules (Supplementary Figure 6). SDS14KS4 was particularly grouped in the same clade as KS domains from cisAT chainextension modules, such as McyD (Tillett et al. 2000). We propose that a cisAT chainextension module harboring SDS14KS4 may recruit malonylCoA through the cognate AT domain and load it onto the ACP domain. The ACP may deliver this exten der unit to SDS14KS4 for a decarboxylative condensation reaction with the installed intermediate (Weissman 2015).
In contrast to SDS14KS4, SDS14KS5 was classified into a different cisAT KS group associated with load ing modules, which is usually fused to the first exten der module (Buntin et al. 2010) (Figure 2). In particu lar, SDS14KS5 was closely related to the KS in HerB, a loading module in herboxidiene biosynthesis that accepts the starter unit propionate . One of the unique conserved residues in SDS14KS5 is a glutamine in place of a cysteine in the motif CSSL, which is highly conserved in some KS domains in loading modules. These so called KSQ domains have no condensation activity but can decarboxylate dicarboxylic acid starters. In the KSQ type loading module harboring this KS, it is assumed that the AT domain might select propionylCoA as a starter substrate and transfers it onto the ACP domain. Subse quent KSQcatalyzed decarboxylation generates a propi onate unit, which is then transferred to the downstream ex tension module for polyketide chain elongation (Miyanaga 2017).

Accession numbers
The partial 16S rRNA gene sequence of the strain ID04 677 obtained in this work was deposited in the GenBank with Accession Number of KX150801. KS and AD se quences isolated from ID 04 677 obtained in this work was deposited in the GenBank database under the following accession numbers: MT676434, MT676435, MT676436, MT676437, and MT676438. KS sequence derived from SDS14LU in this work was deposited under the accession numbers MT676439.

Discussion
In this work, we analyzed biosynthetic potential of two Streptomyces strains derived from geographically differ ent locations in Indonesia, namely Puwardadi (a region in West java) and Mount Tambora (an active stratovolcano Savanna located in the northern part of Sumbawa, West Nusa Tenggara). Phylogenetic analysis indicated the iden tity of the Puwardadi strain (ID 04 677) as S. luridus and the Tambora strain (SDS14LU) as S. luteosporeus. To the best of our knowledge, both species remain poorly inves tigated. Based on MetaMetaDB search (Yang and Iwasaki 2014), their 16S rRNA gene sequences shared >97% iden tity with thoses found predominantly in root, rhizosphere, soil, and ant, suggesting that they live not only as free living soil bacteria, but also as symbiosis with plant roots and insects.
The antimicrobial assays indicated that both strains ex hibited antibacterial activity towards the pathogenic bac terium P. aeruginosa. Since this opportunistic pathogen is resistant to a wide range of antibiotics (Moradali et al. 2017), both strains promise biosynthetic potential to pro duce multidrug resistant antimicrobial compounds. The ability of them to produce antimicrobial metabolites com bined with their predominant distribution in root, rhizo sphere, soil, and ant may indicate their ecological role as biocontrols to protect the plant and ant hosts from micro bial pathogens (van der Meij et al. 2017). This bioecolog ical information provides a strategic direction to explore them from unique habitat types located in geographically different regions for drug discovery program.
Previous studies indicated that S. luridus was a pro ducer of compounds with bioemulsification potential (Lamilla et al. 2018). S. luteosporeus is known to produce thiolutin and indolmycin (Celmer and Solomons 1955). Although some compounds have been described for both strains, such previous studies indicated that their biosyn thetic capacity remain poorly investigated, especially in terms of polyketide and nonribosomal peptide biosyn thetic pathways. Therefore, identification of PKS and NRPS based on KS and AD sequence analysis described in this work is necessary to reveal the biosynthetic potential of both strains.
Our genetic analysis indicated that two KS domains obtained from S. luridus ID 04 677 are phylogenetically derived from different cisAT PKS module types: the first one (ID677KS1) belongs to an extension module with the predictive specificity for an αMe and βketo intermedi ate, and the second one (ID677KS1) is a part of a loading module with the predictive specificity for methylbutyryl CoA. Phylogenetic analysis of two KS domains derived from S. luteosporeus SDS14LU suggests that one of them (SDS14KS4) belongs to a cisAT chain extension module with the predictive specificity for an αMe and βOme in termediate. The other KS (SDS14KS5) seems to be a part of a KSQtype loading module that consists of KSQ, AT, and ACP (Miyanaga 2017) with the predicted specificity for propionylCoA. Three different AD domains of NRPS system were cloned from S. luridus ID 04 677. Their pre dicted specificity are Gly, Thr, and Orn (Röttig et al. 2011).
To the best of our knowledge, no PKS and NRPS pathways harboring such KS and AD features have been reported so far from both strains.
Through genetic and bioinformatic analyses, we have shown the biosynthetic potential of both strains. Further genome sequencing and chemical analyses are necessary to reveal their entire biosynthetic pathways responsible for polyketide and nonribosomal peptide production. The ge netic analysis approach described in this work can be used as a simple guide not only to obtain rapid insight into natu ral product biosynthetic gene diversity in microorganisms (both cultivated and uncultivated), but also to prevent the rediscovery of polyketide and peptide biosynthetic path ways described previously.

Conclusions
Bacterial strain (ID 04677 and SDS14LU) was screened from Purwadadi and Mount Tambora. ID 04677 was identity as Streptomyces luridus and SDS14LU was iden tity as Streptomyces thioluteus. One of the two KS am plicons from S. luridus ID 04 677 exhibited high iden tity (73%74%) with those in cisAT PKS modules. Two different KS sequences from the strain S. luteosporeus SDS14LU were similar with many KS domains from type I PKS system. The antimicrobial assays indicated that both strains exhibited antibacterial activity towards the pathogenic bacterium P. aeruginosa. Genetic analysis is one of technique for obtain natural biosynthetic gene di versity in microorganisms.