A Single Streptomyces Symbiont Makes Multiple Antifungals to Support the Fungus Farming Ant Acromyrmex octospinosus

Attine ants are dependent on a cultivated fungus for food and use antibiotics produced by symbiotic Actinobacteria as weedkillers in their fungus gardens. Actinobacterial species belonging to the genera Pseudonocardia, Streptomyces and Amycolatopsis have been isolated from attine ant nests and shown to confer protection against a range of microfungal weeds. In previous work on the higher attine Acromyrmex octospinosus we isolated a Streptomyces strain that produces candicidin, consistent with another report that attine ants use Streptomyces-produced candicidin in their fungiculture. Here we report the genome analysis of this Streptomyces strain and identify multiple antibiotic biosynthetic pathways. We demonstrate, using gene disruptions and mass spectrometry, that this single strain has the capacity to make candicidin and multiple antimycin compounds. Although antimycins have been known for >60 years we report the sequence of the biosynthetic gene cluster for the first time. Crucially, disrupting the candicidin and antimycin gene clusters in the same strain had no effect on bioactivity against a co-evolved nest pathogen called Escovopsis that has been identified in ∼30% of attine ant nests. Since the Streptomyces strain has strong bioactivity against Escovopsis we conclude that it must make additional antifungal(s) to inhibit Escovopsis. However, candicidin and antimycins likely offer protection against other microfungal weeds that infect the attine fungal gardens. Thus, we propose that the selection of this biosynthetically prolific strain from the natural environment provides A. octospinosus with broad spectrum activity against Escovopsis and other microfungal weeds.


Introduction
Insect fungiculture has been best studied in the attine ants (subfamily Myrmicinae, tribe Attini) whose common ancestor is estimated to have evolved agriculture around 50 million years ago [1]. Attine ants are so dependent on the cultivation of fungus that when a daughter queen leaves to found a new nest she carries a piece of the cultivar fungus in her mouth in order to establish a culture of that fungus in her new nest [1]. Intriguingly, attine ants have also evolved a mutualism with Actinobacteria that produce antibiotics that the ants use as weedkillers to keep their fungal gardens free of other microbes [2,3,4]. The relationship between Actinobacteria, fungal cultivar and attine ant has been intensely studied in the branch of higher attines known as the leaf-cutting ants (genera Atta and Acromyrmex) which harvest fresh vegetation to feed to their highly specialised fungal cultivar, Leucoagaricus gongylophorus [1]. The fungus has evolved lipid and carbohydrate rich hyphae known as gongylidia which the ants harvest and use as food [5]. Pathogens of the fungal garden, most notably fungi of the genus Escovopsis, if left unchecked, can destroy a fungal garden and lead to the collapse of the colony within weeks [6,7].
Two overlapping but conflicting theories have been put forward to explain the evolution of mutualism between attine ants and Actinobacteria. The first suggests co-evolution of attine ants and Actinobacteria belonging to the genus Pseudonocardia. This theory suggests that the fungus garden pathogen Escovopsis has also co-evolved and that Pseudonocardia and Escovopsis are engaged in an evolutionary arms race in which the bacteria evolve compounds that specifically target Escovopsis but do not inhibit the growth of the fungal cultivar [5]. The second model suggests that attine ants select antifungal-producing Actinobacteria from the environment and is consistent with the identification of additional Actinobacterial genera on leaf-cutting ants, including Streptomyces and Amycolatopsis species [3,4,8,9]. However, these theories are not mutually exclusive and evidence suggests attine ants co-evolve with Pseudonocardia bacteria and still select other antifungal producing bacteria from the soil, perhaps to prevent evolution of resistance in the fungal pathogens [9,10]. Indeed, there is good evidence, both direct and indirect, that leaf-cutting ants use multiple antifungals produced by multiple Actinobacteria during the cultivation of their fungal gardens suggesting that both models do apply, at least in the higher attine genus Acromyrmex [3,11]. Recent work demonstrated that Acromyrmex octospinosus are associated with a Pseudonocardia strain that may have co-evolved and a Streptomyces strain that was most likely acquired from the environment relatively recently [9].
One intriguing question that still needs to be addressed in the environmental recruitment model concerns how the ants select beneficial bacteria. Actinobacteria are well known producers of useful secondary metabolites, including around 60% of all known antibiotics [12] and it seems likely that this production capability is key to their success as mutualists. Clearly the production of antifungals makes them useful to the ants and we hypothesise that production of multiple antifungals with different targets by single Actinobacterial species would make them more attractive to the ants as mutualists. In this work we carry out a more in depth analysis of the antifungals made by one of the strains associated with the leaf-cutting ant A. octospinosus, a species of Streptomyces which has been proposed to support fungus growing ants through production of the polyene antifungal candicidin [4,9]. We report the genome sequence and analysis of this strain indicating its capacity to make numerous antibiotics, including at least three antifungal compounds. Curiously, this strain makes both of the antifungals that have been reported in the Streptomyces attine ant mutualists but neither the candicidin or antimycins are obligatory for the inhibition of the co-evolved nest pathogen E. weberi. We propose that additional and as yet unknown antifungal(s) made by this strain specifically target Escovopsis and that candicidin and antimycins offer protection against other microfungal weeds. We propose that the ability of this single species to make multiple antifungal compounds makes it an attractive acquisition for the ants and their fungiculture.

Genome sequencing and analysis
To determine the antibiotic biosynthetic capability of Streptomyces S4 a combination of shotgun, 3 kbp and 8 kbp paired end libraries were constructed and 454 pyrosequenced to generate .335 Mbp of sequence that was assembled into 12 scaffolds containing 211 large contigs. The genome consists of one ,7.5 Mbp linear chromosome, which is within the size range reported for genomes of other sequenced streptomycetes, as well as one linear plasmid (,180 kbp) and one circular plasmid (,2 kbp) [13]. The Streptomyces S4 genome was annotated by a combination of manual and automated methods and multiple biosynthetic gene clusters predicted to produce secondary metabolites were identified. Table 1 summarizes characteristics of predicted natural product biosynthetic gene clusters in Streptomyces S4. As expected based on our previous work, the Streptomyces S4 genome contained a candicidin biosynthetic gene cluster that shares 98% nucleotide sequence identity with the candicidin biosynthetic gene cluster from Streptomyces sp. FR-008 [14]. Other biosynthetic gene clusters of note are a non-ribosomal peptide synthetase (NRPS) biosynthetic gene cluster that is predicted to direct the biosynthesis of an antibacterial similar to mannopeptimycin, as well as a NRPS biosynthetic gene cluster whose predicted product is a gramicidinlike antibacterial. We note these antibacterials could be useful in eliminating competition during colonisation of the ant cuticle. Streptomyces S4 also contains a type II polyketide synthase (PKS) biosynthetic gene cluster that shares 100% nucleotide identity to that of the fredericamycin biosynthetic gene cluster characterized in S. griseus [15]. Fredericamycin is mostly known for its antitumor properties [16]. Additionally, there are six functionally unassigned biosynthetic gene clusters (three NRPS and three hybrid NRPS/ PKS) that are not similar to gene clusters with known products. It is likely that at least some these clusters encode secondary metabolites with antibacterial or antifungal activity.

Mutagenesis of the candicidin biosynthetic gene cluster does not abolish antifungal activity
Evidence that candicidin is not the sole antifungal generated by Streptomyces S4 is demonstrated by the retained antifungal activity of a mutant strain deficient in the biosynthesis of this compound. The candicidin biosynthetic gene cluster was disrupted by deletion of the polyketide synthase gene, fscC, which encodes the candicidin biosynthetic modules 6-10 [14]. LC-MS analysis of butanolextracted culture supernatants from the wild-type strain revealed a molecular ion (m/z 1109.6) consistent with that of candicidin D and showed characteristic polyene absorption bands in its UV spectrum, with absorbance maxima at 360, 380, 403 nm ( Fig. 1). As predicted, the molecular ion m/z 1109.6 was not detected in the DfscC mutant, indicating that candicidin production is abolished in this strain (Fig. 1). Bioassays of the isogenic wildtype and DfscC strains against C. albicans and the nest pathogen Escovopsis weberi demonstrated that loss of candicidin has no effect on the antifungal bioactivity of Streptomyces S4 (Fig. 2). This result suggests that Streptomyces S4 makes at least one additional antifungal compound that has not been identified previously, most likely encoded by one of the other biosynthetic gene clusters identified in this work.

Identification of the antimycin biosynthetic gene cluster
While this work was in progress another group reported that antimycins are produced by a number of the other Streptomyces strains associated with attine ant nests [17]. Antimycins inhibit the respiratory chain and are known to have antifungal activity. We investigated whether Streptomyces S4 is making antimycin compounds in addition to candicidin and hypothesised that antimycins could potentially account for the retained bioactivity against Escovopsis observed for the Streptomyces S4 fscC mutant. LC-MS analysis of culture supernatants of the wild-type strain identified eight compounds with m/z that match those reported for antimycins A1-A4 (Fig. 3). To determine if any of the eight compounds could be antimycins, we co-injected commercially available antimycin standards A1-A4 with our wild-type extract, which revealed that four of the eight compounds possess the same retention time as the antimycin A1-A4 standards (Fig. 3). Four of the eight S4 compounds were identical to the commercially available antimycin standards A1-A4 both in terms of UV absorbance profile and LC retention time. Whilst the remaining four S4 compounds possess the same UV absorbance characteristics as the antimycin standards (Fig. S1), and the same m/z parent ions as those of the standards, they exhibit different retention times (Fig. 3); further experiments are being carried out to identify and characterize these four previously unreported compounds.
To our knowledge the gene cluster that encodes the antimycin biosynthetic pathway has not been identified, despite these compounds first being isolated over 60 years ago [18]. The structure of antimycin suggests that it may be synthesized, at least in part, by an NRPS, and that threonine may be utilized as a substrate (Fig. 3). The Basic Local Alignment Search Tool [19] revealed a region of the Streptomyces S4 genome with 57% amino acid sequence identity to the threonine adenylation domain from the daptomycin biosynthetic protein, DptA [20]. This region of homology enabled us to identify a hybrid NRPS/PKS biosynthetic gene cluster that displays significant amino acid identity to a hybrid NRPS/PKS biosynthetic gene cluster present in both S. albus and S. ambofaciens and potentially encodes for the biosynthesis of antimycins (Fig. 4). Table 2 displays the proposed functions of proteins present in the hybrid NRPS/PKS cluster. In order to determine if this biosynthetic gene cluster can direct the production of antimycin, we disrupted the hybrid NRPS/PKS gene, antC and assessed antifungal activity against C. albicans in a plate bioassay. The DantC mutant displayed dramatically reduced antifungal activity against C. albicans compared to that of the wildtype strain (Fig. 2). This strongly suggested that the product of this cluster was an antifungal compound and is consistent with the hypothesis that this cluster could potentially mediate the biosynthesis of antimycins, compounds known to possess strong antifungal activity against C. albicans (Fig. 2). Confirmation that the hybrid NRPS/PKS encoded by Streptomyces S4 directs the biosynthesis of antimycins was obtained by comparing the LC-MS profiles of the wild-type, DfscC and DantC mutant strains. Extracted ion chromatograms revealed that the DantC mutant does not produce the eight antimycins and that the DfscC mutant retained the ability to produce these antimycin compounds (Fig. 3). A Streptomyces S4 mutant which cannot make candicidin or antimycins still has antifungal activity against E. weberi Our bioassays against E. weberi with the DfscC and DantC mutant strains of Streptomyces S4 did not show a reduction in bioactivity against this nest parasite indicating that key antifungal compound(s) affording protection against the natural fungal pathogen still remain to be identified (Fig. 2). To determine whether additional antifungals are made by Streptomyces S4 we generated a new mutant in which the fscC and antC genes were disrupted in the same strain and assessed the bioactivity of this strain against C. albicans and E. weberi. As we predicted, the antifungal activity of the DfscC DantC double mutant against E. weberi was comparable to that of the wild-type strain, confirming that additional antifungal compound(s) made by Streptomyces S4 account for the majority of the antifungal activity observed against E. weberi in vitro. The additional antifungal(s) may be encoded by one of the five functionally unassigned biosynthetic gene clusters identified in our genome analysis (Table 1) or by a gene cluster not identified in our analysis.

Discussion
Although the attine ant-fungal mutualism has been studied for more than a century, the antibiotic-producing Actinobacterial mutualists were only discovered ,15 years ago and it is only very recently that scientists have started to address the nature of the antibiotics being produced by these bacteria [2,4,9,17]. It has been hypothesised, although not proven, that these antibiotics are used by the ants to kill off contaminated parts of the garden and / or to suppress the growth of fungal pathogens including co-evolved pathogens in the genus Escovopsis and many other microfungal weeds [6,7]. Recent studies have shown that strains belonging to two key genera are typically associated with attine ants, species of Pseudonocardia, which have been suggested to have co-evolved with the ants and to be transmitted vertically by the queens, and species of Streptomyces which have been suggested to be more recently acquired from the environment [3,4,8,9]. Two antifungals have been identified from proposed mutualist species in each genus and both inhibit the nest pathogen Escovopsis in vitro. Pseudonocardia associated with the lower attine A. dentigerum makes dentigerumycin and Pseudonocardia associated with the higher attine A. octospinosus makes nystatin P1 [2,9]. Streptomyces mutualists associated with higher attines of the genus Acromyrmex are known to produce the well-known antifungals candicidin and antimycins and it has been suggested that these compounds account for the bioactivity of this Streptomyces strain against Escovopsis [4,9,17]. To date almost all of this work has been carried out through isolation and mass spectrometry analysis of the antifungal compounds, although we used genome scanning of a Pseudonocardia mutualist to identify a nystatin-like biosynthetic gene cluster and its product which we named nystatin P1 [9].
In this work we have undertaken the first in-depth genome sequence analysis of a proposed attine ant mutualist, in this case a candicidin-producing Streptomyces strain isolated from A. octospinosus garden worker ants collected in Trinidad [9]. Genome sequencing and analysis identified 17 gene clusters that are predicted to encode for known or unknown secondary metabolites, including the known gene cluster for candicidin biosynthesis. Following the discovery of antimycin production by Streptomyces strains isolated from attine ants in a separate study [17] we identified a gene cluster encoding a pathway that is consistent with antimycin biosynthesis. Surprisingly, despite antimycins first being isolated and characterised .60 years ago the antimycin biosynthetic pathway was not known. We identified eight compounds which we assigned as antimycins and then identified and disrupted the hybrid NRPS/PKS gene cluster which we predicted to encode antimycin biosynthesis. The production of the eight antimycin compounds was abolished in the mutant strain providing strong evidence that this gene cluster does indeed encode the antimycin biosynthetic pathway.
In bioassays of the wild-type Streptomyces S4 strain alongside strains which cannot make candicidin, antimycins or either of these antifungal compounds we found that whilst antimycin-and candicidin-deficient strains had reduced activity against the human pathogen C. albicans their activity against the nest pathogen E. weberi was unaltered. This is curious as the Streptomyces S4 strain and E. weberi are thought to have co-evolved in this mutualism [21]. This suggests that despite previous research demonstrating that purified candicidin and antimycin preparations inhibit the growth of Escovopsis in vitro neither compound is responsible for the activity observed in bioassays where Streptomyces S4 is challenged with E. weberi. We conclude that these compounds potentially inhibit the growth of other microfungal weeds found in the ant-fungus gardens while additional and currently unknown antifungal(s) produced by Streptomyces S4 have stronger activity against Escovopsis. We also propose that the combination of antifungals produced by this single Streptomyces strain coupled with the antifungal(s) produced by a Pseudonocardia strain isolated from the same nest provides a broad spectrum of antifungal activity that is used by the ants to farm their fungus. Furthermore, the antibacterials made by Streptomyces S4 potentially help it to outcompete other bacteria for the ant host.
Our data suggest that the Streptomyces strains isolated by other researchers from attine ants are likely to make additional antifungals since they appear to be closely related to Streptomyces S4. It will be important to re-examine the biosynthetic capability of these strains in order to fully understand the chemical basis of their interactions with attine ants and their fungal cultivar. This reflects a common problem in the field of natural product antibiotic discovery, in which the reisolation of known compounds hampers the discovery of new antibiotics. The approach we have outlined here is time consuming and technically challenging, but it is perhaps the only way to determine the entire biosynthetic capability of an antibiotic-producing strain particularly if some of the antibiotics being made, and their biosynthetic gene clusters, are new to science. Future work will be aimed at determining the products of the five unassigned biosynthetic gene clusters in Streptomyces S4 and identifying the additional antifungal compound(s) made by this strain. This is likely to involve significant challenges if, as we predict, these are novel secondary metabolites.
In conclusion, although good progress has been made recently we are still a long way from understanding the chemical basis of the symbioses between antibiotic-producing Actinobacteria and their attine ant hosts. We hope that our study will stimulate further research in this area and the identification of additional antifungal and antibacterial compounds in this system.

Growth media and strains
Streptomyces strains were routinely grown on soya flour mannitol (SFM) agar plates or in liquid TSB/YEME while E. coli strains were grown on Lysogeny both-Lennox (LB) [22]. Media was supplemented with antibiotics as required at the following concentrations: carbenicillin (100 mg/ml), hygromycin B (50 mg/ ml), nalidixic acid (25 mg/ml), apramycin (50 mg/ml). S4 was isolated and identified by 16S rDNA sequencing in a previous study (GenBank accession HM179229). Antifungal bioassays with C. albicans and E. weberi were carried out as described previously [9]. Strains and plasmids are described in Table 3.

Construction of Streptomyces S4 mutant strains
In order to create the DfscC mutant, two 3 kb knockout arms were PCR amplified using GoTaq Polymersae (Promega) with oligonucleotide primers RFS78 and RFS79 (upstream arm) and RFS80 and RFS81 (downstream arm), respectively. Oligonucleotide primers (Integrated DNA Technologies) were engineered at their 59 end to contain restriction sites for cloning (Table S1). The Figure 3. LC/MS analysis of Streptomyces S4 wild-type and mutant strains compared to antimycin standards. The extracted ion chromatograms for antimcyins A1-A4 are shown. Eight compounds consistent with the mass of antimycin A1-A4 were produced by S4 wild-type and S4 DfscC, but were not produced by the DantC mutant. Co-injection of antimycin A1-A4 with the S4 wild-type extract demonstrated that antimycin A1-A4 have the same retention time as four of the eight compounds produced by S4 wild-type. The UV visible spectra and ESI positive mode mass spectra for antimycin A1-A4 and the eight antimycin compounds produced by S4 wild-type are shown in Fig. S1 and Fig. S2, respectively. Antimycin A1: R 1 = CH(CH 3 )CH 2 CH 3 , R 2 = (CH 2 ) 5 CH 3 . Antimycin A2: R 1 = CH(CH 3 ) 2 , R 2 = (CH 2 ) 5 CH 3 . Antimycin A3: R 1 = CH(CH 3 )CH 2 CH 3 , R 2 = (CH 2 ) 3 CH 3 . Antimycin A4: R1 = CH(CH 3 ) 2 R 2 = (CH 2 ) 3 CH 3 . doi:10.1371/journal.pone.0022028.g003 resulting PCR products were cloned into pGEMT-EZ (Promega) and sequenced to verify their identity. The upstream arm was released from pGEMT-EZ with HindIII and BamHI and cloned into pKC1132 (which contained the RK2 conjugal origin of transfer and as well as an apramycin resistance gene [22,23] cut with the same enzymes to result in pKC1132-Up. Next, the downstream arm was released from pGEMT-EZ with BamHI and EcoRI and cloned into pKC1132-Up cut with the same enzymes to result in pKC1132-UpDn. Finally, a hygromycin B resistance cassette was PCR-amplified from pIJ10700 [24] using oligonucleotides RFS94 and RFS95 engineered to contain BamHI sites at their 59 end. The hygromycin resistance cassette was cloned into pGEMT-EZ and subsequently released by BamHI digestion and cloned into pKC1132-UpDn cut with the same enzyme to result in pKC1132-UpHygDn.
The plasmid pKC1132-UpHygDn was electroporated into E. coli strain ET12567/pUZ8002 [25] and transferred to Streptomyces S4 by cross-genera conjugation as previously described [22]. Transconjugants were selected for apramycin resistance. An apramycin-resistant transconjugant was obtained and subsequently replica plated to obtain hygromycin-resistant and apramycinsensitive colonies, a phenotype indicating that the fscC gene had been entirely replaced by the hygromycin resistance cassette and that the plasmid backbone was no longer present. Loss of the pKC1132-UpHygDn plasmid backbone and mutagenesis of the fscC gene in the DfscC strain was confirmed by PCR.
In order to disrupt the antC gene a ,1.5 kb internal fragment of the antC gene was PCR amplified using oligonucleotide primers RFS121 and RFS122 which were engineered to contain BamHI and EcoRI restriction sites at their 59 end, respectfully (Table S1). The resulting PCR product was sequenced to verify its identity and cloned into pGEMT-EZ to result in pGEMT-Ant. The apramycin resistance cassette containing a conjugal origin of transfer (aac(3)IV+oriT) was isolated from pIJ773 as a BamHI fragment and cloned into the BamHI site (provided by RFS121) in pGEMT-Ant to result in pGEMT-AntApr. The pGEMT-AntApr plasmid was electroporated into ET12567/pUZ8002 and mobilized to S4 wild-type and S4 DfscC by conjugation. Transconju-  gants were selected for apramycin resistance, a phenotype indicating that the disruption plasmid has crossed into the chromosome. Disruption of the antC gene was confirmed by PCR amplification using oligonucleotide primers (RFS147 and M13F, and RFS148 and M13R) targeting the DNA sequence upstream and downstream of the expected site of integration.

LC-MS analysis
Streptomyces S4 wild-type and mutant strains were cultivated in mannitol-soya flour liquid medium in a 250 ml flask shaking at 270 rpm. For analysis of candidicin production, cultures were harvested after 10 days of growth, bacterial cells were removed by centrifugation and the supernatant of three biological replicates was combined. Fifty milliliters of the combined supernatant was extracted three times with an equal volume of butanol. Butanol extracts were combined and evaporated to dryness under vacuum and the residue was resuspended in 0.5 ml of 50% aqueous methanol. For analysis of antimycin production, cultures were harvested after 4 days of growth, bacterial cells were removed by centrifugation and the supernatant of two biological replicates was combined and extracted with XAD16 resin (Sigma). Following extraction, the resin was washed twice with ten milliliters of deionized water and Streptomyces S4 metabolites were eluted from the resin with one milliliter of 100% methanol. Prior to LC-MS analysis the methanol elution was diluted with water to a final methanol content of 50%. Antimycin A1-A4 standards were purchased from Sigma Aldrich. Immediately prior to LC-MS analysis samples were spun in a microcentrifuge at maximum speed for 5 minutes to remove insoluble material. Only the supernatant (10 ml) was used for injection into a Shimadzu single quadrupole LC-MS-2010A mass spectrometer equipped with Prominence HPLC system as described previously [9]. For coinjection of antimycin A1-A4 with S4 wild-type extract 5 ml of standard and 5 ml of wild-type extract were mixed immediately prior to injection into the LC-MS.