Open Access Article
Zhiwei
Qin
a,
John T.
Munnoch
b,
Rebecca
Devine
b,
Neil A.
Holmes
b,
Ryan F.
Seipke
c,
Karl A.
Wilkinson
d,
Barrie
Wilkinson
*a and
Matthew I.
Hutchings
*b
aDepartment of Molecular Microbiology, John Innes Centre, Norwich Research Park, Norwich, NR4 7UH, UK. E-mail: barrie.wilkinson@jic.ac.uk
bSchool of Biological Sciences, University of East Anglia, Norwich Research Park, Norwich, NR4 7TJ, UK. E-mail: m.hutchings@uea.ac.uk
cSchool of Molecular and Cellular Biology, Astbury Centre for Structural Molecular Biology, University of Leeds, Leeds, LS2 9JT, UK
dScientific Research Computing Unit, Department of Chemistry, University of Cape Town, Rondebosch 7701, Cape Town, South Africa
First published on 13th February 2017
We report a new Streptomyces species named S. formicae that was isolated from the African fungus-growing plant-ant Tetraponera penzigi and show that it produces novel pentacyclic polyketides that are active against MRSA and VRE. The chemical scaffold of these compounds, which we have called the formicamycins, is similar to the fasamycins identified from the heterologous expression of clones isolated from environmental DNA, but has significant differences that allow the scaffold to be decorated with up to four halogen atoms. We report the structures and bioactivities of 16 new molecules and show, using CRISPR/Cas9 genome editing, that biosynthesis of these compounds is encoded by a single type 2 polyketide synthase biosynthetic gene cluster in the S. formicae genome. Our work has identified the first antibiotic from the Tetraponera system and highlights the benefits of exploring unusual ecological niches for new actinomycete strains and novel natural products.
We have been exploring the chemical ecology of protective mutualisms formed between actinomycete bacteria and fungus-growing insects in order to understand how these associations are formed and to explore this niche as a potential source of new antibiotics.4 In addition to the fungus-growing attine ants of South and Central America, which use actinomycete-derived antibiotics in their fungi-culture,5,6 it was recently discovered that many plant-ants also cultivate fungi.7–9 Plant-ants live in a mutualism with their host plant and provide protection from larger herbivores. In return, the host plants have evolved specialised hollow structures called domatia that house and protect the ants.10 South American Allomerus plant-ants and African Tetraponera plant-ants both grow fungi inside their domatia and they are associated with antibiotic-producing actinomycete bacteria.11,12
We previously reported the isolation of filamentous actinomycete bacteria, including Streptomyces and Saccharopolyspora strains, from the domatia and worker ants of Tetraponera penzigi plant-ants collected in Kenya.12 Genome sequencing of these strains allowed us to identify new species with genomes encoding novel and/or atypically large numbers of BGCs based on antiSMASH analysis.13 We consider strains containing significantly higher numbers of BGCs than typical strains (for Streptomyces sp. this is in the range 30–35) to be ‘talented’ with respect to their potential for yielding new natural products. One such organism, which we designate Streptomyces formicae KY5, also displayed a unique antagonistic activity against pathogenic drug resistant bacteria and fungi, including methicillin resistant Staphylococcus aureus (MRSA) and the multidrug resistant fungal pathogen Lomentospora prolificans.14 Subsequent bioassay guided fractionation using the sensitive test strain Bacillus subtilis led to the isolation and structural elucidation of thirteen new polyketide natural products that share a rare pentacyclic structure, some of which contain up to four chlorine atoms. These compounds fall into two groups. The first group (1–3) have an aromatic C-ring structure with sp2 carbon atoms at C10/C19, and lack any formal chiral centres. We have named these compounds fasamycin C–E respectively given their very close structural similarity to fasamycins A and B described previously from heterologous expression of a clone expressing a type 2 polyketide synthase (PKS) BGC isolated from an environmental DNA derived library.15 In contrast, compounds 4–13 are highly modified compared to the fasamycins with a non-aromatic C-ring and chiral centres at C10 and C19. We have named this group of compounds the formicamycins because they are the first natural products to be characterised from S. formicae and are structurally and biosynthetically distinct from the fasamycins (see below). Supplementation of the growth medium with sodium bromide resulted in the incorporation of bromine to yield three additional formicamycin congeners (14–16).
The formicamycins and fasamycins are active against clinical isolates of MRSA and vancomycin resistant enterococci (VRE), but do not display Gram-negative antibacterial or antifungal activity. The availability of sixteen congeners allowed their structure–activity relationship (SAR) to be examined. We then grew MRSA for 20 generations in the presence of sub-inhibitory concentrations of three formicamycins and re-determined the MICs for MRSA. These assays showed that MRSA does not easily acquire spontaneous resistance to formicamycins, at least under the conditions tested. Finally, we show, using CRISPR/Cas9 genome editing, that biosynthesis of these compounds is encoded by a type 2 PKS BGC in the S. formicae chromosome, and that re-introduction of this BGC restores biosynthesis of formicamycins in S. formicae. Identification of the formicamycin BGC allowed us to propose a plausible biosynthetic pathway. Deletion of forV encoding a putative flavin dependent halogenase abolished the production of any halogenated molecules and stalled the biosynthetic pathway at the fasamycin congener stage (1–3) indicating halogenation is a critical step required for further post-PKS modification to yield the formicamycin scaffold.
16 and Lomentospora prolificans CBS116904 (see below). These results prompted us to examine the relative genetic relationship with sequenced streptomycetes, for which there are now more than 950 complete and draft genome sequences available (ESI Fig. S1†). On the basis of 16S RNA sequence analysis this strain possesses a unique lineage and is most closely related to Streptomyces sp. NRRL S-920, which was originally isolated from a soil sample of unknown origin. A more detailed comparison of atpD, rpoB and three other widely used phylogenetic markers, gyrA (DNA gyrase subunit A), recA (recombination protein) and trpB (tryptophan biosynthesis) revealed a 95% shared nucleotide identity between concatenated atpD-gyrA-recA-rpoB-trpB and Streptomyces sp. NRRL S-920, suggesting this strain represents a new species. Given that it was isolated from Kenyan T. penzigi worker ants, we suggest the name Streptomyces formicae KY5.
![]() | ||
| Fig. 1 Structures of the previously reported fasamycins A & B, the new fasamycin congeners C–E (1–3) and the formicamycins A–M (4–16). | ||
With the structure of 5 in hand we were able to readily assign the remaining structures as described in the ESI.† NOESY correlations allowed us to link the methoxy at C5 with H4 (e.g.4, 6, 8–11 and 13). We could also use NOESY correlations to distinguish H14 and H16 once one was chlorinated, depending on their relationship to the gem-dimethyl group (e.g.7, 8 and 9).
In addition to the formicamycins 4–16, we identified three related compounds (1–3) which lacked the two chiral centres at C10 (tertiary hydroxyl group) and C19 (bridgehead proton), and have an aromatic C-ring structure. These compounds were significantly more yellow than 4–16 with distinct UV spectra (with maxima at 246, 286, 353 and 418 nm) and exhibited significantly different optical rotations to the formicamycins. On the basis of these observations we assigned these compounds as new fasamycin congeners C–E (1–3) respectively. The fasamycins were first reported by Brady and co-workers in 2011
15,20 and 1–3 represent new members of this family. We hypothesise that 1–3 represent biosynthetic precursors of the formicamycin biosynthetic pathway as discussed below.
To unambiguously assign the pentacyclic skeleton of these metabolites and confirm their polyketide origin, we performed a stable isotope labelling experiment. S. formicae was cultivated on MS agar (2 L) in the presence of [1,2-13C2] sodium acetate. After 7 days incubation the agar was extracted and the most abundant congener was isolated (compound 4; 5 mg). The resulting 13C NMR spectra clearly indicated the intact incorporation of 12 acetate derived units, plus an enriched single carbon at C24, in a pattern consistent with a polyketide biosynthetic pathway (see ESI Fig. S3†).
To aid in determining their stereochemistry the electronic circular dichroism (ECD) spectra of fasamycin 3 and formicamycin 5 were calculated using time-dependent density functional theory (TDDFT). First, a systematic conformational analysis of each isomer was carried out using the MMFFs molecular mechanics force field via the Maestro software package.21 The conformers obtained within an energetic range of 3 kcal mol−1 of the lowest energy conformer were further optimized using the PBE1PBE22 exchange-correlation functional at the def2tzvp23 basis set level and with the SMD solvent model24 for methanol using the Gaussian09 program package.25 Frequency calculations were then carried out using these same settings to calculate the relevant percentage of the population of the conformers. The 30 lowest electronic transitions were then calculated using TDDFT and the rotational strengths of each electronic excitation were converted to ECD spectra using a Gaussian function with a half-bandwidth of 0.248 eV. The overall ECD spectra were then generated according to the Boltzmann weighting of each conformer.
For the fasamycins, rotation about the C6–C7 axis means ring-A can be drawn with either the ortho hydroxyl or methyl group pointing forwards which correspond to the S- or R-configurations respectively. Comparison of the experimentally obtained ECD spectra for 3 to those calculated gives excellent agreement with that calculated for the S-configuration (Fig. 3A and ESI Fig. S3†) strongly suggesting this represents the preferred conformation.
![]() | ||
| Fig. 3 Comparison of the experimental and calculated CD spectra of 3 (A) and 5 (B), and the lowest energy conformers of (S)-3 (C) and (10R,19R)-5 (D). The key NOESY correlations for 5 are shown. | ||
For 5 we first compared the predicted structures for the lowest energy conformations of both the (10RS,19RS) and (10SR,19RS) diastereoisomeric pairs to data from NOSEY experiments. As observed in ESI Fig. S6† the (10SR,19RS) isomers with a trans relationship of the C10 and C19 substituents adopt an extended conformation of the four fused rings B–E. In contrast the cis (10RS,19RS) isomers are predicted to adopt a twisted L-shaped conformation (Fig. 3D). From this comparison the methine proton at C19 becomes diagnostic as the (10RS,19RS) isomers should show strong correlations to both methyl groups attached to C18 (methyl-26/27), whereas for the (10SR,19RS) isomers it should only give a correlation to methyl-27. Analysis of the NOESY data shows strong correlations for both methyl groups (26/27), and the remaining correlation data are also consistent with that expected for the (10RS,19RS) isomers (see Fig. 2 and 3D). We then acquired additional NMR datasets for 5 in non-protic solvent (d6-DMSO/d3-acetonitrile) and were able to locate the signal for the exchangeable hydroxyl proton at C10. Analysis of the NOESY spectrum showed clear correlations for this proton to the methine proton at C19 and methyl-27 which is compatible with the cis (10RS,19RS) isomers, but not the trans (10SR,19RS) isomers. NOESY data for the remaining formicamycin congeners was also consistent with the cis (10RS,19RS) configuration in each case. On this basis we were able to rule out the trans (10SR,19RS) isomers and proceeded to analyse the calculated and experimentally determined ECD spectra for the cis (10R,19R) and (10S,19S) enantiomers of 5 (Fig. 3B and ESI Fig. S4 and S5†). These data strongly suggested that the (10R,19R) stereochemistry was correct. Therefore, using combined NOESY NMR and ECD data we assign the (10R,19R) stereochemistry to the formicamycins. However, we are unable to make a definitive statement regarding the chiral C6–C7 axis for the formicamycins.
| Compound | Minimum Inhibitory Concentration (μM) | ||
|---|---|---|---|
| B. subtilis | MRSA | VRE | |
| 1 | <20 | 40 | 40 |
| 2 | 10 | 10 | 10 |
| 3 | 5 | 80 | 80 |
| 4 | 5 | >80 | >80 |
| 5 | 10 | 10 | 10 |
| 6 | 5 | 1.25 | 80 |
| 7 | 10 | 20 | 10 |
| 8 | 10 | 20 | 10 |
| 9 | 5 | 20 | 2.5 |
| 10 | 5 | Not tested | Not tested |
| 11 | 10 | Not tested | Not tested |
| 12 | <2.5 | <2.5 | 1.25 |
| 13 | <20 | 0.625 | 1.25 |
| 14 | <2.5 | 2.5 | 5 |
| 15 | <2.5 | 1.25 | 2.5 |
![]() | ||
| Fig. 4 Growth inhibition curve and MIC determination for Bacillus subtilis in the presence of formicamycin 12 (Apr, apramycin; Amp, ampicillin). | ||
To test whether 1–15 can inhibit drug-resistant Gram-positive bacteria we tested them against clinical isolates of MRSA and vancomycin-resistant Enterococcus faecium (VRE) (see ESI†) and found that the formicamycins are effective inhibitors of these organisms (Table 1). During the course of these experiments we observed that our test strains did not acquire spontaneous resistance when cultured on agar containing formicamycins. To test this further, we grew MRSA for four generations in the presence of no compound (control) and half MICs of compounds 6, 13 and 15. We then repeated the MIC tests and found no difference between the MRSA strains suggesting no resistance had arisen to formicamycins. We repeated the experiment but this time grew the strains for 20 generations and again found no increase in the MICs for these compounds, suggesting they exhibit a high barrier for the selection of resistant mutants, at least under the conditions tested here.
13 identified only one type 2 PKS gene cluster (BGC30) which we designate for (Fig. 5; Table S2;† accession number: KX859301). We used the CRISPR/Cas9 vector pCRISPomyces-2
26 to delete the entire BGC30 and surrounding genes in order to generate the unmarked deletion strain S. formicae Δfor; deletion of the BGC was confirmed by PCR amplification and sequencing (see ESI†). The wild-type strain and four independently generated S. formicae Δfor mutants were then grown in parallel under formicamycin producing conditions and subsequent LCMS(UV) analysis of extracts confirmed that fasamycin/formicamycins were not produced by the mutant strains (Fig. 6B and C). To ensure that loss of fasamycin/formicamycin biosynthesis was due to genome editing, and not other mutational events, we utilized a PAC (P1-derived artificial chromosome) library of the S. formicae genomic DNA which was custom made in pESAC13 by BioS&T Co. (Montreal, Canada). This was screened with three primer pairs (Table S1†), amplifying fragments either side and in the centre of BGC30. A single clone carrying the entire BGC30 (pESAC13-215-G) was introduced into one of the fasamycin/formicamycin-deficient mutants using tri-parental mating.27 LCMS(UV) analysis of the complemented strain alongside wild-type and mutant strains confirmed that fasamycin/formicamycin biosynthesis had been restored (Fig. 6D), and we conclude that BGC30 encodes the biosynthesis of compounds 1–13 in S. formicae.
![]() | ||
| Fig. 5 Organization of the formicamycin (for) BGC and annotation of putative gene products. ACC = acetyl-CoA carboxylase; PKS = polyketide synthase; MFS = major facilitator superfamily. | ||
To investigate its biosynthetic role we deleted the forV coding sequence using CRISPR-Cas9 methodology. Four independently isolated mutants were verified by PCR and sequencing, and extracts of the mutants grown on MS agar were analysed by LCMS(UV) (Fig. 6E). This showed accumulation of the non-halogenated fasamycin C (1) plus a new molecule with the same molecular formulae and UV spectrum indicating that it is a structural isomer of 1 (presumably bearing an O-methyl group at either C5 or C23 rather than at C3). The production levels of 1 by this mutant is approx. 188-fold that observed for the wild-type strain. Notably, no formicamycins could be observed in this extract. These data strongly suggest that ForV is responsible for the introduction of up to four halogen atoms. Genetic (in trans) complementation with the forV gene under the control of the native promoter re-established production of the halogenated compounds 2 and 3 and the formicamycins (Fig. 6F) indicating there was no polar effect or unanticipated genetic mutation introduced by the gene editing.
All of 1–16 contain two methyl groups at C18 which, in conjunction with biosynthetic studies on the related pentangular polyketide benastatin,29 suggests that the first post-PKS step will involve installation of the gem-dimethyl group at C18. Three putative methyltransferases are encoded in BGC30 (ForM, ForT, and ForW), and ForT has the highest sequence shared identity with BenF (66%/49%; CAM58795.1) which catalyses the gem-dimethylation step during benastatin biosynthesis and is likely to catalyse the equivalent reaction during fasamycin/formicamycin biosynthesis; this gene is also present in the fasamycin BGC.15 Our inability to identify and isolate the putative intermediate 19, or indeed any congeners lacking the gem-dimethyl moiety, leaves open the possibility that this molecule may not exist as an enzyme free intermediate and that ForT might actually act upon an ACP-bound intermediate which is then released and decarboxylated. Additionally, we did not isolate any congeners lacking a methoxy-group at C3 which suggests that O-methylation at this position occurs next and will be catalysed by one of the remaining methyltransferases ForM or ForW to yield 1.
The accumulation of only 1 and a new isomer in the forV deletion mutant suggests that chlorination is the next step of the biosynthetic pathway and that it is essential to enable further post-PKS steps to occur in order to produce the formicamycins. This is consistent with the low levels of 1–3 observed from the wild-type organism, and analysis of the chlorination patterns for 2–13 suggests that chlorination at C2 or C22 is essential, with C22 likely being preferred to yield 2.
Introduction of the tertiary hydroxyl group at C10 and modification of ring-C probably occurs next in the biosynthetic sequence. Moreover, as we only identified formicamycins containing both of these changes we propose that the transformations are linked, and may be catalyzed by the combined actions of the flavin dependent monooxygenase ForX and flavin dependent oxidoreductase ForY to yield 20. A second O-methylation at C23 most likely occurs next (to give 21) as all formicamycins contain this change. It is currently unclear when the final O-methylation at C5 occurs.
Finally, the most abundant formicamycin congeners contain either three or four chlorine atoms located on three different rings, and the minor congeners contain mostly two or three chlorine atoms distributed around the various locations; no fasamycins have a chlorine atom on ring E. These observations are consistent with the idea that ForV is a promiscuous enzyme capable of catalysing up to four halogenation reactions on a single molecule, but that there is a preferred, but not absolute, ordering to these modifications.
Comparison to the fasamycin BGC15 fails to identify homologues of certain genes present in BGC30 that we propose may be involved in formicamycin biosynthesis. In contrast others are present in both BGCs that we suggest may be responsible for some of the structural differences observed. Plausible reasons for these differences include differential expression, or a lack of expression in one species, and the involvement of genes that were not captured on the expression cosmid used for production of the fasamycins.15 To address these questions a detailed study of formicamycin biosynthesis is underway in our labs.
Intriguingly, bioinformatics analysis shows that the formicamycin BGC is closely related to an unassigned BGC present in the genome of Streptomyces kanamyceticus (Genbank ID LIQU00000000.1). Further, an approx. 188 kbp region of the S. formicae genome, which encompasses BGC30, is syntenic with the S. kanamyceticus genome (extending approx. 64 kbp upstream and at least 95 kbp downstream, which is as far as the contig LIQU01000034 extends) and we suggest there has been a horizontal gene transfer event. Further bioinformatics analysis and consideration of the biosynthetic pathway leads us to propose that forQ and forCC represent the boundaries of BGC30 (Fig. 4). Additionally, the region of sequence encoding forX to forAA, which is not present on the S. kanamyceticus genome, comprises gene sequences with closest homologues in Actinomadura species, and appears to have been inserted into the S. kanamyceticus syntenic sequence. This suggests the formicamycin BGC may have its origin in multiple horizontal transfer events. Further work, both to understand the origins of the formicamycin BGC, and to delineate their biosynthesis, are underway in our laboratories. We anticipate this data will aid in the application of biosynthetic medicinal chemistry methods to produce further improved molecules with potential application as antibacterial agents.
Footnote |
| † Electronic supplementary information (ESI) available: General remarks; full experimental details; Fig. S1–S6; Tables S1 and S2. See DOI: 10.1039/c6sc04265a |
| This journal is © The Royal Society of Chemistry 2017 |