Wide Distribution of Foxicin Biosynthetic Gene Clusters in Streptomyces Strains – An Unusual Secondary Metabolite with Various Properties

Streptomyces diastatochromogenes Tü6028 is known to produce the polyketide antibiotic polyketomycin. The deletion of the pokOIV oxygenase gene led to a non-polyketomycin-producing mutant. Instead, novel compounds were produced by the mutant, which have not been detected before in the wild type strain. Four different compounds were identified and named foxicins A–D. Foxicin A was isolated and its structure was elucidated as an unusual nitrogen-containing quinone derivative using various spectroscopic methods. Through genome mining, the foxicin biosynthetic gene cluster was identified in the draft genome sequence of S. diastatochromogenes. The cluster spans 57 kb and encodes three PKS type I modules, one NRPS module and 41 additional enzymes. A foxBII gene-inactivated mutant of S. diastatochromogenes Tü6028 ΔpokOIV is unable to produce foxicins. Homologous fox biosynthetic gene clusters were found in more than 20 additional Streptomyces strains, overall in about 2.6% of all sequenced Streptomyces genomes. However, the production of foxicin-like compounds in these strains has never been described indicating that the clusters are expressed at a very low level or are silent under fermentation conditions. Foxicin A acts as a siderophore through interacting with ferric ions. Furthermore, it is a weak inhibitor of the Escherichia coli aerobic respiratory chain and shows moderate antibiotic activity. The wide distribution of the cluster and the various properties of the compound indicate a major role of foxicins in Streptomyces strains.


INTRODUCTION
Plants, marine organisms, protozoans, fungi, and bacteria produce a wide range of different secondary metabolites. They are not essential for normal growth, development, or reproduction of an organism, but play a secondary role. There have been several discussions about the selective advantage of these natural products for their producers (Firn and Jones, 2000). Secondary metabolites may act as signals for differentiation, as communication molecules, or as weapons to defend against food competitors (Demain and Fang, 2000) and thus they often possess vital functions in their ecological habitat. The genes responsible for the biosynthesis of a compound are often located next to each other in so called biosynthetic gene clusters. The clusters often span more than 100 kb and encode more than 30 genes related to biosynthesis, transport, regulation, self-resistance and modification. Due to their antibiotic, antitumor, cholesterollowering, immunosuppressant or antiviral activities, secondary metabolites are invaluable elements of drug discovery research (Vaishnav and Demain, 2011). Approximately 18.000 bioactive secondary metabolites are produced in bacteria, thereof more than 10.000 compounds are synthesized in Streptomyces (Bérdy, 2012). Representatives of the genus Streptomyces have been studied extensively in the last decades (Weber et al., 2015b). A well-known class of bioactive secondary metabolites are polyketides that are assembled by modular megaenzymes called PKSs. The subsequent steps in the assembly process are highly similar to the biosynthesis of fatty acids. A detailed introduction to PKS can be found in Staunton and Weissman (2001). Nonribosomal peptides (NRP) belong to another important class of bioactive compounds. They are synthesized by non-ribosomal peptide synthetase (NRPS) that share certain characteristics with PKS (see Schwarzer et al., 2003 for details).
The genome size of Streptomyces ranges from 8 to 9 Mb and shows a high GC (>70%) content. After sequencing the first Streptomyces genomes it was noticed that unexpectedly, it contained far more secondary metabolite gene clusters than had been predicted earlier from the numbers of previously identified metabolites (Bentley et al., 2002;Ikeda et al., 2003). Under laboratory conditions Streptomyces and other secondary metabolite producers synthesize only a few compounds, whereas more than twenty different secondary metabolite gene clusters are contained within most genomes.
The presence of similar biosynthetic gene clusters in different strains reflects their evolutionary history through vertical as well as horizontal gene transfer from one organism to another, also across species barriers. Individual genes, sub-clusters or whole clusters can be exchanged (Egan et al., 2001;Metsä-Ketelä et al., 2002;Donadio et al., 2005). Consequently, the secondary metabolites from similar clusters may vary as they are often built up of distinct moieties from functional sub-clusters. Therefore, biosynthetic gene clusters are ideal to study evolutionary routes and to gain knowledge of the metabolite's importance to a particular strain (Fischbach et al., 2008).
novel bioactive compounds to meet the growing requirements of modern medicine. Potential approaches to successfully activate a gene cluster are summarized in the following paragraph.
The cloning and heterologous expression of complete clusters is one strategy to get access to the genetic potential of Streptomyces (Gomez-Escribano and Bibb, 2014). In addition, silent secondary metabolite gene clusters can be activated through genetic manipulation, e.g., by overexpression or deletion of proposed global or specific positive or negative regulatory genes (Makitrynskyy et al., 2013;Gessner et al., 2015). Furthermore, the cultivation of a given strain under different fermentation conditions (Bode et al., 2002) or the co-cultivation with bacterial or fungal strains (Schroeckh et al., 2009) might stimulate the expression of silent clusters.
Streptomyces diastatochromogenes Tü6028 is known to produce the antimicrobial compound polyketomycin, a tetracyclic quinone glycoside (Paululat et al., 1999). Recently, we deleted the oxygenase gene pokOIV in this strain, resulting in a polyketomycin non-producing mutant (Daum et al., 2009). However, new natural products were synthesized in this mutant. These metabolites were named foxicins A-D. Possibly, foxicins are also produced by the wild type strain but only in little amounts.
In this study we report on the purification and structural elucidation of foxicin A. The compound consists of an unusual nitrogen-containing quinone moiety linked to a short fatty acid (Figure 1).
Furthermore, we identified the corresponding gene cluster, encoding a type I polyketide synthase (PKS I) and an NRPS in the genome of S. diastatochromogenes Tü6028. By gene disruption we show that this cluster is responsible for the production of foxicin and we propose the biosynthetic pathway of foxicin A. Similar gene clusters were detected in many other Streptomyces genomes. Nevertheless, as no foxicin-like compound has been described so far, we conclude that these clusters are either not expressed or expressed at a very low level under culture conditions.
Foxicin A shows several interesting biological properties: it acts as siderophore binding ferric ions, it shows antibiotic activity and inhibits respiratory electron transfer. The wide distribution of the cluster and the various properties of the compound indicate a major role of foxicin in Streptomyces strains.

Bacterial Culture Condition and Production Analysis
The isolation of polyketomycin of the wild type (wt) strain S. diastatochromogenes Tü6028 has been described previously (Paululat et al., 1999). The mutant S. diastatochromogenes pokOIV (Daum et al., 2009) contains a deletion in pokOIV, a gene encoding an oxygenase involved in the biosynthesis of polyketomycin. Mycelium of the mutant strain S. diastatochromogenes pokOIV was used to inoculate a 300 mL Erlenmeyer flask containing 100 mL of HA medium (yeast extract 0.4%, malt extract 1%, and glucose 0.4% in 1 liter tap water, pH 7.3). The flask was shaken on a rotary shaker (180 rpm) at 28 • C. After 48 h, 3 mL of the pre-culture was used to inoculate a second 300 mL Erlenmeyer flask containing 100 mL of HA medium (main culture). After 6 days foxicin A was isolated.
To determine the time dependency of product formation (foxicin A and polyketomycin) the wt strain was grown in 100 mL HA medium for 6 days. Samples were taken after 0, 24,36,48,60,72,96,120, and 144 h of cultivation.
To check the influence of iron on the production of foxicin A, 0.01, 0.1, or 1 mM of FeCl 3 or FeSO 4 were added to the production media of S. diastatochromogenes pokOIV. The experiment was done in triplicate. The strain was incubated for 6 days and the culture was extracted and analyzed by HPLC. For quantification, the integrals of the corresponding chromatogram peaks were compared.

Isolation of Foxicin A
Mycelium was collected by centrifugation and foxicin A was extracted with acetone (2-3 times the volume of the pellet) by shaking at room temperature for 15 min. After removal of the mycelium by filtration, the extract was evaporated. Finally, this mycelium extract combined with the supernatant was extracted with an equal volume of ethyl acetate. The solvent was removed by evaporation. The crude extract was separated by solid phase chromatography (Oasis HLB 20/35cc) with increasing methanol content (in 10% increments) in the mobile phase. Foxicin A was obtained in the 65 and 70% methanol fraction, foxicin B and C in the 70 and 80% methanol fraction, and foxicin D in the 80% methanol fraction. After thin-layer chromatography in ethyl acetate: formic acid :water (44:3:3) foxicin A (Rf 0.41) was further purified by semipreparative HPLC (Agilent Technologies), equipped with a Zorbax B-C18 (9.4 × 150 mm) pre-column and Zorbax B-C18 (9.4 mm × 20 mm) main column with acetonitrile + 0.5% acetic acid as buffer A and water + 0.5% acetic acid as buffer B and a flow rate of 2 mL/min. A 3-min washing step with 50% buffer A was followed by a 7-min linear gradient from 50 to 70% of buffer A where the substance was collected. The eluate was dried and resulted in a violet powder. The method was completed by a 4-min delay with 95% buffer A and a 4-min reequilibration step with 50% buffer A.

Analysis of Foxicin by HPLC/MS
For analysis, HPLC-MS equipped with a XBridge C18 (3.5 µm; 20 mm × 4.6 mm) precolumn and a XBridge C18 (3.5 µm; 100 mm × 4.6 mm) main column, an UV/visible light detector and a mass spectrometer (Agilent, 1100 Series) was used. A flow rate of 0.5 mL/min was used. A 1-min washing step with 20% buffer A was followed by a 7-min linear gradient from 20 to 60% buffer A and a 16-min linear gradient ranges from 60 to 95% buffer A. After a 5-min delay the method completed with a 1-min gradient from 95 to 20% buffer A and a 5-min reequilibration step with 20% buffer A.
Infrared measurements were carried out on a BRUKER Tensor 27 FT-IR spectrometer equipped with a BRUKER PMA 50 VCD module (Bruker Optik GmbH, Ettlingen). A 100 mM foxicin A solution was prepared in anhydrous CDCl 3 and placed in a BaF 2 cell with a path length of 110 µm. Experimental spectra (4 cm −1 resolution) represent the average of a 6 h measurement in a rotating cell. IR spectra were corrected by subtraction of the solvent spectrum. VCD spectra were background corrected by solvent subtraction and smoothed by Fourier filtering (8 cm −1 resolution). The aperture of the light source was adjusted to a width of 4 mm. Opus 7.0 software (Bruker Corporation) was used to analyze the spectra.

Conformer Search and Quantum Chemical Calculations
The conformer search was carried out at the MMFF level using Spartan 08 (Wavefunction, Inc., Irvine, CA, USA) and gave a set of 84 possible conformers. The five conformers with the highest population (according to Boltzmann weights calculated in respect to relative energies) account for >99% of the calculated Boltzmann distribution. These conformers were chosen for quantum chemical calculations at the DFT level [B3LYP/6-31+G(d,p)] in Gaussian 09, Revision D.01 (Frisch et al., 2013). All calculations were performed in gas phase, vibrational frequencies were uniformly scaled by an empirical factor of 0.975. Theoretical spectra for each geometry were obtained by adding Lorentzian band shapes (width 6 cm −1 ) to the calculated IR and VCD intensities. The dissymmetry factor spectrum defined as VCD over IR absorbance was obtained with CDSpecTech Polavarapu, 2013, 2014).

Bioinformatic Analysis of Foxicin Biosynthetic Gene Cluster and Identification of Similar Biosynthetic Gene Cluster
The draft genome of S. diastatochromogenes Tü6028 was sequenced at the Centrum of Biotechnology, University of Bielefeld (Greule et al., unpublished). Prediction of the gene clusters was performed using antiSMASH 3.0 1 (Weber et al., 2015a). The sequence of foxicin 57.6 kb hybrid PKS I/NRPS biosynthetic gene cluster was further analyzed and annotated using BLAST 2 . Similar gene clusters were identified by antiSMASH and BLAST analysis in other Streptomyces strains. The genomes of these strains were analyzed individually by the above mentioned programs.

Phylogenetic Tree of fox Homologous Clusters
A phylogenetic tree was calculated by Clustal Omega (Sievers et al., 2011) using Neighbor-Joining method of foxBII sequence comparison. For better illustration, the tree is shown without distance correction.

Siderophore Chrome Azurol S (CAS) -Assay
Chrome azurol S medium was modified after Schwyn and Neilands (1987). For the CAS medium 60.5 mg CAS, 72.9 mg hexadecyltrimetyl ammonium bromide, and 30.24 g piperazine-1,4-bis-(2-ethanesulfonic acid) were dissolved in 990 mL water and mixed with 10 mL iron (III) solution (1 mM FeCl 3 × 6H 2 O and 10 mM HCl). Purified foxicin A was pipetted to CAS reagent and a color change of the dark blue solution to violet was noted.

Shift of UV/vis Spectra in the Presence of Iron
Foxicin A was dissolved in MeOH. FeCl 3 and FeSO 4 were added in an amount of 0.5 mM to 100 mM. UV/vis spectra were measured by UviLine 9400 spectrophotometer (SI Analytics).

Isolation of Bacterial Plasma Membranes
Escherichia coli BW25113 cells were grown aerobically (180 rpm) at 37 • C in baffled flasks using LB-medium. The cells were harvested by centrifugation (5700 × g, 10 min, 4 • C, Rotor JLA 8.1000, Avanti J-26 XP, Beckman Coulter) in the late exponential phase yielding approximately 6.5 g cells/L. All further steps were carried out at 4 • C. After centrifugation, 5 g of the cell pellet were resuspended in fourfold volume of buffer 1 (50 mM MES/NaOH, pH 6.0, 50 mM NaCl, 0.1 mM PMSF supplemented with desoxyribonuclease I) and disrupted by passing twice through a French Pressure Cell Press (110 MPa, SLM-Aminco). Cell debris and non-disrupted cells were removed by centrifugation (9500 × g, 20 min, 4 • C, Rotor A8.24, RC-5 Superspeed Refrigerated Centrifuge, Sorvall Instruments). The cytoplasmic membranes were obtained from the supernatant by centrifugation at 252000 × g (60 min, 4 • C, Rotor 70.1Ti, L8-M Ultrafuge, Beckman). The sediment was suspended in an equal volume (1:1, w/v) of buffer 1 and was used directly or frozen in liquid nitrogen and stored at −80 • C.

Determination of NADH Oxidase Activity
The NADH oxidase activity of cytoplasmic membranes was measured with a Clark-type oxygen electrode (RE K1-1, Oxytec) at 30 • C. To calibrate the electrode 2 mL 50 mM MES/NaOH, pH 6.0, 50 mM NaCl, 5 mM MgCl 2 were deoxygenized by adding sodium dithionite and the signal was set to 237 µM oxygen (Weiss, 1970). Each measurement was performed with 2 mL buffer containing 5 µL of the membrane suspension at 30 • C. The reaction was started by adding 1.25 mM NADH. 50-500 µM foxicin A was added to the assay to test its inhibitory action on cell respiration.
To assay the herbicidal property, 0.1, 1, and 5 µM of foxicin A was added to 6 mg of Arabidopsis thaliana wt ecotype Wassilewskija seeds that were plated onto plates. The plants were grown for 2 weeks in a phytochamber with long-day conditions (16/8 h), 100 µE m −2 s −1 light intensity and 25 • C constant temperature.
To check for H 2 O 2 sensitivity, 2, 5, and 10 µL of 5% H 2 O 2 solution were pipetted on paper disks (6 mm diameter) and placed on MS and TSB culture plates of S. diastatochromogenes WT, pokOIV mutant and pokOIV/foxBII::pKC1132 mutant. The strains were incubated at 28 • C for 5 days and inhibition zone was measured.

Cell Viability Assay (MTT-Assay)
The effect on cell viability of foxicin A was tested against cancer cell lines CCRF-CEM, CEM-ADR5000 and Jurkat cells and peripheral blood mononuclear cells (PBMC) using the MTT assay as previously described (Calderón et al., 2014). In brief, cells were seeded in 96-well plates at a density of 4 × 10 4 cells/well and incubated for 24 h with various concentrations of foxicin A. The chemotherapeutic agent doxorubicin was used as a positive control, and DMSO (0.1%) was the solvent control. The data are expressed as the mean ± SD of three independent experiments.

Production and Isolation of Foxicin A from S. diastatochromogenes Tü6028
Streptomyces diastatochromogenes Tü6028 is known to synthesize polyketomycin, which is produced at high levels after 96 h of cultivation in HA medium. During our studies on polyketomycin biosynthesis we deleted the oxygenase gene pokOIV (Daum et al., 2009). The mutant failed to produce polyketomycin, instead we observed the accumulation of novel compounds, which we named foxicin A, B, C, and D (minor compounds) (Figure 2A).
A careful analysis of extracts of the wild type strain showed that foxicins are also produced, but with significantly lower titers. Foxicin production in the wt strain reaches a maximum after 48 h of incubation ( Figure 2B). After 96 h less than 5% of the initial foxicin A concentration was detected. In S. diastatochromogenes pokOIV, foxicin production reaches its maximum after 6 days. The mutant also produces fewer spores and visibly lower amounts of melanin, as indicated by the color of the cultivation medium ( Figure 2C). Cultivation of S. diastatochromogenes pokOIV in 5 L of HA production medium yielded 9.8 mg of foxicin A (1.96 mg/L) and 3.4 mg of foxicin B (0.68 mg/L), and even lower amounts of foxicins C and D.  Figure S1). Foxicin A shows absorption maxima at 246, 280, 316, 390 and a small peak at 462 nm.

Physiocochemical Properties and Structure Elucidation of Foxicin A
NMR-data of foxicin A were recorded in CDCl 3 and DMSOd 6 . In Supplementary Tables S2 and S3, NMR assignments of 1D NMR ( 1 H, 13 C) and 2D NMR ( 1 H-1 H-COSY, HSQC, H2BC, and HMBC) experiments are summarized and Supplementary  Figures S2-S14 show the respective spectra. The 13 C NMR spectrum shows 18 carbon signals, which could be assigned to five methyl, three methine groups and nine quarternary carbon atoms by the use of HSQC. A 1,4-benzoquinone system was identified from typical carbonyl chemical shifts (δ C = 182.3 and δ C = 178.6 ppm). The benzoquinone is substituted with a hydroxyl group in position C-2 and an amino-acetate at position C-3, as established by HMBC correlations C-2/6-H, C-2/1 -NH,  and C-4/1 -NH. Moreover, an additional side chain is attached at C-5 via an amide functionality, as proven by the HMBC signals C-4/1 -NH and C-6/1 -NH. The side chain contains two double bonds, which are both trisubstituted. One double bond is in conjugation to the amide carbonyl as proven by the HMBC signal C-2 /4 -H and bears a methyl group at C-3 as shown by the HMBC cross peaks C-2 /3 -CH 3 , 3 -CH 3 /4 -H, this double bond is in E configuration as proven by the ROESY signal 3 -CH 3 /5 -H. The second double bond was determined as a C-6 = C-7 double bond with two methyl groups at C-7 based on HMBC correlations. The two double bonds are connected via the C-5 methine group, which is methyl substituted as established by COSY couplings 4 -H/5 -H, 5 -H/6 -H and H2BC signals C -5/4 -H and C-5 /6 -H. The important 2D correlations are shown in Figure 3. Based on our 1-and 2D NMR data, the absolute configuration of foxicin A could not be solved unambiguously. Therefore, we analyzed this compound by comparing vibrational circular dichroism (VCD) and infrared (IR) spectra to spectra from quantum chemical calculations. A molecular model of foxicin A was constructed and subjected to a conformer search algorithm employing molecular mechanics (MMFF). The conformer models were then subjected to a geometry optimization at the DFT level [B3LYP/6-31+G(d,p)] and the relative energies and IR absorbance and VCD intensities were calculated. The comparison of the Boltzmann-weighted average of the spectra calculated for five conformers of foxicin A and the experimental VCD spectrum (Figure 4) allowed for the assignment of the absolute conformation as well as the configuration of foxicin A as (S)-2-hydroxy-3-(acetylamino)-5-(3 , 5 S, 7 -trimethyl-hepta-3 E, 6 -dienoylamino)-1,4-benzoquinone. The structure of foxicin A is shown in Figure 1.
Foxicin B has the same mass (346 g/mol) as foxicin A. Foxicin C and foxicin D are only produced in low amounts. Both compounds have a molecular weight of M = 360 g/mol, based on the deprotonated molecular ion peak at m/z 359.2 in the negative ion mode CI-MS spectrum.

Identification and Sequence Analysis of the Biosynthetic Gene Cluster
Due to its chemical structure, foxicin A is likely to be a product of a type I polyketide synthase (PKSI) and a non-ribosomal peptide synthetase (NRPS). Bioinformatics analysis of the 7.9 Mb draft FIGURE 4 | Infrared and VCD spectra of foxicin A. (A) Experimental IR spectra recorded for foxicin A (red) overlayed with IR spectra calculated at the B3LYP/6-31+G(d,p) level for the ortho-quinone (black) and for the para-quinone (blue) with considerably better agreement for para (spectra are offset for better comparison); (B) Experimental (red) VCD and dissymmetry factor spectra in comparison to VCD and dissymmetry factor spectra calculated for (S)-foxicin (blue). The good overall agreement allows for assignment of the S-configuration. genome sequence of S. diastatochromogenes Tü6028 revealed 23 putative secondary metabolite gene clusters, but only one cluster containing genes with both PKSI-and NRPS function. The cluster was assigned to 57.6 kb with an overall GC content of 72.4%. The annotation analysis revealed 41 open reading frames (ORFs) putatively involved in foxicin A-D biosynthesis ( Table 1). The genetic organization of the biosynthetic gene cluster (fox gene cluster) is shown in Figure 5A. The GenBank accession number of the nucleotide sequence is KT440882. In order to verify the correct assignment of the fox gene cluster, we constructed the inactivation plasmid pKC1132_SC_foxBII containing a 2 kb homologous region of foxBII. Conjugation between E. coli ET12567 and S. diastatochromogenes Tü6028 pokOIV and integration of the plasmid into foxBII by single crossover recombination resulted in apramycin-resistant mutants. Integration of the plasmid into foxBII by single crossover recombination was confirmed by PCR. Loss of ability of the mutant strain to produce foxicins confirmed the correct assignment of the fox gene cluster.

Characterization of the Deduced Amino Acid Sequences and Putative Model of the Biosynthesis of Foxicins
Based on bioinformatics analysis of the fox gene cluster of S. diastatochromogenes Tü6028, a putative biosynthetic pathway ATs show specificity for methylmalonyl-CoA as predicted using antiSMASH (Weber et al., 2015a), which is in line with the structure of foxicin A. Foxicin A possesses a double bond between the starter unit and the first extender unit in β,γ-position. Similar structural elements are known from ansamitocin-, rhizoxin-, bacillaen and corallopyronin. It has been shown that special dehydratases or additional shift modules are responsible for the double bond shift in β,γ-position during the biosynthesis of these molecules (Taft et al., 2009;Kusebauch et al., 2010;Moldenhauer et al., 2010;Lohr et al., 2013). In some DH domains, the conserved motif is mutated. The DH domain of module I of the fox cluster shows in contrast to other DH domains a motif of HxxxGxxxxS instead of the conserved HxxxGxxxxP motif, indicating that this domain could introduce the β,γ-double bond during foxicins' biosynthesis.
The FkbH-like domain is likely to incorporate a glyceryl moiety (Chan et al., 2006;Dorrestein et al., 2006;Sun et al., 2008). The three genes foxGI, foxGII and foxGIII, also located in the fox biosynthetic gene cluster, encode for enzymes known to be involved in sugar metabolism. The proposed functions of the enzymes are glyceraldehyde-3-phosphate dehydrogenase, sugar isomerase and aldolase. Most likely the three of them provide bisphosphoglycerate, which is transferred onto the carrier protein of FoxBIII.
The adenylation domain (A) might be encoded by foxBI, which is separated from foxBIII by the PKS gene foxBII. In silico analysis did not indicate an A domain specificity, but based on the structure of foxicin A, a non-proteinogenic amino acid with two amide groups similar to 2,4-diamino-3-oxobutanoic acid might play a role. This moiety is then most likely linked to the glyceryl-CP.
We propose that the biosynthesis starts with the PKS I of FoxBII and the generated polyketide chain is then transferred onto the amino acid bound to the glyceryl-CP of FoxBIII. Finally, the molecule is cleaved from the enzyme by the thioesterase domain (TE) of FoxBIII, followed by the ring formation. The responsible enzyme for the ring formation is unknown, but with BLAST analysis we identified candidate genes as foxEI or foxEII, with proposed esterase activity. N-acetylation catalyzed by FoxA results ultimately in foxicin A (Figure 6). After cleavage from the enzyme complex foxicin A gets further modified.
At each end of the fox gene cluster two ABC transporter genes are located (foxTI-TIII and foxTVIII-X). In addition, foxTIV encodes a transmembrane efflux protein, foxTV an integral membrane transport protein, foxTVI a transporter belonging to the EmrB/QacA family and foxTVII, an export protein. The biosynthetic gene cluster of foxicins also includes 10 regulatory proteins (FoxRI-FoxRX), indicating a complex regulation of the outlined biosynthesis.

Similar fox Biosynthetic Gene Clusters in Various Streptomyces Strains
Similar fox gene clusters were identified in the genomes of 21 additional Streptomyces strains (Figures 5C,D): S. aureus NRRL  (Matsunaga et al., 2015) and Streptomyces sp. OK006 (Klingeman et al., 2015). In these strains homologs of foxBI, foxBII, foxBIII, foxEI, and foxEII are located next to each other with high sequence identities of 73% up to 96% (Supplementary Table S4). The closest homologous cluster is located in Streptomyces sp. 142MFCol3.1. A phylogenetic tree based on foxBII sequence analysis is shown in Figure 5B. In all strains the NRPS/PKS I hydride enzyme complex (foxBI -foxBIII) consists of the same  (Figures 5E,F). The strains Streptomyces sp. JHA19, S. curacoi and S. regalis have an additional PKS/NRPS hybrid gene upstream of foxBI-III with [A-PCP-KS-AT-KR-ACP-C-A-PCP-C] catalytic domains. The phylogenetic tree that is based on foxBII illustrates that the biosynthetic gene clusters of these strains do not originate from one clade.
In most of the strains foxEI and foxEII reside adjacent to foxBI-foxBIII indicating that they might play an important role in the biosynthesis of these secondary metabolites. Additionally, in most of the clusters genes from sugar and amino acid metabolism, as well as methyl-/acetyl transferases were identified.
The described genes in the predicted fox clusters do not agree in detail, therefore we assume that the corresponding compounds might exhibit structural differences. Considering the discrete array of catalytic domains, however, we expect similar polyketide tails in all foxicins A-D-like substances. To the best of our knowledge, none of the mentioned strains FIGURE 6 | Proposed biosynthetic pathway of foxicin. FoxBII is proposed to be involved in the formation of the polyketide side chain (blue), FoxBI and FoxBIII in the incorporation of an amino acid (dark red) and FoxGI-III with FkbH domain of FoxBII in the incorporation of a C2 sugar moiety (green). N-acetylation might be catalyzed by FoxA (magenta). synthesizes a compound related to foxicins. The responsible gene clusters seem to be silent or have, until now, not been studied. In Supplementary Table S4, an overview of the described Streptomyces strains is shown, as well as their known secondary metabolites and the percentage identity of the identified genes with foxBI, foxBII, foxBIII, foxEI, and foxEII.

Activity of Foxicin A
In order to understand the function of foxicin A we attempted to characterize this molecule in greater detail. By means of LC-MS analysis we could show that Foxicin A production in S. diastatochromogenes is inhibited by ferric ions supplemented to the growth medium. The supplementation of either 1 mM FeCl 3 or FeSO 4 to the production medium led to a more than 75-fold decrease in foxicin A formation in S. diastatochromogenes pokOIV ( Figure 7A). The addition of FeCl 3 to foxicin A resulted in a shift of the UV/vis maximum at 320-350 nm ( Figure 7B). Using the CAS assay (Schwyn and Neilands, 1987), we observed a color change from blue to violet (Figure 7C), indicating that foxicin A acts as a siderophore.
As foxicin A contains a quinone moiety similar to molecules involved in various redox processes within the cell (Brunmark and Cadenas, 1989), the compound was tested for its property as an electron acceptor or inhibitor of the respiratory chains. Foxicin A was added to E. coli cytoplasmic membranes and the respiratory chain activity was measured with a Clark-type electrode by starting the reaction with NADH as electron donor. 500 µM foxicin A blocked O 2 consumption to 30% indicating its role as a weak inhibitor of the electron transfer chain (Supplementary Table S5). In contrast to other quinones (Kawamukai, 2002) Figure S15).

foxicin A does not protect cells from oxidative stress induced by H 2 O 2 (Supplementary
To identify the antibiotic potential of foxicin A, an inhibition zone test was conducted. Therefore, 100 µg foxicin A were applied to paper disks that were transferred to different culture plates. The compound showed moderate activity against Streptomyces viridochromogenes, Saccharothrix espanaensis and the cyanobacterial strains Synechococcus sp. and Synechocystis sp. The tested amount of foxicin A did not visibly influence the growth of the bacterial strains E. coli XL1-Blue, Bacillus subtilis, Mycobacterium smegmatis, the fungal strains Candida parapsilosis and Fusarium verticillioides (Supplementary Table  S6) and the plant Arabidopsis thaliana. To determine the effect of foxicin A on cell viability of human cells, leukemia cell lines CCRF-CEM, CEM-ADR5000 and Jurkat cells as well as noncancer PBMCs were stimulated for 24 h with foxicin A at a concentration range of 0.6-80 µM. However, foxicin A showed no significant effect on cell viability, as measured by using the MTT assay (Supplementary Figure S16).

DISCUSSION
As a result of the deletion of the structural gene pokOIV of polyketomycin biosynthesis, we identified novel secondary metabolites in S. diastatochromogenes Tü6028, named foxicins. In the wild type strain, the gene cluster is expressed only at low levels. In contrast, the pokOIV mutant produces higher amounts of the foxicin derivatives A-D, enabling further investigations of these fascinating compounds. Noteworthy, the mutant strain produces less spores and melanin, indicating a crucial role of polyketomycin for the differentiation and stress response.
The structure of foxicin A could not be completely elucidated by NMR analysis. Eventually, the combination of NMR, IR, and VCD spectra analysis led to the solution of the structure of foxicin A. Foxicin A shows several unusual features. It has a para-quinone moiety, with two amide groups on each site, one is further acetylated and a short fatty acid side chain with non-conjugated double bonds.
Quinone moieties are commonly found in natural products. One major group are the ubiquinones, for example, being important electron carriers in respiration and photosynthesis. Furthermore, they are involved in all kinds of redox reactions and play a crucial role as antioxidants (Kawamukai, 2002). Their long isoprenoid chains lead to the ability to penetrate biological membranes. The short polyketide tail of foxicin A indicates that the compound is not located in the membrane. Idebenone, a synthetic quinone with similarities to ubiquinones, but with a much shorter, less lipophilic tail is predominantly active in the cytoplasm and not in cellular membranes. It is a potent antioxidant, prevents lipid peroxidation and protects against ROS-induced damage in multiple systems (Suno and Nagaoka, 1984;Sugiyama et al., 1985;Rauchová et al., 2006). Foxicin A has even a shorter chain than idebenone. The presence of foxicin A in the culture medium supports the assumption that foxicin A activity is not associated with membrane binding.
Like other quinones, foxicin A is able to accept electrons, but does not act as an antioxidant. In vitro, foxicin A inhibits respiratory function in E. coli membranes, without affecting the in vivo viability of the strain. Foxicin A obstructs the growth of other actinomycetes strains, as well as cyanobacterial species. In contrast, the molecule does not visibly influence the growth of Arabidopsis thaliana. Therefore, it may also interact with the photosynthesis machinery, but cannot pass through all types of cell walls. Further studies are needed to support this hypothesis.
The main function of the foxicins is most likely explained by their ability to act as a siderophore. Surprisingly, foxicin A does not inhibit the viability of human cells, even though it is able to bind ions from the medium. Ions are essential for all organisms. The lack of ions, especially ferric ions, often limits the growth of bacteria in their natural habitat. Therefore, siderophores are vital molecules that are released into the medium, and, after scavenging ions, are actively transported back into the cell. The strain S. diastatochromogenes pokOIV produces less foxicins in the presence of ferric ions in the production media. Foxicin A interacts directly with iron as shown by the CAS assay and a shift of the absorbance maximum in the UV/vis spectrum.
Investigations on siderophores are nowadays in the focus of many research groups, in order to obtain new antibacterial compounds, by employing the 'trojan horse' strategy. The outer membrane is an important barrier of Gram-negative bacteria, as well as of mycobacteria. Diarra et al. (1996) and Möllmann et al. (2009) have intriguingly shown that the linkage of an antibiotic to a siderophore can lead to facilitated transport via specific transporters into the cell with subsequent death of the pathogens.
The siderophore yersiniabactin has been found in the plague bacterium Yersinia (Pelludat et al., 1998) and in other bacteria such as the nematode symbiont Photorhabdus luminescens (Duchaud et al., 2003), the plant pathogen Pseudomonas syringae, pathogenic strains of E. coli (Bultreys et al., 2006) and even in the Gram-positive marine bacterium Salinispora tropica (Udwary et al., 2007). The ability to synthesize siderophores gives special benefits to a strain and the evolutionary driving force to keep the biosynthetic gene clusters. The foxicin cluster was identified in more than twenty additional Streptomyces strains, which were isolated at different places around the world, indicating an evolutionary early origin. Until now (11/2016) genome sequences of 844 Streptomyces strains are available on NCBI. This means that the cluster is present in about 2.6% of all Streptomyces strains. It is anticipated that further genome sequencing of many more Streptomyces strains will reveal additional homologous fox biosynthetic gene clusters. Surprisingly, the fox cluster was not detected in any other actinomycetes genus. Although horizontal gene transfer is common in the Streptomyces genus, it is unexpected that the cluster is found in that many other strains. In addition, the organization of the structural genes foxBI, foxBII, foxBIII, foxEI, and foxEII remained the same with high sequence similarity. Therefore, we assume that the respective products should only by slightly different to foxicin A-D. In the three strains comprising of additional PKS/NRPS genes, the product might be more complex, e.g., possessing a second polyketide chain. Many of the investigated strains are known producers of secondary metabolites, but not of compounds similar to foxicins. Therefore, the identified clusters seem to be silent in these strains or the compound is produced only in very little amounts. As a consequence, its presence could have been overlooked in routine natural compound screening, as it had been the case for polyketomycin. The high similarity of the clusters indicates an evolutionary driving force to keep the biosynthetic gene clusters in place and consequently a major role of the compounds for Streptomyces strains.

AUTHOR CONTRIBUTIONS
AG designed and performed experiments, analyzed data, proposed the biosynthesis model and wrote the paper; AG, IP, and SZ purified foxicin A; DD performed CAS assay; MaM and SL conducted and analyzed VCD and IR measurements; SZ, CJ-T, TP, S-ML, MM, and PB interpreted NMR data, SB performed in vitro respiratory assay; CDF conducted MTT assay; AB, IM, and TF administered the experiments; all authors have given approval to the final version of the manuscript.

FUNDING
This work was supported by the Deutsche Forschungsgemeinschaft (RTG 1976) and China Scholarship Council.