Identification of Novel Rotihibin Analogues in Streptomyces scabies, Including Discovery of Its Biosynthetic Gene Cluster

ABSTRACT Streptomyces scabies is a phytopathogen associated with common scab disease. This is mainly attributed to its ability to produce the phytotoxin thaxtomin A, the biosynthesis of which is triggered by cellobiose. During a survey of other metabolites released in the presence of cellobiose, we discovered additional compounds in the thaxtomin-containing extract from Streptomyces scabies. Structural analysis by mass spectrometry (MS) and nuclear magnetic resonance (NMR) revealed that these compounds are amino acid sequence variants of the TOR (target of rapamycin) kinase (TORK) pathway-inhibitory lipopeptide rotihibin A, and the main compounds were named rotihibins C and D. In contrast to thaxtomin, the production of rotihibins C and D was also elicited in the presence of glucose, indicating different regulation of their biosynthesis. Through a combination of shotgun and targeted proteomics, the putative rotihibin biosynthetic gene cluster rth was identified in the publicly available genome of S. scabies 87-22. This cluster spans 33 kbp and encodes 2 different nonribosomal peptide synthetases (NRPSs) and 12 additional enzymes. Homologous rth biosynthetic gene clusters were found in other publicly available and complete actinomycete genomes. Rotihibins C and D display herbicidal activity against Lemna minor and Arabidopsis thaliana at low concentrations, shown by monitoring the effects on growth and the maximal photochemistry efficiency of photosystem II. IMPORTANCE Rotihibins A and B are plant growth inhibitors acting on the TORK pathway. We report the isolation and characterization of new sequence analogues of rotihibin from Streptomyces scabies, a major cause of common scab in potato and other tuber and root vegetables. By combining proteomics data with genomic analysis, we found a cryptic biosynthetic gene cluster coding for enzyme machinery capable of rotihibin production. This work may lead to the biotechnological production of variants of this lipopeptide to investigate the exact mechanism by which it can target the plant TORK pathway in Arabidopsis thaliana. In addition, bioinformatics revealed the existence of other variants in plant-associated Streptomyces strains, both pathogenic and nonpathogenic species, raising new questions about the actual function of this lipopeptide. The discovery of a module in the nonribosomal peptide synthetase (NRPS) that incorporates the unusual citrulline residue may improve the prediction of peptides encoded by cryptic NRPS gene clusters.

In the last decades, several studies focused on elucidating the molecular mechanisms of virulence of S. scabies. The production of thaxtomin A has been shown to be a key pathogenicity determinant (6). The txtABCDEH gene cluster is responsible for the biosynthesis of this 4-nitroindol-3-yl-containing 2,5-dioxo-piperazine. These thaxtomin biosynthetic genes are highly conserved across plant-pathogenic streptomycetes and reside on a pathogenicity island that is mobilized in some species (7). Thaxtomin A primarily targets expanding host tissue by affecting cellulose synthase complex density, expression of cell wall genes, and cell wall composition (8,9).
The production of thaxtomin A is strictly controlled, involving several layers of regulation. Cellobiose, together with cellotriose, is recognized as the main specific elicitor of thaxtomin A biosynthesis in S. scabies (10). Once imported via the CebEFG-MsiK ABC transporter (11), these products of cellulose turnover directly target the pathway-specific transcriptional activator TxtR and the cellulose utilization regulator CebR, which together constitute a double-locking system on the txtABCDE gene cluster. Specific interaction between CebR and cellobiose triggers the release of the repressor from different binding sites within the thaxtomin biosynthetic gene cluster (BGC), including txtR. This results in the transcriptional activation of txtA and txtB, consequently inducing thaxtomin A production and pathogenicity (12,13). In addition, several bld global regulators directing secondary metabolism or morphological differentiation are involved in the regulation of thaxtomin synthesis (14).
It is believed that plant-pathogenic streptomycetes produce other important phytotoxins involved in pathogenicity. Using proteomics, we previously demonstrated that the levels of enzymes involved in the biosynthesis of other secondary metabolites like concanamycin A and coronafacoyl phytotoxins are also dependent on cellobiose levels, a finding later confirmed by measuring altered levels of these compounds in the presence of cellobiose and/or upon cebR deletion (15). Natsume et al. isolated concanamycins A and B from Japanese S. scabies isolates (16). Recently, they demonstrated root growth-inhibitory activity, necrosis-inducing activity, and a synergistic effect with thaxtomin A (17). Coronafacoyl phytotoxins contribute to the development of root disease symptoms and cause hypertrophy of potato tuber tissue (18). The causative agent of russet scab, S. cheloniumii, produces FD-891, which induces necrosis of potato tuber tissue (19). All these data indicate that multiple secondary metabolites are involved in scab disease. Moreover, thaxtomin A-deficient streptomycetes that are also able to cause scab disease have been isolated. Streptomyces sp. strain GK18 produces borrelidin, which is reported to cause severe deep, black holes on potato tuber slices (20,21). Similarly, fridamycin E was isolated from an S. turgidiscabies strain from a netted scab lesion in Sweden. This phytotoxin was demonstrated to reduce or even inhibit the sprouting of in vitro microtubers (22).
Driven by the discovery that under thaxtomin production-inducing conditions, i.e., the addition of cellobiose, multiple proteins potentially involved in secondary metabolism were increased in abundance, we further analyzed extracts from the extracellular medium of Streptomyces scabies RL-34 in order to obtain insight into the metabolites secreted by this organism when grown in the presence of cellobiose. We report here the discovery of two new compounds that were characterized to be variants of the lipopeptide rotihibins A and B, previously discovered by Fukuchi et al. in extracellular extracts of Streptomyces sp. strain 3C02, later designated an S. graminofaciens strain (23,24). A culture filtrate containing rotihibins A and B inhibits the growth of lettuce seedlings, while purified rotihibin A causes shoot stunting in tobacco seedlings at low concentrations (25). Recently, it was demonstrated that rotihibin A acts as a TOR (target of rapamycin) kinase (TORK) pathway inhibitor (26). Therefore, we assessed the plant growth-inhibitory effect of the newly isolated compounds, designated rotihibins C and D, and found a severe effect on growth and photosystem II photochemistry efficiency in Lemna minor L. and Arabidopsis thaliana L. Heynh. Furthermore, we provide data that demonstrate that the biosynthetic machinery to produce rotihibins C and D is encoded by a gene cluster covering a 33-kb segment containing 14 open reading frames (ORFs) that is conserved in both pathogenic and nonpathogenic plant-associated Streptomyces species.

RESULTS AND DISCUSSION
Rotihibin production by Streptomyces scabies. Cellobiose is known to induce thaxtomin A production in Streptomyces scabies (10). We postulated that cellobiose also elicits the production of other virulence factors important for infection of root and tuber crops. When comparing the high-performance liquid chromatography (HPLC) profiles of n-butanol extracts of media from cultures obtained from S. scabies RL-34 grown in International Streptomyces Project medium 4 (ISP-4) either supplemented or not with 0.7% cellobiose, new peaks appeared in the HPLC profile of the extract obtained after growth in the presence of cellobiose (Fig. 1). We focused further analysis on the peaks eluting at 5.52, 5.75, and 6.10 min, and the corresponding metabolites were characterized by highresolution electrospray ionization mass spectrometry (HR-ESI-MS). A distinct ion peak at m/z 439. 16 1 (data not shown). This pattern of m/z values can be attributed to thaxtomin A (theoretical monoisotopic mass, 438.11), confirming that our culture conditions induced the production of the main pathogenic determinant of S. scabies.
The ESI spectrum of the fraction eluting at 5.52 min displayed a major peak at m/z 860.47, while the spectrum of the fraction eluting at 5.75 min displayed a peak at m/z 874.47, which is also present in the 5.52-min sample (see Fig. S1 in the supplemental   (Fig. 2). The bioinformatics tool Insilico Peptidic Natural Products Dereplicator was used to dereplicate the structure through database searching of mass spectra (27). The mass spectrum of rotihibin A, a nonribosomal peptide (NRP) functional as a plant growth regulator in Streptomyces graminofaciens, was found to have the best fit to our data. However, the spectrum of our newly detected compound from S. scabies indicated the presence of a threonine in the NRP backbone instead of a serine (as in rotihibin A) ( Fig. 2A) (25). Indeed, the neutral losses of 118, 130, 302, 309, 101, and 201 mass units correspond to the masses of asparaginol (Asn-ol); hydroxy-asparagine (OH-Asn); (allo)-threonine (aThr), 2,4-diamino butyric acid (Dab), and threonine (Thr); 2-cis-decenoic acid (cis-DA) and citrulline (Cit); threonine (Thr); and aThr and 2,4diamino butyric acid, respectively (Fig. 2B). Previously, Halder et al. established a chemical synthesis pathway of rotihibin A and structural analogues (26). In their structural analogue RotA-D3, the serine residue at position 2 of the amino acid chain was replaced by alanine. This derivative turned out to be more active than rotihibin A, as shown by the overall more pronounced plant growth retardation in an A. thaliana bioassay (26). Interestingly, our data suggest that S. scabies RL-34 would produce a natural rotihibin variant that is modified at the same position.
Fragmentation of the ion at m/z 874.47 displayed the same fragment ions, except for the daughter ions containing the acyl chain (m/z 114), which could be explained by a longer/branched acyl chain. For further description, we named the compound at m/z 860 rotihibin C and the compound at m/z 874 rotihibin D. For each of the compounds, we also see a component that is 2 Da smaller, which we assume to be due to desaturation. We also see a minor component at m/z 844/846, which, according to the MS/MS spectrum (data not shown), is a similar compound with a shorter fatty acyl chain.
Glucose is known to suppress thaxtomin A production in Streptomyces scabies and Streptomyces acidiscabies (28,29). The effect of glucose, a breakdown product of cellobiose, on the production of the rotihibin analogues was tested. Indeed, the peak eluting at 6.1 min, corresponding to thaxtomin A, disappeared in the HPLC profile of the S. scabies RL-34 cultures grown in the presence of glucose. In contrast, compounds eluting at 5.52 and 5.75 min were still present (Fig. 3), and they were confirmed by directinfusion HR-ESI-MS/MS to be rotihibins C and D (data not shown). This is a clear indication that cellobiose is not the trigger for rotihibin production in S. scabies RL-34 as it is for thaxtomin A.
Finally, we verified whether rotihibins C and D are also produced by other S. scabies strains. We performed similar extraction and purification methods on extracellular medium from the hypervirulent S. scabies 87-22 strain (12). The chromatograms and ESI spectra of these components were identical, indicating that this strain is also able to produce the same compounds (data not shown).
To confirm the structure proposed from the MS data, HPLC-purified rotihibins C and D were characterized using 1 H nuclear magnetic resonance (NMR). Superposition of the two-dimensional (2D) total correlation spectroscopy (TOCSY) spectra of both compounds as depicted in Fig. 4 clearly indicates the presence of identical correlation patterns. In addition, analysis of the various spin systems in the TOCSY spectrum showed good agreement with the amino acid residues expected from the MS/MS analysis of the peptide chains of rotihibins C and D. Specifically, it agrees with our suggestion that a threonine in rotihibin C replaces the serine observed in rotihibin A. Since the NMR spectra of purified rotihibins C and D did not reveal structural differences between the peptide chains, the NMR analysis also supports that the 14-Da difference in mass likely results from an extra methylene group for the acyl chain in rotihibin D; however, the spectral overlap between the many CH 2 units does not allow us to clearly establish acyl chain length using NMR.
Identification of the biosynthetic gene cluster responsible for rotihibin production. Apart from the thaxtomin biosynthetic gene cluster (BGC), S. scabies also possesses four other NRP synthase (NRPS)-type BGCs; two of them are cryptic, that is, BGCs for which no biomolecule has yet been associated with the genetic material. Proteins associated with a cryptic NRPS-type BGC (from scab_3221 to scab_3351) (see details below) were found in a previous proteomic experiment (15). We investigated the possible relationship between this BGC and rotihibin production by targeted proteomics using multiple-reaction monitoring (MRM). The proteotypic peptides of SCAB_3241 (LIDEEPYR and ATGLSDEEFLAR), SCAB_3251 (IPVYLAALGPK and IDVGSAVLQIPAR), SCAB_3281 (VTDEQLAALDLSR and EDPLLTDALAGQR), SCAB_3291 (GQLPEGAWR and LGTADLWLR), SCAB_3301 (LYGGAATDIPHVR and SELAGVFADLLR), and SCAB_3321 (AWIDSDLATPVPVTGER, ADTSGDPTFEELLDR, and GGTVPFAVPAALR) were used to evaluate the protein levels. Figure 5 shows the positive correlation between the presence of either glucose or cellobiose and the production of the selected proteins, thereby providing the first evidence that rotihibin production and this cryptic BGC might be linked. Interestingly, the inactivation of the cellulose utilization regulator cebR had no effect on the production of enzymes for this cluster or even instead had the reverse effect for the scab_3221 gene product (15). This indicates that the effect of cellobiose on the expression of the NRPS gene cluster is not dependent on cebR and that there is no coregulation with thaxtomin production.
We then performed a bioinformatic analysis of this gene cluster to investigate whether it effectively contains the information for enzyme machinery allowing rotihibin production. We first analyzed the cryptic gene cluster using antiSMASH (30). It resides on a 70.1-kb segment of the S. scabies genome with 51 open reading frames (ORFs). The region around the largest NRPS gene (scab_3321) was compared with all public genomes using PATRIC (31). This analysis confined the cluster to a 33-kb segment with 14 ORFs (scab_3221 to scab_3351). The presence of a putative integrin protein (SCAB_3201) and a putative transposase (SCAB_3371) at the borders of this cluster is evidence for the possible mobilization of this gene cluster. The sequences of these 14 ORFs were individually submitted to a BLAST search against the UniProtKB/Swiss-Prot database, and putative functions derived from this search are listed in Table 1 and Fig. 6. A preliminary analysis of these functions allowed us to conclude that this gene cluster contains the necessary information to produce the enzymes needed for the biosynthesis of rotihibins C and D, as outlined below. Therefore, this cluster is further referred to as the rotihibin (rth) gene cluster.
Analysis of the genes involved in the rotihibin BGC. (i) NRPS genes for peptide chain assembly. Two NRPSs, RthA and RthB, are present in this cryptic gene cluster. RthA is predicted to contain 5 modules, while RthB contains a single module for the incorporation of amino acids in a peptide chain (Fig. 7A). Different tools were used to predict their NRPS adenylation domain (A domain) specificity (Table 2), and from these results, we propose a biosynthetic path for the production of the rotihibin peptide chain (Fig. 7B). Module 1 of RthA was predicted to be specific for Glu or Ser, depending on the tool used. However, we believe that module 1 is responsible for the incorporation of citrulline, the first amino acid of the peptide: the signature sequence for an A domain that could be specific for citrulline is possibly not included in the prediction tools. The A domain was missing from module 2, while (allo)-threonine should be built into the rotihibin backbone. RthB is the best candidate to introduce threonine into the rotihibin peptide sequence, as its A domain is predicted to be specific for threonine. This phenomenon was previously described in the enduracin biosynthetic gene cluster from Streptomyces fungicidicus. Another NRPS activates and transfers L-allo-threonine to the module with the missing A domain (32). RthD, a type II thioesterase-like (TEII) enzyme, is believed to assist in the shuttling of the activated aThr between the stand-alone A-T didomain module RthB and the A-less C-T module of RthA, as described previously for WS9326A biosynthesis (33). A serine residue was predicted in module 3, but (2,4)-di-amino butyric acid was observed. The prediction of the two last modules to introduce asparagine residues is in line with the rotihibin structure having two asparagine-like residues at its C terminus. Module 5 has a reductase (RE) domain. This domain is known to reduce the peptidyl thioester into its corresponding alcohol, which explains the presence of L-asparaginol (34). This also explains why rotihibin does not form a cyclic structure like many other NRPS-derived lipopeptides. Potentially, RthB has a dual function by transferring L-allo-threonine to module 2 and building an additional L-allo-threonine into the NRP backbone by forming an iso-peptide bond with (2,4)-di-amino butyric acid. This was also described for viomycin and capreomycin, where one A domain is proposed to function twice, to acetylate not only the T domain within its own module but also the T domain of another module (35,36).
The stereochemistry of the different amino acids of rotihibin A was previously determined by Marfey's method (23). 2,4-Diamino butyric acid (Dab), aThr, Asn-ol, and hydroxy-asparagine (OH-Asn) were proven to be in the L-form, and the citrulline (Cit) residue was proven to be in the D-form. The epimerization (E) domain in module 1 confirms the results of this FIG 5 Relative abundances of proteins as part of nonribosomal peptide machinery in response to cellobiose (CB) (Glc 2 ) or glucose (Glc) supply determined by targeted proteomics (LC-MRM). The plot displays the average area under the curve for each peptide used as a marker for the different proteins, normalized to the spike-in standard BSA. These results show significant normalized quantitative peptide abundances (P , 0.05) compared to the wild-type (WT) strain grown in ISP-4 without supplementary oligosaccharides (*) and with cellobiose (**). A statistical two-sided Student t test (homoscedastic) was performed. The error bars plot the standard deviations (SD) from three biological replicates. Note that we already adopted the rotihibin gene cluster annotation (see the text and Table 1). experiment, as it is known to convert amino acids between the L-and D-isomers. Many NRPSs require MbtH-like proteins (MLPs) for the proper folding and activity of the NRPS. This protein is essential for the biosynthesis of, for example, pyoverdine (37), vancomycin (38), coelichelin, and the calcium-dependent antibiotic (CDA) (39). rthL, encoding an MbtH-like protein, is probably necessary for the production of the rotihibins. Li et al. showed that RthL can functionally replace TxtH in the thaxtomin biosynthetic pathway, demonstrating that MLPs from different pathways are able to complement each other (40).
(ii) Biosynthesis of nonproteinogenic amino acids. Three nonproteinogenic amino acids are incorporated into the rotihibin nonribosomal peptide backbone: Cit, Dab, and OH-Asn. rthM encodes an L-asparagine oxygenase (AsnO), which we propose to be involved in converting Asn to OH-Asn. AsnO was previously found in the biosynthetic gene cluster of CDA (41) and A54145 (42). Finally, the unusual amino acid 2,4diamino butyric acid is produced from aspartate b-semialdehyde by the enzyme diaminobutyrate-2-oxoglutarate transaminase (43,44). rthN encodes a homologue of this enzyme and is proposed to be responsible for Dab biosynthesis, which is finally incorporated into rotihibin. There is no specific gene that could be responsible for citrulline synthesis, but this is an intermediate of arginine biosynthesis that could be sufficiently available.
(iii) Formation and attachment of the fatty acid tail. Numerous lipopeptides, synthesized by NRPS complexes, have already been described in Streptomyces species, for example, A54145 (42), CDA (45), daptomycin (46), and enduracidin (32). These clusters  contain genes whose products act together to acylate the first amino acid. Bioinformatic analysis of the rth gene cluster revealed the presence of a gene encoding an acyl carrier protein (rthK), which is immediately flanking rthA. This protein is proposed to transfer medium-chain fatty acids to the N-terminal domain of the Cit-incorporating module (Fig. 8). Fatty acyl-AMP ligases are enzymes establishing the cross talk between fatty acid synthases and NRPSs or polyketide synthetases (PKSs). They initiate the biosynthesis of lipopeptides by the activation of a fatty acyl residue and occur with a high incidence in putative lipopeptide NRPS/PKS clusters  (47). RthH is supposed to recruit a fatty acid and transfer it to the acyl carrier protein RthK. The fatty acid side chain lengths and/or degrees of saturation are the only differences between the rotihibin C and D analogues. The desaturation could be explained by the presence of two acyl-CoA dehydrogenases (RthI and RthJ), similar to the situation in the ramoplanin biosynthetic gene cluster (48). One is expected to introduce the first double bond, while the second additional dehydrogenation is possibly not essential. The minor amounts of other lipopeptides identified via MS, which differ by only 22 Da, could be explained this way (Fig. S1). This phenomenon was previously described in acyl-desferrioxamines of Streptomyces coelicolor (49).
(iv) Self-resistance genes. During antibiotic production, transporter and transporter-associated proteins are important for the import of effector molecules, self-resistance, and guiding/exporting the antibiotic to the extracellular environment. rthF and rthG are predicted to act as the rotihibin-exporting machinery as they encode an ABC-type permease and an ABC transporter ATP-binding protein, respectively, a combination often found in antibiotic biosynthetic gene clusters. RthF shows 47% similarity with the daunorubicin/doxorubicin resistance ABC transporter permease protein DrrB, while RthG shows 61% similarity with the daunorubicin/doxorubicin resistance ATPbinding protein DrrA in Streptomyces peucetius. This DrrAB efflux system in S. peucetius has been shown to be a multidrug transporter with broad specificity (50). RthFG possibly has a similar role in the self-resistance mechanism of S. scabies.
(v) Other genes within the rth cluster. The BGC for rotihibin in S. scabies houses a number of genes not directly associated with rotihibin assembly. The protein product of rthE is predicted to be an F 420 -dependent oxidoreductase, and rthC encodes a shortchain dehydrogenase/reductase. These two enzymes are, for example, directly involved in the production of coronafacoyl phytotoxins in S. scabies (51). The impact of the inactivation of these genes on rotihibin biosynthesis has to be studied to reveal their exact role.
(vi) Inactivation of rthB abolishes rotihibin production. To ascertain whether the rth cluster is responsible for the production of rotihibins, the scab_3221 gene was disrupted and replaced in S. scabies 87-22 by an apramycin resistance cassette (11,13,29). HPLC analysis and targeted metabolomics on n-butanol extracts of the DrthB strain confirmed the absence of rotihibins (Fig. 9). These data support the evidence that the rth BGC is indeed responsible for rotihibin production. Simple complementation of the DrthB strain with an integrative plasmid harboring an intact copy of rthB with its own promoter did not restore the production of rotihibins, which could be attributed to polar effects after the removal of the rthB gene and/or the insertion of the new copy of rthB in a region of the chromosome not optimal for its expression. Biological activity of rotihibins C and D. Since rotihibin A has a plant growth-inhibitory effect, we tested whether the new rotihibins C and D display similar properties. After 4 days of exposure to concentrations ranging from 0.84 to 84.4 mM rotihibin C and from 0.3 to 157.8 mM rotihibin D, the growth of duckweed (Lemna minor L.), especially the fronds, was suppressed compared to that in the control group (a neighboring HPLC fraction that did not display UV absorption) (Fig. 10).
Additionally, we assessed the impact of rotihibins C and D on the photochemistry of photosystem II using the F V /F M ratio as a proxy. In dark-adapted F V /F M measurements, the minimal fluorescence (F 0 ) is measured. Next, an intense light flash is used to close (reduce) all reaction centers and measure the maximum fluorescence, called F M . The difference of F M 2 F 0 is the F V value. The F V /F M ratio represents the maximum potential quantum efficiency of photosystem II. In general, the greater the plant stress, the fewer open reaction centers available, which results in a lowered F V /F M ratio. The RGB and chlorophyll fluorescence images are depicted in Fig. S2 to S5 in the supplemental material. Using the F V /F M ratio as a proxy, the maximal photochemistry efficiency of photosystem II showed a decreasing trend with increasing rotihibin concentrations in the nutrient medium, revealing that rotihibins affect the photosynthesis of L. minor (Fig. 11 and 12).
The increased surface area of L. minor treated with 0.17 mM rotihibin C compared to the HPLC solvent indicates a hormetic effect. Hormesis is a dose-response phenomenon where a low dose of a toxic component promotes plant growth, while at higher doses, inhibition of growth is observed.
Ordinal analysis 96 h after application also showed an effect on the growth and maximal photochemistry efficiency of photosystem II (F V /F M ) (52) of Arabidopsis thaliana plants treated with 0.1 mM rotihibins C and D (Fig. 13). Here, rotihibins were not added to the nutrient medium but were sprayed onto the plants; this explains the high concentration needed compared to the L. minor bioassay. Most herbicides are applied as water-based sprays, the easiest way to protect plant products for farmers. The positive control was a commercial glyphosate formulation (RoundUp Turbo) at a molar concentration of approximately 3 M. At lower concentrations of rotihibin D (0.05 mM and 0.005 mM), we observed growth promotion on Arabidopsis seedlings, which again reflects a hormetic effect (Fig. 14). Recently, Halder et al. chemically synthesized rotihibin A and a number of variants (26). They provide evidence that rotihibin A targets the TOR (target of rapamycin) kinase (TORK) pathway, which explains the observed effects on plants. This highly conserved pathway is involved in the regulation of shoot and root development (53). Phylogenetic and evolutionary analyses of rotihibin biosynthesis. The genomic region surrounding rthA was compared to all publicly available genomes using PATRIC, resulting in 24 Actinobacteria strains harboring a complete putative rotihibin BGC (Table 3). Rotihibin BGCs that split across more than one contig were not considered.
Strains of S. scabies, S. stelliscabiei, S. acidiscabies, S. europaeiscabiei, and S. ipomoeae, all known plant pathogens, are found to contain this rotihibin biosynthetic gene cluster and are thus predicted to produce rotihibin analogues. Other rotihibin BGC-containing species, in contrast, are known as plant-protecting bacteria: S. galbus has been shown to induce disease resistance by the accumulation of a phytoalexin in A. thaliana, thereby protecting it from the anthracnose disease caused by Colletotrichum higginsianum (54). S. caeruleatus is also a biocontrol agent by inhibiting the soybean pathogen Xanthomonas campestris pv. glycines (55). S. geranii sp. nov. was isolated from the root of Geranium carolinianum and shares the highest 16S rRNA sequence similarity to the plant pathogen S. turgidiscabies ATCC 700248 (56). S. hygroscopicus subsp. jinggangensis strains are known to produce the antibiotic validamycin (57), while S. corchorusii produces butalactin (58). The latter shows biocontrol and plant growth-promoting activities and potential as a biofertilizer agent for rice plants (59). The actinomycete Lechevalieria aerocolonigenes has been isolated from soil in Japan and is known to produce rebeccamycin with antitumor properties (60). Curiously, the cluster is found only in plant-associated species but is not restricted to plant pathogens. Rotihibin C and D production was experimentally verified in Streptomyces stelliscabiei NCPPB 4040 (data not shown).
Rotihibin BGCs exist in three different organizations (Table 3), and the gene clusters identified here were classified as E-form, O-form, and X-form clusters, based on the presence of two genes: rthE, encoding, a luciferease-like monogygenase (LLM) class F 420 -dependent oxidoreductase with unknown function, and rthO, which encodes a cytochrome P450 hydroxylase (Fig. 15). The cytochrome P450 OxyD from the vancomycin biosynthetic operon is involved in the biosynthesis of the modified amino acid b-hydroxytyrosine, while the cytochrome P450 TxtC was identified to be required for postcyclization hydroxylation of the cyclic dipeptide thaxtomin A (61, 62). The combination of this additional enzyme with additional modules in the rthA gene, which is discussed later, in Actinobacteria harboring the O-form BGC indicates the putative production of other rotihibin variants. The phylogenetic tree of the complete genomes correlates with the distribution of the different BGC forms (Fig. 16). The nucleotide sequences of the rth genes of the different Actinobacteria were compared with those of the rth genes of Streptomyces scabies 87-22 (Table 4). The presence of two rotihibin BGCs in Streptomyces rameus BK387 is quite remarkable and could be the result of recent intraspecies horizontal gene transfer (HGT) or double interspecies HGT. The loss of rthN in one of these BGCs could be compensated for by the other rthN gene.
The comparison of the rthA genes in the different actinomycetes displayed some NRPSs with more modules and other A-domain specificities, suggesting the production of rotihibin analogues (Table 5). These compounds are interesting and could exhibit different or better activities toward plants.
Cultures were centrifuged at 10,000 Â g for 10 min, and 2 ml of each supernatant was transferred to a new tube. The samples were mixed with n-butanol (1:1), vortexed gently for 10 min, and centrifuged for 15 min at 16,000 Â g. The aqueous phase of the extract was dried and stored at 20°C until further use.
The dried extract was dissolved in 10 mM ammonium formate (NH 4 FA) (pH 3) and analyzed by HPLC, monitoring the absorbance at 220 nm. The sample was eluted with solvent A (10 mM NH 4 FA, pH 3) and solvent B (acetonitrile [ACN]) using a 2-to-90% 7.5-min gradient at 1 ml/min on a Zorbax Eclipse Plus C 18 column (4.6 by 100 mm, 3.5mm). Fractions were collected every 15 s, corresponding to 250ml. The total area under the curve (AUC) for each metabolite was calculated. Student's t test (two tailed, homoscedastic) was executed to evaluate the differential metabolite abundances between the different conditions in a statistical way.
The fractions of interest were loaded into a Triverse Nanomate source (Advion) and, using 1.6 kV and 0.3 lb/in 2 of pressure, electrosprayed into a Waters Synapt G1 mass spectrometer. Scans typically ranged from m/z 100 to 1,200, with a scan time of 1 s, for a total of 2 min. For MS/MS experiments, the collision energy was optimized to achieve good fragmentation spectra.
Generation of the DrthB (scab_3221; scab_RS01465) deletion mutant in Streptomyces scabies 87-22. Deletion of the rthB gene was performed in S. scabies 87-22 using the ReDirect PCR targeting method (63) by replacing the genomic copy of the gene with an apramycin resistance cassette: using a HindIII-EcoRI restriction fragment obtained from pIJ773 (Table 6) as the PCR template, a disruption cassette, containing the resistance gene aac(3)IV and oriT for conjugative transfer, was amplified by PCR with primers BDF33 and BDF34 and gel purified as performed previously for cebR, cebE, msiK, and bglC (11,13,29). In parallel, cosmid 2012 (cos2012), based on Supercos-1 and containing a genomic insert of S. scabies 87-22 ranging from positions 341913 to 386384, including rthB, was introduced by electroporation into Escherichia coli BW25113 carrying plasmid pIJ790 for the arabinose-inducible expression of the l Red recombination machinery. After culture to an optical density (OD) of 0.4 with arabinose at 10 mM to induce the l Red machinery, 50 ml of washed E. coli BW25113/pIJ790 cos2012 cells was electroporated with 100 ng (in 1 ml) of the rthB-targeting disruption cassette, and apramycin-resistant clones were selected on LB agar plates plus apramycin (50 mg/ml) and kanamycin (50 mg/ml). The gene replacement by the disruption cassette on the cosmid was confirmed by PCR using primers BDF35 and BDF36 (Table 6) and Sanger sequencing. The resulting cos2012D3221 construct was purified using a GeneJET plasmid miniprep kit (Thermo Scientific) (according to the manufacturer's guidelines) and transferred by intergeneric conjugation with E. coli ET12567/pUZ8002 (ETpUZ) as the donor strain and S. scabies 87-22 as the recipient strain. After growing ETpUZ to an OD of 0.4 (50-ml culture in LB plus chloramphenicol [30 mg/ml], apramycin [50 mg/ml], kanamycin [50 mg/ml], and ampicillin [100 mg/ml]), cells were washed to remove antibiotics and mixed with S. scabies 87-22 spores for mating. The conjugative mixture was then plated on 30-ml petri dishes of soy flour mannitol medium (SFM) (20 g/liter mannitol, 20 g/liter soy flour, and 20 g/liter agar in tap water and autoclaved twice) plus 10 mM MgCl 2 and overlaid with nalidixic acid (50 mg/ml) and apramycin (40 mg/ml). Exconjugants were transferred to ISP-4 agar plates plus nalidixic acid (25 mg/ml) and apramycin (50 mg/ml) to obtain uniform mutant lines that were then used to prepare spore stocks. Each S. scabies DrthB mutant was checked for apramycin resistance and kanamycin sensitivity on ISP-2 agar plates. Genomic DNA was extracted from 48-h liquid cultures in tryptic soy broth (TSB) using a GenElute bacterial genomic DNA kit (Sigma-Aldrich) (according to the manufacturer's guidelines) and used as the PCR template for mutation confirmation.
Complementation of the DrthB deletion mutant. rthB, including its upstream (2701 bp) and downstream (1501 bp) regions, was amplified by PCR with primers BDF35 and BDF36 to obtain a 3,512bp fragment flanked by two XbaI restriction sites. The PCR product was first cloned into a pJET1.2 vector, which was named pBDF054. Using the XbaI enzyme, rthB and surrounding regions were isolated and cloned into a pAU3-45 (64) integrative vector named pBDF043. Conjugation was performed with S. scabies DrthB clones 1 and 2 on SFM (plus 10 mM MgCl 2 ) overlaid with apramycin (50 mg/ml), nalidixic acid (50 mg/ml), and thiostrepton (12.5 mg/ml). The integration of the plasmid was confirmed by PCR with primers BDF69 and BDF70.
Structural confirmation by NMR spectroscopy. Both rotihibin analogues were dissolved in 0.6 ml of DMSO-d 6 , and the nuclear magnetic resonance (NMR) spectra were recorded at 25°C on a Bruker Avance II 700-MHz spectrometer equipped with a 5-mm Prodigy TCI N 2 cooled cryoprobe. Total correlated spectroscopy (TOCSY) spectra were recorded using the standard pulse sequences from the Bruker library.
Protein extraction and sample preparation. S. scabies RL-34 cells were cultivated in three different growth media (ISP-4 medium, ISP-4 medium supplemented with 0.7% cellobiose, and ISP-4 medium supplemented with 0.5% glucose) for 96 h (28°C at 250 rpm), with three biological replicates under each condition. The cultures were centrifuged at 10,000 Â g for 10 min. The resulting pellet was washed two times and resuspended in 10 ml lysis buffer (1Â phosphate-buffered saline [PBS], 0.1% SDS, protease inhibitor cocktail). The homogeneous mycelium suspension was sonicated (20 times with 30 s on and 30 s off at 30%) on ice using a Branson digital sonifier 450 cell disruptor. The homogenate was centrifuged at 16,000 Â g for 30 min at 4°C. Pellets were discarded, and the supernatant was subjected to trichloroacetic acid (TCA) precipitation. The crude intracellular extract was mixed with a 100% (wt/vol) TCA solution (Sigma-Aldrich) (4:1) and incubated overnight. This mixture was centrifuged at 16,000 Â g for 15 min at 4°C, and the protein pellet was washed two times with ice-cold acetone and solubilized in 2 M urea in 50 mM ammonium bicarbonate. The protein concentration was determined using the Pierce Coomassie protein assay kit.
Targeted proteomic analysis. The resulting peptide solutions were injected (0.1 mg) onto an ultraperformance liquid chromatography (UPLC) M-Class system where the peptides were trapped for 5 min, at 15 ml/min, on a 300-mm by 50-mm, 5-mm, 100-Å Acquity UPLC M-Class symmetry C 18 (Table 7). Data were acquired using MassLynx 4.1 software and processed in Skyline 3.7. The transition curves were subjected to Savitsky-Golay smoothing, and their area under the curve (AUC) was determined and normalized to the spiked BSA standard. Student's t test (two tailed, homoscedastic) was executed to evaluate the differential protein abundances between the different conditions in a statistical way.
Bioinformatics analysis of the rotihibin biosynthetic gene cluster. Prediction of the gene cluster of Streptomyces scabies 87-22 was performed using antiSMASH 3.0 (30). The sequence of the