Forty-five samples of rhizospheric soils were collected from rice-cultivated fields in Tolima (Colombia). In order to conduct this research, the ANLA (Autoridad Nacional de Licencias Ambientales) and the Ministry of Environment (Ministerio de Ambiente y Desarrollo Sostenible) were requested for permission to collect samples and study the recovered bacteria.

Briefly, plants roots were removed from the soil and subsequently handshaked for 10 min to remove bulk soil. The remaining adhering soil was considered as rhizospheric soil, and was collected by handshaking roots for 10 min in a sterile plastic bag. Rhizospheric soil samples (4–5 g for each) were then mixed with 1 g of CaCO₃. Samples were further dried at 45°C for 1 h, as described previously (Gurung et al., 2009). Actinobacteria were subsequently isolated by spread plate technique following the serial dilution of soil samples on starch casein agar (Starch 10 g/L, Casein 1 g/L, K₂HPO₄ 0.5 g/L, Agar 13 g/L) and incubated at 30°C for a week. In order to select Streptomyces-like bacteria, microscopic and macroscopic characteristics of the obtained colonies were assessed resulting in a total of 60 isolates, which were further purified and sub-cultured. Single colonies were characterized by their colony morphology in International Streptomyces Project media (ISP2, ISP3, ISP4), Nutrient agar and Mueller Hinton Agar (Supplementary Figure S1). Catalase, oxidase and their carbon source utilization were tested, as suggested for Streptomyces-like bacteria (Shirling and Gottlieb, 1968; Goodfellow, 2012).

Actinobacteria strains were grown for 5 days in M3.7 liquid media (glucose 5 g/L, Yeast Extract 5 g/L, (NH₄)₂SO₄ 5 g/L, Corn gluten 5 g/L, CaCO₃ 2 g/L, NaCl 2 g/L, FeSO₄ 1 mg/L, starch 10 g/L pH 7.2). Five-day cultures were used for antimicrobial tests. Antibacterial activities against 21 strains from 15 species (Supplementary Table S1) were determined by using the Kirby-Bauer agar well-diffusion method, modified from the CLSI 2011 guidelines (Clinical Laboratory Standards Institute, 2011). Briefly, each strain to be tested was grown overnight in LB media, and its OD₆₀₀ was determined. Bacterial absorbance was then adjusted to 0.25, and a cotton swab was fully immersed in each bacterial dilution. The inoculated swab was used to spread the entire surface of a Mueller Hinton Agar plate, containing 25 mL of agar five times. Seven mm (diameter) wells were perforated in the agar, and 50 μL of each actinobacterial culture were poured into the well. Plates were subsequently incubated at 30 or 37°C (according to the best temperature for each bacteria to be tested). Inhibition zones were measured after 24 h of incubation. Antibacterial assays were also performed against a collection of 48 B. glumae isolates recovered from rice plants exhibiting symptoms of BPB (kindly provided by FEDEARROZ-Colombia).

Antifungal activities against 9 fungal plant pathogens (Supplementary Table S1) were measured as described previously (Kanini et al., 2013). For this assay, fungal strains were grown in Potato Dextrose Agar (PDA) plates for 3 days at 25°C, and two 6 mm disks of mycelium from each phytopathogenic fungi were then placed in opposites edges of a new PDA plate. Following fungal inoculation, two 6 mm wells were open in opposite sides of the PDA plate, containing 50 μL of a 5-day culture for each Actinobacteria strain tested. The plates were incubated at 28°C for 5 days and antagonistic activity was estimated by measuring the growth inhibition zone.

Three strains were selected based on their wide range of antimicrobial activity. Genomic DNA was isolated by using the Salting-out procedure described previously (Pospiech and Neumann, 1995), with few modifications. Briefly, single colonies of each isolate were grown in 30 mL Tryptic Soy Broth, with shaking at 30°C for 48 h. Cells were harvested by centrifugation and washed with 10% sucrose, and suspended in 5 ml of STE Buffer (75 mM NaCl, 25 mM EDTA pH 8.0, 20 mM Tris-HCl pH 7.5), containing heat-treated pancreatic RNase A at 10 mg/mL. Hundred microliters of lysozyme (50 mg/mL) were added to the mixture and incubated for 1 h at 37°C, with gentle tapping at intervals. This was followed by the addition of 140 μL of proteinase K (20 mg/ml) and 600 μL of 10% SDS, with further incubation at 55°C for 2 h. Two mL of 5M NaCl were added and mixed by inversion until the temperature of the suspension reached the 37°C. The samples were cooled to room temperature and extracted twice with 5 ml of chloroform. Aqueous fractions were separated by centrifuging at 10000 g for 10 min at 4°C, and genomic DNA was precipitated by adding one volume of isopropanol, mixed by inversion and centrifuged 15 min at 15000 rpm. The obtained DNA pellet was rinsed with 70% ethanol, air-dried and suspended in 100 μL of DNAse-free sterile water.

Taxonomical identification was first approached by PCR amplification, sequencing, and analysis of the entire 16S rRNA locus by using primers 16S_A and 16S_B described by Cui et al. (2001), and the cleaned PCR products were directly sequenced using universal primers 27F, 500F, 818R and 1492R (Supplementary Table S1). Closely related sequences were obtained from RDP (Ribosomal Database Project) (Cole et al., 2014) and EZTaxon (Kim et al., 2012). Obtained sequences were imported into MEGA6 software and aligned with ClustalW for phylogenetic analysis (Thompson et al., 1994).

Multilocus Sequence Typing was used to elucidate a further taxonomical affiliation (Guo et al., 2008; Rong et al., 2009). Briefly, five loci (gyrB, atpD, recA trpB, and rpoB) were PCR amplified and purified by using the EuroGOLD Gel extraction Kit (EuroCLONE, Italy). Each fragment was then sequenced by using primers described previously (Guo et al., 2008), and listed in Supplementary Table S1. The sequences for all loci for each strain were concatenated head to tail in frame and exported in FASTA format. Sequences were aligned using CLCbio Mainworkbench 6.9.1 (CLC-Bio Qiagen) and MEGA 6.0 (Tamura et al., 2011). A phylogenetic tree was constructed from the concatenated sequences of all six loci, and representative Streptomyces strains available in the Streptomyces MLST database (Jolley and Maiden, 2010)¹ Maximum likelihood algorithm trees were generated with MEGA based on the Tamura-Nei model (Tamura and Nei, 1993), and a tree with the highest log likelihood was generated calculating the percentage of trees in which the associated taxa clustered together. Initial trees for the heuristic search were obtained automatically by applying Neighbor-Join and BioNJ algorithms (Gascuel, 1997) to a matrix of pairwise distances estimated using the Maximum Composite Likelihood (MCL) approach, and then selecting the topology with superior log-likelihood value.

Biological nitrogen fixation, phosphate solubilization, indoleacetic acid production (IAA), ACC deaminase and extracellular enzyme production were evaluated. For this purpose, each isolate was grown in 100 mL M3.7 medium for 5 days, and each assay was performed with three independent cultures for each strain.

An egfp tagged mutant from Streptomyces sp. A20 was generated by intergeneric conjugation with Escherichia coli S17-1 λ pir harboring pIJ8641 plasmid, (kindly provided by M. Bibb from the John Innes Centre-Norwich, United Kingdom) (Sun et al., 1999; Kieser et al., 2000; Willey et al., 2007), and transconjugants were selected into Apramycin ISP3 plates (100 μg/ml). Fluorescent transconjugants were further selected for gnotobiotic inoculation experiments as described before, using either semisolid Hoagland or sterile vermiculite supplemented with Apramycin (100 μg/ml). For these experiments, 15 plants were inoculated with A20pIJ8641, non-tagged Streptomyces A20, and a non-inoculated set of plants was included as a control. To assess the colonization process three roots were evaluated for each treatment at days 1, 2, 3, 10, and 15 days post inoculation. In each evaluation day, the roots were excised from the matrix, washed three times with sterile water and cut in 3 mm × 3 mm fragments. Observations were performed in an Olympus BX41 Microscope with a CCD Infinity 3 Camera (dichroic mirror DM500; excitation filter BP460-490 and barrier filter BA520IF).

Streptomyces A20 was grown in 80 L M3.7 media in a 100 L stirrer tank bioreactor with constant shaking (40 rpm) at 25°C. After 4 days, 1 L of culture was centrifuged and spent supernatants were filter-sterilized using a 0.22 μm membrane. Bioassay-guided fractionation of A20 fermentation sterile supernatants was then performed by a sequential partition with ethyl acetate (2 × 500 mL), followed by butanol (4 × 500 mL) extraction (Supplementary Figure S3), to yield organic (EtOAc), butanolic and residual aqueous fraction (ABP fractions). Each fraction was evaporated to dryness and suspended into sterile water to be subjected to antibacterial assays. Active fractions were first purified by ultra-filtration by using 1, 3, and 10 kDa Amicon filters (Millipore).

Several chromatographic techniques were used to purify antimicrobial compounds produced by strain Streptomyces A20. Briefly, residual aqueous fractions from butanol extraction were dried and suspended into buffer A (sodium acetate 0.05 M pH 5.5) to be loaded into a strong anion exchanger column, Q Sepharose Fast Flow, previously equilibrated and washed with buffer A, using a BioRad Biologic Duoflow chromatography system. Compounds bound to the column were eluted with a linear gradient of increasing NaCl concentration (0–1M). Antimicrobial compounds present in the flowthrough, were purified by loading it into an S Sepharose^TM Fast Flow column (cation exchange chromatography), and antimicrobial compounds were eluted again with a gradient of increasing NaCl concentration. Antimicrobial fractions were then suspended into 20 mM ammonium acetate, containing 0.1% heptafluorobutyric acid (HFBA), and injected onto a semipreparative C18 reversed phase column (1x 25 cm). Three fractions were recovered due to their antibacterial activity and were further analyzed by LC-MS.

Samples for LC-MS were dissolved in 90% methanol and 10 μl injected onto a Gemini C18 column (2 × 150 mm, 5 μm; Phenomenex, Torrance, CA, United States) and separated using a gradient from water to 90% acetronitrile mobile phases containing 0.1% formic acid at 0.2 ml min^-1. The column effluent was delivered to an ESI ion trap mass spectrometer (Amazon SL, Bruker Daltonics, Bremen, Germany) and mass spectra collected in positive ion mode in the range 200–2000 m/z.

Most experiments were performed at least three times and means are given. Statistical analyses included unpaired t-tests and ANOVA with Dunnett’s post-test and were performed with PRISM 4.0 software (GraphPad Software, San Diego CA, United States). A P-value of < 0.05 was considered significant.

16S rRNA nucleotide sequences for strains Streptomyces A20, 5.1, 7.1 and Streptomyces racemochromogenes DSM 40194 were deposited in GenBank/EMBL/DDBJ, under the accession numbers KP082880, KP082881, KP082882, and KP082883.

Sixty actinobacterial-like isolates were obtained from Colombian soils upon enrichment with CaCO₃ however, only three isolates denoted as A20, 5.1 and 7.1 showed antimicrobial activity against either two plant pathogenic bacterial strains of B. glumae and P. fuscovaginae, or fungal phytopathogens. Strains A.20 and 5.1 showed both antibacterial and antifungal activity, whereas strain 7.1 showed only antifungal activity (Table 1).

Biochemical tests indicated that these three strains were catalase positive; oxidase negative, and their carbon source analysis indicated that all strains were able to metabolize glucose, but were unable to use rhamnose, arabinose, xylose or mannitol as carbon source (Table 2). Taxonomical analyses derived from 16S rRNA sequencing confirmed that all three strains belong to the genus Streptomyces and similarity calculations of the 16S rRNA locus, based on Maximum Likelihood and Neighbor Joining analyses indicated that the closest relatives for strain A20 were S. racemochromogenes/S. polychromogenes and S. flavotricini with similarity values of 99.93 and 99.8 % respectively, placing this strain within the previously described clade 38 (Labeda et al., 2012). A similar analysis located strain 7.1 within clades 2 and 71 with similarity values of 99.89 and 99.79% to S. chrestomyceticus, S. corchorusii/S. canarius. Strain 5.1 was related to the S. roseoverticillatus clade 65, having 98.9% similarity to Streptomyces kashimirensis and S. salmonis (Figure 1A).

In order to improve the taxonomical identification, MLST was performed for all three strains, and type strain S. racemochromogenes DSM 40194. MLST analyses for 12 strains, including the closest neighbors for each isolate revealed that all six loci have unique sequences compared to all available allele types in the MLST Streptomyces database. MLST profiles for strains 5.1 and 7.1 defined these two isolates as independent entities, and MLST profiles for close neighbors these two should be obtained to improve their taxonomical identification.

All loci obtained for strain A20 were highly similar to those obtained for strain S. racemochromogenes DSM 40194, and concatenated loci for both strains shared 97.2% identity at sequence level, however evolutionary distance calculated with Kimura 2 parameters, was above the species-definitive MLSA distance of 0.007 (0.028) (Figure 1B) (Rong and Huang, 2012; Labeda et al., 2014). Further studies may be required to generate MLST profiles for other type strains from this cluster in order to verify these results and define if strain A20 corresponds to a known species or may be described as a new species.

In order to study the strength and range of antimicrobial activity, all three selected strains were tested by Kirby-Bauer well-diffusion assays, which indicated that Streptomyces strains A20 and 5.1 could inhibit the growth of a wide range of bacterial and fungal phytopathogens to varying degrees, whereas strain 7.1 was able to inhibit mainly fungal phytopathogens (Table 1). Strain A20 showed the wider range of activity, with antimicrobial effects against 20 out of 21 bacterial species tested, and against all fungal pathogens evaluated (Figure 2). All three Streptomyces strains were also tested against a collection of 48 pathogenic B. glumae isolates isolated from Colombian rice-growing areas. In these assays, Streptomyces strains A20 and 5.1 exhibited antimicrobial activities against 91.48 and 78.72% of pathogens tested, respectively. Remarkably most of the pathogenic B. glumae isolates were inhibited by Streptomyces A20.

In vitro tests suggested that all three strains were able to produce siderophores, as well as proteolytic enzymes. All three strains also showed an ability to produce IAA and ACC deaminase, but none of the Streptomyces isolates could fix nitrogen. Streptomyces strain 7.1 was able to solubilize inorganic phosphate and showed a remarkable cellulolytic ability (Table 3 and Figure 3).

Gnotobiotic inoculation experiments suggested that strain A20 was able to promote rice growth in comparison to non-inoculated plants, since it generated increases in shoot length and fresh weight in roots and plants of two different cultivars being however statistically significant only for cultivar F60 (Figure 4A). Streptomyces 7.1 was able to increase root and shoot fresh weight in the cultivar F733 and did not have any effect on rice cultivar F60. Streptomyces 5.1 did not show any increase in any of the evaluated parameters, in any of the cultivars tested. In soil experiments, Streptomyces 5.1 was able to improve the fresh weight of roots and shoots in cultivar 733, and strains A20 and 7.1 displayed the ability to significantly increase all the biometric parameters evaluated in cultivar F733 in comparison to the non-inoculated control (Figure 4B and Supplementary Figure S4).

Plant growth promotion experiments were performed both with sterile and non-sterile conditions in order to compare the performance of the three strains in such environments and foresee their behavior under natural conditions. In non-sterile soil, plants inoculated with Streptomyces A20 and Streptomyces 7.1 showed increases in their biometric parameters, with the same trend they showed in the gnotobiotic experiments. In contrast, strain Streptomyces 5.1 was able to increase plant length and fresh weight in plantlets from cultivar 733 (Figure 4C). Importantly, plantlets from non-sterile soils were smaller due to the presence of fungal strains, observed along the process (data not shown), which were most likely controlled by antifungal metabolites produced by isolates A20 and 5.1.

Colonization experiments indicated that all Streptomyces strains tested were able to colonize the roots, since they were recovered from the rhizoplane in gnotobiotic experiments with roots from cultivars F733 and F60, even after 31 days post-inoculation (dpi) (Table 4 and Supplementary Material S2). Streptomyces A20 and Streptomyces 7.1 were also able to colonize the root endosphere, as they were recovered from a percentage of surface sterilized plants (80 and 20 %, respectively) grown in a gnotobiotic system.

An egfp tagged mutant was generated for strain A20 and colonization experiments were performed. Our results indicated that egfp expressing strain strongly associated with root hairs (Figure 5). Also, our experiments suggested that Streptomyces A20 first established itself in the main root and on the surface of the rice seeds (1 dpi, Figure 5), followed by the formation of micro-colonies between 2 and 3 dpi, and the colonization of secondary roots (Figure 5). After 10 days of growth, Streptomyces A20 had colonized the points of emergence for secondary roots and the surface of the main and secondary roots. After 15 days an overall colonization from the surface of the root is observed both in vermiculite and semisolid Hoagland media.

Strain A20 was chosen for further characterization of antimicrobial metabolites. Bioassay-guided extraction of A20 spent supernatants with solvents of increasing-polarity suggested that the antimicrobial compounds produced by this strain are hydrophilic ionic compounds since they remain in the aqueous fraction after the butanol extraction (Supplementary Figure S3). Ultrafiltration assays followed by antimicrobial activity assays indicated that the antimicrobial compounds present in residual aqueous fractions have sizes below 3 kDa (Table 5). Cation-exchange chromatography generated three more fractions with antimicrobial activity (namely fractions 21, 32, and 33), which were further submitted to ion-pairing chromatography in order to obtain defined peaks with antimicrobial activity. This technique allowed the resolution of 2 identical peaks in fractions 32 and 33, and one single peak in fraction 21, which possessed antibacterial activity. All the fractions were subsequently subjected to mass spectrometry analysis. Results have indicated that strain A20 produces three compounds with same spectra patterns and m/z values (503.3, 631.4, and 759.4), to those reported for Streptothricin F, D and E, as suggested by comparison of their mass spectra (Figure 6) (Ji et al., 2007; Maruyama and Hamano, 2009). Characterization of the streptothricin D was confirmed as follows: the MS/MS spectra for the parent ion at m/z 759.4 [M+H]⁺, proposed as the pseudomolecular ion for Streptothricine D, presented daughter ions at m/z 589 [M-L₃]⁺, 571[M-L₃-H₂O], 554[M-L₃-H₂O-NH₃]⁺, 553 [M-L₃-H₂O-H₂O]⁺, and 510 [M-L₃-H₂O-H₂O-CONH₂]⁺ where the L₃ represents the chain of three lysine residues of Streptothricin D, as suggested previously (Ji et al., 2007, 2008, 2009). These analyses were also applied for Streptothricin E and F, indicating the presence of an ion m/z 171, a signature peak characteristic of streptolidine lactam for all Streptothricins (Figure 6) (Ji et al., 2009).