Uncovering the potential of novel micromonosporae isolated from an extreme hyper-arid Atacama Desert soil

The taxonomic status, biotechnological and ecological potential of several Micromonospora strains isolated from an extreme hyper arid Atacama Desert soil were determined. Initially, a polyphasic study was undertaken to clarify the taxonomic status of five micromonosporae, strains LB4, LB19, LB32T, LB39T and LB41, isolated from an extreme hyper-arid soil collected from one of the driest regions of the Atacama Desert. All of the isolates were found to have chemotaxonomic, cultural and morphological properties consistent with their classification in the genus Micromonospora. Isolates LB32T and LB39T were distinguished from their nearest phylogenetic neighbours and proposed as new species, namely as Micromonospora arida sp. nov. and Micromonospora inaquosa sp. nov., respectively. Eluted methanol extracts of all of the isolates showed activity against a panel of bacterial and fungal indicator strains, notably against multi-drug resistant Klebsiella pneumoniae ATCC 700603 while isolates LB4 and LB41 showed pronounced anti-tumour activity against HepG2 cells. Draft genomes generated for the isolates revealed a rich source of novel biosynthetic gene clusters, some of which were unique to individual strains thereby opening up the prospect of selecting especially gifted micromonosporae for natural product discovery. Key stress-related genes detected in the genomes of all of the isolates provided an insight into how micromonosporae adapt to the harsh environmental conditions that prevail in extreme hyper-arid Atacama Desert soils.

New natural products, especially antibiotics, are needed to control the spread of multi-drug resistant (MDR) microbial pathogens, as exemplified by MDR-resistant Gram-negative bacteria that are associated with high mortality rates 1,2 . Amongst prokaryotes, filamentous bacteria in the class Actinobacteria 3 of the phylum Actinobacteria 4 have a unique track record as a source of novel specialised (secondary) metabolites 5,6 . Despite this, the costly, repeated rediscovery of known chemical entities from common filamentous actinobacteria contributed to the sharp decline in the search for new clinically relevant antibiotics towards the end of the last century 7,8 . However, the discovery that the genomes of filamentous actinobacteria contained many biosynthetic gene clusters (BGCs) that encode for biosynthetic pathways of known and predicted specialised metabolites sparked a renewed interest in these organisms as a source of new chemical scaffolds 9,10 . Especially "gifted" (sensu Baltz 11 ) actinobacteria known to have large genomes (>8.0 Mb) rich in BGCs, include Streptomyces strains [12][13][14] and representatives of historically understudied taxa, such as the genera Amycolatopsis 15 , Micromonospora 16 and Saccharothrix 17 . New approaches to the selective isolation, dereplication and screening of novel filamentous actinobacteria from neglected and

Results and Discussion
Cultural, chemotaxonomic, morphological and genomic properties of the isolates. In general, the cultural, chemotaxonomic and morphological properties of the isolates were consistent with their classification in the genus Micromonospora 16,28 . The isolates were Gram-stain positive, formed extensively branched, non-fragmented substrate hyphae bearing single, non-motile spores, lacked aerial hyphae, produced orange colonies which turned brown-black on spore formation, contained meso-A 2 pm acid and glucose, mannose and xylose in whole-organism hydrolysates, branched chain fatty acids, hydrogenated menaquinones with nine and/or ten isoprene units and polar lipid patterns containing diphosphatidylglycerol, phosphatidylethanolamine (diagnostic lipid) and phosphatidylinositol (phospholipid type 2 sensu Lechevalier et al. 36 ). The RAPD's profiles of the isolates (Fig. S1) underpinned their genetic diversity; though isolates LB4 and LB41 gave similar profiles.
The draft genomes of isolates LB4, LB19, LB32 T , LB39 T and LB41 have been deposited in GenBank under accession numbers QGSX00000000, QDGB00000000, QGSY00000000, QGSZ00000000, QGTA00000000, respectively, and are publically available. Key characteristics of the genomes are shown in Table 1; the number of contigs ranges from 339 to 1725, and the number of genes from 4976 in isolate LB4 to 7013 in LB39 T . RNA genes represented 1-2% of the whole genome sequences ranging from 56 genes in isolate LB32 T to 68 in isolate LB19. The in silico DNA G + C contents of the genomes fell within a narrow range, namely 70.6 to 72.9%, as was the case with strains in an earlier study 16 . Isolates LB32 T and LB39 T presented similar values with 71.0 and 70.6%, while their closest type strains show values of 71.5 and 71.2 for M. chokoriensis and M. saelicesensis, respectively; results that shown coherence with values established for group IVa strains in the micromonosporal phylogenomic tree presented by Carro et al. 16 .
The positions of the isolates in the Micromonospora 16S rRNA gene tree are shown in Fig. S2 and their relationships with their closest phylogenetic neighbours in Fig. 1. The close relationships found between isolate LB4 and M. chalcea DSM 43026 T , between isolate LB39 T and the type strains of M. chokoriensis and Micromonospora violae 37 and between isolates LB19 and LB32 T and the type strains of M. saelicesensis and Micromonospora ureilytica 38 are in good agreement with those reported by Carro et al. 26  www.nature.com/scientificreports www.nature.com/scientificreports/ was not included in the earlier analysis, was found to have an identical 16S rRNA gene sequence to isolate LB4; each of these isolates showed a corresponding sequence similarity with the type strain of M. chalcea of 99.6%. The taxonomic integrity of the M. chalcea clade is supported by a 99% bootstrap value and by the results from the maximum-likelihood and neighbour-joining analyses (Figs 1 and S1).
The isolates were recovered in two well supported clades based on the concatenated sequences of four housekeeping genes (atpD, gyrB, recA and rpoB) and corresponding 16S rRNA gene sequences (Fig. 2). Isolates LB4 and LB41 belong to a clade that encompasses the type strains of Micromonospora aurantiaca 39 46 ; all of these validly named species were recovered in group 1a in the Micromonospora phylogenomic tree generated by Carro et al. 16 . Isolates LB4 and LB41 were found to have identical concatenated gene sequences and showed an MLSA genetic distance with M. chalcea DSM 43026 T of 0.002% (Table S1), a value well below the species level threshold of ≤0.007 proposed by Rong and Huang 47,48 and equivalent to the 70% DNA:DNA cut-off point recommended www.nature.com/scientificreports www.nature.com/scientificreports/ for the delineation of prokaryotic species 49 . In contrast, the two isolates shared genetic distances above the recommended threshold with all of the other closely related phylogenetic neighbours.
Isolates LB19, LB32 T and LB39 T formed a well delineated clade in the MLSA tree together with the type strains of M. chokoriensis, Micromonospora coriariae 50 , Micromonospora cremea 51 , Micromonospora lupini, Figure 2. Neighbour-joining phylogenetic tree based on multilocus sequence alignment of 16S rRNA, gyrB, rpoB, atpD, and recA gene sequences showing relationships between the isolates and between them and Micromonospora type strains. The numbers at the nodes are bootstrap support values when ≥50%. Asterisks indicate branches of the tree that were also recovered in the maximum-likelihood tree. Catellatospora koreensis DSM 44566 T was used as the outgroup. Bar, 0.02 substitutions per nucleotide position. (2019) 9:4678 | https://doi.org/10.1038/s41598-019-38789-z www.nature.com/scientificreports www.nature.com/scientificreports/ M. saelicesensis 35 and M. zamorensis 51 ; all of these validly named species were recovered in group IVa in the phylogenomic tree of Carro et al. 16 . The type strains of Micromonospora noduli, Micromonospora ureilytica and Micromonospora vinacea 38 can be added to this group as they have been shown to be closely related both to one another and to the M. saelicesensis and M. zamorensis strains in phylogenetic tress based on gyrB, MLSA, and 16S rRNA gene sequences 52,53 .
Isolate LB32 T formed a well-supported lineage in the MLSA tree together with M. saelicesensis Lupac09 T ; isolate LB19 was found at the periphery of this taxon, albeit as a distinct branch (Fig. 2). It is apparent from the MLSA distance score of 0.008 that isolate LB32 T and M. saelicesensis are distinct, but sister, species (Table S1). In contrast, it is clear that isolate LB19 belongs to the species M. ureilytica as the two strains share a distance score of 0.005, well below the cut-off point for assigning strains to the same species 47,48 . On the same basis, it is evident from Table S1 that isolate LB39 T shows distance scores with its nearest relatives above the 0.007 threshold and thereby merits consideration as a new Micromonospora species.
The isolates can be distinguished from one another by a broad range of phenotypic properties providing further evidence that they are not clones (Table 2). Excellent congruence was found between the standard phenotypic tests carried out in duplicate though this was not the case with some of the Biolog tests, many of which were weakly positive. The results of all of the phenotypic tests carried out on the isolates can be compared with those of the reference strains as the latter were recorded using the same media and methods. In general, all of the strains grew well from 20-37 °C, at pH 7 and 8, in the presence of 1% w/v sodium chloride, were catalase positive, active in the API-ZYM tests and oxidised a broad range of carbon compounds.
The close relationship recorded earlier between isolates LB4 and LB41 was underpinned by the results from the chemotaxonomic and phenotypic analyses ( Table 2). The strains were found to have identical profiles for the biochemical, enzymatic and tolerance tests and showed a similar ability to oxidise organic acids and sugars. Whole organism hydrolysates of the isolates contained meso-A 2 pm, glucose, galactose, mannose and xylose; they were also shown to have identical polar lipid patterns. In addition, the major fatty acid of strains LB4 and LB41 was iso-C 16:0 (31.0 and 38.5%, respectively) and the predominant isoprenologue MK-9 (H 4 ) (24.5 and 23.1%). When compared with the profiles of the reference strains isolates LB4 and LB41 were most closely related to the type strain of M. chalcea showing overall phenotypic similarities with the latter of 76 and 77% indicating that all three strains belong to the same taxospecies 54,55 . These strains can be distinguished from all of the other organisms given their ability to metabolise D-aspartic acid and inability to oxidise D-saccharic acid. Similarly, the close relationship found earlier between isolate LB19 and the type strain of M. ureilytica is underpinned by the phenotypic data; these strains have many more unit characters in common than isolate LB19 has with the other reference type strains. Isolate LB19 and M. ureilytica GUI23 T can be distinguished from all of the other organisms given their ability to oxidise D-arabitol. They also share similar whole organism hydrolysate and polar lipid patterns (Table 2).
Isolate LB32 T can be separated readily from the type strain of M. saelicesensis, its closest phylogenetic neighbour, by a broad range of phenotypic properties, as exemplified by its ability to produce β-glucoronidase, oxidise L-serine, D-galacturonic acid, D-glucuronic acid, α-keto-glutaric acid, N-acetyl-neuraminic acid, glycerol and D-mannitol. In turn, M. saelicesensis strain Lupac 09 T , unlike isolate LB32 T , grew at pH 9.0 and 45 °C, showed much greater activity in the API-ZYM tests, was oxidase positive, oxidised glycyl-L-proline and D-sorbitol and grew in the presence of lithium chloride, sodium bromate and sodium formate. These differential characters are underscored by several chemotaxonomic traits, notably differences in fatty acid and whole cell sugar patterns ( Table 2).
Isolate LB39 T can be distinguished from the type strain of M. chokoriensis, its closest phylogenetic neighbour, using a combination of chemotaxonomic and other phenotypic features ( Table 2). The former, unlike the latter, produces α-mannosidase, oxidises L-arginine, D-serine #2, butyric acid and bromo-succinic acid and grows in the presence of minocycline, sodium chloride (4%, w/v) and potassium tellurite. In contrast, only the reference strain grows at pH 9.0 and 45 °C, degrades pectin and oxidises L-histidine, α-hydroxy-butyric acid, D-malic acid, D-mannitol and D-sorbitol. The two organisms can also be distinguished using key chemical markers, as illustrated by differences in menaquinone, polar lipid and whole cell sugar composition.
ANI (average nucleotide identity) and dDDH (digital DNA-DNA hybridization) values were calculated between the isolates and between them and their closest phylogenetic neighbours, as shown in Table 3. It is apparent on both counts that isolates LB4 and LB41 are bona fide members of the species M. chalcea as they share ANI and dDDH values with the latter well above the 99.5-99.6% 47,48 and 70% thresholds 49 used to assign strains to the same genomic species. It is also apparent that isolates LB19 and LB39 T are not closely related to one another or to any of the other strains included in these analyses. The situation with respect to isolate LB32 T appears to be less clear cut as it shares a dDDH value with the type strain of M. saelicesensis of 68.2% though the corresponding ANI value, 96.2%, is above the ANI threshold. Similar anomalies have been observed between other closely related Micromonospora species, as with the type strains of Micromonospora carbonacea and Micromonospora haikouensis which shared dDDH and OrthoANI values of 59.9 and 95.2%, respectively; while the corresponding values for the type strains of Micromonospora inyonensis and Micromonospora sagamiensis were 68.9% and 96.5% 16 . Indeed, some species which have been conclusively shown to belong to different Micromonospora species sport higher dDDH and ANI values, as exemplified by the type strains of Micromonospora noduli and M. saelicesensis which share dDDH and OrthoANI values of 71.2 and 96.8%, respectively 56 . Corresponding data between isolate LB19 and the type strain of M. ureilytica cannot be determined until the whole genome sequence of the latter becomes available though the 99.6% MLSA value found between these strains indicates that they belong to the same genomic species, namely M. ureilytica 38 .
In summary, it can be concluded that isolates LB4 and LB41 exhibit a broad range of taxonomic properties consistent with their assignment to the validly named species, M. chalcea 32,33 , strains of which have been isolated from air, soil and aquatic habitats 28 . It is also evident that isolate LB19 belongs to the recently recognised species, www.nature.com/scientificreports www.nature.com/scientificreports/

(c) Oxidation of organic acids:
Bromo-Succinic acid  38 , the sole representative of which came from a root nodule of Pisum sativum. These results provide further evidence that representatives of Micromonospora species are widely distributed in the environment 16 though there is evidence that micromonosporae are a feature of extreme habitats [57][58][59] . It is also apparent that isolate LB32 T , which forms a sister clade to the type strain of M. saelicesensis can be distinguished from the latter by a rich assortment of chemotaxonomic, genotypic and phenotypic data. Similarly, strain LB39 T merits recognition as a novel Micromonospora species as a wealth of taxonomic data can be weighted to separate it from the type strain
www.nature.com/scientificreports www.nature.com/scientificreports/ of M. chokoriensis. In light of these results it is proposed that isolates LB32 T and LB39 T be recognized as new Micromonospora species for which we propose the names Micromonospora arida sp. nov. and Micromonospora inaquosa sp. nov., respectively.
None of the isolates inhibited the growth of the B. subtilis, E. coli and P. fluorescens strains in previous plug assays 26 , possibly due to the use of an inadequate cultivation media. In contrast, extracts from all of the isolates were shown to be active against the bacterial and fungal indicator strains, as shown in Table 4 where extracts showing the greatest activity are given in bold. In general, the most pronounced activity was seen in fractions eluting at higher concentrations of methanol, as exemplified by the inhibition of the E. coli and K. pneumoniae strains. Extracts showed relatively little activity against the A. baumannii, A. fumigatus and P. fluorescens strains and only moderate inhibition of the methicillin-resistant and methicillin-sensitive strains of S. aureus. Similarly, little activity was found against the C. albicans strain with the exception of extracts from isolate LB41. Interestingly, only extracts from isolates LB4 and LB41 showed pronounced inhibition of human hepatocellular carcinoma (HepG2) cells. These results are not only promising, but also provide further evidence that novel and rare micromonosporae from previously unexplored habitats are a promising source of antimicrobial agents 60,61 . Genetic potential of the isolates to produce specialised metabolites. The draft genomes of all of the isolates were examined using the antiSMASH server to detect putative BGCs. The number of such bioclusters ranged from 28 in the genome of isolate LB4 to 64 in the genomes of isolates LB19 and LB41 though this lower number may be a function of a low quality genome, as shown by the relatively high number of contigs (Table 1). Even so, the number of BGCs found in the genomes of the isolates is well within the range found in those of the Micromonospora type strains examined by Carro et al. 16 . In contrast, the average number of BGCs detected in the genomes of the isolates, namely 54, is more than double the average number reported in the earlier study 16 . However, as in that study, the predominant BGCs types coded for lantipeptides, non-ribosomal peptide synthases, polyketide synthases, siderophores and terpenes (Table S2; Fig. 3).
The genomes of the isolates contained 25 BGCs encoding for compounds that showed some degree of similarity to specialised metabolites not previously found in Micromonospora strains. The genomes of all of the isolates encode for BGCs that showed a genetic correspondence to coumermycin, an amino-coumarin antibiotic, produced by Streptomyces rishiriensis strain DSM 40489 62 , which is known to inhibit DNA gyrase and bacterial cell division 63 . In the same vein the genomes of all of the isolates code for a BGC that shows a similarity, ca. ∼40%, to lymphostin biosynthetic cluster, an immunosuppressant originally found in Streptomyces sp. KY11783 64 . Finally, all of the isolates have the capacity to produce compounds related to diazepinomicin, a small alkaloid molecule that binds to and inhibits Ras kinase with the potential to treat multiple solid tumours 65 ; this compound was first detected in the marine actinobacterium, Micromonospora sp. DPJ12 66 . The genome of isolate LB32 T contained a BGC that showed a relatively low similarity to that of algamycin I, an antibacterial 16-membered macrolide active against Micrococcus luteus and Salmonella typhimurium, a compound initially found in Streptomyces sp. KMA-011 67 .
It is also interesting that 18 out of these 25 BGCs were discontinuously distributed across the genomes of the isolates: 1, 5, 3, 2 and 7 in strains LB4, LB19, LB32 T , LB39 T and LB41, respectively (Table S2; Fig. 3); the hybrid system NRPS-PKS was only found in isolates LB39 T and LB41. Secondary metabolite related genes detected   www.nature.com/scientificreports www.nature.com/scientificreports/ using the SEED server were also discontinuously distributed, as exemplified by genes associated with lanthionine synthases which varied from 4 in isolate LB4 to 16 in isolate LB39 T ( Table 5) while all of the isolates, apart from strain LB32 T , contained genes related to the synthesis of thiazole-oxazole-modified microcins, ribosomally produced peptides with post-translationally installed heterocycles derived from cysteine, serine and threonine residues 68 . In turn, only the genome of isolate LB19 harboured a gene associated with the synthesis of clavulanic acid, which encodes for a clavaldehyde dehydrogenase (contig 9) according to a RAST analysis. In this context it is also interesting that compounds extracted from the isolates varied in their ability to inhibit a variety of indicator micro-organisms and the HepG2 cells ( Table 4).
The genomes of the M. chalcea strains were found to harbour 38 different BGCs that presented some similarity with known compounds, out of which 10 have not been detected previously in Micromonospora strains; 7 of these BGCs were only present in the genome of isolate LB41 and the other three in the other LB strains (LB4, LB32 T , LB39 T and LB41). In addition, 26 out of the 38 BGCs were only found in the genomes of one out of the three M. chalcea strains, 9 in two of them and the remaining three in all of them. These results provide further evidence that M. chalcea strains are a good source of novel antibiotics, notably aminoglycosides, lactones and macrolides 61 . However, none of the M. chalcea strains had the capacity to synthesize tetrocarcin A, a spirotetronate antibiotic produced by M. chalcea NRRL 11289 69 or chalcidin or neomycin produced by M. chalcea sp. 70 and M. chalcea B9-683 71 , respectively though the taxonomic provenance of these strains is questionable.
The results of this study taken together with those reported by Carro et al. 16 show that the genomes of Micromonospora strains are a unique source of BGCs that have the potential to synthesise an array of completely novel and uncharacterised specialised metabolites. It is particularly interesting that the genomes of the novel micromonosporae from the extreme hyper-arid Lomas Bayas soil have the capacity to synthesise a broad range of new bioactive compounds. It is also encouraging that M. chalcea strains LB4 and LB41 showed moderate to pronounced antitumour activity and that M. ureilytica strain LB19 and the putative type strains of M. arida (LB32 T ) and M. inaquosa (LB39 T ) showed promise in restricting the growth of the MRD K. pneumoniae strain. Gifted actinobacterial isolates such as these have a role to play in the search and discovery of new chemical scaffolds using state-of-the-art genome tools 9 , including ones designed to induce the expression of silent BGCs 31,72 . Indeed, novel micromonosporae should feature much more prominently in the search and discovery of new classes of specialised metabolites that are needed to control MRD pathogens which currently threaten to take humankind back to the pre-antibiotic www.nature.com/scientificreports www.nature.com/scientificreports/ days of medicine 73,74 . The search for additional novel and rare gifted micromonosporae from Atacama Desert habitats should include functional metagenomics and the use of isolation procedures known to target members of this taxon 26,60,61 and improved characterisation procedures, notably ones for acquiring reliable phenotypic data 56 .

Stress-related genes encoded in the genome of the strains.
The genomes of all of the isolates contained between 97 and 131 putative genes known to be associated with stress responses, notably ones coding for carbon starvation, heat shock responses, osmoregulation and oxidative stress (Table S3). The genomes of isolates LB19, LB32 T and LB39 T contained osmY the expression of which is known to be induced under hyperosmotic stress 75 . The expression of this gene is associated with the induction of the glycine betaine binding protein (proU), which was found in all of the isolated strains (with an average of 4 genes). The genomes of isolates LB19, LB32 T and LB41 harboured genes involved in mycothiol biosynthesis (mshA, mshB, mshC, mshD), an analogue of glutathione that acts as an electron acceptor/donor and serves as a cofactor in detoxification reactions for alkylating agents, free radicals and xenobiotics 76 . Catalase and peroxidase genes were found in all the genomes confirming the results of the laboratory tests (Table S3). However, the type strain of M. chalcea, which gave a negative result has the capacity to produce catalase 16 . The genomes of all the isolates encode for several RNA polymerase Sigma factors and serine phosphatases that acts as regulators, it is known that Sigma B controls a general stress regulon which is induced when cells encounter growth-limiting conditions 77 .
The world's highest levels of surface ultraviolet (UV) irradiance have been reported from the Atacama Desert 78 hence it is particularly interesting that the genomes of all of the isolates included genes associated with protection against UV-radiation; we have previously shown that these strains grew on M65 agar following exposure to UV light at 100 mJoules/second for 30 minutes 26 . The genomes of all of the isolates contained genes belonging to the uvrABC DNA repair system, associated with excision proteins which have been reported in several bacteria 79 . Specific desiccation stress genes were not detected in any of the genomes though several genes associated with the biosynthesis and uptake of trehalose were present, this sugar has been linked with tolerance to heat and desiccation in bacteria 80 . The assortment of stress related genes outlined above provide an insight into how micromonosporae are able to adapt to severe environmental conditions that prevail in arid Atacama Desert soils. However, a similar complement of stress related genes have been found in the genomes of representative Micromonospora taxa isolated from diverse habitats 16 thereby supporting the view that micromonoporae per se have the capacity to colonize multiple microhabitats 28 , including ones associated with extreme biomes 58,81 . In this context it is also interesting that the genomes of all of the isolates contained genes associated with the production of a range of growth promoters of potential value in phytostimulation 16 . The genomes of all of the isolates also contained methylglyoxal detoxification genes (gloA and gloB) which are associated with increases in plant tolerance to abiotic and biotic stress 82 .
Description of Micromonospora arida sp. nov. Micromonospora arida (a'ri.da L. fem. adj. arida, dry, referring to the isolation of the strain from an extreme hyper-arid soil).
The type strain, LB32 T (=CECT 9662 T = LMG 30765 T ) was isolated from an extreme hyper-arid surface soil (2 cm) collected from the Lomas Bayas region of the Atacama Desert soil in Chile. The genome accession number is QGSY00000000.
Description of Micromonospora inaquosa sp. nov. Micromonospora inaquosa (in.a.quo'sa. L. fem. adj. inaquosa without water, referring to the isolation of the strain from an extreme hyper-arid soil).
The type strain, LB39 T (=CECT 9663 T = LMG 30766 T ) was isolated from an extreme hyper-arid surface soil (2 cm) collected from the Lomas Bayas region of the Atacama Desert soil in Chile. The genome accession number is QGSZ00000000.

Methods
Selective isolation. All of the strains (isolates LB4, LB19, LB32 T , LB39 T and LB41) were recovered from the surface (2 cm) of an extreme hyper-arid soil collected from the Lomas Bayas region of the Atacama Desert (23° 24′ 27″ S, 69°31′03″ W 24.02.2014) by Professor Luis Cáceres (University of Antofagasta) as previously described 26 . Briefly, when transferred to the UK the sample was stored at 4 °C. The strains were isolated using the selective isolation procedure devised by Makkar and Cross 83 ; to this end, aliquots (100 µl) of the 10 −1/2 and 10 −1 dilutions of the soil in ¼ strength Ringer's solution were spread over the surface of starch-casein agar plates 84 supplemented with sterile cycloheximide, nystatin and novobiocin (each at 25 µg/ml). Three replicate plates were prepared per dilution and incubated at 28 °C for 3 weeks when five characteristic orange-coloured Micromonospora colonies were detected. The isolates were maintained on M65 (DSMZ medium) agar plates and as mixtures of hyphal fragments and spores in 20% v/v glycerol at −80 °C.
Extraction of DNA and determination of RAPD profiles. Genetic profiles were generated by PCR using the primer M13 (5′-GAGGGTGGCGGTTCT-3′) 85 . DNA was extracted from all of the isolates using a REDExtract-N.Amp kit (Sigma) and amplified following the manufacturer's recommendations to give a final volume of 20 μl per reaction; the thermal cycling parameters were: 7 min at 95 °C, 35 cycles of 1 min at 94 °C, 1 min www.nature.com/scientificreports www.nature.com/scientificreports/ at 45 °C and 2 min at 72 °C, followed by a 6 min final extension at 72 °C. A 1.5% agarose gel containing ethidium bromide was loaded with 5 µl of each of the PCR products and electrophoresis run at 85 V for 90 minutes in freshly prepared 1x TBE-EDTA buffer at pH 8.0 using a Bio-Rad PowerPac 300 power supply; a DNA molecular weight marker (1 kbp) was used as a molecular size standard. Photographs of the electrophoresis results recovered as TIFF files were aligned using BioNumerics package 6.0 into similarity groups. Phylogenetic analysis. Genomic DNA extraction, PCR-mediated amplification and 16S rRNA gene sequencing were performed as described by Carro et al. 26 . Universal primers 27 F and 1522R 86 were used for PCR amplification in a final volume of 50 μl using Bioline 2x MiFi TM mix following instructions of the manufacturer. The PCR products were purified and sequenced using the EZseq Barcode Service (Macrogen). The manually aligned sequences were compared with those of their closest neighbours retrieved from the EzBioCloud server 87 . Maximum-likelihood 88 and neighbour-joining 89 algorithms were used to generate the phylogenetic trees. In addition, a multilocus sequence analysis (MLSA) based on 16S rRNA, atpD, gyrB, recA and rpoB gene sequences retrieved from whole-genome sequences of the isolates was carried out using established procedures 52 and a micromonosporal MLSA tree generated from the 9165 nucleotides using the neighbour-joining and maximum-likelihood algorithms.
Phenotypic profiles. The isolates were examined for micromorphological, Gram-stain and motility using a phase-contrast microscope (Leica; CTR MIC) and 7-day-old cultures grown on GYM Streptomyces agar (DSMZ medium 65 90 ). They were also examined for their ability to grow in the presence of various concentrations of sodium chloride (1, 2, 5, 7 and 9% w/v) and over a range of pH (4.0-9.0 at one unit intervals) and temperature regimes (4, 10, 20, 28 37 and 40 °C) using GYM as the basal medium. All of these tests were recorded on duplicated cultures after 14 days of incubation. Enzymatic activities of the isolates were determined using API ZYM kits (bioMerieux) according to the manufacturer's instructions. The ability of the isolates to oxidise diverse carbon and nitrogen sources and to show resistance to inhibitory compounds was determined using GEN III microplates in an Omnilog device (BIOLOG Inc., Haywood, USA) and the exported data of the duplicated samples analysed using the opm package for R version 1.06 91,92 . Other phenotypic analyses were determined following Carro et al. 93 .
Biomass for the chemotaxonomic analyses carried out on each of the isolates was prepared in shake flasks (180 rpm) in ISP2 broth 94 following incubation at 28 °C for 14 days, washed twice in sterile saline solution, and freeze-dried. Standard procedures were used to detect the isomers of diaminopimelic acid (A 2 pm) 95 , menaquinones 96 , polar lipids 97 and whole cell sugar composition 98 , using appropriate controls. Cellular fatty acids were extracted, methylated, examined by gas chromatography (Agilent Technologies mod. 7890 A) and analysed using the protocol of the Sherlock Microbial Identification (MIDI) system, version 6.3 99 . The resultant peaks were named using the RTSBA6 database.
Whole-genome sequencing and genomic analyses. A single colony of each of the isolates was used to inoculate 50 ml aliquots of M65 broth and the resultant preparations incubated at 28 °C for 7 days when cells were centrifuged prior to sending to Microbes NG (Birmingham, UK). Genomic DNA extracted from each of the preparations was sequenced on an Illumina HiSeq 2500 instrument with 2 × 250 bp paired-end reads. All of the strains were analysed using a standard pipeline and identified with their closest reference genome using Kraken 100 and by mapping the reads using BWA-MEM 101 . The reads were assembled into contigs using SPAdes 3.90 102 and contigs <500 bp discarded. Variant calling performed on the draft assemblies using VarScan were reordered and reoriented relative to a reference genome based on a MUMmer whole-genome alignment. An automated annotation was performed using Prokka 103 while antiSMASH 4.0 was used to determine and compare BGCs encoding for natural products 104 . The presence of other genes was detected using the SEED viewer 105 following RAST annotation of the genomes 106,107 .
Digital DNA:DNA hybridisation (dDDH) values between the genomes of the isolates and between them and available genomes of their phylogenetic neighbours were calculated using the genome-to-genome distance calculator, GGDC 2.0, using formula 2 of the GGDC web server available at http://ggds.dsmz.de/ggdc.php. In addition, ANI values were determined between the strains using OAT version 0.93.1 108 .
Bioassays with the extract of the isolates. Each of the isolates was shaken in 50 ml of ISP 2 broth 94 at 180 revolutions per minute (rpm) with resin beads (Amberlite XAD-16N, Sigma) at 28 °C for 14 days. Each preparation was centrifuged (4100 rpm) for 15 min and the biomass and XAD-16N resin beads soaked overnight in methanol and filtered through glass wool prior to evaporation of the methanol fraction at 40 °C by nitrogen sparging to generate the extracts. They were fractionated with Solid Phase Extraction (SPE) cartridges, using either 2 g or 5 g of a C18 resin (55 µm, 70 Å, from Strata) depending on the weight of the extract; four column volumes of the following solvents were sequentially used for the fractionation of the samples: 100% water, 25%, 50, 75 and 100% methanol and 100% methanol + 0.01% TFA. The eluted fractions were screened using liquid chromatography-mass spectrometry (LCMS).
Inhibition tests were carried out on each of the fractions using a concentration of 300 µg/ml in 96 well-plates containing a total incubation volume of 200 µl. The screening assays were carried out using a range of indicator microorganisms, namely MRD strains of Acinetobacter baumannii (CL5973), Escherichia coli (ATCC 25922), Klebsiella pneumoniae (ATCC 700603), Pseudomonas aeruginosa (MB5919), Staphylococcus aureus MB5393 (methicillin-resistant) and ATCC 29213 (methicillin-sensitive), as well as MDR Aspergillus fumigatus ATCC 46645. Negative controls were included in the plates for each microorganism tested without extracts. Anti-tumour activity against human hepatocellular carcinoma HepG2 cells was determined in the same system using concentrations of 75 mg/ml for each fraction. Negative and positive controls were included containing dimethyl