Isolation and characterization of Streptomyces bacteriophages and Streptomyces strains encoding biosynthetic arsenals

The threat to public health posed by drug-resistant bacteria is rapidly increasing, as some of healthcare’s most potent antibiotics are becoming obsolete. Approximately two-thirds of the world’s antibiotics are derived from natural products produced by Streptomyces encoded biosynthetic gene clusters. Thus, to identify novel gene clusters, we sequenced the genomes of four bioactive Streptomyces strains isolated from the soil in San Diego County and used Bacterial Cytological Profiling adapted for agar plate culturing in order to examine the mechanisms of bacterial inhibition exhibited by these strains. In the four strains, we identified 104 biosynthetic gene clusters. Some of these clusters were predicted to produce previously studied antibiotics; however, the known mechanisms of these molecules could not fully account for the antibacterial activity exhibited by the strains, suggesting that novel clusters might encode antibiotics. When assessed for their ability to inhibit the growth of clinically isolated pathogens, three Streptomyces strains demonstrated activity against methicillin-resistant Staphylococcus aureus. Additionally, due to the utility of bacteriophages for genetically manipulating bacterial strains via transduction, we also isolated four new phages (BartholomewSD, IceWarrior, Shawty, and TrvxScott) against S. platensis. A genomic analysis of our phages revealed nearly 200 uncharacterized proteins, including a new site-specific serine integrase that could prove to be a useful genetic tool. Sequence analysis of the Streptomyces strains identified CRISPR-Cas systems and specific spacer sequences that allowed us to predict phage host ranges. Ultimately, this study identified Streptomyces strains with the potential to produce novel chemical matter as well as integrase-encoding phages that could potentially be used to manipulate these strains.

Introduction Antibiotic discovery is an international priority requiring immediate action [1]. The increasing prevalence of multi-drug resistant (MDR) bacterial pathogens has resulted in an increased use of last-resort antibiotics [1][2][3]. Microbes that produce natural products are the most prolific source of clinically approved antibiotics [4]. Soil dwelling Actinobacteria, notably Streptomyces, account for two-thirds of the antibiotics currently on the market [5][6][7]. Despite intensive studies, however, the full potential of microbes to produce natural products has not been realized [8]. Genome mining studies have shown that microbes encode many biosynthetic gene clusters (BGCs) that have not yet been characterized [8]. It is widely assumed that many of these clusters produce novel natural products and that further characterization of Streptomyces bacteria increases the probability of identifying molecules with unique chemical structures and new mechanisms of action [9].
In addition to identifying Streptomyces strains containing potentially novel BGCs, it is necessary to improve on the conventional approaches used in natural product antibiotic discovery. One of the major stumbling blocks in natural product discovery is dereplication since the isolation of bioactive molecules often yields antibiotics that have previously been discovered [10]. We recently developed Bacterial Cytological Profiling (BCP) as a new whole-cell screening technique that can be used to rapidly identify the mechanism of action (MOA) of antibiotics [11][12][13][14][15][16]. BCP can accurately identify the pathway inhibited by antibacterial compounds present in unfractionated crude organic extracts and can be used to guide the purification of molecules with specific bioactivities [11,15]. BCP can also be used to screen bacterial strains directly on petri plates to identify and prioritize those strains that produce molecules with desired MOAs [15]. In effect, BCP helps with the problem of dereplication by allowing for the determination of the MOA of antibiotics synthesized by a particular Streptomyces strain before labor-intensive antibiotic purification efforts are performed.
Since many BGCs are not expressed under laboratory conditions, genetic methods are often used to augment their expression and facilitate the identification and purification of their products [17]. Sometimes, increased expression can be achieved using techniques such as CRISPR/Cas or plasmid cloning and overexpression [17]. However, there is still an occasional need to move large chromosomal regions from one strain to another via transduction to engineer strains optimally suited for antibiotic production. Transduction requires a phage capable of infecting the strain(s) of interest. Moreover, because phages generally display narrow host ranges [18] and relatively few Streptomyces phages have been isolated [19] compared to the large number of studied Streptomyces bacteria [20], phages aptly suited for genetic manipulations are not available for the majority of antibiotic producing Streptomyces strains isolated. In addition, phage derived enzymes such as recombinases and integrases can also be used to engineer new strains [21][22][23][24][25]. Thus, studying the phages that infect antibiotic-producing Streptomyces strains could not only yield new transducing phages but potentially also new genetic tools for strain engineering.
Here we describe the isolation and characterization of Streptomyces strains and phages. We used a combination of bioinformatics and BCP to characterize the antibiotic biosynthetic potential of four Streptomyces strains that displayed an ability to inhibit Gram-negative and Gram-positive bacterial growth. Additionally, we isolated four new phages and assessed their abilities to infect our Streptomyces strains, which contained many CRISPRs. The proteins encoded by the phages were subjected to bioinformatic analyses to identify putative integrases that might be used for genetic manipulations. This work highlights a novel set of gene clusters and Streptomyces sp. phages that serve as a starting point for the isolation of potentially novel natural products.

Isolation and antibacterial activities of Streptomyces sp.
To identify Streptomyces strains containing potentially novel BGCs, we collected 28 unique soil samples from sites across San Diego County. From these samples, we isolated a total of eight bacterial strains based on colony morphology. The genus level classification of the eight isolates was confirmed as Streptomyces using the phylogeny of their 16S rRNA sequences as well as data from type strains (Fig 1 and Table 1). Each of the strains isolated in this study were part of a well-supported clade including at least one type strain, These strains (designated JS, DF, QF2, EDE, SK, AH, ELW, and SFW) and two known species (Streptomyces coelicolor A3 (2) and Streptomyces platensis AB045882) were screened using the cross-streak method for their ability to inhibit the growth of wild type E. coli MC4100, an efflux defective mutant E. coli JP313 ΔtolC, and B. subtilis PY79. Since the production of bioactive secondary metabolites is highly dependent on growth conditions, this screen was conducted on actinomycete isolation agar (AIA) as well as Luria Broth (LB) agar. Each of the 10 strains proved capable of inhibiting the growth of E. coli and/or B. subtilis on at least one of the tested media (Fig 2), suggesting that these strains likely produce antibiotics. As expected, however, the production of antibiotics often depended upon whether the strain was grown on AIA or LB agar. For example, strain ELW was incapable of inhibiting the growth of Gram-negative and Gram-positive bacteria when grown on AIA. However, when grown on LB agar, strain ELW inhibited the growth of both Gram-negative and Gram-positive bacteria. Conversely, strains JS and QF2 exhibited growth inhibition regardless of the media on which they were grown.

Mechanistic analysis of natural products produced by four Streptomyces isolates
Strains QF2, JS, SFW and DF all inhibited the growth of E. coli ΔtolC when grown on either AIA or LB agar, but in each case, the mechanism underlying inhibition was unknown. Thus, we utilized BCP to examine the mechanism of growth inhibition exhibited by the antibacterial natural products synthesized by these four Streptomyces isolates. Each of the four strains was grown on three different media: LB, AIA, or International Streptomyces Project-2 media (ISP2) for 5 days to allow for the synthesis and excretion of natural products into the surrounding agar. We then added exponentially growing E. coli cells adjacent to the Streptomyces lawn. After two hours of incubation at 30˚C, the E. coli cells were stained with fluorescent dyes and imaged with high resolution fluorescence microscopy. E. coli cells incubated adjacent to each of the four Streptomyces isolates displayed characteristic cytological profiles that, in some cases, allowed for the classification of these strains according to the MOA of the natural products they produced (Fig 3).
When grown on either LB or ISP2, strain QF2 synthesized an antibiotic that caused the DNA of affected E. coli cells to assume a toroidal conformation (Fig 3). This phenotype is characteristic of bacteria treated with protein synthesis inhibitors such as chloramphenicol [11,26], and thus, we concluded that strain QF2 can synthesize a translation-inhibiting natural product. QF2 also produced a membrane-active secondary metabolite, evidenced by visible membrane abnormalities as well as Sytox Green permeability under all tested nutrient conditions (Fig 3). Strain JS appeared to induce similar phenotypes in E. coli, though under different growth conditions; protein synthesis inhibition phenotypes were observed on AIA and ISP2 but not on LB. Similar to strain QF2, Sytox Green permeability was observed in some cells regardless of the medium on which strain JS was grown. Strain SFW induced distinct phenotypes in E. coli cells under each of the three nutrient conditions (Fig 3). On LB, a significant number of E. coli cells grown in the presence of strain SFW appeared to contain three chromosomes (white arrows), a phenotype that was not present in the untreated control cells. When strain SFW was grown on AIA, the E. coli cells became bent and lost their characteristic rod shape. Finally, strain SFW grown on ISP2 induced

PLOS ONE
substantial swelling in E. coli cells that ultimately led to lysis. Notably, E. coli cells grown in the presence of strain DF exhibited nearly these same phenotypes under identical growth conditions, suggesting that these two strains produce compounds targeting similar pathways.

Genomic analysis of four Streptomyces isolates
To better understand how strains QF2, JS, SFW and DF inhibited bacterial growth, we sequenced their genomes and aligned them to the most similar genomes in the NCBI database ( Fig 4A). Sequence reads for strain DF were assembled into a single contig that was most similar to the genome of S. fulvissimus DSM 40593. Sequencing of strains QF2, JS, and SFW yielded multiple contigs that were aligned to the genomes of S. globisporus C-1027, S. parvulus 12434, and S. pratensis ATCC 33331, respectively.
In order to identify predicted gene clusters associated with secondary metabolism, the assembled genome sequences for strains QF2, JS, SFW and DF were annotated using RASTtk [27] and submitted to AntiSMASH 5.0 [28] (Fig 4B). Each strain encoded between 18 and 37 BGCs, some of which were present in multiple strains ( Fig 4C). Additionally, some of the BCP phenotypes of E. coli JP313 ΔtolC exposed to natural products produced by four Streptomyces soil isolates grown on different solid media. Also displayed, are E. coli JP313 ΔtolC untreated controls grown on the tested media (LB agar, AIA, and ISP2 agar). White arrows indicate cells with three chromosomes. BCP images were collected after staining the cells with FM4-64 (red), DAPI (blue), and SYTOX-green (green). The scale bar represents one micron. https://doi.org/10.1371/journal.pone.0262354.g003

PLOS ONE
encoded clusters closely resembled known BGCs in the MIBiG repository [29]. For example, of the 23 putative BGCs identified in the genome of strain QF2 ( Table 2), one of them (cluster 21) was similar to the viomycin BGC ( Fig 5A). Viomycin inhibits protein synthesis by stabilizing tRNAs in the A site of the bacterial ribosome, inhibiting translocation [30]. According to AntiSMASH, 66% of the genes within the viomycin BGC were similar to genes within cluster 21. However, a global pairwise alignment of cluster 21 and the viomycin BGC revealed that the nucleotide sequence of cluster 21 is actually 98.5% identical over 32.5kb of the 36kb viomycin BGC (Fig 5A). This suggests that a viomycin related molecule is synthesized by strain QF2 and may account for strain's ability to inhibit protein synthesis in E. coli (Fig 3). While some of the other clusters in the QF2 genome (Table 2: clusters 2 , 4, 7, 8, 9, 12, 13, 22) have significant similarity to known BGCs, no other clusters appear to produce known antibiotics. Strain JS contained 18 putative BGCs, six of which shared significant similarity (>60% of genes in common) with a known cluster (Table 3). Of these six, however, only cluster 7 was predicted to produce an antibiotic. All of the genes constituting a known terpene cluster that produces albaflavenone were present in cluster 7 ( Fig 5B). Albalfavenone is capable of inhibiting the growth of B. subtilis by an unknown MOA [35] and has previously been isolated from S. coelicolor A3(2) [36], a close relative of strain JS. Since the MOA of albaflavenone is unknown, it's not clear whether the products of cluster 7 or of a different cluster are responsible for the inhibition of protein synthesis and/or the membrane permeability observed in E. coli (Fig 3).
Of the 26 putative BGCs that were identified in the genome of strain SFW, only one cluster shared a high percentage of genes in common with a known antibiotic-producing cluster ( Table 4). Cluster 1 shared similarity with 62% of the genes within a known BGC that produces carbapenems (Fig 5C), a class of beta-lactam antibiotics that inhibit cell wall biogenesis [41,42]. Additionally, cluster 4 contained a low percentage of genes in common with a BGC involved in the synthesis of clavulanic acid, which inhibits beta-lactamase and consequently The most similar BGCs in the MIBiG database are listed, as well as the percentage of genes in each MIBiG known cluster that have similarity to genes in the corresponding QF2 cluster. In cases where the most similar known BGC produces an antibiotic, the MOA was listed (Showdomycin [31], Ficellomycin [32], Skyllamycin [33], Viomycin [30], Alkylresorcinol [34] strengthens the bactericidal activity of beta-lactams. Cluster 1 (and perhaps cluster 4) could, therefore, contribute to the synthesis of bioactive molecules that account for the inhibition of E. coli cell wall biogenesis on ISP2 media (Fig 3). Strain DF encoded 37 BGCs (Table 5). Despite this rich supply of BGCs, however, we were only able to identify one cluster that likely participates in the synthesis of an antibiotic with a confirmed MOA. According to AntiSMASH v5.0, cluster 29 shared 92% gene identity with a known BGC that produces nonactin, a bioactive ionophore that disrupts membrane potential [52] (Fig 5D). The known clusters could not fully account for the antibacterial activity exhibited by strain DF (Fig 3), suggesting that antibiotics might be produced by novel clusters.

Antimicrobial activity of four Streptomyces isolates against clinically relevant pathogens
To assess the relevance of antibiotics produced by strains JS, DF, SFW, and QF2, we screened their ability to inhibit the growth of three clinically isolated pathogens using the cross-streak method ( Table 6). Both strain QF2 and strain JS inhibited the growth of methicillin-resistant S. aureus (MRSA) and efflux-deficient P. aeruginosa PA01. These strains did not, however, inhibit the growth of the wild-type clinical isolates P. aeruginosa PA01 and P. aeruginosa P4, which were resistant to the antibiotics produced by all four Streptomyces isolates. Strain DF, though incapable of inhibiting the growth of E. coli tolC + (Fig 2), did inhibit the growth of MRSA and efflux-deficient P. aeruginosa PA01. Strain SFW was the least capable of inhibiting the growth of clinical pathogens, producing antibiotics only effective against E. coli tolC + (Fig 2).

Phage isolation and genome sequencing
Phages capable of infecting these newly isolated Streptomyces strains can be used for genetic manipulation. With the goal of identifying genetic tools that could be used to augment expression of the BGCs in our bacterial isolates, we isolated bacteriophages using S. platensis as a host. This species, in particular, was chosen as a host because it is relatively well-characterized, and S. platensis phages capable of infecting our Streptomyces isolates could possibly be used to move (via transduction) BGCs from our isolates into a more genetically manipulatable and familiar background [63,64]. Thus, to increase the probability that our S. platensis phages could be used for this purpose, we performed the isolation using the same soil samples from which our Streptomyces strains were obtained. Four S. platensis actinobacteriophages (Bartho-lomewSD, IceWarrior, Shawty, and TrvxScott) were successfully isolated. These phages were imaged using negative-stain transmission electron microscopy ( Fig 6A) and subsequently characterized as members of the family Siphoviridae due to their long filamentous tails and icosahedral capsids [65,66]. Genome sequencing revealed that BartholomewSD (52.1 kb) and TrvxScott (52.6 kb) are 89% identical ( Fig 6B) and belong to the BD2 subcluster of Streptomyces phages, which currently contains 20 other members [19]. IceWarrior (55.5 kb) clustered in subcluster BI1 (24 members), and Shawty (40.7 kb) clustered in BB1, a subcluster of 7 phages that includes notable members TG1 and phiC31 (Table 7) [19]. The BLASTp-predicted functions of the gene products encoded by these phages are shown in Table 8. The most similar BGCs in the MIBiG database are listed, as well as the percentage of genes in each MIBiG known cluster that have similarity to genes in the corresponding SFW cluster. In cases where the most similar known BGC produces an antibiotic, the MOA was listed (Carbapenem [42], Clavulanic acid [43], Kanamycin [44], Lipopeptide 8D1-1 & 8D1-2 [38], Ficellomycin [32], Lactonamycin [45,46], Istamycin [47][48][49], Cinerubin [50], Tetronasin [51]). https://doi.org/10.1371/journal.pone.0262354.t004

Characterization of CRISPR elements in the genomes of our Streptomyces strains
Prior to testing the ability of our phages to infect the Streptomyces isolates, we decided to examine the strains for complete and functional CRISPR/Cas systems. Our reasoning for this was two-fold. First, the presence of acquired spacers and their specific sequences would allow us to make predictions about whether or not our phages can infect our antibiotic-producing strains. Second, it was conceivable that in doing so we might discover a novel CRISPR/Casbased system. Our bioinformatic analysis identified the presence of Cas enzymes and CRISPR  arrays within the genomes all four of our Streptomyces isolates, but the abundance of CRISPRs in each strain varied greatly (Table 9). QF2 contained the largest number of predicted CRISPRs-38 in total, scattered around the chromosome, each containing between one and 25 spacers (Fig 4B, purple; Table 10). Some predicted spacers within these arrays matched with 94-100% identity to sequences within TrvxScott (7 spacers), BartholomewSD (4 spacers), Shawty (2 spacers), and IceWarrior (5 spacers) (Tables 10 and 11). Spacers targeted a variety of genes including those encoding minor tail proteins, tape measure proteins, deoxycytidylate deaminase, helix-turn-helix DNA binding proteins, endolysin, and capsid maturation protease (Fig 7). The large number of putative spacers in the QF2 genome targeting TrvxScott, Bartho-lomewSD, Shawty, and IceWarrior suggests that strain QF2 has likely previously encountered and acquired resistance to each of these phages. Moreover, strain QF2 was isolated from the same soil sample as BartholomewSD, providing support for these findings. Strain QF2 also encoded seven proteins of a Type IE CRISPR-Cas system [67][68][69][70][71]. The QF2 proteins were distantly related to the enzymes of the canonical Cas3 system in E. coli (Fig 8), but the operon in strain QF2 lacked two genes (Cas1 and Cas2) involved in spacer acquisition. This phenomenon, the absence of Cas1 and Cas2, has previously been reported as a common feature of Streptomycetaceae Type IE systems [72]. The presence in the QF2 genome of a Cas3 system and spacers mapping to essential proteins in each of the genomes of our phages suggests that the strain is likely resistant to all four of our phages, and thus, transduction is unlikely with strain QF2. Specific spacers mapping to some of our phages were also discovered within the genomes of strains DF and SFW (but not JS). Strain DF contained two spacers that mapped to sequences within the genome of BartholomewSD, and one of these spacers also shared sequence similarity with a region in TrvxScott (Tables 11 and 12). Strain SFW contained two spacers-one that shared sequence similarity with Shawty and another that mapped to a sequence in IceWarrior (Tables 11 and 13). Both strains DF and SFW encoded proteins containing regions with similarity to the RuvC and HNH endonuclease domains of known Cas enzymes. However, given the limited similarity of these putative proteins to known Cas proteins, further study is necessary to determine if they constitute novel Cas systems. If these systems are functional, we predict that strain DF is resistant to infection by TrvxScott and BartholomewSD, and strain SFW, resistant to Shawty and IceWarrior.
A curious feature emerged from our analysis of the CRISPRs within the Streptomyces strains. Among the 205 predicted spacers encoded by all four bacterial strains, 18 contained sequence similarity (95-100% identity) with at least one of the four phages ( Table 9). The lengths of the matching sequences (100% identity) within bacterial spacers ranged from 14 to 18 nucleotides and accounted for approximately half the length of a typical spacer. Additionally, a single spacer occasionally appeared capable of targeting two distantly related phages. These spacers contained sequences mapping to two distinct genes encoded by different viral genomes. For example, spacer 106 in CRIPSR 23 of strain QF2 is 32 nucleotides in length, and it contains 14 bases that share 100% identity with a region in the TrvxScott tape measure gene. These 14 bases overlap (by 8 nucleotides) with another sequence that is 14 base pairs in length   and shares 100% identity with a region within the Shawty genome (Table 11). If these spacers functionally serve to resist infection, our analysis suggests that a single spacer may evolve to efficiently target more than one phage, thus providing broad immunity.

Susceptibility of Streptomyces strains to infection by S. platensis phages
With the hope of identifying phages that might serve as tools for transduction, we assessed the susceptibility of our Streptomyces isolates to infection by each of the four S. platensis phages (Fig 9). As predicted, strain QF2, with its Type IE Cas system and many CRISPRs containing spacers against our phages, could not be infected by any of our four phages. Strain DF experienced inefficient infection by TrvxScott (~2.0 x 10 4 -fold reduced plating efficiency compared to S. platensis) and was completely resistant to infection by BartholomewSD. In addition to these results, which were generally predicted by our CRISPR/Cas findings, we also demonstrated strain DF's resistance to infection by Shawty and susceptibility to IceWarrior (~20-fold reduced efficiency). Strain SFW was at least partially resistant to infection by all four phages: Shawty (no infection), BartholomewSD (no infection), IceWarrior (~10 7 -fold reduced efficiency), and TrvxScott (~1.3 x 10 5 -fold reduced efficiency). Finally, strain JS, despite having no spacers specifically targeting our phages, was similarly immune to infection by Shawty and BartholomewSD and partially resistant to infection by IceWarrior and TrvxScott (~10 6 -fold and~10 5 -fold reduced efficiency, respectively). These data are consistent with our predictions regarding the resistance of our Streptomyces isolates to infection by the phages against which they carry spacers, though it is the case that the presence of a spacer did not always confer complete immunity to the phage it targeted. In some cases, strains containing spacers could be infected relatively inefficiently by the targeted phage. For example, strain DF encoded a single

Identification of phage integrases
In analyzing the proteins encoded within the genomes of our phages, we identified site-specific serine recombinases encoded by BartholomewSD, TrvxScott, and Shawty. The integrases of BartholomewSD and TrvxScott were nearly identical and shared similarity to integrases belonging to a number of previously studied phages. The Shawty integrase shared protein sequence similarity with the integrases of Streptomyces phages TG1 and phiC31 (71.3% and 51.9% respectively). The TG1 and phiC31 integrases are distinct recombinases that share 49.7% protein sequence identity and have been used extensively as tools for integrating genes of interest into specific loci within the genomes of a wide variety of organisms, ranging from soil microbes such as Streptomyces to multicellular animals such as Drosophila [21][22][23][24][25]. Thus, as a newly discovered member of this serine integrase family, the Shawty integrase could also potentially be used to move Streptomyces-encoded BGCs between strains to facilitate augmented expression of bioactive natural products.

Conclusion
This work demonstrates a method for effectively studying newly isolated, antibiotic-producing bacteria and phages that may infect them. We have highlighted how BCP can be used to assess the novelty of BGCs encoded within Streptomyces strains, providing another solution to the problem of dereplication. Moreover, our work illustrates how the isolation and genomic analysis of phages that infect antibiotic-producing Streptomyces might yield new genetic tools such

PLOS ONE
as transducing phages or integrases, which can be used to augment the expression of novel antibiotics. Specifically, strains DF and JS are promising candidates for future novel antibiotic discovery. Each is active against MRSA and has the potential to produce new chemistries given their encoded biosynthetic arsenal and BCP phenotypes. These two strains are also infected by TrvxScott, which might be used to genetically manipulate their BGCs to induce production of novel antibiotics. Ultimately, our identification of potentially novel BGCs and phage integrases serves as a foundation for further studies that could lead to the discovery of new antibiotics.

Isolation of Streptomyces
Actinomycete isolation agar (AIA) plates (for one liter: sodium caseinate 2 g, L-Asparagine 0.1 g, sodium propionate 4 g, dipotassium phosphate 0.5 g, magnesium sulphate 0.1 g, glycerol 5 mL, rifampicin 50 μgmL -1 , nystatin 100 μgmL -1 , cycloheximide 100 μgmL -1 , agarose 15 g, pH 8.1) were used to select for Actinobacteria from soil samples. 1 g of soil was added to the agar surface and streaked across the AIA plate and incubated for two days at 30˚C. The plates were investigated for individual colonies with morphologies indicative of Streptomyces (vegetative hyphae, aerial mycelium), and those colonies were picked and purified at least four times on AIA plates containing 100 μgmL -1 cycloheximide (CHX), which was included to prevent unwanted fungal growth.

Phage isolation and purification
Actinobacteriophages were isolated from soil samples with host, Streptomyces platensis JCM 4664 substrain MJ1A1. An enrichment culture was prepared from 1 g soil and 2.5 ml of S. platensis added to 15 ml of Luria-Bertani (LB) medium (for one liter: tryptone 10 g, yeast extract 5 g, NaCl 10 g, agar 15 g, pH 7.0), followed by a 2-day incubation at 30˚C with shaking. Phage were isolated from a 1.2 ml volume of enrichment culture that was centrifuged at maximum speed for 3 min, 1 ml of the resulting supernatant was filtered (0.22 μM filter), and 5 μl of the filtrate was spotted and then streaked onto an LB plate containing 100 μg/ml of cycloheximide. S. platensis (0.1 ml) was mixed with 4.5 ml of LB top agar 0.7%, poured over the streak plate and incubated for two days at 30˚C. Resultant plaques were re-streaked onto new LB plates containing 100 μg/ml of cycloheximide about 3-4 times for phage purification.

Bacterial genomic DNA extraction and quantification for 16S rRNA PCR amplification and sequencing
An adaptation of the DNeasy1 Blood & Tissue Kit (Qiagen) protocol was used for bacterial genomic DNA extraction. Strains were cultured overnight at 30˚C in 5 ml of LB broth while rolling. Cells were pelleted (16,000 x g, 3 min) from 1 ml of culture, re-suspended in 180 μl of lysis buffer (prepared in house), and incubated at 37˚C for 45 min after which 25 μl of proteinase K (20 mg/mL) and 200 μl Buffer AL (Qiagen) was added. The samples were vortexed at maximum speed for 20 sec, incubated at 56˚C for 30 min, and 200 μl of ethanol (96-100%) was added. The samples were vortexed at maximum speed for 30 sec, added to a DNeasy Mini spin column, centrifuged (16,000 x g, 1 min), and the supernatant was discarded. Buffer AW2 (Qiagen) was added (500 μl), followed by centrifugation (20,000 x g, 3 min). The DNeasy Mini spin column was placed into a sterile 2 ml microcentrifuge tube, and the gDNA was eluted in 100 μl of AE Buffer by centrifugation (20,000 x g, 1 min) following a 1 min incubation at room temperature. The gDNA concentration was quantified (1 μl sample volume) with a Thermo Scientific TM NanoDrop TM One Microvolume UV-Vis Spectrophotometer (840274100) and stored at -20˚C.

Bacterial genomic DNA extraction for PacBio whole-genome sequencing
High molecular weight genomic DNA (20-160 kb) was extracted from four Streptomyces strains (DF, SFW, QF2, and JS) with the QIAGEN-Genomic-tip 500/G kit (10262) according to the manufacturer's protocol for bacteria.

Bacterial whole-genome sequencing, assembly, and annotation
The genome sequences of four Streptomyces strains were generated using the Pacific Biosciences RS II (PacBio RS II) single molecule real-time (SMRT) sequencing platform at the IGM Genomics Center, University of California, San Diego, La Jolla, CA. Linear genome sequences were assembled using the HGAP protocol integrated in the PacBio RS II sequencer (smrt analysis v2.3.0/Patch5) resulting in a variable number (n = 1-95) of contigs per genome, and ranged in size from 5.42 to 7.79 Mb. The mauve contig mover was used to order the contigs of three draft genome sequences (genomes of strains SFW, QF2, and JS) relative to a closely related reference sequence (S. pratensis ATCC 33331, S. globisporus C-1027, and S. parvulus 2297 respectively). DNA sequencing of strain DF resulted in a single contig and did not require reordering to restore gene synteny, however PacBio sequences were combined with Illumina paired end reads. Illumina sequences were generated from a Nextera genomic library and sequenced using the NextSeq 550 platform with the 300 Mbs kit at the Microbial Genome Sequencing Center (MiGS; Pittsburgh, PA). DNA for Illumina sequencing was prepared using the aforementioned protocol for 16S rRNA sequencing. A hybrid assembly of PacBio and Illumina reads was generated in Geneious Prime (2019.2.3) with the following; (1) Illumina paired end reads were processed using the Geneious Prime workflow 'best practice for preprocessing NGS reads in Geneious Prime,' (2) Processed reads were mapped to the PacBio genome using the Geneious assembler with default settings, (3) the resulting consensus sequence was exported (.fasta) for downstream analyses. Gene prediction and annotation were made with the Rapid Annotations using Subsystems Technology (RASTtk) platform [73].

Phage genomic DNA extraction
5 μl of RNase A and 5 μl of DNase I were added to 10ml of lysate, incubated at 30˚C for 30 minutes, and then precipitated overnight at 4˚C by the addition of 4 ml of 20% polyethylene glycol 8,000. Samples were centrifuged at 10,000 g's for 30 minutes, and pellets resuspended in Qiagen PB buffer and DNA isolated using a Qiagen plasmid DNA isolation column as recommended by the manufacturer.

Phage genome sequencing, assembly, and annotation
Genomic DNA of 4 actinobacteriophages (TrvxScott, BartholomewSD, Shawty, and IceWarrior) was sequenced using the Illumina MiSeq platform at the Pittsburgh Bacteriophage Institute sequencing facility. The genomes were assembled with Newbler and checked for quality with Consed. The whole genome sequences were submitted to GenBank (Acc No. MH669016, MK460245, MK433266, and MK433259). DNA Master was used for annotation, and NCBI BLASTp was used to determine the potential function of gene products. Whole genome sequence comparisons were performed in Phamerator [74].

16S rRNA PCR amplification and sequencing
16S ribosomal DNA templates (~1,465 bp) were amplified using Q5 high fidelity PCR (New England Biolabs) with the universal primer set 27F (5'-AGAGTTTGATCCTGGCTCAG-3') and 1492R (5'-GGTTACCTTGTTACGACTT-3') [75]. Each PCR mixture (50 μl) contained 100 ng of template gDNA, 500 pmol of each primer, and 200 μM dNTPs. PCR thermocycling conditions were as follows: 30 seconds of initial denaturation at 98˚C, 30 cycles of denaturation at 98˚C for 10 seconds, annealing for 15 seconds at 60˚C, extension at 72˚C for 1.5 minutes, and a final extension at 72˚C for 5 minutes then held at 4˚C. PCR products were purified with the oligonucleotide cleanup protocol as described in the Monarch PCR & DNA Cleanup Kit 5 μg user manual (NEB #T1030). Clean PCR products were sequenced using Sanger methods by Eton Biosciences (https://www.etonbio.com/) and trimmed for quality before analysis.

Phylogenetic analyses of bacterial isolates
16S rRNA sequences were trimmed on both ends, (5' and 3') in Geneious Prime using the Trim Ends function with an error probability limit set at 0.05, which trims regions with more than a 5% chance of an error per base. Sequences were aligned using MUSCLE v3.8.425 with a maximum of 1,000 iterations, then maximum likelihood was performed using RAxML with 100 rapid bootstrap replicates and the GTR+G model. The tree was visualized using FigTree v1.4.2.

Cross-streak method for assessing antibacterial production potential
From a single colony, using sterile Q-tips, Streptomyces isolates were streaked in a broad vertical line (2 inch) onto LB, and AIA, solid agar plates and incubated for one week at 30˚C. The day before the assay, test strains (E. coli JP313 ΔtolC, B. subtilis PY79, and E. coli MC4100) were grown in 5 ml of LB and incubated at 30˚C overnight while rolling. On the day of the antibacterial screen, the overnight cultures of each test strain were diluted (1:100 in 5 ml LB) and grown to log phase OD 600 0.15-0.2 (~1.5 hr at 30˚C while rolling). A volume of 10 μl of each test strain was spotted in distinct lines almost to the edge of the Streptomyces line at a perpendicular angle. The plates were incubated overnight at 30˚C, then investigated for the presence of zones of inhibition which were measured in millimeters.

Bacterial cytological profiling (BCP) on plates
Fluorescence microscopy and BCP on plates was performed as previously described by Nonejuie et al. [15]. Briefly, Streptomyces strains (DF, SFW, QF2, and JS) were streaked in a vertical line down the center of LB, AIA, and ISP2 plates (for one liter: 4.0 g Difco yeast extract, 10.0 g Difco malt extract, 4.0 g dextrose, 20.0 g agar, pH 7.0), incubated for one week at 30˚C. The test strain, E. coli JP313 ΔtolC, was prepared and spotted as described above in the cross-streak method. Following a 2 hr incubation at 30˚C, a 1.5 x 1.5 cm square (~2.5 cm 2 ) piece of agar containing the E. coli test strain was cut and prepared for high resolution fluorescence microscopy. The cut piece of agar was placed on a microscope slide, the E. coli cells were stained with fluorescent dyes, a coverslip was placed on top of the stained cells then imaged.

Host range experiment
The host ranges of 4 phages were determined against the Streptomyces strains: QF2, DF, JS, and SFW. The experiment was blinded by assigning phages numbers i-iv and hosts letters A-D. A lawn of Streptomyces in LB top agar was poured on LB CHX plates. After the top agar solidified, a grid was drawn on the bottom of the plate, and 5 μl of pre-diluted phage samples (10 0 to 10 −10 in phage buffer) were spotted in squares on the grid. Plaques were counted and used to calculate a titer, which was then compared to the titer obtained against S. platensis to calculate the efficiency of infection.

Transmission electron microscopy
10 μl of lysate was applied to a copper grid, stained with 1% uranyl acetate, washed twice with phage buffer, and allowed to dry. Images were collected using a FEI Tecnai Spirit G2 BioTWIN Transmission Electron Microscope equipped with a bottom mount Eagle 4k camera.