Phage Milagro: a platform for engineering a broad host range virulent phage for Burkholderia

ABSTRACT

The Burkholderia cepacia complex (Bcc) causes life-threatening respiratory tract infections in persons with cystic fibrosis (CF). In CF patients, end-stage pulmonary disease often requires lung transplantation, and pre-transplant colonization with antibiotic-resistant Burkholderia is predictive of poor post-transplant outcomes. To address this issue, phage therapy has been proposed as a treatment for these infections. However, the majority of characterized Bcc phages are temperate and are therefore difficult to use as therapeutics, and the few obligately lytic phages that have been isolated have limited host ranges. To overcome these limitations, we have produced a virulent, broad-host range derivative of the temperate Burkholderia cenocepacia phage Milagro. Phage Milagro is a 39.1-kb temperate myophage related to phage KL3 and the paradigm coliphage P2. This phage showed a phenotype of spontaneous autoplaquing on lawns of Milagro lysogens, and these autoplaques were found to be produced by virulent mutants of the parental phage Milagro. Mutations associated with virulence were identified as single base changes, insertions or deletions in the phage lysogeny control region that define potential operator sites required for lysogen maintenance. To improve phage host range, the C-terminal domain of the Milagro tail fiber was replaced with the receptor-binding domain of the broad-host range tailocin (high molecular weight bacteriocin) BceTMilo. A spontaneous virulent mutant of this engineered phage, designated Milagro vir gp20:Milo, exhibited an expanded host range over the parental phage and is able to infect multiple Bcc species including B. cenocepacia, Burkholderia multivorans, Burkholderia gladioli, Burkholderia dolosa, and Burkholderia vietnamensis.

IMPORTANCE

Burkholderia infections are a significant concern in people with CF and other immunocompromising disorders, and are difficult to treat with conventional antibiotics due to their inherent drug resistance. Bacteriophages, or bacterial viruses, are now seen as a potential alternative therapy for these infections, but most of the naturally occurring phages are temperate and have narrow host ranges, which limit their utility as therapeutics. Here we describe the temperate Burkholderia phage Milagro and our efforts to engineer this phage into a potential therapeutic by expanding the phage host range and selecting for phage mutants that are strictly virulent. This approach may be used to generate new therapeutic agents for treating intractable infections in CF patients.

INTRODUCTION

The Burkholderia cepacia complex (Bcc) is composed of more than 20 closely related species that are opportunistic pathogens (1, 2) known to cause severe, life-threatening respiratory tract infections in individuals with cystic fibrosis (CF) (3 – 5). Essentially all Bcc clinical isolates demonstrate broad-spectrum antibiotic resistance in vitro, with many being resistant to most currently available antibiotics (6). This renders effective antibiotic treatment nearly impossible, leaving physicians with few or no options to treat infections in the lungs of CF patients, particularly those with late-stage pulmonary disease (3, 7).

Bacteriophage (phage) therapy, the treatment of bacterial infections with bacteriophages, has emerged in recent years as a potentially viable alternative for treatment of antibiotic-resistant bacterial infections (8 – 11). In CF patients, individual phage treatments have been used for recalcitrant infections caused by Mycobacterium abscessus (12, 13), Achromobacter (14 – 17), Burkholderia multivorans (18), Burkholderia dolosa (19), Pseudomonas aeruginosa (20, 21), and Staphylococcus aureus (21). Clinical outcomes following phage treatments for non-Burkholderia infections have been generally positive. Both treatments using phage against Burkholderia were conducted in post-operative lung transplant patients and were marked by periods of improvement followed by terminal decline; however, these cases were also characterized by multiple sequelae related to their recent lung transplants, long-term infections, or drug toxicity (18, 19).

The ideal phage to be used for phage therapy is generally characterized as obligately lytic (virulent), has a broad yet species-specific host range, is stable at application parameters, and is easily propagated to high titers (22). However, a recent review by Lauman and Dennis (23) listed 27 characterized Bcc phages, of which 21 were temperate. Of the six obligately lytic phages described, only two showed activity on more than one clinical Bcc isolate and two showed activity against only plant pathogenic isolates (23, 24). Therefore, the pool of virulent phages capable of broadly infecting clinically relevant Burkholderia isolates remains limited. Temperate phages have been modified by various methods such as homologous recombination, bacteriophage recombineering of electroporated DNA (BRED), and CRISPR-Cas to make them incapable of repression and lysogen formation for phage therapy applications (25). As an example, an M. abscessus temperate phage was engineered using BRED to delete its lysogenic repressor, allowing it to be used to treat a CF patient (13). In Burkholderia phage KS9, the inactivation of the CI-like repressor produced a virulent mutant that showed rescue activity in a Galleria mellonella model of Burkholderia infection (26). While a recent model suggests that some temperate phages may still provide enough lytic activity to effect treatment (27), we propose that it is still preferable to use obligately lytic phages when they are available, and disabling of lysogeny functions increases the available pool of such phages.

A major determinant of phage host range is the recognition of bacterial cell surface features by the phage tail fiber or spike, and modification of these structures has been proposed as a strategy to improve phage host range for therapeutic applications (28). Multiple approaches have been proposed to modify phage tail fibers, one of which is a “domain swapping” strategy in which the C-terminal portion of a phage tail fiber is replaced by the C-terminal portion of a tail fiber taken from a heterologous phage (28). This approach has been shown to function in multiple systems including Listeria phage PSA (29), coliphages P2 (30) and T2 (31), R-type pyocin (32), and T5-like phages infecting Salmonella (33).

There is a clear need for virulent Bcc phages with broad host ranges to treat antibiotic-resistant infections in CF patients. Here we report on the isolation and engineering of the distantly P2-like B. cenocepacia temperate phage Milagro to produce an obligately lytic derivative with a chimeric tail fiber. The engineered vir phage contains the C-terminal portion of the tail fiber taken from the broad host range Burkholderia tailocin BceTMilo (34) and has an expanded host range that is similar to that of the tailocin.

MATERIALS AND METHODS

Bacterial isolates, strains, and culture conditions

Bacterial strains and plasmids used in this study are listed in Table S1. The panel of Burkholderia isolates used in phage host range and efficiency of plating (EOP) are listed in Table S2. Tryptone nutrient broth (TNB) (5-g/L tryptone, 2.5-g/L yeast extract, 1-g/L dextrose, 8.5-g/L NaCl, and 2-g/L KNO₃) was used for Burkholderia isolates in liquid culture. Tryptone nutrient agar (TNA) (5-g/L tryptone, 2.5-g/L yeast extract, 1-g/L dextrose, 8.5-g/L NaCl, and 20-g/L agar) was used for routine maintenance of Burkholderia cultures. Luria-Bertani medium (10-g/L tryptone, 5-g/L yeast extract, and 10-g/L NaCl) was used for Escherichia coli and mating experiments. For counterselection, the cointegrant was grown in yeast extract-tryptone (YT) broth (10-g/L tryptone and 10-g/L yeast extract). YT broth supplemented with 150-g/L sucrose and 20-g/L agar was used to resolve the cointegrant. Antibiotics were added to the media at the following concentrations as required: 400-µg/mL kanamycin (Km), 20-µg/mL tetracycline (Tc) for B. cenocepacia, and 50-µg/mL Km for E. coli.

All Burkholderia and E. coli isolates were grown at 37°C with aeration in liquid or solid medium unless stated otherwise. Mid-log Burkholderia cultures were prepared by adjusting fresh TNB to an OD₆₀₀ of ~0.08 from an overnight culture (TNB) or overnight growth from a TNA plate. This culture was then incubated at 37°C with aeration until an OD₆₀₀ of 0.4–0.5 was reached.

Phage isolation, propagation, and plating

Phage Milagro was isolated from an enrichment of soil extract from Oswego County, NY, using the host B. cenocepacia isolate MS1. Briefly, a 10-g soil sample was added to 20 mL of phosphate buffer (12.5 mM, pH 7.0) amended with 1% (wt/vol) peptone (BD Bacto). After 2 hours of shaking at 150 rpm at 28°C, the suspension was centrifuged at 10,000 × g for 20 min, and the supernatant was filtered through a 0.22-µm membrane bottle-top vacuum filter system (Corning) and stored at 4°C. For the enrichment of phages, 5 mL of the soil filtrate amended with 2× TNB was added to 20 mL of a mid-log (OD₆₀₀ = 0.5) TNB culture of B. cenocepacia isolate MS1 and incubated at 37°C overnight with shaking (150 rpm). The overnight culture was centrifuged (10,000 × g, 20 min at 4°C), and the supernatant was filtered through a 0.22-µm membrane.

Phages were routinely isolated and propagated using the soft agar overlay method (35, 36) using TNA bottom plates and TNA soft agar (TNA amended with 1-mM MgCl₂ and 0.4% wt/vol agar) for the overlay, where 100 µL of the enriched supernatant and 100 µL of host suspension (OD₆₀₀ = 0.5) were mixed in 5 mL of tempered soft agar and poured onto a TNA plate. After overnight incubation at 37°C, isolated plaques were picked and plaque purified three times to ensure clonality. Stock phage lysates were produced using the confluent plate lysate method or by propagation in TNB liquid culture (35). Phages were routinely diluted in P buffer (100-mM NaCl, 25-mM Tris-HCl, pH 7.4, and 8-mM MgSO₄).

Phage genome sequencing and annotation

Phage genomic DNA was isolated following a modified protocol with the Promega Wizard DNA kit described by Summer (37). DNA libraries were prepared using an Illumina TruSeq Nano kit, and pair-end reads were sequenced on an Illumina MiSeq instrument. The genome sequence was assembled with SPAdes v.3.5.0 (38) with default parameters, and the assembled genome were verified by PCR using primers facing off the contig ends and Sanger sequencing of the resulting PCR product. Genes were called by GLIMMER v.2.36 (39) and MetaGeneAnnotator v.1.0 (40). tRNA genes and terminators were predicted by ARAGORN v.2.36 and TransTermHP v.2.09 (41) (42). The identified genes were assigned putative functions using BLAST v.2.9.0 (43) with the non-redundant, Swiss-Prot databases (44), interProScan v.5.33 (45), and TMHMM v.20 (46). Genome annotation was conducted using tools and workflows hosted on the Center for Phage Technology Galaxy and Apollo instances with default settings (https://cpt.tamu.edu/galaxy-pub) (47, 48).

Lysogen isolation, autoplaquing, and purification of virulent mutant phages

To isolate potential lysogens of phage Milagro, a 1-mL pipette tip was used to touch the center of a turbid phage plaque and streaked on TNA to isolate single colonies. Colonies were picked and subcultured three times to ensure clonality. MS1(Milagro) lysogens were confirmed by colony PCR using the primer set Milagro_030-Sp-F vs Milagro_031-Sp-R (Table S3). Select lysogens were streaked for single colonies and used to make bacterial lawns to observe for autoplaques. The term “autoplaque” is used to describe the spontaneous appearance of virulent mutant plaques on lawns of bacteria alone, in contrast to those that appear when exogenous phages are plated on a sensitive indicator bacterial host. Autoplaquing was observed for Milagro lysogens. Presumed virulent Milagro mutants were single plaque purified three times.

Characterization of vir mutants of phage Milagro

Phage Milagro vir mutants were picked from lawns of six biologically independent replicates with three to four plaques picked per replicate, which resulted in 22 clones. Whole genome sequencing of 12 of these vir mutants was conducted as described above. Mutations were identified by comparing the complete genome sequence of each vir mutant to that of the parental phage Milagro. Based on the locations of the mutations identified in these 12 vir mutants, this region was amplified from the additional 10 vir mutants using the primer set Milagro_030-Sp-F and Milagro_031-Sp-R (Table S3), and the resulting products were sequenced by Sanger sequencing.

Virulence assays with selected vir mutants of phage Milagro

The microtiter plate assay developed by Xie et al. (49) was used for the measurement of phage virulence, which is defined as the ability of a specific phage to control the growth of its host in liquid culture. Briefly, inoculum was prepared from overnight host cultures that were adjusted with TNB to OD₆₀₀ of ~1 (10⁹ CFU/mL). Phage lysates were adjusted to ~10⁷ PFU/mL with growth media. For each assay, 20 µL of adjusted bacterial inoculum and phage lysate was mixed in a 160-µL medium in untreated 96-well transparent plates (Corning Cat. No. 351172) to achieve a multiplicity of infection (MOI) of ~0.01. The plates were incubated at 37°C with double orbital shaking in a Tecan Spark 10-M plate reader (Tecan Group Ltd., Männedorf, Switzerland), and OD₆₀₀ was measured at 30-min intervals for 20 hours. Growth curves were plotted using OD₆₀₀ against time after medium blank baseline adjustment. Three independent replicates for each treatment were conducted in triplicate.

RNAseq of the Milagro lysogen

MS1(Milagro) lysogens were exposed to Milagro vir phage, and MS1(Milagro) mutants resistant to phage Milagro vir were isolated and subcultured to ensure clonality. These phage-resistant lysogens were cultured to the exponential phase of growth (OD₆₀₀ of ~0.3), and cell pellets were obtained and stored at −80 C. Two independently generated cell pellets were sent to Zymo Research Corp. (Irvine, CA) for total RNA-seq, which included sample QC and standardization, rRNA depletion, and stranded RNA library preparation by the Zymo-Seq RiboFree Total RNA Library Kit (Zymo), and sequencing on an Illumina platform. Reads obtained from RNAseq were mapped to the Milagro sequence using Bowtie2 v.2.4.2 (50) at default settings and analyzed using featureCounts v.2.0.1 (51) to count reads per gene. Read counts were normalized to gene length and total mapped reads to calculate read counts as transcripts per kilobase million. These analyses were conducted on the public Galaxy instance located at usegalay.org (48).

Construction of recombinant phage tail fiber

To expand the host range of phage Milagro, the allelic replacement system developed by Hamad et al. (52) and modified by Yao et al. (34) was used to produce recombinant phages containing the C-terminal of tail fiber gene (from amino acid 381 to 884) and the tail fiber assembly gene from tailocin BceTMilo. To accomplish this, a sequence encoding 503 amino acids of the C-terminal end of the BceTMilo tail fiber and the full-length tail fiber assembly gene of BceTMilo with flanking 1 kb of Milagro sequence upstream and downstream of the allelic exchange site (4,241 bp) was synthetically produced (GenScript) and cloned into a carrier vector (pCDNA3.1) with NheI/HindIII cloning sites (Table S4). The synthesized sequence was isolated from the carrier vector, purified by gel electrophoresis, and cloned into the suicide vector pMO130, resulting in pMO130::MiloTF380TFA-MilagroUD (52). Sanger sequencing was used to confirm the MiloTF380TFA-MilagroUD construct using primers Milagro:Milo_TF-Sp-F1, F2, F3, F4 and Milagro:Milo_TF-Sp-R1, R2, R3, R4 (Table S3). Plasmid pMO130::MiloTF380TFA-MilagroUD was introduced into the B. cenocepacia AU41545(Milagro) lysogen (Table S1) using triparental mating with pRK2013 as the mobilizing plasmid. Transconjugants were selected on TNA plates amended with Km 400 µg/mL and Tc 20 µg/mL. The resulting colonies were sprayed with 0.45-M pyrocatechol to identify yellow colonies in which single crossover events had occurred due to insertion of the xylE. Single yellow colonies were picked for resolving through double crossover events with counter selection using 15% sucrose in YT medium. The presumptive resolved cointegrants exhibiting a white phenotype were analyzed by PCR using multiple primer combinations both internal (Milagro_TF-Sq-F/Milagro_TFA-Sq-R) and external (Milagro: Milo_TF-Sp-F1/Milagro:Milo_TF-Sp-R2) to the targeted sequence and by the sequencing of PCR products to confirm the recombination (Table S3). Clones of the resolved chimeric lysogen were plated as overlays to obtain individual plaques from autoplaquing lawns. Individual plaques were plaque purified three times using host AU41545.

Host range of tailocin BceTMilo

The tailocin BceTMilo was prepared and titered, and the host range was determined using spot assays as described previously (34). Briefly, bacterial lawns were made as described above, and fivefold serial dilutions of the tailocin were spotted to lawns, allowed to dry, and incubated overnight at 37°C. Spot assay results are representative of three independent biological replicates. Sensitivity to BceTMilo was denoted as “+,” indicating a clear zone; “T,” indicating a turbid zone; and “−,” denoting no zone observed in the lawn when spotting ~3,125 arbitrary units of BceTMilo as determined using the indicator strain K56-2 ΔwaaL. K56-2 ΔwaaL was used as a positive control in all assays.

Host range and plating efficiency of parental and recombinant phages

Phage host range was determined by the spot method as described previously (53). Briefly, test isolates along with positive controls were incubated at 37°C for 48–72 hours until single colonies were obtained. Three single colonies per isolate were then streaked onto TNA and allowed to incubate at 37°C 18–20 hours. After incubation, growth from each single colony was suspended in TNB to an OD₆₀₀ of 0.4–0.45. Bacterial lawns were made as described above, and 10 µL of 10-fold serial dilutions of phage lysates was spotted onto host lawns, allowed to dry, and incubated at 37°C overnight. Plaques were enumerated, and EOP was calculated as the titer of the phage on the test bacterial host divided by the titer on the phage propagation host (EOP = 1). Spot assay results are represented as the means of three to four independent biological replicates.

Phage adsorption assays

Phage Milagro vir gp20:Milo propagated on host AU41545 was used for all adsorption assays. In brief, fresh overnight TNB cultures of Bcc isolates were subcultured into fresh TNB and allowed to grow to an OD₆₀₀ of 0.45. Ten microliters of phage lysate containing 1 × 10⁸ PFU/mL was added into 1 mL of each culture and incubated for 25 min at room temperature with gentle rotation. Phage added to 1 mL of sterile TNB was used as a control. Following incubation, the samples were centrifuged at 7,200 × g for 5 min to obtain unadsorbed phage in the supernatant, followed by filtration of the supernatant through a 0.2-µm filter. Unadsorbed phages were enumerated by dilution and plating to soft agar overlays using AU41545 as host. Adsorbed phage was calculated as 1 minus the titer of the phage remaining in the supernatant divided by the titer observed in the TNB control.

RESULTS AND DISCUSSION

Milagro isolation

The B. cenocepacia isolate MS1 was obtained in 2017 from a lung transplant recipient at the University of Pittsburgh Medical Center. Phage Milagro was isolated by enrichment from soil samples collected in July 2018 from Oswego, New York, against MS1. Transmission electron microscopy of Milagro showed a typical myophage morphology similar to that of P2 and KL3 (54, 55) (picture not shown). Milagro forms turbid plaques on MS1 and MS1(Milagro) lysogens could be readily isolated from the centers of these turbid plaques.

Phage Milagro genome

The Milagro genome was sequenced by Illumina sequencing and assembled in SPAdes to produce a single contig of 39,088 bp at a k-mer coverage of 113×. The completeness of the assembled Milagro genome was verified by PCR amplification of the assembled contig ends followed by Sanger sequencing of the amplified PCR products. DNA packaging strategy of phage Milagro was determined to be a site-specific cos mechanism, and the genome was reopened near the cos site at a point consistent with restriction enzyme digest patterns observed with Milagro genomic DNA (data not shown); however, the precise cos termini were not determined (56). The annotated Milagro genome was deposited to NCBI GenBank under accession OM638609. Phage Milagro was determined to be 39,088 bases in length, with a 63.9% G + C content. The genome (Fig. 1) was found to encode 56 protein-coding genes, and a detailed annotation of the Milagro genome is shown in Table S5.

Fig 1.

Genome maps illustrating functional annotation and protein relationships of phages Milagro, KL3, P2, and the tailocin BceTMilo. Protein-coding genes are represented as boxes with genes encoded on the plus and minus strands indicated by the pointed ends facing right or left, respectively. Genes are color coded based on predicted function as indicated in the legend; genes encoding major structural functions are labeled. Protein sequence relatedness as determined by Blastclust (>30% identity over >30% protein length) is indicated by the gray shading between organisms. Top panel: phage Milagro vs Burkholderia phage KL3 and coliphage P2; lower panel: phage Milagro vs the Burkholderia tailocin BceTMilo. The map of coliphage P2 is also labeled with its canonical gene names.

BLASTn analysis showed that Milagro shares 70.5% DNA sequence similarity with Burkholderia phage KL3 (55), which is its the closest named relative in the ICTV phage taxonomy (57). This would place Milagro in a new species in the genus Kayeltresvirus in the family Peduoviridae, founded by the well-studied coliphage P2 (54). In addition to KL3, Milagro is also related to the Burkholderia tailocin BceTMilo (34), with 74.8% DNA sequence similarity as determined by BLASTn. Protein-level relationships between Milagro, KL3, BceTMilo, and P2 as determined by blastclust are shown in Fig. 1 and listed in Table S6. Overall, 41 of the 56 annotated Milagro proteins are conserved and syntenic in phage KL3, including all structural, lysis, and lysogeny control functions. The right end of the Milagro genome, which contains the phage integrase, hypothetical proteins, and predicted morons [highly variable genes expressed from the prophage state (58)], is not conserved with KL3. Twenty-one Milagro proteins are also directly conserved in coliphage P2, primarily in the structural components (Fig. 1, light and dark green). Milagro shares 28 proteins including tail structural components, lysis functions, and part of the lysogeny control region with the BceTMilo tailocin (Fig. 1), which suggests that this tailocin is derived from a Milagro-like or KL3-like temperate phage.

Spontaneous virulent mutants of Milagro

Lawns of MS1(Milagro) lysogens exhibited an autoplaquing phenotype capable of forming plaques on lawns inoculated with the lysogen alone (Fig. 2). The number of autoplaques per lawn varied from culture to culture, ranging from no observable plaques in some cultures to confluent lawns in others. Individual autoplaques were picked and plaque purified using MS1 as the host. These phages were found to be true vir mutants, unable to form lysogens, and able to infect MS1(Milagro) lysogens (Fig. 2). As noted by Lynch et al. (55), the P2-like phage KL3 was isolated from a spontaneous single plaque arising on a lawn of B. cenocepacia isolate CEP511. We confirmed that spontaneous autoplaquing events also occur abundantly in cultures of B. cenocepacia strain MS1(Milagro) under normal growth conditions (Fig. 2). Spontaneous autoplaquing has also been observed in other bacterial genera such as Helicobacter (59), Neisseria (60), and Brucella (61); however, the mechanisms underlying these observations were not determined.

Fig 2.

Spotting of temperate phage Milagro and spontaneous autoplaquing Milagro vir mutant phage on lawns of the B. cenocepacia MS1 and the Milagro lysogen MS1(Milagro). The parental MS1 (left) is sensitive to both wt Milagro and Milagro vir. The lysogen is immune to Milagro but is sensitive to Milagro vir. Note the spontaneous autoplaque formation throughout the lawn of MS1(Milagro), the result of spontaneous mutation of the wild-type Milagro to its vir form.

To determine the genetic basis of this autoplaquing, the DNA sequences of 22 autoplaquing Milagro clones collected from lawns of six biologically independent replicates were determined by a combination of whole genome sequencing (12 clones) and targeted sequencing of the phage lysogeny control region (10 clones). In all cases, the only mutations identified were inside or slightly upstream of a predicted promoter designated P_R within the predicted lysogeny control region, with nine unique genotypes identified across the 22 sequenced clones (Fig. 3). The predicted Milagro lysogeny control locus is a set of divergently transcribed genes driven by two sigma70 promoters which we have designated P_R and P_L, arranged in a manner similar to P2 but with different genetic content (Fig. 1, orange genes; Fig. 3). The leftward-facing transcript encodes a predicted lambda CI-like transcriptional repressor (gp30) with an N-terminal lambda-like DNA binding domain (IPR010982 and IPR001387) and a second gene of unknown function. The rightward transcript encodes four proteins, three of which have no predicted function but have DNA binding domains detectable by HHpred (e.g., 2K29 RelB, 95%–99%). The fourth protein (gp34) has detectable similarity to the P2 late transcriptional activator Ogr (IPR007684). Each of these transcripts are terminated by rho-independent terminators, T_L and T_R (Fig. 3).

Fig 3.

The lysogenic switch region of phage Milagro. (Top) Predicted promoters P_R and P_L (green arrows) drive short transcripts terminated by rho-independent terminators T_R and T_L (red octagons). The P_L transcript encodes a predicted transcriptional repressor (gp30) and a gene of unknown function, and the P_R transcript encodes four proteins with predicted DNA binding domains but are otherwise of unknown function. (Bottom) In this zoomed-in view of P_R, locations of mutations found in multiple autoplaquing Milagro vir mutants are aligned to the Milagro wild-type sequence. nine unique mutations were identified in this region, corresponding to two possible operator sites that bind the phage lysogenic repressor. Nucleotide position in the parental Milagro genome is shown below the figure.

No mutations were found within the predicted −10 and −35 sites but were found either between these sites or immediately upstream of the −35 site. Coupled with the protein functions described above, these findings strongly support this region’s role as a key control point for lysogenic repression in phage Milagro. In related temperate phages such as coliphages P2 and 186, operator binding sites are recognizable as repeated DNA sequence motifs in or near phage promoters. In P2, the repressor protein C binds to two perfect 8-bp repeats flanking the −10 site of promoter P_e (54), while in phage 186, the lysogenic repressor CI binds as an octamer to multiple DNA sites composed of two different classes of imperfect inverted repeats (62). There is also precedence for a phage repressor to bind as a monomer to multiple different operator sites via independent DNA binding domains (63). In Milagro, we could not identify any clear repeated motif or inverted repeats in the regions defined by the vir mutations at the P_R promoter, suggesting an asymmetric operator binding site. However, the two regions where vir mutations appear are spaced ~30 bp apart, consistent with the spacing observed in operator sites in coliphage 186 (64). We hypothesize that these mutations disrupt an operator binding site for the gp30 lysogenic repressor, allowing transcription from P_R and rendering the phage unable to maintain a lysogenic state and insensitive to the gp30 repressor produced in the lysogen. This model suggests that the genes encoded on the P_R transcript are responsible for driving the phage lytic cycle, for example, by encoding additional transcriptional activators or antiterminators. The presence of an ortholog of the P2 Ogr protein on this transcript supports this role of the P_R transcript, as Ogr is required for activation of the lytic pathway in P2 infection (65). The pattern of mutations observed in Milagro vir mutants is similar to that observed in the P2-like coliphage 186, in which vir phenotypes were associated with mutations at the P_R promoter between the −10 and −35 sites and upstream of the −35 site (66). In the case of phage 186, however, all vir mutants have multiple mutations in this region (66), whereas in phage Milagro, a single SNP appears to be sufficient to confer the vir phenotype.

Expression of the phage lysogenic repressor and morons in the MS1(Milagro) lysogen

RNAseq was used to determine genes expressed from the uninduced MS1(Milagro) lysogen. This was done to identify genes involved in lysogeny maintenance and potential host fitness factors expressed from the dormant lysogen (i.e., morons). To avoid lytic gene expression from the Milagro vir phages that spontaneously arise in MS1(Milagro) cultures, phage-resistant mutants of the MS1(Milagro) lysogens were used for this experiment. Stranded transcriptional profiling of two independent MS1(Milagro) lysogens (Fig. 4) indicated that most phage genes were transcriptionally inactive in the dormant lysogen with the exception of two small regions of the prophage. Genes 29 and 30, corresponding to the P_L transcript, were found to be expressed at moderate levels. Gene 30, encoding the predicted lysogenic repressor, is expected to be expressed in the dormant lysogen in order to maintain the lysogenic state (67). The function of gene 29 is not known; it may be also involved in lysogeny maintenance as it shares a transcript with gene 30. Mapping of raw reads to the Milagro genome (Fig. S1) showed a similar pattern of expression to that shown in Fig. 4.

Fig 4.

Gene expression levels in the Milagro lysogen as determined by RNAseq. Sequence reads were mapped to the Milagro DNA sequence in Bowtie2 and mapped reads per gene determined in Featurecounts. Read counts were normalized to gene length and total mapped reads to calculate read counts as transcripts per kilobase million (TPKM, Y axis) on a per-gene basis, with genes denoted as 1–56 on the X axis. The results from two biological replicate experiments are shown (black and gray bars). Genes 29 and 30, corresponding to the P_L transcript, are transcribed above background levels. Also highly transcribed are genes 45–49, a region containing the two predicted moron cassettes. The locations of these highly expressed genes are shown in the partial Milagro genome map below the figure.

The second locus of gene expression in the dormant Milagro lysogen was identified as genes 45–49, which appear to encode a pair of two-gene modules separated by a single gene encoding a VSR-like endonuclease. The first module (45, 46) is hypothesized to function as a toxin-antitoxin (TA) system based on the presence of a predicted membrane protein and a protein related to IrrE protease (IPR010359) and the HIGA1 antitoxin (PTHR43236). The second module encodes a clearly identifiable restriction-modification (RM) system, with gene 48 encoding a predicted C5 cytosine methyltransferase (IPR001525) and gene 49 encoding a type II restriction endonuclease (IPR018575). Temperate phages are often found to carry variable host fitness factors that can be expressed from the lysogenic state, which are referred to as morons (58). Such factors may be relatively innocuous such as the antiphage systems carried by phage P2, which exclude infection by phages T4, T5, and lambda (54), or may be potent virulence factors such as the Shiga toxin carried by lambdoid prophage 933W (68). In the case of Milagro, its two expressed modules appear to be of the former category, as both TA and RM systems have been implicated in antiphage host defense by inducing cell death early after phage infection (69) or by cleaving incoming phage DNA (70), respectively.

Virulence of selected Milagro vir mutants

Since multiple different mutations were observed at the P_R locus (Fig. 3), it was of interest to determine if the various mutant genotypes exhibited differences in phage virulence. Five of the mutant genotypes were chosen as representatives, and virulence assays were performed in liquid cultures using MS1 as bacterial host at an MOI of 0.01 at the start of each assay. Results showed no significant differences among the vir mutants with different mutations, indicating that the mutant genotypes are functionally the same (Fig. 5). As expected, the parental temperate phage Milagro showed a significantly different lysis curve as compared to the vir mutants.

Fig 5.

Phage virulence assays showing the ability of the temperate phage Milagro and its vir derivatives to control the growth of B. cenocepacia MS1 in liquid culture. Phages were added to cultures of MS1 at an initial MOI of 0.01, and bacterial growth was monitored by optical density over time. Phage vir mutant genotypes are shown in the legend, error bars are the standard deviation for three replicate experiments. Milagro vir mutants showed a greater ability to control bacterial growth than the parental phage (Milagro wt, light blue).

Engineering of the Milagro tail fiber

The diversity of Bcc clinical isolates and the generally narrow host ranges of Bcc phages (23) necessitates the ability to alter or expand phage host range to target the desired Bcc clinical isolates. As described earlier, phage Milagro is related to phages KL3, P2, and the tailocin BceTMilo. This relationship of phage Milagro to other studied phages and tailocin elements makes this a natural entry point for the development of engineered P2-like virulent phages that can be used as therapeutics against Bcc infections. We hypothesized that this could be accomplished by replacing the C-terminal receptor-binding domain (RBD) of the Milagro tail fiber with that of the broad host-range tailocin BceTMilo, which recognizes the α-glucose of the core oligosaccharide (34). As shown in Fig. S2, the Milagro and BceTMilo tail fibers are closely related with notable sequence divergence between alignment residues ~780 to ~835, a C-terminal location consistent with the expected location of the RBD based on previous studies of phage tail fiber structures (71) and predictions of tail fiber structure by AlphaFold2 (72). In this domain swapping approach, the DNA encoding the C-terminal RBD and tail fiber assembly genes of the broad host range tailocin BceTMilo were fused to the N-terminal virion attachment domain of the Milagro tail fiber gene. Protein sequence alignments guided the localization of the fusion junction with candidate fusion targets between residue K380 and R381 of the Milagro tail fiber as shown in Fig. S2. To construct a recombinant derivative of phage Milagro, it was first necessary to construct a stable lysogen in a suitable host. For this purpose, we used the host B. cenocepacia AU41545, a clinical isolate that is sensitive to phage Milagro and selectable using kanamycin. The Milagro lysogens formed in isolate AU41545 also autoplaqued to produce vir mutants of phage Milagro similar to the MS1(Milagro) autoplaques shown in Fig. 2, indicating that this property is phage-specific rather than host-specific, and made it possible to isolate Milagro vir mutants after tail fiber modification.

The engineered tail fiber fusion was constructed by homologous recombination between a Milagro lysogen and a plasmid construct bearing the Milagro gp20:Milo tail fiber fusion (Fig. 6). The resulting chimeric prophage, designated Milagro gp20:Milo, also produced autoplaques when individual colonies were used to make bacterial lawns, indicating that spontaneous vir mutants of the chimeric phage Milagro gp20:Milo can still infect isolate AU41545. Individual plaques were picked, purified and propagated on host AU41545. One mutant with the vir genotype 24,238 C>T, designated as Milagro vir gp20:Milo, was selected and propagated on AU41545. The presence of the recombinant tail fiber was confirmed by complete sequencing of the Milagro vir gp20:Milo phage.

Fig 6.

Partial map of the Milagro genome showing the location of the BceTMilo tail fiber fusion. Protein-coding genes are shown as directional blocks, with green genes indicating the location of native Milagro sequence and light green indicating the location of BceTMilo sequence, which encompasses the C-terminal portion of the tail fiber gene (gene 20) and the entire tail fiber assembly chaperone (tfa, gene 21). The enlarged region (top) shows the amino acid sequences of the Milagro and BceTMilo tail fiber proteins and the point of fusion, located between K380 and R381 of the Milagro protein.

Host ranges of phage Milagro and its derivatives

The host ranges of Milagro and its derivatives were assessed against a diverse panel of Bcc isolates representing multiple species commonly associated with infections in CF patients. As shown in Table 1, the parental phage Milagro and its vir derivative have relatively narrow host ranges, with Milagro infecting only its original isolation host (MS1), three other B. cenocepacia isolates, one Burkholderia vietnamensis isolate, and one B. multivorans isolate. When temperate phages lysogenize a new bacterial host, they typically confer immunity to infection by the same phage through the action of the phage repressor protein on the incoming phage (73, 74). It was hypothesized that the Milagro vir phage may have an expanded host range in the Bcc due to the fact that its vir phage mutants could ignore lysogenic repression and infect hosts carrying Milagro-like lysogens. However, we observed that the host range of Milagro vir was only slightly expanded compared to the parental phage, gaining the ability to form plaques at low efficiency on an additional B. multivorans and B. cepacia isolate (Table 1). This indicates that the presence of Milagro-like lysogens in clinical Bcc isolates is not a major factor controlling the host range of this phage.

TABLE 1.

Host ranges of tailocin BceTMilo and phages Milagro, Milagro vir, and Milagro vir gp20:Milo ^d

Burkholderia species and strain	Tailocin BceT milo	Milagro wt	Milagro vir	Milagro vir gp20:Milo
B. cenocepacia
MS1	T	1 b	1 b	5.6E-06
K56-2	+	−	−	0.77
K56-2 ΔwaaL	+ a	−	−	0.63
HI2718	+	0.22	0.18	0.45
VGH	+	−	−	0.66
AU41263	+	0.63	1.35	0.9
AU41545	+	1	1.22	1 c
B. multivorans
AU6960	+	−	−	−
AU6654	+	1.95E-05	6.50E-05	1.8E-05
AU42676	+	−	−	1.8E-07
AU41974B	+	−	3.80E-06	3.8E-05
AU44839	+	−	−	4.0E-09
B. gladioli
BgPK	+	−	−	2.9E-05
AU12635	+	−	−	1.6E-02
B. dolosa
AU3271	+	−	−	7.2E-06
AU0158	+	−	−	−
B. vietnamensis
AU5003	+	0.7	0.87	0.66
B. cepacia
AU4880	+	−	−	1.2E-05
AU2769	+	−	3.70E-05	1.2E-06

^a Indicates positive control for tailocin Milo.

^b Indicates propagation host for Milagro wt and Milagro vir = MS1.

^c Indicates propagation host for Milagro vir gp20:Milo = AU41545.

^d The host range of BceTMilo is shown as producing a clearing zone (+), turbid zone (T) or no clearing (-) when spotting 3125 arbitrary units to bacterial lawns. The host ranges of the phages are shown as efficiency of plating (EOP) relative to their propagation hosts.

In contrast to Milagro vir, the engineered phage Milagro vir gp20:Milo gained the ability to infect multiple new isolates of B. cenocepacia, B. cepacia, B. gladioli, B. multivorans, and B. dolosa with a host range pattern that resembles that of the tailocin BceTMilo (Table 1). However, there are multiple isolates for which Milagro vir gp20:Milo had a low plating efficiency as compared to its propagation host AU41545. Among the isolates that were sensitive to tailocin BceTMilo, B. multivorans AU6960 and B. dolosa AU0158 showed no plaque formation when infected with phage Milagro vir gp20:Milo (Table 1). All other isolates tested showed plaque formation by Milagro vir gp20:Milo with EOPs ranging from ~1 to ~10⁻⁹.

To further investigate the effect of phage Milagro vir gp20:Milo adsorption on plating efficiency, phage adsorption experiments were conducted with a subset of the host panel shown in Table 1, representing different species and a range of observed EOPs. Phage Milagro vir gp20:Milo was allowed to adsorb to each host, and the phage remaining in the supernatant was determined and normalized to the phage remaining in cell-free control experiments. As shown in Fig. 7, phage adsorption rates varied widely between isolates, with no clear relationship between adsorption rate and EOP. The phage propagation host AU41545 (EOP = 1) showed a moderate adsorption rate (43% adsorbed in 25 min), whereas AU6960 and AU0158 (no plaque formation) adsorbed >99% of the phage. Strain K56-2 and its isogenic waaL derivative produced similar EOP’s in the host range assay (0.77 and 0.63, respectively) but adsorbed phages at markedly different rates (20% and 95% phage adsorbed, respectively). This indicates that phage adsorption rate is not strongly predictive of the ability of the phage to form plaques on these hosts.

Fig 7.

Adsorption of phage Milagro vir gp20:Milo to Bcc isolates. Results are shown as the percent phage adsorbed after 25-min incubation with the bacteria. Results are normalized to the phage titer of a cell-free negative control. Results shown are the means of three independent replicates with error bars showing standard deviation. The efficiency of plating (EOP) of the phage on each isolate (see Table 1) is shown in the red text above the graph; isolates are presented in descending order of EOP, with EOPs below 0.1 rounded to the nearest order of magnitude for clarity. Phage EOPs in the set shown here range from 1.0 for the phage’s propagation host AU41545 to undetectable (“—”) for isolates AU6960 and AU0158. In general, the rate at which Milagro vir gp20:Milo adsorbs to a host does not correlate to its ability to form plaques on the same host.

These results suggest that phage bearing the engineered BceTMilo tail fibers are able to adsorb hosts that are sensitive to the tailocin even though they are unable to form plaques on some of these hosts and that the Milagro gp20 tail fiber C-terminal domain is sufficient for adsorption. However, these results also show that the same phage tail fiber can exhibit widely variable adsorption rates when presented with a diverse panel of hosts. Additional variability in phage EOP may also be attributable to intracellular antiphage mechanisms such as CRISPR-Cas, restriction, or abortive infection systems (75, 76).

Conclusion

Infections caused by the Bcc are typically difficult to treat due to their high inherent resistance to antibiotic therapy (3, 5). Phage therapy represents a viable strategy for the treatment of these infections. However, unlike phages of enteric bacteria such as E. coli and soil bacteria like Pseudomonas and Bacillus species, phages for Burkholderia are dominated by temperate phages (23). To be considered usable as therapeutics, it is generally agreed that phages should be strictly virulent (unable to form lysogens) and have host ranges that are broad enough within the target species to make them practically useful (22, 74, 77). Phage Milagro is a P2-like temperate phage that exhibits a high spontaneous mutation rate to virulence, making it a natural platform for engineering a therapeutic phage active against the Bcc. While spontaneous vir mutants of Milagro varied slightly by genotype, all had mutations in the predicted lysogeny control region that produced similar phenotypes (Fig. 3 and 5). However, the vir Milagro derivatives did not exhibit a significant increase in their host range over the parental temperate phage (Table 1). Replacement of the C-terminal half of the Milagro tail fiber gp20 with the orthologous domain from the BceTMilo tail fiber was shown to significantly expand the host range of the phage to include multiple additional Bcc species (Table 1). This demonstrates the key role of the gp20 phage tail fiber in determining host specificity in this phage and the utility of engineering approaches for the improvement of potentially therapeutic phages. However, we also found that phage adsorption does not guarantee efficient phage replication, as over half of the Bcc test panel isolates infected by the chimeric phage exhibited severely reduced plating efficiencies (Table 1). This suggests that non-receptor-mediated defense systems are active in many Bcc clinical isolates.

While the Milagro vir gp20:Milo phage described here exhibits greater virulence (Fig. 5) and a broader host range (Table 1) than its temperate parental phage Milagro, the use of temperate phages as platforms for therapeutics brings additional issues that must be addressed. Morons are commonly present in temperate phage genomes, and these elements may carry virulence factors or other genes that would potentially impact host fitness (58). While the morons identified in Milagro do not appear to be virulence factors, as a general rule, we propose it would be prudent to delete these genes unless they are shown to be required for phage function. In addition, deletion of accessory lysogeny functions such as the phage lysogenic repressor, integrase, and attP site would further enhance the safety of engineered temperate phages and reduce the possibility of reversion to a temperate lifestyle via recombination with prophages resident in the target host during therapy. Phage therapy offers a novel approach for the treatment of highly antibiotic-resistant bacterial infections, but in some systems, the utility of phage therapy is limited by the lack of naturally occurring virulent phages that are suitable as therapeutics. In such cases where there is a limited virosphere, we show that engineering can provide a flexible approach to enhance the availability of potentially therapeutic phages.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/jvi.00850-23.

Supplemental figures. jvi.00850-23-s0001.pdf.

Figures S1 and S2.

Tables S1 to S4. jvi.00850-23-s0002.pdf.

Table S1. Bacterial strains and plasmids used in this study. Table S2. Burkholderia isolates used for host range and EOP comparisons. Table S3. Primers used in this work. Table S4. Synthesized DNA fragment for recombinant tail fiber engineering.

Tables S5 and S6. jvi.00850-23-s0003.xlsx.

Table S5. Genes, annotations, and evidence for B. cenocepacia phage Milagro. Table S6. Comparison of protein-coding genes annotated in Burkholderia phages Milagro (OM638609) and KL3 (NC_015266), coliphage P2 (NC_001895), and the Burkholderia tailocin BceTMilo (KY316063).

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.