|Home | About | Journals | Submit | Contact Us | Français|
The Burkholderia cepacia complex (BCC) is made up of at least 17 species of Gram-negative opportunistic bacterial pathogens that cause fatal infections in patients with cystic fibrosis and chronic granulomatous disease. KS9 (vB_BcenS_KS9), one of a number of temperate phages isolated from BCC species, is a prophage of Burkholderia pyrrocinia LMG 21824. Transmission electron micrographs indicate that KS9 belongs to the family Siphoviridae and exhibits the B1 morphotype. The 39,896-bp KS9 genome, comprised of 50 predicted genes, integrates into the 3′ end of the LMG 21824 GTP cyclohydrolase II open reading frame. The KS9 genome is most similar to uncharacterized prophage elements in the genome of B. cenocepacia PC184 (vB_BcenZ_ PC184), as well as Burkholderia thailandensis phage E125 and Burkholderia pseudomallei phage 1026b. Using molecular techniques, we have disrupted KS9 gene 41, which exhibits similarity to genes encoding phage repressors, producing a lytic mutant named KS9c. This phage is incapable of stable lysogeny in either LMG 21824 or B. cenocepacia strain K56-2 and rescues a Galleria mellonella infection model from experimental B. cenocepacia K56-2 infections at relatively low multiplicities of infection. These results readily demonstrate that temperate phages can be genetically engineered to lytic form and that these modified phages can be used to treat bacterial infections in vivo.
The Burkholderia cepacia complex (BCC) is a group of at least 17 Gram-negative species, the first identified strains of which were characterized as onion pathogens by W. H. Burkholder (9). Although these bacteria have a number of beneficial activities, including the promotion of crop growth and the degradation of organic pollutants, they have gained notoriety in the last two decades as serious opportunistic pathogens (19, 21, 25). BCC species, particularly B. multivorans and B. cenocepacia, cause serious respiratory infections in patients with cystic fibrosis and chronic granulomatous disease (42, 7). These infections are especially problematic due to symptom severity, the inherent antibiotic resistance of Bcc species, and the potential for rapid spread through susceptible patient populations (25, 23). Difficulties in treating these infections have led to the unfortunate practice of segregating patients, which has high economic, social, and psychological costs (18).
Because of these clinical difficulties, interest in the isolation and characterization of Burkholderia-specific bacteriophages (or phages) has increased in recent years, with the apparent potential for using phages as therapeutic agents. Phage therapy is the clinical application of phages to prevent and/or to treat infections, which offers a promising alternative to antibiotic treatment for resistant bacteria such as those of the BCC (33, 39). A second benefit of these phage studies is that they may provide insight into the possible mechanisms of BCC virulence. For example, BcepMu, a transposable phage that specifically infects strains of B. cenocepacia, was found to carry genes similar to exeA, involved in toxin secretion, and mdmB and oafA, two acyltransferases (44). Finally, as Burkholderia phages tend to be underrepresented in comparative studies with respect to Escherichia coli and lactic acid bacteria phages, BCC-specific phage studies provide novel information about a relatively uncharacterized group of viruses.
Although phage therapy using temperate virions can be effective (39), there are several reasons why lytic phages are generally considered the most appropriate candidates for use in phage therapy. One of the concerns is that phage integration can lead to lysogenic conversion and enhanced virulence (8). A second concern is that integration of temperate phages results in superinfection immunity due to expression of the phage repressor from the prophage. This protein binds to the operators of infecting phage DNA and represses gene expression, preventing both the initiation of the lytic cycle and the establishment of lysogeny (14). A third concern is that lysogeny affects the kinetics of infection. When a phage infects a cell and undergoes lysogeny instead of entering the lytic cycle, the cell survives, and no new phage particles are released (27). A final problem is that prophages can lead to specialized transduction after induction. Specialized transduction occurs after inexact excision of a prophage from the bacterial chromosome. Bacterial DNA flanking the prophage is packaged into the capsid, and this sequence, which can potentially encode virulence factors, can subsequently recombine into the chromosome of a new host (14).
It has been estimated that more than half of tailed phages have evolved a temperate lifestyle, although some estimates have been greater than 90% (1, 22). This situation makes the isolation of naturally lytic phages extremely difficult, particularly when they must have a specific host range that includes clinically relevant bacterial species, such as B. cenocepacia (24). The use of classical genetics to produce lytic phage variants, for example, by plating temperate phages on lysogens and screening for clear plaque vir mutants, is complicated by the fact that such mutations are undefined.
This report describes the characterization of KS9 (vB_BcenS_KS9), a prophage of Burkholderia pyrrocinia LMG 21824 (41), and its conversion to a lytic phage through specific molecular modification of gene 41 encoding its putative lytic phase repressor. Preliminary characterization of short sequences by Seed and Dennis (41) indicated that the genome of KS9, whose host range includes Bcc B. cenocepacia K56-2, shows similarity to the genomes of two non-BCC Burkholderia phages: E125, a prophage of Burkholderia thailandensis E125 (47), and 1026b, a prophage of Burkholderia pseudomallei 1026b (17). However, no phages closely related to KS9 have been functionally tested to demonstrate that proteins similar to gp41 function as true phage repressors. In the present study, we have used the BCC infection model of Galleria mellonella (40) to assess both the contribution of the KS9 prophage to BCC host virulence and the ability of a genetically modified KS9 to treat B. cenocepacia infections without stably integrating into the host bacterial chromosome as a prophage.
Burkholderia cenocepacia K56-2 and B. pyrrocinia LMG 21824 are members of the BCC experimental strain panel (32) and the updated BCC strain panel (13), obtained from Belgium Coordinated Collection of Microorganisms LMG Bacteria Collection (Ghent, Belgium) and the Canadian Burkholderia cepacia Complex Research and Referral Repository (Vancouver, BC, Canada). K56-2 lipopolysaccharide (LPS) mutants were kindly provided by Miguel Valvano (University of Western Ontario, London, Ontario, Canada). These strains were grown aerobically overnight at 30°C on half-strength Luria-Bertani solid medium or broth. Transformations were performed with chemically competent DH5α (Invitrogen, Carlsbad, CA). Transformed DH5α were plated on LB medium containing 100 μg of ampicillin/ml or 100 μg of trimethoprim/ml and 15 μg of tetracycline/ml and grown overnight at 37°C. Electroporations were performed by using a Bio-Rad MicroPulser (Bio-Rad, Hercules, CA) and were plated on half-strength LB medium containing 300 μg of trimethoprim/ml and 300 μg of tetracycline/ml. Strains were stored at −80°C in LB medium containing 20% glycerol.
Overnight cultures of LMG 21824 were pelleted by centrifugation at 10,000 × g for 2 min. The supernatant was filter sterilized using a 0.45-μm-pore-size filter. This solution was incubated on a carbon grid for 5 min at room temperature and then stained with 2% phosphotungstate. A Philips/FEI (Morgagni) transmission electron microscope with charge-coupled device camera was used to take the KS9 micrograph, with the assistance of the University of Alberta Department of Biological Sciences Microscopy Service Unit.
KS9 was isolated by Seed and Dennis (41) from a single plaque on a lawn of B. pyrrocinia LMG 21824. Phage stocks were prepared in 1.5 ml of suspension medium (50 mM Tris-HCl [pH 7.5], 100 mM NaCl, 10 mM MgSO4, 0.01% gelatin solution) containing KS9 plaques isolated using a sterile Pasteur pipette. After the addition of CHCl3 and incubation for 1 h at room temperature, stocks were stored at 4°C. Propagation and plaque assays of KS9 were performed as described previously (41). A 100-μl portion of phage stock was incubated for 20 min at room temperature with 100 μl of B. cenocepacia K56-2 overnight culture. After the addition of 3 ml of soft nutrient agar, this mixture was poured onto half-strength LB medium and grown overnight at 30°C. For efficiency of plating (EOP) determinations, this procedure was repeated in three separate trials. For DNA isolation, plates of half-strength LB medium showing confluent KS9 lysis were prepared as described above using 100 μl of a KS9 high titer stock, 100 μl of K56-2, and 3 ml of soft nutrient agarose. Phages were isolated from these plates by overlaying with suspension medium, and the DNA was extracted by using the Wizard Lambda Preps DNA purification system (Promega, Madison, WI).
KS9 DNA was digested with EcoRI and SalI (Invitrogen), and the restriction fragments were separated on 0.8% (wt/vol) agarose gels in 1× Tris-acetate-EDTA (pH 8.0). These fragments were purified by using the GeneClean II kit (Qbiogene, Irvine, CA) and ligated into pUC19. Blue/white selection was performed on LB medium containing 100 μg of ampicillin/ml after transformation of the constructs into DH5α. Constructs were isolated by using a QIAprep miniprep kit (Qiagen, Hilden, Germany). The presence of an insert in pUC19 was verified by restriction digest and agarose gel electrophoresis. The sequencing of plasmid inserts was performed on an ABI 3100 genetic analyzer (Applied Biosystems, Foster City, CA) by the University of Alberta Department of Biological Sciences Molecular Biology Service Unit.
Sequences were edited with EditView (Perkin-Elmer, Waltham, MA) and aligned into a single contig by using AutoAssembler (Perkin-Elmer). A total of 327 sequences were used, with an assembled length of 39,896 bp. Primers (Sigma-Genosys, Oakville, Ontario, Canada) designed to amplify internal regions of the plasmid inserts were used for primer walking. Gaps between sequences were filled by PCR amplification and cloning and sequencing of the resulting amplicons using the TOPO TA kit (Invitrogen). This protocol was also used to sequence the KS9 insertion sites (LMG 21824 forward primer, 21824F [CCCACGCGCTACGGTACG]; KS9 reverse primer, KS9R [CCGATGTAGTCCAGGCACACC]; KS9 forward primer, KS9F [CACTGGGCGCCCGTTGAG]; LMG 21824 reverse primer, 21824R [AGGTTGGTGTCGGCCGTCC]). PCR was performed with an Eppendorf MasterCycler gradient DNA thermal cycler (Eppendorf, Hamburg, Germany) using TopTaq DNA polymerase (Qiagen) and the recommended reaction composition and cycling conditions.
To identify the putative overlap region of the KS9 attP site, the sequence between gene 30 and the integrase gene, 31, was analyzed by using BLASTN. This region showed 71% identity with a 207-bp region (positions 1155342 to 1155547) of Burkholderia lata 383 chromosome 2 (also known as ATCC 17760/LMG 22485). This region of similarity is longer than attL because it includes both attL and a downstream region similar to KS9. The Burkholderia lata 383 DNA sequence flanking the similar region was used to design primers for amplification of the junction between the LMG 21824 genome sequence and the 5′ end of the KS9 prophage. The site of integration was then verified by sequencing this amplicon and a second amplicon from the 3′ KS9/LMG 21824 junction. The K56-2/5′ KS9 junction in K56-2::KS9 was also identified by using this procedure.
The assembled sequences were analyzed by using the NCBI suite of programs. Annotation was performed using this suite, including BLASTX (3; http://blast.ncbi.nlm.nih.gov/) and ORF Finder (http://www.ncbi.nlm.nih.gov/gorf/gorf.html), and GeneMark.hmm-P (30; http://exon.biology.gatech.edu/) (Table (Table1).1). In all cases (except genes 12 [expressed by a translational frameshift] and 24′ [a gene embedded in 24]), the manual assignments matched the GeneMark results. Gene and gene product numbers were assigned based on the numbering system of E125/1026b. The predicted KS9 protein most similar to gp2 of these phages was assigned the name gp2 and the gene encoding this product was named gene 2. Subsequent proteins were named gp3, etc. Protein transmembrane domains were identified by using OCTOPUS (46; http://octopus.cbr.su.se/). Stem-loop structures were identified by using mfold (50; http://mfold.bioinfo.rpi.edu/). Signal peptide cleavage sites were identified by using LipoP (26; http://www.cbs.dtu.dk/services/LipoP/). The sequences of KS9, 1026b, and B. cenocepacia PC184 were compared by using the Artemis Comparison Tool (11). The KS9 genome map was prepared by using GenVision (DNASTAR, Madison, WI).
Plasmid pKL1 was created by ligating pALTER-1 (Promega) to an EcoRI/KpnI PCR amplicon of KS9 bp 5054 to 6078 (EcoRIF, AAGAATTCCAGCGCGGCATCG; KpnIR, TTGGTACCCGCCGTGTGCTTG), the 630-bp KpnI fragment of p34S-Tp2 (15, 16), and a KpnI/BamHI PCR amplicon of KS9 bp 6481 to 7580 (KpnIF, AAGGTACCGTCTGCAATTCAATAGC; BamHIR, TTGGATCCTTGGTGCTTTCTCG). LMG 21824 was electroporated with pKL1 and mutants with a single crossover (LMG 21824::pKL1) were selected for on LB medium containing 300 μg of trimethoprim/ml and 300 μg of tetracycline/ml. Construction of LMG 21824 (KS9 32) has been described previously (31). Briefly, an internal gene 32 segment was amplified and ligated into the oriR/TpR of pTnMod-OTp′ (15). This construct was electroporated into LMG 21824, and transformants were selected on LB medium containing 300 μg of trimethoprim/ml. To isolate phages from LMG 21824::pKL1, overnight cultures of these mutants were pelleted by centrifugation at 10,000 × g for 2 min, and the supernatant was filter sterilized using a 0.45-μm-pore-size filter. Plaque assays were performed as described above, and single plaques to be screened were added to 500 μl of suspension medium. PCR screening was performed by using the primers 40F (TCGTGACTGGCTGTTTTCGGAC) and 42R (GCGGCCAATTTCACGAGTCG). The TopTaq DNA polymerase reaction composition and cycling protocol were used with the replacement of template DNA by 1 μl of phage suspension.
To isolate KS9- and KS9c-insensitive K56-2 (including K56-2::KS9), plates of K56-2 and KS9 or K56-2 and KS9c exhibiting confluent lysis were prepared as described above and incubated overnight at 30°C. Plates were overlaid with 3 ml of water and placed on a rocking platform at 4°C for 4 h. The water was recovered, and the cells were pelleted by centrifugation at 10,000 × g for 2 min. The supernatant was removed, and the cells were resuspended in water and plated on half-strength LB medium. Single colonies were isolated and colony PCR was used to screen for lysogeny using primers 21824F and KS9R. In this procedure, prior to addition of Taq in the TopTaq protocol, a colony is added to the reaction mixture, incubated at 99.9°C for 5 min, and cooled on ice.
We followed the procedure for G. mellonella infection and treatment outlined by Seed and Dennis (39, 40). Briefly, 1 ml of an overnight culture of K56-2 or K56-2::KS9 was pelleted by centrifugation at 10,000 × g for 2 min and resuspended in 1 ml of 10 mM MgSO4 supplemented with 1.2 mg of ampicillin/ml. To compare the virulence of K56-2 and K56-2::KS9, serial dilutions from 101 to 107 (bacterial concentrations of ~106 to 100 cells/5 μl, respectively) were made in MgSO4-ampicillin solution. G. mellonella larvae were stored at 4°C and warmed to room temperature prior to infection. Infections were performed using a 250-μl syringe fitted with a reproducibility adapter (Hamilton, Reno, NV). For each larva, 5 μl of serially diluted bacteria was injected into the hindmost left proleg. This procedure was repeated with 10 larvae for each dilution. Ten control larvae were injected with MgSO4-ampicillin solution. Larvae were incubated at 30°C, and mortality was recorded at 48 h postinfection. This protocol was repeated three times for each strain. Standard deviations were calculated by using Microsoft Excel. To assess the activity of KS9 and KS9c in G. mellonella, phage lysates were first passaged through Pierce Detoxi-Gel endotoxin removal gel (Thermo Scientific, Rockford, IL), and dilutions were made in MgSO4-ampicillin solution. Larvae were injected with 5 μl of a 1:104 dilution of K56-2 into the hindmost left proleg and 5 μl of a 1:100, 1:101, or 1:102 dilution of phage into the second hindmost left proleg. Ten control larvae were injected with K56-2 and MgSO4-ampicillin solution, and ten control larvae were injected with undiluted phage and MgSO4-ampicillin solution. Larvae were incubated at 30°C, and mortality was recorded 48 h postinfection. This protocol was repeated three times for each multiplicity of infection (MOI) tested. To isolate K56-2 from KS9- and KS9c-treated larvae, hemolymph was extracted from larvae surviving 48 h postinfection by using a 20-gauge needle. Serial dilutions were made in MgSO4-ampicillin solution, and cells were plated on half-strength LB medium containing 100 μg of ampicillin/ml.
Seed and Dennis (41) originally isolated KS9 from a culture of B. pyrrocinia LMG 21824. Induction with UV light or mitomycin C was not necessary. When propagated on B. cenocepacia K56-2, KS9 forms small clear plaques, 0.3 to 1.0 mm in diameter. To the best of our knowledge, KS9 is unable to form plaques on B. mallei (10 strains tested), B. pseudomallei (10 strains tested), or B. thailandensis (4 strains tested). Transmission electron microscopy indicates that KS9 has the B1 morphotype of the family Siphoviridae in the order Caudovirales (Fig. (Fig.1)1) (2). The virion is comprised of an icosahedral head (with a diameter of 75 nm) and a long, noncontractile tail (with a length of 250 nm). The phage particles were highly fragile, since a large number of intact capsids were visible in which the tail had broken off close to or at the head/tail adaptor. The KS9 virion is larger than that of both E125 (63-nm head diameter, 203-nm tail length) and 1026b (56-nm head diameter, 200-nm tail length) (47, 17).
To determine whether LPS is involved in KS9 adhesion, K56-2 LPS mutant strains were tested in a plaque assay with high-titer stocks of KS9 and KS5 (a second BCC phage isolated by Seed and Dennis ). Although KS5 produced confluent lysis on strains XOA7 (waaL::pGPΩTp, EOP = 0.8), XOA15 (wabR::pGPΩTp, EOP = 1.3), XOA17 (wabS::pGPΩTp, EOP = 1.1), RSF19 (wbxE::pGPΩTp, EOP = 0.5), and K56-2 (EOP = 1), KS9 was unable to form plaques on any of these strains (29, 35). Both phages are likely to use the LPS as a receptor because neither phage was able to lyse the LPS mutants XOA8 (wabO::pGPΩTp) or CCB1 (waaC::pGPΩTp) (35). These results suggest that KS9 requires a relatively complete LPS structure in order to infect K56-2, likely binding to an LPS component located distal to lipid A, whereas KS5 likely binds to a receptor located deeper within the LPS, proximal to lipid A. These results must be interpreted cautiously because, since the mutations have caused significant deficits in LPS structure, the overall organization of the outer membrane may have been altered as well (29). Although these results are consistent with both KS9 and KS5 using LPS components as receptors, further experiments are required in order to exclude the possibility of adhesion to other outer membrane structures.
The KS9 genome is 39,896 bp (bp) in length and encodes 50 putative protein-coding genes (Table (Table1).1). The G+C content of the genome is 60.7%, which is identical to that of 1026b and slightly lower than that of E125 (61.2%) (17, 47). The majority of the start codons are ATG (42 of 50), with fewer GTG (8) present. Most stop codons are TGA (30 of 50), with equal numbers of both TAA and TAG codons (10 each). Each of the predicted KS9 proteins has some degree of similarity to other proteins, as determined by a BLASTX search, except for gp36, gp38, and gp39. The protein with the lowest detectable similarity to others is gp46, putatively involved in replication, which has 23% identity with a phage O protein of Salmonella enterica subsp. enterica serovar Javiana strain GA_MM04042433. The predicted proteins most similar to database entries are gp13, the tail length tape measure protein, and gp18, a tail component protein. Both of these proteins have 97% identity with proteins encoded by the B. cenocepacia PC184 genome.
The KS9 genome exhibits a modular organization (Fig. (Fig.2).2). Module boundaries were assigned based on BLASTX predictions of gene function, so KS9 genes encoding hypothetical proteins were not grouped as part of a module unless the flanking genes encoded two proteins with similar functions (as is the case for genes 6 and 15). The smallest module, which encodes proteins involved in replication, is made up of two genes, 45 and 46. The DNA packaging/head morphogenesis module is made up of genes 2 to 7. Although the predicted product of gene 6 has not been assigned a putative function, it is included as part of this module because of its location. Similarly, genes 10 to 19 comprise the tail morphogenesis module. Gene 15, encoding a hypothetical protein, has been included here because of its position. The fourth module in the genome, involved in lysis, includes genes 22 to 24′, which encode the putative holin, endolysin, Rz and Rz1 proteins. The sequence has been deposited in GenBank, with accession no. NC_013055.1.
The predicted gene products of KS9 are most similar to those of a prophage from B. cenocepacia PC184 (vB_BcenZ_PC184; Table Table11 and Fig. Fig.3).3). As shown in Table S1 in the supplemental material, the genes encoding gp1-4, gp7-19, gp22-24′, gp26-28, gp45, and gp49 all have similar genes in a single locus in PC184, spanning open reading frames (ORFs) BCPG_00009 to BCPG_00033. An exception is KS9 gene 15, which encodes a protein similar to PC184 BCPG_01060. For instances in which the PC184 proteins are more closely related than predicted proteins from any other source, the percent identities of the proteins range from 60% (gp45) to 97% (gp13 and gp18) (Table (Table1).1). These proteins are involved in DNA packaging/head assembly, tail assembly, lysis, replication, and endonuclease activity. Casjens (12) outlines several standards by which one can predict whether a bacterial genomic sequence contains a prophage. These include (i) the presence of phage-related genes (especially those required for morphogenesis), (ii) continuous organization undisturbed by non-phage related genes, (iii) characteristic prophage gene order, and (iv) the presence of genes encoding hypothetical proteins (especially if phage related). Using the standards put forward by Casjens (12) for identification of prophage sequences, this locus in PC184 appears to contain an uncharacterized prophage or prophage element: it contains multiple phage genes (including morphogenesis genes), the organization is continuous, the genes encoding proteins similar to those in KS9 are present in the same order as the genes in the KS9 prophage, and genes encoding proteins similar to phage-related hypothetical proteins (such as KS9 gp1) are present.
Because of the similarities between KS9, E125, and 1026b, it was predicted that the overlap region of the KS9 attP would be located at a similar position in the chromosome (upstream of the integrase gene) and that it would facilitate integration into a similar locus, likely the 3′ end of a tRNAPro-3 coding region (47, 17). Although the position of the KS9 attP is similar to those of both E125 and 1026b (upstream of the integrase gene), it facilitates recombination into a different host target gene. Using a 20-bp attP overlap region (Fig. (Fig.4),4), KS9 integrates into the 3′ end of a gene encoding the LMG 21824 or K56-2 GTP cyclohydrolase II (GCHII) enzyme. GCHII is responsible for the production of 2,5-diamino-6-β-ribosyl-4(3H)-pyrimidinone 5′-phosphate from GTP. This product is the first intermediate in the synthesis of riboflavin, which is necessary for metabolism (38). The GCHII ORF lies immediately upstream of an ORF encoding a GCN5-related N-acetyltransferase (GNAT). That phage integration does not disrupt the sequence of the predicted GCHII ORF (with the attP overlap region replacing the last 19 bp of the gene, including the stop codon) suggests that there is no loss of GCHII function in a KS9 lysogen.
To our knowledge, there are no other published reports of prophage integration into a bacterial GCHII gene. However, we have identified a sequence in GenBank that strongly suggests that prophage insertion can occur at this site in other BCC genomes. In chromosome 2 of B. cenocepacia AU 1054, ORF Bcen_3636 (encoding GCHII) is located 5′ to ORF Bcen_3637 (encoding GNAT), separated by 125 bp. In contrast, the distance separating these genes in Burkholderia lata 383 (Bcep18194_B1027 GCHII [1154713 to 1155363] and Bcep18194_B1047 [1174320 to 1174886] GNAT) is 18,957 bp. Between these two genes in Burkholderia lata 383, the annotated sequence contains an additional 19 genes including a phage integrase (Bcep18194_B1028), a terminase (Bcep18194_B1042), a lytic transglycosylase (Bcep18194_B1044), and 16 other genes without assigned functions.
In order to characterize this region further, we analyzed the sequence of this region beginning with the GCHII start codon and ending with the GNAT stop codon using GeneMark (30). Proteins similar to the proteins encoded by GeneMark-predicted ORFs were found by using BLASTX. Thirty-six ORFs were found by GeneMark, including the GCHII (ORF 1) and GNAT (ORF 36). According to BLASTX analysis, proteins encoded by 16 of the ORFs predicted by GeneMark are similar to phage-related proteins (see Table S2 in the supplemental material). These include proteins from phages that infect both Gram-positive (such as Lactococcus) and Gram-negative (such as Burkholderia) bacteria. Interestingly, predicted products of four of these ORFs (3, 17, 19, and 32) are similar to E125 and 1026b proteins, respectively: gp33/gp32 (ORF 32), gp36 of 1026b (ORF 3), gp60/gp71 (ORF 17), and gp65/gp76 (ORF 19). Only one of these sets, gp33/gp32, is similar to a gene present in KS9, gp28. These data suggest that an uncharacterized prophage has integrated into Burkholderia lata 383 at the same (or at a similar) site as KS9 and that this phage is at least partially related to KS9. Because of its small size, it is not known whether this prophage is defective or whether it has retained all of the elements necessary for a productive lytic cycle. Again, using Casjens (12) parameters for prophage localization, the region we have identified in Burkholderia lata 383 meets these requirements: (i) genes encoding such phage proteins as integrase, terminase, and virion morphogenesis factors are present; (ii) there are no obvious nonphage genes in the almost 19-kbp sequence; (iii) the gene order is consistent with prophage organization, with the genes encoding the proteins similar to 1026b gp36-gp76 to the 5′ of the gene encoding the protein similar to gp32 (as in the 1026b prophage); and (iv) genes encoding proteins similar to phage-related hypothetical proteins are present.
The predicted DNA packaging/head assembly proteins of KS9 are similar to those encoded by PC184, E125, and 1026b, but the major capsid protein itself is dissimilar (see Table 1 and Table S2 in the supplemental material). Each putative KS9 tail protein has a similar protein encoded by PC184, E125, and 1026b. As discussed above, although KS9 gene 15 is assumed to be part of the tail morphogenesis module because of its location, its predicted product does not have an assigned function and so may not be involved in tail production. In many phage genomes, two proteins are encoded between the major tail protein and tail tape measure genes by way of a −1 translational frameshift (48). It is thought that frameshifting allows phages to control the relative expression levels of two proteins during infection (48). In KS9, these two proteins are gp11, a tail assembly chaperone, and gp12, a minor tail protein. These proteins are predicted to have the same N-terminal sequence, but gp12 (expressed via the frameshift) is predicted to be longer and have a different C terminus. We have found a putative frameshift site between bases 20,347 and 20,353. This site contains a stretch of seven adenine residues that can cause the ribosome to slip backwards by one position, resulting in a −1 frameshift. Although AAAAAAA is not the canonical frameshift sequence (XXXYYYZ, where Y is A or T), it is the same sequence that was identified for phages such as c2 of Lactococcus and PSA of Listeria monocytogenes (48). Providing further support that this is the correct frameshift sequence, we have used the mfold program to identify a potential stem-loop structure formed by 65 downstream bases (50) (Fig. (Fig.5).5). Although such structures are not necessarily found at frameshift sites, their presence suggests a mechanism by which the ribosome may stall and subsequently change its reading frame (48).
KS9 encodes four proteins putatively involved in lysis, all of which are similar to proteins from PC184, but not E125 or 1026b (see Table S1 in the supplemental material). The genes encoding these proteins are part of a single module (Fig. (Fig.2).2). The first gene in the module, gene 22, encodes a putative holin. Class I holins, such as λ S, have three transmembrane domains, whereas class II holins only have two (49). Because OCTOPUS analysis (46) indicates that gp22 has three transmembrane domains, we predict that this protein is a class I holin. Phage lysins may be one of three major types: endo-β-N-acetylglucosaminidases or N-acetylmuramidases, endopeptidases, or N-acetylmuramyl-l-alanine amidases (20). We predict that the KS9 lysin is an endopeptidase: the putative lysin, gp23, is similar to M15A peptidases and is predicted to belong to the Peptidase_M15_3 superfamily (E-value = 2e−23). The KS9 Rz/Rz1 pair is the last gene segment in the lysis module. Rz/Rz1 proteins function in lysis by joining the inner and outer membranes following holin insertion and lysin activity (5). The Rz1 gene, 24′, is located out of frame in the Rz gene, 24. Using the LipoP program (26), it is predicted that a signal peptidase II cleavage site is present between residues 22 (alanine) and 23 (cysteine) of gp24′. Cleavage at this site would produce a 65-amino-acid lipoprotein. Of these 65 amino acids, 7 are proline (10.8%), a value consistent with the proportion of proline found in the processed BcepMu (23.1%), SalMu (15.5%), and PhotoMu (13.4%) Rz1 lipoproteins (44).
Depending on the organism, the proteins similar to KS9 gp49 in the GenBank database have been annotated as either putative class I holins or as HNH homing endonucleases. If gp49 were to act as a class I holin, it would likely function with gp22 as part of a holin-antiholin pair. These systems are used to control when the onset of lysis occurs (28). However, OCTOPUS analysis of the gp49 sequence suggests that it has no transmembrane domains, so we predict that this protein is not a holin and that some other protein encoded by KS9 may be involved in controlling lysis timing. A Conserved Domain Search of the predicted gp49 sequence indicates that it belongs to the HNHc superfamily (E-value = 8e−6) found in HNH homing endonucleases. Homing endonuclease genes (HEGs) are common in phage genomes. For example, T4 encodes 15 homing endonucleases, 6 of which (I-TevIII and MobA-E) belong to this family (34). HEGs are referred to as selfish genetic elements. They use a mechanism called homing to copy their sequence from one place in one gene to the same place in the same gene at a different locus (10). In order to complete this process, endonucleases (such as those in the HNH family) are used to cut the DNA at a specific 15- to 30-bp site, which is then fixed by using a double-strand break repair. Further experiments are required to determine whether gp49 has homing endonuclease activity or whether it performs some other function during infection.
In many bacterial species, prophage genes make significant contributions to the virulence of the organism. Although examples have been documented in a wide variety of Gram-negative organisms (including Vibrio cholerae, Escherichia coli, Pseudomonas aeruginosa, Neisseria meningitidis, Salmonella enterica, and Shigella flexneri), there is limited evidence that prophage genes contribute to virulence in Burkholderia species (8, 43). It has been suggested that because Burkholderia are not strictly pathogenic and can instead survive in a variety of both terrestrial and aquatic environments, classical toxin genes may not have provided a strong evolutionary advantage to these species (43). Instead, lysogenic conversion genes of Burkholderia prophages would be more likely to encode proteins that would increase the viability of the cell both in the environment and in vivo (43).
In the KS9 genome, we identified one gene, gene 32, whose product may have an effect on the pathogenicity of LMG 21824 or a K56-2 KS9 lysogen. When the gp32 sequence was subjected to a Conserved Domain Search, the conserved domain “COG3950: predicted ATP-binding protein involved in virulence” was found (E-value = 6e−16). To determine whether the integration of KS9 as a prophage increased the pathogenicity of B. cenocepacia K56-2, we compared the virulence of wild-type and KS9-lysogenized K56-2 (K56-2::KS9) in the Galleria mellonella wax moth model. In this infection model, the 50% lethal dose (LD50) for K56-2 is 900 CFU (40). For both K56-2 and K56-2::KS9, infection with between 104 and 106 CFU resulted in 100% mortality. There was no significant difference between the mortality of larvae infected with either K56-2 or K56-2::KS9 at any of the doses tested. Although gene 32 contains a domain putatively involved in virulence, these data suggest that it has no effect on the pathogenicity of KS9 lysogens.
The method used to convert KS9 into a lytic phage was to insertionally inactivate the putative KS9 phage repressor, gene 41. It is well established that repressor mutant phages cannot stably lysogenize (45). Platt et al. (37) constructed a strain of E. coli expressing λ cI from the chromosome and lysogenized this strain with a λ cI mutant, W30 (4). When this strain was cocultured with E. coli that was sensitive to λ infection, the numbers of sensitive E. coli present decreased. They proposed that this system could be used to provide continuous delivery of lytic phages in vivo. We wanted to expand on this concept in two ways, first by showing that a lytic mutant could be constructed by using genetic engineering and second by showing that this mutant phage could be active in vivo.
Gene 41 encodes a 133-amino-acid protein. According to BLASTP analysis, this protein shows 55% identity with gp52, the putative repressor protein of E125 (E-value = 4e−30), as well as similarity to cI-like proteins of phages BP-4795, H-19B, SfV, cdtI, and a putative transcriptional repressor of phage V10. A nonreplicating knockout plasmid (pKL1) was constructed containing DNA flanking the KS9 gene 41 interrupted by a trimethoprim resistance cassette in a pALTER-1 (Tcr) backbone. After transformation of this construct into B. pyrrocinia LMG 21824, a Tcr TpR mutant (LMG 21824::pKL1, generated by a single crossover) was isolated.
If a double-crossover event occurs that replaces the repressor gene with a TpR cassette, the cells should no longer be viable because the prophage will be induced in the absence of the repressor protein, resulting in lysis. We used two different assays to assess if this induction was occurring in LMG 21824::pKL1. First, we screened 1800 Tcr TpR LMG 21824::pKL1 colonies for loss of tetracycline resistance and maintenance of trimethoprim resistance (i.e., a double crossover without loss of viability). No Tcs TpR colonies were found, suggesting that the double-crossover event was lethal due to prophage induction. Second, we compared the release of K56-2-infecting phages from LMG 21824, LMG 21824::pKL1, and LMG 21824 with a mutation in KS9 gene 32 (LMG 21824 [KS9 32]). The number of phages released from LMG 21824::pKL1 was more than 10 times greater than the number released from the other two strains (Fig. (Fig.6),6), a finding consistent with phage induction occurring in LMG 21824::pKL1 following a double-crossover event. These two experiments provide further evidence that gene 41 is the repressor, since it is required to maintain stable integration of the KS9 prophage.
Phages isolated from culture supernatants of LMG 21824::pKL1 were screened for the presence of the TpR cassette. Using primers flanking gene 41, a PCR product of 603 bp was amplified with wild-type KS9, whereas a product of 832 bp was amplified from the mutated repressor phage with the TpR cassette. A representative phage isolate that tested positive during PCR screening was named KS9c. This isolate showed no defects in activity and was able to produce confluent lysis of B. cenocepacia K56-2 in an agar overlay at titers of 107/ml, similar to that of wild-type KS9. The host range of KS9c was tested, and it was found to be identical to that of wild-type KS9, suggesting that there was no change in the susceptibility of Bcc strains to this phage.
To assess whether KS9c is able to stably integrate into K56-2, plates of K56-2 and KS9 or K56-2 and KS9c exhibiting confluent lysis were overlaid with water, and the surviving cells were isolated. Then, 150 bacterial isolates from KS9c lysates were replica plated onto solid medium with or without trimethoprim. These isolates were unable to grow on trimethoprim medium, suggesting that KS9c had not stably integrated into any of these bacteria. A total of 50 KS9-insensitive and 50 KS9C-insensitive K56-2 isolates were recovered from the phage lysates and were assayed for their (i) growth phenotype, (ii) trimethoprim resistance, (iii) susceptibility to KS9 and KS9c, and (iv) ability to amplify with primers designed to the K56-2/5′ KS9 prophage junction. The majority of KS9-insensitive isolates (41 isolates, 82%) were not lysogenized. These strains grew normally, were TpS, KS9RKS9cR and were PCR negative for KS9 integration. Eighteen percent (9 isolates) were stably lysogenized. These strains grew normally, were TpS KS9R KS9cR, and were PCR positive. A different distribution was observed for the KS9c-insensitive isolates. In this case, all 50 isolates (100%) were nonlysogenized. Like the nonlysogenized KS9-insensitive isolates, these strains grew normally, were TpS KS9R KS9cR, and were PCR negative for KS9 integration. Although almost one-fifth of KS9-insensitive isolates were stably lysogenized, none of the KS9c-insensitive isolates showed this phenotype. Therefore, we have shown that inactivating prophage KS9 gene 41 results in the loss of lysogen viability and increased prophage induction from LMG 21824 and prevention of stable lysogeny after K56-2 infection. Consequently, we conclude that gene 41 encodes the KS9 repressor protein and that KS9 is converted from a temperate to a lytic phage following the loss of this gene.
We chose to test the efficacy of KS9c in an in vivo alternative infection model. Seed and Dennis (39) showed that G. mellonella larvae infected with twice the lethal dose of K56-2 could be rescued after administration of phage KS12 at an MOI of 5,000. We tested the efficacy of treatment with endotoxin-free KS9c lysates in K56-2-infected larvae at MOIs between 0.5 and 100. Treatment with KS9c at MOIs of 50 or 100 increased the survival of infected larvae to that of uninfected controls (Fig. (Fig.7).7). This result is in sharp contrast to untreated infected larvae that had survival rates near zero. MOIs lower than 50 were ineffective, since the survival of larvae after treatment was comparable to that of untreated infected controls. In contrast to larvae treated with higher MOIs, those treated with MOIs lower than 50 that remained alive 48 h postinfection exhibited both reduced movement and signs of melanization, suggesting that the K56-2 infection had progressed but was not yet fatal. As a result, we conclude that phage treatment had little or no effect at the lower MOIs. This result is expected, since higher MOIs of lytic phages are generally more effective for therapeutic purposes (6). Our results indicate that KS9c is active in an in vivo model and can effectively treat experimental B. cenocepacia K56-2 infections when administered at an MOI of 50 or greater.
Because KS9c is unable to stably lysogenize K56-2 and facilitate superinfection immunity, we predicted that KS9c might be a more effective therapeutic agent than KS9. We treated K56-2-infected larvae with either KS9 or KS9c at MOIs of 50, 5, and 0.5 and compared their ability to increase larval survival. In contrast to our prediction, we found no significant differences between the survival of larvae treated with the two different phages (Fig. (Fig.8).8). In order to assess why there was no difference observed between the two treatments, we isolated K56-2 from the hemolymph of infected, KS9-treated larvae and screened these isolates for lysogeny. In contrast to the in vitro screening presented above where nearly 20% of insensitive isolates were stably lysogenized, 0 of 50 KS9R KS9cR K56-2 in vivo isolates analyzed were lysogenized based on PCR screening. Because lysogeny does not appear to make a significant contribution to KS9 resistance in the infected larvae at the MOI tested, this result explains why there was no significant difference in the in vivo efficacy of KS9 and KS9c. Different phage MOIs most likely caused the discrepancy observed in the proportion of lysogenized bacterial isolates obtained in vitro and in vivo. MOIs were initially higher for the in vivo screening (100 in vivo versus 0.1 in vitro), which would tend to favor lysogeny. However, rapid infection, lysis and release of progeny phage would be predicted to increase the effective MOI in vitro more so than in vivo, leading to a greater proportion of lysogenized cells in the lysate than in the larvae (36).
In summary, by knocking out the putative KS9 repressor gene, we have created a functionally lytic phage named KS9c. This phage does not stably lysogenize B. cenocepacia K56-2 and it is effective in treating experimental infections in the G. mellonella model. Although it is unexpected that this phage is only as effective as the wild-type phage in this model (as opposed to more effective), we suggest that this result is due to the low level of lysogeny observed at the MOIs tested in vivo.
We have determined the genome sequence of the siphovirus KS9, a prophage of B. pyrrocinia LMG 21824. The KS9 prophage integrates into both LMG 21824 and B. cenocepacia K56-2 in the 3′ end of the GTP cyclohydrolase II gene. It is 39,896 bp in length and encodes 50 genes, many of which show similarity to genes of the Burkholderia prophages E125 and 1026b and an uncharacterized prophage element of B. cenocepacia PC184. Although one gene was identified with a putative role in pathogenicity, KS9 integration was shown to have no effect on the virulence of B. cenocepacia K56-2 in the G. mellonella model. We were able to show that gp41 encodes the phage lysogenic repressor by developing a lytic mutant of KS9, named KS9c. Unlike KS9, KS9c was unable to stably lysogenize K56-2. Treatment of K56-2-infected G. mellonella larvae with KS9c at MOIs of 50 or greater resulted in larva survival comparable to uninfected controls, indicating that KS9c is an effective antibacterial agent in vivo. As a proof-of-principle, we have shown that temperate phages can be genetically engineered to lytic form and that these engineered phages are active in vivo.
This study was funded by an operating research grant to J.J.D. from the Canadian Cystic Fibrosis Foundation and funds from the Canadian Institutes of Health Research for the “CIHR Team on Aerosol Phage Therapy.” K.H.L. and K.D.S. are indebted to the Natural Sciences and Engineering Research Council of Canada for CGS-D and PGS-D scholarships, respectively.
We thank David DeShazer (U.S. Army Medical Research Institute for Infectious Diseases) for testing the B. pseudomallei and B. mallei host ranges and Miguel Valvano (University of Western Ontario) for providing the B. cenocepacia LPS mutants. We also thank Randy Mandryk, University of Alberta Microscopy Service Unit, and Pat Murray and Lisa Ostafichuk, University of Alberta Molecular Biology Service Unit, for their assistance.
Published ahead of print on 25 November 2009.
†Supplemental material for this article may be found at http://jvi.asm.org/.