Genetically modified filamentous phage as bactericidal agents: a pilot study


Udo Bla¨si, Department of Microbiology and Genetics, Max F. Perutz Laboratories, University Departments at the Vienna Biocenter, Dr Bohrgasse 9, 1030 Vienna, Austria (email:


Aims: To evaluate the ability of a filamentous phage encoding lethal proteins to kill bacteria without host-cell lysis.

Methods and Results: Bacterial survival was determined after infection of a growing Escherichia coli culture with phage M13 encoding either the restriction endonuclease BglII gene or modified phage λS holin genes. The genetically engineered phage exerted a high killing efficiency while leaving the cells structurally intact. When compared with a lytic phage, the release of endotoxin was minimized after infection with the genetically modified phages.

Conclusions: Genetically engineered phage can be used for efficient killing, concomitantly minimizing endotoxin release.

Significance and Impact of the Study: This feasibility study provides a possible strategy for the use of genetically engineered phage as bactericidal agents by optimizing the advantages and minimizing potential risks such as release of pyrogenic cell wall components.


The emergence of multi-drug resistant pathogens like Staphylococcus aureus has sparked renewed interest in the use of bacteriophage for treatment of bacterial infections. Historically, research on phage therapy was abandoned in the western countries after the advent of antibiotics, whereas in some eastern countries, phage therapy was further scrutinized and applied. More recently, a number of animal studies have been conducted, all concluding that phage therapy is highly effective in treating bacterial infections (Smith and Huggins 1983; Berchieri et al. 1991; Soothill 1992; Biswas et al. 2002). Single doses of phage also proved to be more effective than repeated antibiotic treatments in several cases (Smith and Huggins 1982; Levin and Bull 1996), which has been explained by the exponential increase of phage during therapy. Transducing phage can be responsible for increased virulence of their host bacteria as in the case of the Shiga-like toxin-converting phage in enterohaemorrhagic Escherichia coli (O'Brian et al. 1989), necessitating the use of strictly lytic therapeutic phage. Rapid lysis of bacteria and release of large amounts of endotoxin are the consequence, which can lead to undesired side-effects of phage therapy even when the phages are administered orally (Slopek et al. 1981). This may be explained by the observation that orally administered therapeutic phage were readily found in blood (Weber-Dabrowska et al. 1987). Similarly, many β-lactam antibiotics lead to markedly increased levels of endotoxin when used for treatment of systemic Gram-negative infections, which can elicit circulatory shock in some patients, and there is increasing evidence that antibiotic-mediated release of endotoxins may lead to clinical deterioration (Lepper et al. 2002). Hence, there is a need for antimicrobial agents which do not cause this side-effect.

A large number of phage of both Gram-negative and Gram-positive bacteria cause host lysis by employing a dual holin/endolysin system. At the end of the lytic cycle, holins induce a lesion in the cytoplasmic membrane by oligomerization, allowing the endolysin access to its peptidoglycan substrate. Dissipation of the membrane potential by the action of holins is lethal, whereas the release of endolysin, and consequently the degradation of the peptidoglycan causes cellular destruction and the release of cell wall components (reviewed by Young and Bläsi 1995). The phage λS-holin gene has a dual-start motif, which results in the production of two S-polypeptides differing by only two amino acid residues at their N-termini (Bläsi et al. 1989). The two S proteins are 105 (S105) and 107 (S107) amino acids in length and have opposing roles in lysis with S105 being the effector and S107 being the inhibitor for the formation of the membrane lesion (Bläsi and Young 1996). Many mutants of the λS gene with an enhanced killing potential exist (Young 2002). Likewise, a fusion protein (Pf3VIIIΦS105) consisting of the N-terminal sequence of the Pseudomonas aeruginosa phage Pf3 major coat protein pVIII fused to the N-terminal amino acid of S105 was shown to be highly effective in cell killing (Graschopf and Bläsi 1999).

M13 is a F-pilus specific, single-stranded DNA phage of E. coli, which does not encode lysis genes. M13 progeny are extruded from the host without causing disintegration of the cell wall (Russel 1991). In this study genetically modified M13 phage encoding either the λ S105 protein or the Pf3VIIIΦS105 fusion protein were used to test whether these phage can be used for efficient killing of E. coli.

Restriction modification (RM) systems are ubiquitously employed by bacteria to degrade heterologous DNA encountered in the cell. Type II RM systems are composed of a methyltransferase and a restriction endonuclease. The DNA modifying enzyme adds methyl groups to certain bases in a particular DNA sequence, which protects cellular DNA from cleavage by the corresponding restriction endonuclease. Here, we have inserted the BglII R-gene (Anton et al. 1997) into the genome of phage M13 with the reasoning that infection with the resulting M13R phage, and thus synthesis of the phage encoded BglII restriction endonuclease will lead to nonrepairable breaks in double-stranded chromosomal DNA which in turn will result in efficient killing of E. coli.

The killing efficiency of the two M13-holin phages was compared with that of the M13R phage, and it was assessed whether the genetically modified phage differ from a lytic phage with regard to endotoxin release upon host cell killing.

Material and methods

Bacterial strains, plasmids, phage, media and growth conditions

The bacterial strains, plasmids and bacteriophage used in this study are listed in Table 1. Bacteria were grown in Luria-Bertani (LB) medium (Sambrook et al. 1989) at 37°C. Where appropriate, the medium was supplemented with 100 μg ml−1 ampicillin, 30 μg ml−1 tetracycline and 0·003 mol l−1 isopropyl-β-d-thiogalactopyranoside (IPTG), which binds to and inactivates the lac repressor.

Table 1.  Strains, plasmids and phage
 FeaturesReference or source
 MC4100araD139 Δ(argFlac)U169 thiA rpsL150 relA1 deoC1 ptsF25 rbsR flbB5301Silhavy et al. (1984)
 MC4100F′MC4100FproAB lacIqlacZΔM15 Tn10Steiner and Bläsi (1993)
 pMBN2Carries the BglII methylase geneAnton et al. (1997)
 pMRB1Carries the BglII endonuclease/methylase gene cassetteAnton et al. (1997)
 λcILytic λ phageLaboratory stock
 QβMale-specific RNA phageMiyake et al. (1966)
 M13mp18Cloning vector, lacpo, multiple cloning siteYanisch-Perron et al. (1985)
 M13S105Carries the λS105 allele under transcriptional control of the lacpoGraschopf and Bläsi (1999)
 M13VIIIS105Carries the VIII-S fusion gene under transcriptional control of the lacpoThis study
 M13RCarries the BglII endonuclease gene under transcriptional control of the lacpoThis study

M13 variants bearing λS-holin mutants and construction of the M13R phage harbouring the BglII R-gene

The construction of phage M13S105 harbouring the λS105 allele has been described (Bläsi et al. 1989). Phage M13VIIIS105 encoding the Pf3VIIIΦ105 fusion protein was constructed by cloning the BamHI/EcoRV fragment of plasmid pPf3-S105 (Graschopf and Bläsi 1999) into the BamHI and HincII sites of the cloning vector M13mp18 (Yanisch-Perron et al. 1985). The recombinant M13VIIIS105 phage was isolated as described below for the M13R phage. The M13S105 and the M13VIIIS105 phage were propagated on strain MC4100F′ harbouring the lacIq allele to ensure transcriptional repression of the lacpo controlled holin genes.

To clone the BglII R-gene in M13, plasmid pMRB1 (Anton et al. 1997) comprising the BglII RM gene region was used as a template together with the oligonucleotides U14 (5′-AAAAAAAAAAGGATCCAAATTAGACCGCAC-3′) and F14 (5′-TTTTTTTAAGCTTTCTAGATTTAATATGTCACGATTGTTCCTCTTTTCCGACGTCTGG-3′) in a polymerase chain reaction (PCR). Oligonucleotide U14 introduced a BamHI site 5′ of the ribosome binding site of the BglII R-gene and a HindIII site 3′ of its stop codon. The resulting PCR fragment was cleaved with BamHI and HindIII, and the fragment was ligated into the corresponding restriction sites of M13mp18 (Yanisch-Perron et al. 1985). The ligation mixture was transformed into E. coli strain MC4100F′ carrying the BglII methylase encoding plasmid pMBN2 (Anton et al. 1997). The recombinant M13R phage was isolated by PCR screening of pseudo-plaques using primers U14 and F14, and by subsequent restriction analysis of M13R RF-DNA. In the resulting M13R phage, the BglII R-gene is under transcriptional control of the lac promoter/operator. The M13R phage was propagated on strain MC4100F′ (pMBN2) expressing the BglII M-gene to ensure ‘immunity’ against the phage encoded BglII restriction endonuclease. High titre phage stocks were obtained from the supernatant by centrifugation of the respective bacterial cultures at 10 000 gin a table-top centrifuge and by subsequent 0·45 μm sterile filtration.

Bacterial growth and survival

Colony forming units (CFUs) and plaque forming units (PFUs) were determined by duplicate plating of serial dilutions of cultures and lysates. For determination of the killing efficiency of the M13S105 and the M13VIIIS105 phage, E. coli strain MC4100F′ was grown to an O.D.600 of 0·2 in LB medium, containing 0·003 mol l−1 IPTG, and then infected with the respective phages at a different multiplicity of infection (MOI). The killing efficiency of the M13R phage was likewise determined. Total cell counts were determined using a Thoma-chamber at a 1000-fold magnification.

Endotoxin assay

Endotoxin units were determined with the QCL-1000 test kit from Bio Whittaker (Rockland, ME, USA). Samples (1 ml) of noninfected bacteria and cultures infected with an MOI of 10 of either M13R, M13S105 or λcI were withdrawn at an O.D.600 of 0·2 at the time of infection, and 1, 2 and 4 h after infection. After 3 min centrifugation at 12 000 rpm the supernatant was stored at −20°C. The samples were diluted until a linear correlation (≙ a range of 0·1–1 endotoxin units per millilitre) between the absorbance at 405 nm and the endotoxin concentration was reached. The endotoxin units in the samples were calculated from the absorbance values of solutions containing known amounts of endotoxin standard.


Effect of the M13 holin phages on survival of host bacteria

In the first experiments the killing efficiency of phages expressing the S105 and the VIIIS105 holin gene was tested. In both cases a rapid decline in cell viability was observed within the first 2 h upon infection of E. coli strain MC4100F′ with an MOI of 10 of either M13S105 or M13VIIIS105 (Fig. 1). More than 99% of the bacteria were killed within this time. Variations in the MOI between 1 and 100 had little effect on the killing efficiency, the only noticeable difference at very low MOIs was a short delay in onset of killing (data not shown). Growth resumed between 120 and 180 min after infection, albeit slower than growth of cultures infected with the parental M13mp18 phage. The rapid resumption of growth upon phage infection indicated that phage-resistant mutants were likely to be present at the time of infection. We tested whether resistance was because of altered or lost pili, which are required for adsorption of M13 to the host (Russel 1991). Fifty single colonies isolated from bacteria after overnight incubation in the presence of the M13S105 phage were tested for their susceptibility to phage M13mp18 or phage Qβ, a male-specific RNA phage. None of these colonies could be infected by either phage. It should also be noted that none of the overnight survivors produced phage as determined by plating of the respective culture supernatants on susceptible hosts. As another means to test for altered or lost pili, 20 of these phage-resistant mutants were analysed for their ability to transduce the F-plasmid. The F-plasmid could be transduced to recipient cells (data not shown), indicating that the pili had probably been modified but not lost.

Figure 1.

Survival of Escherichia coli MC4100F′ after infection with M13S105 in the absence of isopropyl-β-d-thiogalactopyranoside (IPTG) (bsl00046) and with either M13S105 (bsl00001) or M13VIIIS105 (bsl00066) in the presence of 0·003 mol l−1 IPTG. A multiplicity of infection of 10 was used in all experiments

Killing efficiency of the M13R phage

It has been reported that M13 encoding the restriction enzyme EcoRI infects and kills host bacteria, forming clear plaques compared with the turbid growth inhibition zones (pseudo-plaques) of filamentous phage (Heitmann et al. 1989). The M13R phage also formed small, clear plaques on E. coli MC4100F′ lawns in the presence of IPTG, which was added to induce transcription of the lacpo controlled BglII R-gene. Infection of a growing culture of E. coli MC4100F′ with M13R (MOI = 10) led to a rapid decline in the CFU (Fig. 2). When compared with M13S105 or M13VIIIS105 (Fig. 1), the killing efficiency was increased in a manner comparable with the performance of the lytic λcI phage (Fig. 2). In contrast to the culture infected with λcI the optical density of the culture infected with M13R was found to vary little over an 8-h period (Fig. 2). As assessed in a Thoma chamber, the number of cells in the M13R infected culture remained almost constant over a 6-h period although 99·9% of the cells were no longer viable. In contrast, the number of structurally intact cells mirrored the number of viable cells in the culture infected with phage λcI (Fig. 2). As observed with the M13S105 and M13VIIIS105 phage, growth resumed 180 and 300 min upon infection with phage M13R and λcI respectively.

Figure 2.

Survival of Escherichia coli MC4100 infected with λcI and E. coli MC4100F infected with M13R at a multiplicity of infection of 10. The number of viable cells upon infection with M13R (bsl00001) and λcI (bsl00000) was measured by determining the CFU of serial dilutions on agar plates. The total cell count upon infection with M13R(bsl00046) and λcI (bsl00067) was determined in a Thoma-chamber. The O.D.600 upon infection with M13R (bsl00066) and λcI (bsl00084) is also depicted

Endotoxin release after M13S105, M13R and λcI−−-mediated cell killing

The endotoxin levels in the supernatant of a noninfected culture and in that of cultures infected with M13S105, M13R or λcI with a respective MOI of 10 were measured for 4 h. When compared with the endotoxin levels present at the time of infection (O.D.600 = 0·2), the endotoxin levels in the growing noninfected and M13mp18 infected cultures increased three- and six fold, respectively, over this period (data not shown). This showed that replication of the filamentous phage leads to some release of endotoxin, which presumably is because of a nonspecific deterioration of cells. The effects of the M13S105 and M13R phage on endotoxin release were compared with that of the λcI phage (Fig. 3). When compared with the initial amount present in the culture supernatant at t = 0, the endotoxin levels increased with the lytic λcI phage 18-fold, 1 h upon infection, whereas only a two fold increase was observed at that time upon infection with either M13S105 phage or M13R. After 4 h, a 27-fold increase was observed in the λcI infected culture when compared with a six- and seven fold increase in the supernatant of cultures infected with M13S105 and M13R, respectively.

Figure 3.

Increase of endotoxin levels in the culture supernatants upon infection of Escherichia coli strain MC4100 with phage λcI and strain MC4100F upon infection with phage M13S105 or phage M13R. The endotoxin level in the supernatant of the E. coli cultures at the time of infection (≙8.3 × 106 endotoxin units) with the corresponding phages was set to 1 (bsl00001). The relative increase in the endotoxin levels upon infection with the different phages were determined from 1 ml samples taken 60 min (░) 120 min (▒) and 240 min (▓) after infection and normalized to the endotoxin level at t = 0. The endotoxin units were determined in triplicate. The error bars represent s.d.


In this report we have shown that filamentous phage, encoding lethal but nonlytic proteins can efficiently kill host bacteria, concomitantly minimizing endotoxin release when compared with a strictly lytic phage. Therefore, it seems feasible to use such modified phage to minimize the risk of toxic shock during phage therapy. Although it is rather unlikely that the restriction endonuclease exerts a higher lethality than holins, the M13R phage was found to be more efficient (one order of magnitude) in host cell killing, when compared with both M13-holin phages. This may be explained by the different means employed to confer protection to the lethal proteins in the phage-propagation strains. The M13-holin phages, either bearing the genes encoding the lethal holin variants under transcriptional control of the lacpo, were propagated on a strain producing an enhanced level of the lac repressor. Although this was performed to ensure optimal repression of the lac promoter, a complete repression of the holin genes was apparently not achieved as obvious from the following observation. After 8 h of incubation, the CFU of cultures of MC4100F′ infected with M13S105 was three fold lower when compared with MC4100F′ infected with M13mp18. Low-level leaky expression of the holin genes could exert selective pressure towards M13-holin mutants comprising a defect in either holin gene expression or its function, or both (Raab et al. 1986). Thus, the lysate derived from the propagation strain is likely to contain a substantial number of phage mutants which are unable to kill their host.

The BglII restriction endonuclease gene was likewise placed under transcriptional control of the lacpo and the M13R phage was propagated in a lac repressor overproducing strain. However, this propagation strain contained in addition a plasmid born BglII methylase gene, the product of which can compensate for a leaky BglII synthesis, which in turn should avoid any selective pressure towards the cloned R-gene. In other words, the lysate obtained from the propagation strain should contain a homogeneous, killing-proficient M13R phage population.

The use of genetically modified phage for therapeutic purposes may raise some safety concerns. To meet safety standards, it is conceivable to replace an essential phage gene by a ‘kill gene’, e.g. a restriction endonuclease gene, rendering the recombinant therapeutic phage nonreplicative in but lethal for the target bacterium. Using good laboratory praxis, high titres of such a recombinant modified phage can be obtained by supplying the deleted phage gene in trans in a methylase-producing propagation strain.

Another concern in phage therapy is that phage-resistant bacterial mutants are present at the time of phage application, as observed in the experiments shown in Figs 1 and 2, or that they arise during treatment, leaving a substantial fraction of the pathogenic bacteria immune to treatment with this phage. Although the decrease in CFU observed with the M13 holin or the M13R phages could be well sufficient to allow the immune system to counteract the infection, it is likewise feasible to use phage that target bacterial receptors which are either essential or function as virulence factors. This could provide a means to decrease the virulence of phage-resistant mutants. Filamentous phage attach to pili, and most likely, phage-resistant mutants have either modified pili or even lost their pili (Bradley 1973; Johnson and Lory 1987). As bacteria resistant to the M13S105 phage were as well resistant to phage Qβ, we assume that these survivors indeed possess altered pili. For instance, pili are a major virulence factor in P. aeruginosa required for adhesion to mammalian cells (Hahn 1997). P. aeruginosa cells that have altered pili have been shown to be attenuated in virulence (Farinha et al. 1994).


We are grateful to Dr B. Anton for providing plasmids pMBN2 and pMRB1. This work was supported by a grant (GZ 70048) from the Austrian Ministry of Education, Science and Culture to U.B.