Since not much is known about the presence and carriage rate of bacteriophages in P. acnes, we investigated this in relation to both the site of isolation (superficial or deep infections) and to biotype. To investigate if P. acnes had prophages that could be induced to enter the lytic life cycle, we stimulated 92 different P. acnes isolates (see table 1) with 2 μg/ml mitomycin C to induce prophages, followed by analysis of the plaque forming capacity of lysates on the noninduced parental isolates. Plaques were clear with well-defined edges and had a diameter of 6–7 mm. Bacteriophages could be induced in more than 70% of the isolates examined. In this study, we have used different isolates of P. acnes (Holmberg et al, unpublished) from deep tissue (AD-isolates, mainly isolated from infections of foreign material as hip prosthesis and sternal wires), and from the skin (AS-isolates, from the skin of healthy individuals). AD-isolates and AS-isolates had a carriage-rate of 70.5% respectively 70.8% (see Figure 1A).

Bacterial isolates that did not have any inducible phages using mitomycin C were screened for prophages using a PCR-based approach by amplification of the gene encoding a putative major head protein and recA as a positive control. Only one of the isolates, AS14, was positive in the major head PCR, indicating that only 1/9 of these isolates has phages with similarities to the known major head gene (data not shown).

The carriage rate of inducible temperate phages in the different biotypes was also examined. Biotypes had previously been determined by sequencing of recA (Holmberg et al, unpublished). Biotype IA had a higher carriage rate than IB (p < 0.01), as did biotype II (p < 0.05). Since biotype IB had a lower carriage rate compared to the other two biotypes, we compared the carriage rate of phages in the isolates determined as biotype IB between deep and superficial isolates of P. acnes. The carriage rate in biotype IB was significantly higher in superficial isolates as compared with isolates from deep infections, since none of the deep tissue isolates biotyped as IB had inducible phages (p < 0.05, see Figure 1B).

The classification of bacteriophages is mainly based on phage morphology and the nature of the nucleic acid 30. Though other classification systems such as sequence similarities within genes encoding structural proteins have been proposed, classification based on morphology and the nature of the nucleic acid is still the most accepted system 31. Forty-nine of the bacteriophages were examined using negative staining and transmission electron microscopy. All examined phages have an icosahedral head of approximately 55 nm in diameter, and a tail composed of 33 segments with a total length of 145–155 nm and a width of 9–10 nm. The tail is non-contractile and appears flexible. Most phages have a visible base plate on the tail with attached spikes (see Figure 2, 3). These morphological attributes warrant classification of the phages as Siphoviridae. Thus, this morphology is identical to PA6 29 and very similar to the P. acnes phages studied by Zierdt 32. Also, the phages are very similar morphologically to Siphoviruses isolated from other propionibacteria 242527. This classification was further strengthen by amino acid sequence comparison of a part of a putative major head protein with other known phage proteins, using a BLASTp search against GenBank. The best hit was on gp6 from PA6, and the second and third best hit on gp7 from Mycobacterium phage Che9d and gp7 from Mycobacterium phage Halo. All these phages are classified as Siphoviridae.

Bacteriophages are generally quite restricted in their host range. There are phages that can infect over bacterial species boundaries, but some phages are species specific and in many cases also specific for certain subgroups (subspecies, serotypes or biotypes, strains) within a species. To determine how specific the isolated phages are, different P. acnes isolates were infected with 48 different phages (see additional file 1). Bacterial isolates with inducible phages were generally easy to lyse, with plaque production in nearly 99% (2278/2304) of cases. P. acnes isolate AS12 was significantly more difficult to lyse than other isolates with inducible phages, and only 73% (35/48) of the examined phages caused plaque formation. Similarly, phage PAD8 failed to lyse 19% (9/48) of the examined isolates. In P. acnes isolates without any inducible phages, we could only observe plaque formation in 30% (144/480) of the cases. We also examined the ability of phages to cause lysis in the recently sequenced P. acnes strain KPA171202 (DSM no. 16379) of biotype IB 9. This strain was lysed only by 40% (19/48) of the phage lysates. The different phages are not specific to certain biotypes, but seem to be less lytic against biotype II in isolates with inducible phages, since 23/26 failures to infect and lyse the bacterial isolates were in isolates of biotype II.

Since all of the examined phages have identical morphology, protein pattern and in most cases very high similarity in genes encoding the putative major head protein and an amidase (see Figure 4), we examined if phage host-range could be used as a possible tool to differentiate the phages. Based on the host-range analysis we choose P. acnes isolates KPA171202, AD7, AS1 and AS5 to differentiate the phages and to divide them into different groups. By using these four isolates, we could divide the phages into 9 separate groups (see Fig 5). Sixteen of the examined phages could infect and lyse all four bacterial isolates (host-range group PA I). There is a tendency that phages isolated from biotype II and from skin can infect and lyse all four bacterial isolates, and are classified as host-range group PA I. Also, the only phages that can lyse the prophage-free isolates AD26 and AD27 (PAD21, PAS7 and PAS11) are classified as belonging to the host-range group PA I. Furthermore, none of nine selected phages (PAS2, PAS10, PAS12, PAS40, PAS50, PAD9, PAD20, PAD21 and PAD42) were able to infect and lyse P. avidum, P. granulosum or P. freudenreichii (data not shown). Our results show that isolates lacking inducible temperate phages are more difficult to lyse using phages, than isolates carrying inducible temperate phages and that the phages are specific to P. acnes.

A part of a putative major head gene was sequenced in all isolated phages. The nucleotide sequence was aligned and a phylogenetic tree was reconstructed. The phages could be divided in two distinctly divided groups with the already sequenced phage PA6 forming a third group (see additional file 2). The whole gene encoding a putative major head protein was sequenced in nine of these phages, selected based on their partial sequencing and site of isolation (AD/AS), and the amino acid sequences were aligned and a phylogenetic tree was reconstructed (see Figure 4). The P. acnes phage putative major head protein show high similarity to phages isolated from Mycobacterium, Lactococcus and Streptococcus. The already sequenced phage PA6 forms its own group, while PAD9, PAD42, PAS10, PAS12, PAS40 and PAD20 form a large group with similar sequences, even though PAD20 is slightly different. Phages PAS2 and PAS50 form a third group and PAD21 forms a fourth separated group. This pattern is similar to the pattern seen when only using partial sequencing, except that PAD21 now seems to form its own group.

We further sequenced the gene encoding a putative amidase in these nine phages (see Figure 4C). The phage pattern is similar to what could be seen when aligning the major head protein. However, PA6 is more closely related to PAS2 and PAS50 in this protein, and PAD21 do not form its own group, but have high similarity to the large group of phages. The closest phage similarity is to Mycobacterium phage PG1 gp49, but more closely related is a chromosomally encoded P. acnes amidase.

A total of 92 P. acnes isolates were used. These were divided into two groups: 44 isolates from deep infections (AD-isolates, mainly isolated from prosthesis and sternal wires) and 48 isolates from skin from healthy individuals (AS-strain), described in Holmberg et al, unpublished. All isolates from healthy individuals and AD-isolates 1–16 and 37 are isolated in Lund, Sweden. AD-isolates 17–32 and 38–48 are isolated in Örebro, Sweden, and AD-isolate 33–36 are isolated in Malmö, Sweden. The P. acnes strain KPA171202 (DSM no. 16379), P. avidum (DSM no. 4901), P. granulosum (DSM no. 20458) and P. freudenreichii subsp. freudenreichii (DSM no. 20271) were obtained from DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Braunschweig, Germany).

P. acnes isolates were plated from frozen stocks on Tryptic Soy Broth (TSB, Bacto, Mt Pritchard, NSW, Australia) with 1.5% Agar (TSBA, Bacto), and incubated for two days at 37°C under anaerobic conditions. Isolates were inoculated in 10 ml prereduced TSB (rTSB) followed by incubation for three days. TSB was prereduced by 24 h incubation under anaerobic conditions. The cultures were diluted 1:9 in 1 ml rTSB and incubation continued for 8 h. Mitomycin C (Calbiochem EMD Biosciences, San Diego, CA, USA) was added to a final concentration of 2 μg/ml, and incubation continued overnight. Cultures were centrifugated (10 min, 1,500 g, Eppendorf Centrifuge 5415R) and sterile filtered (Millex-GP 33 mm 0.2 μm, Millipore, Billerica, MA, USA) to obtain a phage stock. The phage stock was stored at 4°C for 7 days to clear the stock from unstable phages 32. Phages were spotted onto TSBA plates at different concentrations (1:1 – 1:10,000) after a bacterial overlay of the host P. acnes isolate had been prepared. Plates were incubated for two days at 37°C under anaerobic conditions and examined for plaques. When single plaques were observed, these were picked with a sterile scalpel and transferred to SM-buffer (20 mM Tris-HCl pH 7.5, 100 mM NaCl, 60 mM MgSO₄(7H₂O), 0.01% gelatine) followed by elution overnight at 4°C. Phages were propagated by spotting phages on a TSBA plate with a bacterial overlay of its host P. acnes isolate and incubated three days at 37°C under anaerobic conditions. The overlay containing all plaques were transferred to SM-buffer, eluted overnight at 4°C, followed by sterile filtration of the SM-buffer. This method generally generated a concentration of 10¹¹–10¹³ pfu/ml. Phages were named after the host bacterium with a 'P' before the bacterial isolate name. Isolates were regarded as having inducible phages if infectious phages were induced using the method described above. All bacterial isolates that did not have any inducible phages were screened for prophages using a PCR-based amplification of the major head gene (primer pair MHF/MHR) and using recA amplification as a positive control.

Forty-eight P. acnes isolates that carried inducible phages and 10 P. acnes isolates that did not carry any inducible phages using mitomycin C were plated on TSBA as an overlay assay. To each plate, 5 μl phage stock (10¹⁰ pfu/ml) from 48 different phages was added. Plates were incubated for 2 days under anaerobic conditions and examined for plaques. Nine phages (PAS2, PAS10, PAS12, PAS40, PAS50, PAD9, PAD20, PAD21 and PAD42) were also applied to P. avidum, P, granulosum and P. freudenreichii subsp. freudenreichii as an overlay assay.

Bacteriophages in a stock concentration of 10¹¹–10¹³ pfu/ml were placed on a carbon coated copper grid and negatively stained (for references see 35). A 0.75% uranyl formate solution was obtained by dissolving 37.5 mg uranyl formate (BDH Chemicals Ltd., Poole, UK) in 5 ml boiling water, and stabilized with 5 μl 5 M NaOH. Grids were rinsed for 45 sec with 100 μl TBS and blotted off with a filter paper. The sample (5 μl) was added to the grid, left for 45 sec and blotted off with a filter paper. The sample was washed twice with two 100 μl H₂O drops and blotted off after each wash with a filter paper. The sample was stained for 3 sec with 100 μl 0.75% uranyl formate and then stained for additionally 15–20 sec with 100 μl 0.75% uranyl formate. Samples were observed using a Jeol JEM 1230 transmission electron microscope operated at 60 kV accelerating voltage, and recorded with a Gatan Multiscan 791 CCD camera.

Phage lysate (10 μl) was boiled for 10 min to release phage DNA from intact phages. A PCR (Eppendorf Mastercycler personal) was run under the following conditions: 95°C 10 min, 35 cycles of 95°C 1 min, annealing temperature 1 min 30 sec and 72°C 1 min 30 sec, ending with 72°C for 10 min. Final concentration in the mixture was 1× buffer, 0.2 mM dNTP mix, 30 mU/μl Taq polymerase, 1.5 mM MgCl₂ and 1 μM each of the primers. Primers used are described in table 2. All samples were run with an annealing temperature of 56.5°C except when using primer pairs PR264/PAR-2 and MHF/MHR where an annealing temperature of 54°C was used. All reagents except for the primers are from Fermentas (Vilnius, Lithuania). PCR products were washed with SpinPrep PCR Clean-up Kit (Novagen, Madison, WI, USA), sequenced using BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA), and analyzed using an ABI 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Sequences were aligned using the Clustal W algorithm 36 and a phylogenetic tree was reconstructed using the MacVector v9.5.2 software package (Cary, NC, USA). The final partial major head gene nucleotide sequences correspond to nucleotides 163–484 in gp6 from P. acnes bacteriophage PA6. The phylogenetic tree was constructed using UPGMA 37 and uncorrected p-values with a bootstrap with 1000 replications. Phylogenetic trees including outgroups with protein comparisons of a putative major head and an amidase were constructed using neighbor joining and poisson-correction with a best tree mode, while phylogenetic trees without outgroups were analysed using uncorrected p-values.

All statistical tests were calculated using the Chi-Square test. All tests were also run with Fischer's exact test with similar results.

All partial sequences of the gene encoding a putative major head were submitted to GenBank. Accession numbers (EU302607–EU302671) are shown in table 1. Whole sequences of the gene encoding a putative major head for phages PAS2, PAS10, PAS12, PAS40, PAS50, PAD9, PAD20, PAD21 and PAD42 have accession numbers EU784673–EU784681 and the gene encoding a putative amidase have accession numbers EU784682–EU784690.