CRISPR/cas Loci of Type II Propionibacterium acnes Confer Immunity against Acquisition of Mobile Elements Present in Type I P. acnes

Propionibacterium acnes is a skin commensal that occasionally acts as an opportunistic pathogen. The population structure of this species shows three main lineages (I–III). While type I strains are mainly associated with sebaceous follicles of human skin and inflammatory acne, types II and III strains are more often associated with deep tissue infections. We investigated the occurrence and distribution of the clustered regularly interspaced short palindromic repeats (CRISPR) in P. acnes, assessed their immunological memory, and addressed the question if such a system could account for type-specific properties of the species. A collection of 108 clinical isolates covering all known phylotypes of P. acnes was screened for the existence of CRISPR/cas loci. We found that CRISPR loci are restricted to type II P. acnes strains. Sequence analyses of the CRISPR spacers revealed that the system confers immunity to P. acnes-specific phages and to two mobile genetic elements. These elements are found almost exclusively in type I P. acnes strains. Genome sequencing of a type I P. acnes isolate revealed that one element, 54 kb in size, encodes a putative secretion/tight adherence (TAD) system. Thus, CRISPR/cas loci in P. acnes recorded the exposure of type II strains to mobile genetic elements of type I strains. The CRISPR/cas locus is deleted in type I strains, which conceivably accounts for their ability to horizontally acquire fitness or virulence traits and might indicate that type I strains constitute a younger subpopulation of P. acnes.


Introduction
The Gram-positive bacterium Propionibacterium acnes is one of the predominant members of the commensal skin microbiota [1,2]. It successfully colonizes sebaceous follicles of healthy human skin, but is also associated with the formation and/or progression of acne vulgaris and with a number of opportunistic infections [3][4][5]. The apparent contradiction between the pathogenic nature and the role as a ubiquitous skin commensal may be partly explained by strain-specific properties. P. acnes strains were categorized as phylotypes I, II and III according to sequence comparison of their tly and recA genes [6,7]. More recently, a multilocus sequence typing (MLST) approach, designated the Aarhus scheme, has been used to further discriminate strains, resulting in the identification of clonal complexes (CC) and 57 sequence types (ST) among 210 strains analyzed [8]. Again, three divisions were identified (I, II and III); division I was further subdivided into I-1a, I-1b and I-2. Subdivision I-1a, including the epidemic clone ST 18 and its descendants (CC 18), comprised significantly more isolates associated with moderate to severe acne. In contrast, other phylotypes of P. acnes are associated with healthy skin and with opportunistic deep tissue infections, in particular type II strains [8]. These findings were recently confirmed by an independent study, which employed an alternative MLST scheme [9]. A comparison of the two MLST schemes revealed a different discriminatory power of the two schemes and underlined the previous observation that the clonal complexes CC 3, CC 18 and CC 31 are associated with acne whereas CC 36, CC 60 and ST 27 are associated with healthy skin, but also with opportunistic infections associated with medical implant devices [10]. P. acnes genome comparison studies revealed the existence of genomic regions with island-like features, indicating that they were horizontally acquired [11,12]. These islands, encoding putative fitness and pathogenic traits, are only present in a subset of P. acnes strains, i.e. in certain CCs. It is unclear so far if these islands are still mobile and could disseminate via horizontal gene transfer (HGT). One powerful system, which restricts the acquisition of incoming DNA, thus of HGT, is a recently discovered 'adaptive immune' system, composed of clustered regularly interspaced short palindromic repeats (CRISPRs) and CRISPR-associated genes (cas) [13][14][15]. CRISPR/Cas systems are found in many bacteria and archaea, conferring immunity to invasion of a variety of foreign mobile elements, i.e. viruses or plasmids. Whether CRISPR loci exist in P. acnes has not been explored so far. Here, we show the exclusive presence of CRISPR loci in type II P. acnes, and reveal their immunological memory by analyzing the spacer sequences. Several spacers originated from mobile genetic elements present in a subset of type I P. acnes strains. Gene content analysis of these elements points to their potential role as virulence and/or fitness traits.

Results and Discussion
Identification of CRISPR/cas in P. acnes genomes We searched the available genome sequences for the existence of CRISPR/cas loci. The completely sequenced genomes of P. acnes KPA171202 (KPA), 266, SK137 and 6609 did not contain a cas gene cluster, nor CRISPR loci. However, the recently decoded genome of P. acnes strain ATCC 11828 [16] as well as some partially sequenced genomes, e.g. strain J139 (reference genome for the Human Microbiome Project (HMP)) contained cas loci, composed of the cas genes cas3-casA-casB-casC-casD-casE-cas1-cas2 ( Figure 1). The gene cluster composition is identical to the CRISPR/cas gene cluster of E. coli K12 W3110 [17]. According to the classification of Makarova et al. [18] the system of P. acnes belongs to type I, since it encodes Cas3, a protein with separate helicase and DNase activities [19]; it can be further classified as a member of subtype I-E. Downstream of cas2, no CRISPR region could be identified in strain ATCC 11828; however, a CRISPR region was identified in strain J139, composed of three identical repeats (GTATTCCCCGCCTATGCGGGGGTGAGCCC) and two spacers. Also other partially sequenced P. acnes genomes, which are sequenced in the framework of the HMP (strains HL001PA1, HL060PA1, HL082PA2, HL110PA3, and HL110PA4) contained CRISPR loci with the same repeat in different copy numbers ( Figure 2).

PCR detection of CRISPR loci in clinical isolates of P. acnes
Since we had no information on the origin and clinical relevance of the five CRISPR/cas-positive P. acnes strains of the HMP, we decided to PCR-screen our strain collection of 108 clinical P. acnes isolates for the existence of CRISPR regions. The strain collection contained clinical isolates from diverse tissue and disease sites, e.g. healthy skin, acne vulgaris, foreign body/joint infections, wound infections, endocarditis, oliocranian bursitis, urinary tract infection, and meningitis (Table S1). All isolates were previously typed according to Aarhus MLST scheme [8,10]; all known phylo-/sequence types of P. acnes were represented in the strain collection as illustrated by a phylogenetic tree based on concatamers of sequences of nine housekeeping genes ( Figure 2). The two primers for CRISPR amplification were derived from sequences up-and downstream of the CRISPR region in strain J139; it was ascertained that the primer sequences are conserved in the CRISPR/cas-positive P. acnes genomes of the HMP.
We obtained PCR products for almost all strains of the sequence types ST 43 to ST 55, representing type III (STs 43, 44) and type II strains (STs 45-53 and ST 55) ( Figure 2, Table S1). No PCR product was obtained with any type I strain, nor with type II strains of STs 54, 56, and 57. The five partially sequenced CRISPR/cas-positive genomes of the HMP are also exclusively type II strains (Figure 2), thus supporting our conclusion, that the CRISPR/cas locus is restricted to type II/type III P. acnes.

Sequence analysis of CRISPR spacer sequences
Subsequently, all PCR products were sequenced. All sequences contained one to nine copies of the repeat sequence GTATTCCCCGCCTATGCGGGGGTGAGCCC. Type III strains (STs 43, 44) contained no spacer, whereas the type II strains had one to maximal eight spacer sequences (Table S2). The standard spacer length was 32 bp. Altogether, 63 spacers in 19 type II strains were identified. A total of 25 distinct spacer sequences could be identified (Table 1) and several spacers detected in different strains were identical; for example spacer no. 4 ((G)AGGGCTACCACGTGGTCGATTTGGACTGTCG) was present in seven different strains (Table S2). All distinct spacer sequences were blasted to nucleotide databases: nine spacers had no hit in any database, four of them matched to P. acnes phages, three spacer sequences matched to mobile elements of other bacteria, one matched to an island-like region in the genome of SK137 (type I, ST 3) and eight matched to a special genomic region in several type I P. acnes genomes. All categories are discussed below.

Phage-DNA specific spacers
The CRISPR/Cas system of type II P. acnes apparently confers immunity against a variety of mobile, ''foreign'' DNA elements. One category of foreign DNA is phages: we detected six spacers (four unique sequences, no. 1, 2, 3, 18) in three different strains whose sequence matched with known P. acnes phages, such as the sequenced siphoviruses PAD20 and PAS50 [20]. P. acnes phages are a very homologous group of viruses, and are distinct from phages isolated from other propionibacteria. It was found that 70% of P. acnes strains have inducible siphoviruses, which most likely have a pseudolysogenic life cycle [20,21]. These phages have a strong lytic activity against all P. acnes isolates with inducible phages, but isolates with no prophages are less susceptible. That might indicate that the phage-resistant strains of P. acnes contain the CRISPR/Cas system. It has to be determined if resistant strains do indeed possess CRISPR loci with phage-derived spacer sequences.
Protospacer within a genomic island present only in type I (CC 3/CC 31) P. acnes strains We found one spacer (no. 4) that was present in seven different strains. The spacer sequence matched to an island-like region in the genome of type I strain SK137 (ST 3). This 20.7 kb island was described previously in a genome comparison study [12], and contains a gene cluster for bacteriocin synthesis ( Figure S1). The island is inserted into the P. acnes core genome, and disrupts a gene (PPA0155 in strain KPA) encoding an ATP-dependent helicase. We blasted this island also against all partially sequenced P. acnes genomes and found it in 13 strains (HL078PA1, HL045PA1, HL007PA1, SK182, HL083PA1, HL005PA1, HL074PA1, HL043PA1, HL053PA1, HL043PA2, HL086PA1, HL056PA1 and HL038PA1) ( Figure 2). Interestingly, these were exclusively type I-1a strains belonging to the two clonal clusters CC 3 and CC 31. We PCR-screened our strain collection for the presence of this bacteriocin island, and found it in nine isolates (Table S1, Figure 2), all of which are CC 3 strains. The PCR screen further revealed that all tested strains of the STs 1-4 were PCR-positive for the bacteriocin gene cluster. Interestingly, five out of nine strains were isolated from acne lesions. We concluded that this island exists in a defined subpopulation of P. acnes; its biological role and its possible acne-associated significance have to be determined in the future. Figure 2. Phylogenetic tree of P. acnes strains used in this study and distribution of mobile genetic elements. The tree was constructed based on 4,287-bp concatamer of partial sequences of nine housekeeping genes, and sequences types (STs) were assigned employing the Aarhus MLST scheme (http://pacnes.mlst.net/). The tree was constructed using the Minimum Evolution algorithm in MEGA version 5.05. The bootstrap values were based on 500 replicates and only values exceeding 50 are shown. Clonal complexes (CCs) were determined according to an eBURST analysis described in Kilian et al. [10]. The distribution of three genomic regions within this strain collection was checked by PCR, namely the CRISPR locus (red circle), the TAD locus (green square), and the bacteriocin (bcn) island (blue triangle Protospacers within a genomic island of type I P. acnes strains encoding a tight adherence (TAD) system Several spacer sequences matched to another genomic region of type I P. acnes strains, which has not been identified previously: 17 spacers (eight unique sequences: no. 6,7,8,10,12,17,22,23 in Table 1) were detected in CRISPR loci of eight different type II strains. BLAST analyses revealed that the protospacers were located on a large genomic region, which is found exclusively in type I P. acnes strains, namely in the partially sequenced genomes of the following 16 HMP strains: HL005PA1, HL007PA1, HL027PA1, HL038PA1, HL043PA1, HL043PA2, HL045PA1, HL046PA2, HL053PA1, HL067PA1, HL072PA1, HL072PA2, HL074PA1, HL078PA1, HL087PA3 and J165 (Figure 2). Within the type I lineage, these strains belong to different CCs (10 strains to CC 3, five strains to CC 18, three strains to CC 28, two strains to CC 31). Again, we PCR-screened our strain collection for the existence of genes located within this island: we found six strains carrying this locus, comprising type I isolates of the CCs 3 and 18 (Table S1, Figure 2). Curiously, also one type II strain (CCUG36609, ST 53), which carries the CRISPR/cas locus, was PCR positive for this genetic element.
In order to in-depth analyze this mobile genetic element, strain 15.1.R1 (ST 3), an acne isolate, was selected for sequencing (Table  S1). After pyrosequencing, spacer sequences were searched against all contigs of the draft genome. Spacers no. 8, 10, 12 and 17 could be located on one sequence contig of 54 kb [contig4483, GenBank: JQ612072]. No core genome insertion boundaries could be detected at the 59 and 39 ends of this contig; it apparently exists as an extrachromosomal element in strain 15.1.R1. Surprisingly, functional analysis of the genes located on this contig revealed that it encodes a secretion system-like complex known as tight adherence (TAD) system (Figure 3, Figure 4). The TAD system has been studied in the Gram-negative Aggregatibacter actinomycetemcomitans [22,23]; the system is dedicated to the assembly and export of Flp pili, and it was shown to be important for host colonization and pathogenesis. Gram-positive Actinobacteria such as Mycobacterium, Corynebacterium and Streptomyces also possess a TAD system composed of TadZABC [23]; homologs of these four proteins were found on the 54 kb element in strain 15.1.R1. The function of this system in Actinobacteria is so far unexplored. Apart from tad-like genes, several other functions are encoded, including toxin/antitoxin systems, a putative autolysin, and ParA family protein, which possibly functions as a plasmid partitioning protein. Interestingly, several protospacer sequences within this element were located in this parA gene ( Figure S2, Figure 3), which underlines the importance of this gene within the mobile element.
Further analyses revealed that the TAD-like island of type I P. acnes is also present in two other species (Figure 4): in the closely related species ''Propionibacterium humerusii'' and in Clostridium leptum. The species tentatively designated as ''P. humerusii'' is yet to be effectively and validly described. ''P. humerusii'' strain P08 was isolated from the humeral membrane of a patient who underwent revision of a failed total shoulder arthroplasty [24]. Three other ''P. humerusii'' strains from the HMP (HL037PA2, HL037PA3 and HL044PA1) also contain the TAD locus. The other species positive for this TAD locus is C. leptum (strain DSM753), belonging to the phylum Firmicutes and a member of a predominant group of bacteria in the gastrointestinal tract [25]. This raises interesting questions about the acquisition and dissemination of this genetic element. In ''P. humerusii'' the TAD locus is encoded on a larger region (Figure 4), which could be inserted into the chromosome; the 59-end of the island-containing contig20 matches to the region PPA1278-PPA1304 of the KPA genome. However, this region is itself an island-like region in the KPA genome [12], and encodes non-ribosomal peptide synthetases. The region also encodes a ParA family protein with similarities to known plasmid partitioning proteins, indicating that this genomic region might be acquired via plasmid insertion. Thus, it is possible that the TAD-containing region is an extrachromosomal element in P. acnes 15.1.R1 as well as in ''P. humerusii''. The origin of this TAD locus remains unclear. Its G+C content is 63% in all three species (P. acnes, ''P. humerusii'', C. leptum), which differs strongly from the C. leptum backbone genome (50.2%) and slightly from P. acnes and ''P. humerusii'' genomes (60%). Thus, it can be ruled out that the TAD locus originated from C. leptum. A phylogram tree was constructed based on all available tadA sequences ( Figure S3). This revealed the higher similarity between tadA of C. leptum and P. acnes, compared to ''P. humerusii'' and P. acnes. Thus, the HGT of the TAD locus between P. acnes and C. leptum must have been a relatively recent event.
Taken together, the data shows that the TAD locus is a mobile genetic element that possibly exists as an extrachromosomal element. In contrast to the bacteriocin island, the TAD locus is disseminated in a rather diverse set of type I P. acnes strains; its dissemination even exceeds the species level. Type II strains with TAD locus-specific spacers in their CRISPR region are apparently protected from the acquisition of this locus.

Protospacers present in other foreign DNA elements
Other spacers were found in CRISPR regions of type II P. acnes; three spacers matched to sequences not related to P. acnes or phages. Spacer no. 19 is identical to a sequence from a microsatellite sequence of the red algae Porphyra haitanensis clone PH-41 and spacer no. 21 is identical to a sequence located on the 271 kb plasmid pACMV1 found in Acidiphilium multivorum AIU301, an acidophilic, aerobic, anoxygenic and phototrophic bacterium isolated from pyritic acid mine drainage [26]. Apparently, P. acnes was exposed to mobile genetic elements harboring these sequences. If they actually originated from these species is highly unlikely, since they do not share the same ecological niche with P. acnes. Spacer no. 13 matched to a sequence within the cse3 gene of Verrucosispora maris AB-18-032, a marine actinomycete [27]; cse3 encodes another Cas family protein. That could indicate that CRISPR/cas-positive P. acnes strains harboring this spacer are protected against the acquisition of an additional CRISPR/cas locus. This underlines that cas genes containing genetic loci are themselves mobile genetic elements, as many CRISPR/cas loci, including the ones found in P. acnes, harbor signatures of their horizontal acquisition [28][29][30]; see below).

Evolutionary aspects: acquisition and deletion of the CRISPR/cas locus in P. acnes
In all analyzed CRISPR/cas-positive P. acnes genomes the CRISPR/cas locus is located on a 16 kb genomic region, which is inserted between genes encoding a histidine ammonia-lyase (HutH = PPA2170 in genome KPA) and a Sir2 family protein (PPA2172) ( Figure 5A). The 16 kb CRISPR/cas-containing genomic region harbors additional genes downstream of the CRISPR locus, including a putative restriction-modification system. This indicates that resistance mechanisms to control incoming DNA are apparently clustered and have likely been horizontally acquired and inserted into the core genome, as it was suggested for CRISPR/cas loci in other bacteria [29][30][31]. Interestingly, there is a genetic fragment in almost all CRISPR/ cas-negative genomes (i.e. type I strains), composed of cas2 and the 39-end of cas1 (PPA2171 in the KPA genome) ( Figure 5A). These cas gene fragments of type I strains are identical to the corresponding genes in the complete CRISPR/cas loci of type II strains, indicating that the CRISPR/cas gene cluster was partially deleted in type I strains. Two deletion events could have taken place, deleting the cas genes located on a 8.4 kb fragment (cas3-casA-casB-casC-casD-casE and the 59-end of cas1) and the down- . Sequence similarity of TAD locus in strains of P. acnes, ''P. humerusii'' and C. leptum. A sequence alignment of the partial TAD-like loci of P. acnes 15.1.R1, ''P. humerusii'' P08 [GenBank: AFAM00000000] and C. leptum DSM753 [GenBank:ABCB00000000] is shown. High sequence similarity (shown in red, .95% nucleotide identity) exists between these loci in different species. The 59 end of the ''P. humerusii'' contig, which has a lower G+C content, matches to a genomic region in the KPA chromosome (PPA1278-PPA1304). The location of CRISPR spacers is indicated; the protospacer ''A'' refers to a spacer found in the CRISPR region of P. avidum. The G+C content is depicted above the open reading frames (sliding window 500 bp); the percentages refer to the average G+C content of the contigs. The figure was made using the Artemis Comparison Tool (ACT, Sanger). doi:10.1371/journal.pone.0034171.g004 stream region of cas2 (7 kb fragment containing CRISPR and downstream genes). A hybrid-like pattern of the CRISPR/cascontaining 16 kb locus is supported by the analysis of the respective locus in Propionibacterium avidum (see below). Looking closer at the second site of deletion, it occurred within the first copy of the CRISPR repeat ( Figure 5B), indicating that the repeats are associated with genomic instability. PCR analyses revealed that a few strains (HL050PA2, 34.1.A1, 39.3.R1 and CCUG33206) did not contain cas gene fragments; these were all type II strains ( Figure 5A). Among the type II strains one strain (CCUG38203, ST 53) was exceptional, since it contained cas1/ cas2 gene fragments as seen in type I strains.
We hypothesize that type II P. acnes strains represent more ancient strains of the species P. acnes, compared to type I strains, which is supported by the significantly higher degree of sequence diversity among type II than among type I strains as illustrated in Figure 2. A possible evolutionary scenario would be that an ancestor type II strain without CRISPR/cas (similar to HL050PA2, 34.1.A1, 39.3.R1, CCUG33206) has acquired this system. Subsequent (partial) loss of the CRISPR/cas system gave rise to a progenitor type I P. acnes strain. Conceivably, the deletion of the CRISPR/cas island facilitated the acquisition of new genetic material, in particular of the two mobile genetic elements described here, thus led to the genomic repertoire of present day type I P. acnes strains associated with acne. ST 53 could constitute a transitional P. acnes class between type I and type II, since some ST 53 strains possess markers of type I strains (TAD locus, cas gene fragments).

CRISPRs in Propionibacterium avidum
CRISPR loci can be found in the genomes of other propionibacteria. P. avidum (strain ATCC 25577) possesses a CRISPR/cas region that differs notably from the locus in P. acnes genomes. The island is slightly bigger (18.7 kb) compared to the P. acnes version. Like in P. acnes, the cas genes are inserted into the backbone P. avidum genome downstream of hutH; they are Figure 5. The CRISPR/cas-harboring locus is a genomic island in P. acnes. (A) A 16 kb gene cluster containing the CRISPR/cas locus is inserted as an island-like region in type II P. acnes genomes (strain J139) between two genes of the core genome, encoding HutH (histidine ammonia-lyase) and a Sir2 family protein (highlighted in yellow). Some type II strains (shown here: HL050PA2 [GenBank:ADYC00000000]) have no insertion at this genomic location. In contrast, all tested type I strains have cas gene fragments (cas2 and the 39-end of cas1) inserted at this site (shown here: KPA), indicating that a functional CRISPR/Cas system was lost in the evolutionary history of type I strains. Red, .97% nucleotide sequence identity. (B) Deletion of the CRISPR locus and downstream genes in KPA occurred within the first CRISPR repeat sequence (red). doi:10.1371/journal.pone.0034171.g005 homologous (78% identity on nucleotide level) to the corresponding genes of P. acnes. However, the region downstream of CRISPR shows no similarity to the P. acnes locus, underlining the hybrid-like nature of the CRISPR/cas locus in propionibacteria. The P. avidiumspecific region encodes an ABC transport system as well as hypothetical proteins. The ABC transport system is similar to macrolide-specific ABC-type efflux carriers (MacAB) [32]. The CRISPR repeat sequence in P. avidum is GTCTTCCCCGCC-TACGCGGGGGTGAGCC, thus has two nucleotide substitutions compared to the repeat in P. acnes. The repeat appears in 28 copies in the P. avidum genome. 26 spacers are predicted; they are specific to P. avidum. Only four show similarities to known sequences: one is similar to a sequence of the Pongo abelii chromosome UNK clone CH276-86H12 (AC188115.1), one to the plasmid pAB510d of Azospirillum sp. B510 (AP010950.1), and one to the gene encoding a major facilitator superfamily protein (VAB18032_27756) in Verrucosispora maris AB-18-032. Interestingly, one spacer (GCTGAATG-GAGAGCGAGCATGAGCACCACCCCT) of the P. avidum CRISPR region matches to a sequence within a gene located on the TAD locus-containing element of strain 15.1.R1 (Figure 4). Thus, P. avidum was also exposed to the TAD locus-containing mobile element or a related element.

Conclusions
The CRISPR/Cas system was detected in a subpopulation of P. acnes strains; these strains were phylogenetically closely related, all part of the type II lineage of P. acnes. Type II strains are often found on healthy skin or in deep tissue infections, they are rarely associated with acne vulgaris [8]. In contrast, no CRISPR/cas locus was identified in type I strains. The CRISPR/Cas system apparently protects against siphovirus infections and the acquisition of two mobile genetic elements. These elements encode a bacteriocin synthesis pathway and a putative TAD system, respectively, which might be important fitness functions. These elements were found in a subset of type I P. acnes strains, namely in strains belonging to the clonal complexes CC 3, CC 18, CC 28, and CC 31. These CCs, except from CC 28, were previously identified as being associated with acne vulgaris. Moreover, CC 36, a clonal complex within type I, which is not associated with acne but with healthy skin, does not possess these mobile genetic elements. Thus, it will be interesting to study the biological role of these mobile genetic elements with respect to their putative contribution to acne formation and/or progression. However, not all acne-associated P. acnes strains harbor these mobile genetic elements. Whereas the bacteriocin-encoding island is restricted to strains of CC 3, the TAD locus-containing element is scattered among diverse type I strains. That could indicate that the locus confers properties to P. acnes which are not directly linked to their association with acne. From the evolutionary perspective, the data suggest that CRISPR/cas-negative P. acnes strains, i.e. type I strains, have lost the CRISPR locus by deletion events, as judged from the presence of remnant cas gene fragments in type I strains. This supports the assumption that type I strains are an evolutionary more recent subpopulation of P. acnes.

Materials and Methods
P. acnes strains and their phylogenetic analysis 108 strains of P. acnes were included in this study, comprising isolates from acne patients and from controls without skin disease, as well as isolates from opportunistic infections [8]. Most of the latter were retrieved from recognized public collections representing clinical isolates from the United Kingdom, U.S.A., Sweden, Norway, and Germany collected between 1920 and 2004. All strains were sequence typed based on allelic profiles of nine housekeeping gene sequences according to the Aarhus MLST scheme (http://pacnes.mlst.net/) [8]; the collection contained isolates from all known sequence types of P. acnes. A phylogenetic tree was constructed based on 4,287-bp concatamer of partial sequences of the nine housekeeping genes using the Minimum evolution algorithm in MEGA version 5.05 [33].

PCR conditions and sequencing
For preparation of bacterial DNA, a loopfull of bacteria was suspended in 100 ml of double sterilized water. Twenty ml of this solution was mixed with 80 ml 0.05 M NaOH and incubated at 60uC for 45 minutes. Subsequently, 9.2 ml of 1 M Tris-HCl, pH 7.0 were added and the solution was diluted 1:100. Five ml of this solution was used for PCR. The CRISPR region was PCRamplified using the following primers: CRISPR-for, 59-TGATCCTGACGAGGGATACG, and CRISPR-rev, 59-CT-TTGCTTGCGGTACTGGAC. In addition, the presence of cas1 was verified by PCR using the primers cas1-for 59-TCGAGC-ACTGCATTGTCAAC and cas1-rev 59-CGTATCCCTCGT-CAGGATCA. The presence or absence of cas1/cas2 gene fragments in the core genome between the genes hutH and sir2 was tested by PCR in all type I and II P. acnes strains using the primers hutH, 59-CGCGGAGTACATGCTCAGTC and sir2, 59-GTCGACTCCGGGCATATGAT. The PCR product was 1.94 kb for strains with cas1/cas2 fragments and 1.22 kb for strains without fragments.

Sequence analyses and software tools
To identify CRISPR regions we used the CRISPR finder tool (http://crispr.u-psud.fr). Spacer sequences were compared to the internal spacer database by blastn. In addition, all spacers were searched against the NR database and the microbial genome database of NCBI. Up to three mismatches were allowed between subject and query sequence. Genome and sequence contig comparisons were done using ACT (http://www.sanger.ac.uk/ resources/software/act/) [34] and WebACT (http://www. webact.org/).
Draft sequencing of the genome of P. acnes strain 15.1.R1 was performed at the Institute for Genome Sciences, University of Maryland School of Medicine, Baltimore, USA, using the 454 FLX Titanium pyrosequencing technology. The draft genome was Supporting Information Figure S1 Bacteriocin island in P. acnes strain SK137 [GenBank: CP001977.1]. The spacer sequence can be found in the 39end of the island (black arrow). The island contains a gene cluster for bacteriocin synthesis. Several genes are homologous to the Sag gene cluster for streptolysin S synthesis in Streptococcus pyogenes. SagD is a scaffold or docking protein, SagC is a cyclodehydratase and SagB participates in the maturation of streptolysin S from a ribosomally produced precursor polypeptide. The Abi genes are probably involved in self-immunity. The cluster is inserted into the backbone genome, within a gene encoding a ATP-dependent helicase (PPA0155 in KPA, upper genome). (PDF) Figure S2 Location of protospacers in the parA gene of the TAD locus. One gene within the TAD locus harbors 4 protospacer sequences. This gene encodes a putative plasmid partitioning protein (Soj/ParA family protein). 3 regions (in yellow) in the coding strand of parA in ''P. humerusii'' P08 (P.hum) are identical to spacers no. 8, 22, 23. An alignment of parA of ''P. humerusii'' and of P. acnes 15.1.R1 is shown. There is one additional protospacer (no. 17, in cyan) in the antisense strand of the parA gene of P. acnes 15.1.R1 (with one mismatch). (PDF) Figure S3 Phylogenetic tree of tadA of P. acnes, ''P. humerusii'' and C. leptum. Only complete tadA sequences were taken into consideration. The tree reveals a high similarity of tadA of C. leptum and P. acnes (genetic distance 0.012), indicative of a recent HGT event. In contrast, the genetic distance between tadA of ''P. humerusii'' and P. acnes is 0.149 (+/20.003) and reflects the overall genetic diversification (0.141+/20.0006 based on the nine genes used in the MLST scheme) of the two species since they separated, i.e. the tadA gene has been present in a common ancestor.

(EPS)
Table S1 Strains used in this study and PCR results for TAD, bacteriocin (BCN) and CRISPR loci. Listed are also the GenBank accession numbers of CRISPR containing sequences. (DOC)