As the vast majority of phages, the S. aureus phages known so far are double-stranded DNA phages belonging to the Siphoviridae family of the Caudovirales order (Table S1) (reviewed in [2,4,6]). In general, they are temperate phages detected as prophage inserted in the chromosome, some of them being lytic due to mutations in the lysogeny functions (e.g., phiIPLA35 and phiIPLA88; [16]; or SA11; [17]). According to the morphological classification previously proposed by Ackermann [18], staphylococcal Siphoviridae are composed of an icosahedral capsid and a non-contractile tail ended by a base-plate structure. Capsids may adopt elongated or isometric shapes, and tail length varies from short (130 nm) to long (400 nm). A small number of S. aureus Podoviridae and Myoviridae phages, also belonging to the Caudovirales order, were described (Table S1). Podoviridae, such as the recently identified SAP-2 phage [19], are composed of a small icosahedral capsid and a short, non-flexible, non-contractile tail. Myoviridae phages, such as the well-known phage K [20], are characterized by an icosahedral capsid and a long contractile tail.

A comparative study of S. aureus phage genomes performed by Pelletier and co-workers [3] revealed several key genomic features, which are globally applicable to all staphylococcal phages described so far. The analysis encompassed 27 genomes from S. aureus phages and prophages belonging to the three morphological families described above.

Genome size extends from less than 20 kb up to more than 125 kb [3]. In contrast to phages from Mycobacterium [21] or Pseudomonas aeruginosa [22], genome sizes are not uniformly distributed, and three categories can be established and used to classify Staphylococci phages (class I: < 20 kb; class II: ≈ 40 kb, class III: >125 kb; see Section 4.1) [3]. Interestingly, genome size categories correlate with the morphological classification, Podoviridae harboring the smallest genomes (class I), Myoviridae the largest ones (class III), and Siphoviridae showing intermediate sizes (class II).

Coding regions are tightly packed with very few and small intergenic regions and a high gene density (1.67 genes/kb in average) [3]. Their GC content is similar to that of the host. S. aureus phages provide an impressive, mainly unexplored source of genetic diversity. On the 2,170 predicted proteins from the 27 phages analyzed in the study of Pelletier and co-workers, a function could be assigned using BLAST to 35% of the ORFs. No match was detected for 44% of these ORFS in Bacteria and Phages Gene Bank databases [3].

Genomes of S. aureus Siphoviridae display the typical structure of the morphological family (Figure 1a) [2,4,23]. Five functional modules are arranged as follows: lysogeny, DNA metabolism, DNA packaging and capsid morphogenesis, tail morphogenesis and host cell lysis. The DNA metabolism module can be divided into replication and regulation functions. When present, virulence factors are generally encoded downstream of the lysis module [2,4]. In some cases, they are inserted between the lysogeny and DNA metabolism modules as reported for phiNM1 to four prophages found in the S. aureus Newman strain [24]. Genes are generally transcribed on the same strand, except for small clusters, such as genes involved in host genome integration [3,4].

While harboring a modular structure as well, organization of Podoviridae genomes is different (Figure 1b) [3,19,25]. One major distinction resides in a smaller number of ORFs, as indicated by a smaller size (20 to 32 ORFs) [3,19,25]. Functional modules encoding DNA packaging and capsid morphogenesis, tail morphogenesis and lysis were identified, in addition to genes of unknown function. In contrast to Siphoviridae, modules are not well defined and tail and lysis genes are overlapping. In addition, the lysogeny module is absent as expected for lytic phages.

Genome organization of staphylococcal Myoviridae (e.g. phages K, G1, Twort) [3,20,26] is similar to E. coli T4 phage, the paradigm for Myoviridae [27]. Genomes are organized into functional modules of conserved genes (replication, structural elements), interrupted by large plastic regions encoding mainly genes of unknown function (Figure 1c). Structural modules found in staphylococcal Myoviridae phages are more closely related (in terms of gene content and organization) to modules found in staphylococcal Siphoviridae phages than to modules of Myoviridae found in different bacterial species.

Phages encode a large proportion of S. aureus virulence factors and provide the pathogen with a large variety of toxins, mainly allowing escaping host immune system (reviewed in [1,6,31]). Many factors have been described and characterized, such as the widespread immune modulator staphylokinase (sak) responsible for host tissue destruction, the chemotaxis inhibitory protein CHIP (chp), the staphylococcal inhibitor of complement SCIN (scn) and several superantigens (sea, seg, sek, sek2, sep, seq). These superantigens are enterotoxins causing food poisoning, toxic shock syndrome and necrotizing fasciitis. In addition, the bi-component cytotoxin Panton-Valentine leukocidin (PVL, encoded by lukF-PV, lukS-PV) and related leukocidins (lukM, lukF)) form pores into leukocytes and cause necrotic infections. Finally, the exfoliative toxin A (eta) is involved in severe skin infections.

In general, phages are carrying single virulence factor genes, although some exceptions have been reported. For examples, phiSa3 phages and relatives, such as phiN315, may encode up to five virulence factors, which form an immune escape complex (IEC) [5,31,32,33]. Virulence factor genes are not strictly associated to a specific phage and appear to be exchanged by horizontal gene transfer and recombination [2,6,31] (see Section 2.2.3). As mentioned above, virulence genes are often located near the attachment site (att) of the prophage, i.e., adjacent to the host chromosome. A possible origin for their acquisition by phages might be aberrant excision events from an ancestral bacterial chromosome [34]. They might also derive from mobile genetic elements, as suggested by the presence of transposase genes flanking toxin genes as in phiPV83 [35]. Thus, phages clearly impact virulence by positive lysogenic conversion, since they provide novel functions and activities to the host. Negative conversion also occurs as prophage insertion might inactivate genes [2,4,31]. In most of the cases, both phenomena occur simultaneously. As an example, phiSa3-related phages integrate in beta-hemolysin gene, rendering lysogens defective for beta-hemolyse but effective for IEC production. The integration site relies on the specificity of the phage integrase [36,37].

Expression of phage-encoded virulence genes is maximal upon entry in the lytic cycle, since latent promoters are activated and phage genome is replicated, leading to an increase in genome copy number [5,38,39], although some expression is detected during lysogeny [34].

Phages are the primary vehicles for horizontal transfer between S. aureus strains. They spread chromosomally-encoded virulence determinants through generalized transduction. In addition, they are also responsible for the mobilization of SaPIs, which encodes major toxin genes, such as the toxic shock syndrome toxin 1 and other superantigens (reviewed in [40]). SaPIs are widespread in S. aureus genomes. They are discrete chromosomal DNA segments that have been acquired by horizontal transfer. They are not mobile by themselves and rely on a helper phage for moving. SaPIs are replicated and mobilized either in response to SOS-induced excision of a helper prophage present in the same strain, either following the infection by a helper phage or by the joint entry of SaPI and helper phage [40]. The underlying molecular mechanism for induction is the specific interaction of a SaPI repressor and a de-repressor encoded by the helper phage. Different proteins of a particular helper phage may be involved in induction of different SaPIs. In particular, phi80alpha is able to mobilize at least five different SaPIs [40,41,42]. Hence, SaPIs mobilization represents a remarkable example of evolutionary adaptation involving pathogenicity islands and phages [42].

Additionally to gene transfer, phages contribute to genetic alterations during infection, providing the species with broad genetic variations [5,38,43,44,45]. In addition to phage acquisition or excision (lysogenic conversion); duplication, ectopic integration and stable extra-chromosomal form of phages have been reported, increasing genetic diversity within bacterial populations [36,44]. Hence, generating heterogeneity within a population upon infection offers different virulence potentials and provides the pathogen with the ability to develop a flexible response to host defenses. It was shown that phage mobilization and atypical genomic integration are favored in pathogenic strains upon infection conditions, compared to colonizing strains in healthy carriers [5,36,44]. Factors causing phage induction in situ are environmental conditions that lead to bacterial DNA damage, such as antibiotics treatments or reactive oxygen species released by macrophages. Phage-mediated phenotypic diversification acts in concert with other mechanisms for genetic variations, such as recombination and mutations, and is under the influence of external factors, such as the presence of co-infecting species [43,45].

The mosaic gene organization is consistent with the theory of modular evolution based on module exchanges by horizontal transfer and recombination events [3,4,73]. This mosaicism can be viewed at either the nucleotide or amino acid level. At the nucleotide level, genome comparisons reveal exchange events that are likely to have occurred recently. Comparisons at the protein level identify homologous proteins that are shared by distantly related phages. They clearly derived from a common ancestor, but have diverged with time, and similarity at the nucleotide level is undetectable [73]. S. aureus phages are highly mosaic, indicating that gene exchange is common within this phage group [3,33,35,37,74,75,76,77,78,79]. Exchanges not only concern single genes, but also protein domains or a group of genes, such as functional modules [3,73]. Interestingly, S. aureus phages often share large, highly similar sequences with at least two other phages and with different phages along their genome. As an example, more than 80% of the ROSA genome is covered by large identical sequences from at least five other S. aureus phage genomes, with remarkable co-linearity. Exchange events are favored between phages of the same genome size range, which correspond to different morphological families (see above Section 2.1.2) [79].

Different molecular mechanisms have been proposed to explain how these exchanges occur [23,73]. A first model invokes homologous recombination events between conserved sequences, which are generally found at gene borders, such as detected in lambdoid E. coli phages [80,81]. Recombination events require large homologous regions that are recognized either by host- or phage-encoded recombinases [80]. A second model proposes alternative recombination mechanisms, occurring randomly or between short sequences [82]. It is likely that these two mechanisms contribute to mosaicism generation. For instance, PVL-related phage genomes are composed of long regions that are shared between different phages, flanked by conserved sequences, called junctions, suitable for homologous recombination events. However, shuffling might occur at other regions, since short regions appear to be shared by phage genomes [35,74,75,76]. Even though these mechanisms certainly lead to a high percentage of non-viable phages, it is considered as a highly creative process that might provide countermeasures against anti-phage mechanisms evolved by bacteria, such as CRISPR sequences, toxin-antitoxin systems and restriction-modification [82].

An interesting aspect of mosaicism is that comparison of a large number of genomes allows the establishment of phylogenetic relationships of specific regions instead of entire phages. Phylogenetic trees of representative genes revealed distinct evolutionary histories for different modules, thereby highlighting extensive mosaicism. This observation thus provides a tool to investigate evolution of S. aureus phages.

General clustering approaches, as used by our group to extend the classification of Staphylococci phages to CoNS species [37], are also useful to represent relationships among phage groups. In order to study the modular evolution of the CoNS StB12, StB27 and StB20 phages, families of homologous proteins were defined and the relative proportion of homologs in the nine clusters shown in Figure 2 was calculated. As previously shown [79], phages belonging to different morphological families do not share homologs. Our study confirms that morphogenesis module genes are in general less prone to horizontal swapping [3,73], although genes involved in host-phage interactions are notable exceptions. Clustering of Staphylococcus Siphoviridae phages relies mainly on structural features, indicating that within the overall similarities of virion structure, subtle variations allow differentiation. This is in accordance with a modular mechanism of evolution in which differentiation processes mainly rely on the exchange of a restricted number of genes or protein domains within structural modules [37]. Importantly, mosaicism encompasses genomes from both Siphoviridae phages from S. aureus and CoNS species, suggesting inter-species gene exchanges, which might be of crucial importance for Staphylococci pathogenesis.