An accessory wall teichoic acid glycosyltransferase protects Staphylococcus aureus from the lytic activity of Podoviridae

Many Staphylococcus aureus have lost a major genetic barrier against phage infection, termed clustered regularly interspaced palindromic repeats (CRISPR/cas). Hence, S. aureus strains frequently exchange genetic material via phage-mediated horizontal gene transfer events, but, in turn, are vulnerable in particular to lytic phages. Here, a novel strategy of S. aureus is described, which protects S. aureus against the lytic activity of Podoviridae, a unique family of staphylococcal lytic phages with short, non-contractile tails. Unlike most staphylococcal phages, Podoviridae require a precise wall teichoic acid (WTA) glycosylation pattern for infection. Notably, TarM-mediated WTA α-O-GlcNAcylation prevents infection of Podoviridae while TarS-mediated WTA β-O-GlcNAcylation is required for S. aureus susceptibility to podoviruses. Tracking the evolution of TarM revealed an ancient origin in other staphylococci and vertical inheritance during S. aureus evolution. However, certain phylogenetic branches have lost tarM during evolution, which rendered them podovirus-susceptible. Accordingly, lack of tarM correlates with podovirus susceptibility and can be converted into a podovirus-resistant phenotype upon ectopic expression of tarM indicating that a “glyco-switch” of WTA O-GlcNAcylation can prevent the infection by certain staphylococcal phages. Since lytic staphylococcal phages are considered as anti-S. aureus agents, these data may help to establish valuable strategies for treatment of infections.

Peptidoglycan-anchored surface proteins are dispensable for host specificity of Podoviridae. The specific host-range of Podoviridae suggests that these phages might fail to infect and lyse certain S. aureus strains due to unique barriers preventing adsorption, infection, or reproduction. Since the commonly used laboratory and podovirus-resistant S. aureus strain RN4220 (see Fig. 1 and Supplementary Fig. S1) lacks R-M systems, prophages, and CRISPR/cas loci previously shown to impede HGT, an intracellular barrier facilitating resistance to Podoviridae seems implausible. More likely, alterations in peptidoglycan modifications, for example specific cell-surface exposed molecules such as peptidoglycan-anchored 'microbial surface components recognizing adhesive matrix molecules' Scientific RepoRts | 5:17219 | DOI: 10.1038/srep17219 (MSCRAMMs), might block adsorption and infection in certain S. aureus. However, S. aureus RN4220 mutants and mutants derived from the clinical CA-MRSA isolate USA300 lacking functional surface proteins (Δ srtA) were resistant to Podoviridae indicating that factors other than MSCRAMMs interfere with the podovirus infection process ( Supplementary Fig. S1).
Thus, S. aureus peptidoglycan-anchored surface proteins do not influence the unusual host-range of staphylococcal Podoviridae.
The S. aureus α-O-GlcNAc WTA glycosyltransferase TarM prevents the lytic activity of Podoviridae. Because all studied staphylococcal phages require WTA polymers or O-GlcNAcylated WTA polymers for adsorption and infection 17 , adsorption of Podoviridae to their designated cell surface receptors may also be influenced by WTA polymers. Of note, all podovirus-susceptible strains were simultaneously susceptible to the WTA-dependent phages Φ K and Φ 812, which excludes that podovirus-susceptible strains fail to produce WTA polymers (Table 1). In line with this assumption, Podoviridae still failed to adsorb to and infect S. aureus RN4220 or USA300 mutants lacking either WTA (Δ tagO) or WTA glycosylation (Δ tarM Δ tarS) (Fig. 1a,b).
While well-studied WTA-GlcNAc dependent S. aureus phages such as phage Φ 11 do not seem to require a specific stereochemistry of WTA O-GlcNAc for infection 16 the tested podoviruses exhibited an unexpected preference for TarS-glycosylated but not TarM-glycosylated WTA. Strikingly, lack of WTA α -O-GlcNAcylation (Δ tarM) resulted in dramatically increased binding capacities of phage Φ P68 and rendered strain RN4220 Δ tarM highly susceptible to podovirus infection (Fig 1a,b). In contrast, lack of tarS did not lead to phage susceptibility of RN4220 (Fig. 1a). Complementation of the WTA-glycosylation deficient Δ tarM Δ tarS mutant with one of the two S. aureus WTA glycosyltransferases TarM or TarS demonstrated that, (i) Podoviridae require TarS-mediated WTA β -O-GlcNAcylation, but (ii) are inhibited by TarM-mediated WTA β -O-GlcNAcylation (Fig 1a,b). Similar results were obtained for S. aureus USA300 strongly suggesting that TarM diminishes the adsorption and infection of Podoviridae to S. aureus (Fig. 1a,b). Because TarM is an intracellular protein it appears highly unlikely that it interferes with podovirus binding directly but impedes podovirus binding by α -O-GlcNAcylated WTA.
Thus, the α -O-GlcNAc WTA glycosyltransferase TarM prevents the adsorption and infection by staphylococcal Podoviridae.
or absence of the genes encoding WTA glycosyltransferases TarM and TarS via PCR or BLASTN of available genomes 21 . Most strains contained tarS except for strains PS187, which produce an entirely different type of WTA 11,12 , and ED133, which does not encode any of the so far described WTA glycosyltransferases (Table 1). In contrast, several strains lacked tarM. As proposed, most tarM-plus tarS-encoding S. aureus strains were podovirus-resistant (Table 1). Conversely, S. aureus strains exclusively encoding tarS and even other staphylococcal species such as Staphylococcus xylosus or Staphylococcus equorum, which encode tarS homologues with high similarity, but lack tarM, were susceptible indicating that Podoviridae specifically sense β -O-GlcNAcylated WTA (Table 1 and Supplementary Fig. S2). In line with this, the designated podovirus propagation strains PS44A (Φ 44AHJD) and P68 (Φ P68) exclusively encoded tarS (Table 1). However, strain PS66 (Φ 66) encoded both WTA glycosyltranserases, TarM and TarS, which did not align with the assumption that tarM interferes with podovirus susceptibility. Nevertheless, even though tarM was expressed at good levels during logarithmic growth phase, tarS was significantly higher expressed than tarM during early growth stages, which probably promotes the infection by Podoviridae ( Supplementary Fig. S3). Moreover, the S. aureus PS66 tarM gene was sequenced and found to contain two non-synonymous point mutations (Q453K and A464E), which may compromise the TarM function and capacity to interfere with podovirus infection (Fig. 2a). Indeed, podovirus resistance of RN4220 ∆tarM, whose WTA contains only β -O-GlcNAc could be restored completely by complementation with a wild-type tarM but only partially by the mutated tarM (Fig. 2b). In addition, deletion of tarS in PS66 resulted in drastically reduced binding capacity of Φ P68 and rendered PS66 resistant to Podoviridae Next, tarM was expressed in various podovirus-susceptible strains, including the Φ 44AHJD and Φ 66 propagation strains PS44A and PS66. Even at very high phage titers, expression of tarM rendered most susceptible strains completely resistant, confirming the importance of tarM in diminishing infection by staphylococcal Podoviridae (Fig. 3). In addition, the expression of a plasmid-born copy of tarM in strain PS66 also caused complete resistance to Podoviridae, further suggesting that the tarM gene of PS66 is most likely non-functional or less active (Fig. 3).
Tracking the evolution of TarM reveals an ancient origin in other staphylococcal species and vertical inheritance during S. aureus evolution. TarM is encoded outside of the S. aureus WTA gene clusters but does not appear to be encoded on a mobile genetic element 22 . Nevertheless, it is tempting to assume that it has been acquired by S. aureus at some point in evolution to interfere with podovirus infection.
To track the emergence of TarM in S. aureus, the genome sequences of 98 S. aureus strains including those of most S. aureus laboratory test strains used in this study were obtained to infer their genetic relatedness (Fig. 4a,b). Of note, the presence of tarM in the most deeply branching S. aureus isolates MSHR1132 and FSA084, which were recently proposed as novel staphylococcal species Staphylococcus argenteus sp. nov. and Staphylococcus schweitzeri sp. nov. 23 , revealed that the presence of tarM is probably an ancient genetic trait of S. aureus (Fig. 4a). Still, homologues of tarM are also encoded by certain coagulase-negative staphylococci (e.g. specific S. epidermidis isolates) and even by non-staphylococcal species such as Exiguobacterium oxidotolerans and Tetragenococcus halophilus. Thus, the early evolution of tarM probably involved an ancient HGT event to the last common ancestor of contemporary S. aureus clones, further supported by the notion that tarM is flanked by a gene possibly related to conjugation (SACOL1042) (Fig. 4c). However, at a later stage of S. aureus evolution, different types of genetic rearrangements occurred in emerging phylogenetic branches such as CC5 or CC398, leading to a deletion of tarM, which rendered these podovirus-susceptible (Fig. 4c).

Discussion
Staphylococcal Podoviridae infect an unusually wide panel of staphylococcal species but remain avirulent for certain S. aureus lineages probably as a result of the activity of the α -O-GlcNAc WTA glycosyltransferase TarM. In tarM-encoding strains, WTA polymers are probably glycosylated preferentially with α -O-GlcNAc, suggesting that TarM might be more active than TarS. Consequently, TarS-mediated β -O-GlcNAcylation is probably affected by the activity of TarM, thus preventing the adsorption and infection of Podoviridae. Even though it cannot be excluded that TarM potentially has additional and undiscovered functions, which may interfere with the adsorption or infection process, the drastically increased adsorption of Φ P68 in isogenic Δ tarM mutants suggests that α -O-GlcNAcylated WTA prevents the adsorption of Podoviridae to S. aureus. Nevertheless, one of the designated podovirus propagation strains (PS66) encoded both WTA glycosyltransferases suggesting that certain strains, despite encoding tarM, are potentially podovirus-susceptible. Here, TarM might be non-functional, dis-regulated, or mutated as observed in PS66, and cannot interfere with the activity of TarS. Nevertheless, this TarM-mediated phenomenon limits the host-range of Podoviridae, and thus, their therapeutic potential compared to other lytic staphylococcal phages such as Myoviridae.
Apart from this, it remains intriguing as to why certain strains and lineages have lost tarM during evolution to become podovirus-susceptible. Since both S. aureus and S. aureus-like species such as S. schweitzeri and S. argenteus encode tarM and tarS, and many human-associated S. aureus lineages have lost tarM during evolution, it can be assumed that tarM is probably not essential for continued adaptation to the human host. This is in agreement with the observation that both types of WTA O-GlcNAcylation, can mediate S. aureus binding to nasal epithelial cells and thus nasal colonization 24 . Also, human sera contain preferentially serum antibodies directed against TarS-dependent β -O-GlcNAcylated WTA, but not against TarM-mediated α -O-GlcNAcylated WTA 25 , suggesting that tarM may be down-regulated or less immunogenic than β -O-GlcNAcylated WTA during infections. It can be assumed that some S. aureus lineages did not eliminate tarM because WTA α -O-GlcNAcylation may provide S. aureus with a fitness benefit, whose basis remains to be identified in the future.
However, bearing tarM and TarM-mediated α -O-GlcNAcylated WTA protects S. aureus at least against the lytic activity of staphylococcal Podoviridae via a modification of the designated phage adsorption receptor. Such alterations of cell-surface structures serving as viral receptors are only one of many bacterial strategies to counteract phage infection and have also been described for other bacterial species 26-28 , but does not seem a general strategy of S. aureus to avoid phage adsorption and infection. Since other lytic staphylococcal phages such as Myoviridae are capable of infecting tarM-encoding S. aureus isolates, prevention of podovirus infection could be the result of a highly specific WTA-dependent mechanism in S. aureus, presumably as the result of adaptation to specific podovirus-rich environmental niches. In addition, altered phage-receptor binding proteins may easily change the host-range of Podoviridae to render tarM-bearing clones susceptible. Whereas bacterial phage resistance mechanisms such as CRISPR interference appear more efficient and widespread in prokaryotes these can also be bypassed, for example, by CRISPR-evading phages 29 suggesting that host-virus interaction is a constantly evolving process.
For complementation studies (or tarM expression in tarM-lacking strain backgrounds), the previously described E. coli/S. aureus shuttle vector pRB474 was used 32 . pRB474-tarM (Q453K; A464E) has been described elsewhere (formerly pRB474-H-tarM) 15 . PCR-typing, sequencing, and multiple locus sequence typing (MLST). For verification (and sequencing) of tarM and tarS in S. aureus genomes, PCR analysis using primers listed in Supplementary  Table S2 was used. MLST typing of podovirus propagation strains PS44A, PS66 and P68 was performed as described previously using published primers 33 . Experiments with phages. All phages used in this study are listed in Supplementary Table S1. Phages were propagated on S. aureus strains P68 or RN4220 ∆tarM (Φ 44AHJD, Φ 66 and Φ P68), or RN4220 wild type (Φ K, Φ 812) as described previously 34 . Briefly, the cognate S. aureus host strains were grown overnight at 37 °C in BM and diluted in phage-containing lysates (approximately 1 × 10 9 plaque forming units (PfU) per milliliter; titrated on cognate host strains) to a final optical density OD 600 nm of 0.4. Subsequently, CaCl 2 was added to a final concentration of 4 mM. The bacteria/phage mixture was incubated for 30 min at 37 °C without agitation and afterwards for at least 3 h at 30 °C with mild agitation until complete lysis occurred. In order to remove cell debris, the lysate was centrifuged for 10 min (5,000 × g, 4 °C). Lysates were filter-sterilized (0.22 μ m) and stored at 4 °C.
Phage adsorption to S. aureus strains was analyzed as described previously 17 . Briefly, the phage adsorption rate was analyzed using a multiplicity of infection (MOI) of 0.01 for phage Φ P68. Adsorption rate (%) was calculated by determining the number of unbound PfU in the supernatant and subtracting from the total number of input PfU as a ratio to the total number of input PfU.
Phylogenetic analysis. The chromosomes of all S. aureus and S. argenteus and S. schweitzeri labelled as complete were obtained from GenBank (Supplementary Table S3) and aligned against the chromosome of S. aureus CC45 strain CA-347 (GenBank accession ID NC_021554) after identification and deletion of duplicated regions using MUMmer v 3.22 35 . The 98 publicly available genomes were aligned using MUMmer. Based on the identified core of ~1,9 Mb (67%) among all strains, a total of 312,427 SPNs was identified, from which the phylogenetic relationship was inferred using the NeighbourNets algorithm in SplitsTree v4.13.1 36 .