HEK293 cells do not express TLR2 or TLR5 on their own, and do not respond to the appropriate TLR ligands (Willcocks et al., 2013; Metcalfe et al., 2014). HEK293 cells stably transfected with bovine TLR2 (GenBAnk Acc. No AY634629; HEK-boTLR2) were generated in house (Willcocks et al., 2013). HEK cells expressing human TLR5 were purchased from Invivogen (293-hTLR5), and have been used before (Metcalfe et al., 2014). Expression of TLRs was maintained using Geneticin (Gibco) at 1 mg ml^-1 for boTLR2 and Blasticidin (Lifetech) at 10 μg ml^-1 for 293-hTLR5. Both cell types were grown in DMEM (Gibco) supplemented with 10% fetal calf serum (PAA). The preparation of pbMECs as cells that have been shown to be affected by S. aureus expression, their culture in RPMI 1640 (Gibco) on collagen-coated plates, and general challenge conditions were previously described (Gunther et al., 2011).

S. aureus strains isolated from various species and disease presentations (Table 1), were grown on 5% Sheep blood agar (Oxoid) and cultured in brain heart infusion medium (Oxoid) at 37°C. Bacterial genomic DNA was isolated using a Wizard^® Genomic DNA Purification Kit (Promega) following manufacturer’s guidelines. Polymerase chain reaction (PCR) was performed with Easy-A High-Fidelity PCR Mastermix (Agilent) using the primers stated on http://saureus.mlst.net/ (Aanensen and Spratt, 2005) for the genes arcC, aroE, glpF, gmk, pta, tpi, and tpi. Each primer pair amplified an internal fragment of the housekeeping gene and allowed accurate sequencing of ∼450-bp fragments of each gene on both strands. For each locus, the sequences obtained from all isolates were compared and the different sequences were assigned allele numbers. For each isolate, the alleles at each of the seven loci defined the allelic profile that corresponded to a ST.

Identification of Tcp-like proteins within S. aureus sequences were performed by searching for homologs of mammalian and bacterial STIR domains in the non-redundant protein sequences database¹; using the PSI-BLAST algorithm (Altschul et al., 1997). Identified proteins were screened initially for structural similarities to known Tcp using SMART (Schultz et al., 1998²) and the similarity of their tertiary structure compared to other proteins using the BioInfoBank Meta Server³. Subsequently, primers (Table 2) were used to amplify the sequences of SaTlp1 and SaTlp2 (GenBank Acc. No. CAQ50581.1 and CAQ50581.2, respectively) which also introduced a mutation to alter the start codons to ATG, allowing transcription in eukaryotic cells. PCR products were excised from 1% agarose gels using a GenElute Gel Extraction Kit (Sigma), cloned into pGEM-T Easy (Promega) and inserts sequenced. Subsequently, SaTlp1, SaTlp2, and TcpB were cloned into the pcDNA3.3 TOPO plasmid (Invitrogen). All plasmids were propagated in TOP10 E. coli (Invitrogen) bacteria, grown in Luria Broth supplemented with 100 μg ml^-1 ampicillin (Invivogen), purified using a PureYield Plasmid Miniprep kit (Promega) and subsequently used for transfection into HEK-boTLR2, 293-hTLR5, or pbMEC cells.

The secondary structure prediction and 3D template suggestions for structural SaTlp1, TcpB, and TLR2 were performed by HHpred (Soding et al., 2005) and used to generate a manually adjusted structural alignment. The PdTIR structure is comprised of four chains with differing structures which were manually aligned with an Arabidopsis thaliana TIR-like protein and human TIR domains from TLR2 (1FYW), TLR10 (1FYX), and MyD88 (2JS7) to the target sequences. The PDB structure 3LD8 was used to span the BB loop insertion which was not present in any of the TIR protein template sequences. MODELER version 9.10 (Fiser and Sali, 2003a) was used to generate 100 models and the poorly defined loop region (residues 43–49) of the highest ranking model refined by MODLOOP (Fiser and Sali, 2003b). Models were validated using PROCHECK (Morris et al., 1992), Verify3D (Eisenberg et al., 1997), and ProQ (Wallner et al., 2003). Sequence and structural similarity were analyzed by PDBeFold (Krissinel and Henrick, 2007).

The generation of HEK cells expressing TLR2, the main receptor for recognition of Gram-positive bacteria as well as the NF-κB reporter gene assay have been described recently (Willcocks et al., 2013). HEK cells expressing either boTLR2 or TLR5 (293-hTLR5; Invivogen) were seeded at a density of 2.5 × 10⁵ and were transfected after 24 h with 250 ng NF-kB-Luc (Firefly luciferase gene under the control of NF-κB promoter; Promega) and 1250 ng pSaTlp1-3.3, pSaTlp2-3.3, or pTcpB-3.3, using TurboFect^TM in vitro Transfection Reagent (Fermentas) or Lipofectamine 2000 (Invitrogen) following manufacturers’ guidelines and returned to the incubator for 24 h. A plasmid where TcpB had been ligated into pcDNA3.3 TOPO in the reverse orientation (pTcpB3.3^INV) was used as a control. Cells were then split into six wells of a 24 well plate with three unstimulated wells and three wells of cells stimulated with either 100 ng ml^-1 of the TLR2/6 ligand FSL-1 (Invivogen), 500 ng ml^-1 of the TLR5 ligand flagellin (Invivogen) or 30 μg ml^-1 heat-killed E. coli. (FBI Dummerstorf). After 24 h gene activation was analyzed using the Luciferase^® Reporter Assay System (Promega) following manufacturer’s guidelines. The cell lysates were made using passive lysis buffer, centrifuged at 16,000×g for 5 min and the OD 280 nm measured using a ND-1000 spectrophotometer (Nanodrop) to allow for normalization, as previously described (Liu et al., 2011).

To measure the effect of intracellular SaTlp1 or TcpB proteins on the production of inflammatory response genes in pbMEC, their mRNA concentration was determined for those cells transfected with the respective expression vector or the negative control plasmid. The pbMECs were grown to 80% confluence into 9 cm plates and were subsequently transfected with 1000 ng of the respective expression vector using Lipofectamine as described above. After an overnight recovery the transfected cells of each 9 cm plate were split into two wells of a six well plate (in order to obtain duplicates for the following inflammatory challenge). Subsequent to an additional recovery of 1 day the pbMEC were stimulated with 30 μg ml^-1 heat-killed E. coli for three hours. After the stimulation total RNA was extracted using Trizol (Invitrogen) and RNA quality was determined as previously described (Gunther et al., 2011). For cDNA synthesis, 1.5 μg RNA was primed with a mixture of gene specific reverse oligonucleotide primers (Table 3) as described (Goldammer et al., 2004). After cDNA purification with HiPure columns (Roche) relative mRNA concentrations were determined with quantitative real-time PCR (RT-qPCR) using the LightCycler instrument and the SYBR green Plus kit (both from Roche), essentially as described previously (Goldammer et al., 2004; Gunther et al., 2011). Sequences of the oligonucleotide primers for the Chloride intracellular channel protein 1 (CLIC1; house-keeping gene), the proinflammatory cytokines TNF-α, IL-1β, IL-6, and CXCL8, inducible Nitric oxide synthase 2 (iNOS2), the bactericidal β-defensin lingual antimicrobial peptide (LAP) and the acute phase protein Serum amyloid A (SAA3) are given in Table 3.

To measure the effect of intracellular SaTlp1 or TcpB proteins on the production of inflammatory response genes in pbMEC, the corresponding supernatants from pbMEC used to generate mRNA were analyzed for the presence of CXCL8 and TNF using ELISA systems as described recently (Metcalfe et al., 2014). IL-1β was analyzed using a bovine IL-1β ELISA kit (Pierce Protein Biology Products) as per manufacture’s recommendation.

In order to delete SaTlp1 and SaTlp2 from the ST398 genome the pIMAY plasmid and protocols were used as previously described (Monk et al., 2012). Deletion constructs were produced by amplifying regions 500-bp upstream and downstream of the genes using primer pairs A/B and C/D respectively. These PCR products were used as templates for the spliced overlap extension (SOE) PCR using A/D primers. The PCR product was purified and cleaved at endonuclease sites added by the A and D primers and cloned into pIMAY using standard protocols. Allelic exchange was achieved by transforming electrocompetent ST398 with the pIMAY construct and plating on brain heart infusion agar (BHIA) with chloramphenicol (Cm) at 28° C (Monk et al., 2012). Potential mutants were identified after antisense secY induction by colony PCR with OUT-F/OUT-R primers. The double mutant was confirmed by sequencing the deleted region from isolated genomic DNA.

The murine peritoneal S. aureus infection is generally considered by be a good model to assess invasiveness and subsequently infectivity of S. aureus. The study was performed at the AAALAC-accredited Michigan State University In vivo Facility (East Lansing, MI, USA, study number 14IVF-IF-1003, IACUC approval No. 01/13-013-00⁴). The study was conducted in accordance with the current guidelines for animal welfare (Guide for the Care and Use of Laboratory Animals, 8th Edn, 2011). The procedures used in this study were reviewed and approved by the Institutional Animal Care and Use Committee. Six- to eight-week-old female CD-1 mice (Charles River, Portage, MI, USA) with an average weight of 23.8 gm were randomly assigned into 13 groups (n = 6 per group). Animals were housed three per cage and provided food and water ad libitum. Animals were acclimated 8 days prior to use. One day prior to infection a parent culture of S. aureus ST398, and the KO strain ΔSaTlp1/ΔSaTlp2 were removed from frozen storage and streaked out onto trypticase soy agar supplemented with 5% lysed sheep blood cells (TSA-5% SB). Cultures were grown overnight at 37°C, at ambient atmosphere. On the day of infection, the cultures were removed from the incubator and used to prepare an inoculum of each organism in trypticase soy broth (TSB). Bacterial concentrations were measured in the broth by forward light scatter measurement at 600 nm (O.D.600) and adjusted to provide a target value of 0.600. A 1:100 dilution was prepared, using TSB supplemented with 10% gastric hog mucin (GHM) as the diluent, to generate a high dose concentration for challenge of 7.0 log10 bacteria in a dose volume of 0.5 mL. Serial 1:10 dilutions of the high concentration were made into TSB-10% GHM to prepare 6.0 log10, 5.0 log10, 4.0 log10, 3.0 log10, and 2.0 log10 challenge doses of each strain, and the inoculum concentration of each dosing preparation was confirmed by plating serial dilutions for CFU enumeration. Animals received mucin control, wild-type S. aureus, or KO S. aureus as a single intraperitoneal injections. Animals were monitored at least twice daily for signs of morbidity. During periods of expected high mortality, mice were monitored a minimum of four times per day. Animals exhibiting clinical signs listed as criteria for implementation of early euthanasia in AUF 01/13-013-00 were euthanized by CO2 asphyxiation, followed by cervical dislocation. Deaths were recorded as the number “died” or “euthanized” on a daily basis, and the study was terminated on Day 6. Any remaining animals were euthanized at that time.

All in vitro experiments were performed at least three times, with samples run in at least duplicates. Data are presented as averages of all experiments, and results were assessed for statistical significance by a two-way ANOVA followed by a Bonferroni t-test between treatment groups using Excel and GraphPad Prism software packages (version 5; GraphPad Software, Inc., La Jolla, CA, USA).

For the in vivo experiment, data were collected on an Excel spreadsheet. Survival curves were compared statistically using Log-rank (Mantel-Cox) and Gehan-Breslow-Wilcoxson tests (GraphPad Prism 5.03). Differences were considered significant at p < 0.05. LD50 curves were generated using GraphPad Prism 5.03. A sigmoidal Emax dose response model derived from the Hill equation (four-parameter logistic equation with variable slope using GraphPad Prism v5.03) was used to determine the relationship between the antibacterial drug dose and the efficacy, as determined by bacterial burden: Y = D +(A-D)/(1 +10(log C–X)*α). For the four-parameter logistic equation, Y is the observed effect, D is the bottom, A is the top, C is the LD50 or 50% of the observed maximum effect, X is the log of the challenge concentration and α is the Hill slope. The bottom value (D) was constrained to 0 and the top value (C) was left unconstrained. Data are shown as Kaplan–Meier Survival Curves.

Initially we tried to identify Tcp in S. aureus using a bioinformatics approach. As Tcps in other bacteria have been shown to show similarities to mammalian TLR TIR domains, we compared different TLR TIR domains against all available S. aureus genomes (taxid:1280) using BLAST. This resulted in the identification of two immediately adjacent proteins (e = 8 × 10^-6 and e = 3 × 10^-5, respectively) containing a DUF1863 domain, located within a recently identified putative transposon. SMART annotation confirmed that these proteins, named S. aureus Tcp-like protein 1 and 2 (SaTlp1 and SaTlp2) contain DUF1863 domains_. SaTlp1 and SaTlp2 share less than 20% sequence identity with other members of the TIR superfamily. However, despite low similarity at the sequence level, secondary structure prediction suggests that at least four of the central beta strands common to TIR-like domains are present in these bacterial proteins. The closest structural homologs identified by HHpred are matches to the DUF1863 (PF08937) and TIR or TIR_2 (PF13676 and PF01582) domain-containing families and proteins sharing a Flavodoxin-like fold (SCOP 31128 and 31130). The closest secondary structure match to both SaTlp1 and TcpB was PdTIR (PDB: 3H16) while the Eubacterium rectale putative signal transduction protein containing a Flavodoxin-like fold (PDB: 3HYN) was the top ranked hit for SaTlp2 with probabilities of 99.9, 100, and 99.7%, respectively. Within eukaryotic TIR domains, three conserved regions called boxes 1, 2, and 3 are proposed as putative adaptor binding sites, essential for signaling. The box 1 motif can be clearly identified in all of the TIR-like domains included in the alignment (Figure 1A). However, both SaTlp1 and SaTlp2 appear to have a helical insertion within the BB loop (box 2) and, unlike the other bacterial proteins shown, seem to entirely lack box 3. Computational modeling of SaTlp1 predicts the conserved TIR-like tertiary structure of four/five central beta strands surrounded by helices (Figure 1B). While the modeled structure of TcpB is most similar to PdTIR (∼0.5 Å), SaTlp1 shows closer structural homology to TLR TIR domains (∼1.9 Å).

Having identified two putative Tcp-like proteins, we next assessed the occurrence of these genes in different S. aureus strains. To do so, a phylogenetic tree of all S. aureus strains used in this study was created by MLST analysis, and the presence of SaTlp1 and SaTlp2 in these strains determined (Table 1). Interestingly, SaTlp1 and SaTlp2 were identified in all ST398 isolates tested, regardless of the species of isolation, clinical symptoms of the host-species or methicillin-sensitivity status. Additionally, BLAST searches on 89 ST398 isolates were performed which revealed that all of these contained these genes. In addition to ST398, BLAST searches revealed these genes were present in S. aureus clonal complex (CC) 75 (also known as S. argenteus), however, they were not detected in any other S. aureus lineage. ST398 and CC75 are only distantly related (Figure 2) and published work on CC75 states that it is a highly divergent lineage, showing an average seven fold greater nucleotide divergence in orthologous genes with typical S. aureus strains than other strains (Holt et al., 2011). Considering this divergent relationship but the similarity of these genes it seems highly likely that the transposon containing them was transferred between the strains post-dating their divergence.

To assess whether the presence of the Tcp-like proteins in this S. aureus strain interferes with the activation of the innate immune response, we tested whether the identified proteins would exhibit similar effects as described for Tcps of other bacteria which seem to inhibit TLR signaling. We analyzed whether transfection of HEK cells expressing boTLR2 with SaTlp1 and SaTlp2 affected the downstream activity of the transcription factor NF-κB after stimulation with the TLR2 ligand FSL-1 (Figure 3A). The levels of NF-κB-induced luciferase were significantly up-regulated following transfection with SaTlp1 and SaTlp2 plasmids (both p < 0.001) when compared to the negative control. Simultaneous transfection of both plasmids also up-regulated NF-κB activation (p < 0.001). In the HEK-boTLR2 system these plasmids up-regulated NF-κB-dependent luciferase between 4.1 (SaTlp1) and 6.1 (SaTlp1/2) fold following stimulation compared to the negative control. In contrast, transfection of HEK-boTLR2 cells with TcpB which had been shown before to block TLR-dependent NF-κB activation (Radhakrishnan et al., 2009) did not induce significant NF-κB stimulation, confirming these data. FSL-1 failed to activate NF-κB in HEK293 not expressing TLR2 (Data not shown).

As the HEK cell system is a somewhat artificial system, we next assessed the effects of SaTlp1 using pbMECs which express TLR2 and have been used before to assess the effect of mastitis causing pathogens on the innate immune response (Yang et al., 2008; Gunther et al., 2011). As seen in HEK293 cells, transfection of SaTlp1 (p = 0.006) caused increased NF-κB activation in pbMEC in response to heat-killed E. coli. As before transfection of pbMECs with TcpB before E. coli stimulation did not activate NF-κB signaling (p < 0.001). However, this up-regulation was overall lower compared to HEK293 cells with SaTlp1 up-regulating NF-κB by 1.4 fold compared to the negative control (Figure 3B).

To assess whether the responses seen in both cell types would be unique to the interaction with boTLR2, we examined the effect of SaTlp1 and SaTlp2 in HEK cells transfected with hTLR5 cells stimulated with flagellin (Figure 4). Transfection of SaTlp1 SaTlp2 (both p < 0.001) and SaTlp1/2 (p < 0.001) led to an up-regulation of NF-κB activity of between 2.0 (SaTlp2) and 2.5 (SaTlp1) fold. In contrast, transfection of HEK cells expressing hTLR5 with TcpB down-regulated its activation in response to flagellin (p < 0.001). These results are therefore similar to those for boTLR2 expressing cells, suggesting the action of both proteins may not be limited to a specific TLR.

Having established that SaTlp1 stimulated NF-κB activation, we investigated whether this was also reflected at the transcriptional level for NF-κB-dependent cytokine production. We analyzed expression of the primary response genes IL-1β, IL-6, CXCL8, and TNFα as well as the secondary response genes iNOS-2, LAP, and SAA3 by qPCR, similar to that described before (Gunther et al., 2011). Stimulation of pbMEC transfected with either empty plasmids or plasmids containing SaTlp1 followed by E. coli stimulation resulted in the up-regulation of IL-1β, IL-6, CXCL8, and TNF-α mRNA, levels to various extents (Figure 5). In contrast, transfection of pbMEC with a plasmid coding for TcpB in general reduced expression of these genes after E. coli stimulation (Figure 5). Transfection of pbMEC with plasmids coding for SaTlp1 and TcpB followed by subsequent stimulation with E. coli increased mRNA expression for the secondary response genes iNOS-2, LAP and SAA3 (Figure 6), with LAP expression being effected more by TcpB.

Corresponding supernatants were analyzed for the presence of CXCL8, TNF-α, and IL-1β. The secretion pattern followed in general the results obtained for qPCR, with values obtained for all three cytokines being in generally low. Similar as to the data obtained by qPCR, supernatants of pbMEC transfected with SaTlp1 before stimulation with E. coli contained the largest amount of all three cytokines (Figures 7A–C), whereas no differences were seen between untransfected pbMEC and pbMEC transfected with TcpB.

As the presence of SaTlp1 and SaTlp2 seemed to increase NF-κB activation and pro-inflammatory cytokine production in response to subsequent stimulation, we next assessed whether the absence of the genes has an impact on the virulence of S. aureus ST398 in vivo. After creation of a ΔsaTlp1/ΔsaTlp2 ST398 KO mutant, the growth rate or colony appearance of both wild-type and mutant S. aureus were compared and showed no difference (data not shown). In order to determine whether there are differences in the in vivo pathogenicity of wild-type and ΔsaTlp1/ΔsaTlp2 ST398 a murine sepsis model was used in order to calculate the median lethal dose (MLD) of the bacteria. CD-1 mice were challenged via i.p. injection with 10²–10⁷ CFU of either wild-type or mutant S. aureus after which the time until their death was monitored. The data showed a trend for increased survival rates for mice infected with the ΔsaTlp1/ΔsaTlp2 mutant, most noticeably in the case of infection with 10⁵ CFU when 2 of 6 mice challenged with mutant bacteria survived until day 6 of the experiment whereas all mice infected with wild-type had died by day 2 (Figure 8). Despite the fact that none of the observed differences reach level of significance, the lowest p-value (p = 0.13) was obtained for 10⁵ CFU. Considering the data from all challenge concentrations, the MLD of the strains was calculated using the Reed-Meunch equation. This showed the MLD of ΔsaTlp1/ΔsaTlp2 ST398 to be 3630 CFU compared to one of 575 for the wild-type ST398 indicating that in order for 50% of challenged mice to be killed by the mutant, an infectious dose of over 6.3 times that of wild-type ST398 is required. This suggests that although differences between wild-type and ΔsaTlp1/ΔsaTlp2 ST398 are not significant at individual concentrations, when is considered together there does appear to be genuine differences in their pathogenicity.