Staphylococcus aureus Lpl protein triggers human host cell invasion via activation of Hsp90 receptor

Staphylococcus aureus is a facultative intracellular pathogen. Recently, it has been shown that the protein part of the lipoprotein‐like lipoproteins (Lpls), encoded by the lpl cluster comprising of 10 lpls paralogue genes, increases pathogenicity, delays the G2/M phase transition, and also triggers host cell invasion. Here, we show that a recombinant Lpl1 protein without the lipid moiety binds directly to the isoforms of the human heat shock proteins Hsp90α and Hsp90ß. Synthetic peptides covering the Lpl1 sequence caused a twofold to fivefold increase of S. aureus invasion in HaCaT cells. Antibodies against Hsp90 decrease S. aureus invasion in HaCaT cells and in primary human keratinocytes. Additionally, inhibition of ATPase function of Hsp90 or silencing Hsp90α expression by siRNA also decreased the S. aureus invasion in HaCaT cells. Although the Hsp90ß is constitutively expressed, the Hsp90α isoform is heat‐inducible and appears to play a major role in Lpl1 interaction. Pre‐incubation of HaCaT cells at 39°C increased both the Hsp90α expression and S. aureus invasion. Lpl1‐Hsp90 interaction induces F‐actin formation, thus, triggering an endocytosis‐like internalisation. Here, we uncovered a new host cell invasion principle on the basis of Lpl‐Hsp90 interaction.


| INTRODUCTION
Many bacterial pathogens trigger internalisation into non-professional phagocytes, which is crucial to virulence because this shields the pathogen from certain immune defenses, antibiotics and enables the proliferation in relatively protected niches. Although previously considered an exclusively extracellular pathogen, Staphylococcus aureus is now regarded as a facultative intracellular pathogen that triggers internalisation by non-professional phagocytic cells (NPPCs) such as endothelial, epithelial, and mammary cells as well as fibroblasts or osteoclasts and can persist intracellularly for various periods of time Lowy, 1998;Sinha et al., 1999). Clinical studies also indicate a possible role for an intracellular staphylococcal reservoir in recurring diseases, such as rhinosinusitis or osteomyelitis (Clement et al., 2005;Kalinka et al., 2014;Mohamed et al., 2014).
A prerequisite for any internalisation into NPPCs is host cell adhesion. This step mainly involves fibronectin (Fn), forming a bridge between α5β1 integrin on the cellular side and Fn-binding proteins on the bacteria (Fowler et al., 2000;Grundmeier et al., 2004;Sinha et al., 1999;Tran Van Nhieu & Isberg, 1993). The FnBP-Fn-α5β1 integrin pathway is widely acknowledged to be the main internalisation process. However, there are various so-called secondary mechanisms. These mechanisms mainly involve bacterial serine aspartate repeat-containing protein D (SdrD), clumping factor A (ClfA), serine-rich adhesin for platelets, and the major autolysin, Atl (Josse, Laurent, & Diot, 2017;Zapotoczna, Jevnikar, Miajlovic, Kos, & Foster, 2013). These proteins are microbial surface components recognising adhesive matrix molecules (MSCRAMMs) and (except for Atl) have a cell-wall anchoring sequence located in their C-terminal portion (Josse et al., 2017). For example, SdrD binds, for example, directly to Desmoglein 1 in keratinocytes, promoting adhesion (Askarian et al., 2016;Corrigan, Miajlovic, & Foster, 2009). Additionally, ClfA can interact through fibrinogen bridges to the alpha-V beta-3 integrin (αV β3) or complex bridge involving von Willebrand factor, the secreted von Willebrand factor binding protein and the αV β3 integrin that promotes adhesion in vascular endothelial cells (Claes et al., 2017;McDonnell et al., 2016). On the other hand, SraP adheres to gp340, a salivary scavenger protein, in A549 cell line (Yang et al., 2014). The major autolysin, Atl, mediates S. aureus internalisation via direct interactions with Hsc70 (Hirschhausen et al., 2010). It has been speculated that the various internalisation mechanisms allow the bacteria to expand their internalisation to changing environmental conditions; for example, in the absence or scarcity of Fn, they can make use of alternative binding partners to trigger invasion (Josse et al., 2017).
Recently, a certain class of lipoproteins, the so-called "lipoproteinlike lipoproteins" (Lpls) were found to induce host cell internalisation (Nguyen et al., 2015). The lpl-genes are clustered on a pathogenicity island called νSaα island (non-phage and non-staphylococcal cassette chromosome genomic island), which is present in all S. aureus strains tested so far (Diep et al., 2006;Shahmirzadi, Nguyen, & Götz, 2016).
The lpl cluster comprises 10 lpl paralogous genes that encode 10 Lpl proteins with high sequence similarity and two accessories genes (Nguyen et al., 2015). When the entire lpl gene cluster is deleted in S. aureus USA300, the mutant showed a marked decrease in invasion of S. aureus into human primary keratinocytes and mouse skin and also showed a decreased pathogenicity in a mouse sepsis model (Nguyen et al., 2015). The Lpl lipoproteins not only trigger host cell invasion, but also delay the G2/M phase transition in HeLa cells . As the number of lpl genes is particularly high in epidemic S. aureus strains, it is assumed that the lpl gene cluster might contribute to increased dissemination and epidemic spreading by shielding the pathogen from the immune defence and antibiotic treatment (Nguyen et al., 2015). However, the mechanism of how Lpl proteins trigger the host cell internalisation was unknown.
Here, we identified the human heat shock protein Hsp90 as the host receptor for Lpl-induced S. aureus USA300 invasion of human keratinocytes using Lpl1 as a model of Lpls for in vitro experiments.
The Hsp90-Lpl interaction triggers a cascade of reactions including ATPase activity and F-actin formation, indicating that the bacterial internalisation underlies an endocytosis-like process.

| Human Hsp90 interacts with S. aureus Lpl1 protein in pull-down experiments
Previously, it has been shown that Lpl lipoproteins from S. aureus USA300 increase the internalisation into HaCaT cells, a human keratinocyte cell line and also in human primary cells (Nguyen et al., 2015;Nguyen, Peisl, Barletta, Luqman, & Götz, 2018). As the lipid moiety is anchored in the cytoplasmic membrane, we expect that the protein part interacts with the potential host cell receptor. To capture the host cell receptor, we used Ni-NTA (nikel-nitrilotriacetic acid agarose)-bound Lpl1-his as a bait. Lpl1 from USA300 was used as our model Lpl protein. It was expressed without the lipo signal peptide and with a C-terminal His-tag in S. aureus SA113 (pTX30::lpl1-his) and purified by Ni-NTA affinity chromatography. Purified Lpl1-his was bound to Ni-NTA and loaded with HaCaT cell lysate. After extensive washing, the HaCaT proteins bound to Lpl1-his were eluted with 250 mM imidazole and 500 mM NaCl.
The elution fraction containing the proteins that interacted with the Lpl1-his and the control fraction (in which the cell lysate passed through Ni-NTA without bound Lpl1-his) were separated by SDS-PAGE followed by Coomassie blue staining ( Figure S1a).
The most prominent band on the SDS-PAGE was Lpl1-his. There were five lower-sized protein bands that were not present in the control lane and were used for further analysis by Nano-HPLC-MS/MS.
The identified HaCaT-specific proteins are listed in Table 1. We only considered HaCaT proteins that were present in the Lpl1-bound Ni-NTA but not in the control column. The most abundant proteins with the highest coverage and posterior error probability (PEP) of 0.01 or lower were the human heat shock Hsp90 alpha (Hsp90α) and beta (Hsp90β) proteins. Both proteins are highly homologous, sharing 94% similarity and 86% identity. Hsp90 proteins are approximately 90 kDa; however, the Lpl1-interacting proteins were about 15 kDa. This indicates that the detected Hsp90 proteins were truncated, most likely due to proteolytic degradation, although a protease inhibitor cocktail was used during the preparation of HaCaT cell lysate.

| Hsp90α and Hsp90β are localised to the cell surface in HaCaT cells
It has been described that both the Hsp90α and Hsp90β isoforms were found on the cellular surface in different cell lines and tissues (Bozza et al., 2014;Eustace et al., 2004;Suzuki & Kulkarni, 2010).
Here, we confirmed that both the Hsp90 proteins were localised to the cell surface of HaCaT cells and co-localised with FM 5-95 stained membrane via immunofluorescence analysis ( Figure 1).

| Hsp90 antibodies block USA300 adherence and invasion
To further confirm the contribution of Hsp90 in Lpl-triggered invasion into HaCaT cells, we blocked Hsp90 proteins with specific antibodies in invasion assays in the presence of fetal serum to mimic in vivo conditions where soluble fibronectin is present. We used monoclonal antibodies specific against Hsp90α (α-Hsp90α) and Hsp90β (α-Hsp90β) and a polyclonal antibody that recognises both Hsp90α and Hsp90β USA300Δlpl, the entire lpl cluster was deleted (Nguyen et al., 2015).
In this strain, the polyclonal α-Hsp90αβ antibodies showed no effect on invasion; however, in the lpl-complemented mutant, the invasion was decreased again ( Figure 2a). As fibronectin binding protein (FnBPs)-fibronectin-α5ß1 integrin pathway is a key mechanism for S. aureus adherence and invasion into the host cells, we verified that the FnBPs expression was similar in the wild type USA300 and the Δlpl mutant ( Figure S1b).
The invasion frequency was normalised to the unblocked HaCaT cells, which was set at 1.0 and was used as a comparator for further experiments. Pre-incubation with monoclonal α-Hsp90α caused a decrease in invasion from 1.0 to 0.20 ± 0.05, whereas pre-incubation with monoclonal α-Hsp90β caused a decrease in invasion from 1.0 to 0.27 ± 0.08; in both cases, the invasion was decreased to fivefold and fourfold, respectively. As a control, we used human IgG that has no effect on internalisation ( Figure 2b).
We also investigated the influence of Hsp90 proteins in human embryonic kidney 293 cell line which is untransfected by TLR2 (HEK-0). The result was similar to our earlier observation; α-Hsp90α decreased the internalisation by fivefold ( Figure S2).
Additionally, we investigated the role of Hsp90α in USA300 invasion on primary keratinocytes by using α-Hsp90α to block Hsp90α.

| Geldanamycin blocks S. aureus invasion in HaCat cells
Geldanamycin is a well-known cell permeable anti-neoplastic compound that competes with ATP binding in Hsp90, thus inhibiting Hsp90 activity (Gorska et al., 2012). Here, we investigated whether this compound affects USA300 internalisation. Indeed, the addition of geldanamycin at a concentration of 5 μM decreased the USA300 invasion in a dose-dependent manner by approximately threefold ( Figure 2d). These results further confirm the role of Hsp90 in USA300 invasion. It should be mentioned that geldanamycin had no growth-inhibiting effect on S. aureus at all used concentrations.

| Silencing of Hsp90α expression by siRNA causes a decrease of USA300 invasion
We have shown that the blocking of both Hsp90α and Hsp90β by antibodies caused a fourfold to fivefold decrease in invasion of S. aureus cells. However, for reasons of simplicity, we focus mainly on Hsp90α in the following experiments. Hsp90α is a protein that is inducible by oxidative and heat stress (Prodromou, 2016;Profumo et al., 2018),  (Figure 3a). As a control, random siRNA was used and no effect on invasion was seen ( Figure 3a). The silencing of Hsp90 expression was also confirmed by Western blot analysis ( Figure S3a).

| High temperature-induced expression of Hsp90α causes an increase of USA300 invasion
Hsp90α is an inducible heat shock protein. Its expression increases when the temperature increases, whereas the Hsp90β isoform is constitutively expressed (Prodromou, 2016). An increased body temperature (fever) is frequent during bacterial infection. Therefore, we investigated if an increased temperature upregulates Hsp90α expression, thus leading to an increased USA300 invasion. In order to study this, HaCaT cells were pre-incubated for 2 hr at 39°C prior to the invasion assay. The higher temperature used resulted in an almost twofold (1.86 ± 0.24) increase in invasion relative to the control at 37°C ( Figure 3b). To confirm that the increased invasion was indeed due to the increased Hsp90α expression, we blocked it with α-Hsp90α; as expected, it caused a decrease in invasion at 39°C (Figure 3b).
The increased Hsp90α expression at 39°C was also confirmed by Western blot analysis ( Figure S3b). Seeded and fixed with 4% paraformaldehyde were 1 × 10 5 HaCaT cells. Immunofluorescence was performed using α-Hsp90α or α-Hsp90β as primary antibody and α-mouse IgG-Alexafluor488 as secondary antibody. Control samples were treated the same as experimental samples but without primary antibody. FM5-95 was used as membrane stain. α-Hsp90α: primary mouse antibody to Hsp90α;α-Hsp90β: primary mouse antibody to Hsp90β.

| USA300 DOES NOT ALTER Hsp90 EXPRESSION OR LOCALISATION IN HaCaT CELLS
We investigated if USA300 or Lpl1-his affected Hsp90α expression by using Western blot analysis. Neither USA300 nor purified Lpl1-his affected Hsp90α expression within a period of 6 hr after the addition of USA300 ( Figure S4a). In addition, we also tested whether USA300 affects Hsp90α localisation by using flow cytometry analysis in non-permeabilized HaCaT cells. The amount of Hsp90 on HaCaT surface was similar after 1.5 hr exposure of the cells to USA300 ( Figure S4b).

FIGURE 2
Hsp90 antibodies or inhibitors block USA300 invasion. (a) 1 × 10 6 HaCaT cells with or without pre-incubation during 1 hr with α-Hsp90αβ or control human IgG (IgG). Cells were further infected with Staphylococcus aureus USA300 strain, USA300 Δlpl mutant strain, or USA300 Δlpl + pTXlpl with a MOI of 30 in DMEM medium, supplemented with 10% FBS. Host cells were infected for 1.5 hr followed by lysostaphin treatment for 1.5 hr. Cells were lysed and USA300 CFU was determined by plating on TSA agar plates. The experiments were performed in at least three independent biological replicates. Error bars indicate standard deviation. The statistical was calculated by using Student t test; **p < .01; *p < 0.05, comparing with the control for each strain. (b) Invasion assays using S. aureus USA300 wild type were performed as described above but specific isoform antibodies (35 μg ml −1 ) were used: α-Hsp90α, α-Hsp90β, and human IgG (IgG) as an extra control. Relative invasion was calculated by normalising USA300 invasion in treated HaCaT cells to USA300 invasion in untreated control cells (C). The experiments were performed at least in three independent biological replicates with at least three technical replicates. Error bars indicate standard deviation. The statistical was calculated by using One-way Anova with multiple comparison to invasion in the control (C) sample, **p < 0.01, *p < 0.05. (c) Primary human keratinocytes were cultured in collagen-coated tissue flasks in epidermal keratinocyte medium. Keratinocytes were differentiated with 1.7 mM CaCl 2 in epidermal keratinocyte base medium 24 hr prior to experiments. Incubated with or without α-Hsp90α or a control human IgG prior to USA300 invasion using an MOI of 30 were 2.5 × 10 5 differentiated keratinocytes. Relative invasion was calculated by normalising USA300 invasion in treated cells to USA300 invasion in cells without treatment (C). The experiments were performed using cells from three independent donors with at least three technical replicates. Error bars indicate standard deviation. The statistical was calculated using one-way ANOVA with multiple comparison with the control (C) sample, *p < 0.05. (d) Pretreated with different concentrations of geldanamycin during 1 hr were 1 × 10 6 cells. Relative invasion was calculated by normalising to invasion in control (C) cells. The experiments were performed in at least three independent biological replicates. Error bars indicate standard deviation. The statistical was calculated by using one-way ANOVA with multiple comparison with the control (C) sample, *p <0 .05, **p < 0.01, ***p < 0.001. USA300, S. aureus USA300. MOI, multiplicity of infection; CFU, colony-forming unit; α-Hsp90α, primary mouse antibody to Hsp90α; α-Hsp90β, primary mouse antibody to Hsp90β; α-Hsp90αβ, primary rabbit antibody to Hsp90αβ
Indeed, F-actin formation could be enhanced by >20%, following treatment of HaCaT cells with Lpl1-his ( Figure 4a). This was confirmed by the addition of α-Hsp90α, which caused a decrease of F-actin formation to the control level ( Figure 4a). These results showed that Lpl1 triggers F-actin formation in an Hsp90-dependent manner.
3.2 | Lpl1-Hsp90α/β interaction has no effect on Hsp90 ATPase activity Next, we investigated if Lpl1 would directly affect Hsp90α ATPase activity. The addition of Lpl1 to a functional Hsp90α did not significantly affect the ATPase activity, whereas the specific inhibitor FIGURE 3 Changes in Hsp90 expression lead to modified USA300 invasion in HaCat cells. (a) Transfected with a commercial siRNA against Hsp90α (siRNA Hsp90α ) were 5 × 10 5 HaCaT cells, a random siRNA (siRNA random ) with or without RNA but with lipofectamin as a control (C) sample. The cells were incubated 24 hr prior to the invasion assay. Relative invasion was calculated by normalising USA300 invasion in treated cells to the control (C) sample. (b) Pre-incubated at 39°C during 2 hr and afterwards incubated with or without α-Hsp90α (35 μg ml -1 ) for 1 hr at 37°C before the invasion assay were 1 × 10 6 HaCaT cells. Control cells were incubated at only 37°C (control: C). Relative invasion was calculated by normalising USA300 invasion in temperature-treated HaCaT cells to USA300 invasion in control (C) cells. All experiments were performed at in least three independent experiments with at least three technical replicates. Error bars indicate standard deviation. The statistical was calculated using oneway ANOVA with multiple comparison with the control (C) sample, *p < 0.05; **p < 0.01; ***p < 0.001. USA300, Staphylococcus aureus USA300; α-Hsp90α, primary mouse antibody to Hsp90α FIGURE 4 Lpl1 interaction with Hsp90α triggers F-actin formation but did not affect ATPase activity. (a) HaCaT cells were seeded into 96-well black microtiter plate for 48 hr. Cells were pre-incubated with or without α-Hsp90α antibodies for 1 hr. Afterwards, 35 μg ml −1 of Lpl1-his protein was added and cells were incubated during 1.5 hr. Factin was stained with ActinGreenTM 488 ReadyProbes® (Thermo Fischer). The amount of F-actin formation in treated cells were determined by the measurement at 495 nm for the excitation and 518 nm for the emission, normalised to the untreated control. Relative Factin formation was calculated to HaCaT cells without treatment (C). (b) ATPase activity was determined in vitro as pmol of phosphate per μg Hps90α per minute and the ATPase activity was normalised to the control (Hsp90α assay). The experiments were performed in three independent experiments with three technical replicates. Statistical significance was calculated by using one-way ANOVA, *p < 0.05; **p < 0.01; ***p < 0.001. F-actin, filamentous actin; α-Hsp90α, primary mouse antibody to Hsp90α; GA, geldanamycin geldanamycin inhibited the ATPase activity by 46% (Figure 4b). These results showed that Lpl1 did not directly affect the ATPase activity of Hsp90.

| Lpl1-his interacts directly with purified Hsp90α and Hsp90β
By using far-western blot assay, we could detect interactions between Lpl1-his and purified Hsp90α and Hsp90β, whereas no interaction was seen when bovine serum albumin (BSA) as a control ( Figure S5). In order to find out which protein domains of Lpl1 are interacting with the Hsp90 proteins, peptides covering the Lpl1 sequences were synthesised and assayed for invasion and F-actin formation (Table S1).
The synthesised peptides covered almost entirely the length of Lpl1 protein. Lpl1 has a conserved "core" region near the N-terminus that shows high similarity with the other Lpl proteins (Nguyen et al., 2015). Because of this similarity, we thought that peptides from the core region might have an effect. However, it turned out that only some C-terminal localised non-core peptides affected internalisation and F-actin formation (Table S1). The peptides P2, P10, and P11 enhanced internalisation and significantly increased F-actin formation (Figure 5a-c). As a control, we showed that α-Hsp90α blocked the stimulating effect of P10 and P11 on invasion (Figure 5a).
We also tested the interaction of Hsp90α with the mentioned peptides that increased S. aureus invasion and F-actin formation using farwestern blot assay. P10 and P11 were found to interact with Hsp90α, as well as with Hsp90β. No interaction with P2 was seen, presumably due to weak binding ( Figure S4). All the other peptides that showed no effect on invasion also exhibited no detectable binding to Hsp90α.

| DISCUSSION
Almost all pathogenic bacteria have developed the ability to directly invade or to trigger invasion into non-phagocytic host cells. Triggering the internalisation of pathogenic bacteria into the host cells is a survival strategy because the human body is constantly patrolled by immune cells and contains antibodies and other molecules that can target and kill the bacteria. Therefore, the environment inside the host cells provides a safe place to avoid detection by the immune system.
There are several reports describing S. aureus surface proteins that interact with specific receptors of the host cell. For example, fibronectin-binding proteins bind via fibronectin to α5β1 integrin (Sinha et al., 2000) or human Hsp60 (Dziewanowska et al., 2000), and Atl binds to the heat shock cognate protein Hsc70 (Hirschhausen et al., 2010). However, the host cell receptor for Lpl lipoproteins was still unclear (Nguyen et al., 2015).
Here, we show that Lpl1, which served as a model   Table  S1) during 1 hr prior to invasion assay using USA300 were 1 × 10 6 HaCaT cells. In some experiments, α-Hsp90α was added and preincubated during 1 hr prior to the peptide incubation. Relative invasion was calculated by normalising USA300 invasion in HaCaTtreated cells to USA300 invasion in untreated cells (C). (b) Preincubated with different peptides of 15 amino acids length (35 μ ml −1 , sequences in Table S1) during 1 hr prior to invasion assay using USA300 were 1 × 10 6 HaCaT cells. Relative invasion was calculated by normalising USA300 invasion in treated HaCaT cells to USA300 invasion in untreated cells (C). (c) F-actin formation. HaCaT were preincubated with or without α-Hsp90α antibodies for 1 hr. Afterwards, 35 μg ml −1 of each peptide was added and cells were incubated for 1.5 hr. F-actin was stained with ActinGreenTM 488 ReadyProbes® (Thermo Fischer). The amount of F-actin formation in treated cells were determined and normalised to the untreated control (C). All the experiments were performed in at least triplicates and in three independent replications. Error bars indicate standard deviation. Statistical significance was calculated by using one-way ANOVA with multiple comparisons with the control (C). *p < 0.05; **p < 0.01; ***p < 0.001. USA300, Staphylococcus aureus USA300; F-actin, filamentous actin; α-Hsp90α, primary mouse antibody to Hsp90α the constitutively expressed Hsp90β (Zuehlke, Beebe, Neckers, & Prince, 2015).
The activities of Hsp90 proteins are diverse and not completely analysed. They are found to be membrane-bound, intracellularly, and are also as secreted form. Hsp90α for example, is membrane-bound but can be secreted in response to tissue injury. A well-characterised function of secreted Hsp90α is to promote cell motility, a crucial event for both wound healing and cancer (Li, Sahu, & Tsen, 2012). The anchoring of Hsp90α and Hsp90β to the plasma membrane involves co-localisation with heparan sulfate proteoglycans (HSPGs) on the cell surface (Snigireva, Vrublevskaya, Afanasyev, & Morenkov, 2015).
Expression of Hsp90α is inducible by oxidative stress and increased temperature. Therefore, we investigated whether increased temperature such as 39°C affects invasion. Interestingly, pretreatment of HaCaT cells at 39°C prior to the invasion assay led to an almost twofold increase in invasion relative to the control at 37°C. This suggests that fever favours S. aureus host cell invasion by upregulating Hsp90α ( Figure S2c). Fever is an evolutionarily conserved response that promotes T-lymphocyte trafficking through Hsp90-induced alpha4 integrin activation and signalling in T cells, thus enhancing immune surveillance during infection (Lin et al., 2019). However, fever induces many factors, among them is the pyrogenic cytokine interleukin-6 (IL-6) which is involved in the mobilisation of lymphocytes to the lymphoid organs that are the staging ground for immune defence (Evans, Repasky, & Fisher, 2015). Although fever increases the migration of T-lymphocytes, some pathogens like S. aureus try to escape the lymphocyte killing by hiding in nonprofessional cells.
At present, we do not know which partners are involved in the Lpl-Hsp90-triggered signalling cascade. Proteins and complexes described as potential clients of Hsp90 are growing constantly and include kinases and receptors (Miyata, Nakamoto, & Neckers, 2013). For example, it is assumed that Hsp90 interacts with the extracellular domain of the Her-2 receptor in the membrane and this interaction triggers signalling in which the cytoplasmic Hsp90 also participates in actin polymerisation (Sidera & Patsavoudi, 2008;Taiyab & Rao Ch, 2011). The Lpl-Hsp90-induced internalisation of S. aureus is probably on the basis of a zipper mechanism in which an intracellular signalling cascade involving the activation of adapter proteins and kinases and the formation of F-actin leading to endocytosis (Colonne, Winchell, & Voth, 2016). Thus, rapid actin polymerisation causes internalisation of the pathogen into non-phagocytic cells.
In addition, we attempted to identify the Lpl1 epitopes that trigger S. aureus internalisation, increased F-actin formation and displayed direct interaction with Hsp90α. We found two overlapping peptides, which consist of 64 and 38 amino acids and both are located at the C-terminal part of Lpl1 ( Figure S5). This finding is reasonable, as this part is the tip of the Lpl proteins and most likely protrudes out of the cell wall. The similarity of the two epitopes to other Lpl-proteins is around 65%.
The lpl gene cluster is localised on a pathogenicity island, termed νSaα. This island is predominant in S. aureus clonal complexes that are spreading worldwide. The key for any additional invasion mechanism lies in the increased ability of a pathogen to hide (and multiply) intracellularly in a safe environment. The internalised cells are protected not only from the host's own innate and adaptive immune response, but also from antibiotics. Interestingly, the number of tandem lpls' is particularly high in S. aureus clones that are known to spread worldwide. For example, the USA300 lineage is spreading rapidly and this lineage is distinguished by a high number of tandem lpl repeats (Carrel, Perencevich, & David, 2015;Tickler et al., 2017). We posit that the Lpls play a role in the rapid spreading of such lineages, as intracellular pathogens are better protected. With the identification of Hsp90 as a receptor for Lpls, we laid the basis for the development of drugs that block Hsp90 mediated invasion.
In conclusion, we propose a model ( Figure 6) in which the membrane-bound S. aureus Lpls interact with the cell surface localised

Hsp90. This interaction triggers a cascade of reactions involving
ATPase in an indirect manner, given that geldanamycin blocked S.
aureus invasion and the ATPase activity and F-actin formation are not affected by Lpl1, resulting in an endocytosis-like engulfment of bacteria.

| Purification of Lpl1-his
Lpl1-his (-sp) named Lpl1-his is an Lpl-1 version that lacks the lipid signal (-sp) and has a his tag at the C-terminal and was isolated from the cytoplasmic fraction of S.aureus SA113 (pTX30::lpl1-his (-sp)), as described previously (Nguyen et al., 2015). Briefly, clones carrying pTX30 were first cultivated in the absence of xylose until OD 578nm of approximately 0.5 was reached. Xylose (0.5%) was added to induce Lpl1-his expression and cultured for 4 hr. Bacterial cells were harvested and the pellet was washed with Tris buffer (20 mM Tris, 100 mM HCl, pH 8.0). Bacterial cells lysis were performed using Tris buffer containing protease inhibitor tablet (Merck, Darmstadt, Germany) and lysostaphin (30 μg/ml, Sigma-Aldrich, Germany) and incubating it at 37°C for 2 hr to disrupt the cell wall. Cytoplasmic fraction was obtained after ultracentrifugation at 235,000 × g for 45 min at 4°C.
The supernatant was incubated with Ni-NTA super flow beads (Qiagen, Germany). After overnight incubation, the NTA beads were intensively washed and eluted with buffer containing 400 mM imidazole. Lpl1 was concentrated via centrifugal ultra-filter unit with a molecular mass cut-off of 10 kDa (Sartorius AG, Göttingen, Germany).
The obtained protein was dialysed using D-Tube Dialyzer Maxi MWCO 6-8 kDa. (Novagen Cod) with DPBS (phosphate buffer saline without calcium or magnesium, Gibco). Finally, the Lpl1-his purification was verified by SDS-PAGE and the total protein amount was determined using a Bradford assay kit.

| Synthesis of Lpl1 derivative peptides
Lpl1 derivative peptides were designed on the basis of the sequence (Nguyen et al., 2015) and structural analysis performed with Phyre2 (Kelley, Mezulis, Yates, Wass, & Sternberg, 2015) and listed in Table   S1. The peptides were synthesised by Apeptide (Shanghai, China) with a purity of >95%. The peptides solution at 1 mg/ml was prepared in water and stored at −20°C.

| Pull-down experiments
Pull-down experiments were performed as described before (Hirschhausen et al., 2010). The elution fraction was precipitated with Strata Clean Resin and consequently analysed by SDS-PAGE. To ensure that proteins from cell lysates do not directly interact with Ni-NTA agarose, a control was performed using the same procedure but without the addition of Lpl1-his. Next, an SDS-page gel was run and stained with Coomasie Blue reagent, and bands present in experimental samples (with Lpl1his) and in the control (without Lpl1-his) were analysed Nano-HPLC-MS/MS analysis.

| Invasion and adherence assays of HaCaT cells
Invasion assay was performed as described previously (Nguyen, Peisl, et al., 2018). For bacterial infection, 2.5 × 10 5 HaCaT cells were seeded into 24-well plates and incubated at 37°C under 5% CO 2 for 48 hr. After 48 hr, cells were washed two times with DPBS and then 1 ml DMEM supplemented with 10% FBS, but without the addition of antibiotics. Afterwards, the cells, in the presence of DMEM supplemented with 10% FBS, were incubated with different antibodies or peptides (Table S1 and Table S2)  In some experiments, cells were pre-incubated for 1 hr with α-Hsp90α, α-Hsp90β, or α-Hsp90αβ or human IgG as a control (Table   S2) in different concentrations (17.5 or 35 μg ml −1 ).
Adherence experiments were performed in the presence of 10% FBS similar to the invasion assays but without the addition of lysostaphin. HaCaT cells were lysed and serial dilutions of the lysates were performed. Ten microliters of the dilutions were seeded on agar plates. Internalised bacteria were not considered because their contribution to adherence is very low. Adhered bacteria were determined as CFU per 10 6 HaCaT cells.

| Hsp90α induction and siRNA experiments
For siRNA-mediated gene silencing, predesigned RNA oligonucleotide targeted against human Hsp90α (siRNA Hsp90α ) or a random siRNA negative control (Silencer™ Negative Control No. 1 siRNA, Thermo Scientific) were used. HaCat cells of 2.5 × 10 6 were seeded in 24-well plates and after 24 hr, cells were transfected with 6 μl of 10 μM dsRNA oligonucleotide or without dsRNA using Lipofectamin RNAiMAX reagent according to the manufacturer's instructions (Invitrogen, Carlsbad, CA). After 24 hr, invasion assay was performed as described above. For heat shock Hsp90α induction, 24-wells plate with 1 × 10 6 HaCaT cells per well was pre-incubated for 2 hr at 39°C prior to the invasion assay.

| Primary human keratinocytes invasion
Primary human keratinocytes were isolated from human foreskin after routine circumcision from the Loretto Clinic in Tübingen as previously described  were incubated with α-Hsp90α or human IgG (17.5 μg ml −1 ) for 1 hr prior to the invasion assay.

| F-Actin measurement
Seeded into black cell culture microplate (Greiner, Germany) were 2.5 × 10 4 HaCat cells in 200 μl for 48 hr prior to incubation with 35 μg ml −1 of Lpl1 or the derivative peptides for 1 hr. In some cases, HaCaT cells were pre-incubated with 35 μg ml −1 of α-Hsp90α antibody for 1 hr prior to the addition of Lpl1 or peptide. F-actin levels were measured as described before (Nguyen, Peisl, et al., 2018), using ActinGreen™ 488 ReadyProbes® (Thermo Fischer). The treated cells were washed with DPBS, permeabilized with 0.25% (v/v) Triton X-100, stained with the dye for 30 min and washed again with DPBS.
Then, the fluorescence was measured at 495 nm for the excitation and 518 nm for the emission using Tecan Reader.

| Immunofluorescence of surface Hsp90
To detect Hsp90α on the cellular surface, an immunofluorescence assay was carried out. Briefly, 2.5 × 10 5 of HaCaT cells were seeded in CellView glass bottom culture dish (Greiner, Germany) for 36 hr.

| Far-western blot experiments
The study of Hsp90α or Hsp90β binding with Lpl1-his proteins was performed by far-western blot assay according to previous study (Nguyen, Peisl, et al., 2018). Briefly, 10 μg of Lpl1-his or BSA (as a negative control) was loaded in a polyacrylamide gel in native conditions.
Proteins were transferred to a PDV nitrocellulose membrane (Bio-Rad, USA) and was blocked with 3% BSA for 1 hr. The blocked membrane was incubated with 20 μg of recombinant Hsp90α or Hsp90β recombinant protein (Abcam) overnight at 4°C. For immunoblotting, monoclonal specific α-Hsp90α or α-Hsp90β antibodies (Abcam) were used as first antibody and goat-α-mouse IgG (Sigma, Germany) as secondary antibody. The detection of the reaction was performed with BCIP®/NBT solution (Sigma, Germany) according to the manufacturer's instructions. In the case of synthetic peptides, 2 μg of each peptide was blotted directly to the PDV nitrocellulose membrane following the same steps described above but with 6 μg of recombinant Hsp90α.
For fibronectin binding proteins (FnBPs) detection, we followed the protocol described previously with modifications (Mongodin et al., 2002). Briefly, cells from S. aureus USA300 wild type and Δlpl strain were lysed with lysostaphin and treated with DNAse. Protein concentration was determinate using Braford assay and 10 μg of proteins were loaded into a polyacrylamide gel in native conditions.
For immunoblotting, the steps described above were followed but the membrane was incubated with 30 μg of fibronectin (Sigma Aldrich, Germany) overnight at 4°C. After extensive washes, the membrane was incubated with monoclonal primary antibody to human fibronectin (clone FN3, eBioscience) overnight at 4°C and with GapA polyclonal primary antibody as a loading control (Nega et al., 2015). Goat-α-mouse IgG (Sigma, Germany) as secondary antibody. The detection of the reaction was performed with BCIP®/NBT solution (Sigma, Germany) according to the manufacturer's instructions.

| Western blot experiments
Western blot experiments were performed using standard techniques.
Briefly, protein concentration in the samples was determined by Bradford assay. Proteins were ran on an SDS-page and then transferred to a PDV nitrocellulose membrane and blocked with 3% BSA.
The blocked membrane was incubated with α-Hsp90α (Thermo Fisher, Table S2) or α-GDPH (Thermo Fisher, Table S2) as a loading control (Thermo Fisher) and goat-α-rabbit IgG or goat-α-mouse IgG was used as secondary antibody. Pre-stained protein ladder (Fermentas) was used as molecular weight marker.

| Flow cytometry assay (FACs)
In a MOI of 30 as described previously, 1 × 10 6 HaCaT cells were exposed to USA300. Cells were washed with DPBS and further incubated for 1 hr with DMEM/F-12, supplemented with lysostaphin and 2 mM of EDTA. Cells were desegregated, centrifuged, and incubated for 30 min with α-Hsp90α in PBS and 1% fetal calf serum (FCS) in 1:100 dilution or with an isotype control (Sigma) or the secondary antibody alone as control. The cells were washed in PBS and 1% FCS and further incubated for 30 min with donkey α-mouse IgG H&L pre-absorbed Alexa Fluor® 488 (Abcam) in 1:1000 dilution. Cells were resuspended in 100 μl of PBS, and 4% paraformaldehyde was added to a final volume of 300 μl. Fluorescence intensity was determined with a BD FACSCalibur. We analysed 10 000 events and the data analysis was performed with FlowJo 10.

| Ethic statements
Keratinocyte isolation from human foreskin was approved by the ethics committee of the medical faculty of the University Tübingen (654/2014BO2) and performed according to the principles of the Declaration of Helsinki.

| Statistical analysis
Student's t tests or one-way analysis of variance (ANOVA) were employed whenever appropriate to compare the difference of means.

ACKNOWLEDGEMENT
We would like thank to Mulugueta Nega for provide the Lpl1 nonrelated peptide. This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG: SFB766 and TR156). PMT is supported by a postdoctoral Alexander von Humboldt fellowship and by CONICET, Argentina. SHF is supported by a PhD fellowship from the German Academic Exchange Service (DAAD) and by the Graduate College GRK1708 (DFG). The funders had no role in study design, data collection and analysis or decision to publish.

COMPETING INTERESTS
The authors declare no competing financial, personal, or professional interests.