Staphylococcal Superantigen-Like Protein 10 (SSL10) induces necroptosis through TNFR1 activation of RIPK3-dependent signal pathways

Nan Jia University of Science and Technology of China, Guo Li Fujian Medical University, Fuzhou Wanbiao Chen University of Science and Technology of China, Chengliang Wang University of Science and Technology of China, Ling Chen University of Science and Technology of China, Xiaoling Ma University of Science and Technology of China, Xuan Zhang University of Science and Technology of China, Yue Tao Jiao Tong University Jianye Zang University of Science and Technology of China, Xi Mo (  xi.mo@shsmu.edu.cn ) Jiao Tong University Jinfeng Hu Fujian Medical University


Introduction
Staphylococcus aureus (S. aureus) is a prevalent and opportunistic pathogen that causes a wide range of diseases such as sepsis, pneumonia, endocarditis, and osteomyelitis, threatening the health of both humans and animals (1). Moreover, S. aureus is among the most clinically challenging pathogens worldwide because of its propensity for rapid development and sharing of antibiotic resistance (2).
Although antibiotic treatments can reduce the case fatality rate of most S. aureus related diseases, some severe infectious diseases still have reported high mortality rates. For example, the case fatality rate for S. aureus bacteremia can range between 15 and 50% (3). However, the mechanisms underlying these poor outcomes for some S. aureus-induced diseases have remained largely unknown.
S. aureus can manipulate host immune response through the expression of a myriad of virulence factors responsible for tissue adherence, immune evasion, cell injury, and organ failure, ultimately promoting its survival and pathogenesis (2,4,5). The activities of some virulence factors have been shown to induce host cell death, especially through apoptosis and necroptosis, which facilitates immune evasion and tissue damage (6, 7). Necroptosis is a programmed form of necrosis that is regulated in a RIPK3 kinase signaling-dependent manner (8). Several types of receptors participate in the initiation stage of necroptosis, including death domain receptors, pathogen recognition receptors, and T cell receptors.
Among these receptors, TNFR1 is well characterized for its role in triggering caspase-independent cell death via activation of RIPK1 and RIPK3 when stimulated by TNFα (9,10). MLKL is an important downstream effector of RIPK3 due to its role in the formation of permeable cell membrane channels that lead to cell death (11). In addition, CaMKII is also phosphorylated by RIPK3, resulting in the opening of mitochondrial permeability transition pores (mPTPs) and subsequent necroptosis in cardiomyocytes, independent of MLKL (12).
Activation of both necroptotic signaling pathways during S. aureus infection has been reported. For example, several of the S. aureus pore forming virulence factors can induce necroptosis in macrophages, which can lead to severe lung damage. Among them, α-hemolysin was demonstrated to instigate necroptosis mediated by MLKL (7). In addition, phagocytosis of S. aureus elicits necroptosis of neutrophils by activating RIPK3 in an MLKL-independent manner (13,14). Moreover, in severe sepsis caused by infection of S. aureus infection, vascular permeation, immunosuppression, and organ failure are usually present, which strongly suggests the occurrence of cell death including necroptosis (5,15,16).
Staphylococcal superantigen like (SSL) proteins comprise a family of 14 member proteins with sequences and structures homologous to superantigen but lacking superantigen activities. The genes encoding SSL proteins are located in a pathogenicity island in the genomes of all tested S. aureus strains (17)(18)(19). SSL proteins adopt conserved structures with an N-terminal OB-fold domain and a C-terminal βgrasp domain, which interact with diverse target factors to manipulate host immune response and interfere with blood coagulation in order to evade host defenses (20). For instance, SSL5 and SSL11 are able to bind sialylated sugar chains of P-selectin glycoprotein ligand 1 (PSGL-1) to inhibit neutrophil activation and rolling (21,22). SSL7 interacts with IgA to block IgA-mediated immunity and binds complement component 5 (C5) to inhibit its activation (23,24). SSL3 targets toll-like receptor 2 (TLR2) to prevent its activation by pathogen-associated molecular patterns (25). In contrast, SSL10 is a novel member of this family with distinct activities to bind CXCR4 and inhibit migration of leukemia cells (26), block the interactions between IgG and complement component C1q that consequently prevent the activation of the classical complement pathway (27,28), and interact with prothrombin and factor Xa to impair blood coagulation (29). Recent publications also show that SSL10 binds to ERK2, phosphatidylserine and apoptotic cells to either interfere host cell in ammation or procoagulant activity (30,31). All these studies suggest that SSL10 possesses multiple functions during S. aureus infection, positing the importance of SSL10.
Although SSL proteins have diverse functions in modulating host response to S. aureus infection, it remains unknown whether SSL family proteins can induce cytotoxicity. In the present study, we demonstrate that SSL10 exhibits potent cytotoxicity towards HEK 293T and HUVEC cell growth by inducing cellular necroptosis and contributes to the cytotoxicity induced by S. aureus. We also propose an underlying mechanism for induction of necroptosis through binding with TNFR1, determined by crystal structure and modeling of protein binding interactions, and activating downstream RIPK3 signal pathways. This work provides evidence that SSL10 is a cytotoxic virulence factor that may serve as a therapeutic target in S. aureus infections.

SSL10 induces cell necrosis
To determine the effects of SSLs on cells, human umbilical vein endothelial cells (HUVEC) were treated with puri ed recombinant SSLs protein, including SSL3, SSL7, SSL8, SSL10 or SSL11 for different time periods. As determined by MTS assay (Fig. S1), SSL10, but not other SSL family proteins we tested, signi cantly reduced the cell activity of HUVEC. Further, SSL10 treatment resulted in a decrease in cell viability in a dose-and time-dependent manner in human embryonic kidney cells (HEK 293T) and HUVEC, as determined by MTS assay (Fig. 1A and 1B). Most inviable cells were PI-positive, but were Annexin Vnegative when detected by ow cytometry, suggesting that cell death induced by SSL10 is most likely to be necrosis ( Fig. 1C and 1D). Cell necrosis was further con rmed by transmission electron microscopy (TEM) in which the cells exhibited a typical necrotic phenotype, including cytoplasmic lightening, swollen organelle, and membrane rupture (Fig. 1E). SSL10 is a virulence factor secreted by S. aureus, so we quanti ed the cytotoxicity of supernatants from ssl10 knock out or complementation strain toward HEK 293T and HUVEC. We found that, compared with TSB medium-treated cells, LDH release was signi cantly greater in treatments with WT S. aureus 8325 supernatant. In contrast, supernatant from the ssl10 knock out strain, but not the ssl10 complemented strain, induced signi cantly decreased LDH release (Fig. 1F).
To further demonstrate that SSL10 induced necrosis rather than apoptosis, cells were treated by pancaspase inhibitor Z-VAD-fmk before exposure to SSL10, which resulted in no signi cant inhibition of cell death, as determined by LDH release (Fig. 1G and 1H). As pyroptosis is also caspase dependent, our data thus suggest that SSL10 does not induce apoptosis or pyroptosis, but instead strongly implies activity by the primary cellular necrosis pathway. SSL10 induces necroptosis via the RIPK3-dependent pathway To explore the underlying mechanism by which SSL10 induced necrosis in HEK 293T and HUVEC cells, we next used different inhibitors to pretreat the cells prior to SSL10 exposure. As determined by LDH release, RIPK1 inhibitors (Nec-1 and Nec-1s) could only partially attenuate the effects of SSL10, while RIPK3 inhibitor (GSK'872) almost completely inhibited the necrotic effects of SSL10 ( Fig. 2A and 2B). Thus, we speculated that SSL10 may induce cellular necroptosis. To further test this hypothesis, we generated knock out cell lines for key genes involved in the necroptosis pathway via CRISPR-Cas9 in both HEK 293T and HUVEC cells (Fig. S2). Consistent with the effects of inhibitor treatment, knock out of RIPK3 but not RIPK1 or MLKL inhibited SSL10-induced necrosis ( Fig. 2C and 2E) in HEK 293T cells. In addition, transient complementation with RIPK3 in RIPK3 −/− HEK 293T cells led to robust necrosis, evident by the release of LDH (Fig. 2F). Similarly, substantially less LDH was released in HUVECs knocked out for RIPK3 (Fig. 2D), indicating that SSL10 could induce RIPK3-dependent necroptosis in both HEK 293T and HUVECs.
CaMKII activation and mPTP opening also contribute to SSL10-induced necroptosis Previous studies have reported that RIPK1 can form a complex with RIPK3, which further activates MLKL, resulting in necroptosis of several types of cells (10,32). However, in the present study, we found that inhibition or knock out of RIPK1 or MLKL could not completely inhibit SSL10-induced necroptosis, suggesting that SSL10-induced necroptosis may also depend on other RIPK3-mediated pathways independent of RIPK1 and MLKL.
In addition to MLKL, RIPK3 has been reported to phosphorylate CaMKII to induce the opening of mPTP channels, for example, leading to necroptosis in cardiomyocytes (12). To explore whether CaMKII is also involved in SSL10-induced necroptosis, HEK 293T and HUVEC cells were treated with KN-93, a selective inhibitor of CaMKII, prior to SSL10 treatment. As assessed by the release of LDH and ATP, inhibition of CaMKII profoundly abrogated SSL10-induced necroptosis ( Fig. 3A-C); phospho-CaMKII levels were also signi cantly increased after SSL10 treatment (Fig. 3D), which suggested the involvement of CaMKII in SSL10-induced necroptosis. To further identify the downstream effector of CaMKII, we pretreated HEK 293T and HUVEC cells with an inhibitor of mPTP opening, CsA, which e ciently blocked SSL10-induced LDH release ( Fig. 3E and 3F). In addition, SSL10 treatment led to mitochondrial depolarization, which was signi cantly hampered in the absence of RIPK3, evident by the decrease in mitochondrial membrane potential (∆Ψ m ) (Fig. 3G), indicating that CaMKII-mPTP is also likely to be a primary candidate downstream pathway for RIPK3 in SSL10-induced necroptosis.

SSL10 induces necroptosis by direct interaction with the TNFR1 extracellular domain (TNFR1 ECD )
Necroptosis is initiated through ligand binding to several receptors including TNFR1 (33)(34)(35). To explore whether SSL10 induces necroptosis by interacting with membrane receptors, SSL10 localization was observed by real-time live-cell analysis and scanning confocal microscopy. Notably, SSL10 was found to be enriched on the cell membrane within the rst 30 min of treatment, suggesting that SSL10 may bind to a cell surface receptor (Fig. S3). TNFR1, as well as other TNF family death receptors, control necroptosis through activating RIPK1 and RIPK3 (36). To test whether TNFR1 was the receptor for SSL10, in vitro MBP pull-down assays were conducted using puri ed SSL10 and MBP-tagged TNFR1 ECD (the extracellular domain of TNFR1 containing amino acids 22-211), which showed that SSL10 can indeedly interact with TNFR1 ECD (Fig. 4A).
To further con rm whether TNFR1 participates in SSL10-induced necroptosis, we knocked out TNFR1 in HEK 293T and HUVEC cells via CRISPR/Cas9, which blocked SSL10-induced necroptosis, as indicated by the signi cantly decreased release of LDH ( Fig. 4B and 4C). We then con rmed that the essential role of TNFR1 in SSL10-induced cytotoxicity by ow cytometry of increased viable cell counts and decreased ∆Ψ m (Fig. 4D and 4E). Consistent with these ndings, SSL10-binding to the HEK 293T cell surface was signi cantly reduced when TNFR1 was knocked out (Fig. 4F). Taken together, these data demonstrate that SSL10 activates cell necroptosis via direct interaction with the TNFR1 ECD .
Overall structure of SSL10 To further understand the molecular mechanisms driving the SSL10 activation of necroptosis through initiation of the TNFR1 signaling pathway, we next solved the crystal structure of SSL10 by molecular replacement at 1.9 Å resolution. X-ray diffraction data and structure re nement statistics are shown in Table 1. SSL10 exists as a monomer in both solution and crystal lattice (Fig. S4). Two SSL10 molecules were observed in one asymmetric unit adopting approximately identical structures, with the RMSD value being 0.254 Å when the two molecules are aligned (Fig. S5A). In light of these results, we select molecule B for further investigation. SSL10 exhibits a typical superantigen-like structure, similar to other SSL family proteins, consisting of two distinct domains separated by a exible linker region. The N-terminal OB-fold domain (residues 43-123) contains one α-helix, eight β-strands, and one 3 10 helix, while the Cterminal β-grasp domain (residues 133-227) consists of one α-helix, seven β-strands, and two 3 10 helices ( Fig. 5A). Several, large, positively charged surface areas were identi ed in the SSL10 structure, whereas negatively charged regions were small (Fig. 5B). Both the N-and C-terminal domains of SSL10 contribute to its cytotoxicity Among the SSL family proteins, SSL7 is the most similar member of SSL10, with the highest sequence identity to SSL10 and the RMSD value being 1.345 Å when the structure of SSL10 was aligned to SSL7 (PDB code: 3KLS) (Fig. S5B).
To investigate which domain or domains of SSL10 may be critical for its cytotoxicity, we generated variants of SSL10 with the N-and C-terminal domains swapped between SSL7. We designated the two newly generated chimeric proteins as SSL7⋅10 and SSL10⋅7, with SSL7⋅10 containing the SSL7 Nterminus and the SSL10 C-terminus, and vice versa (Fig. 5C). We found that both of the two chimeric proteins could induce a marked release of LDH, which was less potent than SSL10, and SSL7 didn't induce any extra LDH release compared with the buffer-treated cells (Fig. 5D). In agreement with these results, MBP pull-down assays showed that both of the chimeric proteins, but not SSL7, could bind to the TNFR1 ECD , and the binding of both proteins was weaker compared to that of SSL10 (Fig. 5E), indicating that both the N-and C-terminal domains participate in SSL10-induced necroptosis via interaction with TNFR1 ECD .
Potential binding site of SSL10 for TNFR1 To understand the molecular mechanisms controlling SSL10 binding to TNFR1, the structure of TNFR1 ECD (PDB code: 1EXT) was docked onto the structure of SSL10 using the HDOCK web server (http://hdock.phys.hust.edu.cn/) (Fig. 6A). In this model, eight residues including H64, K66, N85, S88, Q91, K206, K208, and Y209 of SSL10 suggested a potential binding region for TNFR1 ECD (Fig. 6A and  6B). Among these residues, H64, K66, N85, S88, and Q91 are located in the N-terminal OB-fold domain, while the other three residues are found in the C-terminal β-grasp domain, consistent with the previous observation that both the N-and C-terminal domains of SSL10 contribute to its cytotoxicity (Fig. 5D). We then aligned the sequences of SSL10 and SSL7, which did not interact with TNFR1 in MBP pull-down assays, in order to investigate differences in the eight residues between the two proteins (Fig. 6C). In SSL7, the corresponding residues are N68, S70, K89, D92, K95, Q208, E210, and R211, respectively, which differed from the charge of those amino acids in SSL10.
To further verify the binding model of SSL10 to the TNFR1 ECD , we generated two mutant variants, by replacing the eight residues H64, K66, N85, S88, Q91, K206, K208, and Y209 in SSL10 with either alanine residues (mutant A) or with the corresponding residues from SSL7 (mutant B). The results of LDH assays showed that both mutants exhibited a signi cant reduction in LDH release compared with that in SSL10, and with mutant B having the weakest cytotoxicity (Fig. 6D). Supporting these results, binding by either mutant to TNFR1 ECD was weaker than that of SSL10 in MBP pull-down assays, nearly 67% and 49% compared to SSL10, respectively ( Fig. 6E and 6F). Therefore, the docking and mutagenesis analyses demonstrate that residues H64, K66, N85, S88, Q91, K206, K208, and Y209 are critical for SSL10 binding with TNFR1 ECD and subsequent initiation of the necroptosis signal cascade.

Discussion
Previous studies on SSL10, one of the SSL proteins speci cally expressed in S. aureus, suggested that it contributed to S. aureus infection though inhibiting the classical complement activation pathway, the migration of T cells, the interaction between complement C1q and IgG, and the Fc-receptor-mediated phagocytosis of neutrophils (26-29). In addition, recent publications demonstrated that SSL10 binds to ERK2, phosphatidylserine and apoptotic cells to either interfere host cell in ammation or procoagulant activity (30,31). Here, our data demonstrated the cytotoxicity of SSL10 by triggering necroptosis via activation of two distinct signaling pathways by binding to TNFR1 in HET 293T and HUVEC cells.
Though RIPK3 is expressed at an extremely low level in endothelial cell, including HUVEC, and HEK 293T, its importance in vivo cannot be excluded (37,38). Genetic evidence showed that RIPK3 de ciency leads to reduced endothelial cell permeability or necroptosis, thereby suppressing tumor metastasis (39,40). Furthermore, RIPK3 can be induced or upregulated under certain conditions, which confers cells sensitive to RIPK3-dependent necroptosis (39). Notably, we found an SSL10-induced increase in the protein level of RIPK3 in HEK 293T and HUVEC cells (Fig. S6), indicating the important role of RIPK3 on SSL10-induecd necroptosis in HEK 293T and HUVEC.
The best-characterized mechanism of necroptosis is RIP1-RIP3-MLKL signal pathway, which is induced by the interaction of TNF-α and its receptor TNFR1 (32). More recently, CaMKII has been identi ed as a direct substrate of RIP3 to appropriate downstream effector mPTP in myocardial necroptosis (12).
Opening of the mPTP results in loss of mitochondrial inner membrane potential, disruption of ATP production, increased ROS production, organelle swelling, mitochondrial dysfunction and consequent necrosis (41). Notably, the upstream receptor involved in RIPK3-CaMKII pathway is still unknown. Here, we found that TNFR1 contributed to RIPK3-CaMKII pathway induced by SSL10, but the mechanism underlying how TNFR1 activates RIPK3-CaMKII needs further study.
X-ray crystallography revealed SSL10 possess the characteristic structure of SSL family proteins, which consist of an N-terminal OB-fold and C-terminal β-grasp domain (42,43). The OB-fold domain is responsible for recognition of protein ligands, whereas the β-grasp domain is capable of binding to tetrasaccharide sialyl Lewis X (22,23,25,(42)(43)(44)(45)(46). For example, SSL7 binds to the human IgA1 Fc domain through the N-terminal α helix and the L1 and L4 loops of the OB-fold domain, which results in steric shielding of the FcαRI binding site and inhibits FcαRI-mediated immunity (23). In addition, SSL3 interacts with TLR2 (to prevent its lipopeptide binding and dimerization) via four loops localized in the OB-fold domain, leading to blockade of TLR2 signaling and immune evasion (25). However, different from those previously reported SSL proteins, both the L1 loop and 3 10 1 helix in the OB-fold domain, as well as 3 10 2 helix in β-grasp domain of SSL10 participate in forming a surface for TNFR1 binding, which contributes to the activation of TNFR1 signaling-mediated necroptosis (Figs. 5 and 6). A recent report showed that two sequences from both the N-and C-terminal domains of SSL10 were involved in binding to prothrombin (47). Mapping onto the structure of SSL10 showed that the L1 loop and 3 10 2 helix are localized in these two regions. According to these observations, SSL10 likely employs a different mechanism for binding partner recognition than that of other SSL proteins. Notably, further studies, informed by the work reported here, are needed to solve the complex structure of SSL10-TNFR1 ECD , so that we can identify the exact binding sites and key residues.  (55). Above, in combination with our data, we suggest that SSL10 may be involved in the progress of S. aureus sepsis by different mechanisms.
In summary, we provide the rst demonstration of which we are aware that SSL10 can act as a signal to initiate necroptotic programmed cell death signal via direct interaction with TNFR1 in HEK 293T and HUVEC. Moreover, this signal cascade is activated in a RIPK3-dependent manner but is transduced through two independent signaling pathways (Fig. 7). Using molecular docking with the structure of TNFR1 ECD to SSL10, we identi ed a surface region of SSL10 as the potential binding site for TNFR1.
Thus, this study, in combination with other previous reports, provides strong evidence that SSL10 contributes to S. aureus infection via multiple mechanisms, and suggest this virulence factor may serve as a potentially reliable biomarker and therapeutic target for S. aureus-associated infection and diseases.

Materials And Methods
Reagents and cell culture

Cytotoxicity Assays
HUVECs or HEK 293T cells were seeded in 96-well plate (100 µL per well) at a density of 1 × 10 4 cells/well one day prior to treatment with or without SSL10 resuspended in Opti-MEM reduced serum medium (Thermo Fisher) for 48 h. To detect the effects of inhibitors, the cells were pretreated with various inhibitors 0.5 h before 2 µM SSL10 treatment.
For MTS assay, 20 µL of CellTiter 96® AQueous One Solution reagent was added to each well and incubated for two more hours, and the optical density was measured at 490 nm with a BioTek Synergy/2 microplate reader (BioTek, Winooski, VT).
LDH release was measured according to the manufacturer's manual. Brie y, 50 µL culture media from various treated cells were transferred to a new 96-well at clear bottom plate, and 50 µL of the CytoTox 96® reagent was added to each sample aliquot and incubated in dark for 30 min at room temperature. Finally, 50 µL of stop solution was added to each well and the absorbance at 490 nm was recorded with BioTek Synergy/2. The level of LDH released was expressed as a fold of the control (buffer-treated cells) group after subtracting the background absorbance.
To determine the ATP concentration in the cells, CellTiter-Glo® reagent was added to each well and the plate was incubated for 10 min to stabilize luminescent signal before luminescence being recorded with BioTek Synergy/2. Luminescent signals from blank wells and buffer-treated cells were used as background and maximal luminescence.
Cell death detected by ow cytometry was performed as previously described (56). Brie y, 48 h after treatment with SSL10, the cells were collected and washed twice with ice-cold PBS. The cells were then incubated with 5 µL FITC annexin V and 5 µL PI in 500 µL prepared assay buffer in dark for 10 min at room temperature, and applied for ow cytometry analysis.
Knockout and rescue of ssl10 in S. aureus 8325 To explore the effect of SSL10 on the cytotoxicity of S. aureus, ssl10-knockout strain was constructed using the vector pKOR1 as previously described (57). Brie y, ~ 1, 000 bp fragment upstream and downstream, respectively, of ssl10 was cloned to pKOR1 via lambda recombination (BP clonase enzyme mix, Invitrogen). The resulting plasmid was transferred via electroporation rst to S. aureus RN4220 to modify DNA, and subsequently to S. aureus 8325. For homologous recombination and ingratiation of pKOR1 into the bacterial chromosome, S. aureus 8325 was grown at 43°C on tryptic soy agar (TSA Cm10 ), a non-permissive condition for pKOR1 replication. From the resulting plate, one colony was picked, inoculated into 1 ml TSB Cm10 and incubated at 30°C overnight to facilitate plasmid excision. Cultures were then spread on TSA containing 200 ng/mL anhydrotetracycline and incubated at 30°C overnight for selecting ssl10-knockout S. aureus 8325. To rescue ssl10 in ssl10 knockout strain, the gene sequence of ssl10 was ligated with hprk promoter and then cloned to the plasmid pOS1. The resulting plasmid was transferred via electroporation rst to S. aureus RN4220 to modify DNA, and subsequently to ssl10knockout S. aureus 8325.
Cytotoxicity Assay of S. aureus ssl10-knockout or Rescue Strains S. aureus 8325 wild type, ssl10-knockout and rescue strains were cultured in TSB medium (OXOID) for 8 h at 37°C, 220 rpm. The same number of bacteria cells were inoculated into a new tube of TSB medium respectively, and cultured overnight at 37°C, 220 rpm for protein expression and secretion. Bacteria cells were collected after achieving stagnate phase and the supernatants were obtained by centrifugation at 5, 000 rpm for 15 min at room temperature. After ltration with a 0.22 µm lter, the supernatants were used to treat HEK 293T or HUVEC cells for 48 h after 1:1, 000 dilution with Opti-MEM reduced serum medium, and LDH released from the cells were determined. The level of LDH released was expressed as the fold of the control group (TSB medium-treated cells) after subtracting the background absorbance.   Table S1.

Mitochondrial Membrane Potential Assay
The mitochondrial potential, which re ects mitochondrial depolarization, was detected using the mitochondrial membrane potential assay kit following the manufacturer's instructions (Beyotime, Shanghai, China). Brie y, JC-1 working solution was incubated with the cells in dark for 20 min at 37°C. After three washes with prepared buffer, the cells were detected on a ow cytometer using PE and FITC channels. The value of the JC-1 monomers to aggregates positive cells ratio quanti es the mitochondrial membrane depolarization.

Real-time Live-cell Analysis
HUVECs were seeded in 96-well plate (100 µL per well) at a density of 1 × 10 3 cells/well with or without 2 µM GFP-SSL10/GFP treatment and then observed with live-cell dynamic imaging and analysis system (Incucyte S3, Essen Bioscience, USA).

Laser Scanning Confocal Microscope
HUVECs were seeded on coverslips in a 24-well plate and then incubated for 4 h before being incubated with 2 µM GFP-SSL10 protein resuspended in Opti-MEM medium for 20 min. After treatment, the cells were washed twice with PBS to remove unbound proteins and then xed with 4% paraformaldehyde for 20 min. After xation, cells were washed three times with PBS and then incubated with DAPI. After three additional PBS washes, coverslips were mounted onto slides using antifade mounting medium (Beyotime, Shanghai, China). The cells were visualized using a laser scanning confocal microscope (Leica, Wetzlar, Germany).

Protein Expression and Puri cation
DNA fragments encoding amino acid residues 31-227 of SSL10 or its mutants were ampli ed by PCR from S. aureus strain Mu50 and cloned into the pET-22b (+) vector (Novagen) with a C-terminal 6×His tag. Wild type and mutant SSL10 were expressed in Escherichia coli BL21 (DE3) and induced with 0.4 mM IPTG (isopropyl β-D-1-thiogalactopyranoside) for 4 h at 37 ℃ when OD 600 reached 0.6. The cells were harvested by centrifugation at 6, 000 rpm for 8 min, and lysed in a French press in lysis buffer [50 mM Tris-HCl, pH7.5, 500 mM NaCl, 5% (v/v) Glycerol, 5 mM imidazole, 1 mM PMSF]. The lysate was centrifuged at 15, 000 rpm for 30 min, and the supernatant was incubated with Ni-NTA resin for 30 min.
The eluted protein was concentrated and further puri ed by Superdex 75 10/300 size exclusion column (GE Healthcare) equilibrated with the buffer containing 20 mM Tris-HCl, pH 7.5 and 200 mM NaCl. The entire protein puri cation procedure was carried out at 4°C. Purity of the target protein was veri ed by SDS-PAGE and protein aliquots were stored at -80°C for further use.
GFP-SSL10 used for laser scanning confocal microscopy was cloned into the pET-28a(+) vector (Novagen) with the GFP-and 6×His-tag fused to its N-terminus. The protein was expressed and puri ed using the method similar as above. The crystals of SSL10 were soaked in cryoprotectant buffer consisting 2.1 M DL-Malic acid, pH 7.0 and 20% Glycerol for several seconds and ash-cooled in liquid nitrogen. X-ray diffraction data was collected at beamline BL18U1 of Shanghai Synchrotron Radiation Facility (SSRF). Diffraction data were processed, integrated, and scaled using HKL2000 (58).
The crystal structure of SSL10 was determined by molecular replacement using the program Phaser in the CCP4i suite (59, 60) with Exotoxin SACOL0473 (PDB code 3R2I) as the search model. After several runs of structure re nement using the programs REFMAC5, Phenix and Coot (61-63), the nal model was re ned to 1.9 Å resolution with R work of 20.94% and R free of 24.62%. Data collection and structure re nement statistics are summarized in Table 1. All gures of protein structure were prepared using PyMOL (http://www.pymol.org).

MBP Pull-down Assay
Thirty microgram of MBP-TNFR1 ECD was incubated with 30 µg of wild type or mutant SSL10 for 1 h on ice, and the protein mixture was centrifuged at 15, 000 rpm for 30 min at 4°C to remove precipitates. The

Statistical Analysis
All data were collected from at least three independent experiments, and presented as the mean ± SD.
Comparisons between two groups were analyzed by the multiple t-test or by two-way ANOVA using GraphPad Prism Version 6.0 (GraphPad Inc., La Jolla, CA, USA) software program. p values < 0.05 were considered statistically signi cant.   and the results of quanti cation are shown as a bar graph (right). All data represent the means ± SD calculated from at least three independent experiments. *, p < 0.05; **, p < 0.01; ***, p < 0.001 compared to those without SSL10 treatment (-SSL10-treated cells, i.e., buffer-treated cells) under the same conditions. #, p < 0.05; ##, p < 0.01 compared to those treated with 2 μM SSL10 for 48 h under the same conditions. 293T cells for 30 min, and SSL10 bound to the cell surface was detected by ow cytometry using FITCconjugated anti-His-tag antibody binding. The absence of TNFR1 signi cantly reduced binding of SSL10 to the cell surface. All data represent means ± SD calculated from at least three independent experiments. *, p < 0.05; **, p < 0.01; ***, p < 0.001 compared to those without SSL10 treatment (-SSL10-treated cells,