Peptidoglycan mediates Leptospira outer membrane protein Loa22 to toll-like receptor 2 for inflammatory interaction: a novel innate immune recognition

Leptospirosis is an overlooked zoonotic disease caused by pathogenic Leptospira depended on virulence of Leptospira and the host–pathogen interaction. Kidney is the major organ infected by Leptospira which causes tubulointerstitial nephritis. Leptospira outer membrane contains several virulence factors and an outer membrane protein A (OmpA) like protein (Loa22) is essential for virulence. Pull-down assays suggested that Loa22 was a potential Toll-Like Receptor 2 (TLR2) binding candidates from pathogenic Leptospira. Confocal microscopy was employed to observe the co-localization of TLR2 and Loa22-LPGN (Leptospira peptidoglycan) complexes. Atomic force microscopy (AFM), side-directed mutagenesis, and enzyme-linked immunosorbent assay (ELISA) were performed to investigate the affinity between rLoa22, LPGN, and TLR2. Real time PCR was applied to measure the cytokines expression. Downstream signal transduction components were verified by western blot to evaluate the gene regulations. Mutation of two Loa22 key residues (Asp122 and Arg143) attenuated the affinities for LPGN. rLoa22-LPGN complexes were observed to co-localize with TLR2 and provoked inflammatory responses including CXCL8/IL8, hCCL2/MCP-1, and hTNF-α. Affinity studies suggested that Loa22-LPGN complexes elevated the affinity to TLR2 as compared to Loa22 protein. Downstream signals from TLR2 including p38, ERK, and JNK were regulated under rLoa22-LPGN complexes treatments. This study identified LPGN mediates interactions between Loa22 and TLR2 and induces downstream signals to trigger inflammatory responses. rLoa22-LPGN-TLR2 complexes reveal a novel binding mechanism for the innate immune system.


Results
Identification of TLR2 binding candidates from pathogenic Leptospira. Upon infection by pathogens, innate immune responses are induced to defense against the infection on host cell surfaces. We attempted to identify the TLR2 ligands from pathogenic L. santarosai serovar Shermani and characterize the binding mechanisms of these virulence factors with TLR2. The human TLR2 gene was sub-cloned from plasmid pUNO-TLR2 (Invivogen, San Diego, CA) and inserted into a lentivirus expression vector with a V5 tag at the C-terminus. The packaged virus particles were used to infect to HEK-293 T cells, and a stable clone was selected using blasticidin for over-expression of the full-length human TLR2 protein. HEK-293 T-TLR2 cells were used for full-length human TLR2 protein expression, and protein A-immobilized anti-V5 antibody was used for human TLR2 protein pull-down assays. After incubation of Leptospira outer membrane extractions with HEK-293 T-TLR2 cells for two hours, the cells were lysed and protein A-immobilized anti-V5 antibody was used to pull-down the TLR2 and binding candidates. The pull-down fractions were analyzed by western blot, and proteins were recognized with relative antibodies. Several TLR2 binding candidates were isolated and one of the positive controls, LipL32, was observed by anti-LipL32 antibody recognition (Fig. 1A) 22,23,25,26 . This result suggested that the method used for identification of TLR2-binding candidates searching and identification is suitable. An interesting TLR2 binding candidate, Loa22, was observed in the anti-Loa22 antibody recognition after co-immunoprecipitanting with TLR2. Western blot clearly demonstrated the interaction of Loa22 and TLR2 after co-immunoprecipitantion (Fig. 1A). Loa22 was present in pathogenic Leptospires but not in non-pathogenic Leptospires, indicating that Loa22 protein is probably a virulence factor (Fig. 1B) 10 . Loa22 is anchored to the outer membrane of pathogenic Leptospires and contains a large OmpA domain, known as a peptidoglycan-binding domain (Fig. S1A). Therefore, recombinant Loa22 (rLoa22) was constructed and expressed in E. coli to obtain purified rLoa22.
Protein purification and mutagenesis. The Loa22 protein contains 195 amino acids, and domain prediction indicated the N-terminal signal peptide and C-terminal OmpA domain (Fig. S1A). Sequence alignments PGN binding assay. Loa22 is a lipoprotein with a C-terminal OmpA domain, which is speculated to bind the essential cell wall component, PGN. To verify PGN-binding activity of rLoa22, AFM was used to investigate the interaction between rLoa22 and LPGN. The Leptospira was immobilized on a mica surface and washed three times with PBS buffer containing 0.1% (w/v) Triton X-114 to remove the outer membrane and expose the PGN layer (Fig. S2A,B). The rLoa22-modified AFM tip was used to measure the affinity between rLoa22 and Leptospira cell wall. AFM force-distance curves were recorded to distinguish specific and non-specific interactions ( Fig. 2A). The specific interaction force-distance curves were selected to analyze interactions between rLoa22  After three steps of centrifugation and washes, the pellets were subjected to SDS-PAGE and western blot analysis. The results indicated that rLoa22 protein showed relative high affinity for LPGN (Fig. 2C). An internal control of the purified LPGN was used to recognize by anti-Loa22 antibody and the result indicated that the purified LPGN contained no or less Loa22 in the purification process (Fig. 1B,  www.nature.com/scientificreports/ pathogenic Leptospira with high affinity for rLoa22, while other PGN molecules exhibited relatively low affinity for rLoa22 protein (Fig. 2C). Therefore, LPGN was selected for subsequent studies.
rLoa22 Co-localizes with TLR2 on HEK293-TLR2 Cells. Attachment of Leptospira outer membrane proteins to host cell membrane is the first step invading the host during Leptospira infection. Previous studies showed that the Loa22 is up-regulation when host infection induces high levels of antibody production in the infected patient's serum 10,27 . However, the receptor on host cell membrane which recognizes Loa22 protein is still unknown and needs further investigation. The results mentioned above suggested that TLR2 is a possible receptor on host cell membranes for Loa22. In order to demonstrate the co-localization of rLoa22 and TLR2, purified rLoa22 protein and its variants were incubated with HEK293-TLR2 cells for 4 h and the cells were then washed, fixed, and incubated with conjugated antibodies for confocal microscopy analysis (Fig. 3). Besides, a positive control, Pam 3 CSK 4 Rhodamine, was also used to incubate with HEK293-TLR2 cells to observe the colocolization behavior (Fig. 3E) 22 . rLoa22 and rTLR2 proteins were stained with rabbit polyclonal anti-Loa22 and mouse monoclonal anti-V5 primary antibodies follow by Alexa594 (red) conjugated anti-rabbit and Alexa488 (green) conjugated anti-mouse secondary antibodies, respectively. HEK293 cells lacking TLR2 expression were used as negative controls, with very little or no Alexa 488 fluorescence (Fig. 3A). Small amount of the Alexa 594 fluorescence revealed that the Loa22 bind to the HEK293 cell and this result is consisted with previous study that the Loa22 could bind the ECM molecules 12 . Co-localization of rLoa22WT-LPGN complexes and TLR2 Rhodamine was incubated with HEK293-TLR2 cell. The nucleus was stained by DAPI (blue) and TLR2 was stained by Alexa488 (green). Loa22 was stained by Alexa594 (red). The yellow color indicated that the two proteins were co-localized in HEK293-TLR2 cell. www.nature.com/scientificreports/ receptors on HEK293-TLR2 cell was shown in Fig. 3B. The results indicated that TLR2 receptor and the rLoa22 protein were mostly present on the cell surface, with consistent partial localization in the cytosol. The merged colors in several portions indicated that the two proteins were co-localized on HEK293-TLR2 cells (Fig. 3B). In additions, the PGN molecules from other species including E. coli (EPGN), S. aureus (SPGN), and B. subtilis (BPGN) were used to test the ability to facilitate the interaction between rLoa22 and TLR2 on cell surface. The results indicated that the EPGN, SPGN, and BPGN could not mediate the interaction between rLoa22 and TLR2 ( Fig. S3). In contrast, the two mutated variants reduced the cell binding ability, and the red color was absent in the confocal images (Fig. 3C,D). The results from confocal microscopy clearly showed rLoa22-LPGN complexes directly interacted with TLR2 on HEK293-TLR2 cell surface, while the mutated variants, rLoa22D122A-LPGN and rLoa22R143A-LPGN, of rLoa22 significantly decreased co-localization with TLR2 on the cell surface.
Interaction between TLR2 and rLoa22-LPGN complexes. The purified TLR2 protein was used to measure the interaction between TLR2, rLoa22, and LPGN molecule. In ELISA assays, the LPGN molecule from pathogenic Leptospira showed high affinity to TLR2 protein and the affinity between LPGN and TLR2 significantly increased as compared to BSA control (Fig. 4A). The purified rLoa22WT slightly increased the affinity to TLR2. Interestingly, the rLoa22WT-LPGN complex significantly increased the affinity to TLR2 as compared to BSA control, LPGN, and rLoa22WT, respectively. The results indicated that LPGN molecule might played essential roles in rLoa22 and TLR2 affinity. In the presence or absence of LPGN, the affinities between TLR2 and the two mutation variants (rLoa22D122A and rLoa22R143A) showed similar to BSA control. The affinity between TLR2 and the mutation variants in complex with LPGN showed significantly decreased as compared to rLoa22WT-LPGN. The results demonstrated that two residues (Asp 122 and Arg 143 ) of rLoa22 were essential for the affinity between TLR2 and rLoa22-LPGN complex. In AFM measurements, the interaction force and binding frequency of rLoa22WT to TLR2 were slightly increased as compared to BSA control ( Fig. 4B,C). Interestingly, the interaction force and binding frequency of rLoa22-LPGN complexes to TLR2 showed significantly increased as compared to BSA control ( Fig. 4B,C). These results provided direct evidence that LPGN cooperated with rLoa22 to interact with TLR2. In order to identify which components of rLoa22-LPGN complex were responsible for interaction with TLR2, we used anti-rLoa22 antibody to block rLoa22 in rLoa22-LPGN complex and further treated with rTLR2 protein by using ELISA and AFM. The results indicated that anti-rLoa22 antibody efficiently reduced the affinities between rLoa22-LPGN complex and rTLR2. This result supported our hypothesis that Loa22 interacted with LPGN and induced conformational changes in the protein, which exposed the TLR2-binding domain of Loa22 to interact with TLR2. In addition, the interaction force and binding frequency of TLR2 to rLoa22 mutated variants (D122A and R143A) showed no significantly differences as compared to BSA control. It is not surprising that the rLoa22 mutation variants were loss of function variants that showed significantly decreased interaction force and binding frequency as compared to rLoa22WT-LPGN ( Fig. 4B,C). Further addition of the LPGN molecules to these two variants could not raise the affinity to TLR2.
rLoa22-LPGN complexes induced p38, ERK, and JNK dependent signaling. Leptospira infection induces inflammatory responses through the TLR2-dependent pathway, and downstream signalings were therefore evaluated. The activation of the MAPK pathway was validated by western blot after treatment of HK2 and HEK293-TLR2 cells with rLoa22-LPGN complexes. The phosphorylation of MAPK pathway components including p38, ERK, and JNK was observed using their relevant antibodies. HK2 and HEK-293-TLR2 cells were cultured in serum free medium for 16 h before adding the stimulating agent, rLoa22-LPGN complexes to precise evaluations of cellular function. Different time points were tested to determine the maximum phosphorylation levels of p38, ERK, and JNK. The maximum phosphorylation levels occurred at 1 h after stimulation with rLoa22-LPGN complexes in serum-free HK2 and HEK-293-TLR2 cells. Stimulation of HEK293-TLR2 cells by rLoa22-LPGN complexes significantly increased the phosphorylation of p38, ERK, and JNK as compared to HEK-293 cells (Fig. 5A). For TLR2 antibody neutralization experiments, HK2 cells were pretreated with anti-TLR2 antibody (10 μg/ml) for 1 h, followed by adding of rLoa22-LPGN complexes for stimulation. Results in HK2 cells also revealed that cells pretreated with TLR2 antibody exhibited significantly decreased phosphorylation of p38, ERK, and JNK as compared to non-neutralized controls or non rLoa22-LPGN complexes stimulation controls (Fig. 5B). Besides, we also tested the NFκB nuclear translocation using luciferase assay. We transfected pNFκB luc plasmid (Stratagene, La Jolla, CA) into HEK293-TLR2 cell and further treated the cells with PBS (control), TNF-α, rLipL32, and rLoa22-LPGN to evaluate the NFκB nuclear translocation activity. The luciferase assays demonstrated that rLoa22-LPGN significantly increased the NFκB nuclear translocation activity as compare to PBS treatment (Fig. 5C). All these results suggested that rLoa22-LPGN complexes stimulate the production of inflammatory responses through p38, ERK, JNK, and NFκB signaling pathways.

Inflammatory responses induces by rLoa22-LPGN complexes.
Recognition of bacterial components by host TLRs initiates signaling cascades that stimulate nuclear transcription factor κB (NF-κB) and mitogen-activated protein kinases (MAPKs), and induces expression of chemokines and cytokines 25,26 . rLoa22 was expressed in E. coli ClearColi BL21 (DE3) pLys that contained low levels of endotoxin and was further purified by Ni 2+ -columns, MonoQ, and polymyxin to remove the contaminating endotoxin (Fig. S1C). The mRNA and protein expression levels of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α were measured to investigate the role of rLoa22 HEK293-TLR2 cells. Two hours after incubating with the stimulation agents, the HEK293-TLR2 cells were collected for mRNA analysis. In the ELISA assays, HEK293-TLR2 cells were incubated with the stimulation agents for 8 h, the supernatants were collected for cytokines measurements. Purified LipL32 protein was used as the positive control, and PBS buffer alone was served as negative control 23 . In HEK293-TLR2 cells, LipL32 and Loa22WT significantly increased the mRNA and protein expression of CXCL8/IL8, hCCL2/MCP-1, and www.nature.com/scientificreports/ hTNF-α as compared to the PBS control (Fig. S4). Further confirming the inflammatory cytokines induced by recombinant Loa22 protein, the Loa22 was heat-treated (100 °C, 30 min) and digested with proteinase K (20 µg/ml at 63 °C for 18 h), and the results revealed that the denatured and digested rLoa22 protein significantly decreased the mRNA and protein expression of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α as compared to that of Loa22WT (Fig. S4). The absence of stimulatory effects after heat and proteinase K treatments further demonstrated that the Loa22WT provoked the inflammatory responses. Besides, LPGN and Loa22WT significantly increased the mRNA and protein expression of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α as compared to the PBS control (Fig. 6). Furthermore, the rLoa22-LPGN complexes significantly increased mRNA and protein expression levels of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α as compared to that of Loa22WT (Fig. 6) 22,25 . We www.nature.com/scientificreports/ further investigated the roles of LPGN and the mutated variants of rLoa22 in the stimulation of inflammatory responses. The rLoa22D122A and rLoa22R143A mutated variants with low affinity to LPGN showed relative low ability to stimulate the mRNA and protein expression levels of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α (Fig. 6). The mutation variants (Loa22D122A and Loa22R143A) in complex with LPGN significantly decreased the mRNA and protein expression levels of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α as compared to that of rLoa22WT-LPGN complexes. The results further demonstrated that the LPGN cooperated with rLoa22 to interact with TLR2 and stimulated inflammatory responses. In order to identify the inflammation induced by Loa22-LPGN in primary human cells, human monocytic cell line (THP-1), was used to measure the cytokines production under the stimulation of Loa22-LPGN. THP-1 was cultured to 2 × 10 7 /well and induced the differentiation into macrophages using phorbol-12-myristate-13-acetate (PMA). In THP-1 cells, rLoa22-LPGN significantly increased the protein expression of hCXCL8/IL8, hCCL2/MCP-1, and hTNF-α that indicated the rLoa22-LPGN stimulated cytokines production in primary human cells and transfected cell lines of macrophage (Fig. S5). Previous studies suggested that TLR2 preferred to form heterodimer with either TLR1 or 6 depended on different antigens stimulation 18,28 . In order to identify which heterodimer is important to therLoa22-LPGN complex, we co-transfected of TLR2-TLR1 and TLR2-TLR6 into HEK293 cell and stimulated these cells with rLoa22-LPGN complexes to identify which TLRs cooperated with TLR2 in responses of rLoa22-LPGN complexes. The transfection efficiencies of these genes into HEK-293 cells were similar (Fig. S6). The inflammatory responses induced by rLoa22-LPGN complex were assayed including hCXCL8/IL8, hCCL2/MCP-1, and hTNF-α (Fig. 7). The expression level of mRNA was measured by real time PCR and the expression level of protein was measured by ELISA to identify the simulation of rLoa22-LPGN complex in the transfected HEK-293 cells. The results indicated that rLoa20-LPGN stimulated highest levels of cytokines expression mainly through TLR2 (Fig. 7). In the stimulation of IL-8, the rLoa22-LPGN complex also significantly increased mRNA expression in HEK-293-TLR2-TLR1 cells (Fig. 7A). In the stimulation of MCP-1, the rLoa22-LPGN complex significantly increased mRNA expression in HEK-293-TLR2-TLR6 cells (Fig. 7B). In the stimulation of TNF-α, the rLoa22-LPGN complex significantly increased mRNA and protein expression in HEK-293-TLR2-TLR1 cells (Fig. 7C,F).

Discussion
In Leptospira, Loa22 has been shown as an essential virulence factor, in that deletion of the Loa22 from pathogenic Leptospira attenuates toxicity, whereas the re-expression the gene in Leptospira restores the virulence 11 . Leptospira biflexa serovar Patoc is non pathogenic strain that contains a Loa22-like gene (WP_012390072.1) but the expression of this gene seemed seemed to be downregulated that the protein level can hardly be detected (Fig. 1B) 10 . However, the pathogenic mechanisms and the vital domains of Loa22 are still unclear. In this study, we used pull-down assays to demonstrate the interactions between TLR2 and Loa22 from Leptospira outer membrane extractions. One of the positive controls of this pull-down assay was LipL32, which has been proven as a TLR2 binding protein in pathogenic Leptospira 22,25,26 . Interestingly, domain prediction of Loa22 protein suggested an OmpA-like domain and was speculated to interact with PGN. Therefore, PGN molecules were used to investigate the affinity with Loa22. However, PGN molecules from E. coli (EPGN), S. aureus (SPGN), and B. subtilis (BPGN) showed low affinity for Loa22. PGN molecules from pathogenic Leptospira (LPGN) are the only PGN molecule that showed relative high affinity for Loa22 (Fig. 2D). These results suggested that the structure and composition of LPGN were different from those of EPGN, SPGN, and BPGN, and required further investigation. Combined with AFM force-distance curve studies and site-directed mutagenesis demonstrated that Loa22 directly interacted with LPGN on bacterial surfaces and two vital residues were involved in the interaction between Loa22 and LPGN (Fig. 2). Furthermore, Loa22-LPGN complexes were observed to co-localize with TLR2 on HEK293-TLR2 cell surfaces (Fig. 3). The two mutation variants with low or no affinity for LPGN also revealed that the co-localization behaviors to TLR2 on HEK293-TLR2 cell surface were attenuated (Fig. 3). In addition, the interaction forces and binding frequencies between rLoa22 and TLR2 showed no significant difference as compared to BSA control whereas the interaction forces and binding frequencies between rLoa22-LPGN complexes and TLR2 showed significantly increased. The rLoa22 mutation variants also decreased the affinity for TLR2 in the absence or presence of LPGN as compared to rLoa22WT-LPGN complexes (Fig. 4). We suggested two possibilities models concerning the relationships between rLoa22, LPGN, and TLR2. Firstly, Loa22 could not directly bind to TLR2; rather Loa22 used LPGN to interact with TLR2. According to this model, the concept of defining PGN as the TLR2 ligand is controversial, and whether PGN from pathogenic Leptospira as the TLR2 ligand is still unclear. The unique PGN from pathogenic Leptospira also showed higher affinities for Loa22 as compared to those from E. coli, S. aureus, and B. subtilis (Fig. 2D). Leptospira is an idiographic bacterium that differs from other pathogens and its composition and reaction to host cells required further elucidation. Secondly, Loa22 interacted with LPGN and induced conformational changes in the protein, which exposed the TLR2binding domain of Loa22 to interact with TLR2. The second model is more likely, wherein LPGN induces conformational changes in Loa22 and therefore triggers the interaction between Loa22-LPGN and TLR2. Evidence for this model is that the Loa22 mutation variants exhibited low affinity for LPGN and decreased the affinity for TLR2. Besides, anti-rLoa22 antibody was used to neutralize the rLoa22-LPGN complex and the result showed that the affinity between TLR2 and rLoa22-LPGN were attenuated. This study provides further indication of a role of LPGN participating in mediating Leptospira recognition by the innate immune system and as a potential target for anti-Leptospira treatment.
It has been reported that TLR2 interacts with PGN and induces inflammatory responses 24 . In this study, rLoa22 was demonstrated to interact with LPGN through the two vital residues, and consequently interacted with TLR2 to induce downstream signals and cytokines production. Downstream signals induced by rLoa22-LPGN complexes were explored in HEK293 and HK2 cells. MAPK signal transduction pathway components, including p38, ERK, and JNK, were obviously stimulated. Previous studies have reported that HEK293 cell express no www.nature.com/scientificreports/ TLR2 on cell surfaces, and we constructed HEK293-TLR2 cells to express TLR2 22 . HEK293 cells were used as negative controls. rLoa22-LPGN complexes significantly up-regulated the phosphorylation levels of p38, ERK, and JNK in HEK293-TLR2 cells, but not in HEK293 cells (Fig. 5A). In HK2 cells, adding anti-TLR2 antibody for neutralization procedure blocked the interaction between rLoa22-LPGN and TLR2, and therefore alleviated the activation of the MAPK pathway (Fig. 5B). In the NF-κB nuclear translocation assay, rLoa22-LPGN significantly increased the NF-κB nuclear translocation activity as compare to PBS treatment, similar to the positive control TNF-α (Fig. 5C). The inflammatory responses provoked by rLoa22 in HEK293-TLR2 cells were used to measure the toxicity of this essential virulence factor. Previous studies have shown that Leptospira outer membrane proteins induced expression of nitric oxide, MCP-1, and TNF-α in cells 29 . Tian et al. also demonstrated that outer membrane proteins of L. santarosai Shermani increased collagen and TGF-β in renal proximal tubular cells 2 . These data suggested that proinflammatory cytokines production might be involved in tubulointerstitial nephritis caused by L. santarosai Shermani infection through its outer membrane components. The recombinant protein method is the best way to investigate structural and functional relationships. However, endotoxin contamination from the recombinant protein can interfere with the intrinsic inflammatory properties of the innate immunity system. The endotoxin-free expression system provides the solution to overcome this problem. The ClearColi BL21(DE3) expression system was used to produce endotoxin free rLoa22 and its variants, and the yield of protein expressed was similar to that expressed in the BL21(DE3) expression system 30 . In addition, anion-exchange chromatography (Mono-Q) and polymyxin B resin were further used to remove contaminating endotoxin from purified rLoa22 and its variants. The LAL assay was used to confirm that the endotoxin was removed after the purification processes (Fig. S1C). In terms of inflammatory responses, rLoa22WT protein showed low ability to induce inflammatory responses, whereas when rLoa22 combined with LPGN induced an increased inflammatory response, similar to the major outer membrane LipL32 protein 22 . In fact, LPGN could significantly stimulate the expression of CXCL8/IL8, hCCL2/MCP-1, and hTNF-α (Fig. 6). Muller-Anstett et al. reported that SPGN co-localized with TLR2 and stimulated innate immune responses 24 . The inflammatory responses induced by rLoa22WT-LPGN complex showed significantly increased as compared to that of Loa22WT (Fig. 6). These results demonstrated that rLoa22-LPGN complexes effectively stimulated immune responses much more than LGPN alone, and further demonstrated that the inflammatory responses were increased from Loa22-LPGN complex. On the other hand, the rLoa22 mutation variants also reduced the ability to induce inflammatory responses in the presence or absence of LPGN as compared to that of rLoa22WT-LPGN complex (Fig. 6). These results supported our hypothesis that LPGN bind rLoa22 and induced conformational changes to interact with TLR2 on cell surfaces. Furthermore, TLR2 forms heterodimers with TLR1 or TLR6 to recognize different pathogen antigens and rLoa22-LPGN was also used to interact with different TLR pairs. In this regard, when HEK-293 cells were co-transfected with TLR2-TLR1, TLR2-TLR6, or TLR2 only plasmid DNA into the HEK-293 cell and assayed the inflammatory responses induced by Loa22-LPGN and the results indicated that rLoa20-LPGN complex stimulated cytokines expression mainly through TLR2 (Fig. 7). TLR2 might formed homodimer on cell surface and recognized the rLoa20-LPGN complex to induce the downstream signals 6 .
In summary, our study demonstrated that Loa22 protein is a PGN binding protein and the PGN from Leptospira showed high affinity for Loa22. The LPGN binding activity of Loa22 is an important biological reaction to maintain cell wall stability and to protect against immune attack when infecting host cells. We further demonstrated that two key residues within the OmpA domain, Asp 122 and Arg 143 , were involved in the affinity of Loa22 for LPGN. The interaction of Loa22 and TLR2 was explored by ELISA and AFM and the role of LPGN was verified to mediate Loa22 to interact with TLR2. Finally, Loa22 was proposed to interact with LPGN through two vital residues, and subsequently interacts with TLR2. This study showed that LPGN in Leptospira mediates interactions between Loa22 and TLR2 and increases downstream signals to trigger inflammatory responses. Interactions between Loa22-LPGN-TLR2 reveal a novel binding mechanism for the innate immune system and infection induced by Leptospira. Pull-down assay. HEK293-TLR2 cell was used to express the human TLR2 protein for pull-down assay.

Methods
After 48 h cell culture, HEK293-TLR2 cells were incubated with LOMP (10 μg/ml) for 4 h, followed by three times of PBS washes and centrifugation to remove unbound LOMP. The cells were lysed with Cell Lysis Buffer (Abcam; ab152163) and the supernatant were incubated with mouse anti-V5-tag antibody or mouse IgG (negative control) at 4 °C overnight, and then incubated with protein A-agarose beads (Roche) for 4 h. The beads were washed four times with lysis buffer, and the obtained samples were analyzed by 12% (w/v) SDS-PAGE.
DNA construction and mutagenesis. The loa22 gene (LSS_RS00795; 591 bp) was cloned from pathogenic L. santarosai serovar Shermani genomic DNA using pfu-Turbo DNA polymerase (Stratagene, La Jolla, CA) 23 . The primers used for loa22 gene construction were listed in Table 1. After restriction enzyme double digestion, the PCR product was individually inserted into the expression vector pRSET (Invitrogen, Groningen, Netherlands Expression and purification of Loa22 and antibody preparation. The DNA constructs of Loa22 were individually transformed into expression host cell E.coli ClearColi BL21 (DE3) pLys (Lucigen, Middleton, WI). Loa22 and its variants were expressed and purified by using the methods mentioned previously 22,25,26 . The imidazole was removed by dialysis before assays. rLoa22 antigens were also used to induce the production of the polyclonal antibodies (anti-Loa22 antibody) by customized product (ABclonal Inc., MA, USA).

RNA extraction and real-time PCR.
HEK293 and HEK293-TLR2 cells were incubated at 37 °C in a humidified atmosphere of 5% (v/v) CO 2 for 70% confluence as previously described 6 . Cells were shifted to a serum-free medium for 24 h before adding the stimulation agents to the cell culture medium. Total RNA was extracted according to the guanidinium thiocyanate/phenol/chloroform method (Cinna/Biotecx Laboratories International Inc., Friendswood, TX) 6,37 . Real-time PCR was executed on the basis of the manufacturer's instructions using an ABI Prism 7700 with SYBR green I as a double-stranded DNA-specific dye (PE-Applied Biosystems, Cheshire, Great Britain). The primers of hTNF-α, hCXCL8/IL-8, and hCCL2/MCP-1 were shown in Table 1 and constructed to be compatible with a single reverse transcription-PCR thermal profile (95 °C for 10 min, 40 cycles at 95 °C for 30 s, 60 °C for 1 min, and 72 °C for 3 min). The accumulation of the PCR product was recorded in real time (PE-Applied Biosystems). The results of mRNA levels in different genes are displayed as the transcript levels of the analyzed genes relative to GAPDH (glyceraldehyde-3-phosphate dehydrogenase) transcript level.
Leptospira PGN preparation. L. santarosai serovar Shermani PGN was extracted according to the procedure of previous report 38 . Briefly, 2 L of Leptospira culture were harvested by centrifugation at 10,000 xg for 30 min, washed three times with PBS buffer, and resuspended in 100 ml 1% (w/v) SDS in distilled water. The suspension was gently shaken at 37 °C for 18 h and then centrifuged at 110,000 xg for 60 min. After a second treatment with 1% (w/v) SDS the pellet was washed three times with 6 M urea. The pellet was further resuspended in 100 ml distilled water and collected by centrifugation at 110,000 xg to remove the urea. The pellet was suspended in 10 ml 10 mM Tris-HCl buffer (pH 7.4) containing 0.1 mg/ml trypsin and incubated at 37 °C for 18 h. The pellet was collected by centrifugation at 110,000 xg for 90 min and suspended in 10 ml 10 mM Tris-HCl buffer (pH 7.4) containing 0.1 mg/ml pronase (Sigma) for incubation at 37 °C for 18 h. After digestion at 37 °C for 18 h the pellet was recovered by centrifugation at 110,000 xg for 90 min and then lyophilized. The amount of PGN extracted was measured of the dry weight of pellet and resuspended in distilled water for 1 mg/ ml and stock at -80 °C until use.
Enzyme-linked immunosorbent assay (ELISA). The ELISA methods were used to investigate the interaction between Loa22 and TLR2 and the processes of this method was performed according to previously reports with minor modifications 25,39 . Briefly, rTLR2 (1 μg) protein was coated on ELISA plates and Loa22 and its variants (2 μM) were used to interact with rTLR2. Anti-Loa22 antibody (1:10,000 dilution) was used to detect the amount of rLoa22 protein. The binding data were analyzed via SigmaPlot 10.0 program by fitting to the optimal equation 40 .
AFM Measurement and Analysis. The AFM cantilever tips were functionalized according to previous methods to modify the rLoa22 protein and its variants 25,26,41 . The mica surface was modified for deposition of rTLR2 protein according to previous report 42,43 . The unbound proteins were removed and tips and mica were stored at 4 °C until use. A commercial atomic force microscope (Nanoscope III, Digital Instruments, Santa Barbara, CA) with a J type scanner was employed throughout this study. The distance-force curves and force parameters were obtained according to the methods previously described 25,41 . All the measurements described above were performed with modified tips and showed repeatedly similar results.

Statistical analysis.
All experiments were performed at least three independent processes and the variables are expressed as mean ± SEM and compared by using Student's t-test or one-way ANOVA. p values < 0.05 are considered statistically significant. All analyses were performed using the Graphpad Prism 5.1 (Graphpad, La Jolla, CA).