Active Components of Leptospira Outer Membrane Protein LipL32 to Toll-Like Receptor 2

Proteins belonging to the toll-like receptor (TLR) family, particularly TLR2, are the major components of innate immunity against Leptospira infection. The ligands for TLR2 harbor several conserved patterns such as lipidation molecules, leucine-rich repeat (LRR) domains, TLR2 binding motifs, and TLR2 binding structure. In Leptospira, LipL32 interacts with TLR2 on human kidney cells concomitantly stimulating inflammatory responses. However, the binding mechanism of LipL32 to TLR2 is unknown. The computational prediction suggests that β1β2, loop-α3-loop, and α4 domains of LipL32 play vital roles in LipL32-TLR2 complex formation. To test these predictions, protein truncation experiments revealed that LipL32ΔNβ1β2 significantly decreased the affinity to TLR2 while LipL32ΔCα4 slightly reduced it. Interestingly, LipL32ΔCenα3 retained affinity to TLR2 in the absence of Ca2+ ions, indicating that Cenα3 play a role preventing the interaction between LipL32 and TLR2. Furthermore, the critical residues of LipL32 involved in interacting with TLR2 suggested that V35S, L36S and L263S variants significantly decreased the affinity to TLR2. The results further confirm that LipL32 interacts with TLR2 through Nβ1β2 and Cα4 domains of LipL32 as well as LipL32-TLR2 complex formation results from hydrophobic interactions. This study provides a detailed mechanism of the interaction between LipL32 and TLR2 and the residues involved in complex formation.


Results
Molecular docking. The crystallographic structure of the TLR2/1-Pam 3 CSK 4 complex suggests a mechanism for recognition of lapidated ligands by TLRs 35 ; however, the protein ligands of TLR2 are still unknown. To identify the binding mechanism, we modeled the interaction between LipL32 (2WFK) and TLR2 (2Z7X). The top ten scoring models of TLR2/LipL32 were retrieved from ClusPro 40 protein-protein docking website and were divided into three groups on the basis of LipL32 orientation in interact with TLR2 (Fig. S1A). According to the three groups, LipL32 may interact with TLR2 through its central domain (Model I), N terminal domain alone (Model II), or N terminal domain assisted by the C terminal domain (Model III). From the results of protein docking predictions, three vital domains (including the N terminal β1β2 sheet, central α3 helix, and C terminal α4 helix) were predicted to be involved in interaction with TLR2. Therefore, these three domains were individually deleted in construction of the truncation variants for protein expression and to assay their affinity to TLR2 (Fig. 1A). The three vital domains were shown in the 3D structure as indicated in Fig. 1B (red arrows).
Protein construction, purification, and characterization. TLR2 protein is type I transmembrane protein with three domains including leucine-rich repeat (LRR), transmembrane domain (TMD), and Toll/ interleukin-1 receptor domain (TIR) 31,34 . The LRR domain is responsible for sensing the bacteria components and consequently dimerizes with other LRR domain of TLR1, 2, or 6 to induce downstream signaling. Therefore, the human full-length tlr2 DNA was cloned into the vector pLenti6.3/V5-TOPO with the primers listed in Table 1. The DNA sequence was further confirmed by DNA sequencing. The pLenti6.3/V5-TLR2 DNA was packaged using the ViraPower ™ Lentiviral Expression System according to the manual suggestion and transfected into HEK293FT cell for stable cell line construction (Invitrogen). After selecting by blasticidin, HEK293-TLR2 cell were cultured in the medium of DMEM with 10% FBS and maintained at 37 °C in a humidified atmosphere containing 5% CO 2 . In the protein purification process, the HEK293-TLR2 cell were solublized in the PBS buffer containing 0.5% CHAPS with gently stirred for 1h at 4 °C. Recombinant TLR2 (rTLR2) protein was purified in a series of purification procedures including anti-V5 antibody activated NHS resin affinity purification, mono Q Scientific RepoRts | 7: 8363 | DOI: 10.1038/s41598-017-08743-y ion exchange column, and Superdex 200 gel filtration column (Fig. S2A). The molecular mass of purified rTLR2 protein was about 240 kDa as determined by using Superdex 200 gel filtration column (Fig. 1C). The function of rTLR2 protein was tested by the affinity to its nature ligands including lipoteichoic acid (LTA), Pam 3 CSK 4 , and peptidoglycan (PGN), respectively 25 (Fig. S3). The interaction of the purified rTLR2 protein with its nature ligands showed statistically significant difference indicated that the purified rTLR2 protein is suitable for the ligands selection.
The full-length LipL32 protein containing 272aa with N terminal 20 signal peptide. The mature form of recombinant LipL32 protein (rLipL32; 21-272) and the different fragments including rLipL32ΔNβ1β2 (51-272), rLipL32ΔCenα3 (deletion the residues of 195-200), and rLipL32ΔCα4 (21-247), respectively, were constructed Three dimension structure of LipL32 and its relative domains. Blue, α-helix; Yellow, β-sheet; Gray, random-coil; red arrows, active motifs for TLR2 interaction. (C) Size exclusion chromatography (SEC) of purified TLR2 protein and LipL32 variants. TLR2 protein and LipL32 variants were purified as described in Materials and Methods and SEC was used to identify the integrity and purity of these proteins. (D) Circular dichroism (CD) of purified TLR2 protein and LipL32 variants purified. The secondary structure of purified TLR2 protein and LipL32 variants were measured by CD spectroscopy and the spectra were record between 260 nm and 190 nm. (E) Ca 2+ binding assay by stains-all assay. The Ca 2+ binding activity of LipL32 variants were tested by stains-all and the J band at 660 nm was measured. TLR2 protein was used as negative control. Black, TLR2 protein; Red, LipL32WT; Blue, LipL32ΔCenα3; Pink, LipL32ΔCα4; Green, LipL32ΔNβ1β2 in Figure (C-E).
according to the modeling results and these variants were used to evaluate the affinity to TLR2 (Fig. 1A). In addition, the vital residues within the deletion domains of rLipL32 that potentially involved in TLR2 interaction including F34S, V35S, L36S, D195A, D196A, and L263S were performed by using Q5 ® Site-Directed Mutagenesis and the primers used for site-directed mutagenesis were listed in Table 1. The rLipL32WT and its variants were expressed in the E. coli ClearColi TM BL21(DE3) and BL21(DE3)pLys to obtained the recombinant proteins ( Fig. S2B and C). The recombinant proteins were purified by Ni 2+ -NTA affinity column and eluted in 250 mM imidazole. In order to determine the endotoxin contents, Limulus amebocyte lysate (LAL) assay was used to measure the endotoxin contamination of these recombinant proteins. The rLipL32 proteins from E. coli BL21(DE3)pLys contained high level of endotoxin that >100000EU/mg, while the protein from E. coli ClearColi TM BL21(DE3) significant decreased the endotoxin contamination to 688.5 ± 23.4 EU/mg (Fig. S4). The recombinant proteins from ClearColi TM BL21(DE3) contain relatively low endotoxin and suitable for mammalian cell applications. In order to improve the purify and removed endotoxin contamination of the recombinant proteins, Mono Q ion exchange chromatograph was further used to purify the LipL32 protein and its variants as well as the proteins were assayed by Superdex 200 gel filtration column (Fig. 1C) 41 . Mono Q is a strong anion exchange column with advantage of adsorption the negative net charge of endotoxins, have been extensively used for endotoxins adsorption and removal 42 . The level of endotoxin contamination in this rLipL32 fraction was significantly low (Fig. S4). To reduce the LPS contamination even further, rLipL32WT and its variants were subjected to polymyxin B resin. Endotoxin contamination was notably decreased after purified by mono Q ion exchanged column and polymyxin B resin (<0.1EU/mg; p < 0.001) (Fig. S4). The molecular mass of rLipL32 and its variants were assay by using Superdex 200 gel filtration column (Fig. 1C). The structure stability and integrity of rLipL32 and its relative variants were measured by CD and the results indicated that the central β-sandwich domain of rLipL32 was stable when deletion the domains mentioned above (Fig. 1D). Further identification the structure and function integrity of rLipL32 deletion variants, the Ca 2+ binding ability was measured by stains-all assay and the result represented that the deletion variants of rLipL32 could bind stains-all and the J band at 650 nm were observed (Fig. 1E). The Ca 2+ binding domain of rLipL32 is located at the poly D motif within the central β-sandwich domain of rLipL32 25 and the results indicated that the deletion of the three domains of rLipL32 maintained the Ca 2+ binding ability. These results also showed that the structure integrated and Ca 2+ binding ability of rLipL32 protein when deletion of possible rTLR2 binding domains. Therefore, the variants of the affinity to rTLR2 of rLipL32 truncated variants are resulting from the deletion of the vital domains involved in LipL32-TLR2 complex formation.
LipL32 co-localizes with TLR2 on HEK293 cells. In the infection of host cell with Leptospira, outer membrane proteins of Leptospira are the first line to contact the host cell and interact with the cell membrane components. LipL32 is the most abundant outer membrane lipoprotein and occupied about 75% of the total outer membrane protein profile 43 . Previous studies have illustrated that LipL32 binds to different components of the ECM, including fibronectin, collagen, and laminin, respectively 9,22 . Besides ECM molecules, LipL32 also interacts with innate immune components, such as TLR2, and induces inflammatory responses 25,26 . However, the domains of LipL32 responsible for interacting with TLR2 are unclear. Therefore, confocal microscopy was used to observe the co-localization of truncated rLipL32 variants with the TLR2 receptor on the HEK293 cell surface (Fig. 2). Purified rLipL32 variants were used to incubate with HEK293 or HEK293-TLR2 cells for 4 h, after which the cells were washed three times with PBS buffer to remove non-interacting rLipL32 proteins and fixed for confocal microscopy analysis. rLipL32 and rTLR2 proteins were stained with rabbit polyclonal anti-LipL32 and mouse monoclonal anti-V5 primary antibodies fellow by Alexa594 (red) conjugated anti-rabbit and Alexa488 (green) conjugated anti-mouse secondary antibodies, respectively. HEK293 cells lacking the expression of rTLR2 protein were used as negative control and very little or no fluorescence of Alexa 488 signal ( Fig. 2A). As rLipL32 proteins could bind to the cell surface of these HEK293 cells as well, it seems that rLipL32 proteins associate  with cell surface ECM molecules 9 . In rTLR2 expressing HEK293 cells, most of rLipL32WT proteins co-localized with TLR2 receptors (Fig. 2B). This provides direct evidence that the rTLR2 receptor and rLipL32 proteins were co-localize on the cell surface (Fig. 2B, yellow color). Further confirming the rLipL32 protein co-localized with rTLR2, the rLipL32 protein was treated with proteinase K (PK) for 30 min before incubating with HEK293-TLR2 cell and the results indicated the reducing the ability to bind rTLR2 (Fig. 2C). To compare different truncations of rLipL32 in their interaction with the rTLR2 receptor, HEK293-TLR2 cells were incubated with rLipL32 truncation variants to evaluate co-localization phenomenon. Interestingly, although most of the ΔNβ1β2 variants of rLipL32 were bound to cell surface, they did not co-localize with TLR2 (Fig. 2F). In contrast, the rLipL32ΔCenα3 variant showed similar behavior as compared to rLipL32WT (Fig. 2D). Pearson's Correlation was further used to elucidate the co-localization between rTLR2 and rLipL32 truncated variants. rLipL32ΔCα4 (p < 0.01) and rLipL32ΔNβ1β2 (p < 0.001) significantly decreased the co-localization to TLR2 (Fig. 1G). The results suggested that the N-and C-termini of rLipL32 play a critical role in rTLR2 interaction.
, and osteopontin/SPP1 (SPP1) 8,27 were measured to investigate TLR2 activation by its ligands (Fig. 3). Two hours after adding 2.5 μg/ml stimulating agents (including leptospiral outer membrane protein extraction [LOMP], WT, ΔNβ1β2, ΔCenα3, and ΔCα4 of rLipL32) to HEK293-TLR2 cells, the mRNA levels of IL8, MCP-1, hTNF-α, hMMP7, and GRO-α were changed by addition of the truncated rLipL32 proteins as compared to WT. However, these virulence factors could not stimulate the expression of SPP1. LOMP stimulated the expression of IL-8, MCP-1, TNF-α, MMP-7, and GRO-α and considered as a positive control whereas the PBS buffer acted as mock control ( Fig. 3) 25 . To test the effect of a His 6 -tag on rLipL32, His 6 -tagged and un-tagged rLipL32 were used to induce the inflammatory responses and the results indicated that both forms of rLipL32 stimulated similar level of IL-8, MCP-1, TNF-α, MMP-7, GRO-α, and SPP1 (Fig. S5A). The two forms of rLipL32 showed no significant difference in their ability to induce inflammatory responses. Thus, the presence or absence of His 6 -tag on rLipL32 has no obvious effects on their ability to induce inflammatory responses. We next examined whether downstream signaling was directly induced by rLipL32 proteins, therefore, and the proteins were treated with heat (100 °C, 30 min) and proteinase K (20 µg/ml at 63 °C for 18 h). The results showed that rLipL32 protein treated with heat and proteinase K had a significantly lower level of inflammatory induction compared to that of WT ( Fig. S5B; p < 0.01).
To confirm the N terminal β sheet as the binding domain, the synthesized peptide including N20 peptide (1-20 residues of LipL32), N50 peptide (1-50 residues of LipL32), C25 peptide (248-272 residues of LipL32), and a scrambled peptide with the sequence of VSLEGLTKVLAGSKFSESPILIKTVCDMILGASFANPALKISTASTLFTG, were used to induce inflammatory responses. With the exception of the N50 peptide, all these synthesized peptides stimulated less than 2-fold inflammatory responses compared to that stimulated by the mock control (Fig. 2B). The N50 peptide stimulated the expression of IL8, TNF-α, MMP7 by 2.1-, 2.3-, and 2.3-fold, respectively. The synthesized peptides revealed significantly decreased the stimulation level of IL-8 as compared to WT ( Fig. 2E-H; p < 0.05). Thus, the stimulation of inflammatory responses provoked by the N50 peptide indicates that the binding domain is located within the N terminal region.

The interaction between LipL32 and TLR2 as determined by ELISA. Previous studies indicated
that LipL32 directly interacts with TLR2, and we further investigated the molecular basis of interaction between rLipL32 and rTLR2 25 . Using ELISA, we first analyzed the binding of rTLR2 protein to its ligands, including Lipoteichoic acid (LTA), Pam 3 CSK 4 , and peptidoglycan (PGN), and these results indicating that the rTLR2 protein from our system was suitable for ligand binding assays (Fig. S3A) 25 . To identify the active components involved in the interaction between rLipL32 and rTLR2, the binding efficiency of rLipL32WT and its variants, including rLipL32ΔNβ1β2, rLipL32ΔCenα3, and rLipL32ΔCα4, in associating with rTLR2 were determined (Fig. 4A). rLipL32WT pretreated with proteinase K significantly decreased the affinity to rTLR2 (Fig. 4A, p < 0.01). rLip-L32ΔNβ1β2 (p < 0.001) and rLipL32ΔCα4 (p < 0.05) had significantly decreased binding efficiency and rLip-L32ΔCenα3 showed no significant difference in binding efficiency, as compared to that of rLipL32WT (Fig. 4A). This result indicates that the N terminal β1β2 and C terminal α4 domains may be involved in interaction with rTLR2. We next examined the residues within the two domains of rLipL32 are involved in the interaction with rTLR2. Previous studies indicated that rTLR2 binding proteins may utilize hydrophobic residues to interact with the LRR domain of rTLR2 36,38,39 . In addition, the molecular docking analysis for Model III indicated the residues with the nearest distance to rTLR2, including Phe 34 , Val 35 , Leu 36 , and Leu 263 ; these residues were mutated to Serine and the affinity between mutant variants of rLipL32 and rTLR2 were measured (Fig. S1). The V35S and L263S variants of rLipL32 had significantly decreased affinity to rTLR2 compared to that of rLipL32WT (p < 0.05); the L36S variant showed significantly decreased the affinity to rTLR2 compared to rLipL32WT (p < 0.01) (Fig. 4B). These results further confirmed that the N terminal β1β2 and C terminal α4 domains are involved in interacting with rTLR2 and the residues Val 35 , Leu 36 , and Leu 263 possibly involved in the interaction interface of rLipL32-rTLR2 complex. It also suggests that rLipL32 uses hydrophobic residues to interact with rTLR2.
Scientific RepoRts | 7: 8363 | DOI:10.1038/s41598-017-08743-y The role of Ca 2+ in regulating in the interaction between rLipL32 and rTLR2 was shown by dose-dependent ELISA, and the binding affinity of rLipL32 variants to rTLR2 was measured (Fig. 4C). The dissociation constant (K d ) was calculated as 0.13 ± 0.02 µM from Eq. 1 for the binding of rLipL32WT to rTLR2 proteins (Fig. 4C). A previous study confirmed that Ca 2+ regulates the interaction of rTLR2 and rLipL32 as well as poly-D motif variants of rLipL32, which reduce regulation ability 25 . The role of Ca 2+ ions was further investigated in these deletion variants of rLipL32. In the presence of ethylene glycol tetraacetic acid (EGTA), the affinity of rLipL32WT to rTLR2 was decreased to 0.45 ± 0.12 µM as compared to that in the presence of Ca 2+ ions, demonstrating that sequestration of Ca 2+ ion reduced the binding affinity of rLipL32 to rTLR2 proteins (Fig. 4C) 25 . This also demonstrates that rLipL32 and rTLR2 interact in a Ca 2+ dependence manner. To test the role of Ca 2+ in regulating the rLipL32-rTLR2 interaction, the central α3 domain of rLipL32 was removed (LipL32ΔCenα3 variant), and the affinity of rLipL32 to rTLR2 was measured in the absence and presence of Ca 2+ (Fig. 4C). Interestingly, after the deletion of the central α3 domain, this variant showed similar K d values in either the absence or presence of Ca 2+ ions, indicating that rLipL32ΔCenα3 variant might shifts from Ca 2+ dependent to Ca 2+ independent of affinity for rTLR2 protein. According to the crystal structure of LipL32 in the absence and presence of Ca 2+ ions, the conformational change that occurs in the Cenα3 domain suggests that this domain is the Ca 2+ regulation motif. In addition, Cluspro predicted that the Cenα3 domain was close to the TLR2 protein and that this domain may be involved in interacting with TLR2. However, the affinity of rTLR2 to rLipL32ΔCenα3 showed no significant (dark gray bar), and GRO-α (black bar) on HEK293-TLR2 cell. N20 peptide (1-20 residues of LipL32); N50 peptide (1-50 residues of LipL32); C25 peptide (248-272 residues of LipL32), and a scrambling peptide with the sequence of VSLEGLTKVLAGSKFSESPILIKTVCDMILGASFANPALKISTASTLFTG. *p < 0.05; **p < 0.01.
Scientific RepoRts | 7: 8363 | DOI:10.1038/s41598-017-08743-y difference compared to that in rLipL32WT (Fig. 4A). Therefore, the Cenα3 domain may instead play a role as an insulator that interferes with the interaction of rLipL32 and rTLR2 in the absence of Ca 2+ .
We next tested the importance of Cenα3 for the rLipL32-rTLR2 interaction. Two residues, D195 and D196, within Cenα3 domain were selected and mutated to Ala for affinity analysis. In dose-dependent ELISA, the K d values of D195A and D196A were calculated as 0.10 ± 0.03 µM and 0.16 ± 0.04 µM, respectively, in the presence of Ca 2+ (Fig. 4D). In the treatment of EGTA, the K d values of D195A and D196A were calculated as 0.12 ± 0.04 and 0.16 ± 0.03 µM, respectively. Thus, the binding affinity of these two variants showed similar behavior in their interaction with rTLR2, further confirming the role of the Cenα3 domain. The charged residues, D195 and D196 within the Cenα3 domain may interfere with the interaction between rLipL32 and rTLR2; replacing both of the charged residues to Ala altered the ability of Ca 2+ to regulate of the interaction.
Single-molecule interaction between LipL32 and TLR2 as determined by AFM. To further investigate the mechanisms behind rLipL32-rTLR2 interaction, a powerful tool for investigation of protein-protein interaction, AFM, was employed to determine the role of the active components in the rLipL32-rTLR2 interaction. rTLR2 was first deposited through covalent modification on a mica surface, while the tip was modified with rLipL32 or its variants to investigate binding interactions. The force-distance curves were recorded and tip-only and Pam 3 CSK 4 were used as mock and positive controls, respectively (Fig. 5A). The AFM tip modified with Pam 3 CSK 4 was not only used as a positive control for interaction between rTLR2 and its ligand but also to confirm the activity of rTLR2 on a mica surface. The interaction force in the force-distance curve between rTLR2 and Pam 3 CSK 4 was measured and the maximum distribution force was calculated as 67.4 ± 2.4 pN (Fig. 5A). When neutralizing by anti-TLR2 antibody, the maximum distribution force was significantly decreased to 23.1 ± 1.8 pN (p < 0.05). This result indicated that the interaction directly contributed from the Pam 3 CSK 4 -TLR2 interaction. The force-distance curve was also used to determine the rLipL32WT-rTLR2 interaction, and the maximum distribution force was calculated as 55.3 ± 1.5 pN 25,26 (Fig. 5B). We next examined the interaction forces between rLipL32 truncated variants and rTLR2. rLipL32ΔNβ1β2 and rLipL32ΔCα4 showed significantly decreased interaction forces between rLipL32 and rTLR2 (p < 0.001) (Fig. 5B). In contrast, ΔCenα3 of rLipL32 The different truncated and mutated variants of LipL32 were modified on AFM tip and the interaction forces to TLR2 were therefore measured. The anti-TLR2 antibody was used to neutralize the interaction between LipL32 and TLR2 and the interaction forces were calculated. (C) The binding frequency between LipL32 truncated and mutated variants and TLR2 protein. The binding frequency was calculated from the interaction of different LipL32 variants and TLR2 protein. The anti-TLR2 antibody was used to neutralize the interaction between LipL32 and TLR2 and the binding frequency were calculated. Gray bars, LipL32-TLR2 interaction; white bars, anti-TLR2 antibody neutralization. *p < 0.05; **p < 0.01; ***p < 0.001. revealed no significant difference in interaction force between rLipL32 and rTLR2. The result is consistent with the ELISA assays, which suggested that the binding domains were located at the N-and C-terminus. In addition, an anti-TLR2 antibody was used to neutralize the interaction between rLipL32 variants and rTLR2 and binding interactions were measured by AFM. These results showed that addition of anti-TLR2 antibodies significantly decreased the interaction force (p < 0.05) compared to that in the control (Fig. 5B). In addition, the binding frequency between rLipL32 variants and rTLR2 were determined indicating that rLipL32ΔNβ1β2 significantly diminished the binding frequency between rLipL32 and rTLR2 (p < 0.001) (Fig. 5C). In contrast, rLip-L32ΔCenα3 and rLipL32ΔCα4 showed no significant difference in binding frequency. Together, these results suggest that the N-terminal β1β2 and C-terminal α4 domains play a major role in promoting interaction between rLipL32 and rTLR2.
To further elucidate the interacting domain, point mutations of rLipL32, including F34S, V35S, L36S, and L263S variants, were used and interaction forces were determined from AFM force-distance curves. The results revealed that V35S (p < 0.01), L36S (p < 0.01), and L263S (p < 0.05) showed significantly reduced interaction force compared to WT, whereas F34S showed no significant difference compared to WT measurements (Fig. 5B). Additionally, when using anti-TLR2 antibody to neutralize the LipL32-TLR2 interaction, the interaction forces and binding frequency of rLipL32-rTLR2 interaction were significantly decreased (p < 0.05) as compared to those without antibody treatments (Fig. 5B,C).
Peptides inhibit the interaction between LipL32 and TLR2. We next examined the effect of synthesized peptides on the rLipL32-rTLR2 interaction. Synthesized peptides were based on the sequence of LipL32 and used to inhibit the rLipL32WT-rTLR2 interaction (Fig. 6). ELISA was used to measure the interaction of rLipL32-rTLR2, and 1 μM synthesized peptides including, N20, N50, C25, and scrambled peptides, were used to compete with the interaction between rLipL32WT and rTLR2. The results showed that N20 (p < 0.001), N50 (p < 0.001), and C25 (p < 0.01) significantly inhibited the rLipL32-rTLR2 interaction (Fig. 6A). In addition, dose dependent ELISA was used to measure the half maximum inhibition (IC 50 ) of the rLipL32-rTLR2 interaction. The IC 50 of N20, N50, C25, and scrambled peptides were calculated as 0.09 ± 0.05 μM, 0.07 ± 0.04 μM, 0.14 ± 0.01 μM, and 13.9 ± 1.2 μM, respectively (Fig. 6B). These results suggest the two terminal peptides are rTLR2 binding candidates that compete in binding of rTLR2 to rLipL32. In addition, AFM was used to measure rLipL-32WT-rTLR2 interaction in the presence of competing synthesized peptides. Interaction forces were similar in the presence of the synthesized peptides. However, the binding frequency of the rLipL32-rTLR2 interaction was significantly decreased in the presence of N20 and N50 peptides (p < 0.01). This result suggests that the N20 and N50 synthesized peptides play a major role in inhibiting the rLipL32-rTLR2 interaction. The inhibition mediated by these synthesized peptides also verifies that the N terminal β1β2 domains are critical for interacting with rTLR2.

Discussion
The TLR family comprises pattern recognition receptors and their ligands originate from microbes and contain multiple conserved patterns that can be recognized by the TLR LRR domain 34,44 . TLR2 protein forms homo-or hetero-dimeric structure with TLR1 or TLR6 to specifically recognize a wide range of ligands. Previous study indicated that the HEK293 cell transfected with pUNO-TLR2 could stimulated inflammatory responses by adding the Leptospira LipL32 indicated that TLR2 homodimer were responsible for LipL32 stimulation 27 . Several conserved patterns from bacteria have been identified as TLR2 ligands, including lipidation molecules (e.g. lipoprotein, lipopeptide, PGN, LTA, etc.) 35 , leucine-rich repeat family proteins (BspA) 38 , proteins with TLR2 binding sequence (PorB) 45 , and proteins with the specific structures for TLR2 binding (enterotoxins) 39 . Crystal structure of the TLR2/1 complex associated with the ligand, Pam 3 CSK 4 , revealed that the lipid molecule of the lipoprotein supported the binding to TLR2 35 . Previous studies showed that the LipL32 protein is modified with lipid molecules at the N-terminus [16][17][18] . Interesting, other study showed that non-lipidated polypeptides lack stimulation activity via TLR2 pathway 46 . However, an increasing studies revealed that polypeptides ligands could interact to TLR2 and stimulate or inhibit the inflammatory responses through TLR2 pathway 36,38,39,45 .
In our study, deletion of the signal peptide of rLipL32 did not prevent binding to rTLR2, and it was still able to stimulate inflammatory responses, suggesting that other domains of rLipL32 are responsible for interacting with rTLR2 25,26,28 . The mature form of rLipL32 protein is a lipoprotein in which the first 20 residues are removed. Therefore, the rLipL32ΔN20 construct is considered as wild-type rLipL32 protein for comparison with other truncated variants. Crystal structure of rLipL32, a jellyroll conformation without the LRR domain in its structure [19][20][21] , appears unlikely to use the LRR domain to interact with rTLR2 38 . In addition, structure alignment showed that the structure of rLipL32 is substantially different from that of enterotoxin B 39 . Moreover, sequence alignment of rLipL32 with and other TLR2 binding proteins such as BspA, PorB, and LcrV suggested that the homology of these proteins are low (<10%), and TLR2 binding motifs from BspA, PorB, and LcrV were not found in LipL32. These results indicate that rLipL32 is a unique protein with specific sequence and structure that different from the known TLR2 ligands. As a novel TLR2 ligand, rLipL32 is suggested to utilize its distinctive structure and mechanism to interact with TLR2.
The endotoxin contamination from the recombinant protein interfere the intrinsic inflammatory properties of the innate immunity system. To overcome this problem, the endotoxin locked expression system, ClearColi TM BL21(DE3) was used to produced endotoxin free recombinant proteins 47 . The yield of protein expression are similar to that expressed in BL21(DE3) competent cell. In addition, the anion-exchange chromatography (Mono-Q) and polymyxin B resin were further used to remove contaminating endotoxin from the purified rLipL32 variants 41 . In the LAL assay, it was observed that the endotoxin was significantly removed. In order to prevent the interference of the resident endotoxin, the HEK293-TLR2 cell naturally not expressing TLR4 was used to measure the inflammatory responses. Therefore, the responses from the stimulation of HEK293-TLR2 were directly from the rLipL32 and its variants.
The synthesized peptides derived from LipL32 were used to interact with rTLR2. However, the affinity of synthesized peptides to rTLR2 was low as determined in ELISA and AFM analysis. Previous study in BspA protein revealed that the synthesis truncated peptides (LRR domains) could induce responses to interact with rTLR2. We speculated that the synthesis peptides derived from LipL32 lacking suitable conformation to interact with rTLR2. Interestingly, the synthesized peptides, N20 and N50 peptides were used to compete the interaction between rLipL32 and rTLR2 (Fig. 6). However, the C25 peptide failed to compete this interaction indicated that C25 peptide is not the major domain to interact with rTLR2. The C25 domain of LipL32 is suggested as the assisted role in rLipL32-rTLR2 interaction.
Furthermore, the residues at N-and C-termini of rLipL32 are investigated to find out the residues responsible for interacting to rTLR2. In the predicted complex, several resides of rLipL32 with nearest distance to rTLR2 were replace to Serine residues and the mutation variants were purified to measure the affinity to rTLR2. Interestingly, our prediction rLipL32-rTLR2 complex and experimental data also suggest that three hydrophobic residues, including Val 35 , Leu 36 , and Leu 263 , are involved in rLipL32-rTLR2 complex interface. Similarly, protein complex of SSL3-TLR2 was solved and the residues involved in complex formation were proposed 36 . The interface of the SSL3-TLR2 complex is a hydrophobic region which contains Phe 156 , Phe 158 , Leu 160 , and Pro 194 of SSL3 as well as Phe 349 , Leu 350 , Gln 375 , Tyr 376 , and Asn 379 of TLR2. In addition, the pentameric B subunit of type IIb E. coli enterotoxin (LT-IIb-B 5 ) also suggests its hydrophobic upper-pore region directly binds to TLR2 and four residues, including Met 69 , Ala 70 , Leu 73 , and Ser 74 are responsible for TLR2-binding. Therefore, protein ligands for TLR2 with the hydrophobic patch may play vital roles in interacting to TLR2. In summary, both computational prediction of the complex and experimental data suggest that rLipL32 interacts with rTLR2 through both N and C termini. The predicted model suggested that the N-terminus is inserted into the LRR domain of rTLR2 and C-terminus assists in the binding of LipL32 to TLR2. The two domains of LipL32 form a plier-like structure to interact with the rTLR2 LRR domain. Val 35 , Leu 36 , and Leu 263 at N and C terminus play pivotal roles in interaction to TLR2, respectively. The corresponding binding residues of rTLR2 were predicted using nearest distance prediction and the results show the hydrophobic interaction of Phe 349 and Leu 371 of rTLR2 to Leu 36 of rLipL32; Leu 350 , Gln 357 of rTLR2 to Leu 263 of rLipL32, respectively (Fig. 7). Interestingly, the binding domains of rTLR2 to rLipL32 are the domains that bind other ligands as described previously 35,36 . This plier-like structure of rLipL32 from the pathogenic Leptospira serves as the novel ligand described to interact with rTLR2. Improved understanding of the ligands and binding mechanisms of innate immune components provides further information about how host cells interact with pathogens and suggests potential strategies for drug development.  Table 1. These cloning products from DNA ligation were transformed into JM109 Competent E. coli and the DNA sequences were verified by DNA sequencing.

Reagents.
The human tlr2 gene was amplified by PCR using the primers listed in Table 1 and the template pUNO-hTLR2 purchased from Invivogen (San Diego, CA). The PCR product was verified by agarose gel electrophoresis and the PCR product was cloned into the pLenti6.3/V5-TOPO lentiviral vector (Invitrogen). Following TOPO ligation, One Shot Stbl3 Chemically Competent E. coli were transformed with pLenti6.3/V5-TOPO-TLR2. The real cloned DNA with the right size and right orientation was verified by restriction digests and confirmed by DNA sequencing. The correct clone of p-TLR2-Lenti plasmids was used for stable gene expression in culture HEK293 cells.  ViraPower; Invitrogen) to generate the lentivirus according to manufacturer's protocol. HEK293FT cells were transfected with the lentivirus and stable cell lines were generated by selecting with blasticidin. Cells were collected and named HEK293-TLR2. Cells were grown in DMEM with 10% FBS and maintained in a 37 °C humidified atmosphere containing 5% CO 2 .
Protein expression and purification. The DNA plasmid of lipl32 and its variants were transformed into the expression host cell E. coli ClearColiTM TM BL21 (DE3) pLys (Lucigen, Middleton, WI). The transformants of His 6 -LipL32 and its variants were grown in Luria broth medium with 50 μg/ml ampicillin at 37 °C to an OD 600 of ~0.6. Isopropyl 1-thio-β-galactopyranoside (500 μM) was subsequently added to induce the expression of LipL32 protein for additional 24 h at 16 °C. E. coli cells were harvested by centrifugation at 4,000 x g for 15 min and then sonicated in phosphate buffered saline (PBS buffer). The cell debris was discarded after centrifugation at 14,000 xg for 30 min and the supernatant absorbed to Ni 2+ -nitrilotriacetic acid (Ni 2+ -NTA) agarose resin (GE Healthcare Life Sciences, Chalfont) for affinity chromatography purification 25,26,48 . rLipL32 protein and its variants were eluted in PBS buffer containing 250 mM imidazole and the elution fractions were dialysis against the PBS buffer to remove the imidazole and applied to mono Q column ion exchanged column (GE Healthcare Life Sciences, Chalfont) for further removal the contaminated protein and endotoxins. Furthermore, to detach His 6 -tag, the rLipL32 protein and its variants were incubated with 0.2 mg/ml enterokinase (EC 3.4.21.9; Invitrogen) at 37 °C for 16 h 25,49 . The tag peptide was further removed by size exclusion chromatography using Superdex 200 pg 16/600 column (GE Healthcare Life Sciences, Chalfont). The purified rLipL32 and its variants (100-150 μg) were injected into the column and the flow rate was set at 1 ml/min at 4 °C. To validate the inflammatory effects of rLipL32, the protein were subjected to polymyxin B (Invitrogen), heat, and proteinase K (Invitrogen) treatments, respectively, as described previously 11,25 . The Limulus Amebocyte Lysate (LAL) test was further used to detect the recombinant proteins and measured the endotoxin contamination according to the manufacturer's guidelines (Lonza, Switzerland).
HEK293-TLR2 was grown in the medium of DMEM with 10% FBS and maintained in a 37 °C humidified atmosphere containing 5% CO 2 for human rTLR2 protein expression. The cell was suspended in PBS buffer containing 0.5% CHAPS for rTLR2 protein solubilization. The cell debris was removed after centrifugation at 14,000 × g for 30 min and the supernatant absorbed to anti-V5 antibody activated NHS resin (GE Healthcare Life Sciences, Chalfont) for affinity chromatography purification. rTLR2 protein was eluted in 25 mM Glycine-HCl buffer (pH 2.5) and following by neutralization in the PBS buffer containing 0.5% CHAPS. Circular dichroism. The secondary structures of rLipL32WT and deletion variants were analyzed at 25 °C on an CD Aviv 202 spectrometer (Aviv Biomedical, Lakewood, NJ) using a 1-mm path length cuvette. The resulting spectra were corrected to remove buffer signal and smoothened using SigmaPlot 10.0. The obtained CD spectra were reported as averages of no less than three scans for each sample.

Stains-all binding assay.
Stains-all is a carbocyanine probe and could be used to detect the Ca 2+ binding protein. rLipL32WT and its variants (15 μM) were mixed with stains-all dye solution (500 μM) in 2mM MOPS (pH 7.2) containing 30% (v/v) ethylene glycol and incubated for 10 min. The CD spectra was used to scan from 700 to 400 nm and the characteristic peak at 660 nm (J band) was recorded. The obtained CD spectra were reported as averages of no less than three scans for each sample.

RNA extraction and real-time PCR (RT-PCR).
Total RNA was extracted according to the previous protocol with minor modifications 27,28 . Real-time PCR was performed following manufacturer's instructions using an ABI Prism 7700 with SYBR green I as a double-stranded DNA-specific dye (PE-Applied Biosystems, Cheshire, UK). Primers of IL-8, TNF-α, MCP-1, and hMMP7 were listed in Table 1 and were constructed to be compatible with a single RT-PCR thermal profile (95 °C for 10 min, 40 cycles at 95 °C for 30 s, and 60 °C for 1 min). The accumulation of the PCR product was recorded in real time (PE-Applied Biosystems). In addition, the TaqMan gene expression assays were used to measure GRO-α (Hs00236937_m1, Applied Biosystems) and SPP1 (Hs00959010_m1, Applied Biosystems) mRNA expression. The results of the mRNA levels for the different genes are displayed as transcript levels of the analyzed genes relative to GAPDH. The statistical analyses were performed with Student's t test. Statistically significant were considered as p-values of ≤0.05. Molecular docking. The PDB files corresponding to Ca 2+ -bound (PDB:2WFK 20 ) and Ca 2+ -free (PDB:3FRL 19 ) LipL32 and TLR2 LRR domain (PDB: 2Z7X, chain A 35 ) were submitted to automated docking using the ClusPro 2.0 server (available on-line 40,[50][51][52]. Each protein was analyzed both as receptor and ligand. The 10 highest scoring models were analyzed using PyMOL (available on-line). The 10 possible binding models were divided in to 3 groups according to the binding domains of LipL32 interacted with TLR2.
Enzyme-linked immunosorbent assay (ELISA). The interaction between rLipL32 and rTLR2 was investigated by ELISA according to the prior method with minor modifications 20,25 . The ELISA plates (Nunc-Immuno Plate MaxiSorp surface) were primarily coated with 1 μg rTLR2 protein in 100 μl PBS buffer and incubated for 2 h at 37 °C. The wells were washed with PBST (PBS containing 0.05% (v/v) Tween 20) and then blocked with 1% (w/v) bovine serum albumin (BSA) for 1 h at 37 °C. The plates were incubated overnight at 4 °C. Protein samples (rLipL32WT and variants; from 0 to 2 μM) in PBS buffer were added to the attached rTLR2 protein for 90 min at 37 °C and then washed with PBST. The bound rLipL32 protein was detected by adding 100 μl anti-His 6 antibody (1:5,000 dilution) (Proteintech, Rosemont, IL) incubated for 1 h and then washed with PBST. Subsequently, 100 μl horseradish peroxidase conjugated donkey anti-rabbit immunoglobulin G (1:5,000 dilution) was added and incubated for 1 h at 37 °C. After washing with PBST, the 3,3′,5,5′-tetramethylbenzidine (TMB) was added for color development. The reaction was allowed to proceed for 15 min and then terminated by adding 50 μl 2 N H 2 SO 4 . The absorbance was measured at 450 nm in an ELISA reader (BioTek, Winooski, VT). These experiments were accomplished in three independent experiments and the mean absorbance ± S.D. of three independent data sets was calculated. The absorbance at each data point was corrected by subtracting a negative control from BSA coated well. The dissociation constant (K d ) according to the prior protocol 53 , based on the following equation: max where A is the absorbance at a set protein concentration, A max is the maximum absorbance for the ELISA plate reader (equilibrium), [protein] is the protein concentration, and K d is the dissociation equilibrium constant for a given absorbance at a set protein concentration (ELISA data point). The binding data were assayed via SigmaPlot 10.0 program by fitting the data to the optimal equation 54 assuming that ligands bound to one independent site. The statistical analyses were performed with Student's t test. Statistically significant were considered as p-values of ≤0.05.

AFM tip and mica functionalization.
Silicon nitride AFM cantilever, PNP-TR (NanoWorld, Neuchâtel, Switzerland), was cleaned by sonication in a series of solvents and functionalization in 3-aminopropyl-trimethoxysilane and glutaraldehyde (Grade II, Sigma) according to the method reported previously with minor modifications 25,26,55 . For anchoring proteins, the tips were rinsed with PBS to remove the glutaraldehyde remainders and subsequently incubated with proteins, including BSA, Pam 3 CSK 4 , rLipL32 and its variants for 1 h at room temperature. The functionalized AFM tips were then rinsed three times with PBS to remove the unbound proteins 56 . The mica surface was modified for better deposition of proteins according to previous reports 57, 58 . The unbound proteins were removed and the proteins were fixed on the mica for later AFM measurements.
AFM observation. The functionalized cantilever tips used in this work had a spring constant in the range of 0.02-0.08 N/m as determined from the amplitude of their thermal vibrations. A commercial atomic force microscope (Nanoscope III, Digital Instruments, Santa Barbara, CA) with a E type scanner was employed throughout this study. The force volume software takes a force curve at each of the 999 points during a two dimensional scan over a sample surface. The X-Y scan size was 150 μm, and the Z scan distance was 5 μm at a rate of 1 Hz. The force applied to the protein modified mica surface was kept below 400 pN. The distance-force curves and force parameters were obtained according to the methods described previously 25,26 . For antibody neutralization analysis, the protein modified mica was treated with anti-TLR2 antibodies (diluted 1:1,000 in PBS) (Invivogen, San Diego, CA) for 1 h followed by five washes with PBS to remove unbound antibodies. For the binding events, force curves from at least 10 areas on each surface were independently selected for analysis. All the measurements described above were performed with modified tips and showed repeatedly similar results. The statistical analyses were performed with Student's t test. Statistically significant were considered as p-values of ≤0.05.

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).