Toll-Like Receptor 4 Decoy, TOY, Attenuates Gram-Negative Bacterial Sepsis

Lipopolysaccharide (LPS), the Gram-negative bacterial outer membrane glycolipid, induces sepsis through its interaction with myeloid differentiation protein-2 (MD-2) and Toll-like receptor 4 (TLR4). To block interaction between LPS/MD-2 complex and TLR4, we designed and generated soluble fusion proteins capable of binding MD-2, dubbed TLR4 decoy receptor (TOY) using ‘the Hybrid leucine-rich repeats (LRR) technique’. TOY contains the MD-2 binding ectodomain of TLR4, the LRR motif of hagfish variable lymphocyte receptor (VLR), and the Fc domain of IgG1 to make it soluble, productive, and functional. TOY exhibited strong binding to MD-2, but not to the extracellular matrix (ECM), resulting in a favorable pharmacokinetic profile in vivo. TOY significantly extended the lifespan, when administered in either preventive or therapeutic manners, in both the LPS- and cecal ligation/puncture-induced sepsis models in mice. TOY markedly attenuated LPS-triggered NF-κB activation, secretion of proinflammatory cytokines, and thrombus formation in multiple organs. Taken together, the targeting strategy for sequestration of LPS/MD-2 complex using the decoy receptor TOY is effective in treating LPS- and bacteria-induced sepsis; furthermore, the strategy used in TOY development can be applied to the generation of other novel decoy receptor proteins.

Human TLR4 contains a 608-residue extracellular domain, a single transmembrane domain, and a 187-residue intracellular domain [13]. Crystal structural analysis has shown that TLR4 adopts the characteristic horseshoe-like shape of the LRR superfamily, with N-terminal (amino acids , central (amino acids 203-348), and C-terminal (amino acids 349-582) domains [14,15]. MD-2 binds to the concave surface of the N-terminal and central domains of TLR4 [14]. In addition to TLR4-bound MD-2, the MD-2 protein is also secreted into the extracellular milieu in a soluble form, which is present in circulating blood [16,17]. MD-2 has a b-cup fold structure that forms a hydrophobic pocket for LPS binding [14,18]. Binding of the LPS/MD-2 complex to TLR4 causes TLR4 dimerization, and results in the activation of NF-kB leading to acute and severe inflammation and sepsis [7][8][9]19].
In this regard, blocking TLR4 signaling activation using a decoy receptor could be an effective way to prevent LPS-or Gramnegative bacteria-induced sepsis if applied prior to or after challenge. Although two research groups have tried to generate the extracellular domain of TLR4 protein as a decoy receptor for MD-2 [17,20], it proved difficult to generate a substantial amount of the protein because it is insoluble, its production rate is extremely low, and it is hard to purify due to its intrinsic biochemical properties. Therefore, the trial of blocking TLR4 using a decoy receptor in preventing sepsis under in vivo conditions has not been achieved until now. To overcome these problems, we recently developed a novel method, 'the Hybrid LRR Technique [14]', to generate a massive amount of soluble extracellular domains of TLR4 protein. Variable lymphocyte receptors (VLRs) are a new type of immune receptors in jawless fish. These receptors resemble the adaptive immune receptors in jawed vertebrates. VLRs and TLRs commonly contain the LRR domain in the extracellular fragment, which is composed of a signal sequence, an N-terminal cap (LRRNT), several LRR modules, and a C-terminal cap (LRRCT) [14,21]. Therefore, stable TLR4-VLR hybrid proteins can be generated in large amounts without any loss of the intrinsic structural integrity of TLR4 by replacing some LRR modules and LRRCT of TLR4 with those of VLR, termed TV3 and TV8 [14]. The TV3 and TV8 proteins are able to complex with MD-2 and Eritoran (a synthetic LPS antagonist) [22]. By fusion of the Fc domain of IgG1 to TV3 and TV8, we were able to generate dimeric TLR4 decoy receptor TOY, which is effective in treating LPS-and bacteria-induced sepsis.

Results and Discussion
We fused human IgG-Fc to the carboxy-terminal portion of TV3 and TV8, and named the resulting products TLR4 decoy receptor-3 and -8 (TOY3 and TOY8) ( Figure 1A). As a control, we also produced a fusion of IgG-Fc to the TLR4 full-length ectodomain (TFE; Figure 1A). We used computer modeling based on crystal structures to illustrate the potential structures of the three constructs ( Figure 1B). We were able to produce the TFE, TOY3, and TOY8 proteins in CHO cells at rates of 0.1, ,20, and ,15 mg/L. The production rates of TOY3 and TOY8 are currently amplified by the methotrexate selection process [23]. The purified TFE, TOY3, and TOY8 under reducing conditions  revealed predominantly single bands of the expected molecular  masses of ,110, ,65 and, ,105 kDa, respectively ( Figure 1C). Under non-reducing conditions, the recombinant proteins were present as disulfide-linked dimers due to the presence of the Fc domain ( Figure 1C). In vitro binding analysis revealed that TFE, TOY3, or TOY8 could interact not only with MD-2 but also with LPS/MD-2 complex (Figure 2 and Figure S1). Surface plasmon resonance analyses revealed that all three proteins directly interacted with MD-2, and the K D of TFE, TOY3, and TOY8 binding to MD-2 was ,81, ,76, and ,56 nM ( Figure 2C). Thus, the VLR component in TOY did not substantially alter the binding affinity for MD-2. In vitro analyses demonstrated that both TOY3 and TOY8 largely inhibited LPS-induced NF-kB activation in primary cultured lymphatic endothelial cells ( Figure 3A and 3B) and also diminished LPS-induced TNF-a secretion in macrophages ( Figure 3C).
The isoelectric point (pI) of a recombinant protein affects its bioavailability and pharmacokinetics in vivo. High-pI proteins suffer from poor bioavailability because they adhere nonspecifically to the negatively charged proteoglycans of the extracellular matrix (ECM). While the theoretical pI values of TFE, TOY3, and TOY8 are 5.7, 11.8, and 6.0, their actual pI values are 5.0, 5.2, and 5.5, possibly due to abundant N-linked glycosylation ( Figure 4). Indeed, the ECM binding assay demonstrated that TFE, TOY3, and TOY8 had minimal binding to ECM, whereas Flt1-Fc (pI 9.1) bound strongly to ECM ( Figure 5A). To examine the in vivo pharmacokinetic profiles of the proteins, we performed a single intraperitoneal injection (5 mg/kg) of TOY3, TOY8, or Flt1-Fc recombinant protein into mice. TOY3 and TOY8 proteins were rapidly absorbed from the peritoneal cavity into systemic circulation. The proteins reached maximum levels in blood at 1,2 hr after injection, and their half-lives (t 1/2 ) were ,2 days ( Figure 5B). In mice (n = 4), TOY3 had a maximal concentration (C max ) of 7.0060.13 mg/ml and AUC (total ''area under the curve of concentration'') of 20.5060.96, TOY8 had a C max of 7.0360.85 mg/ml and AUC of 18.2761.24, and Flt1-Fc had C max of 5.2860.49 mg/ml and AUC of 9.9660.84. Thus, the pharmacokinetic profiles of the proteins were correlated to their in vitro ECM adhesion properties. TOY3 and TOY8 have relatively high bioavailability and excellent pharmacokinetic profiles in vivo, raising the possibility that TOY could block LPS/MD-2/TLR4 signaling in vivo by providing a decoy for TLR4.
To explore whether the TOY proteins could ameliorate sepsis in vivo, eight experiments were performed. For the first and second experiment, 20 mg/kg of TOY3 or TOY8 was given to mice at 30 min prior to or at 1 hr after administration of LPS (15 mg/kg) ('preventive model' or 'therapeutic model'). Both TOY3-and TOY8-treated mice had an extended lifespan compared to Fctreated mice (50% survival: ,58 hr versus ,22 hr, P,0.001 in the preventive model; ,49 hr versus ,22 hr, P,0.001 in the therapeutic model) ( Figure 6A and 6B). For the third and fourth experiment, 20 mg/kg of TOY3 or TOY8 was given to the mice at 1 hr prior to or at 1 hr after generation of acute peritonitis by cecal ligation and puncture (CLP) ('preventive model' or 'therapeutic model'). Both TOY3-and TOY8-treated mice had a prolonged lifespan compared to Fc-treated mice (50% survival: . The x-axis numbering represents the number of each well in (A). (C) Sensorgrams for the association and dissociation of MD-2 on immobilized TFE, TOY3, or TOY8. One microgram of TFE, TOY3, or TOY8 was immobilized on a Sensor Chip CM5 (BIAcore) using N-hydroxysuccinimide (NHS) and 1-ethyl-3(3-dimethylaminopropyl) carbodiimide (EDC) amine coupling reagent at approximately 2,000 resonance units (RU). As a control, BSA protein was immobilized on another portion of the same chip. Recombinant MD-2 proteins were then applied onto the immobilized TFE, TOY3, or TOY8 surfaces, and the amount captured was recorded in sensorgrams as RU. All samples were in running buffer to minimize bulk effects. The kinetic parameters of the binding interactions were calculated from the sensorgrams by nonlinear curve fitting using BIAEVALUATION software (BIAcore). RU represents resonance units. doi:10.1371/journal.pone.0007403.g002 ,74 hr versus ,29 hr, P,0.001 in the preventive model; ,60 hr versus ,32 hr, P,0.001 in the therapeutic model) ( Figure 6C and 6D). Thus, surprisingly, both TOY3 and TOY8 showed prominent preventive and therapeutic effects in the LPS-and CLP-induced sepsis models. For the fifth experiment, 20 mg/kg of TOY3 was given to the mice at 1 hr and 12 hr after administration of LPS (15 mg/kg) ('repeated treatment to therapeutic model'). The TOY3-treated mice had a markedly extended lifespan compared to Fc-treated mice (50% survival: ,60 hr versus ,26 hr, P,0.001), and the 40% of TOY3-treated mice were completely rescued to live ( Figure 6E). For the sixth experiment, 20 mg/kg of TOY3 was given to the mice at 1 hr and 12 hr after generation of CLP ('repeated treatment to therapeutic model'). The repeated administration of TOY3 led to a dramatic increase in survival up to 96 hr after CLP (TOY3 versus Fc; 60% versus 0%, P,0.001) ( Figure 6F). Thus, the repeated treatments with TOY3 markedly attenuated lethality in both the LPS-and CLP-induced sepsis models compared to the single treatment of TOY3 (50% survival: ,60 hr versus ,50 hr, P,0.01 in the LPSinduced sepsis model; 70% survival: ,60 hr versus ,50 hr, P,0.01 in the CLP-induced sepsis model). For the seventh and eighth experiments, 20 mg/kg of TOY3 was given to mice at 6 hr and 12 hr after administration of LPS (15 mg/kg) or the CLP procedure ('clinically relevant model'). These repeated administrations of TOY3 also significantly improved survival compared to the controls in both the LPS-and CLP-induced sepsis models (50% survival: ,69 hr versus ,21 hr, P,0.001 in the LPSinduced sepsis model; ,84 hr versus ,32 hr, P,0.001 in the CLP-induced sepsis model) ( Figure 7A and 7B). These repeated treatments with TOY3 markedly attenuated lethality in both the LPS-and CLP-induced sepsis models compared to the single treatment of TOY3 (50% survival: ,70 hr versus ,50 hr, P,0.01 in the LPS-induced sepsis model; ,84 hr versus ,60 hr, P,0.01 in the CLP-induced sepsis model). Therefore, repeated treatments with TOY3 could ameliorate sepsis even in a more clinically relevant situation. Gram-negative bacterial sepsis is characterized by increased secretions of proinflammatory cytokines and thrombus formation in blood vessels, and both effects are brought through activation of NF-kB [2,24]. To determine whether TOY attenuates NF-kB activation and subsequent secretion of proinflammatory cytokines and thrombus formation, we monitored these parameters in the LPS 'therapeutic model' using NFkB-RE-luc [25] and wild-type mice. At 24 hr after LPS (1 mg/kg) administration, the luminescence signal examined by IVIS imaging system was markedly upregulated in most regions of NFkB-RE-luc mice, whereas the signal was barely detected in PBS-treated NFkB-RE-luc mice ( Figure 8A and 8B). Both TOY3 and TOY8 significantly attenuated the LPS-induced up-regulation of the signal in the NFkB-RE-luc mice ( Figure 8A and 8B). At 6 or 24 hr after LPS (7.5 mg/kg)  administration, levels of TNF-a, IL-1b, and IL-6 in plasma and thrombi in the blood vessels of liver, adrenal cortex, lung, and brain were profoundly increased in the wild-type mice ( Figure 8C and 8D). The mice treated with either TOY3 or TOY8 had markedly reduced levels of TNF-a, IL-1b, and IL-6 in plasma ( Figure 8C), and fewer thrombi in the blood vessels ( Figure 8D). Thus, TOY attenuated LPS-induced NF-kB activation, which increases the secretion of proinflammatory cytokines and results in thrombus formation in multiple organs.
In this study, we report for the first time the production of a TLR4 decoy receptor, TOY, which can be easily generated in large amounts. We also show for the first time that a modified decoy receptor of TLR4, TOY, is effective not only for prevention, but also for treatment of LPS-and bacteria-induced sepsis in mice. Because TOY would be used only once or twice during treatment of a life-threatening septic condition, the immune reaction against to TOY is not being considered. TOY3 and TOY8 produced almost identical effects in preventing sepsis; since TOY3 bears only the 'A patch' of the TLR4, this result implies that the 'A patch' of the TLR4 ectodomain is sufficient for MD-2 binding in vivo [14]. These findings give us additional structural information in TLR4 biology which shows that further application of TLR4 can be achieved with only the MD-2 minimal binding portion. Considering that TOY3 has a slightly higher production rate and smaller molecular size, it will be the most favorable construct for therapeutic use of TOY in the future. Recently, Roger et al. reported that a blocking antibody against the N-terminal and central domains of TLR4 also produced preventive and therapeutic effects in LPS-and bacteria-induced sepsis models [26]. Thus, targeting TLR4 either by the blocking antibody or decoy receptor such as TOY could be an amenable tool to relieve LPS-and bacteria-induced sepsis [26,27]. In comparison, several lipid A analogs that are competitive inhibitors of LPS [22,28] are able to target only LPS-free MD-2, but not LPS-bound MD-2, whereas TOY could sequester both free and bound forms of MD-2. Furthermore, since LBP and CD14 play accessory roles in LPS recognition by TLR4/MD-2 while MD-2 binds to LPS directly and induces TLR4 dimerization [8,9], the efficacy of targeting TLR4/MD-2 using TOY could be different from that of blocking LBP [29] or CD14 [30]. Therefore, the effectiveness of TOY and blockers of LBP or CD14 should be carefully compared in future studies. Taken together, TOY would be an effective alternative therapeutic molecule for treatment of patients with bacterial sepsis, and our method provides a new platform biotechnology to generate novel decoy receptor from TLR proteins.

Generation of recombinant proteins
Gene constructs encoding different sizes of the ectodomain of human TLR4 [TFE (amino acid residues 27-631), TOY3 (amino acid residues 27-227), and TOY8 (amino acid residues 27-527)], the LRR module of hagfish VLR-B.61 (VLR), and the Fc domain of human IgG (Fc) were cloned into the pCMV-dhfr vector. Recombinant Chinese hamster ovary (rCHO) cells expressing TFE, TOY3, or TOY8 were established following a previously described method [23]. Briefly, the cells were established by transfection of a vector containing the dihydrofolate reductase (dhfr) and TFE, TOY3, or TOY8 gene into dhfr-deficient CHO cells (CRL-9096, American Type Culture Collection). This was followed by dhfr/methotrexate (MTX)-mediated gene amplification. Stable rCHO cells secreting the highest amount of TFE, TOY3, or TOY8 were selected with serially-increasing concentrations of MTX (0.001-0.08 mM, Sigma-Aldrich). Then the cells were grown and maintained in HyQ SFM4CHO (Hyclone) supplemented with 1% dialyzed fetal bovine serum (Invitrogen) and 0.08 mM MTX. After 4 days, the culture media containing recombinant proteins were harvested, and the recombinant TFE, TOY3, or TOY8 proteins were purified by using Protein-A sepharose affinity chromatography with subsequent acid elution and neutralization. After purification, each protein was quantitated using the Bradford assay and confirmed with Coomassie blue staining of an SDS-PAGE gel.
In vitro binding assay 200 ng of BSA or MD-2 was coated onto 96-well plates. After blocking with 1% BSA for 1 hr, 40 mg/ml of Fc, TFE, TOY3, or TOY8 was incubated in each well for another 1 hr with or without FITC-labeled LPS (10 mg/ml). An HRP-conjugated anti-Fc antibody was incubated in each well for 1 hr, and then HRP substrate was added to each well. The fluorescence signal was measured by IVIS imaging, and the absorbance was measured by microplate reader (Bio-Rad).

Surface plasmon resonance assay and isoelectric focusing
Binding between TFE, TOY3, or TOY8 and MD-2 were analyzed with the BIAcore 3000 (BIAcore AB). One microgram of TFE, TOY3, or TOY8 was immobilized on a Sensor Chip CM5 (BIAcore) using N-hydroxysuccinimide (NHS) and 1-ethyl-3(3dimethylaminopropyl) carbodiimide (EDC) amine coupling reagent at approximately 2,000 resonance units (RU). As a control, BSA protein was immobilized on another portion of the same chip. Recombinant MD-2 proteins were then applied onto the immobilized TFE, TOY3, or TOY8 surfaces, and the amount captured was recorded in sensorgrams as RU. All samples were in running buffer to minimize bulk effects. The kinetic parameters of the binding interactions were calculated from the sensorgrams by nonlinear curve fitting using BIAEVALUATION software (BIAcore). To measure isoelectric points, 20 mg of each protein sample and standard marker proteins were loaded on an IsoGel Agarose IEF Plate pH 3-10 strip (Cambrex) and run at 50 mA constant current for 3 hr using 1 M phosphoric acid at the anode and 1 M sodium hydroxide at the cathode. The gels were stained with Coomassie blue.

Glycosylation analysis of recombinant proteins
Two mg of TFE, TOY3, or TOY8 was incubated with Oglycosidase or PNGase F (QA-Bio) according to manufacturer's protocol, and then the protein was run on SDS-PAGE and the glycosylation pattern was analyzed by the migration shift after Coomassie blue staining.
In vitro assays for NF-kB and TNF-a NF-kB activity was assessed by immune-localization of p65 in nuclei of primary cultured lymphatic endothelial cells according to methods previously described [31]. Briefly, LEC (primary lymphatic microvascular endothelial cells derived from human adult dermis) were purchased from Cambrex Inc. (East Rutherford) and maintained in endothelial cell basal medium-2 with growth supplements (EBM-2 MV). Passage 4-6 LEC were incubated in EBM-2 MV containing 1 % FBS for 8 hr and then with 1 mg/ml of Fc, TFE, TOY3, or TOY8 for 15 min, and then NFkB-RE-luc mice were injected intraperitoneally with control PBS (C) or 20 mg/kg of Fc, TOY3 (T3), or TOY8 (T8) at 1 hr after intraperitoneal administration of C or LPS (1 mg/kg). At 24 hr later, the luminescence signals from the whole body were examined by IVIS imaging (A). The luminescence was quantified and expressed as radiance (photon/sec/cm 2 /steradian) (B). Bars represent means 6 S.D. (n = 3-4). *, P,0.05 versus C+C; # , P,0.05 versus C+LPS or Fc+LPS. (C and D) Wild-type mice were injected intraperitoneally with C or 20 mg/kg of Fc, T3, or T8 at 1 hr after intraperitoneal administration of LPS (7.5 mg/kg) or C. At 6 and 24 hr later, plasma levels of TNF-a, IL-1b, and IL-6 were measured (C). Bars represent means 6 S.D. (n = 4-5). *, P,0.05 versus Fc+LPS. At 24 hr after treatment, the mice were anesthetized, and samples of liver, adrenal gland, lung, and brain were H&E stained (D). Arrows indicate thrombi in the blood vessels. Scale bars, 100 mm. doi:10.1371/journal.pone.0007403.g008 the LEC were treated with LPS (500 ng/ml) for 30 min. Primary cultured macrophages from mouse peritoneal cavity were used to assay TNF-a secretion. Three days after intraperitoneal injection of 1 ml of 3% thioglycolate into 9-week-old male C57BL/6 mice, we harvested macrophages by peritoneal lavage with PBS. We incubated 1610 6 macrophages per experimental condition in RPMI 1640 (Lonza) supplemented with 10% dialyzed fetal bovine serum (Invitrogen). The macrophages were pre-incubated for 30 min with 1 mg/ml of Fc, TOY3, or TOY8, and then treated with LPS (100 ng/ml) for 4 hr. Culture media were sampled, and levels of TNF-a were measured by ELISA (R&D systems).

ECM binding assay and pharmacokinetic analysis
ECM-coated 96-well plates (Becton Dickinson) were used for ECM binding assays. Each recombinant protein (100 mg of Fc, TOY3, or TOY8) was injected subcutaneously into 8-week-old male C57BL/6 mice (,25 g body weight), then the amount of each recombinant protein in blood was measured at the indicated times by sandwich ELISA.

Animals
Specific pathogen-free 8-9-week-old male C3H/HeN mice and NFkB-RE-luc [25] (Balb/C) mice were used. Animal care and experimental procedures were performed under approval from the Animal Care Committees of KAIST. To generate the LPSinduced sepsis model, mice were injected intraperitoneally with 15 mg/kg of LPS (E. coli O111:B4; List Biological Laboratories). To generate the CLP-induced sepsis model, mice were anesthetized, and ,75% of the cecum was ligated and punctured with a 21-gauge needle. The mice received 20 mg/kg of Fc, TOY3, or TOY8 into the peritoneal cavity prior to or after the generation of sepsis.
Monitoring NF-kB activation, measurement of proinflammatory cytokines, and histology NFkB-RE-luc [25] and wild-type mice were given 1.0 and 7.5 mg/kg of LPS 1 hr prior to administration of 20 mg/kg of Fc, TOY3, or TOY8. At 24 hrs after the LPS administration, the luminescence signal was examined by the IVIS imaging system (Xenogen). Plasma was sampled at 6 and 24 hrs, and levels of TNF-a, IL-1b, and IL-6 were measured by ELISA (R&D systems). Mice were sacrificed 24 hours after LPS treatment, and their livers, adrenal glands, lungs, and brains were harvested for histological studies. Harvested organs were fixed using 4% PFA dissolved in PBS at 4uC overnight and embedded in paraffin blocks. Sections (4-mm) were stained with H&E and analyzed under a phase-contrast light microscope.

Statistics
Values are presented as means 6 S.D. Significant differences between means were determined by analysis of variance followed by the Student-Newman-Keuls test. For analysis of survival curves, a log-rank test was performed. Figure S1 In vitro binding analysis reveals that TFE, TOY3, or TOY8 could interact not only with MD-2 but also with LPS/MD-2 complex. (A) BSA or MD-2 was coated onto 96-well plates and 40 mg/ml of Fc, TFE, TOY3, or TOY8 was incubated in each well with or without FITC-labeled LPS. An HRP-conjugated anti-Fc antibody was incubated in each well, and then HRP substrate was added. The fluorescence signal of each well is shown. (B) Fluorescence and absorbance were measured. Fluorescence is expressed as radiance (photon/sec/cm2/steradian) on the left yaxis, and absorbance is shown on the right y-axis. Bars represent means 6 S.D. (n = 4). *, P,0.05 versus hMD-2+Fc (5); #, P,0.05 versus hMD-2+FITC-LPS+Fc (9). The x-axis numbering represents the number of each well in (A).