Selective Antibody Intervention of Toll-like Receptor 4 Activation through Fc γ Receptor Tethering

Background: Dysregulated leukocyte activation via Toll-like receptor 4 (TLR4) is central to numerous inflammatory disorders. Results: A novel mechanism of action involving Fc γ receptor tethering allows anti-TLR4 blocking antibodies to achieve increased potency on inflammatory leukocytes. Conclusion: This novel mechanism of action allows selective targeting of TLR4 activation during inflammation. Significance: The data provide a novel mechanism to dampen TLR4-mediated inflammatory disorders. Inflammation is mediated mainly by leukocytes that express both Toll-like receptor 4 (TLR4) and Fc γ receptors (FcγR). Dysregulated activation of leukocytes via exogenous and endogenous ligands of TLR4 results in a large number of inflammatory disorders that underlie a variety of human diseases. Thus, differentially blocking inflammatory cells while sparing structural cells, which are FcγR-negative, represents an elegant strategy when targeting the underlying causes of human diseases. Here, we report a novel tethering mechanism of the Fv and Fc portions of anti-TLR4 blocking antibodies that achieves increased potency on inflammatory cells. In the presence of ligand (e.g. lipopolysaccharide (LPS)), TLR4 traffics into glycolipoprotein microdomains, forming concentrated protein platforms that include FcγRs. This clustering produces a microenvironment allowing anti-TLR4 antibodies to co-engage TLR4 and FcγRs, increasing their avidity and thus substantially increasing their inhibitory potency. Tethering of antibodies to both TLR4 and FcγRs proves valuable in ameliorating inflammation in vivo. This novel mechanism of action therefore has the potential to enable selective intervention of relevant cell types in TLR4-driven diseases.


Inflammation is mediated mainly by leukocytes that express both Toll-like receptor 4 (TLR4) and Fc ␥ receptors (Fc␥R). Dysregulated activation of leukocytes via exogenous and endogenous ligands of TLR4 results in a large number of inflammatory disorders that underlie a variety of human diseases. Thus, differentially blocking inflammatory cells while sparing structural cells, which
are Fc␥R-negative, represents an elegant strategy when targeting the underlying causes of human diseases. Here, we report a novel tethering mechanism of the Fv and Fc portions of anti-TLR4 blocking antibodies that achieves increased potency on inflammatory cells. In the presence of ligand (e.g. lipopolysaccharide (LPS)), TLR4 traffics into glycolipoprotein microdomains, forming concentrated protein platforms that include Fc␥Rs. This clustering produces a microenvironment allowing anti-TLR4 antibodies to co-engage TLR4 and Fc␥Rs, increasing their avidity and thus substantially increasing their inhibitory potency. Tethering of antibodies to both TLR4 and Fc␥Rs proves valuable in ameliorating inflammation in vivo. This novel mechanism of action therefore has the potential to enable selective intervention of relevant cell types in TLR4-driven diseases.
TLRs 6 are part of a broad range of sensors forming the basis of our immune system's response against infectious pathogens (1). These receptors recognize pathogen-associated molecular patterns that are invariant among classes of pathogens and well conserved through evolution (2). Furthermore, the TLR family recognizes damage-associated molecular patterns (also known as danger-associated molecular pattern molecules) that initiate and perpetuate immune responses during inflammatory responses (3).
TLR4 is an extensively studied member of the TLR family (4). The first identified and best characterized TLR4-activating ligand is lipopolysaccharide (LPS), the major structural component of the cell wall of Gram-negative bacteria. The ability to react to LPS involves multiple proteins, both soluble and transmembrane in nature. The glycosylphosphatidylinositol membrane-anchored protein CD14 transmits LPS to the soluble protein MD-2 (myeloid differentiation factor-2), allowing for LPS association with TLR4 (5). Interaction of LPS with the MD-2⅐TLR4 complex induces structural rearrangements that trigger the dimerization and migration of TLR4 to lipid rafts with the subsequent formation of "activation clusters" facilitating signal transduction through the TLR4 cytoplasmic domain (6 -8). TLR4 is expressed on leukocyte subpopulations, such as monocytes, macrophages, dendritic cells, and neutrophils, which play pivotal roles in inflammatory processes (9,10). In addition, TLR4 is also expressed on structural cells (e.g. epithelial cells) and plays a role in tissue homeostasis (11).
In addition to TLRs, other receptors act as mediators of inflammation. For example, Fc␥Rs are expressed on cells responsible for initiating and maintaining inflammatory responses, such as monocytes, macrophages, dendritic cells, neutrophils, natural killer cells, and B cells. Fc␥Rs interact with the Fc portion of IgG. As such, Fc␥Rs play a pivotal role in antibody effector functions, including phagocytosis, the release of inflammatory mediators, and antibody-dependent cellular cytotoxicity (12). The human Fc␥R family includes Fc␥RI (CD64), Fc␥RIIA (CD32A), Fc␥RIIB (CD32B), Fc␥RIIC (CD32C), Fc␥RIIIA (CD16A), and Fc␥RIIIB (CD16B). Thus, Fc␥R biology represents a complex effector function system that has evolved to be exploited mainly by cells of the immune system (13).
We previously reported the characterization of a mouse IgG 1 anti-human TLR4 antibody, 15C1, that exploits a novel Fc␥RIIA-binding mechanism (14). In the current study, we demonstrate that a humanized version (of the original mouse antibody) with an engineered human IgG 1 backbone (Hu 15C1) engages both Fc␥RI and Fc␥RIIA, but not Fc␥RIII, increasing its inhibitory potency on inflammatory cells to block TLR4 signaling. The contribution of Fc engagement to the increased effect is dependent on the clustering of TLR4 with Fc␥Rs in lipid rafts following agonist ligation and is independent of signaling through Fc␥Rs. The Fc␥R contribution also increases the duration of inhibition of TLR4 activity. The benefit of this mechanism of action involving TLR4-Fc␥R co-engagement is demonstrated in vivo in a murine model of inflammation.

EXPERIMENTAL PROCEDURES
Reagents-Recombinant antibodies were produced in house using the Lonza CHO-GS mammalian expression system (Lonza Biologics, Slough, UK). Anti-human Fc␥RIIA mAb IV.3 was purchased from StemCell Technologies. Ultrapure LPS from Salmonella minnesota R595 (Re) and Ultrapure LPS from Escherichia coli 055:B5 were obtained from List Laboratories. The antibodies used for FRET studies were as follows. Anti-TLR4 mAb HTA125 was obtained from Hycult. Non-blocking CD32 (clone 2E1) and CD64 (clone 1D3) antibodies were obtained from Acris and Abnova, respectively. The antibodies were digested using papain into Fab fragments and subsequently conjugated to either Cy3 or Cy5 using labeling kits from GE Healthcare. IL-6 and IL-8 ELISA kits were purchased from R&D Systems or eBioscience, and the mouse cytokine Luminex kit was purchased from Millipore. The RedImune IVIG was from CSL Behring.
Antibody Engineering-Anti-human TLR4 mAb 15C1 (mouse IgG 1, ) and anti-mouse TLR4 mAb 5E3 (rat IgG 2b ) have been described previously (14,15). A chimeric version of 5E3 on a mouse IgG 2a, backbone was generated using standard molecular biology techniques by fusing the VH and VL regions of 5E3 onto mouse ␥2a and constant regions, respectively. 15C1 was humanized by CDR grafting and framework optimization. Two human VH and two human VL germ lines were selected for CDR grafting as follows: 4-28 and 3-66 (VH); L6 and A26 (VL). The humanized 4-28/A26 version of 15C1 on a human IgG 1, backbone was selected for further development because it retained the highest binding affinity for TLR4. Two amino acids in the CH2 domain were replaced (N325S and L328F) to abrogate binding to both Fc␥RIII and complement factor C1q. The two mutations also increased binding to Fc␥RII and retained high affinity binding to Fc␥RI. The final humanized 15C1 antibody was designated Hu 15C1.
Surface Plasmon Resonance-Affinity and kinetic parameters were determined on a Biacore 2000 system (GE Healthcare). Recombinant human TLR4-MD-2 and Fc␥Rs were acquired from commercial sources (R&D Systems or MyBioSource). The neonatal Fc receptor (FcRn) was produced in house as described previously (16). Fc␥Rs were diluted in 10 mM acetate buffer at the appropriate pH according to the pH scouting and then coupled to a CM5 or CM4 sensor chip by amine coupling, following the manufacturer's instructions (GE Healthcare). An 800 -1600-reference unit immobilization signal was reached, depending on the ligand. The FcRn was biotinylated in vivo and captured on a streptavidin-coated CM5 chip for orientated immobilization of the ligand (16). Five concentrations of Hu 15C1 were injected, in duplicate and randomly, on immobilized receptors. The concentration range was adapted to the expected K D for the different ligands. Data were collected at 25°C in HBS EP buffer (GE Healthcare) at a flow rate of 30 l/min. A regeneration scouting was performed when necessary, and a 10 mM glycine, pH 3.0, solution was used for Fc␥RI and Fc␥RIIA 131 R receptors. Because IgG binding to FcRn occurs in a pH-dependent manner, the HBS EP running buffer was adjusted to pH 6.0 for Hu 15C1 dilution and baseline, association, and dissociation steps and to pH 8.0 for regeneration. According to the kinetic profiles, the steady state model was used for low affinity interactions (Fc␥RIIA and Fc␥RIIB), whereas the 1:1 Langmuir binding model was applied to higher affinities (Fc␥RI and FcRn). Data were evaluated using BIAevaluation version 4.1.1 software. A double referencing was applied for analysis to subtract buffer signal drift on coating surface and nonspecific background signal on a reference channel. All experiments were performed in duplicate.
For TLR4 binding, after dilution in 10 mM acetate, pH 4.5, the recombinant receptor was immobilized on a CM5 sensor chip. IgGs were injected at five concentrations starting from 667 nM, randomly and in duplicate. Hu 15C1 was diluted in HBS EP running buffer and injected at 25°C at a flow rate of 50 l/min. Regeneration was assessed using a 10 mM glycine, pH 3.0 solution. Data were analyzed, after double subtraction, using the 1:1 Langmuir fitting model on the BIAevaluation software.
FACS-based Binding Assay-Antibodies (Mu 15C1 or Hu 15C1) were incubated at different concentrations with CHO cells expressing human TLR4 and MD-2. After a washing step, fluorescent anti-15C1 idiotype was added to the cells. Finally, cells were washed and then acquired on a FACSCalibur (BD Biosciences).
Human Macrophage Assay with LPS and Nickel-Monocytes obtained from human blood were cultured with 100 ng/ml M-CSF (Peprotech) for macrophage differentiation. Five days later, the differentiated macrophages were incubated with a dose range of Hu 15C1 together with ultrapure LPS from S. minnesota (1, 10, or 100 ng/ml; List Biological Laboratories) or nickel chloride (0.5 mM; Sigma) for 24 h. Supernatants were harvested for IL-6 ELISA (eBioscience).
U937 Cell Assays-U937 cells were differentiated with medium containing 25 nM phorbol myristate acetate for 48 h. After differentiation, the cells were pretreated with Hu 15C1 in the presence or absence of 50 g/ml IVIG and/or 50 g/ml anti-Fc␥RIIA antibody IV.3 for 30 min. The cells were subse-quently incubated with 1 ng/ml ultrapure LPS from S. minnesota for 20 h, and IL-6 in the supernatants was quantified using the Quantikine ELISA kit (R&D Systems).
Confocal Microscopy and FRET Analysis-THP-1 cells either untreated or treated with ultrapure LPS from E. coli (100 ng/ml) in the presence or absence of different concentrations (0.1, 1, or 10 g/ml) of Hu 15C1 (or 15C1 D265A), for 30 min or 1 h, were rinsed twice in PBS, 0.02% BSA. The cells were fixed and stained with specific detection Fab fragments for TLR4 and Fc␥Rs directly conjugated to fluorophores, prior to imaging on a Carl Zeiss, Inc. LSM510 META confocal microscope (with an Axiovert 200 fluorescent microscope) using a 1.4 numerical aperture ϫ63 Zeiss objective. The images were analyzed using LSM 2.5 image analysis software (Carl Zeiss, Inc.). For FRET analysis, THP-1 cells were labeled with 100 l of a mixture of donor Cy3-conjugated Fab and acceptor-conjugated Cy5 Fab against the molecules of interest. FRET was measured using a method described previously (8).
THP-1 Blue Prolonged Blockade Assay-THP-1 Blue cells (Invivogen) were pretreated with 10 g/ml Hu 15C1 or 15C1 D265A for 30 min. Excess antibody was removed by washing cells twice in PBS. The cells were resuspended in fresh medium and incubated for 0, 1, or 24 h prior to stimulation with 5 ng/ml ultrapure LPS from S. minnesota for 24 h. NF-B reporter gene activation, represented by secreted embryonic alkaline phosphatase activity in the cell culture supernatant, was determined by colorimetric analysis using photospectrometry at A 650 nm .
LPS Lung Instillation-C57BL/6 mice (female, 8 weeks old) from Charles River Laboratories were injected with the indicated antibodies intravenously. One hour later, mice were anesthetized and instilled intranasally (i.n.) with 10 g of ultrapure LPS from S. minnesota. 24 h post-LPS challenge, mice were sacrificed, and bronchoalveolar lavage (BAL) was performed with a volume of 500 l of PBS/BSA (0.2%). Luminex and ELISA were performed to measure cytokine and antibody concentrations in BAL fluid. Antibody concentration in BAL fluid was measured by an in house sandwich ELISA using anti-5E3 idiotype antibody (DA4).

RESULTS
Fc Engagement Promotes the Potency of Hu 15C1-We have previously described a mouse anti-human TLR4 mAb, 15C1, with which the observation was made that the engagement of the mouse Fc with human Fc␥RIIA added neutralization benefits (14). To further explore Fc␥R assistance, the antibody was humanized. Following conventional CDR grafting of 15C1 into a human IgG1 backbone, the Fc portion was altered by inserting N325S and L328F substitutions within the CH2 domain to eliminate Fc␥RIII binding. The modification, although abolishing this binding, maintained high affinity for Fc␥RI and resulted in intermediate affinities for Fc␥RII (Table 1). FcRn binding was unaffected by the Fc engineering ( Table 1). The engineered antibody was designated Hu 15C1. Humanization did not modify the binding to human TLR4 because Hu 15C1 and the parental murine antibody (Mu 15C1) demonstrated overlapping binding profiles for human TLR4/MD-2 expressed on CHO cells (Fig. 1A). This maintenance of TLR4 engagement capacity aligned with the results of surface plasmon resonance analysis, in which the dissociation constant (K D ) of the parental Mu 15C1 was 3.2 nM, whereas that of Hu 15C1 was 2.1 nM for human TLR4.
When comparing the ability of Hu 15C1 versus the parental Mu 15C1 to inhibit TLR4 activation on human myeloid cells, Hu 15C1 was more potent (Fig. 1B). The IC 50 of Hu 15C1 was 6 ng/ml, whereas that of Mu 15C1 was 29 ng/ml. To then investigate whether Fc␥R engagement played a role in enhancing the activity of Hu 15C1, experiments were designed in which the receptors were blocked individually or in combination. For Fc␥RI, a polyclonal preparation of human IgG (IVIG), and for Fc␥RIIA, the antibody, IV.3, were used. Adding both reagents allowed the establishment of baseline TLR4 blockade without Fc␥R interaction, which resulted in an IC 50 of 2200 ng/ml. The occurrence of either Fc␥R interaction (when IVIG or IV.3 was added) increased the potency of Hu 15C1 ϳ100-fold, with a slightly larger contribution from Fc␥RI than Fc␥RIIA (IC 50 of 30 ng/ml in the presence of IVIG alone versus 18 ng/ml in the presence of IV.3 alone) (Fig. 1C). The availability of both Fc␥Rs further increased the potency of Hu 15C1 and restored the IC 50 to 6 ng/ml (Fig. 1C), demonstrating the contribution of both Fc␥Rs in enhancing TLR4 inhibition.
Hu 15C1 Inhibits TLR4 Activation by Interfering with Receptor Dimerization-Dissecting the mechanism of action, it has been shown that the epitope of 15C1 is made up of at least three sites on human TLR4 (14) between amino acid residues 289 and 375 ( Fig. 2A). This stretch is part of the interface between the two TLR4 molecules that interact upon ligand-induced dimerization (17), an event required for TLR4 activation. Thus, we hypothesized that Hu 15C1 interferes with receptor dimerization and therefore would present a non-competitive profile. To evaluate this, the TLR4 agonists, nickel, which directly triggers receptor dimerization (18), and LPS were used. As shown in Fig. 2B, Hu 15C1 inhibited nickel-induced IL-6 production in a dose-dependent manner. When assessing a range of LPS concentrations (i.e. 1-100 ng/ml), overlapping inhibitory profiles were observed (Fig. 2C). The IC 50 values for Hu 15C1 were similar when using either nickel or LPS to activate TLR4 (Fig. 2). Taken together, these data demonstrate that the mechanism of action by which Hu 15C1 interferes with TLR4 signaling is ligand type-and concentration-independent, as would be expected for a molecule blocking receptor dimerization.  (8,19). However, it is not clear whether receptor dimerization is required for migration to lipid rafts or whether dimerization occurs postmigration. Using confocal microscopy in conjunction with FRET, we investigated ligand-dependent TLR4 trafficking to lipid rafts and the subsequent association of TLR4 and Fc␥Rs in the presence of varying concentrations of Hu 15C1 (Fig. 3). At 10 g/ml, Hu 15C1 blocked LPS-induced TLR4 clustering (Fig. 3, A-D). Hu 15C1 at this concentration also blocked the migration of TLR4 to lipid rafts (Fig. 3F), and full inhibition of LPS-stimulated IL-6 secretion was observed (Fig. 3E). The clustering data together with the FRET data suggest that TLR4 clustered in lipid rafts. At 0.1 g/ml, although only partially inhibiting TLR4 clustering (Fig. 3C) and migration to lipid rafts (Fig. 3F), Hu 15C1 completely abrogated LPSstimulated IL-6 secretion (Fig. 3E). At 0.01 g/ml, Hu 15C1 only reduced the migration of TLR4 to lipid rafts by ϳ30% (Fig.  3F), but this concentration was still sufficient to fully block IL-6 secretion (Fig. 3E). Further investigating the potential of trimolecular interactions of the antibody with TLR4 and Fc␥Rs within the lipid rafts, the Fc␥R-non-binding variant of Hu 15C1, designated 15C1 D265A, which has a point mutation in the Fc portion abrogating interaction with Fc␥Rs (20), was employed. When assessing this variant, at 0.01 and 0.1 g/ml, again only partial inhibition of TLR4 association with lipid rafts was measured, similar to that observed with Hu 15C1 (Fig. 3F). In contrast, 15C1 D265A was not able to block TLR4 signaling (Fig. 3E). When a higher dose (i.e. 10 g/ml of 15C1 D265A) was used, both TLR4 association with lipid rafts (Fig. 3F) and TLR4 signaling (Fig. 3E) were completely impaired. These results demonstrate that, depending on the dose, the co-engagement of Hu 15C1 with TLR4 and Fc␥Rs renders the antibody capable of inhibiting TLR4 activation even following ligand-dependent TLR4 migration to lipid rafts. Because both variants of Hu 15C1, which interfere with TLR4 dimerization, abrogate TLR4 association with lipid rafts at higher concentrations, these results suggest that receptor dimerization occurs prior to TLR4 transfer to lipid rafts.
Blockade of TLR4 Dimerization Abrogates Ligand-dependent Association between Fc␥Rs and TLR4 in Lipid Rafts-To further dissect the interplay between Fc␥Rs, TLR4, and lipid rafts on the cell surface, we investigated their association in the presence or absence of LPS. FRET analysis revealed Fc␥RI to be constitutively associated with lipid rafts, independently of the presence of LPS, Hu 15C1, or 15C1 D265A (Fig. 4A). Fc␥RI was not associated with TLR4 under resting conditions, and neither Hu 15C1 nor 15C1 D265A induced association between TLR4  and Fc␥RI. However, association between Fc␥RI and TLR4 occurred upon LPS stimulation, which was blocked by the addition of Hu 15C1 or 15C1 D265A in a dose-dependent manner (Fig. 4, A and C). Fc␥RIIA was not constitutively associated with lipid rafts (Fig. 4B). LPS stimulation induced the association between Fc␥RIIA, TLR4, and lipid rafts, whereas either antibody alone did not. Interestingly, neither antibody affected LPS-dependent association of Fc␥RIIA and lipid rafts, yet both antibodies blocked Fc␥RIIA and TLR4 interaction in a dose-dependent manner with no significant difference between the two antibodies (Fig. 4, B and D). Thus, TLR4 and Fc␥Rs cluster in lipid rafts upon LPS exposure, producing a microenvironment whereby anti-TLR4 antibodies, such as Hu 15C1, co-engage TLR4 and Fc␥Rs.
Next we determined whether engagement by Hu 15C1 stimulates signaling through Fc␥Rs. To this end, THP-1 cells were incubated with Hu 15C1 for 30 or 60 min, and the inhibitory immunoreceptor tyrosine-based activation motif (ITAMi) pathway (21) was investigated. The Fc-engineered antibody Hu 15C1 failed to induce phosphorylation of SHP-1 (Fig. 5). In contrast, the positive control, anti-CD89 (22), stimulated SHP1 phosphorylation at both 30 and 60 min. Thus, activation of the ITAMi pathway (21) is not implicated in the mechanism of action of Hu 15C1. Fig. 1C and Fig. 3E, engagement of Fc␥Rs increased the potency of Hu 15C1 on Fc␥R-bearing cells. This mechanism of action suggests that Hu 15C1 would have higher potency on Fc␥R-bearing cells than on Fc␥R-negative cells. Consistent with this prediction, Hu 15C1 had a much higher potency on HEK293 cells transfected with TLR4⅐MD-2⅐Fc␥RIIA than on those transfected only with TLR4⅐MD-2 (Fig. 6A). Furthermore, Hu 15C1 and its Fc-null variant, 15C1 D265A, had a similar IC 50 on Fc␥R-negative human umbilical vein endothelial cells (Fig. 6B), confirming the lack of Fc contribution to Hu 15C1 potency on these Fc␥R-negative cells.

Co-engagement of TLR4 and Fc␥Rs Enhances the Potency and Prolongs the Inhibitory Duration of Hu 15C1 on Fc␥R-bearing Cells-As demonstrated in
The beneficial effect afforded by the novel mechanism of action of Hu 15C1 was further explored. The Fc␥R-bearing THP-1 Blue reporter cell line was incubated with a dose range of either Hu 15C1 or 15C1 D265A for 30 min. After washing to remove unbound antibody, cells were cultured in fresh medium for various lengths of time before being subjected to LPS stimulation (Fig. 6C). When LPS was added immediately after anti- , or a human IgG 1 isotype control (10 g/ml). Cells were incubated in the presence or absence of LPS (100 ng/ml) and mAbs for 1 h before being subjected to FRET analysis. The assessment of the proximity of TLR4 to GM1 was used as a marker for clustering in lipid rafts, whereas that of TLR4 to MHC class I served as a baseline control to provide values for molecular non-association. Data shown represent the mean Ϯ S.D. (error bars) of four independent experiments. The data were analyzed using Student's t test. The statistical difference between the LPS group and the LPS ϩ mAb groups is shown on individual bars. No difference was observed between the Hu 15C1 and 15C1 D265A groups at the concentrations tested. ***, p Ͻ 0.001; ns, not significant. body incubation, a 10 g/ml concentration of the Fc-null mutant, 15C1 D265A, produced maximal blockade, whereas the effect was reduced at lower concentrations (Fig. 6D). When LPS was added 1 h after removing unbound antibody, the inhibitory effect seen at 10 g/ml was reduced to less than 40% and reduced to less than 20% when LPS was added at 24 h (Fig. 6D). In contrast, a substantially lower concentration of Hu 15C1 (i.e. 0.1 g/ml) produced maximal blockade when LPS was added immediately after antibody incubation. The inhibitory effect was maintained at more than 80% when LPS was added 1 h later and more than 70% when LPS was added 24 h after removing unbound antibody. Even at the lowest concentration of Hu 15C1 tested (0.01 g/ml), inhibition was ϳ40% when LPS was FIGURE 4. Blockade of TLR4 dimerization abrogates its clustering with Fc␥Rs but has no effect on Fc␥R clustering. A and B, FRET analysis of the association of Fc␥RI (A) and Fc␥RIIA (B) with lipid rafts (GM1) as well as with TLR4 in response to LPS (100 ng/ml), Hu 15C1 (0.1, 1, and 10 g/ml), and 15C1 D265A (0.1, 1, and 10 g/ml). Cells were incubated in the presence or absence of LPS and mAbs for 1 h before being subjected to FRET analysis. The assessment of the proximity of Fc␥RI and Fc␥RIIA to GM1 was used as a marker for clustering in lipid rafts, whereas their proximity to MHC class I served as a baseline control to provide values for molecular non-association. Data shown represent the mean Ϯ S.D. (error bars) of four independent experiments. The data were analyzed with Student's t test. The statistical difference between the LPS group and the LPS ϩ mAb groups is shown on individual bars. No difference was observed between the Hu 15C1 and 15C1 D265A groups at the concentrations tested. ***, p Ͻ 0.001; ns, not significant. C and D, photomicrographs depict the association of Fc␥RI (C) and Fc␥RIIA (D) with TLR4 in response to LPS and Hu 15C1. Cells were incubated in the presence or absence of LPS (100 ng/ml) and Hu 15C1 (10 g/ml) for 1 h before being subjected to fluorescence labeling for imaging by confocal microscopy. FIGURE 5. Lack of Fc␥R-mediated ITAMi signaling in response to Hu 15C1. THP1 cells were incubated with 10 g/ml Hu 15C1, 15C1 D265A, positive control (anti-CD89 F(ab)Ј2, clone A77), or negative control (isotype control) mAbs for 0, 30, or 60 min. Cell extracts were prepared and subjected to Western blotting (IB). SHP-1 phosphorylation (pSHP-1) was used as a marker for ITAMi activation. The data shown represent three independent experiments. added 24 h after antibody removal. These results demonstrate that the Fc-dependent mechanism of action of Hu 15C1 on Fc␥R-bearing cells not only enhances its inhibitory potency but also substantially increases the duration of its inhibitory effect.
Fc␥R Engagement Also Enhances in Vivo Efficacy-To explore whether the Fc␥R-dependent mechanism of action afforded additional benefit in TLR4-driven experimental models of inflammation, the previously described rat anti-mouse TLR4 monoclonal antibody (15), 5E3, was reengineered. To ensure Fc␥R engagement on mouse cells, the antibody's Fv portion was reconstructed onto a mouse IgG 2a, backbone (5E3 mIgG 2a ), providing high affinity for mouse Fc␥RI (13). A D265A variant of 5E3 mIgG 2a , which does not bind to any Fc␥R, was also generated. In vitro, 5E3 IgG 2a was found to be ϳ100fold more potent than 5E3 D265A in blocking LPS-stimulated TLR4 activation, demonstrating the potential to exploit this mechanism of action involving Fc␥R co-engagement in a murine system (Fig. 7A).
Next, to investigate beneficial effects of Fc␥R engagement in vivo, these mAbs were used in a mouse model of LPS-induced acute lung inflammation. Intranasal administration of LPS induced inflammatory cytokines (i.e. IL-6 and TNF␣) in the bronchoalveolar lavage (BAL) fluid (Fig. 7, B and C). Both 5E3 mIgG 2a and its D265A variant blocked IL-6 and TNF␣ in a dose-dependent manner. However, 5E3 mIgG 2a demonstrated a greater potency in blocking the cytokines as compared with the Fc␥R-null mutant, 5E3 D265A (Fig. 7, B and C). Significant differences were observed between the two antibodies at all three doses, with 5E3 IgG 2a being ϳ10-fold more potent than the D265A mutant. To confirm that the Fc mutation had not affected the ability of the 5E3 variant to reach the alveolar space, antibody concentrations in the BAL fluid were measured by ELISA. At all doses tested, both mAbs (5E3 mIgG 2a and the D265A mutant) reached the lung compartment in a dose-dependent manner, and no difference in BAL fluid concentrations was detected (Fig. 7D). These results indicate that the difference in potency between the two 5E3 variants is due to their differential ability to engage Fc␥Rs rather than differences in their in vivo distribution.

DISCUSSION
In this study, we have demonstrated that the anti-human TLR4 antibody, Hu 15C1, exploits both Fc␥RI and Fc␥RIIA to increase its potency on Fc␥R-bearing inflammatory cells. This

. Fc␥R engagement renders Hu 15C1 potent with long lasting inhibitory effects on TLR4-and Fc␥R-co-expressing cells but not on Fc␥R-negative cells.
A, blockade of TLR4 activation by Hu 15C1 in HEK293 cells transfected with TLR4⅐MD-2 or TLR4⅐MD-2⅐Fc␥RIIA. Transfected HEK293 cells were incubated in the presence of 100 ng/ml LPS and Hu 15C1 for 24 h. IL-8 levels in supernatants were measured by ELISA. An F test was used to compare the fitted curves of the inhibitory profile of Hu 15C1 with the two different transfected cell lines. ***, p Ͻ 0.001. Data shown represent three independent experiments. B, blockade of TLR4 activation by Hu 15C1 and 15C1 D265A in human umbilical vein endothelial cells. Cells were incubated in the presence of 100 ng/ml LPS and Hu 15C1 or 15C1 D265A for 24 h. IL-8 levels in supernatants were measured by ELISA. An F test was used to compare the fitted curves of different groups. ns, not significant. C and D, Hu 15C1 has a longer lasting inhibitory effect than 15C1 D265A on TLR4 activation in THP-1 Blue cells. The experimental procedure is depicted in C, and the results are shown in D. THP-1 Blue cells were loaded with Hu 15C1 or 15C1 D265A for 30 min, and free mAb in the medium was removed by washing. The cells were then cultured at 37°C for various periods as indicated in the panels (0, 1, or 24 h) and subjected to LPS stimulation (5 ng/ml) for 24 h. Student's t test was used to compare differences between groups. Results are expressed as the mean Ϯ S.E. (error bars). ***, p Ͻ 0.001; **, p Ͻ 0.01; *, p Ͻ 0.05; ns, not significant.
contribution from Fc␥Rs required ligand-dependent clustering of TLR4 with Fc␥Rs in lipid rafts (Fig. 8). Lower concentrations of Hu 15C1 (i.e. below 1 g/ml) were not sufficient to completely prevent LPS-induced clustering of TLR4 with Fc␥Rs. We consider, therefore, that binding to Fc␥Rs in the lipid raft under these conditions increases the local concentration of Hu 15C1 and renders it capable of interacting with incoming TLR4 dimers. Our data suggest that binding of Hu 15C1 to TLR4 in the lipid raft either disrupts the TLR4 dimers or blocks further aggregation among TLR4 dimers and thus blocks TLR4 signaling. The co-engagement of Hu 15C1 to TLR4 and Fc␥Rs in the lipid raft subsequently increases the avidity of Hu 15C1 binding to the cell surface and thus further increases its ability to block TLR4 activation. At higher Hu 15C1 concentrations (i.e. Ͼ1 g/ml), the clustering of TLR4 with Fc␥Rs was totally blocked. Because TLR4 is not in the proximity of Fc␥Rs under these conditions, Hu 15C1 is not able to co-engage TLR4 and Fc␥Rs. Thus, Fc␥R engagement is not expected to contribute to the inhibitory effect of Hu 15C1 at higher antibody concentrations. The ability of Hu 15C1 to engage Fc␥Rs also enables the antibody to block TLR4 activation for a longer duration. The relatively high Fc␥RI and Fc␥RIIA densities on the surface of inflammatory cells compared with TLR4 may allow preferential Hu 15C1 binding at lower antibody concentrations. We therefore speculate that the interaction with Fc␥RI and possibly Fc␥RIIA provides another means for Hu 15C1 to remain on cell surfaces when free antibody in the solution is removed, directly translating into a longer period of inhibition.
Beyond increasing the avidity of Hu 15C1 to the cell surface, it is plausible that Fc␥R engagement could also induce antiinflammatory signaling cascades capable of down-regulating TLR4 activation, hence increasing the potency of Hu 15C1 on Fc␥R-bearing cells. Indeed, monovalent ligation of Fc receptors has been demonstrated to induce ITAMi activation, which subsequently induces the phosphorylation of SHP1 and down-regulates proinflammatory responses (21). However, no increase in SHP1 phosphorylation was detected when THP-1 cells were treated with Hu 15C1, either in the presence or absence of LPS. These results are consistent with our hypothesis that increased local receptor concentration and subsequent antibody binding avidity, rather than Fc-induced signaling events, are solely responsible for the increased potency of Hu 15C1 upon Fc␥R engagement.
The Fc␥R-dependent mechanism of action was also observed with an anti-mouse TLR4 antibody, 5E3. Because the epitope of 5E3 is also located within a region involved in TLR4 dimerization (15), this finding suggests that the mechanism could be widely applicable to TLR4 blocking antibodies directly interfering with receptor dimerization or even more generally to all anti-TLR4 blocking antibodies. Two other anti-human TLR4 antibodies, H4 and H52, were reported to block TLR4 activation via an Fc␥R-independent mechanism (23). Nevertheless, high antibody concentrations (10 g/ml) were used in the above report, which, from our own experience, may mask any potential contribution from Fc␥R engagement. Indeed, antireceptor antibodies with mechanisms of action involving beneficial Fc␥R engagement have previously been reported. For example, Fc␥RIIB engagement has been reported to enhance the function of an anti-CD40 agonistic antibody targeting cells co-expressing Fc␥RIIB and CD40 (24). The underlying mechanism was not clarified in the report. We speculate that the anti-CD40 antibody may use a similar mechanism as described here for TLR4 to enhance its effects. Additionally, an Fc contribution was recently demonstrated in the mechanism of action of an anti-CD115 mAb (25).
The clustering of TLR4 upon stimulation has been reported previously (8,19) and confirmed in this study. In addition to its association with apparent activation partners upon agonist stimulation, TLR4 also clusters in lipid rafts with other proteins not directly associated with the stimuli, such as Fc␥Rs (19). Because Fc␥RI is constitutively present in lipid rafts, it is anticipated that TLR4 will cluster with Fc␥RI upon migration of TLR4 to lipid rafts. In contrast, no association between Fc␥RIIA and lipid rafts or between Fc␥RIIA and TLR4 was detected under resting conditions, whereas the association was triggered following LPS stimulation. Interestingly, blockade of TLR4 dimerization did not affect LPS-induced association between Fc␥RIIA and lipid raft structures but totally abrogated the asso- ciation between Fc␥RIIA and TLR4. These results suggest that the association between Fc␥RIIA and lipid rafts requires LPS but not TLR4. Further studies are necessary in order to elucidate the precise mechanism by which LPS induces Fc␥RIIA migration to lipid rafts independently of TLR4.
The fact that TLR4 clusters with Fc␥Rs upon activation suggests that TLR4 activation may provide a platform to facilitate Fc␥R activation. Indeed, the ability of IgG immune complexes to induce cytokine release in neutrophils and macrophages was lost in TLR4 mutant mice (26), confirming a role for TLR4 in Fc␥R activation. Furthermore, signaling through Fc␥RIII was shown to modulate TLR4 signaling pathways in human monocytes (27). Due to the lack of Fc␥RIII binding by Hu 15C1, potential cross-talk was not tested in the current study. On the other hand, such clustering may also enable Fc␥Rs to modulate innate immune responses, which may play a role in preventing the detrimental effects of uncontrolled inflammation. In support of this, TLR responses have been reported to be enhanced in dendritic cells from Fc receptor ␥ chain-deficient mice (28), and high avidity ligation of Fc␥Rs has been shown to dampen TLR signaling under certain circumstances (29). Both reports demonstrated a negative role for Fc␥Rs in TLR signaling. Although the clustering between TLR4 and Fc␥Rs has been established, it is not clear whether the activation of other TLRs induces similar clustering, and as such, this represents a topic for future investigational work.
It has been established that TLR4 dimerization and clustering in lipid rafts are critical events in TLR4 signaling. However, the relationship between dimerization and clustering has not been well elucidated. LPS-triggered TLR4 dimerization was reported to be only detected in lipid rafts, and such dimerization was abrogated by a lipid raft disruptor (30). On the other hand, artificially forced TLR4 dimerization is by itself sufficient to trigger TLR4 signaling (31,32). We demonstrate here that blockade of TLR4 dimerization with Hu 15C1 totally abrogates TLR4 clustering in lipid rafts, suggesting that TLR4 dimerization is a prerequisite for TLR4 clustering.
The mechanism of action of Hu 15C1 demonstrates that TLR4 activation can be modulated even after receptor dimerization and during receptor clustering in lipid rafts. This suggests that Hu 15C1 may be able to disrupt established TLR4 dimers. Alternatively, further aggregation of TLR4 in lipid rafts is required after ligand-induced dimerization for maximal TLR4 activation, and Hu 15C1 is able to interfere with this amplification event. Supporting these concepts, we found that Hu 15C1 could efficiently block TLR4 activation when added several h after ligand exposure.
Dysregulated activation of TLR4 has been implicated in numerous acute inflammatory, chronic inflammatory, and autoimmune disorders, such as infection, asthma, acute lung injury, ischemia/reperfusion injury, inflammatory bowel disease, graft rejection, transplantation, rheumatoid arthritis, and multiple sclerosis (33). Because uncontrolled leukocyte activation plays a pivotal role in the maintenance of chronic inflammation (34), blocking the activation of these cells is considered beneficial to such diseases. On the other hand, TLR4 signaling on epithelial cells has been demonstrated to play a positive role in tissue homeostasis (11). A prolonged total shutdown of TLR4 signaling on epithelial cells may therefore be detrimental to tissue repair. Because inflammatory cells express Fc␥Rs, whereas epithelial cells do not, the ability to engage Fc␥Rs bestows a higher and long lasting potency upon Hu 15C1 toward inflammatory cells but a lower and shorter lived potency toward epithelial cells. This feature provides the prospect for Hu 15C1 to preferentially block the dysregulated activation of TLR4 on inflammatory cells while sparing some beneficial TLR4 signaling on structural cells. Furthermore, this mechanism highlights events that may occur with other antibodies that target cell surface receptors that traffic to lipid rafts.