Lipopolysaccharide (LPS)-binding protein stimulates CD14-dependent Toll-like receptor 4 internalization and LPS-induced TBK1–IKKϵ–IRF3 axis activation

Toll-like receptor 4 (TLR4) is an indispensable immune receptor for lipopolysaccharide (LPS), a major component of the Gram-negative bacterial cell wall. Following LPS stimulation, TLR4 transmits the signal from the cell surface and becomes internalized in an endosome. However, the spatial regulation of TLR4 signaling is not fully understood. Here, we investigated the mechanisms of LPS-induced TLR4 internalization and clarified the roles of the extracellular LPS-binding molecules, LPS-binding protein (LBP), and glycerophosphatidylinositol-anchored protein (CD14). LPS stimulation of CD14-expressing cells induced TLR4 internalization in the presence of serum, and an inhibitory anti-LBP mAb blocked its internalization. Addition of LBP to serum-free cultures restored LPS-induced TLR4 internalization to comparable levels of serum. The secretory form of the CD14 (sCD14) induced internalization but required a much higher concentration than LBP. An inhibitory anti-sCD14 mAb was ineffective for serum-mediated internalization. LBP lacking the domain for LPS transfer to CD14 and a CD14 mutant with reduced LPS binding both attenuated TLR4 internalization. Accordingly, LBP is an essential serum molecule for TLR4 internalization, and its LPS transfer to membrane-anchored CD14 (mCD14) is a prerequisite. LBP induced the LPS-stimulated phosphorylation of TBK1, IKKϵ, and IRF3, leading to IFN-β expression. However, LPS-stimulated late activation of NF-κB or necroptosis were not affected. Collectively, our results indicate that LBP controls LPS-induced TLR4 internalization, which induces TLR adaptor molecule 1 (TRIF)-dependent activation of the TBK1–IKKϵ–IRF3–IFN-β pathway. In summary, we showed that LBP-mediated LPS transfer to mCD14 is required for serum-dependent TLR4 internalization and activation of the TRIF pathway.


Toll-like receptor 4 (TLR4) is an indispensable immune receptor for lipopolysaccharide (LPS), a major component of the Gram-negative bacterial cell wall. Following LPS stimulation, TLR4 transmits the signal from the cell surface and becomes internalized in an endosome. However, the spatial regulation of TLR4 signaling is not fully understood. Here, we investigated the mechanisms of LPS-induced TLR4 internalization and clarified the roles of the extracellular LPS-binding molecules, LPSbinding protein (LBP), and glycerophosphatidylinositol-anchored protein (CD14). LPS stimulation of CD14-expressing cells induced TLR4 internalization in the presence of serum, and an inhibitory anti-LBP mAb blocked its internalization. Addition of LBP to serum-free cultures restored LPS-induced TLR4
internalization to comparable levels of serum. The secretory form of the CD14 (sCD14) induced internalization but required a much higher concentration than LBP. An inhibitory anti-sCD14 mAb was ineffective for serum-mediated internalization. LBP lacking the domain for LPS transfer to CD14 and a CD14 mutant with reduced LPS binding both attenuated TLR4 internalization. Accordingly, LBP is an essential serum molecule for TLR4 internalization, and its LPS transfer to membrane-anchored CD14 (mCD14) is a prerequisite. LBP induced the LPSstimulated phosphorylation of TBK1, IKK⑀, and IRF3, leading to IFN-␤ expression. However, LPS-stimulated late activation of NF-B or necroptosis were not affected. Collectively, our results indicate that LBP controls LPS-induced TLR4 internalization, which induces TLR adaptor molecule 1 (TRIF)-dependent activation of the TBK1-IKK⑀-IRF3-IFN-␤ pathway. In summary, we showed that LBP-mediated LPS transfer to mCD14 is required for serum-dependent TLR4 internalization and activation of the TRIF pathway.
The innate immune system defends against invading pathogens by recognizing conserved pathogen-associated molecular patterns (PAMPs) 4 (1,2). The induction of cytokines and chemokines by the innate immune system causes inflammation and initiates adaptive immunity (3). The recognition of PAMPs relies on pathogen recognition receptors (PRRs) on the cell surface or in intracellular compartments (1,2). Among the PRRs, Toll-like receptors (TLRs) have been intensively studied and broadly recognize pathogenic organisms (4,5). The subcellular locations where TLRs initiate signals determine the outcomes of such events (2,6,7). Therefore, TLR responses must be spatially and temporally regulated.
TLR4 is an indispensable receptor of LPS, which is a unique component of the Gram-negative bacterial cell wall (8). LPS forms a complex with MD-2, a secreted glycoprotein with an LPS-binding pocket, via the extracellular domain of TLR4 (9,10). Following LPS binding via MD-2, TLR4 induces myeloid differentiation 88 (MyD88)-dependent signaling from the cell surface, which activates the NF-B/MAPK pathway (11). TLR4 is then internalized and initiates Toll/IL-1 receptor (TIR)domain-containing adapter-inducing IFN-␤ (TRIF)-dependent signaling and subsequent NF-B/IRF3 activation (12,13). However, the spatial regulation of TLR4 signaling is poorly understood.
Multiple LPS-binding proteins have been found in serum and cell membranes (5,14,15). Among those, CD14, a secreted and glycerophosphatidylinositol (GPI)-anchored protein, and the LPS-binding protein (LBP) are required for LPS recognition by TLR4 and MD-2 (5, 16 -18). The catalytic mechanism underlying LPS transfer by LBP to CD14 was recently revealed (17). LBP binds LPS micelles via the N-terminal basic patch and forms transient ternary complexes with secreted (sCD14) or membrane-bound GPI CD14 (mCD14) on the C terminus (17). Following the generation of CD14/LBP/LPS micelles, CD14 dissociates from LBP and receives monomeric LPS (17). The LPS transfer to mCD14 enhances the LPS sensitivity of TLR4/ MD-2 in innate immune cells, including macrophages and monocytes. Thus, this is the first line of defense against bacterial invasion (19 -21). Conversely, the transfer of LPS to sCD14 confers a response in CD14-deficient cells (22,23). In addition to its classical role of enhancing LPS sensitivity, CD14 expressed on the cell surface plays an essential role in the internalization of TLR4 and TRIF-mediated signaling from the endosome (13,24). In CD14-positive cells (e.g. macrophages), TLR4 is internalized in endosomes where it dissociates from MyD88 and interacts with TRIF to activate IRF3 (7,12).
LBP belongs to an LBP/bactericidal/permeability-increasing protein/palate, lung, and nasal epithelial clone protein superfamily and is produced primarily in liver and epithelial cells of the lungs and gastrointestinal tract as an acute phase serum glycoprotein (25). The family members constitute physical barriers against bacterial infection in innate immunity (25,26). For example, the bactericidal/permeability-increasing proteins, which show the highest structural homology to LBP, neutralize LPS and also elicit bactericidal activity against Gram-negative bacteria (27). Among the LBP/bactericidal/permeabilityincreasing protein family, LBP is unique in that it binds LPS micelle and transfers a monomer to CD14, which enhances the TLR4/MD-2 response (17,25). Serum LBP increases ϳ10-fold during infection (28), peaking in the acute phase, which dampens the inflammatory response to LPS in some cases (29,30). LBP also interacts with other lipopeptides and enhances TLR2 responses (31,32). Thus, it could be designated a soluble PRR.
We previously showed that LBP is an essential mediator of TLR4/MD-2 dimerization in mCD14-expressing cells and responded to LPS (16). We also found that substantial LPS was bound to surface TLR4/MD-2 in mCD14-positive cells stimulated with LBP (16). Because TLR4 internalization relies on mCD14 expression (13,24), LBP may transfer LPS to TLR4 and transduce signals from the endosome via TLR4 internalization. Notably, LPS-induced type I IFN production was increased in monocytes stimulated with LBP (33), suggesting that LBP initiated TRIF-dependent IFN production. However, the mechanistic details for such events have not been resolved. In this study, we investigated the serum components that mediated TLR4 internalization and demonstrated an essential role for LPS transfer by LBP to mCD14, which was required for TLR4 internalization and activation of the TRIF-mediated pathway.

LPS-induced TLR4/MD-2 internalization is mediated by serum components
To identify the molecular mechanisms underlying LPS-induced TLR4 internalization, we examined the requirement for serum components using HEK293 cells expressing human TLR4/MD-2 and CD14. The surface expression of TLR4/MD-2 was assessed using flow cytometry in LPS-stimulated cells. In the presence of FCS and human serum, TLR4 was internalized upon LPS stimulation; however, internalization did not occur in the absence of serum, even in cells with mCD14 surface expression (Fig. 1, A and B). Serum dependence was also observed in the mCD14-positive macrophage cell line, RAW264, and in bone marrow-derived macrophages (BMMs) (Fig. 1, C and D). In contrast to the reduced surface expression, LPS stimulation did not change the total amount of TLR4 (Fig.  1, E and F). These findings indicate that LPS-induced TLR4/ MD-2 internalization was regulated by serum components.

LBP is essential for LPS-induced TLR4/MD-2 internalization
Serum contains multiple LPS-binding molecules (5,14,15), including sCD14 and LBP, which are involved in LPS sensing by TLR4/MD-2 (5, 16 -18). We previously showed that LBP mediates mouse TLR4/MD-2 dimerization in mCD14-expressing cells (16). This was confirmed by co-immunoprecipitation (IP) assay using human TLR4/MD-2-expressing cells with or without CD14 (Fig. S1). Additionally, we showed that LPS was bound to surface TLR4/MD-2 in the mCD14-expressing cells stimulated with LBP (16). Thus, we investigated whether LBP directly mediates the dimerization of the TLR4 extracellular domain using co-IP assays. The soluble extracellular domains of human TLR4 (sTLR4F or sTLR4G), which were complexed with HA-tagged MD-2 but lacked the transmembrane (TM) and intracellular signaling domains, were FLAG-or GFPtagged and incubated with LPS in the presence of FLAG-tagged sCD14 and LBP. sTLR4G was immunoprecipitated with an anti-GFP Ab, and sTLR4F was detected by Western blotting using an anti-FLAG mAb. sTLR4F co-precipitated with sTLR4G following LPS incubation with LBP and sCD14 (Fig. 2). Conversely, incubation of LPS with sCD14 or LBP alone did not result in co-precipitation. Dimerization also failed to occur in the absence of LPS. These findings indicate that LBP-mediated LPS-induced TLR4 dimerization via the extracellular domain in a CD14-dependent manner. Thus, we hypothesized that LBP is an essential serum component that mediated TLR4 internalization following receptor dimerization.
To test this hypothesis, HEK293 cells expressing TLR4/MD-2/CD14 were stimulated in serum-free medium with recombinant human LBP prepared from CHO cells. LBP restored LPSinduced TLR4 internalization to levels comparable with those seen in the presence of complete serum (Fig. 3, A and B). Internalization was also observed when BMM and RAW264 cells were stimulated with exogenous LBP (Fig. 3, C and D) without apparent changes in the total amount of TLR4 (Fig. 3, E and F).
To confirm the internalization of surface TLR4, we also performed pulse-chase experiments with anti-TLR4 mAb in HEK293 cells expressing TLR4/MD-2/CD14. Following LPS

LBP mediates LPS-induced TLR4 internalization and signaling
stimulation with LBP, immunofluorescence of mAb-labeled surface TLR4 was detected as punctate foci in cytoplasmic regions (Fig. S2). Some internalized TLR4 was co-localized with the early endosome marker, EEA1, but not with the plasma membrane marker, wheat germ agglutinin (WGA), at 60 min of LPS stimulation (Fig. 3, G and H). On the other hand, most cell-surface TLR4 in unstimulated cells was co-localized with WGA but not EEA1. Cytoplasmic localization was rarely observed when stimulated in the absence of LBP (Fig. S2). Based on these results, TLR4 is internalized from the cell-surface in an LBP-dependent manner.
The addition of sCD14 was also investigated because this molecule binds and responds to LPS in CD14-deficient cells (22,23). Exogenous sCD14 restored TLR4 internalization, albeit in a delayed manner compared with LBP (Fig. 3A). The concentrations of sCD14 and LBP required for TLR4 internalization were also evaluated to identify the most potent molecule. sCD14 required 100 -1000 ng/ml to cause internalization, whereas 1-10 ng/ml of LBP was required (Fig. 3B).
The contribution of LBP from human serum to TLR4 internalization was also evaluated. Notably, TLR4 internalization was inhibited in HEK293 cells incubated with human serum that had been pretreated with LBP-blocking mAb (Fig. 4A). A similar finding was also observed in BMMs, albeit to a lesser extent (Fig. 4B). The addition of human CD14-blocking mAb that inhibits LPS binding to human sCD14 in serum, but not that of mouse mCD14 on the surface of BMMs, was ineffective (Fig. 4B). These findings suggest that LBP is the serum component mediating LPS-induced TLR4 internalization, whereas the contribution of sCD14 is minimal.

LPS transfer from LBP to mCD14 is required for TLR4/MD-2 internalization
Reconstitution assays were performed to determine the interaction between LBP and CD14. HEK293 cells were stimulated with a truncated version of LBP that lacked the C-terminal half (LBP-N). This version, which retains LPS binding, but not its transfer to CD14 (10, 16, 34), failed to internalize TLR4. The dotted and open histograms represent staining of unstimulated cells with or without primary mAb, respectively. The mean fluorescent intensity (MFI) is presented as the changes relative to unstimulated cells. RAW264 cells (C) and BMMs (D) were stimulated with LPS (100 ng/ml) in the presence or absence of 5% FCS. Cells were stained with biotinylated anti-mouse TLR4 mAb and PE-conjugated Stv and analyzed as in A and B. Data are representative of at least three independent experiments. E, RAW264 cells were stimulated as in C. WCL was subjected to IP with protein G-conjugated anti-TLR4 mAb and then analyzed using Western blotting with anti-TLR4 mAb. As a control for sample preparation, WCL was analyzed using Western blotting with anti-GAPDH mAb. The band intensities of TLR4 normalized to that of GAPDH are represented below the blot images as the changes relative to unstimulated cells in the absence of FBS. F, BMMs were stimulated as in D. WCL was prepared and analyzed using Western blotting with anti-TLR4 and anti-GAPDH, respectively. Data are representative of three independent experiments.

LBP mediates LPS-induced TLR4 internalization and signaling
Conversely, full-length LBP caused TLR4 internalization (Fig.  5A). TLR4 internalization was also investigated in CD14-negative HEK293 cells to determine whether LBP alone was sufficient for TLR4 internalization, or if sCD14 was required. The exogenous addition of neither LBP nor sCD14 was sufficient for internalization in CD14-negative cells, whereas their combined addition was marginally effective (Fig. 5B). This finding was consistent with the finding that serum containing bovine LBP and sCD14 was also ineffective (Fig. 5C).
A D10K mutation in the LPS-binding pocket of CD14 that introduced electrostatic repulsion decreased LPS binding ( Fig. 5D and Ref. 22). Cells transfected with the D10K mutant exhibited attenuated TLR4 internalization in the presence of LBP (Fig. 5E). These findings suggest that LBP requires CD14 for LPS binding to mediate TLR4 internalization. Moreover, internalization was conferred by mCD14, but not sCD14. Thus, LPS transfer from LBP to mCD14 is required for TLR4 internalization.
To determine whether the GPI moiety of CD14 was required for LBP-mediated TLR4 internalization, HEK293 cells were generated to express CD14 in which the GPI moiety was replaced with a TM domain. Cells stimulated with LPS in the presence of LBP exhibited the same rates of internalization as those expressing WT CD14, suggesting that GPI anchoring is not required for TLR4 internalization (Fig. 5, F and G).
LBP also binds lipoproteins derived from Gram-positive bacteria and induces TLR2 responses (31,32). TLR2 induces type I IFN and proinflammatory cytokine production (35); therefore, the role of serum and LBP on TLR2 internalization was also evaluated. Flow cytometry revealed that BMMs stimulated with the synthetic TLR2 agonist, Pam2CSK4, resulted in TLR2 internalization independently of serum and LBP (Fig. 6). These findings suggest that LBP selectively mediated LPS-stimulated TLR4 internalization.

LBP mediates LPS-induced TLR4 internalization and signaling
in the absence of LBP, whereas phospho-IRF3 was undetectable. Consistent with these findings, LPS stimulation significantly induced IFN-␤ expression only in the presence of LBP (Fig. 7B). In contrast, MyD88-dependent TNF␣ and IL-6 expression were not significantly affected but were slightly delayed in the absence of LBP (Fig. 7B). The role of LBP in late-phase activation of NF-B was also assessed. In contrast to the early phosphorylation (5-15 min) observed for IB when The dotted and open histograms represent staining of unstimulated cells with or without primary mAb, respectively. RAW264 (C) and BMMs (D) were stimulated with LPS (100 ng/ml) in the presence of LBP (100 ng/ml). TLR4 internalization was analyzed as described in the legend to Fig. 1, C and D. Data are representative of at least three independent experiments. RAW264 (E) cells and BMMs (F) were stimulated as in C and D, and total expression of TLR4 was analyzed as described in the legend to Fig. 1, E and F, respectively. Data are representative of three independent experiments. G and H, HEK293 cells transfected with mouse TLR4/MD-2/ CD14 were incubated with anti-TLR4 mAb and then stimulated with LPS (1 g/ml) for 60 min in the presence of LBP (100 ng/ml). G, following stimulation, cells were fixed, stained with CFா488A-conjugated WGA (green), and then permeabilized with 0.1% Triton X-100. Internalized Ab-bound cell-surface TLR4 was reacted with Alexa 546-conjugated F(abЈ) 2 goat anti-mouse IgG (red). H, cells were fixed, permeabilized, and stained with rabbit anti-EEA1 Ab, followed by Alexa 488-conjugated goat anti-rabbit IgG (green) and Alexa 546-conjugated F(abЈ) 2 goat anti-mouse IgG (red). Nuclei were counterstained with DAPI (blue). The arrow indicates cell-surface TLR4 on WGA-positive plasma membrane. Arrowhead indicates internalized TLR4 on EEA1-positive endosomes. Original magnification, ϫ40 objective. Scale bar, 5 m. Images are representative of two (G) and three (H) independent experiments.

LBP mediates LPS-induced TLR4 internalization and signaling
stimulated with LBP, late-phase phosphorylation (60 -120 min) was independent of LBP ( Fig. 7C, Fig. S3). Additionally, LPS slightly induced NF-B activation in the absence of LBP, as assessed using a reporter assay with TLR4/MD-2/CD14expressing Ba/F3 cells, although the response was attenuated compared with the response in the presence of LBP (Fig. 7D). To gain further insights into LBP-mediated TLR4 signaling, we also investigated the impact of LBP on the phosphorylation of MAPKs (e.g. ERK, p38, and JNK) in LPS-stimulated BMMs (Fig.  7E). ERK and p38 were weakly phosphorylated in the absence of LBP and significantly augmented by the addition of LBP. The phosphorylation of JNK depended on the presence of LBP.
TRIF was involved in the necroptosis of LPS-stimulated macrophages when caspase-8 was inhibited, whereas the same was not true for MyD88 (38). Therefore, the role of LBP in cell death was investigated in BMMs and resident peritoneal macrophages. Cytotoxicity and cell viability were assayed by lactate dehydrogenase (LDH) release and intracellular ATP content in cells stimulated with LPS when LBP was or was not added to culture. Cell death was induced by LPS stimulation with the pan-caspase inhibitor Z-VAD (Fig. 7, F and G). However, LBP was not required for LDH release or ATP content. Consistent with these results, the phosphorylation of RIP1 was induced independent of LBP (Fig. 7C). Based on

LBP mediates LPS-induced TLR4 internalization and signaling
these results, TRIF-mediated necroptosis occurred independently of LBP.

LBP mediates Myddosome and Triffosome assembly
To explore the mechanism underlying LBP-mediated TLR4 signaling, we investigated the assembly of Myddosome and Triffosome in stable transfected cells expressing either MyD88-GFP or TRIF-GFP. Co-IP assays using an anti-GFP mAb in LPS-stimulated TLR4/MD-2/CD14 -expressing HEK293 cells showed the LBP-dependent association of MyD88 with endogenously modified IRAK1 with a higher electrophoretic mobility (Fig. 8A). In addition, TRIF-TRAF3, TBK1, and TRAF6 association were also detectable when TLR4/MD-2/CD14-and TRIF-GFP-expressing Ba/F3 cells were stimulated with LBP (Fig. 8B). Only TRAF6 was associated with TRIF to a lesser extent in the absence of LBP. These results suggest that LBP efficiently mediates Myddosome and Triffosome assembly.
To gain further mechanistic insights underpinning how LBP regulates TLR4 signaling, we employed confocal microscopy to investigate the dissociation of surface TLR4 from the plasma membrane-anchored TIR domain-containing adaptor protein (TIRAP) fused to GFP using pulse-chase experiments with TLR4/MD-2/CD14-and TIRAP-expressing HEK293 cells (Fig. S4A). Although, irrespective of LBP, surface-labeled TLR4 co-localized with TIRAP-GFP on several plasma membranes in resting and stimulated cells, TLR4 was also observed in intracellular compartments in LPS-stimulated cells with LBP. To confirm these results, we further investigated the association of TLR4 with TIRAP-GFP using co-IP (Fig. S4B). Consistent with the microscopic observation, TIRAP was constitutively associated with TLR4 independently of LBP. These results suggest that LBP did not promote the dissociation of TLR4 from the plasma membrane of TIRAP in LPS-stimulated cells.

LPS-induced signal from TLR4 enhances LBP-mediated internalization
LPS triggers LBP-mediated TLR4 internalization; therefore, we investigated whether a specific TLR4 signal was required for internalization. C3H/HeJ-derived BMMs harboring a point mutation in the TIR signaling domain of TLR4 that rendered it unresponsive to LPS (39) failed to induce a signal, even if LPS was transferred from LBP and mCD14 to the TLR4 extracellular domain. Internalization of TLR4 in the mutated BMMs was attenuated compared with that of the WT C3H/HeN-derived BMMs during stimulation and at multiple concentrations of LPS (Fig. 9, A and B). These findings indicate that the TIRmediated signal from TLR4 enhances LBP-mediated internalization, but is not essential.
LPS-induced TLR4 internalization was reportedly mediated by GTPase dynamin, which is involved in most endocytic processes (12,40). Therefore, we investigated whether dynamin is also involved in LBP-mediated TLR4 internalization. The pandynamin inhibitor Dynasore significantly inhibited LBP-mediated TLR4 internalization in LPS-stimulated BMMs at 30 min but did so only modestly at 4 h compared with vehicle treatment (Fig. 9C, Fig. S5). Consistent with this result, Dynasore modestly inhibited or delayed the phosphorylation of IRF3 (Fig.  9D). These findings indicated LBP-mediated TLR4 internalization and IRF3 signaling were, at least in part, dynamin-dependent processes.

Discussion
This study revealed that LPS-induced TLR4 internalization required serum and mCD14. LBP-mediated LPS induced TLR4/MD-2 internalization, whereas recombinant LBP caused LPS-induced internalization in serum-free culture, provided that CD14 was expressed on the surface. Thus, interaction between LBP and mCD14, which transfers LPS to surface TLR4/MD-2, was required for internalization. The addition of LBP-N lacking the LPS transfer domain or the expression of mCD14 with decreased LPS binding attenuated internalization. Our present and previous studies showed that, when stimulated with LBP, TLR4 binds substantial amounts of LPS via MD-2 and induces the dimerization of TLR4 in mCD14-positive cells (16). Thus, TLR4/MD-2 dimers may form extensively in the presence of LBP and be a prerequisite for TLR4 internalization.
Similar to the case with TM receptors, TLR4 internalization is under the endocytic control of dynamin, clathrin, and their associated proteins (12,13,41). However, surface-exposed LPS-binding molecules, such as mCD14 and MD-2, regulate LPS-induced TLR4 internalization (13,24). Extracellular interactions between CD14 and MD-2 with LPS initiate internalization (13,24). Although LBP also binds lipopeptides other than LPS (31,32), we showed that TLR2 internalization was independent of serum and LBP. Collectively, these findings suggest that the extracellular interaction between LBP with LPS is a selective mechanism for TLR4 internalization. Additionally, we showed that LBP-mediated TLR4 internalization depends, at least in part, on the dynamin-mediated process because treatment with Dynasore significantly inhibited the internalization. However, the degree of inhibition was marginal at 4 h after stimulation, although it was more effective at 30 min. Consistent with our results, efficient but not complete inhibition was reported in a previous study in which internalization was assessed at 30 min by flow cytometry using BMMs (12). Significant but partial inhibition by Dynasore at 4 h may be due to incomplete inhibition of dynamin. Therefore, we increased the concentration of Dynasore up to 160 M; however, this concentration was insoluble, and crystal deposits formed in the cul- ture. In addition, the Dynasore appeared toxic to the cells at this concentration (data not shown). Previous studies showed the involvement of dynamin in TLR4 internalization (12,40), and the dynamin-dependent process is a crucial part of the LBPmediated TLR4 internalization mechanism; however, it is difficult to completely exclude the possibility of a dynamin-independent process.
As reported previously (13,24), membrane expression of CD14 was required for TLR4 internalization. A TLR4-independent signal is responsible for internalization in response to LPS (7,13), which may be transduced by CD14 via a lipid raft. Such a mechanism is unlikely, however, because the TM-type CD14 caused similar levels of internalization as the GPI-anchored CD14. Unexpectedly, C3H/HeJ BMMs had reduced TLR4 internalization compared with the C3H/HeN strain, which was inconsistent with previous findings indicating that the MyD88/ TRIF-mediated TLR4 signal does not participate in internalization (13). Thus, it may be necessary to test the hypothesis that a TIR-mediated TLR4 signal contributes to LBP-mediated internalization in a MyD88-and TRIF-independent manner. It is also possible that the LPS-induced TLR4 signal enhances internalization through a CD14-mediated TLR4-independent signal (13).
In addition to LBP, exogenous sCD14 also facilitated TLR4 internalization. However, the concentration required was higher than that for LBP. An LBP blocking mAb decreased serum-mediated internalization in mCD14-positive HEK293 cells, whereas a CD14-blocking mAb failed to inhibit internal-

LBP mediates LPS-induced TLR4 internalization and signaling
ization in BMMs. These findings indicate that LBP is the primary serum component responsible for LPS-induced TLR4 internalization, whereas CD14 plays little or no role.
We previously showed that sCD14 does not compensate for LBP in LPS binding or TLR4/MD-2 dimerization (16). This distinct function may be attributable to the binding and catalytic properties of LBP and CD14 for LPS. LBP binds to LPS micelles and repeats monomeric LPS transfer to CD14 multiple times (17). Conversely, sCD14 is not accessible to micelles, and directly binds LPS monomers via LBP. Thus, LBP may load LPS on TLR4 and induce a large cluster of TLR4 oligomers on the cell-surface, triggering its internalization. In pulse-chase experiments, relatively large blots, presumably due to the accumulation of antibody-bound surface TLR4 oligomers, were observed in intracellular compartments, whereas distribution of TLR4 was dispersed on the cell surface in stimulated cells without LBP.
This study showed that LBP-mediated internalization leads to the phosphorylation of TBK1, IKK⑀, and IRF3. This pathway is also inducible by the TRIF-mediated signal from endosomes (42,43). Additionally, IRF3 dimerized upon LPS stimulation in LBP-and serum-dependent manners, which resulted in enhanced IFN-␤ production (33). Consistent with this finding, the induction of IFN-␤ mRNA occurs upon LPS stimulation with LBP in BMMs. In addition to the essential role of LBP on TBK1-IKK⑀-IRF3-mediated IFN-␤ production due to TLR4 internalization, we also revealed that signaling varied with each TRIF-mediated event. LBP did not impact late-phase NF-B activation or necroptosis, and phosphorylation of RIP1 was mostly independent of LBP. TRIF plays an essential role

LBP mediates LPS-induced TLR4 internalization and signaling
through an IRF3-independent mechanism for both processes (38). TRIF controls signal transduction from the endosomes where TLR4 is internalized (1,7,12,13). The signals for delayed NF-B activation and the RIPK-mediated necroptosis may emanate from the cell-surface or other intracellular organelles.
We showed the induction of the MyD88-dependent cytokines, TNF␣, and IL-6 was not significantly impaired in the absence of LBP (i.e. TLR4 was not internalized). LPS-induced expression of TNF␣ and IL-6 was reportedly mediated by a TRIF signal (37) as well as by a MyD88 signal (44). Therefore, based on our results, TRIF may transduce certain types of signals from the plasma membrane unless TLR4 is internalized into the endosome. In fact, we detected the TRIF-TRAF6 association in stimulated cells in the absence of LBP, although it was

LBP mediates LPS-induced TLR4 internalization and signaling
weaker than that in the presence of LBP. Phosphorylation of ERK and p38 MAPKs is augmented by LBP as TBK1 and IKK⑀. JNK was phosphorylated only when stimulated with LBP. LPSinduced phosphorylation of MAPKs is temporally regulated in the early phase by a MyD88 signal, and it is regulated in the late phase by a TRIF signal (1,37,44). Therefore, the potentiation of MAPK activation by LBP can be explained by several possible mechanisms. An intracellular TRIF signal could be initiated from TLR4 internalized by LBP, as suggested by the LBP-dependent association of TRIF with TRAF3 and TBK1, which constitutes a critical signaling complex for IRF3-dependent IFN-␤ production. Alternatively, the sensitivity and intensity of TLR4 signaling by the catalytic transfer of LPS to CD14 (and then TLR4/MD-2) by LBP (17) is possibly increased, which should enhance both MyD88-and TRIF-dependent pathways. As shown in our results, MyD88-dependent signaling was augmented; the association of IRAK1 with MyD88, an upstream event of the MyD88-dependent pathway, was detected when stimulated with LBP. Alternatively, a MyD88 signal could be induced from intracellular compartments in which TLR4 is internalized via the association with TIRAP, as described below. A comprehensive understanding of the spatiotemporal regulation of TLR4 signaling that includes LBP-regulated signaling will require further studies to identify the location of signal induction from TLR4 as well as to elucidate the dependence of adaptor proteins. To the best of our knowledge, the subcellular location for signal induction leading to late-phase activation of NF-B and MAPKs remains unidentified.
In pulse-chase experiments, we showed that, irrespective of LBP, TIRAP co-localized with surface TLR4 on the plasma membrane in resting and stimulated cells but did not dissociate TLR4 from TIRAP. These findings, biochemically confirmed by co-IP assays, are plausible because TIRAP acts as a sorting adaptor for MyD88 to induce the signal from the plasma membrane (11). Notably, TIRAP co-localized with internalized TLR4 in the intracellular compartment in response to LPS stimulation only in the presence of LBP. Recently, TIRAP was demonstrated to mediate MyD88 signaling from endosomal TLRs in addition to cell-surface TLRs (e.g. TLR4) (45). Therefore, internalized TLR4 may have induced the MyD88 signal that led to the activation of MAPKs.
We confirmed that internalized TLR4 was not co-localized on the plasma membrane in LBP-stimulated cells, and that some are located on EEA1-positive endosomes. However, we unexpectedly found that internalized TLR4 was not always positive for EEA1. From this finding, we speculate that the subcellular compartments of internalized TLR4 may be heterogeneous and contribute to different signaling events.
In conclusion, we demonstrated that LBP mediates TLR4 internalization by transferring LPS to mCD14, which activates the TBK1-IKK⑀-IRF3 pathway via a Triffosome assembly. In addition to the unspecific intracellular regulation, the internalization of TLR4 is uniquely regulated by extracellular machinery composed of LBP, CD14, and MD-2. These findings detail a mechanism by which TLR4 induces divergent signals from spatially different subcellular locations.

Cells
Mouse macrophage RAW264 cells (RCB0535) and the human embryonic kidney cell line, HEK293 (CRL-1573), were purchased from Riken Cell Bank (Tsukuba, Japan) and American Type Culture Collection (Rockville, MD), respectively, and maintained in Dulbecco's modified Eagle's medium supplemented with 10% FCS. CHO-DG44 cells were provided by Dr. Fukudome (Saga University, Saga, Japan) and maintained in

Construction of expression vectors
The CD14 D10K mutant vector was generated by site-directed mutagenesis of the pEFBOS construct carrying the WT CD14 ORF (48). Primers were as follows: 5Ј-CTTGTGAGCT-GGACaagGAAGATTTCCGCTGCGTC-3Ј and 5Ј-GACGCA-GCGGAAATCTTCcttGTCCAGCTCACAAG-3Ј. A pEFBOS vector expressing human CD14 with the GPI anchor replaced with the TM domain was generated as follows. The TM domain (amino acids 218 -244) containing a stop codon (49) and BamHI (5Ј) and NotI (3Ј) restriction sites was synthesized by FASMAC (Atsugi, Japan) and subcloned into a pEFBOS construct (16) expressing human sCD14 tagged with FLAG sequences at the C terminus. The resulting construct expressed the signal peptide and amino acids 1-325, which are present in the GPI-anchored form of CD14, and is fused to the TM domain of the tissue factor.
The pCAGGS1 construct expressing C terminally FLAG/ His 6 -tagged mouse MD-2 was generated by subcloning from the pEFBOS construct (a kind gift from K. Miyake) at XhoI and NotI sites. The pEFBOS construct expressing mouse CD14 was generated by subcloning the SalI-and BamHI-digested ORF fragment of the EST clone (Open Biosystems, clone ID 4981093) into a XhoI-and BamHI-digested pEFBOS vector. A cDNA fragment coding for human MyD88 was amplified using PCR with a primer set (5Ј-agggtcgacaggaagcgctggcagacaATG-CGACC-3Ј and 5Ј-cagggatccGGGCAGGGACAAGGCCTTG-GCAAG-3Ј). Amplified fragments were digested with SalI and BamHI and subcloned into an XhoI-and BamHI-digested pEFBOS vector with modifications to introduce a C-terminal EGFP tag. The MyD88-GFP fragment was further subcloned into pcDNA3.1zeo(Ϫ) vector at XbaI and NotI sites.
The pEFBOS construct expressing human TRIF-GFP was generated as follows. The TRIF ORF was amplified using PCR with a primer set (5Ј-ctcctcgagATGGCCTGCACAGGCCCA-TCACTTCC-3Ј and 5Ј-acgggtaccTTCTGCCTCCTGCGTC-TTGTCCTCGG-3Ј). Amplified fragments were digested with XhoI and KpnI and subcloned into a pEFBOS vector with modifications to introduce a C-terminal EGFP tag. Four crucial amino acids (VQLG) in the RHIM domain of TRIF were mutated to alanines by site-directed mutagenesis using a primer set (5Ј-CCACCACGCACAGATGGcAgcGgcGGcaCT-GAACAACCACATGTGG-3Ј and 5Ј-CCACATGTGGTT-GTTCAGtgCCgcCgcTgCCATCTGTGCGTGGTGG-3Ј).

Preparation of sTLR4/MD-2-complexed protein
Expression constructs for sTLR4 tagged at the C terminus containing FLAG/His 6 or EGFP/His 6 tandem tags were generated as follows. A fragment of sTLR4 was subcloned into a pEFBOS vector, which was modified to introduce a FLAG/His 6 tandem tag at the XhoI and BamHI sites at the C terminus. An EGFP/His 6 tandem tag was amplified from a pEGFP-N1 (Clontech, Palo Alto, CA) vector by PCR with a primer set (5Ј-gtcggatccATGGTGAGCAAGGGCGA-3Ј and 5Ј-ggggcggccg-ctcaGTGGTGGTGGTGGTGGTGCTTGTACAGCTCGTC-CATGCCGAGAGTG-3Ј), and subcloned into the pEFBOS vector described above. The FLAG/His 6 -or EGFP/His 6 -tagged sTLR4 sequences were subcloned from modified pEFBOS constructs into a pCAGGS1 vector (sTLR4FH/pCAGID and sTLR4GH/pCAGID), which was also modified by introducing the IRES and DsRed-Express sequences, at XhoI and NotI sites. A cDNA fragment coding for human MD-2 was amplified by PCR with a sense primer (5Ј-ttctcaagcctcagacagtgg-3Ј), and an antisense primer (5Ј-aatagatctATTTGAATTAGGTTGGTG-TAGG-3Ј) (46). Amplified fragments were digested with XhoI and BglII and subcloned into an XhoI-and BamHI-digested pEFBOS vector with modifications to introduce a C-terminal HA tag (MD-2HA/pEFBOS). CHO-DG44 cells were transfected with sTLR4FH/pCAGID or sTLR4GH/pCAGID expression constructs using a Lipofectamine 2000 reagent. Stable clones secreting sTLR4F and sTLR4G were established by G418 selection. Additional co-transfection of MD-2HA/pEFBOS and LBP mediates LPS-induced TLR4 internalization and signaling pBABEpuro was performed in stable clones and selected with puromycin. CHO clones producing sTLR4F/MD-2 and sTLR4G/MD-2 were cultured for 2 to 3 weeks in serum-free medium. The conditioned medium was collected and dialyzed in PBS, and proteins were purified by affinity chromatography using a HisTrap HP column (GE Healthcare, Buckinghamshire, UK). The eluted proteins were dialyzed in PBS and passed through a 0.22-m sterile Millex-Gv filter (Millipore, Billerica, MA).

Detection of sTLR4 dimerization
sTLR4 dimerization was detected by co-IP using FLAG and GFP Abs. Following 30 min of incubation of sTLR4F/MD-2 and sTLR4G/MD-2 (10 g/ml each) with LPS (5 g/ml) in 100 l of PBS, rabbit anti-GFP Ab-immobilized Affi-Gel 10 (25 l) (16) was incubated for 1.5 h with gentle agitation, washed three times with 20 mM Tris buffer, pH 7.5, 150 mM NaCl, 1 mM EDTA, and 0.1% Triton X-100 and boiled in Laemmli buffer. Co-precipitated TLR4F and TLR4G were detected by Western blotting using anti-FLAG and GFP antibody followed by HRPconjugated secondary antibodies.

Flow cytometry
Cells were stained on ice with primary (5 g/ml) or biotinylated Ab (2 g/ml) in staining buffer (PBS containing 2% FCS, 10 mM EDTA, and 100 units/ml of penicillin), washed three times with staining buffer, and incubated with a PE-conjugated secondary antibody or Stv. Flow cytometry was performed using a FACS caliber (BD Biosciences) or CytoFLEX (Beckman Coulter Life Sciences, Brea, CA). Data were analyzed using the WinMDI program (J. Trotter, Scripps Research Institute, La Jolla, CA) or FlowJo (Tree Star, Ashland, OR). For the detection of LPS, cells were incubated with FITC-conjugated LPS (20 g/ml) for 20 min, washed, and analyzed as described above.

Cell cytotoxicity and viability assay
BMMs (1 ϫ 10 5 cells) were inoculated in a 96-well plate, cultured overnight in 100 l of the serum-free medium, and then stimulated with LPS in the presence of 20 -40 M Z-VADfmk or DMSO (0.1%). Resident peritoneal macrophages and BMMs were stimulated without overnight incubation. Following 24 h of stimulation, LDH release to the culture medium was assayed using the LDH Cytotoxicity Detection Kit (Takara, Shiga, Japan). Intracellular ATP content was determined by CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI).

Immunofluorescence analysis
HEK293-transfected cells expressing mouse TLR4/MD-2/ CD14 (1-2 ϫ 10 5 cells) were inoculated on poly-L-lysinecoated coverslips in 24-well plates and cultured overnight in 500 l of the serum-free medium. Following a 15-min incubation with anti-TLR4 mAb (UT49, 5-10 g/ml), cells were stimulated with LPS (1 g/ml) in the presence or absence of LBP (100 ng/ml) in serum-free medium at 37°C, allowing internalization of antibody-bound TLR4. At different time points, cells were washed twice with ice-cold PBS, fixed with 4% paraformaldehyde in PBS for 30 min, permeabilized with 0.1% Triton X-100 in PBS for 5 min, washed twice, and blocked with normal goat serum for 30 min at room temperature. The cells were incubated with rabbit anti-EEA1 Ab (1:400) followed by Alexa 488-conjugated goat anti-rabbit IgG and Alexa 546-conjugated F(abЈ) 2 goat anti-mouse IgG (4 -8 g/ml each) for 1 h. For

LBP mediates LPS-induced TLR4 internalization and signaling
plasma membrane staining, the cells were fixed with 4% paraformaldehyde in PBS for 30 min, washed twice with HBSS, and then incubated with CF488A-conjugated WGA (5 g/ml) in HBSS for 30 min at room temperature. After washing three times with HBSS, the cells were permeabilized with 0.1% Triton X-100, blocked, and incubated with Alexa 546-conjugated F(abЈ) 2 goat anti-mouse IgG, as above. Nuclei were stained with DAPI (1 g/ml) for 10 min at room temperature. Secondary Ab and DAPI were diluted in 5% BSA/PBS. After the incubation, cells were washed four times with 0.05% Tween 20 in PBS and mounted with Fluoromount Plus mounting medium (Diagnostic BioSystems, Pleasanton, CA). The specimens were visualized under a confocal laser scanning microscope LSM 700 (Carl Zeiss Microscopy, Jena, Germany) equipped with a Zeiss Axio Imager upright microscope with a ϫ40/0.95 Korr Plan-Apochromat objective. Fluorescent images were analyzed using the Zen lite software. The specificity of the staining was confirmed using isotype control mAb.