Transfer of Monomeric Endotoxin from MD-2 to CD14

Potent Toll-like receptor 4 (TLR4)-dependent cell activation by endotoxin depends on sequential transfer of monomers of endotoxin from an aggregated form to CD14 via the lipopolysaccharide-binding protein and then to MD-2. We now show that monomeric endotoxin can be transferred in reverse from MD-2 to CD14 but not to lipopolysaccharide-binding protein. Reverse transfer requires a ∼1000-fold molar excess of CD14 to endotoxin-MD-2. Transfer of endotoxin from MD-2 to extracellular soluble CD14 reduces activation of cells expressing TLR4 without MD-2. However, transfer of endotoxin from MD-2 to membrane CD14 (mCD14) makes cells expressing MD-2·TLR4 sensitive to activation by the endotoxin-MD-2 complex. An endotoxin-mutant (F126A) MD-2 complex that does not activate cells expressing TLR4 alone potently activates cells expressing mCD14, MD-2, and TLR4 by transferring endotoxin to mCD14, which then transfers endotoxin to endogenous wild-type MD-2·TLR4. These findings describe a novel pathway of endotoxin transfer that provides an additional layer of regulation of cell activation by endotoxin.

Endotoxins are unique, highly abundant surface glycolipids of Gram-negative bacteria. They have the capacity to potently induce proinflammatory responses of many multicellular hosts. With the discovery of the acute phase protein, lipopolysaccharide-binding protein (LBP), 2 CD14, and, later, the Toll-like receptor 4 (TLR4) and MD-2, it has become clear that each of these proteins plays a key role in the ability of many mammalian hosts to mount highly sensitive responses to endotoxin (1)(2)(3)(4)(5). MD-2 is required for TLR4-dependent cell activation of mammalian cells by endotoxin, whereas LBP and CD14 are needed for maximal sensitivity. This sensitivity is needed to permit host cells to respond to pg/ml concentrations of endotoxin and thereby trigger defensive host responses to small numbers of Gram-negative bacteria soon after bacterial invasion, before the host is overwhelmed by bacterial growth.
We have made use of metabolically radiolabeled purified endotoxin to better define how LBP, CD14 and MD-2 together make possible potent TLR4-dependent cell activation by endotoxin (6 -8). On the basis of these and many other studies, we and others have proposed that potent TLR4-dependent cell activation by endotoxin depends on sequential protein-endotoxin and protein-protein interactions between endotoxin, LBP, CD14, MD-2, and TLR4 (1-4, 9 -11). Thus, LBP binds to endotoxin as presented natively in the bacterial outer membrane or as aggregates after extraction and purification (4,(12)(13)(14)(15). In so doing, LBP catalyzes delivery and transfer of individual molecules of endotoxin to CD14 (either as a soluble extracellular protein (sCD14) or GPI-linked membrane protein (mCD14)) (4,7,(13)(14)(15)(16). The monomeric endotoxin-CD14 complex that is formed is the preferred substrate for MD-2, whether present in soluble extracellular form (sMD-2) or membrane-associated as a complex with TLR4 (2,9,11). The transfer of endotoxin from CD14 to MD-2, coupled with binding of MD-2 to TLR4, is apparently required for TLR4 activation by endotoxin (2,11).
Consistent with this ordered pathway of protein/endotoxin and protein/protein interactions, cells expressing heterodimeric MD-2⅐TLR4 complexes (without mCD14) can be potently activated by monomeric endotoxin (E)⅐CD14 but not by purified endotoxin aggregates with or without LBP or by monomeric E⅐MD-2 complex (2,16). In contrast, cells expressing TLR4 without MD-2 (e.g. airway epithelial cells) can be potently activated by E⅐MD-2 complex but not by E⅐CD14 (17). The inability of the monomeric E⅐MD-2 complex to potently activate cells expressing MD-2⅐TLR4 is consistent with the inability of endotoxin in E⅐MD-2 to be readily transferred to MD-2 (bound to TLR4) or the soluble E⅐MD-2 complex to readily exchange with MD-2 that is bound to TLR4 (9). Thus, under circumstances (e.g. infection and inflammation) in which increases in MD-2 expression (17-20) may lead to both increased formation of extracellular E⅐MD-2 and increased occupation of TLR4 by MD-2, the fate of the extracellular E⅐MD-2 is unclear.
In this study, we have considered the possibility that CD14 could provide an alternative target for the monomeric E⅐MD-2 complex and thereby provide an alternative route for TLR4-dependent cell activation and, perhaps, for clearance of E⅐MD-2 complexes. Using endotoxin (lipooligosaccharide; LOS) from an acetate auxotroph of Neisseria meningitidis, which we have previously described (6), we show direct transfer of endotoxin from MD-2 to CD14 that, depending on the properties of the host cell and surrounding extracellular fluid, can either attenuate or promote cell activation by endotoxin. These findings thus describe a novel pathway of endotoxin transfer that provides an additional layer of regulation of cell activation by endotoxin. Recombinant Proteins-Insect-derived soluble MD-2 containing a hexapolyhistidine tag on the C-terminal end was prepared as previously described (11). Preparative amounts of the mutant MD-2 F126A were generated from infections of High Five insect cells with baculovirus containing the gene for human MD-2 F126A inserted into pBAC11. MD-2 cDNA was isolated, linearized, and inserted, using NcoI and XhoI-sensitive restriction sites, into the baculovirus transfection vector pBAC11 (Novagen) that provides a 6-residue polyhistidine (His 6 ) tag at the carboxyl-terminal end of MD-2 and 5Ј-flanking signal sequence (gp64) to promote secretion of the expressed protein. DNA encoding MD-2 F126A was sequenced in both directions to confirm fidelity of the product.

Materials
The fragment of human TLR4 corresponding to the predicted ectodomain, amino acid residues 24 -634, (TLR4 ECD ) was generated by transient transfection of HEK293T cells. HEK293T cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum. Expression vectors containing DNA of interest for production of FLAG-TLR4 ECD , amino acids 24 -634 (pFLAG-CMV-TLR4), and MD-2-FLAG-His (pEF-BOS) have been previously described and characterized (21). HEK293T cells were transiently transfected with the plasmids (pEF-BOS, pFLAG-CMV) encoding MD-2 or TLR4 ECD fused in frame to the coding region of the signal sequences within the plasmid. Conditioned medium was harvested after a 48-h culture. The medium containing the expressed proteins was concentrated 10ϫ using a Millipore Centricon-10 concentrator (Amicon). Control conditioned medium was obtained from HEK293T cells transiently transfected with an empty vector. Cells (ϳ80% confluence in 6-well plates or T75 flask for preparative amounts of material) were transfected with 4 g (6-well plate) or 20 g (T75 flask) of DNA using PolyFect reagent (Qiagen). After 12-16 h, plates were rinsed in phosphate-buffered saline, and 2 ml of serum-free medium (293 SFM; Invitrogen) plus 0.4% HSA was added.
Media containing expressed proteins were collected 24 -48 h later and stored at 4°C until used. The reactivity of conditioned medium containing secreted TLR4 ECD /MD-2 maintained reactivity with [ 3 H]LOS⅐sCD14 for at least 6 months at 4°C.
Preparation of Metabolically Labeled Endotoxins-Lipooligosaccharide ([ 14 C]LOS (600 cpm/ng) or [ 3 H]LOS (5000 cpm/ ng)) was isolated after growth and metabolic labeling of an acetate auxotroph of N. meningitides serogroup B as previously described (6). An msbB derivative of N. meningitidis serogroup B, NMBA11K3cap Ϫ , was obtained from Dr. Michael Apicella (University of Iowa). The strain was grown in minimal medium supplemented with radioactive [ 3 H]acetate or [ 14 C]acetate, and LOS (500 or 80 cpm/ng, respectively) was isolated as previously described (22,23). After extraction of LOS with hot phenol/ water, precipitation with ethanol and resuspension by sonication in distilled water, LOS aggregates were further purified by ultracentrifugation (16,22,23). [   To minimize contamination of LOS⅐MD-2 with any small remaining amount of LOS⅐sCD14, the peak and downslope fractions of LOS⅐MD-2 were repurified by Sephacryl S200 using a long column (1.6 ϫ 70 cm) to further increase resolution of LOS⅐MD-2 from LOS⅐sCD14.
Samples were applied in Յ1-ml fractions and collected (flow rate, 0.3-0.5 ml/min) at room temperature using AKTA or AKTA Purifier FPLC (GE Healthcare). Aliquots of the collected fractions were analyzed by liquid scintillation spectroscopy using a Beckman LS liquid scintillation counter to detect radioactive LOS; recoveries of LOS were Ն70%. All solutions used were pyrogen-free and sterile-filtered. Radiochemical purity of [ 3 H]LOS aggregates was confirmed by Sephacryl S500 chromatography, and that of [ 3 H]LOS⅐sCD14 and of [ 3 H]LOS⅐MD-2 was confirmed by Sephacryl S200 chromatography (6,11). After chromatography, selected fractions to be used in bioassays were pooled and sterile-filtered (0.22-m pore size) with Ͼ90% recovery. Fractions were stored under sterile conditions at 4°C until needed. Sephacryl S200 columns were calibrated with Bio-Rad gel filtration standards that included thyroglobulin, globulin, ovalbumin, myoglobin, vitamin B 12 , and human serum albumin.
Cell Activation Assays-Human umbilical vein endothelial cells were cultured on collagen-coated plasticware (Costar, Cambridge, MA) at 37°C, 5% CO 2 , and 95% relative humidity in endothelial basal medium supplemented with 5% fetal bovine serum, 12 g/ml bovine brain extract, 10 ng/ml human endothelial growth factor, 1 g/ml hydrocortisone, and 50 g/ml gentamicin. HEK293-derived cell lines have been extensively characterized and were cultured, as has been described (18) or according to recommendations of Invivogen Inc. Cells were subcultured and grown to confluence (ϳ3-5 days) in 96-well plates. Before use in bioassays, confluent monolayers were washed twice with warm phosphate-buffered saline, pH 7.4, to ensure that incubation mixtures were serum-free. Human peripheral blood mononuclear cells (PBMCs) were obtained from heparinized venous blood from healthy donors, with informed consent. Purified PBMC contained ϳ80% lymphocytes, 20% monocytes, and Ͻ2% polymorphonuclear leukocytes, as previously described (6,24). Washed cells (1 ϫ 10 5 ) were incubated for 4 -20 h at 37°C in 5% CO 2 and 95% humidity in Dulbecco's modified Eagle's medium, 0.1% HSA (200 l) with the indicated supplements in 96-well plates. Activation of cells was assessed by measuring accumulation of extracellular IL-8 by enzyme-linked immunosorbent assay (Clontech).

RESULTS
Transfer of Endotoxin from MD-2 to CD14-To test the hypothesis that endotoxin could be transferred from a monomeric E⅐MD-2 complex to CD14, we first examined the effect of increasing concentrations of sCD14 on the ability of purified LOS⅐MD-2 to activate HEK293/TLR4 cells. In contrast to E⅐MD-2, monomeric E⅐sCD14 does not activate these cells (2,11). Thus, transfer of endotoxin from E⅐MD-2 to sCD14 should be accompanied by reduced cell activation.
As shown in Fig. 1A, sCD14 produced a dose-dependent inhibition of cell activation by LOS⅐MD-2 in HEK/TLR4 cells. Nearly complete inhibition by added sCD14 required a 1000 -3000-fold molar excess of sCD14 to E⅐MD-2. There was virtually quantitative conversion of [ 3 H]LOS⅐MD-2 (M r ϳ 25,000) to [ 3 H]LOS⅐sCD14 (M r ϳ 60,000) under these conditions, as demonstrated by gel filtration chromatography (Fig. 1B) and immuno-co-capture of the generated product with an anti-CD14 antibody (18E12) (16) that reacts with CD14 outside of its endotoxin binding site (data not shown). In contrast to the effects of sCD14, LBP had no effect on cell activation by LOS⅐MD-2 (Fig. 1A) and, even at 3000-fold molar excess, showed no interaction with purified [ 3 H]LOS⅐MD-2 (Fig. 1B).
CD14-dependent Activation by E⅐MD-2 of Cells Expressing mCD14, MD-2, and TLR4-Although transfer of endotoxin from E⅐MD-2 to sCD14 inhibited TLR4-dependent activation of cells expressing TLR4 without MD-2 or mCD14, it seemed possible that the same intermolecular transfer of endotoxin could promote activation of cells that expressed mCD14 as well as MD-2 and TLR4. To test this hypothesis, we compared the ability of LOS aggregates plus substoichiometric concentrations of LBP, monomeric LOS⅐sCD14, and LOS⅐MD-2 complexes to activate HEK293/ mCD14/MD-2/TLR4 cells. Remarkably, LOS⅐MD-2 was nearly as potent as LOS⅐sCD14 and at least as potent as LOS aggregates plus LBP in inducing activation of these cells ( Fig. 2A). Because mCD14 is very highly expressed in these stable cell transformants, we repeated the same experiments with primary cells (i.e. PBMCs containing human monocytes) expressing mCD14/MD-2/TLR4. Fig. 2B shows that LOS⅐MD-2 also induced dose-dependent secretion of IL-8 from PBMC (monocytes), although the potency of LOS⅐MD-2 toward these cells was 3-10-fold less than that of LOS⅐sCD14. Activation of monocytes by LOS⅐MD-2 as well as by LOS aggregates plus LBP and by LOS⅐sCD14 at 10 pM LOS was nearly completely inhibited by an antibody to CD14 (MY-4) that blocks the endotoxinbinding site of CD14 (Fig. 2C). This is consistent with a role for endotoxin transfer from MD-2 to CD14 in activation of monocytes by monomeric LOS⅐MD-2. Similar observations were made with cultured endothelial cells that express MD-2 and TLR4 and low levels of mCD14 (data not shown) (6,16,25). These cells were potently activated by LOS⅐sCD14 but much less by LOS⅐MD-2 (Fig. 2D). Cell activation by LOS⅐MD-2 was further reduced by pretreatment of the endothelial cells with anti-CD14 antibody (data not shown), strongly suggesting that transfer of LOS from LOS⅐MD-2 to mCD14 was necessary in these cells as well for maximal activation by LOS⅐MD-2.
Contrasting Properties of LOS⅐MD-2 Complexes Containing either Mutant LOS or Mutant MD-2-We have recently characterized mutant LOS⅐MD-2 complexes containing either pentaacylated LOS or MD-2 F126A that have markedly reduced TLR4 agonist properties when tested with cells expressing only TLR4 (e.g. HEK293/TLR4 cells) (Fig. 3A) (23, 37). These mutant complexes inhibit TLR4 activation by wild-type (WT) LOS⅐MD-2, indicating that the mutant LOS⅐MD-2 complexes bind TLR4 but do not efficiently trigger TLR4 activation. If activation of cells expressing mCD14/MD-2/TLR4 by LOS⅐ MD-2 is mediated by transfer of LOS from added LOS⅐MD-2 to mCD14 and then to membrane MD-2⅐TLR4, rather than by direct interaction of added LOS⅐MD-2 with TLR4, a complex containing an underacylated endotoxin species such as LOS msbB ⅐MD-2 WT should still show reduced agonist properties relative to LOS WT ⅐MD-2 WT . In contrast, LOS WT ⅐MD-2 F126A should activate monocytes as potently as LOS WT ⅐MD-2 WT , since LOS WT will be transferred to endogenous MD-2 WT . Data presented in Fig. 3B are consistent with this scenario. Dose-dependent activation of PBMCs by LOS WT ⅐MD-2 WT and LOS WT ⅐MD-2 F126A was nearly identical, whereas that by LOS msbB ⅐MD-2 WT was reduced. These findings strongly suggest that monomeric LOS⅐MD-2 can potently activate cells expressing mCD14, MD-2, and TLR4 by transferring LOS to mCD14. This conclusion was further supported by a demonstration that activation of PBMC by all three LOS⅐MD-2 complexes (LOS WT ⅐MD-2 WT , LOS WT ⅐MD-

sCD14-dependent Transfer of [ 3 H]LOS from [ 3 H]LOS⅐MD-2 to
MD-2⅐TLR4 ECD -The findings described above strongly suggest that endotoxin from E⅐MD-2 can be delivered to MD-2⅐TLR4 in a CD14-dependent manner. To demonstrate this more directly, we made use of a novel cell-free assay (9) (Fig. 4).
The addition of sCD14 to the incubation mixture markedly increased the yield of this M r ϳ 190,000 complex with a corresponding reduction in [ 3 H]LOS⅐MD-2 (Fig. 4). These findings demonstrate directly the ability of CD14 to facilitate delivery of endotoxin from E⅐MD-2 to MD-2⅐TLR4, presumably via formation of an E⅐CD14 complex.

DISCUSSION
We have demonstrated, for the first time, intermolecular transfer of endotoxin from MD-2 to CD14. Many intermolecular transfer reactions involving endotoxin have been demonstrated before, most notably from LBP to CD14 and from CD14 to MD-2, reactions that are key steps in potent TLR4-dependent cell activation by endotoxin. In addition, delivery of endotoxin via E⅐sCD14 to plasma lipoproteins and via LBP-coated endotoxin aggregates to scavenger receptor(s) and lipoproteins has been demonstrated (26 -30). Whether endotoxin itself or the intact endotoxin-protein complex/aggregate is transferred in those reactions has not been established. The various reactions of endotoxin involving LBP and CD14 have underscored the complexity of host/endotoxin interactions and demonstrated that a single endotoxin-protein complex/aggregate may have multiple host acceptors and, as a result, be linked to different functional outcomes (e.g. MD-2⅐TLR4-dependent cell activation versus scavenger receptor or lipoprotein-mediated endotoxin clearance).
The reverse transfer of endotoxin from MD-2 to CD14 documented in this study reveals that E⅐MD-2 has at least two host acceptors, TLR4 and CD14 (Fig. 5). Depending on the abundance and reactivity of acceptors of E⅐CD14 (e.g. MD-2⅐TLR4, LBP, and lipoproteins), the transfer of endotoxin may, as we have shown, lead either to inhibition (Fig. 1) or promotion (Figs. 2 and 3) of TLR4-dependent cell activation by endotoxin (Fig.  5). The latter effect may be more likely in tissues or under inflammatory conditions where levels of secreted MD-2 (resulting from synthesis of MD-2 in molar excess to that of TLR4) relative to sCD14, LBP, and lipoproteins or other endo-

Endotoxin Transfer from MD-2 to CD14
toxin scavengers may favor formation of E⅐MD-2. Our findings indicate that both cells expressing TLR4 without MD-2 (such as airway epithelial cells) and cells containing mCD14 and MD-2⅐TLR4 (e.g. macrophages) could be targets of E⅐MD-2, the former by direct interaction of E⅐MD-2 with TLR4 and the latter by transfer of endotoxin from extracellular E⅐MD-2 to membrane CD14 and then MD-2⅐TLR4 (Fig. 5). The efficiency of transfer of endotoxin from extracellular E⅐MD-2 to mCD14 is likely to depend on the abundance of mCD14 (Fig. 2). These properties suggest that E⅐MD-2 could provide, pharmacologically, a particularly advantageous vehicle for delivery and transfer of endotoxin agonists and antagonists to desired tissue sites of action, especially if its relatively small mass and favorable aqueous solubility permit penetration into tissue more rapidly than transfer of endotoxin from E⅐MD-2 to CD14 in the circulatory system. In comparison with E⅐sCD14, E⅐MD-2 is much more stable and soluble in aqueous environments. These properties suggest tighter binding of endotoxin to MD-2 than to CD14, with sequestration of the fatty acids of endotoxin within a hydrophobic cavity that is deep in MD-2 (31-33) but wide and shallow in CD14 (34). In support of this view, E⅐MD-2 is relatively resistant to the endotoxin deacylase, acyloxyacyl hydrolase, whereas E⅐sCD14 is a favorable substrate for this enzyme (13). Based on these biochemical and physical chemical properties, we have previously predicted that transfer of endotoxin occurs much less readily, if at all, from MD-2, in contrast to the facile transfer of endotoxin from CD14 (2,11,13). The findings presented in this study support this view. Although the transfer of endotoxin from CD14 to MD-2 occurs readily at molar ratios of MD-2 to E⅐sCD14 of Ͻ10 (2,11,23), the reverse transfer requires molar ratios of CD14 to E⅐MD-2 of Ն300 (Fig. 1). Moreover, endotoxin exchange occurs readily between CD14 molecules (e.g. sCD14 and mCD14) but to only a very limited extent, if at all, between E⅐MD-2 and MD-2⅐TLR4 (Fig. 4). The facile exchange of endotoxin between CD14 molecules (i.e. sCD14 and mCD14) may explain the remarkably potent activity of LOS⅐sCD14 toward PBMCs (Fig. 2, B and C). Half-maximal cell activation is produced in that setting by 10 pM LOS⅐sCD14, well below the apparent K d (ϳ130 pM) of LOS transfer from LOS⅐sCD14 to MD-2⅐TLR4 (9). The very limited reaction of LOS⅐MD-2 with conditioned medium containing MD-2 and TLR4 ECD may reflect 1) limited transfer of endotoxin from E⅐MD-2 to MD-2⅐TLR4, 2) limited exchange of E⅐MD-2 with MD-2 associated with TLR4, or 3) limited amounts of free TLR4 ECD in culture medium containing MD-2 and TLR4 ECD . In any case, the addition of sCD14 in molar excess substantially increases the net transfer of endotoxin from E⅐MD-2 to MD-2⅐TLR4, fully consistent with the role of mCD14 in facilitating activation by E⅐MD-2 of cells containing MD-2⅐TLR4. The accumulation of (E⅐MD-2⅐TLR4 ECD ) 2 , but little or no E⅐sCD14, in these incubation mixtures (Fig. 4) despite a 100-fold molar excess of sCD14 underscores how favorable the transfer of endotoxin from CD14 to MD-2 is in comparison with the reverse reaction. This characteristic is consistent with and presumably necessary for the remarkable potency of cell activation by endotoxin, including in cells that may have a substantial molar excess of mCD14 to MD-2⅐TLR4. The efficiency of transfer of endotoxin from CD14 to MD-2⅐TLR4 is probably further increased when mCD14 and MD-2⅐TLR4 are present within the same membrane. This may explain why, at very low LOS⅐sCD14 concentrations, cell activation is dependent on mCD14 (i.e. blocked by anti-CD14 antibodies). At higher (Ͼ200 pM) concentrations of LOS⅐sCD14, cell activation is not blocked by anti-CD14 antibodies that block transfer of endotoxin between sCD14 and mCD14 (e.g. MY-4 and MEM18), probably reflecting at these higher concentrations direct transfer of LOS from extracellular LOS⅐sCD14 to membrane MD-2⅐TLR4.
In contrast to the ability of endotoxin to be transferred from MD-2 to CD14, we observed no reaction of E⅐MD-2 with LBP even at a 3000-fold molar excess of LBP. This suggests that the molecular requirements for transfer of endotoxin from MD-2 mirror the requirements for delivery of endotoxin to MD-2. Endotoxin is readily transferred to MD-2 from E⅐CD14 but not from LBP-coated aggregates of purified E or from complexes of E⅐sCD14 that have been induced to aggregate by reaction with LBP (11,30). It seems likely that transfer of endotoxin between endotoxin-binding proteins is preceded by docking between the endotoxin-protein complex or aggregate and the recipient endotoxin-binding protein and thus may be driven by proteinprotein as well as protein-endotoxin interactions. Therefore, the apparently specific exchange of endotoxin between CD14 and MD-2 may depend on (weak) interactions between CD14 and MD-2 that are not present between LBP and MD-2. The ability of added LBP to dampen cell activation by endotoxin and shift endotoxin away from association with MD-2(TLR4) (26,30,35,36) probably depends on the ability of mCD14 to act as an intermediate in transfer, perhaps by virtue of its ability to interact with both LBP and MD-2. Whether or not transfer of endotoxin from MD-2 to TLR4 follows engagement of E⅐MD-2 with TLR4 (or of endotoxin with MD-2⅐TLR4) and is an important step in endotoxin-induced TLR4 activation remains an open question.