Binding of Bacterial Peptidoglycan to CD14*

The hypothesis that soluble peptidoglycan (sPGN, a macrophage-activator from Gram-positive bacteria) binds to CD14 (a lipopolysaccharide (LPS) receptor) was tested. sPGN specifically bound to CD14 in the following three assays: binding of soluble32P-CD14 (sCD14) to agarose-immobilized sPGN, enzyme-linked immunosorbent assay, and photoaffinity cross-linking. sCD14 also specifically bound to agarose-immobilized muramyl dipeptide or GlcNAc-muramyl dipeptide but not to PGN pentapeptide. Binding of sCD14 to both sPGN and ReLPS (where ReLPS is LPS from Salmonella minnesota Re 595) was competitively inhibited by unlabeled sCD14, 1–152 N-terminal fragment of sCD14, sPGN, smooth LPS, ReLPS, lipid A, and lipoteichoic acid but not by dextran, dextran sulfate, heparin, ribitol teichoic acid, or soluble low molecular weight PGN fragments. Binding of sCD14 to sPGN was slower than to ReLPS but of higher affinity (K D = 25 nm versus41 nm). LPS-binding protein (LBP) increased the binding of sCD14 to sPGN by adding another lower affinity K D and another higher B max, but for ReLPS, LBP increased the affinity of binding by yielding twoK D with significantly higher affinity (7.1 and 27 nm). LBP also enhanced inhibition of sCD14 binding by LPS, ReLPS, and lipid A. Binding of sCD14 to both sPGN and ReLPS was inhibited by anti-CD14 MEM-18 mAb, but other anti-CD14 mAbs showed differential inhibition, suggesting conformational binding sites on CD14 for sPGN and LPS, that are partially identical and partially different.

and arthritis. Most of these effects are due to the release of cytokines and other mediators from macrophages and other cells (2)(3)(4)(5)(6)(7). To prevent deleterious effects of these bacterial products on the host, it is essential to understand in detail the mechanism of cell activation by PGN and LPS. Since the first step in cell activation is the interaction of the activator with the cell surface receptor, we decided to identify and characterize the receptors for PGN on macrophages.
CD14 is a glycosylphosphatidylinositol-linked 55-kDa protein present on the surface of macrophages and polymorphonuclear leukocytes that functions as the cell surface receptor for LPS (8 -14). We have recently shown that CD14 also functions as a receptor for PGN, because PGN-unresponsive CD14negative cells become responsive to PGN following transfection of these cells with CD14 and expression of CD14 on their cell surface (15), and because anti-CD14 mAbs inhibit activation of CD14-positive cells by PGN (16) and also inhibit binding of PGN to CD14-positive cells (17). Other studies have also shown that CD14 may function as a receptor for other bacterial products, such as mannuronic acid polymers from Pseudomonas (18), insoluble cell wall fragments from several Gram-positive bacteria (15,19), mycobacterial lipoarabinomannan (15,19,20), rhamnose-glucose polymer from Streptococcus (21), and lipoteichoic acid (LTA) (22) or LTA-like molecule (23) from Gram-positive bacteria.
It was, therefore, proposed that CD14 functions as a "pattern recognition receptor" (19) that can recognize shared features of microbial cell surface components and enable the host to respond to pathogenic bacteria but not to a great variety of other non-microbial polysaccharides. This model, which assumes that CD14 can discriminate between different ligands and can control the specificity of macrophage responses, was recently questioned (24) based on the inability of CD14 to discriminate between agonistic and antagonistic derivatives of LPS (25). The second alternative model was proposed (24), according to which CD14 would function as an albumin-like carrier molecule that binds a large variety of molecules without recognition specificity and that it would then transfer these molecules to another, as yet unidentified recognition/cell-activating molecule in the cell membrane. In a third "combinatorial" model (24), both CD14 and the putative recognition/cell-activating molecule would contribute to the specificity of the response. In addition, other possible models can be proposed, e.g. bacterial cell wall products other than LPS (such as PGN), could bind to another molecule, which in turn would activate CD14, or CD14 could be a part of a receptor complex, and this complex would bind PGN without physical binding of PGN to CD14.
To discriminate between these models and to validate the direct function of CD14 as the receptor for a variety of bacterial products, detailed analysis of binding of these bacterial products to CD14 is needed. Because to date detailed studies of binding to CD14 were only performed with enterobacterial LPS (10 -13, 26 -35), the aim of this study was to determine if PGN directly and specifically binds to CD14. We show that PGN binds with high affinity to both soluble and membrane CD14 and that this binding can be competitively inhibited not only by LPS and sPGN but also by LTA. Because the binding of LPS to CD14 is facilitated by the LPS-binding protein (LBP) present in plasma (8 -13, 26 -28, 30, 33), we have also determined the effect of LBP on the binding of PGN to CD14.

EXPERIMENTAL PROCEDURES
Materials-Soluble PGN (sPGN) is a polymeric uncross-linked PGN of approximate average M r ϭ 125,000, released from Staphylococcus aureus grown in the presence of ␤-lactam antibiotics (5,36). sPGN was purified by vancomycin affinity chromatography from three different S. aureus strains (Rb, 845, and 3528) (5,36,37). Vancomycin specifically binds to the D-Ala-D-Ala part of PGN, the structure uniquely found in uncross-linked PGN and biosynthetic PGN precursors (this accounts for the specificity of vancomycin inhibition of peptidoglycan synthesis). PGN precursors are of low molecular weight and were removed by extensive dialysis. By quantitative chemical analysis, the sPGN was more than 98% pure (5,36), i.e. known PGN amino acids and amino sugars accounted for more than 98% of the mass of the sPGN preparation. sPGN contained Ͻ24 pg of endotoxin/mg determined by the Limulus lysate assay (37). sPGN from three strains yielded the same results in the binding assays. Ribitol teichoic acid was obtained from S. aureus cell walls by trichloroacetic acid extraction as described (5) and after hydrolysis contained glucosamine, anhydroribitol, and alanine.
Purified LPS, obtained from Escherichia coli O113 by phenol/water extraction (refined endotoxin standard, approximate average M r ϭ 15,000), and diphosphoryl lipid A from Salmonella typhimurium Re mutant (M r ϭ 1797) were obtained from Ribi Immunochem Research (Hamilton, MT). LPS from Salmonella minnesota Re 595 (ReLPS, a minimal naturally occurring endotoxic structure of LPS, M r ϭ 2000 -3000), obtained by phenol/chloroform/petroleum ether extraction, was purchased from Sigma, and its purity was analyzed as described before (37). Lipid A and ReLPS were dissolved at 2.5 mg/ml in 0.2% triethylamine (37).
sCD14 and LBP-Recombinant human full-length (residues 1-323) sCD14, containing at the C terminus five amino acids that constitute protein kinase A phosphorylation site followed by an affinity tag of six histidines, was expressed in a baculovirus system, affinity purified on nickel-agarose, and analyzed as described (40). Recombinant human 1-152 sCD14 fragment, tagged at the C terminus with six histidines, was expressed in a baculovirus system, affinity purified on nickelagarose, and analyzed as described for the full-length sCD14 (40). Recombinant human LBP was purified by affinity chromatography from supernatants of CHO transfectants as described (41).
Binding of 32 P-sCD14 to sPGN-Agarose or ReLPS-Agarose-Five g of sCD14 or soluble dephosphorylated casein from bovine milk as a control (C-4765, Sigma) were labeled with 32 P by incubating with 0.25 mCi of [␥-32 P]ATP (NEN Life Science Products) and 5 units of catalytic subunit of cAMP-dependent protein kinase from bovine heart (P2645, Sigma) in a total volume of 50 l for 2 h at 37°C (40). 100 g of bovine serum albumin (BSA) was then added, and the reaction mixture was dialyzed twice at 4°C against total 700 ml of PBS, 3.5-kDa cutoff. The specific activities of 32 P-labeled sCD14 and casein were 12-36 and 2.7-5.7 Ci/g, respectively, and their purity was checked by polyacrylamide gel electrophoresis (PAGE) and autoradiography (see Fig. 5 below). sPGN, MDP, GlcNAc-MDP, pentapeptide, or ReLPS were coupled to agarose (Sepharose CL-4B, Sigma) by mixing 2.7-3.3 mg/ml sPGN or 5 mg/ml MDP, GlcNAc-MDP, pentapeptide, or ReLPS with 50% CNBractivated (42) agarose in 0.15 M phosphate buffer, pH 8.3, at 4°C for 5 days, followed by blocking with 1 M ethanolamine and extensive washing with PBS, pH 7.2, with 0.1% NaN 3 . Coupling efficiency was 1.45-1.5 g of sPGN/l of agarose, determined by scintillation counting of sPGN biosynthetically labeled with [ 14 C]Ala. Control agarose was treated in the same way, but the preparations were omitted. Heating the agarose after coupling with 4% SDS at 80°C for 30 min, followed by extensive washing, did not change the sCD14 binding, indicating that the preparations were covalently coupled to agarose.
For the standard binding assay, 2 l of sPGN-agarose, MDP-agarose, GlcNAc-MDP-agarose, or 1 l of ReLPS-agarose, or 1 or 2 l of control agarose were incubated with 2 ng of 32 P-labeled sCD14 or casein, with or without 1 g/ml LBP, in Dulbecco's PBS without Ca 2ϩ and Mg 2ϩ supplemented with BSA and gelatin (1 mg/ml each) in a total volume of 60 l, for 60 (sPGN) or 20 min (ReLPS) at 37°C with frequent mixing. The incubation was ended by adding 190 l of ice-cold PBS with NaCl (to yield 0.5 M final NaCl concentration); the agarose suspension was overlaid onto 0.8 M sucrose in Dulbecco's PBS without Ca 2ϩ and Mg 2ϩ in two 400-l polypropylene tubes and centrifuged for 5 min at 3000 ϫ g. The tubes were frozen, and the tips, containing the sedimented agarose with bound radioactivity, were cut off and placed in the scintillation vials for counting. In the saturation binding experiments, the unbound radioactivity (free ligand) in the tube tops was also counted. In some experiments, the amount of agarose, incubation time, or sCD14 concentration were varied, as indicated under "Results." In the competitive inhibition experiments, inhibitors were first added to sPGN-agarose-, MDP-agarose-, or ReLPS-agarose-containing tubes, followed by the addition of 32 P-sCD14, except for inhibition with unlabeled sCD14 or 1-152 sCD14 fragment, which were first mixed with 32 P-sCD14, followed by addition of sPGN-agarose or ReLPS-agarose. In mAb inhibition experiments, 32 P-sCD14 was first incubated with mAbs for 60 min at 37°C, followed by addition of sPGN-agarose or ReLPS-agarose.
The apparent dissociation constant, K D , defined as the concentration of free ligand required to saturate half of the available binding sites, and B max (the maximal binding at saturation) were calculated by curve fitting to the hyperbolic function (B max ϫ [free sCD14])/(K D ϩ [free sCD14]) using SigmaPlot software (Jandel Scientific, San Rafael, CA).
Cross-linking of 125 I-ASD-sPGN or 125 I-ASD-LPS to CD14 -125 I-ASD (2-p-azido-salicylamido-1,3Ј-dithiopropionate)-derivatized sPGN and LPS were the same as before (36,43) and had specific activities of 6.9 -17 and 2.1-3.6 Ci/g, respectively. The following lines of evidence indicated that the label was indeed bound to sPGN and LPS (and not to some minor otherwise undetectable contaminants or to minor subfractions of PGN and LPS) (36,37): (i) the patterns of the labeled sPGN and LPS on PAGE were unique for each preparation and corresponded to the patterns of biosynthetically labeled sPGN or silver-stained LPS; and (ii) the label was solubilized (converted to low molecular weight migrating on SDS-PAGE with the buffer front) by digestion of 125 I-ASD-sPGN (but not 125 I-ASD-LPS) with muramidase SG, lysozyme, or lysostaphin but not with heparinase, RNase, DNase, or trypsin, or removed by treatment of 125 I-ASD-LPS (but not 125 I-ASD-sPGN) with polymyxin B-agarose or tachyplasin-agarose. Before the cross-linking studies, enzyme-digested 125 I-ASD-sPGN and 125 I-ASD-LPS were dialyzed as described (36,37).
For cross-linking to sCD14, 2 g of sCD14 (or BSA as a control) were incubated with 0.225 g of 125 I-ASD-sPGN or 0.336 g of 125 I-ASD-LPS, with or without 29 g of sPGN or 34 g of ReLPS, in 30 l of Hanks' balanced salt solution with 10 mM HEPES and 0.05% gelatin for 4 h at 37°C in the dark, followed by 8 min exposure to UV light at 22°C (36,44), incubation with 4 g of anti-CD14 63D3 mAb (or anti-BSA 033 mAb) for 18 h at 4°C, and incubation with 20 l of anti-mouse IgGagarose for 4 h at 22°C. The agarose was washed five times with PBS with 0.05% Tween 20 and boiled with sample buffer with 2% SDS and 10% 2-mercaptoethanol, and the supernatant was subjected to SDS-PAGE on 11% gels and autoradiography as described (36,44).
For cross-linking to mCD14, differentiated THP-1 cells were washed four times with Hanks' balanced salt solution with 0.05% gelatin, and 240 ϫ 10 6 cells in 2 ml of Hanks' balanced salt solution with 10 mM HEPES and 0.05% gelatin were incubated with 2.25 g of 125 I-ASD-sPGN for 15 min or 3.36 g of 125 I-ASD-LPS for 5 min at 37°C in the dark (26), then exposed to UV light for 10 min at 4°C (35), washed three times by centrifugation, solubilized at 4°C in 6 ml of solubilizing buffer (6) with 1% Nonidet P-40, and centrifuged at 6000 ϫ g. The supernatant was incubated with 4 g of anti-CD14 63D3 mAb (or control IgG) for 18 h at 4°C, followed by incubation with anti-mouse IgG-agarose, washing, denaturing, reduction, PAGE, and autoradiography as described above for sCD14. The reduction after cross-linking cleaves the cross-linker and eliminates the ligand but retains the label cross-linked to the ligand-binding sites.
TNF-␣ Assay-RAW264.7 cells were cultured and stimulated with the indicated concentrations of sPGN, LTA, or ReLPS, and the concentrations of TNF-␣ in the culture supernatants were determined by the bioassay as described (6).

RESULTS
Binding of CD14 to PGN-Agarose and Determination of Binding Affinity and the Effect of LBP-To test the hypothesis that CD14 binds to PGN, we have developed a binding assay in which we used sPGN covalently bound to agarose and recombinant affinity purified human full-length sCD14, containing at the C terminus five amino acids that constitute protein kinase A phosphorylation site followed by an affinity tag of six histidines, labeled with 32 P to a very high specific activity. To compare CD14 binding of sPGN and LPS, we developed a similar binding assay with ReLPS covalently bound to agarose. Both sPGN-agarose and ReLPS-agarose bound 32 P-sCD14 (Fig.  1). The binding was proportional to the amount of sPGN-or ReLPS-agarose with a very high signal-to-noise ratio (i.e. binding to agarose itself was negligible, Fig. 1A). The binding was dependent on CD14, because similarly labeled 32 P-casein showed a very low binding (Fig. 1B).
Approximately 50% of binding of sCD14 to sPGN was completed in 10 min, and over 85% of binding was completed in 60 min (Fig. 1C). Binding to ReLPS was more rapid, with over 80% of binding completed within 10 min (Fig. 1D). This binding of sCD14 to ReLPS-agarose was much faster than the previously described binding of LPS to CD14 in the absence of LBP (10), which may be due to differences in the conditions of the assay, i.e. interaction of two soluble molecules in the work of Hailman et al. (10) and binding to a solid-phase ligand in our assay. It is possible that the kinetics of CD14 binding to solid-phase ligands is different than to soluble ligands.
Binding of sCD14 to sPGN-agarose was enhanced by LBP; approximately twice as much sCD14 bound to sPGN in the presence than in the absence of LBP (Fig. 1C). LBP diminished the amount of sCD14 bound to ReLPS-agarose (Fig. 1D). Although this effect seems different from the usual enhancing effect of LBP on LPS binding to CD14 (8 -13, 26 -28, 30, 33), it only reflects the amount of sCD14 bound to ReLPS (not the affinity, which is increased by LBP, see below).
Binding of sCD14 to sPGN or ReLPS was specific for both sPGN or LPS and CD14, because it could be competitively inhibited by an excess of either of the soluble ligands, i.e. sCD14 (Fig. 2) or sPGN or LPS ( Fig. 3 and Fig. 6 below). The binding was saturable within the concentration range shown in Fig. 3. In the absence of LBP, the binding of sCD14 to sPGNand ReLPS-agarose could be fitted to a hyperbolic curve and into a single straight line in a Scatchard plot and yielded an apparent K D ϭ 24.9 Ϯ 4.7 nM (B max ϭ 24.3 Ϯ 4.8 ng/ml) for sPGN and K D ϭ 41.2 Ϯ 3.2 nM (B max ϭ 102 Ϯ 2.2 ng/ml) for ReLPS (Fig. 3).
In the presence of LBP, the binding of sCD14 to sPGN-and ReLPS-agarose could be fitted to two hyperbolic curves and into two straight lines in a Scatchard plot and yielded two apparent K D . For sPGN, the presence of LBP did not significantly change the first K D and B max (K D1 ϭ 20.4 Ϯ 3.4 nM, B max1 ϭ 25.9 Ϯ 5.4 ng/ml) but added an additional lower affinity binding constant with a higher B max (K D2 ϭ 71.6 Ϯ 12 nM, B max2 ϭ 59.8 Ϯ 10 ng/ml) (Fig. 3A). For ReLPS, the presence of LBP increased the affinity of binding by yielding two significantly lower K D but lower B max (K D1 ϭ 7.1 Ϯ 2.7 nM, K D2 ϭ 27.3 Ϯ 7.2 nM, and B max1 ϭ 14.1 Ϯ 5.2 ng/ml, B max2 ϭ 28.9 Ϯ 8.3 ng/ml) (Fig. 3B).
Therefore, LBP increased the binding of sCD14 to sPGN by adding another lower affinity binding constant and by increasing the B max . By contrast, LBP increased the affinity of sCD14 binding to LPS by lowering the K D . Lower total amount of binding of sCD14 to LPS in the presence of LBP was due to lower B max for sCD14, which was most likely due to binding of LBP to LPS and competitive inhibition of binding of sCD14 to LPS by LBP (since LPS binds to LBP with a higher affinity than to CD14, Ref. 30). These results demonstrate different mechanisms of enhancement of CD14 binding to sPGN and LPS by LBP.
Binding of Biotinylated-PGN to Solid-phase CD14 -To confirm the binding of PGN to CD14 by another independent method, we have developed an enzyme-linked immunosorbent assay, in which wells of a microtiter plate were sequentially coated with anti-CD14 63D3 mAb and sCD14 and then incubated with biotin-labeled sPGN (or ReLPS). The binding of , and the amounts of 32 P-sCD14 bound to agarose and remaining unbound (free 32 P-sCD14) were measured. Total binding was the amount of radioactivity ( 32 P-sCD14) associated with agarose; nonspecific binding was the amount of radioactivity associated with agarose in the presence of 1 mg/ml of sPGN (A) or LPS (B); specific binding was total binding minus nonspecific binding. The results are means of three to four experiments. The Scatchard plots were fitted using the Cricket Graph software (R 2 ϭ 0.945 Ϯ 0.015, mean Ϯ S.E.); mean apparent dissociation constants Ϯ S.E., calculated using curve fitting to a hyperbolic function, are also shown. sPGN-biotin (or ReLPS-biotin) was detected spectrophotometrically using streptavidin-phosphatase. This assay also demonstrated dose-dependent binding of sPGN to sCD14, but the sPGN binding, in contrast to ReLPS binding, was not enhanced by LBP (Fig. 4A). Similar results were obtained with smooth LPS-biotin, but the absorbance values were 50% lower than with ReLPS-biotin (not shown).
To confirm that biotin was coupled to the representative fraction of PGN, LPS, and ReLPS, biotin-labeled sPGN, LPS, and ReLPS were subjected to SDS-PAGE and blotted onto Immobilon. Probing the blots with peroxidase-labeled streptavidin (Fig. 4B) revealed exactly the same staining patterns as the silver staining or fluoroautoradiography of biosynthetically labeled PGN (Fig. 1 in Ref. 36) or the silver staining of LPS or ReLPS (Fig. 1 in Ref. 37), indicating that biotin was indeed bound to the representative fraction of PGN, LPS, and ReLPS. Digestion of sPGN-biotin with lysozyme, muramidase SG, or lysostaphin (but not with trypsin) completely converted the label into a low molecular weight species that ran off the SDS-polyacrylamide gel with the buffer front (Fig. 4B), which further confirms that biotin was indeed bound to PGN (and not to some minor undetectable contaminant of sPGN). Treatment with the same enzymes had no effect on LPS-biotin and ReLPSbiotin (Fig. 4B).
Binding of sPGN-biotin to CD14 was abolished by digestion of sPGN-biotin with specific PGN-lytic enzymes (lysozyme, muramidase SG, or lysostaphin) but not with trypsin (Fig. 4C), whereas binding of ReLPS-biotin (Fig. 4C) or LPS-biotin (not shown) was not affected by similar treatment with any of these enzymes. These results further confirm that this assay indeed measured binding of PGN to CD14 and not of some minor undetectable non-PGN biotin-labeled contaminant. These results confirm the binding of sPGN to CD14 and confirm the difference in the effect of LBP on the binding of sPGN and LPS to CD14.
Cross-linking of 125 I-ASD-sPGN to CD14 -To confirm further the binding of PGN to sCD14 by yet another method and to test if PGN also binds to membrane CD14 (mCD14) present on the cell surface, we used a photoaffinity cross-linking procedure. Both 125 I-ASD-sPGN and 125 I-ASD-LPS cross-linked to sCD14, detected on autoradiograms of PAGE gels as a doublet (26,30,40), which co-migrated with the standard 32 P-labeled sCD14 (Fig. 5). The photoaffinity labeled sCD14 band was clearly distinct from the 70-kDa photoaffinity labeled albumin band. Cross-linking of both 125 I-ASD-sPGN and 125 I-ASD-LPS to sCD14 was completely inhibited by an excess of underivatized sPGN and partially inhibited by an excess of underivatized ReLPS (Fig. 5). Since these cross-linking studies were done in the absence of serum or LBP, these competitive inhibition results confirm our data ( Fig. 3 and Fig. 6 below) on higher affinity of binding of sPGN than LPS to sCD14 in the absence of LBP.
Digestion of 125 I-ASD-sPGN with PGN-lytic enzymes (lysozyme, muramidase SG, or lysostaphin), but not with other enzymes (heparinase, RNase, and DNase), completely abolished cross-linking of 125 I-ASD-sPGN to sCD14, whereas digestion of 125 I-ASD-LPS with the same enzymes had no effect on its cross-linking to sCD14 (data not shown). These results confirmed that this cross-linking procedure detected binding of Both 125 I-ASD-sPGN and 125 I-ASD-LPS also cross-linked to mCD14 when incubated with vitamin D 3 -differentiated THP-1 cells, detected on autoradiograms of polyacrylamide gels as a characteristic 45-55-kDa doublet (26,30) that immunoprecipitated with anti-CD14 Abs (Fig. 5). These results further confirm that sPGN binds to sCD14 and demonstrate that it also binds to mCD14.
Specificity of CD14 Binding-To determine further the specificity of CD14 binding to sPGN and ReLPS and to define the role of LBP in CD14 binding, we compared the competitive inhibition of binding of 32 P-sCD14 to sPGN-agarose or ReLPSagarose by sPGN, LPS, ReLPS, and lipid A in the presence or absence of LBP (Fig. 6). An excess of sPGN completely or almost completely inhibited the binding of sCD14 to sPGN-or ReLPS-agarose, with an IC 50 ϭ approximately 20 g/ml (160 nM), and this inhibition was not significantly affected by LBP (Fig. 6A). An excess of smooth LPS, ReLPS, and lipid A also inhibited the binding of sCD14 to sPGN-or ReLPS-agarose (Fig. 6, B-D), but complete inhibition was not obtained, most likely due to the aggregation of these preparations at high concentrations and nonspecific binding of these aggregates to the agarose (at 0.1 to 1 mg/ml ReLPS or lipid A, no inhibition or an increase of sCD14 binding was observed, not shown).
LBP greatly enhanced (by 10 to 1000 times) the ability of LPS, ReLPS, and lipid A to inhibit the binding of sCD14 to both sPGN-and ReLPS-agarose (Fig. 6, B-D), and the effect of LBP was the greatest with the most hydrophobic preparations (i.e. ReLPS and lipid A), which tend to aggregate more in solution than smooth LPS. These results are consistent with higher affinity binding of LPS in the presence of LBP and with the function of LBP as a transfer molecule that can disaggregate LPS and transfer single LPS molecules to sCD14 (10,13,33).
To determine which part of PGN binds to CD14, we studied binding of sCD14 to synthetic PGN fragments coupled to agarose. 32 P-sCD14 readily bound to MDP-agarose (Fig. 7) and a disaccharide-dipeptide (GlcNAc-MDP)-agarose (not shown) but did not bind to PGN pentapeptide-agarose (Fig. 7A), indicating that the glycan part of PGN is required for sCD14 binding. Binding of 32 P-sCD14 to MDP-agarose (similarly to the binding to sPGN-agarose, Figs. 7 and 8) was competitively inhibited by sPGN, LPS, and LTA but not by monomeric soluble MDP.
These data further confirm that sCD14 binds to PGN itself (and not to a minor undetectable contaminant of PGN), since it is highly unlikely that synthetic MDP, synthetic GlcNAc-MDP, and natural PGN would all contain the same undetectable contaminant that binds CD14. The lack of inhibition of sCD14 binding to MDP-agarose by monomeric soluble MDP (Fig. 7B) indicates that polymeric, aggregated, or solid-phase bound PGN or MDP is needed for CD14 binding.
To confirm further the above findings, we tested whether a number of low molecular weight PGN fragments, such as PGNderived pentapeptide, muramic acid, MDP, or GlcNAc-MDP, could inhibit binding of sCD14 to PGN-or ReLPS-agarose (Fig.  8A). None of these low molecular weight compounds significantly inhibited binding of sCD14 to PGN or ReLPS (expect for a slight inhibition of binding of sCD14 to ReLPS by GlcNAc-MDP at a high concentration of 1 mg/ml). These results further indicate that polymeric PGN is required for high affinity binding to CD14.
Since it has been suggested recently that CD14 can bind a variety of polymers (15)(16)(17)(18)(19)(20)(21)(22)(23)45), we tested the ability of some of these molecules to inhibit binding of sCD14 to sPGN and ReLPS (Fig. 8A). Non-sulfated non-charged polymers (such as dextran) or sulfated negatively charged molecules (such as dextran sulfate or heparin) did not inhibit or very poorly inhibited binding of sCD14 to sPGN and ReLPS (Fig. 8A). Cell wall ribitol teichoic acid (at 1 mg/ml) also did not inhibit binding of sCD14 to sPGN-agarose or ReLPS-agarose (not shown). By contrast, polyglycerol LTAs from four different bacteria were all very good inhibitors (often better than LPS) of sCD14 binding to sPGN and ReLPS (Fig. 8A). These results confirm the broad but limited specificity of CD14 for a number of polysaccharide polymers.
The ability of sPGN to competitively inhibit binding of sCD14 to sPGN-agarose or ReLPS-agarose was abolished by digestion of sPGN with a PGN-lytic enzyme, lysostaphin, whereas the ability of LPS and LTA to inhibit binding of sCD14 to sPGN-agarose and ReLPS-agarose was not affected by digestion with lysostaphin (Fig. 8B). These results further confirm that PGN itself (and not a minor undetectable non-PGN contaminant) binds to CD14 and also that polymeric PGN is required for CD14 binding.
Regions of CD14 Involved in PGN Binding-Since we have recently shown that the N-terminal 151 amino acids of mCD14 were sufficient for the cellular responsiveness to PGN (15), we next tested if the similar N-terminal sCD14 fragment binds to sPGN. Indeed, the unlabeled 1-152 N-terminal fragment of sCD14 inhibited binding of full-length 32 P-sCD14 to both sPGN and ReLPS as efficiently as the unlabeled full-length sCD14 (Fig. 2). These results indicate that the binding site(s) for both sPGN and ReLPS is located in the N-terminal 152-amino acid region of CD14.
To define further the regions of CD14 that bind sPGN and LPS, we studied the inhibition of binding of sCD14 to sPGN or ReLPS by a number of anti-CD14 mAbs with known specificities for different amino acid sequences of CD14 (11,29,31,35). All antibodies were used at 33 g/ml, which was the highest concentration that did not cause nonspecific inhibition of binding (Fig. 9A). MEM-18, whose epitope is located between amino acids 51 and 64, was the most efficient inhibitor of binding of sCD14 to both sPGN and ReLPS (Fig. 9B). The second most effective antibody was CRIS-6, whose epitope (amino acids 58 -62) is located within the MEM-18 epitope. Several other mAbs (such as 3C10, 60bca, RPA, biG 4, biG 14, and MY-4, whose epitopes are located more N-terminal to the MEM-18 epitope) were also quite effective (although not as effective as MEM-18) in inhibiting sCD14 binding to ReLPS. However, most of these antibodies (except MY-4) were less effective, or completely ineffective, in inhibiting sCD14 binding to sPGN. Combining three of these mAbs (3C10, MY-4, and CRIS-6) caused greater inhibition of binding of sCD14 to ReLPS than each of these mAbs alone, but it did not further decrease the binding of sCD14 to sPGN (beyond the inhibition caused by CRIS-6 alone). Only one mAb (Leu-M3) inhibited sCD14 binding to sPGN but did not inhibit sCD14 binding to ReLPS (Fig. 9B). DISCUSSION Our results demonstrate that PGN directly binds both soluble and membrane CD14 and together with our previous results (15)(16)(17) prove that CD14 acts as a true PGN receptor. Both in the absence and presence of LBP, CD14 binds PGN with a slower kinetics than LPS. CD14 binds PGN with high affinity both in the absence and presence of LBP (K D ϭ 24.9 Ϯ 4.7 and 20.4 Ϯ 3.4 nM, respectively), and this affinity is higher than the affinity of binding of CD14 to LPS in the absence of LBP (K D ϭ 41.2 Ϯ 3.2 nM). LBP increases the binding of CD14 to PGN by adding an additional low affinity binding constant (K D ϭ 71.6 Ϯ 12 nM) and by increasing the B max . By contrast, LBP increases the affinity of CD14 binding to LPS by yielding two significantly lower dissociation constants (K D1 ϭ 7.1 Ϯ 2.  , and LTA (B). A, binding of 32 P-sCD14 to sPGN-agarose, MDPagarose, pentapeptide-agarose, or control agarose was measured as in Fig. 1. The results are means from three experiments; the S.E. were less than 15% and are not shown. B, competitive inhibition of binding of 32 P-sCD14 to MDP-agarose by 0.5 mg/ml of the indicated competitors was measured. The results were calculated as the percent of binding without a competitor and are means Ϯ S.E. from three experiments.
(31.9 nM) for the binding of LPS in the presence of LBP to recombinant CD14 expressed on the cell membrane (28,35) or sCD14 (30).
Our results, therefore, demonstrate different mechanisms of enhancement of CD14 binding by LBP to sPGN and LPS. The increase in the affinity of binding of sCD14 to LPS by LBP is consistent with the previously proposed mechanism (10,13,33), in which LBP acts as a transfer molecule, i.e. binds LPS and transfers it to sCD14. This increased affinity of binding of LPS results in the lowering of the cell-activating concentration of LPS in the presence of LBP (8 -10, 46, 47). By contrast, LBP does not enhance CD14-dependent cell activation by PGN (15,16), because it only increases the low affinity binding of PGN to CD14, which apparently does not result in enhanced cell activation.
Our results demonstrate that the glycan part of PGN is required for CD14 binding and that the minimal structure of PGN that binds CD14 is MDP. However, CD14 binds only to MDP immobilized on agarose, which indicates that polymeric PGN or polymeric-like immobilized MDP is required for the binding. This requirement for polymeric PGN is supported by the loss of CD14-binding capacity of PGN following digestion with PGN-lytic enzymes. These results differ from the previous report showing inhibition of fluorescein isothiocyanate-sPGN binding to cellular CD14 by soluble monomeric MDP (17). The reason for this discrepancy is not clear, but the fluorescein isothiocyanate-sPGN binding assay does not stringently show binding to CD14, because the target is the whole cell and it is only presumed that this assay detects binding of fluorescein isothiocyanate-sPGN to CD14 (because this binding is inhibited by anti-CD14 mAbs) (17). Most recent results from Dr. Ulmer's laboratory confirm our current findings and indicate that monomeric soluble MDP does not bind to sCD14 and does not inhibit binding of sPGN or LPS to sCD14 in a gel shift assay. 2 Our results of competitive inhibition experiments support the notion (15-24) that CD14 binds other polymeric macrophage activators of microbial origin (besides LPS and PGN), since LTAs from four bacterial species were as active as LPS and sPGN in inhibiting CD14 binding to LPS and PGN. Contrary to the previous report (45), however, other polymers, such as dextran sulfate or heparin and also cell wall ribitol teichoic acid, do not inhibit or only marginally inhibit binding of sCD14 to PGN or LPS.
Since it was recently suggested (48) that a PGN contaminant might have been responsible for PGN-induced CD14-dependent cell activation, we have carefully examined if PGN itself bound to CD14. The following lines of evidence indicate that PGN itself, and not an undetectable minor contaminant, binds to CD14. (i) CD14 binds to natural sPGN and also two synthetic PGN fragments, MDP and a disaccharide GlcNAc-MDP, immobilized to agarose, and this binding is inhibited by PGN, LPS, and LTA. It is highly unlikely that synthetic MDP, synthetic 2 A. J. Ulmer, unpublished observations. GlcNAc-MDP, and natural PGN would contain the same undetectable contaminant that binds CD14. (ii) Binding of PGN to CD14 and the inhibitory capacity of PGN for CD14 binding to PGN and LPS are lost when the PGN preparations are digested with PGN-lytic enzymes (lysostaphin, muramidase SG, and lysozyme) but are not affected by digestion with other enzymes (heparinase, RNase, DNase, and trypsin). Gel electrophoresis and blotting results also show that the labels (ASD and biotin) are bound to representative molecular species of PGN and that the PGN-lytic enzymes also solubilize ASD-or biotin-labeled PGN. By contrast, the binding of LPS to CD14, the inhibitory capacity of LPS or LTA, and the labeling of LPS with ASD and biotin are not diminished or destroyed by the same treatment with PGN-lytic enzymes. (iii) In a gel-shift assay, which measures binding of unlabeled sPGN to unlabeled sCD14, the stoichiometry of the binding of sPGN to sCD14 was approximately 1:1 (Fig. 9 in Ref. 17). (iv) Binding of sCD14 to immobilized sPGN was inhibited by sPGN with an IC 50 ϭ 20 g/ml (160 nM) (Fig. 6A), which is of similar magnitude as the two K D values that we have obtained for the binding of sCD14 to immobilized PGN (20 -72 nM, Fig. 3A). If a minor contaminant, undetectable by chemical analysis, was responsible for CD14 binding, the IC 50 of sPGN in this experiment would have been at least 2 orders of magnitude larger. (v) We also previously excluded the possibility that sPGN, LPS, and LTA form hetero-aggregates with each other (Fig. 8 in Ref. 37). Such an aggregation could have been responsible for the inhibition of binding of sPGN by LTA and LPS, and of LPS by sPGN and LTA. Moreover, if a contaminant was responsible for the binding of PGN and LTA to CD14, it would be highly unlikely that the same contaminant was present in insoluble PGN, sPGN, and LTA from several bacterial species, given entirely different methods of purification of LTA, insoluble PGN, and sPGN. (vi) The sPGN preparation used in this study was of the highest purity available (see "Experimental Procedures").
We do not agree with the recent conclusion of Kusunoki and Wright (48) that a minor undetectable contaminant of insoluble PGN was responsible for its cell-activating effect for the following reasons. (i) As indicated in Ref. 48, the activity of PGN was abolished by treatment with SDS. We have treated the same insoluble PGN preparation with SDS (5%, 80°C, 30 min), and after removal of SDS by 10 washes with PBS, the SDS-treated PGN was as active or more active than the untreated PGN in induction of TNF-␣ production in RAW264.7 cells. 3 Most likely in Ref. 48, SDS was not adequately removed from the insoluble PGN, and since SDS is toxic to cells, the activity was lost. (ii) As indicated in Ref. 48, the activity of PGN was abolished by hot phenol/water extraction. In this procedure, after the extraction, the phenol/water/PGN mixture was centrifuged, and the insoluble PGN, sedimented in the bottom of the phenol layer, was not active (whereas the "active component" was supposedly extracted, similar to LPS or LTA, into the water phase). However, the procedure used in Ref. 48 did not include removal of phenol (which is toxic to cells) from insoluble PGN sedimented in the phenol phase, and this most likely accounted for the lack of activity of treated PGN. We have repeated this phenol/water extraction (45% phenol, 60°C, 60 min), but then we removed residual phenol from the PGN by ethyl ether extraction, and the recovered PGN had full activity (for induction of TNF-␣ production in RAW264.7 cells), equal to the activity of PGN before the extraction. Moreover, the water fraction obtained from the hot phenol/water extraction of this PGN had no activity (for induction of TNF-␣ production), indicating that no active component was removed from PGN by phenol/water extraction. 3 In Ref. 48, the authors did not test the activity of the water fraction, which should have contained the "active fraction" if it had been removed from the PGN preparation. (iii) In Ref. 48, when insoluble PGN was incubated with sCD14 and then spun down, the remaining supernatant containing sCD14 was more active in activating neutrophils than the sedimented PGN or untreated PGN. The authors concluded that the putative active contaminant from PGN became bound to sCD14 and remained in the supernatant in an active sCD14-bound form. This experiment, however, is inconclusive, because it does not show what was active in the supernatant, and there are several alternative explanations of these results, such as release of soluble PGN fragments aggregated with the insoluble PGN or binding of a large portion of sCD14 to insoluble PGN, which would make the sCD14 remaining in the supernatant more active (because high concentrations of sCD14 inhibit cell activation and low concentrations enhance cell activation). Moreover, sedimented PGN was equally active as the untreated PGN, which indicates that no significant activity was removed from the PGN by sCD14. Another unexplained aspect of this experiment is the requirement for an overnight incubation of PGN with sCD14 at 37°C, which is inconsistent with much faster binding of sCD14 to PGN shown here by us. Another possibility includes contamination of the PGN-sCD14 mix that would produce enough endotoxin during an overnight incubation to activate cells in conjunction with sCD14. None of these possibilities were formally excluded (48). (iv) In Ref. 48, only insoluble PGN (isolated from the cell walls by enzyme digestions and trichloroacetic acid extraction) was used, and the affinity purified soluble PGN (used in all experiments in this paper) was not tested (48). (v) In our experiments, the cellactivating capacity (for induction of TNF-␣ production in RAW264.7 cells) of sPGN is lost following digestion with PGNlytic enzymes and dialysis (6). All these results indicate that PGN itself activates cells.
The difference in the affinity of binding of CD14 to LPS and PGN in the presence of LBP (K D , 7.1 nM versus 20.4 nM) does not fully explain the 4 log difference (in g/ml, or 200 times difference on per mol basis) between the effective macrophageactivating concentrations of ReLPS and sPGN (Fig. 10). This difference cannot be fully attributed to LBP, because even in the absence of serum (and serum-derived LBP), LPS is a more 3 R. Dziarski, unpublished observations. effective macrophage activator than sPGN (6). Therefore, one possibility is that this difference is related to a much greater ability of CD14 to transfer LPS (than sPGN) to another, as yet unidentified (13,24), cell-activating molecule. LTA, which has an intermediate ability to activate macrophages (Fig. 10), would then also have an intermediate ability to be transferred from CD14 to the putative cell-activating molecule.
Our results demonstrate that less than half of the CD14 molecule (N-terminal 152 amino acids) is sufficient for PGN binding and suggest that the sequences that are most critical for CD14 binding to both PGN and LPS are located within the N-terminal region between amino acids 51-64. This region appears to be necessary for both PGN and LPS binding, because MEM-18, an anti-CD14 mAb specific for this region, inhibited CD14 binding to ReLPS by over 95% and to PGN by over 80%. It is also possible, however, that the effect of MEM-18 could be through a conformational change of CD14, but this would also indicate similar primary binding sites for both PGN and ReLPS (since the same conformational change would abolish binding for both ligands). This region may not be sufficient for binding of LPS or PGN, because mAbs specific to other regions of CD14 also partially inhibited binding of LPS and PGN. Of note, these other sequences that contribute to PGN and LPS binding are different, because there are several mAbs that inhibit LPS binding but not PGN binding (3C10, specific for amino acids 7-14, and 60bca and RPA specific for amino acids 34 -38) and, significantly, one mAb that inhibits PGN binding but not LPS binding (Leu-M3, specific for amino acids 135-146). But again, it is also possible that these effects might have been due to conformational changes induced by these mAbs. In summary, it appears that PGN and ReLPS bind to conformational rather than linear CD14 epitopes that are partially identical (amino acids 51-64) and partially different.
This conclusion is consistent with the results of Juan et al. (29) indicating that amino acids 57-64 are the primary LPSbinding site, but it does not fully agree with the conclusion of Stelter et al. (35) that amino acids 39 -44 are the main LPSbinding site on CD14 and with the conclusion of Shapiro et al. (34) that Glu-37 was necessary for LPS binding. However, Stelter et al. (35) also noticed full inhibition of LPS binding with MEM-18 mAbs and concluded that the LPS-binding site is a conformational site that includes amino acids 9 -13, 39 -44, and 51-63.
We have noted some differences in the ability of anti-CD14 mAbs to inhibit binding of sCD14 to ReLPS, compared with the previously reported ability of the same mAbs to inhibit binding of LPS to membrane or soluble CD14. In particular, it was reported that mAbs 3C10 (31), RPA (11), and biG 4 (35) did not inhibit LPS binding to sCD14 (3C10) or mCD14 (RPA, biG 4), whereas we have observed significant inhibition of sCD14 binding to ReLPS by all three mAbs. Most likely these differences are due to the differences in the assay systems, the use of soluble versus membrane CD14, the presence or absence of LBP, or the use of ReLPS versus smooth LPS, which was shown to affect the interaction of LPS with CD14, LBP, and anti-CD14 mAbs (47). Indeed, other studies (26) showed inhibition of LPS cross-linking to mCD14 by 3C10 mAb.
In general, our binding results are consistent with our previous activation results showing that similar but not identical sequences in mCD14 were critical for the responsiveness of CD14-transfected cells to PGN and LPS (15) and that MEM-18 is the mAb that is most active in both inhibiting PGN and LPS binding to CD14 and in inhibiting PGN-and LPS-induced CD14-mediated cell activation (16). 3 Detailed comparison of the sequences that are most crucial for the binding and cell activation is difficult at the moment, because the activation analysis was done using CD14 deletion mutants (15) and the current binding studies were done using mAbs, whose epitopes are somewhat different than the epitopes deleted in those mutants.
Our conclusion that PGN and LPS bind to conformational sites can accommodate both the similarities (␤1-4-or ␤1-6linked N-acetylated glucosamine residues with closely located carbonyl groups) and differences (presence of fatty acids and phosphate groups in ReLPS and presence of amino acids and a long glycan chain in PGN) in the structures of PGN and ReLPS (37). This conclusion is also consistent with the hypothesis that CD14 binds to the glycan part of LPS and other ligands (49). Our data, showing that agarose-immobilized MDP (but not PGN derived pentapeptide) binds CD14, further support this hypothesis.
In summary, our results support the hypothesis that CD14 does have a defined specificity and that it is not merely a nonspecific transfer molecule. This specificity of CD14 is different from the specificity of other molecules that bind multiple ligands, such as albumin or the scavenger receptor (37,44). These results also support the hypothesis that CD14 functions as a "pattern recognition receptor" and suggest that recognition of different patterns is encoded in somewhat different regions of CD14. However, our results also do not exclude the possibility that additional specificity in cell activation may be determined by other events occurring after the CD14 binding step.