MD-1 and MD-2 are secretory glycoproteins that exist on the cell surface in complexes with transmembrane proteins. MD-1 is anchored by radioprotective 105 (RP105), and MD-2 is associated with TLR4. In vivo studies revealed that MD-1 and MD-2 have roles in responses to LPS. Although the direct binding function of MD-2 to LPS has been observed, the physiological function of MD-1 remains unknown. In this study, we compared the LPS-binding functions of MD-1 and MD-2. LPS binding to cell surface complexes was detected for cells transfected with TLR4/MD-2. In contrast, binding was not observed for RP105/MD-1-transfected cells. When rMD-2 protein was expressed in Escherichia coli, it was purified in complexes containing LPS. In contrast, preparations of MD-1 did not contain LPS. When rMD-2 protein was prepared in a mutant strain lacking the lpxM gene, LPS binding disappeared. Therefore, the secondary myristoyl chain attached to the (R)-3-hydroxymyristoyl chain added by LpxM is required for LPS recognition by MD-2, under these conditions. An amphipathic cluster composed of basic and hydrophobic residues in MD-2 has been suggested to be the LPS-binding site. We specifically focused on two Phe residues (119 and 121), which can associate with fatty acids. A mutation at Phe191 or Phe121 strongly reduced binding activity, and a double mutation at these residues prevented any binding from occurring. The Phe residues are present in MD-2 and absent in MD-1. Therefore, the LPS recognition mechanism by RP105/MD-1 is distinct from that of TLR4/MD-2.
Innate immunity provides early defenses against infection, before the development of an adaptive immune response ( 1). TLRs play a critical role in the recognition of various pathogen-associated molecular patterns in LPS, peptidoglycan, unmethylated CpG nucleotides, double-strand viral RNAs, etc. ( 2). Ten molecules have been identified as members of the TLR family (TLR1 to -10) ( 3), and each TLR recognizes a restricted pathogen-associated molecular pattern. TLR4 is involved in LPS recognition ( 4, 5). However, TLR4 recognition of LPS is distinct from that of other TLRs. Genetic ( 5, 6, 7) and biochemical experiments ( 8) indicated that TLR4 is the signal-transducing component of the LPS receptor. Despite its requirement for LPS signaling, MD-2 was identified as a TLR-4 binding protein that is necessary for TLR4 function. MD-2 is required for the cell surface expression of TLR4 ( 9), and also for LPS recognition ( 9). Moreover, MD-2 directly binds to LPS ( 10, 11, 12). MD-2 is a small soluble glycoprotein that is anchored to the cell surface via TLR4. TLRs are transmembrane proteins that contain a Toll/IL-1R domain in their cytoplasmic region. Signaling by LPS binding to MD-2 is transduced through an associated TLR4. The role of MD-2 in LPS recognition in vivo has been examined using gene-disrupted mice ( 9). MD-2-deficient mice did not respond to LPS. They were resistant to endotoxin shock induced by LPS, and were susceptible to infection by the Gram-negative bacterium Salmonella typhimurium ( 9). These phenotypes in MD-2-deficient mice are similar to those of TLR4-deficient mice.
Originally, MD-2 was identified as a homolog of MD-1 ( 8). MD-1 associates with radioprotective 105 (RP105),3 which is a B cell-specific cell surface transmembrane protein ( 13, 14, 15). The extracellular domain of RP105 contains multiple leucine-rich repeats that occur as TLRs. Gene disruption of RP105 in mice results in a decrease in immune response against LPS ( 16). The same phenotype was observed in mice lacking the MD-1 gene ( 17). MD-1 shares several structural features in common with MD-2. One conserved function is that they are required for the cell surface expression of TLR4 and RP105, respectively ( 9, 17). To determine the existence of any other functional similarities, we compared the binding activities of these molecules to LPS. Direct binding of LPS to MD-2 has been reported ( 10), and an amphipathic structure present in MD-2 was proposed to be the LPS binding site ( 11). However, this structure is unique to MD-2. We found that MD-2 and MD-1 have different functional activities for LPS recognition.
Materials and Methods
Full-length human TLR4 cDNA was kindly provided by Dr. R. Medzhitov (Yale University, New Haven, CT). The coding region of TLR4 and MD-2 were cloned into the expression vector pCAGGS (provided by Dr. J. Miyazaki (Osaka University, Osaka, Japan); Ref. 19). The cDNA fragments of human and mouse MD-1 and human MD-2 were amplified by the PCR, using the following primers: P1, 5′-cgcggatccgGGAGGCGGCGGTGGGAAAGC-3′; P2, 5′-cgcgcggccgcCTATTATCAGGAGCACATGATAGTAGCATTGG-3′; P3, 5′-cgcggatccgGACCATGGCAGCGAAAATGGTTGG-3′; P4, 5′-cgcgcggccgcctattaTCAGGAGGAGGTGACAGTGGCATTGGC-3′; P5, 5′-cgcgcggccgcgGAAGCTCAGAAGCAGTATTGGGTC-3′; and P6, 5′-cgcctcgagtcattactaATTTGAATTAGGTTGGTGTAGG-3′. P1 and P2 were used to amplify mouse MD-1; P3 and P4 were used for human MD-1; and P5 and P6 were used for human MD-2. The mouse and human MD-1 PCR products were digested with BamHI and NotI and cloned into a pET-32b vector, which possesses 6 × His and thioredoxin (Trx) tags (Novagen). The human MD-2 PCR fragment was digested with NotI and XhoI, and cloned into the same expression vector. The identities of all constructs were confirmed by DNA sequencing using an ALF-express sequencer (Amersham Biosciences).
The IL-3-dependent murine cell line Ba/F3 ( 20) and subsequently derived stable transfectant cell lines were maintained in RPMI 1640 medium supplemented with 10% FBS, 100 U/ml mouse IL-3, and 50 μM 2-ME. Ba/F3 transfectants expressing RP105 and MD-1 (BaHRPMD) were gifts from Dr. K. Miyake (Tokyo University, Tokyo, Japan) ( 18). We established a cell line from Ba/F3 that expressed human TLR4 and MD-2. The pEF-BOS expression vector ( 21) was used for TLR4, and pCAGGS was used for MD-2 tagged with FLAG and His epitopes at the C terminus.
Mutant human MD-2 constructs were prepared using a site-directed mutagenesis system (TaKaRa) according to the manufacturer’s protocol. The mutation primers were as follows: 5′-CAACAATATCAGCCTCCTTCAAG-3′ (for F119A), 5′-CATTCTCCGCCAAGGGAAT-3′ (F121A), 5′-CAACAATATCAGCCTCCGCCAAGGG-3′ (F119A and F121A), and 5′-GGGAAAATACAAAGCTGTTGTTG-3′ (C133A). The sequences were confirmed by DNA sequencing.
Transfectants cells were stained with anti-RP105, anti-TLR4, or anti-FLAG M2 mAb followed by FITC-conjugated anti-mouse IgG (Southern Biotechnology Associates). Staining was detected using a flow cytometer (FACScan; BD Biosciences), and analyzed using the WinMDI program (J. Trotter (The Scripps Research Institute, La Jolla, CA)).
LPS pull-down assay
LPS (Escherichia coli 0111:B4) was purchased from Sigma-Aldrich and purified by phenol extraction as previously described ( 22). Cells (108/10 ml of medium) were incubated with 5 μg/ml LPS at 37°C for 30 min. After washing, cells were lysed at 4°C in 10 ml of lysis buffer, PBS containing 1 mM PMSF and 1% Triton X-100. Bound LPS to the cell surface MD proteins with the FLAG epitope tag was precipitated by 25 μl of anti-FLAG M2 agarose gel (Sigma-Aldrich). After washing, gels were boiled in Laemmli’s sample buffer, and the eluted materials were subjected to SDS-PAGE. LPS was detected by Western blotting using the anti-LPS WN1 222-5 mAb as described ( 23).
Monomeric LPS:soluble CD14 (sCD14) complex was prepared by incubation with plasma and ultrafiltration as described previously by others ( 24). LPS from Origami B (DE3) was extracted from acetone-dried bacteria using a modified phenol-water technique ( 25). Purified LPS (1 mg) was incubated at 37°C for 1 h with RPMI 1640 containing 20% human plasma, and subjected to ultrafiltration using Amicon Ultra-15 centrifugal filter device with 100-kDa cutoff membrane (Millipore). The fraction of LPS across the ultrafiltration was estimated by Western blotting comparing with known quantities of LPS. Approximately 15 μg of LPS passed the membrane. Cells were incubated with the fraction at a final concentration of 0.5 μg/ml LPS, and LPS-pull down assay was conducted as described above.
Expression and purification of MD-1 and MD-2 from E. coli
Origami B (DE3; Novagen) or BL21 (DE3) lpxM− ( 26) was transformed with the plasmids (see above). Transformed bacteria were grown at 37°C until an OD (A600) of 0.7 was achieved. They were then cultured overnight at 25°C in the presence of 1 mM isopropyl-β-d-thiogalactopyranoside. The culture fluid (1.2 liters) was collected and resuspended in 180 ml of lysis buffer (20 mM Tris (pH 8.0), 50 mM NaCl, 5 mM 2-ME, 1 mM PMSF, 20 μg/ml DNase, and 100 μg/ml lysozyme) and incubated at 37°C for 15 min. Extraction was performed by sonication, and the supernatants were applied on a nickel adsorbed Chelating Sepharose column (3 × 8 cm; Amersham Biosciences). The column was extensively washed with a buffer (50 mM phosphate (pH 7.0), 0.3 M NaCl, 5 mM 2-ME, 0.1 mM PMSF, 0.5% Triton X-100, and 100 mM imidazole), and the adsorbed proteins were eluted with elution buffer (50 mM phosphate buffer (pH 7.0), 0.3 M NaCl, 5 mM 2-ME, 0.1 mM PMSF, 0.5% Triton X-100, and 500 mM imidazole). They were dialyzed against a buffer (50 mM NaCl, 5 mM EDTA, and 5 mM 2-ME). The protein concentration was estimated by protein assay (Bio-Rad), and the purity was estimated by SDS-PAGE and Coomassie brilliant blue (CBB) staining.
LPS binds to cell surface TLR4/MD-2 but not to RP105/MD-1
We first examined LPS binding to the cell surface complexes using two cell lines established from Ba/F3. One was positive for RP105/MD-1, and the other was positive for TLR4/MD-2. Comparable expression of each complex was seen by FACS analysis (Fig. 1,A). After incubating the cells with LPS followed by washing, bound LPS was analyzed by immunoblotting using anti-LPS WN1 222-5 mAb. LPS from a smooth strain of E. coli 0111:B4, which has O-Ag repeat, was used in this experiment. A significant amount of LPS was detected as a ladder-like pattern in the TLR4/MD-2-positive cells, when 5 μg/ml LPS was added. In contrast, LPS was not present in the cells positive for RP105/MD-1 (Fig. 1,B). The serum components, sCD14 and LPS binding protein, contribute to the efficient binding of LPS to the TLR4/MD-2 complex ( 12, 27). Therefore, we also examined binding of plasma-pretreated LPS. We used LPS isolated from Origami B (rough strain), which lacks O-Ag, in this experiment. LPS was incubated with plasma, and monomeric LPS:sCD14 complex was collected by ultrafiltration (Fig. 1,C). The binding LPS lacking O-Ag also bound to TLR4/MD-2 but not RP105/MD-1 (Fig. 1 D). In this case, lower concentration (0.5 μg/ml) of LPS was required for the binding. Complex formation with sCD14 appeared to enable effective transfer of LPS to TLR4/MD-2.
LPS binds to MD-2 but not to MD-1
Recombinant MD-2 proteins produced using a baculovirus or mammalian expression system have been shown to bind to LPS ( 10, 27). We expressed recombinant MD-1 and MD-2 in E. coli as Trx-fusion proteins that carry a His tag. Each fusion protein and a recombinant Trx control protein were purified from bacterial lysate using a nickel column. The identities of the purified proteins were then verified by SDS-PAGE (Fig. 2,B). We detected a considerable amount of LPS in the human MD-2 fusion protein preparations (Fig. 2,A). LPS exclusively bound to MD-2, because the Trx control preparation without MD-2 did not exhibit LPS bands in a Western blot analysis (Fig. 3, lane 3). In contrast, specific LPS binding was not observed for human MD-1 or mouse MD-1 (Fig. 2 A).
Penta-acylated LPS produced in lpxM− strain does not bind to MD-2
LPS is composed of an O-Ag repeat, a polysaccharide core, and lipid A ( 28). Lipid A is the component responsible for the TLR4/MD-2 recognition. The fatty acids contents of lipid A vary from species to species. In the case of E. coli, it possesses six acyl chains. The lipid A precursor in which four chains are present is known as lipid IVa ( 29, 30, 31). Further incorporation of single myristate and laurate moieties in the final step of biosynthesis occurs in E. coli. Human TLR4/MD-2 complexes do not respond to LPS produced in E. coli that possess a disrupted lpxM gene ( 26). The lpxM gene product promotes acylation of the (R)-3-hydroxymyristoyl chain located at the 3′ position of the glucosamine disaccharide backbone leading to addition of the secondary myristoyl chain. As a result, E. coli lacking the functional lpxM gene produces LPS with only five fatty acid chains (penta-acylated LPS) ( 32). A mutant MD-2 protein, which substituted Cys133 to Ala, was prepared in lpxM− or control Origami B strains. The mutant protein demonstrated comparable binding activity to the wild-type hexa-acylated LPS (see Fig. 5). In contrast, the binding disappeared when it was prepared in lpxM mutant strain (Fig. 3). This result was not due to a lack of reactivity by the Ab against the variant LPS, because WN1 222-5 Ab recognizes the core common to LPS ( 33), and reacted to the penta-acylated LPS equally well compared with that of wild-type LPS (data not shown). Therefore, MD-2 could not bind to the mutant penta-acylated LPS.
Phe119 and Phe121 are conserved in MD-2 but not in MD-1, and are required for LPS binding
Because fine structural differences in fatty acid chains are recognized by MD-2, we especially focused on the hydrophobic residues in the amphipathic structure (Fig. 4). Two Phe residues (Phe119 and Phe121) are conserved in human and mouse MD-2. A single substitution of Ala at Phe119 significantly reduced the LPS binding activity of MD-2 (Fig. 5,A). A stronger inhibitory effect was observed for a mutation at Phe121. Only a trace amount of LPS was detected in preparations of these mutant MD-2 proteins. Binding activity was completely abolished when both Phe residues were substituted with Ala (Fig. 5 A). In contrast, a mutation at Cys133 as a control showed no effect against LPS binding. Therefore, Phe residues 119 and 121 are important for LPS-binding function by MD-2.
We detected specific binding of LPS to rMD-2 proteins expressed in E. coli; however, the binding to MD-1 was not detectable. Both MD-1 and MD-2 possess two potential N-glycosylation sites, but the sugar chains appear to be nonessential for LPS-binding function. In a previous study, substitution of Ala for the Asn residue at position 26 at the first glycosylation site of MD-2 did not affect its function, and substitution at the second site, Asn114, only caused a minor modification. Considerable reduction in LPS signaling was observed for dual mutations at the Asn residues, but LPS binding was not affected ( 34). Recently, Visintin et al. (27) reported that recombinant soluble MD-1 expressed in mammalian system that contained oligosaccharide chains also did not bind to LPS. LPS also did not bind to RP105/MD-1 complex expressed in mammalian cells (Fig. 1). These results indicate that the ability of LPS to bind to MD-1 is not comparable with that to MD-2.
The secondary myristoyl chain attached to the (R)-3-hydroxymyristoyl chain at the 3′ position of the glucosamine backbone in LPS added by LpxM is required for binding to MD-2 (Fig. 3). The secondary acyl chains have been demonstrated to be involved in the LPS recognition mechanism by the TLR4/MD-2 complex ( 35). Especially, the penta-acylated E. coli LPS produced by LpxM mutant lost the reactivity almost completely ( 26). This may be explained by the loss of the binding activity for the penta-acylated LPS to MD-2. Like the other LPS-binding proteins, basic amino acids of MD-2 (Lys128 and Lys132) have been shown to be involved in the binding function ( 27). However, affinity and stability of MD-2 for the binding were not comparable with those of the other LPS-binding proteins ( 12, 23). In addition to the charge interaction, the other structures may be required for the high-affinity binding of MD-2. In fact, Phe119 and Phe121 in MD-2 were also required for the binding (Fig. 5). The phenyl side chains may be required for interaction with the hydrophobic alkyl chain present in the secondary myristoyl chain. Neither the basic residues nor hydrophobic residues exist in the corresponding region of MD-1 (Fig. 4). These results obtained with mutant proteins expressed in E. coli should be confirmed in a cellular assay system. The other LPS-binding proteins such as CD14 and LPS binding protein may affect the binding specificity of MD-2.
Although our binding assay system allowed us to observe the specific binding of LPS to MD-2, MD-1 binding activity was not detected. The structural features of MD-2 required for LPS binding are absent in MD-1. Our present results indicate that MD-1 is not an LPS-binding protein. However, it is clear from the gene-knockout mouse experiments that RP105/MD-1 contributes to LPS recognition ( 16, 17). We speculate that the signal(s) transduced by an unknown ligand through RP105/MD-1 may somehow acts collaboratively on the LPS-activation pathway through TLR4/MD-2. Further studies will reveal the physiological ligand(s) for RP105/MD-1 and the mechanisms for the collaborative activities of RP105/MD-1 in the LPS response.
We thank Drs. K. Miyake (Tokyo University), K. Hatakeyama (University of Pittsburgh, Pittsburgh, PA), and K. Watanabe (Saga University) for helpful discussion.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by grants from the Ministry of Education, Science, Sports, and Culture of Japan (to K.F., N.T., and M.K.), and from the Takeda Science Foundation (to K.F.).
Abbreviations used in this paper: RP105, radioprotective 105; Trx, thioredoxin; sCD14, soluble CD14; CBB, Coomassie brilliant blue.