Portal de Periódicos da CAPES

Lipopolysaccharide (LPS) Neutralizing Peptides Reveal a Lipid A Binding Site of LPS Binding Protein

1995; Elsevier BV; Volume: 270; Issue: 30 Linguagem: Inglês

10.1074/jbc.270.30.17934

1083-351X

Alexander H. Taylor, George A. Heavner, Mark Nedelman, David Sherris, Eva Brunt, David Knight, John Ghrayeb,

Immune Cell Function and Interaction

Endotoxic shock follows a cascade of events initiated by release of lipopolysaccharide during infection with Gram-negative organisms. Two overlapping 15-mer peptides were identified, corresponding to residues 91-108 of human lipopolysaccharide binding protein that specifically bound the lipid A moiety of lipopolysaccharide with high affinity. The peptides inhibited binding of lipopolysaccharide to lipopolysaccharide binding protein, inhibited the chromogenic Limulus amebocyte lysate reaction, and blocked release of tumor necrosis factor α following lipopolysaccharide challenge both in vitro and in vivo. These results suggest lipopolysaccharide binding protein residues 91-108 form at least part of the lipopolysaccharide binding site. Moreover, derivatives of lipopolysaccharide binding protein residues 91-108 might modulate lipopolysaccharide toxicity in the clinical setting. Endotoxic shock follows a cascade of events initiated by release of lipopolysaccharide during infection with Gram-negative organisms. Two overlapping 15-mer peptides were identified, corresponding to residues 91-108 of human lipopolysaccharide binding protein that specifically bound the lipid A moiety of lipopolysaccharide with high affinity. The peptides inhibited binding of lipopolysaccharide to lipopolysaccharide binding protein, inhibited the chromogenic Limulus amebocyte lysate reaction, and blocked release of tumor necrosis factor α following lipopolysaccharide challenge both in vitro and in vivo. These results suggest lipopolysaccharide binding protein residues 91-108 form at least part of the lipopolysaccharide binding site. Moreover, derivatives of lipopolysaccharide binding protein residues 91-108 might modulate lipopolysaccharide toxicity in the clinical setting. Infection with Gram-negative organisms results in the specific activation of mononuclear phagocytes by the bacterial cell wall constituent lipopolysaccharide (LPS)1 1The abbreviations used are: LPSlipopolysaccharideLBPlipopolysaccharide binding proteinLBP-Bbiotinylated LBPELISAenzyme-linked immunosorbent assayLALLimulus amebocyte lysate (assay)TNFαtumor necrosis factor αPBMCperipheral blood mononuclear cellsBPIbactericidal permeability increasing proteinHPLChigh pressure liquid chromatographyPBSphosphate-buffered saline. (reviewed in (1Tobias P.S. Mathison J. Mintz D. Lee J.-D. Kravchenko K. Pugin J. Ulevitch R.J. Am. J. Respir. Cell Mol. Biol. 1992; 7: 239-245Google Scholar)). Overstimulation with LPS, a frequent consequence of Gram-negative sepsis, can result in systemic flooding with potent proinflammatory cytokines that include interleukin-1, interleukin-6, and tumor necrosis factor α (TNFα). This contributes directly to the development of endotoxic shock, multiple organ failure, disseminated intravascular coagulopathy, and acute respiratory distress syndrome; these are collectively referred to as Gram-negative sepsis syndrome. More than 120,000 cases of Gram-negative sepsis syndrome occur in the United States annually(2Ziegler E.J. et al.N. Engl. J. Med. 1991; 324: 429-436Google Scholar). lipopolysaccharide lipopolysaccharide binding protein biotinylated LBP enzyme-linked immunosorbent assay Limulus amebocyte lysate (assay) tumor necrosis factor α peripheral blood mononuclear cells bactericidal permeability increasing protein high pressure liquid chromatography phosphate-buffered saline. LPS stimulation of mononuclear cells is driven by the cellular LPS receptor CD14 (3Wright S.D. Ramos R.A. Tobias P.S. Ulevitch R.J. Mathison J.C. Science. 1991; 249: 1431-1433Google Scholar, 4Hailman E. Lichenstein H.S. Wurfel M.M. Miller D.S. Johnson D.A. Kelley M. Busse L.A. Zukowski M.M. Wright S.D. J. Exp. Med. 1994; 179: 269-277Google Scholar) and is potentiated as much as 1000-fold by prior complexing of LPS with the serum protein LBP(5Mathison J.C. Tobias P.S. Wolfson E. Ulevitch R.J. J. Immunol. 1992; 149: 200-206Google Scholar). Because serum LPS concentrations found in the clinical setting are very low(3Wright S.D. Ramos R.A. Tobias P.S. Ulevitch R.J. Mathison J.C. Science. 1991; 249: 1431-1433Google Scholar), it appears that LPS toxicity largely depends on the presence of LBP, a hypothesis that has been borne out in in vivo models of LPS toxicity(6Gallay P. Heumann D. LeRoy D. Barras C. Glauser M.P. Proc. Natl. Acad. Sci. U. S. A. 1993; 90: 9935-9938Google Scholar). Thus, we were interested in better understanding the different structure-function relationships underlying the formation of specificity of LBP for LPS. Here, we report the identification of an LPS binding site of LBP, determined by scanning the entire primary structure of LBP for the presence of small LPS binding domains. Only one domain, corresponding to LBP-(91-108), was found to bind LPS tightly. This region most likely comprises at least part of the LPS binding site of LBP. An Advanced ChemTech model 350 peptide synthesizer was used for the simultaneous synthesis of up to 96 peptides. The double coupling FMOC (N-(9-fluorenyl)methoxycarbonyl) protocol was followed, and synthesis products were worked up according to the manufacturer's specification. Crude nested peptides were screened for LPS blocking activity as described below. Preparative scale peptide syntheses were performed essentially as previously described (7Heavner G.A. Falcone M. Kruszynski M. Epps L. Mervic M. Riexinger D. McEver R.P. Int. J. Pept. Protein Res. 1993; 42: 484-489Google Scholar) on 4-methyl-benzhydrylamine resins with an ABI 431A synthesizer using version 1.12 of the standard Boc software. Following HPLC purification, all peptides had the correct amino acid analysis and molecular ion (fast atom bombardment mass spectroscopy or plasma desorption mass spectroscopy) and were greater than 95% pure as determined by two different HPLC gradients. Peptide sequences are given in the one letter code, with the following modifications: Ac- denotes N-acetylated derivatives, -NH2 denotes COOH-terminal amidated derivatives, and d- denotes D-isomer substitution of the amino-terminal residue only. The LBP-immunoglobulin fusion protein was comprised of human LBP coding sequences fused to human IgG1 CH1 sequences and secreted from SP2/O murine myeloma cells in association with a human Ck domain from which the Vk region had been deleted. The LBP-Ig was purified to homogeneity by means of protein A chromatography. Acid (0.1 M citrate, pH 3.5)-eluted material was immediately neutralized with a molar excess of pH 7.8 Tris buffer and dialyzed extensively against PBS. Only material with low endotoxin content (<10 enzyme units/ml), as measured by the chromogenic Limulus amebocyte lysate assay (Bio Whittaker), was used in subsequent studies. LBP was released from the LBP-Ig by limited proteolysis with papain, and the immunoglobulin constant regions were removed by passage over a protein A affinity column. The released LBP (LBP60k) appeared as a 60-kDa band that was indistinguishable from native LBP by SDS-polyacrylamide gel electrophoresis. This material was subsequently biotinylated using D-biotin-N-hydroxysuccinimide ester following the manufacturer's instructions (Boehringer Mannheim). LPS-horseradish peroxidase conjugate was obtained from Alercheck, Inc. (Portland, ME). According to the manufacturer, it was prepared from Escherichia coli J5 LPS (rough LPS) by direct coupling of labeling grade horseradish peroxidase (Boehringer Mannheim) to periodate-activated LPS. The horseradish peroxidase:LPS coupling ratio was 2.5:1. LBP-Ig was Fc-captured by coating ELISA plates with 10 μg/ml goat anti-human Fc antibodies (Jackson ImmunoResearch) in bicarbonate buffer. After washing, purified LBP-Ig was added to the wells at 1 μg/ml in PBS containing 1% nonfat dry milk (PBSM), and allowed to incubate at room temperature for 30 min. Following washing, individual competitors and an equal volume of 250 μg/ml LPS-horseradish peroxidase were added to the ELISA wells and allowed to incubate at room temperature for 1 h. The plates were washed with PBS, and the chromogenic substrate O-phenylenediamine dihydrochloride was added and processed as described by the manufacturer (Sigma). In experiments using the preincubation format, the inhibitors are added alone to the assay wells and allowed to incubate prior to washing and addition of LPS-horseradish peroxidase. For ELISA assays measuring the binding of biotinylated LBP to solid phase LPS, ELISA plates were coated with 50 μg/ml E. coli 0111:B4 LPS in bicarbonate buffer essentially as described(8Tobias P.S. Soldau K. Ulevitch R.J. J. Biol. Chem. 1989; 264: 10867-10871Google Scholar). After washing with PBSM, different inhibitor dilutions (in PBSM) were incubated alone (preincubation experiments) or with an equal volume of 1 μg/ml (in PBSM) biotinylated LBP (competitive inhibition) and were incubated for 1 h at room temperature. For the preincubation assays, the plates were further washed, and 0.5 μg/ml (8 nM in PBSM) biotinylated LBP (LBP-B) was incubated and washed as before. Bound LBP was detected in both formats with 1/1000 Z-avidin-horseradish peroxidase conjugate (Zymed) and processed as described by the manufacturer. Human PBMC were purified from normal donor blood by differential centrifugation using mono-poly resolving medium (Flow-ICN) and following the instructions of the manufacturer. In a 200-μl final volume of Iscove's serum-free medium (JRH Biosciences), 5 × 105 PBMC were combined with different concentrations of E. coli 0111:B4 LPS and inhibitor peptides. LBP potentiated cultures were supplemented with 200 ng/ml (3 nM) purified human LBP (LBP60k, described above). Cultures were incubated for 3 h in a 37°C humidified CO2 incubator; supernatants were collected and assayed for human TNFα content by means of the previously described WEHI bioassay(9Espevik T. Nissen-Meyer J. J. Immunol. Methods. 1986; 95: 99-105Google Scholar). Purified human recombinant TNFα (Genzyme) was used as standard. An ELISA-based competitive inhibition assay was used to detect LPS binding peptides by means of their ability to inhibit binding of LPS-horseradish peroxidase to LBP-Ig. Since LBP binds different forms of LPS via the common lipid A moiety N,O-acylated β1,6-D-glucosamine disaccharide 1,4′-bisphosphate(8Tobias P.S. Soldau K. Ulevitch R.J. J. Biol. Chem. 1989; 264: 10867-10871Google Scholar), this competitive inhibition ELISA could potentially identify lipid A-specific peptides. Indeed, we found that LBP-Ig can specifically bind LPS-horseradish peroxidase in a way that is inhibitable by different LPS- and LBP-specific blocking molecules:2 2A. H. Taylor, unpublished observations. (i) smooth (E. coli 0111:B4) and rough (E. coli J5) unlabeled LPS and lipid A (Salmonella minnesota R595) could competitively inhibit binding of LPS-horseradish peroxidase to the Fc-captured LBP-Ig, each reaching 50% inhibition in the 10 μM range; (ii) polymyxin B (Roerig), a potent lipid A binder, also could inhibit the LPS-horseradish peroxidase with an IC50 of 0.05 μM; (iii) 8 of 26 different anti-human LBP monoclonal antibodies were found to inhibit the binding of LPS-horseradish peroxidase to LBP-Ig; (iv) a control LBP-Ig fusion protein in which the COOH-terminal-half of the LBP domain had been deleted displayed no LPS binding activity. We synthesized a panel of 146 overlapping 15-mer peptides corresponding to the entire LBP primary structure (10Schumann R.R. Leong S.R. Flaggs G.W. Gray P.W. Wright S.D. Mathison J.C. Tobias P.S. Ulevitch R.J. Science. 1990; 249: 1429-1431Google Scholar) and screened the crude peptides for LPS blocking activity in the competitive inhibition assay. Interestingly, only two of the nested peptides, 31 (WKVRKSFFKLQGSFD-NH2), corresponding to LBP residues 91-105, and 32 (RKSFFKLQGSFDVSV-NH2), corresponding to 94-108, could completely inhibit binding of LPS-horseradish peroxidase by LBP-Ig (Fig. 1A). The two active peptides were selected for further study. Three peptides from the LBP-(91-108) region were synthesized and purified to homogeneity: LBP-(91-105)-NH2 (WKVRKSFFKLQGSFD-NH2), LBP-(94-108)-NH2 (RKSFFKLQGSFDVSV-NH2), and LBP-(91-108)-NH2 (WKVRKSFFKLQGSFDVSV-NH2). All three peptides were freely soluble in distilled water. The peptides were assessed for the ability to block binding of LPS-horseradish peroxidase to Fc-captured LBP-Ig as before and were found to exhibit similar IC50 values, each in the 10-20 μM range (Fig. 1B). A fourth analog was synthesized, LBPAc-(91-105)-NH2 (Ac-WKVRKSFFKLQGSFD-NH2); interestingly, it was found to be totally inactive, at least within the concentration range tested. This relatively inactive analog was used as a negative control in several of the experiments described below. To define further the specificity of the LBP-(91-108)-related peptides for LPS, two types of preincubation experiments were carried out. First, as expected, preincubation of unlabeled LPS with Fc-captured LBP-Ig, followed by washing, inhibited subsequent binding of LPS-horseradish peroxidase (Fig. 2A). However, preincubation of the peptides with Fc-captured LBP-Ig, followed by washing, had no effect on subsequent binding of LPS-horseradish peroxidase; this indicated that the peptides do not interact with LBP-Ig to inhibit LPS binding. Second, binding of biotinylated LBP (LBP-B) to LPS-coated plates could be inhibited in a dose-dependent manner by preincubation of the plates with various analogs of LBP-(91-108) (Fig. 2B) or by polymyxin B. As before, LBP Ac-(91-105)-NH2 did not block binding of LBP to LPS. These results show that the LBP-(91-108) peptides specifically inhibit the LPS-LBP interaction by binding to LPS. We calculated the LPS binding affinity of the LBP-(91-108) peptides after determining the relative ability of the LBP peptides and polymyxin B to compete with LBP-B for binding to solid phase LPS. The signal obtained from binding 8 nM LBP-B to solid phase LPS was inhibited, by 50%, by simultaneous addition of 1 μM polymyxin B, indicating a 125-fold difference in relative affinity for LPS. Since polymyxin B has a Kd = 8.6 × 10-8M, the relative affinity of LBP-B for LPS = (8.6 × 10-8)/(125) = 6.9 × 10-10M, in good agreement with a previous report(8Tobias P.S. Soldau K. Ulevitch R.J. J. Biol. Chem. 1989; 264: 10867-10871Google Scholar). Values for the peptides were calculated using the same method and ranged between 0.1 and 0.04 of the affinity held by polymyxin B (Table 1). Interestingly, the affinities of the LBP peptides were found to be improved by as much as 12-fold by replacing the amino-terminal amino acid residues with D-isomers (Table 1). Thus, the LPS binding affinities of the LBP peptides approach that of polymyxin B, a potent but highly toxic inhibitor(11Yow E. Moyer J. Smith C. Arch. Intern. Med. 1958; 92: 248-257Google Scholar).TABLE I Open table in a new tab Two models were utilized to determine if the LBP-(91-108) peptides could also block complex biological responses to LPS challenges in vitro. The chromogenic limulus amebocyte lysate assay (LAL) is sensitive to minute amounts of LPS (as little as 1 pg/ml) and is specific for smooth and rough LPS forms, as well as for lipid A. It can be seen (Table 2) that the same LBP-(91-108) peptides that were found to block binding of LPS by LBP also can neutralize the LAL reaction to lipid A; LBP-(91-105)-NH2 and LBP-(91-108)-NH2 displayed similar lipid A neutralizing potencies as found for polymyxin B (Table 2). These results show that the LBP peptides can completely block the exquisitely sensitive LAL reaction to lipid A and accordingly indicate that the LBP-(91-108)-related peptides are specific for lipid A.TABLE II Open table in a new tab A large degree of LPS toxicity is a consequence of TNFα released by CD14+ monocyte/macrophage lineage cells following exposure to LPS•LBP complexes(12Tracey K.J. Lowry S.F. Adv. Surg. 1990; 23: 21-56Google Scholar, 13Beutler B. Milsark I.W. Cerami A. Science. 1985; 229: 869-871Google Scholar). Accordingly, we wished to determine if the LBP peptides could block the LBP-dependent LPS response of human mononuclear cells in vitro. Fig. 3 shows the LBP-dependent TNFα response of purified human PBMC to different doses of LPS and the blocking ability of a representative LBP peptide. In the absence of human serum, PBMC are completely unresponsive to 1 ng/ml LPS. As previously shown(5Mathison J.C. Tobias P.S. Wolfson E. Ulevitch R.J. J. Immunol. 1992; 149: 200-206Google Scholar, 10Schumann R.R. Leong S.R. Flaggs G.W. Gray P.W. Wright S.D. Mathison J.C. Tobias P.S. Ulevitch R.J. Science. 1990; 249: 1429-1431Google Scholar), potentiation with added purified LBP restored the ability of the PBMC to release robust amounts of TNFα in response to 1 ng/ml LPS. Addition of only 3 μM LBP-(94-108)-NH2 inhibited this LBP-dependent TNFα response to 1 ng/ml LPS by 50%, and 30 μM LBP-(94-108)-NH2 completely abolished the TNFα response. In fact, 30 μM LBP-(94-108)-NH2 could substantially inhibit the PBMC response to 10 ng/ml LPS, even though much of this response appears to be LBP independent. In separate experiments, similar results were also obtained using a serum-stable analog of LBP-(94-108)-NH2 (viz. LBP-(d-94-108)-NH2) when tested with 1% normal human serum-potentiated PBMC response to 1 ng/ml LPS.2 The results show that 1-10 μM LBP-(94-108) peptides can block both purified LBP- and normal human serum-potentiated LPS responses to physiologically relevant doses of LPS. The LBP-(91-108) peptides were also tested for the ability to block TNFα response to LPS challenge in vivo. When mice were coinjected with 1 ng of LPS/galactosamine and individual LBP-(91-108) peptides, each of the LBP-(91-108) peptides could significantly inhibit the TNFα response (Fig. 4, A-E). All of the peptides could inhibit by at least 50%; two of the peptides, LBP-(d-94-108)-NH2 and LBP-(d-91-108)-NH2, could completely inhibit the TNFα response when administered at about 10 and 40 μg/mouse, respectively. Fig. 4, panel F, shows the relative LPS neutralization potencies of the peptides. A 500-fold range of potencies was evident, with 0.1 μg/mouse of the most potent peptide blocking by 50% the TNFα response to 1 ng of LPS. Because the maximum serum LPS levels generally found among septic patients do not exceed 0.1 ng/ml, the results augur well for the ability of these peptides to inhibit LPS toxicity in vivo (studies in progress). Few examples exist in the literature that describe potent blocking peptides with Kd in the submicromolar range. This is true even when detailed knowledge of the three-dimensional structure of the receptor and its ligand are available. Therefore, it is remarkable that two overlapping 15-mer LPS binding peptides were identified that can completely block the high affinity interaction between LPS and LBP. Our success with this strategy may reflect essential differences in how protein receptors recognize small ligands (such as the β1,6-D-glucosamine disaccharide 1,4′-bisphosphate that most likely represents the moiety of lipid A recognized by LBP), as opposed to the large discontinuous interactive surfaces characteristic of protein-protein interactions. Accordingly, it is possible that the approach described here might be of general use when the interactive portion of the ligand is predicted to be relatively small. The results suggest that in the intact protein, LBP residues 91-108 play a critical role in the formation of specificity to LPS, forming at least part of the LPS binding site. This proposal is supported by several previously reported observations. (i) Recent modeling experiments have attempted to predict the LPS binding site of human LBP based on the three-dimensional crystal structure of the Limulus-anti-LPS factor and certain sequence similarities between Limulus-anti-LPS factor and LBP(14Hoess A. Watson S. Siber G.R. Liddington R. EMBO J. 1993; 12: 3351-3356Google Scholar). Interestingly, LBP-(91-108) comprises the loop and the second β-strand of the amphipathic loop structure predicted to be the LPS binding site in these modeling experiments. (ii) Rabbit LBP, when cleaved at a unique plasmin hypersensitivity site between residues 99 and 100, loses all LPS binding activity(1Tobias P.S. Mathison J. Mintz D. Lee J.-D. Kravchenko K. Pugin J. Ulevitch R.J. Am. J. Respir. Cell Mol. Biol. 1992; 7: 239-245Google Scholar). (iii) The amino-terminal fragment of LBP (residues 1-197) retains the LPS binding activity of the native molecule(15Theofan G. Horwitz A.H. Williams R.E. Liu P.-S. Chan I. Birr C. Carroll S.F. Meszaros K. Parent J.B. Kasler H. Aberle S. Trown P.W. Gazzano-Santoro H. J. Immunol. 1994; 152: 3623-3629Google Scholar). (iv) A 15-mer peptide derived from bactericidal permeability increasing protein (BPI) has been described that neutralizes LPS in the LAL reaction and is bactericidal(16Little R. Kelner D.N. Lim E. Burke D.J. Conlon P.J. J. Biol. Chem. 1994; 269: 1865-1872Google Scholar), two properties of native BPI(17Marra M.N. Wilde C.G. Griffith J.E. Snable J.L. Scott R.W. J. Immunol. 1990; 144: 662-666Google Scholar, 18Weiss J. Muello K. Victor M. Elsbach P. J. Immunol. 1984; 132: 3109-3115Google Scholar). LBP and BPI display considerable sequence identity (10Schumann R.R. Leong S.R. Flaggs G.W. Gray P.W. Wright S.D. Mathison J.C. Tobias P.S. Ulevitch R.J. Science. 1990; 249: 1429-1431Google Scholar) and share the ability to bind LPS with high affinity. After sequence alignment (which does not require the introduction of gaps through residues 122)(10Schumann R.R. Leong S.R. Flaggs G.W. Gray P.W. Wright S.D. Mathison J.C. Tobias P.S. Ulevitch R.J. Science. 1990; 249: 1429-1431Google Scholar), the bactericidal/LPS neutralizing BPI-related peptide (BPI 85-99) shares partial identity with the active LBP oligomers described here (LBP 91-108). However, unlike BPI(85-99), preliminary experiments indicate that the LBP-(91-108)-related peptides do not inhibit growth of Gram-negative organisms.2 The results described here suggest high resolution site-directed mutagenesis experiments that, together with monoclonal antibody epitope mapping experiments and extensive analog peptide structure-function analysis, will lead to a more detailed understanding of the LBP-LPS interaction (studies in progress). We thank B. Scallon for kindly providing the LBP-Ig and purified recombinant LBP used in these studies and Marian Kruszynski, Robert Weber, and Margret Falcone for synthesis, purification, and physical characterization of the LBP-related peptides.