Analysis of Unmodified Endotoxin Preparations by 252Cf Plasma Desorption Mass Spectrometry

Transcription

1 rhe JOURNAL OF BIOLOGICAL CHEMISTRY c: 1991 by The American Society for Biochemistry and Molecular Biology, Inc. Val. 266, No. 28, Issue of Octoher 5. pp Printed in U.S.A. Analysis of Unmodified Endotoxin Preparations by 252Cf Plasma Desorption Mass Spectrometry DETERMINATION OF MOLECULAR MASSES OF THE CONSTITUENT NATIVE LIPOPOLYSACCHARIDES* (Received for publication, March 1, 1991) Martine Caroff#, Claude Deprun8, Doris KaribianS, and Ladislas Szabo# From the SEquipe Endotoxines du Centre National de la Recherche Scientifique, Institut de Biochimie, Universite de Paris Sud, and the Equipe Interaction Ion-Surface et Materiaux, Institut de Physique Nucleaire, F-91405, Orsay, France Nine unmodified endotoxin preparations constituted of Re-, Rd-, and Rc-type lipopolysaccharides (2 to 5 glycoses), representing four species of enterobacteria were analyzed by Cf plasma desorption mass spectrometry. The constituent lipopolysaccharides were characterized by the ion pair: (M - H)- and its corresponding lipid fragment ion. The lipid fragment ion is produced by cleavage of the glycosidic bond of the 3- deoxy-~-manno-oct-2-u~osonic acid unit that substitutes 0-6 of the glucosamin~l -6glucosamine ( lipid A backbone ) disaccharide of the lipid A moiety. These lipid fragment ions were identical to the (M- H)- ions seen in the spectra of homologous isolated lipid A preparations that were obtained by hydrolysis (ph 4.5, 100 C) promoted by sodium dodecyl sulfate. Since the molecular components present in the endotoxin preparations analyzed are known, the ion pair (M - H)-- lipid fragment ion defines the molecular compositions of each individual lipopolysaccharide. Heterogeneity of the R-type endotoxin preparations analyzed was due almost exclusively to differing lipid A moieties. In three Salmonella minnesota 595 Re endotoxin preparations 10 different lipopolysaccharides were identi- fied, only two of which were common to all three preparations. Of the nine lipopolysaccharides identified in two S. minnesota R7 endotoxin preparations, only two were present in both. ( monophosphoryl lipid A ) of endotoxic lipopolysaccharides (17-23). The material thus analyzed was obtained from endotoxin preparations by treatment with acid, followed by chro- Endotoxic lipopolysaccharides are amphipathic macromolmatographic purifications. With one exception (21) for all ecules that are obligate (1, 2), major (3) antigens (4,5) of the such studies the material was derivatized with diazomethane. outer membrane of Gram-negative bacteria. The fundamental PDMS was used to establish (19) the molecular mass element of their hydrophobic ( lipid A ) domains is a bis- and to confirm previously proposed structures, based on degphosphorylated, pl,g-linked disaccharide of D-glucosamine radative (24) and NMR (25) studies, of a single Re-type (i.e. substituted by fatty acids (6), 3R-3-hydroxytetradecanoic, constituted of lipid A and 2 KDO residues) lipopolysaccharide dodecanoic, tetradecanoic, and hexadecanoic acids being fre- (26). To this end a major component of the endotoxin prepaquently encountered. In the hydrophilic domain which is ration was isolated by column chromatography and derivaconstituted of various acidic, basic, and neutral glycoses two regions can be distinguished the core, proximal to lipid A The abbreviations and trivial name used are: KDO, 3-deoxy-oand the side-chain. The biosynthetic pathways and the manno-oct-2-ulosonic acid; C12, dodecanoic acid; C14, tetradecanoic overall structure of these regions are different (7,8). Core and acid; C140H, hydroxytetradecanoic acid; C16, hexadecanoic acid; lipid A are invariably joined by the glycosidic bond of a 3- Etn, ethanolamine (2-amino-ethanol); lipid A, the hydrophobic domain of endotoxic lipopolysaccharides; isolated lipid A, the hetero- * This work was supported by the Universite de Paris-Sud (Grant A.I.89.14) and the Fondation pour la Recherche Medicale. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked aduertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact deoxy-~-rnanno-2-octulosonic acid (KDO) which, in certain strains, may be the sole component of the hydrophilic domain. Being among the most efficient activators of the human immune system, endotoxin preparations are extensively used to induce immunological reactions in whole animals and isolated cells, and numerous attempts have been made to establish structure/activity relationships. However, because of the heterogeneity of the inducer, the data obtained cannot be considered as unambiguous. It has been demonstrated by chromatography (9) and by electrophoresis (10,ll) that endotoxin preparations ( endotoxin, LPS ), as obtained by extraction (12) of bacterial cells, are heterogeneous, and represent ensembles of related, but constitutionally distinct, molecular species. Chemical analysis of fragments produced by hydrolysis revealed that heterogeneity may be due to the variability of both domains (13, 14). Depending on the genus, the species, the strain, and the culture conditions of the cells, the molecular masses of the individual lipopolysaccharides present in any given endotoxin preparation may vary considerably (15). As estimated by indirect methods, they appear to range from 2 to 20 kda (16); however, none have been determined accurately. So far no endotoxin preparation has been described in terms of unmodified individual chemical structures of its constituent lipopolysaccharides. Different forms of mass spectrometry (fast atom bombardment, laser desorption, 2f PDMS) have been used to determine molecular masses and exact chemical structures of major components of hydrophobic domains ( diphosphoryl lipid A ) or fragments geneous precipitate isolated after treatment of endotoxin preparations with acid lipid base, a group of constituents found to be common to most of the lipopolysaccharides present in the endotoxin preparations analyzed in this study; LPS, endotoxic lipopolysaccharide of bacterial origin; n.i., not identified; n.o., not observed; obs., observed PDMS, plasma desorption mass spectrometry; PenN, pentosamine; Phosph., phosphoric acid; HPLC, high performance liquid chromatography.

2 18544 Cf Analysis PDMS tized, and then the derivatized material was further purified by HPLC. The hexamethyl ester of the lipopolysaccharide thus obtained was then analyzed by 252Cf PDMS in the positive ion mode. Isolated lipopolysaccharides with glycose chains consisting of more than two KDO units have not been analyzed by this technique so far, nor have endotoxin preparations been defined in terms of their constituent lipopolysaccharides. It has now been found that unmodified Re- to Rc-type (27) endotoxin preparations, the constituent lipopolysaccharides of which have short glycose chains (up to 5 sugars), can be analyzed by Cf PDMS, and the preparations described as ensembles of lipopolysaccharides of defined composition. The results thus obtained accounted for all constituents previously shown to be present in the endotoxin preparations and agree with the proposed general architecture of endotoxic lipopolysaccharides. The interpretation of lipopolysaccharide spectra was confirmed by spectra obtained for homologous isolated lipid A preparations. EXPERIMENTAL PROCEDURES~ RESULTS of Endotoxin Preparations Negative-ion PDM spectra (Fig. 1) were obtained for nine conditions of hydrolysis ADP released 0.5 mol of phosphate different, commercially available, unmodified endotoxin prepper molof ADP. Relatively weak signals corresponding to arations derived from Re-, Rd-, and Rc-type mutant cells, and ions containing one instead of two KDO units (m/z 2018/ for their respective isolated lipid A fragments. Since the 2020) were also detected in both endotoxin preparations. This molecular components that are released upon acid-, or basetype of ion has been observed repeatedly (see below). catalyzed hydrolysis of such enterobacterial endotoxins are The other endotoxin preparations were far more heterogewell known, possible molecular masses of individual lipopolyneous. Thus, endotoxins of S. minnesota 595 Re (Sigma, lots saccharides could be computed. They were found to account, 55 and 28) (C and D, respectively), were mixtures of at least almost without exception, for signals observed at high (mlz 5 lipopolysaccharides, while that of S. minnesota 595 Re List > 2200)values. Consequently, these signals were taken to (E) contained at least 7 (Fig. 1 and Table 1). Although some represent (M- H)- ions. Those appearing in the lower range lipopolysaccharides were present in all three endotoxins (m/ of the spectrum were taken to be due to fragment ions: indeed, z and ), their relative amounts varied considalmost all of them coincided with (M- H)- ions observed in erably. Others appeared in one preparation only: m/z spectra of homologous isolated lipid A preparations. The (1 in D); 2556, , and (2, 4, and 5 in E). spectra thus interpreted revealed that all endotoxins exam- As expected, all six lipopolysaccharides identified in the ined were mixtures of a variable number of lipopolysaccha- endotoxin of Salmonella typhimurium SL1181, an Rd2 mutant rides (Table 1). The expression lipid base used here refers (27) (erroneously labeled in the 1990 Sigma catalog as an Re to a group of nine constituents (2 GlcN, 2 phosphate, 4 mutant), contained one heptose unit (Table 1, F). The hydro- C140H, 1 C12) that was found to be present in all (save F6) phobic domain of lipopolysaccharide 6 of this endotoxin was lipopolysaccharides identified. This group is, however, not unusual inasmuch as it contained only 3 instead of 4 C140H necessarily the common set of constituents for other endo- (incomplete lipid base). toxins. Lipid base is not equivalent to lipid A. The latter is Spectra of the two Rdl endotoxin preparations of S. minthe name extensively used to designate the hydrophobic do- nesota R-7 (G, List, H, Sigma) revealed considerable differmain of endotoxic lipopolysaccharides. ences between them. Lipopolysaccharides and lipid fragments Since no data are available concerning the fragmentation containing aminopentose or ethanolamine and additional of unmodified lipopolysaccharides in the conditions of 252CF phosphate (1, 3, 4, 4, a) were present in H. In G, molecular PDMS, a synthetic model, allyl 2-deoxy-2-[(3R)-3-hydroxy- species and lipid fragments containing supplementary phostetradecanamido]-6-0-(3-deoxy-a-~-manno-2-octulopyra phate (2,3 and 2,3 ) or an additional C140H (a, 2,2, 4,4,4 ) nosidono)-p-~-glucopyranoside)4-phosphate (32), was used to ascertain that in oligosaccharides the glycosidic bond of KDO units was in fact preferentially cleaved (Fig. 2). Spectra of the endotoxins of Escherichia coli D31m4 and Shigella flexneri (both Re mutants) (Fig. 1 and Table 1: A and B) were very similar. The two most intense signals of each spectrum (m/z 2318 and 2238), taken to be molecular ions, indicated that the corresponding lipopolysaccharides had, within experimental error, the same molecular masses. Their compositions are given in Table 1. The lipopolysaccharides Portions of this paper (including Experimental Procedures, Table 1, and Figs. 1 and 2) are presented in miniprint at the end of this paper. Miniprint is easily read with the aid of a standard magnifying glass. Full size photocopies are included in the microfilm edition of the Journal that is available from Waverly Press. with (M - H)- m/z 2318 differed from those with (M - H) only by the presence of an additional phosphate (+80 atomic mass units), that was probably part of a pyrophosphate group (24, 25). Other lipopolysaccharides, apparently minor components, were also present in both endotoxins; e.g. in that of E. coli an (M - H)- ion appeared at m/z 2264 due to a lipopolysaccharide species containing C16 instead of C14. No such minor components were detected by their (M - H)- in the spectrum of the Sh. flexneri endotoxin. Their existence was attested however by unresolved signals (at m/z ) visible among the lipid fragment ions of the spectra of both the Sh. flexneri and the E. coli endotoxins, and also in the spectra of their isolated lipid A preparations. Fragment ions constituted of lipid base + C16 (m/z ) or + C140H (m/z ) appear in the region m/z Homologs of these ions, containing one additional phosphate group, were found among the lipid fragment ions of the E. coli endotoxin 80 atomic mass units higher (m/z 1905 and 1893, incompletely resolved). In the spectra of the isolated lipid A of both endotoxins the signals, due to lipid fragments carrying the presumed pyrophosphate group (m/z ), are weak. Since the isolated lipid A preparations were obtained by acid hydrolysis, (31) loss of pyrophosphate in this and other isolated lipid A preparations is not unexpected. Under the same were observed. Rc-, Rdl-, and Rd2-type lipopolysaccharides were identified among the six detected in the endotoxin of S. minnesota R-5, which is described (33) as an Rc mutant, (Fig. 1, Table 1, I). Indeed, while four of these (1, 2, 3, 4) contained 2 heptoses plus 1 hexose, lipopolysaccharide 5 contained only 2 heptoses, and lipopolysaccharide 6 only 1 heptose in addition to the 2 KDO units. Both basic (aminopentose, ethanolamine) and acidic (phosphate) components were present, but no aminopentose-containing lipopolysaccharide was detected by its (M - H)- ion. DISCUSSION Characteristics of the Method-While pure, synthetic compounds, constituted essentially of a single molecular species,

3 gave good spectra rapidly, production and quality of the spectra of endotoxins were critically dependent on the procedures used for the preparation of both sample and target; even for experiments with the same material, the time required to accumulate statistically sufficient data varied considerably (12-44 h). Reasons for these phenomena are not known with certainty, but the presence of cations such as Ca2+ and Mg', invariably present in endotoxin preparations and very difficult to remove quantitatively (34), may be at least partly responsible, as even small amounts of such ions may interfere with the production of 252CF PDM spectra and alter their quality (35). Endotoxins gave useful spectra as negative ions only; positive ion spectra were either not produced or reflected extensive fragmentation. For isolated lipid A preparations, both positive and negative ion spectra were obtained; the latter were better for the purpose of identification of molecular peaks. As a rule, isolated lipid A preparations were ionized/desorbed much faster than endotoxins. Since both endotoxins and isolated lipid A preparations form heteromolecular aggregates, both in solution and on the target, the concentration of monomolecular species is diminished. Signals produced by less abundant and/or less readily ionized/ desorbed lipopolysaccharides will be particularly affected. Accordingly, their intensity may be reduced below the threshold of detection, since the intensity of the signals declines sharply when the concentration falls below a certain Moreover, production and breakdown of charged particles from aggregates during the phase of acceleration will tend to blur the spectra by worsening the signal-to-noise ratio and preclude detection of lipopolysaccharides present in small amounts. These phenomena may explain the appearance of lipid fragment ions and (M - H)- ions in spectra of isolated lipid A preparations even though the (M- H)- ion of the corresponding lipopolysaccharide was not observed in the spectrum of the endotoxin. In spectra of isolated lipid A preparations the absence of signals representing structures particularly labile to acid (e.g. pyrophosphate, aminopentose, 0-phosphoryl ethanolamine) may also be due to hydrolytic cleavage. Experiments with synthetic glycolipids4 established that, within a limited range, increasing amounts of material pro- duced progressively stronger signals, but the relationship was, at best, semiquantitative. Therefore, in any given spectrum, the intensities of signals are merely suggestive of the relative amounts of individual lipopolysaccharides present. Validity of the Method-It has been firmly established (24, 25) that two (presumably major) lipopolysaccharides are present in the endotoxin of E. coli D31m4, an authentic Re mutant (26). Both are made up of lipid A plus 2 KDO units, but one has an additional phosphate as part of a monoesterified pyrophosphate group (24). In the PDM spectrum of this endotoxin (Fig. la), the numerical value of the signal at m/z is equal to that calculated ( ) for a lipopolysaccharide composed of 2 KDO units, lipid base, and 1 tetradecanoic acid, i.e. constituents previously identified (24) in this endotoxin. Within experimental error, the numerical value of the signal appearing at m/z corresponded to that calculated ( atomic mass units) for the expected lipid A fragment containing the above-mentioned elements minus the 2 KDO units. In this same spectrum m/z equalled the sum of the above-mentioned constituents plus one additional phosphate. Finally, the chemical molecular weight ( ) assigned to the major lipopolysaccharide of this endotoxin was equal to that calculated from the data published (18) for a painstakingly purified, methylated derivative :I C. Deprun, unpublished observation. C. Deprun, M. Caroff, and D. Karibian, unpublished observation. "'Cf PDMS Analysis of Endotoxin Preparations of the major lipopolysaccharide of the same endotoxin. Thus the molecular composition of the main constituents of this endotoxin determined by PDMS was found to correspond to the structure established previously by chemical and physical methodology. The putative molecular peaks (M- H)- and the presumably corresponding lipid fragment peaks were correlated in the same way in the spectra of all the other endotoxins. The (M - H)- peaks present in the spectra of homologous isolated lipid A preparations concurred. Hence it was concluded that, in the conditions used to produce the spectra, the glycosidic bonds of KDO units were ruptured easily and selectively and that this fragmentation was a characteristic feature of PDM spectra of endotoxic lipopolysaccharides. The conclusion was corroborated by the PDM spectrum (Fig. 2) of a disaccharide, a close, synthetic analog of the region that embodies the junction core-lipid A in lipopolysaccharides, in which the main signals observed were (M - H)- and ((M - KDO) - H)-. Taken together these data also prove that the method of hydrolysis (31) used afforded lipid A preparations in which the hydrophobic domains of the lipopolysaccharides were particularly well preserved. The coherence of all these data justified the assignment of high field signals (m/z > 2200) to molecular peaks of individual lipopolysaccharides of a given endotoxin preparation, and oflow field signals to fragment ions corresponding to the hydrophobic domains and fragment ions thereof; it established, moreover, that 252Cf PDMS was a valid method of analysis of R-type endotoxin preparations. Patterns of Fragmentation-The signals at m/z and at in the endotoxin spectra of E. coli D31m4 and of Sh. flexneri, respectively, could either be (M - H)- ions containing one KDO unit instead of the usually encountered two, or may be fragment ions formed from the lipopolysaccharide (2237.6) through loss of a KDO unit. This type of ion was also observed in the endotoxin spectra C (6*,T**), D (6*), and G (6*). If 6* of spectrum G is in fact a fragment ion, it must have been produced by loss of an extracatenary KDO, as Rdl-type lipopolysaccharides have 2 KDO and 2 heptose units. Lipopolysaccharides containing 3 KDO units have not been discovered, but they may have escaped detection either because of the fragility of the glycosidic bond of this constit- uent (Fig. 2) in the conditions of 252Cf PDMS or because of their low abundance. Fragmentation of the hydrophobic domain has also been observed. Indeed, in all spectra in which the ion m/z 1797 (lipid base + C14) appeared, the ion m/z 1587 was also present. The numerical difference (210 atomic mass units) between the two signals corresponds to formal loss of tetradecanoic acid. However, without independent proof, an ion appearing at m/z 1587 should not be interpreted as being the daughter ion of m/z 1797 and taken to establish the presence of an ester-bound tetradecanoic acid in the endotoxin prepa- ration. Indeed, at least in enterobacterial endotoxins, the same signal may also be due to the lipid base produced as a primary fragment. For instance, this was the case for the signal H 6' which could be attributed to that primary fragment without ambiguity. In all of the individual lipopolysaccharides (F 6 of S. typhimurium excepted) identified in the endotoxin preparations analyzed, the nine molecular constituents of the lipid base were invariably present. The constant presence of C12 and the frequent absence of c14 were unexpected inasmuch as the latter is considered to be a characteristic component of endotoxic lipopolysaccharides. In one lipopolysaccharide, namely 6 of S. minnesota R-7 (H), the hydrophobic domain

4 18546 Cf Analysis PDMS consisted of lipid base alone, and in another, 6 of the endotoxin of S. typhimurium SL 1181 (F), the lipid base lacked the fourth C140H and contained no additional constituent. Thus the former had a penta-, the latter a tetra-acyl lipid A region. Lipopolysaccharide Composition of the Endotoxins Analyzed-Heterogeneity appeared in both the hydrophilic and the hydrophobic domains of the S. minnesota R-5 (Rc type) endotoxin. In all other endotoxin preparations heterogeneity was restricted to the hydrophobic domain and was due solely to those structural elements that are usually referred to as constituents not present in stoichiometric amounts. It is nevertheless noteworthy that, although not very numerous, these constituents generate considerable heterogeneity even in endotoxin preparations constituted of lipopolysaccharides that are strictly homogeneous as regards their hydrophilic domain (Table 1: C-H). Moreover, in different batches of nominally identical endotoxin preparations, different lipopolysaccharide species may predominate (Fig. 1: C, D, and E G and H). This second type of heterogeneity may be due to such parameters as culture conditions, method of extraction, etc.; if so, its appearance is potentially avoidable. Re-type endotoxins derived from S. minnesota 595 Re (Table 1, C-E) were among the most heterogeneous preparations. Of 10 different lipopolysaccharides identified, only two (m/z calc , and ) were present in all of them, and considerable variation was observed even for different batches of the same strain from the same source (cf. C and D). Hepta-acyl (C, 1,2; D, 1,2; E, 1-3) and hexa-acyl (C, 3-5; D, 3-5; E, 4-7) lipopolysaccharide molecular ions, the corresponding lipid fragment ions and isolated lipid A molecular ions have been recognized. Hepta-, hexa-, and penta-acyl monophosphoryl lipid A fragments have been obtained previously from s. minnesota 595 Re endotoxin by hydrolysis with 0.1 M HC1; they were isolated by silicic acid chromatography, purified by HPLC, and their structures established by fast atom bombardment-mass spectrometry (17). The dimethyl monophosphoryl hepta-acyl lipid A [(M + H) = m/z observed by these investigators corresponds in Table 1, C and E, to the lipid A (M - H)- ions C and E 3 [mlz (calc )], to the lipid fragment ions C 2 and E 3 which are the lipid A fragments of the lipopolysaccharides C 2 and E 3, [(M - H)- = m/z , calc This LPS was not detected in the endotoxin of S. minnesota 595, Sigma, lot 28 (spectrum and Table D). The dimethyl monophosphoryl hexa-acyl lipid A [(M + H) = m/z identified by the same investigators corresponds to the lipid fragment ions C 5, 7, D 5 and E 7 and the lipid A (M - H)- ions m/z C 5, 7, D 5, and E 7 and consequently to LPSs C 5, D 5 and E 7 [(M - H)- observed m/z , , and , respectively; calc. m/z This lipopolysaccharide, one of the simplest, was present in all three S. minnesota 595 endotoxin preparations. No penta-acyl lipopolysaccharide or isolated lipid A fragment was detected in the spectra of the three endotoxin preparations of S. minnesota 595 analyzed, although one such lipopolysaccharide was present in S. minnesota R-7 endotoxin (H 6). Aminopentose (36), ethanolamine (37), and additional phosphate (24) are frequently encountered components of Salmonella endotoxin preparations. Lipopolysaccharide molecular ions having hepta-acyl (C, 1; D, 1, 2; E, 1, 2) or hexa-acyl (C, 3; D, 3, 4; E, 4,5) lipid A regions and one or several of these constituents were also found in the spectra. Because of their very small, simple, and well characterized hydrophilic domain, Re-type endotoxins are extensively used as immunological tools. Studies with synthetic compounds of Endotoxin Preparations identical to those known (38) to be part of isolated lipid A preparations have demonstrated that seemingly small differences may alter considerably their biological potencies (38). In view of the heterogeneity and variability of their hydrophobic domains, Re-type endotoxin preparations should be used with caution for immunological studies. While lipopolysaccharides of S-type endotoxins are, for the time being, not amenable to analysis by 252Cf PDMS, molecular compositions of their hydrophobic domains can be easily established. Indeed, as the lipid fragment ions that appeared in the spectra of the untreated endotoxin preparations almost invariably coincided with those detected in the isolated lipid A preparations, it follows that hydrolysis of the glycosidic bond of the KDO unit at ph 4.5 in the presence of sodium dodecyl sulfate (31) is not accompanied by noticeable destruction and/or loss of constituents of the hydrophobic domain. This method of hydrolysis can even be applied to lipopolysaccharides in which the glycosidic bond of KDO is relatively stable, e.g. LPS-2 of the Bordetellu pertussis endotoxin (39). Since both preparation of the material and obtention of the spectra are relatively rapid and simple, the method seems to be well adapted for performing comparative and taxonomical studies on isolated lipid A fragments (40). Acknowledgments-We thank Dr. B. C. Das (Centre National de la Recherche Scientifique, Gif-sur-Yvette) who suggested the use of PDMS for the determination of molecular masses of lipopolysaccharides and Dr. Y. Le Beyec (Institut de Physique NuclBaire, Orsay) for his interest, help, and encouragement. REFERENCES 1. Rick, P. D., and Osborn, M. J. (1972) Proc. Natl. Acad. Sci. U. S. A.69, Rick, P. D., and Osborn, M. J. (1977) J. Biol. Chem. 252, Lugtenberg, B., and Van Alphen, L. (1983) Biochem. Biophys. Acta 737, Morgan, W. T. J. (1937) Biochem. J. 31, Westphal, O., Jann, K., and Himmelspach, K. (1983) Progr. Allergy 35, Hase, S., and Rietschel, E. Th. (1976) Eur. J. Biochem. 63, Osborn, M. J., Rosen, S. M., Rothfield, L., Zelecnik, L. D., and Horecker, B. L. (1964) Science 115, Wright, A., Dankert, M., and Robbins, P. W. (1965) Proc. Natl. Acad. Sci. U. S. A. 54, Buttke, T. M., and Ingram, L. 0. (1975) J. Bncteriol. 124, Jann, B., Reske, K., and Jann, K. (1975) Eur. J. Biochem. 60, Palva, E. T., and Makela, P. H. (1980) Eur. J. Biochem. 60, Westphal, O., Luderitz, O., and Bister, F. (1952) 2. Naturforsch. 7B, Jansson, P. E., Lindberg, A. A,, Lindberg, B., and Wollin, R. (1981) Eur. J. Biochem. 115, Cheng, C. H., Johnson, A. G., Kasai, N., Key, B. A., Levin, J., and Nowotny, A. (1973) J. Infect. Dk. 128, Weiss, J., Hutzler, M., and Kao, L. (1986) Infect. Immun. 51, Shands, J. W., Jr., and Chun, P. W. (1980) J. Biol. Chem. 256, Qureshi, N., Mascagni, P., Ribi, E., and Takayama, K. (1985) J. Biol. Chem. 260, Takayama, K., Qureshi, N., Hyver, K., Honovich, J., Cotter, R. J., Mascagni, P., and Schneider, H. (1986) J. Biol. Chem. 261, Qureshi, N., Takayama, K., Mascagni, P., Honovich, J., Wong, R., and Cotter, R. J. (1988) J. Biol. Chem. 263, Qureshi, N., Honovich, J. P., Hara, H., Cotter, R. J., and Takayama, K. (1988) J. Biol. Chem. 263, Qureshi, N., Takayama, K., Heller, D., and Fenselau, C. (1983) J. Biol. Chem. 258,

5 "'Cf PDMS Analysis of Endotoxin Preparations Qureshi, N., Takayama, K., and Ribi, E. (1982) J. Biol. Chem. 31. Caroff, M., Tacken, A., and Szabo, L. (1988) Carbohydr. Res. 257, , Cotter, R. J., Honovich, J., Qureshi, N., andtakayama, K. (1987) 32. Auzanneau, F.-I., Mondange, M., Charon, D., and Szabo, L. Biomed. Enuiron. Mass Spectrom. 14, Rosner, M. R., Tang, J., Barzilay, I., and Khorana, H. G. (1979) (1991) Carbohydr. Res., in press 33. Schmidt, G., and Luderitz, 0. (1969) Zentralbl. Bakteriol. Para- J. Biol. Chem. 254, sitenk. Abt. 1 Orig. 210, Rosner, M. R., Khorana, H. G., and Satterthwait, A. C. (1979) J. 34. Field, A. M., Rowatt, E., and Williams, J. P. (1989) Eiochem. J. Bid. Chem. 254, Bowman, H. G., and Monner, D. A. (1975) J. Eacteriol. 121, 263, Deprun, C., and Szaho, L. (1989) Rapid Commun. Mass Spectrom , Wilkinson, R. G., Gemski, P., Jr., and Stocker, B. A. D. (1972) J. 36. Volk, W. A., Galanos, Ch., and Luderitz, 0. (1970) FEES Lett. Gen. Microbiol. 70, , Della Negra, S., Deprun, C., and Le Beyec, Y. (1986) Reu. Phys. 37. Ikawa, M., Koepfli, J. B., Mudd, S. G., and Niemann, C. (1953) Appl. 21, SOC. Chem. J. Am. 75, McNeal, C. J., MacFarlane, R. D., and Thurston, E. L. (1979) 38. Takada, H., and Kotani, S. (1989) CRC Crit. Reu. Microbiol. 16, Chem. Anal. (6), 51, Jonsson, G. P., Hedin, A. B., Hakansson, P. L., Sundquist, B. U. 39. Le Dur, A., Caroff, M., Cbaby, R., and Szabo, L. (1978) Eur. J. R., Save, B. G. S., Nielsen, P. F., Roepstorff, P., Johansson, K.-E., Kamensky, I., and Lindberg, M. S. L. (1986) Anal. Chem. Biochem. 84, Karibian, D., Deprun, C., Szabo, L., Le Beyec, Y., and Caroff, M. 58, (1991) Internatl. J. Mass Spectrom. Zon Processes., press in E U

6 Cf Analysis PDMS of Endotoxin Preparations,,..

7 2s2Cf PDMS Analysis of Endotoxin Preparations 18549