262Cf Plasma Desorption Mass Spectrometry Applied to the Analysis of Underivatized Rough-type Endotoxin Preparations*

Plasma desorption mass spectrometry has recently been used with success to characterize native, underivatized Re- to Re-type endotoxins in terms of their constituent lipopolysaccharides. The spectra give masses for the major molecular species of lipopolysac- charide present from which their probable compositions could be inferred using the overall composition determined by chemical analyses. Moreover, the rela- tive intensities of the signals are roughly proportional to the abundance of their corresponding molecular spe- cies. Native Rc-, Rb-, and Ra-type enterobacterial endotoxins with 5-10 core sugar units have been ren- dered amenable to plasma-desorption mass spectrometry analysis by improvement in their solubility and the use of cellobiose as an additive. The spectra of four Salmonella and Escherichia endotoxin preparations demonstrated heterogeneity in acylation and phos- phorylation. Since these sources of heterogeneity are critical for many biological activities, the spectra underline the need to define the composition of each preparation of endotoxin used in structure-function studies. Endotoxins

Plasma desorption mass spectrometry has recently been used with success to characterize native, underivatized Re-to Re-type endotoxins in terms of their constituent lipopolysaccharides. The spectra give masses for the major molecular species of lipopolysaccharide present from which their probable compositions could be inferred using the overall composition determined by chemical analyses. Moreover, the relative intensities of the signals are roughly proportional to the abundance of their corresponding molecular species. Native Rc-, Rb-, and Ra-type enterobacterial endotoxins with 5-10 core sugar units have been rendered amenable to plasma-desorption mass spectrometry analysis by improvement in their solubility and the use of cellobiose as an additive. The spectra of four Salmonella and Escherichia endotoxin preparations demonstrated heterogeneity in acylation and phosphorylation. Since these sources of heterogeneity are critical for many biological activities, the spectra underline the need to define the composition of each preparation of endotoxin used in structure-function studies.
Endotoxins play an important role during infections of animals and man by Gram-negative bacteria. They cause such symptoms as fever and endotoxic shock, and they also activate the immune system. Each endotoxin is an ensemble of related lipopolysaccharides (LPS)' the structures of which depend on the bacterial source and its culture conditions (1)(2)(3). The LPS of most endotoxins contain a highly conserved lipid moiety called lipid A (4), consisting of bisphosphorylated diglucosamine substituted by three to seven hydroxylated and nonhydroxylated fatty acids. The lipid moiety is attached via a Kdo unit to a "core" oligosaccharide of 8 to 12 sugar units to which may be linked a polysaccharide chain of variable length (the 0-antigen) which consists of repeating sugar units and characterizes the serotype. Molecular weights of the LPS, as * This work was supported by the Universiti. de Paris-Sud (BQR 1991). 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. This paper is dedicated to the memory of Professor Bernard Fournet (University of Lille, France).
More recently, with improved sample preparation, native, underivatized endotoxins having up to five core sugar units (Rc-type) have given interpretable PDM spectra (12). These spectra give signals for the major LPS molecular species whose probable compositions can be inferred from the overall endotoxin composition determined by chemical analyses. Appearing downfield in the spectra are signals that correspond to the lipid A produced by rupture of the weak Kdo ketosidic bond linking it to the LPS core region. These signals testify generally to the heterogeneity that is due to variations in fatty acid composition (13) as well as to substitutions on the lipid A phosphate groups. The two sets of signals (LPS molecular-ion and lipid A fragment-ion) often resemble each other in the number and relative spacing and intensities of the signals. All this information is important not only for structural analysis but also for structure-function studies since considerable variation exists among the results reported from different laboratories regarding endotoxin structure and function. It was therefore essential to adapt PDMS techniques to elucidate the structure of endotoxins having a complete core. New conditions of sample preparation and detailed analysis of the resulting spectra obtained with Rdl-to Ratype endotoxins are presented here. The postulated structures of these endotoxic lipopolysaccharides are shown in Fig. 1.

EXPERIMENTAL PROCEDURES
Negative-ion spectra were obtained on a Depil ' "Cf time-of-flight mass spectrometer (TOF 21, IPN, Orsay, France) with a drift distance of 45 cm and an accelerating voltage of 15 kV (14). Counting time varied from a few minutes to several hours. Masses were determined from peak centroid calculation.
Endotoxins were purified by extraction of phospholipids, treatment with proteases and nucleases, and centrifugation until thin layer chromatography (15) and UV spectra showed no detectable contaminants.
Structured Bilayers-Using a Langmuir-Blodgett apparatus as previously described (19), bilayers of a mixture of endotoxin and hexadecanoic acid (1:l) were deposited onto a target of gold foil 0.075pm thick.
Dissolution Aids-To obtain clear solutions of endotoxins, the preparations were treated for various periods ( a ) in a Bransonic 220 ultrasound bath at room temperature; ( b ) at 90-100 'C; or (c) in a Moulinex Microchef FM 1515 microwave oven.
The method adopted for sample preparation was the following: 100 pg of purified endotoxin was added to 25 pl of 4% aqueous cellobiose solution and treated for 2 min in the microwave oven on position 11. 25 p1 of isopropyl alcohol-methanol (12) and a few grains of Dowex-50-8X Me,NH+ were added and stirred. Then 40 pl of the supernatant were electrosprayed onto an Al-on-mylar target and put into the mass spectrometer.

RESULTS AND DISCUSSION
The solubility of LPS in the chloroform-methanol-water solvent used in our earlier work is inversely proportional to the length of their core domain. Therefore, with LPS having four or more core sugar units, the first step was to use a more polar solvent. A mixture of isopropyl alcohol/methanol/water 1) The resin treatment using Dowex 50-Me2NH+ gave the best results and was more convenient and rapid than other methods of desalting endotoxins (see "Experimental Procedures").  including octyl p-glucoside, bovine serum albumin, and lactose, improved the spectra somewhat but cellobiose proved t o be the most promising.
3) The optimal proportions were established to be one part (by weight) of LPS to 10 parts of cellobiose.

4)
LPS-cellobiose solutions treated briefly with microwave irradiation gave better spectra than those treated with heat or ultrasound which are frequently used to facilitate solution.

)
T h e optimal thickness of sample on a target was determined by comparing spectra from targets electrosprayed with 10-80 pl of sample preparation.
On the basis of spectral quality, sample preparation as described under "Experimental Procedures" was adopted. The three somewhat stronger signals at the left are from the corresponding lipid A fragment ions. As has been noted before (12), the weak ketosidic bond of Kdo is the principal fragmentation point in endotoxins. It often gives rise to a pattern of lipid A signals ( m / z 1200-2100) mimicking that of the molecular ions (m/z > 2100). In this case, the lipid A peaks are 987 amu (mass of the pentasaccharide core consisting of 2 Kdo, 2 heptose, and 1 glucose unit) downfield from their respective molecular counterparts. The m/z 1797 signal, which can be attributed to the hexaacylated lipid A structure shown in Fig.  1, has been found in the spectra of all enterobacterial lipid A examined so far, as well as of some closely related genera (13).
The less-acylated LPS species giving signals at m/z 2347 and 2573 (Fig. 2) represent heterogeneity in the preparation and not fragmentation. This was determined from spectra of homogeneous LPS and synthetic lipid A preparations which give a single peak (not shown).
The improvement in the method can be seen by comparing the two spectra of the same endotoxin of S. minnesota R7 in Fig. 3. The upper spectrum ( A ) was obtained by the earlier method. It has few and poorly resolved molecular-ion signals   signal. This was attributed, as before (121, to a replacement of the tetradecanoate unit (C14, 210 arnu)' with a fifth hydroxytetradecanoate (C140H, 226 amu)'. Each of the four mentioned LPS signals ( m / z 2185, 2411.4, 2622, and 2637.3) is followed by another, 80 amu higher, indicating the presence of molecular species having a third phosphate.
The major Rb2-type lipopolysaccharides of S. minnesota R345 endotoxin (Fig. 4) ( m / z 3502, 3291, and 3067) differ from those of the Rc-type of E. coli F583 by 718 atomic mass units. This is consistent with their having two hexoses (two galactoses in this case, 162 amu each), one heptose (192 amu),* and a pyrophosphorylethanolamine (203 amu)' more than the Retype. The hexa-acyl molecular species is most abundant  in this preparation. The small signals at m / z 3720 and 3940 may represent molecular ions containing 3 and 4 Kdo units, respectively. It has been reported that the LPS of this strain contains 3 Kdo units and that those of other strains of S. minnesota contain nonstoichiometric amounts between 2 and 3 (22). We have analyzed the spectra of seven endotoxin preparations of other strains of this species, and this is the first evidence obtained by PDMS of LPS containing more than 2 Kdo units Even though some fragmentation is expected at the Kdo ketosidic bonds it is not obvious, a priori, that it would be more frequent than that between Kdo and the glucosamine of lipid A. This may be another case of variability between different endotoxin preparations from the same bacterial strain (12). In any event, it would seem prudent to define each endotoxin on the basis of analyses of several different preparations.
The lipid A part of the spectrum (Fig. 4) contains several peaks that are unaccountable by the usual LPS analyses. The peaks in question do not appear in the spectrum of lipid A isolated from this endotoxin preparation (not shown). They are apparently lost during the mild hydrolysis and/or extractions involved in the preparation of lipid A. This suggests that they are contaminants. Thin layer chromatography of the endotoxin supported this idea by revealing the presence of several compounds that migrate faster than the LPS in a hydrophobic solvent. Fig. 5 is the spectrum of an Ra-type (full core) endotoxin of S. typhimurium. The signals at m / z 4030, 3804, and 3595 correspond to the hexa-, penta-, and tetra-acylated lipid As with their 10-sugar-unit cores substituted with a pyrophos-M. Caroff phorylethanolamine and 2 more phosphate units (unplaced). The tetra-acylated species dominate the two parts (lipid A fragment-ion and molecular-ion regions) of the spectrum. Since artificial mixtures of different homogeneous lipid As gave signals roughly proportional to their abundance in the sample3 and, in this spectrum, the molecular-ion region resembles the lipid A fragment-ion region we conclude that the tetra-acylated LPS is most abundant in the preparation examined. Table I sums up the masses of the principal LPS molecular species detected in the four spectra, with the respective contributions from their lipid and saccharide domains. The less acylated species of LPS may represent incomplete synthesis or in vivo degradation of the endotoxin. If the former, the differences in mass values m / z tell us that, in strains R7 and R345 of S. minnesota and F583 of E. coli, the last and next-to-last fatty acid units added to the diglucosamine backbone are, respectively, tetradecanoic (210 amu) and hydroxytetradecanoic acid (226 amu), except, of course, when the LPS has no tetradecanoate (the case of some R7 LPS). In the S. typhimurium LPS examined, the order would be reversed since the penta-acylated molecular species ( m / z 3804) contains tetradecanoate.
The new method gave completely satisfactory results when applied to lipid A and to the Re-type endotoxin of S. flexneri.
However, since the previously described method is simpler and requires smaller quantities of material it is to be preferred for these more hydrophobic compounds. The spectra indicated, and chemical analysis confirmed, that none of the preparations had appreciable amounts of hexadecanoic acid in their lipid domains despite the idea current among many immunologists that this fatty acid is characteristic of Salmonella LPS and distinguishes them from E. coli LPS.
To explain the effect of the cellobiose-plus-microwave treatment on LPS solutions, light scattering and buoyant density measurements are being carried out. Preliminary results concur with the idea that the treatment causes considerable disaggregation of the LPS.