Structural relationships and sialylation among meningococcal L1, L8, and L3,7 lipooligosaccharide serotypes.

Eighteen of 34 endemic meningococcal case strains were of the L8 lipooligosaccharide (LOS) type; four of these were both L3 and L7 (L3,7), and seven were L1. L1 structures arose by alternative terminal Gal substitutions of lactosyl diheptoside L8 structures, as determined by electrospray ionization and other mass spectrometric techniques, and enzymatic and chemical degradations (Structures L1 and L1a). [see text for structure] The more abundant molecule, designated L1, had a trihexose globosyl alpha chain; the less abundant one, designated L1a, had a beta-lactosyl alpha chain and a parallel alpha-lactosaminyl gamma chain. A P(k) globoside (Galalpha1-->4Galbeta1-->4 Glc-R) monoclonal antibody bound 9/10 L1 strains, but a P(1) globoside (Galalpha1-->4Galbeta1-->4GlcNAc-R) mAb bound none of them. alpha-Galactosidase caused loss of both L1 structures and creation of L8 structures; beta-galactosidase caused loss of the L8 determinant. The L1/P(k) glycose was partially sialylated. Some LOS also had unsubstituted basal beta-GlcNAc additions. These structural relationships explain co-expression of L8, L1, and L3,7 serotypes.

The outer membrane lipooligosaccharides (LOS) 1 of Neisseria meningitidis have been divided into 11 serotypes (1), but this system of antigenic discrimination does not adequately account for the observed molecular heterogeneity of these glycolipids (2)(3)(4)(5) nor for the fact that the several different LOS molecules made by each meningococcus may be of different serotypes. Monoclonal antibodies (mAbs) largely have replaced rabbit antisera as the LOS typing reagent (6,7), but a set of mAbs that recognizes most of the important LOS antigens is not available, and many of the available mAbs lack a complete structural definition of their cognate epitopes (8 -10). Furthermore, mAb LOS typing has been applied mostly to epidemiologically related strains (6,7), which may have introduced biases that have led to underestimates of the prevalences of some LOS types among circulating organisms. For these reasons we have used a variety of mass spectrometric and standard immunochemical techniques to investigate the structural basis of LOS serotype heterogeneity and mAb binding among meningococci that caused sporadic disease. We began with the L1, L8, and L3,7 types.
Neisseria LOS structurally resemble human glycosphingolipids (GSL) which often are sialylated (2,11,12). Group B and C N. meningitidis endogenously sialylate lacto-N-neotetraose LOS structures (13)(14)(15). We now report that the ␣ chain of the L1 LOS that mimics the human P k globosyl GSL also is partially sialylated and that L1 strains make a second, alternative LOS structure that we have designated L1a. Whether the L1a structure also is sialylated was not determined.
We used LOS made by Neisseria gonorrhoeae strain MS11 mk D to mark the positions of LOS in SDS-PAGE gels (21). We used LOS made by N. gonorrhoeae strains 1291, and its pyocin-selected mutants, and by MS11 mk D as LOS M r standards (22)(23)(24).
LOS Serotyping-We used the SPRIA inhibition technique developed by Zollinger and Mandrell (1), with the original polyclonal rabbit antisera, to determine the LOS serotype(s) of 34 consecutive and unique blood and cerebrospinal fluid N. meningitidis isolates from epidemiologically unrelated pediatric patients with endemic disease in Houston, TX, from February 1977 through March 1978 (18,19,25,26). This system arbitrarily sets 35% of the inhibition of a type-specific antiserum by the homologous prototype strain as the threshold criterion for a serotype designation (1). We then compared the L types of strains, defined by this criterion, with their binding of "serotype-specific" monoclonal antibodies (mAbs), as determined with use of a whole cell dot blot enzyme-linked immunosorbent assay as described (27).
Preparation of LOS-Bacteria were grown on supplemented GC agar in a CO 2 candle extinction jar, harvested, dehydrated with acetone, and stored at 4°C until ready for use. We extracted LOS from rehydrated acetone-powdered organisms by a modification of the hot phenol/water method (31,32).
Mass and Composition of the L1 and L8 LOSs-We estimated the mass of the L1 and L8 LOS by electrophoresing SDS-disaggregated LOS of meningococcal strain 126E, which makes both, with those of gonococcal strains, MS11 mk D and the 1291 series (3). MS11 mk D makes L8 and L3,7 LOS (21); 1291 b and 1291 c make L1 and L8 LOS, respectively (22), and 1291 wt makes an L3,7 LOS. We used mAbs to locate serotype-specific LOS in the gels.
The observed molecular weights of the (ϪKdo) oligosaccharides (OS) released by acid hydrolysis from LOS made by strain 1291 and its mutants are as follows: 1291 wt , 1679; 1291 a , 1517; 1291 b , 1476; 1291 c , 1314; 1291 d , 1152; and 1291 e , 990 (22). (The 1291 d and 1291 e mutants each make LOS with (ϪKdo) OS of molecular weights of 1152 and 990 but in different abundances (22). Although only ϳ20% of the molecules made by 1291 d have (ϪKdo) OS of M r 1152, as compared with ϳ50% of those made by 1291 e (22), we assigned the smaller molecule to 1291 e to avoid confusion); we added the mass of Kdo (220) to each. The mass of the OS of the "5.4-kDa" LOS made by MS11 mk D (21) was calculated by adding the mass of GalNAc (203) to that of the 1291 wt OS (22,24 The coefficient of correlation (r) between the electrophoretic mobilities of the gonococcal LOS and their derived molecular weights was 0.989. We used the function that described this correlation (M r ϭ (158.5 Ϫ rd)/0.029) to calculate an estimated molecular weight for each of the four 126E LOS bands in the SDS-PAGE gel and then calculated the nominal triply charged ESI-MS negative ion that LOS of each of these estimated molecular weights would yield after deacylation. We used the formula shown in Equation 1, where 760 is the mass of the ester-linked laurate residues and 42 the mass of the ␥ GlcNAc O-acetate groups that are removed during deacylation (10,36). We then sought each of these ions in the triply charged region of ESI-MS spectra of 126E LOS.
We analyzed samples of O-deacylated LOS by ESI-MS with a VG Bio-Q mass spectrometer with an electrospray ion source. Scans were taken in the negative ion mode over the m/z range of 300 -2000 at a resolution of ϳ500. LOS samples were dissolved in distilled water at a final concentration of 1-3 g/l, and 3-l aliquots were injected via a Rheodyne injector into a constant stream of H 2 O/acetonitrile (75:25%, v/v) containing 1% acetic acid as a solvent (36). A flow rate of 2-4 l/min was maintained throughout the analysis with an Isco syringe pump. Mass calibration used horse heart myoglobin as a reference, and the software was supplied by VG Bio-Q; the instrument was tuned in the negative ion mode with sulfated cholecystokinin-8 (Peninsula Laboratories). Spectra were averaged over several scans and digitally smoothed (36). For relatively abundant peaks the accuracy was Ϯ0.3 m/z; for poorly defined, low abundance peaks it was Ϯ0.4 -0.8 m/z. O-Deacylated LOS form deprotonated species that are preferentially in charge state z ϭ 3 and to a lesser extent in charge state z ϭ 2 (36). Triply charged molecular ions appear as doublets, 6 mass units apart, because ions that have lost a water molecule (m/z ϭ 18/3 ϭ 6) are formed by variable lactonization of a Kdo residue (Kdo 2 ) during ionization (36); lactonized ions are indicated in the text by an asterisk.
Neisserial LOS OS are composed of hexoses (Hex), N-acetylhexosamines (HexNAc), heptose (Hep), Kdo, sialic acid (NeuNAc), phosphoethanolamine (PEA), and an O-acetyl group. Two Hep residues, two Kdo residues, and the O-acetylated ␥-GlcNAc are conserved (2,13,14,22,24,36). We used a computer program to generate all compositions that would have a molecular weight that was consistent with that of each O-deacylated LOS molecular ion that was resolved sufficiently in the 126E LOS ESI-MS spectrum. We considered only those compositions that were comprised of known LOS constituents, contained two Hep residues, two Kdo residues, an O-deacylated GlcNAc residue, and an O-deacylated, diphosphoryl lipoidal moiety and that would yield a triply charged ion that was within 1 m/z of each observed ESI-MS ion (36).
Release of OS from LOS-OSs were released from LOS by hydrolysis with 1% acetic acid (100°C, 2 h) (9). We centrifuged the hydrolysate twice for 20 min at 4°C and 7000 rpm; the supernatant with released OS was lyophilized, redissolved in 52 mM pyridinium acetate (5-10 mg/ml), and separated by chromatography through 180 cm of Bio-Gel P-4 (Ͻ400 mesh) equilibrated with 52 mM pyridinium acetate. The eluate was monitored by refractometry and collected in 1.67-ml fractions.
Dephosphorylation-We dephosphorylated LOS and OS by hydrolysis with HF. LOS and OS were reacted with a 48% (aqueous) HF solution (10 g of LOS per l) in the dark at 4°C for 16 -20 h (OS and O-deacylated LOS) or 36 -40 h (intact LOS) (9). After evaporation of HF, the residues were resuspended in H 2 O and lyophilized.
Carbohydrate Composition Analysis-For composition analysis of neutral sugars, we dissolved 20 g (ϳ10 nM) of dephosphorylated OS in 200 l of H 2 O, added 200 l of 4 M trifluoroacetic acid, and heated the mixture for 4.25 h at 100°C (40). For quantitation of HexNAcs, we substituted concentrated HCl for 4 M trifluoroacetic acid (40). We evaporated the hydrolysates to dryness in a Speed-Vac concentrator, and we carried out monosaccharide separation and quantitation with use of high pH anion-exchange chromatography with pulsed amperometric detection, as described (39 -41). To elute the monosaccharide components, the gradient was modified slightly as follows: (i) 20 mM NaOH for 22 min, (ii) linear to 50 mM NaOH in 10 min, (iii) linear to 100 mM NaOH and 100 mM sodium acetate in 3 min, and (iv) linear to 160 mM sodium acetate in 15 min (with NaOH kept constant at 100 mM). A monosaccharide mixture that contained known quantities of fucose, GalNH 2 , GlcNH 2 , galactose (Gal), and glucose (Glc) (Dionex, Sunnyvale, CA) was used as a standard. The authentic monosaccharide liberated from the OS of Salmonella typhimurium Ra mutant LPS was used as a standard for the identification of L-glycero-D-manno-heptose (41).
Carbohydrate Sequence and Linkage Analysis-We used a combination of liquid secondary ion (LSIMS) and tandem (MS/MS) mass spectrometry and methylation analysis to confirm compositions and determine the sequences and linkages of the component sugars (13,22). Oligosaccharides were dissolved in H 2 O, and aliquots were transferred to a stainless steel probe tip that had been charged with approximately 1 l of a glycerol/thioglycerol (1:1) matrix. O-Deacylated HF-treated LOS were first dissolved in H 2 O/methanol (1:1). Samples for LSIMS were analyzed with a Kratos MS-50 double focusing mass spectrometer operating at a mass resolution of 1500 -2000 (m/⌬m, 10% valley). Samples were sputtered and ionized with a Cs ϩ ion primary beam of 10 keV; secondary ions were accelerated at 6 -8 kV. Scans were acquired at 300 s/decade (22). Ultramark 1621 was used manually to calibrate spectra to an accuracy of better than Ϯ0.2 Da.
MS/MS spectra were gotten with a four-sector Kratos Concept II HH mass spectrometer with an optically coupled diode array detector on MS-2 (22). Spectra were acquired in the negative mode; a Cs ϩ ion primary beam energy of 18 keV was used to sputter and ionize the oligosaccharides. The deprotonated molecular ions (M Ϫ H) Ϫ were selected in MS-1 and passed through to the helium collision cell floated at 2 keV, and pressure was adjusted to attenuate the precursor molecular ion by approximately 65%. Fragment ions were separated in MS-2 with use of a magnet jump with successive 4% mass windows and a constant B/E ratio. Fragment ions were detected on a one-inch optically coated diode array detector, and masses were assigned automatically by a Kratos Mach 3 data system that used CsI as an external reference for both MS-1 and MS-2.
The mass spectra were interpreted on the basis of the established structures of gonococcal LOS that bind mAbs 2-1-L8 (23) and 9-2-L379 (24). Nominal masses, based on the isotopically pure 12 C component of the natural isotopic distribution, are given for ions.
Sialic acid residues were removed from LOS by treatment with neuraminidase (13). We suspended 50 g of purified LOS in 25 l of phosphate-buffered saline (pH 6.0) and mixed the suspension with an equal volume of the same buffer that also contained 50 milliunits of active enzyme or the same buffer that contained 50 milliunits of enzyme that had been inactivated at 100°C for 5 min. We incubated the reaction mixtures at 37°C for 2 h and then stopped the reaction by adding 50 l of SDS-PAGE sample buffer (34).
We suspended ␣and ␤-galactosidases and ␤-N-acetylglucosaminidase in 200 mM sodium acetate buffer of pH 4.5; ␣-galactosidase and ␤-N-acetylglucosaminidase were predialyzed in the buffer to remove ammonium sulfate in the enzyme preparations (8). We added appropriate amounts of each enzyme in NaAc buffer (10 l; 2 units) to a 20-g solution of LOS in 40 l of 0.2% sodium taurodeoxycholate (Sigma). We kept the reaction mixture at 37°C for 16 h and then inactivated the enzymes by boiling the mixture for 5 min. We analyzed the effects of the enzymes by SDS-PAGE and immunoblot (9).

Identification of the L1 LOS-
The three reference L1 strains each shared two LOS components; the faster migrating (lower) bound mAb 17-2-L1 (Fig. 1). This LOS was the next higher 126E LOS (LOS 3 ) from the LOS that bears L8 (second component from the bottom, or LOS 2 ). The N. lactamica strain, 1207-0, bound mAb 17-2-L1 to an LOS that was slightly smaller than the meningococcal L1 LOSs. N. meningitidis strains 8529 and 8532, which are not L1, did not make LOS of these electrophoretic mobilities and did not bind mAb 17-2-L1. Strain 190I made two additional and higher molecular weight LOS, the smaller of which bound the L3,7 mAb (Fig. 1). We assigned the L1 epitope to LOS 3 of strain 126E.
LOS Serotypes of Endemic Strains (Table I)-Each of the 34 endemic strains made some amount of more than one "typespecific" LOS; only one was of a single SPRIA inhibition L-type (L2). L-type was not a uniformly distributed attribute. L5 was not found, and 25 strains were L3 (1), L7 (2), or both (22). L-type also was not an independent attribute (degrees of freedom ϭ 11, 2 ϭ 19.9 -43.5, p ϭ 0.05 to Ͻ0.0001 for the independence of L1, L2, L3, L4, L7, and L8 from all other L-types; degrees of freedom ϭ 11, 2 ϭ 18.3, p ϭ 0.07 for the independence of L6 from all other L-types).
Identification of the L1 and L8 LOS in ESI-MS Spectra-Since strain 126E is both L1 and L8, we used its LOS to compare the L1 and L8 structures. Fig. 3 shows the electrophoretic mobility of the four visible 126E LOS components relative to those of the gonococcal LOS molecular weight "standards," and Table II gives their estimated molecular weights. A comparison of the triply charged, deacylated ESI-MS ion that LOS of each of these molecular weights would yield and the corresponding observed ion in the ESI-MS spectrum (Fig. 4) also are given in Table II. In general, there was good concordance between the calculated and observed ESI-MS ions. Ions that were within 6 m/z of the calculated values for the two fastest migrating SDS-PAGE LOSs, LOS 1 and LOS 2 , were easily identified (A and B, respectively); they were the highest abundance ions in their respective regions of the spectrum. The most abundant ion in the entire spectrum, C, had the closest mass to that expected for the most abundant 126E LOS, LOS 3 (862 m/z versus 844   L1  L2  L3  L4  L5  L6  L7  L8   L 1  7  0  3  0  0  0  3  7  L 2  0  8  1  4  0  3  2  0  L3  3  1  23  6  0  4  22  14  L4  0  4  6  10  0  5  7  2  L 5  0  0  0  0  0  0  0  0  L 6  0  3  4  5  0  7  6  1  L7  3  2  22  7  0  6  24  14  L8  7  0  14  2  0  1  14  18 m/z). The expected ion for the highest mass, or slowest migrating, 126E LOS, LOS 4 (889 m/z), was the same as that of the D ion in the ESI-MS spectrum (889.1 m/z), but the D ion was present in much lower abundance than would be expected for LOS 4 , and its mass was less than that of either the E or F ion, both of which were present in equal or greater abundance than the D ion. The F ion (959.1 m/z) was the highest mass ion that was present in good abundance; its ϩ97 m/z ϭ 3 suggested that it arose by the addition of sialic acid (M r 291) to the 862.2 m/z C ion. By assigning the F ion to LOS 4 , the relative abundances of the A, B, C, and F ions matched the relative densities of the four 126E LOSs that were visible in SDS-PAGE. The compositions of the A, B, C, and F ions were readily deduced from the ESI-MS spectrum (Fig. 4), and the compositions of their oligosaccharides are given in Table II. The composition of the OS of the B, or LOS 2 , ion, was consistent with the lactosyl diheptoside L8 structure. The C ion (LOS 3 ; L1) was a single Hex residue larger than the B ion (Hex) 3 ; the A ion (LOS 1 ) was a Hex residue smaller (Hex) 1 . The F ion (LOS 4 ) was the sialylated form of the C ion.
Resolution of the ESI-MS Spectrum (Fig. 4)-The ESI-MS spectrum contained several prominent ions in addition to those that could be assigned to the four LOS that were visible in SDS-PAGE and many other low abundance and sometimes incompletely resolved ions. The multiple molecular species could be resolved into series, each of which had the same pattern of one, two, or three hexose constituents as the ABC molecular series, but each of which had different basal constituents (Table III). The (Hex) 3 molecule of each series usually also was present as a sialylated molecule.
The most abundant ABCF series had a single HexNAc residue ((HexNAc) 1 ) and one PEA residue ((PEA) 1 ). The (Hex) 3 D (883.1* m/z) and H (979.9* m/z) ions lacked a basal PEA ((PEA) 0 ) but had a second HexNAc residue ((HexNAc) 2 ); the H ion arose by sialylation of the D molecule (ϩ97 m/z). The Ϫ54 m/z (Hex) 2 and the Ϫ108 m/z (Hex) 1 ions of this second prominent series were present at 829.4* and 781.6 m/z, respectively, albeit in low abundances. The presence, at 727 m/z, of a very low abundance of (Hep) 2 (HexNAc) 2 (Kdo) 2 ion in this series confirmed that the second HexNAc was a basal addition. The relative abundances of the four largest molecules of the ABCF and DH basal series were quite similar, as judged by signal height. The sialylated F and H ions accounted for 30 and 26%, respectively, of their respective molecular series and their nonsialylated (Hex) 3 precursors, C and D, for 49 and 47%, respectively.
Although (HexNAc) 2 molecules were most abundant as the (PEA) 0 series, they also were found with one or two PEA residues (Table III and Fig. 4). Ions representing the (HexNAc) 2 (PEA) 1 series were present at 923.5* m/z ((Hex) 3 ) and 1027.6 m/z (NeuNAc3(Hex) 3 ); the G ion (963.6* m/z) was the (Hex) 3 molecule of the (HexNAc) 2 (PEA) 2 basal series. Smaller ions in the latter series were not resolved from ions of more abundant series. The E ion could have been either the sialylated (Hex) 3 molecule of the (HexNAc) 1 (PEA) 0 basal series or the (Hex) 2 ion of the (HexNAc) 2 (PEA) 2 series (Table III); desialylation (see below) showed it to be the former. The (Hex) 3 ion of this series was at m/z 815.
LOS Sialylation-We confirmed that L1 LOSs were partially sialylated by treating those of various strains with neuraminidase (Fig. 5). Neuraminidase degradation enhanced the electrophoretic mobility of the largest LOS made by each of the L1 N. meningitidis strains and increased the densities of their L1 LOSs. Strain 190I partially sialylated both its M r 3338 L1 LOS and its M r 3595 L3,7 LOS. ESI-MS analysis showed that neuraminidase degradation of 126E LOS caused the loss of the E, F, and H ions, as well as that of the 1027.6 ion (Fig. 6), thus confirming that these ions arose by sialylation of lower mass precursors. A, B, C, and D ions were preserved.
Analysis of Released Oligosaccharides-OS released from 126E LOS by mild acid hydrolysis contained galactose and glucose in the molar ratio of 3:2 and GlcNAc as the sole Hex-NAc. The 3:2 Gal/Glc ratio in the mixture of OSs is consistent with the addition of galactose to lactosyl (Gal␤134Glc) molecules. This was confirmed by methylation analysis of the re- leased OSs that showed the presence of Gal and GlcNAc as terminal non-reducing residues and of 1,4-Glc and 1,4-Gal as internal hexose residues. These linkages would be expected if Gal residues were added 134 to the now internal Gal residue of the lactosyl moiety of (Hex) 2 L8 molecules to form (Hex) 3 L1 molecules. Most of the Hep residues were tri-linked, as expected (Hep 1 , 1,3,4-linked; Hep 2 , 1,2,3-linked), but 1,2-Hep from (PEA) 0 basal series molecules was present in low abundance.
The released OSs eluted from a P-4 polyacrylamide column as two peaks of equal volume (7.5 ml); each 1.67-ml fraction was analyzed by LSIMS. A (ϪH) molecular ion of mass 1475 was present in all nine fractions. The composition of this ion,   (Fig. 3) in an ESI-MS spectrum (Fig. 4)    native and dephosphorylated ions revealed the presence of two isobaric (Hex) 3 ions, shown as a composite M r 1638 molecule with the fragmentation sites shown in Fig. 7. The more abundant M r 1476 molecule partitioned into the first P4 volume only; its three Hex residues were in sequence (Hex3 Hex3 Hex3 Hep 1 ), as shown by the presence of 1,5 A/X 3␤ (m/z 458/1017), Y 3␤ (m/z 988), and Z 3␤ (m/z 972) fragmentation ions in MS/MS spectra (Fig. 7). Its O-acetylated HexNAc residue was linked to C 2 of the second Hep residue (B/Y 1␣ fragmentation ions; m/z 246/1229), which was phosphorylated on C 3 ( 1,3 A 2␣ (m/z 427) and 1,5 A 2␣ (m/z 531) fragmentation ions). We designated it L1. Both P4 volumes contained a second m/z 1475 ion, in lower abundance, that had two Hex residues linked to Hep 1 in se-quence and a third Hex residue linked to C 4 or C 6 of its O-acetylated HexNAc residue, as shown by the presence of 0,2 A/X 2␣ (m/z 323/1152), 1.3 A/X 3␣ (m/z 589/886), and 1,5 A/X 3␣ (m/z 693/782) fragmentation ions that each differed by m/z 162 from their respective L1 A/X ␣ fragmentation ions (Fig. 7). We designated this molecule, with an elongated ␥ chain, L1a.
Glycosidase Degradations-Hydrolysis of 126E LOS with ␣-galactosidase caused the total collapse of LOS 3 into LOS 2 and the loss of the L1 epitope (Fig. 8). The now more abundant LOS 2 bound mAb 2-1-L8. Degradation of strain 6940 LOS with ␣-galactosidase also caused the total loss of its L1 LOS and the appearance of a very prominent, and previously inapparent, LOS that bound mAb 2-1-8. These results confirmed that all components of LOS 3 arose by the addition of ␣-Gal residues to LOS 2 molecules and that L1 is determined by an ␣-Gal substitution of the L8 structure.
Addition of Gal␣134 to an LOS lactosyl moiety would create an ␣-OS with the same structure as that of the human P k GSL oligosaccharide. A P k mAb bound strains 126E, 190I, 6940 and the six Houston L1 strains that bound mAb 17-2-L1. In contrast, only strains 126E, 6940, and 8022 (93% L1 SPRIA inhibition), and the gonococcal 1291 b LOS, bound a Gal␣134Gal-R mAb as well as they bound the P k mAb, and only strain 1291 b bound a P 1 mAb (Gal␣134Gal␤134GlcNAc-R). These results confirm the Gal␣134Gal␤134Glc structure of the L1 LOS ␣ chain and that it is this structure that binds mAb 17-2-L1.
␤-Galactosidase hydrolysis of 126E LOS caused the partial loss of LOS 2 and a marked increase in LOS 1 (Fig. 8), thus confirming that LOS 2 terminates with a ␤-Gal residue. The binding of mAb 2-1-L8, however, was not affected visibly, either because hydrolysis was incomplete or because the L1a molecule, which would run as LOS 2 after removal of the terminal ␤-Gal residue from its ␣ chain (Table IV), bound the mAb.
␤-Galactosidase degradation of strain 6940 LOS caused the partial loss of LOS 4 and LOS 3 and the appearance of prominent LOS 2 and LOS 1 molecules, such that the treated 6940 LOS had the same SDS-PAGE profile as the native 126E LOS. Thus, the principal difference between the LOS of these two L1 strains appeared to be the presence of terminal ␤-Gal residues on higher molecular weight LOS of strain 6940. Treatment of this same LOS with ␤-galactosidase after prior hydrolysis with ␣-galactosidase caused the appearance of a much more prominent LOS 1 , as would be expected if the ␣-galactosidase removed a terminal ␣-Gal residue from an L1 molecule and uncovered a precursor ␤-Gal residue (Fig. 4).
Hydrolysis of 126E LOS with ␤-N-acetylglucosaminidase caused a marked increase in the density of LOS 2 and in its binding of mAb 2-1-L8. Since GlcNAc was the only HexNAc detected in 126E LOS, the second HexNAc in the (HexNAc) 2 basal series must be a terminal GlcNAc as the ␤-anomer. ␤-N-Acetylglucosaminidase degradation of strain 6940 LOS caused the appearance of a new LOS 2 and a reduction in the densities of higher molecular weight LOS, some of which also must have been (HexNAc) 2 basal series molecules with terminal ␤-GlcNAc residues.
Structures for the different basal series L8 and the two alternative L1 molecules could be derived from these data, and published structures and are given in Table IV. DISCUSSION These studies reveal both quantitative and qualitative differences in LOS immunotype expression and find that meningococal LOS are more heterogeneous, and LOS serotyping less  Fig. 4. The presumptively sialylated E, F, and H ions are lost after neuraminidase treatment; the A-C and D ions remain. The fact that neuraminidase treatment caused the heights of the C and D ions to increase proportionately relative to those of the A and B ions confirmed that the former were partially sialylated. The position of the G ion is marked in the desialylated spectrum; it was not in sufficient abundance to be regarded as a signal. discriminatory, than previously thought.
To our knowledge, this is the first application of LOS serotyping to strains that caused discrete, sporadic disease during a prolonged period in a single location and that were of diverse protein serotypes (19,26,43). With use of mAbs, Jones et al. (6) found that 97% of hyperendemic ET-5 complex strains isolated from cases in the United Kingdom expressed the L3,7 immunotype and that only 13% of these invasive "L3,7,9" strains also expressed "L1, 8,10." In contrast, 70% of contemporaneously carried but non-invasive ET-5 complex strains expressed the L1,8,10 immunotype, and only 24% expressed L3,7,9 alone. We found that L3 and L7 also were expressed commonly by the diverse strains (19,26,43) that cause endemic disease (25/34; 74%), but these endemic L3,7 strains always expressed other LOS types, as well, including L8 (14) and L1 (3).
The structural basis for the association between L8, L1, and L3,6,7 is readily apparent, as the latter two arise by alternative, and mutually exclusive, additions to the lactosyl diheptoside L8 structure as follows: L1 by the addition of an ␣-Gal residue to create a globoside-like structure, and L3,7 by the addition of lactosamine to form the lacto-N-neotetraosyl oligosaccharide of paragloboside. Thus, L8 structures act as "toggles" or biosynthetic decision points. Some strains only add ␣-Gal (126E, strain 6940), whereas others only add ␤-GlcNAc (strain 8529 and others), and still others add both (190I and 8022). Whether a strain types as L8 depends on how many of its L8 LOS molecules remain unsubstituted. This biosynthetic relationship between lactosyl, globosyl, and paraglobosyl molecules also has been noted among gonococcal LOS (22)(23)(24).
The 126E L1 LOS structure was found previously with use of NMR spectroscopy (37), but our application to this strain's LOS of the mass spectrometric techniques that we developed to structure gonococcal LOS (22,36,44) enabled us to appreciate additional structures. It is reasonable to expect that still more structures will be found as more neisserial LOSs are studied. For example, terminal ␤-Gal residues on higher molecular weight LOSs of strain 6940 were not expected and are not explained by known meningococcal LOS structures.
ESI-MS proved to be particularly useful (36). By electrophoresing 126E LOS with gonococcal LOS whose structures have been established (21)(22)(23)(24), we could estimate the molecular weights of 126E LOS components with sufficient accuracy to align their SDS-PAGE profile with their ESI-MS spectrum and gain structural information about SDS-PAGE-visualized components. SDS-PAGE resolved 126E LOS into only four components, even with use of longer gels to maximize resolution (45); ESI-MS resolved these same LOS into more than 20 molecules of various abundances that had six different basal regions, four ␣ chains, and an extended ␥ chain.
The LPS made by Salmonella isogenic mutants that we used as molecular weight standards in the past (3) systematically overestimate the molecular weight of neisserial LOS, presumably because of the different branching patterns of the two sorts of glycolipids (42). Our previous estimates of the size of the L1 and L8 LOSs were too great by ϳ610 (4000 versus 3391) and ϳ370 (3600 versus 3229), respectively (3,5). We now can replace the LPS-based estimates of the size of the most abundant LOSs that can be visualized in SDS-PAGE with more exact masses.
The different PEA basal series were not unexpected (46). The site of the second PEA substitution is not known, but meningococcal LOSs may have single PEA substitutions at C 6 and C 7 of Hep 2 (47)(48)(49), so one of these two sites may bear the second PEA residue. Kulshin et al. (50) reported that a diphosphoryl  (22) proposed by Domon and Costello (54) in which ␣ and ␤ denote the precedence of branching, not chain length. As Hep 2 is attached to C 3 of Hep 1 , it precedes the C 4 -attached ␤-Glc residue and defines the ␣ branch; this usage is distinct from the use of ␣, ␤, and ␥ to denote the three glycose chains that can arise from the basal region of neisserial LOS, a usage that was based on the original Domon and Costello (54) proposal rather than precedence of branching (2). Assignment of A-C fragments is for the composite structure that would have two triglycose branches and an m/z of 1637. The A-C designations of fragments from the diglycose branches of the two m/z 1475 ions would be reduced by 1 (e.g. B 3␤ 3 B 2␤ ). Spectra were obtained for precursor ions found in each of nine 1.67 ml of OS-containing P-4 volumes before and after HF hydrolysis to remove PEA residues. Nominal masses for precursor and fragmentation CID fragments were used. meningococcal LOS lipoidal moiety was substituted at both ends with PEA. Lipid PEA substitutions would have confounded interpretation of the ESI-MS spectra, but we have found no evidence of pyrophosphoryl groups on deacylated neisserial LOS.
Neither the presence of an isobaric L1a molecule nor that of a ␤-GlcNAc basal region addition was expected. Di Fabio et al. (37) reported that the OS released by acid hydrolysis from their preparation of strain 126E LOS eluted from 90 cm of P-4 in a single broad peak that could be narrowed by deacylation with NH 4 OH and that strain 126E made only a single LOS molecule with the (GlcNAc) 1 (PEA) 1 L1 structure. One explanation why Di Fabio et al. (37) missed the heterogeneity of 126E LOS would be that NMR spectroscopy is not sensitive to the presence of structurally related oligosaccharides in a mixture (51). In addition, we were able to resolve 126E OS into two fractions with use of 180 cm of P-4.
The L1a structure was confirmed by the presence in tandem spectra of A/X ␣ fragmentation ions that differed by m/z 162. That these were ␣ and not ␤ fragments was clear from the difference in mass between in-line Hex (m/z 162) and OAcHex-NAc (m/z 245) residues and the contribution of the O-acetate moiety (m/z 42) to the 0,2 A/X 2␣ fragment. The site of Gal substituion of the ␥-GlcNAc could not be unambiguously assigned in the absence of 2,4 A 2␣ fragments, but as in previous studies (22), tandem spectra confirmed O-acetylation of C 3 of the ␥-GlcNAc, which leaves only C 4 and C 6 as possible sites. Methylation analysis did not resolve this question, as di-linked GlcNAc was not seen, presumably because of the relatively low abundance of L1a molecules. C 6 substitutions of neisserial LOS Hex residues have not been found, whereas C 4 substitutions are common, so we assigned the Gal residue to C 4 . This assignment is consistent with the unambiguous assignment of the L1 additional Hex to C 4 of the preceding Hex.
Extensions from the ␥ GlcNAc, although predicted (2), have not been reported previously. Because the L1 complexes of both 126E and 6940 were entirely degraded by ␣-galactosidase, we assigned the ␣-anomeric configuration to the additional Gal residue of the L1a structure. This would create an ␣-lactosaminyl ␥ chain in parallel with a ␤-lactosyl ␣ chain (Table IV and Fig. 7).
The linkage of the basal region ␤-GlcNAc could not be determined because of the low abundance of the m/z 1555 LSIMS ion. C 3 of Hep 2 normally is substituted, either with PEA (14, 22-24, 37, 48) or an ␣-Glc residue (42,47,49), so this would be a likely site of ␤-GlcNAc substitution, as well. This supposition is supported by the absence of a PEA residue on the most abundant (HexNAc) 2 basal series and the predominance of 1,3,4-and 1,2,3-tri-linked Hep residues. We presume that phosphoethanolaminylated (HexNAc) 2 molecules had Hep 2 C 6 or C 7 PEA substitutions (48,49), but such species were in such low abundance that their tetra-linked Hep residues would have been missed.

L8
␣-OS Gal␤134Glc␤134Hep␣13(Kdo) 2  gardless, sialogloboside-like LOSs now can be added to sialoparagloboside-like ones in the list of human glycoconjugates that neisserial LOS mimic. Strain 190I, an epidemic strain that expresses a unique bactericidal serotype (19), made both L1 and L3,7 LOS and partially sialylated both molecular species. It will be interesting to learn whether the same enzyme sialylates both molecules and, if so, whether the kinetics of sialylation differ.
Strains 126E and 6940 LOS had very different SDS-PAGE profiles, primarily due to the absence of LOS 1 and LOS 2 (L8) from the LOS of strain 6940. Both strains used basal region ␤-GlcNAc substitutions, but strain 6940 also used a ␤-Gal substitution that was not used by 126E. Treatment of strain 6940 LOS with ␤-galactosidase resulted in an SDS-PAGE profile that was nearly the same as that of 126E LOS. Solving strain 6940 LOS structures will require additional analyses.
In summary, strains of N. meningitidis use different kinases and glycosyltransferases to substitute lactosyl diheptoside LOS in a variety of ways, some mutually exclusive and some not, that create a set of structurally related glycolipids. The use of multiple sets of modifying enzymes by a strain results in expression of more than one LOS type. When multiple LOS types are expressed, one may predominate, reflecting preferential use of the responsible enzymes. Whereas the relative abundances of the different LOS that an individual strain can make is a fairly stable attribute of laboratory grown organisms (6,52), they can vary greatly among different strains of the same L-type. Thus LOS type is neither a stable nor discrete phenotype.
How these organisms express LOS in vivo, and whether certain LOS molecules that are uncommonly made by in vitro grown organisms are necessary for pathogenesis, remains to be explored.