On the antigenic determinants of the lipopolysaccharides of Vibrio cholerae O:1, serotypes Ogawa and Inaba.

Monoclonal, murine IgG1s S-20-4, A-20-6, and IgA 2D6, directed against Vibrio cholerae O:1 Ogawa-lipopolysaccharide exhibited the same fine specificities and similar affinities for the synthetic methyl alpha-glycosides of the (oligo)saccharide fragments mimicking the Ogawa O-polysaccharide (O-PS). They did not react with the corresponding synthetic fragments of Inaba O-PS. IgG1s S-20-4 and A-20-6 have absolute affinity constants for synthetic Ogawa mono- to hexasaccharides of from approximately 10(5) to approximately 10(6) M-1. For IgG1s S-20-4, A-20-6, and IgA 2D6, the nonreducing terminal residue of Ogawa O-PS is the dominant determinant, accounting for approximately 90% of the maximal binding energy shown by these antibodies. Binding studies of derivatives of the Ogawa monosaccharide and IgGs S-20-4 and A-20-6 revealed that the C-2 O-methyl group fits into a somewhat flexible antibody cavity and that hydrogen bonds involving the oxygen and, respectively, the OH at the 2- and 3-position of the sugar moiety as well as the 2'-position in the amide side chain are required. Monoclonal IgA ZAC-3 and IgG3 I-24-2 are specific for V. cholerae O:1 serotypes Ogawa/Inaba-LPS.1 The former did not show binding with members of either series of the synthetic ligands related to the O-antigens of the Ogawa or Inaba serotypes, in agreement with its reported specificity for the lipid/core region (1). Inhibition studies revealed that the binding of purified IgG3 I-24-2 to Ogawa-LPS might be mediated by a region in the junction of the OPS to the lipid-core region of the LPS. cDNA cloning and analysis of the anti-Ogawa antibodies S-20-4, A-20-6, and 2D6 revealed a very high degree of homology among the heavy chains. Among the light chains, no such homology between S-20-4 and A-20-6 on the one hand, and 2D6 on the other hand, exists. For the anti-Inaba/Ogawa antibodies I-24-2 and ZAC-3, their heavy chains are completely different, with some homology among the light chains.

We used synthetic mono-to hexasaccharides that mimic the fragments of the O-antigen of Ogawa and Inaba O-polysaccharides (2)(3)(4), together with certain analogs of their monosaccharides to evaluate specificity. The binding of three immunoglobulins G (two specific for Ogawa and one specific for Ogawa/ Inaba) and of two immunoglobulins A (one specific for Ogawa and one specific for Inaba/Ogawa) were characterized by ligand-induced fluorescence titration or ELISA inhibition. The cDNA sequences of these antibodies are also presented in this report.

MATERIALS AND METHODS
Monoclonal Antibodies-Murine ascites fluids of A-20-6 and S-20-4 both contain vibriocidal IgG 1 specific for Ogawa-LPS. I-24-2, in contrast, contains IgG 3 specific for both serotypes Ogawa and Inaba-LPS, and it has low vibriocidal activity (5) (clone S-20-4 comes from the same hybridoma cells as clone S-20-3 described in this reference). Murine ascites fluid 2D6 and ZAC-3 contain IgA specific for Ogawa-LPS and IgA specific for Inaba/Ogawa-LPS, respectively. The latter two hybridomas were gifts from Drs. Marian Neutra, Harvard Medical School, and Dr. Richard Weltzin, Oravax, Cambridge, MA (1,6) and were grown in BALB/c mice. IgGs were purified using ImmunoPure® (G) IgG purification kits (Pierce). Briefly, ascites fluid (2 ml, clarified by centrifugation) was mixed with ImmunoPure® (G) binding buffer (2 ml) and applied to a protein G column. After washing the column with 5 ϫ 2-ml aliquots of the ImmunoPure® (G) binding buffer, the bound IgG was eluted with 6 ml of ImmunoPure® (G) elution buffer, dialyzed against PBS, pH 7.4 (2000 ml) for three changes at 0°C, frozen, and labeled. The purified A-20-6, S-20-4, and I-24-2 showed a single arc of precipitation versus goat anti-mouse IgG 1 and IgG 3 , respectively (heavy chainspecific), and goat anti-whole mouse serum (Sigma) by immunoelectrophoresis. IgAs were purified from ascites fluid by 40% ammonium sulfate precipitation and anion-exchange DEAE-Sephadex A-25 chromatography (7). Monomeric IgA was obtained by mild reduction with 5 mM 1,4-dithiothreitol (Sigma) and alkylation with 11 mM 2-iodoacetamide (Sigma), followed by re-adsorption of the sample on DEAE-Sephadex A-25 and elution with PBS, pH 7.4. The purity of IgAs was also verified by immuno-electrophoresis against anti-mouse IgA and serum and SDS-polyacrylamide gel electrophoresis.
Positive PCR products were purified on 1.0% low melting agarose gels using AgarACE enzyme (Promega) to extract the DNA and ligated to pCRII vectors as described by the manufacturer (Invitrogen). White colonies were selected and screened by restriction analysis for cloned inserts. The isolated V L and V H clones were grown in YT medium to produce phage, and the single-stranded DNA was purified (21). The DNA was then subjected to single-stranded DNA sequencing according to the protocol provided by the manufacturer (Sequenase, U. S. Biochemical Corp.).
For the hybridoma cells producing A-20-6, S-20-4, and I-24-2 antibodies, the total RNA was extracted by a modification of a known method (22). The reaction mixture (100 l) containing maximally 10 g of total RNA, 20 mM mercaptoethanol, 80 units of RNasin (RNase inhibitor, Promega) 1 ϫ first strand buffer (Life Technologies, Inc., Basel, Switzerland), 8 mM dithiothreitol, 0.1 g/ml primer p(dT) 15 (Boehringer Mannheim GmbH, Mannheim, Germany), 500 M each dNTP, and 400 units of Moloney murine leukemia virus reverse transcriptase (Promega, using their protocols) was incubated for 2 h at 42°C. The reaction mixture was then used directly for PCR amplification. For that, each cDNA mixture so obtained was divided into, respectively, eight (for the VH) and ten (for the VL) separate mixtures, each one of which was then treated with the appropriate, separate, back primers. A 50-l reaction mixture containing 2 l of the reaction mixture from the cDNA synthesis, 30  The back primers were designed to hybridize to the partially conserved sequences in the leader or the FR1 regions of the V H or V L . The forward primers corresponded to the N-terminal beginning of the hinge region and the C-terminal part of the C L region, respectively.
Direct sequencing of these PCR products was performed on an Applied Biosystem Apparatus ABI 373A (Genome Express, Grenoble, France) using dye terminators. Sequence data were analyzed and comparisons performed, with software from the Genetics Computer Groups, Inc. (Madison, WI) and the GeneBank™ (Los Alamos, NM) and EBI (Heidelberg, Germany) Fluorescence Titration-Purified immunoglobulins were diluted with PBS, pH 7.4, to solutions having an A 280 of 0.04. Protein solutions (1.1 ml) were added to each of two cuvettes, one for ligand addition, the other a reference. A third cuvette filled with 1.1 ml of PBS was used as a blank. Temperature was maintained at 25°C by a circulating thermostatic bath. Ligand solutions were verified to show no fluorescence effects by themselves by addition to a suitable nonbinding protein. To obtain a good distribution of points, ligand concentrations were adjusted so that the protein would be saturated after the addition of ϳ20 l of ligand solution to the test cell. Affinity constants (K a ) for the association between ligands and antibodies were obtained by monitoring the ligand-induced tryptophanyl fluorescence of the antibody as a function of ligand concentration (25,26). We used a Perkin-Elmer LS 50 luminescence spectrophotometer and where K a is the affinity constant, is the fraction of antibody sites bound to ligand (measured as the change in antibody fluorescence due to ligand addition, divided by the maximal ligand-induced fluorescence change), 1 Ϫ is the fraction of free antibody sites, and C L is the concentration of free ligand. A Scatchard plot of /C L versus gives a line whose slope is equal to the K a . Two representative Scatchard plots are shown in Fig. 2 percent inhibition was calculated as follows: (1 Ϫ A I ) ϫ 100 ϭ % inhibition.

Antigenic Epitopes on Ogawa-LPS-IgG 1 s, A-20-6 and
S-20-4, specific for Ogawa-LPS (vibriocidal titer: 1280), protect mice from mortal challenge with three times the LD 50 of V. cholerae O:1 serotype Ogawa strain 920139 (5). Thus, a study of their combining sites is of interest, as oligosaccharide fragments of the O-antigen that bind immunoglobulins maximally will be more likely, when linked to a carrier protein, to elicit antibodies reactive with the parent polysaccharides (28). Each synthetic mono-to hexasaccharide fragment of the Ogawa Oantigen bound the two antibodies, with binding constants of from ϳ10 5 to ϳ10 6 M Ϫ1 as measured by fluorescence titration ( Table I). The free energies of association for oligosaccharide with each of the two antibodies were close, indicating that these two vibriocidal antibodies have the same fine specificity for the epitope on the Ogawa O-PS. The terminal, nonreducing Ogawa-monosaccharide (1 O ) contributed ϳ90% of the maximal binding energy shown by the entire hexasaccharide. This finding differs from those in some other (homo)polysaccharideantibody systems (29), where monosaccharide binding accounts for only 50 -60% of the maximum binding energy that is shown by the binding of four to six sugars. Methyl 4,6-dideoxy-4-(3deoxy-L-glycero-tetronamido)-2-O-ethyl-␣-D-mannopyranoside (1 O -Et) showed an affinity constant for both these antibodies that was an order of magnitude less than that of the Ogawa monosaccharide. Neither the corresponding monosaccharide with a free OH group at C-2 (1 I ), nor its 2-deoxy derivative (1-2d), 3-deoxy derivative (1 O -3d), 2-O-propyl derivative (1 O -Pr), or the one deoxygenated at the 2Ј-position in the tetronic acid moiety (1 O -2d), exhibited binding. Removal of the primary hydroxyl group in the tetronic acid group gave (1 O -4d), which bound to IgGs A-20-6 and S-20-4 with higher affinity than did 1 O itself (Table I)   It is not unusual that small determinants can dictate an immune response in mice (30). Antigenic Epitopes on Inaba-LPS-Synthetic oligosaccharide fragments reflecting the structural difference between Ogawa and Inaba O-antigens were used to study the specificity of monoclonal IgA ZAC-3, obtained from a lymphocyte of the mouse's Peyer's patches following immunization with V. cholerae O:1 serotype Inaba (1), and monoclonal IgG 3 I-24-2, obtained following immunization of mice with a lysate from V. cholerae O:1 serotype Inaba (5). Others showed (1) that the IgA ZAC-3 dimer and polymer bound all the fragments (including the 3-4-kDa fragment) of V. cholerae LPS (Inaba), indicating that the determinant epitope for this IgA is located in the lipid A or core region, and not in the O-specific side chain. These workers also showed by microcalorimetric measurements that the K a for IgA ZAC-3 and detergent-solubilized LPS was ϳ6 ϫ 10 5 M Ϫ1 . Their proposed specificity of ZAC-3 for the lipid/core (1) would accommodate our results described below, namely that none of our Ogawa or Inaba (oligo)saccharides show binding to that IgA. IgG 3 I-24-2 is reported to possess weak agglutination titers against Ogawa and Inaba organisms and a weak vibriocidal titer against either organism (5). No ligand-induced fluorescence change was observed with antibody IgG 3 I-24-2 and either 1 O to 6 O or 1 I to 6 I or with IgA ZAC-3 and either 1 O to 4 O or 1 I to 4 I . The absence of binding of these ligands to these two (ZAC-3 and I-24-2) purified antibodies was verified by ELISA inhibition systems (antibody/Ogawa-or Inaba-LPS). Indeed, these saccharides also failed to show interaction with either antibody in the ELISA system employing either LPS-Ogawa or LPS-Inaba as the capturing agent, except for Inaba monosaccharide 1 I , which could moderately inhibit (50%) only the IgG 3 I-24-2/Ogawa-LPS system (see Table II), but not the ZAC-3/ Inaba-LPS or ZAC-3/Ogawa-LPS ELISA system. (It is puzzling to us why only the mono-and not the higher oligosaccharides of the Inaba series is able to inhibit the IgG 3 I-24-2/Ogawa-LPS system.) The binding of either IgA ZAC-3 or I-24-2 to ELISA plates, using either the LPS-Ogawa or LPS-Inaba as the capturing agent, could be inhibited by either LPS.
To elucidate the inhibitory effect by various other potential inhibitors in the I-24-2/Ogawa-LPS ELISA system we examined the ligands shown in Table II (9) consists of the core region that is flanked at the upstream end by the base-stable remnant of its short O-PS, namely two carbohydrate residues: a threo-hex-4-enuronopyranosyl residue linked to 2-amino-2,6-dideoxy glucosyl residue, while at the downstream end of the core two phosphorylated glucosaminyl residues remain. That product did not react with ascites fluid, but did weakly inhibit the interaction of purified IgG 3 I-24-2 with  Ogawa-LPS as the capturing reagent. O/R-DeOAc-Ogawa-LPS is a substance that has only its core destroyed by the periodate oxidation (borohydride reduction was executed to render the aldehydo groups formed chemically unreactive), since the core has vicinal hydroxyl groups (33). Its O-PS is unaffected by the periodate oxidation, as we showed by measuring its K a values with IgGs A-20-6 and S-20-4 (K a values of 0.5 ϫ 10 6 and 1.3 ϫ 10 6 , respectively), showing them to be nearly the same as the K a values these antibodies have with the hexameric O-PS saccharide 6 O . This O/R-DeOAc-Ogawa-LPS was able to inhibit IgG 3 I-24-2 on ELISA (see Table II) only 70%, whereas the LPS itself could do so nearly completely (90%). Since it is unlikely that the lipid plays a role, this could indicate a partial specificity of IgG 3 I-24-2 for the core, as that is the only part of the molecule that would be destroyed by periodate oxidation. Dextran 10T, D-glucose, methyl ␣-glucopyranoside, methyl ␤-Dglucopyranoside, D-mannose, and methyl ␣-D-mannopyranoside (the latter two are structurally related to the heptose in the core region) showed moderate to significant (50 or 70%) inhibition. The inhibition by D-galactose and methyl ␣-galactopyranoside also remains unexplained. Although from the above it does appear to recognize determinants in the core region of the LPS, an immunoblotting experiment (5), shows IgG 3 I-24-2 to bind only to the high molecular weight LPS from both Inaba and Ogawa V. cholerae O:1, and not to the low molecular weight fractions, presumably made up of lipid-core region only (34).
In summary, for the immunoglobulins with specificity for Ogawa/Inaba, the behavior of the IgG 3 I-24-2 indicates partial specificity for the core as well as for a single residue of 4,6dideoxy-4-(3-deoxy-L-glycero-tetronamido)-␣-D-mannopyranoside. Such a residue is linked, as the first of many in the O-PS, to the core in the Inaba-LPS. Nevertheless the behavior of IgG 3 I-24-2 remains somewhat puzzling. Our observations on IgA ZAC3 are entirely consistent with the specificity proposed by Lü llau et al. (1) for the LPS' lipid/core region. For the binding behavior of the three anti-Ogawa monoclonal antibodies, IgG 1 S-20-4, IgG 1 A-20-6, and IgA 2D6, our data consistently indicate the immunodeterminant to be the upstream, terminal residue of the Ogawa O-PS.
The above results were corroborated by the cDNA-derived amino acid sequences. It is clear from the V H sequences of the anti-Ogawa immunoglobulins (Fig. 4) that all three belong to the same family, as they show a high degree of homology. It is interesting that IgGs A-20-6 and S-20-4 have identical () V L regions, whereas IgA 2D6, possessing a L-chain, has a V L sequence that is therefore quite different. Since they show identical specificity patterns, that may be dictated more by the H-than by the L-chain. It is noteworthy that the IgGs A-20-6 and S-20-4 show significant ligand-induced tryptophanyl fluorescence change, while IgA 2D6 does not. The former two (IgGs) both possess a tryptophanyl residue at position L-91, while the latter (IgA) carries a glycine at that position. Tryptophan at that position of the L-chain has been correlated before with ligand-induced changes in the tryptophanyl fluorescence (29).
In comparing the cDNA sequences of IgA ZAC-3 and IgG 3 I-24-2 (Fig. 5), we observe the reverse. Here, the V H sequences differ extensively, while the V L regions show some degree of homology. It should be noted, however, that there, most of the differences occur in the hypervariable regions of the V L .