Unusual lipopolysaccharide antigens of a Salmonella typhi oral vaccine strain expressing the Shigella sonnei form I antigen.

Salmonella typhi 5076-1C, a potential live, oral vaccine for protection against typhoid fever and Shigella sonnei shigellosis, expresses the S. sonnei form I antigen and normal S. typhi somatic antigens. Polysaccharide antigens of this galactose epimeraseless genetic derivative strain were hot phenol-water extracted from cells grown with (+gal) and without (-gal) galactose. Ultracentrifugation of the aqueous layer from (+gal) cells resulted in a lipopolysaccharide (LPS) pellet having core-linked S. typhi O-antigen but no core-linked form I antigen; the LPS from (-gal) cells lacked O-antigen. The form I antigen, obtained from the supernatant, was purified by alcohol precipitation and ion exchange chromatography. Unlinked form I and S. typhi O-polysaccharide antigens, both present in the (+gal) supernatant, were further separated by gel filtration. Chemical analyses revealed the 5076-1C form I antigen to be a polymer (Mr = 14,000-20,000) having O-disaccharide repeating units comprised of 2-acetamido-4-amino-2, 4,6-trideoxy-D-galactose and 2-acetamido-2-deoxy-L-altruronic acid. Unlike parental S. sonnei form I LPS, the 5076-1C form I antigen lacked core lipid A, had low phosphorus content, and migrated in polyacrylamide gels with lower relative mobility. In contrast to current concepts of LPS assembly, these data indicate that 5076-1C form I antigen is transported to the cell surface without covalent linkage to core lipid A, and exists as a polymerized, antigenic surface entity.

2 To whom correspondence and reprint requests should be addressed.
The abbreviations used are: LPS, lipopolysaccharide; Ps, polysaccharide; galE, galactose epimeraseless; KDO, 3-deoxy-~-mnno-octulosonic acid; Hep, L-glycero-D-manno-heptose; 4-n-DFucNAc, 2acetamido-4-amino-2,4,6,-trideoxy-~-galactose; L-AltNAcUA, 2-acetamido-2-deoxy-~-altruronic acid; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; ACL, antigen carrier lipid; BHI, brain heart infusion; EDAC, 1-(3-dimethylaminopropyl)-3ethylcarbodiimide; LAL, Limulus amoebocyte Lysate. tinal mucosal disease. In an attempt to construct a vaccine that would stimulate local intestinal immunity to S. sonnei, the genes determining the Shigella form I antigen have been transferred to an attenuated mutant strain of Salmonella typhi, Ty2la (3); strain Ty2la had previously been shown to act as a safe, effective live oral vaccine for protection against typhoid fever (4)(5)(6). One resulting Ty2la derivative, 5076-1c, has been found to express both the Shigella form I antigen and typical S. typhi somatic antigens. Strain 5076-1C appears from other preliminary studies to be a potential bifunctional oral vaccine for protection against these two important human intestinal disease agents (3, 7).' Due to the great practical importance of this oral vaccine strain, the present investigation was initiated to examine the physiochemical nature of the LPS antigens of this genetic derivative strain. We report here the isolation and purification of the 5076-1C LPS antigens and compare their properties to normal S. sonnei and S. typhi LPS. Previous serological studies of strain 5076-1C indicated the presence of normal form I LPS, but the chemical data presented herein demonstrate that the 5076-1C form I antigen is not covalently bound to core lipid A. Instead, this form I antigen exists as an independent, highly polymerized, immunogenic entity on the cell surface. In contrast to current concepts of LPS assembly (8-lo), this unexpected finding indicates that transport of polymerized 0-antigen, possibly attached to a phosphorylated carrier lipid, from the cytoplasmic membrane to the outer membrane does not require covalent attachment to core lipid A. EXPERIMENTAL PROCEDURES3

RESULTS
Isolation and Purification of Form I Antigen and Somatic LPS Antigens from 5076-IC Cells-Ultracentrifugation of the aqueous phases derived from hot phenol-water extraction of (-gal) and (+gal) 5076-1C cells afforded LPS-R and LPS-S, respectively, (1.5 and 2.3% yields based on acetone-dried cells; Fig. 1). The form I antigen did not pellet but remained in the supernatant extracts. After nuclease treatment, the form I antigen was alcohol-precipitated, dialyzed, and passed through a DEAE-Sephacel column to remove residual nucleic acids as described under "Experimental Procedures." At this 5076.1C grawn "*"i LPS Antigens of S.  stage, the form I Ps antigen (0.10% yield) derived from (-gal) 5076-1C cells was judged homogeneous due to the low amount (0.05%) of nucleic acids, protein, and neutral carbohydrates detected. The (+gal) supernatant extract contained S. typhi 0-Ps in addition to S. sonnei form I Ps antigen (Fig. 1). These two antigens, which were retained in a dialysis membrane with a molecular mass cutoff of 14 kDa, co-precipitated during ethanol addition and co-eluted during the subsequent DEAE-Sephacel chromatography, but separated on a Sephacryl gel permeation column in EDTA-containing buffer (Fig. 2). The form I Ps antigen (0.14% yield), identified by immunodiffusion assays, eluted before the S. typhi 0-Ps peak (0.12% yield, fractions 47-55), which was detected colorimetrically by phenol-sulfuric acid reagent. Three other positive phenolsulfuric acid peaks were also evident. The first two peaks, one eluting at the void volume and the other before the S. sonnei form I Ps-containing fractions, absorbed maximally at 258 nm and yielded ribose during acid hydrolysis, thus indicating nucleic acids that were not retained by the DEAE-Sephacel column. The peak eluting at the bed volume contained free glucose detected by the sugar analyzer. Based on the color yield of a standard glucose solution, the free glucose content (0.1 mg) was considered a minor contaminant constituting typhi Derivative 9029 less than 0.4% of total weight (26 mg) of both the isolated S. typhi 0-Ps and the form I 0-Ps antigen. The form I antigen (14 mg) and the S. typhi 0-Ps (12 mg) derived from (+gal) cells (10 g dry weight) contained less than 0.1% nucleic acids and protein.
The form I Ps antigens, isolated from (-gal) and (+gal) cells, proved to be equivalent antigens by immunoassays, SDS-PAGE, chemical modification studies, and NMR spectroscopy. The term form I Ps antigen is used t o represent either of these two indistinguishable antigen preparations. During gel filtration in disaggregating buffer (Fig. 2), it eluted in a region behind dextran T-20 which has an average M , = 20,000. Lipid A Analysis-To ascertain whether lipid A is a component of the 5076-1C somatic antigens, we employed the Limulus amoebocyte lysate assay (Table I) and also performed chemical analyses of acid-hydrolyzed products. LPS-R and LPS-S gave high LAL end points corresponding to levels observed for the control Escherichia coli LPS and LPS of parental S. sonnei and S. typhi organisms. The 5076-1C form I Ps antigen and S. typhi 0-Ps were about 10S-fold less effective in promoting LAL gelation, thus indicating the absence of the toxic lipid A. Moreover, treatment of both antigens with hot 1% acetic acid, known to cleave acid-labile KDO linkages in LPS (8,9), did not result in lipid A flocculation, which did occur in similar treatment of LPS-R and LPS-S.
Neither the 5076-1C form I Ps antigen nor S. typhi 0-Ps contained glucosamine or fatty acids, which were readily detected in LPS-R and LPS-S ( Table 11). The phosphorus contents of the 5076-1C form 1 Ps antigen and S. typhi 0-Ps were 4-to 10-fold lower than LPS-R and LPS-S. The fatty  typhi (8,9,281, and S. sonnei (1,2). Tyu, tyvelose; R h , rhamnose; acid profiles of LPS-R and LPS-S, which revealed a predominance of &OH myristic acid (about 70%), exemplified the pattern seen for normal enterobacterial LPS (28), including the parental S. typhi and S. sonnei LPS (Table 11).
Monosaccharide Analysis-The monosaccharide constituents which comprise the 0-specific Ps and core regions of S. sonnei and S. typhi LPS are schematically depicted in Fig. 3. Except for the amino sugars and the two unusual sugars of the S. sonnei 0-repeat unit, the neutral monosaccharides, as well as KDO, in these two LPS structures are readily detected by the sugar analyzer (11).
The sugar chromatograms of LPS-R, LPS-S, and S. typhi 0-Ps are presented in Fig. 4. While glucose and mannose are not separated by the chromatographic system, they were distinguished by treating a portion of sample hydrolysate with glucose oxidase to remove glucose prior to analysis (29). In this manner, the glucose/mannose peak of LPS-R at 57 min in Fig. 4A contained 90% glucose and 10% mannose. In contrast, the glucose/mannose peak of LPS-S (Fig. 4B) contained 20% glucose and 80% mannose. In s. typhi 0-Ps, the 57-min peak contained only mannose. The chromatograms of LPS-R and LPS-S differed markedly. In LPS-R (Fig. a), the core sugars heptose and glucose were predominant compared to the S. typhi 0-chain components: tyvelose, rhamnose, mannose, and galactose. The reverse was true for LPS-S (Fig.  4B) where the S. typhi 0-specific chain sugars were predominant. The core sugar KDO was also detected in both chromatograms even though a substantial proportion of it was presumably destroyed during hydrolysis (11). The chromatogram of the nontoxic S. typhi 0-Ps (Fig. 4C) revealed only the presence of 0-side chain components: tyvelose, rhamnose, mannose, and galactose. Based on their color yields, these monosaccharides were present in equimolar amounts. The S. sonnei 0-antigen sugars were not detected on the sugar analyzer, possibly due to their acid lability (1) and/or retention on the analytical column. The acid hydrolysate of S. sonnei form I LPS yielded a sugar chromatogram containing peaks only for the core components, KDO, glucose, galactose, and heptose. In contrast, acid-hydrolyzed 5076-1C form I Ps antigen yielded a blank chromatogram, suggesting the absence of the inner and outer core components (data not shown).
Chemical Modification and Analysis of 5076-lC Form I Antigen-To identify its sugar components, the 5076-1C form I Ps antigen was chemically modified, as described under "Experimental Procedures," to yield stable amino sugar derivatives detectable with the amino acid analyzer. The main reaction course for deamination of the S. sonnei O-disaccharide repeat structure (Fig. 3) minyl derivative (1). Reduction of the carboxyl group of L-AltNAcUA residues would lead to an altrosaminyl structure. As expected, the acid hydrolysate of deaminated form I Ps antigen gave an amino sugar peak eluting at the quinovosamine position (Table 111), which was not present in the hydrolysate of unmodified form I Ps antigen. The carboxylreduced, deaminated 5076-1C form I Ps antigen gave a second amino sugar peak eluting in the position of 2-amino-2-deoxyaltrose (Table 111). These results indicated that the 5076-lC-derived form I Ps antigen contains the expected Shigella 0-disaccharide repeat units.
Acid-hydrolyzed, parental S. sonnei form I LPS gave a glucosamine peak (Tables I1 and 111), whereas its deaminated, carboxyl-reduced form yielded the expected quinovosamine and 2-amino-2-deoxyaltrose peaks in addition to glucosamine. On the other hand, only glucosamine could be detected in   6). The arrows correspond to the top (7') and bottom ( R ) of the SDS gel.
both untreated and chemically treated LPS-R, LPS-S, and S. typhi Ty2la LPS (Table 111). Thus, the S. sonnei form I 0constituents are not covalently bound to 5076-1C LPS-R or LPS-s. The 5076-1C form I Ps antigen (Lanes A and I) was not visualized on gels by the silver-staining reagent, possibly because of its failure to enter the SDS gel or because of the peculiarity of its 0-repeat structure. Thus, we performed an immunoblot of a gel containing electrophoresed 5076-1C form I Ps antigen along with the LPS of S. sonnei form I, S. typhi 5076-1C, and S. typhi Ty2la strains. The resulting autoradiogram (Fig. 6) indicated that both the 5076-1C form I Ps antigen (Lanes A and B) and native S. sonnei form I LPS (Lane D) were serologically reactive with the form I-specific antiserum. The LPS-R from strain 5076-1C (Lane C) and S. typhi Ty2la (Lane E) were not reactive. Moreover, we have observed that the 5076-1C LPS-S, when electroblotted, failed to react with the form I-specific antiserum (data not shown). These data demonstrate that the Shigella form I 0-Ps is not incorporated as part of the 5076-1C LPS structure.

SDS-PAGE and Immunoblot
The radioactive profile of the parental S. sonnei form I LPS corresponded to its silver-stained pattern (Fig. 5). The radioactivity observed in Lanes A and B of Fig. 6 indicated that the 5076-1C form I Ps antigen did indeed migrate into the SDS gel even though it was not silver stained. Its putative location on SDS-PAGE (indicated by the asterisks, Lane I, from the immunoblot autoradiogram (Fig. 6). It should be noted that Shigella 0-Ps, derived from S. sonnei form I LPS by mild acid treatment, failed to electrophorese under similar SDS-PAGE conditions (data not shown). Thus, even though the 5076-1C form I Ps antigen and the S. sonnei form I 0-Ps share the same antigenic determinants, they are markedly different with respect to electrophoretic behavior (see "Discussion").
I 3 C NMR Spectroscopy of Form I Antigen-The presence of S. sonnei 0-disaccharide repeat units in the 5076-1C form" I Ps antigen was confirmed by FT-NMR spectroscopy. The I3C NMR spectra of 5076-1C form I Ps antigen (Fig. 7A) and of S. sonnei 0-Ps (Fig. 7B), obtained from S. sonnei form I LPS by mild acid hydrolysis, are similar. Most prominent are the I3C signals for carbonyl carbons at 176.6 and 176.2 ppm, two anomeric carbons at 105.4 and 103.5 ppm, acetamido group(s) (CH3CO) at 24.9 ppm, and a methyl group signal of a 6deoxyhexose at 18.2 ppm. Signals appearing at 53. 5,54.1, and at 57.3 (weak) ppm may be assigned to carbon atoms substituted with amino or acetamide groups. Secondary hydroxyl carbon atoms were also observed at 69. 8, 70.2, 78.1, and 79.8 ppm. The chemical shifts observed in the two spectra (Fig. 7, A and B) are consistent with those reported by Kenne et al. (1) for Shigella 0-Ps and are indicative of an 0-disacchariderepeating structure comprised of L-AltNAcUA and 4-n-DFUCNAC. No other major component was revealed by the NMR spectra.

DISCUSSION
The LPS antigens obtained from the S. typhi vaccine derivative strain 5076-1C, which expresses the Shigella form I Ps antigen, were compared to those derived from parental S. sonnei and S. typhi strains. Chemical studies demonstrated typical core-linked 0-Ps from the parental strains and confirmed that the form I 0-specific disaccharide repeat unit contains L-AltNAcUA and 4-n-~FucNAc (1,2). All LPS materials, in these studies, that sedimented during ultracentrifugation displayed typical behavior: 1) treatment with hot 1% acetic acid resulted in lipid A flocculation; 2) LPS-R and LPS-S promoted Limulus lysate gelation (Table I); and 3) chemical analyses revealed lipid A components glucosamine and P-OH myristic acid and inner core sugars heptose and KDO (Table 11). Because galE mutants are unable to convert UDP-glucose to UDP-galactose when grown in a galactosedeficient medium, the 5076-1C strain should produce an incomplete LPS of an R, chemotype (8). In fact, LPS-R was a rough, incomplete LPS as verified by the sugar chromatogram which showed the predominance of the core sugars glucose and heptose ( Fig. 4; the core sugar KDO was detected in low amounts due to its acid lability). In the presence of exogenous galactose,galE mutants should produce LPS indistinguishable from that of the wild type bacteria (8). Sugar analysis of LPS-S, derived from (+gal) 5076-1C cells, revealed a predominance of 0-antigen components (tyvelose, rhamnose, mannose, and galactose; Fig. 4B) as expected.
Besides synthesizing LPS-S, 5076-1C (+gal) cells produced the form I Ps antigen in an unusual form, differing from classical LPS in these respects: 1) it did not sediment during ultracentrifugation; 2) it gave low Limulus amoebocyte lysate activity; 3) it contained no core lipid A components; and 4) it exhibited lower mobility than native S. sonnei LPS on SDS gels. The inability to silver stain the 5076-1C form I Ps mtigen on SDS gels was attributed to the lack of core lipid A components, since the LPS of both S. sonnei form I and I1 organisms were stained (Fig. 5). As evidenced by immunoblot analysis, the 5076-1C form I antigen did electrophorese on SDS gels, although at a much lower relative mobility compared to native S. sonnei form I LPS (Fig. 6). Two factors may explain its low mobility. First, the 5076-1c form I antigen is a high molecular weight polymer; dialysis membrane and gel filtration results indicate M, = 14,000-20,000. Second, the antigen may have a phosphorylated lipid carrier end. Phosphorylated ACL such as undecaprenol pyrophosphate-linked moieties, are known to be involved in the biosynthesis of 0-Ps chains (8)(9)(10). Since native LPS mobility in SDS gels depends on their relative lipid A content (30,31), a P-ACL attached to 5076-1c form I Ps antigen would be the only hydrophobic region that binds with SDS, thus allowing electrophoresis into an SDS gel. A lipid requirement is based on the observation that parental S. sonnei 0-Ps, denuded of lipid A, failed to enter the SDS gel under similar conditions (data not shown). The presence of P-ACL is also supported by the detection of a low phosphorus content in the 5076-1C form I 0-Ps (Table 11). Unequivocal proof of a phosphorylated lipid carrier attached to the form I Ps antigen, however, must await further chemical analysis.
The (+gal) 5076-1C strain also produced some core-free S. typhi 0-Ps (0.12% yield, Fig. 4C). Like the form I Ps antigen, this 0-antigen exhibited low Limulus activity (i.e. lipid Afree), remained in the supernatant during centrifugation, and exceeded 14 kDa in molecular mass as measured by dialysis.
Its low phosphorus content (Table II), together with the above data, suggests that this S. typhi 0-Ps antigen is linked through a P-ACL, as discussed above for the form I antigen. Evidently, transfer of P-ACL-linked S. typhi 0-Ps to the core lipid A structure is an inefficient process. This is not unexpected since, during galactose uptake in the galE mutant, intracellular UDP-galactose accumulates to a toxic level (3, 4), possibly interfering with normal LPS synthesis. Note that corefree 0-Ps antigen is not produced in wild type s. typhi or s. sonnei cells (data not shown); apparently, all of the 0-Ps chains are covalently integrated into LPS molecules, signifying an efficient translocation process.
Recent studies have demonstrated that the core-linked 0specific polysaccharide chain of one bacterial species could be replaced by an 0-antigen of another species via genetic manipulation, e.g. conjugal mating or transduction (9,(33)(34)(35). However, no evidence of covalent union between the form I Ps antigen and the Salmonella core was found by chemical or by immunoblot analyses. We conclude that the 5076-1C form I Ps antigen does not form an integral part of the Salmonella LPS molecule, but exists as a highly polymerized (i.e. an estimated 30-45 0-repeat units) autonomous cell surface entity. Lipid A-free, 0-Ps antigens are also known to be synthesized by rfu mutant bacteria (8,36). Defective in R-core biosynthesis, these rfa mutants elaborate 0-Ps antigens that are not transported out to the outer membrane, but remain attached to P-ACL at the cytoplasmic membrane, thus explaining the inability of these rfa mutant cells to agglutinate in the presence of anti-0 antibodies (8,37,38). However, the lipid A-free, 5076-1C form I antigen is expressed on the cell surface, based on cell agglutinability in form I-specific antisera (3,12). Hence, in contrast to the current theory of LPS assembly in the Enterobacteriaceae, the 5076-1C form I PS antigen does not require translocation to core lipid A at the cytoplasmic membrane for transport to the outer membrane.
According to present knowledge, the biosynthesis of 0specific Ps in the Enterobacteriaceae is controlled by the chromosomal rfb, rfc, and rfe gene clusters (8,9). The form 1 antigen genes of s. sonnei, however, exist on a 120-MDa plasmid in natural isolates of this species. This plasmid was transferred to S. typhi during the construction of strain 5076-1C. Although previous serological studies of the 5076-1C strain indicated that this plasmid encodes the structural genes for the polymerized form I 0-antigen, the above chemical data provide further support for this conclusion. In addition, the S. sonnei plasmids may encode other undefined traits involved in 0-antigen expression. Recently, the S. sonnei plasmid has been introduced into an E. coli K12 strain4 that elaborates a complete LPS core, a prerequisite for smooth LPS synthesis.
The derivative E. coli strain is agglutinable with the S. sonnei form I-specific antiserum. In contrast to results obtained with strain 5076-1C, SDS-PAGE analysis revealed that the form I antigen is covalently bound to E. coli core-lipid A.' These results are not altogether surprising since the E. coli K12 core is more akin to the Shigella core than to that of Salmonella. Further genetic transfer of the form I 0-antigen genes to other defined LPS mutants of enteric bacteria should provide additional insights into the mechanism of LPS assembly.