Optimization of Virulence Functions Through Glucosylation of Shigella LPS

See allHide authors and affiliations

Science  25 Feb 2005:
Vol. 307, Issue 5713, pp. 1313-1317
DOI: 10.1126/science.1108472


Shigella, the leading cause of bacillary dysentery, uses a type III secretion system (TTSS) to inject proteins into human cells, leading to bacterial invasion and a vigorous inflammatory response. The bacterium is protected against the response by the O antigen of lipopolysaccharide (LPS) on its surface. We show that bacteriophage-encoded glucosylation of Shigella O antigen, the basis of different serotypes, shortens the LPS molecule by around half. This enhances TTSS function without compromising the protective properties of the LPS. Thus, LPS glucosylation promotes bacterial invasion and evasion of innate immunity, which may have contributed to the emergence of serotype diversity in Shigella.

Pathogenic bacteria have evolved mechanisms to occupy specific niches within hosts while avoiding elimination by innate immune killing. Shigella is one of several pathogens that express a type III secretion system (TTSS), a needle-like structure through which effector molecules are translocated from the bacterium directly into human cells (1). This molecular syringe is required by Shigella for entry into epithelial cells (24). Bacterial invasion is critical for gastrointestinal (GI) disease because it elicits an intense inflammatory response with subsequent disruption of the epithelial barrier and formation of abscesses in the mucosa. Thus, Shigella both inflicts damage to host tissues and survives the vigorous innate immune response. We used signature-tagged mutagenesis (5) to understand the genetic basis of GI infection by the Shigella flexneri serotype 5 strain, M90T. Out of 2900 mutants analyzed in the rabbit ligated ileal loop model of shigellosis (6, 7), 15 colonization-defective mutants were identified, with 7 resulting from loss of genes involved in lipopolysaccharide (LPS) biosynthesis (Fig. 1, A and B).

Fig. 1.

(A) Genes disrupted in colonization-defective mutants. The competitive index (C.I.) assesses the ability of each mutant to colonize the rabbit ileal loop in direct comparison with the wild-type strain. A C.I. of 1 indicates no attenuation. (B) Tricine-SDS–polyacrylamide gel electrophoresis (TSDS-PAGE) of LPS from the mutants. Mode A and mode B LPS (indicated) contain around 15 and 100 subunits of O antigen, respectively. (C) Representation of S. flexneri 5a LPS (N-ag, N-acetylglucosamine). (D) Sensitivity of the colonization-defective mutants to the antimicrobial peptide, HBD-1; the percent survival relative to M90T after a 1-hour incubation is shown. Error bars indicate SEM.

LPS of enteric bacteria consists of three regions: lipid A, an inner core, and repeating O-antigen subunits. The Shigella O-antigen subunit is a tri-rhamnose (rha)-N-acetyl glucosamine (N-ag) tetrasaccharide (8), modified by the addition of glucose and/or acetate, depending on the serotype (Fig. 1C). The modal number of subunits is about 15 (mode A, Fig. 1B), with another prevalent population of around 100 repeats (mode B). The waaJ, waaD, and waaL mutants, which cannot synthesize the inner core or ligate the O antigen to the inner core, were markedly attenuated (Fig. 1A), which probably reflects their failure to resist being killed by antimicrobial peptides expressed in the GI tract (9) (Fig. 1D; fig. S1). This probably also accounts for the attenuation of the rfbA mutant, which fails to convert glucose-1-phosphate to deoxythymidine diphosphate–rhamnose (dTDP-rha), a step required for the efficient synthesis of O antigen. Likewise the cld mutant has a truncated LPS molecule lacking mode B of the O antigen (Fig. 1B) (10). The colonization capacity of these mutants was restored by complementation with the corresponding wild-type allele (table S1 and fig. S2).

Two mutants were isolated with insertions in the O-antigen glucosylation operon. This operon consists of gtrA, gtrB, and gtrV present on a resident bacteriophage (11, 12), and confers serotype-specific glucosylation of LPS. The gtrA and gtrB mutants had subtle alterations in their LPS profile and expressed a number of O-antigen subunits equivalent to the wild-type strain. However, each subunit had a slightly reduced molecular mass (Fig. 1, B and C) through lack of glucosylation on the second rhamnose residue (rhaII) of the O antigen (13). To confirm the role of the gtr operon in colonization, we constructed a gtrV deletion mutant. gtrV is predicted to encode the glucosyltransferase, which adds glucose to the O antigen in a serotype-specific manner (14). The strain lacking gtrV (M90ΔgtrV) had a substantial disadvantage for survival in the GI tract (Fig. 2A). Complementation of each gtr mutant with pNW83, which contains all gtr genes (Fig. 2), demonstrated that expression of serotype 5a LPS confers a marked survival advantage compared with isogenic strains expressing unglucosylated LPS.

Fig. 2.

(A) Competitive index of mutants with defects in genes of the gtr operon. (B) Construction of strains expressing LPS from different serotypes. (C) TSDS-PAGE analysis of LPS from various strains. The arrow indicates enhanced glucosylation of M90TΔgtrp5a compared with M90T. (D) Histological analysis of infected ileal loops stained with hematoxylin, eosin, and safranin or with Evans blue to demonstrate the location of bacteria (indicated by arrows). The infecting strain is shown. (E) The invasive capacity of strains expressing LPS from different serotypes. The average number of invasive bacteria per epithelial cell is indicated (SD in parentheses).

To establish whether O-antigen modifications of other Shigella serotypes also confer a competitive advantage, we constructed isogenic strains expressing LPS from different serotypes (Fig. 2B). The gtr operon was deleted from M90T (generating M90TΔgtr), then the gtr operons from S. flexneri 1a, 2a, and 5a (13) were introduced on plasmids resulting in M90TΔgtrp1a, M90TΔgtrp2a, and M90TΔgtrp5a, respectively (Fig. 2B). Furthermore, the entire gtr operon from S. flexneri 2a was introduced chromosomally into M90TΔgtr, resulting in M90T2a. The serotype replacements were confirmed by agglutination assays and LPS profiling (table S2 and Fig. 2C). The O-antigen subunits of M90TΔgtrp5a were more extensively glucosylated than those in M90T, probably because of a gene dosage effect. Glucosylation of the O antigen restored the virulence of S. flexneri in the GI tract (competitive index of 0.88, 0.92, and 1.1 for M90TΔgtrp1a, M90TΔgtrp2a, and M90TΔgtrp5a, respectively), which demonstrates that LPS glucosylation promotes the fitness of Shigella in vivo, irrespective of serotype.

One potential explanation for the attenuation of the gtr mutants is that they are defective in avoiding host clearance by innate immune mechanisms. However, the gtr mutants were fully able to withstand the adverse effects of bile salts, hyperosmolarity [to 300 mosmol (15)], complement-mediated lysis (fig. S3), and antimicrobial peptides (Fig. 1D; fig. S1). Furthermore, the survival disadvantage resulting from loss of gtrA was not evident in a noninvasive strain of S. flexneri, BS176, which lacks the TTSS (Fig. 2A). Thus, the decreased fitness of the gtr mutants was only manifest after mucosal invasion.

We next investigated whether LPS glucosylation contributes to the proinflammatory host response, the hallmark of shigellosis. Infection of ileal loops with M90T led to dramatic alterations of mucosal tissues with rupture and destruction of the intestinal epithelium (Fig. 2D). The lesions, typical of acute shigellosis, included extensive zones of epithelial detachment and loss of villi. A massive polymorphonuclear leukocyte (PMN) response was present within the lamina propria and in the edematous submucosal tissues. Numerous invasive bacteria were seen in abscesses and in the lamina propria. Infection with the noninvasive isolate, BS176, did not result in significant alteration of the mucosa, and M90TΔcld (which expresses a truncated O antigen) also had attenuated virulence (Fig. 2D). M90TΔgtr caused consistently less pathological damage than M90T. Intestinal villi were only shortened and swollen, and zones of epithelial destruction were restricted, with abscesses limited to the base of villi (Fig. 2D). The number of bacteria and infiltrating PMNs was much lower than observed in M90T-infected tissues. Total restoration of the virulent phenotype was observed after infection with M90TΔgtr expressing a glucosyltransferase operon, irrespective of serotype (Fig. 2D).

We hypothesized that LPS glucosylation contributes to the virulence of S. flexneri through an effect on its invasive potential. Consistent with this, each gtr mutant had a substantially reduced ability to invade epithelial cells when compared with M90T (Fig. 2E); this defect was entirely corrected by reintroducing the gtr operon from any serotype. Furthermore, strains with increased LPS glucosylation (e.g., M90TΔgtrp5a) showed enhanced invasion compared with M90T (Fig. 2E). We considered whether changes in surface hydrophobicity of the strains could explain this finding. The increased hydrophobicity of M90TΔgalU, which expresses a truncated LPS, might be responsible for the increased adherence of bacteria to eukaryotic cells (table S3). However, the slight variations in hydrophobicity among M90T, M90TΔcld, M90TΔgtr, and the complemented strain could not account for the differences in invasion (table S3). Furthermore, the initial adhesion of the gtr strains to epithelial cells was not affected (table S4), which indicated that loss of O-antigen glucosylation did not affect exposure of any tissue-specific adhesin.

Three-dimensional molecular models of serotype 5a of the O antigen based on nuclear magnetic resonance (NMR) data give structures adopting a right-handed, threefold helix with the branched glucosyl residues pointing outward (16). Glucosylation of serotype 5a LPS is predicted to induce a transition from a linear to helical conformation with the glucosyl residue exposed on the exterior of the helix, forming a more compact structure (Fig. 3A) than unglucosylated LPS (17). This would dramatically shorten the O antigen, halving the distance it extends beyond the outer membrane. In the mode A LPS, the O antigen would extend around 21 nm from the outer membrane in the absence of glucosylation, but only 11 nm for a glucosylated O antigen. To test these predictions, bacteria were visualized by transmission electron microscopy (Fig. 3). M90T had a dense surface material extending about 35 nm beyond the outer membrane (between the two arrowheads, Fig. 3B). In contrast, the exterior of M90TΔgtr was composed of more diffuse, filamentous material that extended around 70 nm from the outer membrane (Fig. 3B). The surface of M90TΔgtrp5a, which displays enhanced glucosylation, was even more compact than M90T. As the average length of protruding needles of the Shigella TTSS is 60 nm (18), LPS glucosylation might affect TTSS function, which would account for the changes in cell invasion (Fig. 2E).

Fig. 3.

(A) Models of S. flexneri unglucosylated (a) and glucosylated (b) O antigens formed by 15 repeating units depicted with branched glucosyl residues colored in magenta. The distances between the C1 atoms of the first and last subunits are 21.2 nm and 11.3 nm, for unglucosylated and glucosylated O antigens, respectively. View of space-filling models, a′ and b′ (shades of green, rhamnopyranosyl residues; yellow, glucopyranosyl). (B) Surface topology of strains revealed by transmission electron microscopy after cryofixation or treatment with ruthenium red. The electron-dense material at the bacterial surface is indicated between the arrows. In M90TΔgtr, the surface material is less compact (reaching about 70 nm beyond the outer membrane) than M90T (extending around 35 nm). Surface staining is more intense and condensed on M90TΔgtrp5a than on M90T. The regions within boxes have been expanded.

To establish whether LPS glucosylation affects the exposure of TTSS on the cell surface, we examined bacteria by scanning electron microscopy (Fig. 4A). TTSS needles were seen clearly on M90T but were absent in M90TΔmxiD. A significantly lower number of TTSS needles were detected on M90TΔgtr compared with M90T, and the defect was entirely restored when the gtr operon was introduced (Fig. 4B). To confirm that these structures are TTSS needles, we performed immunolabeling with a monoclonal antibody against IpaB, which is both secreted and exposed at the TTSS tip (1921). This revealed surface structures on M90T that were much less frequently seen in M90TΔgtr and were absent in the TTSS-null mutant. This reduced staining was not due to impaired secretion, because secretion of IpaB and IpaC was identical in M90T and M90TΔgtr (Fig. 4D). Furthermore with negative-staining electron microscopy, similar numbers of TTSS needles were observed emerging from the outer membrane of M90T and M90TΔgtr (Fig. 4E), which shows that glucosylation does not affect the assembly of TTSS in the outer membrane.

Fig. 4.

(A) Scanning electron microscopy of bacteria demonstrating the abundance of TTSS secretions (arrowed) on the cell surface; scale bars and strains are shown. (B) Number of visible secretions per bacterium (error bars show SD). (C) Immunogold labeling of IpaB at the bacterial surface, and (D) secretion of IpaB and IpaC into culture supernatants detected by Coommassie are not affected by LPS glucosylation; secretion was induced by the addition of Congo Red (+). (E) Negative-staining electron microscopy showed that the number of TTSS emerging from the bacterial outer membrane was not altered in the strain lacking the gtr operon. (F) The interaction between LPS and TTSS. Hyperinvasive strains that have truncated LPS molecules are susceptible to being killed by the innate immune response in vivo. Bacteria expressing unglucosylated LPS are compromised for invasion and therefore attenuated. Glucosylation of the O antigen halves the length of the LPS molecule, which allows efficient function of the TTSS while it retains resistance to antimicrobial factors in vivo.

Our results show that LPS dictates the key function of the Shigella TTSS of mediating bacterial invasion into host cells. Strains with truncated LPS are highly proficient at invading cells in vitro, possibly through enhanced access of the TTSS to host cells and altered hydrophobicity (Fig. 4E). However, this advantage is entirely offset in vivo where the bacterium is more susceptible to innate immune factors. In strains expressing full-length LPS in the absence of glucosylation, although Shigella is fully able to resist innate immune killing, the extended LPS isoform impairs TTSS function, which leads to reduced virulence within the GI tract. In both instances, the bacterium is at a competitive disadvantage through a tradeoff between its invasive capacity and its ability to survive innate immune effectors.

Evidently, glucosylation of LPS facilitates invasion of target cells by altering the conformation of LPS to optimize the exposure of TTSS needles while retaining resistance against host defences. The relationship between LPS and the function of the TTSS can be considered similar to a sword and shield, in which there is a balance between the length of LPS O side chains to protect the bacterium against innate immune effectors (“the shield”), and the influence on the function of the TTSS needle (“the sword”): both are essential for Shigella virulence.

Supporting Online Material

Materials and Methods

Figs. S1 to S3

Tables S1 to S6

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article