Salmonella Pathogenicity Island 2-Dependent Evasion of the Phagocyte NADPH Oxidase

See allHide authors and affiliations

Science  03 Mar 2000:
Vol. 287, Issue 5458, pp. 1655-1658
DOI: 10.1126/science.287.5458.1655


A type III protein secretion system encoded bySalmonella pathogenicity island 2 (SPI2) has been found to be required for virulence and survival within macrophages. Here, SPI2 was shown to allow Salmonella typhimurium to avoid NADPH oxidase–dependent killing by macrophages. The ability of SPI2-mutant bacteria to survive in macrophages and to cause lethal infection in mice was restored by abrogation of the NADPH oxidase–dependent respiratory burst. Ultrastructural and immunofluorescence microscopy demonstrated efficient localization of the NADPH oxidase in the proximity of vacuoles containing SPI2-mutant but not wild-type bacteria, suggesting that SPI2 interferes with trafficking of oxidase-containing vesicles to the phagosome.

The central importance of the phagocyte NADPH (nicotinamide adenine dinucleotide phosphate) oxidase to innate host defense is vividly demonstrated in chronic granulomatous disease. Mutations in any of the subunits comprising the NADPH oxidase predispose patients to recurrent infections with fungi and bacteria, including Salmonella (1). The NADPH oxidase catalyzes the univalent reduction of oxygen to superoxide, an oxidizing species and precursor to potent antimicrobial molecules such as hydrogen peroxide, hydroxyl radical, and peroxynitrite (2,3). Pathogenic microbes have developed strategies to resist the antimicrobial effects of the NADPH oxidase, including the production of molecular scavengers, antioxidant enzymes, repair systems, and expression of specific antioxidant regulons (2). For example, the OxyR and SoxRS regulons enableEscherichia coli to resist the effects of hydrogen peroxide and superoxide, respectively (4). However, S. typhimurium does not require a functional OxyR or SoxRS regulon for virulence (5), suggesting that Salmonella may use alternative strategies to avoid exposure to high concentrations of phagocyte-derived oxidants in vivo.

A cluster of genes at centisome 30 of the S.typhimurium chromosome, designated Salmonellapathogenicity island 2 (SPI2), encodes a type III secretion system required for virulence and intracellular survival (6,7) and believed to translocate bacterial proteins into the cytosol of host cells. We have used immunodeficient mice to identify the specific host defenses targeted by products of the SPI2 genes. Salmonella typhimurium strains deficient at any of several SPI2 loci (ssrA, ssaJ, ssaV,sseB) (8) were found to be highly attenuated for virulence in C57BL/6 mice (Fig. 1A) (9). Virulence of these SPI2-mutant strains was not restored by administration of aminoguanidine, an inhibitor of inducible nitric oxide synthase (iNOS) (10), or by genetic abrogation of interferon-γ (IFN-γ) (11) or interleukin-12 (12) production (Fig. 1, B and C). In contrast, all four SPI2 mutants were able to cause lethal infection of congenic C57BL/6 mice deficient in the gp91phox subunit of the phagocyte NADPH oxidase (gp91phox knockout mice) (13) (Fig. 1D). Thus the SPI2 genes are not required for virulence in the absence of a phagocyte respiratory burst and might play a specific role in avoiding bacterial interaction with the NADPH oxidase.

Figure 1

Abrogation of the NADPH phagocyte oxidase restores virulence to SPI2-deficient S.typhimurium mutants. Survival curves are shown for wild-type C57BL/6 mice (A), wild-type mice fed drinking water containing the iNOS inhibitor aminoguanidine (B), congenic immunodeficient IFN-γ knockout mice (C) or gp91phox knockout mice (D), following intraperitoneal challenge with wild-type S.typhimurium or isogenic strains with mutations atssaJ::Tn5, ssaV::Tn5,ssrA::Tn5, or ΔsseB::aphT. An isogenicaroA-mutant S. typhimurium strain with attenuated virulence in mice (29) was included as an additional control. These experiments used 4 to 14 mice per group.

Killing of isogenic S. typhimurium strains carrying mutations in various SPI2 genes (ssaJ,sseA, sseB, ssrA) was examined in macrophages from wild-type or respiratory burst-deficient mice (14). SPI2-mutant bacteria had increased susceptibility to killing by periodate-elicited murine peritoneal macrophages from C57BL/6 mice (Fig. 2A), but this enhanced susceptibility was abrogated in macrophages from congenic gp91phox knockout mice (Fig. 2B). Parallel experiments using the NOS inhibitorN G-monomethyl-l-arginine or macrophages from iNOS knockout mice did not restore wild-type levels of macrophage survival to the SPI2 mutants (15). Similar levels of NO-derived nitrite production were measured from macrophages infected with either wild-type or SPI2-mutant bacteria, indicating that SPI2 does not interfere with NO synthesis.

Figure 2

Abrogation of the NADPH phagocyte oxidase restores the ability of SPI2-deficient S.typhimurium mutants to survive in macrophages. The survival of wild-type S. typhimurium 12023 or isogenic SPI2-deficient strains with mutations atssaJ::Tn5, ssrA::Tn5, ΔsseA::aphT, and ΔsseB::aphT was measured in macrophages from C57BL/6 (A) and congenic gp91phox knockout mice (B). These data represent the mean ± SEM of three separate experiments.

The SPI2 mutants were not more susceptible in vitro to hydrogen peroxide, the superoxide-generator methyl viologen, or the peroxynitrite-generator SIN-1 in disk diffusion assays and were not more susceptible to hydrogen peroxide killing in liquid medium (16). Thus SPI2 does not appear to directly enhance bacterial resistance to macrophage-derived oxidants. Furthermore, C57BL/6 peritoneal macrophages infected with a similar inoculum of wild-type or SPI2-mutant Salmonella exhibited comparable lucigenin-dependent chemiluminescence (18) (Fig. 3A). Infection with Salmonellawas sufficient to stimulate a respiratory burst in periodate-elicited peritoneal macrophages. Superoxide production by macrophages infected with wild-type or sseB-mutant bacteria was approximately 0.013 and 0.009 nmol/hour per 105 macrophages, respectively, as measured by reduction of cytochrome c. Thus the SPI2 gene products might interfere with the localization rather than the activation of the phagocyte NADPH oxidase.

Figure 3

Exclusion of oxyradical formation fromSalmonella-containing vacuoles. Macrophages from C57BL/6 mice challenged with wild-type or sseB-mutant bacteria produced comparable quantities of superoxide as measured by reduction of lucigenin (arbitrary units for chemiluminescence) (A). NADPH oxidase activity was visualized as cerium perhydroxide precipitate in periodate-elicited macrophages from C57BL/6 mice challenged with either avirulent E. coli W3110 (B), wild-type S. typhimurium(C, E, and G), or isogenicsseB-mutant S. typhimurium(D and F). Cerium perhydroxide precipitate was localized to the plasma membrane (G) and to uninfected vacuoles in macrophages harboring wild-type Salmonella. In some experiments (B to D and G), the macrophages were activated in vitro with 20 ng/ml PMA (30). (B) and (D) are magnified ×10,752 (original magnification, ×49,250); (C), (E), (F), and (G) are magnified ×7968 (original magnification, ×36,500). The percentage of bacteria-containing vacuoles colocalizing with cerium perhydroxide precipitate as an indication of NADPH oxidase activity is shown in (H). These data represent 190 vacuoles from eight separate experiments.

To visualize NADPH oxidase activity in relation toSalmonella-containing vacuoles, we performed ultrastructural studies of infected macrophages using cerium chloride (19). In the presence of hydrogen peroxide, cerium chloride is converted to electrodense cerium perhydroxide precipitate. Electron micrographs revealed efficient colocalization of cerium perhydroxide with vacuoles containing nonpathogenic E. coli W3110 (Fig. 3B) or SPI2-mutant S. typhimurium (Fig. 3, D and F), but not wild-type S. typhimurium (Fig. 3, C, E, and G). Approximately 50% of the vacuoles containingsseB-mutant S. typhimurium colocalized with cerium perhydroxide, contrasting with only 5% of those containing wild-type bacteria (Fig. 3H). The reduced tendency of vacuoles containing wild-type Salmonella to colocalize with cerium perhydroxide persisted despite augmented phagocyte stimulation with the potent NADPH oxidase activator phorbol 12-myristate 13-acetate (PMA) (Fig. 3, B, C, D, G, and H). Evidence of NADPH oxidase activity was seen in macrophages containing wild-type bacteria but was localized to empty vacuoles or to the plasma membrane (Fig. 3G). No cerium perhydroxide was detected in infected macrophages obtained from gp91phox knockout mice (15).

Intracellular distribution of the NADPH phagocyte oxidase following infection of macrophages with Salmonella was visualized by immunofluorescence microscopy (20). In quiescent cells, the NADPH oxidase p22 and p47 subunits appeared to be preferentially distributed in the plasma membrane and cytosol, respectively (Fig. 4, A and B), and were mobilized to the periphery upon activation with PMA (Fig. 4, C and D). In Salmonella-infected cells, the p22phox and p47phox subunits appeared to be localized within compartments that coalesced in the proximity of phagosomes containing green fluorescent protein (GFP)–expressing SPI2-mutant bacteria (Fig. 4, E and F). In contrast, the compartments containing the NADPH oxidase remained diffusely distributed within cells infected with GFP-expressing wild-type Salmonella (Fig. 4, G and H), or aggregated at intracellular locations remote from the bacteria (Fig. 4G). Localization of NADPH oxidase components was seen in the vicinity of 56% of sseB-mutant and 10% of wild-type bacteria, respectively, correlating well with the electron microscope studies. Thus, functional NADPH oxidase can be localized within discrete intra- cellular compartments in macrophages as has been described in human neutrophils (21). Intraphagosomal oxyradical production appears to require both recruitment of cytosolic components to the membrane-associated flavocytochrome and intracellular vesicular trafficking to deliver active oxidase to the phagosome.

Figure 4

Exclusion of the NADPH phagocyte oxidase fromSalmonella-containing vacuoles. Immunofluorescence microscopy of p22 and p47 NADPH oxidase subunits (red) was performed in quiescent (A and B) and PMA-activated (Cand D) macrophages. Both NADPH oxidase p22phox (E) and p47phox (F) subunits coalesced in the vicinity of GFP-positive (green) sseB-deficientS. typhimurium, contrasting with the dispersed pattern of vesicular distribution in the cytoplasm of macrophages infected with GFP-expressing (green) wild-type bacteria (Gand H). (I) and (J ) show p22phox and p47phox staining in macrophages from gp91phox knockout (phox KO) mice infected in vitro with sseB-mutantS. typhimurium. Magnification of fluorescence micrographs is ×1900 (original magnification, ×4000). These data are representative of 150 cells from 14 separate experiments.

Other type III secretion systems have been shown to target the host cytoskeleton (22), and it is possible that SPI2 gene products originating from intraphagosomal bacteria (23) interfere with localization of the NADPH oxidase in phagocyte vacuolar membranes by blocking cytoskeletal rearrangements (24). The effects of this action are not necessarily limited to the NADPH oxidase and, indeed, could provide a common mechanism to explain observations suggesting that S.typhimurium interferes with the fusion of phagosomes with secondary lysosomes carrying markers such as the mannose-6-phosphate receptor and cathepsin (25). Such oxidase-independent effects of SPI2 might help to explain the slight delay in mortality caused by SPI2-mutant bacteria in gp91phox knockout mice (Fig. 1D), and the modest survival defect of SPI2 mutants in some cell lines lacking detectable production of reactive oxygen intermediates (7).

Overall, SPI2 appears to prevent the phagocyte NADPH oxidase from trafficking toward Salmonella-containing vacuoles, both reducing the oxidant stress encountered by Salmonella and potentially enhancing collateral oxidative damage to host tissues. Interference with the localization of the NADPH oxidase is coupled with more conventional antioxidant strategies such as scavengers, detoxifying enzymes, and repair systems (2,26). The ability of Salmonella to limit its exposure to high concentrations of toxic phagocyte-derived oxidants may help to explain the dispensability of catalase (17) and the SoxRS and OxyR oxidative stress regulons (5) for virulence. It is possible that other intracellular pathogens pursue similar strategies. For example,Mycobacterium tuberculosis is susceptible to reactive oxygen intermediates in vitro (27), but lacks a functional OxyR locus and appears to be relatively protected from effects of the NADPH oxidase in vivo (28). Exclusion of the NADPH oxidase from phagosomes may be an important contributor to the virulent nature of intracellular pathogens.

  • * Present address: Centre for Veterinary Science, University of Cambridge, Cambridge CB3 0ES, UK.

  • To whom correspondence should be addressed. E-mail: ferric.fang{at}


Stay Connected to Science

Navigate This Article