Escape of Intracellular Shigella from Autophagy

See allHide authors and affiliations

Science  04 Feb 2005:
Vol. 307, Issue 5710, pp. 727-731
DOI: 10.1126/science.1106036


The degradation of undesirable cellular components or organelles, including invading microbes, by autophagy is crucial for cell survival. Here, Shigella, an invasive bacteria, was found to be able to escape autophagy by secreting IcsB by means of the type III secretion system. Mutant bacteria lacking IcsB were trapped by autophagy during multiplication within the host cells. IcsB did not directly inhibit autophagy. Rather, Shigella VirG, a protein required for intracellular actin-based motility, induced autophagy by binding to the autophagy protein, Atg5. In nonmutant Shigella, this binding is competitively inhibited by IcsB binding to VirG.

During the multiplication of microbes within host cells, bacteria become sequestered in membrane-bound organelles such as phagosomes (13). This event is a key component of host defense against invading microbes. Nevertheless, some invasive bacteria such as Legionella, Salmonella, Mycobacteria, and Brucella can block or alter the maturation of the phagosome and can reside in vacuoles (27). Some others such as Shigella (8, 9), Listeria monocytogenes (10), and Rickettsia conorii (11) can escape from phagosomes into the cytoplasm, multiply, and disseminate into neighboring cells by eliciting actin polymerization. Cytoplasmic pathogens may thus circumvent autophagic events.

IcsB, one of the Shigella flexneri effectors, is secreted by means of the type III secretion system (TTSS) of cytoplasmic bacteria and located on the bacterial surface (12). The icsB mutant is fully invasive and able to escape from the vacuole but is defective in spreading within host cells (12).

To clarify the role of IcsB in promoting infection, we investigated the intracellular behaviors of the icsB mutant (ΔicsB), YSH6000 (wild type; WT), and ΔicsB/pIcsB (the icsB complement strain). In baby hamster kidney (BHK) cells, although mutants lacking IcsB multiplied as normal for about 3 hours, their growth plateaued 4 hours after invasion (fig. S1A). To characterize intracellular bacteria, we introduced green fluorescent protein plasmid (pGFP) into ΔicsB and WT then investigated BHK cells infected with bacteria 4 hours after infection. ΔicsB/pGFP colocalized with markers for acidic lysosomes (Lysotracker) or autophagosomes [monodancyl-cadaverin (MDC)], where the bacterial morphology was indistinct. (fig. S1, B and C). WT cells, on the whole, did not colocalize with the same markers: 37.2% of ΔicsB bacteria colocalized with lysosomes compared with only 10.2% of WT. Furthermore, when BHK cells expressing GFP-LC3, an autophagosome-specific marker (13, 14), were infected with ΔicsB or WT, ∼40% of ΔicsB was associated with LC3 signal; bacterial shape was also indistinct compared with WT (fig. S1D). To further characterize the ΔicsB defect, we exploited MDCK cells (epithelial cells from dog kidney) expressing GFP-LC3 (MDCK/pGFP-LC3 cells), which made it feasible to visualize cytoplasmic organelles and bacteria (Fig. 1A). The number of LC3-positive ΔicsB was greater than that of WT throughout the 1 to 6 hours after infection. The LC3-positive population of ΔicsB had increased 50% by 6 hours, whereas that of WT remained at 10 to 15% (Fig. 1B). Two hours after infection, WT and ΔicsB had similar numbers of actin tails. After 4 hours, however, the population was decreased in ΔicsB (fig. S2), presumably because ΔicsB was within autophagosomes. The LC3-positive population of the ΔicsB/pIcsB was decreased: it fell to a level as low as that of WT (Fig. 1B). Autophagic events can be triggered by amino acid starvation (13). MDCK/pGFP-LC3 cells were infected with ΔicsB or WT, under amino acid–starved conditions. LC3-positive bacteria in MDCK cells were significantly increased from 10 to 16% (WT) and from 23 to 36% (ΔicsB) in response to amino acid deprivation (fig. S3). Conversely, when MDCK cells were treated with known inhibitors of autophagy or of lysosomes, such as Wortmannin, 3-methyladenine (3-MA) or bafilomycin-A1 (Baf-A1), the LC3-positive ΔicsB population was markedly decreased (Fig. 1, C and D). In the presence of Baf-A1, fusion of lysosomes with autophagosomes containing ΔicsB was blocked, which would have allowed the bacteria to escape into the cytosol. Consistently, despite the smaller diameter (<0.15 mm) of plaques formed by ΔicsB 2 days after infection than that of plaques formed by WT (∼0.5 mm), the plaque-forming capacity of ΔicsB was restored by treatment with Baf-A1 (fig. S4). Another investigation was made in atg5-knockout mouse embryonic fibroblasts (atg5–/– MEFs), which are defective in autophagy. When atg5–/– MEFs or normal MEF cells (atg5+/+ MEFs) expressing GFP-LC3 were infected with ΔicsB, even though LC3-positive ΔicsB is detectable in normal MEFs, signals were barely detected in the atg5-knockout MEFs (Fig. 1, E and F). Consistently, intracellular growth of ΔicsB was recovered to the level of WT in atg5–/– MEF cells (Fig. 1G), which provided further evidence that the defective intracellular phenotype of ΔicsB was associated with autophagy.

Fig. 1.

Shigella icsB mutant trapped by autophagy. (A) MDCK/pGFP-LC3 cells infected with WT or ΔicsB for 4 hours were stained with antibody against Shigella (blue) and rhodamine phalloidin (red) and (B) the percentages of GFP-LC3–positive bacteria in (A) were quantified. *P < 0.0001. (C) In the presence of autophagy inhibitors, MDCK/pGFP-LC3 cells infected with ΔicsB for 4 hours were stained with antibody against Shigella (red). In (D), cells described in (C) are compared with cells infected with WT under the same conditions to determine the percentages of GFP-LC3–positive bacteria. *P < 0.0001. (E) atg5-knockout MEFs or normal MEFs expressing GFP-LC3 infected with ΔicsB for 4 hours were stained with antibody against Shigella (red) and compared with those infected with WT. (F) Under the same conditions, they were stained with antibody against Shigella. The percentages of GFP-LC3–positive cells were quantified. *P < 0.0001. (G) atg5-knockout or normal MEF were infected with WT or ΔicsB for indicated periods. Intracellular multiplication of bacteria is determined. *P < 0.0001. Scale bars (A), (B), and (E), 2 μm.

MDCK cells infected with ΔicsB or WT for 4 hours were examined by thin-section electron microscopy (EM). ΔicsB was frequently enclosed by lamellar membranous structures, in striking contrast to the phagocytic membrane surrounding an invading bacterium (Fig. 2A). Occasionally, some bacteria enclosed by lamellar membranes were also surrounded by structures like onion skin (Fig. 2A, arrows). Six hours after invasion, some bacteria enclosed by an onion skin–like structure had become indistinct (fig. S5A). At the same stage of infection, WT bacteria generally lacked the lamellar membranes. Instead, most had long actin tails, as seen in motile bacteria (fig. S5B). To confirm that the lamellar membranes surrounding ΔicsB were autophagic membranes, MDCK/pGFP-LC3 cells infected with ΔicsB 4 hours after infection were analyzed by immunogold EM with antibody against GFP. The lamellar membrane around ΔicsB in MDCK/pGFP-LC3 cells was specifically labeled with immunogold (Fig. 2B, arrows). Onion skin–like membranous structures associated with the bacterium were also highly labeled with immunogold (Fig. 2B, arrowheads). Thus intracellular Shigella lacking IcsB readily succumbs to autophagy within epithelial cells.

Fig. 2.

Multilamellar structures enclosing ΔicsB in MDCK cells. (A) MDCK cells infected with ΔicsB were observed by EM 4 hours later. Multilamellar structures surround the bacterium (arrows). (B) GFP-LC3 (arrows) is localized in the multilamellar structures (arrowheads) surrounding the bacterium, under immunogold EM. Scale bars, 0.5 μm.

To explore the role of IcsB, we assessed whether IcsB is linked to autophagic proteins. We constructed COS-7 transfectants expressing GFP-Atg5, GFP-Beclin (Atg6), GFP-LC3 or Myc in complex with the vacuolar protein-sorting protein Vps34, that is Myc-Vps34 (type III phosphatidylinositol 3-kinase), and each cell lysate was placed in a pull-down assay with glutathione S-transferase (GST) in complex with IcsB. However, under these experimental conditions, none was precipitated by GST-IcsB. Intriguingly, we noted that the LC3 and Atg5 signals in the vicinity of the bacterial body of ΔicsB tended to be distributed asymmetrically, with signals occasionally accumulating at one pole of the bacterium. Furthermore, lamellar membranes associated with ΔicsB in EM were also frequently seen in the area at one end of the bacterial body, which reminded us of the asymmetric distribution of VirG (IcsA) on Shigella required for actin-based intracellular motility (15). We tested whether some autophagic component(s) might directly associate with VirG and so possibly trigger the development of autophagy around the bacterium. Cell lysates prepared from COS-7 cells expressing GFP-Atg5, GFP-Beclin, GFP-LC3, or Myc-Vps34 were pulled down with GST-VirGα1 (the surface-exposed VirG portion) (16), and the bound proteins were analyzed by immunoblotting. Note that only Atg5 was reproducibly precipitated with GST-VirGα1 (fig. S6A). Furthermore, the binding capacity was also confirmed by performing immunoprecipitation experiments in 293T cells expressing Atg5-Myc and GFP-VirGα1 (fig. S6B), which raised the possibility that Atg5 might have some affinity for VirG. To pursue this, BHK cells expressing GFP-Atg5 (BHK/pGFP-Atg5) were infected with ΔicsB for 4 hours and investigated for association of the bacterium with GFP-Atg5. The GFP-Atg5 signals were occasionally confined to one pole of the bacterium. It is noteworthy that when BHK/pGFP-Atg5 cells were infected with virG mutant (ΔvirG), the association of the Atg5 signal was barely detectable, but the signal was restored after infection with ΔvirG/pVirG (Fig. 3A). The Apg5 signal occasionally associated with VirG at one pole of the bacterium (Fig. 3B).

Fig. 3.

VirG affinity for Apg5 is inhibited by IcsB. (A) BHK/pGFP-Atg5 cells infected with indicated Shigella strains for 4 hours were stained with antibody against Shigella (red). (B) BHK/pGFP-Atg5 cells infected with ΔvirG/pVirG for 4 hours were stained with antibody against VirG (red). (C) MDCK/pGFP-LC3 cells infected with indicated Shigella strains for 4 hours were stained with antibody against Shigella. The percentages of GFP-LC3–positive bacteria were quantified. *P < 0.0001. (D) Analysis of IcsB inhibition for Atg5 binding to GST-VirGα1 with the use of cleared lysates of 293T cells expressing Atg5-Myc. (E) Dose-response analysis of IcsB inhibition for Atg5 binding to GST-VirGα1. (F) Determination of VirG domain involved in binding with Atg5 and IcsB (top) and schematic representation of the virGα1 derivatives (bottom). (G) MDCK/pGFP-LC3 cells infected with ΔvirG/pVirG or ΔvirG/pD10-VirG2 for 4 hours were stained with antibody against Shigella (red). Scale bars (A), (B), and (G), 2 μm.

To verify the concept, MDCK/pGFP-LC3 cells infected with WT, ΔicsB, ΔvirG, ΔvirG/pVirG, or ΔvirGicsB (double mutant) were also examined for association of the LC3 signal with the bacteria. LC3-positive ΔvirG accounted for only 1.5%, whereas LC3-positive ΔicsB made up 37% (Fig. 3C). Although LC3-positive ΔvirG/pVirG was increased to 23%, exceeding that from WT (15%), LC3-positive ΔvirGicsB was still less than 2% (Fig. 3C). Consistently, similar results were obtained from the experiment using normal MEFs expressing GFP-LC3 (fig. S7, A and B). Then, we investigated the multiplicity of WT, ΔicsB, ΔvirG, and ΔvirGicsB within BHK cells. ΔvirG grew slightly better than WT within cells and found that the deficiency of intracellular growth of ΔicsB was recovered in ΔvirGicsB, which suggests that VirG is critical for Shigella to trigger autophagy (fig. S8). Atg5 may play a role as the “seed” of isolation membranes in the autophagic process (17). VirG present on Shigella could thus be the target for autophagy, in which IcsB would apparently act as the anti-autophagocytic agents. To assess this possibility, immobilized GST-VirGα1 with or without IcsB was incubated with a lysate of 293T cells expressing Atg5-Myc, and the bound proteins were examined by immunoblotting. IcsB was examined in a pull-down assay with GST-VirGα1 in the presence of Atg5; Atg5-Myc was precipitated in a pull-down assay only when IcsB was absent (Fig. 3D). The extent of His-Atg5 binding to GST-VirGα1 1 decreased as we raised the amount of purified IcsB added, which suggests that Atg5 binding to VirG was inhibited by IcsB in a dose-dependent manner (Fig. 3E). To define the VirG region involved in binding, various truncated VirGα1 versions (VirGα1 through to VirGα8) (16) were tested for the ability to bind to either His-Atg5 or IcsB by using the GST-VirG pull down assay. Both His-Atg5 and IcsB were pulled down by the same series of VirG truncations, which suggests that they share the amino acid residues 320 through 433 of VirG, which are involved in binding (Fig. 3F). Indeed, in MDCK/pGFP-LC3 cells infected with ΔvirG/pD10-VirG2 (pVirG lacking VirG internal amino acid residues 319 to 507) but not ΔvirG/pVirG (full-length VirG), the GFP-LC3 signals were barely visible around bacteria (Fig. 3G). Thus the interaction of IcsB with VirG by means of the putative VirG targeting sequence for Atg5 might be a mechanism for eluding the development of autophagy in the vicinity of the bacterial surface in mammalian cells.

To demonstrate whether VirG itself was a target for autophagy, BHK cells expressing Myc-LC3 (BHK/pMyc-LC3) were infected with an E. coli K-12 strain expressing VirG plus GFP (E. coli/pSU-GFP/pUC-VirG) or GFP (E. coli/pSU-GFP) together with ΔvirGicsB (invasion-positive but autophagy-negative S. flexneri used as the carrier, which allowed E. coli to move into the host cytoplasm). The association with LC3 was investigated. In cells infected with E. coli/pSU-GFP/pUC-VirG but not E. coli/pSU-GFP, bacterial signals were localized with Myc-LC3 signals, and the VirG-associated signals merged with the GFP-LC3 signals (Fig. 4). It is thus highly likely that VirG presented on the Shigella surface in the host cytoplasm triggers autophagy.

Fig. 4.

VirG triggers autophagy. BHK/pMyc-LC3 coinfected with E. coli/pSU-GFP/pUC-VirG and ΔvirGicsB at a ratio of 10:1 for 2 hours were stained with antibody against Myc (red) (top). As a control, E. coli/pSU-GFP was used. (Bottom) BHK/pGFP-LC3 cells were coinfected with E. coli/pUC-VirG and ΔvirGicsB as described above, and stained with antibody against VirG (red) (bottom). Scale bar, 2 μm. Note that as E. coli/pUC-VirG expressed a high level of surface VirG, the VirG signal was visible over the entire bacterial surface.

Shigella VirG can thus be targeted for autophagy perhaps through its affinity for Atg5, and IcsB can interfere with this autophagic process. The role of IcsB in bacterial infection differs from other known TTSS-secreted bacterial effectors, in that its role is to camouflage its own bacterial target molecule (VirG) from the autophagic host defense system.

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Table S1


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article