A system for identifying post-invasion functions of invasion genes: requirements for the Mxi–Spa type III secretion pathway of Shigella flexneri in intercellular dissemination


  • Raymond Schuch,

    1. Department of Microbiology and Immunology, F. Edward Hébert School of Medicine, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814-4799, USA.
    Search for more papers by this author
  • Robin C. Sandlin,

    1. Department of Microbiology and Immunology, F. Edward Hébert School of Medicine, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814-4799, USA.
    Search for more papers by this author
  • Anthony T. Maurelli

    1. Department of Microbiology and Immunology, F. Edward Hébert School of Medicine, Uniformed Services University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814-4799, USA.
    Search for more papers by this author

Anthony T. Maurelli. E-mail amaurelli@usuhs.mil; Tel. (+1) 301 295 3415; Fax (+1) 301 295 1545.


Invasion and intercellular spread are hallmarks of Shigella pathogenicity. Invasion of the eukaryotic cell cytosol requires a type III secretion system (Mxi–Spa) and its cognate set of secreted Ipa invasins. Once intracellular, the IcsA protein directs a form of actin-based motility that helps to drive intracellular bacterial movement, formation of cellular protrusions and cell-to-cell spread. Work in our laboratory has focused on identifying additional factors required for this intercellular form of dissemination. In this study, we sought to identify novel contributions of the type III secretion pathway to post-invasion-specific processes, distinct from its previously characterized roles in invasion. Studies of post-invasion Ipa and Mxi–Spa functions are complicated by an absolute requirement for these virulence proteins in invasion. To circumvent this problem, we developed a system called TIER (for test of intracellular expression requirements), whereby specific ipa, mxi or spa loci are transiently expressed before infection of tissue culture cell monolayers (thus supporting invasion), but then repressed after invasion in the intracellular environment. Such invasive type III secretion mutants (called TIER mutants) were severely restricted in their ability to spread intercellularly and form plaques in confluent tissue culture cell monolayers. Intercellular spread defects were associated with the repression of most type III pathway components examined, including structural (MxiM and Spa33), secreted effector (IpaB, IpaC and IpaD) and regulatory elements (VirF and VirB). A kinetic analysis of bacterial growth in L2 cell monolayers showed that each of the TIER mutants was defective with respect to long-term intracellular proliferation and viability. Examination of TIER mutant-infected monolayers by electron microscopy revealed that the type III pathway was required for a late step in intercellular spread — bacterial escape from protrusion-derived, double-membrane-bound vacuoles. The TIER mutants were eventually degraded in a process involving vacuolar acidification. Based on these findings, we propose that Ipa secretion via Mxi–Spa is required in the protrusion vacuole for double-membrane lysis.


Shigella flexneri is one of the causative agents of bacillary dysentery, an invasive disease of the human colonic epithelium marked by an intense inflammatory reaction and subsequent mucosal destruction. The molecular and cellular basis for the pathogenesis of S. flexneri has been investigated primarily using in vitro tissue culture cell infection assays. These studies helped to define a multistep infection cycle in epithelial cells, whereby extracellular shigellae are internalized, escape from the phagocytic vacuole, multiply intracellularly and ultimately elaborate a form of actin-based movement (Sansonetti, 1998). Actin-based motility directs random intracytoplasmic movement and eventually generates bacteria-tipped cellular protrusions that facilitate intercellular dissemination. During cell-to-cell spread, protrusions containing bacteria are engulfed by adjacent uninfected cells, leaving the bacteria transiently contained within double host membrane-bound vacuoles. Lysis of the protrusion vacuoles releases Shigella into the cytosol of the secondarily infected cell and allows initiation of another round of intercellular spread. A similar strategy for intracellular movement and intercellular spread is used by Listeria, Rickettsiae and Vaccinia virus (Dramsi and Cossart, 1998).

Central to the invasive phenotype of Shigella and its ability to induce disease is the Mxi–Spa type III secretion pathway and its set of four secreted Ipa substrates (Parsot and Sansonetti, 1996). Type III secretion pathways are multicomponent, envelope-associated protein complexes of Gram-negative bacteria that are dedicated to the transmembrane traffic of pathogenic molecules (Hueck, 1998). Sets of well-conserved structural elements, comprising distinct type III pathways, have now been identified in a broad range of bacteria, including Shigella spp., Yersinia spp., Salmonella enterica, Pseudomonas aeruginosa, enteropathogenic and enterohaemorraghic Escherichia coli, Chlamydia spp., Bordetella bronchiseptica and a range of plant pathogenic bacteria. In general, type III secretion provides the means of delivering bacterial effector proteins to the eukaryotic cell surface or intracytoplasmic targets. Typically, effectors disrupt eukaryotic signal transduction pathways and elicit alterations in host cell morphology or physiology. These alterations serve to promote bacterial survival in either host cells or host tissues.

The ipa, mxi and spa loci are part of a group of 32 genes encoded within a 30 kb ‘invasion region’ of the 200 kb virulence plasmid of Shigella (Maurelli et al., 1985; Sasakawa et al., 1988). Expression of invasion genes is co-ordinately induced in response to environmental stimuli by the transcriptional regulators, VirF and VirB (Dorman and Porter, 1998). The Mxi–Spa secretory system, once synthesized and assembled, is maintained in a nearly inactive state until bacteria physically contact target host cells (Ménard et al., 1994; Watarai et al., 1995), probably in the colonic or rectal mucosa. This interaction triggers massive Ipa release into the extracellular milieu where they associate, either individually or as a complex, with the host cell membrane (Ménard et al., 1994; De Geyter et al., 1997; Sansonetti, 1998). Host cytoskeletal and membrane rearrangements ensue, which drive internalization of extracellular bacteria within a vacuole (Ménard et al., 1996a,b). Products of the ipaB, ipaC and ipaD loci then probably degrade the vacuolar membrane and complete the process of cytoplasmic invasion (Sansonetti et al., 1986; High et al., 1992). Not surprisingly, most derivatives of Shigella that lack a complete type III pathway (i.e. ipa, mxi, spa, virF and virB mutants) cannot mediate internalization and direct penetration of the host cytosol (Allaoui et al., 1993; 1995; Ménard et al., 1993; Sasakawa et al., 1993; Schuch and Maurelli, 1999). Whether or not Ipa secretion via Mxi–Spa is required for processes subsequent to invasion is unknown.

Type III secretion by most animal pathogens is required either for inhibition of phagocytosis/survival in the extracellular environment or for invasion/survival in the phagosomal environment. Mxi–Spa of Shigella, which mediates direct bacterial entry into the host intracytoplasmic matrix, is an exception. This novel capacity may be related to another unique feature of the Mxi–Spa pathway: it does not appear to inject type III effectors directly into the host from extracellular positions. For these reasons, Shigella presents a system in which to study type III secretion functions and host cell manipulations not observed with other pathogens. In the study presented here, we sought to identify another novel function of the Mxi–Spa type III secretion system. A method was developed whereby specific contributions of type III secretion to the process of intercellular dissemination could be studied (independent of any effects on invasion). This method, called the TIER system (for test of intracellular expression requirements), consists of a period of induced expression of an intact type III secretion system (thus allowing invasion), followed by a period of repressed expression. The net result is delivery of type III secretion mutant derivatives of Shigella into the cytosol of non-phagocytic host cells. Using the TIER system, we identified absolute requirements for the entire type III pathway of S. flexneri in the process of intercellular spread. In the absence of a functional type III secretion pathway, intracellular shigellae are unable to escape from the double-membrane-bound vacuole formed upon protrusion engulfment and eventually lose viability. Our results implicate the Ipa proteins, secreted via Mxi–Spa, as specific effectors of protrusion vacuole lysis during the intercellular dissemination phase of Shigella infection.


Development of a system for analysing post-invasion functions for Shigella type III secretion pathway components

We hypothesized that post-invasion Ipa secretion via the Mxi–Spa type III system contributes to the intercellular dissemination of S. flexneri in a significant manner. Identification of such contributions, however, is complicated by the non-phagocytic nature of tissue culture cell lines used to evaluate intercellular spread defects and the fact that mutant derivatives lacking type III secretion pathway components (including products of the ipa, mxi, spa, virF and virB loci) are generally non-invasive. To circumvent this problem, we developed transiently invasive Shigella derivatives that, after entry into the target cell cytosol, downregulate the expression of specific type III secretion pathway components. The effects of this repression on the intercellular spread phenotype were evaluated using a modified version of the Shigella plaque assay (see Experimental procedures).

The study of transient invasion gene expression in the course of a plaque assay is referred to herein as the TIER system (for test of intracellular expression requirements) and is based on the use of plasmid-borne fusions of wild-type ipa, mxi, spa, virF and virB loci to the arabinose-inducible/glucose-repressible PBAD promoter. These fusions are introduced into their respective ipa, mxi, spa or vir null mutant backgrounds and maintained in trans. The resulting strains, called TIER mutants, are phenotypically wild type when grown in LB containing arabinose. After growth in LB with arabinose, the TIER mutants are then transferred to a glucose-rich infection medium (DMEM) lacking arabinose and applied to tissue culture cell monolayers. Subsequent to infection, two events are expected to occur in order: (i) rapid bacterial internalization owing to type III pathway expression during growth in LB with arabinose; and (ii) repression of PBAD-directed expression of a complementing wild-type ipa, mxi, spa, virF or virB gene (i.e. conversion to a type III pathway mutant). The effects of this conversion on the ability of intracellular bacteria to spread intercellularly can then be evaluated.

The validity of the TIER system as a tool for studying post-invasion gene expression requirements was investigated, in part, by analysis of PBADgfp fusion activity in an invasive, wild-type 2457T background (BS586) both before and after L2 cell infection. Before invasion (0 min samples), high levels of arabinose-induced green fluorescent protein (GFP)-directed fluorescence were observed in extracellular BS586 (Fig. 1B). Subsequent to invasion (120 min samples), however, in media lacking arabinose, very little fluorescence remained (Fig. 1D). Diminished fluorescence was not a general feature of GFP in intracellular shigellae, as a 2457T derivative bearing a PLACgfp fusion (BS587) was highly fluorescent at both 0 and 120 min after infection (Fig. 1A and C respectively). Because of the absence of lacI from both S. flexneri and the PLACgfp fusion-bearing vector, expression of GFP in BS587 should be constitutive. The decreased fluorescence observed with intracellular BS586 was therefore attributable to the absence of arabinose after infection and subsequent repression of PBADgfp.

Figure 1.

. PBADgfp expression is repressed in the intracellular environment. Semi-confluent L2 cell monolayers were infected as described in Experimental procedures with either BS586 or BS587 (in the absence of arabinose). GFP expression was controlled by the PLAC promoter of pBluescript in BS587 (A and C) and the PBAD promoter of pBAD18 in BS586 (B and D). Monolayers were fixed immediately after infection (0 min infection samples; A and B) or 120 min after infection (C and D) and stained with Evans blue. The fluorescence images shown (each at 1000 ×) were characteristic of the entire monolayer and represent equivalent exposure times. Bacteria on top of or within L2 cells appear yellow in these images. In (D), there are equivalent numbers of intracellular bacteria to that shown in (C); the low level of GFP-directed fluorescence associated with intracellular BS586 is almost completely masked by the Evans blue fluorescence.

We also analysed a Shigella virulence gene, icsA, which is required for intercellular spread and the formation of plaques in confluent tissue culture cell monolayers (Bernardini et al., 1989). After growth in LB with arabinose, an icsA strain (BS543) bearing a PBADicsA+ fusion in trans yielded high levels of unipolarly localized IcsA (data not shown). While fully induced at the outset of infection, the subsequent ability of this strain to form plaques required the post-invasion addition of arabinose (Table 1). When arabinose was not added after invasion, plaque formation was not observed. As expected, the icsA strain bearing a constitutive PLACicsA+ fusion, which should maintain IcsA expression after invasion, was fully complemented with respect to plaque formation (Table 1).

Table 1. . Requirements for Shigella virulence loci in intercellular spread (TIER analyses results).a a. All indicated values represent the average of at least three to five independent experiments.b. Wild-type S. flexneri backgrounds are listed above their respective mutant derivatives used in this study.c. Plasmids encoding the indicated promoter–virulence gene fusions were introduced into each strain. PBAD fusions are in plasmid pBAD18. PLAC fusions are in pBluescript. The ‘none’ designation indicates that no plasmid is present in that background.d. Indicates abilities of each strain to invade confluent and semi-confluent L2 cell monolayers. +, denotes the ability to invade efficiently. ±, denotes the ability to invade at levels 50% that of the wild-type. −, denotes the complete inability to invade L2 cells.e. This number was calculated using the formula (number of plaques/number of input bacteria) × 100.f. Average size of ≈ 50 independent plaques.g. NA, not applicable; ND, not done.Thumbnail image of

Thus, the PBAD promoter fusion to icsA, which was fully induced at the start of infection, could be repressed in the intracellular environment as a result of the absence of arabinose in the plaque assay overlay. This condition was sufficient to prevent complementation of the icsA intercellular spread defect. By extension then, preinvasion expression of PBAD in TIER analyses of ipa, mxi, spa, virF or virB loci should be sufficient to support bacterial internalization. The subsequent absence of arabinose in the plaque assay should cause expression to fall below threshold levels required to complement putative post-invasion activities.

Requirements for Mxi–Spa type III secretion pathway components in intercellular spread

The TIER system was used to analyse post-invasion functions for three classes of type III pathway genes: those encoding (i) membrane-associated structural elements (mxiM and spa33); (ii) secreted Ipa effectors (ipaA, ipaB, ipaC, ipaD); and (iii) transcriptional regulators of type III system induction (virF and virB). All mutant backgrounds analysed (except for the ipaA::Tn5 strain BS572) were non-invasive and unable to form plaques (Table 1). Introduction of either PBAD or PLAC promoter::complementing wild-type gene fusions into these mutants (except ipaA) restored invasion to wild-type-like levels. The PBAD fusion-bearing backgrounds, however, were completely unable to form plaques. PBAD promoter activity was probably insufficiently maintained to support an essential post-invasion process. Complementation of this activity was observed only through the post-invasion addition of arabinose or use of the constitutively expressed PLAC fusion-bearing vectors.

The induction/repression of PBAD activity had little effect on the invasion and intercellular spread of the ipaA TIER mutant (Table 1). An ipaA mutant has been shown previously to be both invasive and competent for intercellular spread (Tran Van Nhieu et al., 1997). Our findings indicate that there is nothing inherent in the TIER system (such as the induction or repression of PBAD), which alone prevents plaque formation. In agreement with this, the intracellular repression of a PBADgfp fusion in BS586 had no effect on the efficiency of plaque formation (data not shown).

These results demonstrate that type III secretory pathway components, which are induced during growth and required for invasion, must also be expressed after invasion to support intercellular spread. The Ipa proteins, secreted via the Mxi–Spa system, probably catalyse at least one of the essential steps required for intercellular spread.

PBAD-directed gene expression can be efficiently repressed in intracellular bacteria

To confirm that PBAD promoter activity in each of the type III secretion TIER mutants is repressed after invasion in the absence of arabinose, we compared levels of PBAD and PLAC fusion-encoded proteins in whole-cell bacterial extracts isolated just before L2 cell infection (extracellular bacteria) or 3 h after infection (intracellular bacteria). As a control, we prepared and analysed similarly samples from wild-type strains, in which expression was directed by native virulence plasmid promoters.

In extracellular bacteria, the levels of virulence proteins expressed from PBAD and PLAC were surprisingly similar to each other and to that expressed from native promoters (Fig. 2, lanes 1–3). In each of the backgrounds examined (except for the virF TIER mutant), expression was evaluated using antisera specific for the vector fusion-encoded protein. For the virF TIER mutant, expression of VirF was evaluated indirectly using antisera to IcsA, IpaB and IpaC. VirF, which is a regulator of virulence gene transcription, must be examined in this manner in the absence of VirF antisera. These findings indicated that, at the time each strain was added to tissue culture cell monolayers, each produced high levels of the virulence proteins of interest.

Figure 2.

. Expression of PBAD–virulence gene fusions is repressed in the intracellular environment. Bacterial populations were obtained just before infection of confluent L2 cell monolayers (extracellular bacteria; lanes 1–3) or from the L2 intracellular environment 180 min after infection (intracellular bacteria; lanes 4–6). No arabinose was present during these infections. Samples were isolated from a wild-type parental strain (lanes 1 and 4) or from a mutant background, indicated at the upper lefthand corner, bearing either a PBAD (lanes 2 and 5) or a PLAC (lanes 3 and 6) fusion in trans. Total bacterial proteins from these populations were analysed by immunoblot using antisera against the protein listed next to each arrow. In each blot, protein from equivalent numbers of bacteria was loaded per lane (for analysis of MxiM, 1 × 106 per lane; for IpaB, 2.5 × 105 per lane; for IcsA, 5 × 107 per lane; for IpaC, 2.5 × 105 per lane; and for IpaD, 2.5 × 106 per lane). The arrows indicate the positions of each of the proteins analysed.

In intracellular bacteria recovered 3 h after infection, expression levels directed by the native and PLAC promoters (Fig. 2, lanes 4 and 6 respectively) still remained similar to each other and were only slightly lower than that observed before invasion. Expression directed by the PBAD promoter in the mxiM, ipaB, ipaC and ipaD TIER mutant backgrounds was barely detectable (Fig. 2, lane 5). This repression did not impair the expression of other virulence factors in these backgrounds: each of the ipa mutants still yielded wild-type-like levels of MxiM, while the mxiM mutant yielded wild-type-like levels of IpaB (data not shown). In the virF TIER mutant, IpaB and IpaC levels were reduced or absent, while IcsA expression remained unchanged. This finding suggested that the intercellular spread defect of the virF TIER mutant was mediated through the repression of type III secretion (like the other TIER mutants) and not through IcsA-directed actin polymerization.

As predicted, the PBAD promoter is quite active at the outset of infection (in extracellular bacteria) and is repressed at later time points after the removal of arabinose (in the intracellular environment). This regulated expression probably accounts for the initial invasiveness of type III secretion TIER mutants and the subsequent intercellular spread defects.

Intracellular growth of type III secretion mutants

To determine whether the Mxi–Spa type III secretion pathway is required for intracellular survival and multiplication, a kinetic analysis of bacterial growth was performed in semi-confluent L2 cell monolayers. Both the wild-type strain 2457T and the mxiM TIER mutant grew steadily for the first several hours after infection (Fig. 3A), with a doubling time (40 min) similar to that reported previously for Shigella (Sansonetti et al., 1986). At time points up to 9 h after infection, similar numbers of viable bacteria were recoverable from each strain. As these experiments were performed in the absence of arabinose, MxiM was not required for intracellular growth in primarily infected L2 cells. Similar results were also obtained using TIER mutants lacking other type III pathway components (data not shown).

Figure 3.

. Kinetic analysis of intracellular bacterial growth. At the indicated time points, infections were stopped, and the number of recoverable bacterial colony-forming units was determined. A. The intracellular growth rates of 2457T (□) and the mxiM TIER mutant (○) in semi-confluent monolayers. B. The intracellular growth rates of 2457T (□) and the mxiM (⋄) and virF (○) TIER mutant derivatives in confluent monolayers. C. The intracellular growth rates of M90T (□) and the ipaC (⋄) and ipaD (○) TIER mutant derivatives in confluent monolayers. No arabinose was present during these infections. All values are averages of at least three independent experiments.

Intracellular bacterial growth kinetics were also followed after infection of confluent L2 cell monolayers. Tight interactions exist between tissue culture cells in confluent monolayers, which are required for the passage of Shigella protrusions. As such, confluent monolayers support cell-to-cell spread and plaque formation, while semi-confluent monolayers do not. The wild-type strains 2457T and M90T both grew steadily for a 21 h period in the intracellular environment of confluent L2 cells (Fig. 3B and C). In the absence of arabinose, the mxiM, virF, ipaC and ipaD TIER mutants sustained wild-type-like growth rates for only 2.5–5 h after infection. After longer incubations, a pronounced loss of bacterial viability became apparent. By 21 h, the number of recoverable TIER mutants was roughly 100-fold lower than that of wild-type strains. These growth defects were a consequence of the lack of arabinose, as the post-invasion addition of arabinose promoted wild-type-like growth over the duration of the experiment (data not shown). These results indicated that a functional type III Mxi–Spa pathway, while not required for intracellular multiplication per se, was required for long-term survival and proliferation in confluent monolayers.

Requirements for the Mxi–Spa pathway in protrusion membrane escape

To investigate the inability of intracellular type III secretion mutants to spread intercellularly, L2 cells infected with either wild-type bacteria or TIER mutant strains were analysed by transmission electron microscopy (TEM). Infections proceeded for 5 h in the absence of arabinose, before sample processing and visualization. For each strain, intracellular bacteria were examined and classified as belonging to one of four different classes (Table 2). Most of the wild-type bacteria (2457T and M90T) observed were either free within the host cytosol (cytoplasmic class) or in the process of intercellular spread (internalized protrusion class). Those spreading were generally within protrusions containing only one or two bacteria; in most cases, the bacterial envelope was closely apposed to the surrounding double-membrane structure (Fig. 4A). Most of the type III secretion mutants observed were also free within the host cytosol; however, a few fell within the internalized protrusion class (Table 2). Of the mutants within internalized protrusions, most were in protrusions bearing multiple bacteria (Fig. 4B). The decrease in the internalized protrusion class was associated with a corresponding increase in the number of mutants observed in abortive protrusion structures (protrusions that were either non-internalized or appeared to enclose dead or dying bacteria). Intracellular bacteria in this latter class generally appeared to be undergoing either a process of cytoplasmic condensation [shrinkage of the cytoplasmic matrix away from the bacterial envelope (seen with the mxiM TIER mutant in Fig. 4C)] or membrane lysis (data not shown). Additionally, the protrusion vacuoles containing such bacteria accumulated a very dense or osmiophilic, granular substance that contained extensive membranous material. These structures were observed not only with the mxiM TIER mutants, but with the virF, ipaC, ipaD and spa33 TIER mutants as well (Fig. 4D and E and data not shown respectively). The appearance of these compartments somewhat resembles that of phagolysosomes containing degraded shigellae within enterocytes of ligated rabbit ileal loops (Polotsky et al., 1994). Shigella associated with such phagolysosomal structures are ultimately killed and degraded. This was also the fate of the intracellular type III secretion mutants, as observed in monolayers examined 7.5 h after infection (data not shown). At this extended time point, numerous structures were observed that appeared to be double-membrane-bound vacuoles containing the remnants of bacteria (cytoplasmic debris and membrane material).

Table 2. . Classes of intracellular type III secretion TIER mutants.a a. For each strain, at least 100 intracellular bacteria were classified as cytoplasmic (bacteria not associated with host membrane structures), in an internalized protrusion (bacteria completely or partially surrounded by a double host cell membrane), in a non-internalized protrusion (bacteria projecting from the surface of an infected cell and surrounded by one cellular membrane) or in an internalized protrusion containing bacteria undergoing gross morphological alterations (double host membrane completely or partially encompassing dense osmiophilic material, membranous matter and bacteria that appear to be condensing or lysing).Thumbnail image of
Figure 4.

. Transmission electron micrographs of L2 cells infected with wild-type and TIER mutant strains. A. Wild-type strain 2457T partially surrounded by a double membrane at 5 h after infection. B. Three mxiM TIER mutants in a double-membrane-bound vacuole at 5 h after infection. C. Three mxiM TIER mutants in a double-membrane-bound vacuole at 5 h after infection. The bacterial cytoplasm is condensing away from the bacterial envelope (note region near arrowhead). Dense osmiophilic material and membranous structures fill the vacuolar matrix. D. Two virF TIER mutants in a double-membrane-bound vacuole at 5 h after infection. E. One ipaC TIER mutant in a double-membrane-bound vacuole at 5 h after infection. F. Two ipaD TIER mutants in a double-membrane-bound vacuole at 5 h after infection. No arabinose was present during these infections. The bar in each field is 500 nm.

Inhibition of host V-ATPases blocks the intracellular growth defect of the mxiM TIER mutant

Our results suggested that type III secretory functions were required for escape from the protrusion vacuole. In the absence of type III secretion, the TIER mutants die within structures that are similar in appearance to that formed upon vacuolar maturation and fusion with vesicles of the endocytic pathway. To determine whether a process associated with vacuolar maturation (i.e. acidification) contributed to the decreased viability of intracellular type III secretion mutants, we examined the impact of the antibiotic bafilomycin A1 on the intracellular growth rate of 2457T and the mxiM TIER mutant. Bafilomycin A1 specifically inhibits host vacuolar-type H+-ATPases (V-ATPases) and prevents the acidification of lysosomes and the activation of lysosomal hydrolases (Yoshimori et al., 1991). The addition of bafilomycin A1 to confluent L2 cell monolayers 4 h after infection did restore the growth rate of the mxiM TIER mutant to near wild-type-like levels (Fig. 5). These results suggested that acidification of protrusion membrane vacuoles promotes the loss of viability associated with intracellular type III secretion mutants.

Figure 5.

. The effect of bafilomycin A1 on the intracellular growth rate of wild-type and mxiM TIER mutant strains. At 4 h after infection (indicated by the arrow), bafilomycin A1 was added to 2457T (□)- and mxiM TIER mutant (○)-infected confluent monolayers. The intracellular growth rate of the mxiM TIER mutant to which no bafilomycin A1 was added is also shown (Δ). At the indicated time points, infections were stopped, and the number of recoverable bacterial colony-forming units was determined. No arabinose was present during these infections. All values are averages of at least three independent experiments.

Filamentous structures link the bacterial envelope with the protrusion vacuole

The TEM analysis of intracellular 2457T revealed an intimate association between the bacterial envelope and the inner face of host protrusion membranes (Fig. 6A). This tight association is necessary for efficient cell-to-cell spread (Sansonetti et al., 1994; Allaoui et al., 1995). In confluent L2 monolayers examined 2.5 h after infection with the mxiM TIER mutant strain, the bacterial–host interaction within protrusions was only loosely maintained over the bacterial surface and was often separated by a thin electron-transparent halo (data not shown). This interaction was completely absent in samples examined 5 h after infection. These results suggested that the force maintaining tight bacterial–host interactions in the protrusion vacuole is slowly lost after repression of the Mxi–Spa pathway. This force may be provided directly by elements of the type III secretion pathway. Consistent with this hypothesis, in the 2.5 h after infection samples, the zones of loose interaction between the bacterial envelope and the inner protrusion membrane were often connected by a loose network of filamentous material (Fig. 6B). These structures extended up to ≈ 120 nm in length and were ≈ 5 nm wide. S. flexneri has neither flagella nor pili, but is known to elaborate filamentous structures via the Mxi–Spa pathway in response to signals that mimic host cell contact (Parsot et al., 1995). EPEC and S. typhimurium also elaborate surface filamentous structures via their type III pathways, which contact the host cell surface and may translocate effector molecules into the host (Ginocchio et al., 1994; Knutten et al., 1998). The filamentous material that we observed within the protrusion vacuoles may therefore represent structures elaborated by Mxi–Spa, which serve to mediate bacterial–host interactions and/or deliver pathogenic molecules to either the protrusion membrane or the host cytosolic matrix.

Figure 6.

. Transmission electron micrographs of 2457T and the mxiM TIER mutant in double-membrane-bound vacuoles. At 2.5 h after infection (in the absence of arabinose), infected confluent monolayers were fixed and processed. A. Wild-type strain 2457T in close contact over much of its surface with the inner face of the protrusion vacuole. B. The mxiM TIER mutant is separated from the protrusion membranes by an electron-transparent halo interspersed with thin filamentous structures (note region near arrowhead). These fibrils were visualized within many of the protrusion vacuoles examined at this time point. The bar in each field is 200 nm.


Host–pathogen interactions characteristic of the onset of shigellosis are controlled, in part, by the Ipa invasins and their dedicated type III secretion pathway, consisting of ≈ 20 Mxi–Spa components. Genetic and/or biochemical studies have implicated the Ipas exclusively in the early stages of infection, which include the induction of cytoskeletal rearrangements responsible for bacterial internalization and rapid escape from the phagocytic vacuole formed upon entry (Sansonetti, 1998). Unlike entry, few effectors of intercellular spread are known. IcsA is the best-characterized protein required for cell-to-cell spread and is responsible for the unipolar polymerization of actin filaments and intracellular bacterial movement. An icsA mutant and mutants that alter IcsA surface presentation or distribution are unable to spread properly intercellularly and do not form plaques in tissue culture cell monolayers (Bernardini et al., 1989; Sandlin et al., 1995; Egile et al., 1997). The icsB locus has been implicated in intercellular spread as well, but is no longer believed to be required for such a function (P. J. Sansonetti, personal communication). MxiG is also required for the process of intercellular spread, although its effects are not exerted through alterations in Ipa secretion (Allaoui et al., 1995). A MxiG derivative in which an internal RGD motif is replaced by RAD can secrete the Ipas and is invasive, but is unable to mediate proper bacterial contact with the protrusion membrane and, as such, forms abnormally small plaques. Other mutations that affect cell-to-cell spread include those that specifically inhibit intracellular septation (MacSíomóin et al., 1996), alter cytochrome bd expression (Way et al., 1999), destabilize the cytoplasmic membrane or peptidoglycan layer (Hong et al., 1998) or prevent periplasmic oxidative protein folding (Yu, 1998).

No experiments have yet been reported that specifically implicate Ipa secretion in any post-invasion functions within epithelial cell or epithelial cell-like lines. Ménard et al. (1994) did show that host cell contact-induced Ipa secretion does not release all cytoplasmic Ipas, suggesting that Ipa pools may be retained for intracellular functions. Additionally, Demers et al. (1998) have shown that activity of the mxi operon promoter, as well as virulence gene promoters that require Ipa secretion for activation, are well maintained in the intracellular environment. These findings suggest that Ipa secretion may occur in the intracellular environment and are in agreement with our immunoblot analyses in Fig. 2, in which we observed comparable levels of type III system elements in wild-type bacteria recovered at both pre- and post-invasion time points. If the Mxi–Spa system is maintained in the intracellular environment, it is probably required for some post-invasion function.

Post-invasion roles for Mxi–Spa are technically challenging to investigate, as the standard assay for evaluating intercellular spread, the Shigella plaque assay, requires that strains examined are invasive. As ipa, mxi, spa, virF and virB mutant strains are generally non-invasive, we sought to construct type III secretion mutant derivatives that were more amenable to analysis. Towards this end, we developed transiently invasive Shigella strains that, after invasion of the host cell cytosol, convert phenotypically into ipa, mxi, spa, virF or virB mutants. Essentially, the TIER system provides the means to deliver invasion mutants to the intracellular environment of non-phagocytic cells.

In TIER analyses of GFP and IcsA, we found that PBAD fusions were induced to high levels before invasion and repressed at later time points in the intracellular environment (repressed to a degree sufficient to block complementation of the icsA intercellular spread defect). This repression was a consequence of the absence of arabinose in post-invasion incubations (i.e. in the plaque assay). The TIER system therefore proved to be suitable for the examination of post-invasion Mxi–Spa functions. Seven type III secretion loci required for invasion, including mxiM, spa33, virF, virB, ipaB, ipaC and ipaD, were subsequently examined. The Ipas are secreted effectors of invasion, while MxiM and Spa33 are required for Mxi–Spa assembly and function. VirF and VirB upregulate transcription of ipa, mxi and spa expression. Induction of these proteins from the PBAD promoter to wild-type-like levels supported invasion by the TIER mutants into either confluent or semi-confluent L2 cell monolayers. The subsequent ability of these strains to sustain PBAD expression in the intracellular compartment and spread intercellularly was completely dependent on the addition of arabinose after L2 cell invasion. In the absence of arabinose, little expression from PBAD was detectable within intracellular bacteria, and no plaque formation was observed. In the presence of arabinose, expression of PBAD was maintained, and intracellular bacteria were capable of forming plaques. The Mxi–Spa type III secretion pathway was therefore required to support an intercellular spread function.

Repression of PBAD in the TIER system probably requires a few rounds of infection before protein levels fall below the threshold level needed to support cell-to-cell spread. Not surprisingly then, microscopic analysis of Giemsa-stained confluent L2 cell monolayers 24 h after inoculation with TIER mutants revealed small foci of up to approximately five lightly to moderately infected cells (data not shown). This block in intercellular spread was unlikely to be related to effects on IcsA expression based on the following observations: (i) TIER mutant strains recovered from L2 cells 3 h after infection produced wild-type-like levels of IcsA (Fig. 2; data not shown); (ii) immunolocalization and actin tail staining experiments indicated no alterations in IcsA localization or activity, respectively, associated with intracellular TIER mutants (data not shown). Rather, the intercellular spread block observed with the type III secretion mutants resulted from a specific defect in bacterial escape from the double host membranes of an internalized protrusion. First, in a kinetic analysis of intracellular TIER mutant growth, a defect in prolonged survival and proliferation was observed in confluent L2 monolayers (which support protrusion engulfment and bacterial spread) but not in semi-confluent monolayers (which do not support the spread process). Secondly, TEM analyses revealed that type III secretion mutants remain trapped in protrusion vacuoles where they multiply and eventually undergo processes of cytoplasmic condensation and membrane lysis. Bacterial killing within protrusion membrane vacuoles probably accounts for both the loss of intracellular bacterial viability observed and the block in plaque formation.

The killing of intracellular shigellae in non-phagocytic cultured cell lines has been observed previously. Electron microscopic studies have been used to show that the dsbA mutant of Shigella lyses in association with protrusion structures in HeLa cells (Yu, 1998). TEM analysis of intracellular TIER mutants suggested a form of death different from that observed with the dsbA mutant, however. Unlike the TIER mutants, the dsbA derivative is killed rapidly upon protrusion internalization, leaving distinct holes throughout the monolayer. The fate of the TIER mutants appeared to be most similar to that of S. sonnei within enterocytes of ligated loops of rabbits immunized with a Shigella polyantigen vaccine 10 days before challenge (Polotsky et al., 1994). These enterocytes bore large osmiophilic phagolysosomes containing extensive membranous matter and partially degraded Shigella, which appeared to be similar in structure to protrusion vacuoles bearing the TIER mutants. Whether the TIER mutants were trapped within phagolysosomes is unknown, although it seems likely that the osmiophilic material we observed within protrusion vacuoles was bacteriocidal. When we added bafilomycin A1, the inhibitor of vacuolar acidification and protein degradation within lysosomes, to TIER mutant-infected monolayers, we no longer observed loss of intracellular bacterial viability. These results suggested that, when type III secretion is blocked, such as in the TIER mutants, Shigella remains trapped in the protrusion and is eventually killed by exposure to toxic host products. A functional type III pathway serves to mediate vacuolar escape and avoidance of any such killing mechanism.

Several lines of evidence suggest that the Ipa proteins, particularly IpaB and IpaC, interact with and possibly disrupt host cell membranes (High et al., 1992; Zychlinsky et al., 1994; Ménard et al., 1996b; De Geyter et al., 1997). For these reasons, we postulated that the inability of the TIER mutants to escape from protrusion vacuoles may reflect a role for secreted Ipas in physically lysing the protrusion membrane. The complete inhibition of plaque formation observed with the TIER mutants is consistent with such an essential role for the Ipas. Listeria spp. escape from a similar type of protrusion structure using a phospholipase (Marquis et al., 1997), for which no homologue in Shigella has been found. It is likely then that Shigella disrupts the double-membrane structure in a manner distinct from that used by Listeria and that involves the Ipas.

The mechanism by which the Mxi–Spa pathway functions in the environment of the protrusion vacuole may have been identified in our electron microscopic analysis of the mxiM TIER mutant. The normally tight bacterial interaction with the inner protrusion membrane observed with wild-type Shigella was not observed in the mxiM TIER mutants. Instead, the bacterial envelope of the mutant was withdrawn from the protrusion membrane and connected only by small, pili-like appendages. These structures appeared to be similar to the extracellular filaments elaborated by the type III systems of EPEC (Knutten et al., 1998) and S. typhimurium (Ginocchio et al., 1994), which are involved in host cell contact and possibly protein translocation into the host. Shigella produces similar appendages upon type III system induction (Parsot et al., 1995) and may therefore also elaborate them within the protrusion vacuole. Of particular interest is the similarity between the putative Mxi–Spa fibrils and the outer membrane surface projections of the type III system-bearing pathogen, Chlamydia psittaci (Matsumoto, 1988; Bavoil and Hsia, 1998). Components of these structures in Chlamydia are predicted to be type III system substrates that function as physical bridges between intravacuolar chlamydiae and the phagosomal membrane (used either to deliver bacterial products to the host or to take up host nutrients). The type III system appendages of Shigella may also be used for interactions between intravacuolar bacteria and the host. Our results suggest that these intravacuolar appendages could serve as protrusion membrane anchors, as their appearance and disappearance (at 2.5 and 5 h after infection respectively) coincided with increasingly diminished interactions between the bacterial envelope and the protrusion inner membrane. The tight bacterial–host interaction within protrusion vacuoles could be necessary for proper Ipa delivery and function with respect to membrane lysis.

In conclusion, the results presented in this report extend our knowledge regarding the Mxi–Spa type III secretion system of Shigella and its role in pathogenesis. The Mxi–Spa system is more versatile than previously thought, required not only for invasion, but also for intercellular spread. Use of the TIER system has enabled us to identify this novel role for type III secretion in Shigella, and it will be exploited further to analyse the process of intercellular spread. Specifically, we hope to examine more closely the interactions between the Ipas and/or putative Mxi–Spa fibrils and the protrusion membranes and to identify additional effectors of intercellular spread. Finally, we also believe that the versatility of the TIER system makes it useful for analyses of other invasive pathogenic bacteria, including Salmonella, Legionella and Listeria, for which post-invasion functions of invasion genes may exist.

Experimental procedures

Bacterial strains and culture conditions

S. flexneri strains used in this study are described in Table 3. All mutants used were derivatives of either 2457T or M90T. Mutant backgrounds bearing a PBAD promoter fusion in pBAD18 are referred to throughout the text as TIER mutants (for example, mxiM/PBAD-mxiM+ is the mxiM TIER mutant). The Escherichia coli strain used for all cloning steps was DH5αλpir (Miller and Mekalanos, 1988).

Bacteria were grown either in L broth (LB) with aeration at 37°C or on L agar plates (1.5% agar) in a 37°C incubator. The tissue culture cell medium used in all infections was Dulbecco's modified Eagle medium (DMEM; Gibco BRL). Antibiotics were used at the following concentrations: ampicillin, 100 μg ml−1; gentamicin, 50 mg ml−1; kanamycin, 50 μg ml−1; bafilomycin A1, 125 nm. Activity of the PBAD promoter was induced during growth in LB supplemented with 0.2% arabinose.

Plasmid and strain constructions

The plasmids and the plasmid-borne PBAD and PLAC promoter fusions used in this study are described in Table 3. Analysis of DNA, plasmid constructions and the transformation of S. flexneri and E. coli were performed according to standard protocols (Sambrook et al., 1989). Polymerase chain reaction (PCR) amplifications for cloning and plasmid-screening purposes were performed using Pfu (Stratagene) and Taq (Qiagen) DNA polymerases, respectively, in a DNA Thermal Cycler 480 (Perkin-Elmer). To confirm the fidelity of PCR reactions, some PCR-generated plasmid inserts were sequenced. Templates for DNA sequencing were prepared using the ABI Prism Dye Terminator Cycle Sequencing Core Kit and analysed using an ABI Prism 377 DNA sequencer (Biomedical Instrumentation Center at USUHS).

TIER analysis

The TIER system is based upon the Shigella plaque assay (Oaks et al., 1985). Wild-type and TIER mutant backgrounds were grown overnight in LB supplemented with the appropriate antibiotics. On the following day, samples were diluted 1:100 in LB containing antibiotics (if necessary) and arabinose, and grown to an OD600 of ≈ 0.8. This growth allowed the induction of PBAD promoter activity. Culture aliquots were removed, washed in phosphate-buffered saline (PBS), resuspended in DMEM and applied to confluent monolayers of mouse L2 fibroblasts in 60 mm tissue culture dishes. The multiplicities of infection (MOIs) were ≈ 6.0 × 10−3 bacteria per cell (an amount sufficient for the wild type to yield 100–300 distinct plaques per dish). For those mutants unable to form plaques, the MOI was increased up to six bacteria per cell. Infected monolayers were incubated on a rocking platform at 37°C under a 6% CO2 atmosphere for 90 min (this is referred to as the invasion period). The monolayers were then washed with PBS and overlaid with a 0.5% agarose solution containing DMEM, 10% fetal calf serum (FCS) and the appropriate antibiotics (including gentamicin). Arabinose was either added or omitted from the overlay as indicated during this post-invasion period. Plaque formation was evaluated after a 2 day incubation at 37°C under 6% CO2 by staining with neutral red.

Invasion assays

Invasion assays using either semi-confluent or confluent L2 cell monolayers were performed and assessed in a manner similar to that described previously (Oaks et al., 1985; Sandlin et al., 1996). Ninety minutes after infection, monolayers were washed with PBS, incubated for 30 min in DMEM containing gentamicin and extensively washed again with PBS. The L2 cells were then lysed with 0.1% Triton X-100 in PBS. Bacteria released from the intracellular environment were collected by low-speed centrifugation and plated for enumeration.

Recovery and analysis of protein from extracellular and intracellular bacteria

Protein expression was assessed during the course of a modified Shigella plaque assay. Multiple aliquots were removed from exponential bacterial cultures (grown in the presence of arabinose), washed with PBS and used for infection of confluent L2 monolayers, a viable bacterial count or to prepare total protein extracts. Bacteria collected for the preparation of protein were resuspended in protein sample buffer. Infected monolayers were incubated for 90 min to allow invasion, washed with PBS and incubated for an additional 90 min in the presence of gentamicin (without arabinose). After extensive washes with PBS, the infected monolayers were lysed with 0.1% Triton X-100 in PBS. Bacteria released from the intracellular environment were recovered by low-speed centrifugation, plated for enumeration and resuspended in protein sample buffer.

Boiled protein samples prepared from the preinfection and post-invasion pools were subjected to SDS–PAGE. Separated proteins were transferred to polyvinylidene difluoride (PVDF) membranes (Schleicher & Schuell) and incubated with a blocking agent. Immunodetection was performed using primary mouse anti-IpaB (Mills et al., 1988), anti-IpaC (Mills et al., 1988), anti-IpaD (E. Oaks) or anti-IcsA (Sandlin et al., 1996) serum and primary rabbit anti-MxiM (Schuch and Maurelli, 1999) serum. The activity of a mouse- or rabbitspecific alkaline phosphatase-labelled secondary antibody was then visualized using the chemiluminescent substrate CDP-Star (Boerinhger Mannheim), as described by the manufacturer.

Fluorescence microscopy

The fluorescence emission of a red-shifted GFP variant called GFPmut2 (Cormack et al., 1996), expressed from either PBAD or PLAC in 2457T, was followed in a modified invasion assay. GFPmut2 emits bright green light (λMAX = 507 nm) when excited with blue light (λMAX = 481 nm). After the invasion assay protocol, strains BS586 and BS587 were grown to mid-log phase in the presence of arabinose. Culture aliquots containing ≈ 7 × 108 cells were then applied to semi-confluent L2 cell monolayers in duplicate 35 mm dishes. After centrifugation of bacterial samples onto the monolayers, one each of the duplicate dishes was immediately washed three times with PBS and fixed with 3% formalin in PBS at 4°C. These dishes are referred to as the 0 min infection samples. The remaining dishes were incubated for 90 min at 37°C to allow for bacterial invasion. After invasion, the monolayers were washed with PBS, incubated in DMEM containing gentamicin for 30 min, washed again with PBS and fixed with 3% formalin in PBS at 4°C. These dishes are referred to as the 120 min infection samples. After fixation, both the 0 and 120 min infection samples were stained with 0.0015% Evans blue (Sigma) in PBS for 15 min at room temperature. Samples were washed with PBS and visualized with an Olympus BX50 fluorescence microscope (using a mercury lamp as the light source). An Olympus U-MWIB filter cube was used to co-visualize the green fluorescence from GFPmut2 and the orange–red fluorescence from Evans blue.

Assays for growth of intracellular bacteria

Cultures were grown to mid-late log phase in LB containing arabinose. Standardized aliquots were removed, washed with PBS and applied to either semi-confluent (MOI of ≈ 100) or confluent (MOI of ≈ 10) L2 cell monolayers in DMEM. After a 45 min invasion period, samples were washed with PBS and incubated further in DMEM containing gentamicin (with or without arabinose, as indicated) for the indicated lengths of time. All incubations were carried out at 37°C under a 6% CO2 atmosphere. At the indicated intervals, monolayers (in triplicate) were lysed with 0.1% Triton X-100 in PBS, and the recovered bacteria were enumerated.

Inhibition of vacuolar acidification, mediated by bafilomycin A1, was examined during the course of bacterial growth in confluent L2 monolayers. The inhibitor was prepared, stored and used as described (Marquis et al., 1997).

Transmission electron microscopy

Confluent L2 cell monolayers were infected with 10 bacteria per cell as described above for the intracellular growth studies. At either 5 or 7.5 h after infection (in the absence of arabinose), samples were fixed and processed for analysis essentially as described by Mounier et al. (1990). Ultrathin sections, collected on copper grids, were examined and photographed using a Phillips CM-100 electron microscope.


  1. The opinions or assertions contained herein are the private ones of the authors and are not to be construed as official or reflecting the views of the Department of Defense or the Uniformed Services University of the Health Sciences.


This work was supported by grant AI24656 from the National Institute of Allergy and Infectious Diseases, and grant RO7385 from the Uniformed Services University of the Health Sciences. We thank Ed Oaks (WRAIR) for the IpaB, IpaC and IpaD antisera, Gertrude Goping (USUHS) for technical assistance in EM analysis, Yuri Polotsky (WRAIR) for assistance in the interpretation of electron micrographs, and Reinaldo E. Fernández (USUHS) for assistance in L2 cell maintenance.