Structure and composition of the Shigella flexneri‘needle complex’, a part of its type III secreton


  • Ariel Blocker,

    Corresponding author
    1. Unité de Pathogénie Microbienne Moléculaire, INSERM U389, Institut Pasteur, 25–28 rue du Dr Roux, 75724 Paris Cedex 15, France.
    • *For correspondence. E-mail; Tel. (+33) 1 40 61 37 71; Fax (+33) 1 45 68 53 98. †Present address: Sir William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford OX1 3RE, UK. ‡The first three authors contributed equally to this work.

    Search for more papers by this author
  • Noureddine Jouihri,

    1. Université Libre de Bruxelles, Faculté de Médecine, Laboratoire de Bactériologie Moléculaire, Bâtiment G/E, CP 614, B-1070 Brussels, Belgium.
    Search for more papers by this author
  • Eric Larquet,

    1. Station Centrale de Microscopie Electronique et
    Search for more papers by this author
  • Pierre Gounon,

    1. Station Centrale de Microscopie Electronique et
    Search for more papers by this author
  • Frank Ebel,

    1. Unité de Génétique Moléculaire, CNRS 1773, Institut Pasteur, 25–28 rue du Dr Roux, 75724 Paris Cedex 15, France.
    Search for more papers by this author
  • Claude Parsot,

    1. Unité de Pathogénie Microbienne Moléculaire, INSERM U389, Institut Pasteur, 25–28 rue du Dr Roux, 75724 Paris Cedex 15, France.
    Search for more papers by this author
  • Philippe Sansonetti,

    1. Unité de Pathogénie Microbienne Moléculaire, INSERM U389, Institut Pasteur, 25–28 rue du Dr Roux, 75724 Paris Cedex 15, France.
    Search for more papers by this author
  • Abdelmounaaïm Allaoui

    1. Université Libre de Bruxelles, Faculté de Médecine, Laboratoire de Bactériologie Moléculaire, Bâtiment G/E, CP 614, B-1070 Brussels, Belgium.
    Search for more papers by this author


Type III secretion systems (TTSSs or secretons), essential virulence determinants of many Gram-negative bacteria, serve to translocate proteins directly from the bacteria into the host cytoplasm. Electron microscopy (EM) indicates that the TTSSs of Shigella flexneri are composed of: (1) an external needle; (2) a transmembrane domain; and (3) a cytoplasmic bulb. EM analysis of purified and negatively stained parts 1, 2 and a portion of 3 of the TTSS, together termed the ‘needle complex’ (NC), produced an average image at 17 Å resolution in which a base, an outer ring and a needle, inserted through the ring into the base, could be discerned. This analysis and cryoEM images of NCs indicated that the needle and base contain a central 2–3 nm canal. Five major NC components, MxiD, MxiG, MxiJ, MxiH and MxiI, were identified by N-terminal sequencing. MxiG and MxiJ are predicted to be inner membrane proteins and presumably form the base. MxiD is predicted to be an outer membrane protein and to form the outer ring. MxiH and MxiI are small hydrophilic proteins. Mutants lacking either of these proteins formed needleless secretons and were unable to secrete Ipa proteins. As MxiH was present in NCs in large molar excess, we propose that it is the major needle component. MxiI may cap at the external needle tip.


Shigella flexneri, the aetiological agent of the endemic form of bacillary dysentery, causes disease by invading the colonic mucosa (Hale, 1998). All bacterial genes necessary for entry into host cells are clustered in 30 kb of the Shigella virulence plasmid (Parsot, 1994). This region carries two types of genes: the ipa and ipg genes encoding the entry-mediating proteins (Ipa or invasins); and the mxi and spa genes encoding a type III secretion apparatus (TTSS or secreton) required to deliver invasins into the target cells.

Type III secretons are found in many other pathogenic Gram-negative bacterial species that have developed diverse strategies to survive within their hosts (Hueck, 1998). The major function of TTSSs is to transport proteins from the bacteria cytoplasm into the host cell plasma membrane or cytoplasm upon contact with host cells (Cornelis and Wolf-Watz, 1997; Galan and Collmer, 1999). In S. flexneri, the mxi and spa operons and the ipaipg operon are expressed at 37°C, but Ipa proteins remain in the bacterial cytoplasm until the secretion machinery is activated by host cell contact or external, presumably surrogate, signals such as serum or the small amphipathic dye molecule Congo red (Ménard et al., 1994; Bahrani et al., 1997). Physical contact between the bacterium and the host cell induces the insertion of two Ipas (IpaB and IpaC) into the host membrane to form a 25 Å pore that might be used to translocate the other invasins into target cells (Blocker et al., 1999). The Ipas then catalyse the formation of a localized, actin-rich, macropinocytic-like ruffle on the host cell surface, which internalizes the bacterium (Bourdet-Sicard et al., 1999; Tran Van Nhieu et al., 1999; Niebuhr et al., 2000). Bacterial internalization initiates a cycle of intra- and intercellular spreading (Niebuhr and Sansonetti, 2000).

Many components of type III secretons share sequence similarities with components of flagellar basal bodies. Kubori et al. (1998) recently identified and isolated a macromolecular structure, which they called the ‘needle complex’ (NC), formed by the Salmonella TTSS1 (SPI1). The NC resembles flagellar basal bodies and comprises two parts: (i) a 7- to 8-nm-wide, 60-nm-long external needle; and (ii) a shorter cylinder, formed by plates (20–40 nm in diameter) that presumably traverse both bacterial membranes and the peptidoglycan. Electron microscopic (EM) analysis of Shigella indicated that its type III secretons are composed of three parts: an external needle, a transmembrane neck domain and a large proximal bulb that is 44 nm wide and 27 nm high and is presumably located on the cytoplasmic side of the inner membrane. The secretons (50–100 copies cell−1) are constitutively assembled at 37°C before any host contact, and their morphology does not appear to change upon activation of secretion (Blocker et al., 1999).

Kubori et al. (1998) reported that the Salmonella NC contained only three proteins, PrgH, PrgK and InvG, which are predicted to be inner or outer bacterial membrane proteins. Very recently, Tamano et al. (2000) have published the isolation of the Shigella NC and identified four of its components, MxiG, MxiJ, MxiD and MxiH, which are the homologues of Salmonella TTSS1 proteins PrgH, PrgK, InvG and PrgI respectively. They also showed that deletion of spa47, encoding the Shigella TTSS homologue of the Salmonella flagellar ATPase FliI (which is essential for secretion; Macnab, 1996), or mxiH leads to needleless and inactive secretons. They concluded that the needle component is secreted by the TTSS itself, that the needle component is MxiH and that the needle is essential for secretion. Even more recently, Kubori et al. (2000) demonstrated that both InvC, the Salmonella TTSS1 FliI homologue, and InvA, the Shigella MxiA homologue are required for needle, but not transmembrane, region assembly (the bulb could not be distinguished in their images) and that PrgI is the major needle component of Salmonella TTSS1 and is required for effector protein secretion. Lastly, Kimbrough and Miller (2000) have shown that the prgJ (mxiI in Shigella) gene is required for assembly [in addition to those identified by Kubori et al. (2000)] of the Salmonella needle complex and produced images of NC subcomplexes composed of PrgH and PrgH + PrgK. However, none of these groups performed a detailed structural analysis of the complete NC.

The relationship of the NC to the longer, thicker Salmonella surface appendages called ‘invasomes’ (Ginocchio et al., 1994) is unclear, and the dependency of the latter on the TTSS1 in this bacterium is controversial (Reed et al., 1998). However, EspA filaments in enteropathogenic Escherichia coli (EPEC; Ebel et al., 1998; Knutton et al., 1998) and ‘pili’ composed of the HrpA or HrpY proteins in the plant pathogens Pseudomonas syringae and Ralstonia solanacearum (Roine et al., 1997; Van Gijsegem et al., 2000) are bona fide TTSS-dependent surface appendages. In these cases, the unrelated filament-forming proteins are encoded within the TTSS-encoding operons. The filaments are 10 nm thick and several microns long. EPEC require EspA filaments for attachment to eukaryotic host cells and for bacterial protein translocation into these cells (Ebel et al., 1998; Wolff et al., 1998). The relationship of these appendages to NCs is not understood.

Finally, there remains the major question of how TTSSs mediate the insertion and translocation of proteins into the host cells. There is evidence for a 2–3 nm canal within the bacterial flagellum and basal body (Francis et al., 1994; Morgan et al., 1995), through which unfolded flagellin may transit to the tip of the structure where it inserts into the growing flagellum (Emerson et al., 1970). By analogy, the type III secreton might allow protein traffic through an internal canal, with insertion of invasins into the host membrane or their translocation into the host cytoplasm across a bacterially inserted pore in the host membrane corresponding to the incorporation of flagellin at the flagellar tip.

Here, we present the isolation of the Shigella NC and its structural analysis, which allows the visualization of a central canal within its length for the first time. We also identified five NC components, including two (as Kimbrough and Miller 2000), rather than one (Kubori et al., 2000; Tamano et al., 2000), that probably originate from the needle itself.

Results and discussion

NCs show structural similarity to the bacterial flagellar basal body and are traversed by a central 2–3 nm canal

NCs were purified by a modification of protocols for the isolation of the Salmonella TTSS1 NC (Kubori et al., 1998) and flagellar basal bodies (Aizawa et al., 1985; Khan et al., 1992; Zhao et al., 1995). A negatively stained sample of purified complexes is shown in Fig. 1A. After multireference alignment of 868 particles, five stable classes were obtained (Fig. 1C). Using a hierarchical ascendant classification, the similarity relationship between the five classes, analysed on the dendrogram, allowed their merger into three final classes of 432, 341 and 95 particles. Two-dimensional averages and variances were generated for each class. The average two-dimensional projection of the major particle class, which had the lowest variance, is labelled by an asterisk in Fig. 1C and D. The resolution of the average image (Fig. 1D) is 17 Å. The NC structure is very reminiscent of that of the bacterial flagellar body. Thus, given that the two-dimensional projections were identical for all classes (despite the slightly higher variances) and because the bacterial flagellar basal body shows cylindrical symmetry (Francis et al., 1994; Thomas et al., 1999), we assumed that the NC also displays cylindrical symmetry. We saw a few bottom views of NCs with broken needles, which supported our assumption (not shown; S.-I. Aizawa, personal communication). Using this assumption and the average two-dimensional image, a three-dimensional reconstruction of the NC was generated. A surface representation of the volume is shown in Fig. 1E. Figure 1F and G represents the central axial section of the volume and the half volume of the object respectively.

Figure 1.

Structural analysis of the NC by electron microscopy.

A. Negative staining of isolated NCs; arrows point to incomplete NCs, lacking the base.

B. CryoEM of isolated NCs; three typical and enlarged NCs are shown at the top. The bar is 100 nm.

C. Results of hierarchical ascendant classification with Ward's criterion. The diagram is read from bottom to top. Each node-connecting branch represents a merged class. For instance, the classes 4 (n = 267) and 1 (n = 74) are merged into a new class with n = 341. The vertical height of the horizontal lines (in arbitrary units, a.u.) is a measure of the gain in intraclass variance obtained through the merging; if that gain is small, then the two classes are very similar. For each class, the average (above) and the variance (below) image is shown. The asterisk marks the major class for which the variance is the lowest and the particle number is the highest (n = 432).

D. Enlargement of the average image of the major particle class obtained in (C).

E. Surface representation of the NC volume obtained assuming cylindrical symmetry.

F. Central axial section of the volume obtained in (E).

G. Presentation of half of the volume obtained in (E).

H. The same as (F) but with an interpretative model representation of the three parts of the NC; a = base; b = upper ring doublet; c = needle. The parts of the full type III secreton defined by Blocker et al. (1999) are represented for reference; 1 = needle, 2 = transmembrane neck domain, 3 = bulb.

Observation of the average two-dimensional image and of successive slices of the NC volume, as well as threshold analysis of the volume (not shown), suggested the subdivision of the NC structure into three major parts, presented in the interpretative model in Fig. 1H: (a) a triangular base; (b) an upper ring doublet; and (c) the needle. Also shown on this figure are the overlapping positions of the parts we previously called 1 (needle), 2 (transmembrane neck) and 3 (cytoplasmic bulb). Given the present data, it is difficult to decide where the base and the ring doublet meet along the needle shaft. We presume that they interdigitate within the periplasm (see also below) to form a tight seal, as type III secretion occurs without a periplasmic intermediate. In Fig. 1D, the needle appears to insert down into the structure as far as the upper part of the base. This finding is supported by the presence of incomplete NCs lacking the base (arrows in Fig. 1A), which show that the needle extends beyond the upper ring doublet. The length of the extension was constant and is sufficient to reach the ‘bulge’ that connects the needle with the base (best visualized in Fig. 1F). The averaged NC image also shows that the needle contains a central 2–3 nm canal that connects to a conduit of similar diameter formed at the bottom end of the base (Fig. 1D, F and G). This canal and its diameter are directly visible on every NC observed by cryoEM (Fig. 1B). Alignment of the 231 NC particles observed by this method was impossible because of the presence of contaminating material in the preparation. Finally, there is another substructure visible at the base of the NC, a sort of plate at the entrance of the NC canal.

The ring structure is probably embedded in the outer membrane, whereas the periplasmic face of the inner membrane is probably located immediately below the proximal end of the needle. The data supporting the latter conclusion are that mutants lacking the spa47 ATPase have a relatively normal bulb and a full neck but lack the needle (Tamano et al., 2000; M.-P. Sory, P. Gounon, A. Allaoui and A. Blocker, manuscript in preparation). This suggests that the needle component(s), like the rod and hook components of the flagellum, are secreted by the TTSS and are therefore located beyond the inner membrane. The proposed position of the bulb relative to the cytoplasmic membrane and the NC in Fig. 1H is based on the fact that several putative inner membrane proteins, MxiA, Spa40, Spa33, Spa29, Spa24 and Spa9, and putative cytoplasmic proteins, at least Spa47, Spa13, Spa15, MxiK and MxiN and perhaps Spa32, MxiC and MxiL (the latter three also being secreted under certain conditions; Buchrieser et al., 2000), are encoded within the mxi/spa operons but were not recovered in NCs (see below). The former proteins have a number of obvious α-helical transmembrane regions (TMs) and would probably contribute to the NC base and/or bulb, positioning it immediately below the inner membrane.

The NC is composed of at least MxiD, MxiG, MxiJ, MxiH and MxiI

In order to identify the components of the NC, they were separated by SDS–PAGE, and all the major bands were submitted to N-terminal or internal peptide sequencing (Fig. 2). Many of the proteins identified derived from outer membrane vesicles/fragments (TolC, the OmpC porin, OmpA and SlyB, a lipoprotein) or from cytoplasmic chaperone complexes (GroEL), which were contaminants of the preparation visualized by EM (Fig. 1A and B). The other contaminants were abundant cytoplasmic enzymes [dihydrolipoamide dehydrogenase (DldH) and glycogen synthase, (GlgA)]. For comparison, only 40 out of 174 proteins identified by mass spectrometry in an affinity-purified yeast nuclear pore complex (NPC) preparation were ultimately found to be associated with the NPC (Rout et al., 2000). In our case, only five of the proteins identified were encoded within the TTSS region of the Shigella virulence plasmid (Maurelli et al., 1985; Sasakawa et al., 1988): MxiD, MxiG, MxiJ, MxiH and MxiI. Estimation of the amount of material from the first cycle of Edman degradation indicated a relative stoichiometry of 9:7:1:40:2. However, this estimate is subject to error because of incomplete NCs in the preparation (Fig. 1A), uneven transfer of proteins onto the sequencing membrane and differential efficiency of sequencing of each amino acid. The last is likely to be particularly true for the first amino acid of MxiJ, which is modified (see below).

Figure 2.

Identification of NC components. An NC sample purified as described in Experimental procedures was denatured with an equal volume of water-saturated phenol and precipitated with three volumes of cold acetone. After centrifugation, the pellet was rinsed with ethanol and resuspended in SDS–PAGE loading buffer. The sample was separated on a 20–10% acrylamide gradient gel and blotted onto a PVDF membrane. The proteins were stained with amido black. All the listed sequences were determined by N-terminal sequencing except those marked by an asterisk, which derived from internal peptides, obtained by trypsin digestion of the proteins excised from another gel and then separated by high-performance liquid chromatography (HPLC); ND, not defined. The peptide sequences, yields and identified proteins are shown on the right.

MxiD belongs to the family of outer membrane proteins called ‘secretins’ and is predicted to contain many amphipathic β-sheets. Its signal peptide is cleaved at position 23 (Fig. 2). Secretins form well-defined 10- or 12-mers that appear in a side view as ‘two stacked rings’ with a periplasmic extension (Nouwen et al., 1999; 2000). Hence, MxiD probably forms the upper ring doublet and part of the periplasmic needle shaft. We used the known stoichiometry of secretins, our estimated number of secretons in each bacterium (50–100), and the amount of OmpC in our preparation (105 copies cell−1; Nikaido, 1996) to estimate the enrichment of our preparation at about 1000-fold. The secretin stoichiometry also leads us to propose that the number of each Mxi molecule per NC may be about 1–1.5 times the relative stoichiometry determined above, except for MxiJ, which is likely to have been severely underestimated (see above). Indeed FliF, a protein within flagellar basal bodies with some homology to MxiJ, was determined to be present in 26 copies per structure (Jones et al., 1990).

MxiJ and MxiG are predicted to span the inner membrane once (Allaoui et al., 1995). MxiG is predicted to have a hydrophobic TM called an ‘internal signal sequence’ region within the first third of its sequence and thus a large periplasmic C-terminal domain and a smaller cytoplasmic N-terminal domain. MxiJ has a predicted hydrophobic α-helical ‘stop transfer sequence’ close to its C-terminus and a signal peptide that is cleaved in front of the Cys at position 18 at a consensus lipoprotein signal peptidase cleavage site (Fig. 2). However, as we and others have been able to sequence this protein or its homologue (Kubori et al., 1998; Tamano et al., 2000), its N-terminus is not blocked by the N-acyl group normally present in lipoproteins from Gram-negative bacteria. Therefore, these proteins might carry a single diacylglyceride attached to the SH group of the N-terminal Cys. As MxiJ has a Glu at position +2, it should be located in the inner membrane with its lipids in the periplasmic leaflet of the outer membrane (Yamaguchi et al., 1988; Seydel et al., 1999). However, when we aligned the 21 homologues of MxiJ in protein databases, we found that three MxiJ homologues (YsaJ, SctJ and SsaJ) carry an Asp at position +2, the canonical signal for inner membrane localization of the lipids of lipoproteins (Seydel et al., 1999). Furthermore, five MxiJ homologues (PrgK, YscJ, PscJ, NolT and EscJ) have a Lys at position +2, whereas only two out of 95 putative E. coli lipoproteins have a Lys at this position. In a model lipoprotein, the presence of a lysine at position +2 impedes processing by lipoprotein signal peptidase (Seydel et al., 1999). Thus, the exact nature of the lipid modification on MxiJ and its membrane localization need to be determined experimentally, but it is worth considering that MxiJ might belong to a new class of lipoproteins. However, given the images of Kimbrough and Miller (2000, Figs 4 and 5) showing quite flat complexes of PrgH (MxiG in Shigella) and PrgH + PrgK (MxiJ in Shigella) expressed in E. coli, it is likely that MxiJ's lipid moiety is inserted into the inner membrane and that MxiD's periplasmic region reaches down (as seen by Nouwen et al., 1999) to bind to the periplasmic domains of MxiG and MxiJ. In summary, we propose that MxiG and MxiJ form most of the triangular base of the NC, with MxiG contributing to the cytoplasmic and periplasmic parts and MxiJ mostly to the periplasmic part. Accordingly, none of the three proteins discussed above can be components of the external needle.

Figure 4.

Analysis of Ipa protein secretion in the mxiH and mxiI mutants. Cultures of S. flexneri wild-type strain M90T (WT), the mxiH and mxiI mutants, complemented strains mxiH + pJN34 (a plasmid containing the entire mxiH gene) and mxiI + pJN16 (a plasmid harbouring the mxiI gene) were harvested by centrifugation, resuspended in PBS and incubated in the presence of Congo red for 30 min at 37°C to induce secretion. Samples were centrifuged, and identical amounts of bacterial lysates and supernatants were analysed by SDS–PAGE and immunoblotting using monoclonal antibodies specific for IpaB and IpaC.

Figure 5.

EM analysis of TTSS morphology in mxiH and mxiI mutants. Bacteria were prepared for whole-mount negative stain as explained in Experimental procedures.

A. mxiH mutant.

B. Complemented mxiH mutant.

C. mxiI mutant; arrowhead shows two other incomplete secretons.

D. Complemented mxiI mutant.

Arrows in (A) and (C) indicate particularly informative views of incomplete secretons in mxiH and mxiI mutants, where the connection of the complex to the outer membrane is best visualized. Bar equals 100 nm.

There are a number of components of the Shigella NC that, given what is known of the composition of the flagellar basal body and its similarity to the NC, we could have expected but did not find (S.-I. Aizawa, personal communication). These include components of the ‘bottom plate’ and of the periplasmic part of the needle. These substructures are probably equivalent to the C-rod and periplasmic rod, respectively, in the flagellar basal body. Our putative C-rod components are Spa33, Spa24, Spa9 and Spa29 (the homologues of FliO, FliP, FliQ and FliR, respectively, in the flagellum' Ohnishi et al., 1997), whereas our periplasmic rod equivalent(s) cannot easily be identified by sequence homology with periplasmic rod components of the flagellum (Homma et al., 1990). The flagellar C- and periplasmic rod components are all present in low copy number within the basal body (two to six copies), and we are likely to have missed them in our NC preparations, because they are also small proteins and thus are probably poorly visible upon amido black staining. We found MxiI in our preparations only because it co-migrated with the large MxiH band.

MxiH is conserved in TTSSs from a wide variety of species, whereas MxiI is only found in Shigella and Salmonella TTSS1 (SPI1)

MxiH and MxiI are predicted to be hydrophilic, globular proteins and could thus be needle components. Indeed, both proteins lack a signal sequence and could, in principle, be substrates for TTSSs, like the rod and hook components in flagellar biogenesis (Macnab, 1996). A sequence database search revealed that genes encoding MxiH homologues are present in Salmonella typhimurium SPI1 and SPI2, the Pseudomonas aeruginosa TTSS region, the EPEC LEA region and on the Yersinia virulence plasmid (Fig. 3), but not in the TTSS from Bordetella bronchiseptica, Chlamydia pneumoniae, the plant pathogens or the flagellar basal body-encoding genes. MxiH is most similar to the Salmonella SPI1 PrgI protein (68% identity). The only close MxiI homologue detected is the Salmonella SPI1 PrgJ protein (37% identity; Fig. 3). Salmonella SPI1 is the TTSS that is most similar in sequence to the Shigella TTSS (Hueck, 1998).

Figure 3.

Sequence analysis of MxiH and MxiI and construction of the non-polar mutants in mxiI and mxiH.

A. The sequence searches were performed with blast2 and the sequence alignment with clustalw on the EMBL server using standard parameters.

B. Construction of the mxiH and mxiI mutants. The plasmids are shown in linear form, and part of the S. flexneri virulence pWR100 plasmid DNA encompassing the mxiH and mxiI genes is shown at the top. The aphA-3 cassette is indicated by a filled bold arrow. Small arrows indicate the extent and orientations of several genes. Only positions of relevant restrictions sites are shown: B, BglII; Bs, BstEI; E, EcoRI; Ev, EcoRV; M, MscI; S, SmaI; Sf, SfuI.

The MxiH and MxiI protein families are distantly related. Indeed, multiple sequence alignment showed that the proteins are likely to be composed of two equal-sized domains separated by a short variable hinge flanked by two Pro or a Pro and other turn-inducing amino acids. The N-terminal domain is specific to each protein family, whereas the C-terminal domain is conserved between them (Fig. 3A). When the residues shown in bold were used to derive a consensus sequence that was then used to search protein databases, no other sequence except those of the MxiH and MxiI families was found (not shown). Therefore, this consensus is specific to these families of proteins and, in particular, it is not found in either the HrpA or the HrpY pilus proteins from the plant pathogens, in the EPEC EspA or in the flagellar rod or hook components. Tamano et al. (2000) identified two putative short coiled-coil regions in MxiH using the coils software. We did not obtain highly significant coiled-coil scores for either MxiH or MxiI using this or any other program. However, most of these programs are designed to detect relatively long, solvent-exposed, left-handed two-helix coiled-coils and perform poorly on the kind of α-helices found in flagellar components, i.e. the buried multiple-helix bundles that allow polymerization and stability of the flagellum (Macnab, 1996). MxiH and MxiI do each contain two 10- to 15-amino-acid-long regions predicted to form charged α-helices, one in the N-terminal domain and one in the C-terminal domain (not shown).

mxiH and mxiI mutants produce needleless secretons

We reported previously that non-polar mutants in mxiG, mxiJ and mxiD displayed no visible TTSS structures at their surface (Blocker et al., 1999). This observation is consistent with the present identification of MxiG, MxiJ and MxiD as intrinsic membrane-bound components of the NC. In order to understand the role of MxiH and MxiI in NC assembly and function, we generated non-polar mutants in both genes. Both mutants were unable to invade HeLa cells (not shown) and to secrete Ipa proteins upon the addition of Congo red (Fig. 4) or during overnight culture (not shown). The latter result points to a very basic defect in the secreton machinery itself. The mxiH and mxiI mutants complemented in trans with the respective gene were fully restored in their ability to invade HeLa cells (not shown) and to secrete Ipa proteins (Fig. 4 and not shown).

We examined the secretons produced in the mxiH and mxiI mutants by EM. In both mutants, the type III secretons displayed only the bulb and transmembrane regions (Fig. 5A and C). Figure 5A and C shows views of how these needleless secretons connect to the outer membrane, and the empty transmembrane needle shaft, initially seen in Fig. 1, can be discerned (arrows). Secretons appeared normal in the complemented strains (Fig. 5B and D). The absence of a needle in these mutants suggested that MxiH and MxiI are required for needle assembly and/or are needle component(s). MxiH is the major NC component (see above), being approximately 20-fold more abundant than MxiI in NC preparations. We therefore propose that MxiH is the major component of the NC needle, the equivalent of the hook protein FlgE in bacterial flagella.

What might be the function of MxiI? Given what is known of bacterial flagellar basal bodies (S.-I. Aizawa, personal communication), MxiI is unlikely to be the homogeneously sized periplasmic rod (Fig. 1) component(s) for two reasons. First, the flagellar rod proteins, FlgB, FlgC, FlgF and FlgG, which bridge the periplasm between the inner and outer membrane rings and the outside hook component, are each present in six copies (approximately one rod ring each; Jones et al., 1990) and are quite homologous to each other at the sequence level (Homma et al., 1990). Indeed, they share α-helical regions that are thought to allow them to co-polymerize with each other and with the hook component in an ordered manner. Thus, the periplasmic rod proteins are quite different from the hook (FlgE) and hook cap (FlgD) proteins at the sequence level. Secondly, because the MxiI–MxiH ratio is 1:20 in the NC (Fig. 2), MxiI alone cannot account for the length of the periplasmic rod in the NC, which is nearly one-third of the length of the total needle (Fig. 1). Thus, we propose that MxiI is the functional equivalent of the flagellar hook cap, FlgD (Ohnishi et al., 1994).

FlgD shows sequence homology to FlgE and has been termed the ‘hook scaffolding protein’ because it is initially positioned at the rod tip and allows hook protein insertion between it and the terminal rod protein. FlgD is required for hook assembly, but not for hook protein secretion, as in the flgD mutant, FlgE is found in the medium (Ohnishi et al., 1994). In addition, it is not present in the complete flagellum and is probably lost from the hook tip immediately before or after FlgK (HAP1) becomes secreted and associated with the hook tip (S.-I. Aizawa, personal communication). We noted that MxiI and MxiH share significant sequence homology (Fig. 3), as do FlgD and FlgE. Thus, we searched for secreted MxiH in the overnight growth medium of the mxiI mutant, but found none. MxiH might require MxiI for its stability in Shigella cytoplasm or for its secretion, or perhaps MxiH polymerizes in the growth medium in the absence of MxiI and is hence lost from the growth medium supernatant before trichloroacetic acid (TCA) precipitation. However, there is good reason to believe that needle length control operates similarly to hook length control in bacterial flagella. For instance, when the regulator of flagellar hook length, FliK, is absent, flagella have ‘polyhooks’ and, recently, Kubori et al. (2000) have shown that deletion of invJ, the homologue of fliK in the Salmonella TTSS1-encoding region and of Shigella spa32 (J. Magdalena, P. Gounon, A. Blocker and A. Allaoui, in preparation), leads to highly elongated needles.

Implications for type III secreton function

An important conclusion of our findings [and those of Tamano et al. (2000) and Kubori et al. (2000)] is that the external needle is required for the Shigella TTSS and the Salmonella TTSS1 to secrete effector proteins into the medium. This implies that either there is a feedback mechanism to sense that the machine is incomplete and to shut it off or that the incomplete machine cannot be activated to secrete. The needle might therefore be required to transmit the activating signal to the inner membrane.

MxiH is conserved in most of the bacterial species that have TTSSs and are pathogenic in animals. Therefore, it is surprising that a mxiH homologue was not reported in B. bronchiseptica and Chlamydia. The Bordetella TTSS region is still poorly characterized but, in Chlamydia, the probable TTSS has a similarly sized needle to that produced by Shigella (Miyashita et al., 1993). However, the Chlamydia TTSS machinery is encoded by genes in at least four different chromosomal regions (Subtil et al., 2000), and open reading frames (ORFs) coding for proteins below 10 kDa, such as MxiH, are often ignored during genome annotation. Yet, the ‘needle-like structure’ is clearly variable in sequence, as it is not found in plant pathogens in which the TTSS-encoding regions are grouped and are well characterized (Hueck, 1998). Another mystery is why a MxiI homologue is found only in Salmonella SPI1. Perhaps MxiI is part of the ‘sensor’ that transmits the activation signal to the inner membrane apparatus and is therefore variable because it senses different signals?

The results and conclusions on MxiH presented here agree with those of Tamano et al. (2000). However, we see no reason to invoke needle elongation during Shigella contact with the smooth apical domain of M-cells or the basolateral membrane of enterocytes in vivo (Niebuhr and Sansonetti, 2000), as they did upon finding that MxiH overexpression leads to µm-long needles (Tamano et al., 2000). Indeed, we have shown that the morphology of the TTSSs does not change grossly upon activation of secretion by Congo red (Blocker et al., 1999). Furthermore, freeze-fracture, deep-etched views of bacteria and red blood cells after contact haemolysis was triggered revealed secreton needles of the same length as those seen on bacterial ghosts fixed before host contact (P. Gounon and A. Blocker, unpublished). Finally, in EPEC, the length of the additional EspA appendage, but not that of the needle, varies. For this organism, haemolysis does not require the intimate contact we reported for Shigella but will occur without centrifugation (Shaw et al., in press).

Is there a needle in those TTSSs that have an MxiH homologue and a pilus-like structure, such as the EspA filament in EPEC? (Shaw et al., in press) have shown EspA labelling along the filaments up to about 50–100 nm from the outer membrane, suggesting either that the antibody cannot access this region because of the lipopolysaccharide (LPS) (which is probably only 10–15 nm thick; B. Finlay, personal communication) or that another protein structure exists close to the membrane, perhaps the MxiH homologue EscF? It is possible that some activated TTSSs that need to traverse long distances before reaching their host cell membrane (for instance, for EPEC, which attaches to the apical face of enterocytes, the mucus, glycocalyx and microvilli) polymerize a superstructure on top of the needle upon encountering signals indicating host cell vicinity and only then secrete effectors once tight contact with the host cell has been established.

Finally, we would like to emphasize that the NC does not contain the Spa47 ATPase that we proposed was positioned within the cytoplasmic bulb of the TTSS (Blocker et al., 1999). A cytoplasmic structure, the C-ring, is known to be associated with the bacterial flagellum (the ‘blob’ in Fig. 5 of Katayama et al., 1996) and is thought to contain at least FliI and associated components. Although the secreton phenotype of the spa47 mutant (relatively normal bulb and neck but no needle; Tamano et al., 2000; M.-P. Sory, P. Gounon, A. Allaoui and A. Blocker, manuscript in preparation) suggests that the ATPase might not be a large structural part of the bulb, our current data suggest that the needle complex is only a hollow protein conduit. The way in which the presumed translocation motor, Spa47, attaches to the bulb and NC and how the complete machine works now needs to be addressed.

Experimental procedures

Bacterial strains and growth media

S. flexneri strains were derivatives of the wild-type strain M90T (serotype 5; Sansonetti et al., 1982). Strain M90T-Sm (SmR) has been described previously (Allaoui et al., 1993). Bacteria were grown in Luria–Bertani (LB) medium or tryptic casein soy broth (TCSB) at 37°C. The antibiotics used were: ampicillin (100 µg ml−1), kanamycin (50 µg ml−1) and streptomycin (100 µg ml−1).

Purification of the needle complex

TCSB (1.6 l) was seeded with Shigella diluted 1:25 from overnight precultures at 37°C. Bacteria were shaken in flasks for 2.5 h at 37°C until OD600 of 1, then harvested at low speed and washed in PBS. Bacteria were spheroplasted in 80 ml of 0.5 M sucrose, 50 mM Tris, pH 8, 8 mM EDTA and 2 mg ml−1 chicken lysozyme (Sigma) at 37°C for ≈ 1 h. Complete spheroplasting was essential to obtain a high yield of NCs, which otherwise remain trapped in the peptidoglycan. Spheroplasting was visualized by phase-contrast microscopy. Spheroplasts were round and disintegrated entirely (leaving no ‘ghosts’) upon osmotic shock. The spheroplasts were lysed in 1% Triton X-100 (the solution became clear and viscous) in the presence of protease inhibitors (Complete EDTA free; Boehringer Mannheim). MgSO4 was added to 35 mM along with 100 µg ml−1 DNase, and the sample was left at 37°C until it was no longer viscous (a white precipitate sometimes developed). The solution was centrifuged at 20 000 g for 20 min at 4°C to remove unlysed bacteria and precipitated material. The supernatant was then centrifuged at 70 000 g for 1 h at 4°C. The pellets were resuspended in 5 ml of 150 mM KCl, 5 mM EDTA, 1% Triton X-100 in 10 mM Tris (pH 8) and passed through 21-, 25- and then 27-gauge syringe needles. The mixture was then centrifuged at 20 000 g for 20 min, and its supernatant was subsequently centrifuged at 70 000 g for 1 h at 4°C. The pellet was resuspended in 5 ml of 50 mM Tris, pH 8, containing 5 mM EDTA and 0.1% Triton X-100 (TET buffer), filtered through a 0.2 µm filter and centrifuged again at 70 000 g for 1 h. The pellet was resuspended in 500 µl of TET buffer and passed over a 2.5 ml bed volume of Sephacryl S-1000 Superfine (Amersham Pharmacia Biotech) at 1 ml 6 min−1 in a 2.5 ml syringe plugged with glass wool and fitted with a 26-gauge needle. Fractions (500 µl) were collected, and the NCs were concentrated by centrifugation at 70 000 g for 1 h, followed by resuspension in 20–40 µl of TET. NC enrichment was estimated by EM after negative staining. Because spheroplasting was not always reproducible, strain M90T was transformed with pLysS (Novagen) carrying the bacteriophage T7 lysozyme gene. The bacteria were grown as above and frozen at −80°C after the PBS wash. Upon thawing into warm spheroplasting buffer, the bacteria lysed in the absence of added lysozyme.

Construction of the mxiH and mxiI mutants

The non-polar mutants were engineered by insertion of the aphA-3 cassette (Ménard et al., 1993) into virulence plasmid pWR100 by allelic exchange. Plasmid pAB10i was constructed by cloning the 5484 Bgl II fragment of pHS5103 (Baudry et al., 1987) into the Bam HI site of pUC19 (Fig. 3B). Plasmid pANJ1 was obtained by digesting pAB10i with Bst EII and SmaI, filling in the ends and religating. pANJ2 was constructed by inserting the 850 bp SmaI DNA fragment (aphA-3 cassette) of pUC18 K into the unique Sfu I site of pANJ2 located within the mxiH gene. In this construct, the aphA-3 cassette introduced a translational stop at codon 22 of mxiH and a translational start codon in frame with the last 41 codons of the gene. Mutator plasmid pANJ3 was then constructed by cloning the 4040 bp KpnI–Sal I fragment into the corresponding sites of the pGP704 suicide vector. To generate mxiI mutant, plasmid pANJ4 was constructed by cloning the 1313 bp EcoRI–BspEI fragment of pAB10i into the EcoRI–XmaI site of pTZ18R. Plasmid pANJ5 was constructed by replacing the 122 bp EcoRV–MscI segment DNA (internal to mxiI) with the 850 bp SmaI DNA fragment (aphA-3 cassette). In this construct, the mxiI gene was interrupted at codon 22, codons 23–64 were deleted and the last 33 codons were inserted in frame with the translational start codon located at the 3′ of the aphA-3 cassette. pANJ6 was obtained by inserting the 1995 bp XbaI–EcoRI DNA fragment into the corresponding sites of pGP704. To inactivate the mxiH and mxiI genes on pWR100 by allelic exchange, pANJ3 (mxiH) and pANJ6 (mxiI) were transferred to S. flexneri M90T-Sm by conjugation, with selection for the transconjugants on plates containing streptomycin and kanamycin. Clones in which a double recombinational event had replaced the wild-type gene by the mutated copy from the recombinant plasmids were identified by screening for ampicillin resistance. The structure of the resulting pWR100 plasmids carrying the mxiH and mxiI single mutations was confirmed by polymerase chain reaction (PCR) analysis and by complementation experiments. The resulting strains were designated SB116 (mxiH) and SB125 (mxiI).

Preparation of secreted Ipa proteins, SDS–PAGE and immunoblotting

Overnight cultures grown at 37°C were diluted to an OD600 of 0.02 in 10 ml of TSB and grown at 37°C to an OD600 of 2. The culture supernatant was obtained by centrifugation (10 000 g for 10 min) and passage through a 0.45 µm filter, and soluble proteins were precipitated with 10% TCA. Bacterial pellets were washed with PBS, resuspended in 500 µl of PBS containing 100 µg of Congo red (Serva) and incubated at 37°C for 15 min to induce Ipa protein secretion. The bacteria were then harvested by centrifugation (20 000 g for 20 min) at 4°C and resuspended in 500 µl of PBS. Each sample (15 µl) was loaded on a 12% SDS–PAGE gel, the proteins were transferred to polyvinylidene difluoride (PVDF) membranes and immunoblotting was carried out with mouse monoclonal antibodies directed against IpaB and IpaC (a gift from A. Phalipon). Horseradish peroxidase-labelled goat anti-mouse antibodies (Sigma) were used as secondary antibodies, and the detected proteins were visualized by enhanced chemiluminescence (Amersham Pharmacia Biotech).

Electron microscopy

Whole-mount negative stain of lysed bacteria for transmission electron microscopy (TEM).  Observation of whole-mounted negatively stained bacteria was performed as described by Blocker et al. (1999) except that the 20 µl of bacteria in PBS was vortexed 4 × 10 s at maximum speed with a small amount (insufficient to cover the liquid) of glass beads (400–600 µm in diameter, acid washed; Sigma) to increase the number of properly lysed bacteria for observation. This treatment left the bacterial ghosts mainly intact, while removing much of the cytoplasm through the small holes formed in their membranes.

Negative staining of isolated NCs for TEM.  Purified NCs were diluted 1:5 in 20 mM Tris, pH 7.4. An aliquot of 5 µl was deposited for 1 min on glow-discharged carbon-coated copper grids. After removal of the excess liquid with filter paper, the sample was stained with a drop of 1% uranyl acetate (pH 7.5).

Vitrification of isolated NCs for cryoEM.  A 300-mesh copper grid coated with a holey carbon film was inserted into tweezers mounted on a plunger (Dubochet et al., 1982; Adrian et al., 1984). Purified NCs (4 µl) were applied to the grid. After blotting the excess of solution (≈1–3 s) with filter paper, the plunger immediately immersed the specimen into liquid ethane cooled close to its freezing point using liquid nitrogen. The vitrified specimens were transferred, without reheating, into a Gatan 626 cryospecimen holder via a cryotransfer system and introduced into the electron microscope.

Observation and image acquisition of single particles.  The particles were observed in low-dose electron beam conditions on a Philips CM12 TEM at 100 kV. Micrographs were recorded at 45 000 × magnification on Kodak SO163 films. Defocus values of −600 nm and −1.2 µm were used for negative staining and cryoEM respectively.

Digitization and image processing.  Micrographs were digitized on a Hi-Scan rotary drum microdensitometer using a 10 µm2 scanning aperture, equivalent to a final 5 Å pixel−1 image resolution on the computer screen. After contrast normalization, 868 particles were selected for two-dimensional image analysis. Processing of the particle images was performed using spider software and web interactive selection program. To generate an average projection image of the needle complex, the whole set of selected images was subjected to a multireference alignment process called ‘alignment through classification’ (Boekema et al., 1986). For each alignment step, the subsequent classification revealed several projection classes. These classes were used as reference images in the next multireference alignment step, classification using Ward's merging criterion was repeated and new classes formed. The procedure was reiterated until stable classes were obtained. A classification tree or dendrogram was constructed from the hierarchical ascendant classification. The resolution limit was calculated as the average between two values obtained independently using differential phase-residual (Frank et al., 1981) and Fourier ring correlation (Saxton and Baumeister, 1982) criteria. For three-dimensional reconstruction, cylindrical symmetry was assumed, and the final average (two-dimensional projection) was converted to a three-dimensional map using an iterative back projection procedure (Frank, 1996).


We dedicate this paper to Chi Aizawa, in memory of a marvellous week in Paris (25–28 September 2000) when new friendships started. We thank Jacques D'Alayer and Marilyne Davi at the Institute Pasteur protein sequencing facility for identifying the NC components, especially MxiI. F.E. and A.B. thank Ülf Nehrbass for long-term use of his lovely TC facility during monoclonal antibody generation and lots of tight espressos. E.L. and A.B. are grateful to Claude Rolin for expert EM negative development and NC particle selection for image analysis. E.L. thanks Nicolas Boisset for advice on the optimization of the three-dimensional reconstruction of the NC. A.B. is especially grateful to Tony Pugsley for lessons on secretins and lipoprotein processing (and for ‘putting up with her’). T. Pugsley and Dana Philipott are thanked for critical readings of the manuscript. A.B. thanks the Pugsley laboratory as a whole for sharing secreton fondness, Geneviève Milon for the kindest moral support, and Ira Mellman for magical mentoring via E-mail and other such personal treasures throughout 1999–2000. The latter are both thanked also for paving A.B.'s way to Oxford. A.B. was supported by a postdoctoral fellowship from the NIAID and a Roux fellowship from the Institute Pasteur. A.A. and N.J. were supported in part by grants from the EEC (also shared by P.S.), the Belgian Fonds National de la Recherche Scientifique Médicale and the Direction Générale de la Recherche Scientifique de la Communauté Française de Belgique. F.E. was supported by a research fellowship from the Deutsche Forschungsgemeinschaft.


  1. Present address: Sit William Dunn School of Pathology, University of Oxford, South Parks Road, Oxford OX1 3RE, UK.

  2. The first three authors contributed equally to this work.