Characterization of a Type III secretion substrate specificity switch (T3S4) domain in YscP from Yersinia enterocolitica



The length of the needle ending the Yersinia Ysc injectisome is determined by YscP, a protein acting as a molecular ruler. In addition, YscP is required for Yop secretion. In the present paper, by a systematic deletion analysis, we localized accurately the region required for Yop secretion between residues 405 and 500. As this C-terminal region of YscP has also been shown to control needle length it probably represents the substrate specificity switch of the machinery. By a bioinformatics analysis, we show that this region has a globular structure, an original α/β fold, a P-x-L-G signature and presumably no catalytic activity. In spite of very limited sequence similarities, this structure is conserved among the proteins that are presumed to control the needle length in many different injectisomes and also among members of the FliK family, which control the flagellar hook length. This region thus represents a new protein domain that we called T3S4 for Type III secretion substrate specificity switch. The T3S4 domain of YscP can be replaced by the T3S4 domain of AscP (Aeromonas salmonicida) or PscP (Pseudomonas aeruginosa) but not by the one from FliK, indicating that in spite of a common global structure, these domains need to fit their partner proteins in the secretion apparatus.


Injectisomes are multicomponent nanomachines involved in molecular trans-kingdom communication between bacteria and eukaryotic cells. They are encountered in Gram-negative bacteria that are either pathogenic for animals and plants or are symbionts (Viprey et al., 1998; Dale et al., 2001; Dale et al., 2002). They allow extracellular bacteria to inject proteins across the plasma membrane of eukaryotic cells and they enable intracellular bacteria to inject proteins across the host membrane limiting their compartment. For reviews see References (Cornelis and Wolf-Watz, 1997; Anderson and Schneewind, 1999; Galan and Collmer, 1999; Cornelis and Van Gijsegem, 2000; Plano et al., 2001; Cornelis, 2002).

Based on a phylogeny analysis made on three highly conserved proteins, the injectisomes can be grouped into five major families (Foultier et al., 2002). The injectisomes from animal pathogens cluster in three families. The plasmid encoded Ysc injectisome common to Yersinia enterocolitica, Yersinia pestis and Yersinia pseudotuberculosis represents an archetype for the largest family, which also includes the injectisomes from Pseudomonas aeruginosa (Psc), Aeromonas salmonicida (Asc), Bordetella pertussis, parapertussis and bronchiseptica (Bsc), Photorhabdus luminescens (Lsc and Sct) and an unnamed injectisome from Vibrio parahemolyticus (called here Vsc). A second family includes the Shigella spp. Mxi-Spa injectisome, the Salmonella spp. Inv injectisome (encoded by SPI-1 in S. enterica), the Ysa injectisome from Y. enterocolitica, an injectisome from Burkholderia pseudomallei and one from Sodalis glossinidius. A third family includes the Ssa injectisome from S. enterica (encoded by SPI-2), the Esc injectisome from enteropathogenic (EPECs) and enterohaemorrhagic (EHECs) E. coli and a chromosome-encoded injectisome from Y. pestis. The injectisomes from plant pathogens cluster in two families called Hrp1 (Erwinia amylovora and Pseudomonas syringae) and Hrp2 (Xanthomonas campestris, Ralstonia solanacearum).

The amino acid sequence of many injectisome constituents has significant similarity to proteins from the flagellum with whom they share Type III secretion (T3S) (Allaoui et al., 1994; Fields et al., 1994; Van Gijsegem et al., 1995; Aizawa, 2001; Blocker et al., 2003). Both structures have a similar basal body consisting of a pair of rings that span the inner and outer bacterial membranes, hold together by a short tube. In most animal pathogens, the injectisome ends up with a needle that protrudes outside the bacterium. Therefore, the injectisome is also called needle-complex (Kubori et al., 1998; Blocker et al., 1999; Kimbrough and Miller, 2000; Kubori et al., 2000; Sukhan et al., 2003; Marlovits et al., 2004). A long flexible pilus terminates the Hrp and Esc injectisomes (Roine et al., 1997; Knutton et al., 1998; Van Gijsegem et al., 2000). In the flagellum, the filament is connected to the basal body via a hook. The length of this hook is genetically determined and a protein called FliK has been shown to play the key role in this process (Patterson-Delafield et al., 1973; Suzuki and Iino, 1981). Mutants deficient in FliK make extra-long hooks (called polyhooks) and no filament. Extragenic suppressive mutations restoring filament assembly on polyhook structures (polyhook-filament phenotype) have been mapped in flhB, a gene encoding a major component of the export apparatus (Suzuki and Iino, 1981; Kutsukake et al., 1994; Williams et al., 1996; Fraser et al., 2003a). The study of these fliK suppressors mutations led to the proposal that the C-terminal domain of FlhB has two substrate specificity states and that a conformational change, mediated by FliK and accompanied by a proteolytic cleavage of FlhB is responsible for the specificity-switching process allowing secretion of flagellin once assembly of the hook is completed (Minamino and Macnab, 2000; Makishima et al., 2001; Fraser et al., 2003b). The length of the injectisome-needle is also genetically determined. Mutations in spa32 from Shigella, in invJ from Salmonella (SPI-1) (Kubori et al., 2000; Magdalena et al., 2002; Tamano et al., 2002) and in yscP from Yersinia (Journet et al., 2003) lead to needles with undefinite length and no substrate secretion. The homologues of FlhB in injectisomes form the well-conserved YscU family (Allaoui et al., 1994). Introduction into YscU of the FlhB substitutions that suppress the fliK phenotype restores Yop secretion in yscP mutants (Edqvist et al., 2003). Moreover, like FlhB, YscU undergoes a proteolytic cleavage (Lavander et al., 2003). Thus, very much like FliK, YscP is presumably involved in the substrate specificity switch, interacting with the basal-body component YscU to stop secretion of the YscF needle subunits and to start Yop secretion. In addition to this switch function, YscP was recently shown to act as a molecular ruler determining the length of the needle (Journet et al., 2003). YscP is thus a protein with a dual function. In the molecular ruler model, the N-terminus and C-terminus of YscP are proposed to act as anchors of the central ruler part, one extremity being attached to the basal body and the other one to the growing tip of the needle (Journet et al., 2003). However, no information exists regarding the regions of YscP involved in its switch role.

Here, by systematic deletion mutagenesis of YscP and a bioinformatics analysis, we localized accurately the switch domain in the C-terminal domain of the protein and we show that this switch domain is globular and conserved among the proteins controlling length in injectisomes as well as in the flagellum. In contrast to most of the proteins of the injectisomes and flagella, the relationship between Spa32/InvJ, YscP and FliK had remained to date unrevealed because of the too high divergence in their sequences (in the range of 10% identity).


YscP385–500 is required for Yop secretion

We showed previously that in frame deletions and insertions in the central part of YscP (36–96 and 222–306) defined a ruler region (Journet et al., 2003), and that in frame deletions in the N-terminus (16–25 and 26–35) as well as in the C-terminus (385–424 and 467–515) identified regions of the protein that are necessary for YscP to exert its needle length control function (Journet et al., 2003). However, the deletion mutants were not analysed for their capacity to secrete Yops. In this work, we generated new deletions in the yscP gene that together with the ones generated before (see above) encompass almost the entire gene (Fig. 1A) and we tested the mutants for their capacity to secrete Yops. The ruler region of YscP is characterized by a duplication of 60 central residues (Stainier et al., 2000), which hampers the engineering of deletions. For this reason, we could only construct one large deletion between residues 222 and 306 of YscP, and not smaller ones as in the rest of the protein. We cloned the mutated alleles downstream from an arabinose-inducible promoter (pBAD) and the recombinant plasmids were introduced in Y. enterocolitica E40 (pLJ4036), a new yscP null mutant constructed for this study. In the pLJ4036 pYV plasmid, the yscP gene is completely removed, from start to stop, in order to prevent any interference from left-over domains or sites. All the recombinant bacteria, incubated at 37°C in Ca2+-deprived medium containing arabinose, synthesized YscP proteins of the expected size (Fig. 1B). We then analysed the phenotype of all the mutants with regard to Yop secretion. As shown in Fig. 1C, the deletions removing parts of the protein before residue 381 and after residue 500 had no impact on Yop secretion. In contrast, deletions between amino acids 385–500 affected Yop secretion. This indicates that this C-terminal region (385–500) is required for both functions, length control (Journet et al., 2003) and Yop secretion, and hence is responsible for the substrate specificity switch function. Interestingly, there was no strict correlation between the phenotype of Yop secretion deficiency and tight length control. Indeed, deletions affecting the N-terminus (YscPΔ16–25 and YscPΔ26–35) (Journet et al., 2003) led only to a loss of length control but not to a loss of Yop secretion.

Figure 1.

A. Schematic representations of the yscP in frame deletions mutants together with their ability to secrete Yops. The highlighted box in YscP (aa 403–492) represents the T3S4 domain.
B. Western blot analysis (total cells; polyclonal antibody) of the various YscP proteins produced by Y. enterocolitica MRS40(pLJ4036), i.e the yscP mutant (–, control); and by Y. enterocolitica MRS40(pLJ4036) carrying the plasmids (listed on the left) encoding the various yscP + genes.
C. Yop proteins secreted by the same Y. enterocolitica E40 strains as in B (Coomassie-stained SDS-PAGE). Note that MRS40(pLJ4036), the yscP knockout background is not completely negative for Yop secretion.

The YscP C-terminal region is conserved and defines a new family of proteins

YscP and its flagellar equivalent, FliK, have similar functions but their global identity at the primary sequence level is only within the 10% range. We identified the switch function in the C-terminus of YscP. In the model proposed for FliK, the domain that switches the substrate specificity is also localized in the C-terminus (Williams et al., 1996; Minamino et al., 2004). We therefore then wondered if the C-terminal regions of YscP and FliK were sharing similarities and if we could point out an equivalent region in proteins found in other injectisomes or T3S systems.

To begin with, we used the C-terminal domain of YscP (from aa 385 to the C-terminus) as query and performed PSI-blast searches (Altschul et al., 1997) on the non-redundant database (nr) at the National Center for Biological Information (NCBI) using an E-value inclusion threshold of 0.005. By iteration 4, searches converged to identify counterpart proteins of YscP in T3S systems from the YscP family: AscP, PscP, LscP and SctP. As suggested by the letter code, these proteins are encoded by genes that occupy the same locus as yscP in the operons encoding the different injectisomes. In addition, we found the VscP protein from Vibrio harveyi[GenBank identifier (gi) 41834182]. All these proteins turned out to have a conserved C-terminal domain (34% identity) preceded by variable regions. We called the conserved domain the T3S4 domain after Type III secretion substrate specificity switch.

We also found a suggestive match (E-value 0.19, 35% identity on a 117 aa overlap) with the flagellar hook-length control protein FliK from Y. pestis (gi 45442809). Integrating the Y. pestis FliK sequence within the position specific score matrix (PSSM) led to significantly detect all the members of the FliK family. One can note that the similarity between YscP and FliK proteins was directly found to be significant using a longer sequence of YscP as query (aa 363–515). The Y. pestis FliK region (aa 251–393) was also used as a query for reciprocal PSI-blast searches (same parameters). By iteration 2, LscP from P. luminescens was significantly retrieved (E = 0.005), whereas other members of the YscP family were retrieved by iteration 3 (E-value of 4 × 10−5 for Y. enterocolitica YscP). Proteins of the HrcP family were retrieved just below the threshold value after convergence by iteration 22 (e.g. R. solanacearum HpaP using the Y. pestis YscP sequence as query; E-value of 0.13, 18% sequence identity over 90 aa) or by iteration 14 (e.g. R. solanacearum HpaP using the V. parahaemolyticus LafE sequence as query; E-value of 0.023, 18% sequence identity over 90 aa).

Although marginal, the potential similarity between YscP/FliK and HrcP proteins was further supported at the 2D level by using hydrophobic cluster analysis (HCA) (Gaboriaud et al., 1987; Callebaut et al., 1997) (Fig. 2A and B and legend). The aligned sequences were also searched against databases using HMMER (Eddy, 1998), but such searches did not identify proteins other than those highlighted before. Interestingly the T3S4 domain is localized in the C-terminus in every member of the family, suggesting that this specific position might be relevant to the mechanism of length control (Fig. 3).

Figure 2.

A. Comparison of the HCA plots of T3S4 domains of the YscP, FliK and HrcP families, highlighting conserved regular secondary structures. Protein sequences are shown on a duplicated alpha helical net on which the encircled hydrophobic amino acids (V, I, L, M, F, Y, W) form clusters. The positions of these clusters statistically match those of the regular secondary structures (alpha helices and beta strands). This analysis gives access to 2D signatures, which are much more conserved than 1D sequences and thus help sequence comparison at high levels of divergence. Guidelines to the use of HCA are given in the study by Gaboriaud et al. (1987) and by Callebaut et al. (1997) whereas recent publications can be found at the following URL ( The way to read the sequences and special symbols are indicated in the inset. Representative sequences of the three families are shown: YscP from Y. enterocolitica[GenBank identifier (gi) 17839593], FliK from Y. pestis (gi 16121061) and RspP from P. fluorescens (gi 15042139). Cluster similarities are indicated in green, sequence identities in orange. Amino acids shaded pink within the box in the YscP sequence (‘repeated sequence’) correspond to identities relative to a sequence fragment found upstream (aa 243–255). Vertical bars indicate cluster correspondences (labelled from 1 to 7). Clusters labelled 2, 3, 4 and 7 are highly representative of β-strands structures (Hennetin et al., 2003; K. Le Tuanet al. in preparation). The sequence of InvJ from S. entericaLT2 (gi 16766198) is shown at the bottom, below the YscP, FliK and RspP sequences, in order to illustrate the putative correspondence that can be identified using HCA with cognate T3S4 domains. In yellow are indicated some identity ‘clusters’ that can be identified relative to the FliK sequence, strengthening the putative relationship.
B. Representative multiple alignment of T3S4 domains, as deduced from PSI-blast analysis and adjusted manually, in particular considering Hydrophobic Cluster Analysis (HCA) (see before). Species and protein names are given on the left; GenBank identifiers (gi) are on the right. N- and C-termini amino acid positions are also indicated. Stars indicate the C-terminal ends of the sequences. The T3S4 domains are boxed, whereas acidic and polar amino acids (D, E, N, Q, S, T) C-terminal to the T3S4 domains are coloured in red, highlighting the presence of acidic/polar regions downstream the T3S4 domains in the YscP and FliK families. These forms ‘hinge’ regions separating T3S4 domains from short C-terminal globular sequences, evidenced by the presence of hydrophobic clusters (shaded grey). The SctP protein is specific of the subspecies laumondii TTO1 of Photorhabdus luminescens. Conserved hydrophobic amino acids (V, I, L, F, M, Y, W) or amino acids that can substitute them (S, T, A, C) are boxed green, whereas other conserved positions are shaded using other colours. The secondary structure prediction, performed using JPred (Cuff and Barton, 1999), is shown beneath the alignment. H and E stand for helix and extended (β-strand) respectively.

Figure 3.

Schematic representation of T3S4 domain-containing proteins equivalent to YscP from the different TTSS.

Finally, because it was anticipated that the T3S4 domain, detected in the YscP, FliK and HrcP families, would also be present in the Spa32 and InvJ family, we screened these last sequences using HCA for the presence of 2D markers of the T3S4 domain, which could have been missed by other search methods. This led to identify clear markers of the T3S4 domain in the Spa32/InvJ sequences (Fig. 2A and B, see also below), suggesting that these proteins could indeed share the T3S4 domain with the YscP, FliK and HrcP families.

Characterization of the T3S4 domain

The T3S4 domain defined in the preceding section between aa 385 and 500 of YscP, has all the characteristics of a globular domain, as visualized using HCA (Callebaut et al., 1997) (Fig. 2A). It contains approximately one-third of hydrophobic amino acids, organized in clusters, the lengths of which are typical of those of regular secondary structures. Based on the YscP sequence, the limits of the globular domain can be refined on both sides (Fig. 2A): (i) it starts at aa 403 as the preceding sequence (∼aa 391–403) corresponds to a repeated fragment, the first copy of which can be found upstream in the YscP sequence (∼aa 243–255) and it ends up at aa 492, where a long sequence devoid of hydrophobic amino acids starts. The latter sequence likely defines a hinge separating the switch domain from a small region, which might organize as an α-helix.

The T3S4 domain (arrow in Fig. 2A and boxed in Fig. 2B) has an α /β predicted fold, with a core region having almost no insertion-deletion. Seven hydrophobic clusters, as defined using HCA and numbered up to the alignment of Fig. 2A and B, delineate the predicted secondary structures, for which buried positions are highlighted by conservation of hydrophobic amino acids (boxed green in Fig. 2B). A conserved proline and a conserved glycine (P-x-L-G) between the predicted β-strands β1 and β2 seems to represent a signature of the T3S4 domain. No clear conservation is observed outside from the hydrophobic amino acids and loop markers, rendering a catalytic function unlikely.

Strikingly, in proteins from the YscP and FliK subgroups, the T3S4 domain is followed by a clear hinge region, particularly rich in acidic residues (represented in red in Fig. 2B). A hydrophobic cluster or a group of hydrophobic clusters (shaded grey on Fig. 2B) follows this hinge region and ends the sequences in the YscP and FliK subgroups respectively. These additional secondary structure(s) (likely α-helices) might regulate the T3S4 domain function. However, deletion of this 15-residue tail from YscP had no major impact on the needle length (data not shown) and on Yop secretion (Fig. 1). In the group of proteins from plant pathogens, this tail is missing and the predicted last β-strand of the T3S4 domain ends the sequence.

Finally, additional secondary structures appeared to be present N-terminal to the first alignment block, suggesting that variable sequences might complete the T3S4 domain core.

Fold prediction methods [3DPSSM (Kelley et al., 2000), FUGUE (Shi et al., 2001)] did not highlight any known 3D structures, which might be compatible with the T3S4 domain structure.

The closest T3S4 domains are exchangeable

The overall structure of the T3S4 domain seems thus to be highly conserved among the different members of the YscP/FliK family. Not only the expected fold but also the size of the domain is the same, even though the full-length proteins are of different sizes (Fig. 3). This suggested that the T3S4 domains of different proteins could be exchanged. We tried to replace the T3S4 domain of YscP by the one of AscP (Aeromonas salmonicida) (Burr et al., 2003), PscP (Pseudomonas aeruginosa) and the one of FliK from S. enterica serovar Typhimurium and from Yersinia pestis. To do this, we first introduced two restriction sites on both sides of the region encoding the predicted T3S4 domains in the yscP gene cloned downstream from pBAD. The resulting YscP protein was still fully functional for both needle length control (data not shown) and Yop secretion (Fig. 4), suggesting that the proposed delineation of the T3S4 domain was correct and that both insertions were performed outside the domain. We then cloned the region encoding the T3S4 domain from the different proteins in the inserted sites and the recombinant plasmids were introduced in the yscP strain. Secretion of Yops was triggered in vitro and monitored. The YscP hybrid proteins, in which the T3S4 domain was replaced by the T3S4 domain of AscP or PscP (40% and 38% sequence identity respectively), could restore a wild-type phenotype regarding Yop secretion. So it appears that T3S4 domains of AscP and PscP are capable of playing the switch role of YscP although their sequence identity is limited. In contrast, the protein carrying the FliK T3S4 domain either from S. enterica ser Typhimurium or from Y. pestis could not complement the yscP mutant. Thus, T3S4 domains could be exchanged between proteins from the same family of injectisomes. But when it comes to exchange T3S4 domains between proteins involved in different systems such as flagellum and injectisome, the function cannot be performed any more.

Figure 4.

Swapping of the T3S4 domains.
A. Coomassie stained SDS-PAGE of culture supernatants of Y. enterocolitica MRS40(pLJ4036) complemented with 1 pLJ6 (wild-type YscP, control); 2 pCA88 (introduction of a XbaI restriction site); 3 pCA89 (XbaI and BglII restriction sites); 4 pCA90 (T3S4 of AscP); 5 pCA91 (T3S4 of FliKS.t.); 6 pCA92 (T3S4 of PscP); 7 pCA93 (T3S4 of FliKY.p.). The size of YscP is indicated by an arrow.
B. Histograms of the needle length measurements, and electron micrographs of yscP + pCA90 (T3S4 of AscP) and yscP + pCA91 (T3S4 of FliK). M, mean of the lengths; N, number of needles measured.

We then analysed if the hybrid proteins YscP-AscP and YscP-PscP conferred needle length control. We observed by electron microscopy that the length of most needles was controlled but the control was leaky: the histograms of length distribution show a clear wt peak around 55 nm but some needles are longer (Fig. 4). Thus, it looks as though in hybrid proteins, the substrate specificity switch works but is not well controlled by the ruler.

Site-directed mutagenesis of the T3S4 domain

Based on the alignment performed using HCA and ClustalW combined, some residues appeared to be conserved. Most of them are the so-called topohydrophobic residues that are of great importance for the structure of domains and therefore their mutation would dramatically affect the structure. But there are also a few other positions conserved in all the T3S4 domains (R418, Q488), which are probably not as critical for the structure of the domain. We decided to mutate them by alanine replacement to perhaps identify residues essential for the function. These site-directed mutations were again engineered on the yscP+ gene cloned in the pBADMycHisA vector, downstream from the arabinose promoter. The different constructs obtained were then transformed in the yscP strain and secretion was monitored at 37°C after Ca2+ chelation. The mutants had a wild-type phenotype regarding Yop secretion (Fig. 5). Even though these two positions were conserved in all the T3S4 domains, they were not affected by an alanine replacement.

Figure 5.

Site-directed mutagenesis of T3S4.
A. Western blot analysis (total cells; polyclonal anti-YscP antibody) of 1 yscP + pLJ6 (YscP wt); 2 yscP + pCAP77; 3 yscP + pCAP78; 4 yscP + pCA79; 5 yscP + pCAP80; 6 yscP + pCAP81; 7 yscP + pCAP82; 8 yscP + pCAP85.
B. Yops proteins secreted by the same Y. enterocolitica E40 strains as in A (Coomassie stained SDS-PAGE).

Considering that swapping the T3S4 domains between YscP, AscP and PscP yielded functional hybrids, we decided also to focus on positions conserved in the T3S4 domains of only this subgroup of proteins (Q472, E479, R480, Q482 and P486). None of the five single alanine replacements performed led to a Yop secretion mutant phenotype. The lack of effect of alanine substitutions strengthen the importance of the structure of the T3S4 domain for the function. Indeed no essential residues were highlighted beside the hydrophobic ones. Finally, we considered to mutate constitutive amino acids of the domain signature (P-x-L-G). So we performed alanine replacements of residues P440 and L442 although we were aware of the fact that a replacement of the proline could impact the local structure of the β-turn between the predicted β1 and β2 strands. The phenotype observed for the L442A mutant was wild-type for both functions (Fig. 6). In contrast, the P440A mutant secreted uniformly less Yops than wild-type bacteria. Hence, we also analysed needle length control in this mutant and found an intermediary phenotype (Fig. 6). Indeed a wt peak (55 nm) was observed but there were also needles completely deregulated, some being extra-longs. It seems thus that the domain still exerts its function in some individuals but not in others. The decreased efficiency may be attributed to a slight modification of T3S4 fold resulting from the replacement of the proline, usually crucial for the turn structure.

Figure 6.

Site-directed mutagenesis of the signature.
A. Western blot analysis (total cells; polyclonal anti-YscP antibody) of yscP; yscP + pLJ6 (YscP wt); yscP + pCAP83 (YscPL442→A); yscP  + pCAP86 (YscPP440→A).
B. Yops proteins secreted by the same Y. enterocolitica E40 strains as in A (Coomassie stained SDS-PAGE).
C. Histograms of the needle length measurement of the same strains as in A. Note the difference in scale of the x-axes.

All the observations made are not only validating the predicted characteristics of the structure of the T3S4 domain but also pinpointing the importance of its overall structure for the function.


We have previously reported that YscP acts as a molecular ruler and that the first 35 residues as well as the last 130 residues are required for the needle length control (Journet et al., 2003). We postulated from this observation that the two ends of YscP act as anchors. One end would be connected to the growing tip of the needle whereas the other end would be attached to the basal body. When YscP would be fully stretched, it would signal via its internal anchor to the secretion apparatus that would stop exporting YscF and the needle would stop growing (Journet et al., 2003). Here we show that the C-terminal domain, which is required for length control is also required for Yop secretion. Thus, the C-terminal domain is required for the two functions of YscP: control of Yop secretion and control of needle length. Such a dual function has previously been shown for FliK, which controls secretion of flagellin as well as the length of the hook of the flagellum (Williams et al., 1996; Minamino and Macnab, 2000; Minamino et al., 2004). To explain this phenotype, Williams et al. (Williams et al., 1996) suggested that FliK would act on the basal body protein FlhB to switch the substrate specificity of the export apparatus. In other words, FliK would control the length of the hook by switching the export apparatus from hook subunit secretion to flagellin secretion. This interpretation was further supported by the isolation of suppressors of fliK mutations, which mapped in flhB (Minamino and Macnab, 2000; Fraser et al., 2003a). YscU from the basal body of the Yersinia Ysc injectisome is quite similar to FlhB, which prompted the group of H. Wolf-Watz to engineer yscU mutations, which were similar to the suppressive flhB mutations. As expected by the authors, these mutations turned out to be suppressive of yscP mutations (Edqvist et al., 2003) indicating that YscP presumably exerts the same switch function as FliK does. FliK consists of three regions: an N-terminal region, a proline-rich central region, and a C-terminal region highly conserved among FliK proteins from different species (Kawagishi et al., 1996; Williams et al., 1996). This C-terminal region is thought to be the region of FliK that interacts with FlhB (Suzuki and Iino, 1981; Williams et al., 1996; Minamino et al., 1999). We show here that this domain shares structural properties with the C-terminal domain of YscP, the domain which precisely controls Yop secretion. We can thus refine our model and postulate that the N-terminus of YscP is attached to the growing needle while the C-terminus would stay in the secretion apparatus and switch the substrate specificity from YscF to Yops by interacting with YscU.

Orthologues to YscP share less sequence similarity than any other component of injectisomes. Although YscP and FliK seem to perform the same function, they share very little similarity. However, the HCA performed here showed that the switch is a globular domain whose structure is conserved in orthologues, both in injectisomes and in flagella. In addition, according to fold prediction algorithms, this globular domain does not seem to adopt any known 3D structure. This domain that we called T3S4 thus defines a new family of proteins involved in length determination, which we propose to call the ‘FliK/YscP’ family. In good agreement with this conclusion, the T3S4 domain of AscP (from A. salmonicida) and the T3S4 domain from PscP (from P. aeruginosa) could replace the T3S4 domain of YscP and direct Yop secretion. However, the T3S4 domain from FliK was inactive when fused to the ruler part of YscP, showing that the interaction between the T3S4 and the secretion apparatus is rather specific. Interestingly, all the conserved residues between T3S4 domains appear to be hydrophobic, which suggests that the T3S4 domain has no catalytic activity. Proteins of the YscP/FliK family would thus not be directly responsible for the proteolytic cleavage of FlhB and YscU, which has been shown to occur when the substrate specificity changes (Fraser et al., 2003a; Lavander et al., 2003).

A recent biochemical analysis showed that FliK is monomeric in solution and has an elongated shape (Minamino et al., 2004). In addition, the C-terminal switch domain is more globular than full-length FliK. Our results are in perfect agreement with these data. Moreover, the authors cite unpublished results from T. Minamino and R. MacNab demonstrating that FliK interacts with the hook-capping protein FlgD and propose for FliK a model very similar to the one we proposed for YscP, namely that the N-terminal half of FliK within the central channel of the hook-basal body transmits the hook length information to the switch domain that then interacts with FlhB (Minamino et al., 2004). The study of YscP and FliK thus nicely converge to a similar general structure and a similar working model.

An interesting observation that arose from the experiments described in this paper is that the two phenotypes of yscP null mutants, loss of length control and failure to secrete Yops can be dissociated. Indeed, unlike mutants in the T3S4 domain, mutants that are deprived of the N-terminus of YscP do secrete Yops but fail to control length of the needle (Journet et al., 2003). One possible explanation would be that when the YscP constructs are deprived of one of their anchors the switch could operate at random, independently of the trigger given by the stretching of the ruler. In bacteria harbouring such constructs, the model would predict that the needle size would be variable but Yop secretion would occur from needles that have stopped growing. This is compatible with the observed phenotype but it could not be taken as a proof for the model unless one could demonstrate that only needles that have stopped growing do secrete Yops.

Although the model can account for the phenotype of mutants that do no longer tightly control the needle length but still secrete Yops, it predicts that mutants with the opposite phenotype (normal length but no Yop secretion) are unlikely to occur. Indeed, an arrest in the elongation of the needle implies that the secretion of YscF is switched off and thus that the switch is functional. None of the yscP mutants analysed to date displayed such a phenotype. In the flagellum also no fliK mutants controlling hook length but unable to initiate filament assembly were ever described (Williams et al., 1996).

According to our present study of the domain, a full understanding of its functions will probably have to await the determination of its 3D structure as well as that of its potential interacting partner YscU.

Experimental procedures

Sequence analysis

Similarity searches were performed using PSI-blast (Altschul et al., 1997) and HMMer(Eddy, 1998). The bidimensional HCA (Gaboriaud et al., 1987; Woodcock S., 1992; Callebaut et al., 1997) was used for refining the proposed similarities.

Secondary structure prediction was performed using Jpred (Cuff and Barton, 1999). FUGUE (Shi et al., 2001) and 3D-PSSM (Kelley et al., 2000) were used for fold recognition.

Induction of the yop regulon and Yop secretion analysis

Bacteria were routinely grown on Luria–Bertani agar plates and in liquid LB medium. For the induction of the yop regulon, Y. enterocolitica bacteria were inoculated to an OD600 of 0.1 and cultivated in brain–heart infusion (BHI; Remel) supplemented with 4 mg ml−1 glycerol, 20 mM MgCl2 and 20 mM sodium oxalate (BHI-Ox) 2H at 22°C, then shifted to 37°C and incubated for 4 h. Expression of the different yscP genes cloned downstream from the pBAD promoter was induced by adding 0.2% arabinose to the culture just before the shift at 37°C, and again 2 h later. Ampicillin was used at a concentration of 200 µg ml−1 to select for the expression plasmids.

Proteins from the supernatant were precipitated overnight at 4°C with trichloroacetic acid 10% (w/v) final. Electrophoresis was carried out in 12% or 15% (w/v) polyacrylamide gels in the presence of SDS (SDS-PAGE). Proteins secreted by 3 × 108 bacteria were loaded per lane. For the total bacterial cells, the proteins from 107 bacteria were loaded per lane. After electrophoresis, proteins were stained with Coomassie brilliant blue (Pierce) or transferred by electroblotting to a nitrocellulose membrane. Immunoblotting was carried out using a polyclonal rabbit anti-YscP antibody (MIPA57).

Detection of immunoblots was performed with a secondary antibody conjugated to horseradish peroxidase (1:2000; Dako) before development with supersignal chemiluminescent substrate (Pierce).

Electron microscopy

Visualization of the needle-like structures at the cell surface of the bacteria was performed by electron microscopy as described by Hoiczyk and Blobel (Hoiczyk and Blobel, 2001). After 4 h of induction at 37°C, bacteria were harvested at 2000 g and resuspended gently in 20 mM Tris-HCl, pH 7.5. Droplets were applied for 1 min to freshly glow-discharged, formvar-carbon coated grids, and negatively stained with 2% (w/v) uranylacetate. Bacteria were visualized in a Philips Morgagni 268D electron microscope at a nominal magnification of ×44 000 or 27 000 and an acceleration voltage of 80 kV. Sizes were measured with the ‘Soft Imaging System’ software (Hamburg, Germany).

Construction of plasmids

The full list of plasmids used in this study is given in Table 1. DNA amplification for cloning purposes was made using the oligonucleotides listed in Table 2 and the Vent polymerase (Biolabs). Deletions were generated by inverse polymerase chain reaction, using the Pfu turbo polymerase (Stratagene). Both strands of each construct were sequenced using 3100-Avant genetic analyser (ABI Prism).

Table 1. Plasmids used in this work.
PlasmidsEncoded proteinGenotype or descriptionSource or reference
 pYV40 Wild-type virulence plasmid from strain Y. enterocolitica E40 Sory et al. (1995)
 pLJ4036 pYV40 yscPThis work
 pBADMycHisA  Invitrogen
 pCA1YscPΔ1−15 pBADMycHisA – yscP Δ1−15 Journet et al. (2003)
 pCA18YscPΔ385−424 pBADMycHisA – yscP Δ385−424 Journet et al. (2003)
 pCA5YscPΔ16−25 pBADMycHisAyscPΔ16−25 Journet et al. (2003)
 pCA6YscPΔ26−35 pBADMycHisA – yscP Δ26−35 Journet et al. (2003)
 pCA7YscPΔ36−45 pBADMycHisAyscPΔ36−45 Journet et al. (2003)
 pCA79YscPQ482→AReplacement of Q482 by alanine by inverse PCR on pLJ6 using oligonucleotides 3467 and 3468This work
 pCA83YscPL442→AReplacement of L442 by alanine by inverse PCR on pLJ6 using oligonucleotides 3460 and 3461This work
 pCA86YscPP440→AReplacement of P440 by alanine by inverse PCR on pLJ6 using oligonucleotides 3458 and 3459 
 pCA88YscPΩ1Introduction of a XbaI site between codons 405 and 406 of yscPentero by inverse PCR on pLJ6 using oligonucleotides 3478 and 3479This work
 pCA89YscPΩIntroduction of a BglII site between codons 495 and 496 of yscPentero by inverse PCR on pCA88 using oligonucleotides 3480 and 3481This work
 pCA9YscPΔ46−96 pBADMycHisA – yscP Δ46−96 Journet et al. (2003)
 pCA90YscPΩT3S4AscPInsertion of aa 149 to aa 238 from AscP (Aeromonas salmonicida), amplified using oligonucleotides 3487 and 3488, into YscPΩThis work
 pCA91YscPΩT3S4FliKS.t.Insertion of aa 268 to aa 353 from FliK (Salmonella typhimurium LT2), amplified using oligonucleotides 3476 and 3477, into YscPΩThis work
 pCA92YscPΩT3S4PscPInsertion of aa 263 to aa 348 from PscP (Pseudomonas aeruginosa), amplified using oligonucleotides 3569 and 3570, into YscPΩThis work
 pCA93YscPΩT3S4FliKY.p.Insertion of aa 261 to aa 348 from FliK (Yersinia pestis), amplified using oligonucleotides 3567 and 3568, into YscPΩThis work
 pCAP19YscPΔ425−464Deletion from pLJ6 using oligonucleotides 3244 and 3245This work
 pCAP47YscPΔ97−137Deletion of codons 97–137 of yscPentero from pLJ6 by inverse PCR using phosphorylated oligonucleotides 3318 and 3319, followed by a ligationThis work
 pCAP48YscPΔ137−177Deletion of codons 137–177 of yscPentero from pLJ6 by inverse PCR using phosphorylated oligonucleotides 3320 and 3321, followed by a ligationThis work
 pCAP49YscPΔ177−197Deletion of codons 177–197 of yscPentero from pLJ6 by inverse PCR using phosphorylated oligonucleotides 3322 and 3323, followed by a ligationThis work
 pCAP50YscPΔ197−216Deletion of codons 197–216 of yscPentero from pLJ6 by inverse PCR using phosphorylated oligonucleotides 3338 and 3339, followed by a ligationThis work
 pCAP56YscPΔ465−485Deletion from pLJ6 using oligonucleotides 3340 and 3341This work
 pCA57YscPΔ485−500Deletion from pLJ6 using oligonucleotides 3342 and 3343This work
 pCAP77YscPQ488→AReplacement of Q488 by alanine by inverse PCR on pLJ6 using oligonucleotides 3462 and 3464This work
 pCAP78YscPQ472→AReplacement of Q472 by alanine by inverse PCR on pLJ6 using oligonucleotides 3465 and 3466This work
 pCAP80YscPP486→AReplacement of P486 by alanine by inverse PCR on pLJ6 using oligonucleotides 3469 and 3470This work
 pCAP81YscPE479→AReplacement of E479 by alanine by inverse PCR on pLJ6 using oligonucleotides 3472 and 3473This work 
 pCAP82YscPR480→AReplacement of R480 by alanine by inverse PCR on pLJ6 using oligonucleotides 3474 and 3475This work
 pCAP85YscPR418→AReplacement of R418 by alanine by inverse PCR on pLJ6 using oligonucleotides 3456 and 3457This work
 pLJ11YscPΔ222−306 pBADMycHisA – yscP Δ222−306 Journet et al. (2003)
 pLJ6YscPwt pBADMycHisAyscPE40 Journet et al. (2003)
 pLJ7YscPΔ222−381 pBADMycHisA – yscP Δ222−381 Journet et al. (2003)
 pLJC12YscPΔ501−515Cloning of the YscPΔ501−515 coding sequence, amplified by a PCR on pYV40 using oligonucleotides 3064 and 3072 (introducing NcoI and EcoRI sites respectively), in NcoI and EcoRI sites of pBADMycHisAThis work
Table 2. Oligonucleotides used in this work.
CodesOligonucleotidesUnderlined sites


We thank M. Dürrenberger, G. Morson and U.M. Spornitz for use of electron microscopy facilities. We also thank J.M. Meyer for providing Pseudomonas aeruginosa PAO1 strain, K. Hughes for Salmonella typhimurium LT2 strain and J. Frey for Aeromona salmonicida JF2267. This work was supported by the Swiss National Science Foundation (Grant 32-65393.01).