Flagellar proteins and type III-exported virulence factors are the predominant proteins secreted into the culture media of Salmonella typhimurium


Shin-Ichi Aizawa. E-mail aizawa@nasu.bio.teikyo-u.ac.jp; Tel. (+81) 28 627 7211; Fax (+81) 28 627 7184.


We analysed all major proteins secreted into culture media from Salmonella typhimurium. Proteins in culture supernatants were collected by trichloroacetic acid precipitation, separated in SDS–polyacrylamide gels and analysed by amino acid sequencing. Wild-type strain SJW1103 cells typically gave rise to nine bands in SDS gels: 89, 67, 58, 52, 50, 42, 40, 35 and (sometimes) 28 kDa. A search of the sequences in the available databases revealed that they were either flagellar proteins or virulence factors. Six of them were flagella specific: FlgK or HAP1 (58 kDa), FliC or flagellin (52 kDa), FliD or HAP2 (50 kDa), FlgE or hook protein (42 kDa), FlgL or HAP3 (35 kDa) and FlgD or hook-cap protein (28 kDa). The other four bands were specific for virulence factors: SipA (89 kDa), SipB (67 kDa), SipC (42 kDa) and InvJ (40 kDa). The 42 kDa band was a mixture of FlgE and SipC. We also analysed secreted proteins from more than 30 flagellar mutants, and they were categorized into four groups according to their band patterns: wild type, mot type, polyhook type and master gene type. Virulence factors were constantly secreted at a higher level in all flagellar mutants except a Δmot (motAB deletion) mutant, in which the amounts were greatly reduced. A new morphological pathway of flagellar biogenesis including protein secretion is presented.


Bacteria secrete various substances into culture media not just as results of metabolism, but for more purposeful reasons (for general reviews, see Wandersman, 1996; Hueck, 1998): as the means of communication among themselves (Egland and Greenberg, 1999); as toxic substances such as the bacteriocines against other organisms; as invaders into eukaryotic cells (Galan, 1996); and as spill-over during the construction of extracellular organelles (Aizawa, 1996; Macnab, 1996). Substances to be secreted could be either small chemical compounds or rather bulky protein molecules.

For secretion of proteins, special export systems are required to guide proteins through the physical barrier of membranes (Pugsley, 1993). As Gram-negative bacteria have two different kinds of membranes, the inner and outer membrane, the export system penetrates both membranes. Consequently, many genes are necessary for elaborating such a complicated system. Three types of export systems are currently well known (Pugsley, 1993).

Recently, it has been shown that so-called type III export systems in many pathogenic bacteria play an essential role in pathogenicity or, more specifically, virulence; often, one or more gene clusters called pathogenicity islands encode both virulence factors and structural components of the export apparatus through which the factors are secreted (Lee, 1997; Hueck, 1998).

In S. typhimurium, there are two kinds of type III secretion systems: one for flagellar biogenesis and one for the export of virulence factors. The flagellum is an extracellular structure that grows distally, and the component proteins are exported through its central channel to the growing end. There is a hole (about 2 nm in diameter) running along in the centre of a flagellar filament (Mimori et al., 1995). The passage initiates at the basal body, where a gating mechanism at the cytoplasmic side selects flagellar proteins only out of an overwhelmingly large number of other proteins. It should be noted that the flagellar basal structure is engaged in protein secretion, before it works as the flagellar motor later on.

The transportation of flagellar proteins should not be considered as secretion in a general sense, because most of the proteins to be transported are not intentionally emitted into media. But they are mistakenly emitted; there must be some leaks or spill-over during construction. For example, flagellin is secreted into media as a monomer in the absence of FliD, a cap protein, which helps flagellin to polymerize into flagellar filaments (Ikeda et al., 1996). Hook proteins also leak into medium in the absence of FlgD, a hook-cap protein (Ohnishi et al., 1994). This kind of leakage might occur, albeit briefly, during flagellar construction even in the wild-type strain. The only exceptions among the flagellar proteins are FlgM and FliK. FlgM is an antisigma protein and is purposely emitted to lower its own concentration in the cytoplasm (Hughes et al., 1993). FliK is involved in length control of the hook, but its exact role has been unclear. Just recently, it has been shown that FliK was secreted into media and, after the FliK secretion, the substrate specificity of the export apparatus would change from the rod–hook mode to the flagellin mode (Minamino et al., 1999).

As for pathogenicity, several virulence factors (Sip, Sop and InvJ proteins) are known to be secreted into media (Hueck et al., 1995; Kaniga et al., 1995a,b). The exact roles of these secreted proteins for invasion in host cells have been under intense investigation, and part of the roles of SipA (Zhou et al., 1999) and SipB (Hersh et al., 1999) have been uncovered.

The export apparatus for virulence factors in S. typhimurium was identified as the needle complex (Kubori et al., 1998). The needle complex looks similar to the flagellar basal body by electron microscopy. Several genes from the two systems share common sequence features, suggesting that they might originate from a common ancestor.

In this paper, we describe the complete set of proteins secreted from S. typhimurium. They appear as major proteins in the culture supernatant. We have also analysed the secreted proteins from more than 30 flagellar mutants, which fell into four groups according to their band patterns in SDS gels. As a summary, a schematic model of the order of secreted proteins during flagellar construction is presented.

It is not clear whether flagellar biogenesis and expression of virulence are totally independent events or somehow related to each other in Salmonella. As pathogenicity islands were left genetically intact in this study, we will be able to observe the effects, if any, of flagellar biogenesis on the secretion of virulence factors.


How many proteins are secreted into media?

We analysed protein components secreted from S. typhimurium into culture media. Proteins in culture supernatants were collected by trichloroacetic acid (TCA) precipitation and examined by SDS–PAGE. One millilitre of culture supernatant was enough to observe protein bands by Coomassie brilliant blue (CBB) staining. We have established conditions under which consistent band patterns were obtained.

Supramolecular structures, such as flagella or pili, are released into media during growth. As these multimeric protein structures, especially flagella, might be accumulated during a prolonged incubation, they were carefully removed by high-speed centrifugation (18 500 × g for 15 min). Ultraspeed centrifugation (65 900 × g for 20 min) gave a similar pattern (data not shown), indicating that the supernatant does not contain supramolecules after high-speed centrifugation.

Incubation time was not crucial; basically, the same band pattern was obtained at any time point of growth from 4 h after inoculation (Fig. 1A, lane 1) to overnight incubation (Fig. 1A, lane 2). Protein concentration in the supernatants increased with increasing cell density.

Figure 1.

. SDS gel patterns of secreted proteins from S. typhimurium SJW1103. A. The culture grown with mild aeration was clarified by high-speed centrifugation: after 4 h incubation (lane 1) and after overnight incubation (lane 2). For an overnight culture with vigorous aeration, 5 ml of culture in a 100 ml flask was vigorously shaken, and the supernatant was obtained by high-speed centrifugation or ultraspeed centrifugation (lane 3). On the right, an arrow indicates flagellin, and three arrowheads indicate proteins whose amounts were greatly reduced after overnight incubation. B. Molecular weights of secreted proteins. The sample from SJW1103 [the same as in (A), lane 1] was stained with CBB (lane 1). The flagellar hook–basal body (HBB) complex isolated from SJW1103 was used as the protein marker (lane 2). Molecular weights of the secreted proteins were estimated from the HBB band pattern and are indicated on the left of the gel. A 12.5% gel was used.

Aeration was crucial; the amount of flagellin increased greatly when cells were grown overnight with vigorous agitation, resulting in overloading and distortion of adjacent bands in SDS gels (Fig. 1A, lane 3). Even after ultraspeed centrifugation, the band pattern remained the same (as lane 3; data not shown); flagellin in the supernatant was still most abundant among the secreted protein, indicating that flagellar filaments would depolymerize into monomeric flagellin in prolonged contact with tiny air bubbles formed during vigorous agitation. On the other hand, the densities of three major bands were appreciably reduced (arrowheads in Fig. 1A). This phenomenon will be discussed intensively later. We routinely used culture media after either 4 h incubation with vigorous aeration or overnight incubation with mild aeration.

A wild-type strain, SJW1103, typically gave rise to eight CBB-stained bands in gels. Their apparent molecular weights were estimated from the HBB band pattern as follows: 89, 67, 58, 52, 50, 42, 40 and 35 kDa (Fig. 1B, lane 2). With silver staining, several minor bands appeared in the low-molecular-weight region: 32, 30, 28 and 14 kDa (data not shown). These bands were seen even with CBB staining in some mutants (see below). As the 28 kDa protein was often detected with CBB staining, as seen in 4Fig. 4A, we counted it as a major protein.

Figure 4.

. SDS gel patterns of secreted proteins from various flagellar mutants of S. typhimurium. The samples were prepared from overnight culture with mild aeration unless otherwise mentioned. 12.5% gels were used. A. Band patterns of secreted proteins from various flagellar mutants were categorized into four groups. Typical patterns of the four groups are shown: wild type SJW1103 (group1; lane 1), Δmot mutant SJW2241 (group2; lane 2), fliK mutant SJW107 (group3; lane 3) and flhD mutant SJW1457 (group4; lane 4). As a control, the culture supernatant from a double mutant (flhD, invA) SB912 was run in lane 5. The positions of flagellar proteins (arrows) and virulence factors (arrowheads) are shown on the left. B. Band patterns of members of group 1:ΔfliC mutant SJW2536 (lane 2); fliD mutant SJW2152 (lane 3); flgL mutant SJW2176 (lane 4); and flgK mutant SJW2160 (lane 5). Each mutant lacks a protein (indicated on the right) corresponding to its mutation. C. Band patterns of members of group 2 (1): wild-type SJW1103 (lane 1); motA mutant SJW3003 (lane 2); motB mutant SJW2971 (lane 3); and Δmot mutant SJW2241 (lane 4). Lanes in the left half show proteins from 4-h culture, and those in the right half are from overnight culture. D. Immunoblotting for detection of hook protein (FlgE) in the 42 kDa band of members of group 2 (2): wild-type SJW1103 (lane 1); fliG/mot mutant SJW2274 (lane 2); fliM/mot mutant SJW1764 (lane 3); and fliN/mot mutant SJW1784 (lane 4). Proteins in SDS gel (top) were transferred onto PVDF membrane (bottom) and immunostained with anti-FlgE antibody. E. Band patterns of hook-related mutants: wild-type SJW1103 (lane 1); flgE mutant SJW1473 (lane 2); fliK mutant SJW107 (lane 3); flhB mutant SJW3124 (lane 4); and flgD mutant (lane 5). Numbers on the right indicate the positions and sizes of minor proteins in low molecular ranges. 15% gel was used. F. Band patterns of export apparatus mutants: wild-type SJW1103 (lane 1); fliH mutant SJW2748 (lane 2); fliI mutant SJW2684 (lane 3); fliJ mutant SJW104 (lane 4); fliO mutant SJW1431 (lane 5); fliP mutant SJW1515 (lane 6); fliQ mutant SJW185 (lane 7); and fliR mutant SJW183 (lane 8). G. Band patterns of rod and PL ring mutants: wild-type SJW1103 (lane 1); flgB mutant SJW1461 (lane 2); flgH mutant SJW1460 (lane 3); and flgJ mutant SJW1456 (lane 4); and fliE mutant SJW1372 (lane 5). H. Immunoblotting for detection of a rod protein (FlgC) in various flagellar mutants: wild-type SJW1103 (lane 1); fliC (lane 2); flgK mutant (lane 3); flgD mutant (lane 4); flgE mutant (lane 5); flgH mutant (lane 6); flgJ mutant (lane 7); flgB mutant (lane 8); fliE mutant (lane 9); and flhD mutant (lane 10). Proteins were transferred onto PVDF membrane and immunostained with anti-FlgC antibody. Upper bands are FlgC signal (arrowhead), and the lower bands are non-specific. 15% gel was used.

Identification of secreted proteins by amino acid sequencing

These major proteins were separated in SDS gels (Fig. 2A) and blotted onto polyvinylidene difluoride (PVDF) membranes (Fig. 2B). Most proteins were transferred to the membrane, whereas the 42 kDa band remained behind in the gel (Fig. 2C), suggesting that this protein is extremely basic (see next section).

Figure 2.

. Band patterns of secreted proteins blotted. A. Secreted proteins from wild type were separated in an SDS gel and stained with CBB (left lane). Sample from a flhD mutant was run in the right lane for comparison. B. Band patterns blotted from (A) to PVDF membrane. Note that the 42 kDa protein is missing (arrowhead). C. Proteins remaining in the original gel were revealed with CBB staining. The 42 kDa protein and a few more bands were not transferred but remained behind.

Blotted proteins were applied to an amino acid sequencer. The first five to 10 amino acids of each protein were sequenced, and the sequences obtained were screened for identification in the SWISSPROT protein database. Each protein was uniquely identified even with a sequence of five amino acids by three strict criteria: (i) it is a protein of S. typhimurium; (ii) the sequence starts from or near the N-terminus; and (iii) the molecular size has to match with the predicted one. Interestingly, the secreted proteins from S. typhimurium were predominantly either flagellar proteins or virulence factors (Table 1).

Table 1. . Summary of secreted proteins detected in culture supernatants of S. typhimurium.Thumbnail image of

The 89 kDa protein is SipA.

SipA is a protein secreted through the needle complex (a type III secretion apparatus encoded in the inv and spa loci) and has a predicted molecular weight of 73 776 Da (a polypeptide of 684 residues). Strangely, SipA was not necessarily required for bacterial entry into host cells (Kaniga et al., 1995b). It turned out that SipA interacts with skeletal actin filaments in the host cell, enhancing the process of bacterial uptake (Zhou et al., 1999).

It should be noted that the apparent molecular weights of Sip proteins previously described by Kaniga et al. (1995b) are different from ours; our marker proteins are flagellar proteins, whereas they used standard marker proteins commercially available.

The 67 kDa protein is SipB.

SipB is a polypeptide of 593 residues with a predicted molecular weight of 62 411 Da. Interestingly, the central domain contains the potential membrane-spanning regions, suggesting its ability to bind to the host cells. Unlike SipA, SipB is essential for bacterial entry into host cells (Kaniga et al., 1995a).

The 58 kDa protein is FlgK.

FlgK is a junction protein between the hook and the flagellar filament, and is termed the hook-associated protein 1 (HAP1; Homma et al., 1985). FlgK is thought to form one layer at the tip of the hook, when the hook length reaches its mature length (≈ 55 nm). As FlgK is detected in the HBB complex after acid treatment of intact flagella, it seems to bind firmly to the hook (Aizawa et al., 1985). Therefore, FlgK detected in the culture supernatant could be those monomers that were spilt over during or after the formation of the FlgK zone on the hook.

The 52 kDa protein is FliC.

FliC is flagellin. In wild-type strains of S. typhimurium, there are two flagellin genes; the alternate expression of the two flagellin species (a phenomenon called phase variation) gives cells a survival advantage in vivo (Lederberg and Iino, 1956). In strain SJW1103, only one flagellin gene is expressed by deleting the other gene, introducing simplicity for in vitro experiments (Yamaguchi et al., 1984).

The 50 kDa protein is FliD.

FliD has been called HAP2, as it was originally found attached to the hook in some filament-deficient mutants. However, it is a cap protein binding at the tip of the filament and plays an essential role in the polymerization of flagellin molecules into a filament; without FliD, flagellin is spilt into media as monomers (Ikeda et al., 1996). From its role as a chaperone moving at the growing tip, the binding affinity of FliD to the filament should not be too strong; it is very likely that FliD occasionally falls off from its working position. The question how often FliD is supplied to the growing tip has not been answered yet.

The 42 kDa protein is a mixture of SipC and FlgE.

The protein that was hardly transferred to PVDF membranes but remained in the gel was SipC. As a matter of fact, SipC is the only basic protein among the secreted proteins; its predicted pI is 9.53. Although the efficiency of transferring SipC to the membrane was low, the blotted SipC protein was practically enough to sequence.

Another protein detected from the band was FlgE, the hook protein. Although the molecular weights of these two proteins are slightly different, they often overlapped in one band when overloaded for sequencing.

The sipC gene locates downstream of the sipB gene. SipC is a polypeptide of 409 residues with a predicted molecular weight of 43 997 Da. SipC as well as SipB is essential for bacterial entry into host cells (Kaniga et al., 1995a). The amount of SipC often varied from preparation to preparation. Especially when oxidized by vigorous aeration, SipC as well as SipA or SipB was reduced to a negligible amount (Fig. 1A, lane 3). This anomalous behaviour of SipC will also be discussed in the following sections.

The 40 kDa protein is InvJ.

InvJ, previously termed SpaN, is required for entry into the host cell. The predicted molecular weight of InvJ is 36 415 Da (Collazo et al., 1995). InvJ has homologous sequence profiles to FliK, one of the secreted proteins from the flagellar export apparatus (see Discussion).

The 35 kDa protein is FlgL.

FlgL is another junction protein between the hook and filament, and therefore called HAP3. As FlgL is seldom detected in the purified HBB complex, it might bind weakly to HAP1 (Homma et al., 1985).

The 28 kDa protein is FlgD.

FlgD is a hook-cap protein. It works for hook formation in the same way as FliD for filament formation; it stays at the tip of the growing hook. However, when the hook grows to the mature length, it is replaced by HAP1; therefore, FlgD is a scaffolding protein (Ohnishi et al., 1994).

The major proteins secreted from S. typhimurium were predominantly either flagellar components or virulence factors. This conclusion is consistent with the observation that the wild-type strain SJW1103 that we used in experiments has expressed both flagella and needle complexes at the same time on a cell (Fig. 3). These two supramolecular structures are the major outlets of protein secretion systems in S. typhimurium. This is confirmed by the fact that there were no significant amounts of proteins secreted from a double mutant of flhD and invA, in which neither flagellum nor needle complex is expressed (Fig. 4A, lane 5). Without these two export systems, obviously no proteins can penetrate the membranes.

Figure 3.

. Electron micrograph of the cell surface of an SJW1103 cell. Both flagellar basal body and needle complex (arrows) are seen on the cell surface of an osmotically shocked cell. Negatively stained with 1% PTA (pH 7). Bar represents 100 nm.

Secretion proteins from various flagellar mutants

We have also examined secreted proteins from various flagellar mutants (Table 2). The band patterns of secreted proteins from these mutants were categorized into four groups. Briefly, the characteristics of the secretion patterns of these groups are as follows.

Group 1, typified by the wild-type strain, secretes Sip/Inv proteins, hook protein, HAPs and flagellin (Fig. 4A, lane 1). Group 2, consisting of mot mutants, shows some strange secretion patterns of Sip/Inv proteins (lane 2). Group 3, consisting of polyhook mutants, secretes not only the flagellar major proteins, but also the flagellar minor proteins in the small molecular region (lane 3). Group 4, typified by the master gene mutant, secretes Sip/Inv proteins only, but none of the flagellar proteins (lane 4). Now, we examine the secretion pattern of each mutant in detail.

Group 1 (wild type)

We expected that this group would include not only the wild type but also any strains that produce intact flagellar structures, for example mot mutants, which produce structurally intact flagella with paralyzed function. However, one of the mot mutants showed unexpected band patterns, and therefore mot mutants make their own group (see group 2). Instead, we added flagellin and HAP mutants to this group, because only flagellin or the corresponding HAP protein was lacking in the secreted proteins from these mutants (Fig. 4B): (i) wild type: SJW1103; (ii) flagellin mutants: SJW2536 (ΔfliC); (iii) HAP mutants: SJW2152 (fliD), SJW2160 (flgK ) and SJW2176 (flgL).

According to the current theory regarding HAPs, a mutant with a defect in one of the HAPs should not interfere with the secretion of the other HAPs and flagellin. All the HAP mutants we examined obeyed to this rule, except that secreted proteins from these mutants naturally lack the band corresponding to each HAP protein. Virulence factors are secreted as much as the wild type.

However, the amount of flagellin was different. HAP1 (flgK ) and HAP3 (flgL) mutants secreted excess amounts of flagellin, whereas HAP2 (fliD) mutants produced as little as the wild type (Fig. 4B, lanes 3–5). Downstream of the fliD gene are the fliS and fliT genes, which encode chaperones specific for flagellin and HAPs respectively (Yokoseki et al., 1996; Fraser et al., 1999). Our fliD strains used in the experiments might have a polar effect over those genes (see Discussion).

The leakage of appreciable amounts of flagellin from these mutants suggests that the export apparatus is fixed to the flagellin mode, and the leakage continues during growth.

Group 2 (mot type)

As any mot mutants, by definition, can make paralyzed but intact flagellar structures, we expect that they would secrete all flagellar proteins and virulence factors. Most of the single mot mutants we examined behaved as expected, giving the same band pattern as the wild type. However, a double mot mutant gave rise to an unexpected result (Fig. 4C).

Authentic mot mutants: SJW3003 (motA), SJW2971 (motB) and SJW2241 (Δmot).

Secreted proteins from motA and motB mutants were similar to those of the wild type, whereas a Δmot (motAB deletion) mutant secreted more flagellin than any other mot mutants. Furthermore, the amounts of Sip/Inv proteins secreted from this mutant were greatly reduced; only trace amounts were detected in the first 4 h of growth (Fig. 4C, left lane 4). However, the amounts of Sip/Inv proteins recovered to the normal level overnight (Fig. 4C, right lane 4). This is peculiar, as there is no obvious connection between virulence and motility found in Salmonella. The possible reasons for this phenomenon will be discussed later (see Discussion).

Switch (Mot) mutants: SJW2274 (fliG/mot), SJW1764 (fliM/mot) and SJW1784 (fliN/mot).

The fliG, fliM and fliN genes belong to the early genes of flagellar biogenesis. The null mutants in these genes do not make any flagellar structures, except the MS ring complex. The mot phenotype of these genes occurs mostly by single amino acid substitution, allowing the flagellar construction to proceed as in the authentic mot mutants (Sockett et al., 1992; Irikura et al., 1993).

Species of the secreted proteins from the switch (Mot) mutants were the same as the wild type (data not shown). However, amounts of secreted proteins varied occasionally. Here, we show one example in which the 42 kDa proteins of a fliM/mot mutant were much less than those of the other strains (Fig. 4D, top). As the 42 kDa protein is a mixture of the hook protein and SipC, we measured the amount of hook protein secreted from these mutants by immunoblotting. The amounts of hook protein were the same, indicating that the decreased protein in the band was SipC (Fig. 4D, bottom). In fact, the amount of SipC often varied from preparation to preparation. This anomaly of SipC will be discussed later.

Group 3 (polyhook type)

Defects in the fliK gene give rise to abnormally long hooks, polyhooks, and to suppression of the flagellar late genes. Therefore, HAPs and flagellin are not included in the secreted proteins from fliK mutants. Instead, some low-molecular-weight proteins (32, 30, 28 and 14 kDa) were secreted more than those in the wild type that were barely detected. All virulence factors were secreted in the same amounts as in the wild type.

Authentic polyhook mutant: SJW107 (fliK).

It is noteworthy that the smaller proteins are secreted in higher amounts from this mutant than those from any other mutants in other groups (Fig. 4A, lane 3). Amino acid sequencing showed that they were rod proteins and hook-cap protein: 32 kDa (FlgF or proximal rod protein); 30 kDa (FlgG or distal rod protein); 28 kDa (FlgD or hook-cap protein); and 14 kDa (FlgC or proximal rod protein). These results reinforce the conclusion of Minamino and Macnab (1999) that rod and hook proteins belong to a single class of secretion (see Discussion).

Subsidiary polyhook mutants: SJW3124 (flhB/phf).

This type of polyhook mutant has never been found among spontaneous flhB mutants. It should be noted that flagellar construction is completely halted in the null flhB mutant. However, the flhB gene can also suppress fliK mutations, so that authentic polyhooks are turned into polyhook filaments (phf), recovering their motility. A second site mutation in the flhB gene alone gives rise to a phf mutant (Williams et al., 1996). Secreted proteins from this phf mutant were basically the same as those of the wild type (Fig. 4E, lane 4).

Group 4 (master gene type)

This group secretes only virulence factors (SipA, SipB, SipC and InvJ). Mutants retaining defects in genes involved in the early steps of flagellar biogenesis do not secrete any flagellar proteins for several reasons: shut-off of flagellar genes; an incomplete export apparatus; and defects in the passage of export. There are five subgroups falling into this group.

Master gene mutants: SJW1457 (flhD).

No flagellar proteins were secreted from flhD mutants, because the expression of all the flagellar genes was shut off. The amounts of secreted virulence factors stayed at the same level as the wild type (Fig. 4A, lane 4).

Export apparatus mutants: SJW1466 (flhA), SJW1383 (flhB), SJW2748 (fliH) and SJW2684 (fliI), SJW104 (fliJ), SJW1431 (fliO), SJW1515 (fliP), SJW185 (fliQ), SJW183 (fliR).

Defects in export machinery resulted in no secretion of flagellar proteins at all (Fig. 4F).

Rod mutants: SJW1461 (flgB), SJW1392 (flgC), SJW1496 (flgF) and SJW1419 (flgG), SJW1456 (flgJ), SJW1371 (fliE).

FlgB, FlgC, FlgF and FlgG are the structural components of the rod. Recently, it has been shown that FlgJ has muramidase activity and is essential for rod formation (Nambu et al., 1999). FliE is a component of the HBB complex and necessary for rod formation (Kubori et al., 1992). Defects in the rod resulted in no secretion of flagellar proteins at all (Fig. 4G).

PL ring mutants: SJW1460 (flgH) and SJW1351 (flgI).

Both these mutants fail to form the PL ring complex, a physical hole in the outer membrane. Without the complex, no flagellar proteins can go out through the membrane (Fig. 4G, lane 3). As the rod is intact in these mutants, rod and proteins are secreted into the periplasmic space (Minamino and Macnab, 1999).

Hook mutants: SJW155 (flgD) and SJW1473 (flgE).

In these cells, the hook is not formed. However, as the PL ring complex is formed, the distal end of the rod is supposed to be exposed to media. However, no flagellar proteins were detected from these mutants (Fig. 4F). Does this fact indicate that the tip of the rod is still closed?

The duration of the rod-hook mode would be shorter than that of the flagellin mode, as the latter mode dominates the population in a prolonged incubation. Therefore, it is not surprising to find rod proteins much less than HAP proteins. In order to determine whether any rod proteins are secreted or not, we carried out immunoblotting experiments with several key mutants on the pathway of flagella biogenesis. A rod protein, FlgC, was secreted from (flgE, flgD, flgK and fliC) mutants that have basal body structures beyond the PL ring complex (Fig. 4H).

In conclusion, any flagellar proteins will not be transported from the cytoplasm until the export machinery is complete, and will not be secreted into culture media until the outlet of the export apparatus is exposed beyond the outer membrane. The export machinery has substrate specificity and shows at least two modes: the mode for secretion of the rod-hook proteins (the rod-hook mode) to flagellin–HAP proteins (the flagellin–HAP mode). Until the hook is complete, the machinery is fixed to the rod-hook mode.

Once the mode was switched to the latter, HAP proteins and flagellin are transported simultaneously to the growing site. Once filament formation started, useless HAP proteins will be spilt over. Furthermore, the efficiency of polymerization of the axial structures (the hook and filament) is probably not 100%, even with the help of cap proteins, and occasional failures in construction would allow leakage of unused proteins into media.

A schematic model for protein secretion during flagellar morphogenesis

There already exist several versions of the morphological pathway of flagellar biogenesis, which have been revised from time to time for better understanding of a construction mechanism of this complicated structure (for recent versions, see Ohnishi et al., 1994; Aizawa, 1996; Kubori et al., 1992; 1997; Nambu et al., 1999). Here, we present a new version based on the results obtained in this study (Fig. 5). As flagellar construction is not really a simple secretion process, we use the words ‘secretion’, ‘export’ and ‘transport’ as well according to the conditions under which one construction step proceeds.

Figure 5.

. Schematic pathway of protein secretion during flagellar biogenesis. Intermediate flagellar structures are drawn in the order of construction on a cell surface, proceeding in a clockwise direction from the MS ring complex to the mature flagellar structure. Also drawn is the needle structure that is an export apparatus specific for virulence factors. Secreted proteins are divided into three groups: the rod-hook proteins, flagellin-HAP proteins and virulence proteins. The rod-hook proteins are secreted through the flagellar structures before the completion of the hook. The flagellin-HAP proteins are secreted through the flagellar structures beyond the hook. Virulence proteins are secreted through the needle structure all the time independently of the flagellar structure except under special conditions as seen in Δmot mutant (see text).

(i) The MS ring complex self-assembles in the inner membrane (Ueno et al., 1992).

(ii) The export apparatus, presumably the C rod, is formed at the cytoplasmic side of the MS ring complex (Katayama et al., 1996). At the same time, the C ring, consisting of FliG, FliM and FliN, is formed around the C rod. Upon completion of these cytoplasmic structures, secretion of rod proteins begins.

(iii) A rod protein FliE is integrated in the basal structure, and the muramidase FlgJ is secreted into the periplasmic space to make a hole in the peptidoglycan layer (Nambu et al., 1999).

(iv) Other rod proteins (FlgB, FlgC, FlgG and FlgF) are exported to complete the rod (Minamino and Macnab, 1999). The hook formation begins in the periplasmic space with the help of hook-cap protein FlgD (Kubori et al., 1992).

(v) FlgH and FlgI are secreted by the general secretion pathway to form the PL ring complex (not shown). Formation of the outer ring complex allows the rod and growing hook to penetrate the outer membrane.

(vi) The hook keeps growing. During hook construction, the rod, hook and cap proteins are exported and spilt over. FliK is also secreted into media before completion of hook assembly (Minamino et al., 1999).

(vii) When the hook length reaches 55 nm, the substrate specificity of the export apparatus is switched from the rod-hook mode to the flagellin–HAP mode. Rod and hook proteins are not secreted anymore.

(viii) Probably then, the antisigma factor FlgM is secreted, and the class 3 flagellar proteins (flagellin and HAPs) are expressed and exported. FlgD at the hook tip is replaced by HAP1 (Muramoto et al., 1999).

(ix) The HAP1 zone is stabilized by HAP3 (Muramoto et al., 1999). Completion of the HAP zone allows flagellin to polymerize into a filament with the help of the filament-cap protein FliD. HAP proteins and flagellin are spilt over during filament growth.

(x) Mot proteins are added around the basal structure and interact with the C ring to generate the torque. The secretion of flagellar proteins slows down.

(xi) The filament keeps growing for generations (Aizawa and Kubori, 1998).

This study, together with other recent work, show that the flagellar component proteins are secreted in two modes: the rod-hook mode and the flagellin–HAP mode. In one mode, the components belonging to this group are secreted altogether, allowing the spill-over of unused components. In previous schemes, only one component at one time was secreted efficiently to the growing end.

There are still ambiguous steps in this scheme, and detailed studies on these steps are necessary for complete understanding of flagella biogenesis.


We have analysed all major secreted proteins from S. typhimurium and shown that they were exclusively flagellar proteins and virulence factors. In previous studies along the same lines, secreted proteins were mainly detected with specific antibodies against either flagellar proteins or virulence factors, leaving proteins belonging to the other group undetected (Homma et al., 1984; Hueck et al., 1995; Kaniga et al., 1995a; Fraser et al., 1999; Minamino and Macnab, 1999). Our analysis using amino acid sequencing is more direct and unprejudiced than those using immunostaining methods for the identification of major proteins appearing in SDS gels.

The disadvantage of our method is that fair amounts of proteins are necessary for sequencing. Consequently, some minor proteins that were known to be secreted were hardly or seldom detected; for example, antisigma factor FlgM was never detected, and both FliK and SipD were occasionally but not always detected.

Sip proteins were abnormal in appearance in gels; they, especially SipC, often disappeared from gels. We suspect that Sip proteins might be easily oxidized, because they disappeared when incubated with vigorous aeration. In Shigella flexneri, IpaA, B and C (corresponding to SipA, B and C respectively) form filamentous structures and float at the meniscus (Parsot et al., 1995). This possibility is now under examination.

The most unpredicted results of this study were the patterns of secreted proteins from a Δmot mutant; the secretion of virulence factors (Sip/Inv proteins) was greatly reduced after 4–6 h of culture. However, ordinary amounts of Sip/Inv proteins were secreted in overnight culture. Aeration during growth was carefully controlled so that disappearance of those proteins by oxidation would not occur. There is no simple explanation for this phenomenon, but we suspect that global turbulence of the energy level for export (ATP or proton motive force) or some regulatory pathways might affect the kinetics of other export systems. The observation that the amount of flagellin secreted from the Δmot mutant was more than that from the other mot mutants may be related to this possibility.

It should be noted that neither motA nor motB mutant by itself showed such unexplainable secretion. If MotA and MotB made a complex in tandem, which current models prefer, we would expect that both motA and motB mutants would behave like the Δmot mutant. Defects in either gene would result in disruption of the Mot complex, which could be phenologically the same as the Δmot mutant. But this is not the case; two motA and three motB mutant strains examined secreted ordinary amounts of Sip/Inv proteins (data not shown), suggesting that MotA and MotB could interact independently with the basal body. This idea agrees with genetic data obtained from pseudorevertant analysis; suppressors of some motB mutations are found in the fliG gene, suggesting that MotB could interact directly with the C ring (Garza et al., 1995; Togashi et al., 1997).

The junction zone between the hook and filament contains only five or six molecules of both HAP1 and HAP3, which is less than 0.01% of flagellin (Jones et al., 1990). However, the culture supernatants contained relatively high amounts of HAP proteins, indicating that HAP proteins are secreted even during filament elongation where they are useless. As HAP proteins and flagellin belong to the same export mode, the export apparatus may not be able to distinguish flagellin from HAP proteins. The segregation and secretion of substrates are complicated processes, including not only the export apparatus but also chaperones such as FliS, FliT and FlgN (Yokoseki et al., 1996; Fraser et al., 1999).

The HAP2 mutant that we examined first secreted much less flagellin than HAP1 or HAP3 mutants. We also examined six fliD (either spontaneous or Tn10 inserted) mutants in our stock, but none of them secreted the increased amount of flagellin (data not shown). However, it is likely that our mutants have polar effects down to the fliS and fliT genes, judging from a comparison between our data and those of Yokoseki et al. (1995). It is a wise strategy for a cell not to express chaperones for FliD and flagellin when FliD protein is not expressed so that a cell would not waste its energy on abortive protein secretion.

The role of FliK is still controversial; although FliK is involved in hook length control, it is not a molecular ruler (Koroyasu et al., 1998). FliK, together with FlhB, is required in switching of substrate-specific export, but FliK itself is exported into the culture media (Minamino et al., 1999). Our data showed that rod-hook proteins are secreted more in a fliK mutant than in the wild type, indicating that, in the absence of FliK, the export system is fixed indefinitely in the rod-hook mode. The key role of FliK is not its absence (of course!) from the cytoplasm but its interaction with the switching gear when it is exported.

It is interesting to see that FliK has sequence similarity with InvJ, although the similarity as a whole is not high. However, there are homologous sequences here and there, especially Q-rich blocks (QQQxxQQQ) near their C-terminal regions.

Young et al. (1999) showed that a virulence factor, phospholipase, is secreted through the flagellar basal body in Yersinia. There is no obvious relationship between the two systems in Salmonella. As we have learned from this study and other recent work, substrate specificity of the export apparatus is strictly controlled at the gate. We cannot tell whether or not virulence factors could be the substrate of the flagellar export apparatus or vice versa, unless we examine mutants of virulence in the background of various flagellar mutations. This kind of experiment will be one approach to a question in dispute: what is the relationship between flagellar biogenesis and pathogenicity in S. typhimurium?

Experimental procedures

Bacterial strains and growth conditions

All SJW strains used in this study were derived from S. typhimurium SJW1103, and SB912 strain was a gift from Jorge Galan (Yale Medical School) (Table 2).

For culture with mild aeration, 5 ml of LB in a test tube was used, and 5–20 ml of LB in a 100 ml flask or 100 ml of LB in a 500 ml flask was used for culture with vigorous aeration.

Chemicals and culture media

Tryptone, yeast extract and agar were purchased from Difco. All other chemicals used were reagent grade.

L broth (LB) contains 1% tryptone, 0.5% yeast extract and 1% NaCl.

Collection of protein components secreted in media

The cells were removed at 18 500 × g for 15 min. Supernatant (1 ml) was removed into an Eppendorf tube, mixed with prechilled 25% TCA (final concentration 6%), chilled on ice for 15 min and centrifuged at 10 000 × g for 10 min. The pellets were suspended with 0.3 ml of acetone, and the suspension was quickly dissolved by ultrasonic turbulence (Ultrasonic Cleaner Hi-Power SUS-200; Shimadzu) and centrifuged at 10 000 × g for 5 min. Acetone washing was repeated twice to remove TCA from the precipitates completely. The pellets were dissolved in SDS sample buffer and analysed by SDS–PAGE.

From 100 ml of culture media with vigorous aeration, 1.5 ml were collected in Eppendorf tubes at various time points (from 4 h to 16 h) and processed as described above.

Preparation of osmotically shocked cells

In order to observe the flagellar basal structure in the cell wall, cells were depleted of the cytoplasm by osmotic shock treatment as follows.

Cells were harvested at the log phase of growth (between OD650 of 0.5 and 0.7) and resuspended in a small volume (1/100 of the culture volume) of sucrose solution (0.5 M sucrose, 0.15 M trizma base; pH not adjusted). After incubation on ice for 30 min, a large volume (the same as the culture volume) of prechilled water was added to the cell suspension at a burst. Intact cells were removed by low-speed centrifugation (5000 × g for 10 min), and the osmotically shocked cells that remained in the supernatant were collected by high-speed centrifugation (18 500 × g for 20 min). Cells in the pellet were resuspended in water and immediately observed by electron microscopy.


SDS–PAGE was carried out according to Laemmli (1970). Acrylamide concentrations of gels were selected to be 12.5% or 15% depending on the molecular range of the proteins of interest. Gels were stained with silver or CBB. The HBB proteins purified from SJW1103 were used to determine the molecular weights of secreted proteins.

Immunoblotting was carried out to detect hook protein and a rod protein FlgC according to Kubori et al. (1997).

Amino acid sequencing

Proteins in SDS gels were electroblotted onto PVFD membrane in transblotting buffer (10 mM CAPS, 10% methanol, pH 11). The membranes were stained with CBB, and the bands of interest were applied to a Beckman LF3000 protein sequencer equipped with an on-line PTH amino acid analyser (System Gold).

Electron microscopy

Osmotically shocked cells were negatively stained with 1% (w/v) phosphotungstic acid (pH 7.0) and observed with a JEM-1200EXII electron microscope (JEOL). Micrographs were taken at an accelerating voltage of 80 kV.


We thank Mina Yamanoi and Satoshi Kinosihita for preliminary data, Shigeru Makishima for electron microscopy, Yuichi Watabe for technical help, and Takashi Yamamoto and Kazuyuki Yamazaki for figures. We also thank Professor Macnab for discussion, Professor Galan for SB strain, and Professor Kanno for encouragement. This work was supported by a Grant-in-Aid for Scientific Research (B) from The Ministry of Education, Science, Sports and Culture.