N-terminal signal region of FliK is dispensable for length control of the flagellar hook


  • Takanori Hirano,

    1. CREST, Japan Science and Technology Agency (JST), 1064-18 Takahori, Hirata, Takanezawa, Shioya-gun, Tochigi 329-1206, Japan.
    Search for more papers by this author
  • Satoshi Shibata,

    1. CREST, Japan Science and Technology Agency (JST), 1064-18 Takahori, Hirata, Takanezawa, Shioya-gun, Tochigi 329-1206, Japan.
    Search for more papers by this author
  • Kouhei Ohnishi,

    1. Institute of Genetics, Kochi University, 200 Monobe, Nangoku 783-8502, Japan.
    Search for more papers by this author
  • Tomomi Tani,

    1. Department of Molecular Physiology, The Tokyo Metropolitan Institute of Medical Science, Honkomagome 3-18-22, Bunkyo-ku, Tokyo 113-8613, Japan.
    Search for more papers by this author
  • Shin-Ichi Aizawa

    Corresponding author
    1. CREST, Japan Science and Technology Agency (JST), 1064-18 Takahori, Hirata, Takanezawa, Shioya-gun, Tochigi 329-1206, Japan.
      E-mail aizawa@softnano.org; Tel. and Fax (+81) 286 768 510.  
    Search for more papers by this author
    • Prefectural University of Hiroshima, Faculty of Life and Environmental Sciences, 562 Nanatsuka, Shobara, Hiroshima 727-0023, Japan.

E-mail aizawa@softnano.org; Tel. and Fax (+81) 286 768 510.  


The length of the flagellar hook is regulated; it is 55 ± 6 nm long in Salmonella. Five genes involved in hook-length regulation are fliK, flhB, fliG, fliM and fliN. The last four genes encode structural components of the protein export apparatus in the flagellar base, whereas FliK is soluble and secreted during flagellar assembly. The role of FliK, however, remains ambiguous. We constructed two kinds of FliK variants: N-terminally truncated FliK protein and FliK N-terminally fused with cyan fluorescent protein (CFP-FliK). Both N-terminally truncated FliK missing the first 99 amino acids (aa) and CFP-FliK fusion variants partially complemented a fliK null (polyhook) mutant to produce cells with filaments, allowing cells to swim; the hooks, however, were not normal but were polyhooks. When the N-terminally defective FliK variants were expressed at high levels, the average polyhook length was shortened coming close to the length of the wild-type hook, independently of the sizes of the FliK variants. These FliK variants were not secreted. CFP-FliK fusion proteins were observed to homogeneously distribute in the cytoplasm. We conclude that FliK does not need to be exported to control hook length and is unlikely to be a ruler; instead, we conclude that FliK controls hook length by the timely switching of secretion modes of the flagellar type III secretion system  by  the  FliK  C-terminal  domain,  and  that  the  N-terminal region is dispensable for hook length control.


The bacterial flagellum is a motile organelle present in many bacterial species, and the flagellar base works as a rotary motor. The base of the flagellar motor in the cytoplasmic membrane includes a type III secretion system (TTSS) responsible for the export of flagellar protein subunits (Macnab, 1996; Aizawa 2001). The exported flagellar proteins assemble into extracellular structures and, thus, flagellar growth reflects the functional properties of the TTSS more evidently than other systems in which secreted proteins usually diffuse away into the surrounding environment (Aizawa 2004). The flagellar hook, which transmits torque from the motor to the filament, has a closely defined length. Five genes appear directly involved in hook length control: fliK, flhB, fliG, fliM and fliN. (Williams et al., 1996; Makishima et al., 2001).

The fliK gene was the first gene identified as a regulator of hook-length. Mutations in fli K produce hooks of uncontrolled length. These hooks were named superhooks and later renamed polyhooks (Patterson-Delafield et al., 1973). These fliK mutants failed to make the switch from secretion of hook subunits to secretion of late flagellar proteins, which include the hook-filament junction proteins, FlgK and FlgL, the filament subunit proteins, FliC or FljB, the flagellar cap, FliD, and the regulatory protein, FlgM. Second-site suppressors of fliK mutant alleles that restored secretion of late flagellar proteins were located in the flhB gene. Mutations in flhB in the presence of FliK (fliK+) also gave rise to polyhooks with filaments, permitting the cell to swim; the average length of the polyhooks was a little shorter than those of fliK mutants (Hirano et al., 1994; Williams et al., 1996). The fliG, fliM and fliN genes encode proteins involved in switching the direction of motor rotation (Yamaguchi et al., 1986a); the switch proteins are located in the cytoplasmic (C) ring of the motor (Yamaguchi et al., 1986b; Francis et al., 1994). These three proteins are also part of the export apparatus because mutations in fliG, fliM and fliN produce short hooks of defined length, which led us to propose the measuring-cup model for hook length control (Makishima et al., 2001). Because among the long hooks in a polyhook mutant, there is a large population of hooks of wild type length (see below), we concluded that FliK was not directly involved in length control (Koroyasu et al., 1998). However, the bona fide role of FliK has remained elusive.

In Salmonella fliK-deletion mutants, polyhook growth occurs in two phases: the initial rate slows exponentially until the length reaches 55 nm, whereupon hook elongation continues at a slow constant rate (Koroyasu et al., 1998). If the secretion of the hook protein (FlgE) does not stop, the hook grows over many generations to form polyhooks. Hook growth will only cease when FlgE secretion arrested. In wild-type cells, FlgE secretion stops when flagellin (late substrate) secretion begins. Switching between the two modes (FlgE mode and flagellin mode) occurs by cleavage of FlhB, one of the protein components of the flagellar export apparatus (Minamino and Macnab, 2000). This switching is irreversible in that cleaved FlhB allows the secretion of flagellin but not FlgE (Williams et al., 1996; Fraser et al., 2003). In the absence of FliK, FlhB remains intact and FlgE is secreted, thus resulting in polyhooks. The N269A variant of FlhB, which cannot be cleaved even in the presence of FliK, gives rise to polyhooks (Fraser et al., 2003).

In our model, hook length is determined by the capacity of the C ring, which somehow measures the amount of FlgE to be exported (Makishima et al., 2001). However, we don’t know how this measuring cup functions. Hook elongation terminates at a defined length only when secretion is switched from FlgE mode to flagellin mode at the proper time, suggesting that hook length is dependent on the timing of the switching of secretion modes. Thus, FliK seems to work in two ways: (i) by monitoring hook length and (ii) by stimulating FlhB cleavage. Function (i) is assigned to the N-terminal region, and function (ii) to the C-terminal domain (Williams et al., 1996; Minamino et al., 1999; 2004). Because function (i) is vaguely defined, it was proposed that hook length is determined by the physical length of a FliK molecule; i.e. FliK is a molecular ruler (Journet et al., 2003; Minamino et al., 2004; Thomas and Finlay, 2004).

In the TTSS in pathogenic Gram-negative bacteria, which secretes virulence factors through needle structures, the length of the needle is also well controlled (Kubori et al., 1998; 2000; Tamano et al., 2000). Evidence supporting the ruler model invokes a role for YscP, a FliK homologue in Yersinia, in needle length control. Lengthening or shortening the central part of YscP by inserting or deleting amino acids (aa), respectively, causes concomitant changes in needle length (Journet et al., 2003). Whether flagellar hook length can be varied according to the length of FliK has not been tested. Furthermore, the fact that FliK is secreted during flagellar assemble, without serving any obvious extracellular function, would appear to support the ruler model.

Although genetic studies suggest that FliK consists of three domains, an N-terminal secretion domain, a proline-rich central domain, and a C-terminal domain (Williams et al., 1996), gel filtration and analytical ultracentrifugation suggest there are only two (N-terminal and C-terminal) structural domains (Minamino et al., 1999; 2004). The N-terminal domain includes the FliK secretion signal, while the C-terminal domain promotes the substrate-specificity switch in the secretion apparatus. Mutations in the FliK C-terminal domain diminished switching of the secretion modes from FlgE to flagellin (Williams et al., 1996; Minamino et al., 1999). A series of mutant FliK variants with sequential in-frame 10-amino-acid deletions in the first 100 aa showed that three 10 aa segments (Δ5, Δ6 and Δ10) are dispensable (Minamino et al., 1999). Deletion of the other 10 aa segments (Δ1–Δ4, Δ7–Δ9) blocked secretion.

Why does FliK have to be secreted through an N-terminal secretion signal? In this article, we have constructed FliK variants: N-terminally truncated FliK and FliK (either N- or C-terminally) fused with fluorescent proteins (XFP); we used these variants to examine the role of FliK N-terminal domain in the FliK-dependent secretion specificity switch.


N-terminally (99 aa) truncated FliK can complement the fliK mutants

The FliK N-terminal structural domain is 200 aa long (Minamino et al., 2004). We have constructed a set of FliK variants, with deletions of amino acids in the N-terminal domain. This set of FliK deletion variants includes N-terminal truncation of the first 13 aa (Δ13) and the first 19 aa (Δ19) and then further 20-aa segments deletions up to aa 199 (Δ39, Δ59, Δ79, Δ99, Δ119, Δ139, Δ159, Δ179 and Δ199). A second set of FliK fusion variants was also constructed with fluorescent protein reporters fused to either the N- or C-terminus of FliK (see later sections).

The fliK deletions were examined for their ability to complement a fliK null mutation (SJW108). On soft agar plates, cells producing FliK with N-terminal truncations smaller than 100 aa (Δ13–Δ99) showed swarm rings of various sizes. FliKΔ39 or FliKΔ99 caused production of small swarm rings in the absence of IPTG (no induction of fliK expression) (Fig. 1A), and swarm sizes of 30% and 60% of wild-type, respectively, in the presence of IPTG (Fig. 1B). FliKΔ13 and Δ59 almost completely failed to restore swarm ring formations in the absence of IPTG (Fig. 1A) and allowed only the formation of small rings in the presence of IPTG (Fig. 1B). Cells producing FliKΔ19 and Δ79 gave very small but still discernible swarm rings in the presence of IPTG (Fig. 1A) but almost none in the absence of IPTG (Fig. 1B). On the other hand, cells producing FliK with a large N-terminal truncation of more than 100 aa (Δ119–Δ199) did not swarm at all even in the presence of IPTG (Fig. 1B), indicating that complementation did not occur. However, some of these cells were motile when observed by microscopy. This discrepancy between the two observation methods (soft agar plates vs. microscopy) can be explained by the fact that cells with polyhook-filaments reveal ‘erratic swimming behaviour’ because of a defect in the formation of flagellar filament bundles (Williams et al., 1996; also see Discussion).

Figure 1.

Complementation in a fliK mutant strain, SJW108, by plasmids expressing fliK with 5′-truncations was performed on soft agar (motility) plates. Alleles of deleted fliK were inserted into pTrc99AFF4 and introduced into SJW108. Motility plates were inoculated with cells (A) in the absence or (B) in the presence of 0.1 mM IPTG and incubated for 6 h at 30°C. Vector only (pTrc99AFF4) controls in wild-type strain SJW1103 (upper left spots) and SJW108 (upper centre spots) are included. The wild-type allele on the pTrc99AFF4 vector was also included as a control (upper right spSots). See Table 1 for plasmid descriptions.

Swimming behaviour of cells with polyhook-filaments

We have re-examined the swimming behaviour of all cells producing FliK deletion variants. Cells were taken from spots on the agar, suspended in a drop of the medium on a glass slide, and observed by dark-field microscopy. Cells producing FliK with a large truncation (Δ119–Δ199) did not swim at all, whereas cells producing FliK with a smaller truncation (Δ13–Δ99) swam sluggishly.

This swimming phenotype correlated with the amounts of flagellins secreted into media; cells producing FliKΔ13–Δ99 secreted flagellins, whereas cells producing FliKΔ119–Δ199 did not (Fig. 2). The thick band at 52 kDa in Fig. 2A is secreted flagellin (FliC). The amount of secreted FliC was proportional to that in the cells as detected by immunoblotting using the antiflagellin antiserum (Fig. 2E). The amounts of FliK variant proteins in the cells were also examined by immunoblotting using the anti-FliK antiserum (Fig 2B and D). Under inducing conditions, all FliK variant proteins were detected to some extent, but the amounts did not correlate with the amounts of flagellins (Fig 2D and E).

Figure 2.

SDS-PAGE of secreted proteins recovered from culture supernatant fractions of SJW108 cells transformed with the fliK deletion plasmids. Proteins secreted from cells transformed by the vector only (lane 1) and by wild-type fliK plasmid (lane 2) are shown as controls. Supernatant fractions from cells transformed by fliK deletion plasmids (lane 3–15) were examined in the absence (A) and presence (C) of IPTG. The flagellins in the cells were detected by immunoblotting using the antiflagellin antiserum (E). Thick bands at 52 kDa are flagellin protein (FliC), while the bands at 42 kDa are a mixture of hook protein (FlgE) and a virulence factor (SipC). The other bands are all either flagellar proteins or effector proteins. Truncated FliK variants were detected by immunoblotting using the anti-FliK antiserum in the absence (B) and presence (D) of IPTG. Gels (15%) were stained by Coomassie blue. See Table 1 for plasmid descriptions.

The thin band at 42 kDa in Fig. 2A is a mixture of the hook protein (FlgE) and SipC, a secreted TTSS effector protein (Komoriya et al., 1999). The actual amounts of FlgE detected by immunoblotting using the anti-FlgE antiserum were the same in all but wild type cells (data not shown). The amounts of flagellins secreted from cells producing FliKΔ19 were less than the others, in agreement with the swarm sizes on agar plates. We have not yet elucidated why these deletions with intermediate lengths showed such inconsistent results.

FliK with a large truncation (Δ119–Δ199) did not retain any FliK activity, suggesting that this region may be a part of the FliK C-terminal domain. We have further examined cells producing either FliKΔ13 or FliKΔ99. FliKΔ13 lacks the minimum secretion signal, whereas FliKΔ99 lacks half of the N-terminal domain. Microscopy revealed that cells producing FliKΔ13 swam at a rate of 20% of wild-type in the absence of IPTG and at 80% of the wild-type motility rate in the presence of IPTG. Cells producing FliKΔ99 were all (100%) motile both in the presence and absence of IPTG.

N-terminally truncated FliK are not secreted into media

The amounts of FliKΔ13 and FliKΔ99 produced after induction by IPTG and their secretion into the culture media were re-examined by immunoblotting (Fig. 3). To detect trace amounts of proteins, we used a detection kit with the highest sensitivity available (ECL plus). Under inducing conditions, both FliKΔ13 and FliKΔ99 were produced at readily detectable levels, but they were barely detectable in culture media (Fig. 3A). We could not detect FliK variant proteins under inducing conditions, even when the supernatant fractions were concentrated 100 times (see Experimental procedures). FliK itself was found in the supernatant fraction (Fig. 3A) while a motor component FliG remained associated with the cell fraction only (Fig. 3B). Both FliKΔ13 and FliKΔ99 lack the secretion signal so their non-secretion is in line with previous observations (Minamino et al., 1999). The possibility that hook length measured by a single FliK molecule will be discussed later (see Discussion).

Figure 3.

immuno-detection of FliK variants in the cell and supernatant fractions of strain SJW108 (fliK null) transformed by plasmids expressing of fliKΔ13 (pAH12) or fliKΔ99 (pAH17) in the presence of IPTG. Proteins in the supernatant fractions were concentrated 100 times by TCA precipitation. A flagellar structural component, FliG, was identified in the cell fraction as control (lower panel).

FliK-fluorescent protein chimeras also can complement the fliK mutants without being secreted

A second set of FliK variants were constructed so that the FliK N-terminus was fused with the cyan fluorescent protein (CFP). As a control, CFP-fused to the FliK C-terminus (FliK-CFP) was also constructed. Because of the construction strategy, FliK-CFP fusion protein has one extra alanine residue at the fusion joint-point (645 aa), while the N-terminally CFP-fused FliK (CFP-FliK fusion) protein had no additional residues (644 aa). Secondary structures of the fusion protein predicted using ExPASy tools suggested that the original structure of FliK would not be altered in either chimera (data not shown).

The functionality of the FliK chimeras was determined in a fliK mutant (SJW108) as mentioned above for FliK deletion variants (Fig. 4). Cells producing the CFP-FliK chimera produced a small swarm in the absence of IPTG and a large swarm in the presence of IPTG, showing the positive multicopy effect. The level of complementation of CFP-FliK under inducing conditions was as good as that of intact FliK judging from the swarm size (Fig. 4, right middle). On the other hand, the FliK-CFP chimera was non-functional, even in the presence of IPTG (Fig. 4, right bottom). Both proteins were produced at comparable levels (Fig. 5). These results indicate that the FliK C-terminal domain is not functional when masked by CFP.

Figure 4.

Complementation tests in strain SJW108 (fliK null) transformed with CFP-fliK fusion plasmids on motility plates. Alleles of CFP fusions to fliK were inserted into pTrc99AFF4 and introduced into SJW108 (see Experimental procedures). Motility plates with (A) or without (B) IPTG were incubated at 30°C for 6 h. The vector pTrc99AFF4 was introduced into SJW108 (upper left spots) and SJW1103 (upper right spots) as controls. Other plasmids carry fliK (pAH11), CFP (pAH2), CFP-fliK (pAH5) and fliK-CFP (pAH8).

Figure 5.

immuno-detection of CFP/FliK fusion proteins in the cell and supernatant fractions of strain SJW108 (fliK null). Proteins recovered from cells transformed by the vector only (pTrc99AFF4) (lane 1), wild-type fliK plasmid (pAH11) (lane 2), and CFP only (pAH2) (lane 3) are shown as controls. Cell fractions and supernatant fractions from cells transformed by the CFP-fliK fusion (pAH5) (lane 4), and the fliK-CFP fusion (pAH8) (lane 5) were examined (A) in the absence and (B) in the presence of IPTG. A flagellar structural component, FliG, was identified in the cell fraction as control (lower panel).

Cells producing CFP-FliK exhibited erratic or sluggish swimming phenotype in the absence of IPTG and smooth swimming in the presence of IPTG. Cells producing FliK-CFP failed to swim even in the presence of IPTG. In accordance with microscopic observations, flagellar proteins were secreted into the media from cells producing the CFP-FliK but not from cells producing the FliK-CFP (Fig. 6A). As well as in Fig. 2, the thick bands at 52 kDa are flagellin (FliC), and other minor bands are either flagellar proteins or secreted effectors.

Figure 6.

A. SDS-PAGE of secreted proteins recovered from culture supernatant fractions of SJW108 cells transformed with plasmid harbouring fliK/CFP fusion constructs. Proteins secreted from cells transformed by the vector only (pTrc99AFF4) (lane 1), wild-type fliK plasmid (pAH11) (lane 2), and CFP only (pAH2) (lane 3) were shown as controls. Supernatant fractions from cells expressing the CFP-fliK fusion (pAH5) (lane 4), and the fliK-CFP fusion (pAH8) (lane 5) were examined in the absence and presence of IPTG. Thick bands at 52 kDa are flagellin protein (FliC). The gel (12.5%) was stained with Coomassie blue.
B. The flagellar basal protein FliF in the cells as described in (A) was detected by immunoblotting using the monoclonal anti-FliF antibody. The bands at 65 kDa are FliF.

Cells producing the FliK-CFP chimeras had few polyhooks (which resembled those of the parent strain SJW108) but no flagellar filaments (see below). Electron microscopy showed that the number of the basal structures counted on osmotically shocked cells was nearly zero, indicating that this chimera might prevent flagellar assembly altogether. The amount of FliF, as determined by immunoblotting, was the same in all cells (Fig. 6B), indicating that FliK-CFP chimera did not affect the expression of the flagellar early genes. The aberrant behaviour of the FliK-CFP chimera will be discussed later (see Discussion).

Intact FliK is secreted into media during flagellar assembly (Minamino et al., 1999; Fig. 5). CFP/FliK (both CFP-FliK and FliK-CFP) chimeras were detected in the cell fraction but not in the supernatant fraction even under inducing conditions (as detected by SDS-PAGE and immunoblotting), indicating that they are not secreted. The native FliK appears as a band at 42 kDa. The band appearing at around 68 kDa corresponds to the CFP/FliK chimera. As observed in cells producing FliK deletion variants (Fig 2B and D, and Fig. 3), some CFP/FliK chimeras were degraded (Fig. 5B) but the degradation did not give rise to any secreted FliK derivatives, though we rule out the leakage of a few undetectable molecules (see Discussion).

In summary, FliK with defects in N-terminal region were not secreted, but these non-secreted FliK variants still retain FliK function (ii), i.e. the ability to switch to secretion of late secretion substrates after hook completion, such that hook elongation was replaced by filament growth. In order to determine whether these N-terminally defective FliK constructs retain FliK function (i), i.e. the ability to measure hook length, hook structures were observed by electron microscopy.

Flagellar structure of cells producing the N-terminally defective FliK

Cells producing FliK deletion variants swim erratically because they have fewer flagellar filaments. Electron microscopy showed them to retain a mixture of polyhooks alone and polyhook-filaments (Fig. 7). Excluding the flagella that were lost by shearing, we analysed whole flagella with the basal body attached as described (Aizawa et al., 1985). Cells producing FliK Δ99 in the absence of IPTG had a mixture of polyhooks alone and polyhook-filaments at ratio of 1:1 (actual numbers were 25:28) in exponential growth phase (Fig. 7A), but in stationary phase cells most of them were polyhook-filaments (2:26). In the presence of IPTG, the ratio of polyhook-filaments increased up to 80% in cells exponentially growing (Fig. 7B). Because hooks produced in the presence of IPTG appeared shorter than those produced in the absence of IPTG (Fig. 7, lower panels), we need to quantify lengths and we converted hooks to their straight form for accurate hook-length measurements (Hirano et al., 1994). Indeed, histograms showed that the average length of polyhooks produced in the presence of IPTG was shorter than that in the absence of IPTG (Fig. 8).

Figure 7.

Electron microscopic images of flagella isolated from SJW108 transformed by the plasmid expressing of fliKΔ99 (pAH17) (A) in the absence and (B) in the presence of 0.1 mM IPTG. Samples were stained with 2% phosphotungstic acid (pH 7.0). Hook-basal bodies produced under each condition were enlarged in lower panels. Bars at the lower corners represent 200 nm.

Figure 8.

The hook length of polyhook-filaments produced under various conditions. The hook-length distribution was detected for cells producing FliKΔ13 (pAH12) (A), FliKΔ99 (pAH17) (B) or CFP-FliK (pAH5) (C) in the absence (left panels) and presence (right panels) of 0.1 mM IPTG.

The hooks of the polyhook-filaments produced by the cells with FliK Δ99 varied widely in the absence of IPTG (Fig. 8B, left). In the presence of IPTG, FliK Δ99 gave rise to hooks as short as those with FliK (Fig. 8B, right). It should be noted that the FliK Δ99 molecule is the shortest FliK variant of 306 amino-acid long (vs. 405 aa of intact FliK) that still retains the function to control, albeit inefficiently, the hook length.

Cells producing CFP-FliK had a mixture of polyhooks and polyhook-filaments in a similar way to those producing FliK deletion variants. The ratio of polyhook-filaments increased under inducing conditions (data not shown). The range of the hook lengths in the cells overproducing CFP-FliK was similar to that found in wild-type hooks (Fig. 8C, right). On the other hand, FliK-CFP fusion protein interfered with flagellar assembly resulting in a reduced number of flagella (see Discussion).

Quantitative analysis of electron micrographs and histograms of the hook length (Fig. 8) revealed the following:

  • (i) The ratio between polyhooks alone and polyhook-filaments in exponentially growing cells was nearly 1:1.

  • (ii) Filaments grow from polyhooks of any length (Fig. 7A, lower panels).

  • (iii) Most of the polyhooks produced under inducing conditions were shorter than 250 nm (Fig. 7B and right panels in Fig. 8B; also see Discussion).

Observation (i) suggests that polyhooks grow filaments before cell division, because the same number of new hooks has to be added after each cell division to keep the flagellar number constant (Aizawa and Kubori 1998). In stationary phase cells, which no longer produce new hooks, most of the polyhooks have filaments attached. Observation (ii) indicates that filament growth from the hook is a stochastic process; in other words, the timing of filament growth does not depend on the existing hook length. Finally, observation (iii) suggests that the hook can grow up to 250 nm within one generation (see Discussion). These observations tell us that filament growth can occur from any existing polyhooks within one generation. Furthermore, when cells producing an N-terminally defective FliK variant were induced by IPTG, the polyhook length became shorter, close to the wild-type hook length. This indicates that both FliKΔ99 protein and CFP-FliK chimera can be as effective as intact FliK in regulating the secretion substrate specificity if overexpressed (Fig. 7B).

FliK molecules are homogeneously distributed in the cytoplasm

FliK must be localized at the flagellar base, albeit temporarily, if FliK works as a measuring ruler (Minamino et al., 2004). Localization of CFP/FliK fusion proteins was observed by fluorescent microscopy (Fig. 9). As a control, we chose FliG, one of the components of the C ring, fused to YFP. CFP/FliK and YFP-FliG chimeras were coproduced in SJW108 (fliK) cells. Because the peak emission wavelength of YFP and CFP are 527 and 475 nm, respectively, both CFP- and YFP-chimeras were easily distinguishable in the same cell. Cells producing CFP-FliK swam but those producing FliK-CFP did not as described above. In separate experiments, we confirmed that YFP-FliG was functional; fliG-deleted mutant swam when YFP-FliG was produced (data not shown; see also Yorimitsu et al., 1999). Stationary phase cells became abnormally long compared to mid-exponential cells. However, the long cells were still able to swim and did not show any indication of cell lysis. Because fluorescence images of short cells and long cells were similar, we analysed the long cells that are clearer than short cells to determine the location of fluorescent proteins along the cell body.

Figure 9.

Fluorescent microscopic images of CFP-FliK chimeras in SJW108 (fliK null) cells. SJW108 was transformed with the pTrc99AFF4-based (AmpR) plasmid harbouring either N- or C-terminally cfp-fused fliK alleles. The cells were also cotransformed with pACTrc-based (CmR) plasmid harbouring the N-terminally YFP-fused fliG allele (YFP-fliG). Fluorescence of CFP and YFP was observed using a fluorescent microscope (Olympus). Cells containing CFP-FliK and YFP-FliG (left panels) and those with FliK-CFP and YFP-FliG (right panels) were observed through either a CFP filter (Upper row) or a YFP filter (middle row), and the two images were merged for comparison (lower row).

The distribution of CFP/FliK (both CFP-FliK and FliK-CFP) chimeras in the cells appeared more homogeneously (Fig. 9, upper row) than that of YFP-FliG, which localized to several strong spots (presumably) at the flagellar base (Fig. 9, middle row). Careful inspection of the images revealed what appeared to be very faint spots of CFP/FliK but these spots overlapped the YFG-FliG spots, as shown in the merged image (Fig. 9, lower row), indicating that they were derived from the fluorescence of YFP-FliG leaking through the emission filter of CFP. Because the FliK-CFP, which retains the original N-terminal signal, did not localize at the flagellar base, intact FliK might also remain randomly distributed through in the cytoplasm for most of the time.

Conclusion: FliK can control the hook length prior to its secretion

The experiments described in this article are summarized in Fig. 10. The two sets of FliK variants have various molecular sizes different from that of intact FliK (405 aa length): from 306 aa (FliK Δ99) to 643 aa (CFP-FliK). Despite of the difference of molecular sizes (as long as the C-terminal domain remains intact), all FliK variants gave rise to the same results; when they were abundant in the cell, the cells could produce hooks of nearly normal length. In other words, the size of FliK N-terminal region does not correlate with the hook length, but the copy number of intact FliK C-terminal domain is important in the hook length control in the absence of the N-terminal region. We suggest that FliK secretion might be necessary for the efficient action of the C-terminal domain on the secretion apparatus. In this case, the N-terminal domain might be important only when FliK is limited in the cell and poorly accessible to the flagellar base.

Figure 10.

Schematic presentation of FliK variants used in the experiments and results of complementation tests and FliK secretion. Numbers at ends of bars indicate the amino acid numbers based on intact FliK as the standard. Empty bars indicate FliK, and shaded bars are CFP. Swarming ability on motility agar plates was measured by diameters of swarm rings. Swimming ability was examined by microscopy. FliK secretion was estimated from relative band density in Figs 3 and 6. The number of plus (+) indicates either the swarm size or the degree of complementation.


In this report, we show that FliK variants can control the hook length as long as the C-terminal domain remains intact, independently from the molecular sizes of the N-terminal region and whether or not the protein is secreted. All FliK variants altered in the N-terminal region were barely secreted yet allowed the production of hooks of nearly the wild-type length when they were overproduced, leading us to conclude that FliK N-terminal domain is dispensable for hook length control. It is not possible to conclude that all of these variants give rise to polyhooks that are of exactly normal length, because the normal hook length varies over a wide range, with a peak at 55 nm (Hirano et al., 1994) and hooks of any length within this range are considered as normal.

Size of swarm rings does not correlate with swimming ability of the cell

To our surprise, cells producing N-terminally defective FliK variants were able to swim in liquid, despite their inability to swarm. In order to understand this discrepancy, the flagellar structures were observed by electron microscopy. Cells producing these FliK variants had polyhook-filaments of variable length. We assume that the hook length of the polyhook-filaments was too long to produce enough thrust to propel cells in a viscous medium (motility agar). Higher level production of the FliK variants resulted in a shortening the polyhook length to a peak at less than 100 nm (Fig. 8, right panels). Because, in FliKΔ99-producing cells, the hook length but not the number of flagella per cell changed according to the level of induction, we attribute differences in the swarm sizes of these mutants to the differences in hook length.

Are CFP-FliK fusion proteins cleaved?

We fused FliK to various fluorescent proteins, expecting that the fusion might block FliK secretion. Three commercially available fluorescent proteins (GFP, CFP and YFP) were fused to FliK (see Experimental procedures). Because all combinations of fluorescent protein fusions to FliK yielded similar results in complementation tests, the flagellar number and protein secretion (data not shown), we are only reporting the results from the CFP/FliK fusion proteins.

A problem with CFP fusion proteins often pointed out is the cleavage of CFP from the fusion proteins, hence allowing the cleaved but nearly intact proteins to work independently. CFP-FliK seems not to be cleaved, judging from the results of immunoblotting with anti-FliK antiserum (Fig. 5). On the other hand, the FliK-CFP fusion was degraded but did not complement the fliK null mutant, ruling out the possibility that the degradation products are functional. N-terminally truncated FliK variants showed the positive multicopy effect on swarming (Fig. 4, right middle), whereas intact FliK is known to show the negative multicopy effect so that the swarm size becomes smaller (Fig. 1, right top). Thus, if the CFP-FliK fusion were cleaved and the cleaved fragment were functional, this fragment would show the negative multicopy effect on swarming, the reverse of what we observed.

CFP-fliK fusion protein works in a similar way as fliK deletion variants

CFP-fliK complemented a fliK mutation in a similar way to the truncated fliK variants: i.e. it partially complemented at low levels and fully complemented when overexpressed. Quantitative analysis of polyhook-filaments from cells producing CFP-FliK revealed that more than 90% of the polyhooks were shorter than 250 nm (Fig. 8), which is still much longer than the average length of the wild-type hook (55 nm). Strictly speaking, the length was between 200 and 300 nm, and we chose 250 nm for the calculation below. What does a 250 nm polyhook mean? We calculated the time the hook elongates up to 250 nm according to Koroyasu et al. (1996). It takes 3.5 min to grow to the normal length of 55 nm, whereafter it grows at a constant rate of 8 nm min−1. Thus, it would take 28 min (3.5 min + 24.5 min) for the hook to reach 250 nm, i.e. one generation. In cells producing CFP-FliK, the filaments start to grow at random intervals but within one generation, thus preventing the hook's elongating more than 250 nm.

Interestingly, cells producing the FliK-CFP chimera had few polyhooks, though it did not affect the expression of the flagellar early genes as confirmed by measurement of FliF levels. Then how does the FliK-CFP chimera stop flagellar assembly? We wonder if the active N-terminal region might actually enter into the export apparatus and stick at the export gate to prevent other flagellar proteins from being secreted. In agreement with this assumption, overproduced FliK-CFP chimeras also interfered with flagellar assembly in wild-type SJW1103 cells, resulting in a reduction in the flagellar number (data not shown).

Why is intact FliK secreted?

N-terminally truncated FliK variants were not as efficient as intact FliK at limiting hook growth to 55 nm but, according to the arguments we present above, would eventually access all of the flagellar bases within one generation. As the copy number increases, the probability that partially functional FliK can access the secretion apparatus becomes higher, consequently leading to the switching of secretion modes at an earlier time. The FliK N-terminal region might be designed to facilitate its interaction with secretion apparatus so that the C-terminal domain can access FlhB, a part of the secretion apparatus, more easily.

Another possibility is that the N-terminal region affects flagellar gene regulation. Overproduction of FliK gives rise to shorter hooks (Muramoto et al., 1998), suggesting two possible mechanisms: (i) earlier switching of the secretion modes, or (ii) the down-regulation of FlgE synthesis. Bonifield et al. (2000) reported a threefold increase in FlgE production in a fliK mutant and hypothesized post-transcriptional regulation of FlgE in response to the stage of flagellar assembly. In Campylobacter jejuni, FlgE was overexpressed 60 times in the fliK deletion mutant (C.W. Penn, personal communication). These data paradoxically support mechanism (ii). FliK could be secreted to reduce its own copy number in the cytoplasm, in a manner analogous to the secretion of FlgM to release sigma factor 28 (Hughes et al., 1993). We conclude that the FliK molecule does not measure hook length itself, but might control amounts of the hook protein to be secreted from inside the cell. The study on the role of FliK as a regulator of flagellar gene expression is currently under way.

How many FliK molecules are necessary for the hook length control?

The N-terminally defective FliK variants we constructed could not be detected in the supernatants, indicating that they were not secreted. The level of FliK is normally low; 40–80 molecules per cell, that is, 4–8 molecules per flagellum (Muramoto et al., 1998). How many FliK molecules are necessary for the hook length control?

In an extreme case would be one molecule, that goes into the channel, measures the hook length (in line with the ruler model), and then induces the cleavage of FlhB to switch the secretion mode. But, how does this molecule know that the hook length is correct before it really interacts with the growing hook? If the molecule docks too early or too late, the length will be too short or too long respectively. In the measuring cup hypothesis, it is also difficult for a single FliK molecule to detect when the cup no longer contains FlgE. Although there is no way to measure a small number of molecules at the moment, measuring how many FliK molecules are involved is crucial for determining which, if either, of the two hypotheses is correct.

Experimental procedures

Strains and growth conditions

All strains and plasmids used in this study are listed in Table 1. SJW 108, the fliK null mutant, lacks the 388th base (A→C) in its sequence, resulting in A131R and 132stop in the amino acid sequence (Williams et al., 1996). Strains were grown and maintained in Luria–Bertani (LB) broth and incubated at 37°C. When necessary, media were supplemented with Ampicillin (50 µg ml−1), Chloramphenicol (25 µg ml−1), or IPTG (0.1 mM).

Table 1.  Strains and plasmids used in this study.
Strain and plasmidRelevant characteristicsSource or reference
E. coli strain
 NovaBlueGeneral cloning strainNovagen
Salmonella strains
 SJW1103Wild-type strain for motility and chemotaxisYamaguchi et al. (1984)
 SJW108fliKWilliams et al. (1996)
 pTrc99AFF4Cloning vectorOhnishi et al. (1997)
 pblue-sGFP(S65T)-nos3’ KSsGFP(S65T) allele on pBluescript KSChiu et al. (1996)
 pECFP BD Biosciences
 pEYFP BD Biosciences
 pAH1pTrc99AFF4-sGFP(S65T)This study
 pAH2pTrc99AFF4-eCFPThis study
 pAH3pTrc99AFF4-e-YFPThis study
 pAH4pTrc99AFF4-sGFP(S65T)-fliKThis study
 pAH5pTrc99AFF4-eCFP-fliKThis study
 pAH6pTrc99AFF4-eYFP-fliKThis study
 pAH7pTrc99AFF4-fliK-sGFP(S65T)This study
 pAH8pTrc99AFF4-fliK-eCFPThis study
 pAH9pTrc99AFF4-fliK-eYFPThis study
 pAH10pTrc99AFF4-eYFP-fliGThis study
 pAH11pTrc99AFF4-fliKThis study
 pAH12pTrc99AFF4-fliKΔ13This study
 pAH13pTrc99AFF4-fliKΔ19This study
 pAH14pTrc99AFF4-fliKΔ39This study
 pAH15pTrc99AFF4-fliKΔ59This study
 pAH16pTrc99AFF4-fliKΔ79This study
 pAH17pTrc99AFF4-fliKΔ99This study
 pAH18pTrc99AFF4-fliKΔ119This study
 pAH19pTrc99AFF4-fliKΔ139This study
 pAH20pTrc99AFF4-fliKΔ159This study
 pAH21pTrc99AFF4-fliKΔ179This study
 pAH22pTrc99AFF4-fliKΔ199This study
 pAH23pACTrc-YFP-fliGThis study

Construction of FliK variants

All synthetic primers were synthesized by Espec oligo service (Japan) as a PCR grade. The polymerase chain reaction (PCR) was carried out using the GeneAmp PCR system 9700 (ABI) and Pfu DNA polymerase (Stratagene). PCR products were purified with the QIAquick PCR purification kit (Qiagen). Restriction enzymes were obtained from New England Biolabs. T4 DNA ligase was purchased from TAKARA. The plasmid constructs were purified using the QIAprep spin miniprep kit (Qiagen).

In the course of constructing FliK N-terminally truncated variants, reverse primers were designed to begin each deletion construct with a methionine translational start codon, and we introduced a NdeI restriction enzyme site for cloning purposes. In all cases, iKBamFd-1 primer (5′-gaa tcc ctg gcg tta ggc gaa gat atc-3′) was used as a forward primer. The reverse primer for specific N-terminal truncations were as follows: FliKΔ13 (iKNdeRv-13 = 5′-caccgatacccatatgaccgcgggtc-3′), FliKΔ19 (iKNdeRv-19 = 5′-gaccgcgggtcatatgtcaggaaaaa-3), FliKΔ39 (iKNdeRv-39 = 5′-ggcgggcgcgcatatggcagacggcg-3′), FliKΔ59 (iKNdeRv-59 = 5′-tttacaggcgcatatgggcaagttat-3′), FliKΔ79 (iKNdeRv-79 = 5′-ccaggcggtgcatatggccgacctgc-3′), FliKΔ99 (iKNdeRv-99 = 5′-taccgatctgcatatggcgcagcatc-3′), FliKΔ119 (iKNdeRv-119 = 5′-cgctctggcccatatgagtaagacgg-3′), FliKΔ139 (iKNdeRv-139 = 5′-tgaggatcttcatatgctgagcgcct-3′), FliKΔ159 (iKNdeRv-159 = 5′-gcctgtcgcccatatgacgccggctg-3′), FliKΔ179 (iKNdeRv-179 = 5′-cgacatgccacatatgccgcagga ag-3′) and FliKΔ199 (iKNdeRv-199 = 5′-agggaaaacccatatgtc gcttgcgc-3′). PCR-amplified fragments were digested with NdeI and BamHI and then ligated into pTrc99AFF4 vector to make plasmids from pAH11 to pAH22.

To construct FliK fusion proteins, we employed a modified version of GFP (GFP/S65T), which has been known more stable than commercially available GFP (Heim et al., 1995; Chiu et al., 1996). peCFP and peYFP plasmids harbouring alleles encoding enhanced cyan and yellow fluorescent proteins, respectively, were purchased from BD Biosciences. GFP, eCFP and eYFP alleles were PCR-amplified using two flanking primer GFPNdeRv (5′-tct aga gga cat atg gtg agc aag-3′) and GFPBamFd (5′-gca gcc cgg ggA tcc gct tta ctt gta-3′). These fragments were inserted into the pTrc99AFF4 vector between the NdeI and BamHI sites to give pAH1, pAH2 and pAH3 plasmids respectively. The fliK allele was PCR-amplified using two flanking primers, iKBsrRv (5′-gtg tac aag atg atc acc ctg ccc-3′) and iKBamFd-1 (see above). This fragment was inserted into pAH1, pAH2 and pAH3 plasmids between the BsrGI and BamHI sites to make pAH4, pAH5 and pAH6 plasmids for the construction of GFP-fliK, eCFP-fliK and eYFP-fliK in-frame fusion alleles respectively. The fliG allele was PCR-amplified using two flanking primers, iGBsrRv (5′-agt gga tga ctg tac aag atg agt aat ctt-3′) and iGBamFd (5′-cag act tgc gga tcc aat tca tta gac-3′). A BsrGI-BamHI fragment containing the fliG allele was inserted into the BsrGI-BamHI site of pAH3 to give pAH10. For construction of fliK-GFP, fliK-eCFP and fliK-eYFP in frame fusion alleles, fliK allele was PCR-ed using two flanking primers, iKNdeRv (5′-gag gaa acc cat atg atc acc ctg ccc-3′) and iKNcoFd (5′-ctc cat ggc ggc gaa gat atc cac-3′). GFP, eCFP and eYFP alleles were PCR-amplified using two flanking primers, GFPNcoRv (5′-aga gga tcc atg gtg agc aag-3′) and GFPBamFd (5′-gca gcc cgg gga tcc gct tta ctt gta-3′). Three fragments, e.g. pTrc99AFF4 (NdeI–BamHI fragment), fliK allele (NdeI–NcoI fragment) and GFP allele (NcoI-BamHI fragment), were ligated together to construct a pAH7 derivative harbouring a fliK-GFP in-frame fusion allele. The same procedures were used for construction of pAH8 and pAH9 harbouring fliK-eCFP and fliK-eYFP respectively. The sequence of all PCR-amplified fragments was verified by DNA sequence analysis of the final cloned products.

Complementation tests

Complementation tests were carried out on motility (soft) agar plates (1% tryptone, 0.7% NaCl and 0.35% agar). On motility agar plates, non-motile mutant cells form small dots whereas the wild type or complemented mutant strains form spreading rings or swarms. Motility was assayed by inoculation of strains onto motility agar plates with or without 0.1 mM IPTG followed by incubation at 30°C for 6 h.

Alleles for eCFP fusions to fliK were inserted into pTrc99AFF4, and the resulting plasmids were introduced into SJW108 (fliK null strain). The resulting strains were assayed for motility in the presence or absence of 0.1 mM IPTG. The vector pTrc99AFF4, wild-type fliK, and eCFP plasmids were also included as controls.


Proteins secreted into the media were analysed by SDS-PAGE. Briefly, the cell pellet and the culture supernatant were fractionated by differential centrifugation. The cells were suspended into SDS-sample buffer and the proteins secreted into the culture supernatant were precipitated with 7% TCA. Proteins were dissolved in SDS-sample buffer and boiled for 3 min Both fractions were subjected to SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) membranes. Amounts of protein sample applied onto SDS gels were normalized to the cell number, which was measured at OD600.

SDS-PAGE was carried out using a mini-gel kit (ATTO). Acrylamide concentration of the gels was 10%, 12.5%, or 15% depending on the molecular range of the proteins of interest in each experiment. Gels were stained using Coomassie brilliant blue. Molecular weight protein markers were obtained from New England Biolabs.

Immunoblotting analysis

Overnight cultures of SJW108(fliK) harbouring various pTrc99AFF4-based plasmids were inoculated into 5 ml of LB media at a 1/100 dilution. After incubation at 37°C for 5 h, the culture was centrifuged at 14 000× g and separated into a supernatant fraction and a pellet fraction. Alternatively, especially for the samples grown in the presence of 0.1 mM IPTG, the culture was incubated at 37°C for 3 h. Upon the addition of 0.1 mM of IPTG, the culture was further incubated for 2 h. The samples were subjected to SDS-PAGE and electro-blotted to PVDF membranes. Either the polyclonal anti-FliK antibody (Ohnishi et al., 1997) or monoclonal anti-FliF antibody (Homma et al., 1987) was used for immunoblotting. ECL plus Western blotting detection kit and ECL camera used for immuno-detection were purchased from Amersham Bioscience.

Electron microscopy

Samples prepared as previously described (Aizawa et al., 1985) were negatively stained with 2% phosphotungstic acid (pH 7.0), and observed with a JEM-1200EXII electron microscope (JEOL, Tokyo). Micrographs were taken at an accelerating voltage of 80 kV.

To count the number of flagellar basal structures in a given strain, cells of interest were suspended in sucrose solution (0.5 M sucrose, 0.15 M Trizma-base) and abruptly diluted with water to create osmotically shocked cells. After removing non-lysed cells by low-speed centrifugation, osmotically shocked cells in the supernatant were collected by high-speed centrifugation.

To measure the hook length, intact flagella were isolated and the hook basal bodies were prepared according to the conventional method (Hirano et al., 1994).

Fluorescence microscopy

Various strains were inoculated into LB broth containing the appropriate antibiotics and incubated overnight at 37°C. The cells were washed once with PBS and fixed onto glass slides coated with poly lysine. Fluorescence of the CFP and YFP was observed using an epi-fluorescent microscope (IX-70, Olympus) with 100W mercury arc lamp. Cells containing CFP-FliK and YFP-FliG, and those with FliK-CFP and YFP-FliG were observed using the appropriate CFP filter or YFP filter. Fluorescent images of cells were collected using cooled CCD camera (MICROMAX 512BFT, Roper Scientific, Duluth, GA). The resulting 16 bit TIFF images were fed into a personal computer (Endeavor Pro 1000, Epson-direct, Tokyo, Japan).


We thank Kelly Hughes and David DeRosier for their critical reading of the manuscript, May Macnab for the monoclonal anti-FliF antibody, Yasuo Niwa for providing sGFP, Naoko Haga and Noriko Takahashi for their technical help, and Dr Yoshie Harada for discussion.