Localization of components of the chemotaxis machinery of Escherichia coli using fluorescent protein fusions



We prepared fusions of yellow fluorescent protein [the YFP variant of green fluorescent protein (GFP)] with the cytoplasmic chemotaxis proteins CheY, CheZ and CheA and the flagellar motor protein FliM, and studied their localization in wild-type and mutant cells of Escherichia coli. All but the CheA fusions were functional. The cytoplasmic proteins CheY, CheZ and CheA tended to cluster at the cell poles in a manner similar to that observed earlier for methyl-accepting chemotaxis proteins (MCPs), but only if MCPs were present. Co-localization of CheY and CheZ with MCPs was CheA dependent, and co-localization of CheA with MCPs was CheW dependent, as expected. Co-localization with MCPs was confirmed by immunofluorescence using an anti-MCP primary antibody. The motor protein FliM appeared as discrete spots on the sides of the cell. These were seen in wild-type cells and in a fliN mutant, but not in flhC or fliG mutants. Co-localization with flagellar structures was confirmed by immunofluorescence using an antihook primary antibody. Surprisingly, we did not observe co-localization of CheY with motors, even under conditions in which cells tumbled.


Bacterial cells have a signal transduction pathway that allows them to sense and respond to changes in the concentrations of chemical attractants or repellents. In Escherichia coli, the addition of attractant or the removal of repellent promotes counterclockwise flagellar rotation or smooth swimming, which carries cells in a favourable direction. The signal transduction pathway includes a number of membrane-bound receptors, including four methyl-accepting chemotaxis proteins (MCPs: Tsr, Tar, Trg and Tap) and an oxygen receptor (Aer), six cytoplasmic chemotaxis proteins (CheA, CheW, CheR, CheB, CheY and CheZ) and three proteins comprising a switch complex at the cytoplasmic face of the flagellar motor (FliM, FliN and FliG). For reviews and recent structural work, see Bilwes et al. (1999), Djordjevic and Stock (1998), Falke et al. (1997), Kim et al. (1999) and Levit et al. (1998). The chemotactic signal is transmitted by autophosphorylation at His-48 of CheA, phosphoryl group transfer to Asp-57 of CheY (Hess et al., 1988a,b; Borkovich et al., 1989; Sanders et al., 1989) and binding of phosphorylated CheY to FliM (Welch et al., 1993), which stabilizes the clockwise and destabilizes the counterclockwise rotational states (Kuo and Koshland, 1989; Alon et al., 1998; Scharf et al., 1998). Dephosphorylation of CheY is accelerated by CheZ (Hess et al., 1988a).

The assembly of a ternary complex comprising MCP, CheW and CheA was shown to be necessary for CheA activation in vitro (Ninfa et al., 1991; Gegner et al., 1992), where it forms higher order structures (Liu et al., 1997). The formation of a quaternary complex, including CheY, has also been shown in vitro (Schuster et al., 1993). In addition, CheZ was found to interact with the short form of the histidine kinase, CheAS (Wang and Matsumura, 1996; 1997). Immunoelectron microscopy revealed that MCP–CheW–CheA complexes are clustered in vivo, predominantly at the cell poles (Maddock and Shapiro, 1993), but weaker lateral clusters also are observed (Lybarger and Maddock, 1999; Skidmore et al., 2000). CheA and CheW are required for strong clustering, but there is a significant level of CheA-independent clustering (Skidmore et al., 2000). Receptor clustering might be important for the generation of chemotactic signals (Bray et al., 1998; Levit et al., 1998; Duke and Bray, 1999).

The switch complex has been visualized by electron microscopy (Khan et al., 1992; Francis et al., 1994), in which it appears as a cytoplasmic ring (the C-ring) about 45 nm in diameter. The primary components of this ring are FliM and FliN. The remaining component of the switch complex, FliG, appears at the periphery of the MS-ring, where it forms a bridge to the C-ring. FliG is thought to be assembled first on the MS-ring, followed by the co-operative association of FliM and FliN (Zhao et al., 1995; 1996; Kubori et al., 1997; Khan et al., 1998): preparations of flagellar basal bodies that lack FliN also usually lack FliM, and vice versa. However, physical interactions among the three switch proteins, demonstrated both in vitro (Tang et al., 1996) and in vivo (Marykwas and Berg, 1996; Marykwas et al., 1996), suggest that FliM can interact with FliG by itself. Another contradiction is the observation by Oosawa et al. (1994), also made in vitro, that FliM can bind to the MS-ring in the absence of FliG.

Green fluorescent protein (GFP) and its mutants are now widely used to study the localization of proteins in living cells. For general reviews, see Conn (1999); for applications in bacteria, see Margolin (2000). A GFP fusion has already been used to study the localization of one of the chemotaxis proteins, CheZ (M. Manson, personal communication). This construct localizes to the cell poles to what appears to be MCP clusters, in apparent contradiction to the previous result that the interaction of CheZ with CheAS is blocked by the binding of CheAS to CheW (Wang and Matsumura, 1997).

In the present work, we used fusions to YFP (the S65G, V68L, S72A, T203Y GFP mutant) to study the localization in E. coli of the chemotaxis proteins CheY, CheZ and CheA, and the motor protein FliM. The chemotaxis proteins localized in a pattern of clusters in parallel with MCPs, as identified by immunofluorescence. This suggests the existence of a chemotaxis signal complex comprising five proteins: MCPs, CheW, CheA, CheY and CheZ. FliM–YFP localized to the flagellar motors, even in the absence of FliN, but not in the absence of FliG.


YFP fusions to chemotaxis and motor proteins

In order to investigate the localization of chemotaxis and motor proteins in E. coli cells, we designed several fusions of YFP to the proteins CheY, CheZ, CheA and FliM, as summarized in Table 1. For CheY, CheZ and FliM, both C-terminal and N-terminal fusions were constructed. For CheA, two N-terminal fusions, YFP-CheAL and YFP-CheAS, were constructed, corresponding to the long and short forms of the CheA protein (Kofoid and Parkinson, 1991). For all fusions, a short amino acid linker (3× or 5× glycine) was used (see Experimental procedures). The fusion genes were expressed under control of the arabinose promoter (pBAD), which allows tight regulation of expression (Guzman et al., 1995). The existence of full-length fusion proteins was verified by immunoblot (see Experimental procedures). YFP fusions to CheY, CheZ, CheA and FliM were tested for their ability to complement corresponding null mutants for spreading in soft agar (Fig. 1; Table 1). CheY, CheZ and FliM fusions complemented the null mutants in an inducer-dependent manner, suggesting that all these fusions are functional (although not as efficient as the wild-type proteins). However, the YFP–CheAL fusion did not complement the cheA null mutant. The induction level that gave an optimal (or nearly optimal) complementation, 0.002% arabinose for FliM–YFP and 0.005% arabinose for all the other proteins, was used for further experiments. Under these conditions, the levels of expression of the fusion proteins were close to the levels of expression of the native proteins (≈ 70–180%), and the total amount of degraded protein was small, always less than 10% of the amount of native protein (see Experimental procedures). Thus, it seems unlikely that degradation products could account for any of the observed complementation effects.

Table 1. Functionality of YFP fusions.

Fusion protein

Presence of
full-length productb
  • a . Defined by complementation for spreading in soft agar ( Fig. 1) as the ratio of the size of the outer swarm ring of the complementation strain to the size of the outer swarm ring of the wild-type strain, as follows: –, < 0.15; ±, 0.15–0.3; + +, 0.4–0.5; + + +, 0.5–0.66; + + + +, > 0.66.

  • b

    . Tested by immunoblot.

  • c

    . The YFP–FliM fusion complemented the fliM mutant for swimming in liquid media but only poorly for spreading in soft agar.

CheY–YFP+ + ++
YFP–CheY+ ++
CheZ–YFP+ + + ++
YFP–CheZ+ + + ++
FliM–YFP+ + + ++
Figure 1.

Swarm plates illustrating the ability of different YFP fusion proteins to complement the corresponding null mutants. The concentration of arabinose (as a percentage), the inducer of gene expression for the pBAD constructs, is shown at the top.

A. Swarm ring formation by (1) RP437/pBAD18K wild type; (2) cheY/CheY–YFP; (3) cheY/YFP–CheY; (4) cheZ/CheZ–YFP; and (5) cheZ/YFP–CheZ.

B. Swarm ring formation by (1) RP437/pBAD18K wild type; (2) fliM/FliM–YFP; (3) fliM/YFP–FliM; (4) cheA/YFP–CheAL; and (5) cheA/YFP–CheAS. Swarm plates were supplemented with kanamycin (50 µg ml−1).

Association of YFP fusions with MCP clusters

Both CheY–YFP and YFP–CheY showed a similar pattern of localization. Two types of clusters were observed, intense polar clusters and weaker lateral clusters. Usually, in cells grown to late exponential phase, either one or two clusters were seen at the poles and none or one along the sides of a cell (Fig. 2A). This strongly resembles the clustering of MCPs observed previously by immunoelectron and immunofluorescent microscopy (Maddock and Shapiro, 1993).

Figure 2.

Localization of YFP fusions to chemotaxis proteins in different mutant backgrounds.

A. CheY–YFP/cheY.

B. CheY–YFP/trg.

C. CheY–YFP/tsr.

D. CheY–YFP/cheA.

E. CheZ–YFP/cheZ.

F. CheZ–YFP/cheA.

G. YFP–CheAL/cheA.

H. YFP–CheAS/cheA.

I. YFP–CheAS/tar tsr tap trg.

The arabinose concentration was 0.005%. Results for a larger set of mutant backgrounds are given in Table 2.

To verify that CheY–YFP localizes with the MCPs, we analysed the distribution and intensity of CheY–YFP clusters in different mcp, che and fli backgrounds (Table 2; Fig. 2A–D). CheY–YFP clusters were seen in the wild type, cheY, cheZ, tar, tsr and trg backgrounds, but not in a mutant defective in cheA, a mutant defective in cheW or a mutant defective in all four MCPs (tar tsr tap trg). This is consistent with CheY localization on MCP clusters, mediated by CheW and CheA. In tar and tsr strains, which lack either one or other of the major MCP proteins, the intensity of CheY localization at the poles was significantly reduced, and lateral clusters nearly disappeared (Table 2; Fig. 2C). In a trg strain lacking a minor MCP protein, CheY localization was similar to that in the wild type (Table 2; Fig. 2B). Phosphorylation of CheY–YFP by CheA was not required for localization, as shown by the presence of clusters in the cheAH48Q mutant (Table 2). There was no CheY–YFP localization in fliM or cheY fliM mutants, but further investigation showed that this was a result of the negative effect of the fliM deletion on the expression of chemotaxis and receptor genes (Kutsukake and Iino, 1994). In the fliM flgM background, where the gene for the anti-sigma factor FlgM responsible for the negative control was deleted, normal localization of CheY–YFP was restored (Table 2).

Table 2. Localization of YFP fusions.





tar tsr
tap trg




  1. The degree of localization was measured as the ratio of the integrated intensity of fluorescence of the localized protein to the integrated intensity of fluorescence of a region of the same size elsewhere in the same cell. More than 40 cells were measured for each fusion construct/strain combination. The standard errors of the measurements were ≤ 0.05.

  2. NL, non-localized; NI, non-localized with inclusion bodies in some cells; ND, not determined. See text for further comments.


To determine further whether we might see the localization of CheY–YFP on flagellar motors, we phosphorylated CheY–YFP in vivo under conditions that do not allow clustering of MCPs. First, we used acetate, which is known to cause CheA-independent CheY phosphorylation through the formation of acetyl phosphate (Wolfe et al., 1988; Lukat et al., 1992). The addition of acetate to the cheY cheA strain expressing CheY–YFP yielded no CheY–YFP localization (data not shown), although the cells became tumbly compared with the control, suggesting that the CheY–YFP fusion was phosphorylated and could bind to the motor. We also co-expressed CheY–YFP with a constitutively active cytoplasmic fragment of the Tsr receptor (Ames and Parkinson, 1994) in a cheY tar tsr tap trg strain. Again, the cells tumbled, but no CheY–YFP localization was seen on the motor (data not shown).

Two other chemotaxis proteins, CheZ and CheA, were expected to localize with the MCPs. Co-localization of CheZ–GFP and MCP was shown previously in the laboratory of Mike Manson (personal communication). We observed such localization for both YFP–CheZ and CheZ–YFP (Table 2; Fig. 2E). This localization was dependent on the presence of MCP, CheW and CheA, but not on the presence of CheY (Table 2; Fig. 2F). Both YFP–CheAL and YFP–CheAS localized in a pattern identical to that of CheY–YFP. This localization was dependent on the presence of MCP and CheW (Table 2; Fig. 2G-H). As the plasmid expressing YFP–CheAL also expresses unlabelled CheAS, one might expect less fluorescence in the clusters for YFP–CheAL because of competition between the two. This appears to be the case, as indicated in Table 2.

To verify the co-localization of the CheY, CheZ and CheA fusion proteins with MCP clusters, we performed additional staining using primary antibody raised against the signalling domain of Tsr, in combination with Texas red-coupled secondary antibody (Fig. 3). Cell fixation with methanol and treatment with lysozyme, necessary to make E. coli cells permeable to antibodies, led to some loss of CheY–YFP, as evidenced by a reduction in the overall level of green fluorescence. At the same time, some cells were not permeabilized enough to allow MCP staining. However, in most cells, presumably cells that were partially lysed, the remaining CheY–YFP spots co-localized precisely with the MCP clusters (Fig. 3A–C). We tried other fixatives, e.g. paraformaldehyde, formaldehyde and glutaraldehyde, but CheY–YFP localization was lost. However, there was no loss of the other fusion proteins, which remained associated with the MCP clusters (Fig. 3D–K).

Figure 3.

Association of CheY, CheZ and CheA fusion proteins with MCP clusters, demonstrated by immunofluorescence. Cells expressing YFP fusions were fixed with methanol, stained with anti-Tsr antibody and counterstained with anti-rabbit Texas red-coupled antibody, as described in Experimental procedures.

A–C. cheY cells expressing CheY–YFP.

D–F. cheZ cells expressing CheZ–YFP.

G–I. cheA cells expressing YFP–CheAL.

J–L. cheA cells expressing YFP–CheAS.

Left: YFP fluorescence (green); centre: Texas red fluorescence (red); right: superimposition of the images from the left and centre. Orange colour shows co-localization of CheY (C), CheZ (F), CheAL (I) and CheAS (L) fusion proteins with MCP clusters.

Results obtained by immunoelectron microscopy suggest that, although CheW and CheA are important for MCP clustering (Maddock and Shapiro, 1993), there is a significant level of clustering in the absence of CheA (Skidmore et al., 2000). To learn whether the mere delocalization of MCPs could account for the absence of localization of the chemotaxis fusion proteins in the cheW and cheA backgrounds, we compared MCP localization in wild-type, cheA and cheW strains (Fig. 4A–C). In the absence of CheA or CheW, MCPs were still localized to the poles (Fig. 4B and C), although not in tight clusters (Fig. 4A). Thus, it is the missing interaction with CheA (for CheY and CheZ) and with CheW (for CheA) that prevents the fusion proteins from being localized to the poles.

Figure 4.

Localization of MCPs in different mutant backgrounds, demonstrated by immunofluorescence. Wild type (A), cheA (B), cheW (C) and tar tsr tap trg (D) cells were fixed and stained with anti-Tsr antibody, as described in the legend to Fig. 3. In wild-type cells, MCPs are tightly clustered (A). In the cheA or cheW background, they remain localized at the cell poles, but more diffusely (B and C). This localization is not observed in the mcp strain.

FliM–YFP localization on flagellar motors

In addition to the localization of chemotaxis proteins on MCP clusters, we also observed the localization of FliM on flagellar motors. Although FliM–YFP was much more efficient than YFP–FliM in complementing a fliM mutant for spreading in soft agar (Fig. 1B, swarms 2 and 3), both constructs complemented the cells for flagellation and motility and showed the same pattern of localization (Table 2; Fig. 5A). This pattern showed several spots (up to 10), more or less evenly distributed along the cell body. As FliM is known to be a part of the cytoplasmic C-ring of the flagellar motor, these spots probably correspond to motors distributed at random on the surface of the cell. In some cells, a bright fluorescent spot was also observed at the end of the cell, presumably corresponding to an inclusion body. The localization of FliM–YFP was dependent on another constituent of the cytoplasmic C-ring, FliG (Fig. 5C), which is known to be assembled on the C-ring before FliM (Zhao et al., 1995). The absence of yet another C-ring component, FliN, did not abolish FliM–YFP localization (Fig. 5B), in agreement with the observation that FliM binds to FliG in the absence of FliN (Marykwas et al., 1996; Tang et al., 1996). The formation of a single protein aggregate (inclusion body) by FliM–YFP in some fliM cells, in many fliM fliG cells or in a flhC strain suggests that proper targeting to the motor reduces the free concentration of FliM and, hence, the likelihood that it forms inclusion bodies. This is consistent with the finding that, under normal growth conditions, wild-type FliM tends to form aggregates of high molecular weight (Zhao et al., 1996).

Figure 5.

Localization of FliM–YFP in different mutant backgrounds.

A. fliM.

B. fliM fliN.

C. fliM fliG.

The arabinose concentration was 0.002%.

The localization of FliM–YFP to flagellar motors was confirmed by immunofluorescence. We used a primary antibody against flagellar hooks and a secondary antibody coupled with Texas red. These experiments were carried out in a fliC strain that lacks flagellar filaments but has hooks (Fig. 6). An additional mutation, clpP, which inactivates a component of one of the major E. coli proteases, was introduced in the fliM fliC background strain (GP90). This reduces the level of FliM–YFP degradation to < 2% without interfering with FliM–YFP function. As shown in Fig. 6C, green FliM–YFP spots (Fig. 6A) co-localized with red hook spots (Fig. 6B). However, not all of the green FliM–YFP spots had a corresponding red hook spot. This is expected, as many motors that have yet to synthesize hooks contain FliM (Aizawa, 1996). However, the brightest FliM–YFP foci were not associated with the hooks. Currently, we do not have an explanation for this fact. We do not know why incomplete motors should be brighter. And it seems unlikely that the brighter spots are inclusion bodies, because there is generally only one inclusion body formed per cell in the fliG background (Fig. 5C), even when the cells are made filamentous by growth in the presence of cephalexin (data not shown).

Figure 6.

Association of FliM–YFP with flagellar motors, demonstrated by immunofluorescence. fliM fliC cells expressing FliM–YFP were stained with antihook antibody and counterstained with anti-rabbit Texas red-coupled antibody, as described in Experimental procedures.

A. FliM–YFP (green).

B. Flagellar hooks (red).

C. Superimposition of the images in (A) and (B). All the red spots (hooks) are close to or overlap with the green spots (FliM–YFP). Additional green spots represent immature motors (motors with C-rings but without hooks).


Two protein complexes are of particular importance in bacterial chemotaxis: the receptor complex, consisting of MCPs and associated proteins; and the switch complex, assembled on the cytoplasmic face of the flagellar motor. The clustering of receptor complexes adds an additional level of complexity and could be important for signal amplification (Bray et al., 1998; Levit et al., 1998; Duke and Bray, 1999). We used YFP fusions to study the localization of a number of the components of these complexes. Except for CheA, the fusions were functional and restored chemotaxis in the corresponding null mutant background. The present results (together with the work of M. Manson, in preparation) provide direct visualization of receptor and switch complexes in living cells.

Our data imply the existence of a stable complex in which CheW, CheA, CheY and CheZ co-localize with MCP. We did not make a CheW fusion, but CheW was required for association of the other components, in agreement with previous studies (Ninfa et al., 1991; Gegner et al., 1992; Maddock and Shapiro, 1993). The deletion of either of the genes that encodes a major MCP, tar or tsr, decreases the degree of localization of CheA, CheY or CheZ, but it does not abolish it. This is consistent with the suggestion that different MCP proteins are intermixed in the same clusters (Stock and Levit, 2000). None of the fusion proteins formed clusters in strains that were deleted for all of the MCPs or for CheW; therefore, their localization is not an artifact resulting from the formation of inclusion bodies. CheW is required for localization of CheA, and CheA is required for localization of both CheY and CheZ; however, CheY and CheZ localize independently. As MCPs localize at the poles in the absence of CheW or CheA, the localization of CheY and CheZ must result from their binding to the MCP–CheW–CheA ternary complex, not to MCP directly.

We were also able to observe assembly of FliM–YFP on the switch complex. If a motor had a hook, as indicated by immunofluorescence, it also had a switch complex, as indicated by fluorescence of FliM–YFP. However, not all motors had hooks. FliM localization was not observed in a flhC strain, where flagellar synthesis does not occur. Nor was it observed in the absence of FliG, in support of a mediatory role for FliG in FliM attachment. FliN was not required for the FliM localization as long as FliG was present, consistent with the biochemical evidence for FliM, FliG binding (Marykwas and Berg, 1996; Marykwas et al., 1996; Tang et al., 1996). A possible explanation for the absence of FliM in basal body preparations from fliN mutants (Zhao et al., 1995) is that FliN stabilizes the attachment of FliM to FliG, which is otherwise lost in the purification procedure. We did observe somewhat lower intensity of FliM localization in a fliN background.

However, we were not able to observe localization of CheY fusion proteins to the switch complex, even under conditions in which cells tumbled. This might be a signal-to-background problem, as there are many more CheY molecules in the cell than there are FliM binding sites. On the other hand, there are a relatively large number of components with which to build CheW–CheA–MCP complexes and a relatively small number of molecules of FliM (see Bray et al., 1993).

An obvious advantage in the use of YFP (or other GFP) fusions is that the architecture of receptor and motor complexes can be probed in living cells. Our results complement those obtained by immunofluorescence and protein purification.

Experimental procedures

Strains and plasmids

E. coli strains used in this work were derived from strain RP437, a K-12 derivative that is wild type for chemotaxis (Parkinson and Houts, 1982). All strains and plasmids are listed in Table 3. LB (Luria–Bertani) medium or tryptone broth (TB; 1% tryptone, 0.5% NaCl) was used for E. coli cultures. Ampicillin was used at 100 µg ml−1 and kanamycin at 50 µg ml−1. Chemotaxis tests were performed on TB soft agar plates (1% tryptone, 0.5% NaCl, 0.3% Difco agar).

Table 3. Bacterial strains and plasmids used in this study.
Strain or plasmidRelevant genotype/phenotypeReference or source
 RP437Wild type for chemotaxis Parkinson and Houts (1982)
 RP9535ΔcheAJ. S. Parkinson
 RP1616 ΔcheZ J. S. Parkinson
 RP3021 cheW J. S. Parkinson
 RP4532 Δtar J. S. Parkinson
 RP5714 Δtsr J. S. Parkinson
 RP1131 trg::Tn10J. S. Parkinson
 RP3098 Δ(flhC–flhA) J. S. Parkinson
 HCB339 Δ(tar–tap) Δtsr trg::Tn10J. S. Parkinson
 DFB190 fliM null strain, fliM::cam Tang and Blair (1995)
 DFB228 fliM null strain, in frame fliM deletionD. F. Blair
 DFB232 fliM fliN null strainD. F. Blair
 DFB247 fliM fliG null strainD. F. Blair
 GP90 fliM::cam fliC::Tn10 clpP::cam recA::kanP. Danese
 VS100 ΔcheY, in frame cheY deletionThis work
 VS103 ΔcheY fliM::cam ΔflgM This work
 pBAD18KExpression vector, PBAD promoter, KmR; parent of pVS1, pVS5,
 pVS11, pVS30, pVS50, pVS53, pVS56, pVS59
Guzman et al. (1995)
 pBAD30Expression vector, PBAD promoter, ApR; parent of pVS13 Guzman et al. (1995)
 pBAD33Expression vector, PBAD promoter, CmR, pACYC ori; parent of pVS15, pVS17 Guzman et al. (1995)
 pEYFPYFP expression plasmid, ApRClontech
 pRL22CheY expression plasmid, ApR Matsumura et al. (1984)
 pDFB72FliM expression plasmid, ApR Tang and Blair (1995)
 pDV4CheA expression plasmid, ApR Hess et al. (1987)
 pDV4cheA48HQCheA48HQ expression plasmid, ApR Oosawa et al. (1988)
 pVS1CheY–YFP expression plasmid, KmRThis work
 pVS5YFP–CheY expression plasmid, KmRThis work
 pVS11FliM–YFP expression plasmid, KmRThis work
 pVS13FliM–YFP expression plasmid, ApRThis work
 pVS15CheY–YFP expression plasmid, CmRThis work
 pVS17YFP–CheY expression plasmid, CmRThis work
 pVS30YFP–FliM expression plasmid, KmRThis work
 pVS50YFP–CheZ expression plasmid, KmRThis work
 pVS53CheZ–YFP expression plasmid, KmRThis work
 pVS56YFP–CheAL expression plasmid, KmRThis work
 pVS59YFP–CheAs expression plasmid, KmRThis work
 pPA56Tsr290−470 expression plasmid, ApR Ames and Parkinson (1994)
 pAMPTSCloning vector, ts origin of replication, ApR; parent of pVS20, pVS21G. J. Phillips, personal gift
 pVS20 ΔcheY construct, ApRThis work
 pVS21 ΔflgM construct, ApRThis work

DNA methods

E. coli plasmid DNA was purified using the QIAprep Spin Miniprep Kit (Qiagen). Polymerase chain reaction (PCR) was performed in a MiniCycler (MJResearch) using Pwo DNA polymerase (Boehringer Mannheim). PCR primers used in this study are listed in Table 4; they were from Integrated DNA Technologies. Sequencing was carried out at the Harvard MCB sequencing facility.

Table 4. Primers used in this study.
PrimerSequenceaPriming siteb
  • a. 

    Introduced restriction sites (SacI, XbaI or EcoRI) are underlined; sequences encoding glycine linkers are marked in italics.

  • b

    . Relative to the transcriptional start site (+1) of the corresponding gene.

  • c

    . Complementary strand.

 VIC35′-TCAGTTGGAATTCTAGAGTC-3′749→729 (eyfpcompl)
 VIC3(See above) 
 VIC20(See above) 
 VIC11(See above) 
 VIC21(See above) 
 VIC3(See above) 
 VIC20(See above) 
 VIC11(See above) 
 VIC20(See above) 
 VIC11(See above) 
 VIC20(See above) 
 VIC11(See above) 
 VIC63(See above) 
cheY deletion

Strain construction

An in frame deletion in cheY (codons 8–122) was generated in vitro using two PCR steps, as described by Higuchi (1989), with pRL22 as template and the primers listed in Table 4. Outer primers were designed to contain EcoRI (VIC19) and XbaI (VIC14) restriction sites, and these sites were used to clone the fragment into the temperature-sensitive pAMPTS vector. The resulting construct, pVS20, was transformed in E. coli and grown on LB plates with ampicillin at 30°C. Transformants were streaked on LB–ampicillin plates and grown overnight at 42°C, allowing only the growth of cells that integrated the pVS20 construct into the chromosome. From these plates, single colonies were picked, grown for 24 h in LB at 30°C without selection, plated at serial dilutions on LB–ampicillin plates and grown overnight at 30°C. The colonies were then tested for ampicillin resistance and chemotaxis on soft agar plates.

Construction of YFP fusion proteins

YFP fusions to chemotaxis proteins and FliM were constructed using PCR. The target gene and the eyfp gene (Clontech) were amplified using primers with complementary overhangs, encoding either a 3× or a 5× glycine linker (Table 4). The resulting DNA fragments were annealed and amplified in a second round of PCR to form a fragment encoding a fusion protein with either a 3× Gly or a 5× Gly linker between the target protein and YFP. Outer primers were designed to contain SacI and XbaI restriction sites, and these were used to clone fragments into the arabinose-inducible pBAD18K expression vector. The final constructs were sequenced to ensure no PCR mistakes. Expression of full-length fusion proteins was verified by immunoblot using antibodies against FliM, CheY, CheZ and/or YFP (see below). Immunoblots were quantified using the program nih image. The levels of expression of fusion proteins (compared with native proteins) estimated from immunoblots were as follows: CheY–YFP, ≈ 90%; YFP–CheY, ≈ 70%; CheZ–YFP, ≈ 110%; YFP–CheZ, ≈ 70%; YFP–CheAL, ≈ 120%; YFP–CheAS, ≈ 180%; FliM–YFP, ≈ 150%; YFP–FliM, ≈ 100%. The total levels of degradation products were: CheY–YFP and YFP–CheY, < 3%; FliM–YFP and YFP–FliM, < 10%; CheZ–YFP, < 10%; YFP–CheZ, not detected; YFP–CheAL and YFP–CheAS, not detected. FliM–YFP degradation was further reduced to < 2% by the introduction of a clpP mutation (ClpP is a component of the ClpAP protease) in the background strain (GP90) without loss of motor function. Functionality of fusion proteins was tested by the restoration of chemotaxis on soft agar plates supplemented with kanamycin (50 µg ml−1) and arabinose (0.001–0.005%).


Immunoblots were performed as described previously with minor modifications (Scharf et al., 1998). Motile cells expressing fusion proteins were grown as described below. Whole-cell extracts were prepared from 10 ml samples. Cells were washed once with PBS, resuspended in 300 µl of PBS and lysed by sonication. SDS–PAGE loading buffer (3×) was added, samples were boiled at 95°C for 5 min, and 5 µl of each sample was loaded onto an SDS gradient (8–15%) polyacrylamide gel. After separation by electrophoresis, the proteins were electroblotted to a Hybond ECL nitrocellulose membrane using a tank blot device (Bio Labs, Harvard University) for 3 h at 60 V in transfer buffer (25 mM Tris, 192 mM glycine, 0.05% SDS, 20% methanol, pH 8.3). Blots were blocked overnight at room temperature in TBS-T [20 mM Tris-HCl, 140 mM NaCl, 0.1% (v/v) Tween 20, pH 7.6] with 5% blocking reagent (instant non-fat dry milk) on a rocking platform. Blots were then incubated with primary antibodies in TBS-T for 2 h at room temperature. Monoclonal anti-GFP antibodies (Clontech) were used at a 1:1500 dilution, monoclonal anti-CheY and anti-CheZ antibodies (Scharf et al., 1998) at a 1:1000 dilution and polyclonal anti-FliM antibodies (a gift from David Blair) at a 1:1000 dilution. Blots were washed three times (15 min each) with TBS-T, incubated for 2 h with sheep anti-mouse (or anti-rabbit) horseradish peroxidase-linked secondary antibodies (Amersham) diluted 1:2500 in TBS-T, washed again and detected using an ECL kit (Amersham), as described by the manufacturer.

Growth conditions and fluorescence microscopy

Motile cell cultures were grown in tryptone broth (TB) with kanamycin (50 µg ml−1) at 33°C. To obtain motile cells expressing YFP fusions for fluorescence measurements, overnight cultures were diluted 1:100 in TB containing arabinose (0.002–0.005%, as indicated in the text) and allowed to grow for 4 h in a rotary shaker. A cell suspension (100 µl) was applied to a polylysine-coated coverslip, incubated for 5 min and washed three times with tethering buffer (Block et al., 1983) before microscopy.

For double staining with FliM–YFP and polyclonal antihook antibody (Ishihara et al., 1983), cells from a 1 ml culture were resuspended in 100 µl of tethering buffer and incubated with antibody at a 1:500 dilution for 20 min. Cells were washed three times with tethering buffer and incubated in 100 µl of tethering buffer with secondary goat anti-rabbit Texas red-conjugated antibodies (1:300 dilution; Molecular Probes) for another 20 min. After incubation, cells were washed with tethering buffer, applied to a polylysine-coated coverslip, washed with tethering buffer once more and imaged.

For double staining with CheY–YFP and anti-Tsr antibody, cells from a 1 ml culture were fixed using methanol as described previously (Teleman et al., 1998). Fixed cells were placed on a polylysine-coated coverslip, allowed to dry fully and treated with lysozyme (2 mg ml−1) in GTE buffer (50 mM glucose, 25 mM Tris, 1 mM EDTA, pH 7.5) for 10 min. Coverslips were incubated with a blocking solution (2% BSA in PBS, pH 7.5) overnight at 4°C, incubated in 2% BSA–PBS with anti-Tsr antibodies (1:500 dilution; a gift from Sandy Parkinson) for 2 h at room temperature, washed 10 times with PBS, incubated with secondary goat anti-rabbit Texas red antibody (1:300 dilution in 2% BSA–PBS) for 2 h at room temperature, washed with PBS again and imaged.

Fluorescent microscopy was performed using a Delta Vision deconvolution microscope and program package (Applied Precision). YFP images were taken using a bandpass excitation filter (480–500 nm) and either a longpass emission filter (510 nm) or, if cells were also stained with Texas red, a bandpass emission filter (509–547 nm). Texas red images were taken using a bandpass excitation filter (541–569 nm) and a bandpass emission filter (581–654 nm). When necessary, images were quantified using the program nih image. Deconvoluted images were prepared for final publication using Adobe Photoshop 5.5 and a Tektronix Phaser 450 printer.


We thank Paul Danese for inspiration and help during the early phases of this study. This work was supported by grant AI16478 from the National Institute of Allergy and Infectious Diseases and by the Rowland Institute for Science.