Molecular characterization of Escherichia coli FtsE and FtsX


Joen Luirink. E-mail; Tel. (+31) 20 444 7175; Fax (+31) 20 444 7136.


The genes ftsE and ftsX are organized in one operon together with ftsY. FtsY codes for the receptor of the signal recognition particle (SRP) that functions in targeting a subset of inner membrane proteins. We have found no indications for a structural relationship between FtsE/X and FtsY. Evidence is presented that FtsE and FtsX form a complex in the inner membrane that bears the characteristics of an ATP-binding cassette (ABC)-type transporter. FtsE is a hydrophilic nucleotide-binding protein that has a tendency to dimerize and associates with the inner membrane through an interaction with the integral membrane protein FtsX. An FtsE null mutant showed filamentous growth and appeared viable on high salt medium only, indicating a role for FtsE in cell division and/or salt transport.


In Escherichia coli, cell division involves the formation of a septum at the middle of the rod-shaped cell. Genetic studies have identified several conditional lethal mutants that fail to septate at the non-permissive temperature (Donachie, 1993; Bramhill, 1997). Collectively, these mutants have been designated fts (filamentation temperature sensitive). The most extensively studied fts genes, such as ftsQ, ftsA and ftsZ, are clustered in the 2 min region of the chromosome (Bachmann, 1983). These genes play an essential role in cell division, although their precise function is still under study (Donachie, 1993; Bramhill, 1997; Lutkenhaus and Addinall, 1997).

Another fts gene cluster is located at 76 min on the genetic map and consists of three genes, ftsY, ftsE and ftsX, which are organized in one operon and transcribed in the same direction (Crickmore and Salmond, 1986; Gill and Salmond, 1990). Recent evidence shows that FtsY belongs to the family of ‘signal recognition particle (SRP)-type GTPases’ that are involved in protein targeting in prokaryotic and eukaryotic cells (Rapoport et al., 1996). FtsY functions in the targeting of proteins to the E. coli inner membrane as the receptor of the SRP (Luirink et al., 1994; Miller et al., 1994). The SRP is a ribonucleoprotein particle that binds to short nascent (ribosome-associated) polypeptides that expose a hydrophobic targeting signal just outside the ribosome (Valent et al., 1995; 1997). FtsY mediates the release of the SRP from the nascent chain at the cytoplasmic membrane in a GTP-dependent process, so that the nascent chain can enter the membrane at the translocation channel (Valent et al., 1998). This targeting route is used primarily by inner membrane proteins (reviewed in De Gier et al., 1997). Disruption of the SRP pathway may indirectly affect cell division, as many essential proteins that function in cell division are located in the inner membrane.

The direct involvement of FtsE in cell division has been questioned, as the conditional lethal FtsE mutant filaments at the non-permissive temperature only in rich medium (Taschner et al., 1988). Based on its primary sequence, FtsE has been proposed to be a nucleotide-binding protein (Gill et al., 1986).

FtsY, E and X were found to be associated with the inner membrane in a maxicell expression system (Gill and Salmond, 1987). Based on this finding and their co-ordinated expression, it has been argued that their function is related (Gill et al., 1986; Gill and Salmond, 1987). In this study, we have investigated the function and molecular properties of FtsE and FtsX in more detail. FtsE and FtsX appear to display characteristics that are common to ABC transporters, nucleotide-binding proteins that are involved in a variety of transport processes in prokaryotic and eukaryotic cells (Higgins, 1992; Binet et al., 1997). Thus, the hydrophilic FtsE is associated with the inner membrane through an interaction with FtsX, which is shown to be an integral membrane anchor protein. Furthermore, FtsE is a nucleotide-binding protein that has a tendency to dimerize. Surprisingly, FtsE is not an essential gene when cells are grown in medium containing high salt levels, indicating that the FtsE/X complex is directly or indirectly involved in salt transport. A functional or structural relationship between FtsY and FtsE/X could not be established.


Characterization of an ftsE null mutant

The designation filamentation temperature sensitive (fts) was originally used to describe a class of E. coli conditional cell division mutants that are viable at the permissive temperature of 30°C. At 42°C, the cells elongate, filament and die (Donachie, 1993; Bramhill, 1997). The fts designation has since been applied to other genes exhibiting this mutant phenotype, including the ftsYEX operon.

Insertional inactivation of ftsE in strain RG60 (see Experimental procedures) resulted in a filamented cellular morphology, but not in a temperature-sensitive phenotype. It should be noted that, given the operon structure of ftsYEX and the opposite orientation of the kanamycin cassette insertion, this mutation of ftsE is likely to have a polar effect on ftsX expression. Thus, the filamentation phenotype may be a combined effect of the insertional inactivation of ftsE and the downstream polar effect on ftsX. When grown on LB with 1% NaCl, RG60 was viable at both 30°C and 42°C. While moderately filamented at 30°C, the cells exhibited more filamentation at 42°C. A decrease in the concentration of NaCl in the growth medium resulted in extreme cell filamentation and, finally, in cell death (data not shown). Restoration of viability occurred between 0.5% and 0.6% NaCl at both 30°C and 42°C. Taken together, these results suggest that ftsE is a non-essential gene when cells are cultured in the presence of concentrations of NaCl over 0.5% or, alternatively, ftsE may be considered as a conditional salt-dependent essential gene.

Salt and osmotic rescue of RG60 was studied in more detail (Table 1). The viability of RG60 was not restored by high concentrations of sucrose or glycerol. Amino acids important in osmoregulation, such as glutamate, glutamine, proline and the compatible solute glycine betaine, did not rescue RG60. Hence, an increase in osmolarity alone is not a mechanism of rescue in this case, making a cell wall defect unlikely. No pattern of rescue by monovalent versus divalent ions was apparent.

Table 1. . Effects of different compounds on the viability of RG60 at 42°C as determined by disc assays (see Experimental procedures). ‘+’ indicates the presence of visible growth. ‘−’ indicates that no growth was visible. All compounds that rescued RG60 did so at approximately the same level of about 6 mm ring of growth around the paper disc. However, LiCl2 and MgCl2 rescued to a lesser extent, about 2–3 mm ring of growth. CsCl exhibited a zone of killing of about 3 mm with a 3 mm ring of growth beyond that.Thumbnail image of

The ftsY gene, which is located in the same operon as ftsE and ftsX, encodes the SRP receptor. Depletion of FtsY has a strong effect on export of β-lactamase and the assembly of various inner membrane proteins (Luirink et al., 1994; Ulbrandt et al., 1997). To study a functional relationship between FtsY and FtsE/X, we examined the effects of the ftsE null mutation on export of β-lactamase and assembly of the SRP-dependent inner membrane protein leader peptidase (Lep). The precursor form of β-lactamase did not accumulate at any growth temperature or salt concentration in the medium, as evidenced by immunoblotting of cell extracts (not shown). Also, membrane assembly of Lep appeared to be unaffected using an assay that determines the protease accessibility of a periplasmic domain of pulse-labelled Lep (De Gier et al., 1996; data not shown). In conclusion, FtsE appeared not to be required for SRP-mediated protein targeting.

Overexpression of FtsE and FtsX

To examine the effects of overexpression of FtsE and FtsX and to facilitate the subcellular localization and purification of these proteins, the genes encoding FtsE and FtsX were subcloned as separate entities under the control of an inducible bacteriophage T7 promoter. Similarly, the gene fragment encoding both FtsE and FtsX was subcloned without disrupting the operon structure to preserve the natural stoichiometry of expression.

Induction of FtsE or FtsE/X expression resulted in the accumulation of large amounts of FtsE, which migrated as a prominent 23 kDa band upon SDS–PAGE. This is consistent with the predicted molecular mass for FtsE of 24 kDa (Fig. 1A, lanes 2 and 4). Overexpression of FtsE alone did not affect cell growth (Fig. 1B) and led to the formation of large polar inclusion bodies. Most of the overexpressed FtsE could be sedimented by centrifugation at 12 400 × g. Induction of FtsX (alone or in combination with FtsE) resulted in a relatively moderate overexpression, but could be distinguished upon SDS–PAGE at 36 kDa (Fig. 1A, lanes 3 and 4), which is slightly below its expected molecular weight (39 kDa). Overexpression of FtsX led to a strong inhibition of growth (Fig. 1B), cell filamentation and eventually cell death (Fig. 1B). Overexpression of both FtsE and FtsX also affected cell growth and morphology, albeit less than upon overexpression of FtsE or FtsX alone.

Figure 1.

. Overexpression of FtsE and FtsX. A. Coomassie-stained SDS–PAGE of BL21(DE3) cells (lane 1) overexpressing FtsE (lane 2), FtsX (lane 3) or both FtsE and FtsX (lane 4). Cells were grown to an OD660 of 0.3 and induced with 0.4 mM IPTG. Samples were collected 2 h after induction. FtsE-His, present in washed inclusion bodies (lane 5), was affinity purified (lane 6). The positions of molecular weight markers and FtsE monomer, FtsE dimer and FtsX are indicated. B. Growth curves of BL21(DE3) overexpressing FtsE, FtsX or both FtsE and FtsX. Cells were grown to an OD660 of 0.3 and induced with 0.4 mM IPTG at 0 h.

Overexpression of FtsY strongly inhibits the export of β-lactamase, resembling the effect of overexpression of the SRP component P48 (Luirink et al., 1994). To examine the effects of overexpression of FtsE and FtsX on protein export, the accumulation of precursor forms of β-lactamase was monitored in vivo by immunoblotting. A slight accumulation of the β-lactamase precursor form was apparent upon overexpression of FtsE and/or FtsX, but only after prolonged exposure of the chemiluminescent immunoblots (data not shown). Similar results were obtained for the SRP-independent outer membrane proteins OmpA and OmpC. This aspecific effect is most probably caused by general stress imposed on cells by overproduction of membrane proteins and not by interference with SRP-mediated protein targeting.

Purification of FtsE and FtsX

To facilitate purification of FtsE and FtsX, both proteins were provided with a carboxy-terminal 6xHis tag. FtsE-His was purified from inclusion bodies (Fig. 1A, lane 5) by urea extraction, followed by affinity chromatography using Ni-NTA resin. SDS–PAGE analysis of the purified FtsE revealed that a second protein migrating at 46 kDa was co-purified with the FtsE monomer (Fig. 1A, lane 6). To identify this protein and to confirm the identity of FtsE, the N-terminal amino acid sequence of the proteins was determined and appeared to be identical to the predicted sequence of FtsE. Thus, the 46 kDa protein most probably represents a FtsE dimer. The overall purity of both bands together was estimated to be > 95%. The purified protein was used to raise antibodies that reacted with both the FtsE monomer and the FtsE dimer in immunoblots.

FtsX-His was purified from detergent-solubilized membranes, and its identity was confirmed by N-terminal amino acid sequencing. However, the purified protein did not yield specific antibodies. For this reason, a serum was raised against a peptide that consisted of the amino-terminal 15 residues of FtsX.

Dimerization of FtsE

The co-purification of the FtsE monomer and dimer prompted us to investigate the dimerization of FtsE in more detail. Dimerization appeared to depend on the concentration of urea present during the purification process. However, the FtsE dimer was still present after incubation in urea concentrations as high as 8 M, indicating strong association (Fig. 1A, lane 6). The concentration of FtsE present in the sample did not influence dimer formation, indicating that it represents an interaction that is present in the native protein and not a result of the high concentration of FtsE in the purified protein sample. Dimer formation was also observed in SDS–PAGE analysis of crude cell extracts of cells overproducing FtsE (not shown). Denaturation in the presence and absence of dithiothreitol (DTT) indicated that FtsE in the dimer was covalently linked by interchain disulphide bonds (Fig. 2, lanes 3 and 4). FtsE contains a unique cysteine residue at position 49. Substitution of this cysteine for an alanine residue prevented dimer formation, confirming the involvement of the unique cysteine in this process (Fig. 2, lanes 3 and 11). Even though these results may reflect the tendency of FtsE to dimerize, the formation of disulphide bonds in vivo is not expected, given the location of FtsE in the reducing environment of the bacterial cytoplasm. Hence, disulphide bond formation might be an artifact resulting from exposure of the cysteine residues upon denaturation of FtsE during purification and sample preparation. To study this possibility, the effect of denaturation on the dimerization of purified FtsE was determined (Fig. 2). FtsE was diluted to 1 M urea, which favours dimer formation. Subsequently, the samples were denatured in sample buffer (2% SDS) and incubated at 37°C or 100°C in the presence or absence of DTT. The dimer was only converted to the monomeric form after denaturation at 100°C in the presence of DTT (Fig. 2, lane 4). This indicates a stable interaction in the dimer, in which the disulphide bridge is relatively inaccessible. The FtsE Cys-49 → Ala-49 mutant was monomeric under all conditions tested (Fig. 2, lanes 9–12).

Figure 2.

. Dimerization of FtsE. Wild-type FtsE (lanes 1–4), FtsE-K41R (lanes 5–8) and FtsE-C49A (lanes 9–12) were purified over Ni-NTA agarose from urea-extracted inclusion bodies. Samples were analysed by SDS–PAGE under non-reducing (lanes 1, 3, 5, 7, 9, 11) or reducing (lanes 2, 4, 6, 8, 10, 12) conditions without (lanes 1, 2, 5, 6, 9, 10) or with heat treatment (lanes 3, 4, 7, 8, 11, 12) and visualized by Coomassie staining. The position the FtsE monomer and FtsE dimer is indicated.

Subcellular localization of FtsE and FtsX

To study the location of FtsE and FtsX, cells that expressed these proteins at different levels were disrupted and subjected to differential centrifugation to separate unlysed cells and inclusion bodies (low-speed pellet) from membrane-associated material (high-speed pellet) and soluble proteins (supernatant fraction). The fractions were analysed by immunoblotting (Fig. 3). FtsE co-fractionated mostly with membranes at the wild-type expression level (Fig. 3A, lanes 1–3). FtsX could not be detected at the wild-type expression level (Fig. 3B, lanes 1–3) and co-fractionated with the membranes upon overexpression (Fig. 3B, lanes 4–6). Part of FtsE expressed at the wild-type level was present in the soluble fraction (Fig. 3A, lane 2). However, upon overexpression of FtsX, all FtsE was found in the membrane pellet (Fig. 3A, lanes 4–6), indicating that FtsX plays a role in the membrane association of FtsE. Overexpression of FtsE invariably resulted in the formation of aggregates that complicated the interpretation of fractionation data (not shown).

Figure 3.

. Subcellular localization of FtsE and FtsX. Fractionation of BL21(DE3) (lanes 1–3) and BL21(DE3) overexpressing FtsX (lanes 4–6). Cells were grown to an OD660 of 0.3 and induced with 0.4 mM IPTG. Samples were collected 2 h after induction. Fractions derived from 0.1 OD660 units were analysed by immunoblotting using antiserum raised against FtsE (A) or FtsX (B). T, total cell lysate; S, soluble fraction; P, total membrane fraction.

To determine the nature of interaction of both FtsE and FtsX with the membrane, inner membrane vesicles (IMVs) were purified by sucrose density centrifugation from cells expressing FtsE and FtsX at different levels, and subjected to extraction with different agents. Both FtsE and FtsX co-fractionated with the IMVs and were completely extracted from the IMVs by sodium lauryl sarkonisate, suggesting that they are inner membrane associated and are not fractionating with the IMVs as a result of aggregation (data not shown). The IMVs were also extracted with 1 M NaCl, 4 M urea or 0.2 M Na2CO3 to remove peripherally associated membrane proteins using 50 mM NaCl as a control condition. Extracted and residual membrane-associated material was analysed by immunoblotting (Fig. 4). FtsE expressed at the wild-type level was partly extracted with 0.2 M Na2CO3 but was relatively resistant to extraction with 1 M NaCl or 4 M urea (Fig. 4A, lanes 3–8). However, overexpression of both FtsE and FtsX rendered membrane association of FtsE completely Na2CO3 resistant (Fig. 4C, lanes 7 and 8). Most of the overexpressed FtsE was readily extracted already at 50 mM NaCl, indicating that it is loosely associated with the IMVs (Fig. 4B, lanes 1 and 2). Again, when FtsX was concomitantly overexpressed, the extraction of FtsE from the membrane was dramatically reduced (Fig. 4C, lanes 1 and 2).

Figure 4.

. Solubilization of peripheral inner membrane proteins. IMVs of BL21(DE3), BL21(DE3) overexpressing FtsE, FtsX or both FtsE and FtsX were extracted with 1 M NaCl (lanes 3 and 4), 4 M urea (lanes 5 and 6) or 0.2 M Na2CO3 (lanes 7 and 8) with 50 mM NaCl as the control incubation (lanes 1 and 2). Pellet (P) (lanes 1, 3, 5 and 7) and soluble (S) (lanes 2, 4, 6 and 8) fractions were analysed by immunoblotting using antiserum against FtsE (A, B and C) or against FtsX (D and E).

Taken together, these data suggest that FtsE does not contain a membrane integral anchor sequence consistent with computer predictions. Rather, FtsE interacts with FtsX in the IMVs to acquire a stable membrane association. Overexpressed FtsX was not significantly extracted by any of the agents used (Fig. 4D and E, lanes 1–8), suggesting that it is an integral inner membrane protein consistent with the presence of four predicted membrane-spanning hydrophobic segments (Gill and Salmond, 1986).

8-Azido[α-32P]-ATP labelling of FtsE

The predicted amino acid sequence of FtsE contains the Walker A and B consensus motifs (Gill and Salmond, 1986) that contribute to the ATP binding site in proteins belonging to the ABC family of proteins (Saraste et al., 1990; Higgins, 1992). To investigate nucleotide binding to FtsE, purified FtsE was irradiated with UV light after incubation with the photoreactive ATP analogue 8-azido[α-32P]-ATP. As shown in Fig. 5, lane 2, FtsE was labelled with 8-azido[α-32P]-ATP upon UV irradiation, indicating that it is an ATP-binding protein. The labelling of FtsE was inhibited by a 40-molar excess of ATP, ADP or GTP, but not AMP (Fig. 5A, lanes 3–6). The FtsE dimer displayed similar labelling characteristics. The control protein ovalbumin that lacks Walker A and B motifs was not labelled under these conditions (data not shown).

Figure 5.

. 8-Azido[α-32P]-ATP labelling of FtsE. Purified wild-type FtsE (A and B), FtsE-K41R (C and D) or FtsE-C49A (E and F) was preincubated with 8-azido[α-32P]-ATP. Samples were irradiated by UV light in the absence (lane 2) or in the presence of different nucleotides (added at a 40:1 molar ratio over 8-azido[α32P]-ATP; lanes 3–6) with no UV irradiation as the control incubation (lane 1). Samples were analysed by SDS–PAGE and phosphorimaging (A, C and E) or Coomassie staining (B, D and F). Purified protein (3 μg) was used in each reaction.

The Walker A motif (GXXGXGKT/S) is conserved in both GTP and ATP-binding proteins (Saraste et al., 1990). The lysine residue in this motif, also called the P-loop, is most conserved and in direct contact with the β- and γ-phosphate of the bound NTP. In FtsE, the consensus sequence (GHSGAGKS) is located between residues 35 and 42. To confirm that this sequence is involved in ATP binding, the lysine at position 41 was converted into arginine. The mutated protein, FtsE-K41R, was purified and appeared to be unable to bind 8-azido[α-32P]-ATP (Fig. 5C, lane 2), indicating that the assigned Walker A motif in FtsE is indeed involved in nucleotide binding. The mutation did not affect dimerization (Fig. 2, lanes 5–8). Conversely, the Cys-49 → Ala-49 replacement did not alter the nucleotide-binding properties of FtsE (Fig. 5E), indicating that the cysteine residue, which is close to the Walker A motif, is not essential for nucleotide binding. Similar results were obtained when IMVs derived from cells that overexpress (mutant) FtsE were analysed for their nucleotide-binding properties (data not shown).

Interactions between FtsE and FtsX

ABC transporters in bacteria are often composed of separate membrane peripheral ATP-binding proteins and transmembrane proteins (Higgins, 1992; Binet et al., 1997). The subcellular localization data (presented above) suggested that FtsE and FtsX interact to form a membrane-associated complex in which FtsX would serve as the transmembrane protein that ties the nucleotide-binding protein FtsE to the membrane. To gain independent evidence that FtsE and FtsX associate to form a multisubunit complex, two different experimental strategies were used.

First, crude membranes were purified from pulse-labelled cells that overexpress FtsE, FtsX or both FtsE and FtsX. As observed before, both FtsE and FtsX co-localized with the membrane fraction (Fig. 3). The membranes were incubated with the thiol-cleavable, homobifunctional, membrane-permeable cross-linker DSP to allow the formation of covalent bonds between membrane proteins that are in close proximity. Subsequently, the samples were subjected to immunoprecipitation under denaturing conditions using anti-FtsE antibodies. Before SDS–PAGE, the samples were heated with DTT to dissociate cross-linked complexes and analyse the co-immunoprecipitated proteins. Overproduction of FtsE alone results in enhanced immunoprecipitation of three proteins of approximately 24, 60 and 96 kDa (Fig. 6, lane 9). These products were not observed when FtsE was not overproduced or when a control antiserum was used, indicating that they are related to FtsE (Fig. 6, lanes 4, 10 and 14). The 24 kDa product probably represents FtsE, whereas the identity of the other proteins remains to be identified. When both FtsE and FtsX were overproduced, an extra product was co-immunoprecipitated with FtsE that migrated at the position of FtsX (36 kDa), suggesting strongly that FtsE and FtsX are in close proximity in the membrane (Fig. 6, lane 19).

Figure 6.

. Co-immunoprecipitation of FtsE and FtsX. BL21(DE3) (lanes 1–5), BL21(DE3) overexpressing FtsE (lanes 6–10), FtsX (lanes 11–15) or both FtsE and FtsX (lanes 16–20) were grown to an OD660 of 0.3 in M9 medium lacking cysteine and methionine. Expression was induced with 1 mM IPTG for 15 min, and the cells were pulse-labelled for 10 min with [35S]-methionine. Cells were fractionated, and crude membrane fractions were incubated with DSP. Membranes were solubilized, and the supernatant was immunoprecipitated under denaturing conditions using antiserum against FtsE (lanes 4, 9, 14 and 19) or a control antiserum (lanes 5, 10, 15 and 20). Samples were analysed by SDS–PAGE under denaturing conditions and visualized by phosphorimaging. T, total cell lysate; P, total membrane fraction; S, soluble fraction.

Secondly, we tried to purify recombinant protein complexes using FtsX that was modified to introduce a His-tag at its C-terminal end. This tag allows the purification of FtsX by Ni-NTA affinity chromatography under mild conditions that are expected to preserve the interaction with associated proteins. His-tagged FtsX was overexpressed in cells that also overproduce wild-type FtsE. A crude membrane extract was prepared and solubilized with the detergent dodecylmaltoside. Under these conditions, only a small portion of FtsX is solubilized. Complete solubilization of the integral membrane protein FtsX required SDS or Triton X-100, conditions that are likely to disrupt protein interactions. After centrifugation of the insoluble material, the supernatant was subjected to Ni-NTA affinity chromatography. The Ni-NTA-bound material was washed twice with 30 mM imidazole and eluted with 500 mM imidazole. Wash and elution fractions were analysed by SDS–PAGE and immunoblotting to detect the low amounts of solubilized FtsE and FtsX (Fig. 7). As a control for the aspecific binding of both FtsE and FtsX to the resin, a membrane extract was prepared from cells that lack His-tagged FtsX but overproduced wild-type FtsX and FtsE.

Figure 7.

. Co-purification of FtsE and FtsX-His. BL21(DE3) overexpressing both FtsE and FtsX (bottom) or FtsE and FtsX-His (top) were grown to an OD660 of 0.3 and induced with 0.4 mM IPTG. Cells were harvested 2 h after induction. A crude membrane extract was prepared and solubilized with dodecylmaltoside. FtsX-His was purified from the solubilized membranes using Ni-NTA agarose, as described in Experimental procedures. Fractions were TCA precipitated and analysed by immunoblotting using antiserum against FtsE or FtsX as indicated. P, total membrane fraction; S, soluble fraction; FT, flowthrough; W1 and W2, wash fractions; E, eluate.

A small fraction of FtsX-His, but not wild-type FtsX, remained bound after washing of the resin and was eluted at 500 mM imidazole, indicating that it was attached to the resin via its His-tag (Fig. 7A and C, lane 6). A small but reproducible fraction of the co-expressed FtsE co-eluted specifically with FtsX-His at 500 mM, indicating that FtsX-His and FtsE interact physically. It should be noted that repeated washing was necessary to remove untagged FtsX from the column (Fig. 7C, lanes 4 and 5). Likewise, part of FtsE was present in the wash fractions (Fig. 7B and D, lanes 4 and 5). Possibly, part of the overexpressed FtsX is aggregated and does not fully expose the His-tag.


In this study, we have shown that FtsE and FtsX interact and together make up a complex that bears characteristics of an ABC-type system. Mutant cells that lack FtsE show filamentous growth and require a high salt concentration in the medium, indicating that this complex is directly or indirectly involved in cell division and/or salt transport. The exact function and substrate specificity of this system remains elusive.

In E. coli, ftsE and ftsX are located in one operon together with ftsY (Gill and Salmond, 1986). FtsY is a hydrophilic protein that is associated in part with the inner membrane and functions in SRP-mediated protein targeting as the receptor for the SRP analogous to the SRP receptor α-subunit in eukaryotic cells (Luirink et al., 1994; Miller et al., 1994; Valent et al., 1998). It has been proposed that the functions of FtsY, FtsE and FtsX are related (Gill et al., 1986; Gill and Salmond, 1987) and that FtsE/X play a role in the membrane anchoring of FtsY analogous to the function of the β-subunit of the eukaryotic SRP receptor in the membrane of the endoplasmic reticulum (Luirink et al., 1994). We did not obtain any evidence for a functional or structural relationship between FtsE/X and the SRP receptor FtsY. Neither the absence nor the overexpression of FtsE/X had a specific effect on membrane assembly and the export of SRP-dependent proteins. It should be noted that ftsE and ftsX homologues are present and found clustered in most, but not all, known bacterial genome sequences, suggesting a linked function. However, only in E. coli and H. influenzae is the ftsE/X gene cluster preceded by the ftsY gene, whereas in all other species, ftsY homologues are present but encoded by separate genes. Hence, genetic linkage of ftsY and ftsE/X in E. coli may be fortuitous. Interestingly, two transcripts are derived from the E. coli ftsYEX operon: one encoding all three gene products and one encoding FtsE and FtsX, indicating that the expression of FtsY and FtsE/X is partly independent (Gill and Salmond, 1990).

When grown on LB medium containing > 0.5% NaCl, ftsE and ftsX do not appear to be essential genes. Previous reports have characterized ftsE and ftsX as essential genes involved in cell division (Salmond and Plakidou, 1984; Gill and Salmond, 1986). However, those studies were done in medium that contained 0.5% NaCl compared with the 1% NaCl used in this study. Salt repair has been reported for ftsE(Ts) mutants in an early study (Ricard and Hirota, 1973). It was assumed that these were missense mutants, which might be rescued by changes in the conformation, function or stability of the protein induced by electrolytes and osmolytes. The present study, in which the ftsE gene has been knocked out, makes this explanation less likely. Interestingly, it has been shown very recently that membrane insertion of a subset of K+ pump proteins is specifically impaired in an ftsE(Ts) mutant at the non-permissive temperature, which might explain the restoration of viability of this mutant by high KCl concentrations in the culture medium (Ukai et al., 1998).

There is the possibility that the null mutation of ftsE reveals the presence of a second protein whose presence has been suppressed in RG60. This possibility is being tested through screening of RG60 with genomic libraries. Clones that rescue RG60 under the restrictive conditions of 0% NaCl LB at 42°C should contain the complementing gene. Indeed, a clone containing both ftsE and ftsX has been isolated. Additionally, the isolation of multicopy suppressors of the ftsE ::kan mutant may elucidate the function of ftsE.

Based on its sequence, FtsE is predicted to be one of the smallest ATP-binding components of an ABC system (Gill et al., 1986; Higgins, 1992). ABC transporters form a large family of ATP-binding proteins involved in a variety of transport processes in both prokaryotic and eukaryotic cells. Interestingly, several Ts ftsE alleles have been found clustered in a highly conserved region that is often affected in the CFTR protein of cystic fibrosis patients (Gibbs et al., 1992). We have purified FtsE from urea-solubilized inclusion bodies and indeed observed ATP binding after dilution to 1 M urea, which appeared to be necessary to keep this protein soluble. The specificity of nucleotide binding as revealed by competition experiments and site-directed mutagenesis of the Walker A motif, resembled that of other ATPases of this family (Walter et al., 1992; Delepelaire, 1994). Under these slightly denaturing and non-reducing conditions, most of the FtsE formed a homodimer. Dimerization and the formation of the nucleotide-binding fold indicate that the protein refolds into its native conformation upon dilution from the denaturant.

Dimerization of the ATP-binding domain is consistent with the paradigm of two ATP binding sites per ABC transporter and is observed more frequently in bacterial ABC transporters (Higgins, 1992; Binet et al., 1997) and other membrane-bound ATPases such as SecA and PulE (Akita et al., 1991; Driessen, 1993; Possot and Pugsley, 1994). Although the dimers were linked covalently by disulphide bonds, we consider it unlikely that these bonds exist in situ, as FtsE appears to be a peripheral inner membrane protein. However, we cannot entirely exclude the possibility that the cysteine residue protrudes into the periplasm to form interchain disulphide bonds. Other hydrophilic peripheral ATP-binding membrane proteins have been shown to expose sequences that are accessible from the periplasm (Baichwal et al., 1993; Schneider et al., 1995; Ramamurthy and Oliver, 1997). Nevertheless, for FtsE, we consider this possibility unlikely, given the fact that the cysteine residue is located in between the Walker A and B motifs, a region that is most probably exposed towards the cytoplasm.

FtsE was shown to be membrane associated via its interaction with FtsX by independent approaches that included co-localization, co-purification of FtsE and His-tagged FtsX under native conditions and protein cross-linking. FtsX was shown to be an integral inner membrane protein consistent with localization studies carried out in a maxicell expression system (Gill and Salmond, 1987). Notably, FtsX contains only four hydrophobic segments that potentially form transmembrane helices. As a rule, ABC transporters are anchored in the membrane by two times six transmembrane segments that supposedly form the channel through which the substrate molecules cross the membrane (Higgins, 1992). We have not been able to detect dimerization or trimerization of FtsX in the membrane (data not shown).

Future studies will concentrate on the identification of the substrate of the putative FtsE/X complex. The small size of both FtsE and FtsX and the relative ease of purification of folded FtsE makes this ABC system an attractive model for structural studies.

Experimental procedures

General methods

Recombinant DNA techniques were carried out as described previously (Sambrook et al., 1989). Protein concentration was determined according to Bradford (1976). Radiolabelled protein bands on dried polyacrylamide gels were visualized by phosphorimaging using a Molecular Dynamics Phosphor Imager 473 and quantified using the IMAGEQUANT quantification software from Molecular Dynamics.

Strains and media

E. coli TOP10F (Stratagene) was used for routine cloning procedures. BL21 FhsdS gal (DE3) harbouring pLysS was used for the expression of FtsE and/or FtsX cloned in the pET21d expression vector (Studier et al., 1990).

Construction of ftsE knockout strain RG60 was as follows: plasmid pFtsYEX9, containing the ftsYEX operon and ORF4 (Gill and Salmond, 1986), was digested with ScaI and StuI. The resulting 4.9 kbp fragment was ligated to a 1.4 kbp spectinomycin/streptomycin resistance cassette, resulting in plasmid pFts′YEXSp. Making use of a unique MluI restriction site in pFts′YEXSp 201 bp downstream of the start of the ftsE gene, a 1.5 kbp cassette (MluI ends) encoding kanamycin resistance was ligated into ftsE. The orientation of the cassette was determined by restriction digestion to be opposite to the direction of transcription. SacII digestion of pFts′YEXSp resulted in a 2.8 kbp fragment that was ligated to a 5.8 kbp fragment of the original pFtsYEX9. Restriction digestion confirmed that the orientation of the 2.8 kbp fragment restored the entire ftsYEX operon with the kanamycin insertion and ORF4. The 6.0 kbp HindIII fragment of this plasmid, pFtsYE:KanX, was ligated into the HindIII site of pTSA29 (G. Phillips, in preparation), resulting in an ampicillin-resistant plasmid with a temperature-sensitive origin of replication. Using in vivo allelic exchange, the ftsE::kan allele was exchanged onto the chromosome and transduced into wild-type strain MG1655 using P1 transduction. This strain was then made recA by a P1 transduction of the recA::Tn10 allele from CSH126, resulting in RG60.

For whole-cell labelling, cells were grown in M9 medium supplemented with all amino acids except methionine and cysteine (Sambrook et al., 1989). The ftsE knock-out strain RG60 and its isogenic wild-type MG1655 were grown in LB medium supplemented with 1% NaCl and the appropriate antibiotics at 30°C unless mentioned otherwise.

Viability assay

Strain RG60 grows on LB medium supplemented with 1% NaCl, but not on LB medium without NaCl. To study the effects of different salts on growth, RG60 was cultured at 37°C in 1% NaCl LB, back-diluted in 0% NaCl LB broth and added to 0% NaCl top agar. Sterile 1/2-inch analytical paper discs (Schleicher & Schuell) were applied to the surface of the cooled agar. Saturated salt solutions (50 μl) were spotted onto the paper discs. The plates were incubated at 42°C for 16 h. The diameters of growth surrounding the discs were measured. Saturated sucrose, 80% glycerol, 0.1 M glycine betaine, 4.6% proline, 18.7% glutamate and 14.6% glutamine solutions were also assessed in this way for rescue of the growth of RG60 under restrictive conditions.

Mutagenesis and subcloning of ftsE and/or ftsX in expression vectors

To allow overexpression of FtsE, FtsX or FtsEX, the coding sequences were amplified by polymerase chain reaction (PCR) and provided with NcoI and HindIII recognition sequences at their 5′ and 3′ ends respectively. The amplified fragments were cut with NcoI and HindIII and cloned into pET21d. pET21d-FtsE-His and pET21d-FtsX-His were created similarly introducing six histidine codons at the 3′ end just upstream from the stop codon. Plasmid pDB1(Gill et al., 1986), which contains the complete ftsYEX operon, was used as a template in all PCRs. Plasmid pET21d-FtsEX-His was created by replacing the BsmI/PvuI fragment of pET21d-FtsEX with that of pET21d-FtsX-His.

Mutations were introduced in ftsE by overlap extension using PCR (Ho et al., 1989). Mutants were analysed by DNA sequencing using the Taq Dye Primer cycle sequencing kit and the 373A Automated DNA sequencer from Applied Biosystems.

Subcellular localization of FtsE and FtsX

Subcellular fractions were prepared essentially as described previously (Luirink et al., 1994). Inverted inner membrane vesicles (IMVs) were prepared as described previously (De Vrije et al., 1987). Peripheral cytoplasmic membrane proteins were extracted from cytoplasmic membrane vesicles with 1 M NaCl, 4 M urea and 0.2 M Na2CO3 as described previously (Cabelli et al., 1991).

Purification of (mutant) FtsE and FtsE/X complexes

For (mutant) FtsE purification, strain BL21(DE3) harbouring pLysS and pET21d-FtsE-His was cultured in LB to an optical density at 660 nm of 0.3 and induced with IPTG to a final concentration of 0.4 mM. After 3 h of induction, cells were harvested, resuspended in 50 mM Tris, 50 mM NaCl, 1 mM EDTA, pH 8.0 (buffer A), containing DNase (10 μg ml−1) and incubated for 20 min at room temperature. Cells were disrupted by sonication, and inclusion bodies were collected by low-speed centrifugation (10 min at 12 400 × g ). Inclusion bodies were washed with buffer A supplemented with 1% NP-40 and 1 M urea, sedimented by low-speed centrifugation (10 min at 12 400 × g ) and resuspended in 0.1 M NaH2PO4, 8 M urea, pH 8.0. Denatured proteins were incubated with Ni-NTA agarose for 2 h at 4°C. To remove aspecifically bound proteins, the Ni-NTA beads were washed in 0.1 M NaH2PO4, 8 M urea, pH 6.3, containing 30 mM imidazole, until the absorbance at 280 nm was below 0.01. His-tagged protein was eluted in the same buffer containing 250 mM imidazole.

To study the interaction between FtsE and FtsX, strain BL21(DE3) harbouring pLysS and pET21d-FtsEX or pET21d-FtsEX-His was grown to an optical density at 660 nm of 0.3 and induced with IPTG (0.4 mM final concentration) for 2 h. Cells were harvested and resuspended in 50 mM TEA, 250 mM sucrose, 2 mM EDTA, 1 mM DTT, pH 7.5. Cells were subjected to French press treatment twice (5000 psi), cell debris was removed by low-speed centrifugation (5 min at 12 400 × g ), and membranes were collected by high-speed centrifugation (90 min at 165 000 × g ). The membrane pellet was resuspended in PBS and stored at −80°C in aliquots until use. Membranes were solubilized in 1% dodecylmaltoside, 50 mM HEPES/KOH, 150 mM KCl, pH 7.9, for 1 h at 4°C. After centrifugation (10 min at 82 000 × g ) to remove insoluble material, the supernatant was incubated with Ni-NTA beads for 30 min at room temperature. Ni-NTA beads were washed twice with 50 mM HEPES/KOH, 150 mM KCl, pH 7.9, containing 0.1% dodecylmaltoside and 30 mM imidazole, and eluted with the same buffer containing 500 mM imidazole. Fractions were TCA precipitated and analysed by SDS–PAGE and immunoblotting using antiserum directed against FtsE or against FtsX.

Azido-ATP labelling

Purified (mutant) FtsE was labelled with 8-azido[α-32P]-ATP (ICN Biochemicals) in 50 mM HEPES/KOH, 50 mM KCl, pH 7.9, essentially as described previously (Delepelaire, 1994). 8-Azido[α-32P]-ATP was preincubated in the dark for 1 min. Labelling reactions, containing 3 μg of protein, were carried out in a 10 μl drop on parafilm at about 8 cm from a UVGL-100 lamp for 5 min. After irradiation, samples were mixed with 10 μl of SDS sample buffer containing 50 mM DTT and 1 mM unlabelled ATP. The mixtures were heated at 100°C for 5 min and analysed by SDS–PAGE and phosphorimaging.

Co-immunoprecipitation of FtsE and FtsX

BL21(DE3) harbouring pLysS and pET21d-FtsE, pET21d-FtsX or pET21d-FtsEX were grown at 37°C to the early logarithmic phase of growth in 5 ml of M9 medium lacking cysteine and methionine. Expression was induced with 1 mM IPTG for 15 min, and the cells were grown for a further 15 min in the presence of 200 μg ml−1 rifampicin to inhibit transcription by the host RNA polymerase. Cells were labelled for 10 min with [35S]-methionine (50 μCi ml−1), cooled on ice, collected and washed once with 50 mM HEPES/KOH, 50 mM KCl, pH 7.9. Cells were disrupted by sonication, and cell debris was removed by low-speed centrifugation (10 min at 2900 × g ). Crude membranes were collected by high-speed centrifugation (20 min at 270 000 × g ) and resuspended in 200 μl of 50 mM HEPES/KOH, 50 mM KCl, pH 7.9. Membrane fractions (90 μl) were treated with 1 mM DSP for 1 h on ice. Reactions were quenched for 10 min on ice by adding 0.1 volume of 1 M Tris-HCl, pH 8.0. Membrane solubilization and immunoprecipitation were carried out essentially as described by Duong and Wickner (1997). For denaturing immunoprecipitation, membranes were solubilized for 15 min on ice by adding 500 μl of 1% NP-40, 0.5% deoxycholate and 0.1% SDS in 50 mM Tris-HCl, 150 mM NaCl, pH 7.9. Insoluble material was removed by low-speed centrifugation (5 min at 2900 × g ), and specific antibodies were added to the supernatant fraction. After immunoprecipitation, samples were analysed by SDS–PAGE, using sample buffer containing 50 mM DTT, and phosphorimaging.


We thank J.-W. de Gier for carrying out the Lep assembly test, and G. Salmond for clones and strains. We are grateful to N. Harms and P. Scotti for critical reading of the manuscript. This work was supported in part by the Netherlands Organization for Scientific Research (to L.).