The Escherichia coli Tat system mediates Sec-independent export of protein precursors bearing twin arginine signal peptides. Formate dehydrogenase-N is a three-subunit membrane-bound enzyme, in which localization of the FdnG subunit to the membrane is Tat dependent. FdnG was found in the periplasmic fraction of a mutant lacking the membrane anchor subunit FdnI, confirming that FdnG is located at the periplasmic face of the cytoplasmic membrane. However, the phenotypes of gene fusions between fdnG and the subcellular reporter genes phoA (encoding alkaline phosphatase) or lacZ (encoding β-galactosidase) were the opposite of those expected for analogous fusions targeted to the Sec translocase. PhoA fusion experiments have previously been used to argue that the peripheral membrane DmsAB subunits of the Tat-dependent enzyme dimethyl sulphoxide reductase are located at the cytoplasmic face of the inner membrane. Biochemical data are presented that instead show DmsAB to be at the periplasmic side of the membrane. The behaviour of reporter proteins targeted to the Tat system was analysed in more detail. These data suggest that the Tat and Sec pathways differ in their ability to transport heterologous passenger proteins. They also suggest that caution should be observed when using subcellular reporter fusions to determine the topological organization of Tat-dependent membrane protein complexes.
In Escherichia coli, the movement of most periplasmic and membrane-bound proteins from their site of synthesis in the cytoplasm to their final subcellular location is mediated by the Sec protein translocation apparatus (Pugsley, 1993). Proteins are targeted to the Sec system by discrete, often amino-terminal regions of polypeptide termed signal sequences. Transport across or insertion into the membrane is by an aminoterminal to carboxy-terminal threading mechanism, in which the substrate protein is in an extended conformation. Recently, a second general protein transport system has been identified in the cytoplasmic membrane of E. coli and many other bacteria (Settles et al., 1997; Santini et al., 1998; Sargent et al., 1998; Weiner et al., 1998). Targeting to this second pathway is by means of signal sequences bearing a distinctive twin arginine-containing amino acid sequence motif (Berks, 1996; Stanley et al., 2000). In contrast to the Sec apparatus, this twin arginine (Tat) translocase functions to transport prefolded proteins (reviewed by Berks et al., 2000). The majority of Tat substrates in enteric bacteria are the peripheral membrane subunits of membrane-bound electron transfer complexes (Berks, 1996; Gennis and Stewart, 1996). Analogues of the bacterial Sec and Tat pathways are found in the thylakoid membrane of plant chloroplasts (Keegstra and Cline, 1999).
Gene fusions have been extensively exploited as a tool to probe the Sec protein transport pathway and to determine the topological organization of transmembrane proteins assembled by the Sec apparatus (Broome-Smith et al., 1990; Manoil et al., 1990; Schatz and Beckwith, 1990; Traxler et al., 1993; Danese and Silhavy, 1998). This approach involves fusing a protein fragment containing potential targeting determinants to a reporter protein with an activity that is sensitive to subcellular location. Consequently, the activity of the reporter protein indicates the compartment to which the protein has been targeted. The reporter proteins typically chosen have an easily defined phenotype in plate screens, for example antibiotic resistance or colour development, enabling the fusion construct to be used in genetic selection protocols. With the proven success of gene fusion as probes of the Sec pathway, it is of interest to explore carefully the application of gene fusions to the genetic analysis of the Tat pathway as well as to the topological organization of Tat-targeted enzyme assemblages.
A consideration of the different modes of protein translocation performed by the Tat and Sec pathways suggests that it would be dangerous to assume that fu-sion strategies and reporter proteins appropriate to one system will perform in an identical manner when applied to the other. In at least some cases, there appears to be a requirement that Tat substrate proteins are correctly folded before translocation can take place (Roffey and Theg, 1996; Santini et al., 1998; Halbig et al., 1999; Sanders et al., 2001). If this is a general property of the Tat pathway, then fusion proteins in which folding is disrupted (either by removal of the carboxy-terminal portion of the targeting domain or because addition of the reporter protein precludes correct folding of the targeting domain) will not be competent for Tat transport. A similar situation appertains if the reporter protein only folds after transport. For example, folding of the reporter enzyme E. coli alkaline phosphatase (PhoA) requires the formation of intramolecular disulphide bonds by a periplasmic enzyme system (Sone et al., 1997). It is also conceivable that the structure of the chimeric precursor protein might impede interactions between the twin arginine signal peptide and the transport mechanism. In the Tat mechanism, the substrate protein is presumably transported in a single step, implying an upper limit to the size of the fusion proteins that can be translocated. In contrast, the threading mechanism of the Sec pathway probably imposes no such constraint on the substrates. Finally, the most widely used reporter proteins are either substrates of the Sec pathway or cytoplasmic proteins and, thus, cannot automatically be assumed to be competent for translocation by the Tat mechanism. For example, the mature domains of these proteins could recruit chaperones and interfere with Tat targeting and transport.
Here, we show that reporter gene fusions to the periplasmically located subunits of the Tat-dependent and membrane-bound E. coli enzyme formate dehydrogenase-N (FDH-N) imply an incorrect topological organization of the enzyme. We then assess the behaviour of four commonly used subcellular reporter proteins (Bla, β-lactamase; Cat, chloramphenicol acetyltransferase; LacZ, β-galactosidase; and PhoA) when targeted to the Tat pathway. We conclude that heterologous proteins cannot necessarily be assumed to be acceptable Tat substrates and that, in some cases, the transport behaviour is fusion specific.
Results and discussion
The formate-oxidizing subunit of formate dehydrogenase-N is located at the periplasmic face of the cytoplasmic membrane
In the main portion of this study, we have sought to determine whether standard reporter fusion constructs can be applied to the topological mapping of membrane-bound enzyme complexes assembled by the Tat path-way. Our model protein in these studies was the E. coli nitrate-inducible respiratory formate dehydrogenase (FDH-N) encoded by the fdnGHI operon. Respiratory formate dehydrogenases, including FDH-N, oxidize formate to carbon dioxide concomitant with reduction of membrane quinone to quinol. Transfer of each pair of electrons in this reaction is linked to the net translo-cation of two protons from the cytoplasm to the periplasm (Kröger, 1975; Jones, 1980), and the respiratory formate dehydrogenases therefore function as coupling sites.
Localization of FDH-N to the cytoplasmic membrane requires a functional Tat pathway (Bogsch et al., 1998; Sargent et al., 1998; 1999), and the organization of the enzyme is typical of membrane protein complexes as-sembled by the Tat system. In such proteins, two cofactor-binding peripheral membrane subunits complex with a third integral membrane subunit containing the site of interaction with the membrane quinone/quinol pool. Association of the peripheral subunit pair with the integral membrane protein requires a Tat signal peptide located on one of the peripheral subunits, whereas the integral membrane subunit is thought to be inserted into the membrane by the standard SRP/Sec pathway. The peripheral membrane subunits of FDH-N are FdnG and FdnH (Berg et al., 1991a). Formate is oxidized at a molybdopterin cofactor in the FdnG subunit. FdnH is a ferredoxin subunit, and the integral membrane FdnI subunit is dihaem cytochrome bFdn556.
In order to use E. coli respiratory formate dehydrogenase-N (FDH-N) in our study, we needed to be certain of the topological orientation of the enzyme. However, the literature contains conflicting reports on the location of the active site subunits (FdnG homologues) in orthologous formate dehydrogenases. A periplasmic location has previously been inferred for the formate-oxidizing site of the respiratory formate dehydrogenase of Wolinella succinogenes based on the observation that W. succinogenes can oxidize external formate at circumneutral pH even though the cytoplasmic membrane is formate impermeant under these conditions (Kröger et al., 1980). In contrast, a gene fusion study of E. coli respiratory formate dehydrogenase-O (FDH-O) assigned a cytoplasmic location to the formate-oxidizing subunit (Benoit et al., 1998). To resolve this discrepancy, we undertook our own biochemical analysis of the subcellular location of the FdnG subunit.
One indicator of periplasmic localization is Tat signal peptide cleavage. Inspection of the fdnG sequence suggests that a 33-residue twin arginine signal peptide is present culminating at the predicted cleavage site Ala-Leu-Ala ↓ Gln-Ala-Arg-Asn-Tyr. We determined the amino-terminal sequence of the FdnG subunit to be Gln-Ala-Arg-Asn-Tyr. Thus, the predicted Tat signal peptide was post-translationally removed. The twin arginine signal peptide is likewise cleaved from the homologous FdhA protein of W. succinogenes formate dehydrogenase (Bokranz et al., 1991). As the active site of the E. coli LepB signal peptidase is located at the periplasmic side of the cytoplasmic membrane (Dalbey et al., 1997), the observed processing of the signal peptide indicates that at least the amino-terminal portion of FdnG reaches the periplasmic compartment. The amino-terminus of the FdnH subunit was determined to be Ala-Met-Glu-Thr-Gln, indicating that only the initiator methionine residue was post-translationally removed, again congruent with studies of the homologous FdhB protein of W. succinogenes formate dehydrogenase (Bokranz et al., 1991).
To obtain further evidence of a periplasmic localization for FdnG, we constructed a mutant strain with a complete in frame chromosomal deletion of the fdnI gene. Both fdnI+ and ΔfdnI strains were grown anaerobically to midexponential phase on glucose plus nitrate medium to induce FDH-N synthesis. Immunoblotting revealed intact FdnG polypeptide in both strains. Upon subcellular fractionation, FdnG protein was found, as expected, predo-minantly associated with the membrane fraction of the fdnI+ strain (Fig. 1). In contrast, FdnG protein was located almost exclusively in the periplasmic fraction of the ΔfdnI strain (Fig. 1). The small proportion of FdnG antigen that co-sedimented with the crude membrane fraction in the ΔfdnI mutant was completely removed by a salt wash, whereas FdnG was still present in the membranes from the fdnI+ strain after salt treatment (Fig. 1). We conclude from this experiment that, in intact FDH-N, the FdnG polypeptide is located on the periplasmic face of the cytoplasmic membrane.
The formate dehydrogenase activities of the parental and ΔfdnI strains were measured using phenazine methosulphonate (PMS) as electron acceptor. Whereas cells of the fdnI+ strain exhibited strong formate dehydrogenase activity, the ΔfdnI strain produced no detectable formate:PMS oxidoreductase activity (data not shown). The failure of the periplasmically located FdnG to catalyse formate:PMS oxidoreductase activity in the ΔfdnI mutant could result from greater lability of the metallocofactors in the partially assembled enzyme but may, alternatively, indicate that the PMS-interacting site is located on the FdnI and/or the FdnH subunit. In this context, it is notable that the FDH-N protein that accumulates in the cytoplasm of tat mutant strains is also enzymatically inactive in the PMS-linked assay (Bogsch et al., 1998; Sargent et al., 1998; 1999).
A model for the topological organization of FDH-N taking account of the experimental data presented here is given in Fig. 2A. Given that the protons derived from formate oxidation by respiratory formate dehydro-genases are released at the periplasmic side of the membrane and that cytoplasmic protons are used for quinone reduction (Geisler et al., 1994), net proton translocation by these enzymes must arise from the electrogenic movement of electrons from the periplasmic to the cytoplasmic side of the membrane (Fig. 2A). The two haem groups located in the integral membrane subunit are expected to mediate the transmembrane movement of electrons required by this scheme (Fig. 2A; Berks et al., 1995).
The subcellular location of the FdnH protein in the ΔfdnI mutant was not addressed in our experiments. Even though FdnH lacks a cleavable signal peptide (above), this subunit is thought to carry electrons from FdnG to FdnI and should therefore be located at the same side of the membrane as FdnG (Fig. 2A). This would imply that FdnH is carried across the membrane in complex with FdnG by a mechanism analogous to that already described for [NiFe]-hydrogenase biosynthesis (Rodrigue et al., 1999). This predicted complex formation suggests that FdnH should co-localize with FdnG to the water-soluble periplasmic fraction of the ΔfdnI mutant even though FdnH is probably finally secured to the cytoplasmic membrane in the intact FDH-N complex via a carboxy-terminal anchor (Fig. 2A; Berg et al., 1991a).
Topological mapping of formate dehydrogenase-N using standard genetic fusion techniques leads to an erroneous structural model of the enzyme
To test the utility of genetic fusion technology in the topological mapping of Tat-dependent proteins, we undertook a marker fusion analysis of formate dehydrogenase-N. For this study, we adopted the widely used strategy of constructing complementary fusions to the periplasmic marker enzyme PhoA and the cytoplasmic marker enzyme LacZ (Manoil, 1990). PhoA contains two intra-molecular disulphide bonds that are essential for both enzymatic activity and stability (Sone et al., 1997). As the formation of disulphide bonds in wild-type E. coli cells occurs in the periplasmic compartment, PhoA is only active when targeted to the periplasm. LacZ is a cytoplasmic protein containing sequences that hinder transport by the Sec apparatus (Lee et al., 1989) usually leading to a misfolded, and thus inactive, protein. The formation of disulphide bonds within the LacZ polypeptide as it emerges into the periplasm has been implicated in this phenomenon (Bardwell et al., 1991). LacZ activity is therefore normally only observed from fusions targeted to the cytoplasm. Defined gene (translational) fusions were constructed between each of the fdn structural genes and phoA and lacZ separately. The resulting plasmid-borne gene fusions are expressed under the control of the native fdnG promoter; constructs with fusions to fdnH include an intact upstream fdnG gene, and those with fusions to fdnI include intact upstream fdnGH genes. Plasmids bearing these fusion constructs were introduced into a pcnB background to ensure a low plasmid copy number (Lopilato et al., 1986). Reporter enzyme activities were assessed from cultures grown anaerobically in the presence of nitrate to induce expression from the fdnG promoter fully (Li and Stewart, 1992).
The integral membrane protein FdnI is predicted to comprise four transmembrane helices organized such that the amino-terminus of the protein is located at the cytoplasmic side of the membrane (Berks et al., 1995). To test this model, we chose to construct both phoA and lacZ gene fusions at positions located successively in each of the five predicted extramembranous regions (Boyd et al., 1993). Junctions are at the distal portions of the amino-terminal cytoplasmic tail (Asp-14), the periplasmic loop between helices 1 and 2 (Gln-51), the cytoplasmic loop between helices 2 and 3 (Ala-111), the periplasmic loop between helices 3 and 4 (Leu-152) and the car-boxyl-terminal cytoplasmic tail (Glu-215). The relative activities of the resultant PhoA and LacZ fusions were entirely consistent with the model (Table 1; Fig. 2A). Thus, as found for many other Sec-assembled integral membrane proteins, fusions of PhoA and LacZ to FdnI gave apparently reasonable topological information. Experiments with Bla gene fusions yielded an identical topologi-cal model for the homologous FdoI protein (Benoit et al., 1998).
Table 1. LacZ and PhoA enzyme activities expressed from Φ(fdn–lacZ) and Φ(fdn–phoA) gene fusions.a
Plasmid-bearing cultures of strain VJS5833 (ΔlacΔphoA pcnB1) were grown anaerobically in TYEGN medium to mid-exponential phase and harvested for enzyme assays.
fdn codon at the junction with lacZ or phoA.
Activities are expressed in Miller units. Values are averaged from at least two independent experiments.
PhoA and LacZ fusions were also constructed at two positions each in the Tat-dependent FdnG and FdnH subunits, as described in Experimental procedures. Fusions to the FdnG polypeptide were made at codon Lys-44 in fdnG, 11 codons downstream of the leader peptidase cleavage site, and at codon Asp-792, about 78% into the 1016 codon gene. Fusions to the FdnH polypeptide were made just after codons Leu-257 and His-292, which are located on either side of the predicted transmembrane helix.
Results from enzyme assays of cultures expressing these fusions are shown in Table 1. All four fusion junctions yielded high-level LacZ activity and negligible PhoA activity. In a previous study, Φ(fdnG–lacZ) gene fusions constructed at codons Ala-212 and Asp-493 also yielded high-level LacZ activity (Berg et al., 1991b). [Conversely, a Φ(fdnG–lacZ) gene fusion constructed at codon Leu-32, just proximal to the leader peptidase cleavage site, expressed very low levels of LacZ activity (Li and Stewart, 1992)]. Furthermore, in preliminary studies, an in frame TnphoA insertion (Manoil and Beckwith, 1985) at fdnG codon Gly-65 expressed undetectable PhoA activity (S. B. Williams and V. Stewart, unpublished observations). The conventional interpretation of these results would be that both the FdnG and the FdnH polypeptides are localized to the cytoplasm, the exact opposite of the conclusion drawn from the studies of leader peptide cleavage and periplasmic FdnG accumulation described above. Indeed, analysis of Bla fusions to the homologous FdoG and FdoH proteins of E. coli FDH-O were interpreted to reveal a cytoplasmic location for these subunits (Benoit et al., 1998). In summary, analysis of three conventional topology reporters (Bla, LacZ and PhoA) resulted in predictions that are directly contradicted by the results of the biochemical studies presented above.
Investigating the behaviour of Tat-targeted reporter proteins
The unexpected pattern of whole-cell activities measured for FdnG–PhoA and –LacZ fusions implies that either the addition of a Tat signal peptide to these reporter enzymes is insufficient to allow export by the Tat pathway or the enzymes themselves are incompatible with Tat transport. In order to investigate this phenomenon in more detail, we undertook a biochemical analysis of the behaviour of different passenger proteins fused to the FdnG twin arginine signal peptide. As well as PhoA and LacZ, we selected Bla and Cat as passenger proteins for use in these studies. The reporter proteins Bla and Cat provide genetic selections for fusion proteins that are, respectively, exported to the periplasm (the site of β-lactam action) or retained in the cytoplasm (where chloramphenicol is inactivated by acetylation) (Zelazny and Bibi, 1996; reviewed by Broome-Smith et al., 1990). All four types of fusion were constructed at fdnG codon Lys-44, resulting in hybrid proteins containing the FdnG signal peptide together with the first 11 residues of the mature protein. In the experiments described in this, and indeed all other, sections of this report, the quality of the subcellular fractionation was confirmed by an analysis of marker enzyme distribution as detailed in the Experimental procedures.
Alkaline phosphatase activity was undetectable in cells harbouring the Φ(fdnG–phoA)K44 fusion plasmid (Table 1). In addition, we were unable to detect PhoA-immunoreactive material in these cells (data not shown), possibly because cytoplasmically targeted PhoA is unfolded and prone to proteolytic degradation. PhoA protein and alkaline phosphatase activity were also undetectable in tat mutant strains containing the Φ(fdnG–phoA)K44 construct. This observation argues against the possibility that the fusion is transported by the Tat system and subsequently proteolysed in the periplasm as, if this were the case, one might expect the fusion protein to accumulate in the cytoplasmic fraction of the tat mutants.
The Φ(fdnG–lacZ)K44 construct expressed active LacZ enzyme (Table 1). Upon subcellular fractionation, both LacZ activity (Table 2) and LacZ-immunoreactive material (data not shown) were found exclusively in the cytoplasmic fraction. Thus, the LacZ fusion was either not exported or was unstable in the periplasm after export. Targeting of LacZ to the Sec pathway can result in jamming of the Sec translocase (Ito and Beckwith, 1981), and it was therefore conceivable that an analogous phenomenon was preventing export of the Tat-directed LacZ fusion in our experiments. Expression of the Φ(fdnG–lacZ)K44 fusion had no effect on the levels of FDH-N assembled in the membrane or on targeting of the Tat substrate trimethylamine N-oxide reductase (TorA) to the periplasm (data not shown), suggesting that, under these experimental conditions, the fusion did not block the Tat pathway. However, expression of the Φ(fdnG–lacZ)K44 fusion in a pcnB+ background (in which the encoding plasmid is present at high copy number) led to a substantial (80%) drop in periplasmic TorA levels. Taken together, these observations suggest that, although the hybrid protein did not jam the Tat translocase, it did compete with endogenous substrates for access to the Tat transport system. A corollary is that the Tat system must recognize the Tat signal peptide of the Φ(fdnG–lacZ)K44 fusion and, therefore, that the failure to transport LacZ is unlikely to be a consequence of proteolysis or other occlusion of the Tat signal peptide on the fusion.
Table 2. Enzymatic activities of fdnG translational fusions after subcellular fractionation.
Activities of fdnG fusion constructs
Cells were cultured anaerobically on glucose- and nitrate-containing medium to early stationary phase (OD600 = 2.5–3.0) and fractionated as described in Experimental procedures. Activities are given as μmol of substrate hydrolysed min–1 for the volume of the subcellular fraction obtained from 1 g wet weight of cells. The substrates used in these assays are listed in Experimental procedures. All assays were performed on three independent cultures with errors representing SEM. The figures in brackets give the percentage of the total enzymatic activity associated with each subcellular compartment.
ND indicates that the experiment could not be performed (see text).
The activity is that of lysed spheroplasts (i.e. cytoplasmic and membrane fractions combined).
Fractionation of cells expressing the Φ(fdnG–bla)K44 fusion revealed that around 80% of the total Bla activity accumulated in the periplasmic fraction (Table 2). Immunoblotting confirmed that the Bla protein was present in the periplasmic fraction (Fig. 3). The FdnG Tat signal peptide is therefore capable of directing Bla export. It proved difficult to test whether this export was mediated by the Tat system, as tat mutant strains lysed when expression of the Φ(fdnG–bla)K44 fusion was induced in liquid culture. However, the Φ(fdnG–bla)K44 fusion conferred ampicillin resistance (100 μg ml–1) to a Tat+ strain, but not to a ΔtatC mutant, when plated on solid medium. These experiments indicate that the Φ(fdnG–bla)K44 fusion was transported to the periplasm via the Tat pathway. The analogous fusion to the FDH-O orthologue [Φ(fdoG–bla)R36; Benoit et al., 1998] conferred a moderate level of resistance to ampicillin [minimal inhibitory concentration (MIC) = 25 μg ml–1], which was higher than that for any other Φ(fdoG–bla) fusion tested but about 10-fold lower than that for active Φ(fdoI–bla) fusions.
Upon subcellular fractionation of cells expressing a Φ(fdnG–cat)K44 fusion, the bulk of both enzymatic activity and Cat protein accumulated in the periplasmic fraction (Table 2 and Fig. 4). Thus, Cat, like Bla, was targeted for export by the FdnG signal peptide. Interestingly, al-though the spheroplasts retained substantive Cat activity, this cytoplasmically located Cat was undetectable by immunoblotting. This suggests that the exported Cat has a substantially lower specific activity than the material that is retained in the cytoplasm. Transport of the fusion was abolished in a ΔtatAΔtatE background (Table 2), demonstrating that Cat was being transported by the Tat pathway.
Taken together, these data suggest that, although the FdnG signal peptide was able to target the heterologous Bla and Cat proteins to the Tat transporter, some other passenger proteins, including LacZ and PhoA, may be inherently incompatible with the Tat system.
Behaviour of marker proteins targeted to the Tat system by the signal peptide of the E. coli SufI protein
To test whether the observed transport behaviour of the FdnG signal peptide fusions was signal peptide specific or represented a more general feature of the interaction of the passenger proteins with the Tat system, we individually fused PhoA, LacZ and Bla directly behind the 27-amino-acid signal peptide of the E. coli SufI protein. SufI is a water-soluble and, therefore, unambiguously periplasmic Tat substrate (Stanley et al., 2000).
The Φ(sufI–phoA)A27 fusion was enzymatically active, and the PhoA protein was detected almost exclusively in the periplasmic fraction (data not shown). However, the fusion was still active in a ΔtatC background, indicating that the observed export did not require the Tat system. We note that this Tat-independent transport occurs even though the SufI signal peptide contains a c-region basic residue, a feature that has been shown to inhibit the interaction of other Tat signal peptides with the Sec apparatus (Bogsch et al., 1997; Cristóbal et al., 1999).
Subcellular fractionation of E. coli cells expressing the Φ(sufI–lacZ)A27 fusion showed that >90% of the LacZ activity localizes to the cytoplasmic fraction in either a Tat+ or a Tat– background (Table 3). Thus, within the experimental error of this study, the SufI signal peptide does not direct Tat-dependent export of LacZ. The low level of expression of this fusion precluded the detection of LacZ-immunoreactive material by immunoblotting.
Table 3. Enzymatic activities of sufI–lacZ translational fusions after subcellular fractionation.
Activity of Φ(sufI–lacZ)A27 fusion (nmol ONPG hydrolysed min–1 g–1)
Cells were cultured aerobically on LB-glucose medium to early stationary phase and fractionated as described in Experimental procedures. Activities are given as nmol of substrate hydrolysed min–1 for the volume of the subcellular fraction obtained from 1 g wet weight of cells.
Escherichia coli strain MC4100 transformed with the Φ(sufI–bla)A27 fusion plasmid was resistant to 125 μg ml–1 ampicillin on plates, whereas a ΔtatC derivative of MC4100 containing the same plasmid was ampicillin sensitive. Thus, the SufI signal peptide, like the FdnG signal peptide, is capable of directing Tat-specific export of Bla. The low level of expression of the Φ(sufI–bla)A27 fusion from the sufI promoter precluded accurate analysis of the subcellular localization of Bla by enzyme assay or immunoblotting.
In summary, the SufI signal peptide fusions provide examples of the same signal peptide directing Tat-dependent export, Tat-independent export or no export, depending on the passenger protein. Hybrids in which PhoA was fused to the signal peptides of FdnG and SufI showed different transport behaviour resulting, at least in part, from differences in the ability of these Tat signal peptides to interact with the Sec system.
Behaviour of reporter fusions within the mature sequence of a Tat substrate
The experiments described above show that hybrid proteins in which Cat or Bla are fused close behind the Tat signal peptide of FdnG are capable of directing translocation of the passenger protein to the periplasm. We decided to investigate whether there was a difference in transport behaviour if the reporter proteins were fused further into the FdnG protein leaving a region of misfolded protein between the Tat signal peptide and the reporter protein. We chose to make Cat and Bla fusions after FdnG codon 792. This construct removes ≈ 22% of the FdnG polypeptide and would destroy the fourth consensus domain found in homologous molybdopterin cofactor-binding structures (Boyington et al., 1997). Strains expressing the fusions would also lack stoichiometric amounts of the partner FdnH subunit. However, cofactor binding to the truncated FdnG protein might still be feasible as, by analogy with related molybdopterin-binding structures, the deleted portion of FdnG provides few contacts with the molybdopterin cofactor, whereas all the iron–sulphur cluster-binding domain I is still present.
Expression of the Φ(fdnG–bla)D792 construct resulted in an accumulation of β-lactamase activity in the periplasmic fraction (Table 2). However, the periplasmic localization of the β-lactamase activity was unaffected in a ΔtatAΔtatE mutant, indicating that export by a non-Tat pathway was occurring (Table 2). Immunoblotting showed that the periplasmically located Bla molecules have a molecular mass similar to that of the native mature protein, whereas the cytoplasmically located Bla fusion precursor protein is subject to proteolytic degradation (Fig. 3). It is not clear, therefore, whether the species that is transported is the full-length fusion protein or a truncated fragment.
No chloramphenicol acetyltransferase activity could be detected in extracts of cells expressing the Φ(fdnG–cat)D792 fusion. However, the hybrid protein was detected in the cytoplasmic fraction by immunoblotting. This pro-tein was present primarily as a high-molecular-mass species that showed some degradation to a form with an electrophoretic motility close to that of wild-type Cat (Fig. 4). As Cat activity relies on the formation of a homotrimeric structure (Leslie et al., 1988), the lack of activity exhibited by the fusion protein may indicate that the Cat domains of the hybrid protein are unable to oligomerize.
In summary, these experiments demonstrate that a reporter protein that is targeted to the Tat pathway when fused to a Tat signal peptide alone is not necessarily transported when the fusion junction is within the mature portion of a multisubunit metalloprotein using the same signal peptide.
The DmsAB subunits of E. coli DMSO reductase are located at the periplasmic face of the cytoplasmic membrane
Escherichia coli enzyme dimethyl sulphoxide (DMSO) reductase is a three-subunit membrane-bound enzyme in which the two peripheral membrane subunits, DmsAB, are homologous to the FdnGH subunits of FDH-N, and the DmsC subunit is an integral membrane protein (Weiner et al., 1992). The site of DMSO reduction is a molybdopterin cofactor in the DmsA protein. Targeting of the DmsAB subunits into the mature enzyme complex has been shown to require a functional Tat system as well as the processing of the Tat signal peptide found on the DmsA subunit (Weiner et al., 1998; Sambasivarao et al., 2000). Notwithstanding the involvement of the Tat system in the biosynthesis of DMSO reductase, it has been suggested that the DmsAB subunits are located at the cytoplasmic rather than the periplasmic face of the membrane and that DMSO reductase is the archetype of a subclass of Tat substrates for which the Tat apparatus is involved solely in membrane targeting and not in protein translocation (Weiner et al., 1998; Sambasivarao et al., 2000; 2001). The inferred cytoplasmic localization of DmsAB is based in part on the results of certain types of biochemical analysis, but also on the failure to observe active phoA fusions to either dmsA or dmsB (Sambasivarao et al., 1990), an approach that has been shown here to be flawed. The subcellular localization of DmsA and DmsB has also been probed by genetically removing the integral membrane DmsC subunit. An initial report using this strategy found that DmsA and DmsB were located in the periplasm (Weiner et al., 1998), suggesting that DmsAB are normally found at the peri-plasmic side of the membrane (Sargent et al., 1998). However, in subsequent work using the same approach, only cytoplasmic DmsA and DmsB were detected (Sambasivarao et al., 2001).
Establishing the true topological organization of DmsAB is of fundamental importance in understanding the physiological role and mechanism of the Tat pathway as, if DmsAB are indeed located at the cytoplasmic face of the membrane, then there is a subset of Tat substrates for which the pathway has only a membrane targeting, as opposed to a transport, function (Weiner et al., 1998). We have therefore undertaken an independent assessment of the effect of expressing DmsAB in the absence of DmsC.
DMSO reductase is encoded by the dmsABC operon. A strain, TP051, was constructed containing a complete in frame deletion of dmsC together with the addition of a hexahistidine coding sequence to the 3′ end of dmsB. Both TP051 and the parental strain LCB628 were cultured anaerobically in the presence of DMSO to induce expression from the dmsA promoter. The cells were fractionated, and DMSO reductase activity was measured using the non-physiological electron donor benzyl viologen radical (BV•+). DMSO reductase activity in the parental strain was, as expected, mainly (84%) present in the membrane fraction (Table 4). In contrast, the BV•+-dependent DMSO reductase activity was found predominantly (69%) in the periplasm of the ΔdmsC strain (Table 4). A control fractionation of a ΔdmsABC strain suggested that the residual DMSO reductase activity found in the cytoplasmic and membrane fractions of the ΔdmsC mutant does not arise from DmsAB (Table 4). As the molybdopterin cofactor in DmsA is the site of DMSO reduction, these activity data indicate that the DmsA subunit is localized to the periplasm in the absence of the DmsC membrane anchor protein.
Table 4. DMSO reductase activity localizes to the periplasm in a ΔdmsC mutant.
Cells were cultured in CR medium with 0.5% glycerol, 0.5% fumarate, 0.4% DMSO. Activities were measured with TMAO as the electron-accepting substrate.
The subcellular localization of the DmsB subunit in the ΔdmsC mutant was also probed using antibodies directed against the hexahistidine tag that we had engineered onto DmsB in the mutant. The tagged DmsB protein was found predominantly in the periplasmic fraction (Fig. 5). This observation indirectly supports a periplasmic location for DmsA, as DmsB does not have a signal peptide and would need to be transported in complex with DmsA (Berks, 1996).
Taken together, these experiments show that both DmsA and DmsB are directed to the periplasm of a ΔdmsC mutant. The most reasonable interpretation of these data is that DmsAB are located at the periplasmic face of the membrane in the native DMSO reductase complex.
Peripheral membrane subunits of other membrane-bound Tat complexes have been found to be released into the periplasm upon removal of the membrane anchor protein (FdnG, this work; Krafft et al., 1995; Bernhard et al., 1997; Gross et al., 1998a). In these studies, as in our experiment, the membrane subunit was eliminated by an in frame deletion of the encoding chromosomal gene. This approach ensures that expression of the peripheral membrane proteins is maintained at physiological levels. In contrast, in the earlier study of DmsAB localization (Sambasivarao et al., 2001), dmsAB were expressed from a multicopy plasmid in a complete dms deletion background. It is possible that the Tat system cannot cope with such high levels of DmsAB expression and that this led to the observed cytoplasmic accumulation of DmsAB.
To test the subcellular location of DmsAB further, we took advantage of the observation that DMSO reductase will use trimethylamine N-oxide (TMAO) as an alternative substrate to DMSO both in vitro and in vivo (Weiner et al., 1988; 1992). In E. coli and other bacteria, TMAO is predominantly reduced by periplasmic water-soluble enzyme systems. However, in an E. coli strain devoid of peri-plasmic TMAO reductase (TorA), the membrane-bound DMSO reductase will still support anaerobic growth with TMAO as the sole respiratory electron acceptor (Weiner et al., 1992). TMAO falls into a class of organic compounds for which no completely uncharged Lewis structure can be written. Thus, TMAO is a relatively polar molecule, and we reasoned that the cytoplasmic membrane bilayer might have a very limited permeability to this species. With this in mind, we devised an experiment to test the cellular location of Dms-dependent TMAO reduction in whole cells using the membrane-permeant molecule BV·+ (Jones and Garland, 1977) as the electron donor. To avoid interference from the periplasmic TMAO reductase (TorA), the experiments were carried out in a tor background.
BV·+-dependent TMAO reductase activity can be measured in intact cells of a tor strain, and this activity is essentially unchanged after cell disruption (Table 5). This TMAO reductase activity is abolished in a tor dms double mutant, indicating that the activity can be attributed to DMSO reductase (Table 5).
Table 5. Assessing the subcellular location of DMSO reductase using a membrane-impermeant enzyme substrate.
Cells were grown in CR medium with 0.5% glycerol, 0.5% fumarate, 0.4% TMAO, and activities were measured with TMAO as substrate. Cells were broken by sonication in a nitrogen-saturated buffer comprising 50 mM Tris-HCl, pH 7.5, 2.5 mM Na2EDTA, 5 mM 2-mercaptoethanol and Roche protease inhibitor cocktail.
LCB628 (tor–, dms+, tat+)
GB2301 (tor–, dms–, tat+)
GB2303 (tor–, dms+, tatB)
DSS401 (tor+, dms–, tat+)
GBKK22 (torA [R11K,R12K], dms–, tat+)
All workers are agreed that DmsA and DmsB accu-mulate in the soluble cytoplasmic fraction of tatB mutants (Chanal et al., 1998; Weiner et al., 1998; Sargent et al., 1999; Sambasivarao et al., 2001). Negligible BV·+-dependent TMAO reductase activity was observed in whole cells of a tor tatB double mutant (Table 5). However, upon cell lysis, TMAO reductase activity similar to that of the Tat+ Tor– strain is measured (Table 5). This clearly shows that TMAO is normally membrane impermeant, as the substrate can only access the unequivocally cytoplasmically located DmsAB when the membrane barrier is broken.
Our observation that the inner membrane provides a permeability barrier to TMAO is confirmed by experiments with strains synthesizing TorA but not Dms. BV·+-linked TMAO reductase activity is easily measured for intact cells of a strain in which the water-soluble TorA is targeted to its normal periplasmic location (Table 5). In contrast, if the twin arginine motif of the signal peptide of the TorA precursor is mutated such that enzymatically active TorA is retained in the cytoplasm (strain GBKK22 in which the conserved arginine pair of the signal peptide is substituted by two lysine residues; Buchanan et al., 2001), then substantive BVÑ+-dependent TMAO reductase activity is only measured for broken rather than intact cells (Table 5).
The demonstration that TMAO is poorly membrane permeant, together with the observation that the full BVÑ+-dependent TMAO reductase activity of DMSO reductase can be measured in intact Dms+ Tor– cells, indicates that Dms-dependent TMAO reduction is a periplasmic process. The new biochemical evidence presented here, together with spin-coupling studies (Weiner et al., 1993; Rothery and Weiner, 1996) and the invalidation of the PhoA fusion approach (above), overwhelmingly point to a location of the DmsAB subunits of E. coli DMSO reductase at the periplasmic face of the inner membrane. As a consequence, the previous view that the Tat system can function in membrane targeting alone is no longer sustainable. A model for the topological organization of DMSO reductase taking account of the data presented here is shown in Fig. 2B.
We have examined the behaviour of gene fusions between the signal peptides of two Tat-dependent proteins, respiratory formate dehydrogenase-N and SufI protein, and four different passenger proteins. We found that a Tat signal peptide can direct the Tat-dependent export of Bla or Cat but not of LacZ or PhoA. A consideration of the three-dimensional structures of the four marker proteins (Leslie et al., 1988; Kim and Wyckoff, 1991; Jelsch et al., 1993; Jacobson et al., 1994) indicates that the signal peptide should be accessible at the surface of the folded molecule even if, as expected for Cat and LacZ, the protein has oligomerized. This suggests that the failure to transport LacZ or PhoA through the Tat pathway resulted from an innate incompatibility between these proteins and the Tat system. In the case of LacZ, it is probable that the fully assembled LacZ tetramer is too large to be transported by the Tat machinery, as this molecule is bigger (minimal cross-sectional dimensions of ≈ 90 Å by 130 Å) than known substrates of the E. coli Tat system (diameters of about 60–70 Å; Berks et al., 2000). The inability of the Tat system to translocate LacZ is likely to be a general phenomenon, as this protein was also not exported in Zymomonas mobilis when fused to the signal peptide of the Tat substrate glucose-fructose oxidoreductase (Halbig et al., 1999). As observed here for the FdnG and SufI chimeras, both active (Keon and Voordouw, 1996) and inactive (Sambasivarao et al., 1990; Reinartz et al., 1998) fusions have been reported between PhoA and Tat substrates. As the activity of the Φ(sufI–phoA)A27 fusion was shown to be Tat independent, these experiments suggest that PhoA is incompatible with the Tat pathway in most, and possibly all, fusion contexts. This incompatibility may be related to the inability of PhoA to fold in the cytoplasm, unlike Bla (Plückthun and Knowles, 1987) or the native cytoplasmic proteins Cat and LacZ.
The observation that Cat can be translocated by the Tat pathway is intriguing, as there are reports that this protein is not transported when targeted to the Sec system in E. coli or Bacillus subtilis (Gentz et al., 1988; Chen and Nagarajan, 1993). Cat may therefore be the first example of a cytoplasmic protein that is competent for transport through the Tat but not the Sec pathway. We note, however, that green fluorescent protein is folded and active when targeted to the periplasm via the Tat pathway, but is misfolded and inactive when it has been translocated by the Sec system (Feilmeier et al., 2000; Santini et al., 2001; Thomas et al., 2001).
One goal of the research reported here was to investigate the utility of reporter enzymes in the genetic analysis of the Tat pathway. In principle, the Φ(fdnG–cat)K44 fusion could be used to select for tat mutants, as the mutant strains would become chloramphenicol resistant. However, under the experimental conditions used here, some active Cat fusion protein is retained within the cytoplasm at steady state (Table 2), and this is sufficient to render wild-type cells chloramphenicol resistant (data not shown). Further development of the system would therefore be necessary before Cat fusions could be applied to the genetic analysis of the Tat pathway. Bla was also found to be specifically transported by the Tat pathway. As a consequence, the ampicillin resistance conveyed by the Φ(fdnG–bla)K44 and Φ(sufI–bla)A27 fusions described here could provide a positive selection for suppressors of tat mutations. There are, however, potential problems with the general use of Bla as a reporter for Tat activity, because transport of the fusions is not always totally Tat dependent. For example, the Φ(fdnG–bla)D792 fusion exhibited Tat-independent translocation, whereas the Tat signal peptide–Bla hybrid described by Nivière et al. (1992) was still transported when one of the signal peptide consensus arginine residues was mutated to an acidic residue, even though only lysine substitutions have ever been observed to allow export of native Tat substrates (Dreusch et al., 1997; Gross et al., 1999; Halbig et al., 1999; Stanley et al., 2000; Hinsley et al., 2001). An additional difficulty for certain types of experiment is that expression of the Φ(fdnG–bla)K44 fusion in a tat background liquid culture led to cell lysis.
Bacterial strains, plasmids and growth conditions
The strains and plasmids used in this work are summarized in Table 6. During all genetic manipulations, and during the biochemical analysis of sufI fusions, E. coli strains were grown aerobically in Luria–Bertani medium (Sambrook et al., 1989), and antibiotics were added at the concentrations listed by Sargent et al. (1998). In order to ensure synthesis of proteins from the fdnG promoter, biochemical studies were carried out on cells cultured anaerobically in the modified Cohen and Rickenberg medium described by Sargent et al. (1998) supplemented with 0.4% (w/v) glucose and 0.4% (w/v) KNO3. Cultures for the experiment presented in Table 1 were grown anaerobically in TYEGN medium, which consists of Vogel–Bonner defined medium (E salts; Maloy et al., 1996) supplemented with 0.8% tryptone, 0.5% yeast extract, 10 mM glucose and 40 mM NaNO3.
Strain NRS-7 was constructed as follows. A 654 bp polymerase chain reaction (PCR) product covering the region directly upstream of fdnI and the first 8 bp of the fdnI coding region was amplified from MC4100 DNA using the primers 5′-GCGCtctagaGGTTTACGATAACCCCGCCG-3′ (restriction site in lower case) and 5′-GCGCggatccTTACTCATGATGATCCTCCTCG-3′ (start codon underlined). The resulting product was digested with XbaI and BamHI and cloned into pBluescript-KS II to generate plasmid pNR38. The pro-duction of an equivalent PCR fragment covering the region directly downstream of fdnI proved problematic, as this non-coding region contains a region of repetitive DNA that prevented faithful replication of this region. Instead, we used a 563 bp fragment starting a full 194 bp from the fdnI stop codon and amplified using the primers 5′-GCGCatcgatATAACAGTTTCAAATGGCGCTGT-3′ (which introduces a stop codon, underlined, in frame with fdnI in resulting plasmid pNR70) and 5′-GCGCggtaccGCGTAGGAAGGAACAATAATG-3′. This was digested with ClaI and KpnI and ligated into pNR38 to produce plasmid pNR70. The complete insert was then removed by XbaI and KpnI digestion and ligated into pMAK705 for integration into the chromosome according to the method of Hamilton et al. (1989).
Strain TP051 was constructed as follows. A 668 bp fragment covering the region directly preceding the stop codon of dmsB was amplified from MC4100 DNA using primers 5′-GCGCgaattcCGTCACATACAAACCTTGTTCAGG-3′ and 5′-GCGCagatctCACCTCCTTCGGGTTTGCCAG-3′, digested with EcoRI and BglII and cloned into pQE60 to generate plasmid pDMSB1. DNA covering the stop codon of dmsC and the downstream 544 bp was amplified from MC4100 DNA using primers 5′-GCGCaagcttTAATCATAACAACCGGGGTTTCGG-3′ and 5′-GCGCatcgatCGTGCTGGCGCTGGGTGACCTGGC-3′, digested with HindIII and ClaI and cloned into pBluescript-KS II to give plasmid pDMSDEL1. The DNA covering the his-tagged DmsB region was excised from plasmid pDMSB1 by digestion with EcoRI and HindIII and cloned into pDMSDEL1 to give plasmid pDMSDEL2. The DNA covering the his-tagged DmsB-deleted DmsC region was excised from pDMSDEL2 by digestion with KpnI and XbaI and cloned into the polylinker of pMAK705 to give plasmid pDMSDEL3. The mutant allele was recombined into the chromosome of LCB628 using the method of Hamilton et al. (1989).
Gene (translational) fusions to the fdnG gene were made at a BamHI site that was engineered by site-specific mutagenesis at codon Lys-44 in fdnG (44-AAAGAGATC-46 changed to 44-AAGGATCCC-46), and at a native BamHI site overlapping codon Asp-792 (791-TGGGATCCG-793). Fusions to the fdnH gene were made at an engineered SalI site at codon Lys-258 (258-AAACCG-259 changed to 258-TCG ACG-259) and an engineered PstI site at codon His-293 (293-CATGAGTAA changed to 293-CTGCAGTAA). Fusions to the fdnI gene were made at engineered PstI sites at codons Arg-15 (14-GATCGCGCC-16 changed to 14-GACTGCAGC-16), Met-52 (52-ATGGGACGC-54 changed to ATCTGCAGC-54), Gly-112 (111-GCCGGGCAA-113 changed to 111-GCCTGCAGA-113) and Leu-153 (152-CTGCTG-153 changed to 152-CTGCAG-153), and at an engineered HindIII site at codon Gly-216 (215-GAAGGGATA-217 changed to 215-GAAGCTTTA-217). Each Φ(fdn–lacZ) gene (translational) fusion was constructed by isolating fdn fragments that extend from an introduced EcoRI site located 313 nucleotides upstream of the fdnG transcription initiation site (Li and Stewart, 1992) to each of the newly introduced downstream restriction sites. Fragments were cloned into vectors pNM480 or pNM482 (Minton, 1984) as appropriate to establish an in frame fdn–lacZ fusion junction (at lacZ codon Val-10). Thus, fusions to fdnH include an upstream fdnG+ gene, and fusions to fdnI include upstream fdnG+ and fdnH+ genes. All fusions (except those to codon Lys-44) include a change of fdnG codon 196 from UGA (Sec) to UCA (Ser) to ensure that UGA decoding to selenocysteine did not limit fdnGHI expression (Berg et al., 1991b). Each fusion junction includes up to 10 additional codons derived from the pNM48x polylinker.
Each Φ(fdn–phoA) gene fusion was constructed by cloning a ′phoA–kan cassette into the engineered site in each of the Φ(fdn–lacZ) plasmids described above. The cassettes PHOK2 or PHOK3 (Rodríguez-Quiñones et al., 1994) were used as appropriate to establish the reading frame. Fusions were to phoA codons Val-28 (PHOK3) and Ala-33 (PHOK2). Each fdn–phoA fusion junction includes up to 25 additional codons derived from the PHOK linker.
The Φ(fdnG–bla)K44 and Φ(fdnG–cat)K44 gene fusions were constructed from the EcoRI–BamHI (codon 45) fdnG′ fragments described above. The ′bla and ′cat genes were obtained by PCR amplification from plasmids pBR322 (source of bla from transposon Tn3) and pACYC184 (source of cat from transposon Tn9) respectively. The primers used for amplification introduced in frame BamHI sites just upstream of the bla and cat initiation codons (5′-N NNN ATG-3′ changed to 5′-G GAT CCG-3′) and HindIII sites just downstream of the termination codons. Primers were: 5′-GAAggatccGAGTATTCAACATTTCC-3′, bla 5′ primer (restriction site in lower case); 5′-ATGAGTaagcttGGTCTGACAGTTAC-3′, bla-3′ primer (termination codon underlined); 5′-GAAGCggatccGGAGAAAAAAATCACTG-3′, cat 5′ primer; 5′-TTAAAaagcttACGCCCCGCCCTGCC-3′, cat-3′ primer. The EcoRI–fdnG′–BamHI–′bla–HindIII and EcoRI–fdnG′–BamHI–′cat–HindIII fusions were assembled in the pUC19-derived polylinker of plasmid pK19 (Pridmore, 1987). The BamHI–′bla–HindIII and BamHI–′cat–HindIII cassettes were also cloned into the native BamHI site at fdnG codon Asp-792 to generate the Φ(fdnG–bla)D792 and Φ(fdnG–cat)D792 gene fusions.
The Φ(sufI-x)A27 fusions contain DNA encoding the SufI signal sequence up to and including the Ala-Ser-Ala cleavage site. In most cases, the constructs contained 503 bp upstream of sufI including 429 bp of the upstream plsC gene. For the Φ(sufI–lacZ)A27 fusion, only 128 bp upstream of the sufI initiation codon (54 bp of plsC sequence) were included. The Φ(sufI–bla)A27 fusion was constructed as follows. The DNA coding for the signal sequence of SufI and upstream regions was amplified with primers 5′-GCGCtctagaCTTCGGGCAGTTGTACTGGTTAACC-3′ and 5′-GCGCgatatcTGC GCTGGCCTTCAGGGGAAC-3′, digested with XbaI and EcoRV and cloned into pBluescript-KS II to give plasmid pHH15. The β-lactamase gene lacking DNA encoding the native signal peptide was amplified using primers 5′-GCGCctcgagCACCCAGAAACGCTGG-3′ and 5′-GCGCgggcccTTACCAATGCTTAATCAGTG-3′ (stop codon underlined), digested with XhoI and ApaI and cloned into pHH15 to give plasmid pHH16. As the Φ(sufI–bla)A27 construct is present in pBluescript-KS II, it was necessary to subclone the fusion into a vector that did not carry an additional copy of the β-lactamase gene. Therefore, pHH16 was digested with ApaI, end filled with T4 polymerase and then digested with XbaI to excise the fusion region. This was ligated into pDHB5700 that had been digested with Ecl136I and XbaI to give plasmid pHH18. For the construction of Φ(sufI–phoA)A27, DNA coding for the signal sequence of SufI and upstream regions was amplified with primers 5′-GCGCtctagaCTTCGGGCAGTTGTACTGGTTAACC-3′ and 5′-GCGCggatccGCTGCGCTGGCCTTCAGG-3′. The amplified product was digested with XbaI and BamHI and cloned into pSK4158, which had been digested previously with the same enzymes, to give plasmid pHH21. For the construction of Φ(sufI–lacZ)A27, DNA encoding the SufI signal sequence and 128 bp of upstream DNA was amplified by PCR using the following primers 5′-GCGCgaattcCGCCGAGCTCGATAAAGAAGTCGC-3′ and 5′-GGCggatccCCGGCTGCGCTGGCCT TCAGGG-3′, with MC4100 DNA as template. The product was digested with EcoRI and BamHI and cloned into pVJS2213 digested with the same enzymes to give plasmid pHH26.
Construction of strains by transduction using phage P1 followed the methods of Miller (1992).
Subcellular fractions were prepared as described previously (Sargent et al., 1998; Stanley et al., 2000). The DMSO reductase activity of DmsAB in subcellular fractions is normally quite labile but, in our hands, can be significantly stabilized by a combination of anoxic conditions, the use of Tris-based buffers and the addition of a protease inhibitor cocktail (Roche). Fractionation efficiency was monitored using acid phosphatase and glucose-6-phosphate dehydrogenase as periplasmic and cytoplasmic marker enzymes respec-tively (Atlung et al., 1989; Sargent et al., 1998). Only data from fractionations in which the marker enzyme activities were ≥95% correctly localized are reported. β-Lactamase activity was measured spectrophotometrically by following the hydrolysis of 7-(thienyl-2-acetamido)-3-[2-(4-N,N-dimethylaminopheylazo)pyridinium-methyl]-3-cephem-4 carboxylic acid (Calbiochem). β-Galactosidase activity was assessed by the hydrolysis of ONPG (Sigma) using the method described by Miller (1992). Chloramphenicol acetyltransferase activity was determined according to the method of Shaw (1975) by reaction of the Coenzyme-A product with 5,5′-dithiobis-(2-nitrobenzoic acid). Alkaline phosphatase activity was measured spectrophotometrically by following the hydrolysis of p-nitrophenyl phosphate. The activity of DMSO reductase was measured spectrophotometrically as described previously (Silvestro et al., 1988) using benzyl viologen (BV) radical as electron donor and TMAO as electron acceptor.
SDS–PAGE used the buffer system described by Laemmli (1970). Antigens were detected after immunoblotting using either the Protoblot (Promega) or ECL (Amersham-Pharmacia Biotech) detection systems. Antiserum against FDH-N (Graham and Boxer, 1981) was kindly provided by Professor D. Boxer, University of Dundee, UK. Polyclonal antisera directed against TEM β-lactamase, chloramphenicol acetyltransferase, alkaline phosphatase and β-galactosidase were obtained from 5 PRIME → 3 PRIME.
We thank Professor D. H. Boxer for anti-FDH-N antiserum and partially purified FDH-N. V.S. thanks Professor Peter Hinkle for educational discussions. Work in Norwich was supported by the Biotechnology and Biological Sciences Research Council through project grant 88/P09634 and by the CEC through grant QLK3-CT-1999. Work in Ithaca was supported by US Public Health Service grant GM36877 from the National Institute of General Medical Sciences. N.R.S. was the recipient of a Norwich Research Park Studentship. F.S and T.P. are Royal Society University Research Fellows. B.C.B. is R. J. P. Williams Senior Research Fellow at Wadham College, Oxford.
Note added in proof
The structure of E. coli FDH-N has now been determined by X-ray crystallography (Jormakka et al., in press). The FdnG and FdnH subunits are located at the face of the transmembrane FdnI protein that, on the basis of marker fusion analysis in this work and in Benoit et al. (1998), has been assigned a periplasmic location. The FDH-N structure thus confirms the periplasmic localization of FdnG that is deduced herein.