The Tat Protein Export Pathway

TRACY PALMER,^1* FRANK SARGENT,¹ AND BEN C. BERKS²

[SECTION EDITOR: TRACY PALMER]

Posted 3 November 2010

Division of Molecular Microbiology, College of Life Sciences, University of Dundee, Dundee DD1 5EH, Scotland,¹ and Department of Biochemistry, University of Oxford, South Parks Road, Oxford OX1 3QU,² United Kingdom

*Corresponding author. Mailing address: Division of Molecular Microbiology, College of Life Sciences, University of Dundee, Dundee DD1 5EH, Scotland, United Kingdom. Phone: 44–1382-386464, Fax: 44–1382-388216, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it

INTRODUCTION

Proteins that reside partially or completely outside the bacterial cytoplasm require specialized pathways to facilitate their localization. Globular proteins that function in the periplasm must be translocated across the hydrophobic barrier of the inner membrane. While the Sec pathway transports proteins in a predominantly unfolded conformation, the Tat pathway exports folded protein substrates. Protein transport by the Tat machinery is powered solely by the transmembrane proton gradient, and there is no requirement for nucleotide triphosphate hydrolysis. Proteins are targeted to the Tat machinery by N-terminal signal peptides that contain a consensus twin arginine motif. In Escherichia coli and Salmonella there are approximately 30 proteins with twin arginine signal peptides that are transported by the Tat pathway. The majority of these bind complex redox cofactors such as iron-sulfur clusters or the molybdopterin cofactor. In this chapter, we describe what is known about Tat substrates in E. coli and Salmonella, the function and mechanism of Tat protein export, and how the cofactor insertion step is coordinated to ensure that only correctly assembled substrates are targeted to the Tat machinery.

THE BACTERIAL TAT PATHWAY

The Tat pathway in E. coli was only identified in 1998 and did not, therefore, feature in the earlier print editions of this work. Before 1998 it was thought that all globular proteins were exported by the Sec system. However, for cofactor-containing proteins that reside in the periplasm the timing and location of cofactor insertion was uncertain. Several of the more complex metallo-cofactors, such as the nickel-iron active site of hydrogenase and the molybdopterin cofactor, are highly oxygen sensitive. Moreover, a lack of ATP supply in the periplasm means that any cofactor maturation or insertion steps requiring nucleotide hydrolysis must necessarily occur in the cytoplasmic compartment. An additional complicating factor is that several cofactor-containing proteins, the prime example being the nickel-iron cofactor-containing large subunit of hydrogenase, are located at the periplasmic side of the cytoplasmic membrane despite lacking any cleavable N-terminal signal sequence.

Taken together, these observations resulted in speculation that export of such proteins might be incompatible with the Sec pathway and that an additional route of protein export may exist that was dedicated to the translocation of cofactor-containing proteins. Indeed, it was noted that groups of unrelated proteins binding a range of different redox cofactors all shared an unusual and conserved motif in their signal peptides (15, 16, 59, 91, 218). This motif was defined as S-R-R-x-F-L-K where the consecutive arginine residues were always invariant. The presence of this alternative protein export pathway in E. coli was first demonstrated in a study examining the assembly of the membrane-bound dimethyl sulfoxide (DMSO) reductase enzyme (209). This enzyme is a heterotrimeric protein comprising a molybdenum cofactor-containing DmsA subunit, an iron-sulfur cluster-containing DmsB subunit, and an integral membrane DmsC subunit. The DmsA protein harbors a twin arginine signal peptide and along with DmsB resides at the periplasmic side of the membrane. (See EcoSal Chapter S- and N-Oxide Reductases for a more detailed discussion about the topology of DMSO reductase.) A mutant was identified that showed mislocalization of DmsA and that was also devoid of nitrate and trimethylamine-N-oxide (TMAO) reductase activities in the periplasmic fraction. The point mutation mapped to a region of the genome that, despite containing a number of sequencing errors in early genome sequence releases, coded for several well-conserved genes of unknown function. The authors reported that the point mutation fell within the yigT gene, which was reassigned as mttA (for membrane targeting and translocation) (209).

Shortly afterward, a second report describing the same pathway was published. This latter study took advantage of a recent publication identifying a gene, hcf106, which encoded a component of the ΔpH pathway which transports folded proteins across the thylakoid membranes of plant chloroplasts (42, 43, 140, 187). It was apparent that homologues of hcf106 were encoded in the genomes of many bacteria, including E. coli, and that one of these was at the yigT locus. Sargent et al. (182) resequenced the yigT region of the E. coli chromosome to reveal that yigT actually encoded two genes, later reassigned as tatA and tatB, both of which encoded proteins homologous to Hcf106. After constructing defined deletions of tatA and a close homologue at a second chromosomal location, ybeC (later tatE), it was demonstrated that these two components were essential for the export of the cofactor-containing proteins hydrogenases-1 and -2, TMAO and DMSO reductases, and formate dehydrogenase-N, and that they shared overlapping functions in this novel export pathway. Sargent et al. (182) assigned the designation tat (for twin arginine translocation), and this nomenclature has since been adopted as the standard terminology for this protein export system in both prokaryotes and eukaryotes.

This review will cover what is known about the Tat system in the enteric bacteria E. coli and Salmonella. For further reading regarding the Tat system in plant thylakoids, the reader is referred to the recent reviews of Cline and Theg (46) and Aldridge et al. (3).

TAT SIGNAL PEPTIDES

As described above, proteins are targeted to the Tat machinery by N-terminal cleavable signal peptides that contain a consensus twin arginine motif. A schematic representation of Sec and Tat signal peptides is shown in Figure 1. The two types of signal peptides show a similar architecture, with a polar n-region, a hydrophobic h-region, and a c-region containing the recognition sequence for the signal peptidase LepB, which cleaves both Sec and Tat signal peptides (124, 215). This overall similarity may explain why twin arginine signal peptides were overlooked for so long. However, there are important differences between the two classes of signal peptide that are critical for targeting to the correct export pathway, and in some cases to avoid mistargeting to the incorrect machinery.

Fig. 1Schematic representation of Sec and Tat signal peptides. Both Sec and Tat signal peptides show a recognizable tripartite structure with a polar (basic) n-region, a hydrophobic h-region, and a polar c-region that contains the recognition site for signal peptidase (shown as AxA in the figure, although generally residues at the -3 and -1 positions relative to the cleavage site are any amino acid with a small, neutral side chain; reference 157). The vast majority of Sec-targeting signal peptides in E. coli have a length between 15 (for the signal peptide of the lipoprotein CsgG) and 37 amino acids (FimO signal peptide). Some occasionally have longer length, for example, the unusually long signal peptides associated with autotransporter proteins such as Ag43 (55, 197). However, the vast majority of E. coli Sec signal peptides are fewer than 24 amino acids long. Tat signal peptides have a conserved motif, S-R-R-x-F-L-K, that is found at the n-region/h-region boundary and are generally markedly longer than Sec signal peptides, varying in length in E. coli from 25 (YedY signal peptide) to 50 (YagT signal peptide) amino acids. Other differences include the fact that the h-regions of Tat signal peptides are less hydrophobic than Sec signals and that they often contain one or more basic residues in the c-region that are almost never found in Sec signal peptides and that act as a Sec-avoidance signal (26, 47).

The Twin Arginine Motif

The most striking feature of Tat signal peptides is the twin arginine motif. The motif in bacteria can be defined as S-R-R-x-F-L-K, where the consecutive arginines are almost always invariant, the other motif residues occur with a frequency greater than 50%, and the amino acid at the x position is usually polar (15, 191).

The consecutive arginines.

Numerous studies have shown that the twin arginines play a critical role in Tat signal peptide function. Mutation of just one of the arginines, even conservatively for lysine, has a drastic effect on export of the passenger protein, resulting in either a dramatic slowing of export or, in some cases, a complete block (59, 82, 191). Interestingly, however, there are a few natural substrates of the Tat pathway that have variant signal peptides in which one of the arginines is replaced, usually in the first arginine position. For example, the iron-sulfur cluster-containing TtrB subunit of the Salmonella tetrathionate reductase has a lysine residue in the first arginine position (see Table 1). Despite harboring a non-canonical twin arginine motif, the signal peptide is able to mediate efficient Tat-transport when fused to the mature region of the E. coli Tat substrate SufI (90). A similar natural substitution is also seen on the E. coli C3736 protein (which is encoded by most strains of E. coli but is disrupted by frameshift in E. coli K-12), homologues of which from other bacteria have a classical twin arginine motif. A second variant of the Tat signal peptide is also to be found on the E. coli plasmid-encoded protein penicillin acylase (Table 1). In this case, the substitution is also at the first arginine position, this time for asparagine (95).

A comprehensive mutagenesis study by DeLisa et al. (53), using a Tat compatible reporter protein, showed that while a basic residue was preferred at the first arginine position, glutamine and asparagine in addition to arginine or lysine could be tolerated in the second arginine position. Interestingly, a naturally occurring Tat-targeting signal peptide found on a Mycobacterium tuberculosis FAD-containing protein has an arginine-glutamine pairing at the twin arginine position but was shown to be Tat-transport active in a reporter assay (213). However, it is likely that such variant signal peptides are very rare since almost all plausible cofactor-containing Tat substrates in the sequence databases have consecutive arginine residues in their signal peptides.

Although the consecutive arginine residues in the n-regions of Tat signal peptides are their hallmark, Sec-targeting sequences may also have twin arginines in their n-regions. In E. coli examples include the Sec-targeting signal of the c-type cytochrome NrfA and of the penicillin-binding protein 6b (DacD). This fact indicates that the twin arginine motif per se is not incompatible with the Sec pathway and that it is other features of the Tat signal peptide and/or the passenger domain that confer Sec incompatibility.

The remaining conserved residues.

Fewer studies have been undertaken of the other conserved residues in the twin arginine motif. It has, however, been shown that substitution of the amino acid in the "F" position has the next most significant effect on Tat transport. Substitution of the phenylalanine in the E. coli SufI signal peptide or the Streptomyces lividans xylanase C signal peptide, or the leucine that occupies this position in the E. coli DmsA signal peptide, for less-hydrophobic amino acids resulted in a moderate to severe decrease in export of the passenger protein (118, 134, 191). This suggests that the relative hydrophobicity of the residue at the "F" position is an important determinant.  However, the situation is not completely clear cut because substitution of the phenylalanine of the E. coli TorA signal peptide with either alanine or serine was without effect on the export of a green fluorescent protein (GFP) passenger (nevertheless, substitution with the charged residues arginine or aspartate was poorly tolerated in this system) (134).  It should be borne in mind that the effects observed for a particular subsitution may depend upon the experimental system used, for example the nature of the passenger protein or the expression level of the protein analyzed.

A single study has addressed the role of the amino acid at the consensus leucine position. In the SufI signal peptide this position is occupied by isoleucine, and substitution for alanine resulted in a modest decrease in the rate of SufI export. This observation suggests that the hydrophobicity of the amino acid in this position is important for Tat transport, as is the case for thylakoid Tat targeting sequences (29, 191).

In E. coli and Salmonella twin arginine signal peptides (Table 1), lysine occupies the final position in the Tat motif in just over 50% of the sequences, with a glutamine (20%) or another polar amino acid being present in the other cases. The reason for the high frequency of occurrence of lysine at this position is not clear, but it is notable that it breaks up the stretch of hydrophobic residues between the "F" and "L" positions of the motif and the h-region of the signal peptide. Stanley et al. (191) demonstrated that substitution of the glutamine found in this position of the SufI signal peptide for any of alanine, leucine, or lysine slowed down the export of SufI, with the lysine introduction having the most dramatic effect. Perhaps surprisingly, substitution of the lysine in the YacK (now CueO) signal peptide with an alanine was without effect, but replacement with an arginine resulted in a slight increase in export of the protein. It has been suggested (191) that the presence of a lysine may slow down the export of substrates to allow time for cofactor insertion. However, it should be noted that not all substrates with a lysine at this position in the signal peptide are cofactor containing. Interestingly, the Streptomyces group of organisms show a preponderance of alanine or glycine at this position, indicating that there may be some species specificity in the residues found at some of the consensus positions (58, 185, 212).

The amino acid directly preceding the consecutive arginines is most frequently serine. It has been proposed (15, 191) that this residue acts as a helix-capping residue to stabilize a potential α-helix formed by the rest of the signal peptide. However, substitution of the serine residue in the SufI signal peptide with alanine, which is unable to cap α-helices, was without significant effect (191). By contrast, substitution of the conserved serine in either of the TorA or DmsA signal peptides showed, respectively, a severe slowing or a complete block in the export of a GFP passenger protein (134). The reason for this striking difference is not clear.

The ‘"x" position and motif spacing.

Examination of known Tat targeting signals (for example, those listed in Table 1) reveals that almost any amino acid is found at the "x" position, although only very rarely, if ever, is it an amino acid of high hydrophobicity. This observation suggests that there is a requirement for appropriate spacing between the arginines and the hydrophobic residues at the "F" and "L" positions. The best evidence in support of this contention is a study by Buchanan et al. (35), who examined the effect of site-directed mutagenesis of the twin arginine motif on the export of TMAO reductase expressed at native levels. As can be seen in Table 1, the TorA signal peptide has three consecutive arginines, the first two of which occupy the consensus positions and the third of which occupies the "x" position. Mutagenesis of the first arginine resulted in a significant reduction in periplasmic TorA activity, despite the fact that the signal peptide still contained two consecutive arginines.

The Hydrophobic (h-) Region

Comparison of the hydrophobicity of the h-regions of Tat signal peptides reveals that they are significantly less hydrophobic than Sec-targeting sequences (47). They contain more glycine residues and fewer leucine residues than Sec signal peptides, and this difference in hydrophobicity between Sec and Tat signal peptides has functional consequences. An early study by Cristobal et al. (47) showed that increasing the hydrophobicity of the TorA signal peptide h-region resulted in rerouting of the passenger protein (in this case, the P2 domain of signal peptidase) to the Sec pathway. Interestingly, the h-regions of uncleaved Tat targeting signal anchors of bacterial Rieske iron-sulfur proteins, which form membrane-spanning helices in the assembled protein, are significantly more hydrophobic than the h-regions of cleaved Tat signal peptides, although their average hydrophobicity is still less than standard Sec signal peptides and markedly less than Sec-assembled transmembrane segments (5).

It is also likely that the positioning of twin arginine motif relative to the h-region is important for signal peptide function. Although this issue has not, as yet, been addressed experimentally, it is striking to note that the twin arginine motif always directly precedes the hydrophobic region.

The c-Region

This short region of the signal peptide contains the recognition sequence for proteolytic removal of the peptide from the mature protein. This cleavage occurs at the periplasmic side of the membrane, probably at a late stage in the transport process. The cleavage site for twin arginine signal peptides is identical to that found on Sec signal peptides, and both types of signals in E. coli are processed by signal peptidase (124, 215). There is, however, a difference between Sec and Tat signal peptides in the c-region. Tat signal peptides, in contrast to Sec signals, often contain one or more basic residues in this location. Ample evidence has shown that the presence of a basic amino acid in the c-region is not a requirement for recognition by the Tat machinery and may just reflect a tolerance for basic amino acids in this position. However, a basic residue in the Tat signal peptide c-region can act as a Sec-avoidance signal, preventing engagement of the signal peptide with the Sec machinery (26). Thus, it was demonstrated that the pair of consecutive arginines in the c-region of the TorA signal peptide prevents targeting of reporter proteins to the Sec machinery, while their removal allows the reporter to be transported by the Sec pathway with reasonable efficiency (23, 47).

Predicting Twin Arginine Signal Peptides

As described above, there are several features that distinguish Tat targeting sequences from Sec signals. However, in reality it is not always straightforward to differentiate between these two types of signal peptide. To assist in identifying candidate Tat signal peptides, two generally available prediction programs are available. TATFIND was the first such program to be developed and uses a rule-based classification of allowed amino acids at certain positions of the consensus motif in combination with the presence of an uncharged block of at least 13 amino acids positioned directly C-terminal to the consensus motif (58, 174). More recent versions of this program have improved the prediction capacity by increasing the number of allowed residues at the consensus positions. TATFIND can be found at http://signalfind.org/tatfind.html. A second prediction program, TatP (13), uses an approach similar to that of TATFIND to classify allowed amino acids at the consensus positions. The program couples two artificial neural networks, one for recognition of the cleavage site and the second to determine whether a given amino acid belongs to a Tat signal peptide. TatP can be found at http://www.cbs.dtu.dk/services/TatP/.

Both TATFIND and TatP are reasonably good at predicting Tat signal peptides and usually generate partially overlapping lists of substrates when used to search genome sequences. Based on observations with the known Tat substrates of E. coli, and the experimentally verified substrates of Streptomyces coelicolor, those signal peptides that are recognized as Tat signals by both programs are almost certainly correct (13, 58, 212). However, a significant number of genuine Tat substrates (probably at least 20%) are recognized by only one of the two programs. Experimental testing is necessary to be certain that transport of a protein is Tat dependent. Ideally, experimental verification should be done by testing the Tat dependence of the native protein in the native organism. However, if this approach is not feasible (e.g., because of the genetic intractability of the native organism), the Tat-targeting ability of the signal peptide can be assessed by using several reporter systems (e.g., GFP, MalE, or agarase) (23, 177, 202, 205, 212, 213). Export of GFP mediated by the TorA signal peptide is shown in Fig. 2A.

Fig. 2GFP as a reporter for export by the E. coli Tat pathway. Fluorescence light microscopy images of  E. coli strain MC4100 (A) (37; tat⁺) and the cognate tatABCDE deletion strain DADE (B) (211) producing GFP fused to the TorA signal peptide from plasmid pRR-GFP (177). Halos of GFP are observed in a tat⁺ strain, whereas only diffuse cytoplasmic fluorescence is seen when the Tat system is inactive. Note that arabinose-resistant isolates of the two strains were used. Subcellular fractionation reveals that the GFP detected in the tat⁺ strain resides in the periplasmic compartment (data not shown). We thank Dr. Berengere Ize for providing the images.

E. COLI AND SALMONELLA TAT SUBSTRATES

A full list of the known or probable Tat substrates found in E. coli and Salmonella can be found in Table 1. Inspection of the list shows that a significant number of these substrates are cofactor-containing proteins, many of which are involved in anaerobic respiration. In addition to respiration, Tat substrates have diverse roles in other cellular processes, including iron uptake, copper resistance, and cell wall metabolism, and inactivation of the Tat pathway is associated with incorrect cell division and aberrant biofilm formation (99, 100, 190). In this section, the contributions that Tat substrates make to the physiology of E. coli and Salmonella are discussed.

Hydrogenases

E. coli and Salmonella both encode two periplasm-facing membrane-bound hydrogenases, HYD-1 and HYD-2, discussed in more detail in EcoSal Chapter Anaerobic Formate and Hydrogen Metabolism. The enzymes are multisubunit and catalyze the removal of two electrons from a molecule of dihydrogen gas to liberate two protons, with the electrons being passed into the respiratory chain (25). They thus allow E. coli and Salmonella to grow with hydrogen as an electron donor. Hydrogen is produced to relatively high levels (40 μM) in animals through the action of the gut microflora (148). The active site found in the large subunits of HYD-1 and HYD-2 contains nickel and iron atoms. The assembly of the active site is a complex process that requires several accessory proteins, one of which is a protease that removes a short peptide at the C terminus of the large subunit. The large subunit is found in complex with a small subunit that contains iron-sulfur clusters and is involved in the transfer of electrons away from the active site and to a membrane-bound quinone reductase (25).

A Tat signal peptide is found on the small subunits of HYD-1 and HYD-2, whereas the large subunit is devoid of a signal peptide even though it is located at the periplasmic side of the membrane. It has been shown that the large and small subunits are exported as a complex through the Tat machinery using the signal peptide on the small subunit (173). The signal peptides of hydrogenase small subunits have rather long n-regions, and they show significant sequence conservation in addition to the twin arginine motif (17). It is highly likely that this signal peptide is bound by a proofreading chaperone (see "Chaperones on the Tat Pathway," below), allowing insertion of the cofactor and binding of the large subunit before the assembled complex can interact with the Tat machinery (61, 186). The importance of the signal peptide in this process is highlighted by the fact that if the signal peptide of the small subunit of hydrogenase-2 is replaced by that of TorA, the small subunit is correctly targeted to the membrane but the partner large subunit is not. This uncoupling of export of the large and small subunits indicates that the hydrogenase assembly process requires information contained in the hydrogenase signal peptide. Interestingly, deletion of the hybE gene, which is encoded in the hydrogenase-2 structural operon, gave the same outcome as the signal peptide-swapping experiment—the small subunit was correctly targeted, but the large subunit remained behind in the cytoplasm. These observations strongly suggest that HybE acts in concert with the hydrogenase-2 Tat signal peptide to mediate correct targeting of the heterodimeric small and large subunits (101).

Hydrogenases are some of the most complex proteins that are substrates of the Tat system. Not only are they exported as heterodimers but the small subunit also has a transmembrane helix at its C terminus that must also be handled by the Tat machinery (84). In fact, there are five Tat substrates in E. coli that have C-terminal transmembrane segments, the others being the -H subunits of the exported formate dehydrogenases and the HybA protein. These substrates provide the cell with a biosynthetic problem because protein export by the Tat system is strictly posttranslational, meaning that the hydrophobic stretch will be present in the cytoplasm for a time before the protein is exported. Since these hydrophobic regions would be prone to aggregation, it is also likely that there is an additional chaperone required to mask this region of the protein before export. It is not fully understood how the Tat machinery brings about the insertion of these transmembrane segments, although it has been noted by using certain chimeric model substrates that extended regions of polypeptide containing hydrophobic stretches partition into the membrane (44, 171). Unlike most of the cytoplasmic membrane proteins of E. coli, it has been demonstrated that there is no requirement for YidC for the integration of these C-terminal transmembrane segments (84).

Formate Dehydrogenases

In E. coli and Salmonella the two periplasm-facing formate dehydrogenases, FDH-N and FDH-O, also depend on the Tat pathway for their assembly. (See EcoSal Chapter Anaerobic Formate and Hydrogen Metabolism for a review of formate metabolism in E. coli and Salmonella.) These are three-subunit enzymes (14), where the G and H subunits are exported as a heterodimer via the Tat system (courtesy of a twin arginine signal peptide present on the G subunit) and the I subunit, a multi-spanning membrane b-type cytochrome, is most likely assembled by the SRP-Sec pathway (192). The enzymes catalyze the oxidation of formate (a product of E. coli/Salmonella metabolism only under anaerobic conditions, but E. coli may be using these enzymes to scavenge formate produced by other organisms) to carbon dioxide and protons, with the electrons feeding into the respiratory chain. The formate dehydrogenase G and H subunit heterodimers are the largest native substrates for the E. coli Tat pathway, with a collective mass approaching 150 kDa and dimensions of 60 × 60 × 150 Å (107). The enzyme active site is located in the G subunit and contains molybdenum bound by two molecules of the molybdopterin guanine dinucleotide cofactor (Mo-bisMGD), with selenocysteine forming a ligand to the molybdenum atom. It also contains a [4Fe-4S] cluster, whereas the H subunit harbors four such clusters (107).

The E. coli formate dehydrogenase-N can be expressed at high levels, and under anaerobic conditions in the presence of nitrate it can form up to 10% of the total membrane protein (62). Interestingly, the mRNA covering the twin arginine signal peptide-coding region of E. coli fdnG has been shown to fold into a helical hairpin, regulating translation of the gene. It is not clear under what conditions the hairpin is disrupted or whether this translational control of FDH-N is related to the activity of the Tat machinery (163).

The formate dehydrogenase G and H subunits, like the small and large subunits of hydrogenases, also depend upon a biosynthetic chaperone for their assembly. In this case, the FdhE protein, which is encoded at the fdo locus, is required for the assembly of both FDH-N and FDH-O (128). The FdhE protein will be discussed in further detail in the next section.

DMSO Reductases and Related Enzymes

The E. coli and Salmonella DMSO reductases (discussed in detail in EcoSal Chapter S- and N-Oxide Reductases) are heterotrimeric proteins that are anchored at the periplasmic side of the membrane by means of the membrane-integral DmsC protein (21, 176, 192). The DmsA protein contains Mo-bisMGD at the active site, and it also contains a [4Fe-4S] cluster. The DmsB protein, like the H subunit of formate dehydrogenase, contains four [4Fe-4S] centers. The E. coli enzyme has rather broad specificity and will catalyze the reduction of a range of S- and N-oxides, including DMSO, TMAO, and methionine sulfoxide (189). The twin arginine signal peptide is present on the DmsA subunit, with which the DmsB subunit is coexported as a complex (192, 209). The export of DmsAB is policed by the DmsD chaperone, which interacts with the DmsA signal peptide (149).

Additional DMSO reductase family proteins are found in E. coli and Salmonella. The ynfEFGH gene cluster in both organisms encodes a paralogue of DMSO reductase (with two DmsA homologues, YnfE and YnfF) that has a low level of DMSO reductase activity and can form mixed complexes with the Dms proteins which have some functionality (122). However, the in vivo function of one of these proteins, YnfE, is as a selenate reductase (81). The DmsD protein (encoded by the gene formerly known as ynfI that neighbors ynfEFGH) binds the twin arginine signal peptides of YnfE and YnfF and is absolutely required for selenate reductase activity, indicating that, like FdhE, it is required for the maturation of more than one different but related enzyme (81).

In Salmonella there has been even greater expansion of this protein family, and two additional enzymes of the DmsABC type are encoded by the STM4305-STM4307 cluster and by the STM2530-STM2528 cluster. Curiously, the first of these two has a neighboring gene encoding a DmsD homologue (STM4308), whereas the second one does not. The roles of these proteins in Salmonella physiology are currently unknown.

TMAO Reductase

E. coli TMAO reductase, encoded by the torA gene, is one of the best characterized of all Tat substrates. It is a soluble, periplasmic enzyme containing the Mo-bisMGD cofactor. It catalyzes the reduction of TMAO to trimethylamine (TMA), picking up the electrons required for this reaction from the membrane-anchored pentaheme c-type cytochrome TorC (72) (see also EcoSal Chapter S- and N-Oxide Reductases). The torC and torA genes in E. coli form an operon with a third gene, torD, that encodes a chaperone protein in the same family as DmsD (133, 206). The TorD protein is required for insertion of the molybdenum cofactor into TorA and binds to two sites on the TorA precursor—the twin arginine signal peptide, to which it binds with high affinity, and a second site elsewhere on the protein (68, 83, 101, 158).

There appear to be some differences between the maturation of the TorA proteins of E. coli and Salmonella. Deletion of the torD gene in E. coli results in only a moderate reduction (30 to 50%) in TMAO reductase activity, at least for growth at 37°C (66, 158). This residual activity probably represents maturation of TorA without the assistance of chaperones (96, 101). By contrast, a deletion of torD in Salmonella completely abolishes TorA activity and is therefore more in line with the behavior of chaperone gene deletions in other enzyme systems, which are usually essential for activity (79). The reason for the difference in absolute requirement for TorD between these two organisms is not clear.

It should be noted that a homologue of TorA, TorZ, is present in E. coli but not in Salmonella. Its precise function is unknown, but the enzyme, which is also found in the periplasm, has been shown to catalyze the reduction of TMAO. The torY gene neighbors torZ and encodes a pentaheme c-type cytochrome homologous to TorC (73). Surprisingly, there is no torD-like gene in the vicinity of torYZ and its requirement for a biosynthetic chaperone is currently unknown.

Nitrate Reductase

E. coli and Salmonella encode three nitrate reductases. Two of these, NarA and NarZ, are attached to the cytoplasmic face of the membrane,whereas the third, encoded by the nap genes, is found in the periplasm and in E. coli is involved in nitrate respiration under nitrate-limiting conditions (reviewed in EcoSal Chapter Respiration of Nitrate and Nitrite). The catalytic NapA subunit of the periplasmic nitrate reductase contains Mo-bisMGD and a [4Fe-4S] cluster and forms a soluble complex with a diheme-containing NapB subunit (57, 104). Electrons are fed into the NapAB dimer from a membrane-anchored tetraheme cytochrome c, encoded by the napC gene (78).

The NapA protein is exported by the Tat machinery and is synthesized with a twin arginine signal peptide that is slightly unusual in that it has a long string of alanine residues in the h-region (78) (see Table 1). By contrast, export of the partner NapB subunit uses the Sec pathway so that the c-type heme cofactors of NapB can be inserted in the periplasm. Therefore,  NapA and NapB only associate in the periplasm after transport. Biosynthesis of NapA also depends upon a dedicated chaperone, in this case NapD (127, 161). The role of NapD in the maturation and export of NapA is discussed in more detail in the next section.

Tetrathionate and Thiosulfate Reductases

These two enzymes are found only in Salmonella and are not present in E. coli. They confer to Salmonella an ability to respire the sulfur-containing compounds tetrathionate and its reduction product thiosulfate (85, 86) (reviewed in EcoSal Chapter The Aerobic and Anaerobic Respiratory Chain of Escherichia coli and Salmonella enterica: Enzymes and Energetics). Both enzymes have a large catalytic subunit (TtrA/PhsA) containing Mo-bisMGD and a [4Fe-4S] cluster as well as a globular four-[4Fe-4S]-containing partner subunit (TtrB/PhsB), and they are anchored at the periplasmic face of the membrane by a membrane integral protein (TtrC/PhsC) that contains the site of quinol oxidation (86).

It is likely that the biogenesis pathways of these enzymes follow slightly different courses. The PhsB iron-sulfur protein has no signal peptide of its own and is presumably exported as a heterodimer with PhsA. On the other hand TtrB, the analogous subunit of tetrathionate reductase, has a variant Tat signal peptide with a lysine-arginine pair instead of the classical twin arginines. The TtrB signal peptide is recognized by the Tat machinery when fused to a reporter protein, suggesting that it is probably functional in its native context (90). This implies that the TtrA protein, which also has a twin arginine signal peptide, is exported separately from TtrB and that they dimerize in the periplasm. It is not clear what would prevent these partner proteins from forming a complex in the cytoplasm. One possibility is that the TtrB signal peptide sterically blocks formation of the heterodimer interface.

Copper Enzymes

Two Tat-dependent copper proteins are involved in copper homeostasis and resistance in E. coli.  Both play a role in oxidizing Cu^I to the less toxic Cu^II in the periplasmic compartment (169) (see EcoSal Chapter Copper Homeostasis in Escherichia coli and Other Enterobacteriaceae). CueO is the better characterized of the two and is also found in Salmonella. CueO is known to provide resistance to fluctuating copper levels in the micromolar range. Mislocalization of CueO in tat mutant strains accounts, at least in part, for the increased copper sensitivity of these strains (99, 151). PcoA is encoded on the conjugative plasmid pRJ1004. This plasmid was isolated from E. coli in the gut flora of pigs fed a diet supplemented with copper sulfate as a growth promoter (200). Together with the other products of the pco operon, it confers resistance to copper in the millimolar range.

SufI (FtsP)

SufI is a soluble, monomeric protein that does not bind cofactors (198). These properties have made SufI the most popular model substrate for in vivo and in vitro studies of the Tat pathway (1, 191). SufI was first identified as a protein that could suppress the cell division defect of a ftsI strain. More recently, it was shown that sufI in multicopy can also suppress the synthetic lethality of combined ftsE and ftsX mutations (168). Deletion of sufI is without any obvious effect on the growth characteristics of E. coli under standard laboratory conditions. However, sufI mutants show a filamentous phenotype when grown at elevated temperatures in media devoid of sodium chloride (175). GFP tagging of SufI showed that it accumulates at the cell division site and that recruitment to the septal ring depends upon FtsN (198). It has been proposed that SufI be renamed FtsP to reflect its role in the cell division process (175).

The Cell Wall Amidases

Several years ago it was observed that E. coli tat mutants form long chains of cells that have not separated following cell division (reference 190; see Fig. 2B). This chaining phenotype very much resembles that of mutants with defects in the envA gene involved in lipid A biosynthesis. Indeed E. coli tat mutants also show a leaky cell envelope and are particularly sensitive to the presence of sodium dodecyl sulfate (SDS), a commonly used detergent (145, 190). It was initially hypothesized that the cell separation phenotype of the tat strains was due to mislocalized SufI. However, neither the sufI mutant strain, nor indeed strains carrying mutations in genes coding for a range of other Tat substrate proteins, phenocopied the tat mutant (190).

The mystery was solved by isolating transposon insertions that suppressed the SDS sensitivity of a tat mutant. Many of these insertions mapped upstream of an operon that included the gene amiB. It was shown that overexpression of amiB from a multicopy plasmid was sufficient to confer SDS resistance to the tat strains (100). AmiB is a cell wall amidase, one of three such enzymes found in E. coli and Salmonella. Amidases remove the side chains from peptidoglycan by hydrolyzing the bond between the peptide and N-acetylmuramic acid components to allow remodeling of the cell wall, for example at cell division. Inspection of the N terminus of AmiB revealed that it had a standard Sec signal peptide, whereas analysis of the two other cell wall amidases, AmiA and AmiC (which had a misassigned start codon in early releases of the E. coli genome sequence), revealed that they had typical twin arginine signal peptides (19, 100). This observation was rather surprising since AmiB shares approximately 60% sequence similarity and 40% identity with AmiA and AmiC. It is unclear why these structurally related proteins use different export mechanisms. Indeed, it was shown that AmiC can be functionally exported by the Sec pathway when provided with a Sec signal peptide (100). Deletion of amiA and amiC, or just the DNA encoding their signal peptides, was sufficient to produce a cell division defect resembling that of the tat mutant strains (19, 100). Interestingly, it was shown through GFP tagging that AmiC (but not AmiA) localizes to the septal ring in an FtsN-dependent manner. Thus AmiC, like SufI, is implicated in the periplasmic processes required for cell division (19).

Other Tat Substrates

E. coli and Salmonella encode a range of Tat substrates of uncertain physiological role. YdhX, which is not present in Salmonella, is a ferredoxin-like protein encoded in a cluster of genes that are found on the genomes of a diverse range of Gram-negative bacteria. One of these, ydhV in E. coli, encodes a protein related to the tungsten bis molybdopterin-dependent aldehyde ferredoxin oxidoreductases found in hyperthermophilic archaea (41). YdhV has no apparent signal peptide but might be exported to the periplasm through formation of a complex with YdhX. Interestingly, in the genome of the halophilic alkalithermophilic Gram-positive bacterium Natranaerobius thermophilus the YdhY homologue has a twin arginine signal peptide and YdhX is apparently absent. Other conserved gene products of the cluster encode a membrane-bound cytochrome b (YdhU, which may serve as a membrane anchor) or soluble proteins lacking signal peptides (YdhT, YdhW, YdhY). Taken together these observations suggest that the gene cluster encodes a novel periplasmic tungsten enzyme.

YedY contains the mo-MPT-form of the molybdenum cofactor, and it is related to the sulfite oxidase family of molybdoenzymes (121). A membrane-bound-heme b containing protein, encoded by yedZ, acts as its electron transfer partner (30). Assays using purified YedY show low activity with S- and N-oxides as substrates, but the in vivo substrate is unknown (121).

EfeB, formerly YcdB, was shown to be a periplasmic protein containing noncovalently bound heme. Although the heme bestows peroxidase activity on EfeB (196), recent data suggest that heme is actually the substrate of the enzyme, rather than a cofactor, and that the physiological role of EfeB is iron acquisition from heme (117). The other genes in the efeB-containing operon encode a low-pH Fe²⁺ uptake system which is used to import the iron atoms released by EfeB (36).

YagT is a periplasmic iron-sulfur protein found only in some strains of E. coli. It forms a heterotrimer with YagR, which binds the MCD variant of the molybdenum cofactor, and YagS, a FAD-containing protein, together forming a periplasmic aldehyde oxidoreductase (141). All three proteins are probably exported as a heterotrimer through the Tat system by virtue of the YagT Tat signal peptide. A fourth gene, yagQ, encodes a homologue of XdhC and is therefore very likely to be an assembly chaperone for stabilization and insertion of the molybdenum cofactor (142, 143). It may conceivably also be involved in controlling assembly of the heterotrimer and its interaction with the Tat pathway.

MdoD, WcaM, and FhuD are three confirmed substrates of the Tat pathway that have no known cofactors. MdoD is involved in the biosynthesis of osmoregulated periplasmic glucans (OPGs), which are produced at increased levels in low-osmolarity media. Although not essential for OPG synthesis, MdoD appears to control the size of the glucose backbone (116). WcaM is encoded by the final gene in the cps operon that is required for the polymerization, transport, and modification of colanic acid, an integral component of the E. coli capsule and a component of Salmonella exopolysaccharide. The role of WcaM in colonic acid biosynthesis is not known. However, a wcaM mutant strain of Salmonella was markedly impaired for biofilm formation on chicken intestinal epithelium (113). FhuD is the periplasmic binding protein for ferric hydroxamate siderophores, including ferrichrome. E. coli tat mutants are defective in the utilization of ferrichrome as sole iron source, and this is almost certainly due to an inability to export FhuD (99).

CHAPERONES ON THE TAT PATHWAY

From the first descriptions of the bacterial Tat pathway commentators have wondered about the relationship between the signal peptide, the membrane-bound transport machinery, and the transported protein. Are any soluble factors involved? How are the cofactor insertion and protein transport events sequenced? How does the Tat system distinguish between fully folded transport-ready substrates and immature preproteins? Studies of both E. coli and Salmonella have made some progress in addressing these questions.

No generic signal recognition protein equivalent to the SecA component of the Sec system has so far been identified for the Tat pathway. Indeed, Holzapfel et al. (93) showed that the cofactor-less SufI protein can be translocated by the Tat machinery in an in vitro assay in the absence of any soluble cytosolic proteins. Moreover, although the general chaperones DnaK and GroEL bind to twin arginine signal peptides and stabilize Tat substrate proteins, neither is essential for Tat targeting (76, 156, 172).

Oresnik et al. (149) identified a soluble protein from a crude extract of E. coli that bound to the twin arginine signal peptide of DMSO reductase. The protein in question was termed DmsD and was shown to be absolutely required for DMSO reductase activity. Sequence analysis identified DmsD as a member of a large family of peptide-binding proteins present in both bacteria and archaea. This family is termed the TorD family (206) because it includes the TorD protein, which had previously been characterized as a chaperone required for the assembly of TMAO reductase (158). Subsequent studies showed that TorD is also a twin arginine signal peptide-binding protein, in this case specific for the TorA signal sequence (101). It is likely that all signal peptide-binding proteins, at least from the TorD family, have the dual roles of  assisting Tat substrate assembly (cofactor insertion and, where applicable, binding of partner subunits) and masking signal peptide activity until assembly functions are completed. The concept of chaperone-mediated proofreading is presented in Fig. 3, and the dual roles played by Tat chaperones are discussed in more detail below.

Fig. 3Quality control and proofreading processes on the Tat pathway. Tat precursor proteins are believed to interact (a) either directly with the Tat machinery or (b) with the lipid bilayer before diffusing laterally toward the Tat channel (7, 188). The Tat translocase itself is proposed to reject precursor proteins that are not folded (Tat Quality Control) (54). Complex proteins, such as cofactor-containing respiratory enzymes, undergo a second tier of quality control called "Tat Proofreading." Here, the signal peptide is bound tightly by a target-specific cytoplasmic chaperone, shown in green (step 1). In some cases, a second molecule of the chaperone binds elsewhere on the apoenzyme. The redox cofactor is then loaded (step 2), and partner subunits may also bind where appropriate (not depicted) before the chaperones are released (step 3).

The TorD Protein

The E. coli TorD protein is encoded by the final gene of the torCAD operon, which codes for the major periplasmic trimethylamine N-oxide reductase (133). TorD was identified as a twin arginine signal peptide-binding protein using a bacterial two-hybrid system (101). Subsequent in vitro calorimetric binding analyses established that purified TorD interacted tightly and specifically with the TorA signal peptide (83) and displayed an apparent dissociation constant (K_d) of 59 nM (34). In addition, calorimetric and hydrodynamic analyses suggested that one TorA signal peptide was bound by one TorD monomer (34, 83).

The TorA signal peptide can be fused to the N termini of other proteins, and the interaction with TorD is maintained. For example, as shown in Figure 2, the presence of the TorA signal peptide at the N terminus of GFP mediates transport of the fusion protein to the periplasm. However, coproduction of TorD alongside this fusion enhanced the translocation efficiency of GFP more than threefold (119). Furthermore, replacement of the HybO twin arginine signal peptide with that of TorA resulted in a complete loss of hydrogenase-2 activity unless TorD was overproduced, in which case significant enzyme activity was recovered (101). In both cases it is assumed that binding of TorD to the TorA signal peptide delays interaction of the passenger protein with the Tat machinery, allowing more time for the passenger to fold (and in the case of HybO to bind its partner subunit, HybC).

Extensive site-directed mutagenesis coupled with biophysical studies identified the key TorD-binding epitope on the TorA signal peptide (34). A hydrophobic stretch of amino acid residues (27-LGPSLL-32) located at the C-terminal end of the TorA signal peptide h-region (see Fig. 4A) is crucial for binding. Substitution of any of TorA L27, G28, L31, or L32 individually with glutamine was sufficient to impair TorD binding as measured by an in vivo two-hybrid assay, and the apparent K_d for TorD binding to one TorA variant (L31Q) was found to have increased to 1,147 nM (34).

Fig. 4Molecular basis of the Tat proofreading process. (A) The primary sequence of the E. coli TorA signal peptide. The twin arginine motif is highlighted in dark blue, and the TorD-binding epitope is highlighted in cyan. The arrow indicates the signal peptidase I (LepB) cleavage site. (B) Crystal structure of the TorD homologue from Shewanella massilia (PDB ID no. 1N1C). The model shows an intertwined dimer of two protomers (blue and yellow). (C) Crystal structure of the DmsD protein from E. coli (PDB ID no. 3EPF). In this case the model shows a monomeric form of the protein. (D) The primary sequence of the E. coli NapA signal peptide. The twin arginine motif is highlighted in dark blue, and the NapD binding epitope is highlighted in cyan. The arrow indicates the LepB cleavage site. (E) A model of the high-resolution solution NMR structure of the E. coli NapD protein (PDB ID no. 2JSX). (F) A model of the NMR-derived solution structure of a complex between E. coli NapD and residues 1 to 35 of the NapA twin arginine signal peptide (PDB ID no. 2PQ4).

The TorD-binding epitope is close to the natural signal peptidase cleavage site of the TorA signal peptide (Fig. 4A), explaining an earlier observation that binding by TorD prevents opportunistic proteolysis of the signal peptide at R35 under some circumstances (69). Identification of a binding site toward the C terminus of the TorA signal peptide suggests that TorD has little direct contact with the conserved twin arginine transport motif located closer to the N terminus. Indeed, substitution of the conserved arginines for lysines, which abolishes protein export activity (35), had no deleterious effect on TorD binding (83). This lack of involvement of the twin arginine motif in TorD binding is an important point to note. The interaction between TorD and the TorA signal peptide is exquisitely specific—E. coli TorD will not recognize any other signal peptide in E. coli (39, 83). It seems very unlikely, then, that the twin arginines of the consensus motif, which is conserved in at least 27 proteins in E. coli, would be central to chaperone recognition and binding. It therefore makes some biological sense that the TorD binding site on the signal peptide is distinct from the Tat-targeting motif. It is likely that chaperone-binding epitopes will have to be determined empirically for each system under investigation.

While the E. coli TorD protein has proven recalcitrant to crystallization, the TorD homologue from Shewanella massilia has proven more tractable. The S. massilia TorD is functionally identical to E. coli TorD insofar as it is coexpressed in an operon encoding, and involved in the biosynthesis of, the S. massilia periplasmic TMAO reductase (48). Overproduced, recombinant S. massilia TorD was isolated in a number of oligomeric forms, including monomers, homodimers, and homotrimers (204). The crystal structure of the homodimer was solved (reference 203; shown in panel B in Figure 4), and the protein was observed to consist of 10 α-helices, with no significant regions of β-strand. The helices are arranged into two domains (N- and C-terminal) separated by a linker region (or "hinge"). Most strikingly, the TorD homodimer displayed "extreme" domain swapping, in which the N-terminal domain of one protomer is bound directly to the C-terminal domain of the partner protein (Fig. 4B). The structure of a signal peptide-TorA complex has not so far been forthcoming.

The Tat Proofreading System

Tat proofreading is defined as the process by which export of a Tat substrate protein is prevented by shielding the signal peptide from interaction with the Tat machinery until cytoplasmic assembly processes are complete (101, 102). A schematic representation of Tat proofreading mediated by a chaperone protein such as TorD is shown in Figure 3. Currently, the precise molecular functions performed by TorD family chaperones are unclear. Experiments with E. coli TorD have shown that there are at least two binding sites for the chaperone on the TorA precursor, one of which is on the signal peptide and a second being within the mature region of the protein (68, 83, 101, 158). A truncated TorA protein lacking its signal peptide still depends on TorD for full enzyme activity, indicating that the second binding site is important for the function of the chaperone (158). TorD variants have been isolated that retain the ability to interact with the TorA signal peptide but that cannot interact with the mature region of the protein, and these support only low levels of TorA activity (68, 101). Conversely, TorD variants that still interact with the mature region of TorA but no longer appear to interact with the signal peptide have also been identified (101).

Presumably within the cell the binding of TorD to both sites on TorA is required for proofreading activity. Given the observation that proteins of the TorD family can form homo-oligomers, it is unclear whether the chaperone is in a monomeric or dimeric state when it binds to its substrate protein in vivo. It is conceivable, for example, that a homodimer of TorD binds to a single TorA precursor, with one protomer occupying the signal peptide binding site and the second occupying the site within the mature region of the protein. It is also not known how release of the bound chaperone from the assembled precursor protein is achieved. Hatzixanthis et al. (83) demonstrated that purified, monomeric TorD can bind GTP. However, GTP hydrolysis was only detected from a TorD dimer form (80). This raises the possibility that dimerization of TorD occurs during the proofreading process perhaps if monomers of TorD bound at the two separate binding sites are brought into close proximity because of conformational change in the TorA precursor upon binding of molybdenum cofactor. GTP hydrolysis could, therefore, be a mechanism that drives release of the chaperone.  However, the TorD homologue DmsD has been reported to associate with the cell membrane if the TatB and TatC components of the Tat apparatus are present (153). It has also been reported that DmsD comes into close proximity to TatBC in living cells (109). These observations may support an alternative scenario, where chaperone release from the assembled precursor is stimulated by interaction of the chaperone-substrate complex with components of the Tat machinery itself.

It has been demonstrated that TorD can bind to different forms of the molybdenum cofactor and that it may also interact with MobA, the enzyme that catalyzes the final step in the biosynthesis of the molybdopterin guanine dinucleotide (MGD) form of the molybdenum cofactor. These findings have led to the suggestion that TorD acts as a platform on which the final step of MGD synthesis and its immediate insertion into apo-TorA is achieved (67, 68).

The DmsD Protein

The E. coli DmsD protein is closely related to E. coli TorD (181) and binds specifically and tightly (apparent K_d, 200 nM by calorimetry) to the signal peptide of the E. coli DMSO reductase DmsA subunit (214). The Salmonella DmsD protein has been reported to interact with the Salmonella DmsA signal peptide with an apparent K_d of 100 nM (81). Both the E. coli and Salmonella DmsD proteins are unusual because they interact with more than one partner protein. Besides the DmsA signal peptide, in both organisms DmsD also recognizes the signal peptide from the YnfE protein (39, 81), which is a molybdenum-dependent selenate reductase (81), and the YnfF protein (39, 81). In E. coli, YnfF is believed to be a second DMSO reductase isoenzyme and as a result shares significant sequence similarity with DmsA (122). Consistent with the physiological relevance of the observed interaction between DmsD and the YnfE and YnfF signal peptides the dmsD gene is located within the ynfEF operon in both E. coli and Salmonella (122).

High-resolution crystal structures of both the E. coli and Salmonella DmsD proteins have been solved (Fig. 4C). As expected, there is significant structural similarity with TorD, although crystallized DmsD is monomeric and shows no evidence of the domain swapping seen with S. massilia TorD (166, 167, 193). In solution, however, DmsD has been found in a range of oligomeric forms (178).

It is likely that the function of DmsD is more complex than that of TorD because the signal peptide on DmsA mediates export of the DmsA-DmsB heterodimer (192, 209), whereas TorA does not have a co-exported partner protein. Therefore, one of the roles of DmsD is presumably to ensure that the DmsAB heterodimer is correctly assembled before interaction with the Tat machinery. It is currently not known whether DmsD can also interact with the DmsB (and the homologous YnfG) subunit (and indeed whether it mediates insertion of the iron-sulfur clusters into these proteins). Alternatively, it is possible that the binding site for DmsD on DmsA may overlap with the DmsB binding site and that the interaction of DmsA with DmsB may be part of the mechanism by which the chaperone in displaced before transport of the heterodimer through the Tat machinery.

Other Members of the TorD Family

E. coli TorD is the archetypal member of a growing family of related bacterial and archaeal proteins (206). Interestingly, one member of the family, NarJ, is required for the assembly of a nonexported protein, the Nar nitrate reductase (22). Although the Nar enzymes are found on the inside of the cell, there is evidence that they may have evolved from a Tat-targeted periplasmic enzyme (129, 179, 206). The Nar enzymes of archaea are located at the periplasmic side of the membrane, and the catalytic subunits have classical twin arginine signal peptides that mediate Tat transport of a heterologous reporter protein (129, 213). One of the major binding epitopes for E. coli NarJ is within the N-terminal 35 amino acids of NarG. This part of NarG exhibits some features of a twin arginine signal sequence and has been termed a remnant signal (97, 180, 206, 207). The N-terminal region of NarG is not active for Tat transport but can be activated to function as a signal peptide by judicious introduction of a few site-directed mutations. The transport-active NarG N-terminal region is impaired in interaction with NarJ, confirming that it is indeed a binding site for the chaperone (97). As with other members of the TorD family NarJ is essential to facilitate the insertion of cofactors into its target enzyme (97, 111, 207). It has been proposed that the nonexported NarG protein retains the N-terminal remnant signal as a binding site that allows NarJ to carry out its cofactor maturation role.

In total, the E. coli K-12 genome encodes five members of the TorD family: TorD itself, DmsD, YcdY, NarJ, and NarW (206). The interaction partners of YcdY are unclear. Salmonella has two additional TorD family members encoded by STM0610 and STM4308. These additional chaperone genes are found within apparent operons encoding uncharacterized molybdoenzymes that have N-terminal Tat signal peptides (listed in Table 1). It is reasonable to assume, therefore, that they play a role in coordinating assembly and export of these cofactor-containing proteins.

The NapD Family

A second family of Tat proofreading chaperones, unrelated to the TorD family, has also been extensively characterized. The E. coli NapD protein is a small (87-amino-acid), monomeric cytoplasmic protein that is encoded within the structural operon for the Nap periplasmic nitrate reductase and that is essential for the activity of this enzyme (127, 161). Bacterial two-hybrid experiments have shown that NapD interacts with the catalytic subunit of the nitrate reductase, NapA, and that it specifically recognizes the twin arginine signal peptide (127, 144). In contrast to the TorD family of chaperones, there is no evidence that NapD has additional binding sites within the remainder of the NapA precursor (127). In vitro experiments have indicated tight binding of NapD to the NapA signal peptide (K_d of 7 nM estimated by isothermal titration calorimetry). NapD also has strong antitransport activity, fully blocking export of a NapA-signal peptide fusion protein (127). This contrasts with the stimulation of export of TorA signal peptide fusion proteins by TorD (119).

The solution NMR structure of E. coli NapD revealed that the protein has a ferredoxin-like fold completely unrelated to the all-helical structures of TorD family proteins (127). The structural differences between the two chaperones suggest that proofreading of exported molybdoenzymes arose independently for different families of enzymes. Recently, an NMR structure of NapD in complex with the NapA signal peptide was deposited in the Protein Data Bank (138) (Fig. 4F). The NapA signal peptide adopts an α-helical conformation when bound by NapD, with one face of the helix interacting with the β-sheet face of the NapD ferredoxin fold. The region of the NapA signal peptide that interacts with NapD mainly covers the h-region (indicated in cyan on Fig. 4D), and there is apparently no contribution of the twin arginine motif to the binding epitope.

Apart from the NapA signal peptide, no other ligands of NapD have as yet been identified, and it is not clear how NapD would be released following cofactor insertion into the NapA precursor. However, given the major differences between the TorD and NapD families of chaperones in terms of structure, oligomeric status, and number of binding sites on the cognate molybdoenzyme, it is likely that the mechanisms of precursor binding and release by the two different chaperone families are fundamentally different.

Hydrogenase Proofreading Chaperones

Hydrogenases are among the most complex Tat substrates and would, therefore, be expected to be subject to proofreading during their assembly. Indeed both of the exported hydrogenase enzymes in E. coli, hydrogenase-1 and -2, have signal peptide-binding chaperones. Interestingly, however, it seems that the assembly process, at least in terms of chaperone requirements, differs between these two homologous enzymes.

The signal peptide-bearing subunit of hydrogenase-1, HyaA, was shown to interact with the HyaE protein using a two-hybrid system (61). The hyaE gene is highly conserved among operons coding for hydrogenase-1-like enzymes and is essential for their activity (18, 136). Interaction was specifically detected between HyaE and the precursor form of HyaA and was not seen when the mature domain of HyaA was tested (61). This observation implicates HyaE in recognizing, at least in part, the twin arginine signal peptide of HyaA. The interaction of the hydrogenase signal peptide with HyaE was later confirmed in a study of the Ralstonia eutropha hydrogenase-1 homologue, where the twin arginine signal peptide of the small-subunit (HoxK) was shown to interact with a complex of the R. eutropha HyaE and HyaF homologs (HoxO/HoxQ) (186).

Homologs of HyaE and HyaF are not encoded within the structural operon for E. coli hydrogenase-2. Instead this operon codes for an unrelated protein, HybE (135), which probably serves a signal peptide binding role during hydrogenase-2 assembly. It was demonstrated using a two-hybrid approach that HybE interacts with the precursor, but not the mature domain, of the hydrogenase-2 small-subunit HybO (61). Deletion of hybE resulted in premature targeting of HybO to the membrane without binding of the large-subunit partner protein, indicating that the proofreading process for the correct assembly of hydrogenase-2 had been disrupted. A similar phenotype was also noted when the signal peptide of HybO was swapped with that of TorA, i.e., premature targeting of HybO to the membrane without its partner subunit. This phenotype could be partially corrected when TorD was co-overproduced, indicating that signal peptides and chaperones operate in tandem to orchestrate the assembly of complex enzymes (101).

FdhE—a Proofreading Chaperone for Exported Formate Dehydrogenases?

The exported formate dehydrogenases, FDH-N and FDH-O, are further complex substrates of the E. coli and Salmonella Tat systems. The fdhE gene was identified as being required specifically for the activity of exported formate dehydrogenases and was dispensable for the activity of the cytoplasmic FDH-H enzyme (128, 155).  FdhE copurifies with the Tat signal peptide-containing FdnG subunit of FDH-N, presumably in the form of a pre-export complex (123). Two hybrid experiments also showed that FdhE interacted with FdnG and FdoG but only when their signal peptides were present (39). However, experiments to demonstrate direct binding of FdnG and FdoG Tat signal peptides to FdhE were unsuccessful, suggesting that the FdhE binding epitope may require both the signal peptide and a portion of the mature domain of the protein (123).

The FdhE protein, unlike members of the TorD and NapD families of proofreading chaperones, contains a metal cofactor. Analysis of the purified FdhE showed it to be a rubredoxin that binds two iron atoms per FdhE monomer (123, 217). Iron ligation to FdhE is mediated through conserved cysteine residues. Amino acid substitutions of these conserved residues resulted in a complete loss of FdhE activity, indicating that iron binding was essential for FdhE activity and/or stability (123).

The Tat Pathway and Folding Quality Control

Folding quality control refers to the rejection of unfolded and malfolded substrate proteins by the Tat apparatus and is distinct from the chaperone-mediated proofreading of some Tat substrates described above. The term was first introduced by DeLisa et al. (54). These authors demonstrated that reduced PhoA alkaline phosphatase, a non-natively folded protein in which structural disulfide bonds have not formed, was not transported when targeted to the E. coli Tat machinery by a Tat signal peptide. Crucially, however, mutations that increase the oxidizing capacity of the cytoplasm and  permit formation of the native disulfide bond in PhoA allowed PhoA export by the Tat pathway. DeLisa and coworkers proposed that the Tat machinery has an inbuilt folding quality control mechanism that allows discrimination between folded and unfolded substrates, possibly through the detection of hydrophobic regions present in malfolded proteins. Experiments using an in vitro Tat protein transport system confirmed that only the oxidized form of PhoA is translocated. Moreover, molecular contacts between the substrate signal peptide and the TatBC components of the Tat system (see below) were reduced when the PhoA passenger protein was unfolded. This finding suggests that TatBC play a role in sensing the folding state of the substrate protein (152).

The suggestion that the Tat machinery has an intrinsic quality control mechanism has been called into question by Richter and Bruser (170). These authors found that Tat-targeted unfolded PhoA was specifically toxic to E. coli, probably due to a partial dissipation of the transmembrane proton gradient. They proposed that the Tat machinery cannot distinguish between folded and unfolded forms of PhoA and that the toxic effect of Tat-targeted unfolded PhoA arises from release of the protein into the membrane during translocation (170). Experiments with both the bacteria and plant thylakoids have demonstrated that the Tat machinery can transport unstructured proteins provided that they are small and hydrophilic in nature. Thus unfolded proteins per se are not incompatible with transport by the Tat system (44, 171). However, the introduction of just a short stretch of hydrophobic amino acids into the unstructured protein substrate is sufficient to stop translocation. It was proposed that the Tat system does not have intrinsic quality control features but that the hydrophobic stretches on an unfolded protein partition into the membrane lipids during transport (171).

A study by Matos et al. (131) suggested that quality control of incorrectly assembled Tat substrates depends on proteolysis, occurs at a late stage during Tat transport, and requires the TatA and TatE proteins. An additional study also implicated TatD in this process (130). Technical issues with these studies have, however, recently been identified. When these technical shortcomings were corrected, the proposed Tat-dependent proteolysis of incorrectly folded substrates could not be substantiated (120). Clearly, further work is required to clarify the existence and nature of any Tat-dependent folding quality control process.

THE TAT PROTEIN EXPORT MACHINERY

Organization of the tat Genes

The tatA, tatB, tatC, and tatE genes encode components of the Tat machinery in E. coli and Salmonella. The tat genes are encoded at two separate locations on the chromosome. The gene organization around the tat loci is similar in the two organisms, although with some variation in inter-gene spacing (Fig. 5).

Fig. 5Genetic organization of the tatA and tatE loci in E. coli K-12 and S. enterica subsp. enterica serovar Typhimurium LT2. (A and C) Genetic context of the tatA operon in E. coli K-12 (A) and S. enterica LT2 (C). The distance, in base pairs, between genes is indicated. (B and D) The mRNA between tatC and tatD in E. coli (B) and S. enterica LT2 (D) can potentially fold into a stem-loop structure. (E and F) Genetic context of the tatE gene in E. coli K-12 (A) and S. enterica LT2 (C). The distance, in base pairs, between genes is indicated.

The tatA operon is found downstream of a cluster of genes involved in ubiquinone biosynthesis. In E. coli, transcription of tatA initiates from a promoter between the two operons and the transcriptional start site lies 37 bp upstream of the tatA start codon (103).  In E. coli the fourth gene of the tatA operon, tatD, has an overlapping stop codon with a convergent gene, rfaH, encoded on the opposite strand. rfaH is the structural gene of a transcription elongation factor. While the first three genes of the tatA transcription unit are very close together, there is a 41-bp gap between tatC and tatD  in both E. coli and Salmonella.

Analysis of the transcripts originating from the tatA operon in E. coli using Northern blotting revealed a minor transcript covering all four genes and a major transcript including just tatABC (211). Interestingly, the 41-bp tatC-tatD intergenic region of E. coli or Salmonella contains a potential stem-loop structure (reference 211; see Fig. 5). It is possible that the stem loop may act as a Rho-independent transcription terminator or that it stabilizes the 3′ part of the mRNA from RNase-mediated decay.

In both E. coli and Salmonella the tatE gene is monocistronic and bounded by lipA at one side (encoding lipoyl synthase) and the ybeM gene (encoding a predicted amidohydrolase) at the other (see Figure 5). Two major transcriptional start sites were identified 49 and 75 bp upstream of the E. coli tatE initiation codon (103).

Since many of the E. coli and Salmonella Tat substrates are induced under anaerobic conditions, it was anticipated that the Tat machinery would be similarly upregulated under these conditions to increase Tat transport capacity. However, lacZ fusion analysis of the E. coli tatA and tatE promoters showed that they are expressed at a relatively constant level under varying growth regimens (aerobic, anaerobic in the presence of nitrate, or fermentative growth), leading to the conclusion that the machinery is constitutively produced (103). Interestingly, although the level of transcription from tatA and tatE promoters was broadly similar, using translational lacZ fusions it became clear that the translation of tatE mRNA was much less efficient than that of tatA mRNA. The same conclusion was reached when lacZ was used to replace, in frame, the tatA, B, C, D, and E reading frames at their native chromosomal loci. Comparison of the β-galactosidase activity for each of these strains indicated that TatA is produced at a much higher level than any of the other Tat proteins, in at least a 20-fold excess over TatB, approximately 50-fold more than TatC, and 100- to 200-fold more than TatE (103).

The Tat Proteins

The subcellular location and predicted topology of the Tat proteins is shown in Fig. 6. All of the tat genes, with the exception of tatD, code for integral membrane proteins, and each of TatA, TatB, and TatC is known to form homo-oligomers (52, 164). As discussed in the introduction, the first component of the Tat machinery was identified in maize and was designated Hcf106. Three homologues of Hcf106 are encoded in E. coli and Salmonella by the tatA, tatB, and tatE genes. Of the three Hcf106 homologues, TatA and TatE are very closely related, with similar sizes and amino acid sequences. Early experiments by Sargent et al. (182) showed that these proteins are functionally equivalent. Thus, single deletions in either tatA or tatE are not sufficient to fully block the export of proteins with twin arginine signal peptides. However, a combined deletion of the two genes prevented export of all Tat substrates tested (182). In general, deletion of tatA alone results in a much more severe Tat transport defect than deletion of tatE, which is consistent with the observation that TatA is produced at much higher levels than TatE (103, 182). Complementation of the tatA/tatE mutant strain with plasmid-borne tatA or tatE resulted in similar levels of Tat transport being restored, consistent with the idea that the proteins perform equivalent roles in protein transport (184).

Fig. 6Predicted secondary structure and topological organization of proteins encoded by the E. coli tat genes.

The TatB protein is the third Hfc106 homologue produced by E. coli and Salmonella. It shares significantly less sequence similarity with TatA than TatE does. TatB is much longer than TatA or TatE, and it has a distinct role to these proteins in Tat transport. Mutation of tatB alone is sufficient to completely block export of native Tat substrate proteins (184, 209). Moreover, the tatB deletion cannot be complemented by overexpressed tatA or tatE. Likewise, the tatA/tatE mutant cannot be complemented by plasmid-borne tatB (184). It was demonstrated that when extremely sensitive Tat-dependent reporter proteins are used there is a very low level of Tat transport in the absence of tatB (24, 98). This finding prompted Blaudeck et al. (24) to isolate point mutations that enhanced Tat transport in the tatB deletion strain. All of the mutations that they isolated fell within the first six amino acids of TatA (24).

Many Gram-positive bacteria, such as Bacillus subtilis, and archaea encode TatA and TatC but not TatB. These organisms have minimal Tat translocases that use only TatA and TatC components where TatA is apparently bifunctional and playing the role of TatA and TatB (105, 106). It is likely that the ancestral Tat system comprised only TatA and TatC. It is proposed that TatB arose through a gene duplication event and that TatB serves to increase the efficiency of protein transport rather than being mechanistically essential (24, 216). The gene duplication giving rise to TatA and TatB may have occurred prior to the split between Gram-negative and Gram-positive bacteria since the high G+C Gram-positive bacteria have recognizable TatA and TatB proteins (and indeed the Streptomyces coelicolor tatA and tatB genes are able to restore good Tat transport activity to the E. coli tatA/tatE and tatB deletion strains, respectively [88]). The plant thylakoid orthologs of TatA and TatB are Tha4 and Hcf106. They are, however, more closely related in sequence to each other than either is to E. coli TatA or TatB. This observation suggests that TatA duplication followed by diversification into TatA-like and TatB-like proteins may have arisen on more than one occasion (216).

TatA, TatB, and TatE are all predicted to have a similar structure with single membrane-spanning α-helices at their N termini, followed by adjacent amphipathic helices and unstructured C termini (Figure 6). Application of the positive inside rule would suggest that the extreme N termini of the proteins are located at the periplasmic side of the membrane and that the C termini are in the cytoplasm. A cytoplasmic location for the C terminus of TatA and TatB has been confirmed by protease accessibility experiments, which showed that the C terminus of each protein was protease sensitive in inside-out, but not right-side-out, membranes (115, 160). However, studies by Gouffi et al. (74) using the compartment-sensitive marker fusions alkaline phosphatase (PhoA: generally only active in the periplasmic compartment) and β-glucuronidase (UidA: only active in the cytoplasm) fused to the end of the amphipathic helix of TatA gave results which suggested that this region of TatA was exposed at both sides of the membrane. While it cannot be ruled out that the presence of a large reporter protein fused to part of TatA alters its topology, experiments were also carried out where a TEV protease cleavage site was inserted after residue 53 of TatA and this site was also shown to be protease accessible from either side of the membrane. The authors proposed that the TatA amphipathic helix has a dual topology (shown schematically in panels B and C of Figure 7) and that topology changes of this region of TatA are associated with protein transport (74). A more recent publication, by Chan et al. (40), has fuelled further controversy regarding the topological orientation of TatA. The authors used cysteine accessibility experiments with single-cysteine-substituted TatA proteins to probe the localization of regions of TatA in whole cells. The authors saw labelling patterns indicating that the N terminus of TatA was localized in the cytoplasm, while the labeling of a cysteine at the end of the amphipathic helix changed when cells were treated with a protonophore (40). These findings led the authors to propose two alternative topologies of TatA shown in panels D and E of Figure 7, where the N terminus is located on the inside and the amphipathic helix shows the dual topology originally proposed by Gouffi et al. (74). There is, however, one further piece of evidence that would support an N-out topology for the transmembrane helix of TatA. The TatA protein from Providencia stuartii is synthesized with an inactivating N-terminal extension of seven amino acids. This extension is processed by a membrane-embedded protease of the rhomboid family (194). High-resolution structural and biochemical analysis of the E. coli rhomboid family protease GlpG (which is also able to process P. stuartii TatA) reveals that the active site is close to the periplasmic face of the membrane (126, 208), strongly suggesting that the N terminus of TatA is also localized at this side of the membrane.

A final controversy relating to the organization of TatA in E. coli cells is whether TatA has a solely membrane location or whether it can additionally exist in a soluble form in the cytoplasm. It should be noted that there is a similar controversy regarding the localization of the TatA proteins of B. subtilis, where it was observed that the TatA_d protein of this organism (the ‘d’ subscript referring to the fact that this TatA protein is required for the transport of PhoD) in addition to being found in the membrane also exists as a cytoplasmic form of high molecular weight (159, 210). Furthermore, when B. subtilis TatA_y (which is required for the transport of YwbN) was produced in E. coli, some of the protein was also observed in the cytoplasmic fraction, where it formed a large aggregate (of approximately 5 MDa [8]). A study by Berthelmann et al. (20) similarly showed some E. coli TatA in the cytoplasmic compartment when the the Tat components were overproduced. This TatA protein is associated with long hexagonal tubes seen in electron micrographs of the cells.  The tube-like structures were dependent upon TatC (but not TatB) for their formation. However, tube-like structures were not observed when the protein was synthesized at the native level. Indeed, fractionation experiments have shown that all of the native E. coli TatA protein has a membrane localization and that it cannot be extracted with urea or carbonate, indicating that TatA is a classical integral membrane protein (52).

The tatC gene is essential for protein transport by the Tat pathway (27). It encodes the largest and most highly conserved component of the Tat machinery. Bioinformatic analysis strongly suggests that bacterial TatC proteins have six membrane-spanning domains. However, initial experimental analysis of the topology of E. coli TatC suggested that the protein had only four transmembrane helices because reporter protein fusions to both predicted helices 4 and 5 indicated a periplasmic location (75). Nevertheless, subsequent reporter fusion studies supported the original six transmembrane helix model (12, 108). It is thought that the cytoplasmic domain between helices 4 and 5 plays a critical role in determining the topology of helix 4 and that the fusion used in the original study perturbs TatC topology (12, 108). Recent experiments that test the accessibility of cysteine substitutions in the same region of TatC confirm that E. coli TatC has six transmembrane domains (164).

TatD is the only water-soluble protein that is encoded by the E. coli tat genes. Homologues of TatD are widely distributed and are found in organisms that do not encode other Tat components. Indeed, there are two further homologues in E. coli coded by ycfH and yjjV. Deleting all three tatD homologues in E. coli has no significant effect on the Tat-dependent localization of several redox enzymes or the export kinetics of the Tat substrate SufI  (211). A report that Tat substrate precursor proteins are stabilized in a tatD null strain (130) has been shown to be erroneous (120). Purified TatD was shown to exhibit deoxyribonuclease activity. Consistent with this finding a tatD mutant was shown to have a twofold increase in the number of RecA-GFP foci, suggesting that there are significantly more double-stranded DNA breaks when TatD is lacking (38, 165, 211). Taken together, these observations provide no evidence that TatD plays any role in protein export by the Tat pathway.

Fig. 7Alternative topologies for the E. coli TatA protein. (A) The amino acid sequence of TatA is shown, with predicted secondary structure above. α-helical regions are represented as cylinders, and β-sheet is indicated as arrows. The essential glycine residue at the boundary between the transmembrane and amphipathic α-helices is boxed. (B to E) Alternative topological arrangements of the TatA protein proposed in different studies (see main text).

MECHANISM OF PROTEIN EXPORT

Mutational Analysis of Tat Components

Extensive mutational analysis has been carried out on the E. coli Tat components to define functional domains and to investigate the role of conserved amino acid residues in Tat transport. The conclusions drawn from numerous mutational studies on Tat components are sometimes conflicting, with similar mutations having quite different effects depending upon the reporter proteins used to assess functionality of the Tat system and the plasmid vectors from which Tat components are expressed. The stoichiometry with which the different Tat components are produced can also have significant impact on Tat transport activity. This can be most clearly seen in the case of TatB, where multicopy tatB is extremely detrimental to cells and can almost completely inactivate the Tat activity of a wild-type (tat⁺) strain (184).

It has been demonstrated that the transmembrane helix of TatA is essential for the membrane association and oligomerization of the protein (52, 160). Further genetic truncation analysis of TatA has shown that the transmembrane domain and adjacent amphipathic helix are essential for function but that the unstructured C-terminal domain is not and can be removed without abolishing function (although stability of the protein was quite markedly reduced [114]). Consistent with the conclusions from the truncation analysis, a positive selection screen for mutations that inactivated the function of TatA only identified mutations within the first 42 amino acids of TatA corresponding to the transmembrane and amphipathic helices (89).  Several site-directed mutagenesis studies as well as cysteine scanning mutagenesis have indicated that the amphipathic helix of TatA is very intolerant to mutation (9, 10, 77, 87). The only residue that is absolutely conserved between TatA and TatB proteins is an invariant glycine (G21 in E. coli TatA and shown boxed in Figure 7A) that is always found at the junction between the transmembrane and amphipathic helices. This glycine is essential for TatA activity and even substitution for alanine is not well tolerated (87, 89). This finding suggests that a degree of flexibility between the two helices is essential for TatA function and might be consistent with the suggestion that under some circumstances the transmembrane and amphipathic helices are arranged as a helical hairpin (as shown in panels C and E of Figure 7).

Genetic truncation of TatB also showed that the protein could be quite extensively shortened from the C terminus without significant loss of function (even a truncated protein comprising only the first 41 amino acids retained very low but detectable transport activity), implicating the transmembrane and amphipathic helices as the minimal functional unit (114). However, unlike TatA, in which the amphipathic helix in particular is quite intolerant to mutation, no single mutation in TatB is sufficient to fully inactivate the protein (9, 10, 87, 115). The mutations with the largest effect on TatB function all fall within the hinge region where the glycine at position 21 and proline at 22 (E. coli numbering) are invariant in TatB proteins. However, substitution of G21 of E. coli TatB, even for proline, did not completely block TatB function (9, 87). Likewise mutagenesis of the conserved proline at residue 22 for leucine reduced Tat transport activity to a low level, but functionality was not completely lost. These observations are consistent with the idea that, although they share similar features, TatA and TatB have rather different functions.

Domain-swapping experiments where the transmembrane helices of TatA and TatB were exchanged indicated that a construct where the transmembrane helix of TatA was fused to the amphipathic and C-terminal regions of TatB had some very low level of functionality as either a TatA or a TatB (but could not substitute for the loss of both proteins). The converse fusion had no activity as either a TatA or a TatB (114), suggesting that features in the transmembrane helix of TatA are essential for its function.

Several studies have examined the effects of substituting highly conserved residues in E. coli TatC. The reported effects of these substitutions vary widely (4, 33, 92). There are very few substitutions that appear to consistently inactivate TatC. However, one such substitution is the replacement of the phenylalanine at position 94 of TatC with alanine. This residue is predicted to fall within the first cytoplasmic loop of TatC, a region of the protein that shows particularly high sequence conservation. This region of TatC has been speculated to form part of the binding site for the twin arginine signal peptide (33).  Indeed, individual substitutions of F94A, Y100A, and E103A all prevented binding of substrates to TatC-containing complexes in an in vitro assay (92). A role for the first cytoplasmic loop of TatC in signal peptide recognition is also suggested by the work of Strauch and Georgiou (195), who isolated suppressor mutations that permitted the export of a Tat substrate with a twin lysine motif instead of twin arginines. Most of the mutations they isolated fell within this loop region, including one where F94 was substituted for serine. The suppressor mutations isolated in this study are almost completely blocked for the export of native Tat substrate proteins, suggesting that the mutations served to switch the specificity of the Tat machinery from twin arginine-containing signal peptides to twin lysine.

A second study looking for suppressors of variant Tat signal peptides employed a reporter protein in which the consecutive arginines in the signal peptide were changed to a lysine-glutamine pair. Mutations that permitted recognition of this variant signal peptide were shown to fall within the N-terminal cytoplasmic domain of TatC and also towards the N terminus of TatB (110). However, rather than switching the sequence recognized by the Tat system the mutants appear to retain the ability to interact with  standard twin arginine-containing signal peptides. Thus, these mutants confer a relaxed specificity on the Tat machinery and may be analogous to the prl mutations of Sec components (63).

Organization of the Tat Export Machinery

Various chemical cross-linking experiments have shown that the TatA, TatB, and TatC components all form homo-oligomers in E. coli membranes. Membranes of E. coli overproducing these three Tat proteins and treated with either of the cross-linking reagents disuccinimidyl suberate or formaldehyde showed clear homo-oligomers of TatA and of TatB (52). TatB dimers were observed, but higher-order oligomers of TatA could be cross-linked, up to at least tetramers (52, 87, 89). The same homo-oligomeric species could be detected whether the individual Tat protein was produced in the absence or presence of other Tat components, suggesting that homo-oligomerization does not depend on interactions with other Tat proteins (52).

Disulfide mapping studies were used to define regions of TatB that were involved in self-interaction. It was shown that one face of the transmembrane helix of TatB makes significant contact with the same face of neighboring TatB molecules. When doubly cysteine-substituted TatB proteins were analyzed under oxidizing conditions up to TatB tetramers were detected (115). Consistent with the oligomerization of TatB detected by cross-linking studies overproduction of TatB in the absence of other Tat components resulted in a membrane-bound complex of TatB that, as judged by blue native polyacrylamide gel electrophoresis (BN-PAGE), corresponded to at least tetramers (11, 150).

Disulfide mapping studies of TatA containing single cysteine substitutions in the transmembrane helix indicated that extensive self contacts were formed between TatA protomers. These self-self interactions covered more than one face of TatA. Indeed, single cysteine substitutions all along the length of the essential region of TatA (from amino acids 2 to 43) each gave significant levels of cross-linked dimer consistent with the protein being in intimate contact with other TatA molecules (77). Self-interaction of TatA has also been detected after gel filtration analysis of solubilized native membranes of E. coli, where the protein was shown to elute over a wide molecular mass range, and by BN-PAGE analysis of solubilized membranes overexpressing tat components, where ladders of bands of TatA ranged from less than 60 kDa to in excess of 500 kDa (132, 146). Additional in vitro (electron microscopy analysis of purified TatA [71]) and in vivo (single-molecule fluorescence studies on YFP-tagged TatA [112]) studies have confirmed that TatA has a strong propensity for homo-oligomerization and can form large assemblies.

Evidence for a multimeric arrangement of TatC has also been obtained through disulfide cross-linking and BN-PAGE analysis. Disulfide mapping studies revealed several positions in TatC where mutation to cysteine gave strong homodimer formation after oxidation. One of these mutations, a substitution of glycine at residue 144 which falls in the second periplasmic loop of TatC, gave an almost quantitative cross-link that was also detected if the cross-linking reaction was undertaken on ice. This finding was taken as evidence that the cross-link was between TatC monomers located within the same complex (164). BN-PAGE analysis of detergent-dispersed membranes overproducing TatC in the absence of other Tat components showed that the protein forms a discrete complex of estimated mass approximately 220 kDa and can therefore exist as a stable homo-multimer (11, 150).

Although the TatB and TatC proteins clearly show self-self interactions, these associations are within the context of a heteromeric complex between these two proteins. The E. coli TatBC complex was first isolated by Bolhuis et al. (28) and has a mass of approximately 600 kDa by gel filtration analysis (28, 51). TatB and TatC are present at an equimolar ratio within the complex and a covalent fusion where the C terminus of TatB is fused to the N terminus of TatC is functional (28). BN-PAGE analysis of this complex shows that is has an estimated mass of around 440 kDa (147, 170). Thus, the complex contains several copies of TatB and TatC, probably between six and eight (e.g., 28, 51, 147, 170, 199). Given the size constraints on the TatBC complex, the TatB bundle suggested by the disulfide cross-linking experiments of Lee et al. (115) would most likely be in the center of the TatBC complex rather than on the outside.

When all of the Tat components are overproduced, a small and variable amount of TatA co-purifies with TatBC. However, at native levels of expression, or in a strain devoid of TatA/TatE, the TatBC complex is stable and is of a size similar to that of the TatA-containing complex (11, 132, 150). The isolated E. coli TatBC complex has been analyzed by negative stain electron microscopy and shows oval-shaped particles surrounding a small central cavity (147, 199). The complexes are rather heterogeneous and were classified into two different mass classes, one with an estimated protein mass of 380 kDa and one of 420 kDa (199).

Low-resolution electron microscopy structures are also available for the purified TatA complex (71, 183, 199). These show ring-shaped structures with a range of different diameters surrounding a large central cavity. The variability in diameter results from differences in the number of TatA subunits present in the complexes and suggests that complexes change in size by gain or loss of TatA protomers (71). The rings resemble channels with a lid covering one side and a more accessible open face, which is proposed to face the periplasmic side of the membrane (71). The largest complexes were estimated to contain of the order of 30 TatA monomers surrounding a central cavity of 65 to 70 Å. This would be sufficiently wide to accommodate even the biggest Tat substrates (71) and is consistent with the idea that TatA comprises the Tat transport channel.

The Tat Transport Cycle

Figure 8 shows a model for the operation of the Tat system based on current experimental data. It is proposed that when the Tat system is not operating (i.e., under resting conditions [Fig. 8, top]) the Tat translocon is not assembled and the Tat components exist as separate TatA and TatBC complexes. According to the model the TatBC complex is the same as the complex characterized by purification (above) and contains multiple copies of each of TatB and TatC (note that only one copy of each is depicted in Fig. 8 for clarity). The nature of TatA under resting conditions is more speculative, but recent single-molecule fluorescence imaging results suggest that TatA may exist as tetramers. This is consistent with the tetrameric organization of the chloroplast TatA orthologue, Tha4, inferred from disulfide cross-linking experiments (49, 112).

Fig. 8The Tat protein transport cycle. (Top) Under resting conditions the TatBC complex and tetrameric units of TatA exist as separate assemblies in the membrane. Step 1. A folded substrate protein docks at the TatBC complex, binding by virtue of its twin arginine signal peptide. Step 2. TatA tetramers assemble onto the substrate-bound TatBC complex to form the transport channel in a process requiring the proton motive force (Δp). Step 3. The substrate is transported across the membrane through a channel formed by TatA. It is not known whether the proton motive force is needed to drive this step. Step 4. The translocated substrate is released, its signal peptide is cleaved, and the TatABC complex dissociates to give the TatBC complex and separate multimers of TatA. Note that, for clarity, only one TatBC pair in the TatBC complex and only some of the necessary TatA molecules are depicted.

Step 1 in the cycle is the binding of a precursor protein to the TatBC complex. It has been experimentally demonstrated that Tat signal peptides and full-length precursors bind to both purified and membrane-bound TatBC and that these interactions are independent of TatA (e.g., 1, 45, 51, 170, 199). Genetic and biochemical experiments have indicated that the twin arginine motif of the signal peptide is recognized primarily by TatC (1, 110, 195). Binding of substrates to TatBC does not require the proton gradient, although there are indications from the thylakoid Tat system that energization can change both the affinity and mode of signal peptide binding (70). It is known that Tat precursors can interact with phospholipids in the membrane (7, 188). It is thus plausible that substrate proteins associate nonspecifically with the membrane bilayer before binding to TatBC.

Substrate-bound TatBC complexes have recently been isolated from cells producing saturating levels of the substrate protein SufI (132, 199). Single-particle electron microscopy analysis of these complexes (199) showed either one or two molecules of the SufI substrate bound at the periphery of the complex. In complexes containing two SufI molecules the two substrate molecules were always located close together rather than being randomly distributed around the circumference of the particle. The nonrandom and substoichiometric (with respect to number of TatC subunits present) binding of substrate to TatBC led to the suggestion that the TatBC complex binds substrate molecules anti-cooperatively. Comparison of substrate-bound and unliganded TatBC complexes suggested that TatBC undergoes a structural rearrangement on substrate binding and that this rearrangement most likely involves the loss of one or more TatB and/or TatC subunits. Recently, it was demonstrated that multiple substrates could bind simultaneously to the TatBC-analogous complex in plant thylakoids and be collectively transported, suggesting that multiple precursor binding sites may act in concert during protein translocation (125).

Step 2  is primarily supported by experiments on the Tat system of plant thylakoids. It was demonstrated that the plant TatA orthologue could be cross-linked to the thylakoid equivalent of the TatBC complex only when substrate was present (139), suggesting that TatA specifically associates with the TatBC-substrate complex. The cross-linking was only seen when membranes were energized, indicating a requirement for the transmembrane proton electrochemical gradient at this stage of the cycle. Similar conclusions can be drawn from experiments in an in vitro E. coli system in which photoaffinity cross-links between a substrate signal peptide and TatA were only detected in the presence of the proton electrochemical gradient (1). The number of crosslinks that can be formed between thylakoid TatA molecules also increase in the combined presence of substrate and a proton gradient (49, 50), suggesting that interaction with the substrate-bound TatBC complex leads to TatA polymerization. Support for this model comes from direct imaging of TatA-YFP fusion proteins in E. coli cells, which shows that TatA can exist in either dispersed or higher oligomer forms and that the presence of the higher oligomer state requires TatBC (112). The structures of TatBC with bound substrate indicate that substrates bind to the outside of the complex, which would be consistent with the idea that TatA assembles on the periphery of the TatBC complex.

Polymerization of TatA has been proposed to be crucial to providing a pathway for the substrate protein across the membrane. In the bespoke channel model a TatA channel forms only in the presence of substrate and dynamic variation of the TatA polymerization state maintains a tight seal around substrates of different sizes during transport (71). However, no difference between the TatA oligomeric state triggered by a substrate protein and that triggered by a signal peptide was detected by cross-linking in thylakoids, arguing against a correlation between TatA oligomeric state and substrate size (49, 50). Alternative proposals suggest that TatA interacts with the phospholipid membrane and that concentrating TatA by polymerization alters local membrane bilayer structure to allow substrate transport (32, 44, 46, 49, 50, 77).

Step 3 of the cycle is the transport of substrate across the membrane mediated by polymerized TatA. The active form of TatA may be analogous to the channel-like structures formed by TatA in detergent solution (71). The best evidence that TatA forms the transport pathway comes from the work of Panahandeh et al. (152). Here the authors examined the in vitro transport of a reporter protein in which the mature domain of E. coli PhoA alkaline phosphatase was fused to the TorA signal peptide. Although the experiments were carried out under oxidizing conditions, which should result in disulfide-stabilized folding of PhoA, it was shown that a proportion of the passenger protein stalls during translocation, forming a protease-resistant fragment. This protease-resistant fragment could be extracted from the membrane by detergents and was shown to be co-immunoprecipitated with TatA, but not TatB or TatC, antibodies (152). This suggests that the stalled substrate protein is bound to TatA.

It is not known whether Tat transport has an energy requirement after the TatA polymerization (step 3 in Figure 8). Evidence from Bageshwar and Musser (6) has shown that there are (at least) two steps during protein translocation that require the electrical potential component of the proton motive force. One of these occurs early in the transport cycle and requires a relatively large potential, while the second step requires a long-duration potential of lower magnitude. Theoretically, both of these steps could occur during translocon assembly in step 2 (for example if the membrane potential confers a conformational change on TatA such as realignment of the amphipathic helices in addition to being required for assembly of TatA tetramers into larger multimers). Alternatively, it is possible that the second, slower, Δψ-dependent step is required to drive movement of the substrate protein across the membrane.

Although there is mounting evidence that TatA is able to form channel-like oligomers, it is not clear whether these surround an aqueous compartment or whether it is filled with bilayer lipid. Some models of Tat transport propose that substrates pass directly through the lipid bilayer rather than through an aqueous channel and that the assembly of TatA proteins into a patch close to the substrate induces a weakening in the membrane that allows the substrate to be pulled through the bilayer, probably by the action of TatC (32, 44). An alternative model assumes that a lipid-filled channel forms from the transmembrane helices of TatA and that the amphipathic helices of TatA fold into this channel like a trap-door in response to a pulling force on the substrate (46, 50, 77). This model is particularly attractive since the polar faces of the amphipathic helices could provide an aqueous passage of the substrate across the membrane, with the lipids perhaps dissipating into the bilayer between the transmembrane helices of TatA.

The final stage in the transport cycle (step 4 in Figure 8) results in release of mature substrate at the trans side of the membrane. This step is accompanied by cleavage of the signal peptide by signal peptidase (124, 215), which is then presumably degraded by signal peptide peptidase (94). According to the model the translocase dissociates to give the TatBC complex and separate smaller bundles of TatA. Supporting evidence for translocase dissociation again comes from in vitro studies with plant thylakoids, where Mori and Cline (139) demonstrated that cross-linking of the TatA orthologue to the thylakoid TatBC complex was lost once the substrate had transited the Tat pathway.

The precise point at which the signal peptide is cleaved is uncertain. There is no evidence that signal peptidase interacts with any of the Tat components, and it is unlikely that the active site of the peptidase could access the signal peptide cleavage site until transport is complete. In vitro transport assays using inner membrane vesicles derived from E. coli strains overproducing TatABC show that not all of the transported substrates have cleaved signal peptides, presumably due to insufficient signal peptidase being present in the vesicles. This noncleavage of transported substrates does not, however, appear to affect additional substrate molecules being transported, suggesting substrate molecules with uncleaved signal peptides are still released by the translocase after transport rather than remaining bound and blocking the substrate binding site (e.g., 2, 215). Likewise, if Tat substrates are engineered to have noncleavable signal peptides, they are successfully transported across the membrane where they remain membrane anchored at the trans side.  Again there is no evidence that they remain jammed in the Tat machinery (56). One possibility is that substrates are released from the Tat machinery after translocation, remaining anchored to the bilayer by the signal peptide, after which they are cleaved. However, this idea does not fit well with data from protein transport by the thylakoid Tat system, where it has been shown that partially transported and even some nontransported substrates have their signal peptides removed (e.g., 44, 56). Clearly, the timing of signal peptide cleavage is one area that awaits further experimental analysis.

FUTURE DIRECTIONS

While much progress has been made in characterizing the Tat pathway, many outstanding issues remain. Most pressing is an understanding of the protein transport mechanism at the molecular level, which would be greatly aided by high-resolution structural analysis of Tat components. Structural analysis is likely to be particularly challenging for TatA which forms highly heterogeneous complexes and may undergo topological inversion. Both TatA and TatB have extensive regions that are predicted to be unstructured which may further hamper their structural analysis. Additional mechanistic study is required to decipher the roles of the proton electrochemical gradient in driving transport by the Tat pathway.

One of the most controversial issues yet to be resolved is whether the Tat machinery can discriminate between folded and unfolded substrates. This issue is likely to be an area of intense research over the coming years. While the existence of Tat quality control is still equivocal, chaperone-mediated proofreading of Tat substrates is unambiguous, although it remains to be seen whether all cofactor-containing Tat substrates have a dedicated proofreading chaperone. The mechanism of signal peptide binding and release by proofreading chaperones is currently unclear; it would be of interest to determine whether interaction with the Tat machinery is required to dislodge bound chaperones from Tat signal peptides.

The Tat pathway is unique in its ability to transport folded proteins across the bacterial cytoplasmic membrane. This feature of the Tat pathway has not gone unnoticed by biotechnologists, and it offers great potential for exploitation, for example for therapeutic protein production (e.g., 60, 64, 65, 154, 201; see reference 31 for a recent review of this area). Over the coming years it is anticipated that the unique features of the Tat pathway will be capitalized upon in novel and exciting ways. Finally, it is becoming increasingly clear that the Tat pathway is essential for the virulence of many pathogenic bacteria (including both E. coli and Salmonella [137, 162]). The requirement of the Tat pathway for bacterial virulence raises the possibility that small-molecule inhibitors of the Tat system could ultimately be developed into novel antimicrobial compounds. It is likely that the many challenges and opportunities for the study and exploitation of the bacterial Tat system will occupy biologists for years to come.

ACKNOWLEDGMENTS

We thank everyone in their laboratories, past and present, for contributing to research on the Tat system. Max Fritsch is thanked for his assistance in preparing some of the figures; Dr. Govind Chandra is thanked for bioinformatic help.