Characterization of a pre-export enzyme-chaperone complex on the twin-arginine transport pathway.

The Tat (twin-arginine translocation) system is a protein targeting pathway utilized by prokaryotes and chloroplasts. Tat substrates are produced with distinctive N-terminal signal peptides and are translocated as fully folded proteins. In Escherichia coli, Tat-dependent proteins often contain redox cofactors that must be loaded before translocation. Trimethylamine N-oxide reductase (TorA) is a model bacterial Tat substrate and is a molybdenum cofactor-dependent enzyme. Co-ordination of cofactor loading and translocation of TorA is directed by the TorD protein, which is a cytoplasmic chaperone known to interact physically with the TorA signal peptide. In the present study, a pre-export TorAD complex has been characterized using biochemical and biophysical techniques, including SAXS (small-angle X-ray scattering). A stable, cofactor-free TorAD complex was isolated, which revealed a 1:1 binding stoichiometry. Surprisingly, a TorAD complex with similar architecture can be isolated in the complete absence of the 39-residue TorA signal peptide. The present study demonstrates that two high-affinity binding sites for TorD are present on TorA, and that a single TorD protein binds both of those simultaneously. Further characterization suggested that the C-terminal 'Domain IV' of TorA remained solvent-exposed in the cofactor-free pre-export TorAD complex. It is possible that correct folding of Domain IV upon cofactor loading is the trigger for TorD release and subsequent export of TorA.


INTRODUCTION
The Tat (twin-arginine translocation) protein transport pathway exports proteins across the cytoplasmic membranes of bacteria and archaea, and the thylakoid membranes of plant chloroplasts [1,2]. Protein substrates of the Tat pathway are folded before translocation and are targeted to the Tat machinery by cleavable N-terminal signal peptides containing an almost invariant pair of arginine residues [3]. Transport by the Tat pathway is energized solely by the protonmotive force [1,2].
The model bacterium Escherichia coli K-12 produces 28 proteins bearing twin-arginine signal peptides. Approximately two-thirds of these proteins are known or predicted to contain redox cofactors and some of them also form complexes with partner subunits that lack a signal peptide before export [4]. The complexity of these Tat substrates necessitates a pre-export 'proofreading' process to ensure that only folded and assembled proteins are presented for translocation. One hypothesis involves dedicated chaperone proteins that interact directly with twinarginine signal peptides to prevent premature targeting of the substrate before cofactor insertion and binding of any protein partners [5].
The E. coli TMAO (trimethylamine N-oxide) reductase TorA is a soluble periplasmic enzyme containing the MGD (molybdopterin guanine dinucleotide) cofactor at its active site [6]. TorA is a well-characterized substrate of the Tat pathway; indeed, transport of TorA to the periplasm was found to be blocked in a mutant strain that was unable to synthesize MGD, which led to the original hypothesis that a transport pathway for folded proteins may exist in bacteria [7]. The TorA protein is encoded by the torCAD operon ( Figure 1A) where TorC is a haem-containing quinol oxidase and TorD is a cytoplasmic protein [8]. Cellular levels of enzymatically active TorA are significantly reduced in a strain lacking the torD gene [9], and the TorD protein was found subsequently to interact with an unfolded (heat-denatured) form of the mature TorA enzyme [9]. This initial work therefore established that there was a clear binding site for TorD within the mature region of the protein and suggested a role for TorD in loading the MGD cofactor into the TorA apoenzyme [9,10]. In addition to this, a subsequent genetic screen identified TorD as a binding partner for the TorA twin-arginine signal peptide [11]. This was confirmed in vitro when recombinant TorD was shown to bind directly to the isolated TorA signal peptide with a 1:1 stoichiometry and an apparent K d of ∼ 59 nM [12,13]. The binding epitope for TorD has been mapped to the C-terminal end of the TorA signal peptide, close to its junction with the mature portion of the enzyme [12,14]. Interestingly, binding of the TorA signal peptide by TorD seems to be an activity not strictly connected with cofactor loading, since co-expression of torD can enhance the export of synthetic signal peptide fusion proteins on the Tat pathway [11,15]. The TorD protein itself is 22.5 kDa and is known to exist in equilibrium between monomeric and dimeric forms [16][17][18], have a low level of GTPase activity associated with the dimeric form [18], and have the ability to bind the MGD cofactor in vitro [19]. However, although it is clear that the monomeric form of TorD can bind the isolated TorA signal peptide [12,13], the physiological role and significance of TorD dimerization is not clear.
Current models for TorA biosynthesis assume that TorD interacts with the TorA precursor by binding at two distinct sites: one being within the twin-arginine signal peptide with the other, uncharacterized, site lying elsewhere in the mature portion of the protein [5,20]. These two binding events are thought to not only delay the Tat transport event, but also actively facilitate efficient insertion of the MGD cofactor into the apoenzyme. This pre-export 'Tat proofreading' process would be completed by the controlled release of the bound chaperones and subsequent translocation of the folded active enzyme to the periplasm. However, it is not known whether each of the TorD-binding sites on the TorA apoenzyme are bound by separate TorD monomers or a single TorD dimer, or whether the TorD monomer itself contains two distinct TorA-binding sites. Moreover, the degree of folding of the TorA precursor when bound by TorD has not been explored, and it remains a mystery how release of the chaperones is ultimately triggered.
In the present study, a stable and cofactor-free complex of TorD with the TorA apoenzyme has been isolated, both with and without the twin-arginine signal peptide. Using a combination of biophysical and biochemical approaches, we have demonstrated that a single TorD monomer binds directly to a single TorA apoprotein regardless of the presence of the signal peptide. This suggests that a single TorD monomer recognizes two binding sites on the TorA polypeptide and that these epitopes are nearby or overlapping. Finally, we have demonstrated that binding of TorA by TorD changes the conformation of apo-TorA in relation to the mature active enzyme by inducing exposure of the C-terminal domain of the TorA protein.

Plasmid and strain construction
The torAD genes form a transcription unit on the E. coli chromosome with overlapping stop and start codons. To cooverproduce TorA and TorD His , this natural translational coupling was retained. DNA encoding SPA (sequential peptide affinity)tagged TorD along with TorA was amplified from the chromosome of E. coli strain DY330 torD-SPA [21,22] using oligonucleotides TorAMfeIup (5 -GCGCCAATTGCGATAAGAAGGAAGAAA-AATAATG-3 ) and SPASphIdown (5 -GCGCGCATGCCTACT-TGTCATCGTCATCC-3 ). The resulting PCR product was digested with MfeI and SphI and cloned into pQE80 digested previously with EcoRI and SphI. To replace DNA encoding the SPA-tagged variant of TorD with a His-tagged variant, the pQE80torADSPA plasmid was digested with NheI (which removes DNA covering the 3 -end of torD, the SPA-tag-coding sequence and a small region of the vector). Plasmid pQE60torD [13] which codes for a C-terminally His-tagged TorD was also digested with NheI and the fragment covering the 3 -end of torD along with the His-tag-coding sequence was ligated into the digested pQE80torADSPA vector. Resultant clones were analysed by sequencing to ensure that the NheI fragment had been cloned into the correct orientation, and the construct was designated pQE80TorADhis ( Figure 1B).
To produce a C-terminally truncated TorA along with TorD His , a fragment of torA terminating at codon 675 and followed by a stop codon was amplified with oligonucleotides TorAfor (5 -GACCTCAACCTGCCGCTTTGC-3 ) and TorA675STOP HindIII (5 -GCGCAAGCTTATTTCTCAAACCACATCGGAT-GCCC-3 ) and digested with NruI (the recognition site for which is present in the torA gene) and HindIII. A second PCR product covering the torD ribosome-binding site and a fragment of the torD gene was amplified with oligonucleotides TorDHindRBS (5 -GCGCAAGCTTGCAGGTGAAATCATGACCACG-3 ) and TorDrev (5 -GGTCATCAGAAACAGGCCGC-3 ) and digested with HindIII and SacII (the recognition site for SacII occurs naturally in torD). A three-way ligation was carried out with these two fragments and pQE80TorADhis that had been digested previously with NruI and SacII. The resultant clone was designated pQE80TorAtruncDhis ( Figure 1D).
Strain FTH007 encodes a C-terminally hexahistidine-tagged TorA produced from the native chromosomal location and was constructed as follows. A 582 bp fragment of torA DNA finishing immediately before the torA stop codon was amplified with oligonucleotides TorAHis1 (5 -GCGCCCATGGCTCGAAAAC-TATCGCCGATATGAAC-3 ) and TorAHis2 (5 -GCGCAGAT-CTTGATTTCACCTGCGACGCGGGAAC-3 ), digested with NcoI and BglII and cloned into similarly digested pFAT210 [23]. DNA covering the torD ribosome-binding site and approximately 600 nucleotides of downstream DNA was amplified with oligonucleotides TorAHis3 (5 -GCGCAAGCTTTGTTCCCGC-GTCGCAGGTGAAATC-3 ) and TorAHis4 (5 -GCGCTCTAG-AGACGTTATCTGTTTTGGTGGTCGC-3 ), digested with XbaI and HindIII and cloned into the above construct. DNA covering the torAhis allele was subsequently excised by digestion with XbaI and KpnI and cloned into similarly digested pMAK705 [24]. The torAhis allele was then moved on to the chromosome of strain DSS401 (as MC4100, dmsABC::Kan R ) [25] to give strain FTH007.
All constructs were fully sequenced to ensure that no mistakes had been introduced inadvertently during the cloning procedure.

Preparation of recombinant proteins
All plasmid-encoded TorA/TorD His proteins were overproduced and purified in an identical manner. A single colony of freshly was cultured overnight in 5 ml of LB medium supplemented with 100 μg/ml ampicillin. This provided the inoculum for a 500 ml culture, which was grown aerobically at 37 • C and at 140 rev./min in a baffled 2 litre flask to a D 600 of 0.6 before induction of recombinant protein production by addition of 2 mM IPTG (isopropyl β-D-thiogalactopyranoside), followed by a temperature shift to 18 • C. After 16 h, cells were harvested and resuspended (10 ml per 1 g of cells) in buffer A [50 mM Tris/HCl (pH 7.5), 200 mM KCl, 1 mM DTT (dithiothreitol) and 25 mM imidazole]. Protease inhibitor [EDTA-free Complete TM protease inhibitor cocktail (Roche)], lysozyme and DNase were added to the resuspension and cells were lysed using an Emulsiflex C3 high-pressure homogenizer. Cell debris was removed by centrifugation at 18 850 g for 15 min, followed by removal of membranes by ultracentifugation at 45 000 rev./min for 1 h using a Beckman Ti70 rotor. All centrifugation steps were conducted at 4 • C. The supernatant obtained following ultracentrifugation was filtered through a 0.22-μm-pore-size membrane filter (Millipore) before being subjected to IMAC (immobilized metal-ion-affinity chromatography) as follows. The extract was loaded on to a 5 ml HisTrap column (GE Healthcare) equilibrated with buffer A. After extensive washing with buffer A, hexahistidine-tagged proteins were subsequently eluted using a gradient of 25-500 mM imidazole in buffer A. TorA/TorD His -containing fractions were identified by SDS/PAGE (12% gels), pooled and concentrated to 2 ml using a Vivaspin 20 spin concentrator (10 kDa cutoff, Sartorius). The concentrated sample was then loaded on to a calibrated HiLoad 16/60 Superdex 200 Prep Grade (GE Healthcare) size-exclusion column equilibrated with buffer A (50 mM Tris/HCl, pH 7.5, 200 mM KCl and 1 mM DTT), with TorA/TorD His -containing fractions again identified by SDS/PAGE (12% gels).
Chromosomally encoded TorA His was purified from cells that had been cultured anaerobically in 5 litre capped bottles that had been filled with LB medium supplemented with 0.4 % TMAO, 0.5 % glycerol and 50 μg/ml kanamycin. After inoculation with a 5 ml pre-culture, bottles were incubated for 16 h without shaking at 37 • C. Cells were harvested, washed and resuspended to a final concentration of 10 ml/g of cells in 50 mM Tris/HCl (pH 7.5) and 40 % (w/v) sucrose. The suspension was then supplemented with EDTA (5 mM final concentration) and 0.6 mg/ml lysozyme and incubated for 30 min at 37 • C to release the periplasmic fraction. Sphaeroplasts were removed by centrifugation at 12 500 rev./min for 20 min using a Beckman JA-25.50 rotor at 4 • C and the supernatant was retained. The EDTA was removed by dialysis in buffer A before loading the supernatant on to a 5 ml HisTrap column and purifying as above.

SAXS (small-angle X-ray scattering)
SAXS data using purified TorD His , TorA/TorD His and TorA SP /TorD His samples were recorded on the ID14-3 BioSAXS beamline at the European Synchrotron Radiation Facility (ESRF Grenoble, France). Sample-detector distance was 2.43 m for an X-ray wavelength of 0.931 Å (1 Å = 0.1 nm). Then, 50 μl sample and buffer volumes were loaded in a flowthrough quartz capillary cell at 25 • C. The sample volume exposed to the X-ray beam was approximately 10 μl. Samples were checked for radiation damage by using ten successive exposure times of 10 s each. Final exposure time was 100 s for all samples/buffers. The twodimensional diffraction patterns were normalized to an absolute scale and azimuthally averaged to obtain the intensity profiles I(Q), within BSxCuBE (ESRF beamline data collection software). Solvent contributions (buffer backgrounds collected before and after every protein sample) were averaged and subtracted from the associated protein sample using the program PRIMUS [29].

CD analysis
CD analysis of protein samples was provided as a service by the Scottish Circular Dichroism Facility, University of Glasgow. Spectra of protein solutions were obtained using a JASCO J-810 spectropolarimeter. Far-UV CD measurements (185-260 nm) were collected in quartz cells of 0.02 cm pathlength at 25 • C with a scan speed of 10 nm/min, bandwidth of 1 nm, response of 2 s and data pitch of 0.2 nm. Before collection of CD data, proteins were buffer-exchanged into CDappropriate buffer (50 mM Tris/H 2 SO 4 , pH 7.5, 200 mM K 2 SO 4 and 1 mM DTT). Protein concentrations were estimated using a NanoDrop spectrophotometer before the experiment. Near-UV CD measurements (250-320 nm) were obtained in a 0.2cm-pathlength quartz cuvette using the following parameters: temperature, 25 • C; scan speed, 10 nm/min; bandwidth, 1 nm; response, 2.0 s; and data pitch, 0.2 nm. Protein was analysed in appropriate buffer (50 mM Tris/HCl, pH 7.5, 200 mM KCl and 1 mM DTT). The spectra were corrected for protein concentration and cell pathlength before being analysed by DichroWeb (http://dichroweb.cryst.bbk.ac.uk), an online server which hosts the various algorithms used to estimate protein secondary structures.

Metal analysis
Metal content was analysed by ICP-AES (inductively coupled plasma-atomic emission spectrometry)/ICP-MS (inductively coupled plasma-MS) and was provided as a service by the School of Chemistry at the University of Edinburgh.

Limited trypsinolysis of TorA-TorD His and TorA SP -TorD His complexes
Protein samples (1 mg/ml in 50 mM Tris/HCl, pH 7.5, 200 mM KCl and 1 mM DTT; 100 μl final volume) were incubated with 1 % (w/w) trypsin (porcine pancreas, proteomics grade, Sigma-Aldrich) at 37 • C. Aliquots of 5 μl were withdrawn at various intervals and the reaction stopped by addition of 20 μl of Laemmli buffer (Bio-Rad Laboratories) followed by immediate boiling for 10 min. Once the time course was completed, samples were analysed by SDS/PAGE (12% gels).
Proteins were identified by tryptic peptide mass fingerprinting of gel slices performed by the FingerPrints Proteomic Service (College of Life Sciences, University of Dundee).

Analysis of TorA-TorD His complexes by SEC (size-exclusion chromatography)-MALLS (multi-angle laser light scattering)
Size estimates of TorA-TorD His complexes were obtained by SEC-MALLS, which was provided as a service by the University of York Bioscience Technology Facility. Measurements were conducted on a system comprising a Wyatt HELEOS-II multiangle light-scattering detector and a Wyatt rEX refractive index detector linked to a Shimadzu HPLC system (SPD-20A UV detector, LC20-AD isocratic pump system, DGU-20A3 degasser and SIL-20A autosampler). Work was conducted at room temperature (20 + − 2 • C). Solvent was filtered through a 0.2-μm-pore-size filter before use and a further 0.1-μm-poresize filter was present in the flow path. Filtered (0.2 μm pore size) TorA-TorD His samples (100 μl, corresponding to 100 μg of protein) were applied to a Superdex 200 10/300 GL column (GE Healthcare) equilibrated with 50 mM Tris/HCl (pH 7.5), 200 mM KCl and 1 mM DTT at 0.5 ml/min; Shimadzu LC Solutions software was used to control the HPLC, and Astra V software was used for the HELEOS-II and rEX detectors. BSA was used for molecular mass calibration. The Astra data collection was 1 min shorter than the LC solutions run to maintain synchronization. Blank buffer injections were used as appropriate to check for carry-over between sample runs. Data were analysed using the Astra V software. MWs were estimated using the Zimm fit method [30] with degree 1. A value of 0.19 was used for protein refractive index increment (dn/dc).

Isolation of a TorA-TorD His complex
Isolation of a complex between untagged TorA and affinity-tagged TorD was originally demonstrated on a small scale using a Cterminally SPA-tagged form of TorD [22]. In order to scale up the co-expression system to allow detailed analysis of the complex, the original torAD SPA allele was cloned into expression vector pQE80 and the DNA coding for the SPA tag at the C-terminus of TorD replaced with a hexahistidine tag ( Figure 1B). The soluble cell extract containing overproduced TorA and TorD His was subjected to IMAC and fractions containing proteins of the expected size of TorA and TorD His were identified in a single peak at 110 mM imidazole. The IMAC samples were pooled, concentrated and subjected to SEC using a HiLoad 16/60 Superdex 200 Prep Grade column. As shown in Figure 2(A), protein was eluted at 78.5 ml, corresponding to an estimated molecular mass of 114 kDa. Analysis of the peak fraction by SDS/PAGE revealed the presence of two major proteins ( Figure 2B), which were confirmed by tryptic peptide mass fingerprinting to be TorA and TorD His . The overall protein yield from this process was approximately 1.5 mg/g of bacterial cells.
It was notable during the purification of TorAD His , particularly following concentration of the samples, that more than one electrophoretic form of TorA was evident by SDS/PAGE. A low-percentage acrylamide gel was therefore used to separate these forms, which were excised and analysed by tryptic peptide mass fingerprinting. The analysis showed that, whereas the majority of the TorA was present as full-length intact polypeptide, there was a proportion of TorA that had undergone limited proteolysis and was lacking up to 150 amino acids from its Cterminal end (Supplementary Figure S1 at http://www.biochemj. org/bj/452/bj4520057add.htm). In addition, in some instances, up to 12 amino acid residues could be lost from the N-terminus of TorA. It is concluded that, whereas a complex of TorAD His is stable to purification, the extremities of TorA can be subjected to variable levels of proteolysis during isolation.

Removal of the entire Tat signal peptide from the TorA protein (TorA 2-39) still allows isolation of a stable complex with TorD His
As well as signal peptide binding, previous biochemical and genetic experiments pointed to a binding site for TorD within the mature region of the TorA enzyme (e.g. [9,10]). To investigate this further, a vector was designed that would co-overproduce a version of TorA lacking its entire 39-residue Tat signal peptide (TorA SP ) with TorD His (Figure 1C). When TorD His was isolated from the soluble cell fraction by IMAC, a protein of the expected size of TorA SP was also co-purified. The TorD Hiscontaining fractions were pooled, concentrated and applied to a Superdex 200 size-exclusion column. The TorD His eluted at 78.8 ml, corresponding to an estimated molecular mass of 110 kDa ( Figure 2C) and analysis of the concentrated peak fraction by SDS/PAGE revealed the presence of two major proteins ( Figure 2D), which were confirmed by tryptic peptide mass fingerprinting to be TorA SP and TorD His . Thus it is clear that a stable complex of TorD His and TorA can be isolated in the absence of the twin-arginine signal peptide.
TorD His binds the full-length TorA precursor, and the TorA sp species lacking its entire signal peptide ( 2-39), at a 1:1 stoichiometry SEC indicates that the complex of TorD His with TorA elutes at a similar molecular mass regardless of whether the signal peptide was present on TorA ( Figure 2). A more accurate technique is SEC-MALLS, which uses HPLC-linked SEC coupled with static laser light scattering to provide a direct measure of molecular mass. Using this technique, the TorA-TorD His complex was observed to exhibit a mass range of 100-120 kDa and a peak mass of 114.7 kDa (Table 1), which is very close to the calculated mass (117.9 kDa) for a 1:1 complex of the two proteins ( Table 1). The TorA SP -TorD His complex exhibited a mass range of 95-115 kDa and a peak mass of 108.2 kDa (Table 1), which is close to the calculated theoretical mass (113.8 kDa) for a 1:1 complex of the two proteins ( Table 1). The relatively wide mass range seen for both samples reflects the observed proteolytic events leading to heterogeneity of the TorA sample observed by SDS/PAGE ( Figure 2). Overall, the SEC-MALLS analyses point to complexes containing one TorA protein and one TorD protein, and suggest strongly that TorD can remain stably bound to TorA in the absence of the twin-arginine signal peptide.

Figure 3 Limited trypsinolysis of the TorA-TorD His and TorA SP -TorD His complexes
The TorA-TorD His (A) and TorA SP -TorD His complexes (B) were isolated. Each purified complex was incubated with trypsin and 5 μl aliquots removed at 5 min intervals between 0 and 60 min, mixed with Laemmli buffer and immediately boiled to prevent any further digestion. (C) As a control, periplasmic mature TorA His was purified and incubated with trypsin before 5 μl aliquots were removed at the following eight time points (0 min, 10 min, 20 min, 30 min, 1 h, 2 h, 4 h and 16 h). Concomitantly, mature TorA His incubated without trypsin before aliquots were removed at the following six time points (0 min, 10 min, 30 min, 1 h, 4 h and 16 h). Samples were mixed with Laemmli buffer and immediately boiled. At the end of the time courses, the samples were analysed by SDS/PAGE (12 % gels). Molecular masses are indicated in kDa.

The TorA C-terminal domain is protease-accessible in TorAD complexes
To gain more insight into the architecture of the TorAD complexes under investigation in the present study, limited proteolysis experiments were undertaken. First, the purified TorA-TorD His complex was incubated with a low concentration of trypsin over a time course of 60 min. As shown in Figure 3(A), TorD His was almost completely resistant to trypsinolysis under these conditions. In contrast, TorA was cleaved to yield a stable proteolytic fragment of approximately 70 kDa that was apparently resistant to further digestion by trypsin. Peptide mass fingerprinting of the stable 70 kDa TorA proteolysis product revealed that the initial 675 amino acid residues from the Nterminus of the protein were intact and that the trypsin treatment had therefore removed a short section of polypeptide from the C-terminal end of TorA (Supplementary Figure S1). Next, the TorA SP -TorD His complex, where TorA had already had its 39-residue signal peptide deliberately removed, was incubated with trypsin in a similar manner. As shown in Figure 3(B), TorD His was almost completely resistant to protease digestion, whereas the TorA SP was degraded to a smaller stable product. Analysis of this protease-resistant fragment by tryptic peptide mass fingerprinting showed that it was cleaved at Lys 675 , which was the identical position where the full-length TorA precursor was cleaved. As a control, a C-terminally Histagged, but otherwise native, form of TorA was purified from the periplasm of E. coli strain FTH007. The fully folded mature form of TorA was completely resistant to digestion with trypsin, even after a 16-h incubation period ( Figure 3C). These results are consistent with the suggestion that the TorA-TorD His and TorA SP -TorD His complexes have a similar overall architecture, both being susceptible to limited trypsinolysis in an identical manner. In both cases, a C-terminal stretch of immature TorA is obviously exposed to the aqueous phase in a conformation that is clearly not adopted by the mature fully assembled enzyme.

The C-terminal domain of TorA is not involved in stable TorAD complex formation
TorA is a member DMSO reductase family of molybdoenzymes. Proteins in this family adopt a fold that can be divided into four structurally distinct domains [31], and, of these, Domain IV is formed by the extreme C-terminal part of the protein and is the only domain in this fold that comprises a clear contiguous stretch of polypeptide chain (Supplementary Figure S2 at http://www.biochemj.org/bj/452/bj4520057add.htm). Although the crystal structure of the E. coli TorA protein has not been solved, structure predictions based on the Shewanella massilia TorA crystal structure [32] suggest the position of the trypsin cleavage site in the E. coli TorA-TorD His complexes lies within an unstructured loop region that connects Domain III to Domain IV (Supplementary Figure S2). To investigate this further, a construct was designed that would co-overproduce TorD His together with the first 675 amino acids of TorA, including the signal peptide (designated TorA CT ). First, the construct was expressed in a strain lacking endogenous TMAO reductase activity under strict anaerobic conditions, which would induce expression of the MGD biosynthetic machinery. Then a crude cell extract was prepared and the TMAO reductase activity assayed. The resultant data showed that the C-terminally truncated form of TorA was completely devoid of enzymatic activity (Supplementary  Table S1 at http://www.biochemj.org/bj/452/bj4520057add.htm), establishing that this domain of the protein is absolutely critical for function. Next, TorD His was isolated from the soluble cell fraction by IMAC, which resulted in a protein of the expected size of TorA CT co-purifying ( Figure 4). The two proteins remained fully associated during SEC through a Superdex 200 column ( Figure 4A), with the complex eluting with an estimated molecular mass of 100 kDa ( Figure 4A). Analysis of the peak fractions by SDS/PAGE confirmed that both proteins were present in the sample, but this time very little degradation of the deliberately truncated TorA was apparent ( Figure 4B). SEC-MALLS analysis of the purified complex (Table 1) confirmed that the TorA CT -TorD His complex had a narrow mass range of 97-107 kDa, reflecting the relative homogeneity of the sample, and a peak mass of 99.7 kDa, which is close to the calculated mass (99.2 kDa) for a 1:1 complex of the two proteins (Table 1). It can be concluded that, although the C-terminal 173 amino acids are essential for the ultimate enzymatic activity of TorA, they do not influence the formation or stability of the pre-export complex between TorA and TorD.

The TorAD complexes are devoid of MGD, but retain most of their secondary structure
Whereas it has generally been assumed that a Tat substrate protein is maintained in an 'unfolded conformation' when in complex with its pre-export chaperone, the degree of folding of such Tat precursors has not been examined in detail. In the present study, recombinant TorD His , active TorA His , and the TorA-TorD His and TorA SP -TorD His complexes were analysed by CD spectroscopy. The CD spectra recorded for each sample are shown in Figure 5 and the calculated percentages of helix, strand, turn and disorder are given in Supplementary Table S2 (at http://www.biochemj.org/bj/452/bj4520057add.htm). Consistent with the highly α-helical structure of TorD [17], the CD spectrum of purified TorD His shows a high proportion of helix ( Figure 5 and Supplementary Table S2). Enzymes of the DMSO reductase family are mixed α/β proteins [31][32][33][34] and this is reflected in the CD spectrum of purified TorA His ( Figure 5 and Supplementary Table S2). Interestingly, CD analysis of the TorA-TorD His and TorA SP -TorD His complexes showed that both had a significant degree of secondary structure and  that the observed percentage of disorder was no higher than that of isolated fully assembled TorA His . Using the parameters determined for TorA His and TorD His (Supplementary Table S2), the predicted percentages of helix, strand, turn and disorder that would be expected for a simple 1:1 mixture of these two proteins can be calculated (Supplementary Table S2). It is notable that these predicted values are a close match to those experimentally determined for the purified TorA-TorD His and TorA SP -TorD His complexes, with the values particularly close for the TorA-TorD His complex (Supplementary Table S2). Taken together, these results indicate that the TorA component of the TorA-TorD His complex has probably adopted its native secondary structure. Moreover, whereas the TorA His protein was isolated from an anaerobic culture and exhibited a straw colour in solution, the TorA-TorD His and TorA SP -TorD His complexes examined were produced from cultures grown under aerobic conditions without added molybdate. As a result, they would be expected to be largely devoid of cofactor and indeed were colourless and devoid of enzymatic activity. Metal content in the complexes by ICP-MS showed that the molar ratio of molybdenum to TorA/D complex was less than 1:1000. Taken together with the trypsinolysis experiments (Figure 3), these data suggest that the cofactor-free TorD-bound species of the TorA apoenzyme adopts a fold very similar to that of its fully assembled counterpart, save for an exposed C-terminal Domain IV.

SAXS analysis of full-length TorA-TorD His and signal peptide-less TorA SP -TorD His complexes reveals shared structural features
SAXS is a powerful technique that can be used to compare the overall shapes and sizes of proteins in solution. We analysed the TorA-TorD His and TorA SP -TorD His complexes, together with isolated TorD His , since the high-resolution structure of TorD from S. massilia suggested that the protomer could adopt a highly extended conformation [17]. Representative SAXS scattering curves from each of the three samples are shown in Figure 6. It is clear that the scattering curves for the TorA-TorD His and TorA SP -TorD His complexes are almost indistinguishable, indicating that they have an identical overall shape. An overview of the general parameters extracted from the SAXS data including the radii of gyration (R g ; obtained from the Guinier approximation [35]), the distance distribution function, P(r) and the maximum particle size (D max ; obtained from the GNOM [36] analysis) are given in Table 2.
A low-resolution model of TorD His , obtained using the program DAMMIN [37], is shown in Figure 7(A). The overall shape of TorD His is compact and elliptical with dimensions of approximately 36 Å×50 Å×34 Å. From the crystal structures of S. massilia TorD [17] and the TorD family protein DmsD  from E. coli [38][39][40], scattering envelopes can be calculated [41]. It is clear from Supplementary Figure S3 (http://www. biochemj.org/bj/452/bj4520057add.htm) that neither the S. massilia TorD domain-swapped dimer nor one of the protomers involved in forming that dimer would give scattering curves similar to that measured for TorD His in solution. By contrast, the scattering curve predicted from the X-ray structure of DmsD is very similar in shape, and the electron density of monomeric DmsD, which can be readily docked into the TorD His SAXS envelope ( Figure 7B). Since the scattering curves indicated that the TorA-TorD His and TorA SP -TorD His complexes were indistinguishable, a single SAXS-derived shape corresponding to either the TorA-TorD His or the TorA SP -TorD His complex is shown in Figures 7(C) and 7(D). The complex is highly elongated, with an overall length of approximately 120 Å. There is a large central area of density with one relatively large lobe of density protruding from one side of the central region and a smaller lobe of density protruding from the opposite side. The large central area has approximate dimensions of 70 Å×74 Å×44 Å and is a reasonably close match to the dimensions of the crystal structure of S. massilia TorA (Figures 7C and 7D).

DISCUSSION
The process of co-ordinated assembly ('proofreading' or 'quality control') is an important consideration in the biosynthesis of all complex cofactor-containing Tat-dependent enzymes [5]. The simplest model system used to study this process is the E. coli TMAO reductase, which has a single MGD cofactor and a single biosynthetic chaperone, TorD. Whereas early research has seen analyses of the TorD-signal peptide interaction in isolation [12,42], the present study has focused on the characterization of the entire pre-export TorAD complex.
Design of co-overproduction vectors established that untagged TorA formed a tight and stable complex with hexahistidinetagged TorD. Surprisingly, however, despite the strong evidence for two separate TorD-binding sites on TorA [9,11], we have clearly demonstrated using the accurate SEC-MALLS technique in conjunction with other experiments that TorD binds TorA with a 1:1 stoichiometry. Even more surprisingly, given the well-characterized interaction between TorD and the TorA signal peptide [11][12][13][14][15]42], a truncated TorA enzyme completely lacking its entire 39-residue signal peptide retained the ability to interact in a stable manner with TorD. Whereas the present study clearly establishes that there is a high-affinity TorD-binding site within the mature portion of TorA, it also raises questions about the role and importance of the signal peptide interaction.
These findings can be interpreted in two possible ways. First, that the signal peptide-binding activity of TorD is of a much lower affinity than that of the second binding site within the mature TorA sequence, so much so that there is a strong bias towards the higher-affinity species during the purification protocol. Although this scenario is not impossible, recent studies place the apparent K d for TorD binding to the isolated TorA signal peptide at 59-330 nM [12,42], which is indicative of very tight binding such that it is likely that a detectable portion of TorA would not fail to have two TorD molecules bound. The second possibility is that the TorD-binding site within the mature portion of TorA is very close to the signal peptide-binding epitope such that a single TorD protein could interact with both simultaneously. This hypothesis would require there to be two different TorA-binding sites on a single TorD protein. Although the location of the TorA signal peptide-binding site on TorD has not yet been reported, there is a larger body of evidence available on the location of the signal peptide-binding site on another member of the TorD family, DmsD. The E. coli DmsD protein was the first twin-arginine signal peptide-binding chaperone to be described [43]. It is essential for the assembly of molybdenum-dependent DMSO reductase and selenate reductase enzymes in E. coli and binds to the signal peptides of those enzymes with apparent K d values in the range 10-100 nM [43][44][45]. Genetic, biochemical and structural analyses predict the signal peptide-binding site to occupy only one face of DmsD (primarily involving residues of the N-terminal helices) and so suggests that a significant proportion the DmsD protein is not involved in signal peptide recognition [39,45,46]. This leaves the possibility, at least, that a second protein-or peptiderecognition site may be present on this type of chaperone. Indeed, variants of E. coli TorD have been described (C79R and L83P) Rigid body modelling was conducted using the homologous component parts of the complex, i.e. TorD homologue DmsD (PDB code 3CW0, shown in green), and TorA (PDB code 1TMO) split into two parts: Domains I-III (i.e. amino acids 5-629; shown in cyan) and Domain IV (amino acids 631-798; shown in magenta). Modelling was conducted using SASREF [47]. The generated model is docked into the TorA-TorD His ab intio model for comparison.
that retain signal peptide-binding activity, but are impaired in their binding to mature TorA [10].
Further evidence for overlapping TorD-binding sites near the Nterminus of TorA come from analysis of the SAXS data obtained in the present study. To investigate the possible arrangement of TorA and TorD in the TorAD complex, rigid body modelling was undertaken using the SASREF program [47] and the structures of S. massilia TorA and E. coli DmsD as building blocks. To facilitate this analysis, TorA was split into two parts to give the proteolytically stable N-terminal part (Domains I-III) and the proteolytically sensitive Domain IV identified in the present study. A constraint was then applied that the last amino acid of N-terminal part should be within 10 Å of the first amino acid of Domain IV, in order to incorporate flexibility into the TorA component of the complex, again consistent with the protease accessibility results of the present study. The results from the rigid body modelling show a similar overall shape to that obtained during ab initio analysis (the fit of the model with the experimental data is shown in Supplementary Figure  S4 at http://www.biochemj.org/bj/452/bj4520057add.htm). The rigid body model is shown docked into the ab initio SAXSderived envelope of TorAD in Figure 8. When docked into the ab initio model, the density for TorD fits very well into the larger protruding lobe, whereas the remainder of the ab initio envelope is of an appropriate volume to accommodate TorA. The C-terminal Domain IV of TorA distal from TorD fits into the smaller protruding lobe, which is consistent with the biochemical data of the present study that Domain IV is not required for interaction of TorD with TorA. Bearing in mind that the SAXS scattering curves for both signal peptide-containing and signal peptide-free TorAD complexes are identical, this must be taken as good evidence that the binding epitope for TorD on the mature portion of TorA must lie near the N-terminus and thus also near the second epitope within the signal peptide.
The CD and SAXS analyses suggest that there is a very high degree of folding by the TorA protein even in the absence of the MGD cofactor. However, it is clear that the TorA-TorD His and TorA SP -TorD His complexes are not in the native conformation and that the C-terminal Domain IV of the TorA protein is exposed. This could suggest that the locking of Domain IV into position represents the final act in cofactor insertion. One possible role of TorD could be to prevent premature closing of the Domain IV 'flap' before MGD has bound, or, alternatively, that closing of Domain IV after cofactor loading is the trigger that releases TorD from the now mature enzyme. The molecular mechanism of how TorD senses the events surrounding cofactor loading remains to be unearthed.  The C-terminal Domain IV is shaded in blue and the approximate position of the trypsin-cleavage site determined for E. coli TorA in the TorA-TorD His and TorA SP -TorD His complexes is indicated. The image was created using PyMOL (http://www.pymol.org) with PDB code 1TMO.

Figure S3 Comparison of SAXS scattering curves
Theoretical curves were calculated from S. massilia TorD (PDB code 1N1C) monomer (blue line) and dimer (red line) and E. coli DmsD (PDB code 3CW0) monomer (black line) using CRYSOL and compared with experimental scattering curves of TorD His (green dots). Fitted curve of TorAD His rigid body modelling conducted using the TorA structure (PDB code 1TMO) (split into two components: Domains I-III and Domain IV) and the DmsD structure (PDB code 3CW0), performed using SASREF. The fitted curve is shown in comparison with the E. coli TorAD His scattering curve.