An Escherichia coli twin-arginine signal peptide switches between helical and unstructured conformations depending on the hydrophobicity of the environment Miguel San Miguel 1 , Rachel Marrington 2 , P. Mark Rodger 2 , Alison Rodger 2 and Colin Robinson 1, * 1 Department of Biological Sciences and 2 Department of Chemistry, University of Warwick, Coventry, UK The Tat system catalyzes the transport of folded globular proteins across the bacterial plasma membrane and the chloroplast thylakoid. It recognizes cleavable signal peptides containing a critical twin-arginine motif but little is known of the overall structure of these peptides. In this report, we have analyzed the secondary structure of the SufI signal peptide, together with those of two nonfunctional variants in which the region around the twin-arginine, RRQFI, is replaced by KKQFI or RRQAA. Circular dichroism studies show that the SufI peptide exists as an unstructured peptide in aqueous solvent with essentially no stable secondary structure. In membrane-mimetic environments such as SDS micelles or water/trifluoroethanol, however, the peptide adopts a structure containing up to about 40% a-helical content. Secondary structure predictions and molecular modelling programs strongly suggest that the helical region begins at, or close to, the twin-arginine motif. Studies on the thermal stability of the helix demonstrate a sharp transition between the unstructured and helical states, suggesting that the peptide exists in one of two distinct states. The two non- functional peptides exhibit almost identical spectra and properties to the wild-type SufI peptide, indicating that it is the arginine sidechains, and not their contribution to the helical structure, that are critical in this class of peptide. Keywords: signal peptide; twin-arginine translocation; Tat system; protein transport; SufI. The twin-arginine translocation (Tat) system operates in the cytoplasmic membranes of most free-living bacteria and in the thylakoid membranes of plant chloroplasts [1–3]. Operating alongside Sec-type translocases, the Tat system functions in the transport of proteins bearing cleavable N-terminal signal peptides (RR-signal peptides) in which a twin-arginine motif plays a central role [4,5]. The substrate proteins are recognized by a membrane-bound translocase and subsequently transported, at least in some cases, in a fully folded state [6,7]. The prime role of the Tat system appears to be in the transport of proteins, which are either obliged to fold prior to translocation, or which fold too tightly for the Sec system to accommodate. Examples of the former category include periplasmic proteins that are exported only after binding any of a range of complex redox cofactors, such as molybdopterin or FeS centres [8–11]. These cofactors are inserted in the cytoplasm by complex enzymatic processes, and it has been argued that this must necessitate the export of these proteins in a largely, if not fully, folded form. Substrates bearing RR-signal peptides are recognized by a membrane-bound translocase that consists minimally of TatABC in most cases. Critical genes encoding these subunits have been identified in bacteria, particularly Escherichia coli [10–13], and in plants [14–16], and a TatABC complex has been purified from detergent-solubi- lized E. coli membranes [17]. The size of the purified complex has been estimated to be in the order of 500– 600 kDa, suggesting the presence of numerous copies of each subunit [17]. The TatBC subunits form a tight core subcomplex in a strict 1 : 1 ratio [17] and studies on the thylakoid system indicate that this subcomplex forms the initial binding site for substrates, with TatA recruited at a later stage [18,19]. Consistent with this scenario, important residues in E. coli TatC have been identified on the cytoplasmic side of the membrane [20,21]. Although the importance of the twin-arginine motif has been firmly established by mutagenesis studies, RR- signal peptides have not been characterized in structural respects. The peptides are probably too small to form folded globular domains but with other types of targeting signal it has been shown that functionality is strictly dependent on the formation of specific secondary struc- tures. Prominent examples include Sec-type signal pep- tides and the presequences of imported mitochondrial proteins (e.g [22,23]). In this report, we have analyzed a typical E. coli RR-signal peptide and we show that the structure of the peptide differs dramatically according to environment. Implications for the translocation process are discussed. Correspondence to C. Robinson, Plant Biochemistry Laboratory, Department of Plant Biology, Royal Veterinary and Agricultural University, 40, Thorvaldsensvej, DK-1871 Frederiksberg C, Copenhagen, Denmark. Fax: + 44 2476523701, Tel.: + 44 2476523557, E-mail: CRobinson@bio.warwick.ac.uk Abbreviations: MD, molecular dynamics; TFE, trifluoroethanol. *Present address: Plant Biochemistry Laboratory, Department of Plant Biology, Royal Veterinary and Agricultural University, 40, Thorvaldsensvej, DK-1871 Frederiksberg C, Copenhagen, Denmark. (Received 15 March 2003, revised 2 June 2003, accepted 9 June 2003) Eur. J. Biochem. 270, 3345–3352 (2003) Ó FEBS 2003 doi:10.1046/j.1432-1033.2003.03710.x Experimental procedures Materials Peptides, purified by HPLC, were purchased from Alta Bioscience (Birmingham, UK). Circular dichroism Circular dichroism (CD) spectra are the simplest indication of protein and peptide secondary structure. CD spectra were collected using a Jasco J-715 spectropolarimeter equipped with a single sample Peltier thermostatting unit. Spectra were averaged over eight scans collected with 1 nm data intervals, 1 nm bandwidths, 0.5 s response time and 200 nmÆs )1 scan speed. The CD spectra as a function of temperature were collected by monitoring at a single wavelength (222 nm, 2 nm bandwidth) with an 8-s response time and a ramp rate of 1 °CÆmin )1 . At each of 20, 30, 40, 50, 60, 70, 80, and 90 °C, the temperature ramp was held while a single wavelength scan with 4-s response time was collected. All samples were made up to a concentration of 0.10 mgÆmL )1 by weighing a peptide sample on a seven- figure balance and adding the appropriate volume of solvent. In the case of the mixed organic–aqueous solvents, the solvent was fully mixed prior to adding to the dry peptide. Organic solvent volumes were measured using a syringe. Aqueous solution volumes were measured using micropipettes. a-Helical content was estimated using a De value of  12 mol )1 Ædm )3 Æcm )1 for a 100% helical peptide [24] and also by applying the protein CD structure-fitting program, CD sstr [25]. The latter approach may only be used as a guide as its database does not properly account for random coil structures. Modelling Energy minimization calculations were performed using QUANTA / CHARMM version 28 (Accelerys Inc., Cambridge UK) and molecular dynamics simulations with the DL _ POLY package [26]. Coordinates and force field were constructed within QUANTA / CHARMM and then exported and converted to DL _ POLY format using in-house software. Test calcula- tions were performed on representative single configurations using both CHARMM and DL _ POLY to check that no differences were observed between any of the energy calculations. Initial configurations for the peptide were constructed in two different ways: (a) using the PHD structure prediction method [27], and (b) from the lowest energy configuration obtained from minimizing configura- tions generated during a vacuum molecular dynamics (MD) simulation at 500 K. The PHD method gives an a-helix covering 16 amino acids, while the MD method gives a random coil structure (see Figs 4 and 5). Initial structures were optimized using the conjugate gradient method ( CHARMM ). The peptide was then inserted into a solvent box (65 A ˚ and previously equilibrated at 300 K, 1 atm), all solvent molecules that overlapped with the peptide removed, and three additional solvent molecules converted to Cl – ions to compensate for the +3 charge of SUFI; this resulted in 3173 water molecules, or 990 trifluoroethanol (TFE) molecules, in a periodic truncated octahedral simulation box. The system was then relaxed by (a) performing a 5-ps MD simulation at 300 K, 1 atm in which the peptide was treated as a rigid body, and then (ii) performing a 2-ps MD simulation with a fully flexible peptide at 2 K; these stages served to remove any strain introduced on solvation without destroying the initial secondary structure. A further 6 ns simulation was then accumulated to study the response of the secondary structure. Secondary structure was analyzed using the STRIDE program [28]. All MD simulations were performed at constant temperature and pressure (NPT) using the Nose ´ –Hoover method with thermostat and barostat relax- ation constants of 0.5 and 1.0, respectively, and a timestep of 2 fs. The peptide and TFE were modelled with the CHARMM potential, and water with the SPC model. We note that the published TFE potentials do have some inadequa- cies [29], and that the CHARMM potential can underestimate the stability of a-helices relative to other force fields [30], but overall the model is reasonable. Long range forces were truncated at 10 A ˚ , and the reaction field method used to correct for long-range electrostatic effects. Secondary structure prediction Secondary structure was predicted using the PSIPRED [31], JPRED [32], PROF [33] and PHD [27] secondary structure prediction programs. Results Structures of wild-type and nonfunctional SufI signal peptides The overall aim in this study was to analyze the structural characteristics of a wild-type RR-signal peptide in both aqueous solvent and membrane-mimicking environments, as well as in some intermediate environments. The objective was to identify the secondary structure charac- teristics immediately after synthesis (in aqueous medium, as the targeting process is post-translational at least in the vast majority of cases) and once bound either to the membrane or the translocase (see below). A possibility addressed in this study was that the twin-arginine motif somehow may contribute significantly to the secondary structure, and we therefore also analyzed mutant variants that are not recognized by the Tat system. In this way, we sought to determine whether the significance of the twin- arginine motif stems from the nature of the sidechains and/or its ability to promote a given form of secondary structure. We chose SufI as an example of a typical E. coli Tat substrate. SufI is synthesized with an apparently typical RR-signal peptide with the sequence shown in Fig. 1. Two variants are also shown; in one (SufI-KK), the twin-arginine motif is replaced by twin-lysine, which causes either a complete or near-complete block in translocation by the Tat pathway in both chloroplasts and bacteria [4,5,34]. In the other, the region around twin-arginine motif, RRQFI, is replaced by RRQAA (SufI-AA). While the effects of this double mutation have not been tested in bacterial systems, studies on the thylakoid system have demonstrated the critical importance of a highly hydrophobic residue at the second or third positions after the twin-arginine motif [35]. 3346 M. San Miguel et al. (Eur. J. Biochem. 270) Ó FEBS 2003 Bacterial RR-signal peptides invariably contain such a residue, typically phenylalanine, leucine, isoleucine or valine, at one or both of these positions [8] and the SufI- AA peptide shown in Fig. 1 is therefore strongly predicted to be nonfunctional. The SufI peptide has been synthesized chemically and shown to competitively inhibit the Tat- dependent transport of proteins into inverted E. coli vesicles [36], and to interact with Tat complexes in detergent [37]. This indicates that the peptide is functional in isolation and we used the same peptide in the present study in order to examine the structure of the signal at various stages. The KK mutant peptide does not affect binding of wild-type substrate suggesting that the RR motif is required for substrate binding. The SufI RR-signal peptide is unstructured in aqueous solvent but a-helical in membrane-mimetic environments The secondary structure of the SufI RR signal and the mutant variants was analyzed by CD. Figure 2 shows their structures in aqueous solvent (pH ¼ 7) and two membrane- mimetic environments, namely SDS micelles and water/ TFE systems. A range of alcohols in which the peptide was soluble were also investigated and spectra in methanol and ethanol shown for SufI. The data obtained with the wild- type peptide in water (Fig. 2A) show a spectrum character- istic of an unstructured peptide with a negative maximum below 200 nm and no signal at 220 nm. The spectra in the other solvent systems are typical of a-helices, though of Fig. 1. Structure of the SufI signal peptide. Shown are the amino acid sequences of the E. coli SufI twin-arginine signal peptide, together with the sequences of two mutant ver- sions used in this study. The essential twin-arginine motif is shown as bold italics. Fig. 2. CD spectra of SufI signal peptides in aqueous and apolar solvents. (A) CD spectra were taken using the wild-type SufI signal peptide in solvents as indicated in the figure. Further details are given in Experimental procedures. (B) Comparison of the wild-type peptide with the SufI-KK and SufI-AA pep- tides (denoted as ÔKKÕ or ÔAAÕ, respectively, in inset to graph). CD spectra were taken in water, water/TFE (50 : 50 v/v) and TFE. Ó FEBS 2003 Structural characteristics of Tat-specific targeting signals (Eur. J. Biochem. 270) 3347 magnitude significantly less than expected for 100% helix (De  12 mol )1 Ædm )3 Æcm )1 at 208 nm and 222 nm) even in TFE where the helical content is  45%. The alcohols ethanol, propan-1-ol and butan-1-ol gave the same spec- trum within experimental error down to 195 nm, with an a-helix content of  25%. Methanol induces a noticeably more helical content ( 40%), similar in magnitude to a 50% TFE/water solvent. A 0.5% w/v SDS solution resulted in a structure similar to that observed with ethanol. Figure 2B shows the spectra obtained with the SufI-KK and SufI-AA peptides. In both cases, the spectra are almost identical to those of the wild-type indicating that the twin- arginine motif does not contribute specific a-helical prop- erties in this context. Compared with SufI, the 208 nm region of the spectra indicates a slight increase in helical character for SufI-AA and a slight decrease for SufI-KK in TFE, whereas at 222 nm in water/TFE SufI-AA has less intensity and SufI-KK is the same as SufI. This suggests that the mutant helices are perhaps of slightly different form. The SufI-AA random coil spectrum (water) also differs slightly from the others. Thermal stability of the a-helix: evidence for two distinct states The CD spectra of SufI collected in 50 : 50 water/TFE as a function of temperature show a well-defined nonzero isosbestic point at 201 nm (Fig. 3A). This indicates that the system is in one of two states, presumably a-helical and random coil. The fact that there is no sharp transition during the temperature ramp (Fig. 3B) suggests that we have a temperature dependent equilibrium between the two states rather than any sort of concerted transition. The situation with the two mutants is very similar, though there is only an approximate isosbestic point in each case, with KK being slightly the worse (data not shown). Although there are not enough data to enable us to determine the cause of this, the most likely cause is a slight variation in helical forms present in the solution. Location of the helical segment(s) in hydrophobic milieu The SufI RR-peptide clearly contains substantial amounts of helical structure in apolar environments and, because the CD data do not indicate the location of this structure within the 27-residue peptide, we used additional methods in order to pinpoint the likely location(s). First, the PSIPRED , JPRED , PROF AND PHD secondary structure prediction programs were used, and typical predictions are shown in Fig. 4A. All four programs predict substantial a-helical content (26, 48, 56 and 59%, respectively) and it is notable that these regions encompass the twin-arginine motif in each case, with the RR motif usually positioned at or near the beginning of the helical section. As a complementary technique we carried out a detailed molecular modelling simulation of SufI RR in both water and TFE. Simulations were performed using two different initial peptide conformations, one taken from the PHD structure prediction (59% a-helix; diagram 1 of Fig. 4B) and one with essentially a random coil secondary structure (the lowest energy structure from the conformational search; not shown). The simulated secondary structure from these two initial configurations was found to have converged after about 1 ns, indicating that the subsequent structures were stable and not merely an artefact of the simulation timescale, nor of the starting geometry. The results of the simulations are in semiquantitative agreement with the CD experiments. No a-helix was found to persist in water, and indeed the 16 amino acid helix in the PHD initial structure completely disappeared within  200 ps. Typical peptide conformations from the end of a simulation in water are shown in diagrams 3 and 4 of Fig. 4B. In contrast, the long time behaviour of the simulations in TFE showed about 20% a-helix (not shown). This value is lower than the CD experiments suggest ( 45%). Given the reservations about the TFE potential discussed earlier, this probably indicates that the TFE potential is slightly too hydrophilic in character so that only the most significant a-helix-forming tendency emerges in the simulation. While some fluctuation in the length of the helix was observed, the RR motif was consistently found at the beginning of the helix. Discussion Numerous RR-signal peptides have been characterized in terms of primary sequence and it is now clear that, in most cases, they conform to a standard basic model in which three distinct domains can be identified: a charged N-terminal (N-) domain, hydrophobic core (H-) domain and more polar C-terminal domain terminating with a consensus motif that specifies cleavage by the processing peptidase. Key structural determinants are located at the boundary between the N-domain and the hydrophobic core. In plants, RR-signal peptides contain an essential twin-arginine motif plus a highly hydrophobic residue two or three residues towards the C-terminus. Typically, the sequence RRXXL/F/I/M or RRXL/F/I/M is found [35]. In E. coli, a slightly different consensus is observed, namely RRxFLK [8]. However, E. coli RR-signal peptides are efficiently recognized by plant thyalkoids indicating a highly conserved interaction between the signal peptide and the Tat translocation apparatus [38,39]. A surprising point is that Tat-dependent signal peptides closely resemble Sec-type signals in overall terms. Sec signals likewise contain the three domains described above and the only immediately notable difference is that the basic residue(s) at the N–H domain junction can be either arginine or lysine. In bacteria, Sec signals also tend to be more hydrophobic than RR-signal peptides. Nevertheless, these similarities raise the possibility that the signals may not always direct targeting by the correct translocation path- way, at least initially. While the primary sequences of Tat-type signal peptides are now relatively well established, very little is known of their secondary structures. This is an important issue for two reasons. Firstly, it is now clear that the twin-arginine motif plays a crucial role in the translocation process and this could be either (a) because its sidechains are specifically recognized by the Tat translocase, and/or (b) because it confers a specific secondary structure determin- ant that is important in the context of the remainder of the peptide. Second, the targeting of Tat substrates is in most (if not all) cases an obligatorily post-translational 3348 M. San Miguel et al. (Eur. J. Biochem. 270) Ó FEBS 2003 process, raising the possibility that the signal may adopt different structures before and after reaching the target membrane. In this study, we have used the SufI signal peptide as a model and the data show that this signal can adopt two radically different structures. In aqueous solution, our data show that the signal contains essentially no stable secondary structure. CD is particularly sensitive for the detection of a-helices, and the data show quite clearly that the signal has no stable helical elements, presumably because of compe- tition for hydrogen bonding by water molecules. After release from the ribosome, Tat signal peptides are therefore likely to exist as unstructured peptides. In more hydro- phobic environments, namely 50 : 50 water/TFE or SDS micelles, the structure is very different and the wild-type SufI peptide contains approximately 40 and 25% a-helical structure, respectively. Both secondary structure predictions and MD simulations of SufI in TFE (explicit solvent) strongly suggest that the helical structure is located in the centre section of the peptide, starting either just before or at Fig. 3. Thermal stability of the helical regions in SufI signal peptides. (A) The wild-type SufI peptide in TFE/water (50 : 50 v/v) was ana- lyzed by CD and spectra were taken at the temperatures indicated. (B) Similar CD spec- tra were recorded using the SufI-AA and SufI- KK peptides (not shown) and the graph shows plots of the change in De at 201 nm as a function of temperature for all three peptides. Ó FEBS 2003 Structural characteristics of Tat-specific targeting signals (Eur. J. Biochem. 270) 3349 the twin-arginine motif. The MD simulations also indicate that the formation of an a-helix is very rapid in a hydrophobic environment. While the twin-arginine is strongly predicted to lie at or near the starting point for the helical domain, our data indicate that it does not itself dictate the formation or extent of the helical sequence in a specific manner because a mutant version of the peptide containing twin-lysine contains almost exactly the same amount of a-helical structure. This finding strongly suggests that the real significance of the twin-arginine motif is concerned with protein–protein interactions involving the two extensive arginyl sidechains. On the basis of these data we suggest the following model for the structures of Tat signal peptides during the overall translocation process (Fig. 5). Immediately after synthesis, the signal peptide probably exists in equilibrium between unstructured peptide and a more structured peptide con- taining substantial a-helix in the core region. While in aqueous solvent, the unstructured peptide predominates, to the point where the helical form is essentially undetectable. There is no evidence for the involvement of any soluble Fig. 4. Predicted secondary structure in the SufI signal peptide. (A) Secondary structure for the wild-type SufI peptide (WT) predicted using the PSIPRED , JPRED , PROF and PHD programs. C indicates random coil, H an a-helix and E an extended or sheet motif. The percentage helical content of the peptides is shown on the right. (B) 3D representations (ribbons mode on the left and bonds mode on the right) of the initial structures for the WT peptide obtained from the PHD program (1 and 2) and final configurations from an MD simulation in water (3 and 4); the twin-arginine motif has been depicted using a space-filling representation. 3350 M. San Miguel et al. (Eur. J. Biochem. 270) Ó FEBS 2003 chaperone-type proteins and we therefore believe that the next protein–protein interaction is with the Tat translocase, probably by one of two pathways as shown in Fig. 5. In pathway A, the helical structure is promoted by interaction with the target membrane. There is now good evidence that a-helix formation is strongly promoted within the mem- brane interfacial region [40] and the precursor protein may then migrate laterally on the membrane surface until it contacts the translocase. In pathway B, the signal interacts directly with the translocase and the helical conformation is generated either during entry into the interfacial region or after the initial interaction with the binding site. Any form of typical binding site/groove probably also favours the formation of secondary structure, and the enormously specific signal peptide–translocase interaction would be difficult to achieve if the peptide were largely unstructured. We therefore contend that the signal peptide must be in the helical form when docked onto the translocase binding site. A precedent is the interaction of mitochondrial targeting peptides with the Tom20 receptor, where the targeting signal is likewise unstructured in aqueous solution but in the form of an amphipathic a-helix when bound to the receptor [41,42]. Further work should determine whether the same applies to the Tat-dependent translocation process. Acknowledgements This work was supported by Engineering and Physical Sciences Research Council grant GR/R36503 to CR and PMR. References 1. Robinson, C. & Bolhuis, A. (2001) Protein targeting by the twin-arginine translocation pathway. Nat. Rev. Mol. Cell. Biol. 2, 350–355. 2. Dalbey, R.E. & Kuhn, A. (2000) Evolutionarily related insertion pathways of bacterial, mitochondrial, and thylakoid membrane proteins. Annu. Rev. Cell Dev. Biol. 16, 51–87. 3. Yen, M.R., Tseng, Y.H., Nguyen, E.H., Wu, L.F. & Saier, M.H. Jr (2002) Sequence and phylogenetic analyses of the twin-arginine targeting (Tat) protein export system. Arch. Microbiol. 177, 441–450. 4. Chaddock,A.M.,Mant,A.,Karnauchov,I.,Brink,S.,Herrmann, R.G., Klo ¨ sgen, R.B. & andRobinson, C. (1995) A new type of signal peptide: central role of a twin-arginine motif in transfer signals for the DpH-dependent thylakoidal protein translocase. EMBO J. 14, 2715–2722. 5. Stanley, N.R., Palmer, T. & Berks, B.C. (2000) The twin arginine consensus motif of Tat signal peptides is involved in Sec- independent protein targeting in Escherichia coli. J. Biol. Chem. 275, 11591–11596. 6. Clark, S.A. & Theg, S.M. (1997) A folded protein can be trans- ported across the chloroplast envelope and thylakoid membranes. Mol. Biol. Cell 8, 923–934. 7. Hynds, P.J., Robinson, D. & Robinson, C. (1998) The Sec- independent twin-arginine translocation system can transport both tightly folded and malfolded proteins across the thylakoid membrane. J. Biol. Chem. 273, 34868–34874. 8. Berks, B.C. (1996) A common export pathway for proteins binding complex redox cofactors? Mol. Microbiol. 22, 393–404. 9. Santini, C L., Ize, B., Chanal, A., Mu ¨ ller, M., Giordano, G. & Wu, L F. (1998) A novel Sec-independent periplasmic protein translocation pathway in Escherichia coli. EMBO J. 17, 101–112. 10. Weiner, J.H., Bilous, P.T., Shaw, G.M., Lubitz, S.P., Frost, L., Thomas, G.H., Cole, J.A. & Turner, R.J. (1998) A novel and ubiquitous system for membrane targeting and secretion of cofactor-containing proteins. Cell 93, 93–101. 11. Sargent, F., Bogsch, E.G., Stanley, N.R., Wexler, M., Robinson, C., Berks, B.C. & andPalmer, T. (1998) Overlapping functions of components of a bacterial Sec-independent export pathway. EMBO J. 17, 3640–3650. 12. Bogsch, E.G., Sargent, F., Stanley, N.R., Berks, B.C., Robinson, C. & Palmer, T. (1998) An essential component of a novel bac- terial protein export system with homologues in plastids and mitochondria. J. Biol. Chem. 273, 18003–18006. 13. Sargent, F., Stanley, N.R., Berks, B.C. & Palmer, T. (1999) Sec- independent protein translocation in Escherichia coli.Adistinct and pivotal role for the TatB protein. J. Biol. Chem. 274, 36073– 36082. 14. Voelker, R. & Barkan, A. (1995) Two nuclear mutations disrupt distinct pathways for targeting proteins to the chloroplast thyla- koid. EMBO J. 14, 3905–3914. 15. Settles, M.A., Yonetani, A., Baron, A., Bush, D.R., Cline, K. & Martienssen, R. (1997) Sec-independent protein translocation by the maize Hcf106 protein. Science 278, 1467–1470. 16. Walker,M.B.,Roy,L.M.,Coleman,E.,Voelker,R.&Barkan,A. (1999) The maize tha4 gene functions in sec-independent protein transport in chloroplasts and is related to hcf106, tatA, and tatB. J. Cell. Biol. 147, 267–276. 17. Bolhuis, A., Mathers, J.E., Thomas, J.D., Barrett, C. & Robinson, C. (2001) TatB and TatC form a structural and functional unit of the twin-arginine translocase of Escherichia coli. J. Biol. Chem. 276, 20213–20219. 18. Cline, K. & Mori, H. (2001) Thylakoid DpH-dependent precursor proteins bind to a cpTatC-Hcf106 complex before Tha4-depend- ent transport. J. Cell Biol. 154, 719–729. Fig. 5. Models for the structures of signal peptides during the overall Tat-dependent translocation process. Immediately after synthesis, the signal peptide is exposed to an aqueous environment where it is in equilibrium between the unstructured (1) or substantially a-helical (2) conformations. The equilibrium is strongly in favour of the unstruc- tured peptide, probably because of competition for H-bonding by water molecules. Thereafter, targeting to the Tat translocase can occur by one of two routes. In pathway A, the peptide interacts with the membrane and a-helix formation is promoted within the inter- facial region, after which the precursor protein migrates laterally until contact with the translocase is made. In B, the precursor protein interacts directly with the translocase, in which a-helix formation may be promoted within the binding site. Ó FEBS 2003 Structural characteristics of Tat-specific targeting signals (Eur. J. Biochem. 270) 3351 19. Mori, H. & Cline, K. (2002) A twin arginine signal peptide and the pH gradient trigger reversible assembly of the thylakoid Delta pH/Tat translocase. J. Cell Biol. 157, 205–210. 20. Allen,S.C.H.,Barrett,C.M.L.,Ray,N.&Robinson,C.(2002) Essential cytoplasmic domains in the Escherichia coli TatC pro- tein. J. Biol. Chem. 277, 10362–10366. 21. Buchanan, G., Leeuw, E., Stanley, N.R., Wexler, M., Berks, B.C., Sargent, F. & Palmer, T. (2002) Functional complexity of the twin- arginine translocase TatC component revealed by site-directed mutagenesis. Mol. Microbiol. 43, 1457–1470. 22. Roise, D., Theiler, F., Horvath, S.J., Tomich, J.M., Richards, J.H., Allison, D.S. & Schatz, G. (1988) Amphiphilicity is essen- tial for mitochondrial presequence function. EMBO J. 7, 649–653. 23. Jones, J.D., McKnight, C.J. & Gierasch, L.M. (1990) Biophysical studies of signal peptides: implications for signal sequence func- tions and the involvement of lipid in protein export. J. Bioenerg. Biomembr. 22, 213–232. 24. Woody, R.W. (1994) Circular Dichroism: Principles and Applica- tions (Nakanishi, K., Berova, N. & Woody, R.W., eds). VCH, New York. 25. Johnson, W.C. (1999) Analyzing protein circular dichroism spec- tra for accurate secondary structures. Proteins Struct. Funct. Genet. 35, 307–312. 26. Smith, W., Yong, C.W. & Rodger, P.M. (2002) DL_POLY: application to molecular simulation. Mol. Sim. 28, 385–471. 27. Rost, B. (1996) PHD: predicting one-dimensional protein struc- ture by profile-based neural networks. Methods Enzymol. 266, 525–539. 28. Frishman, D. & Argos, P. (1995) Knowledge-based protein sec- ondary structure assignment. Proteins 23, 566–579. 29. Chitra, R. & Smith, P.E.J. (2001) A comparison of the prop- erties of 2,2,2-trifluoroethanol and 2,2,2-trifluoroethanol/water mixtures using different force fields. J. Chem. Phys. 115, 5521– 5530. 30. Rodziewicz-Motowidio, S., Brzozowski, K., Legowska, A., Liwo, A., Silbering, J., Smoluch, M. & Rolka, K.J. (2002) Conforma- tional solution studies of neuropeptide gamma using CD and NMR spectroscopy. J. Pept. Sci. 8, 211–226. 31. McGuffin, L.J., Bryson, K. & Jones, D.T. (2000) The PSIPRED protein structure prediction server. Bioinformatics 16, 404–405. 32. Cuff, J.A. & Barton, G.J. (1999) Evaluation and improvement of multiple sequence methods for protein secondary structure pre- diction. Proteins 40, 502–511. 33. Ouali, M. & King, R.D. (2000) Cascaded multiple classifiers for secondary structure prediction. Prot. Sci. 9, 1162–1176. 34. Henry, R., Carrigan, M., McCaffrey, M., Ma, X. & Cline, K. (1997) Targeting determinants and proposed evolutionary basis for the Sec and the delta pH protein transport systems in chloroplast thylakoid membranes. J. Cell Biol. 136, 823–832. 35. Brink, S., Bogsch, E.G., Edwards, W.R., Hynds, P.J. & Robinson, C. (1998) Targeting of thylakoid proteins by the DpH-driven twin- arginine translocation pathway requires a specific signal in the hydrophobic domain in conjunction with the twin-arginine motif. FEBS Letts 434, 425–430. 36. Alami, M., Trescher, D., Wu, L.F. & Muller, M. (2002) Separate analysis of twin-arginine translocation (Tat)-specific membrane binding and translocation in Escherichia coli. J. Biol. Chem. 277, 20499–20503. 37.deLeeuw,E.,Granjon,T.,Porcelli,I.,Alami,M.,Carr,S., Muller, M., Sargent, F., Palmer, T. & Berks, B. (2002) Oligomeric properties and signal peptide binding by Escherichia coli Tat protein transport complexes. J. Mol. Biol. 322, 1135–1146. 38. Mori, H. & Cline, K. (1998) A signal peptide that directs non-Sec transport in bacteria also directs efficient and exclusive transport on the thylakoid Delta pH pathway. J. Biol. Chem. 273, 11405– 11408. 39. Wexler, M., Bogsch, E.G., Palmer, T., Robinson, C. & Berks, B.C. (1998) Targeting signals for a bacterial Sec-independent export system direct plant thylakoid import by the delta pH pathway. FEBS Letts. 431, 339–342. 40. White, S.H. & Wimley, W.C. (1999) Membrane protein folding and stability: physical principles. Annu. Rev. Biophys. Biomol. Struc. 28, 319–365. 41. Abe, Y., Shodai, T., Muto, T., Mihara, K., Torii, H., Nishikawa, S., Endo, T. & Kohda, D. (2000) Structural basis of presequence recognition by the mitochondrial protein import receptor Tom20. Cell 100, 551–560. 42. Epand, R.M., Hui, S.W., Argan, C., Gillespie, L.L. & Shore, G.C. (1986) Structural analysis and amphiphilic properties of a chemi- cally synthesized mitochondrial signal peptide. J. Biol. Chem. 261, 10017–10020. 3352 M. San Miguel et al. (Eur. J. Biochem. 270) Ó FEBS 2003 . An Escherichia coli twin-arginine signal peptide switches between helical and unstructured conformations depending on the hydrophobicity of the environment Miguel San Miguel 1 ,. representations (ribbons mode on the left and bonds mode on the right) of the initial structures for the WT peptide obtained from the PHD program (1 and 2) and final configurations from an MD simulation. method with thermostat and barostat relax- ation constants of 0.5 and 1.0, respectively, and a timestep of 2 fs. The peptide and TFE were modelled with the CHARMM potential, and water with the SPC