Efficient type-III secretion depends on cytosolic molecular chaperones, which bind specifically to the translocators and effectors. In the past there has been a tendency to shoe-horn all type-III-secretion chaperones into a single structural and functional class. However, we have shown that the LcrH/SycD-like chaperones consist of three central tetratricopeptide-like repeats that are predicted to fold into an all-alpha-helical array that is quite distinct from the known structure of the SycE class of chaperones. Furthermore, we predict that this array creates a peptide-binding groove that is utterly different from the helix-binding groove in SycE. We present a homology model of LcrH/SycD that is consistent with existing mutagenesis data. We also report the existence of tetratricopeptide-like repeats in regulators of type-III secretion, such as HilA from Salmonella enterica and HrpB from Ralstonia solanacearum. The discovery of tetratricopeptide-like repeats in type-III-secretion regulators and chaperones provides a new conceptual framework for structural and mutagenesis studies and signals a potential unification of prokaryotic and eukaryotic chaperone biology.
Many bacterial pathogens use a complex multi-protein apparatus, the type-III-secretion system (TTSS), to deliver so-called ‘effector’ proteins into the cytosol of eukaryotic cells, which then subvert host cell signalling or cytoskeletal functions [1–3]. Efficient type-III secretion depends on cytosolic molecular chaperones, which bind specifically to the effectors and to the secreted proteins required for translocation (the ‘translocators’) [4,5]. Sequence identity between type-III-secretion chaperones is low, although some authors have stressed common features such as small size (100–150 residues), a C-terminal amphipathic helix and a tendency towards an acidic pI[5–7]. The structures of four chaperones (SicP, SycE, SigE, CesT) have been reported and the clear structural homology between them — each showing a complex alpha–beta fold with a helix-binding groove formed from a highly twisted beta-sheet — has tended to reinforce the view that all type-III-secretion chaperones can be shoe-horned into a single structural and functional class [8–12].
The chaperone hypothesis — that proteins often require other proteins to assist them to reach their fully folded and active confirmation, particularly when they have to traverse membranes to reach their target environment – now encompasses many aspects of bacterial and eukaryotic biology. However, no links have so far been established between TTSS chaperones and eukaryotic chaperones. The tetratricopeptide repeat (TPR) is an imperfect typically 34-amino-acid repeat often arranged in tandem arrays [14–16] that is found in eukaryotic molecular chaperone complexes involving HSP70 and HSP90, where the immunophilins Cyp40 and FKBP52 or the Tom70 mitochondrial import receptor use TPRs to bind the heat shock proteins [17–21]. Although the sequence is poorly conserved, similar residue types are usually found at three canonical positions (residues 8, 20, 27). Several protein structures confirm that individual TPRs adopt a helix-turn-helix conformation and that tandem TPRs fold into concertina-like helical arrays that form a continuous peptide-binding groove [22–25]. Given the link between chaperones and TPRs in eukaryotes, we performed a search for TPR-like repeats in proteobacterial protein sequences. The results of this search show that tetratricopeptide-like (TPR-like) repeats characterise the LcrH/SycD class of type-III-secretion chaperones and also occur in some transcriptional regulators of type-III secretion.
Materials and methods
Initial PSI-BLAST search for proteobacterial TPR-like repeats
A PSI-BLAST search  of the NR database was performed on the NCBI site (http://www.ncbi.nlm.nih.gov/BLAST/) on September 26th, 2002, with the three TPRs of known structure from the human serine/threonine protein phosphatase 5 as the query sequence (residues 28–129 of SwissProt entry PPP5_HUMAN). The default parameters for the service were used (BLOSUM 65 matrix; E-value for inclusion in subsequent iterations 0.005; word size 3; gap costs: existence 11, extension 1; no filter), except that the number of descriptions and alignments to be reported was set at 1000, the search was limited by Entrez query to ‘proteobacteria’ and ‘composition-based statistics’ set to off. Additional PSI-BLAST searches were performed on the ViruloGenome web site (http://www.vge.ac.uk) using exemplary proteins from each class of TPR-like-repeat proteins from TTSSs. Multiple alignments were performed and edited using ClustalW and Jalview at the EBI (http://www.ebi.ac.uk/clustalw/) and were shaded using the Boxshade server (http://www.ch.embnet.org/software/BOX_form.html).
Homology modelling of TPR-like repeats in chaperones
The energy-minimised model of SycD/LcrH (and of the three other homologues) were generated using Modeller version 4 , based on the coordinates of the highest scoring template and the sequence alignment returned by 3D-PSSM. The modelling was restricted to the core region of the alignment (SycD/LcrH: 32–165; SicA: 38–157; SycDCm: 93–240; CesD: 36–149), with no gaps within the three TPR-like repeats. The model was refined by conjugate-gradient minimisation.
The rigid body docking used programmes of the 3D-Dock software suite (http://www.bmm.icnet.uk/docking/index.html). Rigid body docking was performed with FTDOCK  probing 9230 different translations/rotations of the YopD helix. The docking algorithm evaluates geometric surface complementarity and an electrostatic filter. The resulting complexes were evaluated by residue level pair potential scoring using the programme RPSscore  and ranked according to this score.
TPR-like repeats in proteins from TTSSs
In an initial search of proteobacterial proteins with TPRs from human protein phosphatase 5, PSI-BLAST reported, by the third iteration, that there were 4080 BLAST hits to the query sequence. Hits to at least a thousand proteins had significance levels below the Expect value threshold of 0.005 (data not shown). Scrutiny of the results revealed highly significant hits (Expect values <104) to many type-III-secretion proteins, which fell into three broad classes: known and putative type-III-secretion chaperones, predominantly in the SycD/LcrH family; transcriptional regulators similar to HilA from the Salmonella enterica SPI1 system; and transcriptional regulators from plant pathogens similar to HrpB from Ralstonia solanacearum. The TPR-like repeat proteins from TTSSs were subjected to additional PSI-BLAST searches, which in most cases confirmed the link to known TPR proteins and also retrieved similar sets of type-III-secretion proteins to those found by the initial search (data not shown).
TPR-like repeats in SycD-like chaperones
A PSI-BLAST search with the sequence of the YopB/D chaperone SycD/LcrH from Yersinia pseudotuberculosis[6,30,31] revealed strong near-full-length homology with over a dozen other proven or putative type-III-secretion chaperones (Fig. 1), as well as confirming the presence of TPR-like repeats in all members of this family by finding hits to proven TPR proteins. Two other proteins less strongly related to the SycD/LcrH — YscY from Yersinia enterocolitica and Pcr4 from Pseudomonas aeruginosa– were also reported with significance in this search. A multiple alignment (Fig. 1) confirmed that there are three tandem TPR-like repeats in the SycD/LcrH-like chaperones (including YscY and Pcr4) that can be aligned with minimal gaps with the three known TPRs from protein phosphatase 5 and show near-complete conservation of residue type at the canonical sites in the TPR-like repeats (small residues at positions 8 and 20, small or hydrophobic residues at position 27). These three TPR-like repeats comprise nearly the full length of the protein sequence in YscY and Pcr4. The central block of TPR-like repeats (residues 36–137 in the 168-residue SycD/LcrH) is flanked in the other proteobacterial chaperones by short N-terminal and C-terminal sequences (35 and 31 residues respectively in SycD/LcrH) and in the chlamydial chaperones by additional longer lysine-rich extensions (Fig. 1). Also reported in the PSI-BLAST search with SycD/LcrH, but with a marginal E-value (0.021), was a TTSS protein of unknown function, HrpK from R. solanacearum.
Homology modelling of TPR-like repeats in chaperones
Threading the sequence of SycD/LcrH onto 7890 structural templates and scoring for compatibility, returned, with high confidence (E-value 0.004), a homology model based on the crystal structure of the TPR domain of protein phosphatase 5 . A refined energy-minimised model of LcrH appears plausible in that it is free from steric clashes and shows all non-glycine residues in the allowed regions of the Ramachandran plot. Similar procedures were applied to SicA from S. enterica[35,36], a SycD/LcrH homologue from Chlamydia muridarum (SycDcm in Fig. 3)  and CesD from enteropathogenic Escherichia coli.
The structure-based superimposition of the three TPR-like repeats in these four homology models yielded an alignment in agreement with the PSI-BLAST results. The three TPR-like repeats in the SycD/LcrH model (and in the other three models) form a groove capable of binding an extended peptide from the cognate partner protein (Fig. 2). This groove, formed by an all-alpha-helical array, is utterly different from the helix-binding groove formed by the highly twisted beta-sheet in the SycE class of ‘chaperones of one effector’[8–12]. The base of the groove is lined with residues from the first helix of TPR2 (Phe-72, Arg-71, Leu-74, Gly-75, Ala-78, Cys-79, Gln-81, and Ala-82) and the first helix of TPR3 (Pro-104, Arg-105, Phe-108, His-109, Ala-111, Glu-112, Leu-115). The N-terminal side of the peptide-binding groove is almost exclusively formed by aromatic residues in the first TPR-like repeat (A-helix: Tyr-40, Phe-44, Tyr-47, B-helix: Tyr-52, Phe-59), which show a high degree of sequence conservation across the TTSS chaperones aligned in Fig. 1. This largely hydrophobic peptide-binding interface is in agreement with mutagenesis studies on YopD and SycD/LcrH, which suggested a predominantly hydrophobic interaction . However, three charged residues, Arg-71, Arg-105, and Glu-112, prominently dot the surface of the groove. They show a distinctly higher degree of variability across the TTSS chaperones than the series of aromatic residues in TPR1, suggesting that they might confer peptide-binding specificity.
We tested the model by examining the distribution of SycD/LcrH mutations from a pre-existing set of randomly generated mutants that were already known to retain YopD binding . As one would predict from the model, none of the 12 mutations (K20E, E30G, I31V, L42F, N45I, Y86H, H91Y, I98M, M99V, M99T, D136G, V165I) affected the hydrophobic residues lining the base of the groove.
The alignment of the four homology models revealed an extended patch of residues with a high degree of sequence similarity across the chaperone sequences aligned in Fig. 1 (see Fig. 3A). This surface patch runs almost the entire length of the peptide-binding groove and is mainly, but not exclusively, formed by the aromatic residues located in the first TPR-like repeat. This extended ‘low diversity’ patch is complemented by two smaller ‘low diversity’ patches on the TPR-3 side of the groove. Only a single patch of sequence conservation, formed by Tyr-93, is found elsewhere on the protein, which is on the back side of the peptide-binding groove. Thus, in this homology model of SycD/LcrH positions of sequence conservation map almost exclusively to the base and the sides of the peptide-binding groove. Furthermore, three single site mutations in SycD/LcrH (F72L, C79Y, H67Y) known to disrupt the interaction with YopD  coincide with or map very closely to the extended surface patch of conserved residues (Fig. 3B).
Next, we predicted the binding site of the amphipathic helix of YopD on the homology model of LcrH using RPScore. Rigid body docking of the nuclear magnetic resonance (NMR) structure of the amphipathic helix of YopD  onto the static surface of the LcrH homology model yielded a large set of putative complexes ranked by a residue level pair-potential function. Four of the five highest scoring orientations placed YopD with its helix axis roughly parallel to the extended ‘low diversity patch’, forming an extensive contact interface with this part of the peptide-binding groove. The complex ranked fourth, ‘Complex 4’, is particularly intriguing, since its position and orientation is consistent with yeast-2-hybrid data and NMR spectroscopy data. These data identified three hydrophobic residues (Phe-280, Met-281, Ile-288) in YopD as having a severe effect on binding to SycD/LcrH when mutated to alanine or an amino acid with contrasting physical properties . Furthermore, the NMR study observed line broadening for Tyr-291 and Val-292 upon binding to SycD/LcrH . In the orientation of Complex 4, the hydrophobic residues shown to disrupt yeast-2-hybrid interactions between SycD/LcrH and YopD, are all in contact with residues of the ‘low diversity’ patch and their side chains become fully (Ile-288) or partially buried (Met-281, Phe-280) upon complex formation. Likewise, Tyr291 and Val292 in Complex 4 contact the surface of SycD/LcrH, with Val-292 fully buried in the complex. Furthermore, Asp-283, the only polar side chain with a severe effect in the yeast-2-hybrid assay  is located in the immediate vicinity of the guanido groups of Arg-71 and Arg-105, two basic residues that feature prominently in the peptide-binding groove and might play a role in mediating specificity. While Asp-283 does not form a salt bridge with either of the two Arg side chains in Complex 4, only very minor adjustments in helix orientation or side chain conformations would be needed for such an interaction. In conclusion, our homology models yield a structural framework that readily accommodates the functional data previously obtained for the LcrH:YopD interaction [31,40]. All models have been deposited in the PDB database.
TPR-like repeats in TTSS regulators
HilA is a transcriptional regulator responsible for the activation of the SPI-1 TTSS in S. enterica. Intriguingly, several hits in the PSI-BLAST search with protein phosphatase 5 were located in the C-terminal region of the HilA protein suggesting the presence of TPR-like repeats. A follow-up PSI-BLAST search with this region revealed hundreds of hits to known TPR proteins. Scrutiny of the PSI-BLAST alignments to proteins with multiple tandem TPRs (e.g. the human Bardet–Biedl syndrome 4 protein) allowed identification of the canonical TPR residues in the HilA TPR-like repeats. Using this approach, nine TPR-like repeats were detected, spanning residues 223–548, and thus encompassing almost all of the protein C-terminal to the known regulatory domain (Fig. 4). Using similar approaches, a TTSS-associated HilA homologue from E. coli, YgeH, was found to possess TPR-like repeats 1–5 but to lack TPR-like repeats 6–9, as compared to HilA (Fig. 4). Hits in the PSI-BLAST search with protein phosphatase 5 also identified three tandem TPR-like repeats in the N-terminal region of HrpB protein, a TTSS transcriptional regulator from R. solanacearum.
TPR-like repeats and type-III-secretion chaperones
The discovery of TPR-like repeats in type-III-secretion chaperones lends credence to the classification of such chaperones recently proposed by Page and Parsot . On sequence homology and structural grounds, it is now clear that type-III-secretion chaperones can no longer all be lumped into one amorphous category, but that, instead, there are two major classes — the SycD/LcrH-like TPR-like repeat all-alpha class and the SycE-like alpha–beta class — and that these map nicely onto the functionally defined classes of ‘chaperones of the translocators’ and ‘chaperones of one effector’. The structural demarcation between the ‘chaperones of the translocators’ and ‘chaperones of one effector’ is consistent with them having differing functions, shepherding their substrates for secretion at different points in a temporal hierarchy of secretion – it makes sense for the translocators, which must be secreted first, to be handled or guided in a different way from the effectors.
TPR-like repeats and type-III-secretion regulators
The recruitment of TPR-like domains into transcriptional regulators of type-III secretion appears to have occurred independently in subsets of animal and plant pathogens, as the domain organisation of HilA/YgeH (N-terminal trans_reg_c domain, C-terminal TPR-like repeats) differs dramatically from that of HrpB/HrpX (N-terminal TPR-like repeats, C-terminal HTH_AraC subdomains). As all known helical-repeat folds (TPRs, armadillo repeats, HEAT motifs, leucine-rich variant motifs) have been implicated in protein–protein interactions , the presence of TPR-like repeats in these TTSS regulatory proteins strongly implicates them in signal-dependent heterologous protein–protein interactions and/or oligomerisation.
While the repeats in the type-III chaperones are almost certain to be true TPRs in that all but one of them are 34 amino acids long, the repeats in these regulators are more variable in length, although the evidence from the PSI-BLAST searches and the alignment in Fig. 4 still support the designation ‘TPR-like’. Helical repeats similar to TPRs have recently been discovered in the crystal structure of domain III of the ATP-dependent transcriptional activator of the maltose regulon, MalT, where they appear to form a maltotriose-modulated protein–protein interaction interface (these TPR-like repeats were also identified in our PSI-BLAST search with protein phosphatase 5), . Similarly, TPRs have been identified by sequence analysis within GutR, the ATP-dependent activator of the glucitol operon in Bacillus subtilis. In neither regulator have the functions of these repeats been well defined, although the MalT repeats appear to form a potential maltotriose-binding site. Although the HilA-like and HrpB-like classes of TPR-like-repeat-bearing TTSS regulators differ from MalT and GutR in being ATP independent, it is reasonable to speculate that their TPR-like repeats perform similar functions, mediating protein–protein interactions, and that the studies on any of these regulators will illuminate the mechanism of action of all of them. A more tentative speculation is that TPR-mediated protein–protein interactions in the HilA-like and HrpB-like regulators are modulated by the binding of a small molecule, as appears to be the case for MalT and GutR [46,47]. Additional evidence for this idea comes from the recent structure of the atypical penicillin-binding protein, HcpB from Helicobacter pylori, in which TPR-like repeats form a suprahelical groove that is thought to bind penicillin and N-acetylmuramic acid .
In discovering and describing TPR-like repeats in chaperones and regulators from bacterial TTSSs, we have provided a novel conceptual framework for future studies. Furthermore, the discovery of TPR-like repeats in both bacterial and eukaryotic chaperones signals a potential unification of prokaryotic and eukaryotic chaperone biology, so that researchers in both fields might be able to draw on each other's work. The multiple alignments and homology models presented here facilitate the parsing of the TPR-like-repeat-bearing chaperones and regulators into their constituent domains (TPR 1, 2, 3 and N-terminal and C-terminal extensions in the case of the chaperones) and even secondary structural features, so that truncation-mutagenesis and domain-swapping studies can now be performed on a rational basis for each chaperone or regulator. Similarly, our predictions as to which residues are (or are not) important in the structural integrity of the TPR-like repeats (the canonical residues in both chaperones and regulators) or in substrate binding (those in the groove for chaperones) provide a firm foundation for future site-directed mutagenesis experiments. These kinds of experiments can now proceed in advance of the availability of a definitive structure, as they have for numerous TPR domains in eukaryotic proteins [21,49–55]. Finally, the new paradigm of TPR-like repeats mediating bacterial chaperone and regulator protein–protein interactions provides added impetus to the efforts to obtain crystal or NMR structures.
M.P. thanks the BBRSC for funding the ViruloGenome project. M.F. wishes to thank Swedish Research Council and the Foundation for Medical Research at Umeå University for research funding.