PDZ Domains Facilitate Binding of High Temperature Requirement Protease A (HtrA) and Tail-specific Protease (Tsp) to Heterologous Substrates through Recognition of the Small Stable RNA A (ssrA)-encoded Peptide*

The Escherichia coli protease HtrA has two PDZ domains, and sequence alignments predict that the E. coli protease Tsp has a single PDZ domain. PDZ domains are composed of short sequences (80–100 amino acids) that have been implicated in a range of protein:protein interactions. The PDZ-like domain of Tsp may be involved in binding to the extreme COOH-terminal sequence of its substrate, whereas the HtrA PDZ domains are involved in subunit assembly and are predicted to be responsible for substrate binding and subsequent translocation into the active site. E. coli has a system of protein quality control surveillance mediated by the ssrA-encoded peptide tagging system. This system tags misfolded proteins or protein fragments with an 11-amino acid peptide that is recognized by a battery of cytoplasmic and periplasmic proteases as a degradation signal. Here we show that both HtrA and Tsp are able to recognize the ssrA-encoded peptide tag with apparent K D values of ∼5 and 390 nm, respectively, and that their PDZ-like domains mediate this recognition. Fusion of the ssrA-encoded peptide tag to the COOH terminus of a heterologous protein (glutathioneS-transferase) renders it sensitive to digestion by Tsp but not HtrA. These observations support the prediction that the HtrA PDZ domains facilitate substrate binding and the differential proteolytic responses of HtrA and Tsp to SsrA-tagged glutathioneS-transferase are interpreted in terms of the structure of HtrA.

able with alanine. In a model proposed by Keiler et al. (1), SsrA charged with alanine can bind to stalled ribosomes. After the contribution of the ssrA-encoded alanine, translation switches to a short open reading frame in ssrA that encodes the COOHterminal peptide tag degradation signal. This system provides a general quality control mechanism to dispose of incomplete protein fragments and avoid the build-up of ribosomes stalled on defective mRNA molecules (2)(3)(4)(5)(6)(7). Cytoplasmic proteases that respond to this system are largely ATP-dependent (e.g. the ClpXP and ClpAP proteases) whereas SsrA-tagged proteins with signal sequences are directed to the periplasm and degraded there by ATP-independent proteases.
Tsp is a periplasmic serine protease of E. coli that was purified on the basis of its ability to degrade a variant of the NH 2 -terminal domain of the bacteriophage lambda repressor. The wild-type repressor domain, which is not degraded, contains the polar COOH-terminal sequence Arg-Ser-Glu-Tyr-Glu, whereas the variant repressor protein contains the apolar sequence Trp-Val-Ala-Ala-Ala, and it is this that allows degradation by Tsp (1,8,9). Tsp preferentially degrades substrates that are not stably folded, by digestion at several sites that have broad primary sequence specificity (1). Tsp is thought to bind to the COOH terminus of the protein in question, with no proteolysis occurring until spontaneous unfolding makes the polypeptide available to the protease active site (2,10).
HtrA (also known as DegP) is a second periplasmic protease of E. coli, which can also act as a chaperone (11), and was originally identified as a heat shock-induced protein. It is located on the periplasmic side of the inner membrane and has the conserved triad of His, Ser, and Asp residues that is characteristic of the serine proteases. Homologues of HtrA have been found in a wide range of species including Gram-negative and Gram-positive bacteria, cyanobacteria, yeast, and humans. A natural substrate for HtrA is the periplasmic MalS protein, but HtrA is also able to use ␤-casein as a proteolytic substrate in vitro as the latter is largely unordered in solution (11,12). At low temperatures HtrA acts predominantly as a chaperone and in vitro is able to stimulate the refolding of chemically denatured proteins. The proteolytic activity associated with HtrA predominates at high temperatures. Null mutants in the htrA gene are thermosensitive and have a decreased ability to degrade abnormal periplasmic proteins. A mutant form of HtrA in which Ser 210 has been changed to Ala is correctly folded but proteolytically inactive, and has a reduced ability to suppress the thermosensitive phenotype in a strain deleted for the wildtype htrA gene. These observations link the loss of proteolytic activity with the thermosensitive phenotype (13)(14)(15)(16)(17)(18)(19)(20)(21). This mutant HtrA retains chaperone activity (21). Transcription of the htrA gene is highly regulated by a complex interaction with inner membrane proteins and a more global system of phosphoprotein phosphatases (22).
The crystal structure of a protease-deficient mutant HtrA has been determined, and is a hexamer formed by staggered association of trimeric rings; the proteolytic sites are located in a central cavity that is only accessible laterally (23). HtrA monomers are composed of three domains, an NH 2 -terminal protease/chaperonin domain and two PDZ 1 domains (PDZ1 and 2). PDZ domains are composed typically of 80 -100 amino acids and have been reported in many proteins involved in a range of protein:protein interactions. The acronym PDZ derives from three eukaryotic proteins (post-synaptic density (PSD) protein, disc large and zo-1 (zonula occludens)) in which they were first described (24). In the HtrA multimer, 12 PDZ domains form mobile side walls, with PDZ1 domains interacting with one another suggesting they function as the main gatekeeper to the inner chamber (23). Tsp contains a single PDZ-like domain (1), and is a member of a family of proteases (clan SF, family 41; Ref. 25) that includes the Scenedesmus obliquus D1 COOHterminal processing protease (D1P). The structure of the S. obliquus D1P has been determined, and its ␤-domain shows a high degree of similarity at the secondary and tertiary structural level with the PDZ domains of the human homologue of the Drosophila discs large tumor suppressor gene product (D1gA), nitric-oxide synthase, and the third PDZ domain of PSD-95 (26,27).
It has been proposed that the PDZ domains of HtrA are also involved in substrate binding, thereby directly coupling substrate binding and translocation within the HtrA multimer. It is speculated that this coupled binding and translocation may be facilitated by the PDZ domains acting as tentacular arms capturing substrates and transferring them into the inner cavity (23).
To address the prediction that the HtrA PDZ domains facilitate substrate binding, we tested the ability of HtrA and a protease (Tsp) with no chaperonin activity to bind to the ssrAencoded peptide. Here we show that Tsp and HtrA bind to the isolated ssrA-encoded peptide via their PDZ-like domains. Tsp and HtrA bind the SsrA peptide with different affinities (K D for Tsp ϭ 390 nM; K D for HtrA ϭ 4.9 nM) and fusion of this peptide to the GST protein enhances proteolysis by Tsp but not HtrA.

EXPERIMENTAL PROCEDURES
Materials-Chemicals and solvents were purchased from local suppliers and were of AnalaR or greater purity. Enzyme substrates were purchased from Sigma, and molecular biology reagents (which were used in accordance with the manufacturer recommendations) were purchased from Invitrogen, Amersham Biosciences, or BCL. Three peptides of sequence Ala-Ala-Asn-Asp-Glu-Asn-Tyr-Ala-Leu-Ala-Ala (peptide 1; wild-type SsrA sequence tag), Tyr-Asn-Ala-Leu-Asn-Ala-Asp-Ala-Ala-Ala-Glu (peptide 2; randomized SsrA sequence tag), and Ala-Ala-Asn-Asp-Glu-Asn-Trp-Val-Ala-Ala-Ala (peptide 3; a modified SsrA sequence with the Trp-Val-Ala-Ala-Ala motif used to identify the Tsp protease), all containing a biotin molecule linked via standard carbodiimide condensation, were synthesized in the University of Newcastle upon Tyne Facility for Molecular Biology. A fourth peptide of wild-type SsrA sequence, but lacking the biotin molecule, was also synthesized.
Molecular Cloning for Protein Overproduction in E. coli-The complete coding regions of the E. coli and Salmonella typhimurium htrA genes, the E. coli tsp gene, and the sequences encoding the PDZ do-mains were amplified by the polymerase chain reaction (PCR) and subcloned into E. coli expression vectors (see Table I). PCR amplification was performed according to the following conditions: cycle 1, 94°C for 2 min, 50°C for 2 min, and 72°C for 4 min; cycles 2-30, 94°C for 1 min, 50°C for 2 min, and 72°C for 4 min. All PCR reactions used the Expand high fidelity Taq polymerase (Roche Molecular Biochemicals). Similarly, the gene encoding the GST protein was amplified by the PCR with COOH-terminal extensions of either Ala-Ala-Asn-Asp-Glu-Asn-Tyr-Ala-Leu-Ala-Ala or Tyr-Asn-Ala-Leu-Asn-Ala-Asp-Ala-Ala-Ala-Glu. Site-directed mutagenesis, as previously described (28), using a mutagenic oligonucleotide of sequence AACCGTGGTACCGCCGGTGGTG-CGCTG and a non-mutagenic oligonucleotide of sequence GATCGCT-GCATCGGTCTGGAT and plasmid pTR130 as template were used to generate the HtrA Ser 210 3 Ala mutant. A mutagenic oligonucleotide of sequence GAACCGACGTTTGGCGCCGGCACCGTTCAG and a nonmutagenic oligonucleotide of sequence ACCCACAACCAGCGCACGAC-CGTAATC and plasmid pTR147 as template were used to generate the Tsp Lys 455 3 Ala mutant. The amino acids encoded by the various plasmid constructs and the oligonucleotides used in each PCR are shown in Table I. The correct sequences and the absence of PCRgenerated artifacts in the cloned sequences were verified by directly sequencing the double-stranded plasmid DNA on an ABI/PerkinElmer Life Sciences 377 automated DNA sequencer.
Protein Overproduction and Purification-Recombinant plasmids designated pTR129 and130 were used to transform the E. coli expression strain BL21 (DE3) to screen for E. coli and S. typhimurium HtrA overproduction in the presence of 0.2 mg ml Ϫ1 IPTG. Soluble overproduction of both HtrA proteins was achieved, and analysis by SDS-PAGE showed that greater than 90% of HtrA was processed by removal of the signal peptide. For a typical purification of native HtrA, 4 liters of cells grown at 37°C in 500-ml batches in rich medium were induced by the presence of 0.2 mg ml Ϫ1 IPTG when the cells were in mid logarithmic growth and harvested by centrifugation at 2,500 ϫ g. Following disruption by sonication in 450 ml of 50 mM potassium phosphate, pH 7.2, 1 mM DTT, 5 mM EDTA (buffer 1), the cell suspension was clarified by centrifugation at 2,500 ϫ g for 30 min at 4°C. The clarified supernatant was applied to a DEAE-Sephacel column and eluted with 500 ml of the same buffer. The HtrA-containing column flow-through was combined, made 1.0 M with ammonium sulfate, and applied to a phenyl-Sepharose column. After washing with buffer 1 containing 1.0 M ammonium sulfate, the phenyl-Sepharose column was eluted with a 1-liter gradient consisting of 500 ml of 1.0 M ammonium sulfate in buffer 1 connected to 500 ml of 10 mM potassium phosphate buffer, pH 7.2, 1 mM DTT (buffer 2). HtrA protein was then eluted from the column in a single batch wash of 500 ml of buffer 2 after completion of the ammonium sulfate gradient. The HtrA pool was applied to a hydroxyapatite column and washed with 500 ml of buffer 2. The column was then eluted with an 800-ml gradient consisting of 400 ml of buffer 2 connected to 400 ml of 400 mM potassium phosphate, pH 7.2, 1 mM DTT (buffer 3). At the end of the gradient, purified HtrA was eluted with a batch wash of 400 ml of 400 mM potassium phosphate buffer, pH 6.6, 1 mM DTT (buffer 4). Typical yields from this procedure were 150 mg of HtrA from 4 liters of original culture. The same procedure was used in conjunction with plasmid pTr141 to purify the Ser 210 3 Ala mutant HtrA.
Plasmids encoding GST fusion proteins (see Table I) were used to transform the E. coli strain BL21 DE3 leading to the IPTG-inducible soluble overproduction of the appropriate proteins. Typically the cells from two 500-ml cultures, grown at 37°C and induced in mid-logarithmic growth with 0.2 mg ml Ϫ1 IPTG, were harvested by centrifugation at 2,500 ϫ g and sonicated in 200 ml of 0.1 M Tris-HCl, pH 8.0, 1 mM DTT (buffer 5). Following clarification by centrifugation at 2,500 ϫ g, the supernatant was applied to a glutathione-substituted Sepharose column, and the column washed with 500 ml of buffer 5. Fusion proteins were eluted by a batch wash of 100 ml of buffer 5 containing 10 mM glutathione and dialyzed into 20 mM Tris-HCl, pH 8.0, prior to use in surface plasmon resonance measurements.
Proteins containing a His 6 COOH-terminal tag were purified by immobilized metal affinity chromatography as described above but with the following modifications; harvested cells were sonicated in 50 mM potassium phosphate, pH 7.2, 0.5 M NaCl, 1 mM DTT, 1 mM benzamidine (buffer 6) and the clarified supernatant applied to a 21-ml chelating Sepharose column charged to 30% capacity with zinc. Following a 100-ml wash with buffer 6 containing 2.0 M glycine, the column was equilibrated with 100 ml of buffer 6 to remove residual glycine. The column was then eluted with a 100-ml 50 mM potassium phosphate linear pH 6.0 to 4.0 gradient. Following analysis by SDS-PAGE, fractions containing the desired protein were dialyzed into 20 mM Tris-HCl, pH 8.0.
Protease Assays-Protease assays using HtrA and Tsp with ␤-casein or GST as the substrate were carried out according to the following protocol; 0.1 M protease was added to 3.7 M substrate (␤-casein or GST) in a final volume of 100 l of 20 mM Tris-HCl, pH 8.0. The reaction was incubated at 37°C with Tsp for 2.5 h and at 42°C with HtrA for 4 h, using a Hybaid thermal reactor. Control experiments (data not shown) demonstrated that HtrA had greater proteolytic activity at 42°C than at 37°C. The fragments produced by the proteolysis were analyzed by SDS-PAGE using a 12% separating gel (31). Prior to digestion GST proteins were unfolded by heating at 90°C for 10 min using a Hybaid Thermal Reactor.
Surface Plasmon Resonance (SPR) Measurements-The interactions among HtrA, Tsp, and their PDZ subfragments with three synthetic peptides were monitored by SPR measurements using the BIAcore TM 2000 from Amersham Biosciences. The synthetic peptides were used as the ligand and the HtrA, Tsp, and PDZ subfragments as the analyte. The concentrations of purified proteins and the synthetic peptides were determined spectrophotometrically from their calculated molar extinction co-efficients. For HtrA binding, peptides in the range of 7-30 resonance units (RUs) were immobilized by linkage via a biotin molecule to the streptavidin layer of a SA biosensor chip. Because of the lower affinity of Tsp binding to the peptides, 250 RUs of peptide were bound to facilitate a measurable signal. This immobilization was performed at pH 7.4 in HBS buffer (10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, 0.005% v/v P20 surfactant). GST proteins with and without the wild-type and scrambled SsrA peptide tag were immobilized via amine groups to the surface of CM5 chips at pH 7.4 in HBS buffer in the range 160 -180 RUs. The modified biosensor chips were then equilibrated at 25°C with a running buffer consisting of 20 mM Tris-HCl, pH 8.0. Using the BIAcore KINJECT function, 15 l of the desired analyte was injected at 5 l min Ϫ1 at varying concentrations. Previous experiments using a range of flow rates showed no evidence of mass transport limitation effects. The regeneration buffer, HBS, was injected at 5 l min Ϫ1 for 60 s. For HtrA, a data set of 30 points was obtained by using concentrations of 5, 10, 20, 30, 40, and 50 nM and repeating this for a total of five times. For Tsp, concentrations of 600, 650, 700, 750, and 800 nM were used six times to generate a data set of 30 points. The base lines of the sensorgrams for the experimental and reference flow cells were adjusted to zero immediately prior to injection and the specific changes in the experimental sensorgram measured by subtracting the values from the reference cell containing peptide 2 (the randomized SsrA sequence). Kinetic analysis was performed using BI-Aevaluation 3.0 software (BIAcore AB) following the recommendations of the manufacturer for acceptable 2 values, and fitting the to 1:1 Langmuir model (29). This analysis showed that the analyte failed to give interpretable sensorgrams with peptide 2 (randomized SsrA sequence), indicating that the proteases did not recognize this peptide. Initially the data were fitted to a range of models supplied in the BIAevaluation software package, and the 1:1 Langmuir model adequately described the experimental data for peptides 1 and 3.

RESULTS AND DISCUSSION
The HtrA and Tsp Proteases Can Bind to the ssrA-encoded 11-Amino Acid Sequence-HtrA and Tsp proteases were purified according to the protocol under "Experimental Procedures," and their biological activity confirmed by their ability to digest ␤-casein in vitro (see Fig. 2). To test directly the hypothesis that HtrA and Tsp could bind to the ssrA-encoded peptide tag, we monitored the interactions between the proteins and various peptides related to the SsrA peptide by SPR measurements. The proteases were used as the analytes and three different peptides as the ligands: the wild-type SsrA peptide (COOH-terminal sequence Tyr-Ala-Leu-Ala-Ala), a variant with the COOH-terminal sequence Trp-Val-Ala-Ala-Ala (the sequence originally used to identify and purify Tsp), and a variant consisting of wild-type amino acid composition but randomly assembled sequence. The randomized sequence was used as a control to check that any protease:peptide interactions were sequence related, and the Trp-Val-Ala-Ala-Ala variant was included as a positive control for the Tsp protease. Fig.  1 shows typical sensorgrams for the binding of Tsp (panel A) and E. coli HtrA (panel B) to the wild-type SsrA peptide, and Table II summarizes the kinetic data.
None of the proteins (Tsp, Tsp(Lys 3 Ala), HtrA, and HtrA(Ser 3 Ala)) bound above background values to the randomized peptide, but all four bound to the wild-type and three (Tsp Lys 455 3 Ala was not tested) bound to the Trp-Val-Ala-Ala-Ala variant (Table II). The binding of wild-type and mutant E. coli HtrA proteins to the wild-type and Trp-Val-Ala-Ala-Ala variant peptides is similar with an average apparent equilibrium dissociation constant, K D , of the order of 18 nM. The S. typhimurium HtrA bound to the wild-type SsrA peptide with an apparent dissociation constant, K D , of the order of 3 nM. This is the first demonstration that HtrA binds to the SsrA peptide and that an active proteolytic site in HtrA is not required for SsrA-peptide binding. The value for the apparent K D for the binding of wild-type Tsp to the Trp-Val-Ala-Ala-Ala variant peptides is 3.8 M. The difference of ϳ100-fold in the values of K D for E. coli HtrA and Tsp for the same peptide sequence implies that HtrA binds more tightly than Tsp.
The observation that HtrA and Tsp do not discriminate between the wild-type and Trp-Val-Ala-Ala-Ala variant peptides and do not recognize the randomized peptide shows that, although binding is sequence-related, it is not sequence-specific. A comparison of the SsrA-encoded peptide homologues in a range of organisms shows that the prime requirement for recognition is a pentapeptide apolar COOH terminus (10). The randomized peptide we used as a control has glutamate as the COOH-terminal residue, violating the requirement for an apolar COOH terminus, and likely contributes to the inability of HtrA and Tsp to bind to it.
The PDZ Domains of HtrA and Tsp Bind to the Isolated ssrA-encoded Peptide-To test specifically the hypothesis that PDZ domains of HtrA and Tsp are responsible for recognizing and binding to the SsrA peptide, we purified a series of COOHterminal PDZ-containing fragments (Table I). A deletion analysis of this type has the intrinsic problem that use of the protein fragments may produce atypical kinetic results (in the extreme case, an inability to bind) simply because the protein fragments produced either fail to fold correctly or are inherently unstable. To address this potential problem, a series of overlapping HtrA-PDZ-encoding sequences were cloned and expressed from both E. coli and S. typhimurium based on amino acid sequence alignments. Previous work has shown that some PDZ domains are sufficiently stable to allow crystallization and structural determination (11).
Preliminary SPR experiments showed that the individual E. coli PDZ domains did not bind to the SsrA peptide; however, the S. typhimurium PDZ domains did bind, and were used for the collection of a full data set (see Table II). In each case (for both E. coli and S. typhimurium), PDZ domain 1 was truncated by 4 -9 amino acids at the carboxyl terminus as defined by the recently published x-ray structure (23). Interestingly, the E. coli PDZ domain 2 construct incorporated the full PDZ domain and showed no SsrA binding, whereas the equivalent S. typhimurium sequence was truncated at the carboxyl terminus by 4 amino acids from the full domain and was able to bind the SsrA sequence. The inability of the E. coli PDZ domains to bind to the SsrA peptide may be because the small differences in amino acid sequence compared with S. typhimurium cause the isolated PDZ domains to be unstable. The isolated Tsp PDZ sequence was also stable and bound to the wild-type SsrA peptide. However, caution must be exercised in interpreting the apparent K D values shown in Table II, as the PDZ domains were fused to GST, and it has been argued that the use of GST fusion proteins can overestimate the affinity of the measured interaction (31). This has been proposed to be the result of GST fusion dimers binding simultaneously to two ligand molecules at the sensor chip surface; this divalent attachment is described as an avidity effect (31). However, despite this caveat, HtrA and Tsp can bind to the SsrA-encoded peptide, the biological relevance of this observation was tested through binding to, and proteolysis of, a heterologous protein. We used as substrates both wild-type GST and variants modified by the addition to the COOH terminus of the wild-type and randomized SsrA peptides. The three GST variants were purified and assayed for sensitivity to HtrA and Tsp as described under "Experimental Procedures" and found to be inherently resistant to both proteases. However, Tsp and HtrA are only active on substrates that are significantly unfolded (4); therefore, we repeated the experiments with heat-denatured substrates. In this case, the GST variant modified with the wild-type SsrA peptide became hypersensitive to digestion with Tsp at a concentration of 0.1 M, but the other two variants remained insensitive. These digestions were carried out multiple times, and typical digestion profiles for these experiments are shown in Fig. 2A. Addition of the randomized peptide leads to a marginal increase in sensitivity to Tsp, and this may be the result of nonspecific loosening of the GST tertiary structure exposing susceptible peptide bonds. However, all three GST variants remained insensitive to digestion with HtrA at protease concentrations up to 0.5 M, but showed similar levels of digestion at concentrations in excess of this (data not shown). This is a surprising result, as the SPR experiments with the three variant peptides as ligands implied that HtrA bound with greater affinity. These SPR experiments did not rule out the possibility that this binding to the peptide was biologically irrelevant and that HtrA was unable to recognize the peptide when it was attached to a heterologous protein.
To test this possibility, the SPR experiments were repeated using unmodified (wild-type) GST and the two COOH-terminally modified variants as the ligand. Both proteases bound to the GST variant with COOH-terminal wild-type SsrA peptide. Table II summarizes the kinetic data and shows that HtrA bound to the GST variant with an apparent K D value (37.3 ϫ 10 Ϫ9 M) similar to that for the isolated peptide. Tsp, on the other hand, showed enhanced binding with an apparent K D (2.5 ϫ 10 Ϫ10 ) 10 3 lower than for the isolated peptide.
This observation is consistent with the enhanced sensitivity of the modified GST to proteolysis by Tsp but not HtrA. Tsp has previously been shown to bind to a peptide of sequence Ala-Ala-Arg-Ala-Ala-(fluorescent marker)-Glu-Asn-Tyr-Ala-Leu-Ala-Ala. This sequence is a substantially modified derivative of the SsrA peptide (Ala-Ala-Asn-Asp-Glu-Asn-Tyr-Ala-Leu-Ala-Ala), incorporating an artificial substrate cleavage site (between alanine and arginine) and a fluorescence marker ([N ⑀ - [4-[4-(dimethlyamino)phenylazo][benzoyl]lysyl-(6-aminocaproyl) 2 ) (32). In the same report, the Tsp PDZ domain was shown to bind to the peptide Gly-Arg-Gly-Tyr-Ala-Leu-Ala-Ala, leading to the conclusion the Tsp PDZ domain is responsible for substrate recognition. However, the wild-type SsrA peptide is not itself a substrate for proteolysis; its function is to tag a larger protein or peptide for proteolysis, and this larger protein or peptide is the true substrate. The simplest interpretation of the data we present here is that Tsp binds initially to the modified GST via the PDZ domain interacting with the SsrA peptide and subsequently increases its affinity of binding by a second interaction with the GST protein. To test this interpretation experimentally, we purified the proteolytically inactive Tsp Lys 455 3 Ala mutant protein and used it as the analyte in SPR experiments with the wild-type SsrA peptide and the GST protein modified with the wild-type SsrA peptide. The results summarized in Table II show that the mutant Tsp Lys 455 3 Ala protein bound to the modified GST protein with a K D over 2 orders of magnitude greater than the wild-type Tsp. This observation is consistent with the interpretation that an active protease domain is required to increase the affinity of wild-type Tsp binding through an interaction with the GST protein.
Clues to the reason that HtrA does not show an increased proteolytic activity toward the SsrA-modified GST are provided by its structure. The protease domain forms a chamber the inner cavity of which contains possible binding sites for misfolded proteins. Access to the proteolytic active site is controlled by the action of three mobile loops that interact and must undergo large conformational changes to adopt the classical catalytic model of serine proteases (23). HtrA has apparently recruited PDZ domains to a gating function, and this may allow the coupling of substrate to translocation within the HtrA multimer (23). Consequently it has been speculated that PDZ domains may act as tentacular arms that bind substrates and transfer them to the inner cavity (23). Once in the chamber, the fate of the substrate (proteolysis versus refolding) will depend on the interplay of the three mobile loops that interact to form the classical serine protease conformation. It is possible that the nature of the secondary structure elements translocated into the inner chamber can influence the interaction between these loops in the HtrA protease domain and hence the balance between protease and chaperonin activity for different protein substrates.
The data presented here show that the HtrA PDZ domains can bind to the isolated SsrA peptide and that addition of this peptide to a heterologous protein facilitates binding of the full-length HtrA.