Crystal structure of the protease domain of a heat-shock protein HtrA from Thermotoga maritima.

HtrA (high temperature requirement A), a periplasmic heat-shock protein, functions as a molecular chaperone at low temperatures, and its proteolytic activity is turned on at elevated temperatures. To investigate the mechanism of functional switch to protease, we determined the crystal structure of the NH(2)-terminal protease domain (PD) of HtrA from Thermotoga maritima, which was shown to retain both proteolytic and chaperone-like activities. Three subunits of HtrA PD compose a trimer, and multimerization architecture is similar to that found in the crystal structures of intact HtrA hexamer from Escherichia coli and human HtrA2 trimer. HtrA PD shares the same fold with chymotrypsin-like serine proteases, but it contains an additional lid that blocks access the of substrates to the active site. A corresponding lid found in E. coli HtrA is a long loop that also blocks the active site of another subunit. These results suggest that the activation of the proteolytic function of HtrA at elevated temperatures might occur by a conformational change, which includes the opening of the helical lid to expose the active site and subsequent rearrangement of a catalytic triad and an oxyanion hole.

Protein quality control, which is essential for cell viability, is tightly controlled by proteases and molecular chaperones (1). Especially under stress conditions such as high temperature, heat-shock proteins, which are mostly proteases or molecular chaperones, are induced to protect cells from toxic denatured proteins (2). Molecular chaperones bind to the hydrophobic patches on denatured proteins to prevent further aggregation and help them to fold back into their native states (3). The regulatory subunits of heat-shock proteases (4) recognize the hydrophobic surface on unfolded proteins and eliminate them mostly in an ATP-dependent manner (5).
High temperature requirement A (HtrA, 1 also called DegP or protease Do) is a heat-shock protease localized in the periplas-mic space of bacteria (6). It shows an ATP-independent proteolytic activity and plays an important role in the degradation of misfolded proteins accumulated by heat shock or other stresses (7). Therefore, its activity seems to be essential for bacterial thermotolerance and for cell survival at high temperatures (7). HtrA is also involved in pathogenesis of Gram-negative and Gram-positive bacteria by degrading damaged proteins that are produced by reactive oxygen species released from the host defense system (8). Therefore, HtrA is considered as a target for development of broad-spectrum antibiotics (8).
In addition to proteolytic activity, HtrA is known to have a molecular chaperone activity (9,10). The chaperone function is dominant at low temperatures, whereas the proteolytic activity is turned on at elevated temperatures (9). This temperaturedependent functional switch is necessary for controlling protein stability as well as eliminating denatured proteins to maintain cellular viability (9). HtrA is a highly conserved protein found in species ranging from bacteria to humans. Two known human homologues of bacterial HtrA (HtrA1 and HtrA2) are also expected to be involved in mammalian stress response pathways (11,12). However, because HtrA2 showed proteolytic activity even at room temperature, temperature-dependent activation of proteolytic activity seems to be absent from mammalian HtrAs (13).
HtrA is a serine protease with a catalytic triad in its active site. Recent crystal structure analyses revealed that Escherichia coli HtrA forms a hexameric complex composed of two trimers (14) and human HtrA2 forms a homotrimer (15). Each subunit is composed of one protease domain at the amino terminus and one or two PDZ (named after three proteins, PSD-95, Discs-large, and ZO-1) domains at the carboxyl terminus. The protease domain of E. coli HtrA fully retains the molecular chaperone activity, although the proteolytic activity is absent (9). The PDZ domains, also found in the Clp/Hsp100 family of heat-shock proteins, are known to play a role in substrate recognition (16). In the crystal structures of E. coli HtrA and human HtrA2, PDZ domains are proposed to mediate the initial binding of substrates (14) or to be involved in modulation of protease activity (15). However, PDZ domains do not participate in multimerization in both E. coli and human HtrAs (14,15). Unlike other proteases of the Clp/Hsp100 family, HtrA does not have a regulatory component or an ATP binding domain because it is an ATP-independent heat-shock protease.
So far two crystal structures of HtrAs have been reported (14,15), and they seem to differ in structural architecture, multimerization, and activation mechanism. For a better understanding of the dual role of HtrA and the activation mechanism of the proteolytic function, we have solved the crystal structure of the protease domain (PD, residues 24 -262, Fig. 1) of Thermotoga maritima HtrA (Tm HtrA), which displays both molecular chaperone and proteolytic activities. The crystal structure indicates that the rearrangement of the active site of bacterial HtrA is necessary for the proteolytic activity and that oligomerization architecture of HtrA might vary depending on the presence of the lid covering the active site.

EXPERIMENTAL PROCEDURES
Protein Preparation and Crystallization-The protease domain of HtrA from T. maritima (PD, residues 24 -262, Fig. 1) was cloned, purified, and crystallized as described elsewhere (18). The putative signal sequence (residues 1-23) was deleted in the construct. For the translational start, a methionine residue was added in front of Asp 24 . Intact HtrA was also prepared in the same way as HtrA PD (18). HtrA PD was crystallized in the cubic space group P2 1 3, with the unit cell parameters a ϭ b ϭ c ϭ 120.55 Å by the hanging drop vapor diffusion method at 22°C from a reservoir solution containing 100 mM phosphate-citrate (pH 4.4), 110 mM Li 2 SO 4 , and 5% (v/v) PEG 1000 (18). There are two molecules in an asymmetric unit.
Light-scattering Measurement-The chaperone-like activities of HtrA and HtrA PD were measured as described previously (19) with some modifications using pig heart citrate synthase (CS) as the substrate (Sigma). CS (final 65 M monomer) was denatured in a solution containing 100 mM Tris-HCl (pH 8.0), 100 mM NaCl, 6 M guanidinium hydrochloride, and 40 mM dithiothreitol for at least 2 h at room temperature. Chemically denatured CS was rapidly diluted 250-fold into a refolding buffer containing 10 mM Tris-HCl (pH 8.0) and HtrA or HtrA PD to reach the indicated molar ratios (Fig. 2). Light scatterings from aggregated proteins were monitored by measuring the absorbance at 320 nm with Spectra MAX Plus (Molecular Device) at 25°C. Initial velocities of aggregate formation were calculated by measuring the absorbance change for the initial 15 s where the rate of aggregate formation is constant.
Proteolytic Activity Measurement-Thirty micrograms of the reduced form of ␣-lactalbumin (Sigma) and 28 g of HtrA PD or 48 g of HtrA (at 1:2 molar ratio of protease to substrate) were incubated in 60 l of reaction buffer containing 5 mM Tris (pH 7.5) and 1.5 mM dithiothreitol at 25, 45, 65, and 85°C. The reduced form of ␣-lactalbumin was prepared by incubating 3 mg/ml protein in 5 mM Tris (pH 7.5) containing 10 mM dithiothreitol at 4°C for 2 days. HtrA and HtrA PD were preincubated at each indicated temperature prior to the addition of the substrate protein. The reaction was performed in 100-l tubes to minimize the evaporation effect. After incubation for 30 min at each temperature, 17 l of 5ϫ SDS-PAGE sample buffer was added to stop the reaction. Then, the samples were analyzed by 17% SDS-PAGE.
Structure Determination and Refinement-Multiwavelength anomalous diffraction data were collected from a frozen crystal of HtrA PD at the Pohang Accelerator Laboratory beamline 6B with a MacScience 2030b area detector. Data collected at three wavelengths (edge, peak, and remote) were processed and integrated by DENZO and scaled by SCALEPACK using the HKL program suite (20). Native data of HtrA PD were also collected at the Pohang Accelerator Laboratory and pro-cessed using the HKL program suite (Table I). Two selenium sites (one methionine in one subunit) of PD were found and used for phase calculation in the program SOLVE (21). A relatively low figure of merit (0.25) is explained by the presence of only one selenium atom per 238 residues in the subunit.
Solvent flattening and 2-fold noncrystallographic symmetry (NCS) averaged by RESOLVE (22) resulted in a high quality electron density map sufficient for model building. Amino acids were assigned using the program O (23). Several cycles of rigid body refinement, positional refinement, and simulated annealing were performed at 3.0-Å resolution with CNS (24). The refinements were continued at 2.8-Å resolution using the data collected from the native HtrA PD. Successive refinement with temperature factors and addition of solvents resulted in an R-value of 22.2% and an R free value of 28.4%, with a bulk solvent correction and overall anisotropic thermal factor refinement. R free was calculated with 10% of the reflections. NCS restraints were enforced during the refinement except flexible regions (LA and L2), in which two subunits in an asymmetric unit showed different conformations.
The final model includes residues 24 -48 and 51-251 and 58 water molecules (Table I). Structural evaluation of the refined model using PROCHECK (25) reveals that the structure has good geometric parameters (Table I), and no residue falls in the disallowed region of the Ramachandran plot. Statistical analysis of B-factor distribution was performed by t test and the Wilcoxon rank sum test. p Ͻ 0.001 was considered to be significant. The figures in the article were drawn using the programs MOLSCRIPT (26) and GRASP (27). The final coordinates and structure factors have been deposited in the Protein Data Bank (PDB; accession number 1L1J).

Biochemical Activities of Tm HtrA and Tm HtrA PD-Tm
HtrA and Tm HtrA PD were overexpressed in E. coli and purified. The chaperone-like activities of both Tm HtrA and Tm HtrA PD were measured by their abilities to suppress the aggregation of CS, which has been widely used for molecular chaperone assays (19). The aggregation of CS was monitored by light scattering at 320 nm after chemically denatured CS was diluted in the refolding buffer. Tm HtrA or Tm HtrA PD suppressed CS aggregation, decreasing the initial velocity of aggregate formation (Fig. 2, A and B). By addition of 2-and 4-fold molar excesses of Tm HtrA to CS, the initial velocities of aggregation were decreased to 40.6 and 13.7%, respectively, compared with the initial velocity in the absence of HtrA or HtrA PD. It was decreased to 68.3% when a 4-fold molar excess of Tm HtrA PD was added, whereas 2-fold addition of Tm HtrA PD essentially did not change it (data not shown). An 8-fold excess of either Tm HtrA or Tm HtrA PD was enough to decrease the aggregation rate to about zero. Under the same assay conditions, the initial velocity of the reaction is not de- creased when bovine serum albumin was used for the control protein (Fig. 2, A and B). These results indicate that Tm HtrA PD as well as the intact Tm HtrA have the chaperone-like activity to inhibit the aggregation of CS. However, because the enzyme activity of CS was not recovered (data not shown), it seems that Tm HtrA does not assist the refolding of CS, although the aggregation of denaturated CS was completely suppressed by incubation with Tm HtrA. It can be inferred that Tm HtrA exhibited only chaperone-like activity against CS, as is observed for other heat-shock proteins such as ␣-crystallin (28).
The molar ratios of chaperones to substrates required for suppression of CS aggregation are different between intact HtrA and HtrA PD (Fig. 2), which is also observed for E. coli HtrA and E. coli HtrA PD. In the case of E. coli HtrA, a higher concentration of the protease domain was required for refolding of MalS protein than intact HtrA (9). Such a requirement for higher molar ratios of HtrA PD might be explained by the absence of PDZ domains, which are known to be involved in substrate recognition (14 -16).
␣-Lactalbumin, which is commonly used for the assays of heat-shock proteases such as 20 S proteosome (29) and E. coli HtrA (30), was employed as a substrate to investigate the proteolytic activity of Tm HtrA. Clearly, Tm HtrA displayed the proteolytic activity at elevated temperatures, and maximal activity was observed at 85°C (Fig. 2C). As reported for E. coli HtrA (9, 31), the proteolytic activity of Tm HtrA increased with temperature, and Tm HtrA is autodegraded at high temperatures (Fig. 2C).  q)). The aggregation of CS was monitored by measuring the apparent light scattering (absorbance at 320 nm). Bovine serum albumin was used as a negative control. C, proteolytic activities of Tm HtrA PD and intact Tm HtrA. Proteolytic activity was monitored at each temperature by incubating the reduced form of ␣-lactalbumin (LA) with Tm HtrA PD or intact Tm HtrA. The molar ratio of Tm HtrA (or Tm HtrA PD) to ␣-lactalbumin was adjusted to 1:2. The first and sixth lanes contain molecular weight markers in kilodaltons.
Compared with Tm HtrA, Tm HtrA PD showed a relatively weak proteolytic activity, although its activity also increased with temperature. The degradation products of Tm HtrA PD were long enough to be visible on 17% SDS-PAGE, whereas intact Tm HtrA completely degraded the substrate into short peptides that are not shown in the gel (Fig. 2C). Such a weak proteolytic activity of Tm HtrA can also be explained by the absence of a PDZ domain. In both E. coli HtrA and human HtrA2, PDZ domains play key roles in substrate binding and formation of the chamber near the active site (14,15). Tm HtrA PD generates longer degradation products, because substrates are freely released from Tm HtrA PD after cleavage. In contrast, within the chamber of intact Tm HtrA, substrates are cleaved into small peptides simultaneously at adjacent active sites.
Overall Structure of Tm HtrA PD-The crystal structure of Tm HtrA PD (residues 24 -262) has been solved by multiwavelength anomalous diffraction at 3.0 Å resolution and refined to 2.8-Å resolution ( Table I). The experimental electron density map calculated with multiwavelength anomalous diffraction phases and improved by solvent flattening and NCS averaging was of sufficient quality to locate most main chains and some side chains. Tm HtrA PD is composed of two ␤-barrel domains connected by a long loop between ␤6 and ␤7 (Figs. 3 and 4A). Residues in the catalytic triad, Asp 127 -His 97 -Ser 206 , are located in the cleft of two ␤-barrels. The structural comparison by DALI server (32) reveals that its topology is similar to proteases in the chymotrypsin family (33). Among the members of the chymotrypsin family, ␣-lytic protease (PDB accession number 1qq4; Ref. 34) can be superimposed on Tm HtrA PD with the lowest r.m.s.d. of 2.5 Å for 161 C␣ atoms of 198 C␣ atoms of ␣-lytic protease (Figs. 4B and 5). However, its fold is different from those of proteolytic cores of ATP-dependent proteases such as ClpP (35) and HslV (36).
It is notable, however, that several structural differences exist between Tm HtrA PD and ␣-lytic protease. The most significant of them is the length of LA (loop A connecting ␤1 and ␤2, according to the nomenclature in Ref. 33) in two proteins (Figs. 4 and 5). LA of Tm HtrA (residues 47-76), including an amphipathic helical lid (␣2, residues 55-66) and ␣3, is located on top of the catalytic residue Ser 206 (Figs. 4A and 6A). Interestingly, the residues corresponding to the helical lid of Tm HtrA PD are found only in bacterial HtrAs, not in human homologues (Fig. 1), and this lid is expected to have the impor-  Crystal Structure of the T. maritima HtrA Protease Domain tant functional or structural roles in bacterial HtrAs. There are differences in several other loops connecting ␤-strands (Fig. 5). Among them, structural changes near the loop containing the putative oxyanion hole and the catalytic residue Ser 206 (residues 202-206) seem to be significant for explaining the functional differences of Tm HtrA and ␣-lytic protease (Figs. 5 and 6). Three conserved disulfide bonds found in almost all members of the chymotrypsin family of serine proteases are absent from all known HtrAs (Fig. 4).  7), in which three subunits of Tm HtrA PD related by 3-fold crystallographic rotation symmetry are tightly packed by hydrophobic interactions. A hydrophobic patch composed of the residues near ␣1, ␤7, ␤8, and ␤11 is involved in the hydrophobic packing in a trimer (Figs. 4A and 7A). Those hydrophobic residues are quite well conserved in most other HtrAs (Fig. 1), implying that hydrophobic packing in a trimer is a general feature of HtrAs. By trimerization, the 6700 Å 2 surface area of Tm HtrA PD is buried, which is comparable with the 6044 Å 2 in human HtrA2 (15). Taken together, it appears that the Tm HtrA forms a trimer by hydrophobic interaction mediated by the protease domain, as observed in E. coli HtrA and human HtrA2 (14,15). Analytical ultracentrifugation and gel filtration experiments also support the existence of Tm HtrA PD as a trimer in solution (data not shown).

Tm HtrA Trimer and Structural Comparison with Other
The main differences among HtrAs might be the size and conformation of LA (Figs. 1, 4, and 5). E. coli HtrA has a long loop reaching the active site of the opposite subunit (Figs. 4C,  6C, and 7B). In addition, ␤1 and ␤2 in E. coli HtrA is long enough to make a ␤-sheet with two other ␤-strands from the trimer in the other side, leading to a hexamer structure (Fig.  7B) (14). In contrast, LA in Tm HtrA PD is mainly composed of a helix (␣2) covering the active site of the same subunit in the current structure (Figs. 4A, 6A, and 7A) and is shorter than its counterpart in E. coli HtrA (Figs. 1 and 7). Interestingly, human HtrA2 has a very short LA, which is not involved in the dimerization of trimers or the covering of the active site (Figs.  4D, 6D, and 7C).
Two molecules of HtrA PD in an asymmetric unit related by 2-fold NCS are associated by the minimal hydrophobic interactions among a few residues (Tyr 25 , Pro 28 , Val 32 , and Ala 35 ; figure not shown) in the NH 2 -terminal helix (␣1). Therefore, the presence of two molecules in the asymmetric unit seems to have no biological relevance. The two molecules in the asymmetric unit show identical conformations except the regions near LA and L2 (loop 2 connecting ␤11 and ␤12), suggesting those regions are relatively flexible.
Hydrophobic Patches on the Surface of HtrA PD-Most molecular chaperones have hydrophobic substrate binding sites on their surfaces to recognize and bind to the exposed hydrophobic patches of substrates (3). However, in Tm HtrA PD trimer, most of the hydrophobic surface near the active site is buried and no noticeable hydrophobic region is exposed (Fig. 7D). Therefore, certain conformational changes might occur to expose the hydrophobic substrate binding site when Tm HtrA or Tm HtrA PD shows the chaperone-like activity.
Proteolytic Active Site of HtrA PD-Most hydrophobic residues in the helical lid of LA form wide contacts by hydrophobic interactions with Leu 80 in ␤2, Pro 163 and Leu 164 in LD (loop D connecting ␤7 and ␤8), Pro 203 and Gly 204 in L1 (loop 1 connecting ␤9 and ␤10), and Ala 223 and Ile 224 in L2 (Fig. 6A). Because LA* of E. coli HtrA makes intimate contact with L1 and L2 (the asterisk denotes the loops in the neighboring subunit, see Figs. 4C, 6C, and 7B; Ref. 14), the hydrophobic interactions between the loops near the active site and lid seem to be common in bacterial HtrAs. However, in other HtrAs residues interacting with LA appear to vary depending on the size of the lid (Fig. 1). Possible substrate binding sites of Tm HtrA (S3, S2, S1, S1Ј, S2Ј, and S3Ј defined in Ref. 33) are completely blocked by the lid and inaccessible to the solvent (Fig. 6A). Because the average temperature factor of the residues in LA is relatively higher (64.9 Å 2 ) than that for the whole protein (54.8 Å 2 ), it appears that the lid is rather flexible and its interaction with the loops are not tight. B-factor difference between LA and other regions is significant (p Ͻ 0.001). Flexibility of the lid is also inferred from different conformations of the lids of two molecules in an asymmetric unit. When the molecules in the asymmetric unit are superposed, 28 C␣ atoms in LA (residues 47, 48, and 51-76) give a r.m.s.d. of 1.52 Å, whereas other C␣ atoms (except another flexible region at residues 225-232) give 0.45 Å (Fig. 5). We also suspect that residues 49 and 50 do not show electron density because of the flexibility of LA. Structural flexibility of LA found both in Tm and E. coli HtrAs (14) suggests that this loop could undergo a conformational change in bacterial HtrAs.
In addition to the fact that the active site is blocked by the helical lid, several residues that are essential for proteolytic activity are positioned differently from those in the ␣-lytic protease (Fig. 6, A and B). For hydrolytic cleavage of a peptide bond, the residues in the catalytic triad need to be aligned close enough for electron transfer from Asp to Ser through His. However, in the current crystal structure of Tm HtrA PD, distances between the N ⑀2 atom of His 97 and the O ␥ atom of Ser 206 of each molecule in the asymmetric unit are 3.69 and 3.40 Å (Fig. 6A, Table II), whereas the N ⑀2 atom of His 36 and the O ␥ atom of Ser 143 in the ␣-lytic protease are hydrogenbonded at a distance of 2.96 Å (Fig. 6B, Table II). Consequently, His 97 is not expected to function as a general base to remove the proton from Ser 206 . In the crystal structures of E. coli and human HtrAs, the distance between the N ⑀2 atom of His and the O ␥ atom of Ser in the catalytic triad cannot be measured because Ser mutated to Ala (Fig. 6, C and D) (14,15). However, the distance between the N ⑀2 atom of His and the C ␤ atom of the mutated Ala in catalytic triads of E. coli HtrAs is too long compared with the corresponding distance in the ␣-lytic protease (Fig. 6, Table II), although a little conformational change in the Ala mutant is expected. Therefore, it is obvious that both Tm and E. coli HtrAs are inactive because of the distortion of the loops near the active site, whereas ␣-lytic protease is in an active state.
Another crucial factor for the proteolysis by chymotrypsinlike proteases is the stabilization of a negative charge of carbonyl oxygen on the reaction intermediate (oxyanion hole) by hydrogen bonds from the amide nitrogen atoms of two peptide bonds in the backbone. In Tm HtrA PD, nitrogen atoms of Ser 206 and Gly 204 are assumed to be the hydrogen donors to the putative oxyanion hole. However, the NH group of Gly 204 in Tm HtrA PD and its counterpart in the ␣-lytic protease (the NH group of Gly 141 ) point to opposite directions (Fig. 6, A and B). In this conformation it is impossible for Tm HtrA PD to form an oxyanion hole. This unique conformation of the loop near the oxyanion hole is also found in two other HtrAs (Fig. 6, C and D). Similarly, triacylglycerol lipase and Staphylococcus aureus epidermolytic toxin A do not have pre-formed oxyanion holes (37,38), leading to inactive states, as seen in HtrA. DISCUSSION Considering the helical lid, catalytic triad, and oxyanion hole in the crystal structure of Tm HtrA PD, it can be referred that the current crystal structure determined at room temperature represents an inactive conformation of Tm HtrA PD (Figs. 4A and 6A). This is also true in E. coli HtrA, in which the distortion of the active site loops and the intervening LA* are assumed to prevent the proper position of the catalytic triad and the formation of an oxyanion hole, resulting in an inactive conformation of the protease domain (Figs. 4C and 6C). However, because Tm HtrA PD shows protease activity at high temperatures (Fig. 2C), a structural change should occur to activate its proteolytic activity. LA appears to be flexible because of the high temperature factors in the crystal structure and different conformations in two molecules in the asymmetric unit; therefore, it is tempting to propose that LA becomes more flexible and flips up at elevated temperatures. By these plausible conformational changes, Tm HtrA becomes ready for proteolysis by exposing the substrate binding site and rearranging the residues in the active site. A similar temperaturedependent activation mechanism involving conformational changes was suggested based on the biochemical properties of E. coli HtrA (9). In addition, a conformational change of the loop covering the catalytic site of HtrA is also expected from the crystal structure of E. coli HtrA (14). In support of our assumptions, an infrared spectroscopic study (31) and 1-anilino-8-naphthalenesulfonate binding experiment (30) of E. coli HtrA suggest that a conformational change and exposure of the hydrophobic region could occur when HtrA is activated. In human HtrA, which does not have the helical lid ( Figs. 1 and 6D), proteolytic activity is evident even at room temperatures because of the exposed active site (13,15). However, human HtrA might also experience some conformational changes near the active site by binding to the substrate, because its active site also seems to be imperfectly formed in the crystal structure (Fig. 6D) (15). The proposed activation mechanism of HtrA is reminiscent of that found in lipase, whose hydrolytic action is achieved by activation at an oil-water interface (37). Although triggered by different factors, both lipase and HtrA may undergo quite similar conformational changes: the opening of the lid and rearrangement of the active site.
In the crystal structures of E. coli and human HtrAs, the protease domains exclusively are involved in oligomerization (14,15). Therefore, we expect that the oligomerization interaction only occurs in the protease domain of Tm HtrA (Fig. 7A). Structural differences among the three HtrAs mainly originate from the size and conformation of LA (or LA* in E. coli HtrA). This loop participates in the dimerization of HtrA trimers in E. coli (Fig. 7B). But, LA forms a helix and covers the active site of the same subunit in Tm HtrA PD (Fig. 7A) and is very short in human HtrA2 (Fig. 7C). Consequently, E. coli HtrA forms a hexamer, whereas Tm HtrA PD and human HtrA2 are found as trimers (14,15). Therefore, trimer formation of HtrA is mediated by hydrophobic residues in its protease domain and further multimerization occurs by way of LA and neighboring ␤-strands, depending on their size and structure. Furthermore, the proteolytic activity of HtrA may be modulated by the structure of LA and the following multimerization state.
Current crystal structure cannot provide the structural basis for the chaperone activity of Tm HtrA PD, because no remarkable substrate binding sites are found on the surface of Tm HtrA PD (Fig. 7D). To display the chaperone-like activity seen under the experimental conditions ( Fig. 2A), we propose that a hydrophobic groove necessary for chaperone function is exposed by certain conformational changes. The most important finding in the crystal structure of the protease domain of Tm HtrA is the presence of the helical lid covering the active site; this lid is expected to open at high temperatures for proteolytic action. However, we cannot rule out the possibility that LA may not have the same conformation in intact HtrA as the current crystal structure of protease domain, in which LA extends to the opposite subunit instead of covering its own active site and contributes to the formation of hexamer in solution. Similarly, in E. coli HtrA, LA* blocks the active site of the subunit in the opposite trimer (Figs. 6C and 7B) (14). Considering the common hydrophobic patches composed of the loops near the active site in bacterial HtrAs, however, the active site of Tm HtrA could be blocked by either LA in the same subunit or LA* in the other subunit, and the active site is moved away for proteolytic activity. Therefore, a tem-perature-dependent activation mechanism might be a general feature of bacterial HtrAs. To explore this possible conformational change of LA and to confirm the proposed activation, the crystal structure of intact Tm HtrA and its biochemical characterization will be required.