A new native EcHsp31 structure suggests a key role of structural flexibility for chaperone function

Abstract

Heat shock proteins and proteases play a crucial role in cell survival under conditions of environmental stress. The heat shock protein Hsp31, produced by gene hchA at elevated temperatures in Escherichia coli, is a homodimeric protein consisting of a large A domain and a smaller P domain connected by a linker. Two catalytic triads are present per dimer, with the Cys and His contributed by the A domain and an Asp by the P domain. A new crystal Form II confirms the dimer and catalytic triad arrangement seen in the earlier crystal Form I. In addition, several loops exhibit increased flexibility compared to the previous Hsp31 dimer structure. In particular, loops D2 and D3 are intriguing because their mobility leads to the exposure of a sizable hydrophobic patch made up by surface areas of both subunits near the dimer interface. The residues creating this hydrophobic surface are completely conserved in the Hsp31 family. At the same time, access to the catalytic triad is increased. These observations lead to the hypothesis for the functioning of Hsp31 wherein loops D2 and D3 play a key role: first, at elevated temperatures, by becoming mobile and uncovering a large hydrophobic area that helps in binding to client proteins, and second, by removing the client protein from the hydrophobic patch when the temperature decreases and the loops adopt their low-temperature positions at the Hsp31 surface. The proposed mode of action of flexible loops in the functioning of Hsp31 may be a general principle employed by other chaperones.

Molecular chaperones are conserved and ubiquitous proteins that play an important role in the conformational quality control of the proteome by mediating a wide range of processes including protein folding, targeting to membranes, disaggregation, and degradation (for recent reviews, see Gottesman et al. 1997; Wickner et al. 1999; Ben-Zvi and Goloubinoff 2001; Houry 2001; Lund 2001; Dougan et al. 2002; Hartl and Hayer-Hartl 2002). Many chaperones are heat shock proteins (Hsps) that are induced when cells are exposed to a variety of stresses, including temperature upshift. They generally perform their function by interacting with hydrophobic segments exposed to the solvent by nonnative proteins and either stabilize folding intermediates until stress has abated or promote their active remodeling or refolding via cycles of binding and release. This protein rescue system is complemented by a disposal system embodied by the heat shock proteases and responsible for the degradation of irreversibly misfolded proteins (Dougan et al. 2002).

A recent Escherichia coli transcriptome analysis identified a number of heat inducible loci without known function (Richmond et al. 1999). One of these genes, hchA (formerly yedU), encodes a 31-kD protein (Hsp31) that is conserved in a number of pathogenic eubacteria (Sastry et al. 2002). EcHsp31 has relatives in all prokaryotic and eukaryotic kingdoms (Quigley et al. 2003), including the ThiJ family of bacterial enzymes, the PfPI family of putative intracellular proteases (Halio et al. 1996; Kim et al. 1998; Du et al. 2000) and human DJ-1, a protein involved in Parkinson disease (Tao and Tong 2003; Wilson et al. 2003). Dimeric EcHsp31 exhibits chaperone activity and has been shown to interact with early unfolding intermediates and to release properly folded proteins once thermal stress has abated (Sastry et al. 2002; Malki et al. 2003). The 1.6 Å structure of the Se-Met EcHsp31 has revealed that the homodimer contains a system of hydrophobic patches, canyons, and grooves, which may serve to stabilize partially unfolded protein substrates (Quigley et al. 2003). Each subunit consists of two domains termed A and P that are connected by a linker region. Whereas the A domain shares an α-β sandwich with the DJ-1 family, which contains 286 proteins (Bateman et al. 2002), the P domain exhibits no homology with other known folds (Quigley et al. 2003). The P domain is responsible for dimer formation through extensive intersubunit interactions with the A′ and P′ domains of the neighboring subunit. The P domain interface also creates a large concave surface, known as the “hydrophobic bowl” measuring ~20 Å in diameter. The flexible extended linker connecting the A and P domains is formed by residues Pro 32 to Tyr 45 and has been proposed to play a role in conformational flexibility and substrate binding. Finally, the presence of a well-conserved, yet quite buried, Cys-His-Asp triad in each subunit suggests that EcHsp31 may also have a hydrolytic function (Quigley et al. 2003). Initial testing of EcHsp31 for protease and peptidase activity has yielded negative results (Malki et al. 2003).

Three classes of Hsp31 orthologs have been identified by structure-guided sequence alignments. Family members are distinguished by differences in three features: the presence of the P domain, the architecture of the putative catalytic triad, and the quaternary structure (Quigley et al. 2003). EcHsp31 is representative of Class I homologs, which have a complete P domain, form dimers, and use an intrasubunit aspartate to complete their putative catalytic triad. Class II Hsp31 homologs have an incomplete P domain, use glutamate as the third member of the catalytic triad, and have an unknown oligomerization state. Finally, Class III homologs, represented by PhPI and PfPI (Halio et al. 1996; Du et al. 2000) completely lack the P domain, use an intersubunit glutamate to complete their catalytic triad, and are hexameric.

Although the recent structure of C-terminal His-tagged EcHsp31, obtained from monoclinic Form I crystals containing one molecule per asymmetric unit, has revealed the overall architecture of this class of dimeric chaperones (Quigley et al. 2003), little structure–function relationship information is currently available. Here we present the 2.7 Å structure of EcHsp31 in Form II crystals containing four dimers in a triclinic cell. Disorder in a number of key regions, and specifically in the intradomain linker, suggests a key role for flexibility in the mechanism of chaperone action.

Results

Structure of the Form II EcHsp31 monomer

The large P1 cell containing eight EcHsp31 subunits provided us with an opportunity to study multiple conformations of the same protein within a single crystal. Superposition of the 238 Cα atoms from subunits A through H resulted in root mean square deviations between pairs of subunits of less than 1 Å. Despite the high degree of similarity between protomers, each subunit exhibited several disordered regions that could be grouped into five regions of flexibility termed D1 through D5 (Fig. 1 ▶): (1) Region D1 contains N-terminal residues Met 1 through Ser 6; (2) Region D2 encompasses residues Gln 27 to His 49 and includes the linker that connects the A and P domains of each subunit (Quigley et al. 2003); (3) Region D3 (residues Met 109 to Lys 115) corresponds to the loop connecting A domain helices α3 and α4 in the original structure; (4) Region D4 consists of residues Leu 140 to Ser 144 which link α5 to β8 in the A domain; and, finally, (5) region D5 encompasses the C-terminal residues Tyr 281 to Gly 283. Because the N- and C-terminal residues often vary in their visibility from subunit to subunit (Smith et al. 1989; Kumar et al. 2002), disorder in regions D1 and D5 was not unexpected and these regions are therefore not given further attention. Comparison of B factors for all Cα atoms of the eight Form II subunits and the Form I subunit (Fig. 2 ▶) shows that all disordered regions, and particularly D2 and D3, have consistently higher B factors than the remainder of the polypeptide chain in all structures.

Figure 1.

Disorder in EcHsp31 monomer. The P domain is green, the A domain is blue, and regions of disorder are labeled D1–D5. (Top) Stereodiagram of EcHsp31 with disordered regions of Form II shown in red and gray (D1–D5). (Bottom) Regions of disorder (red circles) for Form I and II EcHsp31 superimposed onto sequence and secondary structure. A and P domain boundaries are shown over the sequence.

Figure 2.

B-factor plots of Cα atoms Form II (subunits A–H) and Form I EcHsp31.

Quaternary structure of Form II EcHsp31

The four independent copies of the dimer were superimposed onto one another using only the 238 Cα atoms of one subunit to calculate the rotational and translational parameters. This allows for the evaluation of differences in subunit–subunit arrangements among the dimers. The resulting root mean square deviation over all 476 Cα atoms for each pair of dimers was less than 1 Å, the same deviation as obtained for pairwise superposition of single subunits. This exercise revealed no apparent relative motion of subunits within the EcHsp31 dimers. Biochemical evidence has shown that EcHsp31 forms a stable dimer in solution and binds as a dimer to unfolded substrate proteins (Sastry et al. 2002; Malki et al. 2003). Previously, we have proposed that the dimer is a biologically relevant form of EcHsp31 for the following reasons (Quigley et al. 2003). First, the 12 residues that make up the dimer interface are conserved within the entire Class I of high similarity EcHsp31 homologs. Second, these residues contribute the most extensive dimer interaction in the Form I crystals, burying over 1350 Ų of surface area compared to 800 Ų for the next largest crystal contact. Because the same dimers are observed in the Form II EcHsp31 crystal, the relevance of this quaternary structure is further confirmed.

However, the superposition revealed that areas of disorder cluster on the face of the dimer that includes the hydrophobic bowl. The hydrophobic bowl is made up of 14 amino acid side chains provided by seven residues from each subunit (Val 11, Ile 13, Ala 14, Phe 19, Phe 20, Leu 26, and Tyr 106) that create a concave surface measuring ~20 Å in diameter (Quigley et al. 2003). Four of the above residues (Ala 14, Phe 19, Phe 20, and Leu 26) are completely conserved in Class I Hsp31 homologs, whereas the remaining residues, Val 11, Ile 13, and Tyr 106, are replaced by other hydrophobic amino acids. When the flexible regions are mobile as in Form II, they uncover more hydrophobic surface area adjacent to the conserved hydrophobic bowl. Nineteen residues (Pro 75, Ile 76, Leu 79, Leu 80, Tyr 83, His 84, Leu 85, Ala 88, Phe 90, Phe 104, Trp 107, Ala 108, Phe 120, Ile 220, Tyr 222, Pro 263, Ala 266, Leu 272, and Tyr 281) contribute to the architecture of the uncovered hydrophobic patches. This unmasking of residues increases the total percentage of solvent-exposed nonpolar surface area from 55.1% to 58.4% for each monomer (Fig. 3 ▶).

Figure 3.

Comparison of Form I EcHsp31 (top) and Form II EcHsp31 dimer (bottom). Molecular surface of dimers shown with hydrophobic patches in green. Ribbon representation of dimers are shown on the right in the same orientation. Red ribbons in Form I EcHsp31 highlight regions that are disordered in the Form II EcHsp31. Note that the differences (red) between Form I EcHsp31 and Form II EcHsp31 dimer result in the exposure of more hydrophobic regions (green patches) in Form II EcHsp31. These areas are circled in black.

Differences between triclinic Form II and monoclinic Form I

A difference between Form I and Form II EcHsp31 crystals is that a C-terminal His-tagged version of Hsp31 was used to obtain the former (Quigley et al. 2003) whereas the native protein was employed in the present study. The second difference is that Form I crystals were grown at 14°C from a 14 mg/mL protein solution in a reservoir containing 25% PEG 3350, 300 mM MgCl₂, 50 mM MES (pH 6.5) and 4 mM DTT, whereas Form II crystals were grown at 22°C from a 2 mg/mL protein solution with a reservoir solution of 26% PEG 6000, 50 mM NaCl, and 100 mM Tris HCl (pH 7.5). The V_M of crystal Form I is 2.2 ų per dalton and that of crystal Form II, 2.4 ų per dalton. This is not a very significant difference. We suspect that the combination of a lower temperature, a different pH and ionic strength, and possibly other factors were necessary to reduce protein flexibility and allow acquisition of the high-resolution Form I structure. Because native protein and more physiological conditions were used to grow the triclinic crystal, the regions of flexibility observed in the Form II structure are likely to be biologically relevant and may reflect the difference between active and inactive conformations of the EcHsp31 dimer.

Discussion

Loop flexibility of EcHsp31 exposes hydrophobic surfaces

The five flexible regions of EcHsp31 identified in the Form II crystal structure provide profound insight into the functioning of this heat shock protein. In particular, the most prominent flexible loops, D2 and D3, will be analyzed here in detail. Region D2 encompasses the entire linker (residues 32–45) and in most subunits extends within the A and P domains (Fig. 1 ▶). Based on its lack of secondary structure and high B factors in Form I crystals, we have previously suggested that the linker is flexible and may shift position in a temperature-dependent fashion (Quigley et al. 2003). The present study provides strong support for this hypothesis. One of the most prominent effects of flexibility in the D2 region is the exposure of normally buried residues that form a hydrophobic patch along the outside of the monomer and adjacent to the hydrophobic bowl. Of the 19 residues that make up the patch, 13 residues are completely conserved in Class I Hsp31 orthologs, whereas the remaining 6 are conservatively replaced only by other hydrophobic amino acids, supporting a functional role. Overall, these data are fully consistent with circular dichroism and bis-ANS labeling experiments showing that, as the temperature increases, EcHsp31 progressively exposes structured hydrophobic patches to the solvent without losing secondary structure (Sastry et al. 2002).

D2 flexibility has a second important consequence. In Form I crystals, the Cys 185/His 186/Asp 214 triad resides at the bottom of a deep and restricted pocket and Tyr 29 plays a major role in obscuring it (Fig. 4 ▶; Quigley et al. 2003). Because the D2 region spans as far as Gln 28 in the P domain and His 49 in the A domain of certain subunits, movement of the flexible part of helix α1 that contains Tyr 29 greatly increases the solvent accessibility of the putative catalytic triad (Fig. 4 ▶).

Figure 4.

Molecular surface of Form II EcHsp31 with the linker region from Form I EcHsp31 superimposed to show how Tyr 29 limits access to the active site pocket. The surface of Form II EcHsp31 is colored by domain—A (blue), and P (green), and Cys 185 (yellow). Regions D2 and D3 are modeled in red and orange ball and stick representation from their relative position Form I EcHsp31 structure. Close-up view of the access to Cys 185 of the triad is shown for both Form I and Form II.

The second flexible area of interest is the D3 region that contains the loop connecting helices α3 and α4. D3 is located in close proximity to D2 and, in Form I EcHsp31, packs behind the linker in an interaction stabilized by four hydrogen bonds. Specifically, in Form I, the backbone amides of Lys 112 and Asp 113 from D3 hydrogen bond to the backbone carbonyl of Pro 32 from D2, whereas the OD2 of Asp 113 hydrogen bonds to backbone amide and the hydroxyl moiety of Ser 24. In half of the Form II subunits, the D2 stretch of residues is disordered (Fig. 1 ▶), suggesting that movement of D2 and D3 are coupled in such a way that a positional change in the linker induces flexibility in D3 or pushes it out of the way. Interestingly, the D3 region is mostly α-helical in EcHsp31, whereas β-sheets dominate in the Class III homolog PhPI (Quigley et al. 2003). This lack of structural homology may indicate that D3 plays a unique role in the activity of EcHsp31 and other Class I Hsp31 homologs.

Chaperone activity of EcHsp31 dimer in light of flexibility

Dimeric EcHsp31 has been assigned a chaperone function based on its ability to suppress the aggregation and increase the recovery yields of model proteins that have been unfolded by thermal or chemical treatment, and to form a stable binary complex with reduced carboxymethylated α-lactalbumin (R-CMLA; Sastry et al. 2002^; Malki et al. 2003). Typically, molecular chaperones bind nonpolar regions exposed to the solvent by partially folded client proteins via hydrophobic–hydrophobic and ionic interactions (Hartl and Hayer-Hartl 2002). Although much of the EcHsp31 surface is electrostatically uncharged, it contains a large concave hydrophobic bowl that was proposed to serve as a binding site for nonnative folding intermediates (Quigley et al. 2003). When viewed down the twofold axis of the dimer, the D2 and D3 regions of flexibility are located on either side of the bowl (circled in Form II EcHsp31 of Fig. 3 ▶). As a result, the nonpolar residues that become unmasked upon D2/D3 movement are adjacent to the bowl and contribute to the formation of a large hydrophobic surface encompassing an entire side of the protein. It is tempting to hypothesize that this feature corresponds to a high-affinity binding site for nonnative proteins that becomes fully accessible when an increase in temperature induces linker movement. Interestingly, the enlarged bowl of EcHsp31 does not contain deep hydrophobic grooves that in other bacterial chaperones, such as DnaK (Zhu et al. 1996) and SecB (Xu et al. 2000), are involved in the tight capture of solvent-exposed hydrophobic segments from relatively unstructured substrates. This may explain the preference of EcHsp31 for early unfolding intermediates (Sastry et al. 2002) that retain most of their structure but expose an improperly packed and fluctuating hydrophobic core to the solvent.

Flexibility and the putative triad

All members of the Hsp31 family contain an absolutely conserved catalytic triad, that is intrasubunit in Class I and II homologs and intersubunit in Class III orthologs (Du et al. 2000; Quigley et al. 2003). Although the putative triad is “technically” solvent accessible in Form I EcHsp31, it resides at the bottom of a deep and fairly restricted pocket (Quigley et al. 2003). Comparison of Form I and II EcHsp31 structures shows that when the D2 linker region is flexible, the triad becomes significantly more solvent accessible (Fig. 4 ▶). If disorder in the crystal truly represents flexibility in the protein under physiological conditions, the D2 region may sample a distribution of conformations in solution, several of which considerably open up the active site.

The catalytic triad of many proteases is adjacent to a large substrate-binding groove (Branden and Tooze 1991; Ay et al. 2003) that is not present in our structures. Using purified native EcHsp31, we have been unable to detect proteolytic activity against gelatin, casein, azocasein, and several fluorogenic or chromogenic substrates at 37°C or 45°C in the presence or absence of ATP, Mg²⁺, Mn²⁺, Co²⁺, Zn²⁺, Fe²⁺, K⁺, or various combinations of these additives (M. Sastry and F. Baneyx, unpubl.). Malki et al. (2003) have used nine additional substrates in the presence or absence of ATP and Mg²⁺ to determine if Hsp31 exhibited endo-, amino-, or carboxy-peptidase activity at 24°C. All experiments were negative. In light of the facts that proteins rarely conserve a feature without a need for function and that the Class III homologs PfpI and PhpI exhibit peptidase activity (Halio et al. 1996; Du et al. 2000), the observed lack of protease activity is puzzling. It remains possible that EcHsp31 specifically cleaves a yet-to-be-discovered substrate, that the triad performs another hydrolytic function, or that its presence is unrelated to hydrolysis such as the binding of a specific ligand under certain conditions.

A possible model for EcHsp31 chaperone function

Taking both Form I and Form II structures as representative snapshots of the dynamic process of chaperoning, we propose the following mechanism for EcHsp31 function (Fig. 5 ▶). At low temperatures, represented by Form I, only the hydrophobic bowl is available to interact with folding intermediates. In this conformation, the catalytic triad is essentially inaccessible for peptide or protein substrate, and EcHsp31 exhibits low affinity for partially folded protein substrates. However, it is still able to interact with solvent-exposed hydrophobic segments as evidenced by the fact that the presence of the chaperone increases the recovery yields of chemically unfolded model substrates at room temperature (Sastry et al. 2002; Malki et al. 2003). Under heat shock conditions, represented by the Form II structure, temperature-induced movement of the D2/D3 flexible regions exposes two additional hydrophobic patches on each side of the bowl, leading to the formation of a “high”-affinity binding site for early unfolding intermediates. The latter are stabilized by this interaction for the duration of the high temperature step, although they may be in dynamic equilibrium between chaperone-bound and free forms if the linkers sample different conformations. This would explain why stable binary complexes between EcHsp31 and certain model substrates cannot be isolated at high temperatures. As heat shock subsides and the temperature returns to physiological values, reordering of the flexible regions leads to the ejection of the client protein, which can now refold under permissive conditions (Fig. 5 ▶). Because the triad is more solvent accessible at high temperatures, it is also possible that under appropriate conditions (for example, prolonged stress or cofactor binding) and with specific, though still unidentified substrate(s), cleavage of the partially folded protein takes place (Fig. 5 ▶). Hence, our new native structure in crystal Form II has revealed a key role of flexibility for the chaperone function of EcHsp31, and such temperature modulation of flexible loops might be a more general mechanism by which certain heat shock proteins perform their function in a variety of living cells.

Figure 5.

Possible mechanism of chaperone activity of EcHsp31 and possible hydrolytic functions at high temperatures.

Materials and methods

Expression and purification

The full-length EcHsp31 was expressed and purified as previously described (Sastry et al. 2002) using a Q-Sepharose Fast Flow (Amersham Biosciences) column followed by a Macro-prep hydroxyapatite Type 1 (Bio-Rad) column. The protein was concentrated and further purified by size exclusion chromatography and eluted from a Superose 12 (Amersham Biosciences) column as a single peak (Sastry et al. 2002). The protein was concentrated to 2 mg/mL in a buffer of 25 mM Tris-HCl (pH 7.4), 4 mM DTT, and 1 mM EDTA.

Crystallization

Form II EcHsp31 crystals were grown by vapor diffusion at room temperature. The drops, containing 2 μL of protein solution (2 mg/mL 50 mM Tris-HCl at pH 7.4, 100 mM NaCl, 1 mM EDTA, and 4 mM DTT) and 2 μL of the reservoir solution (26% PEG 6000, 50 mM NaCl, 100 mM Tris at pH 7.5) yielded crystals within 1 wk. After exposure to xenon gas in a pressured chamber (Vitali et al. 1991), the crystal was flash frozen in a solution of artificial mother liquor supplemented with 15% ethylene glycol as the cryo-protectant. The crystals belong to space group P1 with cell dimensions of a = 54.5 Å, b = 99.0 Å, c = 116.8 Å, α = 102.9°, β = 101.5°, γ = 94.2°. They contain eight molecules per asymmetric unit and have a solvent content of 54% and a corresponding V_M of 2.4 ų/Da (Matthews 1968).

Structure determination

The triclinic crystal form of EcHsp31 was obtained prior to the monoclinic crystals that eventually allowed the structure to be elucidated (Quigley et al. 2003). Because the triclinic form contains sulfur methionines, a number of methods were used to obtain useful phasing information. Heavy metal soaks with mercury and gold compounds resulted in complete loss of diffraction. However, when crystals were treated in a pressure cell with xenon gas in an attempt to incorporate Xe, diffraction was still observed. Xenon has been shown to bind nonspecifically to hydrophobic patches in proteins without significantly altering crystal packing and has been useful in gaining phase information with soft X rays (Vitali et al. 1991; Cianci et al. 2001; Wiess et al. 2001). A data set for the xenon-treated crystal was collected at 100 K on beamline 8.2.1 at the Advanced Light Source at the relatively long wavelength of 1.77 Å to optimize the xenon anomalous signal, but no xenon signal was detected. Although 360° of data could be collected on this crystal, the last 180° had to be truncated due to severe radiation damage, probably resulting from the absorption effect of soft X rays. The data set was processed and scaled with HKL2000 (Otwinowski and Minor 1997). Attempts to obtain phases by the S-SAD procedure were not successful.

In the absence of useful phase information from heavy metal or anomalous scatterers, the recent multiwavelength anomalous dispersion (MAD) EcHsp31 structure (Quigley et al. 2003), Form I, allowed eventually for the solution of the triclinic P1 cell using molecular replacement with AMORE (Navaza 1994). Initial attempts with the monomer as search model failed to provide a clear solution to the rotation function. However, when the EcHsp31 dimer was used as search model, four clear placements within the unit cell were obtained with AMORE. Subsequently, the eight monomers were refined as rigid bodies to an R value of 37.4% and a correlation coefficient of 76.6%. Inspection of the model and initial maps at that stage of refinement indicated that certain regions of density were missing for various parts of the different subunits. Because the missing regions varied from subunit to subunit, no density averaging was used during refinement. Solvent flattening was performed in RESOLVE version 1.05, which improved the maps considerably (Terwilliger 2000). Alternate rounds of refinement and model building were performed with REFMAC5 (Murshudov et al. 1997) and XTALVIEW (McRee 1999) until the final model was obtained with an R_work and R_free of 22.2 and 28.7%, respectively (Table 1).

Table 1.

Structure determination statistics

Data collection
Wavelength (Å)	1.77
Resolution (Å)	2.7
Unique reflections	60105
Redundancy	1.9
Completeness (%)	96.1 (94.1)a
Rsym (%)	7.5 (33.9)a
Average I/σ	9.4 (2.0)a
Refinement
Resolution (Å)	20–2.7
Rcryst (%)	22.2%
Rfree (%)	28.7%
Rmsd bonds (Å)	0.004
Rmsd angles (°)	0.793
Residues in Ramachandran plot (%)
Most favored	88.3
Additional	11.0
Disallowed	0.7
Number of residues	2151
Average B factor of protein (Å2)	45.5
Number of waters	72

^a Values in parentheses correspond to the outer shell.

Data deposition

The coordinates have been deposited in the Protein Database (www.rsbc.rutgers.edu) with the PDBID code of 1PV2.