Newly folded substrates inside the molecular cage of the HtrA chaperone DegQ.

The HtrA protein family combines chaperone and protease activities and is essential for protein quality control in many organisms. Whereas the mechanisms underlying the proteolytic function of HtrA proteins are well characterized, their chaperone activity remains poorly understood. Here we describe cryo-EM structures of Escherichia coli DegQ in its 12- and 24-mer states in complex with model substrates, providing a structural model of HtrA chaperone action. Up to six lysozyme substrates bind inside the DegQ 12-mer cage and are visualized in a close-to-native state. An asymmetric reconstruction reveals the binding of a well-ordered lysozyme to four DegQ protomers. DegQ PDZ domains are located adjacent to substrate density and their presence is required for chaperone activity. The substrate-interacting regions appear conserved in 12- and 24-mer cages, suggesting a common mechanism of chaperone function.

Introduction

Cell viability depends on the proper structure and function of the proteome. For protein quality control, all cells have developed elaborate systems of molecular chaperones and proteases[1,2]. Failure of protein homeostasis leads to the accumulation of misfolded or aggregated proteins, a malfunction associated with fatal protein-folding diseases[3]. Members of the ‘High temperature requirement A’ (HtrA) family play a central role in protein quality control in a wide range of organisms as they combine proteolytic and remodelling activities of aberrant proteins in a highly regulated and ATP-independent mechanism[4]. Disturbances in the function of human HtrA proteins (HTRA1, HTRA2) are associated with severe disorders including Alzheimer’s and Parkinson’s diseases and cancers[5-7]. Prokaryotic HtrAs are essential for bacterial virulence and for survival after exposure to various environmental and cellular stresses[8]. In Escherichia coli three HtrA proteins contribute to maintenance of protein quality in the periplasm (DegP, DegQ and DegS)[9-13], with DegQ in particular considered as a model of the HtrA family due to its high sequence identity with many HtrA members[14].

HtrA proteins are composed of a chymotrypsin-like protease domain and one (DegS, HTRA1, HTRA2) or two PDZ domains (DegP, DegQ)[14]. Three protease domains interact tightly to constitute the trimeric building blocks of all HtrA complexes. Whereas membrane-anchored HtrA proteases such as E. coli DegS and human HTRA2 are active as trimers[13,15], several soluble HtrA proteins have been shown to form larger oligomers. Human HTRA1 trimers assemble into 12-mers in the presence of non-native polypeptides[16], whereas E. coli DegP switches between a hexameric resting state without substrate to active 12-, 15-, 18-, 24- and 30-mer states in presence of a substrate that can be refolded or degraded[9,17-20]. Similarly to E. coli DegP, E. coli DegQ changes its oligomeric state from hexamers to either 12- or 24-mers depending on the concentration of unfolded substrate. In addition, E. coli DegQ forms 12-mers in the absence of substrate at acidic pH[10].

Two types of higher oligomeric structure have been described for E. coli DegP: soluble cages (12- and 24-mer) and bowl-shaped structures bound to liposomes (12-, 15- and 18-mers)[9,17,18,20]. For E. coli DegQ, full length 12- and 24-mers remain uncharacterized, but the DegQΔPDZ2 12-mer has been described as a cage-like structure of 135 Å in diameter[10]. This structure provides a model for the dodecameric forms of soluble HtrA proteins containing only one PDZ domain, such as human HTRA1. In addition, the recent crystal structure of a Legionella falloni DegQ dodecamer shows a divergent organization and a smaller size compared to E. coli HtrA proteins (140 Å in diameter for DegQLf vs 165 Å for DegP Ec)[21].

While some HtrA cage-like structures have been obtained in the presence of substrates to be folded or degraded, the protein ligands are not visible in crystallographic structures and symmetrized cryo-electron microscopic (cryo-EM) maps, likely due to conformational and positional flexibility[9,17,18,22]. Only short peptides bound to PDZ1 and protease domains have been resolved in X-ray crystal structures. An asymmetric cryo-EM reconstruction of DegP 12-mer showed a folded outer membrane protein (OMP) encapsulated within the cage, but its low resolution (28 Å) precluded analysis of the DegP–OMP interaction[9].

How the HtrA proteins bind and fold their substrates, central to understanding their chaperone activity, thus remains to be characterized. In order to probe HtrA chaperone function, we carried out a structural and biochemical analysis of the protease deficient S187A mutant of E. coli DegQ in complex with several model substrates. Here, we present cryo-EM structures of DegQ 12- and 24-mer cages encapsulating these substrates. Remarkably, the DegQ 12-mer can accommodate and fold up to 6 lysozymes. The interaction of a bound lysozyme in close-to-native conformation with the DegQ cage is revealed by an asymmetric reconstruction of the DegQ 12-mer–lysozyme complex. PDZ domains are located close to substrate density and a refolding assay indicates that they are required for DegQ chaperone activity.

Results

Cryo-EM structure of DegQ24–β-casein complex

To gain insights into the chaperone function of DegQ, we determined the cryo-EM structures of DegQ-substrate complexes. DegQ 24-mers were formed by incubating proteolytically inactive hexameric DegQS187A with β-casein (a 24 kDa natively unstructured protein used as a model substrate)[10] (Supplementary Fig. 1). Imaging the purified samples by EM reveals mainly large particles approximately 215 Å in diameter and minor complexes of 165 Å (Supplementary Fig. 2). The large complexes correspond to 24-mers and the smaller ones represent 12-mer cages. Cryo-EM images of the 24-mers exhibit 4-, 3- and 2-fold symmetry (Supplementary Fig. 2), and a three-dimensional (3D) reconstruction with octahedral symmetry was obtained by angular reconstitution at 7.5 Å resolution (Supplementary Fig. 3). Overall, the cryo-EM map displays a hollow spherical shape formed by eight trimeric building blocks, similar to the DegP 24-mer[9,17]. The quality of the map allows localization of secondary structure elements (Fig. 1a).

Figure 1

The DegQ 24-mer cryo-EM map reveals a potential β-casein binding site

(a) Two-fold view of the DegQ24 atomic model fitted into the 7.5 Å DegQ24–casein cryo-EM reconstruction with octahedral symmetry. Each trimer is displayed in a different color. (b) Enlargement of contact regions. Cage formation is mediated by a tight interaction between PDZ1 and PDZ2’ domains from adjacent trimers. PDZ1 and PDZ2 domains from the same protomer are connected by an elongated linker spanning 24 Å. Density which is not accounted for by the fitted DegQ atomic model is shown in orange. (c) Overlay of DegP and Q protomers from the 24-mer cages, aligned via their protease domains. DegQ is colored by domain with PDZ1 in cyan, PDZ2 in blue and protease in dark blue and DegP is shown in light green. The 11° and 70° orientation differences observed between DegP/DegQ PDZ1 and PDZ2 domains are indicated. (d) Enlargement of the boxed area in b. DegQ is colored by domain as in c. Residues in the vicinity of the additional orange density belong to domains of three different protomers (labeled domain, domain’ and domain”), namely helix 251-257 of PDZ1 (colored in yellow), loop 408-413 of PDZ2’ (colored in brown) and potentially the LA loop of the protease” domain (residues 31-33 and 58-62 colored in magenta and the LA loop shown as a magenta dashed line).

To obtain a pseudo-atomic model of the complex, a homology model of the DegQ PDZ2 domain was generated, based on the DegP crystal structure, and combined with the DegQ protease-PDZ1 trimer crystal structure[10,19]. Eight copies of the resulting trimer model were docked into the cryo-EM map and their positions refined by flexible fitting, allowing hinge movements between domains[23]. The pseudo-atomic model reveals a strong interaction between PDZ1 and PDZ2 from neighboring trimers, mediated mainly by hydrophobic residues. PDZ1 and PDZ2 of the same protomer are around 24 Å apart, separated by an extended 8-amino acid-long linker (residues 333-341) (Fig. 1b).

Whereas DegP and DegQ 24-mers share a similar global organization, the DegQ complexes are slightly expanded, with a diameter of 210 Å compared to 195 Å for DegP and larger pores on the 2- and 4-fold axes (Supplementary Fig. 4). These differences arise from divergences in DegP/DegQ PDZ domain orientations. When the protease domains of the two structures are superimposed, DegP PDZ1 must be rotated by 11° to fit the DegQ24 cryo-EM map (Fig. 1c). The PDZ2 domain position is substantially different, with a 70° rotation and a 5 Å translation observed between DegP and DegQ. The PDZ domain orientations found in DegQ, unambiguously identified in the cryo-EM map, create a cavity defined by protease, PDZ1 and PDZ2 domains from three different protomers. The cleft thus created is surrounded by helix 251-257 of PDZ1, loop residues 408-413 from PDZ2 and residues 31-33 and 58-62 from the protease domain (Fig. 1d). Additional density (colored in orange in Fig. 1b and d) is present in this cavity and cannot be accounted for by the DegQ model. Its volume is 7,300 A3, corresponding to 6 kDa. We propose that this density corresponds to either (i) the LA loop comprising residues 34-57 of the protease domain, previously shown to be an important regulator of the protease activity but not modeled in DegQ 24-mer due to its flexibility or (ii) to part of the β-casein substrate. Visualization of only 6 kDa out of 24 kDa for β-casein would be consistent with the disordered, natively unstructured state of β-casein. Given that the additional density is buried in a cavity, we suggest that it is more likely to correspond to the ligand density.

Cryo-EM structure of a DegQ 12–peptide complex

To discriminate between ligand and DegQ densities, we determined 3D reconstructions of the DegQ cage with and without high molecular mass substrate. To obtain a homogeneous preparation of DegQ 12-mers devoid of substantial ligand density, we incubated DegQ hexamers with a 20 amino acid long peptide previously shown to bind to the DegQ PDZ1 domain (SPMFKGVLDMMYGGMRGYQV)[10,24]. DegQ12-peptide complexes were purified by size-exclusion chromatography (SEC) and imaged by EM, revealing a preparation of hollow, round particles with a diameter of ~165 Å (Supplementary Fig. 1 and 2). The tetrahedral symmetry of the 12-mer complexes, consistent with the observation of two- and three-fold views in cryo-EM particle averages (Supplementary Fig. 2), was unambiguously identified by angular reconstitution. A 3D cryo-EM map was reconstructed at 7.5 Å resolution (Supplementary Fig. 3) allowing precise fitting of the domains and identification of secondary structure elements (Fig. 2a and b). Previous crystallographic analyses suggest that polypeptides are cooperatively bound by PDZ1 and protease domains, based on the observation of short segments of peptide binding[10,18,24]. Our cryo-EM map is compatible with this binding mode, but the peptide was omitted from the model as the map resolution prevents peptide accurate positioning.

Figure 2

DegQ 12-mer complexes with a peptide and lysozyme

(a,b) Three-fold (a) and two-fold (b) views of the DegQ12 atomic model, colored by trimer, fitted into the 7.5 Å resolution DegQ12–peptide cryo-EM map with tetrahedral symmetry. Regions in the vicinity of the suggested β-casein density of DegQ 24-mer are shown in red. No substrate is visible at this position in the DegQ12–peptide complex, indicated by dashed orange ellipses. The position of the peptide C-terminus from the crystal structure[24] is indicated with a star (*). (c,d) Three-fold (c) and two-fold (d) views of the tetrahedral map of DegQ12–lysozyme at 13 Å resolution. The extra density relative to DegQ12–peptide is colored in orange. Six folded lysozymes are fitted into this additional density. They are colored in different shades of purple, magenta and orange. (e,f) Cut-away of three-fold (e) and two-fold view (f) highlighting the lysozyme fitting.

DegQ 12-mer formation is mediated by the interaction of four DegQ trimers via PDZ1 and PDZ2’ domains from neighboring trimers. The cage is thus formed of four structural units, each of them containing protease-PDZ1 domains of three subunits tightly bound via hydrophobic interactions to three PDZ2’ domains from neighboring protomers. The overall organization of the DegQ 12-mer is reminiscent of the arrangement previously observed in a DegP12–lysozyme cryo-EM map[17]. However there are marked differences in PDZ positions between DegP and DegQ. As in the DegQ 24-mer, PDZ2 is rotated by 70° relative to DegP (Supplementary Fig. 5). In addition, the DegQ 12-mer PDZ1 domain is rotated and shifted 7 Å towards the interior of the cage compared to its position in DegP, with appreciable flexibility of PDZ1 helix 251-257. En-bloc movement of PDZ1 relative to the protease domain is observed in both DegQ 12 and 24-mers. This conformational change was unexpected as PDZ1 was observed in a conserved position in all previously determined E. coli HtrA cage structures[9,10,18] (Supplementary Fig. 6).

Comparison of DegQ 12- and 24-mer cages reveals that the DegQ regions located adjacent to the suggested β-casein density in the DegQ 24-mer (34-57 from the protease domain, 251-257 from PDZ1 and 408-413 from PDZ2, see preceding section) are close to each other in the DegQ12–peptide map (Fig. 2b). They protrude towards the interior of the cage, in a position compatible with substrate binding, near the predicted peptide binding site. The DegQ12–peptide map is devoid of additional density, further suggesting that the additional density in the DegQ 24-mer corresponds to β-casein rather than to the LA loop.

Five or six folded lysozymes bind inside the DegQ 12-mer

We then investigated the positioning of substrates within the DegQS187A 12-mer using reduced and chemically denatured lysozyme (14.3 kDa) as a model substrate. Upon incubation with purified hexameric DegQS187A at 37°C, DegQ12–lysozyme complexes form, which were subsequently purified by SEC (Supplementary Fig. 1). Negative stain and cryo-EM images reveal the presence of cages similar in size to the ones observed for DegQ12–peptide (Supplementary Fig. 2). On the basis of this comparison, the volume of the 12-mer cage appears independent of substrate composition and molecular mass.

The DegQ12–lysozyme images show additional density filling the cages, likely representing the substrate (compare Supplementary Fig. 2b and c). A cryo-EM 3D reconstruction with tetrahedral symmetry was obtained at 13 Å resolution and shows the same DegQ structure as in the peptide complex (Fig. 2c and d, Supplementary Fig. 3). The protease domains remain in the same positions and only a slight opening of the PDZ1-PDZ2’ domain contact is observed. A difference map between DegQ12–lysozyme and DegQ12–peptide reveals the substantial lysozyme density inside the cage, interacting with the inner surface of DegQ (Fig. 2c, d, e, f). The additional density has a volume of 80,000 Å3 corresponding to 66 kDa, indicating the binding of around 5 folded lysozymes (total molecular mass ~70 kDa).

To establish the exact number of lysozymes bound inside the DegQ 12-mer, we conducted mass spectrometry (MS) experiments. We initially recorded a denaturing MS spectrum of DegQ12–lysozyme to determine the mass of the DegQ monomer. The most abundant species has a molecular mass of 44,835.4 ± 11.0 Da, which is smaller than the theoretical mass, likely due to proteolysis at the unstructured termini of DegQ. To determine the precise stoichiometry of DegQ–lysozyme assemblies, we analyzed the apo- and substrate-bound DegQ 12-mer complexes by native mass spectrometry. To obtain DegQ 12-mers in the absence of substrate, we incubated DegQ at pH 5.5, yielding an equilibrium mixture of hexamers and 12-mers, as shown by SEC[10] and negative stain EM (Supplementary Fig. 2). Native mass spectrometry reveals components with the expected masses for DegQ hexamers and 12-mers, (270,135.0 ± 29.2 and 553,105.8 ± 156.5 Da respectively) (Fig. 3a). To determine the number of bound lysozymes, MS spectra of the DegQ12–lysozyme complexes were initially recorded but showed a very broad peak preventing unambiguous mass determination (data not shown). To overcome this, a tandem MS approach was applied. Tandem MS has previously been used to resolve overlapping charge states arising from polydisperse samples and from the presence of different substrate-bound complexes[25-27]. As shown in Figure 3b, the peak at 11050 m/z from the DegQ12–lysozyme precursor was selected using the quadrupole mass filter and subjected to collision-induced dissociation. The peak series in the low m/z region corresponds to a highly charged, ejected DegQ monomer subunit, and the peak series at higher m/z corresponds to the charge-stripped DegQ 11-mer with lysozyme bound. At this region there is a greater separation between charge states facilitating mass assignment[28]. Two predominant charge state series are present in the charge-stripped complex region, corresponding to a mass of 596,555 ± 249 Da and 585,788 ± 365 Da, with the former being the dominant species. These masses correspond to DegQ 11-mer bound to 6 or 5 lysozymes respectively.

Figure 3

Five or six folded lysozymes are bound to the DegQ 12-mer

(a) Mass spectrum of the DegQ protein without substrate bound. Peaks corresponding to different multimeric states of apo-DegQ are labeled in different colors. The most abundant charge state for each of the peak series is indicated. (b) Tandem MS spectrum of the DegQ protein bound to lysozyme. The peak at m/z 11050 corresponds to the precursor ion that was selected for dissociation. The peak series in the lower m/z region corresponds to the ejected DegQ monomer. The two peak series at higher m/z represent charge-stripped DegQ 11-mer bound to 6 (major), and 5 (minor) lysozyme molecules, and are labeled in green and purple respectively. (c) Tryptophan fluorescence spectra (left panel) and box plot of the maximum emission wavelength (right panel) for folded lysozyme (green), unfolded lysozyme (red) and DegQ12–lysozyme (blue). As expected, there is no detectable fluorescence from DegQ alone (yellow), indicating that DegQ12–lysozyme signal corresponds to lysozyme fluorescence.

Consistently, six folded lysozyme molecules can be fitted into this density without clashes (Fig. 2c, d, e, f). In order to confirm that the bound lysozyme substrates are folded as implied by the cryo-EM density, we took advantage of the absence of tryptophan in DegQ to monitor the average tryptophan fluorescence of bound lysozymes[29,30]. The maximum emission of DegQ12–lysozyme was observed at 342.5 nm, whereas the maximum emission is 341.5 nm for folded lysozyme and 352.5 nm for unfolded lysozyme (Fig. 3c). The fluorescence intensity is also increased in the unfolded state (Fig. 3c, with the same concentration of lysozyme used for all spectra). Therefore, the spectra suggest that the bound lysozymes are in a close-to-native state. The small difference in emission maximum and intensity between folded lysozyme and DegQ12–lysozyme could arise from modification of the tryptophan environment by an interaction with DegQ or from a not completely native conformation of the lysozyme. Since cage formation is triggered by denatured but not by folded lysozyme, these results indicate that lysozymes fold within the DegQ cage.

In the tetrahedrally symmetric map, the substrate density level is only about one third of that of DegQ, suggesting that the lysozyme arrangement is most likely asymmetric or disordered. The calculation of an asymmetric reconstruction of DegQ12–lysozyme is thus important for a more accurate description of lysozyme density in the DegQ 12-mer cage.

Substrate binding regions in DegQ cages

An asymmetric 3D reconstruction of DegQ12-lysozyme was obtained at 14.2 Å resolution (Fig. 4, Supplementary Fig. 3), revealing the presence of two separate substrate densities. One density is present in the middle of the complex and does not show any direct interaction with DegQ (Fig. 4a and b, density colored in purple). Its volume corresponds to 2.4 folded lysozymes and is likely to represent mobile substrates sequestered by the DegQ12-mer cage. The second density is in contact with DegQ and has a volume corresponding to 10 kDa (Fig. 4a and b, colored orange). It presents a two-lobed shape, compatible with the two domains of a folded lysozyme (Fig. 4c). A folded lysozyme can be fitted into this density without creating clashes, although it cannot be precisely positioned at the resolution of this EM map. On the other hand, the DegQ domain positions are well defined by the comparison with DegQ12–peptide. Therefore the reliable fitting of DegQ reveals the sites of its interaction with the lysozyme density. The ordered parts of the protease domain LA loop are oriented towards the lysozyme, suggesting involvement of the LA loop in lysozyme binding (Fig. 4c). In addition, lysozyme is adjacent to the protease and PDZ2 domains of two DegQ subunits, as well as the PDZ1 domains of another two DegQ subunits. The PDZ regions that appear to be involved in lysozyme binding in DegQ 12-mer, namely helix residues 251-257 of PDZ1 and loop 408-413 of PDZ2, are the ones located close to the additional density attributed to β-casein in the DegQ 24-mer (Fig. 4d).

Figure 4

Lysozyme–DegQ interaction

(a,b) Three-fold (a) and two-fold (b) views of the asymmetric map of DegQ–lysozyme at 14.2 Å resolution. Two internal densities corresponding to lysozymes are visible inside DegQ and are colored in orange and purple. A folded lysozyme colored in red is fitted into the orange density. (c) Zoomed-in view of the interaction between DegQ and lysozyme in the same orientation as in b, with DegQ colored by domain as in Figure 1. (d) Regions close to substrates are identical in DegQ 12-mers (left) and 24-mers (right). They involve protease, PDZ1 and PDZ2 domains from different subunits. Substrate positions are indicated by orange circles.

PDZ domains are needed for DegQ chaperone activity

The proximity between PDZ regions and substrate prompted us to investigate the role of PDZ domains in DegQ chaperone function. Refolding of α-amylase (MalS), a natural periplasmic substrate of DegP, was used to monitor the chaperone activity of several DegQ constructs. In this assay, DegQ was incubated with chemically denatured MalS and the MalS substrate p-nitrophenylhexaoside (PNP6). When folded, MalS cleaves PNP6 yielding the chromogenic p-nitrophenol that absorbs at 405 nm. The rate of MalS folding in the presence of HtrA chaperones can thus be readily monitored and compared to its spontaneous folding. With this refolding assay, we show that, under the in vitro conditions used, the protease-deficient DegQS187A mutant has a slightly higher chaperone activity than its ortholog DegPS210A (Fig. 5a and b). Deletion mutants of DegQS187A (ΔPDZ2 and ΔPDZ1+2) have a much lower chaperone activity than DegQS187A. A role for PDZ domains in chaperone activity is in accordance with the observation of substrate densities adjacent to these domains in the cryo-EM structures of E. coli DegQ cages. The low chaperone activity of DegQS187AΔPDZ1+2 may also arise from its inability to form cages[10].

Figure 5

Chaperone activity of DegQ

(a) MalS refolding assay. The effects of protease-deficient DegPS210A/DegQS187A constructs on MalS refolding were examined: full length DegQSA in blue, DegQSAΔPDZ2 in green, DegQSAΔPDZ1+PDZ2 in orange and full length DegPSA in purple. Lysozyme (in black) was used as a non-specific solute for the negative control. The linear time course of 405 nm absorbance indicates that the rate of MalS refolding is constant over time. (b) MalS refolding activity of different DegQ and DegP constructs, and lysozyme control, relative to spontaneous MalS refolding in buffer. (c) Steady-state levels of OmpA, OmpC and OmpF in wild type (WT), degP-null and degQ-null mutant strains. Outer membranes were prepared from equivalent numbers of wild-type, degP-null and degQ-null cells.

We then attempted to determine which PDZ domains residues are implicated in substrate binding and folding. PDZ2 loop 408-413 contains poorly conserved hydrophobic residues. Therefore, PDZ2 might be dispensable for substrate binding in other organisms, as reported for L. fallonii DegQ[21]. We thus focused our analysis on the PDZ1 helix 251-257, which contains two hydrophobic residues, namely I253 and F257. In addition, we noticed that residue F266 was correctly oriented to potentially interact with substrates. Unfortunately, the triple mutant I253A F257A F266A, the double I253A F257A mutant and the three corresponding single mutants interfere with 12-mer formation in the presence of lysozyme substrate, preventing the analysis of DegQ12 mutant chaperone activity (Supplementary Fig. 1).

Discussion

The DegQ–substrate complexes presented here reveal new information about the chaperone function of HtrA proteins. The combination of single particle cryo-EM, native mass spectrometry and fluorescence provides strong evidence for the folding of 5 or 6 lysozymes inside the DegQ 12-mer. It suggests that, in the context of the cell, DegQ can capture and enclose multiple small, unfolded substrates that are subsequently refolded within its cavity. Remarkably, six folded lysozymes can fit inside the DegQ 12-mer complex. It is instructive to compare DegQ12–lysozyme with the structure of a closed chaperonin cage containing a newly folded substrate, GroEL–gp31 bound to the T4 bacteriophage capsid protein gp23 (gp31 is the T4 bacteriophage homolog of GroES)[32]. The mechanisms of cage assembly differ: chaperonin cage formation is regulated by ATP binding, whereas DegQ cage assembly is ATP-independent and triggered by substrate binding. The cage architecture is also different, with two compartments alternately used for folding in chaperonins, while DegQ forms a single, larger cage. Substrate packing inside the molecular chaperone cages is very dense for both chaperonins and DegQ. The fraction of substrate that is visible, presumably because of ordered packing in a restricted volume, is also comparable, around 70% for GroEL-gp23 and 78 % for DegQ12–lysozyme. A similar packing density has been observed for tubulin inside the CCT chaperonin[31]. However, chaperonins encapsulate only one substrate at a time per compartment, in contrast to the present finding of up to 6 lysozymes inside one DegQ cage. Folding of multiple substrates within the same compartment, along with the combination of proteolytic and chaperone activities, might have evolved in response to the direct exposure of the periplasm to environmental stresses.

The asymmetric map of DegQ12–lysozyme identifies regions adjacent to lysozyme (Fig. 4c). They originate from protease, PDZ1 and PDZ2 domains of four different protomers. In addition, we show that PDZ-deletion mutants of DegQ have low chaperone activity in MalS refolding assays (Fig. 5a and b). These data suggest that cage formation and/or interaction of PDZ domains with the substrates are required for chaperone activity. It is striking that the regions of DegQ close to lysozyme (helix 251-257 of PDZ1, loop 408-413 of PDZ2 and loop LA of the protease domain) are also adjacent to the proposed β-casein density in DegQ 24-mer (Fig. 4d). Thus our maps suggest that not only the global structural organization, but also the binding mode of chaperone substrates are conserved in the two cages. The cryo-EM maps of DegQ assemblies reveal a conserved organization of cage-like complexes in E. coli DegP and DegQ. PDZ1 and PDZ2’ from different protomers are involved in 12- and 24-mer cage assembly. In a previous study, we proposed a different domain arrangement, based on the fitting of OMP-bound DegP into a low-resolution, asymmetric cryo-EM map. In light of the results presented here and the DegP 12-mer structures[9,17], we have revised our fitting of the OMP-bound DegP 12-mer, leading to a consensus for E. coli DegP and DegQ cage architecture (Supplementary Fig. 7) (4A8D for PDB accession code of new fit).

Interestingly, while the overall organization of E. coli DegQ and DegP 12-mer structure is conserved, the recently published DegQ 12-mer structure of Legionella fallonni shows a different assembly[21]. Whereas a PDZ2 domain of DegQEc only interacts with a PDZ1’ domain of another protomer, a PDZ2 domain in the 12-mer of DegQLf additionally interacts with two PDZ2 and one protease domain of neighboring protomers. Consequently, the DegQLf 12-mer forms a smaller cage (140 Å in diameter vs 165 Å for DegQEc). It will be interesting to determine the physiological implications of these architectural differences between DegQLf and DegQEc in future studies.

Although the global structure of E. coli DegP and DegQ cages is preserved, the positions of their PDZ domains differ. PDZ1 domains of DegQ 12- and 24-mer deviate from their positions in DegP cages, with marked rotation relative to the protease domain. The PDZ1 densities are less well defined in the DegQ 12-mer asymmetric map, implying that they are mobile. PDZ2 orientation differs by 70° between DegP and DegQ. As a consequence, DegP and DegQ cages differ slightly in shape, size and electrostatic potential. These structural divergences might be related to the differences in DegP and DegQ function, for example regarding OMP biogenesis in E. coli. Indeed, we observe that isolated outer membranes from degQ- null strain show no detectable alterations of OMP composition compared to wild type E. coli cells, unlike the degP-null mutant, in which the levels of some OMPs including OmpA, OmpC and OmpF are markedly decreased (Fig. 5c). Consistent with this observation, DegQ is also reported to be dispensable for OMP folding in Neisseria meningitidis[33]. Our data thus support the model of divergence of substrate specificity between DegP and DegQ and suggest directions for further investigation.

In conclusion, this study pinpoints substrate-binding regions within the cavity of E. coli DegQ cages. As many HtrA members are DegQ homologs, the results presented here provide insights on how the HtrA protein family encapsulate and fold substrates.