The Legionella HtrA homologue DegQ is aself-compartmentizing protease thatforms large 12-meric assembliesRobert Wrasea, Hannah Scotta, Rolf Hilgenfelda,b,c,1, and Guido Hansena

aInstitute of Biochemistry, Center for Structural and Cell Biology in Medicine, University of Lübeck, Ratzeburger Allee 160, 23538 Lübeck, Germany;bLaboratory for Structural Biology of Infection and Inflammation, Deutsches Elektronen Synchrotron, Building 22a, Notkestrasse 85, 22603 Hamburg,Germany; and cShanghai Institute of Materia Medica, Chinese Academy of Sciences, 555 Zu Chong Zhi Road, Shanghai 201203, China

Edited* by John Kuriyan, University of California, Berkeley, CA, and approved May 13, 2011 (received for review January 29, 2011)

Proteases of the HtrA family are key factors dealing with foldingstress in the periplasmatic compartment of prokaryotes. In Escher-ichia coli, the well-characterized HtrA family members DegS andDegP counteract the accumulation of unfolded outer-membraneproteins under stress conditions. Whereas DegS serves as a fold-ing-stress sensor, DegP is a chaperone-protease facilitating refold-ing or degradation of defective outer-membrane proteins. Here,we report the 2.15-Å-resolution crystal structure of the secondmajor chaperone-protease of the periplasm, DegQ from Legionellafallonii. DegQ assembles into large, cage-like 12-mers that formindependently of unfolded substrate proteins. We provide evi-dence that 12-mer formation is essential for the degradation ofsubstrate proteins but not for the chaperone activity of DegQ.In the current model for the regulation of periplasmatic chaper-one-proteases, 6-meric assemblies represent important protease-resting states. However, DegQ is unable to form such 6-mers,suggesting divergent regulatory mechanisms for DegQ and DegP.To understand how the protease activity of DegQ is controlled, weprobed its functional properties employing designed protein var-iants. Combining crystallographic, biochemical, and mutagenicdata, we present a mechanistic model that suggests how proteaseactivity of DegQ 12-mers is intrinsically regulated and how deleter-ious proteolysis by free DegQ 3-mers is prevented. Our study shedslight on a previously uncharacterized component of the prokaryo-tic stress-response system with implications for other members ofthe HtrA family.

X-ray crystallography ∣ protein quality control ∣ oligomerization ∣PDZ domain ∣ molecular switch

Protein quality control is essential for all living cells, and com-plex molecular mechanisms have evolved to ensure correct

folding and efficient removal of damaged or misfolded proteins(1–3). In the periplasm of many prokaryotes, proteins of the con-served HtrA family deal with folding stress (4). For pathogenicbacteria, which often encounter a hostile environment insidetheir host cells, HtrA proteins represent important virulencefactors promoting intracellular survival (5).

In Escherichia coli, three HtrA family members, DegS, DegP,and DegQ, have been identified (6). These three proteins share acommon modular domain organization comprising an N-terminaltrypsin-like protease domain and one (DegS) or two (DegP,DegQ) C-terminal PDZ domains. DegS and DegP are well char-acterized (7, 8), and structures for both proteins have beenreported (9–11). DegS is a membrane-associated, homotrimericprotease acting as a folding-stress sensor (12). Activated DegStriggers a signal-transduction pathway that ultimately inducesthe expression of compartment-specific chaperone and proteasegenes including degP (12). DegP is a bifunctional protein withtightly regulated protease and chaperone activities, facilitatingthe degradation or refolding of misfolded periplasmatic proteins(13). In its resting state, DegP forms compact 6-mers composedof two 3-mers arranged in a face-to-face manner (9). Every

monomer of the DegP 6-mer harbors an extended loop, desig-nated LA, which reaches into the proteolytic center of an oppos-ing monomer. This arrangement stabilizes the 6-mer and rendersall six protease sites inactive (9). In the presence of substrate pro-teins, 6-mers reassemble into large 12- and 24-meric complexes,which represent protease-active forms of DegP (10, 14).

In contrast to DegS and DegP, little is known about DegQ. Anamino acid sequence identity of 59% with DegP (Fig. S1) and thepresence of a typical signal sequence indicate DegQ as a secondmajor chaperone-protease of the periplasm. In fact, upon over-expression, DegQ is able to rescue a degP-deficient E. coli strainat elevated temperatures (15). Interestingly, the length of loopLA, crucial for the stabilization of the 6-meric resting state ofDegP, is dramatically reduced in DegQ (for E. coli: 18 residuesin DegQ vs. 41 residues in DegP; see Fig. S1). Many bacterialgenomes encode only two HtrA proteins: a DegS homologueand, judging from length and amino acid sequence of the LAloop, a DegQ homologue (6). Thus, DegQ alone seems to be re-sponsible for maintaining protein homeostasis in the periplasm ofmany prokaryotes. According to previous studies, DegQ assem-bles almost exclusively into higher-order oligomers, most likely12-mers, even in the absence of substrates (16). This raises thequestion if the higher-order DegQ oligomer replaces the 6-meras a resting state or represents a protease-active form, analogousto DegP. In lack of the biochemical and structural data thatallowed the development of detailed functional models for DegPand DegS, it is completely unclear how the protease activity ofDegQ is regulated.

Here, we report the 2.15-Å-resolution X-ray crystal structureof the 12-meric DegQ complex from Legionella fallonii. Thisspecies is closely related to Legionella pneumophila, the causativeagent of Legionnaires’ disease, a severe pneumonia with a highfatality rate (17). Together with structures of three DegQ variantsthat illustrate the formation of substrate complexes and theinactivation of DegQ 3-mers by domain rearrangement, we pro-vide experimental support for a mechanistic model. The crystalstructure of the DegQ 12-mer reveals an assembly mode thatdiffers from available models for the DegP 12-mer based oncryo-EM studies (10, 14).

Author contributions: R.W., R.H., and G.H. designed research; R.W. and H.S. performedresearch; R.W., H.S., and G.H. analyzed data; and R.W., R.H., and G.H. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.

Data deposition: The atomic coordinates have been deposited in the Protein Data Bank,www.pdb.org (PDB ID codes 3PV2, 3PV3, 3PV4, and 3PV5).1To whom correspondence should be addressed. E-mail: hilgenfeld@biochem.uni-luebeck.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1101084108/-/DCSupplemental.

10490–10495 ∣ PNAS ∣ June 28, 2011 ∣ vol. 108 ∣ no. 26 www.pnas.org/cgi/doi/10.1073/pnas.1101084108

Dow

nloa

ded

by g

uest

on

June

10,

202

0

ResultsLegionella DegQ Assembles into Large Complexes in Solution.MatureDegQ from L. pneumophila (DegQLp) and L. fallonii (DegQLf)were produced, purified, and analyzed by size-exclusion chroma-tography (SEC) and dynamic light-scattering (DLS). SEC-elutionprofiles consistently showed a predominant peak indicating thepresence of a large complex with an apparent molecular weightof 320 to 440 kDa (Fig. 1A, Table S1). As for large protein com-plexes molecular-weight determination by SEC is often inaccu-rate, we estimated the oligomer to consist of at least sevenDegQ molecules. According to DLS analysis, the hydrodynamicradius of the particle is approximately 7 nm (Table S1). A smallerfraction of DegQ from both Legionella species formed 3-mers(156 to 164 kDa) and monomers (48 to 50 kDa) (Fig. 1A andTable S1). In contrast to a constant fraction of monomers, apH-dependent dynamic equilibrium between the 3-mer andthe larger complex was observed with the large complex prevail-ing at acidic to neutral conditions and the 3-mer favored at pH 9.5(Fig. S2). Interestingly, the replacement of the active-site serineby alanine (DegQ°) in DegQLp or DegQLf, which inactivates theproteases, led to the sequestering of virtually all molecules intothe large complex (Fig. 1A). It is possible that the formation ofDegQ oligomers might be influenced by associated substrateproteins or peptides originating from the expression host thatcould not be degraded by the DegQ° variants. Indeed, it has beenreported that the equivalent DegP°Ec variant copurified withouter-membrane proteins (OMPs), and assembled into 12- or24-mers (10). Although SDS-PAGE analysis of LegionellaDegQ° (or DegQ) did not indicate the presence of OMPs or othersubstrate proteins coeluting with the large complex (or the lowermolecular-weight species) (Fig. 1B), it cannot be excluded thatsmaller peptides escaped detection.

To explore the role of the PDZ domains in oligomerization,truncated DegQLf variants were produced that lack both PDZdomains (DegQLfΔPDZ1&2) or the C-terminal PDZ2 domainalone (DegQLfΔPDZ2). Both variants were highly soluble, stable,and did not show any signs of misfolding as confirmed by SEC(Fig. S3) and DLS (Table S1). The truncated proteins werenot able to assemble into large complexes and formed 3-mersand monomers only (Fig. S3), implying that the protease domainalone is sufficient for 3-mer formation whereas PDZ2 is essentialfor the assembly of the higher-order oligomers.

Crystal Structure of the DegQ 12-mer. To gain detailed insights intothe structural organization of the observed higher-order oligomerand its functional implications, DegQLf was crystallized and itsthree-dimensional structure determined by X-ray crystallographyto a resolution of 2.15 Å (Table S2). In all four molecules of theasymmetric unit, the protease domain (residues 1–240; residuesof the catalytic triad: S193, H84, and D114), the PDZ1 domain

(residues 241–340), a short linker region (residues 341–352), andthe PDZ2 domain (residues 353–439) were well defined by elec-tron density. Sections of loops LA (residues 31–59), LD (residues152–157), L3 (residues 170–179), and L2 (residues 212–218) weretoo flexible to be traced in at least one molecule (for nomencla-ture of protease loops see Figs. S1 and S4). Low rms deviationsbetween 0.44 and 0.81 Å after superimposition of equivalent Cαatoms indicate that all DegQLf monomers exhibit very similarconformations.

Analysis of the crystal lattice revealed the presence of a highlysymmetric DegQLf 12-mer (Fig. 2A–C), which should correspondto the large complex observed in SEC. The DegQLf 12-merdisplays tetrahedral (332) symmetry and is composed of fourhomotrimers as basic building blocks, each stabilized by extensivecontacts between the three protease domains. Every 3-mer inter-acts with the three remaining 3-mers of the 12-mer, giving rise toa spherical particle with an outer diameter of approximately140 Å. Because of its cage-like organization, it encloses an inter-nal cavity (Fig. 2D) with an average diameter of approximately70 Å that lacks defined electron-density features. The active sitesof the protease domain lining the inner wall of the 12-mer areaccessible only from the interior of the particle (Fig. 2D). Sixlateral pores (approximately 14 Å × 28 Å) connect the internalcavity with bulk solvent (Fig. 2C). Strands β1, β2, and β4 oftwo juxtaposed DegQLf monomers partially cover each pore fromthe inside, restricting the size of the opening. Two LA loops con-necting β1 and β2 are thus positioned in direct vicinity of everylateral pore, although the flexible loop itself could not be tracedin the electron density.

In the DegQLf 12-mer, PDZ1 and PDZ2 are integral parts ofthe protein shell. The peptide-binding groove of PDZ1 is acces-sible from the inner cavity of the DegQLf 12-mer; yet, electron-density maps show no evidence for bound peptides or substratemolecules. It is unlikely that PDZ2 is able to bind substratemolecules because its peptide-binding groove is inaccessible andlacks a positively charged amino acid residue to stabilize theC-terminal carboxyl group of substrates. Instead, PDZ2 is respon-sible for the structural integrity of the 12-mer by mediatinginteractions between adjacent 3-mers (Fig. 2 B, E, and F).Accordingly, in solution DegQ variants lacking a PDZ2 domain(DegQΔPDZ1&2 and DegQΔPDZ2) assemble into 3-mers andare incapable of forming higher-order oligomers (Fig. S3).

The 12-mer Is the Active Form of the Protease. Wild-type DegQLfwas able to proteolytically degrade β-casein and unfolded BSA(via DTT treatment) but not native BSA (Fig. 3A), indicating thatsuitable DegQLf substrates must contain partially unfolded re-gions. The absence of distinct cleavage intermediates suggests aprocessive degradation of substrate proteins into small peptides.Unlike DegPEc (18), DegQLf was unable to process reductivelyunfolded lysozyme (Fig. 3A). In quantitative protease assays,DegQLf was efficiently degrading resorufin-labeled β-casein,whereas, as expected, DegQ°Lf did not show any activity (Fig. 3B).

Deletion of the LA loop (DegQLfΔLA; residues 28–61 re-placed by a single glycine) did not affect formation of 12-mersin solution (Fig. S3) or proteolytic activity (Fig. 3B). In contrast,DegQLf variants incapable of 12-mer formation (DegQLfΔPDZ2,DegQLfΔPDZ1&2) were completely inactive (Fig. 3B). To verifythat oligomerization of DegQLf and not the mere presence of aPDZ2 domain is critical for proteolytic activity, we designed atruncated protein variant lacking only the nine C-terminal resi-dues of PDZ2 (DegQLfΔC9). The DegQLf structure shows thatthese residues should be important for the stability of the 12-merby mediating numerous interactions between PDZ domains ofadjacent 3-mers. Indeed, DegQLfΔC9 formed exclusively 3-mers(Fig. S3) that were proteolytically inactive (Fig. 3B), furthersupporting that the 12-mer represents the protease-active formof DegQLf.

A B

Fig. 1. Legionella DegQ forms large oligomeric complexes. (A) SEC profilesof L. fallonii and L. pneumophila DegQ and DegQ° preparations. Peaks cor-responding to 12-mers, 3-mers, andmonomers are indicated. Elution volumesof marker proteins are shown at the top. (B) SDS-PAGE of purified DegQLf

and DegQ°Lf (S) and fractions obtained after SEC. No copurified substratemolecules were identified after loading comparable amounts of protein from12-mer, 3-mer, and monomer fractions (labeled 12, 3, and 1).

Wrase et al. PNAS ∣ June 28, 2011 ∣ vol. 108 ∣ no. 26 ∣ 10491

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

10,

202

0

To test if DegQLf possesses chaperone-like activity as re-ported for DegPEc, we evaluated the protective effect of DegQLfon heat-induced denaturation of citrate synthase. DegQLf,DegQ°Lf, and DegQLfΔLA, as well as the truncated variantsDegQLfΔC9 and DegQLfΔPDZ2, showed comparable results(Fig. 3C), demonstrating a chaperone-like activity of DegQLfindependent of 12-mer formation or the presence of PDZ2.

Reorientation of PDZ1 Is Inactivating 3-meric DegQ Variants. It is notobvious why protease activity is completely abolished in DegQvariants able to form 3-mers but not 12-mers. Assuming thatthe overall structure of the 3-meric building blocks is independentof incorporation into 12-mers, the exposed protease-active sitesof free 3-mers should in fact promote the degradation of sub-strate molecules. To understand the molecular basis of the inac-tivation of DegQ 3-mers, we crystallized the DegQLfΔPDZ2

variant and determined its three-dimensional structure at 3.1-Åresolution (Table S2). As expected, the protease core of theDegQLfΔPDZ2 3-mer is identical to that observed in the struc-ture of the DegQLf 12-mer (rmsd for Cα atoms of protease3-mers in DegQLf and DegQLfΔPDZ2 is approximately 1.3 Å).Although the overall fold of the PDZ1 domain is also preserved,it is rotated by approximately 180° relative to the protease domaincompared to its orientation in the 12-mer (Fig. 4A). This rotationplaces the peptide-binding cleft of PDZ1 and the protease-activesite on opposite faces of the 3-mer. Furthermore, the linkerregion that connects PDZ1 and protease domain (residues241–249) is inserted into the peptide-binding cleft of PDZ1 inan extended conformation, antiparallel to the central β-sheetof this domain, mimicking a bound substrate molecule (Fig. 4B)(19). The position of PDZ1 with respect to the protease domain isfurther stabilized by hydrogen bonds between E112 and H244

Fig. 2. Structure of the DegQLf 12-mer. (A) View along the threefold axis at the protease interface with protease domain (“Prot,” blue), PDZ1 domain(orange), and PDZ2 domain (green). Individual DegQLf protomers are indicated by white contours. (B) View along the threefold axis at the PDZ2 interface.(C) View along the twofold axis at the lateral pore of the 12-mer. These pores are located in the center between the protease domains of neighboring trimers.(D) Sliced view of the 12-mer. Residues of the catalytic triad (red) lining the inner wall of the central cavity are located in close proximity to the pores.(E) Orientation as in Cwith one protomer displayed in cartoon representation and highlighted by a red contour. (F) Schematic representation of domain inter-actions. The highlighted protomer is shown in the same orientation as in E. Noncovalent interactions are indicated by dashed lines. The position of the pore isshown by a gray ellipse. Symmetry axes are indicated as follows: protease threefold (blue triangle), PDZ2 threefold (green triangle), pore twofold (black ellipse).

A B C

Fig. 3. Assembly of DegQLf 12-mers is essential for protease, but not for chaperone activity. (A) Degradation of BSA and lysozyme. SDS-PAGE analysis ofsamples before (S) and after incubation at 42 °C with (+) or without (−) DTT. Unfolded BSA was processed, whereas native BSA and lysozyme (with or withoutDTT) could not be degraded by DegQLf. (B) Quantitative protease assay with DegQLf proteins using resorufin-labeled casein as substrate. Residual activities ofDegQLf variants (c–h) are compared to wild-type DegQLf (b) and a control sample without protease (a). (C) Chaperone activity of DegQLf proteins. Citratesynthase (CS) was heat-inactivated for the indicated period of time in the presence of DegQLf proteins or BSA (control), and residual CS activity was determined.

10492 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101084108 Wrase et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

and between Q236 and S275. In DegQLfΔPDZ2, loops of theprotease domain exhibit a higher flexibility than in the 12-mer.Furthermore, the catalytic triad is disrupted, as H84 is highly flex-ible and is positioned at a distance of 4.3 Å from the nucleophilicS193, rendering the protease inactive. Although the positioningof PDZ1 in the crystal lattice might be influenced by a Cd2þ ionthat mediates a crystal contact to an adjacent DegQLf 3-mer(CdCl2 was used as an additive in crystallization), the structureshows that PDZ1 is not rigidly attached to the protease domainin DegQLfΔPDZ2. It is tempting to speculate that a very similar

rotation of PDZ1 might switch off protease activity of full-lengthDegQLf 3-mers in solution.

Plasticity of the Active Site and Intrinsic Regulation of ProteaseActivity. In the DegQLf 12-mer, the residues of the catalytic triadassume a protease-competent conformation, but the S1 pocketand the oxyanion hole are blocked by loop L1 (Fig. 5 A and B).Interestingly, electron-density maps of this area demonstrate thatL1 is flexible and adopts a second conformation with a lower oc-cupancy. In this conformation, a rotation of the entire peptidebond between P190 and G191 by approximately 180° leads tothe reconstitution of the oxyanion hole and opens up the S1 pock-et by displacing the side chain of P190 to allow for substrate bind-ing. This suggests L1 as an intrinsic switch element with definedON and OFF conformations existing in an equilibrium. The OFFconformation seems to be favored in DegQLf. To get further in-sights into its function, we disrupted the L1 switch by replacingP190 and the preceding N189 by glycines. The 2.4-Å-resolutionstructure of the resulting variant, DegQLfL1 (Table S2), revealedthat the modified L1 loop adopts a unique conformation differentfrom the ON and OFF conformations and is unable to form afunctional oxyanion hole (Fig. 5D). Accordingly, DegQLfL1 isproteolytically inactive (Fig. 3B). These findings corroboratethe critical importance of an intact L1 switch element for the pro-tease activity of DegQLf.

Details on the activation mechanism of DegQLf were eluci-dated by the 3.1-Å-resolution crystal structure of the protease-inactive DegQ°Lf variant in complex with a peptide substrate(Table S2). Like the wild-type enzyme, this variant formed 12-mers in the crystal lattice. Additional electron density was locatedin the protease-active site of DegQ°Lf, unambiguously indicatinga bound substrate molecule that was copurified with the protein(Fig. 5 E and F). Main-chain and Cβ atoms for six to eight aminoacid residues of the peptide were clearly defined and included inthe model. Superimposition of DegQLf and DegQ°Lf revealeddistinct conformational changes in loops L1, L2, L3, and LD.In the peptide complex, the switch element L1 adopts the ONconformation (Fig. 5C) and residues I188, P190, and N192 alongwith N208 and I211 of L2 shape the S1 specificity pocket of the

A

B

Fig. 4. Structure of DegQLfΔPDZ2. (A) Compared to DegQLf (blue), the PDZ1domain of DegQLfΔPDZ2 (red) is rotated by approximately 180° relative tothe protease domain. The PDZ2 domain of DegQLf is not shown. (B) (Left)The PDZ1 peptide-binding groove of DegQLfΔPDZ2 variant (orange). Theloop connecting protease and PDZ1 (residues G241–G249, shown in red) isinserted into the peptide-binding groove in a substrate-like manner. InDegQLf (equivalent loop shown in green) the peptide-binding groove isaccessible. (Right) For comparison, the PDZ1 domain of DegPEc [gray, PDBID code 3CS0, (10)] bound to a short substrate peptide (blue).

LD

L1

H84

S193

P190Ox

DegQLf

S1

B

LD

L1

H84

S193P190G

Ox

DegQLfL1

S1

D

LD

L1

H84

S193A

P190

Ox

DegQ°Lf

S1

C

LD

LD

LD

L3

PDZ1 PDZ1

PDZ1

L3

L3

L1

L1L1

L2

L2

L2

A

E

L1

L2

LD

H84

S193A

D114

P1

P2 P3

P4

P1‘

P2‘P3‘

L2

L1

F

Fig. 5. Plasticity of the protease-active site of DegQLf. (A) Overview on the active-site loops L1 (orange), L2 (green), L3 (blue), LD (red) of the protease domain ofDegQLf with flexible regions indicated by dashed lines. Residues of the peptide-binding groove of the PDZ1 domain are colored in magenta. (B–D) Comparisonof active-site loops L1 and LD in DegQLf, DegQ°Lf, and DegQLfL1. Positions of the oxyanion hole (Ox) and the S1 pocket are indicated. The oxyanion hole and theS1 pocket are colored in red if malformed and colored in green in the functional conformation. (E and F) Peptide bound to the active site of DegQ°Lf. Peptideresidues P3 and P4 are involved in β-sheet-like contacts with L2. The side chain of P1 is pointing to the S1 pocket formed by residues of L1 and L2.

Wrase et al. PNAS ∣ June 28, 2011 ∣ vol. 108 ∣ no. 26 ∣ 10493

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

10,

202

0

protease (Fig. 5F). This rather restricted pocket is able to accom-modate small, hydrophobic residues. In contrast, the primedsubsites S1′ to S3′ and the nonprimed subsites S2 to S4 lackwell-defined binding pockets and seem to be less discriminatory(Fig. 5F and Fig. S4). Peptide binding is further stabilized bymain-chain hydrogen-bonding interactions between the substrate(P1 and P3) and residues of L2 (T209 and I211; Fig. S5). In thepeptide complex, large portions of L3 are defined by electrondensity, except residues 172–178 located in close proximity tothe substrate binding cleft of PDZ1. A rearrangement of L3enables a hydrogen-bonding interaction between the guanidi-nium group of R170 and the main-chain carbonyl of L151 resid-ing in loop LD of an adjacent molecule (LD*; the asteriskindicates loops of neighboring DegQ molecules). LD* in turnhas moved by 6–7 Å from its position in the peptide-free structure(Fig. 5 B and C) and stabilizes the ON conformation of the L1*switch element via a main-chain hydrogen bond between residuesP190 and F149. Thus, the reorganization of the active-site loopssuggests an interplay between PDZ1 and protease domain ofadjacent monomers. It is easily conceivable that upon bindingof an allosteric activator to PDZ1, a cascade of conformationalrearrangements is initiated along L3 and LD* that finally stabi-lizes loops L1* and L2* in a proteolytically competent state toallow for efficient degradation of substrate molecules.

DiscussionMembers of the HtrA-protein family have been extensively stu-died over more than two decades. Based on biochemical andstructural analysis, functional models have been developed thatshed light onto mechanism and regulation of these intriguingenzymes. In contrast to DegPEc and DegSEc, the third HtrA pro-tein in E. coli, DegQ, has not been characterized in great detail.The notion that numerous bacteria including many pathogensencode only two HtrAs, a DegS and a DegQ homologue (6),prompted us to study DegQ from Legionella.

The predominance of stable 12-mers as the major oligomericform of DegQLf suggests an important biological function for thisassembly. Using our structural data as a framework, we probedthe features of DegQLf 12-mers by designing protein variants thatwere subsequently characterized with regard to oligomerizationbehavior and protease as well as chaperone activity. We show thatprotease but not chaperone activity is dependent on 12-mer for-mation. Large, cage-like 12- and 24-mers have been reported tobe the proteolytically active oligomeric species in E. coli DegP(10, 14). However, DegPEc 12- and 24-mers are formed only tran-siently, dependent on the presence of partially unfolded proteins,whereas DegQLf 12-mers assemble independently of substrateand represent the predominant oligomeric species over a broadrange of environmental conditions. In DegQLf 12-mers as well asin DegPEc 12- and 24-mers, interactions between protease do-mains stabilize the 3-meric building blocks, and PDZ domainsmediate contacts between neighboring 3-mers. Interestingly, thetwo available models for the DegPEc 12-mer based on cryo-EM

data published by two independent groups show discrepanciesregarding the assembly mode of the particle (10, 14). One modelsuggests that interactions between two PDZ2 domains of adja-cent 3-mers stabilize the 12-mer (10). The higher-resolutionreconstruction based on 8-Å-resolution cryo-EM data implicatesthat 12-mer formation is supported by contacts between PDZ1and PDZ2 domains (14). Our structural and biochemical datashow that intact PDZ2 domains are critical for the assembly ofDegQLf 12-mers. However, here PDZ2 is simultaneously inter-acting with three adjacent 3-mers via contacts to two PDZ2,one PDZ1, and one protease domain (Fig. 2F). Thus, a tightlyinterconnected, stable network is formed that is fundamentallydifferent from the models proposed for DegPEc 12-mers, suggest-ing that the general architecture of DegQLf and DegPEc 12-mersmay be different.

The structure of DegQLfΔPDZ2, a variant incapable of assem-bly into 12-mers, suggests that the protease activity of DegQLf3-mers can be switched off by reorientation of the PDZ1 domain.Based on these results, we propose a simple working modelfor function and regulation of DegQLf (Fig. 6). In vivo, DegQLf12-mers are the predominant oligomeric species with a smallerfraction forming 3-mers as indicated by our SEC experiments.As the protease-active sites of free 3-mers are exposed and couldpotentially degrade periplasmatic proteins in an uncontrolledmanner, it is essential that the protease activity is switched offin this state. In the homologous DegPEc, free 3-mers are readilyassembled into inactive 6-mers (9). Yet, 6-mers were absent in allour DegQ preparations, and the shortened LA loop would notsupport the formation of such a resting-state oligomer. Our datasuggest that free DegQ 3-mers are inactivated by a large-scaledomain movement of PDZ1 that might represent a unique safetymechanism protecting the cell from deleterious proteolytic activ-ity. Upon (re-)integration into 12-mers, PDZ1 is reoriented topromote proteolytic activity of DegQLf and allow for degradationof substrate molecules. However, in the absence of suitable sub-strates, the protease-active sites of DegQLf 12-mers are distortedand a distinctive OFF conformation of loop L1 is stronglyfavored. Based on our structural data, we propose that DegQLfis activated via a cascade of conformational changes in L3, LD,L2, and L1, which are most likely initiated by binding of anallosteric activator to PDZ1. The fact that, in our crystal struc-ture, PDZ1 is free of peptides might be attributed to a releaseof the allosteric effector after triggering the activation cascade.As protein substrates could act as allosteric effectors promotingtheir own degradation, a release from the PDZ1 domain aftercleavage is necessary to allow for processive substrate degrada-tion. Similar allosteric activation cascades have been describedfor DegSEc 3-mers (20) and very recently also for DegPEc 24-mers(21). Our findings strongly suggest that the general mechanismfor the intrinsic regulation of HtrA protease activity is conservedin DegQ, DegP, and DegS. A peptide-bound PDZ domain is as-sumed to be the prerequisite for protease activation in DegS andDegP, yet the detection by L3 seems to be remarkably diverse. In

inactive 3-mer active 12-mer inactive 12-mer

Prot

ProtProt

PDZ1 PDZ2

Prot

Prot

Prot

PDZ1 PDZ2

Prot

ProtProt

PDZ1

PDZ2

Fig. 6. Model for DegQLf protease activation. The association of protease-inactive DegQ 3-mers into 12-mers (illustrated by a gray shading) leads to arepositioning of PDZ1 liberating its peptide-binding groove (white triangle). Substrate proteins are trapped during (re-)association of 12-mers, or threadedinto preassembled 12-mers through pores in the protein shell. Subsequent binding of substrate proteins or effector peptides (green triangles) to PDZ1 allos-terically activates the protease domain via a cascade of conformational changes along protease loops L3 → LD� → L1�∕L2�.

10494 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1101084108 Wrase et al.

Dow

nloa

ded

by g

uest

on

June

10,

202

0

DegSEc, peptide-binding causes a reorientation of the PDZ do-main, which in turn relieves inhibitory contacts between the PDZand the protease domain (21). For DegPEc, it has been postulatedthat loop L3 senses the locked position of PDZ1 in the active24-mer, rather than the PDZ1-bound peptide itself (21). Basedon our data, we propose that the allosteric regulation of DegQLfand DegPEc is similar, yet we cannot specify if L3 directly inter-acts with the PDZ1-bound peptide or with the PDZ1 domainbecause the tip of L3 is flexible in our crystal structure.

The structural and biochemical data provided in this studycharacterize DegQ as a unique member of the HtrA familywith distinct features. The high-resolution crystal structure of a12-meric HtrA protein reveals aspects of architecture and regu-lation that also should be relevant for other family members,especially for the DegP 12-mer, although there are differencesfrom models based on cryo-EM data of the latter (10, 14).Our data allowed the development of a preliminary mechanisticmodel, on the basis of which a number of important questions areraised. Firstly, how do substrates gain access to the protease-ac-tive sites that line the inner wall of the 12-mer? On the one hand,DegQLf 12-mers could transiently disassemble into 3-mers encap-sulating substrates upon reassembly; on the other hand, it is alsopossible that substrates are threaded through the large lateralpores present in the shell of the DegQ cage. Secondly, detailsof the allosteric control, especially peptide binding to thePDZ1 domain, remain to be elucidated. A crystal structure ofDegQ in complex with peptides bound to both PDZ1 and pro-tease domain would verify our proposed model of the intrinsicregulation of protease activity in DegQ. Finally, as for DegPEc,the molecular basis for the chaperone activity of DegQ remainsobscure. Experimental results providing answers to these ques-tions are highly desirable and will further our understanding ofthis fascinating protein family.

Materials and MethodsProtein Production and Purification. Details on production and purificationof Legionella DegQ proteins will be described in detail elsewhere. Briefly,N-terminally His-tagged DegQ proteins lacking the signal sequence wereproduced in degP-deficient E. coli KU98 harboring the pQE-31 (Qiagen)derivative pGDR and purified by nickel-affinity chromatography. Fractionscontaining the recombinant proteins were pooled, dialyzed against proteinstorage buffer [20 mM sodium acetate (pH 4.5), 200 mM sodium chloride],and concentrated to 6–20 mg∕mL.

SEC. Analytical SEC was performed using a Superdex 200 HR 10∕30 column(GE Healthcare) equilibrated with two column volumes of running buffer(50 mM Hepes, 200 mM NaCl, pH 7.5). After injection of the protein sample(approximately 1 mg), fractions of 0.5 mL were collected and subsequentlyanalyzed by SDS-PAGE. The SEC column was calibrated with marker proteinsranging from 670 to 29 kDa.

Protein Activity Assays. DegQLf mediated degradation of BSA (GERBU) andlysozyme (GERBU) was analyzed in presence or absence of DTT. The reactionmixture included 20 μMDegQLf, 1 mg∕mL BSA or lysozyme in protein storagebuffer with or without 20 mM DTT. The assay was performed overnight at42 °C, and resulting samples were analyzed by SDS-PAGE. Quantitativeprotease assays using resorufin-labeled casein (Roche) were performed asdescribed previously (13). All measurements were conducted as duplicates.The chaperone activity assay was modified after Buchner et al. (22).

Crystallization, Data Collection, and Structure Determination. DegQLf

constructs were crystallized using the sitting-drop vapor-diffusion technique.Details of methods used for crystallization are provided in SI Materials andMethods. X-ray diffraction data were collected at BESSY (Berlin, Germany)and MAX-lab (Lund, Sweden), integrated with MOSFLM (23), and scaled andmerged with SCALA (24). Initial phases were obtained by molecular replace-ment with Phaser (25) using individual domains of DegPEc [Protein Data Bank(PDB) ID code 3CS0 (10)] as search models. Subsequent model building andrefinement was performed with Coot (26) and REFMAC (27), respectively.Data collection and refinement statistics are summarized in Table S2.

ACKNOWLEDGMENTS. We thank B. Schwarzloh and S. Schmidtke for expert

technical assistance, and U. Müller (BESSY, Berlin, Germany) and T. Ursby(MAX-lab, Lund, Sweden) for assistance with synchrotron data collection.We acknowledge access to beamline BL14.1 of the BESSY II storage ring(Berlin, Germany) via the Joint Berlin MX-Laboratory sponsored by theHelmholtz Zentrum Berlin für Materialien und Energie, the Freie UniversitätBerlin, the Humboldt-Universität zu Berlin, the Max-Delbrück Centrum andthe Leibniz-Institut für Molekulare Pharmakologie. Experiments at MAX-labwere supported by the Integrated Infrastructure Initiative “IntegratingActivity on Synchrotron and Free Electron Laser Science” of the EuropeanCommission (EC) (Contract R II 3-CT-2004-506008). Optimization of crystalswas performed within the OptiCryst project of the EC (LSH-2005-037793;www.opticryst.org). R.H. acknowledges support by the Deutsche Forschungs-gemeinschaft Cluster of Excellence “Inflammation at Interfaces” (EXC 306)and by the Fonds der Chemischen Industrie. He is also supported by a ChineseAcademy of Sciences Visiting Professorship for Senior International Scientists(Grant 2010T1S6).

1. Hengge R, Bukau B (2003) Proteolysis in prokaryotes: Protein quality control andregulatory principles. Mol Microbiol 49:1451–1462.

2. Duguay AR, Silhavy TJ (2004) Quality control in the bacterial periplasm. BiochimBiophys Acta 1694:121–134.

3. Miot M, Betton JM (2004) Protein quality control in the bacterial periplasm. MicrobCell Fact 3:4.

4. Clausen T, Southan C, EhrmannM (2002) The HtrA family of proteases: Implications forprotein composition and cell fate. Mol Cell 10:443–455.

5. Ingmer H, Brøndsted L (2009) Proteases in bacterial pathogenesis. Res Microbiol160:704–710.

6. Kim DY, Kim KK (2005) Structure and function of HtrA family proteins, the key playersin protein quality control. J Biochem Mol Biol 38:266–274.

7. Meltzer M, et al. (2009) Structure, function and regulation of the conserved serineproteases DegP and DegS of Escherichia coli. Res Microbiol 160:660–666.

8. Ortega J, Iwanczyk J, Jomaa A (2009) Escherichia coli DegP: A structure-drivenfunctional model. J Bacteriol 191:4705–4713.

9. Krojer T, Garrido-Franco M, Huber R, Ehrmann M, Clausen T (2002) Crystal structure ofDegP (HtrA) reveals a new protease-chaperone machine. Nature 416:455–459.

10. Krojer T, et al. (2008) Structural basis for the regulated protease and chaperone func-tion of DegP. Nature 453:885–890.

11. Wilken C, Kitzing K, Kurzbauer R, EhrmannM, Clausen T (2004) Crystal structure of theDegS stress sensor: How a PDZ domain recognizes misfolded protein and activates aprotease. Cell 117:483–494.

12. Alba BM, Leeds JA, Onufryk C, Lu CZ, Gross CA (2002) DegS and YaeL participatesequentially in the cleavage of RseA to activate the sigma(E)-dependent extracytoplas-mic stress response. Genes Dev 16:2156–2168.

13. Spiess C, Beil A, Ehrmann M (1999) A temperature-dependent switch from chaperoneto protease in a widely conserved heat shock protein. Cell 97:339–347.

14. Jiang J, et al. (2008) Activation of DegP chaperone-protease via formation of largecage-like oligomers upon binding to substrate proteins. Proc Natl Acad Sci USA105:11939–11944.

15. Waller PRH, Sauer RT (1996) Characterization of degQ and degS, Escherichia coli genesencoding homologs of the DegP protease. J Bacteriol 178:1146–1153.

16. Kolmar H, Waller PRH, Sauer RT (1996) The DegP and DegQ periplasmic endoproteasesof Escherichia coli: Specificity for cleavage sites and substrate conformation. J Bacteriol178:5925–5929.

17. Fields BS, Benson RF, Besser RE (2002) Legionella and Legionnaires’ disease: 25 years ofinvestigation. Clin Microbiol Rev 15:506–526.

18. Jomaa A, et al. (2007) The inner cavity of Escherichia coli DegP protein is not essentialfor molecular chaperone and proteolytic activity. J Bacteriol 189:706–716.

19. Doyle DA, et al. (1996) Crystal structures of a complexed and peptide-free membraneprotein-binding domain: Molecular basis of peptide recognition by PDZ. Cell85:1067–1076.

20. Sohn J, Grant RA, Sauer RT (2009) OMP peptides activate the DegS stress-sensorprotease by a relief of inhibition mechanism. Structure 17:1411–1421.

21. Krojer T, Sawa J, Huber R, Clausen T (2010) HtrA proteases have a conserved activationmechanism that can be triggered by distinct molecular cues. Nat Struct Mol Biol17:844–852.

22. Buchner J, Grallert H, Jakob U (1998) Analysis of chaperone function using citratesynthase as nonnative substrate protein. Methods Enzymol 290:323–338.

23. Leslie AGW (1992) Recent changes to the MOSFLM package for processing film andimage plate data Joint CCP4 + ESF-EAMCB Newsletter on Protein Crystallography 26.

24. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D BiolCrystallogr 62:72–82.

25. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Crystallogr 40:658–674.26. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.

Acta Crystallogr D Biol Crystallogr 66:486–501.27. Winn MD, Isupov MN, Murshudov GN (2001) Use of TLS parameters to model aniso-

tropic displacements inmacromolecular refinement.Acta Crystallogr D Biol Crystallogr57:122–133.

Wrase et al. PNAS ∣ June 28, 2011 ∣ vol. 108 ∣ no. 26 ∣ 10495

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

June

10,

202

0