LapG, Required for Modulating Biofilm Formation by Pseudomonasfluorescens Pf0-1, Is a Calcium-Dependent Protease

Chelsea D. Boyd,a Debashree Chatterjee,b Holger Sondermann,b and George A. O’Toolea

Department of Microbiology and Immunology, Geisel School of Medicine at Dartmouth, Hanover, New Hampshire, USA,a and Department of Molecular Medicine,College of Veterinary Medicine, Cornell University, Ithaca, New York, USAb

Biofilm formation by Pseudomonas fluorescens Pf0-1 requires the cell surface adhesin LapA. We previously reported that LapG, aperiplasmic cysteine protease of P. fluorescens, cleaves the N terminus of LapA, thus releasing this adhesin from the cell surfaceand resulting in loss of the ability to make a biofilm. The activity of LapG is regulated by the inner membrane-localized cyclic-di-GMP receptor LapD via direct protein-protein interactions. Here we present chelation and metal add-back studies demonstrat-ing that calcium availability regulates biofilm formation by P. fluorescens Pf0-1. The determination that LapG is a calcium-de-pendent protease, based on in vivo and in vitro studies, explains the basis of this calcium-dependent regulation. Based on thecrystal structure of LapG of Legionella pneumophila in the accompanying report by Chatterjee and colleagues (D. Chatterjee etal., J. Bacteriol. 194:4415– 4425, 2012), we show that the calcium-binding residues of LapG, D134 and E136, which are near thecritical C135 active-site residue, are required for LapG activity of P. fluorescens in vivo and in vitro. Furthermore, we show thatmutations in D134 and E136 result in LapG proteins no longer able to interact with LapD, indicating that calcium binding re-sults in LapG adopting a conformation competent for interaction with the protein that regulates its activity. Finally, we showthat citrate, an environmentally relevant calcium chelator, can impact LapG activity and thus biofilm formation, suggesting thata physiologically relevant chelator of calcium can impact biofilm formation by this organism.

Bacteria sense and respond to fluctuating environmental nutri-ents and signals to coordinate adaptive changes in metabolic

pathways and physiological outputs. Integration of environmen-tal cues affords bacteria the ability to make important decisionsregarding how to respond to their constantly changing environ-ments. When coming in contact with a surface, whether in natu-ral, industrial, or clinical settings, bacteria must evaluate whetherthe surface and the local environment are favorable for attach-ment.

Initial colonization of a surface is the first step in biofilm for-mation. This transition from a planktonic to a sessile lifestyle isoften in response to a variety of environmental cues, such as os-molarity, pH, carbon, iron availability, oxygen tension, and tem-perature (14, 15, 19, 32–35, 37, 40, 43). Colonization of a surfaceand the subsequent development of a biofilm involve an array ofcellular factors and diverse molecular mechanisms. However, aunifying theme across bacterial species is that synthesis of the cel-lular signaling molecule bis-(3=-5=)-cyclic dimeric GMP (c-di-GMP), a central player in the signaling networks that control bio-film formation, coincides with the transition to a sessile lifestyle.

Pseudomonas fluorescens Pf0-1, a model organism for biofilmresearch, is a biological control agent that promotes plant growthby forming biofilms on plant roots (10). The environmental nu-trient Pi has been shown to be an important signal regulating theswitch between a planktonic and a sessile lifestyle in P. fluorescens(23, 24). Work by our group identified a large adhesive protein,LapA, required for the attachment of P. fluorescens (12). Subse-quent studies have elucidated a c-di-GMP signaling system thatregulates biofilm formation through posttranslational modifica-tion of LapA (26–28). In short, under conditions that do not favorbiofilm formation, such as low Pi, the periplasmic cysteine pro-tease LapG cleaves LapA from the cell surface, thus releasing theadhesin and preventing attachment (28). LapG activity is regu-lated by the inner membrane c-di-GMP effector protein LapD.

LapD binds c-di-GMP and undergoes conformational changes(26, 27), and through an inside-out signaling mechanism that isdependent upon c-di-GMP, LapD binds LapG, preventing LapG-dependent cleavage of LapA from the cell surface. Thus, underconditions that favor biofilm formation, LapA remains at the cellsurface, promoting biofilm formation (28).

While we have a detailed picture of the LapD-LapG c-di-GMPcontrol circuit, from the environmental Pi input to the LapA out-put that regulates biofilm formation in P. fluorescens Pf0-1, muchless is known about how P. fluorescens regulates biofilm formationin response to other environmental nutrients. Calcium (Ca2�)regulates cellular function in bacteria and is involved in manydifferent processes, such as adhesion, cell cycle and cell division,pathogenesis, motility, chemotaxis, and quorum sensing (6, 8, 13,22, 30, 31, 41). Ca2� has been shown to both inhibit and stimulatebiofilm formation in bacterial species that also encode large adhe-sion proteins that mediate biofilm formation (2, 42).

In this study and in the accompanying report by Chatterjee andcolleagues (3), we describe the analysis of the LapG protease. LapGbelongs to the domain of unknown function 920 (DUF920) (orCOG3672) family. Aside from previous studies by our group dem-onstrating that LapG is a cysteine protease that functions to cleaveLapA from the cell surface (28), little is known regarding the func-tion of LapG. Here we present genetic and biochemical studiesdemonstrating that the LapG protease of P. fluorescens Pf0-1 is a

Received 16 April 2012 Accepted 8 June 2012

Published ahead of print 15 June 2012

Address correspondence to George A. O’Toole, georgeo@dartmouth.edu.

Supplemental material for this article may be found at http://jb.asm.org/.

Copyright © 2012, American Society for Microbiology. All Rights Reserved.

doi:10.1128/JB.00642-12

4406 jb.asm.org Journal of Bacteriology p. 4406–4414 August 2012 Volume 194 Number 16

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

on S

eptember 12, 2020 by guest

http://jb.asm.org/

Dow

nloaded from

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

Ca2�-dependent enzyme. These findings are supported and ex-tended by the accompanying report by Chatterjee and colleagues,which presents the structure of the LapG homolog of Legionellapneumophila (3).

MATERIALS AND METHODSStrains and media. The strains and plasmids used in this study are listed inTable S1 in the supplemental material. SMC 4798, described previously(23, 28, 29), is the wild-type (WT) P. fluorescens strain used in this study.P. fluorescens strain SMC 4798 carries three hemagglutinin (HA) epitopetags within a fully functional, chromosomally encoded LapA protein. P.fluorescens and Escherichia coli were routinely cultured with liquid LBmedium in a test tube or on solidified LB with 1.5% agar and grown at30°C or 37°C, respectively. Gentamicin was used at 30 �g/ml for P. fluo-rescens and at 10 �g/ml for E. coli. Tetracycline was used at 30 �g/ml onplates and at 15 �g/ml in liquid culture for P. fluorescens. Kanamycin wasused at 30 �g/ml for P. fluorescens. Ampicillin (Amp) was used at 100�g/ml for E. coli. Arabinose was used to induce expression from pMQ72 at0.2% for complementation and overexpression experiments. Isopropyl-�-D-thiogalactopyranoside (IPTG) at 1 mM was used to induce expres-sion from pET21. P. fluorescens was grown in K10T-1 medium for biofilmassays, LapA localization, protein expression, and coprecipitations.K10T-1 was previously described as a phosphate-rich medium (24).

Biofilm formation assays. Biofilm formation assays were performedexactly as previously described (29). EDTA (Fisher), EGTA (Sigma-Al-drich), calcium chloride (CaCl2; Sigma-Aldrich), magnesium chloride(MgCl2; Fisher), ferric chloride (FeCl3; Sigma-Aldrich), cupric chloride(CuCl2; Sigma-Aldrich), zinc chloride (ZnCl2; Sigma-Aldrich), and man-ganese chloride (MnCl2; Sigma-Aldrich) were added to biofilm assays atthe beginning of the assay at a final concentration of 500 �M. Sodiumcitrate (Fisher) was added at the start of the biofilm assay at a final con-centration of 0.4% (wt/vol). Staining, imaging, and quantification of at-tached biofilm biomass were also performed as previously described (23).

Constructs for clean deletion of Calx-� and complementation andoverexpression of LapG variants. Saccharomyces cerevisiae strain InvSci(Invitrogen) was used to construct plasmids for clean deletion, comple-mentation, and overexpression experiments through in vivo homologousrecombination as previously described (39). All deletion mutants wereconstructed by the same basic strategy as previously described (29).

Complementation and overexpression studies utilized the pMQ72plasmid as previously described (29). LapG-D134A and LapG-E136A site-directed mutations were made as previously described (29). Briefly, twofragments were amplified by PCR either from the LapG complementationplasmid template containing a six-histidine (6H) tag or from theLapG-HA LapD-6H complementation plasmid template (28) using aprimer on the vector backbone and a primer divergent from and exclud-ing the codon for either Asp or Glu in the calcium-binding site. Divergentprimers included additional bases encoding Ala replacement codons, aswell as sequence to facilitate yeast recombination. Yeast was transformedwith pMQ72 and both PCR fragments, resulting in a plasmid identical tothe parent, except for the point mutation at the calcium-binding residue.Sequencing confirmed the successful construction of each plasmid andchromosomal alteration.

LapA localization. For Western and dot blot assays of the LapA pro-tein, we used strains with internal HA tags in the chromosomal lapA gene.Culturing conditions, preparation of clarified cell extracts, and analysis ofsamples for LapA localization were performed exactly as previously de-scribed (23, 28). Detection of cell surface LapA by dot blotting was alsoperformed as previously described (27, 28). To test the effects of the ad-dition of EGTA, EDTA, CaCl2, or MgCl2 on LapA localization, each com-pound was added at the beginning of the 6-h subculturing period to a finalconcentration of 500 �M.

Protein purification. The periplasmic, signal sequence-processedmodule of P. fluorescens LapG (residues 50 to 251) was cloned, expressed,and purified by standard molecular biology and chromatography tech-

niques as described previously (26). Briefly, the gene was amplified bystandard PCR and cloned into a pET21 expression plasmid (Novagen)producing a C-terminally 6H-tagged protein. LapG point mutants forbiochemical studies were generated by using a QuikChange site-directedmutagenesis kit (Stratagene) and following the manufacturer’s instruc-tions. The WT and mutant proteins were expressed in T7 Express cells(New England BioLabs) and grown at 37°C in terrific broth medium sup-plemented with 100 �g/ml Amp. At an optical density at 600 nm of �1,the temperature was reduced to 18°C and protein expression was inducedby the addition of 1 mM IPTG. After 16 h of expression at 18°C, cells wereharvested by centrifugation, resuspended in Ni-nitrilotriacetic acid(NTA) buffer A (25 mM Tris-HCl [pH 8.5], 500 mM NaCl, 20 mM imi-dazole), and flash frozen in liquid nitrogen. Following thawing and celllysis by sonication, the cell debris was removed by centrifugation and theclear lysates were incubated with Ni-NTA resin (Qiagen) that was pre-equilibrated with Ni-NTA buffer A. After the lysates were allowed to flowthrough, the resins were washed with 20 column volumes of buffer A andproteins were eluted with 5 column volumes of Ni-NTA buffer B (25 mMTris-HCl [pH 8.5], 500 mM NaCl, 300 mM imidazole). The eluted pro-teins were subjected to size exclusion chromatography on a Superdex 200column (GE Healthcare) preequilibrated with gel filtration buffer (25 mMTris [pH 8.5], 150 mM NaCl). The proteins were concentrated on Amiconfilters with an appropriate molecular size cutoff, flash frozen in liquidnitrogen, and stored at �80°C. Protein purity was determined by SDS-PAGE and was estimated to be �95%. Purification of N-Term-LapA-6Hfrom E. coli was done by standard nickel affinity chromatography tech-niques as previously described (20).

LapG activity assays. Bacterial cultures were grown in the same man-ner, and clarified cell extracts were prepared as for LapA localization, asdescribed previously (23, 28). LapG activity assays with mini-LapA andN-Term-LapA were performed as previously described (28). Assays in-corporated 375 ng of LapG, D134A LapG, or E136A LapG. Activity assayswere performed in the presence of 500 �M EGTA, CaCl2, or MgCl2 or0.4% (wt/vol) sodium citrate, as indicated.

Coprecipitations. Proteins were precipitated from clarified cell ex-tracts prepared in the same manner as described above for LapA localiza-tion (23, 28). Coimmunoprecipitation with nickel resin was preformedexactly as previously described (28).

Statistical analysis. Statistical analysis of the data presented utilizedthe Student t test, analysis of variance with Tukey’s honest significantdifference as a posttest, or linear models, as appropriate. Significance wasset at the traditional 95% confidence interval.

RESULTSEGTA chelation induces a hyperadherent biofilm phenotype.Calcium (Ca2�) has been shown to both inhibit and stimulatebiofilm formation in bacterial species that encode large adhesionproteins that mediate biofilm formation (2, 42). To test the re-quirement of Ca2� for biofilm formation in P. fluorescens, we in-dividually supplemented the biofilm medium K10T-1 with thechelating agent EDTA or EGTA. EDTA has a strong affinity foriron ions and a much weaker affinity for Ca2� (4), while EGTA hasa strong affinity for Ca2� and a much weaker affinity for magne-sium ions (Mg2�) (44).

WT P. fluorescens biofilm biomass in the static microtiter dishbiofilm assay was not statistically significantly different (P � 0.05)in the presence of EDTA, but the biofilm formed by the WT sup-plemented with EGTA was significantly increased (P � 0.05) andmore than twice that of the WT grown in K10T-1 (Fig. 1A). Che-lation by EGTA impacts the initial stage of biofilm formation, asan accumulation of biofilm biomass begins at the 2-h time pointand continues through the endpoint of the assay at 6 h (see Fig. S1in the supplemental material). Growth of P. fluorescens was unaf-fected by the addition of these chelating agents at the concentra-

LapG is a Ca2�-Dependent Protease

August 2012 Volume 194 Number 16 jb.asm.org 4407

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

tions used here (data not shown). Given the high affinity of EGTAfor Ca2�, our findings are consistent with a role for Ca2� in bio-film formation. These results suggest that decreasing the Ca2�

concentration stimulates biofilm formation and induces a hyper-adherent biofilm phenotype.

The Calx-� domain of LapA is not required for calcium-de-pendent changes in biofilm formation. LapA contains a Calx-�Ca2�-binding domain, indicating that LapA may directly bindCa2�, thus possibly explaining the increased biofilm formationseen after the addition of EGTA. The Calx-� domain was initiallyidentified in Na�-Ca2� exchangers of eukaryotic systems, whichexpel Ca2� from the cytoplasm (1, 21, 38). Given the presence ofthe Calx-� domain, we hypothesized that this domain may berequired for Ca2�-mediated effects on biofilm formation.

To assess the requirement of the Calx-� domain for biofilmformation, we deleted the domain from LapA (this domain spansamino acids T4049 to 4094D). Deletion of the domain did notresult in a change in biofilm biomass compared to that of the WT(Fig. 1B). Consistent with no apparent impact on biofilm forma-tion, we also observed no difference in the levels of LapA in thecellular, supernatant, or cell surface fractions in the WT comparedto those in the strain with the LapA variant with the Calx-� do-main deleted (Fig. 1C and D). A control study demonstrating thespecificity of LapA detection in these fractions is presented in Fig.S2 in the supplemental material. Furthermore, supplementationof the K10T-1 medium with CaCl2, EDTA, or EGTA did not affectthe biofilm phenotype of the Calx-� mutant compared to that ofthe WT (see Fig. S3A in the supplemental material) or alter thelocalization of LapA in the Calx-� mutant (see Fig. S3B and C inthe supplemental material). Together, these data suggest that the

Calx-� domain of LapA is not required for chelator- or Ca2�-mediated effects on biofilm formation.

Calcium rescues the chelation-induced hyperadherent bio-film phenotype. Given that EGTA has an affinity for both Ca2�

and Mg2�, we hypothesized that addition of CaCl2 or magnesiumchloride (MgCl2) to the biofilm assay might reverse the effects ofthe EGTA-induced hyperadherent biofilm phenotype. The WThyperadherent biofilm phenotype was reversed in the presence ofEGTA and 500 �M CaCl2 compared to that in the presence ofEGTA alone (Fig. 2A) but not when 500 �M MgCl2 was added tothe EGTA-treated WT strain (Fig. 2B; see Fig. S4 in the supple-mental material).

K10T-1 contains 610 �M MgSO4 to support bacterial growth.To chelate the Mg2� contained in the K10T-1 medium, chelationexperiments were also performed with 1 mM EGTA. Addition of 1mM or 2.5 mM MgCl2 did not reverse the WT hyperadherentbiofilm phenotype (see Fig. S4A in the supplemental material),suggesting that changes in the Mg2� concentration are not re-sponsible for the alteration in biofilm formation. Other metal ionswere tested for the ability to reduce the EGTA-induced hyperad-herent biofilm. Ferric chloride, cupric chloride, zinc chloride, andmanganese chloride, all at 500 �M, did not reverse the effects ofthe EGTA-induced hyperadherent biofilm (data not shown).Growth of P. fluorescens was unaffected by the addition of metalions at the concentrations used here (data not shown). We alsosupplemented the medium (in the absence of EGTA) with CaCl2or MgCl2 to assess the effects on biofilm formation of the WT.Neither Ca2� nor Mg2� significantly affected biofilm biomass(Fig. 2A and B). Together, these data suggest that EGTA chelationcauses a hyperadherent biofilm phenotype due to the loss of Ca2�.

FIG 1 Chelation by EGTA stimulates biofilm formation. (A) The chelator EGTA stimulates biofilm formation. Shown is the biofilm formation of WT P.fluorescens Pf0-1 grown in K10T-1 medium alone or supplemented with EDTA or EGTA at 500 �M. Assay mixtures were incubated for 6 h at 30°C before beingstained with crystal violet and quantified. The y axis shows the optical density at 550 nm (OD 550) of the extracted crystal violet used to determine the bacterialbiofilm biomass. (B) Deletion of the Calx� domain has no impact on biofilm formation. Shown is a biofilm formation assay of the WT and a strain with a LapAvariant with the Calx� domain deleted. The biofilm assay was performed as outlined for panel A. (C) Levels and localization of LapA in whole-cell andsupernatant fractions. Shown is a Western blot assay of the levels of WT LapA and the LapA �Calx� variant in the whole-cell (Cell) and supernatant (Supe)fractions. Here and in all subsequent figures, LapA was detected via an engineered HA tag as described in Materials and Methods. (D) Cell surface levels of LapA.Shown are a representative dot blot assay (top) and quantification of the pixel density of six to eight replicates (bottom) to assess the level of cell surface LapA ofthe WT or the strain expressing the LapA �Calx� variant.

Boyd et al.

4408 jb.asm.org Journal of Bacteriology

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

We also examined the impacts of Ca2� alone, EGTA alone, andEGTA plus Ca2� on LapA levels and localization in the WT. Thesetreatments did not alter the cellular level of LapA (Fig. 2C), butconsistent with the phenotypes shown in Fig. 2A, treatment withEGTA markedly increased LapA on the cell surface and reducedsupernatant LapA, and these effects were partially reversed by theaddition of Ca2� (Fig. 2C and D).

We next assessed the potential role of Ca2� in mediating bio-film formation via factors known to impact biofilm formation,such as LapG, LapD, and LapA. As expected, the �lapG mutantformed a hyperadherent biofilm phenotype in the presence ofEGTA alone or in medium with EGTA and CaCl2 (Fig. 2A), as wellas in the presence of EGTA and MgCl2 (Fig. 2B). The lack of anyimpact of these treatments on the phenotype of the �lapG mutantis reflected in the lack of impact on the level or localization of LapA(Fig. 2C and D).

Interestingly, the �lapD mutant does not form a biofilm inK10T-1 medium, but the biofilm biomass of the �lapD strain is

significantly increased and approximately three times greater inthe presence of EGTA (Fig. 2A), a phenotype reversed by Ca2�

(Fig. 2A) but not Mg2� (Fig. 2B). These data suggest that theEGTA-induced hyperadherent biofilm phenotype is independentof LapD.

The hyperadherent biofilm phenotype of the WT and the�lapD mutant strain treated with EGTA phenocopies that ob-served in the �lapG mutant, indicating that Ca2� may mediatebiofilm formation via LapG. As a control, we showed that the lapAmutant strain does not form a biofilm under any of the conditionstested (Fig. 2A and B), supporting the conclusion that the hyper-adherent biofilm phenotypes observed are dependent on the LapAadhesin.

LapG cleavage of LapA requires calcium. One explanation forthe similarity in phenotypes between Ca2� chelation and the�lapG mutant is that the LapG protease requires Ca2� for func-tion. To test this hypothesis, we assessed the ability of LapG tocleave N-Term-LapA in the presence of EGTA and/or Ca2� by

FIG 2 Calcium chelation stimulates biofilm formation. (A) Calcium depletion by EGTA stimulates biofilm formation. Shown are biofilm assays of the WT andthe �lapG, �lapD, and lapA mutants grown in K10T-1 medium or in this medium supplemented with CaCl2 (500 �M), EGTA (500 �M), or both. The y axisshows the optical density at 550 nm (OD 550) of the extracted crystal violet used to determine the bacterial biofilm biomass. (B) Magnesium does not reverseEGTA-stimulated biofilm formation. Assays were performed as described for panel A but with MgCl2 (500 �M) in place of CaCl2. (C) Levels and localization ofLapA in whole-cell and supernatant fractions. Shown is a Western blot assay of the levels of LapA in the whole-cell (Cell) and supernatant (Supe) fractions of thestrains indicated under the treatment conditions indicated. (D) Cell surface levels of LapA. Shown is a representative dot blot assay (top) and quantification ofthe pixel density of six to eight replicates (bottom) to assess the levels of cell surface LapA fractions of the strains indicated under the treatment conditionsindicated.

LapG is a Ca2�-Dependent Protease

August 2012 Volume 194 Number 16 jb.asm.org 4409

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

using an in vitro activity assay. Using this assay, we have previouslyshown that LapG cleaves 10 kDa from the so-called N-Term-LapAconstruct, which consists of the first 235 amino acids of LapA andcontains a 6H epitope tag at its C terminus (28). In this assay,purified LapG and purified N-Term-LapA are coincubated in thepresence of EGTA and/or CaCl2 and the cleavage of N-Term-LapA is assessed by SDS-PAGE, followed by Western blotting withan anti-His antibody (28).

In the presence of EGTA, LapG is unable to cleave N-Term-LapA, while in the presence of EGTA and CaCl2, cleavage of N-Term-LapA by LapG is restored. Consistent with a specific role forCa2�, cleavage is not restored upon the addition of MgCl2 toEGTA-treated assay mixtures (Fig. 3). LapG is able to cleave N-Term-LapA in the presence of CaCl2 or MgCl2 (Fig. 3). These dataindicate that LapG-dependent cleavage of N-Term-LapA requiresthe presence of Ca2�.

Residues D134A and E136A of LapG are required for func-tion. In the accompanying report, Chatterjee and colleaguessolved the structure of the LapG homolog from L. pneumophila(3). L. pneumophila and P. fluorescens LapG show 47% and 66%sequence identity and similarity, respectively. Additionally, aswith P. fluorescens LapG, L. pneumophila LapG cleaves N-Term-LapA (3), indicating that the high sequence conservation betweenthese two proteins is reflected in functional conservation of activ-ity. Based on the studies presented here, Chatterjee and colleaguesalso crystallized L. pneumophila LapG in the presence of Ca2� andafter EGTA treatment. Ca2� could be modeled at a site adjacent toLapG’s catalytic triad only when the protein was crystallized in thepresence of this cation. Isothermal titration calorimetry was usedto validate and quantify calcium binding (3). Comparison of theCa2�-bound and Ca2�-free LapG structures indicated that resi-dues D134 and E136 in P. fluorescens LapG might be required forCa2� binding. As shown in the supplemental data of the accom-panying report, D134 and E136 flank the catalytic cysteine resi-dues and are 100% conserved in 24 bacterial species (3) and thelarger family of DUF920 domain-containing proteins or bacterialtransglutaminase-like cysteine proteases (9).

To determine if residues D134 and E136 are required for P.fluorescens LapG function, Ala substitutions were constructed and

the biofilm phenotype of strains expressing WT LapG, LapG-D134A, or LapG-E136A from a multicopy plasmid was tested. WTLapG and the LapG-D134A and -E136A variants, which were de-tected via their C-terminal 6H epitope tag, are stably expressed invivo (Fig. 4A, bottom).

Expression of WT LapG from a multicopy plasmid results inno observed biofilm, as expected (28), because overexpression ofLapG causes release of LapA into the supernatant from the cellsurface (Fig. 4A to C). In contrast, strains expressing the LapG-D134A and LapG-E136A variants resulted in a hyperadherent bio-film phenotype and LapA localization profiles similar to that ofthe �lapG mutant strain carrying an empty vector (Fig. 4A to C).These results suggest that residues D134 and E136 are required forLapG function.

Given the apparent lack of function of the LapG-D134A and-E136A mutant proteins in vivo, we hypothesized that the Ca2�-binding mutants would be unable to cleave LapA in vitro. In thesestudies, we first prepared a cell extract from the lapG mutant car-rying a plasmid expressing mini-LapA. Mini-LapA is a smallerconstructed variant of LapA consisting of the N and C termini ofthe protein flanking an internal myc epitope tag (28). This mini-LapA extract served as a substrate to monitor the N-terminalcleavage of LapA. Cell extracts were also prepared from the �lapGmutant carrying an empty vector, the �lapG mutant carrying aplasmid overexpressing LapG, or the LapG-D134A or -E136Amutant protein. The mini-LapA- and LapG-containing cell ex-tracts were mixed, and reactions were then analyzed by Westernblotting to reveal that mini-LapA cleavage occurs only in the pres-ence of WT LapG and not in the presence of the D134A or E136ALapG variant (Fig. 4D), suggesting that the Ca2�-binding residuesare required for mini-LapA cleavage.

We also tested the activity of WT LapG, LapG-D134A, andLapG-E136A by using purified LapG enzymes and substrate. Pu-rified N-Term-LapA was incubated with purified LapG variants.In contrast to WT LapG protein, the LapG-D134A and LapG-E136A variants were unable to cleave N-Term-LapA (Fig. 4E).Taken together, these data support the conclusion that Ca2�-binding residues are required for LapG protease activity.

LapG calcium point mutants cannot interact with LapD. Wenext assessed if mutation of D134 or E136, in addition to the lossof protease activity, might disrupt the c-di-GMP-dependent in-teraction between LapG and LapD. Operon constructs of WTLapG and LapD, D134A LapG and WT LapD, or E136A LapG andWT LapD were introduced on a multicopy plasmid into a strainwith clean unmarked deletions of the lapG and lapD genes (�lapG�lapD). LapG variants contain an internal HA epitope tag, andLapD contains a 6H epitope tag at its C terminus. Note that incontrast to the data shown in Fig. 4, wherein LapG alone is ex-pressed on a plasmid, here the coexpression of LapD and LapGfrom a plasmid results in the restoration of a WT phenotype to the�lapG �lapD mutant. Cell extracts were prepared to examinethe expression of LapG variant and LapD proteins, showing thatthe proteins are stably expressed (Fig. 5A, bottom).

Because LapG-D134A and -E136A do not complement the�lapG hyperadherent biofilm phenotype (Fig. 4A), we predictedthat LapG-D134A or -E136A introduced on a multicopy plasmidwith WT LapD into the �lapG �lapD mutant strain would alsonot be able to complement the �lapG �lapD mutant strain’s hy-peradherent biofilm phenotype. In the �lapG �lapD mutantstrains, the expression of LapG-D134A or -E136A cannot comple-

FIG 3 Calcium chelation inhibits LapG protease activity in vitro. Shown is aWestern blot assay developed with antibodies directed against purified, 6H-tagged N-Term-LapA. N-Term-LapA contains the LapG cleavage site mappedto the N terminus of LapA, as described previously (28). Cleavage of N-term-LapA by LapG results in a shift of the substrate from �25 kDa to �15 kDa. Theleftmost lane shows the substrate-only control. Under the conditions usedhere, LapG is able to cleave �50% of the substrate. The addition to each assaymixture is indicated above the lane. All additions were made to 500 �M.

Boyd et al.

4410 jb.asm.org Journal of Bacteriology

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

ment the hyperadherent biofilm phenotype. In contrast, the�lapG �lapD mutant strain overexpressing WT LapG does show areduction of the hyperadherent biofilm phenotype (Fig. 5A).

Using a coimmunoprecipitation assay, we previously demon-strated that LapD and LapG interact in a c-di-GMP-dependentmanner (28). Under the same conditions, a nickel resin was usedto pull down LapD-6H and LapG-HA coprecipitation was as-sessed. WT LapG coprecipitated with WT LapD in the presence of

c-di-GMP, with only a weak interaction in the absence of thedinucleotide (Fig. 5B). Neither D134A LapG nor E136A LapGcoprecipitated with LapD in the presence or absence of c-di-GMP,suggesting that mutation of the Ca2�-binding residues disruptsthe c-di-GMP-dependent interaction between LapD and LapG.

Citrate, a potential environmental source of calcium chela-tion. The data presented here demonstrate that removal of Ca2�

by EGTA chelation and mutation of the LapG Ca2�-binding res-

FIG 4 LapG residues D134 and E136 are required for LapG function in vivo and in vitro. (A) The LapG-D134A and -E136A mutants are stable but nonfunctionalin vivo. Shown are biofilm assays (top) of the �lapG mutant strain carrying the vector control or plasmid-expressed WT LapG protein or the LapG-D134A orLapG-E136A mutant protein. The y axis shows the optical density at 550 nm (OD 550) of the extracted crystal violet used to determine the bacterial biofilmbiomass. Western blotting was used to detect the level of His-tagged WT or mutant protein (bottom). (B) Levels and localization of LapA in whole-cell andsupernatant fractions. Shown is a Western blot assay showing the levels of LapA in the whole-cell (Cell) and supernatant (Supe) fractions of the strains indicated.(C) Cell surface levels of LapA. Shown is a representative dot blot assay (top) and quantification of the pixel density of six to eight replicates (bottom) to assessthe levels of cell surface LapA of the strains indicated. (D) Mutant LapG proteins cannot cleave mini-LapA in crude extracts. Shown is a Western blot assaydetecting mini-LapA with anti-Myc antibodies. Full-length mini-LapA ran at �145 kDa, while mini-LapA cleaved by LapG migrated at �130 kDa. Extractsprepared from a �lapG mutant carrying a plasmid expressing mini-LapA were mixed with the extracts indicated above the blot. (E) Purified LapG mutantproteins are not active in vitro. Shown is a LapG activity assay with purified WT or mutant LapG proteins. In this assay, a purified His-tagged portion of the Nterminus of LapA was used as the substrate as described in Materials and Methods. The full-length substrate migrated at �25 kDa, and the cleaved productmigrated at �15 kDa. The LapG protein added to each reaction is indicated above the blot.

LapG is a Ca2�-Dependent Protease

August 2012 Volume 194 Number 16 jb.asm.org 4411

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

idues induce a hyperadherent biofilm and a nonfunctional LapGprotease. These results raise the possibility that absence of Ca2�,potentially due to chelation of this cation, may serve to stabilize abiofilm in the natural environment for this microbe, that is, in soilor on plant roots. We reported previously that citrate, an organicacid and chelating agent (16) stimulated biofilm formation by thelapD mutant (11) in a manner analogous to that of EGTA. Fur-thermore, citrate is found in appreciable levels in the roots androot exudate of plants colonized by P. fluorescens (16, 18). Thus,we hypothesized that citrate may serve as a potential environmen-tal source of calcium chelation. To determine whether an environ-mental and physiologically relevant chelator induced an accumu-lation of biofilm biomass similarly to EGTA, we assessed biofilmformation in the presence of citrate. A study by Kamilova et al.

reported the concentration of citric acid in root exudate of to-mato, cucumber, and sweet pepper plants, and high-performanceliquid chromatography analysis showed that citric acid concentra-tions vary depending upon the plant and the growth substrate.Concentrations ranged from 10 to 110 �g/plant (18). Jones re-ported that the total concentration of organic acids in roots is �10to 20 mM (16), while de Weert et al. reported that “studies ontomato exudate have shown that its major components are or-ganic acids (with citric [55.2%], malic [15.3%], and lactic [10%]acids as major components)” (5). Given the variable concentra-tions of citrate present in the soil and/or in the rhizosphere, weelected to use 0.4% citrate (13.6 mM), the concentration we re-ported in our previous article (11) and a concentration within therange reported for plant roots (16).

Only a minimal increase in biofilm formation by the WT or the�lapG mutant is seen in the presence of citrate. Interestingly, forthese two strains, the small increase in biofilm formation observedwith added citrate is not reduced upon the addition of Ca2�. Wedo not understand the significance of this finding, but it suggeststhat for these two strains, the small increase in biofilm formationobserved upon citrate addition may be Ca2� independent.

A substantial, significant (P � 0.05) increase in biofilm forma-tion is observed in the �lapD mutant in the presence of addedcitrate (Fig. 6A), suggesting that chelation by citrate is responsiblefor increasing biofilm formation. To determine whether the cit-rate-induced hyperadherent biofilm in the �lapD mutant is due toCa2� chelation, biofilm formation was assessed in the presence ofcitrate and CaCl2. As observed with EGTA and CaCl2 (Fig. 2A),CaCl2 reverses the effects of the citrate-induced hyperadherentbiofilm formation in the �lapD mutant, suggesting that citrate isspecifically chelating Ca2�.

Because citrate induces a hyperadherent biofilm in the �lapDmutant, we hypothesized that citrate, like EGTA, would preventLapG cleavage of N-Term-LapA in vitro. In the presence of citrate,LapG is unable to cleave N-Term-LapA, and furthermore, addi-tion of CaCl2 restored LapG cleavage of N-Term-LapA (Fig. 6B).These data suggest that citrate-mediated Ca2� chelation can effec-tively inhibit LapG activity in vitro.

DISCUSSION

The data presented here and in the accompanying report by Chat-terjee and colleagues show that the periplasmic cysteine proteaseLapG is a Ca2�-containing enzyme and furthermore that bindingof this cation is required for LapG activity. This conclusion issupported by chelation studies, genetic analyses, in vitro activityassays, and biochemical studies here, as well as analysis of theLapG structure and in vitro studies in the accompanying report(3). These studies extend our understanding of this relativelypoorly characterized family of DUF920 domain-containing bac-terial proteases.

Interestingly, when LapG is unable to bind Ca2�, it also doesnot interact with LapD in a c-di-GMP-dependent manner. Theseresults suggest that when LapG cannot bind calcium, it also cannotinteract with LapD. Presumably, absence of c-di-GMP-dependentLapD regulation of LapG is inconsequential under these circum-stances, as the LapG calcium point mutants are nonfunctional. Wedo not understand why loss of Ca2� impacts the LapD-LapG in-teraction. It is possible that the site of Ca2� binding and LapDinteraction with LapG are proximal, and thus, the observed localalterations in protein structure upon loss of Ca2� could have sec-

FIG 5 Mutation of LapG residues D134 or E136 to alanine results in loss ofinteraction with LapD. (A) Biofilm assays with strains used for pulldown stud-ies. Shown are representative biofilm wells (top), quantification of the biofilmassay (middle), and levels of LapG and LapD proteins detected by Western blotassay using anti-HA and anti-6H antibodies (bottom). As outlined in the text,the strains have the expected phenotypes based on the data presented in Fig. 4,and LapG-HA and LapD-6H can be detected at similar levels in the strainsexpressing WT or mutant LapG. (B) The LapG-D134A and LapG-E136A mu-tant proteins do not interact with LapD. Shown are Western blot assays fromNi-resin pulldown of LapD-6H using an anti-HA antibody to detect LapG-HA(top) and blots showing the inputs used in these assays (bottom). The strainused to generate the extract used in each lane is indicated above the blot, andthe presence or absence of 5 �M c-di-GMP is indicated.

Boyd et al.

4412 jb.asm.org Journal of Bacteriology

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

ondary effects on LapD binding of LapG. Alternatively, the muta-tions introduced to disrupt Ca2� binding may have global effectson protein structure resulting in loss of LapD interactions withLapG, although the apparent stability of the LapG mutant variantsand the crystallographic analysis of the L. pneumophila orthologargue against gross changes in protein structure. Finally, thisCa2�-dependent interaction with LapD may be an important partof the overall mechanism evolved by P. fluorescens to regulateLapG activity and thus biofilm formation. Until we understandmore about LapD-LapG interactions, it will not be possible todistinguish among these possibilities.

It is important to note that the periplasmic levels of Ca2� re-flect levels in the external environment. Direct measurements ofcalcium in the periplasm of E. coli have been conducted (17). Joneset al. reported that when the external medium contains 0.1 �Mfree Ca2�, Ca2� is concentrated in the periplasm and the cyto-plasm, as the measured concentrations of free Ca2� were 2 to 3�M and 1 �M, respectively. When the external medium con-

tained 6 �M free Ca2�, the free Ca2� level in the periplasm was 6�M, while the cytoplasm contained 1 �M free Ca2�. At submicro-molar concentrations of free Ca2�, the periplasm and cytoplasmare capable of concentrating free Ca2� (17). The authors state thatthese results are consistent with the outer membrane being morepermeable to Ca2� influx than the inner membrane. Thus, fromthe studies by Jones et al., it is apparent that Ca2� concentrationsfluctuate in the cell in response to the external concentration ofCa2�. Finally, while we do not know the concentration of Ca2�

available in our laboratory medium, there is clearly a sufficientamount of this metal present to allow LapG function, as we ob-serve LapG-mediated detachment of biofilms under these labora-tory growth conditions.

The finding that citrate, an organic chelator found in nichesoccupied by P. fluorescens, could impact LapG activity via Ca2�

chelation suggested the very exciting possibility of an additionalenvironmentally relevant mechanism to regulate biofilm forma-tion by this microbe. It is not clear whether citrate is chelatingCa2� specifically in the soil environment, but in addition toEGTA, this environmentally relevant chelator does exert an effecton biofilm formation and LapG activity, suggesting that the effectsobserved here for Ca2� chelation are not limited to EGTA. Thispoint is especially intriguing given that the plant root and rootexudate environment in which P. fluorescens forms biofilms con-tains many organic acids (5, 16) and that the biofilm matrix isknown to be composed of several components, such as extracel-lular polysaccharide and extracellular DNA, which have substan-tial Ca2�-chelating activity (7, 25, 36). Thus, exploring the possi-bility that environmentally relevant organic acids and the biofilmmatrix may inhibit cellular functions via the chelation of impor-tant metal ions, such as calcium, which contribute to biofilm dis-persal is an interesting line for future experimentation.

ACKNOWLEDGMENTS

We thank P. D. Newell for thoughtful discussions and Thomas Hamptonfor assistance with statistical analysis.

This work was supported by the National Institutes of Health (T32-GM08704 predoctoral fellowship to C.D.B., R01-GM081373 to H.S., andR01 AI097307-01 to H.S. and G.A.O.), a National Science Foundationgrant (MCB-9984521 to G.A.O.), and a PEW scholar award in BiomedicalSciences (to H.S.).

REFERENCES1. Alonso-GarcíA N, Ingles-Prieto A, Sonnenberg A, de Pereda JM. 2009.

Structure of the Calx-� domain of the integrin �4 subunit: insights intofunction and cation-independent stability. Acta Crystallogr. D Biol. Crys-tallogr. 65:858 – 871.

2. Arrizubieta MJ, Toledo-Arana A, Amorena B, Penades JR, Lasa I. 2004.Calcium inhibits Bap-dependent multicellular behavior in Staphylococcusaureus. J. Bacteriol. 186:7490 –7498.

3. Chatterjee D, Boyd CD, O’Toole GA, Sondermann H. 2012. Structuralcharacterization of a conserved, calcium-dependent periplasmic proteasefrom Legionella pneumophila. J. Bacteriol. 194:4415– 4425.

4. Dawson RMC, Elliott DC, Elliott WH, Jones KM. 1986. Data for bio-chemical research. Clarendon Press, Oxford, United Kingdom.

5. de Weert S, et al. 2002. Flagella-driven chemotaxis towards exudatecomponents is an important trait for tomato root colonization by Pseu-domonas fluorescens. Mol. Plant Microbe Interact. 15:1173–1180.

6. Dominguez DC. 2004. Calcium signalling in bacteria. Mol. Microbiol.54:291–297.

7. Fang Y, et al. 2007. Multiple steps and critical behaviors of the binding ofcalcium to alginate. J. Phys. Chem. B. 111:2456 –2462.

8. Geesey GG, Wigglesworth-Cooksey B, Cooksey KE. 2000. Influence ofcalcium and other cations on surface adhesion of bacteria and diatoms: areview. Biofouling 15:195–205.

FIG 6 Citrate chelation stimulates biofilm formation. (A) Impact of addedcitrate on biofilm formation. Shown are biofilm assays of the WT and the�lapG, �lapD, and lapA mutants grown in K10T-1 medium or in this mediumsupplemented with CaCl2 (500 �M), citrate (0.4%), or both. The y axis showsthe optical density at 550 nm (OD 550) of the extracted crystal violet used todetermine the bacterial biofilm biomass of the indicated strains. (B) Citrateinhibits LapG activity in vitro. Shown is a Western blot developed with anti-bodies directed against purified, 6H-tagged N-Term-LapA. N-Term-LapAcontains the LapG cleavage site mapped to the N terminus of LapA as describedpreviously (28). Cleavage of N-Term-LapA by LapG results in a shift of thesubstrate from �25 kDa to �15 kDa. Under the conditions used here, LapGwas able to cleave �75% of the substrate. The addition to each assay mixture isindicated above the lane. CaCl2 was added to 500 �M, and citrate was added to0.4%.

LapG is a Ca2�-Dependent Protease

August 2012 Volume 194 Number 16 jb.asm.org 4413

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

9. Ginalski K, Kinch L, Rychlewski L, Grishin NV. 2004. BTLCP proteins:a novel family of bacterial transglutaminase-like cysteine proteinases.Trends Biochem. Sci. 29:392–395.

10. Haas D, Defago G. 2005. Biological control of soil-borne pathogens byfluorescent pseudomonads. Nat. Rev. Microbiol. 3:307–319.

11. Hinsa SM, O’Toole GA. 2006. Biofilm formation by Pseudomonas fluo-rescens WCS365: a role for LapD. Microbiology 152:1375–1383.

12. Hinsa SM, Espinosa-Urgel M, Ramos JL, O’Toole GA. 2003. Transitionfrom reversible to irreversible attachment during biofilm formation byPseudomonas fluorescens WCS365 requires an ABC transporter and a largesecreted protein. Mol. Microbiol. 49:905–918.

13. Holland IB, Jones HE, Campbell AK, Jacq A. 1999. An assessment of therole of intracellular free Ca2� in E. coli. Biochimie 81:901–907.

14. Jackson DW, Simecka JW, Romeo T. 2002. Catabolite repression ofEscherichia coli biofilm formation. J. Bacteriol. 184:3406 –3410.

15. Jackson DW, et al. 2002. Biofilm formation and dispersal under theinfluence of the global regulator CsrA of Escherichia coli. J. Bacteriol. 184:290 –301.

16. Jones DL. 1998. Organic acids in the rhizosphere—a critical review. PlantSoil 205:25– 44.

17. Jones HE, Holland IB, Campbell AK. 2002. Direct measurement of freeCa2� shows different regulation of Ca2� between the periplasm and thecytosol of Escherichia coli. Cell Calcium 32:183–192.

18. Kamilova F, et al. 2006. Organic acids, sugars, and L-tryptophane inexudates of vegetables growing on stonewool and their effects on activitiesof rhizosphere bacteria. Mol. Plant Microbe Interact. 19:250 –256.

19. Lugtenberg BJ, Kravchenko LV, Simons M. 1999. Tomato seed and rootexudate sugars: composition, utilization by Pseudomonas biocontrolstrains and role in rhizosphere colonization. Environ. Microbiol. 1:439 –446.

20. MacEachran DP, Stanton BA, O’Toole GA. 2008. Cif is negatively reg-ulated by the TetR family repressor CifR. Infect. Immun. 76:3197–3206.

21. Matsuoka S, et al. 1995. Regulation of the cardiac Na�-Ca2� exchangerby Ca2�. Mutational analysis of the Ca2�-binding domain. J. Gen. Physiol.105:403– 420.

22. Michiels J, Xi C, Verhaert J, Vanderleyden J. 2002. The functions ofCa2� in bacteria: a role for EF-hand proteins? Trends Microbiol. 10:87–93.

23. Monds RD, Newell PD, Gross RH, O’Toole GA. 2007. Phosphate-dependent modulation of c-di-GMP levels regulates Pseudomonas fluore-scens Pf0-1 biofilm formation by controlling secretion of the adhesinLapA. Mol. Microbiol. 63:656 – 679.

24. Monds RD, Newell PD, Schwartzman JA, O’Toole GA. 2006. Conser-vation of the Pho regulon in Pseudomonas fluorescens Pf0-1. Appl. Envi-ron. Microbiol. 72:1910 –1924.

25. Mulcahy H, Charron-Mazenod L, Lewenza S. 2008. Extracellular DNAchelates cations and induces antibiotic resistance in Pseudomonas aerugi-nosa biofilms. PLoS Pathog. 4:e1000213. doi:10.1371/journal.p-pat.1000213.

26. Navarro MV, et al. 2011. Structural basis for c-di-GMP-mediated inside-out signaling controlling periplasmic proteolysis. PLoS Biol. 9:e1000588.doi:10.1371/journal.pbio.1000588.

27. Newell PD, Monds RD, O’Toole GA. 2009. LapD is a bis-(3=,5=)-cyclic

dimeric GMP-binding protein that regulates surface attachment by Pseu-domonas fluorescens Pf0-1. Proc. Natl. Acad. Sci. U. S. A. 106:3461–3466.

28. Newell PD, Boyd CD, Sondermann H, O’Toole GA. 2011. A c-di-GMPeffector system controls cell adhesion by inside-out signaling and surfaceprotein cleavage. PLoS Biol. 9:e1000587. doi:10.1371/journal.p-bio.1000587.

29. Newell PD, Yoshioka S, Hvorecny KL, Monds RD, O’Toole GA. 2011.Systematic analysis of diguanylate cyclases that promote biofilm forma-tion by Pseudomonas fluorescens Pf0-1. J. Bacteriol. 193:4685– 4698.

30. Norris V, et al. 1991. Calcium in bacteria: a solution to which problem?Mol. Microbiol. 5:775–778.

31. Norris V, et al. 1996. Calcium signalling in bacteria. J. Bacteriol. 178:3677–3682.

32. O’Toole GA, Kolter R. 1998. Initiation of biofilm formation in Pseu-domonas fluorescens WCS365 proceeds via multiple, convergent signallingpathways: a genetic analysis. Mol. Microbiol. 28:449 – 461.

33. O’Toole GA, Gibbs KA, Hager PW, Phibbs PV, Jr, Kolter R. 2000. Theglobal carbon metabolism regulator Crc is a component of a signal trans-duction pathway required for biofilm development by Pseudomonasaeruginosa. J. Bacteriol. 182:425– 431.

34. Prigent-Combaret C, Vidal O, Dorel C, Lejeune P. 1999. Abiotic surfacesensing and biofilm-dependent regulation of gene expression in Esche-richia coli. J. Bacteriol. 181:5993– 6002.

35. Prigent-Combaret C, et al. 2001. Complex regulatory network controlsinitial adhesion and biofilm formation in Escherichia coli via regulation ofthe csgD gene. J. Bacteriol. 183:7213–7223.

36. Sarkisova S, Patrauchan MA, Berglund D, Nivens DE, Franklin MJ.2005. Calcium-induced virulence factors associated with the extracellularmatrix of mucoid Pseudomonas aeruginosa biofilms. J. Bacteriol. 187:4327– 4337.

37. Sauer K, et al. 2004. Characterization of nutrient-induced dispersion inPseudomonas aeruginosa PAO1 biofilm. J. Bacteriol. 186:7312–7326.

38. Schwarz EM, Benzer S. 1997. Calx, a Na-Ca exchanger gene of Drosophilamelanogaster. Proc. Natl. Acad. Sci. U. S. A. 94:10249 –10254.

39. Shanks RM, Caiazza NC, Hinsa SM, Toutain CM, O’Toole GA. 2006.Saccharomyces cerevisiae-based molecular tool kit for manipulation ofgenes from Gram-negative bacteria. Appl. Environ. Microbiol. 72:5027–5036.

40. Singh PK, Parsek MR, Greenberg EP, Welsh MJ. 2002. A component ofinnate immunity prevents bacterial biofilm development. Nature 417:552–555.

41. Smith RJ. 1995. Calcium and bacteria. Adv. Microb. Physiol. 37:83–133.42. Theunissen S, et al. 2010. The 285 kDa Bap/RTX hybrid cell surface

protein (SO4317) of Shewanella oneidensis MR-1 is a key mediator ofbiofilm formation. Res. Microbiol. 161:144 –152.

43. Thormann KM, et al. 2006. Control of formation and cellular detach-ment from Shewanella oneidensis MR-1 biofilms by cyclic di-GMP. J. Bac-teriol. 188:2681–2691.

44. Tsien RY. 1980. New calcium indicators and buffers with high selectivityagainst magnesium and protons: design, synthesis, and properties of pro-totype structures. Biochemistry 19:2396 –2404.

Boyd et al.

4414 jb.asm.org Journal of Bacteriology

on Septem

ber 12, 2020 by guesthttp://jb.asm

.org/D

ownloaded from

Structural Characterization of a Conserved, Calcium-DependentPeriplasmic Protease from Legionella pneumophila

Debashree Chatterjee,a Chelsea D. Boyd,b George A. O’Toole,b and Holger Sondermanna

Department of Molecular Medicine, College of Veterinary Medicine, Cornell University, Ithaca, New York, USA,a and Department of Microbiology and Immunology, GeiselSchool of Medicine at Dartmouth, Hanover, New Hampshire, USAb

The bacterial dinucleotide second messenger c-di-GMP has emerged as a central molecule in regulating bacterial behavior, in-cluding motility and biofilm formation. Proteins for the synthesis and degradation of c-di-GMP and effectors for its signal trans-mission are widely used in the bacterial domain. Previous work established the GGDEF-EAL domain-containing receptor LapDas a central switch in Pseudomonas fluorescens cell adhesion. LapD senses c-di-GMP inside the cytosol and relays this signal tothe outside by the differential recruitment of the periplasmic protease LapG. Here we identify the core components of an or-thologous system in Legionella pneumophila. Despite only moderate sequence conservation at the protein level, key features con-cerning the regulation of LapG are retained. The output domain of the LapD-like receptor from L. pneumophila, CdgS9, bindsthe LapG ortholog involving a strictly conserved surface tryptophan residue. While the endogenous substrate for L. pneumophilaLapG is unknown, the enzyme processed the corresponding P. fluorescens substrate, indicating a common catalytic mechanismand substrate recognition. Crystal structures of L. pneumophila LapG provide the first atomic models of bacterial proteases ofthe DUF920 family and reveal a conserved calcium-binding site important for LapG function.

Bacteria sense and respond to their environment by a multitude ofphysiological programs, allowing them to adapt to changing and

often hostile conditions. Biofilm formation is one such mechanismthat is used widely by many pathogenic and environmental bacteria(16). c-di-GMP, a molecule unique to bacteria, has emerged as animportant intracellular second messenger that regulates the forma-tion of biofilms at multiple levels (17). The majority of the bacterialgenomes sequenced to date encode enzymes for the production andturnover of c-di-GMP, diguanylate cyclases with GGDEF domains,and phosphodiesterases with EAL or HD-GYP domains, respectively(12). Receptors for c-di-GMP are a less-well-defined group that in-cludes receiver domains in transcription factors, PilZ domain-con-taining proteins, riboswitches, and proteins with catalytically inactiveGGDEF or EAL domains constituting a distinct class (36). These pro-teins exploit their degenerate active sites or regulatory c-di-GMP-binding sites to sense the cellular concentration of the dinucleotideand to solicit a specific response.

One such receptor, the transmembrane protein LapD fromPseudomonas fluorescens, contains both catalytically inactiveGGDEF and EAL domains, with the latter serving as the exclusivec-di-GMP binding module (31). Previous work established LapDas a prototypical receptor for mediating inside-out signaling fromthe cytosol to the periplasm. LapD responds to a rise in c-di-GMPlevels triggered by a nutritional signal, namely, the availability ofPi, by undergoing a conformational change from a nucleotide-free, autoinhibited state to a c-di-GMP-activated state (19, 28–30)(Fig. 1A). Only the activated state is capable of sequestering aperiplasmic protease, LapG, to LapD’s output domain which, inturn, stabilizes LapG’s substrate, LapA (30). LapA is a large adhe-sin protein that is embedded in the outer membrane by an un-known mechanism (18). Proteolytic processing by LapG releasesLapA from the membrane, leading to biofilm dispersion when Pi islimiting (30). In conjunction with revealing the signaling and out-put system for biofilm formation in P. fluorescens, we determinedthe crystal structures of the functional domains of LapD spanningalmost the entire receptor (29). These studies revealed the main

regulatory features controlling LapD function. Based on a bioin-formatic analysis, it also allowed us to predict orthologous signal-ing systems in many other bacteria, which may hinge on a LapD-like c-di-GMP receptor and a LapG-like protease. Two of thosehave been characterized independently in Pseudomonas putidaand Shewanella oneidensis (14, 37).

Another orthologous system that we predicted, based onbioinformatics, is that of Legionella pneumophila, the causativeagent of Legionnaires’ disease (29). Unlike Pseudomonas, L. pneu-mophila is a facultative intracellular bacterium that is also able togrow in biofilms. Its genome encodes 21 predicted proteins withGGDEF and/or EAL domains and a single PilZ protein, and asubset of these have been shown to impact intracellular growth,motility, or biofilm formation (5, 25, 34). However, the underly-ing signaling mechanisms and networks and their regulation arelargely unknown for the majority of these proteins.

In this study, we focused on the periplasmic protease LapG ofL. pneumophila. LapG belongs to the domain of unknown func-tion 920 (DUF920) (or COG3672) family, and hence, little isknown regarding its molecular mechanism. Sequence and foldrecognition methods classified these proteins as bacterial trans-glutaminase-like cysteine proteases (BTLCPs) and predicted acysteine-histidine-aspartate (C-H-D) catalytic triad and corestructural motifs at the active site (13). Recently, O’Toole andcolleagues corroborated the notion that LapG functions as a cys-teine protease (30). In an effort to further our mechanistic under-standing of LapG and related proteases, we determined crystal

Received 16 April 2012 Accepted 8 June 2012

Published ahead of print 15 June 2012

Address correspondence to Holger Sondermann, hs293@cornell.edu.

Supplemental material for this article may be found at http://jb.asm.org/.

Copyright © 2012, American Society for Microbiology. All Rights Reserved.

doi:10.1128/JB.00640-12

August 2012 Volume 194 Number 16 Journal of Bacteriology p. 4415–4425 jb.asm.org 4415

structures of the LapG ortholog from L. pneumophila. In a parallelstudy, the O’Toole group noted a dependence of P. fluorescensLapG activity on calcium ions (4), and the structures allowed us toidentify a strictly conserved calcium-binding site in LapG andBTLCPs. In addition, we demonstrate that the L. pneumophilaLapD-LapG system utilizes an output mechanism similar to thatwhich we previously described in P. fluorescens (29, 30).

MATERIALS AND METHODSProtein expression and purification. DNA fragments encoding L. pneu-mophila LapG lacking the signal peptide (lpg0828; residues 52 to 244) andthe periplasmic output domain of the LapD ortholog CdgS9 (lpg0829;residues 22 to 152) were amplified from genomic DNA by standard PCRand cloned into a bacterial expression vector based on pET28a (Novagen)that adds an N-terminal, cleavable His6-SUMO tag (see Table S1 in thesupplemental material).

Native and selenomethionine-derivatized proteins were overex-pressed in Escherichia coli BL21 T7 Express or T7 Express Crystal cells(New England BioLabs), respectively. For the expression of native pro-teins, cultures were grown at 37°C in terrific broth medium supplementedwith 50 �g/ml kanamycin. At an optical density at 600 nm (OD600) of �1,the temperature was reduced to 18°C and protein expression was inducedby adding 1 mM IPTG. Selenomethionine-derivatized proteins were ex-pressed in cells grown at 37°C in M9 minimal medium supplemented with50 �g/ml kanamycin, vitamins (1 �g/ml thiamine and 1 �g/ml biotin), a

carbon source (0.4% glucose), trace elements, and amino acids (each ofthe 20 amino acids at 40 �g/ml, with selenomethionine substituting formethionine). Protein expression was induced at an OD600 of 0.4 to 0.5. Inboth cases, protein expression proceeded for 16 h at 18°C, after which cellswere harvested by centrifugation, resuspended in Ni-nitrilotriacetic acid(NTA) buffer A (25 mM Tris-HCl [pH 8.5], 500 mM NaCl, 20 mM imi-dazole), and flash frozen in liquid nitrogen. Cell suspensions were thawedand lysed by sonication. Cell debris was removed by centrifugation, andthe clear lysates were incubated with Ni-NTA resin (Qiagen) that waspreequilibrated with Ni-NTA buffer A. The resin was washed with 20column volumes of buffer A, followed by protein elution with 5 columnvolumes of Ni-NTA buffer B (25 mM Tris-HCl [pH 8.5], 500 mM NaCl,300 mM imidazole). The eluted proteins were buffer exchanged into alow-salt buffer (25 mM Tris-HCl [pH 8.5], 150 mM NaCl) on a fastdesalting column (GE Healthcare). Proteins were subjected to size exclu-sion chromatography on a Superdex 200 column (GE Healthcare) pre-equilibrated with gel filtration buffer (25 mM Tris-HCl [pH 8.5], 150 mMNaCl). Where indicated, the His6-SUMO moiety was cleaved off by usingthe yeast protease Ulp-1 following desalting. Ulp-1, uncleaved protein,and the cleaved fusion tags were removed by Ni-NTA affinity chromatog-raphy prior to the final gel filtration. Purified proteins were concentratedon Amicon filters with an appropriate size cutoff to concentrations of �25mg/ml, flash frozen in liquid nitrogen, and stored at �80°C.

The expression and purification of the corresponding proteins from P.fluorescens were described previously (29). The construction, expression,and purification of P. fluorescens LapANterm were described elsewhere(30). Site-directed mutagenesis was carried out by using the QuikChangekit (Agilent Technologies) and following the manufacturer’s instructions,followed by validation through DNA sequencing.

Crystallization, data collection, and structure solution. Crystalswere obtained by hanging-drop vapor diffusion mixing equal volumes ofprotein (10 to 30 mg/ml) and reservoir solution, followed by incubation at4°C. For the native, apo-state crystal form, the reservoir solution con-tained 0.14 M ammonium tartrate dibasic and 20% polyethylene glycol3350. The selenomethionine-derivatized protein crystallized in the samecondition supplemented with 0.1 M Bis-tris (pH 7.0). Crystals for calci-um-bound and EGTA-treated LapG were obtained in the same conditionas the apoprotein, supplemented with 2 mM CaCl2 and EGTA, respec-tively. For cryoprotection, crystals were soaked in reservoir solutionsupplemented with 25% xylitol. Cryopreserved crystals were flash fro-zen and stored in liquid nitrogen. For optimal diffraction, crystals ofLapG-Ca2� were grown at 20°C and transferred to 4°C for 1 h prior tocryoprotection and freezing. Data on frozen crystals were collected at100 K using synchrotron radiation at the Cornell High Energy Syn-chrotron Source (CHESS; Cornell University, Ithaca, NY).

Data reduction was carried out with the software package HKL2000(33). Experimental phases for the initial structure determination wereobtained from single-wavelength anomalous diffraction (SAD) experi-ments with crystals grown from selenomethionine-derivatized proteinsby using the software package PHENIX (2). Refinement in PHENIX andCOOT (11) yielded the final models. Data collection and refinement sta-tistics are summarized in Table S2 in the supplemental material. Illustra-tions were made in Pymol (Schrödinger). Alignments were generated us-ing ClustalW2 (24) and formatted with ESPript (15). Sequence logos weregenerated using WebLogo (8, 35).

Protein pulldown assay. His6-SUMO-tagged L. pneumophila LapG orHis6-tagged P. fluorescens LapG was incubated with Ni-NTA resin (Qia-gen) for 1 h at 4°C in binding buffer (25 mM Tris-HCl [pH 8.5], 75 mMNaCl, 25 mM KCl, 40 mM imidazole). Following the removal of unboundprotein by three washing steps with 5 column volumes of binding buffereach, LapG-bound resin (corresponding to �50 �g of protein) was incu-bated with an excess of the untagged output domain (250 �g or 20 �M) ofeither L. pneumophila or P. fluorescens LapD for 30 min at 4°C. After theresin was washed three times with 5 column volumes of binding buffer,proteins were eluted from the resin with elution buffer (25 mM Tris-HCl

FIG 1 The LapADG signaling system. (A) Model of LapD-mediated regula-tion of biofilm formation in P. fluorescens. The c-di-GMP receptor LapD lo-calizes to the inner membrane (IM), where it senses cytoplasmic c-di-GMPlevels. It controls the stability of the large adhesin LapA in the outer membrane(OM) by sequestration of the periplasmic protease LapG. The underlying sig-naling pathway is controlled by the availability of Pi. (B) Genetic map of theLapD ortholog-containing operon in L. pneumophila.

Chatterjee et al.

4416 jb.asm.org Journal of Bacteriology

[pH 8.5], 500 mM NaCl, 300 mM imidazole). Eluates were analyzed usingstandard denaturing SDS-PAGE and Coomassie staining. All incubationswere carried out under gentle agitation in spin columns.

LapA cleavage assay. Purified P. fluorescens LapG variants (wild typeor D136A; 0.4 �M) and L. pneumophila LapG variants (wild type, D136A,E138A, or D139A; 40 to 150 �M) were incubated at the indicated concen-trations with purified LapANterm (2 �M) in reaction buffer (25 mM Tris-HCl [pH 8.5], 150 mM NaCl, 20 mM MgCl2) overnight at room temper-ature in the presence or absence of 10 mM EGTA. The reaction productswere separated by SDS-PAGE and analyzed by Western blotting using amonoclonal antibody raised against pentahistidine (Qiagen) which wasdetected by a horseradish peroxidase-coupled anti-mouse antibody. Blotswere developed by using enhanced chemiluminescence (GE Healthcare)and exposed to film.

SEC-MALS. Size exclusion chromatography-coupled multiangle lightscattering (SEC-MALS) measurements were carried out by injecting pu-rified proteins (100 �M) onto a WTC-030S5 gel filtration column (WyattTechnology) preequilibrated with gel filtration buffer (25 mM Tris-HCl[pH 8.5], 150 mM NaCl). The SEC system was coupled to an 18-angle,static light scattering detector and a refractive index detector (DAWNHELEOS-II and Optilab T-rEX, respectively; Wyatt Technology). Datawere collected at 25°C every second at a flow rate of 1 ml/min and analyzedwith the software ASTRA, yielding the molecular mass and mass distribu-tion (polydispersity) of the samples. For data quality control and normal-ization of the light scattering detectors, monomeric bovine serum albu-min (Sigma) was used.

ITC. Isothermal titration calorimetry (ITC) was used to determine theapparent dissociation constants (Kd) and stoichiometry of interactionsusing a VP calorimeter (Microcal). Calorimetric titrations of calcium (2mM in the syringe; 10-�l injections) and wild-type or mutant L. pneumo-phila LapG (200 �M in the cuvette) were carried out at 20°C in gel filtra-tion buffer (25 mM Tris-HCl [pH 8.5], 150 mM NaCl) with a delay of 300s between injections. The data obtained were analyzed by integrating heateffects normalized to the amount of injected protein and curve fittingbased on a single-site binding model by using the Origin software package(Microcal).

Protein structure accession numbers. Coordinates and structure fac-tors have been deposited in the Protein Data Bank (PDB) and assignedaccession numbers 4FGQ, 4FGP, and 4FGO.

RESULTSThe Lap operon in L. pneumophila. We previously predicted theexistence of proteins with sequence similarity to P. fluorescensLapD and LapG in several bacterial species, including L. pneumo-phila (29). In silico genomic analysis indicates that both genes mapto an operon containing at least five genes (Fig. 1B). It encodes apredicted type I secretion outer membrane protein (lpg0827/TolC), a LapG-like protease (lpg0828), a LapD ortholog with adegenerate GGDEF-EAL domain module (lpg0829/CdgS9), andtwo putative proteins (lpg0830, predicted thioesterase/lipase ac-tivity; lpg0831, predicted flavin-containing monooxygenase).While the functional relevance of the latter two gene productswithin this cluster remains to be established, a type I secretionsystem is required in P. fluorescens for the translocation of theLapG substrate to the outer membrane (18), and L. pneumophilaTolC may fulfill a similar function. Considering the limited mech-anistic characterization of DUF920-containing proteins and theimportance of LapG as a part of a c-di-GMP-dependent signalingsystem (29, 30), we set out to determine the molecular mechanismand structure of a LapG ortholog.

Conservation of the LapD-LapG interaction in L. pneumo-phila. To establish that the LapG and LapD orthologs indeed forma complex as part of a regulatory system, we cloned, expressed, and

purified the periplasmic output domain of L. pneumophila CdgS9(the LapD ortholog) and LapG. Both proteins were expressedmost stably with cleavable, N-terminal hexahistidine-SUMO(His6-SUMO) tags. For comparison, we used the respective pro-tein constructs from P. fluorescens, with the exception that LapGcontained a C-terminal His6 tag instead of the His6-SUMO tag(29). We previously demonstrated that the interaction between P.fluorescens LapG and LapD relies on a strictly conserved, surface-exposed tryptophan residue that is present at the tip of a beta-hairpin motif in LapD’s output domain (29) (Fig. 2A). Mutationof this residue (e.g., to alanine or glutamate) abolished LapG bind-ing and signaling through LapD, serving as an invaluable specific-ity control.

His6-tagged (or His6-SUMO-tagged) LapG orthologs werebound to Ni-NTA Sepharose, washed, and incubated with thepurified, untagged output domain of LapD or CdgS9, respectively.Proteins were eluted and analyzed by SDS-PAGE. Despite overalllow sequence conservation of LapD orthologs (�23% across theentire receptor), P. fluorescens LapG adsorbed not only the cog-nate LapD output domain but also equally efficiently the corre-sponding domain of the L. pneumophila ortholog (Fig. 2B). Asimilar complex formation by L. pneumophila LapG and the cor-responding LapD output domain was observed. Binding of L.pneumophila LapG to the output domain of P. fluorescens LapDwas detectable but weaker. All interactions were sensitive to a non-conservative (W-to-E) mutation at the critical tryptophan residueat the center of the output domain beta-hairpin motif, indicatingthat the mode of binding is specific and conserved across distantlyrelated bacterial species (Fig. 2B).

The crystal structure of L. pneumophila LapG. The LapG or-tholog from L. pneumophila (residues 52 to 244; lacks the signalpeptide) was expressed in E. coli as a soluble protein and purifiedby standard liquid chromatography. Upon crystallization (spacegroup P21212; two molecules per asymmetric unit), the high-res-olution structure was determined by SAD phasing with selenome-thionine-substituted protein crystals (Fig. 3A; see Table S2 in thesupplemental material).

The structure reveals a bilobal fold of LapG (Fig. 3A). TheN-terminal lobe is formed by five �-helices folding into a globularstructure. In contrast, the C-terminal lobe consists of a five-stranded anti-parallel �-sheet. The three central strands are longerthan the two flanking strands. The extreme C terminus folds intoa helix that is connected to the bulk of the protein by a flexiblelinker and is buttressed by the N-terminal lobe via largely hydro-phobic interactions. The strictly conserved active site, the C-H-Dcatalytic triad (C137, H172, D189; Fig. 3B), is located at the interfacebetween the two halves of the protein, with the histidine and as-partate residues being contributed by the C-terminal lobe and thecatalytic cysteine residue by the central helix �5 of the N-terminallobe (Fig. 3A). The hydroxyl group of a serine residue (S190) pointstoward the active site and engages in a hydrogen bond with D189 ofthe catalytic triad.

The catalytic triad is equally conserved within the LapG sub-group and all BTLCPs. The residue corresponding to position 190in L. pneumophila LapG can be either a serine or an asparagineresidue in the LapG subset, as well as the wider BTLCP family (Fig.3B). In order to more globally visualize the conservation of LapG-type proteases, we mapped the conservation scores of 24 LapGorthologs onto the accessible surface of the protease fold.Orthologs from distantly related species, including Pseudomonas,

Crystal Structure of L. pneumophila LapG

August 2012 Volume 194 Number 16 jb.asm.org 4417

Legionella, and Vibrio species, were used to create the alignmentfor this analysis (see Fig. S1 in the supplemental material). Inter-estingly, not only is the catalytic triad strictly conserved, but wealso noted a fairly conserved surface patch extending from theactive site (Fig. 3C). While the functional relevance remains to beestablished, the hydrophobic nature of this region may suggest arole as an interaction interface, for example, for substrate binding,considering its close proximity to the active site.

A structural role can be attributed to several hydrophilic resi-dues. Consistent with the bioinformatic and modeling study ofBTLCPs (13), in addition to the invariant catalytic triad, there areseveral conserved, polar, or charged residues that form a hydrogenbond network stabilizing some of the core secondary structureelements adjacent to the active site (N91, N95, K130, and N173 inBTLCP; N102, N106, K144, and the aforementioned S190 in L. pneu-mophila LapG) (Fig. 4). Residues with similar function but morespecific to the LapG subfamily of BTLCPs are R201 and D203 (Fig.4B), located in the C-terminal lobe (Fig. 4A). Positioned by D203,R201 coordinates D189 of the catalytic triad. Together, these resi-dues are part of the hydrogen bond network that coordinates D189

at the active site.A comparison against the entire PDB using the DALI server

(20) was used to identify structurally related proteins. With thestructure of L. pneumophila LapG as the search model (Fig. 5A),we identified eukaryotic protein transglutaminases as some of theclosest structural neighbors (Fig. 5B; PDB codes 1g0d, 1kv3, 1ggt,and 1nud) (3, 27, 32, 40). In addition, the search identified anarylamine N-acetyltransferase (Fig. 5C; PDB code 2bsz) (21) andputative bacterial cysteine proteases (Fig. 5D and E; PDB codes3isr and 3kd4) with deposited structures but no associated publi-

cation as proteins that contain a similar fold. All but one protein(the putative bacterial protease with PDB code 3kd4) display acatalytic triad that is conserved in sequence (C-H-D) and positionrelative to that of LapG, with the cysteine residue being located atthe tip of the central helix (Fig. 5). Notably, the activity of severaltransglutaminases depends on the presence of calcium ions (1,41), and calcium-binding sites have been identified in a subset ofcrystal structures and by modeling approaches (e.g., PDB code1nud; Fig. 5B) (3, 6, 9). These studies identified three distinctcalcium-binding sites in the human transglutaminase 3 enzyme,with site 1 being located adjacent to the active site. For transglu-taminase, this site has been proposed to stabilize the enzyme, yet asimilar role in LapG could not be established. However, P. fluore-scens LapG is sensitive to EGTA treatment and requires calciumions for the specific cleavage of its substrate, LapA (see the accom-panying report [4]), suggesting a crucial function during catalysis.

Crystal structures of calcium-bound and EGTA-treatedLapG. Incited by the observation that P. fluorescens LapG’s activitydepends on calcium (4) yet no bound calcium ion was apparent inthe crystal structure, we initially used pattern prediction algo-rithms with the apo-LapG structure as the input. The MUG calci-um-binding site prediction server (38) identified a patch of fourconserved, negatively charged residues adjacent to the catalytictriad that had the potential to accommodate a calcium ion (Fig.6A). While not identical in exact position, the predicted site is asclose to the active site as is calcium-binding site 1, which wasfound to be crucial for transglutaminase function (Fig. 5B). InLapG, three of the residues, D136, E138, and D139, flank C137 of thecatalytic triad, whereas the fourth residue (D120) is less conservedand farther away with regard to the primary sequence (Fig. 6A).

FIG 2 Conserved interaction of LapD and LapG. (A) Sequence conservation of a loop in the periplasmic output domain of LapD (PDB code 3PJV) that is criticalfor LapG interaction in P. fluorescens. (B) Interaction between LapD’s output domain with LapG. Purified, His6-tagged LapG was bound to Ni-NTA andincubated in the absence or presence of the untagged LapD output domain. Proteins from P. fluorescens (Pfl) or L. pneumophila (Lpg) were used. A specific outputdomain mutation (W125E in P. fluorescens LapD; W126E in L. pneumophila LapD) was included. Eluted complexes were analyzed by SDS-PAGE, followed byCoomassie staining.

Chatterjee et al.

4418 jb.asm.org Journal of Bacteriology

However, other aspartate or glutamate residues are located inclose proximity to D120 and could function redundantly (Fig. 6A).While both the N- and C-terminal lobes contribute residues to thecatalytic triad, the calcium-binding site is located entirely in theN-terminal lobe. This motif is not unique to LapG-like proteinsbut appears to be conserved in the entire BTLCP family on thebasis of sequence alignments (Fig. 6A), indicating a general mo-lecular mechanism of these proteases.

To confirm the prediction, we crystallized L. pneumophilaLapG in the presence of calcium ions, yielding a new crystal form.We solved the structure by SAD phasing (space group P43212; onemolecule per asymmetric unit; see Table S2 in the supplementalmaterial). While the maximum resolution was 1.9 Å, crystals dif-fracted X-rays anisotropically with a resolution of �2.8 Å in theworst orientation. This observation is consistent with poor pack-ing interactions along one crystal axis and with the two N-termi-nal helices being poorly resolved (with high B factors for residues58 to 88; data not shown).

The overall fold of LapG in the new crystal form was preserved(see Fig. S2 in the supplemental material; root mean square devi-ation [RMSD] with or without calcium of 0.6 Å). We observedclear density around the residues predicted to form the calcium-binding site (Fig. 6B, inset). A model with a calcium ion at that siterefined well. Furthermore, the coordination of the ion by the pro-tein side chains is in good agreement with other calcium bindingmotifs, revealing seven ionic or polar interactions, including onewater-mediated contact and a carbonyl backbone contact with

Y122 (Fig. 6B). To accomplish this, only the side chain of D120 hadto adopt an alternative rotamer conformation, flipping toward thebinding site, while the other three residues are essentially in thesame conformation as in the apo-state structure. We noticed onlya small shift of the D120-presenting loop relative to the body ofLapG. However, we cannot distinguish whether this change is cal-cium induced or due to the altered packing interactions in thiscrystal form.

We also crystallized LapG treated with EGTA (see Table S2 inthe supplemental material). This protein preparation crystallizedin the original condition. While we obtained crystals within thesame space group as LapG without EGTA treatment, it also crys-tallized in a second space group under identical conditions andsimilar crystal packing. We solved the structure by molecular re-placement using the first LapG crystal structure as the searchmodel. The structure of the EGTA-treated monomer is almostidentical to the initial calcium-free state (RMSD of 0.12 Å; see Fig.S2 in the supplemental material) and very similar to the calcium-bound state despite differences in crystal contacts.

Calcium binding by L. pneumophila LapG. To further vali-date and quantify calcium binding, we turned to isothermal titra-tion calorimetry. The titrations of calcium into a solution of LapGprotein are shown in Fig. 7. Purified protein bound calcium sub-stoichiometrically with 0.6 calcium molecule per molecule ofLapG and with an affinity of 6.2 �M (Fig. 7A; Table 1). EGTAtreatment and repurification of LapG by gel filtration in EGTA-free buffer as the mobile phase yielded protein that bound calcium

FIG 3 Crystal structure of L. pneumophila LapG. (A) Cartoon presentation of the LapG fold. The N- and C-terminal lobes are shown in slate and cyan,respectively. The conserved catalytic triad (cysteine-histidine-aspartate; C-H-D) is shown as sticks with the carbon atoms in orange. The electron density map atthe active site (bottom left inset) has amplitudes of 2�Fo� to �Fc�, with Fo and Fc being the observed and calculated structure factors. The electron density contouris at 1.8 . Water molecules are shown as red spheres. A hydrogen bond network involving residues of the catalytic triad is shown (bottom right inset). (B)Conservation of the catalytic triad based on the sequence alignment shown in Fig. S1 in the supplemental material. Separate sequence logos are shown for theLapG subfamily and the wider family of bacterial transglutaminase-like cysteine proteases (BTLCPs), which was generated based on an alignment that covers arepresentative, nonredundant set of sequences. (C) Surface conservation and hydrophobicity. Based on the alignment of 24 sequences of LapG orthologs, thesequence conservation was mapped to the accessible surface (left). In the right panel, the surface is colored according to the hydrophobicity of accessible residues.Hydrophobic residues are shown in green, and polar and charged residues are gray and pink, respectively.

Crystal Structure of L. pneumophila LapG

August 2012 Volume 194 Number 16 jb.asm.org 4419

with the same affinity and negative change in enthalpy as the start-ing material but with a stoichiometry of close to 1 (Fig. 7A; Table1). Together, these data indicate that LapG purifies partiallybound to calcium, which can be extracted by EGTA treatment,yielding calcium-free LapG. Based on the titration data, we expectonly one calcium-binding site per LapG monomer. Given that nobound calcium was observed in the initial crystal structure, weinfer that the crystallization conditions are sensitive to the calci-um-bound state of LapG. Finally, mutations in the key residues forcalcium coordination (D136A, E138A, or D139A) that were revealedby the crystal structure significantly reduce calcium binding byfactors of �50 (Fig. 7B), well within the range one would expecton the basis of similar mutations in other calcium-binding pro-teins (39). In summary, the structural studies elucidate the overallfold, catalytic core machinery, and a calcium-binding site as con-served features of LapG-like proteases and members of the widerBTLCP superfamily.

Oligomeric state of LapG in solution. It can be assumed thatthe transmembrane receptor LapD, which interacts with LapG,forms a constitutive homodimer mediated at least by its HAMPdomains, a two-helix module that forms a four-helix bundlewithin a dimeric assembly as the functional unit. This promptedus to analyze the oligomeric state of LapG in the crystals and insolution.

Two different crystal forms of calcium-free LapG were ob-tained, both with two molecules in their asymmetric units (RMSDof protomer A versus protomer B, 0.08 Å). A potential dimer

interface can be identified on the basis of crystal packing interac-tions. Although the protomer interactions are identical in the twoforms (RMSD considering dimeric assemblies, 0.17 Å), the inter-face is marginal, spanning only 670 Å2. In both crystal forms,protomers interact via a hydrophobic interface formed by the�-sheets of the C-terminal lobe (Fig. 8A). Another part of theinterface in the crystallographic dimer is presented by the loopprotruding from the N-terminal lobe that contains a residue in-volved in calcium binding (D120), in addition to the conservedresidue, D113. However, the overall packing and calculated bind-ing enthalpies suggest that the dimerization interface is ratherweak overall (23).

We next evaluated LapG’s oligomerization propensity in solu-tion by using purified proteins. We used SEC-MALS, a methodthat yields the absolute molecular mass of a protein as it is elutedfrom a gel filtration column (10). L. pneumophila LapG purifiedfrom E. coli is monomeric in solution with a molecular mass of�25 kDa (theoretical molecular mass based on sequence, 21.6kDa; Fig. 8B). There was no change in its monomeric state whenthe protein was incubated with calcium prior to SEC-MALS anal-ysis or even when calcium was added to the mobile phase, so thatcalcium was present throughout the experiment.

While we observed only monomeric LapG in solution, thecrystallographic dimer interface is moderately conserved acrossvarious bacterial strains (Fig. 3C; see Fig. S1 in the supplementalmaterial), which could suggest that a subset of LapG-like or-thologs may utilize this surface. Also, considering that the receptorLapD forms constitutive dimers (29), such a mode of LapGdimerization may still be relevant when LapG is bound to theoutput domains of intact LapD.

Conservation of the basic catalytic mechanism of LapG. Un-like in the case of P. fluorescens, the LapG-containing operon in L.pneumophila lacks a putative LapA-like substrate protein. Al-though the identity of the substrate for the L. pneumophila or-tholog of LapG remains unknown, we evaluated LapG’s catalyticactivity by using P. fluorescens LapA as a model. LapA is a secreted,large adhesin protein. Previous experiments established thatshorter constructs including the LapG cleavage site, a conservedTAAG motif, are processed efficiently by P. fluorescens LapG invitro (30). The model substrate we use here comprises a 235-resi-due-long, N-terminal fragment of LapA with a His6 tag at its Cterminus (LapANterm) (Fig. 9). Proteolysis of LapANterm by puri-fied P. fluorescens LapG produces a fragment of 15 kDa that can bedetected by Western blotting with primary antibodies that recog-nize the His6 tag. Mutation of D134 to an alanine residue in P.fluorescens LapG, which corresponds to the calcium-binding mu-tation D136A in L. pneumophila LapG, abolishes catalytic activitycompletely, corroborating a calcium-dependent mechanism forLapG function.

Similarly, L. pneumophila LapG successfully processed LapANterm,albeit less efficiently than P. fluorescens LapG. In these assays, weused 100-, 250-, or 375-fold more LapG than in the samples incu-bated with the P. fluorescens protease. While we were able to detecta cleavage product with a molecular mass similar to that producedby P. fluorescens LapG, indicating that the same cleavage site wasrecognized, reaction mixtures containing L. pneumophila LapGstill contained uncleaved substrate. In addition, we observed asecond, minor cleavage product with a slightly higher molecularmass at the two highest protease concentrations, which may indi-cate an aspecific activity under these conditions and with this

FIG 4 Conservation of a hydrogen bond network around the catalytic triad.(A) Residues from the N- and C-terminal lobes involved in a hydrogen bondnetwork around the active site are shown as sticks. (B) Sequence conservation.Sequence logos for LapG orthologs and the entire BTLCP family are shown.

Chatterjee et al.

4420 jb.asm.org Journal of Bacteriology

model substrate. However, proteolysis efficiency was LapG con-centration dependent. Furthermore, the calcium-binding single-point mutations (D136A, E138A, and D139A in L. pneumophilaLapG) produced markedly reduced catalytic activity. Consis-tently, both protease orthologs failed to cleave LapANterm whenincubated with EGTA, indicating that calcium is essential forLapG’s function as a periplasmic protease.

Altogether, we validated our prediction of the existing LapD-LapG system in L. pneumophila. More importantly, we showedthat it utilizes an output and proteolytic mechanism similar to thatof its P. fluorescens counterparts.

DISCUSSION

The superfamily of DUF920 domain-containing proteins or BTLCPsconsists of more than 400 members and is conserved across thebacterial domain (13), yet functional and structural informationhas been scarce. We recently identified LapG as a functional ex-ample of this family, assigning it a biological role as a calcium-dependent, periplasmic protease that is involved in biofilm dis-persal of P. fluorescens (4, 30). Since we were unable to obtaincrystals of P. fluorescens LapG thus far, we instead determined thehigh-resolution crystal structure of the LapG ortholog from L.pneumophila in its apo and calcium-bound states. Consideringthe sequence conservation in functionally important motifs, thestructures yield prototypic models of LapG-like proteins and the

wider family of BTLCPs. In addition, the structures confirmedtheir structural relationship to eukaryotic transglutaminases. Inanalogy to transglutaminases, LapG also shows calcium depen-dence, yet the exact molecular mechanism remains elusive in bothcases despite the available structural data.

Calcium has been known to play myriad roles in modulatingprotein stability and function. In order for a protein to interactwith calcium, mainly through electrostatic forces, it usually hasnegative charges on its surface. In the case of LapG, the N-terminallobe displays residues D136, E138, and D139, creating a highly neg-ative surface potential with which Ca2� can coordinate. It is pos-sible that calcium binding to LapG aids in catalysis by providingincreased access to the substrate or by altering enzyme structure ordynamics. While three of the four calcium-coordinating residuesare strictly conserved, the fourth residue (D120) appears morevariable. On the other hand, this residue is located in a flexibleloop that often harbors other, more conserved aspartate or gluta-mate residues in close proximity (Fig. 6), which may have a redun-dant calcium-coordinating function. It is possible that crystallo-graphic packing biases the use of one position over another. Theuse of alternative aspartate or glutamate residues would likely im-pact the conformation of the loop, which could potentially alterthe active or substrate-binding site. A similar mechanism has beendiscussed for mammalian tissue transglutaminases (26).

Another effect of calcium binding on enzyme catalysis could

FIG 5 Structural similarities of LapG to transglutaminases, arylamine N-acetyltransferases, and putative bacterial proteases. A DALI protein structure databasesearch highlighted features structurally conserved between LapG (A) and distinct classes of proteins. Regions corresponding to the N and C lobes of LapG are inred and pink, respectively, in panels B to E. Conserved catalytic (C-H-D) triads are shown as orange spheres. Calcium ions are shown as green spheres. (B) Humantransglutaminase 3. Z score, 7.4; RMSD, 3.6; aligning fragment spans 135 of 673 residues. Several transglutaminases were identified. The particular structure waschosen to highlight the positions of calcium-binding sites in these enzymes. (C) Mesorhizobium loti arylamine N-acetyltransferase 1. Z score, 7.8; RMSD, 3.6;aligning fragment spans 127 of 267 residues. (D) Putative cysteine protease from Cytophaga hutchinsonii. Z score, 9.4; RMSD, 3.1; aligning fragment spans 121of 285 residues. (E) Putative protease from Parabacteroides distasonis ATCC 8503. Z score, 7.8; RMSD, 3.2; aligning fragment spans 121 of 505 residues.

Crystal Structure of L. pneumophila LapG

August 2012 Volume 194 Number 16 jb.asm.org 4421

simply be a change in electrostatic potential close to the active site.Calcium binding is also known to influence the specific subcellu-lar localization of proteins. For example, in the case of calpain,calcium is required for the association of the protein with cellularmembranes (7). In the case here, calcium binding may preferen-

tially place LapG in close proximity to its substrate, LapA, which isanchored in the outer membrane.

The identity of the physiological target of L. pneumophila LapGis unknown. While the structural similarity of LapG to transglu-taminases may indicate different catalytic functions, no transami-

FIG 6 Structural identification of a conserved calcium-binding site in LapG. (A) Position of a putative calcium-binding site in LapG. Sequence conservationindicates that at least 3 of the 4 residues predicted to form an ion-binding site are conserved in LapG and BTLCPs. (B) Crystal structure of calcium-bound LapG.The crystal structure of LapG (gray) was determined in the presence of calcium ions (pink sphere). The left inset shows uninterpreted electron density that wasobserved only in the presence of calcium. The density has amplitudes of �Fo� to �Fc� and is contoured at 3.6 prior to the inclusion of calcium in the refinement.The structure of non-calcium-bound LapG was superimposed. On the right, a close-up view of the calcium-binding site and calcium coordination is shown.

FIG 7 Isothermal titration calorimetry data for calcium binding to LapG. (A) Binding of calcium to wild-type L. pneumophila LapG. Calorimetric titration forcalcium (2 mM) titrated into LapG (0.2 mM) is shown (top, raw data; bottom, binding isotherms). LapG purified from E. coli was used with or without EGTAtreatment during its purification. (B) Binding of calcium to mutant LapG forms. Single-point mutations were introduced into the calcium-binding site of LapGand analyzed as described for panel A. Data analyses are summarized in Table 1.

Chatterjee et al.

4422 jb.asm.org Journal of Bacteriology

dase activity was detected in an in vitro assay (D.C. and H.S., un-published data). However, L. pneumophila LapG can cleaveLapANterm and thus is likely to function as a periplasmic cysteineprotease. The reduced activity of L. pneumophila LapG for

LapANterm compared to that of the P. fluorescens enzyme couldhave several causes. Although we previously determined a consen-sus site for proteolytic cleavage of LapA by LapG, we also noticedthat a minimal peptide containing that motif evades processing bythe enzyme, suggesting that substrate recognition requires otherparts of LapA (Peter D. Newell and G.A.O., unpublished data).Furthermore, L. pneumophila has a more complex life-style since,unlike P. fluorescens, it can grow autonomously or within hostcells, increasing the potential target pool vastly. The notion that aputative target may be encoded outside the LapG-containingoperon is supported by the presence of an incomplete type I se-cretion system at this L. pneumophila locus. Furthermore, the se-cretion system may be complemented by subunits encoded out-side the lapDG-containing operon. Intracellular growth andexposure to different environments (e.g., a different pH or tem-perature) may also have an impact on the optimal point at whichL. pneumophila LapG is active. Altogether, these observationshamper the identification of potential L. pneumophila substratesby using bioinformatics.

Yet, an interesting speculation is the possibility that LapG or re-lated proteases could degrade host proteins or have a role not only intheir own biofilm formation but also in those formed by competing

TABLE 1 Calcium binding measured by isothermal titration calorimetrya

L. pneumophila LapG nb Kd (�M) H (cal/mol) S (cal/mol/degree)

Wild type 0.59 � 0.003 6.17 � 0.22 �1,615 � 10.0 18.3EGTA-treated wild type 0.91 � 0.005 4.98 � 0.32 �1,720 � 13.0 18.4D136A 0.95 � 0.142 349.7 � 59.3 �2,700 � 569.9 6.6E138A NDc ND ND NDD139A ND ND ND NDa Experiments were performed at 20°C.b n, reaction stoichiometry.c ND, not detected.

FIG 8 Oligomerization of LapG. (A) Crystallographic dimer. Analysis of crystalpacking interactions reveals a putative LapG dimer that forms via partially hydro-phobic interactions between protomers. (B) Oligomeric state in solution. SEC-MALS was used to determine the absolute molecular mass of LapG in solution.The effect of calcium on LapG’s oligomeric state was assessed by adding calcium tothe purified protein and by additionally adding calcium to the chromatographymobile phase. The averaged molecular masses were 25.1 � 0.3 kDa for apo-LapG,25.7 � 0.3 kDa for LapG�Ca2�, and 22.5 � 0.2 kDa for LapG with calcium in themobile phase.

FIG 9 Proteolytic activity of L. pneumophila LapG. (A) A model substratebased on P. fluorescens LapA. Previous studies of P. fluorescens identified theLapG-cleavage site of LapA within the first 235 residues of the adhesin. Aminimal, His6-tagged construct containing this region (LapANterm) was pro-teolyzed specifically and efficiently by P. fluorescens LapG. (B) L. pneumophilaLapG activity. Site-specific cleavage of purified LapANterm was monitored byusing SDS-PAGE, followed by Western blotting with pentahistidine-specificantibodies. L. pneumophila LapG with the E138A or D139A mutation was usedonly at the highest concentration. P. fluorescens LapG and a correspondingcalcium-binding mutant were included as positive and negative controls, re-spectively. Sensitivity to EGTA treatment was also assessed. wt, wild type.

Crystal Structure of L. pneumophila LapG

August 2012 Volume 194 Number 16 jb.asm.org 4423

bacteria. Substrate topology and protease access may determine thefeasibility of such a model. At least in P. fluorescens, LapG localizationto the periplasm is crucial for its regulatory effect on biofilm forma-tion, as externally added protease does not appear to alter biofilmphenotypes (C.D.B. and G.A.O., unpublished data).

Recently, a LapADG-like system was identified in S. oneidensis(37). In this system, a large outer membrane protein, BpfA, iscrucial for surface attachment and subsequent biofilm formation.The gene that encodes BpfA is located in a seven-gene operonwhich also encodes a type I secretion system, a GGDEF-EAL do-main-containing protein similar to LapD, and a hypothetical con-served BTLCP-like protein homologous to LapG. An interestingfinding pertains to a positive effect of calcium on S. oneidensisbiofilm formation, which is BpfA dependent. Considering thepresence of putative calcium-binding sites in LapA and BpfA,these large adhesin proteins could function as sinks for calciumions that, in turn, are required for proteolytic processing viaLapG-like proteases. Such a mechanism is supported by the ob-servation that bacteria have the ability to concentrate calcium intheir periplasm to levels that are comparable to or above the mea-sured affinity of LapG for calcium (22). Although the exact mech-anism by which these proteins interact and function has yet to beestablished, it is quite possible that they are regulated in a fashionvery similar to that of the Lap system in P. fluorescens (4, 29, 30).

Although the corresponding regulators and mechanism re-main elusive for the orthologous system in L. pneumophila, agrowth phenotype in macrophages has been observed upon theoverexpression of CdgS9, the LapD ortholog in L. pneumophila,indicating that the system is operational in this opportunisticpathogen (25). In particular, cells showed a moderate or severegrowth reduction in rich broth or in macrophages, respectively.Our structural and mechanistic studies described here may facili-tate future research to uncover the physiological role of this sys-tem in L. pneumophila and other bacteria. The protease and itsinteraction with the receptor LapD/CdgS9 also provide an attrac-tive angle for the development of specific inhibitors.

ACKNOWLEDGMENTS

We are grateful to Howard Shuman for providing genomic DNA and toWilliam Horne for providing access to isothermal titration calorimetry.

This work is based upon research conducted at the CHESS. The facilityis supported by grant DMR-0225180 from the National Science Founda-tion (NSF) and grant RR-01646 from the National Institutes of Health(NIH). Our work was supported by the NIH under grants R01-GM081373 (H.S.), R01-AI097307 (G.A.O. and H.S.), and T32 GM08704(C.D.B.); by the NSF under grant MCB-9984521 (G.A.O.); and by a PEWscholar award in Biomedical Sciences (H.S.).

REFERENCES1. Achyuthan KE, Greenberg CS. 1987. Identification of a guanosine

triphosphate-binding site on guinea pig liver transglutaminase. Role ofGTP and calcium ions in modulating activity. J. Biol. Chem. 262:1901–1906.

2. Adams PD, et al. 2010. PHENIX: a comprehensive Python-based systemfor macromolecular structure solution. Acta Crystallogr. D Biol. Crystal-logr. 66:213–221.

3. Ahvazi B, Boeshans KM, Idler W, Baxa U, Steinert PM. 2003. Roles ofcalcium ions in the activation and activity of the transglutaminase 3 en-zyme. J. Biol. Chem. 278:23834 –23841.

4. Boyd CD, Chatterjee D, Sondermann H, O’Toole GA. 2012. LapG,required for modulating biofilm formation by Pseudomonas fluorescensPf0-1, is a calcium-dependent protease. J. Bacteriol. 194:4406 – 4414.

5. Carlson HK, Vance RE, Marletta MA. 2010. H-NOX regulation of c-di-

GMP metabolism and biofilm formation in Legionella pneumophila. Mol.Microbiol. 77:930 –942.

6. Casadio R, et al. 1999. The structural basis for the regulation of tissuetransglutaminase by calcium ions. Eur. J. Biochem. 262:672– 679.

7. Croall DE, Ersfeld K. 2007. The calpains: modular designs and functionaldiversity. Genome Biol. 8:218. doi:10.1186/gb-2007-8-6-218.

8. Crooks GE, Hon G, Chandonia JM, Brenner SE. 2004. WebLogo: asequence logo generator. Genome Res. 14:1188 –1190.

9. Datta S, Antonyak MA, Cerione RA. 2006. Importance of Ca(2�)-dependent transamidation activity in the protection afforded by tissuetransglutaminase against doxorubicin-induced apoptosis. Biochemistry45:13163–13174.

10. De N, Navarro MV, Wang Q, Krasteva PV, Sondermann H. 2010.Biophysical assays for protein interactions in the Wsp sensory system andbiofilm formation. Methods Enzymol. 471:161–184.

11. Emsley P, Cowtan K. 2004. Coot: model-building tools for moleculargraphics. Acta Crystallogr. D Biol. Crystallogr. 60:2126 –2132.

12. Galperin MY, Nikolskaya AN, Koonin EV. 2001. Novel domains of theprokaryotic two-component signal transduction systems. FEMS Micro-biol. Lett. 203:11–21.

13. Ginalski K, Kinch L, Rychlewski L, Grishin NV. 2004. BTLCP proteins:a novel family of bacterial transglutaminase-like cysteine proteinases.Trends Biochem. Sci. 29:392–395.

14. Gjermansen M, Nilsson M, Yang L, Tolker-Nielsen T. 2010. Character-ization of starvation-induced dispersion in Pseudomonas putida biofilms:genetic elements and molecular mechanisms. Mol. Microbiol. 75:815–826.

15. Gouet P, Courcelle E, Stuart DI, Metoz F. 1999. ESPript: analysis ofmultiple sequence alignments in PostScript. Bioinformatics 15:305–308.

16. Hall-Stoodley L, Costerton JW, Stoodley P. 2004. Bacterial biofilms:from the natural environment to infectious diseases. Nat. Rev. Microbiol.2:95–108.

17. Hengge R. 2009. Principles of c-di-GMP signalling in bacteria. Nat. Rev.Microbiol. 7:263–273.

18. Hinsa SM, Espinosa-Urgel M, Ramos JL, O’Toole GA. 2003. Transitionfrom reversible to irreversible attachment during biofilm formation byPseudomonas fluorescens WCS365 requires an ABC transporter and a largesecreted protein. Mol. Microbiol. 49:905–918.

19. Hinsa SM, O’Toole GA. 2006. Biofilm formation by Pseudomonas fluo-rescens WCS365: a role for LapD. Microbiology 152:1375–1383.

20. Holm L, Rosenstrom P. 2010. Dali server: conservation mapping in 3D.Nucleic Acids Res. 38:W545–W549.

21. Holton SJ, et al. 2005. Structure of Mesorhizobium loti arylamine N-acetyltransferase 1. Acta Crystallogr. Sect. F Struct. Biol. Cryst. Commun.61(Pt 1):14 –16.

22. Jones HE, Holland IB, Campbell AK. 2002. Direct measurement of freeCa(2�) shows different regulation of Ca(2�) between the periplasm andthe cytosol of Escherichia coli. Cell Calcium 32:183–192.

23. Krissinel E, Henrick K. 2007. Inference of macromolecular assembliesfrom crystalline state. J. Mol. Biol. 372:774 –797.

24. Larkin MA, et al. 2007. Clustal W and Clustal X version 2.0. Bioinfor-matics 23:2947–2948.

25. Levi A, Folcher M, Jenal U, Shuman HA. 2011. Cyclic diguanylatesignaling proteins control intracellular growth of Legionella pneumophila.mBio 2:e00316 – 00310. doi:10.1128/mBio.00316-10.

26. Li B, Cerione RA, Antonyak M. 2011. Tissue transglutaminase and itsrole in human cancer progression. Adv. Enzymol. Relat. Areas Mol. Biol.78:247–293.

27. Liu S, Cerione RA, Clardy J. 2002. Structural basis for the guaninenucleotide-binding activity of tissue transglutaminase and its regulationof transamidation activity. Proc. Natl. Acad. Sci. U. S. A. 99:2743–2747.

28. Monds RD, Newell PD, Gross RH, O’Toole GA. 2007. Phosphate-dependent modulation of c-di-GMP levels regulates Pseudomonas fluore-scens Pf0-1 biofilm formation by controlling secretion of the adhesinLapA. Mol. Microbiol. 63:656 – 679.

29. Navarro MV, et al. 2011. Structural basis for c-di-GMP-mediated inside-out signaling controlling periplasmic proteolysis. PLoS Biol. 9:e1000588.doi:10.1371/journal.pbio.1000588.

30. Newell PD, Boyd CD, Sondermann H, O’Toole GA. 2011. A c-di-GMPeffector system controls cell adhesion by inside-out signaling and surfaceprotein cleavage. PLoS Biol. 9:e1000587. doi:10.1371/journal.pbio.1000587.

31. Newell PD, Monds RD, O’Toole GA. 2009. LapD is a bis-(3=,5=)-cyclic

Chatterjee et al.

4424 jb.asm.org Journal of Bacteriology

dimeric GMP-binding protein that regulates surface attachment byPseudomonas fluorescens Pf0-1. Proc. Natl. Acad. Sci. U. S. A. 106:3461–3466.

32. Noguchi K, et al. 2001. Crystal structure of red sea bream transglutami-nase. J. Biol. Chem. 276:12055–12059.

33. Otwinowski Z, Minor W. 1997. Processing of X-ray diffraction datacollected in oscillation mode. Macromol. Crystallogr. A 276:307–326.

34. Plate L, Marletta MA. 2012. Nitric oxide modulates bacterial biofilmformation through a multicomponent cyclic-di-GMP signaling network.Mol. Cell 46:449 – 460.

35. Schneider TD, Stephens RM. 1990. Sequence logos: a new way to displayconsensus sequences. Nucleic Acids Res. 18:6097– 6100.

36. Sondermann H, Shikuma NJ, Yildiz FH. 2012. You’ve come a long way:c-di-GMP signaling. Curr. Opin. Microbiol. 15:140 –146.

37. Theunissen S, et al. 2010. The 285 kDa Bap/RTX hybrid cell surfaceprotein (SO4317) of Shewanella oneidensis MR-1 is a key mediator ofbiofilm formation. Res. Microbiol. 161:144 –152.

38. Wang X, Kirberger M, Qiu F, Chen G, Yang JJ. 2009. Towards predict-ing Ca2�-binding sites with different coordination numbers in proteinswith atomic resolution. Proteins 75:787–798.

39. Yang JJ, Gawthrop A, Ye Y. 2003. Obtaining site-specific calcium-binding affinities of calmodulin. Protein Pept. Lett. 10:331–345.

40. Yee VC, et al. 1994. Three-dimensional structure of a transglutaminase:human blood coagulation factor XIII. Proc. Natl. Acad. Sci. U. S. A. 91:7296 –7300.

41. Zhang J, Lesort M, Guttmann RP, Johnson GV. 1998. Modulation of thein situ activity of tissue transglutaminase by calcium and GTP. J. Biol.Chem. 273:2288 –2295.

Crystal Structure of L. pneumophila LapG

August 2012 Volume 194 Number 16 jb.asm.org 4425