Probing the Role of Divalent Metal Ions in a Bacterial Psychrophilic Metalloprotease: Binding Studies of an Enzyme in the Crystalline State by X-Ray Crystallography

The psychrophilic alkaline metalloprotease (PAP) produced bya Pseudomonas bacterium isolated in Antarctica belongs to theclan of metzincins, for which a zinc ion is essential for catalyticactivity. Binding studies in the crystalline state have beenperformed by X-ray crystallography in order to improve the understandingof the role of the zinc and calcium ions bound to this protease.Cocrystallization and soaking experiments with EDTA in a concentrationrange from 1 to 85 mM have resulted in five three-dimensionalstructures with a distinct number of metal ions occupying theion-binding sites. Evolution of the structural changes observedin the vicinity of each cation-binding site has been studiedas a function of the concentration of EDTA, as well as of time,in the presence of the chelator. Among others, we have foundthat the catalytic zinc ion was the first ion to be chelated,ahead of a weakly bound calcium ion (Ca 700) exclusive to thepsychrophilic enzyme. Upon removal of the catalytic zinc ion,the side chains of the active-site residues His-173, His-179and Tyr-209 shifted $~$ 4, 1.0, and 1.6 Å, respectively. Ourstudies confirm and also explain the sensitivity of PAP towardmoderate EDTA concentrations and propose distinct roles forthe calcium ions. A new crystal form of native PAP validatesour previous predictions regarding the adaptation of this enzymeto cold environments as well as the proteolytic domain calciumion being exclusive for PAP independent of crystallization conditions.

INTRODUCTION

Top

Introduction

The serralysin family constitutes a group of bacterial metalloproteaseswith a molecular mass $~$ 50 kDa, including psychrophilic alkalineprotease (PAP) from Pseudomonas sp. strain TAC II 18 (11). Thesemetalloproteases are further classified into the clan of metzincins(http://merops.sanger.ac.uk/), encompassing astacins, matrixmetalloproteinases (collagenases), and snake venom proteinases(5). The metzincins are characterized by the zinc-binding consensussequence HEXXHXXGXXHZ (residues 169 to 180 in PAP), where thehistidine residues are the zinc ligands and the Z residue ischaracteristic for the different subfamilies (in the case ofthe serralysins, Z = Pro), and by a conserved methionine residue(Met-207 in PAP). Although the sequence identity between thedifferent subfamilies of the metzincins is very low, their catalyticdomains do have similar three-dimensional (3D) structures (2-4,6, 7, 19-21).

The serralysins are secreted as inactive zymogens into the mediumvia the hemolysin pathway, involving an ABC transporter, whichrecognizes a C-terminal secretion signal. Activation occursin the medium autocatalytically in the presence of divalentions, especially Zn²⁺ and Ca²⁺ (12-14).

The implication of zinc metalloproteases in many diseases, e.g.,cancer or arthrosis (15), makes structural studies of theseenzymes highly interesting.

Here we report the results from X-ray crystallographic studiesof PAP, an alkaline protease produced by the psychrophilic Pseudomonassp. strain TAC II 18, which originated in Antarctica. This strain,identified as a gram-negative bacterium (18), is one of numerousbacteria living at temperatures close to 0°C. To surviveand grow under such conditions, these organisms have evolvedseveral adaptations of their cellular components, especiallyof their enzymes (16-18, 30). Thus, PAP is three times moreactive at 20°C than a mesophilic counterpart from Pseudomonasaeruginosa but rapidly inactivated at 45°C. The cold-adaptedenzyme also displays a greater sensitivity to moderate EDTAconcentrations, and fluorescence studies have indicated enzymeunfolding after 2 h of incubation in 2 mM EDTA (11).

The PAP molecule is made up of 463 amino acid residues distributedin two domains, and its tertiary structure is comparable tomesophilic homologues from P. aeruginosa alkaline protease (AP)and Serratia marcescens protease (SMP) (1-3). Four ${alpha}$ -helicesand a mixed five-stranded ß-sheet form the catalyticN-terminal domain to which the active-site Zn²⁺ ion and a Ca²⁺ion (Ca 700, PAP numbering) are bound. This Ca²⁺ ion is notpresent in its mesophilic counterparts, AP and SMP, but it shouldbe noticed that a calcium ion has been observed in the 3D structureof the metzincin adamalysin II in a position similar to thatof Ca 700 (20).

The C-terminal domain which may play a role in the folding ofthe molecule after transmembrane translocation forms an extendedß roll with 20 successive ß-strands woundin a right-handed spiral. The number of Ca²⁺ ions bound to thisC-terminal domain varies from six to seven in the differentorganisms and crystal forms.

When we compared the 3D structure with those of the mesophiliccounterparts, AP and SMP, the most remarkable differences werefound in the proteolytic domain. First, the active site is moreaccessible due, in part, to deletions in surrounding loops andto a region in the catalytic domain of PAP (residues 107 to116) that undergoes a conformational change with a loop movementas large as 13 Å induced by the binding of the extra calciumion, Ca 700. This change does not seem to be related to crystalpacking effects, since it was observed in crystal forms obtainedunder very different crystallization conditions. The role ofthis extra calcium ion in the activity or folding of the PAPmolecule still needs to be elucidated. Its weak chelation inthe proteolytic domain and its close proximity to the solventcould partly explain the sensitivity of PAP to EDTA, which isalso supported by a higher temperature factor for this calciumion compared to other calcium ions in the 3D structure. Thissensitivity could also be related to the lower affinity, comparedto AP and SMP, for some calcium ions located in the C-terminaldomain.

In order to address these questions, we solved the structuresof PAP (form 2) in the presence of various concentrations ofEDTA, as well as a new native form (form 3) by X-ray crystallography.

MATERIALS AND METHODS

Top

Materials and Methods

Crystallization. PAP has been crystallized under three different conditions—form1 (31), form 2 (1), and form 3 [0.2 M (NH₄)₂SO₄, 22% PEG 4000,and 0.1 M sodium acetate buffer at pH 4.6 at 16°C]—bythe hanging-drop vapor diffusion method.

Cocrystallization experiments of PAP have been performed underconditions derived from form 2 [1.6 M (NH₄)₂SO₄ and 0.1 M HEPES(pH 7) at 16°C]. EDTA was added to a final concentrationof 1 mM (experiment A), 5 mM (experiment B), and 10 mM (experimentC) in the drops. Crystals grew within 1 week to a final sizeof about 300 by 100 by 100 µm and were left another 2months in the drops before data collection in experiments Aand C. In experiment B the crystals were left an additional5 days (B1) or 2 months (B2) in the droplets, respectively,before data collection.

Soaking was achieved by directly adding a solution of EDTA toa final concentration of 85 mM into a drop containing form 2crystals and leaving it 4 h at 16°C (experiment D). Higherconcentrations of EDTA have been tested, but the crystals crackedat concentrations of EDTA greater than 85 mM.

Cocrystallization or soaking with EDTA did not modify the spacegroup, but small variations in the unit cell dimensions wereobserved (Table 2); these were certainly due in part to flashfreezing of the crystals. The large difference in unit celldimensions between B1 and the remaining data is a consequenceof the B1 data being collected at 288 K and the rest being collectedat 100 K.

View this table:
[in this window]
[in a new window]
TABLE 2. Data collection and refinement statistics for experiments A, B1, B2, C, and D

Data collection, structure solution, and refinement. (i) Native form. Diffraction data on form 3 crystals (Table 1) were collectedwith a charge-coupled device camera by using synchrotron radiationat the beam-line D2AM at the European Synchrotron RadiationFacility (ESRF; Grenoble, France) under cryogenic conditions(100 K). Initial data reduction and space group determinationwere carried out with the program XDS (23), and further reductionof the data was performed by using programs from the CCP4 package(10).

View this table:
[in this window]
[in a new window]
TABLE 1. Data collection and refinement statistics for crystal form 3 of PAP

The known structure of the AP (3) served as model in a molecularreplacement search performed with the program AMoRe (26). Refinementof the structure was done by using the program CNS (9), applyingan overall anisotropic temperature factor correction algorithm,followed by manual fitting. A summary of the refinement statisticscan be found in Table 1.

Due to poor or missing electron densities, as for form 2 (1),residues 1 to 2 and residues 184 to 188 were not inserted intothe electron density map.

(ii) Experiments A, B, C, and D. X-ray diffraction data of crystals from experiments A, B2, andC were collected at the BM30A beamline at the ESRF at cryogenictemperatures (100 K) at a wavelength of 0.98 Å on a MarCCD.X-ray data on crystals from experiments B1 and D were collectedin house at 288 K and at cryogenic temperatures, respectively,on a MARresearch 345 image plate system associated with a NoniusFR591 rotating anode (CuK radiation) operating at 44 kV and100 mA and coupled to Osmic confocal mirrors.

Diffracted intensities in experiments A, B2, C, and D were integratedwith the program MOSFLM (25), as implemented in the CCP4 softwarepackage (10) and scaled with SCALA (10), whereas experimentB1 reflection data were processed and merged with DENZO/SCALEPACK(27).

Collection and processing statistics of experiments A, B, C,and D are summarized in Table 2, and structures resulting fromthese experiments will henceforward be referred to as structureA, B1, B2, C, and D, respectively. Structures B1 and B2 weresolved by the molecular replacement method by using the AMoResoftware as implemented in the CCP4 suite (10), and the nativeform 2 was resolved to a 1.96-Å resolution (1) as a searchmodel. Withdrawing diffraction data in a resolution range of15 to 4 Å, a unique solution was obtained with a correlationcoefficient of 58.8% and an R factor of 37.5% for B1 and a correlationcoefficient of 54.0% and an R factor of 37.5% for B2.

Since the crystals from experiments A, C, and D were isomorphouswith B2 crystals, their 3D structures were determined by directphasing with the 2.0-Å resolution B2 structure, for whichions as well as water molecules in the active site had beenremoved in order to avoid bias. The refinement of all modelswas done by using the simulated annealing protocol as implementedin the software package CNS (9), alternately with visual examinationof electron density maps and manual building by using the graphicsoftware TURBO-FRODO (29).

Water molecules present at similar positions in the respectivestructures have the same numbering.

In order to confirm the presence or absence of calcium or zincions, anomalous difference Fourier maps were calculated by usingthe program CNS, in which calculated phases are derived froma refined model depleted of metal ions.

For all data set, free and conventional R factors were monitored(8) to avoid overrefinement.

Compared to the native structure of form 2 (PDB entry 1G9K),residues not inserted in the electron density map due to pooror missing electron densities were residues 50 to 51 in structuresA, B2, C, and D and Tyr-183 in all five structures. Moreoverside chain residues for Asn-52, Asp-53, and Asn-184 were notobserved in electron density maps of structures A, B2, C, andD. Double conformations for Ser-15, Ser-138, and Ser-332 werefound in structure B2. Model qualities were examined with PROCHECK(24) and WHATCHECK (22). The details of the refinement statisticsand model quality are given in Table 2.

Relative B-factors were calculated by dividing the mean B-factorof every residue by the mean B-factor of the whole protein.Due to the rather poor quality of structure D data, B-factorswhere initially set to 20 Å² and left out of the refinement.

Superimposition of structures was done by using the "rigid"option in TURBO-FRODO. The figures shown here were renderedby using TURBO-FRODO and VIEWERLITE 5.0 (freeware from Accelrys,Inc., San Diego, Calif.).

Coordinates. The coordinates and structure factors are deposited in the theprotein data bank as 1O0Q (A), 1O0T (B1), 1OM6 (B2), 1OM8 (C),1OM7 (D), and 1OMJ (form 3).

RESULTS

Top

Results

Native form 3. In the crystal structure of form 3, 456 amino acid residues,one Zn²⁺ ion, and 8 Ca²⁺ ions could be located in the electrondensity map. The superposition of 3D structures of forms 2 and3 using all C ${alpha}$ atoms gives a root mean square deviation (rmsd)of 0.19 Å and thus two nearly identical structures, despitequite different crystallization conditions. The main differencebetween the two forms is an additional calcium-binding sitein the C-terminal domain of form 3, which was observed in form1 (1), as well as a loop region around Ca 700 that seems moreflexible in form 3, as judged from the discontinuous electrondensity. Another difference is the presence of a continuouselectron density in the active site of form 3 that can be interpretedas being either a peptide or a network of mobile water molecules.

Effect of EDTA on ion binding. (i) Overall structures. Comparative studies of native PAP (Fig. 1) with structures A,B1, B2, C, and D show that the overall 3D structures are verysimilar. The rmsd values between C ${alpha}$ ’s in these five structuresand the native form 2 of PAP were 0.42, 0.17. 0.43, 0.41, and0.37 Å, respectively. No density corresponding to a moleculeof EDTA can be seen in any of the calculated electron densitymaps. Significant differences, however, were observed in thenumber of metal ions ligated in the structures (Table 2). Infact, B1 reveals the structure of the apo-form of PAP, sincethe electron density map calculated on the basis of the 2.5-Ådata clearly showed that the active-site zinc ion had disappearedbut that all of the calcium ions were present relative to thenative structure, as confirmed by an anomalous difference Fouriermap.

View larger version (38K):
[in this window]
[in a new window]
FIG. 1. Global view of native PAP form 2. Regions of interest for relative B-factors are colored as a function of the degree of flexibility: yellow (residues 47 to 57) and cyan (residues 169 to 191) indicate areas that undergo major changes upon EDTA ligation, dark blue represents "middle-flexibility" zones, and pink (region from residues 320 to 380) indicates the most rigid part of the molecule.

The electron density maps calculated for structures A, B2, C,and D clearly showed that the catalytic zinc ion and the calciumion located in the catalytic domain (Ca 700) were lacking. However,in the C-terminal domain all calcium ions could be located andwere inserted in the electron density of structures A, B2 andC. In structure D, one calcium ion, Ca 702, is replaced by awater molecule.

(ii) Changes in the Zn(II)-binding pocket. In the active site of native PAP, the Zn²⁺ ion is coordinatedin a trigonal-bipyramidal geometry to four or five ligands dependingon the position of the Tyr-209. The Zn²⁺, when penta-coordinated,is bound to the side chains of three histidines (His-169, His-173,and His-179), Tyr-209, and a water molecule (1).

Electron density maps calculated for experiments A, B1, B2,C, and D, clearly show that no metal is present in the active-siteregion and that no additional water molecule replaces the zincion in the binding site (Fig. 2). As a consequence, displacementsin the side chains of the zinc-ion ligands are observed, includingHis-173 and His-179. In the apo-form, the N ${varepsilon}$ 2 from His-173 undergoesa 4.1-Å shift and the imidazole ring of His-179 rotates40° around its Cß-C ${gamma}$ , leading to a 1.0-Ådisplacement of its N ${varepsilon}$ 2. Similar changes for these histidinesare observed in the other structures. Upon removal of the catalyticzinc ion, the phenol ring of Tyr-209 rotates ca. 60° aroundthe Cß-C ${gamma}$ bond. The side chain of His-169 does notundergo significant conformational changes. Interestingly, amajor part of the water molecules constituting an organizednetwork in native PAP have disappeared, including Wat 1225,which ligates the zinc ion in the holo form (Fig. 2).

View larger version (273K):
[in this window]
[in a new window]
FIG. 2. Comparison of the Zn-binding sites in structure B1 (apo-form protein) and native form 2. (A) 2Fo-Fc electron density maps contoured at a 1 ${sigma}$ level with the native protein (left panel) and the apo-protein (right panel), which lacks electron density for the zinc ion and water molecules. (B) Superimposition of the apoprotein in blue on the structure of native form 2 PAP in yellow, indicating a displacement of the zinc ion ligands in the apo form.

Excepting the network of water molecules, similar significantstructural changes (rmsd > 0.2 Å) have been noticedin structures A, B2, C, and D. In structure A, Wat 1225, 1102,and 1104 are present, whereas Wat 1102, 1104, and 1304 are foundin the catalytic site of structure B2, and Wat 1102 and 1134are found in structures C and D.

(iii) Calcium-binding sites. Of the seven calcium ions that are bound in the native form2, Ca 700 is found in the proteolytic domain where it is easilyaccessible to the solvent, and the others are found in the C-terminaldomain (Fig. 1).

Next to the proteolytic domain, Ca 701 and 702 are heptacoordinated,and the four remaining calcium ions (Ca 703 to Ca 706) are boundin the core of the parallel ß-roll in a hexacoordinatedmanner. These are supposed to be essential determinants of theparallel ß-roll structure (1, 3).

The loss of Ca 700 from the proteolytic domain in structuresA, B2, C, and D induces structural changes, including that ofAsn-57, which moves 1.5 Å toward the former calcium position.Moreover, 2Fo-Fc electron density was lacking for residues 50and 51 and for the side chains of residues 52 and 53, all ofwhich are located in the loop surrounding the Ca 700 bindingsite (Fig. 3). The predominant change caused by EDTA chelatingCa 700 is a partial disorganization of this loop.

View larger version (65K):
[in this window]
[in a new window]
FIG. 3. Ca 700 binding sites in the native form 2 of PAP (left panel) and in structure B2 (right panel) with 2Fo-Fc electron density maps contoured at a 1 ${sigma}$ level. It can be seen that electron density is lacking for the calcium ion and for residues 50 and 51 in B2 where a water molecule has been inserted in the former calcium-binding site.

From the highest-resolution structures (B2 and C at 2.0 Å),it can be concluded that a water molecule replaces the calciumion; hence, Wat 1500 has been inserted (Fig. 3).

Although the Ca 700 binding site is accessible to the solventdue to the disorganization of the loop, no additional watermolecules were found in the close vicinity in any of the structures.The 13-Å loop movement (residues 107 to 116) upon bindingof Ca 700 in native PAP compared to its mesophilic counterpartsis not reversible since no significant changes are observedin the backbone or in the side chain conformations of the loopresidues when this Ca²⁺ is lacking. It should be noted thatAsp-114 is one of the ligands of Ca 700. With regard to structureB2, in which Ca 700 has not been chelated, the distances tothe ligands have increased, especially for Asn-51-O ${delta}$ 1 (Table3), suggesting a weaker chelation. The six calcium ions boundin the C-terminal domain and their respective ligands are invariantin structures A and B1.

View this table:
[in this window]
[in a new window]
TABLE 3. Distances between the Ca²⁺ in the catalytic domain of PAP and its ligands^a

However, when higher concentrations of EDTA were used, structuralchanges, especially around Ca 701 and Ca 702, were observed(Table 4). In structures B2 and C, it appears that Ca 701 andCa 702 are weakly bound. The increased distance to Asp-282 suggeststhat these ions go from a hepta- to a hexa-coordinated state.

View this table:
[in this window]
[in a new window]
TABLE 4. Distance between Ca 701 and Ca702 and their ligands in native form 2 and in structure B2, C, and D^a

Particularly in structure D, Ca 702 is replaced by a water molecule(Wat 702) (Table 4). The electron densities corresponding tofive calcium ion sites located in the C-terminal domain wererelatively strong and observed in the 2Fo-Fc map up to 3 ${sigma}$ forCa 706 and to 4 ${sigma}$ for Ca 701, 703, 704, and 705. As for the electrondensity corresponding to Ca 702, this site was only observedin a 2Fo-Fc map contoured at a 1 ${sigma}$ level and, when a calcium ionis inserted at this site, a strong negative density appearsin the Fo-Fc electron density map. Ca 702 was therefore replacedby a water molecule in structure D. Furthermore, despite a mediumresolution of the structure, the distances measured betweenthis water molecule and its neighbors suggest that these interactionshave hydrogen bonding character rather than being coordinationbonds and thus confirms the substitution of Ca 702 by a watermolecule.

As a consequence of the loss of Ca 702, a small reorganizationof the Ca 701 ligands has taken place (Table 4). One shouldnote that Asp-287 ligates both calcium ions.

Moreover, significant structural changes (rmsd > 0.2 Å)have occurred in the Ca 703 and Ca 704 binding sites (Table5), suggesting that these ions are weakly bound.

View this table:
[in this window]
[in a new window]
TABLE 5. Distances between the Ca 703 and Ca 704 and their ligands in native form 2 and in structure D^a

(iii) Crystallographic temperature factors. Crystallographic B-factors to a certain extent reflect the flexibilityof a protein structure but are correlated to the data collectiontemperatures, data quality, and refinement procedure. In orderto pinpoint the more flexible parts of the molecule, relativeB-factors were calculated for each residue. We have comparedthe temperature factors of the native form 2 with those of A,B1 (apo-form), B2, and C, and differences were observed in fourregions (Fig. 1 and 4). The first stretch which displays highrelative B-factors compared to the native structure comprisesresidues 47 to 57 in structures A, B2, and C. The loss of Ca700 might therefore be partly responsible for the higher flexibilityof this loop, which is corroborated by the lack of electrondensity for residues 50 and 51. The second region concerns residues169 to 191 in the vicinity of the active site, with relativelyhigh B-factors observed for residues 173 and 179 in structuresA, B1, B2, and C, a finding consistent with side chain conformationalchanges induced by the loss of the zinc ion. It is likely thatthe removal of the zinc ion by the chelator has conferred ahigher flexibility to the active-site region. Considering theC-terminal domain, Fig. 4 shows that residues 282 to 287 displayhigh relative B-factors, thus confirming the structural changesof residues 282 and 287, which bind to Ca 701 and Ca 702.

View larger version (32K):
[in this window]
[in a new window]
FIG. 4. Relative B-factors for native form 2 and structures A, B1, B2, and C, indicating the four regions displaying major changes: residues 47 to 57, 169 to 191, 250 to 290, and 320 to 380. Gaps in the curves correspond to residues for which electron density was lacking.

Finally, mobile regions in the neighborhood of Ca 703 and Ca704 include residues 341 to 344, 346 to 353, and 366 to 373.

The relative B-factors for the calcium ions are given in Table6. It appears that relative B-factors of Ca 702, followed byCa 703 and Ca 704, increase with the concentration of EDTA.Despite the absence of structural changes observed for Ca 706and its ligands, this ion displays high relative B-factors instructures B2 and C, which may be due to increased exposureto the solvent, as emphasized by increased temperature factorsand Ca 706 having a water molecule as a ligand.

View this table:
[in this window]
[in a new window]
TABLE 6. Relative B-factors for calcium ions

DISCUSSION

Top

Discussion

The new crystal form (form 3) confirmed that the differencesbetween the psychrophilic enzyme and its mesophilic counterparts,and especially the binding of the extra calcium ion in the proteolyticdomain, are independent of crystallization conditions or crystalpacking.

Experiments A to D contribute to elucidate the role of divalent metal ions in PAP. In light of the conditions of experiments B1 and B2 (contacttimes with EDTA of 12 days and 2 months, respectively), it clearlyappears that although the zinc ion was more strongly coordinatedthan Ca 700 (1), it was the first to be chelated by EDTA. Moreover,the complexation constant of Zn²⁺ by EDTA is weaker than forCa²⁺ (10^10.7 and 10^16.3, respectively, at 25°C).

Notwithstanding chelation of Ca 700 by EDTA appears favorablefrom a thermodynamic point of view, our experiments suggestthat the action of EDTA on PAP predominantly is controlled bykinetics in agreement with studies on the ß-roll performedby Rose et al. (28). Thus, in order to explain the increasedsensitivity of PAP compared to mesophilic counterparts, kineticparameters such as steric hindrance should be considered withthe increased accessibility to the active site in PAP as thekey factor.

Our studies of all calcium-binding sites in structures A, B1,B2, C, and D and in the native form 2 of PAP have led us toanalyze their different roles in the global stability of themolecule and their potential implication in the folding of PAP.

Structures B1 and B2 indicate that Ca 700 is present duringthe crystallization process and hereby when crystals are formedand is only chelated by EDTA after crystal formation. One canspeculate that if this site is not occupied, the molecule istoo flexible for crystal formation. It should also be stressedthat the major part of the intermolecular interactions to theproteolytic domain is performed with residues from the regionsurrounding this calcium ion. These findings, taken togetherwith the fact that the 13-Å loop movement upon bindingof Ca 700 is nonreversible upon chelation in structures A, B2,C, and D, are indeed in agreement with an increased accessibilityto the active site, caused by this same loop movement, beinga key factor of the adaptation of this enzyme to cold environments.Thus, Ca 700 does not appear to improve the global stabilityof PAP, but rather the stability of the proteolytic domain asstrengthened by the partial disorganization of the loop surroundingthe depleted calcium site.

On the basis of these findings, we have proposed a classificationof calcium ions from the least to the most tightly bound. After2 months, a 1 mM concentration of EDTA appears to be sufficientfor chelating the calcium ion located in the N-terminal proteolyticdomain (Ca 700). When the concentration of EDTA was increasedto 5 mM, ligands of Ca 702 move away from the calcium-bindingsite. This weakly bound calcium ion also displays a high B-factorand is thought to be partially occupied, especially in structureC. This effect increases with the concentration of EDTA and,at 85 mM, Ca 702 seems to be replaced by a water molecule. Oneligand (Asp-287) is shared by Ca 702 and Ca 701. When the Ca702 binding site is modified, structural changes in the Ca 701binding site are concomitant. After 2 months in 5 mM EDTA, therelative B-factor of Ca 706 increases by 60%, indicating a partialoccupation of this site.

Ca 703 and Ca 704 in turn seem to be affected when the concentrationof EDTA was increased to 10 mM (structure C), as indicated bytheir high relative B-factors. Structure D, for which conformationalmovements among the ligands of Ca 703 and Ca 704 have been observed,confirms this observation.

As for the C-terminal domain, it can be concluded that Ca 702and hereafter Ca 701 are the less tightly bound, followed byCa 706, Ca 703, and Ca 704. Finally, Ca 705 emerges as the mosttightly bound ion since its binding site undergoes no significantchanges in any of the experiments.

This classification is coherent with coordination states andspatial positions in the structures. Ca 701 and Ca 702, whichare located in one extremity of the C-terminal domain and nextto the catalytic domain, are relatively accessible to the solvent.Although Ca 706 is bound more internally in the central regionof the ß-roll, this ion had one water molecule asligand and then was more exposed to the solvent. Finally, thelast three calcium ions (Ca 703 to 705) seem to hold the structuretogether, thus confirming their putative implication in thefolding of the molecule after transmembrane translocation.

Our studies confirm that PAP is sensitive to low concentrationsof EDTA, as described previously (11), since 1 mM EDTA in experimentA and 5 mM EDTA in experiment B1 appear to be sufficient forchelating the zinc ion in the N-terminal proteolytic domainand hence to inactivate the protein.

However, our results do not explain previous fluorescence studiesthat indicated enzyme unfolding after 2 h of incubation in 2mM EDTA (11) but, because the experiments described here takeplace within a crystal, direct comparisons are not possible.

From an overall structural point of view, relative exposureof some calcium ions located in the C-terminal domain to thesolvent seems to allow a higher conformational flexibility dueto the lack of protein-protein interactions by the bias of crystalcontacts.

The rigid nature of the C-terminal domain, which probably playsan important role in the folding of the molecule, and the factthat Ca 703 to Ca 705 are essential determinants of the parallelß-roll structure (3), a finding supported by the presentstudy, promote restrictions in conformational changes withinthis domain. This can probably explain why the 3D structureof PAP remains folded in high concentrations of EDTA (up to85 mM as found in the soaking experiments). Possibly, the chelationof Ca 705 would result in the unfolding of the protein, whichin turn may explain why form 2 crystals cracked in the soakingexperiments when the EDTA concentrations were beyond 85 mM.

ACKNOWLEDGMENTS

This work was supported by the EU contracts BI04-CT96-0051 andBI04-CT97-0131, the Centre National d'Etudes Spatiales, andthe Centre National de la Recherche Scientifique.

We thank C. Gerday for providing the enzyme for the crystallographicstudies and R. Kahn for assistance with data collection at theESRF.

FOOTNOTES

* Corresponding author. Mailing address: Laboratoire de BioCristallographie, Institut de Biologie et Chimie des Protéines, UMR 5086, CNRS-UCBL1, 7 Passage du Vercors, 69367 Lyon, Cedex 07, France. Phone: 33(0)472722608. Fax: 33(0)472722616. E-mail: r.haser{at}ibcp.fr.

REFERENCES

Top