Structural and Functional Analysis of the Metal-binding Sites of Clostridium thermocellum Endoglucanase CelD

Crystallographic analysis indicated that Clostridium thermocellum endoglucanase CeID contained three Ca 2 + -binding sites, termed A, B, and C, and one Zn 2 +_ binding site. The protein contributed five, six, and three of the coordinating oxygen atoms present at sites A, B, and C, respectively. Proteins altered by mutation in site A (CeIDD246A)'B (CeIDD36IA)' or C (CeIDD523A) were com pared with wild type CeID. The Ca2 + -binding isotherm of wild type CeID was compatible with two high affinity sites (K a = 2 x 106 M-1) and one low affinity site (K a < 105 M-1). The Ca 2 + -binding isotherms of the mutated pro teins showed that sites A and B were the two high affin ity sites and that site C was the low affinity site. Atomic absorption spectrometry confirmed the presence of one tightly bound Zn 2 + atom per CeID molecule. The inacti vation rate of CelD at 75°C was decreased 1.9-fold upon

M-1). The Ca 2 + -binding isotherms of the mutated proteins showed that sites A and B were the two high affinity sites and that site C was the low affinity site. Atomic absorption spectrometry confirmed the presence of one tightly bound Zn 2 + atom per CeID molecule. The inactivation rate of CelD at 75°C was decreased 1.9-fold upon increasing the Ca 2 + concentration from 2 x 10-5 to 10-3 M. The K m of CeID was decreased 1.8-fold upon increasing the Ca 2 + concentration from 5 x 10-6 to 10-4 M. Over similar ranges of concentration, Ca 2 + did not affect the thermostability nor the kinetic properties of CeID D523 A" These findings suggest that Ca 2 + binding to site C stabilizes the active conformation of CeID in agreement with the close vicinity of site C to the catalytic center.
Clostridium thermocellum synthesizes a multienzymatic cellulase complex with a molecular mass of 2-4 MDa, termed cellulosome 0, 2). Endoglucanase CelD is a component of the cellulosome, which can be easily purified in large amounts from inclusion bodies produced in recombinant Escherichia coli (3). CelD belongs to the family E of cellulases (4,5). The threedimensional structure of CelD 1 has been determined by x-ray crystallography (6). The protein contains two distinct structural domains that are closely associated: a small amino-terminal f3-barrel domain and a larger, mostly a-helical domain, whose amino acid sequence is similar in all catalytic domains of family E cellulases (4,7). The COOH terminus ofCelD consists of a duplicated segment of 23 residues that is involved in anchoring the protein to the scaffolding component of the cellulosome (8,9). The part of the protein visible in the electron density map terminates 10 residues upstream from the begin-* This work was supported by Contract AIR1-CT-0321from the Commission of the European Communities and by research funds from the University of Paris 7. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
ning of the COOH-terminal duplication. A cleft on the surface of the a-helical domain constitutes the active site. According to structural analysis (6) and mutagenesis data (0), the two residues participating in acid-base catalysis are Asp-201 and Glu-555.
We have previously shown that Ca 2 + binds to CelD, thereby stabilizing the enzyme against thermal denaturation and increasing its substrate binding affinity (11). Three putative Ca 2 + -binding sites and one putative Zn 2 + -binding site were identified in the catalytic domain of the CelD crystal structure (6).
This paper reports the structural analysis of the Zn 2 + -binding site and of the three Ca 2 + -binding sites of C. thermocellum CelD. The presence of Zn 2 + in CelD was assayed by atomic absorption spectrometry. CelD proteins carrying mutations in each of the Ca 2 + -binding sites were purified and characterized to assess the contribution of each site to Ca 2 + binding. The rate of inactivation at 75°C and the kinetic parameters of wild type CelD were determined in the presence of varying Ca 2 + concentrations to correlate changes in these parameters with the occupancy of high or low affinity Ca 2 + -binding sites. The same assays were performed with CelD mutated in the low affinity Ca 2 + -binding site.

MATERIALS AND METHODS
Crystallographic Analysis-Two isomorphous crystal forms of CelD were grown using ammonium sulfate ti.e. no added calcium) or 300 mM calcium chloride as precipitants. Structure determination and independent refinement of the two forms at 2.3 A resolution have been described elsewhere (6). The present models comprise residues 36-574 and include three calcium ions, one zinc ion, and 221 (ammonium sulfate) or 204 (calcium chloride) water molecules. The final agreement factors between observed and calculated structure factor amplitudes in the resolution range 6-2.3 Awere 17.0%for 33,211 observed reflections with F > 5 u(F) (ammonium sulfate) and 17.4% for 29,797 observed reflections (calcium chloride). Root mean squares deviations of bond lengths and angles from ideality were 0.007 A and 1.6 0 , respectively, in both crystal structures.
Purification of Wild Type and Mutant Forms of CelD-E. coli cells harboring the appropriate plasmids were grown to stationary phase at 37°C in Luria Bertani broth (14) containing 100 ug/ml ticarcillin. Wild type and mutant forms of CelD were purified from inclusion bodies as previously described (3). Low and high M; forms ofCelDD24sA (CelD-A*) 9757 This is an Open Access article under the CC BY license. and CelDOS6iA (CelD-B*) were separated on a Mono-Q anion exchange column using a fast performance liquid chromatography system (Pharmacia Biotech Inc.). Up to 4 mg of purified protein was loaded on a Mono-Q HR5/5 anion exchange column (1 ml) equilibrated with 20 mM Tris-HCI, pH 7.7, at a rate of 1 ml/min. Elution was performed at 0.7 mlJmin using a linear gradient from 100 to 220 mM NaCI in the same buffer. The low M r and high M r peaks were eluted at 150 and 180 mM NaCI, respectively, and concentrated by ultrafiltration using a YM10 Amicon membrane. All samples were dialyzed against 40 mM Tris-HCI, pH 7.7.
Zinc Assay-The zinc content of wild type CelD was assayed by flame atomic absorption spectroscopy at 213.9 nm using a Varian AA-1275 spectrophotometer (Varian Techtron, Springvale, Australia), with a single element hollow-cathode lamp for zinc (16).
Ca 2 + -binding Assay-Binding of 45Ca to purified proteins was assayed by monitoring the release of 45Ca from Chelex-100 (Bio-Rad) previously equilibrated with various concentrations of 45Ca (11).
Enzyme and Protein Assays-All reagents used in assays performed in the presence of controlled concentrations of Ca 2+ were kept in disposable plasticware (Sterilin) and were handled with disposable plastic pipettes or pipette tips. Divalent metals were removed from 50 mM Na-MOPS buffer, pH 6.3, and from 20 mMp-NPC, dissolved in the same buffer, by shaking with 10% (w/v) Chelex-100. The resin was removed by centrifuging at 1,000 x g for 2 min. Ca2+ was removed from CelD by shaking in the presence of 10% Chelex-100 followed by decantation. Alternatively, the enzyme was diluted in Chelex-treated buffer so that the contribution of protein-bound Ca 2+ in the assay medium was less than 5 x 10-8 M, assuming 3 mol of Ca 2+ bound/mol of CelD. No difference was observed between the results obtained with either procedure, even when no Ca'+ was added (data not shown).
Enzyme activity was assayed at 60 DC in 50 mMNa-MOPS buffer, pH 6.3. containing CaCI" EGTA, or ZnCI, as indicated for each experiment and 0.5-20 mMp-NPC as substrate. The reaction was stopped after less than 5% of the substrate had been hydrolyzed by adding Ys vol 1 M Na,CO s. 1 unit of activity is defined as the amount of enzyme liberating 1 umol ofp-nitrophenol (E = 1.61 X 10 4 em" X mol-i) per min. Protein concentration was measured using the Coomassie Blue reagent supplied by Bio-Rad (17), with bovine serum albumin as a standard.
Thermostability-Proteins were either treated with Chelex-100 or diluted so that their contribution to the concentration of Ca·+ in the inactivation reaction was less than 1.5 X 10-7 M. No difference was observed between the results obtained with either procedure. even when no Ca 2+ was added (data not shown).
Proteins were incubated at 75°C at a concentration of 3-5 10-8 Min 50 mM MOPS buffer, pH 6.3, containing CaCI., EGTA, or ZnCI. as indicated for each experiment. Temperature control was ascertained by checking the temperature inside of a plastic vial similar to those in which the inactivation reaction was performed. Samples were withdrawn at several time intervals and chilled on ice, and ZnCl 2 and CaCl 2 were added to a final concentration of 1 mM (2 mMCaCI. in the case of samples containing 1 mM EGTA). Residual activity was assayed as described above, using 0.9 mMp-NPC.
Computations-Kinetic constants (including the 95% confidence interval) for the rate of inactivation were computed from linear regressions of log (residual activity) versus time, using the Instat MaC® program (version 2.0, GraphPad Software). K m and k ca t values were calculated by non-linear regression using the KaleidaGraph® program (version 2.1, Abelbeck Software).

Crystallographic Analysis of Ca 2 + -binding Sites in CeLD-
The three-dimensional structure of CelD revealed four metalbinding sites occupied by atoms heavier than water in the crystal. A first internal site is located immediately behind a protein loop involved in substrate binding and catalysis (Zn sphere in Fig. 1). The tetrahedrical coordination by two Cys and two His side chains and the displacement by Hg suggests that this site is occupied by a Zn 2 + ion (6). The three other metal binding sites are located close to the molecular surface in different regions of the protein (spheres A, B, and C in Fig. 1). From the coordination geometry, these three positions could be identified as Ca 2 + -binding sites.
The coordination of the Ca 2 + ion bound at site A appears as a slightly distorted octahedral arrangement with a water molecule at one of the vertices ( Fig. 2A). Protein groups donate the five other oxygen ligands: two main chain carbonyls at positions 236 and 241 and the side chains of residues Asn-239, Asp.243, and Asp·246. The loop forming this site protrudes into the solvent and appears to be stabilized by calcium.
Seven oxygen atoms chelate the Ca 2 + ion at site B. In this case, the coordination polyhedron appears as a distorted pentagonal bipyramid with Asp-362 and a main chain carbonyl at position 401 on the vertices, or alternatively as a distorted octahedral arrangement with one bidentate ligand, Asp-361 (Fig. 2B). In addition to the aspartate residues, protein oxygens involved in Ca 2 + binding include the side chain ofThr-356 and the main chain carbonyl groups at positions 358 and 401. As shown in Fig. 2B, this site appears to have a structural role in linking together two different regions of the protein.
The protein loop forming site C is completely exposed to the solvent, with three out of the six oxygen ligands donated by water molecules (Fig. 2C). Main chain carbonyls at positions 520 and 525 and the carboxylate group of Asp-523 complete the calcium coordination polyhedron. Unlike sites A and B, the protein loop forming binding site C is partially involved in intermolecular interactions in the crystal. The side chain of Arg-314 from a neighbor molecule is stacked against Trp-526, and the carbonyl group at position 524 forms an intermolecular hydrogen bond with the guanido group of Arg·416 (data not shown). each other in loop conformation as well as in the side chains and the number of water molecules involved in the coordination polyhedra.
Overall, only small structural differences were observed for the structure of CelD at 0 and 300 mM calcium. The coordination geometry ofthe three sites was essentially the same within experimental error (Table 1). Only the temperature factors of the calcium atoms bound at sites A and C were different in the two crystal forms (the temperature factors for the three calcium atoms were 27, 25, and 32 A2, respectively, at 300 mM CaCI, and 43, 28, and 47 A2 at 0 mM CaC}), suggesting partial calcium occupancy of sites A and C in ammonium sulfategrown crystals.
Separation ofHigh and Low M; Forms ofCelD-A * and CelD-B*-SDS·PAGE analysis indicated that the wild type and the three mutant proteins were mainly composed of 65-kDa CeID, with 68-and 63-kDa CelD being present as minor species in some of the preparations (Fig. 3A). Previous work has shown that proteolysis accounts for some heterogeneity of the COOH terminus of CeID. However, cleavage does not affect the catalytic domain of the protein, and the 68-, 65-, and 63-kDa species were shown to share very similar catalytic properties (9,11,19).
In non-denaturing electrophoresis (Fig. 3B), CelD-C* displayed the same mobility as wild type CelD, which is a monomeric protein (3). However, CelD-A* and CelD-B* could be separated into a form with a mobility similar to that of the wild type monomer and a slower migrating, higher Me form, presumably resulting from self-association. The two forms could be separated by ion exchange chromatography on a Mono-Q column (Fig. 3B) or by gel filtration on a TSK G2000 column (data not shown) but tended to reequilibrate over a period of a few days. This explains the partial contamination of one form by the other seen in Fig. 3B.
Presence of Zn 2 + -Atomic absorption spectroscopy showed the presence of 1.0::':: 0.2 mol ofZn 2+/mol of wild type CelD. No change in Zn 2 + content was detected when the enzyme was incubated for 15 min at room temperature or at 60 DC in the presence of 10% (w/v) Chelex-l00, but incubation with Chelex at 75 DC for 9 min resulted in total loss of detectable enzyme- Sites Band C are close to either end of the substrate-binding groove and are expected to have some influence on the catalytic activity of CelD. On the opposite site of the a-barrel, the Ca 2 + ion bound at site A stabilizes a helix-connecting loop with no obvious role in enzymatic activity. As a general rule, the conformation of the loops forming the three Ca 2 + -binding sites does not follow the EF-hand pattern observed in many Ca 2 + . binding proteins (18). Moreover, they differ significantly from bound Zn 2 + (dat a not sh own ). Dissociation of Zn 2 + was correlated with an increase in the inactivation rate of the protein (see below ).
Ca 2 + -Bind ing Parameters- Fig. 4 sh ows the Scatchard analysis of Ca 2 + binding to the wild type and variou s mutant for ms of CelD . The binding isoth erm of wild type CelD was compatibl e with the presenc e of two high affinity sites (K a = 2 X 10 6 M -1 ) a nd one low affinity site (K a = 0.66 X 10 5 M -1 ) per molecule. CelD-A* (F ig. 4A) and CelD -B* (F ig. 4B ) eac h di splayed one high affi nity site with K a = 5.1 X 10 6 M-1 an d K a = 3.2 X 10 6 M -1, resp ectively . CelD-C* displayed two high affinity sites with K a = 3.1 X 10 6 M-1 (F ig. 4C ). By contrast to the wil d type, no low a ffinity site was detected in any of the mutant CelD proteins. Ca 2 + -bin ding isotherms were t he same for the low and high M; forms of CelD-A* and CelD-B* (Fig. 4, A  a n d B ).
Effect of Ca 2 + on Kineti c Parameters-Previous data in dicated that Ca 2 + decreased the K", but had little effect on the k en t of CelD (11). Fig. 5A confirms that ad dit ion or removal of Ca 2 + h ad little effect on the k ent of CelD and indicates that the str onge st decrease in K m (fr om 6.2 to 3.5 rna) occurred when the Ca 2 + concentration was increa sed from 5 X 10 -6 to 10 -4 M . As a con sequence, there was a concomitant incr ease in catalytic effici enc y ken/K'n" Addition of 1 mM EGTA had little effect on CelD after Ca 2 + ion s had be en removed by dilution in Ca 2 + -free buffer. Fig. 5B shows that the kinetic paramet ers of CelD-C* were not affecte d by EGTA nor by Ca 2 + in the range of conce ntrations tested.
Thermostability- Fig. 6 shows the kin et ic r a t e of inactivatio n ki n net of wild type CelD a n d of CelD-C* inc u bated at 75°C in t he presence of 1 mM EGTA or va rio us Ca 2 + conce ntrations. Addition of 1 mM EGTA in t he inactivation reaction after Ca 2 + ions h ad been removed by dilution in Ca 2 + -fr ee buffe r or by Che lex treatment at room te mperature resu lted in a 2.4-fold increa se in the rate of inactivation of both enzymes. Addition of Chelex-100 at 75°C produced a simi lar effect (dat a not shown).
For the wild type en zyme, increasing the concentration of Ca 2 + u p to 5 X 10-5 M h ad no significant effect on the rate of inactivation. However, a l.8-fold decr ea se in k in net was observed upon increasing the concentration of Ca 2 + from 5 X 10 -5 to 10 -3 M. Over the same range ofCa 2 + concentration , the inactivation rate of CeID-C* was not affected. DISCUSSION The presence of one Zn 2 + ion/m ol ofCelD, predicted from the crystallographic analysis of the protein, was confirmed by biochemical a nalys is . Zn 2 + binding a ppeared quite stable at room temperature and at 60°C, and dissociati on of Zn 2 + at 75°C was accompanied by rapid den atura tion of the enzyme. By contrast, Ca 2 + could be dissociated fro m CelD without den aturing the protein.
Previous interpretation of Ca 2 + bin ding dat a had led to the conclusion that CelD contained two high affinity Ca 2 +-binding sites (11). Points extending beyo n d two sites/molecu le in t he Scatchard plots were not cons idered in t he analysis. However , crystallographic a nalysis revealed the presen ce of three pu t ative Ca 2 +-binding sites in CelD (6). The presen ce of three functional Ca 2 +-bin ding sites was confi r me d by t he a nalysis of CelD-C *, whose mutation affects site C. The Ca 2 +-binding isotherm of CelD-C * displ ayed two high a ffinity sites si milar to those of t he wild type, but, in contrast to the wild ty pe, bin ding did not exceed 2.1 mol of Ca 2 + bou nd/mol of pr otein . This suggests that in t he wild ty pe, poin t s extendi ng betw een 2 and 3 mol of Ca 2 + bound/mol of pr ot ein were du e to the presen ce of sit e C, which be haved like a low affinity site. High affinity Ca 2 + binding to sit es A and B was confi rmed by a nalys is ofCelD-A* and CelD-B *. The Ca 2 + -bin ding isother ms of both prot ein s showed t hat eac h mutation abo lis he d high affinity binding to on e site. The r elati ve affi n it ies of sites A, B, and C were consistent with the fact that in sites A and B, the protein contribut es five a nd six , respectively, of t he coordinating oxygens but only three of t he coor dinating oxyge ns of site C.
Mutagenesis of site A or B seeme d to abolish bin ding to site C, as if site C cou ld form only whe n both sites A and B a re occu pied . Why this should be the ca se is not obvious from structural analysis.
Invest igat ion of the kinetic parameter s of CelD in dicated that the change in K m of t he enzyme as a fu nctio n of t he Ca 2 + concentration was strongest between 5 X 10-6 an d 10 -4 M. Th is r a nge is most like ly acco unted for by the increased occupancy of t he low affi nity sit e C ra th er t han t he high affi n ity sites A an d B. The fact that the kinetic parameter s of CelD-C* were n ot affected by Ca 2 + confirms t his interpretation.
The stabilization of wild ty pe CelD occurred at concentratio ns that were a n order of magn itude high er t han t hose requ ired to affect catalytic pa r am et ers . This may be ex plained by the fact that changes in catalytic properties in duce d by Ca 2 + dissociat ion are reversible, whe reas the rmal den a tur ati on is not. The Ca 2 + concentrations at wh ich stabilization was observed were consistent with a requirement for occu pa n cy of site C rather than site A and B. Accor dingly, inactivation of site C abolished Ca 2 +-induc ed stabili zation of CelD . The fact that Ca 2 + binding to site C enhanced the substrate binding affinity and stabilized the conformation of the catalytic site is consistent with the close vicinity of the two sites. The FIG. 6. Inactivation rate k i nac t at 75°C of wild type CelD and CelD·C* as a function of divalent metal concentration. The firstorder inactivation rate was determined as described under "Materials and Methods." Results are presented using a double logarithmic scale. Closed circles, wild type CelD; open circles, CeID-C*. Error bars indicate the 95% confidence interval for each determination. Ca 2 + was removed from wild type CelD by treating with Chelex-100 and from CelD-C* by diluting into Chelex-100-treated buffer (contribution of Ca 2 + initially bound to the enzyme added to the assay was < 1.5 X 10-7 M). Except for the EGTA-treated samples, all samples contained 1 !JMZnCl 2 in addition to the Ca 2 + concentrations indicated. loop containing the Ca 2 + -coordinating residues Ser-520, Asp-523, and l1e-525 is connected to the substrate-binding residues His-516 and Arg-518. His-516 and Arg-518 formed hydrogen bonds with hydroxyl groups of the inhibitor o-iodobenzyl-I3-Dcellobioside in the crystal structure of the enzyme-inhibitor complex (6). In addition, chemical modification and mutagenesis studies identified His-516 as an important residue of the catalytic center (20).
The self-association of monomeric CelD-A* and CelD-B* into a high Mr> presumably dimeric form was not correlated with the occupancy of Ca 2 + -binding sites. For both proteins, addition of Ca 2 + or EGTA during non-denaturing electrophoresis failed to alter the proportion of the two forms (data not shown). Both forms displayed very similar Ca 2 + -binding isotherms. Self-association did not seem to influence thermostability nor kinetic parameters (data not shown). However, the compound effects of site A and B mutations on site C precluded a straightforward analysis of the influence of Ca 2 + on the stability and kinetic properties of the mutant enzymes.
Unlike catalytic residues, none of the residues involved in Ca 2 + binding is strictly conserved among all catalytic domains of family E cellulases. At present, it is difficult to predict from sequence analysis which of the other members of family E may be stabilized in a similar manner by Ca 2 + . It would be of interest to know whether the presence of functional Ca 2 + _ binding sites is correlated with the thermostability of the enzymes.