Calcium Affinity, Cooperativity, and Domain Interactions of Extracellular EF-hands Present in BM-40*

The structure and function of cytosolic Ca2+-binding proteins containing EF-hands are well understood. Recently, the presence of EF-hands in an extracellular protein was for the first time proven by the structure determination of the EC domain of BM-40 (SPARC (for secreted protein acidic and rich in cysteine)/osteonectin) (Hohenester, E., Maurer, P., Hohenadl, C., Timpl, R., Jansonius, J. N., and Engel, J. (1996) Nat. Struct. Biol. 3, 67–73). The structure revealed a pair of EF-hands with two bound Ca2+ ions. Two unusual features were noted that distinguish the extracellular EF-hands of BM-40 from their cytosolic counterparts. An insertion of one amino acid into the loop of the first EF-hand causes a variant Ca2+ coordination, and a disulfide bond connects the helices of the second EF-hand. Here we show that the extracellular EF-hands in the BM-40 EC domain bind Ca2+ cooperatively and with high affinity. The EC domain is thus in the Ca2+-saturated form in the extracellular matrix, and the EF-hands play a structural rather than a regulatory role. Deletion mutants demonstrate a strong interaction between the EC domain and the neighboring FS domain, which contributes about 10 kJ/mol to the free energy of binding and influences cooperativity. This interaction is mainly between the FS domain and the variant EF-hand 1. Certain mutations of Ca2+-coordinating residues changed affinity and cooperativity, but others inhibited folding and secretion of the EC domain in a mammalian cell line. This points to a function of EF-hands in extracellular proteins during biosynthesis and processing in the endoplasmic reticulum or Golgi apparatus.

The EF-hand is a highly conserved Ca 2ϩ -binding motif found in a large number of intracellular proteins. EF-hands, as present in calmodulin, troponin C, and calcineurin, are responsible for signal transduction by the second messenger Ca 2ϩ , leading to the activation or inactivation of numerous and diverse target proteins and thus influencing a wide range of cellular processes (1)(2)(3)(4). The term EF-hand was introduced by Kretsinger and Nockolds (5) in describing the structure of parvalbumin. Two helices (E and F in parvalbumin) flank a loop of 12 amino acids in which the Ca 2ϩ ion is 7-fold coordinated in a pentagonal-bipyramidal arrangement. The Ca 2ϩ ion is coordinated by side chain oxygen atoms at the X, Y, and Z vertices, by a backbone carbonyl at ϪY, by a water molecule at ϪX, and by a carboxylate (mostly glutamate) at ϪZ, which makes a crucial bidentate interaction with the metal ion (see Fig. 1). This canonical EF-hand structure was subsequently found in a number of proteins of the Ca 2ϩ second messenger system. Usually, EFhands occur in pairs, although in some cases one or even both EF-hands in a pair may no longer be functional because of mutations of Ca 2ϩ -coordinating residues (for reviews see Refs. 6 -8).
Several variations of the canonical Ca 2ϩ -binding mode are known. In the S100 subfamily of EF-hand proteins, two amino acids are inserted in the N-terminal EF-hand loop, while the C-terminal EF-hand is of the canonical type. The insertions lead to a rearrangement in the loop, such that the Ca 2ϩ ion is coordinated by backbone carbonyl oxygens at positions X, Y, Z and -Y, with the bidentate ligand at -Z providing the only side chain oxygen atoms (9). Despite the unusual coordination of this site, the affinities of the two EF-hands for Ca 2ϩ , e.g. in calbindin D9k, are nearly identical (10). A further variation of the canonical EF-hand was found in myosin essential light chain. The loop in the first EF-hand of the EF-hand pair also contains two inserted amino acids but at different positions compared with S100 proteins. In essential light chain the insertions result in a more pronounced rearrangement of Ca 2ϩ ligands, with several residues coordinating Ca 2ϩ with both their backbone and side chain oxygens (11,12). Finally, in the Ca 2ϩ -binding domain of calpain, the first of the five EF-hands binds Ca 2ϩ noncanonically. This EF-hand lacks the typical side chains at positions X and Y and uses a carbonyl oxygen and a second water molecule instead (13,14).
BM-40 (SPARC (for secreted protein acidic and rich in cysteine)/osteonectin) is the only example of an extracellular EFhand protein that is secreted by virtue of a classic signal peptide and for which a structure determination has proved the presence of EF-hands (15,16). BM-40 is an abundant glycoprotein and has been suggested to participate in the modulation of cell-matrix interactions, bone mineralization, wound repair, and angiogenesis (17). It is highly expressed in some malignant tumors and has been reported to play a crucial role in the tumorigenicity of human melanomas (18). Developmental anomalies are induced by overexpressing or suppressing BM-40 in nematodes (19,20). Furthermore, microinjection of BM-40 RNA, protein, peptides, or antibodies against BM-40 interfere with Xenopus embryonic development (21,22). In contrast, deletion of the BM-40 gene in mice resulted in a relatively mild phenoptype, namely late onset cataract formation in the eye, whereas development proceeded normally (23,24). However, further defects, such as severe osteopenia, are under investigation (25). It is suspected that homologous pro-teins, such as SC1, QR1, TSC-36, testican-1, or testican-2 (26 -30) may compensate for several of the reported functions of BM-40 during murine embryonic development.
BM-40 is a modular protein built from three domains. A short N-terminal region (domain I), rich in glutamic acid residues, binds several Ca 2ϩ ions with low affinity (K d ϭ 5-10 mM) (31). This is followed by a region homologous to follistatin (FS domain) and an autonomously folding C-terminal Ca 2ϩ -binding EC domain. A single EF-hand was predicted in the EC domain of BM-40 (32), and Ca 2ϩ binding was interpreted as being to a single site (31,33,34). However, the structures of the EC domain and the pair of FS and EC domains revealed a pair of EF-hands in the EC domain, each with a bound Ca 2ϩ ion (15,16). Both EF-hands in BM-40 are of unusual structure; a oneresidue insertion into the Ca 2ϩ -coordinating loop of the first EF-hand is accommodated by a cis-peptide bond, which allows a peptide carbonyl to substitute for a Ca 2ϩ -binding side chain. This novel variation explained why even sophisticated search algorithms failed to detect both EF-hands in BM-40. The second EF-hand has a canonical pattern of Ca 2ϩ -coordinating residues but is circularized by a disulfide bond between the flanking helices. A collagen-binding site in the EC domain of BM-40, which is only active in the presence of Ca 2ϩ , has been characterized by an alanine mutagenesis screen (35).
One objective of the present investigation was to determine the Ca 2ϩ binding properties of the EF-hands of BM-40. Both the variant and the canonical EF-hand were found to bind Ca 2ϩ with high affinity, and there is positive cooperativity between the two EF-hands. In addition, interactions with the neighboring FS domain strongly influence the Ca 2ϩ affinity of the EC domain. Some mutations in the EF-hands designed to abolish Ca 2ϩ binding interfered with protein secretion, indicating an essential role of Ca 2ϩ binding for folding and secretion.

EXPERIMENTAL PROCEDURES
Nomenclature of Proteins and Mutants-The sequence numbering of human BM-40 (36) used starts at the N terminus obtained after cleavage of the signal peptide. Secreted recombinant full-length human BM-40 thus corresponds to amino acid residues 1-286. The designation FS-EC (residues 51-286) is used for the fragment encompassing the FS and EC domains but lacking the N-terminal acidic domain. The EC domain encompasses residues 136 -286. The sequence of both EF-hand loops within the EC domain and the point mutations introduced are shown in Fig. 1. The nomenclature for the point mutants includes the domain context, the corresponding EF-hand, and the substitution. For instance, EC-EF1-E234K describes a point mutation introduced in the EC domain in which Glu 234 in EF-hand 1 (EF1) occupying position ϪZ of the pentagonal bipyramid is changed to lysine. The fragment FS-EC-⌬C lacks helix C (residues 196 -203) of the EC domain (35).
Construction of BM-40 Mutants-The point mutant EC-EF1-E234K was produced by the method of Deng and Nickoloff (37) using the transformer site-directed mutagenesis kit (CLONTECH) and the cDNA encoding the EC domain of human BM-40 (34). The mutation primer 5Ј-CCCACACCAAGCTGGCGCCACTGCGT contained the change of the GAG codon for Glu 234 to the AAG codon of lysine. An additional silent T to G mutation resulted in a novel NarI site which was used to identify mutated clones. Mutagenesis was performed as described by the manufacturer (CLONTECH, CA). Production and purification of full-length BM-40, deletion mutants FS-EC and EC, and mutants BM-40-EF2-D257K and BM-40-EF2-E268K was done as described previously (33,34). cDNA fragments (269 base pairs) containing the mutations in EF2 were obtained by KpnI/XhoI restriction cleavages of BM-40-EF2-D257K and BM-40-EF2-E268K (33) and were cloned into the cDNA of the EC domain resulting in EC-EF2-D257K and EC-EF2-E268K. The double mutant EC-EF1,2-E234K,E268K was produced by subcloning the 136-base pair NcoI/XhoI fragment of EC-EF2-E268K into EC-EF1-E234K fragment. NheI/XhoI fragments of EC-EF1-E234K, EC-EF2-D257K, EC-EF2-E268K, and EC-EF1,2-E234K,E268K were excised and ligated with the NheI/XhoI restricted episomal expression vector pCEP-Pu (38). This vector contains the signal peptide of BM-40 and enables the secretion of recombinant protein into the cell culture media. Correct mutagenesis and cloning was verified by cycle sequencing using the Alfexpress DNA sequencer (Amersham Pharmacia Biotech).
Human embryonic kidney cells (EBNA-293) were transfected with 20 g of pCEP-Pu vectors using LipofectAMINE (Life Technologies, Inc.) or electroporation. Cells were selected with 0.5 g/ml puromycin in Dulbecco's modified Eagle's medium/Ham's F-12 medium with 10% fetal calf serum for 14 days as described (38). For purification of proteins from the cell culture media, cells were maintained under serumfree conditions and culture medium was collected every 2 days.
Reverse Transcription-Polymerase Chain Reaction of Transfected Cellular mRNA-Transfected and nontransfected EBNA-293 cells were trypsinized from confluent plates, and 10 7 cells were used for isolation of mRNA performed according to the manufacturer's protocol (Amersham Pharmacia Biotech). mRNA concentrations were determined using UV spectroscopy. RT-PCR 1 was performed in a one-tube reaction by addition of 1 unit of reverse transcriptase and 1 unit of Taq polymerase in RT-PCR buffer (0.1 M Tris/HCl, pH 8.4, 0.5 M KCl, 25 mM MgCl 2 , 1 mM dithiothreitol, 0.2% gelatin, 6.25% W1, 2.5 mM of each dNTP). After addition of 100 ng of mRNA, 50 pmol of sense primer 5Ј-GTTCCCAG-CACCATGAGGG and 50 pmol of antisense primer 5Ј-GATCCTCGAGT-TAGATCACAAGATCC reverse transcription was allowed to proceed for 25 min at 37°C. Five cycles of polymerase chain reaction were then performed at an annealing temperature of 50°C followed by 30 cycles at an annealing temperature of 65°C. Denaturation and elongation temperatures were 95 and 72°C, respectively. Half of the reaction product was subjected to electrophoresis on a 1.5% agarose gel and stained with ethidium bromide. In controls reverse transcriptase was omitted.
Immunoblot of Transfected EBNA-293 Cells-Transfected EBNA-293 cells of confluent plates (diameter, 10 cm) were washed with phosphate-buffered saline and lysed with 1 ml of 0.5% Triton X-100. Aliquots of serum-free culture medium (1 ml) were precipitated with trichloroacetic acid. The pellets were resuspended in reducing sample buffer. Resuspended pellets and cell lysates were subjected to electrophoresis in 15% SDS-PAGE gels according to Laemmli (39). Proteins were transferred onto nitrocellulose, and blocking was done with 5% milk powder solution. A rabbit polyclonal antiserum against human BM-40 (40) was used as a primary antibody, and detection was performed with a goat antibody against rabbit IgG coupled to peroxidase (DAKO, Denmark) and the enhanced chemoluminescence kit (Amersham Pharmacia Biotech).
Purification of EC-EF1-E234K and EC-EF2-D257K-Serum-free cell culture media (about 2.5 liters) were dialyzed against 50 mM Tris/HCl, pH 8.6, and subjected to anion exchange chromatography on DEAE-Sepharose. Elution was performed with a linear salt gradient from 0 to 1 M NaCl. After dialysis of pooled fractions, recombinant EC domains were separated on MonoQ (Amersham Pharmacia Biotech) equilibrated in the same buffer. Elution was done with a linear salt gradient (0 -0.5 M NaCl, containing 2 M urea). EC domains eluted in a broad peak between 0. 15  Residues coordinating Ca 2ϩ with side chain oxygen atoms are boxed in black, whereas residues using backbone carbonyl oxygens are boxed in gray (16). The Ca 2ϩ coordination of canonical EF-hands is indicated by letters denoting the position of the coordinating residues in the pentagonal-bipyramid. The glutamic acid at position ϪZ chelates Ca 2ϩ with both oxygens of its side chain. Note that EF1 is a variant of the canonical EF-hand motif. The insertion of His 224 induces the coordination of Ca 2ϩ via the backbone carbonyl oxygen of Pro 225 . was achieved on a Resource Q column followed by extensive dialysis against the appropriate buffer.
Circular Dichroism-Proteins were dialyzed against 5 mM Tris/HCl, pH 7.4, or 50 mM Tris/HCl, pH 7.4, 150 mM NaCl (TBS). Protein concentrations were calculated from the absorption at 280 nm as described (34). For the EC domain and EC domain mutants, an extinction coefficient of 1.39 ml mg Ϫ1 cm Ϫ1 was used (41). CD spectra in the far UV region were recorded in a thermostated quartz cell of 1-mm optical path length in a Jasco 715 CD spectropolarimeter. Spectra were measured with the protein dissolved in the appropriate dialysis buffer, after adding 2 mM CaCl 2 , and after subsequent addition of 3 mM EDTA. Molar ellipticities [⌰] (expressed in degrees cm 2 dmol Ϫ1 ) were calculated on the basis of a mean residue molecular mass of 110 Da. Ten spectra were accumulated to improve the signal/noise ratio. Spectra of buffers were subtracted. The percentage of change in circular dichroism at 222 nm was calculated as ⌬ Ca 2ϩ titrations were performed with the EGTA/CaEGTA buffering system as described by Tsien and Pozzan (42). 100 mM EGTA and 100 mM CaEGTA stock solutions were purchased from Molecular Probes. An equilibrium dissociation constant K d ϭ 35 nM for the CaEGTA complex was used for the titrations in TBS. Proteins (about 1 M) were dissolved in buffer containing 10 mM EGTA, and fluorescence intensity was measured. Defined aliquots of the solution were removed and substituted by the same volume of protein dissolved at the same concentration in 10 mM CaEGTA, thus increasing the free Ca 2ϩ concentration while keeping the protein concentration constant. This was repeated until finally the fluorescence intensity was measured for protein in 10 mM CaEGTA. Alternatively, proteins were dissolved in buffer containing 10 mM EGTA, and small aliquots of 100 mM CaEGTA stock solutions were added. In this instance, fluorescence intensity was corrected for dilution. Recombinant BM-40 mutants with low affinity for Ca 2ϩ (K d Ͼ20 M) were directly titrated with CaCl 2 in the absence of EGTA. The degree of saturation Y was calculated from the signal S as Y ϭ (S Ϫ S 0 )/(S Ca Ϫ S 0 ), where S Ca and S 0 represent the signals at Ca 2ϩ saturation and zero Ca 2ϩ , respectively.
Data Evaluation-Two potential mechanisms for binding of Ca 2ϩ to BM-40 and its mutants were considered. The single-site model describes the change of the measured signal as a function of the binding of a single Ca 2ϩ ion to the protein as shown in the following equation.
represents the free Ca 2ϩ concentration and K d represents the equilibrium dissociation constant. For a protein containing two binding sites, which may be either independent (no cooperativity present) or interacting (cooperativity present), four microscopic equilibrium dissociation constants are needed to fully describe the situation. However, the spectroscopic methods used do not allow for discrimination of Ca 2ϩ binding to the individual sites, and thus the four dissociation constants cannot be resolved (44). We therefore used a macroscopic two-site model with two Ca 2ϩ -binding sites, in which binding of one Ca 2ϩ is described by the macroscopic equilibrium dissociation constant K D1 followed by binding of a second Ca 2ϩ with K D2 .
This model also allows detection of cooperativity between the two sites. If positive cooperativity is present, K D2 will be less than 4 ϫ K D1 , because the second Ca 2ϩ ion binds with higher affinity when the first site is occupied than when it is empty. If there is no positive cooperativity, K D2 will be equal or greater than 4 ϫ K D1 .
The program COSY (43) or Grafit (Erithacus Software) were used to fit both models to the experimental data with nonlinear least square fit procedures. Cooperativity in a two-site system can be characterized by ⌬⌬G, which gives the difference in free binding energies between the binding to a given site when the other site is occupied and when the other site is empty. For calculation of this free energy of interaction, ⌬⌬G, knowledge of the microscopic equilibrium dissociation constants is required. However, these are not directly accessible from Ca 2ϩ titrations. A lower limit of the free energy of interaction ⌬⌬G can be calculated from the macroscopic K D values with ⌬⌬G ϭ1 ϭ ϪRT ln (4⅐K D1 /K D2 ) (44). ⌬⌬G ϭ1 is identical to ⌬⌬G if the microscopic affinity of both sites is identical. If microscopic affinities are not identical, ⌬⌬G is always smaller than ⌬⌬G ϭ1 .

BM-40 Binds Ca 2ϩ with High Affinity and Cooperativity-
The affinity of EF-hands for Ca 2ϩ varies for the different cytosolic members of the EF-hand family with dissociation constants (K d ) ranging from nM to M values. Our first aim was to determine the affinities of the EF-hands of BM-40. We were unable to obtain BM-40 in a sufficiently Ca 2ϩ -free form either by dialysis against EDTA or Chelex-100 or by chromatography on an EDTA-agarose column (data not shown). We therefore used an EGTA/CaEGTA system (42) to follow Ca 2ϩ binding. Briefly, this method relies on the use of two stock solutions of 100 mM EGTA and 100 mM CaEGTA, respectively. The latter is exactly titrated using a pH-metric method. Using different ratios of EGTA and CaEGTA, the free Ca 2ϩ concentrations in a solution can be adjusted with high precision to [Ca 2ϩ ] free ϭ K d ϫ [CaEGTA/EGTA]. Knowledge of the concentrations of residual Ca 2ϩ from buffer solutions and bound to proteins (usually between 1-20 M) is not necessary because this is buffered by the excess EGTA (normally 10 mM) in the system. For a residual Ca 2ϩ concentration of 20 M the error in the calculated free Ca 2ϩ concentration is less than 2%.
The reversible change in intrinsic fluorescence of full-length BM-40 (36) was used to monitor Ca 2ϩ binding. A simple model of one Ca 2ϩ ion binding to BM-40 could not sufficiently describe the measured values, which evidently follow a curve with a steeper slope around the midpoint than predicted from the single-site model. An appropriate description was achieved with the two-site model, which enabled the determination of two macroscopic K D values of K D1 ϭ 490 nM and K D2 ϭ 57 nM, respectively ( Fig. 2A). The fact that K D2 is much smaller than 4 ϫ K D1 clearly demonstrates that positive cooperativity is present because of interaction between the sites.
Domain Interactions with the FS Domain Influence Ca 2ϩ Affinity-The change in fluorescence of the deletion mutant FS-EC was also investigated in the EGTA/CaEGTA system. Cooperative high affinity binding of Ca 2ϩ to FS-EC was observed (Fig. 2B). The affinity of FS-EC was very similar to that of full-length BM-40 with macroscopic K D values of 470 and 26 nM. This indicates that the acidic domain I does not significantly influence Ca 2ϩ binding to the EC domain. In contrast, removal of both domain I and the FS domain, resulting in the isolated EC domain, strongly decreased Ca 2ϩ affinity. Macroscopic K D values of 11750 and 56 nM were obtained as best fit parameters for the EC domain (Fig. 2B). Ca 2ϩ affinity is on average 6-fold lower if one compares (K D1 K D2 ) 0.5 , which corresponds to a loss in total free binding energy of 9.9 kJ/mol (Table I).

Point Mutations in EF2 Disturb Folding and Secretion of the EC Domain-To investigate the influence of the individual
EF-hands on Ca 2ϩ affinity and cooperativity we mutated the conserved glutamic acid at position ϪZ to lysine, a mutation recently shown to strongly decrease affinity of EF-hands (45)(46)(47)(48). Because the two disulfides of the EC domain are essential for proper folding, all proteins were expressed in a eukaryotic expression system. When analyzing cell culture supernatants of transfected EBNA-293 cells recombinant EC domains with mutations EC-EF1-E234K and EC-EF2-D257K were readily seen in an immunoblot (Fig. 3). However, EC-EF2-E268K as well as the double mutant EC-EF1,2-E234K,E268K could not be detected in the medium (Fig. 3). The presence of similar amounts of endogenous BM-40 shows that the absence of the two mutants was not caused by cell death or a general inhibition of the secretory pathway. To prove successful transfection and selection of the EBNA-293 cells, we isolated mRNA from transfected cells and performed RT-PCR with primers specific for the recombinant EC domain. For all transfected cell lines the 540-base pair cDNA fragment could be amplified demonstrating the presence of the respective mRNAs (Fig. 4A). mRNA of endogenous BM-40 was not amplified as the reverse primer partially spanned noncoding region of the vector sequence. Next we prepared cell lysates of the nonsecreting EC-EF2-E268K and EC-EF1,2-E234,268K cells. In an immunoblot a faint band for both proteins could be detected (Fig. 4B). The intensity of the bands was comparable with that of the endogenous BM-40 and presumably represents the small amount of protein that is translated and processed in the endoplasmic reticulum and/or Golgi apparatus.

Point Mutations in EF-hands Abolish Cooperativity of Ca 2ϩ
Binding-Both EC-EF1-E234K and EC-EF2-D257K were produced and could be purified to homogeneity from cell culture media (Fig. 5), even though expression levels were decreased compared with the nonmutated EC domain. The proteins eluted in broad peaks from ion exchange columns, and as a consequence, yields were very low (Ͻ25 g/liter of medium). Nonetheless, circular dichroism spectra demonstrated that the mutated EC domains in the Ca 2ϩ -free form had similar folds to that of the nonmutated EC domain (Fig. 6). Upon addition of Ca 2ϩ , the EC domain showed a 39% decrease in molar ellipticity at 222 nm, whereas only a 28% decrease was observed for EC-EF1-E234K. The conformational change was even more compromised in mutant EC-EF2-D257K, where the signal changed by only 10%.
In fluorescence studies, EC, EC-EF1-E234K, and EC-EF2-D257K showed emission maxima at about 340 nm when excited at their excitation maxima at 280 nm (data not shown). Titration of EC-EF1-E234K and EC-EF2-D257K revealed no change in fluorescence at submicromolar Ca 2ϩ concentrations. However, changes in intrinsic fluorescence intensities could be observed when Ca 2ϩ concentrations were raised significantly (Fig. 7). In contrast to results obtained with the EC domain, analysis of Ca 2ϩ titrations of EC-EF1-E234K and EC-EF2-D257K revealed no sign of cooperativity. A model assuming a single binding site with K d ϭ 45 M for EC-EF1-E234K and K d ϭ 205 M for EC-EF2-D257K, respectively, could adequately describe the data (Fig. 7).
The mutations EF2-D257K and EF2-E268K have been introduced previously into full-length BM-40 (BM-40-EF2-D257K  and BM-40-EF2-E268K) (33). To analyze cooperativity in the context of the full-length protein, we reinvestigated Ca 2ϩ binding to these mutants. The fluorescence change upon Ca 2ϩ addition could be described with a model assuming a single binding site, and a K d ϭ 0.22 M for BM-40-EF2-D257K was evaluated (Fig. 8), confirming the earlier results (33). For BM-40-EF2-E268K, the Ca 2ϩ affinity was clearly decreased and not cooperative (K d ϭ 1.2 M). In the mutant FS-EC-⌬C, the helix C is lacking in the EC domain. This mutation leads to enhanced collagen binding to the EC domain (35). The structure of FS-EC-⌬C was recently solved and showed two Ca 2ϩ ions bound to the EF-hands (35). The affinity of FS-EC-⌬C for Ca 2ϩ was 10-fold reduced relative to FS-EC when (K D1 K D2 ) 0.5 are com-pared, but the cooperativity between the sites was enhanced by the deletion (Fig. 8 and Table I).

DISCUSSION
Cytosolic EF-hand proteins are designed to fulfill different functions at low Ca 2ϩ concentrations in the resting cell (about 0.1 M Ca 2ϩ ) and when the cell has been activated (about 10 M Ca 2ϩ ). Their affinity for Ca 2ϩ is accordingly tuned to this concentration range, allowing the proteins to switch from Ca 2ϩfree to Ca 2ϩ -bound conformations upon activation. BM-40 is the only example so far of an extracellular protein containing EF-hands of which the three-dimensional structure is known. The role of Ca 2ϩ -binding to extracellular EF-hands is rather mysterious, because the extracellular free Ca 2ϩ concentration is high (about 1.2 mM) and believed to be constant (49). The experiments presented here were designed to shed some light on the possible function(s) of extracellular EF-hands.
We have used a Ca 2ϩ buffering system that allows precise adjustment of Ca 2ϩ concentrations in the nanomolar range to examine Ca 2ϩ binding by BM-40. The two EF-hands in BM-40 were found to bind Ca 2ϩ with a free energy of Ϫ77.3 kJ/mol. We used two models for analyzing the binding data. The first model assumes a single Ca 2ϩ binding site and is described by the microscopic equilibrium dissociation constant K d . The second model assumes two binding sites and is described by two macroscopic dissociation constants, K D1 and K D2 . The values obtained for K D1 and K D2 allow to detect positive cooperativity between the sites. If Ca 2ϩ binding to any site is independent from the occupation of the other site (i.e. no cooperativity), K D2 will be equal or greater than 4 ϫ K D1 . If positive cooperativity is present between the sites, K D2 is smaller than 4 ϫ K D1 because the second Ca 2ϩ ion binds with higher affinity when one site is already occupied by a Ca 2ϩ ion. This model is an alternative to the use of the Hill equation but allows more detailed insight into the binding process and the energetics of binding (44).
The experimental results for calcium binding to BM-40 could only be fitted satisfactorily with the two-site model. The macroscopic dissociation constants, K D1 ϭ 490 nM and K D2 ϭ 57 nM, demonstrate that binding of the second Ca 2ϩ ion is facilitated by positive cooperativity. A minimum interaction energy of Ϫ8.8 kJ/mol can be calculated from these values (Table I). Because a signal specific for EF1 or EF2 cannot be obtained by circular dichroism and fluorescence spectroscopy, it is not pos- sible to know whether the EF-hands become occupied sequentially in a defined order or whether they bind Ca 2ϩ in parallel. The macroscopic dissociation constants can therefore not be attributed to the individual EF-hands.
The failure to detect cooperative binding in our previous studies (33,34) is now seen to have been caused by incomplete removal of Ca 2ϩ from BM-40 and buffer solutions. Thus, Ca 2ϩ titrations were started with partially saturated proteins, and only the second halves of the saturation curves were recorded. At the time, only one binding site had been predicted for BM-40 (32,50), and we therefore used a simple single-site model for data evaluation, which happened to describe the experimental data reasonably well. Of course, the data could also be fitted with more complex models including cooperativity, but these models did not produce a significantly better fit. In the present study, high accuracy in the low nanomolar range could be achieved through the use of the EGTA/CaEGTA system, finally allowing the single-site and the two-site model with cooperativity to be clearly distinguished.
BM-40 is half-saturated at a free Ca 2ϩ concentration of 170 nM. As already mentioned, free Ca 2ϩ concentrations in serum and several other extracellular tissue fluids are much higher (1.2 mM), but local gradients or changes in the extracellular Ca 2ϩ concentration down to 0.1 mM have been observed (49,51,52). Nevertheless, even at the lowest extracellular Ca 2ϩ concentrations reported to date, the EF-hands in BM-40 will be in the fully Ca 2ϩ -saturated form. This indicates that Ca 2ϩ bound to the EF-hands in BM-40 is likely to play a structural rather than a regulatory role. It is known that the structure maintained by Ca 2ϩ is essential for the binding of BM-40 to collagens, and the collagen-binding site has indeed been mapped to the EC domain (34,35).
Nearly identical Ca 2ϩ binding profiles were observed for the FS-EC domain pair and for full-length BM-40, showing that the acidic N-terminal domain I does not influence the affinity and cooperativity of Ca 2ϩ binding to the EC domain. In marked contrast, when both domain I and the FS domain were deleted the free energy of Ca 2ϩ binding was increased by 7.8 kJ/mol to Ϫ69.5 kJ/mol, demonstrating a strong interaction between the FS and EC domains. The first macroscopic dissociation constant, K D1 , was increased about 20-fold compared with fulllength BM-40, whereas the second constant, K D2 , was not affected. Interestingly, the cooperativity between the EF-hands, as measured by ⌬⌬G, is higher in the isolated EC domain relative to FS-EC and full-length BM-40 (Table I). Presumably, the strong interaction between the FS and EC domain in FS-EC restricts the conformational changes in the EF-hands that accompany Ca 2ϩ binding, whereas in the isolated EC domain binding of the first equivalent of Ca 2ϩ leads to a relatively larger rearrangement. The second equivalent then binds to similar structures in both FS-EC and EC, as evidenced by a similar value of K D2 .
The interface between the FS and EC domain is revealed in the structure of the FS-EC domain pair (16). The relatively small domain interface (550 Å 2 ) is largely formed by ␤-strand 5 of the FS domain and ␣-helix E of the EC domain, the second helix of EF1 (Fig. 9). A further contact between the FS domain and EF1 involves His 62 and His 224 , which participates in the unusual cis-peptide bond in EF1. There are no direct contacts between the FS domain and EF2. A comparison of the EF-hand FIG. 6. Conformation of mutated EC domains. Far UV circular dichroism spectra of the EC domain and its mutants EC-EF1-E234K and EC-EF2-D257K were recorded in 5 mM Tris/HCl, pH 7.4. Spectra were measured in the presence of 2 mM Ca 2ϩ (solid line) and after addition of 3 mM EDTA (dashed line).

FIG. 7. Ca 2؉ binding to EC domains with mutated EF-hands.
Ca 2ϩ binding was monitored by intrinsic fluorescence as described in the legend to Fig. 2. Macroscopic constants K D1 ϭ 11750 nM and K D2 ϭ 56 nM resulted from the best fit of the model with two binding sites to the experimental data of wild type EC domain (q). Best fit curves for EC-EF1-E234K (OE) and EC-EF2-D257K (‚) were obtained with a model with one binding site. Corresponding dissociation constants were K d ϭ 45 M and K d ϭ 205 M, respectively. The loss of cooperativity in the mutants is evident from the reduced slope of the binding curves.
FIG. 8. Ca 2؉ binding to full-length BM-40 with mutated EFhands. Ca 2ϩ binding was monitored by intrinsic fluorescence as described in the legend to Fig. 2. Dissociation constants K D1 ϭ 490 nM and K D2 ϭ 57 nM resulted from the best fit of the model with two binding sites to the experimental data for full-length BM-40 (Ⅺ). The best fit curve for BM-40-EF2-D257K () assumed one binding site with K d ϭ 220 nM. Dissociation constants K D1 ϭ 21540 nM and K D2 ϭ 52 nM were obtained for FS-EC-⌬C (E). The best fit curve for BM-40-EF2-E268K (ࡗ) assumed one binding site with K d ϭ 1200 nM. The loss of cooperativity in the mutants EF2-D257K and EF2-E268K is evident from the reduced slope of the binding curves. structures in FS-EC domain pair and the isolated EC domain does not reveal any significant differences that could explain the large effect of the FS domain on Ca 2ϩ binding. This is perhaps not surprising, because both structures were determined under saturating concentrations of Ca 2ϩ . The corresponding Ca 2ϩ -free forms have so far defied a structure determination, but we assume that because of the intimate contacts between the FS domain and EF1, the EF-hands in Ca 2ϩ -free FS-EC adopt a different structure than in the Ca 2ϩ -free EC domain. EF1 is presumably predisposed for Ca 2ϩ binding in FS-EC, perhaps because the cis-trans equilibrium of the His 224 -Pro 225 peptide bond is pushed toward the cis-conformation required for Ca 2ϩ binding or because Glu 234 is positioned appropriately. It is thus tempting to associate K D1 in the interpretation of our binding data with EF1 and K D2 with EF2, although we stress again that the macroscopic dissociation constants are not necessarily linked to microscopic sequential events.
To analyze the contributions of individual EF-hands to Ca 2ϩ binding, we replaced the conserved glutamic acid at position ϪZ in each EF-hand by lysine (Fig. 1). Although the mutant EC-EF1-E234K was readily secreted from the human kidney cells used for recombinant expression, the EC domain with the corresponding mutation in the second EF-hand, EC-EF2-E268K, as well as the double mutant EC-EF1,2-E234K,E268K, could not be detected in the cell culture supernatant. However, the corresponding mRNA as well as small amounts of both proteins were present in cellular extracts (Fig. 4). Apparently, mutation of the bidentate Ca 2ϩ ligand Glu 268 in EF2 interfered with proper folding in the endoplasmic reticulum, resulting in protein degradation. These results point to a function of EFhands in the folding and secretion of extracellular EF-hand proteins. It remains to be elucidated whether chaperones are involved in this process. To determine whether Ca 2ϩ binding to EF2 is strictly required for folding and secretion of BM-40, we produced mutant EC-EF2-D257K, in which the aspartic acid at position X was replaced by lysine. This mutant was secreted from the cells. No cooperativity could be detected in the Ca 2ϩ titrations, and a model assuming a single binding site with a dissociation constant of K d ϭ 205 M fitted the data well. In the mutant EC-EF1-E234K, cooperativity was also lost, and a dissociation constant of K d ϭ 45 M was obtained. Most probably, the mutations decreased the Ca 2ϩ affinity of the affected site to such an extent that binding could no longer be detected at the Ca 2ϩ concentrations used. The remaining Ca 2ϩ binding thus represents binding to the nonmutated EF-hand when the other hand in the pair is mutated. The dramatic, several 100-fold reduction in Ca 2ϩ affinity of the nonmutated EF-hand is additional evidence for a strong coupling between the two EFhands, which underlies the cooperativity of Ca 2ϩ binding. Further insight into the energetics of Ca 2ϩ binding can be gained by comparing the effects of mutations in the EC domain with those in full-length BM-40. Although the EF2-D257K mutation strongly decreased the Ca 2ϩ affinity when introduced in the EC domain, the same mutation had only a small effect when introduced in full-length BM-40 (Table I), supporting the interpretation that the FS domain increases Ca 2ϩ affinity in the EC domain mainly via interactions with EF1. Interestingly, the interaction of the FS and EC domains allows the mutant BM-40-EF2-E268K to be folded and secreted, whereas the same mutation in the isolated EC domain abolished secretion.
In conclusion, our results show that the EF-hands in the extracellular protein BM-40 bind Ca 2ϩ cooperatively and with high affinity, just like their counterparts in cytosolic proteins. However, an additional layer of complexity is introduced by the presence of the FS domain, which physically interacts with the variant EF-hand 1 and, therefore, is intimately involved in Ca 2ϩ binding to the EC domain.