Calbindin D28k Exhibits Properties Characteristic of a Ca2+ Sensor*

Calbindin D28k is a member of the calmodulin superfamily of Ca2+-binding proteins and contains six EF-hands. The protein is generally believed to function as a Ca2+ buffer, but the studies presented in this work indicate that it may also act as a Ca2+ sensor. The results show that Mg2+ binds to the same sites as Ca2+with an association constant of ∼1.4·103 m −1 in 0.15 m KCl. The four high affinity sites in calbindin D28k bind Ca2+ in a non-sequential, parallel manner. In the presence of physiological concentrations of Mg2+, the Ca2+ affinity is reduced by a factor of 2, and the cooperativity, which otherwise is modest, increases. Based on the binding constants determined in the presence of physiological salt concentrations, we estimate that at the Ca2+ concentration in a resting cell calbindin D28k is saturated to 40–75% with Mg2+ but to less than 9% with Ca2+. In contrast, the protein is expected to be nearly fully saturated with Ca2+ at the Ca2+ level of an activated cell. A substantial conformational change is observed upon Ca2+ binding, but only minor structural changes take place upon Mg2+ binding. This suggests that calbindin D28k undergoes Ca2+-induced structural changes upon Ca2+activation of a cell. Thus, calbindin D28k displays several properties that would be expected for a protein involved in Ca2+-induced signal transmission and hence may function not only as a Ca2+ buffer but also as a Ca2+sensor. Digestion patterns resulting from limited proteolysis of the protein suggest that the loop of EF-hand 2, a variant site that does not bind Ca2+, becomes exposed upon Ca2+binding.

Calbindin D 28k is a Ca 2ϩ -binding protein expressed in brain as well as in kidney, bone, pancreas, and other tissues (1). In some tissues it is exceptionally abundant. For example, calbindin D 28k constitutes between 0.1 and 1.5% of the total soluble protein in brain, and protein levels in auditory neurons are estimated to reach concentrations of up to 2 mM (2). Calbindin D 28k contains 261 amino acid residues forming six EF-hands, which are organized in a single globular domain ( Fig. 1) (3, 4). The protein is a member of the calmodulin superfamily (5).
Some members of this superfamily are Ca 2ϩ sensors that undergo Ca 2ϩ -induced conformational changes resulting in the exposure of a hydrophobic surface. The hydrophobic patch typically serves as a binding surface for target molecules, which become activated or attenuated upon complex formation (6). The targets include many membrane transport proteins and enzymes (7). In this way, intracellular Ca 2ϩ influx triggers the regulation of cellular processes, such as muscle contraction and the phosphoinositide cascade.
Another class of Ca 2ϩ -binding proteins are the Ca 2ϩ buffers or signal modulators. Parvalbumin and calbindin D 9k are examples of Ca 2ϩ -buffer proteins. Unlike Ca 2ϩ sensors, the Ca 2ϩbuffer proteins do not expose hydrophobic surfaces upon Ca 2ϩ binding. In fact, the exposure of hydrophobic surfaces would be unfavorable for these proteins because it may limit their stability and Ca 2ϩ affinity. Ca 2ϩ buffer proteins are thought to be involved in deactivation of signal transducers and/or quenching of Ca 2ϩ signals, thus protecting the cell against toxic effects of Ca 2ϩ , such as the formation of insoluble calcium phosphates (for a review see Ref. 8). The Ca 2ϩ affinity is generally higher for buffer proteins than for sensor proteins, and the rates of Ca 2ϩ binding and release are lower in the buffer group. Calbindin D 28k is thought to prevent sustained elevations of Ca 2ϩ by acting as an intracellular Ca 2ϩ buffer. There is evidence that calbindin D 28k has a Ca 2ϩ -buffering function in neurons. For example, increased intracellular levels of calbindin D 28k causes blunted intracellular Ca 2ϩ elevations (9), and Ca 2ϩ transients are increased in calbindin D 28k null mutant mice (10). A number of examples where Ca 2ϩ buffering by calbindin D 28k has an effect on the electrophysiological behavior of cells have also been reported (11)(12)(13). The notion that calbindin D 28k acts as a Ca 2ϩ -buffering system in the cytoplasm has led to the hypothesis that calbindin D 28k may protect neurons against large fluctuations in free intracellular Ca 2ϩ and, hence, prevent cell death (14 -17). However, experiments using calbindin D 28k -null mutant mice subjected to cerebral ischemia did not support a cytoprotective effect of the protein (18). Although most reports deal with the Ca 2ϩ -buffering function of calbindin D 28k , several lines of evidence suggest that the protein also acts as a Ca 2ϩ sensor. Spectroscopic investigations (19) and in vitro studies using antibodies (20) have shown that calbindin D 28k undergoes a conformational shift upon Ca 2ϩ binding. Additionally, a number of putative Ca 2ϩ -dependent interactions with target proteins or brain membrane fractions have been reported (20 -23), and erythrocyte membrane Ca 2ϩ -ATPase and 3Ј,5Ј-cyclic nucleotide phosphodiesterase have been shown to be stimulated in a dose-dependent, saturable manner with calbindin D 28k (24). Moreover, the finding that a fraction of calbindin D 28k (9 -55%) (20,(25)(26)(27) is specifically associated with particulate structures in the cell indicates that the protein plays other roles in addition to its function as a mobile Ca 2ϩ buffer.
In our previous study (28), we showed that although a conformational change occurs upon Ca 2ϩ binding, both the Ca 2ϩfree and Ca 2ϩ -loaded forms of calbindin D 28k have exposed hydrophobic surfaces. Thus, the protein behaves neither like a classical Ca 2ϩ sensor nor like a Ca 2ϩ buffer. The aim of the present study was to further characterize the Ca 2ϩ -induced conformational change and to determine whether calbindin D 28k is likely to respond structurally to changes in the intracellular concentration of Ca 2ϩ in the presence of physiological levels of Mg 2ϩ and salt.
Chemicals-CaCl 2 and MgCl 2 of Pro Analysi quality were from Merck; quin 2 was from Fluka Buchs, Switzerland, and 5,5Ј-Br 2 -BAPTA 1 was from Molecular Probes, Eugene, OR. All chemicals were of highest grade commercially available. Buffers were made Ca 2ϩ -free by incubation with a dialysis tube filled with Chelex 100 (Bio-Rad) for 2 weeks.
Preparation of Apoprotein-All apo solutions were made from Ca 2ϩdepleted protein, which was produced as follows. Purified protein was dissolved in 1 ml of doubly distilled water, containing an excess of EGTA (10 -20 eq) at pH 8. It was then applied to a 3.4 ϫ 20-cm Sephadex G-25 superfine gel filtration column. To abolish EGTA binding, the protein was applied to the column after 15 ml of saturated NaCl had been allowed to penetrate the top of the column. The NaCl solution had been depleted from residual Ca 2ϩ by dialysis against Chelex 100 resin. During the gel filtration, the protein was passed through the NaCl zone and eluted, now free from EGTA, with doubly distilled Ca 2ϩ -free water. The final product contained between 0.2 and 0.6 molar eq of Ca 2ϩ , as determined from Ca 2ϩ titrations in the presence of the chelator quin 2 or by high resolution inductively coupled plasma mass spectrometry (analysis performed by SGAB, Luleå, Sweden). 1 H NMR spectroscopy was used to verify that the apo samples were free from EGTA.
Near UV CD Spectroscopy-Near-UV CD spectra (250 -300 nm) were obtained using a Jasco J-720 spectropolarimeter at 25°C (thermostated) and quartz cuvettes with a path length of 10 mm. A spectrum was first recorded for the apoprotein ( where V i and V 0 are the total volumes at point i and before adding the first CaCl 2 aliquot, respectively. A min and A max are the respective absorbances that the Ca 2ϩ -chelator complex, and free chelator would have at no dilution. K D is the dissociation constant of the chelator-Ca 2ϩ complex. The free Ca 2ϩ concentration, Y i , was solved using the Newton-Raphson method from Equation 3, where K 1 through K N are the N macroscopic binding constants and CQ i , CP i , and CCa i are the total chelator, protein, and Ca 2ϩ concentrations, respectively, at point i corrected for the dilutions due to the CaCl 2 additions. The initial chelator concentration, CQ 0 , was determined by withdrawing an aliquot of the solution and recording the absorbance at 239.5 nm in the presence of excess Ca 2ϩ (using ⑀ ϭ 4.2 ϫ 10 4 liter mol Ϫ1 cm Ϫ1 ). The initial Ca 2ϩ concentration, CCa 0 , was determined by high resolution inductively coupled plasma mass spectrometry (at SGAB, Luleå, Sweden). The initial protein concentration, CP 0 , was determined by amino acid analysis after acid hydrolysis on a withdrawn aliquot of the protein/chelator solution that had been lyophilized in a hydrolysis tube (Biomedical Centre, Uppsala, Sweden). The adjustable parameters in the fit were A min , A max , and the N macroscopic binding constants. The parameter F was either fixed at 1.0 or could be used as an adjustable parameter to analyze if the assumed stoichiometry (number of macro- scopic binding constants used) is correct (F will end up close to 1.0) or not (F will deviate from 1.0). For presentation, the data and fitted curves were normalized according to Equation 4, The apparent Ca 2ϩ -binding constants in the presence of Mg 2ϩ were derived using the same method as above but with 5,5Ј-Br 2 -BAPTA instead of quin 2. The chelator 5,5Ј-Br 2 -BAPTA has a high level of selectivity against Mg 2ϩ (log K Ca ϭ 5.65; log K Mg Ͻ1, at 0.15 M KCl). Hence, differences in titration curves obtained in the interval 0 -10 mM Mg 2ϩ are due to Mg 2ϩ effects on the protein. The resulting binding constants are apparent Ca 2ϩ binding constants for the protein in the presence of Mg 2ϩ . These constants result from the Ca 2ϩ affinity, the Mg 2ϩ affinity, and competition or coupling between the two events. Determination of Ca 2ϩ and Mg 2ϩ Binding Constants by Fluorescence Spectroscopy-Mg 2ϩ or Ca 2ϩ binding constants were determined by monitoring the intrinsic (tryptophan) fluorescence during Ca 2ϩ or Mg 2ϩ titration in 0.15 M KCl, 2 mM Tris, pH 7.3. Fluorescence spectra were obtained using a PerkinElmer Life Sciences Luminescence Spectrometer LS 50 B connected to a Julabo F25 water bath at 25.0°C. Quartz cuvettes with a path length of 10 mm were used. The excitation wavelength was 280 nm, and the emission scanned was between 300 and 450 nm. The protein concentration was 25 M. Ca 2ϩ -free protein was titrated with increasing amounts of metal ion (Ca 2ϩ or Mg 2ϩ ) beyond saturation. The intensity at each point was corrected for dilution. Ca 2ϩ and Mg 2ϩ titrations were also performed in the presence of 120 M ANS (excitation at 385 nm, emission scanned between 400 and 600 nm) with 10 M calbindin D 28k in 10 mM DTT, 0.15 M KCl, 2 mM Tris, pH 7.1. In some experiments, MgCl 2 was added to a final concentration of 1 or 5 mM. The excitation wavelength was 385 nm (bandwidth 3 nm), and the emission was scanned between 400 and 600 nm. The experimental data were fitted according to Equation 5, where Y is the free Ca 2ϩ or Mg 2ϩ concentration; K A is the apparent binding constant, and I 0 and I p are the intensities for the free and bound state, respectively. Limited Proteolysis of Calbindin D 28k -Calbindin D 28k was dissolved in 50 mM Tris, 150 mM KCl, containing either 1 mM CaCl 2 , 1 mM EDTA, or 2.5 mM MgCl 2 ϩ 0.5 mM EGTA (EGTA has a very high selectivity for Ca 2ϩ over Mg 2ϩ ) at a protein concentration of 0.5 mg/ml. The pH was adjusted to 7.5. Sequencing grade modified trypsin (Promega) was dissolved in the supplied buffer, yielding a stock solution with a concentration of 0.5 mg/ml. Proteolysis was initiated by mixing 100 l of calbindin D 28k (0.5 mg/ml) with 1 l of the trypsin stock at room temperature. Aliquots of 5 l were withdrawn at various time points, and the digestion was blocked by the addition of 1 l of soybean trypsin inhibitor (1 mg/ml) (Roche Molecular Biochemicals). The digested fragments were then separated by SDS-PAGE (15%). Following electrophoresis, the protein bands were either stained by Coomassie Blue and excised for mass spectrometry analysis (see below) or blotted onto a poly(vinylidene difluoride) membrane (Immobilon, Millipore) and subjected to N-terminal amino acid sequencing (automated Edman degradation using an Applied Biosystems 477 A sequencer with on-line detection of phenylthiohydantoin-derivatives with an Applied Biosystems 120A high pressure liquid chromatography).
Sample Preparation for Mass Spectrometry-Coomassie-stained protein fragments were excised from the gel and washed with water, followed by 40% acetonitrile in 25 mM NH 4 HCO 3 , pH 7.8, until the gel piece was transparent. The gel piece was dried in a SpeedVac vacuum centrifuge. Reduction using 10 mM DTT at 48°C for 30 min was followed by alkylation in 55 mM iodoacetamide for 30 min in darkness at room temperature. The gel piece was washed and dried again before digestion with sequencing-grade trypsin (Promega) in 25 mM NH 4 HCO 3 overnight at 37°C. The digestion was terminated, and the peptides were eluted by adding 10 l of 2% trifluoroacetic acid. Peptides were purified from buffer using C-18 reversed phase tips (Ziptips, Millipore).
Mass Spectrometry-Mass spectrometric studies were performed using a Bruker Scout 384 Reflex III matrix-assisted laser desorption/ ionization time-of-flight (MALDI-TOF) mass spectrometer. The instrument was operated in the positive ion mode with delayed extraction and an acceleration voltage of 25 kV. Peptide samples were analyzed using the reflector detector with 2,5-dihydroxybenzoic acid as matrix, whereas larger fragments and intact protein were measured in the linear mode using ferulic acid as matrix. Improved signal-to-noise ra-tios were obtained by the accumulation of 50 -150 single shot spectra. Autolysis fragments of trypsin were used for internal calibration, whereas an external calibration standard was used for the analysis of intact protein fragments.
NMR Spectroscopy-NMR spectra were acquired on a 600-MHz Varian Inova spectrometer at 310 K. 15 N HSQC experiments utilizing pulsed field gradient and preservation of equivalent path (32, 33) employing water flip-back (34) were performed on calbindin D 28K at different Mg 2ϩ or Ca 2ϩ concentrations. The GARP-1 decoupling sequence (35) was used for 15 N decoupling during acquisition. All spectra were acquired with spectral widths of 2000 and 8000 Hz in the F1 and F2 dimensions, respectively, and sampled using 128 and 1024 complex points in the t1 and t2 dimensions, respectively, with 16 transients acquired for each free induction decay. The data were processed and analyzed using the Felix97 Software (Micron Separations, San Diego). The data were multiplied by exponential (F2) and cosine-bell (F1) functions prior to Fourier transformation, and zero-filled to generate matrices of 2048 ϫ 2048 real points. Proton chemical shifts were referenced relative to the water signal, which resonates at 4.64 ppm from the sodium 2,2-dimethyl-2-silapentane sulfonate (DSS) at 310 K. Nitrogen chemical shifts were referenced indirectly relative to DSS, using the nitrogen to proton frequency ratio (36).
The NMR samples contained 0.73 mM apocalbindin D 28k (with less than 0.05 eq of Ca 2ϩ ), 0.1 mM DSS, 10 mM deuterated DTT and were prepared by dissolving the lyophilized protein and deuterated DTT in 620 l of 93% 1 H 2 O, 7% 2 H 2 O. The pH was adjusted to 6.8 with 0.1 M KOH or HCl. Aliquots from a 73 mM CaCl 2 stock solution was added to one sample directly in the NMR tube using a Hamilton syringe (5 l for each point of the titration). The protein concentration was determined by amino acid analysis following acid hydrolysis. The concentration of the CaCl 2 stock solution was determined by inductively coupled plasma mass spectrometry. HSQC spectra were also acquired for a 0.8 mM protein sample with 2 or 10 mM MgCl 2 as well as with 10 mM MgCl 2 and 5 mM CaCl 2 .

RESULTS
Ca 2ϩ Binding Determined from Competition with Chromophoric Chelators-Titration with Ca 2ϩ in the presence of a chromophoric chelator was used to determine the macroscopic Ca 2ϩ binding constants (30,37). To minimize experimental errors, care was taken to determine accurately the Ca 2ϩ concentrations before and after titration (by high resolution inductively coupled plasma mass spectrometry), as well as the protein concentration before titration (by amino acid analysis after acid hydrolysis). Reliable acid hydrolysis results were obtained only when an aliquot of the sample was freeze-dried in the hydrolysis tube (glass) before shipment to the analysis station. Samples sent as liquid in Eppendorf tubes yielded unreasonably low and inconsistent protein concentrations, probably due to protein adhesion to the plastic.
The stoichiometry of Ca 2ϩ binding was analyzed by fitting each titration in a simplified way using Equations 1-3 with n ϭ 2. CP 0 was set to the value obtained by acid hydrolysis, and the parameter F was used as a variable parameter. After fitting, the stoichiometry was calculated as 2F. By using this method, the number of high affinity Ca 2ϩ -binding sites in the intact protein are 4.03 Ϯ 0.11 (average and S.D. of 9 titrations). Thus the data convincingly show that human calbindin D 28k binds 4 Ca 2ϩ ions with high affinity. To obtain the values of the four macroscopic binding constants, each titration was then fitted using Equations 1-3 with n ϭ 4. CP 0 was set to the value obtained by acid hydrolysis, and F was fixed at 1.0. The intact protein displayed macroscopic binding constants of 1.5⅐10 7 to 3.5⅐10 8 M Ϫ1 at low ionic strength and 8.9⅐10 5 to 7.9⅐10 6 M Ϫ1 in 0.15 M KCl (Table I).
The Ca 2ϩ titration curves display a close to linear decrease of the absorbance as a function of total Ca 2ϩ concentration. The lack of sigmoidal shape may indicate a low degree of cooperativity of Ca 2ϩ binding. More quantitative information on the cooperativity is derived from the relationships between the macroscopic binding constants. For a protein with four Ca 2ϩbinding sites, values of lgK 1 Ϫ lgK 2 Ͻ 0.426, lgK 2 Ϫ lgK 3 Ͻ 0.352, lgK 3 Ϫ lgK 4 Ͻ 0.426, and lgK 1 Ϫ lgK 4 Ͻ 1.204 imply positive cooperativity (38). The results obtained here for the intact protein in the absence of Mg 2ϩ yield the following: lgK 1 Ϫ lgK 2 ϭ 0.393; lgK 2 Ϫ lgK 3 ϭ 0.275; lgK 3 Ϫ lgK 4 ϭ 0.281; and lgK 1 Ϫ lgK 4 ϭ 0.948. Hence there is only a low degree of cooperativity of Ca 2ϩ binding. In the presence of 2 mM Mg 2ϩ , however, the values obtained are as follows: lgK 1 Ϫ lgK 2 ϭ 0.18; lgK 2 Ϫ lgK 3 ϭ 0.55; lgK 3 Ϫ lgK 4 ϭ 0.03; and lgK 1 Ϫ lgK 4 ϭ 0.80. This indicates that the cooperativity of Ca 2ϩ binding is greater in the presence of Mg 2ϩ , which is the relevant physiological condition. In summary, our results clearly show that calbindin D 28k has four high affinity Ca 2ϩ -binding sites, which act as a cooperative unit. Hence the situation is very different from calmodulin, in which the four sites are distributed into two groups with different affinity (30).

Ca 2ϩ Binding Monitored by Trp Fluorescence-Calbindin
D 28k contains two Trp residues located in the first helix of EF-hands 1 and 3, respectively. Analysis of the Ca 2ϩ titration curve monitored by Trp fluorescence for calbindin D 28k yields an apparent binding constant of around 1⅐10 6 M Ϫ1 in 0.15 M KCl, which is similar to that obtained by Ca 2ϩ titrations with chromophoric chelators. Although the titration monitored by Trp fluorescence does not provide the high precision obtained using the competitive chromophoric chelator method, it does confirm that the binding constants obtained by the chelator method are in the correct range.
Ca 2ϩ Binding as Monitored by NMR Spectroscopy-Ca 2ϩ binding to calbindin D 28k was monitored by heteronuclear twodimensional 15 N-1 H HSQC NMR spectroscopy for samples containing 0 -6 eq of Ca 2ϩ . In the absence of Ca 2ϩ , the signals are broad, and there are few well resolved resonances ( Fig. 2A). Addition of Ca 2ϩ to the protein sample causes significant changes in the HSQC spectrum. The HSQC spectrum of the Ca 2ϩ -saturated form shows predominantly sharp resonances and improved spectral resolution (Fig. 2D). In the absence of Ca 2ϩ , no aggregation or dimerization of the protein is observed by size exclusion chromatography when 100 l of a 1 mM protein sample is injected on a Superdex 75 column (1.1 ϫ 30 cm; Amersham Biosciences, data not shown). Therefore, the NMR signals of the apo state are probably broadened due to exchange between at least two conformational states with a fast-to-intermediate rate on the NMR chemical shift time scale. During the titration, the cross-peaks initially remain broad- ened (Fig. 2, B and C) but start to become sharper after the addition of 3 eq of Ca 2ϩ . Between 3 and 4 eq, very large changes in the spectrum are seen, both in terms of line widths and chemical shift dispersion. No further changes are observed beyond 4 eq of Ca 2ϩ .
The NMR signals of the apo form disappear, and new ones appear during the Ca 2ϩ titration. This shows that Ca 2ϩ dissociation from calbindin D 28k occurs in the slow exchange regime of the NMR chemical shift time scale. The Ca 2ϩ binding process was therefore followed by measuring the intensity of individual cross-peaks, well resolved in the HSQC spectrum as a function of Ca 2ϩ concentration. The signals are grouped according to their behavior between 0 and 4 eq of Ca 2ϩ in Fig. 3, A-C. A small number of resonances are affected already by the addition of the 1st eq of Ca 2ϩ ; others start to increase during the addition of the 2nd or the 3rd eq of Ca 2ϩ . However, the majority of the signals do not appear until addition of the 3rd or 4th eq. Ca 2ϩ binding to a four-site system may be a multibranched process involving one apo state, four Ca 1 states, six Ca 2 states, four Ca 3 states, and one Ca 4 state. Using the macroscopic calcium binding constants, determined by the chelator method, we have calculated for each point of the HSQC Ca 2ϩ titration the fraction of the protein in (i) the apo, (ii) any of the Ca 1 states, (iii) any Ca 2 state, (iv) any Ca 3 state, and (v) the Ca 4 state (Fig. 3D) to saturation results in a 10% increase of the fluorescence intensity at 333 nm and gives a small but reproducible blue shift of the intensity maximum from 333.5 to 332.5 nm (not shown). Addition of excess Ca 2ϩ to the Mg 2ϩ -saturated sample results in an additional 15% increase in the fluorescence intensity (not shown). This shows that, although Mg 2ϩ binds to the protein, it does not invoke the same structural response as Ca 2ϩ binding. The Mg 2ϩ titration curve in the absence of Ca 2ϩ yields an apparent Mg 2ϩ binding constant of 1.4 ϫ 10 3 M Ϫ1 in 0.15 M KCl (average of four independent experiments). The stoichiometry of Mg 2ϩ binding cannot be deduced from the experiment, because the Mg 2ϩ concentration at half-saturation is almost 2 orders of magnitude higher than the protein concentration. However, the obtained binding constant implies that under physiological intracellular concentrations of Mg 2ϩ (0.5-2 mM) and resting levels of Ca 2ϩ , the Mg 2ϩbinding sites in calbindin D 28k are 40 -75% occupied by Mg 2ϩ .
Mg 2ϩ Binding as Monitored by NMR Spectroscopy-Mg 2ϩ binding to calbindin D 28k was also monitored by heteronuclear two-dimensional 15 N-1 H HSQC NMR experiments. Fig. 4 shows spectra recorded using 0.8 mM Ca 2ϩ -free protein samples containing 2 and 10 mM Mg 2ϩ , corresponding to 2.5 and 12.5 eq of Mg 2ϩ , respectively. Addition of Mg 2ϩ to the apoprotein causes an increase in the chemical shift dispersion with a number of shifted peaks, although the effects are not as pronounced as during addition of Ca 2ϩ . Even at 10 mM Mg 2ϩ , most of the peaks are still broadened and clustered between 8.5 and 7 ppm in the proton dimension and between 121 and 115 ppm in the nitrogen dimension. Addition of 3.2 mM Ca 2ϩ (4 eq) to the sample with 10 mM Mg 2ϩ yields an HSQC spectrum that is identical to the one in Fig. 2D (data not shown). This shows that Ca 2ϩ efficiently displaces Mg 2ϩ , suggesting that Mg 2ϩ binds to the same sites as Ca 2ϩ .
Mg 2ϩ Binding to Individual EF-hands as Monitored by CD Spectroscopy-The far-UV CD spectra were recorded for six synthetic peptides, comprising EF-hands 1-6, at a concentration of 20 M. The six peptides appear mostly unfolded in the apo form as well as in the presence of 10 or 200 mM MgCl 2 (data not shown). This is in striking contrast to the spectra obtained in the presence of Ca 2ϩ , which show a high degree of helicity and are consistent with Ca 2ϩ binding for EF1, EF3, EF4, EF5, and EF6 (39). The results suggest that although Mg 2ϩ binds to the intact protein, the interaction between Mg 2ϩ and the indi- vidual EF-hands is not strong enough to induce secondary structure in these peptides.
Ca 2ϩ Binding in the Presence of Mg 2ϩ and a Chromophoric Chelator-Ca 2ϩ binding in the presence of Mg 2ϩ was determined using a chromophoric Ca 2ϩ chelator (5,5ЈBr 2 -BAPTA), which has a high level of discrimination against Mg 2ϩ . Fig. 5 shows the experimental data for Ca 2ϩ titrations of intact calbindin D 28k in the presence of 0 or 2 mM Mg 2ϩ , together with the curves of best fit. Already from the appearance of the curves it is evident that the presence of Mg 2ϩ reduces the Ca 2ϩ affinity for calbindin D 28k . The fitting of the experimental data shows that the Ca 2ϩ affinity of calbindin D 28k is reduced approximately 2-fold to yield K ϭ 3.5⅐10 6 -5.6⅐10 5 M Ϫ1 in the presence of 2 mM MgCl 2 (Table I). Only very small ionic strength effects are expected for the addition of 2 mM MgCl 2 into 0.15 M KCl. Therefore, the data again suggest that Mg 2ϩ competes with Ca 2ϩ for the same binding sites.
Changes in the Tertiary Structure as Monitored by Near UV-CD-Ca 2ϩ binding induces changes in the near-UV CD spectra of calbindin D 28k , which indicates a rearrangement of the tertiary structure (not shown, see also Ref. 28), in agreement with the NMR data. In contrast, addition of 10 mM MgCl 2 to apocalbindin D 28k does not alter the near-UV CD spectrum to any significant degree (not shown).
Limited Proteolysis-Limited tryptic digestion of human recombinant calbindin D 28k was performed in the presence of EDTA, Ca 2ϩ , or Mg 2ϩ . The reaction was quenched at different time points ranging from 1 min to 14 h and was analyzed by SDS-gel electrophoresis (Fig. 6). Both the rate and pattern of tryptic digestion are strongly Ca 2ϩ -dependent. The Ca 2ϩloaded form of calbindin D 28k is clearly more resistant to proteolysis than the Ca 2ϩ -free form, although many of the potential trypsin cleavage sites are protected in both forms (Fig. 1). Significant digestion of the apo form is seen already in the 1st min, whereas fragments of the Ca 2ϩ -loaded form appear at a significantly lower rate. The identities of the tryptic fragments were determined by mass spectrometry and N-terminal amino acid sequencing. The deduced cleavage sites are indicated by the arrows in Fig. 7, and the identity of the major fragments are summarized in Table II. Both apo-and Ca 2ϩ -bound calbindin are cleaved at Lys-59 and Lys-235 in the N-terminal helices of EF-hands 2 and 6. The apoprotein is cleaved also at Arg-169 in the C-terminal helix of EF-hand 4. In contrast, the Ca 2ϩ -bound protein is not cleaved at this site but mainly at Lys-72 in the loop of EF-hand 2 and at Lys-98 in the linker between EF-hands 2 and 3. In addition, a minor cleavage site in the Ca 2ϩ form is found at Lys-245 in the loop of EF-hand 6.
At longer incubation times, a number of small fragments appear. These are mainly due to further cleavage of larger fragments and may therefore represent cleavage sites that are inaccessible in the intact protein. It is noteworthy that in EDTA a small fragment containing residues 181-261 (EFhands 5 and 6) is fairly resistant to further cleavage as its intensity increases all the way up to the last time point (120 min, band A8 in Fig. 6).
Limited proteolysis of calbindin D 28k in the presence of 2 mM Mg 2ϩ results in nearly the same digestion pattern as for the apoprotein (Fig. 6). Thus, all identified fragments were equiv-alent in the EDTA and Mg 2ϩ samples, except for a very small amount of a peptide (residues 60 -261), 2 which was not found in the EDTA samples. Mg 2ϩ seems to offer some protection to proteolytic digestion, although digestion is much faster than in the presence of Ca 2ϩ .
Ca 2ϩ Binding to F123 and F456 as Monitored by the Chromophoric Chelator Quin 2-Ca 2ϩ binding to F123 and F456, 2 Fragments A3 and Mg3 also start at residue 60. The molecular weight as estimated from SDS-PAGE (25-26 kDa) of fragments A3 and Mg3 is higher than the maximum possible weight of a polypeptide starting at residue 60. It is possible that A3 and Mg3 contain more than one fragment that together form a complex strong enough to resist separation during the conditions of the gel electrophoresis. each comprising one-half of calbindin D 28k , was measured with the quin 2 method. The stoichiometry of Ca 2ϩ binding to each half was analyzed as described for the intact protein, yielding n ϭ 2.01 Ϯ 0.22 for F123 and 2.14 Ϯ 0.24 for F456. Hence, there are two high affinity sites in each half of calbindin D 28k . One site in each half was found to retain the Ca 2ϩ affinity of the intact protein (K 1 Ϸ1 ϫ 10 8 M Ϫ1 ). The second site in each half binds Ca 2ϩ with ϳ50-fold lower affinity (K 2 Ϸ2 ϫ 10 6 M Ϫ1 ).
Ca 2ϩ -induced Conformational Changes as Monitored by ANS Fluorescence-The fluorescent probe ANS can be used as a sensitive reporter of solvent-exposed hydrophobic patches in proteins. In an attempt to investigate whether F123 and F456 undergo Ca 2ϩ -induced conformational changes at the tertiary level, we measured the Ca 2ϩ dependence of ANS binding for F123 and F456. In EDTA, both F123 (472 nm) and F456 (485 nm) yield a significant blue shift and intensity increase relative to ANS in buffer (530 nm), indicating binding of ANS. When Ca 2ϩ is added, a further blue shift in the fluorescence maximum from 485 to 475 nm is observed for F456 with a 30% increase in intensity (Fig. 8B). In contrast, addition of Ca 2ϩ to F123 yields a 43% decrease in intensity, with a shift from 472 to 483 nm. The results clearly indicate that both halves undergo structural changes after binding of Ca 2ϩ . The response is different in the two halves with the local environment around the bound ANS becoming less polar in F456 but more polar in F123.
ANS fluorescence spectra were also recorded for an equimolar mixture of F123 and F456 and for the intact protein. These spectra were compared with one another and to the sum of the ANS spectra recorded individually for F123 and F456. No spectral differences were observed between these three cases, nei- ther in the absence nor in the presence of Ca 2ϩ . The data suggest that ANS binds to the same exposed hydrophobic patches in the intact protein as in the two halves. Earlier work (4) has shown that each half-fragment has a tendency to form homodimers, hence the contact surface between the two halves in the intact protein would be hidden in the separate fragments and not accessible to ANS binding. DISCUSSION The results of the present work clearly demonstrate that the tertiary structure of calbindin D 28k is Ca 2ϩ -dependent. The large Ca 2ϩ -induced effects observed by two-dimensional ( 1 H-15 N) HSQC NMR spectroscopy (Fig. 2) indicate that the protein undergoes a conformational change. Whereas the spectrum of the apo form shows limited chemical shift dispersion and broad lines, the Ca 2ϩ -saturated protein displays shift dispersion and narrow lines typical of a well folded helical protein. Far-UV CD data (28) indicate a significant degree of helicity both for the apo and Ca 2ϩ -loaded states, suggesting that the limited chemical shift dispersion of the apo form represents a loosely organized tertiary structure, similar to a molten globule. As seen in Figs. 2 and 3, there is not a progressive change in the spectrum throughout the Ca 2ϩ titration. Rather, fairly small spectral changes are observed until 4 eq of Ca 2ϩ are added, at which point there is a dramatic change with a large number of sharp resonances appearing. For example, four signals appear in a region where the glycine residue in position 6 of the Ca 2ϩbound EF-hand loop is typically found (␦ 1 H ϭ 10 -10.5 ppm, ␦ 15 N ϭ 110 -114 ppm). These signals most likely represent the corresponding four glycine residues of the regular EF-hands 1 and 3-5, because synthetic peptides corresponding to these four sites also display 1 H chemical shifts of these glycine residues between 10 and 10.5 ppm in the Ca 2ϩ -bound form (39). Whereas data on the rat protein have been interpreted as one site having much higher affinity than two or three other sites (40), we do not observe sites differing in affinity. Rather, we have strong evidence from the quin 2 and NMR titrations that all four sites bind Ca 2ϩ in a parallel fashion with similar affinity and positive cooperativity.
Limited tryptic proteolysis of the apo-and Ca 2ϩ proteins (Figs. 6 and 7 and Table II) provides further details on the Ca 2ϩ -induced structural changes. In accordance with the HSQC results, the protein appears more loosely folded in the apo state, because it is considerably more sensitive to proteolysis than in the Ca 2ϩ -loaded state. The cleavage patterns are also distinctly different. Whereas the apoprotein is mainly cleaved at Arg-169, the Ca 2ϩ form is cleaved at Lys-72 and Lys-98. Arg-169 is located in the middle of the second helix of EF-hand 4, whereas Lys-72 and Lys-98 are found in the loop of EF-hand 2 and in the N-terminal helix of EF hand 3, respectively. This suggests that the conformational changes upon Ca 2ϩ binding involve burial or rotation of the second helix in EF-hand 4 and exposure of the loop of EF-hand 2 and of the N-terminal helix of EF hand 3. EF-hand 2 is a variant site that does not appear to bind Ca 2ϩ (39), and its Ca 2ϩ -induced exposure to solvent is intriguing. One could speculate that EF-hand 2 is part of a target-binding surface that becomes exposed upon Ca 2ϩ binding.
It is interesting to compare the limited proteolysis of calbindin D 28k to the digestion patterns of calretinin. Calretinin is a hexa-EF-hand protein that is homologous to calbindin D 28k (58% sequence identity), but in contrast to calbindin, it binds Ca 2ϩ to EF-hand 2. In calretinin, regions in EF-hands 2 and 3 are not cleaved in the presence of Ca 2ϩ , which indicates that this part of the protein is not exposed to solvent in a Ca 2ϩ -dependent way (41,42). Instead, cleavage occurs in EF-hand 6 of calretinin, which corresponds to a minor Ca 2ϩ -induced cleavage point in calbindin D 28k . EF-hand 6 represents a non-canonical sequence in both proteins. The dissimilarities in digestion pattern may reflect functional and evolutionary differences between the two proteins to adapt to different targets.
The heteronuclear two-dimensional 15 N-1 H HSQC NMR experiments show that Ca 2ϩ efficiently displaces Mg 2ϩ , suggesting that Mg 2ϩ binds to the same sites as Ca 2ϩ . The results from limited proteolysis, near-UV CD, and NMR spectroscopy show that the structural changes upon Mg 2ϩ binding are much smaller than upon Ca 2ϩ binding. This indicates that although Mg 2ϩ binds to calbindin D 28k , it coordinates in such a way that is does not drive any substantial conformational change. This is consistent with structural data on other EF-hand proteins, which show that the coordination of Mg 2ϩ and Ca 2ϩ to the same site generally differs (43). The bidendate Ca 2ϩ -ligating glutamate at position 12 in the EF-hand loop is important for inducing conformational changes in Ca 2ϩ sensor proteins. In the Mg 2ϩ -loaded form, the ion is often coordinated by none or only by one of the side chain oxygens of this glutamate side chain.
Each of the two isolated halves, comprising EF-hands 1-3 and 4 -6, respectively, was found to contain two high affinity Ca 2ϩ -binding sites. In each of the two halves, one site retains the Ca 2ϩ affinity as seen for intact calbindin D 28k , whereas the other site has lost affinity by a factor of 50. Hence, truncation of the protein leads to loss of important interactions between EF-hands at the interface between the two halves. Thus, the Ca 2ϩ -binding sites of calbindin D 28k are clearly not grouped into two largely independent domains, as in calmodulin. The b The same bands were also in-gel digested with trypsin, and the amino acid sequences of the bands were determined based on the cleavage pattern identified by MALDI-TOF mass spectrometry. two halves of calmodulin retain native-like Ca 2ϩ binding behavior even when separated (44). The present Ca 2ϩ binding data, obtained for F123 and F456, support earlier findings that all six EF-hands in calbindin D 28k are part of one globular domain (3,4).
Calbindin D 28k appears to have evolved through triplication of an ancestral gene coding for two EF-hands (5). EF-hands often associate in pairs as they occur in the sequence, but in calbindin D 28k there are also extensive interactions beyond the pairwise contacts (3,4). As isolated synthetic peptides, EFhands 1 and 3-5 in calbindin D 28k are the main Ca 2ϩ -binding sites, whereas EF-hand 2 fails to bind Ca 2ϩ and EF-hand 6 does so only weakly (39). Hence, it seems reasonable to assume that EF-hands 1 and 3 are active in the Ca 2ϩ titrations of the N-terminal half and EF-hands 4 and 5 of the C-terminal half. An attractive model for the intact protein is that EF hands 3 and 4 are paired with one another. A cut in this pair would then have major influences on the Ca 2ϩ affinity for both EF-hands 3 and 4, which could explain the reduced Ca 2ϩ affinity of one site in each half. In this model, the conformational properties and high Ca 2ϩ affinities of EF-hands 1 and 5 depend mostly on interactions within the respective half.
The physiological role of calbindin D 28k as a neuroprotector in conditions related to Ca 2ϩ overload, and as a facilitator of neuronal Ca 2ϩ signaling and renal Ca 2ϩ resorption, has been suggested to be linked to the Ca 2ϩ -buffering capacity of the protein. However, Ca 2ϩ -induced conformational changes, reported in an earlier study (28) and further analyzed in the present work, suggest that some of the physiological findings could be due to Ca 2ϩ sensor activity of calbindin D 28k . All cells have transport systems for the extrusion of Ca 2ϩ . The intracellular Ca 2ϩ concentration in a resting cell is therefore relatively low, Ͻ0.1 M. The cytosolic Ca 2ϩ concentration can be abruptly raised to 1-10 M when the cell is activated by influx of Ca 2ϩ . In the activated cell, Ca 2ϩ is bound in a highly selective way to Ca 2ϩ -signaling proteins (Ca 2ϩ sensors) and induce conformational changes, which lead to an altered activity of target proteins. As discussed above, calbindin D 28k clearly undergoes a Ca 2ϩ -induced conformational change; however, to be classified as a Ca 2ϩ sensor, several other conditions also need to be fulfilled. The intracellular concentration of Mg 2ϩ , a potential competitor to Ca 2ϩ , is kept at a relatively constant level around 0.5-2 mM. Consequently, a Ca 2ϩ sensor protein must have ϳ1000-fold selectivity for Ca 2ϩ over Mg 2ϩ . This condition is met by calbindin D 28k as the protein binds to Mg 2ϩ with an affinity constant around 1.4⅐10 3 M Ϫ1 in 0.15 M KCl, where the average Ca 2ϩ binding constant is 2.5⅐10 6 M Ϫ1 . A Ca 2ϩ sensor must further be able to respond structurally within a biologically relevant range of intracellular Ca 2ϩ concentrations. This is also true for calbindin as its average Ca 2ϩ dissociation constant (K d ,av ϭ 7⅐10 Ϫ7 M) falls within the physiological range of intracellular free Ca 2ϩ concentrations. In contrast, a Ca 2ϩbuffer protein would be expected to have higher affinity (K d Ͻ10 Ϫ7 M), whereas Ca 2ϩ -binding proteins with a lower affinity (K d Ͼ10 Ϫ5 M) cannot act as sensors because they are unable to detect the changes in intracellular free Ca 2ϩ concentrations that normally occur in cells. Fig. 9 shows the Ca 2ϩ saturation curves for calbindin D 28k , calmodulin, and parvalbumin calculated from the Ca 2ϩ binding constants in 0.15 M KCl and 1-2 mM MgCl 2 . Parvalbumin is considered to be a Ca 2ϩ -buffering protein and binds Ca 2ϩ with ϳ20 times higher affinity than calbindin D 28k . It is clear from Fig. 9 that parvalbumin cannot be a Ca 2ϩ sensor because it is almost fully saturated with Ca 2ϩ at resting concentrations of Ca 2ϩ . By contrast, calbindin D 28k and calmodulin are only Յ9 and Յ3% saturated with Ca 2ϩ , respectively, at resting concentrations of Ca 2ϩ (Յ0.1 M). At Ca 2ϩ concentrations similar to those that follow Ca 2ϩ activation of a cell, the saturation levels of calbindin D 28k and calmodulin are changed dramatically, as would be expected for Ca 2ϩ sensors.
If calbindin D 28k is a Ca 2ϩ sensor, by which mechanism could it work? As mentioned above, the Ca 2ϩ -binding sites of calbindin D 28k are clearly not grouped into two largely independent domains, as seen in calmodulin. Therefore, it does not follow the calmodulin paradigm of having two independent domains, which expose hydrophobic surfaces only upon Ca 2ϩ binding, and that together clamp around the recognition sequences of target enzymes (45). Instead, the tryptic digestion patterns suggest that the Ca 2ϩ -induced conformational changes involve the rotation of helices and the reorganization of the interface between EF-hands and the linker loops. The mechanism may resemble the Ca 2ϩ -induced changes observed for recoverin (46), a protein with 4 EF-hands organized in a single globular domain. Alternatively, calbindin D 28k may function by a unique mechanism, possibly shared with other hexa-EF-hand proteins.