The Metal-binding Properties of DREAM

DREAM, an EF-hand protein, associates with and modulates the activity of presenilins and Kv4 potassium channels in neural and cardiac tissues and represses prodynorphin andc-fos gene expression by binding to DNA response elements in these genes. Information concerning the metal-binding properties of DREAM and the consequences of metal binding on protein structure are important in understanding how this protein functions in cells. We now show that DREAM binds 1 mol of calcium/mol of protein with relatively high affinity and another 3 mol of calcium with lower affinity. DREAM binds 1 mol of magnesium/mol of protein. DREAM, pre-loaded with 1 mol of calcium, binds 1 mol of magnesium, thus demonstrating that the magnesium-binding site is distinct from the high affinity calcium-binding site. Analysis of metal binding to mutant DREAM protein constructs localizes the high affinity calcium-binding site and the magnesium-binding site to EF-hands 3 or 4. Binding of calcium but not magnesium changes the conformation, stability, and α-helical content of DREAM. Calcium, but not magnesium, reduces the affinity of apo-DREAM for specific DNA response elements in the prodynorphin andc-fos genes. We conclude that DREAM binds calcium and magnesium and that calcium, but not magnesium, modulates DREAM structure and function.

DREAM (downstream regulatory element antagonist modulator, also known as calsenilin or KChIP3) is an EF-hand protein, which binds to the carboxyl-terminal portion of presenilin 2 and the amino-terminal portion of Kv4 potassium channels and thereby modulates the activities of these proteins (1)(2)(3)(4)(5)(6)(7)(8). DREAM reverses presenilin-mediated enhancement of calcium signaling in neuronal cells (7). Because familial forms of Alzheimer's disease are often associated with mutations in the presenilin (PS) 1 1 and 2 genes (9 -18), and mutations in PS1 are associated with a relative increase in A␤42 in amyloid plaques seen in this disease, it is conceivable that DREAM modulation of PS function may be altered in this disease (12-15, 17, 20 -24). DREAM also is a member of a family of proteins that interact with the amino terminus of calcium-activated Kv4 potassium channels (8). These channels form fast-inactivating outward currents in neurons and transient outward currents in the heart, which are critical for normal function in these cells. In its metal-free form (henceforth referred to as the apo-form), DREAM acts as a direct transcriptional repressor of the prodynorphin and c-fos genes (1,25). In the presence of calcium, DREAM dissociates from the DNA response element, thus allowing the activity of these genes to increase (1). Caffeine, via calcium-dependent signaling pathways, relieves DREAM-mediated repression of the prodynorphin and c-fos for promoters in cells maintained in culture (1).
DREAM has four EF-hand motifs with unknown metal-binding properties. It is not known if the protein binds calcium in all four EF-hands or whether it binds metals other than calcium. Mandel and Goodman (5) have suggested that the presence of an aspartate residue instead of a glutamate residue at the ϪZ position of the second EF-hand might allow the protein to bind magnesium. Knowledge of the metal-binding properties of DREAM and how such binding affects DREAM structure and function might be important to our understanding of the pathogenesis of some forms of Alzheimer's disease and in understanding of the modulation of potassium channel function in neural and cardiac tissues. We show for the first time that both calcium and magnesium are bound by DREAM at different sites and that the binding of calcium, but not magnesium, to the protein is associated with distinct structural changes in the protein that affect the affinity of the protein for DNA.

EXPERIMENTAL PROCEDURES
General-Protein amino acid composition, protein amino-terminal and carboxyl-terminal sequencing, and DNA sequencing were carried out as described (26 -28). Oligonucleotide synthesis (29) was performed using an Applied Biosystems DNA/oligonucleotide synthesizer (Applied Biosystems, Foster City, CA). UV spectra of proteins and nucleic acids were recorded using a model DU-70 or DU-640 Beckman spectrophotometer (Beckman Instruments, Fullerton, CA). Protein concentrations were determined by amino acid analysis, by the Bradford method (30), or by measuring UV absorbance using the molar absorptivity of DREAM at 281 nm of 29.6 Ϯ 0.3 mM Ϫ1 cm Ϫ1 obtained by quantitative determination of nitrogen by the indophenol blue method (31). Spectra were corrected for turbidity. SDS and non-denaturing PAGE was carried out using a PhastGel apparatus (Amersham Biosciences) and precast gels.
Full-length DREAM and mutant DREAM proteins were biosynthesized in Escherichia coli BL21 cells, transformed with pGEX-6P-1 plasmid (Amersham Biosciences) containing the DREAM complementary DNA sequence (GenBank TM accession number AJ131730) using procedures described previously (32)(33)(34). The primers shown in Scheme I were used to reverse-transcribe and amplify DREAM mRNA from human brain total RNA (CLONTECH, Palo Alto, CA) using the Titan One Tube Reverse Transcriptase-PCR System (Roche Molecular Biochemicals) by PCR methods (35,36). The underlined sequences in the forward and reverse primer (shown in Scheme 1) represent BglII site and XhoI sites, respectively. The cDNA product was purified and ligated into the TA-cloning site of pCR 2.1 (Invitrogen); the chimeric DREAM pCR 2.1 plasmid was amplified in TOP 10 E. coli cells. The full-length DREAM cDNA served as the template for amplification of appropriate cDNAs for mutant forms of DREAM. The forward and reverse oligonucleotide primers (the underlined sequences in the forward and reverse primer represents BglII sites and XhoI sites, respectively) used in each case are shown in Table I.
BglII and XhoI enzymes were used to excise the insert DREAM cDNA from DREAM pCR 2.1. The purified insert was ligated into BamHI-and XhoI-treated plasmid pGEX-6P-1. Chimeric DREAM pGEX-6P-1 plasmid was used to transform E. coli BL21 cells. Transformed cells were grown in 2ϫ YT medium at 37°C to an A 600 of 1.0 absorbance units. Protein expression was induced with 1.0 mM isopropylthiogalactoside at 20°C. After 7 h the cells were collected by centrifugation at 2,000 ϫ g and lysed using 0.1-mm glass beads and a Bead Beater (Biospec Products, Bartlesville, OK). The clarified lysate was applied to a column of glutathione-Sepharose resin (CLONTECH, Palo Alto, CA). The fusion GST-DREAM protein was eluted in glutathione elution buffer containing 1.5 mM EDTA (30) and treated with PreScission protease (Amersham Biosciences) at 4°C. The protease-treated protein was concentrated, dialyzed against Mono-Q start buffer (50 mM Tris, pH 8.0, 1.5 mM EDTA, 10 mM dithiothreitol (DTT)), and purified by chromatography on a Mono-Q HR10/10 column (Amersham Biosciences). A NaCl gradient (0 -1 M in Mono-Q start buffer) was devel-oped over 1 h. Fractions containing DREAM were pooled and dialyzed against buffer (50 mM Tris, 5 mM DTT, 1.5 mM EDTA, pH 7.5). For the ⌬3,4 DREAM construct and the 1-101 DREAM construct on-column PreScission Protease treatment was carried out, and pure proteins obtained in this manner were used for analysis.

Matrix-assisted Laser Desorption/Ionization-Mass Spectrometry (MALDI-MS) and On-line Membrane Preconcentration-Capillary High
Pressure Liquid Chromatography-MS/MS Analysis (mPC-LC-MS/ MS)-DREAM protein sequence was verified by tryptic digestion of the intact protein, followed by mapping of the resulting tryptic peptides by MALDI-MS. MALDI-MS data was obtained on a Voyager DE-STR system (PE Biosystems, Framingham, MA) using the reflector mode over mass range 400 -4,500 Da with 20-kV accelerating voltage.
Sequence information was obtained from the identified peptides using on-line mPC-LC-MS/MS performed with a Micromass Q-Tof II mass spectrometer (Micromass, Danvers, MA) equipped with a modified Micromass nano-ESI interface (37). Membrane PC-LC-MS spectra were recorded over mass range 300 -1900 every 2 s. Mass resolution was 8000 -10,000. The mass axis was calibrated over the range of 172.88 to 1821.72 with NaI. Precursor ions were automatically chosen from MS survey scans over mass range 375-1500. Collision energy values were automatically chosen as a function of precursor ion charge state and m/z value. Argon was used as the collision gas, at collision energies varying from 15 to 50 eV.
Microelectrospray Ionization Mass Spectrometry (ESI-MS)-ESI-MS analyses were carried out on a Finnigan MAT 900 double focusing mass spectrometer (MAT, Bremen, Germany) of E (electrostatic analyzer)-B (magnet) geometry. Analyses of DREAM and the metal ion titration experiments were performed in positive ion mode, using a modified ESI source introduction interface described previously (38). Measurements were carried out in the presence of SF 6 . The mass spectra were scanned from mass to charge (m/z) 1000 -6000, at a scan  Intrinsic Protein Fluorescence and CD Measurements-Fluorescence measurements of apo-DREAM (5-10 M) were performed in 60 mM MOPS, 2 mM EGTA, 0.5 mM DTT, pH 7.0, buffer before and after the addition of CaCl 2 or MgCl 2 . The entire corrected fluorescence emission spectrum from 300 to 450 nm was measured using a SPEX 1680 spectrofluorimeter (SPEX, Edison, NJ). Final values were corrected for dilution and normalized.
CD spectra of DREAM (10 -20 M) in 60 mM MOPS, 2 mM EGTA, 0.5 mM DTT, pH 7.0, were collected on a J-715 spectropolarimeter (JASCO, Japan). Spectral and temperature-dependent measurements were performed at a bandwidth of 2 nm using a U-type quartz cell of path length 0.233 mm (for far-UV measurements, 188 -250 nm) and with a rectangular 1-cm cell (for near-UV measurements, 245-320 nm) in a thermostated cell holder. CD spectra were recorded at 10°C using 5 accumulations, each at scan speed of 20 nm/min and response time of 2 s. The continuous temperature dependence of ellipticity at 222 nm was measured using a scan rate of 50°C/h and a response time of 8 s. Solvent evaporation was prevented by placing a drop of oil on the sample in the cell. CD spectra were smoothed using a Jasco noise reduction routine. CD data are presented in units of molar ellipticity per residue. Secondary structure determinations were carried out as described (39 -46).
Double-stranded DNAs were prepared by annealing complementary DNA strands to the above noted oligonucleotides. Interaction of DREAM protein with DNA was carried out in 50 mM Tris, pH 7.5, 0.5 mM DTT, 0.15 g/ml poly(dI-dC)⅐poly(dI-dC), 50 M EDTA, 10% glycerol with appropriate amounts of Ca 2ϩ acetate or Mg 2ϩ acetate. In some instances, buffer containing 10 mM HEPES, pH 7.8, 10% glycerol, 0.1 mM EDTA, 8 mM MgCl 2 , 1 mM DTT, 0.15 g/ml poly(dI-dC)⅐poly(dI-dC) was used. Supershift analysis was carried out by adding a specific anti-DREAM antibody raised in rabbits against biosynthetic full-length DREAM (Cocalico Biologicals, Reamstown, PA).
Equilibrium-binding Assay with Radiolabeled Calcium and DREAM-The K d values for calcium in DREAM were determined by adding increasing amounts of 45 Ca to 100 pmol of purified DREAM in 1 ml of 0.1 M Tris, pH 7.5, and incubating the reaction mixture for 90 min at 37°C. 100 l of anti-DREAM polyclonal antibody (1:50 dilution) were added to the mixture, and the reaction was allowed to proceed overnight at 4°C. 100 l of a 10% suspension of IgSorb protein A in 0.1 M Tris, pH 7.5 (Enzyme Center, Bedford, MA), was added to each tube, and the mixture was vortexed. After 2 min the tubes were centrifuged at 1,000 ϫ g, and radioactivity in the pellet was determined by scintillation spectrometry. Nonspecific binding of 45 Ca to DREAM was determined by measurement of binding in the presence of 0.1 M CaCl 2 . The K d values for the protein were determined using established methods for fitting equilibrium binding data (47). Nondenaturing

RESULTS
Protein Identification-Amino acid composition and aminoterminal and carboxyl-terminal sequencing afforded data consistent with the predicted sequences of all constructs (Fig. 1, A  and B). Analysis of full-length DREAM by ESI-MS afforded a multiply charged ion series, which on transformation revealed a relative molecular mass of 29,643 Da (expected 29,641 Da, Fig. 2A and Table II). This experimental value is well within the 0.01% accuracy for such measurements. Tryptic peptides  derived from DREAM were analyzed directly by MALDI-MS. A tryptic peptide map was obtained covering ϳ75% of the expected protein sequence. The tryptic peptide mixture was analyzed by mPC-LC-MS/MS, and a total of eight peptides were subjected to collision-induced dissociation, and the resulting product ion spectra were submitted to the Sequest program (48). All eight peptides were identified as being derived from human DREAM protein. One of these peptides (MH ϩ ϭ 985.5) had the sequence GPLGSMQPAK, identifying it as the amino terminus of the protein. The first five amino acids are derived from the PreScission Protease cleavage site and the linker region. Thus, the biosynthetic protein was unambiguously characterized as human DREAM. Subsequently, all mutant DREAM proteins were subjected to ESI-MS. M r values were all consistent with predicted sequences, and this is summarized in Table II.
Metal Ion Binding Studies-We have demonstrated previously that ESI-MS can be used to detect protein-metal ion interactions (33, 34, 49 -51). Furthermore, using this approach, it is possible to determine exact stoichiometries of metal ion binding to the protein, because uptake of specific metal ions does not significantly alter protein ionization efficiencies (49 -51). To investigate the metal-binding properties of DREAM, we analyzed both apo-DREAM and metal ion-titrated DREAM solutions in positive ion ESI-MS. In all cases, we carried out systematic titration studies adding to ϳ25 M DREAM, con-taining 25 M EDTA, metal ions to concentrations of 10, 20, 50, 100, 125, 150, 175, 200, 300, and 500 M. As noted above, mass spectral analysis of apo-DREAM revealed a mass of 29,643 Da ( Fig. 2A). This corresponds to monomeric DREAM; there was no detectable evidence for the presence of dimeric or other multimeric forms of the protein. However, as increasing amounts of calcium are added to the protein solution, a significant change in the ESI spectrum is noted. For example, addition of calcium to 125 M results in the uptake of 4 eq of Ca 2ϩ ions/mol of DREAM (Fig. 2B). Uptake of 1 Ca 2ϩ ion (peak ion abundance 100%) and 2 Ca 2ϩ ions (peak ion abundance ϳ80%) are the predominant ion species. However, there is also clear evidence for uptake of an additional 3 Ca 2ϩ ions (ϳ60% peak ion abundance) and 4 Ca 2ϩ ions (ϳ30% peak ion abundance) (Fig. 2B), indicating that all four EF-hands bind Ca 2ϩ ions. When the amount of calcium added to apo-DREAM reaches or exceeds 175 M, additional binding of small amounts of calcium occur, up to 6 mol eq/mol of protein. This observation was highly reproducible, but increasing the ESI source voltages that affect skimmer collision-induced dissociation results in facile loss of these additional Ca 2ϩ ions, indicating that they were nonspecifically bound metal ion-protein adducts (51). Binding of the first four calciums occurs within the EF-hands of the protein, as others have shown (1) that mutant DREAM, in which the acidic residues within EF-hand binding loops are mutated to alanine, fails to dissociate from a DNA response element in a calcium-dependent manner. It is also interesting to note that when the DREAM protein solution containing 125 M calcium (Fig. 2B) is passed through a cation exchange micro-cartridge, stripping of 2 mol eq of Ca 2ϩ ions is observed (Fig. 2C). This indicates that there are two distinct classes of binding sites present in the DREAM protein. The predominant forms of metal ion-bound protein correspond to the uptake of 1 and 2 mol of Ca 2ϩ ion/mol eq of DREAM. We also conducted equilibrium binding assays with radiolabeled Ca 2ϩ and DREAM and show that there are two classes of binding sites for calcium in DREAM, the first one with higher affinity (K d ϭ 6 Ϯ 1 ϫ 10 Ϫ7 M), and a second class with a much lower affinity.
We next added magnesium to apo-DREAM (Fig. 3A). As can be seen in Fig. 3B, the apoprotein binds 1 mol eq of Mg 2ϩ ion/mol of protein. Increasing the amount of magnesium did not result in additional specific uptake of magnesium by the protein. Interestingly, when the protein containing the single magnesium atom (as seen in Fig. 3B) was passed through a cation exchange micro-cartridge, the metal ion was readily stripped away to give apo-DREAM as shown in Fig. 3C. This indicates that the relative metal ion binding affinity of Mg 2ϩ ion is similar to that of the third and fourth binding sites for Ca 2ϩ ions as shown in Fig. 2C. Calcium metal ions bound to these sites are readily stripped away on passage of calcium-loaded DREAM through a cation exchange micro-cartridge. To investigate further the binding of magnesium and calcium to DREAM, both metals were added to apo-DREAM. Although there was still detectable apo-DREAM, two other distinct ions corresponding to uptake of 1 Ca 2ϩ ion, and 1 Ca 2ϩ ion plus 1 Mg 2ϩ ion were observed (Fig. 3D). It appears that the relatively higher affinity Ca 2ϩ ion-binding site first takes up the Ca 2ϩ ion, followed by uptake of Mg 2ϩ ion at another EF-hand-binding site. We next examined the effects of increasing amounts of magnesium on calcium already bound to the "high affinity" Ca 2ϩ ion-binding site. DREAM titrated with 75 M calcium showed predominantly the high affinity Ca 2ϩ ion-binding site filled with metal ion (ϳ1:1 apo-DREAM, 1 calcium-DREAM ion abundance signals). Subsequently, 75, 150, and 300 M magnesium was titrated into the 1 calcium-DREAM sample. Although the Mg 2ϩ ion-binding EF-hand was also filled, no reduction in the binding of calcium to DREAM was observed (data not shown). We also assessed the binding properties of DREAM in the presence of a series of other metal ions including Tb 3ϩ , Zn 2ϩ , Mn 2ϩ , and Cu 2ϩ . No binding of these metals to DREAM was observed.
To determine which EF-hands might play a role in Ca 2ϩ ion binding and Mg 2ϩ ion binding, we constructed a series of DREAM mutants lacking the amino terminus and various EFhands. Deletion of the NH 2 -terminal region of DREAM (the first 94 amino acids) has little effect on the Ca 2ϩ ion or Mg 2ϩ ion-binding properties of the protein (Fig. 4, A-C). The apo-sFL DREAM has a M r of 19,198 (Fig. 4A) and addition of 150 M calcium shows the adduction of 1-4 Ca 2ϩ ions (Fig. 4B). The addition of 150 M magnesium shows binding of predominantly 1 mol of magnesium ion/mol of protein (Fig. 4C). Additional magnesium (300 M) does not show any more significant uptake of metal ions (data not shown). All these data are consistent with the notion that the amino-terminal portion of the protein plays no significant role in metal binding.
Binding of metals generally occurs in pairs of EF-hands (6,54). To ascertain which of the EF-hand pairs was important for metal binding, we deleted either EF-hands 1 and 2 or EF-hands 3 and 4 from the protein. ⌬3,4 DREAM lacking EF-hands 3 and 4 and retaining EF-hands 1 and 2 has a mass of 9,360 Da as the apoprotein (Fig. 5A). The addition of 150 M calcium results in only a small uptake of metal ion (ϳ30% of apo⌬3,4 DREAM, data not shown). Even with the addition of 300 M calcium, only ϳ30% of the protein contains 1 calcium/mol of protein (Fig.  5B). As seen in Fig. 5C, the addition of 200 M magnesium results in even less metal ion uptake. Addition of 300 M magnesium only results in ϳ15% of the magnesium-bound protein compared with apo⌬3,4 DREAM (data not shown).
The data obtained from analysis of ⌬3,4 DREAM were in distinct contrast to those seen with the ⌬1,2 DREAM. This construct contains EF-hands 3 and 4, and affords a mass of 12,544 Da in the metal-free state (Fig. 5D). Addition of 150 M calcium results in the ready uptake of predominantly 1 mol of Ca 2ϩ ion/mol of protein, with some amount of 2 calcium eq/mol of protein also detected (Fig. 5E). Addition of 300 M calcium results in uptake of additional Ca 2ϩ ions to afford the 2 calcium proteins, with the ratio of apo⌬1,2:1 calcium ⌬1,2:2 calcium ⌬1,2 DREAM of ϳ60:100:90 (data not shown). We also detected some small nonspecific binding of 3-6 Ca 2ϩ ions/mol of protein. The addition of 150 M magnesium shows some significant Mg 2ϩ ion binding, representing ϳ45% of apo⌬1,2 DREAM ion abundance (Fig. 5F). The additional ion responses detected at ϳ12,584 Da (see * in Fig. 5F) correspond to nonspecific adduction of NH 4 ϩ or Mg 2ϩ ions to the 1 magnesium ⌬1,2 DREAM. Upon addition of 300 M magnesium, the uptake of 1 Mg 2ϩ ion by ⌬1,2 DREAM increases to ϳ60% ion abundance of the apo⌬1,2 DREAM (data not shown). All these data clearly demonstrate that the pair of EF-hands 3 and 4 contains the high affinity Ca 2ϩ ion-binding site and the Mg 2ϩ ion-binding site.
The binding of calcium to EF-hand proteins generally occurs when EF-hands are present as pairs (6). To determine whether this occurs with DREAM, we investigated the uptake of both Ca 2ϩ and Mg 2ϩ ions by proteins containing only 3 EF-hands by examining ⌬1 and ⌬4 DREAM. In the case of ⌬4 DREAM, addition of either 150 M calcium or magnesium resulted in insignificant metal ion uptake. For both titration studies, uptake of 1 calcium or 1 magnesium was Ͻ15% of the apo⌬4 DREAM present (Fig. 7, A-C). Even the addition of 300 M calcium or magnesium did not enhance metal ion uptake (data not shown). Similar data were also obtained in the analysis of ⌬1-DREAM (Fig. 7, D-F). The data in Fig. 7, A-F, indicate that there were effects of end EF-hands of EF-hand pairs 1 and 2 and pairs 3 and 4 on global protein structure that preclude efficient metal ion binding.
Metal-induced Conformational Changes in DREAM-Previously, we have shown that ESI-MS can be used to rapidly ascertain whether a protein undergoes gross tertiary or secondary structural change on metal ion uptake (50,51). We demonstrated that the changes in near-UV CD and/or fluorescence spectra correlate with changes in the ESI-MS charge state distribution of calbindin-D 28K on uptake of Ca 2ϩ ions and the charge state distribution of the DNA binding domain of vitamin D receptor on the uptake of 2 mol eq of Zn 2ϩ ions (50,51). As discussed previously (52) the ESI multiply charged spectrum of a protein is directly related to the number of ionizable side chains at or near the protein surface. As a protein undergoes a conformational change, this number of side chains can increase or decrease and is reflected in a charge state shift in ion distribution in the ESI spectrum. To determine whether metal ions induce conformational changes in the tertiary structure of DREAM, we analyzed the multiply charged spectral data of DREAM in the presence of both calcium and magnesium. The ESI-MS analysis of apo-DREAM reveals an ion series from ϩ19 to ϩ7 (m/z ϳ1,500 -4,250), centered around the ϩ12, ϩ11 charge states (Fig. 8A). However, on titration of 175 M calcium into the DREAM protein solution, a subtle but reproducible change occurs in the charge state distribution (Fig. 8B). The charge state ions shift to a bimodal distribution with the most abundant ions now being centered around the ϩ16 and ϩ15 charge states. This is indicative of a conformational change occurring for DREAM on uptake of 4 mol eq of calcium (see Fig.  2B for the transformed spectrum of this sample). However, when calcium-loaded DREAM is then subjected to cation exchange resin chromatography to afford DREAM containing predominantly only one calcium, the charge state distribution of DREAM reverts to a pattern identical to apo-DREAM (compare Fig. 8C with Fig. 8A). Also, when the charge state distribution of DREAM containing only 1 mol eq of calcium is analyzed, it closely resembles that seen with apo-DREAM. This indicates that uptake of only 1 mol eq of calcium does not significantly alter the gross tertiary structure of DREAM. It is only when the protein is fully loaded with 4 mol eq of calcium that a change in conformation occurs. Finally, titration of magnesium with DREAM resulting in the uptake of 1 mol eq of the metal ion results in a much smaller (relative to changes induced by calcium) but reproducible change in the charge state distribution pattern of multiply charged ions (Fig. 8D).
The data on the effects of metal ions on DREAM conformation was further assessed by examining the mobility of DREAM on non-denaturing polyacrylamide gels (Fig. 9). As seen in Fig.  9, lane 1, the calcium-free form of DREAM migrates as a single band. Addition of increasing amounts of calcium causes a progressive decrease in protein mobility (Fig. 9, lanes 2 and 3). In contrast, the addition of magnesium has little effect on protein mobility (lanes 4 and 5). The addition of magnesium to calciumbound protein has no effect on the decrease in protein mobility brought about by calcium (Fig. 9, lane 6).
Optical Spectroscopy Data-Measurements of intrinsic tryptophan fluorescence of DREAM were conducted in the presence of different amounts of calcium and/or magnesium. The average titration curves with corresponding error bars are shown in Fig. 10. Addition of calcium to apoprotein leads to a decrease in fluorescence intensity in the region of pCa 7-4. The absence of any shoulders on the broad sigmoidal curve does not allow us to distinguish between separate transitions corresponding to particular Ca 2ϩ ion-binding sites. Magnesium titration results in a slight decrease of fluorescence intensity. To determine whether either calcium or magnesium influences the binding of the other metal to apo-DREAM, we carried out sequential titration experiments. DREAM was titrated with calcium to a pCa of 6.6; titration was then continued with magnesium. The resultant titration curve coincided within the experimental error with one corresponding to that seen with magnesium alone (Fig. 10). When we performed the calcium titration of DREAM-saturated with magnesium (pMg 3.0), the resulting curve was identical with that seen with calcium alone.
We used far-UV CD spectroscopy to discern differences be- tween the secondary structures of the apo-and calcium-or magnesium-bound protein. Two CD spectra representing the apo-and calcium-saturated state of DREAM are shown in Fig.  11. Table III presents the results of the protein secondary structure calculations obtained by two different methods for fitting the CD data (39 -46). Apo-DREAM is ϳ35% ␣-helical (Table III). However, calcium binding increases ␣-helical content (by ϳ6%) at the expense of a decrease in ␤-turn content (by ϳ5%). The number of terminal residues (␣-distorted) in ␣-helices is not changed upon Ca 2ϩ ion binding, and thus the number of ␣-helices (11 segments) remains the same. The increase in the number of residues in the central part of the helical segments (␣-regular) extends the average helix length by 1 residue (Table III). Addition of magnesium to the apoprotein does not change the protein secondary structure (data not shown). The effect of calcium on fully magnesium-charged DREAM secondary structure is identical to the one seen with calcium addition to the apoprotein. CD spectra in the near-UV region (spectra not shown) exhibit rather weak induced optical activity of DREAM aromatic residues both in the apo-state and in the presence of calcium and/or magnesium. DREAM thermostability was determined by measuring the temperature dependence of CD signals at 222 nm at different free Ca 2ϩ concentrations (Fig. 12). These data demonstrate an increase in protein stability upon metal ion binding. The mid-point of heat denaturation transition shifts from ϳ52°C in the apo-state to ϳ55°C at pCa 7.0, ϳ65°C at pCa 5.0, and to Ն85°C in the calcium-saturated state (pCa 2.5). The heat denaturation of DREAM is irreversible. CD spectra at high temperatures or after cooling do not show any increase in the content of ␤-structure, suggesting that aggregation is not responsible for the observed irreversibility. Increasing the concentration of DTT does not alter the stability of the protein or affect the irreversibility of heat denaturation. Unlike calcium, magnesium titration minimally stabilizes the protein; at pMg 5.0, the mid-point of heat denaturation transition shifts to ϳ55°C. In the presence of excess Mg 2ϩ the shape of the heat ødenaturation profile changes significantly with an increase of the mid-point of transition to 80 -85°C. Similar changes for calcium and magnesium at saturating concentrations suggest that the observed effects are nonspecific.
Effect of Metals on DREAM-DNA Interaction-We next investigated the binding properties of DREAM to DNA response elements found in the prodynorphin and c-fos gene promoters. Apo-DREAM binds to the prodynorphin DREAM DNA response elements (Fig. 13, 1st panel, arrows 1 and 2). The addition of 1 and 10 mM calcium diminishes the binding of DREAM to prodynorphin DNA response elements (Fig. 13, 3rd and 5th panels, arrows 1 and 2). In contrast, the addition of 10 mM magnesium has no effect on DREAM/DNA binding (Fig. 13, 4th panel, arrows 1 and 2). Similar results were obtained with the c-fos DNA response element (data not shown). Of note, binding of the 1-101 NH 2 -terminal fragment of DREAM to the prodynorphin DNA response element was observed.

DISCUSSION
The stoichiometry of calcium binding and the metal-binding properties of DREAM have not been determined previously. Various authors (18 -20, 24) have speculated that the DREAM binds between 2 and 3 mol of Ca 2ϩ /mol of protein. A recent report (53), using an ultrafiltration technique, shows that DREAM binds 3 mol of calcium/mole protein with a dissociation constant of 14 M. By using ESI MS methods, we now show that the protein binds 4 mol of calcium/mol of protein. The four Ca 2ϩ ion-binding sites identified by ESI-MS have different relative affinities for calcium. One calcium-binding site has a higher affinity for calcium than the second site (as by retention of calcium binding following ion exchange chromatography). Two other sites have affinities that are about equivalent but considerably less than those of the first two. By using radiolabeled 45 Ca-binding methods with separation of metalbound protein with a specific antibody, we have demonstrate that the affinity (K d ) of the high affinity site is ϳ0.6 M. Other sites have lower affinities for 45 Ca.
We have shown that the high affinity calcium-binding site is located in EF-hand 3 or 4. This is consistent with the predicted poor affinity of EF-hand 1 and possibly EF-hand 2 for calcium (24). Furthermore, the magnesium-binding site is also located in these EF-hands. Although Mandel and Goodman (5) had hypothesized that EF-hand 2 might be the Mg 2ϩ ion-binding site, our analysis shows that this is clearly not the case. The binding of calcium to DREAM has significant effects on the structure of the protein. The addition of the 1st mol of calcium to DREAM has little or no effect on the ESI-MS charge distribution. However, the addition of 3-4 mol of calcium significantly alters the ESI-MS charge distribution that indicates a significant change in either the secondary or the tertiary structure of the protein. These data are confirmed by examining the changes in intrinsic protein fluorescence observed following the addition of calcium to the apoprotein. The initial addition of calcium to apo-DREAM is not associated with a change in intrinsic protein fluorescence; however, the addition of calcium beyond pCa 7.0 is associated with a decrease in intrinsic protein fluorescence. This is supported by equilibrium-binding data that show a class of binding sites with a K d in the 10 Ϫ6 to 10 Ϫ7 M range. Secondary structure determinations show that the protein has significant ␣-helical content. The addition of calcium results in an increase in the average length of ␣-helices by approximately 1 amino acid residue. Changes in protein mobility on non-denaturing gels seen on adding calcium to the apoprotein further support the notion that DREAM structure changes on calcium addition. It is possible that the Ca 2ϩ ionbinding sites with a K d in the range of 10 Ϫ6 to 10 Ϫ7 M are filled with calcium at all times within the cell. This is analogous to our findings with calbindin-D 28K (48), in which high affinity Ca 2ϩ ion-binding sites with a K d ϳ10 Ϫ9 to 10 Ϫ10 M are present and hence filled with calcium in the intracellular milieu. Alternatively calcium, on rising above 10 -100 nM resting calcium levels, may bind to DREAM-binding sites signaling subsequent structural changes and biochemical events (24).
The binding of calcium to DREAM not only affects the structure of the protein but also significantly affects its ability to bind to specific DNA response elements (1). In the absence of calcium, DREAM readily binds to the prodynorphin and c-fos DREAM DNA response elements. The addition of excess calcium is associated with a dissociation of the DREAM protein-DNA complex. This is in accord with the observations of others (19,20) demonstrating that calcium has an effect on the binding of DREAM protein to specific DNA response elements in the promoters of DREAM-responsive genes. Of interest, the amino-terminal portion of the protein binds DNA. Signaling in the nucleus via Ca 2ϩ ions has been the subject of several recent reviews (25,54,55). Although there is considerable debate as to how nuclear Ca 2ϩ ion concentrations are regulated, Ca 2ϩ ions directly or indirectly alter the transcription of several genes (25,54,55).
DREAM binds magnesium in addition to binding calcium. Our data show that DREAM binds only 1 mol of magnesium/ mol of protein. Addition of large amounts of magnesium does not further increase the amount of magnesium bound to the protein. The Mg 2ϩ ion-binding site appears to be distinct from the first Ca 2ϩ ion-binding site because when DREAM is loaded with 1 mol of calcium/mol of protein, it is still capable of binding 1 mol of magnesium. The affinity of DREAM for Mg 2ϩ ions is lower than that for Ca 2ϩ ions in the first Ca 2ϩ ionbinding site.
We observed only slight changes in the ESI-MS charge distribution of the protein upon the addition of magnesium. This is supported by the absence of changes in intrinsic protein fluorescence observed upon the addition of magnesium. Additionally, there is no change in the mobility of the protein on non-denaturing gels. Hence, it is likely that magnesium associates with the protein but does not significantly alter its struc- ture. Not only does magnesium fail to alter the structure of the protein upon binding to it, but it also fails to alter the ability of DREAM to bind to specific response elements, electrophoretic mobility of the DREAM protein-DNA response element complex is not significantly altered upon the addition of magnesium in concentrations as high as 10 mM.
We also demonstrate that DREAM does not bind other biologically relevant metal ions such as zinc, copper, and manganese. Interestingly, DREAM does not bind terbium, a fluorescent lanthanide that is often used to study the properties of calcium-binding proteins (49,56). Other neuronal calciumbinding proteins, such as the neuronal calcium sensor-1, also fail to tightly bind Tb 3ϩ ions (57). The exact reason why certain EF-hand proteins fail to bind Tb 3ϩ ions is not known but indicates subtle differences in the conformation of EF-hand metal-binding sites.
In conclusion, DREAM is a biologically important EF-hand protein that binds both calcium and magnesium. All four of the EF-hands within the protein appear to be capable of binding calcium. EF-hand 3 or 4 binds calcium with much higher affinity than the others. EF-hand 3 or 4 is involved in magnesium binding. The affinity of the magnesium-binding EF-hand for magnesium is considerably lower than the affinity of the high affinity Ca 2ϩ ion-binding site for calcium. The protein undergoes significant structural change upon binding calcium but not upon the binding of magnesium. Calcium binding alters the ability of DREAM to bind to specific DNA response elements in the prodynorphin and c-fos genes.

TABLE III
Secondary structure of DREAM in the apo (pCa9.5) and the Ca-saturated state (pCa2.5) ␣-Regular or ␤-regular indicates fraction of residues (%) in central part of helical segments or strands; ␣-distorted or ␤-distorted indicates fraction of terminal residues (%) in ␣-helices (2 at each end of helix; totally 4/helical segment) of ␤-strands (1 residue at each end of strand; totally 2/strand), total content of ␣or ␤-structure is equal to the sum of regular and distorted residues; average length of ␣-helix or ␤-strand, average number of residues per ␣-helix or ␤-strand. Secondary structure was calculated from CD spectra using the CDPro software package (39) and modified versions of four methods as follows: SELCON3 (40), CONTIN (41), CONTINLL-CONTIN method in locally linearized approximation (42) and CDSSTR (43) using two sets of reference proteins: all 48 proteins (44), or 23 ␣ϩ␤ reference proteins which were selected by the CLASTER program. Tertiary structure class was determined by CLASTER program (45) of the CDPro package according to methods reported (46). Averages from all methods and reference set values of secondary structure Ϯ root mean square deviations are presented.  Arrows refer to bands 1 and 2 that are retarded upon addition of DREAM protein.