Calcium-Calmodulin-induced Dimerization of the Carboxyl-terminal Domain from Petunia Glutamate Decarboxylase

The acidic, bilobed protein calmodulin (CaM; molecular mass of 16.7 kDa) can activate some 40 distinct proteins in a calcium-dependent manner. The majority of the CaM-binding domain regions of the target proteins are basic and hydrophobic in nature, are devoid of multiple negatively charged residues, and have a propensity to form an α-helix. The CaM-binding domain in the C-terminal region of petunia glutamate decarboxylase (PGD) is atypical because it contains five negatively charged residues. Therefore, we chose to study the binding of calcium-CaM to a 26-residue synthetic peptide encompassing the C-terminal region of PGD. Gel band shift assays, fluorescence spectroscopy, and NMR titration studies showed that a single unique complex of calcium-CaM with two PGD peptides is formed. The formation of a 1:2 protein-peptide complex is unusual; normally, calcium-CaM forms 1:1 complexes with the majority of its target proteins. Circular dichroism spectroscopy showed that the bound PGD peptides have an α-helical structure. NMR studies of biosynthetically [methyl-13C]methionine-labeled CaM revealed that all the Met side chains in CaM are involved in the binding of the PGD peptides. Analysis of fluorescence spectra showed that the single Trp residue of the two peptides becomes bound to the N- and C-terminal lobes of CaM. These results predict that binding of calcium-CaM to PGD will give rise to dimerization of the protein, which may be necessary for activation. Possible models for the structure of the protein-peptide complex, such as a dimeric peptide structure, are discussed.

recent reviews, see Refs. [1][2][3][4]. These proteins include several protein kinases, phosphatase 2B (calcineurin), multiple components in smooth muscle contraction, ion channels, etc. In the x-ray structure, Ca 2ϩ -CaM adopts a dumbbell-shaped structure, with the two lobes connected by a 26-residue-long ␣-helix (5). Each lobe contains two helix-loop-helix Ca 2ϩ -binding motifs that are interconnected by a small ␤-sheet between the two Ca 2ϩ -binding loops. In solution, the middle portion of the central ␣-helical linker region is flexible, as demonstrated by NMR spectroscopy (6). Methionine-rich hydrophobic patches become exposed on each lobe of CaM following the binding of four Ca 2ϩ ions (5), whereas these two hydrophobic patches are absent or not fully exposed in apo-CaM structures (7,8). When the concentration of the secondary messenger Ca 2ϩ increases inside cells from ϳ10 Ϫ7 to ϳ10 Ϫ5 M, four Ca 2ϩ ions will bind to CaM, and CaM will expose its two hydrophobic patches for target protein binding. This sequence of events is confirmed by high resolution x-ray and NMR structures of the complexes between Ca 2ϩ -CaM and peptides derived from CaM-binding domains of target proteins (9 -11).
CaM-binding domains in typical target proteins do not have any amino acid sequence homology. They often comprise a stretch of ϳ20 -25 amino acid residues; this region has the potential to form a positively charged amphipathic ␣-helix (2,3,12). CaM-binding domains usually do not contain a significant number of negatively charged residues such as Asp or Glu, which could give rise to unfavorable electrostatic repulsion with CaM. CaM is highly acidic, as it carries a net negative charge of 17 even after it has bound four Ca 2ϩ ions. Synthetic peptides encompassing CaM-binding domains of distinct target proteins usually bind to Ca 2ϩ -CaM in a 1:1 complex with an amphipathic ␣-helical motif (9 -11). The hydrophobic face of the peptide interacts with the two Met-rich hydrophobic patches in Ca 2ϩ -CaM, and the positively charged Lys and Arg residues of the peptide bind to several negatively charged residues (Asp and Glu) in Ca 2ϩ -CaM (9 -11). The flexible central linker region of CaM can bend to a different extent to accommodate the binding of diverse peptides, and the two Met-rich hydrophobic patches also create a highly adjustable binding surface to allow binding of peptides of distinct amino acid sequence (9 -11). The interactions between Ca 2ϩ -CaM and the peptides are predominantly hydrophobic in nature, and the interface exclusively involves amino acid side chain-side chain interactions (9 -11).
Here we characterize an unusual CaM-binding domain derived from petunia glutamate decarboxylase. Glutamate decarboxylase catalyzes the conversion of glutamate to CO 2 and ␥-aminobutyric acid. ␥-Aminobutyric acid is an important inhibitory neurotransmitter in mammalian cells, and the ␥-aminobutyric acid level is crucial for the development of plant cells (13). Petunia GAD is a 58-kDa cytosolic protein that has a 34-amino acid extension at its C terminus that is absent in mammalian or Escherichia coli GAD (14). This part of the protein has been shown to interact with petunia CaM as well as with mammalian [ 35 S]Met-labeled CaM, CaM-Sepharose, or biotinylated CaM (14). The CaM-binding domain was further mapped to a 26-residue peptide (15). Thus, in contrast to mammalian GADs, plant GADs are CaM-binding proteins, and the CaM-binding properties of plant GAD have now been demonstrated in different plant species such as petunia, Arabidopsis thaliana, fava bean, and soybean (14 -17). The amino acid sequences of the CaM-binding domains of petunia and A. thaliana have been reported (13,14). The most interesting feature of these CaM-binding domains is that they contain four (for A. thaliana) or five (for petunia) Asp or Glu residues at conserved positions ( Fig. 1) (15). The presence of this many negatively charged residues in a CaM-binding domain peptide is unique. CaM is an acidic protein, and hence, it is not designed to interact with negatively charged residues in the peptide. We have studied the interaction of Ca 2ϩ -CaM and a 26-residue PGD peptide derived from GAD by nondenaturing urea-PAGE, circular dichroism, steady-state Trp fluorescence, and two-dimensional NMR. The peptide is a monomer in aqueous solution in the absence of CaM. However, we have found that two ␣-helical peptides bind simultaneously to Ca 2ϩ -CaM. To the best of our knowledge, the Ca 2ϩ -CaM-induced dimerization of plant GAD represents a new Ca 2ϩ -CaM⅐protein-binding motif.

EXPERIMENTAL PROCEDURES
Materials-A synthetic gene encoding the mammalian CaM sequence was cloned and expressed in E. coli as described previously (18,19). The purification of wild-type CaM and [methyl- 13 C]]Met-labeled CaM followed established procedures (19,20). The purification of CT-CaM (a CaM mutant with all four Met residues in the C-terminal lobe of CaM mutated to Leu residues), SeMet-CaM (a CaM variant in which all nine Met residues are substituted with the unnatural amino acid SeMet), and SeMet-CT-CaM (CT-CaM with the remaining five Met residues substituted with SeMet) has been described in detail elsewhere (21); SeMet was incorporated to 87%. A 26-residue peptide (NH 2 -HKKTDSEVQLEMITAWKKFVEEKKKK-CONH 2 ) that corresponds to the CaM-binding domain in petunia GAD (amino acid residues 470 -495) (15) was chemically synthesized by the Peptide Synthesis Core Facility at the University of Calgary (Calgary, Canada). The composition and purity of the peptide were confirmed by analytical high pres-sure liquid chromatography, amino acid analysis, and mass spectrometry. Ultrapure CsCl was purchased from Life Technologies, Inc. The soluble nitroxide spin label TEMPOL (4-hydroxyl-2,2,6,6-tetramethylpiperidinyl-1-oxy; also called HyTEMPO) was purchased from Sigma. The concentration of CaM was determined by ultraviolet spectroscopy using an extinction coefficient ⑀ 276 1% ϭ 1.8. The concentration of the PGD peptide was determined by weight and by using a molar extinction coefficient ⑀ 280 ϭ 5500 M Ϫ1 cm Ϫ1 , which was consistent with the outcome of quantitative amino acid analysis.
Nondenaturing Urea-PAGE-Nondenaturing urea-PAGE gel band shift assays were performed using a published procedure (22). The urea is normally included in the gel to prevent the formation of nonspecific interactions.
CD Spectroscopy-CD spectra were acquired on a Jasco J-715 spectropolarimeter. All experiments were performed at room temperature (22°C) using a 1-mm path length cylindrical quartz cuvette. The parameters used were as follows: 0.2-nm step resolution, 50-nm/min scanning speed, 2-s response time, 1-nm bandwidth, and 20-millidegree sensitivity. All spectra shown were the average over 10 scans. The concentration of the CaM or PGD peptide used was 10 M in 10 mM Tris-HCl, pH 7.2, with 2 mM CaCl 2 or 2 mM EDTA. The total volume of each sample was 200 l. The background signals from the buffer were subtracted, and each spectrum was smoothed and converted to molar ellipticity using Jasco software. The CD spectra were reported either as molar ellipticity or as mean residue molar ellipticity. The ␣-helical content of the peptide was calculated according to Scholtz et al. (23).
Fluorescence Spectroscopy-Steady-state Trp fluorescence and CsCl quenching experiments were carried out on a Hitachi 2000 spectrofluorometer as described previously (21). The Trp fluorescence was excited at 295 nm to reduce the excitation of the two Tyr residues in CaM. Emission wavelength scans were recorded from 300 to 450 nm. Fluorescence samples contained 10 M PGD peptide in 10 mM Tris-HCl, pH 7.2, and 100 mM KCl in the presence of 1 mM CaCl 2 or 5 mM EDTA. The total sample volume was 1 ml. The CaM concentration was 5 or 10 M for a PGD peptide/CaM ratio of 2:1 or 1:1, respectively. The fluorescence titration of the PGD peptide with Ca 2ϩ -CaM was performed by continuously adding 2 l of a 200 M CaM stock solution into 1 ml of the PGD peptide in 10 mM Tris-HCl, pH 7.2, 100 mM KCl, and 1 mM CaCl 2 . The total volume of CaM added to the cuvette was 20 l. Since we monitored only the shift of the maximum emission wavelength in these experiments, the increase in the total sample volume could be neglected.
NMR Spectroscopy-All NMR experiments were carried out on a Bruker AMX-500 spectrometer using a 5-mm broadband, z axis gradient-shielded probe at 298 K. Two-dimensional 1 H-13 C heteronuclear multiple quantum coherence (HMQC) spectra were acquired with pulsed field gradient selection (24). Quadrature detection in the F 1 dimension was obtained using the time-proportional phase incrementation technique. The sweep width was 8 ppm in the 1 H dimension and FIG. 1. Helical wheel presentation of the PGD peptide. The hydrophobic residues are shaded, and the charged residues (Lys, Glu, and Asp) are labeled with positive or negative signs. Please note the hydrophobic face of the peptide centered by Trp-16. Also shown are the amino acid sequences of the C termini of petunia and Arabidopsis (15). The conserved amino acid residues are boxed, and the negatively charged Asp and Glu residues are in boldface.
6 ppm in the 13 C dimension, with the 1 H carrier set at 500.1388 MHz and the 13 C carrier at 125.7613 MHz. The size of the HMQC spectra was a 1024 ϫ 128 real data matrix with eight scans for each experiment. The TEMPOL titration experiments were carried out as described. 2 Two-dimensional homonuclear 1 H-1 H NMR spectra were acquired for the spectral assignment of the PGD peptide in H 2 O. The peptide sample contained 2 mM PGD peptide in 90% H 2 O and 10% D 2 O, pH 5.0 (not corrected for the isotope effects). COSY and NOESY spectra were acquired with pulsed field gradient experiments (25,26). Total correlation spectra were acquired according to Griesinger et al. (27). The typical size of a spectrum was a 2048 ϫ 400 real data matrix with 64ϳ128 scans for each experiment. NOESY spectra were acquired with two different mixing times (100 and 250 ms) to check for spin diffusion effects. All spectra were acquired at two different temperatures (289 and 298 K) for cross-checking the assignments. The assignment obtained at 298 K is reported under "Results." NMR spectra were processed using nmrPipe and nmrDraw software (28). All spectra were zero-filled once in both dimensions, and 90°and 75°sine square shifted window functions were applied to the F 2 and F 1 dimensions, respectively, before Fourier transformation. Proton chemical shifts are referenced to 2,2-dimethyl-2-silapentane-5-sulfonate as 0 ppm. Carbon-13 chemical shifts are referenced indirectly to 2,2-dimethyl-2-silapentane-5-sulfonate using the converting ratio 13 C/ 1 H ϭ 0.251449530 as suggested by Wishart et al. (29). The ␣-1 H and ␣-13 C random coil chemical shifts for the amino acid residues are taken from Wishart and Sykes (30).
Prediction of the ␣-Helical Content of Peptides-The ␣-helical content of the peptide was calculated using the Agadir program, which is available on the web site of EMBL (31). This program has been developed for predicting peptide rather than protein secondary structure, and it has had considerable success when compared with experimental determinations of secondary structure (31).

RESULTS
Nondenaturing Urea-PAGE-First, we used a gel band shift assay to assess the interaction between Ca 2ϩ -CaM and the PGD peptide. At increasing ratios of the PGD peptide to Ca 2ϩ -CaM, we observed a large band shift due to the formation of a complex between the PGD peptide and Ca 2ϩ -CaM (Fig. 2). This phenomenon is normally observed for high affinity CaM-binding peptides (1,3). We found that the band for CaM did not disappear until 2 eq of the PGD peptide were added (Fig. 2). No changes occurred when Ͼ2 eq of PGD peptide were added. These results suggest that two PGD peptide molecules bind simultaneously to one CaM molecule. This observation is surprising in view of the fact that most of the well characterized CaM-binding peptides, such as peptides derived from myosin light chain kinase (MLCK), constitutive nitric-oxide synthase, and CaM-dependent protein kinase I, form only a 1:1 complex with Ca 2ϩ -CaM ( Fig. 2 and data not shown) (1,3). Also the band for the Ca 2ϩ -CaM⅐PGD peptide complex runs much slower on the gel than the bands of either the Ca 2ϩ -CaM⅐MLCK peptide complex (Fig. 2) or the Ca 2ϩ -CaM⅐constitutive nitricoxide synthase peptide complex (data not shown). Although there is no direct correlation between the band migration distance and the molecular mass of proteins on the nondenaturing urea-polyacrylamide gel, the mobility of the band on such a gel usually suggests either that the Ca 2ϩ -CaM⅐PGD peptide complex has a higher molecular mass or that this complex has a very different shape compared with the Ca 2ϩ -CaM⅐MLCK peptide complex (Fig. 2). We also performed the band shift assay without including urea in the gel and samples and obtained exactly the same results (data not shown).
CD Spectroscopy-We acquired CD spectra to study the secondary structural changes that occur upon binding of the PGD peptide to Ca 2ϩ -CaM. The PGD peptide itself has ϳ31% ␣-helix in aqueous solution, and this percentage agrees very well with the 38% predicted by the Agadir program (Fig. 3). The observation of such an extent of ␣-helix formation is not uncommon in 20ϳ25-residue-long monomeric peptides that contain potential ion pairs in (i, i ϩ 3) or (i, i ϩ 4) positions (32,33). When the PGD peptide was added to a Ca 2ϩ -CaM solution, the negative molar ellipticity at 208 and 222 nm increased dramatically upon adding 1 eq of the PGD peptide. This suggests an increase in the ␣-helical content of the Ca 2ϩ -CaM⅐PGD peptide complex (Fig. 3). Since Ca 2ϩ -CaM usually does not gain any ␣-helical structure upon peptide binding (9 -11), the increase in ␣-helix in the Ca 2ϩ -CaM⅐peptide complex can be attributed to the bound peptide (1,3). Therefore, we conclude that the PGD peptide has ϳ72% ␣-helix when complexed with Ca 2ϩ -CaM. Interestingly, when we added a second equivalent of peptide, we observed a further increase in the amount of ␣-helix in the complex, which is more than would be obtained from addition of unbound PGD peptide (Fig. 3). Our CD results are consistent with the idea that Ca 2ϩ -CaM is able to bind 2 eq of the PGD peptide, both in an ␣-helical conformation. In addition, isotope-  (22). Note that the Ca 2ϩ -CaM band did not disappear until 2 eq of the PGD peptide were added. Also note the decreased mobility of the band from the Ca 2ϩ -CaM⅐PGD peptide complex compared with the band from the 1:1 Ca 2ϩ -CaM⅐MLCK peptide complex. The Ca 2ϩ -CaM⅐constitutive nitricoxide synthase peptide complex (1:1 ratio) ran identical to the MLCK complex (data not shown). filtered Fourier transform infrared spectroscopy (34) also confirmed that the PGD peptide adopts an ␣-helix when bound to Ca 2ϩ -CaM. 3 We next performed a trifluoroethanol (TFE) titration experiment to study the ␣-helix-forming potential of the PGD peptide. The ␣-helical content of the PGD peptide increases with increasing TFE concentration up to 20% TFE, after which it levels off (data not shown). The percentage ␣-helix formed in the PGD peptide is ϳ68% in 20% TFE aqueous solution. Since TFE is a well known ␣-helix-stabilizing solvent (35,36), the observation that the ␣-helix of the PGD peptide can be further stabilized by TFE is as expected. In many of our studies concerning peptide binding to CaM, we have observed that the extent of ␣-helix formation induced in the peptide by binding to CaM or by addition of TFE is identical (36). Also for the PGD peptide, these two values are in close agreement (72 and 68%, respectively).
Fluorescence Spectroscopy-The outcome of the steady-state Trp fluorescence experiments are presented in Fig. 4. The PGD peptide has a maximum emission wavelength at 353 nm, which is typical for a solvent-exposed Trp residue in a peptide (Fig.  4A). When Ca 2ϩ -CaM was titrated into a PGD peptide solution, we found a significant blue shift of the maximum emission wavelength (note that there are no Trp residues in CaM). The maximum emission wavelength recorded for the complex is 335 nm (Fig. 4A). This relatively large blue shift indicates that the Trp residue(s) in the peptide move from a solvent-exposed environment to a hydrophobic environment in the complex, which is quite common for CaM-binding peptides (21). A change in the microenvironment around the Trp residues in the PGD peptides upon binding to Ca 2ϩ -CaM was also observed by near-UV CD spectroscopy (data not shown). However, we noted that the fluorescence intensity of the CaM⅐PGD peptide complex experienced only a small increase; this is quite different from most of the CaM-binding peptides, which typically experience a doubling or tripling of the fluorescence quantum yield upon binding to CaM (21). This could be due to the fact that the PGD peptide contains negatively charged residues (Asp and Glu) as opposed to normal CaM-binding peptides and that the carboxyl side chain of Asp and Glu residues can have quenching effects on the Trp fluorescence. Another possibility is that the two Trp residues in the bound dimeric PGD peptide may sit in a unique environment upon binding to Ca 2ϩ -CaM, so the higher Trp fluorescence, which is normally seen when it resides in a hydrophobic environment, is quenched. Titration experiments of the PGD peptide with Ca 2ϩ -CaM showed that the maximum emission wavelength already reached a plateau (337 nm) when only 0.5 eq of Ca 2ϩ -CaM was added (Fig. 4, A and B). This result indicates again that the PGD peptide binds to Ca 2ϩ -CaM in a 2:1 ratio (Fig. 4B). In addition, we did not find any changes in the Trp fluorescence of the PGD peptide upon adding apo-CaM (with 5 mM EDTA in the sample) (data not shown). Therefore, the interactions between CaM and the PGD peptide take place only in the presence of calcium.
Fluorescence quenching experiments with small molecules such as CsCl were also performed. The Trp residue in the PGD peptide was effectively quenched by increasing concentrations of CsCl, indicating a fully solvent-exposed Trp residue in the peptide. At 1:1 and 2:1 ratios of PGD to Ca 2ϩ -CaM, the Trp residues in the bound peptide were inaccessible to the quenching agent (Fig. 4C). These data again indicate that the PGD peptide binds to Ca 2ϩ -CaM in a 2:1 ratio and that the indole rings of the Trp residues from both peptides are shielded from the solvent in the complex.
To obtain further information about the potential arrangement of the two PGD peptides when bound to Ca 2ϩ -CaM, we recorded the fluorescence spectra of the PGD peptide complexed with four CaM variant proteins with different Met substitutions in the presence of calcium. We have successfully used the same approach to determine the orientation of the CaMbound MLCK and CaM-dependent protein kinase I peptides (21). As expected, the maximum emission wavelength shifted from 353 to 335 nm when the PGD peptide bound to wild-type-CaM. A small increase in the fluorescence quantum yield was also observed (Fig. 5). We then used CT-CaM, the CaM mutant in which all four Met residues in the C-terminal lobe of CaM (Met-109, Met-124, Met-144, and Met-145) are mutated to Leu residues (37). Because of the absence of the sulfur atoms in its C-terminal domain, CT-CaM is less efficient at quenching, compared with WT-CaM, when the Trp residue of the CaMbinding peptide interacts predominantly with the C-terminal lobe of CaM (21). In Fig. 5, we see that the fluorescence intensity of the complex between CT-CaM and the PGD peptide is greatly enhanced compared with that of the WT-CaM⅐PGD peptide complex; hence, this spectrum suggests that at least one of the Trp residues of the bound PGD peptides interacts with the hydrophobic patch in the C-terminal lobe of CaM. Subsequently, SeMet-CaM was used for fluorescence spectroscopy. The selenium atoms in SeMet can act as very efficient quenchers of proximal Trp residues (21). Indeed, significant quenching effects were observed for the bound PGD peptides (Fig. 5), suggesting that both Trp residues in the bound PGD peptides are in contact with the hydrophobic surfaces in SeMet-CaM, where the SeMet residues can directly interact with both Trp residues. Finally, we used the CaM variant SeMet-CT-CaM, in which the five Met residues in the N-terminal lobe of CaM (Met-36, Met-51, Met-71, Met-72, and Met-76) are substituted with SeMet and the four Met residues in the C-terminal lobe of CaM are mutated to Leu residues. The fluorescence spectrum of the complex between the PGD peptide and SeMet-CT-CaM was similar to that of the WT-CaM⅐peptide complex (Fig. 5). This spectrum, taken together with the previous ones, suggests that the PGD peptide can perhaps form a dimer when bound to Ca 2ϩ -CaM (see "Discussion" for alternative interpretation). In this complex, one Trp residue interacts predominantly with the C-terminal lobe of CaM, whereas the other Trp residue is bound to the N-terminal lobe of CaM. In the complex with SeMet-CaM, the fluorescence of both Trp residues is quenched; the SeMet residues in the N-terminal lobe of SeMet-CT-CaM quench the fluorescence of the one Trp residue located in the N-terminal lobe of CaM, whereas the fluorescence of the other Trp residue bound to the C-terminal lobe of CaM is enhanced (as it is in CT-CaM) because of the replacement of the sulfur atoms with carbon atoms in this domain. This proposed structure for the bound PGD dimer seems likely because Ca 2ϩ -CaM possesses a nearly perfect 2-fold symmetry when it binds to some target peptides (9 -11). However, it is difficult to rigorously exclude other models on the basis of the above fluorescence data alone, and the final explanation of these results must await the determination of the high resolution solution structure of the Ca 2ϩ -CaM⅐PGD peptide dimer complex by NMR.
NMR Spectroscopy-We have studied the interaction between the PGD peptide and Ca 2ϩ -CaM using [methyl-13 C]Metlabeled CaM and two-dimensional 1 H-13 C HMQC NMR spectroscopy. Because the Met side chains make up almost 50% of the exposed surface area of the two hydrophobic patches in Ca 2ϩ -CaM (12), this approach is very useful for the study of complexes of Ca 2ϩ -CaM and CaM-binding peptides (1, 3). In the absence of the PGD peptide, the NMR spectrum for Ca 2ϩ -CaM showed nine Met methyl resonances, which were assigned to specific residues by studying nine single Met 3 Leu mutants of CaM (Fig. 6) (38). When the PGD peptide was added, we detected nine new NMR resonances for the complexes that are more spread out, in addition to the nine original peaks from Ca 2ϩ -CaM. The appearance of new peaks suggests slow exchange on the 13 C NMR time scale, indicating relatively tight binding. When 1 eq of the peptide was added, the intensity of all 18 resonances in the spectrum was identical. After 2 eq of the PGD peptide were added, the nine resonances from Ca 2ϩ -CaM had completely disappeared, and the nine resonances for the complex were left in the spectrum (Fig. 5). The remaining nine resonances originated from the nine Met methyl groups of CaM in the Ca 2ϩ -CaM⅐PGD peptide complex (1:2 ratio). Further titration of the PGD peptide up to 3 eq imposed no further changes on the NMR spectrum (data not shown). Given that only nine resonances were observed, it appears that only one unique protein-peptide complex was formed. Again the NMR titration experiment showed that the PGD peptide bound to Ca 2ϩ -CaM at a 2:1 ratio and that binding to the N-and Cterminal lobes of CaM occurred simultaneously. As no intermediate species were observed in the titration experiment, models other than the binding of a dimer species seem unlikely.
The nine Met methyl groups of Ca 2ϩ -CaM are solvent-exposed, whereas almost all of them become fully buried inside the protein upon binding to some CaM-binding domain peptides (2, 9 -11). The exposure of the Met methyl groups in the 1:2 Ca 2ϩ -CaM⅐PGD peptide complex has been studied using the soluble spin label TEMPOL in combination with two-dimensional 1 H-13 C HMQC NMR spectroscopy. 2 No significant linebroadening effects were observed for the complex when 4 eq of TEMPOL were added (Fig. 7); this indicates that all nine Met methyl groups of CaM become buried inside the Ca 2ϩ -CaM⅐PGD peptide complex. In contrast, the nine Met methyl  There is no difference between the CsCl quenching profiles of the 1:1 (छ) and 1:2 (E) ratios of the Ca 2ϩ -CaM⅐PGD peptide complex. Please see "Results" for explanation. groups of Ca 2ϩ -CaM are fully exposed since addition of only 0.4 eq of TEMPOL already dramatically broadened these nine resonances (Fig. 6). From the two-dimensional NMR experiments, we can conclude that all nine Met methyl groups of CaM are involved in the interaction with the PGD peptide. Thus, the interactions between the PGD peptide and Ca 2ϩ -CaM are hydrophobic in nature, as shown before for other CaM-binding peptides (9 -11).
Since the PGD peptide showed a significant amount of ␣-helix (ϳ31%) in aqueous solution (Fig. 3), we attempted to assign the spectra of the peptide in aqueous solution by standard two-dimensional homonuclear and heteronuclear NMR techniques (39). The NOE summary and the ␣-1 H and ␣-13 C chemical shift deviations for the PGD peptide in 90% H 2 O and 10% D 2 O, pH 5.0, at 298 K are shown in Fig. 8. Continuous NN (i, iϩ1) NOE and numerous medium range NOEs (␣N (i, iϩ3) , ␣N (i, iϩ4) , and ␣␤ (i, iϩ3) ) indicate that the peptide adopts a partially ␣-helical structure in aqueous solution. The 1 H NMR chemical shifts of the ␣-1 H protons of the majority of the residues in the peptide move upfield compared with the corresponding random coil chemical shifts. This trend confirms the ␣-helical structure of the PGD peptide. Similarly, the downfield shift trend for the majority of the ␣-13 C chemical shifts supports an ␣-helical structure for the isolated PGD peptide as well (Fig. 8). Combining the NOE summary and the chemical shift data, we found that the PGD peptide adopts an ␣-helix from Asp-5 to Lys-24, with a possible extension to Lys-26. There were no indications that the peptide existed as a dimer in aqueous solution since there were no line width changes when the peptide was diluted to 0.2 mM (data not shown), and no intermolecular NOEs were observed in the NOESY spectra. The sample contained 2 mM PGD peptide in 90% H 2 O and 10% D 2 O, pH 5.0 (uncorrected for the isotope effects). The mixing time of the NOESY experiment used for the assignment was 250 ms. A NOESY spectrum with a 100-ms mixing time was also run to check for spin diffusion effects (data not shown). The chemical shift values of the amino acid residues in random coil structures were taken from Wishart and Sykes (30).

DISCUSSION
In this study, we have characterized a 26-residue CaM-binding domain peptide from petunia glutamate decarboxylase. This CaM-binding peptide is unusual because it contains five negatively charged residues (one Asp and four Glu residues), which are normally absent in CaM-binding domain peptides (2,3). We studied the interactions between the PGD peptide and Ca 2ϩ -CaM by gel band shift assays, CD, fluorescence, and NMR spectroscopy. The band shift results and the Trp fluorescence and two-dimensional NMR titration experiments indicated that two PGD peptide molecules bind simultaneously to one Ca 2ϩ -CaM molecule (Figs. 2, 4B, and 6). The two-dimensional NMR spectra further suggest that the PGD peptides bind to both lobes of Ca 2ϩ -CaM simultaneously and that one unique complex is formed (Fig. 6). The formation of a unique 1:2 Ca 2ϩ -CaM⅐PGD peptide complex was also detected by electrospray mass spectrometry. 4 The CD spectra demonstrated that the PGD peptides bind to Ca 2ϩ -CaM in an ␣-helical conformation (Fig. 3). The emission wavelength blue shift of the Trp fluorescence spectra, the CsCl quenching experiments and the near-UV CD spectroscopy results indicate that upon binding to Ca 2ϩ -CaM, the Trp residue of the PGD peptides becomes buried and shielded from the solvent (e.g. Fig. 4). The fluorescence data with the CaM variants show that the Trp residues of the two bound PGD peptides are associated with the N-and C-terminal lobes of the protein (Fig. 5). The NMR TEMPOL results pinpoint the complete burial of all nine Met side chains upon formation of the Ca 2ϩ -CaM⅐PGD peptide complex (1:2 ratio) (Fig. 7). Since the alternating positive and negative charges on the hydrophilic face of the PGD peptide could give rise to ion pair formation in the peptide dimer interface, it is possible that the PGD peptide binds as a dimer, where the hydrophobic faces on both PGD peptide monomers interact with the Met-rich surfaces of Ca 2ϩ -CaM (Fig. 1). In this way, Ca 2ϩ -CaM can avoid the unfavorable negative charges on the PGD peptide. This mode of binding would represent a new motif for Ca 2ϩ -CaM-peptide interactions, and this complex could potentially be formed because of the high flexibility of the central linker region of CaM, which allows the two lobes of the protein to orient with respect to each other in different manners (9 -11).
If dimerization of the PGD peptide occurs, it seems to be induced by Ca 2ϩ -CaM binding. It might be insightful to first discuss whether the PGD peptide does form a dimer in solution. In fact, there is strong evidence to suggest that the PGD peptide itself is a monomer in aqueous solution. First, the ␣-helical content of the PGD peptide in aqueous solution measured by CD spectroscopy (ϳ31%) is close to the ␣-helical content predicated by the Agadir program (ϳ38%), which assumes that the peptide is monomeric in solution. Second, the CD spectra of the PGD peptide in aqueous solution did not change when we measured at concentrations ranging from 10 M to 1 mM (data not shown). Third, no changes in NMR line width were detected when the PGD peptide was diluted 10-fold between 2 mM and 200 M. Fourth, no peptide dimer species were detected by electrospray mass spectrometry when the PGD peptide was studied under benign buffer conditions, where its 2:1 complex with Ca 2ϩ -CaM was detected. 4 Finally, a single species of the PGD peptide was detected on a Superdex-200 peptide column, and its molecular mass falls in the range of similar sized monomeric peptides, although no strict linear molecular mass relationship could be established for these peptides (data not shown). Taken together, we feel that the evidence points toward the PGD peptide being a monomer in aqueous solution and that dimerization of the peptide may be induced by Ca 2ϩ -CaM binding. It is of course still possible that a monomerdimer equilibrium exists in solution, with a weak association constant where the equilibrium favors the monomer. CaM could then select the dimer, rather than the monomer, from solution because of its more favorable electrostatic properties.
Normally, the binding of a Trp residue of the peptide to the C-terminal domain of CaM initiates the peptide complex formation (1,21,40,41). Subsequently, the N-terminal lobe binds to the hydrophobic and positively charged residues of the region of the peptide 10 -12 residues away from the Trp residue. In the case of PGD, the required recognition site (WKK) for binding to the C-terminal domain of CaM is present, which would initiate binding of a PGD monomer, as it does for other peptides. The negative charges in the PGD peptide are, however, disposed in such a way that interaction of the N-terminal lobe of CaM with the region either N-or C-terminal to the peptide would not be energetically favorable. Therefore, the binding of a second peptide to the hydrophobic surface in the N-terminal region of CaM could now occur. Clearly, sequential binding of the PGD peptides to the C-and N-terminal lobes, respectively, of CaM was not observed in our experiments. As indicated above, binding of an ion-paired PGD peptide dimer would be consistent with our data. However, there is one other possibility that we cannot exclude at this time. Since the twodimensional NMR data show that only one unique complex is formed, binding of PGD to the N-terminal lobe would have to be a positive cooperative process in which binding of the first PGD peptide to the C-terminal lobe facilitates the binding of the second peptide to the N-terminal domain. Only in this way can we explain why only the complex of two peptide molecules and one CaM molecule is observed in our experiments. Evidence for communication between both lobes of CaM has recently been reported by several groups (42)(43)(44)(45)(46). In this way, it would not be necessary for the two bound PGD peptides to interact with each other in the complex.
In most of the well characterized CaM⅐peptide complexes studied to date, Ca 2ϩ -CaM always binds to a single ␣-helical peptide (9 -11). 5 However, some studies have reported that oligomerization of the target proteins takes place in the presence of Ca 2ϩ -CaM. For example, it has been reported that CaM promotes dimerization of the oxygenase domain of human endothelial nitric-oxide synthase and that the dimerization is important for the activation process of endothelial nitric-oxide synthase (47). However, fluorescence titrations with dansyl-CaM indicated that the CaM-binding domain of endothelial nitric-oxide synthase oxygenase binds to Ca 2ϩ -CaM with a 1:1 stoichiometry (47). CaM can also regulate phosphofructokinase activity by preferentially binding to the dimer protein and preventing the formation of active tetramer protein, but also in this case, CaM is thought to bind to each subunit of the phosphofructokinase dimer (48). Smooth muscle MLCK can also bind to Ca 2ϩ -CaM in an oligomeric form, and it is known that oligomerization of smooth muscle MLCK decreases its enzymatic activity (49). Unfortunately, the details of the binding of Ca 2ϩ -CaM to oligomeric smooth muscle MLCK remain to be clarified (49), and in fact, a high resolution x-ray structure of 4 J. Chen, T. Yuan, H. J. Vogel, and G. Lajoie, unpublished results. 5 Studies in which two peptides bind to CaM have been reported, such as for phosphorylase kinase (53)(54)(55) and caldesmon (56 -58). However, in all these cases, two peptides with different amino acid sequence have been found to interact simultaneously with CaM. This study appears to be the first in which a complex of CaM with two full-length identical CaM-binding peptides bound simultaneously. Studies with relatively short peptides, encompassing only partial CaM-binding sequences, have also shown the capacity of CaM to form 2:1 complexes with one peptide sequence (e.g. Ref. 59), although these represent relatively weak binding. the complex of a CaM-binding peptide derived from smooth muscle MLCK and Ca 2ϩ -CaM shows that it forms a 1:1 complex (10). One of the most extensively studied CaM target proteins is human erythrocyte plasma membrane Ca 2ϩ -ATPase. The active form of the Ca 2ϩ -ATPase is either a dimeror monomer-CaM complex as determined by equilibrium ultracentrifugation (50). Interestingly, Vorherr et al. (51,52) demonstrated that the CaM-binding domain of this protein is indispensable for the oligomerization and stimulation of Ca 2ϩ -ATPase. Be that as it may, the CaM-binding domains in all four target proteins discussed in here are typical CaM-binding peptides with the potential to form a basic amphipathic ␣-helix (2). They do not contain negatively charged residues, unlike the PGD peptide, which has a total of five negatively charged residues (Fig. 1). Motifs in which two distinct peptides derived from different segments of the amino acid sequence of a target protein bind simultaneously to CaM have also been reported in the case of phosphorylase kinase (53)(54)(55) and caldesmon (56 -59), for example. This is the first report of two identical peptides that bind simultaneously to CaM with high affinity.
The fact that Ca 2ϩ -CaM can bind to two PGD peptides simultaneously should have implications for the activation mechanism of plant GAD, i.e. binding of Ca 2ϩ -CaM should facilitate the dimerization or oligomerization of plant GAD, and only dimerized or oligomerized GAD has the enzymatic activity. Recent work by Fromm and co-workers (13) has provided some evidence for this hypothesis. They reported that GAD forms an ϳ500-kDa complex with Ca 2ϩ -CaM, which was detected in plant extracts by nondenaturing PAGE. Notably, this complex was not formed if the C-terminal CaM-binding domain was deleted from the protein. In addition, in the presence of Ca 2ϩ , the specific GAD activity in the extracts of a GAD transgenic tobacco is ϳ40-fold higher than in a GAD⌬C transgenic plant, which lacks the CaM-binding domain in its GAD gene construct. The above two lines of evidence support, but do not prove, our hypothesis about the role of Ca 2ϩ -CaM-induced oligomerization in the activation process of GAD. Further studies using intact plant GAD and Ca 2ϩ -CaM are necessary. Nevertheless, our present studies of the Ca 2ϩ -CaM⅐PGD peptide complex clearly demonstrate that it represents a novel motif of CaM-peptide interaction. Determination of the high resolution solution structure of this complex will improve our understanding of CaM-induced dimerization/oligomerization processes that seem to accompany the activation of some of CaM target proteins.