Identification of Calmodulin Isoform-specific Binding Peptides from a Phage-displayed Random 22-mer Peptide Library*

, Plants express numerous calmodulin (CaM) isoforms that exhibit differential activation or inhibition of CaM-dependent enzymes in vitro ; however, their specificities toward target enzyme/protein binding are uncertain. A random peptide library displaying a 22-mer peptide on a bacteriophage surface was constructed to screen peptides that specifically bind to plant CaM isoforms (soybean calmodulin (ScaM)-1 and SCaM-4 were used in this study) in a Ca 2 (cid:1) -dependent manner. The deduced amino acid sequence analyses of the respective 80 phage clones that were independently isolated via affinity panning revealed that SCaM isoforms require distinct amino acid sequences for optimal binding. SCaM-1-binding peptides conform to a 1-5-10 ((FILVW) XXX (FILV) XXXX (FILVW)) motif (where X denotes any amino acid), whereas SCaM-4-binding

CaM 1 is a ubiquitous intracellular Ca 2ϩ receptor involved in transducing a variety of extracellular signals (1)(2)(3). In contrast to mammals, many plant species belong to a CaM multigene family that encodes various CaM isoforms (4 -5). Over 30 genes encoding CaM isoforms are found in the complete nucleotide sequence of Arabidopsis thaliana. 2 CaMs function by modulating or regulating the activities of their target proteins (CaMbinding proteins). More than 30 CaM-binding proteins have been identified, including enzymes such as kinases, phosphatases, and nitric-oxide synthase, as well as receptors, ion channels, G-proteins, and transcription factors (7)(8)(9). Plants seem to have evolved a unique repertoire of CaM targets whose homologues in animals do not appear to be modulated by CaMs. The shear number of CaM isoforms and the diversity of CaM targets imply that these proteins in plants likely modulate a broad spectrum of processes.
CaM is dumbbell-shaped with the N-and C-terminal globular domains separated by a flexible central helix. Each lobe contains two helix-loop-helix Ca 2ϩ -binding motifs referred to as "EF-hands" that are interconnected by a small ␤-sheet between the two Ca 2ϩ -binding loops (10). Ca 2ϩ binding to CaM induces conformational changes that expose hydrophobic amino acid residues on the surface of both lobes. This creates two hydrophobic pockets that are important for target peptide binding. CaM binds to a large number of proteins through interactions with specific CaM-binding domains. Therefore, one of the most intriguing questions concerning the interaction of CaM with target proteins is: how does a phylogenetically conserved protein like CaM specifically interact with so many different target sites? The determination of the three-dimensional structures of the CaM-peptide complexes greatly aided in answering this question (11)(12)(13)(14)(15).
Many known Ca 2ϩ -dependent CaM-binding proteins possess a region that is often characterized by an amphipathic helix consisting of ϳ20 amino acid residues. This region contains two hydrophobic/aromatic residues that are separated by 12 intervening residues that anchor the peptide to the two lobes of CaM (2). However, sequence analyses based upon this criteria do not always identify the CaM-binding region of a protein, and therefore, the CaM-binding regions of target proteins do not always fit these criteria. Furthermore, the presence of a novel class of CaM-binding proteins that assume a non-helical conformation in the CaM complex has been suggested (16). The CaM-binding domains of these proteins are characterized by the dominance of basic amino acids in contrast to the canonical motif in which hydrophobic amino acids are dominant. An attempt was made to examine the elements common to the many reported CaM-* This work was supported by Grant 2000-2-20900-001-1 from KO-SEF, the National Research Laboratory (2000), and BK21 Program. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
binding regions from different proteins. From sequence comparisons of CaM-binding peptides and CaM-binding domains within the CaM-regulated proteins, three classes of CaM-binding motifs have emerged (17). They include a modified variant of the IQ motif as a consensus for Ca 2ϩ -independent binding and two related motifs for Ca 2ϩ -dependent binding, termed 1-8-14 and 1-5-10 (based on the conserved hydrophobic residues within these motifs). Based on these structures and sequence information, a web-based data base is available for quickly searching for a potential CaM-binding site in a given protein sequence (18).
In plant cells, multiple CaM genes code for numerous CaM isoforms in wheat (19), potato (20), Arabidopsis (21), and soybean (22). We have recently cloned five CaM isoforms from soybean (SCaM-1-5). Although some of these isoforms (i.e. SCaM-1-3) are more than 90% identical to mammalian CaM, two (SCaM-4 and SCaM-5) exhibit only a 78% homology with SCaM-1 and are therefore the most divergent isoforms reported thus far in the plant and animal kingdoms. SCaM-4 is considered a bona fide CaM isoform based on the following characteristics. At primary structure level, SCaM-4 has conserved four putative EF-hands and a central linker region, hallmarking structural features of CaM (22). In addition, most of amino acid exchanges occur outside EF-hands, and the number of total amino acid residues are also conserved (148 amino acid residues for SCaM-1 and mammalian CaM, and 149 amino acid residues for SCaM-4 (22)). When compared with the preferred consensus amino acid residues of EF-hands derived from known Ca 2ϩ -binding proteins (23), the residues in all four Ca 2ϩ -binding loops of SCaM-4 conform to the consensus (Fig.  1), suggesting that SCaM-4 can bind four Ca 2ϩ molecules. All SCaM isoforms including SCaM-4 are ubiquitously expressed in various plant tissues and show similar subcellular localization patterns to the highly conserved SCaM-1 (22,24). Intriguingly, the cellular level of SCaM-4 rapidly rises upon specific stimuli such as pathogen infection in the same cells constitutively expressing SCaM-1 (25). Furthermore, SCaM-4 has the ability to modulate activity of many CaM-dependent enzymes.
However, SCaM-4 has distinguished ability from SCaM-1 in the activation of CaM target enzymes, which can be categorized into three different types (22, 26 -29) as follows: 1) enzymes activated equally by both SCaM-1 and SCaM-4 (e.g. phosphodiesterase (PDE), plant Ca 2ϩ -ATPase, plant glutamate decarboxylase, and CaM-dependent protein kinase II); 2) enzymes activated only by SCaM-1 (e.g. calcineurin, myosin light chain kinase, red cell Ca 2ϩ -ATPase, and plant NAD kinase); and 3) enzymes activated only by SCaM-4 (e.g. nitric-oxide synthase). SCaM-1 and SCaM-4 also exhibit differences in their Ca 2ϩ concentration requirements for target enzyme activation (28). Studies defining CaM-binding sequences and analyzing their interactions with CaM have shed light on the specificity of the interactions between CaM and the particular target molecules.
Here we have adapted an approach for defining the SCaM-1 and SCaM-4-binding peptide sequences using a random peptide bacteriophage display library. Bacteriophage display and affinity selection of phage-displayed peptide libraries, a technique based on screening a library of foreign peptides displayed on the surface of M13 bacteriophage, has proven to be a very useful tool for characterizing a number of protein-protein interactions. Due to the physical linkage of the expressed peptide with its genetic sequence, libraries numbering from 10 8 to 10 10 peptides have been rapidly screened for a wide variety of applications. This useful tool has been used to map antibody epitopes and to discover peptide ligands for membrane receptors and cytosolic proteins in recent years (30). In this work, we have searched SCaM isoform (SCaM-1 and -4)-favored peptide sequences from a phage display library, and we defined novel SCaM-1-and SCaM-4-specific binding sequences. Furthermore, we have analyzed the interactions between SCaM-1 and SCaM-4 and their respective binding peptides using a variety of biochemical techniques including gel overlay, gel mobility shift, fluorescence spectroscopy, and phosphodiesterase competition assays. Our data show that SCaM isoforms possess different sequences for optimal binding to specific targets. This may provide a molecular basis for CaM isoform-specific function in plants.

EXPERIMENTAL PROCEDURES
Construction of a Random 22-mer Phage Library-The phagemid vector pCANTAB5E (Amersham Biosciences) was modified to construct a vector for our library ( Fig. 2A). Two oligonucleotides (5Ј-CCGGCCG-GCGATATCAGCGGC-3Ј and 5Ј-GGCCGCGCCGCTGATATCGCCGG-CCGGCT-3Ј) were mixed, heated to 75°C for 10 min, and allowed to hybridize by cooling slowly at room temperature for 30 min. The hybrids were inserted at the NotI and SfiI site of pCANTAB5E to introduce NotI and EcoRV sites and a sequence coding for junction amino acid residues. The resulting phage peptide library vector was named pCANTA-B5F. These vectors were digested with NotI and EcoRV to clone double-stranded DNA coding for random peptides. This process is shown in Fig. 2B. Single-stranded DNA encoding 22-residue random peptides (5Ј-(NNK) 22 GCGGCCGCAGGTGCGC-3Ј) and one oligonucleotide (5Ј-G-CGCACCTGCGGCCGC-3Ј), in which N is A, C, G, or T (equimolar), and K is C or T (equimolar) were hybridized and converted into doublestranded DNA using Klenow DNA polymerase. The DNA was cut by NotI to produce one blunt end and one NotI site and then ligated into pCANTAN5F. The resulting DNA sequence and amino acid sequence of the N-terminal region of gpIII are shown in Fig. 2C. Twenty electroporations, each using 0.5 g of the DNA constructed as above and 100 l of electrocompetent Escherichia coli TG1 cells, yielded 4.2 ϫ 10 8 transformants producing infectious phages. The helper phage M13K07 was used to produce a phage library from the bacteria harboring library phagemids.
Isolation of SCaM Isoform Binding Phages by Biopanning-The library was screened either with SCaM-1 or SCaM-4 isoforms (22) coated on 35 ϫ 10-mm plastic Petri dishes. These Petri dishes were coated with SCaM isoforms by incubating them with 1 ml of the SCaM protein (100 g/ml in TBS) for 1 h at 37°C. The residual binding capacity of the dish was blocked with 0.1% BSA. About 2. Consensus shows the most frequently observed amino acid residue in each position of the Ca 2ϩ -binding loop region of EF hand, which is derived from the comparison of known Ca 2ϩ -binding proteins (23). Underlined amino acid residues indicate higher than 70% conservation. Ca 2ϩ -binding ligands are denoted by asterisks, and residues conformed to the consensus are indicated by boldface letters.
0.1% Tween 20 were preincubated with 0.1% BSA for 1 h at room temperature to remove BSA-binding phages and were bound to the SCaM-coated dishes by incubating them overnight at 4°C. The plates were washed ten times for 5 min with 1 ml of TBS, 0.1% Tween 20 to remove unbound phages. Phages bound to SCaM-1 or SCaM-4 were eluted in TBS, 0.1% Tween 20 with 2 mM EGTA. The eluted phages were amplified by infecting logarithmic phase E. coli TG1. The helper phage M13K07 was introduced into infected cells to produce phages. The resulting phages were subjected to an additional two rounds of biopanning. SCaM-1 or SCaM-4-binding phages obtained by three rounds of the selection process were plated with E. coli TG1 to raise colonies. Each clone was grown in 1 ml of 2ϫ YT medium supplemented with 100 g/ml ampicillin at 37°C, shaking to a late log phase. After adding the Helper phage M13K07, the culture was further incubated overnight, and a clear phage supernatant was collected following a 5-min centrifugation. Each of the phages was transferred into a 96-well microtitration plate coated with the SCaM-1 and -4 isoforms as in the above biopanning method. These were tested for the binding affinities of the selected phages to SCaM-1 and -4 by enzyme-linked immunosorbent assay using an HRP-conjugated anti-M13 antibody (Amersham Biosciences). The reaction of HRP with 2Ј,2Ј-azino-bis(3-ethylbenzthiazoline-6-sulfonic acid) was quantified by measuring the absorbance at 410 nm in a microplate spectrophotometer.
Expression of Recombinant GST Peptide Fusion Proteins in E. coli-Inserts of selected clones were amplified by PCR using modified primers both upstream and downstream of the gene III sequence (5Ј-TTAGC-CATGGCTTTCTATGCGGCCCAG-3Ј and 5Ј-AGCGGACCCGGGTACG-GCACCGGCGCA-3Ј), where the underlined nucleotides represent the SmaI and NcoI recognition sequence. PCR was performed using a temperature program consisting of 25 cycles of 1 min at 95°C, 1 min at 55°C, and 1 min at 72°C, followed by one cycle of 5 min at 72°C. After amplification, the PCR products were digested with SmaI and NcoI and cloned in SmaI-and NcoI-cleaved GST fusion vector, pGEX-KG (31). The ligated DNAs were used to transform E. coli BL21 cells. DNA was prepared as above from the cultures of individual colonies, and recombinants were identified by digestion with SmaI and NcoI and confirmed by DNA sequence analysis. Recombinant DNA techniques including transformations, plasmid preparations, gel electrophoresis, ligations, and enzyme digestions were all carried out according to Sambrook et al. (32). Purification of DNA from agarose gels was performed with a QIAEX II purification kit (Qiagen).
For the preparation of GST fusion proteins, transformants were grown in liquid cultures with ampicillin selection to an optical density of 0.5 at 600 nm. 1 mM isopropyl-1-thio-␤-D-galactopyranoside was then added, and the cultures were further incubated for 18 h at 37°C. Cells were collected by centrifugation and stored at Ϫ20°C or less until used. Frozen cells were thawed on ice and resuspended in buffer M (20 mM Tris-HCl, pH 7.5, 0.1 mM EDTA, 5 mM 2-mercaptoethanol, 2 mM phenylmethylsulfonyl urea, 0.1 mM p-aminobenzamidine), and the cell suspensions were sonicated for 1 min at a 20% pulse. After clearing the lysate by centrifugation at 15,000 rpm for 10 min at 4°C, the supernatant (designated as the crude CaM-PEP preparation) was used for the binding assays.
CaM Gel Overlay Assay-To prepare horseradish peroxidase (HRP)conjugated CaMs, we conjugated SCaMs with a maleimide-activated HRP using the EZ-Link maleimide-activated HRP conjugation kit (Pierce) according to the manufacturer's instructions. Before conjugation, SCaMs were incubated in 50 mM Hepes, pH 7.0, 0.1 M dithiothreitol at 55°C for 1.5 h to reduce cysteine residues. The reduced SCaMs were washed and concentrated in degassed phosphate-buffered saline buffer using Centricon C10s (Amicon). After conjugation, the efficiency of SCaMs conjugation to HRP was determined by SDS-PAGE. Conjugation efficiency was usually greater than 90%. Unconjugated residual CaM was removed by gel filtration. CaM gel overlays were performed as described previously (33). Ten g of GST fusion protein was separated on an 11% SDS-polyacrylamide gel.
Peptide Synthesis-Peptides (Pep A, APAHALFHWGVLGSLIRLV-FLS, and Pep B, CNRLLLRSLRYWGYVVLSALRL) were synthesized at the Peptide Synthesis Facility (Peptron, DaeJun, South Korea) using an Applied Biosystems model instrument. Peptides were desalted and purified by reverse-phase high pressure liquid chromatography using a C 18 column, and their amino acid sequences were analyzed.
Gel Fluorescence Spectroscopy-The changes in the microenvironments of the peptide Trp residues upon binding to SCaM were monitored by fluorescence spectroscopy using a PerkinElmer Life Sciences LS50 luminescence spectrometer. The excitation and emission slit widths were 2 nm, and emission spectra scanning was carried out at 10 nm/min with a 1-cm path length cuvette. The Trp residues of Pep A and B incubated in Ca 2ϩ buffer (50 mM Mops, pH 7.5, 0.1 M KCl, 0.1 mM CaCl 2 ) or in EGTA buffer (50 mM Mops, pH .5, 0.1 M KCl, 2 mM EGTA) at room temperature for 30 min were excited at a 295 nm excitation wavelength, and the fluorescence emission spectra in the range 295-560 nm were recorded. An excitation wavelength of 295 nm was used to decrease Tyr fluorescence in SCaM. SCaM was added to the same cuvette from a highly concentrated stock solution to maximize the dilution effects.

FIG. 2.
Construction of the random peptide library. Recognition sequences and restriction sites of enzymes are indicated by boxes and edged lines, respectively. The letter N stands for an equal mixture of A, G, C, and T, and K stands for an equal mixture of Cys and Thr. A, procedure is described for construction of the peptide library vector pCANTAB5F. The amino acid sequence at the N terminus of gpIII and the corresponding DNA sequences are shown. Boldface letters correspond to the substituted or inserted residues. B, procedure is depicted for construction of double-stranded DNA coding for random peptides. C, amino acid sequence is listed for the random peptide, indicated as X 22 . The corresponding DNA sequence is also shown.
Circular Dichroism-CD spectra were obtained on a Jasco J-720 spectropolarimeter using a 1-mm cell at 25°C. The peptide concentration used for CD measurements was 50 M, and the solutions included 10 mM Tris-HCl, pH 7.5, and 50% 2,2,2-trifluoroethanol (TFE) (v/v). Secondary structure contents were calculated using the Circular Dichroism Deconvolution program (34).
Phosphodiesterase (PDE) Competition Assay-Cyclic nucleotide PDE assays were performed using commercially available bovine heart CaMdeficient PDE (Sigma). The initial 100-l reaction volume contained buffer (100 mM imidazole HCl, 2.56 mM cAMP, 5.13 mM MgSO 4 , 1.28 mM CaCl 2 ) and varying concentrations of SCaMs (5-200 nM) in the presence (100 nM of either Pep A or Pep B) or absence of peptides. The reaction was started by the addition of PDE (0.3 milliunits/l). The basal level of enzyme activity was determined in the absence of SCaM, and stimulated activity was determined in the presence of either SCaM-1 or SCaM-4 and CaCl 2 . After an incubation at 37°C for 30 min, the reactions were stopped by placing the reaction tubes in a boiling water bath for 5 min and then on ice for 2 min. Following a brief centrifugation step, 50 l of alkaline phosphatase (10 units) was added, and the samples were incubated for 10 min at 37°C. The reactions were stopped by adding 500 l of 10% trichloroacetic acid. After vortexing, the precipitates were pelleted, and the supernatants (400 l) were transferred to new tubes. One ml of phosphate reagent (22) was then added, and samples were incubated for 15 min at 37°C and assayed for P i content at OD 660 . The dissociation constants (K d ) of SCaM-1 and -4 for Pep A and B were calculated from the concentration of SCaM (nM) required to obtain half-maximal PDE activity either in the presence (100 nM) or absence of the peptides. The following equation (35) was used to calculate the dissociation constants:

Construction and Characterization of the Random 22-mer
Peptide Phage Display Library-The 22-mer phage display random peptide libraries were constructed using a modified pCANTAB vector, as described in Fig. 2. Random peptides of 22 amino acids in length are expressed at the N terminus of the M13 protein III bacteriophage. Initial transformation of the library produced 4.8 ϫ 10 8 independent clones and subsequent amplification yielded 2.3 ϫ 10 11 pfu/ml. To confirm the diversity of peptide sequences in the library, we randomly picked 80 individual phage clones and determined nucleotide sequences of random peptide regions. In Fig. 3A, the distribution of amino acids in the 22-mer random peptide library is reported as a percentage of the total. All amino acids were uniformly distributed in this library, taking into account the bias inherent in the codon usage ratio. The distribution of amino acid types that appears at each residue position of the sequenced inserts is shown graphically in Fig. 3B. Acidic (Asp and Glu), basic (Lys, Arg, and His), polar (Ser, Thr, Cys, Tyr, Asn, and Gln), and nonpolar amino acids (Gly, Ala, Val, Leu, Ile, Pro, Phe, Trp, and Met) are relatively proportional to their presence in the genetic code and generate not only linear sequence diversity but also the biochemical diversity desired in a library for screening.
Isolation of SCaM-1-and SCaM-4-binding Peptides-In an effort to study the binding sequence preferences of the SCaM-1 and SCaM-4 isoforms, we screened a phage display peptide library by biopanning with purified SCaM-1 and SCaM-4 proteins, respectively (see under "Experimental Procedures"). Briefly, SCaM-1 or SCaM-4 proteins immobilized on the surface of plastic Petri dishes were incubated with phages in the presence of Ca 2ϩ . Bound phages were eluted with EGTA after extensive washing. After three rounds of selection, isolated phage clones were tested for specific binding to SCaM-1 or SCaM-4 via an enzyme-linked immunosorbent assay-based assay (see "Experimental Procedures"). Approximately 90% of the isolates exhibited strong SCaM-1-or -4-binding activity (data not shown). Each of the 80 independent phages reacting strongly in the enzyme-linked immunosorbent assay were selected from the SCaM-1-and SCaM-4 biopanning, and their nucleotide sequences were determined. In Fig. 4, their deduced amino acid sequences are shown in random peptide regions, which are aligned to reveal consensus sequences. For convenience, we designated SCaM-1-binding peptide sequences as the ␣-series and SCaM-4-binding peptides as the ␤-series. Despite the considerable sequence diversity observed among these clones, their careful alignment yields two distinct binding motifs. Based on the positions of conserved hydrophobic residues, SCaM-1-binding peptides can be fit into a 1-5-10 motif, possessing a consensus sequence of ((FILVW)XXX(FILV)) (where X denotes any amino acid). In contrast, SCaM-4-binding peptides conformed to a 1-8-14 motif of ((FILVW)XXXXXX(FAIL-VW)XXXXX(FILVW)). In both cases, most of these peptide sequences contain one or more basic amino acid residues, which results in an overall positive net charge for these peptides. Identification of SCaM-1-or -4-specific Peptides by Gel Overlay Assay-To verify the specificity of SCaM-binding phage isolation using biopanning, a CaM gel overlay assay was performed (24). First of all, three of the most frequently isolated peptides obtained from each of the SCaM-binding phages (i.e. ␣1, ␣2, and ␣3 for SCaM-1 and ␤1, ␤2, and ␤3 for SCaM-4) were expressed as GST fusion proteins using the vector pGEX-KG (30). As shown in Fig. 5, these GST fusion proteins exhibited molecular masses of ϳ29 kDa by SDS-PAGE, consistent with the predicted values of ϳ3 kDa for the SCaM peptide plus ϳ26 kDa for the GST protein. In the gel overlay assays, all ␣ peptide fusion proteins exhibited positive signals with SCaM-1-HRP (Fig. 5A), whereas SCaM-4-HRP resulted in no signal. Conversely, for the ␤ peptide fusion proteins, only SCaM-4-HRP exhibited positive binding. Interestingly, CaM binding signals for these fusion proteins correlated with their frequency of isolation in the biopanning. For example, ␣1 and ␤1, the most frequently isolated sequences for SCaM-1 and SCaM-4 binding, respectively, produced the strongest signals among the tested ␣ and ␤ peptides in the gel overlays (Fig. 5, A and B). In either case, the negative control (i.e. GST protein alone) exhibited no binding to either of the SCaM-HRP probes. In addition, no significant signal was detected when the gel overlay assays were performed in the presence of an EGTA-containing buffer (data not shown). Thus, these results indicate that these peptide sequences specifically bind to the corresponding SCaM-1 or SCaM-4 in a Ca 2ϩ -dependent manner.
Gel Mobility Shift Assays-The gel overlay assays confirmed the binding of SCaM to the isolated peptide sequences. However, they did not provide quantitative information for these interactions. To examine further the SCaM-binding specificity of the isolated peptide sequences, the most frequently isolated peptide sequence was selected from each SCaM-1 or SCaM-4binding peptide sequence. These are designated as Pep A (APA-HALFHWGVLGSLIRLVFLS) and Pep B (CNRLLLRSLRYW-GYVVLSALRL), respectively. The synthesized peptides were then tested by gel mobility shift to assess their interactions with the SCaM isoforms. Complexes formed between the SCaM isoforms and either Pep A or Pep B were confirmed by gel electrophoresis in the presence of 4 M urea. Urea was used since it dissociates low affinity and nonspecific complexes, leaving only the higher affinity and higher specificity complexes behind. As the ratio of peptide to Ca 2ϩ -SCaM increased, we observed a band shift due to the formation of a complex between the peptide and Ca 2ϩ -SCaM (Fig. 6, A and D). In the absence of peptide, SCaM migrates as a single band (data not shown). When the ratio of peptide to SCaM was equal, a mo-

FIG. 4. Deduced amino acid sequences derived from the phage display library by affinity selection with SCaM-1 (A) or SCaM-4 (B).
Phages from a 22-mer random peptide display library were subjected to three rounds of selection on SCaM-1 or SCaM-4 affinity panning. Each of the 80 independent phages selected for its strong enzymelinked immunosorbent assay signal was selected from SCaM-1-or SCaM-4 biopanning. Their nucleotide sequences were determined, and the deduced amino acids are compared. Boldface residues indicate the three hydrophobic amino acids that are grouped into one of two types (the 1-5-10 ((FILVW)XXX(FILV)XXXX(FILVW)) motif represents the SCaM-1-binding peptides, and the 1-8-14 ((FILVW)XXXXXX(FAIL-VW)XXXXX(FILVW)) motif represents the SCaM-4-binding peptides) based on the positions of conserved hydrophobic residues. The frequency of appearance (n) and electrostatic charge of each of the isolates is indicated as shown.
FIG. 5. SCaM gel overlay of isolated peptides. The GST fusion proteins were prepared as described under "Experimental Procedures." Equal amounts (10 g) of total E. coli protein extracts were subjected to SDS-PAGE (11% gel), and stained by Coomassie Brilliant Blue (left 1st panels). The protein gels were transferred to nitrocellulose membranes and probed with anti-GST antibody (Ab) to detect the expression of GST fusion peptides (2nd panels from left) and probed further with either SCaM-1-HRP (3rd panels from left) or SCaM-4-HRP (right panels). The bound SCaM-1-HRP or SCaM-4-HRP was detected using an ECL system (Amersham Biosciences). M indicates protein molecular weight marker, A shows the SCaM-1-favored peptides, and B shows the SCaM-4-favored peptides. bility shift was observed for nearly all of the SCaM proteins. At a peptide to SCaM molar ratio of 1.5, virtually no free SCaM was detected. Taken together, these observations indicate that the peptides bind to the Ca 2ϩ -bound SCaM isoforms at a 1:1 molar stoichiometry. This is consistent with the previous observation (36) that most of the well characterized CaM-binding peptides, including peptides derived from myosin light chain kinase, constitutive nitric-oxide synthase, and CaM-dependent protein kinase I form a 1:1 complex with Ca 2ϩ -CaM. No mobility shift was observed when 5 mM EGTA was added to the samples, indicating that the interactions between the SCaM isoforms and the peptides are Ca 2ϩ -dependent (Fig. 6, B and  E). Interestingly, we could detect neither a SCaM-1-Pep B nor a SCaM-4-Pep A complex in the mobility shift assay (Fig. 6, C  and F). These results suggest that in the presence of Ca 2ϩ , Pep A and Pep B form a 1:1 complex with SCaM-1 and SCaM-4, respectively. Additionally, these peptides possess a certain degree of specificity toward the SCaM isoforms.
Trp Fluorescence Emission Spectra-The peptide interactions with SCaM-1 or -4 were also examined by taking advantage of the intrinsic fluorescence of the Trp residue in the peptide sequences. Fluorescence spectroscopy is a convenient method for determining the SCaM-peptide complex. The peptides have excitation and emission wavelengths of 295 and 295-560 nm, respectively. The maximum emission of the peptide in the absence of CaM occurred at 353 nm (Fig. 7). Upon addition of SCaM-1 or -4 to the reaction mixtures containing Pep A or Pep B in the presence of Ca 2ϩ , the maximum fluores-cence emission of the synthetic peptides shifted to shorter wavelengths (i.e. for SCaM-1 it was from 353 to 333 nm and for SCaM-4 it was from 353 to 335 nm) and exhibited ϳ5-fold increase in intensity. This relatively large blue shift indicates that the Trp residues were in a hydrophobic environment such as the interior of SCaM, which is typical for CaM-binding peptides (37). The addition of 2 mM EGTA to the same reaction mixture completely reversed the shift in the emission spectrum (Fig. 7). These results revealed that each of the SCaM peptides bind directly to its corresponding SCaM isoform.
Circular Dichroism Spectra-To examine the molecular basis of the binding specificity for the plant CaM isoforms, we characterized the secondary structures of the model peptides. Because shorter peptides tend to exist as flexible coils in pure water, we determined the secondary structures of the model peptides in the presence of 50% TFE, a compound that stabilizes secondary structures (38). In Fig. 8, the CD spectra for Pep A and Pep B obtained at 25°C are presented. As can be seen, the spectra are very similar to each other. Both contain a minimum at 206 -208 nm with a shoulder at around 222 nm, indicating the dominant presence of ␣-helices. From the CD spectra, the ␣-helix contents of Pep A and Pep B are estimated to be 72 and 76%, respectively. Thus, both Pep A and Pep B are likely to bind CaM isoforms as helices. In addition, in their SCaM-1-bound forms, both peptides showed increases in their ␣-helical content, further supporting this theory (data not shown).
Peptide-dependent Inhibition of CaM-activated PDE Activity-Finally, the relative affinities of Pep A and Pep B for the SCaM isoforms were determined by using a competition assay for PDE, an enzyme that is activated equally by both SCaM-1 and SCaM-4 (22). In Fig. 9, A and B, the effect of increasing peptide concentrations on PDE activity is shown for a fixed concentration (120 nM) of SCaM-1 or SCaM-4, a concentration sufficient for the maximal activation of PDE. A gradual decrease in PDE activity was observed with increasing concentrations of the relevant peptides. Half-maximal inhibition of PDE activation by SCaM-1 and SCaM-4 was obtained at ϳ120 nM Pep A and ϳ80 nM Pep B, respectively. Up to a concentration of 600 nM, Pep A had no effect on the activation of PDE by SCaM-4 and a concentration of more than 1 M was required to reach half-maximal inhibition. Pep B, on the other hand, was a slightly more potent inhibitor of PDE activation by SCaM-1, inhibiting at concentrations above 400 nM, and although similar to Pep A, half-maximal inhibition was not reached at concentrations below 1 M. To determine K d values for these peptides in the activation of PDE by SCaMs, the CaM dose-dependent activation of PDE was determined in the presence (100 nM) or absence of the peptides (Fig. 9, C and D). The activation curves shifted to the right in the presence of the peptides, indicating a competition occurring between PDE and the peptides for the SCaM isoforms. The K d values for Pep A and Pep B for their inhibition of the activation of PDE by SCaM-1 or SCaM-4 was determined to be 31.0 and 28.5 nM, respectively (Fig. 9, C and D). These experiments suggest that Pep A and Pep B have at least a 10-fold specificity for SCaM-1 and SCaM-4, respectively. DISCUSSION We have shown previously (22, 26 -28, 33) that SCaM isoforms exhibit differences in their abilities to activate target enzymes. Importantly, some CaM isoforms competitively inhibit the activation of certain enzymes by other CaM isoforms, exhibiting a reciprocal regulation of target enzymes (26,28). In these in vitro enzyme assays, the SCaM isoforms often possessed significantly different affinities for a given enzyme. For example, with the activation of plant Ca 2ϩ -ATPase, SCaM-1 has ϳ20-fold higher affinity than SCaM-4 (28). Furthermore, in gel overlay assays of various plant CaM-binding proteins, these isoforms exhibited different binding intensities, suggesting that different CaM isoforms might require distinct primary sequences for optimal target binding (24). In this investigation, in which we took advantage of a phage-displayed random peptide library, we identify the peptide sequences that interact with specific SCaM isoforms. Surprisingly, two distinct families of peptides that specifically bind to SCaM-1 or SCaM-4 isoforms in a Ca 2ϩ -dependent manner were identified in this study. These peptides fit into a 1-5-10 ((FILVW)XXX(FIL-V)XXXX(FILVW)) motif for the SCaM-1-binding peptides and a 1-8-14 ((FILVW)XXXXXX(FAILVW)XXXXX(FILVW)) motif for the SCaM-4-binding peptides. Rhoads and Friedberg (17) originally identified these motifs as Ca 2ϩ -dependent brain CaMbinding sequences, based upon the positions of conserved hydrophobic amino acid residues. These hydrophobic residues play an important role in anchoring peptides to target binding pockets of CaM via a hydrophobic interaction. It is also noteworthy that these Ca 2ϩ -dependent CaM-binding motifs require basic residues, which are important in stabilizing the binding interaction by forming salt bridges. Consistent with this, all of the isolated peptides in this study were determined to be basic.
It is quite remarkable that plant CaM isoforms exhibit differing optimal sequence preferences for their target interactions in vitro. This is most likely due to structural differences between CaM isoforms. For example, SCaM-1 and SCaM-4 are ϳ22% different in regard to their primary amino acid sequences. What is the significance of this finding in terms of the function of CaM in plant cells? One intriguing possibility is that certain CaM isoforms may require specific binding targets thereby leading to unique cellular responses. This is consistent with the idea that these CaM isoforms evolved independently following segregation from a progenitor CaM, as predicted from phylogenetic analyses of these CaM isoforms (22). However, we should also point out that these isolated peptides could be the optimal binding sequences for given CaM isoforms, which are not found in nature. Indeed, in CaM gel overlay assays using plant cell extracts, we could hardly find binding proteins specific for the CaM isoforms (24). In addition, so far none of the enzymes tested in vitro in our laboratory exhibit specific binding to particular CaM isoforms. Furthermore, the differences in binding affinity between Pep A and Pep B to the SCaM isoforms were no greater than 10-fold. Therefore, it is reasonable to speculate that a large number of CaM-binding proteins bind to both isoforms, given a certain degree of affinity differences, and therefore only a few CaM isoform-specific binding proteins may exist in plants.
We reported previously that SCaM isoforms are different in their abilities to activate various target enzymes (28, 33). We examined whether there is a correlation between the differential target binding preferences for the activation profiles of SCaMs. Interestingly, in the case of ␣-CaMKII, which has a 1-5-10 CaM-binding motif, the 1-5-10-favoring SCaM-1 activated this enzyme and at a much lower concentration than SCaM-4 (ϳ10-fold less, K act 22 versus 275 nM). Conversely, SCaM-4 was Ͼ4-fold better than SCaM-1 in the activation of nitric-oxide synthase, a 1-8-14-type enzyme (28). These positive correlations further support the biological relevance of our findings.
Previously, two other groups (39,40) reported isolating CaMbinding sequences using similar strategies. Dedman et al. (39) pioneered this type of work using a random 15-mer library, shorter than the typical CaM-binding domains, to screen for CaM-binding peptides. Among the 28 isolated peptides, Trp was always present and often (11 peptides) present in the first variable position of the random peptide inserts. Additionally, 17 of the peptides contained Trp-Pro sequences. Nevalainen et al. (40) used a random 8-mer peptide library and similarly found the Trp residue and the Trp-Pro combination in the isolated CaM-binding peptides. Because they used a random peptide library consisting of lengths shorter than those of the naturally occurring CaM-binding sequences, the peptides binding to CaM might be the flanking amino acid sequences present in the vector, potentially posing a significant bias. In this regard, the 22-mer random peptide library used in this study has merit over those peptides possessing pre-determined constraints, and it is also long enough to span the length of the hydrophobic binding surface of CaM. These advantages might prove useful in the isolation of CaM-binding sequences that are more like naturally occurring CaM-binding peptides. The majority of the 30 plus CaM-binding domains identified thus far consists of stretches of 16 -35 amino acid residues. Of the 33 peptides isolated in this study, Trp residues are found present at random positions (6 of them are actually devoid of the Trp residue), and only two of the peptides (i.e. ␣12 and ␤16) contain the Trp-Pro configuration. In addition, the 1-5-10 and 1-8-14 consensus motifs found in this study were originally identified in natural CaM-binding sequences, further arguing that the peptides isolated in this study resemble naturally occurring sequences.
What is the importance of our findings in regard to the structural perspective of the CaM-target peptide interaction? To address this, we built model structures of SCaM-1 bound to either M13 (of the 1-8-14 motif) or to the CaMKII CaM-binding sequence (of the 1-5-10 motif). The same models were prepared for SCaM-4. We then looked for any steric hindrance between SCaM and each of the peptides, hoping to discover any structural reason that would support our observation of the differential binding specificity. In short, we could not find any obvious reason that would clearly explain the basis of the binding discrimination of the SCaMs. However, we should note that this kind of modeling study assumes that both of the SCaM isoforms adopt the same conformation as that of target-bound mammalian CaM. Thus, the lack of a positive answer resulting from the model studies suggests two possibilities. First, SCaM isoforms may contain subtle differences in their side chain conformations of the residues directly interacting with target peptides despite the fact that their overall structures are similar to that of mammalian CaM. This would be similar to the structure of yeast CaM which is only 60% identical to mammalian CaM at the amino acid level (41). Indeed, we noticed that many Met residues in the SCaM isoforms were substituted for bulky hydrophobic amino acids such as Ile, Leu, and Val, which are involved in stabilizing the CaM-target peptide complex, by forming hydrophobic interactions and are also important for conferring flexibility to accommodate and recognize a wide range of targets (12). Alternatively, the overall structure of the SCaM isoforms might be different from that of mammalian CaM and may use different mechanisms for target binding than that of mammalian CaM. Therefore, only by high resolution three-dimensional structure determination of the complex with different peptides can the exact mechanism underlying the differential binding specificity of the SCaMs be determined.
The potential merits of this study are severalfold. First, the isolated CaM-binding sequences may serve as good references for searching for CaM-binding proteins in data bases or for mapping the CaM-binding domains of certain proteins. Second, these peptides may serve as useful tools for studying CaM isoform-specific functions in plant cells. For example, in regard to each other, Pep A and Pep B exhibit at least a 10-fold higher specificity for SCaM-1 and SCaM-4, respectively, in both gel overlay and PDE inhibition assays. Another advantage of these techniques is that one can study the function of CaM in different subcellular compartments by tagging the inhibitor peptide to various subcellular targeting sequences. For example, its nuclear function can be studied by fusing it to a nuclear localization signal (6). Currently these studies are underway in our laboratory.