Identification and Characterization of a Calmodulin-binding Domain in Ral-A, a Ras-related GTP-binding Protein Purified from Human Erythrocyte Membrane*

A 28-kDa protein (p28) has been purified from Triton X-100 extracts of human erythrocyte plasma membrane by calmodulin affinity chromatography. Based on internal peptide sequencing and its protein amino acid composition, this protein has been shown to be highly related, if not identical to, Ral-A, a Ras-related GTP-binding protein. This protein assignment is consistent with the findings that p28 binds [32P]GTP specifically and has low GTPase activity. In this study we describe the identification and characterization of a calmodulin-binding domain in Ral-A. The Ca2+-dependent interaction of p28 with calmodulin was first detected by a calmodulin affinity column. Gel overlay experiments of both p28 and recombinant Ral-A with biotinylated calmodulin provided strong evidence that Ral-A is a calmodulin-binding protein. A peptide of 18 residues (P18) with the sequence SKEKNGKKKRKSLAKRIR has been identified as a putative calmodulin-binding domain in Ral-A, because it comprises a basic/hydrophobic composition with the propensity to form an amphiphilic helix. P18 was synthesized, and its interaction with calmodulin by gel overlay was shown to be Ca2+-dependent. Circular dichroism analysis demonstrated that this interaction results in less α-helical content upon calmodulin complex formation. These results indicate that Ral-A is a calmodulin-binding protein, raising the possibility that it may be associated with Ca2+-dependent intracellular signaling pathways.

The Ras superfamily of low molecular mass (20)(21)(22)(23)(24)(25)(26)(27)(28)(29) GTP-binding proteins is involved in the regulation of a wide variety of cellular functions and in signal transduction (1)(2)(3). Functioning as a molecular switch the protein transduces signals in the active GTP-bound form and is converted to an inactive form when the bound GTP is hydrolyzed to GDP. Its intrinsic GTPase activity is regulated by guanine nucleotide exchange factors such as GDP dissociation stimulator (GDS), 1 GDP dissociation inhibitor, and GTPase activating protein (GAP).
Ral proteins represent a distinct family of Ras-related GTPbinding proteins. They share more than 50% amino acid sequence identity with Ras, their nucleotide binding and GTP hydrolysis activities being comparable with those of Ras. The Ral genes were originally isolated from a cDNA library of immortalized simian B-lymphocytes (4). The isolation of additional Ral cDNAs from human pheochromocytoma and HL-60 leukemia libraries has revealed the existence of two forms of the gene, ral-A and ral-B, which are about 85% identical but differ essentially in their C-terminal region (5). Ral gene products are expressed in human platelet membrane (6), as well as in most cell types, with particularly high levels in brain and testis (7,8). Ral proteins have a diverse subcellular localization, not only in plasma membrane, but also in cytoplasmic vesicles, including clathrin-coated vesicles and secretory vesicles (9,10). Like other members of the Ras superfamily, Ral proteins have their own set of highly specific regulatory factors. RalGAP has been identified and characterized from the cytosolic fraction of rat brain and mouse testis (11) as well as human platelet (12). RalGDS was found to interact with ras p21 and to function as a putative effector protein in Ras signaling pathways (13)(14)(15). Recently, several groups have reported that Ral proteins and RalGDS constitute a distinct downstream pathway from Ras that can induce cellular transformation in parallel with activation of the Raf/mitogen-activated protein kinase cascade (16 -18). This signaling pathway from Ras to Ral through RalGDS is selectively regulated by Rap1 (19,20). In addition, the effector protein of Ral, RLIP (Ral interacting protein), has been shown to contain a GAP region related to RhoGAP domains and to have GAP activity acting upon CDC42 and Rac (21). Moreover, Ral-A has been documented to be involved in the tyrosine kinase-mediated activation of phospholipase D, suggesting that the signaling pathway from Ras to Ral leads to the regulation of phospholipid metabolism (22). Ral proteins are biochemically well characterized GTPases. However, little is known about their physiological functions, because no biological activity associated with its overexpression in cells has been detected (11).
In this study, we present evidence that a monomeric GTPbinding protein with molecular mass of 28 kDa (p28) purified from human erythrocyte membranes is highly related, if not identical to, Ral-A, as determined by internal peptide sequencing and amino acid composition analysis. This has also been confirmed by the measurement of its GTP binding and hydrolysis. The data show for the first time that Ral-A is a calmodulin-binding protein. A putative calmodulin-binding domain in Ral-A was identified in its C-terminal region based on its basic/hydrophobic amino acid composition and propensity to form an amphiphilic helix. An 18-amino acid peptide (P18) with the sequence corresponding to this putative calmodulin-binding domain has been synthesized, and its interaction with calmodulin has been characterized. Very recently, Fischer et al. have reported the presence of a calmodulin-binding domain in the C-terminal region of Gem/Kir, a subfamily of Ras-related GTP-binding proteins (23). Rin, a neuron-specific, Ras-related GTP-binding protein, has been also shown to bind calmodulin through a C-terminal binding motif (24). The present study identifying a calmodulin-binding domain in Ral-A supports these two findings and in addition shows the calmodulin binding domain to be present in a Ras-related protein containing the CAAX motif, lacking in the previously identified examples. Work is currently in progress to investigate the function of this calmodulin-binding domain in Ral-A. Preliminary reports of this work have appeared previously in a conference proceeding (25).

EXPERIMENTAL PROCEDURES
Materials-Packed red blood cells were obtained from the Red Cross blood bank, Sydney (NSW, Australia). [␣-32 P]GTP (3000 C i /mmol) and [ 14 C]formaldehyde (40 -60 mCi/mmol) were purchased from DuPont NEN. Calmodulin-agarose, ATP, GTP, 5Ј-adenylylimidodiphosphate, avidin-alkaline phosphatase, nitro blue tetrazolium and 0.05% 5-bromo-4-chloro-3-indolyl phosphate, and the reagents for SDS-PAGE were from Sigma. Calmodulin and biotinylated calmodulin were obtained from Calbiochem. PVDF membranes were from Bio-Rad. An 18-amino acid peptide (P18) with the sequence of SKEKNGKKKRKSLAKRIR was synthesized by Chiron Mimotopes Pty. Ltd. (VIC, Australia). Other materials and chemicals were the highest grade available from commercial sources. Recombinant Ral-A was a generous gift from Drs. Hiroshi Koide and Yoshito Kaziro at the Faculty of Bioscience and Biotechnology, Tokyo Institute of Technology, Japan.
Protein Purification-p28 protein was purified from human red blood cell membrane as described previously (26), with the following modifications. Briefly, calmodulin-depleted human plasma membranes were prepared according to the method of Wang et al. (27). The membranes (0.5-1.0 g of protein) were solubilized for 60 min at 4°C in 200 mM KCl, 1 mM MgCl 2 , 200 M CaCl 2 , 20 mM HEPES, pH 7.4, 0.55% (w/v) Triton X-100, and 20% (v/v) glycerol. After centrifugation at 50,000 ϫ g for 30 min, the supernatant collected from the solubilisate was applied to a calmodulin-agarose column (10 ϫ 1.5 cm) pre-equilibrated with the solubilization buffer. The column was washed rapidly with 500 ml (about 50 column volumes) of washing buffer (200 mM KCl, 1 mM MgCl 2 , 200 M CaCl 2 , 20 mM HEPES, pH 7.4, 0.1% (w/v) Triton X-100, and 20% (v/v) glycerol) until no protein was detected in the wash. The protein was eluted from the column by a gradient of increasing concentrations of EDTA from 2 to 5 mM in the elution buffer (200 mM KCl, 20 mM HEPES, pH 7.4, 0.1% (w/v) Triton X-100, and 20% (v/v) glycerol). The fractions were collected and stored at Ϫ80°C. For further purification, the fractions containing p28 were pooled, dialyzed against three changes of 2 liters of elution buffer without KCl and EDTA to reduce the salt concentration, and then loaded onto a DEAE-cellulose column (5 ϫ 1.5 cm) pre-equilibrated with equilibration buffer (10 mM KCl, 50 mM Tris-HCl, pH 7.4, 20% (v/v) glycerol). The column was washed with 100 ml of equilibration buffer. The protein was eluted with a buffer of 1 M KCl, 50 mM Tris-HCl, pH 7.4, 20% (w/v) glycerol, and 0.05% (w/v) Triton X-100. The fractions were collected and analyzed by 6 -14% SDS-PAGE.
Internal Peptide Sequencing-Protein samples were analyzed by SDS-polyacrylamide 6 -14% slab gel electrophoresis under alkaline conditions (pH 8.8) as described by Laemmli (28). Protein enzyme digestion, peptide sequencing, and protein identification were carried out by Drs. K. Mitchelhill and B. E. Kemp at the Jack Holt Protein Structure Laboratory, St. Vincent's Institute of Medical Research (VIC, Australia). Briefly, p28 protein bands were excised from the gels and digested in 100 mM Tris-HCl, pH 8.0, 10% CH 3 CN containing 2 g of modified sequencing grade trypsin (Boehringer Mannheim). The tryptic peptides were separated by chromatography in a linear gradient from 0.01 to 0.085% trifluoroacetic acid in 80% CH 3 CN using a Hewlett Packard 1090 HPLC and sequenced on a Hewlett Packard G1000 protein sequencer using version 3.0 ethyl acetate chemistry.
Protein Amino Acid Composition Analysis-Protein samples were subjected to 6 -14% SDS-PAGE and transferred to PVDF membrane in transfer buffer (10 mM CAPS, pH 11.0, 10% methanol) at a constant current of 250 mA for 3 h. Amino acid composition analysis was performed by Jun X. Yan at the Macquarie University Center for Analytical Biotechnology (NSW, Australia). The PVDF membrane-bound proteins were hydrolyzed in 6 M HCl in a Pierce digestion tube at 110°C for 22 h. Amino acids were recovered using a single step extraction in 60% (v/v) CH 3 CN and 0.02% (v/v) trifluoroacetic acid. Fmoc (N-(9-fluorenyl)methoxycarbonyl) derivatization was performed prior to chromatography, and the derivatization mix was injected onto a reversed phase Hypersil C18 column. The relative abundance of each amino acid (Asx, Glx, Ser, His, Gly, Thr, Ala, Pro, Tyr, Arg, Val, Met, Ile, Leu, Phe, and Lys) was determined by measuring its concentration compared with an internal standard (125 pmol/amino acid; calibration kit, Sigma). However, during the hydrolysis process, tryptophan was destroyed by hydrochloric acid, and cysteine and lysine were oxidized. These three amino acids could not therefore be detected. Moreover, aspartic acid and asparagine as well as glutamic acid and glutamine were eluted as one peak instead of separate peaks. Thus, asparagine and glutamine were converted to their corresponding acids during hydrolysis so that Asp and Asn (Asx) and Glu and Gln (Glx) were counted together. Glycine values are inaccurate due to contamination from the gel running buffer after protein samples are analyzed by SDS-PAGE and electrotransferred to PVDF membrane. A special search program (Constellation 4) in SWISS two-dimensional PAGE amino acid composition analysis (available through the ExPasy server at the University of Geneva, accessed via World Wide Web) has therefore been developed that omits glycine and uses the remaining 15 amino acids for comparison of composition values. The score in this program is calculated by computing the euclidian distance between the search protein amino acid composition and the amino acid composition of all proteins in the SWISS-PROT data base.
Guanine Nucleotide Binding-p28 GTP-binding activities were detected by two assays: (i) Photoaffinity labeling of p28 with [␣-32 P]GTP was carried out using the method of Nakaoka et al. (29). The pure protein (about 1 g) was incubated with 5 Ci of [␣-32 P]GTP (3000 Ci/mmol) in buffer containing 200 mM KCl, 20 mM HEPES, pH 7.4, 2.5 mM EDTA, 0.1% (w/v) Triton X-100, 20% (v/v) glycerol, 100 M 5Јadenylylimidodiphosphate, 5 mM MgCl 2 , placed in an ice bath, and irradiated with a UV lamp (Mineralight model UVGL-58, 254 nm) at a distance of 8 cm for 30 min. When the specific binding of [␣-32 P]GTP to p28 was studied, the sample was preincubated with competing substrate GTP (100 M) at room temperature for 10 min. After irradiation, the samples were precipitated with 10% trichloroacetic acid and subjected to 6 -14% SDS-PAGE and then transferred to PVDF membrane. The PVDF membranes were air dried and exposed to a Storage Phosphor Screen for 72 h. The [␣-32 P]GTP-binding proteins were detected by autoradiography with a Phosphor Imager (Molecular Dynamics). (ii) [␣-32 P]GTP binding was followed after electroblotting onto PVDF membranes. For the detection of [␣-32 P]GTP-binding proteins on PVDF membranes, the modified procedure of Bhullar and Haslam (30) was followed. The PVDF membranes were first washed with 50 mM Tris-HCl, pH 7.4, containing 2 mM MgCl 2 and 0.3% Tween 20, and then incubated for 30 min at room temperature in the same buffer containing 1 Ci/ml [␣-32 P]GTP (3000 Ci/mmol). The PVDF membranes were washed extensively with the same buffer, air dried, and then exposed to a Storage Phosphor Screen (Molecular Dynamics) for 72 h. The [␣-32 P]GTP binding proteins were revealed by autoradiography on a Phosphor Imager (Molecular Dynamics).
Biotinylated Calmodulin Overlay-About 1-2 g of purified p28 and recombinant Ral-A were subjected to 12% SDS-PAGE and electrophoretically transferred to PVDF membranes. The PVDF membranes containing p28 or recombinant Ral-A were incubated for 3 h with blocking buffer (5% (w/v) powdered skim milk, 50 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 0.2% (w/v) Tween 20) with gentle shaking at room temperature and washed three times with the same buffer. The membranes were then incubated with blocking buffer containing 80 ng/ml of biotinylated calmodulin and 1 mM CaCl 2 for 2 h with continuous shaking at room temperature. In the control, 1 mM CaCl 2 was replaced by 5 mM EGTA in the above steps. After three further washes with blocking buffer for 5 min each, the membranes were incubated for 2 h with 1:1000 diluted avidin-alkaline phosphatase (Sigma) in blocking buffer at room temperature and then washed three times with blocking buffer and finally with substrate buffer (100 mM Tris-HCl, pH 9.5, 100 mM NaCl, and 50 mM MgCl 2 ). The color development was achieved by incubating the membranes in a reaction mixture containing 0.01% nitro blue tetrazolium and 0.05% 5-bromo-4-chloro-3-indolyl phosphate in 10 ml of substrate buffer. The reaction was allowed to proceed at room temperature until the color change was complete. The PVDF membrane-bound p28 and recombinant Ral-A were also visualized by staining with 0.1% (w/v) Amido Black, 40% (v/v) methanol, and 10% (v/v) acetic acid and destaining with 40% (v/v) methanol and 10% (v/v) acetic acid.

Identification of a Putative Calmodulin-binding Domain in Ral-A-
Computer analysis of Ral-A sequence was carried out by Dr. Jose Martin-Nieto (Instituto de Investigaciones Biomedicas, Madrid, Spain). PLOT.A/GGR and PLOT.A/GOR computer programs assign the predicted secondary structure for each residue in a protein sequence and provide approximate boundaries for the helical region. PLOT.A/HEL program draws a putative helical stretch by projecting the positions of the amino acids onto a plane perpendicular to the helix axis.
To quantitate the information presented on the helical wheel projection, the mean hydrophobic moment for this sequence was calculated according to the method of Erickson-Viitanen and DeGrado (31) using the hydrophobicity scales of Eisenberg et al. (32).
Hb i is the hydrophobicity of the ith amino acid, and n is the length of the sequence of interest. The term 100°corresponds to the angle at which successive side chains emerge from an ␣-helix when viewed down the axis.
Calmodulin Overlay with [ 14 C]P18 -A 18-amino acid peptide (P18) with the sequence SKEKNGKKKRKSLAKRIR corresponding to a putative calmodulin-binding domain in Ral-A was synthesized by Chiron Mimotopes Pty. Ltd. P18 was then radioactively labeled with [ 14 C]formaldehyde (DuPont NEN) by the method of Rice and Means (33). The specific radioactivity of 14 C-labeled P18 was estimated to be 7.0 ϫ 10 6 cpm/mg protein from a plot of scintillation counts/min versus peptide concentration. 5 g of calmodulin was subjected to 6 -16% SDS-PAGE and transferred to PVDF membranes. One blot was incubated with 100 l of [ 14 C]P18 and 1 mM CaCl 2 in 10 ml of washing buffer (50 mM Tris-HCl, pH 7.4, and 0.3% (w/v) Tween 20); the other was incubated with 100 l of [ 14 C]P18 and 1 mM EDTA in 10 ml of washing buffer separately for 30 min at room temperature. The blots were soaked in 20 ml of washing buffer and washed five times. The PVDF membranebound calmodulin was visualized by staining with 0.1% (w/v) Amido Black, 40% (v/v) methanol, and 10% (v/v) acetic acid and destaining with 40% (v/v) methanol and 10% (v/v) acetic acid. Blots to be counted were sliced transversely into 2-mm strips. The slices were soaked in 5 ml of Econofluor overnight and then counted in a liquid scintillation analyzer (1900A, TRI-CARB, Packard Pty. Ltd.).
Competitive Binding Assay of p28 and P18 with a Calmodulin-Agarose Resin-Constant amounts of p28 and calmodulin-agarose resin were used in all experimental procedures. Samples of p28 (about 1.5 g) were incubated with 0.5 ml of calmodulin-agarose resin in Eppendorf tubes in a buffer containing 250 mM KCl, 50 mM Tris-HCl, pH 7.4, 20% (w/v) glycerol, 0.05% (w/v) Triton X-100, 1 mM MgCl 2 , and 1 mM CaCl 2 in the presence and in the absence of 10 g of P18 or 10 g of calmodulin at 4°C overnight. The samples were then centrifuged at 10,000 ϫ g in a microcentrifuge for 10 s, supernatants were removed, and proteins (referred to as supernatant) were subjected to SDS-PAGE analysis. The calmodulin-agarose resin samples were washed with 1.5 ml of washing buffer (200 mM KCl, 50 mM Tris-HCl, pH 7.4, 20% (w/v) glycerol, and 0.05% (w/v) Triton X-100) at 4°C for 10 min. The washing solutions were discarded after centrifugation at 10,000 ϫ g in a microcentrifuge for 10 s. This washing step was repeated five times. The calmodulinagarose resin samples were then incubated with 1 ml of elution buffer (200 mM KCl, 50 mM Tris-HCl, pH 7.4, 20% (w/v) glycerol, 0.05% (w/v) Triton X-100, and 5 mM EDTA, pH 7.4) at 4°C for 5 h. After centrifugation at 10,000 ϫ g in a microcentrifuge for 10 s, the supernatants were removed, and proteins (referred to as eluent) were subjected to SDS-PAGE analysis as described above.
Circular Dichroism Spectroscopy Studies of Calmodulin, P18, and Calmodulin-P18 Complex-CD spectroscopy was performed on a Jasco J-720 Spectropolarimeter (Japan) connected to a Neslab RTE-111 waterbath, using a cell path length of 0.1 cm. Spectra of observed ellipticity, (mdeg) versus wavelength were collected between 190 and 260 nm at 15°C. A base-line spectrum was always collected, using the same parameters, with buffer in the sample cell. Each spectrum was averaged over 8 scans, and final spectra were corrected, from the base-line spectrum, for specious signals generated by the buffer. CD spectra were measured in 5 mM sodium borate buffer, pH 8.0, and 0.5 mM CaCl 2 with 6 M calmodulin, or 6 M P18, or calmodulin-P18 complex at 1:1 ratio, respectively. Calmodulin concentration was determined spectroscopically using an extinction coefficient of 3060 M Ϫ1 cm Ϫ1 (31). For quantitative analysis, data were expressed as mean residue ellipticity, [] MRW (deg dmol Ϫ1 cm 2 ).

Purification of p28 from the Solubilized Human Erythrocyte
Membrane-Ca 2ϩ -ATPase is the major calmodulin binding protein in erythrocyte membrane. When a Triton X-100 extract of human erythrocyte membrane was applied to a calmodulinagarose column, a polypeptide of molecular mass 28 kDa on SDS-PAGE was observed to be co-eluted with Ca 2ϩ -ATPase following gradient elution with a buffer containing 2-5 mM EDTA. Ca 2ϩ -ATPase was eluted in the first one or two fractions, whereas p28 appears in the later fractions (see Fig. 1). For further purification of p28, a DEAE-cellulose column was used to limit possible contamination that is not detectable in SDS-PAGE. After passing through a DEAE-cellulose column, p28 appeared to be a clear single band in SDS-PAGE (results not shown). The yield of p28 is about 50 -100 g from 0.5 to 1.0 g of membrane protein.
Internal Peptide Sequencing-Eight tryptic peptides selected from the protein sample (p28) were separated by reversed phase HPLC and sequenced. One of them gave a sequence of VKEDENVPFLLVGN at an initial yield of 13.5 pmol and a repetitive yield of 84.4%. The other gave a sequence of AEQWN at an initial yield of 2.7 pmol and a repetitive yield of 79.7%. A search of the SWISS-PROT data base identified the peptides as having originated from the 23.6-kDa Ras-related protein Ral-A (SWISS-PROT accession number P11233). The two peptides came from Ral-A sequence positions 114 -127 and 146 -150, respectively, and were consistent with expected tryptic digestion products. Based on the absolute homology of internal peptide sequence, p28 was identified as Ral-A.
Protein Amino Acid Composition Analysis-The p28 amino acid composition was used to search and obtain a score in the SWISS-PROT data base. The protein at the top of the closest SWISS-PROT entries in human species as well as the closest SWISS-PROT entries in cross-species was the Ras-related, GTP-binding protein, Ral-A. Ral-A was distinguished from other protein candidates by its significant differences in score. Moreover, the protein molecular mass (which is estimated to be 28 kDa from SDS-PAGE in Fig. 1) and pI value (which is calculated to be 7.0 based on its amino acid composition) provided strong additional information to rule out incorrect candidates from the list. With these guidelines, it was concluded that Ral-A was the correct candidate for p28, because Ral-A appeared to be the highest match in both single and crossspecies data base searches. Fig. 2 shows the amino acid composition comparison of p28 with Ral-A obtained from SWISS-PROT (accession number P11233). Guanine Nucleotide Binding-p28 GTP-binding activities were detected by photoaffinity labeling and gel overlay with [␣-32 P]GTP (see Fig. 3). Panel A in Fig. 3 shows that p28 protein was photoaffinity labeled with [␣-32 P]GTP. The specificity of this labeling was evident from the ability of unlabeled 0.1 mM GTP to prevent the covalent labeling. The p28 protein band on the autoradiogram, as well as on the PVDF membrane (not shown), appeared broader than normal on SDS-PAGE, but this could be due to p28 oxidation under irradiation. Increasing covalent labeling of p28 with [␣-32 P]GTP corresponded to increasing irradiation time. Panel B in Fig. 3 shows that p28 can be specifically labeled by [␣-32 P]GTP. The specificity of labeling was shown from the finding that the presence of competing substrate 0.1 mM GTP completely blocked p28 labeling with [␣-32 P]GTP, whereas 0.1 mM ATP did not abolish GTP-binding. When overlaid, the band on the autoradiogram fell exactly on the Amido Black-stained band detected on the PVDF membrane. Low molecular mass GTP-binding proteins are known to renature after being electrophoretically transferred to PVDF membranes or nitrocellulose blots (unlike heterotrimeric G␣ protein) and exhibit specific binding to guanine nucleotides following incubation of the blots with [␣-32 P]GTP (30,36). Our results suggest that a renatured form of p28 on PVDF membrane retains the ability to specifically bind to GTP. It was therefore concluded that p28 is a low molecular mass GTPbinding protein, consistent with its identification as Ral-A or a highly related Ral-A protein.
Biotinylated Calmodulin Overlay-Using biotin as a reporter ligand to detect biotinylated calmodulin bound to renatured calmodulin-binding proteins (34), biotinylated calmodulin overlay showed binding of calmodulin to purified p28 (Fig. 4, lane  1). No binding of biotinylated calmodulin was observed when 5 mM EGTA replaced 1 mM CaCl 2 (Fig. 4, lane 2), indicating that the interaction of biotinylated calmodulin with p28 is Ca 2ϩ -dependent. Recombinant Ral-A, generously supplied by Drs. Hiroshi Koide and Yoshito Kaziro and expressed in Escherichia coli by using the plasmid pGEX-2T, which contains the rat Ral-A cDNA, was also found to bind biotinylated calmodulin in the presence of 1 mM CaCl 2 (Fig. 4, lane 5). Recombinant Ral-A appears as a 25-kDa protein on 12% SDS-PAGE (Fig. 4, lane 6).
Identification of a Putative Calmodulin-binding Domain in Ral-A-A putative calmodulin-binding domain with the se-quence of SKEKNGKKKRKSLAKRIR arranged in an amphiphilic ␣ helix was identified in the C-terminal region of Ral-A at sequence position 183-200. Fig. 5 shows an axial helical wheel projection generated from the PLOT.A/HEL program based on the sequence of SKEKNGKKKRKSLAKRIR, corresponding to the putative calmodulin-binding domain of Ral-A. In this sequence there are four hydrophobic residues (Leu, Gly, Ile, and Ala) on one side of the wheel and nine strongly basic residues (Lys, Lys, Arg, and Lys, Lys, Lys, Lys, Arg, and Lys) segregated on the opposite side of the wheel. Although the contribution of the hydrophobic side of the amphiphilic helix may not be very high because of the interspersed charges, the overall features of two sets of adjacent positively charged residues, together with the hydrophobic residues on the other side, display the characteristic features of calmodulin-binding domains: a hydrophobic/basic composition with the propensity to form an amphiphilic helix.
The mean hydrophobic moment (͗ H ͘) and the mean hydrophobicity (͗Hb͘) of calmodulin-binding domains from various target enzymes and proteins were calculated according to Erickson-Viitanen and DaGrado (31). Fig. 6 shows the plot of the magnitude of the mean hydrophobic moment, ͗ H ͘, versus mean hydrophobicity, ͗Hb͘, which gives simultaneous visualization of the hydrophobicity and amphiphilicity of a helix. Partitions define regions of ␣-helices from globular, surface, and transmembrane proteins according to Eisenberg et al. (32). The ͗ H ͘ and ͗Hb͘ values of P18 are calculated to be 0.31 and Ϫ0.91 kcal/mol, respectively. As shown in Fig. 6, P18, as well as most calmodulin-binding peptides, tend to fall into the category of "globular helices" with the range of ͗Hb͘ from Ϫ1.0 to 0.5 kcal/mol and the range of ͗ H ͘ from 0.1 to 0.7.
Calmodulin Blot Overlay with 14 C-labeled P18 -Blots of calmodulin overlaid with [ 14 C]P18 in the presence and the absence of Ca 2ϩ were stained, sliced, and counted. Fig. 7 shows the correspondence between the distribution of radioactivity and the stained bands. A radioactive peak was observed corresponding to the calmodulin band (17 kDa) in the presence of 1 mM CaCl 2 , whereas no distinct radioactive peak was found in the presence of 1 mM EDTA. The results suggest that renatured calmodulin on PVDF membrane still retains its functional peptide-binding sites that can bind to [ 14 C]P18 in a Ca 2ϩ -dependent manner. A difference in basal scintillation counts between the two blots may be due to the different nonspecific radioactive binding induced by the different incubation media in the assay (one with 1 mM CaCl 2 and the other with 1 mM EDTA).
Competitive Binding Assay of p28 and P18 with a Calmodulin-Agarose Resin-The effect of P18 on binding of p28 to the calmodulin-agarose resin was determined. After p28 was incubated with a calmodulin-agarose resin in the presence of 1 mM CaCl 2 , the supernatant was removed, and the calmodulin-agarose resin sample was washed and eluted with an elution buffer containing 5 mM EDTA. Both supernatant and eluent were subjected to SDS-PAGE. As shown in Fig. 8, where lanes 2 and 3 represent control samples, p28 is observed in the eluent eluted with EDTA, whereas there is no p28 detected in the supernatant. The results indicate that p28 binds to calmodulinagarose in a Ca 2ϩ -dependent manner and can be eluted by chelation of Ca 2ϩ with EDTA. In the same experiments done in the presence of P18 (lane 4 and 5) and calmodulin (lane 6 and 7), p28 is observed both in the eluent and in the supernatant. This suggests that the binding of p28 with Ca 2ϩ /calmodulin is reduced in the presence of P18, similarly to that found in the presence of Ca 2ϩ /calmodulin itself. Thus, P18 can compete with p28 for binding to calmodulin in the presence of Ca 2ϩ , indicating that p28 and P18 share at least in part the same binding sites for calmodulin. However, under the conditions of the experiment, neither P18 nor calmodulin can completely prevent p28 binding to the calmodulin-agarose resin.
CD Spectroscopy Study of Calmodulin and Calmodulin-P18 Complex-To further characterize the calmodulin-binding domain in Ral-A, the binding of P18 to calmodulin was studied by CD spectroscopy. As shown in Fig. 9, P18 by itself has no obvious secondary structure but exists as a random coil in aqueous solution. However, calmodulin shows a CD spectrum typical of helical proteins, with minimum ellipticity at 222 and 208 nm (35). Upon the formation of calmodulin-P18 complex, a further decrease in negative mean residue ellipticity at 222 and 208 nm is observed. This suggests that addition of P18 to calmodulin at 1:1 ratio reduces to a significant extent the ␣-helical content in the complex. The CD spectroscopy results provide direct evidence for an interaction between P18 and calmodulin and show that this interaction results in a less ␣-helical content in the calmodulin-P18 complex (Fig. 9).

DISCUSSION
In this paper we have shown that a protein (p28) purified from Triton X-100 extracts of human erythrocyte plasma membrane by calmodulin affinity chromatography is highly related, if not identical, to Ral-A, a Ras-related GTP-binding protein, on the basis of internal peptide sequencing and amino acid composition analysis. This assignment was also confirmed by the characteristic of p28 to bind to [ 32 P]GTP specifically and from FIG. 4. Detection of calmodulin-binding proteins by biotinylated calmodulin overlay. About 1-2 g of purified p28 and recombinant Ral-A were subjected to SDS-PAGE and transferred to a PVDF membrane. The PVDF membranes containing p28 or recombinant Ral-A were blocked with blocking buffer and then incubated with 80 ng/ml of biotinylated calmodulin and 1 mM CaCl 2 . In the control, 1 mM CaCl 2 was replaced by 5 mM EGTA in the blocking buffer in the above steps. The membranes were then incubated with 1:1000 diluted avidinalkaline phosphatase (Sigma) in blocking buffer. The color development was achieved by incubating the membranes in a reaction mixture containing nitro blue tetrazolium and 0.  (48). The mean hydrophobic moment (͗ H ͘) and the mean hydrophobicity (͗Hb͘) of calmodulin-binding domains were calculated according to Erickson-Viitanen and DeGrado (31). Partitions define regions of ␣-helices from globular, surface, and transmembrane proteins according to Eisenberg et al. (32). ࡗ shows the hydrophobic moment plot of various calmodulinbinding domains. E represents the hydrophobic moment plot of the putative calmodulin-binding domain in Ral-A (p28) with the sequence SKEKNGKKKRKSLAKRIR.
its low intrinsic GTPase activity (results not shown). Our previous work has demonstrated that this protein is an oligomeric protein with a native molecular mass of 660 kDa, composed of a single subunit of 28 kDa on SDS-PAGE. It is firmly associated with the membrane phospholipid bilayer rather than with the cytoskeletal component (26).
Although the Ral-A protein purified from human erythrocyte membrane has an apparent molecular mass of approximately 28,000 Da on SDS-PAGE, the full length human Ral-A cDNA encodes a protein with a predicted molecular mass of 23,567 Da. The discrepancy of molecular mass may be due to post-translational modification. In agreement with our finding, the Ral-A gene products isolated from human platelet membrane have molecular mass of 28 kDa (36), whereas the human Ral-A gene protein expressed in E. coli is a 29-kDa protein (37). We have interpreted this as evidence that Ral-A binds to plasma membrane through groups associated with its post-translational modification. Ral proteins undergo sequential posttranslational modifications that occur at a CAAL motif in the C-terminal region of the molecules (38). Ral-A translation products in reticulocyte lysates have been found to be modified by 20-carbon isoprenyl groups (39). Very recently, post-translational modification of Ral has been shown to enhance the activities of RalGDS, indicating a functional role of this modification for transmitting its signal effectively (40).
A striking and novel finding in our study is that Ral-A has been identified as a calmodulin-binding protein. The interaction with calmodulin was first detected on the basis of the ability of Ral-A to bind to a calmodulin affinity column in a Ca 2ϩ -dependent manner and to be released upon elution by chelation of Ca 2ϩ with EDTA. The indication of Ral-A binding to calmodulin was strongly supported by further evidence that calmodulin could block Ral-A phosphorylation by cAMP-dependent protein kinase, cGMP-dependent protein kinase, and Ca 2ϩ /phospholipid-dependent protein kinase. 2 Moreover, biotinylated calmodulin overlay experiments have shown that biotinylated calmodulin binds to renatured Ral-A in a Ca 2ϩ -dependent manner, confirming our finding that Ral-A is a calmodulin-binding protein. After these results were completed, it was reported that two other Ras-related GTP-binding protein groups, Kir/Gem and Rin, contain calmodulin-binding domains in their extended polybasic C-terminal regions (23,24). However, unlike Ral-A, Kir/Gem and Rin lack a typical CAAX isoprenylation motif in their C termini.
Based on inspection of Ral-A sequence with respect to the presence of a basic/hydrophobic composition with the propen- Samples of p28 (about 1.5 g) were incubated with 0.5 ml of calmodulinagarose resin in the buffer containing 1 mM CaCl 2 in the presence and in the absence of 10 g of P18 or 10 g of calmodulin (CaM) at 4°C overnight. The samples were then centrifuged at 10,000 ϫ g in a microcentrifuge, supernatants were removed, and proteins (referred to as supernatant) were prepared for SDS-PAGE analysis. The calmodulin-agarose resins were washed five times and then incubated with 1 ml of elution buffer containing 5 mM EDTA at 4°C for 5 h. After centrifugation at 10,000 ϫ g in a microcentrifuge, supernatants were removed, and proteins (referred to as eluent) were subjected to SDS-PAGE analysis. Lanes 1 and 8 show molecular mass standards (M.W.). Lanes 2 and 3 show samples obtained from the supernatant and eluent, respectively, after p28 was incubated with calmodulin-agarose resin. Lanes 4 and 5 show the supernatant and eluent, respectively, obtained after p28 was incubated with calmodulin-agarose resin in the presence of P18. Lanes 6 and 7 show the supernatant and eluent, respectively, obtained after p28 was incubated with calmodulin-agarose resin in the presence of calmodulin.
FIG. 9. CD spectra of P18, calmodulin, and calmodulin/P18 mixture at 1:1 ratio. CD spectra were measured in 5 mM sodium borate buffer, pH 8.0, and 0.5 mM CaCl 2 with 6 M calmodulin, or 6 M P18, or calmodulin/P18 mixture at 1:1 ratio, respectively, as described under "Experimental Procedures." For quantitative analysis, spectra of observed ellipticity, (mdeg) versus wavelength were expressed as spectra of calculated mean residue ellipticity, [] MRW (deg dmol Ϫ1 cm 2 ) versus wavelength. q shows the CD spectra of 6 M calmodulin (CaM); E shows the CD spectra of calmodulin/P18 mixture at 1:1 ratio; ࡗ shows the CD spectra of 6 M P18. sity to form an amphiphilic helix, a putative calmodulin-binding domain (P18) with the sequence SKEKNGKKKRKSL-AKRIR was identified within the C-terminal region of Ral-A. The ␣-helical structure of this domain shows the typical basic, amphiphilic features that are consistent with the criterion for the calmodulin-binding peptides established by DeGrado and co-workers (31,41). This amphiphilic helix model has now been demonstrated to underlie the calmodulin-binding properties common to a variety of peptides by NMR and CD spectroscopy studies (42)(43)(44)(45). The most unusual feature of the calmodulinbinding domain identified in Ral-A is that it is rather more hydrophilic than hydrophobic, because it contains 9 of 18 hydrophilic amino acids (50%) but only four hydrophobic residues (22%). By contrast, the mastoparans, high affinity calmodulinbinding peptides, have 21-28% hydrophilic amino acids and 36 -50% hydrophobic residues (46). Although it has been established that the interaction of calmodulin with its target proteins is predominantly hydrophobic, it is complemented by acidic side chains from the calmodulin EF-hands interacting with basic residues of the target proteins (47). For instance, the high affinity calmodulin-binding site (K d ϭ 20 nM) of the Na ϩ /H ϩ exchange isoform 1 with the segments of RNNLQK-TRQRIRSYNRHT contains 7 of 18 hydrophilic amino acids (39%) but only 2 hydrophobic residues (11%) (48). The replacement of 4 positively charged residues with negative or neutral ones is found to inhibit the binding of this region to calmodulin. The characteristic of the calmodulin-binding domain in Ral-A is reminiscent of domains such as those in Na ϩ /H ϩ exchange isoform 1. This indicates that apart from the hydrophobic interaction, the binding of calmodulin with target proteins involves strong electrostatic interaction, which arises from the charged residues of the target peptide interacting with acidic side chains from calmodulin EF-hand motifs.
The calmodulin-binding domain in Ral-A is somewhat hydrophilic, with the mean hydrophobicity ͗Hb͘ ϭ Ϫ1.08 kcal/mol and hydrophobic moment ͗ H ͘ ϭ 0.31, suggesting that it belongs to the globular helices group. According to Erickson-Viitanen and DeGrado (31), several calmodulin-binding peptides with dissociation constants of less than 5 nM have been reported to fall in the most amphiphilic "surface seeking" region, with a range of ͗Hb͘ values of 0 to 0.4 kcal/mol and values of ͗ H ͘ of 0.5 or greater. However, this criterion is only suitable for those calmodulin-binding peptides with the sequence of 12 residues and an overall net positive charge above ϩ4. Considering the lack of amino acid sequence homology and the substantial difference in length in calmodulin-binding peptides, ͗ H ͘ and ͗Hb͘ for various calmodulin-binding peptides with available sequences were calculated. According to our calculation, the hydrophobic moment plot shows that P18, as well as most calmodulin-binding domains, tend to fall into the globular helices region with the range of ͗Hb͘ from Ϫ1.0 to 0.5 kcal/mol and the range of ͗ H ͘ from 0.1 to 0.7. Despite its usefulness of visualizing the hydrophobicity and amphiphilicity of a calmodulin-binding domain, the hydrophobic moment plot cannot be used for prediction of calmodulin-binding sequences because it ignores the contribution of basicity to calmodulin binding, and there is no predictable relation of the calculated hydrophobic moment to the affinity constants of different calmodulin-binding peptides (31).
Further characterization of this proposed calmodulin-binding domain in Ral-A was undertaken by using a synthetic peptide (P18) corresponding to the sequence H-SKEKNGK-KKRKSLAKRIR-OH and three independent methods: calmodulin overlay with 14 C-labeled P18, a Ral-A competitive binding study with calmodulin in the absence and in the presence of P18, and CD spectroscopy study of calmodulin-P18 complex.
We have shown that P18 interacts with calmodulin in a Ca 2ϩdependent manner, thereby further supporting its role as a calmodulin-binding domain. It has been reported for numerous calmodulin-binding peptides that a significant increase in negative mean residue ellipticity at 208 and 222 nm is observed, indicating the formation of additional ␣-helical structure upon calmodulin-peptide complex formation (31,44,45,49). Because in the presence of a target peptide the long central helix, which connects two globular domains in calmodulin, is disrupted and unravelled to accommodate peptide binding, Ca 2ϩ /calmodulin generally does not gain secondary structure (42, 43, 50,). Thus, the increased ␣-helical content is exclusively attributed to binding of the peptide to calmodulin. In contrast, CD studies of the calmodulin-P18 complex show a decrease in negative mean residue ellipticity at 208 and 222 nm, indicating that the interaction of P18 with Ca 2ϩ /calmodulin results in less helical content in calmodulin-P18 complex. This behavior closely resembles that of Phk13, one of the high affinity (K d ϭ 5 nM) calmodulin-binding domains in the ␥ subunit of glycogen phosphorylase kinase (52). The addition of Phk13 to calmodulin also results in a decrease in negative mean residue ellipticity at 222 and 208 nm. The structure of calmodulin-Phk13 complex has also been studied by tryptic digestion and radiationless energy transfer measurements and a model of calmodulin interacting with Phk13 has been proposed in which calmodulin has an extended conformation resembling its crystallographic structure, whereas Phk13 exists in a nonhelical conformation, making contact with both the N-and C-terminal lobes and lying roughly parallel to the long axis of calmodulin (52). Whether the interaction of P18 with calmodulin will match this model needs further structure studies. It is very difficult to give a precise interpretation based on CD data alone because the diagnostic ␣-helical bands for the protein and the peptide overlap.
Calmodulin is a highly conserved, small (148 amino acid), acidic Ca 2ϩ -binding protein. It is considered to be the major regulator of Ca 2ϩ -dependent signaling pathways in eukaryotic cells, mediating a wide variety of physiological processes, including glycogen metabolism, secretion, muscle contraction, and cell division (53,54). The regulatory effects induced by calmodulin are mediated by numerous calmodulin-binding proteins. Calmodulin-binding proteins therefore constitute a major group of signal transducing proteins. The studies presented here display new experimental evidence that Ral-A is a calmodulin-binding protein. This finding has provided the new possibility of modulation of Ral-A by Ca 2ϩ or a functional role of Ral-A protein in Ca 2ϩ signaling pathways. Although it has been firmly established that heterotrimeric G proteins are involved in Ca 2ϩ release through the regulation of phospholipase C, adenylate cyclase and K ϩ channels (55), there is an increasing amount of information on the involvement of low molecular mass GTP-binding proteins in intracellular Ca 2ϩ mobilization. Rap1 has been found to regulate platelet plasma membrane Ca 2ϩ transport (56). As for Ca 2ϩ release from the intracellular store, Chong et al. (57) have shown that activation of Rho pathways stimulate production of a phospholipase C substrate, thereby enabling Ca 2ϩ release triggered by platelet-derived growth factor. In addition, Rho has been shown to play a role in helping to decrease increased Ca 2ϩ levels without affecting the machinery for Ca 2ϩ release from inositol 1,4,5-triphosphate-dependent Ca 2ϩ stores (58). A Rac-dependent Ca 2ϩ influx pathway has also been reported (51). This accumulating evidence, together with our finding that Ral-A interacts with Ca 2ϩ /calmodulin, shows that Ras-related GTP-binding proteins may be modulated by Ca 2ϩ and may have a potentially significant function in Ca 2ϩ signaling pathways. It also raises the possi-bility of cross-talk between signal transduction pathways mediated by Ca 2ϩ /calmodulin and Ras proteins.