The Calmodulin-binding Site in a-Fodrin Is Near the Calcium-dependent Protease-I Cleavage Site*

of of fodrin 0.1 mM CaCI2, 1 mM 2-me, pH 7.5, to insure a constant free calcium concentration. Digests were terminated at various times by adding diisopropyl fluorophosphate (500 mM in isopropyl alcohol) final followed by either the addition of solubilizing buffer for SDS-PAGE or by lyophilization for isoelectric focusing/SDS-PAGE. In separate experiments, diisopropyl rophosphate from Sigma) was found to these was used protease

Several signals regulating the above processes probably act directly on fodrin. The regulation of glutamate sequestration in postsynaptic membranes, a postulated event in the development of long-term synaptic potentiation, requires the proteolysis of fodrin by calcium-dependent proteases , Siman et al., 1984, 1985. A similar proteolysis of fodrin also occurs with platelet activation (Fox et al., 1987), and the primary site of CDP-I' proteolysis in. uitro is near the center of the CY subunit . A second regulatory mechanism of fodrin involves the calcium-dependent binding of calmodulin, also to a site near the center of the a subunit . Calmodulin enhances the susceptibility of fodrin to CDP-I proteolysis (Seubert et al., 1987), and both this cleavage and calmodulin reduce the ability of fodrin to cross-link actin filaments in Despite the probable importance of these activities, and the clear association between the CDP-I susceptibility of fodrin and calmodulin binding, the precise relationship between the sites of these events in fodrin remains uncertain.
In the present report, intermediate-sized peptides of a fodrin have been prepared either by proteolytic digestion or by chemical cleavage at cysteine residues using NTCB. The position of these fragments within the parent protein has been determined by two-dimensional peptide mapping and their calmodulin binding activity measured by gel overlay using 1251-calmodulin and by their ability to bind to a calmodulin-agarose affinity column under nondenaturing conditions. Gas-phase amino acid sequencing of calmodulin-binding peptides derived from CDP-I and chymotryptic digestion identified a sequence of 24 amino acids in which the calmodulin-binding site must reside. This sequence, beginning 16 residues distal from the site of CDP-I cleavage, bridges the eleventh and twelfth repetitive units of a-fodrin.

MATERIALS AND METHODS
Protein Preparation-Human fodrin was prepared as described (Harris et al., 1986) with the following modifications. Ion-exchange chromatography was performed at 4 "C on a 1 X 10-cm column of Accell-QMA (Millipore) using a 170-ml linear gradient of 25-750 mM NaCl in 20 mM Tris-HCI, 0.5 mM EDTA, 1 mM 2-me, pH 7.8. The fractions containing purified fodrin (monitored by A m and SDS-PAGE) were dialyzed overnight a t 4 "C versus 20 mM Tris-HCI, 25 mM NaC1, 0.5 mM EDTA, 1 mM 2-me, pH 8.0 (buffer A), and concentrated using a 1 X 2-cm column of DEAE-cellulose (DE52, Whatman) eluted with 20 mM Tris-HCI, 1 M NaCl, 0.5 mM EDTA, 1 mM 2-me, pH 7.5. The fodrin from the column (typically at 1-2 mg/ ml) was dialyzed against buffer A and stored at 0 "C. Bovine fodrin, prepared from approximately 400 g of frozen brains, was prepared as described for human fodrin (Harris et al., 1986), with the following modifications. The frozen brain was thawed and homogenized in a Waring blender for 30 s. After preparing the washed membranes, they were resuspended in approximately 2 liters of 0.1 mM EDTA, pH 9.0, and 3 M KC1 was added to achieve 0.6 M KC1 in a total volume of 3.6 liters. EGTA, dithiothreitol, and diisopropyl fluorophosphate were added from stock solutions to final concentrations of 5, 2, and 0.5 mM, respectively. The pH was adjusted to 9.0 with 3 M Tris, and the protein extracted at 37 "C for 30 min. The extract was cleared of membranes by centrifugation at 30,000 X g for 60 min, and precipitated by 50% saturated ammonium sulfate at 4 'C. The precipitated protein, recovered by centrifugation was resuspended in 20 mM Tris-HC1, 150 mM NaCI, 0.5 mM EDTA, 1 mM 2-me, pH 8.0 (TBS), and centrifuged at 30,000 X g for 30 min. The supernatant was made 10% in sucrose and centrifuged for 16 h at 4 "C at 110,000 X g in a Beckman SW 27 rotor. The floating lipid was removed, and the clear supernatant was fractionated on a 5 X 95-cm Sepharose CL-4B column equilibrated in TBS. Bovine fodrin was further purified and concentrated as described above for the human material.
Calmodulin was prepared from frozen and fresh bovine brain by the method of Burgess et al., 1980. Purified calmodulin was incubated in 10 mM diisopropyl fluorophosphate overnight at 4 'C, and dialyzed uersus several changes of 20 mM Tris-HC1, 25 mM NaCl, 0.1 mM CaCI2, 1 mM 2-me, pH 7.5, prior to use.
Protein Digestions-Fodrin was cleaved at cysteine residues with NTCB (Eshdat and Lemay, 1979). Fodrin at 1 mg/ml in 7.5 M guanidine HCI (or 8 M urea), 200 mM Tris-HCI, 1 mM EDTA, pH 8.0, was reacted with 513 mM NTCB for 1 h at room temperature. Cleavage was initiated after the pH was raised to 9.0 with 0.5 M Tris-NaOH, pH 12, followed by incubation at 37 "C for 16 h. The reaction was terminated by the addition of 2-me to 25 mM, and the material was dialyzed versus several changes of 10 mM Tris-HCI, 0.5 mM EDTA, 1 mM 2-me, pH 8.0, at 4 "C, and lyophilized. Proteolytic digests were done in buffer A, except as noted, at protein concentrations of 0.4-1.5 mg/ml at 0 or 23 "C. Trypsin (N-p-tosyl-l-phenylalanine chloromethyl ketone treated-trypsin, Worthington), a-chymotrypsin (Worthington), endoprotease Lys-C (Boehringer Mannheim) or endoprotease Glu-C (Boehringer Mannheim) at 1 mg/ml was added to the protein at the enzyme/substrate ratios indicated. CDP-I, prepared from bovine heart (Croall and DeMartino, 1984) was used at an enzyme/substrate ratio of 1:25 (mol/mol) at free calcium concentrations of 0.1 mM. Digestions with CDP-I were performed after overnight dialysis at 4 "C of the fodrin against 20 mM Tris-HCI, 25 mM NaCI, 0.1 mM CaCI2, 1 mM 2-me, pH 7.5, to insure a constant free calcium concentration. Digests were terminated at various times by adding diisopropyl fluorophosphate (500 mM in isopropyl alcohol) to a final concentration of 5-10 mM, followed by either the addition of solubilizing buffer for SDS-PAGE or by lyophilization for isoelectric focusing/SDS-PAGE. In separate experiments, diisopropyl fluorophosphate (obtained from Sigma) was found to inhibit CDP-I as well as EGTA in these experiments, and therefore it was used as the sole protease inhibitor.
Affinity Isolation of Calmodulin-binding Peptides-NTCB-generated fodrin peptides were dialyzed against 130 mM KCI, 20 mM NaCI, 10 mM HEPES, 1 mM CaC12, and 1 mM 2-me, pH 7.3, and loaded onto a 1.25-ml calmodulin affinity column (10 mg of calmodulin/ml of gel, Bio-Rad) equilibrated in the same buffer. After the nonadherent protein had eluted, the column was washed with 6 M urea in 95 mM KC1, 15 mM NaC1, 7.5 mM HEPES, 1 mM CaCI2, 1 mM 2me, pH 7.3, to remove nonspecifically bound proteins. Calciumdependent calmodulin-binding peptides were eluted from the column with the above urea containing buffer in which the calcium was replaced with 10 mM EGTA. The fractions containing the eluted proteins were lyophilized, desalted on a 0.6 X 22-cm Sephadex G-15 column in 50 mM ammonium bicarbonate buffer and lyophilized prior to two-dimensional isoelectric focusing/SDS-PAGE.
Two-dimensional Peptide Mapping-Two-dimensional cellulose Iz5I-peptide mapping of Coomassie Blue-stained peptides was performed using established procedures (Elder et al., 1977, Speicher et al., 1982. Briefly, the peptides were cut from polyacrylamide gels and labeled with "' 1 using chloramine T. After removal of the free lz5I, the gel slices were digested with a-chymotrypsin (Sigma, 50 pg in 50 mM ammonium bicarbonate) for 24 h at 37 "C. Peptides were recovered in the liquid phase and lyophilized. The peptides were dis-solved in electrophoresis buffer, spotted onto 20 X 20-cm cellulose plates and separated by high voltage electrophoresis in the horizontal dimension and by ascending liquid chromatography in the vertical dimension. Peptide maps were visualized by autoradiography at -70 "C using Kodak XRP film and fluorescent intensifying screens.
Amino Acid Sequencing-Digested protein was separated by twodimensional isoelectric focusing/SDS-PAGE and transferred to polyvinylidene difluoride membranes (Immobilon, Millipore) according to Matsudaira (1987). Peptides were visualized by Coomassie Blue staining. The peptides to be sequenced were cut from the membrane and stored at -20 "C. Polyvinylidene difluoride membranes were not precycled and were placed directly in the cartridge block of the protein sequenator. Peptides were sequenced on an Applied Biosystems model 470A Sequencer with on-line phenylthiohydantoin analysis (Applied Biosystems 120A high pressure liquid chromatography) using the standard program (03RPTH) without modification.
Gel Electrophoresis and Other Procedures-SDS-PAGE and twodimensional isoelectric focusing/SDS-PAGE were performed by the method of Laemmli (1970) and O'Farrell (1975), respectively, except that 3[ (3-chloramidopropyl)-dimethylammonio)]l-propane sulfonate was substituted for Triton X-100 in the isoelectric focusing dimension. All two-dimensional gels are presented with the basic region to the left. Proteins were visualized by Coomassie Blue staining. Protein concentrations were estimated by the method of Lowry et al., (1951). Calmodulin was labeled with Iz5I using immobilized lactoperoxidase and glucose oxidase (Enzymobeads, Bio-Rad) as previously described (Anderson and Morrow, 1987). 12sI-calmodulin overlays were performed as previously described (Carlin et al., 1983). Protein transfers to nitrocellulose or polyvinylidene difluoride membranes were performed according to Tobin et al., (1979).

Proteolytic Fragments of Fodrin That Bind Calmdulin-
Since the calmodulin binding activity of fodrin is rapidly destroyed by trypsin (Glenney et al., 1983, Harris and, chymotrypsin, endoprotease Glu-C and endoprotease Lys-C were tested for their ability to produce fodrin fragments smaller than 150 kDa which retained the ability to bind calmodulin. The results with chymotrypsin using 1251-calmodulin gel overlays to detect calmodulin-binding peptides are shown in Fig. 1. Digestion of fodrin with chymotrypsin under mild conditions yielded a 145-kDa peptide that was active (lane 2), although the affinity of this peptide for calmodulin appeared to be diminished relative to the intact molecule or to the 150-kDa natural fragment ( l a n e 1) of the CY subunit  4 ) more extensively digested fodrin (20 pg fodrin, 1 and 16-h digests respectively; 1:l mol/mol, enzyme/ substrate; 0 "C). B, autoradiogram of the gel in A after lZSI-calmodulin overlay. The calmodulin-binding fragments at 150 and 32 kDa arise from the a subunit, while the peptide at 25 kDa arises from the 0 subunit (see text). a-Fodrin (which is also generated by CDP-I, see below, and Harris and Morrow, 1988). More extensive digestion with chymotrypsin led to the loss of the 145-kDa peptide and the appearance of two smaller peptides of 32 and 25 kDa that both bound 1251calmodulin in the gel overlay assay (lanes 3 and 4 ) . Peptide mapping (data not shown) of the 32-kDa peptide demonstrated that it was derived from the 145-kDa peptide and was similar to domain I11 of a-fodrin , indicating that this chymotryptic calmodulin-binding peptide was derived from the center of the fodrin a subunit. Interestingly, the peptide map of the 25-kDa peptide indicated that it was derived from the 8-fodrin subunit (data not shown). This was surprising since the intact 8-fodrin subunit did not bind calmodulin as determined by 1251-calmodulin gel overlay assay (Harris et al., 1986) and as shown in lane 2 where the protein remaining near 240 kDa after mild digestion is intact 8-fodrin .
Fodrin digested with endoprotease Lys-C rapidly lost all calmodulin binding activity (data not shown). Proteolytic digestion of fodrin with endoprotease Glu-C cleaved the a subunit efficiently at the hypersensitive site and yielded two peptides (145-150 kDa), the smaller of which bound calmodulin in the gel overlay assay (similar to lane 2). Prolonged digestions with this enzyme failed to generate smaller peptides that retained calmodulin binding activity (data not shown). No calmodulin binding was detected in any of the peptides in the absence of calcium (data not shown).
NTCB Cleavage of Fodrin Generates Several Reluted Calmodulin-binding Peptides-Chemical cleavage of fodrin at cysteine residues with NTCB yielded numerous peptides with calmodulin binding activity. An analysis of these peptides by two-dimensional IEF/SDS-PAGE is shown in Fig. 2A. An autoradiogram of an electrotransfer (of a gel identical to that shown in 2 A ) overlaid with 1251-calmodulin is shown in Fig.  2B. More than 80 peptides were present in this NTCB digest ( Fig. 2 A ) , and six bound calmodulin avidly enough to be detected by the overlay technique ( B ) . Mapping of the 27-kDa calmodulin-binding peptide from the gel in A indicated that it was derived from domain I11 of a-fodrin. The other calmodulin-binding peptides in this experiment were not mapped since their assignment to specific Coomassie Bluestained peptides in two-dimensional gels (A) could not be made unambiguously.
Since it was possible that the gel overlay or blotting techniques might identify peptides that did not bind calmodulin under nondenaturing conditions, it was important to confirm the calmodulin binding ability of the NTCB fragments identified in Fig. 2 by a second technique using more physiologic conditions. This was accomplished by demonstrating the ability of these peptides to bind in a calcium-dependent fashion to a calmodulin-agarose affinity column. In initial experiments, NTCB digests of human or bovine fodrin were applied to a calmodulin affinity column as described under "Materials and Methods" and eluted by the replacement of Ca2+ in the buffer with 10 mM EGTA. The peptides that bound to the column in a calcium-dependent manner were identified by two-dimensional peptide mapping (after separation by IEF/ SDS-PAGE). These peptides were all derived from the a subunit of fodrin but from multiple domains (data not shown). This suggested that "non-calmodu1in"-binding peptides bound to the column via noncovalent interactions with the "specific" calmodulin-binding peptides. Such noncovalent interactions between spectrin peptides have been previously observed (Morrow et al., 1980). Thus, more stringent conditions were employed to separate the "nonspecific" and specific calmodulin-binding peptides. NTCB-generated peptides (2 mg, in Fig. 2A) were applied to a 1.25-ml calmodulin-agarose column (Bio-Rad) under nondenaturing conditions in 1 mM Ca2+ as outlined under "Materials and Methods." The results are shown in Fig. 3. The majority of these peptides failed to bind the column (Fig. 3A). In order to exclude from the analysis the peptides which associated with the column by nonspecific adsorption as noted above, the column was washed with 6 M urea in the presence of calcium as indicated in the profile shown in A. An analysis of the peptides eluted by the urea wash is shown in B. Peptide mapping of the most abundant of these peptides near 39, 55, and 89 kDa demonstrated that they were derived from the terminal portions (NH2 and COOH) of the a subunit (data not shown). NO 8 subunit-derived peptides were identified in the fraction. The calcium-dependent calmodulin-binding peptides were then eluted from the column in the presence of 6 M urea and 10 mM EGTA (A). Analysis of these peptides by two-dimensional IEF/SDS-PAGE (C) demonstrated that they were a unique subset of the total digest, corresponded closely to those identified by '251-calmodulin gel overlay (cf. Fig. ZB), and were distinct from the peptides eluted in the presence of urea and calcium (cf. Figs. 3, B and C). Peptide mapping (Fig. 4)  strated that they all were derived from the a subunit and that they all shared common features (arrows, Fig. 4) which were also present in the map of domain I11 . The maps of the peptides near 60 kDa in Fig. 3C also contained these common spots (data not shown). Although several of the peptides such as those labeled B and C in Fig.   3C, typically had a range of isoelectric points (but identical molecular weight), no differences were seen in their peptide maps. By comparing these peptide maps with the domain maps of the a-fodrin subunit , the placement of these peptides within fodrin could be deter-

Chymotrypsin and CDP-I Cleave a-Fodrin at Different Res-a-Fodrin
idues within the Hypersensitive Site-The region joining domains I1 and I11 is unusually susceptible to proteolytic cleavage  and CDP-I rapidly and stoichiometrically cleaved the a subunit of fodrin at this hypersensitive site . CDP-I proteolysis of a-fodrin resulted in the generation of complimentary halves of the CY subunit that differed in their PI but displayed identical and anomalous molecular weights as determined by IEF/SDS-PAGE. Transfer of CDP-I-digested fodrin onto PVDF membranes resulted in quantitative transfer of the a subunit-derived peptides, but virtually no transfer of the intact / 3 subunit (Fig. 5A). Presumably the failure of the /3 subunit to transfer is related to the tendency of this subunit to aggregate (Woods and Lazarides, 1986). The more basic 150-kDa CDP-I-generated peptide, which is the calmodulinbinding fragment , was cut from the membrane and subjected to gas-phase sequencing. Reliable information was obtained for 18 of the first 23 residues; the results are presented in Fig. 6. No secondary sequences were observed in this material, indicating the cleavage by CDP-I at this site is highly specific.
Cleavage of a-fodrin at the hypersensitive site with chymotrypsin also generated two peptides which were complimentary halves of the a subunit. (These peptides are unresolved in Fig. 1, lane 2, due to their similar molecular weight).  , and the more basic peptide, which is competent for calmodulin binding, was sequenced. A Coomassie Blue-stained PVDF membrane is shown. B, identical amounts of intact fodrin (left), fodrin digested with CDP-I (center), fodrin digested with chymotrypsin (right), under conditions in which the only cleavage was at the hypersensitive site, were separated by SDS-PAGE, and overlaid with 1251-calmodulin. An autoradiogram of the resulting gel is shown. C, quantitation of the 1251-calmodulin binding to intact or digested fodrin. Digestion of fodrin with CDP-I reduced the net calmodulin binding by 17%; chymotryptic cleavage reduced the net binding by 66%. The values are the mean ? standard deviation of six determinations.
A two-dimensional IEF/SDS-PAGE analysis of this digest revealed a pattern nearly identical to that in Fig. 5A, . However, the cleavage site of chymotrypsin was not identical to that of CDP-I, since the chymotryptic fragments had reduced avidity of '251-calmodulin (Figs. 1 and 5). Gas-phase sequencing of the 150-kDa calmodulin-binding chymotryptic fragment yielded reliable information for 27 residues (Fig. 6). As with CDP-I, no secondary cleavages were detected in this sequence, indicating the high specificity of chymotryptic cleavage under the conditions used. The effect of chymotrypsin and CDP-I proteolysis on the calmodulin binding activity of fodrin is quantitated in Fig. 5, B and C. Equal quantities of fodrin, either undigested (Fig. 5B, left  lane), CDP-I digested (Fig. 5B, center lane), or chymotrypsindigested (Fig. 5B, right lane) were analyzed for '251-calmodulin binding activity by gel overlay. Cleavage by CDP-I did not significantly diminish ( p > 0.05) the quantity of 1261-calmodulin bound by the molecule and its fragments; however, cleavage by chymotrypsin 16 residues distal to the CDP-I cleavage site significantly decreased ( p < 0.001) the quantity of '251-calmodulin bound.
The Culmodulin-binding Site of a-Fodrin Bridges the Domain 11-111 Junction-The domain structure of the a-fodrin subunit as defined by proteolytic digestion with trypsin (Harris and ) is shown in Fig. 6A. This domain structure is aligned with the sequence repeat structure defined by Speicher and Marchesi (1984), and as deduced from McMahon et al. (1987) and Let0 et al. (1988) (Fig. 6B). Although most of the molecule is comprised of regular 106residue repeating units, repeats 10 and 11 are abnormally short and long, respectively. Based on the data presented here the eleventh repeat must contain the site of calmodulin binding as indicated in Fig. 6, B and C. An enlargement of this region demonstrates the alignment of the three smallest calmodulin-binding peptides generated by NTCB digestion (C and D). All of the calmodulin-binding peptides span portions of repeats 11 and 12 and the amino-terminal portion of repeat 13. The 32-kDa chymotrypsin-generated calmodulin-binding fragment is also derived from this region (not shown in Fig.  6). These peptides span the domain 11-111 junction, the region previously associated with calmodulin binding . The amino acid sequence of the terminal portion of repeat 11 and the first 6 residues of repeat 12 is shown in Fig. 6E for human brain spectrin, along with the cleavage sites of CDP-I and chymotrypsin (arrows, E ) derived from the sequencing studies reported here. The alanine and serine immediately below the sequence shown in E represent substitutions found in the bovine brain material (that was sequenced) as compared to the human sequence. CDP-I cleaved the a subunit on the carboxyl side of tyrosine 104 while chymotrypsin cleaved on the carboxyl side of tryptophan 120. Since chymotryptic cleavage at this latter residue significantly reduces the calmodulin binding activity of fodrin, it is likely that the residues in this region contribute to the binding site. Similarly, trypsin rapidly eliminates the calmodulin activity of fodrin, presumably due to cleavage at either Lys'*'; Arg124, or Lys' (of repeat 12). In conclusion, the sequence Ser-Lys-Thr-Ala-Ser-Pro-Trp-Lys-Ser-Ala-Arg-hu-Met-Val-His-Thr-Val-Ala-Thr-Phe-Asn-Ser-Ile-Lys appears to represent the calmodulin-binding site within a-fodrin.

DISCUSSION
These results identify a single high affinity calmodulinbinding site and the specific site of CDP-I and chymotrypsin cleavage within the hypersensitive site of a-fodrin. Several lines of evidence indicate that this is a physiologically relevant

CI IYMO
FIG. 6. Localization of the sites of CDP-I cleavage and calmodulin binding within fodrin. A , the proteolytic domain structure of brain spectrin as defined by tryptic digestion . The Roman numerals indicate the domains and their sizes (kDa) as determined by SDS-PAGE as indicated. B , the corresponding 106-residue repeat structure (defined by Speicher and Marchesi, 1984, and as deduced from McMahon et al., 1987, and Let0 et al., 1988. Calmodulin binds near the center of the molecule, in the carboxyl half of repeat 11. C, the alignment of the major calmodulin-binding NTCB peptides, with their apparent masses in kDa, in an expanded view of the center of the molecule (repeats 9-13). D, the positions of the cysteine residues are indicated ( S H ) to demonstrate the origin of the NTCB-generated calmodulin-binding peptides. The smallest calmodulin-binding NTCB fragment, N14, presumably arises from material that was previously cleaved by CDP-I in uiuo and then by NTCB. E, the sequence of the terminal portion of repeat 11 and the beginning of repeat 12 of human brain spectrin is shown. (The amino acid substitutions indicated below the peptide (in parentheses) were found in the bovine fodrin protein sequence). The peptide SKTASPWKDARLMVHTVATFNSIK (from residue 114 to the carboxyl terminus of repeat 11 and the first 6 residues of repeat 12) is likely to include the actual calmodulin binding region (see text).

TABLE I All sequences compared to SKM MLCK using "Bestfit." Capital letters represent longest region that yielded the best fit, while lower case letters represent additional amino acids in the sequences identified as the calmodulin binding domains. Vertical lines indicate similar amino acids as compared to SKM MLCK.
Protein 10  * CaATPase, human erythrocyte Ca2+ pump from James et al. (1988). e SKM PPK, rabbit skeletal muscle phosphorylase kinase y subunit from Lukas et al. (1986). e CAM 11, rat brain type I1 Ca'+/calmodulin-dependent protein kinase p subunit from Bennet and Kennedy 'B-SPEC, human erythrocyte spectrin p subunit.' SKM PFK, rabbit skeletal muscle phosphofructokinase from Buschmeier et al. (1987). (1987).

CDP-I Cleavage Sites in a-Fodrin
calmodulin binding domain. 1) All a-fodrin peptides that bind mation for cleavage at the Tyrlo4 G1y1O5 bond by CDP-1. calmodulin in a calcium-dependent manner in solid-phase gel The sequences of the calmodulin binding region in several overlay assays also bind calmodulin-agarose affinity columns calmodulin-binding proteins are aligned in Table I. While in a calcium-dependent manner under nondenaturing condi-some of these peptides share limited homology, no clear tions. 2) All of these calmodulin-binding peptides encompass consensus sequence can be appreciated, and the fodrin sea common region of the a subunit. 3) The CDP-I derived 150-quence reported here shows the least degree of similarity. It kDa peptide retains full calmodulin binding activity, while an has been suggested that a common secondary structure in overlapping chymotrypsin-generated 150-kDa peptide devoid calmodulin-binding peptides is an amphipathic a helix of 16 amino-terminal residues has reduced calmodulin binding flanked by basic residues (Cox et al., 1985, Malencik et al., activity. 4) The sequence of the putative calmodulin binding 1986, O'Neil et al., 1987). The calmodulin binding domain of domain is similar to other calmodulin-binding peptides. The a-fodrin, while amphipathic, contains a proline residue that calmodulin-binding site is in the center of the a subunit, appears to be essential for full activity. In addition, the overlapping structural repeats 11-12. This location is consist-secondary structure predicted for this region based on the ent with that determined by peptide mapping studies (Harris criteria of Chou and Fasman (1978) (using the program Pep- , immunologic data (Glenney et al., 1983), tidestructure, from the Wisconsin Molecular Biology Comand calmodulin binding studies on fusion proteins generated puter Group) is that of a rigid turn connected to a weak helix from Drosophila cDNA clones (Byers et al., 1987). The cal-or p sheet. Andersen and Malencik (1986) demonstrated that modulin-binding site in a-fodrin is distinct in terms of both a wide variety of unrelated calmodulin-binding peptides were sequence and location from that in human erythrocyte spec-predicted to consist of a helix, p strand, anti-parallel p sheet trin, which binds calmodulin near the amino terminus of the and/or undefined structures and introduced the idea that an p subunit4 (Anderson and Morrow, 1987). amphipathic a helix is not the only structure that can bind It is difficult to reconcile these findings with the report of calmodulin with high affinity. Therefore, if the structure of Tsukita et al., (1983), in which the calmodulin-binding site fodrin is as predicted, the results presented here reinforce the was reported to be 10-20 nm from the midpoint of a 220-nm notion that an a helix is not necessary for a high affinity tetrameric molecule. This translates to approximately 24-48 calmodulin-binding site and extends the concept to proteins kDa from the amino terminus of the molecule, which would that bind calmodulin. place the binding site in the middle of domain I (Harris and The close proximity of the CDP-I cleavage site to the site . We cannot demonstrate that domain I (gen-of calmodulin binding is significant since these two activities erated by trypsin digestion) binds calmodulin, nor were any are functionally linked. The binding of calmodulin to a-fodrin fragments that bound "specifically" to the calmodulin affinity accelerates the rate of CDP-I proteolysis of both subunits of column in a calcium-dependent manner derived solely from fodrin3 (Seubert et al., 1987). Significantly, after a-fodrin has domain I. However, since these two studies have used dissim-been cleaved by CDP-I, its ability to cross-link actin filaments ilar approaches to identify a calmodulin-binding site, it is becomes reversibly regulated by ~almodulin.~ Thus, calmopossible that under different conditions additional calmodulin dulin and CDP-I exert a synergistic regulatory effect on fodrin binding sites may become a~p a r e n t .~ Similarly, as noted ear-function; the observations reported here provide a firm struclier (Fig. I), a 25-kDa peptide derived from the / 3 subunit by tural basis for these functional observations. In the future, it chymotrypsin digestion bound '251-calmodulin in a gel overlay will be of interest to identify the sites of subunit-subunit assay. The binding of calmodulin to this peptide is probably interaction, since a-P interactions are clearly altered by CDPnot physiologically relevant, since the intact p subunit (top I proteolysis and calmodulin binding.3 band of lane 2, Fig. 1) does not bind '251-calmodulin (see also Harris et al., 1986). Thus, it appears that peptides may acquire