Ca 2 1 -dependent and Ca 2 1 -independent Calmodulin Binding Sites in Erythrocyte Protein 4.1 IMPLICATIONS FOR REGULATION OF PROTEIN 4.1 INTERACTIONS WITH TRANSMEMBRANE PROTEINS

,

Calmodulin (CaM) 1 is a highly conserved Ca 2ϩ -binding protein that modulates the functional activities of different structural and transport proteins and the activities of many Ca 2ϩdependent enzymes (for a review, see Refs. 1 and 2). In most cases, Ca 2ϩ is essential for the initial binding of CaM to these proteins and for subsequent modulation of their function. However, a small subset of proteins including erythroid 4.1 (4.1R) bind CaM in a Ca 2ϩ -independent manner but require Ca 2ϩ for manifesting changes in function. Of particular relevance to the present study is the role played by CaM in modulating various protein-protein interactions in the erythrocyte membrane involving 4.1R.
CaM is present in human erythrocytes at micromolar concentration (3)(4)(5). In these cells, CaM binds to Ca 2ϩ -ATPase (6, 7) with a dissociation constant on the order of 10 nM (8), while it binds to membrane skeletal proteins, 4.1R, and adducin with a dissociation constant on the order of 0.1-0.2 M (3, 9 -12). Through its interaction with Ca 2ϩ -ATPase, CaM enables the erythrocyte to maintain a submicromolar concentration of Ca 2ϩ . Since only 5% of the CaM in erythrocyte is involved in interaction with Ca 2ϩ -ATPase (13), 95% of CaM is available for interaction with other membrane proteins and modulation of their functional interactions. In fact, recent evidence suggests that Ca 2ϩ /CaM reduces the affinities of the spectrin-actinadducin interaction (9,14) and of the spectrin-actin-4.1R interaction (11,15). Ca 2ϩ /CaM also decreases the affinity of interaction between 4.1R and band 3 and between 4.1R and glycophorin C (16). By modulating the affinities of these different protein-protein interactions, Ca 2ϩ /CaM can play a significant role in regulating the function of the erythrocyte membrane (10,11,15,16).
Biochemical and biophysical studies have facilitated construction of a structural and functional map of the 4.1R molecule (for a review, see Ref. 17). Four major structural domains of 4.1R with apparent molecular masses of 30, 16, 10, and 22-24 kDa were identified. 4.1R interacts with spectrin and actin through its 10-kDa domain and with integral membrane proteins glycophorin C and band 3 through its 30-kDa domain (reviewed in Ref. 17). CaM has previously been shown to bind to the 30-kDa domain of 4.1R at a molar ratio of 1:1 in a Ca 2ϩ -independent manner (11). However, there is an absolute Ca 2ϩ requirement for CaM-induced regulation of 4.1R interactions with other membrane proteins (15,16). To understand the molecular basis for this Ca 2ϩ requirement, we explored the possibility that the presence of both Ca 2ϩ -dependent and Ca 2ϩindependent CaM binding sites in the 30-kDa domain may account for CaM regulation of protein-protein interactions involving protein 4.1R.
In the present study, we quantitated the affinity of interaction between CaM and 4.1R, its 30-kDa domain, and various synthetic peptides. CaM bound to the 30-kDa domain of 4.1R in Ca 2ϩ -independent manner at a molar ratio of 1:1 with a dissociation constant on the order of 0.1 M, consistent with previous findings (11). We identified two different CaM-binding sequence motifs in the 30-kDa domain of 4.1R: one encoded by exon 9 (AKKLSMYGVDLHKAKDL; peptide 9) and the other by exon 11 (AKKLWKVCVEHHTFFRL; peptide 11) (18). These sequences share conserved sequence motifs with the CaM binding sequence of Ca 2ϩ -ATPase (6). Importantly, while peptide 11 bound CaM with high affinity in the absence of Ca 2ϩ , binding of CaM to peptide 9 was Ca 2ϩ -sensitive. Serine 185 in peptide 9 appeared to play a critical role in Ca 2ϩ -dependent CaM binding, since replacement of this serine by tryptophan resulted in loss of Ca 2ϩ -dependent binding of CaM to 4.1R. Ca 2ϩ decreases the affinity of band 3 binding to the 30-kDa domain in the presence of CaM. However, Ca 2ϩ had no effect on the affinity of the band 3 interaction with the 30-kDa domain when either the Ca 2ϩ -independent or both the Ca 2ϩ -dependent and -independent CaM binding sites were deleted. Thus, regulation of 4.1R binding to membrane proteins by Ca 2ϩ and CaM requires binding of CaM to both Ca 2ϩ -dependent and Ca 2ϩ -independent sites in 4.1R. Based on these findings, we propose that two distinct domains in 4.1R are responsible for CaM binding and that one of these domains is responsible for Ca 2ϩ -sensitive regulation of 4.1-R interactions with membrane proteins.
Synthesis of Peptides-Peptides corresponding to the internal sequence of the 30-kDa domain of protein 4.1R (18) and peptides 9, 11 ( Fig. 1B), 9-a (A 181 KKLSMYGV), and 9-b (D 190 LHKAKDL) were chemically synthesized using the peptide synthesizer (PerSeptive Biosystems 9050 Plus, Boston, MA) according to the Fmoc method (23). The synthesized peptides were purified by reverse phase high performance liquid chromatography. The molecular mass of the different synthetic peptides was determined by mass spectrometry (matrix-assisted laser desorption-ionization time-of-flight mass spectrometry; REFLEX II TM , Bruker Instruments, Switzerland) (24,25) and compared with the theoretically calculated mass to validate the chemical composition of the synthesized peptides. The chemically synthesized peptides were coupled to biotin via the spacer sequence, SGSG, at the amino terminus of these peptides followed by biotinylation using succinimide ester (26).
Preparation of Recombinant Proteins-Recombinant 30-kDa domain of 4.1R (r30kDa), r30kDa from which sequences encoded by exons 5, 9, or both exons 9 and 11 had been deleted (⌬Ex.5, ⌬Ex.9, and ⌬Ex.9/11, respectively; Fig. 1A), and a recombinant protein corresponding to the cytoplasmic domain of band 3 (27) were produced as glutathione Stransferase fusion proteins in Escherichia coli. cDNAs encoding these different sequences were amplified by polymerase chain reaction using specific primers, and the minigenes were constructed in modified pGEX-KG vector and sequenced (19). The fusion proteins were expressed as described previously (28,29) and purified from bacteria lysates by affinity column chromatography using glutathione-Sepharose 4B. The purified fusion proteins were cleaved with thrombin. The replacement of Ser 185 by Trp 185 , of Phe 277 -Phe 278 by Ala 277 -Ala 278 , and of Trp 268 and Phe 277 -Phe 278 by Ala 268 and Ala 277 -Ala 278 in r30kDa in the pGEX-KG vector was performed using the QuikChange TM site-directed mutagenesis kit. r30kDa was further purified by affinity column chromatography using CaM-Sepharose CL-4B (16), while the other recombinant proteins were purified using glutathione-Sepharose 4B to remove cleaved glutathione S-transferase. The purity of the recombinant proteins was assessed by SDS-polyacrylamide gel electrophoresis (15% gel) analysis. The protein concentrations were determined using the following relationship: protein concentration (mg/ml) ϭ 1.45 A 280 Ϫ 0.74 A 260 .
Binding Assay by Resonant Mirror Detection-Protein-protein interactions and protein-peptide interactions were studied using the resonant mirror detection method (30 -32) of the IAsys TM system (Affinity Sensors, Cambridge, UK). To measure the ability of different 4.1R peptides and 4.1R recombinant proteins to bind CaM, the biotinylated peptides or proteins were immobilized to the aminosilane cuvette according to the manufacturer's instructions with slight modifications (32). All experimental procedures were carried out at 25°C with constant stirring. In brief, an aminosilane cuvette was activated with 2 mM of bis(sulfosuccinimidyl)suberate. After extensive washing with distilled water, proteins suspended in phosphate-buffered saline (10 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 7.4, containing 0.15 M NaCl) at a concentration of 0.1 mg/ml were added to the cuvette and incubated for 30 min at 25°C. After washing with phosphate-buffered saline, 2 mg/ml of bovine serum albumin in phosphate-buffered saline was added to the cuvette to reduce nonspecific binding. The biotinylated peptides were immobilized onto the cuvette surface through streptavidin. Cuvettes with immobilized bovine serum albumin alone or streptavidin alone were prepared to serve as negative controls for binding studies. To quantitate the effects of Ca 2ϩ and CaM on the binding of different r30kDa proteins to transmembrane proteins, the recombinant cytoplasmic domain of band 3 was immobilized on the surface of the aminosilane cuvette. In some of the mutant constructs, sequences encoded by specific exons are deleted (⌬Ex.5, ⌬Ex.11, and ⌬Ex.9/11), while in others specific amino acid residues indicated by open circles are mutated (Ser 185 replaced with tryptophan and Trp 268 or Phe 277 -Phe 278 replaced by alanine). The locations of two CaM binding peptides encoded by exon 9 (peptide 9) and exon 11 (peptide 11) are shown as solid bars. B, amino acid sequences of peptides 9 and 11. The mutated amino acids are represented as outlined letters.
lyzed using the software package FASTfitTM (Affinity Sensors, Cambridge, UK). The dissociation constant from this form of kinetic analysis (termed K (D)kin ) is then calculated as follows: K (D)kin ϭ k d /k a , where k a is the association rate constant and k d is the dissociation rate constant. The dissociation constant by Scatchard plot analysis (termed K (D) K (D)Scat ) was also derived from the binding data (33). In the present study, the K (D)Scat derived under all experimental conditions closely matched the corresponding K (D)kin values calculated.
The stoichiometry of 4.1R binding to CaM in the presence and absence of Ca 2ϩ was determined using the IAsys TM system. Maximal binding (B max ) of 4.1R represented as arc seconds was obtained from the Scatchard plot as described previously (33). The amount of immobilized CaM on the aminosilane cuvette was determined as the difference of arc seconds between bis(sulfosuccinimidyl)suberate and CaM under equilibrium conditions. The stoichiometry of 4.1R binding to CaM was calculated according to the following equation described in the Method Guide of the IAsys TM system: Stoichiometry of 4.1R:CaM ϭ (B max of 4.1R/80,000):(amount of immobilized CaM on aminosilane cuvette/ 18,000), where 80,000 and 18,000 are apparent molecular weights of 4.1R (18) and CaM (1, 2), respectively.
The cuvettes were reused after cleaning with HCl. Original binding curves could be replicated after HCl washes, implying that the washing procedure used did not denature the bound ligands.

30-kDa Domain of 4.1R Binds CaM-
The ability of CaM to bind to native 4.1R and to its four major structural domains (30-, 16-, 10-, and 22/24-kDa domains) was determined using the IAsys TM system. All binding studies were performed at physiologic ionic strength. Ca 2ϩ dependence of binding was evaluated at defined concentrations of Ca 2ϩ . Analysis of binding response curves (Fig. 2) obtained at varying concentrations of 4.1R in the presence of EGTA ( Fig. 2A) provided k a and k d of 6.1 ϫ 10 4 M Ϫ1 s Ϫ1 and 2.0 ϫ 10 Ϫ2 s Ϫ1 , respectively. From these measured values of k a and k d , a K (D)kin value of 0.33 M was derived for Ca 2ϩ -independent CaM binding to 4.1R. Ca 2ϩ had little effect on the affinity of 4.1R binding to CaM. The stoichiometry of 4.1R binding to CaM was calculated as described under "Experimental Procedures." Using the values of B max (970 arc seconds) in the presence and absence of Ca 2ϩ and the amount of immobilized CaM on the aminosilane cuvette (200 arc seconds) (Fig. 3), a molar ratio of 4.1R binding to CaM was derived to be 1:1 both in the presence and absence of calcium. This finding is consistent with earlier findings on stoichiometry of Ca 2ϩ -independent binding of 4.1R to CaM (11,16). CaM also bound to the N-terminal 30-kDa domain isolated from native 4.1R and to r30kDa with a K (D)kin value of 0.35 M (Fig. 2B,   Table I) at a molar ratio of 1:1. Again, Ca 2ϩ had little effect on the affinity or stoichiometry of CaM binding to the 30-kDa domain. In contrast, no binding signal could be detected for interaction between CaM and 16-, 10-, or 22/24-kDa domains of 4.1R (Fig. 2C). Thus, CaM binds to the 30-kDa domain of 4.1R in a Ca 2ϩ -independent fashion, consistent with earlier reports (11,16). However, these results do not rule out the presence of additional Ca 2ϩ -dependent binding sites, of equal or lower affinity, whose existence could be masked by the high affinity Ca 2ϩ -independent binding site.
CaM Binding Sites in the 30-kDa Domain of 4.1R-To identify CaM binding sites in the 30-kDa domain of 4.1R, we quantitated CaM binding to r30kDa from which sequences encoded either by exon 11 (⌬Ex.11) or both exons 9 and 11 (⌬Ex.9/11) had been deleted. As noted before, CaM binds wild type r30kDa with a K (D)kin value on the order of 0.1 M either in the presence or absence of Ca 2ϩ . Deletion of sequences encoded by both exons 9 and 11 (⌬Ex.9/11) completely abolished the binding of the mutant r30kDa peptide to CaM either in the presence or absence of Ca 2ϩ (Table I). This finding implies that CaM binding domain sequences are encoded by these two exons.
Interestingly, the mutant r30kDa ⌬Ex.11 polypeptide exhibited Ca 2ϩ -sensitive CaM binding with a K (D)kin value of 30 M in the absence of Ca 2ϩ and a K (D)kin value of 5.1 M in the presence of Ca 2ϩ (Table I). These data imply that sequences encoded by exon 9 exhibit Ca 2ϩ -sensitive CaM binding. It should be noted that the affinity of the interaction of this mutant protein with CaM was markedly decreased compared with the wild type protein, suggesting that sequences encoded by exon 11 may constitute the high affinity binding site in 30-kDa domain for CaM. We were unable to study CaM binding to ⌬Ex.9 r30kDa, since this recombinant protein could not be obtained in a soluble form. Deletion of sequences encoded by exon 5 (⌬Ex.5) from the r30kDa polypeptide had no effect on CaM binding in the presence or absence of Ca 2ϩ (data not shown). Taken together, these data imply that two distinct CaM binding sites encoded by exons 9 and 11 exist in 4.1R.
Identification of Sequence Motifs in the 30-kDa Domain of 4.1R That Bind CaM-We identified the putative CaM binding sequence motif, AKKLXXXXVXX(X)HXXXXL (where the XX(X) indicates that there may be 2 or 3 unknown residues) in the sequences encoded by both exons 9 and 11. The sequence encoded by exon 9 is AKKLSMYGVDLHKAKDL (peptide 9), while the one encoded by exon 11 is AKKLWKVCVEHHTF-FRL) (peptide 11) (Fig. 1B). To test directly whether these sequences can bind CaM, the corresponding synthetic peptides were made, and their binding to CaM was assayed (Table II). CaM bound to peptide 11 with a K (D)kin value of 0.54 M both in the presence and absence of Ca 2ϩ . CaM binding to peptide 9 was highly sensitive to Ca 2ϩ . CaM bound to peptide 9 with a K (D)kin value of 1.7 M in the presence of Ca 2ϩ but with a much higher K (D)kin value of 40 M in the absence of Ca 2ϩ . Peptide 9-a (A 181 KKLSMYGV) also showed Ca 2ϩ -sensitive CaM binding, while binding of peptide 9-b (D 190 LHKAKDL) to CaM was not Ca 2ϩ -sensitive (data not shown). It is noteworthy that there was an absolute requirement for Ca 2ϩ for interaction of CaM with peptide 9-a. These data suggest that the Ca 2ϩ -sensitive CaM binding site in peptide 9 requires sequences in peptide 9-a.
The binding assays thus revealed two distinct CaM binding sites in 4.1R: a Ca 2ϩ -independent site in peptide 11 and a Ca 2ϩ -dependent site in peptide 9. Further analysis of the data showed that although both peptide 9 and peptide 11 bind CaM in the absence of Ca 2ϩ , the affinity of Ca 2ϩ -independent binding of CaM to peptide 11 was almost 80-fold higher than that of CaM binding to peptide 9. This difference was mainly due to differences in the k a of CaM interactions with these two peptides (k a values of 6.5 ϫ 10 2 M Ϫ1 s Ϫ1 for peptide 9 and 5.4 ϫ 10 4 M Ϫ1 s Ϫ1 for peptide 11). Furthermore, while Ca 2ϩ had no effect on the interaction of CaM with peptide 11, it had a marked effect on the interaction of CaM with peptide 9. In fact, Ca 2ϩ increased the affinity of the CaM-peptide 9 interaction 25-fold to a value very close to that of the interaction between CaM and peptide 11 in either the presence or absence of Ca 2ϩ . This large increase in affinity induced by Ca 2ϩ is in large part due to its effect on the k a of the CaM-peptide 9 interaction. We suggest that this Ca 2ϩ -induced increase in k a values is most likely due to Ca 2ϩ -induced change in the conformation of CaM.
In effect, these binding data suggest that CaM may bind preferentially to peptide 11 in the absence of Ca 2ϩ and that increased Ca 2ϩ may stimulate binding to a second site in peptide 9. To further validate a role for Ca 2ϩ in modulating the interaction of CaM with peptide 9, we measured the ability of a preformed CaM-peptide 11 complex to bind peptide 9 in the presence and absence of Ca 2ϩ (Table II). As with native CaM, the CaM-peptide 11 complex bound peptide 9 (Table II) as well as peptide 9-a (data not shown) in a Ca 2ϩ -dependent manner.  (D)kin for the interaction between CaM and 4.1R or r30kDa (wild type), exon 11-deleted r30kDa (⌬Ex.11), and exons 9 and 11-deleted r30kDa (⌬Ex.9/11), in the presence and absence of Ca 2ϩ are shown. 4.1 R and r30kDa proteins (50 nM to 1 M) were incubated with CaM immobilized on the aminosilane cuvette in the presence of either 0.1 mM EGTA (EGTA) or 1.1 mM CaCl 2 and 1.0 mM EGTA (ϩCa 2ϩ ) in buffer A as described under "Experimental Procedures." From the binding curves obtained by resonant mirror detection method, K (D)kin was determined using the software package FAST-Fit™. K (D)kin represents mean Ϯ S.D. (n ϭ 3-5).    Binding to peptide 9-b once again was not Ca 2ϩ -sensitive (data not shown). The affinity between the CaM-peptide 11 complex and peptide 9 was an order of magnitude higher than the affinity between native CaM and peptide 9 (Table II), suggesting potential cooperativity between peptide 9 and peptide 11 sequences in modulating the binding of the 30-kDa domain to CaM.
CaM bound to peptides 9 and 11 at physiologically relevant ionic strength (0.15 M NaCl). However, CaM began to dissociate from these peptides at higher ionic strengths (0.2 M NaCl and greater) with complete dissociation observed at 0.6 M NaCl (data not shown). CaM did not bind to various synthetic peptides representing sequences surrounding the CaM binding sequence motifs identified in the 30-kDa domain. In particular, CaM did not bind to peptides encoding the following sequence motifs: E 198 GVDIIL (sequence downstream of peptide 9), T 281 STDTIPK (sequence downstream of peptide 11), or I 241 RPGEQEQYESTIGFKLPSYRA (sequence upstream of peptide 11) either in the presence or absence of Ca 2ϩ (data not shown). In addition, CaM did not bind to bovine serum albumin or streptavidin in our assay system (data not shown).
Serine 185 in Peptide 9 Is Critical for Ca 2ϩ -dependent Binding to CaM-To determine the molecular basis for the differences in Ca 2ϩ sensitivity for CaM binding to peptides 9 and 11, we performed binding studies using peptides 9 and 11 with defined mutations. Although the first four amino acids are identical in both peptides (A 181 KKL in peptide 9 and A 264 KKL in peptide 11), the fifth amino acid residue is different (Ser 185 in peptide 9 and Trp 268 in peptide 11) (Fig. 1). Replacing this Ser 185 residue in peptide 9 with Trp in effect converted the Ca 2ϩ -dependent binding site into a Ca 2ϩ -independent site, mainly by enhancing CaM binding in the absence of Ca 2ϩ (Table II). Similar Ca 2ϩ -independent binding was noted using the r30kDa domain in which Ser 185 was mutated to Trp 185 (data not shown). More importantly, it was possible to impart Ca 2ϩsensitive binding to peptide 11 by replacing residues Trp 268 -Lys 269 in peptide 11 with Ser 268 -Leu 269 . This effect is due to a decreased k a of the Ser 268 -Leu 269 mutant for CaM in the absence of Ca 2ϩ (k a values of 5.4 ϫ 10 4 M Ϫ1 s Ϫ1 for wild type peptide and 5.6 ϫ 10 2 M Ϫ1 for mutant peptide). Interestingly, replacing just the tryptophan residue in peptide 11 with the serine residue was not as effective in imparting Ca 2ϩ -sensitive binding of peptide 11 to CaM. These data strongly imply that the serine residue in the CaM binding motif of 4.1R plays a critical role in imparting Ca 2ϩ -sensitive binding.
Aromatic Amino Acids in Peptide 11 Are Critical for Ca 2ϩindependent Binding to CaM-It has been previously documented that tryptophan and phenylalanine play a critical role in Ca 2ϩ -dependent binding of CaM (1). In order to determine if aromatic amino acids also play an important role in Ca 2ϩindependent binding of CaM to 4.1R, we measured its binding to peptide 11 in which different amino acid substitutions were made. Replacement of Trp 268 by alanine markedly decreased the affinity for CaM (Table III). Whereas the wild type peptide bound with a K (D)kin of ϳ0.5 M, the W268A mutant had an increased K (D)kin of ϳ35 M, predominantly due to a marked decrease in the k a (k a values of 5.4 ϫ 10 4 M Ϫ1 s Ϫ1 for wild type peptide and 4.2 ϫ 10 2 M Ϫ1 s Ϫ1 for mutant peptide). Similar changes in the affinity of the interaction between peptide 11 and CaM were measured in mutant F277A/F278A. Complete abrogation of CaM binding to peptide 11 was noted when the tryptophan residue (Trp 268 ) as well as the two phenylalanine residues (Phe 277 -Phe 278 ) in peptide 11 were replaced by alanine residues. These findings imply that the aromatic amino acids in peptide 11 of 4.1R play a critical role in Ca 2ϩ -independent binding to CaM. Data obtained using r30kDa domain in which  (Table IV) or Ca 2ϩ alone (data not shown) had no effect on this binding affinity. However, in the presence of both Ca 2ϩ and CaM, the affinity of binding of r30kDa to band 3 decreased 10-fold (Table IV). Deletion of CaM binding motifs did not alter the intrinsic affinity of r30kDa for band 3 but did abolish the Ca 2ϩ -CaM-induced decrease in binding affinity (Table IV). Wild type r30kDa, ⌬Ex.11, and ⌬Ex.9/11 all bound to cytoplasmic domain of band 3 with K (D)kin values ranging from 0.1 to 0.2 M in the absence of Ca 2ϩ , but only r30kDa showed reduced binding affinity in the presence of Ca 2ϩ (Table IV). Most importantly, while replacement of Ser 185 by Trp 185 in exon 9 of r30kDa did not alter the intrinsic affinity of the mutant protein to the cytoplasmic domain of band 3, it completely abolished the Ca 2ϩ -CaM-induced decrease in binding affinity (Table IV). These results strongly imply that Ca 2ϩ and CaM regulation of the interaction of 4.1R with band 3 requires both CaM binding sites and that Ser 185 plays a critical role in modulating this interaction.

Regulation of Band 3-Protein 4.1 Interaction by Ca 2ϩ and CaM-
The Ca 2ϩ concentration dependence of the CaM-modulated interaction between 4.1R and band 3 was quantitated (Fig. 4). At Ca 2ϩ concentrations greater than 0.1 M (pCa Ͻ 7), the extent of r30kDa binding to band 3 started to decline, and maximal inhibition of binding was noted at Ca 2ϩ concentrations of 100 M and higher (pCa Ͻ 4). Half-maximal effect was seen at a Ca 2ϩ concentration of ϳ2 M.

DISCUSSION
The present study has identified two distinct CaM binding sites in 4.1R encoded by two different exons. These two sites exhibit interesting differences in their affinity for CaM in the presence and absence of Ca 2ϩ . While the site encoded by exon 11 binds CaM with high affinity in the absence of Ca 2ϩ , the site encoded by exon 9 binds CaM with high affinity only in the presence of Ca 2ϩ . The measured stoichiometry of CaM binding to 4.1R (1:1) implies that one molecule of CaM binds to two distinct binding sites in the same molecule of 4.1R. Thus, 4.1R is a unique CaM-binding protein in that it has distinct Ca 2ϩ - No binding dependent and Ca 2ϩ -independent high affinity CaM binding sites. While in the vast majority of CaM-binding proteins the binding domain is believed to be contained in a single contiguous sequence motif, a limited number of proteins, including phosphorylase kinase and caldesmon, have been shown to contain two distinct noncontiguous CaM-binding domains (34,35). 4.1R belongs to this latter class of CaM-binding proteins with one significant difference. In contrast to phosphorylase kinase and caldesmon, which bind CaM with high affinity only in the presence of Ca 2ϩ , 4.1R binds CaM with high affinity even in the absence of Ca 2ϩ (11). Although 4.1R can bind CaM in a Ca 2ϩindependent manner, modulation of 4.1R interactions with its binding partners requires both Ca 2ϩ and CaM (11,16). Our finding that 4.1R possesses both a Ca 2ϩ -dependent and a Ca 2ϩindependent CaM binding site and that high affinity binding of CaM to both of these sites is critical for regulation of 4.1R binding to membrane proteins provides a mechanism to explain earlier observations regarding the effect of Ca 2ϩ /CaM on membrane protein interactions with 4.1R (11,15,16).
By quantitating the binding of CaM to various synthetic peptides that represent the two distinct CaM binding domains, we have gained insight into the nature of the interaction between CaM and 4.1R. CaM bound to peptide 9 (AKKLSMYG-VDLHKAKDL) with K (D)kin of 1.7 M in the presence of Ca 2ϩ and 40 M in its absence. This large increase in affinity induced by Ca 2ϩ is mostly due to its effect on the k a , suggesting that a Ca 2ϩ -induced conformational change in CaM may be responsible for its increased affinity to peptide 9. In contrast, CaM binding to peptide 11 (AKKLWKVCVEHHTFFRL) is not sensitive to Ca 2ϩ , suggesting that Ca 2ϩ has no effect on the conformation of the binding domain of CaM for peptide 11. It is also interesting to note that while both peptide 9 and peptide 11 can bind CaM in the absence of Ca 2ϩ , the affinity of Ca 2ϩindependent binding of CaM to peptide 11 is almost 80-fold higher than that of CaM binding to peptide 9. This finding may  Fig. 1. Ca 2ϩ concentration inside red cells is normally maintained at less than 1.0 M. At this Ca 2ϩ concentration, CaM is bound predominantly to the Ca 2ϩ -independent site (peptide 11). With elevated Ca 2ϩ concentrations, CaM binding affinity to the Ca 2ϩdependent site (peptide 9) is increased, causing a conformational change in the binding site for band 3 (encoded by exon 5) and a resultant decrease in the affinity of 4.1R for band 3. This model implies that Ca 2ϩ and CaM effect on protein 4.1R binding to transmembrane proteins requires CaM binding to both Ca 2ϩ -independent and Ca 2ϩ -dependent binding sites in 4.1R.

TABLE IV
Ca 2ϩ /CaM down-regulation of r30kDa-band 3 interaction K (D)kin for the interaction between cytoplasmic domain of band 3 and r30kDa (wild type), r30kDa in which Ser 185 was mutated to Trp 185 , exon 11-deleted r30kDa (⌬Ex.11), and exons 9 and 11-deleted r30kDa (⌬Ex.9/11) in the presence and absence of Ca 2ϩ are shown. r30kDa protein (50 nM to 1 M) was preincubated with CaM (5 M) and either 0.1 mM EGTA (EGTA) or 1.1 mM CaCl 2 and 1.0 mM EGTA (ϩCa 2ϩ ) for 30 min at 25°C in buffer A. The complex of CaM and r30kDa protein was incubated with the cytoplasmic domain of band 3 immobilized on the aminosilane cuvette. Binding assays were carried out as described under "Experimental Procedures." K (D)kin represents mean Ϯ S.D. (n ϭ 3-5). r30kDa Condition Serine 185 in 4.1R was identified as the critical residue responsible for Ca 2ϩ -sensitive binding of peptide 9 with CaM based on studies with synthetic peptides and with the entire 30-kDa domain. This conclusion is reinforced by the finding that while Ca 2ϩ had no effect on CaM binding to peptide 11, which does not contain the serine residue, replacement of Trp 268 -Lys 269 with Ser 268 -Leu 269 in this peptide induced Ca 2ϩsensitive binding of peptide 11 to CaM. Our findings further suggest that nonpolar amino acids next to the serine residue, such as Met 186 in peptide 9 and Leu 269 in the mutant peptide 11, may be required for Ca 2ϩ -sensitive binding to CaM. Aromatic amino acids, Trp 268 and Phe 277 -Phe 278 , appear to play a critical role in the high affinity binding of peptide 11 to CaM in the absence of Ca 2ϩ . Further structural studies are needed to define the role of these different residues in the interaction of 4.1R with CaM.
4.1R binds to transmembrane proteins band 3 and glycophorin C in erythrocytes. The binding site in 4.1R responsible for its interaction with band 3 is the sequence motif LEEDY (36), encoded by exon 5 (18). 4.1R binding to band 3 results in the loss of the high affinity binding sites for ankyrin on band 3 with resultant alterations in membrane mechanical properties (37). Since Ca 2ϩ /CaM can dissociate 4.1R from band 3, they can indirectly modulate the ankyrin-band 3 interaction and hence the membrane mechanical properties (37). The effect of Ca 2ϩ / CaM on 4.1R binding to different transmembrane proteins is yet to be quantitatively evaluated. In the present study, we document that Ca 2ϩ and CaM regulate the interaction of 4.1R with band 3. In the absence of either CaM or Ca 2ϩ , r30kDa bound to band 3 with high affinity (K (D)kin of 0.18 M), but in the presence of both Ca 2ϩ and CaM, the affinity of interaction was reduced by an order of magnitude (K (D)kin of 2.0 M). However, Ca 2ϩ /CaM had no effect on the affinity of the interaction between band 3 and the 30-kDa domain of 4.1R, from which either the Ca 2ϩ -independent or both the Ca 2ϩ -dependent and -independent CaM binding sites were deleted. These data strongly imply an absolute requirement for high affinity binding of CaM to both the Ca 2ϩ -dependent and the Ca 2ϩindependent sites in 4.1R for regulation of 4.1R binding to band 3 by Ca 2ϩ and CaM. Furthermore, the importance of Ser 185 in regulating this interaction was reinforced by the finding that replacement of Ser 185 by Trp 185 in the 30-kDa domain abolished Ca 2ϩ /CaM-dependent regulation of 4.1R interaction with cytoplasmic domain of band 3.
Based on these findings, we propose the following model for the Ca 2ϩ /CaM regulation of protein 4.1R interaction with band 3 in the erythrocyte membrane (Fig. 5). Ca 2ϩ concentration inside red cells is normally maintained at less than 1.0 M. At this Ca 2ϩ concentration, CaM should bind with high affinity to the Ca 2ϩ -independent site in peptide 11 and only weakly to the Ca 2ϩ -dependent site in peptide 9. With triggered increases in Ca 2ϩ concentration, CaM binds to the Ca 2ϩ -dependent site with high affinity, causing a conformational change in 4.1R that affects the binding sites for band 3 encoded by exon 5 (18,36), resulting in decreased affinity of 4.1R for this transmembrane protein. Indirect evidence in support of such a CaMinduced conformational change in 4.1R has been provided by Tanaka et al. (11). However, direct evidence to support this model awaits elucidation of the structure of the membrane binding domain of 4.1R.
Recently, two homologues of 4.1R, designated 4.1G and 4.1N, have been identified and characterized (38,39). These proteins share high degrees of sequence similarity in different functional domains, such as the 30-kDa membrane binding domain, but exhibit different patterns of tissue expression. The two CaM binding domain sequences we have identified in 4.1R are highly conserved in the homologues of 4.1R (Fig. 6), suggesting that the interactions of these newly identified proteins with their membrane partners are likely to be regulated by Ca 2ϩ / CaM. In marked contrast, these CaM binding sequences are not conserved in other previously identified 4.1R related proteins such as ezrin (40,41), moesin (42), and radixin (43) (Fig.  6). These findings suggest that Ca 2ϩ /CaM regulation of protein-protein interactions is a distinct feature of 4.1R and its closely related homologues, and not a general feature of all members of the 4.1 superfamily.
4.1R and its homologues are present in a variety of nonerythroid cells and interact with a diverse group of transmembrane proteins in these cells. It is likely that these proteinprotein interactions may be regulated by Ca 2ϩ /CaM. For example, CD44, a transmembrane cell adhesion molecule, has recently been shown to bind to 4.1R in keratinocytes, and the interaction is modulated by Ca 2ϩ /CaM (33). After ligand binding, the cytosolic Ca 2ϩ concentration increases, activating many Ca 2ϩ -dependent cellular functions including those modulated by CaM. We suggest that Ca 2ϩ /CaM regulation of the interactions between transmembrane proteins and 4.1R and its new homologues may play a major role in cytoskeletal reorganization during cell signaling in erythroid and nonerythroid cells.