The Plant Plasma Membrane Ca2+ Pump ACA8 Contains Overlapping as Well as Physically Separated Autoinhibitory and Calmodulin-binding Domains*

In plant Ca2+ pumps belonging to the P2B subfamily of P-type ATPases, the N-terminal cytoplasmic domain is responsible for pump autoinhibition. Binding of calmodulin (CaM) to this region results in pump activation but the structural basis for CaM activation is still not clear. All residues in a putative CaM-binding domain (Arg43 to Lys68) were mutagenized and the resulting recombinant proteins were studied with respect to CaM binding and the activation state. The results demonstrate that (i) the binding site for CaM is overlapping with the autoinhibitory region and (ii) the autoinhibitory region comprises significantly fewer residues than the CaM-binding region. In a helical wheel projection of the CaM-binding domain, residues involved in autoinhibition cluster on one side of the helix, which is proposed to interact with an intramolecular receptor site in the pump. Residues influencing CaM negatively are situated on the other face of the helix, likely to face the cytosol, whereas residues controlling CaM binding positively are scattered throughout. We propose that early CaM recognition is mediated by the cytosolic face and that CaM subsequently competes with the intramolecular autoinhibitor in binding to the other face of the helix.

In plant Ca 2؉ pumps belonging to the P 2B subfamily of P-type ATPases, the N-terminal cytoplasmic domain is responsible for pump autoinhibition. Binding of calmodulin (CaM) to this region results in pump activation but the structural basis for CaM activation is still not clear. All residues in a putative CaM-binding domain (Arg 43 to Lys 68 ) were mutagenized and the resulting recombinant proteins were studied with respect to CaM binding and the activation state. The results demonstrate that (i) the binding site for CaM is overlapping with the autoinhibitory region and (ii) the autoinhibitory region comprises significantly fewer residues than the CaM-binding region. In a helical wheel projection of the CaM-binding domain, residues involved in autoinhibition cluster on one side of the helix, which is proposed to interact with an intramolecular receptor site in the pump. Residues influencing CaM negatively are situated on the other face of the helix, likely to face the cytosol, whereas residues controlling CaM binding positively are scattered throughout. We propose that early CaM recognition is mediated by the cytosolic face and that CaM subsequently competes with the intramolecular autoinhibitor in binding to the other face of the helix.
Ca 2ϩ acts as a secondary messenger in eukaryotic cells. One of the key proteins that mediate Ca 2ϩ signals is calmodulin (CaM) 2 a small, ubiquitous, and highly conserved protein found in all eukaryotes (1,2). Upon Ca 2ϩ binding, CaM changes conformation, which enables CaM to bind to and regulate a wide array of target enzymes. The interaction between CaM and the CaM-binding domain (CaMBD) of the target enzyme is mainly hydrophobic in which two bulky hydrophobic residues in the CaMBD, the so-called anchor points, are especially important to anchor the protein to CaM. In addition, the complex is stabilized by electrostatic interactions between negatively charged glutamates in CaM and basic residues in the CaMBD (3)(4)(5). No consensus CaMBD exists for proteins that are targets of CaM, but CaMBDs typically have a hydrophobic and basic nature consisting of 15-30 amino acid residues that have a tendency to form an ␣-helix. Based on the position of the two hydrophobic anchor points, the majority of Ca 2ϩ -dependent CaMBDs is divided into three classes, namely 1-10, 1-14, and 1-16 (6,7). Many CaM-regulated enzymes are autoinhibited with autoinhibition released by CaM. The autoinhibitory domain is often located either adjacent to or overlapping with the CaMBD (5,8).
The P 2B Ca 2ϩ -ATPases including PMCAs (plasma membrane Ca 2ϩ -ATPases) from animals and ACAs (autoinhibited Ca 2ϩ -ATPases) from plants are autoinhibitory proteins regulated by CaM. The regulatory region consisting of a CaMBD and an autoinhibitory domain is located in the C terminus in animals and N terminus in plants (9,10). These two domains are likely to be at least partly overlapping in both plant and mammalian Ca 2ϩ -ATPases, as has been shown for the human PMCA4b (isoform 4, splice variant b) (11), Arabidopsis ACA2 (isoform 2) (12,13), and ACA8 (14). In PMCA4b, two residues, Trp-1093 and Asp-1080, are involved in autoinhibition as well as CaM binding (11,(15)(16). In ACA2, a mutagenic study has identified 11 residues in the N-terminal CaMBD of ACA2 that, when substituted, cause pump deregulation (17). However, even though residues important for both autoinhibition and CaM binding in CaM-regulated autoinhibitory enzymes have been identified, little is still known on how autoinhibition is released by CaM.
In the present study, the mechanism behind CaM activation of an autoinhibited enzyme has been studied using the P 2B plasma membrane Ca 2ϩ -ATPase ACA8 from Arabidopsis (18). Both the autoinhibitory domain and the CaMBD have recently been localized to the N terminus of ACA8 (14,18,19). The CaMBD in ACA8 has been suggested to be localized between residues Ile-41 and Leu-66 (18,20). Alignment of this amino acid stretch with other known CaMBDs ( Fig. 1) indicates that ACA8 contains a CaM-binding motif belonging to the 1-14 class, in which the two hydrophobic residues, Trp-47 and Phe-60, could function as anchor points for CaM binding, or a 1-18 CaM binding motif like PMCA4b (21), with Trp-47 and Leu-64 as anchor points. Therefore, amino acid residues in the stretch from Arg-43 to Lys-68, comprising the putative CaMBD, were analyzed for their role in autoinhibition and/or CaM binding. It is demonstrated that residues involved in recognition of CaM and pump autoinhibition are separated in a helical wheel presentation of the CaMBD. This allows us to present a simple model for the mechanism by which CaM activates this autoinhibited Ca 2ϩ pump.

MATERIALS AND METHODS
Plasmid Constructs-Site-directed mutagenesis of ACA8 was performed by overlap extension polymerase chain reaction (22) to produce mutant genes encoding the individual alanine substitutions aca8R43A throughout to aca8L54A. In the case of Ala-50, Ala-51, and Ala-56, these residues were substituted with Ser. Three nucleic acid substitutions, G156A, G216C, and A217C, were introduced as silent mutations in the coding sequence of ACA8 to generate unique sites for the restric-tion enzymes SpeI and XmaI, respectively. This allowed for introduction of mutations by a cassette mutagenesis strategy to generate mutant genes encoding aca8N55A and aca8S57A throughout to aca8K68A. All mutations were verified by DNA sequencing. Wild type ACA8 and mutants were inserted into the yeast expression vector, pYES2 (Invitrogen), under control of a galactose inducible promoter. For localization purposes, a green fluorescent protein (GFP) was fused to the 3Ј ends of ACA8, ⌬74-aca8, and aca8W47A in the pYES2 vector just in front of the stop codon of the genes. ⌬74-aca8 does not encode the first 74 amino acid residues of the regulatory N terminus (19).
To produce recombinant N-terminal domain that could later be purified by NTA affinity chromatography, the coding sequences for the N termini (amino acid residue 1-143) of wild type and mutants of ACA8 were inserted into the Escherichia coli expression vector pET15b (Novagen) in this way fusing a (His) 6 tag to the N terminus of the peptide.
Complementation of the Yeast Mutant K616-Complementation tests were performed as previously described (25).
Localization of GFP Fusion Proteins in Yeast Cells-Ura ϩ yeast colonies from transformation of K616 with the GFP fusion proteins of ACA8, ⌬74-aca8, and aca8W47A, were inoculated in 2% (w/v) galactose (Gal), 50 mM succinic acid, pH 5.5, 0.7% (w/v) yeast nitrogen base, 40 g/ml adenine, and 10 mM CaCl 2 for 3 h at 30°C. The cells were pelleted, washed, and resuspended in H 2 O. Three l of a 1:1 mixture of yeast cells of A 600 ϭ 1 and 0.5% agarose were mounted on a polylysinecoated microscope slide (Teflon-printed multiwell slide). Fluorescence was measured with a confocal laser scanning microscope (Leica TCS SP2/MP; Leica Microsystems, Bernsheim, Germany) with GFP filter setting, excited at 488 nm, and emission recorded at 500 -540 nm.
Isolation of Yeast Membranes-Plant pumps were expressed in yeast grown in 1% (w/v) bacto-yeast extract, 2% (w/v) bacto-peptone, 2% (w/v) Gal, 40 g/ml adenine, and 10 mM CaCl 2 , for 24 h at 30°C. Microsomes were harvested as described (26) with slight modifications. The expression level of the Ca 2ϩ pumps in microsomes (40 g) was analyzed by Western blotting using a specific antibody against the N terminus of ACA8 (18).
ATPase Assay-ATPase activity was determined as described (27). Five g of membrane proteins from K616 expressing wild type or mutant ACA8 were assayed for 30 min at 30°C in the following four buffers: 50 mM KNO 3 , 5 mM NaN 3 , 0.25 mM sodium molybdate, 5 mM (NH 4 ) 2 SO 4 , 40 mM bis-Tris-Hepes, pH 7.5, 0.1 mg/ml Brij 58, 1 M A23187, 3 mM MgSO 4 , 3 mM ATP, 1 mM EGTA, with and without 20 g/ml bovine brain CaM (1.2 M; Sigma) and with and without CaCl 2 to give a free Ca 2ϩ concentration of 20 M. Free Ca 2ϩ concentrations were estimated by the program WEBMAXCLITE version 1.15 (28). Activities without Ca 2ϩ were subtracted from those with Ca 2ϩ to give the Ca 2ϩ -ATPase activity. Addition of CaM in the absence of Ca 2ϩ did not result in any change in ATP hydrolytic activity.
Fusion Protein Purification-Vectors expressing fusion proteins were expressed in E. coli strain BL21(DE3)pLysS (Novagen) by standard pro-cedures. Fusion proteins were purified from inclusion bodies on NTAagarose (Qiagen) and renaturated while bound to the affinity matrix by a procedure from Amersham Biosciences (product code 18-1134-37) with the following modifications. The pellet was disrupted by sonication in 20 mM Tris-HCl buffer. All buffers included 0.5 mM phenylmethylsulfonyl fluoride and 1 mM dithiothreitol instead of 2-mercaptoethanol. Finally, fusion proteins were eluted with 150 -200 mM imidazole in 20 mM Tris-HCl, 0.5 M NaCl, pH 8.0.
Protein Determination-Protein concentrations were determined by the method of Bradford (29). If not indicated otherwise, ␥-globulin was used as a protein standard.
CaM Overlay Assays-Two g of purified N termini were subjected to CaM overlay assays as described previously (25). As a protein control, 2 g of the purified N termini were examined by Western blotting using Penta-His antibody (Qiagen).
Surface Plasmon Resonance Spectroscopy-Surface plasmon resonance spectroscopy was achieved with a BIAcoreX optical biosensor instrument (Biacore AB, Uppsala, Sweden) using a NTA sensor chip (NTA bound to a solid support). Surface plasmon resonance buffers were as follows: (i) regeneration buffer: 10 mM Hepes, 0.15 M NaCl, 0.35 M EDTA, 0.005% (v/v) Surfactant P20, pH 7.4; (ii) immobilization buffer: 10 mM Hepes, 0.15 M NaCl, 0.005% (v/v) Surfactant 20, pH 7.4; (iii) nickel buffer: 500 M NiCl 2 in immobilization buffer; (iv) eluent buffer: 1 mM EGTA, 40 mM bis-Tris-Hepes, pH 7, 50 mM KCl, 0.1 mM ammonium molybdate, 5 mM (NH 4 ) 2 SO 4 , 0.1 mg/ml Brij 58, 3 mM MgSO 4 , 1 mM inosine 5Ј-triphosphate, and CaCl 2 to give a free Ca 2ϩ concentration of 1 M. All reactions were done at 25°C. The sensor chip contains two flow cells that are run in parallel of which one provides background corrections for nonspecific binding of CaM. After extensive washing with regeneration buffer followed by immobilization buffer, the NTA surface was saturated with Ni 2ϩ by loading nickel buffer into the two flow cells (1-min pulse at 5 l/min). The (His) 6 -tagged N termini were injected into one of the flow cells in immobilization buffer (20-min pulse at 5 l/min) until a resonance response unit of 250 -270 was obtained. The buffer was changed to eluent buffer and a concentration range (60 -600 nM) of bovine brain CaM (Sigma) in eluent buffer was injected into both flow cells (80 s pulse at 75 l/min). After the dissociation phase, the NTA chip was regenerated by injection of regeneration buffer (1-min pulse at 5 l/min for 3 times). The data were analyzed using BIAevaluation 3.0 software (Biacore AB) and kinetic analyses of primary sensorgrams were carried our by global fitting using a 1:1 Langmuir binding model.

Identification of Deregulated Ca 2ϩ Pumps by Functional Complementation of a Yeast
Mutant-In a number of autoinhibited proteins, such as CaM-activated pumps and kinases, the binding site for the activating protein overlaps with the autoinhibitory region of the molecule.  (20) is aligned with three known CaMBDs from a skeletal muscle myosin light chain kinase (skMLCK) (33), a CaM-dependent protein kinase I (CaMKI) (45), and a human plasma membrane Ca 2ϩ -ATPase (PMCA4b) (21). The alignment was made with ClustalW using standard parameters. The hydrophobic anchor points for skMCLK, CaMKI, and PMCA4b are marked in black and the putative anchor points for ACA8 are marked in gray.
To understand the mechanism of enzyme activation, a detailed characterization and comparison of residues involved in protein-protein interaction and autoinhibition is a prerequisite. Such an analysis has to our knowledge not been carried out for any autoinhibited protein activated by CaM. To investigate the relationship between autoinhibitory and CaM binding residues in a CaM-activated Ca 2ϩ pump we focused on ACA8, an Arabidopsis plasma membrane Ca 2ϩ -ATPase with a CaMbinding site at its N terminus (18).
First, a series of ACA8 mutants carrying amino acid substitutions in the presumed N-terminal CaMBD (Fig. 1) were studied for their ability to functionally complement a yeast strain (K616; Ref. 23) devoid of PMR1 and PMC1, the two Ca 2ϩ -ATPases present in this organism. It has previously been shown that a plant P 2B Ca 2ϩ -ATPase is not able to complement pmr1 and pmc1, unless the pump molecule is deregulated either by removal of the regulatory N terminus (12, 19, 25, 30 -32) or by point mutations in the regulatory region (17).
Mutant ACA8 pumps carrying Ala substitutions at six amino acid positions (Trp-47, Arg-48, Leu-52, Asn-55, Arg-58, and Phe-60) were able to support growth of K616 under calcium stress conditions (Fig. 2), indicating that mutations of these residues result in pumps that can functionally replace yeast Ca 2ϩ -ATPases.

Localization of Plant Ca 2ϩ -ATPases in Transgenic Yeast Cells-Yeast
Ca 2ϩ pumps PMC1 and PMR1 are localized to internal membranes (23). Thus, improved ability of ACA8 mutants to complement for pmc1 and pmr1 could have resulted from re-localization of the heterologous protein within the yeast cells. Indeed, whereas ACA8 is known to be expressed in the plasma membrane of plant cells (18), Bonza et al. (19) showed that ⌬74-aca8 expressed in yeast is enriched in a fraction containing membrane derived from the endoplasmic reticulum.
To localize the recombinant plant proteins in transgenic yeast cells, ACA8 and two mutants, ⌬74-aca8 and aca8W47A, were equipped with a GFP tag and expressed in yeast. Addition of the GFP tag to ACA8, ⌬74-aca8, and aca8W47A resulted in pumps that supported growth of K616 to the same degree as non-tagged versions (results not shown). GFP fluorescence in transgenic yeast cells was visualized by confocal laser scanning microscopy (Fig. 3). For all three pumps, fluorescence was intense in the periphery of cells and around the nucleus, indicating localization at the plasma membrane and endoplasmic reticulum, respectively. The pattern of GFP fluorescence was identical for all three plant pumps indicating that targeting of recombinant plant Ca 2ϩ -ATPase was not affected by the mutations.
Measurements of Ca 2ϩ Pump Activity -Increased ability of mutant plant Ca 2ϩ -ATPases to complement their yeast counterparts could be the result of differential expression. However, this possibility could be ruled out as all the pumps were expressed at similar levels (Fig. 4B). Alternatively, improved complementation could be related to increased specific activity of mutant pumps. To test this possibility, the ATP hydrolytic activity in yeast microsomal membranes expressing ACA8 and derived mutants were analyzed in the presence and absence of CaM.
Hardly any ATP hydrolytic activity could be measured for wild type ACA8 in Ca 2ϩ buffer without CaM (v ϭ 11 Ϯ 1.1 nmol of P i /min/mg of membrane protein; Table 1), whereas addition of CaM resulted in increased activity of the pump by ϳ20-fold (v ϭ 210 Ϯ 11.3 nmol of P i /min/mg of membrane protein; Table 1). N-terminal truncated ATPase (⌬74-aca8) had high basal activity that could not be stimulated further by CaM (Table 1), as has been shown for other P 2B pumps illustrating that truncation of the N terminus releases autoinhibition of pump activity.
Six mutants, W47A, R48A, L52A, N55A, R58A, and F60A showed increased activity in Ca 2ϩ buffer without CaM compared with wild type (v ϭ 48 -95 nmol of P i /min/mg of membrane protein, Table 1). The  activity of these mutants in Ca 2ϩ buffer was between 50 and 100% of the activity in the Ca 2ϩ /CaM buffer (Fig. 4A). The deregulated mutants were all able to functionally complement yeast pmc1 and pmr1 (Fig. 2). In addition to these six mutants, three mutants, L54A, Y62A, and L66A, also showed an increased activity in Ca 2ϩ -buffer (v ϭ 30 -66 nmol of P i /min/mg of membrane protein; Table 1) corresponding to 28 -35% of the CaM-stimulated activity. None of these three mutants, although, were able to functionally complement the yeast Ca 2ϩ -ATPases. This would indicate that a certain activity threshold is required for complementation of pmc1 and pmr1 to occur.
CaM Overlay Assays to Detect CaM Binding to Mutated ACA8 N Termini-After having identified residues in the putative CaMBD of importance for pump autoinhibition, we decided to map in detail residues of importance for CaM binding. For this purpose the N-terminal region of ACA8 comprising residues 1 to 143, as well as variants carrying single amino acid substitutions at every position in the stretch comprising Arg-43 and Lys-68, were expressed in E. coli as (His) 6 -tagged proteins. CaM binding was next examined by CaM overlay assays to purified, immobilized fusion proteins (Fig. 5). Binding to CaM was severely reduced in the case of four mutants (W47A, R48A, A56S, and F60A). Furthermore, eight mutants (A51S, V53A, L54A, R58A, R61A, L64A, K67A, and K68A) showed reduced binding of CaM compared with wild type. In all cases CaM binding was Ca 2ϩ dependent, as interaction with peptide was never observed in the absence of Ca 2ϩ (results not shown).
Binding Kinetics of CaM to Wild Type and Mutants of ACA8-Overlay assays suggested that the number of residues involved in binding of CaM (12 residues) was significantly higher than the number of residues involved in autoinhibition (6 residues). To confirm this finding and to analyze the observed protein-protein interactions in more detail, binding of CaM to the recombinant N termini of ACA8 was further analyzed by surface plasmon resonance spectroscopy. This gave us the association rate (k a ) and dissociation rate (k d ) for CaM binding to ACA8 N termini, and then the binding affinity (K d ) could be estimated as the ratio between the kinetic constants of the dissociation and association reaction.
Although binding kinetics were slightly altered for most mutants, four mutants, W47A, R48A, A56S, and F60A, showed a severely reduced CaM binding by surface plasmon resonance spectroscopy. In fact, it was not possible to estimate kinetic and equilibrium parameters for W47A and F60A, as the signal was too low, even at the highest CaM concentration tested (Fig. 6), whereas mutants R48A and A56S both exhibited a strongly reduced affinity for CaM (K d Ն 150 nM) compared with wild type ACA8 (K d ϭ 26 nM) ( Table 2). The reduced binding observed by surface plasmon spectroscopy was well in agreement with our results from the CaM overlay assays (Fig. 5).
In addition to these four mutants, we found that the binding affinity for CaM compared with wild type ACA8 was significantly decreased for the following nine mutants: L44A, A51S, V53A, L54A, R58A, R61A, L64A, K67A, and K68A (K d Ն 40 nM, indicating more than 50% reduced affinity). The reduced binding affinity primarily resulted from a faster dissociation of CaM from the N termini of these mutants (k d Ն 5.3 ϫ 10 Ϫ3 s Ϫ1 ) than from the N terminus of wild type ACA8 (k d ϭ 2.9 ϫ 10 Ϫ3 s Ϫ1 ) ( Table 2). Accordingly, these residues seem to be important for the stability of the complex between CaM and the CaMBD of ACA8. One of these nine residues, Val-53, may also be important for the recognition of CaM because its mutant exhibited an increase in its association rate toward CaM (ϳ50% compared with wild type ACA8). The effect of this increase in association rate on CaM affinity was counteracted by an increase of the dissociation rate of CaM ( Fig. 6 and Table 2).
In addition to V53A, three other mutants, R43A, Q46A, and K49A had a faster association of CaM (k a Ն 1.5 ϫ 10 5 M Ϫ1 s Ϫ1 ) than wild type ACA8. As a result of the increased association rate, the binding affinity of CaM to these three N termini was slightly increased (K d Յ 18 nM) compared with wild type ACA8. Among the 26 mutants investigated, FIGURE 5. CaM binding to wild type and mutated N termini of ACA8 measured by CaM overlay assay. Binding of biotinylated bovine brain CaM to purified peptides (2 g) derived from wild type and mutant ACA8 N termini was examined in a CaM overlay assay. As a control for protein loading, the same amount of proteins (quantified using bovine serum albumin as a protein standard) were subjected to Western blotting and immunodecorated using a Penta-His antibody (His). only substitution of Arg-43, Gln-46, Trp-47, Lys-49, Val-53, and Phe-60 significantly affected the association rate of the protein toward CaM, which would suggest that these residues are directly involved in CaM recognition. Two mutants, S57A and T63A, also showed a slight increase in CaM affinity (K d ϭ 12 and 17 nM, respectively), but this was caused by small alterations in both the association and dissociation rate ( Table 2), which did not seem to be notably different from wild type. Likewise, all the remaining mutants (Q45A, A50S, L52A, N55A, R59A, Y62A, D65A, and L66A) had reaction kinetics similar to those of wild type resulting in K d values ranging between 20 and 38 nM.
Mutants that demonstrated reduced CaM binding affinity by surface plasmon resonance spectroscopy likewise showed reduced interaction in CaM overlay assays, except for the mutant L44A. This mutant showed a slight decreased CaM binding affinity (K d ϭ 46 nM) by surface plasmon resonance but apparently similar CaM binding as wild type ACA8 by the CaM overlay assay. Similarly, in the CaM overlay assay it was not possible to see any noteworthy differences in CaM binding to mutants R43A, Q46A, K49A, and S57A, although they showed a small increase in CaM binding affinity by surface plasmon resonance spectroscopy. These differences probably reflect a reduced sensitivity of CaM overlay assay compared with surface plasmon resonance spectroscopy.
Correlation of CaM Stimulation with CaM Binding-It is worth noting that all mutants, with the exception of R58A, were stimulated by CaM in the ATPase assays (Table 1). Even the four mutants, W47A, R48A, A56S, and F60A, that exhibited extremely low affinity for CaM were stimulated by CaM (Table 1, Fig. 4A). In the soybean Ca 2ϩ -ATPase, SCA1, two CaMBDs have been identified in the N-terminal domain (31). We therefore tested the possibility that ACA8 could contain an additional CaMBD downstream of Ala-143 and therefore not present in the (His) 6 -tagged N termini of ACA8. CaM overlay assays on yeast microsomes expressing full-length pumps were performed and the results were similar to those of the CaM overlay assays on the N termini. Thus, only an extremely weak binding of CaM could be observed for W47A, R48A, F60A, and A56S (results not shown). No binding of CaM was observed for ⌬74-aca8 arguing against a second CaMBD outside this region in ACA8. CaM was found to activate the wild type pump with a half-maximal concentration of 30 nM (results not shown), which explains why 1.2 M CaM is sufficient to cause at least partial activation even of low affinity mutants.

DISCUSSION
Features of the CaM-binding Domain of ACA8-In this work all residues in the presumed CaMBD of ACA8 (Arg-43 to Lys-68) were substituted to investigate the functional consequences for CaM binding and pump autoinhibition.
Mutations of Trp-47 and Phe-60 to alanines both resulted in severely reduced affinity for CaM supporting the hypothesis that these residues function as anchor points in the CaMBD of ACA8. Trp-47 and Phe-60 are conserved in the Arabidopsis subfamily of P 2B Ca 2ϩ -ATPases and mutation of these two residues in the closely related ACA9 also results in abolished binding of CaM. 3 Phylogenetically, ACA8, ACA9, and ACA10 form a small subgroup of Ca 2ϩ -ATPases in which Trp and Phe are spaced by 12 amino acid residues as in common 1-14 CaM binding motifs. In the remaining Arabidopsis P 2B pumps, two hydrophobic and bulky putative anchor points are spaced by 13 residues suggesting the presence of a novel 1-15 CaM-binding motif. The human plasma membrane Ca 2ϩ -ATPase, PMCA4b, has a CaM binding motif in which two presumed anchor points are spaced by 16 amino acid residues (21). These differences illustrate the diversity of CaMBDs even within the same subfamily of P-type ATPases.
Besides Trp-47 and Phe-60, 11 other residues (Leu-44, Arg-48, Ala-51, Val-53, Leu-54, Ala-56, Arg-58, Arg-61, Leu-64, Lys-67, and Lys-68) were found to be important for CaM binding by stabilizing the complex  between CaM and the CaMBD of ACA8. Thus, substitution of these residues resulted in decreased binding affinity of CaM primarily because of an increased dissociation rate (Fig. 5 and Table 2). Common for these 13 residues is that they are either hydrophobic or basic. The importance of hydrophobic or basic amino acid residues in CaM binding has been revealed in several structures of complexes between CaM and a diverse group of CaMBDs (33)(34)(35), and hydrophobic and electrostatic interactions between CaM and the CaMBD of the target proteins have become some of the main characteristics for the complexes.
CaM is suggested to bind to the CaMBD of the target enzyme through a multistep binding mode, in which CaM initially binds to the N-terminal hydrophobic anchor point that is exposed to the solvent and subsequently CaM can bind to the C-terminal end of CaMBD (21,36,37). The initial binding of CaM only to the N-terminal end of the CaMBD of PMCA4b has been suggested to take place at low Ca 2ϩ concentrations making the enzyme ready to react rapidly to an elevated cytosolic Ca 2ϩ concentration caused by a Ca 2ϩ signal (38 -40). In ACA8, the C-terminal anchor point (Phe-60) appears to be important as well, because substitution with Ala resulted in a marked decrease of CaM affinity. Furthermore, four residues, Arg-43, Gln-46, Lys-49, and Val-53, could be important for negative regulation of CaM recognition, as Ala substitutions result in a considerable increase in the association rate of CaM compared with the wild type. Similar changes were not observed for any of the other mutants. This would suggest that the role of these amino acid residues is to slow down the rate of CaM binding.
The Autoinhibitory Domain Is Overlapping with the CaM-binding Domain-The mutations made in the CaMBD of ACA8 were also analyzed for their role in autoinhibition of the pump, as the CaMBD and autoinhibitory domain have been shown to be at least partially overlapping in Ca 2ϩ -ATPases of the P 2B subfamily of both animals and plants (11)(12)(13)(14). Six mutants, W47A, R48A, L52A, N55A, R58A, and F60A, did have increased CaM-independent activity compared with the wild type pump (Figs. 2 and 4). This confirms that the CaMBD and autoinhibitory domain also are overlapping in ACA8, indicating that this could be a common feature of P 2B Ca 2ϩ -ATPases. Not only are the CaM-binding and autoinhibitory domains overlapping, they also involve identical residues such as Trp-47, Arg-48, Arg-58, and Phe-60.
In this study, we identified both hydrophilic, basic and non-charged, as well as hydrophobic residues to be involved in autoinhibition. Two hydrophobic residues involved in autoinhibition represent the two conserved anchor points Trp-47 and Phe-60. Hydrophobic CaM anchor point residues are also involved in autoinhibition of a related Arabidopsis Ca 2ϩ pump, ACA9 (Trp-61 and Phe-74) 3 and human PMCA4b (Trp-1093) (15,42). Likewise, in CaM-dependent protein kinase I, mutagenic and structural studies have revealed that only hydrophobic residues are involved in autoinhibition (36,41). However, residues in the CaMBD of the Arabidopsis Ca 2ϩ pump ACA2, that when substituted with Ala result in pump deregulation, were all basic (17), suggesting that a positively charged surface area could be an important feature of the autoinhibitory interaction. Thus the mode of interaction between the autoinhibitory domain and the catalytic core is not conserved among different kinds of CaM-regulated enzymes, not even in enzymes belonging to the same subfamily of P-type ATPases.
Amino Acid Residues Involved in Autoinhibition or Recognition of CaM Are Separated in an ␣-Helical Representation of the CaM-binding Domain of ACA8-It is a characteristic for CaMBDs that they have a tendency to attain an ␣-helical conformation (33,44,45). Because a peptide derived from the CaMBD of cauliflower BCA1, a CaM-activated Ca 2ϩ -ATPase closely related to ACA8, has an ␣-helical conformation when it binds to different CaM isoforms (46) it seems reasonable to assume that the CaMBD of ACA8 folds as an ␣-helix as well. In many target proteins, the CaMBD is unstructured until it binds CaM (7), however, this does not seem to apply to ACA8, because when ACA8 was denatured, renaturation increased ϳ10-fold the degree of CaM binding (results not shown). Thus it is likely that the CaMBD of ACA8 forms an ordered structure before binding of CaM.
When the amino acid stretch representing Arg-43 to Lys-68 in ACA8 was inserted into a helical wheel (Fig. 7A), residues involved in autoinhibition assembled mainly on one side of the ␣-helix. Likewise, in ACA2, the amino acid residues that have been shown to be involved in autoinhibition by alanine substitutions also cluster to only one side of an ␣-helix that represents the CaMBD of the pump (Fig. 7B). In contrast, the amino acid residues important for CaM binding to ACA8 were scattered all around the ␣-helix. Four of these residues (Trp-47, Arg-48, Ala-56, and Phe-60) that exhibited the most severe reduction of CaM affinity when substituted were located to the same side in the ␣-helix as the autoinhibitory residues. However, the four residues (Arg-43, Gln-46, Lys-49, and Val-53) that are involved in negative regulation of CaM recognition were located on the ␣-helical side, nearly opposite to the side with the autoinhibitory residues (Fig. 7A).
Based on these findings, we suggest a model on how CaM binds to and activates an autoinhibited Ca 2ϩ pump, such as ACA8. In the resting state, the Ca 2ϩ pump is kept in its autoinhibited form. This is mediated by the ␣-helical CaMBD but only by residues situated on one side of the ␣-helix that make interactions to an intramolecular receptor in the pump. In ACA8, an intramolecular site of interaction for the N terminus has been identified in the small cytoplasmic loop (14). The other side of the ␣-helix, presumed to be exposed to the cytosol, contains residues with low affinity to CaM preventing CaM binding at low cytosolic Ca 2ϩ concentration. Following a rise in cytosolic Ca 2ϩ concentration, CaM FIGURE 7. Helical wheel projection of the CaMbinding domains of ACA8 and ACA2. A helical wheel projection was made using the program Binding Site Analysis in The Calmodulin Target Data base (7) on: A, the amino acid stretch from Arg-43 to Lys-68 in ACA8; and B, the amino acid stretch Val-20 to Lys-46 from ACA2 that corresponds to the CaMBD of ACA8. Residues important for CaM binding are shown in gray of which the four most important residues are underlined. Residues important for CaM recognition are boxed and residues involved in autoinhibition are circled. In ACA2, only autoinhibitory residues (17) are circled as residues important for interaction between CaM and ACA2 have not been identified. makes contact with anchor point residues in the CaMBD of ACA8. These are situated on the edges of the ␣-helical face exposing residues involved in negative regulation of CaM binding. Initial binding is followed by interaction with the other face of the ␣-helix in competition with the intramolecular receptor in the pump. When binding to both sides of the ␣-helix is achieved, the N terminus is released from the intramolecular receptor and the pump is activated.
This model only includes autoinhibitory residues identified in the CaMBD of ACA8, although additional autoinhibitory residues may exist in the N terminus outside the CaMBD (14,17). Furthermore, the model only describes how CaM can activate an autoinhibited enzyme, although other regulatory mechanisms may exist. For instance, several studies have shown that phosphorylation in or close to the CaMBD of target enzymes affect the CaM binding and/or the basal activity of the enzyme depending on the nature of the protein kinase and its target (37,43,(47)(48)(49)(50). This would add further complexity to regulation of the CaM target, but would not affect the basic principle of protein regulation.