Arabidopsis calmodulin-like protein CML36 is a calcium (Ca2+) sensor that interacts with the plasma membrane Ca2+-ATPase isoform ACA8 and stimulates its activity

Calmodulin-like (CML) proteins are major EF-hand–containing, calcium (Ca2+)–binding proteins with crucial roles in plant development and in coordinating plant stress tolerance. Given their abundance in plants, the properties of Ca2+ sensors and identification of novel target proteins of CMLs deserve special attention. To this end, we recombinantly produced and biochemically characterized CML36 from Arabidopsis thaliana. We analyzed Ca2+ and Mg2+ binding to the individual EF-hands, observed metal-induced conformational changes, and identified a physiologically relevant target. CML36 possesses two high-affinity Ca2+/Mg2+ mixed binding sites and two low-affinity Ca2+-specific sites. Binding of Ca2+ induced an increase in the α-helical content and a conformational change that lead to the exposure of hydrophobic regions responsible for target protein recognition. Cation binding, either Ca2+ or Mg2+, stabilized the secondary and tertiary structures of CML36, guiding a large structural transition from a molten globule apo-state to a compact holoconformation. Importantly, through in vitro binding and activity assays, we showed that CML36 interacts directly with the regulative N terminus of the Arabidopsis plasma membrane Ca2+-ATPase isoform 8 (ACA8) and that this interaction stimulates ACA8 activity. Gene expression analysis revealed that CML36 and ACA8 are co-expressed mainly in inflorescences. Collectively, our results support a role for CML36 as a Ca2+ sensor that binds to and modulates ACA8, uncovering a possible involvement of the CML protein family in the modulation of plant-autoinhibited Ca2+ pumps.


Calcium (Ca 2ϩ ) is a crucial second messenger in plants,
where it couples the perception of endogenous and environmental signals to plant responses (1)(2)(3)(4). Ca 2ϩ signals are transmitted by stimulus-specific cytosolic Ca 2ϩ elevations that result from the concerted action of both Ca 2ϩ influx (channels) and Ca 2ϩ efflux (pumps and carriers) systems, which temporally shape and spatially define Ca 2ϩ dynamics, commonly referred to as the "Ca 2ϩ signature" (3)(4)(5)(6). These signatures are believed to encode information from primary stimuli and, therefore, to contribute to the stimulus specificity of the biological response. Nonetheless, another layer of specificity in Ca 2ϩ signaling is achieved via the function of Ca 2ϩ sensors, which detect alterations in intracellular free Ca 2ϩ concentration decoding specific Ca 2ϩ signatures into downstream physiological responses (7). Most Ca 2ϩ sensors harbor the EF-hand motif, a helix-loop-helix structure binding one Ca 2ϩ ion. The EF-hand binds Ca 2ϩ through a pentagonal bipyramidal arrangement, and the chelating residues are notated either based on their linear positions (1, 3, 5, 7, 9, and 12) or on their tertiary coordination geometry (X, Y, Z, ϪY, ϪX, and ϪZ) (8). In the canonical EF-hand, positions 1 (X), 3 (Y), 5 (Z), and 12 (ϪZ) are side-chain oxygen ligands, 7 (ϪY) is a backbone carbonyl ligand, and 9 (ϪX) is a water ligand hydrogen-bonded to one loop residue (8). Ca 2ϩ binding modifies the conformation of EF-hand proteins resulting in changes in enzymatic activity of the Ca 2ϩ sensor itself or allowing it to interact with and alter the target protein activity (9). Besides calmodulin (CaM), 2 which has been wellconserved during evolution, plants contain a unique family (50 members in the model plant Arabidopsis thaliana) of CaM-like proteins (CMLs) that greatly differ from canonical CaMs in sequence (they share at least 16% overall amino acid identity with CaM2 of Arabidopsis), length (some CMLs possess N-or C-terminal extensions), and number of EF-hands (from 1 to 6) (10 -12). Functional analyses of various plant CMLs have provided clear indications for the key roles of these proteins in plant development and abiotic and biotic stress responses, as overviewed in several recent reviews (12)(13)(14)(15). Altogether, these studies do not support a functional redundancy hypothesis for CMLs, which could be key positive and/or negative regulator components in the coordination of plant responses to different environmental stimuli (15). However, to date only a few of these proteins have been clearly shown to function as Ca 2ϩ sensors by describing their Ca 2ϩ -binding (e.g. dissociation constants, stoichiometry, ligand specificity) and structural properties and by identifying their biological targets. Thus, a detailed characterization of CMLs at the biochemical and structural levels combined with the identification of distinct downstream targets is essential in determining the high level of specificity that allows the plant cell to transduce diverse stimuli into proper physiological responses.
In the present study, we characterized CML36 from A. thaliana. In particular, using different complementary biophysical and biochemical methods, we analyzed whether CML36 possesses the structural and binding properties of Ca 2ϩ sensor proteins. We found that CML36 behaves as a Ca 2ϩ sensor, in that it experiences Ca 2ϩ -dependent changes in secondary and tertiary structure and in the exposure of hydrophobic regions. Moreover, we identified a biological target, providing evidence that CML36 interacts directly with Arabidopsis Ca 2ϩ -ATPase isoform 8 (ACA8), a type IIB Ca 2ϩ pump localized at the plasma membrane (PM) that contributes to the overall Ca 2ϩ homoeostasis and to the control of intracellular Ca 2ϩ signaling by actively extruding Ca 2ϩ from the cytosol into the apoplast (16 -18). We identified the specific CML36-binding site at the N terminus of ACA8 and demonstrated that the interaction between ACA8 and CML36 promotes the pump Ca 2ϩ -dependent hydrolytic activity in vitro. Collectively, our findings allowed us to classify CML36 as a Ca 2ϩ sensor, uncovering a possible involvement of CML36 in the modulation of the plant autoinhibited Ca 2ϩ pump.

CML36 is a Ca 2؉ sensor
Recombinant production of full-length CML36 (CML36 f.l.) resulted in a highly unstable protein (with a high propensity to aggregate and precipitate) at the concentrations required for efficient protein-based biophysical and structural analyses. CML36 contains an extended N-terminal region (residues 1-60) and a Ca 2ϩ -binding region (residues 61-209) with four EF-hand motifs. The extended N-terminal region was found to be remarkably intrinsically disordered using the GeneSilico MetaDisorder Service (19), and therefore a truncated variant of CML36 was created by deleting the first 60 residues. The resulting mutant (residues 61-209, named CML36-C) was very stable even at high concentrations and was therefore used for all biophysical and structural analyses.
Typical features of Ca 2ϩ sensors include Ca 2ϩ -induced conformational changes that often expose hydrophobic surfaces necessary for target recognition. Thus, we examined Ca 2ϩ -dependent conformational changes and surface hydrophobicity of recombinant CML36-C by using an array of biophysical and structural techniques such as nuclear magnetic resonance (NMR), circular dichroism (CD), 1-anilino-8-naphthalene sulfonate (ANS) fluorescence spectroscopy, and size-exclusion chromatography (SEC). Moreover, because of the high concentration of free Mg 2ϩ in plant cells (0.5-2 mM) (20,21) we also analyzed the behavior of CML36-C upon the addition of Mg 2ϩ , to determine the specificity of CML36 in discriminating cytosolic Ca 2ϩ signals against a ϳ10 2 -10 4 -fold excess of the chemically similar divalent cation Mg 2ϩ . Mg 2ϩ binding to EF-hands is physiologically important; distinct roles have been proposed for the binding of this ion (8,22), such as a structural role, by conferring structural stability to an otherwise poorly defined molten globule apo-state, or roles in the Ca 2ϩ -dependent regulation of cellular processes by modulating the affinity of EF-hands for Ca 2ϩ .
NMR spectroscopy was used to investigate the structural rearrangements of CML36-C upon the addition of metal ions. The signals in a two-dimensional 1 H-15 N heteronuclear singlequantum coherence (HSQC) NMR spectrum belong to the HN groups of the protein. The frequency of HN signals (chemical shift) correlates with the chemical environment of the observed nuclei and is influenced by both local and global structure.
The 1 H-15 N HSQC spectrum of apoCML36-C is characterized by the presence of fewer peaks than expected, severe line broadening, and poor chemical shift dispersion, especially in the proton dimension (Fig. 1A). These spectral features suggest that apoCML36-C undergoes conformational fluctuations typical of a molten globule state, similar to other EF-hand proteins in their apoforms (23)(24)(25). A change in the HSQC spectrum of CML36-C was observed upon the addition of a saturating concentration of Mg 2ϩ (Fig. 1B). The spectrum is characterized by more uniform peak intensities and an increased chemical shift dispersion, indicating that Mg 2ϩ -bound CML36-C adopts a more stable tertiary structure compared with the apoprotein. However, the number of peaks in the Mg 2ϩ -bound form remained much lower than expected considering the non-proline residues in the amino acid sequence of CML36-C (150 residues), and many peaks in the central area of the HSQC spectrum still appeared rather broad, meaning that some portions of the Mg 2ϩ -bound protein may not be structured or may still possess a high degree of flexibility. Remarkably, the addition of saturating Ca 2ϩ concentrations to the apoCML36-C resulted in a more dramatic change in the CML36-C HSQC spectrum, where ϳ120 peaks with good chemical shift dispersion became visible (Fig. 1C), indicating that the addition of Ca 2ϩ promotes a global structural rearrangement throughout the protein. By comparing the position of the peaks in the two metal-bound forms, it was evident that Ca 2ϩ induced conformational changes in CML36-C distinct and separate from those induced by Mg 2ϩ , which is in line with a putative Ca 2ϩ sensor function of the protein.
The far-and near-UV CD analyses were consistent with the interpretation of the apoCML36-C as a molten globule state, as in apoCML36-C the secondary structure is conserved to a large extent, whereas the tertiary structure is highly fluctuating. Panels A and B of Fig. 2 show the far-and near-UV regions of the CD spectra of apoCML36-C and Ca 2ϩ -or Mg 2ϩ -bound CML36-C, respectively. As evident in Fig. 2A, the far-UV CD spectra for the apoform and both holoforms of CML36-C show two negative bands with minima at 208 and 222 nm, characteristic of proteins with a high ␣-helical content. However, the reduced 222/208 ratio observed in apoCML36-C (0.67 for apoCML36-C versus 0.83 and 0.82 for Ca 2ϩ -CML36-C and Mg 2ϩ -CML36-C, respectively) suggests the presence of some regions of high flexibility or random coil structures.
Despite the absence of Trp residues, CML36-C has two Tyr and eight Phe residues, and the corresponding bands are wellresolved in the near-UV CD spectra (Fig. 2B). Signals in the 275-290-nm range can be associated with the microenvironment of Tyr residues, whereas the bands observed in the 250 -270-nm region, with minima at 262 and 268 nm, are ascribable to Phe residues. The switch of CML36-C from the holo-to the apoform produces considerable changes in this part of the CD spectrum ( Fig. 2B), suggesting a partial collapse of the tertiary structure.
A common feature of Ca 2ϩ sensors is the increase in surfaceexposed hydrophobicity upon Ca 2ϩ binding, which we investigated by the use of the fluorescent probe ANS. This dye is a hydrophobic compound that shows a blue-shift in the maximum emission and a considerable increase in fluorescence upon binding to hydrophobic regions of proteins. In the presence of apoCML36-C, we detected a ϳ7-nm blue-shift and a modest increase in the emission of ANS in comparison with ANS alone (Fig. 2C), suggesting that apoCML36-C possesses exposed hydrophobic portions. The addition of Mg 2ϩ did not cause further changes to the spectrum, whereas upon the addition of Ca 2ϩ a further blue-shift (ϳ12 nm) and a 1.6-fold increase in fluorescence emission was observed, indicating that CML36-C undergoes a Ca 2ϩ -dependent exposure of hydrophobic surfaces.
Analytical SEC, together with pulsed-field gradient diffusion NMR studies, was conducted to examine the hydrodynamic radius (R h ) of CML36-C in the apo-state and the Mg 2ϩ -bound

CML36 binds to and activates ACA8
and Ca 2ϩ -bound states (Table 1). Notably, in all of these states, the obtained R h resulted in an apparent molecular mass overestimation, which has been observed in various Ca 2ϩ sensor proteins (26 -28) and is related to their extended conformation. Moreover, the binding of Ca 2ϩ , and to a less extent of Mg 2ϩ , to CML36-C determined a decrease in R h , as found elsewhere for CaMs (26,27,29).
Altogether, these structural studies strongly suggest that CML36-C functions as Ca 2ϩ sensor, indicating that apo-CML36-C is in a dynamic molten globule form that undergoes a large conformational change upon binding Ca 2ϩ . Moreover, CML36-C displays a higher electrophoretic mobility in the presence of Ca 2ϩ both in denaturing and native conditions (Fig.  2, D and E), which is a well-documented phenomenon for CaM proteins (29 -32).

CML36 contains four Ca 2؉ -binding sites
Isothermal titration calorimetry (ITC) was used to monitor Ca 2ϩ and Mg 2ϩ binding to CML36-C, which allows the determination of accurate values of apparent dissociation constants (K d ) for multiple binding sites and related changes in enthalpy (⌬H). Representative ITC isotherms are shown in Fig. 3, and the thermodynamic binding parameters obtained are listed in Table 2. Titration of CaCl 2 into apoCML36-C at physiological salt concentrations (Fig. 3A, black line) produced a multiphasic binding isotherm with an initial exothermic phase followed by an endothermic phase. The data best fit a four-site sequential binding model that reflected the presence of two classes of binding sites (Table 2): the first class comprises two high-affin-

Table 1 Determination of hydrodynamic radius of apo-and holoCML36-C by pulsed-field gradient diffusion NMR analysis and SEC
The mean values Ϯ S.E. from triplicate experiments are presented. D, diffusion coefficient; R h , hydrodynamic radius. Table 2 Thermodynamics of Ca 2؉ binding to CML36-C in the absence and presence of Mg 2؉

Pulsed-field gradient diffusion NMR
The mean values Ϯ S.E. from triplicate experiments using at least two different CML36-C preparations are presented.

CML36 binds to and activates ACA8
ity sites (K d1 ϭ 91 Ϯ 17 nM and K d2 ϭ 122 Ϯ 13 nM, respectively) with a heat release between Ϫ12 and Ϫ10 kcal/mol of CaCl 2 ; and the second class likely represents the binding of two Ca 2ϩ ions with lower affinity (K d3 ϭ 104 Ϯ 9 M and K d4 ϭ 345 Ϯ 24 M) with ⌬H ϳ 3 kcal/mol of CaCl 2 . The Ca 2ϩ titration was also performed in the presence of 5 mM Mg 2ϩ (Fig. 3A, red line, and Table 2). Mg 2ϩ did not greatly affect the affinity constants associated with Ca 2ϩ binding to sites 3 and 4 of CML36-C, indicating no binding of Mg 2ϩ to these sites, whereas the K d values of sites 1 and 2 were shifted to lower affinity. Moreover, Mg 2ϩ had an effect on the ⌬H of sites 1 and 2, as the process was less exothermic than that without Mg 2ϩ . This behavior, likely suggesting that in the presence of Mg 2ϩ these two sites could be formed and structurally stable, prompted us to quantitate the direct binding of Mg 2ϩ to CML36-C. Titration of MgCl 2 into apoCML36-C resulted in an exothermic binding isotherm ( Fig.  3B), giving the best fit with the binding of two Mg 2ϩ ions to protein with a K a value of 9.2E3 Ϯ 0.7E3 M Ϫ1 (which corresponds to a K d value of 109 Ϯ 8 M), an enthalpy ⌬H of Ϫ3.9 Ϯ 0.7 kcal mol Ϫ1 . The stoichiometry of Ca 2ϩ and Mg 2ϩ binding was further confirmed by analysis of the 1 H-15 N HSQC NMR spectra of CML36-C. Downfield-shifted NMR peaks at ϳ10.5 ppm are typical of conserved glycines at position 6 of the EF-hands occupied by divalent metals (25,33,34). In Ca 2ϩ -saturated CML36-C, the presence of four downfield-shifted peaks corresponding to Gly-84 (EF-1), Gly-121 (EF-2), Gly-157 (EF-3), and Gly-194 (EF-4) confirms that Ca 2ϩ is bound to the four EFhands, as seen by ITC analysis (Fig. 4A). The addition of saturating amounts of Mg 2ϩ to apoCML36-C caused the appearance of two downfield-shifted NMR peaks, assigned to Gly-157 and Gly-194 (Fig. 4B, green), indicating that Mg 2ϩ is bound to EF-3 and EF-4. Notably, the addition of Ca 2ϩ to Mg 2ϩ -bound CML36-C caused the two downfield peaks at 10.55 ppm (Gly-157) and 10.45 ppm (Gly-194) to disappear and the appearance of the four glycine peaks in the position typical of the Ca 2ϩbound protein (Fig. 4B, red), suggesting that the Ca 2ϩ ion could displace Mg 2ϩ from EF-3 and EF-4.

CML36 interacts with ACA8 and stimulates its Ca 2؉ -dependent hydrolytic activity
To fully understand the roles of Ca 2ϩ sensors, the identification of physiologically relevant targets is essential. A bioinformatics search for possible interactors of CML36 using string database (http://string-db.org/), which gives extensive coverage and ease of access to both experimental and predicted interaction information (35), found several promising candidates with various confidence scores. Among all of the putative predicted interactors, we decided to focus on the PM-localized type IIB Ca 2ϩ -ATPase isoform ACA8 because: (i) it is a prominent in vivo regulator of cellular Ca 2ϩ dynamics (18); (ii) CML36 transcript has been found to be altered in an Arabidopsis double mutant lacking both ACA8 and ACA10, another PMlocalized Ca 2ϩ -ATPase isoform (36); and (iii) in a phosphoproteomic study, CML36 has been shown to be associated with the PM fraction, which might facilitate the interaction with ACA8 in vivo (37). ACA8 possesses an extended cytosolic N-terminal regulatory domain with autoinhibitory function (16, 38 -40). The best-known regulator of ACA8 is CaM. CaM binding to the N terminus induces a conformational change that suppresses autoinhibition, thus determining the activation of the pump. The N terminus is the target of modulation not only by CaM but also by other regulators of the pump, such as acidic phospholipids, which counteract its autoinhibitory action (41). In addition, the phosphoregulation of N-terminal serine residues by influencing autoinhibition (42) is associated with alterations of the intracellular Ca 2ϩ dynamics in response to mechanical wounding (18). As a first step toward experimental validation of the putative interaction between CML36 and ACA8, we performed an in vitro overlay assay by spotting increasing amounts of recombinant CML36-C onto nitrocellulose and testing its ability to bind purified ACA8. We used bovine CaM and lysozyme as the positive and negative controls, respectively. Moreover, we tested another member of the Arabidopsis CML family, CML14 (43), which was not predicted to interact with ACA8. Specific interaction of ACA8 with CML36-C and CaM was evident (Fig. 5A). A signal of interaction with CML14 was

CML36 binds to and activates ACA8
visible, but a similar signal intensity was also observed when the nitrocellulose was incubated in the absence of ACA8 (Fig. 5A), strongly suggesting nonspecific binding of the anti-ACA8 antibody to CML14.
The specificity of the ACA8-CML36 interaction was also supported by native PAGE analysis (Fig. 5B). We produced in recombinant form the N-terminal portion of ACA8 corresponding to the regulatory domain (residues 1-116, named ACA8-N), and we checked the interaction between ACA8-N and both CML36-C and CML36 f.l. Recombinant Arabidopsis CaM1 (AtCaM1) was used as positive control and two other CMLs, i.e. CML14 and CML19, as negative controls. When ACA8-N and CML36 (both CML36-C and CML36 f.l. variants) were mixed in the presence of Ca 2ϩ , an additional band appeared on the native gel, likely representing the CML36-ACA8-N complex. An extra band was also visible for the AtCaM1-ACA8-N complex, whereas it was missing in the CML19-and CML14-containing samples, indicating the inability of these two CML isoforms to form a complex with ACA8-N. These results clearly confirmed the ability of CML36 to interact with ACA8 and localized the CML-binding domain to a region in the N terminus of the enzyme.
To elucidate the relevance of the CML36-ACA8 interaction, we subsequently examined its effect on ACA8 activity by measuring, in the presence of increasing concentrations of Ca 2ϩbound CML36, the hydrolytic activity of ACA8 overexpressed in the yeast triple mutant K616, which is defective in highaffinity endogenous Ca 2ϩ pumps (38,40,44). Fig. 6A shows that the Ca 2ϩ -dependent hydrolytic activity of ACA8 was significantly stimulated by CML36-C in a concentration-dependent manner. The concentration of CML36-C giving half-maximal stimulation of ACA8 activity (K 0.5 ) was 534 Ϯ 42 nM. The activation of the Ca 2ϩ -ATPase was specific for CML36-C, as the addition of similar concentrations of two other CMLs, CML14 (Fig. 6A) and CML19 (data not shown), was completely ineffective. To exclude the possibility that the CML36-C might lack portions with a potential regulatory role, we also performed the same experiment with CML36 f.l. The results showed that the ACA8 activation curve in the presence of increasing concentrations of CML36 f.l. was comparable to that obtained using the truncated variant, CML36-C (supplemental Fig. S1).
The effect of CML36 on ACA8 activity was further evaluated in the presence of saturating CaM. As expected, the enzymatic activity of ACA8 was increased by CaM (42,45). The addition of increasing concentrations of CML36-C did not influence CaMstimulated enzymatic activity (Fig. 6A), revealing that the two effects were not additive. Thus, in accordance with the native PAGE results, we determined that CML36 and CaM interact within the N-terminal region of ACA8. Consistently, the response to CML36-C of the two N-terminal truncated mutants of ACA8 (⌬74ACA8 and ⌬109ACA8, from which the high-affinity and both the high-and low-affinity CaM-binding sites, respectively, have been deleted (38,46)) was completely suppressed (Fig. 6B).
Because the above-described experiments were carried out on yeast microsomes overexpressing ACA8, to further corroborate our results we evaluated the response to CML36-C of ACA8 following a one-step purification on CaM-Sepharose. Fig. 6C shows that, under the experimental conditions used, the ACA8 activation curve resembled that obtained using the microsomal fraction, with an estimated K 0.5 value of 797 Ϯ 168 nM. Moreover, we confirmed the specificity of the CML36-C effect, as no ACA8 activation was detected in the presence of the other two CMLs, i.e. CML14 (Fig. 6C) and CML19 (data not shown). As shown in Fig. 6A, ACA8 was not further stimulated by CML36-C when activated by saturating CaM. We then tested the ability of 4.2 M CML36-C to stimulate the ATPase activity in the presence of increasing, but subsaturating, CaM concentrations. The results, reported in Fig. 6D, indicate an additive effect of CML36-C and CaM that was evident only in the presence of low CaM concentrations. This finding is in accordance with competition between the two activating proteins.

Expression pattern analysis of CML36 in different Arabidopsis organs
Quantitative RT-PCR performed on adult Arabidopsis plants indicated that the CML36 gene is ubiquitously expressed but is

CML36 binds to and activates ACA8
more abundant in roots and inflorescences ( Fig. 7A and supplemental Fig. S2A). This expression pattern showed a similar trend when either actin (Fig. 7A) or a subunit of phosphatase 2A (47) (supplemental Fig. S2A) was used as a reference gene for normalization. Our data agree with the Arabidopsis microarray expression data available at AtGen-Express Visualization Tool. ACA8 expression was found in all organs examined, with the highest levels in inflorescences ( Fig. 7B and supplemental Fig. S2B) (48).

Discussion
Given the abundance of CaM/CML in plants, the definition of their properties as Ca 2ϩ sensors and the identification of their downstream targets represent an impressive challenge. Herein, we have presented structural and biochemical studies demonstrating that recombinant CML36 from A. thaliana displays the typical properties of proteins that function as Ca 2ϩ sensors and can stimulate the hydrolytic activity of the Arabidopsis ACA8 isoform by specifically binding to the N-terminal regulative domain of ACA8. CML36 possesses the characteristics of regulatory EF-hand Ca 2ϩ -binding proteins including, upon Ca 2ϩ binding: (i) an increase in ␣-helical content; (ii) a shift in electrophoretic mobility; (iii) the adoption of a stable tertiary structure; and (iv) the exposure of a hydrophobic pocket that can interact with an ANS probe as a result of the Ca 2ϩ -induced conformational changes, which is often a critical process for the function of many Ca 2ϩ sensor proteins. In addition, we obtained information on the site-specific Ca 2ϩ binding affinities of CML36, which are crucial because Ca 2ϩ affinity is often attuned to the physiological role of the Ca 2ϩ sensor protein. We found that CML36 possesses two mixed Ca 2ϩ /Mg 2ϩbinding sites of high affinity (K dCa Ͻ 0.2 M and K dMg ϳ 100 M) and two Ca 2ϩ -specific sites of low affinity (Ն100 M). The NMR data indicate that CML36 binds the two Mg 2ϩ ions to EF-3 and EF-4; therefore, EF-3 and EF-4 represent the Ca 2ϩ / Mg 2ϩ high-affinity sites, whereas EF-1 and EF-2 are the lowaffinity sites. Notably, all four EF-hands of the protein are capable of sensing Ca 2ϩ ions. This finding was unexpected, as only three EF-hands were predicted to be functional Ca 2ϩ -binding sites by PROSITE-ProRule annotation (49). Indeed, EF-2 (residues 108 -127) possesses an arginine residue at position 9 of the loop, where a hydrophilic or negative side chain is usually hydrogen-bonded to a Ca 2ϩ -ligand water molecule, and thus it is not predicted to have functional Ca 2ϩ binding, although the canonical Ca 2ϩ -coordinating groups at positions 1, 3, 5, 7, and 12 are present. It is tempting to attribute this EF-hand to the lowest-affinity site. Further, the reduced affinity of EF-1 could be ascribed to the replacement of a conserved glutamate residue with an aspartate at position 12 (8). Future studies of Ca 2ϩbound CML36 X-ray structures will be helpful in elucidating the difference in the microenvironment of the EF-hands at the

CML36 binds to and activates ACA8
atomic level, which will facilitate understanding of these sitespecific Ca 2ϩ affinity properties.
Notably, Mg 2ϩ binding to high-affinity EF-3 and EF-4 sites also induces a global conformational change throughout CML36 that results in the stabilization of its tertiary structure. Cation binding to the Ca 2ϩ /Mg 2ϩ mixed sites seems to control an important structural transition from a molten globule state to a compact globular conformation. Indeed, we found that, similar to other Ca 2ϩ -binding proteins (24,25,28,50), apoCML36 likely assumes a molten globule structure and the transition from this molten globule to the compact state clearly occurs when the high-affinity sites are occupied by either Ca 2ϩ or Mg 2ϩ .
The observed binding constants of Ca 2ϩ /Mg 2ϩ mixed sites for Mg 2ϩ and for Ca 2ϩ imply that under resting conditions these two EF-hands are always occupied by a divalent cation, ensuring a folded ion-bound structure of CML36 at any cytoplasmic Ca 2ϩ concentration. Next, in response to the rapid increase in intracellular Ca 2ϩ concentrations upon a stimulus, the Ca 2ϩ -bound state becomes the dominant state of the protein, able to displace Mg 2ϩ from these EF-hands.
Beyond the description of the structural and metal-binding properties of CML36, a significant advancement in the knowledge about this protein has been obtained by the identification of ACA8 as a CML36 downstream target. As a strategical isoform of PM-localized Ca 2ϩ pumps in the shaping of intracellular Ca 2ϩ dynamics (18), ACA8 is the target of regulation by multiple Ca 2ϩ -dependent effectors (17,18,42). We have provided direct evidence that ACA8 is also able to interact with the Ca 2ϩ sensor CML36 in vitro and that this interaction results in the activation of its Ca 2ϩ -dependent hydrolytic activity with a K 0.5 value in the submicromolar range. Neither this interaction nor the activation of ACA8 is detectable with other isoforms of CMLs such as CML19 and CML14, suggesting that ACA8 may be specific for CML36. Moreover, native PAGE analysis allowed us to identify the binding site for CML36 at the N-terminal regulative domain of ACA8. The interaction of CML36 with the N terminus of ACA8 is also supported by our observations that the effects of CML36 and CaM on ACA8 activity are not additive and that variants of ACA8 without the N-terminal CaM-binding domains are completely insensitive to CML36. Nevertheless, further investigations will be necessary to pre-cisely localize the CML-binding site within the N-terminal portion.
The capacity of both CaM and CML36 to bind ACA8 raises the question of which mechanisms are capable of avoiding promiscuous interactions in cell-regulatory processes. The choice of the proper signaling pathway involving CML36 or CaM may be related to many modulating factors: (i) the abundance of each protein in a specific microdomain, cellular compartment, tissue and organ; (ii) the time-dependent expression in differentiation and development; (iii) the temporal and spatial features of the Ca 2ϩ signal; (iv) the properties of the EF-hand proteins, such as the Ca 2ϩ affinity, the number of functional binding sites, and the ability to discriminate between Ca 2ϩ and Mg 2ϩ ; and (v) site-specific phosphorylation. In this scenario, it is worth noting that the predicted affinities of CML36 for Ca 2ϩ differ significantly from those observed for mammalian and Arabidopsis CaM1 (29,51); this would influence the ability of these Ca 2ϩ sensors to respond to cytosolic Ca 2ϩ increases induced by the perception of different stimuli and thus their ability to activate their downstream targets as ACA8. Moreover, the CML36 and ACA8 genes are co-expressed mainly in shoot inflorescences. Furthermore, CML36 displays an elongated N terminus, which may represent a targeting sequence. Indeed, a mass spectrometry-based quantitative phosphoproteomics approach identifies CML36 as a protein associated with the PM fraction (37), which might facilitate the interaction with ACA8 in vivo.
PM-localized Ca 2ϩ pumps are likely to play a key role in returning cytosolic Ca 2ϩ to resting concentrations and in modulating the dynamics of a Ca 2ϩ signature (2,5). Analysis of ACA8 expression, which is up-regulated upon cold stress and abscisic acid treatment, supports its involvement in crucial processes such as response to abiotic stresses and hormonal regulation (48,52). Furthermore, genetic evidence using KO plant mutants also indicates additional roles for ACA8 in developmental processes and plant immune response (36). Expression of the CML36 gene also suggests a possible role in organ development for CML36. Interestingly, the CML36 transcript has been found to be altered in an Arabidopsis double mutant lacking both ACA8 and ACA10, another PM-localized Ca 2ϩ -ATPase isoform (36). In addition, ACA8 has been recently identified as a prominent in vivo regulator of cellular Ca 2ϩ

CML36 binds to and activates ACA8
dynamics which crucially functions in the termination of Ca 2ϩ signals induced by mechanical wounding and related stimuli (18). In this scenario, CML36 would act in concert with other Ca 2ϩ sensors such as CaM (16,39,53,54), CDPK (42), and the CBL-CIPK complexes (18) to regulate ACA8 activity. In addition, other CMLs may be involved in the modulation of ACA8 and other different isoforms of autoinhibited Ca 2ϩ pumps, which might contribute to the elaborated/intricate Ca 2ϩ -dependent regulatory network for the fine-tuning of intracellular Ca 2ϩ homeostasis.

Protein production
The coding sequence for CML36 (At3g10190) was PCR-amplified using pUNI51-U23411 vector as template obtained from The Arabidopsis Information Resource (TAIR). The following primers were employed: forward, 5Ј-CATATGCACCAC CACCACCACCACATGAAACTCGCCAAAC-3Ј; and reverse, 5Ј-GAATTCTCAACGCTGGAGATCCATCATTCGTGAG-3Ј. The forward primer carried a NdeI restriction site and a His 6 tag, and the reverse primer contained a stop codon and an EcoRI restriction site. The verified cDNA sequence was then cloned into the corresponding restriction sites of pET21a vector for expression in Escherichia coli BL21(DE3) cells. Cells were grown in LB medium at 37°C to A 600 nm ϭ 0.6, and expression of CML36 protein was then induced at 25°C for 16 h by the addition of 0.4 mM IPTG. To express 15 N-labeled CML36 for NMR studies, M9 minimal medium supplemented with 15 NH 4 Cl (1 g liter Ϫ1 ) was employed. The bacterial pellet obtained after low-speed centrifugation of a 1-liter culture was solubilized in 20 mM Tris-HCl, pH 7.5, 150 mM KCl, 10 mM imidazole, and 1 mM DTT lysis buffer containing 0.1 mg ml Ϫ1 lysozyme, stirred at room temperature for 30 min, and then sonicated on ice. The lysate was centrifuged at 30,000 ϫ g for 10 min at 4°C, and the supernatant was loaded onto a nickelaffinity column pre-equilibrated in 20 mM Tris-HCl, pH 7.5, 150 mM KCl, 1 mM DTT, and 10 mM imidazole buffer. A gradient of imidazole from 10 to 500 mM was applied to the column, and the His-tagged CML36 was eluted around 100 mM imidazole. The fractions containing CML36 were pooled and washed extensively at 4°C with 20 mM Tris-HCl, pH 7.5, 150 mM KCl, and 1 mM DTT using Vivaspin concentrators (Sartorius, Goettingen, Germany) to remove the imidazole. Protein concentration was determined using the Bio-Rad assay.
cDNA encoding N-terminally truncated CML36 was generated by PCR using the complete cDNA of CML36 as template, a forward primer containing a NdeI site (5Ј-CATATGCATCAT-CATCATCATCACGAGGTTCCTTCTCCTTATTC-3Ј) and a reverse primer containing a EcoRI site (5Ј-GAATTCT-CAACGCTGGAGATCCATCATTCGTGAG-3Ј). The resulting DNA fragment was inserted into pET21a using the NdeI/ EcoRI sites for directional cloning. The newly generated plasmid was checked by DNA sequencing and used to transform E. coli BL21(DE3) cells. The conditions for expression and purification of the deletion mutant were as described for CML36 f.l. The yield was about 6 mg and 100 mg for a 1-liter culture for the full-length protein and the N-terminally trun-cated variant, respectively. Purified CML36 variants were treated with 2 mM EGTA, washed extensively against decalcified buffer and then against decalcified NH 4 HCO 3 buffer, lyophilized, and stored at Ϫ80°C until use. Buffers were decalcified by using Chelex-100 ion-exchange resin (Sigma) according to the manufacturer's instructions. The residual Ca 2ϩ concentration was determined using 5,5Ј-Br2-BAPTA (5,5Ј-dibromobis-(o-aminophenoxy)ethane-N,N,NЈ,NЈ-tetraacetic acid) as described (55,56).
AtCaM1 and CML14 were produced as described (43,57). CML19 was expressed and purified as reported in the supplemental "Methods". The production of recombinant ACA8-N, the N-terminal domain of ACA8 (residues 1-116) was performed as described (54,58).
The Ca 2ϩ -induced mobility shift of CML36-C was performed on a 12.5% Tris-glycine native PAGE and a 15% SDS-PAGE after incubating CML36-C with 5 mM CaCl 2 or 5 mM EGTA for 30 min at room temperature.

Nuclear magnetic resonance spectroscopy
NMR spectra were collected on a 600-MHz Bruker Avance III spectrometer (Bruker, Karlsruhe, Germany) equipped with a cryogenic probe. A standard 1 H-15 N HSQC pulse sequence was used with pulsed-field gradients for suppression of the solvent signal. 1  To achieve partial backbone atom assignment of CML36-C bound to Ca 2ϩ or Mg 2ϩ ions, standard triple resonance NMR experiments were acquired on 15 N, 13 C-labeled samples (59). The PFG-STE (pulsed-field gradient-stimulated echo) experiments, employing bipolar gradients (60), were conducted and analyzed as described previously (61). DSS (1 mM final concentration) was added to the protein samples and used as an internal standard (62). Delays of 25 and 180 ms between the defocusing and refocusing gradient elements were employed for experiments on DSS and protein samples, respectively. Spectra were acquired with 8000 complex points and 128 transients. The R h values of the proteins were calculated based on the rela- where R b ref is the hydrodynamic radius reported for DSS (3.43 Å) and D DSS and D prot are the diffusion coefficients calculated as described for the DSS and the protein, respectively. The spectra were recorded at 25°C in 20 mM Tris-HCl, 1 mM DTT, pH 7.5, at a protein concentration of 500 M in the presence of 5 mM EGTA, 5 mM MgCl 2 , or 5 mM CaCl 2 and analyzed with Topspin3.2 (Bruker, Karlsruhe, Germany).

Size-exclusion chromatography
The hydrodynamic radius of apoCML36-C and Mg 2ϩ -and Ca 2ϩ -bound CML36-C was estimated by SEC using a Superose 12 column (10/300GL, GE Healthcare) as described previously (43,63). Each experiment was performed at least in triplicate, and the reported values represent means Ϯ S.E.

Spectroscopic measurements
CD spectra were collected on a Jasco J-710 spectropolarimeter at 25°C in 20 mM Tris-HCl,150 mM KCl, and 1 mM DTT, pH CML36 binds to and activates ACA8 7.5, in the presence of 5 mM EGTA, 5 mM CaCl 2, or 5 mM MgCl 2 . Protein concentration was ϳ90 M in a cuvette with a path length of 1 cm for near-UV wavelengths, and for far-UV spectra the CML36-C concentration was 10 M in a cuvette with a 0.1-cm path length. Three spectra, averaged automatically, were recorded for each sample as described (29). The ANS fluorescence spectra were recorded on a Jasco FP8200 spectrofluorometer as described (29).

Isothermal titration calorimetry
ITC experiments were performed on a TA Instruments Nano ITC as described (29,43). All solutions were degassed prior to the rinsing and loading of both the titration and sample cells. The buffer used was decalcified 20 mM Tris-HCl, 150 mM KCl, pH 7.5. Lyophilized CML36-C was resuspended in decalcified buffer.
For Ca 2ϩ titrations into 200 M apoCML36-C or Mg 2ϩ -saturated CML36-C, 4 mM CaCl 2 was injected with a 0.3-l injection to displace air from the syringe followed by 40 ϫ 1.25-l injections. Mg 2ϩ -saturated CML36-C was prepared by incubating apoCML36-C for 15 min at 25°C with 5 mM MgCl 2 . For experiments in which Mg 2ϩ was injected into the apoCML36-C, a 5 mM MgCl 2 stock was added to 170 M apoCML36 by a 1-l injection to displace air from the syringe followed by 24 ϫ 2-l injections.
The heats of dilution, collected by injecting Ca 2ϩ or Mg 2ϩ into buffer without the protein in the sample cell, were subtracted from the raw integrated heats in the presence of protein. Data were fitted using the best-fitting model for each experiment (sequential, two-site, or one-site model). The reported values represent the mean Ϯ S.E. of at least three independent titrations.
The fitted K a values were converted to K d values using the equation,

Native PAGE analysis
The formation of the complex following ACA8 binding was analyzed under native conditions as described previously (29).
Briefly, CML36 f.l., CML36-C, Arabidopsis CaM1, CML14, and CML19 were dissolved at a 10 M concentration in 100 mM Tris, pH 7.5, 5 mM CaCl 2 , and 4 M urea and incubated in the presence of an excess of recombinant ACA8-N for 1 h at room temperature. Samples were assayed on a 12.5% continuous gel containing 5 mM CaCl 2 and 4 M urea. Urea was added either to the gel and the incubation buffer to prevent any nonspecific interactions.

ACA8 purification from microsomes
Yeast microsomes were solubilized as reported (45) and incubated (45 mg of microsomes) overnight at 4°C on 4 ml of CaM-Sepharose 4B gel (GE Healthcare). Purification was performed as described (45). The 1 mM EGTA-eluted fraction, added with stoichiometric CaCl 2 , was concentrated about 70-fold on Vivaspin ultrafiltration spin columns, with a 30-kDa cutoff (Sartorius, Goettingen, Germany), and immediately used for testing Ca 2ϩ -ATPase activity or for overlay analysis. For quantification, 1 l of the concentrated eluate was loaded onto a precast Tris-Tricine polyacrylamide gel (4 -12% linear gradient; Anamed, Darmstadt, Germany) with increasing amounts of ␤-galactosidase used as the standard. After staining with Coomassie Blue, signal quantification was carried out using AlphaEaseFC software by Alpha-Innotech (MMedical, MI, Italy). ACA8 was purified to virtual homogeneity.

Ca 2؉ -ATPase activity assay
ACA8 activity in yeast microsomes (0.4 mg of protein/ml) and in the EGTA-purified fraction (0.03 mg of protein/ml) was assayed as Ca 2ϩ -dependent MgITP or MgATP hydrolysis, respectively, as described (40,42). Free Ca 2ϩ concentration was buffered at 10 M with 1 mM EGTA. Bovine testes CaM (Sigma-Aldrich) and A. thaliana purified Ca 2ϩ -bound CML36 or CML14 were added at the specified concentrations. Assays were performed by preincubating CMLs with microsomes or ACA8-purified protein in assay buffer for 20 min at 25°C before starting the reaction by the addition of nucleotide substrates. Samples were then incubated at 25°C for 60 min. MgITP or MgATP hydrolysis measured in the presence of 1 mM EGTA without added Ca 2ϩ was subtracted from the reported data. Assays were performed at least four times, with three replicates for each condition, using three independent biological samples and at least two different CML36 preparations. The standard error did not exceed 5% of the mean. Data analysis and curve fitting were performed using SigmaPlot software. K 0.5 for CML36 (ϮS.E.) was estimated using nonlinear regression. Student's t test was conducted to assess potential differences between K 0.5 estimated from data obtained using yeast microsomes and purified ACA8 (p ϭ 0.1173).

Overlay assay
Different amounts of purified Ca 2ϩ -bound CML36 (0.6, 0.12, and 0.03 nmol), 0.6 nmol of purified Ca 2ϩ -bound CML14, 0.6 nmol of lysozyme (Sigma-Aldrich), and 0.03 nmol of bovine testes CaM (Sigma-Aldrich) were spotted onto 0.2-m nitrocellulose. Proteins were first cross-linked to the membrane by incubating them for 45 min at room temperature in 0.2% (v/v) glutaraldehyde freshly prepared in KP buffer (25 mM KH 2 / K 2 HPO 4 buffer, pH 7). The membrane was incubated thereafter for 16 h in blocking solution (3% (w/v) BSA, 0.15 M NaCl, 2 mM MgSO 4 , 0.2% (v/v) polyoxyethylene (20) sorbitan monolaurate, and 20 mM Tris-HCl, pH 7.4). After washing, the overlay was conducted by incubating the membrane for 2 h at room temperature in blocking solution with 1.5 M purified ACA8 or, in parallel as a negative control, in the same solution without ACA8. Following a fast wash in blocking solution, an additional 45-min cross-linking step was carried out. Immunodetection

Plant material, growth conditions, and gene expression analysis by quantitative reverse transcription-PCR
To evaluate CML36 gene expression in various adult organs, A. thaliana ecotype Columbia (Col-0) seeds were sown on soil in a greenhouse under a 16-h light/8-h dark photoperiod at 24°C and 20°C, respectively; the plants were grown to maturity. The isolation of total RNA was performed using the RNeasy mini kit (Qiagen, Hilden, Germany) from 100 mg of frozen organs, and then the RNA samples were treated with RQ1 RNase-free DNase (Promega, Madison, WI). cDNA synthesis was obtained with oligo-(dT) primer and carried out using the GoScript reverse transcription system (Promega). SYBR Green qPCR Supermix-UDG (Invitrogen) was used to amplify cDNA on the StepOnePlus TM real-time PCR system (Applied Biosystems, Foster City, CA). Quantitative RT-PCR was performed using the following cycling conditions: 2 min at 50°C, 2 min at 95°C, 40 cycles of 95°C for 30 s, 59°C for 30 s, 72°C for 30 s, and finally 72°C for 3 min. The actin gene was used as an endogenous control and to normalize all quantifications. Three cDNA samples obtained from three independent RNA extractions were examined. Relative quantitation of transcript levels was performed as reported (64). The following forward (F) and reverse (R) primers were employed: for CML36 (At3g10190), F, 5Ј-CATCTTCCGGCCAAGACTGT-3Ј, and R, 5Ј-ACCTCCGGGAGAATGCTAGT-3Ј; for ACA8 (At5g57110), F, 5Ј-AAGTCCTTACCGTAGCGGTTACAA-3Ј and R, 5Ј-ATTGAATAGGCAAGGGTTAAAGTAACA-3Ј (48). Two internal reference genes, an actin gene (At3g18780) and a subunit of phosphatase 2A (PP2A At1g13320), were used for data normalization (47,65). The following primers were employed: for actin, F, 5Ј-TGTTCTCTCCTTGTACGC-CAGT-3Ј, and R, 5Ј-CAGCAAGGTCAAGACGGAGGA-3Ј; and for PP2A, F 5Ј-TAACGTGGCCAAAATGATGC-3Ј, and R, 5Ј-GTTCTCCACAACCGCTTGGT-3Ј (47).