Evidence for differing roles for each lobe of the calmodulin-like domain in a calcium-dependent protein kinase.

Calcium-dependent protein kinases (CDPKs) are structurally unique Ser/Thr kinases found in plants and certain protozoa. They are distinguished by a calmodulin-like regulatory apparatus (calmodulin-like domain (CaM-LD)) that is joined via a junction (J) region to the C-terminal end of the kinase catalytic domain. Like CaM, the CaM-LD is composed of two globular EF structural domains (N-lobe, C-lobe), each containing a pair of Ca(2+) binding sites. Spectroscopic analysis shows that the CaM-LD is comprised of helical elements, but the isolated CaM-LD does not form a conformationally homogeneous tertiary structure in the absence of Ca(2+). The addition of substoichiometric amounts of Ca(2+) is sufficient to stabilize the C-terminal lobe in a construct containing J and CaM-LD (JC) but not in the CaM-LD alone. Moreover, as J is titrated into Ca(2+)-saturated CaM-LD, interactions are stronger with the C-lobe than the N-lobe of the CaM-LD. Measurements of Ca(2+) affinity for JC reveal two cooperatively interacting high affinity binding sites (K(d)(,mean) = 5.6 nm at 20 mm KCl) in the C-lobe and two weaker sites in the N-lobe (K(d,mean) = 110 nm at 20 mm KCl). The corresponding Ca(2+) binding constants in the isolated CaM-LD are lower by more than 2 orders of magnitude, which indicates that the J region has an essential role in stabilizing the structure of the CDPK regulatory apparatus. The large differential affinity between the two domains together with previous studies on a plasmodium CDPK (Zhao, Y., Pokutta, S., Maurer, P., Lindt, M., Franklin, R. M., and Kappes, B. (1994) Biochemistry 33, 3714-3721) suggests a model whereby even at normally low cytosolic levels of Ca(2+), the C-lobe interacts with the junction, but the kinase remains in an autoinhibited state. Activation then occurs when Ca(2+) levels rise to fill the two weaker affinity binding sites in the N-lobe, thereby triggering a conformational change that leads to release of the autoinhibitory region.

at normally low cytosolic levels of Ca 2؉ , the C-lobe interacts with the junction, but the kinase remains in an autoinhibited state. Activation then occurs when Ca 2؉ levels rise to fill the two weaker affinity binding sites in the N-lobe, thereby triggering a conformational change that leads to release of the autoinhibitory region.
On the basis of sequence homology, CDPKs are most closely related to calmodulin-dependent protein kinases (CaMKs) (6,1). The majority of CaM-regulated kinases contain a CaM binding region near their C terminus. The CDPKs are similar, except they contain their own calmodulin-like regulatory apparatus (CaM-LD) C-terminal to the CaM binding region of the catalytic domain (K) (Fig. 1). The CaM-LD consists of two structural domains (termed the N and C "lobes" in this manuscript), each containing two EF-hand helix-loop-helix Ca 2ϩ binding motifs. The region that joins the kinase and CaM-LD is referred to as the junction (J). The junction contains an autoinhibitor and a target sequence for intramolecular binding of the CaM-LD (7).
The similarities between the CaMK and CDPK systems are numerous. For example, both CaMKs and CDPKs are regulated by an autoinhibitor located immediately C-terminal to the kinase domain. In both cases, autoinhibition is relieved by interaction with CaM or the CaM-LD. The CaM-LD of CDPK has 40% sequence identity with CaM, which suggests a potential similarity in structure and function. In fact, the CaM-LD can be functionally replaced by a CaM sequence in a chimeric CDPK (8). Additionally, both an exogenous CaM and a CaM-LD were shown to be able to activate a truncated CDPK, although to no more than half of the extent of the holoenzyme (9). However, what distinguishes CDPKs and CaMKs is the cova-* This work was supported by National Institutes of Health Grant GM 40120 (to W. J. C.) and National Science Foundation Grants DBI-9975808, MCB-9408101, and MCB-9723539 (to J. F. H.). Access to NMR and mass spectrometry facilities were supported in part by National Institutes of Health Grant P30 ES00267 (to the Vanderbilt Center in Molecular Toxicology). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains Supplemental Fig. 1 lent tethering of the CaM-LD in a CDPK. Evidence from mutational analyses indicates that this tether provides a structural constraint that is essential for the mechanism of intramolecular activation. An insertion of three glycines into the tether, which would be expected to increase its flexibility, results in a 95% disruption in calcium activation (9). The exact structural features of the tether that help mediate activation by the CaM-LD are unclear.
To more fully elucidate the mode of action of the regulatory apparatus of CDPK, we have undertaken a series of biophysical analyses of the CaM-LD and its interactions with the J region. The structural character of the CaM-LD was examined by CD and NMR. An estimate of the strength of the interaction between J and the CaM-LD was obtained by measuring the affinity between the isolated CaM-LD and a peptide corresponding to the J region. Because such measurements do not take into account the effects associated with the tethering of J to the CaM-LD, the linked thermodynamics of Ca 2ϩ and J binding (10,11) were characterized by measuring the Ca 2ϩ affinities of the CaM-LD and JϩCaM-LD (JC) construct. The results obtained from these studies indicate that the two lobes of the CaM-LD do not function in an equivalent manner. Despite the finding that the C-lobe makes the predominant contacts with the junction region, the implications of our findings suggest that the N-terminal lobe of the CaM-LD functions as the calcium sensor for the activation of CDPK. This latter finding is consistent with previous studies on a plasmodium CDPK isoform (12) where mutational analysis showed that only the N-lobe calcium sites were essential for calcium-dependent activation.

EXPERIMENTAL PROCEDURES
Protein Expression-The CDPK isoform, CPK-1, used in this work was cloned from Arabidopsis thaliana (13). Constructs encoding the regulatory apparatus of CDPK (residues Lys-428 -Gly-591) were subcloned from CPK-1 and sequenced to verify the absence of mutations resulting from subcloning. The pET-28bϩ plasmid (Novagen) was used for constructs of the CDPK regulatory domain. This has an N-terminal His tag and a thrombin cleavage site and was used with Escherichia coli BL21(DE3) as the host bacterial strain.
Standard growth conditions were used; 2xYT media for unlabeled preparations and M9 minimal media containing [ 15 N]ammonium chloride and/or [ 13 C]glucose (Cambridge Isotopes Laboratories, Inc.) as the sole nitrogen and carbon sources as needed. Media were supplemented with the antibiotic kanamycin (50 g/ml) and vitamins (Basal Eagle Vitamin Mix, Invitrogen). After isopropyl-␤-D-thiogalactopyranoside induction, the expression of CDPK constructs accounts for the majority of total cell protein. High expression levels were achieved (ϳ60 mg/liter with 2xYT medium and 5-20 mg/liter in minimal media). Although much of the expressed protein was in the periplasmic fraction, the high expression levels resulted in a significant amount (typically 50%) of the target protein in inclusion bodies.
Protein Purification-CDPK constructs were purified by suspending the cell pellet in a minimum volume of lysis buffer (20 mM Tris-HCl, 100 mM NaCl, 0.2 mg/ml lysozyme). Cells were lysed using gentle sonication, and after centrifugation the resulting supernatant was purified by nickel-affinity chromatography (nickel-agarose resin, Qiagen) followed by thrombin cleavage of the purification tag. Thrombin (Calbiochem) was purified before use via a Mono Q ion-exchange column (Amersham Biosciences). Thrombin inactivation was achieved via a benzamidine-Sepharose column (Amersham Biosciences) and with phenylmethylsulfonyl fluoride. Further purification steps included either Superdex-75 or 200 gel filtration using fast protein liquid chromatography (AKTA system, Amersham Biosciences). Protein masses were confirmed by matrix-assisted laser desorption ionization mass spectrometry. Further assessments of purity were by SDS-PAGE and analytical gel filtration (TSK2000SW gel filtration, Tosohaas) using HPLC. CDPK proteins concentrations were determined by amino acid analysis (Protein and Nucleic Acid Facility, University of Cambridge).
A significant amount of protein was expressed in inclusion bodies, which were extracted using standard urea denaturation in the constant presence of thiol reducing agent (10 mM DTT). The denatured protein was refolded by slow dialysis and then subjected to the purification procedure described above. The structural integrity of refolded protein was assayed by two-dimensional 1 H, 15 N HSQC NMR, which showed the refolded protein to be identical to protein purified from soluble fractions.
Peptides-Peptides corresponding to the junction region were obtained from Genemed Synthesis, Inc. Analytical reverse-phase HPLC and electrospray mass spectrometry showed Ͼ95% purity. Peptide concentrations were determined by absorbance at 215 nm relative to bovine serum albumin.
Circular Dichroism-Experiments were performed on Jasco J-720 and J-820 spectropolarimeters. Two scans were averaged. A bandwidth of 1 nm and a scan step size of 0.25 nm were employed with a 1-cm path length. Spectra were recorded at 298 K at 5-10 M protein concentrations in 2 mM Tris-HCl, 18 mM KCl, and 0.1 mM DTT at pH 7.4. Spectra were processed, and plots were created using Kaleidagraph (Synergy Software).
For studies where a strict Ca 2ϩ -free environment was required, all buffers were passed through pre-equilibrated Chelex resin (iminodiacetic acid on agarose support, Aldrich). All glassware and consumables were washed with EDTA and then thoroughly rinsed with Chelextreated water before use. Protein samples were rendered metal-free by bringing 1 mg/ml solutions to 50 mM EDTA. This was followed by repeated dialysis against Chelex-treated buffer.
NMR Spectroscopy-Experiments were recorded on Bruker Avance 500, 700, and 800 spectrometers at 310 K. Samples contained 0.5-1.0 mM protein in 50 mM Tris-d 11 (Cambridge Isotope Laboratories Inc.) and 100 mM KCl with 10 mM DTT and 5% (v/v) D 2 O. Titrations were monitored by two-dimensional 1 H, 15 N HSQC spectra acquired with quadrature detection via the TPPI technique (14). Sweep widths were ϳ13 and 30 ppm in the 1 H and 15 N dimensions, respectively, and acquisition parameters ranged from 64 to 256 increments for each of 8 -32 scans. Three-dimensional HNCO and HNCA spectra were recorded at 800 MHz with constant time in the 15 N dimension (15). 2048, 88, and 80 complex points were recorded in the 1 H, 15 N, and 13 C dimensions, respectively.
NMR resonance assignments have been reported previously for Ca 2ϩ -loaded JC (BioMagResBank Entry 5324, www.bmrb.wisc.edu (16)). For this study the resonances of Ca 2ϩ -loaded CaM-LD were assigned by comparison with JC. Although broadly similar, there were some significant differences in the two-dimensional 1 H, 15 N HSQC profile of JC and CaM-LD. Consequently, three-dimensional HNCO and HNCA spectra were recorded on 13 C, 15 N-labeled Ca 2ϩ -CaM-LD both in the absence and presence of the unlabeled junction peptide and compared with those of JC.
Felix2000 (Accelrys, Inc.) was used for processing and display of spectra. For the three-dimensional experiments, linear prediction was used to double the number of points in each of the indirect dimensions before zero-filling. An un-shifted cosine function was applied to each dimension. The Assign module of Felix2000 was employed to assist in assigning resonances and to create overlay plots.
Ca 2ϩ -affinity Determinations-Macroscopic Ca 2ϩ binding constants (K 1 -K 4 ) were determined for CDPK constructs using the method of Linse et al. (17), which involves competitive chelation using 5,5Ј-Br 2 BAPTA (Molecular Probes) monitored via the absorbance at 263 nm as a function of added calcium. 2 Titrations were performed at 298 K with 37.3 M 5,5Ј-Br 2 BAPTA in 18 mM KCl and 2 mM Tris-HCl at pH 7.2. The concentrations of JC and CaM-LD were 12.4 and 11.6 M, respectively, as determined by amino acid analysis. The K d of the chelator under these conditions was found to be 0.22 Ϯ 0.01 M. At least two determinations were made for each protein preparation. Individual binding constants were determined using the CaLigator software (19).

RESULTS
To examine the significance of a covalent linkage between the junction and the CaM-LD, expression constructs were engineered as shown in Fig. 1. The JC construct incorporating the CaM-LD with the N-terminally tethered junction region contains the full Ca 2ϩ regulatory apparatus of CDPK. Studies of this construct were complemented by a bimolecular version of JC, made up of the isolated CaM-LD plus the free J peptide (JϩCaM-LD). Binding events were monitored by CD, heteronuclear NMR, and calcium-affinity measurements.
The Global Fold of the CDPK Regulatory Apparatus-Far UV CD spectra in the absence and presence of Ca 2ϩ for JC and the CaM-LD are shown in Fig. 2. As expected from the sequence and structural homology with CaM, the dominant contribution is from helical secondary structure (minima at 208 and 222 nm). These results are consistent with the analysis of the NMR chemical shifts of Ca 2ϩ -loaded JC (16), which showed that the distribution of elements of secondary structure in CaM-LD ( Fig. 1B) is very similar to that of CaM.
Although the CD spectrum of the Ca 2ϩ -free state of JC clearly indicates considerable helical structure is present, its 1 H, 15 N HSQC spectrum is characterized by broad peaks and very limited chemical shift dispersion (Fig. 3). Together, the CD and NMR data indicate that the protein maintains much if not all of the native secondary structure but lacks a stable tertiary fold in the absence of Ca 2ϩ (20). The addition of Mg 2ϩ (20 -100 mM) in an attempt to stabilize the tertiary structure of the apo protein had no effect on the NMR spectrum. These observations are consistent with the lack of an effect of Mg 2ϩ on the binding of Ca 2ϩ to soybean CDPK (8).
Significantly higher intensity is found in the CD spectrum of apoJC relative to that of apoCaM-LD (Fig. 2), implying that the J region in JC has partial helical character. Because the J peptide is unstructured on its own, this observation suggests that the CaM-LD interacts with the J region even in the absence of Ca 2ϩ . This contrasts with the tethered CaM-M13 (binding sequence from myosin light chain kinase) hybrid sys- tem, where no interaction was observed between the M13 peptide and CaM in the absence of Ca 2ϩ (21).
CD has been used to determine whether structural changes occur upon Ca 2ϩ binding. Fig. 2 shows there is an increase in intensity of far UV CD spectra of JC upon the addition of Ca 2ϩ . This change in apparent helicity is not attributed to increased helical content but, rather, to a significant change in the spatial distribution of the helices associated with the Ca 2ϩ -induced transition from a closed to an open conformation as shown previously for troponin C (22) and calmodulin (23).
To obtain insights into the regulatory apparatus, the intrinsic affinity of the J region for the CaM-LD was determined. This involved using NMR to monitor structural changes as J peptide (residues 428 -450, numbering as in Fig. 1B) was titrated into a CaM-LD solution, fitting chemical shift changes to a standard binding equation. Changes in the 1 H, 15 N HSQC spectra saturate at 1.0 (Ϯ0.3) mol eq of added J peptide, indicating a 1:1 stoichiometry. Saturation at an approximately equimolar ratio was consistent with the K d of ϳ1 M derived from surface plasmon resonance measurements (7).
Upon the addition of J peptide to Ca 2ϩ -saturated CaM-LD, an increase in CD intensity was observed. A similar observation has been reported for soybean CDPK (24) and is consistent with helix formation in the J peptide upon binding to the Ca 2ϩ -CaM-LD. A transition from disorder to ordered ␣-helix is characteristic of many target peptides that interact with CaM (25,26). However, the magnitude of the change in CD for JϩCaM-LD is so large it cannot be explained solely by J adopting ␣-helical conformation. We propose that binding of J further alters the disposition of the CaM-LD helices. Note that the CD profile of CaM-LD in the presence of J peptide and Ca 2ϩ is similar to that observed for Ca 2ϩ -loaded JC. This suggests that the unimolecular and bimolecular CaM-LDϩJ systems adopt similar structures in the fully calcium-loaded state.
Ca 2ϩ -affinity Determinations-The Ca 2ϩ binding constants (K d ) for JC and CaM-LD (Table I) were determined by competition against BAPTA, a standard Ca 2ϩ chelator with known Ca 2ϩ affinity, using the approach of Linse et al. (17). K 1 -K 4 are the macroscopic binding constants for the binding of four ions of Ca 2ϩ to JC. The assignment of these binding constants to specific sites in the protein requires very detailed analysis, but assignment to individual domains is somewhat easier, particularly when one domain has significantly higher binding constants than the other (27). For JC, two pairs of co-operative binding events were observed (Fig. 4), characterized by dissociation constants of log K 1,2,mean ϭ log(K 1 ⅐K 2 ) 1/2 ϭ 8.3 and log K 3,4,mean ϭ log(K 3 ⅐K 4 ) 1/2 ϭ 6.9. The binding constants for the two stronger sites in CaM-LD are ϳ2 orders of magnitude lower in CaM-LD than in JC. In the two weaker sites, the affinities in CaM-LD are so drastically reduced by the absence of the J region that their dissociation constants are in the millimolar range and could not be accurately measured. This binding affinity is well below that of Ca 2ϩ sites  Table I. Error bars shown on the Ca 2ϩ concentration correspond to the uncertainty in the absorption measurement of 0.003 absorbance units.  participating in Ca 2ϩ signal transduction pathways and points to the great importance of the J region in the CDPK regulatory apparatus.
The C-lobe Contains the Higher Affinity Ca 2ϩ Sites-To learn more about the mode of action of the CaM-LD, NMR was used to characterize the Ca 2ϩ binding events in a site-specific manner. For these experiments, Ca 2ϩ was titrated into a solution of JC, and two-dimensional 1 H, 15 N HSQC spectra were recorded at each step. The intensities of 45 well-resolved residues scattered across the protein sequence could be monitored in these spectra.
At Ca 2ϩ levels below 0.1 mol eq there is severe attenuation of already weak signals from apoJC, presumably because of conformational exchange between the apo and various subsaturated Ca 2ϩ states. Above 0.1 mol eq of Ca 2ϩ , a number of signals from the C-lobe begin to appear at the same positions as in the spectrum of fully Ca 2ϩ -loaded JC (Fig. 3B). These signals continue to increase in intensity throughout the Ca 2ϩ titration. The observation of this subset of relatively narrow signals implies the species to which it corresponds is in slow exchange with the heterogeneous ensemble of other states that we presume have partial loading of Ca 2ϩ . The putative complete filling of sites at substoichiometric levels of Ca 2ϩ suggests binding occurs in a cooperative manner.
As noted above, the first signals to appear are all from the C-lobe. In fact, differences in the relative intensity of signals from the N-and C-lobes were observed over the full course of the titration. Fig. 5 shows selected results for residues from the four calcium binding loops and for Leu-441 in the J region. At Ca 2ϩ concentrations below 0.5 mol eq, only resonances from residues in the C-lobe (Ala-530 and onward) are observed, and these continue to have significantly higher intensity until the latter stages of the titration. These observations indicate that the high affinity Ca 2ϩ sites are located in the C-lobe. Because the affinities differ by a factor of 25, the assignment of two pairs of binding constants to the individual CaM-LD domains is reliable.
The Junction Region Interacts Predominantly with the Clobe-NMR was also used to structurally characterize the in- FIG. 5. Residue-specific responses to Ca 2؉ binding. Normalized cross-peak intensity plotted against mol eq of Ca 2ϩ for selected residues in the junction region (A) and residues in Ca 2ϩ binding sites 1-4 (B-E), respectively. Lines are drawn solely to indicate trends. In B-E, the same symbols are used for residues in corresponding Ca 2ϩ coordination loop positions (e.g. Gly at position 6 and Ile at position 8).
teraction of J with the CaM-LD. Titration of the J peptide into a solution of CaM-LD caused significant perturbations in a number of CaM-LD resonances. Interestingly, plots of the chemical shift changes of backbone amides against the sequence (Fig. 6A) show large changes (ϾϮ0.1 ppm) occur almost exclusively for residues in the C-terminal lobe, with the most significant effects on residues in and around helix 7. This observation implies that the binding interactions of J with the C-lobe are stronger. Additional insight was obtained from a qualitative assessment of the rates of amide proton exchange with solvent (Fig. 6). When a sample of Ca 2ϩ -loaded JC was lyophilized then dissolved in D 2 O, the vast majority of signals remaining in the 1 H, 15 N HSQC spectrum arise from the C-lobe, indicating more protection of this domain. The data are, thus, consistent with the J region interacting predominantly with and stabilizing the C-lobe of the CaM-LD.
One additional point of interest is that the NMR chemical shifts at the end of the titration of J into Ca 2ϩ -saturated CaM-LD (Fig. 6A) were not all identical to those of Ca 2ϩsaturated JC (Fig. 6B), although the overall trend to a greater effect on the C-lobe than the N-lobe is found in both systems. Some of the differences between the unimolecular JC and bimolecular JϩCaM-LD systems can be attributed to the absence of the covalent link between J and the first helix in the N-lobe in the bimolecular systems. In addition the sensitivity of amide resonances to the inevitable differences in the experimental conditions must be considered. However, the observed differences extend beyond what is expected from these factors. The most striking aspect in comparing the data in Figs. 6, A and B, is the larger perturbations in the N-lobe of JC relative to JϩCaM-LD. These observations suggest that, although the bimolecular system is an excellent model for studying CDPK, the linking of J to the CaM-LD plays a more significant role than simply tethering together two independent regions of the protein. dramatically altered by the presence of a covalently tethered J region; for the JC protein, the C-lobe calcium affinity is 5.6 nM, 100-fold stronger than in that of the isolated CaM-LD. Second, the functional properties of the N-and C-lobes of the CaM-LD are not equivalent, with measured Ca 2ϩ binding affinities for the C-lobe ϳ20-fold stronger than for the N-lobe (Table I). Distinct binding constants (0.6 and 55 M) have also been observed in a tomato CDPK isoform (28). These affinities are significantly lower than those determined for the isolated constructs, although again there is a large affinity difference (ϳ 90-fold) between the two lobes of the tomato CaM-LD. Variability in calcium sensitivity has been shown for soybean CDPK, where overall calcium affinities ranged from 51 to 1.6 M for three isoforms. Furthermore, the presence of different substrates caused varying increases (ϳ80-fold) from these values. In the present case, the high affinity observed for the C-lobe suggests that it may be loaded at basal cytosolic levels of Ca 2ϩ , whereas the weaker affinity of the N-lobe suggests that it can respond to Ca 2ϩ fluxes expected during Ca 2ϩ signaling. The following discussion elaborates on these features.
Are J and CaM-LD Pre-associated at the Basal Level of Ca 2ϩ ?-The observation that the C-lobe has a high Ca 2ϩ affinity when the CaM-LD is tethered to a J region suggests that the CaM-LD and J form a stable interaction at basal Ca 2ϩ concentrations. At such low levels of Ca 2ϩ , only the resonances of residues in the C-lobe of JC are visible (Fig. 5), which suggests that the interaction involves the C-lobe of the CaM-LD and the J region. Indeed the C-lobe appears to have stronger interactions with J at all levels of Ca 2ϩ loading (Fig. 6).
For deeper insight into the mechanism of action of the CaM-LD regulatory apparatus, the macroscopic binding constants for CaM-LD and JC were used to simulate 3 the overall Ca 2ϩ loading and the populations of molecules with 1, 2, 3, or 4 ions bound (p 1 -p 4 ) (Supplemental Fig. S1). Fig. 7 shows a plot of the populations of the various states of Ca 2ϩ occupancy as a function of the free Ca 2ϩ level based on the measured binding constants. The dissociation constant for the C-lobe is far below the values found for typical Ca 2ϩ sensors, indicating it would be saturated at the basal level of cytosolic Ca 2ϩ (ϳ0.1 M). This would imply the C-lobe is pre-associated with the J before Ca 2ϩ activation.
The values of 0.006 and 0.1 M measured for the C-and N-lobes of CDPK both fall below the ϳ1 M range of a typical Ca 2ϩ sensor. However, it is important to note that these K d measurements were made at low ionic strength. Increases in K d as ionic strength is increased are well documented, and for CaM, an increase of about a factor of four has been reported when going from 25 to 100 mM KCl (17). Thus, under physiological conditions, the actual K d values are expected to be at least 5 times higher, bringing the values to Ͼ0.03 and 0.6 M, fully consistent with pre-association of the C-lobe and assignment of Ca 2ϩ sensor activity to the N-lobe.
Such pre-association of a Ca 2ϩ sensor protein and its target may be more prevalent than initially anticipated. Small angle X-ray scattering experiments show that CaM interacts with skeletal muscle myosin light chain kinase at substoichiometric Ca 2ϩ levels (29). It has long been known that the C-lobe of troponin C (30), which regulates troponin I in a Ca 2ϩ -dependent manner, has a Ca 2ϩ affinity 2 orders of magnitude greater than the N-lobe and is constitutively bound. In this case, the C-terminal lobe anchors the sensor to its target, and the Nterminal lobe serves as the Ca 2ϩ regulatory element, as we propose here for CDPK.
The Ca 2ϩ binding properties reveal that the mechanism of Ca 2ϩ activation of CDPK is clearly distinct from CaMKs. Our data imply that the Ca 2ϩ -dependent response of the CaM-LD is triggered by the binding of Ca 2ϩ to the N-lobe. The differences in Ca 2ϩ affinity between the two EF-hand regulatory domains in CDPK are significantly larger than the differences observed between the domains of CaM in the complex with myosin light chain kinase (Table I). For CaM, the transition from 0 to 4 Ca 2ϩ ions bound occurs over a narrow Ca 2ϩ concentration range (Table I), and all binding events appear to contribute to Ca 2ϩdependent activation of the kinase. In contrast, binding to the N-lobe of CDPK alone seems likely to serve as the trigger for activation of the kinase.
Interdomain Communication between the N-lobe and C-lobe of the CaM-LD-The NMR analysis of the regulatory apparatus provides detailed information about the coupling of the various Ca 2ϩ and J binding events. For example, in all NMR spectra at Ca 2ϩ levels above 0.1 mol eq, the only resonances observed correspond to fully Ca 2ϩ -loaded JC. Although this is due in part to the absence of peaks as a result of exchange broadening, the data strongly support cooperative binding of Ca 2ϩ . Consider the intensities of the resonances of three residues, Ile-471, Ile-543, Ile-577, which occupy sites in the Ca 2ϩ binding loops that are known to be highly sensitive probes of Ca 2ϩ binding (31). The similarity in the intensity profiles for the C-lobe pair (Ile-543 and Ile-577) (Fig. 5) agrees well with the cooperative Ca 2ϩ binding in the C-lobe, whereas the slower response of Ile-471 in the N-lobe is consistent with a response to Ca 2ϩ binding events in a lower affinity domain. Leu-441 in the J region correlates well with the C-lobe residues, consistent with association of J upon binding of Ca 2ϩ to the C-lobe.
Analysis of the populations of different Ca 2ϩ -bound states based on the simulations noted above indicate that the peak intensity for resonances in the N-lobe should be negligible until the Ca 2ϩ levels are sufficiently high to mostly saturate the C-lobe, i.e. at ϳ1.6 mol eq of Ca 2ϩ . However, for the majority of the N-lobe, resonances begin to appear at ϳ0.4 mol eq of Ca 2ϩ . These data strongly suggest additional levels of coupling of the Ca 2ϩ and J binding events.
One possible explanation for this observation is that the N-lobe may populate the open conformation of the Ca 2ϩ -saturated state before Ca 2ϩ is bound. In this context it is important to consider that the open conformation with exposed hydrophobic surface will be stabilized when J is bound. Moreover, once J is engaged by the C-lobe, the local concentration of J in the vicinity of the N-lobe is very high. A calculation using a simplified model gives an estimate of the maximum possible effective concentration of J equivalent to ϳ6 ϫ 10 9 M free junction peptide (20). Thus, although there is a very small equilibrium population of the N-lobe open state resulting from binding of Ca 2ϩ , the high concentration of J could trap this conformation. As a consequence, the N-lobe residues could exhibit chemical shifts consistent with the fully engaged JC complex even though the equilibrium population of the Ca 2ϩ -saturated Nlobe state is very small. Our results strongly imply that interdomain communication between the C-and N-lobes of CDPK is mediated through interactions with the adjacent J region.
Activation of CDPKs Differs from a Typical Calmodulin-dependent Protein Kinase-As expected, there are many similarities between the structures and properties of the CaM-LD and CaM. First, as in CDPK, CaM interacts with a CaM binding segment at the C terminus of the catalytic domain of several kinases (32). Second, diffusion NMR studies on the bimolecular complex of soybean CaM-LD in complex with a J peptide (33) and corresponding small angle X-ray scattering studies with the Arabidopsis JC used in this work 4 indicate the Ca 2ϩ -saturated state of JC has a compact structure like CaM target complexes. Third, the NMR (Fig. 1b) and CD (Fig. 2) analyses presented here reveal similarity in structure. Fourth, CaM can activate CDPKs in truncation mutants lacking the CaM-LD, i.e. analogous to a calmodulin-dependent protein kinase (although never to a fully activated state) (8). Fifth, there is precedent for CaM interacting with targets predominantly via the C-lobe, e.g. the plasma-membrane pump (34).
Despite these many similarities, our experiments indicate that the CaM-LD has several distinct features that result in a unique mode of action. Among these are the covalent attachment of the J region and the large difference in Ca 2ϩ affinities between the N-lobe and the C-lobe. These greatly alter the manner in which the CaM-LD interacts with J relative to how CaM is anticipated to interact with its targets. The high local concentration due to covalent attachment of the CaM-LD and J implies that the binding properties of the CaM-LD are finely tuned to enable it to serve in a regulatory capacity. The C-lobe of the CaM-LD makes the predominant interactions with J, and the flexible attachment of the two lobes of the CaM-LD by a short linker provides a means for coupling their independent activities. In fact, our studies reveal a significant degree of coordinated action of the two lobes of the CaM-LD. The measurement of the Ca 2ϩ and target binding affinities has, therefore, elucidated some of the key subtleties associated with the interaction with the autoinhibitory J region and activation of the kinase. In light of the likely occupation (closed circles) of the C-lobe sites at basal cytosolic Ca 2ϩ the "Basal State" (middle) shows the CaM-LD in a preformed complex with the junction. Activation of the kinase domain occurs when an increase in the levels of cellular Ca 2ϩ cause occupation of the N-lobe sites of the CaM-LD, relaying a conformational change to disengage the autoinhibitor from the kinase active. The autoinhibitor action line above the autoinhibitor is solid and dotted in the autoinhibited and active states, respectively, further depicting this contact. Two alternative conformations of the fully Ca 2ϩ -loaded state are presented; these schematically indicate that although the C-lobe interactions with the junction predominate, the extent of the N-lobe interactions with J at Ca 2ϩ -saturation are not fully understood.
Implications for the Mechanism of the Regulatory Apparatus-The biophysical and structural analyses presented here indicate that the CaM-like regulatory apparatus of CDPK undergoes multiple structural transitions. A model for CDPK activation, which summarizes the biophysical data described in this work, is shown in Fig. 8. The model uses the difference in affinity between the two domains to show separate functions for the lobes; at basal levels of Ca 2ϩ in the cell, the C-lobe of CDPK is Ca 2ϩ -bound and associated with the J region, whereas the N-lobe is free to respond to intracellular Ca 2ϩ signals and, thus, serves as the critical Ca 2ϩ -dependent regulatory element. A Ca 2ϩ sensor assignment for the N-lobe is consistent with the studies described above on a tomato CDPK and mutational studies conducted on a CDPK from P. falciparum (pf-CPK1) (4), where activation and a Ca 2ϩ -dependent alteration in conformation were both abolished by mutations that disrupt Ca 2ϩ binding residues in EF hands 1 and 2. In contrast, similar mutations that disrupted the C-lobe EF-hands had only minor effects on activation.
Although the C-lobe may not function directly as the Ca 2ϩ sensor, the two EF-hands are nevertheless highly conserved in most CDPKs. It is anticipated that the pre-association of J with the C-lobe is a general feature of CDPKs. It is conceivable that this interaction is a means to stabilize the regulatory apparatus by reducing the population of the conformationally heterogeneous Ca 2ϩ free state.
One obvious limitation of our studies is the characterization of Ca 2ϩ -dependent properties of the regulatory apparatus in isolation from the kinase domain it regulates. Ideally, the Ca 2ϩ affinities need to be measured for the intact protein and correlated with experiments to determine Ca 2ϩ thresholds for kinase activation. However, the analysis of such a complex system at a fundamental molecular level would be extremely complicated, particularly since there are no structures available of any Ca 2ϩ regulatory element bound to a kinase. The approach we and others have taken is to attack the problem by characterizing smaller fragments, with the goal of building up the whole from the parts.
In analyzing the binding data for CDPK, it is essential to consider that the interaction of the J region with the kinase has the potential to compete with the binding between the J region and the CaM-LD. Competition in the context of the holoenzyme will weaken the Ca 2ϩ affinities determined here for an isolated regulatory domain. In fact, as detailed above, lower Ca 2ϩ affinity than we have measured would not compromise the ability of the CaM-LD to serve as a Ca 2ϩ sensor.
A second critical consideration is that Ca 2ϩ activation thresholds of kinases are actually fairly variable. The clear dependence on different peptide substrates shows there is a strong coupling between the peptide substrate, catalytic domain, the junction region, and the CaM-LD. For example, with soybean CDPK isoform ␣, the lowest Ca 2ϩ concentration observed for half-maximal (K 0.5 ) activity was 60 nM, using syntide-2 as a substrate (35). The activation threshold for this same isoform increased by 2 orders of magnitude (i.e. 6 M) when provided histone as a substrate. Given the potential of interconnecting domains to alter the activation threshold, our analysis of the isolated JC represents the foundation upon which other studies of Ca 2ϩ -dependent regulation of CDPK can be built.
In addressing the functional role of plant CDPKs, it is important to ask why so many isoforms co-exist in a single species (34 in the Arabidopsis genus). A potential explanation is that different isoforms have co-evolved with biochemically distinct CaM-LDs that provide different Ca 2ϩ activation thresholds (1). Different Ca 2ϩ activation thresholds have been reported for different isoforms. Some CDPK isoforms have one or more divergent EF hands, with eight Arabidopsis CDPK-related kinases having all four EF-hands predicted to be non-Ca 2ϩ binding. In considering the diversity of CDPKs and related kinases, the biophysical analysis presented here emphasizes the likely role of the differing function of the two lobes of the CaM-LD.