Tobacco Calcium-dependent Protein Kinases Are Differentially Phosphorylated in Vivo as Part of a Kinase Cascade That Regulates Stress Response*

In vivo phosphorylation sites of the tobacco calcium-dependent protein kinases NtCDPK2 and NtCDPK3 were determined in response to biotic or abiotic stress. Stress-inducible phosphorylation was exclusively located in the variable N termini, where both kinases were phosphorylated differentially despite 91% overall sequence identity. In NtCDPK2, serine 40 and threonine 65 were phosphorylated within 2 min after stress. Whereas Thr65 is subjected to intra-molecular in vivo autophosphorylation, Ser40 represents a target for a regulatory upstream protein kinase, and correct NtCDPK2 membrane localization is required for Ser40 phosphorylation. NtCDPK3 is phosphorylated at least at two sites in the N terminus by upstream kinase(s) upon stress stimulus, first at Ser54, a site not present in NtCDPK2, and also at a second undetermined site not identical to Ser40. Domain swap experiments established that differential phosphorylation of both kinases is exclusively determined by the respective N termini. A cell death-inducing response was only observed upon expression of a truncated variant lacking the junction and calcium-binding domain of NtCDPK2 (VK2). This response required protein kinase activity and was reduced when subcellular membrane localization was disturbed by a mutation in the myristoylation and palmitoylation site. Our data indicate that CDPKs are integrated in stress-dependent protein kinase signaling cascades, and regulation of CDPK function in response to in vivo stimulation is dependent on its membrane localization.

respectively. They possess a conserved modular structure of four domains (VKJC): an N-terminal variable domain (V), which for many isoforms contains a myristoylation and palmitoylation motif mediating membrane localization; a serine/ threonine protein kinase domain (K); an autoinhibitory junction peptide (J); and a calcium-sensing calmodulin-like domain (C) at the C terminus (1). These structural features as well as published biochemical and functional data suggest that the regulatory switch for CDPK in vivo activation includes (i) the binding of calcium leading to the release of the autoinhibitory domain from the active site, (ii) CDPK protein (auto)phosphorylation, and (iii) specific subcellular localization (2)(3)(4)(5)(6).
CDPKs have been associated with the regulation of diverse processes of plant growth and development, nutrient primary metabolism, and signaling in biotic and abiotic stress responses and water/ion transport (7)(8)(9). Only recently, the analysis of genetic knock-out mutants in Arabidopsis allowed the assignment of a biological function in abscisic acid signaling, stomatal aperture,andsaltstresstolerancetoArabidopsisCPK4(calciumdependent protein kinase 4) and CPK11 (10), CPK3 and CPK6 (11), and CPK23 (12), respectively. However, none of these studies addressed the biochemical mechanism of how the CDPK enzyme itself is regulated by (auto)phosphorylation in planta.
In vitro autophosphorylation of CDPKs using recombinant or purified enzymes was observed for almost all CDPKs studied so far. The corresponding in vitro autophosphorylation sites (P-sites) have been mapped for several enzymes (13)(14)(15)(16) and were found with high frequency in the variable N termini (17). It has not yet been experimentally addressed whether CDPK autokinase reactions follow an intramolecular (unimolecular: kinase and P-site on the same polypeptide) or intermolecular (bimolecular: kinase and P-site on different polypeptide) mechanism or both. In vivo P-sites of CDPKs have not been mapped comprehensively, although some P-sites of Arabidopsis CDPKs have been identified in general plant phosphoproteomics analyses (18,19). For NtCDPK2, one in vivo P-site has been described (20). The in vivo relevance and function of CDPK phosphorylation remains to be shown. It has been proposed that CDPK phosphorylation is required for enzyme activation (21)(22)(23). For CDPKs from soybean and groundnut, it was demonstrated that in vitro autophosphorylation did not relieve kinase activity from calcium dependence (22,24) as observed for mammalian calmodulin-dependent kinase II (25) or calmodulin-dependent kinase VI (26).
CDPK isoforms from tobacco (Nicotiana tabacum) were shown to participate in plant defense signaling. Treatment of tobacco plants or cell cultures containing the Cf-9 (Cladosporium fulvum resistance gene 9) disease resistance gene from tomato with fungal pathogen-derived AVR9 (C. fulvum avirulence gene 9) peptide resulted in a rapid phosphorylation and activation of NtCDPK2 within a few minutes after elicitation (21,27). Transient loss-of-function studies in Nicotiana benthamiana showed that plants silenced for the NtCDPK2 subfamily by virus-induced gene silencing were compromised in the induction of AVR9/CF-9-dependent hypersensitive response-like cell death symptoms (21). In complementary gain-of-function experiments, the ectopic expression of a truncated NtCDPK2-VK (VK2) variant lacking its regulatory junction-and calcium-binding domains triggered enhanced stress responses upon a mild hypo-osmotic shock (28).
Here we present a detailed analysis of stress-dependent phosphorylation of NtCDPK2 and NtCDPK3 proteins in vivo. Transient expression in N. benthamiana, rapid StrepII tag-mediated purification, and quadrupole time-of-flight mass spectrometry analyses were coupled with site-directed mutagenesis approaches to map phosphorylated amino acids, to study the time course and hierarchical order of phosphorylations, and to further characterize the phosphorylations. P-sites of both NtCDPKs were found to reside mainly in the variable N termini. Inter-and intramolecular autophosphorylation events as well as phosphorylations catalyzed by other kinases were identified, the latter requiring an intact myristoylation and palmitoylation motif. Our data demonstrate that CDPKs are integrated in protein kinase signaling cascades. Ectopically expressed in N. benthamiana, P-site variants still triggered a cell death response when introduced into VK2. Membrane localization of the enzyme was required for this response.

EXPERIMENTAL PROCEDURES
Cloning and Constructs-Clones for NtCDPK2 full-length constructs with C-terminal HA-Strep tag for plant expression and His tag for expression in Escherichia coli were described previously (14,20). A triple-Myc-tagged NtCDPK2 construct was also described as well as CDPK silencing and the corresponding construct (21). The VK2 construct (amino acids 1-375) used in this study was generated by PCR with primers 3 and 5 (supplemental Table S1) using a full-length NtCDPK2 clone as template and cloned into pXCS-HAStrep (20) via the EcoRI and XmaI sites. To prevent bacterial expression, the internal MfeI-NcoI fragment of the VK2 cDNA was exchanged against the corresponding genomic fragment containing the first two NtCDPK2 introns. The NtCDPK3 construct was amplified with primers 3 and 271 (supplemental Table S1) and cloned into pXCS-HAStrep using the EcoRI and XmaI sites. The VK3 construct (amino acids 1-377) was generated by PCR using primers 3 and 296 and cloned into pXCS-HAStrep using the EcoRI and XmaI sites. Because this construct was toxic to the agrobacteria, two introns amplified from genomic tobacco DNA with primers 344 and 345 were cloned into the cDNA using the MfeI and NcoI restriction sites present in the NtCDPK3 coding sequence. Site-directed mutagenesis was performed according to Ref. 29. The PEST domain deletions (amino acids [27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42] were introduced by overlap PCR (30). N-terminal domain swap constructs were generated by mutually exchanging an EcoRI-DraI fragment coding for the complete N terminus and part of the kinase domain between NtCDPK2 and NtCDPK3 (in pXCS-HAStrep). Within the exchanged part of the kinase domain, all amino acids are identical between both kinases. Supplemental Table S1 lists all primers used in this study.
Separation of microsomal fractions and supernatants for the analysis of myristoylation and palmitoylation mutants was achieved by ultracentrifugation of the extract obtained in buffer A in a Beckman TLA 120.1 rotor at 220,000 ϫ g for 30 min (4°C). The pellet was resuspended in extraction buffer B by sonication and centrifuged again at 20,000 ϫ g for 10 min.
For in vitro kinase reactions, 20 l of enzyme affinity purified from leaves was mixed with 4 l of water and 6 l of 5ϫ kinase buffer with or without ATP (200 mM HEPES, pH 7.5, 0.5 mM CaCl 2 , 50 mM MgCl 2 , 50 mM ATP or water) and incubated for 15 min at 30°C. The reaction was stopped by adding 15 l of 3ϫ SDS-loading buffer.
Mass Spectrometry and Site-directed Mutagenesis-NtCDPK2 and NtCDPK3 protein purification from plant material for mass spectrometry (MS) analysis was performed as described (20). The isolation of NtCDPK2 from E. coli was carried out according to Glinski et al. (14). Mass spectra were scanned manually for the presence of potential phosphopeptides. Candidate peptides were analyzed by tandem MS to verify and map the phosphorylated amino acid. In an alternative approach, tandem MS analysis was performed for all peptides that contain a serine or threonine, taking into account one or two possible phosphorylations. The resulting data were compared manually to spectra from the non-phosphorylated peptides, where available, or to a theoretical prediction of fragment ions.
Autophosphorylation sites in kinases isolated from leaves were identified by comparing spectra of catalytically inactive NtCDPK2 and NtCDPK3 variants (D241A and D238A mutants, respectively) with spectra of the corresponding wildtype forms. In order to confirm autophosphorylation sites, Histagged NtCDPK2 was expressed in E. coli and purified as described (14). The isolated recombinant enzyme was subjected to in-gel tryptic digest and quadrupole time-of-flight MS analysis. Previous in vitro incubation with ATP was not necessary because NtCDPK2 autophosphorylated during the bacterial expression. NtCDPK3 could not be expressed in E. coli, probably because it was toxic to the bacteria (also observed for PiCDPK2 (31)). P-sites found in NtCDPK2 expressed in E. coli that were absent in kinase-inactive mutants expressed in planta represent intramolecular auto-P-sites. Those P-sites found in both types of material are intermolecular auto-P-sites, although they may additionally also be phosphorylated by an intramolecular mechanism.
In cases where tandem MS analysis could not resolve the site of phosphorylation in a peptide, variants of NtCDPK2 with alanine substitutions of possible P-sites were subjected to MS analyses. The following variants were analyzed: S18A, S20A, S97A, T102A, S97A/T102A, and S104A. Semiquantitative evaluation of mass spectra to determine inducibility of phosphorylation by stress was performed by comparing ratios of phosphorylated peptide signals with total peptide (phosphorylated and non-phosphorylated) signals before and after stress.
Phenotypic Analysis of VK Constructs-The first fully expanded leaf on 3-4-week-old N. benthamiana plants was infiltrated with one of the following constructs: VK2, VK2-D241A (kinase inactive), VK2-G2A, VK2-AA (S40A/T65A), and VK2-DD (S40D/T65D). Agrobacteria carrying the respective kinase constructs at an A 600 nm of 0.2 were mixed with a concentrated solution of agrobacteria carrying the p19 silencing suppressor from tomato bushy stunt virus (32) reaching a final optical density of 0.3 for the mixture. Infiltration of this mixture was carried out in the evening, and plants were placed in a long day (16 h of light) growth chamber with strong lighting (300 microeinsteins m Ϫ2 s Ϫ1 ) on the following day. One leaf disc was taken on day 2 as control for protein expression assayed by Western blot. On day 3, the developing cell death phenotype was assessed by documenting autofluorescence with a macroscope (Leica Z16APOA) equipped with a UV lamp and a long pass green fluorescent protein filter (I3) and a Leica DSC420C camera. Rela-tive cell death was quantified by using the histogram function of Adobe Photoshop. For each image, the sum of pixels in the green channel with brightness of 20 or above was divided by the total number of pixels in this channel. Areas where the infiltration process had damaged the leaf were excluded. The resulting ratios were multiplied by 1000, and mean and S.D. values were calculated. Statistical significance was evaluated using an analysis of variance coupled with a Newman-Keuls post-test.

NtCDPK2 and NtCDPK3 in Vivo Phosphorylation and Dephosphorylation Kinetics after Hypo-osmotic Stress
The in vivo phosphorylation of NtCDPK2 after biotic or abiotic stress results in an altered electrophoretic mobility of the kinase (21,27). Here we show that HA-StrepII-tagged NtCDPK2 as well as NtCDPK3, transiently expressed in N. benthamiana, each display two electrophoretic mobility shifts after hypo-osmotic stress caused by the infiltration of water into the leaves ( Fig. 1), suggesting that at least two distinct amino acids per kinase are subject to stress-dependent phosphorylation. In NtCDPK2, both mobility shifts were observed simultaneously within 1-2 min after stress (Fig. 1A, s2␣ and s2␤). Such fast kinetics was seen for only one of the shifts of NtCDPK3 (s3␣), whereas the second (s3␤) appeared later between 8 and 16 min. Both kinases returned to their resting state within 2-4 h (Fig. 1B), which may be achieved either by proteolytic degradation of the phosphorylated CDPK coupled with de novo protein synthesis or by enzymatic dephosphorylation. Protein degradation was not observed because the overall amount of CDPK remained unchanged over time courses (Fig.  1B), even when de novo protein synthesis was inhibited by cycloheximide (not shown). However, the reversal of NtCDPK2 and NtCDPK3 to the non-phosphorylated state was sensitive in vivo to the phosphatase inhibitor cantharidin (30 M; Fig. 1C). For NtCDPK2, only one site was affected (corresponding to s2␣), whereas dephosphorylation of both sites was affected in NtCDPK3. A complete shift of NtCDPK2 and NtCDPK3 into the s2␣ and s3␤ forms, respectively, was accomplished with 90 M cantharidin after 2-4 h (not shown). The dose effect of phosphatase inhibitor on the phosphorylation status of the kinases suggests that phosphorylation and dephosphorylation are in a dynamic equilibrium. This may explain why several CDPK phosphorylation states can be observed in vivo.

Identification of in Vivo Phosphorylation Sites (P-sites) in NtCDPK2 and NtCDPK3
In truncated variants of NtCDPK2 and NtCDPK3 only containing the variable and kinase domains (VK2 and VK3), phosphorylation-dependent mobility shifts were still observed, indicating that the corresponding P-sites reside within these domains. Interestingly, phosphorylation became stimulus-independent in VK2 but not in VK3 (not shown). NtCDPK2 and NtCDPK3 contain a PEST domain in their N termini. Such domains were implicated in protein phosphorylation followed by protein degradation (33). A deletion of the 16-amino acid PEST domain in VK2 (amino acids 27-42) resulted in a lack of the phosphorylation shift s2␣ ( Fig. 2A). An alanine scan of all possible P-sites within this region showed that phosphorylation at Ser 40 is responsible for the s2␣ shift because it is absent in the S40A mutant enzyme (Fig. 2B). Thus, NtCDPK2 is phosphorylated in vivo at Ser 40 in a stress-dependent manner, whereas other serines and threonines in the vicinity have no influence on this phosphorylation. A deletion of the PEST domain in NtCDPK3 (amino acids [27][28][29][30][31][32][33][34][35][36][37][38][39][40][41][42] resulted in a lack of s3␤ (Fig. 2C).
For a more general P-site search, we resorted to nanospray quadrupole time-of-flight MS. The kinases were isolated via the StrepII tag (20) from unstressed and stressed leaves and prepared for MS. The resulting spectra were scanned for peptides that displayed the characteristic mass increase associated with phosphorylation. Tandem mass spectrometry was used to confirm and map phosphorylation sites. By comparing spectra from elicited and non-elicited kinase, we identified differential stress-dependent phosphorylation events. An example for NtCDPK3 is shown in Fig. 2D. Here peaks for the phosphorylated triply charged peptide T5-6 (amino acids 43-69; m/z 1033.18; sodiated form at m/z 1040.51) are only observed in the spectrum from the kinase isolated from an elicited leaf. The corresponding non-phosphorylated peptide (m/z 1006.51) is observed in both spectra, although it appears to be more abundant in the spectrum obtained from the unstressed leaf. From the four possible P-sites on peptide T5-6P (Thr 45 , Ser 48 , Ser 54 , and Thr 66 ), only serine 54 was mapped as an exclusive site of modification by tandem MS analysis (supplemental Fig. S1). Ser 54 represents an in vivo P-site in NtCDPK3 phosphorylated in a stimulus-dependent manner. Serine 54 is unique to NtCDPK3 and not present in NtCDPK2 (Fig. 2E).
In total, five in vivo phosphorylation sites were found in NtCDPK2 (supplemental Table S2; coverage 81.6%), and four in vivo sites were found in NtCDPK3 (supplemental Table S3; coverage 76.0%). We have used a variety of approaches to characterize these phosphorylations (see below and see "Experimental Procedures" for details), including the expression of kinase-inactive variants in planta to distinguish autophosphorylation from phosphorylation by different kinases and MS analysis of kinases expressed in E. coli (allowing only autophosphorylation). Most identified P-sites are located in the N terminus (Fig.  2E), whereas no modifications were found in the kinase and junction domains. The pattern of phosphorylation in the N terminus is different between both kinases despite an 82% amino acid identity in this region (overall, both kinases are 91% identical and 94% similar). The only common N-terminal phosphorylation site is serine 97 of NtCDPK2 (Ser 94 in NtCDPK3; Fig. 2E). In NtCDPK3, phosphorylation at Ser 20 was only observed after an in vitro autokinase reaction. This autophosphorylation is also present in the Arabidopsis orthologue AtCPK1 (17). A constitutive phosphorylation in the calmodulin-like domain at Ser 572 (NtCDPK2) and Ser 569 or Ser 567 (NtCDPK3) has also been reported for AtCPK16 at Ser 568 (17).

Characterization of Stress-inducible P-sites by Site-directed Mutagenesis
NtCDPK2-The altered electrophoretic mobility of the kinases as a result of phosphorylation may indicate significant structural changes that could be important in regulation. The wild-type enzyme of NtCDPK2 displayed two mobility shifts after stress (Fig. 3A, lanes 1 and 2), whereas the S40A and T65A variants lacked the lower (s2␣; lanes 3 and 4) or upper shift (s2␤; lanes 5 and 6). This demonstrates that Ser 40 and Thr 65 are phosphorylated independently of each other. The S40A/T65A double mutant lacked both shifts (lanes 7 and 8), whereas a phosphorylation mimic by S40D and T65D either in individual (not shown) or combined variants caused a constitutive shift (lanes 9 and 10).
NtCDPK3-For NtCDPK3, two shifts were observed within 15 min after stimulus (Fig. 3B, lanes 1 and 2). Introducing the mutation S40A (as identified in NtCDPK2) delayed but did not abolish the mobility shift s3␤ (Fig. 3, B (lanes 3 and 4) and C). Only a deletion of the complete PEST domain abolished s3␤ (Fig. 2C). Unfortunately, further fine mapping by MS (possible P-sites are Ser 28 , Thr 32 , Ser 34 , Thr 35 , Thr 36 , and Thr 45 ) was not possible because the corresponding peptides were not observed in the spectra. Introducing the mutation T66A (corresponding to Thr 65 auto-P-site responsible for s2␤ in NtCDPK2) had no impact on the mobility shifts (Fig. 3B, lanes 5 and 6). In addition, tandem MS analysis of peptide T5-6P (supplemental Fig. S1) comprising residue Thr 66 clearly demonstrated exclusive stress-dependent phosphorylation at Ser 54 and not Thr 66 . Thus, despite nearly identical amino acid context around this threonine in both kinases (Fig. 2E), it is only accepted as an autophosphorylation site in NtCDPK2 and not in NtCDPK3. However, in AtCPK1, an Arabidopsis orthologue of NtCDPK2 and NtCDPK3, in vitro autophosphorylation at the corresponding threonine was observed (17). This demonstrates that for CDPKs, primary sequence conservation alone is not sufficient for P-site prediction. Introducing the mutation Ser 54 had a strong effect on the NtCDPK3 shift pattern, leading to the lack of both stress-dependent mobility changes, s3␣ and s3␤ (Fig. 3B, lanes 7 and 8), whereas the PEST domain deletion mutant lacking amino acids 27-42 (thus containing Ser 54 ) still displayed s3␣ (Fig. 2C), appearing within 5 min after stress. In summary, our data suggest that phosphorylation at Ser 54 cor- responds to s3␣ and is a prerequisite for further timely delayed phosphorylation of the PEST domain corresponding to mobility shift s3␤. In contrast to NtCDPK2, where phosphorylation events at Ser 40 and Thr 65 are independent of each other (Fig.  3A), in vivo phosphorylation in NtCDPK3 appears to follow a hierarchical order.
NtCDPK2 and NtCDPK3 were identified in plant defense signaling in a Cf-9/Avr9 gene-for-gene interaction. Rapid NtCDPK2 phosphorylation was observed upon elicitation of Cf-9 tobacco with AVR9 (21,27). Therefore, we expressed StrepII-tagged NtCDPK2 and NtCDPK3 plus their variants with mutations in the identified phosphorylation sites transiently in Cf-9 tobacco plants (N. tabacum) and investigated the phosphorylation pattern by Western blot after infiltration of intercellular fluid containing the fungal elicitor AVR9 into the apoplast. All NtCDPK2 and NtCDPK3 mobility shifts discussed above (and their absence when sites were blocked by mutation) observed in the N. benthamiana system after hypo-osmotic stress could be confirmed in N. tabacum in the context of the Cf-9/Avr9 gene-for-gene interaction (data not shown).

Intramolecular NtCDPK2 Autophosphorylation at Thr 65
Non-elicited NtCDPK2 purified via the StrepII tag and subjected to an in vitro autokinase reaction displayed the form s2␤ (phosphorylation at Thr 65 ) but not s2␣ (phosphorylation at Ser 40 ; Fig. 4C). When a kinase-inactive variant of NtCDPK2 (mutant D241A) was expressed and the leaf was subjected to a stress stimulus, only s2␣ (phosphorylation at Ser 40 ) was observed, whereas s2␤ (phosphorylation at Thr 65 ) was absent (Fig. 3A, lanes 11 and 12). This identifies Thr 65 (but not Ser 40 ) as an NtCDPK2 autophosphorylation site. In addition, it shows that Thr 65 in NtCDPK2-D241A is not accepted for intermolecular autophosphorylation mediated by endogenous NbCDPK2 from N. benthamiana. In planta co-expression experiments of kinase-inactive NtCDPK2-D241A (HA-Strep-tagged) together with kinase-active NtCDPK2 (triple-Myc-tagged) confirmed that Thr 65 autophosphorylation is an intramolecular reaction because the inactive variant was not phosphorylated in trans (supplemental Fig. S2). An alternative model fitting this data is that an unidentified intramolecular autophosphorylation event is required to make Thr 65 accessible for phosphorylation.

NtCDPK2 Phosphorylation at Ser 40
In planta phosphorylation at Ser 40 (s2␣), in contrast to Thr 65 , still occurs when the kinase-inactive NtCDPK2-D241A variant is expressed (Fig. 3A, lanes 11 and 12). This excludes an intramolecular autophosphorylation of the enzyme. We next investigated whether an intermolecular autophosphorylation mediated by endogenous NbCDPK2 may occur. In a combined reverse genetics approach, the closely related endogenous CDPKs in N. benthamiana, NbCDPK2 and NbCDPK3, were silenced by virus-induced gene silencing targeted to the calmodulin-like domain (21). Subsequently, a kinase-inactive truncated variant of NtCDPK2-VK-D241A lacking the targeted domain was expressed in the silenced leaves. In two NbCDPKsilenced plants, in vivo stress-dependent phosphorylation at Ser 40 (s2␣) could still be detected (Fig. 4A, s1 and s2) like in a non-silenced control plant (ns). This documents that Ser 40 of NtCDPK2 is phosphorylated by an unknown upstream kinase in response to a stress stimulus. Consistently, we never observed Ser 40 phosphorylation in NtCDPK2 subjected to in vitro protein kinase reactions (Fig. 4C). Remarkably, Ser 40 phosphorylation is constitutively present in the truncated NtCDPK2-VK2 variant ( Fig. 2A) but only observed after elicitation in the corresponding kinase-inactive VK2-D241A mutant (Fig. 4A). These data suggest that NtCDPK2 activity feeds back on the kinase or phosphatase that regulates Ser 40 phosphorylation and that the cell has no alternative means for CDPK inactivation when the regulatory CDPK domains are absent. Consistently, the active VK2 variant caused enhanced plant stress and defense responses in gain-of-function experiments even in the absence of elicitation (28).

NtCDPK3 Phosphorylation at Ser 54 Only Occurs in Vivo
NtCDPK3 phosphorylation at Ser 54 (possibly s3␣) and within the PEST domain (possibly s3␤) are probably catalyzed by an upstream kinase(s) because both shifts were detected in a kinase-inactive variant of NtCDPK3 (Fig. 3B, lanes 9 and 10), and MS analysis confirmed the presence of phosphorylated Ser 54 in the kinase-inactive NtCDPK3 variant (not shown). In accordance, mobility shifts of the VK3 variant remained stimulus-dependent (not shown), although this truncated kinase can autophosphorylate constitutively. We purified NtCDPK3 from non-elicited and elicited leaves and subjected both preparations to an in vitro kinase reaction. Phosphorylation at Ser 54 could not be generated in vitro, as indicated by the lack of the s3␣ mobility shift (supplemental Fig. S3) and by MS analysis (not shown). Similarly, PEST domain phosphorylation (s3␤) was not observed in vitro. However, in vitro autophosphorylation caused a mobility shift never observed in vivo (supplemental Fig. S3), highlighting the potential risk of generating artifacts by in vitro kinase reactions often employed to map autophosphorylation sites. The data validate our approach to identify the in vivo phosphorylation pattern.

The Phosphorylation Pattern Is Governed by the N-terminal Variable Domains
We next conducted domain swap experiments in which the N-terminal variable domains of NtCDPK2 (amino acids 1-116) and NtCDPK3 (amino acids 1-113) were mutually exchanged, generating two hybrid kinases NtCDPK2-3 and NtCDPK3-2 (supplemental Fig. S4). Upon transient expression in N. benthamiana and stress exposure, the phosphorylation patterns visualized by the shifts were identical to those of the CDPK from which the N-terminal domain originated and did not depend on the protein kinase domain present (supplemental Fig. S5). In agreement, in NtCDPK2-3 and NtCDPK3-2 hybrid variants, the mutations S40A and T65A (NtCDPK2) and S54A (NtCDPK3) compromised the corresponding phosphorylation shifts (not shown). These data suggest that structural constraints of the N termini determine the phosphorylation pattern independent from the kinase domain.

Membrane Localization Is Required for NtCDPK2 and NtCDPK3 Phosphorylation by Upstream Kinases
When the myristoylation or palmitoylation motif of NtCDPK2 was interrupted by introducing a G2A or C5S mutation, respectively, the kinase no longer remained associated with the microsomal pellet but was mainly located in the supernatant (Fig. 5). Remarkably, phosphorylation at Ser 40 was lost in these acylation mutant variants. Correct subcellular membrane localization is therefore required for in vivo stimulus-dependent NtCDPK2 phosphorylation by the upstream kinase. A similar result was obtained for the NtCDPK3 (G2A) mutant, although partial Ser 54 phosphorylation (but not PEST domain phosphorylation) was retained (not shown).

Impact of Phosphorylation Sites for NtCDPK2 in Vitro Kinase Activity
To elucidate whether the identified in vivo phosphorylation sites affect the enzyme activity of NtCDPK2 and NtCDPK3, wild-type enzymes and variants carrying site-specific mutations mimicking or blocking these phosphorylations were isolated from stressed leaves and were subjected to an in vitro kinase assay using syntide 2 as substrate. Kinase activity toward syntide 2 was not altered by the mutations (supplemental Fig. S6).

NtCDPK2 Function in Cell Death Response Requires Protein Kinase Activity and Membrane Localization
We could previously show that the expression of a truncated variant VK2 (but not VK3) resulted in enhanced plant defense responses in N. benthamiana, including the induction of hypersensitive response-like cell death symptoms (28). These responses were accelerated by a secondary mild abiotic stress stimulus. To test whether the identified phosphorylations affect NtCDPK2 in vivo function, we generated VK2 variants carrying amino acid substitutions that either block or mimic phosphorylation at S40A/T65A (VK2-AA) and S40D/T65D (VK2-DD), a delocalized variant VK2-G2A, and the kinaseinactive control VK2-D241A. These constructs were expressed in N. benthamiana, and leaves were assessed for the induction of cell death symptoms in the absence of any further stress stimulus by the autofluorescence of necrotic lesions (supplemental Fig. S7, A and B). In two biological replicates, each with 10 plants per construct, a loss of VK2mediated cell death development was observed with the kinase-inactive VK2-D241A variant, and only weak symptoms developed with the delocalized variant. In contrast, slightly enhanced cell death symptoms were observed by the phosphorylation mimic VK2-DD variant and, remarkably, also by the variant VK2-AA. However, due to the large biological variance, the enhanced response of the phosphorylation mutants could at best be observed with marginal statistical significance (analysis of variance). These data show that kinase activity and correct membrane localization mediated by N-terminal acylation are required for biological function of NtCDPK2. Additionally, these results indicated that amino acid substitutions of Ser 40 and Thr 65 phosphorylated in vivo may alter NtCDPK2-dependent stress responses. This suggests that autophosphorylation and upstream kinase-mediated phosphorylation may be functionally important for stimulus-dependent CDPK regulation.

DISCUSSION
The mapping of in vivo P-sites performed in this study showed that NtCDPK2 and NtCDPK3 were almost exclusively phosphorylated in their respective N termini in a differential and stimulus-dependent manner. In addition, the phosphorylation patterns were determined by the respective N termini and not by the kinase domain, even in the case of autophosphorylations (supplemental Fig. S5). These data confirm the notion that specificity determinants of CDPKs are encoded within the variable N termini of these kinases (34).
We show that NtCDPK2 and NtCDPK3 are not only autophosphorylated but are also substrate to other (upstream) protein kinases and therefore are components of protein kinase signaling cascades. Although calcium binding is a necessary prerequisite for CDPK activation, our data indicate that also kinase phosphorylation is mechanistically involved for the CDPK to become biologically functional. The phosphorylation of the NtCDPKs by unidentified upstream kinase(s) was observed to be stimulus-dependent (Fig. 1), suggesting that these kinases will also be activated during a biotic or abiotic stimulation. Additionally, the upstream kinases are probably membrane-associated because the corresponding phosphorylations were compromised when the membrane localization of the NtCDPKs was disturbed (Fig. 5). Potential candidates for CDPK-phosphorylating kinases may be found among the receptor-like kinases or the cytoplasmic kinases associated with membrane receptors (required, for example, in the case of the Cf-9 disease resistance gene in the Avr9/Cf-9 interaction (35)). Alternatively, members of the calcineurin B-like protein-interacting protein kinase/calcineurin B-like protein family may phosphorylate CDPKs. Membrane localization of calcineurin B-like protein-interacting protein kinases occurs via myristoylation and palmitoylation of the calcineurin B-like protein calcium sensors, which form stimulus-specific complexes with selected calcineurin B-like protein-interacting protein kinases (36,37).
What is the functional role of CDPK phosphorylation? At our test conditions, full-length NtCDPK2 enzymes carrying specific amino acid substitutions at P-sites displayed no difference in phosphorylating activity toward the synthetic substrate syntide 2. However, it is possible that phosphorylation events so far unidentified are required for full enzyme activation. In other stress-related protein kinases (e.g. calcineurin B-like proteininteracting protein kinases or mitogen-activated protein kinases), phosphorylation occurs within the activation loop and causes an increase in kinase activity. Furthermore, CDPK phosphorylation may facilitate changes in the calcium binding affinity of the enzyme (e.g. allowing target substrate phosphorylation at lower calcium concentrations (24), rendering the enzyme in a sensitized state). Interestingly, exposure to a primary stimulus (incubation with abscisic acid) appears to leave guard cells in a status of enhanced calcium sensitivity (priming) for stomatal closure (38), and Arabidopsis CPK3 and CPK6 were shown to affect S-type anion channel activity in guard cells (11). One may speculate that CDPKs are part of a calciumsensing regulatory circuit during ion channel regulation (39) and that the CDPK phosphorylation state (controlled by autophosphorylation and/or an upstream kinase) may be part of the postulated priming mechanism. Alternatively, phosphorylation could cause structural changes (as indicated by the electrophoretic mobility shifts) that may be a prerequisite to access and sequester in vivo phosphorylation targets (or to release them). Unfortunately, in vivo substrates for NtCDPK2 and NtCDPK3 are unknown.
A recent study also based on the N. benthamiana expression system identified plasma membrane NADPH-oxidase isoforms as targets for potato StCDPK4 and StCDPK5 (40). The ectopic expression of a truncated StCDPK5-VK variant resulted in an enhanced synthesis of reactive oxygen species in N. benthamiana leaves not observed in a variant lacking an intact myristoy-lation or palmitoylation motif. Like NtCDPK2, the potato CDPK may require correct localization for phosphorylation mediated by upstream kinases and to gain access to the target. However, unlike NtCDPK2-VK, StCDPK5-VK expression alone did not trigger cell death symptoms.
A mild stimulus (e.g. apoplast flooding) can accelerate cell death mediated by the VK2 variant but has no additional effect on VK2 phosphorylation at Ser 40 and Thr 65 (Fig. 5). Such a stimulus probably activates additional parallel signaling pathways, and the observed accelerated cell death may be the result of a concerted response reaction.
Our data suggest that the truncated gain-of-function variant NtCDPK2-VK (but not NtCDPK3-VK) feeds back on the upstream kinase (and/or phosphatase), causing a constitutive partial Ser 40 (and Thr 65 ) phosphorylation and adopting a kinase-active conformation also in the absence of a stimulus ( Figs. 2A and 5). Continuous CDPK signaling and finally cell death development are the result. These responses appear on average slightly enhanced in the phosphorylation mimic variant VK2-DD but also in variant VK2-AA (supplemental Fig. S7), although these changes could not be measured with statistical significance. Gain-of-function activities have also been reported for both the T271D and the T271A amino acid substitution at the autophosphorylation site in the calcium-and calmodulin-regulated protein kinase DMI3 from Medicago truncatula. Both variants resulted in functional proteins and were able to complement the dmi3 mutant for nodulation development (41). Furthermore, a T265I substitution in the corresponding calcium-and calmodulin-regulated protein kinase protein from Lotus japonicus was identified as the causal mutation resulting in spontaneous nodulation of the snf1 mutant (42). This indicates that phosphorylations do not only modulate protein properties by simply introducing a negative charge, and several functional models may be proposed in the case of NtCDPK2. (i) Phosphorylation may be required for effective kinase inactivation (e.g. by binding to an inhibitor). (ii) Structural changes in the N termini may alter the kinase binding affinity to the target(s). (iii) Binding affinity to other interactors (e.g. the upstream kinase) may be changed, reducing or enhancing the signal flux in feedback regulatory circuits.
The identification of in vivo phosphorylation sites is a crucial first step in understanding the regulation of CDPK function within a biological context. In the absence of tobacco mutant lines in CDPK genes, experiments in this report were based on gain-of-function analyses relying on ectopic expression in planta. Using cell death readout, this experimental system is at its limits for resolving altered biological responses of phosphorylation mutants that do not directly affect kinase activity. Currently, the number of reports in which Arabidopsis mutant lines in CDPK genes are linked to stress-related phenotypes is slowly increasing (10 -12). Thus, future analysis of posttranslational modifications of CDPKs will probably include stable expression of kinase variants in the background of the corresponding mutant plants. The biological function of CDPK phosphorylation can then be addressed in vivo within the native signaling network triggering plant stress responses.