S

Journal of Molecular Modeling (2024) 30:248 https://doi.org/10.1007/s00894-024-06037-5 BRIEF REPORT Structural and molecular dynamics simulation studies of CBL‑interacting protein kinase CIPK and its complexes related to plant salinity stress Prabir Kumar Das1 · Tanya Bhatnagar2 · Sanhita Banik2 · Sambit Majumdar2 · Debajyoti Dutta2 Received: 19 January 2024 / Accepted: 20 June 2024 © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2024 Abstract Context Calcium-dependent signaling in plants is responsible for several major cellular events, including the activation of the salinity-responsive pathways. Calcium binds to calcineurin B-like protein (CBL), and the resulting CBL-Ca2+ complex binds to CBL-interacting protein kinase (CIPK). The CBL-CIPK complex enhances the CIPK interaction with an upstream kinase. The upstream kinase phosphorylates CIPK that, in turn, phosphorylates membrane transporters. Phosphorylation influences transporter activity to kick-start many downstream functions, such as balancing the cytosolic N ­ a+-to-K+ ratio. 2+ The CBL-CIPK interaction is pivotal for ­Ca -dependent salinity stress signaling. Methods Computational methods are used to model the entire Arabidopsis thaliana CIPK24 protein structure in its autoinhibited and open-activated states. Arabidopsis thaliana CIPK24-CBL4 complex is predicted based on the protein–protein docking methods. The available structural and functional data support the CIPK24 and the CIPK24-CBL4 complex models. Models are energy-minimized and subjected to molecular dynamics (MD) simulations. MD simulations for 500 ns and 300 ns enabled us to predict the importance of conserved residues of the proteins. Finally, the work is extended to predict the CIPK24-CBL4 complex with the upstream kinase GRIK2. MD simulation for 300 ns on the ternary complex structure enabled us to identify the critical CIPK24-GRIK2 interactions. Together, these data could be used to engineer the CBL-CIPK interaction network for developing salt tolerance in crops. Keywords Calcineurin B-like protein · CBL-interacting protein kinase · Protein kinase · Protein–protein interactions · Plant salt tolerance Introduction Calcium-dependent signaling in plants is responsible for several major cellular events, including the activation of salinity-responsive pathways [1, 2]. Calcium and Data citation The older version of the manuscript, containing part of the current version, is currently in the Research Square preprint server (https://​www.​resea​rchsq​uare.​com/​artic​le/​rs-​ 37601​68/​v1). * Prabir Kumar Das Prabirk@migal.org.il * Debajyoti Dutta debajyoti.47@gmail.com; debajyoti.dutta@thapar.edu 1 MIGAL Galilee Research Institute, Kiryat Shmona, Israel 2 Department of Biotechnology, Thapar Institute of Engineering and Technology, Patiala, India calcineurin B-like protein (CBL) complex binds to CBLinteracting protein kinases (CIPKs) to propagate signals in plants. CBL-CIPK complex targets membrane ion transporters that belong to the ion homeostasis membrane protein group [3, 4]. Although the targeted membrane transporters control the ­Na+-to-K+ ratio in the cytoplasm, some recent data suggests that the CBL-CIPK complex also regulates C ­ a2+ transporting membrane proteins [3, 5]. Interaction between CBL and CIPK is the pivotal phenomenon in ­C a2+-dependent salinity stress signaling in plants. Different yet specific combinations of CBL-CIPK are possible that target a particular group of membrane ion transporters. For example, Arabidopsis thaliana CBL4 (SOS3) interacts with CIPK24 (SOS2) to regulate plasma membrane ­Na +/H + transporter SOS1 [6]. In the same plant, CBL8 also interacts with CIPK24 to activate SOS1, but the interaction is apparent during higher salt stress [7]. CBL10, which was first reported to combine with CIPK24 Vol.:(0123456789) 248 Page 2 of 13 for targeting a yet unknown vacuolar transporter, can also complex with CIPK8 to activate the plasma membrane SOS1 transporter [8, 9]. Homologous interactions between CIPK24 and its CBL partners are variable among plants. The yeast two-hybrid method shows that physical interactions between CIPK24 and CBL4 are possible in cassava and rapeseed plants, whereas the same interaction is not valid in eggplants [10–12]. CIPK24 and CBL10 interactions are also not universal in plants [11]. CIPK consists of two distinct domains joined by a flexible loop (Fig. 1a). The N-terminal domain is a kinase domain (KD), and the C-terminal domain is a protein–protein interaction or regulatory domain (PPI/RD). Between the domains, a flexible loop is present that binds to the CBLCa2+ complex. It has been predicted that the PPI domain interacts with KD when CIPK is not phosphorylated, resulting in inactivation [13]. Upon binding to the CBL-Ca2+ complex, the CIPK obtains an open conformation, allowing the complex to bind to various regulatory molecules. For example, the CBL4-CIPK24 complex interacts with nucleoside diphosphate kinase 2 and catalase for redox homeostasis in plants [14]. Others have proposed a glutamine-based modulation of the CBL4-CIPK24 complex for the regulation of SOS1 [15]. But, most importantly, CBL4-CIPK24 needs activation by upstream kinase and deactivation by the protein phosphatases. CIPK is phosphorylated by the upstream kinases like Geminivirus Rep-Interacting Kinases (GRIK1/2) primarily at its phosphate-binding loop [16]. Phosphorylated CIPK is deactivated by the abscisic acidinsensitive 1/2 (ABI 1/2), which is a protein phosphatase 2C group of phosphatases [17]. CBLs are predominantly membrane localized by N-terminal myristoylation or by possessing N-terminal transmembrane helices [8, 18]. A CBL-CIPK complex is directed to the membrane for phosphorylating the target membrane transporters. Some reports suggest that Fig. 1  Crystal structures of AtCIPK24 domains: a AtCIPK24 is composed of three distinct domains: kinase domain (KD), NAF/FISL domain, and the protein–protein interaction domain (PPI/RD). Values represent the amino acid ranges; b crystal structure of AtCIPK24 kinase domain (KD) (PDB ID 4d28); c crystal structure of the AtCIPK24 NAF/FISL domain and the protein–protein interaction (PPI) domain from the AtCIPK24 C-terminal domain bound to the AtCBL4 crystal structure (PDB ID 2ehb). The N- and C-terminals of each domain and the A ­ sp168 of the active-site loop are shown Journal of Molecular Modeling (2024) 30:248 CIPK can phosphorylate CBL at its C-terminal region to further enhance its activity [19, 20]. CIPK belongs to a conserved protein kinase group. All plant genomes known to date contain more than one CIPK coded in their genomes [5]. In Arabidopsis thaliana (At), there are 26 CIPK proteins encoded in its genome [21]. These proteins share 39–77% sequence identities; hence, the overall protein structure of CIPK must be the same (Suppl. Figure 1.). Only a few structures of individual domains of plant CIPK proteins are determined at atomic resolution (Fig. 1b, c). These include the crystal structures of AtCIPK14-PPI domain complex with AtCBL2 [22]. AtCIPK24-PPI domain complex with AtCBL4 [23] and the KD structures of AtCIPK24 and AtCIPK23 [24]. Due to the lack of detailed CIPK structural information, there are several unanswered questions. First, what does the autoinhibited CIPK24 structure look like? Second, what is the most probable conformation of the CBL4-CIPK24 complex? Third, how does the CBL4-CIPK24 complex enable CIPK24 interaction with the upstream kinases? In this work, the attempt is made to computationally build CIPK models in their autoinhibited and uninhibited forms, followed by computational prediction of the CBL-CIPK complex and the CBL-CIPK24-GRIK2 complex. The predicted structures are supported by the known data. Together, these data could be used to develop salt tolerance in plants. Methodology Bioinformatics Arabidopsis thaliana CIPK24 (Q9LDI3) and CBL4 (O81223) sequences were retrieved from the UniProt database. The templates used in the study are summarized in Journal of Molecular Modeling Page 3 of 13 (2024) 30:248 Suppl. Table 1. Multiple sequence alignment was done using MAFFT [25] and presented using ESPript [26]. As required, structural editing of PDBs was done using PyMOL [27], COOT [28], or Chimera [29]. All data plotting figures were generated using MATLAB (https://​in.​mathw​orks.​com). Computational modeling The templates for homology modeling were identified from the I-TASSER server [30]. Required atomic coordinates were obtained from Protein Data Bank (https://​www.​rcsb.​ org/), manually edited, and used for homology modeling in MODELLER software [31]. For making the CBL-CIPK complex, the CIPK open conformation was first modeled using MODELLER software. The best-modeled structure was selected using the DOPE score. CBL4 docking onto CIPK24 open conformation was done using multiple protein–protein docking servers, including ClusPro [32], PatchDock [33], Haddock [34], and ZDock [35]. Multiple docking servers were used to find a consensus docking position. For making the CIPK-GRIK2 complex, the GRIK2 structure was obtained from the AlphaFold model available in the UniProt Server (https://​www.​unipr​ot.​org/). Protein–protein docking was done as mentioned above. The available structural and biochemical data supported the docking positions in complex structures. The modeled structures were refined and energy-minimized using the Rosetta relax protocol and validated using Procheck [36]. The validated structure was used for MD analysis. Rosetta structure refinement Rosetta’s relax protocol was employed for initial structure refinement [37–40]. The protocol optimized protein structures through a systematic multi-step process. The protocol was integrated within the Rosetta force field framework. It executed multiple iterations of amino acid side-chain reorganization and energy minimization procedures to explore and identify the most energetically favorable conformation for a given protein structure, effectively navigating local energy barriers. The relaxation process first introduced small perturbations to the protein structure via adjustments in backbone torsion angles followed by side-chain repacking and energy minimization calculations and optimizing the structure’s torsional angles (Φ, Ψ, and χj). Notably, each perturbation step operated under a predefined weight for the van der Waals repulsive component of the all-atom scoring function (fa_rep) [41]. The orchestrated process enabled the structure to traverse the local energy landscape, facilitating transitions toward new energy minima. Concurrently, clashes arising during each cycle were systematically resolved. The resultant structure with the lowest energy was selected as the input structure for subsequent molecular dynamics (MD) 248 analyses, ensuring a refined starting point for further computational investigations. Molecular dynamics simulations The molecular dynamics (MD) simulation was performed using AMBER20 software with the ff19SB force field [42–45]. The system was solvated with TIP3P water molecules in a cubic box and was neutralized by adding counter ions using the “tleap” module of AMBER. Simulation involving calcium ions was done using the frcmod.ions 234lm_1264_tip3p parameter file. Initially, the system’s energy was minimized using a dual-step algorithm, employing the steepest descent method followed by the conjugate gradient method. The energy minimization stabilized the system and eliminated unfavorable contacts that could lead to unrealistically high energies or aberrant trajectories. Identifying and removing these detrimental contacts were imperative to ensure the reliability and accuracy of subsequent simulations. The key parameters governing this optimization involved the system undergoing minimization with 50,000 steps of steepest descent followed by 50,000 steps of conjugate gradient methods. During minimization, positional restraints (“ntr = 1”) with a restraint weight of 100.0 were utilized. The heating process was carried out under constant volume (ntb = 1) conditions without any pressure control (ntp = 0). The heating simulation was run for 50,000,000 steps (nstlim) using a time step of 0.002 ps (dt). The SHAKE algorithm was employed to constrain bonds involving hydrogen atoms. A Langevin thermostat with a collision frequency of 1.0 ­ps−1 (gamma_ln) was applied. Additionally, positional restraints were applied to prevent large conformational changes during the heating process. Additional positional restraints with a weight of 100 (restraint_wt) were applied to atoms. During the first 40 million steps of the heating process, the temperature was gradually increased from 100 to 300 K, and for the remaining 10 million steps, the temperature was held constant at 300 K. During equilibration, the simulation utilized periodic boundary conditions (iwrap = 1) to confine positions within the primary unit cell. Under these conditions, constant pressure was applied (ntb = 2) alongside pressure control (ntp = 1). A total of 500,000 steps were executed (nstlim) with a time step size of 0.002 ps. The treatment of bond lengths and angles employed the SHAKE algorithm (ntc = 2) for hydrogen atoms and utilized the Amber force field for interactions (ntf = 2). For temperature control, Langevin dynamics were employed (ntt = 3) with a collision frequency of 1.0 ­ps−1 (gamma_ln). The relaxation time for pressure coupling was set to 0.1 ps (taup). Positional restraints of 100 were applied during equilibration to maintain specific atom positions within the system. The simulation was maintained at a target temperature of 300 K. The trajectory was analyzed 248 Page 4 of 13 to ensure that the system had reached a stable equilibrium state. During the production phase of the simulation, various parameters were applied to ensure accurate modeling of the system. Constant pressure was maintained (ntb = 2) along with isotropic position scaling (ntp = 1), facilitating a stable environment for the simulation. A considerable duration of 50,000,000 steps was executed with a time step size of 0.002 ps. The treatment of bond lengths and angles utilized the SHAKE algorithm (ntc = 2) for hydrogen atoms and applied the Amber force field for interactions (ntf = 2). Temperature control employed Langevin dynamics with a collision frequency of 1.0 ­ps−1 and a relaxation time of 0.1 ps. The simulation commenced at an initial temperature of 300.0 K. These settings ensured a prolonged and stable simulation for the production phase, allowing detailed observations and analyses of the system’s molecular dynamics behavior. For all calculations, a cutoff distance value of 10 was used for non-bonded interactions, beyond which no interactions were calculated. Finally, the trajectory was analyzed using CPPTRAJ and Chimera software to extract relevant data, such as RMSD, RMSF, and specific analysis. All variables were explained in the Amber Reference Manual (https://​amber​md.​org/​Manua​ls.​php). Result Structure of autoinhibited CIPK The autoinhibited conformation of CIPK24 is modeled based on the human C ­ a2+/calmodulin-dependent protein kinase, MARK1 (MAP/microtubule affinity-regulating kinase) (PDB ID 6c9d), which shows a tight association between kinase domain (KD) and regulatory domain (KA1) [46]. The structural similarity between CIPK and MARK has been observed before [23]. CIPK-KD (1–300) and the MARK-KD Fig. 2  Computationally predicted structure of autoinhibited AtCIPK24. A Cartoon representation of the computationally predicted AtCIPK24 structure. The different domains are shown in colors. The active-site loop (magenta) and the interdomain interacting αD helix (cyan) are shown. B RMSD vs. time plot of the 500 ns molecular dynamics simulation is shown highlighting the drifts at 128 ns and 243 ns. Inset is the RMSF plot of the residues during the simulation run Journal of Molecular Modeling (2024) 30:248 (53–314) show 45% sequence identity, whereas CIPK24-PPI (341–426) and MARK-KA1 (703–751) share 29%. Superposition of CIPK24-KD (1–300) and CIPK24-PPI (341–426) crystal structures on corresponding domains of MARK1 structure (PDB ID 6c9d) shows 1.14 Å and 1.28 Å RMSD, respectively, indicating significant structural homologies between two proteins. Building the KD and PPI domainconnecting loop (301–339) is done based on the orientations of the C-terminal of KD and the N-terminal of PPI domains. The secondary structures of the domain-connecting loop are built based on the known crystal structure information and the predicted secondary structures using PSIPRED [47]. The edited model is used as a template in MODELLER, and the best-predicted structures are chosen for subsequent MD simulation. The final CIPK24 structure after simulation is shown in Fig. 2A. Procheck validation shows that the structure has 99.7% residues in the most favored and allowed regions. The overall structures of the two domains are essentially the same as in the crystal structure, with some local rearrangements of the loops. The PPI domain sits at the interface of the N- and C-lobes of the KD. The NAF/ FISL loop runs parallel and in opposite directions, where it becomes close to the N-terminal end of the protein. A total 500 ns MD simulation run shows that the structure gradually reaches stability (Fig. 2B). However, at 128 ns and 243 ns, drifts are observed in the RMSD vs. time graph. RMSF calculations of the autoinhibited CIPK24 reveal that many PPI domain regions undergo more fluctuations than the kinase domain (Fig. 2B inset). Analysis of the intermediate structures reveals significant structural variation around the active-site loop residue ­Val163, the NAF/FISL loop, and the PPI domain residues ­Asn369 and ­Ala382. Comparison between the initial and final CIPK structure (after simulation) reveals reorientation of the active-site loop spanning the residues 160–172. Measurement shows that after simulation, the loop shifts closer to the N-terminal 20GTFA23 motif and the PPI domain (Fig. 3A). The GTFA motif is Journal of Molecular Modeling (2024) 30:248 Page 5 of 13 248 Fig. 3  Conformational rearrangements of the AtCIPK24 autoinhibited structure: A 500 ns simulation reveals the steady shift of the active-site loop (magenta) that shifted toward the 20GTFA23 sequence and the PPI domain. The active-site residues L ­ eu165 and ­Thr169 are used to measure the distances to T ­ hr21 of the 20 GTFA23 motif and ­Gly408 of the PPI domain respectively. B The 500 ns simulation also shows the only NAF/FISL loop helix 324LSALF328 retains its helical structure conserved in CIPK proteins and can contribute to reorienting the active-site loop (160–172). During simulation, another major readjustment occurs to the domain-connecting NAF/ FISL loop (301–339). RMSF calculation demonstrates that readjustment of the loop occurs during the first 100 ns cycle and shows lesser structural fluctuations during the rest of the simulation cycles. Two regions of the loop 313FEMITL318 and 324LSALF328 are initially modeled as helices based on the predicted secondary structure and the x-ray crystallography structure available (PDB ID 2ehb). After 500 ns simulations, the helical structure of only the 324LSALF328 segment is retained. The result indicates that during the autoinhibited state, 313FEMITL318 exists as a loop structure that attains helical conformation upon CBL binding (Fig. 3B). The simulation data revealed that the interaction between KD and the PPI domain is weak yet stable. The average centroid distance between the domains is maintained at 25 Å. Structural comparison before and after simulation manifests subtle reorientations of the N-lobe and the PPI domains. Interdomain interactions are mediated by hydrogen bonding, electrostatic, and hydrophobic interactions. αD helix of the kinase domain is primarily involved in interdomain ­ is102 belong to the interactions. ­Glu95, 98Asp-Arg99, and H αD helix and interact with the C-terminal residues of the domain-connecting NAF/FISL loop (­Arg337 and G ­ ln338) 337 338 (Fig. 4A). Although A ­ rg and G ­ ln fall within the range of the NAF/FISL loop, this region is also proposed to be a part of regulatory domain [14]. Transient interactions are also observed between ­Lys405 and H ­ is414 with the αD helix residues. The corresponding αD helix of MARK1 also participates in intra-domain interaction [46]. The N-terminal of αD helix in AtCIPK24 comprises the most conserved residues 93GGELFDRIVH102 in AtCIPKs and, therefore, can have a role in interdomain interactions (Suppl. Figure 2.). The 500 ns simulation further shows a steady shift of the active-site loop toward the PPI domain. The active-site loop movement toward the PPI domain is monitored using the distances between ­Cys170 and ­Glu409. ­Cys170 is in the activesite loop, closest to the PPI domain, and G ­ lu409 in the PPI domain shows relatively less atomic fluctuation (Fig. 2B, ­ lu409 distance inset). The result shows that the C ­ ys170 to G 170 is decreased (Fig. 4B, C). ­Cys is a conserved residue, whereas glutamates or aspartates are found in the G ­ lu409 position in AtCIPKs (Suppl. Figure 2.). Interestingly, ­Glu409 is located in the β5-α2 loop in the PPI domain, yet it undergoes less flexibility. We propose that G ­ lu409 has an activesite loop-engaging role in the CIPK autoinhibited state. The stable interactions are summarized in Suppl. Table 2. The MARK1/KA1 domain mediates phospholipid interactions [48]. The cluster of charged residues in MARK1 is suggested to be G ­ lu756, ­Arg771, ­Lys773, and A ­ rg774. The MARK1/KA1 and CIPK4-PPI superimposition identified 248 Page 6 of 13 Journal of Molecular Modeling (2024) 30:248 Fig. 4  Interdomain interactions involved in AtCIPK24 autoinhibited structure: αD helix residues interact with the C-terminal residues of the NAF/FISL loop conferring the interdomain interactions. B The steady shift of the active-site loop is observed by analyzing the 0 ns (top) and 500 ns (bottom) structures. Distances between the ­Glu409 and active-site loop residue ­Cys170 are shown. C Distances measured between ­Cys170 and ­Glu409 and the same between ­Thr169 and ­Thr21 during the simulation time scale endorse the active-site loop moving toward the PPI domain and the N-terminal 20GTFA23 motif the corresponding amino acids G ­ lu390, ­Asp402, ­Arg404, and 405 ­Lys (Suppl. Figure 2.). Although the interactions between the CIPK24/PPI domain and the membrane are not known, it is possible that CBL-mediated CIPK targeting the membrane opens up the PPI domain and interacts with the membrane. the CBL4-CIPK24 complex is predicted by protein–protein docking servers. Results from different protein–protein docking servers show a consensus CBL docking mode. In the final CBL4-CIPK24 complex, the PPI domain is located on the opposite side of the KD active site, keeping the phosphorylation loop accessible to other proteins (Fig. 5). The CBL4-CIPK24 complex model is energy-minimized and validated using Procheck. Procheck validation shows that the structure has 99% residues in the most favored and allowed regions. A 300 ns MD simulation run is performed, showing that the complex reaches stability at the end of the run (Fig. 6A). The CBL-CIPK complex makes a transient complex, which becomes weakened during the following simulation runs. RMSF calculations show that two particular regions of the CIPK encompassing the residues ­Asp283-Ile291 and ­Asp333 undergo significant atomic fluctuations (Fig. 6B, C). ­Asp283-Ile291 harbors a short helical segment that flanks out and opens the NAF/FISL loop for CBL binding. ­Asp333 is located in the C-terminal of the CBL4-binding segment of CIPK24. Interactions between the CBL4-CIPK24 are predominantly driven by hydrophobic interactions (Fig. 7). For example, the ­Phe313 remains within hydrophobic interacting distance with CBL4/Ala160, CBL4/Leu133, and CBL4/ Met 156. ­P he 313 is part of the conserved 311NAFEMI 316 motif of the NAF/FISL loop. P ­ he 328 hydrophobic Structure of CIPK‑CBL binary complex The CIPK24-CBL4 complex is modeled in two steps, first by modeling the CIPK24 open conformation and then by predicting the CBL4 docking onto the CIPK24 open conformation. CIPK24 open conformation is modeled based on the human adenosine monophosphate-activated protein kinase α subunit (AMPKα; PDB IDs 7myj and 4rer) [49–51]. Sequence alignment between the CIPK24-KD and the AMPKα-KD shows 45% sequence identity, and the same between CIPK24-PPI and AMPKα-CTD is 20%. Superposition of CIPK24-KD (1–300) and CIPK24-PPI (340–426) crystal structures on corresponding domains of AMPKα structure (PDB ID 7myj) shows 1.31 Å and 1.33 Å RMSD, respectively, indicating significant structural homology. In the CIPK24 open conformation model, the relative positions of the NAF/FISL loop and the PPI domain are somewhat similar to that of the CBL-complex structure (PDB ID 2ehb) (Suppl. Figure 3.). The structure is validated using Procheck. After obtaining the CIPK24 open conformation, Journal of Molecular Modeling (2024) 30:248 Page 7 of 13 248 Fig. 5  Computationally predicted structure of AtCBL4AtCIPK24 complex: CBL represents AtCBL4 that is bound to the NAF/FISL loop of the AtCIPK24. The distinct domains of AtCIPK24 are represented using different colors Fig. 6  RMSD vs. time and the RMSF vs. residue plots of the AtCBL4-AtCIPK24 binary complex: A RMSD vs. time plot of the binary CBL4-CIPK24 complex showing the stability of the structure is reached after 300 ns. The major RMS deviation occurs from 31 to 85 ns. B and C Residue-wise RMSF values of the individual protein structures in the complex showing the most flexible regions interaction with ­Tyr195 and L ­ eu193 does not change more than 0.5 Å during the 300 ns simulations. ­Phe328 is also a conserved residue in the CIPK24 NAF/FISL loop. The conserved ­Leu324 and CBL4/Phe83 hydrophobic interaction also do not change more than 0.2 Å during the simulation timeframe. A few hydrogen bond interactions are observed and remain consistent during the simulation, including the interaction between the conserved ­Glu314 with CBL4/ Ser186 and S ­ er319 with CBL4/Glu143. In the CBL4-CIPK24 complex structure, the CIPK24 residues P ­ he313 and S ­ er319 are found to be interacting with CBL4. Other CBL4CIPK24 interactions, such as ­Arg331 with CBL4/Asp198, ­Lys371 with CBL4/Glu145, ­Glu393 with CBL4/Arg71, and ­Arg404 with CBL4/Glu145, are mostly confined within the PPI domain of CIPK24. The interactions are summarized in Suppl. Table 2. CBL4 shows lesser atomic fluctuations owing to its compact structure. However, the C-terminal region around the residue L ­ eu148 shows the highest fluctuation (Fig. 6C). Structure of CBL‑CIPK‑GRIK ternary complex Computational prediction of GRIK2 binding to CIPK24 is done on the CBL4-CIPK24 complex. The active site of the CBL4-CIPK24 complex is open and can be accessed by the upstream kinases. Arabidopsis GRIK2 structure was predicted using I-TASSER and compared with the AlphaFold structure and used for docking. The result shows that GRIK2 docks on the CIPK24-KD head-to-tail fashion, comparable to the CIPK24-KD dimer in the crystal structure. GRIK2 shows a typical Ser/Thr protein kinase fold like CIPK. Hence, it is assumed that the GRIK2-CIPK24 heterodimer mimics the CIPK-CIPK homodimer (Fig. 8). A total 300 ns simulation shows that the structure is stable throughout the simulation (Fig. 9A). RMS fluctuation (Fig. 9B–D) shows that the significant structural rearrangement of CIPK24 is limited to the vicinity of the CBL4-binding region. On the other hand, only a few atomic fluctuations are observed in its kinase domain. GRIK2, a 248 Page 8 of 13 Journal of Molecular Modeling (2024) 30:248 other homologous protein kinases suggests that the charged or polar residue occupies the L ­ ys120 position. These interactions, therefore, have a role in stabilizing the active-site loop. Discussions Fig. 7  The interface of AtCBL4-AtCIPK24 complex: The intermolecular interactions of CBL4 and CIPK24 are driven by multiple hydrophobic interactions and a few hydrogen bond interactions. Residues belonging to CBL are indicated as the subscript “CBL” along with the residue names. CIPK24 residues are colored green, whereas CBL24 residues are shown in cyan protein kinase, structure was compared with the closest sequence homolog murine NF-κβ-inducing kinase (NIK) structure (PDB ID 4g3f) to identify the important amino acids [52]. The GRIK2 consists of N- and C-lobes. C-lobe comprises α helices and undergoes relatively higher structural fluctuations than the N-lobe. Functional amino acid segments in GRIK2 are identified using multiple sequence alignment and are clustered at the interface of its N- and C- lobes (Suppl. Figure 4.). GRIK2 functional segments, such as the phosphate-binding P-loop 113IGSGSYGKV121 and the catalytic residues 236HGDIK240, are located at the N- and C-lobe interface. CIPK24 active-site loop residues ­Thr168 and ­Thr169 are close to the GRIK2 P-loop ­Gly116 and ­Ser117 residues (Fig. 10A). Simulation data reveals that the two regions have close interactions and shows the reorientation of the CIPK24 active-site loop and the ordering of the GRIK2 P-loop (Fig. 10B, C). The reorientation and the ordering of the loop are not artifacts, as these regions are involved in phosphate exchange. A few other interactions surrounding the active-site loop may be involved in holding the active-site loop in position. For example, ­Glu95 and ­Glu138 of CIPK24 are observed interacting with the P-loop residue ­Lys120. ­Glu95 and ­Glu138 positions in CIPKs are occupied with the negatively charged residue (Suppl. Figure 2.). In GRIK2, L ­ ys120 is located at the C-terminal of the P-loop. Multiple sequence alignment of GRIK2 with Transient interactions between CBL and CIPK proteins regulate a multitude of protein activation pathways in plants. CIPK remains in an autoinhibited state, but structural alteration occurs to make a complex with a specific CBL protein partner. The transient CBL-CIPK complex is phosphorylated by upstream kinases, and the phosphorylated CIPK can regulate membrane transporters. Current understandings of the CIPK autoinhibited and complex structures are limited. Because, the partial CBL and CIPK structures are known and the structures of the CIPK homologous are known, an attempt is made to predict the autoinhibited and active CIPK structures. The autoinhibited CIPK structure is predicted based on the human ­Ca2+/calmodulin-dependent protein kinase MARK1 due to its significant structural homology with CIPK. While the relative orientations of the KD and PPI domains have not been confirmed experimentally, structural analysis and simulation studies support the predicted structure. The deduced relative orientations of the KD and PPI domains from the autoinhibited MARK1 template are further supported by an independent computational docking experiment using separate KD and PPI domains. The best PPI docking clusters on the KD domain closely mimic the relative orientations of the KA1 and KD domains in MARK1. In MARK1, substitutions of residues G ­ lu143, ­Asp146, ­Val149, and ­His151 reveal that these residues are critical for KA1 trans-inhibiting the KD domain [46]. Corresponding residues in the CIPK-KD ­Glu95, ­Asp98, ­Val101, and ­His102 are located at the interdomain interface and are found in close contact with the PPI in the docking conformation. These KD domain residues are critical as they interact closely with the PPI. The CIPK autoinhibited structure remains stable and converges during simulations. Plotting the radius of the gyration value during the simulation reveals that the structure remains compact. Analysis shows that the interaction between the KD and the PPI domain is weak, but the PPI domain does not move away from the KD structure. Interdomain interaction is maintained due to the interactions between the αD helix and the C-terminal residues from the NAF/FISL loop. Simulation data suggest a steady shift of the active-site loop toward the PPI domain, and the 20GTFA23 manifests the autoinhibition mechanism. Membrane transporters and protein-lipid interaction in the membrane are suggested to be one of the pivotal phenomena during salinity stress in plants [53, 54]. The activation of ion-transporting membrane proteins by CIPKs must occur in close proximity Journal of Molecular Modeling (2024) 30:248 Page 9 of 13 248 Fig. 8  The CBL4-CIPK24GRIK2 ternary complex: GRIK2 (cyan) and CBL4 (magenta) are shown in color. Different domains of CIPK24 are colored and indicated. Arrows show CIPK24 activesite residues and the P-loop of GRIK2 Fig. 9  Three hundred nanoseconds of simulation data of the ternary complex: A RMSD vs. time plot shows the stable ternary complex. B–D RMSF data of the individual protein entities are shown indicating the most flexible residues to the membrane. Recent findings reveal that CIPKs interact with membrane phospholipids through a lysine residue present in the KD domain [55]. Additionally, CBL proteins facilitate CIPK membrane localization. Most CBLs bind to the membrane via N-myristoylation or acylation, while some have an integral membrane helix. Interestingly, the MARK1 template, used for modeling CIPK24, has an intrinsic affinity for phospholipids through its KA1 domain. Structural comparisons show positively charged residues in the CIPK24-PPI domain. However, in AtCIPKs, the positions of these positively charged residues are sparsely conserved. The PPI domain in CIPK corresponds to the KA1 domain of MARK1. Further studies are needed to determine if CIPKs interact with the membrane through the PPI domain. The CIPK24 open conformation is built based on the activated conformation of the AMP-activated protein kinase (AMPK). AMPK autoinhibitory domain (AID) and the regulatory-subunit-interacting motif (RIM) control the phosphorylation-dependent activation of the KD [56]. During inactive conformation, AID remains in folded conformation, whereas upon activation, it obtains an open conformation freeing the KD active site for phosphorylation [57]. The structural modeling assumes that in CIPK24 open conformation, the NAF/FISL loop brings the PPI domain to the opposite face of the CIPK that of the autoinhibited face. This conformational change is facilitated by the CBL4 binding to the NAF/FISL loop. MD simulation endorses the stability of the CBL-CIPK complex. It is interesting to observe that the NAF/FISL loop exhibits minimal secondary structure in its autoinhibited state, but its helicity increases upon forming a complex with CBL. This suggests that the loop becomes more ordered when associated with CBL. Our energy calculations indicate that the CBL-CIPK complex has a negative energy value during the simulation, supporting the formation of a stable complex. Interaction between CBL and CIPK is mostly due to the hydrophobic interactions, and a few conserved hydrogen bonding interactions are observed, such as ­ er319-CBL4/Glu143. CBL4/Glu148 ­Glu314-CBL4/Ser186 and S 248 Page 10 of 13 Journal of Molecular Modeling (2024) 30:248 Fig. 10  GRIK2 and CIPK24 interactions: A GRIK2 and CIPK24 are involved in N-to-C-lobe interactions. Primarily, the CIPK24 active-site loop residues are in close proximity to the GRIK2 P-loop (113IGSGSYGKV121) residues. Residues from GRIK2 are named with the “GRIK2” subscript. B MD simulation confers that the CIPK24 active-site loop has interaction with the GRIK2 P-loop as demonstrated by comparing the 0 ns and 300 ns structures. CIPK24 activesite loop reorients itself, and the GRIK2 P-loop becomes ordered. C and D CIPK24 active-site ­Thr169 becomes closer to the P-loop is located in a loop, showing the highest atomic fluctuations (Fig. 6C). The region is a juxtaposition of the PPI domain, and another negatively charged residue in that loop ­Glu145 is involved in interactions with two positively charged residues of the PPI domain: ­Lys371 and ­Arg404. It is not known if PPI has any role in CBL4 binding, and its interaction with the PPI domain is not evident from the crystal structures. PPI is involved in protein–protein interaction; hence, CBL-CIPK interaction might occur through the PPI domain. Crystal structures of CBL2 and CBL4 complexes with their cognate CIPK partners reveal the role of backbone interactions between ­Gln320-CBL4/Glu142 [23]. A close backbone interaction between G ­ ln320 and CBL4/Glu143 is also observed in some intermediate simulation structures endorsing the predicted complex structure. Another observation is the CBL4-CIPK24 complex exposes the 340RF341 motif allowing it to interact with the other regulatory proteins such as ABA-INSENSITIVE2 Ser/Thr protein phosphatases [17]. In the case of the autoinhibitory state, the CIPK24 αD helix maintains close contact with the region ­Gln338, hence keeping 340RF341 not accessible to other molecules. AMPKs in mammals are phosphorylated by the upstream kinases LKB1 and CaMKKβ [58–60]. BLAST results have identified that mammalian LKB1 and CaMKKβ share the highest homology with the Arabidopsis thaliana GRIK1 and GRIK2. It has been shown that both GRIK1 and GRIK2 kinases of Arabidopsis thaliana can phosphorylate sucrose non-fermenting 1 (SNF1)-related protein kinases such as SnRK1 [16, 61]. It is also evidenced that the SnRK kinase domain is sufficient to bind to GRIK proteins. SOS2 is a class 3 SnRK which was shown to be phosphorylated by GRIK2 [16]. For this study, GRIK2 is used because the GRIK2 AlphaFold model (Q5HZ38) shows the least RMSD value when compared to the I-TASSER GRIK2 model. The CBL4-CIPK24-GRIK2 ternary complex essentially uses the model of the CBL4CIPK24 complex. GRIK2 and CIPK24 both are protein kinases and therefore share a similar fold. GRIK2 docking on CIPK shows a head-to-tail association similar to the CIPK dimer association in the crystal structure. Simulation data shows that the C-lobe of CIPK has a stronger interaction with the N-lobe of GRIK2 that shows lesser atomic fluctuations. Analyses of the GRIK2 protein sequences have identified critical motifs of the protein including the phosphate-binding P-loop and the catalytic bases (Suppl. Fig. 4). In GRIK2, these motifs are 113IGSGSYGKV121 and 236 HGDIK240. The 113IGSGSYGKV121 motif is regarded as a P-loop, which is a phosphate-binding loop located in close proximity to the CIPK active-site loop. Analyzing the ternary structures before and after the simulations revealed the CIPK active-site loop reorientation and the ordering of the GRIK2 P-loop. In another part of the active site, the catalytic residues 236HGDIK240 also substantially moved toward the CIPK24 active-site loop (Fig. 8). Therefore, it is possible that the coordinated movements of the loops are required for the phosphate transfer. Interactions between the G ­ lu95 and Journal of Molecular Modeling Page 11 of 13 (2024) 30:248 248 an active state upon binding with CBL and subsequently interacts with other proteins like GRIK2 to modulate its activity and function within the signaling pathway. Conclusions and future direction Fig. 11  The proposed mechanism of CBL-CIPK-GRIK interaction can be described as follows: During the autoinhibition state, the CIPK-PPI domain (blue) engages with the CIPK-KD (red) active-site helix (lime green), resulting in the inhibition of CIPK. When the CBL (magenta) binds to the NAF/FISL motif (green), it induces a conformational change that relocates the PPI domain (blue) to the opposite face of the KD (red). This conformational shift causes an approximate 140° rotation of the PPI domain relative to the KD domain. As a result, the active-site helix (lime green) is exposed. This open conformation of the active site can then interact with other protein regulators, such as GRIK (cyan), facilitating its phosphorylation and subsequent downstream signaling events ­Glu138 with L ­ ys120 are important for holding the active-site loop in position. Based on the structural data, a series of protein–protein interaction events involving CIPK activation can be hypothesized (Fig. 11). During autoinhibition, the PPI (protein–protein interaction) domain of CIPK binds close to the kinase domain (KD) active site. The active-site loop has an inherent affinity toward the PPI domain and moves closer to the KD active-site loop during the simulation time frame. The interaction between the PPI and KD domains is weak, suggesting that CIPK may exhibit a basal level of activity even in its autoinhibited state. When the NAF/FISL loop binds to the CBL (calcineurin B-like) protein, the interaction between the PPI and KD domains is broken. CBL binding to the NAF/FISL loop induces a conformational change, moving the PPI domain to the opposite side of the active site face of CIPK. The measurement indicates that the PPI domain rotates at approximately 140°. In this open conformation, CIPK can bind to various interacting proteins, including the upstream protein kinase GRIK2 (Geminivirus Rep-Interacting Kinase2). The interaction between CIPK and GRIK2 shows a headto-tail association, with the CIPK active-site loop in close contact with the GRIK2 phosphate-binding loop and the catalytic loop. This interaction likely facilitates the transfer of signals or substrates necessary for the downstream effects of CIPK activity. These hypothesized interactions suggest a dynamic regulation mechanism for CIPK, where it transitions from a basal activity state in autoinhibition to CIPK is the major Ser/Thr protein kinase contributing to the activations of many critical membrane transporters essential for controlling plant biotic and abiotic stresses. CIPK is a mediator connecting the C ­ a2+ sensors and their targets while acting as a hub of many protein–protein interactions. Besides interacting with the C ­ a2+ sensors, CIPK activation and deactivation are controlled by the upstream kinases and the phosphatases, respectively. Simulation data of computationally predicted CIPK24 autoinhibited structure reveals the interactions of the PPI domain with both the NAF/FISL loop and the active-site loop. The two interactions might have a role in CIPK24 autoinhibition. Simulation data of the modeled CBL4-CIPK24 complex shows a stable complex primarily driven by the hydrophobic interaction. The complex shows that the PPI domain may be involved in CBL4 interaction. Molecular dynamics simulation further endorses the CIPK24-CBL4-GRIK2 ternary complex, identifying critical amino acids in GRIK2 responsible for phosphate transfer. The results could be used to design CBL, CIPK, and GRIK constructs for further structural and functional studies of the individual proteins or their complexes. Supplementary Information The online version contains supplementary material available at https://d​ oi.o​ rg/1​ 0.1​ 007/s​ 00894-0​ 24-0​ 6037-5. Author contribution T.B., S.B., and S.M. performed the sequence alignment, initial model building, protein–protein docking, and interaction analysis. P.K.D. performed the simulation experiments. D.D. conceptualized the study, finalized the models, and analyzed the simulation data. P.K.D and D.D. wrote the manuscript. Funding The work is supported by the Science and Engineering Research Board (SERB) Government of India Start up Research Grant no. SRG/2022/000035 dt. 26/10/2022 to Debajyoti Dutta. Sanhita Banik is supported by an institutional fellowship from the Thapar Institute of Engineering and Technology. Data availability The coordinate files of the autoinhibited CIPK24, CIPK-CBL complex and the CIPK-CBL-GRIK2 complex can be available upon request. Declarations Competing interests The authors declare no competing interests. References 1. Luan S (2009) Trends Plant Sci 14:37–42 248 Page 12 of 13 2. Weinl S, Kudla J (2009) New Phytol 184:517–528 3. Sanyal SK, Mahiwal S, Nambiar DM, Pandey GK (2020) Biochemical Journal 477:853–871 4. Tang R-J, Wang C, Li K, Luan S (2020) Trends Plant Sci 25:604–617 5. Ma X, Li Q-H, Yu Y-N, Qiao Y-M, Haq Su, Gong Z-H (2020) Int J Mol Sci 21, 5668 . 6. Gong D, Guo Y, Schumaker KS, Zhu J-K (2004) Plant Physiol 134:919–926 7. Steinhorst L, He G, Moore LK, Schültke S, Schmitz-Thom I, Cao Y, Hashimoto K, Andrés Z, Piepenburg K, Ragel P (2022) Developmental Cell 57:2081-2094. e2087 8. Kim BG, Waadt R, Cheong YH, Pandey GK, DominguezSolis JR, Schültke S, Lee SC, Kudla J, Luan S (2007) Plant J 52:473–484 9. Yin X, Xia Y, Xie Q, Cao Y, Wang Z, Hao G, Song J, Zhou Y, Jiang X (2020) J Exp Bot 71:1801–1814 10. Li J, Jiang M-M, Ren L, Liu Y, Chen H-Y (2016) Mol Genet Genomics 291:1769–1781 11. Liu W-Z, Deng M, Li L, Yang B, Li H, Deng H, Jiang Y-Q (2015) Biochem Biophys Res Commun 467:467–471 12. Mo C, Wan S, Xia Y, Ren N, Zhou Y, Jiang X (2018) Front Plant Sci 9:269 13. Guo Y, Qiu Q-S, Quintero FJ, Pardo JM, Ohta M, Zhang C, Schumaker KS, Zhu J-K (2004) Plant Cell 16:435–449 14. Verslues PE, Batelli G, Grillo S, Agius F, Kim Y-S, Zhu J, Agarwal M, Katiyar-Agarwal S, Zhu J-K (2007) Mol Cell Biol 27:7771–7780 15. Miranda RdS, Alvarez-Pizarro JC, Costa JH, Paula SdO, Prisco JT, Gomes-Filho E (2017) Plant Signal Behav 12:522–536 16. Barajas-Lopez JdD, Moreno JR, Gamez-Arjona FM, Pardo JM, Punkkinen M, Zhu JK, Quintero FJ, Fujii H (2018) Plant. J. 93:107–118 17. Ohta M, Guo Y, Halfter U, Zhu J-K (2003) Proc Natl Acad Sci 100:11771–11776 18. Ishitani M, Liu J, Halfter U, Kim C-S, Shi W, Zhu J-K (2000) Plant Cell 12:1667–1677 19. Hashimoto K, Eckert C, Anschütz U, Scholz M, Held K, Waadt R, Reyer A, Hippler M, Becker D, Kudla J (2012) J Biol Chem 287:7956–7968 20. Li K-L, Tang R-J, Wang C, Luan S (2023) Nat Commun 14:360 21. Kolukisaoglu U, Weinl S, Blazevic D, Batistic O, Kudla JR (2004) Plant Physiology 134:43–58 22. Akaboshi M, Hashimoto H, Ishida H, Saijo S, Koizumi N, Sato M, Shimizu T (2008) J Mol Biol 377:246–257 23. Sánchez-Barrena MJ, Fujii H, Angulo I, Martínez-Ripoll M, Zhu J-K, Albert A (2007) Mol Cell 26:427–435 24. Chaves-Sanjuan A, Sanchez-Barrena MJ, Gonzalez-Rubio JM, Moreno M, Ragel P, Jimenez M, Pardo JM, MartinezRipoll M, Quintero FJ, Albert A (2014) Proc Natl Acad Sci 111:E4532–E4541 25. Katoh K, Standley DM (2013) Mol Biol Evol 30:772–780 26. Robert X, Gouet P (2014) Nucleic Acids Res 42:W320–W324 27. DeLano WL (2002) CCP4 Newsl. Protein Crystallogr 40:82–92 28. Emsley P, Cowtan K (2004) Acta Crystallogr D Biol Crystallogr 60:2126–2132 29. Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, Ferrin TE (2004) J Comput Chem 25:1605–1612 30. Yang J, Yan R, Roy A, Xu D, Poisson J, Zhang Y (2015) Nat Methods 12:7–8 31. Webb B, Sali A (2016) Current Protocols in Bioinformatics 54, 5.6. 1–5.6. 37. https://​doi.​org/​10.​1002/​cpbi.3 32. Kozakov D, Hall DR, Xia B, Porter KA, Padhorny D, Yueh C, Beglov D, Vajda S (2017) Nat Protoc 12:255–278 Journal of Molecular Modeling (2024) 30:248 33. Schneidman-Duhovny D, Inbar Y, Nussinov R, Wolfson HJ (2005) Nucleic Acids Res 33:W363–W367 34. De Vries SJ, Van Dijk M, Bonvin AM (2010) Nat Protoc 5:883–897 35. Pierce BG, Wiehe K, Hwang H, Kim B-H, Vreven T, Weng Z (2014) Bioinformatics 30:1771–1773 36. Laskowski RA, MacArthur MW, Moss DS, Thornton JM (1993) J Appl Crystallogr 26:283–291 37. Burman SSR, Yovanno RA, Gray JJ (2019) Structure 27:10411051.e1048 38. Khatib F, Cooper S, Tyka MD, Xu K, Makedon I, Popović Z, Baker D (2011) Proc Natl Acad Sci 108:18949–18953 39. Nivón LG, Moretti R, Baker D (2013) PLoS ONE 8:e59004 40. Tyka MD, Keedy DA, André I, DiMaio F, Song Y, Richardson DC, Richardson JS, Baker D (2011) J Mol Biol 405:607–618 41. Alford RF, Leaver-Fay A, Jeliazkov JR, O’Meara MJ, DiMaio FP, Park H, Shapovalov MV, Renfrew PD, Mulligan VK, Kappel K (2017) J Chem Theory Comput 13:3031–3048 42. Case DA, Aktulga HM, Belfon K, Cerutti DS, Cisneros GA, Cruzeiro VWD, Forouzesh N, Giese TJ, Götz AW, Gohlke H (2023) J Chem Inf Model 63:6183–6191 43. Case DA, Cheatham TE III, Darden T, Gohlke H, Luo R, Merz KM Jr, Onufriev A, Simmerling C, Wang B, Woods RJ (2005) J Comput Chem 26:1668–1688 44. Pabis A, Geronimo I, York DM, Paneth P (2014) J Chem Theory Comput 10:2246–2254 45. Salomon-Ferrer R, Case DA, Walker RC (2013) Wiley Interdiscip Rev: Comput Mol Sci 3:198–210 46. Emptage RP, Lemmon MA, Ferguson KM, Marmorstein R (2018) Structure 26:1137-1143. e1133 47. McGuffin LJ, Bryson K, Jones DT (2000) Bioinformatics 16:404–405 48. Moravcevic K, Mendrola JM, Schmitz KR, Wang Y-H, Slochower D, Janmey PA, Lemmon MA (2010) Cell 143:966–977 49. Li X, Wang L, Zhou XE, Ke J, De Waal PW, Gu X, Tan M, Wang D, Wu D, Xu HE (2015) Cell Res 25:50–66 50. Ovens AJ, Gee YS, Ling NX, Yu D, Hardee JP, Chung JD, Ngoei KR, Waters NJ, Hoffman NJ, Scott JW (2022) Biochemical Journal 479:1181–1204 51. Xiao B, Heath R, Saiu P, Leiper FC, Leone P, Jing C, Walker PA, Haire L, Eccleston JF, Davis CT (2007) Nature 449:496–500 52. de Leon-Boenig G, Bowman KK, Feng JA, Crawford T, Everett C, Franke Y, Oh A, Stanley M, Staben ST, Starovasnik MA (2012) Structure 20:1704–1714 53. Banik S, Dutta D (2023) J Membr Biol 256:109–124 54. Dutta D (2023) Cell Biochem Funct 41:399–412 55. Li J, Shen L, Han X, He G, Fen W, Li Y, Yang S, Zhang Z, Yang Y, Jin W, Wang Y, Zhang W, Guo Y (2023) Embo J 42:e112401. https://​doi.​org/​10.​15252/​embj.​20221​12401 56. Crute BE, Seefeld K, Gamble J, Kemp BE, Witters LA (1998) J Biol Chem 273:35347–35354 57. Pang T, Xiong B, Li J-Y, Qiu B-Y, Jin G-Z, Shen J-K, Li J (2007) J Biol Chem 282:495–506 58. Hawley SA, Boudeau J, Reid JL, Mustard KJ, Udd L, Mäkelä TP, Alessi DR, Hardie DG (2003) J Biol 2:1–16 59. Hawley SA, Pan DA, Mustard KJ, Ross L, Bain J, Edelman AM, Frenguelli BG, Hardie DG (2005) Cell Metab 2:9–19 60. Woods A, Dickerson K, Heath R, Hong S-P, Momcilovic M, Johnstone SR, Carlson M, Carling D (2005) Cell Metab 2:21–33 61. Shen W, Reyes MI, Hanley-Bowdoin L (2009) Plant Physiol 150:996–1005 Publisher's Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Journal of Molecular Modeling (2024) 30:248 Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted Page 13 of 13 248 manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. Suppl Figure 1: The pairwise sequence identitites of Arabidopsis thaliana CIPKs. In X-axis, the proteins are arranged in same sequence. Supp. Table 1. Template structures used in this study PDB ID (chain used) 2EHB (chain A) Description of the PDB Comment Reference The C-terminal domain of AtCIPK24 (SOS2) bound to AtCBL4 (SOS3) Used for the building of the NAF sequence binding conformation of CBL4 Used for the building of the PPI domain of CIPK; used for the modeling of the NAF sequence Used for the building of the kinase domain of CIPK Used to model the relative positions of the CIPK-KD and the CIPK-PPI domains (Sánchez-Barrena et al., 2007) 2EHB (chain B) 4D28 (any of the A, B, C, D chains) 6C9D (chain A) The kinase domain of AtCIPK24 (SOS2) MARK1 kinase autoinhibited by the tightly associated with its regulatory domain (KA1) Uniprot (CIPK24) Q9LDI3 4RER (chain A) AlphaFold Model 7MYJ (chain C) Uniprot (GRIK2) Q5HZ38 4G3F Phosphorylated human holoAMPK (α1β2γ1) complex bound to AMP and cyclodextrin Full length human AMPK (α1β2γ1) in complex with a small molecule activator MSG011 AlphaFold Model Murine NF-κB inducing kinase bound to 2-aminothiazoly phenol (cmp2) Used to model CIPK with different PPI conformation Used to model activated CIPK; the CBL-bound conformation of CIPK. Used to model activated CIPK; the CBL-bound conformation of CIPK (Chaves-Sanjuan et al., 2014) (Emptage, Lemmon, Ferguson, & Marmorstein, 2018) (Li et al., 2015) (Ovens et al., 2022) Used to GRIK2 reference model Used to model and compare the GRIK2 important residues (de Leon-Boenig et al., 2012) Sequence alignment continued ... Suppl Figure 2: Multiple sequence alignment of Arabidopsis thaliana CIPK sequences showing the important amino acids and motifs identified in CIPK24 from this work. At the top, the secondary structures of AtCIPK24 are shown. The important amino acid clusters GTFA motif, αD helix and the active site loop are shown. Amino acid residues found important in this study are highlighted by the down arrow. The consensus amino acids are shown below the sequence alignment. Sequence alignment is done using MAFFT (Katoh & Standley, 2013) and the diagram is prepared using ESPript (Gouet, Robert, & Courcelle, 2003). Suppl Figure 3: Comparison of the relative orientations of the bound CBL4 and the PPI domain of the CIPK24. (A) Relative orientations of the PPI domain and the CBL4 in presence of KD. (B) Relative orientation of the PPI domain and the CBL4 in absence of KD. The major PPI shift is due to the presence of KD. Suppl Table 2. Interactions identified in the structures Residues in KD Residues in PPI Comments Arg337 and Gln338 2. 3. 4. Glu95, Asp98 – Arg99, His102 Gly104 Glu138 Cys170 5. Leu208 Leu411 Pairs of interactions Glu95 – Arg337 (~ 2.0 Å), Asp98-Arg337 (~2.7 Å), and His102-Gln338 (~2.5 Å) ~ 2.0 Å ~ 2.3 Å Distance Cys170- Gly409 used to determine the shift of the active site loop May be engaged in Leu208 – Leu411 hydrophobic interactions (~4.2 Å) 1. His414 Lys405 Gly409 CBL4-CIPK24 complex in binary complex CIPK24 residues CBL4 residues 313 1. Phe Leu133, Met156, Ala160 328 2. Phe Leu193 and Tyr195 3. Glu314 Ser186 4. Ser319 Glu143 5. Arg331 Asp198 6. Lys371 Glu145 393 7. Glu Arg71 8. Arg404 Glu145 CIPK24-GRIK2 complex in ternary complex * 1. Glu19 Lys198, Ser115 2. Glu95 Lys120 98 3. Asp Arg109 5. Glu138 Lys120 7. Thr169 Gly116 Phe313 mediated hydrophobic interactions are also found in crystal structure May be engaged in hydrophobic interactions ~1.8 Å Ser319 interaction is also found crystal structure PPI domain mediated CBL4 interaction PPI domain mediated CBL4 interaction PPI domain mediated CBL4 interaction PPI domain mediated CBL4 interaction CIPK24 interaction with GRIK2 P-loop Possibly GRIK2 P-loop stabilizing interaction CIPK24 αD mediated interaction with GRIK2 Possibly GRIK2 P-loop stabilizing interaction CIPK24 active site loop and GRIK2 P-loop interaction * Critical amino acids involved in the GRIK2 and the CIPK24 are mentioned. Suppl Figure 4: Multiple sequence alignment of AtGRIK2 (Uniport: Q5HZ38) with its homologues: 4G3F_NF-kB : murine NF-kB, 2QNJ_MARK3_UAD: MARK3 ubiquitin associated domain, 4CFH_AMPK: AMP kinase, 3HX4_TgCDPk1: Toxoplasma gondii Cyclin Dependent Kinase 1, 3LIJ_CpCDPK3: Cryptosporidium paruvum Iowa II Cyclin Dependent Kinase 3, 3Q5I_PbCDPK1: Plasmodium berghei Cyclin Dependent Kinase 1, 4C0T_CaPDK1: Candida albicans Protein Dependent Kinase 1, 5LVO_Human_PDK1: Human Protein Dependent Kinase 1. The important regions discussed in the literature are shown such as the P-loop, Catalytic loop and the metal-binding region. Sequence alignment is done using MAFFT (Katoh & Standley, 2013) and the diagram is prepared using ESPript (Gouet, Robert, & Courcelle, 2003). [note: The AtGIRK2_Q5HZ38 should be read as AtGRIK2_Q5HZ38] References Chaves-Sanjuan, A., Sanchez-Barrena, M. J., Gonzalez-Rubio, J. M., Moreno, M., Ragel, P., Jimenez, M., . . . Albert, A. (2014). Structural basis of the regulatory mechanism of the plant CIPK family of protein kinases controlling ion homeostasis and abiotic stress. Proceedings of the National Academy of Sciences, 111(42), E4532-E4541 de Leon-Boenig, G., Bowman, K. K., Feng, J. A., Crawford, T., Everett, C., Franke, Y., . . . Starovasnik, M. A. (2012). The crystal structure of the catalytic domain of the NF-κB inducing kinase reveals a narrow but flexible active site. Structure, 20(10), 1704-1714 Emptage, R. P., Lemmon, M. A., Ferguson, K. M., & Marmorstein, R. (2018). Structural basis for MARK1 kinase autoinhibition by its KA1 domain. Structure, 26(8), 1137-1143. e1133 Gouet, P., Robert, X., & Courcelle, E. (2003). ESPript/ENDscript: extracting and rendering sequence and 3D information from atomic structures of proteins. Nucleic acids research, 31(13), 3320-3323 Katoh, K., & Standley, D. M. (2013). MAFFT multiple sequence alignment software version 7: improvements in performance and usability. Molecular biology and evolution, 30(4), 772-780 Li, X., Wang, L., Zhou, X. E., Ke, J., De Waal, P. W., Gu, X., . . . Xu, H. E. (2015). Structural basis of AMPK regulation by adenine nucleotides and glycogen. Cell research, 25(1), 50-66 Ovens, A. J., Gee, Y. S., Ling, N. X., Yu, D., Hardee, J. P., Chung, J. D., . . . Scott, J. W. (2022). Structure-function analysis of the AMPK activator SC4 and identification of a potent pan AMPK activator. Biochemical Journal, 479(11), 1181-1204 Sánchez-Barrena, M. J., Fujii, H., Angulo, I., Martínez-Ripoll, M., Zhu, J.-K., & Albert, A. (2007). The structure of the Cterminal domain of the protein kinase AtSOS2 bound to the calcium sensor AtSOS3. Molecular cell, 26(3), 427-435