Total RNA was extracted from leaves of 15-day-old maize B73 plants using PureLink RNA mini kit (Life Technologies, Carlsbad). Briefly, leaves were collected, immediately frozen in liquid nitrogen, and then powdered using a mortar. Two hundred mg of frozen powder was added to 1 mL of TRIzol (Invitrogen, Carlsbad). After 10 min incubation at room temperature with gentle shaking, 200 μL of chloroform was added and the mixture was homogenized under vigorous agitation using a vortex. After centrifugation for 15 min at 12,000 ×g at 4°C, the upper aqueous phase was collected and applied to a PureLink RNA (Life Technologies, Carlsbad) column. The column was washed with Wash buffer I followed by Wash buffer II and eluted in RNAse-free water as per the manufacturer’s instructions. cDNA was synthesized using SuperScript III reverse transcriptase (Invitrogen, Carlsbad). Two microgram of total RNA were incubated with oligo dT20 and dNTP for 5 min at 65°C and then cooled on ice. Buffer, DTT and SuperScript enzyme were added and incubated for 60 min at 50°C. The enzyme was inactivated for 15 min at 70°C and the cDNA stored at -20°C until use.

Four different N-terminal truncations of the ZmSIRK1 kinase domain (residues His623-Ser1045; Ser737-Ser1045; Ser751-Ser1045; and Ser760-Ser1045) were cloned into pNIC28a-Bsa4 (Savitsky et al., 2010). Forward primers used were: sirk1-H623-F (TACTTCCAATCCATGCATTGGAAGATCAGTAGCTGGAAA AG); sirk1-S737-F (TACTTCCAATCCATGTCAGTTGTGTTCA CGGCTGAAG); sirk1-S751-F (TACTTCCAATCCATGTCT CCTGATAAACTGGTTGGGG); sirk1-S760-F (TACTTCCA ATCCATGTCTTGTGCCCCTGCTGAG) and reverse primer used was sirk1-S1045-R (TATCCACCTTTACTGTCACGAC GATAAGGACAAGAGATC). The cDNA synthesized from total RNA of maize B73 leaves was used as a template for PCR. After PCR amplifications, amplicons were treated with T4-DNA polymerase for ligase-independent cloning (LIC) (Savitsky et al., 2010). Positive clones for the four constructs were transformed into BL21(DE3)-R3-pRARE2 and grown in 500 μL in 96 × 1 mL well-block. After overnight growth, individual cultures were diluted 1:100 in LB medium with kanamycin (50 μg/mL) and incubated at 37°C with shaking. When the OD₆₀₀ reached 2, the temperature was lowered to 18°C, 0.1 mM of IPTG was added and the culture incubated overnight with shaking. Cultures were then centrifuged at 3,500 ×g for 20 min at 4°C and lysed using 200 μL of Lysis buffer (50 mM HEPES pH 7.5; 500 mM NaCl; 10% glycerol; 10 mM imidazole; 500 μM TCEP; 0.1% dodecyl maltoside; 1 mM MgCl₂; 1:200 protease inhibitor; 0.5 mg/mL lysozyme; 50 units/mL benzonase). After lysis, cultures were centrifuged at 3,500 ×g for 10 min at 4°C and the supernatant was incubated for 1 h at 18°C with 50 μL of Ni²⁺-sepharose beads (GE Healthcare, Uppsala). After washing with wash buffer (50 mM HEPES pH 7.5; 500 mM NaCl; 10% glycerol; 30 mM imidazole; 500 μM TCEP), purified proteins were eluted with 50 μL of 50 mM HEPES pH 7.5; 500 mM NaCl; 10% glycerol; 300 mM imidazole; 500 μM TCEP. Expression and solubility were verified in 12.5% SDS-PAGE (Laemmli, 1970).

Vector pNIC28a-Bsa4 harboring construct ZmSIRK1^737-1045 was transformed into competent BL21(DE3)-R3-pRARE2 Escherichia coli cells. Pre-culture was grown in 20 ml of LB media grown overnight and then inoculated into 1.5 L of Terrific Broth at 37°C until OD₆₀₀ of 1.5. The culture was cooled down to 18°C, 0.2 mM of IPTG was added to the medium and growth resumed overnight. Cells were collected by centrifugation (15 min at 7,500 ×g at room temperature). Cell pellet was suspended in 2× binding buffer (1× binding buffer is 500 mM HEPES; 500 mM NaCl; 5% glycerol; 10 mM imidazole; 1 mM TCEP) with protease inhibitors (1:200) and frozen at -80°C until use. Suspended cell pellets were thawed and sonicated for 9 min at 4°C (5 s ON; 10 s OFF; Amp 30%). One ml of 5% polyethyleneimine (pH 7.5) was added per 30 ml of lysate and the sample was centrifuged at 53,000 ×g for 45 min at 4°C. The supernatant was loaded onto an IMAC column (5 ml HisTrap FF Crude) and washed in binding buffer with 30 mM imidazole. Recombinant protein was eluted with elution buffer (binding buffer with 300 mM imidazole). To remove the 6xHis-tag, eluted protein was incubated with TEV protease and the tag removed using nickel beads. The protein solution was loaded onto a size exclusion HiLoad 16/60 Superdex 200pg (GE) column equilibrated in a gel filtration buffer (binding buffer without imidazole). Fractions of 1.8 mL were collected and verified for protein purity in a 12.5% SDS-PAGE gel. Purified fractions were pooled together and stored at -80°C.

A mixture containing equimolar quantities of adenylyl-imidodiphosphate (AMP-PNP) and MgCl₂ was added to purified ZmSIRK1^737-1045 (850 μM) at threefold molar excess. The solution was incubated on ice for approximately 30 min. The mixture was centrifuged at 20,000 ×g for 10 min at 4°C prior to setting up 150 nl volume sitting drops at three ratios of the protein-inhibitor complex to reservoir solution (2:1, 1:1, or 1:2). Crystallization experiments were performed at 20°C. The best-diffracting crystals grew under the conditions described in Table 1, first identified from the Morpheus Crystallization screen (Gorrec, 2015). Crystals were cryoprotected in reservoir solution supplemented with 20–25% glycerol before flash-freezing in liquid nitrogen for data collection. Diffraction data were collected at the Advanced Photon Source (APS), integrated using XDS (Kabsch, 2010) and scaled using AIMLESS from the CCP4 software suite (Winn et al., 2011). Molecular replacement (MR) was performed with Phaser (McCoy et al., 2007) using the kinase domain of BAK1 interacting RLK 2 (BIR2) as the search model (PDB ID 4L68) (Blaum et al., 2014). Automated model building was performed with Buccaneer (Cowtan, 2006) following density modification with Parrot (Zhang et al., 1997). Automated refinement was performed in Buster (Global Phasing Ltd.). Coot (Emsley et al., 2010) was used for manual model building and refinement. Structure validation was performed using MolProbity (Chen et al., 2010). Structure factors and coordinates (Table 1) have been deposited in the PDB.

Thermal stabilization (DSF) assays were performed essentially as described (Niesen et al., 2007), with the following modifications. Purified ZmSIRK1^737-1045 was screened against a library of 378 structurally diverse and cell-permeable ATP-competitive kinase inhibitors library from Selleckchem (Houston, TX, United States; catalog No. L1200). DSF experiments were performed in a 384-well plate format. Each well contained 25 μL of 1 μM kinase in DSF buffer (100 mM K-phosphate, 150 mM NaCl, 10% glycerol, pH 7.5) and the Protein Thermal Shift dye at the recommended concentration of 1:1000 (Applied Biosystems; the composition of the buffer and the dye solutions are not disclosed). Compounds (10 mM) in DMSO were added to 10 μM final concentration (0.1% final DMSO). Plates were sealed using optically clear films and transferred to a QuantStudio 6 qPCR instrument (Applied Biosystems, Singapore). Fluorescence intensity data were acquired in a temperature gradient from 25 to 95°C at a constant rate of 0.05°C/s and protein melting temperatures were calculated based on a Boltzmann function fitting to experimental data, as implemented in the Protein Thermal Shift Software (Applied Biosystems, Singapore). Protein in 0.1% DMSO was used as a reference. Compounds displaying a positive temperature shift (ΔTm) of 2°C or higher compared to the control were considered positive.

Solutions containing 50 μM purified ZmSIRK1^737-1045 protein (cell) and 1 mM AMP-PNP (injectant) were prepared in ITC buffer (50 mM K-phosphate; 500 mM NaCl, 5% glycerol; 1 mM TCEP; 50 mM MgCl₂). For ITC measurements, 50 μM of each compound (cell) and 500 μM purified ZmSIRK1^737-1045 (injectant) were prepared in ITC buffer without Mg²⁺. ITC experiments were performed in MicroCal AutiITC200 (GE, Northampton) and titrations were carried out at 25°C with a stirring speed of 750 rpm and 300 s between each 2 μL injection. Controls with buffer and ligand alone were performed to verify the dilution heat and were subtracted from protein-ligand data. Data were analyzed in Origin 7 software package.

The kinase domain of ZmSIRK1 was used to search related proteins within UniProt Knowledge database (e-value of 10^-4; limit: 1,000 sequences). To verify the prevalence of putative phosphorylation sites within the activation segment of ZmSIRK1 homologs we developed a script in java. The script performed a search within the identified proteins and reported on the presence/absence of the conserved DYS motif. If the DYS motif was present, then the script reported on the presence or absence of possible phosphorylation sites (Tyr, Ser, or Thr residues) within SIRK1 activation segment.

Constructs harboring the kinase domain of ZmSIRK1 were tested for recombinant expression in E. coli to obtain high yields of soluble protein. Four N-terminal truncations of the ZmSIRK1 cytoplasmic domain were designed and cloned into pNIC28a-Bsa4 (Figure 1A and Supplementary Figures S1A,B). Large-scale protein expression using construct ZmSIRK1^737-1045 yielded ∼110 mg of ≥98% pure recombinant protein using 1.5 L of culture (Supplementary Figures S1C–E). The identity of recombinant ZmSIRK1 kinase domain was verified by mass spectrometry (data not shown).

Crystallization conditions for purified ZmSIRK1 kinase domain were identified using commercially available sparse matrix crystallization screens, resulting in large ZmSIRK1-AMP-PNP co-crystals (Table 1 and Supplementary Figure S1F). The crystal structure of ZmSIRK1 kinase domain bound to the non-hydrolysable ATP analog, AMP-PNP, was solved by molecular replacement using the kinase domain of AtBIR2, an RLK from Arabidopsis, as the search model (PDB ID 4L68; 35% sequence identity) (Blaum et al., 2014) (Table 1). Overall, the ZmSIRK1 kinase domain has the canonical kinase topology, with an N-terminal lobe composed mostly of β-sheets connected by a flexible linker (hinge region) to a C-terminal lobe predominantly composed of α-helices (Figure 2A). In the crystal structure, AMP-PNP is bound to the ATP-binding pocket of the ZmSIRK1 kinase domain, located in a cleft between N- and C-terminal lobes. Electron density maps allowed placement of most residues in ZmSIRK1. The final model consisted of residues 744–972, 977–1000, and 1009–1045. Omitted regions of the polypeptide chain in the final model located to the protein N-termini or to loop regions in the C-terminal lobe, which are likely to be disordered due to flexibility.

We compared the structure of ZmSIRK1 against those in the PDB using the DALI server (Holm and Rosenstrom, 2010). This search identified the kinase domains of AtBIR2 (BAK1-interacting RLK 2) and AtBSK8 (brassinosteroid signaling kinase 8) as the closest structural neighbors to ZmSIRK1. Both AtBIR2 and AtBSK8 are known to be pseudo-kinases and to lack catalytic activity (Grutter et al., 2013; Blaum et al., 2014). The next closest structural neighbors to ZmSIRK1 were the human protein kinase IRAK4 (interleukin-1 receptor-associated kinase 4) (Wang et al., 2006) and the kinase domain of the brassinosteroid insensitive membrane receptor from Arabidopsis (AtBRI1) (Bojar et al., 2014) (Figures 2A–E).

Compared to its closest structural neighbors, the most notable feature of ZmSIRK1 kinase domain is the unusual architecture of its activation segment (highlighted in orange in Figure 2). The activation segment of kinases extends from the DFG (⁹⁰⁷DYS⁹⁰⁹ in ZmSIRK1) to the APE (⁹³³PPE⁹³⁵ in ZmSIRK1) motifs. In most protein kinases, this region is composed of a flexible loop. The position of this loop controls the activations state of the enzyme. Kinases in a fully active state have activation segments in an extended or open conformation. By contrast, in the inactive kinase conformation, the position of the activation segment occludes the peptide-binding site. In ZmSIRK1, the activation segment is highly structured. The region immediately C-terminal to the DYS motif forms a single turn 3₁₀ helix that packs against αC. This helical turn is followed by a short loop region and an α-helix (P916^ZmSIRK1-L929^ZmSIRK1). The position and organization of the ZmSIRK1 activation segment are stabilized by hydrogen bonds to residues from both lobes of the protein (Figure 3A). Residue Gln922 within the α-helix hydrogen bonds to residues in both the glycine-rich (H772^ZmSIRK1) and catalytic loops (G887^ZmSIRK1). Moreover, Y931^ZmSIRK1, located at the C-terminal end of the activation loop, hydrogen bonds to catalytic loop residue S891^ZmSIRK1 and to E959^ZmSIRK1 from α-helix F (subdomain IX in PKA). These structures partly occlude the kinase ATP-binding site and peptide substrate-binding pocket, hallmarks of the inactive state (Figure 3B). An equivalent 3₁₀ helix has been observed in the inactive state of mammalian kinases, such as ABL, SRC, and CDK2 (De Bondt et al., 1993; Schindler et al., 1999; Xu et al., 1999; Levinson et al., 2006). However, in these proteins, the 3₁₀ helix is followed by a mostly disordered region, whereas in ZmSIRK1 we observe a short loop followed by an α-helix (Figure 3C).

In most kinases, phosphorylation of regulatory Ser/Thr residues within the activation segment stabilizes the active state conformation of the kinase by counteracting an Arg residue within the HRD motif. Within the ZmSIRK1 activation segment, there are three putative phosphorylation sites (Figure 4A). The first, T915^ZmSIRK1, is located in the loop between the helical elements within the activation segment. The next two possible phosphorylation sites, Y931^ZmSIRK1 and S932^ZmSIRK1, are found immediately after the α-helix. There are also three other possible phosphorylation sites immediately C-terminal to the activation segment (residues 937–940). Inspection of electron density maps did not suggest these residues are phosphorylated in the ZmSIRK1-AMP-PNP crystal structure. Moreover, the HRD motif in ZmSIRK1 is degraded and the positively charged Arg is replaced with a Gly. In some of ZmSIRK1 closest structural neighbors, both from plant and mammals, phosphorylation of Ser/Thr residues within the activation domain is required for activity (Hantschel and Superti-Furga, 2004; Wang et al., 2005; Cheng et al., 2007; Kuglstatter et al., 2007; Yan et al., 2012) (Figure 4B).

We reasoned that if these putative phosphorylation sites are involved in protein activation, they should be conserved amongst SIRK1 homologs. A search within UniProt using the kinase domain of ZmSIRK1 revealed 124 proteins, exclusively from plants, bearing the unusual DYS motif present in maize SIRK1 (sequence identities between 29 and 98%). We then interrogated these sequences for the presence of putative phosphorylation sites (Thr/Tyr/Ser residues) at equivalent positions within and immediately C-terminal of ZmSIRK1 activation segment – +6, +22, +23, and +29–31 positions C-terminal from the DYS motif. We found that 119 out of 124 proteins with a DYS motif in Uniprot displayed a Thr residue at position +6, which locates to the spacer loop between the 3₁₀ and α helices within ZmSIRK1 activation segment. We also found that all 124 proteins displayed a Tyr residue at positions +22 – located immediately C-terminal to the α-helical segment in ZmSIRK1; in addition to having a Thr residue at position +6. Conservation at positions +29 (Ser), +30 (Ser/Thr), and +31 (Ser/Thr) was also high (ranging from 80 to 90%), whereas sites at +23 and +28 were markedly less conserved (<21%). Overall, co-conservation of the most prevalent residues occurred in 79 out of the 124 identified proteins (∼64%). These analyses also identified 209 plant SIRK1 homologs with the DYC motif seen in AtSIRK1. Conservation of possible phosphorylation sites within these proteins is similar, but not identical to those identified above for ZmSIRK1 (Supplementary Table S1).

In many kinases, the positions of αC and of the DFG motif in the active and inactive states are also different. In the active conformation, αC swings inward toward the ATP-binding pocket (αC-in conformation) and facilitates the formation of a salt bridge between a conserved glutamic residue within αC (E805^ZmSIRK1) and the catalytic lysine (K789^ZmSIRK1). This interaction is key for enzyme activity. By contrast in the inactive αC-out conformation observed for a few mammalian proteins, this conserved glutamic residue is solvent-exposed and makes a salt link to an Arg residue from the kinase activation segment (Levinson et al., 2006). The interaction of the conserved aspartic residue within the DFG motif to the Mg²⁺ ion is also critical for transphosphorylase activity. This interaction is only possible in the so-called DFG-in conformation. In this configuration, the hydrophobic residue within the DFG motif takes part in a series of hydrophobic contacts that stabilizes the active state conformation of the enzyme (R-spine) (Kornev et al., 2006). In an active kinase, both DFG-in and αC-in configurations are usually present.

The kinase domain in the ZmSIRK1-AMP-PNP co-structure displays a DYS-in conformation and the side chain atoms from D907^ZmSIRK1 within this motif participates in the octahedral coordination of the Mg²⁺ ion bound to AMP-PNP. However, coordination of the divalent ion is unusual (see below) and the position of Y908^ZmSIRK1 in the ZmSIRK1-AMP-PNP co-structure prevents assembly of the hydrophobic residues in the kinase domain core characteristic of active state proteins. The αC in ZmSIRK1 is away from the ATP-binding site (αC-out) and the side chain of E805^ZmSIRK1 is solvent exposed and makes a hydrogen bond to R912^ZmSIRK1 in the activation segment. Moreover, the hydroxyl group from Y908^ZmSIRK1 hydrogen bonds to the main chain carbonyl group from L809^ZmSIRK1 in αC, further stabilizing the observed “twisted-out” configuration of this secondary structure (Figure 5A). The DFG-in/αC-out out configuration has been observed for a few mammalian kinases, such as ABL, SRC, and CDK2 (Figure 5B), but thus far not in any plant kinase domains. For the mammalian kinases, phosphorylation of activations segment Ser/Thr residues can stabilize the active, DFG-in/αC-in conformation (Figure 5C).

In the co-crystal structure, the ATP analog adopts an unusual conformation. The ribose is found in a C2^′-endo conformation, compared to the more commonly observed 3^′-endo puckering. The position of the β-phosphate is also quite distinct and, consequently, so are the position and coordination of the Mg²⁺ ion (Figure 5D). The conformation observed for AMP-PNP here is similar to that in the inactive state structure of the mammalian protein CDK2 (Schulze-Gahmen et al., 1996) (Figure 5E) and contrasts with that observed for kinases in their active state, especially in what concerns the position of the nucleotide phosphate β (Figure 5F).

The position of the ATP-analog within the ZmSIRK1 binding pocket is also unusual. The two lobes of a kinase domain are connected by a flexible “hinge” region. In ZmSIRK1, the hinge region spans residues S836-L844^ZmSIRK1. In general, ATP and its analogs are anchored to the kinase domain via two hydrogen bonds. The adenosine N6 amine and N1αα imine interact with the backbone of residues located at +1 and +3 positions relative to the so-called gatekeeper residue (GK, S836^ZmSIRK1), respectively. These interactions ensure the nucleotide sits deep within the enzyme binding pocket. However, in ZmSIRK1-AMP-PNP co-crystals, the N6 amine from the ligand interacts with the more solvent-exposed residue at GK+3 (A839^ZmSIRK1). This is the only hydrogen bond between ligand and hinge residues. As a result, AMP-PNP sits closer to the solvent than it is normally observed for other kinases.

In most kinases, proper orientation of ATP phosphate groups within the kinase nucleotide-binding pocket relies on the close contacts afforded by glycine residues within the protein P-loop and is crucial for catalysis. As for other plant RLKs, the P-loop of ZmSIRK1 does not display the canonical consensus sequence (GxGxxG). In ZmSIRK1, the second Gly within this motif is replaced with S770^ZmSIRK1. In the ZmSIRK1-AMP-PNP co-structure, the side chain from this serine residue hydrogen bonds to the γ-phosphate in AMP-PNP. The bulkier side chain of S770^ZmSIRK1 also prevents the nucleotide β-phosphate to interact closely with main chain groups from the P-loop. Instead, the phosphate groups from AMP-PNP interact mostly with protein residues in the protein C-terminal domain, opposite to the N-terminal P-loop.

In Arabidopsis, the expression of AtSIRK1 is induced by sucrose and the enzyme phosphorylates an aquaporin regulating water content inside cells. Identification of potent antagonists of ZmSIRK1 may help elucidate the role of this kinase during drought stress. Some of the closest structural homologs of ZmSIRK1 are pseudo-kinases, including BIR2 which does not bind ATP (Blaum et al., 2014). We thus determined if purified recombinant ZmSIRK1 could bind the non-hydrolysable ATP analog AMP-PNP using isothermal titration calorimetry (ITC). The ITC results indicated that the protein can bind the AMP-PNP in a magnesium-dependent manner with a K_D of 2.8 μM (Figure 6).

We next used thermal denaturation assays (DSF) to identify small molecules within a 378-compound collection of chemically diverse human kinase inhibitors that could interact with ZmSIRK1. We identified three compounds with temperature shifts, ΔTms, over 2.0°C, an arbitrary cut-off for a positive hit in this experiment (Table 2). DSF results for these compounds were further confirmed using ITC. The compounds PP242, INK 128 and PP121 bound ZmSIRK1 with K_Ds of 0.46, 1.76, and 1.99 μM, respectively (Table 2 and Supplementary Figure S2).

The ZmSIRK1-ATP-PNP co-structure showed a partly closed ATP-binding site. To assess if compounds identified by DSF and verified by ITC would be able to fit into the ATP-binding pocket of ZmSIRK1, we performed a simple rigid-body docking of PP121 to the ZmSIRK1-AMP-PNP co-structure. This exercise used the available co-structure of PP121 bound to the human kinase domain SRC (PDB ID 3EN4) (Apsel et al., 2008) and suggested PP121 can fit in the ATP-binding pocket of ZmSIRK1 and occupies the bottom of this cavity with few steric clashes to protein residues (Supplementary Figure S3). PP242 and INK 128 are closely related to PP121 (Table 2) and we expect they could also fit without major clashes in the nucleotide-binding pocket of ZmSIRK1. We have not yet been able to obtain co-crystal structures with these synthetic ligands.