Arabidopsis Protein Kinases GRIK1 and GRIK2 Specifically Activate SnRK1 by Phosphorylating Its Activation Loop 1

SNF1-related kinases (SnRK1s) play central roles in coordinating energy balance and nutrient metabolism in plants. SNF1 and AMPK, the SnRK1 homologs in budding yeast and mammals, are activated by phosphorylation of conserved threonine residues in their activation loops. Arabidopsis GRIK1 and GRIK2, which were first characterized as geminivirus Rep interacting kinases, are phylogenetically related to SNF1 and AMPK activating kinases. In this study, we used recombinant proteins produced in bacteria to show that both GRIKs specifically bind to the SnRK1 catalytic subunit and phosphorylate the equivalent threonine residue in its activation loop in vitro. GRIK-mediated phosphorylation increased SnRK1 kinase activity in autophosphorylation and peptide substrate assays. These data, together with earlier observations that GRIKs could complement yeast mutants lacking SNF1 activation activities, established that the GRIKs are SnRK1 activating kinases. Given that the GRIK proteins only accumulate in young tissues and geminivirus infected mature leaves, the GRIK–SnRK1 cascade may function in a developmentally regulated fashion and coordinate the unique metabolic requirements of rapidly growing cells and geminivirus-infected cells that have been induced to reenter the cell cycle.


INTRODUCTION
Protein kinases play central roles in signal transduction and regulatory pathways in eukaryotes.
They often function as cascades in which upstream kinases activate downstream kinases by phosphorylating a serine, threonine or tyrosine residue in the activation loop of the kinase domain. Phosphorylation induces a conformational change that moves the activation loop and allows access to the kinase active site. This mechanism is highly conserved for protein kinases as exemplified by the well-characterized cyclin-dependent kinase and mitogen-activated protein kinase (MAPK) cascades. More recently, sucrose non-fermenting-1 (SNF1), a kinase that modulates sugar metabolism in budding yeast, has been shown to be activated by three partially redundant kinases (for review, see Hardie, 2007). In animals, the SNF1 homolog, AMP-activated protein kinase (AMPK), is activated by two kinases that are phylogenetically related to the yeast SNF1 activating kinases (Hardie, 2007). In plants, DNA sequence analysis and yeast complementation assays have implicated the GRIKs, which were originally identified as geminivirus Rep interacting kinases (Kong and Hanley-Bowdoin, 2002), as the upstream activators of SNF1-related kinases (SnRK1) (Shen and Hanley-Bowdoin, 2006;Hey et al., 2007). An analysis of the relationship between the GRIKs and SnRK1 activation represents an important first step in understanding the role of this putative protein kinase cascade in plants. SnRK1, SNF1 and AMPK belong to a conserved family of protein kinases consisting of an α catalytic subunit and β and γ regulatory subunits (for review, see Polge and Thomas, 2007).
These kinases play central roles in regulating and coordinating carbon metabolism and energy balance in eukaryotes (for review, see Hardie et al., 1998;Halford et al., 2003 and2004;Hardie, conditions (Baena-Gonzalez et al., 2007). Many SnRK1-regulated genes are involved in plant primary and secondary metabolism, and catabolic pathways are generally up-regulated while biosynthetic pathways are down-regulated (Baena-Gonzalez et al., 2007).
SnRK1 also impacts metabolic processes during development and disease. Studies in grains, legumes and tuberous plants showed that loss of SnRK1 alters seed maturation, longevity and germination, retards root growth, and reduces starch accumulation (Zhang et al., 2001;Lovas et al., 2003;Radchuk et al., 2006;Lu et al., 2007;Rosnoblet et al., 2007), while SnRK1 over expression increases starch accumulation (McKibbin et al., 2006). Reduced expression of the SnRK1 β subunit is responsible for sugar reallocation to roots during herbivore attack (Schwachtje et al., 2006). SnRK1 has also been implicated in resistance to geminivirus infection (Hao et al., 2003). However, there is evidence that SnRK1 has additional functions beyond modulating metabolism. Studies in Arabidopsis thaliana showed that SnRK1 is essential for viability, and that plants silenced for both SnRK1.1 and SnRK1.2 are severely stunted and impaired for flowering (Baena-Gonzalez et al., 2007). Importantly, unlike Physcomitrella patens SnRK1 mutants (Thelander et al., 2004) We have designated these truncated proteins as SnRK1.1(KD) and SnRK1.2(KD), respectively.
Both proteins include 22 amino acids at their C-termini that are in the equivalent position to the AMPKα autoinhibitory region (Crute et al., 1998;Pang et al., 2007). However, it is unlikely that these residues are functionally equivalent in SnRK1 and AMPK because of their low sequence identity and lack of similarity in their predicted secondary structures. His 6 -tagged SnRK1.1(KD, K48A) and SnRK1.2(KD, K49A) were mutated in their ATP binding motifs to inhibit autophosphorylation activity. We also produced His 6 -tagged, inactive forms of full-length SnRK2.4(K33A) and the SnRK3.11 kinase domain (residues 1-334, K40A) as representatives of the functionally distinct SnRK2 and SnRK3 kinase subfamilies (Hrabak et al., 2003). All of the recombinant proteins were produced in Escherichia coli to preclude co-purification of potential regulatory partners that are conserved across eukaryotes (Hardie, 2007;Polge and Thomas, 2007).
Various combinations of recombinant GRIK and SnRK proteins were incubated in the presence of [γ-32 P]ATP, and phosphorylation was monitored by autoradiography after SDS-PAGE and transfer of the proteins to nitrocellulose membranes. We detected GRIK autophosphorylation in the reactions containing wild-type GRIK1 ( Figure 1A, lanes 1-4) and GRIK2 ( Figure 1B, lanes 1-4) but not in those containing the corresponding kinase-inactive mutants (lanes 5 and 6).

GRIK1 and GRIK2 Phosphorylate a Threonine Residue in the SnRK1 Activation Loop
SNF1 and AMPK activating kinases phosphorylate a conserved threonine that is located 11 residues upstream of the invariant subdomain VIII glutamic acid in the activation loops of SNF1 and AMPK (Hawley et al., 2003 and2005;Hong et al., 2003;Woods et al., 2003b and2005;Shaw et al., 2004;Hurley et al., 2005). The activation loop sequences of SnRK1 are nearly identical to SNF1 and AMPK, including the conserved threonine residue (Figure 2A). To determine if the GRIKs phosphorylate this threonine, we substituted alanines in place of the  We then asked if antibodies that recognize a human AMPKα phospho-T172 activation loop peptide (pT172 antibodies) can cross-react with GRIK-phosphorylated recombinant SnRK1 on immunoblots ( Figure 2D). The pT172 antibodies recognized SnRK1.1(KD, K48A) (cf. lanes 1 and 5, lanes 7 and 11) and SnRK1.2(KD, K49A) (cf. lanes 3 and 6, 9 and 12) after incubation with wild-type GRIKs but not the kinase-inactive forms. The pT172 antibodies also failed to detect SnRK1.1(KD, T175A) (lanes 2 and 8) or SnRK1.2(KD, T176A) (lanes 4 and 10) in reactions containing the wild-type GRIKs. These data established that the GRIKs phosphorylate the conserved threonine in the SnRK1 activation loop.
The activation loops of SnRK2 and SnRK3 kinases contain threonine or serine residues at the equivalent positions to the SnRK1/SNF1/AMPK loops but the flanking sequences are less conserved ( Figure 2A). We tested three SnRK2.4 and SnRK3.11 activation loop mutants with alanine substitutions in place of the threonine or serine residues in the GRIK kinase assays using The specificity of a protein kinase for its substrate is in part reflected by its affinity for the unphosphorylated substrate (Ubersax and Ferrell, 2007). To investigate GRIK−SnRK1 interactions, we asked if His 6 -tagged, wild-type SnRK1.1(KD) co-purified with GST-tagged GRIK1 or GRIK2 bound to glutathione-Sepharose resin. When equal molar concentrations of GST-GRIKs and SnRK1.1 were mixed, approximately 20% of the His 6 -SnRK1. bound fraction. Neither SnRK1.1 nor SnRK2.4 was in the bound fraction in the GST control (lanes 1 and 6). Together, these results showed that the GRIKs form stable complexes with SnRK1.1 but not SnRK2.4 in vitro, consistent with their phosphorylation specificities.
Interestingly, both kinase-inactive GRIK1(K137A) and GRIK2(K136A) displayed substantially less binding capacity for SnRK1 (lanes 3 and 5) relative to wild-type GRIKs, suggesting that the GRIK autophosphorylation site and/or the SnRK1 transphosphorylation site is important for the GRIK-SnRK1 interaction.

GRIK1 and GRIK2 Activate SnRK1
We asked if GRIK-catalyzed phosphorylation of SnRK1 impacts its kinase activity in vitro by examining the activity of His 6 -tagged, wild-type SnRK1.1(KD) in the presence or absence of functional GRIK. SnRK1.2(KD) was not included in these studies because of technical problems expressing the wild-type recombinant protein in E. coli. Initially, we examined the effect of GRIK1 ( Figure  We next used a SnRK1 peptide substrate to measure the magnitude of the effect of GRIKcatalyzed phosphorylation on SnRK1 activity in vitro. The peptide was derived from spinach (Spinacia oleracea) SPS, which is inactivated by SnRK1 phosphorylation (Huber and Huber, 1996). Other studies established that the SPS peptide, KGRMRRISSVEMMK, is phosphorylated at S158 (underlined) in a Ca 2+ -independent manner by SnRK1 purified from spinach (Huber and Huber, 1996;Sugden et al., 1999b;Huang and Huber, 2001 monitor 32 P-labeling of the SPS peptide ( Figure 4C), we detected a low, but measureable level of kinase activity for wild-type SnRK1.1(KD) by itself but not for the kinase-inactive SnRK1.1(KD, K48A). The activation loop mutant SnRK1.1(KD, T175A) had trace activity that was lower than its wild-type counterpart. When functional GRIK1 or GRIK2 was included in the reactions, wild-type SnRK1.1 activity was increased 7-fold, while no increase was observed in the reactions containing the kinase-inactive and activation loop mutant forms. Neither GRIK1 nor GRIK2 alone phosphorylated the SPS peptide, and their kinase-inactive forms did not impact wild-type SnRK1 activity. Taken together, these data established that the GRIKs activate SnRK1 by specifically phosphorylating the conserved threonine residue in its activation loop.
We examined the biochemical properties of GRIK phosphorylation of SnRK1 using the pT172 antibodies in immunoblot assays. For both GRIK1 and GRIK2, there was no detectable difference in SnRK1.1 phosphorylation in reactions containing Mg 2+ and Mn 2+ ( Figure 5A, cf. lanes 1 and 2). The presence or absence of Ca 2+ also had no effect on SnRK1.1 phosphorylation (cf. lanes 3 and 4). We detected slight differences in SnRK1.2 phosphorylation in the presence or absence of the various divalent cations, with the highest level occurring in the presence of Mg 2+ (lane 5). 5'-AMP had no detectable effect on GRIK-mediated phosphorylation of SnRK1.1 and SnRK1.2 ( Figure 5B), consistent with previous reports that 5'-AMP enhances SnRK1 activity by inhibiting its dephosphorylation (Sugden et al., 1999a). SnRK1.1 and SnRK1.2 phosphorylation levels were also not impacted by 20 μM STO-609 ( Figure 5C), which inhibits CaMKKβ but not LKB1 activity in vitro (Hawley et al., 2005).

GRIK Expression and SnRK1 Phosphorylation Overlap in Planta
Earlier studies showed that GRIK1 and GRIK2 proteins only accumulate in the SAM and very young leaves (< 0.5 cm in length) of healthy Arabidopsis rosette plants and in older leaves of geminivirus-infected plants (Shen and Hanley-Bowdoin, 2006). Using the pT172 antibodies, we detected activated SnRK1 on immunoblots of protein extracts from the SAM and young leaves ( Figure 6A, lane 1) and leaves infected with Cabbage leaf curl virus (CaLCuV) ( Figure   6B, lane 2). GRIK1 was also detected in these extracts by its antibodies. However, unlike GRIK1, activated SnRK1 was also found in extracts from expanding ( Figure 6A difference in the amount of activated SnRK1 in mock versus CaLCuV-infected leaves ( Figure   6B, cf. lanes 1 and 2). Based on these results, we can conclude that the GRIK accumulation and and activated SnRK1 patterns overlap, consistent with the GRIKs acting upstream of SNRK1 in young tissues and during geminivirus infection. Our data also suggested that a protein kinase(s) unrelated to the GRIKs phosphorylates SnRK1 in mature tissues.

DISCUSSION
SnRK1 plays a central role in coordinating energy balance and nutrient metabolism in plants (for review, see Halford et al., 2003 and2004;Polge and Thomas, 2007;Baena-Gonzalez and Sheen, 2008). It promotes cellular catabolism and inhibits macromolecular synthesis in response to nutrient limitation and energy deprivation by directly regulating a number of enzymes related to carbon and nitrogen metabolism and by regulating the transcript levels of more than a thousand genes (Baena-Gonzalez et al., 2007). The SnRK1 homologs, SNF1 and AMPK, are also major players in cellular processes related to energy and carbon source regulation in budding yeast and mammals (for review, see Hardie et al., 1998;Hardie, 2007). Both SNF1 and AMPK are activated by a phylogenetically related group of upstream kinases that phosphorylate conserved threonine residues in their activation loops (Hawley et al., 2003 and2005;Hong et al., 2003;Woods et al., 2003a and2005;Shaw et al., 2004;Hurley et al., 2005). Earlier studies showed that SnRK1 is also phosphorylated at the equivalent threonine residue in vivo (Sugden et al., 1999a), but the identities of the plant kinases that activate SnRK1 have proven elusive. In this report, we demonstrate that the Arabidopsis kinases, GRIK1 and GRIK2, are SnRK1 activating kinases.
The SnRK family consists of three subfamilies -SnRK1, SnRK2 and SnRK3 (Hrabak et al., 2003 GRIK substrates, and mutations in their activation loops had minimal impact on the outcome of the in vitro assays. We cannot formally rule out the possibility that other members of the SnRK2 or SnRK3 subfamilies are GRIK substrates but this seems unlikely given the divergent nature of the activation loop sequences of the three SnRK subfamilies (Figure 2A). The ability of the GRIKs to form stable protein complexes with SnRK1.1 but not SnRK2.4 in vitro also underscores their specificity. SNF1 and AMPK are activated by phosphorylation of the threonine residues in their activation loops. Two lines of evidence indicated that GRIK-catalyzed phosphorylation of the threonine in the SnRK1.1 activation loop also resulted in kinase activation. First, SnRK1.1 autophosphorylation activity was detected only in the presence of the GRIKs. Second, SnRK1.1 phosphorylation of the SPS peptide, a cognate SnRK1 substrate (Huber and Huber, 1996;Sugden et al., 1999b;Huang and Huber, 2001), was elevated 7-fold in the presence of either GRIK1 or GRIK2. Both of these assays depended on the presence of the threonine residue in the SnRK1 activation loop, indicating that its phosphorylation by the GRIKs is essential for full activation. There are reports describing active SnRK1 in vitro in the absence of the GRIKs (Barker et al., 1996;Sugden et al., 1999b;Hao et al., 2003). These studies used either SnRK1 purified from plant tissues or produced in eukaryotic expression systems. In both cases, it is possible that the SnRK1 protein was already activated at the time of purification by endogenous upstream kinases, consistent with the observation that protein phosphatase treatment significantly reduced the SnRK1 activity isolated from spinach (Sugden et al., 1999a and1999b).
Our use of a bacterial expression system precluded prior activation of SnRK1 by endogenous kinases and enabled us to establish unequivocally that the GRIKs are SnRK1 activating kinases. GRIK1 and GRIK2 are phylogenetically related to SNF1 and AMPK activating kinases and share some biochemical properties with them (Wang et al., 2003;Shen and Hanley-Bowdoin, 2006;Hey et al., 2007). GRIK-mediated activation of SnRK1 is not affected by 5'-AMP, which greatly enhances AMPK activation by LKB1 and to lesser extent by CaMKKβ ( Figure 5 We also did not see a change in activated SnRK1 levels in geminivirus-infected mature leaves ( Figure 6B). Earlier studies showed that the GRIKs only accumulate in virus-positive cells (Kong and Hanley-Bowdoin, 2002;Shen and Hanley-Bowdoin, 2006), which constitute less than 10% of the cells in a CaLCuV-infected Arabidopsis leaf (Ascencio-Ibañez et al., 2008). Thus, it was not unexpected that there was no detectable difference in the overall levels of phosphorylated SnRK1 in mock-inoculated and infected leaves. However, the levels of activated SnRK1 may vary between virus-positive cells and adjacent uninfected cells and differentially impact metabolic processes and host gene expression in the two cell populations. This idea is supported by the observation that geminiviruses encode two proteins that interact with the GRIK-SnRK1 cascade (Kong and Hanley-Bowdoin, 2002;Hao et al., 2003;Shen and Hanley-Bowdoin, 2006 control SnRK1 signaling during these diverse processes. The GRIK-SnRK1 cascade may play a key regulatory role in young tissues and geminivirus-infected cells, both of which can support DNA replication and are likely to have high metabolic requirements. The cascade may ensure adequate energy and nutrient supplies to rapidly growing cells and impact plant cell cycle controls by modulating sucrose levels (Riou-Khamlichi et al., 2000). In animals, LKB1 was originally identified as a tumor suppressor gene (Boudeau et al., 2003). The LKB1-AMPK pathway is required for normal mitotic processes and has been implicated in determining cell polarity (for review, see Koh and Chung, 2007;Williams and Brenman, 2008). In addition to phosphorylating and activating SnRK1 in young tissues, the GRIKs may influence substrate recruitment. We recently identified 5 Arabidopsis transcription factors that bind to the GRIKs in a yeast two-hybrid screen (unpublished data). The GRIKs may serve as bridging proteins between the transcription factors and SnRK1, which in turn may phosphorylate these factors. It is also possible that the GRIKs phosphorylate other proteins in addition to SnRK1 and impact developmental processes independent of SnRK1 signaling. Future experiments will address the roles of the GRIK-SnRK1 cascade and the GRIKs alone during development and geminivirus infection.

Plant growth and protein and RNA preparations
Arabidopsis (Arabidopsis thaliana) Col-0 plants were grown in soil at 20°C in a Percival reachin chamber with 8/16-h light/dark cycle and a light intensity of 15,000 Lux. Leaves at different developmental stages were collected from 6-week old rosette plants. CaLCuV-infected leaves (0.5-1.5 cm long) were collected 12 dpi from 7-week old plants agroinoculated with pNSB1090 and pNSB1091, which contain partial tandem copies of CaLCuV A and B DNAs, respectively Immunoblots of plant proteins were visualized using horseradish peroxidase-conjugated secondary antibodies and the SuperSignal West Pico chemiluminescencent Substrate (Pierce).

Protein kinase assay
For assays with protein substrates, a 50 μL reaction containing approximately 250 nM of the kinase and the substrate in 25 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM EGTA, 1 mM DTT, to SDS-PAGE. Separated proteins were transferred to nitrocellulose membrane. The 32 P-labeled proteins were visualized by autoradiography followed by immunoblotting of total proteins on the membrane. When pT172 antibodies were used to assay SnRK1 phosphorylation, only unlabeled ATP was included in the reaction and phosphorylation was detected by immunoblotting. The kinase inhibitor STO-609 was from Sigma-Aldrich. When the plant SPS peptide acetyl-KGRMRRISSVEMMK (GenScript, Scotch Plains, NJ) was used as substrate (Huang and Huber, 2001), the reactions were stopped by transferring 40 μ L aliquots to P81 filter paper discs, which were rinsed three times with 75 mM H 3 PO 4 and dried. Bound radioactivity was measured in a liquid scintillation counter.

Supplemental Materials
The following materials are available in the online version of this article.    and 10), as indicated above the top panel. In B, C and D, total GST-GRIK and His 6 -SnRK were visualized using anti-GST (middle panels) and anti-His 6 antibodies (lower panels).  were visualized using anti-GST (middle panels) and anti-His 6 antibodies (lower panels). C. His 6tagged wild-type (wt) SnRK1.1 kinase domain, its kinase-inactive form (m), or activation loop mutant (m) was incubated alone or in the presence of GST-tagged wild-type (wt) or kinaseinactive (m) GRIK1 or GRIK2. SnRK1 kinase activity was monitored by 32 P-labeling of a peptide substrate derived from sucrose-phosphate synthase. The mean activities and standard deviations for three experiments using different SnRK preparations are shown.