Phosphorylation and 14-3-3 binding of Arabidopsis trehalose-phosphate synthase  5 in response to 2-deoxyglucose


*(fax +44 1382 223778; e-mail


Trehalose-6-phosphate is a ‘sugar signal’ that regulates plant metabolism and development. The Arabidopsis genome encodes trehalose-6-phosphate synthase (TPS) and trehalose-6-phosphatase (TPP) enzymes. It also encodes class II proteins (TPS isoforms 5–11) that contain both TPS-like and TPP-like domains, although whether these have enzymatic activity is unknown. In this paper, we show that TPS5, 6 and 7 are phosphoproteins that bind to 14-3-3 proteins, by using 14-3-3 affinity chromatography, 14-3-3 overlay assays, and by co-immunoprecipitating TPS5 and 14-3-3 isoforms from cell extracts. GST–TPS5 bound to 14-3-3s after in vitro phosphorylation at Ser22 and Thr49 by either mammalian AMP-activated protein kinase (AMPK) or partially purified plant Snf1-related protein kinase  1 (SnRK1s). Dephosphorylation of TPS5, or mutation of either Ser22 or Thr49, abolished binding to 14-3-3s. Ser22 and Thr49 are both conserved in TPS5, 7, 9 and 10. When GST–TPS5 was expressed in human HEK293 cells, Thr49 was phosphorylated in response to 2-deoxyglucose or phenformin, stimuli that activate the AMPK via the upstream kinase LKB1. 2-deoxyglucose stimulated Thr49 phosphorylation of endogenous TPS5 in Arabidopsis cells, whereas phenformin did not. Moreover, extractable SnRK1 activity was increased in Arabidopsis cells in response to 2-deoxyglucose. The plant kinase was inactivated by dephosphorylation and reactivated by phosphorylation with human LKB1, indicating that elements of the SnRK1/AMPK pathway are conserved in Arabidopsis and human cells. We hypothesize that coordinated phosphorylation and 14-3-3 binding of nitrate reductase (NR), 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (F2KP) and class II TPS isoforms mediate responses to signals that activate SnRK1.


The disaccharide trehalose is widespread in bacteria, fungi and insects. It is not thought to be produced by mammals (Elbein et al., 2003). Although trehalose is produced in large quantities by anhydrobiotic organisms under stress (Elbein et al., 2003), Arabidopsis plants contain only trace amounts. Consequently, trehalose was only discovered in these plants recently (Goddijn et al., 1997; Müller et al., 2000, 2001; Vogel et al., 2001). Since then, a compelling case has emerged for considering its biosynthetic precursor, trehalose-6-phosphate (T6P), to be a ‘sugar signal’ that exerts growth factor-like regulatory effects on nutrient assimilation, carbon partitioning and development (van Dijken et al., 2004; Eastmond et al., 2002; Goddijn and van Dun, 1999; Kolbe et al., 2005; Paul et al., 2001; Pellny et al., 2004; Schluepmann et al., 2003).

The discovery of multiple Arabidopsis genes encoding trehalose-synthesizing enzymes suggests that trehalose metabolism could be highly regulated (Leyman et al., 2001). Trehalose comprises two glucose units with an α,α-glycosidic linkage. In one major pathway of trehalose synthesis, glucose is transferred from UDP-glucose to glucose-6-phosphate (G6P) to form trehalose-6-phosphate (T6P) and UDP. This step is followed by the dephosphorylation of T6P to free trehalose (Elbein et al., 2003). The Arabidopsis genome encodes four trehalose-6-phosphate synthase-like proteins (TPS isoforms 1–4). Of these, TPS1 was shown to have T6P-synthesizing activity (Van Dijck et al., 2002; Vogel et al., 2001). Moreover, Arabidopsis has ten trehalose-6-phosphatases (Schluepmann et al., 2004), and seven class II proteins (TPS isoforms 5–11) with both N-terminal TPS-like and C-terminal trehalose-6-phosphatase(TPP)-like domains (Leyman et al., 2001; Müller et al., 2001; Schluepmann et al., 2004; Vogel et al., 2001). Whether the class II proteins actually have TPS and/or TPP activities is unknown (Vogel et al., 2001).

TPS1 plays critical roles throughout the life cycle of a plant: at the stage of embryogenesis when cells expand and storage reserves are laid down (Eastmond et al., 2002), during normal vegetative growth, and at the transition to flowering (van Dijken et al., 2004). However, seedlings that over-express TPS1 have slightly elevated trehalose and T6P levels. This, in turn, causes phenotypic effects on the plant, including lack of the usual repression of both photosynthesis and growth in response to exogenous glucose, and some insensitivity to ABA (Avonce et al., 2004). The levels of TPS1 mRNA, and the mRNAs for several class II proteins, are highly regulated by stimuli including nitrate, abiotic stress, ABA, sugars and the circadian cycle (Leonhardt et al., 2004; Scheible et al., 2004; Schluepmann et al., 2004; Wang et al., 2003). For example, the genes for AtTPS6, 8, 9 and 10 behave as immediate early genes whose mRNAs are expressed quickly when sugar starvation is induced in leaves by extending the night period (Leonhardt et al., 2004; Thimm et al., 2004). In contrast, sucrose feeding of seedlings rapidly induces AtTPS5 to high levels (Schluepmann et al., 2004). External feeding of either trehalose or sucrose increases internal T6P levels (Schluepmann et al., 2004). This is probably due to the changes in expression of TPS and TPP enzymes and/or other mechanisms. In any case, T6P is well placed to be an ‘indicator’ of plant sugar status.

How might T6P exert its downstream regulatory effects? In Saccharomyces cerevisiae, T6P potently inhibits hexokinase. Without T6P, ATP is expended faster by hexokinase than it can be replenished by later steps of glycolysis (Blazquez et al., 1993; Thevelein and Hohmann, 1995). Therefore, the effect is critical for growth of yeast on glucose. On the other hand, plant hexokinase(s) appear to be insensitive to T6P (Eastmond et al., 2002; Wiese et al., 1999), suggesting other sites of T6P action. Trehalose feeding was found to induce expression of the ApL3 subunit of ADP-glucose pyrophosphorylase (AGPase) in Arabidopsis (Fritzius et al., 2001). In addition, AGPase was markedly activated by thioredoxin-mediated reduction when T6P levels were increased by gene expression, or by feeding sugars to plants or isolated chloroplasts (Kolbe et al., 2005). These results indicate that T6P is synthesized in the cytosol as a ‘reporter’ of cytosolic sugars, enters the chloroplast and activates AGPase to promote starch biosynthesis (Kolbe et al., 2005). In view of the wide-ranging physiological effects of T6P, it seems likely that other molecular targets for T6P exist (Fritzius et al., 2001; Müller et al., 2000).

The activation of AGPase in response to T6P in whole plants depends on the sucrose non-fermenting 1-related protein kinase (SnRK1) (Kolbe et al., 2005). Interestingly, gene microarray studies have shown that increased levels of T6P in transgenic and wild-type Arabidopsis plants correlate with increased expression of the SnRK1 family gene AtKIN11 (Schluepmann et al., 2004). Moreover, previous work has implicated SnRK1s in signal transduction affecting plant sugar utilization (Bhalerao et al., 1999). This is mediated by the diurnal regulation of cytosolic enzymes including nitrate reductase (NR) (Huber et al., 2002), sucrose-phosphate synthase (SPS) (Moorhead et al., 1999; Toroser et al., 1998) and 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (F2KP) (Kulma et al., 2004). F2KP is a bifunctional enzyme that both generates and hydrolyses another ‘sugar signal’, fructose-2,6-bisphosphate. NR and F2KP that have been phosphorylated by SnRK1 bind to 14-3-3 proteins. However, the calcium-dependent protein kinases (CDPKs) can also phosphorylate the 14-3-3-binding sites, and the physiological circumstances under which each kinase operates are not yet certain. Under dark conditions, NR activity in leaves is rapidly inhibited by phosphorylation and 14-3-3 binding. In transgenic plants carrying non-phosphorylatable NR, nitrite accumulates because the NR cannot be inhibited by 14-3-3s, and photosynthetic reductant is unavailable to drive the assimilation of nitrite into amino acids (Lea et al., 2004). Although the evidence that SnRK1 is the physiological dark-activated NR kinase is fairly circumstantial, mosses lacking their SnRK1-like enzymes require continuous light to survive. Such a finding is consistent with an essential role for these enzymes in the dark (Thelander et al., 2004).

Plant SnRK1 is related to the mammalian AMP-activated protein kinase (AMPK), and both kinases display similar specificity in phosphorylating peptide and protein substrates (Dale et al., 1995). Mammalian AMPK functions as a sensitive indicator of the cell's energy status. It is activated when ATP levels fall and AMP levels rise, by direct allosteric regulation by AMP and also ATP. Activation also involves phosphorylation of the AMPK by two upstream kinases, LKB1 and Ca2+/calmodulin-dependent kinase kinase (CaMKK) (Hawley et al., 2003, 2005). LKB1, in complex with its regulatory subunits STRAD and MO25, is essential for activation of the AMPK in cellular responses to exercise, metformin and phenformin (Hawley et al., 2003; Sakamoto et al., 2005). Metformin is an anti-diabetic biguanide drug that is thought to inhibit complex 1 in the mitochondrial respiratory chain, thereby decreasing cellular ATP, increasing AMP levels, and activating the AMPK. One effect of the AMPK is to promote uptake of glucose into tissues (for energy production), which reduces blood sugar levels. This effect is beneficial in patients with type II diabetes, whose cells do not absorb glucose properly in response to insulin. Phenformin has a similar action to metformin, except that it is too potent to use as a therapeutic drug. The AMPK is also activated in response to 2-deoxyglucose (Hardie, 2005). Substrates of AMPK include metabolic enzymes and effectors of gene expression, so the overall effect of AMPK phosphorylation is to activate ATP-producing catabolic pathways, while switching off ATP-consuming anabolic processes (Hardie, 2005).

Sugars, T6P, 14-3-3 proteins and the SnRK1 family of protein kinases are all implicated as components of regulatory networks that influence plant metabolism and development. For this reason, we were intrigued to identify TPS5, TPS6 and TPS7 amongst (phospho)proteins that were isolated from Arabidopsis cells by 14-3-3 affinity chromatography. Our findings indicate that members of the SnRK1 family phosphorylate the 14-3-3-binding site(s) that we have identified at the N-terminal tail of TPS5, both in vitro and in cells.


Phosphorylated TPS5 extracted from Arabidopsis cells binds directly to 14-3-3 isoforms

Arabidopsis cell extracts were passed though a 14-3-3 affinity column and specifically bound proteins were eluted with the synthetic 14-3-3-binding phosphopeptide, ARAApSAPA. The elution pool was highly enriched in proteins that bound directly to 14-3-3s in an overlay assay (Figure 1a). The ARAApSAPA elution pool was run on SDS–PAGE, and proteins identified by matrix-assisted laser desorption ionisation-time of flight (MALDI-TOF) tryptic mass fingerprinting included intact (120 kDa) and proteolysed forms of nitrate reductase (NR1 and NR2 isoforms; both known 14-3-3-binding phosphoproteins). Mixed with the proteolytic fragments of NR were peptides from TPS5, TPS6 and TPS7 (Table S1).

Figure 1.

 Phosphorylation-dependent binding of Arabidopsis TPS5 to 14-3-3s.
(a) Clarified Arabidopsis cell extracts were chromatographed on 14-3-3-Sepharose (see Experimental procedures). Extracts and column flow-throughs (40 μg of each), and ARAApSAPA elution pools (5 μg) from two separate preparations were run on NuPage 4–12% Bis-Tris SDS gels, transferred to nitrocellulose membranes and analysed with Sypro Ruby protein stain (top panel), DIG-14-3-3 overlay (second panel), and Western blotting with N-terminal-TPS5 antibody (third panel) and C-terminal-TPS5 antibody (bottom panel).
(b) Specific co-immunoprecipitation of 14-3-3s and TPS5 from extracts of Arabidopsis cells. Antiserum raised against spinach 14-3-3s (1 ml) was covalently coupled to protein G–Sepharose (approximately 2 ml). This matrix (60 μl) was used to immunoprecipitate 14-3-3s from 20 μg of Arabidopsis cell extract, which was pre-cleared with protein G–Sepharose alone. The indicated peptides were included in the immunoprecipitating suspension and in subsequent washes. The immunoprecipitate was extracted in SDS sample buffer without reducing agent, because, when reducing agent was included, antibody protein obscured the signals from TPS5 and 14-3-3s on the subsequent blot (not shown). The samples were run on SDS–PAGE, transferred to nitrocellulose and probed with anti-C-terminal-TPS5 (bottom panel) and anti-14-3-3 antiserum (top panel).
(c) Direct binding of 14-3-3s to phosphorylated TPS5. TPS5 was immunoprecipitated from Arabidopsis cell extract (1.6 mg) using 50 μl anti-TPS5 affinity-purified antibody (raised against GST–TPS5) that had been cross-linked to protein G–Sepharose. The purified protein was incubated for 30 min at 30°C in the presence of 50 mU ml−1 protein phosphatase  2A (PP2A), with or without its inhibitor microcystin-LR (3 μm), as indicated. The samples were run on SDS–PAGE, transferred to nitrocellulose and probed for binding to DIG-14-3-3s in an overlay assay, anti-pThr49-TPS5 antibodies, anti-N-terminal-TPS5 antibodies, and anti-GST–TPS5.

Antibodies were raised against synthetic peptides corresponding to sequences in the N-terminus (N-terminal-TPS5 antibody) and the C-terminus of AtTPS5 (C-terminal-TPS5 antibody). Western blotting with both antibodies confirmed that AtTPS5 was highly enriched in the phosphopeptide elution pool from the 14-3-3 column, but barely detected in the cell extract, flow-through, salt wash, or in samples that were mock-eluted with phosphopeptides that do not bind 14-3-3s (Figure 1a, and data not shown). Thus, AtTPS5 can bind directly or indirectly to 14-3-3 proteins.

When 14-3-3s were precipitated from Arabidopsis cell extracts using anti-14-3-3 antibodies, TPS5 was co-precipitated (Figure 1b). Peptides that do not bind to 14-3-3s had no effect on the TPS5/14-3-3 interaction. In contrast, the 14-3-3-binding phosphopeptide, ARAApSAPA, selectively blocked the co-precipitation of TPS5 with 14-3-3s (Figure 1b). These findings confirm that endogenous TPS5 binds specifically to the phosphopeptide binding site(s) on 14-3-3s in Arabidopsis cell extracts.

When TPS5 was immunoprecipitated from cell extracts with either the C-terminal-TPS5 antibody (not shown) or an antibody raised against GST–TPS5 (Figure 1c), the isolated TPS5 was able to bind directly to DIG-14-3-3s in an overlay assay (Figure 1c). Generally, 14-3-3s bind to phosphorylated serine or threonine residues in target proteins (MacKintosh, 2004). Dephosphorylation of the immunoprecipitated TPS5 with protein serine/threonine phosphatase  2A (PP2A) abolished binding of the immunoprecipitated TPS5 to DIG–14-3-3s (Figure 1c). In the control, TPS5 still bound to DIG–14-3-3s after incubation with a mixture of PP2A and its inhibitor microcystin-LR (Figure 1c), suggesting that phosphorylated site(s) on TPS5 bind directly to the phosphopeptide-binding site on 14-3-3s.

The Arabidopsis cell extracts contain at least nine 14-3-3 isoforms, of which several are co-immunoprecipitated with TPS5

The Arabidopsis genome encodes 15 14-3-3 genes, of which 12 have mRNA transcripts and the remaining three may be pseudogenes (Rosenquist et al., 2001; Sehnke et al., 2002). We determined which isoforms were present in the cells as a first step towards identifying the TPS5-interacting 14-3-3s. The isolated 14-3-3s ran as four major protein bands on one-dimensional SDS–PAGE that were resolved into four ‘lines’ of multiple spots on two-dimensional gel electrophoresis (Figure 2). The 14-3-3 isoforms upsilon and phi were present in the upper series of spots; the second series contained chi and nu; the third line of spots contained only omega; while the lowest series comprised lambda, psi, mu and epsilon isoforms. Thus, the Arabidopsis cells contained at least nine of the twelve known isoforms. In two-dimensional gels, the upsilon, lambda and mu isoforms were present as single spots, whereas the other isoforms were each represented by series of two to five spots that may have been generated by post-translational modifications (Figure 2). Although there were insufficient 14-3-3s in TPS5 precipitates and too much antibody light chain to identify the isoforms by mass spectral analysis, the TPS5-binding 14-3-3s ran as a broad band on SDS gels, indicative of heterogeneity (data not shown).

Figure 2.

 Isolation and identification of multiple 14-3-3 isoforms from Arabidopsis cell extracts.
Arabidopsis 14-3-3 proteins isolated by a series of column chromatography steps (following the procedure used for spinach 14-3-3s described by Moorhead et al., 1999) were resolved by two-dimensional electrophoresis with isoelectric focusing between pH 4.5 and 5.5 in the first dimension, and SDS–PAGE on 12–14% gels in the second. Proteins were stained with colloidal blue, excised, digested with trypsin, and identified by MALDI-TOF MS analysis. Peptides unique to different 14-3-3 isoforms were identified in the spots as indicated.

14-3-3s bind directly to GST–TPS5 that is phosphorylated by mammalian AMPK in vitro

We failed to identify 14-3-3-binding phosphorylated site(s) by mass spectral analysis of tryptic digests of TPS5 isolated from cell extracts by immunoprecipitation or 14-3-3 affinity chromatography. However, NR, which was more abundant in the pool of 14-3-3 –affinity-purified proteins, yielded two phosphopeptides. One was the known 14-3-3-binding site on NR2 [SV(pS534)TPPMNTTAK, where pS is phosphoserine]. The other was a novel doubly phosphorylated site in the N-terminal acidic tail of NR2 (54VHDDDEDV(pS62)(pS63)EDENETHNSNAVYK77), a region previously pinpointed as important for NR regulation. When this N-terminal acidic patch is deleted (ΔNR), the enzyme can still be phosphorylated at Ser534 and bind to 14-3-3s. However, unlike the wild-type NR, ΔNR was not inhibited by the bound 14-3-3s, and potato plants containing ΔNR have altered metabolism and increased biomass (Djennane et al., 2004).

We were, therefore, forced to take an in vitro approach to identifying the phosphorylated site(s) on TPS5. Therefore, we expressed GST–TPS5 in Escherichia coli. The protein isolated on glutathione–Sepharose comprised intact GST–TPS5 and E. coli GroEL, which could only be partially removed by incubating with Mg-ATP. The GST–TPS5 in this preparation could be phosphorylated by Ca2+-independent protein kinase activity in plant cell extracts (not shown) and the mammalian AMPK. The resulting phosphorylated GST–TPS5 bound to 14-3-3s (Figure 3a). When Mg[32P-γ]ATP was used as substrate, tryptic digests of the 32P-labelled GST–TPS5 contained two major phosphopeptides (P1 and P2) that were separated by HPLC (Figure 3b). Using MALDI-TOF mass spectrometry and solid-phase sequencing, P1 was identified as 47VA(pT49)VTGVLSELDDDNNSNSVCSDAPSSVTQDR78 (where pT represents phosphorylated Thr49), whereas P2 was 22(pS22)YSNLLDLASGNFHSFSR39 (where Ser22 is the phosphorylated residue) (Figure 3c,d).

Figure 3.

 Mapping of the sites on GST–TPS5 phosphorylated by mammalian AMPK in vitro.
(a) Bacterially expressed GST–TPS5 (10 μg) was incubated with Mg-ATP in the absence (0 U ml−1) and presence of AMPK (4 and 40 U ml−1) for 45 min at 30°C. Samples were analysed by probing with DIG–14-3-3s in an overlay assay.
(b) Bacterially expressed GST–TPS5 was phosphorylated in vitro with Mg[32P-γ]ATP and AMPK, resolved on SDS–PAGE, digested with trypsin and chromatographed on a Vydac 218TP5215 C18 column (Hichrom, Reading, UK) equilibrated in 0.1% v/v trifluoroacetic acid in water. The column was developed with an acetonitrile gradient (broken lines) and 32P-radioactivity was counted using an online Berthold LD509 detector (Berthold, Redborn, UK) (solid lines). Phosphopeptides P1 and P2 were eluted at 27.5% and 35.8% acetonitrile, respectively. The flow rate was 0.2 ml min−1 and 0.1 ml fractions were collected. Of the 60 000 cpm loaded onto the HPLC column, the 32P radioactivity recovered in each peak is indicated (Cerenkov counting).
(c, d) Identification of phosphopeptides P1 and P2. All residues were identified by a combination of MALDI-TOF mass spectrometry and solid-phase Edman sequencing in which a release of 32P radioactivity was detected at the cycle corresponding to the phosphorylated residue. The MALDI-TOF spectrum of peak P1 contained ions whose masses matched both the mono-phosphorylated and unphosphorylated forms of peptide 47VATVTGVLSELDDDNNSNSVCSDAPSSVTQDR(I)79, corresponding to residues 47 to 79 in TPS5 (not shown). Residue 46 is arginine, consistent with the specificity of trypsin. When subject to solid-phase Edman degradation, Cerenkov counting of the released ATZ-amino acids showed a burst of counts at cycle 3, which shows that the residue corresponding to Thr49 was phosphorylated (c, left panel). Similarly, P2 was identified as 22(pS)YSNLLDLASGNFHSFSR(E)40, where Ser22 was phosphorylated. Again the preceding residue 21 is arginine. Upon solid-phase Edman sequencing, the major release of radioactivity occurred at cycle 1, consistent with phosphorylation of Ser22 of TPS5 (c, right panel). After Edman degradation, of the 2328 cpm that was coupled to the membrane, 70 cpm remained after Edman sequencing of P1, while 48 cpm of the original 1616 cpm remained for P2. Amino acid residues are numbered as in the native TPS5 protein (accession number H71447), not counting the GST.

Phospho-specific antibodies that recognize pSer22 and pThr49 of TPS5

Antibodies were raised that recognized synthetic peptides containing pSer22 and pThr49 of TPS5 only when these sites are phosphorylated (Figure 4a). When tested on TPS5 isolated from Arabidopsis cell extracts by immunoprecipitation (Figure 1c) or 14-3-3 affinity chromatography (Figure 4b), the pThr49-TPS5 antibody gave a strong signal. The pSer22-TPS5 antibody also gave a signal, albeit much weaker (Figure 4b). Consistent with phospho-specificity, the pSer22-TPS5 and pThr49-TPS5 signals were abolished by pre-incubation with the cognate phosphopeptides but not the unphosphorylated peptides (Figure 4b).

Figure 4.

 Use of anti-pSer22-TPS5, anti-pThr49-TPS5 and anti-N-terminal-TPS5 antibodies to analyse the phosphorylation of TPS5 14-3-3-affinity-purified from Arabidopsis cell extracts.
(a) Characterization of phospho-specific antibodies that recognize the pSer22 and pThr49 sites on TPS5. The indicated amounts (from 1 to 100 ng) of phosphopeptides and unphosphorylated peptides were conjugated to ovalbumin and spotted on to ECL-Hybond nitrocellulose membranes. The peptide conjugates were immunoblotted with 10 μg ml−1 anti-pSer22-TPS5 (upper blot), 2 μg ml−1 anti-pThr49-TPS5 (middle blot), and 2 μg ml−1 N-terminal-TPS5 antibody (lowest blot). The weak reactivity of the anti-pThr49 antibody towards the unphosphorylated Thr47-TPS5 peptide could be selectivity eliminated by pre-incubation of the antibody with that peptide (b, and data not shown). The sequences of these peptides are given under Experimental procedures. The prefix p denotes a phosphorylated peptide.
(b) (Phospho)proteins were isolated from Arabidopsis cell extracts (2 g protein) by 14-3-3 affinity chromatography. The ARAApSAPA elution pool was run on SDS–PAGE, transferred to nitrocellulose and analysed for binding to the pSer22-TPS5 and pThr49-TPS5 antibodies, in the presence or absence of the indicated competitor peptides. The identity of the 67 kDa phosphoprotein that is recognized by the pSer22-TPS5 antibody is unknown.

The N-terminal-TPS5 antibody precipitated TPS5 from cell extracts, although 14-3-3s were not co-precipitated with this antibody (not shown). The anti-N-terminal-TPS5 was raised against a non-phosphorylated peptide encompassing Thr49 (Figure 1a), and blots showed that this antibody binds to unphosphorylated Thr49 peptide, but barely gives a signal with the pThr49-phosphorylated counterpart (Figure 4a, bottom panel). Similarly, when TPS5 was immunoprecipitated from cell extracts, its recognition by the anti-N-terminal antibody was markedly enhanced when the TPS5 was dephosphorylated (Figure 1c). Thus, 14-3-3s and the N-terminal-TPS5 antibody have mutually exclusive specificities, with 14-3-3s binding to pThr49 and the N-terminal-TPS5 antibody binding to unphosphorylated Thr49 (Figure 1c). This gave further confirmation that pThr49 of TPS5 binds to 14-3-3s.

Heterologously expressed TPS5 is phosphorylated on Thr49 when the AMPK is activated by phenformin or 2-deoxyglucose in human HEK293 cells

We tested whether TPS5 could act as an AMPK substrate inside human cells, to provide a frame of reference for studying the regulation of TPS5 phosphorylation in plant cells. Human HEK293 cells were transiently transfected with either GST–TPS5 or GST alone, and stimulated with 2-deoxyglucose or phenformin to activate AMPK via LKB1 (Hardie, 2005; Sakamoto et al., 2005). As expected, 2-deoxyglucose promoted an increase in phosphorylation of Thr172 in the activation loop of the kinase, accompanied by AMPK activation and phosphorylation of acetyl CoA carboxylase (Figure 5a). Similarly, phenformin-activated AMPK and promoted phosphorylation of acetyl CoA carboxylase (Figure 5b). In the cells transfected with GST–TPS5, 2-deoxyglucose and phenformin also stimulated phosphorylation of Thr49 on the heterologously expressed plant protein (Figure 5a,b).

Figure 5.

 Effect of phenformin and 2-deoxyglucose on Thr49 phosphorylation of GST–TPS5 expressed in human HEK293 cells and endogenous TPS5 in Arabidopsis cells.
(a) Untransfected human HEK293 cells, and HEK293 cells expressing GST or GST–TPS5 were incubated with 2-deoxyglucose (20 mm) for 30 min. Lysates were prepared and run on SDS–PAGE and proteins transferred to blots that were probed using anti-pThrAMPK, anti-phosphoacetyl-CoA carboxylase (pACC), anti-GST, and anti-pThr49-TPS5 antibodies, as indicated.
(b) Untransfected HEK293 cells, and HEK293 cells expressing GST or GST–TPS5 were incubated with phenformin (2 mm) for 1 h. Lysates were prepared and AMPK activity (expressed as nmol min−1 mg−1 total lysate protein) determined in immunoprecipitates prepared using a mixture of anti-AMPKalpha1 and anti-AMPKalpha2 antibodies (top panel). Lysates were also run on SDS–PAGE and proteins transferred to blots that were probed using anti-AMPK alpha1/alpha2, anti-phosphoacetyl  CoA carboxylase (pACC), anti-GST, anti-TPS5-N-terminal peptide and anti-pThr49 TPS5 antibodies, as indicated.
(c) Arabidopsis cells (40 ml of a 5-day-old culture) were collected by centrifugation (65 g) and transferred to 40 ml of fresh medium containing glucose (100 mm), 2-deoxyglucose (100 mm), 3-O-methylglucose (100 mm) or sucrose (88 mm) as indicated, and harvested after 1 h. TPS5 was immunoprecipitated from clarified extracts using the antibody raised against GST–TPS5. Immunoprecipitates were analysed by Western blotting with anti-pThr49-TPS5 antibodies and anti-TPS5 antibodies. Two representative pThr49-TPS5 blots and an anti-TPS5 blot from one of the experiments are shown.

Thr49 of TPS5 is phosphorylated in response to 2-deoxyglucose in Arabidopsis plant cells

Arabidopsis cells were stimulated in different ways in order to identify signalling pathway(s) that regulate phosphorylation of the endogenous TPS5. Amongst the stimuli tested, 2-deoxyglucose markedly enhanced the phosphorylation of Thr49 (Figure 5c). However, in contrast to its effect in human cells (Figure 5b), phenformin had no obvious effect on Thr49 phosphorylation of TPS5 in Arabidopsis cells at concentrations up to 16 mm for 60 min (data not shown). After 16 h exposure to 20 mm phenformin, this compound had a marked toxic effect on the cells, which became chlorotic.

Mutation of either Ser22 or Thr49 to alanine prevents binding of GST–TPS5 to 14-3-3s

In contrast to wild-type GST–TPS5, proteins with either a Ser22Ala or Thr49Ala mutation, and the double Ser22Ala/Thr49Ala mutant, did not bind to 14-3-3s when phosphorylated by the AMPK either in vitro or in vivo (Figure 6, and data not shown). These experiments also revealed a new regulatory interaction between the two sites. In human cells, compared with the wild-type protein, the Ser22Ala-TPS5 was consistently more highly phosphorylated on Thr49, even in the absence of stimuli that activate the AMPK (Figure 6). Either the alanine in position 22 makes the Thr49 a substrate for the AMPK or another kinase, or the Ser22Ala mutant is defective in its ability to be dephosphorylated at pThr49. The signals with the pSer22-TPS5 antibody were too weak to be useful (data not shown). These results indicate that both Ser22 and Thr49 must be phosphorylated for 14-3-3s to bind to TPS5.

Figure 6.

 Mutation of either Ser22 or Thr49 prevents binding of TPS5 to 14-3-3s.
Wild-type GST–TPS5, GST–Ser22Ala-TPS5, GST–Thr49Ala-TPS5 and the GST–Ser22Ala/Thr49Ala-TPS5 double mutant proteins were expressed in a suspension culture of human HEK293 cells. Proteins were purified by glutathione–Sepharose chromatography, and analysed before (S) and after in vitro phosphorylation with AMPK (S + K = 40 U ml−1 AMPK; S + K2 = 4 U ml−1 AMPK). Blots were analysed for protein (using Ponceau  S stain), DIG–14-3-3 overlay, pThr49-TPS5 antibody and anti-GST–TPS5 antibody. This experiment used a cell suspension that can be grown to high density for isolating relatively large amounts of protein. It is commonly found that the AMPK is activated while harvesting these cells by centrifugation (presumably because of transient anoxia), which explains the basal pThr49 phosphorylation of the extracted wild-type TPS5 (WT, lane S). In addition, the ‘basal’ Thr49 phosphorylation was consistently higher in the Ser22Ala mutant than in wild-type TPS5. In contrast, Figure 5 used cell monolayers adhered to dishes, which can be lysed rapidly without activating the AMPK in the unstimulated controls.

2-deoxyglucose activates SnRK1-like kinase(s) in Arabidopsis cells, and mammalian LKB1 activates plant SnRK1 in vitro

When Arabidopsis cells were transferred to a medium containing 2-deoxyglucose, there was a rapid activation of Ca2+-independent kinase activity that could be detected in ATP-agarose eluates prepared from cell extracts (Figure 7a). The 2-deoxyglucose-activated enzyme was found to have the properties of the SnRK1 family of protein kinases, being Ca2+-independent, using the AMARA peptide as substrate, and being inactivated by dephosphorylation with protein phosphatases 2A (Figure 7a) or 2C (not shown). The AMARA kinase that was isolated by ATP-agarose and dephosphorylated could be reactivated by incubation with Mg-ATP (Figure 7c). However, after further purification by anion-exchange chromatography and then inactivation with PP2A, the AMARA kinase could not be reactivated by incubation with Mg-ATP alone (Figure 7c, and data not shown). This means that the AMARA kinase was reactivated by an endogenous kinase(s) that was present in the ATP-agarose elution pool, but had been separated from the AMARA kinase on the anion-exchange column. The more-purified preparation of AMARA kinase was reactivated by incubation with mammalian LKB1 complexes and Mg-ATP. However, it could not be reactivated with kinase-dead LKB1 in which the catalytic Asp194 residue was mutated to Ala (Figure 7c) (Lizcano et al., 2004). The AMARA kinase and pThr49-TPS5 kinase activities co-purified through the ATP-agarose and anion-exchange chromatography steps (Figure 7d, and data not shown).

Figure 7.

 Changes in SnRK1-like activity in response to 2-deoxyglucose in plant cells and LKB1 in vitro.
(a) Arabidopsis cells (40 ml of a 5-day-old culture) were collected by centrifugation (65 g) and transferred to 12 ml of fresh medium containing glucose (100 mm), 2-deoxyglucose (100 mm), 3-O-methylglucose (100 mm) or sucrose (88 mm) as indicated, and harvested after 30 min. ATP-agarose eluates were prepared from the cell extracts and assayed for kinase activity using the AMARA peptide as substrate (see Experimental procedures). AMARA kinase activities are expressed as mean (± standard deviation) for three separate extractions.
(b) ATP-agarose eluates from cells that had been transferred into 2-deoxyglucose were incubated with 50 mU ml−1 PP2A for 10 min at 30°C in the presence or absence of 1 μm microcystin inhibitor (MC), and reactions were stopped by adding 5 μm microcystin. The AMARA kinase activity of 15 μl aliquots was assayed.
(c) Dephosphorylated ATP-agarose eluates (prepared as for b) and dephosphorylated anion-exchange fraction 26 (from d) were incubated in the presence of Mg-ATP, LKB1/STRAD/MO25 or kinase-dead Asp194Ala-LKB1/STRAD/MO25 complexes, which were prepared as described by Lizcano et al. (2004).
(d) Six ATP-agarose eluates, prepared from 2-deoxyglucose-treated cells, were combined and chromatographed by anion exchange on a Mono Q column (see Experimental procedures). Fractions (100 μl) were assayed in duplicate for AMARA kinase activity and used to phosphorylate GST–TPS5, which was run on SDS–PAGE, transferred to nitrocellulose, and phosphorylation analysed by Western blotting with the pThr49-TPS5 antibodies. AMPK was used as the positive kinase control.


In this paper, we found that TPS5, TPS6 and TPS7 were isolated by 14-3-3 affinity chromatography of Arabidopsis cell extracts. This led us to identify Ser22 and Thr49 on TPS5 as the residues whose phosphorylation is essential for binding to 14-3-3s in vitro and in vivo. The simplest hypothesis to explain the data is that phophoSer22 binds to one side of the binding groove on a 14-3-3 dimer, while phosphoThr49 docks into the other side of the same dimer. One or both of these sites are conserved in TPS6, TPS7, TPS9 and TPS10, suggesting that regulation by phosphorylation and 14-3-3s may extend to these isoforms (Figure S1a).

The anti-pThr49-TPS5 antibody gave a strong signal that facilitated a screen for physiological stimuli that regulate the phosphorylation status of this site. In human cells, Thr49 on TPS5 was phosphorylated in response to 2-deoxyglucose and phenformin, two stimuli that activate AMPK (Figure 5a,b). There is little information about the acute mechanisms that regulate the AMPK-related plant SnRKs. However, we found that 2-deoxyglucose induced an increase in Thr49-TPS5 phosphorylation (Figure 5c), and a parallel increase in the activity of a plant AMPK-like activity (Figure 7a). Plant SnRKs comprise three subfamilies, SnRK1, SnRK2 and SnRK3 (Hrabak et al., 2003), of which the SnRK1s appear to be functional orthologues of yeast Snf1 and mammalian AMPKs (Boudsocq et al., 2004; Halford et al., 2004). Thus, it seems likely that the plant 2-deoxyglucose-activated TPS5 kinase detected in this study is an SnRK1, although its precise assignment will require immunoprecipitating antibodies to be raised for each SnRK isoform, and/or comparison of TPS5 phosphorylation in wild-type and SnRK1 knock-out plants (Halford et al., 2004).

We are particularly keen to determine whether TPS5 becomes phosphorylated in leaves during a light to dark transition. It has been shown that NR and F2KP bind to 14-3-3s after phosphorylation by SnRK1-like kinases (Huber et al., 2002; Kulma et al., 2004). Moreover, it is well established that NR is rapidly inhibited by phosphorylation and binding to 14-3-3s in the dark (Huber et al., 2002). Previously, we found that 2-deoxyglucose inhibits NR by inducing its phosphorylation and binding to 14-3-3s, and causes a fall in cellular levels of fructose-2,6-bisphosphate (Kulma et al., 2004). In this study, we have shown that, under the same conditions, SnRK1-like activity is stimulated, Thr49 of TPS5 becomes phosphorylated and TPS5 binds to 14-3-3s. Thus, we hypothesize that a common kinase (SnRK1?) phosphorylates the 14-3-3-binding sites on all three enzymes, NR, F2KP and TPS5, in response to 2-deoxyglucose, and under dark conditions in leaves. For these studies, we are working up methods to extract TPS5 quickly from leaves in sufficient quantity for analysis by the pThr49-TPS5 antibody and 14-3-3-binding assays. Extractions have to be performed rapidly because the TPS5 kinase appears to be activated during the harvesting and extraction of plant cells and tissues [see basal SnRK1activity (Figure 7) and basal TPS5 phosphorylation (Figure 5c) under ‘control’ conditions; and data not shown], as has been found for mammalian AMPK and the yeast Snf1 (Halford et al., 2004; Hardie, 2005; also Figure 6 legend).

We do not yet know how 14-3-3s affect TPS5 and the other class II enzymes. Although they have both TPS- and TPP-like domains, enzymatic activity has not yet been demonstrated for these proteins (Vogel et al., 2001; data not shown). We have not detected TPS or TPP activity from either endogenous TPS5 isolated from Arabidopsis cells or bacterially expressed GST–TPS5 under a variety of conditions under which TPS and TPP enzymes from S. cerevisiae were active (not shown). However, we did observe quenching of tryptophan fluorescence of TPS5, and binding signals using isothermal calorimetry, when UDP-glucose concentration was increased, indicating that the protein can bind to UDP-glucose. These findings are consistent with the fact that the UDP-glucose binding residues in E. coli OtsA are conserved in TPS5 (Gibson et al., 2002, 2004; Figure S1b). The suspected glucose-6-phosphate binding region, close to the N-terminal tail where the 14-3-3s bind, is less well conserved (Figure S1b). Thus, we do not know whether or how TPS5 influences intracellular T6P levels. However, it is tempting to suggest that SnRK1 phosphorylation of TPS5 might comprise part of the same signal transduction pathway by which thioredoxin-mediated activation of AGPase, and hence starch biosynthesis, is triggered by T6P (Kolbe et al., 2005). Such a response was stimulated by the presence of 2-deoxyglucose (Tiessen et al., 2003) and was dependent on SnRK1 (Kolbe et al., 2005). Indeed, the evidence pointing to some physiological connection between the SnRK1 and trehalose-based signalling has already prompted speculation that members of the TPS or TPP families might be direct substrates of SnRK1s (Glinski and Weckwerth, 2005; Halford et al., 2003).

Early studies showed that, like its mammalian counterpart, the activity of the plant SnRK1 is controlled by upstream kinase kinase(s) (MacKintosh et al., 1992). In this study, we extend these observations, by showing that the phosphorylation and activation state of the plant kinase(s) are increased by 2-deoxyglucose, and, further, that mammalian LKB1 can also phosphorylate and activate the plant enzyme(s). Looking upstream for potential plant SnRK1 kinases, the genes At5g60550 and At3g45240 encode the Arabidopsis kinases that are most similar to LKB1. There are also plant genes similar to STRAD and MO25, which are two regulatory subunits of the mammalian LKB1 (Hawley et al., 2003). People with mutated LKB1 can get cancers (Hardie 2005). Furthermore, the LKB1/AMPK pathway is activated by the anti-diabetic drug metformin, and more potently activated by phenformin (Hardie, 2005; Sakamoto et al., 2004). Thus, it will be fascinating to elucidate the upstream regulators of plant SnRK1 to give new insights relevant to plant physiology, and perhaps find clues that might impinge on cancer and diabetes. For example, we found that 2-deoxyglucose activates plant SnRK1, but have not yet observed any effect of phenformin on the plant enzyme. Does the lack of response of the plant kinase to phenformin reflect differences in the physiological settings within plant and human cells, or is it due to specific differences between the plant and human kinases themselves? Studies of how the plant LKB1-like enzymes perform within LKB1-deficient human cells, as well as in plants, should help to distinguish the common mechanisms of LKB1/AMPK signalling in all eukaryotes from those that are specialized to one system.

Experimental procedures


Reagents for plant cell culture were obtained from Sigma (Poole, UK) and for human cells from Invitrogen (Paisley, UK). Complete protease inhibitor mixture and protein sequencing grade trypsin were acquired from Roche Molecular Biochemicals (Roche Diagnostics Ltd, Lewes, East Sussex, UK). Enhanced chemiluminescence (ECL) reagent was obtained from Amersham Pharmacia Biotech (Little Chalfont, UK). The ATP-agarose, comprising ATP linked to agarose via its γ-phosphate, was KinaseBindTM with high ligand density (8–12 mmol ml−1) acquired from Innova Biosciences (Cambridge, UK).

Cell culture

Arabidopsis cell suspensions (cv. Erecta) were originally derived by May and Leaver (1993) and cultured in Murashige and Skoog medium with minimal organics (MSMO), α-naphthalene acetic acid (0.5 mg l−1; auxin), kinetin (0.05 mg l−1; cytokinin) and 3% w/v sucrose (pH 5.7) in 500 ml conical flasks under continuous light (approximately 20 μE m−2 sec−1) at 20°C, with rotary shaking at 150 rev min−1. Cells were harvested rapidly by vacuum filtration onto Miracloth (Calbiochem, Nottingham, UK), immediately frozen in pre-weighed tubes and stored at −80°C. Lysates were prepared as described by Kulma et al., 2004.

Human embryonic kidney 293 (HEK293) cells were cultured in the presence of 10% v/v fetal calf serum in Dulbecco's modified Eagle's medium under standard conditions. Cells cultured on 10 cm diameter dishes in medium containing 10% v/v serum were transfected and stimulated as indicated. Cells in 10 cm diameter dishes were lysed by scraping in 1 ml of ice-cold lysis buffer comprising 50 mm Tris (pH 7.5), 1 mm EDTA, 1 mm EGTA, 1% v/v Triton X-100, 10 mmβ-glycerophosphate, 50 mm sodium fluoride, 1 mm sodium orthovanadate, 5 mm sodium pyrophosphate, 0.27 m sucrose, 1 mm benzamidine, 0.2 mm PMSF, 10 μg ml−1 leupeptin, 1 μm microcystin-LR, and 0.1% (by volume) 2-mercaptoethanol. Lysates were centrifuged at 13 000 g for 10 min at 4°C. The supernatants were frozen at −80°C until use. Protein concentrations were determined by the Bradford method using bovine serum albumin as standard. HEK293 cell suspensions were grown in conical flasks with shaking.


Sheep anti-AtTPS5 antibodies were raised at Diagnostics Scotland (Penicuik, UK) against the synthetic peptides 42KRFPRVATVTGVLS55C (near the N-terminus of AtTPS5 + Cys for coupling; termed N-terminal-TPS5 antibody) and 821KPSKAKYYLDDTAEI835-Cys (near the C-terminus of TPS5 + Cys; termed C-terminal-TPS5 antibody) and affinity-purified in columns of the immobilized peptides. A third anti-TPS5 antiserum was raised by injecting sheep with bacterially expressed GST–TPS5. Phospho-specific antibody recognizing TPS5 phosphorylated at Ser22 (anti-pSer22-TPS5) was raised in sheep against the peptide 16RDMVSRpS22YSNLLD28, corresponding to residues 16–28 of Arabidopsis TPS5. The antibodies were affinity-purified on CH–Sepharose® covalently coupled to the phosphorylated peptide, then passed through a column of CH–Sepharose coupled to the non-phosphorylated peptide. Similarly, the anti-pThr49-TPS5 antibody was generated in sheep against the phosphopeptide 43RFPRVApT49VTGVLS55. Antibodies that recognize several, or possibly all, plant 14-3-3 isoforms were raised against 14-3-3s purified from spinach leaves (Moorhead et al., 1999). The specific AMPKalpha1 and AMPKalpha2 antibodies and phospho-specific antibodies recognizing AMPK phosphorylated on the T-loop have been described previously (Hawley et al., 2003).

Western blotting and DIG-14-3-3 overlays

For blots of total cell lysates, 40 μg protein was run on SDS–PAGE on NuPage 4–12% Bis-Tris gels (Invitrogen), and transferred to ECL Hybond nitrocellulose membranes (GE Healthcare Biosciences, Amersham, UK), which were blocked in 5% w/v fat-free milk (Marvel, Premium Brands UK, Spalding, UK) in 25 mm Tris–HCl pH 7.5, 0.5 m NaCl, and then immunoblotted at 4°C for 16 h using the indicated antibodies. Detection was performed using horseradish peroxidase-conjugated secondary antibodies (Promega, Southampton, UK) and the enhanced chemiluminescence reagent (ECL®; Amersham Biosciences). The DIG-14-3-3 overlay procedure has been described by Kulma et al. (2004).

14-3-3 affinity chromatography

The procedure described by Kulma et al. (2004) was followed. Briefly, clarified cell lysates (2.0–2.5 g protein) were mixed with 2 ml Sepharose linked to BMH1 and BMH2 (the S. cerevisiae 14-3-3 isoforms at 2 mg ml−1 of gel), washed extensively and mock-eluted with a synthetic peptide that does not bind 14-3-3s (1 mm ARAASAPA) before proteins that bind to the phosphopeptide binding site of 14-3-3s were eluted with 1 mm ARAApSAPA phosphopeptide.

Two-dimensional electrophoresis

The Amersham Pharmacia Biotech IPGphor system comprising immobilized Immoboline pH 4.5–5.5 gradients for isoelectic focusing and 12–14% SDS gels was used, according to the manufacturer's instructions.

Protein identification and phosphorylation site analysis

Proteins were identified from ‘in-gel’ tryptic digestion by MALDI-TOF MS of (Invitrogen) SDS–PAGE gel bands stained with colloidal blue, and 32P-labelled phosphopeptides were separated by reverse-phase HPLC and analysed by a combination of mass spectrometry and solid-phase Edman degradation as described previously (Campbell and Morrice, 2002).

Generation of GST fusions of AtTPS5 for expression in bacterial and human cells

The TPS5 ORF was amplified from Kazusa clone 02g04 with oligonucleotides MP724 (GCGGATCCATTAGTGATTATCTTCGACATTGTATCTACTTCACTTGCTGTAGAGATATGG) and MP725 (GCGAATTCTTAAAACAGATCTTTAGTTGGAACAGTGGCGG) using the GC-rich PCR system (Roche). The resulting product was cloned into pCR2.1TOPO (Invitrogen) and sequenced to completion. The clone was then either digested with BamHI and SalI and cloned into the same sites in pGEX6P-1 (for expression in E. coli BL21 (DE3) pLysS), or with BamHI and cloned into the BamHI site in EBG6P (for expression in human cells) (see Figure S1 for a note about the N-terminal sequence of TPS5). Thr49Ala and Ser22Ala mutations were produced using the Quickchange site-directed mutagenesis kit (Stratagene, Amsterdam, The Netherlands) with oligonucleotides MP1774/MP1775 and MP1776/MP1777 respectively:MP1774 GATATGGTATCAAGAGCTTATTCAAACCTCTTG; MP1775 CAAGAGGTTTGAATAAGCTCTTGATACCATATC; MP1776 GGTTTCCAAGAGTAGCAGCTGTCACTGGTGTCTTATC; MP1777 GATAAGACACCAGTGACAGCTGCTACTCTTGGAAACC.

Extraction and assay of human AMPK and plant SnRK1

AMPK was assayed after immunoprecipitation from human cell extracts as follows. Cell protein (300 μg) was incubated at 4°C for 1 h on a shaking platform with 5 μg of anti-alpha1/alpha2-AMPK antibody, which had been previously conjugated to 5 μl of protein G–Sepharose. The immunoprecipitates were washed twice with human cell lysis buffer containing 1 m NaCl and twice with 50 mm HEPES–NaOH pH 7.0, 1 mm DTT and 0.02% v/v Brij-35. Kinase activity was assayed by measuring the incorporation of 32P from [γ-32P]ATP into the AMARA peptide in a total assay volume of 25 μl in 50 mm Tris–HCl pH 7.0, 0.02% Brij-35, 1 mm dithiothreitol, 5 mm MgCl2, 200 μm AMP, 200 μm [γ-32P]ATP (200–500 cpm pmol−1) and 10  μg AMARA synthetic peptide substrate (Hawley et al., 2003). Reactions were initiated by adding [γ-32P]ATP-Mg. After 10 min at 30°C, 15 μl aliquots were spotted onto Whatman P81 phosphocellulose papers, which were washed three times in 75 mm phosphoric acid to remove residual [γ-32P]ATP, immersed in acetone, dried and the radioactivity measured by Cerenkov counting.

For plant AMARA kinase assays, Arabidopsis cells were harvested and extracts prepared rapidly in ice-cold lysis buffer. Where indicated, extracts were shaken with 200 μl ATP-agarose in the presence of 20 mm MgCl2, which was washed with 5 ml ice-cold plant cell lysis buffer (Kulma et al., 2004) in Spin-X columns (Costar, Corning, NY, USA). Buffer was drained and proteins were eluted by shaking in 200 μl AMARA kinase assay buffer containing 1 mm ATP, but omitting protein phosphatase inhibitors. The AMARA kinase in 10 μl aliquots of ATP-agarose eluates was measured in the presence of [γ-32P]ATP-Mg, with a final specific radioactivity of at least 3000 cpm nmol−1. Reactions were initiated by adding AMARA peptide substrate in 50 μl assays, from which 40 μl aliquots were spotted onto phosphocellulose. Where indicated (Figure 7) six ATP-agarose eluates from 2-deoxyglucose-treated cells were combined and run on an anion-exchange Mono Q 1.6/5 column (Amersham Biosciences) in 25 mm HEPES/KOH containing 1 mm dithiothreitol (where fractions 1–15 were the flow-through and wash), and developed with a 1.5 ml gradient from 0 to 0.5 m NaCl (fractions 16–31). The flow rate was 0.1 ml min−1 and fractions of 0.1 ml were collected.


We thank the UK BBSRC and MRC for supporting this research, the Wellcome Trust for a studentship to B.H.C.W., and David B. Collinge of the Royal Veterinary and Agricultural University, Copenhagen, Denmark, for sponsoring a visit of J.B. to Dundee. We thank the pharmaceutical companies supporting the Division of Signal Transduction Therapy (DSTT) at Dundee (AstraZeneca, Boehringer-Ingelheim, GlaxoSmithKline, Merck, Merck KgaA and Pfizer) for financial support, and the DSTT antibody production team coordinated by Dr James Hastie for affinity purification of antibodies, Leanne Brown and Colin Bell for tissue culture support, the Sequencing Service (University of Dundee; for DNA sequencing, Dr Simon Hawley, University of Dundee, for performing the AMPK activity assays in Figure 5b, Dr Gursant Singh Kular and Professor Dario Alessi, MRC Unit, University of Dundee, for active and kinase-dead forms of LKB1, and Dr John Lunn, Max Planck Institute for Molecular Plant Physiology, Golm, Germany, for helpful discussion about the N-terminal sequence of TPS5.