SEARCH

SEARCH BY CITATION

Keywords:

  • CRMP;
  • GSK3;
  • knockout;
  • neuron;
  • splice variant;
  • tau

Abstract

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

J. Neurochem. (2010) 115, 974–983.

Abstract

Mammalian glycogen synthase kinase-3 (GSK3) is generated from two genes, GSK3α and GSK3β, while a splice variant of GSK3β (GSK3β2), containing a 13 amino acid insert, is enriched in neurons. GSK3α and GSK3β deletions generate distinct phenotypes. Here, we show that phosphorylation of CRMP2, CRMP4, β-catenin, c-Myc, c-Jun and some residues on tau associated with Alzheimer’s disease, is altered in cortical tissue lacking both isoforms of GSK3. This confirms that they are physiological targets for GSK3. However, deletion of each GSK3 isoform produces distinct substrate phosphorylation, indicating that each has a different spectrum of substrates (e.g. phosphorylation of Thr509, Thr514 and Ser518 of CRMP is not detectable in cortex lacking GSK3β, yet normal in cortex lacking GSK3α). Furthermore, the neuron-enriched GSK3β2 variant phosphorylates phospho-glycogen synthase 2 peptide, CRMP2 (Thr509/514), CRMP4 (Thr509), Inhibitor-2 (Thr72) and tau (Ser396), at a lower rate than GSK3β1. In contrast phosphorylation of c-Myc and c-Jun is equivalent for each GSK3β isoform, providing evidence that differential substrate phosphorylation is achieved through alterations in expression and splicing of the GSK3 gene. Finally, each GSK3β splice variant is phosphorylated to a similar extent at the regulatory sites, Ser9 and Tyr216, and exhibit identical sensitivities to the ATP competitive inhibitor CT99021, suggesting upstream regulation and ATP binding properties of GSK3β1 and GSK3β2 are similar.

Abbreviations used:
CK2

casein kinase-2

CRMP

collapsin response mediator protein

DYRK2

dual-specificity tyrosine-(Y)-phosphorylation regulated kinase

GSK3

glycogen synthase kinase-3

GST

glutathione-S-transferase

Inh-2

inhibitor-2

KO

knockout

pGS2

phospho-glycogen synthase 2

SDS–PAGE

sodium dodecyl sulfate–polyacrylamide gel electrophoresis

Glycogen synthase kinase-3 (GSK3) was originally identified by its ability to phosphorylate glycogen synthase. Since then more than fifty protein substrates have been proposed for this enzyme, many of which are implicated in the action of growth factor receptor tyrosine kinases, Wnts, Semaphorins and sonic hedgehog. Mammalian GSK3 is generated from two genes, GSK3α and GSK3β, while a splice variant of GSK3β (GSK3β2) is enriched in neurons. In resting cells GSK3 activity is relatively high, and in response to a variety of stimuli this activity is normally reduced (for review see Frame and Cohen 2001). Regulation of activity occurs by different mechanisms, including (i) phosphorylation at an N-terminal serine (Sutherland et al. 1993b), (ii) through disruption of the axin-β-catenin multiprotein complex (Cadigan and Liu 2006), (iii) through phosphorylation of a tyrosine residue (Lochhead et al. 2006) and (iv) in the case of GSK3β, through phosphorylation of a C-terminal serine residue (Thornton et al. 2008). The other unusual property of GSK3 is that most of its substrates require prior phosphorylation (priming) at a residue 4 or 5 amino acids C-terminal to the target residue (Frame and Cohen 2001). That said, there are examples of ‘unprimed’ substrates reported.

Deletion of the GSK3β gene in mice is lethal (Hoeflich et al. 2000; Liu et al. 2007), and while GSK3β heterozygous (+/−) mice are viable, they exhibit a wide range of neurological abnormalities, including reduced aggression, increased anxiety, reduced exploratory activity, poor memory consolidation and reduced responsiveness to amphetamine (O’Brien et al. 2004; Beaulieu et al. 2008; Kimura et al. 2008). Conversely, over-expression of GSK3β results in hyperactivity and mania (Prickaerts et al. 2006). Mice lacking GSK3α are viable and relatively normal (MacAulay et al. 2007), exhibiting some improvement in whole body insulin sensitivity and glucose tolerance. Interestingly, mice lacking GSK3α specifically in neurons display reduced exploratory activity and aggression, similar to the GSK3β heterozygotes, but in addition have decreased locomotion, impaired co-ordination and a deficit in fear conditioning (Kaidanovich-Beilin et al. 2009). These distinct phenotypes suggest non-redundant functions of the GSK3 genes in the brain, while the overlapping behavioural problems between GSK3α neuronal knockout (KO) and GSK3β (+/−) mice suggest some common substrates.

Loss of both GSK3 isoforms specifically in the brain results in increased self-renewal of neuronal progenitor cells, but reduced neurogenesis (Kim et al. 2009), while GSK3α/β double knockin mice (both isoforms replaced by mutant proteins with Ser to Ala alterations at Ser21 and Ser9 respectively, preventing repression by growth factor signalling) show impairment of neuronal precursor cell proliferation (Eom and Jope 2009). Taken together this indicates that proper regulation of expression and activity of GSK3 is required for maturation of these cells during mammalian brain development.

The expression of GSK3 in neurons is further complicated by alternative splicing of GSK3β between exon 8 and 9, giving rise to two main variants (Mukai et al. 2002). GSK3β1 is the most widely expressed isoform, however, GSK3β2 (which contains a 13 amino acid insert because of use of exon 8A) is highly enriched within the brain (Mukai et al. 2002). The inserted sequence is within the kinase domain near to the substrate binding pocket, however, the effect on kinase activity, substrate specificity or requirement for priming of substrates remains unclear, although GSK3β2 phosphorylates non-primed residues on tau and MAP1B to a lesser extent than GSK3β1 (Mukai et al. 2002; Wood-Kaczmar et al. 2009).

In this report, we examine phosphorylation of several proposed GSK3 substrates in cortical tissue lacking either or both GSK3 genes, and examine substrate phosphorylation by the brain enriched GSK3β2 splice-variant.

Experimental procedures

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Materials

pCMV5-FLAG-tau, CT99021, collapsin response mediator protein (CRMP)2, CRMP4, Inhibitor-2, extracellular stimuli regulated protein kinase (ERK)2, glutathione-S-Transferase (GST)-c-Myc, c-Jun, and Anti-c-Myc pThr58 antibody were provided by the Division of Signal Transduction Therapy, University Dundee. cDNA for GSK3β2 was generated as described (Wood-Kaczmar et al. 2009). Phospho-glycogen synthase 2 (pGS2) peptide [Catalog # 12–241, YRRAAVPPSPSLSRHSSPHQ(pS)EDEEE], GST-p35/His-Cdk5, and casein kinase-2 (CK2) were purchased from Millipore, Watford, UK. Silver Stain kit was from Bio-Rad, Hemel Hempstead, Herts, UK and antibodies to FLAG-M2, Inhibitor-2 pThr72, and Actin were purchased from Sigma Aldrich Inc., Poole, Dorset, UK, to β-Catenin and GSK3α/β pSer9/21 from Cell Signalling Technology (Hitchin, Hertfordshire, UK), to c-Myc (9E10) from Santa Cruz Biotechnology, Santa Cruz, CA, USA, to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and c-Jun pThr239 from Abcam PLC (Cambridge, MA, USA), to GSK3β pTyr216 and GSK3β from BD Biosciences (San Jose, CA, USA) to sheep IgG coupled to Alexa Fluor®680 from Invitrogen (Paisley, UK) and to rabbit IgG coupled to IRDye™800 from Rockland Immunochemicals Inc. (Gilbertsville, PA, USA).

Optimisation of substrate priming and characterisation of purified GSK3 activity

Kinase assays contained a mixture of purified kinase, protein or peptide substrate (final concentrations indicated in figure legends), 10 mM MgCl2, 0.1 mM [γ-32P] ATP (approximately 106 cpm/nmol) and kinase buffer [50 mM Tris–HCl, 0.03% (v/v) Brij-35, 0.1% (v/v) β-mercaptoethanol, pH 7.4]. All assays were performed at 30°C with shaking for the times indicated in text or figure legends. Reactions were terminated in one of two ways; either by precipitation by addition of trichloroacetic acid (final concentration 20% v/v in the presence of 50 μg bovine serum albumin), or by adsorption onto 2 cm2 of p81 paper (Whatman) (Sutherland et al. 1993a). Phosphate incorporation into substrate was measured by Scintillation counting. One unit of kinase activity is defined as that amount catalysing the incorporation of 1 nmol of phosphate into peptide substrate in 1 min (specific activity).

Linked GSK3 assays

Substrate priming was carried out using non-radiolabeled MgATP as above, for the times established to result in maximal priming. Subsequently, GSK3β was added along with additional MgCl2 and ATP ([γ-32P] 106 cpm/nmol) to a final concentration of 10 mM and 0.1 mM respectively, for the times indicated. Reactions were terminated by addition of 4× sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) loading buffer (Invitrogen), proteins separated by SDS–PAGE (4–12% Novex gels, Invitrogen, Paisley, Scotland, UK) and visualised by Coomassie staining. Substrate proteins were excised from the gel and the phosphate incorporation assessed by Scintillation counting.

GSK3 KO animals/tissue

The GSK3α and β brain specific KO mouse was generated by crossing GSK3α KO (exon 2 deletion) mice and GSK3β loxP/loxP (exon 2 flanked with loxP sites) mice, followed by crossing the homozygous offspring with a nestin-cre mouse, which drives the expression of Cre recombinase from embryonic day 10 in CNS progenitors (Kim et al. 2009). Tissue lysates were prepared in ristocetin-induced platelet agglutination buffer [50 mM Tris–HCl (pH 7.4), 1% NP-40 (Sigma Aldrich, Dorset, UK), 0.25% Na-deoxycholate, 150 mM NaCl, 1 mM EDTA, 1× protease inhibitor cocktail (Sigma), 1× phosphatase inhibitor cocktail (Sigma)] as described previously (Kim et al. 2009).

Results

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Phosphorylation of proposed GSK3 substrates in tissue lacking GSK3α, GSK3β or both

Cortical brain tissue was isolated from mice lacking either or both GSK3 isoforms, and age-matched controls (embryonic day 15). Protein lysates were prepared from tissue and analyzed for GSK3 expression and phosphorylation by western blot (Fig. 1a). A relatively small amount of each GSK3 isoform was detectable in the double KO samples, probably arising from non-neuronal tissue contaminating the samples. However, this represents less than 5% of the amount of each isoform detectable in the wild-type tissue (relative to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) internal control –Fig. 1a). GSK3β (47 kDa) appears to be the predominant isoform in this tissue at this age (Fig. 1a, pTyr216, long exposure, 3rd panel). The pTyr216 antibody was raised to a peptide sequence that is fully conserved between each GSK3 isoform thus the affinity of both isoforms for the antibody is identical. This modification occurs during synthesis of the polypeptide (Lochhead et al. 2006) and is generally highly phosphorylated (Cole et al. 2004a) hence it can be used to give an approximation of isoform ratio.

image

Figure 1.  Analysis of CRMP phosphorylation by GSK3α and GSK3β. Cortical protein was isolated from mice lacking GSK3α, GSK3β or both, as well as from wild type mice (WT). (a) The expression (total) and phosphorylation (pTyr) of each GSK3 isoform and (b) the expression and phosphorylation (at the priming site – p522, and the GSK3 target site – p509) of two isoforms of CRMP was assessed in each genotype by western blot using the antibodies indicated. (c) Following priming with Cdk5, purified CRMP2 and CRMP4 were incubated in vitro with either GSK3α or GSKβ1 plus Mg [γ-32P]-ATP for the times indicated, then the reactions separated by SDS–PAGE. The phosphorylated CRMP was visualised by autoradiography (representative image presented). (d) The phosphorylation stoichiometry of (c) was quantified by subjecting each isolated band to scintillation counting, and the data presented as moles of phosphate incorporated per mole of CRMP (average of two separate experiments performed in duplicate), for CRMP2 (C2) and CRMP4 (C4) after incubation with GSK3β1 (β1) or GSK3α (α) for the times indicated.

Download figure to PowerPoint

Pharmacological inhibition of GSK3 reduces phosphorylation of CRMP2 and CRMP4 at Thr509 (Cole et al. 2004b, 2007). Consistent with this, CRMP2 and CRMP4 phosphorylation at Thr509 was almost undetectable in the GSK3β and GSK3 double KO cortex tissues (Fig. 1b), while CRMP isoform expression, and phosphorylation at Ser522 (phosphorylated by Cdk5) was similar in all animals. This confirms Thr509 (and almost certainly Thr514 and Ser518) of CRMP2 and CRMP4 are substrates for GSK3βin vivo, and that Ser522 (priming site) is not regulated by GSK3. Surprisingly, CRMP phosphorylation in the GSK3α knockout tissue was similar to control (Fig. 1b), while CRMP phosphorylation at Thr509 was absent in the GSK3β single KO tissue (Fig. 1b), suggesting that GSK3β is the main isoform regulating CRMP2 and CRMP4 phosphorylation at these sites in vivo. This may be partly explained by relatively higher expression of GSK3β compared with GSK3α (based on pTyr216 blot in Fig. 1a). However, when CRMP2 and CRMP4 were incubated with GSK3α or GSK3β1 (matched for pGS2 peptide kinase activity) and [γ32-P]-ATP in vitro (Fig. 1c), GSK3β1 phosphorylated both CRMP isoforms at a higher rate than GSK3α, consistent with CRMP being a better substrate for GSK3β1. The rate of phosphorylation of CRMPs by GSK3β1 was around four times faster than with an equivalent amount of GSK3α under initial rate conditions (Fig. 1d).

Next, we analysed inhibitor-2 (Inh-2), c-Myc, and β-catenin, as these proteins are also proposed substrates of GSK3 (Liu et al. 2002; Gregory et al. 2003). Phosphorylation of Inh-2 at Thr72 was identical in all samples, questioning the role of GSK3 in its phosphorylation, at least in this tissue (Fig. 2a). Meanwhile, c-Myc and β-catenin protein levels increased in the double KO tissue relative to control, with little effect of knocking out either isoform individually, consistent with previous data (Kim et al. 2009). GSK3 phosphorylation promotes degradation of these proteins. This suggested that these two substrates can be phosphorylated by both isoforms of GSK3 in vivo, in contrast to CRMP (Fig. 1b).

image

Figure 2.  GSK3 substrate phosphorylation in cortical tissue lacking GSK3α and GSK3β. (a) The same cortical brain lysates from Fig. 1(a) were analysed by western blot for phosphorylation of Inh-2, or expression of β-catenin and c-Myc (both proteins that are degraded following phosphorylation by GSK3). (b) The expression (tau-5) and phosphorylation of tau (at residues indicated) in wild-type (WT), GSK3α knockout (αKO), GSK3β knockout (βKO) and total GSK3 knockout (α/β KO) was compared by western blot. A non-specific band of 52 kDa (NS) was seen in some longer exposures and is also found in control blots with no primary antibody (data not shown). (c) Quantification of phosphorylation relative to tau expression (tau5) was performed on three different sets of mice, and data are presented relative to WT animals.

Download figure to PowerPoint

Finally, we analysed phosphorylation of the microtubule associated protein tau (Fig. 2b and c), another well characterised GSK3 substrate (Hanger et al. 2009). The phosphorylation of tau at Ser404, Ser199, Ser202 and Thr231 was similar between all animals examined, suggesting the phosphorylation of these residues did not require GSK3, at least in healthy animals of this age (Fig. 2c). However, phosphorylation of tau at the dual epitope AT180 (pT231/pT235), and the single site, pSer396, was dramatically reduced in the double KO tissue and also in the tissue lacking only GSK3β, relative to control (Fig. 2b), suggesting these residues required GSK3 activity for phosphorylation. Interestingly, AT180 and Ser396 were phosphorylated normally in the GSK3α KO tissue, suggesting that they are phosphorylated predominantly by GSK3β. Finally, loss of GSK3β but not GSK3α increased the mobility of tau in SDS–PAGE (Fig. 2b, tau-5 antibody), suggesting that the phosphorylation event(s) that reduced mobility of tau on SDS–PAGE is specific to GSK3β (the AT180 site is a candidate for this).

Hence differential expression of the two GSK3 genes clearly alters substrate phosphorylation profile in intact neuronal tissue. Next we investigated whether the neuronally enhanced splice variant of GSK3β also had differential substrate phosphorylation preferences. This had to be performed in vitro because of the current lack of in vivo models with specific deletion of each of the splice variants.

GSK3β1 and GSK3β2 expression and purification from HEK293 cells

GST-GSK3β1 or GST-GSK3β2 fusion proteins were produced in HEK293 cells in order to obtain sufficient amounts to perform in vitro phosphorylation studies. Expression levels were similar for each isoform and there was no difference in phosphorylation of Tyr216 [autophosphorylated residue required for kinase activity (Cole et al. 2004a)], whether the cells were cultured with or without serum (Figure S1). Serum withdrawal from the cells prior to isolation of protein completely removed the inhibitory Ser9 phosphorylation and this was routinely used to generate maximally active GSK3 isoforms (Figure S1).

Each GST-GSK3 was purified from transfected HEK293 lysates on GSH-sepharose beads, and the human GSK3β released from the GSH-sepharose beads by cleavage with PreScission protease (see Appendix S1). Western blot demonstrated similar amounts of GSK3β1 and GSK3β2 could be purified from the HEK293 lysates (Figure S1).

Characterisation of purified GSK3β1 and GSK3β2 kinase activity in vitro

The specific activities of the purified GSK3β1 and GSK3β2 were estimated using a peptide substrate based on the sequence of glycogen synthase (pGS2, Figure S2a). At two different substrate concentrations (3 and 15 μM) GSK3β1 consistently phosphorylated pGS2 at a higher rate than an equivalent amount of GSK3β2 (for several different preparations of each enzyme). This was not because of variation in Tyr216 or Ser9 phosphorylation (Figure S1). The specific activity of GSK3β1 was approximately 11 Units/mg, while GSK3β2 exhibited lower activity toward the peptide of around 6 Units/mg. The Km of the pGS2 peptide was approximately 3.5 μM for GSK3β1 and 10 μM for GSK3β2. Therefore, the 13 amino acid insert in GSK3β2 reduces affinity for the peptide substrate. CT99021 is a cell permeable, ATP competitive, highly specific inhibitor of GSK3 activity (Wagman et al. 2004; Bain et al. 2007). The IC50 of CT99021 against GSK3β1 and GSK3β2 (assessed in vitro against pGS2 peptide) was around 50 nM for both (Figure S2b). This suggests that the ATP binding pocket of each isoform is structurally very similar, and that this compound will not distinguish between the isoforms in cells.

CRMP2 and CRMP4 are better substrates for GSK3β1 than GSK3β2 in vitro

Glycogen synthase kinase-3 β1 phosphorylation of CRMP isoforms requires prior phosphorylation of the substrates at Ser522 (Cole et al. 2004b). Incubation of CRMP2 or CRMP4 with p35-Cdk5 and ATP for 1 h phosphorylated both substrates to 0.2 mol/mol, almost exclusively at Ser522 (data not shown). CRMP2 and CRMP4 (with or without prior phosphorylation by Cdk5) were then incubated with GSK3β1 or GSK3β2 in the presence of [γ-32P] ATP. CRMP isoforms were only phosphorylated by either GSK3β isoform following priming by Cdk5 (Fig. 3a). GSK3β1 introduced more phosphate into CRMP2 and CRMP4 than GSK3β2 even though the amount of each GSK3β enzyme used was matched for activity towards peptide (Units of pGS2 kinase activity). Identical non-radioactive reactions (analysed by western blot) confirmed that GSK3 did not phosphorylate either CRMP without prior priming at Ser522, and that GSK3β1 phosphorylated both primed CRMP isoforms more efficiently than GSK3β2 (Fig. 3b).

image

Figure 3.  Comparison of the rate of phosphorylation of CRMP isoforms by GSK3β1 and GSK3β2. (a) 1 μM FLAG-CRMP2 or FLAG-CRMP4 was incubated with p35-Cdk5+ non-radiolabeled MgATP for 1 h at 30°C in order to phosphorylate CRMP at Ser522 (PRIMED). Primed and unprimed CRMP were then incubated with GSK3β1, GSK3β2 or no kinase (control) in the presence of Mg [γ-32P] ATP. Aliquots were removed at 15 and 180 min, separated by SDS–PAGE, stained with Coomassie, and subjected to autoradiography. A representative of two independent experiments is presented. (b) The same experiment was performed using identical reagents except with non-radiolabelled MgATP prior to SDS–PAGE and this time the proteins were analysed by western blot using antibodies as indicated. A representative of two independent experiments is shown.

Download figure to PowerPoint

In order to establish whether this differential phosphorylation by the GSK3β isoforms was specific for this substrate we examined phosphorylation of other reported GSK3 substrates in a similar fashion.

Inhibitor-2 phosphorylation is differentially regulated by GSK3β1 and GSK3β2

Inhibitor-2 is a potent inhibitor of PP1, however, phosphorylation of Inh-2 at Thr72 by GSK3 promotes a conformational change thus inducing PP1 activity (Picking et al. 1991). Phosphorylation of Ser86 of Inh-2 by CK2 augments GSK3 phosphorylation of Inh-2 at Thr72, a highly unusual priming event (Holmes et al. 1986). Inh-2 (1 and 5 μM) was primed by incubation with CK2 and ATP, promoting phosphorylation to a stoichiometry of > 1.5 mol/mol (data not shown). This high level of stoichiometry reflects CK2’s ability to phosphorylate multiple sites on Inh-2, however, Ser86 (priming site) is the major site targeted (Park et al. 1994). GSK3β1 phosphorylated unprimed Inh-2 to around 0.08 (15 min) and 0.36 (180 min) mol/mol (Fig. 4a, left panels). GSK3β2 phosphorylation of Inh-2 was considerably lower at 0.03 (15 min) and 0.05 (180 min) mol/mol. GSK3β1, therefore, phosphorylated Inh-2 three to seven times faster than GSK3β2, suggesting GSK3β1 was a more efficient unprimed Inh-2 kinase than GSK3β2 in vitro. GSK3β1 phosphorylation of primed Inh-2 reached 0.15 (15 min) and 0.64 (180 min) mol/mol (Fig. 4a, right panels). GSK3β2 phosphorylation of primed Inh-2 was again considerably lower than GSK3β1, but enhanced over GSK3β2 phosphorylation of unprimed Inh-2. GSK3β1 phosphorylates primed Inh-2 two to three times faster than GSK3β2. The data confirmed that Inh-2 can be phosphorylated by GSK3β independent of priming, but that priming (albeit in an unusual sequence) enhanced GSK3β phosphorylation of Inh-2.

image

Figure 4.  Comparison of the rate of phosphorylation of inhibitor-2 by GSK3β1 and GSK3β2 in vitro. (a) Inh-2 (1 μM or 5 μM) was primed by incubation with CK2 and Mg[ATP]. Primed or unprimed Inh-2 was then incubated with GSK3β1, GSK3β2 or no kinase (control) in the presence of Mg [γ-32P] ATP. Aliquots were removed at 15 and 180 min, separated by SDS–PAGE, stained with Coomassie and subjected to autoradiography. A representative of two independent experiments is presented. (b) The same experiment was performed using identical reagents except with non-radiolabelled MgATP prior to SDS–PAGE. Samples were removed at 15 and 180 min and subjected to SDS–PAGE followed by western blot using a phospho-specific antibody against Thr72 of Inh-2, and silver staining to visualise Inh-2 loading. A representative of two independent experiments is shown.

Download figure to PowerPoint

These experiments were repeated and monitored by western blot using an Inh-2 Thr72 phosphorylation specific antibody (Fig. 4b). Consistent with the phosphate incorporation data, unprimed Inh-2 was phosphorylated at Thr72 by GSK3β1, priming enhanced phosphorylation by both GSK3β isoforms, and GSK3β1 was a more efficient Inh-2 Thr72 kinase than GSK3β2, with or without priming.

GSK3β1 and GSK3β2 phosphorylate c-Jun and c-Myc to similar extents in vitro

c-Jun is a proto-oncogene and was the first transcription factor identified as a GSK3 substrate (Boyle et al. 1991). Priming of c-Jun at Ser243 by DYRK2 [dual-specificity tyrosine-(Y)-phosphorylation regulated kinase] allows the phosphorylation of Thr239 by GSK3 (Morton et al. 2003). DYRK2 phosphorylated c-Jun to 0.75 mol/mol, and subsequent incubation with GSK3β1 or GSK3β2 promoted phosphorylation of Thr239 of c-Jun (Fig. 5a). Phosphorylation of Thr239 was observed in the absence of GSK3 (control), and DYRK2 (or a contaminating kinase in the purified DYRK2 or c-Jun preps) was the most likely reason for this background phosphorylation. Addition of GSK3β to the incubation increased the phosphorylation of Thr239 above this background, and both GSK3β isoforms increased c-Jun phosphorylation at this site by a similar extent.

image

Figure 5.  Phosphorylation of c-Myc and c-Jun by GSK3β1 and GSK3β2 in vitro. (a) c-Jun tagged by fusion with Maltose Binding Protein (c-Jun Mal BP, 1 μM) was primed by incubation with DYRK2 in the presence of non-radiolabeled MgATP for 30 min at 30°C. The primed substrate was then incubated with GSK3β1, GSK3β2 or no kinase (control) in the presence of non-radiolabeled MgATP. Samples were removed at 15 and 180 min and subjected to western blot, using a phosphorylation specific antibody against c-Jun pThr239 while total c-Jun Mal BP was visualised by silver staining. (b) GST-c-Myc was primed by incubation with ERK2 in the presence of non-radiolabeled MgATP for 30 min at 30°C. The primed substrate was then incubated with GSK3β1, GSK3β2 or no kinase (control) in the presence of non-radiolabeled MgATP. Samples were removed at 15 and 180 min and subjected to SDS–PAGE followed by western blot, using a phosphorylation specific antibody against c-Myc pThr58, and a total GST-c-Myc antibody.

Download figure to PowerPoint

c-Myc is a proto-oncogene and transcription factor that is phosphorylated by GSK3 at Thr58 following priming at Thr62 by p42/p44 MAPK, leading to degradation of c-Myc (Gregory et al. 2003). Incubation of c-Myc and p42MAPK resulted in phosphorylation of c-Myc to around 1 mol/mol after 30 min (data not shown). Subsequent to priming GSK3β1 and GSK3β2 both phosphorylated c-Myc at Thr58 to a similar extent (Fig. 5b). These data suggest c-Jun and c-Myc are equally phosphorylated by the two GSK3β splice variants, at least in vitro.

Phosphorylation of GSK3β substrates following over-expression of the two isoforms in HEK293 cells

To investigate phosphorylation of these GSK3 substrates in cells, GSK3β1 or GSK3β2 was transiently expressed in HEK293 cells for 32 h, then cells were serum starved for 16 h in order to maximally activate GSK3, and finally cells were incubated with the GSK3 inhibitor CT99021 (2 μM) for 2 h prior to cell lysis and isolation of total cellular protein.

Over-expression of GSK3β1 produced a 1.73-fold significant increase in endogenous Inh-2 Thr72 phosphorylation while over-expression of GSK3β2 did not significantly change phosphorylation of this residue compared with the control (Fig. 6). Incubation of the HEK293 cells with CT99021, almost completely removed Thr72 phosphorylation, whether GSK3β isoforms were over-expressed or not (Fig. 6a). These cell-based studies implied that Inh-2 was a better substrate for GSK3β1 than GSKβ2, in support of the in vitro data (Fig. 4). They also indicated that in HEK293 cells at least, Thr72 of Inh-2 was a GSK3 target (as the relatively specific GSK3 inhibitor reduced phosphorylation).

image

Figure 6.  Expression of GSK3β1 and GSKβ2 in HEK293 cells. (a) HEK293 cells were transiently transfected with one of GST-GSK3β1, GST-GSK3β2 or control plasmid without (left panel) or with FLAG-tau (right panel). After 32 h the cells were serum starved for a further 16 h, prior to incubation ± 2 μM CT99021 for 2 h and harvest of cellular protein. Samples were subjected to SDS–PAGE and western blot for expression and phosphorylation of the GSK3 substrates, using antibodies as indicated. (b) Immunoreactive bands were quantified from two experiments performed in duplicate, and the mean signal ± SEM is shown relative to control for each antibody. *p < 0.05, comparison of GSK3β1 expressing cells and control; **p < 0.05, comparison of GSK3β1 and GSK3β2.

Download figure to PowerPoint

Over-expression of GSK3β1 and GSK3β2 slightly increased the phosphorylation of endogenous β-catenin (at Ser33/37 and Thr41) to a similar extent, suggesting that β-catenin is equally phosphorylated in cells by these two isoforms (Fig. 6a and b). However, GSK3β over-expression did not significantly affect endogenous c-Jun phosphorylation (Ser239), total c-Myc (which is degraded upon phosphorylation by GSK3) or total β-catenin protein levels, compared to control (Fig. 6a and b). The inhibition of GSK3 using CT99021 (Fig. 6a) resulted in dephosphorylation of endogenous c-Jun, Inh-2 and β-catenin and dramatically increased c-Myc and β-catenin expression. This confirmed that these proteins are indeed regulated by GSK3 activity in HEK293 cells.

It should be noted that when priming of GSK3 substrates is a limiting factor phosphorylation would not be induced by co-expression of GSK3, however, phosphorylation would be reduced by GSK3 inhibition. In addition, it would be difficult to increase phosphorylation (by over-expression of GSK3) of any substrates that are phosphorylated to high stoichiometry by endogenous GSK3, yet these would also be dephosphorylated in cells incubated with a GSK3 inhibitor.

The microtubule associated protein, tau has been proposed as a GSK3 substrate although the precise residues on tau phosphorylated by GSK3 in health or disease remain controversial. Our analysis of GSK3 KO brain suggested that the phosphorylation of several residues on tau requires GSK3β activity (Fig. 2). Therefore, we co-expressed human FLAG-tagged tau (isoform 2) and the GSK3β isoforms in HEK293 cells, and analysed phosphorylation of specific sites (Fig. 6a, right panel). Inhibition of GSK3 with CT99021 for 2 h promoted dephosphorylation of Ser396, consistent with loss of phosphorylation of this site in the GSK3 KO brain (compare Figs 2 and 6a). The phosphorylation of Ser396 and the AT180 epitope (Thr231/Thr235) increased when GSK3β1 or GSK3β2 were co-expressed, and this was more apparent with GSK3β1 than GSK3β2 (Fig. 6b), suggesting they were more efficiently phosphorylated by the shorter isoform of GSK3β, similar to CRMPs. We were unable to detect phosphorylation of the AT8 epitope of human tau expressed in HEK293 cells, even when tau was co-expressed with GSK3β. Inhibition of GSK3 with CT99021 had little effect on the relative level of phosphorylation of Ser199 or Ser404 of human tau, while over-expression of either GSK3β isoform did not induce the phosphorylation of these residues. This is consistent with the GSK3 KO data and suggested that these are unlikely to be GSK3 regulated sites. Taken together these data demonstrate the importance of studying specific residues on tau during functional studies, rather than assuming tau ‘hyperphosphorylation’ is a single entity.

Discussion

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Validating physiological substrates of GSK3

There are over fifty proposed substrates of GSK3, and this number continues to grow. The tools (KO animals and highly selective inhibitors) are now available to establish whether selective loss of GSK3 activity affects phosphorylation of proposed substrates. The use of highly unspecific inhibitors such as lithium, or in vitro studies with no comparison of stoichiometry or phosphorylation rate can be very misleading and may account for a high degree of inaccuracy in the ‘GSK3 target proteome’. Even with the aid of targeted deletion of GSK3 isoforms care must be taken in interpretation of phosphorylation data. For example in the GSK3 KO tissue Inh-2 phosphorylation at Thr72 was the same as wild type tissue (Fig. 2), yet the phosphorylation of this residue was acutely reduced in cells incubated with a highly selective GSK3 inhibitor (Fig. 6a), and Inh-2 is a relatively good substrate for GSK3 in vitro (Fig. 4). Clearly there are many potential reasons for the discrepancy between the genetic and pharmacologic data. Deletion of GSK3 could induce compensatory expression of an alternative Thr72 kinase that takes more than a few hours to occur. Similarly, there could be tissue specific Thr72 kinases, or CT99021 may not be a completely specific inhibitor of GSK3. In contrast, the phosphorylation of CRMPs is reduced in both the GSK3β KO tissue and in cells treated with CT99021, and CRMPs are phosphorylated by GSK3 at a comparatively high rate in vitro. The evidence is therefore very strong that these are physiological GSK3 substrates. Meanwhile, the phosphorylation/stability of β-catenin, c-Myc, c-Jun and some sites on tau are all affected by loss of GSK3 activity genetically or pharmacologically. This suggests that these are all bona fide substrates of GSK3. However, even this relatively small number of proposed GSK3 substrates demonstrate distinct sensitivities to the GSK3 isoforms.

Comparison of GSK3β isoform substrate specificity

The ATP binding site of GSK3β sits in the hydrophobic pocket between the two lobes, enclosed by a glycine rich loop (aa 60–70) and a hinge region (aa 134–139) while the substrate binding domain is within the C-terminal region. The GSK3β2 isoform contains a 13 amino acid insert, which lies within the catalytic domain (between aa 303 and 304 of GSK3β1), in a linker region flanked by α-helices (between domain X and XI). This region is known to reduce the binding affinity for axin, and is very close to the binding region of frequently rearranged in advanced T-cell lymphoma (FRAT) (Wood-Kaczmar et al. 2009; Castano et al. 2010). It was therefore possible that the insert of GSK3β2 would confer altered substrate specificity/affinity of a subset of known GSK3β substrates.

We present data demonstrating that the insert does not alter the regulation of GSK3 by phosphorylation (at Ser9 or Tyr216), and does not affect inhibition by the ATP competitive inhibitor CT99021. However, the specific activity of the purified GSK3β2 is almost 2.5 times lower than GSK3β1, when measured using a standard peptide substrate for GSK3 (pGS2). Interestingly, even after correcting for this reduced peptide phosphorylating activity, GSK3β2 has a reduced ability to phosphorylate CRMP2, CRMP4 and Inh-2 in vitro, suggesting that the difference in substrate affinity for these full length substrates is even more significantly affected by the insert than pGS2 peptide. In addition, phosphorylation of tau at two epitopes (AT180 and Ser396) was more pronounced when GSK3β1 was co-expressed in cells, compared to GSK3β2. This is consistent with a previous proposal that tau is a better substrate in vitro for the GSK3β1 splice variant (Mukai et al. 2002), although the sites on tau analysed in that study were Ser404 and Ser199, which are not affected by deletion of GSK3β or inhibition using CT99021 (Figs 2 and 6). Our findings with tau are similar to experiments where tau was co-expressed with green fluorescent protein (GFP)-GSK3β1 and GFP-GSK3β2 fusion proteins (Castano et al. 2010). Finally, GSK3β1 and GSK3β2 phosphorylate the transcription factors c-Myc and c-Jun at a comparable rate in vitro and when co-expressed in HEK293 cells, suggesting that these substrates are equally phosphorylated by these GSK3 isoforms in vivo.

Therefore, there are GSK3 substrates with preference for GSK3β1 over GSK3β2, and others that hold no distinction, suggesting different modes of interaction with GSK3. It is reasonable to predict that there are also substrates not studied here with higher affinity for GSK3β2 over GSK3β1. One assumes that these are likely to be neuronal proteins because of the expression profile of this splice variant. Importantly, our data indicates that the insert does not affect the relative affinity for primed over non-primed substrates, as the preferential targeting of substrates by GSK3β1 does not correlate with priming.

Physiological significance of GSK3β2 expression

In neurons, the axons and dendrites are clearly distinguishable from each other, whereas glial cells possess irregular less defined processes. Interestingly, GSK3β2 is not expressed in glial cells but is highly expressed in neurons, particularly within neurites and growth cones during neuronal differentiation (Goold and Gordon-Weeks 2001; Wood-Kaczmar et al. 2009). Specific knockdown of GSK3β2 using shRNA has clearly established a role for this variant during the development of neuronal polarity, and the growth of axons (Castano et al. 2010). It will be interesting to establish the GSK3 substrates whose phosphorylation changes during the establishment of neuronal polarity, and those with a preference for GSK3β2 over GSK3β1 are highly likely to play a role in neuronal polarity.

Potential for isoform specific therapeutic intervention

Glycogen synthase kinase-3 activity and substrate phosphorylation [e.g. glycogen synthase, tau, CRMP2 and amyloid precursor protein (APP)] are reported to be abnormally high in both Type 2 diabetes and Alzheimer’s disease (Frame and Cohen 2001; Bhat et al. 2004; Cole et al. 2007). This has led to the development of selective inhibitors of GSK3, however, development as therapeutics has been hampered by the wide range of cellular substrates of this enzyme, increasing the risk of unwanted side effects of GSK3 inhibition. Therefore, the establishment of distinct substrate pools for the different GSK3 isoforms provides the opportunity to investigate whether inhibition of a specific GSK3 isoform is worth pursuing. Clearly developing isoform specific inhibitors is not a trivial task, however, a GSK3β2 specific inhibitor is unlikely to have side effects outside of the CNS. In the short term, discovering more detail on the determinants of substrate recognition may lead to novel ideas on how to target substrate specific interactions, thereby focussing on those phosphoproteins clearly involved in disease progression.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

The work was supported by the MRC (capacity building studentship to MS, and MRC Protein Phosphorylation Unit). All authors declare no conflict of interest.

References

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Experimental procedures
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  8. Supporting Information

Appendix S1. Supplementary Materials and Methods.

Figure S1. Expression and purification of recombinant active GSK3β1 or GSK3β2 HEK293 cells were transiently transfected with human GST-GSK3β1 or GST-GSK3β2. After 32 h half of the cells were serum starved for 16 h prior to protein harvest. Protein lysates were separated by SDS–PAGE and visualised by western blot using total and phosphorylation specific (pSer21/9 or pTyr216) GSK3 antibodies. A representative of two different preparations is shown.

Figure S2. Assessment of the specific activity of purified recombinant human GSK3β1 and GSK3β2. (a) Purified GSK3β1 and GSK3β2 were assayed in duplicate against pGS2 peptide (3 μM and 15 μM) in vitro in the presence of Mg [γ-32P] ATP for 15 min at 30°C. [γ-32P] phosphate incorporation into pGS2 peptide was determined by scintillation counting and data are presented as moles phosphate incorporated per mole of peptide substrate (mean ± SD). (b) 4 mU of GSK3β1 and GSK3β2 peptide kinase activity were incubated with 15 μM pGS2 peptide in the presence of Mg [γ-32P] ATP and the GSK3 inhibitor CT99021 (at the concentrations indicated) for 15 min at 30 C. The reaction was stopped and quantified as in (a). Data are presented relative to activity in the absence of CT99021.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JNC_6988_sm_Figs1-2.pdf288KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.