AS160 (Akt substrate of 160 kDa) and TBC1D1 are related RabGAPs (Rab GTPase-activating proteins) implicated in regulating the trafficking of GLUT4 (glucose transporter 4) storage vesicles to the cell surface. All animal species examined contain TBC1D1, whereas AS160 evolved with the vertebrates. TBC1D1 has two clusters of phosphorylated residues, either side of the second PTB (phosphotyrosine-binding domain). Each cluster contains a 14-3-3-binding site. When AMPK (AMP-activated protein kinase) is activated in HEK (human embryonic kidney)-293 cells, 14-3-3s bind primarily to pSer237 (where pSer is phosphorylated serine) in TBC1D1, whereas 14-3-3 binding depends primarily on pThr596 (where pThr is phosphorylated threonine) in cells stimulated with IGF-1 (insulin-like growth factor 1), EGF (epidermal growth factor) and PMA; and both pSer237 and pThr596 contribute to 14-3-3 binding in cells stimulated with forskolin. In HEK-293 cells, LY294002 inhibits phosphorylation of Thr596 of TBC1D1, and promotes phosphorylation of AMPK and Ser237 of TBC1D1. In vitro phosphorylation experiments indicated regulatory interactions among phosphorylated sites, for example phosphorylation of Ser235 prevents subsequent phosphorylation of Ser237. In rat L6 myotubes, endogenous TBC1D1 is strongly phosphorylated on Ser237 and binds to 14-3-3s in response to the AMPK activators AICAR (5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside), phenformin and A-769662, whereas insulin promotes phosphorylation of Thr596 but not 14-3-3 binding. In contrast, AS160 is phosphorylated on its 14-3-3-binding sites (Ser341 and Thr642) and binds to 14-3-3s in response to insulin, but not A-769662, in L6 cells. These findings suggest that TBC1D1 and AS160 may have complementary roles in regulating vesicle trafficking in response to insulin and AMPK-activating stimuli in skeletal muscle.
Rabs are GTP-activated proteins that regulate the flow of vesicle traffic between organelles by interacting with effectors of vesicle formation, motility, docking and fusion . Inactive GDP-Rabs are held by Rab GDIs (GDP-dissociation inhibitors), and upon release GEFs (guanine-nucleotide-exchange factors) convert GDP-Rabs into the active GTP-Rabs. GAPs (GTPase-activating proteins) inactivate Rabs by enhancing their intrinsic GTP-hydrolysing activities .
AS160 (Akt substrate of 160 kDa)/TBC1D4 and TBC1D1 are related RabGAPs implicated in regulating translocation of GLUT4 (glucose transporter 4) glucose transporters from intracellular GSVs (GLUT4 storage vesicles) to the plasma membrane [3,4]. It has been proposed that AS160 and/or TBC1D1 must be inhibited for GSV-bound Rabs to stay loaded with GTP so that GLUT4 can move to the cell surface and import glucose into cells in response to stimuli such as insulin and exercise [5,6]. The tissue distribution of AS160 and TBC1D1 is wider than GLUT4 however [7,8], suggesting that these RabGAPs also regulate trafficking of vesicles containing other unknown cargoes.
The mechanisms of control of AS160 are still being defined. At least eight residues are phosphorylated in response to insulin, IGF-1 (insulin-like growth factor 1), EGF (epidermal growth factor), PMA and the AMPK (AMP-activated kinase) activator AICAR (5-aminoimidazole-4-carboxamide-1-β-D-ribofuranoside), though the extents of phosphorylation of different sites vary with each stimulus [3,9,10]. In IGF-1-, EGF- and PMA-stimulated cells, 14-3-3 proteins have been found to bind with high affinity to pThr642 (where pThr is phosphorylated threonine) on AS160, whereas pSer341 (where pSer is phosphorylated serine) provides a secondary 14-3-3-binding site [10,11]. These findings suggest the hypothesis that a single 14-3-3 dimer docks on to pThr642 and pSer341, thereby affecting the function of the PTB (phosphotyrosine-binding domain), which is located between these two residues. Ectopic expression of non-phosphorylatable forms of AS160 that cannot bind to 14-3-3s is sufficient to impair insulin-stimulated glucose uptake into cells [11,12], consistent with the possibility that phosphorylation and 14-3-3 binding across the PTB domain inhibits AS160.
The signatures of AS160 phosphorylation by protein kinases in vitro, and effects of protein kinase inhibitors in cells, are consistent with Thr642 of AS160 being phosphorylated by PKB (protein kinase B) in response to IGF-1, and Ser341 and Thr642 being phosphorylated by RSK (p90 ribosomal S6 kinase) in response to PMA. However, the kinase that mediates phosphorylation of Ser341 in response to IGF-1 is unknown . Ser588 is strongly phosphorylated by AMPK in vitro and in response to the AMPK activator AICAR in cells. However, the more potent AMPK activator phenformin does not promote AS160 phosphorylation. Thus it is possible that an AICAR-activated kinase other than AMPK phosphorylates Ser588in vivo .
Less is known about TBC1D1, except that a mutation in TBC1D1 has been linked with a strong familial predisposition to severe obesity in females [8,13,14], and TBC1D1 is similar to AS160 with respect to its specificity towards Rabs in vitro, phosphorylation on a threonine residue (Thr596, within a potential PKB recognition sequence) in response to insulin, and wide tissue distribution . One distinction between AS160 and TBC1D1 is that ectopic expression of wild-type TBC1D1 inhibits GLUT4 translocation, whereas only non-phosphorylatable forms of AS160 have this effect . Stemming from the identification of TBC1D1 among proteins isolated from HEK (human embryonic kidney)-293 cell extracts by 14-3-3-affinity purification , in the present study we investigate its regulation. We find similarities, but also striking differences, in the patterns of phosphorylation and 14-3-3-binding of TBC1D1, compared with what is known about AS160. Our findings give new insight into how clusters of multisite phosphorylation may modulate 14-3-3 binding to target sites.
MATERIALS AND METHODS
Synthetic peptides were from Graham Bloomberg (Department of Biochemistry, University of Bristol, Bristol, U.K.), oligonucleotides were from MWG-Biotech, IGF-1 was from Biosource, microcystin-LR was from Linda Lawton (School of Life Sciences, The Robert Gordon University, Aberdeen, Scotland, U.K.), Vivaspin concentrators were from Vivascience, tissue culture reagents and EGF were from Life Technologies, protease-inhibitor cocktail tablets (catalogue number 1697498) and sequencing-grade trypsin were from Roche Molecular Biochemicals, AICAR was from Toronto Research Chemicals, 6-[4-(2-piperidin-1-yl-ethoxy)-phenyl)]-3-pyridin-4-yl-pyrrazolo[1,5-a]-pyrimidine (Compound C), A23187 and calyculin A were from Calbiochem, metformin, phenformin and PMA were from Sigma–Aldrich, insulin was from Novo-Nordisk, BI-D1870 was from Boehringer Ingelheim Pharma GmbH & Co., precast SDS/polyacrylamide gels were from Invitrogen, and DSP [dithiobis(succinimidyl propionate)] was from Perbio. Protein G–Sepharose and other chromatographic matrices were from GE Healthcare. All other chemicals were from BDH Chemicals or Sigma–Aldrich.
Antibodies and protein kinases
Anti-HA (haemagglutinin) was raised in sheep against the synthetic peptide YPYDVPDYA. The antibodies that recognize phosphorylated sites in TBC1D1 were raised against the following synthetic phosphopeptides: CPMRKSFpSQPGLRS (cysteine for coupling, plus residues 231–243 of human TBC1D1, where pS represents pSer237) and CFRRRANpTLSHFPI (cysteine for coupling plus residues 590–602 of human TBC1D1, where pT represents pThr596). The peptides were conjugated via the added cysteine residues to BSA and keyhole-limpet haemocyanin and injected into sheep at Diagnostics Scotland. The antibodies were affinity-purified on CH-Sepharose coupled to the same peptide. Sheep antibodies for immunoprecipitating TBC1D1 were raised against a GST (glutathione transferase)–TBC1D1-C-terminal fragment (residues 640–1168) that was expressed in Escherichia coli and purified, and antibodies used for immunocytochemistry were raised against full-length GST–TBC1D1. Phosphospecific antibodies for sites in AS160 were reported in . The pan-14-3-3 antibody was K19 from Santa Cruz; the antibodies that recognize pThr172 on AMPK and phosphorylated Thr308 on PKB, anti-PAS [p-(Ser/Thr)-Akt/PKB substrate] antibody, and anti-ERK1/2 (extracellular-signal-regulated kinase 1/2) antibody were from Cell Signaling Technology; and the antibody that recognizes pSer80 on human ACC (acetyl-CoA carboxylase) (pSer79 on the rat protein) and anti-AS160 were from Millipore.
Purified recombinant protein kinases, generated in the DSTT (Division of Signal Tranduction Therapy, University of Dundee, Dundee, Scotland, U.K.), were His–PKBα-S473D (residues 118–480 of the human protein) expressed in insect cells and activated by PDK1 (phosphoinositide-dependent kinase 1); and bacterially expressed GST–AMPK T172D (residues 3–308 of rat). Native AMPK was purified from rat liver by Kevin Green (Division of Molecular Physiology, College of Life Sciences, University of Dundee, Dundee, Scotland, U.K.). PKB was used at 1 unit/ml and AMPK at 10 units/ml, where 1 unit is nmol of phosphate incorporated per min at 30 °C into the substrate peptides Crosstide (GRPRTSSFAEG) for PKB and AMARA peptide (AMARAASAAALARRR) for AMPK. Reactions were carried out at 30 °C for 30 min when a single kinase was used. Sequential phosphorylation with two kinases was carried out by starting the reaction with the first kinase for 15 min followed by the second kinase for a further 30 min.
TBC1D1 cDNA was generated by RT (reverse transcription)–PCR amplification from IMAGE consortium EST 5262043, and cloning into the vector pCMV5.HA-1. Bacterial expression plasmids for GST–TBC1D1-full-length and GST–TBC1D1-C-terminus (encoding residues 640–1168) were generated by cloning into the vector pGEX6P-1 using standard procedures. Site-directed mutations were introduced into the TBC1D1-coding sequences following the QuikChange® protocol (Stratagene) using the KOD HotStart DNA polymerase (Novagene). Sequences of all DNA constructs were checked by the service managed by Nick Helps (University of Dundee, Dundee, Scotland, U.K.; www.dnaseq.co.uk).
Cell culture, stimulations, cross-linking, lysis and immunoprecipitations
Human HEK-293 cells cultured on 10-cm diameter dishes in medium containing 10% (v/v) foetal bovine serum were used 30 h after transfection with the indicated plasmids. Rat L6 myoblasts were maintained in medium containing 10% (v/v) foetal bovine serum and differentiated into myotubes in medium containing 2% (v/v) horse serum for 4 days. Cells were serum-starved for 4 h (unstimulated), then stimulated as indicated with IGF-1 at 50 ng/ml for 15 min, insulin at 100 nM for 30 min, serum at 10% (v/v) for 15 min, calyculin A at 100 nM for ∼5 min, AICAR at 2 mM for 1 h, phenformin at 2 mM for 1 h, A-769662 at 50 μM for 1 h, EGF at 50 ng/ml for 15 min, PMA at 100 ng/ml for 30 min and A23187 at 10 μM for 1 h. Where indicated, cells were incubated with LY294002 (100 μM for 1 h), wortmannin (100 nM for 1 h), Go6983 (1 μM for 30 min), BI-D1870 (10 μM for 30 min) and AMPK inhibitor Compound C (20 μM for 1 h) prior to stimulation with IGF-1 and other stimuli. After stimulation, cells were lysed in 0.3 ml of ice-cold lysis buffer as in .
TBC1D1–14-3-3 interactions were stabilized during cell lysis and immunoprecipitated using a reversible chemical cross-linker. Cells were rinsed with ice-cold PBS, lysed in 0.3 ml of lysis buffer containing DSP (2.5 mg/ml from a 250 mg/ml stock in DMSO) for 30 min on ice, and unreacted cross-linker was quenched with 75 μl of 1 M Tris/HCl (pH 7.4) with a further 30 min incubation [10,17]. For immunoprecipitations with anti-HA and anti-TBC1D1, 3 μg of antibody/mg of lysate was mixed at 4 °C for 2 h, then Protein G–Sepharose (30 μl of a 50% suspension in lysis buffer) was added and mixed for a further 1 h. The suspension was centrifuged at 12000 g for 30 s between washes. Immunoprecipitates were extracted into SDS-sample buffer lacking reducing agent, and centrifuged (12000 g for 1 min at 4 °C) to clarify before adding reducing agent (Invitrogen) to the supernatant.
14-3-3 overlays and Western blots
Membranes were incubated in 50 mM Tris/HCl (pH 7.5), 0.15 M NaCl and 0.2% (v/v) Tween 20 containing 5% (w/v) dried skimmed milk powder (Marvel) and were immunoblotted at 4 °C for 16 h using the indicated antibodies at 1 μg/ml. Detection was performed using horseradish-peroxidase-conjugated secondary antibodies (Promega) and ECL® (enhanced chemiluminescence reagent; Amersham Biosciences) for all Western blots of endogenous proteins and DIG (digoxigenin)-14-3-3 overlays (which use DIG-labelled 14-3-3 in place of primary antibody) . Western blots of HA–TBC1D1 used IR dye-labelled secondary antibodies that were detected using the Odyssey Infrared Imaging System (LI-COR).
Identification of phosphorylated residues in HA–TBC1D1 extracted from cells growing in medium containing serum
HA–TBC1D1 was immunoprecipitated from lysates of cells growing in medium containing serum. After SDS/PAGE, the colloidal Coomassie-Blue-stained bands were digested with trypsin and analysed by LC-MS on a 4000 Q-TRAP system using precursor ion scanning . If the site of phosphorylation could not be assigned from the MS/MS spectra acquired on this mass spectrometer, a second aliquot of the tryptic digest was analysed by LC-MS on a Thermo-Electron LTQ-orbitrap mass spectrometer coupled to a Dionex 3000 nano liquid chromatography system. Precise masses for the phosphopeptides detected by the precursor ion scanning on the 4000 Q-TRAP were entered into an inclusion list and these ion masses were preferentially selected for MS/MS with multistage activation. The resultant data files were searched against the TBC1D1 sequence, using Mascot run on an in-house server, with a 10 p.p.m. mass accuracy for precursor ions and phosphorylation of serine/threonine or tyrosine as variable modifications. The individual MS/MS spectra for the phosphopeptide ions were inspected using Xcalibur 2.2 software to assign the site(s) of phosphorylation.
Except for MS experiments, results shown are representative of at least two similar experiments.
Phylogenetic analysis of TBC1D1 and AS160, and identification of phosphorylated sites on TBC1D1
The phylogeny of AS160 and TBC1D1 sequences from animal species suggests that TBC1D1 is the most ancient gene, whereas AS160 emerged with the vertebrates with subsequent sequence divergence generating the distinct TBC1D1 and AS160 that exist today (Figure 1A).
Sequence alignment and phylogenetic analysis of AS160 (TBC1D4) and TBC1D1
HA–TBC1D1 was extracted from transfected HEK-293 cells that were cultured in medium containing serum. Precursor ion scanning of tryptic digests of the extracted protein revealed the presence of eight phosphopeptides, which were then identified by MS/MS as seven singly phosphorylated peptides and one triply phosphorylated peptide (see Supplementary Table 1 at http://www.BiochemJ.org/bj/409/bj4090449add.htm). Six phosphorylated residues could be assigned with confidence, namely Ser237, Ser263, Ser507, Ser565, Ser566 and Thr596, whereas less clear data indicated that Ser585 might also be phosphorylated under these conditions (Supplementary Table 1 and Figures 1B and 1C).
Some of the phosphorylated sites in TBC1D1 were conserved in the amino acid sequence of AS160; Thr596 on TBC1D1 looked similar to pThr642-AS160 (both potential PKB sites), and Ser507 on TBC1D1 looked similar to pSer570 on AS160 (Figures 1B and 1C). There is sufficient divergence, however, that matches for the other phosphorylated sites could not have been inferred simply by comparing the aligned sequences.
Identification of pSer237 and pThr596 as 14-3-3-binding sites on TBC1D1
HA–TBC1D1 immunoprecipitated from transfected cells was able to bind directly to 14-3-3 proteins in an overlay (Far-Western) assay (Figure 2A). The binding to 14-3-3s was abolished by dephosphorylation of the TBC1D1 with the protein phosphatase PP2A in vitro, and this was prevented by the PP2A inhibitor microcystin-LR (Figure 2A).
Phosphorylation-dependent binding of 14-3-3 to HA–TBC1D1 extracted from transfected HEK-293 cells
HA–T596A-TBC1D1 mutant protein extracted from cells growing in serum still bound to 14-3-3s, albeit to a lesser extent than the wild-type protein, indicating that 14-3-3s bound to both pThr596 and a second 14-3-3-binding site on TBC1D1 (Figure 2B). Binding of 14-3-3s to HA–T596A-TBC1D1 was also abolished by dephosphorylation, showing that the second interaction with 14-3-3s was also phosphorylation-dependent (Figure 2C).
The PAS antibody recognizes proteins that have been phosphorylated within RXRXXpS/T motifs. Both the anti-PAS and 14-3-3-binding signals for wild-type HA–TBC1D1 were increased when the protein was extracted from IGF-1-stimulated cells, compared with unstimulated cells (Figure 2D). The T596A mutation largely abolished binding of the anti-PAS antibody and also blunted the IGF-1-induced 14-3-3 binding (Figures 2B–2D). Previously, we had identified two 14-3-3-binding sites flanking the second PTB domain on AS160 , and we hypothesized that this might also be true for TBC1D1. Of the phosphorylation sites identified (see Supplementary Table 1), pSer237 conformed to the Mode 1 14-3-3-binding consensus (http://scansite.mit.edu/motifscan_seq.phtml and ) and was an obvious candidate for the second 14-3-3-binding site on TBC1D1. Consistent with this possibility, the double mutant HA–S237A/T596A-TBC1D1 was unable to bind to 14-3-3s (Figure 2D).
pThr596 mediates 14-3-3 binding in response to IGF-1, whereas phosphorylation of Ser237 is required for 14-3-3 binding in response to IGF-1/LY294002
The anti-PAS and 14-3-3 overlay signals, and amounts of co-immunoprecipitating endogenous 14-3-3s, were increased when HA–TBC1D1 was extracted from serum-, IGF-1- and calyculin A-stimulated HEK-293 cells, compared with unstimulated (serum-starved) cells (Figure 2E). In contrast, the binding of 14-3-3s to HA–T596A-TBC1D1 was not increased by serum and IGF-1 (Figure 2E).
The trace anti-PAS signal seen in the lanes containing HA–T596A-TBC1D1 could be attributed to co-immunoprecipitation of the endogenous AS160, which was detected by Western blotting (Figure 2E). HA–TBC1D1 and GST–AS160 were also found to co-immunoprecipitate from extracts of doubly transfected cells (see Supplementary Figure 1 at http://www.BiochemJ.org/bj/409/bj4090449add.htm), though whether this represents a physiologically significant interaction is not known.
Consistent with phosphorylation of Thr596 being stimulated via PI3K (phosphoinositide 3-kinase), the IGF-1-induced anti-PAS signal for HA–TBC1D1 was markedly decreased when cells were pre-incubated with the PI3K inhibitor LY294002 (Figure 2E). This inhibitor blocked the IGF-1-induced phosphorylation of Thr308 of PKB, as expected. In contrast, LY294002 did not inhibit the binding of 14-3-3s to wild-type HA–TBC1D1 as determined in the overlay assay, and by co-precipitation of endogenous 14-3-3s with the extracted HA–TBC1D1 (Figure 2E). Moreover, 14-3-3 binding to HA–T596A-TBC1D1 was actually enhanced by treatment of cells with LY294002 (Figure 2E). As reported previously , we found that LY294002 induced phosphorylation of the activating site on AMPK in HEK-293 cells (Figure 2E). Thus these findings suggested the hypothesis that 14-3-3 binding might be enhanced by LY294002-activated phosphorylation of Ser237 by AMPK. Phosphorylated Ser237 is in a site that conforms to the consensus for phosphorylation by AMPK .
To investigate the phosphorylation of Thr596 and Ser237 in more detail, we raised phosphospecific antibodies against these sites. Both the pThr596-TBC1D1 and pSer237-TBC1D1 antibodies specifically recognized the appropriate phosphopeptide immunogens, but not the corresponding dephosphorylated peptides (see Supplementary Figures 2A and 2B at http://www.BiochemJ.org/bj/409/bj4090449add.htm) nor the phosphopeptide corresponding to the other site (results not shown). The pThr596-TBC1D1 antibody showed some cross-reactivity with AS160 extracted from IGF-1-stimulated cells, presumably because pThr642 on AS160 is within a similar sequence (see Supplementary Figure 2E at http://www.BiochemJ.org/bj/409/bj4090449add.htm). It is not a problem to distinguish TBC1D1 and AS160 however, because AS160 runs above TBC1D1 on SDS/PAGE and can be detected with the anti-AS160 antibodies, which do not react with TBC1D1 (Figure 2E and Supplementary Figure 2D at http://www.BiochemJ.org/bj/409/bj4090449add.htm). When tested on HA–TBC1D1 extracted from serum-grown cells, the signal for each phosphospecific antibody was abolished when the appropriate residue was mutated to an alanine residue (Supplementary Figure 2C).
Use of the pThr596 antibody confirmed that phosphorylation of Thr596 in TBC1D1 was induced by serum, IGF-1 and calyculin A, and abolished by LY294002 (Figure 2E). In contrast, the pSer237-TBC1D1 antibody results indicated that phosphorylation of Ser237 is increased in response to LY294002 (Figure 2E), consistent with possible phosphorylation by AMPK and with pSer237 being responsible for LY294002-induced 14-3-3 binding (Figure 2E).
Because of the peculiarity of LY294002 in activating AMPK in HEK-293 cells, we compared its effects with those of a second PI3K inhibitor, wortmannin. In cells stimulated with IGF-1 without inhibitors, the binding of 14-3-3s to HA–TBC1D1 mirrored the phosphorylation of Thr596 and 14-3-3 binding was lost in the T596A mutant (Figure 3, left-hand side). Wortmannin markedly reduced both phosphorylation of Thr596 and binding of 14-3-3s to TBC1D1 (Figure 3, right-hand side). No obvious activation of AMPK was observed with wortmannin (see below). In contrast, the wild-type protein extracted from cells treated with LY294002 displayed strong 14-3-3 binding despite inhibition of Thr596 phosphorylation (Figure 3, middle lanes). Moreover, the binding of 14-3-3s to HA–TBC1D1 from LY294002-treated cells was unaffected by the T596A mutation, but lost in the S237A mutant (Figure 3). Thus pThr596 is the highest affinity 14-3-3-binding site generated in response to IGF-1, and phosphorylation of Thr596 is blocked by both PI3K inhibitors (Figures 2E and 3). In addition, LY294002 induces 14-3-3 binding by promoting phosphorylation of Ser237 (Figures 2E and 3), and AMPK, which is activated by LY294002 in these cells (Figure 2E), became a candidate Ser237-TBC1D1 kinase.
Comparison of the effects of LY294002 and wortmannin on the phosphorylation and 14-3-3 binding of HA–TBC1D1
Regulation of Ser237 and Thr596 phosphorylation and 14-3-3 binding of TBC1D1 in response to different stimuli in HEK-293 cells
Using a wider variety of cellular stimuli, the binding of 14-3-3s to HA–TBC1D1 was enhanced when cells were stimulated with serum, the growth factors IGF-1 and EGF, AMPK activators AICAR and phenformin, PMA, the adenylate cyclase activator forskolin, and the calcium ionophore A23187 (Figure 4). The strongest 14-3-3-binding signals were seen with HA–TBC1D1 from cells stimulated by IGF-1 and phenformin. The 14-3-3 overlay signal for extracted HA–TBC1D1 mirrored the results for co-immunoprecipitation of endogenous 14-3-3s (Figure 4).
Effects of different cellular stimuli on the multisite phosphorylation and 14-3-3 binding of HA–TBC1D1
The various stimuli induced distinct patterns of phosphorylation of Ser237 and Thr596 of HA–TBC1D1. The phosphorylation of Thr596 was increased in response to serum, IGF-1, EGF, PMA and forskolin. In contrast, phenformin and A23187 promoted a stronger phosphorylation of Ser237, compared with Thr596 (Figure 4 and results not shown). These findings are consistent with the possibility that AMPK phosphorylates Ser237. These results also indicate that binding of 14-3-3s to HA–TBC1D1 may depend on either pSer237 and/or pThr596 to different extents in cells stimulated in different ways.
Mutation of Ser237 and Thr596 has differential effects on 14-3-3 binding of TBC1D1 in response to different stimuli in HEK-293 cells
HA–TBC1D1 with mutated Ser237 and/or Thr596 was tested for its binding to 14-3-3s after extraction from cells exposed to the various stimuli (Figure 5). A general observation is that the double mutant HA–S237A/T596A-TBC1D1 did not co-precipitate with endogenous 14-3-3s in response to any of the stimuli tested (Figures 5A–5G). Similarly, the HA–S237A/T596A-TBC1D1 mutant extracted from the various stimulated cells was unable to bind to 14-3-3s in the overlay assay (results not shown).
Relative contributions of pSer237 and pThr596 to 14-3-3 binding of HA–TBC1D1 in cells stimulated in different ways
The analysis of the mutants indicated that basal binding of 14-3-3s to HA–TBC1D1 in the unstimulated (serum-starved) cells is primarily attributable to basal phosphorylation of Ser237, with a trace of Thr596 phosphorylation (Figures 4 and 5A). Similar to the results seen with IGF-1 (Figure 3), the binding of 14-3-3s to HA–TBC1D1 extracted from cells stimulated with serum, EGF and PMA was most severely abrogated by mutation of Thr596, whereas the S237A mutation had a lesser effect (Figures 5B–5D). These results are consistent with Thr596 phosphorylation being induced by EGF and PMA (Figure 4). For HA–TBC1D1 from forskolin-treated cells, the binding of 14-3-3s was partially reduced by mutation of either Ser237 or Thr596 (Figure 5E). In contrast, the interaction between 14-3-3s and HA–TBC1D1 from phenformin-stimulated cells was most strongly inhibited by mutation of Ser237 (Figure 5F). This result is similar to that seen with IGF-1/LY294002 (Figures 2E and 3), and is consistent with AMPK being the Ser237 kinase. Binding of 14-3-3s to TBC1D1 extracted from cells treated with the calcium ionophore A23187 was also dependent on Ser237 (Figure 5G and results not shown).
Effect of inhibitors on TBC1D1 phosphorylation and 14-3-3 binding TBC1D1 in response to different stimuli in HEK-293 cells
Consistent with phosphorylation by AMPK, the AMPK inhibitor Compound C  partially inhibited the phosphorylation of Ser237 and 14-3-3 binding in response to phenformin (Figure 6A, last two lanes), and Compound C also partially inhibited the lesser responses to metformin and AICAR (Figure 6A). Wortmannin inhibited both the IGF-1- and EGF-induced phosphorylation of Thr596 (Figure 6B). In contrast, the RSK-specific inhibitor BI-D1870  and non-specific PKC (protein kinase C) inhibitor Go6983 prevented the PMA-induced phosphorylation of Thr596, consistent with the PKC/ERK/RSK pathway mediating the regulation of this site in response to PMA. In cells stimulated with EGF and PMA, BI-D1870 caused a slight increase in phosphorylation of Ser237 of TBC1D1 concomitant with a stimulation of AMPK and ACC phosphorylation (Figure 6B and results not shown).
Effects of protein kinase inhibitors on the multisite phosphorylation and 14-3-3 binding of TBC1D1 extracted from cells stimulated in different ways
In vitro phosphorylation of GST–TBC1D1 by PKB and AMPK
Purified bacterially expressed GST–TBC1D1 comprised full- length protein and truncated proteins that were missing various sized portions from the C-terminus (Figure 7A). Native and recombinant AMPK phosphorylated several residues on GST–TBC1D1 (Ser237, Ser487 or Thr489, Ser503 or Thr505, Ser558 or Ser559, Thr596), as determined using phosphospecific pSer237-TBC1D1 and pThr596-TBC1D1 antibodies (Figure 7) and mass spectral analysis of tryptic digests (results not shown). The AMPK-phosphorylated forms of GST–TBC1D1 that were >50 kDa could bind strongly to 14-3-3s, including fragments that were too small to contain pThr596 (Figure 7D). We also noticed that AMPK induced a slight upwards band-shift in the GST–TBC1D1 fragment that runs at ∼110 kDa (most obvious in Figure 7C).
In vitro phosphorylation of GST–TBC1D1 by PKB and AMPK
In contrast, PKB phosphorylated Ser235 (mass spectral data; results not shown) and Thr596 (Figure 7C), but did not phosphorylate Ser237 (Figure 7B), and did not induce the GST–TBC1D1 to bind to 14-3-3s (Figure 7D). Moreover, prior phosphorylation with PKB inhibited subsequent AMPK-induced phosphorylation of Ser237 (Figure 7B) and binding of GST–TBC1D1 to 14-3-3s (Figure 7D), and PKB also prevented the AMPK-induced upwards band-shift in the ∼110 kDa protein (Figure 7C). In contrast, GST–TBC1D1 that was phosphorylated first by AMPK and then by PKB could bind to 14-3-3s (Figure 7D), and displayed the upwards band-shift of the ∼110 kDa fragment (Figure 7C).
The in vitro phosphorylations were repeated with GST–TBC1D1-S235A mutant protein. In contrast with the wild-type protein, prior phosphorylation of the S235A-TBC1D1 mutant with PKB did not prevent subsequent AMPK-mediated Ser237 phosphorylation and 14-3-3 binding (see Supplementary Figure 3 at http://www.BiochemJ.org/bj/409/bj4090449add.htm). These results indicated that in vitro phosphorylation of Ser235 by PKB was responsible for preventing the AMPK-induced phosphorylation of Ser237 and/or AMPK-induced binding of wild-type GST–TBC1D1 to 14-3-3s.
Complementary phosphorylation and 14-3-3-binding of endogenous TBC1D1 and AS160 in response to insulin and AMPK activators in rat L6 myotubes
Using an antibody raised against the C-terminal 529 amino acids of the protein, sufficient endogenous TBC1D1 for analysis could be immunoprecipitated from rat L6 myotubes (Figure 8A), but not from HEK-293 cells or adipocytes. From a Coomassie-Blue-stained SDS-gel of the immunoprecipitate, four protein bands of ∼140–150 kDa were identified as forms of TBC1D1 (results not shown). These forms of TBC1D1 corresponded to the four bands seen in the top four panels of Figure 8(A). In addition, a protein of 110 kDa was identified as TBC1D2, another GAP protein that is likely to cross-react with the anti-TBC1D1 antibody (results not shown). When the myotubes were stimulated with insulin, the activating Thr308 of PKB and Thr596 of TBC1D1 were phosphorylated (Figure 8A), and both phosphorylations were inhibited by wortmannin. However, insulin did not promote any obvious binding of 14-3-3s to the TBC1D1 (Figure 8A). In contrast, phenformin promoted a robust activation of AMPK, phosphorylation of Ser237 of TBC1D1, 14-3-3-overlay signal and co-precipitation of 14-3-3s (Figure 8A). A-769662, a recently described small-molecule AMPK activator [22,23], also promoted a strong phosphorylation of Ser237 of the endogenous TBC1D1, together with strong binding of 14-3-3s and co-precipitation of 14-3-3s with the isolated TBC1D1 (Figure 8A).
Phosphorylation, 14-3-3 binding and subcellular localization of endogenous TBC1D1 isolated from rat L6 myotubes
Co-localization of endogenous TBC1D1 and GLUT4 in rat L6 myotubes
In immunocytochemistry experiments with unstimulated rat L6 myotubes, the anti-TBC1D1 and anti-GLUT4 antibodies both gave signals that were associated with perinuclear structures (Figure 8C). With the caveat that the anti-TBC1D1 antibody also recognized TBC1D2, these findings are consistent with TBC1D1 and GLUT4 being located near to each other. When cells were stimulated with either insulin or A-769662, GLUT4 staining was seen at the cell surface in a proportion (∼10–15%) of cells. Presumably, the other cells were not sufficiently differentiated to be fully responsive to these stimuli. It was of note, however, that cells with cell-surface staining for anti-GLUT4 also displayed more prominent cell-surface localization for anti-TBC1D1, suggesting that in these cells TBC1D1 accompanied GLUT4 during its well-characterized translocation to the plasma membrane (Figure 8C).
Phylogenetic analysis suggests that AS160 evolved with the vertebrates, whereas TBC1D1 is more ancient, being found in all animals examined (Figure 1A). We speculate that AS160 arose to provide more versatile upstream regulation of the Rabs that may be common substrates of these RabGAPs ; the GAP domains are conserved, whereas AS160 and TBC1D1 sequences diverge most in the N-terminal tail and within the two distinct clusters of phosphorylated sites that are located either side of the second PTB domain in both AS160 and TBC1D1 (Figures 1B and 1C). Notable exceptions are the identical short sequences surrounding the phosphorylatable Thr596 in TBC1D1, which corresponds to Thr642 in AS160; and Ser507 in TBC1D1, which is equivalent to Ser570 in AS160 (Figure 1B). Thr568, which can be phosphorylated by SGK (serum- and glucocorticoid-induced protein kinase) in AS160 , also finds a match in Ser505 of TBC1D1, though whether Ser505 can be phosphorylated is unknown.
14-3-3s pick out specific binding sites within each phosphorylated cluster, and the simplest hypothesis is that a 14-3-3 dimer straddles the PTB2 domain by binding to pSer237 and pThr596 in TBC1D1, and pSer341 and pThr642 in AS160 (Figure 1B). There are, however, clear differences between TBC1D1 and AS160 with respect to the signalling pathways that target these sites and the relative importance of each site for 14-3-3 binding.
For AS160, pThr642 is the high-affinity site for binding to 14-3-3s in response to IGF-1, EGF and PMA in HEK-293 cells [10,11], whereas pSer341 is a lower affinity 14-3-3-binding site . Consistent with previous findings , insulin stimulated phosphorylation of Ser341 and Thr642 and binding of 14-3-3s to AS160 in L6 cells, but AMPK activators have little effect (Figure 8B).
In contrast, it appears that either pSer237 or pThr596 of TBC1D1 can take the lead in mediating 14-3-3 binding in response to different stimuli. 14-3-3 binding to TBC1D1 is dominated by phosphorylation of Ser237 in HEK-293 cells containing active AMPK, whereas Thr596 is more important for 14-3-3 binding in response to IGF-1, EGF and PMA (Figures 3–5). In contrast with AS160, insulin stimulated the phosphorylation of Thr596 of TBC1D1, but did not induce phosphorylation of Ser237 or binding to 14-3-3s in L6 cells. Thus there is a complementarity in that insulin promotes binding of 14-3-3s to AS160, whereas AMPK activators promote binding of 14-3-3s to TBC1D1 in L6 cells.
It is not clear why IGF-1 can promote 14-3-3 binding to TBC1D1 via pThr596 in HEK-293 cells, and yet insulin stimulated the phosphorylation of Thr596 of TBC1D1, but not its 14-3-3 binding, in L6 cells. One possibility is that the phosphorylation status of other residues in the clusters have positive or negative effects on the phosphorylation and/or binding to 14-3-3s. We found, for example, that 14-3-3s cannot bind to TBC1D1 that has been phosphorylated in vitro by PKB, despite this kinase promoting robust phosphorylation of Thr596 (Figure 7). Moreover, both the pSer237 and 14-3-3-binding signals of AMPK-phosphorylated TBC1D1 were markedly inhibited by prior phosphorylation by PKB. Our findings suggest that in vitro phosphorylation of Ser235 by PKB prevents subsequent phosphorylation of Ser237 by AMPK. We do not yet know whether Ser235 is a physiological phosphorylation site, nor do we understand why phosphorylation of Thr596 by PKB does not allow binding of 14-3-3s, even in the form of TBC1D1 where Ser235 is mutated to an alanine residue (Supplementary Figure 3). These results indicate that there is much to learn about the hierarchies and regulatory interactions among the sites of multisite phosphorylation of TBC1D1.
Another finding is that combinations of IGF-1/LY294002, EGF/BI-D1870 and PMA/BI-D1870 promote phosphorylation of the activating Thr172 of AMPK and hence phosphorylation of the AMPK substrate ACC (Figures 4 and 6). To our knowledge, these effects have not been reported previously, though inhibitors of the MAPK (mitogen-activated protein kinase) pathway  and human EGF receptor 2 in combination with EGF receptor tyrosine kinase inhibitors  have been reported to activate AMPK. We have made similar observations when studying another protein that is targeted for multisite phosphorylation, and the mechanisms of AMPK activation by these inhibitors are under investigation in HEK-293 and other cell types (J. Murphy and C. MacKintosh, unpublished work). Another AMPK-related puzzle is why phenformin does not promote phosphorylation of Thr596 of TBC1D1 (Figures 4 and 8A), whereas AMPK can phosphorylate this residue in vitro (Figure 7 and Supplementary Figure 2).
Finding that TBC1D1 and GLUT4 co-localize with each other in unstimulated, and insulin- and A-769662-responsive rat L6 myotubes (Figure 8C) is consistent with a role for this RabGAP in GLUT4 trafficking. The relative roles of TBC1D1 and AS160 are not yet known, but finding that they are subject to complementary regulation suggests that these two proteins may co-operate to mediate responsiveness of glucose uptake to insulin, the anti-diabetic drug metformin, and exercise. The effects of forskolin on AS160  and TBC1D1 (Figure 4) suggest that it might also be worth revisiting the question of whether adrenaline acting via PKA can influence glucose uptake into skeletal muscle . Whether the TBC1D1 phosphorylation in response to the calcium ionophore A23187 is mediated via AMPK or by direct phosphorylation by a calcium-activated protein kinase is not yet known.
AS160 and TBC1D1 have a wider cellular and tissue distribution than GLUT4 however, which means that GSVs are unlikely to be the only vesicles whose trafficking is regulated by these RabGAPs. For example, PKB is implicated in trafficking of CD89 to class II-containing vesicles and class II Ag presentation , and it is possible that AS160 and TBC1D1 may be PKB targets in this and/or other processes. The antibodies generated in the present study should be useful for defining the relative roles and regulation of TBC1D1 and AS160 in different cells and tissues.
This work was supported by Diabetes UK, the U.K. Medical Research Council, the U.K. Biotechnology and Biological Sciences Research Council (grant BB/C511613/1 for MS), and the companies who support the DSTT (Division of Signal Transduction Therapy) at the University of Dundee, namely AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Merck and Co, Merck KGaA and Pfizer. We thank Natalia Shpiro (MRC Protein Phosphorylation Unit, University of Dundee, Dundee, Scotland, U.K.) for synthesis of A-769662. Thanks to Claire Balfour for tissue culture support, the DSTT antibody and protein production team co-ordinated by James Hastie for purification of antibodies and bacterially expressed GST-fusion protein, and Rachel Naismith for secretarial assistance.
(AMP-activated protein kinase
Akt substrate of 160 kDa (also known as TBC1D4)
epidermal growth factor
extracellular-signal-regulated protein kinase 1/2
glucose transporter 4
GLUT4 storage vesicle
human embryonic kidney
insulin-like growth factor-1
protein kinase B (also known as Akt)
protein kinase C
p90 ribosomal S6 kinase