AKT and AMP-activated protein kinase regulate TBC1D1 through phosphorylation and its interaction with the cytosolic tail of insulin-regulated aminopeptidase IRAP

In skeletal muscle, the Rab GTPase-activating (GAP) protein TBC1D1 is phosphorylated by AKT and AMP-activated protein kinase (AMPK) in response to insulin and muscle contraction. Genetic ablation of Tbc1d1 or mutation of distinct phosphorylation sites impairs intracellular GLUT4 retention and GLUT4 traffic, presumably through alterations of the activation state of downstream Rab GTPases. Previous studies have focused on characterizing the C-terminal GAP domain of TBC1D1 that lacks the known phosphorylation sites, as well as putative regulatory domains. As a result, it has been unclear how phosphorylation of TBC1D1 would regulate its activity. In the present study, we have expressed, purified, and characterized recombinant full-length TBC1D1 in Sf9 insect cells via the baculovirus system. Full-length TBC1D1 showed RabGAP activity toward GLUT4-associated Rab8a, Rab10, and Rab14, indicating similar substrate specificity as the truncated GAP domain. However, the catalytic activity of the full-length TBC1D1 was markedly higher than that of the GAP domain. Although in vitro phosphorylation of TBC1D1 by AKT or AMPK increased 14-3-3 binding, it did not alter the intrinsic RabGAP activity. However, we found that TBC1D1 interacts through its N-terminal PTB domains with the cytoplasmic domain of the insulin-regulated aminopeptidase, a resident protein of GLUT4 storage vesicles, and this binding is disrupted by phosphorylation of TBC1D1 by AKT or AMPK. In summary, our findings suggest that other regions outside the GAP domain may contribute to the catalytic activity of TBC1D1. Moreover, our data indicate that recruitment of TBC1D1 to GLUT4-containing vesicles and not its GAP activity is regulated by insulin and contraction-mediated phosphorylation.

TBC1D1 and TBC1D4 (Tre-2/Bub2/Cdc16 domain family member 1 and 4) 2 are Rab-GTPase-activating proteins (Rab-GAPs) involved in insulin signaling, GLUT4 translocation (1)(2)(3), and therefore may play roles in obesity and type 2 diabetes (4 -6). Both TBC1D1 and TBC1D4 contain two N-terminal phosphotyrosine-binding (PTB) domains, a calmodulin-binding domain (CBD), and a C-terminal catalytic RabGAP domain (7)(8)(9). The RabGAP domain promotes GTP hydrolysis of active Rabs leading to the conversion of Rabs into their inactive GDPbound form (9). The arginine residue located in the GAP domain (Arg 854 in mouse TBC1D1, Arg 973 in mouse TBC1D4), previously designated "arginine finger," is required for the full catalytic activity of the RabGAPs and substitution of this arginine residue resulted in loss of the RabGAP activity (9 -11). TBC1D1 and TBC1D4 are similar and share over 79% sequence homology in their RabGAP domains. Both RabGAP domains display Rab substrate specificity toward Rab8a, Rab10, and Rab14 in vitro and are involved in GLUT4 trafficking in vivo (9,(12)(13)(14). TBC1D1 is a downstream target for AKT and AMPK (AMP-activated protein kinase) and it has been suggested that phosphorylation of TBC1D1 on Thr 590 by AKT in response to insulin and Ser 231 by AMPK in response to muscle contraction, and subsequent 14-3-3 binding is essential for GLUT4 translocation (9,(15)(16)(17)(18). Moreover, mutation of four phosphorylation sites in the related TBC1D4 (4P) led to inhibition of insulinstimulated GLUT4 translocation in vivo (17,19). Interestingly, a naturally occurring mutation of an arginine residue (R125W) located in the first PTB domain of TBC1D1 has been linked to familial obesity in humans and impaired glucose transport in mice through an unclear mechanism (5,6,20). Recent studies have investigated the structure and function of the C-terminal GAP domains of TBC1D1 and TBC1D4 in vitro (13,17,21,22). Although the overall three-dimensional structure of the GAP domains were similar to that of the previously described yeast homolog Gyp1p (11,23), large differences were described for their catalytic activities and interfaces potentially involved in substrate recognition and enzymatic specificity (11,22). Moreover, as these studies focused on the GAP domains expressed in Escherichia coli, which were lacking the known phosphorylation sites, as well as all other putative regulatory domains, it has been not possible to study how phosphorylation would regulate the GAP activity of the enzyme. In fact, several studies have provided evidence that TBC1D4 might be regulated through phosphorylation of the PTB domains and alterations in protein-protein interactions (24,25). In the present study, we have for the first time expressed, purified, and characterized fulllength TBC1D1 in vitro. Our results indicate that phosphorylation may regulate subcellular targeting of TBC1D1 rather than altering the GAP activity of the enzyme.

Expression and purification of His 6 -TBC1D1
We generated a baculovirus coding for the long isoform of the murine TBC1D1 that is predominantly expressed in skeletal muscle (4). The construct included an N-terminal His 6 tag, two N-terminal PTB domains, a CBD, and the C-terminal GTPaseactivating (Tre-2/Bub2/Cdc16-or GAP)-domain (Fig. 1A). We generated a mutant, TBC1D1-R854K, that lacks the arginine finger required for full catalytic activity (11,23) in addition to a mutation in the first PTB domain (R128W). Baculovirus-in-fected Sf9 cells were cultured, harvested, and the recombinant proteins were purified by immobilized metal affinity chromatography (IMAC) using Ni-NTA resin as described under "Experimental procedures." Typically, the purity of the protein was ϳ35% as determined by densitometry of Coomassiestained gels, and the yield after IMAC was approximately 0.1 to 0.3 mg/10 8 Sf9 cells. The purified protein showed an apparent molecular mass of ϳ180 kDa in SDS-PAGE (Fig. 1B). However, using size exclusion chromatography (SEC), the apparent molecular mass of the purified protein corresponded to ϳ600 kDa (Fig. 1C). Interestingly, a minor proportion (ϳ5%) of the protein eluted in an SEC fraction corresponding to ϳ180 kDa (Fig. 1C, inset). In our study, we used affinity-purified fulllength TBC1D1 that contained both, the high M r and the low M r forms.

GAP activity of full-length TBC1D1
To further characterize the enzyme, we investigated the GAP activity of TBC1D1 toward Rab10, an apparent natural substrate for the enzyme (9). As the production of GST-Rab10 in E. coli had low yield, we also generated a baculovirus to express and purify sufficient amounts of GST-tagged mouse Rab10 in Sf9 cells. Purified GST-Rab10 was loaded with [␥-32 P]GTP and then incubated either alone, or with full-length TBC1D1, or the GAP mutant (R854K) of TBC1D1 as described under "Experimental procedures." To determine the GTPase activity, we mea- Figure 1. Domain structure of mouse TBC1D1, expression and apparent molecular mass. A, domain structure of the 1255-amino acid isoform of the murine TBC1D1 (predicted molecular mass 147 kDa). Annotated domains include two N-terminal protein-tyrosine binding (PTBD) domains, a CBD domain, and the C-terminal GTPase-activating (Tre-2/Bub2/Cdc16-or GAP) domain. Reported phosphorylation sites for AKT and AMPK, Ser 231 , Thr 499 , Thr 590 , and Ser 621 (corresponding to human sites Ser 237 , Thr 505 , Thr 596 , and Thr 627 ) are indicated. Arrows show the position of the mutated residues, R125W and R854K. The dotted lines indicate the coordinates for the truncated GAP and PTB domains expressed as GST fusion proteins. B, 8% Coomassie-stained SDS-PAGE of purified wildtype (WT) TBC1D1 and TBC1D1-R854K mutant after IMAC using Ni-NTA resin. C, elution profile of SEC with Ni-NTA-purified TBC1D1 on a SEC650 column. Inset: Western blotting using antibodies against TBC1D1 of selected SEC fractions.

Regulation of recombinant TBC1D1 in vitro
sured the rate of production of [ 32 P]phosphate from the hydrolysis of [␥-32 P]GTP to GDP. As shown in Fig. 2A, purified GST-Rab10 exhibited some endogenous GTPase activity. GTP hydrolysis of GST-Rab10 was substantially increased by TBC1D1, but much less by the R854K mutant. We then compared the relative activities of the truncated GST-GAP domain and the full-length TBC1D1 toward Rab10. The GAPs were incubated with increasing concentrations of [␥-32 P]GTPloaded GST-Rab10 and the amount of released [ 32 P]phosphate was determined. Compared with the truncated GAP domain, the full-length protein displayed markedly higher activity toward Rab10 as substrate (Fig. 2B). Using nonlinear fitting to Michaelis-Menten kinetics, we estimated a moderate decrease in apparent K m (30.4 M) of the full-length TBC1D1 compared with the GST-tagged GAP domain (60.5 M) for GST-Rab10 as substrate, and a ϳ10-fold increase in apparent V max for the full-length TBC1D1 ( Fig. 2B and Fig. S1). We and others have shown previously that both Rab8a and Rab14 are substrates for the truncated GAP domain of TBC1D1 in vitro (9,26). We therefore compared the GAP activities toward the different GTP-loaded Rabs. As illustrated in Fig. 2C, the full-length GAP exhibited similar Rab substrate specificities as shown previously for truncated GAP domain (13,17).

Mapping of AKT2-and AMPK-dependent phosphosites in TBC1D1
Previous studies have demonstrated in vitro and in vivo phosphorylation of at least six Ser/Thr residues in TBC1D1, including mouse positions Ser 231 , Ser 499 , Thr 590 , Ser 621 , Ser 660 , and Ser 700 (7,8,15,20,27). In particular, Ser 231 and Thr 590 (human Ser 237 and Thr 596 ) were predicted as potential phosphorylation sites for AKT and AMPK in response to stimulation with insulin and muscle contraction (7,20). We phosphorylated affinity purified TBC1D1 in vitro with purified kinases, AKT2 or AMPK, and mapped Ser/Thr phosphorylation sites by MS and Western Blots using phospho-TBC1D1 antibodies as described under "Experimental procedures." As shown in Table 1, we identified Ser 489 and Ser 590 as direct targets for AKT2, and Ser 231 , Thr 404 , Thr 590 , Ser 660 , and Ser 700 as direct targets for AMPK in vitro.

Phosphorylation of TBC1D1 by AKT2 or AMPK does not alter GAP activity
Insulin, AICAR, and contraction-mediated phosphorylation of Ser/Thr residues in TBC1D1 and TBC1D4 has been implicated to regulate the activation state of downstream Rab GTPases, including Rab8, Rab10, and Rab14 (9,13). We therefore investigated the direct impact of AKT2 and AMPK-mediated phosphorylation on the GAP activity of TBC1D1 using Rab10 as substrate. IMAC purified TBC1D1 was phosphorylated by either AKT2 or AMPK for 30 min, then incubated with purified, [␥-32 P]GTP-loaded Rab10 and the amount of released [ 32 P]phosphate was determined as described under "Experimental procedures." As determined by Western blots using phospho-TBC1D1 antibodies, AKT2 and AMPK substantially increased phosphorylation of Thr 590 , whereas Ser 231 was phosphorylated by AMPK and not AKT2 (Fig. 3, A and B). However, we did not observe any relevant reduction in the GTP-hydrolysis rate of Rab10 when exposed to phosphorylated TBC1D1 (Fig. 3C). Similar results were obtained with TBC1D1 immunoprecipitated from insulin-or AICAR-stimulated cells (Fig. S3). Moreover, phosphorylation of TBC1D1 by either AKT2 or AMPK in vitro had no effect on the apparent M r of the protein as determined by SEC (data not shown). We determined the extent of phosphorylation by pulldown assays to be typically about 70% of the total TBC1D1 (Fig. S4).

Phosphorylation of TBC1D1 by AKT2 or AMPK increases 14-3-3 binding but does not alter the GAP activity
Phosphorylation of TBC1D1 in response to insulin stimulation or AICAR has been shown to increase binding of 14-3-3 proteins to the RabGAP. To test whether phosphorylation-induced 14-3-3 binding alters the GAP activity, TBC1D1 was phosphorylated by AKT2 or AMPK, and association of GST-14-3-3 was determined by GST-pulldown assay as described under "Experimental procedures." As shown in Fig. 4, A-C, phosphorylation substantially increased binding of 14-3-3 to TBC1D1. However, phosphorylation of TBC1D1 by AKT2 or AMPK in the presence of 14-3-3 did not alter the GAP activity toward Rab10 (Fig. 4D). Likewise, addition of Ca 2ϩ /calmodulin, which had been shown to bind to the were performed with Rab10 in the presence of full-length WT TBC1D1 and R854K mutant. Affinity-purified GST-Rab10 (0.6 -1 pmol) loaded with [␥-32 P]GTP was incubated in the absence or presence of 2 pmol of purified TBC1D1 or inactive TBC1D1-R941K mutant as described under "Experimental procedures." After 10 min at 30°C, aliquots were filtrated through activated charcoal and radioactive [ 32 P]phosphate was determined by scintillation counting. Data represent mean Ϯ S.E. from three independent experiments. ***, p Ͻ 0.001, Rab alone versus TBC1D1 and TBC1D1-R854K (two-way ANOVA). B, purified truncated GAP domain (GST-GAP) and full-length TBC1D1 were incubated with increasing concentrations of [␥-32 P]GTP-loaded GST-Rab10 for 10 min and the amount of released [ 32 P]phosphate was determined. Phosphate production resulting from the endogenous GTP hydrolysis activity of Rab10 was subtracted. C, purified [␥-32 P]GTP-loaded GST-Rabs (0.6 -1 pmol) were incubated in the absence and presence of 2 pmol of purified full-length TBC1D1 for 30 min and the amount of released [ 32 P]phosphate was determined. Data represent mean Ϯ S.E. from three independent experiments. ***, p Ͻ 0.001, Rab alone versus GAP (two-way ANOVA).

Regulation of recombinant TBC1D1 in vitro
related TBC1D4 (28) did not alter the GAP activity of TBC1D1 (Fig. S2).

R125W mutation does not alter phosphorylation, 14-3-3 binding, or GAP activity of TBC1D1
The R125W mutation maps in the first PTB domain of TBC1D1 and has been associated with familial obesity in humans (5,6). To investigate whether and how this mutation may affect TBC1D1 function, we generated a corresponding baculovirus, and expressed and purified the protein from Sf9 cells. Expression levels, yield, purity, and apparent M r in size exclusion chromatography of TBC1D1-R125W were not different from WT TBC1D1 (not shown). Purified TBC1D1-R125W was incubated with [␥-32 P]GTP-loaded GST-Rab10 and the amount of released [ 32 P]phosphate was determined as described under "Experimental procedures." No difference in GAP activity toward Rab10 was observed between the R125W mutant and WT TBC1D1 (Fig. 5A). Also, addition of a 20-fold excess of a GST fusion protein with the two N-terminal PTBD domains of TBC1D1 did not alter the GAP activity of neither WT nor the R125W mutant protein (Fig. 5A). Next, purified TBC1D1-R125W was incubated with AKT2 and AMPK as described under "Experimental procedures," and phosphorylation of Thr 590 was determined by Western blots using phospho-TBC1D1 antibodies. No difference in phosphorylation of R125W mutant and WT TBC1D1 was observed (Fig. 5B). Also, 14-3-3 binding to the phosphorylated R125W mutant was not altered (not shown). Last, TBC1D1-R125W was phosphorylated by either AKT2 or AMPK for 30 min, then incubated with purified, [␥-32 P]GTP-loaded Rab10 and the amount of released [ 32 P]phosphate was determined as described under "Experimental procedures." The nonphosphorylated and phosphorylated TBC1D1-R125W exhibited the same GAP activity toward Rab10 as the WT protein (Fig. 5C).

Phosphorylation of TBC1D1 regulates IRAP interaction
The insulin-regulated aminopeptidase IRAP has been identified previously as a resident membrane protein in GLUT4 vesicles, highly co-localizing with the glucose transporter. To explore a possible interaction of IRAP with TBC1D1, we conducted pulldown and co-immunoprecipitation experiments in vitro. HEK293 cells coexpressing HA-tagged IRAP and FLAGtagged TBC1D1 were lysed and the proteins were phosphorylated by adding purified AMPK or AKT2 and Mg 2ϩ ATP. As shown in Fig. 6B, immunoprecipitation of IRAP resulted in coprecipitation of TBC1D1 in the absence but not presence of AMPK/AKT2. To further map the TBC1D1-binding domain in IRAP, we fused the entire 110-amino acid cytoplasmic C-terminal tail of IRAP to the C terminus of GST to yield GST-cIRAP. Sf9 cells expressing TBC1D1 were lysed and the proteins were phosphorylated by adding purified AKT2, AMPK, and Mg 2ϩ -ATP as described under "Experimental procedures." Pulldown of GST-cIRAP co-precipitated TBC1D1 in the absence but not presence of either AKT2 or AMPK, respectively (Fig. 6C). To map the IRAP-binding domain in TBC1D1, we fused the two PTB domains of TBC1D1 (amino acids 1-132) to the C terminus of GST to yield GST-PTBD. Binding of FLAG-tagged TBC1D1 to GST-cIRAP was abolished by excess  A and B, 2 pmol of affinity-purified TBC1D1 was phosphorylated in vitro with purified AKT2 or AMPK as described under "Experimental procedures." Aliquots were analyzed for TBC1D1 phosphorylation by Western blotting using phospho-specific antibodies for Thr 590 and Ser 231 . C, nonphosphorylated TBC1D1, AKT2-phosphorylated, and AMPK-phosphorylated TBC1D1 were incubated with purified [␥-32 P]GTP-loaded GST-Rab10 (0.6 -1 pmol) for 30 min and the amount of released [ 32 P]phosphate was determined as described under "Experimental procedures." Phosphate production resulting from the endogenous GTP hydrolysis activity of Rab10 was subtracted. Data represent mean Ϯ S.E. from three independent experiments. *, p ϭ 0.0350 phosphorylated versus nonphosphorylated TBC1D1 (two-tailed unpaired t test).

Regulation of recombinant TBC1D1 in vitro
GST-PTB but not excess GST (Fig. 6D). Similar to the WT TBC1D1, GST-cIRAP co-precipitated TBC1D1-R125W in the absence but not presence of the kinases (Fig. 6E). Moreover, addition of excess GST-cIRAP did not affect the GAP activity of TBC1D1 toward Rab10 (Fig. 6F).

Discussion
Insulin-and contraction-mediated Ser/Thr phosphorylation of the ϳ150 -160-kDa RabGAPs TBC1D1 and TBC1D4 by AKT and AMPK has been associated with alterations in the activation state of downstream Rab GTPases, suggesting that the GAP activity might be directly or indirectly regulated by phosphorylation (7,17,29). However, previous enzymatic studies on both structure and function of the GAP domains have been conducted mostly with C-terminal-truncated fragments of the RabGAP proteins that lack the relevant phosphorylation sites, as well as other putative regulatory elements. As a result, the direct impact of RabGAP phosphorylation on the GAP  A, RabGAP assays were performed with Rab10 in the presence of full-length WT TBC1D1 or TBC1D1 R125W mutant. Affinity-purified GST-Rab10 (0.6 -1 pmol) loaded with [␥-32 P]GTP was incubated in the presence of 2 pmol of purified TBC1D1, and TBC1D1-R125W with or without 10-fold excess of GST-PTBD as described under "Experimental procedures." After 30 min at 30°C, aliquots were filtrated through activated charcoal and radioactive [ 32 P]phosphate was determined by scintillation counting. Phosphate production resulting from the endogenous GTP hydrolysis activity of Rab10 was subtracted. Data represent mean Ϯ S.E. from three independent experiments. B, representative Western blots of phosphorylation reactions. 2 pmol of affinity-purified TBC1D1-R125W was phosphorylated in vitro with purified AKT2 or AMPK as described under "Experimental procedures." Aliquots were analyzed for TBC1D1 phosphorylation by Western blotting using phosphospecific antibodies for Thr 590 . C, nonphosphorylated TBC1D1-R125W, AKT2-phosphorylated and AMPK-phosphorylated TBC1D1-R125W were incubated with purified [␥-32 P]GTP-loaded GST-Rab10 (0.6 -1 pmol) for 30 min and the amount of released [ 32 P]phosphate was determined as described under "Experimental procedures." Phosphate production resulting from the endogenous GTP hydrolysis activity of Rab10 was subtracted. Data represent mean Ϯ S.E. from three independent experiments.

Regulation of recombinant TBC1D1 in vitro
activity has been unclear. In the present study, we have for the first time purified and characterized TBC1D1 as a full-length RabGAP from the TBC1D family.
TBC1D1 was affinity purified from baculovirus-transduced Sf9 cells using the N-terminal His tag. Interestingly, the purified protein appeared to be mostly in an oligomeric state as indicated by an apparent M r in size exclusion chromatography about 3 times higher than predicted from the amino acid sequence. This is consistent with a previous study that showed dimerization of the related TBC1D4 GAP domain through a ϳ100 amino acid C-ter-minal region predicted to adopt a coiled-coil motif (21). A similar C-terminal sequence is also found in TBC1D1, which might contribute to oligomerization of the protein. However, oligomer formation has not been found to have an impact on the GAP activity of the truncated domain (21), implicating other possible functions except involvement in catalysis. Further experiments are required to compare the enzymatic properties of the high M r form and the low M r form of full-length TBC1D1.
Previous studies have investigated the activity and substrate specificity of the truncated GAP domain of TBC1D1 (9). In vitro Cleared lysates from HEK293 cells coexpressing full-length HA-IRAP and FLAG-TBC1D1 were incubated in the absence and presence of AMPK for 20 min and co-precipitations were carried out using magnetic beads coated with anti-HA monoclonal antibodies as described under "Experimental procedures." Samples were separated by SDS-PAGE and co-precipitated FLAG-TBC1D1 was analyzed by Western blotting. C, GST-pulldown of the cytoplasmic tail of IRAP and TBC1D1. GST-cIRAP was immobilized on GSH-Sepharose beads and incubated with 5 pmol of purified nonphosphorylated, AKT2-phosphorylated, or AMPKphosphorylated TBC1D1 from Sf9 cells for 1 h at 4°C as described under "Experimental procedures." Immobilized GST was used as a negative control. Eluted samples were separated by SDS-PAGE and analyzed by Western blotting using TBC1D1 antibodies and phospho-TBC1D1 antibodies against Thr 590 and Ser 231 . Quantification of GST-cIRAP was phosphorylated TBC1D1. Data represent mean Ϯ S.E. from three independent experiments. ***, p Ͻ 0.001 TBC1D1 versus TBC1D1 ϩ GST-cIRAP (two-way ANOVA). D, competition of HA-IRAP/FLAG-TBC1D1 co-immunoprecipitation with excess PTB domains of TBC1D1. Cleared lysates from HEK293 cells coexpressing full-length HA-IRAP and FLAG-TBC1D1 were incubated in the absence and presence of GST-PTBD or GST for 1 h and co-precipitations were carried out using magnetic beads coated with anti-HA monoclonal antibodies as described under "Experimental procedures." Samples were separated by SDS-PAGE and co-precipitated FLAG-TBC1D1 was analyzed by Western blotting with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as loading control. Quantification of HA-IRAP/FLAG-TBC1D1 co-immunoprecipitation was with excess GST-PTB. Data represent mean Ϯ S.E. from three independent experiments. ***, p Ͻ 0.001 TBC1D1 versus TBC1D1 ϩ PTB (one-way ANOVA). E, GST-pulldown of GST-cIRAP and TBC1D1-R125W. Immobilized GST-cIRAP was incubated with 5 pmol of purified nonphosphorylated, AKT2-phosphorylated, or AMPK-phosphorylated TBC1D1-R125W from Sf9 cells for 1 h at 4°C as described under "Experimental procedures." Immobilized GST was used as a negative control. Eluted samples were separated by SDS-PAGE and analyzed by Western blotting. F, GAP activity of TBC1D1 in the presence of GST-cIRAP. Purified [␥-32 P]GTP-loaded GST-Rabs (0.6 -1 pmol) were incubated in the absence and presence of 2 pmol of purified full-length TBC1D1 and 10 pmol of GST-cIRAP for 30 min and subsequently the amount of released [ 32 P]phosphate was determined. Data represent mean Ϯ S.E. from three independent experiments.

Regulation of recombinant TBC1D1 in vitro
measurements of GTP hydrolysis rates of recombinant GSTtagged Rab GTPases in the absence and presence of the intact GAP domain resulted in identification of several Rabs that are considered to be substrates, including Rab8a, Rab10, Rab14, and Rab28 (26). Similar results were reported for the truncated GAP domain of the related TBC1D4 (13). Interestingly, catalytic activities for the GAP domains of TBC1D1 and TBC1D4 using Rab14 as the substrate were 1-2 orders of magnitude lower than that of the yeast homolog Gyp1p toward the native yeast substrate Ypt1p or the mouse Rab33B (22). As a result, it has been questioned whether and to what extent the truncated GAP domains reflect the properties of the full-length RabGAPs. Our results demonstrate that the substrate specificity of fulllength TBC1D1 is similar to that of the truncated GST-tagged GAP domain, when Rab8a, Rab10, and Rab14 are considered. However, by comparing the activities of both enzymes we found that full-length TBC1D1 was markedly more active toward Rab10 than the truncated form. Although our radioactive filtration-based assay was not suitable for precise determination of kinetic constants, we found that the apparent K m for Rab10 was similar for both enzymes (approximately 30 M for TBC1D1 and 60 M for GST-GAP) and well in the range of the K m values (25-35 M) previously determined for Gyp1p and Rab33B (11). In contrast, we found a ϳ10-fold higher apparent V max of the full-length TBC1D1 compared with the truncated GAP domain, which could translate into 1-2 orders of magnitude increase in the k cat /K m value (Fig. 1S). It is possible that the catalytic efficiencies of the truncated domains underestimate the values of the in vivo reaction. Likewise, other regions outside of the GAP domain may participate in structural alignment of the GAP and enzyme/substrate interactions. However, further studies are required to address these issues.
Phosphorylation of Ser/Thr residues in response to insulin stimulation and contraction has been implied to regulate the GAP activity of both TBC1D1 and TBC1D4 (7,17). Using purified TBC1D1 and kinases, we confirmed a previous observation that AMPK targets more phosphosites than AKT2 (7). With 86% sequence coverage in the Lumos-Orbitrap analysis, we found that two sites, Ser 489 and Thr 590 , are direct targets for AKT2, and five sites, Ser 231 , Thr 404 , Thr 590 , Ser 660 , and Ser 700 , are direct targets for AMPK in vitro. Phosphorylation of Thr 404 in TBC1D1 by AMPK has not been reported before. By analyzing TBC1D1 immunoprecipitates from insulin-and AICARstimulated tibialis anterior muscle lysates, a previous study described a total of 7 phosphorylation sites in TBC1D1: phosphorylation of Ser 231 , Thr 499 , Ser 660 , and Ser 700 was found mainly after AICAR stimulation, Thr 253 was found phosphorylated only after stimulation with insulin, and phosphorylation of Thr 590 and Ser 621 was observed after stimulation with both, AICAR or insulin (7). Although we could confirm Thr 590 as AKT target, and Ser 231 , Thr 590 , Ser 660 , and Ser 700 as targets for AMPK, we did not detect phosphorylation of Ser 253 and Thr 499 , indicating that kinases other than AKT and AMPK may phosphorylate these sites in vivo in response to insulin and AICAR.
Nevertheless, as all known phosphorylation sites do not map into the GAP domain, it has been unclear how those Ser/Thr residues would directly or indirectly regulate the GAP activity. Mutation of phosphorylation sites in the RabGAPs has been linked to altered biological activity and the activation state of Rab10 in vivo (29). For the related TBC1D4 it has been shown that insertion of a constitutive binding site for 14-3-3 restores the biological activity of a mutant TBC1D4 lacking the four phosphorylation sites (4P) in 3-T3-L1 adipocytes (19). It was suggested that 14-3-3 proteins may modulate targeting or activity of the GAP domain in TBC1D4.
Our results clearly demonstrate that phosphorylation by AKT2 and AMPK, or subsequent binding of 14-3-3 proteins to the phosphorylated full-length protein does not have a direct impact on the intrinsic GAP activity of TBC1D1 toward Rab10, Rab8, and Rab14 in vitro. Moreover, phosphorylation of TBC1D1 did not alter the M r in size exclusion chromatography, indicating that the oligomeric state of the proteins may not be changed acutely by kinases (data not shown). Of note, quantitative assessment of TBC1D1 binding to 14-3-3 proteins indicated a substantial extent of phosphorylation after incubation with purified kinases. The CBD in the related TBC1D4 has been shown to bind calmodulin in a calcium-dependent manner (30), but the impact if this interaction on regulating GLUT4 translocation and glucose uptake has been discussed controversially (28,30). The RabGAP activity of TBC1D1 toward Rab10 was not different in the presence of Ca 2ϩ /calmodulin in vitro, indicating that this interaction does not directly alter the activity of the enzyme.
As TBC1D1 likely targets Rab GTPases bound to GLUT4 vesicles, we tested the hypothesis whether recruitment of the RabGAP might be regulated by phosphorylation. Our data provide evidence that TBC1D1 interacts with the GLUT4 vesicle-resident protein IRAP in a phosphorylation-dependent manner: (i) co-precipitation of TBC1D1 occurs with both fulllength IRAP and the 110-amino acid cytoplasmic tail (cIRAP); (ii) this interaction is strongly reduced after phosphorylation of TBC1D1 by AMPK or AKT2; (iii) interaction of full-length TBC1D1 with IRAP is blocked with an excess of the PTB domains of TBC1D1. Binding of IRAP does not affect the GAP activity of TBC1D1. We therefore propose a mechanism consistent with the findings from other groups by which phosphorylation by AKT2 and AMPK does not regulate the intrinsic GAP activity but rather prevents recruitment of TBC1D1 to the Rabs located in GLUT4 vesicles. In fact, interaction of the related TBC1D4 with IRAP has been demonstrated in previous studies (25,(31)(32)(33). However, recruitment of TBC1D4 to GLUT4 vesicles has been reported to occur even in the absence of IRAP and was not impaired by constitutive binding of 14-3-3 proteins, indicating a more complex requirement of RabGAP targeting of the GLUT4 vesicles (19,34).
The R125W coding variant in TBC1D1 has been linked to an increased risk for familial obesity in humans (5,6). Modeling studies based on related PTB domain structures suggested that the R125W exchange might affect certain ligand binding properties of the first PTB domain in TBC1D1 and AMPK interaction (35,36). However, in this study we found no evidence that addition of excess PTB domains alters the GAP activity of TBC1D1 in vitro. Interestingly, overexpression of the R125W mutant in mouse tibialis anterior muscle was reported to reduce glucose uptake in response to a glucose bolus injection, although not affecting basal glucose transport (20). In this

Regulation of recombinant TBC1D1 in vitro
study, we also expressed and purified the R125W variant of TBC1D1. Taken together, we did not observe any differences in phosphorylation pattern, GAP activity, or phosphorylation-dependent IRAP binding in vitro.
In summary, our results implicate a plausible mechanism for the regulation of RabGAP function, and contribute novel insights into the function of RabGAPs. Further studies are required to understand the subcellular distribution of TBC1D1 targeting GLUT4 vesicles.

DNA constructs
The baculovirus expression vector pAcGS2 was purchased from BD Bioscience (Heidelberg, Germany). A His 6 -tagged cDNA fragment for the long mouse isoform of Tbc1d1 (1255 amino acids; NP_001297540.1) was cloned into pAcGS2 to generate a baculovirus as described (37). A baculovirus for a GST fusion protein of mouse Rab10 was constructed as described (37). The constructs for TBC1D1-R854K and -R125W mutations were generated using PCR-based QuikChange Site-directed Mutagenesis Kit (Agilent Technologies, CA). The cDNA for the hemagglutinin (HA)-tagged insulin-regulated aminopeptidase (IRAP) from rat was cloned into the cytomegalovirus pCIS2 expression vector (38). For production of GST-cIRAP, a cDNA fragment coding for the cytoplasmic tail of IRAP (Met 1 -Thr 110 ) was cloned into pGEX-3X. For generation of GST-PTBD, a cDNA fragment coding for the N-terminal PTB domains of TBC1D1 (Met 1 -Gly 382 ) was cloned into pGEX-3X. Cloning of GST-Rab8a, Rab14, GST-RabGAP domain of TBC1D1, and FLAG-TBC1D1 (all from mouse) was as described before (4,26). Plasmids for GST fusion proteins of human 14-3-3␥ were generous gifts from Dr. Michael B. Yaffe (Massachusetts Institute of Technology, Cambridge, MA). All constructs were confirmed by DNA sequencing.

Animals
Mice were kept in accordance with the NIH guidelines for the care and use of laboratory animals, and all experiments were approved by the Ethics Committee of the State Ministry of Agriculture, Nutrition and Forestry (State of North Rhine-Westphalia, Germany) and the State Agency of Environment, Health, and Consumer Protection (State of Brandenburg, Germany). Three to six mice per cage were housed at 22°C and a 12-h light-dark cycle with ad libitum access to food (standard diet; Ssniff, Soest, Germany) and water.

Antibodies
TBC1D1 and IRAP antibodies were obtained from Cell Signaling Technology (Danvers, MA). FLAG antibody was purchased from Sigma. Phospho-TBC1D1 (Thr 596 and Ser 237 ) antibodies were kindly provided from Dr. Hilary McLauchlan, University of Dundee, UK.

Mass spectrometry
In vitro phosphorylated TBC1D1 using ATP (ϩ79.966 Da; Ser, Thr, Tyr) or "heavy" [␥- 18 O 4 ]ATP (ϩ85.966 Da; Ser, Thr, Tyr; Cambridge Isotope Laboratories (40)) was separated by SDS-PAGE, followed by Coomassie staining. Visible protein bands corresponding to TBC1D1 were excised and subjected to in-gel protein digestion. Samples were measured using LC-MS instrumentation consisting of an Ultimate 3000 separation LC system (ThermoFisher Scientific, Schwerte, Germany), which was combined with an EASY-spray ion source and Orbitrap Fusion TM Lumos TM Tribrid TM mass spectrometer (Thermo-Fisher Scientific). Peptides were trapped on an Acclaim Pep-Map C18-LC column (inner diameter: 75 m, 2 cm length; ThermoFisher Scientific) and separated via EASY-Spray C18 column (ES802; inner diameter: 75 m, 25 cm length; Thermo-Fisher Scientific). Mass spectrometry raw files were analyzed with Proteome Discoverer TM 2.2 software (ThermoFisher Scientific) and HTsequest search was done against the FASTA database (SwissProt Mus musculus (TaxID ϭ 10090), 25,103 sequences (version 2017-10 Ϫ25 )), enzyme was set to trypsin with maximum 2 missed cleavage sites allowed, a fragment mass tolerance of 0.02 Da, cysteine carbamidomethylation as fixed, and N-terminal acetylation, methionine oxidation, and serine, threonine, or tyrosine phosphorylation as variable modification were used. For validation of phosphorylated peptides the ptmRS-node included in Proteome Discoverer TM were applied and only phosphosites with site probability of more than 100 or Ͼ75% were annotated to protein.

Kinase assays
Recombinant human constitutively active AKT2 was purified from Sf9 lysates as described previously (37) and purified trimeric AMPK was obtained from Invitrogen. Two l of the purified enzyme (about 4 pmol of protein) were used to phosphorylate about 2 pmol of purified TBC1D1 for 20 min at room temperature in the presence of 2 mM ATP, 40 mM Tris-HCl, pH

GST pulldown reactions
E. coli BL21-CodonPlus (DE3)-RIL cells expressing GST-cIRAP and GST-14-3-3 were lysed in lysis buffer (10 mM Tris, pH 8, 2.5 mM MgCl 2 , 5 mM ␤-mercaptoethanol, 1 mM DTT and protease inhibitor mixture) and centrifuged at 20,000 ϫ g for 20 min. 1 ml of cleared lysate was incubated with 150 l of GSH-Sepharose beads (GE Healthcare) for 2 h at 4°C with mixing. As a negative control the same amount of the beads were incubated with GST alone. The beads were washed with 500 l of a buffer containing 50 mM Tris, pH 8, and 150 mM NaCl and incubated with 1 pmol of nonphosphorylated and AKT2/ AMPK-phosphorylated TBC1D1 for 1 h at 4°C with gentle mixing. The beads were washed intensively with the same buffer and eluted either with a buffer containing 10 mM reduced GSH (Sigma) or boiled in sample buffer. The eluted fractions were analyzed in parallel with TBC1D1 and phospho-TBC1D1 antibodies.

TBC1D1 and IRAP co-immunoprecipitation
HEK-293 cells were seeded at a density of 3 ϫ 10 5 cells/ 35-mm dish, and were transfected 1 day later with 4 g of pCIS-HA-IRAP and 4 g of pcDNA3-3xFLAG-TBC1D1 using Lipofectamine 2000 (Invitrogen). 2-3 days post-transfection, cells were harvested, lysed in 150 mM NaCl, 1% Triton X-100, 50 mM Tris-HCl, pH 8.0, EDTA-free protease inhibitor tablets (Roche Diagnostics), and cleared by centrifugation at 20,000 ϫ g for 10 min. Phosphorylation was carried in buffer containing 2 mM ATP, 40 mM Tris-HCl, pH 7.4, 8 mM MgCl 2 , 200 M AMP, and 5 microunits of purified AMPK for 20 min at room temperature. Magnetic beads pre-conjugated with anti-HA monoclonal antibodies (MACS HA-beads, Miltenyi Biotec GmbH, Bergisch Gladbach, Germany) were used according to the manufacturer's instructions. Antibody-coupled beads and cleared cell lysates were incubated for 30 min at 4°C and subsequently beads were washed with 50 mM Tris, pH 8, and 150 mM NaCl and the protein complexes were eluted using SDS sample buffer. For competition assays ϳ100 pmol of purified GST-PTBD protein was added prior to co-immunoprecipitation.

RabGAP assay
GST-Rab proteins bound to GSH-Sepharose were loaded with [␥-32 P]GTP (Hartmann Analytic, Braunschweig, Germany) in a buffer containing 50 mM Tris-HCl, pH 8, 2.5 mM DTT, 5 mM MgCl 2 . The matrix was washed with the same buffer to remove unbound GTP, and [␥-32 P]GTP bound Rab proteins were eluted in 20 mM Tris-HCl, pH 7.5, 1 mM DTT, 10 mM reduced GSH, 2.5 mM MgCl 2 and then incubated with GST-GAP domains or full-length TBC1D1 at 30°C for different time points. Aliquots were stopped with 0.5 M EDTA, radioactive [ 32 P]phosphate was separated by filtration through activated charcoal and measured by scintillation counting as described (26). Unless indicated otherwise, [ 32 P]phosphate produced was normalized to the amount of radioactivity of [␥-32 P]GTP bound Rab proteins.

Data analysis
Data are expressed as mean Ϯ S.E. unless indicated otherwise. Nonlinear fitting to Michaelis-Menten was performed by using GraphPad Prism (GraphPad Software, Inc., La Jolla, CA). p values were calculated by two-tailed unpaired t test or twoway ANOVA followed by Bonferroni post hoc test using GraphPad Prism. Statistical significance was defined as p Ͻ 0.05.