Phosphorylation of the N-terminus of Syntaxin-16 controls interaction with mVps45 and GLUT4 trafficking in adipocytes

Cells and antibodies

HEK293T cells (Cat# CRL-3216 RRID:CVCL_0063) and 3T3-L1 fibroblasts (Cat# CCL-92.1, RRID:CVCL_0123) were obtained from the American Tissue Culture Collection and grown and differentiated into adipocytes as outlined in (Sadler et al., 2019). HeLa cells expressing HA-GLUT4-GFP were as described previously (Morris et al., 2020). AMPK α1/ α2 knockout mouse embryonic fibroblasts (MEFs) and control MEFs were cultured as described previously (Laderoute et al., 2006). Antibodies used: anti-GLUT4 (Thermo Fisher Scientific Cat# PA1-1065, RRID:AB_2191454), anti-Sx16 (Synaptic Systems Cat# 110 162, RRID:AB_887799), anti-mVps45 (Novus Cat# NB100-2431, RRID:AB_2272935), anti-VAMP4 (Synaptic Systems Cat# 136 002, RRID:AB_887816), anti-SNAP23 (Synaptic Systems Cat# 111 202, RRID:AB_887788) and anti-Syntaxin 4 (Synaptic Systems Cat# 110 042, RRID:AB_887853). A phospho-specific antibody raised in rabbits against the N-terminal 16 amino acids of murine Sx16 with Thr-7 chemically phosphorylated and recognizing Thr-7 in the Sx16 N-terminus only when phosphorylated, was generated by Antagene Inc (Mountain View, CA, USA) and purified using affinity chromatography. Anti-AKT (pan) (Cell Signaling Technology Cat# 2920, RRID:AB_1147620) and anti-AKT Phospho-S473 (Cell Signaling Technology Cat#4058), RRID:AB_331168) were used as outlined (Bremner et al., 2022).

Plasmids and viruses

HA-GLUT4-GFP in the pRRL-PGK plasmid was from Cynthia Mastick (Brewer et al., 2011; Muretta & Mastick, 2009) and pCOG70, driving the over-expression of Vps45p harboring an HA tag at its carboxy terminus was a generous gift of Dr. Lindsay Carpp (Carpp et al., 2006; Carpp et al., 2007). The V107R (Sx16 binding deficient) mutant of mVps45 (corresponding to V107R) was generated using a QuickChange kit (Thermo Fisher Milton Park, UK) and fully sequenced on both strands prior to expression in yeast. T7A and T7D mutants in a truncated Sx16 lacking the transmembrane domain were generated by PCR and fully sequenced. Constructs comprising the first 33 residues of Sx16 (either wild-type or containing the F10A mutation previously shown to disrupt binding to mVps45) fused to Protein-A were generated by PCR and fully sequenced on both strands and sub-cloned into pcDNA3.1 for expression in mammalian cells. All other cDNAs were as previously outlined (Morris et al., 2020; Proctor et al., 2006; Sadler et al., 2019). Full length Sx16 containing the T7A to T7D mutations were synthesised by GenScript (Piscataway, NJ, USA) and sub-cloned into pCDH vectors for lentivirus production using the ViraPower™ Lentiviral system (Invitrogen, Waltham, MA, USA) following manufacturer’s instructions. The sequence used corresponds to Q8BVI5 in UniProt and is the largest of the proposed splice variants of Sx16 in mouse tissues, and is the only version thought to contain a transmembrane domain. 3T3-L1 adipocytes were incubated with lentivirus for 12 h on day 5 or 6 post-differentiation. The media was changed, and the cells were assayed between day 7 and 9 after differentiation (48–72 h post infection).

Recombinant protein production

Sx16 lacking its transmembrane domain, or the T7A or T7D mutants, were sub-cloned in-frame with dual protein-A tags, expressed in E. coli and purified on IgG-Sepharose (GE Healthcare). Protein levels were determined using Micro-BCA kits (Pierce, Waltham, MA, USA).

2-deoxyglucose transport, adipocyte fractionation, immunofluorescence, and immunoblotting

2-Deoxy-D-glucose transport was assayed in cells exactly as outlined (Bremner et al., 2022; Roccisana et al., 2013) in except 96-well plates were employed to minimise virus requirements. To determine the levels of GLUT4, adipocyte lysates were generated as outlined (Roccisana et al., 2013; Sadler et al., 2019). SDS-PAGE and immunoblotting were performed as described (Sadler et al., 2019). Immunoblot signals were quantified using LICOR software (Bremner et al., 2022). Immunofluorescence to quantify cell surface GLUT4 staining in HeLa cells was performed exactly as outlined in Morris et al. (2020). Confocal microscopy of HA-GLUT4-GFP-expressing cells was performed on a Zeiss Exciter equipped with a 63x lens (Morris et al., 2020). Images were captured and processed using Zeiss software.

Phosphorylation and analysis

Adipocyte cytosol was prepared from day-10 adipocytes. Cell plates were washed three times in ice-cold buffer with protease inhibitors (50 mM Hepes pH 7.4; 5 mM EDTA, 10 mM sodium pyrophosphate, 10 mM NaF, 150 mM NaCl, 2 mM sodium orthovanadate, 50 mM beta-glycerophosphate and 1 mM DTT) and homogenised by passing through a needle twenty times. The homogenate was centrifuged at 12,000 g for ten minutes and the fat cake carefully aspirated. The supernatant was used in a series of in vitro kinase assays; supernatant was stored at −80 °C prior to use. In brief, 30 µg of cytosol extract was incubated in 50mM HEPES pH 7.5, 10 mM MgCl₂, 2 mM MnCl₂, 0.2 mM DTT, 50µM ATP.Mg²⁺ (±[γ-³²P]-ATP) in the presence of 5 µg of recombinant Protein A-tagged Sx16 (residues 1-279 of Sx16, fused to two copies of Protein-A via a thrombin cleavage site), protein A alone or myelin basic protein at 30 °C for 2 h. Samples were analysed by autoradiography.

Analysis of the phosphorylation sites was performed at the FingerPrints facility, University of Dundee. Phosphorylation reactions were carried out exactly as outlined above but using non-radioactive ATP; control reactions lacking cytosol and/or ATP were performed. Bands corresponding to Sx16-PrA were excised and trypsinised, followed by liquid chromatography-electrospray ionisation mass spectrometry on an ABI QTrap 4000 instrument. Potential phospho-peptide ions present in samples but absent from controls were submitted to ESI-MS/MS to determine their amino acid sequence and locate the phosphorylated residue.

A total of 2 µg of purified ProteinA-Sx16 or Protein A-Sx16-T7A was incubated with the indicated purified recombinant protein kinases (5 mU; see figure legends for details) in the presence of 10 mM MgCl₂, 0.1 mM [γ-³²P]-ATP (approx. 0.5 × 10⁶ cpm/nmol) and kinase buffer (50 mM Tris-Cl, 0.03% (v/v) Brij-35, 0.1% (v/v) β-mercaptoethanol) at 30 °C in a total volume of 50 µl. A 20 µl aliquot of reaction mix was taken at the times indicated on the figure legends and the reactions terminated by addition of 4xNovex SDS PAGE loading buffer and heating at 70 °C for 15 mins. The Sx16 was separated from [ γ-³²P]-ATP by SDS-PAGE on Novex 4–12% gradient gels using a MOPS buffer. After staining with Coomassie Brilliant Blue to visualise the protein bands, the gels were dried and subjected to autoradiography. Finally, the Sx16 bands were excised, and the incorporated radioactivity was determined by scintillation counting.

In experiments using anti-phospho-T7 for detection, in vitro kinases assays typically contained 0.5 mg of bacterially expressed Sx16 incubated with 30 µg of cytosol in the presence or absence of 25 mM ATP for 60 min at 30 °C as indicated. Samples were then immunoblotted with anti-Sx16 antibodies or anti-phospho-T7 antibodies as described in the figure legends.

Statistical analysis

Result are expressed as ± SD with the N-number indicated. Significant differences were determined using either t-test or ANOVA as indicated in the figure legends, with p < 0.05 being deemed as significant. Statistical analysis was carried out using GraphPad Prism software.

Interaction of mVps45 and Sx16 is functionally important for GLUT4 sorting

Syntaxin proteins are subject to regulation by cognate SM proteins through multiple distinct but evolutionarily conserved binding mechanisms (Morey et al., 2017; Munson & Bryant, 2009; Sudhof & Rothman, 2009; Yu et al., 2013). One of the best characterized examples of this is binding of yeast Vps45p by the Syntaxin Tlg2p via its N-terminal peptide which is necessary for productive SNARE complex formation (Carpp et al., 2006; Carpp et al., 2007; Furgason et al., 2009; Struthers et al., 2009). This binding mechanism is conserved for mammalian Sx16 which binds mVps45 in an analogous manner (Struthers et al., 2009) and involves the insertion of the Sx16 N-terminus into a hydrophobic pocket located on the surface or mVps45 (Carpp et al., 2006; Dulubova et al., 2002; Munson & Bryant, 2009). This interaction is known to mask a lower affinity but functionally important interaction which does not involve the Sx16 N-terminus and involves Sx16 binding to mVps45 in a conformation incompatible with productive SNARE complex formation (Furgason et al., 2009; Munson & Bryant, 2009; Struthers et al., 2009).

Given the previously demonstrated a role for both Sx16 and mVps45 in GLUT4 traffic (Bremner et al., 2022; Proctor et al., 2006; Roccisana et al., 2013; Shewan et al., 2003), here we asked whether the interaction between these two proteins is required for their function(s) in this process. To this end, we undertook two different approaches to disrupt this interaction. We produced a protein fusion between the N-terminal 33 residues of Sx16 and Protein-A and expressed this in HeLa cells expressing HA-GLUT4-GFP. We reasoned that introduction of this N-terminal peptide into cells would inhibit the interaction between Sx16 and mVps45, prevent functional SNARE complex formation and perturb GLUT4 trafficking to the IRVs (Munson & Bryant, 2009) and elected to study this initially using the readily tractable HeLa cell system which we and others have shown behaves similarly to adipocytes and muscle (Bryant & Gould, 2020; Morris et al., 2020). Consistent with this, we observed significantly reduced insulin-stimulated translocation of HA-GLUT4-GFP to the plasma membrane in cells staining positive for the Protein-A fusion. By contrast, a mutation in this peptide (F10A), previously shown to reduce binding of Sx16/Tlg2p to mVps45/Vps45p (Dulubova et al., 2002; Struthers et al., 2009), was without effect (Figs. 1A, 1B).

Studies in yeast have shown that mutation of a conserved hydrophobic residue predicted to be on the outer surface of Vps45 is required for interaction with the N-peptide of Tlg2 and Sx16, with its mutation abrogating interaction (Carpp et al., 2006; Dulubova et al., 2002; Struthers et al., 2009). We therefore created this mutation in mVps45 and determined its effect on GLUT4 distribution. The rationale for this approach being that if cycles of association/dissociation of mVps45/Sx16 are needed for effective GLUT4 sorting, the over-expression of a mutant defective in this interaction should manifest itself by impairment of GLUT4 sorting. Transient over-expression of (V107R)-mVps45 in HeLa cells increased cell surface GLUT4 levels measured in the absence of insulin, detected using anti-HA antibodies in non-permeabilised cells, consistent with disruption of the Sx16/mVps45 interaction perturbing trafficking of GLUT4. Over-expression of wild type mVps45 was without effect. This data is shown in Fig. 2 in which cells over-expressing wild-type HA-mVps45 or (V107R)-mVps45 were identified by immunostaining with anti mVps45 (we are unable to use anti-HA to quantify mVps45 expression because of the presence of HA-tagged GLUT4 in these cells). This approach stains both endogenous and over-expressed proteins, hence in the visual analysis shown, cells expressing higher than average mVps45 staining were assigned to be positive for the transfected DNA (Fig. 2A); similar levels of the mVps45 species were expressed as judged by immunoblot analysis (Fig. 2B). To mitigate against bias in this subjective analysis, we also measured the cell surface GLUT [HA staining in non-permeabilised cells]/GFP ratio for entire populations of transfected cells (i.e., containing both non-transfected and transfected cells) and found that over-expression of wild-type HA-mVps45 did not change the cell surface GLUT[HA staining in non-permeabilised cells]/GFP ratio compared to non-transfected cells, but over-expression of (V107R)-mVps45 consistently increased this ratio by 20% (Fig. 2C). Collectively, these data, together with our previous observations of depletion of either protein separately (Bremner et al., 2022; Proctor et al., 2006; Shewan et al., 2003), supports the hypothesis that an interaction between the N-terminus of Sx16 and mVps45 is required for the correct trafficking of GLUT4.

Sx16 is phosphorylated on T7

Previous work from this lab has shown that Sx16 is a phosphoprotein (Perera et al., 2003). We therefore sought to identify potential sites of phosphorylation of this key SNARE. Attempts to identify phosphosites using combinations of co-IP or purification from adipocytes proved unsuccessful. However, we consistently observed phosphorylation of recombinant Sx16 in vitro using an adipocyte cytosolic extract (Fig. 3A). Therefore, this phosphorylated recombinant Sx16 was subjected to phospho-site mapping. Phospho-mapping hits (FingerPrints ProteomicsFacility, University of Dundee) were assigned a score based upon probabilities that the fragment detected was a true phospho-peptide (Table 1; Fig. S1). The highest score, and the only hit, was assigned to fragments containing Threonine-7 (hereafter T7) suggesting that under these conditions, T7 is a major phospho-acceptor site. We therefore generated a phospho-selective antibody against the N-terminal 16 amino acids of murine Sx16. This antibody recognized a band of the expected molecular weight in in vitro phosphorylation reactions, the intensity of which increased significantly when ATP was included in the reactions (Fig. 3B). Consistent with this antibody recognising the phosphorylated forms of Sx16, we showed that the antibody cross-reacted strongly with recombinant Sx16-T7D but significantly less so with Sx16-T7A or Sx16 purified from bacteria (Fig. 3C).

Phospho-mimetic mutation of T7 perturbs binding to mVps45

Interaction of Sx16 with mVps45 involves the extreme N-terminus of Sx16 (Carpp et al., 2006; Dulubova et al., 2002; Furgason et al., 2009; Munson & Bryant, 2009), hence the identification of a potential phospho-acceptor site in this region could suggest a potential regulatory mechanism to control Sx16/mVps45 interaction and function. To test this, we employed recombinant Sx16-PrA to pull-down HA-tagged human mVps45 expressed in yeast and compared wildtype Sx16 (which is not phosphorylated when expressed in bacteria) with Sx16-PrA containing the phospho-mimetic T7D mutation (Fig. 4). As shown wild-type Sx16 effectively pulled-down HA-tagged mVps45 from yeast lysate. Although Sx16-T7D was still able to bind HA-mVps45, this was reduced in both magnitude and affinity (Fig. 4). This data strongly suggests that phosphorylation of T7 could significantly impact Sx16 function by preventing or reducing the affinity of its interaction with mVps45.

Sx16-T7D perturbs insulin-stimulated glucose transport

Previously, we have shown that perturbation of either Sx16 or mVps45 function modulated GLUT4 sorting in adipocytes. Hence, if the interaction between Sx16 and mVps45 is regulated by phosphorylation at T7, as suggested by the data in Figs. 3 and 4, we would predict that phosphorylation of this site would be expected to modulate GLUT4 trafficking in vivo.

To test this hypothesis, full length Sx16 wild-type, or T7A and T7D mutants were over-expressed in 3T3-L1 adipocytes, a well-used model of insulin-stimulated GLUT4 translocation and glucose transport. We consistently observed reduced insulin-stimulated 2-deoxyglucose uptake in cells over-expressing Sx16-T7D compared to either wild-type or Sx16-T7A (Fig. 5A). To facilitate comparison of datasets from biological replicates in which levels of over-expression of Sx16 varied and the magnitude of the insulin response varied, the data from multiple experiments expressed as a % of the maximal insulin response is shown in Fig. 5A. A representative immunoblot of lysates from a typical experiment is shown in Fig. 5B; for reasons that we have been unable to identify, the overexpressed proteins migrate faster than endogenous Sx16. The apparent reduction in total GLUT4 levels in T7D-expressing cells in this dataset was not recapitulated consistently in all experiments. Importantly, insulin-stimulated activation of Akt was not modulated by expression of the different Sx16 mutants, indicating that a deficit in insulin-signalling is unlikely to be the mechanism responsible for reduced insulin-simulated glucose transport (Fig. 5C).

Sx16 is a substrate for AMP-dependent protein kinase

The observation that Sx16-T7 is a potential regulatory site for the control of GLUT4 trafficking prompted an investigation of the kinase(s) that might phosphorylate this site. The peptide sequence around T7 resembles the proposed consensus sequence for the AMP activated protein kinase (AMPK) and several of its related kinases (Bright, Thornton & Carling, 2009). AMPK activation enhances insulin-stimulated glucose transport in skeletal muscle (Kleinert et al., 2017). However, it should be noted that in 3T3-L1 adipocytes, the converse is true: activation of AMPK inhibits insulin-stimulated glucose transport and GLUT4 translocation to the plasma membrane (Salt, Connell & Gould, 2000). Using recombinant Sx16 as a substrate for in vitro phosphorylation reactions revealed that AMPK phosphorylated Sx16 in vitro, and that the extent of this phosphorylation was comparatively reduced in Sx16-T7A under identical initial rate phosphorylation conditions. Importantly phosphorylation was enhanced by inclusion of AMP in the kinase assay mix (Fig. 6). Sx16 was also phosphorylated by the AMPK-related kinase SIK2 (Bright, Thornton & Carling, 2009; Patel et al., 2014) and the extent of this phosphorylation was consistently reduced in the Sx16-T7A mutant (Fig. 6); MARK1 (Bright, Thornton & Carling, 2009) also phosphorylates Sx16 but this was not reduced in Sx16-T7A and hence was not studied further (Fig. S2). By contrast the AMPK-related kinase BRSK2 (Bright, Thornton & Carling, 2009) (Fig. S2) was without effect on Sx16 phosphorylation. It should be noted that phosphorylation of Sx16-T7A by AMPK is also observed, indicating that further phospho-acceptor sites are present within this SNARE protein.

Our initial report identifying Sx16 as a phosphoprotein indicated that total Sx16 phosphorylation levels were slightly reduced in insulin-stimulated 3T3-L1 adipocytes compared to controls (Perera et al., 2003). Hence, we adopted several routes to address this in intact 3T3-L1 adipocytes here focussing on phospho-T7 levels using our phospho-selective antibody. We have been unable to measure any change in phospho-T7 levels in 3T3-L1 adipocytes in response to insulin (Fig. 7A) or 5-aminoimidazole-4-carboxamide ribonucleotide (AICAR) a known activator of AMPK, (Fig. 7B). One possible explanation for this is provided by the data in Fig. 7C in which lysates from Mouse Embryonic Fibroblasts (MEFs) engineered to lack AMPK compared to their wildtype counterparts were used in in vitro phosphorylation reactions with or without the addition of ATP. Lack of AMPK modestly reduced the phosphorylation of Sx16 (Fig. 7D), but the effect was small, indicating that other, presently unidentified, kinases may regulate this site. This is consistent with the autoradiography in Fig. 6 which indicates substantial phosphorylation of Sx16-T7A by AMPK, suggesting that other sites within Sx16 may be important, and with our previous data (Perera et al., 2003). Future work should address whether phosphorylation status of Sx16 is altered in disease or conditions of insulin resistance.