ARTICLE Sensing of nutrients by CPT1C controls SAC1 activity to regulate AMPA receptor trafficking Maria Casas1, Rut Fadó1, José Luis Domı́nguez1, Aina Roig1, Moena Kaku2, Shigeru Chohnan2, Montse Solé3, Mercedes Unzeta3, Alfredo Jesús Miñano-Molina3,4, José Rodrı́guez-Álvarez3,4,5, Eamonn James Dickson6, and Núria Casals1,7 Carnitine palmitoyltransferase 1C (CPT1C) is a sensor of malonyl-CoA and is located in the ER of neurons. AMPA receptors (AMPARs) mediate fast excitatory neurotransmission in the brain and play a key role in synaptic plasticity. In the present study, we demonstrate across different metabolic stress conditions that modulate malonyl-CoA levels in cortical neurons that CPT1C regulates the trafficking of the major AMPAR subunit, GluA1, through the phosphatidyl-inositol-4-phosphate (PI(4)P) phosphatase SAC1. In normal conditions, CPT1C down-regulates SAC1 catalytic activity, allowing efficient GluA1 trafficking to the plasma membrane. However, under low malonyl-CoA levels, such as during glucose depletion, CPT1C-dependent inhibition of SAC1 is released, facilitating SAC1’s translocation to ER-TGN contact sites to decrease TGN PI(4)P pools and trigger GluA1 retention at the TGN. Results reveal that GluA1 trafficking is regulated by CPT1C sensing of malonyl-CoA and provide the first report of a SAC1 inhibitor. Moreover, they shed light on how nutrients can affect synaptic function and cognition. Introduction α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid re- capacity to influence GluA1 subunit abundance at the PM. At the ceptors (AMPARs) are ionotropic glutamate receptors that me- molecular level, they do so by tuning intracellular trafficking diate the majority of excitatory synaptic transmissions in the patterns. To date, three trafficking pathways have been docu- mammalian central nervous system. Functional AMPARs are mented as regulating GluA abundance at the PM: (1) the ca- tetramers, with each subunit having four transmembrane nonical ER–TGN secretory pathway, which requires passage domains, two extracellular regions that bind glutamate, and a through an ER-Golgi intermediate compartment; (2) a recycling cytosolic C-terminal regulatory domain that determines inter- pathway, in which endocytosed GluA1 subunits are recycled actions with other binding proteins. GluA1 is the major AMPAR back to the PM through recycling endosomes; and (3) the Golgi- subunit mediating fast excitatory neurotransmission in the independent route, in which GluA1-containing vesicles exiting brain and plays a critical role in synaptic strength and plas- the ER are routed to recycling endosomes without transiting ticity, with key features encoding many brain functions such as through the Golgi apparatus (Hangen et al., 2018; Bowen et al., learning and memory (reviewed by Jacobi and von Engelhardt, 2017). Despite this information, there have been few molecular 2018). For this reason, GluA1 surface expression at the plasma targets identified within these trafficking pathways that influ- membrane (PM) is tightly regulated, since its down-regulation ence GluA1 expression at the PM. Uncovering such mechanisms depresses neuronal activity and overexpression leads to ex- is essential for understanding how our brain responds and be- citotoxicity (Henley and Wilkinson, 2016). haves under energy deficiencies, such as in metabolic disorders External environmental signals, including situations of oxy- or anorexic subjects. gen and glucose deprivation, caloric restriction or high-fat diets, One intracellular protein that has recently been found be part and exposure to orexigenic and anorexigenic hormones (Moult of AMPAR macrocomplexes is the phosphoinositide phosphatase et al., 2010; Ribeiro et al., 2014; Dennis et al., 2011; Fernandes Suppressor of Actin 1 (SAC1; Schwenk et al., 2012; Chen et al., et al., 2014; Spinelli et al., 2017; Ouyang et al., 2017), have the 2014). SAC1 is a ubiquitously expressed, ER-localized, ............................................................................................................................................................................. 1 Basic Sciences Department, Faculty of Medicine and Health Sciences, Universitat Internacional de Catalunya, Sant Cugat del Vallès, Spain; 2Department of Food and Life Science, Ibaraki University College of Agriculture, Ami, Ibaraki, Japan; 3Biochemistry and Molecular Biology Department, Institute of Neurosciences, Universitat Autònoma de Barcelona, Bellaterra, Barcelona, Spain; 4Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas, Madrid, Spain; 5Department of Neuroscience, Albert Einstein College of Medicine, New York, NY; 6Department of Physiology and Membrane Biology, University of California, School of Medicine, Davis, CA; 7Centro de Investigación Biomédica en Red de Fisiopatologı́a de la Obesidad y la Nutrición, Instituto de Salud Carlos III, Madrid, Spain. Correspondence to Rut Fadó: rfado@uic.es; Núria Casals: ncasals@uic.es. © 2020 Casas et al. This article is distributed under the terms of an Attribution–Noncommercial–Share Alike–No Mirror Sites license for the first six months after the publication date (see http://www.rupress.org/terms/). After six months it is available under a Creative Commons License (Attribution–Noncommercial–Share Alike 4.0 International license, as described at https://creativecommons.org/licenses/by-nc-sa/4.0/). Rockefeller University Press https://doi.org/10.1083/jcb.201912045 1 of 24 J. Cell Biol. 2020 Vol. 219 No. 10 e201912045 phosphatidyl-inositol-4-phosphate (PI(4)P) phosphatase that that it may possess an additional or differing cellular role than tightly regulates PI(4)P levels at different cellular compart- its family counterparts. Additionally, the N-terminal regulatory ments, including the TGN, by dephosphorylating PI(4)P into domain of CPT1C is known to undergo a malonyl-CoA–dependent phosphatidylinositol (PI). Since PI(4)P at the TGN is necessary switch between the Nα and the Nβ conformations (López-Viñas for the recruitment of coat molecules and transport carriers et al., 2007; Rao et al., 2011; Samanta et al., 2014). Malonyl-CoA that control cargo segregation toward endosomal destinations is an intermediate in the novo synthesis of long-chain fatty or directly to the PM, it follows that SAC1 plays a crucial reg- acids and is synthetized by acetyl-CoA carboxylase (ACC), a ulatory role in vesicular trafficking. Indeed, under stressful downstream target of the master energy sensor AMP-activated conditions, such as glucose or serum deprivation, SAC1 accu- protein kinase (AMPK). Crucially, malonyl-CoA levels undergo mulates at Golgi membranes to deplete PI(4)P pools in order to brain-specific fluctuations depending on nutritional state, be- halt secretory transport toward the PM (Faulhammer et al., coming reduced during fasting and increased following feeding 2007; Blagoveshchenskaya et al., 2008). Specifically related to (Tokutake et al., 2010). Given the capacity of CPT1C to weakly AMPAR trafficking, it has been described that upon N-methyl- bind malonyl-CoA, several authors have proposed that CPT1C D-aspartate (NMDA)–induced stress, which causes excitotoxicity, is a malonyl-CoA sensor (Price et al., 2002; Wolfgang et al., SAC1 is retained within the TGN regions, causing Golgi PI(4)P 2006; Casals et al., 2016). Despite this, nothing is known dephosphorylation and a down-regulation of GluA1 trafficking to about nutritional-dependent regulation of CPT1C function or the PM in neurons (Yang et al., 2013). It remains controversial downstream molecular interactions that enable it to control whether SAC1 modulates the TGN pool of PI(4)P in a cis- or in a GluA1 trafficking. trans-conformation, with evidence that SAC1 can act in trans by Here, we demonstrate that CPT1C regulates surface GluA1 directly dephosphorylating PI(4)P in the TGN across ER-TGN abundance in cortical neurons in response to metabolic chal- junctions (Venditti et al., 2019) or in cis combined with the ac- lenges, such as glucose depletion, through the sensing of tions of Oxysterol-binding protein (OSBP; Mesmin et al., 2013). malonyl-CoA levels. Moreover, we uncovered that CPT1C con- Despite numerous reports exploring the spatial localization of trols SAC1 translocation to ER-TGN contact sites, as well as SAC1 across subcellular compartments, little is known about SAC1 PI(4)P phosphatase activity in a malonyl-CoA–dependent possible mechanisms that may influence its cellular localization manner. In basal conditions, CPT1C decreases SAC1 activity, or even its phosphatase activity. which maintains normal PI(4)P levels at the TGN and GluA1 A recently described interactor of SAC1 is the carnitine pal- trafficking to the PM. However, under energy stressful con- mitoyltransferase 1C (CPT1C) protein, which is also part of ditions, CPT1C releases SAC1 inhibition and promotes SAC1 AMPAR macrocomplexes (Brechet et al., 2017). CPT1C belongs to translocation to ER-TGN junctions, where it decreases TGN a family of enzymes classically involved in regulating the PI(4)P pools to halt GluA1 trafficking. Collectively, we describe transport of long-chain fatty acids into mitochondria for their a novel regulator of SAC1 phosphatase activity in neurons and β-oxidation. However, the CPT1C isoform has no catalytic ac- unravel a nutritionally dependent mechanism by which CPT1C tivity, is found only in the ER (not in mitochondria), and is ex- regulates GluA1 trafficking. clusively expressed in mammalian neurons, cancer cells, and stem cells (Price et al., 2002; Sierra et al., 2008; Carrasco et al., 2012; Roa-Mansergas et al., 2018; Zaugg et al., 2011). In neurons, Results CPT1C together with ferric-chelate reductase 1-like drives mul- CPT1C regulates GluA1 trafficking through malonyl-CoA timerization of GluAs into functional tetrameters at the ER sensing (Schwenk et al., 2019). Moreover, it is necessary for proper Since CPT1C has been proposed to be a malonyl-CoA sensor that trafficking of GluA1-containing AMPARs to the PM (Fadó et al., regulates the function of other proteins (Price et al., 2002; 2015; Gratacòs-Batlle et al., 2018, 2015). In the brain, CPT1C has Wolfgang et al., 2006; Casals et al., 2016) and is well known to been found to be critical for correct regulation of food intake and interact with the AMPAR GluA1 (Fadó et al., 2015; Gratacòs- peripheral lipid metabolism in the hypothalamus (Gao et al., Batlle et al., 2015), we decided to study if CPT1C regulates 2011, 2009; Ramı́rez et al., 2013; Pozo et al., 2017; Rodrı́guez- GluA1 trafficking under different nutrient and hormonal chal- Rodrı́guez et al., 2019). In addition, its rich expression in the lenges known to modify intracellular malonyl-CoA levels in cerebral motor cortex and cerebellum makes it relevant in motor cortical cultured neurons. coordination and activity (Carrasco et al., 2013), underscored by We first analyzed the effect of glucose deprivation. Glucose human CPT1C mutations being associated with hereditary deprivation for 2 h remarkably decreased the amount of surface spastic paraplegia (Rinaldi et al., 2015; Hong et al., 2019). Finally, GluA1 in cultured cortical neurons. Interestingly, CPT1C- due to its abundance in the hippocampus and in the cerebral silenced neurons showed lower basal GluA1 levels at the PM, cortex, CPT1C also plays a pivotal role in cognition (Carrasco as previously published (Fadó et al., 2015), but they did not et al., 2012). Here, CPT1C regulates spatial learning via mod- further decrease with glucose depletion (Fig. 1 A; Fig. S1 A; and ifications in spine morphology and synaptic plasticity. Fig. 2, A and B). Similar findings were obtained with 15 mM CPT1C contains two transmembrane domains connected by a 2-deoxy-D-glucose (2DG) treatment, a glucose analogue that short loop that allows the N-terminus and C-terminus to face the cannot be metabolized and competitively inhibits glycolysis cytoplasm. The C-terminal region of CPT1C is around 30 resi- (Fig. 1 B and Fig. S1 B). Glucose deprivation or 2DG treatments dues longer than that of the other 1A and 1B isoforms, suggesting are known to reduce intracellular malonyl-CoA levels through Casas et al. Journal of Cell Biology 2 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 phosphorylation of ACC (Wolfgang et al., 2007). Consistent To address this, a mutant CPT1C insensitive to malonyl-CoA with previous results, ACC, the enzyme that synthesizes (CPT1CM589S; Morillas et al., 2004; Rodrı́guez-Rodrı́guez et al., malonyl-CoA, showed elevated levels of phosphorylation under 2019) and a truncated CPT1C in which the last 39 aa were re- glucose depletion or 2DG treatment (Fig. 1, A and B, right placed by the FLAG tag (CPT1CΔCterm-FLAG) were transfected graphs), which is indicative of its inhibition. Importantly, in into HEK293 cells. Coimmunoprecipitation assays showed that both cases, these treatments resulted in no change in neuronal malonyl-CoA does not seem to facilitate CPT1C-GluA1 interac- survival (Fig. S2, C and D). Next, we analyzed the effect of tions, since no differences were observed after TOFA treatment leptin addition, an anorexigenic hormone that is known to (Fig. 2 A) or with the CPT1CM589S mutant compared with the increase malonyl-CoA levels (Gao et al., 2007; Wolfgang et al., WT CPT1C (Fig. 2 B). However, CPT1C lacking the C-terminus 2007). Consistent with malonyl-CoA levels playing a role in (CPT1CΔCterm) hardly coimmunoprecipitated with GluA1 (Fig. 2 GluA1 surface expression, 50 nM leptin for 30 min promoted B), suggesting that CPT1C was interacting with this AMPA an increase in GluA1 surface levels (Fig. 1 C, left; and Fig. S1 C), subunit through its C-terminal domain. When only the last 36 aa as previously reported (Moult et al., 2010), and a decrease in of CPT1C were fused directly to GFP (GFP-Cterm), the coim- ACC phosphorylation (Fig. 1 C, right). This effect was abol- munoprecipitation was not observed (Fig. 2 C), indicating that ished in neurons where CPT1C was knocked down (Fig. 1 C). the last 36 aa alone are necessary but not enough for the inter- These results are consistent with the hypothesis that malonyl- action. Similar interaction results were observed using fluo- CoA regulates GluA1 trafficking and, specifically, that rescence resonance energy transfer (FRET) assays in HEK cells: a malonyl-CoA down-regulation leads to a CPT1C-dependent positive interaction between GluA1 and CPT1C or CPT1CM589S, decrease in GluA1 at the PM. but not with CPT1CΔCterm-FLAG, was observed (Fig. 2 D). To test this hypothesis, we decided to directly down-regulate To further identify a putative interaction region between the malonyl-CoA levels through treatment with an ACC inhibitor, two proteins, in silico molecular dynamic simulations were 5-tetradecyloxy-2-furoic acid (TOFA). TOFA is an acetyl-CoA conducted taking the C-terminal cytosolic domain of GluA1 and analogue that cannot be further metabolized to malonyl-CoA. the C-terminal plus the catalytic globular cytosolic domain of In consequence, TOFA does not modify ACC phosphorylation CPT1C under consideration. 12 different starting positions, each levels. We verified that 20 µg/ml TOFA treatment for 2 h did of them differing in the initial GluA1 fragment positioning rel- reduce malonyl-CoA levels in cortical neurons (Fig. 1 D, right) ative to the CPT1C fragment and referred to as “approaching and that it did not produce neuronal death (Fig. S2 E). Fig. 1 D vectors,” were employed in order to most efficiently map all the and Fig. S1 D show that neurons treated with TOFA had signif- potential contact spaces (Knapp et al., 2018). Fig. 2 E shows that icantly reduced levels of GluA1 at the PM compared with control five of the approaching vectors presented a contact time per- cells. Importantly, this response was not detected in CPT1C si- centage >5%, which according to previous literature can be lenced neurons. The regulation of GluA1 trafficking by TOFA considered a statistically significant amount of interaction, treatment was reversible and was also observed in cultured enough to determine that contact exists (see Convertino et al., hippocampal neurons (Fig. S2, F and G). To increase rigor and 2009; Domı́nguez et al., 2015). A ribbon representation for the further underscore these results, immunocytochemistry and GluA1 and CPT1C interaction frame from the approaching surface biotinylation studies were also conducted in cortical vector 7 (the one with the highest contact time) is shown and neurons derived from CPT1C knockout (KO) mice embryos provides further evidence supporting a physical interaction (CPT1C−/−), with comparable results observed (Fig. S3, A and B). between GluA1 and CPT1C. Interestingly, this regulation was specific to the GluA1 subunit, Next, we tested if (1) the CPT1C C-terminal domain was re- as no changes in surface levels of AMPAR subunit GluA2 or quired for GluA1 transport to the PM and (2), even though NMDA receptor subunit 2A (GluN2A) were observed after TOFA malonyl-CoA binding is not a determinant for CPT1C-GluA1 in- treatment (Fig. S3, C and D). Collectively, these results strongly teractions, if its sensing is necessary to properly regulate GluA1 link CPT1C and malonyl-CoA levels as part of a rheostat for trafficking. To this end, CPT1C−/− neurons were infected at controlling GluA1 subunit surface expression. 7 d in vitro (DIV) with lentiviral vectors carrying the trun- cated CPT1CΔCterm-FLAG, the CPT1CM589S mutant, or the C-terminal domain of CPT1C is necessary for interaction with WT CPT1C. The expression of these constructs was con- GluA1 firmed a week later (Fig. 3 A). Transfected neurons were Given that CPT1C strongly interacts with GluA1 (Schwenk et al., then treated with TOFA or vehicle, and GluA1 surface levels 2012; Fadó et al., 2015) and, as outlined above, malonyl-CoA is a were analyzed by immunostaining. As expected, KO neurons regulator of GluA1 trafficking in a CPT1C-dependent manner, we showed a robust reduction of GluA1 levels at the PM com- next analyzed whether malonyl-CoA could regulate CPT1C– pared with WT, which was rescued by full-length CPT1C GluA1 interaction. Moreover, since the last 36 aa of the protein (Fig. 3 B). However, neither the C-terminal–truncated nor are one of the main differences between CPT1C and the mito- the malonyl-CoA–insensitive mutated forms were able to chondrial isoforms CPT1A and CPT1B, which lack that region restore GluA1 at the PM. Likewise, TOFA-mediated down- (Price et al., 2002; Sierra et al., 2008), we considered the regulation of GluA1 surface levels was only effective in C-terminus as a possible region facilitating CPT1C-GluA1 full-length CPT1C-expressing neurons. These data support interactions. the hypothesis that the C-terminal domain of CPT1C and Casas et al. Journal of Cell Biology 3 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 1. CPT1C is required for metabolic regulation of GluA1 surface levels. (A–D) Surface GluA1 quantification. CPT1C-silenced (shCPT1C; see Fig. S2 A) or control (short hairpin-loop RNA with a random sequence; shRandom) cortical neurons at 14–15 DIV were submitted to glucose deprivation for 2 h (A), 2DG treatment (15 mM, 2 h; B), leptin treatment (50 nM; 30 min; C), or TOFA treatment (20 µg/ml, 2 h; D), respectively. Then, nonpermeabilized cells were fixed and processed by immunocytochemistry using an anti-GluA1 antibody against the N-terminal domain (extracellular). Lentivirus plasmids (pLVTHM-Random-IRES- GFP and pLVTHM-CPT1C-IRES-GFP) used in these experiments produced soluble GFP; therefore, it was feasible to draw masks for the dendrites using Imaris software. Representative images are shown. In each condition, we analyzed 40 neurons from two independent experiments performed by biological duplicates. Results in left graphs are given as the mean ± SEM (two-way ANOVA followed by Bonferroni’s comparison test; **, P < 0.05; ***, P < 0.001). (A) shRandom Casas et al. Journal of Cell Biology 4 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 25 mM glucose (1.00 ± 0.08, n = 50), shRandom 0 mM glucose (0.55 ± 0.08, n = 45); shCPT1C 25 mM glucose (0.58 ± 0.08, n = 40), and shCPT1C 0 mM glucose (0.82 ± 0.08, n = 44). (B) shRandom + vehicle (1.00 ± 0.07, n = 53), shRandom + 2DG (0.73 ± 0.056, n = 50), shCPT1C + vehicle (0.36 ± 0.05, n = 44), and shCPT1C + 2DG (0.78 ± 0.05, n = 51). (C) shRandom + vehicle (1.00 ± 0.14, n = 51), shRandom + leptin (1.65 ± 0.23, n = 52), shCPT1C + vehicle (0.69 ± 0.11, n = 52), and shCPT1C + leptin (0.66 ± 0.09, n = 52). (D) shRandom + vehicle (1.00 ± 0.10, n = 57), shRandom + TOFA (0.46 ± 0.07, n = 52), shCPT1C + vehicle (0.49 ± 0.06, n = 50), and shCPT1C + TOFA (0.64 ± 0.09, n = 55). Treatment efficiency was determined by phosphorylated ACC (pACC) using Western blotting method in cortical neurons under glucose deprivation (n = 6; 0.96 ± 0.23 for 25 mM glucose and 1.94 ± 0.57 for 0 mM glucose), 2DG treatment (n = 3–6; 0.97 ± 0.28 for vehicle and 2.09 ± 0.21 for 2DG), or leptin treatment (n = 5; 1.16 ± 0.50 for vehicle and 0.37 ± 0.12 for leptin). Results in right graphs are the mean ± SD (Mann- Whitney U test; **, P < 0.01 and *, P < 0.05). TOFA efficiency was evaluated in cortical neurons by malonyl-CoA recycling assay. Values are the mean ± SD of four independent experiments, three of them performed by biological duplicates (n = 7 samples per condition; two-sided Student’s t test; P = 0.049; 1.00 ± 0.05 for vehicle and 0.83 ± 0.06 for TOFA). Scale bars = 5 µm. malonyl-CoA binding are necessary for the interaction and of the TGN lipid phosphatase SAC1, which is a strong regulator trafficking of GluA1. of PI(4)P (reviewed by Del Bel and Brill, 2018) and has been reported to interact with CPT1C (Brechet et al., 2017). Thus, we CPT1C silencing or decreased malonyl-CoA production trigger explored a potential role for malonyl-CoA sensing and the CPT1C GluA1 retention at the TGN by decreasing PI(4)P C-terminus in tuning CPT1C-SAC1 interactions. Next, we wanted to determine the precise molecular mechanism Coimmunoprecipitation studies were performed in through which CPT1C regulates GluA1 trafficking. To begin, we cortical neurons with endogenous proteins and in HEK focused on the TGN, a central trafficking hub for GluA1 receptors cells overexpressing SAC1 and WT CPT1C, CPT1CM589S, or on their way to the PM (reviewed by Jacobi and von Engelhardt, CPT1CΔCterm-FLAG. Results presented in Fig. 5 A confirm the 2018; Yang et al., 2013). Treating cortical neurons with TOFA to previously published interaction between SAC1 and CPT1C. decrease malonyl-CoA production resulted in the significant ac- Treatment with TOFA (to reduce malonyl-CoA; Fig. 5 A) or cumulation of GluA1 in the TGN (Fig. 4 A). Additionally, knock- CPT1CM589S expression (Fig. 5 B) both failed to alter interac- down of CPT1C resulted in a similar GluA1 retention pattern at the tions between CPT1C and SAC1. Moreover, the last 36 aa of TGN. These data point to the TGN as a potential location where CPT1C were not required for an efficient interaction between CPT1C and malonyl-CoA exert influence over GluA1 trafficking. CPT1C and SAC1 (Fig. 5 B). Similar results were obtained using Previously, it was demonstrated that TGN PI(4)P controls FRET assays (Fig. 5 C). These results indicate that the interac- GluA1 vesicular release from the TGN (Yang et al., 2013); tion of CPT1C with SAC1 is not modulated by malonyl-CoA and therefore, we asked if CPT1C or malonyl-CoA fluctuations can that the C-terminal domain of CPT1C does not appear to play a alter TGN PI(4)P levels. To test this hypothesis, WT cortical critical role. neurons were treated with TOFA or the short hairpin-loop RNA of CPT1C (shCPT1C), and TGN PI(4)P levels were quantified using SAC1 is a key modulator of GluA1 trafficking downstream of immunocytochemistry. As shown in Fig. 4 B, both knocking the malonyl-CoA/CPT1C axis down CPT1C or treatment with TOFA resulted in a significant Next, we wanted to confirm whether SAC1 was actually regu- decrease of PI(4)P levels at the TGN. TOFA-dependent decreases lating surface GluA1 levels downstream of the malonyl-CoA/ in TGN pools of PI(4)P were detected only in WT neurons, with CPT1C axis. To this end, surface GluA1 levels were quantified by no further decrease observed in CPT1C-silenced neurons. immunostaining in SAC1-silenced and SAC1-overexpressing Changes in PI(4)P occurred independently of alterations in the cortical neurons submitted to TOFA treatment to block the size of the TGN (Fig. 4 C). Underscoring these data, similar re- synthesis of malonyl-CoA. sults were obtained when comparing WT and CPT1C−/− neurons In agreement with Yang et al. (2013), SAC1-silencing en- and using a specific PI(4)P biosensor (mCherry-P4M; Fig. S4 A). hanced Golgi PI(4)P pools, as well as GluA1 surface levels (Fig. 6, To further corroborate the involvement of CPT1C in controlling A–C). Interestingly, the silencing of SAC1 abrogated the effects of Golgi PI(4)P pools in a malonyl-CoA–dependent manner, we TOFA on surface GluA1 expression. Conversely, SAC1 over- expressed either the insensitive malonyl-CoA CPT1CM589S, the expression reduced both TGN PI(4)P levels and GluA1 levels at the truncated CPT1CΔCterm-FLAG, or the WT form of CPT1C in KO PM compared with control neurons. Moreover, TOFA treatment neurons. As expected, the full-length CPT1C form was able to did not have a significant effect on SAC1-overexpressing cells rescue PI(4)P levels in the TGN compartment (Fig. 4 D), while (Fig. 6, D–F). Neither SAC1 silencing nor SAC1 overexpression neither the malonyl-CoA–insensitive nor the truncated form changed TGN volume in neurons (Fig. S4 B). These data indicate could restore Golgi PI(4)P pools in CPT1C−/− neurons. that SAC1 is downstream of the malonyl-CoA/CPT1C axis, which is Taken together, our data indicate that CPT1C sensing of crucial for the maintenance of GluA1 levels at the PM in neurons. malonyl-CoA regulates GluA1 retention at the TGN through the regulation of PI(4)P levels. Moreover, results suggest that the CPT1C regulates SAC1 translocation to ER-TGN contacts C-terminal domain of CPT1C is necessary for this regulation. Previous studies described SAC1 translocation to the TGN under stressful conditions, which correlated with a decrease in Golgi Characterization of CPT1C–SAC1 interaction PI(4)P levels and impaired vesicular transport to the PM As mentioned above, CPT1C regulates PI(4)P levels in the TGN (Blagoveshchenskaya et al., 2008; Yang et al., 2013). For this compartment. This finding can be explained by the involvement reason, we hypothesized that CPT1C could promote SAC1 Casas et al. Journal of Cell Biology 5 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 2. CPT1C–GluA1 interaction is dependent on CPT1C C-terminal domain but not on malonyl-CoA levels. (A) Quantification of coimmunopreci- pitated GluA1 after CPT1C immunoprecipitation in TOFA-treated cortical neurons. Results are the mean ± SD from three independent experiments performed Casas et al. Journal of Cell Biology 6 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 by biological duplicates (n = 5 or 6 samples per condition; Mann-Whitney U test; P = 0.6991; 1.00 ± 0.57 for vehicle and 0.97 ± 0.24 for TOFA). (B) Coim- munoprecipitated CPT1C quantification versus immunoprecipitated GluA1 in HEK cells cotransfected with GluA1 and WT CPT1C, CPT1CM589S (a mutated form insensitive to malonyl-CoA), or CPT1CΔCterm-FLAG (a truncated form in which the last 39 aa have been substituted by the tag FLAG). Whole lysates (input) were used to corroborate a proper transfection. Western blots using antibodies against CPT1C, FLAG tag, and GluA1 were performed. Values are the mean ± SD from three independent experiments performed by biological replicates (one-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001). GluA1 + WT CPT1C (1.00 ± 0.34, n = 8), GluA1 + CPT1CM589S (0.85 ± 0.28, n = 4), and GluA1 + CPT1CΔCterm-FLAG (0.12 ± 0.14, n = 6). (C) Coimmunoprecipitated GFP, CPT1C-GFP, or GFP-Cterm (GFP fused with the last 36 aa of CPT1C) in GluA1 immunoprecipitated samples from transfected HEK cells. Representative Western blots from four independent experiments are shown. (D) Percentage of FRET effciency between CFP-GluA1 and different YFP-CPT1C forms in transfected HEK cells (blue bars). KDEL (a short sequence with ER localization)/CPT1C and calnexin (an ER resident protein)/CPT1C pairs were used as negative controls for CPT1C interaction (black bars). GluA1/WT CPT1C pair was used as a positive control (solid blue). The dashed line represents the higher percentage of FRET of the negative control pairs. Results are the mean ± SEM from three independent experiments performed by biological duplicates (one-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001 versus KDEL/CPT1C pair; ###, P < 0.001 versus calnexin/CPT1C pair; $$$, P < 0.001 versus GluA1/CPT1C pair). KDEL + CPT1C (2.13 ± 0.31, n = 36), calnexin + CPT1C (5.18 ± 0.30, n = 36), GluA1 + CPT1C (8.70 ± 0.65, n = 30), GluA1 + CPT1CΔCterm-FLAG (5.87 ± 0.59, n = 30), and GluA1 + CPT1CM589S (9.51 ± 0.56, n = 26). Scale bars = 10 µm; scale bars of inset magnifications = 2 µm. ns, not significant. (E) C-terminal CPT1C region encompassing the catalytic globular domain and the cytosolic GluA1 domain was used for protein–protein interaction simulations. The percentage of contact time for each of the 12 approaching vectors analyzed is shown in the table. The two approaching vectors with the highest contact time are shown in bold. A ribbon representation of the GluA1 (orange) and CPT1C (blue) interaction frame from approach vector 7 is shown. The main contact region is highlighted by the oval and corresponds to residues 843–851 for GluA1 and 786–792 for CPT1C (localized in the C-terminal domain, which ends in residue 803). transport to the TGN in low malonyl-CoA conditions and, activity. Interestingly, in this context TOFA treatment did not consequently, decrease both Golgi PI(4)P pools and GluA1 exert any effect in control cells, but it partially abolished the trafficking to the PM. However, our data do not support such effects of CPT1C on PI(4)P levels. These results suggest that model, since SAC1 did not increase either at the TGN after TOFA CPT1C may be regulating not only SAC1 subcellular distribution treatment or in CPT1C knockdown neurons (Fig. 7 A). but also its PI(4)P phosphatase activity in response to malonyl- More recently, other authors have demonstrated that SAC1 CoA changes to control GluA1 trafficking patterns to the PM. can act in trans on the TGN PI(4)P pool, across ER-TGN junctions (Venditti et al., 2019). To address whether CPT1C can modulate CPT1C regulates SAC1 PI(4)P phosphatase activity in a SAC1 recruitment to these contact sites under low malonyl-CoA malonyl-CoA–dependent manner levels, we studied SAC1 in ER-TGN junctions using CRISPR- To further investigate a role for CPT1C in regulating SAC1 Cas–edited HEK cells that have endogenous SAC1 tagged with phosphatase activity, we took several independent approaches. GFP (Zewe et al., 2018). Our results clearly demonstrate that First, we directly measured the specific PI(4)P phosphatase ac- TOFA treatment favors ER-TGN junction creation in a CPT1C- tivity of SAC1 (Mani et al., 2007) in Hela cells stably expressing independent manner (Fig. 7, B and C; and Fig. S4 C), with en- CPT1C or CPT1CM589S under different malonyl-CoA conditions. dogenous SAC1 only being enriched at these contact sites under As shown in Fig. 9 A, CPT1C expression decreased SAC1 activity, low malonyl-CoA levels and CPT1C overexpression (Fig. 7 D). a reduction that was antagonized by TOFA treatment and was Considering that the CPT1C–SAC1 interaction is not affected by absent following the expression of the malonyl-CoA–insensitive malonyl-CoA (Fig. 6), we decided to analyze whether CPT1C also CPT1C (CPT1CM589S; Fig. 9 B). In parallel experiments per- accumulated at ER-TGN contact sites under low malonyl-CoA formed in mice brains, we determined that SAC1 phosphatase levels. In line with this hypothesis, CPT1C levels at ER-TGN activity is significantly increased in CPT1C−/− brains compared junctions increased after TOFA treatment (Fig. 7 E). The per- with WTs (Fig. 9 C). Moreover, brains from 16 h–fasted animals, centage of CPT1C found in the ER-TGN junctions (<10%) was a protocol that strongly decreases malonyl-CoA levels in the similar to that found in the TGN (Fig. S4 D), suggesting that cortex, hippocampus, hypothalamus, and cerebellum (Tokutake CPT1C does not translocate to the TGN but accumulates at the ER et al., 2010), had increased SAC1 phosphatase activity (Fig. 9 D). side of the ER-TGN junctions. These findings suggest that low As a second, indirect approach to measure SAC1 activity, we malonyl-CoA levels increase ER-TGN contact sites favoring measured the intensity and distribution of a cellular PI(4)P CPT1C/SAC1 recruitment to these junctions. biosensor (mCherry-P4M) in live HEK293 cells overexpressing To verify that GluA1 trafficking is regulated by ER-TGN CPT1C, with and without the addition of TOFA. Quantitative junctions, we next decided to artificially enhance ER-TGN con- analysis from these cells revealed that CPT1C overexpression tacts using the rapamycin-induced dimerization system (Dickson increased total cellular PI(4)P levels, consistent with CPT1C in- et al., 2016, 2014). Increasing ER-TGN junctions resulted in the hibiting SAC1 activity, with this increase in PI(4)P inhibited expected decrease of TGN PI(4)P levels following the addition of following TOFA treatment (Fig. 9 E). The increase in cellular rapamycin (Fig. 8 A; and Fig. S5, A and B) and a corresponding PI(4)P was paralleled by a significant increase in PM PI(4,5)P2 decrease in the levels of GluA1 at the PM (Fig. 8 B), indicating (Fig. 9 F) and suggests that TGN PI(4)P is a source for the that GluA1 trafficking can be influenced by ER-TGN–dependent maintenance of PM PI(4,5)P2 (Dickson et al., 2014). To confirm regulation of TGN PI(4)P levels. that the CPT1C-dependent control of SAC1 can regulate PI(4)P Surprisingly, Myc-CPT1C expression totally abolished and PI(4,5)P 2 levels in the brain, ultra performance liquid rapamycin-dependent PI(4)P decay and surface GluA1 decrease. chromatography–tandem mass spectrometer (UPLC MS/MS) This result suggests that CPT1C could be an inhibitor of SAC1 was performed on lipids extracted from WT and CPT1C−/− Casas et al. Journal of Cell Biology 7 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 3. CPT1C C-terminal and binding to malonyl-CoA are needed to regulate GluA1 trafficking. (A and B) Surface GluA1 level recovery by CPT1C reexpression in KO neurons. CPT1C KO cortical neurons were transduced at 7 DIV with lentiviral vectors carrying CPT1C, CPT1CM589S, CPT1CΔCterm-FLAG, or an EV. WT neurons transduced with the EV were used as the control. At 14-15 DIV, cells were treated with TOFA (20 µg/ml, 2 h) or the vehicle (DMSO; 1:500). Surface GluA1 levels were detected and quantified as in Fig. 1. Proper overexpression of different CPT1C forms was corroborated by Western blot in A. Data in B represent the mean ± SEM from three independent experiments performed by biological duplicates (one-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001). WT + EV + vehicle (1.00 ± 0.02, n = 17), KO + EV + vehicle (0.55 ± 0.05, n = 80), KO + EV + TOFA (0.54 ± 0.069, n = 63), KO + CPT1C + vehicle (1.13 ± 0.07, n = 85), KO + CPT1C + TOFA (0.35 ± 0.04, n = 68), KO + CPT1CΔCterm-FLAG + vehicle (0.52 ± 0.08, n = 54), KO + CPT1CΔCterm-FLAG + TOFA (0.57 ± 0.07, n = 56), KO + CPT1CM589S + vehicle (0.38 ± 0.05, n = 52), and KO + CPT1CM589S + TOFA (0.38 ± 0.07, n = 33). Scale bars = 5 µm. brains. Consistent with biosensor data, mass-spec analysis of different nutritional conditions. In the present work, we address phosphoinositides revealed that CPT1C−/− brains have signifi- both questions and demonstrate that CPT1C regulates GluA1 cantly reduced levels of phosphatidylinositol monophosphate trafficking from the TGN toward the PM by modulating SAC1 (PIP) and phosphatidylinositol bisphosphate (PIP2), relative to phosphatase activity in a malonyl-CoA–dependent manner (see WT animals (Fig. 9, H–J; and Fig. S6). the proposed model at Fig. 10). Collectively, these data support the hypothesis that CPT1C is a We demonstrate that glucose depletion, glycolysis inhibition, metabolic regulator of SAC1 PI(4)P phosphatase activity in a or leptin addition—metabolic challenges known to modify in- malonyl-CoA–dependent manner. tracellular malonyl-CoA levels (Gao et al., 2007; Wolfgang et al., 2007)—modulate GluA1 surface expression in cortical neurons. Moreover, direct down-regulation of malonyl-CoA levels Discussion through the inhibition of ACC recapitulates these effects, con- AMPARs mediate excitatory neurotransmission in the brain and firming the important role of malonyl-CoA in GluA1 trafficking. play a key role in most brain functions, including cognition. Interestingly, CPT1C-deficient neurons did not respond to any of Recent studies have demonstrated the involvement of CPT1C in the treatments above, and the malonyl-CoA–insensitive CPT1C GluA1 trafficking, the main component of AMPARs (Brechet was unable to restore GluA1 at the PM in CPT1C KO neurons. et al., 2017; Fadó et al., 2015; Gratacòs-Batlle et al., 2015, 2018). These findings clearly indicate that surface GluA1 is finely reg- However, little is known about the precise molecular mecha- ulated by CPT1C-sensing of malonyl-CoA. Importantly, malonyl- nism and whether this function is linked to CPT1C’s capacity to CoA is an intermediary shared by the metabolism of lipids bind and sense malonyl-CoA fluctuations in the brain across and glucose, which makes it a precise indicator of nutritional Casas et al. Journal of Cell Biology 8 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 4. CPT1C regulates GluA1 and PI(4)P levels at the TGN in a malonyl-CoA–dependent manner. (A) GluA1 retention at the TGN. CPT1C was silenced in WT cortical neurons by the transduction of shCPT1C-carrying lentivirus. A short hairpin-loop RNA with a random sequence (shRandom) was used as a Casas et al. Journal of Cell Biology 9 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 control. At 14–15 DIV, cells were treated with TOFA (20 µg/ml, 2 h), fixed, and permeabilized in order to analyze the levels of GluA1 in the TGN by coim- munocytochemistry. TGN38 protein was used to localize and define the TGN compartment. The mean intensity of GluA1 in the TGN compartment per each neuron soma was analyzed. Graph shows the mean ± SEM of 20 cells from two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; **, P < 0.01 and ***, P < 0.001). shRandom + vehicle (1.00 ± 0.02, n = 19), shRandom + TOFA (1.12 ± 0.03, n = 27), shCPT1C + vehicle (1.29 ± 0.06, n = 23), and shCPT1C + TOFA (1.17 ± 0.06, n = 21). Scale bars = 7 µm. (B) Quantification of PI(4)P levels in the TGN compartment by immunocytochemistry in control or CPT1C-silenced neurons after TOFA treatment (20 µg/ml, 2 h). PI(4)P was detected by immunocytochemistry using a specific antibody purchased from Echelon (Z-P004). TGN46 protein was used to localize and define the TGN compartment. Below, black-and-white images show the ROI generated by TGN-PI(4)P colocalizing pixels. The mean intensity of PI(4)P in the TGN compartment per each neuron soma was analyzed. Data represent the mean ± SEM from four independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and ***, P < 0.001). shRandom + vehicle (1.00 ± 0.02, n = 42), shRandom + TOFA (0.95 ± 0.01, n = 48), shCPT1C + vehicle (0.88 ± 0.03, n = 39), and shCPT1C + TOFA (0.94 ± 0.04, n = 37). Scale bars = 7 µm. n.s, not significant. (C) Quantification of the TGN volume per each neuron soma in the same experiments and conditions as in B. Values are the mean ± SEM from two independent experiments performed by biological replicates (two-way ANOVA; P > 0.05). shRandom + vehicle (1.00 ± 0.09, n = 22), shRandom + TOFA (1.12 ± 0.10, n = 23), shCPT1C + vehicle (1.23 ± 0.11, n = 19), and shCPT1C + TOFA (1.29 ± 0.13, n = 18). (D) Quantification of PI(4)P levels in the TGN compartment in CPT1C KO cortical neurons overexpressing or not CPT1C full or different mutated CPT1C forms and submitted to TOFA treatment (20 µg/ml, 2 h). Analysis was performed as in Fig. 4 B. On the right, black-and-white images show the ROI generated by TGN-PI(4)P colocalizing voxels. Results are shown as the mean ± SEM from two independent experiments performed by biological duplicates (one-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001). WT + EV + vehicle (1.00 ± 0.05, n = 7), KO + EV + vehicle (0.86 ± 0.01, n = 43), KO + EV + TOFA (0.84 ± 0.02, n = 40), KO + CPT1C + vehicle (0.98 ± 0.02, n = 36), KO + CPT1C + TOFA (0.81 ± 0.03, n = 46), KO + CPT1CΔCterm-FLAG + vehicle (0.71 ± 0.06, n = 26), and KO + CPT1CM589S + vehicle (0.54 ± 0.08, n = 25). Scale bars = 7 µm. conditions. In fact, malonyl-CoA levels in the brain highly de- energy balance and supply to the brain, including mechanisms crease by fasting, while they increase after refeeding (Tokutake independent of malonyl-CoA. In any case, our results clearly et al., 2010). Therefore, we can say that nutrients impact GluA1 demonstrate that GluA1 trafficking in primary cultured cortical trafficking through CPT1C-sensing of malonyl-CoA. Since the neurons is sensitive to cell-autonomous intracellular malonyl- energy sensor AMPK phosphorylates and inhibits the activity of CoA fluctuations. ACC, the enzyme that catalyzes the conversion of acetyl-CoA to The second crucial result of our work is that CPT1C regulates malonyl-CoA, it suggests that the CPT1C/SAC1 axis is down- GluA1 trafficking through SAC1-dependent modulation of TGN stream of the AMPK pathway. To date, very few articles have PI(4)P levels. To our knowledge, this is the first report of a reported the mechanism by which nutrients regulate GluA1 negative regulator of SAC1 enzymatic activity (for an exhaustive trafficking. A recent paper demonstrates that a long-time high- review on SAC1, see Del Bel and Brill, 2018). SAC1 PI(4)P fat diet (during 6 wk in mice or for 24 h in cultured neurons) phosphatase activity assays in cell lines and brain tissues hinders GluA1 trafficking to the PM by a mechanism involving demonstrate that SAC1 activity is regulated by CPT1C in an GluA1 palmitoylation (Spinelli et al., 2017). Here, we demonstrate energy-dependent manner. When malonyl-CoA levels are de- that short-term fluctuations of malonyl-CoA, such as glucose creased (by TOFA treatment in cultured neurons or by fasting depletion for 2 h, regulates GluA1 trafficking by a mechanism in whole animals), the CPT1C-mediated inhibition of SAC1 is involving CPT1C regulation of SAC1. released and phosphatase activity restored. Moreover, SAC1 Although the brain only accounts for 2% of body mass, it activity regulation correlates with PI(4)P levels in cell lines and consumes up to 25% of total glucose metabolized by the body. in the total brain. This is of crucial importance, since SAC1 has Moreover, neurons use 85% of this energy mainly to maintain been involved in the regulation of other PI(4)P pools and glutamatergic transmission (Harris et al., 2012). Therefore, functions in the cell, such as the regulation of cholesterol neurons need to sense their energy status in a very precise way transport at the ER-TGN contacts (Mesmin et al., 2013), phos- to turn on or turn off synaptic activity and preserve brain phatidylserine transport at the ER-PM contacts (Chung et al., function. Interestingly, one of the most important cell energy 2015; Von Filseck et al., 2015), the retromer function at the ER- sensors and regulators during synaptic activation is AMPK endosome contacts (Dong et al., 2016), or ceramide de novo (Marinangeli et al., 2018). Pathological conditions that undergo synthesis at the ER (Breslow et al., 2010). Therefore, it is rea- with deficits in energy supply to the brain have been associated sonable to think that CPT1C sensing of malonyl-CoA could also with impaired cognitive performance. Examples of this are se- regulate some other functions of SAC1. In fact, this could ac- vere undernutrition in anorexic patients (reviewed by Beilharz count for the reduced levels of ceramide found in brain tissues et al., 2015), exhaustive exercise (Rosa et al., 2007; Zimmer et al., of CPT1C KO mice (Carrasco et al., 2013). 2017), brain metabolic hypoactivity in Alzheimer’s disease, and One controversial issue that remains is how SAC1 dephos- sleep deprivation (Dworak et al., 2010; Nikonova et al., 2010). phorylates PI(4)P at the TGN. Some authors claim that SAC1 Recent integrative analyses drawing on multiple sources of data dephosphorylates PI(4)P in trans at membrane contact sites, imply that AMPAR trafficking is one of the crucial determinants while others suggest it acts in cis after its translocation to the underlying the negative effects of insufficient sleep (Diessler TGN. Additionally, these two models of SAC1 activity do not have et al., 2018). However, it is well known that aerobic exercise to be mutually exclusive, with a clear compromise between the and caloric-restricted diets promote synaptic incorporation of two models being that SAC1 can act in both configurations. Our GluA1 and improve cognition (Thacker et al., 2019; Ouyang et al., stressful conditions in cortical neurons were not able to induce a 2017). This contradiction can be explained by the existence of shuttling of SAC1 from the ER to the TGN, contrary to some several mechanisms for organ and location-specific control of previous studies performed in nonexpressing CPT1C cells Casas et al. Journal of Cell Biology 10 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 5. CPT1C–SAC1 interaction is dependent on neither malonyl-CoA levels nor CPT1C C-terminal domain. (A) Quantification of coimmunopreci- pitated CPT1C after SAC1 immunoprecipitation in TOFA-treated cortical neurons. Results are the mean ± SD from two independent experiments performed by biological replicates (n = 5 samples per condition; Mann-Whitney U test; P = 0.4206; 1.00 ± 0.12 for vehicle and 0.84 ± 0.36 for TOFA). (B) Quantification of coimmunoprecipitated CPT1C by SAC1 immunoprecipitation in cotransfected HEK cells. Western blot using antibodies against CPT1C, FLAG tag, and SAC1 was performed. Values are the mean ± SD from three independent experiments (one-way ANOVA followed by Bonferroni’s comparison test; P > 0.05; SAC1 + WT CPT1C 1.00 ± 0.26, n = 5; SAC1 + CPT1CM589S 1.08 ± 0.63, n = 3; and SAC1 + CPT1CΔCterm-FLAG 0.97 ± 0.28, n = 5). (C) Percentage of FRET efficiency between SAC1 and different CPT1C mutated forms in transiently transfected HEK293 cells (orange bars). KDEL/CPT1C and calnexin/CPT1C pairs were used as negative controls for CPT1C interaction (black bars). The GluA1/WT CPT1C pair was used as a positive control (blue bar). The dashed line indicates the highest Casas et al. Journal of Cell Biology 11 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 percentage of FRET corresponding to a negative interaction. Results are the mean ± SEM from three independent experiments performed by biological replicates (n = 30 cells per condition; one-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001 versus KDEL/CPT1C pair; ##, P < 0.01, and ###, P < 0.001 versus calnexin/CPT1C pair). KDEL + CPT1C (1.91 ± 0.33, n = 27), calnexin + CPT1C (5.58 ± 0.32, n = 31), GluA1 + CPT1C (8.70 ± 0.65, n = 30), SAC1 + CPT1C (8.94 ± 0.48, n = 31), SAC1 + CPT1CΔCterm-FLAG (7.98 ± 0.45, n = 36), and SAC1 + CPT1CM589S (10.42 ± 0.62, n = 31). Scale bars = 10 µm; scale bars of inset magnifications = 2 µm. ns, not significant. (Faulhammer et al., 2007; Blagoveshchenskaya et al., 2008; The last interesting issue is that both CPT1C and SAC1 form Yang et al., 2013). Our data align with recent literature re- part of the AMPAR macrocomplexes at the ER in neurons. Re- porting the ability of SAC1 to metabolize the TGN PI(4)P pool cently, Schwenk et al. (2019) showed that CPT1C, specifically across ER-TGN junctions (Venditti et al., 2019). Underscoring bound to ferric-chelate reductase 1-like, drives the assembly of this statement, artificial induction of ER-TGN contacts using GluAs into functional tetramers at the ER. Moreover, a role for the rapamycin platform induced TGN PI(4)P depletion and a CPT1C has been suggested in the palmitoylation state of GluA1 consequent reduction of GluA1 at the PM, which was abolished (Gratacòs-Batlle et al., 2018). Here, we complete the picture of by CPT1C only under high malonyl-CoA levels. Notably, we CPT1C function and describe a new role of CPT1C linked to SAC1 found that inhibition of malonyl-CoA synthesis by TOFA in- at the ER-TGN contacts. We show that the CPT1C–SAC1 complex creased ER-TGN contacts and triggered CPT1C translocation to regulates the trafficking of GluA1-containing AMPAR from the ER-TGN junctions together with SAC1, suggesting that both TGN to the PM in response to nutritional changes and unravel proteins translocate together. These findings indicate that the molecular mechanism involved. Interestingly, SAC1 is not GluA1 trafficking in neurons is regulated by SAC1 activity at found in AMPAR macrocomplexes of CPT1C KO mice (Brechet ER-TGN junctions under the control of CPT1C. et al., 2017), indicating that the role of SAC1 is necessarily The fact that GluA1 abundance at the PM decreased by 50% linked to CPT1C. The interacting region between CPT1C and after TOFA treatment or in CPT1C-silenced neurons while at the SAC1 is currently unknown. Using experimental and in silico TGN it only increased around 10% suggests that CPT1C not only approaches, we have identified the C-terminal region of regulates the canonical trafficking pathway but also may be CPT1C as being essential for GluA1 interaction and its trans- involved in other pathways, such as the Golgi-bypass route. port to the PM. Other authors have demonstrated (using nu- Additionally, it has been published that PI(4)P contributes to the clear magnetic resonance spectroscopy) that the N-terminal PI(4,5)P2 pool at the PM (Dickson et al., 2014), which is known to domain of CPT1C (a peptide of 50 residues) can switch be- directly bind and stabilize surface GluA1 and therefore avoid its tween two different conformations depending on malonyl- internalization (Seebohm et al., 2014). The fact that CPT1C favors CoA (Rao et al., 2011; Samanta et al., 2014). Collectively, re- an increase in PI(4,5)P2 at the PM in a malonyl-CoA–dependent sults support a model in which CPT1C binds GluA1 through its manner reveals that CPT1C probably modulates GluA1 surface C-terminal domain, while its N-terminal domains acts as a expression by also regulating its PI(4,5)P2-dependent stabiliza- sensor of malonyl-CoA. CPT1C will change its conformation in tion at the PM. response to nutritional scarcity or energy supply deficits, Surprisingly, neither GluA2 nor GluN2A surface levels in which will regulate SAC1 activity and, in consequence, block cortical neurons were altered after 2 h of TOFA treatment, the exit of GluA1 from the TGN. suggesting that CPT1C-mediated regulation of GluA1 trafficking In summary, our study reveals that CPT1C sensing of is specific to this AMPAR subunit. Therefore, we think that malonyl-CoA regulates GluA1-containing AMPAR trafficking in CPT1C is regulating specifically GluA1 homomeric AMPARs, neurons through the regulation of SAC1 activity. Moreover, it which have higher calcium permeability and larger channel discloses the first inhibitor of SAC1 PI(4)P phosphatase activity conductance than heteromeric receptors containing calcium- and supports the model of trans-activity of SAC1 at the ER-TGN impermeable GluA2 subunits. In the case of GluA2, it is estab- contact sites. Considering that CPT1C has been involved in lished that the exit from the ER toward the PM is slower than synaptogenesis, glutamatergic transmission, and learning pro- GluA1 (Henley and Wilkinson, 2013); for GluN2A, not much is cesses, the present work provides new molecular evidence re- known about the timing of its turnover. Perhaps longer times garding how CPT1C senses nutritional conditions to control of TOFA treatment are required to observe changes in their synaptic function and cognition. transport to the PM, but TOFA treatment can become toxic over prolonged treatment periods, impairing our ability to test this idea. It has been suggested that coat molecules and Materials and methods post-Golgi carriers may be sensitive to different PI(4)P levels Animals and pools at the TGN, leading to a selective vesicular releasing CPT1C KO mice were developed as described previously (Carrasco (Venditti et al., 2019). Moreover, different motor proteins are et al., 2012). After eight backcrosses with C57BL/6J mice, litter- involved in the vesicular transport of GluA1, GluA2, and mate homozygous CPT1C KO and WT mice were crossed sepa- NMDA receptor (Kneussel et al., 2014), which could also favor rately to obtain a line for each genotype that was used in the individual/selective trafficking. Clearly, additional experi- experiments. All animal procedures met the guidelines in ments are warranted to further unravel the mechanism of this Spanish legislation (Boletı́n oficial del estado 32/2007) and were selective trafficking. approved by the local ethics committee. Casas et al. Journal of Cell Biology 12 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 6. SAC1 is downstream of malonyl-CoA/CPT1C axis in cultured cortical neurons. (A and B) Increasing volumes of shSAC1 lentiviruses were used to knock down SAC1 expression in cultured cortical neurons at 8 DIV. A short hairpin-loop RNA with a random sequence (shRandom) was used as a control. shSAC1 silencing efficiency was checked at 15 DIV by quantifying the decrease of SAC1 protein levels by Western blot (A) and by the increase in TGN PI(4)P levels measured by immunocytochemistry (B; scale bars = 7 µm). TGN46 was used as a marker of TGN. Values are the mean ± SD from two independent experiments (n = 8–12 cells per condition; two-sided Student’s t test; ***, P = 0.0009; 1.00 ± 0.23 for shRandom and 1.35 ± 0.18 for shSAC1). (C) SAC1 silencing increased GluA1 surface levels and abolished the effect of TOFA treatment. Surface GluA1 after TOFA treatment (20 µg/ml, 2 h) was detected in non- permeabilized neurons by immunocytochemistry. Results are the mean ± SEM from two independent experiments performed by biological duplicates (two- way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and **, P < 0.01). shRandom + vehicle (1.00 ± 0.09, n = 35), shRandom + TOFA (0.44 ± 0.06, n = 32), shSAC1 + vehicle (0.90 ± 0.13, n = 34), and shSAC1 + TOFA (1.48 ± 0.09, n = 20). Scale bars = 5 µm. n.s, not significant. (D) SAC1 overexpression in cortical neurons. A lentiviral vector carrying SAC1 was used to infect the cells at 7 DIV. At 14–15 DIV, SAC1 overexpression was corroborated by Western blot. Casas et al. Journal of Cell Biology 13 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 (E) PI(4)P levels in the TGN compartment decreased by SAC1 overexpression. TGN46 was used as a marker of TGN. Values are the mean ± SD from two independent experiments performed by biological replicates (n = 16–22 cells per condition; two-sided Student’s t test; *, P = 0.0269; 1.00 ± 0.13 for EV and 0.89 ± 0.14 for SAC1). (F) SAC1 overexpression decreased surface levels of GluA1 and abolished the effect of TOFA treatment in cortical neurons. Neurons, transduced as explained above, were treated with TOFA at 14–15 DIV. Results are the mean ± SEM from two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001). EV + vehicle (1.00 ± 0.05, n = 36), shRandom + TOFA (0.27 ± 0.03, n = 36), SAC1 + vehicle (0.46 ± 0.04, n = 39), and shSAC1 + TOFA (0.36 ± 0.05, n = 36). Scale bars = 5 µm. Cell culture and treatments pmTurquoise2 (mTurquoise2-SAC1 and mTurquoise2-GluA1). Primary cortical mouse neurons were prepared from E16 WT or SAC1 was also cloned in the lentiviral pWPI-IRES-GFP plasmid CPT1C KO embryos, cultured, and maintained as described (pWPI-SAC1-IRES-GFP). For CPT1C silencing, the pVTHM previously (Fadó et al., 2015) in Neurobasal (21103049; Gibco) vector encoding a silencing sequence was used (Fadó et al., supplemented with B27 (17504044; Gibco), glutaMAX 2015). For SAC1 silencing, the sequence 59-gtagcaaatcacatggat (35050061; Gibco), and antibiotics. Cell lines were grown ac- g-39 (Yang et al., 2013) was cloned in the vector psi-LVRU6MP cording to American Type Culture Collection. HeLa and HEK (Tebu-Bio) to generate the plasmid short hairpin-loop RNA cells were maintained in DMEM (D5671; Sigma-Aldrich) sup- of SAC1 (shSAC1); and as a control (shRandom), the ran- plemented with 10% FBS (F7524; Sigma-Aldrich), 2 mM gluta- dom sequence 59-gcttcgcgccgtagtctta-39 was used (CSHCTR001- mine (X0551-100; Biowest), 100 U/ml penicillin, and 100 mg/ LVRU6MP; Tebu-Bio). All plasmids were verified by sequencing. ml streptomycin (P0781; Sigma-Aldrich) at 37°C with 5% CO2. FuGENE6 (E2692; Promega) was used for transient transfec- Cell lines were authenticated by genotyping and regularly tion in HeLa cells, the calcium phosphate method in HEK293T tested for mycoplasma contamination. For glucose starvation, cells, and Lipofectamine 2000 reagent (11668019; Invitrogen) cell medium was replaced by DMEM without glucose (A14430- in both HEK-SAC1GFP cells and cortical neurons at 13–14 DIV. 01; Gibco) supplemented with (control cells) or without All transfections with cell lines were conducted at 80% of (glucose-depleted cells) 25 mM glucose for 2 h. To inhibit confluence. malonyl-CoA synthesis, 20 µg/ml TOFA (ab141578; Abcam) was added for 2 h. 15 mM 2DG (D8375; Sigma-Aldrich) was Lentiviral infection added for 2 h to inhibit glycolysis. For the leptin treatment, For protein overexpression in cultures of primary cortical leptin (450–31; Bionova PrepoTech) was added to a final neurons, pWPI-SAC1-IRES-GFP, pWPI-CPT1C-IRES-GFP, pWPI- concentration of 50 nM for 30 min. For rapamycin-dependent CPT1CΔCterm-FLAG-IRES-GFP, pWPI-CPT1CM589S-IRES-GFP, recruitment, 5 µM rapamycin (R8781; Sigma-Aldrich) was and pWPI-IRES-GFP (empty vector; EV) were used to generate used at room temperature for 15–30 min. lentiviral particles, as previously described (Fadó et al., 2015). To silence CPT1C and SAC1, the pVTHM and LVRU6MP plasmids Plasmids and transfection described above were employed. Neurons were transduced with Human CPT1C-SYFP2 and mTurquoise-CPT1C were gifts from low virus titles (MOI = 2) at 8 DIV. Dr. Serra’s group (School of Pharmacy and Food Sciences, Universitat de Barcelon, Barcelona, Spain). The mTurquoise- Generation of stable cell lines TGN38-FRB plasmid, the PI(4)P probes mCherry-P4M and YFP- HeLa stably expressing mouse CPT1C or CPT1CM589S was P4M, and the PI(4,5)P2 probe YFP-PHPLCδ1 were gifts from E. generated by lentiviral transduction of pWPI-CPT1C-IRES-GFP Dickson’s group (School of Medicine, University of California, (CPT1C cells) and pWPI-CPT1CM589S-IRES-GFP (CPT1CM589S Davis, CA). mTurquoise-calnexin (#55539), mTurquoise-ER cells), respectively. Cells were transduced with low virus titles (#55550), mCherry-ER-FKBP (#72904), and iRFP-Sec61β (MOI = 5), and stable cells showing green fluorescence were (near-infrared fluorescent protein-Sec61β; #108125) were isolated by cell-sorter (FACS Aria Fusion; Becton Dickinson). purchased from Addgene. pcDNA3-Myc-CPT1C (Sierra et al., GFP was excited at 488 nm and green fluorescence collected at 2008) and pWPI-CPT1C-IRES-GFP (Fadó et al., 2015) plasmids 530/30 nm. pWPI-IRES-GFP EV was used to obtain control cells were engineered in our laboratory as published. CPT1CM589S- (EV cells). HEK-SAC1GFP cells were genetically modified using mTurquoise2 and pWPI-CPT1CM589S-IRES-GFP (Rodrı́guez- CRISPR-Cas (CRISPR-associated proteins) technology as previ- Rodrı́guez et al., 2019) encoded a malonyl-CoA–insensitive ously described (Zewe et al., 2018). CPT1C form (generated by site-directed mutagenesis at me- thionine 589). Plasmids containing the C-terminal truncated Viability assay form of CPT1C (mTurquoise2-CPT1CΔCterm-FLAG and pWPI- To test cellular viability, the 3-(4,5-dimethylthiazol-2-yl)-2,5- CPT1CΔCterm-FLAG-IRES-GFP) were constructed by replacing diphenyltetrazolium bromide (MTT) assay was performed in the last 39 aa (772–810) of the CPT1C sequence for the FLAG cortical neurons at 14–15 DIV. After treatments, MTT solution epitope (DYKDDDDK). SAC1 and GluA1 cDNAs were obtained by was added at a final concentration of 0.2 mg/ml, and cells were RT-PCR from mice brain tissue, using TRIzol (15596018; Life incubated for 1 h at 37°C. The mitochondrial activity of live cells Technologies) for RNA isolation, Mu-MLV (#3022-1; Lucigen) was measured by spectrophotometry at an absorbance wave- as a reverse transcription, and the Q5 DNA polymerase length of 570 nm. Cellular viability was given as a percentage (M0491L; New England Biolabs). SAC1 and GluA1 were cloned in versus control cells. Casas et al. Journal of Cell Biology 14 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 7. CPT1C regulates SAC1 recruitment to TGN-ER contacts. (A) CPT1C knockdown or TOFA treatment does not increase SAC1 in the TGN. Quantification of SAC1 levels at the TGN in shCPT1C-transduced cortical neurons by immunocytochemistry. A short hairpin-loop RNA with a random Casas et al. Journal of Cell Biology 15 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 sequence (shRandom) was used as a control. All data represent the mean ± SEM from two independent experiments performed by biological duplicates (two- way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05; **, P < 0.01). shRandom + vehicle (1.00 ± 0.02, n = 40), shRandom + TOFA (0.91 ± 0.03, n = 38), shCPT1C + vehicle (0.95 ± 0.02, n = 39), and shCPT1C + TOFA (0.88 ± 0.04, n = 29). Scale bars = 7 µm. (B–D) SAC1 localization at ER-TGN contact sites. HEK-SAC1-GFP cells were transfected with CPT1C or an EV together with expressing plasmids for ER (iRFP-Sec61β) and TGN (mTurquoise-TGN38-FRB) markers. 24 h later, cells were treated with TOFA (20 µg/ml, 2 h) or vehicle and live imaged. Representative images are shown in B. Black-and-white images show the ROI generated by TGN-PI(4)P colocalizing pixels. ER-TGN contacts were quantified as the total area of positive colocalization between TGN and ER signals (C). Endogenous SAC1-GFP mean intensity at these junctions was quantified (D). Data represent the mean ± SEM from two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and **, P < 0.01). Scale bars = 7 µm. (C) EV + vehicle (1.00 ± 0.12, n = 37), EV + TOFA (1.73 ± 0.24, n = 24), CPT1C + vehicle (1.44 ± 0.15, n = 37), and CPT1C + TOFA (1.62 ± 0.09, n = 33). (D) EV + vehicle (1.00 ± 0.06, n = 37), EV + TOFA (0.97 ± 0.05, n = 24), CPT1C + vehicle (0.94 ± 0.06, n = 37), and CPT1C + TOFA (1.33 ± 0.12, n = 33). (E) CPT1C mean intensity at ER-TGN contact sites in HEK293 cells overexpressing CPT1C-SYFP2. Contact sites were determined as in B. Representative images of each condition are shown. The graphs on the left represent the linear fluorescence profiles of the black line drawn on enlarged images: CPT1C in green, ER (Sec61β) in blue, and TGN (TGN38) in red. The right graph shows the mean intensity of CPT1C in the ER-TGN ROIs, and data represent the mean ± SD from two independent experiments performed by biological duplicates (n = 15–12 cells per condition; Mann-Whitney U test; **, P = 0.0037; 1.00 ± 0.09 for vehicle and 1.14 ± 0.11 for TOFA). Scale bars = 7 µm; scale bars of inset magnifications = 1.75 µm. Analysis of malonyl-CoA levels permeabilization (see Fig. S1 B). For surface GluA1 analysis, Intracellular malonyl-CoA levels were quantified in cortical isolated primary dendrites close to the soma of the neuron were neurons at 15 DIV. After treatment, neurons were washed with selected. Rabbit anti-GluA1 (1:30; PC246) was from Calbiochem. PBS, scraped, and collected. After centrifugation at 500 ×g for Mouse anti-PI(4)P (1:100; Z-P004) was from Echelon. Mouse 3 min, the cell pellets were frozen in liquid nitrogen. Malonyl- anti-TGN38 (1:100; NB300-575) was from Novus Biologicals. CoA levels in the collected cells were determined using the acyl- Rabbit anti-TGN46 (1:100; ab16059) and chicken anti-GFP CoA cycling method (Tokutake et al., 2010) and normalized by (1:1,000; ab13970) were from Abcam. Rabbit anti-SAC1 (1:100; cellular wet weight (milligram wet weight). Briefly, intracellular 13033–1-AP) was from Proteintech. Incubation with the ap- malonyl-CoA was extracted by adding 0.6 M sulfuric acid per propriate Alexa Fluor–conjugated secondary antibodies was each 100 mg wet cells. After being kept at 4°C overnight and also performed. Anti-mouse Alexa633 (1:100; A21052), anti- centrifuged, 1 M Tris (0.05 volume of supernatant) was added to rabbit Alexa488 (1:300; A11008), anti-chicken Alexa488 (1: the supernatant, and the pH of the solution was adjusted to 1,000; A11039), and anti-mouse Alexa568 (1:150; A11011) were around 6.5 with NaOH on ice. Malonyl-CoA levels in the extracts from Invitrogen. Coverslips were mounted with Fluoromount were measured by the enzymatic cycling assay using malonate Aqueous Mounting Medium (F4680; Sigma-Aldrich). decarboxylase after elimination of acetyl-CoA by citrate syn- thase. The reaction mixture for the citrate synthase treatment Immunoblotting was composed of 50 mM Tris-HCl (pH 7.2), 2 mM oxaloacetate, Western blot was performed as previously described (Fadó 5 mg recombinant citrate synthase from Bacillus subtilis, and the et al., 2015). Rabbit anti-CPT1C was developed in our labora- tissue extracts in 1 ml. The reaction was performed at 25°C for tory (1:1,000; Sierra et al., 2008). Rabbit anti-ACC (1:1,000; 3676) 20 min, and an aliquot of the mixture was transferred to the and rabbit anti-phosphorylated ACC (Ser79; 1:2,000; 3661) reaction mixture for the acyl-CoA cycling method. The mixture were from Cell Signaling. Rabbit anti-SAC1 (1:500; 13033–1-AP) contained 50 mM Tris-HCl (pH 7.2), 1 mM 2-mercaptoethanol, was from Proteintech. Mouse anti–β-actin (1:1,000; ab6276) 10 mM MgSO4, 50 mM malonate, 10 mM ATP, 1 U of malonate and chicken anti-GFP (1:5,000; ab13970) were from Abcam. decarboxylase, and the solution from citrate synthase treatment Rabbit anti-GluA1 (1:1,000; ab1504) and rabbit anti-GluA2 (1: in 400 µl. The mixture was incubated at 30°C for 20 min, and 1 U 1,000; ab397) were from Millipore. Rabbit anti-GluN2A (1: of recombinant acetate kinase from Escherichia coli was added. 1,000; G9038), mouse anti–β-tubulin (1:2,000; T5201), and After 20 min of incubation, 0.2 ml of 2.5 M neutralized hy- goat anti-FLAG (1:500; F3165) were from Sigma-Aldrich. HRP- droxylamine was added, and the incubation was continued for conjugated secondary antibodies anti-mouse or anti-rabbit an additional 20 min at 30°C. The reaction was terminated by (1:10,000) were from DAKO. Blots were developed using Lu- adding 0.6 ml of 10 mM ferric chloride dissolved in 25 mM minata Forte Western HRP substrate (Millipore). Semiquan- trichloroacetic acid–1 M HCl. The A540 of the acetohydroxamate titative analysis was performed using densitometry with Fiji formed was measured. software. Immunostaining Cell-surface biotinylation assay Immunostaining of proteins and lipids was used in fixed cells Surface protein levels were measured in cortinal neurons at 14- with 4% (wt/vol) paraformaldehyde. For surface proteins, non- 15 DIV by the biotin covalent binding method and processed as permeabilized cells were directly blocked with goat serum and described previously (Miñano-Molina et al., 2011). Briefly, live then incubated with the appropriate primary antibody. For in- cells were transferred to ice-cold PBS-Ca2+-Mg2+ buffer (pH 7.4, ternal proteins and lipids, cells were permeabilized with 0.1% 1 mM CaCl2, and 0.1 mM MgCl2), followed by incubation with (vol/vol) Triton X-100 before blocking and incubation with 0.5 mg/ml biotin (b10185; Life Technologies). Free biotin the primary antibody. Control experiments were perfor- was quenched by three washes in cold PBS-Ca2+-Mg2+ + 0.1 M med to make sure that chemical fixation did not induce glycine. Cells were next scraped in cold 1% Triton X-100 Casas et al. Journal of Cell Biology 16 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 8. Artificially formed ER-TGN contacts decrease surface GluA1 levels. (A) PI(4)P at the TGN after artificial ER-TGN contact creation. HEK cells overexpressing or not Myc-CPT1C were transfected with the mCherry-pHR-TcRb-FKBP and CFP-TGN38-FRB plasmids and treated with rapamycin (5 µM, 30 min) in order to induce ER-TGN contacts. PI(4)P was labeled with the lipid probe YFP-P4M. A Myc EV was used as a control. Cells were treated with TOFA 24 h after transfection. Cells were live imaged with confocal microscopy. Values are the mean ± SD from four independent experiments (n = 6 cells per condition; two-way ANOVA followed by Bonferroni’s comparison test; **, P < 0.01 and ***, P < 0.001). (B) Surface GluA1 levels after artificial ER-TGN contact creation. Live HEK cells were transfected and treated as in A. Surface GluA1 was imaged by TIRF microscopy. Data represent the mean ± SD from two independent experiments (n = 4–8 cells per condition; two-sided Student’s t test; *, P < 0.05 in YFP control cells and P = 0.0503 in YFP-CPT1C overexpressing cells). EV + vehicle (1.00 ± 0.76, n = 8), EV + rapamycin (0.29 ± 0.10, n = 6), CPT1C + vehicle (1.00 ± 0.85, n = 8), and CPT1C + rapamycin (1.85 ± 0.87, n = 4). Scale bars = 7 µm; scale bars of inset magnifications = 1.75 µm. n.s, not significant. homogenization buffer (NaCl, 50 mM; EDTA, 10 mM; EGTA, 10 supernatants were pulled down with 40 µl of 50% avidin- mM; Na3VO4, 1 mM; NaF, 50 mM; NaPPi, 25 mM; glycero- agarose beads (53150; Pierce) overnight at 4°C. Beads were phosphate, 1 mM; PMSF, 1 mM; and Hepes, 50 mM, pH 7.5) pelleted, washed three times with homogenization buffer, and supplemented with phosphatase and protease inhibitor cock- heated in 100 µl of 2 x SDS buffer (surface fraction). 75 µl of the tails. Homogenates were centrifuged at 10,000 ×g for 20 min to supernatants was mixed and heated with 25 µl of 2× SDS buffer pellet insoluble fraction. 75 µl of the supernatants was mixed (internal fraction). Equal volumes of the input, internal, and and heated with 25 µl of 2× SDS buffer (input). The remaining biotinylated fractions were processed by immunoblotting. Casas et al. Journal of Cell Biology 17 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure 9. CPT1C negatively regulates SAC1 PI(4)P phosphatase activity. (A) SAC1 phosphatase activity regulation by CPT1C and TOFA. Hela cells per- manently expressing CPT1C or the EV were treated by TOFA (20 µg/ml, 2 h) or vehicle. After TOFA treatment, SAC1 was immunoprecipitated from cellular lysates, and PI(4)P phosphatase activity was measured as the Pi released from PI(4)P substrate using a green malachite assay. All data were normalized by the amount of immunoprecipitated SAC1 in each sample detected by Western blot. Values are the mean ± SD of three independent experiments performed by technical replicates (one-sample t test; *, P < 0.05). EV + vehicle (1.00 ± 0, n = 6), EV + TOFA (0.56 ± 0.37, n = 6), CPT1C + vehicle (0.92 ± 0.42, n = 4), and CPT1C + TOFA (1.19 ± 0.33, n = 4). (B) SAC1 phosphatase activity regulation by CPT1C-sensing of malonyl-CoA. HeLa cells permanently expressing CPT1C, malonyl- CoA–insensitive CPT1C form (M589S), or EV were analyzed for PI(4)P phosphatase activity as in A. Results are the mean ± SD of three independent experiments performed by technical replicates (n = 4–8 samples per condition; one sample t test; **, P < 0.01). EV (1.00 ± 0.00, n = 9), CPT1C (0.50 ± 0.31, n = 9), and CPT1CM589S (3.31 ± 0.1.96, n = 5). n.s, not significant. (C) SAC1 phosphatase activity in brains from WT and CPT1C KO adult mice. Total brain lysates were processed for SAC1 immunoprecipitation. PI(4)P phosphatase activity was measured in immunoprecipitated pellets as in A. Values are the mean ± SD (n = 5 Casas et al. Journal of Cell Biology 18 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 animals per genotype; two-sided Student’s t test; *, P = 0.0353; 1.00 ± 0.08 for WT and 1.50 ± 0.27 for KO). (D) SAC1 phosphatase activity in brains from fed and fasted mice. WT mice were subjected to 24 h of fasting. PI(4)P phosphatase activity was measured as in C. Values are the mean ± SD (n = 9 or 10 animals per condition; one-sided Student’s t test; *, P = 0.0256; 1.00 ± 0.66 for fed and 1.77 ± 0.91 for fasted). (E) Total PI(4)P levels were quantified with a lipid probe (mCherry-P4M) in live HEK cells overexpressing or not CPT1C after TOFA or vehicle addition. Graph shows the mean ± SEM from three independent experiments performed by biological triplicates (two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and ***, P < 0.001). EV + vehicle (1.00 ± 0.08, n = 16), EV + TOFA (1.17 ± 0.08, n = 18), CPT1C + vehicle (1.27 ± 0.08, n = 22), and CPT1C + TOFA (0.79 ± 0.04, n = 20). Scale bars = 10 µm. (F) Total PI(4,5)P2 levels were quantified with a lipid probe (YFP-PHPLCδ1) in live HEK cells overexpressing or not CPT1C after TOFA or vehicle addition. Graph shows the mean ± SEM from three independent experiments performed by biological triplicates two-way ANOVA followed by Bonferroni’s comparison test; **, P < 0.01 and ***, P < 0.001). EV + vehicle (1.00 ± 0.08, n = 18), EV + TOFA (0.78 ± 0.04, n = 18), CPT1C + vehicle (1.62 ± 0.23, n = 18), and CPT1C + TOFA (1.46 ± 0.08, n = 18). Scale bars = 10 µm. (G) Summary histogram of UPLC-MS/MS PI measurements from WT and CPT1C−/− brains (n = 3 brains per condition; 894.51 ± 208.68 for WT and 999.26 ± 273.00 for KO). (H) Same experiments as in G only measuring PIP. n = 3 brains per condition; two-sided Student’s t test; **, P < 0.01; 393.40 ± 52.77 for WT and 202.40 ± 24.76 for KO. (I) Same experiment as in G only measuring PIP2. n = 3 brains per condition; two-sided Student’s t test; P = 0.07; 428.60 ± 32.77 for WT and 300.80 ± 63.45 for KO. Graphs in G, H, and I show the mean ± SD from one experiment performed by biological replicates. (J) Heat map detailing the relative changes in each PI isoform from CPT1C−/− brains relative to WT brains. Immunoprecipitation assay unspecifically. Meanwhile, 2 µg of anti-CPT1C, 1 µg of anti- Cells were washed in ice-cold PBS and immediately scraped in GluA1, or 2.5 µg of anti-SAC1 antibodies was incubated with cold lysis buffer (500 µl/2 × 100-mm plate; 1% Triton buffer for 10 µl of protein G–Sepharose beads (Ab-beads). Both samples CPT1C immunoprecipitations, Tris 50 mM, NaCl 150 mM, and were centrifuged at 4,000 g for 2 min. Clean supernatants were Triton X-100 1% at pH 8; or CL-47 buffer for Sac1 im- immunoprecipitated with the beads attached to the antibody munoprecipitations). Homogenates from cultures were solubi- (Ab-beads) overnight at 4°C, since protein G–Sepharose beads lized 30 min at 4°C in the orbital and centrifuged at 10,000 ×g (preclear fraction) were mixed with 100 µl of 2× SDS buffer and for 10 min to pellet insoluble fraction. 75 µl of the supernatant 1 µl of DTT at 1 M and heated. The beads were pelleted, and 75 µl was mixed and heated with 25 µl of 4× SDS sample buffer (in- of the supernatant (unbound fraction) was mixed and heated put). 1 mg of total protein from the remaining supernatants was with 25 µl of 4× SDS sample buffer. The beads were then rinsed incubated 1 h in the orbital with 10 µl of slurry protein- three times with lysis buffer and heated in 100 µl of 2× SDS G–Sepharose beads in order to retain protein that bind beads sample buffer and 1 µl of DTT at 1 M (immunoprecipitation Figure 10. A working model for CPT1C regulation of GluA1 trafficking. CPT1C and SAC1 form part of the AMPAR complex at the ER of neurons. We do not discard the existence of CPT1C/SAC1 pairs in the ER that do not form part of AMPAR complexes. In normal nutrient conditions, malonyl-CoA–bound CPT1C down-regulates SAC1 activity, allowing the accumulation of PI(4)P at the TGN. This favors the trafficking of GluA1-containing AMPARs from the TGN to the PM through the secretory pathway. In energy stress conditions that decrease malonyl-CoA levels, such as glucose deficiency, ER-TGN contacts are increased and CPT1C favors the transport of SAC1 to these junctions. At the same time, malonyl-CoA–free CPT1C releases its inhibition on SAC1 activity. This results in dephosphorylation of PI(4)P at the TGN by SAC1 in trans, which halts GluA1-containing AMPAR trafficking. The figure was drawn using BioRender web-based software. Casas et al. Journal of Cell Biology 19 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 fraction; IP). 30 µl of the preclear fraction and IP and 10–20 µl DMi8 using a ×63 1.4 NA oil objective. Approximately 18 stacks of the input and the unbound fractions were detected by with 0.5-µm distance were taken per each imaged cell using immunoblotting. identical settings, and all quantifications were done on a re- constructed 3D image with IMARIS 9.2 software (Bitplane). For Protein–protein interaction simulations quantification, sets of cells were cultured and stained simulta- For the CPT1C-GluA1 interactions, the GluA1 cytosolic region neously and imaged using identical settings. For surface GluA1 (residues from 827 to 903) was modeled with the available quantification, primary dendrites were taken, being considered GluA2 crystal structure (Protein Data Bank accession no. 3KG2), as the prolongations first branching from soma. The axon was and the CPT1C C-terminal plus its globular regions (residues identified and discarded according to morphological criteria from 157 to 803) was modeled with the COT crystal structure (thickness and position relative to the nucleus). A 15 µm × 5 µm (Protein Data Bank accession no. 1XL7). The whole process was ROI was randomly selected, and integrated intensity was mea- performed on a simulated aqueous solution (tip4p-D cubic wa- sured inside it using the Spot tool. For the measurements of terbox; optimization by Best et al., 2012) with physiological SAC1, GluA1, and PI(4)P intensities at the TGN compartment, the concentration of NaCl and enough ions to neutralize the charge mean intensity of each staining in the 3D TGN-englobing ROI of the system. The mathematical function Charmm36 was em- close to the soma was quantified using the Surface tool of Imaris ployed. It was calculated as interaction time percentage (when software. The same threshold values were used in all the low energies are required for protein approach). Next, a mini- samples. mization function with 10,000 steps was used in order to get rid of bad contacts or unsuitable torsion angles from both inter- Live-cell imaging acting molecules that may have been inevitably introduced when Live-cell imaging was used to detect lipids or proteins in non- setting up the whole system. A final two-step equilibration process fixed transfected HEK-SAC1GFP cells (kind gift of Dr. Gerry was also employed. We submitted the CPT1C-GluA1 system to an Hammond, University of Pittsburgh, Pittsburgh, PA). 1 d after aggregated 300 ns of production runs in which 12 different transfection, cells were transferred from culture medium to a starting points were employed, each of them differing in the initial recording chamber containing modified Krebs–Ringer solution GluA1 fragment positioning relative to CPT1C fragment. The 12 (160 mM NaCl, 2.5 mM KCl, 2 mM CaCl2, 1 mM MgCl2, and 8 mM approaching vectors were split into two groups of six each, where glucose, pH 7.4). Cells were kept in this solution up to ∼1 h and each of the group presented one of two “faces” of the GluA1 cy- 30 min while imaging. tosolic fragment (as defined by an arbitrary plane intersecting lengthwise with the protein fragment) to the CPT1C fragment, and Total internal reflection fluorescence (TIRF) microscopy each approaching vector on the group was displaced 60 degrees in TIRF imaging was performed in GluA1-transfected live HEK- a clockwise manner from the previous. The utility of multiple SAC1GFP cells. Footprints were acquired in a Nikon TiE micro- replicas is discussed in Knapp et al. (2018). The molecular dy- scope equipped with a TIRF 60×/1.25 oil differential interference namics package used for the minimization, equilibration, and contrast H objective and Photometrics QuantEM electron- production processes was GROMACS (Van Der Spoel et al., 2005). multiplying charge-coupled device camera (Nikon). The inci- dent angle settled in all experiments was 110 degrees. All images FRET analysis were quantified with the ImageJ/Fiji (National Institutes of All FRET imaging experiments were performed on a Leica TCS Health) software and surface GluA1 mean intensity was mea- SP2 microscope. A 63×/1.40 NA objective with immersion oil and sured per cell. argon lasers with 458 and 514 lines (for donor, mTurquoise2, and acceptor, YFP, respectively) was used for all image acquisitions. Super-resolution imaging The pinhole size was set to 2 (airy units), and the region of interest Super-resolution studies were conducted using HEK-SAC1GFP (ROI) to 5 × 5 (mm). 458-nm laser power was set up to 23%–60% cells transfected with TGN and ER organelle markers (mTur- and 514-nm laser up to 5%–10% so that no pixel saturation oc- quoise-TGN38-FRB and iRFP-Sec61β, respectively) and a CPT1C curred. The number of bleaches was between three and four, all at overexpressing plasmid (CPT1C-SYFP2). Images were acquired 40%. For the calculation of FRET efficiency from acceptor to do- in a ZEISS LSM 880 with Airyscan using 63× oil-immersion nor, we used the following formula operated at pixel basis: objective. Quantification of contact sites, endogenous SAC1,   and overexpressed CPT1C at these junctions was performed cell E  Ifret Id  (Ia − Idcrossem − Icrossex ) Id , per cell in one-stack images by using ImageJ/Fiji software. The where Ifret refers to FRET intensity, Id refers to donor intensity in amount of ER-TGN contacts was measured as a positive coloc- the absence of acceptor, Ia is the acceptor intensity in the alization between the ER and the TGN signals and was always presence of donor, Idcrossem is the fraction of donor emission normalized to ER area. To measure SAC1 or CPT1C in the ER- cross-emitted into the acceptor channel, and Icrossex is the frac- TGN contacts, we drew two 2D ROIs corresponding to the ER and tion of acceptor intensity cross-excited by donor excitation. the TGN signals and then applied the colocalization plugin of ImageJ/Fiji software in order to determine the ER-TGN contact Confocal microscopy area per cell and calculate the mean intensity of SAC1 or CPT1C Confocal imaging was performed in fixed cortical neurons with in that colocalized ROI. For the analysis of linear fluorescence the confocal laser-scanning microscope ZEISS LSM 700 or Leica profiles, a line 1 µm wide and 5 µm long was drawn inside the Casas et al. Journal of Cell Biology 20 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 cell, and the sum intensity histograms of CPT1C, TGN marker, by biological duplicates and are given as the mean ± SEM or SD. and ER marker from 2D images was obtained using ImageJ/Fiji When individual cells were analyzed, we decided to evaluate software. around 30 per condition for FRET assays and between 20 and 40 cells for immunostaining studies (at least) seeded in two inde- Rapamycin-dependent ER-TGN contact site creation pendent coverslips per experiments. Exceptionally, the outliers To explore ER-TGN contact sites, the rapamycin system was were excluded if the values were above or below the mean ± two used to produce a tight association between the FKBP and FRB SDs. Statistical analysis was performed using PRISM (GraphPad domains (Bohdanowicz and Fairn, 2011). Based on this, trans- Software). Data normality was determined in all the experi- fection of HEK-SAC1GFP cells with mCherry-pHR-TcRb-FKBP, ments according to the Shapiro-Wilk test and D’Agostino and CFP-TGN38-FRB, and YFP-P4M was performed followed by ra- Pearson test. Significance between groups was detected using pamycin addition (5 µM at room temperature; R8781; Sigma- Student’s t test (one or two sided, as indicated in each figure Aldrich). This synthetic system allows ER recruitment to the legend) or a Mann-Whitney U test (indicated in each figure TGN and a consequent decrease in Golgi PI(4)P (Dickson et al., legend) according to normality and a 95% confidence interval. 2014). Under constant perfusion, cells were tracked for ∼15 min For comparisons among three or four groups, ANOVA was with images taken every ∼5 s. Positive ER-TGN junction creation performed (one-way or two-way; indicated in each figure leg- was only considered when the ER signal localized at the TGN in- end), followed by the Bonferroni’s multiple comparison post- creased with respect to its signal outside the TGN. This increase test. was measured with ImageJ/Fiji software as described above in one-stack images. Mean PI(4)P intensity was also quantified in the Online supplemental material TGN compartment in each cell by creating a TGN-englobing mask Fig. S1 presents whole cell images of cortical neurons showing and then applying it on the PI(4)P image. PI(4)P inside the TGN GluA1 surface levels under different treatments. Fig. S2 shows was always normalized to the cytoplasm signal in order to discard that glucose deprivation, 2DG, and TOFA treatments have no any difference due to the transfection efficiency. effect on cortical neuron survival. Fig. S3 shows TOFA treatment decreases GluA1 but not GluA2 and GluN2A at the surface of PI(4)P-dependent phosphatase activity assay of SAC1 cortical neurons in a CPT1C-dependent manner. Fig. S4 shows The activity assay of SAC1 was performed in stable transfected the effects of TOFA treatment or SAC1 modulation on PI(4)P HeLa cells and in mice brain tissues. In cell lines, SAC1 was levels, TGN volume, and TGN-ER occurrence. Fig. S5 shows overexpressed by FuGENE6 transfection 48 h before the assay. rapamycin-induced ER-TGN contacts. Fig. S6 shows that CPT1C Mice brain tissues were obtained from 4-mo-old male mice. deficiency decreases the brain levels of different PIP and PIP2 HeLa cells or brain tissue was homogenized with chilled CL-47 species. lysis buffer supplemented with a protein inhibitor cocktail. 2 mg or 10 mg of protein (for HeLa cells and brain tissue, respectively) was immunoprecipitated with anti–SAC1 conjugated beads as Acknowledgments described above. The immunoprecipitated protein-containing This work was partially supported by Ministerio de Economia y pellet was processed as described previously (Mani et al., Competitividad from Spain (grant SAF2017-82813-C3-3R to N. 2007). Released free phosphate, an indicator of SAC1 activity Casals, grant SAF2017-89271-R to J. Rodrı́guez-Álvarez, cofunded on PI(4)P, was measured at 650 nm and was always normalized by the European Regional Development Fund), Centro de In- to immunoprecipitated SAC1 detected by Western blot. vestigación Biomédica en Red Fisiopatologı́a de la Obesidad y la Nutrición (grant CB06/03/0001 to N. Casals), Centro de In- UPLC MS/MS vestigación Biomédica en Red sobre Enfermedades Neuro- Lipids were extracted with butanol and chloroform from age- degenerativas (grant CB06/05/0042 to J. Rodrı́guez-Álvarez), and sex-matched WT and CPT1C −/− brains. Samples were Generalitat de Catalunya (2014SGR465 to N. Casals, SGR2014- neutralized by methylation with trimethylsilyldiazomethane, 0984 to J. Rodrı́guez-Álvarez), Fundació La Marató de TV3 infused with sodium formate, and analyzed using a Waters XEVO (grant 87/C/2016 to N. Casals, grant 2014-3610 to J. Rodrı́guez- TQ-S MS/MS in multiple reaction monitoring mode using elec- Álvarez), University of California funds (to E.J. Dickson), and trospray and positive ion mode (Vivas et al., 2019). Elution pro- National Institutes of Health grant R01 GM127513 (to files plotting the intensity (arbitrary units) as a function of elution E.J. Dickson). time were analyzed to determine relative changes between sam- The authors declare no competing financial interests. ples. Total integrated areas under peaks from samples and Author contributions: M. Casas conducted conceptualization, standards were quantified using MassLynx software (Research formal analysis, investigation, methodology, validation, visual- Resource Identifier: SCR_014271; Waters). Samples were normal- ization, and review and editing. R. Fadó conducted conceptual- ized to total brain protein levels and nonphysiological lipid ization, formal analysis, investigation, validation, visualization, standards: PI(4)P, 37:4, and PI(4,5)P2, 37:4 (Avanti Polar Lipids). supervision, original draft, and review and editing. J.L. Domı́n- guez and A. Roig conducted investigation and methodology. M. Statistical analysis Kaku and S. Chohnan conducted methodology, investigation, re- As indicated in each figure legend, the results were obtained sources and review, and editing. M. Solé and M. Unzeta conduc- from two to four independent experiments performed (at least) ted review and editing. A.J. Miñano-Molina, J. Rodrı́guez-Álvarez, Casas et al. Journal of Cell Biology 21 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 and E.J. Dickson conducted conceptualization, methodology, re- AMPA receptors on hippocampal CA3 neurons mediated by mGluR1 and adenosine A3 receptors. J. Neurosci. 31:11941–11952. https://doi.org/10.1523/ sources, supervision, and review and editing. N. Casals conceived JNEUROSCI.1183-11.2011 the study and conducted conceptualization, formal analysis, Dickson, E.J., J.B. Jensen, and B. Hille. 2014. Golgi and plasma membrane pools funding acquisition, project administration, resources, supervi- of PI(4)P contribute to plasma membrane PI(4,5)P2 and maintenance of KCNQ2/3 ion channel current. Proc. Natl. Acad. Sci. USA. 111:E2281–E2290. sion, and review and editing. https://doi.org/10.1073/pnas.1407133111 Dickson, E.J., J.B. Jensen, O. Vivas, M. Kruse, A.E. Traynor-Kaplan, and B. Submitted: 9 December 2019 Hille. 2016. Dynamic formation of ER-PM junctions presents a lipid phosphatase to regulate phosphoinositides. J. Cell Biol. 213:33–48. Revised: 21 April 2020 https://doi.org/10.1083/jcb.201508106 Accepted: 18 June 2020 Diessler, S., M. Jan, Y. Emmenegger, N. Guex, B. Middleton, D.J. Skene, M. Ibberson, F. Burdet, L. Götz, M. Pagni, et al. 2018. A systems genetics resource and analysis of sleep regulation in the mouse. PLoS Biol. 16. e2005750. https://doi.org/10.1371/journal.pbio.2005750 References Domı́nguez, J.L., F. Fernández-Nieto, M. Castro, M. Catto, M.R. Paleo, S. Porto, Beilharz, J.E., J. Maniam, and M.J. Morris. 2015. Diet-Induced Cognitive F.J. Sardina, J.M. Brea, A. Carotti, M.C. Villaverde, et al. 2015. Computer- Deficits: The Role of Fat and Sugar, Potential Mechanisms and Nutri- aided structure-based design of multitarget leads for Alzheimer’s dis- tional Interventions. Nutrients. 7:6719–6738. https://doi.org/10.3390/ ease. J. Chem. Inf. Model. 55:135–148. https://doi.org/10.1021/ci500555g nu7085307 Dong, R., Y. Saheki, S. Swarup, L. Lucast, J.W. Harper, and P. De Camilli. 2016. Best, R.B., X. Zhu, J. Shim, P.E.M. Lopes, J. Mittal, M. Feig, and A.D. Mackerell, Endosome-ER Contacts Control Actin Nucleation and Retromer Func- Jr.. 2012. Optimization of the additive CHARMM all-atom protein force tion through VAP-Dependent Regulation of PI4P. Cell. 166:408–423. field targeting improved sampling of the backbone φ, ψ and side-chain https://doi.org/10.1016/j.cell.2016.06.037 χ(1) and χ(2) dihedral angles. J. Chem. Theory Comput. 8:3257–3273. Dworak, M., R.W. McCarley, T. Kim, A.V. Kalinchuk, and R. Basheer. 2010. https://doi.org/10.1021/ct300400x Sleep and brain energy levels: ATP changes during sleep. J. Neurosci. 30: Blagoveshchenskaya, A., F.Y. Cheong, H.M. Rohde, G. Glover, A. Knödler, T. 9007–9016. https://doi.org/10.1523/JNEUROSCI.1423-10.2010 Nicolson, G. Boehmelt, and P. Mayinger. 2008. Integration of Golgi Fadó, R., D. Soto, A.J. Miñano-Molina, M. Pozo, P. Carrasco, N. Yefimenko, J. trafficking and growth factor signaling by the lipid phosphatase SAC1. Rodrı́guez-Álvarez, and N. Casals. 2015. Novel Regulation of the Syn- J. Cell Biol. 180:803–812. https://doi.org/10.1083/jcb.200708109 thesis of α-Amino-3-hydroxy-5-methyl-4-isoxazolepropionic Acid Bohdanowicz, M., and G.D. Fairn. 2011. Rapamycin-based inducible translo- (AMPA) Receptor Subunit GluA1 by Carnitine Palmitoyltransferase 1C cation systems for studying phagocytosis. Methods Mol. Biol. 748: (CPT1C) in the Hippocampus. J. Biol. Chem. 290:25548–25560. https:// 183–193. https://doi.org/10.1007/978-1-61779-139-0_13 doi.org/10.1074/jbc.M115.681064 Bowen, A.B., A.M. Bourke, B.G. Hiester, C. Hanus, and M.J. Kennedy. 2017. Faulhammer, F., S. Kanjilal-Kolar, A. Knödler, J. Lo, Y. Lee, G. Konrad, and P. Golgi-independent secretory trafficking through recycling endosomes Mayinger. 2007. Growth control of Golgi phosphoinositides by recip- in neuronal dendrites and spines. eLife. 6. e27362. https://doi.org/10 rocal localization of sac1 lipid phosphatase and pik1 4-kinase. Traffic. 8: .7554/eLife.27362 1554–1567. https://doi.org/10.1111/j.1600-0854.2007.00632.x Brechet, A., R. Buchert, J. Schwenk, S. Boudkkazi, G. Zolles, K. Siquier-Pernet, I. Fernandes, J., M. Vieira, L. Carreto, M.A.S. Santos, C.B. Duarte, A.L. Carvalho, Schaber, W. Bildl, A. Saadi, C. Bole-Feysot, et al. 2017. AMPA-receptor and A.E. Santos. 2014. In vitro ischemia triggers a transcriptional re- specific biogenesis complexes control synaptic transmission and intellec- sponse to down-regulate synaptic proteins in hippocampal neurons. tual ability. Nat. Commun. 8:15910. https://doi.org/10.1038/ncomms15910 PLoS One. 9. e99958. https://doi.org/10.1371/journal.pone.0099958 Breslow, D.K., S.R. Collins, B. Bodenmiller, R. Aebersold, K. Simons, A. Gao, L., W. Chiou, H. Tang, X. Cheng, H.S. Camp, and D.J. Burns. 2007. Si- Shevchenko, C.S. Ejsing, and J.S. Weissman. 2010. Orm family proteins multaneous quantification of malonyl-CoA and several other short- mediate sphingolipid homeostasis. Nature. 463:1048–1053. https://doi chain acyl-CoAs in animal tissues by ion-pairing reversed-phase .org/10.1038/nature08787 HPLC/MS. J. Chromatogr. B Analyt. Technol. Biomed. Life Sci. 853: Carrasco, P., I. Sahún, J. McDonald, S. Ramı́rez, J. Jacas, E. Gratacós, A.Y. 303–313. https://doi.org/10.1016/j.jchromb.2007.03.029 Sierra, D. Serra, L. Herrero, A. Acker-Palmer, et al. 2012. Ceramide Gao, X.F., W. Chen, X.P. Kong, A.M. Xu, Z.G. Wang, G. Sweeney, and D. Wu. levels regulated by carnitine palmitoyltransferase 1C control dendritic 2009. Enhanced susceptibility of Cpt1c knockout mice to glucose in- spine maturation and cognition. J. Biol. Chem. 287:21224–21232. https:// tolerance induced by a high-fat diet involves elevated hepatic gluco- doi.org/10.1074/jbc.M111.337493 neogenesis and decreased skeletal muscle glucose uptake. Diabetologia. Carrasco, P., J. Jacas, I. Sahún, H. Muley, S. Ramı́rez, B. Puisac, P. Mezquita, J. 52:912–920. https://doi.org/10.1007/s00125-009-1284-0 Pié, M. Dierssen, and N. Casals. 2013. Carnitine palmitoyltransferase 1C Gao, S., G. Zhu, X. Gao, D. Wu, P. Carrasco, N. Casals, F.G. Hegardt, T.H. deficiency causes motor impairment and hypoactivity. Behav. Brain Res. Moran, and G.D. Lopaschuk. 2011. Important roles of brain-specific 256:291–297. https://doi.org/10.1016/j.bbr.2013.08.004 carnitine palmitoyltransferase and ceramide metabolism in leptin hy- Casals, N., V. Zammit, L. Herrero, R. Fadó, R. Rodrı́guez-Rodrı́guez, and D. pothalamic control of feeding. Proc. Natl. Acad. Sci. USA. 108:9691–9696. Serra. 2016. Carnitine palmitoyltransferase 1C: From cognition to can- https://doi.org/10.1073/pnas.1103267108 cer. Prog. Lipid Res. 61:134–148. https://doi.org/10.1016/j.plipres.2015.11 Gratacòs-Batlle, E., N. Yefimenko, H. Cascos-Garcı́a, and D. Soto. 2015. AM- .004 PAR interacting protein CPT1C enhances surface expression of GluA1- Chen, N., N.J. Pandya, F. Koopmans, V. Castelo-Székelv, R.C. van der Schors, containing receptors. Front. Cell. Neurosci. 8:469. https://doi.org/https:// A.B. Smit, and K.W. Li. 2014. Interaction proteomics reveals brain doi.org/10.3389/fncel.2014.00469 region-specific AMPA receptor complexes. J. Proteome Res. 13:5695–5706. Gratacòs-Batlle, E., M. Olivella, N. Sánchez-Fernández, N. Yefimenko, F. Mi- https://doi.org/10.1021/pr500697b guez-Cabello, R. Fadó, N. Casals, X. Gasull, S. Ambrosio, and D. Soto. 2018. Chung, J., F. Torta, K. Masai, L. Lucast, H. Czapla, L.B. Tanner, P. Nar- Mechanisms of CPT1C-Dependent AMPAR Trafficking Enhancement. ayanaswamy, M.R. Wenk, F. Nakatsu, and P. De Camilli. 2015. PI4P/ Front. Mol. Neurosci. 11:275. https://doi.org/10.3389/fnmol.2018.00275 phosphatidylserine countertransport at ORP5- and ORP8-mediated Hangen, E., F.P. Cordelières, J.D. Petersen, D. Choquet, and F. Coussen. 2018. ER–plasma membrane contacts. Science. 349:428–432. https://doi.org/ Neuronal Activity and Intracellular Calcium Levels Regulate Intracel- 10.1126/science.aab1370 lular Transport of Newly Synthesized AMPAR. Cell Rep. 24:1001–1012.e3. Convertino, M., R. Pellarin, M. Catto, A. Carotti, and A. Caflisch. 2009. 9,10- https://doi.org/10.1016/j.celrep.2018.06.095 Anthraquinone hinders beta-aggregation: how does a small molecule Harris, J.J., R. Jolivet, and D. Attwell. 2012. Synaptic energy use and supply. interfere with Abeta-peptide amyloid fibrillation? Protein Sci. 18: Neuron. 75:762–777. https://doi.org/10.1016/j.neuron.2012.08.019 792–800. https://doi.org/10.1002/pro.87 Henley, J.M., and K.A. Wilkinson. 2013. AMPA receptor trafficking and the Del Bel, L.M., and J.A. Brill. 2018. Sac1, a lipid phosphatase at the interface of mechanisms underlying synaptic plasticity and cognitive aging. Dia- vesicular and nonvesicular transport. Traffic. 19:301–318. https://doi logues Clin. Neurosci. 15:11–27. .org/10.1111/tra.12554 Henley, J.M., and K.A. Wilkinson. 2016. Synaptic AMPA receptor composition Dennis, S.H., N. Jaafari, H. Cimarosti, J.G. Hanley, J.M. Henley, and J.R. in development, plasticity and disease. Nat. Rev. Neurosci. 17:337–350. Mellor. 2011. Oxygen/glucose deprivation induces a reduction in synaptic https://doi.org/10.1038/nrn.2016.37 Casas et al. Journal of Cell Biology 22 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Hong, D., L. Cong, S. Zhong, L. Liu, Y. Xu, and J. Zhang. 2019. A novel CPT1C of AMPA receptors in the hippocampus. Proc. Natl. Acad. Sci. USA. 111: variant causes pure hereditary spastic paraplegia with benign clinical E149–E158. https://doi.org/10.1073/pnas.1313798111 course. Ann. Clin. Transl. Neurol. 6:610–614. https://doi.org/10.1002/ Rinaldi, C., T. Schmidt, A.J. Situ, J.O. Johnson, P.R. Lee, K.L. Chen, L.C. Bott, R. acn3.717 Fadó, G.H. Harmison, S. Parodi, et al. 2015. Mutation in CPT1C Asso- Jacobi, E., and J. von Engelhardt. 2018. AMPA receptor complex constituents: ciated With Pure Autosomal Dominant Spastic Paraplegia. JAMA Neurol. Control of receptor assembly, membrane trafficking and subcellular 72:561–570. https://doi.org/10.1001/jamaneurol.2014.4769 localization. Mol. Cell. Neurosci. 91:67–75. https://doi.org/10.1016/j.mcn Roa-Mansergas, X., R. Fadó, M. Atari, J.F. Mir, H. Muley, D. Serra, and N. .2018.05.008 Casals. 2018. CPT1C promotes human mesenchymal stem cells survival Knapp, B., L. Ospina, and C.M. Deane. 2018. Avoiding False Positive Con- under glucose deprivation through the modulation of autophagy. Sci. clusions in Molecular Simulation: The Importance of Replicas. J. Chem. Rep. 8:6997. https://doi.org/10.1038/s41598-018-25485-7 Theory Comput. 14:6127–6138. https://doi.org/10.1021/acs.jctc.8b00391 Rodrı́guez-Rodrı́guez, R., C. Miralpeix, A. Fosch, M. Pozo, M. Calderón- Kneussel, M., A. Triller, and D. Choquet. 2014. SnapShot: receptor dynamics Domı́nguez, X. Perpinyà, M. Vellvehı́, M. López, L. Herrero, D. Serra, at plastic synapses. Cell. 157:1738–1738.e1. https://doi.org/10.1016/j.cell et al. 2019. CPT1C in the ventromedial nucleus of the hypothalamus is .2014.06.002 necessary for brown fat thermogenesis activation in obesity. Mol. López-Viñas, E., A. Bentebibel, C. Gurunathan, M. Morillas, D. de Arriaga, D. Metab. 19:75–85. https://doi.org/10.1016/j.molmet.2018.10.010 Serra, G. Asins, F.G. Hegardt, and P. Gómez-Puertas. 2007. Definition by Rosa, E.F., S. Takahashi, J. Aboulafia, V.L.A. Nouailhetas, and M.G.M. Oli- functional and structural analysis of two malonyl-CoA sites in carnitine veira. 2007. Oxidative stress induced by intense and exhaustive exer- palmitoyltransferase 1A. J. Biol. Chem. 282:18212–18224. https://doi.org/ cise impairs murine cognitive function. J. Neurophysiol. 98:1820–1826. 10.1074/jbc.M700885200 https://doi.org/10.1152/jn.01158.2006 Mani, M., S.Y. Lee, L. Lucast, O. Cremona, G. Di Paolo, P. De Camilli, and T.A. Samanta, S., A.J. Situ, and T.S. Ulmer. 2014. Structural characterization of the Ryan. 2007. The dual phosphatase activity of synaptojanin1 is required regulatory domain of brain carnitine palmitoyltransferase 1. Bio- for both efficient synaptic vesicle endocytosis and reavailability at nerve polymers. 101:398–405. https://doi.org/10.1002/bip.22396 terminals. Neuron. 56:1004–1018. https://doi.org/10.1016/j.neuron.2007 Schwenk, J., N. Harmel, A. Brechet, G. Zolles, H. Berkefeld, C.S. Müller, W. .10.032 Bildl, D. Baehrens, B. Hüber, A. Kulik, et al. 2012. High-resolution Marinangeli, C., S. Didier, T. Ahmed, R. Caillerez, M. Domise, C. Laloux, S. proteomics unravel architecture and molecular diversity of native Bégard, S. Carrier, M. Colin, P. Marchetti, et al. 2018. AMP-Activated AMPA receptor complexes. Neuron. 74:621–633. https://doi.org/10.1016/ Protein Kinase Is Essential for the Maintenance of Energy Levels during j.neuron.2012.03.034 Synaptic Activation. iScience. 9:1–13. https://doi.org/10.1016/j.isci.2018 Schwenk, J., S. Boudkkazi, M.K. Kocylowski, A. Brechet, G. Zolles, T. Bus, K. .10.006 Costa, A. Kollewe, J. Jordan, J. Bank, et al. 2019. An ER Assembly Line of Mesmin, B., J. Bigay, J. Moser von Filseck, S. Lacas-Gervais, G. Drin, and B. AMPA-Receptors Controls Excitatory Neurotransmission and Its Plas- Antonny. 2013. A four-step cycle driven by PI(4)P hydrolysis directs ticity. Neuron. 104:680–692.e9. https://doi.org/10.1016/j.neuron.2019 sterol/PI(4)P exchange by the ER-Golgi tether OSBP. Cell. 155:830–843. .08.033 https://doi.org/10.1016/j.cell.2013.09.056 Seebohm, G., E. Wrobel, M. Pusch, M. Dicks, J. Terhag, V. Matschke, I. Miñano-Molina, A.J., J. España, E. Martı́n, B. Barneda-Zahonero, R. Fadó, M. Rothenberg, O.N. Ursu, F. Hertel, L. Pott, et al. 2014. Structural basis of Solé, R. Trullás, C.A. Saura, and J. Rodrı́guez-Alvarez. 2011. Soluble PI(4,5)P2-dependent regulation of GluA1 by phosphatidylinositol- oligomers of amyloid-β peptide disrupt membrane trafficking of 5-phosphate 4-kinase, type II, alpha (PIP5K2A). Pflugers Arch. 466: α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid receptor con- 1885–1897. https://doi.org/10.1007/s00424-013-1424-8 tributing to early synapse dysfunction. J. Biol. Chem. 286:27311–27321. Sierra, A.Y., E. Gratacós, P. Carrasco, J. Clotet, J. Ureña, D. Serra, G. Asins, F.G. https://doi.org/10.1074/jbc.M111.227504 Hegardt, and N. Casals. 2008. CPT1c is localized in endoplasmic retic- Morillas, M., E. López-Viñas, A. Valencia, D. Serra, P. Gómez-Puertas, F.G. ulum of neurons and has carnitine palmitoyltransferase activity. J. Biol. Hegardt, and G. Asins. 2004. Structural model of carnitine palmitoyl- Chem. 283:6878–6885. https://doi.org/10.1074/jbc.M707965200 transferase I based on the carnitine acetyltransferase crystal. Biochem. J. Spinelli, M., S. Fusco, M. Mainardi, F. Scala, F. Natale, R. Lapenta, A. Mattera, 379:777–784. https://doi.org/10.1042/bj20031373 M. Rinaudo, D.D. Li Puma, C. Ripoli, et al. 2017. Brain insulin resistance Moult, P.R., A. Cross, S.D. Santos, A.-L. Carvalho, Y. Lindsay, C.N. Connolly, impairs hippocampal synaptic plasticity and memory by increasing A.J. Irving, N.R. Leslie, and J. Harvey. 2010. Leptin regulates AMPA GluA1 palmitoylation through FoxO3a. Nat. Commun. 8:2009. https:// receptor trafficking via PTEN inhibition. J. Neurosci. 30:4088–4101. doi.org/10.1038/s41467-017-02221-9 https://doi.org/10.1523/JNEUROSCI.3614-09.2010 Thacker, J.S., Y. Xu, C. Tang, A.R. Tupling, W.R. Staines, and J.G. Mielke. Nikonova, E.V., N. Naidoo, L. Zhang, M. Romer, J.R. Cater, M.T. Scharf, R.J. 2019. A Single Session of Aerobic Exercise Mediates Plasticity-Related Galante, and A.I. Pack. 2010. Changes in components of energy regu- Phosphorylation in both the Rat Motor Cortex and Hippocampus. lation in mouse cortex with increases in wakefulness. Sleep. 33: Neuroscience. 412:160–174. https://doi.org/10.1016/j.neuroscience.2019 889–900. https://doi.org/10.1093/sleep/33.7.889 .05.051 Ouyang, J., I. Carcea, J.K. Schiavo, K.T. Jones, A. Rabinowitsch, R. Kolaric, S. Tokutake, Y., N. Onizawa, H. Katoh, A. Toyoda, and S. Chohnan. 2010. Co- Cabeza de Vaca, R.C. Froemke, and K.D. Carr. 2017. Food restriction enzyme A and its thioester pools in fasted and fed rat tissues. Biochem. induces synaptic incorporation of calcium-permeable AMPA receptors Biophys. Res. Commun. 402:158–162. https://doi.org/10.1016/j.bbrc.2010 in nucleus accumbens. Eur. J. Neurosci. 45:826–836. https://doi.org/10 .10.009 .1111/ejn.13528 Van Der Spoel, D., E. Lindahl, B. Hess, G. Groenhof, A.E. Mark, and H.J.C. Pozo, M., R. Rodrı́guez-Rodrı́guez, S. Ramı́rez, P. Seoane-Collazo, M. López, Berendsen. 2005. GROMACS: fast, flexible, and free. J. Comput. Chem. D. Serra, L. Herrero, and N. Casals. 2017. Hypothalamic Regulation of 26:1701–1718. https://doi.org/10.1002/jcc.20291 Liver and Muscle Nutrient Partitioning by Brain-Specific Carnitine Venditti, R., M.C. Masone, L.R. Rega, G. Di Tullio, M. Santoro, E. Polishchuk, Palmitoyltransferase 1C in Male Mice. Endocrinology. 158:2226–2238. I.C. Serrano, V.M. Olkkonen, A. Harada, D.L. Medina, et al. 2019. The https://doi.org/10.1210/en.2017-00151 activity of Sac1 across ER-TGN contact sites requires the four-phos- Price, N., F. van der Leij, V. Jackson, C. Corstorphine, R. Thomson, A. Sor- phate-adaptor-protein-1. J. Cell Biol. 218:783–797. https://doi.org/10 ensen, and V. Zammit. 2002. A novel brain-expressed protein related to .1083/jcb.201812021 carnitine palmitoyltransferase I. Genomics. 80:433–442. https://doi.org/ Vivas, O., S.A. Tiscione, R.E. Dixon, D.S. Ory, and E.J. Dickson. 2019. 10.1006/geno.2002.6845 Niemann-Pick Type C Disease Reveals a Link between Lysosomal Ramı́rez, S., L. Martins, J. Jacas, P. Carrasco, M. Pozo, J. Clotet, D. Serra, F.G. Cholesterol and PtdIns(4,5)P2 That Regulates Neuronal Excitability. Cell Hegardt, C. Diéguez, M. López, et al. 2013. Hypothalamic ceramide Rep. 27:2636–2648.E4. https://doi.org/10.1016/j.celrep.2019.04.099 levels regulated by CPT1C mediate the orexigenic effect of ghrelin. Von Filseck, J.M., A. Čopič, V. Delfosse, S. Vanni, C.L. Jackson, W. Bourguet, Diabetes. 62:2329–2337. https://doi.org/10.2337/db12-1451 and G. Drin. 2015. Phosphatidylserine transport by ORP/Osh proteins is Rao, J.N., G.Z.L. Warren, S. Estolt-Povedano, V.A. Zammit, and T.S. Ulmer. driven by phosphatidylinositol 4-phosphate. Science. 349:432–436. 2011. An environment-dependent structural switch underlies the reg- https://doi.org/10.1126/science.aab1346 ulation of carnitine palmitoyltransferase 1A. J. Biol. Chem. 286:42545–42554. Wolfgang, M.J., T. Kurama, Y. Dai, A. Suwa, M. Asaumi, S. Matsumoto, S.H. https://doi.org/10.1074/jbc.M111.306951 Cha, T. Shimokawa, and M.D. Lane. 2006. The brain-specific carnitine Ribeiro, L.F., T. Catarino, S.D. Santos, M. Benoist, J.F. van Leeuwen, J.A. Es- palmitoyltransferase-1c regulates energy homeostasis. Proc. Natl. Acad. teban, and A.L. Carvalho. 2014. Ghrelin triggers the synaptic incorporation Sci. USA. 103:7282–7287. https://doi.org/10.1073/pnas.0602205103 Casas et al. Journal of Cell Biology 23 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Wolfgang, M.J., S.H. Cha, A. Sidhaye, S. Chohnan, G. Cline, G.I. Shulman, and metabolic stress. Genes Dev. 25:1041–1051. https://doi.org/10.1101/gad M.D. Lane. 2007. Regulation of hypothalamic malonyl-CoA by central .1987211 glucose and leptin. Proc. Natl. Acad. Sci. USA. 104:19285–19290. https:// Zewe, J.P., R.C. Wills, S. Sangappa, B.D. Goulden, and G.R. Hammond. 2018. doi.org/10.1073/pnas.0709778104 SAC1 degrades its lipid substrate PtdIns4P in the endoplasmic reticulum Yang, G., X. Zhou, J. Zhu, R. Liu, S. Zhang, A. Coquinco, Y. Chen, Y. Wen, L. to maintain a steep chemical gradient with donor membranes. eLife. 7. Kojic, W. Jia, et al. 2013. JNK3 couples the neuronal stress response to e35588. https://doi.org/10.7554/eLife.35588 inhibition of secretory trafficking. Sci. Signal. 6:ra57. https://doi.org/10 Zimmer, P., S. Binnebößel, W. Bloch, S.T. Hübner, A. Schenk, H.G. .1126/scisignal.2003727 Predel, P. Wright, C. Stritt, and M. Oberste. 2017. Exhaustive ex- Zaugg, K., Y. Yao, P.T. Reilly, K. Kannan, R. Kiarash, J. Mason, P. Huang, S.K. ercise alters thinking times in a tower of london task in a time- Sawyer, B. Fuerth, B. Faubert, et al. 2011. Carnitine palmitoyltransferase dependent manner. Front. Physiol. 7:694. https://doi.org/10.3389/ 1C promotes cell survival and tumor growth under conditions of fphys.2016.00694 Casas et al. Journal of Cell Biology 24 of 24 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Supplemental material Casas et al. Journal of Cell Biology S1 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S1. Whole-cell images of cortical neurons showing GluA1 surface levels under different treatments. (A–D) CPT1C-silenced or control cortical neurons were submitted to glucose deprivation for 2 h (A), 2DG treatment (15 mM, 2 h; B), leptin treatment (50 nM, 30 min; C), or TOFA treatment (20 µg/ml, 2 h; D). A short hairpin-loop RNA with a random sequence (shRandom) was used as a control. Surface GluA1 was detected by immunocytochemistry in nonpermeabilized cells. Lentiviral infection was confirmed by the detection of the GFP tag. Complete images used in Fig. 1 are shown. The selected ROI in each image is identified with a rectangular box. Scale bars = 7 µm for A, B, and D and 10 µm for C. Casas et al. Journal of Cell Biology S2 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S2. Glucose deprivation, 2DG, and TOFA treatments have no effect on cortical neuron survival. (A) CPT1C silencing in cortical neurons. Cells were infected at 7 DIV with a shCPT1C-carrying lentivirus or a short hairpin-loop RNA with a random sequence (shRandom) was used as a control. At 14–15 DIV, CPT1C down-regulation was confirmed by Western blot. A representative image is shown. Graph shows the mean ± SD from three independent experiments performed by biological replicates (n = 8 per condition; two-sided Student’s t test; ****, P < 0.0001; 1.00 ± 0.15 for shRandom and 0.32 ± 0.17 for shCPT1C). (B) GluA1 and TGN38 detection in nonpermeabilized and permeabilized neurons. To make sure that chemical fixation with 4% of paraformaldehyde does not induce permeabilization itself, GluA1 and TGN38 were immunodetected by immunocytochemistry in cultured cortical neurons permeabilized or non- permeabilized with 0.1% (vol/vol) of Triton X-100. As expected, only surface GluA1 was stained in nonpermeabilized neurons, since TGN38 (a Golgi-resident protein) and total GluA1 (surface and internal) were detected in permeabilized cells. Scale bars = 20 µm. (C–E) Cell survival after 2 h of glucose starvation (C), 2DG treatment (15 mM; D), or TOFA treatment (20 µg/ml; E) in cortical neurons was determined by MTT assay. Values are the mean ± SD from two inde- pendent experiments performed by biological replicates (two-sided Student’s t test; P > 0.05). (C) 25 mM glucose (100.00 ± 33.24, n = 17) and 0 mM glucose (75.25 ± 17.33, n = 15). n.s, not significant. (D) Vehicle (100.00 ± 13.22, n = 12) and TOFA (91.55 ± 12.63, n = 11). (E) Vehicle (100.00 ± 18.99, n = 10) and 2DG (104.7 ± 17.45, n = 7). (F) TOFA treatment (microgram per milliliter, 2 h) down-regulates GluA1 surface levels in hippocampal neurons. All data represent mean ± SEM of two independent experiments (n = 20 cells per condition; two-sided Student’s t test; ***, P = 0.0004; 1.00 ± 0.12 for vehicle and 0.47 ± 0.06 for TOFA). (G) Reversibility of TOFA-induced effects on GluA1 surface levels. Cultured cortical neurons were treated at 14 DIV with TOFA (microgram per milliliter) or vehicle (DMSO; 1:500). 2 h later, TOFA was removed, replacing the medium with conditioned media (Cond. medium) from sibling plates, and cells were in- cubated for an additional 4 h. The rest of the wells were fixed at 2 h after treatment. In all conditions, surface GluA1 levels were detected by immunocy- tochemistry. Data represent mean ± SEM of one experiment performed by biological duplicates (one-way ANOVA; **, P < 0.01 and ***, P < 0.001). Vehicle (1.00 ± 0.15, n = 33), TOFA (0.39 ± 0.05, n = 37), and reversion (0.87 ± 0.07, n = 34). Scale bars = 5 µm. ns, not significant. Casas et al. Journal of Cell Biology S3 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S3. TOFA treatment decreases the levels of GluA1 but not GluA2 and GluN2A at the surface of cortical neurons in a CPT1C-dependent manner. (A) Surface GluA1 quantification by immunochemistry in nonpermeabilized WT and CPT1C KO cortical neurons treated with TOFA. Values are the mean ± SEM of two independent experiments (two-way ANOVA followed by Bonferroni’s comparison test; **, P < 0.01 and ***, P < 0.001). WT + vehicle (1.00 ± 0.12, n = 67), WT + TOFA (0.41 ± 0.08, n = 45), KO + vehicle (0.53 ± 0.11, n = 36), and KO + TOFA (0.52 ± 0.10, n = 38). (B) Surface GluA1 levels measured by the biotinylation assay in WT and CPT1C KO cortical neurons. Results are given as the mean ± SD from three independent experiments performed by biological replicates (n = 4 samples per condition; two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 for surface levels). WT + vehicle (57.54 ± 7.47 for surface and 42.46 ± 7.47 for internal, n = 4), WT + TOFA (13.05 ± 7.96 for surface and 86.95 ± 7.96 for internal, n = 4), KO + vehicle (28.31 ± 16.55 for surface and 71.69 ± 16.55 for internal, n = 4), and KO + TOFA (31.95 ± 14.05 for surface and 68.07 ± 14.05 for internal, n = 4). (C) Surface GluN2A and GluA2 quantification by biotinylation in cortical neurons. Results are mean ± SD from two independent experiments performed by biological replicates (two-way ANOVA followed by Bonferroni’s comparison test; P > 0.05). GluN2A vehicle (12.78 ± 7.17 for surface and 87.22 ± 7.17 for internal, n = 4), GluN2A TOFA (12.55 ± 6.87 for surface and 87.45 ± 6.87 for internal, n = 4), GluA2 vehicle (22.59 ± 10.24 for surface and 77.40 ± 10.24 for internal, n = 5), and GluA2 TOFA (19.22 ± 12.43 for surface and 80.78 ± 12.43 for internal, n = 5). (D) Surface GluA2 quantification by immunocytochemistry in WT and CPT1C KO cortical neurons. Results are mean ± SEM of one experiment performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; ***, P < 0.001). WT + vehicle (1.00 ± 0.10, n = 37), WT + TOFA (0.92 ± 0.07, n = 34), KO + vehicle (0.41 ± 0.06, n = 32), and KO + TOFA (0.37 ± 0.04, n = 24). Scale bars = 5 µm. Casas et al. Journal of Cell Biology S4 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S4. Effects of TOFA treatment or SAC1 modulation on PI(4)P levels, TGN volume, and TGN-ER occurrence. (A) Quantification of PI(4)P levels using the specific plasmid mCherry-P4M at the TGN compartment (TGN46) in WT or CPT1C KO neurons after TOFA treatment (20 µg/ml, 2 h). Data represent the mean ± SEM from two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and ***, P < 0.001). WT + vehicle (1.00 ± 0.03, n = 26), WT + TOFA (0.87 ± 0.03, n = 26), KO + vehicle (0.92 ± 0.02, n = 20), and KO + TOFA (1.00 ± 0.02, n = 28). Scale bar = 7 µm. (B) SAC1 modulation (overexpressing, OE, or knocking down, KD) does not change TGN volume. The same images used in Fig. 6, B and E were taken to analyze TGN volume in KD or OE SAC1 neurons. Values are the mean ± SD from two independent experiments (Mann-Whitney U test and P = 0.8444 for KD experiment; two-sided Student’s t test and P = 0.3062 for OE experiment). For KD experiment, short hairpin-loop RNA with a random sequence (shRandom; 1.00 ± 0.90, n = 10) and shSAC1 (0.99 ± 1.11, n = 11). For OE experiment, EV (1.00 ± 0.33, n = 15) and SAC1 (1.20 ± 0.64, n = 18). (C) TGN- ER occurrence increases upon TOFA treatment. The same images used in Fig. 7, B–D were taken to analyze the percentage of TGN that was forming contacts with the ER. Results are the ratio between the TGN-ER contact area and the TGN area per cell, given as a percentage. Data represent the mean ± SD of two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; *, P < 0.05 and **, P < 0.01). EV + vehicle (0.49 ± 0.13, n = 13), EV + TOFA (0.71 ± 0.08, n = 13), CPT1C + vehicle (0.56 ± 0.23, n = 13), and CPT1C + TOFA (0.67 ± 0.11, n = 13). (D) CPT1C intensity at ER-TGN contact sites. The images used in Fig. 7 E were also used for analyzing CPT1C intensity inside the ROIs of the ER, TGN, and ER-TGN contacts (CSs) relative to the total CPT1C intensity in the whole cell (expressed in percentage). The graph represents the mean ± SD from two independent experiments performed by biological duplicates (two-way ANOVA followed by Bonferroni’s comparison test; ****, P < 0.001). ER vehicle (50.37 ± 11.83, n = 13), ER TOFA (52.22 ± 11.69, n = 15), TGN vehicle (7.74 ± 2.73, n = 13), TGN TOFA (21.99 ± 7.02, n = 15), ER-TGN vehicle (7.04 ± 2.72, n = 13), and ER-TGN TOFA (19.43 ± 7.66, n = 15). Casas et al. Journal of Cell Biology S5 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S5. Rapamycin-induced ER-TGN contacts. (A) No artificial contacts were formed without the addition of rapamycin. HEK cells were transfected and treated as in Fig. 8 A. PI(4)P levels in the TGN were live imaged with confocal microscopy. TOFA addition in these conditions does not have any effect. Data represent the mean ± SD from one experiment performed by biological replicates (n = 5 or 6 cells per condition). Scale bars = 7 µm; scale bars of inset magnifications = 1.75 µm. (B) Effect of CPT1C expression and TOFA treatment on PI(4)P levels at the TGN after induction of artificial ER-TGN contacts. HEK cells were transfected with mCherry-pHR-TcRb-FKBP and CFP-TGN38-FRB plasmids to induce artificial ER-TGN by rapamycin (5 µM, 30 min). PI(4)P was labeled with the lipid probe YFP-P4M. Cells were treated for 1 h with TOFA or vehicle 24 h after transfection. Cells were live imaged with confocal microscopy. PI(4)P, TGN, and the ROI of PI(4)P inside the TGN are shown. Individual cells are the same as the ones analyzed in Fig. 8 A. Casas et al. Journal of Cell Biology S6 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045 Figure S6. CPT1C deficiency decreases the brain levels of different PIP and PIP2 species. (A–C) Mass spectrometry analysis for PI, PIP, and PIP2 species. Histogram of the most abundant PI (A), PIP (B), and PIP2 (C) species in WT (gray bars) and CPT1C−/−(blue bars) brains (mean ± SEM; n = 3). Casas et al. Journal of Cell Biology S7 Regulation of GluA1 trafficking by CPT1C/SAC1 axis https://doi.org/10.1083/jcb.201912045