Regulation of AMPK Activation by CD36 Links Fatty Acid Uptake to β-Oxidation

AMPK plays an important role in muscle fuel preference and flexibility (1,2). AMPK is normally quiescent, and its activation in response to energy stresses suppresses ATP consumption while upregulating nutrient uptake and catabolism. AMPK activation in muscle involves phosphorylation of the catalytic α-subunit threonine 172 (T172) by upstream kinases, primarily LKB1 (liver kinase B1). Activated AMPK induces sarcolemmal recruitment of the glucose and fatty acid (FA) transporters GLUT4 and CD36 (3), and also upregulates long-chain FA β-oxidation by inactivating acetyl-CoA carboxylase 2, reducing levels of the β-oxidation inhibitor malonyl-CoA. Dysregulation of AMPK signaling in skeletal muscle associates with a diminished capacity to adjust FA oxidation to FA availability, leading to lipid accumulation and insulin resistance (4).

Similar to AMPK, CD36, which facilitates sarcolemmal FA transport, is critical to the adaptive response of muscle FA metabolism. Fasting, muscle contraction, and metabolic signals (FoxO1, AMPK), which are associated with increased energy needs, recruit CD36 to the sarcolemma (3,5). The resulting coordinated increases of FA uptake and oxidation contribute to the adaptability of FA metabolism (3,5,6). The muscle of insulin-resistant humans and that of obese rodents display persistent sarcolemmal CD36 relocation and intracellular lipid accumulation (1).

CD36 plays an important role in modulating events related to lipid use (5). FA binding to CD36 increases levels of cAMP and cytosolic calcium, and mediates taste cell neurotransmitter release for fat taste perception (7) or gut peptide secretions that optimize fat absorption (8). In the current study, we examined the possibility that sarcolemmal CD36 signaling might engage AMPK to coordinate FA uptake and oxidation.

FA-free BSA, sodium palmitate, and other chemicals were from Sigma-Aldrich (St. Louis, MO); and human highly oxidized LDL (oxLDL) was from Kalen Biomedical (Montgomery Village, MD). All phospho-antibodies were from Cell Signaling Technology (Danvers, MA), anti-Fyn was from Santa Cruz Biotechnology (Santa Cruz, CA), GAPDH was from Sigma-Aldrich, and anti-LKB1 was from Millipore (Billerica, MA). IRDye800 anti-rabbit or anti-goat IgG and IRDye700 anti-mouse IgG were from Rockland (Gilbertsville, PA). Anti-human and anti-mouse CD36 were from R&D Systems (Minneapolis, MN), rat monoclonal anti-mouse CD36 was from AbD Serotec (Raleigh, NC), and mouse monoclonal anti-human CD36 was from Abcam (Cambridge, MA). Alexa Fluor IgGs (594 anti-rabbit, 488 goat anti-mouse, and 488 donkey anti-goat) were from Life Technologies. Mouse CD36 small interfering RNAs (siRNA) (siCD36–1: 5′-AAACCCAGATGACGTGGCAAA-3′; siCD36–2: 5′-AACGACATGATTAATGGCACACCTGTCTC–3′) were from Life Technologies; and pcDNA3-FLAG-LKB1 (catalog #8591), pRK5 c-Fyn (catalog #16032), and pRK5 DN-Fyn (catalog #16033) were from Addgene.

Age-matched (3- to 4-month-old), male, wild-type (WT) C57BL/6 and CD36^−/− mice (9) backcrossed onto the C57BL/6 line for more than seven generations were fed ad libitum or fasted for 18 h. All protocols were approved by the Animal Studies Committee of Washington University.

Respiratory quotient (RQ; Vco₂/Vo₂ ratio) was determined by indirect calorimetry (ULTRAMAT/OXYMAT 6 analyzer; Siemens) (10), heart triglyceride (TG) content was determined by multidimensional mass spectrometry (TSQ Quantum Ultra mass spectrometer; Thermo Electron) (11), and quadriceps TG content was determined using a kit from Wako Diagnostics (Richmond, VA).

C2C12 myoblasts grown in high-glucose DMEM with 10% FBS (Life Technologies) supplemented with l-glutamine (2 mmol/L) were differentiated in DMEM containing 2% horse serum. CD36 knockdown in C2C12 myotubes used transfection (Lipofectamine RNAiMAX) with siRNAs (20 nmol/L final), anti-mouse CD36 (Life Technologies), or anti-GFP (controls). All siRNA-transfected myotubes appeared morphologically normal and were used 72 h after transfection.

CHO cells stably expressing CD36 (12) were transiently transfected (Lipofectamine 2000; Life Technologies) then 48 h later were serum starved and treated as indicated. C2C12 myotubes or CHO cells were serum starved (16–18 h) in buffer A (low-glucose, 5 mmol/L DMEM with 2 mmol/L l-glutamine, 100 μmol/L MEM nonessential amino acids, 100 units/mL penicillin, and 100 μg/mL streptomycin). FAs were added with BSA at a ratio of 2:1.

Tissue extracts and cells were lysed (30 min) in ice-cold buffer containing 60 mmol/L Octyl (octylβ-d-glucopyranoside) or 1% Brij99 (polyoxyethylene [20]-oleylether) when indicated. For immunoprecipitation (IP), equal amounts of lysate proteins were incubated (4°C) with antibodies overnight before adding protein G or A agarose beads (Thermo Scientific) (1–3 h). After five washes with cold lysis buffer, bound proteins were eluted by boiling for 5 min in SDS sample buffer, and cleared (10,000g, 3 min) supernatants were resolved by SDS-PAGE.

For immunofluorescence (IF), cells grown on coverslips were fixed with ice-cold methanol (−20°C, 20 min), blocked (1 h, PBS with 0.05% Tween 20, 1% BSA, 5% goat serum) at room temperature, and incubated overnight with primary antibodies then for 1 h with Alexa Fluor 488 and/or 594 secondary antibodies. Images were acquired using an inverted microscope (Axiovert 200 M; Zeiss, Thornwood, NY) and ImageJ software (National Institutes of Health).

Palmitate (100 μmol/L, 2:1 with BSA)-supported respiration was measured in intact or digitonin (Dig)-permeabilized (10 μg/10⁻⁶ cells) C2C12 myotubes (Oxygraph-2k; OROBOROS Instruments) at 37°C with room air saturation (13). Cells were serum starved for 16 h, trypsinized, and resuspended in DMEM buffer A (for intact cells) or in MIR05 buffer with 1 mmol/L ATP, 5 mmol/L ADP, 2 mmol/L malate, and 2 mmol/L l-carnitine (for permeabilized cells) (13).

Statistical significance was determined using an unpaired Student t test or one-way ANOVA followed by Tukey post hoc test for multiple comparisons.

Long-chain FA addition (1 h) to L6 myotubes enhanced LKB1-mediated AMPK phosphorylation (14). In C2C12 myotubes, we confirmed that 5- to 15-min incubations with palmitic acid (PA) enhance T172 phosphorylated AMPK (pAMPK) and S79 phosphorylation of the AMPK target acetyl-CoA carboxylase 2 (data not shown). We examined the CD36 dependence of AMPK activation by FA using two siRNAs, siCD36–1 and siCD36–2, which markedly reduced CD36 levels (Fig. 1A). CD36 knockdown robustly increased pAMPK, suggesting that CD36 expression suppresses AMPK phosphorylation (Fig. 1A).

Palmitate addition increased pAMPK levels in control myotubes, but not in CD36-depleted myotubes where basal pAMPK was already high (Fig. 1B). Similar results were obtained with oleate (data not shown). CD36 knockdown decreased PA-supported respiration by myotubes (Fig. 1C), which is consistent with its impairing FA uptake (3,5). However, knockdown increased respiration when myotubes were permeabilized with Dig (13) to bypass CD36-mediated FA uptake (Fig. 1C), which is consistent with knockdown enhancing FA oxidation. Thus, CD36 depletion increases basal pAMPK and FA oxidation. In nondepleted cells, AMPK activation by exogenous FA requires CD36.

AMPK phosphorylation was increased in skeletal muscle (quadriceps) and heart from CD36^−/− mice fasted for 18 h compared with WT mice (Fig. 1D), indicating that CD36 deletion activates muscle AMPK in vivo. Since CD36 deficiency reduces the availability of exogenous FA for oxidation (3), we examined whether AMPK activation would enhance FA oxidation from endogenous stores. We first compared the RQ (ratio of CO₂ produced to O₂ consumed) of fed and fasted (18 h) WT and CD36^−/− mice (Fig. 1E). RQ was similar across genotype in the fed state, as previously reported (10), but fasting-induced reduction of RQ was significantly more pronounced in CD36^−/− mice (Fig. 1E), which is consistent with enhanced FA oxidation.

We next measured muscle TG content. Fed CD36^−/− mice had lower myocardial TG content than WT mice, while quadriceps TG content was similar (Fig. 1F). Fasting increased the TG content of WT hearts 2.6-fold, as previously reported (15), while TG content of CD36^−/− hearts decreased (Fig. 1F). Quadriceps TG content trended higher in WT mice, whereas it decreased in CD36^−/− mice (Fig. 1F). This suggested that in the fasted CD36^−/− mouse muscle, endogenous lipids support FA oxidation, depleting intramuscular TG content.

Several kinases phosphorylate LKB1, the major AMPK kinase in muscle, including the src kinase Fyn (16), which interacts with CD36. Fyn tyrosine phosphorylates LKB1, resulting in its nuclear sequestration and reduced access to cytosolic AMPK (16). We examined the effect of CD36 expression on LKB1 phosphorylation and its Fyn dependence using CHO cells stably expressing CD36 and transiently transfected with native Fyn or dominant-negative Fyn (FynD) (17). CD36 expression reduced pAMPK levels (Fig. 2A). LKB1 IP (Fig. 2B) and probing with PY100 phosphotyrosine antibody showed increased phosphorylated LKB1 in CD36 cells transfected with Fyn, and this increase was abolished in cells expressing FynD. Thus, CD36 expression promotes Fyn-dependent LKB1 phosphorylation.

IF in CHO (Fig. 2C) and HeLa cells (data not shown) with or without stable CD36 expression showed that CD36 promotes nuclear LKB1 localization. LKB1 was distributed throughout the cytoplasm in controls (Fig. 2C, top panels) but was predominantly nuclear in CD36 cells (Fig. 2C, bottom panels) (nuclear/cytosolic LKB1 in controls and CD36 cells, respectively: 0.25 ± 0.02 and 1.20 ± 0.06; P < 0.001). CD36 regulation of LKB1 was observed in C2C12 myotubes (Fig. 2D). In controls, LKB1 was mostly nuclear (Fig. 2D, top panels), while CD36 depletion redistributed LKB1 to the cytosol (Fig. 2D, bottom panels). FA addition (PA, 15 min) to CHO-CD36 cells (Fig. 3A) redistributed LKB1 from the nucleus to cytosolic structures that coincided with CD36 (Fig. 3A, inset). Similar PA redistribution of LKB1 occurred in HeLa (data not shown) and in C2C12 cells (Fig. 3B). Thus, FA addition recruits LKB1 from nuclei to cytosolic CD36-positive vesicles.

The LKB1-CD36 association was confirmed by CD36 IP in CHO-CD36 cells (Fig. 3C) solubilized with Brij99, which preserves hydrophobic interactions between membrane proteins, including CD36 (18). CD36 IP contained LKB1, and content was increased by palmitate. Flotillin-1, a lipid raft marker, was recovered from CD36 IP, suggesting that the CD36-LKB1 association might occur in lipid rafts (Fig. 3C). CHO-CD36 cells were then solubilized with Octyl, which efficiently extracts membrane proteins (19) (Fig. 4). Flotillin-1 was then absent from CD36 IP (Fig. 4A), but LKB1 was still recovered, and PA increased this recovery (Fig. 4B). These data supported CD36-LKB1 association independent of membrane lipid domains and FA enhancement of this interaction.

We next examined whether the effect of FA on LKB1 might involve modulation of the CD36 association with Fyn, and potentially with AMPK. The CD36 ligand oxLDL enhances CD36 association with src kinases in platelets (20). The effect of FA is unknown, so CD36 immunoprecipitates from CHO-CD36 cells incubated with or without PA or oxLDL were probed for coassociated Fyn and AMPK.

The addition of palmitate did not alter the amounts of Fyn found in cell lysates (Fig. 4B), but markedly decreased Fyn recovery in CD36 IP, suggesting a reduced CD36-Fyn association. AMPK was identified in CD36 IP, and its recovery increased after PA (Fig. 4B). In contrast to PA, incubation of CD36-expressing cells with oxLDL (Fig. 4C) increased the amount of Fyn pulled down by CD36, as in platelets (20), without altering the amount of LKB1 recovered. Thus, PA addition specifically decreases the CD36-Fyn interaction while increasing the CD36-LKB1-AMPK interaction.

Our findings support CD36 control of LKB1 and AMPK activity, which might have relevance to the maintenance of cellular FA homeostasis. CD36 regulation of AMPK appears to be linked to optimal AMPK responsiveness to FA supply and is exerted through the assembly of a CD36/Fyn/LKB1/AMPK protein complex (Fig. 4D). Under low exogenous FA availability, this complex maintains quiescence of AMPK by promoting Fyn-dependent LKB1 phosphorylation and LKB1 relocation to the nucleus. As exogenous FA become more available, FA binding to CD36 dissociates Fyn, impairing its access to CD36-associated LKB1. The ensuing cytosolic enrichment in LKB1 promotes LKB1 activation of AMPK. The data support the concept of CD36 functioning as a component of signaling protein complexes with CD36 ligands differentially modulating protein associations within these complexes and consequently the access of CD36 partner kinases to downstream phosphorylation targets.

Evidence of CD36 regulation of AMPK in vivo was shown in fasted CD36^−/− mice from the basally increased muscle and heart pAMPK, from the low RQ, and from the depletion of endogenous TG. Thus, while oxidation of exogenous FA, measured using FA analogs, was shown to be reduced in CD36-deficient mice and humans due to impaired FA uptake (3,5), oxidation of FA from endogenous supply would be enhanced. CD36 dysfunction, as occurs with chronic excess FA (21) or diabetes (22), might alter the lipid-sensing ability of AMPK, contributing to the etiology of lipid-associated metabolic complications.

In summary, this study documents the close coordination between CD36 and AMPK in the regulation of cellular FA metabolism. CD36 maintains basal AMPK quiescence, whereas FA-induced CD36 signaling activates AMPK, which disinhibits β-oxidation and recruits more CD36 to the sarcolemma. The positive reinforcement of FA catabolism during high FA supply would help to maintain cellular FA homeostasis. However, its dysfunction might contribute to the abnormalities of FA metabolism observed in diabetes and to the association of common CD36 variants with the metabolic syndrome (23).

Acknowledgments. The authors acknowledge the assistance of Alice Tong with mouse husbandry and genotyping and the Mouse Phenotyping Core and the Adipocyte Biology and Molecular Nutrition Core of the Nutrition and Obesity Research Center.

Funding. This study was supported by National Institutes of Healthhttp://dx.doi.org/10.13039/100000002 (NIH) grants R01-DK-033301, R01-DK-060022, and R01-DK-097608 and by National Natural Science Foundation of China grant 31371437. The Mouse Phenotyping Core and the Adipocyte Biology and Molecular Nutrition Core of the Nutrition and Obesity Research Center were supported by NIH grant P30-DK-056341.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. D.S. designed the study, conducted the research, analyzed the data, and wrote the manuscript. J.S. and T.P. conducted the research and analyzed the data. R.W.G., R.H.E., X.S., and P.D.S. contributed to the data analysis or provided reagents. N.A.A. designed the study and wrote the manuscript. D.S. and N.A.A. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.