Lipophagy-derived fatty acids undergo extracellular efflux via lysosomal exocytosis

ABSTRACT

The autophagic degradation of lipid droplets (LDs), termed lipophagy, is a major mechanism that contributes to lipid turnover in numerous cell types. While numerous factors, including nutrient deprivation or overexpression of PNPLA2/ATGL (patatin-like phospholipase domain containing 2) drive lipophagy, the trafficking of fatty acids (FAs) produced from this pathway is largely unknown. Herein, we show that PNPLA2 and nutrient deprivation promoted the extracellular efflux of FAs. Inhibition of autophagy or lysosomal lipid degradation attenuated FA efflux highlighting a critical role for lipophagy in this process. Rather than direct transport of FAs across the lysosomal membrane, lipophagy-derived FA efflux requires lysosomal fusion to the plasma membrane. The lysosomal Ca2+ channel protein MCOLN1/TRPML1 (mucolipin 1) regulates lysosomal-plasma membrane fusion and its overexpression increased, while inhibition blocked FA efflux. In addition, inhibition of autophagy/lipophagy or MCOLN1, or sequestration of extracellular FAs with BSA attenuated the oxidation and re-esterification of lipophagy-derived FAs. Overall, these studies show that the well-established pathway of lysosomal fusion to the plasma membrane is the primary route for the disposal of FAs derived from lipophagy. Moreover, the efflux of FAs and their reuptake or subsequent extracellular trafficking to adjacent cells may play an important role in cell-to-cell lipid exchange and signaling.

Abbreviations: ACTB: beta actin; ADRA1A: adrenergic receptor alpha, 1a; ALB: albumin; ATG5: autophagy related 5; ATG7: autophagy related 7; BafA1: bafilomycin A1; BECN1: beclin 1; BHBA: beta-hydroxybutyrate; BSA: bovine serum albumin; CDH1: e-cadherin; CQ: chloroquine; CTSB: cathepsin B; DGAT: diacylglycerol O-acyltransferase; FA: fatty acid; HFD: high-fat diet; LAMP1: lysosomal-associated membrane protein 1; LD: lipid droplet; LIPA/LAL: lysosomal acid lipase A; LLME: Leu-Leu methyl ester hydrobromide; MAP1LC3B/LC3: microtubule associated protein 1 light chain 3 beta; MCOLN1/TRPML1: mucolipin 1; MEF: mouse embryo fibroblast; PBS: phosphate-buffered saline; PIK3C3/VPS34: phosphatidylinositol 3-kinase catalytic subunit type 3; PLIN: perilipin; PNPLA2/ATGL patatin-like phospholipase domain containing 2; RUBCN (rubicon autophagy regulator); SM: sphingomyelin; TAG: triacylglycerol; TMEM192: transmembrane protein 192; VLDL: very low density lipoprotein.

Introduction

Autophagy is a cellular recycling mechanism that provides cells a source of energy during nutrient deprivation [1]. Autophagy can be classified into three categories based on lysosomal cargo delivery: macroautophagy [1], chaperone-mediated autophagy [2], and microautophagy [3]. The role of autophagy in triacylglycerol (TAG) hydrolysis was unclear until Singh et al. showed that inhibiting autophagy promotes LD accumulation and attenuates oxidation of LD-derived FAs [4]. Since these findings of autophagy-mediated lipid mobilization, termed lipophagy, a plethora of studies have shown the existence of lipophagy in diverse cell types such as macrophages [5], neurons [6], lymphocytes [7], and among organisms [8]. Lipophagy has been most studied in the liver, where is a critical factor regulating LD turnover.

PNPLA2/ATGL (patatin-like phospholipase domain containing 2) was discovered in 2004 and was initially identified to be the major TAG lipase in adipose tissue and heart [9]. Since this time, numerous studies have expanded these original findings to show that PNPLA2 regulates TAG turnover in many other tissues, including the liver, intestine, skeletal muscle, and immune cells [10–13]. Within the context of the liver, shRNA knockdown or genetic ablation of hepatic PNPLA2 promotes steatosis [10,14], whereas overexpression of PNPLA2 in the liver alleviates steatosis [15]. Numerous studies show that PNPLA2 specifically channels hydrolyzed fatty acids (FAs) to oxidative pathways and does not influence very-low-density lipoprotein (VLDL) secretion [10,14]. Thus, in addition to its prominent roles in tissues such as adipose and heart, PNPLA2 also appears to play a critical role in regulating hepatic energy metabolism and signaling.

Since PNPLA2-catalyzed lipolysis and lipophagy may both contribute to hepatic LD degradation, more recent studies have provided insight into the crosstalk between the two pathways. Chaperone-mediated autophagy has been shown to contribute to hepatic LD catabolism via its degradation of the LD proteins PLIN2 (perilipin 2) and PLIN3, which subsequently allows PNPLA2 access to LDs to promote lipolysis [16]. Additionally, ablating the LC3-interacting regions on PNPLA2 prevents its colocalization with LDs suggesting that autophagy and cytosolic lipolysis are tightly intertwined [17]. We have extended this work to show that PNPLA2 promotes autophagy/lipophagy and that PNPLA2-mediated bulk degradation of LDs requires lipophagy [18]. Consistent with these findings, hepatic ablation of PLIN2, which promotes PNPLA2-catalyzed lipolysis, requires lipophagy to facilitate LD degradation [19]. More recent work also shows that PNPLA2 acts on large lipid droplets to initiate catabolism, while lipophagy drives the degradation of smaller lipid droplets [20]. While these studies highlight the importance of lipophagy as a major contributor to LD turnover, the metabolic channeling of the FAs derived from this process have not been characterized.

A major function of autophagy is to produce nutrients (amino acids, glucose, and fatty acids) that can be oxidized to provide cells with energy during nutrient deprivation. While lysosomal transporters for glucose and amino acids have been identified [21–23], the mechanism through which lipophagy-derived fatty acids exit lysosomes for subsequent oxidation or re-esterification has not been explored. MCOLN1/TRPML1 (mucolipin 1) is a lysosomal cation channel protein predominantly located on the late-endosomal-lysosomal transmembrane. Accumulating evidence suggests that MCOLN1 is a critical regulator of lysosome function, including exocytosis, and thus contributes to the regulation of lipid metabolism [24–26]. Therefore, we examined the involvement of MCOLN1 in fasting and PNPLA2-mediated lipophagy. Herein, we demonstrate that hepatic PNPLA2 and nutrient deprivation promote the expulsion of FAs, which requires lipophagy and lysosomal fusion with the plasma membrane.

Results

Nutrient deprivation and PNPLA2 promote FA efflux

Pulse-chase experiments employing radiolabeled FAs are a standard procedure for measuring lipid turnover in cells. However, these studies commonly use media that is completely or largely devoid of albumin (ALB), the primary FA carrier outside of cells. Similarly, studies utilizing nutrient deprivation conditions that promote autophagy/lipophagy routinely remove FBS, a source of various lipid-binding proteins including ALB. The exception in such studies is the measurement of lipolysis in adipose tissue, where FA efflux is expected and ALB is included in the media to bind the extracellular FAs and limit reuptake. To explore this pathway in more detail in hepatocytes, we adapted a pulse-chase assay using BODIPY-conjugated C16 FA to measure effluxed FA in cell culture media.

It should be noted that we used bovine serum albumin (BSA) (FA-free) at 2% w/v (~300 μM), which is below the normal concentration of ALB (~400 μM) in mouse serum [27]. The presence of BSA allowed the detection of effluxed FA under fasting media conditions (serum deprivation), which represented ~10-fold change compared to the absence of BSA (Figure 1A). We conducted similar pulse-chase assays in mouse embryo fibroblasts (MEF), Hep3B, and HepG2 cells (Figure 1B–D). While the extent of FA efflux varied, all cell types tested exhibited FA efflux in the presence of BSA. While BSA did not alter fasting-mediated autophagy flux, as measured by using the GFP-LC3-RFP-LC3ΔG plasmid, in primary hepatocytes, it did attenuate autophagy flux in the other cell types tested indicating that the increased FA efflux in response to BSA is not due to increased autophagy (Fig. S1A). We also observed attenuated MAP1LC3B/LC3 (microtubule associated protein 1 light chain 3 beta)-II:I ratios, independent of alterations in signaling (Fig. S1B and C), which may be explained by reduced LD surface area (Figure 5A) needed for MAP1LC3B lipidation/flux [28]. To further explore which nutrient influence FA efflux, we conducted a pulse-chase experiment with [¹⁴C] oleate and removed one or more nutrients in the chase media (Figure 1E). Depletion of glucose accompanied by serum starvation induced FA efflux about 7-fold and further removal of amino acids increased FA efflux 10-fold without altering cell viability (Fig. S2A). These results suggest that nutrients and growth factors in the serum, as well as glucose, are the major factors regulating FA efflux.

Figure 1.

Nutrient deprivation and PNPLA2 promote FA efflux. (A–D) Sequestration of media BODIPY C16 in response to fasting and/or BSA in (A) primary mouse hepatocytes, (B) MEFs, (C) Hep3B and (D) HepG2 cells. (E) Effects of nutrient removal on the efflux of ¹⁴C-oleate from hepatocytes with BSA present in the media. (F–G) Effects of Pnpla2 overexpression (AdPnpla2) on FA efflux with BSA present in the media. (H) Effects of silencing Pnpla2 by the administration of shPnpla2 on BODIPY C16 FA efflux in the presence of BSA under fasting conditions in primary mouse hepatocytes. (I) Inhibition of PNPLA2 using 20 μM ATGListat (Astat) on BODIPY C16 FA efflux in the presence of BSA in primary mouse hepatocytes. All experiments were performed at least three times with n = 3, mean±SEM. Statistical differences among groups were determined using one-way ANOVA followed by Dunnett’s post hoc test in A-E, and I; or a two-way ANOVA followed by Turkey’s post hoc test in F-G; or student t-test in G-H. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001 were compared to control within groups unless specified otherwise. ^####P < 0.001 were compared to the fasted group

Figure 5.

Blocking FA reuptake decreases intracellular TAG levels. (A) Representative images of LipidTOX stained intracellular LDs under fed or fasted media conditions along with either 2% BSA or CB16.2 (10 μM) in mouse hepatocytes. Scale bars: 20 μm. (B) Quantifications of LD area from 6 images in A. (C) Intracellular TAG levels in mouse hepatocytes under either fed or fasted conditions in the presence or absence of 2% BSA were measured and quantified. (D) Effect of transient overexpression of dsRed2-Dgat1 under indicated media conditions in MEF cells. Representative images of BODIPY C12 FA-labeled LDs. Scale bars: 10 μm. (E) Quantification of LD area from 6 images in C. (F) Effect of DGAT1 and DGAT2 inhibitors on BODIPY C16 FA efflux with BSA present in the chase media in mouse hepatocytes. All experiments were performed at least three times with n = 6. Mean±SEM. Statistical differences among groups were determined using two-way ANOVA followed by Turkey’s post hoc test in B, C, E; or a one-way ANOVA followed by the Dunnett’s post hoc test in F. *P < 0.05, **P < 0.01, ****P < 0.001 were compared to the fed group. ^##P < 0.01, ^###P < 0.005, ^####P < 0.001 was compared to BSA negative group or null group

Given the role of PNPLA2 in promoting LD catabolism, in part through lipophagy, we sought to determine the extent to which overexpression of PNPLA2 affects FA efflux. We overexpressed Pnpla2 using an adenovirus (Fig. S2D) and conducted pulse-chase experiments with [¹⁴C]oleate in which BSA was present or absent in the chase media. Adenovirus overexpression of Pnpla2 (AdPnpla2) resulted in a robust increase in media FAs in the presence of BSA, but no differences in media FAs were detected when media was devoid of BSA (Figure 1F). Pnpla2 overexpression also increased the efflux of BODIPY C16 FAs (Figure 1G). In contrast, knocking down Pnpla2 (shPnpla2) robustly decreased media FAs (Figure 1H and Fig. S2E), and chemical inhibition of PNPLA2 negated the fasting-induced efflux of FAs (Figure 1I) consistent with reduced LD catabolism and FA turnover as we have documented previously [10,18]. Neither overexpressing nor knocking down Pnpla2 altered cell viability(Fig. S2B and C). Together, these data show that the presence of ALB at sub-physiological concentrations is sufficient to realize FA efflux in response to Pnpla2 overexpression or nutrient deprivation.

Lipophagy contributes to PNPLA2 and fasting-initiated FA efflux

Since previous work from our laboratory has shown that PNPLA2 acts to induce lipophagy to degrade TAG, we next explored if lipophagy contributes to FA efflux in response to Pnpla2 overexpression. Inhibition of macroautophagy via knockdown of Atg5 (autophagy related 5; siAtg5) (Fig. S2F) or chemical inhibition of PIK3C3/VPS34 (phosphatidylinositol 3-kinase catalytic subunit type 3) attenuated FA efflux in response to Pnpla2 overexpression (Figure 2A,B). Complete ablation of PNPLA2-mediated FA efflux was observed with inhibition of lysosomal function with chloroquine, or genetic (siRNA) or pharmacological (LAListat1) inhibition of LIPA/LAL (lipase A, lysosomal acid type), the major lipase in lysosomes (Figure 2C–E and S2G). In response to fasting, autophagy inhibitors bafilomycin A1 (BafA1), chloroquine (CQ), or VPS34-IN1 blunted FA efflux in mouse hepatocytes (Figure 2F). Pharmacological inhibition of LIPA or knockdown of Atg7 (autophagy related 7) also reduced fasting-induced FA efflux (Figure 2F,G and S2H). Consistent with the role of autophagy/lipophagy in FA efflux, knocking down Rubcn (rubicon autophagy regulator), which leads to a BECN1 (beclin 1)-mediated increase in autophagy [29], significantly increased FA efflux in both fed and fasted conditions (Fig. S3A). Similarly, several well-known small-molecule autophagy inducers, including caffeine, curcumin, and calcium channel blocker and ADRA1A (adrenergic receptor, alpha 1a) antagonist niguldipine significantly induced FA efflux in mouse hepatocytes under fed condition (Fig. S3B). To further test the importance of lipophagy in mediating FA efflux, we conducted FA transfer assays using BODIPY C12-loaded donor cells co-incubated with GFP expressing acceptor cells, as shown in Figure 2H. Both fasting and Pnpla2 overexpression (AdPnpla2) in donor cells resulted in the detection of transferred BODIPY C12 FA in the acceptor cells (Figure 2I,J). Consistent with an important role of lipophagy in FA efflux, Lipa knockdown (siLipa) abrogated fasting-mediated FA transfer (Figure 2K,L). Together, these results indicate that lipophagy contributes to FA efflux in response to PNPLA2 or nutrient deprivation.

Figure 2.

Lipophagy contributes to PNPLA2 and fasting-initiated FA efflux. (A–E) Media chase ¹⁴C-oleate in response to Pnpla2 overexpression (AdPnpla2) and (A) knockdown of Atg5, (B) inhibition of PIK3C3 (VPS34-IN1, 5 μM) (C) chloroquine (CQ, 5 μM), (D) knockdown of Lipa, or (E) inhibition of LIPA using LAListat1 (10 μM) in mouse hepatocytes. (F) Media BODIPY C16 FA in cells cultured in fed or fasted conditions along with bafilomycin A1 (BafA1, 100 nM) or the above inhibitors. (G) Effect of knockdown of Atg7 on BODIPY C16 FA efflux in primary hepatocytes. The effluxed FAs were measured in the chase media containing 2% FA-free BSA in A-G. (H) Experimental design of FA transfer assay for I-L. (I) FA transfer assay measuring BODIPY C12 FA transfer from donor MEF cells treated with adenoviruses to the GFP-labeled receptor cells. Scale bars: 10 μm. (J) Quantification of the BODIPY-labeled LD in receptor cells from 6 images for I. (K) Effects of Lipa knockdown on fasting-mediated FA transfer in MEF cells. Scale bars: 10 μm. (L) Quantification of the BODIPY-labeled LD in receptor cells from 6 images for K. The receptor cells are outlined. All experiments were repeated at least three times with n = 3, mean±SEM. Statistical differences among groups were determined using two-way ANOVA followed by the Turkey’s post hoc test in A-E, G, J, L; or a one-way ANOVA followed by the Dunnett’s post hoc test in F. *P < 0.05, **P < 0.01, ***P < 0.005, ****P < 0.001 were compared to control groups (AdCtrl in A-E; Fed in F, G, J, and L). ^#P < 0.05, ^##P < 0.01, ^####P < 0.001 were compared to siCtrl in A, E, G, and L, compared to Vehicle in B-D, compared to Fasted in F, compared to AdCtrl in J

FA efflux occurs through lysosomal fusion with the plasma membrane

It is well-accepted that lysosomal degradation of lipids produces FAs used for energy production or other cellular needs. However, the mechanisms through which FAs exit the lysosome have not been identified. It is well-documented that lysosomes can fuse with the plasma membrane and efflux their contents, including various lysosomal proteins [30]. This fusion process appears to be dependent upon the lysosomal Ca²⁺ channel protein MCOLN1 [31–33]. Thus, we employed the MCOLN1 antagonist vacuolin-1 to block lysosomal fusion to the plasma membrane [34]. Immunofluorescence staining of luminal epitope of LAMP1 (lysosomal associated membrane protein 1), specifically developed to measure lysosomal exocytosis [35], showed that fasting-induced the abundance of LAMP1 at the cell surface, whereas vacuolin-1 attenuated fasting-induced cell surface LAMP1 accumulation demonstrating the efficacy of vacuolin-1 on preventing lysosomal fusion to the plasma membrane (Fig. S4A-C).

Similarly, fasting increased the protein levels of the lysosomal protease CTSB (cathepsin B) in the media, but this effect was blocked by vacuolin-1 treatment, further validating the importance of MCOLN1 on lysosomal exocytosis (Fig. S4D). Next, we examined whether blocking lysosomal fusion with the plasma membrane affects FA efflux. The presence of vacuolin-1 during the chase period abrogated the increases in media FAs under either fasting conditions or following Pnpla2 overexpression (Figure 3A,B). In addition, silencing Mcoln1 using shRNA (Fig. S2I) modestly reduced FA efflux under fed conditions while abolishing the fasting-induced FA efflux in mouse hepatocytes (Figure 3C). Consistent with reduced efflux, vacuolin-1 also prevented the transfer of BODIPY C12 FA from donor cells to GFP-labeled acceptor cells (Figure 3D,E). To further elucidate its role in FA efflux, we transiently overexpressed MCOLN1 in mouse hepatocytes and examined the media FA levels in either fed or fasted conditions during the chase period. Overexpression of MCOLN1 significantly increased media FA in fed conditions and induced FA efflux even more under fasting conditions (Figure 3F). Thus, these results show that that MCOLN1 is the key lysosomal protein that mediates FA release through lysosomal fusion with the plasma membrane.

Figure 3.

FA efflux occurs through lysosomal fusion to the plasma membrane. (A) Effect of vacuolin-1 (1 μM) on the efflux of BODIPY C16 FA under fed or fasted conditions in mouse hepatocytes. (B) Media chase ¹⁴C-oleate in response to Pnpla2 overexpression and vacuolin-1 treatment in mouse hepatocytes. (C) Effect of knocking down Mcoln1 (shMcoln1) on the efflux of BODIPY C16 FA under fed or fasted conditions in AML12 cells. The effluxed FAs were measured in the chase media containing 2% FA-free BSA in A–C. (D) Effects of vacuolin-1 on fasting-mediated FA transfer. The receptor cells are outlined. Scale bars: 10 μm. (E) Quantification of the BODIPY-labeled LD in receptor cells from 6 images for D. (F) Effect of transient overexpression of MCOLN1 (TRPML1-YFP) on BODIPY C16 FA efflux in mouse hepatocytes. (G) Workflow for measurement of intracellular FFA using biosensor ADIFAB. (H) Fold-change of intracellular FFA level measured by ADIFAB in AML12 cells. (I) The workflow of lysosome isolation and measurement of lysosomal FFA using ADIFAB. (J) ADIFAB measured the lysosomal FFA level in AML12 cells. (K) Effluxed FFA from in situ liver perfusate (n = 6). All experiments were repeated at least three times with n = 4 unless specified otherwise. Mean±SEM. Statistical differences among groups were determined using two-way ANOVA followed by Turkey’s post hoc test in A-C, F, J, K; or a one-way ANOVA followed by the Dunnett’s post hoc test in E, H. **P < 0.01, ***P < 0.005, ****P < 0.001 were compared to fed or control group. ^#P < 0.05, ^##P < 0.01, ^####P < 0.001 were compared to vehicle (A-B), or shScr (C), or Null (F), or Fasted (H, and J)

While the above studies suggest that FA undergo efflux following lysosomal exocytosis, we wanted to explore further if, alternatively, FAs could be directly transported out of lysosomes into the cytosol. To overcome the technical difficulties of imaging the fluorescence-labeled FA inside of lysosomes due to low pH-caused fluorescence quenching, we employed biosensor assays to measure intracellular or lysosomal FA levels quantitatively. The FA biosensor ADIFAB is a small FA-binding fluorescent protein that can be excited at wavelength 385, but the emission wavelength of FA-bound ADIFAB is 505, whereas the emission wavelength of unbound ADIFAB is 432. The free FA level can be calculated by the conversion of the signal of FA-bound ADIFAB and unbound ADIFAB [36,37]. We first measured the intracellular FA level using AML12 cells in which we electroporated the ADIFAB protein (Figure 3G). Fasting significantly increased intracellular FA levels, and the presence of BSA, which sequesters FAs and attenuates reuptake, abolished fasting-induced intracellular FA accumulation (Figure 3H). Treatment of vacuolin-1 also abolished the fasting-mediated accumulation of intracellular FAs, which suggests that inhibition of MCOLN1 traps FAs inside lysosomes. To more definitively measure the lysosomal release of FAs, we isolated lysosomes from cells with stable expression of the HA-tagged lysosomal protein TMEM192 (transmembrane protein 192) [38] followed by the incubation with lysosome permeabilization agent Leu-Leu methyl ester hydrobromide (LLME) to release lysosomal FAs as measured with ADIFAB (Figure 3I). FA levels were not detectable in incubations of lysosomes derived from either fed or fasted cells suggesting that lysosomes do not directly export FAs in the absence of fusion to the cell membrane. However, lysosomes isolated from fasted cells treated with vacuolin-1 did release FAs, which may be due to damage of swollen lysosomes following isolation procedures. Importantly, incubating lysosomes with the lysosome permeabilization agent, LLME triggered a robust increase in FA release from lysosomes, especially from those isolated from cells treated with vacuolin-1 (Figure 3J). These findings are consistent with FA accumulation in lysosomes if fusion and efflux at the plasma membrane do not occur. To extend our findings to an in vivo model, we characterized FA efflux using in situ perfused livers from mice fed with either a control chow diet or a high-fat diet (HFD) for 12 weeks. Livers were perfused sequentially with basal media alone or that containing BSA, CB16.2, a FATP2 inhibitor that attenuates FA uptake, or CB16.2 co-treatment with vacuolin-1. Outflow liver perfusates from mice fed the control diet had increased FAs during the perfusion of BSA and CB16.2, but the addition of vacuolin-1 negated the increased hepatic FA efflux (Figure 3K), similar to what was observed in cell models. In contrast, the mice fed the HFD had attenuated efflux consistent with reduced autophagy in this model [4,39].

Previously, we performed lipidomic analysis (unpublished) on liver samples isolated from mice administered control or Pnpla2 shRNA (shPnpla2) adenovirus. One of the more significant changes observed was higher levels of several sphingomyelin (SM) species in samples from mice treated with Pnpla2 shRNA (shPnpla2) (Figure 4A and B). The relevance of these findings is that SMs are known to be negative regulators of MCOLN1 [26]. Based on this data, we treated cells with SM(d18:1/16:0), the species most affected by Pnpla2 knockdown, and observed that it prevented PNPLA2-mediated FA efflux (Figure 4C). Moreover, SM(d18:1/16:0) also reduced fasting-mediated FA efflux in mouse hepatocytes (Figure 4D). To test the importance of MCOLN1 in mediating this response, we cultured control and shMcoln1 cells with SM(d18:1/16:0) under fasting conditions and measured FA efflux. Knockdown Mcoln1 or administration of SM(d18:1/16:0) independently reduced fasting-induced FA efflux, yet the combination of SM(d18:1/16:0) and absence of MCOLN1 did not lower FA efflux further than either treatment alone (Figure 4E). Together, these studies identify a previously uncharacterized mechanism through which SMs influences hepatic FA trafficking.

Figure 4.

Sphingomyelins regulate FA efflux. (A and B) The lipidomic analysis showed that relative abundances of SM species are upregulated with knockdown Pnpla2 (shPnpla2) in mouse livers (n = 8). (C) SM(d18:1/16:0) reduced Pnpla2 overexpression-mediated FA efflux in mouse hepatocytes. (D) SM(d181/16:0) decreased fasting-induced FA efflux in mouse hepatocytes. (E) SM(d181/16:0) failed to further decrease the reduced FA efflux in AML12 cells treated with shMcoln1 to knockdown Mcoln1. The effluxed FAs were measured in the chase media containing 2% FA-free BSA in C-E. FA efflux assay was repeated at least three times with n = 4. Data presented as mean±SEM, n = 4. Statistical differences among groups were determined using two-way ANOVA followed by Turkey’s post hoc test in B, C, E; or a one-way ANOVA followed by Dunnett’s post hoc test in D. *P < 0.05, ****P < 0.001 were compared to shScr or Vehicle or Fed. ^####P < 0.001 were compared to AdCtrl or Fasted or shMcoln1.

Blocking FA reuptake decreases intracellular LDs

FAs liberated from TAG stored in LDs are thought to be used in numerous downstream metabolic pathways, including re-esterification to TAG for storage in LDs. Our findings indicate that hydrolyzed FAs are largely effluxed, which suggests that FAs would have to undergo reuptake before further metabolism. To further interrogate this pathway, we quantified LDs in response to manipulations of FA efflux. We utilized confocal imaging to analyze the intracellular LDs using a neutral lipid dye LipidTOX in mouse primary hepatocytes that were cultured in the fed or fasted media in the presence or absence of BSA or CB16.2 for 4 h before imaging. As shown in Figure 5A,B, LD area did not significantly change when cells were fasted in the absence of BSA. However, LDs were significantly decreased in fasting condition when BSA was present in the media to sequester effluxed FA or when CB16.2 was present to inhibit the reuptake of FFA. To precisely measure the intracellular TAG level, we measured TAG using a colorimetric kit. As shown in Figure 5C, BSA significantly blunted the fasting-induced intracellular TAG level. Re-esterification of hydrolyzed FAs back to TAG is a common pathway in numerous cell types and acts to prevent the accumulation of fatty acids and the associated lipotoxicity. The DGATs (diacylglycerol O-acyltransferase enzyme) are responsible for the terminal step in TAG synthesis, the acylation of DAG to form TAG. DGAT1 is responsible for the incorporation of exogenous FAs into TAG or FA re-esterification following lipolysis, whereas DGAT2 is thought to play a larger role in the esterification of de novo synthesized FAs [40,41]. To further explore the extent to which FA efflux and reuptake influence fasting-induced TAG synthesis, we transiently transfected MEFs with a dsRed2-Dgat1 plasmid and subsequently exposed them to fasting conditions with or without BSA. DGAT1 increased the area of LDs under fasting conditions as expected, but the presence of BSA ameliorated the LD accumulation in both non-transfected and Dgat1-transfected cells (Figure 5D,E). Consistent with these data, DGAT inhibitors did not alter fasting-induced FA efflux (Figure 5F). These data suggest that lipophagy-derived FAs undergo a cycle of efflux followed by reuptake before their subsequent reincorporation into cellular LDs.

FA efflux precedes reuptake and channeling to oxidation pathways

During nutrient deprivation, autophagy/lipophagy occurs as a means to provide cells with substrates for oxidation to meet energy needs. In particular, most cell types turn to FA oxidation under such conditions. Thus, we tested the importance of FA efflux in regulating the oxidation of endogenous FAs. As we have reported previously [10], Pnpla2 overexpression increased the channeling of LD-derived FAs to oxidative pathways (Figure 6A). However, the presence of BSA prevented the increase in FA oxidation consistent with its role in sequestering effluxed FAs. Furthermore, the addition of LAListat1 to inhibit lipophagy or addition of vacuolin-1 to block lysosomal exocytosis also attenuated FA oxidation in both control cells and those with Pnpla2 overexpression (Figure 6B).

Figure 6.

Fasting and PNPLA2-mediated FA efflux require reuptake for channeling toward oxidative pathways. (A–B) Fold change of chase ASM level under either overexpression of Pnpla2 along with vacuolin-1 or LAListat1 treatment in mouse hepatocytes. (A) Effect of overexpression of Pnpla2 (AdPnpla2) on ASM. (B) Effect of vacuolin-1 or LAListat1 on ASM while overexpression of Pnpla2 (AdPnpla2). (C) Fold change of chase ASM under fasting conditions in MEF cells. (D) Representative images showing the co-localization of BODIPY C12 FA-labeled LDs with mitochondria. Scale bars: 10 μm. (E) Quantification for D, co-localization was quantified from 6 images using Mander’s overlap analysis. (F) BHBA level was determined from in situ liver perfusates. All experiments were performed at least 3 times with n = 4, mean±SEM. Statistical differences among groups were determined using two-way ANOVA followed by Turkey’s post hoc test in A–C, F; or a one-way ANOVA followed by the Dunnett’s post hoc test in E. *P < 0.05, ***P < 0.005, ****P < 0.001 were compared to AdCtrl or Fed. ^#P < 0.05, ^##P < 0.01, ^###P < 0.005, ^####P < 0.001 were compared to treatment control or fasted group

Similarly, the presence of BSA reduced the fasting-mediated increase of FA oxidation, and the addition of 500 μM oleate in chase media failed to normalize the reduced oxidation of endogenous FAs in response to BSA under fasting conditions (Figure 6C). Previous studies have used BODIPY C12 FA colocalization with mitochondria as a means to measure FA transfer from LDs to the mitochondria [42]. As expected and consistent with the FA oxidation data, fasting promoted a significant increase in BODIPY C12 FA that colocalized with mitochondria (Figure 6D,E). However, the addition of BSA during the fasting period prevented BODIPY C12 FA trafficking to mitochondria. In the same in situ liver perfusion assay,described in Figure 3, beta-hydroxybutyrate (BHBA) was measured in collected perfusates. In the chow-fed group, sequestering FFA with BSA reduced ketogenesis, and addition of CB16.2, which blocks FA reuptake, and vacuolin-1 further attenuated ketogenesis (Figure 6F). Consistent with overall reduced FA efflux in the HFD-fed mice (Figure 3K), this group also showed reduced FA oxidation compared to the control fed mice. Together, our findings suggest that lipophagy-mediated FA efflux and their subsequent reuptake contributes to fasting and PNPLA2-induced FA oxidation and TAG synthesis (Figure 7).

Figure 7.

Working model. Nutrient deprivation and PNPLA2 activation drive lipophagy and FA generation in lysosomes. FFA are effluxed extracellularly through lysosome fusion with the plasma membrane, which is regulated by the lysosome calcium channel protein MCOLN1. Subsequently, effluxed FA undergo reuptake and then either channeled to mitochondria for oxidation or are re-esterified to TAG and stored in LDs

Discussion

The past decade has seen significant advances in our knowledge of autophagy and its pathophysiological role in multiple disorders. Despite a long-known linkage between LIPA deficiency and steatosis in humans [43], lipophagy has only recently been shown to contribute directly to hepatic lipid catabolism [4]. Since these initial findings, a rapidly growing body of literature shows that a host of genetic perturbations and small-molecules regulate hepatic LD accumulation through alterations in lipophagy [44]. However, the mechanistic details through which autophagic machinery recognizes LDs and how lysosomal lipid degradation affects cellular metabolism and signaling have not been extensively studied. Lysosomal-mediated degradation of cargo yields glucose, amino acids, and FAs for subsequent use in cellular metabolism. While the transport of glucose and amino acids has been characterized [45], the mechanism through which FAs leave the lysosome is not well understood. Groener et al. suggested that the acyl chains of cholesterol esters contained in low-density lipoproteins, which undergo endocytosis and require lysosomal lipolysis for their catabolism, undergo extracellular efflux before the subsequent incorporation into cellular TAG in fibroblasts [46]. These data strongly support our findings that FAs derived from intracellular TAG degradation (i.e., lipophagy), which was observed in numerous cell types, also undergo efflux rather than direct transport into the cytosol.

It is well-established that LDs interact with numerous organelles, including mitochondria [47–49], the latter of which is speculated to facilitate the transfer of FAs for their subsequent oxidation [49,50]. Indeed, Rambold et al. demonstrate that, in MEFs pulsed with BODIPY FA, 24 h of nutrient starvation triggers the appearance of BODIPY FA in mitochondria suggesting the transfer of BODIPY FA from LDs to the mitochondria [42]. Moreover, the authors note that FA efflux, as measured by donor/acceptor cell transfer experiments, increases in response to mitochondrial fission, which is presumed to inhibit FA oxidation. However, several points must be made. First, BSA was not added during any of the experiments; thus, the amount of FA that can be effluxed is limited by solubility. Second, while BODIPY FA appearance in the mitochondria in response to fasting shows that it originated from LDs, it does not show that the BODIPY FA is transferred directly from LD to adjacent mitochondria (i.e., efflux or indirect trafficking was not considered). Along these lines, inhibition of FA oxidation also inhibits FA uptake [51]; thus, it is difficult to ascertain if the increased FA efflux/transfer in response to reduced oxidation is due to increased efflux or decreased reuptake. While we cannot rule out some contribution of intracellular FA flux to the mitochondria directly from LDs, as widely speculated, our data support a model where a significant amount of FAs stored in LDs undergo lipophagy-mediated efflux before the subsequent oxidation or entrance into other metabolic pathways (Figure 6). Consistent with our findings, previous studies in hepatocytes show the abundant transfer of FAs between cells, with cells having more LD accumulation transferring FAs to neighbor cells with fewer lipid stores [52].

Similarly, DGAT1 inhibition prevents re-esterification of lipophagy-derived FAs but increases their oxidation providing evidence against the direct transfer of LD to mitochondria [41]. Finally, it has recently been reported that mitochondria-associated with LDs preferentially burn carbohydrates to fuel lipid synthesis at LD-mitochondrial contacts [53]. In contrast, mitochondria not in contact with LDs have a higher oxidative capacity. Thus, these studies further challenge the dogma that LD directly transfer FAs to mitochondria for oxidative purposes. Much remains unraveled regarding the regulation and role of LD-mitochondria contact sites, and we must be careful not to define function simply by association.

Why do lysosomes export FAs from the cell? Intuitively, it would seem logical and most efficient that lysosome-derived FA is trafficked directly to mitochondria for beta-oxidation or ER/LDs for re-esterification. However, intracellular FAs are highly toxic and are a defining characteristic of lipotoxicity. Indeed, the re-esterification of lipophagy-derived FAs by DGAT1 is critical to prevent excess oxidation and acyl-carnitine-induced mitochondrial dysfunction [41]. In agreement with these studies, Herms et al. suggested that the heterogeneity of LDs and FA transfer amongst hepatocytes occurs as a mechanism to prevent lipotoxicity [52]. Thus, it is plausible that the efflux of FAs allows cells to gauge reuptake based on metabolic needs/capacity and, thus, acts as a protective mechanism to reduce lipotoxicity or as a means to redistribute energy to adjacent cells. Thus, when the capacity to handle the high rates of FA influx, derived from adipose, that occurs during fasting becomes limiting, slowing the uptake of exogenous FAs and promoting efflux of lipophagy-derived FAs may help alleviate the FA burden.

The liver is highly zonated with diverse gene and protein expression profiles between periportal and perivenous hepatocytes [54,55]. While zonation of autophagy in the liver during fasting periods has not been characterized, it is plausible that FA efflux could occur in a subset of hepatocytes to redistribute FA among the liver or, if occurring centrally, provides a mechanism for the liver to lessen its FA load. The liver is well-established to be an organ responsible for the disposal of circulating fatty acids with rates of uptake dictated by the concentration of fatty acids in the blood [56]. It remains unclear if FAs effluxed from the liver readily equilibrate with exogenous FAs circulating through the liver or if most FAs remain within the liver and are further metabolized by reuptake into the same or neighboring cells in which they were effluxed. Nonetheless, given the highly fenestrated nature of the liver, it would seem possible that a proportion of the FAs make it to the bloodstream, as we observed in studies utilizing perfused livers. While we can only speculate, it would seem that the quantity of FAs effluxed from the liver that reaches the bloodstream would be a minor contributor to circulating FA levels during fasting when FA release from adipose tissue is high. It should also be noted that efflux was observed in non-hepatocyte cells (MEFs), suggesting that FA efflux and subsequent local or distal reuptake may contribute to normal lipid handling across numerous cell types and tissues. A focus of future studies will likely address the significance of FA efflux in tissue homeostasis and disease development.

MCOLN1 is a critical factor in regulating lysosomal fusion to the plasma membrane [57]. Lysosomal exocytosis plays an important role in the export of Fe²⁺ and Zn²⁺ and lysosomal enzymes, antigen presentation, transmitter release, and membrane repair [30]. Mutations in MCOLN1 that reduce or ablate catalytic activity result in mucolipidosis type IV storage disease, which is characterized by the accumulation of gangliosides in lysosomes [58]. While MCOLN1 has been suggested to play a key role in the lysosomal efflux of lipid contents [26], its role in lipophagy and FA trafficking has not been reported. Results from the current study reveal a critical role for MCOLN1 in mediating FA efflux consistent with its known role in regulating lysosomal exocytosis. However, MCOLN1 is also known to influence endosome trafficking and autophagosome/lysosome fusion; thus, the contribution of these to the observed effects on FA efflux is currently unknown. We also found sphingomyelins, which are known to negatively regulate MCOLN1 [26,59], increased in response to PNPLA2 inhibition. Moreover, PNPLA2 was unable to promote efflux in the presence of exogenous sphingomyelin, and sphingomyelin was unable to further suppress FA efflux in cells lacking MCOLN1. Together, these data suggesting that alterations in sphingomyelin levels may contribute to MCOLN1-mediated lysosomal FA efflux.

In summary, this study highlights the importance of FA efflux as an intermediate step in the catabolism of lipids in numerous cell types. These data also show the importance of MCOLN1 and lysosomal fusion to the plasma membrane as important events that are critical for the downstream trafficking of FAs in response to nutrient deprivation or PNPLA2 overexpression (Figure 7). Overall, these data provide new insights into the regulation of LD catabolism and FA trafficking and warrant future studies further defining the contribution of lysosomal-mediated FA efflux to the etiology of diseases characterized by alterations in lipid metabolism.

Materials and methods

Chemicals, reagents, and plasmids

The PNPLA2 inhibitor ATGListatin was purchased from Xcess Biosciences (MC0150-2s) and used in cells at 20 µM in DMSO for 4 h (for imaging studies) and up to 8 h for pulse-chase experiments. Chloroquine was purchased from Sigma-Aldrich (c6628) and vacuolin-1 from Cayman Chemicals (20,425). LAListat 1 was purchased from Tocris (6098). CB16.2 was purchased from D&C Chemicals (DC11011). All autophagy inhibitors were used at the specified concentrations for the duration of the chase. Pnpla2 overexpression (AdPnpla2) and shRNA knockdown (shPnpla2) adenoviruses, as well as the corresponding control (AdCtrl; empty adenovirus) or scrambled shRNA, were provided by Dr. Andrew Greenberg (Jean Mayer USDA Human Nutrition Research Center on Aging, Tufts University) [60,61]. Atg5 siRNA (siAtg5) was custom synthesized from Qiagen. Atg7 siRNA (siAtg7) was purchased from Dharmacon (ON-TARGETplus Mouse Atg7 siRNA-SMARTpool, 74244). Lipa siRNA (siLipa) was procured from Sigma-Aldrich (EMU075891). Dgat1 was inserted into dsRed2-N1 vector (Addgene, 54493; gift from Michael Davidson) using EcoRI and KpnI sites to construct the dsRed2-Dgat1 plasmid. MCOLN1-YFP was a gift from Craig Montell (Addgene, 18826). pMRX-IP-GFP-LC3-RFP-LC3ΔG was a gift from Noboru Mizushima (Addgene, 84572). LAMP1-YFP was a gift from Walther Mothes (Addgene, 1816). Rubcn-ASO and control-ASO were provided by Ionis Pharmaceuticals. All vectors were administered using Invitrogen^TM Lipofectamine 2000 Transfection Reagent (Thermo Fisher Scientific, 11668019).

Animals and diet

All animal protocols were approved by the University of Minnesota Institutional Animal Care and Use Committee. Eight- to 10-week-old C57Bl/6J male mice were housed and acclimatized as previously described [10]. All mice had free access to water and were fed with either a purified control diet (Harlan Teklad Premier Laboratory, TD 94045) or a 45% high-fat diet (Harlan Teklad Premier Laboratory, TD09404).

Cell culture, adenoviral infection, and radiolabeling

Hepatocytes were isolated as described previously [10] and were transduced with adenoviruses 4 h after plating. Experiments measuring lipid incorporation (pulse) to measure TAG turnover or media FA efflux were performed in M199 media (Thermo Fisher Scientific, 31100) containing 26 mM sodium bicarbonate, 23 mM HEPES (Fisher Scientific, BP310-500), 50 IU/mL penicillin and 50 µg/mL streptomycin (Thermo Fisher Scientific, 15140122), 100 nM dexamethasone (Sigma, D8893), 11 mM glucose, 20 mM L-carnitine (Sigma, C0283), 100 nM insulin (Sigma, I1882). The following morning, cells were pulsed with 500 µM oleate plus [14]Coleate (Perkin Elmer, NEC317250UC), which was conjugated to 2.1 mM FA-free BSA (Gemini, 100-107P) at 37°C and added to the above media for 2 h. Some cells were harvested at the end of the pulse period to measure radiolabel incorporation into cellular lipid fractions. The remaining cells were washed with phosphate-buffered saline (PBS; Thermo Fisher Scientific, 10010023) and replaced with fresh complete M199 medium (as described above) lacking insulin for an additional 6–8 h (chase period) followed by the collection of media and cells for lipid extraction. FAs oxidized or effluxed during the chase period are expressed as a percentage of the pulse [¹⁴C]TAG. All experiments to determine autophagy induction in hepatocytes were performed in complete M199 without hormones unless specified otherwise. Media FAs were extracted using Dole reagent (isopropanol:heptane:1 M sulfuric acid, 40:10:1), hexane and water in a 5:3:2 ratio [62]. Lipid samples were separated into different fractions by thin-layer chromatography (TLC) and analyzed, as described previously [63].

FA transfer assay

MEF cells (ATCC, SCRC-1040) were cultured under normal conditions and sub-grouped as donor cells and receptor cells. The donor cells were either transduced with Pnpla2 adenovirus (AdPnpla2, 3.2 × 10⁷ viral particle/mL) for 36 h in fasting medium or transfected with Lipa siRNA (siLipa) for 48 h in the complete medium as indicated. The receptor cells were transduced with GFP adenovirus (AdGFP, 1 × 10⁸ virus particle/mL) for 48 h. Subsequently, the donor cells were lipid-loaded with 500 µM oleate and 5 µM C12 BODIPY FA 558/568 (Thermo Fisher, D3835) that conjugated to 2.1 mM FA-free BSA at the last 16 h of adenovirus transduction or siRNA transfection period. The donor cells and receptor cells were then trypsinized and combined with a 1:1 ratio in the following morning and allowed to attach on coverslips in 12-well plate for 4 h. The combined of donor cells and receptor cells were cultured under indicated conditions for 4 h before further processing for confocal imaging.

ADIFAB transfection and FFA binding assay

AML12 cells (ATCC, CRL-2254) were cultured in either complete DMEM/F12 (1:1) media (Thermo Fisher, 11320082) (Fed), or HBSS buffer (Fasted; 137 mM NaCl, 5.4 mM KCl, 250 µM Na₂HPO₄, 440 µM KH₂PO₄, 1.3 mM CaCl₂, 1 mM MgSO₄, 4.2 mM NaHCO₃) in the presence or absence of 2% BSA (FA free, w/v) or 1 µM vacuolin-1 for 4 h. Subsequently, cells were collected in a microcentrifuge tube and transfected with small FA binding protein ADIFAB (FFA Sciences, ADI-1-200) via electroporation following kit instructions (Neon™ Transfection System 100 µL Kit; ThermoFisher, MPK10025). Electroporated cells were immediately transferred to previously indicated cell culture media then the fluorescence intensity of ADIFAB (Excitation at 385, emission of FA-bound ADIFAB is 505, emission of unbound ADIFAB is 432) were measured. The fluorescence intensity of non-electroporated cells under the same culture conditions was also measured as a control.

Isolation of lysosomes and detection of lysosomal FFA

AML12 cells that stably express TMEM192 were transduced with lentivirus, followed by puromycin (Thermo Fisher Scientific, A1113803) selection [38]. The plasmid Tmem192-3xHA was a kind gift from Dr. Sabatini and acquired from Addgene (102930) and the lentivirus was constructed in the Viral Vector Core at the University of Minnesota. Cells (4 × 10⁶) were cultured in 10 cm plate under either complete DMEM/F12 (1:1) media (Fed) or HBSS buffer (Fasted) in the presence or absence of 1 µM vacuolin-1 for 4 h. Subsequently, cells were harvested, followed by lysosome isolation [38]. Anti-HA magnetic beads (Thermo Fisher, 88836) immuno-captured lysosomes were then incubated with 3 mM of Leu-Leu-O-methyl ester (LLME; Sigma, L7393), which permeabilizes lysosome membranes [64], for 1 h to release lysosomal metabolites including FFA. Then FFA level was measured and quantified using ADIFAB binding assay, as described above.

Mouse liver perfusion

Livers were perfused at a flow rate of 8 mL/min as performed in similar experiments with Krebs-Henseleit Buffer as described previously [65]. BSA (1%) was added to the perfusion buffer after 10 min of perfusion. In addition to 1% BSA, 10 µM CB16.2 was added after 20 min of perfusion. During 30–60 min of perfusion, mice received 1 µM vacuolin-1 along with the BSA and CB16.2. Samples were collected during the last 2 min of each perfusion period. FAs were extracted as described above and measured via a colorimetric kit from Wako Diagnostics (995–34,791). Beta-hydroxybutyrate was analyzed via a colorimetric kit from Wako Diagnostics (415–73,301).

Triacylglycerol assay

TAG levels were measured following the manufacturer’s instructions (Stanbio, SB2100-430). In general, mouse hepatocytes were cultured in the indicated media conditions for 4 h. Cells were then harvested in 1 mL of distilled water, followed by homogenization. A portion of the cell lysates (25 μL) was removed for subsequent protein determination and the remaining cell lysates were subjected to lipid extraction with chloroform:methanol (2:1). Extracted and nitrogen dried lipids were resuspended in 100 μL of isopropanol with 1% Triton X-100 (Sigma, X100-500mL). Samples were vortexed and left at room temperature for 1 h, then 20 μL of the samples were assayed.

Western blotting

Protein extraction and immunoblotting were performed, as described previously [66]. LIPA antibody was obtained from Abcam (ab36597). ATG5 (D5F5U; 12994S), ATG7 antibody (8558), PNPLA2/ATGL (2439S), RPS6KB1 (9202), p-RPS6KB1 (9205), ACTB/beta actin (4970) antibodies were purchased from Cell Signaling. MCOLN1 antibody was purchased from Novus Biologicals (NB110-82375). MAP1LC3B antibody (M1867) was purchased from MBL. CTSB antibody was purchased from R&D System (AF965).

Confocal imaging

For immunofluorescence microscopy, cells were grown on coverslips and fixed with 4% paraformaldehyde for 10 min and blocked with 5% BSA in PBS (with 1 mM calcium and 1 mM magnesium). Incubation with primary (overnight at 4°C) and secondary (1 h at RT) antibodies were done in PBS (with 1 mM calcium and 1 mM magnesium) containing 1% BSA. The human CD107a (luminal LAMP1) antibody (eBio1D4B) was purchased from Invitrogen (14-1079-80). The CDH1/E-cadherin antibody was purchased from Novus Biologicals (AF648). The secondary antibody anti-mouse conjugated to Alexa Fluor 555 was purchased from Cell Signaling (4409). The secondary antibody donkey anti-goat conjugated to Alexa Fluro 647 was purchased from Abcam (ab150131). For LD staining, cells were incubated with LipidTOX™ Green Neutral Lipid Stain (1 μM; Invitrogen, H34475) for 60 min at room temperature after fixation. For mitochondria staining, cells were incubated with 20 nM MitoTracker Green FM (Invitrogen, M22426) for 60 min at 37°C before live-cell imaging. Nuclei were stained with DAPI (4′, 6-diamidino-2-phenylindole) (Cell Signaling, 4083S) for 10 min followed by mounting onto slides for visualization. To evaluate the specificity of the CD107 antibody against luminal LAMP1, the cells were fixed then divided into two groups. One group of cells was blocked with 5% BSA in PBS (with 1 mM calcium and 1 mM magnesium) while the other group of cells was blocked with 5% BSA in PBS (with 1 mM calcium and 1 mM magnesium) with 0.2% Triton X-100 for permeabilization. The difference for imaging cell surface LAMP1 is showing in Fig. S4C. Images were acquired with Nikon A1Rsi Confocal with SIM Super Resolution microscope with either 60x oil immersion objective for fixed cell imaging or 60x water immersion objective for live-cell imaging, and Zeiss spinning disk confocal microscope with 63x water immersion objective for live-cell imaging. Images were analyzed using ImageJ (NIH) and Cell Profiler [67]. Images from 6 different fields per well were captured, and experiments were performed in triplicate.

Lipidomic analysis

Lipid extraction and liquid chromatography-mass spectrometry (LC-MS) lipidomic analysis were conducted based on previous methods [68]. Briefly, lipids extracted from 50 mg liver samples were reconstituted in 0.5 ml n-butanol (Sigma, 281549). A 5 µL aliquot of liver lipid sample was injected into an Acquity^TM UPLC system (Waters, Milford, MA) and separated by a gradient of mobile phase ranging from water to 95% aqueous acetonitrile containing 0.1% formic acid over a 10-min run. LC eluents were introduced into a Xevo-G2-S mass spectrometer (Waters) for accurate mass measurement and ion counting. Capillary voltage and cone voltage for electrospray ionization were maintained at 3 kV and 30 V for positive-mode detection, respectively. Source temperature and desolvation temperature were set at 120°C and 350°C, respectively. Nitrogen was used as both cone gas (50 L/h) and desolvation gas (600 L/h), and argon as the collision gas. For accurate mass measurement, the mass spectrometer was calibrated with sodium formate solution (range m/z 50–1000) and monitored by the intermittent injection of the lock mass leucine enkephalin ([M + H]⁺ = 556.2771 m/z) in real-time. Mass chromatograms and mass spectral data were acquired and processed by MassLynx^TM software (Waters) in centroided format. Additional structural information was obtained tandem MS fragmentation with collision energies ranging from 15 to 30 eV, and database search (Lipid Maps: http://www.lipidmaps.org/).

Statistical analyses

Statistical differences among experimental groups were determined using either a one-way ANOVA followed by Dunnett’s post hoc test or a two-way ANOVA followed by Turkey’s post hoc test. Moreover, statistical differences between the two experimental groups were determined using the Student’s t-test. All data are presented as Means ± SEM and statistical significance was declared at p < 0.05.

Funding Statement

This work was supported by grants from the National Institutes of Health (R01AG055452, R01DK108790, R01DK114401) and the American Diabetes Association (1-16-IBS-203) to DGM.

Disclosure statement

No potential conflict of interest was reported by the authors.

Supplementary material

Supplemental data for this article can be accessed here.