Lysosomal Acid Lipase Drives Adipocyte Cholesterol Homeostasis and Modulates Lipid Storage in Obesity, Independent of Autophagy

Lysosomal acid lipase (LAL) is the sole nonpolar lipid esterase within lysosomes. Because of human gene mutations, partial inactivation of LAL (as in cholesterol storage disease) or total absence (as in Wolman syndrome) promotes ectopic lipid accumulation, hyperlipemia, inflammation, and multiorgan (particularly liver) failure (1). According to substrate specificity and pH spectrum, LAL is involved in the use of exogenous lipids that enter through receptor-mediated endocytic pathways. As such, LAL regulates macrophage inflammation downstream of lipoprotein endocytosis and efferocytic clearance of apoptotic bodies (2). It is also part of the gene program driving the metabolic switch from a glucose-using proinflammatory M1-like macrophage toward an M2-like lipid-oxidizing phenotype that promotes tissue remodeling (3). More globally, genome-wide association study analysis of gene variation identified LAL as a susceptibility locus for cardiovascular disease (4), suggesting a role in metabolic regulation.

In the context of diabetes and metabolic diseases, interest in lysosomal lipid hydrolysis has increased since the discovery of lipophagy (5), an LAL-dependent autophagic process in cytoplasmic lipid droplets (LDs) in which lysosomes ultimately break down triglycerides (TGs). Indeed, lipophagy is now recognized as a critical step in the control of hepatocyte steatosis, and selective inhibition of liver autophagy produces fatty liver disease in mice (6). It is now recognized that hypothalamic neurons (7) and macrophages (8) also degrade their TG stores by lipophagy, highlighting preference for acid versus neutral lipolysis in some cell types.

Adipocytes, which respond to nutritional inputs by orchestrating lipid mobilization and fatty acid fluxes across the body, are essential to maintain metabolic health. Adipose tissue (AT) is equipped with highly regulated cytoplasmic lipases that become activated at the LD surface to fine tune lipid release (9,10). Despite the well-established importance of AT cytoplasmic lipolysis, high LAL expression is also found in fat cells, which suggests additional roles besides TG mobilization.

Starting from the observation of decreased adipocyte LAL expression in human obesity, we have investigated the impact of LAL gain and loss of function in cultured fat cells and in mice, and highlighted an inverse relationship to adipocyte storage, linked to reorientation of fat-cell cholesterol metabolism and thermogenesis, independent of lipophagy. Our data are the result of an extensive analysis of the impacts of adipocyte LAL modulation and reveal beneficial metabolic effects of LAL stimulation through cholesterol dynamics.

Subcutaneous AT (ScAT) from the periumbilical region or omentum was excised from obese patients during bariatric surgery at La Pitié-Salpétriere Hospital (Paris, France), in accordance with local ethics committee recommendations (no. 0811792). All participants provided written informed consent. Clinical descriptions of the two groups is provided according to adipose samples use (Table 1): direct congelation (group 1, n = 45) or collagenase treatment (group 2, n = 21). Periumbilical AT from eight nonobese, healthy persons was obtained by lipoaspiration (Clinique Remusat, Paris, France).

We induced 3T3-L1 cells to differentiation with 1.25 µmol/L dexamethasone (catalog no. D1756; Sigma-Aldrich), 250 µmol/L IBMX (catalog no. 15879; Sigma-Aldrich), and 250 nmol/L insulin (catalog no. 19278; Sigma-Aldrich) for 3 days, followed by insulin alone. The 3T3-L1 cells were transduced with different lentiviral shRNA vectors (1-TRCN000076829, 2-TRCN000076831, 3-TRCN000076832, 4-TRCN000288217, and 5-TRCN000295618; Sigma-Aldrich) when fully differentiated or as undifferentiated preadipocytes, subsequently selected with puromycin (catalog no. A1113802; Thermo Fisher Scientific) to establish stable cell lines.

Cultured cells were serum starved for 6 h before addition or not of the following lysosome inhibitors: 0.1 µmol/L bafilomycin A1 (BAF) (catalog no. B1793; Sigma-Aldrich) or chloroquine (CQ) 25 µmol/L (catalog no. C6628; Sigma-Aldrich) for 4 h. P62 and LC3-II were probed by Western blotting.

TetO-CMV-hLAL mice were obtained by additive transgenesis (Pasteur Institute) by injecting into C57/Bl6J embryos a tetracycline-responsive promoter element–cytomegalovirus promotor upstream of human LAL (hLAL) cDNA, purified from a pTRE2hyg plasmid. Adipose-specific LAL–overexpressing mice were generated by breeding Adpn-rtTa mice with TetO-CMV-hLAL mice and genotyping with 5′-AGCCCAGTGTAAAGTGGCCC-3′ 5′-CTGGACAAGAGCAAAGTCAT-3′ and 5′-TGTGCCTTAACCGAATTCCT-3′ 5′-CTGGTTTGGGACCTTTGTCA-3′. Weaned littermates were fed a high-fat diet (HFD) (catalog no. D12492i; Research Diet) with free access to drinking water containing 1 g/L doxycycline (D9891; Sigma-Aldrich).

Mice were housed at room temperature (22°C) or thermoneutrality (30°C) in a climate chamber (PhenoMaster; TSE). Body composition was analyzed by nuclear magnetic resonance (Minispec LF90; Bruker). Blood and fat pads (gonadal (i.e., ovarian (OvAT) and epididymal AT, ScAT, and brown AT (BAT)) were frozen in liquid nitrogen, fixed with formalin (HT501128; Sigma Aldrich), or incubated with collagenase, as described previously (11). An oral glucose tolerance test was started by gavage with 2.0 g/kg glucose in mice fasted for 6 h. Glycemia was followed with a glucometer (Accu-Chek; Roche). Serum steroid profiling was measured by LC/MS-MS according to established methods (12).

Mice were acclimated in individual cages for 2 days before measurement of Vo₂, Vco₂, locomotor activity, and food and water consumption over five consecutive days at 22°C (PhenoMaster; TSE). The HFD and doxycycline dosing were maintained.

Total RNA was extracted using an RNeasy Mini Kit (catalog no. 74104; Qiagen), followed by quantitative real-time polymerase chain reaction using Master Mix (Applied Biosystems). Lysates in RIPA buffer (catalog no. R0278; Sigma Aldrich) containing proteinase inhibitors (Complete Mini Protease Inhibitor Cocktail; Roche) were prepared (Precellys 24; Bertin Technologies) and cleared for 10 min at 10,000 rpm. Protein concentration was measured (BCA Protein Assay Kit; Thermo Fisher Scientific). Western blotting was performed as described (13).

Cultured adipocytes or ScAT fragments were incubated in DMEM containing 1% BSA (catalog no. A8806; Sigma-Aldrich) for 2 h at 37°C with or without 10 µmol/L isoproterenol (catalog no. 1351005; Sigma-Aldrich). Released glycerol was measured with free glycerol reagent (F6428; Sigma-Aldrich). Frozen tissues were used to determine cholesterol (Amplex Red Cholesterol Assay Kit, catalog no. A12216; Invitrogen) or TG (Triglyceride Colorimetric Assay Kit, catalog no. 10010303; Cayman) contents.

Unpaired two-tailed t tests, ANOVA analysis, Mann-Whitney nonparametric tests, or Spearman correlations were performed with GraphPad Prism 7 (GraphPad Software).

The TetO-CMV-hLAL mouse strain generated during this study is available from the corresponding author upon request.

AT LAL expression is poorly documented. In mice, a positive association with body weight was reported, with elevated Lipa mRNA in obese compared with lean fat pads (14). We also have observed increased Lipa mRNA in the fat tissue of mice made obese by HFD feeding compared with that of lean animals fed chow (Fig. 1A). Moreover, microarray analysis of human ScAT LIPA expression replicated in two independent groups of obese (BMI >30 kg/m²) and nonobese persons (BMI <30 kg/m²) indicated significantly upregulated LIPA signal in obese patients (fold change: 1.55 and 1.07, respectively; q = 0.031 and 0.005, respectively).

To pinpoint the origin of this regulation, we first analyzed Lipa expression in the main cell subpopulations in mice AT (i.e., mature adipocytes or the different cell types in the stromal vascular fraction separated by FACS, as previously described (15). We observed comparable Lipa expression in floating adipocytes, endothelial cells (Cd45⁻/Cd31⁺), and resident macrophages (Cd45⁺ F4/80⁺), but very low mRNA levels in progenitor cells defined as Cd45⁻/Cd31⁻/Cd140⁺ (Fig. 1B). This suggested that cell composition, especially the adipose versus nonadipose cell ratio, might be a strong determinant of tissue LAL expression. In nonobese persons or patients with massive obesity (mean BMI, 47 kg/m²) (Table 1), stromal vascular fraction LIPA mRNA level was independent of the BMI of the donor on a per-cell basis (Fig. 1C). This suggested that upregulation of LAL expression with obesity is linked to the number of stromal cells present in fat, especially infiltrated macrophages. Importantly, contrary to stromal LAL, obesity modulated floating adipocyte expression and LIPA mRNA levels were lower in obese compared with lean adipocytes (Fig. 1D). Within the bariatric obese cohort in which two distinct fat-tissue depots (subcutaneous and omental) could be sampled in the same patient, the adipocyte LIPA mRNA level was highly concordant between fat locations (Fig. 1E). Unfractionated AT biopsy specimens, available in a separate group of obese patients with similar characteristics (Table 1), did not reveal links between LIPA expression and indicators of obesity severity (i.e., body weight, BMI, or total fat mass) but showed positive correlation with markers of obesity-related dysfunction, such as android fat distribution or circulating serum AST (Fig. 1F). These findings indicate that total AT LAL is a marker of metabolic impairment in obesity, rather than an indicator of obesity severity.

Thus, these experiments revealed a complex pattern of obesity-related adipose-LAL regulation, with a strong dependence on tissue composition with regard to nonadipose cells, concomitant with a deficit of LAL expression in mature adipocytes. Our findings highlight the need for cell type–specific studies to address the question of the significance of adipocyte LAL.

To explore the consequences of obesity-linked adipocyte-LAL downregulation, we modulated LAL expression using lentiviral vectors to reduce LAL in fully differentiated, lipid-laden 3T3-L1 cells, a recognized model for fat cells. Five days after differentiation, when clearly visible LDs had accumulated, cells were transduced with five different shRNA lentiviral stocks or no virus, and LAL expression was assessed after an additional 7 days. Among five sequences, two (sh3 and sh4) inhibited LAL expression with >50% efficiency, whereas the other three (sh1, sh2, and sh5) showed slight or no inhibition (Fig. 2A). Cells treated with sh3 and sh4 were analyzed together and thereafter denominated Lipa shRNA, whereas cells treated with inactive sh1, sh2, and sh5 were considered shRNA controls. Similar to that reported in genetically or pharmacologically LAL-deficient macrophages (16), adipocyte LAL downregulation increased total-cholesterol cell contents (Fig. 2B). Furthermore, high doses of oxidized LDL, known to inhibit LAL activity (17), increased cholesterol contents in LAL-expressing adipocytes up to the level of Lipa shRNA cells (Fig. 2B). Thus, as a hallmark of LAL functional impairment, cholesterol loading is seen in both oxidized LDL treatment and lentivirus-mediated Lipa knockdown.

TGs are LAL substrates, as are cholesterol esters. Despite unaffected expression of cytoplasmic lipases and LD-coating Plin1 (Supplementary Fig. 1A), or adipocyte-specific genes (Supplementary Fig. 1B), we found that TG contents were increased in Lipa shRNA cells compared with control shRNA or untransfected cells (Fig. 2C). Perilipin1 protein content increased in Lipa shRNA compared with control cells (Fig. 2D), further indicating lipid-droplet accumulation. Thus, LAL downregulation favors adipocyte TG storage.

We next generated stable shRNA Lipa knockdown by puromycin selection, in which low levels of LAL mRNA were maintained before (D0) or after (D7) full differentiation (Supplementary Fig. 1C). Overall fat-cell conversion was unaffected, as indicated by normal adipocyte markers expression (Supplementary Fig. 1D). Lysosome integrity and density were conserved, as shown by the protein contents of lysosome chaperones Lamp2A and Hsc70 (Supplementary Fig. 1E) and pulse-chase labeling with dextran–Texas Red dye (Fig. 2E). Autophagosome marker (LC3-II) or autophagic substrate (p62) contents were also similar in control and LAL-deficient cells (Supplementary Fig. 1F). Treatment with lysosome inhibitors (CQ and BAF) to assess autophagic flux showed equivalent accumulation rates of p62 and LC3-II (Fig. 2F–H), indicating unaffected lysosomal clearance in LAL-deficient conditions. Regarding lipolysis, the presence of BAF did not affected basal glycerol release, which was also not different in LAL-deficient cells and in control cells (Fig. 2I–J). BAF only slightly inhibited isoproterenol-stimulated lipolysis (<25%) in control adipocytes (Fig. 2K), as also reported previously (12). This effect was also observed in LAL-deficient cells (Fig. 2K), suggesting a lipophagy-unrelated action. Of note, lysosome inhibitors did not affect perilipin1 contents (data not shown), indicating that the adipocyte-specific isoform is not degraded in lysosomes, contrary to that shown for nonadipose perilipin-2 and -3 (18). Thus, the impact of LAL downregulation in adipocytes is autophagy independent.

We next investigated potential changes in cell-cholesterol homeostasis, revealed by sterol regulatory element binding protein (SREBP) status. As a cholesterol-regulated transcription factor, autoregulated SREBP2 and its target genes (HMG-CoA reductase, LDL receptor) are induced in response to cholesterol depletion sensed in the ER. Independently generated, stable transfectants were stratified into three groups on the basis of residual levels of Lipa expression (Fig. 3A), in which Hmgcr, Srebf2, and Ldlr mRNA levels gradually increased with stronger LAL knockdown, suggesting a state of ER-sensed cholesterol depletion in LAL deficiency (Fig. 3B). Indeed, residual Lipa expression negatively correlated with Srebf2 gene expression, the lower Lipa mRNA levels, and the higher Srebf2 expression (Fig. 3C), and also varied inversely to Fasn, a SREBP-regulated lipogenic gene (Supplementary Fig. 2A). Cell replenishment with a bolus of a free cholesterol–supplemented medium, followed by a 4-h chase in standard medium, was sufficient to abolish differences in Srebf2 and Hmgcr mRNA between Lipa groups (Fig. 3D and Supplementary Fig. 2B), suggesting that restrained LAL-dependent, free-cholesterol delivery to the ER was most likely the cause of upregulated expression of cholesterol-regulated genes.

Because cholesterol-sensitive ER is also the site of LD assembly and budding, we probed for altered topology of TG accumulation by LAL deficiency. Concomitant with elevated TG contents and LD-associated perilipin fluorescence in Lipa shRNA cells compared with shRNA control cells (Fig. 3E), we also found increased FAS protein contents in Lipa-deficient cells, which is indicative of active lipogenesis (Fig. 3F). Unexpectedly, the mean diameter of LDs proportionally declined with Lipa inhibition. ANOVA analysis of >10,000 single LDs from independent cell pools indicated significant negative contribution of LAL deficiency to LD size (Fig. 3G). This infers that the number of LDs per cell increases in LAL-deficient adipocytes, because more lipid was accommodated in smaller LDs. However, because LDs are so numerous and heterogeneous in differentiated 3T3-L1 adipocytes, direct measure of the total number of LDs per cell requires three-dimensional reconstitution of images. Although not directly evaluated, we suggest that an increase in the number of LDs could be linked to facilitated LD budding from cholesterol-depleted ER. Other ER functions, such as the unfolded protein response, remained unaffected by LAL deficiency, because a normal response to an unfolded protein response inducer (i.e., dithiothreitol), which disrupts oxidative protein folding in the ER lumen, was mounted irrespective of Lipa expression (Fig. 3H and I). Together, these data underline LAL dependence of a specific cholesterol lysosome to ER axis in regulating adipocyte homeostasis. We propose that defective cholesterol-ester hydrolysis in LAL deficit could limit free-cholesterol exit from lysosomes and induce a chronic state of cholesterol demand in the ER in favor of LD budding.

We next generated mice with adipocyte-specific LAL overexpression to investigate the impact of adipocyte LAL modulation at the whole-body level. We bred Adpn-rtTA mice with TetO-CMV-hLAL mice and fed littermates an HFD for 12 weeks with free access to doxycycline-supplemented drinking water (1 g/L). We used primers recognizing both the hLAL transgene and the endogenous mouse Lipa mRNA to confirm significant overexpression of LAL in ATs (Fig. 4A). Compared with control single transgenics (Adpn⁺/LAL⁻ or Adpn⁻/LAL⁺), a significant increase in LAL expression was seen in Adpn⁺/LAL⁺ female ATs but not in the liver, as expected. In male mice, because of high levels of endogenous LAL in the epidydimal fat, overexpression was seen only in BAT and subcutaneous ScAT depots (Fig. 4B). Adpn⁺/LAL⁺ livers of male mice expressed unpredicted high levels of LAL mRNA compared with livers of control mice. Thus, considering sex-specific responses, we decided to limit analysis to female mice only, in which transgene-driven adipocyte LAL expression was robust and consistent (Fig. 4C).

LAL-overexpressing Adpn⁺/LAL⁺ female mice gained less weight than Adpn⁻/LAL⁺ or Adpn⁺/LAL⁻ control mice fed an HFD for 12 weeks (Fig. 4D). Analysis of body composition indicated decreased fat mass, but not lean mass or fluids (Fig. 4E), and less developed gonadal AT and ScAT depots (Fig. 4F). Because LAL has TG lipase activity, we checked for stimulation of adipocyte TG mobilization to explain less fat deposition. Basal glycerol release by ex vivo ScAT fragments was similar regardless the donor mice genotype or diet (chow diet or HFD). Maximally stimulated lipolysis with isoproterenol was also unaffected (Fig. 4G). Thus, the lean phenotype of LAL-overexpressing mice is unrelated to adipocyte lipolysis and TG mobilization.

In line with their lower fat mass, LAL-overexpressing mice were also less prone to HFD-dependent, visceral AT inflammation and expressed less Cd68, Il6, and Tnfa mRNA than did control mice (Fig. 5A). We also found elevated Scd1 (stearoyl CoA desaturase 1) mRNA expression in the OvAT of Adpn⁺/LAL⁺ mice (Supplementary Fig. 3A), which translated into a greater proportion of monounsaturated fatty acids at the expense of saturated fatty acids (Supplementary Fig. 3B and C). A high ratio of monounsaturated to saturated fatty acids is metabolically beneficial, as are low levels of AT inflammation, which prompted us to check glucose homeostasis in LAL-overexpressing mice. Compared with control mice, Adpn⁺/LAL⁺ mice had lower fasting blood-glucose levels (Fig. 5B), and their glycemic response to an oral glucose load (i.e., the oral glucose tolerance test) was less (Fig. 5C). Noticeably, a highly significant correlation between glycemia and circulating cholesterol concentrations was found (Fig. 5D), and Adpn⁺/LAL⁺ mice had lower blood cholesterol levels than did control mice (Fig. 5E), despite comparable triglyceridemia (Fig. 5F). Together, these data indicate a beneficial impact of adipocyte LAL overexpression to damper the deleterious metabolic consequences of a HFD.

To investigate the cause of LAL-mediated obesity resistance in mice fed an HFD, we assessed the energy balance components by indirect calorimetry. We compared Adpn⁺/LAL⁺ mice with Adpn⁻/LAL⁺ control mice and found similar food intake and locomotor activity in the two groups (Fig. 5G), but the energy expenditure profiles in Adpn⁺/LAL⁺ mice were higher than that of control mice (Fig. 5H). This suggested a role for thermogenesis, a prominent energy expense module required to defend mouse core temperature in standard housing conditions (22°C). LAL expression in BAT inversely correlated with body weight response to the HFD (Fig. 5I), which further suggests BAT contribution.

To decipher the role of BAT-related thermogenesis in the obesity-resistant phenotype of LAL-overexpressing mice, we reevaluated mice fed an HFD, as described in the previous paragraph, but housed under thermoneutral conditions (i.e., 30°C instead of 22°C) to shut off a thermogenic response. LAL overexpression persisted in thermoneutrally housed Adpn⁺/LAL⁺ mice (Fig. 6A), and BAT morphology showed expected differential lipid accumulation after 4 weeks of conventional versus thermoneutral housing (Fig. 6B). Within this 4-week period, we did not observe differences in cumulative weight gains between thermoneutral and conventional control groups, but body weights poorly segregated among genotypes in thermoneutral mice, whereas differences had clearly established in mice raised conventionally (Fig. 6C). Furthermore, even if a small difference in body weight still persisted between the thermoneutral Adpn⁺/LAL⁻ and Adpn⁺/LAL⁺ groups, this was not related to lower fat mass but rather to a decrease in lean mass existing before thermoneutrality was applied (Fig. 6D). Lowering of cholesterolemia in conventionally raised, LAL-overexpressing mice (Fig. 5E) was no longer present after thermoneutral housing (Fig. 6E), nor was amelioration of glucose tolerance (Fig. 6F). Thus, the LAL mouse phenotype is highly dependent on housing temperature, which demonstrated a role for adipocyte thermogenic activity and led us to focus on brown/beige AT depots as preferential functional targets of LAL overexpression.

Investigating BAT closer, we found no significant change in TG contents between Adpn⁺/LAL⁺ and control mice (Fig. 7A), but free-cholesterol levels were increased in the BAT of LAL-overexpressing mice (Fig. 7B). Free cholesterol represents the major fraction of total cholesterol in BAT (81.7% ± 6.1% and 83.0% ± 3.7% (mean ± SD) in Adpn⁺/LAL⁺ and control mice, respectively) and is present in much higher levels than the esterified cholesterol form (Fig. 7B insert).

Selective impact on BAT was suggested because no change with LAL overexpression was found in the white OvAT (Supplementary Fig. 4). In BAT of Adpn⁺/LAL⁺ mice, cholesterol-regulated Ldlr and Hmgcr mRNA expression genes were downregulated compared with those of control mice, and Srebf2 expression followed the same trend, although it did not reach the statistical significance threshold (Fig. 7C). Genes controlling cholesterol efflux (i.e., Abca1 and Scarb1) were not induced (Fig. 7D), consistent with tissue free-cholesterol accumulation. We also observed that compared with control BAT, LAL-overexpressing BAT expressed lower levels of ER stress markers, suggesting ameliorated metabolic adaptation (Fig. 7E). In agreement, BAT Ucp1 mRNA expression was positively associated with LAL expression (Fig. 7F). Regarding ScAT, in a white fat depot in which brown-like adipocytes can develop (a process called fat beiging), we also detected positive Ucp1 mRNA expression in 40% of the Adpn⁺/LAL⁺ mice, compared with <10% of the control mice (Fig. 7G). Also, there was a trend toward a higher proportion of fat tissue positive for Cidea and Dio2 mRNA in ScAT of Adpn⁺/LAL⁺ mice (Fig. 7G), suggesting a higher potential for beiging. Together, these data indicate LAL overexpression improved thermogenic potential.

BAT activation fits with a leaner phenotype of LAL-overexpressing mice fed an HFD but surprisingly coexists with tissue free-cholesterol accumulation, which is usually considered a deleterious condition for mitochondrial function (19). Nonetheless, the inner mitochondrial membrane is equipped with a steroidogenic enzyme machinery, which transforms free cholesterol into steroids, a pathway that has the potential to alleviate mitochondrial cholesterol overload. The gene encoding the first step in steroidogenesis, the mitochondrial P450 side-chain cleavage enzyme Cyp11a1, which converts cholesterol into pregnenolone, is expressed in brown adipocytes (20). We found that the expression of this steroidogenic gene, Cyp11a1, as well as its upstream transcriptional regulator Nr5a1 (alias steroidogenic factor 1), were upregulated in BAT of LAL-overexpressing mice (Fig. 7H). In addition, BAT Nr5a1 expression was positively associated with LAL (Fig. 7I) and negatively linked to body weight gain in mice fed an HFD (Fig. 7J). Mass spectrometry profiling of serum steroids indicated a trend toward higher pregnenolone and 17-OH pregnenolone concentrations in LAL-overexpressing mice compared with control mice (Fig. 8A and B). Furthermore, levels of 5α-dihydrotestosterone and testosterone, downstream products of pregnenolone in the sex hormone branch of steroidogenesis, were significantly higher in Adpn⁺/LAL⁺ mice serum (Fig. 8C and D). Also, 17OH-pregnenolone and 5α-dihydrotestosterone positively correlated with BAT Nr5a1 mRNA expression (Fig. 8E and F). Other serum metabolites in the corticosteroid or mineralocorticoid pathways remained unchanged in the two groups of mice (data not shown).

These data reveal induction of BAT mitochondrial steroidogenesis, which likely is an adaptive response to accelerated cholesterol use in BAT following LAL overexpression. This potentially explains why mitochondrial function is preserved in cholesterol-laden BAT of LAL-overexpressing mice. It also provides a molecular basis for the role of BAT as a cholesterol sink, which remain to be further characterized.

The findings of our study reveal novel insights about the unappreciated role of adipocyte lysosomal lipase in energy balance regulation. Previous report in mice with LAL global loss of function demonstrated lipotoxic fattening of organs, especially liver and intestine (21). As models for life-threatening Wolman disease in patients, these mice could be normalized by hepatocyte LAL rescue (22). Also, liver-specific gene deletion of LAL revealed that cholesterol esters were the most prominent lipids accumulated in hepatocytes—far more than TGs and retinyl esters, other in vitro LAL substrates (23). Global LAL deficiency was reported to cause disappearance of AT stores (21), which led to the conclusion that LAL positively regulated AT storage. However, lipoatrophy in global LAL deficiency is paradoxical, largely because of malabsorptive cachexia and liver failure at the end of disease development. Fat mass was still preserved in young LAL-deficient mice, it was shown recently, but hypothermic events and defective BAT thermogenesis were reported (24), which are in line with our present observations.

Our data reveal a metabolic role of adipose LAL modulation. We demonstrate that LAL regulates energy expenditure and subsequent AT TG storage, primarily as a gatekeeper of lysosome cholesterol exit for delivery into adipocytes. When adipocyte LAL expression decreases, cholesterol delivery from lysosomes to ER is reduced, which, in turn, activates cholesterol-dependent SREBP and subsequent fat storage. On the other hand, LAL overexpression, by increasing the turnover of cholesterol through the lysosomal-ER axis, promotes beneficial brown fat activity and a mitochondrial steroidogenic pathway. LAL knockdown in adipose cell lines and overexpression in mice led to apparently different end points (i.e., increased fat storage or activated BAT thermogenesis), which can appear unrelated. However, a common pattern in these two opposite situations is the modulation of cholesterol-dependent SREBP pathway (Fig. 8G).

A limitation of our study is lack of information on LAL protein levels or enzyme activity. We could not validate consistent LAL immunoreactivity in AT extracts despite attempts with several commercial antibodies. Accurate measurement of enzyme activity was also confounded by the unusual status of AT esterase activities, which comprises, in addition to LAL, a large variety of other abundant lipases that can still exhibit residual activity at acidic pH. In such a context of lipase diversity, the accuracy of a “specific” LAL inhibitor compound is not validated. Further, the artificial LAL substrate designed in the LAL assay competes with uncontrolled adipose TG, which further confounds enzyme activity determination.

Our data highlight LAL-dependent regulation of adipocyte cholesterol balance, in line with findings of previous studies on the importance of ABCA1 (25) and ABCG1-mediated cholesterol efflux (26) in fat-cell lipid storage. LAL-dependent cholesterol regulation is closely linked to adipocytes’ unique low capacity for endogenous cholesterol synthesis (27,28), leading to inability to restore homeostasis with de novo synthetized cholesterol. The LAL-dependent pathway for exogenous cholesterol delivery to adipocytes is responsive to nutritional overload in obesity. We have shown here that in mice fed an HFD, BAT cholesterol content is LAL dependent, as is body weight gain. Although direct subcellular localization of cholesterol accumulation upon LAL manipulation would have reinforced our conclusions, indirect biochemical quantifications confirm that free cholesterol indeed, is accumulated in LAL-overexpressing fat cells. Together with BAT cholesterol accumulation in LAL-overexpressing mice, we found lower cholesterolemia, which might suggest that activated BAT serves as a metabolic sink for cholesterol, as it does for glucose, fatty acids (29), and branched-chain amino acids (30). Indeed, protective brown fat activation against hypercholesterolemia has been proposed (31).

LAL is required for lipophagy, but we found autophagy-independent, LAL-dependent regulation in adipocytes. This fits with a minor role of adipocyte autophagic degradation of LDs in the face of the prominent importance of cytoplasmic lipases for adipocyte TG mobilization, even if autophagic clearance of other cell components is central for adipocyte maintenance and response to obesity (13,32,33).

By exploring LAL status of human adipocytes, we show that when the confounding contribution of AT macrophage LAL is eliminated, LAL expression is reduced in obese adipocytes, which opens the avenue of therapeutic LAL activation. With brown fat as a preferred adipose target, LAL activation could ameliorate metabolic condition in obese people if combined with interventions aimed at favoring brown/beige fat stimulation. Furthermore, beneficial lowering of blood cholesterol levels could be expected, as indicated in the mouse study. The association between cholesterolemia and adipocyte LAL in the obese population explored here could not be investigated because most patients were receiving statin treatment. Altogether, our observations argue for considering LAL activation in the setting of obesity as a strategy to implement protective effects of brown fat activation against metabolic disorders.

Acknowledgments. Adpn-rtTa mice were provided by P. Scherer (UT Southwestern Medical Center, Dallas, TX). The authors thank the members of the Mouse Genetics Engineering of the Institut Pasteur for technical support with transgenic mice. Marie Lhomme (ICAN Analytics Lipidomic facility, La Pitié-Salpetrière, Paris) is acknowledged for providing fatty acid analysis.

Funding. The funding of the National Research Agency (ANR-14-CE12-0017, LIPOCAMD) is acknowledged.

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

Author Contributions. C.G., C.R., A.L., and F.L.-V. performed the experiments. K.C., E.L.G., L.Y.-C., and I.D. designed the experiments and wrote the manuscript. I.D. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.