D. Galaris, Laboratory of Biological Chemistry, University of Ioannina Medical School, 451 10 Ioannina, Greece Fax: +30 26510 07868 Tel: +30 26510 07562 E-mail: firstname.lastname@example.org
The aim of the present study was to investigate in detail the molecular mechanisms by which free fatty acids induce liver toxicity in liver cells. HepG2 and Huh7 human liver cell lines were exposed to varying concentrations of stearate (18:0), oleate (18:1), or mixtures of the two fatty acids, and the effects on cell proliferation, lipid droplet accumulation and induction of endoplasmic reticulum stress and apoptosis were evaluated. It was observed that: (a) stearate, but not oleate, inhibited cell proliferation and induced cell death; (b) stearate-induced cell death had the characteristics of endoplasmic reticulum stress-mediated and mitochondrial-mediated apoptosis; (c) the activation of stearate in the form of stearoyl-CoA was a necessary step for the lipotoxic effect; (d) the capacity of cells to produce and accumulate triacylglycerols in the form of lipid droplets was interrupted following exposure to stearate, whereas it proceeded normally in oleate-treated cells; and (e) the presence of relatively low amounts of oleate protected cells from stearate-induced toxicity and restored the ability of the cells to accumulate triacylglycerols. Our data suggest that interruption of triacylglycerol synthesis in the endoplasmic reticulum, apparently because of the formation of a pool of oversaturated intermediates, represents the key initiating event in the mechanism of saturated fatty acid-induced lipotoxicity.
RNA-dependent protein kinase-like endoplasmic reticulum eukaryotic initiation factor-2α kinase
saturated fatty acid
unsaturated fatty acid
Dietary habits in the Western world have changed drastically during the last few decades, and this change correlates with increasing levels of obesity, implying that diet may be associated with the development of insulin resistance, type 2 diabetes, cardiovascular disease and other pathologies in the general population . Consumption of food rich in fat causes qualitative and quantitative changes in serum free fatty acid (FFA) levels, and increases the rate of uptake and accumulation of lipids in nonadipose tissues such as the liver, which is the main lipid-metabolizing organ. Inappropriate accumulation of excess lipids in liver cells in the form of lipid droplets has been proposed to lead to dysfunction of hepatocytes and, consequently, to serious pathological complications [2,3]. Nonalcoholic fatty liver disease (NAFLD) is a term used to characterize a spectrum of pathological changes ranging from simple fatty infiltration (steatosis) to hepatic steatosis accompanied by inflammation, fibrosis, and cirrhosis (nonalcoholic steatohepatitis) [4,5]. Despite the high prevalence of NAFLD and its potential for serious complications, the underlying molecular mechanisms that determine the progression to liver damage remain poorly understood and need further investigation.
A number of recent in vitro and in vivo studies have shown that different forms of fatty acids exert remarkably different effects. Exposure of a variety of cell types, including hepatocytes, to long-chain saturated fatty acids (SFAs) led to increased expression of proinflammatory cytokines, inhibition of insulin signaling, induction of endoplasmic reticulum (ER) stress, and promotion of cell death, mainly by apoptosis [6–12]. On the other hand, unsaturated fatty acids (UFAs) were not toxic at the same concentrations and, in addition, their presence protected cells from SFA-induced effects [6,13–16]. A protective role for endogenously generated UFAs was also indicated by in vivo experiments using genetically modified mice bearing an inactivating mutation in the gene encoding the enzyme stearoyl-CoA desaturase 1. This enzyme is responsible for the introduction of a double bound into long-chain SFAs . However, the exact molecular mechanisms underlying these events remain partially understood, and the data obtained, as well as the explanations provided, are often controversial. For instance, it has been suggested that SFAs can influence important cellular signaling pathways either directly or indirectly through the generation of reactive oxygen species , ceramides , or accumulation of saturated triacylglycerols (TAGs) , leading to cellular dysfunction and ultimately to cell death. The precise mechanisms lying beneath these processes remain elusive, and the key elements that determine the induction of toxicity have not been identified yet.
The aim of the present investigation was to perform a detailed evaluation of several aspects concerning SFA-induced lipotoxicity in order to define the key event(s) involved in this mechanism. For this purpose, human hepatoblastoma cells were exposed to varying combinations of saturated (stearate, 18:0) and unsaturated (oleate, 18:1 cis) fatty acids for various time periods, and cell proliferation, toxicity, induction of ER stress and apoptosis and lipid droplet accumulation were evaluated.
SFAs inhibit proliferation and induce cell death
Exposure of HepG2 cells to 0.3 mm of the SFA stearate (18:0), but not to the monounsaturated fatty acid oleate (18:1, cis), induced a transient inhibition of cell proliferation during the first 24 h (Fig. 1A). This observation was also confirmed by analysis of bromodeoxyuridine (BrdU) incorporation into DNA, which decreased by more than 50% after 24 h of stearate treatment. However, cells regained their normal proliferation capacity at longer incubation periods, whereas coadministration of oleate (0.3 mm) prevented the transient inhibition of proliferation induced by stearate alone (Fig. 1A,B).
When higher concentrations of FFA (0.6 mm) were used, cells could not recover following stearate treatment, and toxic effects were apparent (Fig. 1C). Again, coadministration of oleate protected cells from stearate-induced toxicity and restored the capacity of cells to proliferate (Fig. 1C,D). The protection offered by oleate was apparent even when it was administered 12, 24 or 36 h following stearate administration (Fig. 1E) or in ratios of oleate to stearate lower than 1 : 1 (Fig. 1F). Essentially the same results were obtained when another human hepatocellular carcinoma cell line (Huh7) or other unsaturated (linoleic acid, 18:2 cis) or saturated (palmitate) fatty acids were used instead of HepG2 cells, oleate, and stearate, respectively (results not shown).
Stearate treatment induces ER stress and mitochondrial-mediated apoptosis
As shown in Fig. 2, treatment of HepG2 cells with 0.6 mm stearate led to the appearance of condensed and fragmented nuclei (Fig. 2A,B), hypodiploid sub-G1 DNA (Fig. 2C), DNA internucleolar fragmentation (Fig. 2D), caspase-3 cleavage (Fig. 2E), and the appearance of cytochrome c in the cytosol (Fig. 2F), which are clear characteristics of mitochondrial-mediated apoptotic cell death. In all cases, oleate was not toxic by itself, and its coadministration with stearate prevented the appearance of these apoptotic markers.
Different Bcl-2 family members serve as proapoptotic and antiapoptotic mitochondrial regulators under certain circumstances [21,22]. As shown in Fig. 3A, the relative amount of the antiapoptotic protein Bcl-2 was gradually increased in HepG2 cells during the initial 22 h of stearate treatment, but this increase was interrupted thereafter, and the Bcl-2 concentration stabilized at a somewhat lower level. On the other hand, the proapoptotic Bcl-2-like protein Bax was activated, as indicated by its translocation from the cytosolic fraction to the mitochondrial fraction following stearate administration (Fig. 3B). The presence of oleate inhibited this translocation, indicating the involvement of Bax activation in stearate-induced mitochondrial destabilization and apoptosis.
In order to further evaluate specific molecular mechanisms contributing to induction of mitochondrial destabilization, we analyzed changes in specific markers of ER stress, which has been proposed to be involved in SFA-induced lipotoxicity [8,13,16,23]. As shown in Fig. 4A, increased phosphorylation of eukaryotic translation initiation factor 2α (eIF2α) was apparent after 16 h of stearate treatment, reached a peak at 22 h, and declined thereafter. Concomitantly with eIF2α phosphorylation, dramatic elevations in the expression of activating transcription factor 4 (ATF4) and of CCAAT/enhancer-binding protein homologous protein (CHOP) proteins downstream of eIF2α were observed (Fig. 4A), indicating the initiation of ER stress-induced apoptosis. Oleate was unable to induce the expression of the proapoptotic protein CHOP, and its coadministration with stearate inhibited CHOP expression induced by stearate alone (Fig. 4B). These observations indicate that activation of the RNA-dependent protein kinase-like ER eukaryotic initiation factor-2α kinase (PERK) branch of ER stress is triggered following exposure of HepG2 cells to stearate, and the presence of oleate prevented this activation.
Along with PERK branch activation, the phosphorylation of c-Jun N-terminal kinase (JNK) was also increased, displaying a strong peak after 22 h of stearate treatment (Fig. 4C). In oleate-treated cells, JNK phosphorylation increased slightly after 16 h of treatment, but oleate, in contrast to stearate, did not exhibit the sharp increase at 22 h. In addition, oleate prevented stearate-induced phosphorylation of JNK when the two agents were coadministered (Fig. 4C). Taken together, these results indicate that the location of the protective action of oleate was upstream of ER stress activation.
Whether the protection offered by oleate was specific for SFA-induced ER stress or represented a more general phenomenon was also investigated. The presence of oleate was unable to prevent the toxic effects induced by thapsigargin or tunicamycin (two classical ER stress inducers), indicating that oleate is not a general inhibitor of the ER stress response (results not shown).
Stearate treatment interrupts TAG synthesis and lipid droplet accumulation
In addition to its role in proper protein folding, the ER is responsible for lipid synthesis. In particular, excess availability of FFAs, as is the case in the present experimental model, leads to increased formation of TAGs, which are either released from the cells as very low density lipoprotein or stored in the cytosol as lipid droplets. It was observed that the accumulation of lipid droplets was efficient in oleate-treated cells, whereas stearate-treated cells contained fewer and smaller lipid droplets after 24 h of treatment (Fig. 5A). Moreover, coadministration of oleate restored the capacity of stearate-treated cells to accumulate lipid droplets. This observation was further confirmed by using TLC analysis, which showed that stearate-treated cells contained much lower amounts of TAG than oleate-treated and oleate plus stearate-treated cells (results not shown). These observations indicate that the magnitude of cellular steatosis as such is not responsible for lipotoxicity. Rather, it is likely that diversion of fatty acids into inert TAG stores contributes to cell survival and preserves cellular functions.
In order to further investigate these fundamentally different effects of different FFAs, we performed long-scale time-course experiments. Cells were treated with FFAs for time periods of 3, 6, 12, 24 and 36 h, before staining of the accumulated neutral lipids with Nile red and analysis of the fluorescence intensity of individual cells by flow cytometry. Cell fluorescence increased progressively in oleate-treated and oleate plus stearate-treated cells, whereas it was significantly lower in stearate-treated cells during the first 3 and 6 h of treatment (Fig. 5B). Interestingly, the fluorescence in stearate-treated cells started to decrease gradually at exposure times longer than 6 h, giving rise to a distinct cell population with basal levels of fluorescence (Fig. 5B). After 36 h, almost the entire cell population was devoid of lipid droplets.
It is obvious from these results that TAG synthesis was initially hindered following stearate administration and was completely interrupted at longer incubation periods. Interestingly, the interruption of TAG synthesis preceded the appearance of toxic effects, supporting the notion that it constitutes the initiating event in the process of lipotoxicity.
Stearate has to be activated in order to be toxic
The first enzyme involved in metabolism of FFAs after their uptake into liver cells is the long-chain acyl-CoA synthetase (ACS), which activates fatty acids by linking them to coenzyme A. As shown in Fig. 6, triacsin C (TrC), a specific competitive inhibitor of ACS [24,25], was not toxic by itself, whereas it inhibited the accumulation of lipid droplets following exposure of cells to either stearate or oleate (Fig. 6A,B). At the same time, TrC protected cells from stearate-induced death, as indicated by estimation of cell viability (Fig. 6C), annexin-V plus propidium iodide (PI) staining (Fig. 6D), and other ER stress markers (results not shown). This protective effect could not be attributed to a nonspecific inhibitory effect on ER stress, as TrC was not able to prevent CHOP induction by thapsigargin or tunicamycin, two classic ER stressors (Fig. 6E). These observations show that it is not stearate as such that is responsible for inducing cell toxicity, but one or more of its metabolic intermediates in the pathway of TAG synthesis. In addition, these results show that the properties of the metabolic intermediates of stearate and oleate must be fundamentally different.
The results presented in this investigation, in agreement with previously reported observations, revealed fundamentally different effects of SFAs and UFAs on liver cells [8,11–13]. In an attempt to identify the key event(s) responsible for these differences, we examined the main steps involved, following the uptake of saturated and unsaturated FFAs into the cells. After their internalization, FFAs are converted to fatty acyl-CoA, a reaction catalyzed by ACS. Fatty acyl-CoAs are activated forms of fatty acids that can be either oxidized in mitochondria or utilized in the ER as substrates for the synthesis of phospholipids, cholesterol esters, and TAGs [26–28]. The observation in this investigation that inhibition of ACS by TrC abolished both FFA-induced lipid droplet accumulation and SFA-induced toxicity (Fig. 6) indicates that SFA activation is essential for the manifestation of toxicity. It has to be noted that TrC was not able to inhibit thapsigargin-induced or tunicamycin-induced ER stress (Fig. 6E), thus excluding the possibility of nonspecific inhibition of ER stress. When the available acyl-CoAs are in excess, they are channeled towards TAG synthesis.
The main findings of the present investigation were the observations that TAG synthesis in liver cells exposed to excess stearate was interrupted, and that this interruption preceded the appearance of toxic effects (Fig. 5A,B). In sharp contrast, oleate-treated cells, which continued to proliferate normally, were able to produce TAGs continuously and accumulate them in the form of lipid droplets. Moreover, coadministration of oleate restored the ability of stearate-treated cells to synthesize TAGs and prevented cell toxicity. These findings are in agreement with previous observations from the Schaffer group, indicating increased incorporation of palmitate (16:0) into the TAG pool only in the presence of oleate .
The above results raise two main questions: (a) what is the cause of TAG synthesis inhibition, and (b) what is the exact nature of the events that ultimately lead to cell toxicity?
Regarding the first question, it is obvious that one or more steps (following acyl-CoA formation) in the cascade of TAG formation that take place in ER membranes are defective in SFA-treated but not in UFA-treated cells. It has been previously shown that the degree of saturation of fatty acyl chains in TAG synthesis intermediates, such as phosphatidic acid and diacylglycerol (DAG), can influence their physicochemical properties, and in this way modulate their interactions with specific proteins [29–32]. Oversaturated DAGs, for example, were unable to interact with protein kinase C, and this effect was attributed to the formation of gel-like domains (instead of liquid-crystalline domains) in the membranes, making these molecules unavailable for the required interactions [30,33]. Addition of UFAs could restore the liquid-crystalline phase, making DAG molecules accessible to the interacting proteins . We propose that a similar mechanism can satisfactorily explain the results reported in this work, as well as the majority of previous observations from other laboratories.
Regarding the second question, the induction of ER stress and apoptosis observed in SFA-treated cells can be explained by modulation of the physicochemical properties of ER membranes by saturated lipid intermediates, such as PA and DAG. Excessive saturation accompanied by the formation of gel-like domains can influence the rigidity and fluidity of ER membranes, thus compromising the functional integrity of these organelles. It has to be stressed here that the ER is especially vulnerable, as its membranes require higher concentrations of UFAs in order to be functional . In support of this notion, previous investigations have shown major irregularities in the morphology of the ER in SFA-treated but not UFA-treated cells [9,20,35]. Moffitt et al.  suggested that accumulation in the ER lumen of oversaturated TAGs, which cannot be further processed, because of their inappropriate physicochemical properties (high melting point), is the main cause of toxicity. This proposal, however, is not consistent with the disappearance of lipid droplets from stearate-treated cells, as observed in this investigation (Fig. 5B).
The ER is the site of synthesis of all secretory proteins and resident proteins of the membrane system, and any perturbation that compromises the protein-folding capacity of the organelle can lead to ER stress [36–38]. ER stress is a general, integrated stress response displayed by mammalian cells. This response can be divided in two phases according to the intensity and the duration of the stress. An initial adaptive response culminates in the temporary inhibition of protein synthesis, providing cells with the opportunity to recover and restore normal homeostasis. The data presented in this work demonstrate that cells exposed to stearate are moved initially towards such an adaptive state, as indicated by the transient inhibition of cell proliferation (Fig. 1A,B) and the early phosphorylation of eIF2α (Fig. 4A). When the stress is more intensive and prolonged, secondary events, such as ATF4 and CHOP protein expression and JNK activation, were induced, leading ultimately to cell death by apoptosis (Fig. 4A,C). Prolonged ER stress and JNK activation, as observed in this study, usually stimulate apoptosis by several pathways, including the translocation of Bax to mitochondria, and CHOP-regulated inhibition of the expression of antiapoptotic proteins, such as Bcl-2 [23,39]. A schematic representation of the events observed in FFA-supplemented cells is presented in Fig. 7.
Although the conditions prevailing in this cellular model are quite different from those prevailing in vivo, previous experiments with rats fed a diet enriched in SFAs demonstrated similar characteristics of ER stress activation and apoptosis in the liver . Moreover, in accordance with the results presented in this article, the above characteristics were not apparent in animals fed a control diet or a diet containing UFAs, although steatosis developed. In addition, phosphorylation of eIF2α, which is characteristic of PERK branch activation of ER stress, has been demonstrated in humans with NAFLD and nonalcoholic steatohepatitis .
In conclusion, it is proposed that the key event determining SFA-induced lipotoxicity is the interruption of TAG synthesis. It is suggested that the creation of a pool of oversaturated lipid intermediates makes these molecules inaccessible to the enzymes of TAG synthesis, whereas a certain degree of unsaturation can restore normal TΑG formation. Excessive saturation compromises the functional integrity of the ER, leading ultimately to ER stress and apoptosis. Unraveling the exact molecular mechanism(s) of lipotoxicity may lead to new strategies for the management of NAFLD.
Cell culture and treatment
Human hepatocellular HepG2 (ATCC, HB-8065) and Huh7 (Health Science Research Resources Bank, JCRB0403, Osaka, Japan) carcinoma cells were grown in DMEM containing 10% heat-inactivated fetal bovine serum, 2 mm glutamine, 100 U·mL−1 penicillin, and 100 ng·mL−1 streptomycin, at 37 °C in air with 5% CO2. Cells were seeded and left under normal conditions for 24 h before any further treatment. Stock solutions of FFAs (100 mm) were prepared in isopropanol by heating to 70 °C, and the desired concentrations were prepared in growth medium supplemented with BSA, as described previously . Briefly, a 5% (w/v) BSA solution in DMEM was filtered and mixed with the fatty acid stock solution, giving a concentration of 5 mm FFA (FFA/BSA molar ratio of 6.6 : 1). The solution was left for 30 min at 50 °C, and diluted with DMEM, giving the desired concentrations.
Estimation of cell viability
Following FFA treatment, cell numbers were assessed by Trypan blue exclusion. Floating and attached cells were collected, centrifuged at 200 g for 5 min, and resuspended in DMEM plus Trypan blue (0.4% in NaCl/Pi) at a ratio of 1 : 1. Viable cells were counted with a hemocytometer, and cell numbers were expressed as percentages of the respective control, unless otherwise indicated.
Estimation of DNA synthesis
Cells were seeded in 24-well plates onto 11-mm glass coverslips at a density of 5 × 104 cells per well. After 24 h, cells were supplemented with media containing the indicated concentrations of FFA for 24 and 48 h. In the final 8 h, cells were supplemented with 100 μm BrdU (Sigma, St. Louis, MO, USA) and analyzed by indirect immunofluorescence. Briefly, cells were fixed with 3.7% paraformaldehyde, quenched with 50 mm ammonium chloride for 15 min, and permeabilized with 0.1% Triton X-100 for 4 min, before being treated with 1.5 m hydrochloric acid for 10 min. Incorporation of BrdU into newly synthesized DNA was detected with an antibody against BrdU (Sigma), and analysis was performed with a Leica TCS-SP scanning confocal microscope. Cell nuclei were detected by Hoechst 33342 staining (Sigma). More than 300 cells per sample were counted, and BrdU-positive cells were expressed as a percentage of the total cell number.
Detection of lipid accumulation
Lipid droplet accumulation was detected by Nile red staining as previously described . Cell imaging for Nile red staining was performed by confocal microscopy. Quantification of lipid droplets was performed by flow cytometric analysis of the distribution of Nile red fluorescence in individual cells. Briefly, cells were seeded in 24-well plates onto 11-mm glass coverslips for confocal microscopy, or in six-well plates for flow cytometry. After FFA treatment, cells were fixed with 3.7% paraformaldehyde for 10 min, washed twice, and stained with Nile red (Sigma) solution (final concentration, 2 μg·mL−1). Samples were kept for 45 min in the dark at 37 °C to allow equilibrium with the dye. Coverslips were mounted in Mowiol, and viewed with a Leica TCS-SP scanning confocal microscope, equipped with an argon/krypton laser and Leica TCS software. Flow cytometric analysis (15 000 events per sample) was carried out with a CyFlow ML (Partec) equipped with an argon laser, in the FL1 channel (logarithmic scale).
Determination of TAG levels
Cells seeded in six-well plates were treated with FFA medium for 24 h. Cells were harvested, and lipids were extracted twice with CHCl3/MeOH/ddH2O (1 : 1 : 0.9). Lipid extracts were dried under a nitrogen stream, redissolved in chloroform/methanol (2 : 1), and separated by TLC (silica gel/TLC; Fluka), with hexane/diethyl ether/acetic acid (70 : 30 : 1). For visualization of lipids, the TLC plates were immersed in a solution of cupric sulfate (10%, w/v) in aqueous phosphoric acid (8%, v/v), allowed to dry at room temperature, and charred at 150 °C for 10 min. TAGs were identified with glyceryl tripalmitate standard (Sigma).
Estimation of apoptosis
For nuclear morphological observations, cells were fixed with 3.7% paraformaldehyde, neutralized with 50 mm NH4Cl, and stained with Hoechst 33342. Nuclear morphology was observed under a fluorescence microscope (Axiovert S 100; Zeiss, Ontario, NY, USA) equipped with a UV filter.
Flow cytometric analysis of cellular hypodiploid sub-G1 DNA content was performed according to . Cells were harvested, fixed with 70% ethanol, resuspended in NaCl/Pi for 20 min at 37 °C, and stained with Hoechst 33342 (30 min, 37 °C). Annexin V–fluorescein isothiocyanate (FITC) (BD Pharmigen, San Diego, CA, USA) binding and PI staining were performed according to . Following incubation with 1.0 μg·mL−1 annexin V and 4.0 μg·mL−1 PI for 20 min, cells were analyzed by flow cytometry.
Preparation and analysis of DNA and protein extracts
After the appropriate treatment, cellular DNA was isolated from 3 × 106 cells per sample, and analyzed for internucleolar fragmentation by agarose gel electrophoresis. Preparation of mitochondrial and cytosolic fractions was achieved by differential centrifugation, as described previously . For western blot analysis, cell lysates (40–50 μg of protein) were subjected to SDS/PAGE, and the separated proteins were demonstrated by immunoblotting after being transferred to nitrocellulose membranes. Antibodies against cytochrome c (sc-13156), Bax (2D2, sc-20067), Bcl-2(100) (sc-509), ATF4 (sc-200) and CHOP/GADD153 (R-20, sc-793) were from Santa Cruz Biotechnology. Antibodies against phospho-JNK (#9251), total JNK (#9252) and phospho-eIF2α (#9721) were from Cell Signaling. Horseradish peroxidase-conjugated antibody against caspase-3 (#610325) was from BD Pharmigen, and antibody against β-actin (A5441) was from Sigma.
All data are expressed as the mean ± standard deviation (SD). Differences between groups were compared by one-way ANOVA followed by a post hoc Bonferroni correction test for multiple comparisons, using originpro 8 software (OriginLab). Differences were considered to be statistically significant at P < 0.05.