Busulfan induces steatosis in HepaRG cells but not in primary human hepatocytes: Possible explanations and implication for the prediction of drug-induced liver injury

1 INTRODUCTION

Busulfan (1,4-butanediol dimethanesulfonate) is an alkylating drug used for the treatment of chronic myelogenous leukemia as well as a myeloablative agent administrated prior to hematopoietic stem cell transplantation. The usual protocols of busulfan treatment comprise 2 to 16 intravenous administrations [1]. Once administrated, this anticancer drug spontaneously hydrolyzes leading to the formation of a reactive carbonium ion that alkylates DNA, thus causing cell death in rapidly dividing cells (Figure 1) [2, 3]. This carbonium ion is further hydrolyzed to tetrahydrofuran (THF) and methanesulfonic acid (MSA) (Figure 1) [2]. Busulfan can also be conjugated to reduced glutathione (GSH) by different glutathione S-transferase (GST) isoenzymes with GSTA1 playing a predominant role, whereas GSTM1, GSTP1, and GSTT1 appear to be involved only to a lesser extent (Figure 1) [2, 5, 6]. In keeping with this report, other studies showed that GSTA1 polymorphisms significantly affect busulfan clearance in treated patients [7-9].

Busulfan can induce various adverse effects involving different tissues and organs such as bone marrow, gastrointestinal tract, liver, lung, skin, bladder, central nervous system, and ovary [3, 10]. Regarding the liver, busulfan can induce different types of hepatic lesions including intrahepatic cholestasis, hepatocellular necrosis, nodular regenerative hyperplasia, and sinusoidal obstruction syndrome (SOS), which can be severe and even fatal [11, 12]. However, to the best of our knowledge, steatosis (i.e., lipid accumulation) has never been reported in patients. In rodents, investigations in mice treated with 33- and 40-mg/kg/day busulfan reported diffuse microvesicular steatosis, which was dose-dependent [13]. Nevertheless, as underlined by the authors, this liver lesion might not be specific because severe body weight loss was concomitantly observed in treated mice [13]. Hence, busulfan could be classified as a non-steatogenic drug, at least in humans. Alternatively, busulfan-induced steatosis might occur only in a few patients and/or might be mostly asymptomatic.

The mechanisms of busulfan-induced hepatotoxicity are still poorly understood. Investigations in cultured mouse hepatocytes showed that busulfan-induced cytotoxicity (as assessed by LDH release) required GST activity, thus suggesting that the sulfonium cation conjugate might be toxic (Figure 1) [4]. This study also showed that busulfan induced oxidative stress as reflected by significant GSH depletion and increased levels of oxidized glutathione (GSSG), which were exacerbated when hepatocytes were cultured in a medium without sulfur amino acids [4]. Busulfan-induced oxidative stress was also characterized by lipid peroxidation as assessed by increased levels of thiobarbituric acid reactive substances (TBARS) [4]. In keeping with these in vitro results, investigations in mice showed that repeated administration of 33-mg/kg busulfan induced a significant decrease in hepatic GSH levels [13]. Finally, earlier investigations in cultured rat hepatocytes reported that busulfan induced DNA damage [14].

In a recent study performed in HepaRG cells and primary human hepatocytes (PHH), we studied 12 different drugs reported to be steatogenic in treated patients [15]. Among these drugs, seven were found to induce a significant accumulation of neutral lipids in both differentiated HepaRG cells and PHH [15]. Further experiments in HepaRG cells also disclosed that in the absence of severe mitochondrial dysfunction, these seven drugs can induce steatosis by different mechanisms including impairment of mitochondrial fatty acid oxidation (FAO), increased de novo lipogenesis (DNL), inhibition of very low-density lipoprotein (VLDL) secretion, and possibly enhanced fatty acid uptake [15]. These investigations and others showed that the HepaRG cell line could be a suitable model to investigate drug-induced steatosis and related mechanisms [15-18].

In parallel to these investigations, we also treated HepaRG cells with three drugs that have never been reported to induce steatosis in patients, namely, busulfan, amoxicillin, and glimepiride [11, 12]. To our surprise, whereas glimepiride and amoxicillin caused little or no steatosis in HepaRG cells, busulfan induced a marked concentration-dependent increase in neutral lipids. In sharp contrast, this drug did not induce steatosis in six different batches of PHH. Hence, we decided to perform further experiments in differentiated HepaRG cells to get more insight into the mechanism(s) whereby this antineoplastic agent caused lipid accumulation in these cells. We also examined transcriptomic databases in order to understand why busulfan could be steatogenic in HepaRG cells but not in PHH. Our study highlights the need to use PHH to evaluate the ability of drugs and other xenobiotics to induce steatosis whenever their steatogenic potential in humans is still unknown.

2 MATERIALS AND METHODS

3 RESULTS

In our previous study, drug-induced steatosis in differentiated HepaRG cells was assessed with drug concentrations that did not induce severe mitochondrial dysfunction [15]. This allowed us to disclose that drug-induced lipid accumulation could be induced by at least four mechanisms, namely, moderate impairment of mitochondrial FAO, increased DNL, inhibition of VLDL secretion, and possibly enhanced fatty acid uptake [15]. Noteworthy, severe mitochondrial dysfunction was defined as a loss of cellular ATP levels in treated HepaRG cells greater than 30% of the control cells [15], in accordance with previous investigations [26-28]. Hence, in the present study, we first assessed the effect of different concentrations of busulfan on ATP levels in HepaRG cells. Our results showed that busulfan concentrations above 750 μM led to a concentration-dependent loss of cellular ATP levels. Indeed, ATP levels were reduced by 23%, 53%, 75%, and 99% in HepaRG cells treated with 1000, 1250, 1500, and 2000 μM of busulfan, respectively (Figure 2a). Hence, a maximum concentration of 1000 μM was used for subsequent experiments in order to avoid severe mitochondrial dysfunction.

Next, we evaluated the effects of busulfan on neutral lipids in HepaRG cells (Figure 2b). Neutral lipid level was significantly enhanced by 29%, 41%, and 59% in HepaRG cells treated with 500, 750, and 1000 μM, respectively (Figure 2c). In sharp contrast, busulfan did not induce steatosis in six different PHH batches (Figure 3). This drug even tended to lower neutral lipids in some PHH batches, as for PHH 2, 3, and 5 (Figure 3). Although we did not assess busulfan-induced cytotoxicity in these PHH batches, recent investigations in three different PHH batches showed that busulfan induced significant cytotoxicity (assessed with the CellTiter-Blue assay) with a median EC₅₀ greater than 1000 μM for 1 and 2 days of treatment and equal to 210 μM after 7 days [29]. Hence, these data suggested that the absence of steatosis in PHH treated by busulfan could not be attributed to a lack of responsiveness of these cells to the antineoplastic drug.

In order to determine the mechanisms whereby busulfan could induce steatosis in HepaRG cells, fatty acid uptake, mitochondrial FAO, DNL, and VLDL secretion were investigated by measuring the corresponding metabolic fluxes. Fatty acid uptake was significantly enhanced by 92%, 85%, 84%, and 69% with 250, 500, 750, and 1000 μM of busulfan, respectively (Figure 4a). Mitochondrial FAO was significantly increased by 10%, 21%, and 37% with 500, 750, and 1000 μM of busulfan, respectively (Figure 4b), whereas DNL was significantly decreased by 54%, 70%, 75%, and 81%, with 250, 500, 750, and 1000 μM of this drug, respectively (Figure 4c). Finally, apoB and apoC3 secretion in the culture medium was strongly reduced by busulfan (Figure 4d), thus indicating a severe impairment of VLDL secretion [15]. Hence, our results suggested that busulfan-induced steatosis could be secondary to a dual mechanism, namely, increased fatty acid uptake and reduced VLDL secretion. Of note, increased mitochondrial FAO and reduced DNL fluxes in the face of drug-induced steatosis in HepaRG cells might represent an adaptation in order to limit lipid accumulation [15]. In keeping with this hypothesis, the lack of significant increase in neutral lipids induced by 250-μM busulfan (Figure 2b) might be due to the presence of a strong reduction of DNL that was concomitant to increased fatty acid uptake and reduced apoB secretion (Figure 4). Nonetheless, further investigations at different time points might be of interest to confirm whether higher mitochondrial FAO and lower DNL fluxes represent adaptive mechanisms.

By measuring gene expression, we found that FAT/CD36 mRNA levels were markedly increased by 5.5- and 8.2-fold with 750- and 1000-μM busulfan, respectively (Figure 5a), thus suggesting that FAT/CD36 upregulation might be involved in the stimulation of fatty acid uptake induced by busulfan (Figure 4a). Next, we measured the mRNA expression of several key proteins (APOB, APOC3, ANGPTL3) and enzymes (MTTP, P4HB) involved in VLDL formation and secretion [15]. Notably, we found a significant and sometimes profound reduction in the mRNA levels of APOB (coding for apolipoprotein B), APOC3 (coding for apolipoprotein C3), MTTP (coding for microsomal triglyceride transfer protein), and ANGPTL3 (coding for angiopoietin-like 3) (Figure 5b), whereas the expression of P4HB (coding for prolyl 4-hydroxylase subunit beta) was unchanged (data not shown).

Since our recent investigations unveiled that drug-induced impairment of VLDL secretion was always associated with endoplasmic reticulum (ER) stress [15], we assessed the mRNA levels of five different proteins involved in ER stress-induced unfolded protein response (UPR), namely, heat shock protein family A member 5 (HSPA5, also known as BIP), DNA damage-inducible transcript 3 (DDIT3, also known as CHOP), endoplasmic reticulum to nucleus signaling 1 (ERN1, also known as IRE1α), eukaryotic translation initiation factor 2 alpha kinase 3 (EIF2AK3, also known as PERK), and activating transcription factor 6 (ATF6). Interestingly, the mRNA abundance of BIP, CHOP, IRE1α, and PERK was significantly enhanced by 750 and 1000 μM of busulfan, albeit with different magnitudes (Figure 5c). In contrast, ATF6 expression was not significantly increased (data not shown).

Next, we wished to understand why busulfan induced steatosis in HepaRG cells but not in PHH. Because previous investigations in mouse hepatocytes suggested that busulfan-induced toxicity might be linked to GST-catalyzed formation of the sulfonium cation conjugate [4], we compared the expression of GSTA1, GSTM1, GSTP1, and GSTT1 between HepaRG cells and 12 different batches of PHH by using two different transcriptomic datasets, namely, GSE25844 [23] and GSE41289 [24]. Notably, investigations in HepaRG cells used to generate these datasets were performed by investigators from our laboratory with the same native HepaRG cells used in the present study. GSTA1 expression was much higher in HepaRG cells than in PHH, while GSTP1 expression was much lower in HepaRG cells than in PHH (Table 3). GSTT1 expression was lower in HepaRG cells than in PHH, but the difference was significant only in the GSE41289 dataset (Table 3). Finally, results for GSTM1 expression showed opposite results between both datasets (Table 3). Hence, busulfan-induced steatosis in HepaRG cells might be related to high GSTA1 expression, which in turn could lead to the formation of high level of sulfonium cation conjugate.

In order to test this hypothesis, levels of reduced GSH were measured in HepaRG cells treated with different concentrations of busulfan. Interestingly, 1000-μM busulfan significantly decreased the levels of reduced GSH by 25%, whereas lower concentrations were without effect (Figure 6a). We also measured reduced GSH levels in cells treated with 1000-μM busulfan in the absence or presence of ethacrynic acid (ETA), a diuretic drug known to inhibit different human GST isoenzymes, in particular GST of the alpha-class [30, 31]. Importantly, ETA almost fully prevented the decrease in reduced GSH levels induced by 1000-μM busulfan (Figure 6b), thus suggesting that this reduction could be secondary to the formation of busulfan GSH-conjugate. In addition, the chemical chaperone tauroursodeoxycholic acid (TUDCA) partially but significantly alleviated busulfan-induced lowering of reduced GSH (Figure 6b).

We next determined the potential protective effect of ETA on busulfan-induced lipid accumulation. Furthermore, since busulfan appeared to induce ER stress, we wished to determine whether TUDCA could protect against busulfan-induced steatosis as it did for the pharmaceuticals allopurinol, indomethacin, and rifampicin and for the prototypical ER stress inducer tunicamycin [15]. Interestingly, both ETA and TUDCA alleviated lipid accumulation induced by 1000-μM busulfan (Figure 7a).

We also tested the ability of ETA and TUDCA to restore the reduction of VLDL secretion and the increase in fatty acid uptake stimulation induced by 1000-μM busulfan. Only TUDCA partially but significantly alleviated ApoB secretion (Figure 7b), whereas ETA tended (P = 0.09) to reduce fatty acid uptake (Figure 7c). Of note, both ETA and TUDCA alleviated busulfan-induced cytotoxicity (as assessed with ATP), although the protective effect was significant only for ETA (Figure 7d).

4 DISCUSSION

In this study, the anticancer busulfan induced a marked accumulation of neutral lipids in HepaRG cells, while no steatosis was observed in six different batches of PHH. Indeed, busulfan caused significant neutral lipid accumulation in HepaRG cells in a concentration-dependent manner from 500 to 1000 μM, while the same concentrations in PHH had no effect or tended to decrease intracellular lipids. Of note, the lack of steatosis in PHH is in line with the fact that busulfan-induced steatosis has not been reported so far in treated patients [11, 12]. However, we cannot exclude the possibility that more PHH batches would have been necessary to uncover the possible steatogenic potential of this antineoplastic agent. Hence, we wished to determine the mechanism(s) whereby busulfan induced lipid accumulation in HepaRG cells and understand why this drug was not steatogenic in PHH.

Busulfan-induced steatosis in HepaRG cells was observed for concentrations ranging from 500 to 1000 μM. Importantly, the upper values of the maximum plasma concentrations (C_max) of busulfan in patients were found to be around 15 μM (i.e., 3.7 μg/L) [32, 33]. Hence, 1000-μM busulfan is below 100 × C_max, a threshold classically used in toxicological studies pertaining to drug-induced liver injury [15, 34, 35]. Of note, concentrations of busulfan between 625 and 2500 μM were previously used to study the mechanisms of its toxicity in cultured murine hepatocytes [4].

The initial goal of this study was to determine the mechanism(s) whereby busulfan caused steatosis in HepaRG cells. Our data suggested that a first mechanism of busulfan-induced steatosis could be an impairment of VLDL secretion (Figure 8), as reflected by reduced levels of both apoB and apoC3 in the cultured medium. Impairment of VLDL secretion might be linked to reduced mRNA levels of different proteins and enzymes playing a significant role in VLDL assembly, particularly APOB, APOC3, MTTP, and ANGPTL3, although we cannot rule out other mechanisms such as direct apoB protein degradation [15]. In addition, reduced VLDL secretion was associated with enhanced expression of four proteins involved in ER stress-induced UPR, namely, HSPA5 (BIP), DDIT3 (CHOP), ERN1 (IRE1α), and EIF2AK3 (PERK), thus extending our recent data with seven steatogenic drugs showing that impairment of VLDL secretion in HepaRG cells was always associated with ER stress [15]. ER stress could alter VLDL secretion by different mechanisms such as specific degradation of mRNA of key proteins involved in VLDL assembly through regulated IRE1α-dependent decay (RIDD), decreased MTTP expression and activity, and apoB degradation via both proteasomal and non-proteasomal pathways [15, 36-38]. Importantly, the chemical chaperone TUDCA significantly alleviated busulfan-induced steatosis and partially restored apoB secretion, thus indicating that ER stress could be involved, at least in part, in VLDL secretion impairment and triglyceride accumulation (Figure 8). Of note, previous investigations in different experimental models showed that reducing ER stress in hepatocytes can alleviate intracellular lipid accumulation [39-41].

Although ETA alleviated steatosis, this GST inhibitor did not restore VLDL secretion. Hence, we surmise that busulfan-induced steatosis might involve a second mechanism, possibly increased uptake of fatty acids (Figure 8). In keeping with this hypothesis, busulfan markedly enhanced fatty acid uptake, possibly via upregulation of FAT/CD36 expression. However, although several investigations reported a close relationship between ER stress and high FAT/CD36 expression [42-44], TUDCA did not significantly reduce fatty acid uptake. Several reasons might explain this discrepancy: (a) ER stress might have a limited impact on the kinetics of fatty acid uptake; (b) the protocol of TUDCA exposure might not have been optimal to prevent ER stress-induced stimulation of fatty acid uptake; (c) TUDCA could have alleviated adverse events other than ER stress that might also play a role in busulfan-induced steatosis. For instance, TUDCA was shown to afford protection against mitochondrial dysfunction and oxidative stress and to activate several intracellular signaling pathways secondary to its binding to membrane-bound integrins [45-48].

In addition to reduced triglyceride secretion and increased fatty acid uptake, other mechanisms of busulfan-induced steatosis cannot be excluded. For instance, it would be interesting to determine whether busulfan could inhibit lipase-mediated lipid droplet lipolysis and/or lipophagy (i.e., a specific type of autophagy targeting lipid droplets), which would favor lipid accumulation [49-51] (Figure 8). Interestingly, previous investigations suggested that in some circumstances, ER stress can impair lipophagy/autophagy, thus contributing to fatty liver but also to hepatocyte death [40, 52, 53].

Of note, previous investigations showed that busulfan induced ER stress in mouse testis and spermatogonial stem cells [54, 55], thus suggesting that this antineoplastic drug can trigger ER stress in different tissues. Moreover, busulfan was shown to induce oxidative stress in murine hepatocytes, as assessed by increased GSSG levels and lipid peroxidation [4]. Because there are possible interrelationships between oxidative stress and ER stress [56, 57], further investigations would be needed in order to determine whether busulfan-induced ER stress is secondary to oxidative stress or vice versa.

The second goal of the present study was to understand why busulfan triggered steatosis in HepaRG cells but was not steatogenic in PHH. Our working hypothesis was that high GST expression in HepaRG cells might generate the sulfonium cation conjugate, which was deemed to be the toxic busulfan metabolite in murine hepatocytes (Figure 1) [4]. To this end, we first compared the mRNA expression of GSTA1, GSTM1, GSTP1, and GSTT1 between HepaRG cells and PHH. We found that GSTA1, which plays a predominant role in busulfan conjugation [2, 5, 6], was much more expressed in HepaRG cells than in PHH (Table 3). Importantly, a global proteomic analysis using liquid chromatography–tandem mass spectrometry (LC–MS/MS) previously reported a higher abundance of the GSTA1 protein in HepaRG cells compared with cryopreserved PHH [58]. Hence, HepaRG cells might be able to accumulate more sulfonium cation conjugate than PHH, thus triggering ER stress, impairment of lipid metabolism, and steatosis (Figure 8). In line with this hypothesis, GST inhibition with ETA significantly alleviated busulfan-induced decrease in reduced GSH levels and steatosis, although this was not associated with a restoration of apoB secretion. As previously mentioned, this result suggests that impairment of VLDL secretion might not be the primary mechanism of lipid accumulation induced by busulfan (Figure 8). It is also noteworthy that ETA was able to prevent busulfan-induced cytotoxicity in HepaRG cells. Hence, ER stress might also be involved, at least in part, in busulfan-induced cytolysis (Figure 8). In keeping with this assumption, TUDCA partially restored busulfan-induced cytolysis and this was associated with a partial restoration of GSH levels. This suggested that ER stress might also be involved in GSH level reduction, as previously reported [57, 59]. Thus, busulfan-induced decrease in reduced GSH levels could be secondary to its consumption for the generation of the sulfonium cation conjugate and also to ER stress. The lack of GSH level reduction in HepaRG cells treated with 750-μM busulfan (that induced steatosis and ER stress) might be secondary to a compensatory enhancement of GSH synthesis, as previously described under some conditions of cellular stress [60, 61].

Our results showed that busulfan did not induce steatosis in PHH. Moreover, this anticancer agent even tended to decrease neutral lipids in some PHH batches. Because busulfan increased mitochondrial FAO and reduced DNL in HepaRG cells these metabolic changes could also occur in these PHH batches, thus leading to fewer intracellular lipids. However, this hypothesis does not exclude the possibility that higher mitochondrial FAO and lower DNL also participate in the adaptive response aiming to limit neutral lipid accretion in HepaRG cells, as already mentioned.

As mentioned above, busulfan-induced steatosis has never been reported so far in treated patients [11, 12]. However, this does not mean that macrovacuolar steatosis never occurred in some patients. Indeed, this liver lesion can develop without clinical and biological signs, and thus, it can be silent for years [62, 63]. Furthermore, busulfan-induced steatosis might occur only in patients with high hepatic GSTA1 expression, as the one found in the female patient whose liver tumor was used for the isolation of the HepaRG cell line [19]. Notably, various factors can significantly modulate GSTA1 expression including genetic polymorphism [64], hormones [65], and cytokines [66]. Whether high GSTA1 expression is uncommon in the general population would require further investigations. It is also worth mentioning that different xenobiotics including drugs, environmental toxins, and plant compounds can induce the hepatic expression and activity of GSTs [67-69]. Hence, patients exposed to these xenobiotics might be at risk of busulfan-induced hepatotoxicity.

Previous studies showed that GST-mediated biotransformation of some xenobiotics can generate glutathione conjugates that can be cytotoxic [70, 71]. For instance, this has been shown with dibromomethane, perfluoropropene, hexachlorobutadiene, trichloroethene, and bromobenzene [70, 71]. Hence, for some molecules including busulfan, GST-catalyzed GSH conjugation is associated with bioactivation rather than detoxication.

Notably, busulfan-induced cellular effects in HepaRG cells were quite similar to those induced by tunicamycin and thapsigargin. Indeed, like both prototypical ER stress inducers [15], busulfan induced ER stress, impairment of VLDL secretion, and steatosis. Moreover, busulfan increased mitochondrial FAO and reduced DNL, similar to tunicamycin and thapsigargin [15]. Further investigations are warranted to determine whether sulfonium cation conjugate (i.e., busulfan GSH conjugate) impairs the glycosylation of newly synthesized proteins like tunicamycin and/or inhibits the sarcoplasmic/endoplasmic reticulum calcium ATPase (SERCA) like thapsigargin [56].

Previous investigations showed that the HepaRG cell line is an invaluable tool to investigate drug-induced steatosis [15-18] and other types of hepatotoxicity such as cholestasis and hepatocyte cell death [72-75]. Nevertheless, the present study suggests that this might not be always an appropriate model to predict drug-induced hepatotoxicity. In addition, this cell line could identify molecular mechanisms that are potentially not relevant in the clinic, or that might occur only rarely in the general population. Thus, we strongly recommend carrying out in vitro toxicological investigations in both HepaRG cells and PHH in order to avoid drawing wrong conclusions on the potential hepatoxicity of xenobiotics including drugs. Nevertheless, while both types of cells can be used to search for different patterns of hepatotoxicity such as steatosis, cytolysis, and cholestasis, HepaRG cells are better suited for mechanistic investigations than PHH, whose availability is increasingly limited [76, 77]. Finally, our study highlights that novel in vitro tools should require thorough characterization in order to determine the potential artefactual expression of some proteins or the presence of metabolic pathways that could differ from the in vivo situation.

AUTHOR CONTRIBUTIONS

Bernard Fromenty drafted the original manuscript. Julien Allard, Simon Bucher, Pierre-Jean Ferron, and Youenn Launay performed the experiments and critically amended the manuscript. All the authors have read and approved the final version of the manuscript.

CONFLICT OF INTEREST STATEMENT

J. Allard, S. Bucher, P-J. Ferron, Y. Launay, and B. Fromenty declare that they have no conflict of interest in relation to this work.