Nelfinavir Induces Necrosis of 3T3F44-2A Adipocytes by Oxidative Stress

Protease inhibitor treatment strongly diminishes mortality in HIV-infected patients. This treatment has also been associated with lipodystrophy and has been shown to alter adipocyte differentiation. The protease inhibitor nelfinavir has been indirectly implicated in the appearance and development of lipodystrophic syndrome, as well as in adipocyte cell death. The aim of this study was to evaluate the effects of nelfinavir on the 3T3-F442A adipocyte cell line. Nelfinavir (30 μM) induced cell death of 3T3-F442A adipocytes by a necrotic process that was not mediated by TNF-α. Treatment of cells with this protease inhibitor led to a significant increase in expression of the heme oxygenase-1 gene that could be reduced by 100 μM of the antioxidant ascorbate. Moreover, ascorbate had a protective effect on nelfinavir-induced necrosis, decreasing the percentage of necrotic cells by 70%. Our results show that nelfinavir induces necrosis of adipocytes mediated by a cellular increase of reactive oxygen species. This deleterious effect could be counterbalanced by ascorbate.

Current antiretroviral treatment of HIV replication in infected patients generally combines nucleoside reverse transcriptase inhibitors with protease inhibitors (PIs). Thus, highly active antiretroviral treatment has dramatically improved the prognosis of patients by drastically suppressing HIV load, increasing CD4 cell counts, and reducing opportunistic infections associated with AIDS.^1-3 Nevertheless, it is now recognized that ∼80% of patients receiving highly active antiretroviral treatment develop metabolic abnormalities, which include alterations of glucose and lipid metabolism,^4-6 insulin resistance, and, sometimes, diabetes mellitus.⁷ Clinical symptoms generally include peripheral lipoatrophy, with a marked loss of facial and limb adipose tissue,⁸ as well as central adiposity.^9,10 The etiology of this lipodystrophic syndrome is not completely understood at the present time, although it is thought to be linked to a combination of different classes of drugs.¹¹ However, the involvement of PIs has been strongly suspected since the first description of this syndrome¹² and indeed has subsequently been widely documented. These studies have implicated 2 PIs in particular, indinavir and nelfinavir that are currently administered during highly active antiretroviral treatment.

Investigation of the deleterious effects of PIs in patients receiving several classes of antiviral therapy is difficult to assess and thus necessitates the use of in vitro models. Previous in vitro studies, using both human and murine preadipocytes, have demonstrated an inhibition of preadipocyte to adipocyte differentiation by PIs.^11,13-16 Avery thorough study by Caron et al¹⁷ of the murine cell line 3T3-F442A demonstrated that indinavir inhibits preadipocyte differentiation and induces insulin resistance by acting upon nuclear localization of SREBP-1. Another study by Dowell et al¹⁶ also indicated down-modulation of the nuclear cleaved form of SREBP-1 by nelfinavir. These results were also confirmed by a recent study on hepatic cells.¹⁸ Thus, nelfinavir has been indirectly implicated in the development of lipodystrophy.¹⁹ Moreover, several reports have shown a major lipolytic effect of nelfinavir,^20-23 as well as its inhibitory influence on lipogenesis.^16,23 Finally, Dowell et al¹⁶ demonstrated the marked cell death of mature adipocytes (3T3-L1) upon incubation with nelfinavir, providing a possible explanation for the development of some lipodystrophies observed during highly active antiretroviral treatment.

This study was undertaken with the aim of extending the above-mentioned findings, particularly with respect to the role played by nelfinavir in cell death.

MATERIALS AND METHODS

Materials

Cell culture sera were from Calbiochem (Meudon, France). Trizol, cell culture medium, and enzymes were from Invitrogen (Cergy-Pontoise, France). Nelfinavir and indinavir were kindly provided by Roche (Meylan, France) and Merck (Darmstadt, Germany) Laboratories, respectively, and were resuspended in ethanol.

Cell Culture and Treatment

All experiments were conducted on the 3T3-F442A cell line. Cells with passage numbers 11 to 16 were used in all studies. During the growth phase, the fibroblast monolayer was maintained in Dulbecco modified Eagle medium with 100 μg/L penicillin/streptomycin, 800 μg/L biotin-pantothenate (medium 1), and 10% newborn calf serum at 37°C in 5% CO₂. When cell confluence reached 70%, the medium was replaced by medium 1 supplemented with 5% newborn calf serum and 5% fetal calf serum. At confluence (day 0), the medium was replaced by medium 1 plus 10% fetal calf serum and insulin (100 μg/L). All experiments were performed 7 days after confluence. The incubation medium was changed every 2 days. Because the compounds used in the study were solubilized in ethanol, treated and control cells were cultured in the presence of 0.1% ethanol.

Apoptosis and Necrosis Detection

Postconfluence (day 7) adipocytes were incubated without (control) or with different compounds (nelfinavir, N-acetylcysteine [Sigma], ascorbate [Sigma], nelfinavir + N-acetylcysteine, and nelfinavir + ascorbate). The medium was deprived of fetal calf serum 12 hours before PI treatment. At the times indicated (6, 12, 18, and 36 hours, Fig. 1A; and 24 hours, Fig. 4), the cell monolayer was washed with phosphate-buffered saline and trypsinized. Apoptotic and necrotic cells were detected using the Annexin V-FLUOS Staining Kit (Roche) as per the manufacturer's instructions. Cells were then analyzed by flow cytometry with a Facscan flow cytometer (Becton Dickinson) using an excitation wavelength of 488 nm. The percentage of necrotic versus apoptotic cells was determined by propidium iodide staining versus Annexin V/propidium iodide staining, respectively.

Figure 1

Figure 4

RNA Extraction

On day 7 and at the times indicated (9, 12, 15, 18, and 24 hours), cells were rapidly washed twice with phosphate-buffered saline, lysed by adding 1 mL of Trizol reagent, scraped from the plates, and transferred to Eppendorf tubes. Chloroform (500 μL; Merck) was added, which was followed by centrifugation for 15 minutes at 10,000 g. Superior layers were carefully removed and then mixed with 200 μL of isopropanol (Merck). RNAwas collected by centrifugation for 10 minutes at 9000 g, washed twice with 75% ethanol, and treated with DNase (10 U per assay) for 30 minutes. RNA was extracted once more by using a phenol/chloroform/isoamyl alcohol mixture (25:24:1) and then rinsed twice with 75% ethanol. RNA was recovered in diethylpyrocarbonate-treated water and quantified by spectrophotometry at 260 nm (any aliquot with an A₂₆₀:A₂₈₀ ratio of <1.8 was discarded).

Real-Time Quantitative RT-PCR Analysis

Total RNA (1 μg per assay) was then submitted to nonpreferential reverse transcription (200 U RT) in the presence of random hexanucleotides. Reverse transcribed heme oxygenase-1 (HO-1) mRNA was amplified on an ABI PRISM 7700 thermal cycler (Applied Biosystems, Courtaboeuf, France) using the SYBR green fluorescence method. Primers used are indicated in Table 1. HO-1 amplification was checked for efficiency equal to that of the 18S rRNA gene used as a reference. Real-time amplifications were then analyzed using Sequence Detector software (Applied Biosystems), and quantification of target mRNA was carried out by comparison of the number of cycles required to reach reference and target threshold values (ΔΔCT method).

Table 1

Cell Extract Preparation

Nuclear, cytoplasmic, and membranous cell extracts were prepared at the indicated times by washing cell monolayers once in phosphate-buffered saline. Cells were scraped, lysed in 10 mM Hepes (pH 7.9), 10 mM KCl, 1 mM dithiothreitol, PI cocktail (Sigma), 0.5% Igepal, and incubated on ice for 15 minutes with intermittent vortexing. The mixture was centrifuged for 5 minutes at 2000 g, and the supernatant was collected (cytoplasmic and membranous extract). The pellet was lysed in 10 mM Hepes (pH 7.9), 420 mM NaCl, 1 mM dithiothreitol, PI cocktail, and incubated on ice for 30 minutes with intermittent vortexing. The mixture was centrifuged for 10 minutes at 10,000 g, and the supernatant was collected (nuclear extract). Protein concentrations of all samples were determined using the Bradford assay (BioRad).

Immunoblot Assay

Cytoplasmic and membranous protein extract (100 μg per lane) was subjected to SDS-polyacrylamide gel electrophoresis (12% acrylamide). After transferring to a nitrocellulose membrane (Hybond ECL, Amersham Biosciences), the blot was probed with rabbit antimouse TNF-α antibody (Abcam, Ltd., Cambridge, UK). Protein detection was performed by ECL plus Western blotting detection reagent as per the manufacturer's instructions (Amersham Biosciences).

Electrophoretic Mobility Shift Assays

Labeling of the neosynthesized double-stranded oligonucleotides was performed with Escherichia coli DNA polymerase (Klenow fragment) in the presence of [α-³²P]dCTP. Sequences of the consensus p53 and NF B oligonucleotides are indicated in Table 1 (lowercase nucleotides represent those that were used as templates for radiolabeling with E. coli DNA polymerase 1). In all experiments, the amount of labeled oligonucleotides used was 1 ng (∼20 × 10³ cpm), and in the negative control, the amount of added unlabeled oligonucleotide was 100-fold greater than that of the labeled oligonucleotide. EMSAs were performed as previously described.²⁴ Briefly, 7.5 μg of protein from nuclear extracts was preincubated at 25°C for 30 minutes in 20 mM Tris-HCl, (pH 7.9), 5 mM MgCl₂, 0.5 mM EDTA, 20% (vol/vol) glycerol, and 0.5 mM dithiothreitol. The addition of ³²P-labeled oligonucleotides was followed by an additional 30-minute incubation at 25°C in a total volume of 50 μL. The oligonucleotide-protein complexes were then resolved using a nondenaturing 6% (wt/vol) polyacrylamide gel. Electrophoresis was carried out at 100 V for 4-6 hours in 0.09 M Tris, 0.009 M borate, 2 mM EDTA, pH 8.3. Gels were then dried and underwent autoradiography.

Preparation of Microsomes

On day 7 and 24 hours after treatments, cells were rinsed with ice-cold phosphate-buffered saline and scraped on ice in phosphate-buffered saline containing 1 mM EDTA. All subsequent extraction and centrifugation procedures were performed at 4°C. Cells were pelleted by centrifugation at 800 rpm and resuspended in 2 mL of solution A (0.25 M sucrose, 20 mM Tris-HCl, pH 7.4, containing PIs), and transferred into 1.5-mL tubes. Suspensions were homogenized by sonication (2 × 10 seconds) and centrifugation at 15,000 g for 20 minutes. The supernatants were mixed into polyallomar Beckman ultracentrifuge tubes with 2 mL of solution containing PIs and ultracentrifuged for 65 minutes at 105,000 g. Microsomal pellets were drained, resuspended in 100-300 μL of solution A containing PIs, and transferred to microcentrifuge tubes. The suspensions were homogenized by sonication and centrifuged for 20 minutes at 15,000 g. The protein concentration of the supernatant was determined using the Pierce BCA kit, with BSA as a reference.

HO Activity Assay

Five hundred micrograms of microsomal proteins, 2 mM glucose-6-phosphate, 1 mM β-NADPH, and 1 U of glucose-6-phosphate dehydrogenase were assembled on ice in a final volume of 100 μL. The reactions were started by adding, under dark conditions, 25 μM substrate (hemin), vortexing for 5 seconds, and incubating for 1 hour at 37°C.

Reactions were stopped by adding 100 μL of 0.4 μM mesoporphyrin prepared in ethanol/DMSO (95:5). Samples were vortexed for 10 seconds and centrifuged for 10 minutes at 15,000 g. The supernatant content was analyzed by HPLC Waters LC Module I, equipped with an autosampler. Samples were separated on a Lichrocart reverse-phase C18 column (25 × 4 mm) using a linear gradient from 100% solvent A to 100% solvent B over 14 minutes, at a total flow of 1.5 mL/min. Solvent A consisted of 100 mM ammonium acetate, pH 5.1, 60% methanol. Solvent B was 100% methanol. Compounds were detected by absorption at 405 nm.

HO activity was expressed as (picomolar bilirubin equivalents [bilirubin + biliverdin amounts in a same sample] · mg protein)/h.

Statistical Analysis

Statistical analysis was performed using the SPSS 11.0 statistical software package. Differences between the different drugs and control groups were tested for significance (P < 0.001) by the unpaired Student t test.

RESULTS

To examine the effect of nelfinavir on cell viability, differentiated 3T3-F442A adipocytes were treated with this drug for 3 to 36 hours. Cells were analyzed by flow cytometry, and the percentage of necrotic versus apoptotic cells was determined by propidium iodide staining versus Annexin V/propidium iodide staining, respectively. Figure 1A shows that 30 μM nelfinavir induced massive necrosis (67%) of cells after 36 hours, while no significant variation in the percentage of apoptotic cells was observed. Nelfinavir-induced necrosis was observed as soon as 12 hours after addition of nelfinavir (31% vs. 17% for the control). The high rate of apoptosis and necrosis in the controls is explained by the removal of fetal calf serum 12 hours before addition of PIs. The same experiment carried out with indinavir did not lead to a similar result, because the cells treated with this compound (from 20 to 100 μM) did not significantly differ from control cells after 6 days of culture (previous findings¹¹ and data not shown). To confirm the nonapoptotic pathway leading to cell death induced by nelfinavir, we studied the expression of the proapoptotic protein p53.²⁵ As shown in Figure 1B, no variation of the p53 protein could be observed from 3 to 12 hours after addition of nelfinavir. Moreover, neither DNA fragmentation nor rescue of cell death by a caspase 3 inhibitor could be observed in any of the cells treated with nelfinavir (data not shown).

Because expression of TNF-α was higher in adipose tissue from lipodystrophic patients²⁶ and because this cytokine has been implicated in the development of lipodystrophic syndrome,^27,28 we hypothesized that an increase in TNF-α levels could lead to the necrosis induced by nelfinavir. Figure 2A shows that from 3 to 12 hours after treatment, no difference in membranous and cytoplasmic TNF-α protein expression could be observed between control and nelfinavir-treated cells. EMSA detecting NFκB activation was carried out to indirectly determine the level of soluble TNF-α. Figures 2B and C indicate that the addition of nelfinavir did not influence the amount of TNF-α in the culture media between 3 and 12 hours of incubation.

Figure 2

Advances made in the comprehension of lipodystrophic syndrome have established a link between disease onset and reactive oxygen species (ROS) production.²⁹ Moreover, a recent study of the mechanisms of cholesterol metabolism modulation by PIs indicated that nelfinavir could generate endoplasmic reticulum (ER) stress. We therefore examined whether nelfinavir-induced necrosis was due to oxidative stress. Because HO-1 gene expression has been extensively linked to the generation of cellular stress by ROS, we measured the expression of this gene upon treatment of cells with nelfinavir.

Figure 3A shows that, when measured from 3 to 24 hours, a highly significant increase in expression of the HO-1 gene was observed in nelfinavir-treated cells, with a peak at 18 hours (>8-fold the basal expression). This effect only began to attenuate 24 hours after treatment. Indinavir-treated cells also had increased HO-1 gene expression, although this was less pronounced than that observed with nelfinavir (no more than 3-fold the basal expression). HO-1 activity after 24 hours, as determined by bilirubin dosing, followed the measurement of gene expression Fig. 3B.

Figure 3

Because ascorbate and N-acetylcysteine can act as antioxidants and have been shown to decrease HO-1 expression in response to prooxidative compounds, we verified, in our cell model, if these compounds could decrease nelfinavir-induced HO-1 expression. Ascorbate (100 μM) counteracted, in a highly significant manner, the prooxidative effect of nelfinavir in the mouse 3T3-F442A cell line, while N-acetylcysteine had no effect (Fig. 3A). After 18 hours of treatment, the level of HO-1 gene expression in cells treated by nelfinavir and ascorbate was comparable with that in indinavir-treated cells (no more than 3-fold the basal expression).

To examine whether ascorbate or N-acetylcysteine counteracts nelfinavir-induced necrosis of 3T3-F442A, the protective effect of these antioxidants on cell viability was studied by flow cytometry. Figure 4 shows that while ascorbate alone had little effect on the percentage of necrotic cells 24 hours after treatment, nelfinavir was twice as effective. Furthermore, the combination of ascorbate with nelfinavir restored a level of necrosis almost comparable with that of the control, while the combination with N-acetylcysteine had no effect.

DISCUSSION

Treatment of HIV-infected patients with PIs has led to a dramatic drop in the incidence of opportunistic infections and, hence, in the death of patients treated by this group of molecules. However, these PIs (with or without nucleoside reverse transcriptase inhibitors) can lead in 60% to 80% of cases to various conditions grouped under lipodystrophic syndrome.^4-6 One of these molecules, nelfinavir, has been indirectly but highly implicated in lipodystrophic syndrome,¹⁹ due to its deleterious effect especially on adipose tissue^11,16 as well as by its inhibitory and activatory actions on lipogenesis^16,23 and lipolysis, ^20,21,23 respectively. In a previous study,¹¹ we showed that nelfinavir induced 3T3-F442A cell death by an undetermined mechanism, whereas other PIs were not cytotoxic. These results confirmed the findings of Dowell et al¹⁶ on 3T3L1 cells, where an apoptotic effect of nelfinavir was inferred but not definitively demonstrated.

We provide here evidence that nelfinavir can induce, at concentrations similar to those found in the serum of treated patients,³⁰ massive necrosis of mature adipocytes (3T3-F442A cells). It clearly appears that the process of cell death is nonapoptotic, because neither early (Annexin V staining) nor late (DNA fragmentation) markers of apoptotic death were observed. Because there remained ambiguities between our results and those of Dowell et al,¹⁶ we also looked at variations in levels of p53, a proapoptotic protein par excellence. However, we could not detect any variation in the levels of p53 at the protein level using EMSA. Finally, the use of an inhibitor of caspase 3, which is implicated in adipocyte apoptotic death,³¹ was inoperative on nelfinavir-induced cell death. Therefore, these results would suggest a necrotic process. Our results partially confirm those of Dowell et al, with some of the differences observed between the 2 studies, such as DNA fragmentation, being explicable at least partially by the choice of in vitro model. In contrast, Domingo et al³² showed that apoptosis was increased in adipose tissue of HIV-infected patients treated with PIs. This discrepancy could be explained by the difference of models used (patient vs. cell line) and the presence of high viral loads in HIV-infected patients. Furthermore, in our study, we used only 1 drug, while the treatment of HIV infection requires the combination of different PIs and/or nucleoside reverse transcriptase inhibitors.

In agreement with our results, other studies demonstrated that PIs did not induce apoptosis in vitro,^{11,13,14,16,22} supporting the possibility that tissue environment or viral load may affect the effect of the drug.

Nelfinavir has been shown to inhibit apoptosis in T cells,^33,34 while we demonstrated a clear and dramatic necrotic effect on adipocytes. This is consistent with the findings of Dowell et al,¹⁶ who showed increased cell death on adipocytes. Taken together, these results suggest that this drug could be specific for this cell type.

A detailed analysis of the necrotic effect of nelfinavir was conducted to identify potential mechanisms. There is some evidence that TNF-α, a cytokine that is strongly synthesized by adipose tissue and plays an important role in the regulation of metabolism (notably in cell death), is implicated.^35,36 TNF-α expression was augmented in adipose tissue of lipodystrophic patients²⁶ and has also been implicated in the development of lipodystrophic syndrome.^27,28 We therefore verified whether nelfinavir-induced necrosis was linked to augmentation of adipocyte TNF-α secretion. Because the transcription factor NFκB is an obligatory mediator of most TNF-α responses,³⁵ we indirectly monitored TNF-α secretion by measuring NFκB activation. We clearly found that the NFκB pathway is functional in 3T3-F442A cells and can be triggered by lipopolysaccharide, an activator of TNF-α secretion by the adipose tissue.³⁷ However, no variation in NFκB was detectable upon nelfinavir treatment (from 3 to 24 hours), as confirmed by Western blotting of membrane and cytoplasmic TNF-α, which showed steady levels of both forms of TNF-α (Figs. 2A-C). It was therefore evident that nelfinavir-induced necrotic death was not mediated through TNF-α up-regulation.

Cell necrosis is often due to oxidative stress generated by ROS.³⁸ Nerurkar et al,²⁹ in a recent study, made the assumption that the generation of ROS in adipocytes may be an early and critical event in highly active antiretroviral treatment-associated toxicity leading to cell death, partially explaining the mechanism underlying lipoatrophy. Following this reasoning, we investigated whether nelfinavir induces oxidative stress in treated cells. We accordingly studied the gene expression and enzymatic activity of HO-1, an enzyme considered as a reliable marker for oxidative stress.^39,40 Confirming the general hypothesis of Nerurkar et al, we found that nelfinavir can generate a strong oxidative stress, leading to steep up-regulation of HO-1 expression (up to 800%). This augmentation of expression was followed by an increase in enzyme activity (Fig. 3B). It should be noted that this effect was not specific to nelfinavir, because indinavir can also induce expression of the HO-1 gene, but to a lesser extent, and the increase of expression occurred much later than was observed with nelfinavir (Fig. 4A). Moreover, indinavir did not induce an increase in the activity of HO-1, while nelfinavir brought about a 2-fold increase in activity within 24 hours (Fig. 3B). If we take into account that indinavir did not lead to necrosis, because the cells treated with this compound (from 20 to 100 μM) did not significantly differ from control cells after 6 days of culture (previous work¹¹ and data not shown), this leads us to think that nelfinavir's effect upon the HO-1 gene is specific to this drug and that the action of indinavir occurs via an entirely different process. These results are compatible with those from a recent study on Hep-G2 cells¹⁸; this study showed that nelfinavir, at concentrations similar to those used in our study, induced up-regulation of the heat shock proteins GRP78 and GRP94, revealing that this HIV PI induces ER stress.

It is clearly established that antioxidants, such as vitamin C or N-acetylcysteine, can inhibit or limit HO-1 activation, provided that it is as a result of ROS production. Moreover, it has been demonstrated that ascorbate and tocopherol treatment can reverse the effects of stavudine on AKR/J mice liver oxidative stress genes.⁴¹ Addition of vitamin C to the culture medium upon nelfinavir treatment should therefore diminish HO-1 activation, but also limit PI-induced cell death, through the reduction of ROS production. Confirming our hypothesis, our results indeed demonstrated that ascorbate significantly diminished the nelfinavir-triggered activation of the HO-1 gene, thus confirming the induction of a strong oxidative stress when cells are treated by nelfinavir. In parallel, ascorbate has a protective effect on nelfinavir-treated 3T3-F442A cells. Thus, nelfinavir-triggered cell death is diminished by 70% in the presence of ascorbate. It should be noted that the addition of N-acetylcysteine (1 μM) under the same conditions does not bring about diminution of HO-1 activation, nor does it limit nelfinavir-induced necrotic cell death (Figs. 3 and 4). This demonstrates the specificity of ascorbate with respect to the deleterious effects of nelfinavir, which can likely be explained by the fundamental differences between the mechanisms of action of these 2 molecules.⁴²

The mechanism by which nelfinavir induces augmentation of ROS production in the adipocyte remains unclear. However, Strickland et al¹⁸ showed that nelfinavir triggers ER stress that activates the unfolding protein response pathway. This suggests that nelfinavir could favor accumulation of misfolded proteins within the ER.⁴³ Accumulation of these proteins might then induce cell necrosis, possibly through augmentation of ROS inside the ER.⁴⁴ However, it could not be ruled out that nelfinavir could induce an increase of ROS production at the mitochondrial level, leading to necrosis. At present, the published literature does not allow us to firmly establish a link between the ER or mitochondria and the necrotic effect of nelfinavir.

This study therefore shows that nelfinavir is an atypical PI, because it triggers massive necrosis of adipocytes. This necrosis is not mediated through augmentation of secreted TNF-α but through ROS production by adipocytes. Thus, nelfinavir-induced cell death can be almost completely reversed by ascorbate. Despite the difficulties of comparing a 36-hour in vitro exposure to nelfinavir with the long-term treatment of HIV-infected individuals, our results suggest that it would be interesting to study the effects of ascorbate on patients with lipoatrophy linked to nelfinavir administration.

ACKNOWLEDGMENT

The authors gratefully acknowledge the expert helpful discussion with and careful manuscript review by Dr. C. Lefebvre-d'Hellencourt.