Hypoxia causes triglyceride accumulation by HIF-1-mediated stimulation of lipin 1 expression

Exposure of human tissues or cells to reduced oxygen concentration (hypoxia) is encountered both during physiological and pathological processes (Semenza, 2011) and causes dramatic changes in gene expression. Essential to this response are the hypoxia-inducible transcription factors HIF-1 and HIF-2 (EPAS1), recent major anti-cancer targets (Majmundar et al., 2010). Under normal oxygen conditions the regulatory alpha subunit (HIFα) of HIFs is continuously produced and destroyed, in a process involving modification by prolyl hydroxylases (PHDs), VHL-mediated polyubiquitination and subsequent proteasomal degradation (Epstein et al., 2001; Ivan et al., 2001; Kallio et al., 1999; Kamura et al., 2000; Maxwell et al., 1999; Turkish and Sturley, 2007). When oxygen concentration is low, hydroxylation is impaired and HIFα is stabilized, translocates to the nucleus, dimerizes with ARNT (HIFβ) to form HIF and binds to hypoxia-responsive elements (HREs) in the promoters/enhancers of its target genes.

Tissue-specific and differential control of HIFs also involves several oxygen-independent molecular mechanisms. For HIF-1α these include regulation of its mRNA synthesis by NF-κB (Belaiba et al., 2007; Rius et al., 2008) or the PKR–Stat3 axis (Papadakis et al., 2010), control of its stability by phosphorylation (Flügel et al., 2007) or protein–protein interactions (Liu et al., 2007), MAPK-dependent regulation of its nuclear transport (Mylonis et al., 2008) and CK1-mediated control of its heterodimerization (Kalousi et al., 2010).

One of the major adaptive changes in response to hypoxia is HIF-mediated reprogramming of cellular metabolism. HIF-1 activates genes responsible for switching from oxidative to glycolytic metabolism such as the ones coding for glucose transporters, glycolytic enzymes, lactate dehydrogenase and pyruvate dehydrogenase kinase 1 (Majmundar et al., 2010). Furthermore, hypoxia has been shown to stimulate lipid storage and inhibit lipid catabolism in cultured cardiac myocytes and macrophages (Boström et al., 2006; Huss et al., 2001), but the involvement of HIFs in these processes was not investigated. More recent studies using transgenic mice have suggested that artificial or diet induced activation of liver HIF signalling can augment lipid accumulation by modifying hepatocyte lipid metabolism (Kim et al., 2006; Nath et al., 2011; Qu et al., 2011; Rankin et al., 2009). Although these studies have implicated both HIF-1α and HIF-2α in liver fat accumulation and the subsequent maladaptive pathologies, their particular roles have remained largely unclear or controversial and no direct HIF targets have been so far identified among the major lipidogenic enzymes.

A key role in lipid biosynthesis is played by lipins, a family of proteins with Mg²⁺-dependent phosphatidate phosphatase (PAP1) activity that catalyze the conversion of phosphatidic acid (PA) to diacylglycerol (DAG) in the penultimate step of triglyceride (TG) synthesis (Donkor et al., 2007; Han et al., 2006). TGs form the core of lipid droplets (LD), the major neutral lipid stores in adipocytes as well as other cell types. Mutations in the lipin 1 gene (Lpin1), which is expressed predominantly in adipose tissue and skeletal muscle, are the cause of lipodystrophy in the fatty liver dystrophic (fld) mouse and, conversely, overexpression of lipin 1 in the adipose tissue of transgenic mice promotes obesity (Péterfy et al., 2001; Phan and Reue, 2005). Moreover, lipin 1 as well as lipin 2, a liver-enriched isoform, interact with and/or modulate the activity of several transcription factors, including members of the peroxisome proliferator-activated receptor (PPAR) family and SREBP that control the expression of genes involved in fatty acid and lipid metabolism (Donkor et al., 2009; Finck et al., 2006; Koh et al., 2008; Peterson et al., 2011). Apart of its apparently important role in adipogenesis, relatively little is known about the function of lipin 1 in cells other than adipocytes. Various stimuli such as glucocorticoids, insulin or fasting alter lipin 1 expression and/or activity, regulation of which is incompletely understood and involves transcriptional control, post-translational modifications and compartmentalization (reviewed by Harris and Finck, 2011).

We now report that the lipin 1 gene is directly controlled by HIF-1 and its up-regulation is linked to TG and LD accumulation in cells of non-adipocyte origin, suggesting that lipin 1 is an important mediator of the metabolic adaptation to hypoxia at the cellular level. Moreover, inhibition of HIF-1 by pharmacological or genetic means can suppress the hypoxic induction of lipin 1 and TG or LD production.

To investigate the effect of hypoxia on lipid biosynthesis we analysed human hepatoblastoma (Huh7) and cervical adenocarcinoma (HeLaM) cells exposed to normoxia or hypoxia (1% O₂). As expected, hypoxia caused a robust expression of nuclear HIF-1α that could be detected by immunofluorescence microscopy (Fig. 1A,B, left panels). Furthermore, Nile Red staining revealed that in both cell types hypoxia caused a significant and time-dependent increase in LD accumulation, which was comparable to that caused by oleic acid (OA) treatment used as positive control under normoxia (Fig. 1A,B, middle panels). The induction of neutral lipid accumulation by hypoxia was confirmed in Huh7 cells by determining the cellular TG content using a biochemical assay (Fig. 1C). Also in this case, TG accumulation under hypoxia was similar to the accumulation triggered by oleic acid.

Analysis of Huh7 cells by western blotting showed that in addition to HIF-1α, the protein levels of lipin 1, which is considered to be the minor liver isoform, also increased in response to hypoxia (Fig. 2A). On the contrary, lipin 2, normally the major hepatic isoform, remained unaffected indicating an isoform-specific effect of hypoxia on lipin expression. Similar results were also obtained with HeLaM cells treated with hypoxia or dimethyloxalyl glycine (DMOG), a PHD inhibitor that stabilizes HIFα and activates the HIFs (Fig. 2B). Induction of lipin 1 by hypoxia was also confirmed using immunofluorescence microscopy in both Huh7 (Fig. 2C) and HeLaM (Fig. 2D) cells. The overexpression of endogenous lipin 1 under low oxygen conditions allowed its visualization, which was previously not easily attainable in this type of cells. In Huh7 cells, hypoxically induced lipin 1 displayed a predominant cytoplasmic, both diffuse and reticular, localization. Nuclear staining was weak but became more evident at later time points (Fig. 2C). Similar images were obtained in HeLaM cells, in which, however, cytoplasmic spots and nuclei were stained more intensely at early and late time points, respectively (Fig. 2D). Therefore, hypoxia causes an increase in the cytoplasmic protein levels of lipin 1, which, under prolonged treatment, also appears to migrate inside the nucleus.

To test whether the effects of hypoxia on lipin 1 expression and lipid droplet accumulation can also be reproduced in normal, non-cancer cells we used primary, non-transformed human bronchial smooth muscle (hBSM) cells. Analysis of these cells by western blotting showed that both HIF-1α and lipin 1 increased in response to hypoxia, especially after 48 hours of treatment, whereas lipin 2 expression was not detectable (Fig. 3A). Hypoxic induction was confirmed using immunofluorescence microscopy (Fig. 3B), which additionally showed that lipin 1 is predominantly cytoplasmic in hBSM cells under normoxia but also translocates inside the nucleus under hypoxia. Finally, Nile Red staining revealed that 48 hours of hypoxic treatment causes significant LD accumulation also in hBSM cells (Fig. 3C), suggesting that hypoxia affects lipin 1 expression and neutral lipid content in both cancer and non-cancer cells.

The analysis was continued using Huh7 cells, which, as derived by hepatocytes, are more appropriate for studying the relationship between lipin 1 induction and lipid accumulation. First, to investigate the increase in lipin 1 protein levels, we determined its mRNA levels using qPCR. Lipin 1 mRNA significantly increased in Huh7 cells subjected to hypoxia for 8 hours, whereas lipin 2 mRNA was little affected (Fig. 4A). In the same experiment, hypoxia caused robust mRNA expression of VEGF, a well-known HIF target. Interestingly, lipin 1 mRNA levels decreased after 24 hours of hypoxic treatment (although still remaining higher than normoxia) suggesting a transient effect or the operation of a negative feedback loop, possibly related to the nuclear entry of lipin 1 that was also observed at later time points. Since Lpin1 appears to be an early hypoxia-inducible gene, we addressed its involvement in hypoxic TG accumulation by siRNA-mediated silencing. Knocking down of lipin 1 was effective both under normoxia and hypoxia as judged by western blotting (Fig. 4B, upper panel). Depletion of lipin 1 under normoxia did not significantly affect the TG content of the cells (Fig. 4B, lower panel). In contrast, suppression of lipin 1 expression completely blocked hypoxia-inducible TG accumulation. To a lesser extent, lipin 1 knockdown also diminished TG accumulation triggered by OA treatment, in agreement with a general role of lipin 1 in response to lipidogenic stimuli in hepatocytes. Therefore, lipin 1 is not only transcriptionally controlled by hypoxia but is also required for the ensuing overproduction of triglycerides.

We then investigated the role of HIFs in the hypoxic induction of lipin 1 and TG synthesis. Overexpression of either HIF-1α or HIF-2α in Huh7 cells increased the protein levels of lipin 1, with HIF-1α being considerably more effective (Fig. 5A). In contrast, neither HIF-1α nor HIF-2α overexpression affected lipin 2 expression. Quantification of the fluorescent signal of Huh7 cells overexpressing HIFα and stained with Nile Red showed that HIFα overexpression also causes LD accumulation (Fig. 5B). In accordance to its effect on lipin 1 expression, HIF-1α exerted a stronger effect on LD content than HIF-2α. Therefore, HIF-1 and, to lesser extent, HIF-2 activation is sufficient to induce both lipin 1 and LD production even under normoxic conditions.

The involvement of HIFs in the hypoxic induction of lipin 1 was further tested by siRNA-mediated repression of HIF-1α or HIF-2α in Huh7 cells treated with DMOG. Knockdown of HIF-1α greatly reduced lipin 1 expression (Fig. 5C). HIF-2α silencing also had a negative, but considerably weaker, effect on lipin 1. Neither knockdowns affected lipin 2. These data imply that hypoxic lipin 1 induction and subsequent TG and LD accumulation is predominantly mediated by HIF-1, with HIF-2 playing a possible minor role. Knocking down HIF-1α using a shRNA plasmid also largely eliminated the capacity of hypoxia to induce lipin 1 (Fig. 5D) and, moreover, caused a modest but significant reduction in hypoxic TG accumulation (Fig. 5E), while it did not affect TG accumulation caused by OA treatment.

Oxygen tension and hydroxylation regulate HIF-1α protein stability. However, the ability of HIF-1α to enter the nucleus and activate transcription is additionally controlled by phosphorylation. We have previously shown that CK1δ targets HIF-1α and inhibits its activity by impairing its binding to ARNT (Kalousi et al., 2010), while phosphorylation by p44/42 MAPK stimulates HIF-1α activity by blocking its nuclear export (Mylonis et al., 2008). To correlate lipin 1 expression to HIF-1α activity rather than just its protein levels, we used kinase inhibitors. Treatment of Huh7 cells with the CK1δ inhibitor IC261, which stimulates HIF-1α activity (Kalousi et al., 2010), potentiated hypoxic induction of lipin 1 without significantly affecting lipin 2 or HIF-1α protein levels (Fig. 5F). Inversely, treatment of Huh7 cells with the flavonoid kaempferol, which suppresses HIF-1α activity by impairing its nuclear accumulation through MAPK inhibition (Mylonis et al., 2010), diminished lipin 1 expression under hypoxia (Fig. 6A,B). The same treatment also significantly reduced the amount of LDs in Huh7 cells grown under hypoxia (Fig. 6B,C). Therefore, both gain-of-function and loss-of-function experiments strongly suggest that hypoxic induction of lipin 1 and neutral lipid accumulation involves predominantly HIF-1.

A direct role of HIF-1 in Lpin1 regulation requires the contribution of an HRE. Examination of the 3.5 kb promoter region of Lpin1 revealed the presence of 8 potential HRE sites conforming to the consensus XCGTG (X = A or G or C). To test the functionality of these potential HREs, we constructed luciferase reporter plasmids containing the whole region (3298 bp; HRE_1–8) or 5′ truncated forms containing 7 (HRE_2–8), 4 (HRE_5–8), 2 (HRE_7–8) or 1 (HRE₈) proximal potential HREs (Fig. 7A, upper panel) and introduced them in Huh7 cells. The full-length region (HRE_1–8) exhibited the highest promoter activity under normoxia and, more importantly, it was the only one responding to and activated by hypoxia (Fig. 7A, upper panel), suggesting that only the most distal HRE (HRE₁) is functional. This was also positively shown by using two 3′-truncated forms of the Lpin1 promoter region containing only HRE₁ or four distal HRE sites (HRE_1–4): both of these promoter fragments could also be activated by hypoxia (Fig. 7A, lower panel). HRE₁ is conserved (Fig. 7B) as it is perfectly matched by a potential distal HRE in the mouse Lpin1 promoter region and is found close to the only other known distal regulatory elements, the SRE and NF-Y sites (Ishimoto et al., 2009). Finally, mutation of HRE₁ (Fig. 7B), in the context of both full-length and truncated Lpin1 promoter regions (producing mHRE_1–8, mHRE₁ and mHRE_1–4), led to complete loss of responsiveness to hypoxia (Fig. 7C). Therefore, transcriptional activation of Lpin1 under hypoxia is mediated by a single distal HRE site in its promoter.

Interaction of HIFs with HRE₁ in the Lpin1 promoter was tested by introducing wild-type or mutant reporter constructs HRE_1–8, HRE₁ and HRE_2–8 in Huh7 cells overexpressing HIF-1α or HIF-2α. HIF-1 could activate transcription from wild-type full-length or single HRE₁ promoter constructs but not from constructs lacking HRE₁ (HRE_2–8) or containing its mutant form (mHRE_1–8 and mHRE₁; Fig. 8A). HIF-2 also activated the wild-type full-length promoter, albeit weaker than HIF-1, but failed to activate transcription of the single distal HRE construct (HRE₁) suggesting that its role in Lpin1 activation is minor and involves additional regulatory elements aside of HRE₁.

Involvement of endogenous HIF-1 in Lpin1 regulation was tested in cells transformed with the full-length Lpin1 promoter construct and overexpressing the wild-type or SA mutant form of the HIF-1α MTD under hypoxia. Overexpression of wild-type MTD (MAPK-target domain), a 42 amino acid long fragment of HIF-1α containing its MAPK modification sites, specifically inhibits HIF-1α by blocking its phosphorylation and triggering its nuclear export, while the mutant SA form of MTD lacks the MAPK sites and is inactive (Mylonis et al., 2008). In cells expressing wild-type MTD the hypoxic activation of Lpin1 promoter was virtually abolished but persisted in those cells expressing the mutant SA MTD form (Fig. 8B), providing further evidence that HIF-1 is the major transcriptional activator of Lpin1 gene under hypoxia.

Direct binding of HIF-1α to HRE₁ was tested by chromatin immunoprecipitation using Huh7 cells cultured under hypoxia or normoxia. The Lpin1 promoter region containing HRE₁ was greatly enriched in anti-HIF-1α immunoprecipitates of DNA-protein complexes isolated from hypoxically treated cells in comparison to rabbit IgG immunoprecipitates or anti-HIF-1α immunoprecipitates from normoxic cells (Fig. 8C,D). Therefore, HIF-1α not only activates but also physically associates with the Lpin1 promoter.

We have revealed a direct link between HIF-1α and lipin 1 or the cellular response to hypoxia and the metabolic pathway that leads to synthesis of triglycerides and formation of lipid droplets (Fig. 8E). Although this connection was identified in cultured human cells, it can mechanistically explain previous observations made with whole transgenic animals. Over-expression of a constitutively active form of HIF-1α in mouse liver led to hepatic fat accumulation and liver steatosis (Kim et al., 2006). In contrast, similar over-expression of HIF-2α did not result in liver steatosis but rather caused vascular lesions. Furthermore, lipid accumulation in mouse hepatocytes caused by alcohol feeding was also shown to involve HIF-1α (Nath et al., 2011). On the other hand, in double knockout studies, hepatic steatosis caused by liver-specific deficiency of pVHL could be avoided by simultaneous inactivation of HIF-2α (Qu et al., 2011; Rankin et al., 2009). However, in these animal models HIF-2 was mainly associated with suppression of fatty acid oxidation, increased lipid storage capacity, serum TG levels and inflammation. In our study with human cells, HIF-2 also seems to be involved in lipin 1 regulation and LD accumulation but its role is minor compared to HIF-1 and probably indirect since it requires more complex regulatory elements. It, therefore, appears that HIF-1α and HIF-2α may have overlapping or complementary but distinct roles in the regulation of hepatic lipid metabolism and the development of related pathologies such as alcoholic and non-alcoholic fatty-liver disease (AFLD and NAFLD, respectively), with HIF-1 predominantly stimulating lipin 1 and triglyceride synthesis under hypoxic or other stress conditions. HIF-1 has also been implicated in cardiac steatosis by controlling expression of PPARγ, a principal transcriptional activator of lipid anabolism (Krishnan et al., 2009). Lipin 1 also co-activates PPARγ (Koh et al., 2008), so the HIF-1–lipin-1 axis may serve to amplify not only PAP1 activity but also the non-catalytic transcriptional lipin 1 functions.

In order for lipin 1 to exert its transcriptional role, it has to translocate inside the nucleus. Although lipin 1 is predominantly cytoplasmic in our tested cell lines under normoxia, hypoxic treatment increases its nuclear pool especially after long-term induction. A recent report has shown that nuclear localization of lipin 1 is regulated by nutrient- and growth factor-responsive kinase mTOR complex 1 (mTORC1)-mediated phosphorylation (Peterson et al., 2011). Inhibition of mTORC1 causes normoxic redistribution of lipin 1 from the cytoplasm to the nucleus, which in turn reduces the nuclear protein levels of SREBP-1 (Peterson et al., 2011), a transcription factor regulating many lipogenic genes, including lipin 1 (Ishimoto et al., 2009). Therefore, our data showing nuclear accumulation of lipin 1 and a decrease of lipin 1 mRNA expression at later points of hypoxic treatment could be due to inhibition of mTORC1 that is known to occur after prolonged exposure to hypoxia (Wouters and Koritzinsky, 2008). The physiological relevance of lipin 1 hypoxic nuclear migration could thus be justified by the operation of the following negative feedback mechanism. After the initial HIF-1-mediated induction of lipin 1 and the subsequent TG accumulation, inhibition of mTOR by prolonged hypoxia causes translocation of lipin 1 inside the nucleus, where, by suppressing SREBP-1 function, it reduces expression of its own and other lipogenic genes and blocks further TG de novo synthesis.

The idea that cells in a stressful situation, such as oxygen deprivation, respond by stimulating, even transiently, an anabolic process, e.g. lipid biosynthesis, is counterintuitive but not without precedent. Hypoxia also stimulates glycogen synthesis in a HIF-1-dependent way (Brahimi-Horn et al., 2011; Fakas et al., 2011; Pescador et al., 2010). Apparently, exposure of cells to mild hypoxia may serve as a way of sensing ensuing ischemia/nutrient limitation and, while it is still available, glucose is stored for later use as anaerobic source of energy. However, this may not be the case for lipids since they can only produce energy through oxidative phosphorylation, a process drastically inhibited by severe hypoxia. A more likely explanation for storing triglycerides in the form of lipid droplets under hypoxia could be that, in this way, cells buffer the toxicity caused by free fatty acids, which build up because of suppressed respiratory chain activity. Increased PAP1 activity and formation of lipid droplets can, thus, be a protective response that prevents intracellular lipotoxicity (Fakas et al., 2011; Turkish and Sturley, 2007) although stored triglycerides could still serve as a source of lipotoxic intermediates if the cells remain for long unable to handle them metabolically (Neuschwander-Tetri, 2010).

Irrespective of its cellular role, the direct involvement of HIF-1 in triglyceride formation may open new avenues for therapeutic interventions in lipid disorders leading to NAFLD and steatohepatitis. Furthermore, LD accumulation in non-adipose tissue (ectopic fat) is a hallmark of the metabolic syndrome and is linked to the pathogenesis of insulin resistance and diabetes (Unger et al., 2010). The extensive recent pharmacological targeting of HIF-1 as means of anti-cancer or anti-ischemic treatment, has led to the identification and availability of many agents that inhibit or activate HIF-1, respectively (Semenza, 2011). Our results with HIF-1 inhibitors, such as kaempferol and MTD, provide proof-of-principle that such agents may also be useful as effective modulators of intracellular lipid metabolism and can form the basis of novel molecular therapies targeting lipin 1.

The human Lpin1 promoter region (nucleotides −3298 to −1; Acc. No.: NT_005334.16) was amplified by PCR using HeLaM total DNA as template, primers hL1prmF1 and hL1prmR1 and KOD Hot start DNA polymerase (TOYOBO Life Science, Tokyo, Japan). The PCR fragment was then introduced into pUC19 vector, sequenced and subcloned into pGL3-basic vector (Promega, Madison, WI, USA) to construct pGL3-HRE_1–8. Truncated promoter forms, pGL3-HRE_2–8 (−2198), pGL3-HRE_5–8 (−1297), pGL3-HRE_7–8 (−602) and pGL3-HRE₈ (−445) were constructed by using as template the pUC19-HRE_1–8 plasmid, and an appropriate set of primers listed in supplementary material Table S1. The occurring PCR fragments were then subcloned into the pGL3-basic vector and sequenced. To construct HRE mutant plasmid pGL3-mHRE_1–8, three nucleotide substitutions (CCGTG to CAACG) were introduced by PCR using hL1pHRE1mF1 and hL1pHRE1mR1 primers listed in supplementary material Table S1. Plasmids pGL3-HRE₁ (−3298 to −2588), pGL3-HRE_1–4 (−3298 to −1585) and their respective mutant forms, pGL3-mHRE₁ and pGL3-mHRE_1–4, were constructed by digesting pGL3-HRE_1–8 with BglII, to yield fragment HRE₁, or HindIII, to yield HRE_1–4, and subcloning to the empty pGL3-vector. pEGFP-HIF-1α, pCMV-FLAG-MTD and pCMV-FLAG-MTD-SA plasmids were previously described (Mylonis et al., 2008). To construct pEGFP-HIF-2α, full-length HIF-2α cDNA was obtained from pcDNA-HIF-2α, kindly provided by S. L. McKnight [University of Texas Southwestern Medical Center (Tian et al., 1997)].

Human HeLaM and Huh7 cells were cultured in DMEM (Biosera) containing 10% FCS and 100 U/ml penicillin-streptomycin (Gibco BRL). Human bronchial smooth muscle cells (Lonza, Switzerland) were cultured in DMEM F-12 containing 10% FCS and 100 U/ml penicillin-streptomycin. When required, cells were treated for 4–24 hours with 1 mM dimethyloxalyl glycine (DMOG; Alexis Biochemicals), with CK1 inhibitor IC261 (2 µM; Sigma, St Louis, MO) or with kaempferol (10–50 µM; Sigma). For oleic acid treatment, the medium was supplemented with 0.4 mM oleic acid (Sigma) bound to 1.5% bovine serum albumin. For hypoxic treatment, cells were exposed for 4–48 hours to 1% O₂, 95% N₂ and 5% CO₂ in an IN VIVO₂ 200 hypoxia workstation (RUSKINN Life Sciences, Pencoed, UK). Transient transfections and reporter gene assays were performed as described previously (Mylonis et al., 2008).

Antibodies used for western blotting, immunofluorescence and chromatin immunoprecipitation were: affinity purified rabbit polyclonal antibodies against HIF-1α (Lyberopoulou et al., 2007), lipin 1 and lipin 2 (Grimsey et al., 2008), rabbit polyclonal antibodies against GFP (Mylonis et al., 2008) or anti-HIF-2α (Abcam, Cambridge, UK) and mouse monoclonal antibodies against actin or tubulin (Millipore, Billerica, MA, USA). Analysis of total cellular proteins by immunoblotting and immunofluorescence microscopy were carried out as previously described (Grimsey et al., 2008; Lyberopoulou et al., 2007; Mylonis et al., 2008; Mylonis et al., 2006). To visualize lipid droplets after usual immunofluorescence microscopy procedure cells were stained with Nile Red (Sigma; 0.1 µg in PBS) for 15 minutes, washed with PBS and mounted on slides (Greenspan et al., 1985).

Huh7 cells were lysed in 150 µl PBS with 0.1% Triton X-100 (Sigma). The mixture was then sonicated (five 10 sec pulses at maximum intensity), boiled for 5 minutes to facilitate the formation of an even suspension (Al-Anzi and Zinn, 2010) and 20 µl aliquots were tested using a biochemical triglyceride determination kit (Biosys, Athens, Greece). Triglyceride values were normalized to protein measured with Bio-Rad Protein Assay (Bio-Rad, Hercules, CA).

Huh7 cells were incubated in serum-free DMEM for 4 hours with siRNA (10 nM) against HIF-1α or HIF-2α (Qiagen, Hilden, Germany) or Lpin1 (target sequences shown in supplementary material Table S1) in the presence of Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA). AllStars siRNA (Qiagen, Hilden, Germany) was used as negative control. shRNA-mediated silencing of HIF-1α was done using pSuper-shHIF-1α, kindly provided by I. Papandreou and N. C. Denko [Stanford University School of Medicine (Papandreou et al., 2006)] and Lipofectamine 2000 (Invitrogen).

Total RNA from Huh7 cells was isolated using the Trizol reagent (Invitrogen) and cDNA was synthesized with the High Capacity Reverse Transcription kit (Applied Biosystems, Carlsbad, CA). Real-time PCR was performed with SYBR GreenER qPCR SuperMix Universal (Invitrogen) in a MiniOpticon instrument (Bio-Rad). The mRNAs encoding lipin 1, lipin 2 and VEGF were amplified using primers listed in supplementary material Table S1. Each sample was assayed in triplicate for both target and internal control. Relative quantitative gene expression was calculated using the ΔΔCt method and presented as relative units.

Huh7 cells were treated with 1.1% formaldehyde for 30 minutes, quenched with 125 mM glycine for 5 min and collected by centrifugation. Following steps were performed as previously described (Braliou et al., 2008) using a polyclonal anti-HIF-1α antibody (Lyberopoulou et al., 2007) or rabbit IgG. The region −2916 to −2686 of the hLPIN1 promoter was amplified from immunoprecipitated chromatin or input samples using primers listed in supplementary material Table S1. The 230 bp product was analyzed by agarose gel electrophoresis or quantified by real-time PCR. The Ct (threshold cycle) of immunoprecipitated samples (with specific anti-HIF-1α antibody or with non-specific IgG) was normalized to the Ct of total input DNA (ΔCt). Results obtained with specific anti-HIF-1α antibody was then normalized to those obtained with non-specific IgG antibodies (ΔΔCt).

Quantification of the surface covered by Nile Red fluorescence was performed with ImageJ public domain software for image analysis and expressed as pixels/cell (Abramoff et al., 2004). Statistical differences between two groups of data were assessed using the unpaired t-test in the SigmaPlot v. 9.0 software (Systat); P<0.05 was considered to be significant.

We are grateful to Drs I. Papandreou, N. C. Denko and S. L. McKnight for generously providing plasmids and to Dr E. Paraskeva, Laboratory of Physiology, University of Thessaly, for providing hBMS cells and instructions for their culture.