Identification of a 16-Nucleotide Sequence That Mediates Post-transcriptional Regulation of Rat CYP2E1 by Insulin*

Insulin directly down-regulates the gene expression of the rat CYP2E1 by altering its mRNA stability (De Waziers, I., Garlatti, M., Bouguet, J., Beaune, P. H., and Barouki, R. (1995) Mol. Pharmacol. 47, 474–479). Because the regulation of CYP mRNA stability was poorly understood, the molecular mechanisms involved in this regulation in the rat hepatoma H4IIEC3 cell line were studied. By using RNase T1 protection methods, the formation of a major CYP2E1 RNA-protein complex was observed. By competition experiments, the binding site of this complex was located on a 16-nucleotide sequence in the 5 (cid:1) -proximal region of the CYP2E1-coding sequence. Insulin did not modify the binding pattern of proteins to this sequence. and transfections of expression vectors or antisense oligonucleotides were undertaken to demonstrate the actual functionality of the 16-mer sequence. The insertion of this sequence in a luciferase gene was sufficient to render the chimeric mRNA sensitive to insulin. Furthermore, transfection of H4IIEC3 cells with antisense oligonucleotide complementary to this sequence blocked the insulin effect on the CYP2E1 mRNA expression, i.e. its rapid degradation. All these results demonstrate

Insulin controls a number of metabolic pathways by regulating the expression of a variety of genes (2). The mechanisms of insulin action have been determined in the case of enzymes involved in glucose and lipid metabolism and were found to be mostly transcriptional. Insulin also controls the expression of several genes encoding cytochromes P450 (CYP). 1 These enzymes are involved in the metabolism of xenobiotics and some endogenous compounds. Both the mechanisms and the biological implications of these regulations are poorly understood. Previous studies indicated that, in contrast to other metabolic genes, insulin does not regulate the transcription of the CYP genes but rather modulates a post-transcriptional step, in par-ticular mRNA stability. The mechanism whereby insulin regulates mRNA stability remains unknown.
CYP gene regulation has been studied mainly at the transcriptional level (3,4). However, CYP isoenzymes are also regulated at a post-transcriptional level, and little is known about the molecular mechanisms involved in this process. Posttranscriptional regulations are mediated by specific RNA-protein interactions that result either in the activation of mRNA degradation or in preventing ribosomal access to the translation start codon. Such interactions often occur in the 5Ј-or 3Ј-untranslated regions of the mRNA and rarely within the coding region (5). Recent studies have shown that some motifs in the 3Ј-untranslated region (UTR) of mRNAs may play an important role in the post-transcriptional regulation of CYP1A2 and -2A5 genes. More precisely, it has been shown that these elements may serve as binding sites for proteins that could play a regulatory role in gene expression (6 -8).
CYP2E1 metabolizes various endogenous compounds of physiological importance such as lipid hydroperoxides (9,10), ketone bodies, and acetone (11,12) in addition to a wide variety of xenobiotics such as ethanol, chlorzoxazone, nitrosamines, and halogenated alkanes (13)(14)(15). CYP2E1 may generate reactive oxygen species (16) known to be involved in the etiology and pathology of many diseases including insulin-dependent diabetes (17). The regulation of the CYP2E1 gene expression is complex and involves several mechanisms. Whereas several xenobiotics increase CYP2E1 protein levels by activating translational efficiency and/or protein stabilization, both CYP2E1 mRNA and protein levels are altered in response to physiopathological conditions (18). Diabetes has been reported to increase CYP2E1 expression at both the mRNA and protein levels in chemically induced and spontaneous diabetic rats (19 -22). This elevation of CYP2E1 mRNA levels in the diabetic state in vivo has been attributed to mRNA stabilization (23) and can be reversed by daily insulin treatment (24).
The rat hepatoma Fao and H4IIEC3 cell lines have been used to study the insulin effect and CYP gene regulation because they exhibit functional insulin receptors by which insulin regulates the expression of several genes (25,26) and because they express some of the inducible CYPs including CYP2E1 (27,28). Previous results showed that insulin directly downregulates the expression of CYP2E1 through a post-transcriptional mechanism in these cells (1).
The aim of the present investigation was to further study the mechanisms of CYP2E1 mRNA stability and its regulation by insulin. The first step was to detect the binding of cytoplasmic proteins to CYP2E mRNA and to localize the binding sequence. We show here that cytoplasmic proteins were able to bind to a 16-nucleotide sequence within the CYP2E mRNA and that this sequence is implicated in CYP2E mRNA destabilization by insulin.

EXPERIMENTAL PROCEDURES
Cell Culture and Treatments-H4IIEC3 cells were derived from the Reuber H35 rat hepatoma (29). They were grown in monolayer culture as described previously (30) for their close parent Fao cells. Cells were exposed to insulin (0.1 M), the usual supplement to primary hepatocyte culture medium, for 6 h during their exponential phase of growth. Untreated cultures were used as controls.
Northern Blot Analysis-Total RNA from H4IIEC3 cells was extracted using the RNeasy Mini kit (Qiagen, Courtaboeuf, France). Subsequently, 10 -20 g of total RNA were subjected to electrophoresis in denaturing formaldehyde, 1.2% agarose gels and transferred to nylon membranes. These membranes were prehybridized and then hybridized with several 32 P-labeled cDNA probes as follows: CYP2E, Renilla luciferase, calmodulin, and 18 S rRNA. After hybridization, the filters were washed at 65°C for 30 min successively with 2ϫ standard saline citrate (SSC) (300 mM sodium chloride, 30 mM sodium citrate, pH 7.0), 0.1% SDS, then 0.5ϫ SSC, 0.1% SDS, and sometimes 0.1ϫ SSC, 0.1% SDS and then were processed for autoradiography. The relative intensities of the hybridization signals were determined by scanning with a Phos-phorImager Storm 840 (Amersham Biosciences).
Preparation of Cytoplasmic Extracts-After trypsinization, a pellet of cells from five culture dishes (30 ϫ 10 6 cells), was resuspended in 25 mM Tris, pH 7.9, 0.5 mM EDTA. After 4 cycles of freezing in ethanol/dry ice and thawing at 37°C for 5 min, the samples were then centrifuged at 4°C at 15,000 ϫ g for 15 min; the supernatant was aliquoted and stored at Ϫ80°C until used. The protein concentration was determined using the bicinchoninic acid (Pierce) procedure with bovine serum albumin as a standard according to the manufacturer's instructions.
Plasmids-Dr. F. J. Gonzalez (NCI, National Institutes of Health, Bethesda) generously provided us with a partial rat CYP2E1 cDNA lacking the 5Ј-UTR. To obtain the full-length CYP2E1 cDNA, this partial cDNA was reamplified by PCR using a primer containing the 5Ј-UTR. EcoRI sites were added in both primers for cloning into pcDNAI/Amp (Invitrogen) (Fig. 1). This plasmid was named pcDNAI/CYP2E1. A fragment of the rat CYP2E1 cDNA from BamHI (base 680) to EcoRI (base 1653) sites was also subcloned into pcDNAI/Amp.
The Renilla reniformis luciferase expression plasmid p␣glob-RL (named pRL), which contains the proximal promoter of the ␣-globin gene, was constructed by Y. Morel (see Ref. 31). To maintain the open reading frame, an 18-mer oligonucleotide (CTTCCCATCCTTGG-GAAC), corresponding to the binding site of cytoplasmic proteins found in the CYP2E1 sequence, was inserted in pRL, in both orientations, in the coding sequence of the Renilla gene at the AflIII site, 183 bases after the ATG codon. The insertions were analyzed by sequencing.
In Vitro Transcription-After linearization of the pcDNAI/CYP2E with either the NotI or Xma or EcoRV restriction enzymes (Fig. 1), the products were transcribed with T7 RNA polymerase in the presence or in the absence of [ 32 P]UTP (800Ci/mmol, Amersham Biosciences) using standard protocols (Stratagene, Amersham Biosciences). Transcripts were separated from unincorporated nucleotides on a G-50 Sephadex column, and specific activity was determined by counting an aliquot in a Packard scintillation counter.
RNA-Protein Binding Reactions (RNase T1 Protection)-The reaction was carried out in a final volume of 20 l. Cytoplasmic proteins (40 g) and 10 5 cpm of 32 P-labeled RNA probe were incubated at room temperature for 10 min in 10 mM HEPES, pH 7.6, 3 mM MgCl 2 , 40 mM KCl, 1 mM dithiothreitol, 5% (v/v) glycerol, and 100 ng/l yeast tRNA to prevent nonspecific binding. 100 units of RNase T1 (Roche Diagnostics) were added to the reaction mixture and incubated for 20 min at 37°C to degrade the mRNA unprotected by protein binding. The mixture was electrophoresed on a 5.3% native polyacrylamide gel (polyacrylamide/ bisacrylamide ratio, 37.5:1) with 1 mM Tris-HCl, 1 mM boric acid, 25 M EDTA, pH 8.5, as running buffer. Protein-32 P mRNA complexes were detected by autoradiography of the dried gel and quantified with a PhosphorImager. In competition experiments, the unlabeled RNA (50fold in excess) was previously incubated at room temperature for 10 min with cytoplasmic proteins before the addition of the radiolabeled probe.
In competition experiments with antisense (AS) oligonucleotides, the antisense (3 g) was preincubated with the 32 P-labeled RNA probe to allow base pairing for 20 min at room temperature before the addition of cytoplasmic proteins. A 16-mer RNA (UUCCCAUCCUUGGGAA) was synthesized by Genset (Paris, France) and 5Ј-labeled with [␥-32 P]ATP by the T4 polynucleotide kinase. Incubations and competition experiments with antisense oligonucleotides were performed as described above for RNAprotein binding reactions without addition of RNase T1.
Transfection Experiments-Transfection experiments were performed in H4IIEC3 cells. Briefly, 1 day before transfection, cells (0.4 ϫ 10 6 cells/well) were seeded into 6-well dishes. The pRL vectors, with or without an 18-mer insertion (1 g), were incubated with the Fu-GENE TM 6 reagent (Roche Diagnostics) (3 l) in 100 l of serum-free medium at room temperature for 20 -30 min according to the manufacturer's instructions. The mixture was added to the cells in 2 ml of culture medium. Four to 6 h later, the culture medium was changed. The following day, the cells were treated or not by insulin (0.1 M) for 6 h. Total RNA from 3 wells by point were extracted using the RNeasy mini kit (Qiagen, Courtaboeuf, France).
One day before the transfection, H4IIEC3 cells (10 6 cells/well) were seeded into a 6-well dish in the culture medium. Four hundred nM concentrated solutions of AS1,2-P and P were prepared in culture medium and added to a final concentration of 200 nM. Thirty min later, cells were treated or not with insulin (0.1 M). After 6 h, total RNA from two wells by point was extracted.

RESULTS
It was previously shown that a 6-h insulin treatment decreased the amount of CYP2E1 mRNA by 50% in the highly differentiated Fao hepatoma cells by a post-transcriptional mechanism (1). The same effect of insulin was observed in H4IIEC3 cells also derived from the Reuber rat hepatoma which were used in the present study (data not shown). In order to determine the mechanisms of CYP2E1 mRNA stability and its regulation, we first studied RNA-protein interactions and mapped the protein-binding sites in the CYP2E1 mRNA. Such binding sites were detected by their ability of bound proteins to protect a labeled RNA probe from complete degradation by RNase. By using H4IIEC3 cells, cytoplasmic proteins, and the full-length CYP2E1 mRNA as a probe, a major CYP2E1 RNA-protein complex was observed (Fig. 2, lanes 3  and 5). Treatment of the binding reaction mixture with proteinase K completely abolished the formation of the complex (data not shown); moreover, the formation of this complex was prevented by a 50-fold excess of unlabeled full-length CYP2E1 mRNA as competitor (Fig. 2, lanes 4 and 6). The pattern of the labeled complex formed was similar with cytoplasmic proteins from either control or insulin-treated cells (Fig. 2, lanes 3 and  5). In order to map the protein-binding site on the CYP2E1 mRNA, various truncated mRNA probes were prepared and used in binding reactions (1-824, Fig. 3A, and 1-191, Fig. 3B). The RNA protein complex was observed with both probes and was competed out in each case by an excess of homologous unlabeled mRNA. Treatment by insulin did not modify the binding pattern. Moreover, in order to confirm the specificity of this binding, CYP2E1 mRNA fragment (680 -1653) was added, and it did not prevent the formation of the complex (Fig. 4, lane  2). Similar results were obtained with cytoplasmic proteins from H4IIEC3 cells treated with 0.1 M insulin (data not shown). All these results suggest that the RNA-protein complex is located in this 1-191 fragment of the CYP2E1 mRNA.
In order to map more precisely the RNA-protein complex, several consecutive or overlapping antisense oligonucleotides covering the 1-191 fragment were synthesized. Each antisense oligonucleotide was complementary to part of the fragment sequence (Fig. 5A). Preincubation of the full-length 32 P-CYP2E1 mRNA with antisense oligonucleotides AS3 to AS6 did not prevent complex formation (Fig. 5B, 5th to 8th lanes). In contrast, oligonucleotides AS2 (3rd lane) and, to a lesser extent, AS1 (2nd lane) prevented protein binding. Similar results were obtained using cytoplasmic proteins from cells treated or not by insulin (data not shown). We next used RNA draw software to predict the putative secondary structure of the 1-191 sequence. Such an analysis allowed us to predict that the secondary structure of the CYP2E mRNA includes a hairpin loop overlapping the fragments complementary to AS1 and AS2 (Fig. 6). It was thus hypothesized that this putative hairpin loop could be the binding site for the cytoplasmic proteins. Therefore, we synthesized a 16-mer antisense oligonucleotide complementary to this hairpin loop (AS1,2). As shown in Fig.  5B (4th lane), AS1,2 was indeed able to block the formation of the CYP2E1 mRNA/cytoplasmic protein complex. These data suggested that the hairpin-forming sequence overlapping regions 1 and 2 was necessary for the formation of the RNAprotein complex. In order to test whether this hairpin alone was sufficient to bind cytoplasmic proteins, a conventional gel mobility shift assay was performed using a 32 P-labeled 16-mer "sense" RNA covering the loop region (147-163) and H4IIEC3 cytoplasmic proteins. As shown in Fig. 7, we observed the formation of the 16-mer RNA-protein complex that was displaced by a specific antisense oligomer (AS1,2) and was unaffected by a nonspecific antisense oligomer (AS6, lane 4).
The functional consequences of the presence of a binding site for cytoplasmic proteins in the CYP2E1 mRNA was then studied. Indeed, RNA-binding proteins could play a role in the destabilization of this mRNA by insulin, although the results shown above suggest that the binding of the cytoplasmic proteins to the CYP2E1 mRNA is not significantly altered by insulin. However, other functions of these proteins could be the target of the hormone. Consequently, additional experiments were undertaken to demonstrate the actual functionality of the 16-mer sequence in the destabilization of CYP2E1 mRNA by insulin. The ability of this sequence to function in a heterologous environment was first tested. We introduced heterologous oligonucleotides in the coding sequence of the Renilla luciferase because this gene is not regulated by insulin and because the half-life of its mRNA is comparable with that of CYP2E1 (halflife 8 h, data not shown). To introduce the oligonucleotide that corresponds to the 16-mer sequence 147-163 of the CYP2E1 mRNA, we used the unique AflII restriction site in the coding region of the Renilla luciferase gene. This insertion was made in both directions: a sense direction and a control antisense direction. An 18-mer insertion was used in order to not modify the reading frame of the luciferase mRNA. Following transfection of the different constructs into H4IIEC3 cells, the effects of insulin (0.1 M) on the Renilla luciferase mRNA expression were evaluated by Northern blot analysis (Fig. 8). Insulin did not modify the wild type Renilla luciferase mRNA expression. Interestingly, it significantly decreased (p Ͻ 0.05) the amount of chimeric Renilla luciferase mRNA containing the sense insertion. In contrast, the chimeric mRNA containing the antisense insertion was unaffected by this treatment.
We next asked whether the 16-mer sequence was indeed able to mediate the insulin effect within the endogenous CYP2E1 mRNA. To investigate this question, we introduced the antisense oligonucleotide (AS1,2) in the H4IIEC3 cells and quantified the CYP2E1 mRNA in the presence or absence of insulin. Calmodulin mRNA, which is not regulated by insulin, was used to correct loading variability. The oligonucleotide AS1,2 was coupled to penetratin in order to ensure an efficient delivery of the oligonucleotides in the cultured cells (32,33). Fig. 9 shows that insulin significantly decreased (p Ͻ 0.01) the amount of CYP2E1 mRNA, in untreated H4IIEC3 cells and in cells treated with free penetratin. In contrast, in cells transfected with the penetratin-linked AS1,2, the CYP2E1 mRNA expression was not affected any longer by insulin, suggesting that the 16-mer sequence was necessary for the insulin action on the amount of the endogenous mRNA. DISCUSSION The regulation of mRNA stability constitutes a critical control step in cellular mRNA level and has been linked to sequence elements and specific RNA-protein interactions. The aim of the present study was to identify the molecular mechanisms involved in rat CYP2E1 mRNA destabilization by insulin as shown previously in the laboratory (1). Insulin is known to destabilize the mRNAs encoding for phosphoenolpyruvate carboxykinase, glucose transporter 4, and glycogen synthase (34), but the pathways by which the stability of these mRNAs is controlled by insulin remain unknown. We have shown that mRNA sequence from base 115 to 131 following the ATG was necessary and sufficient to convey the negative regulation by insulin. This is the first identification of a functional RNA sequence that is the target of insulin action. This sequence constitutes a hairpin loop and binds cytoplasmic proteins. The binding of these proteins does not appear to be regulated by insulin.
Indirect evidence suggests that the human CYP2E1 mRNA is regulated by insulin similarly to the rat CYP2E1 mRNA. Indeed, a hairpin loop with a sequence displaying 88% identity to the one identified in the rat CYP2E1 mRNA is present in the human CYP2E1 mRNA (UUCCCAUCAUCGGGAA). In addition, whereas levels of CYP2E1 were very low to undetectable in human lymphocytes from healthy control subjects, they were elevated in lymphocytes from patients with insulin-dependent diabetes mellitus (35). Therefore, we speculate that this hairpin sequence could be involved in the human CYP2E1 mRNA stability.
In contrast, Peng and Coon (36) showed that insulin downregulated rabbit CYP2E1 mRNA at the post-transcriptional level and that the 3Ј-UTR is involved in the regulation of this mRNA by this hormone. The 16-mer sequence that we identified on the rat CYP2E1 mRNA is not conserved in the rabbit CYP2E1 mRNA. Moreover, the CYP2E1 3Ј-UTR is much longer in the rabbit sequence (480 bp) than in the human (152 bp) or rat (130 bp) sequences. Therefore, it is possible that the rabbit CYP2E1 3Ј-UTR contains a specific sequence involved in the negative regulation by insulin.
Most of the sequences involved in mRNA decay are located in the 3Ј-UTRs and the 5Ј-UTRs (37). One of the best examples of post-transcriptional regulation is observed in iron homeostasis (38). Iron starvation of cells induces high affinity binding of the iron regulatory protein (IRP) to the RNA iron-responsive element (IRE). This binding results in the repression of translation of ferritin mRNA because of IRP binding to the IRE located in the 5Ј-UTR of this sequence. This binding also leads to the stabilization of the transferrin receptor transcript due to the IREs located in its 3Ј-UTR. The IRP-IRE binding presumably prevents the 3Ј-initiated degradation of the transferrin receptor complex. In both cases, i.e. ferritin and transferrin receptor mRNAs, the critical step is the regulation of the binding of IRP to the IRE. Another well studied example of sequence-dependent degradation is the AU-rich element (ARE, AUUUA) present in the 3Ј-UTR of many protooncogenes, lymphokines and cytokines (39,40). A family of ARE-binding proteins has been characterized that influences the decay of ARE-containing mRNAs (40). Some of these proteins appear to promote destabilization of the RNA upon binding to the ARE, whereas the interaction of other proteins with the ARE contributes to the RNA stabilization. The mechanisms of post-transcriptional of regulation of the CYPs have been partly characterized in the case of two mouse CYP2a5 and -1a2. The cis-acting element and trans-acting factors, involved in the stabilization of these CYP mRNAs, have been identified. A protein from mouse liver specifically binds to the 3Ј-UTR of CYP2a5 mRNA; this binding is increased after pyrazole treatment and is associated with mRNA stabilization and elongation of the poly(A) tail (8,41). Likewise, two nuclear proteins bind specifically to the 3Ј-UTR of CYP1a2 mRNA, and their binding is altered by a typical inducer of CYP1a2, 3-methylcholanthrene (6). In these cases too, the critical regulatory step is probably the binding of proteins to RNA target sequences.
Whereas the best characterized RNA regulatory sequences are localized in the 5Ј-and 3Ј-UTR, the regulatory hairpin loop of the CYP2E1 mRNA is localized within the coding region. The mechanisms whereby these sequences exert their regulatory function is still unclear. A few cases of destabilizing sequences in the open reading frame of a number of early response gene mRNAs have been reported, such as in c-fos, c-myc (42)(43)(44), and ␤-tubulin transcripts (45), but the precise localization of the binding sites in these transcripts was not identified.
The binding site of cytoplasmic proteins in the CYP2E1 mRNA constitutes a 16-mer hairpin loop. Both the sequence and structure of regulatory RNA motifs are critical for their function as described for the IRE (46) and for the CYP2a5 mRNA (7). One example of the importance of such a secondary structure is the 3Ј end of histone RNA for which the presence and integrity of a stem loop is essential for the post-transcriptional regulation of the mRNA (47,48).
The contribution of the hairpin sequence to insulin regulation was evaluated in this study. No difference in the localization or the pattern of the RNA-protein complex was observed in experiments performed with cytoplasmic extracts from H4IIEC3 cells treated or not by insulin. Thus, in the case of CYP2E1 mRNA, the critical regulatory steps are probably not the binding of proteins to the target RNA sequence. However, our results (Figs. 8 and 9) demonstrate the actual functional implication of this 16-mer hairpin loop in the CYP2E posttranscriptional regulation. The activity of certain RNA-binding proteins can be modulated by means of conformational changes induced by amino acid modifications such as phosphorylation. It has been reported that proteins involved in the control of mRNA half-life are often post-translationally modified by phosphorylation, which is important for their RNA binding ability and/or function (49,50). Raffalli-Matthieu et al. (6) showed that phosphatase treatment of the nuclear extracts reduced the binding of the 46-kDa protein to the CYP1a2 mRNA. This suggests that the binding activity of the protein may depend on its phosphorylation status or on some intermediary factors necessary for binding. Recently, Ostrowski et al. (51) showed that insulin-induced phosphorylation of the heterologous nuclear ribonucleoprotein K alters its interaction with its target RNA sequences. In the case of CYP2E1, because the binding of the regulatory protein to its target RNA is not regulated by insulin, it is possible that other functions are regulated, such as protein-protein interactions mediating the destabilization of the mRNA. Alternatively, this RNA-binding protein may constitute a scaffold upon which the actual target of insulin can bind and exert a RNA destabilization function.
The insulin receptor is a membrane tyrosine kinase receptor that activates two major signaling pathways, the mitogenactivated protein kinase and the phosphatidylinositol 3-kinase pathways. Woodcroft and Novak (52,53) demonstrated that CYP2E mRNA expression was enhanced in primary cultured rat hepatocytes by either lowering the insulin concentration or completely excluding insulin from the medium. Moreover, insulin produced a concentration-dependent decrease in CYP2E1 mRNA levels by a post-transcriptional mechanism, confirming our own results and a correspondence between insulinreceptor phosphorylation and the decrease in CYP2E1 mRNA levels observed in primary cultured rat hepatocytes (54). Furthermore, they demonstrated that inhibition of phosphatidylinositol 3-kinase, p70S6 kinase, and Src kinase prevented the decrease of CYP2E mRNA levels (54). However, alternative results were obtained by Sidhu et al. (55). These authors conclude that the insulin-mediated suppression of CYP2E1 mRNA expression is not associated with the activation of phosphatidylinositol 3-kinase-dependent pathway, but rather that the CYP2E1 response is probably linked to the activation of intracellular stress pathways. Thus, the mechanism of insulin action remains controversial. The characterization of a regulatory protein that could be the target of insulin will help to determine the actual mechanism of insulin action.
In conclusion, we showed that binding of cytoplasmic proteins to a 16-mer hairpin loop was responsible for CYP2E1 mRNA turnover by insulin. Our future investigations will consist in identifying the proteins that bind to this sequence. These proteins seem predominantly expressed in the cytoplasm because experiments performed with nuclear extracts from H4IIEC3 cells treated or not by insulin did not allow detection of the CYP2E1 mRNA-protein complex (data not shown). Identification and characterization of these proteins could provide a considerable insight into the mechanisms involved in regulated mRNA turnover by insulin, not only for CYP expression but also for other genes that are post-transcriptionally regulated by this hormone.