P-Glycoprotein (P-gp) encoded by the MDR gene is one of the main factors in multidrug resistance. Its expression in cancer cells, which compromises cancer outcome, can be enhanced by some stress signals. Energy depletion, frequently observed in malignant cells, was shown to induce chemoresistance and could be one of these signals. To test this hypothesis, we studied the effect of glucose deprivation on P-gp expression in a rat hepatoma cell line (Fao). Incubation of Fao cells with a glucose-free medium enhanced P-gp mRNA and protein expression in a time-dependent manner, up to 400% at 40 h. This effect was associated with a stimulation of [3H]vinblastine efflux by P-gp despite impaired glycosylation. It was reproduced by inducers of endoplasmic reticulum stress response, such as 2-deoxyglucose (DG), tunicamycin, and thapsigargin. P-gp mRNA induction by DG was preceded by an increase in activator protein binding activity, c-Jun expression, and phosphorylation. In contrast, nuclear factor-κB binding activity was unaffected by DG. The antioxidant N-acetylcysteine partially reversed the increase in P-gp mRNA and protein levels induced by DG, as well as the enhancement of c-Jun phosphorylation and activator protein binding activity. Finally, transient transfection of the cells with a deleted mutant of c-Jun, Tam 67, abolished the DG-induced P-gp mRNA expression and mdr1b promoter activation. In conclusion, glucose deprivation enhances P-gp expression and transport function in liver cancer cells. This effect is mediated by endoplasmic reticulum stress response and involves MDR transcriptional induction through c-Jun activation. These results emphasize the importance of glucose metabolism in chemoresistance.
The resistance to multiple chemotherapeutic drugs is one of the main causes of poor cancer outcome. One of the best-characterized forms of MDR3 is enhanced drug efflux mediated by ABC family transporters such as P-gp (1). P-gp is a 150–180-kDa membrane phosphoglycoprotein encoded by the MDR gene, which functions as an energy-dependent drug transporter with broad specificity (1). High P-gp expression has been described in many tumors, notably in hepatocellular carcinoma in humans and rodents (2, 3). P-gp is also expressed on the apical membranes of various normal epithelia and at the blood-tissue barrier in endothelia (4), but its physiological function is still under debate.
The regulation of P-gp expression has been widely investigated to better understand its pathophysiological implications. In particular, it has been shown that P-gp expression is regulated through transcriptional and post-transcriptional mechanisms and by various endogenous and environmental stimuli that evoke stress response (5). The mediators of these regulations have been only partially elucidated, but ROS (6) and transcriptional factors such as AP-1 and NF-κB have been implicated (5, 7, 8, 9).
Numerous studies have shown that energy depletion can enhance tumor survival and aggressiveness (10, 11, 12). Low oxygen tension and glucose deprivation occur frequently during tumor development because of inadequacies between vascular supply and metabolic requirement (10, 11). Moreover, malignant transformation is accompanied by an increased rate of glycolysis (13), which may exacerbate glucose depletion in tumor cells. Interestingly, it has been shown that glucose depletion leads to resistance to the P-gp substrate doxorubicin (14), but no data are available on the regulation of P-gp by glucose deprivation.
We postulated that glucose depletion could enhance chemoresistance through induction of P-gp expression. To confirm this hypothesis, we studied P-gp expression in a rat hepatoma cell line subjected to glucose deprivation. Our results demonstrate that glucose depletion induced P-gp expression and strongly suggest that these effects are mediated by an ER stress involving ROS. P-gp induction results from MDR transcriptional activation, in which c-Jun/AP-1 activation seems to play a key role.
MATERIALS AND METHODS
Cell Lines and Culture.
The highly differentiated rat hepatoma cell line Fao (15), the human hepatoma cell line HepG2 (16), and the SV40-transfected aortic rat endothelial cell line SVAREC (17) were grown in pyruvate-supplemented DMEM free of glucose (G0) or containing 4.5 g/l glucose (G4.5; Life Technologies, Inc.). DMEM was supplemented with 5% newborn calf serum in SVAREC or 10% FCS in Fao and HepG2 cells.
Plasmids and Transfection.
The mdr1b promoter/luciferase construct (pmdr1), including the mdr1b promoter sequence from −320 to +150, was generated from the luciferase pgGL3 basic vector (Promega). This sequence contains the basal promoter function (18) and an AP-1 site at −263. The plasmid pCMV-67, which contains TAM67, a deleted form of c-Jun missing the NH2-terminal transactivation domain but retaining the DNA binding and leucine zipper domains (19) was a gift from Dr. Michael Birrer, National Cancer Institute (Rockville, MD).
Fao cells seeded in 6-well plates at a density of 0.35 × 106 cells/well were transiently transfected the following day with 2 μg/well of either pCMV-67 or p-CMV mixed with Lipofectamine Plus reagent (Invitrogen, Cergy-Pontoise, France) and cotransfected with 0,4 μg/well pmdr1 for luciferase assays. Six to 24 h after the transfection, cells were transferred to medium containing 4.5 g/l glucose or treated with 6 mg/ml DG for 16–24 h. Mdr1b and c-Jun mRNA levels were measured by RNase protection assay. Luciferase transactivation assays were performed as recommended by the manufacturer (Promega).
Reagents and Antibodies.
Dipyridamole, DRB, DG, TU, THA, curcumin, NAC, and H2O2 (30% solution, w/w) were purchased from Sigma France. Mouse monoclonal anti-P-gp antibody C219 and anti-β-actin were obtained from Calbiochem and Sigma, respectively. Rabbit polyclonal anti-c-Fos and anti-c-Jun and mouse monoclonal anti-P-c-Jun (phosphorylation on Ser63) were purchased from Santa Cruz Biotechnology (Tebu). Rabbit polyclonal anti-GRP78 antibody was obtained from Stressgen (Tebu).
Preparation of Total Protein Extracts.
Cells were first washed with PBS and then lysed in 50 mm Tris-HCl (pH 8.0), 150 mm NaCl, 0.1% SDS, 0.5% sodium deoxycholate, 1% Nonidet, 1 μg/ml aprotinin, and 100 μg/ml PMSF. Nuclear debris was removed by centrifugation, and the supernatant was frozen at −80°C.
Preparation of Nuclear Extracts.
Cell pellets were washed with PBS and then lysed in hypotonic buffer (40 mm HEPES, 3 mm MgCl2, 20 mm KCl, 2 mm EDTA, 1 mm PMSF, 1 mm DTT, and 1 μg/ml each of leupeptin, aprotinin, and pepstatin) containing phosphatase inhibitors (2 mm sodium orthovanadate, 1 mm glycerophosphate, 40 mm sodium fluoride). The pellets were resuspended in 100 μl of high-salt buffer (hypotonic buffer supplemented with 500 mm KCl and 20% glycerol), and samples were shaken on a roller for 30 min at 4°C. After centrifugation, the supernatants (nuclear extracts) were kept in aliquots at −80°C until analyses.
Western blotting of P-gp was performed as described previously (20). Total protein extracts were immunoblotted with the anti-P-gp antibody C219 (1 μg/ml), anti-β-actin (1:6000 dilution), and in some instances, with anti-GRP78 (1:1000 dilution). Immunoblotting of c-Fos, c-Jun, and P-c-Jun was performed on nuclear extracts with the appropriate antibodies, each diluted to 1:5000. Immunoreactive bands were visualized by enhanced chemiluminescence (Amersham Pharmacia France) and quantified by the NIH program.
Electrophoretic Mobility Shift Assay.
Single-stranded oligonucleotides corresponding to the upper and lower strands of the AP-1-binding element (upper: 5′-TAAGTATGACTCACCAGGGAC-3′) or NF-κB-binding element (upper: 5′-TATGTCTGGGGAATTCCAGCTCC-3′) present on mdr1b promoter (consensus sequence in bold) were end-labeled with [γ-32P]ATP and then annealed to produce 32P-labeled AP-1 or NF-κB DNA probes. Nuclear proteins (10 μg) were incubated in a total volume of 20 μl of binding buffer [20 mm HEPES (pH 7.9), 60 mm KCl, 5 mm MgCl2, 20% glycerol, 0.1 mm PMSF, 0.1 μg/ml leupeptin and aprotinin, 2 mm levamisole] containing 2 μg of poly(dI-dC) and 32P-labeled DNA probe (20,000 cpm) for 20 min at 4°C. Samples were loaded on a 5% polyacrylamide gel in 44 mm Tris borate-2.4% glycerol-1 mm EDTA (pH 8.0) buffer and electrophoresed at 200 V for 1.5 h at 20°C.
RNase Protection Assay.
The rat mdr1b and GAPDH probes were synthesized by in vitro transcription as described previously (20). Rat mdr1a probe was constructed by reverse transcription-PCR using primers at positions 1882 and 2132 in the 5′–3′ direction of rat mdr1a cDNA (GenBank accession no. AF286167) and cloned into PGEM Easy vector (Promega France). Rat c-Jun probe was synthesized from the PBSK plasmid containing a cDNA fragment of c-Jun (gift from B. Lardeux, INSERM U327, Faculté X. Bichat, Paris, France). The protected mRNA products were analyzed by 6% denaturing PAGE (42% urea), and radiolabeled bands were visualized and quantified by electronic autoradiography (Instant Imager; Packard). The results were normalized by GAPDH mRNA levels. For mRNA stability measurements, DRB, an analogue of actinomycin D, was added to the medium 24 h after cell treatment. RNA extraction was performed at 3 h (time zero) after DRB and 3 and 6 h later to measure the decreases in mdr1b and mdr1a levels.
[3 H]VBL Accumulation.
P-gp transport activity was assessed by measuring the decrease in accumulation of its labeled substrate, [3 H]VBL (21). Briefly, Fao cells cultured in 24-well plates were incubated with 37 nm [3 H]VBL for 1 h at 37°C and then lysed in 0.5% Triton to count radioactivity. We added 20 μm of a P-gp inhibitor, dipyridamole, to some wells for each condition. Data were calculated as pmol/mg of [3 H]VBL protein that accumulated, which was decreased when the efflux activity of P-gp was increased and reversed by dipyridamole.
Induction of Protein and mRNA P-gp Expression by Glucose Deprivation in Fao Cells.
Fao cells cultured under control conditions expressed high levels of P-gp protein and mRNA. As shown in Fig. 1,A, the P-gp protein content increased progressively with the duration of glucose depletion. During this period, the apparent molecular size of P-gp was reduced from 170 to ∼150 kDa. Mdr1b and mdr1a mRNA levels also progressively rose during the course of glucose depletion (Fig. 1 B), and the increase was comparable between the two transcripts.
Role of Altered Glycosylation in P-gp Expression and Function.
We then examined whether other treatments that induce glucose deficiency and/or inhibition of protein glycosylation could affect P-gp expression. For this purpose, Fao cells were treated with an inhibitor of glucose metabolism, DG (22) or an inhibitor of protein N-glycosylation, TU (23). Fig. 2,A shows that treatment with 6 mg/ml DG or 1 μg/ml TU for 24 h reproduced the increase in P-gp protein content and the decrease in apparent molecular size observed after glucose deprivation. In response to each treatment, mdr1b and mdr1a levels (Fig. 2,B) also rose markedly. These data confirm that P-gp stimulation is related to events that impair protein N-glycosylation. The functional relevance of P-gp induction was examined by measuring P-gp transport activity through changes in [3 H]VBL accumulation. This latter averaged 3.7 ± 0.3 pmol/mg of protein in untreated cells and decreased by ∼50% under glucose depletion or DG or TU treatment (Fig. 2 C). After inhibition of P-gp transport with 20 μm dipyridamole, [3 H]VBL accumulation was enhanced to 10.1 ± 0.7 pmol/mg of protein in the untreated cells and was unaffected by the different treatments (data not shown). These results demonstrate that overexpression of P-gp is associated with increased transport activity under glucose depletion.
Role of UPR in P-gp Expression.
Treatments that alter protein N-glycosylation induce a stress response called UPR (24, 25, 26, 27). UPR is also activated by ER calcium depletion resulting from treatment with the calcium ATPase inhibitor THA (28). We therefore examined P-gp expression in response to THA treatment. When Fao cells were treated for 24 h with 1 μg/ml THA, P-gp protein levels rose to 380 ± 140% above levels of untreated cells, whereas mdr1b and mdr1a levels rose to 500 ± 185% and 400 ± 66% of controls, respectively (n = 3; P < 0.001). We also determined whether P-gp induction was preceded by overexpression of the UPR-inducible chaperone GRP78 (26, 29). After 6 h of treatment with 6 g/l DG, GRP78 protein levels rose significantly over control levels (151 ± 9% of control; n = 8; P < 0,05), whereas P-gp protein levels remained unchanged at this time (104 ± 7% of control). These results strongly indicated a role for UPR in P-gp overexpression induced by glucose deprivation in Fao cells.
Effects of Glucose Depletion and UPR Inducers on P-gp Expression in Other Cell Lines.
UPR also affected P-gp expression in the human hepatoma cell line HepG2 because P-gp protein levels were enhanced to 270 ± 46% and 219 ± 48% over control levels in cells treated for 24 h with 6 g/l DG and 1 μg/ml TU, respectively (n = 3; P < 0.05). This effect was not restricted to hepatoma cells: glucose depletion and UPR dramatically increased P-gp protein and mRNA levels in SVAREC endothelial cells, as shown in Table 1.
MDR Transcript Stability in Response to DG.
UPR-induced P-gp mRNA expression may result from increased mRNA stability and/or from MDR transcriptional activation. We examined the effect of DG on mdr1 transcript stability by measuring mdr1b and mdr1a decay rates after addition of the transcription inhibitor DRB to DG-treated or untreated Fao cells. Fig. 3 shows that the slopes of mRNA decay were comparable for mdr1b and mdr1a and were unaffected by DG treatment. The levels of the two transcripts decreased with an absolute half-life of ∼6 h, whereas GAPGH mRNA levels were not significantly affected during this time. These data indicate that P-gp mRNA stability was unaffected by DG and that P-gp overexpression likely resulted from MDR transcriptional activation.
Role of the Transcriptional Factors AP-1 and NF-κB in DG-Induced P-gp Response.
Several transcriptional factors, including AP-1 and NF-κB, are induced by glucose depletion and UPR (24, 30, 31). Because putative binding sites for AP-1 and NF-κB have been found in the mdr1b promoter (5, 32), we tested whether AP-1 and NF-κB binding activities were affected by DG treatment. Whereas mdr1b and mdr1a levels progressively increased from 2 to 16 h of DG treatment (Fig. 4,A), NF-κB binding activity was not significantly modified by DG treatment in the same conditions. By contrast, AP-1 binding activity was significantly increased, and the increase was roughly parallel to the changes in mdr1 (Fig. 4,B). For each factor, the specificity of DNA binding was assessed by an oligonucleotide competition study (Fig. 4 C).
Effect of DG on the AP-1 Constituents c-Fos and c-Jun.
The proteins c-Fos and c-Jun are the major constituents of the dimeric AP-1 complex (33). As shown in Fig. 5,A, AP-1 binding was reduced by anti-c-Fos and anti-c-Jun antibodies and supershifted with anti-P-c-Jun antibody, indicating the presence of c-Fos, c-Jun, and its active form, P-c-Jun in the AP-1 complex (33, 34). In the course of DG treatment, c-Fos protein levels were unaffected regardless of the time of treatment (Fig. 5,B), whereas c-Jun and particularly P-c-Jun protein levels markedly increased (Fig. 5 C). Similarly, c-Jun mRNA levels rose dramatically during DG treatment, averaging 378 ± 42% and 211 ± 16% of controls after 2 and 16 h of incubation, respectively (n = 3; P < 0.01). The results clearly show that DG treatment enhances c-Jun expression and phosphorylation.
Role of Oxidative Stress in DG-induced AP-1 Activation and P-gp Expression.
ROS are potential sensors of UPR-induced gene activation (24) and can induce both AP-1 and P-gp expression (35, 36, 37), so that they could be the link between UPR response, AP-1 activation, and P-gp induction. We first examined the effect of the oxidant H2O2 on P-gp expression. After 24 h of treatment with 2 mm H2O2, P-gp protein content significantly increased up to 213 ± 39% above control levels, and mdr1b and mdr1a levels rose to 191 ± 17% and 173 ± 26% above controls, respectively (n = 4; P < 0.001). We next examined the role of two antioxidants NAC (38) or curcumin (39) on the DG-induced increase in P-gp expression. Pretreatment of the cells with NAC had little effect on P-gp expression in control conditions but inhibited by ∼40% the DG-induced increases in P-gp protein content (Fig. 6,A) and mdr1b and mdr1a levels (Fig. 6,B). Similar inhibitory effects were obtained when we used curcumin (Fig. 6). Under the same conditions of DG incubation, NAC pretreatment did not significantly affect c-Jun mRNA (Fig. 7,A) and protein levels (Fig. 7,B) but reduced by ∼40% the increase in P-c-Jun levels (Fig. 7,B) and modestly reduced DG-induced AP-1 binding activity (Fig. 7 C). These results suggest that oxidative stress contributes to the DG effects by increasing c-Jun phosphorylation in parallel to MDR induction.
Effect of a Dominant-Negative Mutant of c-Jun (TAM67) on DG-Induced MDR Transcriptional Activation.
To confirm that c-Jun is involved in MDR activation, we examined the effects of transient transfection with the dominant-negative mutant of c-Jun, TAM67, on mdr1b mRNA levels and on AP-1 containing mdr1b promoter activation. In Fao cells overexpressing TAM67, the DG-induced increase in mdr1b mRNA expression was reduced by 42%, whereas basal expression was not significantly affected (Fig. 8,A). Because c-Jun activation induces its own expression (33), we assessed the efficacy of TAM67 transfection by determining its inhibitory effects on c-Jun mRNA expression. As expected, c-Jun mRNA levels were reduced in TAM67-transfected cells regardless of the conditions of treatment (Fig. 8,B). As shown on Fig. 8 C, mdr1b promoter-driven luciferase reporter was activated by DG treatment, and this activation was inhibited by cotransfection with TAM67. All of these results demonstrate that c-Jun activation is directly involved in DG-induced MDR transcriptional activation.
This study shows that glucose depletion induces P-gp expression in hepatocarcinoma cells. The finding that the increase in P-gp mRNA levels preceded that of P-gp protein levels without changes in P-gp transcript stability during DG treatment argues for MDR transcriptional activation in response to glucose depletion.
Stimulation of P-gp expression during glucose depletion was associated with impaired glycosylation, a situation known to alter protein folding within the ER and cause a cell stress response termed UPR (22, 23, 24, 25). Accordingly, these effects were reproduced by use of two well-known UPR inducers: TU, an inhibitor of protein N-glycosylation (23); and THA, an ER calcium ATPase inhibitor (26) that impairs protein folding without affecting glycosylation (28). These findings and the observation that stimulation of the UPR-inducible chaperone GRP78 (26, 29) preceded P-gp induction strongly suggest that UPR mediates P-gp up-regulation.
UPR involves the coordinate transcriptional activation of genes that encode ER chaperones, death signals, and proteolytic factors (27, 29) and could then activate P-gp at the transcriptional level. Available data in the literature indicate that several transcription factors can be activated by UPR (27), including the ATF/CREB family (40) and the ubiquitous and inducible factor NF-κB (30). UPR and hypoglycemia were also shown to induce AP-1 binding activity (31) and c-Jun phosphorylation (41), the major effector of the transcription factor AP-1 (33). Putative AP-1 and NF-κB binding sites are present on human and rodent mdr1 promoter (5, 32). However, UPR did not increase NF-κB activity in our conditions, suggesting that this factor was not likely involved in UPR-induced P-gp activation. By contrast, we found that AP-1 was activated by UPR as evidenced by the DG-induced increase in AP-1 binding activity, in c-Jun expression, and phosphorylation. Moreover, overexpression of the dominant-negative mutant of c-Jun, TAM67, blunted the DG-induced increase in mRNA levels and activation of AP-1 containing mdr1b promoter. These results provide direct evidence that c-Jun/AP-1 complex plays a crucial role in P-gp gene activation during UPR, in accordance with the positive regulation of human MDR by AP-1 described previously (9).
We demonstrated in the present study that oxidative stress is involved in UPR-mediated P-gp induction. Indeed, the oxidant H2O2 enhanced P-gp expression, whereas antioxidants partially reversed the DG-induced increase in P-gp overexpression. These data are supported by previous observations showing that redox modulation influences P-gp expression in hepatic cells (6). It is noteworthy that a broad spectrum of unrelated P-gp-inducing agents trigger the production of ROS, suggesting that oxidative stress plays a major role in mdr1b up-regulation (6, 7, 8, 42). In agreement with previous studies showing that ROS could activate the AP-1 and JNK signaling pathway (35, 37, 43), we observed that antioxidants inhibited the DG-induced increase in AP-1 binding activity and c-Jun phosphorylation. Taken as a whole, our findings suggest that ROS could mediate the UPR effects on P-gp induction by activating JNK and AP-1. Nevertheless, it is likely that other pathways are implicated in MDR activation by DG because reversal of P-pg induction by antioxidants was only partial.
The induction of P-gp by UPR is unexpected because most glycosylated proteins undergo biosynthesis repression and enhanced proteolysis during UPR (27, 44). These are adaptive processes aimed at lowering the load of client proteins the ER must process (44). Likewise, we observed that the activities of the two membrane glycoproteins alkaline phosphatase and γ-glutamyltransferase were reduced by glucose depletion or by DG treatment.4 The present study shows that P-gp overexpression is associated with increased efflux activity of the protein in response to UPR. This demonstrates that newly synthesized P-gp remains active despite defective N-glycosylation, at variance with other membrane glycoproteins (23, 45, 46). The fact that P-gp regulation follows that of ER chaperones suggests its implication in a specific function linked to UPR.
Despite the uncertain physiological function of P-gp, some hypotheses can be put forward. First, several endogenous substrates of P-gp, including amphipathic peptides, have been described (47, 48). P-gp could participate to UPR either directly by transporting some misfolded proteins or indirectly by interacting with chaperones involved in UPR. Second, P-gp has been shown to confer resistance to caspase-dependent apoptosis induced by various stimuli (47, 49, 50). Because UPR triggers caspase-dependent apoptosis (25, 27), the induction of P-gp could provide an antiapoptotic process and counterbalance UPR-mediated cell death if cell injury is survivable.
Beside the physiological implications of our findings, we first demonstrated that the metabolic state of the cell can modify P-gp expression in rodent as well as in human cells and then influence MDR. The in vivo relevance of our findings are reinforced by the observation that fasting induced liver mdr1a expression in mice (51). Our results could have clinical implications because the glucose supply of the tumor and the nutritional state of the patient may influence efficacy of chemotherapy. P-gp stimulation by UPR was not restricted to malignant cells, but was also observed in endothelial cells. This endothelial stimulation could contribute to enhanced chemoresistance in tumors by decreasing drug accessibility. Our findings provide therapeutic perspectives, involving the control of UPR-dependent P-gp activation as a new target for adjuvant treatment of cancer. In this respect, the administration of antioxidants could negatively control the UPR-mediated cascade leading to P-gp induction. This hypothesis is reinforced by previous publications showing the role of antioxidants on cancer outcome (52, 53).
In conclusion, our results show that UPR triggered by glucose deprivation enhances P-gp expression. This effect involves a transcriptional MDR induction through c-Jun activation, partly mediated by ROS. These original findings emphasize the role of glucose supply in conferring the phenotype of MDR in cancer cells.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The work was supported by the CRIT firm.
The abbreviations used are: MDR, multidrug resistance; P-gp, P-glycoprotein; ROS, reactive oxygen species; AP-1, activator protein; NF-κB, nuclear factor-κB; ER, endoplasmic reticulum; DRB, 5,6-dichlorobenzimidazole riboside; DG, 2-deoxy-d-glucose; TU, tunicamycin; THA, thapsigargin; NAC, N-acetyl-l-cysteine; P-c-Jun, phosphorylated c-Jun; PMSF, phenylmethylsulfonyl fluoride; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; VBL, vinblastine; UPR, unfolded protein response.
D. Laouari, personal observations.
|.||P-gp expression (% of control) .||.|
|.||G0 .||DG .|
|P-gp||324 ± 119a||364 ± 66a|
|Mdr1b||856 ± 141b||584 ± 114a|
|Mdr1a||1575 ± 390b||1012 ± 321a|
|.||P-gp expression (% of control) .||.|
|.||G0 .||DG .|
|P-gp||324 ± 119a||364 ± 66a|
|Mdr1b||856 ± 141b||584 ± 114a|
|Mdr1a||1575 ± 390b||1012 ± 321a|
P < 0.05;
P < 0.01.
We thank Drs. D. Bernuau, B. Lardeux, and P. Codogno and Prof. L. Baud for their interesting suggestions in the course of these studies and C. Leroy for technical advice.