If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
To whom correspondence should be addressed: Laboratory of Molecular Biology, University of Ghent, K. L. Ledeganckstraat 35, B-9000 Ghent, Belgium. Tel.: 32-9-264-51-31; Fax: 32-9-264-53-48
Affiliations
Laboratory of Molecular Biology, University of Ghent, and the Flemish Institute for Biotechnology, B-9000 Ghent, Belgium
∗ This research was supported by the Interuniversitaire Attractiepolen, the Fonds voor Geneeskundig Wetenschappelijk Onderzoek, the Nationale Loterij, the Vlaams Instituut voor de Bevordering van het Wetenschappelijk-technologisch Onderzoek in de Industrie, and the Vlaams Instituut voor Biotechnologie. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore by hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Treatment of the mouse fibrosarcoma cell line L929 with tumor necrosis factor (TNF) induces necrotic cell death. A crucial step in the cytotoxic action mechanism of TNF involves perturbation of mitochondrial functions leading to the formation of reactive oxygen intermediates (ROI). L929 cells have energy requirements adapted to a high proliferation rate. Glutamine (Gln) is utilized as a major energy source and drives mitochondrial ATP formation, while glucose is mainly converted to lactate through glycolysis. We investigated the role of the bioenergetic pathways involved in substrate utilization on the cytotoxic action of TNF and established a link between Gln oxidation and TNF-induced mitochondrial distress. Omission of Gln from the medium desensitizes the cells to TNF cytotoxicity, while the lack of glucose in the medium does not alter the TNF response. Sudden depletion of Gln from the culture medium results in a sharp decline in mitochondrial respiration in the cells, which might explain the decreased TNF responsiveness. However, when L929 cells are adapted to long term growth under conditions without Gln, these so-called L929/Gln- cells have restored respiration, but they still display a decreased sensitivity to TNF cytotoxicity. Thus the TNF responsiveness of L929 cells depends on bioenergetic reactions that are specifically involved in the oxidation of Gln. This is further confirmed by the desensitizing effect of specific inhibitors of these Gln-linked enzyme reactions on TNF cytotoxicity in the parental cells, but not in the L929/Gln- cells. Analysis of the induction of mitochondrial ROI formation by TNF in parental and L929/Gln- cells suggests that the effect of Gln on the sensitivity to TNF cytotoxicity involves a mechanism that renders the mitochondria more susceptible to TNF-induced mediators, resulting in enhanced ROI production and accelerated cytotoxicity.
is a pleiotropic cytokine mainly produced by activated macrophages. Besides its role in the host defense against microorganisms and bacterial pathogens, TNF is involved in the pathology of various diseases, such as the systemic inflammatory response syndrome(
). Furthermore, TNF is specifically cytotoxic for many types of transformed cells, especially in the presence of interferon. Most TNF-mediated activities, including cytotoxicity, are initiated by ligand-induced cross-linking of the p55 TNF receptor; only in T-lymphocytes has a role of the p75 TNF receptor in cell proliferation and in cytotoxicity been unambiguously demonstrated(
). A major step in the cytotoxic mechanism is the formation of ROI in the mitochondria. Their crucial role was demonstrated by the interference of specific inhibitors of the electron transport chain with necrotic cell response (
) can be explained by the inability to generate ROI in the mitochondria.
Besides oxygen, another important parameter that affects mitochondrial functionality is the availability of energy substrates. Normal and malignant cells often exhibit different metabolic requirements (
); the exuberant proliferation rate of transformed cells demands an adapted energy metabolism. Therefore, the amino acid Gln, instead of the Glc-derived pyruvate, is used preferentially as a substrate for ATP production by oxidative phosphorylation. The underlying adaptations of the intermediary metabolism include (i) an increased activity of the mitochondrial matrix enzyme glutaminase, converting Gln to glutamate, (ii) formation in the mitochondria of the citric acid cycle intermediate α-ketoglutarate via transamination of glutamate, and (iii) the presence of malic enzyme, generating intramitochondrial pyruvate(
Here we describe the result of the above changes in mitochondrial enzyme composition and substrate utilization to TNF cytotoxicity. We show that L929 cells use Gln and not Glc as the major energy substrate and that this particular energy metabolism promotes the cytotoxic response of the cells to TNF. Our results also demonstrate that the dependence of TNF cytotoxicity on Gln is not due to the overall rate of mitochondrial respiration per se. Enzymatic pathways specifically utilized in mitochondrial oxidation of Gln appear to sensitize the mitochondria to TNF-induced perturbation of their activity and thereby amplify the resulting production of cytocidal ROI.
MATERIALS AND METHODS
Cell Culture
L929, a murine fibrosarcoma cell line, was grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum, 2 mM Gln, 100 units/ml penicillin, and 0.1 mg/ml streptomycin. Adaptation of the cells to Gln-free culture conditions was performed by a stepwise decrease in Gln concentration in the medium to zero over a 2-month period. The resulting L929/Gln- population was routinely maintained in Dulbecco's modified Eagle's medium free of Gln (and glutamate) (Gibco Bio-Cult, Paisley, UK), supplemented with 10% dialyzed heat-inactivated fetal calf serum and antibiotics. Dulbecco's modified Eagle's medium without Gln and Glc was supplemented with Gln (2 mM), Ino (100 μM), dThd (40 μM) and Urd (100 μM), or Glc for use in short term culture experiments. Suspension cultures of the normally adherent-growing L929 and L929/Gln- cells were set up when required for flow-cytometric experiments. To that end, cells were harvested from adherent cultures grown in tissue culture flasks by trypsinization at 37°C, washed, and resuspended in culture medium. The cells were seeded in 90- or 30-mm diameter bacterial-grade Petri dishes at 3-5 × 105 cells/ml and incubated overnight at 37°C in a humidified, 5% CO2 incubator. Under these conditions, the cells no longer adhered to the plastic surface and remained in suspension. TNF sensitivity of the cells was not extensively altered in these suspension cultures(
). The preparation had a specific activity of 1.2 × 108 IU/mg protein and contained <4 ng of endotoxin/mg of protein. TNF activity was determined as described previously(
), using an international standard TNF preparation (code 88/532, obtained from the Institute for Biological Standards and Control, Potters Bar, UK) as a reference. Unless otherwise mentioned, TNF was added at time zero to cell cultures at a final concentration of 1000 IU/ml. Cycloheximide was dissolved in culture medium and added to a final concentration of 50 μg/ml. PI (Becton-Dickinson, Sunnyvale, CA) was prepared as a 100 × stock solution (3 mM) in NaCl/Pi and stored at 4°C. DHR123 (Molecular Probes, Eugene, OR) was prepared as a 5000 × stock solution (5 mM) in dimethyl sulfoxide and stored at −20°C. Amino-oxyacetate (0.2 mM), L-glutamate-γ-hydrazide (5 mM), 6-diazo-5-norleucine (1 mM), and acivicine (1 mM) were purchased from Sigma and dissolved in medium. Butylated hydroxyanisole (Sigma) was prepared as a 1000 × stock solution (100 mM) in ethanol.
Lactate Production
The medium of exponentially growing adherent cultures was replaced with fresh medium, and the cells were cultured for a further period of 10 h at 37°C in a humidified CO2 incubator. The amount of lactate released by the cells into the supernatant was assayed with a commercial diagnostic test (Sigma) on the basis of lactate dehydrogenase activity. The lactate content was expressed in mM normalized to a standard of 106 cells/well.
Oxygen Consumption
Oxygen consumption of intact cells was measured with a Clark-type microelectrode in a 1-ml cell (Strathkelvin Instruments, Glasgow, UK) equilibrated at 37°C. Cells, grown in suspension culture overnight, were harvested, counted, and resuspended in fresh medium saturated with O2. O2 consumption was followed for 5-15 min, depending on the rate of consumption. Background O2 consumption in cell-free medium was 0.25% per min. S.D. on independent preparations was <5%.
PI Exclusion Assay and Quantitative Flow Cytometry
Cell samples were taken from suspension cultures, and PI was added 3-10 min before analysis on an EPICS 753 flow cytometer (Coulter Electronics, Luton, UK). Data acquisition was triggered on a forward angle light scatter signal. Cell debris and multicell aggregates were electronically gated out. The PI dye was excited with a water-cooled argon-ion laser (250 mW) at 488 nm. PI fluorescence was measured above 610-nm wavelength using a long pass filter. Routinely, 5000 cells were analyzed. S.D. on duplicate samples was consistently <5%.
Measurement of ROI Formation by Flow Cytometry
The ROI probe DHR123 was added to suspension cultures at the onset of the experiment. Cell samples were taken from the suspension cultures at regular time intervals and analyzed on an EPICS 753 flow cytometer as described before(
). The DHR123-derived R123 fluorescence was excited with a water-cooled argon-ion laser (250 mW) at 488 nm and detected between 515- and 550-nm wavelength. Cell debris and multicell aggregates were electronically gated out. R123 fluorescence was gated on the viable cell population and was measured on 3000 viable cells/sample.
RESULTS
L929 Fibrosarcoma Cells Use Gln as a Respiratory Substrate
L929 cells are routinely maintained in synthetic medium containing 25 mM Glc and 2 mM Gln. In order to evaluate the balance between the utilization of Gln and Glc as substrates for oxidative phosphorylation, the rate of oxygen consumption and lactate production was measured in cells maintained for 18-24 h before analysis in the presence of either substrate alone or combined. Thus, the measurement of these parameters on cells maintained on either Gln or Glc or both allows to evaluate the contribution of each substrate to the oxidative metabolism and to glycolysis, respectively. Since Glc is an essential precursor of ribose moieties for nucleic acid biosynthesis, Ino, dThd, and Urd were added to the medium, where Glc was omitted in order to replace the anabolic function of Glc(
). Gln, on the contrary, is a nonessential amino acid and can be omitted from the culture medium.
As shown in Fig. 1, the presence of both Glc and Gln, i.e. normal growth conditions, supports a metabolism that combines a high rate of respiration with a high production of lactate. A high lactate release indicates that a considerable part of the pyruvate generated from Glc by glycolysis does not enter the citric acid cycle but is reduced to lactate. Nevertheless, the cells show a high rate of respiration. Apparently, the substrate that fuels this high respiration is Gln, since cells maintained exclusively with Gln showed an equal or even slightly increased respiration compared with the control condition. As expected, lactate production in these cells dropped to background levels due to the absence of substrate suitable for glycolysis. When Gln was omitted, Glc being the only metabolic substrate available to the cells, both the rates of lactate release and of oxygen consumption dropped below 50% of the control value. Apparently, under this culture condition, part of the pyruvate that otherwise is reduced to lactate now enters the citric acid cycle to drive respiration. Finally, when both Gln and Glc were omitted from the medium, all cells died within 48 h (results not shown), indicating that no other energy-providing substrate was available to the cells under this condition. These results clearly indicate that in L929 cells Gln is metabolized in the mitochondria through the citric acid cycle and hereby fuels oxidative phosphorylation, whereas the metabolism of Glc is limited to glycolysis, resulting in a high lactate production. These characteristics are as expected for a tumor cell-type energy metabolism as described before for other cell lines(
Figure 1:Analysis of the energetic pathways used by L929 cells. For lactate production (open bars), 2 × 106 cells were seeded in 6-well dishes in the respective media for 18 h. Then the medium was refreshed in all wells, and the lactate released in the medium was measured after 10 h of incubation. The 100% value of lactate equals a concentration of 5.2 mM/106 cells. For oxygen consumption (filled bars), cells were cultured in suspension cultures for 18 h in the respective media, collected by centrifugation, and resuspended in a small volume of O2-saturated medium to measure oxygen consumption with a Clark-type oxygen electrode at 37°C. The 100% value of oxygen consumption equals 115 ng oxygen/min/106 cells. ITU refers to Ino, dThd, and Urd.
Sensitivity to TNF-induced Cytotoxicity Is Modulated by the Oxidative Substrate
Cells kept overnight in the presence of Glc, Gln, or both were assayed for their sensitivity to the cytotoxic activity of TNF. Cytotoxicity was measured on individual cells by flow cytometry on the basis of loss of plasma membrane integrity, detected by means of uptake of the fluorogenic exclusion dye PI. Loss of plasma membrane integrity has previously been shown to represent an early and irreversible step in the necrotic process induced by TNF in L929 cells (
). As shown in Table 1, replacement of Glc by its anabolic, but energetically neutral, counterparts Ino, dThd, and Urd (thus leaving Gln as the sole energetic substrate) did not affect the sensitivity of the cells to TNF-induced cytotoxicity. However, when Gln was omitted from the culture medium, a markedly decreased cytotoxic response was observed. This desensitization was equally pronounced when cells were treated with TNF in the presence of cycloheximide. This protein synthesis inhibitor is known to enhance the response of L929 cells to the cytotoxic action of TNF, presumably by preventing the synthesis of rescue factors that counteract TNF activity(
). Furthermore, the sensitizing effect of cycloheximide was observed equally in the presence or absence of Gln, suggesting that TNF signaling leads to comparable rescue activity in either culture condition.
Adaptation of L929 Cells to Growth in the Absence of Gln Reversibly Alters Their TNF-induced Cytotoxic Response
As shown above, Gln deprivation results in a drop in the rate of oxygen consumption exceeding 50% (Fig. 1) and hence in mitochondrial activity. Since earlier reports established that mitochondrial electron transport activity is a prerequisite for the TNF cytotoxic mechanism in L929 cells(
), the reduced mitochondrial respiration observed in the absence of Gln might explain the diminished responsiveness of these cells to TNF. In order to verify this explanation, L929 cells were adapted to growth on Glc as sole energy-providing substrate. They were grown with gradually lowered concentrations of Gln, the concentration of Glc remaining unaltered, so that after 2 months a population was obtained that stably grew on Gln-free medium. These adapted L929/Gln- cells grew slower and showed a more elongated, fibroblast-like morphology as compared with the scattered, dedifferentiated morphology of the parental cell population. These changes in morphology are in agreement with previous reports that utilization of Gln may affect the phenotype of cells(
). Also metabolically, L929/Gln- cells differ from the parental cells in that Glc fully supported the respiration, reaching an equal or even slightly elevated rate of oxygen consumption as compared with parental L929 cells; the only difference was that marginal amounts of lactate were produced by the adapted cells. Thus adaptation restored the metabolization pathway of the cells driven by oxidation of Glc instead of Gln. However, despite the restored mitochondrial activity, L929/Gln- cells still were desensitized with respect to TNF cytotoxicity (Fig. 2), similar to the desensitization after a brief Gln deprivation. This impeded TNF responsiveness is a stable feature and cannot be ascribed to an altered genotype of the adapted cells, as demonstrated by the nearly complete restoration of TNF response after 2 weeks of culture in the presence of Gln (Fig. 2). Thus, switching between a sensitive and an insensitive phenotype appears to depend on enzymatic pathways involved in the oxidation of Gln as opposed to Glc and not on the overall electron flow generated by either substrate in the mitochondria.
Figure 2:Comparison of TNF-induced cell death in parental L929 and L929/Gln- cells. TNF-induced cell death in parental L929 cells (♦), L929/Gln- cells (•), and L929/Gln- cells that subsequently had been cultured in the presence of Gln for over 2 weeks (■) is shown. Cells were grown in suspension cultures 18 h before TNF treatment (1000 IU/ml); cell death was detected by PI uptake and flow cytometry.
Specific Inhibitors of the Gln Metabolism Mimic the Effect of Gln Removal
To further substantiate the involvement of Gln-specific, enzymatic pathways in the sensitization of L929 cells to the cytotoxic activity of TNF, we analyzed the interference with TNF-induced cytotoxicity by specific inhibitors of Gln hydrolysis and catabolism. The inhibitors were chosen on the basis of their interference with sequential steps of Gln oxidation; they were assayed on L929 cells, cultured in the presence of both Gln and Glc, as well as on L929/Gln- cells, the latter in the absence of an exogenous source of Gln (Table 2). Acivicine, a Gln analogue that cannot be metabolized and thus acts as a Gln antagonist, drastically decreased the TNF sensitivity of L929 cells. Similarly, inhibition of glutaminase activity by L-glutamate-γ-hydrazide (
) or by 6-diazo-5-norleucine decreased the cytotoxic response of L929 cells to TNF. Glutaminase catalyzes the conversion of Gln to glutamate and is mainly present in the mitochondria. Mitochondrial overexpression of this enzyme is a common feature in tumor cells(
). In a final step, the glutamate formed is converted to the citric acid cycle intermediate α-ketoglutarate by transamination, whereby aspartate or pyruvate act as ammonium acceptors. Inhibition of the transaminase enzymes by amino-oxyacetate again drastically reduced the cytotoxic response of L929 cells to TNF. Contrary to acivicine, L-glutamyl-hydrazide, or 6-diazo-norleucine, amino-oxyacetate also significantly reduced the cytotoxic response of L929/Gln- cells, suggesting that transamination of other amino acid intermediates contributes to the cytotoxic response of L929/Gln- cells. Apparently, transamination of intermediates derived from Gln, being the most abundant amino acid, or from other amino acids in the absence of Gln, represents a key step in the sensitization of L929 cells to the cytotoxic action of TNF.
Gln Metabolism Enhances TNF-induced Mitochondrial ROI Production
We described previously that TNF treatment of L929 cells results in the production of intracellular ROI. These ROI represent an essential step in the TNF-cytotoxic pathway and are of mitochondrial origin(
). Hence, bioenergetic pathways involved in mitochondrial oxidation of Gln might affect ROI formation and thus explain the sensitizing effect exerted by this amino acid on TNF-induced necrotic cell death. In order to verify this hypothesis, we analyzed the levels of ROI generated upon TNF treatment in L929 and L929/Gln- cells. ROI production was probed with the cell-permeable, fluorogenic dye DHR123. Oxidation of the nonfluorescent DHR123 by the reactive oxygen species H2O2 or superoxide anion yields the cationic, fluorescent R123, which is subsequently sequestered by active mitochondria(
). Analysis by flow cytometry of the DHR123-derived R123 fluorescence in TNF-treated cultures showed that the number of cells exhibiting increased ROI levels (Fig. 3A) as well as the amplitude of ROI increment (Fig. 3B) relative to untreated controls were dramatically decreased in the L929/Gln- culture deprived of exogenous Gln. The number of cells exhibiting increased ROI levels evolved in parallel with the cytotoxic response for both cell types (Fig. 3C). The lower ROI levels generated by TNF in the absence of Gln provide an explanation why these cells require more time before sufficient ROI have accumulated to inflict cell death. The link between ROI production and cell death, as reported previously in L929 cells, was also confirmed in L929/Gln- cells by the complete arrest of both responses following addition of the free oxygen radical scavenger butylated hydroxyanisole (Fig. 3, A-C). These results indicate that sensitization of L929 cells to TNF cytotoxicity by oxidative metabolism of Gln involves a mechanism that renders the mitochondria more susceptible to a cytosolic TNF signal, resulting in an enhanced activation of the ROI-producing pathway, which in turn leads to accelerated cytotoxicity.
Figure 3:Analysis of TNF-induced ROI response and cell death. Analysis was performed in parental L929 cells (•) and L929/Gln- cells (■) in the absence (full line) or presence (dashed line) of the radical scavenger butylated hydroxyanisole. A, percentage of TNF-responsive viable cells (DHR123 profiles of TNF-treated cells minus DHR123 profiles of untreated cells at the corresponding time points), showing an increased DHR123-derived R123 fluorescence intensity after TNF treatment. B, TNF-induced increment in ROI formation: mean R123 fluorescence intensity of TNF-responsive viable cells compared with untreated control cells. C, TNF-induced cell death: percentage of PI-positive cells in TNF-treated cultures as compared with untreated cultures. The arrows indicate the time of addition of butylated hydroxyanisole to the cell cultures.
Binding of TNF to its p55 receptor in malignant, nonlymphoid cells triggers a complex signal transduction mechanism leading to cytostasis, necrosis, or apoptosis, depending on the cell type and the physiological condition. In this report we analyzed the necrotic death induced by TNF in the mouse fibrosarcoma cell line L929 and show that one of the factors that modulates the responsiveness of the cells to the cytotoxic activity of TNF is the availability of energy substrate and, more specifically, the species of respiratory substrate used by the cell at the time of TNF treatment. Neoplastic cells have been reported to have a changed metabolic behavior characterized by the preferential use of Gln, instead of Glc, as the major energy source. Gln is oxidized at a high rate in tumor mitochondria, while Glc is primarily converted to lactate. We established that L929 cells grown in culture indeed use Gln preferentially as a substrate for oxidative metabolism and as a corollary produce high amounts of lactate from Glc. Furthermore, we demonstrated that modulating the cellular energy metabolism by altering substrate availability in the culture medium resulted in significant changes in TNF response. More specifically, omission of Gln from the medium shortly before TNF treatment desensitized the cells to TNF cytotoxicity. This treatment led to a sharp decrease in electron flow through the mitochondrial electron transport chain. Since several lines of evidence previously established that TNF cytotoxicity in L929 cells requires functional mitochondria (
), this loss of mitochondrial activity might explain the reduced TNF responsiveness of Gln-deprived L929 cells. Therefore, we adapted L929 cells to grow without Gln, thus enforcing the use of Glc as major respiratory substrate. The resulting stable L929/Gln- population had the appearance of differentiated fibroblasts, as opposed to the dedifferentiated, scattered morphology of parental L929 cells; moreover, it showed a reduced proliferation rate and in fact nearly behaved like untransformed fibroblasts. Also the metabolic parameters, namely low lactate release and high oxygen consumption, corresponded to those of normal cells. These parameters indicate that Glc supports in L929/Gln- cells a restored electron flow through the mitochondrial electron transport chain. Nevertheless, these adapted cells still displayed the TNF-resistant phenotype observed after short term depletion of Gln. We therefore conclude that a factor responsible for sensitivity of the TNF-cytotoxic response in the presence of Gln resides in the activity of enzyme systems linked to the oxidative Gln metabolism and not in the overall mitochondrial electron flow. This conclusion was confirmed by the mimicking effect observed with inhibitors that block, at various steps, the enzymatic oxidation of Gln through the citric acid cycle in the mitochondria. Thus the use of Gln as an energy substrate facilitates TNF signal transduction leading to necrotic cell death.
Previous reports from this laboratory have shown that TNF induces excess ROI production in the mitochondria of L929 cells and that these ROI are directly cytocidal and/or necessary for downstream events leading to cell death. In the absence of Gln, this ROI response was markedly attenuated, whereby both the absolute levels of ROI generated as well as the rate of appearance of cells exhibiting increased ROI levels were abated. Apparently, the oxidation of Gln creates a metabolic condition in the mitochondria that facilitates mitochondrial production of excess ROI upon TNF stimulation and that enhances the cytotoxic response of the cells to TNF. This facilitation is not based on the overall rate of electron flow in the mitochondria, since L929 cells using Glc or Gln as a respiratory substrate showed similar rates of oxygen consumption. Possibly, the kind of substrate used affects the formation of multienzyme complexes of a higher order (
) that are differentially sensitive to regulatory mechanisms (in)activated by TNF. Alternatively, by-products of Gln oxidation through the citric acid cycle, such as citrate(
), a precursor of fatty acid and cholesterol biosynthesis, may affect upstream signaling events and amplify the activation signal delivered to the mitochondria. Further experiments will be needed to clarify this mechanism.
Since both the use of Gln as an oxidative substrate and sensitivity to the cytotoxic action of TNF are tumor cell-specific features (the first facilitating the latter), the validity of a causal link between tumor cell bioenergetics and tumor cell responsiveness to TNF cytotoxicity has to be considered. However, as expected for a very pleiotropic cytokine like TNF, there is no evidence for a strict correlation between the use of Gln as a respiratory substrate and the sensitivity to TNF cytotoxicity. For example, the TNF sensitivity of cells that die in an apoptotic mode, was not diminished after Gln omission (not shown). This may be related to the early inactivation of mitochondrial activity observed during programmed cell death(
), rendering an active contribution of mitochondria impossible. In contrast, when mitochondria actively contribute to the cytotoxic process, such as in L929 or WEHI 164 cl 13 cells (not shown), Gln enhanced the TNF-cytotoxic response. Hence we may conclude that the tumor cell-characteristic use of Gln as an oxidative substrate contributes to the TNF responsiveness of those tumor cells in which mitochondria play an active role in the cytotoxic process. Since Gln is abundantly present in body fluids and since its concentration may alter during the progression of neoplastic disease, maintaining high concentrations in circulation may, for certain tumors, positively affect the therapeutic value of TNF.
Acknowledgments
We thank Drs M. Vincx and J. Vanfleteren for technical assistance.
95323-8&id=gr1.jpg){kind=link}
95323-8&id=gr2.jpg){kind=link}
95323-8&id=gr3.jpg){kind=link}