Astroglial expression of the P-glycoprotein is controlled by intracellular CNTF

Expression of P-gp in primary astrocyte culture

Astrocytes in culture expressed P-glycoprotein as early as 24 h after plating. The RT-PCR performed on a fragment of the P-gp transcript yielded a 167-bp product (Fig. 1a), present in all astrocyte samples and in the MES-SA/MX2 cells used as a positive control. At the protein level, the results varied over time: until the 4^th day in vitro (div), the antibody raised against P-gp detected a band at about 120 kDa; from the 6^th div a doublet was detected at about 160–170 kDa; finally, at 15 div this doublet became predominant and its expression increased (Fig. 1b).

Effects of molecules promoting astrogliosis on P-gp expression

Among the agents known to activate astrocytes ("astrogliotic factors") that were tested for modulating P-gp expression, most failed to alter it (Table 1). This was in particular the case for dBcAMP, IL-1β, LPS, NGF, RA, rhIFNα, rhTGFα and TNFα. In contrast, all the cytokines that belong to the IL-6 family induced a 50% or more increase in expression of P-gp in long-term astrocyte culture. This effect was dose-dependent and was modulated by α receptor subunits, when relevant. For instance, rrCNTF was inefficient when used alone at 30 ng/ml in long-term cultures. However, an increase in the cellular content of P-gp was observed in a significant and reproducible manner when 100, and even further 250 ng/ml of rCNTF were used (Fig. 2a). As previously described for the effects of rCNTF on other astroglial markers of differentiation [25], addition of 200 ng/ml of c-myc-sCNTFRα induced a potentiation of the effects, and an increase of P-gp cellular content was then observed with the lowest concentration of rCNTF studied (10 ng/ml), and maintained for all other concentrations used (data not shown). Addition of rhLIF and rhCT-1, in the culture medium provoked an increase of about 1.5 to 2-fold of P-gp cellular content (Fig. 2b). Addition of rmIL-6 at concentrations below 40 ng/ml in the culture medium was inefficient, but a significant increase of P-gp intracellular content was observed when this concentration was used (Fig. 2c). Addition of rhsIL-6Rα to the culture medium induced a major potentiation of the effects, since, in these conditions, all concentrations used triggered a significant increase in P-gp content (Fig. 3).

Selective blockade of P-gp transporter activity was studied using two different specific blockers, S9788 and verapamyl. When used at concentrations of 5 μM and 10 pM, a complete blockade of the effects of all IL-6 family cytokines was observed. Even the presence of the specific α receptor subunit for IL-6 and CNTF failed to reverse the neutralising effects of S 9788 (Fig. 5) and verapamyl (data not shown).

Besides the IL-6 family of cytokines, the only compound to trigger P-gp overexpression in astrocytes was IFNγ (Fig. 6a). This IFNγ effect was very likely direct since hLIF05, which inhibits the LIF receptor pathway, could not reverse it.

Regulation of P-gp expression in CNTF-/- astrocytes

The progressive maturation of P-gp during development and the regulation of its expression by members of the IL-6 family of cytokines are two features that P-gp shares with CNTF in astrocytes (see discussion section). Moreover, like P-gp, CNTF regulation is very specific, since, among different molecules that normally trigger astrogliosis, only those belonging to the IL-6 family provoked an increase of CNTF (Fig. 7). We, therefore, explored the possibility of a functional link between the two molecules by analysing whether the presence of endogenous CNTF in astrocytes was required for the observed modulation of P-gp expression by IL-6 type cytokines and IFN-γ. This was tested by analysing the effects of the cytokines in astrocytes taken from CNTF knockout mice. In these cultures, the modulation of P-gp by LIF, CT-1 or IL-6 (even in presence of its specific α receptor subunit) was completely lost (Fig. 4). In contrast, addition of rrCNTF (250 ng/ml) produced a significant increase of P-gp intracellular level. In the CNTF -/- astrocytes, addition of IFN-γ to the medium triggered an increase of P-gp cellular level comparable to that observed in wild-type astrocytes, underlining its direct effect on P-gp modulation (Fig. 6b).

P-gp and CNTF may be involved together in a specific type of astrocytic activation

The recent discovery of P-gp expression on astrocytic endfeet [17, 18] and its induction after a stress [5] has raised the issue of a role of this protein in astrocytes. To address this issue, we have undertaken, here, an analysis of the expression of P-gp over time during development and of its stimulation in mature cells. The first feature in this study is the striking similarity between the results obtained for astroglial P-gp and those previously gathered on the astroglial cytokine CNTF. Indeed, P-gp and CNTF seem to share parallel modulations of expression both during development and under specific stimulations. First, like CNTF [26], P-gp is not detectable in the early stages of postnatal development in the central nervous system. Both proteins are detected from P7, and their concentration levels gradually increase, reaching a plateau at P28 [15, 26]. Results obtained in in vitro astroglial cultures are also indicative of a parallel maturation of the two proteins. P-gp, both at mRNA and protein levels, was expressed in our study as soon as after 24 hours in culture, but Western blot studies revealed that the protein detected in those early cultures did not have the 170 kDa mature form [15], that became predominant only after 14 days, the point at which astrocytes exhibit a mature phenotype (see discussion in 25). A very similar time course of expression was observed previously for CNTF [25]. Before 7 days in vitro, immature astrocytes did not express detectable levels of CNTF, whereas at 14 div, when astrocytes form a confluent monolayer and express high levels of GFAP, they exhibit a significant intracellular content of CNTF.

In vitro, astrocytes activation experiments further indicate a parallel regulation of the two proteins. Even though in vivo models of astrogliosis have revealed a concomitant overexpression of both proteins [5, 27], most factors that readily induce astroglial activation in vitro fail to increase their expression [28, 29] and the present study). Indeed, for both proteins, only IFNγ (in the study by Carroll and colleagues, 31, for CNTF) and the IL-6 family cytokines could trigger a quantifiable increase in expression in astrocytes (this study).

All these results suggest that the overexpression of CNTF and P-gp participate in the definition of a particular stage of astroglial differentiation. So-called astrogliosis has long been considered as a discrete "activation" stage of astrocytes, on the basis of its characteristic morphological features. However, more recent studies exploring the expression of specific molecules have questioned this view, drawing attention to the fact that reactive gliosis varies qualitatively and quantitatively depending on both the nature of the injury and the microenvironment of the injury site (see references and discussion 30 and 31). The nature of the injury and the microenvironment necessary for the differentiation stage of astrocytes that involves CNTF and P-gp overexpression is, at this point, a matter of speculation. It is interesting to note, nevertheless, that, in agreement with the in vitro data, intra-cerebral administration of adenovirus recombinant for CNTF produced an increase in the astroglial content of endogenous CNTF without triggering massive "astrogliotic" morphological changes [32].

Intracellular concentration of CNTF as a modulator of P-gp expression in astrocytes

Out of the present study, the relationship between CNTF and P-gp in astrocytes appears as a functional link rather than as a mere co-regulation. Indeed, one main result obtained in this study was the apparent requirement of an intracellular level of CNTF for the expression of P-gp to be modulated by IL-6-type cytokines, since this modulation was not seen in cells that did not express the CNTF gene. The only exogenous cytokine that remained effective in CNTF-/- cells was CNTF itself. This may be explained by the fact that exogenous CNTF, when bound to its receptor, is internalised into the cells [33]. As discussed by these authors, endocytosed CNTF may be active inside the cell, as has been demonstrated for neurotrophins/Trks complexes (see 34 for a review). Through this mechanism, therefore, restoration of an intracellular content of CNTF may become sufficient to trigger a increased P-gp expression.

How an increase in the cellular content of CNTF increases the expression of P-gp was not directly addressed in this study. Nevertheless, a number of elements help to suggest a working hypothesis. A direct transcriptional role of CNTF on the P-gp gene promoter is unlikely for three reasons. First, it has been well demonstrated that the intracellular signalling systems triggered by CNTF are essentially similar to those triggered by LIF [35, 36]. Why would LIF not be effective in CNTF-/- astrocytes when CNTF is, would be difficult to explain in case of a direct transcriptional effect. Second, our results demonstrate that the control of P-gp expression requires a modulation of the intracellular content of the cytokine. Third, the stimulation of P-gp expression by cytokines was blocked with antagonists of the activity of the transporter. This indicates that the regulation of P-gp depends upon its own activity and suggests the existence in astrocytes of a positive feedback of the molecule upon its own expression. Such a suggestion is concordant with the well demonstrated positive feedback of P-gp activity upon its expression in various tumor cells, in the presence of non-organic substrates like cytotoxic drugs [37, 38].

In a mechanistic model based upon this hypothesis, endogenous astroglial CNTF would most likely not appear as a usual cytokine activating signalling pathways after binding to a transmembrane receptor system, but as a "sensor", itself modulated by extrinsic cues. The changes induced in the intracellular concentration of this "sensor" might affect the activity of the P-gp transmembrane transporter and, eventually, lead to a modulation of its gene expression (Fig. 8). Such a scheme suggests a physiological role for CNTF, a cytokine which, because of the absence of a signal peptide and the lack of a quantifiable release in the culture medium of cells that synthesise it [32, 39, 40], was not thought to play a physiological role in the absence of a lesion [41]. Whether P-gp is the only target of the effects of CNTF, or just the first one identified is obviously an open question at this point.

One directly related issue is the role of P-gp in astrocytes. Taking into account the functions of P-gp in other cells, in the protection against potentially harmful chemicals and metabolites, it is plausible that the transporter plays an important role in the response against cell stress [42], as suggested by its increase in reactive astrocytes. In addition, one may consider another role that has been demonstrated for P-gp and other transporters of the mdr family, namely their ability to transport various organic molecules through cell membranes, including various cytokines. Indeed, P-gp has been shown to transport interleukin-2 (IL-2) and interleukin-4 (IL-4) through cell membranes [9, 21, 43]. Whether, among other substrates, P-gp can similarly transport CNTF out of astrocytes is a tempting hypothesis that will require further studies.

Culture conditions

Astroglial cultures were prepared as previously described [44]. After 14 days in culture, the astrocytes had formed a confluent monolayer. Serum containing medium (Minimal Essential Medium containing 2 mM Glutamin, Essential Amino acids, 0.03% glucose, Penicillin-Streptomycin, foetal calf serum -FCS- 10%) was removed and serum free medium added for 24 h. To investigate the effects of various factors on P-gp expression, the following compounds were added for 24, 48 or 72 h: rrCNTF (30, 100, 250 ng/ml); rhLIF (10, 30 ng/ml); rhCT-1 (10 ng/ml), TNFα (10, 100 ng/ml), NGF (10 ng/ml), retinoic acid (10^-7 M, 10^-8 M), dBcAMP (0.5 mM), IFNβ (600 U/ml), LPS (1 μg/ml), TGFα (50 ng/ml), rhIL-1β (10, 50 ng/ml), rmIL-6 (10, 20, 30, 40, 50 ng/ml) with or without rhsIL-6R (100, 200, 300, 400, 500 ng/ml), rhIFN-γ (300 U/ml). In the experiments with soluble CNTFRα, 200 ng/ml of myc-sCNTFRα (kindly provided by Ralph Laufer, IRBM, Italy) were added to the medium without FCS after 14 days in vitro. Thirty minutes later, rrCNTF was added at 0, 10 (4.4 × 10^-10 M), 30, 50 (2.2 × 10^-9 M), 100 or 250 ng/ml (10^-8 M).

Verapamil (5 μM, Sigma) and S9788 (10 pM, Servier, France) were used as blockers of P-gp. They were added in the culture medium 12 h before rrCNTF, rhCT-1 and rhLIF, used at the concentrations indicated above.

Since our preliminary results suggested that ligands of the LIF-receptor (LIFR) may modulate P-gp, a specific inhibitor of this receptor (hLIF05) [45] was added to the culture medium in some experiments (0.5 and 5 μg/ml), 30 min before addition of potential stimulating agents. MES-SA/MX2 cells (ATCC, Biovalley, France) were used as a control. This cell line is a mitoxanthrone-resistant derivative of the human uterine sarcoma cell line MES-SA, that displays features of overexpression of the two classical multidrug resistance P-gps. The cells were cultured in a medium containing 1:1 mixture of Waymouth's MB 752/1 medium and McCoy's 5 a medium, 90%, foetal bovine serum, 10%.

In all cases, the medium was removed at the end of the experiments and each dish was rinsed three times with HBSS (Hank's Balanced Salt Solution, Seromed, Germany). The cells were collected by scraping into 62.5 mM Tris HCl (pH 6.8), 2 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, 0.5 % Triton X-100 and 2.3 % sodium dodecyl sulfate.

Biochemical analysis

Total protein content was determined by the BCA protein assay kit (Pierce, Illinois, USA) with bovine serum albumin as a standard. The proteins were analysed by Western blotting. Briefly, samples were boiled for 5 min after addition of 10 % glycerol, 5 % mercaptoethanol (or 5 % Dithiotreitol for the MAPK-P) and 5 % bromophenol blue, then lysates were electrophoresed on 7.5% SDS polyacrylamid gels. Gels were blotted on nitrocellulose, blocked for one hour in 5 % non fat dry milk in TBS-T (20 mM Tris, pH 7.5/ 500 mM NaCl/0.1% Tween 20) and then probed overnight at 4°C, first with a polyclonal anti-P-glycoprotein antibody (mdr, Ab-1, Immunotech, France, 1/200), then with a monoclonal anti-α tubulin antibody (Sigma-Aldrich, France, 1/5000). After washing with TBS-T, membranes were incubated with a horseradish peroxidase-conjugated donkey anti-rabbit secondary antibody followed by the enhanced chemiluminescent reaction (Amersham, Sweden), according to the manufacturer's instructions. The levels of P-gp were measured by densitometry and normalised to the total protein loaded in each lane.

To limit variations in their processing extracts from the control and all experimental conditions were treated in parallel, on a single sheet, for each specific experiment. In addition, to evaluate the variability between specific experiments, control extracts were subsequently reloaded together on a single nitrocellulose membrane and processed together. Statistical analysis used one-factor ANOVA and unpaired t-test.

PCR experiment

For RT-PCR, RNA isolation was performed with the Trizol method (Life technologies, Cergy-Pontoise, France). PCR was carried out with cDNA derived from 2 μg of RNA, 2.5 unit of AmpliTaq Polymerase and reaction kits (Superscript preamplification systems, Gibco BRL, France) in a final volume of 50 μl. Each cycle of PCR included 30 sec of denaturation at 94°C, 1 min of primer annealing at 55°C, and 2 min of extension/synthesis at 72°C and one cycle of 72°C for 10 min. MDR1-specific sequences were amplified by using the sense-strand primer CCCATCATTGCAATAGCAGG (residues 2596–2615) and the antisense-strand primer GTTCAAACTTCTGCTCCTGA (residues 2733–2752) [46], which yield a 167-bp product. Each primer was added at 10 μM per reaction.