The nucleus, a site for signal termination by sequestration and inactivation of p42/p44 MAP kinases

Upon stimulation of many types of cell surface receptors, p42/p44 MAPKs are activated by the signaling cascade Ras>Raf>MEK (Guan, 1994; Lewis et al., 1998; Reiser et al., 1999; Robinson and Cobb, 1997). In resting cells p42/p44 MAPKs are anchored in the cytoplasm via association with MEK (Bardwell et al., 1996; Bardwell and Thorner, 1996; Fukuda et al., 1997b). Following activation, they dissociate from the cytoplasmic anchoring complex and enter the nucleus, where they accumulate during the mid-G₁ phase (Fukuda et al., 1997b). Although p42/p44 MAPKs nuclear translocation has been shown to be essential for growth factor-induced DNA replication (Brunet et al., 1999; Kim-Kaneyama et al., 2000) and cell transformation (Robinson et al., 1998), the mechanisms of nuclear p42/p44 MAPKs import and export are still largely unknown. Results support the notions that p42/p44 MAPKs freely diffuse across nuclear pores (Adachi et al., 1999; Fukuda et al., 1997c) or that entry is catalyzed by p42/p44 MAPKs dimerization (Khokhlatchev et al., 1998). Nuclear export of p42/p44 MAPKs has been shown to be dependent upon a nuclear export sequence and could be mediated by MEK (Adachi et al., 2000). Finally, it was found that the nuclear accumulation of p42/p44 MAPKs requires persistent activation of p42/p44 MAPKs that is necessary and sufficient for the neo-synthesis of short-lived nuclear anchors (Lenormand et al., 1998). Consequently, nuclear accumulation is observed solely after stimulation by mitogenic agonists, which induce long-term activation of p42/p44 MAPKs that persists for up to 6 hours of G₁ progression (Kahan et al., 1992; Meloche et al., 1992).

Although the mechanisms of activation of p42/p44 MAPKs are relatively well described, little is known about the mechanisms of inactivation. Different phosphatases have been implicated in the rapid inactivation of p42/p44 MAPKs such as the serine/threonine-specific phosphatase PP2A (Alessi et al., 1995; Sohaskey and Ferrell, 1999) and several related tyrosine-specific phosphatases PTP-SL, STEP, He-PTP and LC-PTP, which show a good specificity towards p42/p44 MAPKs (Oh-hora et al., 1999; Pettiford and Herbst, 2000; Pulido et al., 1998; Saxena et al., 1999; Zuniga et al., 1999). However, dual specificity phosphatases can specifically dephosphorylate both the tyrosine and the threonine residues of MAPKs. These MAPK phosphatases (MKPs) have different subcellular localizations and diverse specificities for the MAPKs (Camps et al., 2000; Keyse, 2000). Among them, MKP1 and MKP2 are induced by the p42/p44 MAPK pathway (Brondello et al., 1997), stabilized by p42/p44 MAPKs phosphorylation (Brondello et al., 1999) and located exclusively in the nucleus (Brondello et al., 1995). However, the in vivo implication of these MKPs has never been clearly demonstrated in the long-term regulation of p42/p44 MAPKs.

With the emergence of antibodies that recognize exclusively the dually phosphorylated TEY sequence (Yung et al., 1997) of the activating loop of p42/p44 MAPKs (Anderson et al., 1990; Payne et al., 1991), it was of great interest to focus on the active p42/p44 MAPKs pool and to follow its spatial and temporal regulation during growth factor stimulation. Here, we report that both non-mitogenic and mitogenic agonists induce the rapid entry of phosphorylated p42/p44 MAPKs into the nucleus. However, during long-term stimulation, the active nuclear p42/p44 MAPKs pool progressively decreases and, after 3 hours of stimulation, p42/p44 MAPKs accumulate massively in the nucleus but in an inactive state. We show that the phosphatases involved in this nuclear inactivation have a tyrosine or a dual specificity, they are neo-synthesized following activation of the p42/p44 MAPKs pathway and their action is dependent on their binding to p42/p44 MAPKs via a specific docking site. Presently, the only nuclear phosphatases that fulfill all these criteria are the MKPs. Therefore we propose that they are the best candidates for this inactivation process. Although, p42/p44 MAPKs permanently shuttle between the cytoplasm and the nucleus during mitogenic stimulation, we propose that progressive sequestration of p42/p44 MAPKs in a nuclear ‘anchoring and inactivating center’, away from the cytoplasmic Raf>MEK ‘activating center’ is an efficient mechanism for signal termination.

CCL39 Chinese hamster lung fibroblasts were cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (Life Technologies Inc) and ΔRaf-1:ER cells (Lenormand et al., 1996) were cultured in DMEM without phenol red. Both culture media were supplemented with 7.5% fetal calf serum, glutamax, penicillin (50 units/ml) and streptomycin (50 μg/ml). REF52 cells were cultured on gelatin-coated dishes or glass coverslips in DMEM supplemented with 10% inactivated fetal calf serum, glutamax, 1 mM sodium pyruvate, non-essential amino acids, penicillin (50 units/ml) and streptomycin (50 μg/ml). Cells were maintained in 5% CO₂ at 37°C. For microinjection REF52 cells were incubated in the same medium but containing 20 mM Hepes pH 7.4 and maintained at 37°C. Hypoxic conditions were obtained by placing cells in a sealed ‘Bug-Box’ anaerobic workstation (Ruskinn Technologies, Leeds, UK/Jouan, Saint Herblain, France). The oxygen in this workstation was maintained at 1-2% with the residual gas mixture being 93-94% nitrogen and 5% carbon dioxide. Quiescent cells (G₀-arrested) were obtained by incubating confluent cultures in serum-free medium for 24-48 hours.

After growth on glass coverslips, G₀-arrested cells were treated as described in the Results and in figure legends. Indirect immunofluorescence analysis was performed as previously described (Lenormand et al., 1998). For anti-activated p42/p44 MAPKs staining, cells were fixed with 10% paraformaldehyde and permeabilized in –20°C methanol. For anti-p42/p44 MAPKs staining we used methanol/acetone 70/30 (v/v) at –20°C for fixation and permeabilization. The following antibodies were used: monoclonal antibody anti-activated p42/p44 MAPKs (Sigma #M8159; 1:200); polyclonal antibody anti-p42/p44 MAPKs (UBI #06-182; 1:1000); polyclonal antibody anti-BSA (a gift from J. C. Chambard; 1:300); biotin-coupled anti-rabbit antibody (Amersham; 1:500); Alexa-594 coupled anti-mouse antibody (Molecular Probes; 1:250) and FITC-coupled streptavidin (1:500).

All observations were performed with a DMR Leica microscope using a ×100 lens, except for Fig. 6, for which observations were performed with a ×63 lens. Confocal microscopy was performed using a DMR Leica confocal microscope with a ×100 lens and optical sections were taken every 2.5 μm.

Proteins were separated by SDS-PAGE and analyzed by immunoblotting using a polyclonal anti-HIF-1α antiserum (1:1000) (Richard et al., 1999), a monoclonal antibody anti-activated p42/p44 MAPKs (Sigma #M8159; 1:10,000) and a polyclonal anti-MKP1 antiserum (1:300) (Brondello et al., 1997).

Microinjection was performed using an inject+matic microinjector (Geneva, Switzerland) and an MO-188 micromanipulator (Narishige) mounted on an inverted Nikon diaphot TMD microscope. Capillaries (GC120F-10) were purchased from Harvard Apparatus (Endenbridge, UK). BSA-coupled peptides were dissolved in injection buffer (10 mM Hepes-KOH, pH 7.4) at a concentration of 12 mg/ml.

Phosphatase activity was measured as previously described (Nichols et al., 2000).

Chemicals of the highest purity available were purchased. Leptomycine B was a generous gift from B. Wolff-Winiski, Novartis Forschungsinstitut, Vienna, Austria. bpV(phen) was from Calbiochem and UO126 from Promega. Peptides were synthesized and coupled to BSA (average of seven peptides per BSA molecule) by Neosystem, Strasbourg, France.

Since both mitogenic and non-mitogenic stimuli induce the transcription of immediate early genes (Seuwen et al., 1990), we assessed the subcellular localization of active p42/p44 MAPKs during stimulation. Non-mitogenic stimulation of CCL39 cells with the thrombin receptor peptide agonist TRP (Fig. 1A) (Vouret-Craviari et al., 1993) induced the appearance of phosphorylated p42/p44 MAPKs in the cytoplasm within 2 minutes of TRP stimulation, indicating that p42/p44 MAPKs activation occurred first in the cytoplasm. After a 5 minute stimulation, active p42/p44 MAPKs accumulated markedly in the nucleus. This nuclear signal was decreased at 10 minutes and, after 15 minutes of TRP stimulation, phosphorylated p42/p44 MAPKs were absent from both the nucleus and the cytoplasm, and the signal observed was comparable with that of resting cells.

Stimulation with 10% serum (mitogenic stimulus for CCL39 cells) induced the appearance of phosphorylated p42/p44 MAPKs with an initial time course similar to that observed with the non-mitogenic stimulus (Fig. 1B). Within 5 minutes, active p42/p44 MAPKs could easily be detected in the cytoplasm and in the nucleus. The nuclear accumulation peaked at 10 minutes post-stimulation. However, contrary to that observed with the non-mitogenic stimulus, after 15 minutes of serum stimulation, active p42/p44 MAPKs were still present in the nucleus of most of the cells at levels slightly higher than or equivalent to those of the cytoplasm.

We conclude that both mitogenic and non-mitogenic agonists can trigger the rapid entry of phosphorylated p42/p44 MAPKs into the nucleus with nearly identical time courses. However, the activation of p42/p44 MAPKs is briefer with non-mitogenic agonists with an extinction of the signal both in the nucleus and in the cytoplasm within 15 minutes of stimulation.

Fig. 2A shows serum stimulation up to 9 hours. Whereas p42/p44 MAPKs remained active for up to 6 hours in the cytoplasm, a rapid decline of the nuclear active forms occurred from 10 minutes to 1 hour of stimulation. After 3 or 6 hours of serum stimulation, phosphorylated p42/p44 MAPKs were no longer detectable in the nucleus and gradually disappeared in the cytoplasm. This result was rather unexpected since p42/p44 MAPKs progressively accumulate in the nucleus during serum stimulation (Chen et al., 1992; Lenormand et al., 1998). Therefore, it was of great interest to follow the localization of active p42/p44 MAPKs in cells where the p42/p44 MAPKs activity is artificially maintained at an elevated level. In cells stably expressing the ΔRaf1:ER chimera, the activation of ΔRaf1:ER chimera with estradiol induces persistent activation of the p42/p44 MAPKs pathway (Lenormand et al., 1998; Samuels et al., 1993). Indeed, in these cells, ΔRaf1:ER chimera stimulation induces a persistent staining of active p42/p44 MAPKs in the cytoplasm from 10 minutes up to 9 hours. Nevertheless, in the nucleus, staining for the phosphorylated forms of p42/p44 MAPKs was greatly reduced at 1 hour and not detectable from 3 to 9 hours of ΔRaf1:ER chimera stimulation (Fig. 2B).

We next compared the subcellular localization of total p42/p44 MAPKs (FITC labeling, green) with that of active p42/p44 MAPKs (Alexa-594 labeling, red) (Fig. 2C). In serum-deprived cells, the p42/p44 MAPKs pool was primarily located in the cytoplasm (Fig. 2Ca), where it was inactive (Fig. 2Cc). As described previously, addition of serum for 3 hours induced the relocalization of the p42/p44 MAPKs pool into the nucleus (Chen et al., 1992; Lenormand et al., 1993) (Fig. 2Cb). However, after a 3 hour serum stimulation, the nucleus was nearly void of phosphorylated p42/p44 MAPKs (Fig. 2Cd). Optical sections were also obtained by confocal microscopy to confirm that after 3 hours of serum stimulation, p42/p44 MAPKs accumulated in the nucleus (Fig. 2Ce) but in an inactive form (Fig. 2Cf).

Similar time courses of nuclear p42/p44 MAPKs translocation and specific inactivation of the nuclear ‘pool’ were observed in all cells tested so far, including mouse vascular endothelial cell line (1G11), primary mouse embryo fibroblasts (MEF), rat embryo fibroblasts (REF 52), Hela cells and mouse vascular smooth muscle cells (data not shown).

To characterize the phosphatases responsible for the nuclear inactivation of p42/p44 MAPKs, we used bpV(phen) (1 mM), also called oxovanadate, a potent tyrosine phosphatase inhibitor (Bevan et al., 1995; Posner et al., 1994). After 3 hours of serum stimulation, nuclear p42/p44 MAPKs were inactive (Fig. 3a). Treatment of these cells with 1 mM bpV(phen) led to the reappearance of nuclear phosphorylated p42/p44 MAPKs within 5 minutes (Fig. 3b). After a 10 or 15 minute treatment (Fig. 3c,d), phosphorylated p42/p44 MAPKs accumulated markedly in the nucleus. A lower dose of bpV(phen) (100 μM) or treatment with a high concentration of the tyrosine phosphatase inhibitor sodium orthovanadate (20 mM) (Gordon, 1991) led to the nuclear accumulation of the phosphorylated forms of p42/p44 MAPKs with a slower time course (data not shown).

The sole addition of bpV(phen) was shown to be insulino-mimetic in some cells by inhibiting the phosphatases associated with the insulin receptor (Posner et al., 1994). However, the effect observed in Fig. 3 could not be explained by a global reactivation of the p42/p44 MAPKs pathway since, in stimulated ΔRaf1:ER cells, the activation of p42/p44 MAPKs is maximal and persistent, and nuclear p42/p44 MAPKs are inactive (Fig. 2B). Moreover, activation of ΔRaf1:ER chimera for 15 minutes following a 3 hour stimulation with serum was unable to reactivate p42/p44 MAPKs in the nucleus of ΔRaf1:ER cells (Fig. 3e).

bpV(phen) is an inhibitor of tyrosine-specific phosphatases (Posner et al., 1994). However, to obtain a rapid effect we used high concentrations that may inhibit several types of phosphatases. To test the potential involvement of other classes of phosphatases, we treated 3 hour serum-stimulated cells for 15 minutes with three well-characterized inhibitors of serine/threonine-specific phosphatases. Under these conditions, the nuclear pool of p42/p44 MAPKs could not be reactivated by okadaic acid (1 μM), microcystin-LR (1 μM), nor cyclosporin A (8 μM) (data not shown).

Since sodium orthovanadate leads to inhibition of protein synthesis (Vinals et al., 2001), we tested whether bpV(phen) also alters protein synthesis. Indeed, in cells stimulated for 4 hours with 10% FCS, 1 mM bpV(phen) completely blocked protein synthesis, as potently as treatment with 30 μM cycloheximide (CHX) (data not shown). Nevertheless, treatment of 3 hour serum-stimulated cells for 15 minutes with 30 μM of CHX did not modify the nuclear signal with the anti-activated p42/p44 MAPKs antibody (Fig. 3f).

Taken together, these results indicate that the ability of bpV(phen) to trigger the rapid appearance of phosphorylated p42/p44 MAPKs in the nucleus is not mediated by reactivation of the p42/p44 MAPKs pathway, by inhibition of serine/threonine-specific phosphatases or protein synthesis, but most likely by a direct inhibition of tyrosine- or dual-specific phosphatases (also sensitive to orthovanadate) (Muda et al., 1997).

To further characterize the phosphatases involved in the nuclear inactivation of p42/p44 MAPKs, we investigated the contribution of long-term protein synthesis on the nuclear inactivation of p42/p44 MAPKs. Treatment of serum-deprived cells with CHX for 3 hours by itself did not lead to the activation of p42/p44 MAPKs (data not shown) (Brondello et al., 1997; Lenormand et al., 1998). However, stimulation of the cells with 10% FCS for 3 hours in the presence of 30 μM of CHX had two marked effects: (1) suppression of nuclear accumulation of the p42/p44 MAPKs pool (compare Fig. 4a and Fig. 4b); (2) suppression of nuclear p42/p44 MAPKs inactivation (compare Fig. 4d and Fig. 4e). The same effects were observed with 5 μg/ml of actinomycin D, which inhibits transcription (Fig. 4c,f), thus excluding the possibility that the effects observed with CHX are due to an activation of other signaling pathways such as the stress kinase pathways.

Therefore, the p42/p44 MAPKs nuclear anchoring and inactivation system could be the same entity or two parallel processes that both require p42/p44 MAPKs persistent activation and protein neo-synthesis.

We interpreted the progressive decrease in the phospho-p42/p44 MAPKs signal in the nucleus monitored by immunostaining as an inactivation of nuclear p42/p44 MAPKs. However, to confirm this interpretation we followed the time course of phosphorylation of a nuclear substrate of p42/p44 MAPKs: the hypoxia-inducible factor 1α (HIF-1α) (Richard et al., 1999). HIF-1α was detectable only when quiescent ΔRaf1:ER cells were incubated in hypoxic conditions (Fig. 5, upper blot, compare lanes 1 and 2) and its phosphorylation was evaluated by its mobility shift on a western blot (Richard et al., 1999). ΔRaf1:ER cells were stimulated for different times before the end of a 5 hour hypoxic period. After 15 minutes of ΔRaf1:ER chimera stimulation, the mobility shift up of HIF-1α was detectable (Fig. 5, lane 3) but was maximal between 30 and 60 minutes of stimulation (Fig. 5, lanes 4,5) with an equal ratio between the unphosphorylated and the phosphorylated forms of HIF-1α. After 3 hours of ΔRaf1:ER chimera stimulation, nearly all of the HIF-1α migrated as the lower molecular weight band (unphosphorylated form; Fig. 5, lane 7). At that time, addition of bpV(phen) for 30 minutes triggered the shift up of the whole band (phosphorylated form; Fig. 5, lane 8), which correlates with the strong ‘re-activation’ of nuclear p42/p44 MAPKs observed by immunofluorescence after such treatment (Fig. 3).

Interestingly, there was little variation in the total p42/p44 MAPKs activity from 15 to 180 minutes of ΔRaf1:ER chimera stimulation (Fig. 5, middle blot). The differences in the level of HIF-1α phosphorylation can therefore be explained by the localization of the active pool of p42/p44 MAPKs, which progressively decreases in the nucleus, as observed by immunofluorescence upon ΔRaf1:ER chimera stimulation (Fig. 2B).

Finally, we followed the appearance of the nuclear phosphatases MKP1 and MKP2 (Fig. 5, lower blot). After 120 and 180 minutes of ΔRaf1:ER chimera stimulation, the phosphatases were maximally induced, which correlates with the nuclear inactivation of p42/p44 MAPKs observed by immunofluorescence and with dephosphorylation of HIF-1α.

Therefore, the decrease in the nuclear phospho-p42/p44 MAPKs signal observed by immunofluorescence corresponds to a decline in the nuclear p42/p44 MAPKs activity, possibly due to the dephosphorylation of p42/p44 MAPKs by the dual specificity phosphatases MKP1 and MKP2.

To further characterize the phosphatases involved in the nuclear inactivation of p42/p44 MAPKs, we tested whether these phosphatases are specific and need to bind to the conserved docking site of p42/p44 MAPKs (Sharrocks et al., 2000; Tanoue et al., 2000). To test this hypothesis, we microinjected a peptide corresponding to the docking site for p42/p44 MAPKs on p90^rsk (Gavin and Nebreda, 1999; Smith et al., 1999). This peptide has been shown in vitro to specifically inhibit MKP3 catalytic activation by p42/p44 MAPKs by disrupting the interaction between p42/p44 MAPKs and MKP3, thereby preventing p42/p44 MAPKs dephosphorylation (Nichols et al., 2000). As a control, we microinjected a peptide mutated on the critical residues for binding (Nichols et al., 2000) (Fig. 6A, green). The peptides were coupled to BSA to reduce diffusion after specific microinjection into the nucleus or into the cytoplasm. In an in vitro MKP3 activity assay, the BSA-coupled peptides behaved similarly to the free peptides (Fig. 6B). We then microinjected BSA-p90^rskwt and BSA-p90^rskmut peptides either into the nucleus or into the cytoplasm of REF52 cells after 2 hours of serum stimulation, when inactive p42/p44 MAPKs were maximally accumulated in the nucleus. The cells were fixed after a 20 minute recovery period in serum-containing medium. Microinjected cells were detected with an anti-BSA antibody (FITC labeling, green) and the phosphorylation of p42/p44 MAPKs was detected with the anti-activated p42/p44 MAPKs antibody (Alexa-594 labeling, red).

When microinjected into the nucleus (Fig. 6C), the BSA-p90^rskwt peptide induced a reappearance of active nuclear p42/p44 MAPKs (Fig. 6Cc,d) compared with non-microinjected cells or cells microinjected with the BSA-p90^rskmut peptide (Fig. 6Cf). Moreover, microinjection of the same BSA-coupled peptides into the cytoplasm (Fig. 6D) did not have any detectable effects on the phosphorylation of nuclear p42/p44 MAPKs (Fig. 6Db,d).

These experiments suggest that the phosphatases responsible for the nuclear inactivation of p42/p44 MAPKs are likely to be nuclear phosphatases that specifically bind to p42/p44 MAPKs through a conserved docking site, two features of MKP1 and MKP2.

The rapid reappearance of nuclear phosphorylated p42/p44 MAPKs observed following treatment with bpV(phen) suggested that there was an efficient shuttling of p42/p44 MAPKs between the nucleus and the cytoplasm after 3 hours of serum stimulation.

To get more insight into this dynamic process, we analyzed the effects of Leptomycine B (LMB), an inhibitor of active nuclear export (Fukuda et al., 1997a), on p42/p44 MAPKs localization. In quiescent cells, p42/p44 MAPKs are sequestered in the cytoplasm (Fig. 7Aa). However, a 5 minute treatment with LMB was already sufficient to induce a weak nuclear accumulation of p42/p44 MAPKs (Fig. 7Ab). LMB does not activate the p42/p44 MAPKs pathway (data not shown) and the same accumulation was observed in the presence of the MEK inhibitor UO126 (data not shown). In cells stimulated with serum for 5 minutes we could not detect any nuclear accumulation of p42/p44 MAPKs (Fig. 7Ac), except when the stimulation was performed in the presence of LMB (Fig. 7Ad). Finally, after 3 hours of serum stimulation, p42/p44 MAPKs accumulated massively in the nucleus (Fig. 7Ba), and a 10 minute treatment with LMB reinforced this nuclear accumulation (Fig. 7Bb). Therefore, in quiescent as well as in serum-stimulated cells, there is a constant active export of p42/p44 MAPKs from the nucleus.

We previously showed that removal of growth factors leads to the exit of p42/p44 MAPKs from the nucleus (Lenormand et al., 1993). We thus tested the effect of a rapid and total block of MEK activity by UO126. In cells stimulated with serum for 3 hours, a 10 minute treatment with UO126 led to a rapid and nearly total relocalization of the nuclear p42/p44 MAPKs pool to the cytoplasm (Fig. 7Bc). This suggests that the MEK activity is required to maintain p42/p44 MAPKs accumulated in the nucleus. The rapid exit of p42/p44 MAPK from the nucleus observed in the presence of UO126 is also dependent on an active export mechanism, since concomitant treatment of the cells with LMB and UO126 markedly slowed down nuclear p42/p44 MAPKs exit (Fig. 7Bd).

Therefore we propose that there is a permanent shuttling of p42/p44 MAPKs throughout stimulation. The mechanism mediating the nuclear import of p42/p44 MAPKs are not well understood yet. However, we show that nuclear accumulation of p42/p44 MAPKs requires MEK activity and that nuclear export is dependent on an active mechanism.

The nucleus has been shown to be a critical site for p42/p44 MAPKs localization. Indeed, many of the p42/p44 MAPKs substrates are nuclear transcription factors and p42/p44 MAPKs nuclear translocation is necessary for the progression of cells into the S phase of the cell cycle (Brunet et al., 1999). Furthermore, the forced expression of the constitutively active MEK-p42 MAPK chimeric protein in the nucleus relaxes growth factor requirements for cell differentiation and cell transformation (Robinson et al., 1998). Inversely, senescent cells present a lack of nuclear localization of p42/p44 MAPKs compared with pre-senescent cells and restoring p42/p44 MAPKs localization in the nucleus permits overcoming some of the senescent phenotypes (Kim-Kaneyama et al., 2000). Here, we define a new role for the nucleus in the spatio-temporal regulation of p42/p44 MAPKs. Indeed, we show that the nucleus is also a critical site for the termination of the mitogenic signal by sequestration of p42/p44 MAPKs away from their cytoplasmic activator MEK and by inactivation by specific phosphatases. The nuclear sequestration of p42/p44 MAPKs after long-term mitogenic stimulation has already been described and depends on the neo-synthesis of unidentified nuclear anchors (Lenormand et al., 1998). In the yeast Saccharomyces cerevisiae it has been shown that the nuclear anchor for the MAPK Hog1 (p38 MAPK homologue) is the tyrosine-specific phosphatase Ptp2 (Mattison and Ota, 2000). Thus nuclear anchors and phosphatases could be the same entity.

Our study identified the phosphatases responsible for the nuclear inactivation of p42/p44 MAPKs as tyrosine- or dual-specific phosphatases inhibited by bpV(phen) treatment. Furthermore, we showed that these phosphatases are neo-synthesized in response to activation of the p42/p44 MAPKs pathway, as treatment with either CHX or actinomycine D during stimulation prevents the nuclear inactivation of p42/p44 MAPKs. Finally, using a peptide that corresponds to the specific docking site for p42/p44 MAPKs on their substrates and regulators (Tanoue et al., 2000), we reduced markedly nuclear p42/p44 MAPKs inactivation by disrupting the interaction between p42/p44 MAPKs and their nuclear phosphatases. We thus conclude that the nuclear phosphatases responsible for p42/p44 MAPKs nuclear inactivation need to specifically bind to this docking site in order to dephosphorylate and inactivate p42/p44 MAPKs. MKP1 and MKP2 are the only nuclear phosphatases described so far that have all these properties. They are two nuclear dual specificity phosphatases that show significant specificity towards p42/p44 MAPKs and that possess the consensus sequence for binding to p42/p44 MAPKs (Fig. 6A). Furthermore, they are specifically induced by the p42/p44 MAPKs pathway with a time course that correlates with the nuclear inactivation of p42/p44 MAPKs (Fig. 5) (Brondello et al., 1997). Finally, p42/p44 MAPKs phosphorylate MKP1 and MKP2, thereby slowing down their proteasome-dependent degradation (Brondello et al., 1999). Therefore, this enzymatic system constituted of p42/p44 MAPKs and MKP1/2 possesses all the attributes of an autoregulatory loop capable of setting the spatio-temporal activity of this pathway. However, we have not been able to ablate both MKP1 and MKP2 activity by classical antisense strategies (cDNA or oligonucleotides) to firmly establish that MKP1 and/or MKP2 are the phosphatases that dephosphorylate p42/p44 MAPKs in the nucleus. Moreover, the inactivation of the mkp1 gene does not modify the development of mice nor the time course of p42/p44 MAPKs activation (Dorfman et al., 1996). This suggests either that MKP1 does not play a key role or rather that other phosphatases such as MKP2 are able to compensate for the deficit in MKP1 in these mice. In lower eukaryotic organisms such as the worm Caenorhabditis elegans, which possesses less genetic redundancy, it has been shown that the MKP LIP-1 needs to be expressed in the nucleus to inhibit the RTK/RAS/MAPK signaling pathway and to fully induce a penetrant vulvaless phenotype (Berset et al., 2001). These studies strongly support the role of nuclear MKPs in the negative control of the RAS/MAPK signaling pathway.

Interestingly, this new role for the nucleus as a specific site for signal termination is likely to be conserved for other MAPK pathways. Indeed, McDonald et al., have recently characterized β-arrestin 2 as a MAPK scaffold for the activation of JNK3 (McDonald et al., 2000). Although most of this study was done in cells overexpressing the different partners of the pathway, it clearly shows that activation triggers the nuclear accumulation of a GFP-JNK3 fusion protein but that no nuclear signal is detectable with the anti-phospho-JNK antibody. The authors do not discuss this observation but it might be a regulatory mechanism comparable with that of endogenous p42/p44 MAPKs (Pouyssegur, 2000). Indeed, nuclear MAPK phosphatases specific for the JNK and p38 MAPK pathway have also been identified (Camps et al., 2000; Keyse, 2000).

The present report gives a new view of the dynamics of p42/p44 MAPKs localization throughout stimulation as schematized in Fig. 8. Indeed, we show that p42/p44 MAPKs constantly and rapidly shuttle between the nucleus and the cytoplasm even in quiescent cells deprived of any residual p42/p44 MAPK activity by UO126 pretreatment. This is revealed by treating the cells with LMB, which blocks active nuclear export and induces the nuclear accumulation of p42/p44 MAPKs. This result is in accordance with the model proposed by Adachi et al., in which p42/p44 MAPKs are actively exported from the nucleus by a mechanism that depends on the nuclear export sequence of MEK (Adachi et al., 2000).

Therefore, we propose the following model in which p42/p44 MAPKs localization depends on its interaction with nuclear and cytoplasmic anchors. In quiescent cells, inactive p42/p44 MAPKs accumulate in the cytoplasm, anchored to inactive MEK, while no nuclear anchors are present. Following stimulation, MEK is activated and has a very low affinity for p42/p44 MAPKs, there is then a rapid nuclear translocation of activated p42/p44 MAPKs. This translocation is independent of protein synthesis since it occurs in the presence of CHX (data not shown). Non-mitogenic stimulation does not induce neo-synthesis of nuclear anchors, therefore p42/p44 MAPKs are very rapidly exported from the nucleus. However, after mitogenic stimulation, nuclear anchors with a high affinity for inactive p42/p44 MAPKs are neo-synthesized and trigger a progressive nuclear accumulation of p42/p44 MAPKs, which is maximal after 3 hours of serum stimulation in CCL39 cells. At that time of stimulation, cytoplasmic MEK is still active and has a low affinity for p42/p44 MAPKs. After treatment with UO126, two effects combine to explain the rapid exit of p42/p44 MAPKs from the nucleus: (1) MEK is no longer active and can no longer activate p42/p44 MAPKs, thus preventing the neo-synthesis of nuclear anchors; (2) inactive MEK becomes a potent cytoplasmic anchoring protein (Fukuda et al., 1997c), which competes with nuclear anchors and drives the relocalization of p42/p44 MAPKs into the cytoplasm. Therefore, in our model the subcellular localization of p42/p44 MAPKs during stimulation is a dynamic process driven by the abundance and the affinity of the nuclear and cytoplasmic anchors. However, we cannot exclude the possibility that the rate of shuttling may vary during mitogenic stimulation since it has been shown that global nucleo-cytoplasmic transport increases during growth factor stimulation (Feldherr and Akin, 1993).

Interestingly, the permanent entry of phosphorylated p42/p44 MAPKs into the nucleus after long-term mitogenic stimulation might be sufficient to phosphorylate a very specific set of nuclear substrates. Indeed, nuclear MKP1 and MKP2 are phosphorylated by p42/p44 MAPKs at times when p42/p44 MAPKs are massively inactivated in the nucleus (Brondello et al., 1999). By contrast, substrates such as HIF-1α are phosphorylated in accordance with the time course of nuclear p42/p44 MAPKs activity (Fig. 5). These observations suggest that after long-term growth stimulation, p42/p44 MAPKs may discriminate between different substrates for specific phosphorylation. We are currently investigating this aspect of the spatio-temporal regulation of p42/p44 MAPKs activity.

We thank G. Bossis for his enthusiastic involvement at the very beginning of this project, D. Richard, E. Berra and D. Roux for designing the HIF-1α experiment, D. Grall for efficient assistance in cell culture, C. Brahimi-Horn and G. Pagès for carefully reading the manuscript and all the members of the laboratory for support and helpful discussion. This work was supported by research grants from CNRS, Université of Nice, Sophia Antipolis, INSERM, Association pour la Recherche contre le Cancer, Ligue Nationale contre le Cancer and the European Community (EC Contract B104-CT97-2071).