ERK2 shuttles between the nuclear and cytosolic compartments with a speed that depends on its activation state. We verified whether a similar process occurs with ERK1 by photobleaching the nucleus and imaging the recovery of fluorescence [23], [24]. Since photobleaching is irreversible, the recovery in the nucleus is due to the exchange of protein between cytoplasmatic and nuclear compartments. This process is described by an exponential function with a time constant inversely proportional to the speed of permeation through the nuclear membrane. Surprisingly, we found major differences in the recovery of the fluorescence between ERK1-GFP and ERK2-GFP, indicating that ERK1 moves across the nuclear membrane at a much slower rate (figure 2). In starved fibroblasts, ERK1 turnover was 3.7 times slower than ERK2 turnover (ERK1: τ=653 s; ERK2: τ=178 s). After measuring the shuttling in starved cells, cultures were stimulated with FGF4 and we waited for the translocation response to reach a stable plateau before performing a new set of FRAP recordings. After activation shuttling increased for both proteins, but ERK1 still lagged behind (ERK1: τ=266 s; ERK2: τ=84 s). In average, shuttling accelerated with activation of a factor 2.1 for ERK2 and of a factor 2.4 for ERK1 (figure 2C). Interestingly, figure 2C shows that notwithstanding the difference with ERK2, ERK1 shuttled much faster than a GFP-GFP dimer that had a lower molecular weight (54 kD compared to 71 kD for GFP-ERK1), indicating that facilitated diffusion was occurring for ERK1 as well as for ERK2. In some cases we have measured the fluorescence recovery in the same cell before and after stimulation (figure 2D). Each cell is plotted according to the level of nuclear accumulation (see Methods) and the time constant of recovery measured in starved condition (blue symbols) and after stimulation with FGF4 (green). As shown, the speed of shuttling increased for each monitored cell. This experiment was repeated in PC12 cells differentiated with NGF to their neuron-like phenotype: here we found an even stronger difference between the two kinases (data not shown). In these cells the time constant of ERK2 shuttling varied from 610±40 s (non stimulated) to 180±15 s (stimulated, n=17). In contrast, ERK1 shuttling varied from 780±42 to 640±46 s (n=13). Therefore, after stimulation ERK2 was found to be 3.6 times faster than ERK1 in PC12.

ERK1/2 consist of a central kinase domain flanked by short N- and C-terminal extension [26]. While activation and regulation of ERK1/2 have been extensively studied, little is known about the N-terminal region that resides outside of the kinase domain and encompasses the major sequence divergence between ERK1 and 2 (figure 3). Despite no specific functional motif has previously been mapped within this region we decided to explore the possibility that the N-terminus of ERK1 might be responsible for the altered trafficking. For this purpose we generated three fusion proteins constituted by 5′ truncated clones of ERK1, named respectively E1Δ²⁶, E1Δ³⁹ and E1Δ^7–39, fused to the C-terminal GFP. As shown in figure 4A, these proteins consist of ERK1 mutants in which the N-terminus was progressively shortened, thus eliminating the major difference present at the N-terminus of ERK1 with ERK2. Symmetrically, to examine the property conferred by the residues 1–39 of ERK1 to ERK2, we generated a chimeric cDNA in which the N-terminus of ERK2 was replaced with the N-terminus of ERK1 (Δ³⁹ tagged ERK2, in brief Δ³⁹E2; figure 5A). Cellular localisation and FGF-induced translocation to the nucleus of these mutants was similar to what observed for the wild type ERK fusion protein (data not shown).

We studied the shuttling properties of the N-terminal mutants of ERK1/2 with the same FRAP protocol used above. By imaging the fluorescence recovery we found that the progressive deletion of the N-terminus of ERK1 caused an acceleration of nuclear trafficking (figure 4B). The effects of these deletions were present both in the starved and in the stimulated conditions. The shuttling time constants of the E1Δ³⁹ and E1Δ^7–39 deletion mutants were identical to what was measured for ERK2 (figure 4B). The first seven residues common to ERK1 and ERK2 do not seem to be involved in the modulation of ERK nuclear permeation since the shuttling of E1Δ³⁹ and E1Δ^7–39 were identical. Furthermore, the deletion of the initial 7 residues of ERK2 (E2Δ⁷) did not affect the shuttling rate (data not shown). These experiments indicate that the core domain that determines the difference between ERK1 and 2 are the residues 8–39. The importance of the integrity of the core domain is demonstrated by the experiment on the mutant E1Δ²⁶ where the acceleration of the shuttling speed was only partial. Finally, the fusion of ERK2 with the N-terminus of ERK1 (Δ³⁹E2) strongly slowed down its exchange through the nuclear membrane (figure 5B). In summary, the N-terminus domain (residues 8–39) of ERK1 appears to be necessary and sufficient to slow down the nucleo-cytoplasmic shuttling of ERK. Therefore, in terms of their trafficking properties the chimeric protein E1Δ³⁹ assume the characteristics of ERK2, while Δ³⁹E2 becomes like ERK1: because of the conversion of the trafficking properties of ERK1 into those of ERK2 and viceversa, we labelled these constructs as E1>E2 and E2>E1, respectively. At this stage two questions arise: 1) are there any functional differences between ERK1 and ERK2 that could be attributed to their trafficking differences? 2) In case that such functional difference exists, how are the signalling properties of ERK1/2 affected by modifying the N-terminus? More specifically, is E1Δ³⁹ functionally equivalent to ERK2 and is Δ³⁹E2 equivalent to ERK1?

The results above suggested that ERK2 should activate more efficiently compared to ERK1. To verify this idea we studied the intensity/response relationship in cells treated with increasing concentrations of serum for 15 min (figure 7A). The gels have been probed with the phospho-specific ERK antibody and analyzed with a phospho-imager to ensure linearity. Since we were interested in the relative levels of activation of ERK1 and ERK2 we compared directly the phospho-specific signals rather then normalising on the total amount of protein. Therefore, the relative phosphorylation was computed as the ratio of the pERK2, pERK1 signals. If the two kinases were following the same activation profile we should expect that the ratio of activation would remain constant. On the contrary, we found that at low serum doses, ERK2 activation was larger than ERK1 indicating a better sensitivity to the stimulus. According to our model, this difference might largely be dependent on the action of nuclear phosphatases that affects more ERK1 because of its longer retention time in the nucleus. If this were true one could expect that inhibition of phosphatases would attenuate the difference in activation between ERK1 and 2. This idea was tested by immunoblotting cells which have been harvested in either starved conditions or after 30 min in 20% serum in presence or in absence of phosphatase inhibitors. As predicted, figure 7B shows that the inhibition of dephosphorylation decreased the difference between ERK1 and 2, indicating that in normal conditions ERK1 is more subjected to the action of phosphatases than ERK2. It is unlikely that this is simply due to a differential regulation of ERK1 compared to ERK2, since phosphatases have a low degree of selectivity, even more so when the targets are so similar. Interestingly, recently it has been determined that the interaction of ERK2 with dual specificity phosphatases depends on regions within the common docking domain [32], that are identical for ERK1 and ERK2. In the light of these consideration, we interpret our data as suggesting that the slower trafficking of ERK1 makes it more prone to be dephosphorylated in the nucleus. In consequence of that, ERK1 is less competent than ERK2 in activating the downstream nuclear targets.

For a long time it has been believed that ERK1 and 2 were interchangeable. This notion had originated by their vast sequence homology and by the identity of the upstream activator and of the identified substrates. Only recently it has become clear ERK1 contributes far less than ERK2 to the overall signalling output. Indeed, it has been proposed that ERK1 functions as a partial agonist of ERK2 for MEK binding, thus partially attenuating ERK2 phosphorylation and its stimulus-dependent activation [14]. However, that study did not identify the mechanisms and the domains responsible for the differences between ERK1 and ERK2. Here we believe to have not only identified one of such mechanisms but also determined the exact structural element of ERK1 which accounts for it. Most of the functional domains that have been mapped on ERK kinases, including a consensus sequence for MEK [34], [35], the catalytic domain [36] two consensus sequences for the nucleoporins [37], [38], are present on both ERK1 and 2. Thus, there are no clear reasons why the two proteins should act as differently as they do. The most evident sequence divergence is located at the N-terminus of ERK1 where a sequence of about 39 residues has not been associated previously to any specific function. We demonstrated that this sequence is necessary and sufficient to cause the difference in shuttling rates between ERK1 and 2. This has been demonstrated producing two mutants: E1Δ³⁹ and Δ³⁹E2. The first construct codifies for ERK1 minus its N-terminus and was found to shuttle at the same speed of ERK2: because of the conversion of its trafficking properties into those of ERK2 we labelled it as E1>E2 (figure 8, ,9).9). Symmetrically, the addition of the N-terminus of ERK1 at the beginning of ERK2 (Δ³⁹E2; labelled as E2>E1 in figure 8, ,9)9) slowed down the trafficking to levels undistinguishable from ERK1. There are possible alternative explanations for this effect. In one scenario, the domain might interact with elements of the nuclear envelope that hinder the progression of the protein in the nuclear pore. Alternatively, the folding of the N-terminus of ERK1 might disrupt the interactions with the nucleoporins that facilitate the diffusion of ERK through the nuclear pore.

The nucleus represents the end station for every signalling pathway that controls gene expression in response to changes of extracellular environment. It is becoming clear that the passage of signalling molecules through the nuclear membrane requires a shuttling process that allows a steady flux of biochemical communication between nucleus and cytoplasm [39]. It is conceivable that the biological advantage of this mechanism is that, since the signalling elements are continuously moving between the two compartments, the system has a better temporal responsiveness to the external environment. Furthermore, the speed of shuttling represents a possible locus for the fine tuning of the pathway activity. When the pathway is activated, the accumulation of ERK proceeds because of a combination of factors, including increased affinity to nuclear anchors [18], [24], [40], lowered affinity to cytosolic anchors [41] and changes in the equilibrium between import and export mechanisms [29], [42]. The translocation is crucial for the signalling process: events as diverse as ERK-mediated gene expression, proliferation, differentiation and neurite extension all require the entrance of the activated protein into the nucleus [19]–[22]. ERK trafficking between cytoplasm and the nucleus occur through a variety of mechanisms, both energy dependent and independent [38], [43]–[45], but little is known on the relative contributions of these processes. The influx is mediated mainly by a process of facilitated diffusion in which ERK permeates the nuclear pore following direct interactions with domains of specific nucleoporins [37], [38], [46]. Upon stimulation, an energy dependent process might activate, possibly explaining the increased rate of shuttling [44], [45] and the shift of the equilibrium conditions of the influx/efflux toward nuclear accumulation. The efficiency of these processes is witnessed by the fact that ERK2-GFP shuttles through the nuclear membrane only marginally slower than GFP in spite of being almost three times heavier. The energy independent processes must be thermodynamically reversible. Interestingly, such a reversibility has also been suggested for energy dependent processes [43]. At steady state, either during starvation or during sustained maximal stimulation, the influx is in equilibrium with the efflux. Particularly relevant to the experiments shown in figure 6, is the identity of the system that determines the rapid exit of ERK after inactivation of MEK activity. It has been proposed that ERK is moved out of the nucleus as piggyback with MEK which, in turn, is extruded by the CRM1 mediated export [29], [30]. This system is likely to play an important role in determining ERK localization in starved cells, when the concentration of ERK is lower in the nucleus than in the cytoplasm (see figure 1A). In these conditions the flux across the membrane is small, because of the very low amount of ERK present in the nucleus and of the relatively slow rate of trafficking. However, at the peak of the response (figure 6A), both the steady state concentration of nuclear ERK and the shuttling rate are much larger so the equilibrating outward flux must be far more conspicuous. It is unlikely that the CRM1-dependent export has sufficient carrying capacity to contribute substantially to the large fluxes observed in stimulated conditions. Since the concentration of MEK in the nucleus [25], [27] is very low and one MEK molecule is required to extrude one ERK molecule, there is simply not enough MEK to bind the far more numerous ERK molecules. Indeed, the blockage of the CRM1-mediated export by leptomycin B caused little if any reduction in the rate of ERK2-GFP export after treatment with U0126 [24]. These considerations suggest that the efflux observed after blockage of MEK activation (figure 6) are due to reversible facilitated diffusion.

NIH 3T3 cell line were cultured in Dulbecco modified medium supplemented with 10% FBS and antibiotics (100 units/ml penicillin/streptomycin). Rat pheochromocytoma PC12 cells were cultured in DMEM supplemented with 5% FBS, 10% horse serum and 1× Pen/Strep mixture in 5% CO₂. PC12 were induced to differentiate by adding NGF to a final concentration of 50 ng/ml. The cells were plated on glass disks or on glass-bottomed dishes (Willco) at 60–70% confluence, and transfected using Lipofectamin2000 (Invitrogen) according to the manufacture's instructions. After transfection, cells were left undisturbed for 24 hr before any further experimental manipulation. Starvation was obtained by keeping the cells for 24 hours in 1% FBS. One hour before the beginning of experiment, cells were placed in a saline solution of composition: 130 mM NaCl, 3.1 mM KCl, 1.0 mM K₂HPO₄, 4.0 mM NaHCO₃, 5.0 mM dextrose, 1.0 mM MgCl₂, 2.0 mM CaCl₂, 10 mM HEPES, 1.0 mM ascorbic acid, 0.5 mM myo-Inositol, 2 mM pyruvic acid, pH 7.3.

To understand how the speed of nucleo-cytoplasmic trafficking influence the equilibrium of the activation/inactivation balance of ERK we modelled the ERK system considering that the protein equilibrates among 4 different states regulated by first order kinetics. The simulator computes the temporal evolution of the system starting from arbitrary initial conditions by means of a Montecarlo methods. In brief, the system is modelled as a collection of particles that can exists in any of the states indicates in the diagram of figure 6C. At each time point of the simulation each particle can undergo the transitions admitted by its present state or remain in the present state. The probability of each transition is given by the rate (probability per second) divided by the temporal resolution of the sequence. A random number generator is utilized to decide the particle fate. The data presented have been computed on a system with 10000 particles simulated with a time resolution of 1 s and for a duration of 5000 iterations. This time was sufficient to allow a complete evolution of the system to the equilibrium condition. Convergence to a unique solution was verified as the final state was independent on the values of the initial conditions. At least 4 simulation runs were used for each set of conditions.

Samples were assayed to measure the phospho-transferase activity of our fusion proteins on Myelin Binding Protein (MBP), by using the pMBP assay (17–191, Upstate). The reactions were worked out accordingly to the manufacture procedures. In brief, the immunoprecipitated samples of active ERK preparations were mixed with Mg²⁺/ATP cocktail, the Assay Diluition Buffer I (ADBI, n. cat. 20–108, Upstate) and the MAPK substrate cocktail II (n. cat. 20–166, Upstate); then, the reaction mixtures were incubated for 20–30 min in a shaking incubator kept at 30°C. Samples were then analyzed by immunoblot probed using 1 µg/ml anti-phospho-MBP, clone p12 (n. cat. 20–113, Upstate).

Colony Formation assay cells were performed as previously described [14], [51] with some modifications. In brief, NIH 3T3 cells were plated in 100 mm dishes, 1.5×10⁵ cells per plate and were transfected the following day. After 48 hours, cells were trypsinized and plated on 100 mm plates, 1×10⁴ cells per plate in DMEM containing 10% bovine calf serum and 0.5 mg/ml G418 (Geneticin, Gibco-Invitrogen) for the selection of neomycin-resistant cells. Each transfection sample was plated into three plates. Medium was changed every 3 days and after 10 days clones were washed with PBS and fixed in 10% formaldehyde (Sigma-Aldrich) for 15 min. Plates were then washed once with water and stained for 5 min in 0.5% crystal violet (Fluka, Sigma-Aldrich) in 20% methanol and finally washed with water to remove background staining. Images were acquired with a scanner and all colonies larger then 1.5 mm in diameter were counted.