Many stimuli mediate activation and nuclear translocation of ERK (extracellular-signal-regulated kinase) by phosphorylation on the TEY (Thr-Glu-Tyr) motif. This is necessary to initiate transcriptional programmes controlling cellular responses, but the mechanisms that govern ERK nuclear targeting are unclear. Single-cell imaging approaches have done much to increase our understanding of input–output relationships in the ERK cascade, but few studies have addressed how the range of ERK phosphorylation responses observed in cell populations influences subcellular localization. Using automated microscopy to explore ERK regulation in single adherent cells, we find that nuclear localization responses increase in proportion to stimulus level, but not the level of TEY phosphorylation. This phosphorylation-unattributable nuclear localization response occurs in the presence of tyrosine phosphatase and protein synthesis inhibitors. It is also seen with a catalytically inactive ERK2–GFP (green fluorescent protein) mutant, and with a mutant incapable of binding the DEF (docking site for ERK, F/Y-X-F/Y-P) domains found in many ERK-binding partners. It is, however, reduced by MEK (mitogen-activated protein kinase/ERK kinase) inhibition and by mutations preventing TEY phosphorylation or in the ERK common docking region. We therefore show that TEY phosphorylation of ERK is necessary, but not sufficient, for the full nuclear accumulation response and that this ‘phosphorylation-unattributable’ component of stimulus-mediated ERK nuclear localization requires association with partner proteins via the common docking motif.
Spatiotemporal ERK regulation
ERK (extracellular-signal-regulated kinase) 1 and 2 (referred to as ERK or ERK1/2 herein) are the prototypical members of the MAPK (mitogen-activated protein kinase) family and represent the terminal members of a three-tiered kinase amplification cascade that relays extracellular cues to the cell interior [1–3]. A diverse array of transmembrane receptors can activate the ERK cascade, which is usually initiated through GTP-loading of Ras isoforms, or through activation of second messenger kinases, such as protein kinase C . These signals cause Raf kinase activation (A-Raf, B-Raf or C-Raf), the entire mechanism of which is not fully understood . Activated Raf kinases phosphorylate two serine residues in the activation segment of MEK1/2 (MAPK/ERK kinase 1/2) [5–7]. Activated MEK1/2, in turn, phosphorylate ERK1/2 on threonine and tyrosine residues of a characteristic TEY (Thr-Glu-Tyr) activation motif [8,9]. MEK1/2 are the only established physiological substrates of Raf, whereas ERK1/2 are the only known substrates for MEK1/2, but ERK has hundreds of substrates throughout the cell . As the cellular concentration of Raf/MEK/ERK increases at each tier in the pathway , this allows the amplification of Raf-initiated signals to intracellular targets.
A central puzzle is how activation of the Raf/MEK/ERK pathway can generate so many different responses even when acting within monoclonal cell populations. Changes in the intensity, duration and localization of ERK signalling can specify distinct biological outcomes . A classic example of this is the rat phaeochromocytoma 12 cell line, in which transient ERK activation causes cell proliferation, while sustained activation causes differentiation into neuron-like cells . In many other epithelial and fibroblast cell types, transient ERK activation causes differentiation, whereas sustained ERK activation causes cells to leave quiescence and initiate G1/S transition . Controlling the intensity of this stimulus is critical: inhibition of MEK during sustained phases of activity blocks G1/S transition [15,16], whereas forced hyperactivation of the ERK pathway causes cell-cycle arrest [17,18]. This latter mechanism is thought to underlie how oncogenic Ras and Raf mutations often mediate senescence and spontaneous tumour regression .
In resting cells, ERK is chiefly localized in the cytoplasm, bound to MEK and a number of other cytoplasmic scaffolds [10,20]. Dual phosphorylation and activation of ERK causes its release from these anchors and accumulation in the nucleus [21–23]. This is an essential event for ERK to phosphorylate key transcription factor targets and effect changes in gene expression . Similarly, the regulation of ERK distribution is critical for appropriate substrate targeting in the nucleus (e.g. c-Fos and c-Myc) and cytosol (e.g. ribosomal S6 kinases) , but the mechanisms regulating this are still unclear.
Graded and digital ERK signals
A further question is how a cell generates appropriate ERK signals when exposed to a mixture of different stimuli, which may be acting in competition or opposition. Early single-cell-based studies of ERK regulation in Xenopus oocytes provided key insights in this regard, using Western blotting approaches . These studies revealed that progesterone causes ERK phosphorylation and activation in a digital or ‘all-or-nothing’ manner during oocyte maturation, which arises from bistability in the ERK cascade . That is, the output of the ERK cascade is maximal after the stimulus reaches a defined threshold level and thus prevents weak signals from causing an undesired biological response. Clearly, this kind of insight can only be gleaned from single-cell approaches, as recording only mean values would obscure these different modes of ERK signalling (Figure 1A). More recent studies have used microscopy and high-throughput flow cytometry to study ERK phosphorylation responses in large numbers of intact cells. These methods have shown that ERK phosphorylation in yeast and mammalian cells can be digital or graded in nature [26–29]. Graded-type response refers to when the ppERK (dual-phosphorylated ERK) response in individual cells is proportional to the stimulus input concentration (Figure 1A). The simplest way to distinguish these modes of signalling is by using frequency distribution histograms, plotting the number of cells falling into defined ‘bins’ of phosphorylation level [26,28]. Graded responses characteristically show an increase in mode ppERK value with increasing stimulus, whereas digital responses show a characteristic bimodal distribution irrespective of the stimulus level [26,28] (Figure 1). The biological significance of the distinction may be cell-type-specific. For example, digital signals may be usefully employed where ERK is used to irreversibly commit to cellular processes, such as oocyte maturation or T-cell activation [25,26], whereas graded ERK activation may be used where reversible responses need to be mediated at low ligand concentrations, such as chemotaxis .
Proposed and observed relationships between ERK phosphorylation and nuclear localization
Proposed model for regulation of ERK nuclear localization by phosphorylation and D-domain proteins.
Several mechanisms can underlie these differences in ERK phosphorylation output. These include the non-processive nature of dual ERK phosphorylation and dephophorylation, positive-feedback loops (such as ERK-dependent inhibition of Raf kinase inhibitory protein) and negative-feedback loops (such as ERK-dependent inhibition of Raf) . Scaffold proteins can also play a central role in controlling these relationships. Controlled membrane or cytosolic localization of MAPK pathway components [27,31], as well as recruitment of positive or negative regulatory proteins to ERK cascade scaffolds, can generate different types of ERK phosphorylation output . These studies reveal that both the subcellular localization and binding partner repertoire of ERK pathway scaffolds are essential for correct signal interpretation.
Relationship between ERK TEY phosphorylation and nuclear accumulation
Experimental approaches that yield population average data, such as cell fractionation, have revealed that the subcellular localization of ERK pathway components plays a central role in correct biological output. Surprisingly, however, few studies have attempted to resolve the relationship between ERK phosphorylation and subcellular distribution throughout the full range of ppERK response within cell populations. This appears to be particularly important in the light of work showing that ERK-driven expression of immediate early genes can be digital, even when the ERK phosphorylation response is graded . This has prompted the hypothesis that ERK signalling could be digitized at the level of nuclear translocation, i.e. ERK may have to reach a threshold level of phosphorylation in a cell before nuclear accumulation can occur (Figure 1A). The use of recombinant wild-type GFP (green fluorescent protein)-tagged ERK in permeabilized cell systems has revealed that phosphorylation and mutation of key docking motifs have only minimal effects on ERK's ability to cross the nuclear envelope [33–35]. Further experiments in intact living cells expressing GFP-tagged ERK have used photobleaching approaches to show that rates of ERK shuttling to and from the nucleus are very high, and are not limiting factors during net changes in ERK localization [36,37]. This suggests that the largest influence on ERK nucleocytoplasmic distribution is the availability of binding partners in the cytoplasm and nucleus. In this context, there are several potential mechanisms that could drive a digital-type change in nuclear accumulation that are sensitive to a ppERK threshold. For example, ERK nuclear export machinery could be overcome at a given level of import, phosphorylation of ERK on sites additional to the TEY motif affecting nuclear accumulation could play a role [38,39], or phosphorylation of cytoplasmic anchors of ERK/ppERK [such as PEA-15 (protein enriched in astrocytes-15)] to cause ERK to dissociate and become available for nuclear import [40,41].
We examined spatiotemporal aspects of ERK signalling using immunostaining approaches in conjunction with automated fluorescence microscopy . This system enables cells to be stained and imaged in multiwell plate formats. Automated retrieval and extraction of data from images using predefined image analysis algorithms vastly increases the throughput and removes any user bias at acquisition or analysis stages of microscopy experiments [43,44]. These platforms have therefore been widely adopted by pharmaceutical companies for secondary drug screening and by laboratories attempting genome-wide screening assays, but they are also useful for focused single-cell-based studies where the throughput of conventional microscopy is limiting.
We treated cells with various stimuli before immunofluorescence staining for ERK1/2, ppERK1/2 and DAPI (4′,6-diamidino-2-phenylindole). Using automated image acquisition and analysis, we were able to acquire data from large numbers of attached single cells and establish the relationship between ERK phosphorylation levels, by measuring ppERK1/2 intensity levels per cell, and distribution, by measuring the N/C (nuclear/cytoplasmic) ratio of ERK1/2, during different phases of stimulation. Using frequency histograms to compare single-cell responses to increasing ligand dose, we found that both ppERK1/2 and ERK1/2 N/C responses were graded in nature. That is, the mode value of ppERK1/2 levels and the ERK1/2 N/C ratio increased gradually with increasing stimulus level (Figure 1B). This trend was observed in several cell types, but the increase in the ERK1/2 N/C ratio was highly stimulus-specific. Additionally, the redistribution of ERK1/2 appeared to be consistent across a wide range of ppERK1/2 intensities . For example, in HeLa cells stimulated with 1 μM phorbol ester, there was a roughly 30-fold difference in ppERK1/2 levels across the entire cell population, but the amplitude of the ERK1/2 N/C response encompassed a roughly 3-fold range .
Thus our data did not reveal a scenario where ERK1/2 nuclear accumulation is digital when the ppERK1/2 response is graded, but they did suggest uncoupling of ERK relocalization from ppERK levels. To test for this, we initially plotted ppERK1/2 and ERK1/2 N/C values for each individual cell on the same axes, but the volume of data and normal variation of responses made trends hard to visualize. We therefore sorted the cells into bins according to ppERK1/2 level [each spanning 80 AFU (arbitrary fluorescence units)] and, for each bin, plotted the mean ppERK1/2 value against the mean ERK1/2 N/C ratio in the same cells (Figure 1B). This gave a more convenient measure of the relationship between ERK phosphorylation and distribution, and in unstimulated cells, shows the expected positive relationship between increasing ppERK1/2 and increasing ERK1/2 N/C. If phosphorylation of the TEY motif were the only determinant of ERK redistribution, we would expect a simple extension of this curve when a stimulus is applied to cells, but this was not what is observed. Surprisingly, we find that stimulation with PDBu (phorbol 12,13-dibutyrate) [or EGF (epidermal growth factor) or GnRH (gonadotropin-releasing hormone)] for just 5 min causes a substantial increase in the ERK1/2 N/C ratio, even at comparably low levels of ppERK1/2 (Figure 1B). This is striking when comparing the localization of ERK in only a subpopulation of cells that have comparable levels of ppERK1/2, which demonstrates than the distribution of ERK1/2 is governed chiefly by the stimulus concentration rather that the phosphorylation level . This makes intuitive sense, given the wealth of data showing that the binding partner repertoire of ERK is highly stimulus-specific . Thus our data show that nuclear localization can be uncoupled from ERK phosphorylation levels, and the acute nature of this response precludes the neosynthesis of proteins known to cause this type of uncoupling, such as nuclear phosphatases. Indeed, we find that pre-incubation with protein synthesis or tyrosine phosphatase inhibitors does not affect the immediate ERK response at all . However, it is important to note that, because these are fixed cell assays, we cannot draw the conclusion that ERK nuclear accumulation is independent of TEY phosphorylation level. For example, cells with the same ppERK1/2 level could have substantially different histories, such as rapid phosphorylation and dephosphorylation, or a comparable change in ppERK1/2 level, but starting from a lower baseline. Using highly specific MEK inhibitors to block ERK phosphorylation revealed that phosphorylation on the TEY motif is required for any nuclear accumulation to occur, indicating that TEY phosphorylation is necessary, but not sufficient, to induce the full ERK nuclear localization response .
In order to explore a mechanistic basis for these stimulus-coded signals that were additional to TEY phosphorylation, we used a system where endogenous ERK1/2 are removed from cells using siRNA (small interfering RNA) transfection, and subsequently replaced using adenoviruses expressing GFP-tagged wild-type or mutated ERK2 [42,45,46]. Removal of the TEY phosphorylation sites replicates the effects of the MEK inhibitor, confirming the minimal requirement of TEY phosphorylation for nuclear accumulation . Mutation of the catalytic site and the D (docking)-domain responsible for association with DEF (docking site for ERK, F/Y-X-F/Y-P) motifs in ERK-binding partners affects basal, but not stimulus-induced, accumulation of ERK in the nucleus. These mutations did not affect the uncoupling of phosphorylation from nuclear localization levels seen with the wild-type kinase, but lend support to the view that basal and stimulus-induced ERK nuclear traffic may be regulated by distinct mechanisms . In contrast, mutation of the common D-domain, which abrogates binding to D-domain-containing binding partners, reduced stimulus-induced nuclear accumulation of ERK, specifically by inhibiting the stimulus-induced uncoupling of ERK phosphorylation from nuclear accumulation seen with the wild-type kinase .
In summary, we have not observed a scenario where ERK signalling is digitized at the level of nuclear translocation. Instead, we reveal a different model where ERK phosphorylation on the TEY motif is necessary, but not sufficient, for the full nuclear accumulation response. This additional stimulus-induced component of the ERK translocation response does not rely on protein neosynthesis, catalytic activity or DEF-domain association, but it is dependent on binding to stimulus-regulated D-domain-containing proteins. These studies illustrate how distinct ERK motifs control stimulus-regulated subcellular targeting in a way that can only be revealed by sensitive single-cell imaging of cells. Our next challenge is to find the binding partners responsible and explore the therapeutic potential for manipulating ERK targeting.
Signalling 2011: a Biochemical Society Centenary Celebration: A Biochemical Society Focused Meeting held at the University of Edinburgh, U.K., 8–10 June 2011. Organized and Edited by Nicholas Brindle (Leicester, U.K.), Simon Cook (The Babraham Institute, U.K.), Jeff McIlhinney (Oxford, U.K.), Simon Morley (University of Sussex, U.K.), Sandip Patel (University College London, U.K.), Susan Pyne (University of Strathclyde, U.K.), Colin Taylor (Cambridge, U.K.), Alan Wallace (AstraZeneca, U.K.) and Stephen Yarwood (Glasgow, U.K.).
arbitrary fluorescence units
docking site for ERK, F/Y-X-F/Y-P
green fluorescent protein
mitogen-activated protein kinase
This work was funded by the Wellcome Trust [grant numbers 084588 and 078407].