Differential activation of extracellular signal-regulated kinase 1 and a related complex in neuronal nuclei
- First Online:
- Cite this article as:
- Lundquist, J. & Dudek, S. Brain Cell Bio (2006) 35: 267. doi:10.1007/s11068-008-9018-7
- 63 Views
The extracellular signal-regulated kinases 1 and 2 (ERKs 1/2) are known to participate in regulating transcription in response to moderate depolarization, such as synaptic stimulation, but how the same active enzyme can differentially regulate distinct transcriptional programs induced with abnormal depolarization (high potassium) is unknown. We hypothesized that ERK1 or 2 accomplishes this differential nuclear response through close association with other proteins in stable complexes. In support of this hypothesis, we have found that immunoreactivity for an apparent high molecular weight complex containing phospho-ERK1 increased in response to synaptic stimulation, but decreased in response to high potassium; p-ERK immunoreactivity at 44/42 kDa increased in both cases. Evidence supporting the conclusion that the band of interest contained ERK1 in a complex, as opposed to it being an unrelated protein crossreacting with antibodies against p-ERK, is that ERK1 (p44 MAPK) and 14-3-3 protein were electroeluted from the 160-kDa band cut from a gel. We also found the nuclear complexes to be exceptionally durable, suggesting a role for the crosslinking enzyme, transglutaminase, in its stabilization. In addition, we found other components of the ERK pathway, including MEK, ERK2, p90RSK, and Elk-1, migrating at higher-than-expected weights in brain nuclei. These results describe a novel stable complex of ERK1 in neuronal nuclei that responds differentially to synaptic and depolarizing stimulation, and thus may be capable of mediating gene transcription in a way distinct from the monomeric protein.
The extracellular signal-regulated kinases 1 and 2 (ERKs 1/2; p44/p42 mitogen-activated protein kinases, MAPKs) in neurons respond to a wide variety of stimuli including glutamate receptor activation (Bading et al., 1991), depolarization with potassium (Baron et al., 1996), synaptic activity that induces either long-term potentiation (LTP) or long-term depression (LTD); (English et al., 1996; Dudek et al., 2001; Thiels et al., 2002), and growth factors such as BDNF (Segal et al., 1996). ERKs are thought to participate in regulating cellular transcriptional responses (Waltereit et al., 2001). It is unknown how active ERKs might differentially regulate transcriptional programs induced in response to synaptic activity (moderately depolarizing) as opposed to those induced by potassium (severely depolarizing) stimulation. In one case, ERK would be activated with normal synaptic activity in the brain, possibly during learning (Blum et al., 1999), and in the other only under pathological conditions, such as migraine aura, which is now thought to be due to a spreading-depression like phenomenon (Sanchez-del-Rio et al., 2004). Other calcium-activated enzymes and transcription factor phosphorylation events might play a role in distinguishing the two types of response, but they too are typically activated during both potassium and synaptic stimulation (Bito et al., 1996). To determine how enzymes that are activated in both cases, such as ERK, distinguish between the two stimulation conditions is thus critical to our understanding of activity-regulated transcription. Studying the enzymes in nuclear compartments could shed some light on the issue.
After synaptic stimulation, antibodies to phospho-ERK often stain neuronal nuclei (Patterson et al., 2001), yet most antibodies to the total ERK (types 1 or 2) typically fail to stain the nuclei, even after stimulation. Because phosphorylated ERK has been shown to translocate to the nucleus in response to a variety of stimuli in non-neuronal cells (Chen et al., 1992), the interpretation has been that neuronal activity similarly stimulates ERK translocation to the nucleus. However, translocation of ERK upon activation does not necessarily occur in every case; in PC12 cells both NGF and EGF activate ERK, but only NGF induces ERK to translocate to the nucleus (Traverse et al., 1992). Experiments using GFP-ERK have thus far been inconclusive in showing that neuronal ERK translocates to the nucleus following stimulation (Rosenblum et al., 2002), however. Further, these types of studies are complicated by the need for co-expression of a cytoplasmic retention protein to prevent nuclear localization under baseline conditions (MEK, (Horgan et al., 2003); MKP-3, (Karlsson et al., 2004)). It is also curious that such a robust translocation cannot be detected with conventional antibody methods, particularly because such translocation can be demonstrated easily with other types of stimulation, such as NGF (Traverse et al., 1992; Sano et al., 1995), serum (Chen et al., 1992), or with proteins such as p90rsk (Zhao et al., 1995). In other contexts we might generally propose that the ERK antigen-epitopes were simply inaccessible to the antibodies, possibly due to ERK being bound within a protein complex, thus explaining the lack of nuclear staining. This explanation has led us to hypothesize that a pool of ERK might exist in neuronal nuclei in complexes in both the control and stimulated cases. Supporting the idea that ERK can exist in stable complexes in neuronal nuclei, we have found that immunoblots of nuclear extracts from rat forebrain or hippocampal CA1 mini-slices show not only the expected bands of 42 (ERK2) and 44 kDa (ERK1), but also a novel high molecular weight species of approximately 160 kDa immunoreactive for phospho-ERK1/2 (p-ERK).
That complexes of ERKs exist in neurons is not without precedent; scaffolding is an important method of ERK pathway regulation in yeast and in mammalian cells (Edmunds et al., 2004; Kolch 2005). Here we describe the first case where the complexed enzyme can be regulated independently from its uncomplexed form according to the type of neuronal depolarization. Further, because this complex is always localized in nuclei, it is well positioned to provide a mechanism for different types of neuronal stimulation to regulate alternative complements of gene transcription.
Differential responses to synaptic and depolarizing (K+) stimulation
PERC-160 is an exceptionally stable complex
Most antibodies to ERK1 or ERK2 did not recognize PERC-160. However, many antibodies to ERKs 1 and 2 do not recognize ERK in the nucleus histologically, even after activation when nuclear p-ERK is readily stained. A single ERK1 specific antibody did weakly recognize high molecular weight bands, one of which co-migrates with PERC-160 in preparations of neuronal nuclei (Fig. 4B). When proteins electroeluted from the PERC-160 band were probed with this ERK1 antibody, several lower bands appeared, including one co-migrating with authentic ERK1 at 44 kDa, as stained with antibodies specific for ERK1 (Figs. 4B, C). An antibody against ERK (not specific for ERK1, but also recognizes ERK2) recognized an ERK1-sized band, but not the 160-kDa band present in the eluant, consistent with the idea that most ERK antibodies do not recognize ERKs in neuronal nuclei because the ERKs are complexed. The majority of the PERC-160 band remained intact even after electro-elution and concentration with SDS, as seen when the eluant was probed with p-ERK. Note that a small amount of p-ERK reactivity was present in the eluant at 44 kDa (Fig. 4C). All these lower bands (ERK1, p-ERK, 14-3-3) could have come only from the higher molecular weight entity, as only the 160-kDa region was cut from the gel for electroelution.
Transglutaminase may stabilize the PERC-160 complex
Transglutaminase has a certain degree of specificity in that random proteins are not crosslinked in the presence of transglutaminase. First, the two candidate substrate proteins need to associate with each other. Second, crosslinking requires that the presumed glutamine substrate from one protein be aligned very close to the lysine of the other protein. This pairing between two transglutaminase-crosslinked proteins has been used to develop antibodies that can be used to probe for crosslinked substrates (but see (Johnson et al., 2004)). Figure 5B shows that blots probed with the antibody showed immunoreactivity at 160 kDa as well as at a slightly lower molecular weight. Interestingly, the sequences of ERK1 and ERK2 differ in lysines (K) flanked by the amino acids valine (V) or phenylalanine (F), which are considered favorable for the transglutaminase crosslinking reaction (Fig. 5C) (Grootjans et al., 1995). ERK1 has one potential crosslinking site of near its N-terminus, which is ERK2 has one near its C-terminus. This suggests that both ERK1 and ERK2 could be substrates for transglutaminase, and that the two enzymes contain distinct sites for possible crosslinking or other post-translational modifications.
To determine whether transglutaminase can associate with phosphorylated ERK, antibodies against p-ERK were used for immunoprecipitation experiments. Figure 5D shows that transglutaminase can interact with p-ERK (or the PERC-160 complex) because it was co-precipitated with p-ERK. 14-3-3 can be similarly co-immunoprecipitated with p-ERK (not shown). Due to the tendency of p-ERK and PERC-160 to associate with beads in the absence of antibody, the reverse immunoprecipitation of transglutaminase was not meaningful.
To determine whether transglutaminase could be modifying the protein during the isolation of nuclear proteins, we performed several nuclear preparations (from rat forebrain) with cystamine present in the homogenization buffer and sucrose cushion. No apparent reduction in PERC-160 isolation was observed. Further evidence that transglutaminase already present in nuclei was not responsible for crosslinking during preparation was that lysing neurons in culture directly into sample buffer followed by immediate boiling still revealed the presence of PERC-160. Therefore, the complex was likely present in the nuclei prior to isolation and did not form as a result of disrupting the normal localization of transglutaminase.
Other members of the ERK pathway may similarly exist in nuclear complexes
When antibodies against other phosphorylated members of the ERK pathway were used, similar high molecular weight immunoreactive species were observed, also only in nuclear preparations. Antibodies to a phosphorylated substrate of ERK, Elk-1, labeled a band higher than 250 kDa, which did intensify after theta-burst stimulation (Fig. 6A). Similarly, theta-burst stimulation increased immunoreactivity for phosphorylated MEK, the upstream kinase that phosphorylates and activates ERK, at high molecular weights. In addition, another substrate of ERK, p90rsk, also migrated at a molecular weight higher than expected, at 160 kDa (Fig. 6B). Note that the high-running bands for Elk-1, MEK, and p90rsk were dominant in the nuclear fractions, while labeling at the expected molecular weights (62, 45, and 90 kDa, respectively) was comparatively weak in nuclei. When antibodies against other regions of the molecules—that do not label the phosphorylated forms—were used, labeling was most typically found at the expected molecular weights. An exception was found with one ERK2-specific antibody, which labeled a 250-kDa band in nuclear fractions (Fig. 6B). Therefore, the putative complexes often appear to contain predominantly the phosphorylated, active forms of the proteins. Whether the complexes only become visible upon phosphorylation, as opposed to the proteins associating upon phosphorylation, cannot be differentiated conclusively in our experiments.
In neurons, the MAP-kinases ERK-1 and ERK-2 are activated under a variety of stimulation conditions that include depolarization, synaptic activation, and growth factors such as BDNF acting through Trk-receptors (Segal et al., 1996). The patterns of gene expression, however, for each of the various types of stimuli are likely very different (Nedivi et al., 1993; Choudhuri et al., 2002). It is possible, and indeed likely, that stimulus-specific activation of additional kinases would result in the differential responses. Such an explanation could account for the different responses to depolarization versus growth factors or neuromodulators. However, it is less clear how synaptically induced depolarization is distinguished from depolarization due to hyperkalemic conditions, because most calcium-stimulated enzymes are activated under both conditions (Bito et al., 1996). One way to accomplish a signal cut-off feature, in cases of pathological depolarization, would be to have protein phosphatases closely associated with the active kinase. Associations between MAP-kinases, MAP-kinase-kinases, and appropriate MAP-kinase phosphatases have been well described in yeast (Martin et al., 2005) and rely on scaffolding molecules to bring the specific molecules together (Whitmarsh et al., 1998). Consistent with such an association with MAP-kinase phosphatases, our study shows that a complex-associated ERK1 can behave differently from its monomeric counterpart in that it can be dephosphorylated under conditions where the monomeric ERK is not. It remains unknown, however, which MAP-kinase phosphatase would be associated with PERC-160 and why it would be more active under conditions of potassium depolarization. Curiously, the dephosphorylation of CREB occurs only after NMDA receptor activation and not with potassium depolarization (Sala et al., 2000).
14-3-3 is a small, acidic group of proteins that serves a diverse group of functions. This protein family can either positively modulate enzyme activity, in the cases of serotonin N-acetyltransferase (Obsil et al., 2001) and protein kinase C (Van Der Hoeven et al., 2000), or negatively modulate activity, such as with CaM-kinase kinase (Davare et al., 2004). In the ERK pathways, 14-3-3 does not impact the activity of MEK (Shimizu et al., 1994), but has been reported to stimulate Raf, the enzyme upstream of MEK (Freed et al., 1994; Irie et al., 1994) and associates with other upstream kinases in the pathway (Yamamori et al., 1995). Therefore 14-3-3 is a likely candidate for association with ERK1 in a stable complex. Consistent with this is our finding that along with ERK1, 14-3-3 reactive bands can be electroeluted from PERC-160. The 14-3-3 could only have come from PERC-160 in that the high molecular weight complex was cut out of the gel, with the proteins subsequently removed by electroelution.
The apparent presence of 14-3-3 in the 70 kDa band in both cytosolic and nuclear fractions could represent an incomplete version of the complex or one that is localizing phospho-ERK to non-nuclear targets in addition to nuclear ones. Because this 14-3-3- and pERK-immunoreactive band was observed inconsistently, no conclusions on its function can be made. It is possible that variations in conditions during nuclear purification could have resulted in its leakage from the nucleus. Further work will be required to characterize this possible clue to the composition PERC-160 complex.
For the most part, PERC-160 remains stable through a variety of harsh treatments including boiling in SDS, guanidine, or urea. Therefore, it is curious that immunoreactivity for phospho-ERK at its correct molecular weight (44/42 kDa) increased upon boiling of nuclear fractions, while it decreased slightly in cytosolic fractions (Fig. 3A). This was not apparent when the non-phospho-antibodies were used (Fig. 3B). The most likely explanation is that the apparent increase in phospho-ERK in the nuclear fraction is due to a small amount phospho-ERK (but not unphosphorylated ERK) falling apart from otherwise stable high molecular weight complexes. Interestingly, this was not accompanied by a decrease in the 160-kDa band, and so one must therefore postulate that the increase in phospho-p44/42 ERK is coming from either an unlabeled source, or else from PERC-160, with PERC-160 being replenished by some higher complex falling apart. Decreases in staining for both the phospho- and non-phosphorylated forms of p44/42 ERK in cytosolic fractions may be due to degradation of the protein under such harsh conditions.
How is it that this ERK1-containing complex is otherwise so stable? A possibility is that transglutaminase acts to crosslink the components of the complex, which makes the complex extremely resistant to denaturation. Consistent with this possibility, we found that transglutaminase associated with p-ERK and/or PERC-160 in that it co-immunoprecipitated with p-ERK. Further, an antibody against transglutaminase-mediated isopeptide crosslinks clearly recognized the PERC-160 band in nuclear preparations. Moreover, one study found some isopeptide crosslinks within hippocampal neuronal nuclei (Maggio et al., 2001). It has been shown that transglutaminase can be active in nuclei in response to calcium (Lesort et al., 1998), and that transglutaminase associates with importin molecules (Peng et al., 1999), indicating a possible regulated nuclear function. To date, though, the only nuclear proteins demonstrated to be modified by transglutaminase in response to physiological stimulation (fertilization) are histones (Nunomura et al., 2003). Transglutaminase is activated under conditions of high calcium or low GTP (Lorand et al., 2003), indicating that it may be activated during times of cellular stress. Nuclear localization of transglutaminase attenuates apoptosis (Milakovic et al., 2004). Thus it is unclear whether such a transglutaminase-stabilized complex would be formed in a physiological homeostatic role or inadvertently in a pathological role. These results could provide a mechanism for sustaining ERK activation, and subsequent phosphorylation of CREB, following transient stimulation (Impey et al., 1998; Wu et al., 2001). However, this explanation is unlikely because in slices PERC-160 appears to be dephosphorylated with a time course similar to that of monomeric ERK. A similar complex containing ERK2 may instead be responsible.
What could be the purpose of having a complex stabilized so permanently? Protection from dephosphorylation or proteolysis is unlikely, because PERC-160 and p-ERK are equally sensitive to phosphatase and protease activity (data not shown). Because levels of the complex do not appear to form immediately after neuronal stimulation, transglutaminase-stabilized complexes could form over much longer time frames; this may be one way that a cell could adjust its nuclear signaling to integrate neuronal activity over hours or days. Alternatively, the large complex may allow ERK and MEK to be primed for activity in a form too large for export out of the nucleus: MEK has a nuclear export sequence (Fukuda et al., 1996), and it is thought the unphosphorylated ERK must associate with MEK to leave the nucleus (Adachi et al., 2000). A preformed complex in the nucleus could also facilitate rapid ERK signaling in response to action potentials by positioning it with its substrates like p90RSK (Adams et al., 2005; Zhao et al., 2005). As dimers of phosphorylated ERK1 have the highest levels of ERK activity (Philipova et al., 2005), the purpose of the complex may maintain peak activity. Interestingly, stable complexes containing recombinant ERK1 dimers can be generated by incubation with extracts from sea urchin embryos, but not from HeLa cells (Philipova et al., 2005). We propose that this occurs because 14-3-3 and/or transglutaminase were present in the sea urchin extract.
ERK1 and ERK2 are very similar enzymes, but there is much to be learned about the different roles of the two; further research focusing on PERC-160 may aid in finding how they respond differentially to different types of neuronal stimuli.
Biotinylated pentylamine (cadaverine) was purchased from Pierce. All other chemicals were obtained from Sigma-Aldrich. Anti-ERK antibodies were obtained from Calbiochem, anti-p-ERK antibodies from Promega, anti-tubulin antibodies from Upstate Biotechnology, anti-transglutaminase II antibodies from Lab Vision Corporation, anti-isopeptide crosslink antibodies from CovalAb. All other antibodies were purchased from Cell Signaling Technology.
Acute slice preparation and electrophysiology
Hippocampal slices (350-μm thickness) were cut using a vibrating microtome and were trimmed to include only the CA1 region. The size of the resultant “mini-slice” was such that a single stimulating electrode placed in the center could generate a population spike out to the slice edge. Slices were prepared in ACSF containing 124 mM NaCl, 4 mM KCl, 1.25 mM NaH2PO4, 26 mM NaHCO2, 1 mM CaCl2, 3 mM MgCl2, 10 mM glucose, 2 mM kynurenic acid, and bubbled with 95/5% O2/CO2. Following trimming, mini-slices were maintained in an interface chamber (Haas top design) at 35°C with a 1 ml/min perfusion of ACSF lacking kynurenic acid and with 2.5 mM CaCl2 and 1.5 mM MgCl2. Stimulation was carried out 2-4 h after cutting using concentric bipolar stimulating electrodes (FHC). Pulse duration and intensity, 130 μs at 60 μA, were chosen to produce population spikes across the majority of the mini-slice, resulting in an estimated 60–80% of the pyramidal neurons being activated based on phospho-ERK staining (Dudek et al., 2001). A theta-burst stimulation pattern (TBS; ten 100-Hz bursts of 4 pulses, at 5 Hz; delivered 3 times, 30 s apart), or 20 s at 5 Hz was used for the stimulation of the mini-slices, as both these protocols are effective at stimulating ERK (Dudek et al., 2001). Alternatively, mini-slices were bathed in 60 mM KCl for 5 min after the disappearance of synaptic responses. Mini-slices were subsequently removed from the interface chamber with a brush, frozen in tubes on dry ice and stored at −80°C until nuclear extraction was carried out.
Primary neuronal cultures and stimulation
Cortical neuron cultures were prepared from E18 rat embryos in a serum-free medium protocol as previously described (Brewer et al., 1993). Cultures were maintained for 21–30 days in vitro prior to experimentation in all cases. For stimulation, culture media was supplemented with 60 mM KCl or 50 μM bicuculline with or without 75 μM 4-aminopyridine for up to 20 min. A crude cytosolic fraction was prepared by removing the media and replacing it with lysis buffer (0.6% Igepal CA-630, 10 mM HEPES, 10 mM KCl, 1.5 mM MgCl2, 1 mM DTT along with protease and phosphatase inhibitors). After removing the resultant solution (cytosol), a crude nuclear fraction was generated by scraping the remaining material into LDS-PAGE sample buffer. As the ERK complex was only observed in nuclear fractions, cells were harvested directly into sample buffer in some cases.
Nuclear extract preparation
For the isolation of nuclei from the hippocampal mini-slices, a Nuclei PURE Prep kit from Sigma was utilized, with the following supplements added to all buffers: protease inhibitor cocktail III (Calbiochem), phosphatase inhibitor cocktail II (Calbiochem), 1 mM DTT, and 3 mM cystamine. Using the included procedure as a guide, approximately 12 mini-slices were homogenized in 200 μl of the supplemented lysis buffer containing 0.6% Igepal CA-630. The homogenate was diluted with 270 μl of supplemented 1.6 M sucrose solution and layered upon 150 μl of the same sucrose solution and spun in a centrifuge at 41.5 k × g for 2 h. The sucrose solutions were removed and centrifuged for 140 k × g for 1 h with the supernatant reserved as the cytosolic fraction. The nuclei were resuspended in 100-μl nuclei storage buffer and centrifuged for 30 min at 4 k × g. The supernatant was discarded and the nuclei were resuspended in 33 μl deionized water, snap-frozen on dry ice, and stored at −80°C until analyzed by SDS-PAGE. Large-scale preparations were performed in a similar manner, but scaled up accordingly, starting with forebrain and hippocampi from two rats.
For isolation of the PERC-160 complex prior to electroelution, 250 μl of resuspended nuclei from 2 brains were sonicated (20 one-second pulses, amplitude: 60) with a probe tip. Following nuclear disruption, the sample was brought to 20% saturation with ammonium sulfate and centrifuged for 15 min at 16 k × g. The pellet was discarded and the supernatant was adjusted to 50% saturation with ammonium sulfate prior to being spun in a centrifuge for 20 min at 6 k × g. The precipitate was subsequently diluted with 2.25 ml HEPES buffer (100 mM, pH 7.4), mixed with 100 mg of Reactive Brown 10 agarose (pre-equilibrated in the same buffer) and allowed to incubate with gentle end-over-end mixing for 2 h at 4°C. A Handee spin column (Promega) was used to separate the protein-bound agarose from the binding solution. The resin was then washed twice with HEPES buffer and incubated with end-over-end mixing with 500 μl salt solution (2.5 M NaCl, 0.2 M NH4OH) for 1 h to remove some proteins. Following this extraction, the protein-bound resin was separated from the salt solution, washed with HEPES buffer twice, and the protein of interest was eluted with LDS-PAGE sample buffer (Invitrogen). The resultant solution was diluted with 1.85 ml of acetate buffer (200 mM, pH 2.7), and dialyzed against the same buffer. Following dialysis, 100 mg pre-equilibrated PhosphoSelect Resin was added to the sample, and the slurry was allowed to incubate with gentle end-over-end mixing for 2 h at 25°C. After incubation, the resin was washed twice with acetate buffer and resuspended in LDS sample buffer as before.
All electrophoresis materials (gels, buffers, and membranes) were purchased from Invitrogen and used according to the manufacturer’s instructions. For immunoblot analysis, PVDF membranes were utilized and developed with ECL Plus reagent (Amersham Bioscience). Protein loading was corrected for by immunoanalysis of tubulin. Quantification of visualized bands was accomplished using ImageQuant 5.0 software. All experiments were repeated at a minimum of one additional time with similar results observed. When comparing samples from different conditions, a two-tailed t-test was used to determine statistical significance.
The transglutaminase assay was based on methods previously described (Johnson et al., 1997; Case et al., 2003). Briefly, a solution containing 50 mM HEPES, 0.5 M NaCl, 10 mM CaCl2, 1 mM DTT, 5 μM N,N-dimethylcasein, and 2 mM 5-(biotinamido)pentylamine, pH 7.4, was prepared and the reaction initiated with the addition of a 5% reaction volume of nuclear extract. This reaction proceeded at 25°C with aliquots removed every 30 min and frozen in liquid nitrogen. The inhibitory potential of cystamine, GTPγS, or calcium chelators (EDTA and EGTA) on this reaction was determined by adding the chemicals to achieve final concentrations of 12 mM, 500 μM, or 5 mM, respectively, prior to the addition of the nuclear extract. Samples were analyzed for biotinylation of the N,N-dimethylcasein by probing an immunoblot with a horseradish peroxidase-streptavidin conjugate that was subsequently developed using ECL Plus.
We thank J. Paige Adams and Rachel Robinson for some of the Western blots and Negin Martin, Marc Sommer, and Mariel Birnbaumer for critical reading of the manuscript. This research was supported by the Intramural Research Program of the National Institutes of Health and the National Institute of Environmental Health Sciences.