C-terminal domain small phosphatase 1 and MAP kinase reciprocally control REST stability and neuronal differentiation

Significance

A fundamental process involved in nervous-system formation is the conversion of stem cells into mature neurons. A key transcription factor in this regard is repressor element 1 (RE1) silencing transcription factor (REST), which suppresses the neuronal phenotype in stem cells and must be eliminated to promote the expression of neuronal genes in postmitotic neurons. We find that a phosphatase, C-terminal domain small phosphatase 1, coexpressed with REST in stem cells, dephosphorylates a newly identified site on REST and promotes REST stability. Conversely, we find that epidermal growth factor, an extracellular signaling molecule that promotes neurogenesis, induces phosphorylation by extracellular signal-regulated (ERK/MAP) kinases at the same site on REST. The phosphorylation facilitates elimination of REST during the transition to neurons. Our mechanism helps explain the timing of REST degradation during neuronal differentiation.

Abstract

The repressor element 1 (RE1) silencing transcription factor (REST) in stem cells represses hundreds of genes essential to neuronal function. During neurogenesis, REST is degraded in neural progenitors to promote subsequent elaboration of a mature neuronal phenotype. Prior studies indicate that part of the degradation mechanism involves phosphorylation of two sites in the C terminus of REST that require activity of beta-transducin repeat containing E3 ubiquitin protein ligase, βTrCP. We identify a proline-directed phosphorylation motif, at serines 861/864 upstream of these sites, which is a substrate for the peptidylprolyl cis/trans isomerase, Pin1, as well as the ERK1/2 kinases. Mutation at S861/864 stabilizes REST, as does inhibition of Pin1 activity. Interestingly, we find that C-terminal domain small phosphatase 1 (CTDSP1), which is recruited by REST to neuronal genes, is present in REST immunocomplexes, dephosphorylates S861/864, and stabilizes REST. Expression of a REST peptide containing S861/864 in neural progenitors inhibits terminal neuronal differentiation. Together with previous work indicating that both REST and CTDSP1 are expressed to high levels in stem cells and down-regulated during neurogenesis, our results suggest that CTDSP1 activity stabilizes REST in stem cells and that ERK-dependent phosphorylation combined with Pin1 activity promotes REST degradation in neural progenitors.

The repressor element 1 (RE1) silencing transcription factor (REST) is a transcriptional repressor that suppresses neuronal gene expression in nonneural cells, such as fibroblasts, as well as in neural progenitors (1–3). Its targets represent genes required for the terminally differentiated neuronal cell phenotype, including genes encoding voltage and ligand-dependent ion channels, receptors, growth factors, and axonal-guidance proteins (4–7). Thus, during neurogenesis, REST is progressively down-regulated to allow elaboration of the mature neuronal phenotype (3). Nonetheless, precisely how REST itself is regulated still remains an open question. Relatively little is known about either its transcriptional or posttranscriptional regulation (3, 8, 9). In contrast, several studies have focused on posttranslational regulation of REST (3, 10, 11), but the identity of the signaling molecules involved has received little attention.

In neural progenitors and human embryonic kidney (HEK) cells, rapid REST turnover is mediated by targeting to a proteasomal pathway (3, 10, 11). REST degradation during neuronal differentiation in culture requires interaction with beta-transducin repeat containing E3 ubiquitin protein ligase (βTrCP) for targeting to the proteasome (11). βTrCP was also required for cell-cycle–dependent degradation of REST in HEK cells (10). Two adjacent phosphorylated peptides in the C-terminal domain of REST were identified as βTrCP substrates in these studies, and function as degrons. One kinase responsible for the phosphorylation and degron activity in hippocampus is casein kinase 1, CK1 (12), but whether it functions in other cellular and developmental contexts is not known.

We identify a proline-directed phosphorylation site, ⁸⁶¹SPP⁸⁶⁴SP, in a domain of REST that regulates REST stability. This site lies N terminal to the degrons identified previously. We test whether phosphorylation of serines 861/864 regulates βTrCP binding to REST. Using a reporter peptide containing S861/864, as well as full-length REST protein, we monitor phosphorylation in response to perturbations of the ERK pathway and demonstrate binding and activity of the peptidylprolyl cis/trans isomerase Pin1, and the RNA polymerase C-terminal domain small phosphatase 1, CTDSP1. Finally, we examine the role for the REST S861/864 sites in destabilizing REST and promoting terminal neuronal differentiation of primary cultures of cortical neural progenitors.

Results

Phosphorylation Status of Serines 861/864 Affects REST Protein Stability.

Because REST is a relatively large protein (predicted size 122 kDa), we hypothesized that there may be multiple phosphorylation sites that impact its stability. Therefore, we performed a mass spectroscopic analysis on FLAG-tagged REST protein stably expressed in rat pheochromocytoma, PC12, cells treated with the proteasome inhibitor carbobenzoxy-L-leucyl-L-leucyl-leucinal (MG132). We chose PC12 cells because REST is turned over rapidly in this cell line unless stabilized with MG132 (3). This analysis revealed 14 phosphorylated residues, which included the previously published degron at serine 1013 (10, 11). We focused on phosphorylation sites in the C terminus because previous analysis indicated that control of REST stability was restricted to this half of the molecule. From our analysis, two prominent phosphorylated amino acids, S861 and S864, were identified (Fig. 1A). The significance of the S861/864 site was examined further in HEK293T cells because they exhibit REST turnover that is stabilized by MG132, and for comparison with the previous studies of REST degrons. Mutation of these serine residues to alanines, in the context of full-length REST, prolonged the half-life of REST protein in HEK cells treated with cyclohexamide (CHX), similar to the effect of mutations within the two C-terminal degrons (S1024/1027/1030) (10, 11) (Fig. 1B). To ensure that the S861/864A mutations did not perturb in vivo activity, we measured localization of the epitope-tagged REST to its RE1 consensus-binding site located within the first intron of NPAS4 (13). These results demonstrated that the S861/864A mutation did not impair nuclear import and localization to genomic targets relative to the wild-type (WT) protein (Fig. 1C). Therefore, without disruption of its normal activity, preventing phosphorylation at S861/864 increased the protein stability of REST.

Fig. 1.

Mutating REST serines 861/864 modulate REST protein stability without affecting chromatin binding. (A) Schematic of the predicted primary structure of human REST protein showing the newly identified phosphorylated site (861/864) relative to the previously identified degrons at residues 1009/1013 and 1024/1027/1030. (B) HA-tagged REST phosphorylation mutants are more stable than controls after transfection into HEK293T cells. CHX, 25 mM cyclohexamide. Differences between WT REST and mutant S861/864A REST are significant (SEM; P < 0.05) at every time point. Differences between WT and mutant S1024/1027/1030A REST are significant (SEM; P < 0.05) at 1.5 and 3 h. There is no significant difference between the mutants at any time point (Kruskal–Wallis test). The degradation data were fit with monoexponential functions y = A_max^(−t/τ) + C, with A_max set to 100%, t = time, τ = time constant of the exponential. (C) Histogram showing no change in occupancy of WT and S861/864A mutant FLAG–REST at the RE1 site of the transcription factor NPAS4 gene in HEK293T cells. The four and one-half lim domains 5 gene (FHL5) lacks an RE1 site. The graph represents the average of three technical replicates of a single qPCR experiment.

Phosphorylation of Serines 861/864 Regulates βTrCP Binding to REST.

Previous studies indicated that βTrCP is involved in REST degradation and bound specifically to residues S1024/27/30 in REST (11). To further investigate a role for S861/864 in βTrCP binding, we cotransfected HEK cells with cDNAs encoding a GST–βTrCP fusion and HA-tagged versions of either WT or mutant S861/864A REST. After 48 h, cell extracts were precipitated with glutathione beads and resolved on SDS gels, followed by Western blotting for HA and GST epitopes. WT REST was present in the GST–βTrCP precipitates, as predicted, but REST with the S861/864A mutation was barely detectable, suggesting that preventing phosphorylation at S861 and S864 either disrupted βTrCP binding to this site or to the degrons due to a change in REST conformation (Fig. 2 A and B). To distinguish between these two possibilities, we did the reciprocal experiment to prevent phosphorylation at the previously identified degron. When cDNAs encoding the GST–βTrCP fusion and REST E1009A-S1013/24/27/30A were cotransfected, REST protein did not copurify with GST–βTrCP. This result indicates that βTrCP does not bind to S861/864 directly in the absence of phosphorylation at the more C-terminal degron (Fig. 2C). Taken together, these results suggest that S861/864 does not interact directly with βTrCP, but that phosphorylation at this site is required for optimal binding of βTrCP to the downstream degrons.

Fig. 2.

Mutating REST serines 861/ 864 disrupt βTrCP binding to the downstream degrons. (A) Western blot analysis showing reduced levels of mutant REST compared with WT REST in GST–βTrCP complexes. HA-tagged REST cDNAs were cotransfected into HEK293T cells along with GST–βTrCP and FLAG–dominant negative-Cul1 to stabilize REST. GST–βTrCP complexes were isolated from cell extracts on glutathione agarose and probed with anti-GST and anti-HA antibodies. (B) Quantification of results in A. Band intensities of HA–WT and mutant REST proteins were determined by fluorescence intensity and normalized to GST–βTrCP levels in the glutathione pull-down. (n = 10 for each sample, ***P = 0.0002, Holm–Sidak’s multiple comparisons test; error bars represent 95% confidence intervals.) (C) Coimmunoprecipitation analysis showing that the REST degron mutant, which is WT at S861/864, does not bind to GST–βTrCP. HA–REST cDNAs were transfected into HEK293T cells along with GST–βTrCP, and FLAG–dominant negative-Cul1 to stabilize REST. Cell extracts were collected on glutathione agarose and probed with anti-HA or anti-GST antibodies. (D) Western blot showing increased CoREST in complexes with REST mutant (WT vs. S861/864A, n = 4, **P = 0.0027, ratio paired t test).

A phosphorylation-dependent change in the conformation of the C terminus of REST may also influence binding of REST to other proteins, such as its corepressors. To test this idea, we transfected either WT REST or S861/864A REST into HEK cells and looked for changes in the amount of REST corepressor 1 (CoREST1, also called Rcor1) in REST complexes. We found that preventing phosphorylation of REST increased its interaction with CoREST1, in contrast to reducing its binding to βTrCP (Fig. 2D).

REST–GFP Fusion Peptides Can Be Used to Monitor Phosphorylation Status of Serines 861/864.

To monitor phosphorylation status of specific REST sites under different circumstances, we generated a cDNA reporter that expresses a peptide encoding amino acids 810–910 of REST (Fig. 3A). We generated mutant REST (810–910) cDNAs variously encoding alanine mutations at S861, S864, and at an adjacent site, S856, which was minimally phosphorylated according to our mass spectrometric analysis. We also cloned cDNA for a control REST (amino acids 595–694) peptide that does not contain any known phosphorylated residues. All of these peptides have N-terminal FLAG and C-terminal eGFP tags (Fig. 3A). HEK cell extracts were resolved on native gels 24 h after transfection. The expressed peptides were detected by in-gel GFP fluorescence.

Fig. 3.

A REST–GFP fusion peptide (amino acids 810–910) detects phosphorylation on serines 861 and 864. (A) Schematic showing location of REST (810–910) and control (595–694) peptides. Potentially phosphorylated residues are bolded. (B) The REST peptide is phosphorylated primarily on serines 861/864 in HEK293T cells. Extracts were prepared from cells transfected with cDNAs for REST (810–910), mutated REST (810–910), or control (595–694). Peptides were resolved on native gels and analyzed by direct GFP fluorescence.

Phosphorylation state of the peptides can be visualized by differential migration in native gels due to their charge differences. The WT REST peptide (Fig. 3B, lane 1) migrated predominantly as two distinct bands that were not present upon treatment with lambda phosphatase (Fig. 3B, lane 7). Phosphatase treatment resulted in a single, slower migrating species, which we deduced to be a nonphosphorylated peptide because this band comigrated with the very low abundance species in the nontreated sample (Fig. 3B, Top band, lane 1). The assignment of the doubly phosphorylated species was based upon the mobility of a peptide containing WT S861/864 (Fig. 3B, lane 2, S856A), which disappeared in peptides containing either mutant S864A or S861A (Fig. 3B, lanes 3 and 4, respectively). Peptides containing alanine mutations at both S861 and S864 (Fig. 3B, lane 5) showed primarily the nonphosphorylated species, irrespective of whether S856A was also included (Fig. 3B, lane 6). These results suggest S861 and S864 as major sites of phosphorylation in this cellular context and establish the utility of the reporter peptide to indicate phosphorylation status of REST.

Proline Isomerase Activity at Serines 861/864 Promotes REST Degradation.

The ⁸⁶¹SPP⁸⁶⁴SP site in REST comprises a motif for peptidylprolyl cis/trans isomerase (Pin1) binding (14). We hypothesized that Pin1 interacts with REST at this site. To test this idea, we transfected HEK cells with human Pin1 (hPin1) and cDNAs encoding peptides containing S861/864 or S861/864A sites. Pin1 was present in immunocomplexes of WT REST (810–910) reporter peptides, but not phosphorylation-resistant mutant S861/864A peptides (Fig. 4A). Full-length REST, stabilized with a proteasomal inhibitor, was also copurified with Pin1 immunocomplexes after transfection of both cDNAs in HEK cells (Fig. 4B and Fig. S1), and inhibition of Pin1 activity with the inhibitor PiB at 2 μM and 5 μM increased REST protein levels in PC12 cells (Fig. 4C). Further, βTrCP binding to stabilized (MG132 treated) WT REST was disrupted by treatment with an inhibitor of Pin1 similar to the effect of preventing phosphorylation of REST by the introduction of S861/864A mutations (Fig. 4D).

Fig. 4.

Peptidylprolyl cis/trans isomerase (Pin1) activity at serines 861/864 promotes REST degradation. (A) Coimmunoprecipitation analysis showing Pin1 is present in REST (810–910) immunocomplexes in a phospho-dependent manner. HEK293T cells were transfected with hPin1 cDNA and either FLAG–WT or FLAG–mutant (S861/S864A) REST (810–910) or FLAG–control REST (595–694) cDNAs (Fig. 3). After 48 h, cells were extracted and immunoprecipitated. Western blots were probed with indicated antibodies. (B) Coimmunoprecipitation analysis showing REST in complexes with Pin1. HEK293T cells were transfected with REST and FLAG–Pin1 cDNAs for 48 h, then treated with MG132 for 4 h before extraction and immunoprecipitation. Western blots were probed with indicated antibodies. (C) Western blot showing higher REST protein levels in the presence of the Pin1 inhibitor, PiB. PC12 tet-on FLAG–REST cells were induced 24 h with doxycycline, treated with vehicle (lanes 1 and 3), or 2 μM (lane 2), or 5 μM diethyl-1,3,6,8-tetrahydro-1,3,6,8-tetraoxobenzo (lmn) (3,8) phenanthroline-2,7-diacetate PiB (lane 4) for 7.5 h, and then extracted for analysis. Blots were probed with anti-FLAG (REST) antibody and α-tubulin (loading control). (D) Coimmunoprecipitation analysis showing inhibition of Pin1 results in diminished βTrCP binding to REST, comparable to S861/864A. HEK293T cells were transfected with FLAG–REST (WT or S861/864A) and GST–βTrCP for 48 h before treatment with PiB for 7.5 h and MG132 for 4 h, and then immunoprecipitated. Western blots were probed with anti-FLAG or anti-GST antibodies.

REST Stability Is Regulated, at Least in Part, Through ERK Signaling Acting at Serines 861/864.

Pin1 binding to prolines requires phosphorylation of an adjacent serine or threonine (14). Because both serines 861 and 864 are highly predicted target sites for the proline-directed kinases ERK1 and 2, we hypothesized that ERK and its upstream activators, EGF and the small GTPase Harvey rat sarcoma viral oncogene homolog, H-Ras, would promote phosphorylation at these sites in REST.

We used the reporter peptides to test the effects of EGF treatment on REST phosphorylation in HEK cells. We found that treatment with recombinant EGF alone caused a mobility shift in the S864 peptide indicative of increased phosphorylation (Fig. 5A), but this shift was prevented by prior treatment with PD184352, a mitogen-activated protein kinase kinase (MEK) inhibitor (15). Next we analyzed whether a constitutively active Ras induced phosphorylation of either S861 or S864 in HEK cells. For this analysis, we generated three different reporter peptides, each containing a single potentially phosphorylated serine, at positions 856, 861, or 864. These were introduced into HEK cells and the peptides were monitored by in-gel GFP fluorescence. Only the peptides with individual WT S861 or 864, but not the doubly mutated S861/864A peptide, underwent a mobility shift indicative of Ras-induced phosphorylation (Fig. 5B).

Fig. 5.

EGF, RAS, and ERK signaling phosphorylates serines 861/864 on REST. (A) Native gel showing that EGF induces phosphorylation of S864 in REST that is inhibited by treatment with PD184352, a MEK inhibitor. To interrogate this serine residue, HEK293T cells were transfected with the REST(810–910)–GFP peptide containing WT S864 and alanine mutations at 856/861. Cells were treated or not with PD184352 for 20 min, followed by treatment with recombinant EGF or vehicle for an additional 10 min before analysis. (B) Native gel showing that REST S861/864 is phosphorylated by H-Ras. HEK293T cells were cotransfected with active H-Ras and indicated REST (810–910) constructs. (C) Native gel showing partial loss of S861/864 phosphorylation in WT REST peptide after treating transfected HEK293T cells with 10 μM PD184352 for 4 h. (D) Western blot showing increased FLAG–REST protein levels after treatment with PD184352 or vehicle for 30 min. (E) In vitro kinase assay showing direct phosphorylation of REST by recombinant ERK2. The indicated GST–REST peptides were expressed and purified from bacteria, then incubated with either recombinant ERK2 protein or HEK293T whole cell lysate. The blot was probed with GST antibody and REST phospho-S861/864 antibody.

The increase in stability of full-length REST protein containing alanine mutations at residues 861 and 864 (Fig. 1B) suggests that endogenous kinase activity in HEK cells phosphorylates these residues. To provide further support for this idea, we transfected REST (810–910) reporter cDNA into HEK cells in the presence and absence of the MEK/ERK inhibitor PD184352. Treatment with PD184352 caused a modest loss from the band at the position representing doubly phosphorylated peptide and a reciprocal modest increase in the species representing nonphosphorylated peptide (Fig. 5C and Fig. S2). This result predicts that ERK inhibition should also stabilize REST. Accordingly, we treated HEK cells transfected with FLAG–REST cDNA with PD184352 for 30 min and then analyzed the extracts by Western blot. The inhibitor treatment resulted in increased REST protein levels compared with treatment with vehicle (Fig. 5D and Fig. S2).

To determine whether ERK was acting directly at S861/864, we performed an in vitro kinase assay with activated recombinant ERK2 (16, 17), the isoform implicated in the regulation of neuronal differentiation (18), and bacterially expressed GST–REST (810–910) peptides containing WT S861/864 or mutant S861/864A. The untreated GST–WT REST (810–910) peptide is not recognized by a REST antibody specific for phosphorylated S861/864 (Fig. S3) and becomes detectable only after incubation with the recombinant active ERK or with whole-cell extracts containing endogenous kinase activities (Fig. 5E).

CTDSP1 Activity at Serines 861/864 Stabilizes REST Protein.

Protein stability is often regulated by a balance between kinase and phosphatase activities. CTDSP1 was first identified as a nuclear phosphatase targeting the C terminus of RNA polymerase II (19–21). Subsequently it was implicated in REST repression of neuronal gene expression in P19 embryonal carcinoma cells that have the capacity to differentiate into neurons (22). To test whether this effect on neuronal gene repression might result from CTDSP1 influence on REST protein stability, we transfected HEK cells with FLAG–CTDSP1 along with full-length REST cDNAs encoding either WT S861/864 or glutamate or alanine substitutions at these sites (Fig. 6A). We detected REST in FLAG immunocomplexes only when the site was WT or substituted with glutamate, a phosphorylation mimic. The S861/864A mutant is not in complexes with CTDSP1. Using the REST phospho-specific antibody, we found that exogenously expressed FLAG–CTDSP1 dephosphorylated these sites in full-length REST (Fig. 6B and Fig. S4). Dephosphorylation by CTDSP1 was also observed in the REST reporter peptides as detected by in-gel fluorescence (Fig. 6C). The dephosphorylation by CTDSP1 resulted in REST stabilization, which was not increased further by treatment with the proteasomal inhibitor MG132 (Fig. 6D).

Fig. 6.

CTDSP1 dephosphorylates REST at serines 861/864 and stabilizes REST. (A) Western blot analysis showing REST in immunocomplexes with CTDSP1. HEK293T cells were transfected with either WT or mutant HA–REST cDNAs together with FLAG–CTDSP1 and treated with MG132 for 4 h before FLAG immunprecipitation. Western blots were probed for REST and FLAG. (B) Western blot analysis showing loss of S861/864 phosphorylation in full-length REST after cotransfection of FLAG–CTDSP1. HEK293T cells were transfected with FLAG–REST and FLAG–CTDSP1 or empty vector and treated with MG132 for 4 h to stabilize REST. The blot was probed with FLAG antibody and REST phospho-S861/864 antibody. (C) Native gel showing partial loss of S861/864 phosphorylation on REST–GFP peptide after cotransfection with CTDSP1 cDNA into HEK293T cells. Peptides were visualized by direct GFP fluorescence. (D) Western blot analysis showing stabilized endogenous REST protein after transfection of CTDSP1 cDNA into HEK293T cells. α-Tubulin is the loading control.

Expression of the REST (810–910) Peptide Stabilizes Endogenous REST.

The above studies used the REST (810–910) reporter as a surrogate substrate for signaling molecules that target S861/864 on full-length REST. Here we use the peptide as a signaling decoy to stabilize full-length REST protein in mouse neural progenitors, where REST turnover is enhanced in preparation for terminal neuronal differentiation (3). Cultured E12.5 cortical neurospheres were Sox2-positive and remained Sox2-positive after expression of the mutant REST (810–910) peptide for 24 h (Fig. 7A), demonstrating that transfection does not alter the neural progenitor state of the culture. However, transfection with cDNA encoding WT REST (810–910) peptide (Fig. 7 B, lane 4, and C) significantly stabilized endogenous full-length REST protein compared with transfection with mutant S861/864A or control peptides (Fig. 7 B, lanes 2 and 3, respectively and C). MG132 did not stabilize the REST protein further (Fig. 7 B, lane 5, and C).

Fig. 7.

Expression of REST (810–910) peptide in neurospheres stabilizes endogenous REST protein. (A) E12.5 mouse neurospheres were transfected with REST(810–910)–IRES–GFP cDNA (green) and stained for the neural progenitor marker SOX2 (blue). [Scale bar: Right, 50 μm.] (B) Western blot showing endogenous REST protein levels in E12.5 neurospheres after transfection with empty vector, control REST (595–694) peptide, mutant S861/864A REST (810–910) peptide, WT REST (810–910) peptide, and empty vector with MG132 treatment for 4 h. The blot was probed with anti-mouse–REST and anti–α-tubulin (loading control) antibodies. (C) Quantification of B. There is a significant difference between WT REST (810–910) compared with control (595–694) peptide (n = 5, **P = 0.0021) and WT REST (810–910) compared with mutant S861/864A REST (810–910) (n = 5, *P = 0.046). The difference between MG132 and REST (810–910) is not significant (n = 5, P = 0.7996, Sidak’s multiple comparison test; error bars represent 95% confidence interval).

Expression of REST (810–910) Inhibits Neuronal Differentiation.

Previous studies indicate that persistent REST expression in progenitors inhibits neuronal differentiation. Therefore, we asked whether stabilization of REST, via expression of the REST (810–910) peptide in neural progenitors, would also hinder neuronal differentiation. Mouse neurosphere cultures were immunostained for microtubule-associated peptide 2 (MAP2), a marker of mature neurons, at days 0 and 10 posttransfection. On day 0, there was negligible MAP2 signal and no significant difference between the two conditions. At day 10, there was significantly less MAP2 expression in neurospheres transfected with REST (810–910) compared with the REST (595–694) control peptide (Fig. 8 A and B).

Fig. 8.

Expression of the REST (810–910) peptide inhibits neuronal differentiation. (A) Representative immunolabeled E12.5 neurospheres transfected with FLAG–control REST(595–694)–IRES–GFP or FLAG–REST(810–910)–IRES–GFP peptides, sorted, and then allowed to differentiate for 10 d in culture. White, DAPI-stained nuclei; red, MAP2 antibody. (Scale bar: Right, 100 μm.) (B) Histogram showing that neurospheres expressing REST (810–910) peptide have fewer MAP2-positive cells than cells transfected with control peptide. Results are from three different transfections. Error bars represent standard deviation. Fields were selected from multiple places on the coverslip; total number of cells are indicated on bars (***P < 0.0001, unpaired t test).

Discussion

Transcription factors represent a large important class of developmental regulators. Many of these factors function as gatekeepers to control the timing of expression of genes that, in turn, control the timing of progressive stages of differentiation. When the gatekeepers are repressors, they must be removed rapidly at the end of each stage to crisply promote the expression of genes required for the subsequent stage. The REST repressor is a classic example of such a gatekeeper. Because it represses a large repertoire of genes required at terminal differentiation, it is down-regulated dramatically in neural progenitors, close to the time that the progenitors exit the cell cycle (3, 10). We show that phosphorylation of serines 861 and 864 in the carboxyl terminus of REST, likely through ERK1/2 activities, is required for this rapid degradation in neurons differentiating in culture. We propose a model in which phosphorylation of S861/864 recruits, via Pin1 activity, βTrCP to two downstream degrons. This results in destabilization of REST in neural progenitors. The observation that REST still represses neuronal genes in progenitors (23) may be explained by our result showing that REST is also a substrate for the RNA polymerase II phosphatase, CTDSP1, which may also stabilize REST at neuronal promoters.

REST was shown previously to require Skp/Cullin/F box (SCF)-mediated βTrCP binding for ubiquitin-dependent proteolysis, by binding to phosphorylated serines in two adjacent peptides in the C terminus of REST (10, 11). These peptides were identified as degrons because their phosphorylation resulted in the targeting of REST to the proteasomal pathway and degradation. Our mass spectrometry results revealed the presence of another upstream REST site that contained two additional phosphorylated serines (S861/864) in the neuroendocrine cell line PC12, where REST is extremely unstable. We determined that phosphorylation at S861/864 also results in destabilization of REST, raising the question of whether all three peptides mediate degradation of REST through βTrCP binding to the phosphorylated residues. Our results show that the S861/864 peptide is likely not a direct substrate for βTrCP, because even in its phosphorylated state, if the downstream degrons are unphosphorylated, βTrCP binding is lost. Despite the indirect influence of S861/864 on βTrCP binding, competition with a small peptide containing wild-type S861/864 in neural progenitors stabilizes REST and attenuates neuronal differentiation.

Interestingly, binding of SCF βTrCP to REST leads to different outcomes inside and outside the nervous system. In the nervous system, βTrCP interactions destabilize REST to promote cessation of cell division and terminal neuronal differentiation, whereas in mammary epithelial cells, overexpression of βTrCP promotes tumorigenesis (11). If the latter involves binding to REST, something that has not been directly shown, this might suggest that REST degradation is mediated through different signaling pathways in the different cell types. The kinase responsible for the oncogenic events has not been identified. The S861/864 phosphorylation site is a consensus site for phosphorylation of the MAP kinases ERK1/2. Further, during neurogenesis, REST turnover increases, whereas concomitantly, growth factors such as EGF and FGF mediate transcriptional changes via the Ras–ERK signaling pathway (24). Our results, using growth-factor stimulation in cells as a surrogate for signaling pathways, suggest that S861/864 phosphorylation occurs, at least in part, via ERK1 and ERK2.

Consistent with this finding, we show further that ERK phosphorylates the REST reporter peptide in vitro, and inhibition of activators of ERK1/2 stabilizes REST in cells. Importantly, mice bearing a mutation in ERK2 show delayed neuronal differentiation in the CNS (18). In conditional mouse knockouts of both ERK1 and ERK2 in neural progenitors (25), some populations of neurons, particularly in cortex, were strikingly more like undifferentiated than differentiated neurons. Both mutant ERK phenotypes would be predicted by stabilization of REST through loss of ERK1/2 phosphorylation of S861/864. A similar differentiation-arrested phenotype was also observed after persistent expression of REST in neural progenitors in vivo (26). REST expression has been reported in adult brain (27, 28) as well as in neural progenitors during neurogenesis. Whether the ERK signaling pathway plays an important role in modulating REST, and therefore neuronal gene expression, in adult brain has yet to be determined. However, a recent study indicates that CK1 phosphorylates REST S1013/1024/1030 in adult brain neurons. Further, during an ischemic insult, activation of CK1 leads to βTrCP binding, ubiquitylation, and REST degradation to rescue neuronal cell death (12). It is possible that the recruitment of Pin1 to S861/864 allows accessibility to both CK1 and βTrCP at the downstream sites of REST. Results from experiments to test the relative roles of ERK and CK1 throughout development may prove informative for designing therapeutic interventions of certain brain neuropathologies.

How might transcriptional repressors, which are targeted for degradation by kinases at one developmental stage, be stabilized on chromatin at another stage to prevent expression of their target genes? This is an important question because a stabilized repressor may increase the chromatin resident time of its associated chromatin modifiers. In some cases, temporal control of stability is achieved by mechanisms that shuttle the repressors between the nucleus and cytoplasm (29–31). Although REST has been observed in some cellular contexts in the cytoplasm (28, 31, 32), its half-life in this cell compartment has not been measured, and whether its cytoplasmic complexes contain kinases or phosphatases has not been determined. In contrast, REST has been detected, together with a phosphatase, CTDSP1, on neuronal gene chromatin (33). CTDSP1 was first identified as a phosphatase for the C-terminal domain of RNA polymerase II in vitro (34). The expression of CTDSP1 decreases dramatically with differentiation of stem cells to mature neurons (33), and knockdown of CTDSP1 in a neural progenitor cell line accelerates neuronal differentiation (22). These results led to the hypothesis that REST recruits CTDSP1 to the RNA polymerase II initiation complex where it specifically dephosphorylates Pol II CTD to attenuate transcription (22, 33). However, whether REST or Pol II CTD was the target of CTDSP1 activity on neuronal gene chromatin was not determined. Our results suggest that REST is directly targeted by CTDSP1, such that REST repression of neuronal genes may be enhanced by both maintaining repressive deacetylated histones, while at the same time attenuating PolII transcriptional initiation.

The S861/864 site in REST is a proline-directed protein phosphorylation consensus sequence (pSer/Thr-Pro). In keeping with this finding, phosphorylated S861/864 in our reporter peptide is bound by the peptidylprolyl cis/trans isomerase, Pin1. The reporter peptide experiments show that Pin1 interacts with phosphorylated REST at S861/864 in vitro, and presumably in vivo as well, because treatment of PC12 cells with a Pin1 inhibitor stabilizes REST. The function of Pin1 is to alter the conformation of substrate proteins. Previous studies indicate that such structural alterations can underlie changes in protein–protein interactions. For example, in the case of the mitotic inhibitor 1, Emi1, Pin1 binding to one domain disrupts βTrCP binding at a different domain, which subsequently stabilizes Emi1 (35). Our results suggest an opposite function for Pin1 in that its binding facilitates REST degradation. The interactions of gatekeepers, such as REST, with Pin1 have consequences for neuronal differentiation. In the human neuronal cell line, SY5Y, Pin1 levels increase concomitantly with retinoic-acid–induced neuronal differentiation (36), as REST is down-regulated. In vivo, Pin1 becomes highly expressed in nestin-positive neural progenitors between embryonic days E12.5 and E15.5 (37), precisely the time at which REST is down-regulated in cortical progenitors (3). Further, premature loss of Pin1 in mice, using nestin-driven Cre recombinase, results in fewer βIII tubulin positive neurons in the developing cerebral cortex (37). The authors proposed that Pin1 is required to stabilize β-catenin and thereby terminate expression of Wnt-activated proneural genes, such as neurogenin 1. Our results, together with this study, suggest an intriguing model. It is possible that Pin1 propels neurogenesis by both stabilizing β-catenin, which inactivates early proneural genes, and destabilizing REST, thus allowing activation of genes required for terminal differentiation. For REST, the Pin1 interaction would lead to recruitment of βTrCP and rapid proteolysis. Future tests of this model will provide insight into how early and late states of neurogenesis are coordinated.

Materials and Methods

Plasmids.

HA–REST and mutant REST (E1009A/S1013A) cDNAs were provided by Michele Pagano (New York University School of Medicine, New York) (10). HA–REST and REST mutations (S1024/1027/1030A, E1009/S1013/1024/1027/1030A, S861/864A, and S861/864E) were generated by overlap extension PCR. HA was replaced with FLAG epitope for some experiments. For REST peptide reporter constructs, selected regions of REST were PCR amplified with flanking restriction sites to permit insertion along with FLAG linker into pEGFP-N1 (Clontech). WT and mutant REST regions were also cloned into plasmid CMV-IE/chicken beta actin promoter containing the internal ribosome entry site (pCAG–IRES–GFP) (38) for expression in neural progenitors or into pGEX-3X (GE Healthcare) for bacterial expression. All constructs were confirmed by DNA sequence analysis. GST–βTrCP and dnCul1 (amino acids 1–452) were provided by Xiaolu Cambronne (Oregon Health and Science University, Portland, OR) (39, 40). Human Pin1 and FLAG–Xenopus Pin1 were provided by Rosalie Sears (Oregon Health and Science University, Portland, OR) (41, 42).

Transient Transfections and Cell Culture.

HEK293T cells were transfected using Lipofectamine 2000 (Life Technologies) following manufacturer recommendations. PC12 tet, PC12 tet-on FLAG–REST, and HEK293T cell lines were maintained as described (8). Neurospheres were isolated from E12.5 mouse cortices and cultured as previously described (43) in neurobasal media supplemented with 2 mM l-glutamine, 100 units/mL penicillin, 100 units/mL streptomycin, B27 supplement, 20 ng/mL EGF, and 10 ng/mL FGF-2, all from Life Technologies. Neurospheres were passaged every 3–4 d using Accutase (Sigma). For stability analysis, neurospheres (passage 4/5) were dissociated and 5 × 10⁵ cells were transfected with 3 μg of plasmid DNA using Lipofectamine 2000 (Life Technologies). For differentiation analysis, neurospheres (passage 4/5) were dissociated, transfected with REST–GFP constructs, and 48 h later dissociated again and sorted for GFP. GFP-positive neurospheres (2 × 10⁵) were plated on a coverslip precoated with laminin (20 μg/mL; BD Bioscience) and poly-d-lysine (200 μg/mL; Sigma) and incubated in neurobasal medium containing 0.5% FBS and lacking EGF and FGF-2.

Mass Spectrometry.

Ten 150-mm plates of PC12 tet-on FLAG–REST cells were treated with 10 μM MG132 for 5 h before immunoprecipitation with FLAG antibody (M2 agarose; Sigma). Samples were reduced with DTT, alkylated with iodoacetamide, and digested overnight with the following enzymatic conditions: trypsin + GluC, GluC only, and GluC + Asp-N. For analysis on the mass spectrometer, each protein digest was analyzed by LC-MS using an Agilent 1100 series capillary LC system (Agilent Technologies) and an LTQ linear ion trap mass spectrometer (Thermo Fisher). Electrospray ionization was performed with an ion max source fitted with a 34-gauge metal needle (Thermo Fisher; 97144–20040) and 2.7 kV source voltage. Samples were applied at 20 µL/min to a trap cartridge (Michrom BioResources) and then switched onto a 0.5 × 250 mm Zorbax SB-C18 column with 5 μm particles (Agilent Technologies) using a mobile phase containing 0.1% formic acid, 7–30% acetonitrile gradient over 95 min, and 10 µL/min flow rate.

Data-dependent collection of MS/MS spectra used the dynamic exclusion feature of the instrument’s control software (repeat count equal to 1, exclusion list size of 50, exclusion duration of 30 s, and exclusion mass width of −1 to +4) to obtain MS/MS spectra of the three most abundant parent ions (minimum signal of 10,000) following each survey scan from m/z 400–2,000. The tune file was configured with no averaging of microscans, a maximum MS1 inject time of 200 msec, a maximum MS2 inject time of 100 msec, and automatic gain control targets of 3 × 10⁴ in MS1 mode and 1 × 10⁴ in MS2 mode.

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the PRIDE partner repository with the data set identifier (PXD001127).

Interaction Studies.

GST pull-down experiments and Western blots were performed as described (11).

Chromatin Immunoprecipitation.

ChIP analyses with REST were performed as described previously (3). Briefly, 10⁷ HEK293T cells transfected with constructs expressing versions of FLAG–REST were cross-linked in 1% formaldehyde for 10 min and quenched in 0.125 M glycine for 5 min at room temperature. Cell lysates were sonicated to shear chromatin fragments to a size range of ∼100–750 bp. Chromatin was purified from these lysates using either 5 μg of anti-FLAG M2 (Sigma) or nonspecific mouse IgG. Quantitative real-time PCR (qPCR) measured relative quantities of coimmunoprecipitated NPAS4 and FHL5 genomic regions. ChIP enrichment was determined by using the ΔΔCt method to calculate the fold change between signal measured from immunoprecipitations to signal from total nuclear lysate (10% of immunoprecipitation input), i.e., 10 × 2^{−[Ct(FLAG or IgG) – Ct(Input)]}. Primer sequences (5′→3′) used for qPCR were as follows: NPAS4 (F) CCTGAGCCTAGGGGAACATAG and NPAS4 (R) CATGGACAGAGCCATACACG; FHL5 (F) ACAGGTGCCAAGTTTATCTGC and FHL5 (R) TACCCACCAAGGAGACAGAG.

Coimmunoprecipitation and Western Blot Analysis.

Whole-cell lysates were prepared and immunoprecipitated following the procedures in refs. 8 and 44. Western blots were performed by standard procedures using antibodies described below and analyzed using anti-IgG conjugated to infrared dyes (Thermo Fisher) on a Odyssey infrared fluorescence imager (LiCor).

In Vitro Binding Assay.

Bacterially expressed GST–REST peptides were purified by glutathione agarose affinity chromatography (45). ERK2 in vitro kinase analysis with purified protein was performed as described (46).

Antibodies and Chemicals.

The antibodies used were: anti-REST-C and CoREST1 (3), HA-probe (F-7) and anti-Pin1 (Santa Cruz Biotechnology), anti-FLAG M2 (Sigma), anti-GST (Thermo Fisher), anti–phospho-ERK1/2 (Cell Signaling), anti-ERK1/2 (BD Transduction Laboratories), anti–alpha-tubulin (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA), and anti-MAP2 and anti-Sox2 (Millipore). The chemicals used were: cyclohexamide solution (Sigma), MG132 (Calbiochem), PiB (Sigma), and PD184352 (LC Laboratories).

Imaging.

Detection of in-gel fluorescence on native Tris⋅HCl 10.5–14% acrylamide gels (Bio-Rad) was performed using the Typhoon imaging system (GE Healthcare Life Sciences).