Report

Small CTD Phosphatases Function in Silencing Neuronal Gene Expression

See allHide authors and affiliations

Science  28 Jan 2005:
Vol. 307, Issue 5709, pp. 596-600
DOI: 10.1126/science.1100801

Abstract

Neuronal gene transcription is repressed in non-neuronal cells by the repressor element 1 (RE-1)–silencing transcription factor/neuron-restrictive silencer factor (REST/NRSF) complex. To understand how this silencing is achieved, we examined a family of class-C RNA polymerase II (RNAPII) carboxyl-terminal domain (CTD) phosphatases [small CTD phosphatases (SCPs) 1 to 3], whose expression is restricted to non-neuronal tissues. We show that REST/NRSF recruits SCPs to neuronal genes that contain RE-1 elements, leading to neuronal gene silencing in non-neuronal cells. Phosphatase-inactive forms of SCP interfere with REST/NRSF function and promote neuronal differentiation of P19 stem cells. Likewise, small interfering RNA directed to the single Drosophila SCP unmasks neuronal gene expression in S2 cells. Thus, SCP activity is an evolutionarily conserved transcriptional regulator that acts globally to silence neuronal genes.

The central nervous system is derived from neuroepithelial precursor cells within the ventricular zone that are poised to produce a wide array of specialized neuronal and glial cell types upon receiving the appropriate differentiation signals. The timing and specificity of neuronal gene expression are regulated by activator and repressor systems that gate the proper pattern of transcription. One of the best-characterized transcription factors for controlling neuronal gene expression at a global level is the repressor element 1 (RE-1)–silencing transcription factor/neuron-restrictive silencer factor (REST/NRSF) repressor protein, which is involved in suppressing both premature and nonspecific neuronal gene expression (13). The Zn2+-finger–containing protein REST/NRSF binds the 23–base pair RE-1 DNA element found in many neuronal genes and then nucleates the formation of a multiprotein complex that represses gene transcription by deacetylating histones and methylating both DNA and histone H3 (46). These covalent modifications of chromatin reduce transcription, leading to silencing of neuronal gene expression.

We recently identified a family of three closely related human class-C phosphatases with specificity for serine 5 in the C-terminal domain (CTD) of RNA polymerase II (RNAPII) (7). These small CTD phosphatases (SCPs) are localized to the nucleus in a complex with RNAPII and, like the well-characterized CTD phosphatase FCP1, they negatively regulate RNAPII activity. The x-ray structure of SCP1 has revealed a core fold and an active center similar to those of other phosphohydrolases that share the DXDX(T/V) amino acid signature motif with SCP1 and FCP1 (8). Although a conserved pocket in the SCPs and FCP1 adjacent to the active site is proposed to bind the CTD of RNAPII and confer specificity (8), it remains possible that other proteins involved in gene regulation are also targeted by SCPs. The phosphatase activity of SCP1 is required for gene silencing (7). Elevated expression of SCP1 was found to repress transcription, whereas the phosphatase-inactive mutant, dominant negative SCP1 (dnSCP1), enhanced RNAPII activity (7). However, the mechanisms that recruit SCPs to promoters and the biological function of SCPs have remained unknown.

Northern blot analysis revealed that SCP1 is widely expressed in human tissues (Fig. 1A). The highest levels were observed in skeletal muscle, in contrast to the brain, which contained very low levels of SCP1 (Fig. 1A). To determine whether another SCP family member was expressed in the brain, we also probed for SCP2 and SCP3 (fig. S1). The entire SCP family is largely excluded from the adult nervous system, whereas non-neuronal tissues expressed high levels of SCP1, -2, and -3 RNA.

Fig. 1.

SCP is a component of the REST/NRSF complex (A) Northern blot analysis of the expression of SCP1 in human tissues showing lack of expression in nervous tissue. (B) In situ hybridization analysis of the expression of SCP1 and Isl1 in E10.5 mouse cervical spinal cord. SCP1 is widely expressed in cells surrounding the developing spinal cord and in proliferating neuroepithelial cells within the ventricular zone (open arrow), but not in lateral differentiated cells (solid arrow). Isl1 labels differentiated motor neurons (solid arrow) and dorsal root ganglion sensory neurons (open arrow), which are areas where SCP1 is not expressed. (C) Coimmunoprecipitation of SCP1 and REST/NRSF. HEK293 cell extracts were immunoprecipitated with antibodies to SCP (upper panel) or REST/NRSF (lower panel). Immunoprecipitates were resolved with SDS–polyacrylamide gel eletrophoresis (SDS-PAGE), and associated proteins were identified by Western blot with antibodies to REST/NRSF and SCP1. (D) dnREST/NRSF lacks high affinity for SCP1. P19 cells were transfected with wild-type (wt) His-tagged REST (1082 amino acids) or His-tagged dnREST [DNA binding domain (DBD): 285 amino acids] together with Flag-tagged SCP1. REST-associated proteins were immunoprecipitated with antibody to His-epitope, resolved on SDS-PAGE, and probed with antibody to Flag to detect SCP1. Input controls show that Flag-SCP1 expression levels were comparable in both samples. (E) ChIP with antibody to SCP. ChIP assays of HEK293 cells with antibodies to SCP, REST/NRSF, HDAC1, HP1, or control IgG are shown. PCR primers specific for the RE-1 elements of GAD1, GRIN2A, and SCN2A2 genes and for the 3′ intron-exon region of GAD1 were used for real time (RT)–PCR. (Left) Lanes 1 to 6 show the RE-1 element of the GAD1 gene; lanes 7 to 11 show the 3′ region of the GAD1 gene; lanes 1 and 7 represent load controls (1%); lanes 2 and 8 are control IgG; lanes 3 and 9 used antibody to SCP; lanes 4 and 10 used antibody to REST/NRSF; lanes 5 and 11 used antibody to HDAC1; and lane 6 is antibody to HP1. (Right) Lanes 1 to 6 are the RE-1 element of the GRIN2A gene; lanes 7 to 11 show the RE-1 element of SCN2A2 gene; lanes 1 and 7 are load controls (1%); lanes 2 and 8 are control IgG; lanes 3 and 9 are antibodies to SCP; lanes 4 and 10 are antibodies to REST/NRSF; lanes 5 and 11 are antibodies to HDAC1; and lane 6 is antibody to HP1. (F) ChIP demonstrates that SCP1 is recruited to the RE-1 element within the GAD1 gene (arrowhead), whereas SCP1 is not detected at this site when the truncated dn form of REST/NRSF is present. The asterisk corresponds to a nonspecific primer band.

To define the embryonic pattern of SCP expression, in situ hybridization was carried out on mice at embryonic day 10.5 (E10.5). SCP1 was widely expressed in mesenchymal and ectodermal cells and in undifferentiated neuroepithelial cells (Fig. 1B). Although SCP1 was widely expressed in cervical tissues, we noted two prominent sites where it appeared to be absent: differentiating dorsal root ganglion neurons and postmitotic spinal neurons (Fig. 1B). This pattern of expression does not appear to correspond to proliferating versus nonproliferating cell populations, because SCP1 was expressed broadly in both proliferating mesenchymal and nonproliferating muscle cells. Rather, the pattern of SCP1 to -3 expression paralleled that of REST/NRSF, which is found widely in non-neuronal tissues but is excluded from neuronal cells (fig. S1) (1, 2).

The pattern of exclusion of SCPs from differentiated nervous tissue suggested that these phosphatases might function with REST/NRSF to silence neuronal gene expression. We first examined possible interactions between SCP and REST/NRSF by using coimmunoprecipitation of the endogenous proteins in HEK293 cells. SCP immunoprecipitates were found to contain REST/NRSF, and REST/NRSF immunoprecipitates contained SCP (Fig. 1C). However, a truncated form of REST/NRSF was found to act as a dominant negative (3) and was unable to bind SCP1 (Fig. 1D). The coexpression and interaction between full-length REST/NRSF and SCP1 suggests that they form a physical complex in non-neuronal cells.

To determine whether SCP was targeted to RE-1 elements in vivo, we used chromatin immunoprecipitation (ChIP) with SCP antibodies and polymerase chain reaction (PCR) primers specific for the REST/NRSF binding elements of the Na+ channel II (SCN2A2), glutamate receptor (GRIN2A), and glutamic acid decarboxylase (GAD1) genes (fig. S3) (9). These results were compared with ChIP by using antibody to REST/NRSF and a control immunoglobulin. As shown in Fig. 1E, both SCP and REST/NRSF are associated with RE-1 regulatory sites in the neuronal genes GAD1, GRIN2A, and SCN2A2. As expected, this interaction was not detected with control immunoglobulin antibodies or probes for non–RE-1 sequences in the 3′ end of the GAD1 gene. Likewise, neither REST/NRSF nor heterochromatin protein 1 (HP1) were detected at chromatin sites within the 3′ region of GAD1. The recruitment of SCP1 to RE-1 elements appeared to be mediated directly by REST/NRSF, because SCP1 is not detected at RE-1 elements in the presence of dnREST, which lacks the ability to recruit SCP1 (Fig. 1, D and F).

Next we examined the localization of other known components of the REST/NRSF silencing complex by using ChIP. Antibodies to histone deacetylase 1 (HDAC1), HP1, and MeCp2 detected each of these chromatin-bound proteins at RE-1 elements (Fig. 1E) (10). We did not identify HDAC1 on the SCN2A2 promoter, which agrees with observations that this gene was not activated by the HDAC inhibitor tricostatin A (9, 11). Thus, SCP is included with other known silencer components at RE-1 REST/NRSF binding sites, although cell type and promoter differences in the exact composition of these nucleoprotein complexes probably exist.

P19 mouse embryonic stem cells can be induced to undergo neuronal differentiation by treatment with retinoic acid (12) or the basic helix-loop-helix (bHLH) transcription factor neurogenin 2 (Ngn2) (13, 14). As expected for potential negative regulators of neuronal differentiation, SCP1 and REST/NRSF are both expressed in replicating P19 stem cells but expression of both proteins is extinguished as these cells become neurons (Fig. 2, C and D) (15). The transition point of SCP and REST/NRSF down-regulation is associated with the up-regulation of neuron-specific genes such as that which codes for β-tubulin (Fig. 2E). To examine the role of SCP1 in neuronal differentiation, we generated clonal P19 cell lines constitutively expressing the following: SCP1; a mutant phosphatase-inactive dnSCP1, where Asp96 is replaced by Glu and Asp98 is replaced by Asn (7); REST/NRSF; and green fluorescent protein (GFP) as a vector control. Under defined conditions with retinoic acid, >90% of viable P19 cells undergo morphological differentiation into neurons (Fig. 2, A and B). Expression of REST/NRSF did not affect the proliferation of undifferentiated P19 cells, but neuronal induction triggered the death of the transfected cells (Fig. 2, A and B). In contrast, expression of SCP1 did not block neuronal differentiation or cause cell death (Fig. 2, A and B). Blocking SCP activity with dnSCP1, however, increased neuronal differentiation ∼twofold (Fig. 2, A and B). These results suggest that SCP1 function is dependent on the presence of REST/NRSF. Because REST/NRSF is down-regulated as neurons differentiate (Fig. 2D) (15), the presence of SCP1 is insufficient to disrupt neuronal differentiation. Conversely, it appears that REST/NRSF neuronal-silencing activity in undifferentiated P19 cells is antagonized by dnSCP1, leading to enhanced neuronal differentiation.

Fig. 2.

SCP and REST/NRSF affect retinoic acid–induced neuronal differentiation of P19 cells. (A) Wild-type P19 cells and clonal lines expressing GFP (vector control), SCP1, phosphatase-inactive dnSCP1, or REST/NRSF were induced to differentiate into neuronlike cells (NLCs) by treatment with retinoic acid and growth in neuron-selective medium. (B) Tabulation of the number of NLCs per field. Comparisons were made to control P19 cells with Student's t test. (C) Western blot analysis of SCP expression in undifferentiated and retinoic acid–differentiated P19 cells. Densitometry indicates an ∼85% reduction in SCP levels in differentiated P19 cells. (D) Western blot analysis of REST/NRSF protein expression in undifferentiated and retinoic acid–induced differentiated P19 cells. The loading control, β-actin, is low in P19 cells engineered to express REST/NRSF, which reflects the very low number of differentiated cells. (E) RT-PCR quantitation of β-tubulin expression in undifferentiated and differentiated P19 cells.

P19 cells can be driven to differentiate into postmitotic neurons in the absence of retinoic acid by the neurogenic bHLH transcription factor complex Ngn2:E47. To test whether SCP also influences this neuronal differentation pathway, we coexpressed SCP1, Ngn2, and E47 in P19 cells and monitored their differentiation status by using cell morphology and gene expression. Neuronal differentiation driven by Ngn2:E47 was markedly attenuated in the presence of SCP1 (Fig. 3, B and E). Constitutive expression of REST/NRSF also antagonized Ngn2:E47–induced neuronal differentiation (Fig. 3, C and E); and, when coexpressed, both SCP1 and REST/NRSF were found to cooperate in blocking neuronal differentiation (Fig. 3, D and E).

Fig. 3.

SCP and REST/NRSF control Ngn2-induced neuronal differentiation (A to E) P19 stem cells were transfected with Ngn2–internal ribosome entry site (ires)–GFP plus E47 without or with SCP1 and REST/NRSF. Neurons were identified by morphology and by staining with the antibody to βIII-tubulin TuJ1 (inset, red). The fraction of Ngn2-transfected cells (green) that were TuJ1-positive (red) is tabulated in (E). (F to J) P19 stem cells were transfected with Ngn2 without E47 in the absence or presence of dnSCP1 and dnREST/NRSF. The fraction of Ngn2-transfected cells (green) that were TuJ1-positive (red) is tabulated in (J).

Because SCP1 and REST/NRSF inhibited Ngn2-induced neuronal differentiation, we explored whether interference with SCP1 and REST/NRSF by dn forms of these proteins would conversely enhance neuronal differentiation. By lowering the concentration of Ngn2 and omitting E47, we found that neuronal differentiation was greatly reduced (Fig. 3, F and J). Under these conditions, dnSCP1 and dnREST/NRSF alone or in combination were found to strongly enhance neuronal differentiation (Fig. 3, G to J) (3). Taken together, these results indicate that SCP1 functions in cooperation with REST/NRSF to suppress neuronal differentiation of stem cells, whether driven by retinoic acid or bHLH neurogenic transcription factors.

To determine whether SCP might be part of an evolutionarily conserved pathway for restricting neuronal gene expression, we next examined the role of SCP in Drosophila. A search of the Drosophila genome identified a single dSCP homolog with 75% identity to human SCP1. The dSCP protein expressed in bacteria exhibited phosphatase activity similar to that described for human SCPs (7, 10). Examination of the Drosophila databases revealed no P element insertions or specific deletions in the dSCP locus. Because the level of dSCP expression is relatively stable from the earliest times of analysis of Drosophila embryos, there is likely a strong maternal component of dSCP mRNA during early development (16).

The single SCP gene in combination with a robust RNA interference (RNAi) system in Drosophila prompted us to use an RNA silencing strategy [small interfering RNA (siRNA) dSCP] to knock down dSCP in S2 cells. Transfection of a 700–base pair siRNA dSCP into S2 cells reduced dSCP mRNA levels by >80% after treatment for 24 hours (Fig. 4A). After the knockdown of dSCP, glyceraldehyde phosphate dehydrogenase (GAPDH), ribosomal protein S35, non-neuronal oxygenase, and β-actin mRNA levels remained unchanged. In addition, the growth and morphology of the S2 cells appeared normal after siRNA dSCP treatment. Under these conditions, however, siRNA dSCP triggered an enhancement in the expression of neuronal genes in the S2 cells. Levels of the sodium channel II protein (NaChII), the N-methyl-d-aspartate glutamate receptor (NMDAR), the embryonic lethal abnormal visual (ELAV), and β-tubulin all increased 2 to >10 fold (Fig. 4A and fig. S2). The mammalian orthologs of the genes encoding NaChII, NMDAR, and β-tubulin are classical neuronal genes that contain RE-1 elements (1, 2, 9). Some neuronal genes remained unchanged in response to siRNA dSCP, including those that code for synapsin, stathmin, choline acetyltransferase (ChAT), and neurofilament. Myosin light-chain kinase (MLCK), a muscle-specific protein that was highly expressed in S2 cells, was further increased by siRNA dSCP. The derepression of genes in S2 cells also extended to markers of glial cells such as glial cell missing (GCM) (17). Thus, the down-regulation of SCP and REST/NRSF in both differentiated neurons and glia correlates with their broad role in suppressing a large subset of genes normally found in mature cells of the nervous system (Fig. 1).

Fig. 4.

Silencing of dSCP enhances the expression of neuronal and glial genes (A) Quantitation of transcript amounts was done with real-time quantitative (RT-qPCR). S2 cells were either untreated (–) or treated (+) with siRNA dSCP for 24 hours. Total RNA was prepared and treated with deoxyribonuclease, and primer pairs specific for the coding sequence of each gene were used for qPCR with the dye SYBR Green. Threshold cycle (Ct) values were obtained from triplicate data points, and changes in transcript levels for treated samples were compared with un-treated samples, which were assigned a value of 1. (Inset) Relative amounts of RT-PCR products for dSCP and GAPDH transcript without and with siRNA dSCP treatment. Similar results were obtained in four additional experiments. (B) Enzymatic processes involved in REST/NRSF-dependent neuronal gene silencing.

REST/NRSF is a DNA binding protein that assembles a repressor complex on RE-1 elements present in >1000 neuronal genes (9, 18). The proteins mSin3A/B and HDAC1/2 bind to the N terminus of REST/NRSF, and Co-REST binds to the C terminus (19). In addition to these protein interactions, methylation of Lys9 of histone H3 creates a high-affinity binding site for HP1, providing a mechanism for localization of this protein to RE-1 elements (20, 21). Our study identifies SCPs as functional components of the REST/NRSF silencing complex (Fig. 4B). The negative influence of SCPs on the transcription of neuronal genes may be mediated by dephosphorylation of the CTD of RNAPII; it is possible, however, that other phosphatase substrates mediate these effects (22). Moreover, other related phosphatases with the DXDX(T/V) amino acid signature found in FCP1-class proteins might contribute to gene regulation by using similar mechanisms. For example, the eyes absent (Eya) transcription cofactor is a protein phosphatase belonging to this general family that serves to convert the DNA binding homeo-domain protein sine oculis from a repressor to an activator, promoting eye formation in Drosophila and cell proliferation required for organ formation in mice (kidney and muscle) (23, 24). In addition, SCP2 is reported to dorsalize the ventral mesoderm, indicating it might also help to negatively regulate a subset of non-neuronal genes (25).

These data provide further evidence that a variety of mechanisms are used by REST/NRSF to suppress neuronal gene expression. Previous studies have focused on interactions with deacetylases and methylases that modify chromatin (Fig. 4B). Here we show that a mechanism involving SCPs also contributes to blocking inappropriate neuronal gene expression in developing cells. The mechanisms that inhibit the expression of any particular gene are likely to vary considerably, but our findings suggest that antagonism of the SCP pathway might help to promote neuronal differentiation from the appropriate cell types.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5709/596/DC1

Materials and Methods

Figs. S1 to S3

References and Notes

View Abstract

Navigate This Article