Cacnb4 directly couples electrical activity to gene expression, a process defective in juvenile epilepsy

Abstract

Calcium current through voltage-gated calcium channels (VGCC) controls gene expression. Here, we describe a novel signalling pathway in which the VGCC Cacnb4 subunit directly couples neuronal excitability to transcription. Electrical activity induces Cacnb4 association to Ppp2r5d, a regulatory subunit of PP2A phosphatase, followed by (i) nuclear translocation of Cacnb4/Ppp2r5d/PP2A, (ii) association with the tyrosine hydroxylase (TH) gene promoter through the nuclear transcription factor thyroid hormone receptor alpha (TRα), and (iii) histone binding through association of Cacnb4 with HP1γ concomitantly with Ser¹⁰ histone H3 dephosphorylation by PP2A. This signalling cascade leads to TH gene repression by Cacnb4 and is controlled by the state of interaction between the SH3 and guanylate kinase (GK) modules of Cacnb4. The human R482X CACNB4 mutation, responsible for a form of juvenile myoclonic epilepsy, prevents association with Ppp2r5 and nuclear targeting of the complex by altering Cacnb4 conformation. These findings demonstrate that an intact VGCC subunit acts as a repressor recruiting platform to control neuronal gene expression.

Introduction

Voltage-gated calcium channels (VGCC) are heteromultimeric complexes that translate electric signals into calcium influx (Catterall et al, 2005), thereby controlling synaptic vesicle exocytosis, neuronal excitability and gene expression (Deisseroth et al, 2003; Flavell and Greenberg, 2008; Greer and Greenberg, 2008). In VGCC, the pore-forming subunit is associated to auxiliary subunits, among which the cytoplasmic Cacnb4 (β₄) plays an essential function in channel expression level at the plasma membrane and biophysical properties (Arikkath and Campbell, 2003). Mutations in the genes encoding VGCC induce diverse neuronal pathologies, such as epilepsy, ataxia, autism and migraine (Bidaud et al, 2006). In humans, a mutation of CACNB4, leading to a 38 amino acid truncation of β₄ C-terminus, has been associated to juvenile myoclonic epilepsy (Escayg et al, 2000). In mice, a four-nucleotide insertion into a splice donor site of Cacnb4 results in a truncation of 60% of the protein sequence and a lethargic (lh) phenotype (Burgess et al, 1997). How these mutations may affect VGCC-mediated function still remains an open question. Calcium entering through VGCC has been implicated in gene regulation by activating calcium-binding proteins that propagate the signal to the nucleus (Rosen et al, 1994; Graef et al, 1999; Dolmetsch et al, 2001; Oliveria et al, 2007) or by diffusing to the nucleus with or without signal amplification by intracellular stores (Hardingham et al, 1997; Carrion et al, 1999; Hardingham et al, 2001). Evidence for alternative pathways have been reported. First, an atypical short β₄ splice variant, β_4c, was shown to interact with heterochromatin protein 1 gamma (HP1γ), a nuclear protein involved in gene silencing and transcription regulation (Hibino et al, 2003; Xu et al, 2011). However, β_4c lacks the domain required for association with the pore subunit (Chen et al, 2004) and therefore cannot couple neuronal activity to gene regulation. A second pathway implicates CCAT, a C-terminal fragment of Cacna1c VGCC, as a calcium-regulated transcription factor (Gomez-Ospina et al, 2006). In this case, calcium influx triggers CCAT nuclear export providing a calcium-dependent link between neuronal activity and transcription. The origin of this fragment (channel proteolysis or alternative splicing) remains unknown.

Here, we show that β₄ directly couples neuronal excitability to gene expression along a signalling pathway that is disrupted by the human R482X mutation. We found that electrical activity promotes the formation of a new nuclear complex in which β₄ plays the role of an organizing platform that brings together a transcription factor for DNA binding, a phosphatase for histone dephosphorylation and HP1γ for nucleosome association. Formation of this complex controls gene activity as witnessed by the case of tyrosine hydroxylase (TH) gene.

Results

β₄ nuclear localization

β₄ is highly expressed in the hippocampal dentate gyrus of adult wild-type (wt) C57Bl/6 mice brain but not in lh mice brain (Figure 1A). β₄ labelling is strongest in soma and coincides with NeuN nuclear labelling, suggesting that a fraction of β₄ protein may be localized in cell nuclei, in agreement with previous observations (Subramanyam et al, 2009). Western blot analyses of wt brain lysates show a unique 58 kDa band, corresponding to β₄, absent in lh (Figure 1B). Western blot analyses of cytoplasmic versus nuclear fractions of wt brain lysates show the presence of β₄ in the nucleus (Figure 1C). Electron microscopy (EM) demonstrates the presence of β₄ in the nucleus of CA1 hippocampal neurons (Figure 1D; Supplementary Figure 1). In primary cultures of hippocampal neurons, β₄ is predominantly expressed within the cytoplasm/plasma membrane at 5 DIV (Figure 1E, left panels). In contrast, the protein is densely located in the nucleus at 18 DIV (Figure 1E, right panels). The nuclear to cytoplasmic density ratios (NCR values) of β₄ increase as a function of time from 0.07 (3 DIV) to 3.81 (18 DIV) on average indicating the differentiation-dependent progressive targeting of β₄ to neuronal nuclei (Figure 1F). NCR value increase is accompanied by a histogram broadening witnessing higher β₄ subcellular distribution variability among neurons. β₄ nuclear appearance coincides with that of VGCC currents and electrical activity at 5 DIV (unpublished observation). EM analysis of hippocampal neurons at 5 and 18 DIV also illustrate the increase of β₄ nuclear density at 18 DIV (Figure 1G). Synchronized neuronal differentiation of NG108-15 cells, induced by serum deprivation and 1 mM cAMP, is accompanied by VGCC expression (Chemin et al, 2002). In undifferentiated NG108.15 cells, β₄ is exclusively cytoplasmic, while present in nuclei after 13 days of differentiation (Figure 1H). Mean NCR values illustrate that nuclear β₄ expression is triggered by neuronal differentiation (Figure 1I).

Figure 1.

Presence of β₄ in the nucleus of neurons. (A) Immunohistochemical confocal images of 20 μm coronal sections of adult wt (upper panels) or lh (lower panels) mice brains showing the distribution of endogenous β₄ in the dentate gyrus (green, left panels). Neuronal nuclei were labelled with NeuN (blue, middle panels). Right panels illustrate merged images. The inserts show × 5 images of the CA3 region (initially × 20). (B) Western blots of adult wt and lh mice brain lysates indicating the presence of β₄ only in wt mice. Actin staining was used as an internal control. (C) β₄ immunoblotting from cytoplasmic and nuclear fractions of adult wt mice brain. β-Tubulin and nucleolin are used as indicators of nuclear and cytoplasmic fraction purities, respectively. (D) EM image of an ultrathin cryosection of the CA1 hippocampal region showing the intranuclear presence of β₄ labelled with antibody-coated 15 nm gold particles. Black and white arrows indicate the position of gold particles associated to heterochromatin and euchromatin, respectively. (E) Confocal images of hippocampal neurons from embryonic E18 mice brains in primary culture at 5 DIV (left panels) and 18 DIV (right panels). Green: immunocytochemical staining of β₄; red: membrane staining with concanavalin A (ConA) conjugated to rhodamine; blue: nuclear staining with ToPro3. (F) DIV-dependent evolution of β₄ NCR values in neurons (n=50 cells for each data point). (G) Nuclear density of β₄ in hippocampal neurons at 5 and 18 DIV as assessed by EM (n=28 and 20 nuclei, respectively). (H) Confocal images of NG108.15 cells 1 day before (−1 DIV) and 13 days (13 DIV) after induction of neuronal differentiation with 1 mM cAMP and serum reduction. Colour code is as in (E). Differentiated NG108-15 cells have larger nuclei than non-differentiated cells. (I) Mean NCR values expressed in percent as a function of culture time in vitro. DIV 0 represents the induction time of neuronal differentiation.

β₄ epilepsy mutation alters nuclear targeting

To search for a β₄ nuclear targeting domain, truncated β₄-EGFP constructs were expressed in CHO cells (Figure 2A). In CHO and HEK293 cells, β₄-EGFP spontaneously locates in the nucleus (NCR=10.1±3.3 for CHO cells). In contrast, truncated constructs are deficient in nuclear targeting suggesting that β₄ has no nuclear localization sequence (NLS) or that the NLS depends on β₄ structure integrity (Figure 2B and C). Indeed, five constructs display a distribution similar to EGFP alone (NCR ∼1). Two constructs, that contain β₄ guanylate kinase (GK) domain, show preferential cytoplasm distribution (NCR=0.27±0.2 and 0.26±0.2 for β_216–402-EGFP and β_216–519-EGFP, respectively). β₄-EGFP truncation therefore results in 10- to 40-fold decrease of nuclear targeting efficiency. These data were confirmed for two truncated β₄ constructs by western blot analyses of nuclear and cytoplasmic fractions of transfected HEK293 cells (Supplementary Figure 2a). An earlier report mentioned the importance of a restricted N-terminal domain for β₄ nuclear accumulation (Subramanyam et al, 2009). As observed in this study, mutation of this domain significantly reduced nuclear accumulation albeit by a limited extent (by 1.41-fold, from a mean NCR value of 11.4±0.7 to 8.1±0.7). However, β_{R28A-R29A-S30A}-EGFP still significantly accumulates into the nucleus in contrast to the constructs presented in this manuscript.

Figure 2.

Nuclear targeting and structural integrity of β₄ are disrupted by the human juvenile epilepsy mutation. (A) Schematic representation of different truncated β₄ constructs in pEGFP-C1. The β_1–481-EGFP mutant is also shown. (B) Confocal images of CHO cells expressing some representative constructs, including β₄-EGFP and mutant β_1–481-EGFP. Expression time is 24 h after transfection. (C) Mean NCR values of EGFP fluorescence for each condition in transfected CHO cells. ***P≤0.01. (D) Western blot detecting EGFP in nuclear and cytoplasmic fractions of CHO cells after 2 days of transfection with EGFP, β₄-EGFP or β_1–481-EGFP. Cytoplasmic β_1–481-EGFP is less stable than cytoplasmic β₄-EGFP, as witnessed by the presence of lower molecular weight bands. These truncated constructs are deficient in nuclear accumulation as shown in (C), possibly because containing GK domains. (E) Confocal images of hippocampal neurones transfected at 7 DIV with β_1–481-EGFP (upper panels) or β₄-EGFP (lower panels). Images were acquired at 9 DIV on β-tubulin III-positive cells. (F) NCR values for each construct. ***P⩽0.01. (G) Confocal images of hippocampal neurons transfected with β_1–481-EGFP showing the comparative distribution of endogenous β₄ (red) and exogenous β_1–481-EGFP (green).

We engineered a β_1–481-EGFP mutant (Figure 2A) corresponding to the human juvenile myoclonic epilepsy mutation (Escayg et al, 2000) and assessed its cellular distribution in CHO cells (Figure 2B and C). Nuclear accumulation of β_1–481-EGFP mutant is strongly reduced (NCR=0.98±0.62; 10.3-fold lower than β₄-EGFP NCR). This mutation induces a loss in nuclear targeting as shown by western blot analyses of nuclear and cytoplasmic fractions of CHO cells (Figure 2D). Finally, β_1–481-EGFP mutant is completely excluded from hippocampal neuron nuclei at 7 DIV (NCR=0.32±0.18; Figure 2E and F). This corresponds to a 9.5-fold decrease in NCR value compared to β₄-EGFP (3.15±1.76), a reduction similar to that measured in CHO cells. Double imaging of immunolabelled endogenous β₄ and β_1–481-EGFP indicates the extent of nuclear exclusion of the mutant (Figure 2G). In conclusion, the nuclear targeting of β₄ requires the preservation of its structural integrity. The human epilepsy mutation alters this structural integrity and hinders nuclear targeting.

The SH3/GK interaction controls β₄ nuclear targeting

Interaction between β SH3 and GK domains regulates channel activity (McGee et al, 2004; Takahashi et al, 2005). Its role was therefore investigated on β₄ nuclear targeting. Three constructs were designed: β_1–166-EGFP, β_200–519-myc and β_200–481-myc (corresponding to the juvenile epilepsy deletion) (Figure 3A). In agreement with Figure 2C, none of these constructs show a specific nuclear targeting in CHO cells (Figure 3B). Expressed together, β_1–166-EGFP and β_200–519-myc reconstitute a nuclear-targeting-competent complex (Figure 3C; NCR=6.9±1.6 for β_1–166-EGFP and 7.1±1.1 for β_200–519-myc). In contrast, expression of β_1–166-EGFP with β_200–481-myc fails to form a nuclear complex (Figure 3C, right panels). Co-immunoprecipitation experiments show that β_200–481-myc no longer interacts with β_1–166-EGFP contrary to β_200–519-myc (Figure 3D). These data suggest that the human mutation prevents β₄ nuclear targeting by modulating the SH3/GK interaction. This issue was further addressed by investigating the effect of two mutations (SH3 L125P or GK P225R) known to disrupt the SH3/GK interaction (McGee et al, 2004; Takahashi et al, 2004; Figure 3E). As shown, β_{1–166-L125P}-EGFP no longer interacts with β_200–519-myc (Figure 3F). Also, both β_L125P-EGFP and β_P225R-EGFP are completely excluded from hippocampal neuron nuclei (Figure 3G). NCR values, 0.43±0.15 (L125P) and 0.43±0.11 (P225R), correspond to a 7.3-fold decrease in mutant β₄ nuclear localization (Figure 3H). These results demonstrate that β₄ nuclear targeting requires the internal SH3/GK and reveal a previously unrecognized structural role of the last C-terminal 38 amino-acid sequence in modulating the interaction between the C-terminal containing hemi-β₄ subunit and the SH3-containing β₄ fragment.

Figure 3.

The SH3/GK interaction is required for β₄ nuclear localization. (A) EGFP- or myc-tagged β₄-truncated constructs used for confocal microscopy and co-immunoprecipitation experiments. (B) Confocal images of CHO cells transfected with individual constructs. (C) Confocal images of CHO cells expressing β_1–166-EGFP with β_200–519-myc, or β_1–166-EGFP with β_200-481-myc (upper panels). Mean NCR values of EGFP or anti-myc fluorescence summarizing the effect of the co-expressions on nuclear localization (lower panels). ***P⩽0.001. (D) Immunoprecipitation experiments in CHO cells investigating the association of β_1–166-EGFP with β_200–519-myc or β_200–481-myc. Left panel: immunoprecipitation via anti-myc antibodies and western blot of EGFP, except first lane showing β₄-EGFP expression level in cell lysates. Right panel: western blot with polyclonal anti-β antibody (Bichet et al, 2000) indicating equivalent β_200–519-myc and β_200–481-myc immunoprecipitation levels. β_1–166-EGFP expression confirmed by cell EGFP fluorescence. (E) Schematic representation of β_L125P-EGFP, β_{1–166-L125P}-EGFP and β_P225R-EGFP mutants. (F) Lack of co-immunoprecipitation shows the absence of β_200–519-myc/β_{1–166-L125P}-EGFP interaction. Expression of β_1–166-EGFP and β_200–519-myc was verified by the cell fluorescence and western blot as in (D). (G) Confocal images of hippocampal neurons showing the cytoplasmic localization of β_L125P-EGFP and β_P225R-EGFP. Cells were transfected for 48 h at 6 DIV. (H) Mean NCR values for β₄-EGFP, β_L125P-EGFP and β_P225R-EGFP fluorescence in hippocampal neurons. ***P⩽0.001.

The phosphatase 2A Ppp2r5d (B56δ) subunit contributes to β₄ nuclear localization

Absence of a clear molecular determinant for β₄ nuclear targeting suggests that this process requires at least one protein partner, whereas defective β_1–481, β_L125P and β_P225R nuclear localization indicates that interaction with this (these) partners depends on SH3/GK interaction state. Yeast two-hybrid screenings were performed with a mouse brain complementary DNA library using β₄ as bait. The 62 positive clones were subjected to two-hybrid β-galactosidase assays using β₄ and β_1–481 as baits. B56δ, a 594 amino-acid protein containing a poorly defined NLS at its C-terminus, was revealed to interact with β₄ but not with β_1–481 (Figure 4A). B56δ is one of the two regulatory subunits of phosphatase 2A (PP2A), along with B56γ, known to target PP2A to the nucleus (McCright et al, 1996). Co-immunoprecipitation confirms that β₄-EGFP, but not β_1–481-EGFP, interacts with B56δ-myc expressed in HEK293 cells (Figure 4B). As expected, Cacna1e and RIM1 were also identified as β₄ partners (Kiyonaka et al, 2007). In contrast to B56δ, both proteins interact equally well with β₄ and β_1–481. These yeast two-hybrid assays also reveal that Cacnb3 (β₃), another β isoform, also interacts with B56δ suggesting the existence of redundant signalling pathways (Figure 4C). β_4c, a short β₄ splice variant, known to interact with the HP1γ chromo shadow domain (Hibino et al, 2003), does not interact with B56δ. Conversely, β₄ does not interact with HP1γ in these conditions (Figure 4C). Truncated β₄ constructs were analysed for B56δ interaction by yeast two-hybrid assays (Figure 4D). β_49–519 is the only truncated construct interacting with B56δ indicating that the N-terminus is not essential for this interaction. B56δ does not interact with β_L125P confirming that binding to β₄ requires an intact SH3/GK interaction (Figure 4D). Endogenous B56δ is expressed in both HEK293 and CHO cells in agreement with spontaneous nuclear targeting of β₄ (Supplementary Figure 3a and b). Exogenous B56δ-myc co-localizes with nuclear β₄-EGFP and further increases β₄-EGFP nuclear targeting in CHO cells (Supplementary Figure 3c). Conversely, expression of a B56δ shRNA, reducing the level of endogenous B56δ in CHO cells (Supplementary Figure 3b), leads to a reduction of β₄-EGFP nuclear targeting (NCR value dropping from 11.5±0.7 to 7.3±0.5 or average 4.3-fold redistribution between nuclear and cytoplasmic pools by WB) (Figure 4E and F). In agreement with binding data, B56δ-myc expression has no effect on β_1–481-EGFP distribution in CHO cells (Supplementary Figure 3d). Expression of B56δ-myc in non-differentiated NG108-15 cells produces a marked nuclear re-localization of endogenous β₄ (Figure 4G; 11.8-fold increase in NCR). The influence of B56δ on endogenous β₄ nuclear localization was further investigated using B56δ^−/− mice (Louis et al, 2011). Hippocampal neurons from B56δ^−/− mice show a significant 1.65-fold decrease in NCR value at 11 DIV (Figure 4H). Similarly, the density of immunogold-labelled endogenous β₄ is significantly decreased by 1.61-fold in B56δ^−/− adult hippocampal neurons compared to wt hippocampal neurons (Figure 4I). In conclusion, B56δ is an SH3/GK conformation-sensitive β₄ partner that contributes to nuclear distribution of β₄. Since a fraction of β₄ is still nuclear in the absence B56δ, we conclude that other B56 isoforms and/or nuclear partners also help β₄ accumulation in neuronal nuclei. β_1–481-deficient nuclear targeting originates from its inability to interact with B56δ.

Figure 4.

B56δ/β₄ interaction contributes to β₄ nuclear localization. (A) Yeast two-hybrid results indicating the interaction level of β₄ and β_1–481 with B56δ as a function of time (left). The interactions were scored as the ratio of β-galactosidase activity to His prototrophy. Schematic representation of mouse B56δ showing the NLS amino-acid sequence (right). The position of seven HEAT repeats is shown by homology with B56γ (Magnusdottir et al, 2009). (B) Co-immunoprecipitation experiments determining the interaction between β₄-EGFP or β_1-481-EGFP with B56δ-myc in HEK293 cells. Left panel: expression of B56δ-myc was confirmed by western blot (Mr 69 kDa). Right panel: pull-down with anti-myc antibody, and western blot with anti-EGFP antibody. (C) Interactions of β isoforms with B56δ and HP1γ in yeast two-hybrid assay. (D) Interactions of truncated and mutated β₄ constructs with B56δ using yeast two-hybrid assay. (E) Effect of B56δ shRNA and control shRNA on the NCR value of β₄-EGFP in CHO cells. ***P⩽0.001; NS, P⩾0.1. (F) Effect of B56δ shRNA and control shRNA on the nuclear/cytoplasmic distribution of β₄-EGFP in CHO cells, as assessed by WB. Histone H3 and β-tubulin are used as markers of the nucleus (N) and cytoplasm (C), respectively. Performed in duplicate. Average ratios of 4.2±0.32 (β₄-EGFP) and 0.97±0.81 (β₄-EGFP+shRNA B56δ). (G) Confocal images showing the cell distribution of endogenous β₄ in undifferentiated NG108.15 cells in the absence or presence of B56δ-myc (left panels). B56δ-myc was transfected 24 h before confocal imaging. Panels in green present the endogenous β₄ detected with an anti-β₄. Mean NCR values showing that B56δ-myc induces the redistribution of endogenous β₄ to the nucleus (right panel). ***P⩽0.001. (H) Mean NCR values of endogenous β₄ from wt and B56δ^−/− hippocampal neurons at 11 DIV. ***P⩽0.001. (I) Average immunogold-labelled endogenous β₄ in nuclei from pyramidal neurons from wt and B56δ^−/− mice hippocampus measured by EM (n=36 and 51 nuclei, respectively). ***P⩽0.001. Control counts in lh hippocampus were 0.51±0.36 gold particles/μm².

β₄ forms a complex with B56δ and active PP2A

Immunoprecipitation of endogenous β₄ from wt mice brain results in co-precipitation of B56δ and PP2A (Figure 5A). As expected, neither B56δ nor PP2A are immunoprecipitated from lh mice brain by β₄ antibodies. Similarly, PP2A is not precipitated by β₄ antibodies using B56δ^−/− mice brain indicating that PP2A/β₄ association requires B56δ. In HEK293 cells, overexpression of B56δ-EGFP with β₄-myc strongly enhances endogenous PP2A immunoprecipitation through β₄-myc (Figure 5B). Restricted immunoprecipitation of endogenous PP2A by antibodies against β₄-myc suggests that not all endogenous B56δ (Supplementary Figure 3a) is associated to PP2A in these cells, possibly due to a competition with other B regulatory isoforms unable to associate to β₄-myc. As expected, β_1–481-myc mutant does not precipitate endogenous PP2A in the presence of B56δ-EGFP (Figure 5B). β₄-myc/B56δ-EGFP complex immunoprecipitated from HEK293 cells shows PP2A phosphatase activity (Figure 5C).

Figure 5.

β₄ association to B56δ/PP2A is inhibited by channel expression and activated by membrane depolarization. (A) β₄ immunoprecipitates B56δ and PP2A from wt mice brain but not from lh or B56δ^−/− mice as shown by western blots. (B) Left panel: immunoprecipitation of endogenous PP2A by B56δ-myc or β₄-myc±B56δ-EGFP expressed in HEK293 cells. Right panel: similar experiments using β_1–481-myc±B56δ-EGFP. (C) Endogenous PP2A phosphatase activity associated to immunoprecipitated β₄-myc±B56δ-EGFP expressed in HEK293 cells. Dephosphorylation of p-nitrophenyl phosphate (pNPP), a generic phosphatase substrate, was measured by absorbance of the metabolite at 405 nm. Experiment in duplicate. (D) Confocal images of HEK293 cells 2 days after transfection with β₄-EGFP together with Cacna1a and Cacna2d2 subunits (VGCC) (left panel). Mean NCR values of EGFP fluorescence in various transfection conditions (n=50 for each condition, right panel). (E) Western blot analysis of β₄-EGFP subunit immunoprecipitated by B56δ-myc±VGCC expressed in HEK293 cells. Control IP represents precipitation in absence of anti-myc antibodies. (F) Western blot analysis of B56δ-EGFP immunoprecipitated by β₄-myc (left panels) or β_1–481-myc (right panels) expressed in HEK293 cells together with VGCC, without or with a 30-min depolarization by 140 mM KCl. (G) Effect of HEK293 membrane depolarization (140 mM KCl, 30 min) on β₄-myc or β_1–481-myc complex formation with PP2A in the presence of B56δ-EGFP and VGCC. (H) Mean NCR values of β₄ immunofluorescence in wt or B56δ^−/− hippocampal neurons at 18 DIV (control or 1 μM TTX). TTX was added at 11 DIV in the culture media (n=50 for each condition). (I) Effect of a 1-h application of 40 μM biccuculine and 400 μM 4-AP on average β₄ NCR value in wt and B56δ^−/− hippocampal neurons. NS, non significant; ***P<0.001.

Membrane depolarization triggers β₄ association to B56δ

β₄-EGFP, expressed together with Cacna1a and Cacna2d2, pore-forming and auxiliary subunits of VGCC, loses its nuclear distribution in HEK293 cells (Figure 5D, NCR=0.25±0.1, i.e., 40-fold reduction). Part of β₄-EGFP staining is cytoplasmic instead of at the plasma membrane, possibly by interaction with Cacna1a stuck in the endoplasmic reticulum. As expected, β_1–481-EGFP mutant distribution remains cytoplasmic in the presence of VGCC (Figure 5D, right panel). In the presence of Cacna1a and Cacna2d2, B56δ-myc no longer interacts with β₄-EGFP (Figure 5E). These data indicate that (i) B56δ does not form a higher order molecular complex with the channel and (ii) β₄ association to Cacna1a overrides its interaction with B56δ. In addition, B56δ-EGFP does not affect VGCC kinetics and current densities in HEK293 cells (Supplementary Figure 4a and b). Also, Ca²⁺ currents were not modified in hippocampal neurons from B56δ^−/− mice (Supplementary Figure 4c and d), confirming that B56δ does not associate to the channel complex. Previous reports evidenced that β/channel interaction is reversible (Restituito et al, 2001; Sandoz et al, 2004; Spafford et al, 2004), but the impact of membrane depolarization on this interaction has not been studied. Membrane depolarization of HEK293 cells enhances β₄-myc interaction with B56δ-EGFP, while it does not promote any interaction between β_1–481-myc and B56δ-EGFP (Figure 5F). In the presence of VGCC, the depolarization-induced B56δ-myc/β₄-EGFP interaction requires external Ca²⁺ (Supplementary Figure 5). In these conditions, β₄-myc precipitates endogenous PP2A only upon membrane depolarization (Figure 5G). In contrast, β_1–481-myc does not precipitate PP2A upon depolarization. Finally, blocking electrical activity of 11 DIV hippocampal neurons by 1 μM tetrodotoxin (TTX) produces a significant 2.1-fold reduction of β₄ NCR value measured at 18 DIV (Figure 5H). This TTX effect is not observed in B56δ^−/− hippocampal neurons, demonstrating that β₄ nuclear accumulation induced by electrical activity relies on B56δ. Note the 2.45-fold reduction in NCR value between wt and B56δ^−/− control neurons at 18 DIV in agreement with B56δ role in β₄ nuclear accumulation. Conversely, a 1-h stimulation of neuronal electric activity by a biccuculine/4-AP cocktail produces a significant 23.5±1.9% increase in NCR value that is not observed in B56δ^−/− hippocampal neurons (Figure 5I). Depolarization of the plasma membrane therefore represents an important signalling event that allows β₄/B56δ interaction and favours β₄ nuclear accumulation.

β₄ represses TH gene expression by interacting with the thyroid hormone receptor alpha

To determine whether β₄ regulates transcription of endogenous genes, we analysed a publicly available microarray data set comparing gene expression in wt and lh mice cerebellum (data set GSE6275 from Gene Expression Omnibus; http://www.ncbi.nlm.nih.gov/geo). A volcano plot representation of our statistical analysis identifies 56 upregulated and 38 downregulated genes (over two-fold expression change, P<0.05) in lh mice cerebellum (Supplementary Figure 6). Based on gene ontology annotations, clustering of regulated transcripts in functional groups illustrate that many encode proteins with functions related to signalling and ion transport (Supplementary Figure 6). Among these genes, those coding for TH and for BC031748 cDNA sequence are the most upregulated and downregulated, respectively (Figure 6A). The effect of β₄ on gene expression was confirmed on a set of genes by qRT–PCR experiments using purified mRNA from wt and lh mice cerebellum (Figure 6B). From both approaches, TH gene shows the greatest variation in expression level; an interesting observation considering its relevance in epilepsy (Hess and Wilson, 1991; Bengzon et al, 1999; Donato et al, 2006). The upregulation of TH expression in lh mice cerebellum indicates that β₄ acts as a repressor on this gene. Transcripts upregulation or downregulation is lost in the hippocampus and in the cortex indicating the existence of compensatory mechanisms to the lack of β₄. Interaction of β₄ with the promoter region of TH gene was investigated by chromatin immunoprecipitation (ChIP) experiments. ChIP of the TH gene promoter from hippocampal neurons in culture, using antibodies against β₄, shows specific precipitation at 18 DIV but not at 1 DIV (Figure 6C). As a control, ChIP of the 3′-UTR region of TH gene was negative at 18 DIV. Similar results were obtained for ChIP experiments of the transmembrane protein 100 gene promoter (Supplementary Figure 7), suggesting that the TH promoter is not an isolated case. In wt mice brain, β₄ antibodies immunoprecipitate 1.89±0.42% of TH promoter input, corresponding to a 6.3-fold enrichment over control (Figure 6D). In contrast, no specific TH promoter immunoprecipitation is observed using lh mice brain (Figure 6D). β₄ antibodies also immunoprecipitate 1.37±0.06% of TH promoter from B56δ^−/− mice, indicating that β₄ association with the promoter does not necessarily require the B56δ partner (3.3-fold enrichment; Supplementary Figure 8a). Yeast two-hybrid screening, using β₄ as bait, was used to identify a transcription factor that targets β₄ to the TH gene. The thyroid hormone receptor alpha (TRα) came out as a positive candidate (Figure 6E). TRα is a nuclear receptor that contains a T3 hormone binding domain and a DNA binding domain interacting with Thyroid Receptor Elements (TRE). Interaction of TRα with β₄ was confirmed by co-immunoprecipitation experiments in HEK293 cells (Figure 6F). However, this interaction is barely detectable in wt brain lysates probably because it constitutes a minor interacting fraction of total brain β₄ (not shown). ChIP using TRα antibodies shows that TRα can interact with the TH gene promoter regardless of the presence of β₄ or B56δ (Figure 6G). Enrichments of the TH gene promoter were 4.92-, 4.77- and 4.27-fold for wt, lh and B56δ KO brains, respectively. This result is in agreement with the presence of putative TRE in the promoter region and in the coding sequence of the TH gene. The effect of TRα on TH gene promoter was tested using a luciferase reporter (Figure 6H). TRα has little effect on its own on the expression of luciferase, while its activation by the T3 hormone leads to a 4.4-fold repression of luciferase expression (Figure 6H). Interestingly, β₄, like T3, activates the repressing function of TRα (4.8-fold repression). The effect of β₄ on TRα activity may therefore constitute a first step of TH gene repression in agreement with the upregulation of this gene in the cerebellum in the absence of β₄.

Figure 6.

β₄ regulates gene expression by interacting with a transcription factor. (A) Histogram showing upregulated (in red) and downregulated (in green) genes displaying a >2-fold change in lh versus wt cerebellum mRNA levels as detected with transcriptomic probe sets. (B) qRT–PCR experiments showing upregulation of TH, TAC1, KLB and downregulation of BC031748 and RIF1 in lh versus wt cerebellum. Mean±s.d. of three experiments performed in triplicate. Data were normalized using three housekeeping genes (glyceraldehyde 3-phosphate dehydrogenase, transferin receptor and peptidylprolyl isomerase A). (C) Agarose gel of representative ChIP assay on hippocampal neurons at 1 or 18 DIV with anti-β₄ antibodies. (D) TH promoter immunoprecipitation in presence or absence (control) of anti-β₄ antibodies, expressed as percentage of input, from wt or lh mice brain (n=4–22, left panel). ***P<0.001. (E) Yeast two-hybrid results indicating the interaction level of β₄ with TRα as a function of time (upper panel). The interactions were scored as the ratio of β-galactosidase activity to His prototrophy. Schematic representation of mouse TRα showing the localization of the DNA Binding Domain (DBD) and the Ligand Binding Domain (LBD) (lower panel). (F) Western blot analysis of TRα-EGFP immunoprecipitated by β₄-myc expressed in HEK293 cells. (G) TH promoter immunoprecipitation by anti-TRα antibodies, expressed as percentage of input, from wt, lh or B56δ^−/− mice brain (n=4–20, left panel). ***P<0.001. (H) Schematic representation of the luciferase reporter system under the control of TH promoter (upper panel). Luciferase expression measured in the absence and presence of T3 in HEK293 cells transfected with the luciferase reporter alone or together with TRα±β₄ (lower panel). An everted repeat type 6 TRE is present in the promoter region of TH gene (TGGCCTTGCCTGAGGCCA) at position −341 to −323.

B56δ induces the recruitment of HP1γ by β₄ and contributes to association of the signalling complex with histones

ChIP experiments using B56δ antibodies reveal the presence of B56δ within the β₄ complex associated to the promoter of the TH gene (Figure 7A). Control experiments using B56δ KO mice brain demonstrate the specificity of the signal, while the absence of TH precipitation in lh brain denotes a key role of β₄ in targeting B56δ to the TH promoter. The presence of PP2A in the complex associated to the TH promoter was also confirmed by using PP2A antibodies (Figure 7B). TH promoter precipitation by PP2A antibodies was diminished in lh brain indicating the importance of β₄ in recruiting PP2A to the TH gene (Supplementary Figure 8b). Next, we questioned the role of B56δ and PP2A in TH gene regulation. Phosphorylation of histones regulates chromatin state and gene transcription (Jenuwein and Allis, 2001). Within the nucleus of wt CA1 hippocampal neurons, 45% of β₄ gold-labelled is associated to heterochromatin (black arrows; n=195/440, Figure 1D), whereas the remaining fraction is associated to euchromatin. Expression of B56δ-EGFP together with β₄-myc in HEK293 cells leads to a genome-wide dephosphorylation of Ser¹⁰ of histone H3 (Figure 7C), in agreement with earlier findings that PP2A dephosphorylates histone H3 (Nowak et al, 2003). This effect cannot be attributed to an increased population of transfected cells in G2/M phases of the cell cycle (Supplementary Figure 8c). In addition, immunoprecipitated β₄-myc/B56δ-EGFP/PP2A-HA complex allows dephosphorylation of a Ser¹⁰ phosphorylated histone H3 N-terminal peptide (Figure 7D). Finally, ChIP experiments using anti-histone H3 and anti-phosphorylated H3 Ser¹⁰ antibodies show that the absence of β₄ in lh brain is accompanied by an increase of histone H3 Ser¹⁰ phosphorylation at the TH promoter (Figure 7E). Ser¹⁰ dephosphorylation of histone H3 has been associated to recruitment of HP1γ (Fischle et al, 2005). Interestingly, HP1γ has been shown to interact with a PXVXV motif on β_4c (Xu et al, 2011), a short splice variant of β₄. Although this motif is also found in full-length β₄ (Figure 7F), no interaction with HP1γ has been detected by yeast two hybrid (Figure 4C; Hibino et al, 2003). Nevertheless, we questioned whether the interaction of β₄ with B56δ could control its interaction with HP1γ. Indeed, expression of B56δ-EGFP strongly enhances immunoprecipitation of HP1γ-EGFP by β₄-myc (Figure 7G). The basal level of precipitation of HP1γ-EGFP by β₄-myc is due to the presence of endogenous B56δ in HEK293 cells (Supplementary Figure 3a). The B56δ dependence of HP1γ/β₄ interaction, together with the PP2A-induced dephosphorylation of histone H3 Ser¹⁰, prompted us to investigate the interaction of this complex with histones. Immunoprecipitation of mice brain β₄ leads to the precipitation of histone H3 (Figure 7H). The specificity of this association is witnessed by the lack of histone H3 precipitation from lh brain. Importantly, β₄ association with histone H3 is lost in B56δ^−/− mice brain, in agreement with the specific role of B56δ in β₄/HP1γ interaction. In HEK293 cells, immunoprecipitation of β₄-myc leads to the co-precipitation of all histones (H2B, H3 and H4) (Figure 7I), constitutive of nucleosomes. This is in agreement with the endogenous expression of B56δ (Supplementary Figure 3a) and HP1γ (Supplementary Figure 3d) in these cells. Expression of B56δ further enhances β₄-myc/histone interaction (Figure 7I). Similarly, B56δ-myc requires the expression of β₄-EGFP to precipitate histone H3, in agreement with the fact that HP1γ binds β₄ and not B56δ (Figure 7J). B56δ-myc expressed together with β_L125P-EGFP or β_1–481-EGFP mutants, deficient for B56δ association, no longer precipitate histone H4 further supporting the B56δ dependence of β₄/HP1γ interaction (Figure 7K). Neuronal differentiation strongly increases the association of endogenous β₄ with histone H3 in NG108-15 (Supplementary Figure 9a), which correlates with the differentiation-dependent β₄ nuclear targeting (Figure 1H and I). In the presence of B56δ-EGFP and channel subunits, membrane depolarization increases β₄-myc, but not β_1–481-myc, association to histone H3 (Supplementary Figure 9b). Coherent with its role as PP2A regulatory subunit and its genome-wide effect observed in Figure 7C, B56δ-EGFP expression in HEK293 cells also results in dephosphorylation of Ser¹⁰ of β₄-myc-associated histone H3 (Supplementary Figure 9c). ChIP experiments using anti-HP1γ antibodies demonstrate the association of HP1γ to the TH promoter (Figure 7L). The strong decrease of this association in lh brain illustrates the role of β₄ in recruiting HP1γ to the TH gene. Finally, double mutation of the HP1γ binding motif of β₄ precludes the B56δ-dependent interaction of β₄ with histone H3 demonstrating that HP1γ or HP1γ-like proteins are required for β₄ association to histones (Figure 7M).

Figure 7.

Association of β₄ to HP1γ and histones requires B56δ. (A) TH promoter immunoprecipitation by anti-B56δ antibodies, expressed as percentage of input, from wt, lh or B56δ^−/− mice brain (n=2–8). **P<0.05. (B) TH promoter immunoprecipitation by anti-PP2A antibodies, expressed as percentage of input, from wt mice brain (n=12). **P<0.05. (C) Western blot of H3 and H3 Ser¹⁰P in HEK293 cells expressing β₄-myc alone or with B56δ-EGFP. (D) Mass spectrometry data showing dephosphorylation of the phosphorylated N-terminal histone H3 peptide (amino acids 5–13 phosphorylated on Ser¹⁰) by β₄-myc immunoprecipitated from HEK293 cells, co-transfected with B56δ-EGFP and PP2A-HA. (E) TH promoter immunoprecipitation by anti-Ser¹⁰P H3 or anti-H3 antibodies, expressed as percentage of input, from wt or lh mice brain (n=3–22). *P<0.1. (F) Schematic representation of β_4c and β₄ illustrating the localization of the HP1γ binding motif (critical residues are shadowed). (G) Western blot of HP1γ-EGFP following immunoprecipitation by anti-myc antibodies from HEK293 cells expressing various combinations of HP1γ-EGFP, B56δ-EGFP and β₄-myc. (H) Immunoprecipitation of histone H3 from wt, lh or B56δ^−/− brain by anti-β₄ antibodies. (I) Immunoprecipitation of H3, H2B and H4 histones from HEK293 cells by β₄-myc in the absence or presence of B56δ-EGFP. Control: immunoprecipitation in non-transfected cells (NT). (J) Immunoprecipitation of histone H3 in HEK293 cells by B56δ-myc in the absence or presence of β₄-EGFP. (K) Lack of co-immunoprecipitation of histone H4 and B56δ-myc in HEK293 cells in the presence of β_L125P-EGFP or β_1–481-EGFP mutants. (L) TH promoter immunoprecipitation by anti-HP1γ antibodies, expressed as percentage of input, from wt or lh mice brain (n=4–14). **P<0.05. (M) Mutation of the HP1γ binding motif prevents β₄-myc association to histone H3 in the presence of B56δ-EGFP.

β₄-dependent TH gene repression is mediated by electrical activity

Our results indicate that in neurons there should be a β₄-dependent coupling between electrical activity and TH gene repression. To validate this conclusion, the electrical activity dependence of TH mRNA levels was investigated using cerebellar granule neurons from wt and lh mice. Because these cells required depolarizing conditions to survive (25 mM KCl), which promotes basal level electrical activity, our control condition was defined as TH mRNA level after 2 days of TTX treatment (from 8 DIV to 10 DIV). Promoting electrical activity for 1 day, by removing TTX from 9 DIV to 10 DIV, induces strong TH mRNA level repression in wt (50%) but not in lh neurons (Figure 8A).

Figure 8.

Experimental validation and schematic illustration of the novel excitation-transcription pathway mediated by β₄. (A) Electrical activity reduces TH mRNA levels in primary cultures of cerebellar granule neuronal from wt mice but not from lh mice. Upper panel represents the experimental protocol. First condition in which TTX is added for 2 days (8–10 DIV) is used as control condition without electrical activity (blue colour), whereas second condition in which TTX is washed out at 9 DIV is used as the condition in which electrical activity is restored (red colour). Lower panel illustrates the fold-change in TH mRNA level induced by electrical activity using rpl27 as housekeeping gene and normalization to control condition. **P<0.05. (B) Schematic representation of the new signalling pathway and of the defective steps produced by the epileptic truncation of β₄.

Discussion

Here, we characterized a novel signalling pathway in which, under membrane depolarization, the VGCC β₄ subunit accumulates in the nucleus and acquires a gene regulatory function by successively recruiting the B56δ/PP2A complex, a transcription factor and the heterochromatin protein, HP1γ. Nuclear localization of β₄ was previously shown in mice brain by Subramanyam et al (2009). In contrast to what we observe with endogenous β₄, these authors show that depolarization of cultured neurons expressing V5-tagged β₄ promotes nuclear export of the tagged protein. Besides the possibility that the V5 tag may interfere with the β₄/ B56δ interaction we describe and thus modify β₄ nuclear targeting pathways, these results may highlight an active equilibrium between different pathways of β₄ nuclear import and export. This group identified the N-terminus of β₄ as a possible determinant of nuclear accumulation. In support with the previous hypothesis, our own results show that this domain only partially interferes with β₄ nuclear accumulation suggesting the existence of different pathways controlling β₄ subcellular localization involving different β₄ protein partners. We demonstrate here that the N-terminus of β₄ is not important for its interaction with B56δ suggesting that its contribution to nuclear accumulation occurs through a B56δ-independent pathway. In our work, we focused on the β₄ nuclear pathway that involves B56δ and that is disrupted by the human epileptic mutation of β₄. We show that in-vitro TRα acts as a transcription factor capable to interact with β₄ and repress TH gene expression. The in-vivo implication of TRα in controlling TH gene expression remains to be deciphered. The resulting nuclear platform, structured by β₄, binds to specific gene promoters, dephosphorylates histones associated to these promoters and orients HP1γ interaction with the dephosphorylated histones. The formation of this complex allows gene repression, first by a direct effect of β₄ on TRα, and second by HP1γ recruitment known to promote heterochromatin formation (Figure 8B). This later step may lead to long-term gene repression, while the first step is conceptually more dynamic. In lh brain, a complete disorganization of this signalling pathway is observed in the absence of β₄. The combined reduction in HP1γ levels and greater phosphorylation level of histone H3 Ser¹⁰ are two events that conceptually may favour the upregulation of TH expression. In that respect, further studies on the ability of nuclear β₄ complex to induce other histone post-translational modifications, such as methylation or acetylation, will complete the characterization of the role of this complex in gene regulation. While our data demonstrate the key role of β₄ in organizing TH gene repression, TH expression is not uniformly upregulated in all lh brain areas indicating the existence of compensatory mechanisms. In that respect, B56γ, that also targets PP2A to the nucleus (McCright et al, 1996), and β₃, another VGCC subunit, that is also addressed to the nucleus (Beguin et al, 2006) and binds B56δ (Figure 4C) and TRα (unpublished observation), might participate to this compensation process. The juvenile epilepsy mutant that normally associates with VGCC (Escayg et al, 2000) is defective for B56δ association. This mutant no longer forms a complex with PP2A, and is unable to translocate to the nucleus and to interact with HP1γ and histones (Figure 8B). Therefore, this mutation induces a full destabilization of the β₄ signalling pathway. At the structural level, formation of the B56δ/β₄ complex requires the interaction between the β₄ SH3 and GK modules. The nuclear targeting defect induced by the human mutation resides in the loss of this interaction highlighting an unexpected role of β₄ C-terminal domain in controlling β₄ conformation state. Association of B56δ with β₄ promotes β₄/HP1γ interaction further indicating the exquisite modularity of β₄ as an adaptor protein.

In earlier reports, B56δ/PP2A was shown to regulate gene expression through a complex interplay between DARPP-32 phosphorylation state (Ahn et al, 2007), phosphatase 1 (PP1) activity and histone H3 phosphorylation level (Stipanovich et al, 2008). PP1 complexes have been associated to other histone-modifying enzymes, such as histone deacetylases or kinases, with chromatin remodelling capabilities (Hsu et al, 2000; Murnion et al, 2001) and with transcription repressors or activators (Canettieri et al, 2003; Jin et al, 2003). Here, we illustrate that, like PP1, PP2A belongs to a signalling macromolecular complex which ultimately regulates gene expression by modulating the histone code. This complex is organized by the VGCC β₄ subunit that plays the role of a platform that, besides PP2A, also recruits a nuclear receptor and a heterochromatin protein, HP1γ. By homology with PP1, this protein complex is likely to be enriched by other nucleosome-modifying enzymes in the near future. It should be emphasized that long-term modifications accompanying epileptic disorders have previously been associated to changes in post-translational state of histones and to alterations in transcriptional programs (Huang et al, 2002; Sng et al, 2006). This work thus opens exciting perspectives for understanding the underlying molecular causes of epilepsy in which VGCC subunits have been implicated.

Materials and methods

Mice lines

Lethargic (B6EiC3H-a/A-lh) and wild-type mice (strain B6EiC3H) were obtained from Jackson Labs and genotyped as previously described (Burgess et al, 1997). B56δ^−/− mice were generated by exchanging 146 bp of exon 3, exons 4–14, and 45 bp of exon 15 of the Ppp2r5d gene with a neomycine cassette through homologous recombination (Louis et al, 2011). Animal procedures were run with the approval of the local ethical committee and all efforts were made to minimize both the suffering and number of animal used. Also, the experiments were carried out in accordance with the International Ethical Committee Guidelines (EEC Council Directive 86/609, OJ L 358, 1, December 12, 1987; Guide for the Care and Use of Laboratory Animals, US National Research Council, 1996) for the care and use of laboratory animals.

Electron microscopy

Mice were fixed by intracardiac perfusion during 10 min with 50 ml of fixation buffer (2% paraformaldehyde, 0.2% glutaraldehyde in 0.1 M phosphate buffer pH 7.3). The hippocampus was then isolated and kept in fixation buffer during 2 days. A transversal slice was made in the mid third of the hippocampus and dissection of the CA1/CA3 area in this slice was performed under a binocular magnifier. Hippocampal neurons were cultured in B100 plates during 5 or 18 days and fixed 2 h with fixation buffer. Cells were gently detached with a cell scraper, centrifuged and embedded in 10% gelatine. The cell pellet was then cut in small pieces. In all, 1 mm³ cell or tissue samples were then incubated during 4 h in 2.3 M sucrose and frozen in liquid nitrogen. Cryosections were made at −120°C using a cryo-ultramicrotome (Leica-Reichert) and retrieved with a 1:1 solution of 2.3 M sucrose and 2% methylcellulose (Liou et al, 1996). Cryosections were first incubated with primary anti-β₄ (Everest Biotech, 1:100), then incubated with rabbit anti-goat antibodies (Jackson, 1:400) and revealed with protein A-gold conjugated (CMC, Utrecht). Labelled cryosections were viewed at 80 kV with a 1200EX JEOL TEM microscope and images were acquired with a digital camera (Veleta, SIS, Olympus). Gold particles were counted and surface area measured with iTEM software (analySIS).

Hippocampal neurons, NG108-15 cell cultures and cerebellar granule neurons

Primary cultures of hippocampal neurons were prepared as previously described (Peris et al, 2009). Astrocyte proliferation was blocked by 10 μM arabinoside cytosine (AraC) 4 days after plating. Predominance of neurons was witnessed by Tuj1 labelling and lack of GFAP staining. NG108.15 cells were differentiated 2 days after plating by decreasing fetal calf serum to 1% and addition of 1 mM dibutyryl cAMP. Cell lines were transfected using jetPEI and neurons at 9 DIV using calcium phosphate. Dissociated cerebellar granule neurons were prepared as previously described (Kiyonaka et al, 2007) except that 0.02% (w/v) trypsin (Difco) was used for dissociation of cerebellar tissues. Three hours after cell plating, culture medium was replaced with Neurobasal (Invitrogen) with 2% B-27 (GIBCO), 26 mM KCl, 60 U/ml penicillin, and 60 μg/ml streptomycin, and renewed every 3 days. At 8 DIV, 0.1 μM TTX was added in the culture medium to block electrical activity. At 9 DIV, the TTX was washed out to promote electrical activity or maintained for continued electrical activity block. At 10 DIV, total RNA was isolated using NucleoSpin RNA XS (Macherey-Nagel) according to manufacturer’s instructions.

Plasmid constructions

cDNA constructs coding for EGFP-tagged β₄ fragments were generated by PCR amplification using rat β₄-EGFP plasmid as template (GenBank accession code L02315) and subcloned into pEGFP-C1 using HindIII and KpnI sites. Myc-tagged β₄ fragments were subcloned into pcDNA3.1(+) using HindIII and XbaI sites. For yeast two-hybrid assays, rat β₁, β₂, β₃, β₄ and β₄ mutants were subcloned into pGBK-T7 vector (Kiyonaka et al, 2007). QuikChange® Site-Directed Mutagenesis Kit (Stratagene) or overlap extension method was used to produce β_L125P-EGFP, β_P225R-EGFP, β_1-166-L125P-EGFP, β_P220A-V222A-myc and β_{R28A-R29A-S30A}-EGFP. Mouse B56δ (GenBank accession code NM009358), HP1γ (NM_007624) and TRα (NM_178060) were cloned from mouse brain Marathon-Ready cDNA (Clontech) using PCR and inserted into pCMV-tag3 vector and pEGFP-C2 vector using EcoRI and SalI site. The B56δ and control shRNAs (sense strand sequences: 5′-CATCGCATCTATGGCAAGTTT-3′ and 5′-CAACAAGATGAAGAGCACCAA-3′, respectively) were subcloned into TRC1.5 vector (Sigma). All constructs were sequenced. Human Cacna1a (GenBank accession code AF004883) and rat Cacna1a (code NM012919) were also used for expression experiments. The TH-luciferase reporter gene was as previously described (Robert et al, 1997) and contains 5200, bps of the TH promoter upstream of the Firefly luciferase coding region.

Cytochemical stainings

Cells were fixed with 4% paraformaldehyde in PBS for 15 min. After PBS washing, plasma membranes were stained 5 min with 5 μg/ml concanavalin A-rhodamine. Cells were permeabilized with 0.1% Triton X-100. Myc-tagged constructs were visualized using biotinylated anti-myc antibodies (1:600, Santa Cruz) and Cy3-Streptavidin (1:300, Amersham). Endogenous β₄ was labelled for 1–2 h with a goat antibody directed against its carboxyl-terminus (1/1000 dilution, Everest Biotech), followed by Alexa488 conjugated anti-goat (1:500, Molecular Probes). Cell nuclei were visualized using ToPro3 dye (T3605; Molecular Probes Inc). Images were obtained with a confocal microscope (Leica SP2). NCR values correspond to the average measure of 50 cells. In each cell, fluorescence intensities were measured in 40 different ROI (20 for the cytoplasm and 20 for the nucleus). Merged pictures were obtained by using Adobe Photoshop software.

Yeast two-hybrid assays

Rat β₄, subcloned into pGBK-T7, was used as bait to screen a mouse brain pACT2 library expressed in the yeast strain AH109 (Clontech) (Kiyonaka et al, 2007). In all, 5.0 × 10⁶ transformants were cultured with synthetic medium lacking adenine, histidine, leucine and tryptophan. His⁺ colonies were assayed in yeast cells for β-galactosidase activity by a filter assay. Signal strengths of interactions were evaluated visually after every 1 h of colour development. For mapping of the β₄ region that interacts with B56δ or HP1γ, yeast strain Y187 was co-transformed with each β bait vector and the B56δ or HP1γ prey vector. The interactions were evaluated by β-galactosidase activity on a filter.

Subcellular fractionation and western blotting

Transfected HEK293 cells were lysed with 10 mM Tris pH 7.5, 10 mM KCl, 1.5 mM MgCl₂, 0.5% Triton X-100 and protease inhibitor cocktail (Roche). Nuclei and cytoplasm were separated by 5 min centrifugation at 2000, r.p.m. Pellets corresponding to nuclei were sonicated in lysis buffer. For mice brain subcellular fractionation, brains were homogenized in 0.32 M sucrose, 20 mM HEPES, pH 7.4, 1 mM DTT, 1 mM PMSF and protease inhibitor cocktail (Roche). The homogenate was centrifuged 10 min at 800 g. The resulting crude nuclear pellet was washed twice with homogenization buffer, resuspended in a volume equivalent to the cytoplasmic fraction, and sonicated. Nuclear and cytoplasmic (supernatant) proteins, from equal volumes, were separated by SDS–PAGE and analysed by western blot. EGFP-tagged proteins were visualized using anti-EGFP antibodies (1:300, Santa Cruz or 1:5000 Rockland) and anti-β₄ as described (Kiyonaka et al, 2007).

Immunoprecipitation experiments

Transfected cells were lysed 30 min in 10 mM Tris pH 7.4, 1.5 mM MgCl₂, 60 mM KCl, 15 mM NaCl, 0.5% Triton X-100 supplemented with protease inhibitor cocktail (Roche), sonicated and centrifuged 5 min at 1500 g. Similar procedure applies for mice brain, except that they were homogenized first in lysis buffer. Protein extracts were incubated with streptavidin beads (Dynabeads® Streptavidin M-280; Dynal) precoated with 3 μg of biotinylated anti-myc antibody (Santa Cruz Biotechnology Inc.). For immunoprecipitation of endogenous β₄, protein extracts were incubated with 40 μl of Dynabeads protein G (Dynal) precoated with anti-β₄ antibodies (Everest Biotech). Beads were washed with PBS/0.1% Tween-20. Immunoprecipitated proteins were eluted with denaturating buffer and analysed by western blot. For immunoprecipitation of phosphorylated histones, lysis buffer was supplemented with 50 mM NaF phosphatase inhibitor. For immunoprecipitation and ChIP, primary antibodies were used at 10 μg: anti-β₄ from Everest or as described (Kiyonaka et al, 2007), anti-histone H2B from Upstate, anti-histone H3, anti-phosphorylated Ser¹⁰ histone H3, anti-histone H4, anti-HP1γ, and anti-TRα from Abcam, anti-B56δ as described in Martens et al (2004), anti-PP2A from Bethyl Laboratories Inc., and anti-actin from Sigma-Aldrich.

PP2A activity

PP2A activity of immunoprecipitates was measured using SensoLyte™pNPP protein phosphatase kit from AnaSpec. Incubation lasted 1 h at 37°C in assay buffer in the presence of p-nitrophenyl phosphate (pNPP). OD was measured at 405 nm.

Quantitative RT–PCR

Total RNAs were extracted from wt, lh or B56δ^−/− cerebellum using the RNeasy kit, supplemented with DNase I (Qiagen). cDNAs were synthesized from 1 μg total RNA using random hexamer primers (Promega) and Superscript II (Invitrogen). Primers recognizing mice sequences were designed with the Primer 3 software (http://frodo.wi.mit.edu/) or by Sigma service (Supplementary Figure 10). Real-time PCR was performed with a light cycler 480 instrument using the Lightcycler LC480 SYBR green I master (Roche) according to manufacturer’s instructions. The cycling protocol was 10 min at 95°C followed by 40 cycles of 3 steps each (10 s at 95°C, 5 s at 60°C and 10 s at 72°C). The specificity of the amplification was check by generating a melting curve ranging from 65 to 95°C. Multiple normalization of gene expression (Vandesompele et al, 2002) was performed using three housekeeping genes (glyceraldehyde 3-phosphate dehydrogenase, transferin receptor and peptidylprolyl isomerase A).

For TH expression levels in cerebellar granule neurons, cDNAs were synthesized using AMV reverse transcriptase XL and an oligo dT-adaptor primer (TaKaRa). Real-time PCR and cycling protocol were performed as above, except for the number of cycles (50). TH gene expression was normalized relative to rpl27 (de Jonge et al, 2007). Values were expressed relative to wt treated with TTX, which was given the arbitrary value of 1.

Chromatin immunoprecipitation

Frozen adult mice brain (wt, lh or B56δ^−/−) was first crunched into powder in a mortar filled with liquid nitrogen. This powder was resuspended in 1% formaldehyde (in PBS) and incubated for 20 min at RT. Crosslinking was stopped by addition of glycine (125 mM final concentration) and incubation 5 min. Fixed tissues were thus washed with (i) cold PBS (5 min), (ii) 10 mM HEPES, 0.5 mM EGTA, 10 mM EDTA and Triton X-100 0.25%, pH 7.4 (10 min) and (iii) 10 mM HEPES, 0.5 mM EGTA, 1 mM EDTA and 200 mM NaCl, pH 7.4 (10 min). Samples were then solubilized in 1 ml of buffer containing 50 mM Tris, pH 7.4, 10 mM EDTA, 1% SDS for 2 h at 4°C, sonicated until reaching an average 500–1000, bps DNA fragment size, and centrifuged 5 min at 20 000 g. In all, 100 μl of lysate (OD₂₈₀=0.266) was incubated overnight at 4°C in 1 ml of 0.01% SDS, 1.1% Triton X-100, 1.2 mM EDTA, 16 mM Tris pH 8.0 and 170 mM NaCl, 1 mM PMSF, and protease inhibitor cocktail, with or without 10 μg antibodies. Immunoprecipitation was performed by adding Dynabeads protein G (Invitrogen). Different washing steps and elution were performed as described (Dahl and Collas, 2008). Crosslinks were reversed overnight at 65°C. DNA samples were sequentially treated with RNAse A and proteinase K. DNA was extracted by phenol/chloroform and ethanol precipitated with glycogen as a carrier. DNA pellet was dissolved in water and real-time quantitative PCR analyses were performed on 25 ng of DNA in duplicate with selected primers (Supplementary Figure 10). On cultured hippocampal neurons, ChIP experiments were conducted on 5 × 10⁶ cells using the Magna ChIP^TM (MILLIPORE) according to manufacturer’s protocol. The PCR cycling protocol was 10 min at 94°C followed by 38 cycles of 3 steps each (30 s at 94°C, 30 s at 55°C or 61°C, and 30 s at 72°C), and 7 min at 72°C. Annealing temperature for the promoter region of TH was 55°C and for the 3’UTR of TH was 61°C.

Mass spectrometry

The histone H3 peptide phosphorylated at Ser¹⁰ QTARKS_pTGGC was obtained from Smartox Biotechnologies (Grenoble, France). Immunoprecipitates were washed with (in mM) 40 Tris pH 8.4, 34 MgCl₂, 4 EDTA, 4 DTT, and incubated overnight at 37°C with 1 mM peptide. Dephosphorylation was evaluated by MALDI-TOF (Applied 4800) mass spectrometry. Briefly, 1 μl of peptide solution was spotted directly onto the MALDI target, dried and 1 μl of 20% α-Cyano-4-hydroxy-cinnamic acid (Sigma) was added in 50% acetonitrile, 0.1% trifluoroacetic acid.