|Home | About | Journals | Submit | Contact Us | Français|
The γ-aminobutyric acid (GABA) type A receptor (GABAAR) is the major inhibitory neurotransmitter receptor in the brain. Its multiple subunits show regional, developmental, and disease-related plasticity of expression; however, the regulatory networks controlling GABAAR subunit expression remain poorly understood. We report that the seizure-induced decrease in GABAAR α1 subunit expression associated with epilepsy is mediated by the Janus kinase (JAK)/signal transducer and activator of transcription (STAT) pathway regulated by brain-derived neurotrophic factor (BDNF). BDNF- and seizure-dependent phosphorylation of STAT3 cause the adenosine 3′,5′-monophosphate (cAMP) response element–binding protein (CREB) family member ICER (inducible cAMP early repressor) to bind with phosphorylated CREB at the Gabra1:CRE site. JAK/STAT pathway inhibition prevents the seizure-induced decrease in GABAAR α1 abundance in vivo and, given that BDNF is known to increase the abundance of GABAAR α4 in a JAK/STAT-independent manner, indicates that BDNF acts through at least two distinct pathways to influence GABAAR-dependent synaptic inhibition.
The subunit composition of GABAARs, which mediate most fast synaptic inhibition in the mature brain, determines their function and their response to drugs and other modulators (1, 2). Most GABAARs are heteropentamers composed of isoforms of α, β,and γ subunits, with the most common subtype being α1β2γ2. Alterations in GABAAR subunit expression occur during development (1-5) and in animal models of epilepsy (6-9), alcohol dependence (10), anxiety (8, 11), and stress (12-14), but the molecular pathways that regulate these changes are largely unknown. Prolonged seizures [status epilepticus (SE)] result in alterations in the expression and membrane localization of several GABAAR subunits (α1, α4, γ2, δ) in hippocampal dentate granule neurons (9, 15-17). These alterations, which are associated with changes in phasic and tonic GABAAR-mediated inhibition and in GABAAR modulation by benzodiazepines, neurosteroids, and zinc, begin soon after SE and continue as animals become epileptic (9, 15, 17, 18). After SE in adult rodents, GABAAR α1 subunit messenger RNA (mRNA) expression decreases, α4 subunit mRNA expression increases, and animals develop the recurrent spontaneous seizures that define epilepsy (9, 15). In contrast, neonatal SE (postnatal day 10) results in increased GABAAR α1 subunit expression and does not lead to the subsequent development of epilepsy (19). Increasing GABAAR α1 expression in adults by viral-mediated gene transfer reduces the incidence of spontaneous seizures during the first 4 weeks after SE by 60% (20). Together, these data support a role for GABAAR α subunit changes in the process of epilepsy development (epileptogenesis).
Although the mechanisms mediating regulation of the α4 subunit after SE have been characterized (21, 22), the molecular signals that regulate seizure-induced changes in expression of the gene encoding the GABAAR α1 subunit (abbreviated GABRA1 for the human gene and Gabra1 for the rat gene) have not been identified. The human GABRA1 promoter (GABRA1-p) contains a functional adenosine 3′,5′-monophosphate (cAMP) response element (CRE) (23). Cyclic AMP response element–binding protein (CREB) is a transcription factor that is activated by phosphorylation at Ser133 (24, 25) in response to various stimuli (26-29). Phosphorylation of CREB at Ser133 is often associated with an increase in transcriptional activation of target genes (24, 30, 31); among these are various CREB family members, including inducible cAMP early repressor (ICER), a group of four proteins transcribed from an internal promoter of the gene encoding CRE modulator (CREM) (32, 33). ICER binds CREs to repress target gene expression (34, 35), and transcription from GABRA1-p is bi-directionally regulated by phosphorylated CREB (pCREB) and ICER (23). In transfected cultured cortical neurons, pCREB enhances the activity of GABRA1-p–containing reporter constructs, whereas ICER—or co-binding of ICER and pCREB—represses their activity. Increased phosphorylation of CREB and transcription of ICER may occur after seizures (36-40), but it is not known whether pCREB and ICER mediate a seizure-induced decrease in the activity of the rat GABAAR α1 subunit gene promoter (Gabra1-p). Moreover, the upstream pathways mediating increases in the transcription of ICER in response to seizures have not been identified.
We report that CREB and ICER regulate Gabra1-p activity in vivo, showing an increase in CREB phosphorylation and ICER expression in the hippocampal dentate gyrus (DG) after SE. This leads to enhanced binding of pCREB and ICER to the Gabra1-p CRE, resulting in transcriptional repression and a subsequent reduction in α1 mRNA expression and the number of α1γ2-containing GABAARs. Furthermore, using primary cultured hippocampal neurons, we show that pCREB is required for the ICER-mediated decrease in GABRA1-p activity and that brain-derived neurotrophic factor (BDNF) activation of the Janus kinase (JAK)/signal transducers and activators of transcription (STAT) pathway is the endogenous signal that drives the process. In particular, blockade of the JAK/STAT pathway inhibits SE-induced ICER up-regulation and the subsequent decrease in abundance of GABAAR α1 subunit mRNA in vivo, and small interfering RNA (siRNA) knockdown of STAT3 in primary hippocampal neurons blocks BDNF-induced activation of ICER transcription. Taken together, these data show that BDNF and the JAK/STAT pathway control GABAAR α1 subunit expression in the brain. Identification of signaling pathways regulating the most abundant GABAAR α subunit in the central nervous system (CNS) may facilitate the development of novel molecular therapies for disorders such as epilepsy, alcoholism, anxiety, and stress that are associated with reduced GABAAR α1 expression.
To determine if CREB family transcription factors are involved in regulating Gabra1 expression in response to SE, we looked for evidence of seizure-induced activation of the CREB pathway with the use of an antibody directed against Ser133 pCREB as an indicator. The abundance of pCREB normalized to β-actin in DG increased to 270% of controls [n = 4 controls (CTRL), n = 5 SE, P < 0.05] at 1 hour and 310% of controls (n = 4 CTRL, n = 5 SE, P < 0.05) at 24 hours after SE (Fig. 1, A and B). Western blot analysis indicated no changes in total CREB protein at 1 hour (n = 3 CTRL, n = 5 SE) or 24 hours (n = 4 CTRL, n = 5 SE) after SE (Fig. 1, C and D). We also measured CREB mRNA expression with real-time reverse transcription polymerase chain reaction (RT-PCR). CREB mRNA was not increased at 1 hour (n = 3 CTRL, n = 4 SE), 6 hours (n = 4 CTRL, n = 6 SE), 48 hours (n = 4 CTRL, n = 6 SE), or 1 week after SE (n = 4 CTRL, n = 5 SE) (Fig. 1E). There was an increase in CREB mRNA 24 hours after SE to 220% of controls (n = 4 CTRL, n = 6 SE, P < 0.05), but, as noted above, this was not associated with a significant change in the abundance of CREB protein.
Because SE in adult rats results in dentate granule cell (DGC) neuro-genesis and CREB phosphorylation has been shown in newly born DGCs during development (41), we used immunostaining to see whether pCREB activation occurred predominantly in regions of newborn DGCs. Immunostaining of hippocampal slices with antibody directed against pCREB at 1 and 24 hours after SE showed increased pCREB staining in DGCs throughout the DGC layer (Fig. 1F). pCREB staining extended beyond the inner third of the DGC layer where newborn DGCs are predominantly found, suggesting that the increase in pCREB was not limited to newborn DGCs.
Phosphorylation of most CREB isoforms in the brain is associated with increased transcriptional activation of target genes (25, 29) and is thus unlikely to be solely responsible for decreased transcription of the GABAAR α1 gene (Gabral) after SE. We therefore examined whether increased phosphorylation of CREB is accompanied by increased expression of the CREB-family repressors CREM and ICER. Using RT-PCR, we found robust increases in ICER mRNA abundance in DG after SE (Fig. 2A) starting at 1 hour (1360% of controls; n = 4 CTRL, n = 6 SE, P < 0.05), peaking at 6 hours (3270% of controls; n = 4 CTRL, n = 6 SE, P < 0.01), and continuing at 24 hours (1050% of controls; n = 4 CTRL, n = 6 SE, P < 0.01) and 48 hours (250% of controls; n = 5 CTRL, n = 6 SE, P < 0.01). By 1 week after SE, ICER mRNA expression was similar to that in control animals (n = 4 CTRL, n = 6 SE). ICER protein abundance was also increased 6 and 24 hours after SE as measured by Western blot (at 6 hours: 1900% of controls, n = 5 SE, n = 4 CTRL, P < 0.05; at 24 hours: 3330% of controls, n = 4 SE, n = 4 CTRL, P < 0.05; Fig. 2, B and C) and by immunostaining in the DGC layer (Fig. 2D). As with pCREB immunostaining, enhanced staining for ICER after SE was present diffusely throughout the DGC layer, suggesting that it was not limited exclusively to newborn DGCs. As would be expected if ICER is regulating Gabra1 transcription, the increase in ICER abundance precedes the increase in GABAAR α1 subunit mRNA expression (fig. S1). CREM mRNA abundance was found by RT-PCR to slightly increase 24 hours after SE (Fig. 2E; 137% of controls; n = 4 CTRL, n = 5 SE, P < 0.05), but not at 1, 6, or 48 hours or 1 week. No changes were observed in CREM protein abundance at 6 hours (n = 4 CTRL, n = 5 SE) or 24 hours (n = 4 CTRL, n = 4 SE) after SE (Fig. 2, F and G).
Changes in the expression of the GABAAR α1 and α4 subunits occur in a temporally coincident fashion after SE and have been hypothesized to result in changes in GABAAR composition that underlie deficits in receptor function found in animal models of temporal lobe epilepsy (TLE) (9, 21, 22, 42-44). To confirm whether SE-induced changes in α1 and α4 subunit abundance are associated with a change in the composition of GABAARs, we immunoprecipitated receptors from DG with antibodies directed against γ2 [which confers synaptic localization by virtue of its binding to synaptic scaffolding proteins such as gephyrin (45, 46)] and then determined the abundance of α1γ2- and α4γ2-containing GABAARs by Western blot analysis. Twenty-four hours after SE, the abundance of α1γ2-containing GABAARs decreased by 33%, whereas that of α4γ2-containing GABAARs increased by 41% (Fig. 3, A and B). Taken together with earlier study results that SE produces altered physiological and pharmacological properties and changes in subcellular localization of GABAARs in DGCs (9, 15-18) as well as the finding that α4γ2-containing receptors have kinetics that make them less effective than α1γ2-containing receptors at inhibiting neurons during repetitive stimulation (42), our finding of seizure-induced changes in α subunit composition of γ2-containing GABAARs provides evidence that transcriptional regulation of α subunit genes after SE results in functional changes in GABAAR populations.
A consensus CRE site (GABRA1:CRE) is present in the human and rat α1 promoters (GABRA1-p and Gabra1-p, respectively; Fig. 3C). To determine if SE-induced increases in pCREB and ICER mediate decreased Gabra1 transcription, we investigated whether SE produces a change in the binding of these factors to the rat Gabra1-p:CRE site. We immunoprecipitated the endogenous rat Gabra1–protein complex with antibodies directed against pCREB (Ser133) and detected precipitated Gabra1 genomic fragments by PCR. We used two sets of primers to detect precipitated Gabra1: one pair that flanks the CRE site and one that flanks a potential CCAAT–enhancer binding protein (C/EBP) site that is 545 base pairs (bp) upstream of the CRE. Although both primer pairs detected Gabra1 sequence in DNA subjected to sonication but not immunoprecipitated (called “input” DNA), only the primer pair amplifying the region of Gabra1 that contained a CRE site amplified DNA after precipitation with anti-pCREB antibody (Fig. 3, D and E). The C/EBP primer pair did not amplify any DNA after precipitation with anti-pCREB antibody (Fig. 3F), confirming the specificity of pCREB binding to the region of Gabra1 containing the CRE. Results of these chromatin immunoprecipitation (ChIP) assays show that there is a greater than 200% increase in the amount of pCREB at the Gabra1:CRE site 24 hours after SE (Fig. 3E).
All available antibodies to ICER also recognize CREM variants, making it impossible to specifically identify ICER binding to Gabra1 by standard ChIP. We therefore developed a DNA pull-down assay to purify nuclear proteins in DG tissue that specifically recognize the Gabra1:CRE so that these proteins could be identified by gel electrophoresis. Nuclear binding proteins prepared from the DG of animals 24 hours after SE and controls were incubated with streptavidin beads containing oligonucleotide duplexes corresponding to the Gabra1:CRE. ICER was identified from the bound proteins by size on an SDS–polyacrylamide gel electrophoresis (SDS-PAGE) Western blot. As shown in Fig. 3G (left panel), CREB bound to the Gabra1:CRE site in both the control and SE-exposed animals 24 hours after SE. ICER, however, was barely detectable in control animals but was abundant after SE. Additional CREM variants were not detected at the Gabra1:CRE site (Fig. 3G, right panel). Mutant A1 CRE containing oligonucleotide duplexes do not purify CREB or ICER proteins from nuclear extracts of cultured neurons (fig. S2).
GABRA1-p is differentially regulated by the binding of CREB and ICER to the CRE site, and overexpression of recombinant ICER in neurons decreases the presence of α1 subunits at the cell surface (23). Given that SE causes an increase in CREB phosphorylation and transcriptional induction of ICER and that ICER can bind as a heterodimer or as a homo-dimer, we determined whether pCREB is required for the ICER-mediated decrease in GABRA1 transcription. Overexpression of ICER in the presence and in the absence of a dominant-negative form of CREB (MCREB), which contains a mutation in the Ser133 phosphorylation site of pCREB, suggests that ICER-induced endogenous repression of Gabra1 promoter activity in embryonic hippocampal neurons (Fig. 4A) depends on pCREB. Taken together with the presence of pCREB at Gabra1-p after SE, these results suggest that CREB–ICER heterodimers repress Gabra1 transcription in vivo.
Seizures increase BDNF expression (22, 47-50) and limiting activation of the TrkB receptor inhibits development of kindling, suggesting that BDNF may play a role in epileptogenesis (51). Twenty-four hours of BDNF exposure increases the abundance of GABAAR α4 subunits and decreases the abundance of α1 subunits in primary hippocampal neurons in a manner similar to that seen in animal models of TLE (22). Here, we found that BDNF treatment of hippocampal neurons for 6 hours decreased α1 mRNA expression by 33% (Fig. 4B). We further found that this decrease was associated with an increase in ICER mRNA expression and protein abundance (Fig. 4, C and D), suggesting that BDNF-induced decreases in transcription of α1 may be mediated by an increase in ICER. The protein kinase A (PKA) inhibitor H-89 failed to block induction of ICER, as did inhibitors of calmodulin and phosphoinositide 3 (PI 3)-kinases previously shown to play a role in CREB activation (fig. S3), indicating that BDNF-induced ICER in primary hippocampal neurons is not PKA dependent.
Because the JAK/STAT pathway can be activated by cytokines and has been implicated in multiple diseases associated with the increased release of cytokines (52, 53), we examined whether this signaling pathway controls synthesis of ICER in response to BDNF and SE. BDNF treatment induced the activation of STAT3 (Fig. 4), and pretreatment with the pan JAK inhibitor pyridone 6 (P6, 500 nM) (54) prevented BDNF-induced ICER synthesis (Fig. 5, A, B, and C) and the decrease in α1 subunit mRNA expression (Fig. 5D) in primary hippocampal neurons. P6 failed to inhibit forskolin-induced increases in ICER expression (fig. S5), suggesting that JAK/STAT activation is essential for increases in ICER expression mediated by BDNF but not those mediated by other signaling pathways. P6 alone had no effect on ICER expression (Fig. 5, A, B, and C) or on α1 expression (Fig. 5D), nor did it have any effect on the amount of pCREB with or without BDNF stimulation (Fig. 5, E and F). To directly examine the contribution of the JAK/STAT pathway to the altered gene expression controlled by BDNF, hippocampal neurons were transfected with siRNAs that specifically target STAT3. As shown in Fig. 5G, when compared to scrambled siRNA controls, STAT3 siRNAs specifically attenuated the amount of BDNF-induced ICER mRNAs (43.1 ± 3.6%; n = 6, CTRL, n = 5, STAT3, P < 0.05) in a manner consistent with a 41.3 ± 0.5% transfectional efficiency of siRNA transfection in primary hippocampal neurons (see Materials and Methods).
We next determined whether activation of the JAK/STAT pathway elicits the decrease in transcription of Gabra1 that occurs in vivo in the pilocarpine model of epilepsy. We examined the activation of the JAK/STAT pathway at several time points after SE by measuring phosphorylated STAT3 (pSTAT3) and phosphorylated STAT1 (pSTAT1) abundance (Fig. 6, A to D). pSTAT3 abundance in DG increased to 380% of controls at 1 hour (n = 4 CTRL, n = 5 SE, P < 0.05) to 2190% of controls at 6 hours (n = 4 CTRL, n = 5 SE, P < 0.05), to 890% of controls at 24 hours (n = 4 CTRL, n = 6 SE, not significant), and to 860% of controls at 48 hours (n = 4 CTRL, n = 5 SE, P < 0.05) after SE (Fig. 6, A and B). pSTAT1 abundance in DG did not increase at 1 hour (n = 4 CTRL, n = 5 SE), but increased to 520% of controls at 6 hours (n = 4 CTRL, n = 5 SE, P < 0.05), to 1004% of controls at 24 hours (n = 4 CTRL, n = 6 SE, P > 0.05), and was not significantly different from controls by 48 hours (n = 4 CTRL, n = 5 SE) after SE (Fig. 6, C and D). The time points for these increases in STAT phosphorylation correlate temporally with the observed increases in ICER mRNA and protein abundance and precede α1 subunit changes (Fig. 2 and fig. S1), thereby supporting the hypothesis that the JAK/STAT pathway may be an upstream regulator of ICER-dependent decreases in Gabra1 transcription after SE.
To directly determine if the JAK/STAT pathway mediates ICER synthesis and α1 mRNA expression after SE, we used Alzet (Cupertino, CA) mini osmotic pumps to slowly infuse 1 mM of the JAK/STAT pathway inhibitor P6 directly into the DG beginning 48 hours before SE induction. Given that P6 was administered before SE was induced, we first determined whether P6 affects the acquisition of SE itself and found that pump implantation and slow infusion of P6 did not affect SE (fig. S4). The total dose of pilocarpine required to induce SE (fig. S4A), the latency to develop stage 5 seizures (fig. S4B), and the total dose of diazepam (milligrams per kilogram) required to stop SE (fig. S4C) were not different between the SE only group (SE), the group that was implanted with dimethyl sulfoxide (DMSO, vehicle)–filled pumps before SE (veh-SE), or the group that was implanted with P6-filled pumps before SE (P6-SE) (for all: n = 6 SE, n = 8 veh-SE, n = 8 P6-SE; P > 0.05).
P6 infusion blocked JAK/STAT pathway activation after SE as shown by the absence of an increase in P6-SE animals 48 hours after SE (P6-SE; Fig. 6, E and F). pSTAT3 abundance was 980% higher in the combined group that contained rats that underwent SE only and rats that underwent SE after implantation with DMSO vehicle pumps (SE + veh-SE) than in controls (n = 6 rats per group, *P < 0.05) but there was no significant difference in pSTAT3 abundance between P6-SE (n = 4) and controls (P > 0.05). Slow infusions of P6 also blocked increases in ICER mRNA expression and protein abundance that occur at 48 hours after SE (Fig. 6, E, G, and H). ICER protein was 510% higher in the SE + veh-SE group than in controls (n = 6 CTRL, n = 6 SE + veh-SE, *P < 0.05; Fig. 6, E and G). ICER protein abundance in the P6-SE group was not significantly different from that in controls (n = 6 CTRL, n = 8 P6-SE, P > 0.05; Fig. 6, E and G). ICER mRNA expression in DG increased to 990% of control values in the SE + veh-SE group (n = 6, *P < 0.05) but did not change in P6 pump–implanted rats (P6-SE, n = 4, P > 0.05) compared to controls (Fig. 6H).
Activated STATs bind to STAT consensus sequences in the promoters of target genes to increase their transcription, so we determined whether pSTAT3 might exert its effects on ICER expression by binding to a consensus site on the ICER promoter (ICER-p). We identified one STAT site in the proximal region of the proposed rat and human ICER-p (Fig. 6I) and used ChIP to determine whether there is an increase in associated pSTAT3 after SE (Fig. 6J). pSTAT3 binding to ICER-p increased nearly 1000% after SE (control, 0.016 ± 0.015 relative units; SE, 0.168 ± 0.01 relative units; Fig. 6, J and K), providing further evidence that the JAK/STAT pathway regulates ICER in vivo and identifying pSTAT3 as a mediator of ICER transcription in the brain.
Finally, we asked whether the decrease in Gabra1 transcription after SE was reversed on inhibition of JAK/STAT signaling in vivo. Animals subjected to SE and killed 9 days after implantation of DMSO vehicle or P6-containing Alzet pumps were used to determine whether JAK/STAT inhibition prevented decreases in Gabra1 transcription that persist in animals that become chronically epileptic after SE (Fig. 6L). Animals in the SE + veh-SE group showed a 69% reduction in α1 mRNA expression (n = 6 CTRL, n = 8 SE + veh-SE, P < 0.05), but there was no significant decrease in the P6-SE group (n = 4, P > 0.05; Fig. 6L), suggesting that JAK/STAT pathway activation is necessary for seizure-induced decreases in Gabra1 transcription.
Changes in GABAAR α subunit composition may be important in the molecular pathogenesis of several common CNS conditions including epilepsy (6-9), alcoholism (10, 55), anxiety (8, 11), and stress (12-14). Decreases in GABAAR α1 subunit expression and increases in α4 subunit expression in hippocampal DG after prolonged seizures result in altered GABAAR localization and function, which may facilitate the development of spontaneous seizures (20). BDNF mediates increases in GABAAR α4 transcription by stimulating the protein kinase C–mitogen-activated protein kinase (MAPK) pathway to activate early growth factor 3 (Egr3) binding to the early growth factor response element site of Gabra4 (21, 22); however, the mechanisms underlying the coordinate decreases in transcription of Gabra1 have been unclear. We now elucidate the signaling pathways that regulate Gabra1, showing that increases in pCREB and ICER occurring after SE mediate the repression of Gabra1 transcription through CRE site binding, resulting in reduced expression of α1 mRNAs and a loss of α1γ2-containing GABAARs. We further show that the JAK/STAT pathway drives enhanced ICER transcription after SE and that activation of this pathway is stimulated by BDNF. These findings suggest that the interplay of the CREB, JAK/STAT, and BDNF signaling pathways are critical for GABAAR α1 gene regulation and that these pathways may provide novel therapeutic targets for the treatment of epilepsy and possibly for other CNS disorders in which GABAAR α1 subunit decreases are a pathogenic mechanism.
Modulation of neuronal activity, particularly through the neuro-transmitter receptors for GABAAR, glutamate, and serotonin, can influence α1 subunit abundance (56-59), but the regulatory pathways mediating these effects are not known. A study using microarrays to screen for changes in multiple mRNAs in response to overexpressed dominant-negative CREB (MCREB) in vivo suggested that GABRA1 might be regulated directly or indirectly by CREB (60), but whether this mechanism was important for GABRA1-p regulation after SE was not known.
Either homodimerization of ICER or heterodimerization of CREB with ICER could repress promoter activity by blocking CREB homodimerization that would result in enhanced transcription (34, 35). In the current study, we find that in vivo in DG, SE stimulates increased binding of pCREB and ICER to Gabra1-p, suggesting that CREB and ICER bind as heterodimers at the Gabra1:CRE site to repress transcription. Using transfected GABRA1 promoter–reporter constructs with expression vectors for CREB, MCREB, and ICER, we provide further evidence that the repressive effect of ICER on GABRA1 promoter activity in transfected hippocampal neurons depends on pCREB (Fig. 4A).
No evidence was found that other repressors of the CREB family are involved in the transcriptional regulation of α1 subunit expression. The CREM gene encodes multiple truncated proteins that can bind CRE sites but lack the activation domain necessary to support transcription. The ICER antibody used in our studies detects other CREM isoforms, but no increases in these isoforms occurred after SE. Other CREB family repressors, such as activating transcription factor-4 (ATF-4), were not specifically examined, so their effects on Gabra1 regulation after SE remain unknown.
The excessive neuronal activity associated with SE stimulates many different signaling pathways that could result in enhanced phosphorylation of CREB and expression of ICER (51). Determining which of these pathways mediates the increase in ICER expression critical for decreased transcription of the GABAAR α1 subunit is crucial to understanding how cross talk between different signaling pathways can lead to specific changes in inhibitory neurotransmission. We focused on BDNF as a potential regulator of ICER because BDNF expression increases markedly after SE and because BDNF differentially regulates the abundance of α1 and α4 subunits in cultured neurons (22). We here confirm that BDNF induces ICER expression in primary hippocampal cultures, but that this induction does not depend on PKA activation (fig. S1), as expected based on our previous finding that PKA activation by forskolin increases transcription of ICER and decreases α1 mRNA abundance in cortical cultures (23). BDNF activates several cAMP–PKA–independent signaling pathways that could regulate ICER synthesis, including the Ras-MAPK, calmodulin (CaM)-dependent protein kinase IV, PI 3-kinase–protein kinase C, and phospholipase C-γ pathways (61-64). However, BDNF-induced ICER expression is not blocked by either the PI 3-kinase inhibitor LY294002 or the CaM kinase inhibitor KN93 (fig. S1), suggesting that another pathway must underlie BDNF-mediated ICER induction in primary cultured hippocampal neurons.
Because the JAK/STAT pathway has been implicated in stress-related cellular changes in function and has been shown to be activated in CNS disorders, including ischemia and one form of seizure (65-67), we tested whether BDNF might activate this pathway to induce ICER and decrease GABAAR α1 subunit expression in primary cultures. The JAK/STAT signaling pathway is activated by cytokines binding to their specific receptors, resulting in transphosphorylation of JAK kinases that then lead to phosphorylation of STAT proteins (68-72). Phosphorylation of STATs on tyrosine residues leads to STAT homo- or heterodimerization, trans-location from the cytoplasm to the nucleus, and binding to specific DNA elements (STAT-recognition sites) to regulate target gene expression (69, 70, 73). Such an element is found in the ICER promoter (Fig. 6I) and here we show that pSTAT3 association with this site is enhanced after SE in DG (Fig. 6, J and K). We further show that siRNA knockdown of STAT3 inhibits BDNF-induced ICER expression, as does blockade of JAK/STAT signaling with P6 in primary hippocampal cultures (Fig. 5G). Using P6 as a tool in vivo to explore the relationship of the JAK/STAT pathway to seizure-induced α1 subunit changes, we show that ICER induction and decreased transcription of α1 mRNA are pathway dependent. Our findings are consistent with recent studies showing that BDNF induces phosphorylation of JAK2, STAT1, and STAT3 to promote neurite growth in the major pelvic ganglion of the rat (74).
Numerous changes in GABAAR trafficking, localization, and subunit expression have been reported in TLE models (8, 9, 15, 17, 21, 75). In hippocampal DGCs, decreases in α1 subunit transcription and increases in α4 subunit transcription may play a critical role in epileptogenesis. Increases in α4 subunit expression in DGCs have been reported in different epilepsy models (3, 9, 75-77), and two recent studies have found increased colocalization of α4 and γ2 subunits in synaptic and perisynaptic locations on DGCs in epileptic animals (16, 17). Our current receptor pull-down studies show that α1 is replaced by the α4 subunitin γ2-containing receptors after SE. This provides further support for the concept that during epileptogenesis, altered expression of α subunits leads to changes in synaptic GABAAR composition and localization that may underlie impairments in synaptic inhibition in DGCs, including diminished benzodiazepine sensitivity, enhanced zinc sensitivity, reduced neurosteroid modulation, and diminished phasic inhibition in dendrites (9, 16-18).
The JAK/STAT and CREB/ICER pathways regulate a myriad of genes with diverse functions and modulation of either of these pathways may thus have a multitude of downstream effects, many of which may involve cell- and region-specific responses. Therefore, the final impact of global blockade of these pathways on epileptogenesis may be difficult to predict. Although the enhanced GABAAR α1 subunit expression in DG that results from JAK/STAT pathway blockade and subsequent ICER inhibition would be expected to have an antiepileptic effect, mutant mice lacking ICER have accelerated kindling (78) and develop more severe epilepsy after pilocarpine-induced SE (40). Consistent with this finding, ICER-overexpressing mice show retardation of kindling development (78). Whether the effects of short-term and transient blockade of ICER increases at the time of SE specifically in the hippocampal formation, such as done in the current study, will have a similar effect on epileptogenesis as constitutive under- or overexpression of ICER in the brain examined previously remains to be determined.
In summary, we show that seizure-induced decreases in GABAAR α1 subunit expression are mediated by JAK/STAT pathway stimulation of ICER expression, resulting in binding of CREB and ICER to the Gabra1 promoter. Our findings that BDNF treatment of hippocampal neurons stimulates phosphorylation of STAT3 and induces increases in ICER expression, and decreased transcription of Gabra1 similar to those seen after SE, suggest that seizure-associated increases in BDNF are likely part of the upstream signals stimulating the JAK/STAT pathway. These results are particularly interesting in light of the demonstration that BDNF up-regulates GABAAR α4 expression by stimulating expression of Egr3 via protein kinase C–MAPK pathway activation (21, 22). Taken together, these findings identify BDNF as a “master” transcriptional regulator that works through at least two distinct signaling pathways to coordinately influence GABAAR subunit expression.
Adult rats underwent pilocarpine-induced SE according to standard protocols (9). All rats treated with full-dose pilocarpine were confirmed to have prolonged stage 5 behavioral seizures (79). Control rats were given subconvulsive doses of pilocarpine. Rats were killed at 1, 6, 24, or 48 hours or 1 week after onset of SE and the DG was rapidly dissected and frozen for protein and RNA isolation for Western blots and real-time RT-PCR, respectively, or animals were rapidly perfused with 4% paraformaldehyde for immunocytochemistry and ChIP. All animal procedures were performed in accordance with Institutional Animal Care and Use Committee regulations and approved protocols.
RNA was extracted from microdissected DG with the use of the Trizol reagent protocol (Invitrogen, Carlsbad, CA). To synthesize complementary DNA (cDNA), 1 μg of RNA was separated and processed with the SuperScript II reverse transcription kit (Invitrogen) according to the manufacturer's instructions and then diluted 1:4 for storage and subsequent RT-PCR. For RT-PCR reactions, each sample was run in triplicate and each 25-μl reaction contained 1.25 μl ICER, CREB, CREM, or GABAAR α1 Taqman primer and probe sets from Applied Biosystems (Foster City, CA) or 1.25 μl Taqman cyclophilin probe (Applied Biosystems) with 1.25 μl of each cyclophilin primer (cycloREV, 5′-CCC AAG GGC TCG CCA-3′; cycloFWD, 5′-TGC AGA CAT GGT CAA CCC C-3′; IDT Technologies, Coralville, IA), and 12.5 μl Taqman Master mix, and 10 μl of sample cDNA. RT-PCR was performed on the SDS-7500 PCR machine (Applied Biosystems). The RT-PCR runs consisted of first 1 cycle of 50°C for 2 min, then 1 cycle of 95°C for 10 min, and 40 cycles of 95°C for 15 s and 60°C for 1 min. All values were normalized to cyclophilin expression in the same samples to control for loading variability, then expressed as percent change with respect to mean control values in the same run (defined as 1). Statistical significance was defined as a P value of less than 0.05 and was calculated with InStat software with the Mann–Whitney or Kruskal–Wallis tests as indicated.
Coimmunoprecipitation experiments were carried out with the use of the Pro-found mammalian coimmunoprecipitation kit (Pierce, Rockford, IL). Affinity-purified rabbit polyclonal antibodies (50 μg) raised against the γ2 subunit (Alpha Diagnostics, San Antonio, TX), were coupled to AminoLink Plus gels as specified by the manufacturer. Hippocampi from each rat were homogenized in 1 ml of M-PER buffer (mammalian protein extraction reagent) plus protease inhibitor cocktail (Roche, Indianapolis, IN) 1× and 1 mM phenylmethylsulfonyl fluoride (PMSF). Homogenized lysates were passed five times through a 21-gauge needle fitted to a syringe. Samples were centrifuged at 12,000g for 30 min at 4°C. The supernatant was recovered and protein concentration was determined by the Bradford assay (Bio-Rad Laboratories, Hercules, CA). Supernatants were incubated with the antibody-coupled gel overnight at 4°C with gentle rocking. Gels were washed five times with buffer and complexes were recovered with 100 μl of the elution buffer provided with the kit. Controls were performed by quenching the same gel before coupling antibodies to test for proteins that may bind nonspecifically to the gel. Approximately 35 μl of the eluate was analyzed by Western blot with anti-α1 or anti-α4 (both 1:1000 dilution, Novus Biologicals, Littleton, CO) and γ2 (1:1,000 dilution, Alpha Diagnostic) polyclonal antibodies.
Immunocytochemistry was performed with standard methods (80). Cryostat sections (10 μM) were blocked with horse serum, incubated with hydrogen peroxide, and then treated with rabbit polyclonal antibody raised against CREB phosphorylated at Ser133 (anti-pCREB) (Upstate, Lake Placid, NY; 1:500) or rabbit polyclonal antibody raised against CREM-1 (anti-CREM/ICER) (Santa Cruz Biotechnology, Santa Cruz, CA; 1:1000), which recognizes multiple CREM isoforms including ICER. After overnight incubation, the slices were treated with polyclonal anti-rabbit immunoglobulin G (IgG) antibody labeled with rhodamine (1:1000) or avidin conjugated and reacted with diaminobenzene (Sigma, St. Louis, MO; 1:1000).
Western blot was performed with modifications of published protocols (80). Protein (25 μg) extracted from DG was loaded into 10% (for pCREB and pSTATs) or 12% (for CREM/ICER and pSTATs) SDS–polyacrylamide gels and run for 1.5 hours at 110 V. Blots were then transferred to nitrocellulose membranes and blocked in 2% milk/tris-buffered saline with Tween-20 (TBS-T) (for pCREB) or 5% milk/phosphate-buffered saline (PBS) (for CREM/ICER). Membranes were incubated with rabbit polyclonal antibodies raised against CREB phosphorylated at Ser133 (anti-pCREB) (Upstate, 1:400), total CREB (anti-CREB) (Upstate, 1:500), CREM-1 (anti-CREM/ICER) (Santa Cruz Biotechnology; 1:1000), STAT3 phosphorylated at Tyr705 (anti-pSTAT3) (Cell Signaling Technologies; 1:1000), total STAT3 (anti-STAT3) (Cell Signaling Technologies; 1:1000), STAT1 phosphorylated at Tyr701 (anti-pSTAT1) (Cell Signaling Technologies; 1:1000) or total STAT1 (anti-STAT1) (Cell Signaling Technologies; 1:1000) overnight at 4°C in 1.5% milk/TBS-T (for pCREB or CREB) or 5% milk/PBS (for CREM/ICER), or 5% bovine serum albumin/TBS-T (for pSTATs), then washed and incubated with anti-rabbit antibody (1:5000) conjugated to horseradish peroxidase (HRP) for 2 hours. CREM and ICER bands were differentiated based on size (CREM, 31 kDa; ICER/ICERγ, 19/17 kDa). Protein bands were detected with the use of chemiluminescent solution (Pierce), then membranes were stripped and reprobed with rabbit polyclonal antibody raised against β-actin (1:5000, Sigma) in 5% milk/TBS-T. All values were normalized to β-actin expression in the same samples to control for loading amount variability and then expressed as percent change with respect to mean control values in the same run (defined as 1). (Exception: Fig. 4, E and F data from pSTAT3 Western blot were normalized to total STAT3 expression that does not change with BDNF treatment.) Densitometry was performed with NIH Image version 1.63. Statistical significance was defined as a P value of less than 0.05 and was calculated with InStat software with the Mann–Whitney or Kruskal–Wallis tests as indicated.
After the animal was perfused with 4% formaldehyde, DG tissue was dissected, homogenized, and kept immersed in formaldehyde to cross-link protein–DNA interactions. DNA was sheared, producing about 300-bp fragments. Size was confirmed by gel electrophoresis. DNA was immunoprecipitated with CREB and pCREB antibodies (both 1:150; Cell Signaling), then isolated and dissolved in 100 μl TE or for pSTAT3 antibody (Santa Cruz Biotechnology) in 25 μl nuclease-free water. Real-time PCR was performed to determine amplification in the linear range with the use of primers that flank the CRE site and an internal sequence-specific Taqman probe: forward, 5′TGG TAC CAC CTT CCT TTC TAA AAT AAA3′; reverse, 5′ATA CGT CCC AGC GCA AAC C3′; Taqman probe, 5′TCT CTC TGG CAT GAA GTC ACC GCC T3′. Data were normalized as percentage of antibody/input signal and expressed as percent change with respect to control (defined as 100%). Input is the signal from the DNA preparation before precipitation. Standard PCR for detection of precipitated genomic DNA was performed with the following primer pairs: for CREB binding to Gabra1-p, forward 5′-AGA GGA ATG CTT GCA GCA G-3′ and reverse 5′-GCT TCC CAA TAT CCA ATC TG-3′; for C/EBP binding to Gabra1-p (as a control for DNA fragmentation and site specificity of CREB binding), forward 5′-TGT CTG GCT GTC CTC TAT GC-3′ and reverse 5′AAT GAA CGC CAA GCT ACG AA-GC-3′; for STAT3 binding to ICER-p, forward 5′-GTT CAG TCC CTG AAA TGT GG-3′ and reverse 5′-GCC GTA CCA CAG TCC AAA AT-3′. Nonradioactive PCR products were visualized with standard agarose gel electrophoresis (Gabra1-p) and p33-labeled radioactive PCR products were visualized with 5% polyacrylamide gel electrophoresis followed by exposure to film (ICER-p). Negative controls for nonspecific DNA precipitation were either IgG, standard in the field, or an isotype antibody such as chicken polyhistidine tag (fig. S6). We see no difference between the negative controls for their ability to detect nonspecific binding.
Hippocampi from three control and three SE-exposed rats were individually homogenized in 1 ml of 1× PBS. The cell pellet was gently resus-pended in 500 μl buffer A [10 mM tris–Cl, pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM dithiothreitol (DTT), 1 mM PMSF, 1× Protease inhibitor cocktail] and incubated for 15 min, after which 25 μl of a 10% NP-40 solution (American Bioanalytical, Natick, MA) was added. The cell suspension was vortexed for 5 s and centrifuged at 12,000g for 30 s to pellet nuclei. The supernatant was removed and the nuclear pellet was resuspended in 400 μl of buffer C (20 mM tris–Cl, pH 7.9, 400 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF, 1× Protease inhibitor cocktail) and incubated for 1 hour at 4°C. The nuclear debris was then removed by centrifugation at 12,000g for 5 min at 4°C and the supernatant was transferred to a fresh tube. Glycerol was added to a final concentration of 20%. Before the nuclear extracts were used for DNA pull-down assay, buffer exchange was performed with the use of Zeba Desalt Spin Columns (Pierce). The nuclear extract buffer was changed to DNA pull-down buffer C (20 mM tris, pH 8.0, 1 mM EDTA, 10% glycerol, 1 mM DTT, 50 mM NaCl) and then protein concentration was determined by the Bradford assay (Bio-Rad Laboratories).
The DNA pull-down assay was performed as previously described (81). Oligonucleotide duplexes corresponding to the sequence of the CRE site in the α1 promoter (as well as the mutant form containing a 2-bp mutation; see fig. S2) and its flanking region were covalently linked to a biotin moiety at their 5′ ends. Sequence of sense strand: wild type-CTC TGG CAT GAA GTC ACC GCC TAT; mutant-CTC TGG CAT GAA CAC ACC GCC TAT. Streptavidin Dynabeads (DynaL A.S.A.) were washed three times in buffer A (5 mM tris, pH 8.0, 0.5 mM EDTA, 1 M NaCl). Annealed oligonucleotides were incubated with beads (200 μmol/mg of beads) for 15 min at room temperature in buffer A. Beads were then washed twice with 500 μl buffer A and three times with 500 μl buffer C (20 mM tris, pH 8.0, 1 mM EDTA, 10% glycerol, 1 mM DTT, 50 mM NaCl). Nuclear extract (600 μg) from hippocampal tissues in buffer C was incubated with 1 mg of streptavidin Dynabeads bound to the appropriate oligonucleotides. After 15 min incubation at room temperature, beads were washed three times with buffer C and eluted in 1 M NaCl. Eluted proteins were then analyzed by SDS-PAGE followed by Western blot with anti-CREB (Cell Signaling) and anti-CREM1 antibody (which detects multiple CREM isoforms including ICER) (Santa Cruz Biotechnology).
Primary rat hippocampal cultures were prepared from embryonic day 18 embryos (Charles River Laboratories) and transfected 6 to 8 days after dissociation with a modified calcium phosphate precipitation method (23).
siRNAs corresponding to STAT3 were chemically synthesized by Ambion as were negative control scrambled siRNAs. Transfections were carried out with RNAiFect transfection reagent (Qiagen). Primary cultured hippo-campal neurons were plated in six-well plates as described (21-23). Seven days after plating, siRNAs (1 μg) were diluted 1:100 with buffer EC-R to a total volume of 100 μl. Complex formation was formulated with 12 μlof RNAiFect transfection reagent. After 15 min of incubation, complexes were directly added onto the cells and incubated for 24 hours under normal growth conditions. The culture medium was replaced with Neuro-basal medium. The transfection was repeated for another 24 hours. After the second transfection, cells were treated with BDNF for 4 hours to stimulate ICER induction, and further analysis was performed. Transfectional efficiency of around 40% was measured by quantitation of fluorescent siRNA controls through microscopy.
The human GABRA1 promoter fragment ( 894/+70) in the pGL2 luciferase reporter vector (Promega, Madison, WI) was a generous gift of the D. H. Farb laboratory. The promoter fragment confers full promoter activity on primary neocortical neurons. The dominant-negative MCREB expression vector was a generous gift of the M. E. Greenberg laboratory. Wild-type CREB was generated with the use of MCREB as a template for site-directed mutagenesis reestablishing the Ser133 phosphorylation site. The ICER expression vector was generated from the CREM cDNA. ICER IIγ cDNA was amplified from an IMAGE human cDNA clone (accession no. BC090051, Open Biosystems, Huntsville, AL) by PCR. The sequences of the primers were as follows: ICER IIγ,forward 5′-CGG GAT CCA TGG CTG TAA CTG GAG ATG ACA CAG CTG CCA CTG GTG ACA TGC CAA C-3′; ICER IIγ, reverse 5′-GCT CTA GAC TAG TAA TCT GTT TTG GGA GAA CAA ATG-3′. The amplified fragments were cloned into the BamH I (5′ end) and Xba I (3′ end) sites of the expression vector pcDNA3 (Invitrogen). The construct was sequenced to confirm identity. A Western blot analysis was performed after overexpression of the ICER IIγ cDNA construct in human embryonic kidney–293 cells to confirm expression and identity. Eight micrograms of total DNA was transfected into each well of a six-well dish and luciferase activity was assayed 12 to 16 hours after transfection with the Victor 1420 detection system (E. G. Wallace). Luciferase counts were normalized to protein in each dish (counts per minute per microgram).
Sprague–Dawley male rats were obtained from Zivic Miller (New Castle, PA), preimplanted with mini osmotic pumps (model 1002; Alzet) that were prefilled to deliver 1 mM P6 or DMSO vehicle at a rate of 0.5 μl/hour. Pumps were attached to unilateral cannulae to deliver P6 or vehicle into the right DG at the following bregma coordinates: anterior/posterior, −4.5 mm; medial/lateral, 2.0 mm; dorsal/ventral, 3.8 mm. SE was induced with pilocarpine as described above 48 hours after pump implantation and the start of P6 or vehicle infusion. Rats were killed at 48 hours or 9 days after SE along with subconvulsive pilocarpine control rats that did not undergo surgery (CTRL group) and rats subjected to pilocarpine-induced SE that did not undergo pump implantation surgery (SE group). After decapitation, brains were removed, then hippocampi were rapidly dissected and fast-frozen as described above.
Fig. S1. GABAR α1 mRNA expression is decreased in DG after pilocarpine-induced SE.
Fig. S2. Mutant CRE site does not bind CREB or ICER after DNA pull down.
Fig. S3. BDNF-induced ICER is not dependent on PKA, CaM kinase, or PI 3-kinase activation.
Fig. S4. Inhibition of JAK/STAT pathway via slow infusion of P6 beginning 48 hours before SE does not affect SE.
Fig. S5. No effect of P6 on PKA-induced ICER synthesis.
Fig. S6. BDNF increases binding of pSTAT3 to the endogenous ICER promoter as compared to isotype antibody control.
Citation: I. V. Lund, Y. Hu, Y. H. Raol, R. S. Benham, R. Faris, S. J. Russek, A. R. Brooks-Kayal, BDNF selectively regulates GABAA receptor transcription by activation of the JAK/STAT pathway. Sci. Signal. 1, ra9 (2008).