Chronic glutamate treatment selectively modulates AMPA RNA editing and ADAR expression and activity in primary cortical neurons

Abstract

Adenosine-to-inosine RNA editing is a post-transcriptional process, catalyzed by ADAR enzymes, with an important role in diversifying the number of proteins derived from a single gene. In neurons, editing of ionotropic AMPA glutamate receptors has been shown to be altered under several experimental conditions, including severe pathologies, thus highlighting the potential significance of its modulation. In this study, we treated rat primary cortical cell cultures with a sub-lethal dose of glutamate (10 μM), focusing on RNA editing and ADAR activity. We found that chronic glutamate treatment down-regulates RNA editing levels at the R/G site of GluA2-4 subunits of AMPA receptors and at the K/E site of CYFIP2. These changes are site-specific since they were not observed either for the GluA2 Q/R site or for other non-glutamatergic sites. Glutamate treatment also down-regulates the protein expression levels of both ADAR1 and ADAR2 enzymes, through a pathway that is Ca²⁺- and calpain-dependent. Given that AMPA receptors containing unedited subunits show a slower recovery rate from desensitization compared to those containing edited forms, the reduced editing at the R/G site may, at least in part, compensate for glutamate over-stimulation, perhaps through the reduced activation of postsynaptic receptors. In summary, our data provide direct evidence of the involvement of ADAR1 and ADAR2 activity as a possible compensatory mechanism for neuronal protection following glutamate over-stimulation.

Introduction

Adenosine deaminase that act on RNA (ADAR) enzymes catalyze a post-transcriptional process called RNA editing.^1,2 This molecular mechanism consists of the enzymatic deamination of specific adenosines (A) nucleosides into inosines (I). Since inosine has similar base-pairing properties to guanosine, it is read as guanosine by both splicing and translation machineries, thus generating different RNA molecules from those coded by DNA.³ RNA editing contributes to the diversification of the information that is encoded in the genome of an organism, thereby providing a greater degree of complexity.⁴ Currently, the conversion of A to I is thought to be the most common RNA editing process in higher eukaryotic cells, especially in the brain.⁵

In mammals, 3 members of the ADAR family have been characterized so far. ADAR1 and ADAR2 are active enzymes expressed in many tissues, while ADAR3 is expressed in the CNS⁶ more abundantly in glial cells than in neurons.⁷ To date, no functional RNA editing activity has been attributed to this enzyme.²

The majority of RNA editing sites is located within intragenic non-coding sequences: the 5’UTR, 3′UTR and intronic retrotransposon elements, such as Alu and long interspersed elements.^8,9 However, A-to-I RNA editing in mammalian cells was originally described for a number of protein-coding RNA sequences, resulting in dramatic changes of protein functions.⁴ Given its critical role, ADAR expression and activity must be tightly controlled by cells. The subcellular distribution of ADARs¹⁰ and their interaction with inhibitors¹¹ and activators^12,13 have been shown to influence ADAR activities. Accordingly, the knock-out of ADAR1 or ADAR2 in mice resulted in embryonic lethality or death shortly after birth, respectively,^14-16 clearly indicating that A-I RNA editing is essential for normal life and development.

RNA editing targets are abundant in the central nervous system (CNS)^5,17 where proteins involved in synaptic transmission are recoded.^4,18 Dysregulation of this process might result in profound alterations in neuronal signaling and give rise to severe neurological disorders.^19,20,21-23 RNA editing has profound effects on glutamate neurotransmission given that it recodes several glutamate receptor subunits.^24-25 Glutamate is the most important excitatory neurotransmitter in the CNS and is involved in cognitive functions such as learning and memory. The neurotransmitter can potentially activate 3 types of ionotropic glutamate receptors, namely N-methyl-D-aspartate receptor (NMDA), α-amino-3-hydroxy-5-methyl-isoxazole-propionic acid (AMPA) and kainate receptors.²⁶ RNA editing might modify AMPA glutamate receptor subunits (GluAs) in several specific positions.

The editing positions have been named on the basis of the amino acid substitution, such as the Q/R site in AMPA GluA2 and the R/G site in GluA2, GluA3, and GluA4.^24-25 The amino acid changes alter channel properties.^27,28 For the GluA2 Q/R site, the positively charged amino acid arginine in the inner channel pore makes the receptor channel impermeable to Ca²⁺, reduces its ion conductance and alters its current/voltage relationship^29,30 as well as affecting receptor maturation and cellular trafficking.^31,32 Under physiological conditions, GluA2 is fully edited at the Q/R site and the resulting AMPA-GluA2-containing receptors are Ca²⁺-impermeable, whereas AMPA receptors lacking GluA2 are permeable to Ca²⁺.^33,34

Whether AMPA receptors are permeable or impermeable to Ca²⁺ can greatly affect neuronal plasticity and survival following injury or excitotoxicity events.³⁴ The editing loss at the GluA2 Q/R site creates an increased influx of Ca²⁺ leading to epileptic seizures and death, as reported for ADAR2 null mice.^15,35 In humans, transient ischemia reduced GluA2 subunit mRNA editing and decreased the abundance of ADAR2 mRNA, leading to cell death of pyramidal neurons.³⁶ In addition, in the spinal motor neurons of sporadic amiotrophic lateral sclerosis (ALS) patients, editing at the GluA2 Q/R position was severely decreased, presumably through the down-regulation of ADAR2 levels,^7,21,37,38 leading to highly Ca²⁺ permeable AMPA channels and in turn facilitating motoneuron death, as reported in conditional ADAR2 knockout.³⁹ Exposure of cortical cells to excitotoxic levels of glutamate has been reported to induce cleavage of ADAR2 and a subsequent loss of GluA2 Q/R editing, resulting in excitotoxic cell death.⁴⁰

GluA2, 3, and 4 receptor subunits are also edited at the R/G site, which is located just before the sequences involved in the splicing events forming the flip/flop isoforms, and thus seems to affect both the splicing events and the desensitisation properties of the AMPA receptor channels.^27,41 Edited receptors have an enhanced rate of recovery from desensitization, thus generating ion channels that are able to respond more rapidly to a train of impulses.²⁷ In a previous paper, we showed that glutamate overstimulation, induced by spinal cord injury (SCI), reduced AMPA R/G editing levels,⁴² presumably by dampening post-synaptic excitatory responses to glutamate in an attempt to limit the progression of cell death.

The over-activation of GluAs in several pathological conditions such as ALS, stroke, epilepsy and SCI is well established.⁴³ However, the GluAs functional state is tightly controlled by post-transcriptional modifications and the role of these regulatory events in neurological disorders has not yet been elucidated.

The aim of our work was to evaluate the role of RNA editing of AMPA receptors and ADAR enzymes during glutamate over-stimulation. We found that chronic glutamate treatment in neuronal cells selectively affects the AMPA receptor R/G editing levels through a cellular cascade involving Ca²⁺ and calpain-dependent mechanisms.

Results

Effects of glutamate exposure on cell viability

We evaluated neuronal cell survival (DIV14 cortical cultures) 24 hours after exposure to different glutamate concentrations (10, 50, 100 μM) by simultaneous staining with propidium iodide and antibody anti-NeuN protein. Figure 1 shows that 24 h incubation promoted a dose-dependent reduction in cell viability, expressed as a percentage of NeuN positive cells that present propidium iodide staining. The highest level of cell viability was found with the lowest glutamate concentration (10 μM). Based on these data, we chose 10 μM of glutamate, i.e. the experimental condition showing the lowest cell death, for the following experiments.

Figure 1.

Kinetics of glutamate mediated excitotoxicity. Cortical neurons were treated with different doses of glutamate for 24 h. The graph reports the percentage of NeuN positive cells (a neuronal cell marker) that present propidium iodide staining (cell death marker).

Editing level quantification after chronic glutamate treatment

We exposed cortical neurons to glutamate for 24 h and then collected the cells immediately or 72 h later, for editing analysis. The Q/R site of GluA2 which is almost fully edited under physiological conditions, was not affected after 24 h of chronic glutamate stimulus or 72 h of washout (Fig. 2A), both at the pre-mRNA and mRNA levels (Fig. S1). On the other hand, the R/G site was affected considerably, and the changes were maintained even after washout. Concerning GluA2 R/G editing levels (Fig. 2B), 2-way ANOVA showed an effect of treatment (P < 0.001) in both Flip and Flop variants. We observed an overall reduction in GluA2 R/G editing (−48.1%; P < 0.001), in the Flip variant as well as in the Flop variant (−41.1%; P < 0.001), after 24 h of chronic treatment of mature neurons. Such glutamate-induced downregulation persisted for at least 72 h, with a decrease in editing levels of −52.2% in the Flip (P < 0.001) and −44.5% in the Flop variant (P < 0.001).

Figure 2.

Editing levels of AMPA receptor subunits after chronic glutamate treatment. The measurements were taken after 24 h of continuous treatment (Ctr 24 h and Glu 24 h) and after 72 h of glutamate washout (Ctr+WO and Glu+WO). (A) GluA2 Q/R editing level; (B) GluA2 R/G editing level for the Flip and Flop versions; (C) GluA3 R/G editing level for the Flip and Flop versions; (D) GluA4 R/G editing level for the Flip and Flop versions. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

A similar effect was observed for the GluA3 R/G site (Fig. 2C). Two-way ANOVA for the Flip variant showed both a treatment (P < 0.001) and a time (P < 0.001) effect, whereas for the Flop variant effects of treatment (P < 0.001), time (P < 0.001) and time × treatment interaction (P < 0.05) were found.

After glutamate chronic treatment, a robust decrease in the editing levels of both isoforms was evident at both time points. After 24 hr, the editing of the GluA3 Flip and Flop variants decreased by −46.6% (P < 0.001) and by −45.6% (P < 0.01), respectively. Editing levels remained low even after 72 hr of washout: −48.8% for the Flip variant (P < 0.001) and −61.1% for the Flop variant (P < 0.001).

Two-way ANOVA showed an effect of treatment (P < 0.001) for the GluA4 editing level (Fig. 2D). Significant decreases were observed for the Flip transcript both after 24 hr (−30.8%; P < 0.001) and after washout (−26.2%; P < 0.001). Similarly, the Flop variant showed a significant decrease after 24 hr (−36.4%; P < 0.001) and 72 hr of washout (−47.5%; P < 0.001).

To distinguish whether the treatment effects on editing were selective for the AMPA receptor or were due to a general loss of editing capability, we evaluated the mRNAs encoding for the Cytoplasmic FMR1-interacting protein 2 (CYFIP2), the serotonin 5-HT_2C receptor and the bladder cancer associated protein (BLCAP). We found that the editing site of CYFIP2 was affected by glutamate treatments. Two-way ANOVA showed both a treatment (P < 0.001) and a time (P < 0.001) effect. After 24 hr, editing of the CYFIP2 K/E site decreased after 24 h of treatment (−56.4%; P < 0.001) and after the washout (−65.3%; P < 0.001) (Fig. 3). In contrast, the 5 editing sites of 5-HT_2C and the 3 of BLCAP were not affected by glutamate treatment (Table 1).

Figure 3.

Editing levels of the CYFIP2 K/E site after chronic glutamate treatment. The measurements were taken after 24 h of continuous treatment (Ctr 24 h and Glu 24 h) and after 72 h of glutamate washout (Ctr+WO and Glu+WO). Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

Table 1.

Editing levels of the 5 5HT_2CR editing sites and the 3 BLCAP sites

	Ctr 24 h	Glu 24 h	Ctr + WO	Glu +WO
5HT2CR Site A	84.4 ± 1.17	80.5 ±  0.92	81.9 ± 0.79	78.7 ± 0.96
5HT2CR Site B	86.8 ± 0.32	85.0 ± 0.35	85.5 ± 0.56	83.1 ± 0.34
5HT2CR Site C’	7.5 ± 0.20	6.9 ± 0.67	7.7 ± 0.01	7.4 ± 0.93
5HT2CR Site C	26.5 ± 0.10	23.5 ± 1.00	22.3 ± 0.37	19.6 ± 1.05
5HT2CR Site D	72.3 ± 0.67	70.8 ± 2.32	70.4 ± 0.15	68.0 ± 0.93
BLCAP Y/C Site	51.3 ± 0.56	51.9 ± 0.60	52.8 ± 0.39	52.7 ± 0.19
BLCAP Q/R Site	2.9 ± 0.65	3.8 ± 0.18	4.0 ± 0.65	2.2 ± 0.11
BLCAP K/R Site	5.9 ± 0.10	6.2 ± 0.45	6.7 ± 0.60	5.1 ± 0.40

Modulation of glutamate receptor subunit protein expression after glutamate treatment

To evaluate the effects of glutamate treatment on AMPA subunit protein expression, we focused on GluA1 e GluA2 subunits which are the main AMPA subunits expressed. GluA1 protein expression decreased by about 50% after glutamate treatment (P < 0.001) and remained down-regulated by about 40% after washout (P < 0.01). In contrast, GluA2 protein expression was not affected by glutamate treatment (Fig. 4).

Figure 4.

GluA1 and GluA2 protein expression pattern after chronic glutamate treatment. The measurements were taken after 24 h of continuous treatment (Ctr 24 h and Glu 24 h) and after 72 h of glutamate washout (Ctr+WO and Glu+WO). The graphs report expression data normalized on GAPDH expression and relative to control samples. Representative Western blots are reported below the graphs. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

ADAR1 and ADAR2 expression after chronic glutamate treatment

To verify whether the reductions in the editing levels were related to changes in protein expression of the enzymes responsible for RNA editing, the expression levels of ADAR1 and ADAR2 enzymes were evaluated (Fig. 5). Two-way ANOVA showed an effect of treatment (P < 0.001) for both ADAR1 and ADAR2 protein expression. Both enzymes were downregulated both after 24 hr of chronic glutamate treatment and after 72 hr of washout. ADAR1 protein expression (Fig. 5A) decreased by 58.6% after 24 hr treatment (P < 0.001) and by 45.8% following washout (P < 0.01). The same result was observed for the ADAR2 protein (Fig. 5B), which decreased by 63.3% after 24 hr (P < 0.001) and by 57.9% after 72 hr (P < 0.001).

Figure 5.

ADAR1 and ADAR2 protein expression pattern after chronic glutamate treatment. The measurements were taken after 24 h of continuous treatment (Ctr 24 h and Glu 24 h) and after 72 h of glutamate washout (Ctr+WO and Glu+WO). The graphs report expression data normalized on GAPDH expression and relative to control samples. (A) ADAR1 protein expression and (B) ADAR2 protein expression. Representative Western blots are reported below the graphs. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

ADAR2 RNA editing and alternative splicing after chronic glutamate treatment

One ADAR2 splicing event results in the inclusion of an additional 47 nt cassette at the 5’end of the coding region, changing the reading frame of the mature ADAR2 transcript. The inclusion of this cassette depends on ADAR2 ability to edit its own pre-mRNA to generate a new intronic 3’acceptor site. In order to evaluate the potential change in ADAR2 catalytic activity, the self-editing levels at position -1 were analyzed.

After chronic glutamate treatment, ADAR2 self-editing (Fig. 6) was reduced by −53.4% (P < 0.001). The level of ADAR2 self-editing was also down-regulated (−59.6%), even after 72 hr of washout (P < 0.001). Two-way ANOVA showed an effect of treatment (P < 0.001), time (P < 0.01) and time/treatment interaction (P < 0.05). Following the editing variation, the splicing pattern of the canonical (−47 nt) and alternative forms (+47 nt) changed (Fig. S2).

Figure 6.

Effect of chronic glutamate treatment on ADAR2 catalytic activity. Estimation of ADAR2 self-editing at the intronic position -1. The measurements were taken after 24 h of continuous treatment and after 72 h of glutamate washout. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P <0 ,001).

Glutamate-induced RNA editing and ADAR protein decrease is Ca²⁺- and calpain- dependent

In order to investigate the potential mechanisms responsible for the modulation of ADARs expression and RNA editing level, we set up some experiments to evaluate the effect of Ca²⁺ influx, kinase activity as well as calpain action, which is an important protease activated by Ca²⁺ cascade.⁴⁰ Cortical neurons were treated with BAPTA/am (Ca²⁺ chelator, 20 μM), calpain inhibitor (10 μM), KN93 (CamKII inhibitor, 10 μM) in parallel with glutamate treatment. After glutamate treatment, the activation of calpain was demonstrated by the cleavage of spectrin (Fig. S3).

As in the previous experiments, the GluA2 Q/R editing level was not affected by glutamate. Similarly, neither BAPTA/am, KN93 or the calpain inhibitor treatment did not modify the editing levels at this site (Fig. 7A). As expected, chronic glutamate treatment induced GluA2 R/G editing down-regulation (−71.4%; P < 0.001). Interestingly, both BAPTA/am and calpain inhibitor treatment blocked the decrease in glutamate GluA2 R/G editing (Fig. 7B), but KN93 did not (GluA2 R/G editing downregulation: −47,7%; P < 0.001). However, the different treatments alone induced a partial down-regulation of GluA2 R/G compared to the untreated control.

Figure 7.

Editing levels of GluA2 Q/R (A) R/G sites (B) and CYFIP2 K/E site (C) in DIV14 cortical neuron chronically treated either with glutamate (24 h, 10 microM – gray bar) or without (white bars) in parallel with BAPTA/am (Ca²⁺ chelator), calpain inhibitor and KN93 (CamKII inhibitor) treatment. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after one-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

Considering the CYFIP2 K/E site (Fig. 7C), the glutamate-dependent downregulation (−66,96%, P < 0,001) was completely abolished by BAPTA/am and calpain Inhibitor treatments, but not by KN93 (CYFIP2 K/E editing down-regulation: −57,17%; P < 0.001).

The analysis of ADARs protein expression revealed a similar pattern of changes seen for GluA2 R/G and CYFIP2 K/E editing variations (Fig. 8). After 24 h, glutamate treatment caused a downregulation by −56.7% for ADAR1 (P < 0 .05), and by −67.5% for ADAR2 (P < 0.01). In contrast, BAPTA/am and calpain Inhibitor treatments prevented protein down-regulation, while KN93 did not prevent ADARs downregulation (by −61.7% for ADAR1 (P < 0.01) and by −62.9% for ADAR2 (P < 0.01)).

Figure 8.

Protein expression analysis of ADAR1 (A) and ADAR 2 (B) in DIV14 cortical neuron chronically treated with glutamate (24 h, 10 microM – gray bar) or not (white bars) in parallel with BAPTA/am (Ca²⁺ chelator), calpain inhibitor and KN93 (CamKII inhibitor) treatment. Representative protein gel blots are reported below the graphs. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after one-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

Analysis of ADAR cofactors after chronic glutamate treatment

To further shed light on ADAR activity modulation, we analyzed the expression pattern of several stimulatory factors such as Split Hand/Foot Malformation 1 (SHFM1), and RNA binding protein hnRNP A2/B1 (hnRNPA2/B1), as well as inhibitory factors such as ribosomal protein S14 (RPS14) and Serine/Arginine-Rich Splicing Factor 9 (SRSF9) which can modulate ADAR activity (Fig. 9).

Figure 9.

Expression levels of RPS14, hnRNPA1B2, SRSF9, SHF1 mRNA by qPCR after 24 h of glutamate treatment. Data are reported as Log₂ of the expression ratio R= 2^−ΔΔct (expression level of control sample is equal to 0) as in ⁶³ and represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after one-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

Chronic glutamate treatment induced an up-regulation of the inhibitory factor SRSF9 by a mean factor of 2.39 (log₂R= 1.22 P < 0.05), an increase was also found for the inhibitory factor RPS14, although this was not statistically significant. The two tested stimulatory factors were not affected by chronic glutamate treatment. Inhibition of Ca²⁺ entry, CamKII activity or calpain activity did not alter the expression levels of the 4 ADAR cofactors (Fig. S4).

GluA2 editing level quantification after acute glutamate treatment

We then investigated whether acute glutamate treatment,⁴⁰ had the same effect as the chronic treatment on the RNA editing level and ADAR protein expression. Cortical neurons were treated with 100 μM glutamate and the experiments were performed at different time points. The editing level of GluA2 R/G and Q/R sites was examined immediately after 1 hr of treatment and after 5 hr of washout. An additional 6 hours of continuous treatment was also performed.

Interestingly, the editing level of the GluA2 Q/R site was not affected in the acute treatments as previously observed in the chronic treatment (Fig. 10A). Regarding the GluA2 R/G editing site (Fig. 10B), the editing level was unchanged 1 hr after treatments, whereas after 5 hr of washout, a robust down-regulation (−39.1%; P < 0.001) was observed, which persisted even after 6 hr of treatment (−42.5%; P < 0.001). The two-way ANOVA displayed an impact of treatment (P < 0.001), time (P < 0.001) and time/treatment interaction (P < 0.001).

Figure 10.

Analysis of RNA editing in DIV14 cortical neurons acutely treated with 100 μM glutamate for 1 h; for 1 h of Glu and 5 h of wash out; for 6 h of continuous Glutamate treatment. (A) Editing level of GluA2 Q/R site and (B) editing level of GluA2 R/G site. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P < 0,05; ^**P < 0,01; ^***P < 0,001).

ADAR protein expression after acute glutamate treatment

Concerning ADAR1 (Fig. 11A), no difference in protein expression was observed after 1 hr of treatment while a decrease was seen after 5 hr of washout (−25.9% P < 0.05). Following 6 hr of continuous treatment, a reduction in ADAR1 by −42.1% (P < 0.01) was observed.

Figure 11.

Protein expression analysis of ADAR1 (A) and ADAR 2 (B) in DIV14 cortical neuron acutely treated with 100 μM glutamate for 1 h; for 1 h of Glu and 5 h of wash out; for 6 h of continuous Glutamate treatment. Data represent mean values and standard errors obtained from at least 3 independent evaluations. Bonferroni correction was used after 2-way ANOVA (*P <0,05; ^**P < 0,01; ^***P < 0,001).

With respect to ADAR2 (Fig. 11B), a downregulation was seen immediately after the acute treatment (−29.5%; P < 0 .001), whereas an increase was observed after wash-out (−66.4%; P < 0.001). Following 6 hr of continuous treatment, a profound ADAR2 protein expression reduction was observed (−69.3%; P < 0.001).

Discussion

Here we report that, in primary cortical neurons, chronic glutamate treatment reduces the R/G editing levels of GluA2, GluA3, and GluA4 AMPA receptor subunits, without affecting the GluA2 Q/R editing level. This decrease was accompanied by a parallel loss in ADAR1 and ADAR2 expression, through a signal cascade that is Ca²⁺- and calpain- dependent. However, the protein expression of GluA1, but not GluA2, was also downregulated.

One of the features of the R/G editing site is to modulate the kinetic properties of AMPA receptor channels in association with alternative splicing at the Flip/Flop cassette, thus determining the time course for desensitization and resensitization.^27,41 AMPA receptors containing an edited (G) subunit show a faster recovery rate from desensitization compared with an unedited (R) form. Therefore, it is likely that the reduced editing at the R/G site of the AMPA subunits slows the kinetics of AMPA receptor resensitization, thereby attenuating the response to the continuous glutamate stimulus. In turn, the reduced activation of post-synaptic receptors probably limits the calcium-induced activation of post-synaptic neurons, which is the key step in mediating excitatory cell death.^44,45

In contrast to Mahajan's findings,⁴⁰ the GluA2 Q/R editing level was not affected under our experimental conditions. The same occurred for the 5 5-HT_2CR sites and the BLCAP sites. These results strongly point to a selective effect on the R/G editing level of GluA2, A3, and A4 AMPA receptor subunits, despite the strong down-regulation of both ADAR enzymes.

Interestingly, the glutamate treatment induced a robust downregulation of the CYFIP2 K/E site.⁴⁶ CYFIP2 mRNA has been ubiquitously expressed with the K/E site edited to different extents among human tissues.⁴⁷ In neurons CYFIP2 and its homolog CYFIP1 are localized to the synapse and have been described to interact with FMRP with a possible role in axon guidance and synapse formation.⁴⁸ While it has been demonstrated that the K/E editing site is ADAR2 specific^47,49 and its editing level is regulated in murine development,⁵⁰ no data exist on the functional effect of the K/E editing substitution, and its role in the field of glutamate transmission remains unknown.

Our finding that the GluA2 Q/R site is still fully edited indicates the continuous formation of GluA2 (R) containing AMPA receptors which, being impermeable to Ca²⁺, may counteract the excitotoxic effects of glutamate. Our data suggest that an increased Ca²⁺ influx through unedited GluA2-containing AMPA channels is not involved in the cellular response to in vitro chronic glutamate treatment. In addition, glutamate treatment strongly reduced the GluA1 subunits without affecting the GluA2 subunits, thus indicating the formation of more AMPA channels impermeable to Ca²⁺ thus inhibiting the pathological effects of the treatment. The regulatory changes seen here might therefore represent a compensatory response set in motion by neuronal cells in order to attenuate, at least partially, glutamate excitotoxicity. This possibility is further reinforced by our previous observation that spinal cord injury produces a decrease in AMPA receptor R/G editing, perhaps by reducing post-synaptic excitatory responses to glutamate and limiting the progression of cell death.^25,42 Thus, both our in vitro and in vivo evidence converge to indicate that appropriate regulation in the editing levels is an important step in neuronal cell physiology and survival.

A recent work by Balik et al.⁵¹ reported that enhanced neuronal activity after treatment with the GABA-A channel blocker bicuculline (BIC) increased the level of R/G editing in CA1, but not CA3, hippocampal cells, suggesting a cell type / subfield specificity in neuronal response. These data are in apparent contrast with ours since we found a glutamate-induced down-regulation of R/G editing. This discrepancy might be due to the fact that we used glutamate to directly activate glutamate receptors in a sub-excitotoxic way, while Balik et al. enhanced neuronal activity by inhibiting gabaergic interneurons. The intracellular pathways activated by the 2 paradigms might be different together with the amount of neurotransmitter that activates glutamate receptors. We also used dissociated cortical neurons whereas Balik et al. used hippocampal slices, observing modifications only in the CA1 subfield. It is possible that different neuronal cells respond differently to an exogenous stimulus. However, both our and Balik's data show that the modulation of RNA editing is crucial for neuronal activity.

To further understand why glutamate treatment alters specific editing sites, we analyzed the expression pattern of recently found ADAR cofactors that might inhibit or activate ADAR activity.^11,12 We found that the splicing factor SFRS9, an inhibitor of ADAR activity, is up-regulated after glutamate treatment, and a trend toward an increase (although not statistically significant) could also be seen for the inhibitory factor RPS14. These data could indicate that these co-factors might also be involved in the processes leading to the downregulation of the editing activity. BIC treatment of hippocampal slices, besides inducing a partial upregulation of the R/G site editing and ADAR2 mRNA levels,⁵¹ also determined a strong down-regulation of SFRS9 mRNA,¹¹ thus confirming a possible link between SFRS9 and ADAR2 activity.

Given that we had found a profound reduction in both ADAR1 and ADAR2 as a consequence of chronic glutamate over-stimulation, we evaluated ADAR activity by measuring the levels of ADAR2 self-editing and its splicing isoforms. ADAR2 edits its own pre-mRNA by introducing an alternative proximal 3′ acceptor site. This new acceptor site adds 47 nt to the ADAR2 coding region,⁵² giving rise to a frameshift that codes for a truncated ADAR2 protein lacking deaminase activity.⁵³ Chronic glutamate stimulation reduces ADAR2 self-editing and, given that this editing position is a specific target of ADAR2, our results suggest a potential deactivation of ADAR2 enzymatic activity. If this is true, the selectivity of ADAR2 action might be due to a site-specific cellular response to the treatment which, through changes in R/G editing only, leads to a specific modulation of AMPA receptor functions. However, the decrease in ADAR2 self-editing is not followed by the expected increase in the full length protein which is, instead, profoundly down-regulated. In addition, chronic glutamate treatment induced an up-regulation of the splicing isoform of ADAR2 lacking a 30 nt cassette in the deaminase domain, which is characterized by a low editing activity.⁵²

Taken together our data indicate that glutamate treatment induced a downregulation in the expression of ADAR1 and ADAR2 proteins, together with a decrease in ADAR2 enzymatic activity. Interestingly, this reduction seems to be dependent on Ca²⁺ inflow since the treatment with extracellular Ca²⁺ chelator BAPTA/am reverted ADAR1, ADAR2⁴⁰ as well as GluA2 R/G and CYFIP2 K/E editing level reductions.

Given that ADAR activity might be regulated by phosphorylation,^2,13 we investigated the contribution of CamKII, one of the most important kinases of the glutamatergic system, by inhibiting its activation. We found that the reduction in R/G editing levels was not prevented by CamKII inhibition, suggesting that other modifications or other kinases, take place to regulate ADAR activity.

It has been reported that calpain might cleave ADAR2 after glutamate excitotoxicity treatment.⁴⁰ Our results confirm and extend these data indirectly, showing that the same mechanism takes place after chronic glutamate treatments and might be responsible for the changes in ADAR1 activity. Further analyses are needed to confirm ADAR1 cleavage by calpain. However Mahajan et al. showed a down-regulation of GluA2 Q/R editing⁴⁰ that was not observed in our experiments, either at the mRNA or pre-mRNA levels. Although in our acute treatments, we reproduced the exact same experimental conditions as Mahajan et al., we did not observe GluA2 Q/R editing reductions at any of the points analyzed. We found a decrease in R/G editing and ADAR1 and ADAR2 protein expression also after acute glutamate treatment.

Taken together, our findings show that chronic treatment with sub-excitotoxic doses of glutamate causes a decrease in the R/G editing site of all AMPA receptors, presumably in an attempt to dampen potential excitotoxic effects of the continuous stimulation. However, CYFIP2 K/E is downregulated in a similar pattern to the R/G site. The functional consequence of this modulation remains unknown due to the lack of information regarding the role of the K/E site in the physiology of the protein. The reported effects are selective for these specific editing sites and occur through a reduction in ADAR1 and ADAR2 protein levels in a Ca²⁺- and calpain-dependent fashion. This further points to the need to fine tune these mechanisms under both physiological and pathological conditions.

Material and Methods

Primary neuronal cell cultures and treatments

Our experiments complied with guidelines for the use of experimental animals issued by the European Community Council Directive 86/609/EEC and were approved by the Italian Ministry of Health (Project ID: 320/2010). Rat cortical cultures were prepared as previously described.^54,55 In brief, cerebral cortices from day 18 Sprague-Dawley rat embryos (E18, Charles River Laboratories Inc., Wilmington, MA, USA) were mechanically dissociated by trituration in cold HBSS (Invitrogen) containing 10 mM HEPES (pH 7.4). The suspension was allowed to settle for 5 min, and the top fraction was collected. The neurons were centrifuged for 5 min at 200 g and resuspended in serum-free Neurobasal medium (Invitrogen) supplemented with B27 (Invitrogen), 30 U/ml penicillin (Sigma-Aldrich), 30 μg/ml streptomycin (Sigma-Aldrich, St. Louis, MO, USA) and 0.5 mM Glutamax (Invitrogen). Neurons were plated at a density of 30,000 cells/cm² on poly-D-lysine (Sigma-Aldrich) coated Petri dishes.^56,57 Three days after plating, 50% of the medium was replaced with fresh medium. Subsequently, half of the medium was replaced once a week.

Cell death analysis

Glutamate concentration for the subsequent analyses was chosen after propidium iodide analysis (PI) of DIV14 rat cortical cultures incubated for 24 hr with different concentrations of glutamate (10, 50 and 100 μM). After treatment, neuronal cultures were incubated with 5 μM/ml of propidium iodide (Sigma-Aldrich) in HBSS for 15 min at room temperature to stain the nuclei of dead cells. The propidium iodide is incorporated only by dying or dead cells, emitting red light. NeuN protein staining, a neuronal specific marker, was used to detect neuronal cells. Apoptotic cells presented either a condensed or fragmented nucleus and were labeled, and necrotic cells appeared as characteristic red dots. Neuronal death (Fig. 1) was expressed by the percentage of NeuN-positive cells that presented propidium iodide staining.

The coverslips were fixed in 4% paraformaldehyde in phosphate-buffered saline (4% PFA-PBS, Invitrogen) for 15 min at room temperature and before mounting, DAPI nuclear staining was performed.

Treatment of neuronal cultures

Cortical neuron cells for western blot and RNA editing analysis were chronically treated with 10 μM Glutamate (Sigma-Aldrich) and harvested after 24 hr of treatment.^58,59 After treatment, cells were maintained in normal culture medium for an additional 3 d (washout). BAPTA/am intracellular calcium chelator (Sigma-Aldrich) 20 μM and Calpain inhibitor (Sigma-Aldrich) 10 μM was added 30 min prior to addition of Glutamate and maintained for the 24 h of chronic treatment.⁶⁰

Acute treatments were carried out in 3 different time points: 100 μM Glutamate for 1 hr; 100 μM Glutamate for 1 hr and cells harvested after 5 hr of washout; 100 μM Glutamate for 6 hr of continuous treatment. Each experiment was performed using 3 independent preparations of DIV14 neurons.

Western blot analysis

Cells harvested from primary cortical cultures were solubilized with modified RIPA (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% IGEPAL CA630, 0.25% NaDOC, 0.1% SDS, 1% NP-40 and Roche protease inhibitor tablets) and then sonicated. A portion of the lysate was used for the bicinchoninic acid (BCA) protein concentration assay (Sigma-Aldrich). Equal amounts of protein were applied to precast SDS polyacrylamide gels (4–12% NuPAGE Bis-Tris gels; Invitrogen) and the proteins were electrophoretically transferred to a nitrocellulose transfer membrane (GE Healthcare, Waukesha, WI, USA) for 2 h. The membranes were blocked for 60 min with 3% non-fat dry milk in TBS-T (Tris-buffered saline with 0.1% Tween-20, Sigma-Aldrich) and then incubated overnight at 4°C in the blocking solution with the mouse monoclonal anti-ADAR1 (1:350; Santa-Cruz Biotechnologies, Dallas, Texas, USA. Cod: sc-73408), rabbit polyclonal anti-ADAR2 (1:350; Abcam Cambridge, GB; cod: AB-64830), rabbit polyclonal GluA1 (1:200, Millipore, Billerica, MA, USA cod: AB1504), rabbit polyclonal GluA2 (1:2500, Millipore cod: AB1768–25UG) and mouse monoclonal anti-GAPDH (1:10000, Millipore Billerica, MA 01821; cod: MAB374) primary antibodies.

For detection, after 3 washes in TBS-T, the membranes were incubated for 1 h at room temperature with IR-Dye secondary antibodies. Signals were detected using an Odyssey infrared imaging system (LI-COR Biosciences) and quantified using Odyssey version 1.1 (LI-COR Biosciences).

Data are presented as the ratio of the intensity band of the investigated protein to that of the GAPDH band, and are expressed as a percentage of the controls. Each condition was carried out and analyzed in 3 independent primary culture dishes.

RNA extraction and retro-transcription reaction

Total RNA from cultured neurons was extracted using the ZymoResearch™ kit (Irvine, CA 92614, USA), according to the manufacturer's instructions. Reverse transcription was carried out using Moloney murine leukemia virus-reverse transcriptase (MMLV-RT) provided by Invitrogen. Briefly, 2 μg of total RNA were mixed with 2.2 μl of 0.2 ng/μl random hexamer (Invitrogen), 10 μl of 5× buffer (Invitrogen), 10 μl of 2 mM dNTPs, 1 μl of 1 mM DTT (Invitrogen), 0.4 μl of 33 U/μl RNasin (Promega, Madison, WI, USA) and 2 μl MMLV-RT (200 u/μl) in a final volume of 50 μl. The reaction mixture was incubated at 37°C for 2 h, and then the enzyme was inactivated at 95°C for 10 min.

RNA editing quantification

The editing level for the AMPA receptor (GluA2, GluA3, GluA4), 5-hydroxytryptamine receptor 2C (5-HT_2CR) transcripts, bladder cancer associated protein (BLCAP), CYFP2 and ADAR2 pre- mRNA was quantified by sequence analysis using the same approach.^42,61,62 Briefly, in the electropherogram obtained after RT-PCR and sequencing analysis of a pool of transcripts that might be edited or not, the nucleotide that undergoes the editing reaction appears as 2 overlapping peaks: A from unedited transcripts, and G from the edited ones. The editing level can be reliably calculated as a function of the ratio between the G peak area and A plus G peaks areas. The areas representing the amount of each nucleotide were quantified using Discovery Studio (DS) Gene 1.5 (Accelrys Inc., San Diego, CA, USA), and the means and standard errors (N > 3) for each experimental group were calculated and used for subsequent statistical analysis.

Quantitative real-time PCR

Quantitative real-time PCR (RT-qPCR) was performed to analyze the expression variation of several ADAR cofactors^11,12: Split Hand/Foot Malformation 1 (SHFM1), RNA binding protein hnRNP A2/B1, ribosomal protein 14 (RPS14) and Serine/Arginine-Rich Splicing Factor 9 (SRSF9).

The RNA expression pattern of the genes of interest was analyzed using Applied Biosystems 7500 real-time PCR system (Life Technologies, Foster City, CA, USA). PCR was carried out using TaqMan Universal PCR Master Mix (Life Technologies), which contained AmpliTaq Gold DNA Polymerase, AmpErase UNG, dNTPs with dUTP, passive reference and optimized buffer components. AmpErase UNG treatment was used to prevent the possible reamplification of carry-over PCR products. Thermal cycling was started by incubation at 50°C for 2 min and at 95°C for 10 min for optimal AmpErase UNG activity and activation of AmpliTaq Gold DNA polymerase. After this initial step, 40 cycles of PCR were performed. Each PCR cycle consisted of heating at 95°C for 15 s for melting and 60°C for 1 min for annealing and extension. Then, 20 ng of sample was used in each real-time PCR reaction in a final volume of 20 μl. The expression ratio of the target genes (SHFM1: Rn01455913_m1, Invitrogen; hnRNPA2/B1: Rn.PT.58.24173291 Integrated DNA Technologies; RPS14: Rn00821130_g1, Invitrogen; SRSF9: Rn.PT.58.18075228 Integrated DNA Technologies) in the treated cells, in comparison with the non treated cells, was calculated as described by Pfaffel et al,⁶³ using the geometric mean of 2 housekeeping genes (GAPDH: Rn99999916_m1; H2AFZ: Rn00821133_g1). Each individual determination was repeated in triplicate.

Statistical analysis

Each experiment was performed in at least 3 independent preparations of rat cortical cultures. Statistical analysis of the editing and splicing data as well as the protein levels was performed using 2-way ANOVA followed by Bonferroni's post-test. One-way ANOVA was used for the statistical analysis of protein inhibitor experiments and qPCR experiments. Bonferroni correction was used after ANOVA.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This work was supported through a grant from MIUR (PRIN 2012 A9T2S9_004)

Supplemental Material

Supplemental data for this article can be accessed on the publisher's website.