Abstract

Results

Extinction of conditioned fear results in histone acetylation around BDNF gene promoters within the prefrontal cortex

Behavioral training resulted in significant changes in H3 acetylation around the P1 promoter of the BDNF gene (one-way ANOVA, F_(2,8) = 58.70, P < 0.0001; Fig. 1A). Fear-conditioned mice without extinction showed a significant increase in H3 acetylation compared to naive controls (Bonferroni post hoc analysis; P < 0.01), suggesting an increase in histone acetylation due to fear conditioning. Extinction training reversed this H3 acetylation to below naive levels (P < 0.01 vs. fear conditioned without extinction; P < 0.05 vs. naive). There was no significant effect of learning on H4 acetylation around the P1 promoter of the BDNF gene (Fig. 1B).

Open in a separate window

Figure 1.

Learning-induced histone modifications ±SEM around BDNF gene promoters (A–C) P1 and (D–F) P4. (A) At the BDNF P1, histone H3 acetylation was increased in fear-conditioned animals compared to both other groups (**P < 0.01), and H3 acetylation was decreased in extinguished animals relative to naive controls (*P < 0.05). (B) There were no differences in histone H4 acetylation. (C) Representative gel for the BDNF P1 PCR product. (D) At BDNF P4, histone H3 acetylation was increased in the fear-conditioned without extinction mice compared to naive controls (*P < 0.05). (E) Histone H4 acetylation was significantly increased in extinguished animals compared to both other groups (*P < 0.05), which did not differ from each other. (F) Representative gel for the BDNF P4 PCR product. (G,H) There were no significant learning-induced histone modifications around the promoter for β-tubulin. (I) Representative gel for the β-tubulin PCR product. (n = 4 samples pooled from two animals each.) (N) Naive; (FC-No EXT) fear conditioned without extinction; (EXT) extinction.

Behavioral training led to an increase in acetylated H3 around the P4 promoter of the BDNF gene (one-way ANOVA, F_(2,8) = 5.92, P < 0.05; Fig. 1D). Fear-conditioned without extinction mice showed a significant increase of H3 acetylation at this promoter compared to naive animals (P < 0.05), while increases in extinguished animals were not quite significant. There was no difference in H3 acetylation between fear-conditioned mice without extinction and extinguished animals (P > 0.05). Training also affected acetylated H4 at this promoter (F_(2,8) = 15.13, P < 0.01; Fig. 1E). Specifically, although fear conditioning without extinction caused no change in histone acetylation compared to naive animals (P > 0.05), extinction training increased H4 acetylation around the P4 promoter of the BDNF gene relative to both other groups (P < 0.01). There was no significant learning-induced histone modification around the control promoter for β-tubulin (Fig. 1G–I).

Extinction of conditioned fear results in increased BDNF mRNA expression within the prefrontal cortex

Behavioral training resulted in a significant increase in BDNF exon I mRNA expression (one-way ANOVA, F_(2,9) = 31.32, P < 0.0001; Fig. 2B). Fear-conditioned mice without extinction showed no increase in BDNF exon I mRNA compared to naive controls; however, extinction training increased BDNF exon I mRNA expression relative to both naive and fear-conditioned without extinction mice (Bonferroni post hoc tests, P < 0.01) (Fig. 2B).

Open in a separate window

Figure 2.

Effect of learning on (A,B) BDNF exon I and (C,D) IV mRNA expression in the prefrontal cortex. (A) Representative quantitative RT-PCR plot for the effect of strong (57 CSs) extinction training on BDNF exon I mRNA expression in the prefrontal cortex. (B) Strong extinction training yielded a significant increase in BDNF exon I mRNA expression. (C) Representative quantitative RT-PCR plot for the effect of strong (57 CSs) extinction training on BDNF exon IV mRNA expression in the prefrontal cortex. (D) Strong extinction training yielded a significant increase in BDNF exon IV mRNA expression. (**P < 0.01, n = 4 samples pooled from two animals each.) (N) Naive; (FC-No EXT) fear conditioned without extinction; (EXT) partial extinction.

Behavioral training resulted in a significant increase in BDNF exon IV mRNA expression (one-way ANOVA, F_(2,9) = 22.85, P < 0.001; Fig. 2D). Fear-conditioned mice without extinction showed no increase in BDNF exon IV mRNA compared to naive controls; however, extinction training increased BDNF exon IV mRNA expression relative to both naive and fear-conditioned without extinction mice (Bonferroni post hoc tests, P < 0.01) (Fig. 2D).

Histone deactylase inhibitors, given prior to training, enhance long-term memory for the extinction of conditioned fear

Valproic acid (VPA) has been used for decades as an anticonvulsant and a mood stabilizer (Winterer and Hermann 2000); however, it is also a potent HDAC inhibitor (Phiel et al. 2001) and increases BDNF mRNA and protein levels both in cell culture (Castro et al. 2005) and in the prefrontal cortex (Fukumoto et al. 2001). When administered chronically, VPA reduces fear responding after nonreinforced extinction-like exposures to an aversive context (Li et al. 2006), and a derivative of VPA, pentyl-4-yn-valproic acid, enhances both spatial learning and long-term memory for avoidance conditioning (Murphy et al. 2001; Foley et al. 2004). We therefore injected VPA before a partial extinction protocol as previously described (Cain et al. 2003; Ponnusamy et al. 2005). By itself, this protocol yields no significant long-term memory for extinction.

VPA had no effect on long-term memory (LTM) for fear of the CS (measured by behavioral freezing) when injected on day 2 in the absence of CS presentations (see FC-No EXT groups, Fig. 3A) nor on the rate of within-session extinction across 20 CS exposures (data not shown), but enhanced long-term memory for extinction in a dose-dependent, U-shaped manner (n = 16/group, F_(5,94) = 6.91, P < 0.0001; Bonferroni post hoc analysis, P < 0.05 vehicle vs. 100 mg/kg). The U-shape may have resulted from the sedating effects of the highest dose preventing learning during extinction training. There was a main effect of VPA treatment on pre-CS freezing after partial extinction training (F_(5,94) = 2.92, P < 0.05; Fig. 3A, left); however, when pairwise comparisons were made, Bonferroni post hoc analysis revealed no significant difference in freezing between vehicle-treated and VPA-treated mice In contrast, valpromide, an analog of VPA that shares its anticonvulsant and mood-stabilizing effects but is not an HDAC inhibitor (see chemical structures in Fig. 4), had no effect on LTM for extinction (n = 8–16/group, F_(5,66) = 0.9078, P = 0.48; Fig. 3B). Conversely, sodium butyrate (NaBt), a prototypical HDAC inhibitor with no known anticonvulsant activity, also increased LTM for fear extinction (n = 15–16/group, F_(5,95) = 5.80, P < 0.0001, P < 0.05, vehicle vs. 1000 mg/kg) (Fig. 3C). There was no effect of sodium butyrate treatment on pre-CS freezing scores (Fig. 3C, left).

Open in a separate window

Figure 3.

Mean ± SEM percent freezing 7 d after histone deacetylase (HDAC) inhibitor injection before partial extinction training. None of the drugs had any significant effect on pre-CS freezing compared to mice treated with vehicle and partially extinguished. (A) At a dose of 100 mg/kg, valproic acid potentiates long-term memory for the extinction of conditioned fear. (B) Valpromide has no effect on extinction of conditioned fear. (C) At a dose of 1000 mg/kg, sodium butyrate potentiates long-term memory for the extinction of conditioned fear. (*P < 0.05 compared to mice extinguished with vehicle; n = 8–16 mice/group.) (FC-No EXT) Fear conditioned without extinction.

Open in a separate window

Figure 4.

Chemical structures of compounds used to facilitate extinction.

VPA, combined with partial extinction training, increased histone H4 acetylation around both the BDNF P1 and P4 gene promoters but increased only BDNF exon IV mRNA expression in prefrontal cortex

To provide evidence for a more direct relationship between changes in histone acetylation in the prefrontal cortex and extinction learning, we tested the effect of VPA (100 mg/kg) combined with partial extinction training on histone modifications around the BDNF P1 and P4 promoters as well as on BDNF exon I and IV transcripts using the same extinction protocol used above to analyze the effect of HDAC inhibitors on behavior.

There was no significant effect of drug treatment or training on H3 acetylation at the BDNF P1 promoter (n = 4 samples pooled from two animals each, two-way ANOVA, F_(1,12) = 2.71, ns; Fig. 5A). On the other hand, VPA treatment caused a significant increase in H4 acetylation overall at this promoter (two-way ANOVA, F_(1,12) = 3.48, P < 0.05 for treatment) with a trend toward interaction between training and drug treatment (P = 0.076) (Fig. 5B). Partial extinction training alone yielded no significant increase in H4 acetylation, consistent with the absence of an effect on freezing. However, this VPA-induced increase in H4 acetylation was unrelated to BDNF exon I mRNA expression (Fig. 5D). While there was a significant main effect of partial extinction training on BDNF exon I mRNA expression (n = 6 samples, pooled from two animals each; F_(1,20) = 6.20, P < 0.05), there was neither interaction nor main effect of VPA on the expression of BDNF exon I.

Open in a separate window

Figure 5.

Effect of histone deacetylase inhibition by valproic acid (VPA) on learning-induced histone modifications (±SEM) around the P1 promoter of the BDNF gene and on BDNF exon I mRNA expression in the prefrontal cortex. VPA (100 mg/kg) combined with partial extinction training (20 CS presentations) (A) had no significant effect on H3 acetylation around the P1 promoter, (B) but increased histone H4 acetylation relative to all other treatment groups (*P < 0.05, n = 4 samples pooled from two animals each). (Inset) Raw data. (C) Representative quantitative RT-PCR plot for the effect of VPA and partial extinction training on BDNF exon I mRNA expression in the prefrontal cortex. (D) Extinction training yielded a significant increase in BDNF exon I mRNA expression. (*P < 0.05, n = 6 samples pooled from two animals each.) (FC-No EXT) Fear conditioned without extinction; (EXT) partial extinction.

Neither partial extinction training nor VPA affected H3 acetylation at the P4 promoter (n = 4 samples pooled from two animals each, 2-way ANOVA, F_(1,12) = 1.563, ns; Fig. 6A). However, there was a significant interaction between training and drug treatment for H4 acetylation there (two-way ANOVA, F_(1,12) = 10.66, P < 0.02; Fig. 6B). Partial extinction training yielded no significant increase in H4 acetylation, in parallel with its lack of effect on freezing. VPA alone also had no effect on H4 acetylation. However, VPA significantly increased H4 acetylation in combination with partial extinction training (P < 0.01), again paralleling its behavioral effect in facilitating extinction (Fig. 6A). This increase in H4 acetylation by the combination of VPA and partial extinction training also mimicked the effect of strong extinction training alone. Unlike the result for BDNF exon I mRNA, there was a significant interaction between training and drug treatment for exon IV mRNA (two-way ANOVA, F_(1,20) = 11.97, P < 0.01, n = 6 samples, pooled from two animals each; Fig. 6D). Bonferroni post hoc tests indicate that partial extinction training only increased exon IV mRNA when combined with VPA (P < 0.01). Thus, even though VPA may inhibit HDACs nonspecifically, its effects on histone acetylation at the P4 BDNF promoter and on gene expression of BDNF exon IV in the prefrontal cortex are quite specific to the pairing of VPA with behavioral experience.

Open in a separate window

Figure 6.

Effect of histone deacetylase inhibition by valproic acid (VPA) on learning-induced histone modifications (±SEM) around the P4 promoter of the BDNF gene and on BDNF exon IV mRNA expression in the prefrontal cortex. VPA (100 mg/kg) combined with partial extinction training (20 CS presentations) (A) had no effect on H3 acetylation around the P4 promoter of the BDNF gene, (B) but increased histone H4 acetylation relative to all other treatment groups (**P < 0.01, n = 4 samples pooled from two animals each). (Inset) Raw data. (C) Representative quantitative RT-PCR plot for the effect of VPA and partial extinction training on BDNF exon IV mRNA expression in the prefrontal cortex. (D) VPA (100 mg/kg) yielded no significant (ns) increase in exon IV mRNA expression in retention controls, but a significant increase with extinction. (**P < 0.01, n = 6 samples pooled from two animals each.) (FC-No EXT) Fear conditioned without extinction; (EXT) partial extinction.

Discussion

This study generated four main findings: (1) Acquisition and extinction of conditioned fear results in distinct histone modifications around two BDNF gene promoters in the prefrontal cortex. (2) Extinction of conditioned fear leads to an increase in BDNF mRNA expression in the prefrontal cortex. (3) Histone deacetylase inhibitors, including VPA, given prior to partial extinction training, enhance long-term memory for the extinction of conditioned fear. (4) VPA, given prior to partial extinction training, increases histone H4 acetylation at both the BNDF P1 and P4 gene promoters, reproducing the acetylation pattern of strong extinction on P4, but the combination only increases BDNF exon IV-containing mRNA expression in the prefrontal cortex.

The increase in H4 acetylation around the P4 promoter after extinction (Fig. 1E) is consistent with a previous report that H4 acetylation is generally increased in the hippocampus after latent inhibition of conditioned fear, another inhibitory learning protocol (Levenson et al. 2004). However, the change we see in H4 acetylation is promoter-specific, even in prefrontal cortex. H4 acetylation is significantly increased around the P4 promoter, but only shows a trend to increase around the P1 promoter (Fig. 1B) after extinction training. Also, we have seen no such H4 acetylation increases after extinction in either the hippocampus or amygdala at these BDNF promoters (data not shown). Conversely, while H3 acetylation is increased around both promoters in the prefrontal cortex by fear conditioning without extinction (Fig. 1A,D), extinction training has opposite effects on H3 acetylation at the two promoters. Thus, our data support the hypothesis that there potentially exists a histone code for different forms of learning and memory (Chwang et al. 2006; Wood et al. 2006), since we show specific histone modifications around individual promoters of the BDNF gene in response to specific learning protocols.

While the contrast in histone acetylation between fear-conditioned without extinction and naive animals is striking for H3, it is important to be cautious about the interpretation of these results, since naive animals were exposed to neither the conditioning context, the CS, the US, nor the second context used for extinction. Thus, any of these factors may be influencing the changes in histone acetylation we see in fear conditioning without extinction compared to naive animals in a nonspecific manner regardless of fear conditioning. However, the contrast in histone acetylation between fear-conditioned without extinction and extinguished mice (or, for that matter, the lack of contrast for H3 at P4 and for H4 at P1) cannot depend on such nonspecific factors, since both groups received exactly the same training in context A on the first day, and then spent the same amount of time in context B on the second day. The only difference between these groups is the CS presentations on day 2. So while it remains possible that repeated CS exposures alone generate changes in histone acetylation (regardless of previous training), we believe that these changes are most likely to be due to extinction learning. This is supported by the correlation of such changes with extinction learning and not with CS exposures in the extinction paradigm used in our later experiments (Figs. 4, ​,5,5, ​,6).6). No change in histone acetylation was observed in groups that received only extinction training and vehicle, despite the presentation of a substantial number of CSs to those animals. In fact, histone H4 acetylation only increased around the P1 and P4 promoters under conditions that yielded effective extinction learning after the injection of VPA in combination with partial extinction training (Figs. 4, ​,5,5, ​,66).

Interestingly, when given before extinction training, VPA induced reductions in pre-CS freezing similar to those in freezing to the CS (Fig. 4A). Although not significant, these data suggest the possibility that VPA may influence extinction through contextual rather than cue-based learning. Nevertheless, there is no effect on pre-CS freezing in the fear-conditioned without extinction groups between those treated with vehicle and those treated with VPA, indicating that context exposure alone is not responsible for this reduction of pre-CS freezing. Furthermore, our experiments testing the combination of VPA with partial extinction training indicate that cue exposure is crucial for any VPA effect on H4 acetylation.

The findings that fear conditioning with and without extinction result in distinct patterns of histone acetylation support, and extend to specific promoters of a single gene, the findings of Levenson et al. (2004), who show an increase in H3 acetylation in hippocampus after contextual fear conditioning and in H4 acetylation after latent inhibition, a form of inhibitory learning that shares common cellular substrates with extinction (Barad et al. 2005). However, our data suggest a more complicated story. Acetylated H3 appears to be increased by fear conditioning without extinction at both promoters, but it remains elevated around P4 after extinction, while extinction strongly depresses the amount of acetylated H3 associated with P1. H4 is little affected by fear conditioning without extinction at either promoter, and it remains largely unchanged by extinction training at P1. However, acetylated H4 associated with the P4 promoter is significantly increased by extinction training. Thus, in our experiments, different behavioral training protocols do not appear to affect any histone in general, but rather acetylation of specific combinations of histones associated with specific promoters. These data establish a rationale for future studies to examine specific components of the associative learning process (i.e., the effects of CS alone, or US alone, the contrast between unpaired and paired training, etc.) and to examine training effects on histone modification around the promoters of other candidate learning genes. Such studies will ultimately require increased specificity in the molecular analysis by examining specific populations of brain cells (i.e., glutamatergic or GABAergic neurons, or glia) before and after opposing forms of learning like the acquisition and extinction of conditioned fear.

Our results indicate that the HDAC inhibitors VPA and sodium butyrate facilitate long-term memory for extinction in a dose-dependent fashion, because of their HDAC-inhibiting activity (Fig. 4A,C). First, we see similar behavioral effects for two well-characterized HDAC inhibitors. Second, as a negative control we used valpromide, a compound that shares most of the structure and therapeutic effects of VPA, but is not an HDAC inhibitor. Valpromide had no effect on extinction learning, indicating that VPA potentiates extinction because of its HDAC-inhibitor, but not its anticonvulsant or mood-stabilizing effects (Fig. 4B). Indeed, previous work has demonstrated a role for HDAC inhibitors in enhancing long-term memory for conditioned fear with both systemic (Levenson et al. 2004) and local infusion (Yeh et al. 2004), as well as increasing long-term memory for spatial learning (Wood et al. 2005). An HDAC inhibitor can also reverse some of the effects of early life stress on fear-related behavior (Weaver et al. 2004).

To add to the evidence that VPA may exert its effects on extinction learning through changes in histone acetylation, we looked for further correlative evidence of the interaction between VPA and extinction training at the level of histone acetylation associated with the two BDNF promoters we targeted in this study. Interestingly, while the combination of VPA and extinction caused a synergistic increase of H4 acetylation at both the P1 and P4 promoters, exon I-containing BDNF transcripts were increased as much by extinction alone as by the combination of VPA and extinction. Exon IV-containing transcripts, however, were only increased by the combination of VPA and partial extinction training. Thus, BDNF exon IV mRNA expression is preferentially sensitive to HDAC inhibition and appears to be associated with learning the tone–no shock association (as occurs during extinction). BDNF exon I mRNA expression appears to be more sensitive to CS-alone presentations, and insensitive to the extinction association. This idea is consistent with the observation that higher numbers of CS presentations in the overtraining protocol used to measure the direct effect of behavioral extinction training on histone acetylation and BDNF mRNA expression have a more dramatic effect on BDNF exon I transcripts (Fig. 2B) than fewer CS presentations in the partial extinction protocol (even when combined with VPA treatment; Fig. 5D). In both the overtraining and the partial extinction protocols, the expression of BDNF exon I mRNA is uncorrelated with acetylation patterns of H3 and H4 at the P1 promoter, suggesting that learning-induced changes in acetylation of histones at the P1 promoter are not mechanistically connected to expression of the BDNF exon I mRNA. In contrast, the increase in BDNF exon IV mRNA expression with extinction is not only correlated with extinction learning, but is similar is magnitude between extinction overtraining (Fig. 2D) and the combination of partial extinction and valproate treatment (Fig. 6D). The contrast in the correlation between extinction learning and the expression of BDNF transcripts containing the two different exons suggests three possibilities: (1) the BDNF transcripts containing different exons may have distinct functions within a single cell, (2) the differing BDNF transcripts may have the same molecular function but act in different compartments of the same cell, or (3) the differing BDNF transcripts have the same molecular function but act in different populations of cells. Our data do not allow us to distinguish between these possibilities, but we favor the third, as it would be consistent with a model in which different aspects of learning are subserved by different cell populations.

The effect of HDAC inhibition on fear extinction and on H4 acetylation around P4 appears to be contingent on behavioral training at the time of drug administration, since when HDAC inhibitors are injected into fear-conditioned animals that get no extinction training, there is no effect either on fear responding to the CS at test (Fig. 4), or on H4 acetylation (Figs. 5, ​,6).6). Thus even though HDAC inhibition by these compounds is likely to be both nonspecific and widespread, they can still have specific effects guided by behavioral training. In fact, while VPA blocks histone deacetylation, we see no effect of VPA on H3 acetylation at the P4 promoter (Fig. 6A), and even see decreases in H3 acetylation around the P1 promoter (Fig. 5A).

There are at least 11 known HDACs, each with some degree of regional specificity in the mammalian brain (Sugino et al. 2006). It would of great interest to determine whether specific HDACS show preference for gene promoters associated with H3 or H4, and/or whether there is also neuronal (or glial) subtype specificity further regulating their action. We hypothesize that regional and cell-type specificities of HDAC expression may play a critical role in translating different forms of learning into sustained changes of gene expression. In fact, experiments using viral vectors indicate that HDAC5 activity may play a role in conditioned defeat (Tsankova et al. 2006), and that some class IIa calcium-regulated HDACs facilitate the development of cocaine addiction (Kumar et al. 2005).

In conclusion, our data suggest a relationship between histone H4 modifications at BDNF promoters and long-term memory for extinction of conditioned fear. Histone modifications have been associated with models of neuropsychiatric disorders including schizophrenia, depression, and addiction (Kumar et al. 2005; Sharma 2005; Tsankova et al. 2006), supporting a potential role for epigenetic mechanisms in the regulation of a wide variety of behaviors. Our studies provide new information about the potential role of epigenetic regulation of gene expression in a preclinical model of anxiety disorders and their treatment. The facilitating effects of VPA on extinction are particularly intriguing, since the drug is already in use in psychiatric treatment and can be used immediately as an adjunct to behavior therapy for human anxiety disorders.

Materials and Methods

Acknowledgments

Footnotes

References