|Home | About | Journals | Submit | Contact Us | Français|
It has been established that regulation of chromatin structure through post-translational modification of histone proteins, primarily histone H3 phosphorylation and acetylation, is an important early step in the induction of synaptic plasticity and formation of long-term memory. In this study, we investigated the contribution of another histone modification, histone methylation, to memory formation in the adult hippocampus. We found that tri-methylation of histone H3 at lysine 4 (H3K4), an active mark for transcription, is upregulated in hippocampus one hour following contextual fear conditioning. In addition, we found that di-methylation of histone H3 at lysine 9 (H3K9), a molecular mark associated with transcriptional silencing, is increased one hour after fear conditioning and decreased twenty-four hours after context exposure alone and contextual fear conditioning. Tri-methylated H3K4 levels returned to baseline levels at twenty-four hours. We also found that mice deficient in the H3K4-specific histone methyltransferase, Mll, displayed deficits in contextual fear conditioning relative to wildtype animals. This suggests that histone methylation is required for proper long-term consolidation of contextual fear memories. Interestingly, inhibition of histone deacetylases (HDACs) with sodium butyrate (NaB) resulted in increased H3K4 tri-methylation and decreased H3K9 di-methylation in hippocampus following contextual fear conditioning. Correspondingly, we found that fear learning triggered increases in H3K4 tri-methylation at specific gene promoter regions (Zif268 and bdnf) with altered DNA methylation and MeCP2 DNA binding. Zif268 DNA methylation levels returned to baseline at twenty-four hours. Together, these data demonstrate that histone methylation is actively regulated in the hippocampus and facilitates long-term memory formation.
Experience-dependent transcriptional regulation is critical for the stabilization of long-term memory formation. While the role of transcription factors in the regulation of gene expression in memory have been the focus of intense research, an increasing amount of recent data have identified a role for chromatin remodeling in this process. Specifically, post-translational modifications of histones has been shown to be a mechanism for transcriptional regulation during long-term memory formation (Swank and Sweatt, 2001; Guan et al., 2002; Alarcon et al., 2004; Levenson et al., 2004a; 2005; Kumar et al., 2005; Levenson and Sweatt, 2005; Chwang et al., 2006; Wood et al., 2006; Bredy et al., 2007; Fischer et al., 2007; Lubin et al., 2008; Guan et al., 2009).
The regulation of chromatin structure is complex. This is partly due to the N-terminal tails of histones which are highly accessible for enzymatic transformation and are sites for multiple covalent post-translational modifications, including acetylation, phosphorylation, methylation, ubiquitination, and sumoylation (Peterson and Laniel, 2004; Berger, 2007). Histone modifications such as histone acetylation and phosphorylation facilitate the unraveling of DNA around the histone core resulting in the recruitment of the transcriptional machinery to mediate cell- and promoter-specific gene expression. However, other histone modifications such as histone methylation have different positive or negative effects on gene transcription depending on the amino acid residue modified and the number of methyl groups added (Berger, 2007).
The methylation of lysines can exist in three states: mono-, di-, and tri-methylated. Generally, transcriptionally silent regions contain di- and tri-methylation of histone H3 at lysine9 (H3K9) whereas active genes correlate with di- and tri-methylation of histone H3 at lysine4 (H3K4) (Sims et al., 2003; Margueron et al., 2005; Martin and Zhang, 2005; Vermeulen et al., 2007). Specifically, di-methylation of H3K9 correlates with transcriptional silencing, and tri-methylation of H3K4 is linked to active transcription (Santos-Rosa et al., 2002; Ng et al., 2003; Schneider et al., 2004). Recently histone methylation has been implicated in the regulation of chromatin structure in the nervous system (Tsankova et al., 2006; Huang and Akbarian, 2007; Huang et al., 2007; Kim et al., 2007). However, to date there have been no reports of histone methylation regulation in the nervous system functioning in the process of long-term memory storage.
Thus in the present study, we sought to investigate histone H3 methylation as an experience-driven mechanism of transcriptional regulation in hippocampus during consolidation of fear conditioned memories. We found that contextual fear conditioning induced both tri-methylation of H3K4 and di-methylation of H3K9 in hippocampus, suggesting active increases in gene expression and gene repression during memory formation. Interestingly, treatment of animals with the histone deacetylase (HDAC) inhibitor sodium butyrate (NaB) elevated tri-methylation of H3K4 and decreased di-methylation of H3K9. This suggests that histone acetylation is intricately linked to histone methylation in hippocampus during consolidation of fear conditioned memories. We also found that heterozygous deletion of a known regulator of histone methylation, Mll, led to significant deficits in memory consolidation. Finally, we found that fear conditioning triggered increases in tri-methylation of H3K4 at the Zif268 gene which were associated with altered DNA methylation at this gene promoter. Taken together these observations implicate experience-driven alteration of histone methylation in hippocampus as a mechanism contributing to long-term memory formation.
Young adult, male Sprague-Dawley rats (250–300 g) were used for all experiments with the exception of the Mll;eed mutant mice which were created as previously described by Kim et al. (2007). Briefly, the eed null allele employed in this study contained a N-ethyl-N-nitrosourea-induced L290P substitution in the second WD motif (Schumacher et al., 1996) and was maintained on a mixed genetic background largely derived from 101/R1 and C3Hf/R1. The mutant Mll allele employed in this study harbors a lacZ insertion in the third exon, resulting in a truncated protein (Yu et al., 1995). The mutant Mll allele was maintained on a mixed background with contributions from C57BL/6J, C3H/HeJ, and FVB/NJ. Experiments encompassed single and double heterozygous animals derived from intercrosses of eed and Mll single mutant mice. Homozygosity for the rd mutation was excluded by PCR. Animals were housed under light/dark 12 h/12 h and allowed access to rodent chow and water ad libitum. Five days prior to experiments, animals were handled for 3–5 minutes each and allowed to acclimate to laboratory condition. All procedures were performed with the approval of the University of Alabama at Birmingham Institutional Animal Care and Use Committee and according to national guidelines and policies.
Rats were transported to the laboratory at least 30 min prior to fear conditioning. Fear conditioned animals were allowed to explore the chamber for 2 min, after which they received a series of three electric shocks (1 s, 0.5 mA) at 2 min intervals. As controls, context exposed or Latent Inhibition plus fear conditioned animals were also placed in the novel training chambers. Context exposed animals were placed in the novel training chamber for 7 min without receiving the footshock. Latent Inhibition plus fear conditioned animals were pre-exposed to the context for 2 h before the same 7 min training protocol was administered as described for the fear conditioned animals. In all shocked groups the animals were allowed to explore the novel context (training chamber) for an additional 1 min after the receiving the final footshock prior to being returned to their home cage. Freezing behavior was recoded using Video Freeze software (Med Associates, St. Albans, VT). Another group of age-matched animals that were handled by the experimenter but did not receive any experimental manipulations were used as naïve controls in all experiments.
Exploratory activity, motor function, and sensorimotor processing were performed as described (Paylor et al., 2006). Behavioral experiments of eed x Mll investigation employed 91 female mice at 8–12 weeks of age (n=29 eed+/+;Mll+/+, 26 eed+/−;Mll+/+, 18 eed+/+;Mll+/−, and 18 eed+/−;Mll+/−). In the contextual fear paradigm (McIlwain et al., 2001), mice were placed in a chamber and after 2 min the mice received two conditioned stimulus (CS; 80 dB auditory sound, 30 sec) – unconditioned stimulus (US; 0.7 mA mild foot shock, 2sec) pairings separated by 2 min. Twenty-four hours later, the mice were assayed for freezing behavior to determine associative learning of the footshock with the contextual cues, which constitutes the training environment, and the discretely paired auditory sound. On Test day, the amount of freezing during the 3 min pre-CS period was subtracted from the freezing during the 3 min CS presentation.
Animals were sacrificed using a rodent guillotine. The brain was immediately immersed in oxygenated (95%/5% O2/CO2) ice-cold cutting saline (in mM: 110 sucrose, 60 NaCl, 3 KCl, 1.25 NaH2PO4, 28 NaHCO3, 0.5 CaCl2, 7 MgCl2, 5 glucose, 0.6 ascorbate). The whole hippocampus was then removed and area CA1 was dissected away from other hippocampal subfields under a dissecting microscope. Once isolated, area CA1 was frozen on dry ice then stored at −80C until processed.
Histone extractions were performed as in Lubin et al. (2007). Briefly, all procedures were performed on ice, and all solutions were chilled to 4 °C prior to use unless otherwise indicated. All centrifugation steps were performed at 4 °C. Hippocampal tissue was dounce homogenized in ice-cold homoginization buffer (in mM: 250 sucrose, 50 Tris, pH 7.5, 25 KCl, 0.5 phenylmethylsulfonyl fluoride, 1% protease inhibitor mixture (Sigma), 0.9 Na+-butyrate) using no more than 6 strokes of a glass pestle (Kontes Glass Co, NJ). Hippocampal tissue homogenates were centrifuged at 7,700 × g for 1 min and the supernatant (cytoplasmic fraction) was aspirated. The pellet (nuclear fraction) was then resuspended in 0.5 ml of 0.4 N H2SO4, incubated for 30 min and centrifuged at 14,000 × g for 10 min. The supernatant was transferred to a fresh tube, and proteins were precipitated with the addition of 250 µl of 100% trichloroacetic acid containing 4 mg/ml deoxycholic acid (Na+ salt, Sigma) for 30 min. Acid extracted histone proteins were then collected by centrifugation at 14,000 × g for 30 min. The supernatant was discarded, and the protein pellet was washed with 1 ml of acidified acetone (0.1% HCl) followed by 1 ml of acetone for 5 min each. Protein precipitates were collected between washes by centrifugation (14,000 × g, 5 min). The resulting purified histone proteins were resuspended in 10 mM Tris (pH 8) and stored at −80 °C until processed for western blotting.
Protein concentrations for each sample were determined using a DC protein assay (BioRad). Aliquots of sample were then normalized to ~2 µg/µl. Laemmli sample buffer [final concentration: 6.25 mM Tris, pH 6.8, 2% sodium dodecyl sulfate (SDS), 10% glycerol, 1.25% 2-mercaptoethanol, 0.1% bromophenol blue] was added to histone samples prior to performing SDS- polyacrylamide gel electrophoresis (PAGE). SDS-PAGE gels consisted of a 12% acrylamide resolving gel overlaid with a 4% acrylamide stacking gel. Histone proteins were transferred to polyvinylidene difluoride (PVDF) membranes for immunoblotting which consisted of incubating PVDF membranes in primary antibodies for 1 h at room temperature or overnight at 4 °C followed by incubation with a horseradish peroxidase-conjugated secondary antibody for 2 h at room temperature. Detection of immunolabeled histone proteins was facilitated via electrogenerated chemiluminescence (ECL; Amersham Biosciences; SuperSignal, Pierce) and exposure to BioMax X-Ray film (Kodak Scientific Imaging Systems, Rochester, NY) and digitized (Epson Perfection 1240U). Antibodies—The primary antibodies used, and their dilutions were as follows: anti-H3K4me3 (1:1000, Upstate Biotechnology Inc.), anti-H3K9me2 (1:1000, Upstate Biotechnology Inc.), and anti-H3 (1:1000, Upstate Biotechnology Inc.). The host for all primary antibodies was rabbit. The secondary antibodies were horseradish peroxidase-conjugated goat Anti-IgG heavy and light chain (Jackson ImmunoResearch, West Grove, PA).
ChIP analysis was performed as previously described in Lubin et al. (2007). Briefly, the CA1 regions of hippocampus were microdissected and held in ice-cold PBS solution containing protease inhibitors [1 mM phenylmethysulfonyl fluoride, 1 µg/ml of protease inhibitor cocktail and phosphatase inhibitors (1 mM Na3VO4 and 20 mM NaF)]. Hippocampal tissue was immediately incubated in 1% formaldehyde in PBS at 37 °C for 10 min. Hippocampal tissue was homogenized in SDS lysis buffer (50mM Tris, pH 8.1, 10mM EDTA, 1% SDS). Chromatin was sheared using a Branson Sonifier 250 at 1.5 power and constant duty cycle. Lysates were centrifuged to pellet debris and then diluted 1:10 in ChIP dilution buffer (16.7 mM Tris, pH 8.1, 0.01% SDS, 1.1% Triton X-100, 167 mM NaCl, 1.2 mM EDTA). Immunoprecipitations were carried out at 4 °C overnight with primary antibodies (anti-H3K4me3 or MeCP2) or No antibody (control). Immune complexes were collected with an protein A-agarose bead/salmon sperm slurry and sequentially washed with low salt buffer (20 mM Tris, pH 8.0, 0.1% SDS, 1% Triton X-100, 2 mM EDTA, 150 mM NaCl), high salt buffer (20 mM Tris, pH 8.1, 0.1% SDS, 1% Triton X-100, 500 mM NaCl, 1 mM EDTA), LiCl immune complex buffer (0.25 M LiCl, 10 mM Tris, pH 8.1, 1% deoxycholic acid, 1% IGEPAL-CA630, 500mM NaCl, 2 mM EDTA), and TE buffer. Immune complexes were extracted in 1× TE containing 1% SDS, and protein-DNA cross-links were reverted by heating at 65 °C overnight. After proteinase K digestion (100 µg; 2 h at 37°C), DNA was extracted by phenol/chloroform/isoamyl alcohol and then ethanol- precipitated. Immunoprecipitated DNA was subjected to quantitative real-time PCR using primers using primers specific for 150–200 bp segments corresponding to promoters upstream of the rat Bdnf or Zif268: Bdnf promoter 1 sense, 5’-TGAGAAGTTGTGGGACCGCATA-3’; antisense, 5’- CGTCCTCTTCCCTTGGCTTTTT-3’; Bdnf promoter 4 sense, 5’- GTGGAGGAGAGGTGCCTTTTGA-3’; antisense, 5’- GTCCTCTGGGGACCATTTACCC-3’; Zif268 (ChIP 1) sense, 5’-ATGGGCTGTTAGGGACAGTG −3’; antisense, 5’-TTGGGGATTTAGCTCAGTGG-3’; Zif268 (ChIP 2) sense, 5’-CTCCCTCACTGCGTCTAAGG-3’; antisense, 5’- CCTAAAGAGGGGGACTTTGC-3’. The cumulative fluorescence for each amplicon was normalized to input amplification.
DNA was isolated from hippocampal tissue, purified and processed for bisulfite modification. For all our MSP data, we performed bisulfite treatment of genomic DNA an average of 5–6 times. Bisulfite treatment of genomic DNA converts cytosine to uracil, but leaves methylated 5’Cytosines unchanged. Quantitative real-time PCR was used to determine the DNA methylation status of the bdnf gene. Methylation-specific PCR primers were designed using Methprimer software (http://www.urogene.org/methprimer/)(Li and Dahiya, 2002). Methylation and unmethylated-specific PCR primers were designed to target putative CpG islands detected in silico in promoter or non-promoter regions of the Zif268 gene. Detection of CpG islands (≥ 200 bp) within Zif268 DNA was performed with the following primer sets. For CpG island region 1; Methylated Zif268 sense, 5’-ATTGCGGGGAAGGTTTAGGTC-3’; antisense, 5’- TATATTCCGAAAACCCCAAACG-3’; Unmethylated Zif268 sense, 5’- ATTGTGGGGAAGGTTTAGGTTG-3’; antisense, 5’-ATATTCCAAAAACCCCAAACAC-3’. For CpG island 2; Methylated Zif268 sense, 5’- TTTAGGATGACGGTTGTAGAATTTC-3’; antisense, 5’- GCGTAAAACTAAACGCTCCG-3’; Unmethylated Zif268 sense, 5’- GGATGATGGTTGTAGAATTTTGG-3’; antisense, 5’- CAAACACATAAAACTAAACACTCCA-3’. We also evaluated β-tubulin-4 DNA after context exposure or contextual fear conditioning and found no significant alterations in unmethylated β-tubulin-4 DNA. Thus, we used unmethylated β-tubulin-4 levels for normalization of Zif268 DNA methylation in the studies presented here. B-tubulin-4 sense, 5’-GGAGAGTAATATGAATGATTTGGTG-3’; antisense, 5’-CATCTCAACTTTCCCTAACCTACTTAA-3’. PCR reactions were in a iQ5 real-time PCR system (Bio-Rad, Hercules, CA) using the following cycling conditions: 95 °C for 3 min, 40 cycles of 95 °C for 15 sec and 58–60 °C for 1 min. Detection of the fluorescent products occurred at the end of the 15 sec temperature step. Samples were normalized to β-tubulin-4 and the comparative Ct method was used to calculate differences in gene expression between samples (Livak and Schmittgen, 2001).
Bisulfite-treated samples (as described above) were amplified by primers that amplify the same region of bdnf exon IV DNA, but independent of methylation status. The thermocycler protocol involved an initial denaturation cycle (5 min, 95 °C), 50 cycles of denaturation (1 min, 95 °C), annealing (1 min, 60 °C) and extension (1 min, 72°C), followed by a final extension cycle (5 min, 72°C) terminating at 4 °C resulting in a 258 bp PCR product for exon IV. BSP primer pairs used for the assessment of Zif268 cytosine residues within CpG island region 1 were as follows; Zif2568 sense, 5’-TTGTGAAGGAAGTGTTATTTTG-3’; antisense, 5’- CCAATCTAATAACCCCAAACTT-3’. The BSP products were then purified using a gel extraction kit (Qiagen), and sequenced using the reverse primer at the University of Alabama at Birmingham Genomics Core Facility of the Heflin Center for Human Genetics (http://www.heflingenetics.uab.edu). The percent methylation of each CpG site within the region amplified was determined by the ratio between peaks values of G and A (G/[G+A]), and these levels on the electropherogram were determined using Chromas software.
Behavioral data were analyzed with two-tailed unpaired t tests. Post-hoc comparisons after One-way ANOVA were made using Tukey post-hoc test where appropriate. Data for Mll;eed mutant mice were analyzed using individual one-way ANOVA with LSD post-hoc and simple-effects tests where appropriate for the context test and the CS test. Follow-up comparisons were made using least significance difference test. SPSS 10.0 software was used for statistical analysis of the data. All Western blots were quantified by densitometry (Scion) and GraphPad Prism software was used for statistical analysis of the data. For studies involving the assessment of histone or DNA methylation and mRNA levels were analyzed by two-tailed one-sample t tests when appropriate. Differences in BSP data were analyzed by Two-way ANOVA with Bonferroni post hoc test. Data are shown as the mean ± standard error of mean (SEM). Significance for all tests was set at p ≤ 0.05.
To begin exploring the potential role for histone methylation in memory formation we investigated whether contextual fear conditioning as an associative learning event triggers altered histone methylation within the hippocampus. For behavioral studies animals were fear conditioned in a novel context (training chamber) with a training protocol consisting of three 1-second, 0.5 mA shocks at 2-min shock intervals (Fig. 1A). As a control for our biochemical studies, we first confirmed that the context-plus-shock pairing produced an associative memory. As expected, 24 h after fear conditioning, re-exposing the animals to the training chamber without the footshock triggered recall of the associative memory, as measured by freezing behavior (Fig. 1B). As an additional control, novel context exposure alone produced no significant freezing when the animals were replaced in the training chamber (Fig. 1B).
To evaluate whether histone methylation was altered in response to contextual learning, histone extracts were prepared from area CA1 of hippocampus 1 h after training (Fig. 1A). Thus, we evaluated histone H3 methylation at lysine 4 (K4) as an active marker of transcription and observed a significant increase in tri-methylated H3K4 levels in the hippocampus 1 h after contextual fear conditioning compared to naïve controls (Fig. 1C). No changes in tri-methylated H3K4 levels were observed with context exposure alone (Fig. 1C). These results suggest that H3K4 tri-methylation is regulated in hippocampus after associative fear conditioning training and are consistent with prior observations that active transcription occurs during consolidation of associative contextual memories. These observations also implicate histone methylation as a potential positive regulator of gene transcription in memory consolidation.
We also examined whether or not contextual fear conditioning alters methylation of histone H3 at lysine9 (K9), a repressive marker of transcription, in area CA1 of the hippocampus. Western blotting analysis revealed that 1 h after contextual fear conditioning H3K9 di-methylation significantly increased in area CA1 of the hippocampus (Fig. 1D). Interestingly, context exposure alone produced similar increases in di-methylated H3K9 levels in area CA1 of hippocampus relative to naïve control animals (Fig. 1D). These findings suggest that H3K9 di-methylation is triggered in hippocampus with both novel context learning and associative contextual learning of fear. Together these results are consistent with the idea that histone-methylation-dependent active repression of transcription plays a role in the consolidation of contextual memories. It is interesting that context exposure alone triggered only the transcription-suppressing H3K9 di-methylation mark and not the transcription-activating H3K4 mark (see above); implying that the H3K4 mark may be an associative-learning specific signal.
Next, we determined whether regulation of hippocampal histone H3 methylation triggered with learning about the context or with associative contextual fear conditioning is a dynamic process. To address this, area CA1 of hippocampus was removed from animals 24 h after training and western blotting performed for histone H3 methylation that serve as activation (H3K4) or repressor (H3K9) markers of gene transcription (Fig. 2A). All groups were compared to naïve-handled animals that were not exposed to the training chamber. Twenty-four hours after training (without replacing animals into the training context) western blotting analysis revealed no changes in H3K4 tri-methylation in hippocampus after fear conditioning (Fig. 2B). In addition, levels of tri-methylated H3K4 were unchanged 24 h after context exposure alone when compared to naïve animal controls (Fig. 2B). These results suggest that tri-methylation of H3K4 is reversible, allowing the H3K4 tri-methylation state necessary for memory consolidation to return to basal levels after transient activation. Furthermore, these findings support the idea that a histone demethylase does exist for active demethylation of tri-methylated H3K4 induced in hippocampus after associative fear conditioning training.
Having established that the increase in H3K4 tri-methylation with fear conditioning was reversible, we then sought to determine whether or not the effect of contextual fear conditioning on H3K9 di-methylation was also plastic. Surprisingly, we found that H3K9 di-methylation significantly decreased 24 h after fear conditioning compared to naïve controls (Fig. 2C). We also observed a decrease in H3K9 di-methylation 24 h after exposure to context alone compared to naïve animal controls (Fig. 2C). These results indicate that learning about the context or fear conditioning to the context has effects on H3K9 di-methylation in hippocampus that persist at least 24 h. These findings are the first evidence to indicate a persistent change in a hippocampal histone modification following learning, lasting well beyond the 2–4 h memory consolidation phase.
The results obtained thus far suggest that H3K4 tri-methylation is induced by formation of an associative long-term memory wherein an animal associates a novel context with a footshock. However, it is also possible that the regulation of hippocampal H3K4 tri-methylation was due to a stress response to the footshock alone in the absence of associative memory formation. Therefore, a latent inhibition training paradigm was used to determine whether this event contributed to the regulation of H3K4 tri-methylation in hippocampus after fear conditioning.
The latent inhibition training protocol consists of pre-exposing the animal to a novel context (training chamber) for a long period of time prior to administering the unconditioned stimulus (footshock). This form of contextual learning is unique in that the animal does not form a strong association between the unconditioned stimulus and the novel context, making the actual latent inhibition training paradigm context-specific (Lubow and Josman, 1993; Atkins et al., 1998). Therefore, during a latent inhibition contextual learning the animal forms a spatial memory that attenuates the formation of an associative contextual fear memory. Nevertheless, in this protocol the animal still receives the identical footshock utilized for associative fear conditioning training.
Animals were placed into the novel training context (training chamber) 2 h prior to receiving the three-shock context fear conditioning training (Fig. 3A). The animals were then returned to their home cages. Twenty-four hours later, the animals were re-exposed to the training chamber with no footshock (Test). Latent inhibition fear conditioning inhibited the associative contextual fear memory as assessed by freezing behavior (Fig. 3B). We also found that the increase in hippocampal H3K4 tri-methylation normally observed 1 h after fear conditioning was blocked with the latent inhibition paradigm compared to naïve controls (Fig. 3C). These results suggest that the regulation of H3K4 tri-methylation after training was selective for the formation of robust associative contextual fear memories.
A number of enzymes for methylating or demethylating histones have recently been discovered (summarized in Allis et al., 2007). While few pharmacological agents are currently known to specifically affect these enzymes, genetic knock-out of these chromatin-modifying enzymes or their complexing partners allows us to further elucidate their specific role in learning and memory. Recently, Kim and colleagues reported altered patterns of histone methylation in eed+/−;Mll+/− mice (Kim et al., 2007). Eed interacts directly with the histone-lysine methyltransferase Ezh2 to target H3K27 or H1K26 (Kuzmichev, et al., 2004) and interacts with HDAC 1 and 2 (van der Vlag & Otte, 1999). Mll also interacts with HDAC 1 and 2 (Xia, et al. 2003) and acts as an H3K4-specific methyltransferase (Milne, et al. 2002).
To further assess the role of regulation of histone methylation in the process of memory consolidation, we assessed fear memory in eed and Mll and single and double mutant mice (Paylor et al., 1994; McIlwain et al., 2001; Kim et al., 2007) comparing their behavior to wild-type littermate control animals using an associative contextual fear paradigm. We found significant differences between genotypes during the context behavior test (Fig. 3D). In these comparisons we found that eed+/+;Mll+/− and eed+/−;Mll+/− animals displayed significantly reduced freezing compared to wild-type littermates (all comparisons p < 0.015). The effect on freezing behavior is stronger in heterozygote Mll mice compared to heterozygote eed mice. This is consistent with the idea that Mll is a H3K4-specific methyltransferase whereas Eed is a general component of the methyltransferase complex. Together, these results suggest in-part that the H3K4-specific methyltransferase, Mll, is essential for hippocampal-dependent long-term memory formation. Additionally, we assessed cued fear conditioning in eed+/−;Mll+/+, eed+/+;Mll+/−, and in eed+/−;Mll+/− mice and found no statistically significant differences in freezing behavior among genotypes [F(3,63)=0.664, P=0.57] (Supplemental Fig. 1H). Thus, eed and Mll are important for hippocampus-mediated learning and memory. Together, these findings indicate regulation of hippocampal histone methylation as a molecular mechanism underlying proper memory formation.
As additional behavioral controls, exploratory activity, motor function, and sensorimotor processing were assayed in eed and Mll mutant mice. While eed single heterozygous mice exhibited largely normal behavioral responses, Mll single heterozygotes as well as double heterozygotes exhibited defects in some tests of motor function (Supplementary Fig. 1F). In contrast, exploratory activity, anxiety-related responses, acoustic startle, and prepulse inhibition were normal (Supplementary Fig. 1).
Recent studies have revealed that the methylation state of histone H3 is directly linked to the hyperacetylated state of the histone H3 N-terminal tail in vitro (Zhang et al., 2004). In these studies, Zhang and colleagues report a synchronous relationship between hyperacetylated histone H3 at lysine-9 and hypermethylated histone H3 at lysine-4. Thus, we hypothesized that enhancing histone acetylation via HDAC inhibition might result in an enhancement in histone methylation as well.
To this end, we investigated whether treatment with the HDAC inhibitor sodium butyrate (NaB) might enhance the methylated state of histones in hippocampus during memory formation (Fig. 4A). We first confirmed the reported effect of HDAC inhibition on long-term memory formation. We found that 24 h after the training period, NaB-treated animals displayed significantly more freezing behavior than vehicle-treated controls (Fig. 4B). These results are in agreement with previous reports of the effect of HDAC inhibition on long-term memory formation (Levenson et al., 2004a; Fischer et al., 2007; Lubin and Sweatt, 2007; Guan et al., 2009). We then assayed histone H3 methylation in area CA1 of hippocampus 1 h after training and found that both NaB- and vehicle-treated animals exposed to the fear conditioning paradigm displayed increased H3K4 tri-methylation relative to naïve animal controls (Fig. 4C). However, there was no significant difference in H3K4 tri-methylation from NaB-treated animals relative to that of vehicle-treated controls.
We next examined the effect of enhancing histone acetylation on H3K9 di-methylation in hippocampus following fear conditioning. We found that NaB treatment significantly decreased H3K9 di-methylation in hippocampus relative to vehicle-treated animals (Fig. 4D). These results suggest that elevating histone acetylation via HDAC inhibition leads to suppression of the negative regulator of transcription H3K9 di-methylation. This surprising result suggests that one mechanism whereby HDAC inhibition might enhance memory formation is through secondary suppression of histone H3K9 di-methylation. Regardless of this specific possibility, the result suggests a dynamic cross-talk between histone acetylation and histone methylation in hippocampus.
To further support a role for histone methylation during memory formation, we determined whether fear conditioning triggers changes in chromatin structure directly at the level of individual genes. The Zif268 and bdnf genes are critical for the consolidation of several forms of memories including hippocampus-dependent memory formation (Bramham, 2007; Poirier et al., 2007). Thus, we investigated hippocampal tri-methylated H3K4 levels at the Zif268 and bdnf gene promoters 30 min after training, a time point known to be associated with transcriptional activation of these gene loci. Using chromatin immunoprecipitation (ChIP) assay combined with quantitative real-time PCR, we found that H3K4 tri-methylation increased at the Zif268 promoter after fear conditioning (Fig. 5 A , B). We also found that H3K4 tri-methylation significantly increased at bdnf promoter 1, but not at promoter 4, in area CA1 of hippocampus after fear conditioning, relative to naïve controls (Fig. 5C). Context exposure alone produced no significant changes in H3K4 tri-methylation at the Zif268 or bdnf promoters relative to naïve animal controls (Fig. 5 A–C). The selective increase in histone tri-methylation at the Zif268 promoter and bdnf promoter 1 is in good agreement with fear-conditioning-associated increased transcription of the Zif268 and bdnf genes (Lubin et al., 2008; Alberini, 2009). Together, these results demonstrate active regulation of H3K4 tri-methylation within specific gene promoters, and further demonstrate that H3K4 tri-methylation is regulated in response to fear conditioning.
There is evidence that H3K9 di-methylation represses gene expression via recruitment of DNA methyltransferase enzymes, which mediate methylation of CpG dinucleotides that then recruit repressive chromatin remodeling complexes (Fuks, 2005). In contrast to H3K9 di-methylation, H3K4 tri-methylation mediates active transcription and shows a strong correlation with activated RNA polymerase II and histone acetylation levels at gene promoters (Rice and Futscher, 2000; Nakayama et al., 2001; Rice and Allis, 2001; Li and Zhang, 2004; Zhang et al., 2004; Ruthenburg et al., 2007). Thus, given our finding that H3K4 tri-methylation is increased after fear conditioning, we next determined whether or not DNA methylation was altered during memory consolidation in a fashion that was indicative of an active promoter. Consistent with this idea, we previously have reported that bdnf promoter 1 shows decreased cytosine (DNA) methylation with contextual fear conditioning (Lubin et al., 2008). These previous findings are in good agreement with the idea that H3K4 tri-methylation levels at bdnf promoter 1 (Fig. 5C) correlates with DNA demethylation during fear conditioning. However, we also previously observed that fear conditioning is associated with decreased DNA methylation at bdnf exon 4 at 2 h post-training (Lubin et al., 2008), but this site did not manifest increased histone methylation (at 30 min post-training) in the current studies (Fig. 5D). Thus, the available evidence does not allow us to ascertain whether H3K4 trimethylation is universally correlated with DNA demethylation in the adult CNS.
Therefore in the present studies, we performed additional experiments and examined the DNA methylation state of the Zif268 promoter because increases in H3K4 tri-methylation levels in hippocampus were robust at this promoter region in response to associative contextual fear conditioning (Figure 5B). Furthermore, unlike bdnf DNA methylation (Lubin et al., 2008), the methylation state of the Zif268 gene has not been previously characterized in hippocampus during memory formation.
We first screened the Zif268 promoter region for CpG islands, which are defined as stretches of DNA (≥200bp) where there are a large number of cytosine-guanine dinucleotide sequences linked by phosphodiester bonds (reviewed in Goldberg et al., 2007). Using Methprimer software (http://www.urogene.org/methprimer/)(Li and Dahiya, 2002) two CpG island sites were found within the promoter region of the Zif268 gene immediately preceding the transcriptional start site (Fig. 6A). We first used methyl specific PCR (MSP) to screen DNA methylation changes at each CpG island site 30 min following fear conditioning. Our MSP results demonstrated that fear conditioning triggered changes in DNA methylation at CpG island region 1, but not region 2 in hippocampus (data not shown).
Next, we used direct bisulfite DNA sequencing PCR (BSP) to examine site-specific methylation changes at 20 CpG dinucleotides within Zif268 CpG island region 1 (Fig 6A). The results show that fear conditioning elicited an increase in methylated DNA levels across the region relative to context exposure alone or naïve animal controls (Fig. 6B). However, Zif268 mRNA levels 30 min after training were significantly increased in hippocampus in response to fear conditioning (Fig. 6C). Thus, these results suggest that increased H3K4 tri-methylation levels at the Zif268 promoter is associated with increases in DNA methylation resulting in active transcription of the Zif268 gene after fear conditioning. Overall these data support the idea that H3K4 tri-methylation can be associated with either increased DNA methylation (Zif268), or decreased DNA methylation (bdnf) at genes that are transcriptionally activated in memory consolidation.
Next, we sought to determine the effect of contextual fear conditioning on Zif268 DNA methylation 24 h later. We found that Zif268 DNA methylation and mRNA levels returned to baseline 24 h after fear conditioning compared to naïve controls (Fig. 7A–C). This suggests that the effect of contextual fear conditioning on Zif268 DNA methylation was also transient. Interestingly, we also observed an increase in Zif268 DNA methylation 24 h after exposure to context alone compared to naïve animal controls (Fig. 7A), which corresponds with a decrease in Zif268 gene expression 24 h after context exposure alone (Fig. 7B). These data support the idea that altered DNA methylation can be associated with either increased Zif268 gene expression 30 min after fear memory consolidation or decreased Zif268 gene expression 24 h after context exposure.
Although the findings of increased DNA methylation associated with increase Zif268 gene transcription with contextual fear conditioning were surprising, they are consistent with previous results by Chahrour et al. (2008) suggesting that methyl-binding proteins such as methyl-CpG binding protein 2 (MeCP2) can activate as well as repress gene transcription, specifically suggesting that MeCP2 associates with the transcriptional activator CREB1 at methylated promoter regions of activated gene target. Thus, we sought to determine whether or not the effect of contextual fear conditioning on Zif268 DNA methylation was associated with altered MeCP2 levels at the gene promoter. We found that MeCP2 significantly decreased at the Zif268 promoter region associated with increased H3K4 tri-methylation 30 min after fear conditioning or context exposure compared to naïve controls (Fig. 8A; ChIP1). These results indicate that learning about the context or fear conditioning to the context has effects on the recruitment of MeCP2 levels at the Zif268 promoter in hippocampus. We also observed an increase in MeCP2 at the Zif268 promoter region associated with increased DNA methylation 30 min after fear conditioning, but not after exposure to context alone, compared to naïve animal controls (Fig. 8B; ChIP2). Thus, Zif268 mRNA levels 30 min after training were significantly increased in hippocampus in response to fear conditioning (Fig. 6C). Together, these findings are consistent with the idea that increases in DNA methylation can represent an active signal for gene transcription.
In the studies presented here, we investigated the potential role of histone lysine methylation in memory formation and established several findings. First, we found that fear conditioning triggers changes in H3K4 tri-methylation (a transcriptional active marker) and H3K9 di-methylation (a transcriptional repressive marker) in area CA1 of the hippocampus. Second, we observed that H3K4-specific methyltransferase (Mll) deficient mice have a deficit in long-term memory formation. Third, treatment of animals with the HDAC inhibitor NaB altered histone methylation levels in hippocampus during memory consolidation, suggesting that altered histone methylation is coupled to HDAC inhibition. Finally, H3K4 tri-methylation significantly increased at the promoter of two activated gene targets (Zif268 and bdnf) during memory consolidation. Together, these findings support the role of histone methylation changes in the consolidation of long-term memory formation.
Covalent post-translational modifications of histones serve as an important mechanism for transcriptional regulation during consolidation of long-term memories. For example, increases in histone H3 phosphorylation and acetylation in hippocampus have been shown to be regulated during long-term memory formation (Levenson et al., 2004a; Chwang et al., 2006; Fischer et al., 2007). Additionally, regulation of histone modifications including histone methylation have been implicated in mental disorders including schizophrenia (Deutsch et al., 2008; Akbarian and Huang, 2009). However, investigation of the role of histone methylation in the process of long-term memory formation had not been explored prior to the current studies. Encouragingly, studies on abnormal gene expression in postmortem brain have revealed that histone methylation may be a viable avenue for early detection for some cases of schizophrenia (Akbarian and Huang, 2009). Therefore, the study of histone methylation in the regulation of memory formation is intriguing and will provide further insights into the epigenetic mechanisms that may be dysregulated in cognitive impairments.
We investigated the contribution of hippocampal histone methylation in the consolidation of long-term memory in a rodent model of fear conditioning. We were particularly interested in studying this form of histone modification because it is unique in that various specific sites of methylation of histones can have opposite roles in gene regulation; that is, methylation of H3K4 is associated with active transcription whereas methylation of H3K9 is associated with transcriptional silencing (Strahl et al., 1999; Lachner and Jenuwein, 2002; Schotta et al., 2002). Indeed, the process of storing stable long-term memories is likely to involve coordinated changes in both transcriptional activation and silencing of genes (Ressler et al., 2002; Levenson et al., 2004b). Now with the results presented here, we provide the first evidence for histone methylation changes in hippocampus in response to contextual learning, in both episodic learning and associative contextual fear conditioning. Furthermore, these results support the hypothesis that histone methylation plays a role in the regulation of gene expression changes that are permissive for memory formation.
One fascinating finding from our studies is that HDAC inhibition significantly altered histone methylation levels. Specifically, treatment with the non-selective class I HDAC inhibitor NaB significantly attenuated H3K9 di-methylation in hippocampus after fear conditioning. This suggests the new idea that a possible mechanism for the permissive actions of inhibiting HDACs, resulting in memory enhancement, is through the negative regulation of hippocampal H3K9 di-methylation. However, the specific HDAC isoform (s) inhibited by the HDAC inhibitor NaB to mediate this effect on H3K9 di-methylation is unknown and remains an intriguing focus of ongoing studies. Promisingly, the HDAC2 isoform has recently been identified as a modulator of dendritic spine density, synapse number, and synaptic plasticity which negatively regulates memory formation (Guan et al., 2009). Although very speculative at this point, an intriguing hypothesis is that inhibition of HDAC2 negatively regulates H3K9 di-methylation to allow memory enhancement. Whether this scenario is true or not, our present study is consistent with the idea that inhibition of HDAC isoforms promotes active gene transcription for stable formation of enhanced long-term memories.
Another remarkable finding from these studies is that changes in H3K4 tri-methylation were associated with increased DNA methylation at the Zif268 promoter region, while Zif268 gene expression was active during memory consolidation. In addition, these changes in Zif268 DNA methylation were transient returning to baseline levels 24 h later long after the consolidation period. These observations are in sharp contrast to findings from developmental studies that suggest that DNA methylation is primarily associated with the repression of gene transcription. However, a recent study by Chahrour et al. (2008) indicates that methyl-CpG binding proteins, such as MeCP2, bind to DNA methylation sites with the transcriptional activator cyclic adenosine monophosphate (cAMP) response element binding (CREB) protein 1 to actively regulate gene transcription. Indeed, our results suggest that MeCP2 levels increased at the Zif268 promoter region in association with increased DNA methylation resulting in transcriptional activation of the Zif268 gene following fear conditioning. Correspondingly, we found a cAMP response element (CRE; TCACGTCA) binding site for the transcription factor CREB downstream of the CpG island 1 region and upstream from the transcriptional start site of the Zif268 gene (Fig. 8C). Thus, based on previous studies by Chahrour et al. (2008) and our present data, we propose that dynamic changes in DNA methylation may serve as both a repressive or active marker of transcription in hippocampus during long-term memory formation. This new concept raises several intriguing possibilities for future research. For example, in our study we demonstrated that contextual learning induces global changes in histone methylation as well as altered DNA methylation. However, it remains to be determined what specific cell-types are involved in these observed epigenetic mechanisms.
In summary, our results continue to support a role for epigenetic mechanisms in the process of stable formation of long-term memories. Specifically, our findings demonstrate considerable histone methylation changes in hippocampus in association with memory formation and furthermore suggest that this process is necessary for long-term memory formation. In addition our results provide the first evidence for an association between differential histone methylation and DNA methylation regulation at the Zif268 gene promoter in the adult brain and implicate a possible role for these processes in the formation of long-term memories. Finally, as epigenetic mechanisms continue to be linked to cognitive dysfunctions a better understanding of the complex molecular interaction and regulation of these processes will need to be further refined.
The importance of continued investigation of such mechanisms is underscored by several prior studies. For example, similar to DNA methylation, histone methylation was once considered to be a non-plastic process. The concept of the irreversibility of histone methylation arose from the belief that histone demethylases did not exist. However, this view has recently been challenged with the identification of several histone H3 demethylases including LSD1 and JHDM1 (Shi et al., 2004; Tsukada et al., 2006; Whetstine et al., 2006; Tahiliani et al., 2007) which can specifically demethylate lysine-4 within histone H3. Thus, the discovery of these enzymes may lead to more selective therapeutic interventions that include the use of histone demethylase activators or inhibitors as well as HDAC inhibitors for treatment of cognitive impairments associated with neurological disorders.
(A) Locomotor activity. Open-field assessment of overall locomotor activity and anxiety-related responses displayed no significant difference among genotypes. (B) The center distance ratio score. There was a significant difference amongst the genotypes. Specifically, analysis found that eed+/+;Mll+/− or eed+/−;Mll+/− mice displayed reduced center exploration, suggesting greater anxiety-related responses. (C) Rotarod test. When mice were given four trials on two consecutive days, the overall analysis confirmed that the eed+/+;Mll+/− or eed+/−;Mll+/− mice performed significantly worse on the rotarod compared with eed+/+;Mll+/+ and eed+/−;Mll+/+ littermates. Moreover, this difference was present on trial 2–8. (D–E) Mini-motor test of wire-suspension, vertical-pole and stationary dowel tests. (D) Along with significant differences among the four genotypes in the wire test, the eed+/−;Mll+/− mice performed worse than eed+/+;Mll+/+ and eed+/−;Mll+/+ littermates. (E) The vertical pole test also demonstrated a main effect of genotype and significantly lower scores for eed+/+;Mll+/− animals compared with eed+/+;Mll+/+ and eed+/−;Mll+/+ littermates. (F) Composite motor score of overall performance on four motor tests (rotarod, wire suspension, pole test, and stationary dowel). The composite motor score confirmed both the main effect of genotype in various tests and the overall deficient performance of eed+/+;Mll+/− and eed+/−;Mll+/− mice in contrast to eed+/+;Mll+/+ and eed+/;Mll+/+ littermates. (G) Acoustic startle and prepulse inhibition of the acoustic startle response. Both acoustic startle response and the increase in the level of prepulse inhibition with the increase of the prepulse sound were comparable and not significant between genotypes. (H) Behavioral memory formation measured by contextual and auditory-cue conditioned fear. The left side of the figure indicates the level of freezing for the four genotypes 24 hours following training. In addition to significant differences among the genotypes for the context test, eed+/−;Mll+/+ and eed+/+;Mll+/− animals displayed significantly reduced freezing compared with wild-type littermates. Note the additional decrease in freezing of double heterozygotes in comparison to eed+/−;Mll+/+ and eed+/+;Mll+/− animals, although the latter difference did not reach statistical significance. The right side of the figure displays the level of freezing during the CS test (defined as freezing in the presence of the auditory CS minus the freezing during the pre-CS period). In contrast to the context test, the differences between the genotypes during the CS test did not reach statistical significance. Error bars indicate standard error of mean (SEM).
We thank members of the Lubin laboratory for their thoughtful comments, Dr. M. Han and the University of Alabama at Birmingham Genomics Core Facility for assistance in DNA sequencing, and Dr. A. Schumacher for providing the eed; Mll mutant mice. This work was supported by the National Institute of Health (NS048811, MH082106, MH57014, NS37444, NS013546, AG031722, and NS057098), the Federation of American Societies for Experimental Biology, the National Alliance for Research on Schizophrenia and Depression, the Rotary Clubs Coins for Alzheimer’s Research Trust Fund, and the Evelyn F McKnight Brain Research Foundation.