Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Mar 15.
Published in final edited form as: Biol Psychiatry. 2019 Jun 12;87(6):577–587. doi: 10.1016/j.biopsych.2019.05.022

The SETD6 Methyltransferase Plays an Essential Role in Hippocampus-Dependent Memory Formation

William M Webb 1,*, Ashleigh B Irwin 1,*, Mark E Pepin 2,†, Benjamin W Henderson 3,†, Victoria Huang 1, Anderson A Butler 1, Jeremy H Herskowitz 3, Adam R Wende 2,4, Andrew E Cash 1, Farah D Lubin 1
PMCID: PMC6906268  NIHMSID: NIHMS1531648  PMID: 31378303

Abstract

Background

Epigenetic mechanisms are critical for hippocampus-dependent memory formation. Building on previous studies that implicate the lysine methyltransferase SETD6 in the activation of NF-κB RELA as an epigenetic recruiter, we hypothesize that SETD6 is a key player in the epigenetic control of long-term memory.

Methods

Using a series of molecular, biochemical, imaging, electrophysiological, and behavioral experiments, we interrogated the effects of siRNA-mediated knockdown of Setd6 in the rat dorsal hippocampus during memory consolidation.

Results

Our findings demonstrate that SETD6 is necessary for memory-related NF-κB RELA methylation at Lysine 310 and associated increases in histone H3 lysine 9 dimethylation (H3K9me2) in the dorsal hippocampus; and that SETD6 knockdown interferes with memory consolidation, alters gene expression patterns, and disrupts spine morphology.

Conclusion

Together, these findings suggest that SETD6 plays a critical role in memory formation and may act as an upstream initiator of H3K9me2 changes in the hippocampus during memory consolidation.

Keywords: SETD6, memory consolidation, hippocampus, histone methylation, epigenetics, RELA

Introduction

Memory consolidation in the mammalian brain is a dynamic process in which neuronal gene transcription must be tightly controlled. Epigenetic mechanisms such as posttranslational histone modifications have emerged as critical regulators of this process [1, 2]. While a growing body of literature supports the role of histone modification in memory formation, the processes by which these epigenetic mechanisms are themselves initiated remain poorly understood. One candidate for addressing this gap in knowledge are transcription factors, which, in addition to targeting specific loci throughout the genome also possess the ability to modify chromatin by recruiting epigenetic modifying enzymes [3–6].

One example of such transcription factors is the nuclear factor kappa B (NF-κB). A family of related proteins, NF-κB subunits possess dimerization domains that permit the formation of various dimers, the most abundant of which is the RELA/p50 heterodimer. Classically, RELA/p50 is sequestered in the cytosol by proteins of the family Inhibitor of kappa B (IκB) [7]. Upon activation of the appropriate signaling cascades, IκB degradation occurs, liberating RELA/p50 and allowing it to translocate into the nucleus where it binds to NF-κB binding sites throughout the genome [8]. These binding sites, known as κB sites, consist of variations on the DNA sequence 5-GGGRNYYYCC-3′ (where R is a purine, Y is a pyrimidine, and N is any nucleotide) [9]. In contrast to the classic mechanisms by which NF-κB initiates transcription [10], new studies demonstrate that RELA also mediates epigenetic changes by recruiting chromatin modifiers [11].

NF-κB regulates gene transcription in many contexts and is a critical mediator of learning-induced transcription changes in the brain. In 2004, Levenson showed that κB sites are overrepresented in the promoters of genes with known roles in learning and memory [12]. It has also been shown that RelA−/− mice exhibit deficits in the radial arm maze [13], while the Morris water maze induces NF-κB subunit binding in the hippocampus [14]. Other studies suggest roles for NF-κB in dendritic arborization, GABAergic signaling, and LTP [7, 15]. In the hippocampus, we have shown that NF-κB is involved in the epigenetic control of long-term memory and that fear memory retrieval involves activation of the NF-κB pathway to mediate histone acetylation at gene promoters in the hippocampus [16]. Collectively, these findings suggest that NF-κB plays a role in the epigenetic control of memory consolidation. However, whether NF-κB directly initiates recruitment of epigenetic modifying enzymes to regulate transcription during memory consolidation remains unexplored.

Recent studies have shown that monomethylation of RELA at lysine 310 (K310me1) by the SETD6 methyltransferase causes RELA to recruit the histone methyltransferase EHMT1, facilitating the repressive histone H3 lysine 9 dimethylation mark (H3K9me2) at gene promoters [17, 18]. However, no such interactions have been demonstrated in the brain, nor has SETD6 been shown to play a role in the context of memory or behavior.

In this study, we investigated the role of SETD6 in the hippocampus in response to learning, finding that SETD6 is critical for learning-induced changes in RELA-K310me1, histone methylation, gene transcription, dendritic spine morphology, neuronal electrophysiology, and memory consolidation. In doing so, we begin to address a major gap in our understanding of epigenetic initiation mechanisms in the brain.

Methods and Materials

Animals

All experiments employed male Sprague-Dawley rats between 6–8 weeks old, weighing between 250–300 mg. All animals were housed two-to-a-cage on an alternating 12-hour light/dark cycle and were afforded ad libitum access to food and water in accordance with standards set by the Institutional Animal Care and Use Committee at the University of Alabama at Birmingham.

Cranial Infusions

Region-specific knockdown of target genes was achieved by cranial infusion of short interfering RNA (siRNA) according to previously described stereotactic surgical protocols [19]. See SI for expanded methods.

Behavior Tests

Behavioral tests were conducted as previously described [20–22]. See SI for expanded methods.

Tissue Collection and Protein Extraction

Animals were euthanized by rapid decapitation, and their brains were removed for immersion in O.C.T. (for immunohistochemistry) or flash-frozen on dry ice for later subdissection of the dorsal hippocampus (CA1) and retrosplenial cortex (RSC). Preparation for immunohistochemistry or extraction of nuclear and cytoplasmic fractions was conducted according to previously described protocols [23]. See SI for expanded methods.

Histone Isolation

Histone isolation was performed as previously described [24, 25]. See SI for expanded methods.

Protein Co-immunoprecipitation and GST Pulldown

Protein co-immunoprecipitation was performed using PureProteome Protein A/G Mix Magnetic Beads (Millipore) according to the manufacturer’s instructions and using the following antibodies: anti-SETD6 (Abbexa #005464), anti-NF-κB RELA (p65) (Abcam #16502). For GST pulldown, GST-labeled EHMT1 fusion proteins (Creative Biomart) or GST control proteins (Fisher) were used with glutathione magnetic beads (Pierce) to pull down proteins from nuclear and cytoplasmic extracts, incubated together in wash buffer (75 mM NaCl, 5mM DTT, 8.3% glycerol, and 1.6% Triton X-100) for 2 hours at 4°C. Captured proteins were collected by heating for 5 min at 95°C in SDS loading buffer under agitation at 800 rpm.

Electrophoretic Mobility Shift Assay (EMSA)

Fluorescently labeled IRDye NF-κB consensus oligonucleotides (Licor) were incubated with CA1 nuclear extracts from the following treatment groups: naïve (home cage) rats treated with scrambled, non-targeting siRNA; CFC-trained rats treated with scrambled, non-targeting siRNA; and CFC-trained rats treated with siRNA directed against Setd6. EMSA was performed according to the Odyssey Infrared EMSA Kit (Licor) manufacturer’s instructions.

Chromatin Immunoprecipitation (ChIP)

ChIP was conducted as previously described [26] with modifications. See SI for expanded methods and primer design.

Electrophysiological Recordings

Rats underwent cranial infusion of siRNA followed by five days of recovery, after which their brains were harvested and hippocampal slices were collected for further testing. High frequency stimulation of the Schaffer collateral pathway was conducted using four trains of 100 pulses at 100 Hz, spaced 60 seconds apart. The initial slope of the field excitatory postsynaptic potential (EPSP) was measured as an index of synaptic strength. %fEPSP slopes were averaged after 20 min of baseline recording. Data are reported as means ±SEM, where n = number of slices.

Primary Rat Neuron Culture & Transfection

Rat hippocampal neurons were isolated from E18 rat embryos and cultured at high density on poly-L-lysine-coated glass coverslips as previously described [27]. See SI for expanded methods.

Widefield Fluorescent Microscopy

On DIV 14, neurons were fixed with 2% PFA in 0.1M PBS and washed 2 times with 1X PBS. Coverslips were mounted on microscope slides using Vectashield mounting media (Vector Laboratories). Images were captured on a Nikon (Tokyo, Japan) Eclipse upright microscope. A Nikon Intensilight and Photometrics Coolsnap HQ2 camera were employed to image Lifeact-GFP using the Nikon Elements 4.20.02 image-capture software with a 60X oil-immersion objective (Nikon Plan Apo, N.A. 1.40). Z-series images were acquired at 0.15μm increments through the entire visible dendrite.

Dendritic Spine Image Processing and Analysis

Prior to analysis, captured images were deconvolved using Huygens Deconvolution System (16.05, Scientific Volume Imaging). Semi-automated analysis was then performed with Neurolucida 360 (2.70.1, MBF Biosciences). Dendritic spine analysis was performed as previously described [28] with adjustments. Each dendritic protrusion was automatically classified as a dendritic filopodium, thin spine, stubby spine, or mushroom spine based on morphological measurements using previously described methods [28]. Reconstructions were collected in Neurolucida Explorer (2.70.1) for branched-structure analysis and exported to Microsoft Excel. Spine density was calculated as the number of spines per μm dendrite length. See SI for expanded methods.

Statistical Analysis

With the exception of the RNA sequencing analysis described below, all statistical analyses were performed using Prism® v.7 (Graphpad). Outliers were removed according to Grubb’s outlier test. Unless otherwise specified, comparisons of means between groups were made using two-tailed Student’s t-tests or one-way analysis of variance (ANOVA) followed by Tukey’s post-hoc test for multiple comparisons. One-tailed t-tests were deemed acceptable for confirmatory/knockdown tests where changes in a particular direction were expected. Differences in dendritic spine frequency were determined using the Kolmogorov-Smirnov (K-S) test.

RNA-Sequencing and Analysis

RNA was extracted from hippocampal subdissections using Qiagen’s RNA Isolation protocol and submitted to Novogene (Sacramento, CA) for paired-end RNA sequencing using Illumina HiSeq 2500 with 150 BP read-length. Quality control and analysis were conducted in R according to code provided in the SI. Briefly, sequenced samples were quality-inspected using Multiqc [29] to ensure PHRED > 30. Reads were aligned and annotated to the Ensembl Rattus norvegicus genome (Rnor_6.0.90) using STAR [30], and differential gene expression performed using DESeq2 (v1.18.1) [31] within the R (v3.4.2) computing environment, as previously described [32]. See SI for expanded methods and statistics.

Data Availability

Data included in sequencing analysis are available online from the NCBI Sequence Read Archive underBioProject PRJNA476137. Other data are available from the corresponding author upon reasonable request.

Results

Learning triggers NF-κB RELA methylation in the hippocampus, and SETD6 associates with EHMT1

To investigate the roles of SETD6 and NF-κB RELA methylation in memory consolidation, we developed a custom polyclonal antibody (Covance) against the RELA-K310me1 modification. After troubleshooting the antibody (Fig. S1), we asked whether RELA-K310me1 levels change in area CA1 1 hr post-training in a context fear conditioning (CFC) memory paradigm (Fig. 1a). We found that RELA-K310me1 levels increased 1 hr post-training along with global H3K9me2 levels, but not SETD6 protein expression, according to Western blot analysis (Fig. 1b). This change was also appreciable by immunohistochemistry using the same custom antibody (Fig. 1c) on hippocampal slices. To determine whether SETD6, RELA, or EHMT1 interact in the rat hippocampus in vivo, we used protein co-immunoprecipitation assays and found interactions between SETD6 and RELA in the nuclear, but not the cytoplasmic fractions, of CA1 (Fig. 1d). We next used a recombinant, GST-tagged EHMT1 peptide (Creative Biomart) to pull down EHMT1 in the same region and found an interaction between EHMT1 and SETD6 in the nuclear, but not the cytoplasmic, fractions (Fig. 1d).

Fig. 1. Learning triggers NF-κB -RELA methylation in the hippocampus, and RELA associates with SETD6 and EHMT1.

Fig. 1.

Open in a new tab

(A) Schematic representing context fear conditioning followed by subdissection of CA1 and downstream tests. (B) Immunoblotting revealed no change in SETD6 protein expression following training (cytoplasmic protein fraction, n=4, p=0.5693, unpaired t-test), while RELA-K310me1 (cytoplasmic protein fraction, n=6, p=0.0460, unpaired t-test) expression and H3K9me2 (histones extracted from nuclear fraction, n=8, p=0.0210, unpaired t-test) levels both increased in area CA1. (C) Representative images from dorsal hippocampal slices show increased fluorescent signal when treated with anti-RELA-K310me1 antibody and developed with FITC-conjugated secondary antibody. (D) Protein co-immunoprecipitation and GST-glutathione assays revealed in vivo interactions between RELA and SETD6 as well as EHMT1 in the nuclear, but not the cytoplasmic, fractions of hippocampal (CA1) lysates.

Knockdown of Setd6 prevents learning-associated RELA-K310 methylation and H3K9me2

We next asked whether knockdown of Setd6 gene expression would prevent learning-associated increases in RELA-K310me1 and H3K9me2. After confirming that knockdown of Setd6 mRNA and protein expression was significant and region-specific (Fig. S2), we found that infusion of siRNA directed against Setd6 in CA1 blunted learning-associated increases in RELA-K310me1 (Fig. 2b) and H3K9me2 (Fig. 2c) at 1 hr after CFC. Under naïve conditions, there were no differences in RELA-K310me1 or H3K9me2 with siSetd6 infusion (Fig. S7), suggesting that changes Setd6 knockdown are activity dependent. To determine whether manipulation of Setd6 and RELA-K310me1 affects the intrinsic binding activity of the NF-κB transcription factor, we performed an electrophoretic mobility shift assay (EMSA) and asked whether binding of CA1 nuclear extract proteins to NF-κB consensus oligonucleotides was altered by knockdown of Setd6. We found no significant difference in oligonucleotide binding or migration with Setd6 knockdown in CA1 compared to similar controls infused with scrambled, non-targeting siRNA (Fig. 2d).

Fig. 2. Knockdown of Setd6 prevents learning-associated RELA-K310 methylation and H3K9me2.

Fig. 2.

Open in a new tab

(A) Stereotax-guided cranial infusion of siSetd6 into area CA1 prevents memory consolidation-associated increases in (B) RELA-K310me1 (cytoplasmic protein fraction, n=8, p=0.041, unpaired t-test) as well as consolidation-associated increases in (C) H3K9me2 (histones from nuclear fraction, n=10, p=0.0011, unpaired t-test) compared to scrambled, non-targeting controls. (D). Electrophoretic mobility shift assay showed no significant changes in migration hindrance for NF-κB consensus sequence oligonucleotides between untrained rats infused with scrambled, non-targeting siRNA (Naïve-Scr), context fear conditioned rats infused with non-targeting siRNA (CFC-Scr), or context fear conditioned rats infused with siSetd6 (CFC-siSetd6) (n=3, p=0.3398, oneway ANOVA). Single comparison testing between CFC-Scr and CFC-siSetd6 groups was also insignificant (n=3, p=0.1930, unpaired t-test).

Knockdown of Setd6 results in widespread transcriptional changes in the hippocampus

To better understand the biological significance of SETD6-associated changes in RELA-K310me1 and H3K9me2, we investigated changes in gene expression that accompany knockdown of Setd6 during memory consolidation compared to similarly-trained rats infused with scrambled, non-targeting controls (Fig. 3a). After confirming decreased Setd6 mRNA expression area CA1 of siSetd6-treated rats (Fig. S3a), we collected RNA from the CA1 of rats sacrificed 1 hr after CFC and submitted it for mRNA sequencing. After quality control assessment, alignment, and high-level characterization of the data (Fig. S3b–c), we found robust differences between the transcriptional profiles of both groups, with over 1300 differentially expressed genes (adjusted p < 0.05) (Fig. 3b, 3d). In total, approximately 60% of all differentially expressed genes were overexpressed following knockdown of Setd6, while 72.9% of genes annotated by Ingenuity Pathway Analysis (IPA, Qiagen) for NF-κB regulation were upregulated (Fig. 3c–d). Reactome pathway analysis returned multiple ontologies significant for learning and memory, including “Transmission across Chemical Synapses,” “Neurotransmitter Receptors and Postsynaptic Signal Transmission,” and others. (Fig. 3e; see Datasets S1 and S2 for tables of enriched gene ontologies).

Fig. 3. Knockdown of Setd6 results in widespread transcriptional changes in the hippocampus.

Fig. 3.

Open in a new tab

(A) Stereotax-guided cranial infusion of siSetd6 into area CA1 alters transcription of memory-related and κB-regulated genes during memory consolidation. (B) Heat map depicts clear clustering of upregulated and downregulated gene sets by treatment group (siSetd6 vs. Scr). (C) Following siSetd6 treatment and context fear training, approximately 60.0% of differentially expressed genes were upregulated, while 40.0% were downregulated. 72.9% of genes annotated for regulation by NF-κB were upregulated with treatment, while 27.1% were downregulated. (D) Comparing curated lists of genes annotated for learning and memory to our total list of differentially expressed genes revealed that 114 (8.9%) of the 1277 differentially expressed genes recognized by IPA were known to play a role in learning or memory, and 85 (6.7%) were specifically annotated for regulation by NF-κB. (E) Reactome pathway analysis revealed a number of memory-relevant gene ontologies being dysregulated following treatment with siSetd6 (see Datasets S1, S2 for full tables).

Knockdown of Setd6 alters H3K9me2 enrichment at memory-related gene Comt near RELA binding sites

To confirm that these changes in gene expression occurred alongside changes in H3K9me2 near κB consensus sites, we identified Comt as an upregulated (log2FC +0.56, p=0.0161), memory-relevant gene from our RNA sequencing data and asked whether H3K9me2 occupancy would be changed near putative RELA binding sites using chromatin immunoprecipitation (ChIP) followed by real-time quantitative PCR (RT-qPCR) (Fig. 4a). We designed primers for Comt, using LASAGNA-Search 2.0 to identify possible κB sites within 1kB of the Comt gene transcription start site (Fig. 4b). After confirming that Comt expression was increased according to RT-qPCR performed independently of our sequencing data (p=0.0375), we used ChIP-qPCR to show that knockdown of Setd6 results in decreased H3K9me2 occupancy (p=0.0334) near putative κB sites (Fig. 4c). Analysis of H3K9me2 occupancy near these sites under naïve conditions showed no differences between Setd6 knockdown and control groups (Fig. S8), indicating this an activity dependent change.

Fig. 4. Knockdown of Setd6 decreases H3K9me2 occupancy at memory-related gene Comt near RELA binding sites.

Fig. 4.

Open in a new tab

Knockdown of Setd6 by (A) stereotax-guided cranial infusion of siRNA into area CA1 decreases gene-specific H3K9me2 occupancy near two (B) putative RELA binding sites at (C) exon 1 of Comt (n=4, p=0.0334, unpaired t-test) while Comt gene expression increases (n=4–5, p=0.0375, unpaired t-test).

Knockdown of Setd6 alters dendritic spine morphology in hippocampal neurons

Given the widespread changes observed in the transcriptional profile of CA1 following knockdown of Setd6 during memory consolidation, we next sought to determine whether knockdown of Setd6 affected dendritic spine morphology in hippocampal neurons. Accordingly, we cultured rat hippocampal neurons, treated them with siRNA, and transfected them with plasmid encoding Lifeact-GFP for visualization of F-actin in dendritic spines. Two days later, neurons were fixed with PFA and mounted for imaging and analysis. Fig. 5a depicts representative tracings from Setd6-knockdown neurons and controls. After confirming that RELA-K310me1 protein expression was successfully reduced in siSetd6-treated neurons (Fig. S4), we found that overall spine density remained unchanged with knockdown of Setd6 compared to controls, with trends toward increased density of thin spines (p=0.0718) and filopodia (p=0.0680) (Fig. 5b). Overall spine extent was significantly increased following knockdown of Setd6 (p=0.0055, K-S test) (Fig. 5c), with additional trends toward increased thin spine extent (Fig. 5d). Mushroom-type spine extent (Fig. 5e) and overall spine head diameters (Fig. 5f) were not changed following knockdown of Setd6 compared to controls.

Fig. 5. Knockdown of Setd6 alters dendritic spine morphology in primary hippocampal neurons.

Fig. 5.

Open in a new tab

(A) Representative images of Lifeact-GFP treated neurons show differences in overall spine extent following knockdown of Setd6. (B) Overall spine density remained unchanged after knockdown of Setd6 compared to controls, with trends toward increased density in thin spines (n=6–10, p=0.0718, unpaired t-test) and filopodia (n=6–10, p=0.0680, unpaired t-test) morphology. (C) Overall spine extent was significantly increased following knockdown of Setd6 (p=0.0055, D=0.1687, K-S test), with trends toward significant increase in (D) thin spine extent (p=0.1061, D=0.2527, K-S test). (E) Mushroom spine extent and (F) overall head diameter were not altered.

Knockdown of Setd6 increases LTP at Schaffer collateral synapses

We next asked whether knockdown of Setd6 and its associated changes in dendritic spine morphology would affect neuronal electrophysiology. To test this, we infused siRNA directed against Setd6 or scrambled, non-targeting controls (Scr) into the dorsal hippocampus and harvested hippocampal slices for high frequency stimulation of the Schaffer collateral pathway (Fig. 6a). We found that knockdown of Setd6 actually enhanced LTP immediately upon high-frequency stimulation, persisting 70 minutes, and then again at 120 minutes (Fig. 6b). Meanwhile basal synaptic transmission by input-output function (Fig. 6c) and paired-pulse facilitation between pathways (Fig. 6d) were not significantly altered between groups.

Fig. 6. Knockdown of Setd6 increases LTP at Schaffer collateral synapses.

Fig. 6.

Open in a new tab

Knockdown of Setd6 by (A) stereotax-guided cranial infusion of siRNA into area CA1 enhances (B) early and late LTP at the Schaffer (CA3-CA1) collateral pathway (*=p<0.05). (C) Basal synaptic transmission by input-output function was not affected by treatment. (D) Paired-pulse facilitation between pathways was also unaffected by treatment.

Knockdown of Setd6 impairs long-term memory consolidation, subject to rescue by preservation of H3K9me2 levels

Finally, we sought to determine whether knockdown of Setd6 would result in changes to performance on hippocampus-dependent memory tasks, and whether these changes could be rescued by counteracting changes to H3K9me2, the putative means by which SETD6-mediated RELA-K310me1 mediates epigenetic changes (see Fig. 8). A well-established approach to preserve histone methylation levels is to reduce lysine (K)-specific demethylase 1A (KDM1A) activity. Prior studies have demonstrated preservation of H3K9me2 levels by decreasing KDM1A is sufficient to rescue behavioral deficits due to decreased H3K9me2 levels in the amygdala [26]. Using this approach, we confirmed knockdown of Setd6 and/or Kdm1 mRNA in the relevant experimental groups (Fig. S5a) and asked whether knockdown of Setd6 in CA1 impaired CFC or object location memory. At the same time, we asked whether knockdown of Kdm1 would preserve H3K9me2 levels in light of siSetd6 treatment, thereby rescuing any memory deficits caused by knockdown of Setd6. After confirming that fear memory acquisition during training was not affected by siRNA infusion (Fig. S5b), we found that knockdown of Setd6 significantly impaired fear memory consolidation as evidenced by decreased freezing behavior 24 hrs post-training (Fig. 7a) and that these deficits could be restored when siSetd6 was co-infused with siRNA directed against Kdm1 (siKdm1). Additionally, we observed that co-infusion of siSetd6 and siKdm1a in the hippocampus resulted in similar H3K9me2 levels as siKdm1a infusion alone (Fig. S6). For object location memory tests, we found significant deficits in object discrimination index compared to controls (Fig. 7b, S5c). As with CFC, these deficits were also rescued by co-infusion with siKdm1.

Fig. 8. Summary cartoon depicting the putative role of SETD6 as an epigenetic initiator in the rat hippocampus.

Fig. 8.

Open in a new tab

We hypothesize that SETD6 monomethylates NF-κB RELA at Lysine 310, promoting the recruitment of EHMT1, which mediates H3K9me2 at κB-regulated target genes. This results in transcriptional changes that alter dendritic spine morphology, neuronal electrophysiology, and ultimately memory, leading to behavioral changes that can be measured by memory consolidation tests.

Fig. 7. Knockdown of Setd6 impairs long-term memory consolidation, subject to rescue by preservation of H3K9me2 levels.

Fig. 7.

Open in a new tab

Stereotax-guided cranial infusion of siSetd6 into area CA1 impairs memory consolidation in two hippocampus-dependent tasks. However, these memory deficits can be rescued by co-infusion of siKdm1—siRNA directed against the histone lysine demethylase Kdm1. (A) Context fear memory was impaired by infusion of siSetd6 compared to scrambled, nontargeting controls but was restored when siSetd6 was co-infused with siKdm1 (n=6–8, p=0.0010, one-way ANOVA; significance by Tukey’s multiple-comparison test between groups indicated by *[p<0.05] or **[p<0.01]). (B) Likewise, object location memory was abolished by knockdown of Setd6 and restored by concurrent administration of siKdm1 (n=6–8, p=0.0184, one-way ANOVA; significance by Tukey’s multiple-comparison test between groups indicated by *[p<0.05]).

Discussion

Although SETD6 methylates RELA at lysine 310, thereby facilitating recruitment of EHMT1 in various cell lines in vitro [17, 18], no one has demonstrated this mechanism in the mammalian brain or in the context of learning and memory. In this study, we determined that RELA, SETD6, and EHMT1 all interact in vivo in the mammalian hippocampus (Fig. 1d), a critical first step in establishing the presence of this process in the brain. We established that normal memory consolidation is associated with increased RELA-K310me1 alongside previously characterized increases in H3K9me2 in CA1 following CFC [24]. To determine whether this change in RELA-K310me1 was related to changes in de novo SETD6 expression (versus protein activity), we examined SETD6 protein levels during consolidation and found that it did not change post-training (Fig. 1c). This finding suggests that the activity of SETD6 may be limited to those instances in which NF-κB has already been liberated from IκB in the cytoplasm and has translocated into the nucleus following upstream stimulation of the neuron.

To establish a link between SETD6, RELA-K310me1, and H3K9me2 in the hippocampus, we infused siRNA against Setd6 into area CA1 of the hippocampus, confirming that mRNA knockdown and subsequent changes in SETD6 protein expression were limited to CA1, and not adjacent regions (see Methods). These findings were corroborated by our RNA sequencing data, where the Allen Brain Atlas predicted tissue-specificity to area CA1 of the hippocampus given the set of differentially expressed genes following Setd6 knockdown (Fig. S3d). Additionally, we confirmed by EMSA that knockdown of Setd6 and its associated decrease in RELA-K310me1 did not alter the intrinsic binding capabilities of NF-κB.

Analysis of RNA sequencing data revealed that a majority (roughly 60%) of differentially expressed genes showed increased expression with knockdown of Setd6. This is in keeping with our understanding of H3K9me2 as a transcriptionally repressive mark. Furthermore, of those differentially expressed genes that were also annotated for regulation of NF-κB by IPA, 72.9% genes showed increased expression (Fig. 3c) following knockdown of Setd6—an encouraging finding given our hypothesis. Regarding the non-trivial number of genes whose expression decreased following Setd6 knockdown, we consider that many of our differentially expressed genes are themselves transcriptional repressors. For example, one upregulated gene was Ezh2 (log2FC +0.79)—the functional enzymatic component of the polycomb repressor complex and mediator of the repressive H3K27me3 mark. Pathway analysis of differentially expressed genes identified a number of ontologies that anticipated our later findings in dendritic spine morphology, electrophysiology, and behavior.

In keeping with our hypothesis, ChIP-qPCR showed decreased H3K9me2 occupancy near a putative κB binding site in Comt in concert with that gene’s upregulation following Setd6 knockdown. Comt, a gene responsible for the synaptic catabolism of dopamine (and required for long-term memory), is also known to undergo expression changes with NF-κB activity and has a known κB binding motif immediately upstream of the transcription start site [33, 34]. Although H3K9me2 is generally considered a repressive mark, maps of the epigenetic landscape generated by ChIP-seq show that many active genes exhibit H3K9me2 enrichment at their promoters [35]—a caveat that cautions against attributing exclusive transcriptional control to a single epigenetic mark. H3K9me2 is increased globally in CA1 following CFC, and disruption of this change in CA1 by pharmacologic inhibition of the EHMT1/EHMT2 complex in CA1 impairs long-term memory, compromising expression of critical, memory-related genes [24].

In recent years, a growing body of evidence emerged to suggest a relationship between dendritic spine morphology and synaptic neurotransmission in neuronal plasticity, learning, and memory [36–38]. Classifying dendritic spines on the basis of their 3-dimensional structure yields several categories of dendritic spine, including “stubby,” “mushroom,” “thin,” and “filopodia.” Although live imaging studies underscore the dynamic nature of spine number and shape [39], these morphological categories possess functional implications for the spine-synapse relationship [27]. By studying cultured rat hippocampal neurons in vitro, we were able to employ automated tracing and analysis software and ask whether the changes detected in our electrophysiological studies could be elucidated in terms of morphologic differences in a cell-type specific manner.

We found significant increases in overall spine extent following knockdown of Setd6, with a trend toward increased thin spine extent. These findings suggest an increase in rapid plasticity. Additionally, we saw trends toward increased spine density for filopodia and thin spines, suggesting new spine formation. These interpretations agree with our other finding of increased LTP following Setd6 knockdown; indeed, they corroborate studies in which manipulations resulting in increased thin spine and filopodia density and extent were associated with enhanced LTP in rat hippocampal brain slices [27, 40]. Our findings of gene set enrichment in such categories as “Signaling by Rho GTPase” (Dataset S1) may help to explain the formation and growth of new dendritic spines in terms of expression changes in cytoskeleton-related genes, where Rho GTPases play a critical role in dendritic spine plasticity [41].

LTP in the Schaffer collaterals relies on activation of NMDA receptors during its early phase, while protein synthesis becomes increasingly important for the maintenance of late LTP [42, 43]. Our findings of both immediate and late increases in LTP suggest drastic changes in the transcriptional milieu of neurons following knockdown of Setd6, findings that are corroborated by robust changes in gene expression identified by our RNA sequencing analysis. While increased LTP is usually seen as a cellular correlate of long-term memory formation [44], some studies show that increased LTP is accompanied by behavioral deficits [45] such as we found following knockdown of Setd6 (Fig. 7). Moreover, our findings may actually be anticipated by the previously identified changes in histone methylation, gene transcription, and dendritic spine morphology. Decreased global H3K9me2 levels, increased overall gene transcription (including genes annotated for NMDA receptor function [46] and Rho GTPase activity [47]), increased expression of genes like Camk2d [48], and increased dendritic spine extent all characterize a transcriptional milieu that corroborates increased LTP.

If methylation of RELA at Lys310 precedes RELA’s recruitment of EHMT1, as previously shown [18], and if EHMT1-mediated H3K9me2 is the primary means by which SETD6 exerts epigenetic control of gene transcription during memory consolidation (see Fig. 8), then the ability to intervene at the level of histone methylation and countermand the decrease in histone methylation induced by knockdown of Setd6 is an important step in testing our proposed mechanism of epigenetic control. Interestingly, previous studies have shown that preservation of H3K9me2 by knocking down Kdm1 is sufficient to rescue behavioral deficits originating in compromised H3K9me2 levels in the rodent amygdala [26]. Accordingly, we chose global histone methylation as a point of rescue and asked whether Kdm1 knockdown in the hippocampus would restore behavioral deficits induced by Setd6 knockdown. Our finding that co-infusion siKdm1 with siSetd6 was sufficient to rescue two forms of hippocampus-dependent memory provides even stronger support for the claim that SETD6 mediates its effects on memory chiefly through its effect on H3K9me2 levels. It also corroborates past studies which suggest that H3K9me2 can bidirectionally regulate fear memory formation [38].

In conclusion, our findings suggest a new role for SETD6 as an epigenetic initiator in the transcriptional control of memory, potentially through its effects on the transcription factor NF-κB RELA. We find that knockdown of SETD6 is associated with changes in RELA methylation, post-translational histone modification, gene transcription, dendritic spine morphology, neuronal electrophysiology, and ultimately, hippocampus-dependent memory behavior (see Fig. 8). As many cognitive disorders feature changes in the epigenetic control of gene expression [49] and that NF-κB plays a known role in the epigenetic pathophysiology of disorders like Alzheimer’s disease [50], this study begins to explore ways in which transcription factors initiate epigenetic changes in the brain. Moreover, these findings may eventually identify new therapeutic strategies for mitigating the cognitive symptoms of psychiatric conditions related to traumatic memories such as post-traumatic stress disorder.

Supplementary Material

1
2
3

Acknowledgments

The authors would like to thank Dr. Cristin Gavin and Dr. Jing Wang of the UAB McKnight Synaptic Plasticity Core. This work was supported in part by National Institute of Health (NIH) grants MH097909 to F.D.L. and the Evelyn F. McKnight Brain Institute at the University of Alabama at Birmingham (UAB). W.M.W. is supported by a predoctoral fellowship from the NIH (F30 NS100340). M.E.P. is also supported by a predoctoral fellowship from the NIH (F30 HL137240).

Footnotes

Disclosures: The authors report no biomedical financial interests or potential conflicts of interest.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  • 1.Jarome TJ and Lubin FD, Epigenetic mechanisms of memory formation and reconsolidation. Neurobiol Learn Mem, 2014. 115: p. 116–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Jarome TJ, Thomas JS, and Lubin FD, The epigenetic basis of memory formation and storage. Prog Mol Biol Transl Sci, 2014. 128: p. 1–27. [DOI] [PubMed] [Google Scholar]
  • 3.Brettingham-Moore KH, Taberlay PC, and Holloway AF, Interplay between Transcription Factors and the Epigenome: Insight from the Role of RUNX1 in Leukemia. Front Immunol, 2015. 6: p. 499. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Donati G, Imbriano C, and Mantovani R, Dynamic recruitment of transcription factors and epigenetic changes on the ER stress response gene promoters. Nucleic Acids Res, 2006. 34(10): p. 3116–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Soufi A, et al. , Pioneer transcription factors target partial DNA motifs on nucleosomes to initiate reprogramming. Cell, 2015. 161(3): p. 555–568. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Mayran A and Drouin J, Pioneer transcription factors shape the epigenetic landscape. J Biol Chem, 2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Snow WM, et al. , Roles for NF-kappaB and gene targets of NF-kappaB in synaptic plasticity, memory, and navigation. Mol Neurobiol, 2014. 49(2): p. 757–70. [DOI] [PubMed] [Google Scholar]
  • 8.Cieslik M and Bekiranov S, Genome-wide predictors of NF-kappaB recruitment and transcriptional activity. BioData Min, 2015. 8: p. 37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Chen FE, et al. , Crystal structure of p50/p65 heterodimer of transcription factor NF-kappaB bound to DNA. Nature, 1998. 391(6665): p. 410–3. [DOI] [PubMed] [Google Scholar]
  • 10.Barboric M, et al. , NF-kappaB binds P-TEFb to stimulate transcriptional elongation by RNA polymerase II. Mol Cell, 2001. 8(2): p. 327–37. [DOI] [PubMed] [Google Scholar]
  • 11.de la Fuente V, et al. , NF-kappaB transcription factor role in consolidation and reconsolidation of persistent memories. Front Mol Neurosci, 2015. 8: p. 50. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Levenson JM, et al. , A bioinformatics analysis of memory consolidation reveals involvement of the transcription factor c-rel. J Neurosci, 2004. 24(16): p. 3933–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Meffert MK, et al. , NF-kappa B functions in synaptic signaling and behavior. Nat Neurosci, 2003. 6(10): p. 1072–8. [DOI] [PubMed] [Google Scholar]
  • 14.O’Mahony A, et al. , NF-kappaB/Rel regulates inhibitory and excitatory neuronal function and synaptic plasticity. Mol Cell Biol, 2006. 26(19): p. 7283–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Salles A, Romano A, and Freudenthal R, Synaptic NF-kappa B pathway in neuronal plasticity and memory. J Physiol Paris, 2014. 108(4–6): p. 256–62. [DOI] [PubMed] [Google Scholar]
  • 16.Lubin FD and Sweatt JD, The IkappaB kinase regulates chromatin structure during reconsolidation of conditioned fear memories. Neuron, 2007. 55(6): p. 942–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Chang Y, et al. , Structural basis of SETD6-mediated regulation of the NF-kB network via methyl-lysine signaling. Nucleic Acids Res, 2011. 39(15): p. 6380–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Levy D, et al. , Lysine methylation of the NF-kappaB subunit RelA by SETD6 couples activity of the histone methyltransferase GLP at chromatin to tonic repression of NF-kappaB signaling. Nat Immunol, 2011. 12(1): p. 29–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Webb WM, et al. , Dynamic association of epigenetic H3K4me3 and DNA 5hmC marks in the dorsal hippocampus and anterior cingulate cortex following reactivation of a fear memory. Neurobiol Learn Mem, 2017. 142(Pt A): p. 66–78. [DOI] [PubMed] [Google Scholar]
  • 20.Jarome TJ, et al. , NF-kappaB mediates Gadd45beta expression and DNA demethylation in the hippocampus during fear memory formation. Front Mol Neurosci, 2015. 8: p. 54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Morse SJ, et al. , Environmental enrichment reverses histone methylation changes in the aged hippocampus and restores age-related memory deficits. Biology (Basel), 2015. 4(2): p. 298–313. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Bevins RA and Besheer J, Object recognition in rats and mice: a one-trial nonmatching-to-sample learning task to study ‘recognition memory’. Nat Protoc, 2006. 1(3): p. 1306–11. [DOI] [PubMed] [Google Scholar]
  • 23.Rong Y and Baudry M, Seizure activity results in a rapid induction of nuclear factor-kappa B in adult but not juvenile rat limbic structures. J Neurochem, 1996. 67(2): p. 662–8. [DOI] [PubMed] [Google Scholar]
  • 24.Gupta-Agarwal S, et al. , G9a/GLP histone lysine dimethyltransferase complex activity in the hippocampus and the entorhinal cortex is required for gene activation and silencing during memory consolidation. J Neurosci, 2012. 32(16): p. 5440–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Ryley Parrish R, et al. , Status epilepticus triggers early and late alterations in brain-derived neurotrophic factor and NMDA glutamate receptor Grin2b DNA methylation levels in the hippocampus. Neuroscience, 2013. 248: p. 602–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Gupta-Agarwal S, et al. , NMDA receptor- and ERK-dependent histone methylation changes in the lateral amygdala bidirectionally regulate fear memory formation. Learn Mem, 2014. 21(7): p. 351–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Swanger SA, et al. , ROCK1 and ROCK2 inhibition alters dendritic spine morphology in hippocampal neurons. Cell Logist, 2015. 5(4): p. e1133266. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Boros BD, et al. , Dendritic spines provide cognitive resilience against Alzheimer’s disease. Ann Neurol, 2017. 82(4): p. 602–614. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Ewels P, et al. , MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics, 2016. 32(19): p. 3047–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Dobin A, et al. , STAR: ultrafast universal RNA-seq aligner. Bioinformatics, 2013. 29(1): p. 15–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Love MI, Huber W, and Anders S, Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol, 2014. 15(12): p. 550. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Pepin ME, et al. , Antiretroviral therapy potentiates high-fat diet induced obesity and glucose intolerance. Mol Metab, 2018. 12: p. 48–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Tchivileva IE, et al. , Characterization of NF-kB-mediated inhibition of catechol-O-methyltransferase. Mol Pain, 2009. 5: p. 13. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Hartung JE, et al. , Nuclear factor-kappa B regulates pain and COMT expression in a rodent model of inflammation. Brain Behav Immun, 2015. 50: p. 196–202. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Barski A, et al. , High-resolution profiling of histone methylations in the human genome. Cell, 2007. 129(4): p. 823–37. [DOI] [PubMed] [Google Scholar]
  • 36.Hayashi Y and Majewska AK, Dendritic spine geometry: functional implication and regulation. Neuron, 2005. 46(4): p. 529–32. [DOI] [PubMed] [Google Scholar]
  • 37.Kasai H, et al. , Structural dynamics of dendritic spines in memory and cognition. Trends Neurosci, 2010. 33(3): p. 121–9. [DOI] [PubMed] [Google Scholar]
  • 38.Hering H and Sheng M, Dendritic spines: structure, dynamics and regulation. Nat Rev Neurosci, 2001. 2(12): p. 880–8. [DOI] [PubMed] [Google Scholar]
  • 39.Parnass Z, Tashiro A, and Yuste R, Analysis of spine morphological plasticity in developing hippocampal pyramidal neurons. Hippocampus, 2000. 10(5): p. 561–8. [DOI] [PubMed] [Google Scholar]
  • 40.O’Kane EM, Stone TW, and Morris BJ, Increased long-term potentiation in the CA1 region of rat hippocampus via modulation of GTPase signalling or inhibition of Rho kinase. Neuropharmacology, 2004. 46(6): p. 879–87. [DOI] [PubMed] [Google Scholar]
  • 41.Murakoshi H, Wang H, and Yasuda R, Local, persistent activation of Rho GTPases during plasticity of single dendritic spines. Nature, 2011. 472(7341): p. 100–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Harris EW, Ganong AH, and Cotman CW, Long-term potentiation in the hippocampus involves activation of N-methyl-D-aspartate receptors. Brain Res, 1984. 323(1): p. 132–7. [DOI] [PubMed] [Google Scholar]
  • 43.Frey U, et al. , Anisomycin, an inhibitor of protein synthesis, blocks late phases of LTP phenomena in the hippocampal CA1 region in vitro. Brain Res, 1988. 452(1–2): p. 57–65. [DOI] [PubMed] [Google Scholar]
  • 44.Lee YS and Silva AJ, The molecular and cellular biology of enhanced cognition. Nat Rev Neurosci, 2009. 10(2): p. 126–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Gu Y, et al. , Impaired conditioned fear and enhanced long-term potentiation in Fmr2 knock-out mice. J Neurosci, 2002. 22(7): p. 2753–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Chetkovich DM, et al. , N-methyl-D-aspartate receptor activation increases cAMP levels and voltage-gated Ca2+ channel activity in area CA1 of hippocampus. Proc Natl Acad Sci U S A, 1991. 88(15): p. 6467–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Duman JG, et al. , Mechanisms for spatiotemporal regulation of Rho-GTPase signaling at synapses. Neurosci Lett, 2015. 601: p. 4–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Zalcman G, Federman N, and Romano A, CaMKII Isoforms in Learning and Memory: Localization and Function. Front Mol Neurosci, 2018. 11: p. 445. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Lipsky RH, Epigenetic mechanisms regulating learning and long-term memory. Int J Dev Neurosci, 2013. 31(6): p. 353–8. [DOI] [PubMed] [Google Scholar]
  • 50.Srinivasan M and Lahiri DK, Significance of NF-kappaB as a pivotal therapeutic target in the neurodegenerative pathologies of Alzheimer’s disease and multiple sclerosis. Expert Opin Ther Targets, 2015. 19(4): p. 471–87. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1531648-supplement-1.xlsx (20.5KB, xlsx)
2
NIHMS1531648-supplement-2.pdf (4.8MB, pdf)
3
NIHMS1531648-supplement-3.xlsx (195.3KB, xlsx)

Data Availability Statement

Data included in sequencing analysis are available online from the NCBI Sequence Read Archive underBioProject PRJNA476137. Other data are available from the corresponding author upon reasonable request.

RESOURCES