The DNA modification N6-methyl-2'-deoxyadenosine (m6dA) drives activity-induced gene expression and is required for fear extinction

Abstract

DNA modification is known to regulate experience-dependent gene expression. However, beyond cytosine methylation and its oxidated derivatives, very little is known about the functional importance of chemical modifications on other nucleobases in the brain. Here we report that in adult mice trained in fear extinction, the DNA modification N6-methyl-2'-deoxyadenosine (m6dA) accumulates along promoters and coding sequences in activated prefrontal cortical neurons. The deposition of m6dA is associated with increased genome-wide occupancy of the mammalian m6dA methyltransferase, N6amt1, and this correlates with extinction-induced gene expression. The accumulation of m6dA is associated with transcriptional activation at the brain-derived neurotrophic factor (Bdnf) P4 promoter, which is required for Bdnf exon IV messenger RNA expression and for the extinction of conditioned fear. These results expand the scope of DNA modifications in the adult brain and highlight changes in m6dA as an epigenetic mechanism associated with activity-induced gene expression and the formation of fear extinction memory.

Fig. 1.. m6dA is present in the neuronal genome and accumulates in response to neural activation.

(A) Experimental plan to determine whether m6dA is a functionally relevant base modification in neurons. (B) The Dpn1 enzyme cuts DNA specifically at methylated adenine in GATC linker sequences; Dpn1 digestion reveals the abundance of m6dA in DNA derived from primary cortical neurons, but not in DNA from liver (from the left; lanes 1–3: Dpn1 digested DNA from mouse primary cortical neurons, lane 4: Dpn1 digested DNA from mouse liver, lane 5: DNA ladder, lane 6: Dpn1 digested DNA from E. coli, lane 7: undigested DNA from E. coli. (C) LC-MS/MS detects a neuronal activity-induced global m6dA induction (7DIV, 20mM KCl, 7 h, two-tailed, unpaired student’s t test, t=6.411, df=4, **p=0.003, median: KCl-=12.17 p.p.m; KCl+=45.63 p.p.m); Representative LC-MS/MS chromatograms: control compound (m6dA standard) and isolated RNase-treated gDNA samples, which were extracted from primary cortical neurons, were used to directly quantify the global level of m6dA. (D) Dot blot assay shows global accumulation of m6dA in stimulated primary cortical neurons (7DIV, 20mM KCl, 7 h, two-tailed, unpaired student’s t test, t=2.634, df=4, *p=0.02, median: KCl-=1; KCl+=1.406 ). (All n=3 biologically independent experiments /group; Error bars represent S.E.M.Data distribution was assumed to be normal but this was not formally tested.)

Fig. 2.. Experience-dependent redistribution of m6dA deposition within ILPFC neurons that have been activated by extinction learning.

(A) The Venn diagram shows the learning-induced increase in m6dA sites in retention control (RC) and extinction-trained (EXT) mice. (B) There is no obvious difference with respect to the specific accumulation of m6dA from RC and EXT groups within repetitive elements. (C) Metagene plot shows that m6dA deposition is primarily located in the promoter, 5’UTR and CDS regions. (D) Frequency plots demonstrating that m6dA is enriched at +1bp from TSS, and (E) exhibits a significant increase in deposition at +4bp from the start codon. (F) Representative heat map of genome-wide m6dA enrichment within active neurons after behavioral training (All n=3 pooled samples per group; active neurons derived from 5 individual animals were pooled together). (G) Gene ontology analysis of 10 gene clusters associated with m6dA deposition in RC vs. EXT groups. (H) Representative list of genes that exhibit a significant increase in the accumulation of m6dA and that have been associated with synaptic function, learning and memory.

Fig. 3.. Extinction learning-induced accumulation of m6dA positively correlates with gene expression in activated neurons.

(A) Representative heat map of mRNA expression within activated neurons (EXT+) vs quiescent neurons (EXT-)(n= 4 biologically independent animals for EXT+; n = 3 individual animals for EXT-). (B) Gene ontology analysis was performed through DAVID bioinformatic database. GO results shows gene clusters enriched in the upregulated and differentially expressed genes; neuronal activity-related gene clusters are highlighted by red stars. (C) Extinction learning-induced m6dA sites positively correlate with highly expressed genes (n=4 biologically independent animals for EXT+ group was applied; median for each group from expression low to high: 61.398%, 64.445%, 68.769%, 69.844% and 72.297%).

Fig. 4.. N6amt1 mRNA expression is induced in the ILPFC in response to fear extinction learning, and N6amt1 occupancy increases within gene promoters and 5’UTR.

(A) Extinction-learning leads to increased expression of N6amt1 in the ILPFC (n=4 biologically independent animals per group, two-tailed, unpaired student’s t test, t=2.483, df=6, *p=.0476, RC: median=0.8931, data range:0.794 to 0.964 and EXT: median=1.4, data range:0.957 to 1.902). (B) No significant effect of learning on N6amt2 mRNA expression in the ILPFC (n=4 biologically independent animals per group, two-tailed, unpaired student’s t test, t=3.683, df=6, p=.0609, RC: median=0.8906, data range: 0.832 to 0.981 and EXT: median=1.027, data range: 0.982 to 1.068) (Data distribution was assumed to be normal but this was not formally tested). (C-D) N6amt1 protein level is induced post to extinction (n=4 biologically independent animals per group, two-tailed, unpaired student’s t test, t=2.843, df=6, *p=0.0295, RC: median=1, data range: 0.829 to 1.244 and EXT: median=1.64, data range: 1.116 to 1.991, WB Image is cropped) but not N6amt2 (n=4 biologically independent animals per group, two-tailed, unpaired student’s t test, t=0.2906, df=6, p=.7812, RC: median=1 data range: 0.939 to 1.081 and EXT: median=0.9691, data range: 0.748 to 1.236, WB image is cropped) (Data distribution was assumed to be normal but this was not formally tested). (E) N6amt1 distribution across genome. (F) Extinction training-induced increased in N6amt1 occupancy at promoter and 5’UTR regions. (G) There is a positive relationship between N6amt1 deposition and m6dA sites within promoter and 5’UTR region with more genes showing N6amt1 within 0–200bp of m6dA sites.

Fig. 5.. Extinction learning-induced accumulation of m6dA is associated with an active chromatin landscape and increased bdnf exon IV mRNA expression.

Fear extinction learning (EXT), relative to fear conditioned mice exposed to a novel context (retention control: RC), led to (A) increased m6dA at the previously identified GATC site (two-tailed, unpaired student’s t test, t=4.921, df=8, **p=.0012, RC: median=0.354, data range: 0.065 to 0.739 and EXT: median=1.506, data range: 1.184 to 2.271), (B) a selective increase in N6amt1 occupancy (two-tailed, unpaired student’s t test, t=4.133, df=8, **p=.0033, RC: median=0.9053, data range: 0.699 to 1.089 and EXT: median=1.412, data range: 1.052 to 1.617 ), (C) an increased open chromatin structure was detected by using FAIRE-qPCR (two-tailed, unpaired student’s t test, t=3.76, df=8, **p=.0055, RC: median=1.185, data range: 0.282 to 1.780 and EXT: median=4.771, data range: 2.813 to 7.685), (D) a significant increase in H3K4^me3 occupancy (two-tailed, unpaired student’s t test, t=2.986, df=8, *p=.0174, RC: median=0.2164, data range: 0.098 to 0.284 and EXT: median=0.400, data range: 0.252 to 0.447 ), (E) an increase in the recruitment of YY1 (two-tailed, unpaired student’s t test, t=3.885, df=8, **p=.0046, RC: median=0.1044, data range: 0.019 to 0.162 and EXT: median=0.4216, data range: 0.232 to 0.652), (F) an increase in TFIIB occupancy (two-tailed, unpaired student’s t test, t=6.474, df=8, **p=.0002, RC: median=0.2325, data range: 0.100 to 0.314 and EXT: median=0.6321, data range: 0.579 to 0.742), (G) an increase in Pol II occupancy (two-tailed, unpaired student’s t test, t=4.838, df=8, **p=.0013, RC: median=0.1873, data range: 0.056 to 0.345 and EXT: medium=0.5822, data range: 0.416 to 0.803). (H) a significant increase in bdnf exon IV mRNA expression within the ILPFC (two-tailed, unpaired student’s t test, t=2.941, df=8, *p=.0187, RC: median=0.1.104, data range: 0.922 to 1.285 and EXT: median=2.104, data range: 1.650 to 3.188). (All n=5 biologically independent animals per group, Data distribution was assumed to be normal but this was not formally tested).

Fig. 6.. N6amt1-mediated accumulation of m6dA is required for fear extinction memory and for learning-induced bdnf exon IV mRNA expression in the ILPFC.

(A) Left: representative image of cannula placement in the ILPFC, Right: transfection of N6amt1 shRNA into the ILPFC. (B) N6amt1 mRNA expression analysis in Non-FACS and FACS sorted cells, a significant reduction of N6amt1 expression was only observed post to FACS sorted cells (n=5 biologically independent animals per group. Non-FACS sorted groups: two-tailed, unpaired student’s t test, t=1.559, df=8, RC: mean=0.861, data range: 0.461 to 1.211 and EXT: mean=1.709, data range: 0.936 to 1.289. FACS sorted groups: two-tailed, unpaired student’s t test, t=4.956, df=8, **p=.0011, RC: mean=1.128, data range: 0.671 to 1.510 and EXT: mean=0.2537, data range: 0.105 to 0.537). (C) Schematic of the behavioral protocol used to test the effect of lentiviral-mediated knockdown of N6amt1 in the ILPFC on fear extinction memory. (D) There was no effect of N6amt1 shRNA on within-session performance during the first 15 conditioned stimulus exposures during fear extinction training (n=8 biologically independent animals per group, two-way ANOVA, F_1,210=2.539, p=0.1126; Bonferroni’s posthoc test, all section is P>0.9999). (E) Although there was no effect of N6amt1 shRNA on fear expression in mice that had been fear conditioned and exposed to a novel context without extinction training, N6amt1 knockdown led to a significant impairment in fear extinction memory (n=8 biologically independent animals per group, two-way ANOVA, F_1,28=16.9, p<.0001; Dunnett’s posthoc test: Scrambled control RC vs. Scrambled control EXT, **p=.0019, Scrambled control RC: median=40.21, data range: 13.56 to 63.78; shRNA RC: median=46.13, data range: 34.00 to 64.81; Scramble control EXT: median=10.57, data range: 0.00 to 32.07; shRNA EXT: median=49.68, data range: 30.29 to 85.75 ). (F) Schematic of the behavioral protocol used to test the effect of BDNF Injection within N6amt1 knockdown animals on fear extinction memory. (G) ILPFC infusion of BDNF has minimum effect during the section of extinction training (n=5 biologically independent animals per group, two-way ANOVA, F_1,120=105, p<.0001; Bonferroni’s posthoc test: N6amt1 shRNA+Saline vs. N6amt1+BDNF section 4:*p=0.0121 N6amt1 shRNA+Saline: median=67.2, data range: 38.56 to 95.33 and N6amt1+BDNF: median=19.78, data range: 3.67 to 40.22; 10:**p=0.0033 N6amt1 shRNA+Saline: median=74.87, data range: 26.22 to 98.89 and N6amt1+BDNF: median=22.27, data range: 11.00 to 26.67; 11:**p=0.0022 N6amt1 shRNA+Saline: median=67.41, data range: 10.21 to 96.67 and N6amt1+BDNF: median=13.27, data range: 2.44 to 19.20; 12: **p=0.0092 N6amt1 shRNA+Saline: median=68.78, data range: 10.44 to 91.33 and N6amt1+BDNF: median=20.2, data range: 9.99 to 60.82; 13: **p=0.0035 N6amt1 shRNA+Saline: median=76.58, data range: 54.16 to 96.89 and N6amt1+BDNF: median=24.26, data range: 9.99 to 60.82; 14: *p=0.0429 N6amt1 shRNA+Saline: median=71.47, data range: 19.56 to 97.78 and N6amt1+BDNF: median=29.44, data range: 2.11 to 48.67 and 15: *p=0.0254 N6amt1 shRNA+Saline: median=74.38, data range: 39.29 to 95.12 and N6amt1+BDNF: median=30.06, data range: 7.44 to 62.15), and (H) promotes extinction rescues the N6amt1 shRNA-induced impairment in fear extinction memory (n=5 biologically independent animals per group, two-way ANOVA, F_1,17=13.38, p<.01; Dunnett’s posthoc test: N6amt1 shRNA+Saline vs. N6amt1+BDNF, **P=.0052, N6amt1 shRNA+Saline RC: median=48.82, data range: 24.86 to 67.76; N6amt1+BDNF RC: median=57.16, data range: 26.04 to 73.64; N6amt1 shRNA+Saline EXT: median=43.54, data range: 35.96 to 57.16; N6amt1+BDNF EXT: median=13.107, data range: 4.05 to 22.11).( Data distribution was assumed to be normal but this was not formally tested).

Fig. 7.. N6amt1 knockdown prevents the learning-induced accumulation of m6dA and related changes in chromatin and transcriptional landscape associated with the BDNF P4 promoter.

N6amt1 shRNA blocked (A) the learning-induced increase in N6amt1 occupancy (n=6 biologically independent animals per group, two-way ANOVA F_1,20 = 7.663, p<.05; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, **p=.0041, Scramble control RC: median=0.187, data range: 0.082 to 0.280; Scramble control EXT: median= 0.444, data range: 0.308 to 0.594) and (B) the deposition of m6dA (n=5 biologically independent animals in RC Scramble control group; rest with n=6 biologically independent animals per group, two-way ANOVA F_1,19 = 18.56, p<.0001; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, ****p<.0001, Scramble control RC: median=0.107, data range: 0.066 to 0.133; Scramble control EXT: median= 0.314, data range: 0.217 to 0.468), or (C) open chromatin structure (n=4 biologically independent animals per group, two-way ANOVA F_1,12 = 19.38, p<.001; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, ***p=.0001, Scramble control RC: median=1.089, data range: 0.866 to 1.365; Scramble control EXT: median= 3.103, data range: 2.478 to 3.901), (D) the accumulation of H3K4me³ (n=6 biologically independent animals per group, two-way ANOVA F_1,20 = 9.815, p<.01; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, **p=.0046, Scramble control RC: median=0.126, data range: 0.008 to 0.342; Scramble control EXT: median= 0.421, data range: 0.261 to 0.714), (E) YY1 (n=6 biologically independent animals per group, two-way ANOVA F_1,20 = 17.64, p<.001; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, **p=.0018, Scramble control RC: median=0.183, data range: 0.088 to 0.268; Scramble control EXT: median= 0.587, data range: 0.240 to 0.814), (F) induction of TFIIB occupancy (n=6 biologically independent animals per group, two-way ANOVA F_1,20 = 10.03, p<.01; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, *p=.0258, Scramble control RC: median=0.335, data range: 0.191 to 0.421; Scramble control EXT: median= 0.699, data range: 0.347 to 1.296) and (G) RNA Pol II (n=6 biologically independent animals per group, two-way ANOVA F_1,20 = 8.883, p<.01; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, **p=.0024, Scramble control RC: median=0.092, data range: 0.075 to 0.139; Scramble control EXT: median= 0.226, data range: 0.134 to 0.296) occupancy at the proximal GATC site within the BDNF P4 promoter. Also, (H) N6amt1 shRNA inhibited the bdnf exon IV mRNA expression (n=5 biologically independent animals per group, two-way ANOVA F_1,16 = 20.95, p<.001; Dunnett’s posthoc test: scrambled control RC vs. scrambled control EXT, ***p=.0007, Scramble control RC: median=1.127, data range: 0.638 to 1.396; Scramble control EXT: median= 2.117, data range: 1.683 to 2.850). (Data distribution was assumed to be normal but this was not formally tested).