Fig 2: mH2A1, but not mH2A2 binding at broad domains is dynamically regulated during learning.a Differentially enriched regions (FDR = 0.05) in untrained compared to trained mice were identified using DiffBind analysis of mH2A1 consensus peaks (using epic2 peak caller) 30 min, 1 h and 6 h after fear conditioning. b Differentially enriched regions (FDR = 0.05) in untrained compared to trained mice were identified using DiffBind analysis of mH2A2 consensus peaks (using epic2 peak caller) 30 min or 1 h after fear conditioning. Data are plotted as ChIP FPKM divided by Input FPKM.

Fig 3: Hippocampal depletion of mH2A1 impairs memory more than the depletion of mH2A2.a Validation of mH2A1 (encoded by H2afy gene) and mH2A2 (encoded by H2afy2 gene) depletion in the mouse hippocampus. b During the fear memory training session, all mice show increased freezing after receiving shock, suggesting that learning has occurred in all groups. c Depletion of mH2A1, but not of mH2A2, impairs recall of fear memory 24 h after training (N: shScr = 35, shH2afy = 24, shH2afy2 = 18). d mH2A1 depletion does not affect locomotor activity (N = 10/group). e mH2A1 and mH2A2 depletion each impair memory for Object in Place, but the impairment is greater in mH2A-1 than in mH2A2-depleted mice (n: shScr = 16, shH2afy = 17, shH2afy2 = 18). Data are shown as mean ± SEM. (ns p > 0.05, *p < 0.05).

Fig 4: mH2A1 depletion promotes basal gene expression and blocks learning-induced gene induction.a Experimental design. Mice underwent stereotaxic surgery for intra-hippocampal (CA1) injections of AAV-DJ vectors carrying shRNA constructs against a scramble control sequence, H2afy or H2afy2. After 2 weeks of recovery, the infected region was extracted and RNA was sequenced. b Intra-CA1 depletion of H2afy (encodes mH2A1) alters the hippocampal expression of 2350 genes, majority of which (85%) were upregulated (significance cut-off: p adj < 0.01 and FC > 2). c Intra-CA1 depletion of H2afy2 (encodes mH2A2) alters the expression of 127 genes, 69% of which were upregulated (significance cut-off: p adj < 0.01 and FC > 2). d Experimental design for testing the effects of mH2A1 depletion on training-induced gene expression. Mice underwent stereotaxic surgery for intra-CA1 injections of shScr or shH2afy and were given 2 weeks to recover. After 2 weeks, half of the mice were fear conditioned and half were left undisturbed in their home cage. RNA was extracted from infected tissue and gene expression was compared in trained vs untrained mice for each virus. e In area CA1 of the hippocampus of scramble control mice, fear conditioning altered the expression of 169 genes, 122 of which were upregulated (significance cut-off: p adj < 0.05 and FC > 1.2). f In area CA1 of the hippocampus of mH2A1-depleted mice, fear conditioning altered the expression of 44 genes, 33 of which were upregulated (significance cut-off: p adj < 0.05 and FC > 1.2). n = 3 mice/group.

Fig 5: Learning-induced mH2A1 removal after 30 min occurs on upregulated genes.a Analysis of gene expression 1 h after contextual fear conditioning (n = 3 untrained, 2 trained). Blue and red circles indicate significantly altered genes. 364 of 467 differentially expressed genes (DEGs) were upregulated with fear conditioning. Average profile plot of b mH2A1 and c mH2A2 binding at TSS-flaking regions of upregulated genes. Upon training, mH2A1 signal is significantly reduced on promoters compared to untrained controls. There is no difference in mH2A2 binding on promoters of upregulated genes between trained and untrained mice. Average profile of d mH2A1 and e mH2A2 binding at promoters of randomly selected genes. There is no statistical difference in mH2A1 or mH2A2 binding on randomly selected promoters between trained and untrained mice. f, g Box plots comparing mH2A1 binding RPKM (mean across replicates) normalised to mean input RPKM values at promoters (-2 kb) in trained vs untrained mice on up- and down-regulated genes. Similarly, box plots were generated at h Upregulated DEGs and i Downregulated DEGs for comparison of mH2A2 binding between untrained and trained mice. Only mH2A1 shows a significant decrease in binding in trained compared to untrained samples for upregulated DEGs promoters, while mH2A2 show no significant changes in its binding in trained compared to untrained samples for either up or down regulated DEGs. j Venn diagrams demonstrating significant over-representation (p < 0.01, hyper-geometric test) of reduced mH2A1 signal on promoters (proximal regions) of differentially upregulated compared to promoters of upregulated DEGs. k Venn diagrams demonstrating significant over-representation (p < 0.01, hyper-geometric test) of reduced mH2A1 signal on extended regions (flanking ±100 kb from the centre of the peak) of upregulated compared to downregulated DEGs, implicating distal regulatory elements (e.g., enhancers/silencers). l University of California at Santa Cruz (UCSC) Genome Browser track (mm10 assembly) demonstrating significant decrease in mH2A1 binding 30 min after fear conditioning. Reads are normalised by dividing ChIP by Input. m To validate the learning-induced mH2A1 reduction observed with ChIP-seq, we fear conditioned a separate set of mice (n = 8/group) and measured mH2A1 binding (normalised to H3) 30 min later. As with ChIP-seq, there was a significant reduction in mH2A1 binding at the Arc promoter upon learning. Data are shown as mean ± SEM. (ns p > 0.05, *p < 0.05).