Mapping the epigenomic and transcriptomic interplay during memory formation and recall in the hippocampal engram ensemble

Abstract

The epigenome and three-dimensional (3D) genomic architecture are emerging as key factors in the dynamic regulation of different transcriptional programs required for neuronal functions. In this study, we used an activity-dependent tagging system in mice to determine the epigenetic state, 3D genome architecture and transcriptional landscape of engram cells over the lifespan of memory formation and recall. Our findings reveal that memory encoding leads to an epigenetic priming event, marked by increased accessibility of enhancers without the corresponding transcriptional changes. Memory consolidation subsequently results in spatial reorganization of large chromatin segments and promoter-enhancer interactions. Finally, with reactivation, engram neurons use a subset of de novo long-range interactions, where primed enhancers are brought in contact with their respective promoters to upregulate genes involved in local protein translation in synaptic compartments. Collectively, our work elucidates the comprehensive transcriptional and epigenomic landscape across the lifespan of memory formation and recall in the hippocampal engram ensemble.

Extended Data Fig. 1:. Reactivated cells play a key role in encoding prior experience

(a) Schematic of the Targeted Recombination in Active Populations (TRAP), which requires two transgenes, one that expresses CreERT2 from an activity-dependent Arc promoter (ArcCreERT2) and one that allows expression of the eYFP reporter, in a Cre- dependent manner. Administration of TAM to TRAP mice results in a permanent eYFP label in the initially activated Arc neurons. Without TAM, CreERT2 is retained in the cytoplasm of active neurons in which it is expressed, so no recombination can occur. (b) Representative images and co-localization analysis between endogenous Arc protein (Cyan) and the Arc:eYFP reporter (Magenta), 1.5 h after the FS. Upper panel shows images from the whole hippocampus and lower panel images shows neurons from the DG. Magenta arrows – neurons with only Arc:eYFP reporter signal, Cyan arrows – neurons with only endogenous Arc signal, White arrows – neurons with signal from both populations. Analysis was performed using IMARIS module (co-localization tools) and revealed an average of 84% overlap between the two populations, The scale bar represents 100 μm. n = 3 mice /5 slices per animal, boxplot indicates the mean, interquartile range and the minimum and maximum. (c) Contextual fear conditioning (CFC) freezing test. Percentage of the time freezing levels were measured during habituation (Day 0, Pre-FS) and during the re-exposure (Day 5) to the fear-inducing cue. n = 30 mice, boxplot indicates the mean, interquartile range and the minimum and maximum, two-sided unpaired Student’s t-test, t=15.63, df=16, ****P< 0.0001. (d) Representative images of Activated -late and Reactivated neurons in the hippocampus in two experimental groups; CFC with (TAM) or without (NO-TAM) tamoxifen administration. The scale bar represents 100 μm. (e) Contextual fear conditioning (CFC) freezing test. Percentage of the time freezing levels were measured during the re-exposure to the fear inducing cue (A-A) and during the exposure to a novel neutral environment (context B). n = 15 mice, boxplot indicates the mean, interquartile range and the minimum and maximum, two-sided unpaired Student’s t-test, t=8.506, df=14, ****P< 0.0001. (f) Representative images of Activated -late and Reactivated neurons from the DG in the A-A or A-B group.

Extended Data Fig. 2:. Stable DARs are predominantly enriched for enhancers marks

(a) Workflow for the flow cytometry dissection of different neuronal population from the hippocampus, during memory formation and retrieval. Representative FACS plots showing expression of all population (left panel). Further selection was made on single nuclei and NeuN+/DAPI+ population (middle panel). Last, selection was made on the gated sub-population; GFP+ (adjusted to ~2.5% from all cells), ARC+/GFP+ (~0.15% from all cells) and nuclei were sorted to 1.5 ml Eppendorf tubes coated with 200ul of 1% PBS. (b) Venn Diagram (left) and table (right), which illustrate the overlap between the DARs identified in the different pairwise comparisons during memory formation and recall. (c) Resampling-based statistical analysis was performed to determine if the enrichments of chromatin states over ChromHMM emissions (observed) are statistically significant. Expected enrichment was calculated by performing 10,000 randomized sets of overlaps (permutations) between ‘all accessible sites’ and ‘all histone modifications sites (i.e. all emissions)’ loci and presented as histogram in the figure. Sample size of each randomized set was determined by the size of DARs from each state. The mean and standard deviation of the sample was calculated (Supplementary Table 3). The number of observed overlaps between DARs and each emission was calculated and presented as lines. z-score was calculated as fallows; Z = (observed values (X) – mean of the sample (μ))/(standard deviation of the sample (σ)). z-score Basal vs. Early; S.E 10.5, W.E 6.9. Stable; S.E 16.9, W.E 12.7, all p < 0.0001, z-score Early vs. Late; S.P 91.7, W.P 38.7. Late vs. Reactivated; S.P 26.8, W.P 28.9, all p<0.0001. p-values (Two-Sided) were calculated from z-table. Full analysis is reported in supplement table 3. (d) Pie chart shows percentage of different enhancer states, for all stable regions. Overlap of each individual stable region was performed with previously published H3K4me1 and H3K27ac ChIP-seq data, obtained 1h after FS. Enhancers states were classified as ‘primed’ – overlap with regions marked with only H3K4me1, ‘active’ – with H3K4me1/H3K27ac or ‘latent’ – no overlap. (e) Motifs identified from nucleosome free regions (NFR) on the ATAC-seq tracks from each state (Basal, Activated -early, Activated -late and Reactivated). Peaks were divided into positions that annotated to promoters (5kb from TSS) and enhancers (>5kb from TSS). Circle size indicate percentage of enrichment (1–50%). Color indicates –log(P-value).

Extended Data Fig. 3:. Coordinated priming of the epigenetic state during memory encoding and consolidation facilitates long-range interactions during reactivation

(a) The properties of CHiCAGO-detected interactions in each phase (Basal, Activated -early, Activated -late and Reactivated). Default settings and a score threshold of 5 were used in significant interaction calling, performed jointly on all replicates. (b) Pie chart represent the percentage of all CHiCAGO-detected interactions that are demarcated by either H3K27ac/H3K4me1 (67.5% enhancers marks), H3K4me3/H3K9ac (46.2% promoters mark) or H3K27me3 (1.1% repressive marks). (c) WashU epigenome browser image, encompassing ~ 500 Kb region around the Eif4e2 genes. Arcs shows significant common (red rectangle) and unique (arrowed) enhancers that interacts with promoters (blue rectangle). (d) Examples of interactions called by CHiCAGO. Plots showing all the read counts from bait-other-end (enhancer), within 500–700 kb (upstream and downstream) of the Grink3 and Wwc2 promoters. Significant interactions detected by CHiCAGO (score ≥5) are shown in red, and sub-threshold interactions (3 ≤ score < 5) are shown in blue. Grey lines show expected counts and dashed lines the upper bound of the 95 % confidence intervals. (e) Overlap enrichment analysis between interacting enhancers and DARs, using a permutation procedure on 10,000 randomized sets of accessible sites. Histogram present random sampling distribution of accessible sites for each condition (Basal vs. Activated -early, Activated -early vs. Activated -late, Activated -late vs. Reactivated, Stable). The number of overlapped loci is presented in colored lines from Basal, Activated -early, Activated -late and Reactivated neurons. DARs of BAS vs. Activated -early (Z-score; 7.1, 7.7, 8.5, 10.9). DARs of Activated -early vs. Activated -late (Z-score; −0.1, −0.8, −2.4, −3.2). DARs of Activated -late vs. Reactivated (0.4, 1.8, 0.4, −0.2). DARs of stable (Z-score; 1.7, 2.0, 6.5, 7.3).

Extended Data Fig. 4:. Chromatin changes that occur during the early phase enable transcriptional changes observed at a later time point, primarily in reactivated engram cells

(a) Overlap analysis between gene names from the pair-wise differential analysis and previously published data of: i) activated DG granule cells 1 h after novel exposure ii) 24 h after FS and iii) after prolonged stimulation (6h) of mouse cultured cortical neurons with KCl. Analysis was carried by GeneOverlap R package. (b) Exonic (red) and intronic (blue) reads were quantified separately across all conditions and compared to transcriptional activity as measured by DEseq2. Reads were normalized (RPKM) and the log2FC changes are presented for each state. violin plot indicates the mean, interquartile range and the minimum and maximum, one-way ANOVA (parametric, unpaired), Basal vs. Early; F (5, 248) = 389.9. Early vs. Late; F (5, 2374) = 2183. Late vs. Reactivated. F (5, 1357) = 945.5, All Ps < 0.0001. Bonferroni’s multiple comparisons. n.s = non-significant, ***P < 0.0001. (c) Exon/Intron ratios were measured in each cluster across all conditions (Log2FC scale). violin plot indicates the mean, interquartile range and the minimum and maximum, one-way ANOVA (parametric, unpaired), F (5, 1143) = 260.2, P<0.0001. Bonferroni’s multiple comparisons test. n.s = non-significant, ***P < 0.0001. (d) Overlap analysis between DARs and DEGs during different memory phases. DARs on Intergenic and introns regions were mapped to their respective genes with the pc-HiC interaction maps. Overlap analysis was carried by GeneOverlap R package. P-value (numbers) and odds ratio (color) from Fisher’s exact test are presented in the heatmap. n.s – not significant. (e) Pearson correlation between log2FC values of DARs and log2FC DEGs that were annotated to those region (intergenic regions were mapped via the pc-HiC data set). Chromatin accessibility changes were compared with parsed exonic reads (red line), intronic reads (blue lines) and total transcriptional changes (both intronic and exonic reads) as measured by Desq2 (gray line). All r and p-values are reported in supplementary table 9.

Extended Data Fig. 5:. Transcriptional changes in the activated-late neurons correlated higher with intronic reads and reactivated neurons presented higher correlation with exonic reads

(a) Exonic (red) and intronic (blue) reads were quantified separately across all conditions for each of the clusters identified in Fig 5B. Reads were normalized (RPKM) and the log2FC changes are presented for each cluster. (b) Exon/intron ratios were measured in each cluster across all conditions. Violin plot indicates the mean, interquartile range and the minimum and maximum, n = 3 biologically independent samples one-way ANOVA (parametric, unpaired), Dw –Late cluster; F (3, 968) = 139.4, P<0.0001. Up -Late cluster; F (3, 734) = 15.95, P<0.0001. Stable -cluster; F (3, 652) = 93.97, P<0.0001. Reactivation -cluster; F (3, 1600) = 485.2, P<0.0001. Bonferroni’s multiple comparisons test to Deseq2 reads. ***P < 0.0001.

Extended Data Fig. 6:. Distinct temporally transcriptional programs are being synchronized to maintain neuronal excitability, structural changes and protein translation in synapses in the engram ensemble

(a) Schematics representation of the experimental design. Three weeks prior to the CFC test, ArcCreERT2 mice were bilaterally injected to the DG with AAV9-EF1a-DIO-hChR2-EYFP. In a similar manner to the TRAP system, eYFP reporter is only expressed in a Cre- dependent manner in the presence of tamoxifen. In the right panel, representative IHC images of the DG. Green – AAV-eYFP, Red – endogenous Arc. The scale bar represents 50 μm. (b) Spines morphology assessment during different memory phases. Right panel shows a single eYFP+ dendritic shaft with different types of spines (Stubby, Thin, Mushroom, Enlarged mushroom). The scale bar represents 5 μm. boxplot indicates the mean, interquartile range and the minimum and maximum, Activated -early: n = 4 mice /5 section per animal, Activated -late: n = 4 mice /4 section per animal, Reactivated: n = 4 mice /2 section per animal. one-way ANOVA (parametric, unpaired), Stubby; F (2, 36) = 2.313, P=0.1135. Thin; F (2, 36) = 35.12, P<0.0001. Mushroom; F (2, 36) = 38.42, P < 0.0001. Bonferroni’s multiple comparisons test, ***P < 0.0001. (c) Representative IHC images and quantification of the protein levels of two members of the EIF family; (left) Eif2a and (right) Eif3e. The scale bar represents 10 μm. Data for dendritic shaft is presented as a ratio between number and the length (μM). n = 4 mice /5 section per animal, boxplot indicates the mean, interquartile range and the minimum and maximum, one-way ANOVA (parametric, unpaired) with Bonferroni’s multiple comparisons test, n.s - not significant, Eif2a Shaft; F (2, 20) = 4.484, P=0.0246. Soma; F (2, 21) = 19.58, P < 0.0001. (Activated -early vs. Activated -late *P = 0.0142, Activated -early vs. Reactivated *P < 0.0001, Activated -late vs. Reactivated *P = 0.0303). Eif3e Shaft; F (2, 14) = 1.983, P = 0.1745. Soma; F (2, 23) = 8.309, P = 0.0019, (Activated -early vs. Reactivated *P = 0.0057, Activated -late vs. Reactivated *P = 0.0055). (d) Pie chart present the percentage of enlarged mushroom spines (Dh ≥ 3Dn) and mushroom spines from Activated -late and Reactivated neurons. (e) Representative images (left panel) and quantification (right panel) of Gria1 mRNA levels, during different phases of memory. Data is presented as a ratio between number of puncta and the dendritic shaft length. The scale bar represents 10 μm. N = 4 mice /5 section per animal, boxplot indicates the mean, interquartile range and the minimum and maximum, Shaft; one-way ANOVA (parametric, unpaired) F (2, 15) = 10.41, P = 0.0015. Bonferroni’s multiple comparisons test, **P = 0.0011. lower panel - Soma; one-way ANOVA F (2, 12) = 0.13, P = 0.88.

Extended Data Fig. 7:. Interactions with distinct combinatorial enhancers leads to a directional change in gene expression.

(a) Venn diagrams shows percentage of overlap between chromatin accessibility (DARs) across all memory phases (BAS vs. Activated -early, light green grid circle; Activated -early vs. Activated -late, dark green grid circle; Activated -late vs. Reactivated, orange grid circle) and total transcriptional changes from all identified clusters (Dw –late, Up –late, Stable, Reactivation, blue grid circle). Intergenic and introns DARs were mapped to their respective genes with the pc-HiC interaction maps. Percentage of overlap was calculated from all identified DEGs in the clusters (n = 1095). (b) Overlap analysis between DARs (pairwise) and DEGs from each cluster. Intergenic and introns DARs were mapped to their respective genes with the pc-HiC interaction maps. Overlap analysis was carried by GeneOverlap R package. P-values and Jaccard values (color) from Fisher’s exact test are presented on the heatmap (left). Percentage of overlap was calculated from all identified DEGs in the clusters (right). (c) Representative image of chromatin and transcriptional changes of the Gabrb3 locus from the Dw-late cluster. While early state interactions were between promoter and enhancers with transcriptional activators (Ap1), late state interaction were with transcriptional repressors (Slug). Upper IGV genome browser tracks (purple - Basal, light green - Activated -early, dark green - Activated -late and orange - Reactivated) presenting transcriptional changes (ncRNA-seq), middle tracks shows chromatin accessibility dynamics (ATAC-seq) on promoter (red rectangle) and enhancers (gray rectangle). Significant promoter-enhancer interaction are represented as arcs (WashU browser tracks). Lower track present motifs that were identified via HOMER tools (Slug, Ap1 and Rest). (d) Aggregation plots for individual motifs. The enrichment values (motifs per bp/ per peak) of six selected motifs (two repressors, two activators and two bivalent) was assessed around the center of peaks (−/+ 4000bp) from each cluster.

Extended Data Fig. 8:. Proposed model of the chromatin accessibility, promoter-enhancer interaction and transcriptional dynamics of hippocampal memory engram neurons

Over the course of memory formation and recall. Basal state - gene promoters interact with enhancers that carry transcriptional repressor cargo and express low levels of mRNA. Early phase – leads to a priming event, in which enhancers that harbor transcriptional activator cargo become more accessible, but most of them remain isolated and lack interactions with respective gene promoters. Late phase - gene promoters shift their interaction to the primed regions, which harbor transcriptional activator motifs. This promoter-enhancer reprograming results in increased gene expression that presumably allows the stabilization of the memory. Recall - reactivated engram neurons utilized a subset of primed promoter-enhancer interactions, which is associated with transcriptional changes involved in mRNA transport to synaptic compartments and protein translation. Transcription factor – TF, E(1–3) – different enhancer that interact with the same promoter, Red - transcriptional repressors, Blue - transcriptional activators.

Fig. 1.. Temporal and spatial identification of activated and reactivated engram cells.

(a) Schematic representation of the experimental design. TRAP mice were initially exposed to training in which a conditioned (context) and unconditioned (foot shock, FS) stimulus were paired together within a brief window of 4-Hydroxytamoxifen (TAM) exposure. Four different neuronal populations (Basal, Activated –early, Activated –late and Reactivated) were sorted by flow cytometer and subjected to library preparation for nRNA-seq, ATAC-seq, Hi-C and pc-HiC. (b) Representative images (DG) and whole hippocampus quantifications of eYFP+ neurons after TAM treatment in home cage (naïve, n = 4 mice/6 slices per animal), 1.5 h (Activated –early, n = 4 mice/4 slices per animal) and 5 days (Activated –late, n = 4 mice/8 slices per animal) after exposure after FS. The scale bar represents 200 μm. The dot-plot indicates the mean and error bars (SEM), one-way ANOVA (parametric, unpaired), F (2, 70) = 240.3, P<0.0001. Bonferroni’s multiple comparisons test; HC vs. Early ***P = 0.0007, HC vs. Late ****P < 0.0001, Early vs. Late ****P < 0.0001. (c) Quantifications of Activated –late and Reactivated neurons in the hippocampus in two experimental groups; CFC with (CFC+TAM, n = 4 mice/8 slices per animal) or without (CFC NO-TAM, n = 4 mice/6 slices per animal) tamoxifen administration. The dot-plot indicates the mean and error bars (SEM), two-sided unpaired Student’s t-test, Activated –late; t=25.65, df=47, Reactivated; t=11.00, df=52, ****P < 0.0001. (d) Representative images and quantifications from different regions of the hippocampus. Number of positive nuclei were analyzed using IMARIS (see methods) from the DG, CA1 and CA3. The scale bar represents 100 μm. n = 5 mice/ Activated –early: 6 slices per animal, n = 6 mice/ Activated –late: 6 Reactivated: 6 slices per animal. The dot-plot indicates the mean and error bars (SEM), one-way ANOVA (parametric, unpaired), Early; F (2, 42) = 13.46. Late; F (2, 95) = 556.3. Reactivated; F (2, 89) = 83.53. All Ps < 0.0001. Bonferroni’s multiple comparisons test, *P= 0.029, **P= 0.0075, ***P= 0.0009, ****P< 0.0001. (e) Valuation of context effects on the number and the distribution of hippocampal memory traces. (f) Quantifications of Activated -late and Reactivated neurons in the A-B (n = 6 mice/ 6 slices per animal), compared to the A-A (n = 5 mice/ 6 slices per animal) group. The dot-plot indicates the mean and error bars (SEM), two-sided unpaired Student’s t-test, Activated –late; t=0.08292, df=59, P = 0.9342. Reactivated; t=8.385, df=57, ****P< 0.0001. n.s = non-significant.

Fig. 2.. Chromatin accessibility landscape during memory formation and recall.

(a) Differential accessible regions (DAR)s are shown in a volcano plot (Pairwise comparisons, FDR < 0.01; fold changes > 1.5), n = 3 biologically independent samples. (b) Analysis of common gained- accessible region (stable), acquired 90 min after the FS (Activated -early), that remain accessible in the late state of consolidation (Activated -late) and recall (Reactivated). Overlapping stable DARs were only considered if they appeared in at least 2 biological replicates in a given condition. (c) Representative IGV genome browser tracks of the Arc locus. (d) Annotation analysis (HOMER tools) for all the DARs. Numbers on the y-axis indicate the Log2 ratio values of observed loci (number of DARs) over expected (total size (bp)) of the TTS, exons, introns, intergenic and promoters. P-value were calculated by HOMER tools; Basal vs. Activated -early (Intergenic *P = 1.73 × 10⁻¹⁰³), Activated -early vs. Activated -late (TTS *P = 0.0339, Promoter *P = 1.48 × 10⁻³⁹), Activated -late vs. Reactivated (Promoter *P = 2.48 × 10⁻⁶), stable (Intergenic *P = 0.0358). (e) Histone modification emission states (left) and enrichment analysis of DARs over emissions using ChromHMM. Fold enrichment scale (FE) 0–30. The and p-value was determined by the z-score value (as calculated in Extended Data Fig. 2c), Full analysis is reported in supplement table 3. *P < 0.00001, reported only for enriched values (not depleted) (f) Chip-qPCR. Samples were immunoprecipitated with antibodies against H3K27ac or H3K4me1 and subjected to real-time PCR with two sets of primers for identified putative enhancers. H3K27ac: n = 4, H3K4me1 n = 3, The dot-plot indicates the mean and error bars (SEM), one-way ANOVA (parametric, unpaired) with Dunnett’s multiple comparisons test, Enhancer 1; H3K27ac - F (3, 12) = 4.664, P = 0.02, (Basal versus Activated-late *P = 0.0246, Basal versus Reactivated *P = 0.0232), H3K4me1 - F (3, 8) = 0.845, P = 0.506. Enhancer 2; H3K27ac - F (3, 12) = 5.221, P = 0.0155, (Basal vs. Activated -late **P = 0.0059), H3K4me1 - F (3, 8) = 1.719, P = 0.24. Enhancer 3; H3K27ac - F (3, 12) = 29.91, P < 0.0001, H3K4me1 - F (3, 8) = 10.39, P = 0.0039, (Basal vs. Activated -early *P = 0.0298, Basal vs. Activated -late **P = 0.0021, Basal vs. Reactivated **P = 0.0063). Enhancer 4; H3K27ac - F (3, 12) = 6.273, P = 0.0083, (Basal vs. Activated -late **P = 0.007, Basal vs. Reactivated *P = 0.0152), H3K4me1 - F (3, 8) = 9.901, P = 0.0045, (Basal vs. Reactivated **P = 0.0016).

Fig. 3.. Sub-compartment switching across the different phases of memory formation and recall.

(a) Hi-C correlation heatmap represent compartmentalization of chr6 for Basal, Activated -early and Activated -late cells. Compartments calculated by Juicer at 500 kb resolution (observed–expected normalized). (b) Dynamic changes in compartment state. Positive values of the first component (+PC1) reflecting active chromatin and negative values (-PC1) indicative of inactive chromatin state were compared between all three populations of neuron (Basal, Activated -early and Activated -late). Compartment switch was only considered if a negative value was transformed to a positive value, and vice versa. (c) Representative images from IGV browser that shows stable compartment switch from A to B (upper panel) and B to A (lower panel) across different memory phases. (d) Overlap analysis between DARs in the early phase (Basal vs. Activated –early) and stable switched compartment. Left Venn diagram shows gained DARs and stable compartment switched from B to A. Right Venn diagram shows lost DARs and stable compartment switched from A to B. (e) Illustration of activity induced compartment switch and specific promoter-enhancers interactions.

Fig. 4.. Promoters interact more frequently with a distinct subset of enhancers during each memory phase.

(a) Heatmap represent the interaction score as defined by CHiCAGO, for each population. Basal: n = 3, Activated -early, Activated -late and Reactivated: n = 4 biologically independent samples. (b) Violin plots show the distribution and the intensity (log10 scale) of interaction that were either unique or shared by at least 3 populations (common). One-way ANOVA (parametric, unpaired), F (4, 57922) = 2364, P < 0.0001. Bonferroni’s multiple comparisons test, n.s = non-significant (P = 0.7659), *P < 0.0001. (c) Frequency of promoter-enhancer interaction as measured by 3C-qPCR for selected loci. Four sets of primers encompassing interaction between promoter-enhancer (P/E), and three random loci inside the loop (P/Ein), outside the loop (P/Eout) and on the promoter (P/Pin). Reads were normalized to promoter-enhancer interaction values of the house keeping gene βActin. n = 3 biologically independent samples. Data are presented as mean values with error bars indicate the SEM, two-way ANOVA. Eif3d; population × interaction location (F (9, 32) = 1.917, P=0.0851), effect of interaction location only (F (3, 32) = 22.64, P < 0.0001). Grik3; population × interaction location F (9, 32) = 0.1716. Bonferroni’s multiple comparisons test, P/E Basal vs. Reactivated **P = 0.0019. (d) Representative priming events of the Eif5a locus. IGV genome browser tracks presenting chromatin accessibility dynamics (colored tracks) on promoter (gray rectangle) and enhancers (blue rectangle). WashU browser tracks presenting significant promoter-enhancer interaction via arcs (purple - Basal, light green - Activated -early, dark green - Activated -late and orange - Reactivated).

Fig. 5.. Transcriptional signature of reactivated engram neurons

(a) Pairwise analysis of DEGs from the hippocampus of TRAP mice that were exposed to CFC are shown in a volcano plot (FDR<0.01; log2FC > 2). n = 3 biologically independent samples. (b) Heatmap and (c) Line graphs shows the average transcriptional changes (log2 scale), compared to the expression at the non-activated neurons (basal). Gene-sets were clustered by K-means to i) Dw–Late, ii) Up-Late, iii) Stable and iv) Reactivation clusters. Mean of n = 3 biologically independent samples; P value calculated using one-way ANOVA (parametric, unpaired), followed by Bonferroni’s multiple-comparison test. Dw–Late; F (3, 1032) = 42.19, Up-Late; F (3, 280) = 752.5, Stable; F (3, 764) = 92.28, Reactivation; F (3, 1744) = 597.4. All Ps < 0.0001. multiple comparisons test, n.s = non-significant, ***P < 0.0001. (d) Gene ontology (GO) analysis (ToppFun) for DEGs that were identified in each cluster. (e) Representative DEGs from each cluster. log2FC from the basal-state is presented in three biological replicates. Line represent calculated average. n = 3 biologically independent samples, one-way ANOVA (parametric, unpaired) with Dunnett’s multiple comparisons test, n.s = non-significant. Gad1; F (3, 8) = 4.387, P= 0.04. *P = 0.0268. Gabrb3; F (3, 8) = 6.184, P=0.0177. *P = 0.0441. Rest; F (3, 8) = 23.23, P=0.0003. ***P = 0.0268. Map4k4; F (3, 8) = 6.244, P=0.0172. *P = 0.0151. Nefl; F (3, 8) = 12.06, P=0.0024. **P = 0.0059, **P = 0.0014. Nefh; F (3, 8) = 21.50, P=0.0003. *P = 0.016, ***P = 0.0008, ***P = 0.0002. Eif5a; F (3, 8) = 8.482, P=0.0072. **P = 0.004. Eif4h; F (3, 8) = 26.58, P=0.0002. ***P = 0.0001. (f) Representative images and quantification of Eif4e protein levels from Activated -early, Activated -late and Reactivated. For quantification, two separate surfaces were built on IMARIS; cell soma and dendritic shaft. The number of Eif4e puncta were measured only in those surfaces boundaries. Data for dendritic shaft is presented as a ratio between number and the length (μM). The scale bar represents 10 μm. n = 4 mice /5 section per animal, boxplot indicates the mean, interquartile range and the minimum and maximum, one-way ANOVA (parametric, unpaired), Shaft; F (2, 28) = 5.113, P=0.0128. Soma; F (2, 43) = 3.026, P=0.0589. Bonferroni’s multiple comparisons test; Shaft; Activated -early vs. Reactivated *P = 0.0324, Activated -late vs. Reactivated *P = 0.0324 (g) Representative images (left panel) and quantification (right panel) of the Gria1 protein levels, during different phases of memory. Three populations of neurons (Activated -early, Activated -late and Reactivated) were analyzed, based on the immunoreactivity of eYFP and endogenous Arc. For quantification, IMARIS surfaces were built on the dendritic shaft and the number of GRIA1 puncta were measured only in the surfaces boundaries (<1μm threshold). Data is presented as a ratio between number of puncta and the shaft length. The scale bar represents 10 μm. n = 4 mice /5 section per animal, boxplot indicates the mean, interquartile range and the minimum and maximum, one-way ANOVA, F (2, 16) = 15.22, P = 0.0002. Bonferroni’s multiple comparisons test, **P = 0.002, ***P = 0.0005.

Fig. 6.. Sequential reprograming of chromatin accessibility, promoter-enhancer interactions and gene expression over the course of engram formation.

Line graphs for each K-means cluster shows the average (log2 scale) of (a) transcriptional changes (p-values as reported in Fig. 5c, mean of n = 3 biologically independent samples) (b) chromatin accessibility (normalized RPKM values, mean of n = 3 biologically independent samples) and (c) promoter-enhancer interaction scores (as calculated by CHiCAGO, mean of Basal: n = 3, Activated -early, Activated -late and Reactivated: n = 4 biologically independent samples) Repeated measures one-way ANOVA (between groups). chromatin accessibility; Dw-Late; F (2.359, 5773) = 917.8. Up-Late; F (2.519, 4590) = 638.7. Stable; F (2.437, 4504) = 717.9. Reactivation; F (2.434, 8481) = 1425. All Ps < 0.0001. Interaction scores; Dw-Late; F (2.833, 7054) = 46.86. Up-Late; F (2.94, 5442) = 14.87. Stable; F (2.91, 5702) = 21.5. Reactivation; F (2.86, 11188) = 44.94. All Ps < 0.0001. Bonferroni’s multiple comparisons test, ***P < 0.0001. (d-e) Characterization of motif enrichment on interacting enhancers, in each cluster. ChromHMM was used to build the emission states using HOEMR motif data base (Egr1, Ap1, Slug, Plagl1, Hic2, Rest, CTCF, Yy1, Mecp2) and previously published ChIP-seq data (Ap1-Junb, CTCF, Mecp2). Four states were identified as i) control – random genome sampling, ii) strong repressors iii) bivalent iv) strong activator. Fold enrichment scale (FE) 0.15–20.