Part 2:Mapping The Epigenomic And Transcriptomic Interplay During Memory Formation And Recall in The Hippocampal Engram Ensemble
Mar 15, 2022
for more information:ali.ma@wecistanche.com
Distinct transcriptional programs of the engram ensembles during memory formation and recall
To elucidate how changes in chromatin accessibility and promoter-enhancer interactions were affecting gene expression, we performed nuclear RNA-seq (nRNA-seq) from all the neuronal subpopulations. Our differentially expressed gene (DEG) analysis revealed a low amount of changes in the early phase when compared to basal-state neurons (Basal vs. Early, 88 DEGs, Fig. 5a, Supplementary Table 8). This was especially unexpected given the extensive degree of chromatin organization observed at this time point (Fig. 2a). In contrast, neurons from the late phase presented higher numbers of DEGs despite there being more similarity in chromatin accessibility at these time points (Fig. 5a). To validate our transcriptional findings, we compared our results to previously published data of activated DG granule cells 1 h after novel exposure32, 24 h after FS20, and after prolonged stimulation (6h) of mouse cultured cortical neurons with KCl33 (Extended Data Fig. 4a). Interestingly, although this comparison revealed significant overlap between up-regulated genes from our data set and 1 h after novel exposure32 (P = 0.006) or 24 h after FS20 (P = 0.0005), we did not observe any significant overlap with genes that were up-regulated in the reactivated engram cells and the other data sets, confirming the unique transcriptional landscape of reactivated engram cells during recall.
Click to Cistanche benefits and side effects for memory
Next, we took advantage of the previously published computational approach34 that measures changes in mature RNA (exons) and pre-mRNA (introns) reads. As expected, exonic changes correlate with global transcriptional changes as measured by DEseq2 (both exon and intron reads), across all memory phases (Extended Data Fig. 4b). However, measurement of the exon/intron ratios, indicates that altered transcripts (up-and down-regulated) during memory consolidation (Early vs. Late) are predominantly driven by intronic reads, suggesting that this transcriptional profile mostly reflects immature transcripts that require additional steps of RNA processing (Extended Data Fig. 4c). In contrast, up-regulated genes during recall (Late vs. Reactivated) are predominantly driven from exonic transcripts, thus reflecting a more mature state of RNA (Extended Data Fig. 4c).
To reconcile the discrepancy between the observed chromatin changes and the transcriptional state in each subpopulation, we performed correlation analysis of the DARs and DEGs, identified during the different time points of memory formation and recall. For this analysis, DARs on the non-coding regions (intergenic and intronic) were mapped to their respective promoters by utilizing our pc-HiC data. Overall, we found a low correlation (P = 2.81 × 10−2) between differential chromatin accessibility and gene expression changes that occur in the early phase (Basal vs. Early, Extended Data Fig. 4d). These results were in line with the previous work21, reporting that histone modifications have little correlation with transcriptional alterations during memory acquisition. Unexpectedly, we observed a stronger overlap and similarity between the chromatin changes that occur during memory encoding (Early) with the altered transcriptional landscape during memory consolidation (Late) and reactivated engram neurons (P = 1.13 × 10−8, 2.42 × 10−7, respectively, Extended Data Fig. 4d). To further confirm this analysis, we performed a Pearson correlation between log2FC values from DARs and log2FC values of DEGs that were annotated to that region (intergenic regions were mapped via the pc-HiC data set, Extended Data Fig. 4e, Supplementary Table 9). Our analysis revealed that early chromatin changes (Basal vs. Early, DARs) were significantly correlated with down-regulation of genes in the late phase (Early vs. Late, Extended Data Fig. 4e, gray line, r = −0.14, p < 0.0001). Moreover, after parsing the DEGs to intronic and exonic reads we observed that the correlation was largely driven by intronic reads (Introns r = −0.21, p < 0.0001, Exons r = −0.09, p = 0.02). In contrast, early chromatin changes (Basal vs. Early, DARs) showed positive correlation with DEGs from the reactivated cells, which was predominantly driven by exonic reads (Introns r = −0.007, p = 0.73, Exons r = 0.11, p = 0.04). This data provides further evidence for a priming event and transcriptional lag, in which chromatin changes that occur during the early phase enable transcriptional changes observed at a later time point, primarily in reactivated engram cells.
Memory encoding, consolidation, and recall is a dynamic process requiring several waves of delayed gene transcription12,32,33. Hence, to identify different transcriptional programs across all memory phases, we assessed the expression of each gene from non-activated neurons (Basal) through early (Activated early), late (Activated-late), and neurons that were reactivated during recall (Reactivated). All differential expressed genes (DEGs, n = 1095) that met the cutoff criteria (p-value < 0.01, Log2(fold change) > 1) were unbiasedly clustered (k-mean) (Fig 5b,c). Four major clusters were identified; i) 260 genes
that were down-regulated in the late phase (Dw-late cluster), ii) 207 genes that were up-regulated in the late phase (Up-late cluster), iii) 186 genes that were up-regulated in the early phase, and were sustained throughout all memory phases (Stable cluster) and iv) 438 genes
that were uniquely up-regulated in the reactivated neurons, during memory recall (Reactivation cluster). Similar to our pairwise comparison, transcriptional changes in the activated-late neurons correlated higher with intronic reads, and the transcriptional landscape of reactivated neurons seems to be more mature and presented a higher correlation with exonic reads (Extended Data Fig. 5a,b).
We next performed a gene ontology (GO) analysis for each cluster (Fig. 5d), to identify enriched pathways within different memory phases. The transcriptional signature of-late cluster was enriched for genes involved in regulating ‘Inhibitory synapse function’, which have been shown to play a role in neuronal excitability35,36 (Fig. 5e). In the Up-late cluster, we observed genes sets involved in positive or negative regulation of transcription, such as the v-maf musculoaponeurotic fibrosarcoma oncogene family, protein B (Mafb), snail family zinc finger 2 (Snai2), and RE1-silencing transcription factor (Rest) (Fig. 5e). Interestingly, we found that 15% (32/207) of the gene’s promoter in Dw –late cluster harbor either Snai2 (p-value - 1 × 10−9) or Rest (p-value - 1 × 10−3) binding sites. This dynamic of transcriptional regulators points toward a possible interplay between the Dw- and Up-late clusters. Notably, it was found that functionally distinct neuronal ensembles exist within a single memory engram, in which FBJ osteosarcoma oncogene (Fos)- and neuronal PAS domain protein 4 (Npas4)-dependent ensembles undergo distinct synaptic modifications after CFC and drive memory-guided behaviors in opposite directions37. It was shown that excitatory synaptic inputs promote memory generalization and inhibitory ones promote memory discrimination. Although these distinct ensembles could not be differentiated in the current study, these observations might explain the bi-directional expression of positive or negative transcriptional regulators, as well as up and down-regulated genes involved in neuronal excitability.
Cistanche can improve memory
The Stable-cluster was mostly enriched for transcripts involved in cytoskeletal, axon remodeling, and hippocampal-dependent memory38, including the light and heavy polypeptide neurofilaments (Neal, Neath) and myosin, heavy polypeptide 10, non-muscle (Myh10) (Fig. 5d,e). To visualize and track cytoskeletal-related changes on dendrites and spines, loop-eYFP reporter (TAM dependent) was delivered by adeno-associated virus (AAV) into to the DG ofArc-CreERT2 mice (Extended Data Fig. 6a). Our data revealed that Activated-late and Reactivated neurons harbor significantly higher numbers of mushroom and thin spines (P< 0.0001, Extended Data Fig. 6b), with no significant changes in dendritic shaft diameter (P=0.6), when compared to the Activated -early neurons.
The unique transcriptional signature of reactivated engram neurons (Reactivation cluster) was mostly characterized by categories of ‘RNA transport/binding’ and ‘protein translational mechanisms’ (Fig. 5d). This transcriptional program includes elevated expression of kinesin family members 5a and c (Kif5a, Kif5c), eukaryotic translation initiation (Eif) factor 3 subunit A, D (3a, 3d), factor 4 subunit H (Eif4h), factor 5 subunit A (Eif5a) and ribosomal protein (Rpl27a, Rpl35, Rpl36, Rps19, Rps24, Rosa) (Fig. 5d,e), that were previously reported to play a crucial role in the transport of mRNA to the synaptic compartment and autonomous control of local translation39. To further validate these findings at the translational level, we measured protein expression and localization of three members from the Eif-family by immunohistochemistry (IHC). Serval Eif-members was previously reported to be specifically involved in translation in synaptic compartments39, hence, we separately analyzed protein expression on the soma and the dendrite shafts of activated neurons (Fig. 5f). Reactivated neurons had higher expression ofEif3e proteins in the soma, as compared to the Activated-early and -late neurons (Extended Data Fig. 6c). In contrast, Eif4e was significantly more abundant in the shafts of the Reactivated neurons, when compared to Activated-early and -late neurons (Ps<0.05, Fig. 5f), providing further evidence for increased protein synthesis in those neurons. Cistanche can improve memory.
Increased translation machinery facilitates alterations in spine shape, size, and number, resulting in enhanced synaptic plasticity39–41. Furthermore, spines with larger head diameters are more likely to contain ribosomes for local protein synthesis, larger postsynaptic densities, which anchor more AMPA glutamate receptors and make these synapses functionally stronger41,42. Consistent with those reports, Reactivated neurons presented a higher number of enlarged head diameters (51% from all mushroom spines), as compared to the Activated -late (18%) group (Extended Data Fig. 6d). Moreover, we found a significant increase in both glutamate receptor 1, ionotropic, AMPA1 (Gria1) mRNA (Extended Data Fig. 6e), and Gria1 protein levels only in dendritic shafts of the Reactivated neurons, as compared to the Activated -early and -late neurons (Fig. 5g). Collectively, this data supports the hypothesis that distinct temporally -separated transcriptional programs are being synchronized to maintain neuronal excitability, structural changes, protein translation, and synaptic plasticity, across the lifespan of the engram ensemble.
Regulatory elements with opposing transcriptional effects interact with promoters as memory progresses from encoding to recall
We next asked how long-range promoter-enhancer interaction and chromatin state could lead to the coordinated gene expression changes observed in each of the transcriptional clusters. To achieve that, we measured chromatin accessibility over enhancers that interact with promoters, from each transcriptional cluster (Fig. 6a–c). In addition, the interaction score of those enhancers was plotted together against chromatin accessibility and their respective gene expression (Supplementary Table 10).
Cistanche can improve memory
Overall, all clusters showed a significant incremental increase in chromatin accessibility on enhancers. Importantly, we found that chromatin accessibility occurred before 3D - chromatin rearrangement and gene expression alteration, providing further support of a transcriptional lag (Fig. 6a–c, Extended Data Fig. 7a,b). Specifically, in the Reactivation- cluster, our analysis revealed a biphasic process, in which the early phase of memory formation leads to a priming event, where enhancers become more accessible but fail to show expected changes in gene expression. Reactivation of the engram cells was associated with the reorganization of promoter-enhancer interactions, where primed enhancers were brought in contact with their respective promoters in a higher frequency, and a transcriptional surge was observed (Fig 6a–c). Cistanche can improve memory.
It was previously suggested that promoter-enhancer interactions are highly dynamic, and interactions with distinct combinatorial enhancers lead to a directional change in gene expression, dependent on whether the enhancer harbors a repressor or activator motifs14, 18.Cistanche can improve memory.
To further probe this, we used chrome to unbiasedly segregate the genome into different states defined by motif locations that bind either transcriptional activators1 (early growth response 1 (Egr1), Ap1), transcriptional repressors43,44 (pleiomorphic adenoma gene-like 1 (Plagl1), hypermethylated in cancer 2 (Hic2), methyl CpG binding protein 2 (Mecp2), Snai2 and Rest,) or TFs that bi-directionally impact transcription and involved in the 3D-genome architecture25,44,45 (CCCTC-binding factor (CTCF), E2F transcription factor 1 (E2f), Yy1). This identified three different combinatorial states of TF-binding sites (strong repressor, bivalent and strong activator, Fig. 6d,e). Interestingly, the cluster with a reduction in gene expression (Dw-late) showed a progressively increasing correlation from encoding to retrieval with enhancers that harbor transcriptional repressors and bivalent motifs (Fig. 6d, upper heatmap; Extended Data Fig. 7c,d). This data indicates that despite the observed chromatin accessibility, genes are down-regulated due to repression that is emerging from these regulatory elements. Cistanche can improve memory.
In contrast, in the Reactivation cluster, we observed an increment enrichment for enhancers that harbor transcriptional activators together with motifs that anchor loop structures (Extended Data Fig. 7c,d). Hence, it appears that upon the neuronal activity, several promoters can utilize a wide range of different enhancers to establish novel promoter-enhancer interactions that can differentially coordinate gene expression. In this model, the promoters of down-regulated genes interact with enhancers that harbor transcriptional repressors, while up-regulated genes are the result of promoter-enhancer interactions that bring transcriptional activators in close contact with respective promoters. Collectively, our data reveal a sequential reprogramming of chromatin accessibility, 3D-genome architecture, and promoter-enhancer interactions throughout engram formation (Extended Data Fig. 8).
