Introduction

Cognitive development is profoundly influenced by rearing conditions. For instance, after Donald O. Hebb’s early discovery that domestic rats performed better in problem-solving tests than rats raised in laboratory cages1, numerous studies have employed environmental enrichment (EE) to study the impact of a stimulating setting on cognitive development2. Prolonged exposure to EE, which typically includes opportunity for physical exercise, exploration of novel objects and increased social interactions, has been shown to improve memory in various hippocampal-dependent tasks both in wild type animals and in a range of models of neuronal dysfunction, including brain injury, Huntington’s Disease, Alzheimer’s Disease and intellectual disability disorders3,4,5,6,7,8. Investigation of the underlying cellular and synaptic mechanisms revealed that EE stimulates neurogenesis9,10,11 and increases spine density, dendritic length and dendritic complexity in hippocampus and cortex regions3,7,12. Conversely, an impoverished environment (IE), conceived as individual housing with no exposure to stimulating objects, has been used to model psychiatric disorders in rodents13,14,15, and leads to cognitive impairments, decreased dendritic length and spine density, and defects in glutamatergic transmission16,17,18,19.

The application of EE as a potential intervention for a broad range of human traits20,21, from neurodevelopmental disorders in children22 to cognitive decline in the elderly23, has sparked significant interest in understanding the molecular mechanisms by which housing conditions influence cognition. Transcriptional and epigenetic processes, which are well-established mediators between environmental factors and the genome24,25, play a crucial role in synaptic plasticity underlying learning and memory26,27,28,29. These mechanisms are therefore appealing candidates for translating the impact of rearing conditions into cognitive outcomes. However, genome-wide studies on cortical and hippocampal neurons following EE have identified relatively few consistent changes in gene expression30,31,32,33. The molecular study of IE has received far less attention, leaving the genomic modifications induced by IE largely unexplored. A comprehensive understanding of the epigenetic and transcriptional programs driving environment-induced changes in neuronal plasticity and behavior could be crucial for developing new therapeutic strategies that mimic EE or correct IE-related deficits. Yet, the underlying mechanisms remain elusive. Since cellular heterogeneity in nervous tissue may obscure subtle changes induced by environmental conditions, strategies that mitigate cellular heterogeneity would significantly enhance the sensitivity and specificity of epigenomic and transcriptional analyses, offering unprecedented insights.

In this study, we exposed female mice to EE and IE during the juvenile period to investigate how rearing conditions influence cognitive abilities and to map the associated transcriptional and chromatin changes in the two main hippocampal neuron types: CA1 pyramidal and DG granule neurons. Our behavioral experiments revealed opposite and durable effects on the animals’ cognitive abilities, while the multiomic analysis identified activity-regulated AP-1 as a central player, activated by EE and inhibited by IE, leading to the cell type-specific modulation of synaptic target genes critical for memory functions.

Results

Bidirectional regulation of cognitive capacities by rearing conditions

To investigate the impact of EE on cognitive functions, we developed a paradigm in which a large cohort of 3-week-old C57BL6/J female mice was weaned and housed together in an ample space with opportunities for voluntary exercise, social interaction and exploration of novel objects, where toys and running wheels were changed every two weeks. In parallel, to investigate the impact of social isolation and IE, 3-week-old female mice were weaned and housed individually in small cages, shielded from acoustic, visual, and social stimuli. After 3 months, we compared the performance of the EE and IE cohorts and their littermates housed in standard cages (SC; 4-5 mice per cage) (Fig. 1A).

Fig. 1: Opposite impact of rearing conditions on cognitive capacities.

A Experimental design. At weaning (P21), littermate WT female mice were randomly distributed in impoverished (IE, red), standard (SC, white), or enriched (EE, blue) environment for three months before behavioral analysis. B Total distance traveled in the Open Field test. Mean ± SEM is shown. IE, *n *= 7 mice per group; SC, n = 10; EE, n = 7. No statistically significant differences comparing EE vs SC and IE vs SC by Kruskal-Wallis test followed by Dunn’s multiple comparisons test. C Alternation percentage in the Y-maze. Mean ± SEM is shown. IE, n = 10; SC, n = 10; EE, n = 11. No statistically significant differences comparing EE vs SC and IE vs SC by Kruskal-Wallis test followed by Dunn’s multiple comparisons test. D. Contextual Fear Conditioning. The freezing time before the shock, after the shock, and during the context session was measured. Mean ± SEM is shown. IE, n = 11; SC, *n *= 13; EE, *n *= 15. Kruskal-Wallis test followed by Dunn’s multiple comparisons test comparing EE vs SC (**p *= 0.0111) and IE vs SC (ns, not significant). E Novel Object Recognition. The total time spent exploring the identical objects during the training (left) and the relative time spent exploring the familiar and the novel object during the test session (right) are shown. The green horizontal line across the violin plot represents the median. IE, n = 14; SC, n = 17; EE, n = 18. Mice with a total exploration time inferior to 5 s during the training or the test were excluded from the analysis of the relative exploration time. Total exploration time, Kruskal-Wallis test followed by Dunn’s multiple comparisons test comparing EE vs SC (***p = 0.0004) and IE vs SC (ns). Relative exploration time, Two-tailed Mann-Whitney test comparing familiar and novel object: IE, ns; SC, **p *= 0.0168; EE, ***p *= 0.0050. F Experimental design. At weaning (P21), littermate female mice were randomly distributed in impoverished (IE-SC, red), standard (SC-SC, white), or enriched (EE-SC, blue) environment for one month, and were then kept in standard environment for one additional month before behavioral analysis. G Morris Water Maze. The latency to reach the platform as an indicator of spatial memory was measured for EE-SC vs SC-SC (left) and IE-SC vs SC-SC (right). Mean ± SEM is shown. EE-SC vs SC-SC: SC-SC, n = 14; EE-SC, n = 15. IE-SC vs SC-SC: SC-SC, n = 10; IE-SC, n = 11. Repeated measures two-way ANOVA with the Geisser-Greenhouse correction for multiple comparisons. EE-SC vs SC-SC, visible phase: day, *****p *< 0.0001, rearing conditions, ****p *= 0.0006. EE-SC vs SC-SC, hidden and transfer phase: day, *****p *< 0.0001, rearing conditions, *****p *< 0.0001. EE-SC vs SC-SC: V1, ***p *= 0.0031; H1, ***p *= 0.0019; H2, ***p *= 0.0101; T2, ***p *= 0.0041; T4, **p *= 0.0439; T5, ****p *= 0.0010. IE-SC vs SC-SC, visible, hidden and transfer phase: day, *****p *< 0.0001, rearing conditions, not significant. Source data are provided as a Source Data file.

First, we examined mice basal exploratory activity in an open field (OF) test and did not find any significant change either in the total distance traveled, nor in the time spent in the center and the periphery of the arena (Fig. 1B and Supp. Fig. S1A). Exploration of a Y-maze did not reveal significant differences either (Fig. 1C and Supp. Fig. S1B). Long-term memory was analyzed in a contextual fear conditioning (CFC) task in which the mice received a single foot-shock. EE mice froze significantly longer during recall, indicating better contextual memory in line with previous studies6, while IE mice behaved similarly to SC mice (Fig. 1D). Furthermore, training in a Novel Object Recognition (NOR) task revealed that EE mice spent more time exploring the objects, while IE mice displayed a trend towards reduced exploration (Fig. 1E, left). Both SC and EE mice dedicated more time to explore the novel object in the test session, while IE mice could not discriminate between the familiar and novel object, suggesting impaired recognition memory (Fig. 1E, right, and Supp. Fig. S1C).

To evaluate the persistence of cognitive changes induced by rearing conditions, we examined the performance of EE and IE mice after being returned to SC one month prior to behavioral assessment (henceforth referred to as EE-SC, IE-SC and SC-SC) (Fig. 1F). Given that spatial navigation is known to improve following EE5,7,34, we first assessed the mice using the Morris Water Maze (WMW) test. EE-SC mice demonstrated superior learning compared to the SC-SC group, whereas IE-SC mice did not differ significantly from their controls (Fig. 1G and Supp. Fig. S1D). EE-SC and SC-SC mice swam at comparable speed (Supp. Fig. S1E), and rearing conditions did not affect forelimb strength (Supp. Fig. S1F), suggesting that the reduced latency of the EE-SC group likely reflects improved spatial learning rather than enhanced swimming abilities or physical strength. In the CFC task, EE-SC mice also exhibited significantly stronger associative memory compared to their SC-SC littermates, while IE-SC mice performed similarly to the SC-SC group (Supp. Fig. S1G). Similarly, in a working memory task conducted using the radial maze, EE-SC mice outperformed their SC-SC littermates, while IE-SC mice showed a greater number of errors (Supp. Fig. S1H).

In summary, our findings reveal that experiencing IE, and especially EE, during the juvenile period exerts a lasting impact on hippocampal-dependent memory processes, highlighting the potential existence of enduring molecular changes in hippocampal neurons.

CA1 and DG excitatory neurons exhibit distinct transcriptional and chromatin profiles

The chromatin of CA1 pyramidal neurons and DG granule neurons, the two main neuronal populations in the hippocampus with fundamental yet distinct roles in memory formation35,36, are likely depository for environment-induced molecular changes that influences neuronal responsiveness and hippocampal circuits functioning. To investigate if EE and IE trigger changes in the transcriptome and chromatin of these neurons, we generated mice in which the nuclear envelope of forebrain excitatory neurons is fluorescently tagged with the fusion protein SUN1-GFP upon tamoxifen (TMX) injection (Fig. 2A; hereafter referred to as Sun1-GFP mice). This labeling strategy enabled efficient isolation of these neuronal nuclei from CA1 and DG regions through fluorescence-activated nuclear sorting (FANS) (Supp. Fig. S2A) for subsequent transcriptional and epigenomic analyses29.

Fig. 2: Multiomic analysis reveals profound and correlated differences in transcriptional and epigenetic profiles between CA1 pyramidal neurons and DG granule neurons.

A Genetic strategy for the conditional expression of Sun1-GFP in forebrain excitatory neurons. B At weaning (P21), Sun1-GFP female mice were randomly distributed in enriched (EE, blue), standard (SC, white) or impoverished (IE, red) environment for 3 months. At 1.5 month, mice were administered TMX. C Experimental design. SUN1-GFP+ nuclei were isolated by FANS from manually dissected CA1 and DG regions for downstream multiomic analysis. D PCA of nuRNA-seq, ATAC-seq, H3K27ac CUT&TAG and WGBS DNA methylation profiles of CA1 and DG SUN1-GFP+ neurons. nuRNA-seq, ATAC-seq: n = 3 samples per region per rearing condition. H3K27ac CUT&TAG: CA1 SC, CA1 IE, DG IE, n = 3 samples; CA1 EE, n = 2; DG EE, n = 2; DG SC, n = 4. WGBS DNA methylation: n = 2 samples per region per rearing condition. E Snapshot of nuRNA-seq, ATAC-seq, H3K27ac and DNA methylation genomic profiles of Chrna5, Hunk (CA1 pyramidal neurons markers), Prox1 and Tdo2 (DG granule neurons marker) genes in CA1 (violet) and DG (orange) samples from SC mice. DNA methylation is represented as the percentage of methylation at a given position. DARs, differential H3K27ac peaks and DMRs are highlighted by rectangles. Zoomed-in panels displaying DNA methylation profiles of the boxed regions are shown on the right. F Gene Ontology (GO) enrichment analysis of the upregulated DEGs in CA1 (left) and DG (right) excitatory neurons. Fisher’s one-tailed test with g: SCS (Set Counts and Sizes) correction for multiple comparisons. The 4-5 significant GO terms with higher -log10padj for Molecular Function (MF), Cellular Component (CC), KEGG pathway and Reactome pathway are shown. The minimum number of genes in a statistically significant GO group is 4, and among those displayed in the figure, it is 10. GO terms related to GPCR signaling and ECM are highlighted in bold for CA1 and DG, respectively. G Density distribution of ATAC reads across the DEGs up-regulated in CA1 and DG. H Matrix illustrating the overlap between the genes significantly up-regulated in CA1 or DG neurons, and the genes associated with DARs, H3K27ac peaks and DMRs significantly increased in CA1 or DG neurons. The Jaccard index and the number of shared genes for each comparison, in brackets, are indicated. I Density distribution of H3K27ac reads across the DEGs up-regulated in CA1 and DG. J,K Percentage of cytosine methylation within CG context (J) and CH context (K) across the DEGs up-regulated in CA1 and DG.

Sun1-GFP mice were housed in IE, SC or EE starting at 3-weeks of age, and after 6 weeks, TMX was administered to trigger SUN1-GFP expression (Fig. 2B). After 6 additional weeks in their respective environments, the mice were euthanized, CA1 and DG hippocampal layers were microdissected, and the tissues dissociated and subjected to FANS for multiomic analysis (Fig. 2C). RT-qPCR assays of specific markers of DG (Tdo2, Dsp) and CA1 (Ccn3)37 confirmed the efficiency and precision of the hippocampal subregions microdissection (Supp. Fig. S2B).

Transcriptional changes were assessed using nuRNA-seq, providing insights into gene expression dynamics. Chromatin accessibility was evaluated through ATAC-seq (assay for transposase-accessible chromatin coupled to sequencing), which allowed for the identification of regulatory elements and potential changes in the epigenetic landscape. Histone H3 lysine 27 acetylation (H3K27ac) was investigated using CUT&TAG38, a targeted approach to detect modifications associated with active enhancers and promoters, linking chromatin state to transcriptional activity. DNA methylation patterns were examined via low-input whole-genome bisulfite sequencing (WGBS) to further identify changes in the epigenetic state influencing gene regulation. This multiomic strategy provided a robust framework for unraveling the interplay between housing conditions, gene expression and chromatin modifications. Principal component analysis (PCA) revealed a clear separation between CA1 (violet) and DG (orange) samples in all four genomic screens, while we did not observe any obvious clustering of the samples according to environmental conditions (Fig. 2D).

As a first step in our analyses, we compared CA1 and DG neurons in SC conditions to identify fundamental differences in their transcriptomic and chromatin profiles, before assessing how environmental perturbations modify these cell-type specific programs. nuRNA-seq identified 2552 differentially expressed genes (DEGs; padj <0.05 and ∣log2FC(CA1/DG)∣ > 1), with 1339 genes up-regulated in CA1 and 1,213 genes up-regulated in DG (Supp. Fig. S3A and Supp. Data 1). DEGs included well-known markers of CA1 pyramidal neurons, such as Chrna5 and Hunk, and DG granule neurons, such as Prox1 and Tdo2 (Fig. 2E). GO analysis revealed profound differences between the two neuronal types in terms of cell adhesion molecules and intracellular signaling. In particular, multiple G-protein-coupled (Gprc6A, Gpr101, Gpr139, Gpr26, Gpr161, Gpr3), serotoninergic (Htr1b, Htr2c, Htr5b, Htr1f), and cholinergic (Chrm2, Chrm3, Chrm5, Chrna4) receptors were found up-regulated in CA1 neurons, while extracellular matrix (ECM) proteins were highly overrepresented in DG neurons, including members of laminin (Lama1, Lama2, Lama5), collagen (Col1a1, Col1a2, Col3a1, Col4a5, Col4a6, Col5a3, Col12a1, Col13a1, Col15a1, Col27a1) and ADAMTS (short for a disintegrin and metalloproteinase with thrombospondin motifs) families (Adamts3, Adamts12, Adamts15, Adamts17, Adamts18, Adamtsl2) (Fig. 2F and Supp. Data 2). These broad transcriptomic differences likely underlie the distinct synaptic features and functional specialization of CA1 and DG excitatory neurons35,39.

ATAC-seq peaks were overrepresented in gene bodies and promoters, in line with the association of open chromatin with transcribed genes (Supp. Fig. S3B). Comparative analysis identified 39,084 differentially accessible regions (DARs; padj <0.01 and log2FC(CA1/DG) > 1) between CA1 and DG samples in the SC condition, with 20,465 regions more accessible in CA1 and 18,619 regions more accessible in DG (Supp. Fig. S3D and Supp. Data 3). Chromatin opening is a well-established feature of actively transcribed genes. In agreement, CA1 pyramidal neurons displayed higher accessibility across CA1 up-regulated genes than DG up-regulated genes, and vice versa for DG granule neurons (Fig. 2E, G and Supp. Fig. S3C). Furthermore, the genes annotated to DARs augmented in a specific cell-type showed a highly significant overlap with the DEGs upregulated in the same neuronal type (Fig. 2H).

Similar to ATAC-seq, CUT&Tag profiling of H3K27ac revealed peak enrichment at promoters and gene bodies, consistent with its association with active transcription (Supp. Fig. S3B). When comparing H3K27ac profiles between CA1 SC and DG SC samples, 2426 and 1765 peaks were retrieved as significantly increased in CA1 or DG neurons, respectively (padj <0.05 and log2FC(CA1/DG) > 1) (Supp. Fig. S3E and Supp. Data 4). Like chromatin accessibility, H3K27ac levels at CA1 up-regulated genes were higher than DG up-regulated genes in CA1 pyramidal neurons, and vice versa for DG granule neurons (Fig. 2E, I). Coherently, the genes annotated to H3K27ac peaks that increased in one cell-type largely overlapped with the DEGs upregulated in the same neuronal population (Fig. 2H). Overall, we found widespread changes in chromatin accessibility and H3K27 acetylation between pyramidal and granule neurons, which positively correlate with their cell-type specific transcriptional signatures, consistent with the established association of these chromatin features with active gene expression.

DNA methylation analysis using WGBS revealed even more numerous differences between the two cell types, identifying 91,987 and 82,159 differentially methylated regions (DMRs; FDR < 0.01 and methylation difference > 25%) with higher methylation levels in CA1 and DG neurons, respectively (Supp. Fig. S3F and Supp. Data 5). Promoter DNA methylation is typically associated with gene repression, whereas the effects of gene body methylation on transcription are more complex40. In dividing cells, gene body DNA methylation correlates with gene expression; however, in non-dividing cells such as neurons, it does not correlate with active transcription41,42. We found that genes upregulated in CA1 exhibited regions with lower CG and CH methylation level in the chromatin of CA1 pyramidal neurons across both the promoter and gene body compared to genes upregulated in DG. Conversely, in DG granule neurons, lower DNA methylation levels were observed in the gene bodies of DG-upregulated genes (Fig. 2E, J, K). Consistent with this, and in contrast to genes annotated to ATAC and H3K27ac peaks, the genes associated with DMRs that augmented in one neuronal type significantly overlapped with DEGs upregulated in the other cell type (Fig. 2H). Taken together, our data show that CA1 and DG excitatory neurons exhibit extensive differences in DNA methylation profiles, which inversely correlate with their unique transcriptional programs, indicating a repressive role for this chromatin mark in regulating neuron type-specific gene expression.

To further explore the relationship among chromatin features within each cell type, we categorized differential ATAC-seq peaks between pyramidal and granule neurons as either proximal (<2 kb) or distal (>2 kb) to the nearest TSS and analyzed H3K27ac and DNA methylation profiles across these regions (Supp. Fig. S3G). For all three marks, the profiles were similar between TSS-proximal and -distal regions, suggesting that comparable chromatin processes operate in both contexts. However, we observed marked differences between CA1 and DG neurons in the correlation pattern among chromatin features. In DG neurons, regions with increased chromatin accessibility showed sustained H3K27ac enrichment and reduced DNA methylation, whereas this pattern was much less pronounced in CA1 neurons. These findings suggest a tighter coupling of chromatin features in DG granule neurons than in CA1 pyramidal neurons, pointing to distinct epigenetic mechanisms or greater chromatin heterogeneity in the latter population.

To better understand how the distinct chromatin profiles of CA1 and DG excitatory neurons result into cell-type specific transcriptional programs, we classified all the accessible regions identified by ATAC-seq in the two cell types into promoters or enhancers, based on the presence or absence, respectively, of the epigenetic mark histone 3 lysine 4 trimethylation (H3K4me3) as determined by ChIP-seq analysis of hippocampal excitatory neurons43. Our analysis determined 20,606 promoters and 138,191 enhancers in CA1 pyramidal neurons, and 19,039 promoters and 103,699 enhancers in DG granule neurons (Supp. Fig. S4A). Remarkably, a substantial fraction of promoters and enhancers were exclusive to one cell-type or the other, with 40% of pyramidal neuron enhancers (over 55,000 regions) which were absent in granule neurons, denoting extensive differences in chromatin regulatory landscapes between the two populations (Supp. Fig. S4B). The genes associated to these cell-type exclusive promoters and enhancers largely overlapped with DEGs upregulated in the respective neuronal populations, suggesting that they play a crucial role in shaping the distinct transcriptional profiles of CA1 and DG neurons (Supp. Fig. S4C). Additionally, DMRs with increased DNA methylation in one cell type were enriched in enhancers exclusive to the other, further supporting the repressive role of DNA methylation (Supp. Fig. S4D).

Altogether, our genomic analyses revealed profound differences in the transcriptional and chromatin profiles of the two main neuronal populations in the hippocampus. Furthermore, the strong correlation between transcriptomic and chromatin profiles within each cell population demonstrates the robustness and cell-type specificity of our datasets.

Intrinsic differences in activity-driven transcription between pyramidal and granule neurons

To understand how TFs operate differently in CA1 and DG regulatory regions, we performed a TF footprint analysis of our ATAC-seq dataset within promoters and enhancers shared by CA1 and DG excitatory neurons (p < 0.05 and binding differential score >0.25 or <−0.25). The analysis revealed that the activity-dependent transcriptional complexes AP-1 and NEUROD44,45 exhibited stronger predicted binding to chromatin in pyramidal neurons compared to granule neurons, both within promoters and enhancers (Fig. 3A, B and Supp. Fig. S5A). These results suggest that CA1 neurons exhibit distinct responses to synaptic activation compared to DG neurons, with activity-induced signaling pathways being more strongly activated in CA1 neurons under standard conditions, likely reflecting higher baseline activity in pyramidal neurons.

Fig. 3: CA1 pyramidal neurons exhibit higher expression of the activity-induced transcriptional program than DG granule neurons.

A TF binding prediction in enhancers and promoters comparing CA1 (violet) and DG (orange) excitatory neurons in SC conditions (*p *< 0.05 and binding differential score >0.25 or <−0.25; significance derived from the TOBIAS model). Circle size indicates motif enrichment p-value, and circle color refers to the predicted occupancy by the TFs. n = 3 samples per region. B Digital footprint for AP-1 (FOS motif) in enhancers (top) and promoters (bottom) comparing CA1 and DG excitatory neurons. Values correspond to normalized Tn5 insertions. C Venn diagrams showing the overlap between the DEGs up-regulated in CA1 (left) or DG (right) excitatory neurons and the 200 most up-regulated genes in response to KA treatment29. Cumulative hypergeometric probability is indicated. D Relative expression levels of a panel of IEGs in CA1 and DG excitatory neurons in SC conditions, as measured by nuRNA-seq. Data are expressed as fold change over DG neurons levels. Mean ± SD is shown. n = 3 samples per region. Two-tailed Mann-Whitney test: Arc, Dusp5, Egr1, Homer1, Npas4, Nr4a1, Nr4a2 ****padj <0.0001; Fos, ***padj = 0.0003; Egr2, *padj = 0.0150. E Snapshot of nuRNA-seq, ATAC-seq, H3K27ac and DNA methylation genomic profiles of Homer1 (top) and Dusp5 (bottom) genes in CA1 (violet) and DG (orange) samples from SC mice. DNA methylation is represented as the percentage of methylation at a given position. DARs, differential H3K27ac peaks and DMRs are highlighted by rectangles. Zoomed-in panels displaying DNA methylation profiles of the boxed regions are shown on the right. F Representative immunostaining of FOS (top) and ARC (bottom) proteins in CA1 and DG coronal sections of SC mice. DNA was counterstained with DAPI. Scale bar, 40 µm. The staining was repeated on n = 2 mice. Source data are provided as a Source Data file.

To further investigate the regulation of activity-induced gene expression in the two cell types, we examined the levels of previously identified activity-induced genes29 within our datasets. Interestingly, a significant fraction of the top 200 induced genes was upregulated in CA1 neurons compared to DG neurons in SC (Fig. 3C), including classical inmediate early genes (IEGs) such as Arc, Fos, Dusp5, Egr1, Egr2, Homer1, Npas4, Nr4a1 and Nr4a2 (Fig. 3D). Some of these genes also displayed higher accessibility, increased H3K27ac levels and reduced DNA methylation in CA1 compared to DG (e.g., Homer1 and Dusp5, Fig. 3E). Since IEGs are induced in response to neuronal activity, our results suggest that a larger fraction of CA1 neurons are active in standard housing conditions compared to DG neurons.

Consistent with the multiomic analysis, immunostaining of hippocampal sections revealed a moderate but spread expression of FOS protein in CA1 pyramidal neurons, while only a few sparse granule neurons in the DG layer displayed detectable FOS levels (Fig. 3F, top). Similarly, ARC showed a stronger expression in CA1 compared to DG (Fig. 3F, bottom).

These findings align with differences observed during flow cytometry. While DG nuclei showed a narrow green fluorescence distribution across all conditions, CA1 SC samples displayed a bimodal signal distribution, which was shifted toward higher levels after EE (Supp. Fig. S6A, B). Sun1-GFP expression is driven by the chimeric CAG promoter that contains binding sites for activity-regulated transcription factors (TFs) like the cAMP response element binding protein (CREB) and the activator protein 1 (AP-1)46. Consistently, Sun1-GFP transcripts were significantly increased in animals treated with the glutamate receptor agonist kainic acid (KA) compared to saline29 (Supp. Fig. S6C). Altogether, these results suggest that CA1 pyramidal neurons have a more sustained baseline activity than DG granule neurons, and that EE causes an increase in tonic activity-dependent transcription in CA1 neurons.

EE enhances and IE downregulates the activity-induced transcriptional program in hippocampal neurons

Next, we conducted a detailed analysis of the transcriptional and chromatin differences induced by rearing conditions in the two cell types. In the PCA of CA1 nuRNA-seq samples, the experimental groups did not cluster distinctly, although IE nuclei tended to separate from the other two conditions (Fig. 4A). Indeed, IE was the primary driver of the transcriptional changes, leading to the downregulation of most DEGs (Fig. 4B). Notably, nearly all DEGs were more strongly expressed in CA1 than DG in standard conditions. We identified 58 DEGs (padj <0.1; Supp. Data 6) which displayed a highly significant number of physical and functional interactions among them, including a group of 6 highly interconnected IEGs involved in transcriptional regulation (Supp. Fig. S7A). Remarkably, we noticed that numerous IEGs (Arc, Fos, Fosb, Egr1, Egr2, Npas4, 1700016p03rik) clustered together and were downregulated after IE and upregulated in response to EE (Fig. 4B, C). Consistent with flow cytometry findings, this evidence indicates that rearing conditions bidirectionally regulate the activity-dependent transcriptional program in CA1 pyramidal neurons.

Fig. 4: EE and IE modulate the activity-induced transcriptional program in CA1 and DG excitatory neurons in a cell-type specific manner.

A PCA of nuRNA-seq profiles of CA1 neurons in IE, SC an EE conditions (n = 3 samples per rearing condition). B Heatmap of DEGs in nuRNA-seq profiles of CA1 neurons in IE, SC and EE conditions (LRT test, padj <0.1). The gene expression fold change of each gene in the CA1 SC vs DG SC comparison (Fig. 2) is also shown. IEGs are highlighted in yellow. C Snapshot of nuRNA-seq genomic profiles of Arc and Fos in CA1 samples in IE (red), SC (gray) and EE (blue) conditions. The merged profiles of three independent samples are shown. D PCA of nuRNA-seq profiles of DG neurons in IE, SC an EE conditions (n = 3 samples per rearing condition). E Heatmap of DEGs in nuRNA-seq profiles of DG neurons in IE, SC and EE conditions (LRT test, padj <0.1). The gene expression fold change of each gene in the CA1 SC vs DG SC comparison is also shown. IEGs are highlighted in yellow. F Snapshot of nuRNA-seq genomic profiles of Arc and Nr4a1 genes in DG samples in IE (red), SC (gray) and EE (blue) conditions. The merged profiles of three independent samples are shown. G The effect of EE and IE on the expression levels of the top 200 KA-induced genes in CA1 and DG excitatory neurons is shown. The log2FC of all detected genes and a randomly selected subset of 200 genes are also plotted as controls. The horizontal line across the violin plot represents the median. Kruskal-Wallis test followed by Dunn’s multiple comparisons test, ****p < 0.0001, ns not significant. Source data are provided as a Source Data file.

In DG granule neurons, PCA revealed slight segregation of EE samples, while IE and SC samples were intermingled (Fig. 4D). In agreement with this pattern, IE and SC displayed similar expression levels among the 74 DEGs identified (padj <0.1; Supp. Data 6 and Supp. Fig. S7B), while EE was responsible for most transcriptional changes, leading to the upregulation of the majority of the DEGs (Fig. 4E). Almost all DEGs were enriched in CA1 compared to DG, suggesting that EE induces a subset of the CA1 transcriptional program in DG granule neurons (Fig. 4E). Of the 74 DEGs identified in DG, only three overlapped with the DEGs in CA1, indicating that rearing conditions modulate distinct gene sets in pyramidal and granule neurons. However, we retrieved multiple IEGs (Arc, Nr4a1, Dusp5, Pcsk1, Lbh, 1700016p03rik) upregulated under the EE condition (Fig. 4E, F), demonstrating that EE enhances the activity-driven transcriptional program also in the DG.

Next, to analyze the global behavior of the activity-induced transcriptional program in response to environmental conditions, we examined the EE/SC and IE/SC fold-change of the 200 genes most strongly induced by KA in the hippocampus29. Remarkably, EE led to a significant up-regulation of the activity-induced gene program both in CA1 and in DG, while IE robustly downregulated this set of genes in the CA1 but had no apparent effect in the DG (Fig. [4G](h