Formation of memory assemblies through the DNA-sensing TLR9 pathway

Abstract

As hippocampal neurons respond to diverse types of information¹, a subset assembles into microcircuits representing a memory². Those neurons typically undergo energy-intensive molecular adaptations, occasionally resulting in transient DNA damage^3-5. Here we found discrete clusters of excitatory hippocampal CA1 neurons with persistent double-stranded DNA (dsDNA) breaks, nuclear envelope ruptures and perinuclear release of histone and dsDNA fragments hours after learning. Following these early events, some neurons acquired an inflammatory phenotype involving activation of TLR9 signalling and accumulation of centrosomal DNA damage repair complexes⁶. Neuron-specific knockdown of Tlr9 impaired memory while blunting contextual fear conditioning-induced changes of gene expression in specific clusters of excitatory CA1 neurons. Notably, TLR9 had an essential role in centrosome function, including DNA damage repair, ciliogenesis and build-up of perineuronal nets. We demonstrate a novel cascade of learning-induced molecular events in discrete neuronal clusters undergoing dsDNA damage and TLR9-mediated repair, resulting in their recruitment to memory circuits. With compromised TLR9 function, this fundamental memory mechanism becomes a gateway to genomic instability and cognitive impairments implicated in accelerated senescence, psychiatric disorders and neurodegenerative disorders. Maintaining the integrity of TLR9 inflammatory signalling thus emerges as a promising preventive strategy for neurocognitive deficits.

Fig. 1. Nucleic acid-sensing activity after CFC.

a, Bulk RNA-seq showed increased expression of 441 genes in hippocampi obtained 96 h after CFC (recent, n = 7 mice) compared with those collected 21 days (remote, n = 5 mice) after CFC. Volcano plots demonstrate significant increases in expression of genes related to inflammation and TLR signalling. P adj, adjusted P value. b, TLR9 protein levels and co-localization of TLR9 with the mature vesicle marker LAMP2 at different times after CFC. LAMP2 levels did not fluctuate, TLR9 levels and its co-localization with LAMP2 increased 6 h after CFC, peaking 96 h later (n = 6 mice, 360 neurons per time point; one-way ANOVA; LAMP2: P = 0.3104, F_(3,19) = 1.278; TLR9: P = 0.0005, F_(3,20) = 9.363; co-localization: P < 0.0001, F_(3,20) = 21.27). c, TLR9 and LAMP2 signals in glial cells (revealed by nuclear size), show no significant co-localization (n = 6 mice, 12 glial cells per time point; one-way ANOVA; P = 0.8186, F_(3,20) = 0.3090). Green arrow, LAMP2; orange and white arrows, LAMP2–TLR9 co-localization; red arrow, TLR9. Scale bars: left, 25 μm; right, 40 μm. d, TLR9–vesicle pool co-localization 96 h after CFC (early endosome: EEA1 and RAB7; recycling endosome: RAB11; late endosome: LAMP2) reveals that the highest overlap is with LAMP2 (orange arrows; n = 6 mice, 30 neurons per time point; one-way ANOVA; P < 0.0001, F_(3,20) = 31.53). Note the lack of TLR9 and LAMP2 signals in a glial cell (cyan arrow). Data are mean ± s.e.m. Scale bar, 20 μm. e, Hippocampal cytosolic dsDNA (naive, 24 h or 96 h after CFC) shows no contamination with nuclear DNA, as revealed by lack of ubiquitous amplification of Slc17a7 (which encodes vGlut1) (left). Cloning and sequencing identified genomic dsDNA fragments enriched with non-coding gene GC sequences 24 h and 96 h after CFC (left graph), sized 50–300 bp (right graph). Ctrl, control; miRNA, mitochondrial RNA; ncRNA, non-coding RNA; snoRNA, small nucleolar RNA; snRNA, small nuclear RNA. f, In vitro imaging of primary hippocampal neurons using fluorescent dyes, revealing mobile extranuclear DNA distinct from mitochondrial DNA (Supplementary Video 1). Scale bar, 10 μm. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; NS, not significant; WT, wild type. Source Data

Fig. 2. DNA damage and DDR after CFC.

a, Left, number of neurons showing γH2AX puncta (n = 360 neurons per group; one-way ANOVA, P < 0.0001, F_(5,30) = 65.09) and size of γH2AX foci (n = 360 neurons per group; one-way ANOVA; P < 0.0001, F_(5,30) = 38.17) after CFC. Right, localization of γH2AX (green arrows) in neurons (marked by NeuN; cyan arrows) relative to astrocytes (marked by GFAP; purple arrows) and microglia (marked by IBA1; orange arrows) (two-way ANOVA; factor: cell type, P < 0.0001, F_(2,90) = 673.8; factor: time, P < 0.0001, F_(5,90) = 77.73; cell type × time, P < 0.0001, F_(10,90) = 77.95). Bottom right, number of γH2AX foci. b, Nuclear envelope ruptures coinciding with detection of extranuclear γH2AX (red arrows) and DNA (green arrows) (n = 6 mice; one-way ANOVA; P = 0.0038, F_(5,30) = 4.445). Scale bars: top row, 25 μm; bottom row, 20 μm. c, Extranuclear γH2AX overlapping with TLR9 (orange arrows) (n = 120 total, 75% overlap). Scale bar, 20 μm. d, Pericentrosomal accumulation of γH2AX shown by co-localization with centrin 2 and γ-tubulin (arrowheads). Additional co-recruitment of 53BP1, revealing centrosomal DDR (n = 30–131 neurons; two-tailed Chi-square test; ${\chi }_{\left(4\right)}^2=22.98$, P < 0.0001; post hoc analysis using Bonferroni-corrected α = 0.05, 53BP1: 3 h versus 1 h ^NSP = 0.116, 6 h versus 1 h ****P < 0.0001, 24 h versus 1 h ****P < 0.0001, 96 h versus 1 h ****P < 0.0001; centrin 2: 3 h versus 1 h ^NSP = 0.5061, 6 h versus 1 h ^NSP = 0.2061, 24 h versus 1 h ^####P < 0.0001, 96 h versus 1 h ^####P < 0.0001; adjusted α P < 0.001). e, Co-labelling of γH2AX⁺ (purple arrows) and Fos⁺ (green arrow) neurons (20%). f, Significantly lower number of γH2AX⁺ neurons (orange arrows) relative to PRAM⁺ neurons (green arrows) show memory reactivation (co-labelling with Fos; purple arrows) (n = 216 neurons; two-tailed Chi-square test; ${\chi }_{\left(3\right)}^2=6.518$, P = 0.0384; post hoc analysis using Bonferroni-corrected α = 0.05, γH2AX⁺ versus PRAM⁺ *P = 0.0215, γH2AX⁺PRAM⁺ versus PRAM^{+ NS}P = 0.1007; adjusted α P < 0.025). Data are mean ± s.e.m. Dox, doxycycline; T, total neurons; R, reactivated neurons. Scale bar, 20 μm. Source Data

Fig. 3. Impaired context memory after neuron-specific deletion of hippocampal Tlr9.

a, Left, experimental schematic. Middle, persistent reduction of freezing during context tests of Tlr9^fl/fl mice injected intrahippocampally with Syn-cre (n = 11 mice) compared with the control group injected with Syn-GFP (n = 9 mice; two-way ANOVA with repeated measures; factor: virus, P = 0.0007, F_(1,18) = 16.54; factor: test, P = 0.0007, F_{(1.936,34.84)} = 9.30, virus × test, P = 0.4358, F_(2,36) = 0.85). Right, lack of co-localization of Syn-cre with astrocytic and microglial markers. Scale bar, 40 μm. b, Left, reduction of TLR9 levels (mean optical density per neuron, 60 neurons per mouse, 7 mice per group; two-tailed unpaired t-test; t₁₂ = 17.4700, P < 0.0001) and RELA nuclear signal (60 neurons per mouse, 7 mice per group; two-tailed unpaired t-test; t₁₂ = 3.5679, P = 0.0039) after neuron-specific deletion of TLR9. Right, representative micrographs. Orange arrow indicates TLR9, red arrow indicates RELA. Scale bar, 20 μm. c, Persistent reduction of freezing during context tests of wild-type mice injected intrahippocampally with neuron-specific Tlr9 shRNA (n = 10 mice) compared with scrambled RNA (scrRNA) (n = 8 mice). Two-way ANOVA with repeated measures; factor: virus, P < 0.0001, F_(1,16) = 35.50; factor: test, P = 0.2347, F_(2,32) = 1.517; virus × test, P = 0.0027, F_(2,32) = 7.168. d, After TFC, Tlr9-KO resulted in impaired freezing during context (two-tailed unpaired t-test; t₁₅ = 4.362, P = 0.0006) and tone tests after TFC (two-tailed unpaired t-test; t₁₅ = 3.899, P = 0.0014) (GFP, n = 9 mice; Syn-cre, n = 7 mice), but intact freezing during the tone test after DFC (two-tailed unpaired t-test; t₁₄ = 1.214, P = 0.2448). Data are mean ± s.e.m. Source Data

Fig. 4. Single-cell changes of gene expression after neuron-specific deletion of hippocampal Tlr9.

a, Tlr9^fl/fl mice injected with Syn-cre or Syn-GFP were trained in CFC or left untrained (naive), and 96 h later, dorsal hippocampal nuclei were isolated and processed for snRNA-seq. Robust (more than 1.5-fold) changes of gene expression were found after CFC in mice injected with control virus (Syn-GFP CFC versus Syn-GFP naive) or Cre virus (Syn-cre CFC versus Syn-GFP CFC). b, Gene Ontology (GO) analysis reveals that most genes regulated by CFC and Tlr9-KO involve endoplasmic reticulum (ER), mitochondrial function, IL-6 production and inflammation. CFC induced up-regulation of these genes, whereas Tlr9-KO blocked this effect. c, The most conserved genes up-regulated by CFC and down-regulated by Tlr9-KO across cell clusters include Atp6v0c and Hsp90b1, key regulators of TLR9 function (Extended Data Fig. 6). d, Dcx expression superimposed on uniform manifold approximation and projection (UMAP) analysis of snRNA-seq data from dorsal hippocampal cells. The expression of Dcx in the main neuronal clusters is outlined in orange (excitatory CA1), red (inhibitory) and green (DGGC). e, Cell-specific changes in gene expression demonstrates dominant effects of CFC and Tlr9-KO on gene expression in neurons relative to other cell populations with particularly strong effects of Tlr9-KO in DCX⁺ CA1 neurons. DG, dentate gyrus; oligo, oligodendrocyte; OPC, oligodendrocyte precursor cell. f, Reactome analysis reveals the major functional gene networks affected by CFC and Tlr9-KO in DCX⁻ CA1 and DGGC neurons. Circles are scaled to the percentage effect of Tlr9-KO on the gene expression or pathway relative to Syn-GFP with CFC. TLR cascades, DDR and cilium assembly are enriched among the pathways that are most up-regulated by CFC and down-regulated by Tlr9-KO. RNA Pol II, RNA polymerase II. Source Data

Fig. 5. Impaired DDR, ciliogenesis and PNN formation by neuron-specific deletion of hippocampal Tlr9.

a, Increased number of neurons showing γH2AX signals in mice hippocampally injected with Syn-cre relative to Syn-GFP (n = 5 mice (150 neurons) per group; two-tailed Chi-square test; ${\chi }_{\left(4\right)}^2=54.86$, P < 0.0001; post hoc analysis using Bonferroni correction, α = 0.05; versus GFP: WT cre ****P < 0.0001, Tlr9^fl/fl ****P < 0.0001, Rela^fl/fl ****P < 0.0001 and Ifnar1^fl/fl ****P < 0.0001). This effect was further potentiated by injection of Syn-cre in hippocampi of Tlr9^fl/fl, Rela^fl/fl and Ifnar1^fl/fl mice (versus WT cre: Tlr9^{fl/fl ##}P < 0.002, Rela^{fl/fl ##}P < 0.003 and Ifnar1^{fl/fl ##}P < 0.006; adjusted α P < 0.001). The observed genomic instability was accompanied by centrosomal DDR in wild-type and Ifnar1^fl/fl mice, but blunted in Tlr9^fl/fl and Rela^fl/fl mice injected with Syn-cre (n = 5 mice (150 neurons) per group; two-tailed Chi-square test; ${\chi }_{\left(4\right)}^2=124.1$, P < 0.0001; post hoc analysis using Bonferroni correction α = 0.05; versus WT cre: Tlr9^{fl/fl ####}P < 0.0001, Rela^{fl/fl ####}P < 0.0001 and Ifnar1^{fl/fl NS}P = 0.9681). Right, up-regulated γH2AX signals (white arrows) were seen in CA1 neurons but not in adjacent astrocytes or microglia. Scale bar, 20 μm. b, Illustration of the findings presented in a. To facilitate signal detection of γH2AX–53BP1 overlap, 53BP1 images were pseudocoloured with cyan. Scale bar, 20 μm. c, Top, whereas Syn-cre injection in wild-type mice and Ifnar1 deletion did not affect the number of cilia (n = 6 mice per group, 30 neurons per mouse) and PNNs (n = 6 mice per group, one slice of dorsal CA1 per mouse), Tlr9 and Rela knockout impaired ciliogenesis and PNN formation (top; one-way ANOVA; ACIII: P < 0.0001, F_(4,25) = 97.47; PNNs: P < 0.0001, F_(4,25) = 23.31). Representative micrographs depicting ACIII signals (middle, red arrows; scale bar, 20 μm) and PNNs (bottom, blue arrows; scale bar, 100 μm). Wisteria floribunda lectin (WFN). Data are mean ± s.e.m. Source Data

Extended Data Fig. 1. Up-regulated genes and pathways during memory formation.

a Expanded reactome analysis (String database, STRING-db.org) outlining individual functional clusters of up-regulated genes enriched in inflammation, cytokine release, and cell cycle/DNA repair pathways. b Gene ontology analysis showing enrichment of Tlr signaling. c Up-regulation of the Tlr9 pathway replicated in a separate set of hippocampi collected 96 h or 28 days after CFC with qPCR microarrays customized for immune response genes. d qPCR comparing the level of selected Tlr (Tlr7, Tlr9, and Tlr13) obtained 24, 48, or 96 h after CFC relative to naïve samples showing significant up-regulation of Tlr9 (n = 5 mice/group; one way ANOVA Tlr7: p = 0.6677, F_(3,16) = 0.5306; Tlr9: p = 0.0317, F_(3,16) = 3.783; Tlr13: p = 0.5903, F_(3,16) = 0.6569). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 2. Detection of extranuclear γH2AX signals.

a Perinuclear localization of extranuclear γH2AX signals 1-3 h post CFC in RNA-rich compartments (Syto dye; n = 115 cells, 98%) including ER and ribosomes, but not nucleoli. Measurements of the distance from nuclei revealed that most signals were perinuclear, co-localizing with RNA-rich ER and ribosomal compartments (n = 80 cells; one-way ANOVA p < 0.0001, F_(3,76) = 400.3). b Co-localization of extranuclear dsDNA and γH2AX signals in the vicinity of ruptured nuclei 1-3 h post CFC in 36% of the γH2AX-positive puncta (n = 107 nuclei total). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 3. Relationship between the immediate early gene response, inflammatory response, and γH2AX signals.

a, b, c Representative micrographs of γH2AX and cFos signals with Erg1, CREB, and RelA, respectively. d Regression analyses of γH2AX signals with Erg1, CREB, and RelA, respectively in individual neurons. 1 h after CFC, γH2AX puncta show no correlation with Erg1 (n = 120 neurons, 20 neurons/mouse). Similarly, γH2AX signals did not correlate with CREB levels (n = 120 neurons, 20 neurons/mouse). Significant correlation and neuronal overlap of RelA and γH2AX-positive puncta (n = 120 neurons, 20 neurons/mouse) providing support for an inflammatory component of dsDNA damage and DDR. At time points of optimal RelA detection, cFos levels were expectedly low. The r and P values are depicted in the individual graphs. e Examples of random occurrence of γH2AX puncta relative to individual IEG. f Segregated γH2AX and cFos labeling 96 h after context memory reactivation. Source Data

Extended Data Fig. 4. Responses of hippocampal astrocytes and microglia to viral infusions and diet.

a Immunohistochemistry demonstrating lack of overlap between GFP signals stemming from Syn-Cre expression and markers specific for astrocytes (GFAP) or microglia (Iba1, left). GFAP-driven Cre expression was only found in astrocytes (right). b Increased density of the microglial protein Iba1 (WT, Syn-GFP, and Syn-Cre, n = 8 mice/group, one-way ANOVA, p < 0.0001, F_(2,21) = 69.83) and decreased density of astrocytic protein GFAP (n = 8 mice/group, one-way ANOVA, p = 0.0006, F_(2,21) = 10.65) in mice injected with AAV9-Syn-Cre-GFP or control virus AAV9-Syn-GFP when compared to naïve WT controls. c Intact CFC after hippocampal astrocytic deletion of Tlr9 (left, middle) induced by intrahippocampal injection of astrocyte-specific GFAP-Cre in Tlr9^flox/flox mice when compared to the control group injected with GFAP-GFP (GFAP-GFP and GFAP-Cre n = 9 mice/group, two-way ANOVA RM, Factor: Virus, p = 0.9560, F_(1,16) = 0.0031, Factor: Test, p = 0.5451, F_{(1.716,27.46)} = 0.572, Virus × Test, p = 0.5928, F_(2,32) = 0.5315). Micrograph showing viral expression in astrocytes (right, labeled by GFAP). d WT and Tlr9^flox/flox mice injected intrahippocampally with neuron-specific Syn-Cre and fed for 4 weeks with regular or microglial depletion diet prior to CFC (WT control and diet, Tlr9 KO control and diet n = 9-10 mice/group). While depleting microglia, the diet did not affect the Tlr9KO induced freezing impairments (two-way ANOVA, Factor: Genotype, p < 0.001, F_(1,35) = 25.94, Factor: Diet, p = 0.2611, F_(1,35) = 1.305, Genotype × Diet, p = 0.9731, F_(1,35) = 0.0012). e Decreased density of the microglial protein Iba1 after a 4-week diet containing the CSF1R kinase inhibitor PLX3397 when compared to regular diet prior to CFC (left). Similar reduction was found in WT (control n = 9 mice, diet n = 7 mice) and Tlr9^flox/flox mice (control n = 8 mice, diet n = 9 mice) injected with Syn-Cre (two-way ANOVA, Factor: Genotype, p = 0.1487, F_(1,31) = 2.193, Factor: Diet, p = 0.0001, F_(1,31) = 19.06, Genotype × Diet, p = 0.5433, F_(1,31) = 0.3777) (top right). Additional control experiment showing similar freezing of WT mice after intrahippocampal injection of Syn-Cre and Syn-GFP (n = 10 mice/group; two-tailed unpaired t test, t₁₈ = 0.5571, p = 0.5843, bottom right). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 5. Effect of pharmacological manipulations of Tlr9 and GAS/STING pathways on CFC.

a Cannula localization and individual placements (left). Dose-dependent impairment of CFC by pharmacological inhibition of Tlr9 with ODN2088 (4 nmol/mouse n = 8; 8 nmol/mouse n = 7, vehicle n = 17; middle, two-way ANOVA RM, Factor: Drug, p = 0.0238, F_(2,29) = 4.2630, Factor: Test, p < 0.0001, F_{(1.860,53.54)} = 12.66, Drug × Test, p = 0.0748, F_(4,58) = 2.249), and lack of effect of the GAS and STING inhibitors RU-521 (n = 8 mice/group) and H-151 (n = 8 mice/group), respectively (right, two-way ANOVA RM, Factor: Drug, p = 0.1096, F_(2,34) = 2.361, Factor: Test, p = 0.1708, F_(1,34) = 1.958, Drug × Test, p = 0.6029, F_(2,34) = 0.5136). b Intact CFC in STING knockout mice (n = 7 mice/group) versus WT littermates (n = 5 mice/group; two-way ANOVA RM, Factor: Genotype, p = 0.5132, F_(1,10) = 0.4596, Factor: Test, p = 0.4794, F_(1,10) = 0.5398, Genotype × Test, p = 0.7141, F_(1,10) = 0.1421). c Impaired CFC after dorso-hippocampal injection of DNAse2 shRNA (n = 7 mice/group) when compared to scrambled shRNA (n = 7 mice/group; two-way ANOVA, Factor: Virus, p < 0.0001, F_(1,36) = 48.02, Factor: Test, p = 0.3667, F_(2,36) = 1.032, Virus × Test, p = 0.6051, F_(2,36) = 0.5094). d Overexpression of TREX1 (n = 8 mice/group) did not affect CFC when compared to the GFP control (n = 7 mice/group; two-way ANOVA, Factor: Virus, p = 0.3918, F_(1,39) = 0.7500, Factor: Test, p = 0.2407, F_(2,39) = 1.478, Virus × Test, p = 0.4763, F_(2,39) = 0.7559). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 6. Bioinformatic analysis of snRNA-Seq data.

a Unsupervised algorithm identifying 29 cell clusters (left). Nuclei obtained from hippocampi Tlr9^flox/flox mice injected with Syn-Cre but not exposed to CFC were prepared as described in Fig. 4 but separated using fluorescence-activated cell sorting (FACS) and the separated populations of GFP⁺ and GFP⁻ nuclei were subjected to snRNA-Seq using the same approaches as described for unsorted samples. When compared to the total neuron clusters (left), the gene profiles of GFP⁺ nuclei showed a pattern consistent with neuron-specific expression whereas the gene profiles of GFP^- nuclei showed a pattern consistent with expression in non-neuronal cells (right). b Identification of main cell populations based on known markers for neuronal and non-neuronal cells. c Examination of cosine similarity of Atp6v0c and Hsp90b1 across clusters revealed a coefficient of 0.6 and a significant above chance association (p < 0.01), suggesting co-expression across cell types (top). Violin plots showing up-regulation of Atp6v0c and Hsp90b1 by CFC and down-regulation by Tlr9KO (bottom). Source Data

Extended Data Fig. 7. Cluster-specific gene expression and presence of neuronal markers.

a Up-regulation of gene expression after CFC (top) and down-regulation after Tlr9KO (middle) is predominant in neuronal clusters. A significant correlation was found between CFC increased vs Tlr9 KO decreased changes of gene expression by cluster (top corner). A unique pattern opposite the effect of Tlr9 KO is seen in cluster 26, showing a strong increase of gene expression, most of which were involved in control of axon guidance and cell-matrix adhesion. This cluster uniquely expresses Slc17a6 coding for vGlut2 (see c). Gene ontology analysis revealing similar patterns of up-regulation and down-regulation of functional pathways (bottom). b Neuronal phenotyping with conserved markers revealing presence of Dcx in several excitatory and all inhibitory neuronal clusters in addition to immature DGGC (cluster 25). c The expression of excitatory and inhibitory neuron markers as well as d CA and DGGC markers were consistent with the clustering method and identified clusters 0, 7, 11, 15, 18, 20, 21, 22, 24, 26, 28, and 29 as excitatory CA neurons, clusters 1-4 as DGGC, cluster 25 as immature DGGC, and clusters 13, 14, 19, and 23 as inhibitory neurons. Source Data

Extended Data Fig. 8. Cluster-specific changes of cellular phenotypes after CFC and Tlr9 KO.

a CFC induced decrease of GAP43 from excitatory clusters 11 and 24, but the most dramatic effect was the disappearance of Dcx from most excitatory clusters (7, 11, 20, 24, 29) and inhibitory cluster 19 (gray stars). This CFC effect was prevented by Tlr9 KO (black stars), enabling these clusters to retain Dcx while losing Slc17a6. In response to CFC, the most interesting change was in cluster 28, which shifted from an undifferentiated to a mature DGGC phenotype (Prox1⁺ Calb1⁺ Dcx⁻), an effect prevented by Tlr9 KO. In response to Tlr9KO, significant phenotype changes were noted in immature DGGC (cluster 25), which lost their excitatory and acquired an inhibitory phenotype, and in excitatory cluster 26, which switched from an undifferentiated vGlut2⁺ Calb1⁺ GAP43⁻ NeuroD6⁻ phenotype to a vGlut1⁺ Satb2⁺ GAP43⁺ NeuroD6⁺ phenotype typical of CA neurons.

Extended Data Fig. 9. Validation of the RNA-seq data using RNAscope.

a Low magnification images demonstrating Dcx mRNA expression in newly born DGGC as well as in individual CA1 neurons. b Positive and negative controls (see Methods for detailed description). c High magnification images identifying Dcx⁺ CA1 neurons. d The number of Hsp90b1 signals was significantly higher in hippocampi 96 h after CFC relative to the naïve group (left, representative micrographs; right, n = 4 mice/group, 4 sections/mouse, 180-240 nuclei/mouse, two-tailed unpaired t test, t₃₀ = 2.939, p = 0.0063). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 10. Analysis of the potential contribution of infiltrating immune cells and cell free DNA to the cellular and behavioral effects of CFC and Tlr9.

a Violin plots demonstrating lack of detectable infiltrating lymphocytes and myeloid cells in the hippocampus after CFC. Cells positive for macrophage/monocyte markers were identified as microglia based on the presence of Siglech, Hexb, and Sall1. b Intraperitoneal (i.p.) injections of DNase I before CFC did not affect freezing behavior at test (n = 9 mice/group; two-tailed unpaired t test, t₁₆ = 0.3194, p = 0.7536). c Similar freezing was also found in mice injected with intrahippocampal (i.h.) DNase I or vehicle before CFC (n = 7 mice/group; two-tailed unpaired t test, t₁₂ = 0.9197, p = 0.3758). d Combined i.h. injections of DNase I and S1 nuclease before the memory test did not affect context memory retrieval relative to vehicle controls (n = 8 mice/group; two-tailed unpaired t test, t₁₄ = 0.4799, p = 0.6387). Data represented as mean ± s.e.m., ^nsp > 0.05; ****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05. Source Data

Extended Data Fig. 11. Low magnification images demonstrating the extent of genotype-dependent effects of Syn-Cre on dsDNA breaks and 53BP1 recruitment.

Images obtained from two mice per genotype depicting dsDNA breaks (γH2AX) and DDR in neuron-specific knockout of Tlr9 and its downstream signaling genes RelA and IFNAR1. Red arrows indicate γH2AX signals and white arrows indicate γH2AX/53BP1 co-localization. Note lack of colocalization in hippocampi of Tlr9^flox/flox and RelA^flox/flox mice injected with Syn-Cre.

Extended Data Fig. 12. Low magnification images demonstrating the extent of genotype-dependent effects of Syn-Cre on ciliogenesis and PNN formation.

a Genotype-dependent effects of Syn-Cre on ciliogenesis showing lack of cilia (ACIII labeled filaments, red arrows) in Tlr9 KO and RelA KO hippocampi. b Genotype-dependent effects of Syn-Cre on PNN formation (orange arrows depict individual PNNs, magenta arrows indicate dsDNA breaks) revealing disappearance of PNNs in Tlr9 KO and RelA KO hippocampi. c PNN formation around CA1 clusters of intact WT mice exhibiting dsDNA breaks during memory consolidation (3 h after CFC) indicated with arrows depicting γH2AX and WFA signals according to the immunolabeling color code.