Formation of memory assemblies through the DNA-sensing TLR9 pathway

Memories are encoded by assemblies of neurons across hippocampal and cortical circuits. Prevailing mechanisms for forming and maintaining these assemblies include stimulus-induced long-term potentiation (LTP), as well as contributions of intrinsic neuronal programs such as baseline CREB expression and developmental lineage. Stabilization of assemblies is also influenced by perineuronal nets (PNNs) that modulate inhibitory inputs. The authors posited that an overarching process might integrate stimulus-dependent plasticity with pre-existing intrinsic mechanisms to commit specific neurons to memory assemblies. They investigated whether learning triggers DNA damage in subsets of neurons that, in turn, engage DNA-sensing inflammatory pathways—particularly TLR9—to coordinate DNA damage repair (DDR), centrosome-dependent functions, and structural stabilization, thereby contributing to memory formation and persistence.

Background work shows that hippocampal neurons form microcircuits representing related and opposing memories, with LTP requiring extensive energetic and structural remodeling. Intrinsic predispositions including CREB levels and developmental lineage influence recruitment to memory assemblies. PNNs help stabilize memory circuits. Neuronal activity can cause transient DNA breaks required for immediate early gene (IEG) induction, typically occurring at enhancers and repaired rapidly. Immune-related mechanisms—including the NLRP3 inflammasome—are reported to activate days after contextual fear conditioning (CFC). TLR9 and the CGAS-STING pathway are principal sensors of extranuclear DNA; under non-infectious conditions, they are usually silent but can be activated by mitochondrial or self-DNA fragments under cellular stress. The study builds upon these findings to test whether TLR9-mediated DNA sensing links activity-induced DNA damage to neuronal recruitment and consolidation processes.

- Behavioral paradigms: Contextual fear conditioning (CFC; 3 min context, 2 s 0.7 mA shock), tests at recent (96 h) and remote (21 days) intervals; trace fear conditioning (TFC) and delay fear conditioning (DFC) for dissociating hippocampal dependence. - Transcriptomics: Bulk RNA-seq of dorsal hippocampi 96 h versus 21 days after CFC; validation by qPCR. Single-nucleus RNA-seq (snRNA-seq) from dorsal hippocampi 96 h after CFC across 4 libraries: (1) Tlr9fl/fl + Syn-GFP CFC, (2) Tlr9fl/fl + Syn-Cre CFC, (3) Tlr9fl/fl + Syn-GFP naive, (4) Tlr9fl/fl + Syn-Cre naive (with sorting to determine cell specificity). Unsupervised clustering (29 clusters), GO and Reactome analyses, focus on DCX+ clusters. - Protein localization and imaging: Immunofluorescence in CA1 for TLR9, endosomal markers (EEA1, RAB7, RAB11), LAMP2 (late endosome/lysosome), γH2AX (dsDNA breaks), lamin B1 (nuclear envelope), centrosomal markers (centrin, γ-tubulin), DDR mediator 53BP1, RELA (NF-κB), neuronal (NeuN), astrocytic (GFAP), microglial (IBA1) markers. Criteria for γH2AX foci defined as ≥2 s.d. above average nuclear signal. Time course up to 96 h post-CFC. Live imaging of primary hippocampal neurons using dsDNA and mitochondrial DNA dyes to visualize mobile extranuclear DNA. - Extranuclear DNA isolation: From hippocampi (naive, 24 h, 96 h after CFC); verification of absence of genomic DNA contamination (lack of Slc17a7 amplification); cloning and sequencing of dsDNA fragments; analysis of origin (genomic vs mitochondrial), GC/CpG enrichment, fragment size (50–300 bp). - Genetic manipulations: Neuron-specific deletion or knockdown in dorsal hippocampus using AAV9 vectors under human synapsin promoter (Syn): Syn-Cre in Tlr9fl/fl mice; Syn-driven Tlr9 shRNA versus scrambled control in WT; tests also in Rela fl/fl and Ifnar1 fl/fl lines. Cell-type specificity assessed (lack of astrocyte/microglia co-localization). - Pharmacology and genetic pathway testing: TLR9 antagonist ODN2088; CGAS-STING inhibitors RU-521 and H-151; Sting1 knockout mice. DNase manipulations: Dnase2 shRNA overexpression to limit TLR9-activating fragments; TREX1-GFP overexpression to restrict CGAS-STING activation; systemic/intrahippocampal DNase I, S1 nuclease to address extracellular DNA contributions. - Memory-related neuronal tagging: PRAM-GFP under Fos promoter to permanently label IEG-activated neurons at CFC, allowing assessment of Fos reactivation vs DDR (γH2AX) populations at 96 h. - Structural and matrix assessments: Ciliogenesis via ACIII immunostaining; PNNs via Wisteria floribunda lectin (WFL) staining and complexity assessment. - Statistics: ANOVA, t-tests, Chi-square with Bonferroni corrections as reported (e.g., TLR9/LAMP2 co-localization one-way ANOVA; γH2AX time courses; RELA nuclear signal changes; behavioral freezing across tests; DDR/53BP1 recruitment; ciliogenesis and PNN quantifications). Sample sizes typically n=5–11 mice per group; neuron counts 30–360 per condition where applicable.

- Transcriptome changes at 96 h vs 21 d after CFC: 847 differentially expressed genes (440 upregulated). Immune response genes—particularly TLR9 and downstream NF-κB signaling—were prominent (bulk RNA-seq; qPCR validated). - TLR9 dynamics and endosomal trafficking: Neuronal TLR9 protein increased after CFC, with co-localization to LAMP2 peaking at 96 h (one-way ANOVA: TLR9 P=0.0005, F=9.363; LAMP2 NS; TLR9-LAMP2 co-localization P<0.0001, F=21.27). Co-localization with early/recycling endosomes was lower; strongest overlap with LAMP2. Minimal signal in glia. - Extranuclear dsDNA after CFC: Hippocampal extranuclear DNA fractions lacked genomic contamination (no Slc17a7 amplification). Cloning/sequencing yielded 53 fragments (25 unique genomic sequences), predominantly non-coding nuclear origin, none mitochondrial. CpG/GC-rich sequences increased from 33% (naive) to 77% (24–96 h). Live-cell imaging revealed mobile non-mitochondrial extranuclear DNA in neuronal cytosol. - DNA damage and DDR spatiotemporal profile: Discrete CA1 clusters displayed increased γH2AX foci at 1–3 h post-CFC (ANOVA P<0.0001), then large pericentrosomal γH2AX accumulations at 6–96 h. Nuclear envelope ruptures (lamin B1 discontinuities) increased at 1 h and persisted (P=0.0038), with perinuclear γH2AX co-localizing with TLR9 (~75% of 120 events). Pericentrosomal γH2AX co-localized with centrin/γ-tubulin and 53BP1, indicating centrosomal DDR. γH2AX signals were neuron-specific. - Segregation from IEG responses: γH2AX+ neurons showed low overlap with Fos/CREB/EGR1 at 1 h; only ~20% of γH2AX+ nuclei were Fos+. During reactivation (96 h), γH2AX (DDR) neurons showed significantly less Fos reactivation than PRAM-GFP IEG-tagged neurons (Chi-square P=0.0384), indicating largely non-overlapping populations for IEG vs inflammatory/DDR signaling. - Causal role of neuronal TLR9 in memory: Neuron-specific Tlr9 deletion (AAV9 Syn-Cre in Tlr9fl/fl) reduced context freezing across tests (two-way RM ANOVA, virus factor P=0.0007). Neuron-specific Tlr9 shRNA in WT similarly impaired context memory (virus factor P<0.0001). Tlr9fl/fl KO impaired TFC (context: t=4.362, P=0.0006; tone: t=3.899, P=0.0014) but not DFC (tone: P=0.2448). Astrocytic knockdown and microglial depletion did not impair CFC; Syn-Cre in WT had no effect vs GFP, supporting neuron-specific requirement. - Pathway specificity: TLR9 antagonist ODN2088 impaired CFC; CGAS-STING inhibitors (RU-521, H-151) did not. Sting1−/− mice had intact CFC. Dnase2 shRNA (reducing TLR9-activating fragments) impaired memory; TREX1 overexpression (limiting CGAS-STING) did not. DNase I/S1 treatments targeting extracellular DNA were ineffective, arguing against peripheral/extracellular DNA contributions. - snRNA-seq at 96 h: CFC upregulated conserved gene sets across neuronal clusters (and some non-neuronal) related to ER function, vesicle trafficking/acidification, IL-6/inflammation; key TLR9 function genes Atp6v0c, Hsp90b1 were upregulated. Tlr9-KO blunted most CFC-induced upregulated genes while leaving downregulated genes unaffected; cluster 26 showed paradoxical axon guidance/adhesion increases. DCX+ neurons (CA1 and DGGC) showed stronger conserved responses; Dcx expression and Hsp90b1 induction confirmed by RNAscope. - TLR9 controls DDR, ciliogenesis, and PNNs: Syn-Cre increased neurons with dsDNA breaks in WT and more in floxed lines; Tlr9−/− and Rela−/− mice lost 53BP1 recruitment to nuclear and centrosomal DDR sites; Ifnar1−/− preserved centrosomal DDR. Ciliogenesis (ACIII) and PNN formation were intact in WT and Ifnar1−/− but disrupted in Tlr9−/− and Rela−/− (ANOVA for ACIII P<0.0001; PNNs P<0.0001), indicating RELA/NF-κB as the critical downstream mediator. Overall, learning induces nuclear envelope rupture and release of γH2AX/dsDNA fragments, activating neuronal TLR9 in endolysosomes, which organizes centrosomal DDR, cilia formation, and PNN build-up in discrete CA1 clusters required for context memory formation and persistence.

The findings reveal a neuronal, immune-like mechanism that links learning-induced DNA damage to DDR and structural stabilization via TLR9 signaling. In discrete excitatory CA1 clusters, CFC triggers dsDNA breaks, nuclear envelope ruptures, and release of histone/dsDNA fragments into the endomembrane system, promoting TLR9 activation and NF-κB (RELA)-dependent pathways. This cascade coordinates pericentrosomal DDR complex accumulation, ciliogenesis, and PNN maturation—features that likely support the stability and persistence of context memory representations. The DDR/inflammatory phenotype occurs predominantly in neurons distinct from those exhibiting classic IEG responses, suggesting complementary populations: IEG-responsive neurons may facilitate updating and flexibility during retrieval, whereas TLR9–DDR neurons may maintain stable contextual representations. Pharmacological/genetic dissociation indicates TLR9—not CGAS-STING—is essential for hippocampal context memory, consistent with gene expression signatures emphasizing ER and vesicle trafficking needed for endolysosomal DNA sensing. Although direct single-neuron TLR9–ligand interactions were not demonstrated, increased CpG-rich genomic fragments and histone release post-learning provide plausible endogenous ligands. Disruption of TLR9 (and RELA) compromises DDR assembly, cilia, and PNNs, leading to genomic instability and memory impairment, highlighting neuron-intrinsic TLR9–NF-κB signaling as central to integrating synaptic DNA damage with delayed, extra-synaptic mechanisms that consolidate memory.

This study identifies a TLR9-dependent, neuron-intrinsic pathway that transforms learning-induced DNA damage into coordinated DDR, ciliogenesis, and PNN formation in discrete hippocampal CA1 neuronal clusters, thereby promoting formation and persistence of context memories. Neuron-specific Tlr9 loss impairs context memory, blocks CFC-induced gene programs related to ER/vesicle trafficking and inflammation, disrupts 53BP1 recruitment to DDR foci, and compromises cilia and PNN development, effects mirrored by Rela but not Ifnar1 deletion. The data position neuronal TLR9–NF-κB signaling as a key mechanism for memory assembly recruitment and genomic integrity maintenance. Future research should define the precise endogenous ligands and activation kinetics of TLR9 at single-neuron resolution, map how DDR/inflammatory and IEG-defined assemblies interact across memory phases, and explore therapeutic modulation of neuronal TLR9–RELA signaling to prevent or treat cognitive decline linked to genomic instability.

- Direct demonstration of TLR9 activation by specific DNA fragments at single-neuron resolution was not achieved; evidence is inferential (fragment origin/composition, localization, and downstream signatures). - Viral delivery can induce inflammatory responses; although controls and cell-type specificity were addressed, residual confounds cannot be fully excluded. - snRNA-seq did not detect low-abundance transcripts for some immune mediators (e.g., TLR9, RELA), limiting direct transcriptional confirmation. - Behavioral assessments centered on fear conditioning paradigms; generalization to other memory types remains to be tested. - Some author inferences (e.g., DCX-related maturation state shifts contributing to DDR complex recruitment) are correlative. - The contribution of peripheral immune cells and extracellular DNA was considered unlikely but cannot be absolutely excluded beyond the tested conditions.