MainLoss of heterochromatin is a hallmark of aging and cellular senescence1 and has been linked to derepression of repeat elements, including retrotransposable elements (RTEs)2,3,4. In adult tissues, RTEs are silenced via multiple mechanisms, including loading of the histone variant H3.3 (refs. 5,6). RTEs are co-opted in the response to inflammatory stimuli via their enhancer function and are themselves targets of key proinflammatory transcription factors (TFs)7. More generally, RTE deregulation has been implicated in aging, neurodegenerative diseases, autoinflammatory conditions and cancer5,8,9,10.Microglia are the resident innate immune cells of the CNS11, where they mediate surveillance against infections and various brain pathological states12,13,14. During aging and in neurodegenerative conditions, microglia undergo transition from a homeostatic state to distinct reactive states15,16, which have been reported as either beneficial or detrimental depending on the disease type and/or stage. These transitions are associated with profound alterations of the microglia transcriptome and their function16,17. Although chromatin modifications have been suggested to modulate microglia development and activity18,19, the impact of age-dependent decline in chromatin compaction in microglia remains mostly unexplored. In this respect, trimethylated lysine 27 histone 3 (H3K27me3) was linked to either an increased or decreased inflammatory response20,21,22. Interestingly, loss of heterochromatin has been associated with senescence induced by pathogenic Tau23 and to the ‘dark microglia’ state24.The H3.3 chaperone DAXX25 acts as a repressor of RTEs of viral origin (hereafter referred to as long-terminal repeat RTEs (LTR-RTEs))4,26,27,28 via interaction with H3.3 and the H3K9 methyltransferase TRIM28–SETDB1 complex4. We have shown that loss of Daxx in hematopoietic progenitors causes a reduction in H3K9me3 and RTE derepression, leading to PU.1-mediated neutrophilia27. Moreover, DAXX is upregulated in microglia following interferon-γ (IFNγ) stimulation29, but the functional implications of its induction remain unknown.In the present study, we unravel a role for DAXX in maintenance of the homeostatic state of adult microglia, with implications for our understanding of the relationship between loss of epigenetic barriers, innate immune cell senescence and aging-associated neuroinflammation.ResultsExpression dynamics of RTEs and their regulators during maturation of microglia from development to adult stagesWe sought to investigate the expression of RTEs and their regulators during development and in the adult brain. Microglia originate from yolk sac progenitors, which colonize the CNS and differentiate into specialized resident macrophages11. By mining existing RNA sequencing (RNA-seq) datasets30 and k-mean clustering analysis, we revealed dynamic regulation of LTR- and non-LTR-RTEs across developmental and adult stages (Fig. 1a). While the early-stage cluster E1 displayed higher RTE expression, the E2 cluster was enriched in both early and fetal microglia but was downregulated in adult microglia. By contrast, the adult microglia clusters (A1–A2) were enriched in both embryonic and adult microglia. In accordance with previous studies showing RTE deregulation during aging31,32, a number of LTR-RTE subfamilies (based on ref. 33) were enriched in 24-month-old microglia (Fig. 1b).Fig. 1: Dynamic expression of RTEs and RTE regulators in microglia during development, in the adult brain and during aging.Full size imagea, Heat maps showing k-means clustering (k = 5) of dynamically expressed LTRs and non-LTRs in microglia isolated from the developing brain or CNS across embryonic, postnatal and adult stages (data from ref. 30; n = 1 mouse for E14.5 and E16.5, n = 2 mice per group for the rest). Clusters are named as early microglia cluster 1 (E1), E2, premature, adult microglia cluster 1 (A1) and A2 and are indicated on the left; mon, month; normcount, normalized count. b, Heat maps showing k-means clustering (k = 5) of dynamically expressed LTRs and non-LTRs in microglia isolated from 3- to 24-month-old mouse brains (data from ref. 33; n = 4 mice for 3 and 12 months, n = 2 mice for 6 months, n = 3 mice for 9 and 24 months, n = 5 mice for 16 months). Clusters are named as C1, C2, C3, C4 and C5 and are indicated on the left. c, Heat map showing the relative expression of RTE repressor genes across major brain cell types (data from ref. 34). d, Heat maps showing the relative expression of RTE repressor genes in microglia across embryonic (E), postnatal (P), adult and aging stages (data from refs. 30,33). e, Western blots and quantification analysis of DAXX expression in whole-brain lysates from 2-, 9- and 25-month-old mice (n = 3 mice for each age; data were analyzed by ordinary one-way analysis of variance (ANOVA), F = 5.812, P = 0.0395; Tukey’s multiple comparisons test, P9 month versus 2 month = 0.5353, P25 month versus 2 month = 0.0355, P25 month versus 9 month = 0.1449) and in isolated microglia from brains of 4-, 9- and 24-month-old mice (n = 3 mice for each age; data were analyzed by ordinary one-way ANOVA, F = 57.81, P = 0.0001; Tukey’s multiple comparisons test, P9 month versus 4 month = 0.0005; P24 month versus 4 month = 0.0001; P24 month versus 9 month = 0.1213). β-Actin served as a loading control; Fold ch., fold change. f, Volcano plots showing the association between expression levels of DAXX (on a continuous scale) and those of RTE subfamilies (N = 2,922) in a human cohort using multivariable linear regression models, while applying false discovery rate (FDR) correction to account for multiple comparisons; Model 1 (adjusted for age and sex): higher expression levels of DAXX were associated with lower expression levels of 782 (of 795) RTE subfamilies (FDR-adjusted two-sided P