Before synapses are lost: Do endogenous retroviruses drive microglial overpruning in autism spectrum disorders?

Abstract

Endogenous retroviruses (ERVs) are epigenetically silenced genomic elements whose reactivation can influence host gene regulation. In this issue of Neuron, Chen et al. identify ERV-derived RNA-DNA hybrids that activate neuronal C4b, driving microglial synaptic overpruning and autism-like behaviors.

Endogenous retroviruses (ERVs) are remnants of ancient retroviral infections that integrated into host germline genomes millions of years ago and are now inherited as part of the host DNA. Although once dismissed as “junk DNA,” ERVs constitute ~8% of the human genome and ~10% of the mouse genome. Aberrant activation is normally prevented by silencing of ERVs through epigenetic repression involving histone-modifying complexes that establish heterochromatin at ERV loci.¹ Notably, ERV long terminal repeats are enriched with transcription factor binding sites and can function as regulatory elements controlling both ERV activity and nearby host gene expression. When epigenetic silencing is disrupted, aberrant ERV reactivation has been linked to inflammatory, oncogenic, and autoimmune pathogies.²

Growing evidence suggests that ERV dysregulation may also be relevant to brain development, where epigenetic programs are tightly coordinated with neuronal differentiation and synaptic maturation. Autism spectrum disorder (ASD), which affects approximately 1%–2% of the population,³ is characterized by persistent deficits in social communication and restricted, repetitive behaviors, and it is frequently associated with altered synaptic development in the prefrontal cortex.⁴ Maternal immune activation (MIA), triggered by prenatal infection, autoimmune conditions, or inflammatory exposures, has emerged as a prominent environmental risk factor for ASD.⁵ MIA can reshape neurodevelopment through immune and epigenetic mechanisms, yet how these perturbations intersect with ERV regulation to influence synaptic development has remained poorly understood.

In this issue of Neuron, Chen et al. address this knowledge gap by identifying a mechanistic link between ERV reactivation and microglia-mediated synaptic pruning in ASD models.⁶ The study was motivated by the identification of a de novo C>G variant in the 5′ untranslated region of the histone methyltransferase SETDB1 gene (chr1:150,926,445) in an individual with ASD, accompanied by reduced SETDB1 protein levels in primary lymphocytes. SETDB1 plays a key role in epigenetic repression, particularly in silencing ERVs through deposition of the repressive histone mark H3K9me3.⁷ Consistent with this observation, reanalysis of published MIA datasets revealed significant downregulation of SETDB1 in the fetal cerebral cortex, where it emerged as one of the most dysregulated epigenetic regulators and was accompanied by enrichment of antiviral and autoimmune transcriptional programs. These findings suggest that SETDB1 deficiency, arising from either genetic variation or prenatal immune exposure, may represent a convergent mechanism linking immune activation to ASD pathogenesis.

To test this hypothesis, the authors generated neural Setdb1 knockout mice using hGFAP-Cre and achieved efficient deletion in over 96% of prefrontal cortex neurons. These mice exhibited core autism-like behaviors, including reduced social interaction, diminished interest in unfamiliar mice, and repetitive stereotyped behaviors. Consistent with these behavioral deficits, analysis of the prefrontal cortex—a key brain region implicated in ASD—revealed reduced synaptic density and impaired neural circuit function. Similarly, MIA models induced by poly(I:C), lipopolysaccharide, or valproic acid during pregnancy resulted in sustained downregulation of Setdb1 in the offspring brain, further supporting a link between prenatal immune challenge, SETDB1 loss, and ASD-related neuropathology.

To identify the cellular mediator of synaptic loss, Chen et al. examined microglial phagocytic activity in Setdb1-deficient mice. Quantitative analyses revealed enlarged microglial soma and increased engulfment of postsynaptic proteins, indicative of enhanced synaptic phagocytosis. Importantly, pharmacological depletion of microglia largely restored synaptic density, demonstrating that microglia are the principal mediators of synapse elimination in this context. Mechanistically, loss of SETDB1 reduces H3K9me3 deposition at ERV loci, resulting in chromatin relaxation and ERV reactivation. Reactivated ERVs generate RNA-DNA hybrid intermediates detected by cytosolic innate immune sensors such as TLR3 and cGAS.⁸ Activation of these pathways drives neuronal upregulation of the complement component C4b, which functions as a synaptic “eat-me” signal that promotes excessive microglia-mediated synaptic pruning (Figure 1).

Figure 1. SETDB1-mediated H3K9me3 deposition silences ERVs and limits C4b-driven microglial activation and synaptic overpruning.

(Left) Under maternal immune activation or Setdb1 deficiency, SETDB1 activity is reduced, leading to loss of H3K9me3 at ERV loci and transcriptional depression of endogenous retroviral elements as well as upregulation of neuronal complement C4b, which drives excessive microglial synaptic pruning.

(Right) When SETDB1 is functional, it uses S-adenosylmethionine (SAM) as a methyl (Me) donor to catalyze H3K9 trimethylation at ERV loci, thereby maintaining transcriptional silencing and limiting C4b expression. Genetic deletion of C4b (C4b cKO) equivalently prevents microglial overpruning even when ERVs are reactivated, which confirms C4b as the critical neuronal signal linking ERV dysregulation to synaptic pathology.

Figure created with BioRender.

The identification of C4b as the critical neuronal signal is particularly noteworthy in light of prior work establishing complement-mediated synaptic pruning as a fundamental mechanism of postnatal circuit refinement³ and demonstrating that elevated complement signaling can drive excessive synapse elimination in pathological contexts.⁹ Chen et al. extend this framework to ASD by showing that neuronal ERV activity—rather than extracellular complement deposition per se—initiates the pathogenic cascade. Neuron-specific knockdown of C4b using AAV-hSyn-shC4b confirmed that the signal originates in neurons rather than in glia. Moreover, complete C4b deletion rescued synaptic density, electrophysiological function, and social behavior without disrupting baseline microglial activity or normal development, establishing a non-redundant role for neuronal C4b in this disease-relevant circuit.

Current ASD treatment relies almost entirely on behavioral and supportive interventions, with no approved therapies addressing underlying disease mechanisms.¹⁰ The most readily interpretable aspect of Chen et al.’s findings for translation is therefore the reverse transcriptase inhibitor (RTI) experiment. Nevirapine and emtricitabine are already in widespread clinical use for HIV treatment and have established safety profiles in pregnancy. Administered to pregnant dams carrying Setdb1-deficient embryos beginning at embryonic day 13.5, before the peak period of microglial synaptic pruning, RTIs normalized RNA-DNA hybrid levels, suppressed neuronal C4b induction, attenuated microglial activation, restored synaptic density, and rescued social behavioral deficits. Comparable effects were also observed in a valproic-acid-induced MIA model and following genetic C4b deletion. Together with human endogenous retrovirus subfamily H (HERV-H) upregulation detected in fetal prefrontal cortex tissue from pregnancies complicated by maternal inflammatory disease, these data support the ERV-C4b-microglial pruning axis as a convergent pathway across diverse ASD etiologies.

Several questions remain unresolved. Mechanistically, it is unclear how neurons sustain chronic TLR3 and cGAS activation without overt cytotoxicity. The authors propose that neuronal ERVs preferentially generate RNA-DNA hybrids rather than double-stranded RNA, which can trigger ZBP1-dependent necroptosis. Whether hybrid accumulation represents an adaptive strategy or reflects subthreshold signaling warrants further investigation. The clinical translation pathway also requires clarification. Because the therapeutic window in these experiments was prenatal (E13.5–P7), it remains unclear whether postnatal RTI intervention could still confer benefit or whether synaptic pruning during neurogenesis defines an irreversible trajectory. It also remains to be determined whether RTIs exert direct anti-inflammatory effects on microglia independent of ERV suppression.

Nevertheless, Chen et al. establish SETDB1 as an epigenetic hub through which diverse genetic and environmental insults converge on shared neuropathology in ASD. Using genetic, environmental, and human fetal tissue models, the study shows that loss of H3K9me3 at ERV loci induces neuronal ERV expression, cytoplasmic RNA-DNA hybrid accumulation, innate immune activation, and C4b upregulation, ultimately driving microglia-mediated synaptic overpruning in the prefrontal cortex. The finding that clinically available anti-HIV RTIs can interrupt this cascade prior to synapse loss highlights a mechanism-based strategy for modifying ASD-relevant pathology rather than merely managing symptoms.