Zuranolone mitigates delirium-like bispectral EEG changes, behavioral deficits, and neuroinflammation across surgical and inflammatory mouse models and age groups.

Delirium is an acute, fluctuating brain dysfunction that commonly follows surgery and systemic inflammation, disproportionately affects older adults, and remains difficult to quantify continuously over time and treat pharmacologically. Here, we tested whether the neuroactive steroid zuranolone, a positive allosteric modulator of synaptic and extrasynaptic GABA_A receptors, mitigates delirium-like abnormalities across two complementary murine delirium models, a lipopolysaccharide-induced systemic inflammation (LPS) model and a postoperative delirium (POD) model, primarily in young and aged mice, with selected analyses in super-aged mice. Using continuous EEG with a validated bispectral EEG (BSEEG) metric, we found that zuranolone attenuated delirium-like EEG slowing in the LPS model in young mice in a dose-dependent manner and retained efficacy in aged mice. In the POD model, prophylactic dosing provided limited benefit in young mice, whereas post-surgery treatment reduced postoperative BSEEG elevations. In aged mice, prophylactic dosing suppressed POD-associated BSEEG abnormalities, and in super-aged mice, prophylactic zuranolone improved survival after POD induction. In parallel, zuranolone reduced microglial density and activation markers (IBA1 and CD68 immunoreactivity) at 24 h after POD surgery and after LPS challenge, with effects that were particularly evident in peri-screw site tissue in young POD mice and more broadly distributed across regions in aged mice. Finally, in young mice, zuranolone improved a composite behavioral severity score in the LPS model, whereas behavioral effects in the POD model were modest and domain-specific. Together, these findings support zuranolone as a candidate strategy to reduce delirium-like electrophysiological and neuroimmune abnormalities, with the strongest effects in inflammation-driven and age-vulnerable contexts.