Cellular heterogeneity and multicellular mechanisms in the pathogenesis of late-life depression: insights from single-cell and spatial multi-omics.

Late-life depression (LLD) is a heterogeneous disorder characterised by affective symptoms, cognitive decline, vascular burden, frailty, treatment resistance, and an increased risk of neurodegenerative progression, yet bulk molecular approaches may obscure the cell-type-specific alterations underlying this clinical and biological heterogeneity. This narrative review synthesises single-cell, single-nucleus, spatial, and multi-omic evidence concerning neuronal, glial, vascular, and peripheral immune alterations relevant to LLD, while distinguishing direct findings from clinically defined or older-adult depression cohorts from evidence extrapolated from major depressive disorder, ageing, mild cognitive impairment, Alzheimer's disease, and experimental models. The strongest current evidence implicates selective excitatory and inhibitory neuronal subtypes and astrocytes in prefrontal affective-cognitive circuit dysfunction, whereas endothelial and neurovascular-unit abnormalities, microglial reprogramming, oligodendrocyte-lineage dysfunction, hippocampal plasticity deficits, and peripheral immune remodelling represent biologically plausible but less well-validated mechanisms. Ageing-related cellular alterations, blood-brain barrier disruption, endothelial-astrocyte dysregulation, neuroinflammation, impaired myelin maintenance, and hippocampal vulnerability may interact to contribute to affective, cognitive, vascular, and treatment-resistant phenotypes. However, direct single-cell and spatial multi-omics evidence from clinically defined LLD cohorts remains scarce, and peripheral immune signatures should be interpreted primarily as accessible systemic correlates rather than direct measures of brain pathology. Overall, current evidence supports a hypothesis-generating multicellular framework rather than validated LLD-specific mechanisms or therapeutic targets. Future studies integrating single-nucleus and spatial multi-omics with longitudinal clinical phenotyping, neuroimaging, peripheral immune profiling, in situ validation, and functional perturbation are required to define LLD-specific cellular states, establish causal mechanisms, and identify clinically actionable biomarkers and therapeutic targets.