Microglial Evolution: From Ancestral Hypoxia Tolerance to Derived Synaptic Refinement

A Two-Layer Microglial Identity

We subsetted microglial cells from the 100 analysis-layer species across five vertebrate lineages (mammals, birds, reptiles, amphibians, fishes) and aligned them to a divergence time axis relative to Homo sapiens, reconstructing 20 microglial clusters. Program-level scoring resolved two superimposed layers of microglial identity. One is an ancient layer present in the large majority of species, including cartilaginous fishes and, remarkably, a lamprey telencephalon reference. The other is an amniote-enriched module that is strengthened, not invented, in mammals and birds, consistent with the atlas-wide principle of "ancient genes, new tricks": lineage-specific diversity emerges from the redeployment of deeply conserved genetic modules rather than gene birth.

Ancient Conservation, Amniote Amplification

Program-level scoring of the 20 microglial clusters recovers two superimposed layers of identity. The ancient layer dominates across the tree of life, whereas the mature, amniote-enriched layer is widespread in mammals and birds but traces to an earlier origin, detected in fishes as well.

Ancient Microglial Identity in a Jawless Fish

A key anchor for the ancient layer is the sea lamprey (Petromyzon marinus), a jawless vertebrate. The atlas draws on a telencephalon-only lamprey library of 16,647 cells, kept outside the 100-species analysis layer and used as an external non-whole-brain reference. Detection of the ancient microglial program in this library is a deliberate, cautious test: the lamprey sits at a deep evolutionary distance, and its ortholog mappability is limited (18.7%), so absence calls are not evidence of absence, and the lamprey's low cross-species label comparability means statements built on it warrant explicit caveats.

The lamprey, together with Xenopus, Carassius, Astyanax, Microcebus and Nannospalax, is listed among species with low cross-species label comparability or limited sampling, so cross-species quantitative comparisons involving it warrant caution.

Abundance-Divergence at Deep Branches

At the cluster level, abundance-divergence associations separate ancestral from derived states. These correlations are driven by compositional shifts at deep phylogenetic branches rather than continuous gradual change along the divergence axis, and phylogenetically independent contrasts (PIC) were non-significant for all 20 clusters, indicating the abundance trends reflect branch shifts rather than gradual divergence.

Mic_011 shows elevated expression of the regulator MEF2A and of synaptic refinement genes (C3, PROS1; species-level Mann-Whitney FDR < 0.05 versus fishes), consistent with a tentative mammalian shift toward derived states.

Differential abundance testing identified lineage-associated shifts in microglial composition, of which two signals were robust across both cell-level (pooled Fisher's exact test) and species-level (Mann-Whitney) tests: the derived state Mic_011 enriched in mammals, and the ancestral state Mic_005 enriched in fishes. Other lineage-associated patterns were method-dependent and are not assigned cluster-level functional annotation.

Gating Out CAM-Like Peripheral Macrophages

Microglia and CNS-associated macrophages (CAMs) form the microglia/CAM lineage, with 3 clusters in the 43-cluster atlas-level annotation. Because CAM-like peripheral macrophages can resemble microglia in single-nucleus data, potential contamination was assessed per species using the CAM score together with STAB1 expression, and species flagged for high CAM-like load were excluded from downstream microglial analyses. Four atlas species were gated out on this basis.

Microglial subset delineation underpins the lineage conclusions, yet external corroboration against an independent reference was weak, so the microglial subsets rest primarily on internal evidence.

Microglia / CAMs Rank Highest in TAI

A transcriptome age index (TAI) computed from pseudobulk expression ranks Microglia / CAMs at 1.0042, second only to endothelial cells (1.0047). The full ranking runs from ependymal cells (most ancient profile) to endothelial cells (most derived profile). Given the per-mille effect sizes and the use of log-normalized expression weights in place of TPM, this ranking is treated as descriptive rather than as a definitive ordering.

The high TAI of microglia, close to that of endothelial cells, reflects the lineage's position within the cross-species-retained ortholog universe, where identity genes trace to the Euteleostomi-age stratum; the contrast is descriptive at per-mille scale.