Neuronal Identity: Ancient Genes, New Tricks

A Conserved Core, Recombined Across Lineages

Systematic transcriptomic alignment across the 100 analysis-layer species uncovers both ancestral neuron clusters and lineage-specific neuronal populations. While core cell types remain conserved, lineage-specific cell states emerge within them. Across 3,479,032 neurons in the full matrix, excitatory and inhibitory identities were clearly segregated in 92 species, and the top marker genes of every neuronal subpopulation trace to the oldest phylostratum, as old as the brain-expressed background itself. Contemporary diversity is organized not by gene birth but by the repeated, subcluster-specific redeployment of ancient transcription-factor families through duplication and regulatory divergence.

Excitatory and Inhibitory Programs, 100 Species

Using GAD versus SLC17A scoring to distinguish the inhibitory and excitatory programs, the atlas resolves clear Ex/In segregation across the phylogeny. Of the 100 analysis-layer species, 92 show a clean excitatory / inhibitory split; 5 species were not testable because of ortholog-mapping or library gaps, 2 lacked sufficient neurons for the test, and 1 showed an inhibitory-only recovery (Figure 2A).

The segregation holds across every major vertebrate lineage, from cartilaginous fishes to mammals, underscoring that the fundamental dichotomy between excitation and inhibition is a deeply conserved feature of the vertebrate brain, even as individual neuronal subtypes diversify within each clade.

24 Neuron Subpopulations, Named by Defining Factors

Dedicated re-clustering of the neuronal lineage resolves 24 neuron clusters: 15 excitatory and 9 inhibitory. These 24 neuron subpopulations are lineage-resolved subclusters nested within the 27 neuronal major-type clusters of the 43-cluster atlas-level annotation.

The fish-specific Ex_21_HCK population is named after a marker that is also a myeloid-lineage kinase; because of this, it is treated as tentatively annotated pending orthogonal support and is not interpreted further here. Reciprocal-best-hit comparison of cluster marker sets against public neuronal reference atlases corroborated the cross-species assignments.

Ancient Modules Redeployed Across Clades

Lineage-specific neuronal subtypes offer a powerful lens on how conserved developmental programs were selectively redeployed during vertebrate brain evolution. Two modules in particular illustrate "ancient genes, new tricks": an ancient ZIC-driven excitatory module and ARX-centered inhibitory programs coordinated with DLX-family regulons.

Beyond these, EGR2-marked populations (Ex_14_EGR2 excitatory and In_17_EGR2 inhibitory) in fishes and amphibians point to conserved activity-dependent regulatory states, while BACH2-positive interneurons (In_18_BACH2) in birds and reptiles imply amniote-specific innovations in inhibitory modulation. Collectively, these patterns reveal a core transcriptional logic: POU, ZIC and EGR families drive the diversification of excitatory projection neurons, whereas ARX-centered programs orchestrate inhibitory neuron specification, fine-tuning the excitation / inhibition balance of the vertebrate cerebrum.

Paralog Diversification of Glutamatergic Identity

Excitatory identity itself shows a striking example of paralog-level diversification at the water-to-land transition. SLC17A6 and SLC17A7, the two vesicular glutamate transporter genes defining excitatory identity, show segregated usage across excitatory subtypes between teleost fishes and tetrapods (Figure 2D): a gene-duplication event and subsequent regulatory divergence rewired which paralog marks which excitatory population, reshaping glutamatergic programs at the fish-tetrapod boundary.

Identity Genes as Old as the Brain Background

A phylostratigraphic gene-age enrichment analysis across 15 branches, from Euteleostomi to Homo sapiens, revealed that neuronal subtype identity is encoded by genes drawn from the oldest stratum that already dominates the brain transcriptome. The top marker genes of all 24 neuronal subpopulations were assigned exclusively to the oldest phylostratum (ps1, Euteleostomi-age) at 100% under both ortholog versions, and every subgroup is individually 100% ps1. The marker transcription factors TRPS1, ZBTB2 and RREB1 are likewise all ps1.

A transcriptome age index (TAI) computed from pseudobulk expression differed only subtly across cell types, and the neurons ranked among the most ancient: Ependymal 1.0019 < Astrocytes 1.0024 < OPC 1.0028 < Oligodendrocytes ≈ Neurons 1.0029 < Microglia/CAMs 1.0042 < Endothelial 1.0047. Given the per-mille effect sizes and the use of log-normalized expression weights in place of TPM, this ranking is treated as descriptive. Across the phylogeny, the neuronal transcriptome age index remained close to 1.0 in all 100 species, indicating that ps1-age dominance is a stable, lineage-independent feature of neuronal transcriptomes.

An important scope note applies: the analysis layer is restricted to human-anchored orthologs retained across species, so these conclusions hold within this cross-species-retained gene universe, not as a genome-wide census. Within this conserved functional core, it is the regulatory programs that recombine extensively across lineages, not the identity genes themselves.