ResearchPod Summary
This ambitious preprint from the Allen Institute for Brain Science and global collaborators presents the first comprehensive multiomic cell atlas of the basal ganglia (BG)—a conserved brain region central to motor control, reward learning, emotion, and cognition. Analyzing 1.8 million single nuclei from human, macaque, and marmoset brains across eight BG structures (like striatum, globus pallidus, and subthalamic nucleus), the study integrates RNA-seq gene expression, ATAC-seq open chromatin, and spatial transcriptomics. This reveals a new cross-species taxonomy of cell types, with 61 homologous types conserved over 80 million years of evolution, aligned to mouse and human brain atlases.
The atlas identifies 61 cell types shared across primates, remarkably stable despite massive evolutionary divergence. These include excitatory projection neurons, inhibitory interneurons, and glia, with precise homology to rodent models. A standout discovery is the STRd D2 StrioMat Hybrid medium spiny neuron (MSN)—a novel MSN subtype blending striatal direct-pathway (D2-expressing) and matrix compartment features. MSNs dominate the striatum (80-90% of neurons) and form the core of BG circuits, gating action selection via dopamine-modulated direct/indirect pathways. Conservation suggests these types encode universal BG computations, from habit formation to decision-making.
Single-nucleus multiomics provides a layered view: RNA reveals what cells do, ATAC shows how (via accessible chromatin), and spatial profiling maps where. This uncovers conserved cis-regulatory grammars—DNA sequence motifs that dictate cell identity across species. These 'enhancer codes' are prime targets for genetic tools like AAV viruses, enabling precise circuit manipulation in disease models. For students: think of it as reverse-engineering the brain's software from its hardware blueprint, revealing why certain neurons activate in Parkinson's or addiction.
The BG evolved as a loop linking cortex, thalamus, and midbrain, refining actions and predicting rewards. Divergences (e.g., expanded human striatum) hint at cognitive upgrades, but core circuitry persists. This atlas is a foundational resource for disorders like Parkinson's (dopamine loss disrupts MSNs), Huntington's (striatal degeneration), dystonia, OCD, and addiction—where BG hyperactivity drives compulsions. By aligning primate data to human disease atlases, it accelerates translation from mouse models to therapies, like targeted gene editing or neuromodulation.
Led by equal first authors Nelson Johansen and Yuanyuan Fu, this involved 40+ researchers from Allen Institute, Princeton, Yale, VIB, Sanger, and more. Such scale is essential for rare cell types and technical replicates, setting a gold standard for brain mapping consortia like BRAIN Initiative. Published as a Dec 2025 bioRxiv preprint (DOI: 10.64898/2025.12.15.694496), it builds on Allen's whole-brain atlases, pushing toward a full primate connectome.
Why care? This isn't just data—it's a Rosetta Stone for BG biology, bridging evolution, function, and disease with unprecedented resolution. For neuroscience students, it exemplifies how multiomics transforms messy tissues into interpretable cell maps, fueling precision medicine.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.
Alex: Welcome to another episode of ResearchPod.
Sam: Today we're discussing a preprint from the Allen Institute for Brain Science, posted on bioRxiv in December 2025. It's a detailed map of cell types in the basal ganglia—clusters of neurons deep in the brain that help control smooth movements, form habits through learning, manage emotions, and support thinking. The researchers analyzed data from 1.8 million individual cell nuclei across humans, macaques, and marmosets to create this map.
Alex: So the core puzzle here is that without a shared way to name these cell types across species, it's hard to connect mouse experiments to human brain diseases?
Sam: That's right. Basal ganglia issues link to disorders like Parkinson's or schizophrenia, but cell definitions vary between species and brain regions, stalling progress on treatments. This study integrates gene activity, chromatin openness, and location data—what they call single-nucleus multiomics—to align 61 matching cell types conserved over 80 million years of evolution. One notable find is a hybrid type of medium spiny neuron in the striatum.
Alex: Meaning these cells look and act the same way from mice to monkeys to us?
Sam: Precisely. They used alignment tools to match them across atlases, revealing shared features that could guide therapies targeting specific circuits. The fragmentation before made translating risk genes for diseases like schizophrenia unreliable. This atlas suggests a unified starting point.
Alex: How exactly did they build that shared map of the 61 cell types across species?
Sam: They started by grouping cells with similar gene activity patterns within each species—like sorting kids into teams based on shared hobbies and skills from surveys. Then they refined those groups by checking how well they matched across humans, macaques, and marmosets using a tool that measures similarity in gene expression. Researchers call this iterative clustering followed by MetaNeighbor alignment. That process created a hierarchy: broad neighborhoods, then classes, subclasses, and finally 61 detailed groups.
Alex: So it's like a family tree for brain cells, getting more specific at each level?
Sam: Yes. This hierarchy captures how cells relate evolutionarily and functionally, from major types like projection neurons to finer ones like medium spiny neurons in different compartments. A key discovery was a rare hybrid group in the dorsal striatum's D2 medium spiny neurons—cells that blend features from striosome and matrix zones, plus some eccentric traits. It's conserved across primates, and it shows up enriched for schizophrenia genetic risk.
Alex: Enriched for schizophrenia risk? Meaning these hybrids might be central to why the disease develops?
Sam: The paper suggests so. By mapping disease-associated genes onto these precise groups, rather than broad categories, they found schizophrenia polygenic risk concentrates in ventral striatum MSNs and this hybrid type. This links genetics to specific circuits, unlike before when mismatches stalled mouse-to-human translation. Non-human primates helped spot these rare cells due to denser sampling.
Alex: So the deeper sampling and cross-species matching revealed rarities that single-species studies missed.
Sam: Correct. Integrating with spatial transcriptomics confirmed their locations, and mouse alignments showed most groups have counterparts there too—unifying taxonomies across 80 million years. It's a solid foundation for targeted therapies.
Alex: What makes this STRd D2 StrioMat Hybrid MSN stand out molecularly from regular D2 neurons?
Sam: Imagine cells as houses with specific wiring patterns that define their jobs; this hybrid has wiring blending two neighborhoods—striosome and matrix—in the striatum. It expresses unique markers, like synaptic adhesion molecules that help form connections, shared across primates but distinct from standard D2 medium spiny neurons. The paper notes these features suggest special wiring for circuits, and its schizophrenia risk enrichment points to a role in disease circuits.
Alex: How did they confirm that blending at the DNA level, not just gene activity?
Sam: They mapped open chromatin regions—spots where DNA is accessible for gene control—from over a million such regions per species. These showed the hybrid's accessibility bridging striosome and matrix patterns, with intermediate profiles. To decode the DNA grammar, they trained computer models on human and macaque sequences; these models predict if a DNA stretch will open in a specific cell group just from its code—like pattern-recognizing AIs trained on photos to spot animal breeds.
Alex: Predict accessibility from sequence alone? So the DNA code itself encodes cell identity, conserved across species?
Sam: Yes—the models learned motif grammars, combinations of short DNA patterns that transcription factors bind to turn genes on. For the hybrid MSN, certain motifs stood out, distinguishing it while linking it to similar eccentrics. This grammar matched across human and macaque, and even predicted activity of viral tools for targeting MSNs accurately—about as well as lab tests. It suggests precise sequence rules for therapies.
Alex: For schizophrenia, those risk genes overlap these hybrids and ventral MSNs?
Sam: The analysis showed strongest enrichments there, with variants like one in CACNA1I predicted to disrupt access in specific GABA neurons via motif changes. This maps risks to cells precisely, aiding mouse-to-human translation. The paper cautions it's from current GWAS data, needing more diverse samples for fuller picture.
Alex: How does this unified map fit with the classic ideas of basal ganglia circuits, like direct and indirect pathways for choosing actions?
Sam: The classic view sees basal ganglia as having direct pathways that encourage actions and indirect ones that suppress them, plus compartments like striosomes and matrix that organize these into loops for learning rewards. This atlas recovers that structure exactly, placing projection neurons—which send signals out—alongside interneurons and support cells into the same hierarchy of 61 groups shared across primates. It adds molecular details from paired gene and chromatin data, making the groups both chemically consistent and tied to brain locations.
Alex: So it confirms the old model but sharpens it with new cell details?
Sam: Yes—and it tests how stable those features are over evolution. Dorsal-ventral patterns stayed more similar across primates than the others, suggesting position is locked in evolution while pathways and compartments flex more—like a fixed backbone with adaptable layers.
Alex: What about that hybrid MSN—does it challenge those compartments?
Sam: The STRd D2 StrioMat Hybrid blends striosome and matrix signatures without dopaminergic markers seen in some rodent reports, marking it as a pure medium spiny neuron type. It splits into dorsal-ventral subtypes with unique synaptic genes for connections, and data on shape and electrical firing show it occupies fixed spots, hinting at special circuit roles beyond simple direct-indirect splits.
Alex: Any primate-only surprises?
Sam: A key one is a shell of olfactory bulb-like GABA neurons around the striatum edges—two groups, one dopamine-GABA type matching macaque reports, expressing full dopamine genes like their bulb kin. The other sits more dorsal; both share a gene program from early brain development but are scarce in mice. They might handle reward signals, but need more wiring data.
Alex: And those DNA models—do they predict real tools?
Sam: The models scan enhancer sequences to forecast activity in groups, matching lab tests for viral vectors targeting MSNs—good overall, less sharp for D1 versus D2 due to shared grammars. This guides designing viruses for precise cell hits across species, linking code to therapies. The paper notes limits like few donors, missing rare cells or disease variation, and GWAS biases toward Europeans.
Alex: How did they double-check that these 61 groups really match up reliably across species?
Sam: They tested it by pretending one species' data was unknown and seeing if a computer could correctly label its cells using the others' patterns—like quizzing a student on shared traits from class notes. High scores showed most groups replicate strongly.
Alex: So genes and DNA access both conserved in patterns?
Sam: Yes—they measured how consistently each gene's activity pattern matches its counterpart across species, using scores based on how often a gene's profile ranks high. This highlights genes reliably tied to specific groups over evolution. For chromatin too, they did the same for open DNA regions.
Alex: Does that explain why the hybrid MSN's motifs predict tools so well?
Sam: It does. They scanned those regions for short DNA patterns that control genes, counting matches to known switches for proteins called transcription factors. Models learned the combinations that make a region active in one group versus another, predicting lab tests for viruses targeting cells. The paper suggests this sequence logic holds across primates, aiding precise therapies.
Alex: How does the disease risk mapping actually point to therapies, and what are the real limits here?
Sam: The study tested if genes linked to schizophrenia and other conditions show up more in specific cell groups—like seeing if clues from a crime scene fit certain suspects better. It found the strongest links in ventral medium spiny neurons and that hybrid type for schizophrenia. They also predicted how well virus-based tools—called AAVs—would hit these groups by scanning their DNA codes with the models. Matches were solid for most, suggesting ways to deliver therapies precisely to schizophrenia-linked circuits. The paper notes real limits: few brain donors per species means rare cells might be undercounted, and the genetic risk data leans heavily on people of European ancestry.
Alex: Ties genetics, cells, and tools into a clearer path forward, limits notwithstanding. That's a solid contribution from this collaborative effort at the Allen Institute and beyond—backed by BICAN and posted as a preprint on bioRxiv last December.
Sam: Precisely. It advances understanding of conserved brain circuits for motor control, learning, and disorders like schizophrenia, setting up precise therapies.
Alex: Well said. Thanks for joining me on this look at the basal ganglia atlas, Sam. Listeners, explore the supplements and data yourself. Thanks for listening to ResearchPod.