Co-Conservation of synaptic gene expression and circuitry in collicular neurons

Liu, Yuanming; McDaniel, John A.; Chen, Chen; Yang, Lu; Kipcak, Arda; Savier, Elise L.; Erisir, Alev; Cang, Jianhua; Campbell, John N.

doi:10.1038/s41467-025-64204-5

Download PDF

Article
Open access
Published: 15 October 2025

Co-Conservation of synaptic gene expression and circuitry in collicular neurons

Yuanming Liu¹,
John A. McDaniel²,
Chen Chen²,
Lu Yang¹,
Arda Kipcak²,
Elise L. Savier ORCID: orcid.org/0000-0001-7512-1630¹,
Alev Erisir ORCID: orcid.org/0000-0002-7538-7911²,
Jianhua Cang^1,2^na1 &
…
John N. Campbell¹^na1

Nature Communications volume 16, Article number: 9146 (2025) Cite this article

Subjects

Abstract

The superior colliculus (SC), a midbrain sensorimotor hub, is anatomically and functionally similar across vertebrates, but how its cell types have evolved is unclear. Using single-nucleus transcriptomics, we compared the SC’s molecular and cellular organization in mice, tree shrews, and humans. Despite over 96 million years of evolutionary divergence, we identified ~30 consensus neuronal subtypes, including Cbln2+ neurons that form the SC-pulvinar circuit in mice and tree shrews. Synapse-related genes were among the most conserved in the SC, unlike neocortex, suggesting co-conservation of synaptic genes and collicular circuitry. In contrast, cilia-related genes diverged significantly across species, highlighting the potential importance of the neuronal primary cilium in SC evolution. Additionally, we identified an inhibitory SC neuron in tree shrews and humans but not mice. Our findings reveal that the SC has evolved by conserving neuron subtypes, synaptic genes, and circuitry, while diversifying ciliary gene expression and an inhibitory neuron subtype.

Introduction

Neurons are the most diverse cell type of the body, with thousands of molecular subtypes identified in the human brain so far¹. Yet, much remains unknown about the evolutionary mechanisms shaping their forms and functions, including whether these differ across brain regions. Early brain research relied on comparative anatomy to explore evolutionary changes in brain size, organization, and cellular composition^2,3. Subsequently, electrophysiological techniques provided insights into neuronal function and connectivity in different species^4,5. As evolution works through changes in gene sequence and expression^6,7, many studies have applied DNA/RNA sequencing and other genetic and genomics tools and revealed the molecular mechanisms that differentiate brain development, size, organization, and function across species^8,9,10,11,12. More recently, single-cell transcriptomics has revolutionized the field by providing a high-throughput and real-time view of gene expression, cell type diversity, functional specialization, and developmental trajectories at cellular resolution^{13,14,15,16,17,18,19,20}.

Comparative single-cell transcriptomics has focused mostly on the neocortex, presumably because the most striking difference between the human brain and that of non-primates is the vast expansion of neocortical neurons and increased intracortical connectivity²¹. Several studies have shown that synapse-related genes have diverged across human, non-human primate, and rodent neocortex, consistent with differences in cortical circuitry^{15,22,23,24,25,26,27}. For instance, genes associated with neurotransmitter receptors, cell adhesion, and extracellular matrix diverge most when comparing their expression between human and mouse cortex²⁵. In addition, genes related to synaptic transmission and cell adhesion are the most divergent when comparing the dorsolateral prefrontal cortex across four primate species, including humans²³. Thus, changes in synaptic gene expression during evolution may have facilitated synaptic rewiring of brain regions²⁸. However, it is not clear whether this principle also applies to other brain areas, such as subcortical structures that evolved at a different rate compared to the cortex¹². Furthermore, it remains largely unknown how individual cell types in subcortical brain regions have diverged over evolution and whether synaptic genes have been conserved with synaptic circuitry.

To answer these fundamental questions, we studied a midbrain structure, the superior colliculus (SC). The SC is a major sensorimotor center that integrates multimodal sensory inputs to drive reflexive behaviors, as well as higher cognitive functions such as visual spatial attention and decision-making^29,30,31,32. In contrast to the neocortex, which differs dramatically across vertebrate species, the SC is considered as a conserved structure due to similarities across species in its laminar and topographic organization, cell morphology, and similar input and output connections^30,33. In this study, we specifically focused on the visual layers of the SC (sSC). We used single-nucleus transcriptomics, RNA fluorescence in situ hybridization, and retrograde tracing to compare the sSC’s molecular and anatomical organization between mice (M. musculus) and northern tree shrews (T. belangeri), species which diverged in evolution over 96 million years ago³⁴. The tree shrew is a close relative of primates and an emerging model in neuroscience research^35,36,37. To determine whether the similarities and differences we found between mouse and tree shrew SC also extended to human SC, we incorporated human single-nucleus transcriptomics data into our comparative analysis.

Our results identify ~30 consensus molecular subtypes of SC neurons across the three species. Notably, in contrast to previous comparative studies of the cortex, genes for synapse organization and function are among the most similarly expressed across mouse, tree shrew, and human SC neurons. For example, a specific subtype of glutamatergic neuron, marked by synaptic molecule Cbln2 in all 3 species, provides most of the SC’s projection to the pulvinar in both mice and tree shrews, indicating a co-conservation of synaptic gene expression and synaptic circuitry. Interestingly, transcripts related to cilia function are among the most divergent across the mouse, tree shrew, and human SC, suggesting that the primary cilium, a signaling hub for neurons, might have been a hot spot during SC evolution. Finally, we detected an inhibitory neuron subtype in the tree shrew and human SC but not in the mouse SC, suggesting divergence of inhibitory neuron subtypes during evolution. Together, our results demonstrate that the SC evolved through distinct mechanisms compared to the neocortex. Our results also shed light on how SC cell types have diversified, revealing inhibitory neuron subtypes which diverged highly during evolution and an enrichment of transcripts related to primary cilium function in higher order species.

Results

Cellular composition and anatomy are conserved between mouse and tree shrew SC

We profiled the genome-wide mRNA expression of female and male tree shrew superficial superior colliculus (sSC) using single-nucleus RNA sequencing (snRNA-seq) and integrated the dataset with our recently published one from mouse sSC³⁸ (Supplementary Fig. S1A; hereafter referred to as ‘TM dataset’). We clustered 69,492 cells into 31 consensus cell types based on expression of high-variance genes (Fig. 1A and Supplementary Fig. S1B). Neurons comprised the largest proportion and highest diversity of sSC cells in both species (Fig. 1B and Supplementary Fig. S1C). Compared with the mouse sSC, the tree shrew sSC contained a slightly lower proportion of neurons but a higher proportion of oligodendrocyte lineage cells (i.e., oligodendrocytes and oligodendrocyte precursor cells).

**Fig. 1: Transcriptomically defined cell types and neuronal subtypes in the mouse and tree shrew superficial superior colliculus (sSC).**

We then compared the transcriptomes of each cell type across sex and species. Across all cell types, neurons showed the highest correlation between tree shrew and mice, followed by astrocytes (Fig. 1C and Supplementary Fig. S1D). In contrast, macrophages displayed the greatest divergence in gene expression (Fig. 1C and Supplementary Fig. S1D). In addition, when comparing gene expression between sexes within species, neurons were the most conserved cell type between females and males (Supplementary Fig. S1E–H). Leptomeningeal cells and endothelial cells were the most sexually dimorphic in mice and tree shrews, respectively (Supplementary Fig. S1E–H). Together, our results indicate that neurons are the most conserved across species and between sexes.

To compare neuronal subtypes between species, out of the 69,492 total cells, we subsetted and re-clustered 42,940 neurons into 33 distinct clusters and identified cluster-enriched genes (Fig. 1D, E). This clustering of mouse and tree shrew neurons in the sSC was driven by their molecular similarity and present across a range of clustering parameters (Supplementary Fig. S2A–H). We next assessed conservation of the 33 neuronal subtypes between the two species (Fig. 2A, B). Both species had more clusters of inhibitory neurons than excitatory neurons, based on their relative expression of neurotransmitter phenotype genes Gad2 (glutamate decarboxylase 2) and vGluT2 (vesicular glutamate transporter 2, gene Slc17a6; Fig. 2B), respectively. Both species also displayed a similarly high proportion of inhibitory neurons, which constituted an average of 64% of neurons in the tree shrew and mouse integrated dataset, which were validated by RNA fluorescence in situ hybridization (RNA FISH; Fig. 2C–E). Inhibitory neurons were enriched in the most superficial layers of the SC, whereas excitatory neurons were more prevalent in deeper layers (Fig. 2F). Notably, the overall density of neurons was roughly twice as high in mice than in tree shrews, with the difference largely uniform along SC depth (Fig. 2G). These results suggest that sSC neurons have similar composition and spatial organization in the two species but different density and varying proportions of individual transcriptomic clusters.

**Fig. 2: Neuronal composition and spatial organization in the mouse and tree shrew sSC.**

Synapse-related genes are conserved between mouse and tree shrew sSC neurons

Given the close correspondence of sSC cell types and neuronal subtypes we observed between mice and tree shrews, we investigated the conservation of gene expression between the two species. For each orthologous gene, we calculated expression differences between mouse and tree shrew sSC in log2 fold-change (hereafter referred to as ‘species specificity index (SSI)’; Fig. 3A). The SSI values were normally distributed around 0, indicating that most genes are expressed at similar levels between species (Supplementary Fig. S3A). We then performed Gene Ontology (GO) analysis on the 1600 genes with the smallest absolute SSI values to categorize the genes with the most conserved expression levels. These genes were enriched with GO terms associated with synaptic functions and nervous system development, including: postsynapse organization; synapse assembly; regulation of transporter activity; dendrite development; and the regulation of neurogenesis (Fig. 3B). Synapse-, postsynapse-, and projection organization-related genes, for instance, had a higher percentage of SSI values close to 0 than general genes, indicating greater conservation (Fig. 3C and Supplementary Fig. S3B). Among postsynapse-related genes, 52% had absolute SSI values lower than 1, compared to 45% for synapse-related genes and 34% for all genes, suggesting that the expression of postsynapse-related genes is more conserved than other synaptic genes (Supplementary Fig. S3C). Repeating our analysis with a synapse-focused GO method, SynGO³⁹, confirmed enrichment of genes for synaptic processes and organization among the most conserved between mouse and tree shrew sSC (Fig. 3D and Supplementary Fig. S3D).

We then examined the anatomical expression of conserved synapse-related genes in mouse and tree shrew sSC. For example, expression of calcium–calmodulin–dependent protein kinase II alpha (Camk2a), which is crucial for synaptic plasticity⁴⁰, was detected in both inhibitory and excitatory neurons across most subtypes in both species (Fig. 3E and Supplementary Fig. S3E). This contrasts with the Camk2a expression in the hippocampus and cortex, where it is abundant in glutamatergic neurons. To confirm this, we co-localized Camk2a with the inhibitory neuron marker Gad1 in the tree shrew SC and cortex using RNA FISH. The majority of Camk2a+ cells at the superficial SC in fact co-expressed Gad1 (77.48 ± 2.2%), with this co-expression decreasing with depth in the SC (Fig. 3F, H and Supplementary Fig. S3F). In contrast, very few Camk2a+ cells in the tree shrew cortex co-expressed Gad1 (Supplementary Fig. S3G, I). In the mouse SC, Camk2a expression was widespread and strong, making it difficult to distinguish its expression in neurons. We thus injected AAV-Camk2a-GFP virus into the mouse SC and cortex and co-localized GFP with the inhibitory neuron marker vGAT or the excitatory neuron marker vGluT2 by RNA FISH (Fig. 3G). More than half of GFP + cells in the mouse SC contained vGAT transcripts, and most of the remaining GFP+ cells expressed vGluT2 (Fig. 3G, H). This was again different from the cortex, where nearly all GFP + cells lacked expression of Gad1 (Supplementary Fig. S3H, I). This confirms the snRNA-seq analysis that sSC inhibitory and excitatory neurons both express Camk2a in both species.

In addition, one synapse-related gene, cerebellin 2 (Cbln2), which encodes a secreted neuronal glycoprotein essential for synapse formation and maintenance^41,42,43, showed enriched expression in an excitatory cluster of the integrated atlas, TM01 (Fig. 3I). RNA FISH validated co-expression of Cbln2 and vGluT2 in both mouse and tree shrew sSC (Fig. 3J–L). Furthermore, Cbln2 expression was restricted to the lower SGS/SO sublayer of both species, indicating that Cbln2+ cells are similarly located in the tree shrew and mouse sSC (Fig. 3J, K). Another synapse-related gene, somatostatin (Sst), which encodes a neuropeptide involved in synaptic communication, had enriched expression in an inhibitory neuron subtype, TM32, in both the mouse and tree shrew sSC (Fig. 3I). RNA FISH validated co-expression of Sst and Gad1 in the tree shrew SC and cortex, where almost all Sst+ cells contained Gad1 transcripts, consistent with our previous observations in mice³⁸ (Supplementary Fig. S3J–L). Furthermore, Sst+ cells were similarly distributed in both species, localized in the lSGS and SO sublayers (Supplementary Fig. S3J). In summary, these results demonstrate that synapse-related genes, the neurotransmitter of each subtype, and the spatial organization of at least some neuronal subtypes are conserved between the tree shrew and mouse sSC.

A conserved subtype expressing Cbln2 projects to the pulvinar in both mice and tree shrews

Given the conserved molecular and anatomical features of the sSC between mice and tree shrews, we next compared the neural circuitry of one common neuronal subtype. In mice, Cbln2 (a marker of TM01) is known to mark the so-called wide field vertical cells that project to the pulvinar (also known as the lateral posterior nucleus of the thalamus)^44,45,46,47. To confirm this, we first retrogradely labeled pulvinar-projecting neurons by injecting retrograde AAVrg-Cre into the pulvinar of H2B-TRAP mice, which Cre-dependently express a nuclear fluorescent protein (Fig. 4A.1). This labeled a subset of SC neurons (Fig. 4A.2). We isolated cell nuclei from the labeled neurons, transcriptionally profiled them, and clustered them into nine subtypes, including seven neuronal subtypes, accounting for 93.1% of all labeled cells, as well as two glial cell subtypes (Fig. 4A.4 and Supplementary Fig. S4A, B).

**Fig. 4: A conserved SC neuron subtype expressing Cbln2 projects to the pulvinar in both mice and tree shrews.**

To identify the retrogradely labeled SC cells, we mapped their transcriptomes to the TM dataset. We filtered out cells mapping with a confidence score less than 0.8 (out of 1.0), based on the confidence scores at which almost no glia map to the reference dataset, which contained only neurons (Fig. 4A.5, B). A total of 66% of neurons (329 out of 499) met this criterion. About 70% of these pass-filter neurons mapped to TM01, a percentage 18-fold higher than would be expected from random mapping based on cluster proportion (Fig. 4C–E) and largely independent of the confidence filter (Supplementary Fig. S4C, D). These results thus indicate that the pulvinar-projecting sSC neurons in mice are primarily of the subtype TM01.

Of note, two subtypes of pulvinar-projecting neurons mapped to TM01 (Fig. 4D). We next asked whether these two transcriptomic subtypes correspond to two previously reported wide field vertical cell types displaying dendritic and projection differences -- Type I projects to the bilateral pulvinar and extends dendrites to the SC surface, whereas Type II innervates the ipsilateral pulvinar exclusively and ends their dendrites deeper^48,49,50,51. We isolated cell nuclei from the contralateral SC, performed snRNA-seq, and mapped it with the ipsilateral data to the TM dataset (Fig. 4A). If there was a correspondence between two transcriptomic subtypes and two WFV subtypes, the ipsilateral mapping result would present a distinct cluster which was not seen in the contralateral mapping result, presumably Type II. However, both ipsilateral and contralateral pulvinar-projecting neurons consist of the same Cbln2+ subtypes, suggesting a lack of such correspondence in the mouse SC (Supplementary Fig. S4E).

To validate these findings, we injected a retrograde tracer, Cholera Toxin Subunit B (CTB), into the mouse pulvinar and co-localized its labeling with Cbln2 RNA FISH. CTB injection into the pulvinar labeled neurons in the SO sublayer of the SC (Fig. 4F, G). Most of the labeled neurons (87.5 ± 3.4%) contained Cbln2 transcripts, consistent with our snRNA-seq results (Fig. 4H). To investigate whether the pulvinar in the tree shrew receives input from the same subtype of SC neurons, we injected a retrograde, tdTomato-expressing adeno-associated virus (AAV) into the tree shrew pulvinar and co-localized tdTomato with Cbln2 transcripts in the SC by RNA FISH (Fig. 4I, J). Robust tdTomato expression was detected in the lower SGS/SO sublayer of the tree shrew sSC, with most of these cells (85.3 ± 1.9%) co-expressing Cbln2 (Fig. 4K). Together, our results demonstrate that a conserved molecular neuron type, TM01, is the major source of synaptic output to the pulvinar in both tree shrew and mouse sSC.

Primary cilium-related genes are enriched in the tree shrew sSC

Evolution works in part through changes in gene identities and expression levels^6,7. We therefore investigated the differences in gene expression between mouse and tree shrew SC neurons to identify molecular mechanisms of their evolution. First, we performed GO analysis on the top 1600 genes enriched in the mouse sSC based on SSI values. In terms of biological processes, these mouse-enriched genes were associated with chromosome organization, RNA processing, organelle localization, and cellular metabolic processes (Fig. 5A). On the other hand, our GO analysis of the top 1600 tree shrew-enriched genes in the biological process category revealed enrichment of genes for cilium movement, second messenger-mediated signaling, adenylate cyclase (cAMP)-modulating G protein-coupled receptor (GPCR) signaling pathways, calcium-mediated signaling, and lipid transport (Fig. 5B). We also examined enriched GO terms in the category of cellular components and found that they included motile cilia, secretory granules, and collagen-containing extracellular matrix (Fig. 5C). Reciprocally, genes related to cilium movement, secretory granule, and regulation of cilium movement showed a lower percentage of SSI values near 0 but a higher percentage of positive SSI values than general genes, indicating enrichment in the tree shrew sSC relative to mice (Fig. 5D and Supplementary Fig. S5A).

**Fig. 5: Primary cilium-related genes and cluster markers are divergent between mouse and tree shrew sSC neurons.**

Although genes related to motile cilia and their movement were enriched in tree shrew sSC neurons, motile cilia have not been reported in SC neurons. Instead, primary cilia, which are antenna-like sensory organelles, are prevalent in neurons. Primary cilia are enriched with receptors, particularly GPCRs, to receive extracellular signals, and they use signaling cascades to amplify and transduce signals to the soma via secondary messengers such as cAMP and calcium^52,53,54. Therefore, the motile cilia-related genes identified by our GO analysis are likely related to the function of the neuronal primary cilia. Concurrently, approximately one-third of genes related to primary cilium and primary cilium assembly were enriched in the tree shrew with SSI values greater than 1 (Fig. 5E and Supplementary Fig. S5B, C). Our results, therefore, suggest that the tree shrew and mouse sSC diverged highly in terms of their expression of primary cilia-related genes.

Cluster marker genes are divergent between mice and tree shrews

We next explored whether the species-enriched genes were more likely to occur within certain neuronal subtypes. We calculated the expression differences of each orthologous gene between clusters in log2 fold-change (Cluster Log2FC). Compared to all genes, the species-enriched genes showed a higher percentage of larger Cluster Log2FC values, suggesting a cluster specificity among some divergent genes (Fig. 5F). On the other hand, the top 50 markers for each cluster together displayed a lower percentage of SSI values near zero and a higher percentage of larger absolute SSI values than all genes, providing complementary evidence that certain cluster markers tend to differ between species (Fig. 5G). Furthermore, GO analysis of these top 50 markers revealed functions associated with cell-substrate adhesion, neurogenesis, neurotransmission, and GPCR signaling pathways (Fig. 5H). Together, our findings suggest that even though the neuron types are remarkably conserved in mouse and tree shrew sSC, many marker genes of these molecular neuron types are different in the two species.

To validate these observations, we identified markers for some neuron clusters and confirmed their convergent and divergent expression in the mouse and tree shrew sSC with RNA FISH. For instance, syntrophin beta 1 (Sntb1), which encodes a cytoskeletal protein, showed enriched expression in an excitatory cluster, TM13, in both species (Fig. 5I and Supplementary Fig. S5D). For the same cluster, vasoactive intestinal peptide (Vip), which encodes a signaling neuropeptide, was a marker gene in the mouse, but not expressed in tree shrew sSC (Fig. 5I). RNA FISH validated the co-localization of Sntb1 and Vip in the mouse sSC, but detected only Sntb1, not Vip in the tree shrew sSC (Fig. 5J–L). In contrast, Vip transcripts were detected in the cortex of the same tree shrews, confirming that the lack of Vip expression in the tree shrew sSC was not due to technical issues (Fig. 5M). Another sSC neuronal subtype, TM01, contained Cbln2 transcripts in both species but expressed nephronectin (Npnt), which encodes an extracellular binding protein⁵⁵, in mice but not in tree shrews (Fig. 5N and Supplementary Fig. S5E, F). Furthermore, the spatial organization of cluster TM13 differed between species, in that Sntb1 + cells were restricted to the surface of the mouse SGS, whereas in the tree shrew, they were primarily located in the middle of the SGS (Fig. 5J, L). Overall, our results highlight that neuron-subtype marker genes are more divergent during evolution.

An inhibitory neuron subtype is present in the tree shrew but not the mouse sSC

Our clustering analysis identified one subtype of inhibitory neurons, TM03, present in the sSC of tree shrews but not mice (Figs. 2B, 6A). This previously uncharacterized subtype was closely related to another inhibitory subtype, TM02, which is present in both species (Fig. 1C). The presence of the TM03 neuron cluster and its close relatedness to TM02 were stable across clustering parameters, numbers of cells, and dataset normalization and integration methods, indicating that it is not an artifact of data processing (Fig. 6B, C and Supplementary Fig. S6A–I). To further determine the genes driving divergence between the two clusters, we performed differential gene expression analysis between TM03 and TM02 (Fig. 6D). We then performed GO analysis on these two sets of enriched genes separately. Notably, genes enriched in both TM03 and TM02 are associated with functions such as cell-cell adhesion, synapse assembly, cell junction assembly, cell projection, and dendrite development, suggesting a distinct set of synapse-related genes drives the emergence of TM03 in the tree shrew SC (Fig. 6E, F). For instance, TM02 and TM03 differed in their expression of members of the cadherin family, cell-cell adhesion proteins which specify synaptic connectivity⁵⁶: expression of Cdh2, Cdh6, and Cdh9 was more enriched in TM03, whereas Cdh12 and Cdh18 showed higher expression in TM02, raising the possibility that TM02 and TM03 connect with different synaptic partners. Other transcriptional differences imply that TM03 and TM02 may respond differently to fast-acting neurotransmitters. For instance, relative to TM03, TM02 showed higher expression of the GABA receptor subunits Gabrb2, Gabra1, and Gabrg2, the AMPA receptor subunit gene Gria1, and the ionotropic glutamate receptor gene Grid2. In summary, our observations suggest that TM02 and TM03 are related neuronal subtypes that may have diverged from a common cellular ancestor during sSC evolution.

**Fig. 6: An inhibitory neuron subtype is present in the tree shrew sSC but not the mouse sSC.**

We identified genes with enriched expression in TM03 relative to all other clusters and performed GO analysis to explore their potential functions (Fig. 6G). The top GO terms included GTPase- mediated signal transduction, synaptic organization, and dendrite development (Supplementary Fig. S6J). Moreover, transcripts enriched in this neuron cluster relative to the others tended to be highly specific to tree shrews, whereas transcripts enriched in other clusters tended to show greater expression in the mouse sSC (Fig. 6G, H). These results suggest that a species-specific regulatory mechanism of gene expression, involving both upregulation and downregulation, may contribute to the emergence of this distinct inhibitory subtype.

To anatomically map the TM03 neuron cluster in vivo, we selected a top marker gene, Postn, and performed RNA FISH (Fig. 6G and Supplementary Fig. S6K). Postn+ cells were primarily located in the tree shrew SGS, with 90% of them co-expressing Gad1, confirming that they are inhibitory neurons (Fig. 6I–J and Supplementary Fig. S6L). In contrast, in mouse midbrain sections, Postn expression was detected in the Edinger-Westphal nucleus (EW) but not in the sSC, consistent with our snRNA-seq analysis (Fig. 6K). Overall, our results identify an inhibitory neuron subtype in the tree shrew sSC, which is absent from the mouse and provides insights into the cellular and molecular basis for its evolution.

Conserved and divergent molecular features of the mouse, tree shrew, and human SC

Finally, we investigated the conserved and divergent molecular features of the human SC. We obtained human SC neuron data from a published snRNA-seq study¹ (Supplementary Fig. S7A). Neurons (SYN1 + /RBFOX3 + ) comprised 89% of the cells, which is consistent with the reported percentage in the original human SC dataset (Supplementary Fig. S7B). Integrating neurons from the human SC with our tree shrew and mouse datasets (hereafter referred to as HTM dataset) resulted in 31 clusters, annotated as HTM01 to HTM31 (Fig. 7A). Three human-specific neuronal subtypes, HTM01-HTM03, were excluded from further analysis since their expression of homologous neocortical genes suggests they likely originated from the neocortex, not the SC (Supplementary Fig. S7C–E). The remaining 28 neuronal clusters contained cells from all three species in varying abundance, largely consistent with our previous observations in the TM dataset (Supplementary Fig. S7C). Importantly, neurons from the same clusters in the TM dataset consistently mapped together to corresponding clusters in the HTM dataset, indicating conserved neuronal composition across the mouse, tree shrew, and human SC (Fig. 7B). This conserved neuronal composition was not specific to mammals – we found similar neuron subtypes in the optic tectum, an SC homolog, of pigeons and zebra finch (Supplementary Fig. S7F, G). As we observed when comparing tree shrew and mouse sSC neurons, cluster marker genes associated with functions such as cell-substrate adhesion and the neuropeptide signaling pathway showed a higher likelihood of divergence across species (Fig. 7C, D). These findings were further confirmed by examining markers of two excitatory and two inhibitory neuron subtypes (Supplementary Fig. S7H).

TM03, the unique cluster in the tree shrew SC, along with TM02 and TM21, merged into HTM21, suggesting the presence of subclusters within this HTM cluster. Therefore, re-clustering only the HTM21 cells showed that almost all human HTM21 cells and most tree shrew TM03 cells clustered together, suggesting that a unique SC neuron subtype exists in both human and tree shrew SC but not in mouse SC (Fig. 7E). In contrast, other subclusters of HTM21 contained almost no human SC neurons but were correspondingly mapped with neurons from TM02 and TM21, implying the absence of these two clusters in the human SC. In addition, given the close molecular relatedness among HTM21 subclusters, the proportional differences in neurons across the three species support the possibility that these subclusters diverged from a common cellular ancestor during evolution, with human- and tree shrew-specific neuron subtypes becoming predominant at the expense of other close subtypes enriched in mice.

Next, we performed GO analysis on genes enriched in human SC relative to tree shrew and mouse SC (Supplementary Fig. S7I). Notably, as we previously observed when comparing mouse and tree shrew SC neurons, cilia-related functions were among the top GO terms for human-enriched genes (Supplementary Fig. S7J). Our GO analysis of 1,152 genes which increased in expression from mouse to tree shrew to human SC neurons confirmed enrichment of cilia-related genes, along with genes related to pattern development and left-right asymmetry (Fig. 7F, G and Supplementary Fig. S7J). Among the many ciliary-related genes were those encoding proteins for the assembly, organization, and regulation of cilia (e.g., Mark4, Cep41, Tchp, Bbs9, Wdpcp), implying that the primary cilia of SC neurons may have a more complicated structure in higher-order mammals. Genes for ion or protein transport during signaling transduction in cilia (Ttc21a, Pkd1l1, Scn11a, Ift43, Bbs1, Dync2li1) were similarly upregulated, consistent with more dynamic receptor trafficking and signal responsiveness in the primary cilia. Finally, expression of genes related to dynein arm assembly (Daw1, Ttc29, Dnah5, Dnah8, Dnah12) were also enriched in tree shrew and human relative to mouse, suggesting that SC neuron cilia in higher order mammals may be structurally similar to nodal cilia and so have partial motility^57,58. As before, we presumed that these ciliary genes are related to the neuronal primary cilium, since neurons lack motile cilia. Consistent with this, genes selected from the primary cilium assembly GO term also showed increased expression from mouse to tree shrew to human SC neurons (Fig. 7H).

GO analysis of the genes showing decreased expression from mouse to tree shrew and to human SC neurons were associated with glial cell development, glial cell differentiation, negative regulation of axonogenesis, and axon ensheathment (Supplementary Fig. S7K, L). The decrease in neuronal expression of glia-related genes across the three species suggests that neurons may have become more molecularly distinct from glia over the course of evolution.

Finally, GO analysis of genes similarly expressed by all three species revealed functions related to synaptic processes, such as vesicle organization, vesicle-mediated synaptic transport, and neurotransmitter regulation (Fig. 7I). These results suggest that, during the evolution of sSC neurons, genes related to cilia function and pattern specification were highly dynamic, whereas those related to synaptic function were stable, consistent the conservation of synaptic circuitry. In summary, our results demonstrate that the conserved and divergent features between mice and tree shrews extend to humans, revealing insights into the molecular and cellular evolution of the SC.

Discussion

The SC is often described as a “conserved” midbrain structure. This is based on the many anatomical features that are consistently observed in the SC (or optic tectum) across various vertebrates, including its layered organization, input/output connectivity patterns, topographic maps, and morphologically defined neuron types^29,30,33,59. Whether the molecular features and organization of the SC are conserved across mammalian species has remained unknown until now. Here, we have performed a comparative transcriptomics study of the SC to fill this important knowledge gap and revealed molecular and cell-type divergence in a brain region which is otherwise highly conserved.

Our study yields critical insights into the molecular and cellular evolution of the nervous system. While most neuron subtypes are largely conserved across mouse, tree shrew, and human SC, some of their genes vary significantly in expression across these species. Notably, genes associated with ciliary function are among the least conserved in SC neurons across species, whereas synapse-related genes are among the most conserved. Consistent with the latter, we find that the synaptic circuitry between SC and pulvinar is similar across species – one neuronal subtype, marked by expression of the Cbln2 gene, accounts for most SC projections to pulvinar in mouse and tree shrew. Our study also reveals how the SC has diversified with evolution. We identify a subtype of SC neuron present only in tree shrews and humans, and a closely related subtype which is only in mice and tree shrews, which may represent “sister cell types”⁶⁰. Together, our results suggest that the SC has evolved by (1) molecularly specializing cell types with otherwise conserved identities and circuitry, (2) diverging a subset of inhibitory neurons, and (3) enhancing the role of the primary cilium in neuronal signaling.

In some ways, our findings align with previous comparative studies of neocortical cell types. For instance, neocortical neurons are far more conserved across species than neocortical glia^61,62,63. Similarly, we observed that SC neurons were also more alike across species than SC glia. In addition, the human neocortex contains higher ratios of glia to neurons than the non-human primate neocortex⁶⁴, which we also observed in tree shrew sSC relative to mouse sSC. This suggests that, across brain regions, glia may have adapted to a greater extent than neurons to support species-specific functions.

Our study also identified other evolutionary similarities between the neocortex and SC at the level of neuronal subtypes. Most neocortical neuron subtypes are homologous across mammals^25,65 and even vertebrates¹⁸, with a few exceptions for species-specific cell types⁶⁶. Likewise, we found that almost all neuron subtypes were conserved between mouse, tree shrew, and human SC. The two subtypes that were not conserved were highly similar, raising the possibility that they are ‘sister cell types’, derived from a common cell-type ancestor⁶⁰. Comparative transcriptomic studies of the neocortex have also found species-specific neuronal subtypes^66,67. Interestingly, like the inhibitory neuron subtype we identified in human and tree shrew SC but not mouse SC, the species-specific neuron subtypes in human neocortex are also distinguished from other subtypes by their expression synapse-related genes²². This raises the possibility that neuronal subtypes may diverge during evolution to form novel synaptic circuits¹⁷. Another way in which the primate and mouse neocortex differ is in its layering of neuronal subtypes^25,68. Likewise, we observed differences in the distribution of some neuronal subtypes between mouse and tree shrew SC (e.g., cluster TM13, or Sntb1+ neurons), though other neuronal subtypes showed a more conserved anatomical distribution (e.g., cluster TM01, or Cbln2 + neurons). Finally, genes with neuronal subtype-specific expression in the neocortex tended to also be specific to one species²⁴, which we found was true for SC too and suggests that, across brain regions, expression of individual genes can be subject to evolution at the cell-type level.

Our study also revealed key evolutionary differences between the neocortex and SC. In the neocortex, the proportion of neurons expressing inhibitory markers increases from mouse to marmoset to human-^15,65 but see also ref. ⁶⁹. In contrast, our study finds that the ratio of inhibitory to excitatory neurons is relatively stable across mouse, tree shrew, and human SC. This likely reflects regional differences in the importance of excitatory-inhibitory balance across mammals.

Across mammalian species, neocortical neurons differ highly in their expression of genes related to synapse function^25,63. This is true even when comparing neocortical neurons of humans and non-human primates^23,26. The neocortex also differs across mammalian species in terms of its microcircuitry^27,70. Such differences in synaptic circuitry likely require changes in neuronal expression of synaptogenic and synapse adhesion genes. On the other hand, conservation of synaptic circuitry in other regions or cell types may require similar conservation of synapse-related genes. Our study found that, in contrast to the neocortex, synapse-related genes are highly conserved across mouse, tree shrew, and human SC neurons. Consistent with this, we also found that the molecular identity of pulvinar-projecting SC neurons is highly conserved between mice and tree shrews. Thus, in contrast with the neocortex, the SC exemplifies co-conservation of synaptic circuitry and gene expression. However, an important limitation to note in our study is that we assessed synaptic circuitry by projection patterns, not by connectivity. Future studies comparing the synaptic partners of SC neurons across species will fully determine the conservation of their synaptic circuitry.

On the other hand, by comparing SC neurons across species, our study identified an evolutionary divergence in the expression of genes related to the primary cilium. The primary cilium is a sensory organelle which transduces a variety of extracellular signals into changes in gene expression through a specialized connection with the cell nucleus, the basal body^53,71. Also, primary cilia are located close to chemical synapses and can directly form junctions with axonal segments or interact with synapses to modulate neuronal activity and synaptic homeostasis^52,72. Our study detected a progressive enrichment, from mice to tree shrews to humans, of transcripts related to ciliary signal detection, intracellular transport, and transcriptional regulation, suggesting a more dynamic and complex role for primary cilia in neuronal signaling in the SC of higher mammalian species. Notably, genes associated with cilia organization, assembly, and regulation were enriched in the SC neurons of humans relative to tree shrews, and in tree shrews relative to mice, suggesting that the primary cilium of SC neurons of higher order mammals have evolved a more dynamic and complex structure. In other brain regions, such as the hypothalamus and prefrontal cortex, primary cilia are important for metabolic homeostasis⁷³ and stress responses⁷⁴, respectively. However, while previous studies have identified primary cilia in the SC⁷⁵, their significance and role in SC function has not been explored.

Humans, tree shrews, and mice evolved from a common mammalian ancestor that was most likely nocturnal^76,77. The evolutionary shift to diurnal activity placed new selective pressures on the visual system, including the SC. Since the SC is conserved in many ways between nocturnal and diurnal species, including its laminar organization, input-output circuitry, and morphological cell types, it must have evolved in other ways to meet the increased visual demands of daytime behavior. Previous studies have shown, for instance, differences in SC size and dendritic architecture between nocturnal and diurnal rodents⁷⁸. Our findings suggest that the SC also evolved through the diversification of inhibitory neurons and the elaboration of primary cilia structure and function. These developments may have enhanced and diversified signaling within the SC while largely preserving its cell types and synaptic circuitry. In addition, we observed higher expression of cilia-related genes in the human SC compared to the tree shrew, raising the possibility that the greater reliance on vision in human ancestors drove further evolution of SC primary cilia.

In conclusion, our study provides a foundation for future comparative studies of SC neuron function. For instance, our results demonstrate that molecular markers for cell types and neuron subtypes vary across species to a greater extent than other genes. This underscores the importance of finding conserved genetic access points for investigating homologous cell types, especially given the recent development of enhancer-driven viral tools for genetically manipulating cell types in different species^79,80. Our study also raises questions for future investigation. For instance, since synapse-related genes are conserved in the SC but not in the neocortex, how have circuits between these two regions evolved? Future studies could assess the conservation of synapse specificity transcripts in the cortico-collicular circuit across species. Another question raised by our study is, what is driving the regional and cell type divergence in gene expression in the SC? Comparative epigenetic studies could identify changes in the presence or activity of gene regulatory elements that explain how gene expression has diverged during the evolution of this brain region.

Methods

Animals

A total of 9 adult male and female wild-type tree shrews (locally bred at the University of Virginia animal facility), aged 6 to 16 months, were used in this study. Specifically, one male and one female tree shrew were used for snRNA-seq of the superficial superior colliculus (sSC); three for retrograde tracer injections into the pulvinar; and another four for RNA fluorescence in situ hybridization (RNA FISH). In addition, 21 male and female mice aged 2–8 months were used, including H2b-TRAP (Jackson Laboratory, Strain #029789; RRID: IMSR_JAX: 029789) and C57Bl/6 J (Strain #000664; RRID: IMSR_JAX: 000664). Of these, nine H2b-TRAP mice were used for snRNA-seq of pulvinar-projecting SC neurons; six were used for RNA FISH; three for CTB injections to confirm molecular identity of the pulvinar-projection neurons; and three for injections of AAV-Camk2a-GFP.

All tree shrews and mice were kept on a 12 hr light/dark cycle under standard feeding and housing conditions, with 1 to 2 tree shrews housed per cage and 2 to 5 mice housed per cage. The temperature was kept between 68–79 °F, with relative humidity maintained at 30–70%. All experimental procedures were approved by the University of Virginia Institutional Animal Care and Use Committee under protocol numbers 4172 (mice) and 4244 (tree shrews).

Single-nuclei RNA-sequencing of tree shrew sSC

Tree shrews were anesthetized first with isoflurane and then with a lethal dose of Euthasol. Brains were rapidly dissected, cooled in ice-cold Hibernate-A for 1.5 min, transferred to PBS, and prepared under a dissecting microscope. The cortex and cerebellum were removed to expose the superior colliculus (SC). The sSC was visualized coronally, micro-dissected by knife cuts along its visually approximated anatomical borders and processed into a single-nucleus suspension using a published detergent-mechanical lysis protocol⁸¹. Briefly, the sSC was homogenized in a lysis buffer, and cell nuclei were purified by density gradient centrifugation. After resuspension, nuclei were labeled with propidium iodide (Propidium Iodide Ready Flow™ Reagent, Invitrogen, catalog # R37169) and then counted with a CellDrop automated cell counter (DeNovix, Delaware, USA).

We used the 10X Genomics Chromium Next GEM 3’ kit v3.1 according to the manufacturer’s protocol (user guide revision D) to encapsulate each nucleus’s poly-adenylated RNA and then process the poly-adenylated RNA into cDNA sequencing libraries. We then sequenced the cDNA libraries by Illumina Next-Seq 2000 to an average of 29,857 mean reads per cell. We loaded approximately 16,500 cell nuclei per lane of the 10X chip. A total of 4 single-nuclei samples were processed across 2 10X chips (batches): samples 1 and 2 were aliquots of the same nuclei suspension from a female tree shrew and processed in parallel as batch 1; samples 3 and 4 were aliquots of the same nuclei suspension from a male tree shrew processed in parallel as batch 2. Finally, the 10X Cell Ranger pipeline (version 7.2.0) was used to demultiplex and align sequencing reads to a custom “pre-mRNA” reference genome created from the published tree shrew database (http://www.treeshrewdb.org/) to generate feature-barcode matrices for each batch of tree shrew sSC.

Integrating snRNA-seq datasets from the mouse and tree shrew sSC

The mouse sSC feature-barcode matrices were obtained from our previous publication³⁸. Both the tree shrew and mouse feature-barcode matrices were analyzed using R (version 4.3.1) with the Seurat 5.0.3 software package for R⁸². For the tree shrew samples, quality control was performed on both batches, filtering out genes detected in fewer than three cells and cells with fewer than 200 genes, more than 6500 genes, or over 21,000 transcripts. We filtered the mouse dataset as previously described³⁸ - in general, filtering out genes detected in fewer than three cells and cells with fewer than 200 genes, more than 3500 genes, or over 7200 transcripts. To place cells from the two species into the same gene expression space, the tree shrew and mouse matrices were filtered so that only orthologous genes were used for dataset integration, based on the orthologous gene list provided by Dr. Maosen Ye from the Kunming Institute of Zoology.

Next, we integrated the tree shrew and mouse matrices to correct for technical variance, including batch effects. In brief, we log-normalized the data; selected 2000 most variable genes for each matrix; found anchor correspondences across matrices with canonical correlation analysis (CCA); integrated the matrices using the IntegrateData() function in Seurat; scaled each gene; performed Principal Component Analysis (PCA) to linearly reduce the dimensionality of the highly variable gene set; clustered the cells using the Louvain algorithm, based on Euclidean distance in the PCA space comprising the first 10 PCs and with a resolution value of 1; reordered the cluster ID using the BuildClusterTree() function; and performed non-linear dimensionality reduction by Uniform Manifold Approximation and Projection (UMAP). Cell types were annotated according to their expression of classical cell type marker genes: neurons (Syn1+ , Rbfox3+ ); oligodendrocyte lineage cells (Olig1+ , Cspg4+ ); astrocytes (Agt+ ); endothelial cells (Cldn5+ ); macrophages (Cx3cr1+ , Mrc1+ ); and leptomeningeal cells (Lum+ ; Fig. 1A and Supplementary Fig. S1C. Clusters potentially representing cell doublets were identified based on cluster- and cell-level co-expression of these marker genes. After removing the suspect clusters, the remaining cells were re-clustered in two additional rounds to remove remaining potential doublets, following the steps of log normalization, feature selection, anchor-based integration, PCA, and clustering (using the first 10 PCs based on the elbow plot, and resolution set to 1).

Clusters with relatively high expression of the neuronal marker genes Syn1 and Rbfox3, but little to no expression of other cell type marker genes, were annotated as neurons, then subset and re-clustered for another two additional rounds to remove suspected cell doublets. In the first round of neuronal re-clustering, the top 15 PCs and a resolution value of 1 were used, while the second round used the top 14 PCs with a resolution value of 0.9. A total of 33 neuronal clusters were identified from the tree shrew and mouse integrated dataset (hereafter referred to as the TM dataset). Genes differentially expressed across clusters were identified using the Wilcoxon Rank Sum test and filtered to include only those detected in at least 10% of cells in the cluster. Candidate marker genes were selected based on their expression least detected by other clusters and highly expressed in the given clusters.

Clustering relationships across various parameters and integrative methods

To explore the relationship between neuronal subtypes in the TM dataset and mouse sSC datasets, individual mouse neurons were matched between these datasets according to 10X cell-specific barcodes. The corresponding neurons were visualized using the givsSankey() function based on the ‘googleVis’ package (Supplementary Fig. S2A, B). To reduce the technical bias, the clustering analysis was also performed independently for the mouse or the tree shrew neurons while using the orthologous genes and keeping the same PCs and resolution as in the TM dataset. Corresponding cells of the integrated and species-specific datasets were also visualized with the Sankey plot (Supplementary Fig. S2C–F).

Given that the TM dataset includes more neurons from mice (27,497) than from tree shrews (15,443), we assessed whether the larger sample size might influence the clustering results of the integrated data. An equivalent number of mouse neurons were randomly selected and integrated with tree shrew neurons using the same parameters as in the TM dataset (Supplementary Fig. S2G, H). In addition, resolution parameters were tested varying from 0.1 to 1.8 for clustering with the same 14 PCs, as a high resolution might lead to overfitting (Supplementary Fig. S6A). Moreover, while maintaining the same PCs and resolution, different normalization and integration methods were tested, such as SCTransform normalization with the SCTransform() function, Harmony with the IntegrateLayers() function, label transfer annotation with the TransferData() function, and reciprocal principal component analysis (rPCA) reduction with the FindIntegrationAnchors() function (Supplementary Fig. S6C–G).

Integrating human, tree shrew, and mouse SC datasets

The human SC dataset was obtained from the Human Brain Cell Atlas v1.0 via the CELLxGENE platform (https://cellxgene.cziscience.com/collections/283d65eb-dd53-496d-adb7-7570c7caa443). This dataset was generated in a previous study¹ from three adult male donors using the 10X Chromium Next GEM 3’ reagent kit v3 for sequencing library preparation and sequenced on an Illumina NovaSeq platform. Sequencing reads were mapped to the human reference genome GRCh38.p13 (Gencode V35) to generate a feature-barcode matrix using Cell Ranger version 4.0.0. After obtaining this human SC snRNA-seq dataset, we first converted its features from Ensembl IDs to gene symbols using the “org.Hs.eg.db” package in R, and then further converted it into mouse orthologous genes using the convert_human_to_mouse_symbols() function in the “nichenetr” package⁸³.

Next, we conducted quality control on the human SC dataset based on its quality metric distribution. Cells with more than 8000 detected genes, over 60,000 transcripts, or a mitochondrial gene fraction greater than 0.3 were filtered out. Subsequently, a similar clustering approach was used to identify cell types within the entire human SC dataset, including log normalization, feature selection, data scaling, PCA, clustering (using the top 9 PCs based on the elbow plot and a resolution of 1), and identification of marker gene expression. Clusters with relatively high expression of the neuronal marker genes Syn1 and Rbfox3 were subset and re-clustered to further exclude potential cell doublets, using the top 6 PCs and a resolution of 1.

The human neuronal feature-barcode matrix was then integrated into the TM dataset. This process involved log normalization, feature selection, anchor identification using CCA, matrix integration, scaling, PCA, and clustering (using the top 11 PCs with a resolution of 1). As a result, we identified 31 neuronal subtypes (hereafter referred to as the HTM dataset). Candidate marker genes were selected based on their expression in more than 10% of cells of the given cluster and little to no detection in other clusters. The avian optic tectum integrative snRNA-seq dataset and homologous gene list were kindly shared by Dr. Zhenkun Zhuang⁸⁴. We first matched the bird genes in this dataset to homologous mouse genes, then integrated the datasets into the HTM dataset using the same integrative procedure and parameters, resulting in 25 neuronal subtypes (hereafter referred to as the HTMB dataset).

Analysis of conserved and divergent genes across clusters and species

To compare gene expression across species, we conducted a differentially enriched gene (DEG) analysis using Seurat’s FindMarker() function with the Wilcoxon Rank Sum test. To ensure all genes were included, we adjusted the default settings by setting the minimum fraction of detected genes in either species to 0 and the log2 fold change threshold to 0. Due to minor differences between the list of orthologous genes in mice and tree shrews compared to mice and humans, only orthologous genes were included in this analysis. Specifically, 15,567 orthologous genes between tree shrews and mice were used for the TM dataset; 14,900 orthologous genes across the three species were used for the HTM dataset; and 8234 homologous genes across the zebra finch, pigeon, mouse, tree shrew, and human genomes were used for the HTMB dataset. In addition, we confirmed that using 15,567 orthologous genes (TM dataset) instead of 14,900 orthologous genes (HTM dataset) did not impact the three-species integration results.

The cross-species DEG analysis yielded the log2 fold change of expression means between species (hereafter referred to as “species specificity index (SSI)”), the percentage of cells in which the gene was detected in each species, and a p-value adjusted according to Bonferroni correction. Positive SSI values indicate enrichment in tree shrews, whereas negative SSI values indicate enrichment in mice. To predict the functional terms of the most tree shrew-enriched genes in the TM dataset, we first tried the top 500 genes with the highest SSI (log2FC) for gene ontology analysis – however, this yielded no valid GO term. Next, we increased the number of genes to about 1000, which only resulted in a few GO terms. Therefore, we continued to increase the number of genes to about 1600 by reducing the SSI cutoff to 4. To obtain the same number of divergent and conserved genes, we set the SSI cutoff to − 9 for mouse-enriched genes and absolute SSI values to 0.27 for the most conserved genes between mice and tree shrews. On the other hand, to identify divergent and conserved genes in the HTM dataset, we performed DEG analysis comparing gene expression in human neurons with those in tree shrews and mice. To include a similar number of genes in our HTM analysis as in our TM analysis, we selected the top 1500 genes with the highest SSI values as human-enriched genes, whereas those with the lowest absolute SSI values were identified as shared genes.

To identify genes with progressively enriched expression from mouse to tree shrew to human SC neurons, we performed Seurat’s FindMarker() function between mice and tree shrews and between tree shrews and humans. Then we selected genes with adjusted p-values smaller than 0.05 (False Discovery Rate) and with log2FC values larger than 0.25 (Seurat default cut off, which is about 19% increased expression), resulting in 5911 genes enriched in tree shrews relative to mice and 6942 genes enriched in humans relative to tree shrews. A total of 1152 overlapping genes were identified between the two groups (tree shrew vs. mouse, human vs. tree shrew). On the other hand, we identified 464 genes with gradually decreased expression from the mouse SC to tree shrew, and to human SC by selecting genes with log2FC values smaller than −0.25 and passing the false discovery rate threshold 0.05.

To predict the functions of conserved and species-enriched genes, we performed Gene Ontology (GO) enrichment analysis using the enrichGO() function from the “clusterProfiler” and “org.Mm.eg.db” packages in R⁸⁵. We selected the top 20 GO terms with the smallest adjusted p-value and further selected about 10 of them that show fewer overlapped genes across terms and might be connected to parent predicted terms. Moreover, to confirm that the shared genes were related to synapse function, we conducted a synGO analysis, which is designed specifically for synapse function (https://www.syngoportal.org/)³⁹. On the other hand, we examined distribution of SSI values on selected GO terms between tree shrews and mice or across three species, including: synapse (GO:0045202); postsynapse organization (GO:0099173); cell projection organization (GO:0030030); cilium movement (GO:0003341); secretory granule (GO:0030141); non-motile cilium assembly (GO:1905515); non-motile cilium (GO:0097730); and positive regulation of cilium movement (GO:0003353).

To calculate the log2 fold change for all genes across clusters (cluster Log2FC) in the TM and HTM datasets, the FindAllMarkers() function in Seurat was used with both the minimum detected percentage and log2FC threshold set to 0. The Log2FC was calculated for each gene expressed in a cluster compared to all other clusters, resulting in a table of genes with either positive (higher expression) and negative (lower expression) log2FC values. This process was repeated for each cluster, generated a corresponding number of marker gene tables, and then merged all the tables. Thus, the same genes in the merged table appeared many times, matching the number of clusters, with various log2FC values, including positive and negative ones. Genes in the merged table were then re-ordered based on their Log2FC values, retaining only the first occurrence of duplicated genes with the highest values, and removed the other duplicates. Therefore, only genes with positive log2FC values were retained. In addition to the relative expression level (log2FC), the top 50 cluster markers were selected for each subtype based on the percentage of neurons in which expression of the gene was detected, inside vs. outside the selected cluster. This identified a total of 982 cluster marker genes that were detected in more than 10% of neurons inside the selected cluster and the minimal outside.

snRNA-seq of pulvinar-projecting neurons in mouse SC

Four weeks after pENN.AAV.hSyn.Cre.WPRE.hGH (Addgene, Catalog # 105553-AAVrg) was injected into the pulvinar of H2B-TRAP mice, their brains were harvested following rapid decapitation to minimize the effects of stress on gene expression. The brains were cooled in ice-cold PBS for 1.5 min, placed in a chilled stainless steel brain matrix, and coronally sectioned into 1 mm thick slices containing the sSC (Bregma −3.0 mm to −5.0 mm). The slices were immediately immersed in ice-cold RNAprotect reagent (Qiagen) and incubated at 4 °C overnight to stabilize RNA. The following day, the sSC was visualized under a fluorescent dissecting microscope, micro-dissected using knife cuts, and processed into single-nuclei suspensions using a detergent-mechanical lysis protocol. After resuspending the nuclei pellet, DRAQ5 was added, and mCherry-positive nuclei were sorted using a Becton Dickinson Influx Cell Sorter. Poly-adenylated RNA released from each sorted and encapsulated cell nucleus was used to generate cDNA sequencing libraries with the 10X Chromium Next GEM 3’ reagent kit v3.1, following the user guide revision D protocol. cDNA libraries were then sequenced on an Illumina NextSeq 2000 platform to an average depth of 28,135 reads per cell. After that, Cell Ranger (version 5.0.0) was used to map sequence reads to a custom “pre-mRNA” reference genome, built from the mouse reference genome GRCm38.98, to generate a feature-barcode matrix.

The feature-barcode matrix was analyzed using the Seurat 5.0.3 package⁸² in R (version 4.3.1). Genes detected in fewer than three cells, and cells with fewer than 200 or more than 5000 detected genes, over 12,500 UMIs, or more than 2% mitochondrial content were filtered out. Then, a similar procedure for clustering was conducted, including log normalization, feature selection, scaling, PCA, clustering with the first 9 PCs and a resolution of 0.5, and detected expression of cell type markers. Clusters with enriched expression of the neuronal markers Syn1 and Rbfox3 were subset and re-clustered to identify neuronal subtypes (re-clustering parameters: first 15 PCs and resolution at 0.3). To determine the identities of these neuronal subtypes, we used the label transfer annotation method to map them to the TM dataset, with the FindTransferAnchors() and TransferData() functions in Seurat. A confidence score at 0.8 was applied to minimize technical bias when assigning cells to the reference atlas.

Stereotaxic injection of tracers and viral vectors

Mice were anesthetized with 2% isoflurane both before and during surgery and positioned in a stereotaxic apparatus. A small midline incision was made to expose the skull, and the skull was adjusted to ensure it was flat by aligning Bregma and Lambda (Anterior-Posterior) at the same level, as well as the left and right sides (Medial-Lateral). The injection site was marked based on the targeted coordinates relative to Bregma, and the skull at the marked location was carefully drilled. Three sets of injections were then performed with a pressure injector (Drummond Nanoject II, Cat# 3000204): (1) 13.8 nl of pENN.AAV.hSyn.Cre.WPRE.hGH (Addgene, Catalog # 105553-AAVrg) was injected into the pulvinar (ML: 1.4 mm; AP: −2.3 mm; DV: −2.55 mm) of H2B-TRAP mice to label pulvinar-projecting neurons; (2) three injections of 46 nL of pAAV-CaMKIIa-EGFP (Addgene, Catalog #50469-AAV9) were delivered into the cortex (AP: 0.8 mm; ML: − 3.8 mm; DV: 0.7 mm) and two sites in the SC (ML: 0.8 mm; AP: −3.8 mm; DV: 1.2 & 1.7 mm) of WT mice; and (3) 45 nl of Cholera Toxin Subunit B, conjugated with Alexa Fluor 555 (5 mg/mL; Invitrogen by Thermo Fisher Scientific, Catalog # C34776) was injected into the pulvinar (ML: 1.4 mm; AP: −2.3 mm; DV: −2.55 mm) of WT mice. After that, the holes in the skull were filled with wax, the scalp was closed with sutures, and mice were injected with 2% carprofen.

Tree shrews were initially anesthetized with 5% isoflurane, followed by intraperitoneal injections of pre-surgical medications including midazolam (5 mg/kg), atropine (0.3 mg/kg), and dexamethasone (1 mg/kg). Afterward, the animals were positioned in a stereotaxic apparatus with maintenance anesthesia at approximately 3% isoflurane. EKG pads were attached to the limbs to monitor heart activity, and once stable EKG signals were recorded, the animals were intubated. A similar stereotaxic procedure used for mice was then applied to the tree shrews, including scalp incision, Bregma-Lambda alignment, marking target coordinates, and drilling at the marked location. pAAV-CAG-tdTomato viral vector (Addgene, Catalog # 59462-AAVrg) was injected into three sites of the pulvinar (ML: 3 mm; AP: −4.28 mm; DV: 5.00/5.25/5.55 mm) of the tree shrew, with 200 nL delivered at each depth. The injection sites were sealed with wax, and the scalp was closed with sutures. Following the surgery, the animals were administered post-surgical medications (Buprenorphine: 0.05 mg/kg; Enrofloxacin: 5 mg/kg) and placed on a heating pad until fully recovered from anesthesia.

Perfusion and histology

Tree shrews were initially anesthetized with isoflurane in a chamber, followed by deep anesthesia with an intraperitoneal injection of a lethal dose of Euthasol. Mice were anesthetized directly with intraperitoneal Euthasol. All animals were then perfused with PBS, followed by 4% paraformaldehyde (PFA). Brains were dissected immediately, fixed in 4% PFA overnight, dehydrated in 30% sucrose for three days, embedded in O.C.T. compound, frozen at −20 °C, and sectioned coronally at 40 μm thickness using a cryostat (Leica CM1950). The slices were finally collected and stored in cryoprotectant at −20 °C.

RNA fluorescence in situ hybridization (FISH)

Slices removed from cryoprotectant were washed twice in phosphate-buffered saline (PBS; pH 7.2) for 5 min each, then incubated with hydrogen peroxide for 10 minutes within a well plate. After incubation, the slices were washed three times in PBS to remove bubbles and mounted onto slides (Fisher Cat. 12-550-15). The mouse slices were air-dried overnight in preparation for RNA FISH the next day. In contrast, slices from tree shrews underwent target retrieval before FISH. Specifically, tree shrew slices were baked in an oven at 40 °C for 30 min, incubated in 4% PFA at 4 °C for 15 minutes, and dehydrated sequentially in 50%, 70%, and 100% ethanol at room temperature for 5 min each. The slices were then air-dried for 5 min, incubated in distilled water at 99 °C for 10 s, followed by antigen retrieval solution at 99 °C for 5 min, rinsed in distilled water for 15 s at room temperature, and finally dehydrated in 100% ethanol for 3 min.

On the following day, RNAscope Multiplex v2 (Cat. 323100) was used to stain specific messenger RNA transcripts in the coronal sections, following the manufacturer’s protocol. Briefly, slide-mounted slices were circled with a hydrophobic barrier, rinsed with distilled water, digested with Protease IV at 40 °C for 30 min, and incubated with targeted probes at 40 °C for 2 h. The slices were then treated with amplification reagents (Amp 1-3) at 40 °C. The probes used were: Slc17a6 (vGluT2); Slc32a1 (vGAT); Gad1; Camk2a; Cbln2; Cbln4; Sst; Npnt; Vip; Sntb1; Postn; and tdTomato. Following this, HRP-C1/C2/C3, FITC/Cy3/Cy5 dyes, and HRP blocker were applied to bind to Amp 3 and activate fluorescence. Finally, DAPI was used as a counterstain, and fluorescence images were captured using a confocal microscope (20X/0.80 Plan Apo, Zeiss LSM 800).

RNA FISH analysis

To reveal the spatial organization of gene expression in the tree shrew and mouse superior colliculus (SC), the layering structures were outlined using the Allen Mouse Brain Atlas⁸⁶ (https://mouse.brain-map.org/) as a reference and the stereotaxic coordinates of the tree shrew brain. Based on these structural outlines and the density of detected expression of inhibitory and excitatory markers, a correspondence in depth was identified between the two species (Fig. 2D). Specifically, the SO sublayer in the tree shrew SC begins at approximately 500 µm and ends around 800 µm from the surface, whereas in mice, it starts at about 300 µm and ends around 500 µm (Fig. 2F, G).

To investigate the spatial distribution of inhibitory and excitatory neurons along the depth of the SC in tree shrews, RNA FISH experiments were separately performed for Gad1 and vGluT2 on different tree shrew slices. This was necessary to prevent signal leaking between channels during confocal imaging due to their robust expression. Then, a 1400 μm × 400 μm white rectangle was used to quantify gene expression along the depth of the tree shrew SC, which was further divided into 14 subregions, each measuring 100 μm in length and 400 μm in width. In mice, the corresponding rectangle measured 800 μm × 250 μm and was divided into 16 subregions, each 50 μm in length and 250 μm in width.

Separately performing RNA FISH for Gad1 and vGluT2 in different animals and slices could add variance in the percentage of cells expressing inhibitory or excitatory markers along the depth. Therefore, the percentage of inhibitory or excitatory cells in each subregion was calculated relative to the total detected cells within the entire rectangle. This subregion percentage was then averaged across different slices from multiple animals, thus generating an average percentage of inhibitory/excitatory neurons in each subregion compared to the whole rectangle. Then the number of inhibitory/excitatory neurons in the rectangle was quantified and averaged across slices and species, therefore separately creating the average number of inhibitory or excitatory neurons in the entire rectangle. By multiplying the average percentage of inhibitory/excitatory neurons in each subarea with the average number of inhibitory/excitatory neurons in the whole rectangle, the number of inhibitory/excitatory neurons in each subarea was obtained. Finally, dividing the number of inhibitory/excitatory neurons by their sum yielded the percentage of inhibitory/excitatory neurons in each subarea along the depth. A similar analysis was conducted for mice, though vGAT and vGluT2 expression were detected in the same brain sections. This analysis in mice produced the same percentage of inhibitory and excitatory neurons along the depth of the SC as the direct quantification of vGAT and vGluT2 in the same slices.

To assess the cellular co-localization of transcripts in tree shrews, 500 μm × 500 μm or 300 μm × 300 μm squares (Regions of Interest, ROIs) were outlined from the middle to lateral regions at the surface or deeper layer of the SC. A similar approach was used in mice, where 300 μm × 300 μm squares were outlined from the middle to lateral regions for quantification.

Statistics and reproducibility

All experiments were repeated independently at least twice with similar results.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

● The raw and quantified snRNA-seq data are available at the Gene Expression Omnibus (GEO) repository at GSE287986. ● The processed data are available at Zenodo at this link: https://doi.org/10.5281/zenodo.15932065. ●The clustered datasets are available for user-friendly visualization and exploration through the Broad Single Cell Portal at these links: ○ tree shrew/mouse (TM dataset), https://singlecell.broadinstitute.org/single_cell/study/SCP3231/co-conservation-of-synaptic-gene-expression-and-circuitry-in-collicular-neurons-tm-dataset#study-visualize; ○ human/tree shrew/mouse (HTM dataset), https://singlecell.broadinstitute.org/single_cell/study/SCP3234/co-conservation-of-synaptic-gene-expression-and-circuitry-in-collicular-neurons-htm-dataset. ● All other raw and processed data are available from the corresponding authors upon request. Source data are provided in this paper.

Code availability

The code used to process and analyze the snRNA-seq data is available at Zenodo at this link: https://doi.org/10.5281/zenodo.15932065.

References

Siletti, K. et al. Transcriptomic diversity of cell types across the adult human brain. Science 382, eadd7046 (2023).
Article CAS PubMed Google Scholar
Reiner, A. The triune brain in evolution. role in paleocerebral functions. Paul D. MacLean. Plenum, New York, 1990. xxiv, 672 pp., illus. $75. Science 250, 303–305 (1990).
Article CAS PubMed Google Scholar
Ramón Y. Cajal, S. Histology of the Nervous System of Man and Vertebrates. (Oxford University Press, 1995).
Adrian, E. D. The impulses produced by sensory nerve endings: Part I. J. Physiol. 61, 49–72 (1926).
Article CAS PubMed PubMed Central Google Scholar
Roberts, R. J. V., Pop, S. & Prieto-Godino, L. L. Evolution of central neural circuits: state of the art and perspectives. Nat. Rev. Neurosci. 23, 725–743 (2022).
Article CAS PubMed Google Scholar
Britten, R. J. & Davidson, E. H. Gene regulation for higher cells: a theory. Science 165, 349–357 (1969).
Article ADS CAS PubMed Google Scholar
King, M. C. & Wilson, A. C. Evolution at two levels in humans and chimpanzees. Science 188, 107–116 (1975).
Article ADS CAS PubMed Google Scholar
Wolfe, K. H. & Li, W. H. Molecular evolution meets the genomics revolution. Nat. Genet. 33, 255–265 (2003).
Article CAS PubMed Google Scholar
Sun, T. & Hevner, R. F. Growth and folding of the mammalian cerebral cortex: from molecules to malformations. Nat. Rev. Neurosci. 15, 217–232 (2014).
Article CAS PubMed PubMed Central Google Scholar
Somel, M., Liu, X. & Khaitovich, P. Human brain evolution: transcripts, metabolites and their regulators. Nat. Rev. Neurosci. 14, 112–127 (2013).
Article CAS PubMed Google Scholar
Genome, K. C. O. S. Genome 10K: a proposal to obtain whole-genome sequence for 10,000 vertebrate species. J. Hered. 100, 659–674 (2009).
Zhuang, X. L. et al. Divergent evolutionary rates of primate brain regions as revealed by genomics and transcriptomics. Genome Biol. Evol. 16, evae023 (2024).
Article PubMed PubMed Central Google Scholar
La Manno, G. et al. Molecular diversity of midbrain development in mouse, human, and stem cells. Cell 167, 566–580 (2016).
Article PubMed PubMed Central Google Scholar
Saunders, A. et al. Molecular diversity and specializations among the cells of the adult mouse brain. Cell 174, 1015–1030.e1016 (2018).
Article CAS PubMed PubMed Central Google Scholar
Network, B. I. C. C. A multimodal cell census and atlas of the mammalian primary motor cortex. Nature 598, 86–102 (2021).
Article Google Scholar
Shi, Y. et al. Mouse and human share conserved transcriptional programs for interneuron development. Science 374, eabj6641 (2021).
Article CAS PubMed PubMed Central Google Scholar
Kebschull, J. M. et al. Cerebellar nuclei evolved by repeatedly duplicating a conserved cell-type set. Science 370, eabd5059 (2020).
Article CAS PubMed PubMed Central Google Scholar
Hain, D. et al. Molecular diversity and evolution of neuron types in the amniote brain. Science 377, eabp8202 (2022).
Article CAS PubMed Google Scholar
Caglayan, E. et al. Molecular features driving cellular complexity of human brain evolution. Nature 620, 145–153 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Hahn, J. et al. Evolution of neuronal cell classes and types in the vertebrate retina. Nature 624, 415–424 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Lancaster, M. A. Unraveling mechanisms of human brain evolution. Cell 187, 5838–5857 (2024).
Article CAS PubMed PubMed Central Google Scholar
Berg, J. et al. Human neocortical expansion involves glutamatergic neuron diversification. Nature 598, 151–158 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Ma, S. et al. Molecular and cellular evolution of the primate dorsolateral prefrontal cortex. Science 377, eabo7257 (2022).
Article CAS PubMed PubMed Central Google Scholar
Sousa, A. M. M. et al. Molecular and cellular reorganization of neural circuits in the human lineage. Science 358, 1027–1032 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Hodge, R. D. et al. Conserved cell types with divergent features in human versus mouse cortex. Nature 573, 61–68 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Suresh, H. et al. Comparative single-cell transcriptomic analysis of primate brains highlights human-specific regulatory evolution. Nat. Ecol. Evol. 7, 1930–1943 (2023).
Article PubMed PubMed Central Google Scholar
Peng, Y. et al. Directed and acyclic synaptic connectivity in the human layer 2-3 cortical microcircuit. Science 384, 338–343 (2024).
Article ADS CAS PubMed Google Scholar
Vanderhaeghen, P. & Polleux, F. Developmental mechanisms underlying the evolution of human cortical circuits. Nat. Rev. Neurosci. 24, 213–232 (2023).
Article CAS PubMed PubMed Central Google Scholar
Cang, J., Chen, C., Li, C. & Liu, Y. Genetically defined neuron types underlying visuomotor transformation in the superior colliculus. Nat. Rev. Neurosci. 25, 726–739 (2024).
Article CAS PubMed Google Scholar
Basso, M. A., Bickford, M. E. & Cang, J. Unraveling circuits of visual perception and cognition through the superior colliculus. Neuron 109, 918–937 (2021).
Article CAS PubMed PubMed Central Google Scholar
Basso, M. A. & May, P. J. Circuits for action and cognition: a view from the superior colliculus. Annu. Rev. Vis. Sci. 3, 197–226 (2017).
Article PubMed PubMed Central Google Scholar
Isa, T., Marquez-Legorreta, E., Grillner, S. & Scott, E. K. The tectum/superior colliculus as the vertebrate solution for spatial sensory integration and action. Curr. Biol. 31, R741–R762 (2021).
Article CAS PubMed PubMed Central Google Scholar
Liu, X. et al. The superior colliculus: cell types, connectivity, and behavior. Neurosci. Bull. 38, 1519–1540 (2022).
Article PubMed PubMed Central Google Scholar
Fan, Y. et al. Genome of the Chinese tree shrew. Nat. Commun. 4, 1426 (2013).
Article ADS PubMed Google Scholar
Savier, E., Sedigh-Sarvestani, M., Wimmer, R. & Fitzpatrick, D. A bright future for the tree shrew in neuroscience research: Summary from the inaugural Tree Shrew Users Meeting. Zool. Res. 42, 478–481 (2021).
Article PubMed PubMed Central Google Scholar
Yao, Y. G. et al. Study of tree shrew biology and models: A booming and prosperous field for biomedical research. Zool. Res. 45, 877–909 (2024).
Article PubMed PubMed Central Google Scholar
Campbell, C. B. G. The relationships of the tree shrews: the evidence of the nervous system. Evolution 20, 276–281 (1966).
Article CAS PubMed Google Scholar
Liu, Y. et al. Mapping visual functions onto molecular cell types in the mouse superior colliculus. Neuron 111, 1876–1886 (2023).
Article CAS PubMed PubMed Central Google Scholar
Koopmans, F. et al. SynGO: An evidence-based, expert-curated knowledge base for the synapse. Neuron 103, 217–234 (2019).
Article CAS PubMed PubMed Central Google Scholar
Yasuda, R., Hayashi, Y. & Hell, J. W. CaMKII: a central molecular organizer of synaptic plasticity, learning and memory. Nat. Rev. Neurosci. 23, 666–682 (2022).
Article CAS PubMed Google Scholar
Shibata, M. et al. Hominini-specific regulation of CBLN2 increases prefrontal spinogenesis. Nature 598, 489–494 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Sudhof, T. C. Cerebellin-neurexin complexes instructing synapse properties. Curr. Opin. Neurobiol. 81, 102727 (2023).
Article CAS PubMed Google Scholar
Dai, J., Liakath-Ali, K., Golf, S. R. & Sudhof, T. C. Distinct neurexin-cerebellin complexes control AMPA- and NMDA-receptor responses in a circuit-dependent manner. Elife 11, e78649 (2022).
Article CAS PubMed PubMed Central Google Scholar
Xie, Z. et al. Transcriptomic encoding of sensorimotor transformation in the midbrain. Elife 10, e69825 (2021).
Article CAS PubMed PubMed Central Google Scholar
Choi, J. S. et al. Single-nucleus RNA sequencing of developing superior colliculus identifies neuronal diversity and candidate mediators of circuit assembly. Cell Rep. 42, 113037 (2023).
Article CAS PubMed PubMed Central Google Scholar
Relota, X. J., Ford, A. & Savier, E. L. Behavioral modulation and molecular definition of wide-field vertical cells in the mouse superior colliculus. J. Neurosci. Off. J. Soc. Neurosci. 45, e1816242025 (2025).
Article CAS Google Scholar
Chen, C., Liu, Y. & Cang, J. Accessing genetically defined cell types in the superior colliculus with transgenic mouse lines. iScience 28, 112194 (2025).
Article ADS CAS PubMed PubMed Central Google Scholar
Major, D. E., Luksch, H. & Karten, H. J. Bottlebrush dendritic endings and large dendritic fields: motion-detecting neurons in the mammalian tectum. J. Comp. Neurol. 423, 243–260 (2000).
Article CAS PubMed Google Scholar
Fredes, F., Vega-Zuniga, T., Karten, H. & Mpodozis, J. Bilateral and ipsilateral ascending tectopulvinar pathways in mammals: a study in the squirrel (Spermophilus beecheyi). J. Comp. Neurol. 520, 1800–1818 (2012).
Article PubMed PubMed Central Google Scholar
Luksch, H., Cox, K. & Karten, H. J. Bottlebrush dendritic endings and large dendritic fields: motion-detecting neurons in the tectofugal pathway. J. Comp. Neurol. 396, 399–414 (1998).
Article CAS PubMed Google Scholar
Zhou, N. A., Maire, P. S., Masterson, S. P. & Bickford, M. E. The mouse pulvinar nucleus: Organization of the tectorecipient zones. Vis. Neurosci. 34, E011 (2017).
Article CAS PubMed PubMed Central Google Scholar
Wu, J. Y. et al. Mapping of neuronal and glial primary cilia contactome and connectome in the human cerebral cortex. Neuron 112, 41–55 (2024).
Article CAS PubMed Google Scholar
Jurisch-Yaksi, N., Wachten, D. & Gopalakrishnan, J. The neuronal cilium - a highly diverse and dynamic organelle involved in sensory detection and neuromodulation. Trends Neurosci. 47, 383–394 (2024).
Article CAS PubMed Google Scholar
Youn, Y. H. & Han, Y. G. Primary cilia in brain development and diseases. Am. J. Pathol. 188, 11–22 (2018).
Article CAS PubMed PubMed Central Google Scholar
Tsai, N. Y. et al. Trans-Seq maps a selective mammalian retinotectal synapse instructed by Nephronectin. Nat. Neurosci. 25, 659–674 (2022).
Article CAS PubMed PubMed Central Google Scholar
de Wit, J. & Ghosh, A. Specification of synaptic connectivity by cell surface interactions. Nat. Rev. Neurosci. 17, 22–35 (2016).
PubMed Google Scholar
Nonaka, S. et al. Randomization of left-right asymmetry due to loss of nodal cilia generating leftward flow of extraembryonic fluid in mice lacking KIF3B motor protein. Cell 95, 829–837 (1998).
Article CAS PubMed Google Scholar
Satir, P. & Christensen, S. T. Overview of structure and function of mammalian cilia. Annu. Rev. Physiol. 69, 377–400 (2007).
Article CAS PubMed Google Scholar
Cang, J. & Feldheim, D. A. Developmental mechanisms of topographic map formation and alignment. Annu. Rev. Neurosci. 36, 51–77 (2013).
Article CAS PubMed Google Scholar
Arendt, D. et al. The origin and evolution of cell types. Nat. Rev. Genet. 17, 744–757 (2016).
Article CAS PubMed Google Scholar
Pembroke, W. G., Hartl, C. L. & Geschwind, D. H. Evolutionary conservation and divergence of the human brain transcriptome. Genome Biol. 22, 52 (2021).
Article CAS PubMed PubMed Central Google Scholar
Berto, S. et al. Accelerated evolution of oligodendrocytes in the human brain. Proc. Natl. Acad. Sci. USA 116, 24334–24342 (2019).
Article ADS CAS PubMed PubMed Central Google Scholar
Jorstad, N. L. et al. Comparative transcriptomics reveals human-specific cortical features. Science 382, eade9516 (2023).
Article CAS PubMed PubMed Central Google Scholar
Sherwood, C. C. et al. Evolution of increased glia-neuron ratios in the human frontal cortex. Proc. Natl. Acad. Sci. USA 103, 13606–13611 (2006).
Article ADS CAS PubMed PubMed Central Google Scholar
Bakken, T. E. et al. Comparative cellular analysis of motor cortex in human, marmoset and mouse. Nature 598, 111–119 (2021).
Article ADS CAS PubMed PubMed Central Google Scholar
Boldog, E. et al. Transcriptomic and morphophysiological evidence for a specialized human cortical GABAergic cell type. Nat. Neurosci. 21, 1185–1195 (2018).
Article CAS PubMed PubMed Central Google Scholar
Krienen, F. M. et al. Innovations present in the primate interneuron repertoire. Nature 586, 262–269 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Chen, A. et al. Single-cell spatial transcriptome reveals cell-type organization in the macaque cortex. Cell 186, 3726–3743 (2023).
Article CAS PubMed Google Scholar
Jorstad, N. L. et al. Transcriptomic cytoarchitecture reveals principles of human neocortex organization. Science 382, eadf6812 (2023).
Article CAS PubMed PubMed Central Google Scholar
DeFelipe, J., Alonso-Nanclares, L. & Arellano, J. I. Microstructure of the neocortex: comparative aspects. J. Neurocytol. 31, 299–316 (2002).
Article PubMed Google Scholar
Macarelli, V., Leventea, E. & Merkle, F. T. Regulation of the length of neuronal primary cilia and its potential effects on signalling. Trends Cell Biol. 33, 979–990 (2023).
Article CAS PubMed Google Scholar
Wang, H. et al. The primary cilia are associated with the axon initial segment in neurons. Adv. Sci. 12, e2407405 (2025).
Article Google Scholar
Yang, D. J., Hong, J. & Kim, K. W. Hypothalamic primary cilium: A hub for metabolic homeostasis. Exp. Mol. Med. 53, 1109–1115 (2021).
Article CAS PubMed PubMed Central Google Scholar
Yang, J. et al. Primary ciliary protein kinase A activity in the prefrontal cortex modulates stress in mice. Neuron 113, 1276–1289 (2025).
Article CAS PubMed Google Scholar
Alvarado, J. A. et al. Developmental distribution of primary cilia in the retinofugal visual pathway. J. Comp. Neurol. 529, 1442–1455 (2021).
Article PubMed Google Scholar
Maor, R., Dayan, T., Ferguson-Gow, H. & Jones, K. E. Temporal niche expansion in mammals from a nocturnal ancestor after dinosaur extinction. Nat. Ecol. Evol. 1, 1889–1895 (2017).
Article PubMed Google Scholar
Gerkema, M. P., Davies, W. I., Foster, R. G., Menaker, M. & Hut, R. A. The nocturnal bottleneck and the evolution of activity patterns in mammals. Proc. Biol. Sci. 280, 20130508 (2013).
PubMed PubMed Central Google Scholar
Hilbig, H., Merbach, M., Krause, J., Gartner, U. & Stubbe, A. Dendritic organization of neurons of the superior colliculus in animals with different visual capability. Brain Res. Bull. 51, 255–265 (2000).
Article CAS PubMed Google Scholar
Graybuck, L. T. et al. Enhancer viruses for combinatorial cell-subclass-specific labeling. Neuron 109, 1449–1464 (2021).
Article CAS PubMed PubMed Central Google Scholar
Mich, J. K. et al. Functional enhancer elements drive subclass-selective expression from mouse to primate neocortex. Cell Rep. 34, 108754 (2021).
Article CAS PubMed PubMed Central Google Scholar
Matson, K. J. E. et al. Isolation of adult spinal cord nuclei for massively parallel single-nucleus RNA sequencing. J. Vis. Exp. 140, https://doi.org/10.3791/58413 (2018).
Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol. 42, 293–304 (2024).
Article CAS PubMed Google Scholar
Browaeys, R., Saelens, W. & Saeys, Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat. Methods 17, 159–162 (2020).
Article CAS PubMed Google Scholar
Liao, K. et al. Single-nucleus profiling decoding the subcortical visual pathway evolution of vertebrates. iScience 28, 112128 (2025).
Article ADS CAS PubMed PubMed Central Google Scholar
Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16, 284–287 (2012).
Article CAS PubMed PubMed Central Google Scholar
Lein, E. S. et al. Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168–176 (2007).
Article ADS CAS PubMed Google Scholar

Download references

Acknowledgements

We thank Maosen Ye and YongGang Yao for assisting with tree shrew genome annotation and for sharing the homologous gene list between mice and tree shrews. We thank Sten Linnarsson’s lab for releasing the human snRNA-seq dataset ahead of its publication. We thank Zhenkun Zhuang’s lab for kindly sharing the zebra finch and pigeon optic tectum snRNA datasets. We thank Rachel Satterfield for sharing the protocol for target retrieval in tree shrew brain slices. Cell sorting and cytometry was performed by the University of Virginia Flow Cytometry Core Facility (RRID:SCR_017829), which is partially supported by a National Cancer Center award (P30 CA044579). Sequencing on the Illumina Next-Seq platform was performed by the Genomics Core of the Biology Department at the University of Virginia and by the Genome Analysis and Technology Core of the University of Virginia’s School of Medicine (RRID:SCR_018883). This work was supported by the following funds: US NIH grant EY031783 to E.L.S.; US NIH grant NS122040 to A.E.; US NIH grants (EY026286, EY020950, NS122040) and Jefferson Scholars Foundation funding to J.C.; and the American Diabetes Association Pathway to Stop Diabetes Initiator Award 1-18-INI-14 and US NIH grants HL153916 and EY033528 to J.N.C.

Author information

These authors jointly supervised this work: Jianhua Cang, John N. Campbell.

Authors and Affiliations

Department of Biology, University of Virginia, Charlottesville, VA, USA
Yuanming Liu, Lu Yang, Elise L. Savier, Jianhua Cang & John N. Campbell
Department of Psychology, University of Virginia, Charlottesville, VA, USA
John A. McDaniel, Chen Chen, Arda Kipcak, Alev Erisir & Jianhua Cang

Authors

Yuanming Liu
View author publications
Search author on:PubMed Google Scholar
John A. McDaniel
View author publications
Search author on:PubMed Google Scholar
Chen Chen
View author publications
Search author on:PubMed Google Scholar
Lu Yang
View author publications
Search author on:PubMed Google Scholar
Arda Kipcak
View author publications
Search author on:PubMed Google Scholar
Elise L. Savier
View author publications
Search author on:PubMed Google Scholar
Alev Erisir
View author publications
Search author on:PubMed Google Scholar
Jianhua Cang
View author publications
Search author on:PubMed Google Scholar
John N. Campbell
View author publications
Search author on:PubMed Google Scholar

Contributions

Y.L., J.N.C., and J.C. designed the experiments; Y.L. performed snRNA-seq experiments and data analysis with assistance from J.N.C.; Y.L., J.A.M., and C.C. performed RNAscope experiments and data analysis; Y.L., L.Y., A.K., and A.E. performed retrograde tracer injection into the mouse and tree shrew pulvinar; E.L.S. dissected the tree shrew sSC for the sequencing experiments; Y.L., J.N.C., and J.C. wrote the manuscript with input and/or approval from all authors.

Corresponding authors

Correspondence to Jianhua Cang or John N. Campbell.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Xin Duan, Jason Triplett and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Reporting Summary

Transparent Peer Review file

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Liu, Y., McDaniel, J.A., Chen, C. et al. Co-Conservation of synaptic gene expression and circuitry in collicular neurons. Nat Commun 16, 9146 (2025). https://doi.org/10.1038/s41467-025-64204-5

Download citation

Received: 13 January 2025
Accepted: 08 September 2025
Published: 15 October 2025
DOI: https://doi.org/10.1038/s41467-025-64204-5