Single-cell RNA sequencing of adult primate neocortex reveals the regulatory dynamics of neural plasticity

Abstract

Objectives: To investigate gene expression dynamics in adult primate neocortex and elucidate the molecular basis of these capabilities. Methods: Whole-cell transcriptome data from primary visual cortex (V1) of cynomolgus monkeys were analyzed, focusing on inhibitory neurons derived from the medial ganglionic eminence. Using RNA velocity and scVelo algorithms, we investigated subtle gene expression shifts, identified key driver genes, and conducted functional enrichment analysis. Results: Our analysis revealed that somatostatin (SST)+ neurons are generated prior to parvalbumin (PVALB)+ neurons, with gene expression trajectories resembling those observed in developing animals. However, certain well-established SST+ and PVALB+ neurons, such as long-projecting SST+ chondrolectin (CHODL)+ neurons and PVALB+ unc-5 netrin receptor B (UNC5B)+ neurons, do not share the same lineage, suggesting distinct regulatory programs. Furthermore, we identified a set of genes strongly correlated with these trajectories, including dihydropyrimidinase like 3 (DPYSL3), osteocrin (OSTN), smoothelin (SMTN), meis homeobox 2 (MEIS2), chromodomain helicase DNA binding protein 3 (CHD3), and chromodomain helicase DNA binding protein 5 (CHD5). Functional enrichment analysis suggested that these genes are associated with neural plasticity and axon growth. Conclusions: This study provides novel insights into the gene expression dynamics of the primate neocortex. Genes involved in neuronal development and plasticity may underlie the advanced information-processing capabilities in primates.

Introduction

Mammals develop complex neural networks with high plasticity to adapt to their ever-changing environments, which are closely related to both the basic and higher brain functions. However, neurogenesis is limited to a few specific regions in adult mammals, and most existing neurons undergo a prolonged maturation and refinement process shaped by local signaling [1]. The extent to which adult neural networks can be modified, as well as the core regulatory programs that enable mature neuron subtypes to maintain both their identity and plasticity, remain unclear.

The primary visual cortex (V1) of adult macaques offers an ideal model to explore these questions [2]. As diurnal animals, primates depend heavily on visual information and have evolved trichromatic and stereoscopic vision with high precision. Anatomical and electrophysiological studies have identified primate-specific structures, such as ocular dominance columns and an expanded layer (L4) [3-5]. Visual experience drives network plasticity during the critical period of postnatal visual development, primarily through interaction between local signaling and intrinsic gene-expression regulation [6]. Even in adulthood, the visual system maintains high levels of plasticity in primates [2,7]. Investigating the subtle divergencies and similarities among neurons of the same cell type can provide insights into the core transcriptomic programs that govern plasticity and maintain cellular identity and functions.

The rise of digital technologies has created unparalleled opportunities in the global healthcare sector [8]. Recent advancements in high-throughput sequencing enable high-resolution classification of neuronal cell types [2,9-16]. Transcriptomic signatures have been established for various neuronal populations, including glutamatergic and GABAergic neurons. Glutamatergic neurons are excitatory and classified based on their laminar distribution and predicted projection targets. In contrast, GABAergic neurons are inhibitory and can be further subdivided by their origin into two branches: the lim homeobox 6 (LHX6)+ branch, derived from the medial ganglionic eminence (MGE), and the nuclear receptor subfamily 2 group f member 2 (NR2F2)+ branch, originating from the caudal ganglionic eminence (CGE). The MGE-derived branch includes the parvalbumin (PVALB)+ and somatostatin (SST)+ subclasses, while the CGE-derived branch consists of the lysosomal-associated membrane protein family member 5 (LAMP5)+, vasoactive intestinal peptide (VIP)+, and adenosine deaminase RNA-specific b2 (ADARB2)+/paired box 6 (PAX6)+ (synuclein gamma (SNCG)+) subclasses [2]. GABAergic neurons synapse on local neurons and fine-tune the network in an experience-dependent manner. However, detailed analyses of GABAergic neurons, especially those within the same lineage (such as the MGE-derived PVALB+ and SST+ neurons), and the genes that drive and maintain transcriptomic diversity, have not been fully explored.

In this study, we utilized a whole-cell single-cell RNA sequencing dataset from the primary visual cortex of macaque monkeys to analyze transcriptomic shifts within one of the most well-characterized inhibitory lineages, the MGE lineage. Using the RNA velocity algorithm, which captures transcriptomic shifts at single-cell resolution by assessing both spliced and unspliced RNA abundance, we identified a conserved group of mature adult neurons [17,18]. Enrichment analysis revealed that genes associated with these transcriptomic shifts are linked to neuroplasticity, providing molecular insights into how mature neurons maintain core functions while preserving appropriate levels of plasticity.

Materials and methods

Data acquisition

The Macaque V1 10× Genomics dataset was downloaded from the European Molecular Biology Laboratory’s (EMBL) European Bioinformatics Institute (EBI) database (https://www.ebi.ac.uk/) under the accession code E-MTAB-10459 [2].

Data preprocessing

Cell Ranger (version 3.0.2, 10× Genomics) was used to process the raw output from Chromium Single Cell 3’ RNA-seq. Cells with fewer than 200 features or those exhibiting more than 30% mitochondrial content were removed for quality control. Additionally, features expressed in fewer than 10 cells were also excluded. Data processing, analysis, and visualization were performed using SCANPY (version 1.4.6) [19]. Scrublet (version 0.2.1) was used to identify and exclude potential cell doublets [20], which were detected based on cluster analysis, where a Scrublet score was computed for each cluster, indicating the likelihood of doublet presence. Clusters with high Scrublet scores in t-SNE or UMAP (detailed below) and with marker gene expression patterns overlapping multiple clusters were classified as cell doublets and excluded from further analysis.

Data processing, analysis, and visualization

Data processing, analysis, and visualization were performed sequentially using SCANPY. Data normalization: Gene expression counts were normalized to counts per 10,000 (NC) by dividing the total mapped exonic reads for each gene by the total number of mapped reads in the cell, then multiplying by 10,000. This method generates a value equivalent to one-hundredth of counts per million (CPM). Log transformation: A log2 (NC + 1) transformation was applied to stabilize variance and approximate a normal distribution. Highly Variable Genes (HVG) Selection: HVGs were selected from the log-transformed data for further analysis. Principal Component Analysis (PCA): PCA was conducted on the selected HVGs to reduce dimensionality of the dataset while preserving the majority of variance. If necessary, HVGs and PCA were recalculated for specific cell subsets.

Data integration and correction were conducted using Harmony (version 0.0.5) based on PCA embeddings [21]. The corrected data minimized both technical factors and biological variations.

Dimension reduction and clustering were performed using SCANPY. Neighborhood calculation: The local neighborhoods were calculated using the scanpy.pp.neighbors function, constructing a k-nearest neighbors (knn) graph based on Harmony-corrected PCA embeddings. The number of local neighborhoods was set to 15, using all 50 Harmony-corrected PCs. Uniform Manifold Approximation and Projection (UMAP)/T-distributed stochastic neighbor embedding (t-SNE) [22-25]: Both UMAP and t-SNE were performed using Euclidean distance. UMAP was computed using the scanpy.tl.umap function with the following parameters: minimal effective distance = 0.5, spread = 1.0, initial learning rate = 1.0, negative sample weighting = 1.0, and negative edge sample rate = 5. t-SNE was computed using scanpy.tl.tsne. Louvain clustering [26]: Louvain clustering was applied using the sc.tl.louvain function on the neighborhood graph generated in the previous steps. The analysis was conducted with the resolution parameter set to 2, using the vtraag package, without applying weights from knn graph. The resolution parameter was adjusted between 1 and 5 to achieve finer cluster divisions. Single-Cell Clustering Assessment Framework (SCCAF) [27]: SCCAF (version 0.0.10) was used with the SCCAF_assessment function to evaluate the quality and robustness of the clusters. For five-fold cross-validation training, 200 cells from each cluster were selected. The remaining cells were utilized to evaluate segregation accuracy. Clusters with a segregation accuracy greater than 84% on the hold-out set were considered acceptable.

Differential expression analysis

Differentially expressed genes (DEGs) between clusters were identified using the Model-based Analysis of Single-cell Transcriptomics (MAST), which applies a hurdle model to account for both technical effects and biological variabilities [28]. The top 200 DEGs from each cell cluster were selected for Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses [29-33]. Gene Set Enrichment Analysis (GSEA) and gProfiler were used to perform GO and KEGG enrichment analyses [34-36].

RNA velocity analysis

RNA velocity analysis was conducted using RNA velocyto (version 0.17.15) and scVelo (version 0.2.2) [17,18]. RNA velocyto was run on the 10× Cell Ranger using the run 10× option. The UCSC genome GTF file provided organism-specific annotations, while the macaque genome GTF file served as the reference for genome annotation [37]. After generating loom files, scVelo was applied by merging these files with the AnnData object in SCANPY.

The standard scVelo pipeline was used to calculate RNA velocity. Data preprocessing involved filtering for genes with a minimum of 20 counts (both unspliced and spliced) to retain those with sufficient expression for further analysis. Moments were computed for each cell based on its nearest neighbors. The neighbor graph was constructed using Euclidean distances in the PCA embeddings, utilizing 30 principal components and connecting 30 neighbors in connectivity mode. The UMAP method was then applied using the previously selected highly variable genes. Full splicing kinetics for genes were recovered using the scvelo.tl.recover_dynamics function with default settings. The velocities were estimated using the scvelo.tl.velocity function in dynamical mode. The velocity graph, embedded on the UMAP for visualization, was generated using the scvelo.tl.velocity_graph function with default settings.

Tissue procurement

Macaque V1 tissue samples were generously provided by Professor Hao from the Zhongshan Ophthalmic Center. The V1 regions were fixed in 4% paraformaldehyde (PFA) for subsequent Nissl staining and immunofluorescence analysis.

Nissl staining

Nissl staining was performed using the NeuroTrace™ 530/615 red fluorescent Nissl stain kit (Injeka, N21482). Tissue slices were fixed in 4% PFA at 4°C for 1 h, followed by dehydration through increasing sucrose concentrations (10%, 15%, and 30%) and three washes with phosphate-buffered saline (PBS). Permeabilization was achieved using 0.1% Triton X-100 for 10 min, followed by a 5-minute PBS wash. The slices were then incubated with the NeuroTrace stain for 20 min, washed in PBS for 10 min, and mounted with DAPI (Abcam, AB 104139). Images were captured using a laser scanning confocal fluorescence microscope (Zeiss LSM 880).

Immunofluorescence

Tissues were fixed in 4% PFA at 4°C overnight, and then dehydrated in a 30% sucrose solution at 4°C. They were then embedded in an optimal cutting temperature (OCT; Tissue-Tek) compound and frozen at -80°C. The tissues were sectioned into 30 μm slices using a cryostat microtome (Leica CM 1950). The slices were washed in PBS and blocked for 1 h at room temperature with a solution containing 5% BSA, 1% normal goat serum (Jackson ImmunoResearch, 005000121), and 0.3% Triton X-100 in PBS. Primary antibody incubation was performed overnight at 4°C. After three PBS washes, the slices were incubated with secondary antibodies for 1 h at room temperature. The primary antibodies used included anti-Parvalbumin (1:500, P3088, Sigma, USA), anti-Somatostatin (1:100, ab37008, Abcam, USA), anti-Dihydropyrimidinase Like 3 (1:1000, HPA010948, Sigma, USA), anti-Meis Homeobox 2 (1:100, H00004212-M01, Abnova, USA), and anti-Calcyon (1:500, AB15040, Millipore, USA). The secondary antibodies included Alexa Fluor 647 anti-rat (Invitrogen, A78947, 1:1000), Alexa Fluor 488 anti-rabbit (Invitrogen, A-21206, 1:1000), and Alexa Fluor 555 anti-mouse (Invitrogen, A-31570, 1:1000). After three PBS washes, the slices were mounted with medium containing DAPI (Abcam, ab104139). Images were captured using a laser scanning confocal fluorescence microscope (Zeiss LSM 880).

Results

Molecular taxonomy of MGE-derived inhibitory neurons

To elucidate the transcriptomic dynamics of mature MGE-derived GABAergic neurons, we reanalyzed a well-established dataset of whole-cell transcriptomes from the V1 of cynomolgus monkeys (Macaca fascicularis). The dataset included both spliced RNAs in the cytosol and unspliced RNAs in the nucleus, preserving comprehensive transcriptional information (Figure 1A). The dataset contained 133,454 cells, which were classified into 26 clusters (Figure 1B-D). These clusters were subsequently grouped into the major cell types present in the neocortex: astrocytes, excitatory neurons (Exc), inhibitory neurons (Inh), microglia, oligodendrocyte progenitor cells (OPC), oligodendrocytes (Oligo), and vascular cells (Figure 1C). Further subdivision of the excitatory and inhibitory neurons revealed four and five subclasses, respectively, based on the expression of specific cell markers (Figure 1E, 1F).

Figure 1.

Flow diagram and cell type classification in macaque primary visual cortex (V1). A. Illustration diagram of single-cell transcriptome analysis in macaque V1. The V1 region selected for single-cell isolation is highlighted in green. Illustrations of the macaque and individual cells were created using the BioRender platform (www. BioRender.com). B. tSNE visualization of the 26 original clusters of total cells from the cynomolgus monkey V1. C. tSNE visualization of the total cells, colored by cell type. D. Nissl staining of the macaque V1, showing cortical layers 1-6 (L1-6) of the neocortex. WM, white matter. E. tSNE visualization of the total cells, colored by marker gene expression. Excitatory neuron markers: Hippocalcin-like 1 (HPCAL1), rar-related orphan receptor b (RORB), FEZ family zinc finger 2 (FEZF2), thymocyte selection-associated (THEMIS); Inhibitory neuron markers: parvalbumin (PVALB), somatostatin (SST), vasoactive intestinal peptide (VIP), lysosomal-associated membrane protein family member 5 (LAMP5), adenosine deaminase RNA-specific B2 (ADARB2)_paired box 6 (PAX6); Astrocyte marker: aquaporin 4 (AQP4); Oligodendrocyte marker: myelin oligodendrocyte glycoprotein (MOG); Microglia markers: platelet-derived growth factor receptor alpha (PDGFRA), oligodendrocyte progenitor cell marker; complement c1q B chain (C1QB); Vascular cell marker: FMS-related receptor tyrosine kinase 1 (FLT1). Red circles indicate medial ganglionic eminence (MGE) lineages of PVALB+ and SST+ neurons. F. Dotplot shows marker gene expression in the total cells.

Among the total 133,454 cells, 12,097 inhibitory neurons (9%) were identified based on canonical marker expression. The MGE-derived cells were selected based on the expression of PVALB, SST, and corresponding cell labels, yielding 1,572 PVALB+ and SST+ neurons (Figure 2A). The expression of well-established MGE lineage marker genes, including synaptosome-associated protein 25 (SNAP25), LHX6, and glutamate decarboxylase 1 (GAD1), as well as the layer distribution of the dataset, is illustrated in Figure 2B. Based on these canonical markers, the PVALB+ and SST+ neurons were further subdivided into 6 and 8 transcriptomic cell types, respectively (Figure 2C). The proportions of each cell type were calculated and illustrated in Figure 2D.

Figure 2.

MGE-derived cell taxonomy in the macaque V1. A. tSNE visualization of MGE-derived cells (indicated by circles) using the same color scheme shown in Figure 1C, with all other cells shown in gray. B. tSNE visualization of the expression of the canonical marker genes for MGE-derived cells and the layer distribution within the datasets. C. Dotplot showing the marker gene expression in MGE-derived cells. D. Proportional representations of different MGE-derived cell types.

RNA velocity reveals transcriptomic shifts in MGE-derived inhibitory neurons

To determine where cellular dynamics regulated by driver genes during neuronal development persist in mature primate neurons, we conducted RNA velocity analysis using scVelo. This allowed us to systematically identify the key factors and pathways involved in the development of MGE-derived PVALB+ and SST+ neurons [17,18]. All transcriptomic cell types of the PVALB+ and SST+ neurons were included in future analyses. ScVelo predicted that SST+ neurons give rise to the PVALB+ neurons (Figure 3A). Here, we found that SST+ complexin 2 (CPLX2)+ neurons, originating from SST+ GPM6A+ neurons, differentiate into PVALB+ immunoglobulin superfamily containing leucine-rich repeat 2 (ISLR2)+ neurons, followed by PVALB+ regulator of G protein signaling 5 (RGS5)+ neurons, and ultimately PVALB+ osteocrin (OSTN)+ neurons. In addition, SST+ triadin (TRDN)+ neurons branch from SST+GPM6A+ neurons. Notably, SST+ fucose mutarotase (FUOM)+ neurons differentiate into SST+ CAMP-dependent protein kinase inhibitor alpha (PKIA)+ neurons, then SST+ mitogen-activated protein kinase 6 (MAPK6)+ neurons, and eventually give rise to PVALB+ gamma-aminobutyric acid type A receptor subunit alpha 5 (GABRA5)+ neurons. ScVelo also predicted that SST+ proenkephalin (PENK)+ neurons and SST+ chondrolectin (CHODL)+ neurons are independent of other neuron populations, with no predicted trajectories leading to other neuron types. Similarly, PVALB+ unc-5 netrin receptor B (UNC5B)+ neurons and PVALB+ filamin A interacting protein 1 (FILIP1)+ neurons exhibited distinct origins, suggesting separate developmental pathways.

Figure 3.

Dynamics of the SST to PVALB transition recapitulated by scVelo. (A) scVelo analysis of developmental trajectories for PVALB+ and SST+ subclasses with regulatory dynamics indicated by black arrows. (B) Example of tentative PV-related driver genes identified with high likelihoods by scVelo. The cells follow the same color scheme as shown in (A). (C) Immunostaining of PVALB, dihydropyrimidinase-like 3 (DPYSL3) and SST. DAPI, nuclei. Scale bar, 40 µm (low magnification) and 20 µm (high magnification). (D) Example of tentative PVALB-related driver genes identified with high likelihoods by scVelo. The cells follow the same color scheme shown in (A). (E) Immunofluorescent staining of meis homeobox 2 (MEIS2) and SST. DAPI, nuclei. Scale bar, 40 µm (low magnification) and 20 µm (high magnification).

The selective driver genes correlated with these developmental trajectories are illustrated in Figure 3B, 3D, with representative gene expressions validated in Figure 3C, 3E. Dihydropyrimidinase-like 3 (DPYSL3), OSTN, and smoothlin (SMTN) appear to drive neurons towards the PVALB+ cell fate, while meis homeobox 2 (MEIS2), chromodomain helicase DNA binding protein 3 (CHD3), and chromodomain helicase DNA binding protein 3 (CHD5) seem to drive neurons towards the SST+ cell fate.

Next, we calculated the latent time of the cells in the MGE lineages (Figure 4A). The latent time for SST+ neurons was lower compared to the PVALB+ neurons, which exhibited higher latent times. Putative driver genes correlated with these latent time differences were identified and visualized in a heatmap (Figure 4B). A shared driver gene, PVALB+ SST+ calcyon neuron specific vesicular protein (CALY), was validated in Figure 4C, and gene expression patterns across each cell type are illustrated in Figure 4D.

Figure 4.

Gene expression dynamics revealed by latent time. A. UMAPs colored by the corresponding scVelo-inferred latent time, ranging from early (deep purple) to late (yellow). B. Heatmap of tentative driver genes involved in the maturation process of PVALB+ and SST+ cells, generated by scVelo. The color bar above the heatmap denotes cell types and follows the same color scheme shown in Figure 3A. C. Immunostaining validation of the PVALB SST shared driver gene, calcyon neuron-specific vesicular protein (CALY). DAPI, nuclei. Scale bar, 40 µm (left panel) and 20 µm (right four panels). D. Dotplot showing the putative driver gene expression in MGE-derived cells.

GO biological process enrichment analysis revealed that, in addition to the nervous system development process, the identified genes are involved in cell morphology and projection, cell adhesion, cell-cell communication, and most interestingly, axon genesis [34,35]. These processes are predominantly focused on anatomical structure morphogenesis, cell communication, and signaling. Furthermore, KEGG pathway enrichment analysis revealed that these genes are mainly associated with neurodegeneration pathways, calcium signaling, and axon guidance. These results suggest that major modifications in mature neurons are likely related to axon growth, particularly involving projections to new areas through the formation of new synapses (Figure 5A, 5B).

Figure 5.

Functional enrichment analysis of latent time-related genes reveals their major roles in nervous system development and axon genesis. A. Dotplot showing the Gene Ontology (GO) biological process enrichment analysis. B. Dotplot showing the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis.

Discussion

In this study, we analyzed the transcriptomic shifts within the MGE-derived inhibitory neurons in the macaque primary visual cortex. Our findings revealed that SST+ neurons are generated prior to PVALB+ neurons, mirroring developmental patterns observed in earlier stages of tissue development. However, long-projecting SST+CHODL+ neurons and PVALB+UNC5B+ neurons did not share the same lineage, suggesting distinct regulatory mechanisms. We identified a set of genes correlated with the neural development trajectories, which are linked to neural plasticity and axon growth. These genes offer insights into the enhanced information-processing capabilities of the primate neocortex.

Previous studies have established a detailed genetic cascade within the mouse MGE lineage [17]. Consistent with earlier research on mice, our study found that SST neurons give rise to PVALB neurons [38-40]. Although scVelo analysis is traditionally used in developing tissues, accumulating evidence suggests that gene expression in mature neurons remains under the regulation of transcription factors [41]. Our findings align with these developmental patterns, even in mature neuronal populations.

We identified selective driver genes that correlate with the developmental trajectories of MGE-derived neurons, including DPYSL3, OSTN, SMTN, MEIS2, CHD3, and CHD5. These genes are associated with neuronal development, plasticity, and connectivity. For instance, DPYSL3 plays a critical role in neuronal development and function, including neurite and axonal outgrowth, neuronal differentiation, axonal guidance and regeneration, and proapoptotic signal transduction [42]. OSTN, which is induced by sensory stimuli, regulates activity-dependent dendritic growth in neurons [43]. SMTN is a cytoskeletal protein contributing to cellular structural stability. MEIS2 is essential for regulating neural tube patterning, neural progenitor cell proliferation, cell-fate acquisition, neuronal maturation, neurite outgrowth, and synaptogenesis [44]. CHD5 is involved in early radial migration, while CHD3 is crucial for late migration and laminar specification of neurons [45]. These driver genes likely contribute to neuronal morphological plasticity, functional differentiation, and network integration, thereby enhancing the information-processing capacity of the primate neocortex.

Consistent with previously established maturation processes, our study found that the latent time for SST+ neurons is shorter than that for PVALB+ neurons. The putative driver genes correlated with latent time include CALY, a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTS), protocadherin-19 (PCDH19), and talin 2 (TLN2). CALY is involved in endocytic and recycling processes, playing roles in neuronal development, synaptic function, and neurodegeneration [46]. ADAMTS encodes a secreted enzyme that regulates synaptic and neurite plasticity by influencing the extracellular matrix of the central nervous system [47]. PCDH19 encodes a calcium-dependent adhesion molecule that plays a role in signal transduction at synapses, and is associated with cell adhesion, neuronal migration, and synaptic plasticity [48,49]. TLN2 is crucial for the formation of actin filaments and the spreading and migration of various cell types [50]. Together, these genes likely form a network that supports the development, connectivity, and plasticity of neural circuits, shedding light on the complex mechanisms behind synaptic plasticity and neuronal function in the primate brain.

Evolution has driven significant modifications in both the structure and function of the brain. In primates, complex sensory processing demands have led to evolutionary enhancements across various cortical regions. In macaque V1, specialized structures such as ocular dominance columns, an expanded L4, and enlarged macular receptive fields have been identified [51-53]. These adaptations suggest that specific cell types and intricate neural circuits play a critical role in supporting sophisticated visual functions. The laminar organization of macaque V1, as demonstrated by Nissl staining and molecular patterns at the cellular level, closely resembles that of human V1 [5]. This similarity makes macaque V1 an ideal model for exploring the cellular and molecular mechanisms underlying brain evolution across species, particularly in the context of higher-order primates like humans. These cell types and circuits are shaped by experience, allowing individuals to adapt their behavior to the surrounding environment in unique ways. Unlike rodents, primates are diurnal animals with flat eyes that receive binocular light stimuli. The differences in visual experience between primates and rodents may be attributed to the emergence of novel cell types or the modification of existing networks, which can be elucidated through large-scale, high-throughput analysis of cortical neurons.

From mice to primates, mammals share a basic neural framework for processing sensory information [54]. However, neonatal primates demonstrate very limited, if any, advantages in visual perception over mice, suggesting that innate circuits alone do not account for the advanced sensory processing seen in adult primates. Instead, these innate circuits are extensively shaped by experience [55]. For example, deprivation amblyopia caused by congenital cataracts can disrupt the development of retinal neural circuits [56,57]. Similarly, elevated anxiety symptoms are associated with delays in the development of neural circuits, while music therapy has been shown to mitigate anxiety and potentially promote healthy neural development [58,59]. Once the nervous system reaches full maturity, sensory processing capacity varies significantly among species. Primates, leveraging their prior experience and current state, exhibit greater flexibility in handling complex tasks [60]. Such differences in maturation trajectories indicate that primates have evolved enhanced mechanisms to modify intrinsic neural circuits, necessitating both greater plasticity and robust experience-dependent modulation [60,61].

Interestingly, our results suggest that, despite sharing similar baseline survival mechanisms, primates and rodents develop dramatically different skill sets for handling everyday challenges. This divergence likely stems from differential gene activity related to neuronal development and plasticity. These genes may be crucial to understanding the molecular and cellular mechanisms behind the advanced information-processing capabilities of primates.

RNA velocity and scVelo analyze RNA lineages by comparing the relative abundances of spliced and unspliced RNAs. These methods were developed to study tissue transcriptomic dynamics at the single-cell level and were originally designed for developing tissues. However, their applicability to mature tissue remains uncertain. Despite this, the highly consistent results observed in our study support the idea that RNA velocity and scVelo are reliable tools for inferring transcriptomic shifts in individual neurons. Further validation is required to determine the extent to which these findings can be generalized across different neuronal populations and conditions.

Disclosure of conflict of interest

None.