Molecular architecture of primate specific neural circuit formation

The mammalian cortex is a highly evolved brain region, but we still lack a comprehensive understanding of the molecular mechanisms underlying primate-specific neural circuits formation. In this study, we employed spatial transcriptomics to assess gene expression dynamics in the marmoset cortex during development, focusing on key regions and time points. Spatial transcriptomics identified genes that are sexually, spatially, and temporally differentially expressed in the developing marmoset cortex. Our detailed analysis of the visual cortex unveiled dynamic changes in gene expression across layers with distinct projections and functions. Notably, we discovered numerous axon guidance molecules with spatiotemporal expression patterns unique to the developing marmoset prefrontal cortex (PFC), which control PFC neuronal circuits. Among these molecules, PRSS12 (Protease, Serine, 12 (neurotrypsin, motopsin), when ectopically expressed in the mouse prelimbic cortex, caused similar changes in connectivity as observed in the marmoset A32 area. Furthermore, PRSS12 showed similar expression patterns in both marmoset and human PFC during development, suggesting parallels between marmoset and human brain development. The differential expression of axon guidance molecules in the developing PFC, varying by region, likely contributes to the formation of unique circuits observed in primates.


Introduction
The cerebral cortex, responsible for higher cognitive functions, including the primate-speci c prefrontal cortex, is crucial in evolution 1 .It is believed that the evolution of the cerebral cortex is driven by multiple mechanisms, including gene mutations, selection pressures, and changes in developmental processes 2 .
However, the precise mechanisms involved are still not fully understood.Understanding its development requires analyzing gene expression patterns across mammalian species.While rodent studies offer insight, studying primate cortical development, especially in species like the common marmoset, provides valuable clues due to their brain's similarity to humans [3][4][5][6][7][8][9][10] .Our analysis of gene expression by spatial transcriptomics, in various neocortical regions during marmoset development revealed time-speci c, region-speci c cortical layer-speci c gene expression.We hypothesized that this differential expression, especially of axon guidance factors in the PFC, could reveal molecular processes governing evolutionary diversi cation.Our functional gene expression analysis provide evidence that these species-speci c spatiotemporally expressed genes play important role to generate species-speci c neuronal circuit.

Result
Sample preparation and analysis for 10x spatial transcriptomics Fresh frozen brain slices (coronal plane) were sectioned (10 μm thickness) and placed on 6.5 mm x 6.5 mm, 10x Genomics Visium slides (Fig. 1a).We pro led spatial gene expression patterns of common marmoset cortices encompassing the primary and secondary visual cortex (V1 and V2, respectively), primary auditory cortex (AuA1), primary somatosensory cortex (A3), and prefrontal cortex (PFC) at different postnatal developmental ages: P16, male, 1-month (M), 3-months (M) and 6-months (M), for both sexes 11 (Fig. 1a and Supplementary Table 1).We detected 11,368 ± 5,176 unique molecular identi ers (UMIs) per spot and 4,289 ± 1,273 unique genes per spot from all 47 samples (Supplementary Table 1).High mean rates to exonic regions of mRNAs (mean: 74.4 %) was detected compared with intronic regions (mean: 2.4 %) (Supplementary Table 1).The spots classi ed as white matter, and unreliable/mispositioned spots, and noise spots were excluded from further analysis because of tissue loss, damage, ambiguous clustering.The analysis revealed the molecularly distinct clusters for cortical layers (layer 1 (L1) to L6), as shown in the uniform manifold approximation and projection (UMAP) (Extended Data Fig. 1).Several resources have compiled genes that exhibit laminar-speci c expression across both rodent and human cortices, which were used for analysis 12,13 .Although both overlapping and unique marker genes have been identi ed, these studies used different technologies, examined different developmental periods, and queried different regions of the cortex [13][14][15] .We thus assessed the robustness of these previously identi ed marker genes in our marmoset cortex layer enriched gene expression dataset.We generated aggregated layer-enriched expression pro les for each Visium data using a 'supervised' approach to assign individual spots to each of the six neocortical layers.The comparison between clusters of cortical layers identi ed genes enriched in each layer, which is displayed in the heatmap (Fig. 1a).
We next analyzed 10x Visium samples collected from each developmental period (P16, 1M, 3M, and 6M) for each brain regions.The data from each developmental periods was merged after clustering into each cortical layers at each developmental periods using the Seurat R package 16,17 .The batch correction was conducted using the Harmony R package 18 , followed by UMAP visualization for each developmental period (Extended Data Fig. 1).
First, we performed gene expression analysis to identify genes that exhibit sexually speci c expression in various brain regions.Moreover, we extracted genes which are derived from sex chromosomes, and identi ed genes that exhibit temporal and brain region-speci c expression changes, as indicated in Supplementary Table 2.
Among the analyzed brain regions, we observed strong expression of the gene LOC118150934 in males but minimal expression in females (Fig. 1b).The gene EOLA1 displayed a trend of opposite expression changes in male (decrease) and females (increase) (Fig. 1b).EOLA2 exhibited different temporal expression changes between males and females in each brain regions (Fig. 1b,c).Additionally, we revealed genes such as PCSK1N and TMSB4X that exhibited speci c expression changes in speci c brain region (Fig. 1b).The spatial organization of sexually differentially expressed genes provides valuable insights into their potential interactions within neural circuits and their roles in brain function.By analyzing spatial transcriptomics data, we can investigate co-expression networks and identify gene regulatory networks that are speci c to each sex.The observed differences in gene expression patterns between males and females may contribute to the developmental and functional disparities observed in their respective brains.For example, a recent paper showed female-speci c risk genes involved in depressive-like behaviors by conducting single-cell RNA sequencing (scRNA-seq) and spatial transcriptomic data in non-human primate 19 .
Together, these analyses demonstrate the power of concurrently acquiring histology and gene expression data and highlight the ability of the 10x Visium platform to achieve high-resolution spatial expression pro ling within the developing marmoset neocortex.

Identifying time dependent layer-enriched genes in marmoset visual cortex
The critical period in the primate visual cortex is a crucial period characterized by signi cant changes in the exibility of neurons 21 .This period is marked by speci c genes undergoing functional and expression changes in response to appropriate visual stimuli, especially for the development of connections between neurons and the formation of neural circuits 22,23 .For example, V1 and V2 are distinct regions of the brain that participate in visual processing 24 .Therefore, V1 and V2 differ in terms of developmental timing, cellular specialization, connectivity, and functional integration.To comprehend these differences during development, we compared the expression pro le between V1 and V2 cortices in developing marmoset (Fig. 3a).
As Fig. 2 analysis identi ed clusters for six cortical layers of V1 and V2 during development (Extended Data Fig. 1), we directly compared the gene expression between V1 and V2 cortical layers.Direct comparison for V1 layer 4 and V2 layer 4 clusters indicates signi cant gene differences during development, however, such signi cant gene differences were not detected when other layer clusters of V1 and V2 were compared during development (Fig. 3b).This suggests that the layer 4 is a characteristic cortical layer in visual cortex, therefore, we focused on the gene expression pro les of layer 4 in developing marmoset visual cortex.The gene ontology (GO) analysis of genes obtained from the comparison between V1 layer 4 and V2 layer 4 showed various categories such as "nervous development", "neuron projection", and "ion channels" (Fig. 3d,e).In general, V1 matures earlier, specializes in fundamental visual processing, and establishes speci c connections with both visual and non-visual areas 25 .On the other hand, V2 develops later, demonstrates higher cellular specialization, forms more extensive connections with other visual regions, and engages in more intricate visual processing tasks 26 .Consistent with this, our comparison between V1 layer 4 and V2 layer 4 showed distinct gene expression dynamics during development.(Fig. 3e).The number of differentially expressed genes (DEGs) in V1 at different periods reached its peak after 3M and then decreased at 6M, showing the dynamic gene expression from P16 to 6M (Fig. 3e).On the other hand, V2 also exhibited dynamic changes in gene expression during the P16-3M time period, indicating that V2 undergoes dynamic plasticity during this period (Fig. 3e).
To enhance understanding of our GO analysis, we focused on the comparison between V1 layer 4 and V2 layer 4 at P16, given that gene expression, including axon guidance molecules, appears to be dynamic during early development (Supplementary Table 4).GO analysis at P16 revealed the categories for "neuron projection development" and "ion channels" in V1 and V2 layer 4, which often characterize area-, layer-, or cell-type-speci cities.For example, Semaphorins which are involved in axon guidance exhibited area-, and layer-speci c expression in the visual cortex (SEMA3F, SEMA5A, SEMA5B, and SEMA7A in V1, and SEMA3A in V2) (Supplementary Table 4).Furthermore, genes encoding ion channels exhibited unique expression patterns in the visual cortex (KCNA2, KCNC1, KCNC2, KCNJ3, KCNQ5, KCNS1, KCNS2, SCN1B, and SCN2B in V1, and CACNG3, CACNA1I, GABRA5, GABRG1, KCNA4, KCNAB1, KCNF1, and KCNG1 in V2) (Supplementary Table 4).Primate V1 layer 4 is divided into two sublayers: 4A and 4B.These sublayers process visual information differently, with 4A receiving direct input from the thalamus and relaying information related to visual perception.Neurons in 4A receive inputs from pathways processing motion and color.In contrast, 4B processes orientation and spatial frequency information.Layer 4 in V2 continues processing visual stimuli with more complexity.Understanding these layers' functional properties and connectivity patterns is crucial for understanding visual perception and higher-order visual processing.We investigated genes with spatiotemporal expression patterns at sublayer resolution, focusing on layer 4A and 4B in V1 (Fig. 3f).A direct comparison among V1 layer 4A, V1 layer 4B, and V2 layer 4 identi ed genes with spatiotemporal expression patterns in these layers, along with the area-and layer-speci c genes that exhibit stable expression during development (Fig. 3g).ISH con rmed their stable-and area-speci c expression at sublayer level in visual cortex (V1 layer 4A: MATN4; V1 layer 4B: NTNG1, HTR2A; V2 layer 4: IL1RAP in Fig. 3g,h).For spatiotemporal genes, CDH6 in V1 layer 4A, GPR6 in V1 layer 4B, and GPR88 in V2 layer 4 exhibited transient expression at early periods (Fig. 3g,h).Finally, through these analyses, we identi ed certain genes, like CYP26B1 and WHRN, with expression patterns that swap between V1 and V2 at speci c developmental time points (Fig. 3i).This gradual specialization and maturation of visual processing in these areas, with V1 focusing on initial visual processing and V2 contributing to higher-order visual processing, may give the appearance of swapped gene expression during development.
Analyzing genes speci cally expressed in V1 and V2 at sublayer resolution is crucial for comprehending the developmental processes, cellular differentiation, circuitry, connectivity, functional specialization, and neurobiology of these visual areas.
Identi cation of spatiotemporally expressed genes speci c to area, time, and both area and time in the marmoset PFC The primate prefrontal cortex (PFC) is divided into several regions, including the dorsal prefrontal cortex (dPFC), the medial prefrontal cortex (mPFC), and the ventrolateral prefrontal cortex (vPFC) (Fig. 4a).These cortical areas are highly evolved in primate and they play distinct roles in higher-order cognitive processes and exhibit differences compared to the PFC of mice 27 .The PFC in primates possesses extensive interconnections with various cortical regions, encompassing sensory and association areas 28 .Consequently, the identi cation of genes speci c to area, time, and both area and time holds the potential to unveil on the mechanisms underlying the distinctive connectivity and functional development of the primate-speci c PFC.To identify these marker genes, we compared layer makers in each PFC area as landmarks and identi ed area-speci c and layer-speci c genes in PFC (Supplementary Table 5).We rst looked for genes which expression is stable during the development and have area speci city (Fig. 4b).Although the regional boundaries are less clear in PFC, some genes have relative regional identities (CBLN4 in dlPFC, PCDH17 in mPFC, and SYT17 in vmPFC and vPFC) (Fig. 4b and Supplementary Table 5).The primate PFC is known to be a complex cortical region with distinct subregions and cytoarchitectonic areas.Identifying regional markers in Supplementary Table 5 helps to de ne the boundaries and organization of different PFC areas, providing a structural framework for understanding its functional organization.In addition to the area marker genes, we identi ed genes with temporal expression (THBS1 and INSYN2A, Fig. 4c).Finally, we focused on genes that exhibited speci c layer and regional identities, leading us to identify genes which showed spatiotemporal expression pattern in the marmoset PFC (Fig. 4d).PRSS12 in mPFC and vPFC layer 2, CHRD in dPFC layer 5, SLIT3 in mPFC, FRZB in dPFC, and CCN3 in mPFC layer 2 exhibited transient expression during early developmental stage (Fig. 4d).We further compared the genes that show time-and region-speci c expression in the PFC of marmosets with that in the prelimbic region and infralimbic region (PrL and IL) of mice to determine whether they are involved in the speci city of the PFC of marmosets.The results revealed that number of genes expression speci c to marmoset (Fig. 4e).This suggests that the PFC is a highly evolved region in primates, and differences in gene expression during development may lead to unique circuit formation and function in primates.

Spatiotemporal gene expression controls PFC circuits
To understand gene function involved during early development in the marmoset PFC, we conducted GO analysis between datasets from early developmental stage and later developmental stage of mPFC.The GO analysis revealed the presence of abundant GO terms showing similarity to the category of "nervous system development" in the mFPC at the early developmental stage (Fig. 5a).Notably, a signi cant number of genes that showed spatiotemporal expression in the developing marmoset PFC were discovered to hold pivotal functions in neural circuit formation, including aspects like dendritic morphology and axon guidance (Fig. 5a and Supplementary Table 6).These results strongly suggest that the primate cerebral cortex undergoes dynamic neural circuit formation in the early stages of life.This process entails the region-and layer-speci c expression of genes critical for the development of neural circuits at various stages.However, it remains unclear how these spatiotemporal gene expression controls PFC development.
One of those genes, SLIT3 was highly expressed in marmoset mPFC but not in mouse PrL in early postnatal stage P0 (Fig. 5b).As mPFC receives input from various brain regions, including contralateral mPFC, mediodorsal thalamus (MD), and amygdala, we hypothesized that SLIT3, a known repulsive guidance molecule, controls axon guidance to the mPFC.To investigate this hypothesis, we tested SLIT3 function for axon guidance from MD to mouse PrL 29 .SLIT3 was ectopically overexpressed in mouse PrL layer 2 neurons by in utero electroporation (IUE), then MD axons was labeled by DiI placement on MD at P6 or 7 (Fig. 5c).Consistent with previous observations, MD neurons reached their axons to the mPFC in control animals (Fig. 5c).On the other hand, MD neurons failed to innervate into mPFC cortical layers when SLIT3 was overexpressed in PrL layer 2 neurons (Fig. 5c,d).This may seem different from the mPFC-MD connection in marmosets.However, when we looked at gene expression in MD, ROBO3, which repels SLIT3, is present in mouse MD but not in marmosets (Extended Data Fig. 4).This implies that gene expression in the mPFC might have changed to match the repulsive factor in the MD.Changes in the thalamus during evolution may have in uenced gene expression in the cerebral cortex or other way around.
PRSS12 (Serine Protease 12) exhibited speci c expression in layer 2 of mPFC at just one month of age, however, Prss12 didn't show such speci c expression in the layer 2/3 of the PrL or IL regions of the developing mouse PFC (Fig. 4e-f).PRSS12 is known as a diseases-associated gene, including Intellectual Developmental Disorder, Autosomal Recessive 1, and Autosomal Recessive Non-Syndromic Intellectual Disability 30 .Notably, mouse PrL and IL, which morphologically resemble the primate cingulate cortex, project to the contralateral cortex, amygdala, and substantia nigra (SNR) from postnatal day 10 31,32 (Fig. 5g).In contrast, comparing the projection to region A24 (anterior cingulate cortex) in marmosets reveals a localized projection to amygdala, with no projection to SNR and the contralateral cortex.This illustrates distinct circuitry between mice and marmosets (Fig. 5d and Extended Data Fig. 3).In the marmoset PFC, projections from region A25 (subgenual cingulate area) extend signi cantly to the contralateral cortex and amygdala, but not to the SNR (Extended Data Fig. 3).Consequently, when the Prss12 gene, which is not speci c to the mouse PrL, was overexpressed in the mouse PrL layer 2 through IUE, it resulted in reduced projections to amygdala and SNR, without altering projections to the contralateral cortex (Fig. 5g,h).Remarkably, this projection pattern mirrors that of marmoset A32, where PRSS12 is expressed in a temporally speci c manner: projection to contralateral cortex, amygdala (small projection) and SNR (small projection) (Fig. 5g-i, Extended Data Fig. 3).The homology between mouse PrL and IL and any primate brain region remains a complex question.Introducing a single PRSS12 gene to the mouse PrL, however, results in a connection pattern resembling marmoset A32.Finally, to investigate the similarity of the marmoset and human developing PFC, we examined the expression patterns of the human PRSS12 gene in the developing human PFC 33 .We found that PRSS12 is strongly expressed during early developmental stage in layer 2/3 neurons (Fig. 5j), which suggests that gene expression in the marmoset brain is similar to that in humans in many ways, and that the development of the marmoset brain is similar to that of humans.Understanding how genes are expressed in the developing marmoset brain will help us understand how the human brain develops and, in turn, the causes of developmental disorders.

Discussion
To advance the scienti c understanding of cortical development and function, it is essential to identify precise molecular markers exhibiting activity within speci c temporal periods and regions.A comprehensive analysis using spatial transcriptomics identi ed sexually, spatially, and temporally differentially expressed genes in the developing marmoset cortex.These results provide valuable insights into evolutionary perspectives and contribute to our understanding of the mechanisms underlying evolution, primate development, and developmental disorders.
Temporal marker genes expressed in speci c cortical layers are signi cant for understanding the dynamic processes of brain development and the establishment of functional circuits.The primate brain undergoes intricate developmental processes characterized by the sequential generation and maturation of different cell types and the establishment of neural circuits.Temporally expressed layer marker genes provide insights into the processes underlying circuit formation and plasticity during development, including synapse formation, neural connectivity, and synaptic plasticity, which are essential for establishing and re ning functional circuits.Comparing temporally expressed layer marker genes between species can shed light on the evolutionary modi cations that have shaped cortical development and function in different primate lineages, contributing to our understanding of the evolutionary origins and adaptations of the primate brain.
The results of our study highlight the intricate interplay between gene expression patterns in the marmoset cortex and their implications for biological and evolutionary processes.Through a comparative analysis with mouse and marmoset, we uncovered distinct expression pro les of key genes in different species.
Notably, the temporal expression pro les of several genes in marmoset mPFC differed signi cantly from those in mice.These differences in gene expression are likely important for marmoset-speci c circuit formation.For example, forced expression of Slit3 in mice inhibited projection from the MD.This is because the MD in mice has the receptor for Slit3 (Robo3), while the marmoset MD does not (Extended Data Fig. 4).This difference in gene expression suggests that not only the mPFC but also the thalamus in marmosets has its own gene expression pattern, leading to more species-speci c neural circuits.On the other hand, PRSS12 in the marmoset PFC, particularly its early localization to layer 2 and uncertain expression in PrL and IL regions in newborn mice, demonstrate intricate developmental variations.Interestingly, our observation that overexpression of PRSS12 in mouse PrL layer 2 led to reduced projections to the amygdala and SNR from PrL layer 2, akin to marmoset A32's projection pattern, suggests that speci c gene expression dynamics play a pivotal role in shaping primate-speci c circuits.Despite these valuable insights, the precise homology between mouse PrL and IL and their counterparts in primate brain regions remains a complex query, warranting further investigation.Nevertheless, our study underscores the intricate relationship between gene expression dynamics, brain region speci city, and timing in the establishment of species-speci c neural circuits within the prefrontal cortex.
The 10x Visium technology enables high spatial resolution mapping of gene expression patterns within tissue sections, allowing comprehensive transcriptomic pro ling of thousands of genes in a single sample.This technology preserves tissue morphology, facilitating histological examination or imaging alongside gene expression data.Analyzing and interpreting the large datasets generated by 10x Visium spatial transcriptomics requires specialized bioinformatics tools and expertise.To facilitate easy data access and analysis, a database has been created (accessible at https://gene-atlas.brainminds.jp/genevisium/,pass word: MMTD@7267, pass word will be removed after the acceptance) that offers visualization of the generated data.Recently, database of mouse and macaque spatial transcriptome data has been developed to allow comparison of gene expression patterns in different species 34,35 .These publicly available resources will greatly contribute to our understanding of the mechanisms of brain evolution, primate-speci c neural circuit formation and brain function, and the mechanisms of human brain pathology.

Method Animal
All procedures in marmoset and mice were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee of RIKEN Wako branch (W2022-2-026 and W2022-2-027).Marmosets were derived from a breeding colony at RIKEN Center for Brain Science.ICR (CD1) timed pregnant mice were purchased from Japan SLC.

Data processing for 10x Visium
For area de nition, AuA1, A3, dPFC, mPFC, and dPFC areas were identi ed based on H&E staining, while V1 and V2 areas were determined using area-marker genes and H&E staining.After selecting each area, the data was analyzed using the Seurat package in R 16,17 .The following functions in Seurat package were utilized: SCTransform function for normalization, RunPCA function for dimensional reduction, FindNeighbors function for computation of the nearest neighbors, and FindClusters function for determining clusters.
The unsupervised clustering of the cortex into cortical layers was determined based on characteristic cortical layer markers and the position of H&E staining pattern.Manual removal of unreliable spots from the clusters, such as empty spots and misplaced spots was performed.Additionally, overlapped, and crumpled tissue were excluded from the analysis, as these spots were often associated with unreliable clusters resulting from unsupervised clustering.When merging the data, batch effect correction was conducted using Harmony 18 .
For mouse mPFC analysis, mPFC including PrL and IL were selected by H&E staining patterns.Then, similar analysis described above using Seurat identi ed the clusters for cortical layers (L1, 2/3, L5, and L6).The marker genes for each cluster were determined using the FindAllMarkers function in Seurat.

Analysis
Cluster information obtained from the Seurat analysis was exported into Loup Browser software (10x, RRID: SCR_018555) to identify characteristic layer marker genes within the cortical layers.The top 100 up-regulated genes in each layer's cluster were ordered based on adjusted p-values using Benjamini-Hochberg correction.From these genes, we extracted common genes across the datasets of each area and each developmental period.These common genes were de ned as layer marker genes for each area at each developmental period.For de nition of temporal layer marker genes for each area, overlapping marker genes during developmental period were de ned as stable layer marker genes for the area, representing layer-speci c expression patterns throughout development.On the other hand, nonoverlapping marker genes at each developmental period were de ned as temporal layer marker genes for the area, representing layer-speci c expression patterns to certain developmental periods within the area.Within PFC, stable and temporal layer marker genes for mPFC, vPFC, and dPFC were compared between these areas to identify spatiotemporal layer marker genes, representing area-, time-and layer-speci c expression patterns within PFC.From the above analysis, the expression was manually validated by ISH using MGA.
For visual cortex (V1 and V2), to identify differentially expressed genes (DEGs), layer clusters within two developmental periods (P16 vs 1M, 1M vs 3M, and 3M vs 6M) within the same area were compared using the FindMarkers function in Seurat.The comparison was conducted using Wilcoxon Rank Sum test, with a log fold change threshold greater than 0.3.The same comparison was also performed between V1 and V2 comparison during development.After comparison, gene ontology (GO) analysis for biological process (BP), cellular component (CC), and molecular function (MF) were done using R package gPro ler2 and REVIGO website (http://revigo.irb.hr/) 36,37.For mPFC comparison during development, we compared the gene expression between P16 mPFC and 3M mPFC datasets including all cortical layers, then GO analysis using gPro ler and REVIGO identi ed the categories showing similarity to the category of "nervous system development" for BP.
For the analysis of sex differences, datasets for male and female PFC, A3, AuA1, and visual cortex were used for comparison (Supplementary Table 2).Entire Visium data including gray matter and white matter for male and female at each developmental stage (1M, 3M, and 6M) were compared using FindMarkers function.The comparison was conducted using Wilcoxon Rank Sum test, with a log fold change threshold greater than 0.25.Subsequently, differentially expressed genes between male and female derived from sex chromosomes were extracted.After merging all data for each area, average expression for each developmental stage was used for line graph.Also, their expression patterns were shown in violin plot.
For the plots, ggplot2 R package was utilized 38 .

Single-nucleus RNA sequencing (snRNA-seq) data
To investigate the expression pattern of PRSS12 in the human developing PFC, snRNA-seq data was obtained from publicly available dataset for the developing human PFC 33 .The L2/3-CUX2 cluster annotated in the original paper was extracted from the dataset to obtain the expression in the L2/3-CUX2 cluster during development (Fetal, Neonatal, Infancy, Childhood, Adolescence, and Adult de ned in the original paper).
In situ hybridization (ISH) for mouse ISH was performed as described previously 39 .In brief, mice were deeply anesthetized with a lethal dose of pentobarbitone (150 mg/kg, i.p.) and transcardially perfused with 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS).Brains were removed and post xed in the same xative for 1 h at 4°C and 2 h at room temperature and equilibrated with 30% sucrose in 4% PFA at 4°C overnight.Coronal brain sections of 28 μm thickness were generated with a sliding microtome (SM2020R, Leica).Antisense probes for Prss12, Slit3, and Tacr3 were generated from the template plasmids (FANTOM clone, 4833438J08 for Prss12, M5C1008N08 for Slit3, and B230383K19 for Tacr3).The expression patterns of each gene were validated in three independent brains.

Data acquisition for confocal imaging and quanti cation
Mice were deeply anesthetized with a lethal dose of Secobarbital (150 mg/kg, i.p.) and transcardially perfused with 4% PFA in PBS.Subsequently, the brains were removed, post xed in the same xative overnight at 4°C and equilibrated with 30% sucrose in PBS at 4°C overnight.For DiI labeling, the brains were removed, post xed in the same xative overnight at 4°C.DiI crystal (1, 1'-Dioctadecyl-3, 3, 3', 3'-Tetramethylindocarbocyanine Perchlorate) (D282, Molecular probe) were put into mediodorsal thalamus (MD) of hemi-dissected brains under microscopy, then the brains were incubated with 1% PFA in PBS at 37°C for 12 days.Coronal brain sections (100 μm thickness) were generated using a vibratome (VT1000 S, Leica, RRID:SCR_016495).These sections were counterstained with 1 μg/mL of 4′,6-diamidino-2phenylindole (DAPI) (11034-56, Nacalai Tesque).Fluorescent images were captured using a confocal microscopy (FV3000, Olympus, RRID:SCR_017015) or a uorescence microscopy (BZ-X810, Keyence).The images were imported into Fiji (NIH, RRID:SCR_002285) for the analysis.For the quanti cation of PRSS12 experiments, the average intensity of YPF uorescence was measured for contralateral mPFC layer 2, BLA, and SNR.The background signal intensity of YFP uorescence from adjacent contralateral mPFC layer 1 area, where mPFC layer 2 neurons do not project, was subtracted.Likewise, the background signal intensity of YFP uorescence from adjacent area of BLA and SNR was subtracted.The data were obtained from three independent brains for both control and PRSS12-overexpressed conditions.For SLIT3 experiments, the average intensity of DiI uorescence was measured for cortical layers and white matter in mPFC.The signal of cortical layers was normalized with the signal in white matter.The data were obtained from at least three independent brains for both control and SLIT3-overexpressed conditions.Declarations 39.Grove, E. A., Tole, S., Limon, J., Yip, L. & Ragsdale, C. W. The hem of the embryonic cerebral cortex is de ned by the expression of multiple Wnt genes and is compromised in Gli3-de cient mice.The areas analyzed for mPFC, vPFC, dPFC, A3, AuA1, V1, and V2, enclosed by a black dashed square, were shown in the images of Nissl staining of the 1-month marmoset brain.The brains were collected from postnatal day (P) 16, 1-month (1M), 3-months (3M), and 6-months (6M) marmosets.The target region from a fresh frozen section of the marmoset cortex and placing it on a 6.5 x 6.  (c) The stable cortical layer marker genes were con rmed in the developing marmoset cortex.In situ hybridization (ISH) con rmed the speci c cortical layer expression for NDNF(L1), CALB2 (L2), CPNE8 (L3), PLCH1 (L4), RORB (L4), ETV1 (L5), FEZF2 (L5), TLE4 (L6), and SYT6 (L6) in PFC, A3, AuA1, and V1 at P0. Scale bars: 1 mm.(d) Con rmation of expression patterns for PFC spatiotemporal marker genes within PFC.ISH for PRSS12, SLIT3, CHRD, FRZB, and CCN3 validated area-, time-, layer-speci c expression in developing PFC.Each PFC region is indicated in the schema (i: mPFC, ii: dlPFC, iii: vmPFC, iv: dmPFC).
(e) A dot plot showed distinct cortical layer marker genes in mPFC between mice and marmosets.
Comparison of mPFC layer marker genes between P7 mice and P16 marmosets identi ed conserved and marmoset-speci c layer marker genes in the developing mPFC.The upper panel displays conserved mPFC layer marker genes between species, while the lower panel displays marmoset-speci c mPFC layer marker genes.
Marmoset speci c genes control PFC neural circuits.
(a) GO analysis from the comparison between P16 mPFC and 3M mPFC.The graphs show GO terms showing similarity to the category of "nervous system development" for biological process (BP).

Figures
Figures
5 mm 10x Visium platform.The heat map displays markers across cortical layer cell populations, presenting the expression of the top 10 genes in the 1M AuA1 cortical layers.