Specification of claustro-amygdalar and palaeocortical neurons and circuits

Animals

All experiments were carried out with the protocol approved by the Committee on Animal Research at Yale University. The research methodologies for the use of mice (Mus musculus) and rhesus macaques (Macaca mulatta) were implemented following the guidelines sanctioned by the Yale University Institutional Animal Care and Use Committee (IACUC) and the directives of the US National Institutes of Health. The care and management of the animals were conducted within the precincts of the Yale Animal Resource Center, ensuring controlled environments for both prenatal and postnatal development stages of mouse and primate specimens. The mice were group-housed, maintaining a density of fewer than five individuals per cage, under environmental conditions regulated at 25 °C and 56% relative humidity, complemented by a photoperiod consisting of 12 h of light and 12 h of darkness. Nutritional needs were met ad libitum, coupled with veterinary oversight provided by the centre’s staff. The genetic lineage of the subjects was maintained on a C57BL/6 J strain, with experimental cohorts comprising both genders, randomly designated to respective studies. The detection of a vaginal plug in the mouse subjects was recorded as gestational day 0.5, marking the initiation of the experimental timeline. The Sox4^fl/fl and Sox11^fl/fl mice were a kind gift from V. Lefebvre. The Tfap2d-Lacz mice were a kind gift from M. Moser and the generation and genotyping were previously described³⁰ (Extended Data Fig. 7a). These mice were generated by inserting a Lacz cassette that disrupts the exon 1 of the Tfap2d locus, abolishing the proper expression of the trapped allele, Tfap2d. The Nex1-Cre (Neurod6-cre) were a kind gift from the laboratory of K.-A. Nave and CAT-Gfp mice were procured from Jackson Laboratories⁵². We bred these mice with Sox4^fl/fl and Sox11^fl/fl to obtain controls (Sox4^fl/+; Sox11^fl/+; Neurod6-cre; CAT-Gfp or Sox4^fl/fl; Sox11^fl/fl), Sox4-cKO (Sox4^fl/fl; Sox11^fl/+; Neurod6-cre; CAT-Gfp), Sox11-cKO (Sox4^fl/+; Sox11^fl/fl; Neurod6-cre; CAT-Gfp) and Sox4/Sox11-cdKO (Sox4^fl/fl; Sox11^fl/fl; Neurod6-cre; CAT-Gfp). Further we also bred Tfap2d^fl/fl and obtained wild-type control (Tfap2d^+/+; Neurod6-cre; CAG–CAT-Gfp or Tfap2d^fl/fl) Tfap2d-cHet (Tfap2d^fl/+; Neurod6-cre; CAG–CAT-Gfp) and Tfap2d-cKO (Tfap2dfl/fl; Neurod6-cre; CAG–CAT-Gfp). Refer to Supplementary Table 3 for genotyping details. Although blinding was not relevant for the primary mutant versus control comparison, other aspects of the study required careful design to minimize bias. Randomization was implemented while acquiring the data, including the behaviour data. Littermates (WT, HET and KO) were housed together, to avoid confounding housing effects on statistical analyses. The experimental cohort comprised male and female littermates aged between PD 120 and 180. Including samples from multiple litters further enhanced reproducibility.

Postmortem human and macaque brain tissue

Human brain samples were collected postmortem at PCW17. Rhesus macaque brain samples were collected postmortem at PCD105. Whole slabs or whole hemispheres were post-fixed in 4% paraformaldehyde (PFA) for 48 h and then cryoprotected in an ascending gradient of sucrose (10%, 20%, 30%). Tissue was handled in accordance with ethical guidelines and regulations for the research use of human brain tissue set forth by the NIH (http://bioethics.od.nih.gov/humantissue.html) and the World Medical Association Declaration of Helsinki (http://www.wma.net/en/30publications/10policies/b3/index.html). All experiments using non-human primates were carried out in accordance with a protocol approved by Yale University’s Committee on Animal Research and NIH guidelines.

Generation of Tfap2d-floxed mice

Guide RNA sequences (gRNAs) to insert flox sites were designed using an online program (http://zlab.bio/guide-design-resources)⁵³. gRNAs with the minimum off-target effects were selected, and DNA oligonucleotides carrying guide RNA sequences were cloned into pX330 vector. The sequences of the guide oligonucleotides used are listed in Supplementary Table 3. Cas9 mRNA were in vitro transcribed from vector px330 linearized by restriction enzyme NotI-HF (New England Biolabs, R3189S) and purified by phenol/chloroform extraction. We inserted the flox sites into intron 1 and intron 3. Founders were genotyped and bred at least 3 generations to exclude chimeras. Genotyping of mice for floxed allele was performed using primers listed in Extended Data Fig. 8a,b, Supplementary Fig. 1 and Supplementary Table 3.

In situ hybridization

The cDNA (Tfap2d human: ENSG00000008197; Tfap2d mouse:; Tfap2d macaque: ENSMMUT00000004057; Lmo3: ENSMUSG00000030226; Etv1: ENSMUST00000095767; Mef2c: ENSMUST00000163888, Cyp26b1: ENSMUST00000077705) for generating the human and mice probes was procured from Dharmacon and the plasmid was cut using NotI-HF enzyme. For the Tfap2d mouse probe, three different probes were amplified from PD0 and adult cDNA using the primers listed in Supplementary Table 3. The Tfap2d V3 probe worked best across ages and was used in the manuscript for detection of Tfap2d expression. The macaque probe was generated using respective neocortical tissues cDNA as a template by TA-cloning kit (Invitrogen, K202020). The cDNA clones were purified through the Qiagen clean up kit (Qiagen, 28104) and in vitro transcription (Millipore Sigma- Roche, 10999644001) was performed per the manufacturer’s instructions. Templates were purified by phenol/chloroform extraction and digoxigenin-labelled probes were synthesized using T3 (Roche, RPOLT3-RO) and T7 RNA polymerases (Roche, RPOLT7-RO) respectively, and RNA labelling mix (Roche, 11277073910) according to the manufacturer’s instructions. Probes were purified by ethanol precipitation, quantified, quality controlled and stored at −80 °C until hybridization. For in situ hybridization, slide-mounted cryo-sections at 60–70 μm thickness were processed. In brief, brains were fixed overnight at 4 °C in 4% PFA (Electron Microscopy Sciences, 15711) diluted in Dulbecco’s phosphate buffer (DPBS) (Thermo Fisher Scientific, 14190144), equilibrated for 12 h at 4 °C in 10% sucrose, and another 12 h at 4 °C in 30% sucrose in DPBS. Fixed brains were then embedded in OCT (Scigen, 23-730-571) and sliced on a cryostat (Leica Biosystem, CM1800). Slides were stored at −80 °C until processed for in situ hybridization. Sections were first post-fixed in 4% PFA in PBS for 15 min at room temperature, washed with PBS, and treated with 0.5 μg ml⁻¹ proteinase K solution (15 min at room temperature for P0 and 30 min at room temperature for adults). Slides were post-fixed in 4% PFA in PBS for 15 min at room temperature, washed with PBS. Slides were treated with acetic acid and triethonalamine solution for 10 min, followed by PBS washes. Slides were then submerged in hybridization buffer (5× SSC, 50% formamide, 20% SDS and 250 μg ml⁻¹ of Torula yeast RNA, aBSA (25 mg ml⁻¹), heparin stock (50 mg ml⁻¹)) supplemented with 1,000 ng ml⁻¹ of the appropriate digoxigenin-labelled probe at 70 °C overnight. Sections were washed 3 times for 45 min at 70 °C in 2× SSC, 50% formamide, 1% SDS, and then three times with TBST and incubated overnight at 4 °C with an anti-digoxigenin antibody conjugated to alkaline phosphatase (1:5,000, Roche, 11093274910). Sections were washed and then rinsed in the substrate buffer (100 mM Tris-Cl pH9.5, 100 mM NaCl, 50 mM MgCl₂, 0.1% Tween-20) before being overlaid with NBT/BCIP substrate (Roche). Colour development was done at room temperature in the dark until the desired signal was reached. Finally, sections were rinsed in EDTA solution and DPBS, washed in water and mounted with VectaMount AQ Aqueous Mounting Medium (Vector Laboratories, H-5501-60).

RNA in situ hybridization (RNAscope) analysis on mouse and chicken brain tissue

Brains from the embryonic chicken (PCD17) and mouse (PCD16.5) were fixed in 4% PFA overnight at 4 °C. They were sectioned at 20 μm and slides were stored in −80 °C until use. RNAscope was performed as per RNAscope Multiplex FL v2 protocol and kit (323270) from ACD (Advanced Cell Diagnostics bio) with slight modification. In brief, the slides were thawed and washed with 1× PBS for 5 min at room temperature, followed by incubation at 60 °C for 30 min in the HybEZ hybridization system (321711). Slides were fixed for 30 min at 4 °C, followed by dehydration with ethanol (50%, 70%, and 100%, each for 5 min). Slides were then air dried for 5 min and then incubated at 60 °C for 15 min. Slides were treated with hydrogen peroxide for 10 min and washed with distilled water twice for 2 min. They were again incubated at 60 °C for 15 min and then washed with boiling water for 10 s. Further, target retrieval was performed using the RNAscope target retrieval reagents for 5 min in a boiling beaker. The samples were washed with distilled water again and then incubated with 100% ethanol for 5 min. Slides were incubated at 60 °C for 15 min. Using a barrier pen a boundary was created and the slides are incubated with 5 drops of the RNAscope protease III for 30 min. Samples are then washed with distilled water and incubated with the C1, C2, C3 probes for 2 h at 40 °C. After that samples were incubated sequentially with RNAscope Multiplex FL v2 Amp 1, Amp 2 and Amp 3 for 30 min each with intermittent washes with 1× wash buffer twice for 2 min. The slides were then incubated with horseradish peroxidase–C1, followed by tyramide signal amplification, horseradish peroxidase blocker each with intermittent washes with 1× wash buffer twice for 2 min. These steps were repeated for C2, C3 and slides were mounting. Slides were imaged on the VS 200 microscope (Olympus Microscopy). For mice tissue, milder conditions were used. We used 1 h fixation, 2 min target retrieval treatment and proteinase plus 10 min. We used the following probes: chicken Tfap2d (NPR-0050976, ACD) and mouse Tbr1 (413301, ACD), Ngn2 (417291, ACD) and Sema5a (508091, ACD).

Immunohistochemistry

Brains isolated from prenatal and early postnatal mice were fixed overnight in paraformaldehyde (PFA) at 4 °C. Immunostaining was performed on the sections of 60–70 μm thickness cut on vibratome. Sections were blocked for 1 h at room temperature with blocking buffer consisting of 10% Donkey serum (Jackson ImmunoResearch, AB_2337258), 1% BSA (Millipore Sigma, A4612), 0.3% Triton X-100 (ThermoFisher, A16046.AP) in 1× PBS. After blocking, sections were incubated overnight at 4 °C with primary antibodies for SATB1 (1:500, Santacruz, sc-5989), TBR1 (1:1,000, Abcam, ab183032), NR4A2 (1:500, R&D systems, AF2156), CUX1 (1:1,000, Santacruz, sc-13024), BCL11B (1:2,000, Abcam, ab18465), FOS (1:500, Cell Signalling, 2250), BHLHE22 (1:1,000; Abcam, ab204791), GFP (1:500; Abcam, ab13970) and RFP (1:500, Abcam, ab124754). Sections were washed with washing buffer (0.3% Triton X-100 in 1× PBS) 3 times, 10 min each and were incubated with secondary antibodies (Jackson ImmunoResearch) for 1 h at room temperature, followed by 3 washes with washing buffer for 10 min each. Sections were mounted onto glass slides with vector shield (Vector labs, H-1000) and sealed with nail polish. All the slides were stored at −20 °C for further analyses. Antigen retrieval was performed on the brain sections prior to NR4A2 and TBR1 immunostaining (Dako Antigen Retrieval Solution, GV80511-2). For acquiring the images, LSM 800 (Zeiss Microsystems) and VS 200 microscope (Olympus Microscopy) were used. Analyses of the images was done using ZEN software, Qupath⁵⁴, OlyVIS or Fiji⁵⁵ using BioFormats plugin.

In utero electroporation

In utero electroporation was performed on PCD12.5 Sox4^fl/fl; Sox11^fl/fl timed-pregnant females with pNeurod1-cre and pCalnl-Tfap2d-Gfp on the right-side uterine horns and with pNeurod1-cre and pCalnl-Gfp into the left side uterine horns. Nearly 50% of all the embryos were injected with 0.5 μl DNA preparation (4 μg μl⁻¹ DNA mixed with 0.05% Fast Green FCF dye (Sigma-Aldrich)). The DNA was injected into the lateral ventricle of the embryos and guided for electroporation to the pallial–subpallial boundary using gene paddle electrode (EC1 45-0122 3 × 5 mm, Harvard apparatus) and square-wave pulse electroporator (Harvard Apparatus) at 30–33 V, 5 pulses, 50 ms ON and 950 ms OFF At PCD16.5, viability of the electroporated pups was screened and the embryos in healthy conditions were used for further analysis. GFP expression was grossly determined by NIGHTSEA Full System with UV (EMS, SFA-UV). Embryos with the fluorescence signal were dissected to isolated brains and were fixed for 24 h in 4% PFA overnight at 4 °C, and brains were analysed as previously described. The brains were sectioned at 70 µm into 5 wells of a 24-well plate serially and all the sections from 1 well were stained for GFP, CC3 (1:1,000, Cell Signaling (Asp175) Antibody 9661), which marks dying cells and ADGRE1 (1:500, Bio-Rad, MCA497RT), which labels infiltration and activation of microglia⁵⁶, using the protocol described above. For image acquisition, LSM 800 (Zeiss Microsystems) and VS 200 microscope (Olympus Microscopy) were used. Image analysis was done using ZEN software, Qupath⁵⁴, OlyVIS or Fiji⁵⁵ using the BioFormats plugin.

Quantification of RNAscope and immunostaining

Images were acquired using the VS 200 microscope and analysed with QuPath. The images were opened directly in QuPath, and regions of interest (ROIs) were outlined using the Allen Brain Atlas or Mouse Developing Brain Atlas as references. For quantification, 5–6 coronal sections per brain, spanning from the anterior to posterior axis, were analysed. As we collected 70-µm sections into 5 wells, the sections analysed were nearly 400 µm apart. For RNAscope, the brains were sectioned at 20 µm and collected onto 10 slides. The number of cells above a set threshold in their respective channel and DAPI-positive cells within the ROI were quantified in QuPath and their ratio to DAPI was then calculated. Data from mutant brains were normalized to the wild-type littermates for CC3 and ADGRE1. For the quantification of BHLHE22 intensity, regions were analysed using the straighten plugin in ImageJ. A custom-made ImageJ program was used to divide the straightened PIR into 200 equal bins. The BHLHE22 intensity distribution was determined by calculating the intensity per bin relative to the total intensity across all bins. The data were plotted into graphs using GraphPad Prism version 10.0.0 for Windows (GraphPad Software). For statistics, the two-way ANOVA, one-way ANOVA or pairwise comparisons were directly performed in the GraphPad Prism.

Generation and analysis of RNA-seq data

The cerebral cortex and amygdala from P0 Sox4/Sox11-cdKO, Sox11-cKO, Sox4-cKO and littermate control pups were collected (n = 3). Total mRNA was extracted using mirVANA Kit and DNAse digestion was performed using Tubro DNAse. Libraries were prepared with Illumina TruSeq mRNA preparation kit (Illumina RS-122-2101) for whole cortices, per the manufacturer’s instructions. Libraries were quality controlled by Tapestation/Bioanalyzer analysis and sequenced on the Illumina HiSeq 2000 platform at Yale Center for Genome Analysis (YCGA) to generate 75 bp single reads. Sequencing data were quality controlled by FastQC and aligned to the mouse genome (NCBI37/mm9) using STAR (v2.4.0e) (https://doi.org/10.1093/bioinformatics/bts635). To improve the mapping quality of splice junction reads, mouse gene annotation retrieved from the GENCODE project (version M1) was additionally provided (https://doi.org/10.1101/gr.135350.111). The command line “–sjdbOverhang 74” was used to construct a splice junction library. At least ten million uniquely mapped reads were obtained for each sample. Differential gene expression DEX analysis was performed by the R package DESeq2 (https://doi.org/10.1186/gb-2010-11-10-r106) and principal components analysis was performed by the R package prcomp. Genes with |log₂ fold change| ≥ 0.5 and a false discovery rate < 0.01 were classified as DEX. Integrated analysis was performed to identify the number of common or distinct DEX between Sox4/Sox11-cdKO, Sox11-cKO and Sox4-cKO. In total, 928 uniquely upregulated and 659 uniquely downregulated in Sox4/Sox11-cdKO were selected as candidate downstream targets, as shown in Supplementary Tables 1 and 2.

We further searched for cell types of amygdala using the public scRNA-seq dataset³⁶. To find cell-type marker genes, we used the Seurat FindMarkers function. We performed the Wilcoxon rank sum test, and considered genes with a minimum expression ratio of 0.2, adjusted P value less than 0.05, logfc.threshold greater than 0.25, expression percentages less than 0.2 in other subtypes, and a fold change of expression percentages greater than 1.5 between the Excit_B subtype and the second subtype where the genes are detected as marker genes.

Human gene expression analysis utilizing the Allen Brain Atlas and the BrainSpan Atlas

The images of brain structures presented in Fig. 2d were generated using the Freesurfer software⁵⁷. To identify cortical regions, we used the Desikan–Killiany atlas, a commonly used atlas for human brain mapping⁵⁸. TFAP2D expression values from two probes (A_23_P386973, CUST_2289_PI416261804) were taken from the Allen Brain Atlas³⁴ (adult data) and summed. Then, the average of log₂ expression values across six distinct individuals was computed. To visualize the expression levels of the gene, plot data, and to establish correspondence between the Desikan–Killiany atlas and the Allen Brain Atlas probes, the methods described in a previous study were used⁵¹. Subcortical regions were visualized using data from the Allen Brain Atlas website (http://atlas.brain-map.org). A standardized log₂ gene expression colour bar graph across both cortical and subcortical regions was created using a custom R script developed in-house. To generate the graphs in Fig. 2c (Allen Brain Atlas adult data³⁴) and Extended Data Fig. 4a (BrainSpan Atlas³⁵ midfetal data; www.brainspan.org), the log₂ TFAP2D expression values of the two different probes were averaged within a sample. For Fig. 2c, top-level structure names were changed to follow cortical cytoarchitectonics designations.

Comparative analysis of Tfap2d gene expression across mammalian and non-mammalian species

To assess the expression patterns of TFAP2D across different species, we reanalysed public single-nucleus transcriptome datasets for amygdala in human (Homo sapiens), macaque (M. mulatta), mouse (M. musculus) and chicken (Gallus gallus). We checked the expression level of Tfap2d, Sox4, Sox11, Lmo3, Etv1 and Mef2c across major cell types, space clusters and cell-type clusters, as identified by the original paper³⁸. For cross-species comparison, we only included the excitatory neurons.

Analysis of Tfap2d regulatory regions and SOX11 ChIP–seq data

To determine what regions within the Tfap2d locus can be potential enhancers, we re-processed a single-cell ATAC-seq embryonic dataset of the mouse cortex³³ using Signac version 1.1.0 (https://satijalab.org/signac/), to obtain ATAC-seq peaks that are accessible in migrating excitatory neurons in PCD13.5 onwards. We further refined the list of putative peaks by requiring that they intersect with occurrences of Sox4 or Sox11 JASPAR motifs (MA0867.1, MA0867.2, MA0869.1, MA0869.2). Similarly to regions directly bound by SOX4 and SOX11 within Tfap2d locus, we analysed data generated from ChIP–seq using an antibody against H3K27ac, a histone mark enriched in promoters and enhancers³². The SOX11 ChIP–seq data were directly accessed from Bergsland et al.²⁹.

Chromatin immunoprecipitation and rtPCR

Cortices from PCD15.5 from wild-type and Sox11-cKO mice were isolated and were crosslinked with a formaldehyde solution at a final concentration of 1% at room temperature for 10 min. l-Glycine (AmericanBio) at a final concentration of 125 mM was added and incubated for 5 min at room temperature to quench the cross-linking. The tissue was spun down and lysed in the hypotonic solution (50 mM Tris-Cl pH 7.5, 0.5% NP40, 0.25% sodium deoxycholate, 0.1% SDS, 150 mM NaCl) on ice for 10 min to obtain the nuclei. Nuclei were centrifuged at 600g for 5 min at 4 °C and pellets were resuspended in the SDS lysis buffer (1% SDS, 10 mM EDTA and 50 mM Tris-Cl, pH 8.1) prior to being sheared into 200–500 bp size fragments using a sonicator (M220 Focused-ultrasonicator, Covaris). The sheared DNA was diluted with the ChIP dilution buffer (0.01% SDS, 1.1% Triton X- 100, 1.2 mM EDTA, 16.7 mM Tris-Cl, pH 8.1, 167 mM NaCl) and was precleared with magnetic protein A/G beads (Thermo Scientific) for 1 h at 4 °C. For ChIP, 10 μg anti-SOX11 (Abcam, ab229185), 1 μg Pol II (Sigma-Aldrich, 05-623) and 10 μg of IgG (Diagenode, C15410206) were used. Samples were incubated on constant rotation overnight at 4 °C. Magnetic protein A/G beads (EMD Millipore,166-63) were blocked with 1 mg ml⁻¹ bovine serum albumin (BSA) (Sigma-Aldrich) and tRNAs were added to the chromatin-antibody complexes for 4 h at 4 °C. Beads were washed with low salt (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-Cl, pH 8.1, 150 mM NaCl), high salt (0.1% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-Cl, pH 8.1, 500 mM NaCl), LiCl (0.25 M LiCl, 1% IGEPAL CA630, 1% deoxycholic acid (sodium salt), 1 mM EDTA, 10 mM Tris-Cl, pH 8.1) and TE (AmericanBio), sequentially for 3 min each. ChIP DNA was incubated overnight at 65 °C for reverse cross-linking and subjected to RNAse A treatment (37 °C, 1 h) and proteinase K treatment (55 °C, 2 h), then purified on PCR purification columns. For input control, 5 μg crosslinked chromatin from each sample was also treated by reverse cross-linking, RNAse A (Thermo Scientific) and Proteinase K (Millipore Sigma) together with IP samples and purified by PCR purification columns. DNA amounts were quantified by PicoGreen assay (Thermo Scientific). The samples were eluted in 20 μl of TE. Each sample is diluted 1:5 and processed for rtPCR using the Bio-Rad SYBR Green Mix and Bio-Rad machine. Samples were run in triplicates and fold enrichment was calculated over the input. Two-way repeated measures ANOVA with Tukey’s multiple comparison correction was applied.

Enhancer conservation

We used mafsInRegion⁵⁹ from UCSC tools to obtain MULTIZ60 alignments for each selected enhancer. We stitched maf alignments and converted them to fasta using the script maf_to_concat_fasta.py from bx-python (bx-python). Pairwise alignment distance between species of each of the selected H3K27ac peaks were obtained using the function dist. alignment from the seqinr package in R.

Luciferase assays

For luciferase reporter plasmids, mouse Tfap2d enhancer candidate fragments (E1–E6) were amplified by PCR from genomic DNA and inserted into the pGL4.24 vector (E8421, Promega). Mutant form of each enhancer lacking SOX4 or SOX11 binding sites were also generated. The sequences of the PCR primers and PCR products are listed in Supplementary Table 3. The Neuro2a mouse neuroblastoma cell line was purchased from ATCC. The cell line was authenticated by morphology or genotyping, and no commonly misidentified lines were used. All lines tested negative for mycoplasma contamination, checked monthly using the MycoAlert Mycoplasma Detection Kit (Lonza). Luciferase assay was performed with Neuro2a cells basically as described⁶⁰. Neuro2a cells were transfected using Lipofectamine 3000 (L3000015, Thermo Fisher Scientific) with either mouse pCagig-Sox4−3×Flag, pCagig-Sox11−3×Flag or empty pCAGIG (Addgene #11159), together with one of the pGL4.24 luciferase vectors generated with enhancer sequences as described above. pGL4.73 The Renilla luciferase plasmid (E6911, Promega) was co-transfected as a control transfection efficiency. The luciferase assays were performed 48 h after transfection using the Dual-Luciferase Reporter Assay System (E1910, Promega) according to the manufacturer’s instructions. For each sample, three replicates were measured in a single assay. Luciferase activity was measured and quantified by GloMax Explorer (GM3500, Promega). The values for firefly luciferase/Renillia ratio for Sox11 or Sox4 were normalized to their respective pCagig controls.

Nissl staining

Perfused brains were sectioned at 40 μm thickness and mounted and dried on slides. A 1:12 series through the brain was stained for Nissl. In brief, sections went through a demyelination step including first an ascending series of ethanol followed by a descending series of ethanol, stained with cresyl violet acetate stain for 5 min and then briefly washed with water and ethanol. Sections were destained using a mild differentiating solution (50% ethanol + 5 drops of glacial acetic acid) and further dehydrated in an ascending series of ethanol before being placed in xylene and coverslipped. Slides were scanned with the Leica Aperio digital slide scanner at 20× magnification. Samples were observed for gross morphological alterations due to genotype.

Acetylcholinerase histochemistry and stereological analysis

For each animal, a 1:12 series of sections through the brain was stained with acetylcholinerase⁶¹, which demarcates the LA from the BLA. Each series contained between four and six sections through the anterior–posterior extent of the amygdala. Slides were scanned with the Leica Aperio digital slide scanner (Leica) at 20× magnification. Images used for stereological estimates were cropped using Leica’s Aperio ImageScope software and saved as JPEG. The Cavalieri estimator probe, implemented in Stereoinvestigator (MBF Bioscience) was used to calculate BLA and La volumes using a grid size of 10 μm. Volume estimates used are corrected for overprojection.

Diffusion tensor imaging

Nine postmortem mouse brains (3 wild-type control, 3 Tfap2d-Het and 3 Tfap2d-KO) were perfusion-fixed in with 4% paraformaldehyde solution in 0.1 M PBS for 24 h. Following perfusion–fixation, mouse brains were immersed and stored in 0.1 M PBS for 24 h and were transferred to Fomblin (SPI Supplies 69991-67-9) just before imaging. Diffusion MRI was acquired on a BioSpin 9.4 T MRI (Bruker) machine for each subject using a 3D echo-planner-imaging diffusion sequence (b = 1,500 s mm⁻²; 30 directions) with the following parameters⁶²: repetition time = 1,250 ms; echo time = 26 ms. The resolution was 0.1 mm isotropic. Overall scanning time was 19.5 h.

DTI image processing and tractography

Cerebral cortical and subcortical ROI and TH were manually defined according to Paxinos⁶³ and the Allen Mouse Brain Atlas⁶⁴ (https://mouse.brain-map.org/static/atlas) by D.A., N. Salla and N.K. without prior knowledge of the experimental groups. Whole-brain deterministic tractography^65,66,67 was performed with Diffusion Toolkit (Version 0.6.4.1) software with a fractional anisotropy threshold of 0.02 and an angular threshold of 60 degrees. Fibres that pass through each pair of ROIs were delineated and quantified. Visualization of the tracts and mouse brain was generated using MRtrix361 software⁶⁸ (v.3.0.0-65-g91788533). Circular visualization of the tractography derived connectivity between brain regions was generated using Circos software package⁶⁹.

Retrograde and anterograde neuronal tracing

To determine the effects of BLC deficits associated with Tfap2d loss and confirm the DTI results, we performed the AAV-based tracing from the mPFC, with retrograde and anterograde tracers injected simultaneously into the mPFC of PD120–PD180 mice. In brief, mice were anaesthetized by injecting a ketamine/xylazine solution and head fixed in a stereotactic frame. Thirty minutes before surgery, buprenorphine was administered. After lubricating the eyes and shaving the fur, an incision of <1 mm was made. A craniotomy was made with a round 0.5-mm drill bit at the desired coordinates (for the mPFC: medial–lateral ±0.35, anterior–posterior 2.0, dorsal–ventral 2.5). Using a Hamilton neuros syringe (0.5 ml), we mixed and injected 100 nl of each of AAVrg-CAG-Tdt (59462-AAVrg, Addgene) of AAV1-Camk2a-eGfp (50469-AAV1, Addgene) into the mPFC. To prevent the virus from spreading along the injection tract, the needle was held in place for at least 10 min. After injections, the skin was sutured, and the mice were returned to the cage. Approximately 3 weeks later, the mice were euthanized, and their brains were collected. The brains were coronally sectioned on a vibratome to obtain 70 µm thick sections. After staining the sections with anti-GFP antibody (1:500; Abcam, ab13970) and anti-RFP antibodies (1:500, Abcam, ab124754), the sections were imaged with VS 200 (Olympus Microscopy). The images were analysed using Qupath software⁵⁴ where the outline for the ROI was created and the number of cells projecting to mPFC were counted above threshold and the intensity of the axons in each region was measured using the ImageJ plugin.

Mouse behavioural assessments

We performed a series of tests on the mice to determine their exploratory, innate or learned anxiety-related behaviours^70,71. The experimental cohort comprised of male and female littermates aged between PD120 and PD180. Mice were acclimatized in the room for 30 min to 1 h before the beginning of any behaviour experiments. The experiment began with the apparatus being sanitized using 70% ethanol. In the following trials, water was used for cleaning instead. The first trial featured control mice and to mitigate the strong ethanol scent that might alter their behaviour, cage dust was rubbed on the apparatus. The results from this initial trial were omitted from the final analysis. The tests were performed in the order listed below with a spacing of 24 to 48 h between tests.

Open-field test

The rectangular arena (50 × 50 cm) was cleaned, and using ethovision software it was divided into centre, border and wall zones. Mice were placed in the centre of the open-field arena to allow them to explore the open-field area for 20 min. The activity of the mouse was videographed using Noldus Ethovision XT software. The EthoVision software automatically records the movement of the mouse, tracking the time spent in the centre versus the border zones, as well as the total distance moved and the speed of movement. The collected data were analysed to determine the preference of the mouse for the centre or border of the arena. Less time spent in the centre and more in the border indicates higher anxiety-like behaviour. Data are presented as mean ± s.e.m. Statistical analyses involved two-way ANOVA with Tukey’s multiple comparison correction, utilizing GraphPad Prism software (version 9). n = 16 (wild-type control), 33 (Tfap2d-Het), 29(Tfap2d-KO), 14 (wild-type control), 14 (Tfap2d-cHet) and 15 (Tfap2d-cKO).

O-maze

The elevated O-maze is clean and positioned in a quiet room and the EthoVision system was calibrated to recognize the different sections of the maze (open arms versus closed arms). Mice were gently placed in the centre of the closed arm of O-maze and allowed to explore freely for 10 min. The EthoVision software tracked movements, specifically recording entries into and the time spent in the open and closed arms. The data were analysed to assess the preference of the mouse for closed arms over the open arms, which indicates anxiety-like behaviour. The frequency of entries and the duration spent in each arm type are key metrics for this analysis. Statistical analyses involved two-way ANOVA with Tukey’s multiple comparison correction, utilizing GraphPad Prism software (version 9). n = 14 (wild-type control), 38 (Tfap2d-Het), 27 (Tfap2d-KO), 19 (wild-type control), 13 (Tfap2d-cHet) and 16 (Tfap2d-cKO).

Light-dark box

The assessment of anxiety levels using the light/dark box test was conducted in accordance with the methodology detailed by Bourin et al.⁷². In summary, the experimental setup involved a dual-compartment apparatus, each compartment made of opaque Plexiglas and measuring 18 cm in length, 10 cm in width, and 13 cm in height. The light compartment was illuminated by a 100 W lamp positioned overhead, with the light diffused through a transparent Plexiglas lid. Mice were granted access between the two compartments via a small aperture connecting them. Mice were initially placed within the illuminated section. Observations commenced from the moment a mouse first entered the dark compartment, continuing for a duration of 5 min. The amount of time the mouse spent in the light box was noted. The mice were scored blindly without information about their genotypes. The primary behavioural metrics quantified included the duration of time spent within the light compartment, measured in seconds, and the frequency of compartment transitions. Data are presented as mean ± s.e.m. Statistical analyses involved one-way repeated measures ANOVA with Tukey’s multiple comparison correction, using GraphPad Prism software (version 9). n = 10 (wild-type control), 10 (Tfap2d-Het) and 7 (Tfap2d-KO).

Tail suspension

Depression-like behaviours were measured using tail suspension and forced swim tests^73,74. Mice were suspended by the tail with a paper clip attached with adhesive tape about 5 mm from the end of the tail. Time spent immobile was recorded over the duration of 6 min. The video was recorded during the test. After completion of the test, mice were returned to a holding cage until all cage-mates were tested. After completion of the experiment, all mice were returned to the home cage and transferred back to the holding room. The videos were scored blindly by an independent person for the time mice were mobile and immobility time was calculated. Data are presented as mean ± s.e.m. Statistical analyses involved one-way repeated measures ANOVA with Tukey’s multiple comparison correction, utilizing GraphPad Prism software (version 9). n = 18 (wild-type control), 24 (Tfap2d-Het) and 13 (Tfap2d-KO).

Forced swim

Mice were gently introduced into transparent 4-l glass beakers filled with 2.5 l of water at ambient temperature. Each subject was gently introduced into the water, ensuring their ability to maintain their nostrils above the surface for breathing. On day 1 of the experimental procedure, mice were exposed for a duration of 15 min. Subsequently, on the following day, the immersion time was reduced to a series of 5 min. The videos of the test were recorded and were scored blindly by an independent person for the time mice were mobile and immobility time was calculated. Immobility was characterized by a state of passive floating, wherein the subject performed minimal movements necessary to remain afloat. Data are presented as mean ± s.e.m. Statistical analyses involved one-way repeated measures ANOVA with Tukey’s multiple comparison correction, using GraphPad Prism software (version 9). n = 20 (wild-type control), 23 (Tfap2d-Het) and 13 (Tfap2d-KO).

Fear conditioning

The experimental cohort comprised of male and female littermates’ siblings aged between PD120 and PD180. The method used for contextual and cued fear conditioning was outlined previously^75,76,77,78, with some modifications. The experiment consisted of four parts: habituation, training, contextual memory assessment, and cued memory evaluation. During the habituation phase, mice were introduced to a dimly lit enclosure composed of dark plastic with a cardboard floor, incorporating sawdust from their original housing to establish context A, without the administration of shocks or cues. For the training phase, the subjects were relocated to a differently structured chamber equipped with illuminated stainless steel grids capable of delivering electric shocks, designated as context B, and infused with a cinnamon scent beneath the grids to serve as a unique olfactory stimulus. Following a 160-s acclimatization period, an auditory cue (75 decibels, 2.8 kHz) was emitted for 20 s, with the concluding 2 s overlapping with an electric shock (0.5 mA). The mice experienced 6 instances of tone-shock associations, interspersed with 40-s intervals, and concluded with a 60-s relaxation phase before being returned to their original cages. The subsequent day involved testing for contextual memory in the same chamber (context B) but with the introduction of a vanilla scent beneath the grids to differentiate from the training phase’s cinnamon odour, thereby focusing on contextual rather than olfactory memory recall. Cued memory was assessed in the original context A setup, using the same auditory cues as during training but omitting the shocks. All sessions were video recorded, and the freezing behaviour of the mouse—defined as the absence of movement for each 3-s video frame—was analysed. No significant variances in body weight or shock response were noted. Data are presented as mean ± s.e.m. Statistical analyses involved two-way ANOVA for the training and cued memory test and multiple comparisons with Tukey’s correction, using GraphPad Prism software (version 9). For conditioned memory test, one-way ANOVA with Tukey’s correction was applied. n = 27 (wild-type control), 38 (Tfap2d-Het), 26 (Tfap2d-KO), 16 (wild-type control), 12 (Tfap2d-cHet) and 16 (Tfap2d-cKO).

FOS immunostaining and registration to the mouse brain common coordinate framework

To determine the activity of different brain regions after cued memory test, we performed a whole-mount FOS staining using the LifeCanvas Technologies. We could not utilize the extensively used, targeted recombination in active populations (TRAP) mice⁷⁹ for these experiments because Neurod6-cre is expressed in the cerebral excitatory neurons of the conditional Tfap2d mice, leading to reporter activation during prenatal development, which occurs prior to FOS-driven Cre-ER(T2) expression. Mice were euthanized 60–90 min after the cued memory test using isofluorane followed by transcardial perfusion with ice-cold 1× PBS with 10 U ml⁻¹ heparin until the fluid ran clear, followed by ice-cold 4% PFA. The extracted brains were incubated in 4% PFA solution at 4 °C for 24 h with gentle shaking. The samples were cleared and stained by LifeCanvas Technologies using their protocol⁸⁰. Samples were cleared using a SHEILD OFF solution for 3 days at 4 °C, followed by incubation in the SHIELD ON for 24 h at 37 °C with light shaking. Samples were then processed for delipidation and staining using SmartBatch + . The brains after delipidation were blocked with donkey serum or 2 days, followed by primary antibody incubation for 1–3 days and secondary antibody for incubation 6 h. After each primary and secondary antibody incubations, stringent washes and PFA fixation was performed. The brains were imaged using SmartSPIM at 4 μm z-step and 1.8 μm xy pixel size. After imaging, the samples underwent automated atlas registration to the Allen Brain Atlas via LifeCanvas Technologies using rigid, affine, and b-spline warping algorithms provided by SimpleElastix, for both propidium iodide and FOS (647) channels. For the KO brains where the BLC shape was altered, the protocol was modified and manually registered for atlas registration. For cell detection, custom cell detection networks developed by LifeCanvas Technologies, employing a two-step approach with a U-Net architecture for candidate detection and a 3D ResNet for classification, and were trained on hand-tagged 3D ROIs to accurately identify cell locations. The detected cells were then mapped onto the Allen Brain Atlas for quantitative analysis. Colocalization analysis determined cell co-expression of FOS and propidium iodide by measuring the Euclidean distance between cells in different channels, with co-expressed cells within a 3-voxel distance threshold being counted for each atlas-defined brain region. The output generated would provide the number of FOS-positive cells per mm³ in each region labelled by the Allen Brain Atlas.

Functional network construction

We developed a comprehensive analytical framework to examine the relationship between various regions functionally active after cued memory test based on the FOS expression^81,82. The analysis was performed using Python, with key libraries including Pandas for data manipulation, Seaborn and Matplotlib for visualization, NetworkX for network analysis, and SciPy for statistical tests. We calculated Pearson correlation coefficients to assess the linear relationship between each pair of subregions^41,82. To evaluate the statistical significance of these correlations, P values were computed. Heat maps were generated to visualize the correlation matrix amongst brain regions associated with threat responding, with annotations indicating the significance levels. The upper triangle of the matrix was masked to prevent redundancy, given the symmetric nature of correlation matrices. A diverging colour palette from blue (negative correlation) to red (positive correlation) was applied for visual clarity. Network graphs were created to visualize the connectivity between subregions based on their correlation coefficients. Edges were drawn for correlations surpassing a defined cutoff of 0.85, emphasizing stronger relationships. Edge colours transitioned from blue to red, representing the spectrum of correlation strengths. We compared the correlation coefficients between experimental groups (wild-type, Het and KO) using Fisher’s z-transformation followed by significance testing. First, Pearson correlation coefficients were computed for each group based on the averaged data across grouped regions. To stabilize variance and allow for the comparison of correlations between groups, we applied Fisher’s z-transformation to the correlation coefficients. We then calculated the z-score for the difference between the two z-transformed correlations across groups. The P value associated with the z-score was determined using the cumulative distribution function of the standard normal distribution, and statistical significance was annotated on heat maps of the correlation differences with significance thresholds. The data related to these experiments is available in Supplementary Table 10 and Supplementary Tables 20 and 21. Abbreviations are available in Supplementary Tables 4 and 5.

Statistics and reproducibility

All data are reported as mean ± s.e.m. with data analysis being conducted using one-way and two-way ANOVA with Tukey’s post hoc multiple-comparisons adjustment or unpaired t-tests for comparisons between two groups. Fisher’s exact tests were also used and the details of statistics are presented in the Supplementary Tables 6–21. Details of the number of replicates in each experimental group together with appropriate statistical analyses are shown in the figure legends. For the representative images the numbers of biological replicates used are following. Figure 1a includes data from 3 brains per genotype for Nissl, Lmo3, Etv1 and Mef2c staining, and 4 brains per genotype for GFP and SATB1. Figures. 1d and 3b,c use 4 brains/genotype. Figures 2b and 5f use 2 brains, and Figs. 2e and 5c include data from 3 brains per age. Extended Data Figs. 1a,b and 2a,c incorporate 4 brains per genotype. Extended Data Fig. 11b uses 15 or more samples per genotype. These animals were tested for recombination events for 5′ and 3′ sites of the Tfap2d-floxed allele across multiple generations. We did not observe recombination between them. Finally, Extended Data Fig. 11c and Fig. 12a–c use 3 brains per genotype. Significance threshold was set at P  <  0.05. All statistical analysis and plotting were performed in GraphPad 9 and 10 (GraphPad) or in Python. All of the figures were created using Adobe Illustrator (Adobe Systems).

Reporting summary

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