Category: 3. Business

Animals

Housing, handling, and all experimental protocols for mice were performed in accordance with the Regulations for the Administration of Laboratory Animals of East China Normal University (ECNU) and were approved by the Animal Care and Use Committees of ECNU (Approval ID: 10269). We have complied with all relevant ethical regulations for animal use. Mice were housed in a certified specific-pathogen-free (SPF) facility under standardized conditions: housed in standard filter-top cages (3-5 animals per cage) with free access to water and food. They were maintained on a 12:12-h light/dark cycle (07:00 to 19:00 lighting) at 22°C with relative humidity of 50–60%. The noon of the day when the vaginal plug was found was counted as embryonic day (E) 0.5.

Wild-type C57BL/6 J mice (MGI: 3028467) were purchased from Shanghai Jihui Laboratory Animal Care Co., Ltd. and ICR mice (CD-1, MGI:5649524) were obtained from Hunan SJA Laboratory Animal. Med13l knockout mice were generated by CRISPR/Cas9 in C57BL/6 N zygotes at Cyagen Biosciences (C57BL/6N-Med13l^em1cya/Cya, Cyagen Ltd., Nanjing, China; Stock No. S-KO-14807; MGI: 7835672) and backcrossed > 5 generations to C57BL/6 J, then crossed with ICR mice to establish experimental breeding colonies. All KO experiments used littermate controls from these crosses. Med13l-floxed mice (C57BL/6JGpt-Med13l^em1Gpt/Gpt; Gempharmatech Ltd., Nanjing, China; Stock No. T010181; MGI: 6426082), and Emx1-Cre mice (B6.129S2-Emx1^tm1(cre)Krj/J; Jackson Laboratories, JAX: 005628, MGI: 3617405). Genotyping and sequencing primer sequences for all the aforementioned mouse lines are provided in Supplementary Data 5.

Med13l knockout mice were generated by CRISPR/Cas9-mediated gene editing in C57BL/6 N mouse zygotes³⁸. Guide RNAs (gRNAs) were designed to target Exon 2 (gRNA1: 5′-GCACGTGGGCTGTTCGATACAGG-3′; gRNA2: 5′-GCTTCACACCACCTACGGTTTGG-3′) to delete a fragment of 238 bp. Specificity of the gRNA was evaluated using the online tool available at crispr.mit.edu. A mix of Cas9 mRNA and gRNAs was microinjected into zygotes. An F0 founder carrying a 3187-bp deletion and an 8-bp insertion was identified by Sanger sequencing (Supplementary Fig. 1A; primers listed in Supplementary Data 5), resulting in complete deletion of Exon 2 and a frameshift that introduced a premature stop codon. For the floxed allele, gRNAs targeting Exon 2 (gRNA1: 5′-TAACCATCTGCAACCCAACC-3′; gRNA2: 5′-AGCACTCCCCACGGAGACAT-3′) were used with Cas9 mRNA and Flox Donor DNA; Med13l-floxed mice (Gempharmatech Stock No. T010181) were later crossed with Emx1-Cre (JAX: 005628) mice for forebrain-specific knockout.

Sanger sequencing

Genomic DNA was amplified by polymerase chain reaction (PCR), and the resulting products were purified and directly sequenced using an ABI PRISM 3730 automated sequencer. Sequencing chromatograms were aligned to the mouse reference genome using Chromas software. Due to the 3187-bp deletion, only the junction regions of the deleted allele are presented in Supplementary Fig. 1A.

Histological analysis

For newborn mice (P0), the heads were removed after ice anesthesia, and the brains were dissected and photographed under a stereoscopic anatomical microscope (Leica M165 FC). For adult mice, animals were anesthetized by 0.7% (w/v) pentobarbital sodium solution at 5 mL/kg (35 mg/kg body weight), followed by perfused with precooled phosphate buffer (PBS, pH 7.4) and 4% paraformaldehyde (PFA). The brains were dissected and photographed. Morphological parameters measured included the width of cortex (WC1), longitudinal length of cortex (LC1), oblique length of cortex (LC2), and width of cerebellum (WC2).

For brain slice preparation, brains were post-fixed in 4% PFA (pH 7.4) for 24 h. (P0) or overnight (adult), followed by dehydration in 20% and 30% sucrose solutions in PBS. Brains were embedded in OCT compound (SAKURA, USA). Coronal sections of 40-μm thickness were cut using a cryostat (Leica CM1950). Glass slides with brain sections were treated with Nissl staining solution for 15 min at 55 °C. Microscopic images of Nissl-stained brain slices were collected using the TissueFAXS Plus ST system (TissueGnostics GmbH, Vienna, Austria). The thickness of the motor cortex (M1) and primary somatosensory cortex (S1ULP/S1BF) were measured in microscopic images of Nissl-stained brain slices using Image J software. The volumes of the unilateral cerebellum or striatum were obtained by cumulative integration of areas of all brain slices encompassing the entire brain region.

Immunofluorescence

Glass slides with brain sections were washed with PBS (pH 7.4) three times and incubated for 15 min in 10 mM sodium citrate buffer (pH 6.0) at 99°C for antigen retrieval. Sections were then incubated in blocking buffer (5% BSA, 5% goat serum, and 0.25% Triton X-100 in PBS) for 2 h at room temperature. Subsequently, sections were incubated with specific primary antibodies for 48 h at 4 °C. For primary antibodies, we used mouse anti-BrdU (1:500, Cell Signaling, 5292S), rabbit anti-Pax6 (1:300, MBL, PD022), rabbit anti-Tbr1 (1:1000, Abcam, ab31940), rabbit anti-EOMES (1:1000, Abcam, ab183991), rat anti-EOMES (1:100, Invitrogen, 14-4875-82), rat anti-Ctip2 (1:1000, Abcam, ab18465), rabbit anti-DCX (1:800, Cell Signaling, 4604), rabbit anti-Ki67 (1:200, Cell Signaling, 12202). After rinsing in PBS three times, sections were incubated with secondary antibodies in the dark for 2 hr. at room temperature and then thoroughly rinsed. Secondary antibodies were: goat anti-rabbit, Alexa Fluor 546 (1:1000, Invitrogen, A11035), donkey anti-mouse, Alexa Fluor 647 (1:1000, Invitrogen, A32787), goat anti-rat, Alexa Fluor 647 (1:1000, Cell Signaling, 4418), goat anti-rabbit, Alexa Fluor 488 (1:1000, Invitrogen, A11008), goat anti-rabbit, Alexa Fluor 647 (1:1000, Cell Signaling, 4414). Nuclei were labeled by incubating sections in PBS containing 1 μg/ml 4’,6-diamidino-2-phenylindole (DAPI) (Thermo Fisher Scientific, 62248), and samples were mounted using ProLong Diamond Antifade Mountant (Thermo Fisher Scientific, P36970).

For BrdU labeling, time-pregnant mice were intraperitoneally injected with 50 mg/kg body weight BrdU (Abcam, ab251467). After 1.5 h. (proliferation assay) or 24 h. (differentiation analysis), mice were anesthetized, and the brains of embryos were dissected on ice and fixed with 4% PFA for 4-6 h. Cryostat sections of 14-μm (1.5-h pulse) or 16-μm (24-h pulse) thickness were prepared. Sections were treated with 2 N HCl for 15 min at 37 °C before immunofluorescence staining.

Fluorescence images of brain sections were collected using the TissueFAXS Plus ST system (TissueGnostics GmbH, Vienna, Austria) and further processed with Adobe Illustrator. For cell counting, a 250 μm wide rectangular column was placed perpendicular to the ventricular surface in the cortical area, and the cell number in the column was counted using the cell counter program of Image J software.

Behavior assessments

Behavioral tests and data analyses were conducted according to previously described protocols³⁹. Briefly, 17- to 20-week-old age-matched male and female mice were used. Mice were handled for 5 min each day for 3 days prior to the behavioral tests. All behavioral tests were conducted during the light cycle by an experimenter blind to the mouse genotypes. The surfaces of the behavioral apparatus were cleaned using 75% ethanol before each experiment and between trials, with at least a 5-min wait before the next test to allow ethanol evaporation and odor dissipation.

Open-field test

The Panlab Infrared (IR) Actimeter system was used for the open-field test. The test mouse was gently placed near the wall-side of a 20 cm × 20 cm × 25 cm open-field arena and allowed to explore freely for 30 min. The movement of the mouse was recorded by a video camera and analyzed with the IR Actimeter. The total locomotion distance, the time exploring the central zone (10 cm × 10 cm), and the number of hindlimb rearing events in the central zone were measured.

Balance beam test

The experimental device consisted of a flat surface rod, 100 cm long and 1 cm wide, placed 50 cm above the ground with an upward angle of 30°. A target cage was placed at the upper end of the rod. The experiment was divided into a 3-day training phase and a testing phase. During the training phase, the tested mouse was guided to walk through the balance beam from the lower end toward the upper end once daily. On the fourth day, the mouse was tested, and the time required for each mouse to traverse the balance beam without guidance was recorded. Mouse movement was video recorded from the front of the balance beam for detailed analysis.

Rotarod test

The test was conducted using a 5-channel rotarod device (DigBehv-RRTM, Shanghai Jiliang Software Technology Co., Ltd., Shanghai, China). The total experiment spanned 7 days, with 4 trials per day and a 20-min interval between trials. The speed of the rotating rod was set to smoothly accelerate from 4 rpm to 40 rpm within 5 min. The maximum time and speed which the testing mice kept balance on the rotating rod were recorded.

Hindlimb clasping test

The hindlimb clasping test followed a 2-level scoring system⁴⁰. Each mouse was tested three times, with each trial lasting 10 s. The tail of the mouse was gently lifted to suspend the mouse head down to observe whether the hind limbs moved closer to the abdomen. If the hind limbs were always stretched out away from the abdomen, with both legs open, a score of 0 was given. If the hind limbs moved close to the abdomen for more than 50% of the time, it was considered moderate paw clasping, and a score of 1 was given. The average score of each mouse was calculated.

Gripping strength test

The experiment was conducted using a grasping net (100 cm × 50 cm) placed 50 cm above the ground. The mouse was placed on top of the net, which was then quickly and smoothly turned upside down, so the mouse hung under the net. The time from reversing the net to the mouse falling was recorded as the holding latency. Five repeated trials were performed for each mouse, with a 20-min interval between trials. For latencies longer than 120 s, a value of 120 s was recorded. The average latency of the 5 trials was calculated for each mouse.

Morris water maze

The standard Morris water maze was used to assess the spatial learning and memory abilities of the experimental mice. A stainless water-filled circular tank (210 cm in diameter and 50 cm in height) with non-reflective interior surfaces and ample visual cues was used. White edible emulsifier was added to the water to make it non-transparent. The pool was divided into four quadrants, with a circular hidden platform (13 cm in diameter) placed 2 cm below the water surface in the center of the target quadrant. The pool was surrounded by black shading curtains with spatial cues of different shapes and colors. The water temperature was maintained at 22–24 °C with a heating device. A camera above the pool recorded animal behavior, controlled by SuperMaze software. The experiment spanned 5 days, including 4 days for training and 1 day for testing. Each mouse trained with 4 trials per day, being gently released into the tank with its head toward the inner wall from a random quadrant. If the mouse found the platform within 60 s, the time was recorded, and the mouse stayed on the platform for 20 s. If the mouse did not find the platform within 60 s, it was guided to the platform and allowed to stay for 20 s, with a latency of 60 s recorded. On the 5th day, two independent 60-s tests were conducted without the hidden platform, recording the target quadrant entry times and total swimming time inside the target quadrant. After testing, mice were dried and returned to a dry cage to prevent stress from low body temperature.

Three-chamber social test

A modified three-chamber sociability test was conducted in a transparent rectangular chamber (40 cm × 30 cm × 25 cm) divided into three equal-sized chambers (13.3 cm × 30 cm × 25 cm) with two doors (3 cm × 5 cm) connecting the central chamber to the side chambers. The social partner mouse was tied in the corner of a side chamber with a 25 cm polyacrylonitrile string. In stage 1, the test mouse was placed in the central chamber and allowed to explore freely for 10 min to adapt. In stage 2, the test mouse was placed in the central chamber, a stranger mouse (stranger 1) was tied in one side chamber, and a mouse doll was placed in the other side chamber. The mouse’s behavior was recorded for 10 min. In stage 3, the doll was replaced by another stranger mouse (stranger 2), and behavior was recorded again. The total sniffing time of the test mouse on each stranger mouse or the doll was measured.

Self-grooming

Self-grooming was analyzed as described previously³⁹. The test mouse was placed in the center of an open field arena (20 cm × 20 cm × 25 cm) and allowed to explore freely for 20 min. The first 10 min analyzed self-grooming in an unfamiliar environment, and the last 10 min analyzed self-grooming in a familiar environment. The self-grooming time and frequency were determined by manual review of the video recordings.

RNA isolation and qRT-PCR

Mice were sacrificed by cervical dislocation at the indicated time points. Total mRNA from mouse brain tissues was isolated using the RNAiso Plus kit (Takara, #9109) on ice. Phase separation was achieved with 200 µl or 100 µl chloroform. After centrifugation at 12,000 rpm for 20 min at 4 °C, RNA was precipitated by mixing the aqueous phase with equal volumes of isopropyl alcohol and 1 µl of 20 mg/ml glycogen. The RNA precipitates were dissolved in DNase/RNase-free water. cDNA was synthesized using a cDNA reverse transcription kit (KR116-02, Tiangen, China) from 1 µg of total RNA. qPCR was performed using the SYBR Green qPCR kit (AG11701, Accurate Biology, China) on a CFX96 Real-Time System (Bio-Rad). The thermal profile was 95 °C for 5 min, followed by 40 cycles of 95 °C for 10 s, 60°C for 20 s, and 70 °C for 30 s. Gapdh gene was used as the internal control. The relative expression level of target genes was normalized to the Ct value of Gapdh using the 2^−ΔΔCt relative quantification method. The primers used are listed in Supplementary Data 5.

Bulk RNA-seq

Total RNA was extracted as described above. Bulk RNA-seq was conducted by Shanghai Genergy Bio. Co., Ltd. The concentration and quality of RNA were measured using Qubit fluorometer (Thermo Fisher Scientific, Q33226). RNA-seq libraries were constructed using the VAHTS Universal V10 RNA-seq Library Prep Kit (Premixed Version, NR616). Briefly, mRNA was extracted using poly-A selection with magnetic beads with poly-T and then converted into cDNA through first and second-strand synthesis. The newly synthesized cDNA was purified using AMPure XP beads (1:1) and eluted in 50 μl of nucleotide-free water. RNA-seq libraries were sequenced on NovaSeq XPlus platform with paired-end reads of 150 bp, achieving a sequencing depth of 60 million reads per library.

Quality control of the RNA-seq data was performed using FastQC (version 0.11.9). Paired-end reads were trimmed to remove adapters and low-quality reads using Skewer (version 0.2.2). Clean reads were aligned to UCSC mm10 mouse genome using STAR (version 25.2b) with default parameters. The number of mapped reads was counted using stringtie (version 2.2.1). The resulting read counts were processed with the R package DESeq2 (version 1.18.1) to identify differentially expressed genes (|log2 fold change| > 1 and p-value < 0.05) between groups. Gene expression levels were normalized by fragments per kilobase of transcript per million mapped reads (FPKM). Raw sequencing data were deposited in the Gene Expression Omnibus under accession number GSE277054.

Single-cell RNA sequencing (scRNA-seq)

The scRNA-seq was performed by Shanghai OriginGene-Tech Biotechnology Co., Ltd. Cortices from 5 embryos of the same genotype from 3 litters were pooled into 5 ml tissue storage solution (Miltenyi Biotec, Inc., MD, USA) on ice and carefully cut into small pieces of 1–3 mm. Tissue samples were subjected to enzymatic digestion with 0.25% Trypsin-EDTA (Thermo Fisher Scientific) at 37 °C for up to 30 min. Cell suspensions were centrifuged at 300×g for 5 min at 4 °C. The pellets were further digested in basal media supplemented with 0.2% type II collagenase (Sigma-Aldrich) at 37 °C for up to 4 h. Isolated cells were filtered through 70 μm nylon filters (BD), washed twice with sterile PBS (pH 7.4). Samples with a cell density of 700–1200 cells/μl and a viability higher than 85% were proceeded for scRNA-seq.

Droplet-based scRNA-seq was performed using the 10× Genomics Chromium Single Cell 3′ Reagents Kit (v3.1)⁴¹. The single-cell samples were passed through a 40 μm cell strainer and counted using a hemocytometer with trypan blue. Subsequently, the single-cell suspension, Gel Beads and Oil were added to the 10× Genomics single-cell A chip. After droplet generation, samples were transferred into PCR tubes and reverse transcription was performed using a T100 Thermal Cycler (Bio-Rad). After reverse transcription, cDNA was recovered using a recovery agent, provided by 10× Genomics, followed by silane DynaBead clean-up as outlined in the user guide. Before clean-up using SPRIselect beads, cDNA was amplified for 10 cycles. Libraries were submitted to 150 bp paired-end sequencing on an Illumina NovaSeq 6000 platform, yielding 150 Gb of sequencing data per sample. ScRNA-seq data are available in GEO under accession number GSE277054.

Data analysis was conducted according to previous studies^42,43. CellRanger 6.6.1 was applied for preliminary data analysis. Raw data from single-cell RNA-seq libraries were first trimmed to remove the template switch oligo (TSO) sequence and poly(A) tail sequence. Reads with adaptor contaminants and low-quality bases were removed. Clean reads were aligned to the reference genome (GRCm39/mm39). Gene expression levels were quantified as transcripts per million (TPM). Valid barcodes and Unique Molecular Identifiers (UMIs) were used to generate the gene-barcode matrix for single cells. Seurat v4.1.1 was used for quality control and data analysis in R. Cells with fewer than 200 genes, more than 6,000 detected genes, or with more than 10% of transcripts from mitochondrial genes, were excluded to maintain data quality.

Principal component analysis with variable genes was used to identify significant principal components (PCs) based on the JackStraw function. Twenty PCs were selected as the input for uniform manifold approximation and projection (UMAP) and t-distributed stochastic neighbor embedding (t-SNE) when statistically significant. Cell clusters were annotated manually based on marker gene expression and developmental information. Defined cell clusters included neuroepithelial cells (NEs), cycling and non-cycling radial glial cells (RGCs), cycling and non-cycling intermediate progenitors (IPCs), pyramidal neurons (PyNs), interneurons (INs), Cajal-Retzius cells, ependymal progenitors (EpPs), oligodendrocyte-astrocyte progenitor cells (OAPCs), and microglia, among others, based on their specific marker gene panels (detailed in Fig. 2 and Supplementary Data 2). These clusters were assigned to broad lineages, such as the interneuron lineage, pyramidal lineage, NG2-oligodendrocyte lineage, or microglial lineage. The Monocle3 algorithm was used to infer the developmental trajectories of cell clusters within the pyramidal lineage, including NEs, cycling RGCs, non-cycling RGCs, cycling IPCs, non-cycling IPCs, and PyNs.

Differential expression analysis was conducted independently for each cell cluster. Only genes with a minimum expression of 0.1 TPM and detected in at least 100 cells were included in the differential expression analysis. Significantly dysregulated genes were identified with p-value < 0.05 and log2 fold change (log2FC) > 1 or log2FC < −1. Gene ontology (GO) analysis was performed using Gene Ontology Resource (https://www.geneontology.org/) with the PANTHER overrepresentation test (Fisher’s exact test with FDR). For redundant/overlapping pathways identified by the GO analysis, only one representative GO entry was shown.

For cell clusters within the pyramidal lineage, the transcriptional regulatory targets of MED13L in each cluster were further analyzed through three steps. Firstly, Med13l-correlated genes were identified as those with significant correlation in expression across all cells from WT mice, based on Spearman correlation analysis (p < 0.05), including both positively and negatively correlated genes. Secondly, for each cell cluster, significantly upregulated or downregulated genes in KO mice compared to WT mice were identified. Finally, the gene lists from Steps 1 and 2 were compared to identify overlapping genes as potential targets of MED13L, including positively correlated & down-regulated genes and negatively correlated & up-regulated genes.

Western blotting and immunoprecipitation

For Western blotting, brain tissues were homogenized in lysis buffer (P0013K, Beyotime Biotechnology, Shanghai, China). Proteins were separated by 6% SDS-PAGE and transferred onto a PVDF membrane. The blots were incubated in blocking buffer (5% non-fat dry milk in TBST) at room temperature for 2 h, followed by overnight incubation with the primary antibody (diluted in 0.05% TBST with 3% BSA). After thorough rinsing with TBST, the blots were incubated with HRP-conjugated secondary antibodies at room temperature for 2 h. Chemiluminescence detection of protein bands was performed using the BeyoECL Plus kit (Beyotime Biotechnology, P0018S). Western blot images were acquired using the UVP ChemStudio System (Analytik Jena GmbH Co. KG, Jena, Germany). The band intensity was analyzed using VisionWorks Software (VisionWorks 8.0 associated with the UVP ChemStudio System.

For co-immunoprecipitation (Co-IP), cortical tissues from E13.5 embryos were homogenized in lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% NP-40, and 1 mM EDTA) containing protease and phosphatase inhibitors (P10008, Beyotime Biotechnology). Lysates (500 μl, approximately 800 μg total protein) were incubated with 10 μg MED13L antibody at 4 °C overnight to allow antibody-protein binding. Meanwhile, 25 μl protein-A magnetic beads (1008D, ThermoFisher) were incubated with 3% BSA at 4 °C overnight to block non-specific binding sites. The BSA-saturated magnetic beads were then added to the lysate and antibody mixture and incubated for 1 h. to capture the antibody-protein complex. Beads were washed with lysis buffer to remove any unbound proteins and contaminants. Finally, the immunoprecipitated proteins were eluted from the beads for further analysis using mass spectrometry or Western blotting. Primary antibodies used were rabbit anti-MED13L (1:1000, custom-developed via Abclonal, E19406; antigen: aa2000-2090), mouse anti-β-actin (1:5000, Proteintech, 66009-l-Ig), rabbit anti-α-Tubulin (1:5000, Proteintech, 11224-l-Ap), rabbit anti-MED12 (1:2000, Abcam, ab70842), mouse anti-Pol II (1:1000, Biolegend, 664906), rabbit anti-MED1 (1:1000, Cell Signaling, 51613S), rabbit anti-TASP1 (1:50, Santa Cruz, sc514677), rabbit anti-IgG (1:250, Invitrogen, 02-6102). Secondary antibodies were goat anti-mouse IgG(H + L)-HRP conjugate (1:5000, Absin, abs20163), goat ant-rabbit IgG(H + L)-HRP conjugate (1:5000, Absin, abs20147).

4D-DIA proteomics analysis

Purified immunoprecipitated proteins were subjected to 4D-DIA LC-MS/MS analysis performed by PTM Bio. (Hangzhou) Co., Ltd. (Zhejiang, China). Briefly, tryptic peptides were dissolved in solvent A (0.1% formic acid, 2% acetonitrile in water) and directly loaded onto a homemade reversed-phase analytical column (25 cm length, 75/100 μm i.d.). Peptides were separated using a gradient from 6% to 24% solvent B (0.1% formic acid in acetonitrile) over 70 min, 24% to 35% in 14 min, and increasing to 80% in 3 min, holding at 80% for the final 3 min, all at a constant flow rate of 450 nL/min on a nanoElute UHPLC system (Bruker Daltonics, Massachusetts, USA). The peptides were ionized using a capillary source and analyzed with a timsTOF Pro (Bruker Daltonics) mass spectrometer. The electrospray voltage applied was 1.60 kV. Precursors and fragments were analyzed using the TOF detector, with an MS/MS scan range from 100 to 1700 m/z. The timsTOF Pro operated in parallel accumulation serial fragmentation (PASEF) mode, selecting precursors with charge states 0 to 5 for fragmentation, and acquiring 10 PASEF-MS/MS scans per cycle. The dynamic exclusion was set to 30 s.

The DIA proteomics data were aligned to the Mus_musculus_10090_SP_20230103.fasta database using DIA-NN (v1.8) software. Data were filtered to pass quality control for both peptide length (typically 7–20 amino acids) and peptide counts (at least 2 peptides per protein). Pearson’s Correlation Coefficient, Principal Component Analysis, and relative standard deviation were applied to evaluate repeatability. To compare the normalized intensity of each protein between different groups, fold change (FC) was calculated and transformed to log₂FC. Significantly different proteins were identified based on two criteria: p < 0.05 and KO/WT Ratio < 0.5. The results were visualized using a volcano plot. Identified proteins underwent Gene Ontology (GO) analysis, with the top 20 MED13L-related pathways presented in a bubble plot.

Proteomics analysis of cortical tissues

Proteomic analysis of mouse motor cortex tissue was carried out by Shanghai Biotree Biotech Co., Ltd. Briefly, tissue samples were collected to perform protein extraction, followed by BCA measurement of protein concentration, acetone precipitation, protein reconstitution, reduction, alkylation, enzymatic hydrolysis, SDC removal, and peptide desalination. Then 200 ng of total peptide were separated using the Evosep One nano UPLC liquid phase system and analyzed using a mass spectrometer (timsTOF Pro2, Bruker Daltonics) with a nanoliter ion source. The mass spectrometer used the DDA PASEF (data dependent acquisition) mode for data acquisition, with a scanning range from 100–1700 m/z. Raw data files were searched using SpectroMine (4.2.230428.5,2329; Biognosys AG) software with the Pulsar search engine, followed by further qualitative and quantitative analyses. Proteomics data have been deposited to ProteomeXchange via PRIDE with identifiers PXD055474 and PXD055844.

Bioinformatics

Gene Ontology (GO) enrichment analysis was performed using Gene Ontology Resource (https://www.geneontology.org/). A Sankey plot of the GO enrichment entries, clustering analysis heatmap, and volcano plots was generated using an online platform for data analysis and visualization (https://www.bioinformatics.com.cn). The bubble plot was created using the online tool Hiplot (https://hiplot.cn/). Transcription factor (TF) enrichment analysis of target genes was performed using the web-based tool ChEA3 (https://maayanlab.cloud/chea3/) as described previously⁴⁴. The ChEA3 background database contains a collection of gene set libraries generated from multiple sources, including TF-gene co-expression from RNA-seq studies, TF-target associations from ChIP-seq experiments, and TF-gene co-occurrence computed from crowd-submitted gene lists. The analysis results were ranked based on Fisher’s Exact Test p-values (Score), where lower p-values indicate higher statistical significance and greater relevance to the transcriptional regulation of the target genes. For graphic presentation, top 15 TFs were shown in histograms. PCA plots were produced using the online tool BioLadder (https://www.bioladder.cn/web/#/pro/cloud). Gene co-expression analysis was performed as previously described³⁹. Transcriptomes from different cortical regions, ranging from 8 PCW to 13 PCW, were obtained from the BrainSpan human developmental transcriptome dataset. Spearman’s correlation coefficient was computed between gene pairs to assess their co-expression levels. The gene co-expression matrix was generated using Hiplot.

Statistics and reproducibility

Data are presented as the mean ± SEM. All statistical analyses were performed using GraphPad Prism 9. The sample size “n” denotes biologically independent samples, with exact values reported in the figure legends. At least three biological replicates were analyzed in each experimental group to ensure reproducibility. Data normality was determined by the Shapiro-Wilk test. Statistical significance was determined using one-tailed or two-tailed Student’s t-test for single comparisons and one-way ANOVA with two-tailed Bonferroni post-hoc tests for multiple comparisons. For GO enrichment analysis, statistical significance was set at FDR < 0.05 (with Benjamini-Hochberg correction). Statistical significance was set at p < 0.05 for all comparisons. Detailed test types (including tail specification for t-tests), all p-values, and group sizes are reported in the figure legends and Supplementary Data 6.

Reporting summary

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

On 31 July 2025, South Korea’s government agreed to buy USD100 billion worth of energy from the United States (U.S.) as part of the tariff negotiations between the two countries. Consequently, a 15% tariff will be levied on most South Korean exports to the U.S. While this agreement may have avoided higher tariffs, the fossil fuel purchase commitment could undermine South Korea’s decarbonization goals, increase energy costs, and exacerbate stranded asset risks amid weakening liquefied natural gas (LNG) demand.

U.S. energy import bill could increase significantly

The Institute for Energy Economics and Financial Analysis (IEEFA) estimates that if South Korea buys approximately USD25 billion to USD33 billion worth of U.S. energy for the next three to four years, its U.S. energy import bill could increase between 1.3 and 1.7 times its current level. As of 2024, South Korea imported around USD19 billion worth of U.S. energy products, including crude oil (USD14.25 billion) and natural gas (USD3.09 billion). 

The sensitivity analysis below assumes a three-year timeframe as the base case, consistent with similar energy agreements with other countries. Scenarios for three-and-a-half-year and four-year durations are also included, as official bilateral announcements regarding the energy imports were still pending as of 7 August 2025. 

 

 

According to IEEFA’s scenario analyses, to meet its commitment of procuring around USD25 billion to USD33 billion worth of U.S. energy annually, South Korea would need to increase the volume and value of its energy purchases by approximately 29% to 72%. Assuming the composition of energy products and 2024 prices remain unchanged, this would translate to an estimated increase in annual U.S. crude oil imports to around 30 million to 39 million tonnes and LNG imports to approximately 7 million to 10 million tonnes, depending on the period.

In 2024, South Korea imported around 5.6 million tonnes of LNG from the U.S., representing 12% of its total annual LNG imports of 46.33 million tonnes. Nearly 29 million tonnes, or 68%, was secured through long-term contracts. 

Increasing LNG purchases is not feasible

According to data from energy service provider Bloomberg New Energy Finance (BNEF), South Korean buyers have already signed long-term contracts with various suppliers worldwide for nearly 36 million tonnes of LNG, including about 6.5 million tonnes from the U.S., set to ramp up between 2026 and 2028. Many of these deals have terms spanning 15 to 20 years. 

These contracted volumes mark a significant increase in fixed-take commitments from South Korean buyers. They now account for nearly 72% of total projected national demand, exceeding the historic range of 65% to 68%. Spot purchases of LNG help balance supply with demand, which can fluctuate significantly across the year due to changes in weather or economic conditions. Given existing contractual commitments and the need to maintain flexibility for spot purchases to meet total demand, there is little room for additional long-term contracts under the U.S.-South Korea tariff deal. 

The figure below shows historic LNG imports by supplying country, both long-term and spot (left), and contracted future long-term supplies through 2040 (right). As the projection indicates, room for additional LNG supplies will not materialize until 2029. Even then, buyers will need to consider future demand constraints, as government decarbonization targets under the 11_^th Basic Plan for Long-Term Electricity Supply and Demand (BPLE), adopted earlier in 2025, are set to reduce the LNG needs beginning in the 2030s.

If South Korea attempts to meet the USD33 billion per year purchase commitment, it will need to source its energy exclusively from the U.S. This would require breaking existing contracts and incurring significant penalties. Furthermore, concentrated sourcing is inadvisable; portfolio and supplier diversification are key for energy and economic security. Even under a U.S.-focused scenario, the projected decline in LNG demand and fluctuations in global oil and gas prices make it unlikely that the value of purchases would reach the agreed target. 

Déjà vu of Free Trade Agreement (FTA) renegotiation in 2017

The new tariff deal and the potential increase in U.S. energy imports, especially LNG, resemble the 2017 renegotiations of the Korea-U.S. Free Trade Agreement (FTA), which prompted a substantial increase in U.S. LNG imports. 

LNG imports from the U.S. spiked 70 times year-on-year after the first U.S. export project, Sabine Pass, began operating in 2016. One of the main reasons for South Korea’s LNG imports from the U.S. included efforts to address the trade deficit, which prompted renegotiations of the Korea-U.S. FTA. The media and parliament frequently mentioned LNG as a potential remedy to alleviate the trade disparity and FTA cancellation risks by the U.S.

At the G20 Summit in 2017, South Korea’s administration announced its ‘Coal and Nuclear-free Economy’ policy, counterintuitively recognizing LNG as a form of ‘green energy’, and making it a key component of the country’s energy transition strategy. In December that year, the Ministry of Trade, Industry, and Energy (MOTIE) increased the 2030 LNG share target in the power mix from 16.9% to 18.8% by 2030. By year’s end, South Korean imports of U.S. LNG had more than doubled year-on-year, coinciding with the start of the Korea-U.S. FTA renegotiation.

In December 2020, the targeted share of LNG in South Korea’s power mix increased to 23.3% by 2030, and U.S. LNG imports hit a historic high in 2021. Since then, the U.S. share of South Korea’s LNG imports has remained level, despite overall LNG demand stagnating and declining. In the first quarter of 2025, LNG demand was down 16% compared to the same quarter in the previous year.

Undermining decarbonization, increasing stranded asset risks

Significantly increasing LNG imports to satisfy the U.S. tariff deal would contradict the South Korean government’s ambitious decarbonization targets, including those under the BPLE, which aim to reduce fossil fuel dependence while increasing renewable energy. 

Incentivizing LNG imports could lock in fossil fuel dependence, leading the country to miss its 2038 target of reducing the share of LNG in the power mix to 10.6% from 28% in 2024. This goal requires the current LNG-fired power generation to more than halve to 74.3 terawatt-hours (TWh) by 2038, which means substantially decreasing LNG imports. Consequently, signing new, long-term LNG supply commitments would be inopportune. 

 

Given the imminent global LNG supply glut and the anticipated lower prices in the international market, it is risky to commit to additional long-term purchases of U.S. LNG amid the current energy landscape. Instead, a strategy that favors more flexible cargoes from physical spot markets offering potentially reduced prices would be advantageous. 

There is a widening gap between LNG import infrastructure and demand in South Korea. A growing number of the country’s mega-scale LNG receiving terminal projects are underutilized, while proposed new projects have been scrapped.

South Korea’s commitment to achieving net-zero carbon emissions by 2050 necessitates recognizing that LNG is a fossil fuel and that its use should decline. Additional imports of U.S. LNG could add further stranded asset risks amid the country’s declining demand and the fuel’s diminishing role in the energy transition.

Related IEEFA commentaries:

On Friday, the People’s Bank of China (PBOC) set the USD/CNY central rate for the trading session ahead at 7.1382 as compared to the previous day’s fix of 7.1345 and 7.1742 Reuters estimate.

OMAN:

Shrouded in fog in the lush mountains of southern Oman, a giant chasm plunges into the landscape, echoing with mysterious sounds that have spawned myths and legends among nearby tribes.

This enormous sinkhole is one of four that dot Dhofar governorate, including one of the world’s biggest: the yawning Kahf Teiq, 211 metres deep and 150 metres wide. At the Tawi Atair sinkhole, tourists potter around on concrete paths and stairways.

Not all of the holes are so welcoming, however. The sheer drop of the Sheeheet pit, a 40-minute drive away along mountain roads, is ringed with slippery mud, prompting the authorities to put up a fence and warning signs.

During AFP’s visit, one tourist slipped and slid perilously close to the edge. Dhofar’s governor, Marwan bin Turki Al-Said, gave assurances in a briefing attended by AFP that safety was a priority at the sinkholes.

Tawi Atair means “Well of Birds” in Dhofar’s regional language, a reference to the avian twittering, distorted by echoes, that reverberates off the rock. It lay unknown to the outside world until 1997, when a team of Slovenian researchers working with Oman’s Sultan Qaboos University brought it to international attention.

Now the sinkholes are marketed as a tourist attraction in Dhofar, whose temperate climate draws many visitors from the Gulf during its punishing summers. Long on the margins of the mainstream tourism circuit, Oman as a whole is increasingly attracting attention from international travellers seeking natural beauty and authenticity.

The country welcomed nearly four million visitors in 2024, with the government aiming to triple that figure by 2040 by focusing on sustainable tourism. Dhofar folklore has it that the sinkholes were created by meteorite strikes, direct hits from outer space that gouged the colossal craters.

But Ali Faraj Al-Kathiri, a geologist based in Dhofar, explains that water seeping into the porous limestone forms an acid that then dissolves it, creating the caverns over a period of thousands of years.

The Oman sinkholes are not to be confused with the “Well of Hell”, the foul-smelling, pitch-black Barhout pit across the border in eastern Yemen that is reputed to be a prison for demons.

Uncertainty about the direction of the U.S. economy again has catapulted the price of gold to near all-time highs. Increasing expectations that the Federal Reserve will cut interest rates in September may keep it there.

Spot gold reached a high of $3,418.14 per troy ounce Thursday, within striking distance of its June 13 all-time high of $3,448.50. The price of gold has gained more than 3% since hitting a one-month low of $3,311.80 a week ago, just before the release of employment data that showed the U.S. labor market to be far weaker than previously thought.

Seeking Safety

The latest rally falls in line with gold’s reputation as a safe haven for investors in times of economic uncertainty.

The jobs report last Friday showed that employers hired fewer workers in July than economists had estimated, while the unemployment rate ticked higher to 4.2%. Even more worrisome, employment numbers for the previous two months were revised dramatically lower.

Weakening labor market conditions could portend lower economic growth, something investors have been worried about amid uncertainty about the impact that tariffs will have. Concerns about the economic outlook have helped fuel gold’s 30% price rise year-to-date.

Fed Rate-Cut Expectations Rise

The weak jobs numbers have boosted market expectations that the Fed’s policy committee will cut the benchmark fed funds rate when it meets in September. After trimming the rate a full percentage point in late 2024, the Fed has refrained from cutting rates this year, with officials saying they need more data showing how tariffs affect inflation before adjusting policy. (The Fed has a dual mandate to promote high levels of employment and to maintain price stability.)

While the Fed has stood pat on rates, the European Central Bank has cut interest rates eight times since June 2024. The ECB rate cuts have bolstered the value of gold globally. Because gold does not offer a regular yield payment to investors, it tends to perform better when competing investments, such as bonds, offer lower interest payments.

That’s why Fed rate cuts, were they to occur, could further underpin demand for gold.

Prior to the jobs report on Aug. 1, just 37% of investors expected a September rate cut, based on the fed funds futures market. Now, more than 90% expect a quarter-point cut in the Fed’s benchmark target rate to 4%-4.25%, as well as additional cuts before the end of 2025.

Getty Images

MODerna

Covisian

Bunq