Table of Contents
All animal experiments were conducted following protocols approved by the Administrative Panel on Laboratory Animal Care at Stanford University. The TRAP2:Ai14 mouse line was a gift from the Luo laboratory at Stanford. TRAP210 mice were heterozygous for the Fos2A-icreER allele, and homozygous for Ai14 in the C57BL/6 background. Gt(ROSA)26Sortm1.1(CAG-cas9*,-EGFP)Fezh/J mice were acquired from Jackson Laboratory. Mice were group-housed (maximum 5 mice per cage) on a 12 h light:dark cycle (07:00 to 19:00, light) with food and water freely available. Mice were kept with ambient temperature at 21.1 ± 1.1 °C and humidity at 55 ± 5%. Male mice 49–56 days of age were used for all the experiments. Mice were handled daily for 3 days before their first behavioural experiment. The animal protocol no. 20787 was approved by Stanford University APLAC and IACUC. All surgeries were performed under avertin anaesthesia and carprofen analgesia, and every effort was made to minimize suffering, pain and distress.
The following primers: TCCTGGGCATTGCCTACAAC (forward), CTTCACTCTGATTCTGGCAATTTCG (reverse) and ACCCTGCTGCGCATTG (reporter) were used for genotyping of the Fos2A-icreER allele; CTGAGCTCACCCACGCT (forward), GGCTGCCTTGCCTTCTCT (reverse), ACTGCTCACAGGGCCAG (reporter) for wild-type allele; CGGCATGGACGAGCTGTA (forward), CAGGGCCGGCCTTGTA (reverse) and AATTGTGTTGCACTTAACG (reporter) were used for genotyping of the Rosa-Ai14 allele; TTCCCTCGTGATCTGCAACTC (forward), CTTTAAGCCTGCCCAGAAGACT (reverse) and CCGCCCATCTTCTAGAAAG (reporter) for Rosa wild-type allele.
The fear conditioning training was conducted according to previously described methods9. Each mouse was placed individually in the fear conditioning chamber (Coulbourn Instruments), which was positioned at the centre of a sound-attenuating cubicle. Prior to each session, the chamber was cleaned with 10% ethanol to provide a background odour, while a ventilation fan generated background noise at around 55 dB. The training began with a 2-min exploration period, after which the mice received three tone-foot shock pairings separated by 1-min intervals. Each tone, an 85 dB 2-kHz sound, lasted for 30 s, and was followed by a 2-s foot shock of 0.75 mA, with both ending simultaneously. Following each pairing, the mice remained in the chamber for an additional 60 s before being returned to their home cages. For context recall, the mice were reintroduced to the original conditioning chamber for 5 min, 16 days after the training. Injections of 4-hydroxytamoxifen injections were administered immediately prior to the recall experiments, within 30 min. In the HC and NR groups, 4-hydroxytamoxifen was injected at a similar time to the other two groups during the recall. The behaviour of the mice was recorded and analysed using FreezeFrame software (version 4; Coulbourn Instruments), with motionless bouts lasting over 1 s being considered as freezing.
Brain tissue dissociation and flow cytometry
Basolateral amygdala was microdissected from live sections cut by a vibratome (300 μm thick). Tissue pieces were enzymatically dissociated using a papain-based digestion system (LK003150, Worthington). In brief, tissue chunks were incubated with papain (containing L-cysteine), DNase I, and kynurenic acid for 1 h at 37 °C and 5% CO2. After incubation, tissues were triturated with 300 µm glass pipette tips, then 200 µm glass pipette tips, and 100 µm glass pipette tips. Cell suspensions were then centrifuged at 350g for 10 min at room temperature, resuspended in 1 ml EBSS with 10% v/v ovomucoid inhibitor, 4.5% v/v Dnase I, and 0.1% v/v kynurenic acid, and centrifuged again. The supernatant was removed, and 1 ml artificial cerebrospinal fluid (ACSF) was added to the cells. ACSF contained 2.5 mM KCl, 7 mM MgCl2, 0.5 mM CaCl2, 1.3 mM NaH2PO4, 110 mM choline chloride, 24 mM NaHCO3, 1.3 mM sodium ascorbate, 20 mM glucose, and 3 mM sodium pyruvate, 2 mM thiourea, and 13.2 mM trehalose. Cells were then passed through a 70 μm cell strainer to remove debris. Hoechst stain (1:2,000; H3570, Life Technologies) was added and incubated in the dark at 4 °C for 10 min. Samples were centrifuged (350g for 8 min at 4 °C) and resuspended in 0.5 ml of ACSF and kept on ice for flow cytometry. Live cells were sorted using the BD Vulcan into 384-well plates (Bio-Rad) directly into lysis buffer, oligodT, and layered with mineral oil. After sorting, the plates were immediately snap frozen until reverse transcription.
The Smartseq3 protocol was used for whole-cell lysis, first-strand synthesis and cDNA synthesis, as previously described with modifications. Following cDNA amplification (23 cycles), the concentration of cDNA was determined via Pico Green quantitation assay (384-well format) and normalized to 0.4 ng µl−1 using the TPP Labtech Mosquito HTS and Mantis (Formulatrix) robotic platforms. In-house Tn5 were used for cDNA tagmentation. Libraries were amplified using Kapa HiFi. The libraries were then sequenced on a Novaseq (illumina), using 2 × 100-bp paired-end reads and 2 × 12-bp index reads, with an average of 2 million reads per cell.
Bioinformatics and data analysis for scRNA-seq
Sequences from Nextseq or Novaseq were demultiplexed using bcl2fastq, and reads were aligned to the mm10 genome augmented with ERCC (External RNA Controls Consortium) sequences, using STARsolo 2.7.9a. We applied standard algorithms for cell filtration, feature selection and dimensionality reduction. In brief, genes that appeared in fewer than five cells, samples with fewer than 2,000 genes and samples with less than 50,000 reads were excluded from the analysis. Out of these cells, those with more than 10% of reads as ERCC or more than 20% mitochondrial were also excluded from analysis. Counts were log-normalized and then scaled where appropriate. Canonical correlation analysis (CCA) function from the Seurat70 package was used to align raw data from multiple experiments. The top 20 canonical components were used. After alignment, relevant features were selected by filtering expressed genes to a set of 2,000 with the highest positive and negative pairwise correlations. Genes were then projected into principal component space using the robust principal component analysis. DEG analysis was done by applying the Mann–Whitney–Wilcoxon test on various cell populations.
To find memory-induced genes in each type of neurons, series of strict criteria were applied. First, we removed the background activation by excluding the DEGs resulted from FR versus NF among tdT negative neurons. This guarantees their specificity that DEGs are activity-dependent, rather than a general increase in all cells caused by experience. Second, DEGs must be differentially expressed when FR TRAPed cells are compared to NR and HC controls, ensuring that the DEGs were unique to neuronal ensembles associated with memory recall, and not a result of baseline activity (HC) or activity remaining from the initial fear learning (NR). Finally, each DEG had to meet the criteria of being expressed in a quater of cells and exhibiting at least a 1.75-fold change. By adhering to these standards, a total of 107 DEGs were recognized as ‘remote-memory-associated DEGs’ across 6 distinct neuron types, BLA.Int.Pvalb was not included in the analysis due to insufficient numbers of cells. EnrichR was used for GO, KEGG and REACTOME pathway analysis and classification of enriched genes (log2FC > 0.5 and P < 0.05) in each subpopulation.
scRNA-seq data from mPFC cells were mapped to mm10 genome with full-length tdTomato construct (including Woodchuck Hepatitis Virus Posttranscriptional sequence included in Ai14 line71), which improved the sensitivity in calling tdT+ cells. Data from BLA and mPFC cells were integrated using CCA. TRAPed neurons from the each integrated population were analysed, except B-P.Int.Pvalb and B-P.Int.Gpr88 neurons, due to limited cell number. DEGs with P < 0.05 (Mann–Whitney–Wilcoxon test) were considered as significant DEGs (highlighted in orange in Fig. 5d and Extended Data Fig. 12f).
After unbiased clustering astrocytes, RNA velocyto40 and Monocle341 were applied to infer astrocytic trajectory. DEGs between FR and NF conditions were estimated using Mann–Whitney–Wilcoxon test on each astrocyte population. R, RStudio, Python were used for data analysis.
The RNAscope multiplex fluorescent reagent kit v2 (323100, ACD) and RNAscope 4-Plex probes were used to conduct the RNAscope experiment according to the manufacturer’s guidelines. The probes employed were either obtained from available stocks or specially created by ACD.
Gene selection for MERFISH measurements
We used a combination of single-cell RNA sequencing data and literature to select genes for MERFISH. Our selection criteria involved identifying cell-type-marker genes for a particular cell population using a one-vs-all approach. To do this, we performed a Mann–Whitney–Wilcoxon test for each gene between the cells within the cell population and all other cells not in that population, and corrected the resulting P values for multiple hypothesis testing to obtain false discovery rate-adjusted P values. A gene was considered a cell-type marker for a specific cell population if it met the following criteria: (1) it was expressed in at least 30% of cells within the specified population; (2) the false discovery rate-adjusted P value < 0.001; (3) gene expression in the specified population was at least fourfold higher than the average expression in all cells not in that population; and (4) expressed in a fraction of cells within the specified population that was at least 2 times higher than any other population of cells. We then sorted the marker genes for each population by fold change in expression relative to cells outside the population, and saved the top five marker genes for each population to use for marker selection. In addition to these markers, known genes related to microglia, astrocytes and OPCs from the literature and included. Finally, DEGs from remote memory-associated genes were added to the panel with a total number of 158 genes.
Tissue processing for MERFISH and RNAscope
Brain tissue samples were processed using a fixed-frozen protocol for both MERFISH and RNAscope. In brief, mice were euthanized using CO2 and perfused with cold 4% paraformaldehyde. Brain tissue was dissected and followed by incubation at 4 °C in 4% paraformaldehyde overnight, 15% sucrose for 12 h, and 30% sucrose until sink. Brain tissue was frozen in OCT using dry ice and stored at −80 °C until sectioning. Sectioning was performed on a cryostat at −18 °C. Slices were removed and discarded until BLA region was reached.
Slices with 10 μm in thickness were captured onto Superfrost slides for RNAscope and MERSCOPE slides for MERFISH. The same anatomical region was identified for imaging post hoc after sample preparation, before the start of RNAscope or MERFISH imaging.
Sample preparation and MERFISH imaging
Slides with tissue sections were processed according to MERSCOPE protocol (Vizgen). In brief, slides with tissue sections were washed three times in PBS, and then stored in 70% ethanol at 4 °C for 18 h to permeabilize the tissue. Tissue slices from the same mouse were cut at the same time and distributed onto four coverslips. After permeabilization, the samples were removed from 70% ethanol and washed with Sample Prep Wash Buffer (PN 20300001), then incubated with Formamide Wash Buffer (PN 20300002) at 37 °C for 30 min. Gene Panel Mix (RNA probes) was incubated for 48 h at 37 °C. After hybridization, the samples were washed in Formamide Wash Buffer for 30 min at 47 °C for a total of 2 times to remove excess encoding probes and polyA-anchor probes. Tissue samples were then cleared to remove lipids and proteins that contribute fluorescence background. In brief, the samples were embedded in a thin 4% polyacrylamide gel and were then treated with Clearing Premix (PN 20300003) for 36 h at 37 °C. After digestion, the coverslips were washed in Sample Prep Wash Buffer 2 times and stain with DAPI/PolyT mix for 15 min. Slides were washed with Formamide Wash Buffer followed by Sample Prep Wash Buffer before imaging. Finally, slides were loaded to MERSCOPE Flow Chamber and imaged at both 20× and 63× magnification.
MERFISH data processing
MERFISH imaging data were processed with MERlin72 pipeline with cell segmentation using CellPose73, a deep learning-based cell segmentation algorithm based on DAPI staining. Decoding molecules were then assigned to the segmented nuclei to produce a cell-by-gene matrix that list the number of molecules decoded for each gene in each cell. The MERFISH expression matrix for each sample was concatenated with the normalized, log-transformed with Scanpy74 and integrated using Harmony75 and Leiden76 clustering was applied to separate the cells into distinct clusters. TRAPed neurons were assigned based on tdTomato expression. DEGs from a comparison of FR-TRAPed and NF-TRAPed conditions were estimated using Mann–Whitney–Wilcoxon test. Peri-engram cells were computed as follows: for each engram cell (tdT+), its peri-engram cells were counted within a radius of 30 μm.
CalEx injection and behavioural experiments
AAVs carrying CalEx51 or tdTomato were generated by Addgene based on the vector pZac2.1-GfaABC1D-mCherry-hPMCA2w/b (AAV5, Addgene 111568) or pZac2.1 gfaABC1D-tdTomato (AAV5, Addgene 44332). Stereotaxic procedure of viral microinjection has been described previously. In brief, mice with fear training (within 12 h or after 24 h) were anaesthetized and placed onto a stereotaxic frame (model 1900, KOPF). Mice were injected with Carprofen (5 mg kg−1) subcutaneously before and after surgery. AAVs carrying hPMCA2w/b (CalEx) or control (tdTomato) viruses were loaded via a glass pipette connected with a 10 μl Hamilton syringe (Hamilton, 80308) on a syringe injection pump (WPI, SP101I) Bevelled glass pipettes (1B100–4; World Precision Instruments) filled with viruses were placed into the BLA (1.3 mm posterior to the bregma, 3.4 mm lateral and to the midline, and 4.6 mm from the pial surface). Either 0.3 μl of AAV5 GfaABC1D mCherry-hPMCA2w/b (7 × 1012 viral genomes (vg) per ml) or 0.3 μl AAV5 GfaABC1D tdTomato (7 × 1012 vg ml−1) were injected at 100 nl min−1. Glass pipettes were withdrawn after 10 min and scalps were cleaned and sutured with sterile surgical sutures. Mice were allowed to recover in clean cages for 7 days. behavioural experiments (recall) were performed three weeks after surgeries. Schematic illustrations (Figs. 1a and 4a,f and Extended Data Fig. 7h,o) created with BioRender.com.
Mice were placed in the centre of 40 × 40 cm white box and allowed to freely explore for 15 min. Videos were recorded and analysed by BIOBSERVE III software. The 20 × 20 cm region in the centre was defined as the central zone. The total distance travelled and the activity exploring the centre area were analysed to evaluate the subject’s locomotor ability and anxiety levels.
Oligos and antibodies
For quantitative PCR analysis, specific primers were designed to amplify the Igfbp2 gene: Igfbp2 FW (GTCTACATCCCGCGCTG) and Igfbp2 RV (GTCTCTTTTCACAGGTACCCG). Additionally, for CRISPR–Cas9 gene editing, six gRNAs (Igfbp2 guides 1–6) were selected to target distinct regions of the Igfbp2 gene. These gRNAs were designed based on predicted specificity and efficiency: Igfbp2 guide 1 (CTACGCTGCTATCCCAACCC), Igfbp2 guide 2 (GCCAGACGCTCGGGCGTGCA), Igfbp2 guide 3 (AGAAGGTCAATGAACAGCAC), Igfbp2 guide 4 (GCCCTCCTGCCGTGCGCACA), Igfbp2 guide 5 (CTCTCGCACCAGCTCGGCGC), and Igfbp2 guide 6 (CGTAGCGTCTGGGCGCAGCG).
Antibodies targeting mCherry (Thermo Fisher M11217) and cFOS (Synaptic Systems 226308) were applied for immunostaining following manufacturers’ manuals.
Inclusion and ethics statement
We, the authors of this manuscript, recognize the importance of inclusion and ethical considerations in scientific research. Our work is guided by the principles of fairness, transparency, and respect for human dignity.
We affirm our commitment to promoting diversity and inclusivity in science, recognizing that diverse perspectives, backgrounds, and experiences enrich research and enhance scientific discovery. We have made efforts to ensure that our study is conducted in a manner that respects and includes individuals of all races, ethnicities, genders, sexual orientations, abilities, and other aspects of human diversity.
We have obtained all necessary ethical approvals and have followed appropriate guidelines and regulations for the research conducted. We have taken measures to protect the privacy and confidentiality of research participants, including obtaining informed consent and ensuring data security.
We acknowledge the potential for harm in scientific research and have taken steps to minimize any potential harm to research participants or others affected by our work. We have carefully considered the potential implications of our research and have taken responsibility for ensuring that our work is conducted in a manner that upholds ethical and moral standards.
We recognize that scientific research has the potential to impact society in profound ways and we are committed to engaging in responsible research practices that promote the well-being of individuals and society as a whole.
In summary, we affirm our commitment to inclusive and ethical research practices and recognize our responsibility to conduct research that is conducted with integrity, respect, and social responsibility.
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.