Single-nucleus Atlas of Sevoflurane-induced Hippocampal Cell Type– and Sex-specific Effects during Development in Mice

The animal study protocol was approved by the Ethics Committee for Animal Care and Use of Soochow University, Suzhou, China (approval No.: 202007A098). C57BL/6J mice (8 weeks old, 20 to 25 g, drug- or test-naive) were purchased from the Slaccas Laboratory (China). We report this study according to the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines.

In this study, we conducted experiments on male and female offspring mice separately. All mice were housed in a specific pathogen-free room (22 to 24°C) with a 12-h light/dark cycle. The offspring mice were sequentially numbered with ear tags. Using an online randomization tool (https://www.randomizer.org/), we randomly allocated the mice to the control or sevoflurane group on postnatal days 6, 8, and 10. Mice received 60% oxygen balanced with nitrogen (control treatment) or 3% sevoflurane in 60% oxygen balanced with nitrogen (sevoflurane treatment) for 2 h from 9:00 to 11:00 am. This animal model was described previously.⁶  The gas concentration was monitored using a gas analyzer (Vamos; Dräger Medical, Germany). The rectal temperature of every mouse was maintained at 36.5 to 37.5°C by adjusting the heating pad temperature.⁵  The 3% sevoflurane is a clinically relevant concentration.¹²  This model did not induce adverse changes in the blood gas or electrolytes of mice.¹¹  Mice were treated and assessed in the sequential order. At 9:00 to 11:00 am on postnatal day 37, mice were decapitated under brief sevoflurane anesthesia (3% sevoflurane for 3 min) for harvesting hippocampus samples.

The allocation details were stored in an opaque sealed envelope. The investigators who perform the behavior tests, single-nucleus RNA sequencing, fluorescence in situ hybridization, real-time quantitative polymerase chain reaction, and Western blot were blinded to the group allocation until the completion of final analysis.

The Morris water maze test was performed to confirm the successful establishment of cognitive impairment model in mice.¹¹  To evaluate spatial learning ability, the mice were trained to reach the platform sequentially for 5 consecutive days in the training phase (9:00 to 11:00 am on postnatal days 31 to 35), and the escape latency was recorded. To assess memory function, the platform was removed in the testing phase (9:00 to 11:00 am on postnatal day 36), and the mice were sequentially assessed in 60 s in the Morris water maze. The platform-crossing times, time spent in the platform quadrant, and mean distance from the original platform area were recorded. All behavioral data in the Morris water maze were recorded automatically on a video-tracking system with the ANY-maze software (Stoelting Co., USA).

The hippocampi were rapidly harvested and dissected into 1- to 2-mm³ tissue blocks in ice-cold phosphate buffered saline at 9:00 to 11:00 am on postnatal day 37. The hippocampal tissues were ground into a nuclei suspension with 500 ml lysis buffer. After resuspension, centrifugation, and filtration, the hippocampal nuclei were extracted. The nuclei were counted using Trypan blue staining.

Barcoded single nucleus, Gel Bead-In-Emulsions, and cDNA libraries were generated and processed following the manufacturer’s protocol. Data were analyzed using the 10 × Genomics GemCode Technology (10x Genomics, USA) and OmicShare tools. Every transcript in a single cell was uniquely barcoded using a unique molecular identifier. The single-nucleus RNA sequencing analysis was performed on the Illumina NovaSeq 6000 system (Illumina, USA). The raw sequence data are deposited in the Sequence Read Achieve of the National Center of Biotechnology Information (Bethesda, Maryland; https://www.ncbi.nlm.nih.gov/sra) under accession Nos. PRJNA797582 and PRJNA873858.

Cell Ranger software (version 6.1) was utilized for quality control. We analyzed gene counts, unique molecular identifier counts, and percentage of mitochondrial genes, and then we aligned the reads to the musculus transcriptome (Ensembl_release98) using the R package Seurat.¹³  The criteria for effective cells were as follows: gene counts 500 to 4,000 per cell, unique molecular identifier counts less than 8,000 per cell, and the percentage of mitochondrial genes less than 10%. Low-quality cells were filtered out. We normalized the filtered gene expression using Seurat’s NormalizeData function, as follows: a gene expression level = $log(1+unique molecular identifierAunique molecular identifierTOTAL×10000)$⁠.

After data preprocessing, we used the canonical correlation analysis to normalize and filter the gene matrix. For principal component analysis and dimension reduction, we selected genes with average expression values of 0.01 to 3 and dispersion greater than 0.5.¹⁴  The principal components were applied for t-distributed stochastic neighbor embedding on two dimensions using the RunTSNE function in the Seurat.¹³  Cluster- or subcluster-specific marker genes were defined from our 10x sequencing dataset using a likelihood-ratio test, based on the following criteria: P value ≤0.01, log₂(fold change) 0.361 or greater, and the percentage of cells where the gene is detected in the specific cluster or subcluster 25% or greater (mincell_pct 0.25 or greater).¹⁵  To resolve the identities of the clusters, canonical cell type–specific markers for major hippocampal neurons and nonneuron cell types were retrieved from previous studies.^16–19  Major hippocampal cell clusters were identified by mapping canonical marker genes on the two-dimensional t-distributed stochastic neighbor embedding map.

In addition, we analyzed functional cell subcluster (such as dentate gyrus granule cells, microglia, and cornu ammonis 1 pyramidal neurons) to discover compositional and transcriptional differences for each subcluster between the control and sevoflurane groups. We performed principal component analysis to identify significant principal components for subsequent clustering and t-distributed stochastic neighbor embedding visualization. We identified subclusters using the Seurat’s FindClusters function. We utilized the Seurat’s FindAllMarkers function to confirm differentially expressed genes and specific marker genes for each subcluster.

We used the CellPhoneDB software (version 2.0; https://www.cellphonedb.org)²⁰  to analyze the expression abundance of ligand–receptor interactions among hippocampal cell types. To be included in the cross-talk analysis, the ligands or receptors should be expressed by 30% or greater cells in a specific cluster (mincell_pct 0.30 or greater). We used the permutation test by randomly distributing the cluster labels of all cells 1,000 times and analyzed the average expression of each ligand–receptor pair between two clusters. We defined P value <0.05 as significant ligand–receptor enrichment. To illustrate the regulation relationship between cell types, we constructed a cell–cell interaction network diagram, with boxes representing the interactions from ligands (source cells) to receptors (target cells).

We constructed single-cell development and differentiation trajectory in the pseudotime analysis using the Monocle 2 package (version 2.6.4; Monocle, USA).²¹  We used the normalized expression cells × genes matrix to create a starting subject CellDataSet class.²²  We used the differentialGeneTest function to identify specific genes across multiple cell subtypes and applied the orderCells function for cell ordering in pseudotime. We implemented the DDRTree approach in the reduceDimension function for dimension reduction. We applied the plot_cell_trajectory function to map the trajectory in a two-dimensional space, with the two dimensions together presenting the cell differentiation state. The branches in the single-cell trajectories indicate different biologic functions depending on the gene expression during development. We identified key genes in different cell states using the cutoff value of false discovery rate less than 1e-5.¹⁵ 

We identified the differentially expressed genes within each cell type between the control and sevoflurane groups using the FindMarkers function in the Seurat. The differentially expressed genes were uncovered using the Model-based Analysis of Single-cell Transcriptomics,²³  based on the following criteria: |log₂(fold change)| 0.585 or greater, q value 0.05 or less, and the percentage of cells in the specific cluster greater than 25%.²⁴  We generated the heatmaps of differentially expressed genes using the z-score approach.

In addition, we performed the Gene Ontology enrichment analysis to provide all Gene Ontology terms (Cellular Component, Molecular Function, and Biologic Process) with significant enrichment of peak related differentially expressed genes in each cell type.²⁵  All peak related genes were mapped to Gene Ontology terms in the database (https://www.geneontology.org/). We also performed enrichment analysis of pathways based on the Kyoto Encyclopedia of Genes and Genomes database (https://www.genome.jp/kegg/) to identify significantly enriched metabolic pathways or signal transduction pathways with peak related differentially expressed genes in each cell type.²⁶  For the Gene Ontology and Kyoto Encyclopedia of Genes and Genomes enrichment analyses, the P value is calculated as follows: $p=1−∑i=0m−1(Mi)(N−Mn−i)(Nn)$⁠, in which N is the number of all genes with Gene Ontology/Kyoto Encyclopedia of Genes and Genomes annotation, n is the number of differentially expressed genes in N, M is the number of all genes annotated to the certain Gene Ontology terms or specific pathways, and m is the number of differentially expressed genes in M. The calculated P values were corrected using false discovery rate, with false discovery rate 0.05 or less as a threshold to determine a significantly enriched Gene Ontology terms or Kyoto Encyclopedia of Genes and Genomes pathways.

The fresh hippocampal samples were fixed in 4% paraformaldehyde for 2 h at 4°C. The samples were then embedded in paraffin and sectioned into 8-μm slices at room temperature. The sections were dehydrated with sequential washes with 50%, 75%, and 100% ethanol in phosphate buffered saline for 5 min each. Rehydration was performed with sequential washes with 100%, 75%, and 50% ethanol. The sections were permeated for 2 h in 1% Triton-X and washed three times with 0.1% Tween in phosphate buffered saline for 5 min at room temperature. After prehybridization with probe hybridization buffer for 2 h at 37°C, the sections were hybridized overnight at 37°C with the probes (RiboBio, China) that bound to the following target RNA: Dcc (lnc102941), Grid1 (lnc102942), Shank2 (lnc102939), Ctnnd2 (lnc102943), Pcdh7 (lnc102937), and Sobp (lnc102938). Next, the sections were washed twice with 2 × standard saline citrate solution for 5 min each, 0.5 × standard saline citrate solution for 15 min, and 0.2 × standard saline citrate solution for 15 min at 37°C. The cell nuclei were counterstained with 4’,6-diamidino-2-phenylindole. The sections were treated with strept actividin-biotin complex -fluorescein isothiocyanate for 30 min at 37°C. Fluorescent images were obtained at 400 × and 100 × magnification (Leica Aperio VERSA; Germany) using the Axiovision software (Carl Zeiss Vision, version 4.6). The results were analyzed with the fluorescence in situ hybridization quantification assay using the Image-Pro Plus software (version 6.0, Media Cybernetics, Inc., USA). The mean fluorescence intensity was calculated as integrated optical density/area (unit: pixel).

Total RNA was extracted from mice hippocampi using TRIZOL (ambion, USA). After RNA quantification and evaluation of RNA purity, cDNA was generated from RNA using a reverse transcription kit (TransGen Biotech, China). We performed the real-time quantitative polymerase chain reaction using the SYBR Green MasterMix (TransGen Biotech) in a reaction volume of 20 μl on the 7,500 Fast Real-Time Polymerase Chain Reaction System (Applied Biosystems, USA). The specific primers were designed and manufactured by Sangon Biotech (China) targeting Dcc, Grid1, Shank2, Ctnnd2, Pcdh7, and Sobp. The GAPDH was used as a reference gene. We recorded the cycle threshold value of each gene, and the gene abundance was calculated using the 2 ^{−ΔΔ (cycle threshold)} method. The specific primers are as follows:

Mbp-forward, 5’-CCCACTTGATCCGCCTCTTT-3’

Mbp-reverse, 5’-CTCGGAGCTCACCTTGCC-3’

Mag-forward, 5’-GATGCCCTCGACCATCTCAG-3’

Mag-reverse, 5’-CGAACTGCAAGGTGGTGTTG-3’

Mog-forward, 5’-GGTATCCCATCCGGGCTTTAG-3’

Mog-reverse, 5’-AGGTGCTTGCTCTGCATCTT-3’

Mal-forward, 5’-GTTTGTGAGTTTGATGCAGCCT-3’

Mal-reverse, 5’-GCTTCCAGAACTGAGGCACT-3’

GAPDH-forward, 5’-GGTTGTCTCCTGCGACTTCA-3’

GAPDH-reverse, 5’-TGGTCCAGGGTTTCTTACTCC-3’

Total protein of the hippocampus was extracted using the MT-cellytics and PIC 100X reagents (Biocolor BioScience, China). The proteins were separated using electrophoresis with 10% sodium dodecyl sulfate-polyacrylamide gel and transferred onto polyvinylidene fluoride membranes (Millipore Corp., USA). The membranes were blocked for 2 h at room temperature. Next, the membranes were incubated with the primary antibodies at 4°C overnight: anti-glutamate ionotropic receptor delta type subunit 1 (1:50; 13040-1-AP; Proteintech) for glutamate ionotropic receptor delta type subunit 1 (100 to 112 kd), anti-delta-catenin (1:1,000; ab184917; abcam) for delta-catenin (135 kd) encoded by Ctnnd2, anti-sine oculis binding protein homolog (1:1,000; OAAB00857; Aviva Systems Biology) for sine oculis binding protein homolog (92 kd), and anti-glyceraldehyde-3-phosphate dehydrogenase (1:2000; AF5009; Beyotime) for glyceraldehyde-3-phosphate dehydrogenase (37 kd). The membranes were incubated with horseradish peroxidase–conjugated secondary antibodies for 2 h at room temperature. Glyceraldehyde-3-phosphate dehydrogenase was used as the loading control. The bands were visualized on the Tanon 5,200 system (Tanon, China).

No statistical power calculation was conducted before the study. We determined the sample size based on a recent study (10 mice per group for behavioral tests, four to six mice per group for Western blot and real-time quantitative polymerase chain reaction, and three mice per group for immunostaining experiments).⁵ 

For the single-nucleus RNA sequencing data, we used R v3.5.3 (Foundation for Statistical Computing) for the statistical analyses and graphic generation. Bioinformatic tools were used to analyze cellular clustering, cross-talk, single-cell trajectory, differentially expressed genes, and enrichment. For the experimental data, we used GraphPad Prism software (version 9.0, USA) to perform statistical analyses and graphic plotting. We checked normal distribution of data using the Shapiro–Wilk test. The Morris water maze data are expressed as mean ± SD. We used Student’s t test to analyze the between-group differences in platform-crossing times, time spent in the fourth quadrant, and mean distance from the platform. We used two-way repeated-measures ANOVA with Bonferroni correction to compare the two groups on escape latency, body weight, and body length. The expression of key genes in the trajectory analyses of dentate gyrus cell neurogenesis is expressed as mean ± SD (arbitrary units). We applied mixed-effect ANOVA (restricted maximum likelihood) with Geisser–Greenhouse correction to analyze the interaction between group and pseudotime, and multiple comparisons were corrected using the Bonferroni method. For the 24 common differentially expressed genes in male mice, we used Student’s t test to analyze the between-group differences in their expression in specific hippocampal cell types. The data from fluorescence in situ hybridization quantification, real-time quantitative polymerase chain reaction, and Western blot are expressed as mean ± SD (arbitrary units), and we compared the two groups using Student’s t test, mean difference, and 95% CI. There were no missing data for variables in this study. A P value <0.05 was considered statistically significant. All statistical tests were two-tailed.

Mice underwent control or sevoflurane treatment at postnatal days 6, 8, and 10, and the behavior tests showed that multiple neonatal exposures to sevoflurane resulted in significant learning and memory impairment in young mice of both sexes (Supplemental Digital Content, fig. 1, https://links.lww.com/ALN/D49). All mice survived until postnatal day 37, and then we performed data preprocessing and quality control (Supplemental Digital Content, fig. 2, https://links.lww.com/ALN/D49) and conducted single-nucleus RNA sequencing of the hippocampus.

For the identification of cell clusters, we compared our cluster-specific gene signatures of male mice (Supplemental Data 1, https://links.lww.com/ALN/D50) and female mice (Supplemental Data 2, https://links.lww.com/ALN/D51) to known markers of hippocampal cell types (Supplemental Digital Content, fig. 3, https://links.lww.com/ALN/D49).^16–19  The common markers for both sexes included Slc1a3 (astrocytes), Pex5l (cornu ammonis 1 neurons), Cpne4 (cornu ammonis 3 neurons), Prox1 (dentate gyrus neurons), Bsg (endothelial cells), Gad1 (γ-aminobutyric acid–mediated [GABAergic] neurons), C1qa (microglia), MBP (oligodendrocytes), and Pdgfra (oligodendrocyte precursor cells). Notably, our results revealed sex-specific cell type markers, including Atp1a2 (astrocytes), Tshz2 (cornu ammonis subtype neurons), Reln and Nxph1 (cholinergic neurons), and Acta2 (mural) for male mice, and Rxfp1 (cornu ammonis subtype 1 neurons) and Sulf1 (cornu ammonis subtype 2 neurons) for female mice.

For the male mice, we detected 15 hippocampal cell clusters including 23,806 hippocampal cell nuclei from the control male mice and 13,190 from the sevoflurane male mice (fig. 1A; Supplemental Data 1, https://links.lww.com/ALN/D50). For the female mice, we detected 12 hippocampal cell clusters including 25,402 hippocampal cell nuclei from the control female mice and 23,077 from the sevoflurane female mice (fig. 1B; Supplemental Data 2, https://links.lww.com/ALN/D51).

In male mice, the previously known cell markers exhibited cluster-specific expression, which validated our identification of specific hippocampal cell clusters (fig. 1C). After multiple sevoflurane exposures, the proportion of cornu ammonis 1 neurons was decreased from 79.9% to 32.3%, the proportion of dentate gyrus neurons was increased from 4.2% to 23.4%, and the proportion of oligodendrocytes was increased from 0.6% to 6.9% (fig. 1D). In contrast, no such changes in the proportion of cornu ammonis 1 neurons (from 27.3% to 24.3%), dentate gyrus neurons (from 36.2% to 35.8%), and oligodendrocytes (from 5.9% to 7.8%) were found in females (fig. 1E). For the female mice, the cell type–specific clusters were also validated by known cell markers (fig. 1F).

Our investigations also identified novel sex-specific marker genes for different hippocampal cell types. We showcased the top five highly expressed genes in male mice (Supplemental Digital Content, fig. 4A, https://links.lww.com/ALN/D49). For these genes, we examined cellular identity and visualized hippocampal localization by using the Allen Brain Atlas in situ hybridization images (https://mouse.brain-map.org)¹⁸  (Supplemental Digital Content, figs. 5 to 9, https://links.lww.com/ALN/D49). Based on the expression of these top genes, we identified several novel marker genes in the male mice (Gli2 for astrocytes, Gm49906 for cornu ammonis subtype neurons, Gm2629 for microglia, and 4930420G21Rik for oligodendrocytes; Supplemental Digital Content, fig. 4B, https://links.lww.com/ALN/D49). For the female mice, we presented the top five highly expressed genes (Supplemental Digital Content, fig. 4C, https://links.lww.com/ALN/D49) and identified several novel marker genes (Gm32647 for cornu ammonis 3 neurons, Dlx6os1 for GABAergic neurons, and Prr5l for oligodendrocytes; Supplemental Digital Content, fig. 4D, https://links.lww.com/ALN/D49). These novel marker genes showed sex-specific differences and may help to uncover the characteristics and function of specific hippocampal cell types.

We investigated the hippocampal cellular cross-talk and focused on the trisynaptic circuit of the hippocampus (cornu ammonis 1, cornu ammonis 3, and dentate gyrus) during multiple sevoflurane exposures of both sexes. For the male mice, sevoflurane exposures enhanced the mutual interactions among cornu ammonis 1, cornu ammonis 3, and dentate gyrus, with increased numbers of significantly enriched ligand–receptor pairs in all interactions (from 18 to 42 in cornu ammonis 1–cornu ammonis 3, from 21 to 35 in cornu ammonis 1–dentate gyrus, from 25 to 45 in cornu ammonis 3–dentate gyrus, from 8 to 32 in cornu ammonis 3–cornu ammonis 3, and from 14 to 16 in dentate gyrus–dentate gyrus; fig. 2A; Supplemental Data 3, https://links.lww.com/ALN/D52). Our results showed specific expression patterns of the top 25 ligand–receptor interactions (fig. 2B).

For the female mice, sevoflurane exposures increased the number of significantly enriched ligand–receptor pairs from 15 to 21 in cornu ammonis 1–cornu ammonis 3, from 12 to 20 in cornu ammonis 1–dentate gyrus, and from 17 to 20 in cornu ammonis 3–dentate gyrus, while the number of ligand–receptor pairs remained 20 in cornu ammonis 3–cornu ammonis 3 and 11 in dentate gyrus–dentate gyrus (fig. 2C; Supplemental Data 4, https://links.lww.com/ALN/D53). Sevoflurane exposures altered the expression profile of the top 25 ligand–receptor interactions (fig. 2D).

In addition, we showcased the significantly enriched ligand–receptor interactomes among different hippocampal cell types in male mice (Supplemental Digital Content, fig. 10A, https://links.lww.com/ALN/D49; Supplemental Data 5, https://links.lww.com/ALN/D54) and cell–cell communication network diagrams (Supplemental Digital Content, fig. 10B, https://links.lww.com/ALN/D49). Sevoflurane also resulted in obvious changes in ligand–receptor gene regulatory interactions in female mice (Supplemental Digital Content, fig. 10, C and D, https://links.lww.com/ALN/D49; Supplemental Data 6, https://links.lww.com/ALN/D55).

Neurogenesis of dentate gyrus granule cells throughout postnatal life plays a critical role in various physiologic and pathophysiological conditions.²⁷  However, evidence is scant on the neurogenesis of dentate gyrus granule cells in developmental sevoflurane neurotoxicity. Thus, we explored the effects of multiple sevoflurane exposures on the maturation trajectory of dentate gyrus granule cells. The t-distributed stochastic neighbor embedding plots detected nine subclusters of dentate gyrus granule cells in the male mice (fig. 3A). The sevoflurane exposures changed the proportions of subcluster 0 from 26.1% to 30.2% and subcluster 1 from 26.0% to 22.9% (fig. 3B). To investigate the function of the subclusters, we analyzed their specific marker genes (fig. 3C; Supplemental Data 7, https://links.lww.com/ALN/D56). Based on the literature, dentate gyrus granule cell populations consist of neural stem cells (expressing Apoe, Slc1a3, Clu, Fabp7, Sox9, and Sox6), neural progenitors (expressing Eomes, Draxin, Dcx, Sox4, Ccnd2, and Neurod2), and mature granule cells (expressing Gria1), together constituting the neurogenic lineage.¹⁷  We then analyzed the maturation trajectory of dentate gyrus granule cells, showing the gene expression of subcluster 8 (neural stem cells) and subclusters 4, 2, 3, and 0 (neural progenitors) along with the temporal axis (fig. 3D). Based on subclusters 8, 4, 2, 3, and 0, the pseudotime analysis showed that sevoflurane exposures significantly interrupted the expression of Slc1a3 [F(6,680) = 4,074, P < 0.001], Sox6 [F(6,453) = 18,551, P < 0.001], and Neurod2 [F(6,453) = 2,576, P < 0.001] along the transition from neural stem cells to neural progenitors (fig. 3, E and F; Supplemental Data 8, https://links.lww.com/ALN/D57).

For the female mice, the t-distributed stochastic neighbor embedding plots illustrated 16 dentate gyrus granule cell subclusters (fig. 3G), without obvious changes in the proportions (fig. 3H). We also identified the specific marker genes (fig. 3I; Supplemental Data 9, https://links.lww.com/ALN/D63 and https://links.lww.com/ALN/D64) to analyze the maturation trajectory from subcluster 15 (neural stem cells) to 1, 2, and 12 (neural progenitors) (fig. 3J). Sevoflurane exposures significantly interrupted the expression of Dcx [F(3,35) = 400.2, P < 0.001], Draxin [F(3,33) = 15.1, P < 0.001], and Fabp7 [F(3,62) = 33,779, P < 0.001] in the pseudotime analysis (fig. 3, K and L; Supplemental Data 8, https://links.lww.com/ALN/D57). Thus, sevoflurane exposures led to sex-specific interruption of dentate gyrus granule cell neurogenesis, with significant changes in males but small changes in females.

Microglia are the main neuroimmune cells. The differentiation of microglia determines several essential functions, including synaptic pruning and remodeling, nerve regeneration and repairment, and pro- and anti-inflammatory effects.²⁸  We explored the impact of multiple sevoflurane exposures on microglia differentiation. The t-distributed stochastic neighbor embedding plots showed three microglia subclusters in the male mice (fig. 4A). Sevoflurane exposures increased the proportion of subcluster 0 from 83.4% to 96.3% (fig. 4B). The functions of these subclusters were defined based on known marker genes.²⁸  The subcluster 0 (expressing P2yr12, Cx3cr1, and Trem2) was related to sensing cellular injury, synaptic pruning, and promotion of inflammation, while the subcluster 1 (expressing Cd163 and Igf1) and subcluster 2 (expressing Mrc1 and Cd163) functioned to enhance nerve regeneration and repair (fig. 4C; Supplemental Data 7, https://links.lww.com/ALN/D56).^28–32  In the differentiation trajectory of microglia, different branches correspond to different cell states and functions. Microglia differentiated from the subcluster 0 to 1 and 2 in normal conditions. However, microglia almost all stayed in the subcluster 0 in the sevoflurane group, suggesting that multiple sevoflurane exposures hampered microglia differentiation (fig. 4D). We also identified many key genes that determined the differentiation branches of microglia (Supplemental Digital Content, fig. 11, https://links.lww.com/ALN/D49).

For the female mice, the t-distributed stochastic neighbor embedding plots showed two microglia subclusters (fig. 4E). The proportion of subcluster 0 was slightly changed from 92.4% to 89.4% (fig. 4F). Based on the marker genes (fig. 4G; Supplemental Data 9, https://links.lww.com/ALN/D63 and https://links.lww.com/ALN/D64), we defined the functions of the subclusters. The sevoflurane female group showed almost no changes in the microglia differentiation trajectory (fig. 4H). The Kyoto Encyclopedia of Genes and Genomes pathway enrichment analysis suggested that the subcluster 0 was related to the nervous system, and the subcluster 1 was related to the immune system (Supplemental Digital Content, fig. 12, https://links.lww.com/ALN/D49). Thus, sevoflurane exposures resulted in sex-specific inhibition of microglia differentiation, with remarkable changes in males but almost no changes in females.

Cornu ammonis 1 pyramidal cells exhibit a high level of diversity, which plays key roles in spatial navigation and memory circuit.³³  Whether multiple neonatal sevoflurane exposures affect the diversity of cornu ammonis 1 neurons remains unknown. We discovered the patterns of cornu ammonis 1 cell diversity using the t-distributed stochastic neighbor embedding analysis to reveal 19 cornu ammonis 1 pyramidal cell subclusters in the male mice (fig. 5A). The proportion of the subcluster 0 was remarkably decreased from 41.8% to 5.8% (fig. 5B). We visualized the cornu ammonis 1 pyramidal cell states using the pseudotime approach. There were 11 states in the control mice, suggesting the cell diversity in the normal condition. However, multiple sevoflurane exposures eliminated this diversity, leaving three states only (fig. 5C). To further uncover the key genes underlying this phenomenon, we assessed the expression of the top genes (Cacnb2, Cst3, Dpp10, Lingo2, Lrrc4c, and Tshz2) in different states (fig. 5D; Supplemental Data 10, https://links.lww.com/ALN/D58). Cacnb2 was the top gene in state 1 of both groups. Dpp10, Lingo2, Lrrc4c, and Tshz2 were the top genes in state 2 of the sevoflurane male group and in state 8 of the control male group. Cst3 was the top gene in state 3 of the sevoflurane male group and in state 6 of the control male group. Thus, the three states (1, 2, and 3) in the sevoflurane male group appeared to be representative of the three states (1, 8, and 6) in the control male group.

For the female mice, the t-distributed stochastic neighbor embedding plots showed 14 cornu ammonis 1 pyramidal cell subclusters (fig. 5E), and the proportion of the subclusters remained unchanged after sevoflurane treatment (fig. 5F). By visualizing the cornu ammonis 1 pyramidal cell states, the two female groups had similar cell diversity profiles (fig. 5G). We also assessed the expression of the top genes in different cell states and found that the five states of the sevoflurane female group corresponded to the states of the control female group (fig. 5H; Supplemental Data 10, https://links.lww.com/ALN/D58). Therefore, sevoflurane exposures caused sex-specific reduction in cornu ammonis 1 pyramidal cell diversity, with dramatic changes in males but small changes in females.

Oligodendrocytes arise from oligodendrocyte precursor cells, which influence the generation and myelination of oligodendrocytes;³⁴  however, the sex-specific effects of sevoflurane exposures on the differentiation of oligodendrocytes have not been reported. In the male mice, we analyzed the differentiation trajectory from oligodendrocyte precursor cells to oligodendrocytes (fig. 6A). The results showed that both the control male and sevoflurane male groups had three cell differentiation states (fig. 6B). In addition, we explored the mRNA expression of Mbp, Mag, Mal, and Mog (known cell markers of oligodendrocytes), showing that the expression of these four marker genes was comparable between the two groups (fig. 6C).

For the female mice, we visualized the differentiation trajectory (fig. 6D) to show that the number of oligodendrocyte states increased from three to five in the sevoflurane female group (fig. 6E). The Gene Set Enrichment Analysis enrichment plot suggested that most upregulated differentially expressed genes were enriched in the process of positive regulation of myelination (Gene Ontology: 0031643) (fig. 6F). The sevoflurane female group had significantly increased mRNA expression of Mbp (0.19 ± 0.01 vs. 0.15 ± 0.01; mean difference = 0.04; 95% CI, 0.03 to 0.06; P < 0.001) and Mag (0.16 ± 0.01 vs. 0.12 ± 0.01; mean difference = 0.04; 95% CI, 0.03 to 0.05; P < 0.001) compared to the control female group (fig. 6G). Thus, sevoflurane exposures induced sex-specific effects on oligodendrocyte differentiation, with significant changes in females but very small changes in males.

To determine key genes and signaling pathways during sevoflurane exposures, we identified differentially expressed genes within each cell cluster in the male mice (table 1; Supplemental Data 11, see the Supplemental Digital Content section at the end of the article for links to the data). Annotations of the differentially expressed genes with enriched Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathways in main cell types may help to explain the fundamental pathogenesis of developmental sevoflurane neurotoxicity (Supplemental Data 12, https://links.lww.com/ALN/D59 and Data 13, https://links.lww.com/ALN/D60). We found that the majority of the differentially expressed genes were unique to specific cell types, such as 402 differentially expressed genes in endothelial cells and 262 differentially expressed genes in neurons (fig. 7A). Most differentially expressed genes in neuron cell types were subtype-specific, such as 127 differentially expressed genes in cornu ammonis 3 neurons and 96 differentially expressed genes in cornu ammonis 1 neurons (fig. 7B). By visualizing the expression patterns of these unique differentially expressed genes, we revealed clear differential expression profiles in the endothelial cells, neurons, astrocytes, cornu ammonis 3 neurons, and cornu ammonis 1 neurons (fig. 7C).

For the female mice, we identified differentially expressed genes within each hippocampal cell cluster (Supplemental Digital Content, fig. 13A, https://links.lww.com/ALN/D49; Supplemental Data 14, see the Supplemental Digital Content section at the end of the article for links to the data) and analyzed the enriched Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathways (Supplemental Data 15, https://links.lww.com/ALN/D61 and Data 16, see the Supplemental Digital Content section at the end of the article for links to the data). The majority of the differentially expressed genes were also unique to specific cell types, such as 126 differentially expressed genes in oligodendrocytes and 379 differentially expressed genes in neurons (Supplemental Digital Content, fig. 13B, https://links.lww.com/ALN/D49). Most differentially expressed genes in neuron cell types were subtype-specific, such as 178 differentially expressed genes in cornu ammonis 3 neurons and 86 differentially expressed genes in cornu ammonis 1 neurons (Supplemental Digital Content, fig. 13C, https://links.lww.com/ALN/D49).

In our previous study, we utilized RNA sequencing to identify 314 hippocampal differentially expressed genes from the male mice subjected to multiple neonatal sevoflurane exposures.¹¹  Our current single-nucleus RNA sequencing in the male mice revealed a total of 1,035 differentially expressed genes, many of which were cell type–specific and cannot be retrieved from the whole-tissue analysis. There are 24 common differentially expressed genes between our previous whole-tissue results and the current single-cell genomic analysis (fig. 7D), which may be mechanistic driving forces for the sevoflurane-induced hippocampal cell type–specific effects during development in mice. We evaluated the expression of 24 common differentially expressed genes across main cell types (Supplemental Digital Content, fig. 14, https://links.lww.com/ALN/D49; Supplemental Data 11, see the Supplemental Digital Content section at the end of the article for links to the data).

The fluorescence in situ hybridization quantification assay was used to assess the mRNA expression and localization of six common differentially expressed genes related to neural function and memory. Our results showed increased expression of three common genes, as measured with mean fluorescence intensity: Grid1 (0.12 ± 0.02 vs. 0.07 ± 0.01; mean difference = 0.05; 95% CI, 0.01 to 0.08; P = 0.019), Ctnnd2 (0.13 ± 0.03 vs. 0.06 ± 0.01; mean difference = 0.07; 95% CI, 0.02 to 0.12; P = 0.014), and Sobp (0.13 ± 0.02 vs. 0.09 ± 0.02; mean difference = 0.04; 95% CI, 0.001 to 0.08; P = 0.045) (fig. 8). The Western blot showed that the protein expression of delta-catenin encoded by Ctnnd2 was increased after sevoflurane treatment (1.94 ± 0.30 vs. 0.74 ± 0.23; mean difference = 1.20; 95% CI, 0.74 to 1.67; P < 0.001; Supplemental Digital Content, fig. 15, https://links.lww.com/ALN/D49).

In this single-cell nucleus RNA sequencing study, we identified novel sex-specific cell type marker genes in the hippocampus and demonstrated sex- and cell type–specific effects of sevoflurane on distinct hippocampal cell populations. In male but not in female mice, multiple neonatal sevoflurane exposure altered ligand–receptor interactomes in the trisynaptic hippocampal circuit of cornu ammonis 1, cornu ammonis 3, and dente gyrus, interfered with dentate gyrus granule cell neurogenesis, hampered microglia differentiation, and decreased cornu ammonis 1 pyramidal cell diversity. In contrast, oligodendrocyte differentiation was mainly altered in female but not in male mice.

To identify the cell clusters in our study, we compared our cluster-specific gene signatures to the known and published patterns of hippocampal cell types.^16–19  For the accuracy of the clusters in our study, we showed significant overlap between our defined cluster-specific genes and previously known markers. Next, we validated that our identified cell types fit into the Allen Brain Atlas data set by visualization of their hippocampal location. We also found an unknown cell cluster in either sex of mice, and the cell signatures have not been reported in the literature. Due to the small number of cells (66 cells in male mice and 135 cells in female mice) in these unknown clusters, we assume that they exert limited biologic function only. Nevertheless, further studies are required to assess their identity and function.

Emerging evidence suggests that the pathophysiological bases of many brain disorders can be accounted for by disturbances in the neurologic network, other than a small number of gene mutations or risk factors.³⁵  The perturbations in the interaction patterns among hippocampal neural circuits contributed to cognitive impairment in neurologic injury.¹⁸  Currently, there is no evidence to show the changes in the interactions among different hippocampal neuronal and nonneuronal cell types after multiple neonatal exposures to sevoflurane. We assessed the abundance of ligand–receptor interactions and uncovered the cell–cell communication relationship. We found significant shifts in the expression patterns of the hippocampal neural circuit, implicating a reorganization of the hippocampal network in response to sevoflurane. Based on these findings, our results point toward hippocampal circuit perturbations underlying sevoflurane exposures during development.

Neurogenesis in the hippocampus is restricted to the dentate gyrus subgranular zone and occurs constitutively throughout postnatal life.²⁷  Neural stem cells generate differentiated neural progenitor cells and ultimately produce mature granule neurons, which plays an essential role in cognitive function.³⁶  A recent study suggested that the ongoing hippocampal neurogenesis contributed to sustaining cognitive function throughout life, and the decreased neurogenesis with aging was associated with cognitive impairment or neuropsychiatric diseases.³⁷  In a rat model of anesthesia exposure, repeated neonatal sevoflurane treatments resulted in neurocognitive abnormalities, impaired proliferation of dentate gyrus neural progenitors, and decreased survival of new granule cells.⁷  In our study, different dentate gyrus granule cells (neural stem cells, neural progenitors, and mature granule cells) at different developmental stages were identified according to known hallmark genes. Sevoflurane exposures led to sex-specific interruption of dentate gyrus granule cell neurogenesis, with changes of different key genes related to dentate gyrus cell neurogenesis along the maturation trajectory, such as Slc1a3, Sox6, and Neurod2 in male mice and Dcx, Draxin, and Fabp7 in female mice.

As principal brain immune cells, microglia are related to homeostasis and host defense responses against pathogens in the central nervous system.³⁸  Previous RNA sequencing studies identified the microglia-specific markers in the brain related to essential functions: (1) sensing environment, (2) housekeeping, and (3) protection against infectious pathogens, injurious self-proteins, and tumor cells.^28,29  Microglia sense and scan the surrounding area constantly and polarize toward the injury area rapidly, and they are regulated by genes such as P2yr12 and Itgb5 in the subcluster 0 of the male mice in our results.³⁹  The housekeeping functions consist of synaptic remodeling, migration, and maintaining myelin homeostasis (involving Cx3cr1 and TREM2 in the subcluster 0).^30,31  Moreover, microglia exert protective functions to reduce inflammation and repair cell damage (involving Mrc1, cd163, and Igf1 in the subclusters 1 and 2 of the male mice).³²  Thus, sevoflurane exposures blocked the process of microglia differentiation from the subcluster 0 to subclusters 1 and/or 2 in the male mice, which may lead to persistent neuroinflammation, neurotoxicity, and finally neurodegeneration. For the female mice, we identified two subclusters, with subcluster 0 related to the nervous system and 1 related to the immune system. Sevoflurane exposures led to only mild changes in the differentiation trajectory of microglia in female mice.

Cornu ammonis 1 pyramidal neurons with diverse differentiation states and the cornu ammonis 3–cornu ammonis 1 circuit are essential for memory processing and storage.⁴⁰  The existing clinical evidence showed that cognitive and memory declines were the most characteristic pathologic hallmarks in anesthetic-induced developmental neurotoxicity.⁴¹  We found that multiple neonatal sevoflurane exposures considerably affected the pyramidal cells, particularly the cornu ammonis 1 neurons, with only minor effects on the GABAergic or cholinergic neurons. In the male mice, the differentiation states of cornu ammonis 1 cells were remarkably reduced to only three states after sevoflurane treatment compared with 11 states in the normal conditions. As the three states in the sevoflurane group were probably to be included in the control group, we assumed that sevoflurane exposures resulted in a narrowing of cornu ammonis 1 diversity outcome in male mice. In contrast, the female control and sevoflurane groups exhibited similar profiles of cornu ammonis 1 pyramidal cell diversity. Notably, while there was a dramatic loss of cornu ammonis 1 cell diversity in males, there was almost no change in females.

Oligodendrocytes derive from oligodendrocyte precursor cells, maintaining myelin integrity to retain fast electrical conduction and axonal support.⁴²  Recent studies have reported the effects of sevoflurane on myelinization. Zhang et al. reported that multiple sevoflurane exposures resulted in folate metabolism disruption and subsequently myelination defects in the developmental brain.⁴³  Fan et al. reported that maternal sevoflurane exposure blocked the development and maturation of oligodendrocytes in the hippocampus of offspring mice.⁴⁴  Our results suggested that sevoflurane exposures may affect hippocampal oligodendrocyte myelination in the female mice, but not in the male mice. This is evidenced by increased expression of Mbp and Mag in relation to positive regulation of myelination.

By analyzing the common differentially expressed genes between our previous whole-tissue study and the current single-cell genomic results, we found several multiple cell type differentially expressed genes in relation to neural function and memory. Dcc (the receptor deleted in colorectal cancer) and its ligand netrin-1 are essential for hippocampal synaptic plasticity and spatial memory.⁴⁵ Grid1, a member of the ionotropic glutamate receptor family, is crucial for normal synapse function, learning, and memory.⁴⁶ Shank2, coding a scaffolding protein expressed on the postsynaptic membrane of glutamatergic neurons, regulates the number of synapses, dendrite length and complexity, synaptic plasticity, and neuronal morphogenesis.⁴⁷ Ctnnd2, coding delta-catenin (a brain-specific protein), regulates dendritic complexity and dendritic spine density in the mature brain.⁴⁸ Pcdh7 is highly expressed in the brain, which is related to the regulation of N-methyl-D-aspartate receptor currents and central nervous system disorders.⁴⁹ Sobp, encoding a nuclear zinc finger protein in the limbic system, is involved in intellectual disability.⁵⁰ 

This study has several limitations. First, the Morris water maze test confirmed the successful establishment of cognitive impairment model in mice of both sexes. We showed similar cognitive behavioral disorders between the male and female mice; however, it is plausible that the underlying mechanisms are different at the cellular level. Thus, our results provide novel mechanistical insights for further investigations. Second, we assessed the sevoflurane-induced hippocampal cell type–specific effects at postnatal day 37 only. Including more timepoints for observation would have provided more information in the acute phase after multiple sevoflurane exposures. Third, the relatively low number of animals used in these experiments does not allow statistical comparison between males and females, but our data are suggestive of sex-specific effects. Next, because of the low number of animals per analysis, future experiments are still needed for the replication of our results and cluster solution. Last, the role of the key risk genes in specific hippocampal cell types identified in our study needs further investigation in functional studies, preferably using transgenic mice.

In conclusion, this single-nucleus RNA sequencing study offers a comprehensive hippocampal cell type atlas and transcriptional profiles in mice of both sexes after multiple neonatal exposures to sevoflurane. We discovered sex-specific perturbations in cellular cross-talk, dentate gyrus granule cell neurogenesis, microglia differentiation, cornu ammonis 1 neuron diversity, and oligodendrocyte myelination with novel risk genes that may serve as underlying mechanisms. The results of our study provide a valuable resource for sevoflurane-induced hippocampal neuronal and nonneuronal cell type–specific and sex-specific effects during development.

The authors thank Guangzhou Genedenovo Biotechnology Co., Ltd. (Guangzhou, China), for assisting in sequencing and bioinformatics analysis.

This work was supported by the National Natural Science Foundation of China (Beijing, China; 82001126 to Dr. Song, 82072130 and 81873925 to Dr. Ji), Natural Science Foundation of Jiangsu Province (BK20191171 to Dr. Ji), 333 High-level Talent Training Project in Jiangsu Province (BRA2020089 to Dr. Ji), and Six Talent Peaks Project in Jiangsu Province (WSN-022 to Dr. Ji).

The authors declare no competing interests.

Supplemental Figures: https://links.lww.com/ALN/D49

Supplemental Data 1. Details of cell clusters in male mice, https://links.lww.com/ALN/D50

Supplemental Data 2. Details of cell clusters in female mice, https://links.lww.com/ALN/D51

Supplemental Data 3. Interactions of cornu ammonis 1, cornu ammonis 3, and dentate gyrus in male mice, https://links.lww.com/ALN/D52

Supplemental Data 4. Interactions of cornu ammonis 1, cornu ammonis 3, and dentate gyrus in female mice, https://links.lww.com/ALN/D53

Supplemental Data 5. Ligand–receptor interactomes in male mice, https://links.lww.com/ALN/D54

Supplemental Data 6. Ligand–receptor interactomes in female mice, https://links.lww.com/ALN/D55

Supplemental Data 7. Marker genes of dentate gyrus subclusters in male mice, https://links.lww.com/ALN/D56

Supplemental Data 8. Pseudotime analysis of dentate gyrus neurogenesis, https://links.lww.com/ALN/D57

Supplemental Data 9. Marker genes of dentate gyrus subclusters in female mice, https://links.lww.com/ALN/D63, https://links.lww.com/ALN/D64

Supplemental Data 10. Key genes of cornu ammonis 1 cell diversity in male mice, https://links.lww.com/ALN/D58

Supplemental Data 11. Differentially expressed genes within each cell cluster in male mice, https://links.lww.com/ALN/D65, https://links.lww.com/ALN/D66, https://links.lww.com/ALN/D67, https://links.lww.com/ALN/D68, https://links.lww.com/ALN/D69, https://links.lww.com/ALN/D70, https://links.lww.com/ALN/D71, https://links.lww.com/ALN/D72

Supplemental Data 12. Gene Ontology terms of differentially expressed genes in male mice, https://links.lww.com/ALN/D59

Supplemental Data 13. Kyoto Encyclopedia of Genes and Genomes of differentially expressed genes in male mice, https://links.lww.com/ALN/D60

Supplemental Data 14. Differentially expressed genes within each cell cluster in female mice, https://links.lww.com/ALN/D73, https://links.lww.com/ALN/D74, https://links.lww.com/ALN/D75, https://links.lww.com/ALN/D76, https://links.lww.com/ALN/D77, https://links.lww.com/ALN/D78, https://links.lww.com/ALN/D79, https://links.lww.com/ALN/D80, https://links.lww.com/ALN/D81

Supplemental Data 15. Gene Ontology terms of differentially expressed genes in female mice, https://links.lww.com/ALN/D61

Supplemental Data 16. Kyoto Encyclopedia of Genes and Genomes of differentially expressed genes in female mice, https://links.lww.com/ALN/D82, https://links.lww.com/ALN/D83, https://links.lww.com/ALN/D84, https://links.lww.com/ALN/D85, https://links.lww.com/ALN/D86, https://links.lww.com/ALN/D87, https://links.lww.com/ALN/D88, https://links.lww.com/ALN/D89, https://links.lww.com/ALN/D90, https://links.lww.com/ALN/D91, https://links.lww.com/ALN/D92, https://links.lww.com/ALN/D93