General anesthetic action profile on the human prefrontal cortex cells through comprehensive single-cell RNA-seq analysis

Summary The cellular and molecular actions of general anesthetics to induce anesthesia state and also cellular signaling changes for subsequent potential “long term” effects remain largely elusive. General anesthetics were reported to act on voltage-gated ion channels and ligand-gated ion channels. Here we used single-cell RNA-sequencing complemented with whole-cell patch clamp and calcium transient techniques to examine the gene transcriptome and ion channels profiling of sevoflurane and propofol, both commonly used clinically, on the human fetal prefrontal cortex (PFC) mixed cell cultures. Both propofol and sevoflurane at clinically relevant dose/concentration promoted “microgliosis” but only sevoflurane decreased microglia transcriptional similarity. Propofol and sevoflurane each extensively but transiently (<2 h) altered transcriptome profiling across microglia, excitatory neurons, interneurons, astrocytes and oligodendrocyte progenitor cells. Utilizing scRNA-seq as a robust and high-through put tool, our work may provide a comprehensive blueprint for future mechanistic studies of general anesthetics in clinically relevant settings.


INTRODUCTION
General anesthetics have been used to induce anesthesia state during surgery for more than 150 years. Research being conducted in this field has built the foundation for developing more effective and safer compounds that elicit rapid induction and recovery from anesthesia. Their cellular and molecular modes of action have been investigated primarily through in vitro, in vivo and ex vivo animal models and their potential long-term effects in those settings on the central nervous system have recently been recognised. [1][2][3][4][5][6] General anesthetics have the rapid onset of action owing to their high lipid solubility and imminent effects on cell membrane ion channel conductance, which are currently considered to be the main target of anesthetics. 3,4,7-9 Indeed, general anesthetics were reported to act on voltage-gated Na + , K + and Ca 2+ channels, ligand-gated ion channels such as g-aminobutyric acid (GABA) receptor and N-methyl-D-aspartic acid (NMDA) receptor. [1][2][3][4] A recent study showed that action potential discharge of glutamatergic neurons of the retrotrapezoid nucleus is strongly enhanced by inhalational anesthetic isoflurane. The ionic mechanism includes the activation of an unidentified Na + -dependent inward current as well as the inhibition of a background K + current. 10 Similar findings were made with sevoflurane in mice. 3, 10 However, these studies on neuronal membrane proteins or related secondary messengers in rodents have not provided convincing evidence to explain the pharmacodynamic characteristics of general anesthetics in humans. 11 In this study, we cultured brain cells derived from the human prefrontal cortex (PFC), which governs the highest-order cognitive functions including memory, decision-making capability, and social behavior, [12][13][14][15][16][17][18] of deceased human fetuses, as only brain cells at such developmental stage can be cultured under current technology for experimental utilization. PFC cells were then exposed to a clinically relevant concentration of inhalational anesthetic sevoflurane or intravenous anesthetic propofol followed by single-cell RNA sequencing (scRNA-seq), whole-cell patch clamp, and calcium transient analysis. Our study provides a comprehensive blueprint for understanding the molecular, transcriptome, and functional effects of general anesthetics on the different cell types within the human prefrontal cortex, and facilitates future studies to establish the mechanistic basis of the functional sequelae of general anesthetics in humans.

Brain cell populations affected by sevoflurane and propofol
In this study, the brain cells derived from the prefrontal cortex of deceased fetuses at 28 weeks gestation were cultured, and cell-type specific markers (PTPRC for microglia NEUROND2 for excitatory neurons, GAD1 for interneurons, AQP4 for astrocytes and PDGFRA for oligodendrocyte progenitor cells) were applied 19 to distinguish these cells into five distinct groups by scRNA-seq analyses ( Figures 1A, 1B, 2A, and 2B). Of the 22,153 genes analyzed by scRNA-seq, the top 3 expressed genes based on copy numbers from each cell cluster were mapped and showed no overlap, except STMN2 being highly expressed by excitatory neuron and interneuron clusters. Therefore, the cell-type markers selected sufficiently differentiated the different cell populations in a mixed culture ( Figure 2C).
In situ immunostaining showed that these cells were positive for Tuj1, Map2 or GFAP ( Figures 1C-1E). Exposure to sevoflurane or propofol for 6 h, which is the upper ''threshold'' of routine anesthetic exposure for surgeries, did not change the length of axons of neurons (p > 0.05) when compared to controls ( Figure 1F). ScRNA-seq also revealed that the PTPRC-positive population expanded following sevoflurane or propofol exposure, implying the increased number of microglial cells. A significant increase in microglia population occurred for 6 h propofol exposure and sustained throughout 2 h recovery, but this increase induced by sevoflurane was only remarkable during 2 h recovery relative to the control cultures with mock treatment at each corresponding time point (Figures 2D and 2E). Moreover, the cell transcriptional similarity analysis within each cell cluster revealed that, unlike propofol, sevoflurane decreased microglia transcriptional similarity ( Figure 2F).

Gene and enrichment pathways affected by sevoflurane and propofol
The top 40 regulated genes and top 30 enriched pathways by gene ontology analysis during sevoflurane or propofol exposure and following recovery in five types of PFC cells are presented herein (Figures 3A-3D and S1-S5). In excitatory neurons, sevoflurane strongly upregulated ARID5A (AT-Rich Interaction Domain 5A, an eukaryote conserved transcriptional factor involved in cell growth and tissue-specific gene expression) and the classical neuronal immediate-early genes IER2 (Immediate-Early Response 2) and FOS (Fos Proto-Oncogene). A handful of genes suppressed by sevoflurane included MICU2 (Mitochondrial Calcium Uptake 2, mitochondrial calcium uniport regulator), JAKMIP1 (Janus Kinase And Microtubule Interacting Protein 1, microtubule-dependent transport of the GABA-B receptor), RBM4B (RNA Binding Motif Protein 4B, translational activator of circadian clock mRNA PER1), ATP6V1D (ATPase H+ transporting V1 subunit D, component of vascular ATPase for the acidification of intracellular organelles) and HIF1A (hypoxia-inducible factor 1 subunit alpha, a transcriptional factor that orchestrates metabolic adaptation to hypoxia (Figure 3E and Table S1). Propofol exposure induced a distinct set of genes from sevoflurane, include CNR1 (cannabinoid receptor 1, mediating the mood and cognition alteration effects of cannabinoids. NAV2 (neuron navigator 2, cellular growth and migration), PCLO (piccolo presynaptic cytomatrix protein, component of presynaptic cytoskeletal matrix to enable synaptic vesicle trafficking). On the other hand, propofol potently downregulated SLC25A39 (Solute Carrier Family 25 Member 39, inner mitochondrial membrane transporter), IER3IP1 (Immediate-Early Response 3 Interacting Protein 1, endoplasmic reticulum stress sensor that mediates cell differentiation and apoptosis), RTN4RL2 (Reticulon 4 Receptor-like 2, cell surface receptor inhibiting axon outgrowth), and THY1 (Thy-1 cell surface antigen, immunoglobulin involved in cell adhesion/communication of the nervous system) ( Figure 3F and Table S1). In order to validate the transcriptional changes at the translational level, the protein expression of HIF1A and CNR1 induced by anesthetics were selected for further determination. The immunoblotting data showed that HIF-1a ( Figure 3G) was downregulated by sevoflurane whilst CNR1 expression was upregulated by propofol ( Figure 3H). These were corroborated with the transcriptome finding reported above. Interestingly, after 2 h recovery, there is a tendency for the altered transcriptome to return to baseline, for example ARID5A, IER2, FOS. and CNR1 (Figures S1A and S1B). There was enrichment for pathways responsible for spliceosome, RNA transport, and neuroactive ligand-receptor interaction ( Figures 3B and 3E). After 2 h of recovery, there was also  iScience Article differential enrichment in pathways associated with morphine addiction, axon guidance, and steroid biosynthesis (Figures S1A and S1B).
The heatmaps generated from gene set enrichment analysis (GSEA) showed that sevoflurane and propofol differentially regulated gene expression across 5 different cell types. Sevoflurane had mixed effects (up/ downregulation) on gene expression of excitatory neurons during 6 h exposure, i.E. genes associated with cell cycle and Axon guidance were inhibited and TNF signaling pathway were activated. During 2 h recovery post-sevoflurane exposure the majority of gene expression declined to below baseline levels such that they were downregulated relative to naive controls ( Figure 3I). In contrast, propofol appeared to upregulate gene expression of almost all enriched pathways in excitatory neurons during 6 h exposure, preceding a systemic downregulation during 2 h recovery ( Figure 3J). Upregulated pathways included PI3K-Akt, ECM-receptor interaction, AMPK, TNF, Dorsoventral axis formation, and phosphatidylinositol and focal adhesion appeared to be activated during 6 h of propofol exposure. Only SNARE interaction in vesicular transport, Parkinson's disease-related genes, and oxidative phosphorylation were downregulated. The transcriptome changes in other brain cell types including astrocytes, interneurons, microglia, and oligodendrocyte progenitor cells (OPCs) exposed to sevoflurane and propofol were comparable ( Figure S6).
The additional scRNA sequencing analysis was done to directly compare the differential effects between sevoflurane and propofol on five types of brain cells, as well as enrichment analysis of the top 30 signal pathways. In excitatory neurons, a stark contrast between sevoflurane and propofol is evident, for example CNR1 and NAV2 were enhanced by propofol only, whereas FOS was upregulated by sevoflurane (Figure 4A). Similar to the data reported in Figures 3B and 3D, the HIF-1 signaling pathway, axon guidance, and AMPK signaling pathway were enriched ( Figure 4B).

Pseudo-time-trajectory of brain cells unaffected by either anesthetic
The pseudo-time trajectory analysis on scRNA-seq data from brain cells exposed to sevoflurane or propofol demonstrated that neither anesthetic altered the single-cell trajectories 20,21 on five different brain cell types throughout the 6 h exposure plus recovery period ( Figures 4C, 4D, 4F and 4G and S7-S9). As exemplified by interneurons, despite the substantial expression pattern changes identified in the heatmaps of enriched ontology terms ( Figures 4E and 4H), those changes were not associated with significant cellular deviations from the control state. These may suggest that either propofol or sevoflurane did not considerably disrupt the overall transcriptome of primary human brain cells at the developing stages during and after exposure.

Partial unique effects of general anesthetics on ion channels and calcium transient
Gene set enrichment analysis from 5 types of cells further identified a total of 341 human ion channelrelated genes that responded to anesthetics including those encoding the subunits of voltage-gated Na + , K + , Ca 2+ ion channels, ligand-gated NMDA, AMPA, and GABA receptor. It can be readily appreciated that many of those subunits are putative binding targets of general anesthetics including the GABA A R receptor subunit-alpha1, -alpha2, -beta1, -beta2, and -gamma1 encoded by GABRA4, GABRA2, GABRG2, and GABRB1; all of which exhibited immediate and reversible transcriptional perturbations in response to sevoflurane or propofol exposure (Figures S10 and S11). As exemplified by the excitatory neurons, iScience Article propofol and sevoflurane differentially regulated ion channel-related genes. For example, sevoflurane altered the expression of KCNK12, GRIA1, and KCNF1 (K + ion channels-related genes), ITPR1, GRIN2B, and CACNB3 (Ca 2+ ion channels-related genes) and SCN2A (Na + ion channels-related gene), which are distinct from propofol-associated transcriptional changes in GRIK5, KCNK2, and KCNJ6 (K + ion channels-related genes), PKD1, ITPR1 and GRIN1 (Ca 2+ ion channels-related genes) and GRID2 (Na + ion channels-related gene) ( Figures 5A and 5B).
Sevoflurane suppressed both the outward K + current and the inward Ca 2+ current in a concentrationdependent manner, whereas an increasing propofol concentration effectively blocked the inward Na + current. The inhibitory effect of sevoflurane on K + and Ca 2+ current appeared to be more pronounced than that of propofol ( Figures 5C-5H). This is in line with our previous enrichment of related genes, K + and Ca 2+ ion channel genes were inhibited after sevoflurane and propofol treatment (e.g., KCNK12, CACNB3, and ITPR1) while Na + channel genes change less (e.g., SCN2A and GRID2) ( Figures 5A and  5B). Our enrichment analysis also revealed extensive gene expression changes to ion channel subunits during anesthetic exposure and recovery phase (e.g., sodium voltage-gated channel alpha subunit 2/SCN2A, calcium voltage-gated channel subunit alpha 1A/CCNA1A, potassium voltage-gated channel subfamily D member 2/KCND2 (Figures 5A, 5B, and S10).
Furthermore, we found that propofol and sevoflurane led to transient, dose/concentration-dependent increase of intracellular calcium concentration ( Figures 5I-5L), which rapidly returned to the baseline level when anesthetics were eluted without intracellular calcium overload ( Figures 5M and 5N). This provides evidence that sevoflurane/propofol administered at clinically relevant doses do not disrupt intracellular calcium homeostasis and may be safe for developing human excitatory neurons. However, the gene enrichment heatmap ( Figure S10) identified increasing gene expressions related to inositol 1,4,5-triphosphate receptor (ITPR1/ITPR2) and ryanodine receptor (RYR1/RYR2/RYR3), which are.
Conventional RNA sequencing analysis also revealed that single exposure of propofol or sevoflurane was associated with substantial transcriptional alterations in the PFC mixed culture (Figures 6A-6E). Interestingly, the overall gene expression profile at 2 h recovery after sevoflurane or propofol became indistinguishable ( Figure 6B) albeit substantial divergence in gene expression pattern during the 6 h exposure to either anesthetic. Taken together, the data from scRNA-seq and conventional RNA sequencing with the PFC mixed cells demonstrated that sevoflurane and propofol-induced significant, differential, and temporal changes to gene expression across diverse ontology terms in all five cell types, and the brain cells were capable of ''self-normalising''/reversing such changes after the removal of general anesthetics ( Figures S12-S15).

DISCUSSION
Our single-cell-RNA sequencing study, for the first time, illustrated the comprehensive and diverse gene responses in different human PFC cell types following a single exposure to general anesthetics. Both propofol and sevoflurane at clinically relevant dose/concentration promoted ''microgliosis'' but only  iScience Article sevoflurane changed microglia gene similarity. Propofol and sevoflurane each extensively but transiently altered transcriptome profiling across microglia, excitatory neurons, interneurons, astrocytes, and oligodendrocyte progenitor cells. Within the excitatory neurons and microglia, exemplary ion-gated and ligand-gated ion channels related genes responsive to either anesthetic included SCN1A, CACNAB2, KCNA1, GABRR2 and GRINA1 amongst many others.
Both propofol and sevoflurane at clinically relevant dose/concentration promoted ''microgliosis'' but only sevoflurane changed microglia transcriptional similarity. In line with our findings, sevoflurane was reported to aggravate microglia-mediated neuroinflammation via the downregulation of PPAR-g in the hippocampus. 22 Collectively, our data showed that single exposure to general anesthetics does not alter cell number and similarity for most types of fetal PFC cells except microglia. Propofol and sevoflurane each extensively but transiently altered transcriptome profiling across microglia, excitatory neurons, interneurons, astrocytes, and oligodendrocyte progenitor cells. Of particular interest, sc-RNA seq heatmaps highlighted a large number of previously unidentified genes highly responsive to sevoflurane or propofol, exemplified by JAKMIP1, RBM4B, and THY1, to represent potential molecular targets through which general anesthetics impart amnesia, analgesia and loss of consciousness. Interestingly, only propofol increased CNR1 expression in excitatory neurons which may result in increasing brain endocannabinoid. 23,24 Recently, repeated exposure to sevoflurane in neonatal mice induced cognitive and behavioral deficits which was associated with the hippocampus transcriptome changes. 25 Yamamoto et al. further examined the differential gene expression across PFC, striatum, hypothalamus, and hippocampus in sevofluraneanesthetised mice and found that sevoflurane upregulated angiogenesis and downregulated brain cell differentiation-related genes. 26 Moreover, the dysregulated genes/pathways associated with sevoflurane exposure were also reported in rhesus macaques. 27 However, the work reported here was the first-line evidence on transcriptome changes in human fetal prefrontal cortex cells following general anesthetic exposure using scRNA sequencing technique, providing accurate identification of gene expression patterns within each cell type. Unlike enrichment of KLF4 and NFATC2 in animals (Yamamoto et al., 2020; Cheng et al., 2022), our study identified enrichment in KLF6, KLF10, and NFATC4 in human PFC cells that could be due to species variation.
We, for the first time, determined the effects of general anesthetics on sodium, potassium, and calcium currents from primary human excitatory neurons. Our findings are consistent with previous studies demonstrating a reduction in peak amplitude of potassium currents following sevoflurane exposure in xenopus oocytes. 28 In contrast to other studies showing that sevoflurane activated voltage-gated K + channels and increased outward K + current in the pons and cerebellar granule neurons in rats, 7 suppression of outward K + current following sevoflurane or propofol exposure was found in our study. This difference was likely due to that human prefrontal cortex neurons behave differently to other species. Within the excitatory neurons and microglia, exemplary ion-gated and ligand-gated ion channels related genes responsive to either anesthetic included SCN1A, CACNAB2, KCNA1, GABRR2, and GRINA1 amongst many others. These findings suggest that sevoflurane may have multiple mechanisms of action on brain cells when compared to propofol. This is corroborated with other studies; in vitro studies with spinal neurons showed a complete inhibition of action potential activity with sevoflurane but not propofol. 29 Furthermore, metabolomic profiling studies of parietal cortex of children's brains revealed that sevoflurane exposure led to Compared with the control, the activation threshold potential for I K (delayed rectifier potassium current) did not change significantly (À40 mV), and the IV curve was suppressed significantly with rising concentration of sevoflurane or propofol (p < 0.05; n = 18 neurons from three independent replicates). MAC, minimum alveolar concentration.
(E and F) Left, representative Na + trace recorded from À80 to +60 mV under control, sevoflurane (2 MAC) and propofol (10 mg/mL) conditions. Right, IV showing the relationships of voltage-gated Na + channel currents before and after the application of sevoflurane (1, 2, 5 MAC) and propofol (3, 10, 30 mg/mL) in cultured cortical neurons. Compared with the control, the activation threshold potential of Na + in the three groups did not change significantly (À40 mV); sevoflurane did not alter the IV curve (p > 0.05), whereas propofol dose-dependently flattened the IV curve (p < 0.05; n = 18 neurons from three independent replicates). iScience Article higher glucose consumption and lactate generation than propofol. 5 This suggested greater glycolysis, glutamate-neurotransmitter cycling and lactate shuttling between astrocytes and neurons following sevoflurane exposure. 5 Compelling evidence demonstrated that propofol acts upon gamma-aminobutyric acid type A (GABAA) receptors whereas the contribution of glycine receptors remains uncertain. 29,30 Volatile anesthetic sevoflurane also likely exerts its central depressant effect through the glycine and GABAA receptors, but indirect evidence suggests a mechanism of action distinct from propofol. 29 In the present study, we found that propofol and sevoflurane exposure were associated with significant, distinct gene expression alterations across an extensive list of ion channel-related genes, such as SCN8A, CACNB4, KCNJ3, GLRA2 and GABARA5 (Figures 5A and 5B). In addition, we also disclosed a large pool of previously unidentified ion channel/receptor subunits that may represent novel molecular targets of anesthetics.
Overall, it is yet to be determined whether the perturbations to ion channel subunit expression on single cell level found in our study are accountable for the transient inhibitory effects of anesthetics on ion channel currents and neuronal excitability.

Limitations of study
Our study is not without limitations. First, brain cell cultures were used and this in vitro setting cannot fully recapitulate in in vivo setting and humans. Second, it is unclear to what extent the findings derived from fetal PFC cells can be extrapolated to the adult human brain. However, it should be pointed out that adult human brain cells cannot be cultured for the study type reported herein. Nevertheless, our study is the first one that clearly demonstrated the distinct and significant effects of general inhalational versus intravenous anesthetics on human brain cells at the gene transcriptome, molecular and cellular levels.
The distinct changes to microglial population and reactivity by both anesthetics are of significance, with studies showing that sevoflurane promoted neurofunctional impairment through glial-mediated neuroinflammation, 31 and that propofol inhibited microglial activation to protect traumatic brain injury. 32 Anesthetics have also been shown to increase cancer cell growth, 33 protect neuronal and other type cell death, 34 or even trigger neuronal cell death. 35,36 However, our data likely only represent a ''snapshot'' of the actions of general anesthetics on the human brain cells. Whether transcriptome alterations translate into mid-to long-term phenotypic and functional modifications needs to be addressed in the future. Nonetheless, the plethora of transcriptome changes and pathway enrichment identified by sc-RNA seq reported here may serve as a robust database for future mechanistic studies into general anesthetics.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:   Tissue sample collection and culture PFC tissue from four fetuses were processed and were subsequently hybridized to probe sets and pooled with four samples in a single lane of a Chromium chip. The tissues were collected and dissected in ice-cold Hibernate TM -E medium (Invitrogen, A1247601) with penicillin and streptomycin (Solarbio, P1400). The PFC tissues were chopped into small pieces with a scalpel blade and digested in 0.25% trypsin-EDTA (Solarbio, T1300) at 37 C for 10 min. The cell suspension was further pipetted to be dispersed to single cells, then the cells were passed through 70 mm filter and centrifuged at 1000 rpm for 7 min at 4 C. The supernatant was aspirated, and the cell pellet was resuspended and seeded to 6-well plates, covered with 20 mg/ml paraformaldehyde (PDL) (Sigma, P1524), in Dulbecco's Modified Eagle Medium (DMEM) (Hyclon, SH30022.01) with 10% foetal bovine serum (FBS) (Gbico, 10099141C). After 4 hrs incubation, the medium was aspirated, and the cells were cultured in Neurobasal medium (Gbico, 21103049) supplemented with 2% B27. Every three days half of the medium was replaced by fresh Neurobasal medium. 19,37,38 METHOD DETAILS Anesthetic exposure, immunofluorescence, single cell RNA-seq and immunoblotting analysis Cells were divided into 5 groups: Naive control (NC, collected at 8 hrs); Sevoflurane treatment for 6 hrs (Sevo 6h) and recovery for 2 hrs (Sevo-R2h); Propofol treatment for 6 hrs (Prop 6h) and recovery for 2 hrs (Prop-R2h). The cultured cells were incubated in a 1.5 L airtight temperature-controlled chambers with inlet and outlet valves and an internal electric fan. For sevoflurane (Jiangsu Hengrui Medicine, China) delivery, the chamber inlet was connected to a closed gas delivery system consisting of a calibrated oxygen flow metre and an inline sevoflurane vaporiser (Abbott Laboratories, Maidenhead, UK). The outlet gas was monitored (Datex gas monitor, Helsinki, Finland) to ensure the sevoflurane in chamber reach the target concentration (3.3%), 39 the equivalent of 1.0 minimum alveolar concentration, MAC). Cells were then incubated in the sealed chamber for 6 hrs at 37 C. 40 In the propofol group, the PFC cells were treated with 20 mM propofol (AstraZeneca, UK) for 6 hrs (the equivalent of half maximal effective concentration in vitro). 41,42 An exposure duration of 6 hrs and the anesthetic concentrations were chosen in accordance with previous studies and reflect the concentration and duration for the longest paediatric anesthesia. 43 In any groups either during anesthetic or mock exposure or recovery phase, all cells were treated under identical conditions except with or without anesthetics and exposed to the identical mixture gases of 20% oxygen, 5% carbon dioxide and balanced with nitrogen at 37 C.
The cells isolated from foetal prefrontal cortex were washed in PBS for twice, then followed by 4% paraformaldehyde (Biosharp Life Sciences, BL539A) for 15min. After fixing the cells, the 0.3% triton X-100 (Solarbio Life Science, T8200) were added to the cells to permeate the cell membrane. The penetrated cells were pre-incubated in 5% normal goat serum for 1 h at room temperature, followed by overnight incubation with primary antibody Tuj1 (