Introduction

Despite considerable investigation into the molecular targets, neural circuits, and functional connectivity associated with various anesthetics, our comprehension of their effects on overall brain activity continues to be limited and incomplete [1]. At the molecular level, ketamine (KET) and isoflurane (ISO) interact with N-methyl-D-aspartate (NMDA) and gamma-aminobutyric acid type A (GABAa) receptors, respectively, modulating neuronal excitability and ultimately leading to a loss of consciousness [2]. In systems neuroscience, the neural mechanisms of anesthetic induced unconsciousness involve both top-down and bottom-up processes [3, 4]. As evidenced by in vivo electro-physiology or functional magnetic resonance imaging (fMRI) studies, the top-down paradigm illustrates that anesthetics induce unconsciousness by disrupting corticocortical and corticothalamic circuits responsible for neural information integration, while peripheral sensory information can still be conveyed to the primary sensory cortex [5, 6]. The bottom-up approach, exemplified by ISO, involves the activation of sleep-promoting nuclei like ventral lateral preoptic nucleus (VLPO) and inhibition of arousal centers in the brain-stem and diencephalon, supporting the shared circuits of sleep and anesthesia [7, 8]. However, the limited spatial resolution of fMRI studies and the inability of EEG to capture specific brainstem nuclei hinder the acquisition of comprehensive whole-brain information. Although a substantial body of knowledge has been amassed, our understanding of the reciprocal responses among different brain regions during general anesthesia remains relatively sparse and fragmented. To bridge these gaps, further investigation using advanced techniques that can capture the whole-brain dynamics is needed to elucidate the complex interactions and shared mechanisms between various anesthetics.

Neuronal extracellular stimulation typically results in the elevation of adenosine 3’,5’-cyclic monophosphate (cAMP) levels and calcium influx, ultimately leading to the upregulation of immediate early genes (IEGs) such as c-fos [9, 10]. The translation product of c-fos, c-Fos protein, offers single-cell spatial resolution and has been utilized as a biomarker to identify anesthetic-activated brain regions [11]. Previous investigations of c-Fos expression throughout the brain demonstrated that GABAergic agents inhibited cortical activity while concurrently activating subcortical brain regions, including the VLPO, median preoptic nucleus (MnPO), lateral septal nucleus (LS), Edinger-Westphal nucleus (EW), and locus coeruleus (LC) [12–15]. In contrast, KET was shown to provoke wake-like c-Fos expression and intense augmentation of c-Fos expression in various brain regions at clinical dosages (75-100 mg/kg) [13]. It is important to note that these experiments administered KET at lights-on and GABAa receptor agonists at lights-off, potentially introducing circadian influences for direct comparison of ISO and KET. Moreover, it has been re-vealed that state of general anesthesia is not determined by activity in individual brain areas, but emerges as a global change within the brain. This change involves the activation of lateral habenular nucleus (LHb), VLPO, supraoptic nucleus (SON), and central amygdaloid nucleus (CeA), which are essential for anesthetics induced sedation, unconsciousness, or analgesia [7, 16–18]. However, brain-wide mapping and comparison of distinct anesthetic (KET and ISO) activated nuclei at a cellular level have not been fully elucidated.

In this study, we examined the distribution of nuclei activated by ISO and KET in 987 brain regions using immunochemical labeling and a customized MATLAB software package, which facilitated signal detection and registration to the Allen Mouse Brain Atlas reference [19]. We compared whole-brain c-Fos expression induced by KET and ISO through principal component analysis (PCA) and calculated inter-regional correlations by determining the covariance across subjects for each brain region pair. We then extracted significantly positively correlated brain regions to construct functional networks and performed graph theory-based network analyses to identify hub nodes. Our results uncovered distinct yet overlapping whole-brain activation patterns for KET and ISO.

Results

A comparison of the activation patterns of c-Fos in 53 brain areas in response to ISO and KET

To examine the pattern of c-Fos expression throughout the brain, 1.5% ISO was continuously ventilated, or 100 mg/kg KET was administrated 90 minutes before harvesting (Figure 1A). Raw images of brain slices were aligned to the Allen Mouse Brain Atlas (Figure 1B). The entire brain was divided into 12 main regions in grade four with a total of 53 subregions in supplementary table 1. We first calculated the c-Fos cell density in each brain region across the four groups and then log-transformed the data for principal component analysis (PCA). The top two principal components revealed a clear separation between the KET and ISO groups, with KET induced c-Fos expression closely associated with PC1, accounting for 30.72% of the variance, while ISO was associated with PC2, accounting for 25.89%. The absence of overlap between PC1 and PC2 suggests distinct patterns of brain region activation for each anesthetic (Figure 2A). Figure 2B illustrates the component coefficients of principal component 1 (PC1) and principal component 2 (PC2) for each brain region. The points in the figure represent brain regions with absolute component coefficients within the top 25%. KET, closely related to PC1, predominantly affected cortical regions, as evidenced by top 25% positive PC1 coefficients in MOB, AON, MO, ACA, and ORB, while subcortical areas exhibited negative coefficients. In contrast, ISO, showing a notable association with PC2, displayed top 25% positive coefficients in PALc, LSX of CNU, cortical ILA, and hypothalamic PVZ, with top 25% negative coefficients primarily found in cortical regions. These findings reveal distinct patterns of brain region activation for each anesthetic, with KET primarily affecting cortical regions, and ISO mainly influencing the central nucleus (CNU) and hypothalamus. To reduce inter-individual variability, we normalized c-Fos+ cell counts in each brain area by dividing them by the total c-Fos+ cells in the entire brain (Figure 2C), enabling a more accurate comparison of KET and SO induced activation patterns. In line with PCA results, KET significantly activated the isocortex (16.55 ± 2.31 vs. 40.13 ± 1.97, P < 0.001) and reduced the proportion of c-Fos+ cells in the hypothalamus (14.88 ± 2.18 vs. 4.03 ± 0.57, P = 0.001) and midbrain (21.87 ± 2.74 vs. 6.48 ± 1.06, P = 0.001) compared to the saline group. In contrast, ISO induced c-Fos+ cells were mainly observed in the striatum and hypothalamus compared to the home cage group, specifically in the periventricular zone (PVZ) (2.330 ± 0.759 vs. 6.63 ± 0.84, P = 0.004) and lateral zone (LZ) (2.82 ± 0.3 vs. 5.59 ± 0.75, P = 0.007) of the hypothalamus, as well as STRv (0.59 ± 0.10 vs. 2.77 ± 0.40, P < 0.001) and LSX (1.93 ± 0.59 vs. 5.11 ± 0.41, P = 0.0012) of the striatum (Figure 2C and Table 1). In summary, our findings reveal that KET predominantly affects cortical regions, whereas isoflurane mainly targets subcortical areas, specifically the striatum, and hypothalamus. These results demonstrate distinct patterns of brain region activation for each anesthetic agent.

Figure 1.

Figure 2.

Similarities and differences in ISO and KET activated c-Fos brain areas

To further elucidate the similarities and differences between ISO and KET induced c-Fos expression patterns, we conducted a PCA on the logarithm of cell density across 201 brain regions, aiming to analyze their responses and characterize the variance among the respective experimental groups (Figure 3). The top two principal components displayed well-separated ISO and KET groups (Figure 3A), where the KET group had a strong association with PC1 (accounting for 28.97% of the variance) and the ISO group was associated with PC2 (accounting for 15.54% of the variance), indicating distinct features for the ISO and KET groups. We then further calculated the PC coefficients corresponding to the c-Fos density within each brain region (Figure 3B). The points in the figure represent brain regions with absolute component coefficients within the top 25%, highlighting the most influential regions in the PCA analysis. KET, closely associated with PC1, mainly activates brain regions within the cerebral cortex (CTX), including the visceral area (VISC), claustrum (CLA), dorsal peduncular area (DP), orbital area (ORB), and temporal association areas (TEA), while significantly inhibiting the hypothalamus, midbrain, and hindbrain. In contrast, ISO, closely associated with PC2, predominantly activates the hypothalamic regions, such as the ventrolateral preoptic nucleus (VLPO), suprachiasmatic nucleus (SCH), tuberal nucleus (TU), medial preoptic area (MPO), and supraoptic nucleus (SON), as well as certain nuclei in the central nucleus (CNU), while suppressing the cortex, midbrain, and hindbrain. KET predominantly affects brain regions within the cerebral cortex (CTX), while significantly inhibiting the hypothalamus, midbrain, and hindbrain. This result is in line with the top-down mechanism proposed previously, whereby anesthetic agents suppress consciousness by modulating cortical and thalamocortical circuits involved in the integration of neural information. ISO primarily activates hypothalamic regions, while suppressing the cortex, midbrain, and hindbrain. This aligns with the bottom-up mechanism, where anesthetics suppress consciousness by modulating sleep-wakefulness nuclei and neural circuits in the brainstem and diencephalon that have evolved to control arousal states[4]. We summarized the responses differences between the two anesthetics in Figure 3C, KET acts mainly via the top down mechanism, affecting cortical regions, while ISO operates through the bottom up mechanism, primarily targeting subcortical areas such as the hypothalamus, suggesting that different anesthetic agents may achieve the loss of consciousness by modulating distinct brain regions and neural circuits.

Figure 3.

Identifying brain regions activated by KET

We further employed quantitative methods to identify the brain regions activated by KET across the entire brain. Our findings concur with previous studies, numerous cortical regions associated with somatosensory, auditory, visual, and movement were activated (Figure 4A and C, Supplementary Figure 1, Supplementary Table 2). Additionally, we identified several innovative observations that enrich the current understanding in this field. To provide a clearer overview of our findings, we categorized the activated brain areas based on their functions: (1) Arousal: Several nuclei associated with arousal were activated,includingtheprelimbicarea(PL),infralimbiccortex(ILA), paraventricular nucleus of the thalamus (PVT), sublaterodorsal nucleus (SLD), and dorsal raphe (DR). Prior evidence indicates that the PL/ILA and PVT regions play a role in regulating sleep and arousal in both cortical and subcortical areas [20]. The SLD has been reported to stabilize REM sleep, while dopaminergic neurons in the DR are involved in sleep-wake regulation [21, 22]. (2) Pain modulation, KET significantly activated pain-related areas such as the anterior cingulate cortex (ACA) in the cortex [23], the anterior pretectal nucleus (APN) in the thalamus, which is known to be involved in managing chronic pain [24], and the anterior periaqueductal gray (PAG) region and medial prefrontal cortex (mPFC), both part of the endogenous pain inhibitory pathway [25, 26],. Additionally, the activation of the locus coeruleus (LC) in the midbrain may contribute to KET’s analgesic effects [27]. (3) Neuroendocrine regulation: The paraventricular hypothalamic nucleus (PVH) and supraoptic nucleus (SON), which are neuroendocrine-related regions, were also activated [18, 28]. (4) Movement: Subcortical nuclei associated with movement, such as the subthalamic nucleus (STN) and nucleus incertus (NI), were prominently activated by KET administration [29, 30]. (5) Connectivity: We observed significant activation of the nucleus reunions (Re) located in the thalamus, which receives substantial afferent input from limbic structures and serves as a connector linking the hippocampus to the medial prefrontal cortex [31]. In summary, our study identified extensive activation of cortical and subcortical nuclei during KET anesthesia, encompassing regions related to arousal, pain modulation, neuroendocrine regulation, movement, and connectivity.

Figure 4.

Identifying brain regions activated by ISO

In our study, we aimed to identify brain regions activated by ISO and compare these with those activated by KET. We began by summarizing previously reported ISO activated nuclei, including PIR, LSd/LSv, CeA in the cortex and striatum, and VLPO, MnPO, EW, NTS, LC, ventral group of the dorsal thalamus (VeN), and area postrema (AP) in the hypothalamus and midbrain (Supplementary Table 3). We subsequently conducted a comprehensive MATLAB-based analysis of c-Fos expression throughout the entire brain, uncovering previously undetected activated nuclei (Figure 5A, Supplementary Figure 2). Newly identified activated nuclei in the CTX and the CNU included PL/ILA and ENT, aPVT in the thalamus, TU, ARH, PVi, and PVH in the hypothalamus, and PB in the hindbrain. All nuclei activated by ISO in this study were functionally classified and depicted in Figure 5C. Our results confirmed the activation of several brain regions involved in sleep-wakefulness regulation, such as the prelimbic area (PL) and infralimbic areas (ILA), and the anterior paraventricular nucleus (aPVT), leads to increased NREM sleep and decreased wakefulness [20]. Additionally, we observed activation in previously reported analgesia-related nuclei, including CeA and LC, as well as the parabrachial nucleus (PB) [18, 27, 32]. We also found activation in neuroendocrine function-related nuclei of the hypothalamus, such as TU, PVi, ARH, PVH, and SON. Moreover, we identified activations related to ISO induced side effects, such as in the piriform cortex (PIR) [33] and ENT [34]. which may be stimulated by ISO odor, and the solitary tract nucleus (NTS), potentially responsible for ISO induced vomiting [35]. The only activated nucleus in the midbrain was the Edinger-Westphal nucleus (EW). Recent research has found that sevoflurane activates EW and is involved in sleep induction and maintenance of anesthesia, suggesting its crucial role in general anesthesia [36]. By comparing the ISO and KET induced c-Fos expression, we summarized the brain regions activated by both anesthetics in Figure 5D. Despite variations in molecular targets, the coactivation of regions such as PL/ILA, aPVT, CeA, PVH, SON, EW, PB, LC, and NTS by both ISO and KET suggests an overlapping neuronal circuitry that influences sleep-wake regulation, analgesia, and neuroendocrine functions. This shared neural circuitry may potentially offer a common mechanism across the two anesthetics for the maintenance of general anesthesia.

Figure 5.

Network generation and Hub identification

Previous research has established that general anesthesia is mediated by various brain regions [7, 16–18]. c-Fos expression serves as an indicator of neuronal activity, providing a single index of activation per region per animal. By examining the covariance of c-Fos expression across animals within each group, we can infer interactions between brain regions and identify functional networks engaged during general anesthesia [37]. Highly correlated activities across different brain regions are presumed to constitute components of a functional network, reflecting the complex interplay of brain areas during general anesthesia. We first employed c-Fos expression as a neuronal activity marker and calculated a comprehensive set of interregional correlations for four groups. The matrices exhibited interregional correlations for the number of c-Fos-positive cells in each condition (Figure 6A). Network graphs were generated by extracting Pearson’s coefficients > 0.82, as well as significant positive correlations (P < 0.05), from these matrices to construct functional brain networks (Figure 6B). In comparison to the control group, isoflurane (ISO) slightly decreased interconnections between regions (network density: 0.13 vs. 0.10; edges: 267 vs. 198), with no significant difference in mean interregional correlation coefficients between the ISO group and the home cage group (Fisher Z = -0.018, P = 0.98). Conversely, ketamine (KET) significantly increased the network’s connectivity density (network density: 0.13 vs. 0.47; edges: 265 vs. 1008) and showed a significant increase in mean interregional correlation coefficients compared to the saline group (Fisher Z = 3.54, P < 0.001) (Figure 6C). These findings suggest that ISO may exert a mild inhibitory effect on functional network connectivity, whereas KET appears to enhance interregional correlation following administration (Figure 6D).

Figure 6.

Hubs are nodes that occupy critical central positions within the network, enabling the network to function properly. To determine the brain regions that serve as the hubs for ISO and KET induced functional networks, we calculated each node’s degree (the number of links that connect it) and betweenness centrality (the fraction of all shortest paths in the network that contain a given node). Nodes with a high degree and betweenness centrality typically have a lot of connections. We ranked each node according to its degree and betweenness centrality and extracted nodes with a rank greater than the 80th percentile in its network. Additionally, we segmented the functional network into non-overlapping modules using a spectral community detection algorithm and calculated the within-modal degree Z-score (the degree of nodes within a module, which indicates within-module connectivity) as well as participation coefficients (the distribution of the edges of a node between other modules, which indicates within-module connectivity) [38]. Nodes with relatively high values for both parameters were considered relatively significant within and outside the network module. We classified nodes with participation coefficients > 0.4 and Z scores > 1 as connector hubs [39]. The LC, exhibiting elevated degree and betweenness centrality as well as relatively higher Z-scores and participation coefficients, may play a pivotal role in mediating isoflurane-induced general anesthesia, given its known involvement in arousal, attention, and analgesia. In contrast, the somatosensory cortex (SS) functions as a connector hub in the KET group, indicating its integrative and coordinating role in ketamine-induced dissociative anesthesia (Figure 7D). The LSc served as the central hub in the saline group (Figure 7C). No significant findings were observed in the within-modal degree Z-score and participation coefficient analyses for the home cage group. However, the analysis did reveal a relatively high degree and betweenness centrality of APNs (Figure 7A), suggesting that the within-module connections are relatively independent, or that the pathways of inter-module information transmission do not depend on specific connector nodes.

Figure 7.

Figure 8.

Discussion

In this study, we conducted a comparative analysis of the effects of two general anesthetics, isoflurane (ISO) and ketamine (KET), on c-Fos expression throughout the brain. By employing principal component analysis (PCA), we were able to thoroughly examine the responses of individual brain regions to each anesthetic agent. Our findings reveal that KET dominantly activates the cerebral cortex yet suppresses subcortical regions, reflecting a top down mechanism of action, while ISO predominantly stimulates subcortical brain regions with relative cortical inhibition, substantiating its bottom up mechanism of action [3, 4]. Further functional analysis of brain networks, based on c-Fos expression, identified the somatosensory cortex (SS) and the locus coeruleus (LC) as central nodes for KET and ISO, respectively, highlighting the crucial roles of LC and SS under ISO and KET induced unconsciousness.

Our results demonstrate that ISO activates the sleep-promoting VLPO nucleus, the aPVT-infralimbic loop [20], and broadly inhibits the cortex, supporting a bottom-up mechanism of ISO induced unconsciousness. Nonetheless, our findings also reveal the activation of arousal related nuclei, such as PB and LC, which implies that the influence of isoflurane on consciousness may not solely rely on suppressing arousal centers, but rather through a more intricate relationship than formerly recognized. Identifying cell types and their dynamic changes during anesthesia will be crucial for clarifying their role in ISO induced unconsciousness.

While c-Fos-based functional network analysis offers lower temporal resolution, its single-cell resolution across the entire brain allows for the inclusion of midbrain and hindbrain regions, supplementing previous fMRI analyses. Our observations reveal that ISO mildly inhibits network density, and through graph theoretical analysis, we identify the LC as a highly connected hub, high-lighting the critical role of the brainstem in ISO induced general anesthesia. The LC performs a wide range of functions in mice, including arousal, pain modulation, attention, stress response, and neuroprotection. Studies have shown that chemical activation of the LC increases whole-brain functional connectivity, attributed to its role as the primary source of norepinephrine (NE) and its extensive influence on nearly the entire brain [40]. The significant LC activation and its central position within the functional network underlying ISO induced unconsciousness suggest that the LC plays a crucial part in maintaining and integrating the entire unconsciousness functional network, emphasizing the involvement of LC in the bottom-up paradigm of ISO induced unconsciousness.

Mashour et al. proposed that anesthesia-induced unconsciousness encompasses not only the modulation of lower-level brain activity but also top down neural processing [3, 4]. Within this top down framework, anesthetics diminish consciousness by interfering with cortical and thalamocortical circuits responsible for neural information integration. Our study discovered that KET administration substantially activated cortical and subcortical arousal-promoting nuclei while concurrently causing relative thalamic suppression, with only the RE and TRS exhibiting activation. This suggests that thalamic inhibition may lead to a reduction in thalamocortical communication, which is characterized by the inability to perceive the external environment and results in disconnection from reality. Graph theoretical analysis also identified the somatosensory cortex (SS) as the hub node of the KET induced functional network. As a critical cortical area, SS is responsible for sensory processing, motor control, and cognitive functions [41]. Previous studies have demonstrated that local KET administration to SS recapitulates the effects of systemic KET on both the switch in pyramidal cell activity and dissociative-like behavior, implying that SS may serve as a key target for KET induced dissociation [42]. Our findings that SS acts as a hub node suggest that KET may modulate brain network function by influencing connectivity between SS and other brain regions, thereby affecting the behavior and cognitive states of mice. This further supports the significance of cortical areas during KET anesthesia.

Identifying shared neural features between KET and ISO is essential for understanding anesthetic-induced unconsciousness. The coactivation of sleep-wake regulation-related regions, such as PL/ILA and aPVT, along with analgesia-related nuclei like CeA, PB, and LC, suggests a shared mechanism for sleep-wake regulation and the common pathways for pain relief. This observation provides valuable insights into the fundamental mechanisms of anesthesia-induced hypnosis and analgesia. Additionally, the coactivation of neuroendocrine-related nuclei, including PVH and SO in the hypothalamus, raises questions about the potential influence of anesthetics on hormonal release and homeostatic regulation. Other coactivated nuclei, such as EW and NTS, warrant further investigation of their roles in anesthesia. In summary, the coactivated nuclei imply a potential shared neuronal circuitry for general anesthesia, encompassing common features like unconsciousness, analgesia, and autonomic regulation, regardless of the specific molecular targets of each drug. Future research could examine coactivated brain regions by the two anesthetics or manipulate identified hub nodes to further understand the mechanisms of general anesthesia. In summary, our study reveals distinct and shared neural mechanisms underlying isoflurane and ketamine anesthesia using c-Fos staining and network analysis. Our findings support “top-down” and “bottom-up” paradigms, and the identification of hub nodes and coactivated brain regions suggests shared neurocircuitry for general anesthesia, providing insights into the mechanisms underlying anesthetic-induced unconsciousness and analgesia.

Methods

Animals

All animal experiments were conducted in accordance with the National Institutes of Health guidelines and were approved by the Chinese Academy of Sciences’ Institute of Neuroscience. Adult male wild-type (WT) mice (C57BL/6J) (8-10 weeks old, weight from 22 to 26 g) were purchased from institute-approved vendors (LingChang Experiment Animal Co., China). Mice were individually housed and maintained on a 12 h:12 h light/dark cycle (lights on at 07:00 a.m. and off at 07:00 p.m.) with food and water available ad libitum.

Drug administration

All experiments occurred between 13:00-14:30 (ZT6–ZT7.5). We adapted mice to handling and the anesthesia chamber (10×15×15 cm) for several days to minimize experimental confound-induced c-Fos induction. Adult male mice were handled for the KET group for 10 min per day with normal saline (NS) injected intraperitoneally (i.p.) for three consecutive days at 13:00. On day five, a randomly chosen mouse received an injection of Ketamine (Gutian Medicine, H35020148), and the control groups (n=8) received the same volume of saline. ISO group (n=6) mice were handled and inhaled 1.5% isoflurane (RWD Life Science, 1903715) at 13:00 on day four in the chamber. Meanwhile, the control groups (n=6) were left undisturbed in their home cages prior to sampling. We confirmed the loss of righting reflex at 5 min after anesthetics exposure. For 90 min after KET injection or ISO inhalation, mice were deeply anesthetized with 5% ISO and transcardially perfused with 35ml 0.1 M phosphate-buffered saline (PBS) followed by 35ml 4% paraformaldehyde (PFA). The brains were then removed and postfixed overnight with 4% PFA. Following fixation, the brains were dehydrated for 48 hours with 30% sucrose (wt/vol) in PBS. Coronal sections (50 µm) of the whole brain were cut using a cryostat (HM525 NX, Thermo Scientific) after being embedded with OCT compound (NEG-50, Thermo Scientific) and freezing.

Immunohistochemistry

One out of every three brain slices (100 µm intervals) of each whole brain was washed three times with 0.1 M phosphate-buffered saline (PBS) for 10 min and permeabilized for 30 minutes at room temperature with 0.3% Triton X-100 in PBS (PBST). Slices were incubated for 2 hours at room temperature with 2% normal donkey serum (Sigma, G6767) in PBS overnight at 4°C with c-Fos primary antibodies (226003, Synaptic Systems; 1:500) diluted in PBS with 1% donkey serum. After three washes with PBST, slices were incubated with the Cy3 donkey anti-rabbit (711165152, Jackson; 1:200) secondary antibody for 2 hours at room temperature. Immunostained slices were mounted with VECTASHIELD mounting medium with DAPI and then scanned under a fluorescent microscope equipped with a 10× objective (VS120, Olympus) or a confocal microscope with a 20× objective (FV300, Olympus).

Quantification of c-Fos positive cells

The procedures used for c-Fos analysis were based on previously research [19]. A custom-written software package was employed for the analysis of brain images. The software consists of four modules: atlas rotation, image registration, signal detection, and quantification/visualization.

Rotation module

The rotation module allows the Allen Mouse Brain Atlas to be rotated in three dimensions at arbitrary angles to fit individual samples. Anatomical landmarks were manually selected to estimate rotation angles. The most posterior slices containing the CA3 in each hemisphere were used to calculate the rotation angle about the left-right axis. The most anterior slice with the anterior commissure crossing the midline and the most posterior slice with the corpus callosum crossing the midline were used to calculate the rotation angle about the dorsal-ventral axis. The reference atlas was then rotated using these estimated angles to match the experimental brain’s sectioning angle.

Registration module

The registration module is an image alignment software that uses reference points to align brain section images with a rotated 3D reference atlas for further quantification. First, reference points in both the atlas and the brain image were chosen. The module then applied geometric transformations to the brain section to optimize the match between the brain image and the atlas’s reference points. After the transformation, the image was compared to the atlas, and any necessary adjustments were made manually.

Detection module

The detection module manually counts the position of c-Fos positive cells in each digitized brain section image.

Principal components analysis (PCA) of the activity patterns at whole brain

We employed arithmetic to calculate the cell density in each brain region:

Nr: The number of c-Fos+ cells in each brain region. Vr: The volume of each brain region. The density of c-Fos from 201 brain regions were concatenated for each mouse (Home cage, n = 6; ISO, n = 6; Saline, n = 8; KET, n = 6). Principal components analysis was performed on the concatenated matrix containing data from four conditions, using singular value decomposition. We then selected the first two principal components (PCs), which accounted for 56.6% of the variance in Figure 2A and 44.5% in Figure 3A.

Network generation

To evaluate how functional connectivity changed under general anesthetics in WT mice, we extracted 63 brain regions from major brain subdivisions (cerebral cortex, cerebral nuclei, thalamus, hypothalamus, midbrain, and hindbrain) listed in Supplementary Table 2. Correlation matrices were generated by computing Pearson correlation coefficients from interregional c-Fos expression in the 63 regions. Mean correlations were calculated to assess changes in functional connectivity between these major subdivisions of the brain. Weighted undirected networks were constructed by considering correlations with Pearson’s r ≥0.82, corresponding to a one-tailed significance level of P<0.05 (uncorrected). The nodes in the networks represent brain regions, and the correlations that survived thresholding were considered connections. Theoretical graph analysis was performed using Brain Connectivity Toolbox (https://sites.google.com/site/bctnet/, version 2019-03-03) in MATLAB R2021 (The MathWorks Inc.) [43]. Network visualization was performed using Cytoscape (version 3.2.1) [44].

Hub identification

Network centrality was evaluated using degree, betweenness, within-module z-scores (representing within-module connectivity), and participation coefficient (indicating between-module connectivity). These measures were computed for all nodes to identify potential hub regions [45]. Modularity was assessed employing Newman’s spectral community detection algorithm [46]. Degree represents the number of edges connected to a node, while betweenness denotes the number of shortest paths traversing a given node. Nodes with elevated betweenness centrality are involved in numerous shortest paths.

Statistical analysis

The sample size was determined based on prior studies [13, 14]. Normality of data distribution was assessed using the Shapiro-Wilk W test. Unpaired Student’s t-test was applied for normally distributed data, while the Mann-Whitney U test was employed for non-normally distributed data. Data are presented as mean ± SEM, with all statistical tests being two-sided. Pearson correlation coefficients (R) were transformed into Z-scores using Fisher’s Z transformation before computing group means and conducting statistical comparisons. GraphPad Prism 9.0 (GraphPad Software, USA) and MATLAB R2021 (Mathworks Inc.) were utilized for statistical analyses. A P-value less than 0.05 was considered statistically significant.

Acknowledgements

We express our gratitude to our interns, Chuhang Wong from Imperial College London and Jiale Huang from ShanghaiTech University, for their assistance in cell counting.

Funding

This study was funded by the NSFC (grants 82271292, 81730031 to Y.W.; 82101350 to M.L.) and the Shanghai Municipal Key Clinical Specialty (grant shslczdzk06901 to Y.W.).

Author contributions

Yue Hu, Conceptualization, Formal analysis, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing; Jiangtao Qi, Conceptualization, Software, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing; Zhao Zhang, Resources, Investigation, Supervision, Writing— review and editing; Mengqiang Luo, Resources, Supervision, Funding acquisition, Project administration, Writing— review and editing; Yingwei Wang, Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing—review and editing.