All the experimental protocols were approved by the Institutional Animal Experimental Ethics Committee of West China Hospital of Sichuan University (Chengdu, China). Animal Research Reporting In Vivo Experiments (ARRIVE) guidelines were applied during the study. Adult C57 BL/6J mice (8 weeks old) were housed in humidity- and temperature-controlled cages, underwent a 12-h light–dark cycle (light on from 7 a.m. to 7 p.m.), and had ad libitum access to chow and water. Animals were euthanized using CO₂ and isoflurane after experiments.

Minimum alveolar concentrations (MACs) of isoflurane that induced loss of righting reflex (LORR) and loss of movements to tail-clamping stimulus (LOM) were measured, respectively, as MAC_LORR and MAC_LOM (Fukagawa et al., 2014; Zhou et al., 2015). Isoflurane was delivered in an open-circuit rodent anesthetic system. Carrier gas flow was 2 L/min (100% O₂). For the measurement of LORR, each mouse was placed in a cylindrical anesthetic chamber (15 cm in length and 5 cm in diameter). Isoflurane concentrations were increased starting from 0.6% in a step interval of 0.05%. As the plastic chambers were rotated slowly, LORR was defined as the mouse loss ability to turn itself prone onto all four limbs in 60 s. For the measurement of LOM, mice were placed in a cuboid container (25 cm in length, 15 cm in width, and 12 cm in height) and left the tail outside. Isoflurane concentrations were increased from 0.9% in a step interval of 0.05%. LOM was defined as no purposeful movements to tail-clamping stimulation in 30 s.

Four hand-made EEG electrodes were implanted in each animal, including two recording electrodes, a common electrode, and a grounding electrode. The cortical EEG electrode, common electrode, and grounding electrode were anchor screws (0.6 mm in diameter and 1.5 mm in length), which were fixed on the skull and inserted into the cortex. The thalamic electrode was made by insulating silver wire (0.2 mm in diameter) and connected with an anchor screw. The thalamic electrode was made by insulating silver wire (0.2 mm in diameter). The insulation was removed from one end of the wire, and the other end was cut with a conductive tip for intracranial insertion (Guidera et al., 2017). When implanting EEG screws and thalamic electrodes, mice were anesthetized with 2% isoflurane and placed in a stereotactic apparatus (RWD, Shenzhen, China). A heating pad (RWD, Shenzhen, China) was used to maintain the body temperature of mice. A longitudinal incision was made to expose the bregma and lambda of the skull. The cortical EEG electrode was implanted to primary motor cortex (M1) [anteroposterior (AP): +1.0 mm; mediolateral (ML): +1.5 mm; and dorsoventral (DV): –1.5 mm] to record the frontal cortex EEG. The thalamic electrode was inserted into ventroposterolateral and/or ventroposteromedial thalamic nuclei (VPL/VPM) [AP: –1.5 mm; ML: +1.5 mm; and DV: –3.5 mm], which was close to relay cells to record thalamic LFPs (Landisman and Connors, 2007; Haidarliu et al., 2008) (Supplementary Figure S1). The common electrode was implanted in the left frontal cortex [AP: +1.0 mm; ML: –1.5 mm; and DV: –1.5 mm]. The grounding electrode was planted in the parietooccipital cortex [AP: –3.5 mm; ML: –1.5 mm; and DV: –1.5 mm] to eliminate the disturbance of the surrounding environment. Four electrodes were connected to a miniature plug with silver wires. Then, the electrodes and the plug were secured to the skull with dental acrylic. After the surgery, animals were allowed to recover for 1 week before recordings.

Electrophysiological signals (frontal cortical EEG and/or thalamic LFPs) were recorded by a Pinnacle EEG recording system (Part #8200-SL; Pinnacle Technology, United States) (Gelegen et al., 2021). A preamplifier unit was rigidly attached to the miniature plug, providing the first stage of amplification (×100) and initial high-pass filtering (first-order 0.5 Hz for EEG). The recording sampling frequency of signals was 500 Hz. Raw signals were preamplified, digitized, and recorded using a Sirenia Acquisition system (Part #8206-SL; Pinnacle Technology, United States) and analyzed using MATLAB (version 2006a; MathWorks, United States). The accuracy of the electrode placement was confirmed by visual examination of brain tissue in postmortem.

For the recordings under isoflurane, each mouse was placed in a transparent gastight plastic chamber (20 cm in length, 15 cm in width, and 10 cm in height). A heating pad (RWD, Shenzhen, China) was used to maintain the body temperature of mice. The chamber was ventilated with 2 L/min 100% O₂, and the EEG and LFPs were recorded for 5 min as baseline. Then, the concentrations of isoflurane were continuously increased from 0.9% to 1.5% and then to 2.0% and decreased to 1.5, 0.9, and 0.5%. The concentrations of isoflurane were continuously monitored by an infrared gas monitor (Datex-Ohmeda, WI, United States). Each concentration of isoflurane was maintained for at least 20 min. Frontal cortical EEG and thalamic LFPs were recorded continuously. For the recordings under propofol, propofol (AstraZeneca SpA, London, United Kingdom) (Ou et al., 2017) was intravenously injected into the caudal vein of mice at a dose of 14 mg/kg. After injection, the duration of LORR was recorded. The definition of LORR was the same as described in isoflurane anesthesia. Cortical EEG and thalamic LFPs were continuously monitored throughout.

Power spectra were estimated using multitaper spectral methods implemented in the Chronux toolbox with the “mtspectrumc” or “spectrograms” function in MATLAB (version 2006a; MathWorks, United States) (Bokil et al., 2007). Spectrograms were computed using the continuously recorded EEG signals from 0.5 to 25 Hz. The spectral difference at differential isoflurane concentrations was calculated for particular EEG epochs. Typical frontal cortical EEG and thalamic LFPs were analyzed at differential isoflurane concentrations for the duration of 2 min. For baseline EEG and/or LFPs, stable signals without artifacts were chosen. For each concentration of isoflurane, the signals were chosen at 10 min after stabilization of isoflurane at this concentration. For propofol, 30-s duration signals were chosen after LORR had occurred. The time course of the relative power of slow (0.5–1 Hz), delta (1–5 Hz), and alpha (9–14 Hz) frequency bands was extracted. The absolute EEG power of each 3-s bin was normalized to the sum of the power over the entire analysis range (0.5–25 Hz) as previously described (Gui et al., 2021).

The coherence represents the correlation between two signals at a unique frequency (Bokil et al., 2007; Akeju et al., 2014). Coherograms between thalamic LFPs and cortical EEG were computed using the multitaper method (Bokil et al., 2007) based on the continuously recorded electrophysiological signals. Coherence values were calculated for particular epochs (the same as above described for spectral analysis), and the values were averaged for all animals. Jackknife techniques were used to determine 95% confidence interval (CI) of coherence values (Bokil et al., 2007). The parameters for coherence analysis were set as follows: window length, T = 4 s with 0-s overlap; time–oscillation width product, TW = 3; number of tapers, K = 5; and a spectral resolution of 2 W of 1.5 Hz.

To characterize the coupling between the phase of slow oscillation (0.5–1 Hz) and the amplitude of alpha oscillation (9–14 Hz), we constructed a time-varying phase–amplitude modulogram, which indicated the relative alpha amplitude of a particular phase at each slow oscillation cycle (Mukamel et al., 2011; Purdon et al., 2013). A wavelet packet transform was applied to construct slow and alpha signal narrow oscillations. Then, the Hilbert transform was applied to the signals and low-frequency phase and alpha oscillation amplitudes were computed. Modulogram was computed using the continuously recorded signals. The modulation index (MI) was calculated to quantify the strength of modulation (Mukamel et al., 2011). Signals with the duration of 600 s were selected at 10 min after stabilization of each isoflurane concentration to calculate the total proportions of statistically significant epochs of MI.

A computational model based on the DynaSim MATLAB toolbox (Sherfey et al., 2018) was used to simulate the isoflurane-induced alpha spiking in thalamocortical neurons. The model was based on a thalamic, Hodgkin–Huxley network, which simulates firing frequency of 50 thalamocortical cells (TCs) and 50 reticular single-compartment cells (RE) coupled to each other (the code is available on GitHub, https://github.com/asoplata/propofol-coupling-2017-full) (Soplata et al., 2017). Three paraments were adjusted in this model according to the effects of isoflurane on a single TC neuron: I_GABA-A (current of GABA), I_AMPA (current of AMPA), and gK_Leak (conductance of background potassium channels). The changes in I_GABA-A and I_AMPA induced by isoflurane (Supplementary Table S1) were demonstrated in previous studies (Jenkins et al., 1999; de Sousa et al., 2000; Sebel et al., 2006). The effect of isoflurane on gK_Leak was based on our results of the whole-cell patch-clamping recordings in acute brain slices (Supplementary Figure S1; Supplementary Table S1).

Processed EEG and LFP data were exported from MATLAB (version 2006a; MathWorks, United States) and analyzed by the GraphPad Prism 8.0 software (GraphPad, United States). Data were presented as means with 95% CI intervals. The sample size of animals was calculated by the test for paired means with the PASS 15 software (NCSS, LLC, Kaysville, UT, United States). By the preliminary test (n = 4) on EEG changes of alpha oscillations between baseline and 1.0 MAC_LOM isoflurane, it was observed that the isoflurane enhanced relative alpha power from 12.94% [7.95%, 17.92%] to 25.47% [13.17%, 37.77%]. Therefore, the calculated minimal sample size was 8.0 (α = 0.05; power = 0.90), and a sample size of 10 was chosen. Spectral and coherence differences were evaluated by the one-way repeated-measures analysis of variance (ANOVA) with a Greenhouse–Geisser correction. Followed by the repeated-measures ANOVA, Tukey’s post hoc test was used to determine the difference between baseline and anesthetic states, or between various anesthetic depths. For PAC analysis, a permutation test was conducted (Mukamel et al., 2011). Electrophysiologic data were analyzed using the software packages pClamp 10.2 (Molecular Devices, United States), GraphPad Prism 8.0 (GraphPad, United States), and SPSS 22.0 (IBM, United States). The exact statistical methods employed are indicated in the figure legends, and a statistical significance was deemed to be a p-value < 0.05.

Concentration–response curves of isoflurane (n = 20; Figure 1A) indicated that MAC_LORR was 0.84% [0.83%, 0.85%] and MAC_LOM was 1.09% [1.06%, 1.13%]. Thus, we defined 0.9% isoflurane as light anesthesia (∼1.0 MAC_LORR), 1.5% as surgery anesthesia (∼1.3 MAC_LOM), and 2.0% as deep anesthesia (∼2.0 MAC_LOM). By the repeated-measures ANOVA, it was found that there were significant differences in the relative power of slow oscillations (_rANOVA F _2.035,18.32 = 3.848, p = 0.039; Figure 1F), delta oscillations (_rANOVA F _2.09,18.78 = 11.48, p = 0.0005; Figure 1F), and alpha oscillations (_rANOVA F _{3.15, 28.37} = 12.16, p < 0.0001; Figure 1F) between baseline and anesthetic states. Isoflurane at 0.9% significantly increased the relative power of slow oscillation (p = 0.037 by Tukey’s post hoc analysis; isoflurane vs. baseline, 15.77% [10.40%, 21.15%] vs. 7.46% [5.92%, 9.01%], n = 10; Figure 1E left, Figure 1F and Table 1) and delta oscillation (p = 0.0027 by Tukey’s post hoc analysis; 0.9% isoflurane vs. baseline, 56.43% [50.84%, 62.03%] vs. 44.58% [39.96%, 49.20%], n = 10; Figure 1E left, Figure 1F and Table 1). 1.5% and 2.0% isoflurane significantly enhanced the relative power of alpha oscillations (p = 0.032 by Tukey’s post hoc analysis; 1.5% isoflurane vs. baseline, 20.02% [14.15%, 25.88%] vs. 9.64% [6.86%, 12.43%], n = 10; Figure 1E middle, Figure 1F and Table 1: p = 0.029 by Tukey’s post hoc analysis; 2.0% isoflurane vs. baseline, 20.92% [15.87%, 25.97%] vs. 9.64% [6.86%, 12.43%], n = 10; Figure 1E right, Figure 1F and Table 1). In conclusion, the relative power of cortical slow and delta oscillations increased at 0.9% isoflurane, and the relative power of cortical alpha oscillation was increased at 1.5% and 2.0% isoflurane. As a positive control, when compared to baseline, the cortical alpha power increased during propofol-induced LORR (p = 0.007 by a two-tailed paired t-test; propofol vs. baseline, 459.65 [133.71, 785.62] μV²/Hz vs. 1548.24 [562.83, 2534.54] μV²/Hz, n = 10; Figures 4B,C).

By the repeated-measures ANOVA, it was found that there were no significant differences in the relative power of slow oscillations (_rANOVA F _1.64,16.44 = 3.17, p = 0.076; Figure 2D), but there were significant differences in delta (_rANOVA F _3.33,33.28 = 9.10, p < 0.0001; Figure 2D) and alpha (_rANOVA F _{2.65, 26.54} = 18.17, p < 0.0001; Figure 2D) oscillations between baseline and anesthetic states. Isoflurane at 0.9% significantly increased the relative power of delta oscillations (p = 0.049 by Tukey’s post hoc analysis; 0.9% isoflurane vs. baseline, 49.18% [42.04%, 56.32%] vs. 37.39% [32.73%, 42.05%], n = 10; Figure 2C left, Figure 2D and Table 1). Isoflurane at 1.5% and/or 2.0% significantly enhanced the relative power of alpha oscillations (p = 0.004 by Tukey’s post hoc analysis; 1.5% isoflurane vs. baseline, 16.38% [12.83%, 19.91%] vs. 9.20% [7.40%, 11.00%], n = 10; Figure 2C middle, Figure 2D and Table 1; p = 0.0007 by Tukey’s post hoc analysis; 2.0% isoflurane vs. baseline, 21.32% [17.26%, 25.39%] vs. 9.20% [7.40%, 11.00%], n = 10; Figure 2C right, Figure 2D and Table 1). The relative power density of slow, delta, and alpha oscillations over all experimental conditions is shown in Figure 2D. In conclusion, 0.9% isoflurane increased the relative power of thalamic delta oscillation, and 1.5% and 2.0% isoflurane increased the relative power of thalamic alpha oscillation. The dynamics of thalamic LFPs were highly similar to EEG of the prefrontal cortex. As a positive control, the propofol-induced LORR increased the relative thalamic alpha power when compared to baseline (p = 0.02 by a two-tailed paired t-test; propofol vs. baseline; 3985.04 [1143.32, 6826.34] μV²/Hz vs. 2054.02 [−78.20, 4285.22] μV²/Hz, n = 10; Figures 4E,F).

By the repeated-measures ANOVA, it was found that there was no statistical difference in coherence between cortical EEG and thalamic LFPs in slow band (_rANOVA F _3.34,33.38 = 1.83, p = 0.16; Figures 3C,D) or delta band (_rANOVA F _3.45,34.53 = 0.38, p = 0.80; Figures 3C,D), but there was a significant difference in alpha band (_rANOVA F _{2.70, 26.95} = 4.15, p = 0.018; Figures 3C,D). Isoflurane at 1.5% enhanced the coherence of alpha oscillations between cortical EEG and thalamic LFPs (p = 0.028 by Tukey’s post hoc analysis; 1.5% isoflurane vs. baseline, 0.70 [0.63, 0.77] vs. 0.63 [0.58, 0.68], n = 10; Figure 3C middle, Figure 3D and Table 1). As a positive control, the propofol-induced LORR increased the coherence of alpha oscillation between cortical EEG and thalamic LFPs (p = 0.02 by a two-tailed paired t-test; propofol vs. baseline, 0.58 [0.47, 0.69] vs. 0.48 [0.33, 0.62], n = 10; Figure 4H,I).

The phase and amplitude of slow and alpha oscillations filtered from cortical EEG (Figure 5A) were extracted by the Hilbert transform (Figure 5B) and were calculated by the time-dependent modulogram (Figure 5C). A significant slow–alpha PAC within the cortex was observed under 1.5% isoflurane (p < 0.05 in 65.5 percent of total duration of 220 min by a permutation test; Figures 5D–H and Table 1), while 2.0% isoflurane disrupted the PAC (p < 0.05 in 34.5 percent of total duration of 220 min by a permutation test; Figures 5D–H and Table 1). The cortical alpha amplitude and thalamic slow phase (Figure 5F) were also calculated using the time-dependent modulogram. A significant slow–alpha phase–amplitude modulation between the cortex and the thalamus was observed under 1.5% isoflurane (p < 0.05 in 58.0 percent of total duration of 220 min by a permutation test; Figures 5G,H), while 2.0% isoflurane disrupted the PAC (p < 0.05 in 22.1 percent of total duration of 220 min by a permutation test; Figures 5G,H).

The effects of isoflurane on K_Leak current were measured by the whole-cell patch-clamping recordings in TC (Supplementary Figure S2A). TC neurons displayed a weakly rectifying current–voltage profile, and reversal potential was near E_K (approximately –90 mV), as expected for a K_Leak current (Supplementary Figure S2B) (Lazarenko et al., 2010). An isoflurane-enhanced K_Leak current (Supplementary Figure S2C, top) was associated with a net increase in conductance (Supplementary Figure S2C, bottom). 0.26–0.30 mM isoflurane solution (∼1.0 MAC_LOM) increased K_Leak conductance (gK_Leak) (_rANOVA F_{1.5, 7.7} = 8.4, p = 0.01; p = 0.009 by Tukey’s post hoc analysis; control vs. 1.0 MAC_LOM isoflurane, 30.40 nS [4.48 nS, 56.32 nS] vs. 39.04 nS [4.52 nS, 73.56 nS], n = 7; Supplementary Figure S2D). 0.42–0.50 mM isoflurane solution (∼1.5 MAC_LOM) increased K_Leak conductance (_rANOVA F_{1.2, 8.7} = 8.4, p = 0.002; p = 0.008 by Tukey’s post hoc analysis; control vs. 1.5 MAC_LOM isoflurane, 27.03 nS [14.10 nS, 39.94 nS] vs. 39.29 nS [26.54 nS, 52.04 nS], n = 7; Supplementary Figure S2E). Therefore, 0.26–0.30 mM (∼1.0 MAC_LOM) and 0.42–0.50 mM (∼1.5 MAC_LOM) isoflurane solution enhanced gK_Leak by 25.86% [10.91%, 40.81%] and 47.10% [27.98%, 66.22%], respectively (Supplementary Figure S2F).

To investigate the preliminary mechanism of how isoflurane induced alpha spiking in the thalamic network, a in silico thalamic model was used to simulate the effects of isoflurane on thalamic rhythms (Soplata et al., 2017). Briefly, AMPA, GABA_A, and K_Leak were the main molecular targets in this model. Therefore, the effects of isoflurane on AMPA, GABA_A, and K_Leak were applied in simulation. The effects of isoflurane on AMPA (de Sousa et al., 2000) and GABA_A (Sebel et al., 2006) were demonstrated in previous studies. The effect of isoflurane on gK_Leak was based on our results of the whole-cell patch-clamping recordings on acute brain slices (Supplementary Figure S2). When the potentiation of GABA_A was up to ∼twofold more than the baseline level, the frequency of the TC spikes increased to alpha oscillations (Figure 6A). If the contribution of GABA_A was ignored, the TC spikes decreased along with the potentiation of gK_Leak, while the inhibition of I_AMPA slightly modulated the network frequency (Figure 6B). When the potentiation of I_GABA-A was adjusted to the levels of 1.3% isoflurane (∼1.7-fold potentiation) (Figure 6C) and 2.0% isoflurane (∼twofold potentiation) (Figure 6D), respectively, the TC spikes decreased with the potentiation of gK_Leak (Figure 6B) and were nearly stable with the inhibition of I_AMPA.

On the cellular level, thalamic TC neurons did not display any intrinsic oscillatory activity at the baseline level (Figure 6E). If the overall modulations of isoflurane on GABA_A, gK_Leak, and AMPA were applied simultaneously, isoflurane at both ∼1.3% (Figure 6F) and ∼2.0% (Figure 6G) induced persistent alpha firing of TC neurons. On the network level, no alpha oscillations were induced in a thalamic rhythm at the baseline condition (Figure 6I). With the potentiation of GABA_A effect under ∼1.3% (Figure 6J) and ∼2.0% isoflurane (Figure 6K), respectively, alpha oscillations were induced along with the modulation of gK_Leak and I_AMPA. Isoflurane at both ∼1.3% (Figure 6J) and ∼2.0% (Figure 6K) induced persistent alpha oscillations in thalamic rhythm. In summary, the overall modulation of isoflurane on I_AMPA, I_GABA-A current, and gK_Leak may contribute to alpha spiking of TC neurons under surgery and/or deep anesthesia under isoflurane, which may involve in a thalamic alpha rhythm.