We utilized adult male C57BL/6 J mice (3–4 months old) for the electrophysiological surgery in this study. Prior to the surgery, mice were group-housed, and those implanted with tetrodes were individually housed. The mice received adequate food and water under a 12/12 h light/dark cycle (lights at 8:00 a.m.), maintaining a constant environment at 21–24°C with 50–60% relative humidity. All procedures adhered to institutional animal welfare guidelines and were approved by the Third Military Medical University Animal Care and Use Committee.

Tetrodes were crafted from 25-μm-diameter insulated tungsten wire (California Fine Wire, CFW2002936) (Yang et al., 2023). In brief, a 30-cm-long tungsten wire was folded twice and twisted using an electromagnetic stirrer (Apera Instruments, 801 Magnetic Stirrer) to form a tetrode. The tetrode was then vertically aligned and securely fused by melting the insulation with a heat gun (450°C, 1 min). A 1.2-cm-long silica capillary tube (Polymicro Technologies, TSP100170; ID, 100 μm; OD, 164 μm) was applied and affixed with super glue to the tetrode tip for protection and support. Four tetrodes were arranged in a line with a 200 μm spacing, connecting the other end of the 16 electrodes (4 × 4 tetrodes) to a 16-channel connector (Omnetics, A79016-001) via gold pins (Neuralynx, small EIB pins).

A microdrive, composed of a metal rod, three plastic blocks, and a screw-nut pair, facilitated vertical movement of the tetrodes. The plastic blocks were threaded onto the metal rod and screw, with two blocks secured to the ends using super glue and the third block positioned in the middle. The previously prepared tetrodes were fastened to the middle block, allowing for vertical movement when the screw was turned.

Mice were anesthetized with isoflurane in oxygen (3% induction, maintained at 1–2%, 0.5 L/min). Placed in a stereotactic apparatus with a heating pad maintaining a ~ 37°C temperature, an incision was made to expose the skull at the top of the head after hair removal. A small craniotomy (0.5 × 0.5 mm) was performed above dorsal CA1 (AP: −1.85 mm, ML: 1.10 mm, Figure 1). The prepared tetrodes were inserted through the craniotomy to a depth of 0.9 mm (for CA1 recording) or 1.7 mm (for DG recording). EEG–EMG electrodes were implanted for recording signals. Two stainless screws were inserted into craniotomy holes above the frontal lobe for EEG recording, and two EMG electrodes were inserted into the neck muscles for EMG recording. Dental cement secured the tetrodes and EEG–EMG electrodes to the skull. Post-surgery, mice were returned to warm cages for full recovery and received intraperitoneal injections of dexamethasone sodium phosphate (1 mg/mL, 0.1 mL/10 g/d) and ceftriaxone sodium (50 mg/mL, 0.1 mL/10 g/d) for three consecutive days to minimize inflammation.

Following complete recovery from surgery, as evidenced by an increase in mouse weight, we conducted electrophysiological recordings (Figure 2A). Tetrodes were progressively advanced to a final depth of 1.1 mm for CA1 pyramidal cell recording or 1.85 mm for DG granule cell recording. Recordings were performed during the light phase after mice had been acclimated to the recording cables for 2–3 days. Electrophysiological and EEG–EMG signals were acquired at 20 KHz using an RHD2000 USB board (Intan Technology, C3100) and amplified by a 16-channel digital amplifier (Intan Technology, C3334). Simultaneously, behavioral videos were recorded at 25 Hz. Recordings initially spanned at least 2 sleep-wakefulness cycles (Figure 3A). Subsequently, mice were placed in a chamber for a 5–15-min recording during freely exploring states. Following this, recordings under isoflurane anesthesia were conducted, with mice positioned in a self-made horizontal Plexiglas® cylinder (30 cm long, 15 cm in diameter, Figure 4A). A continuous delivery of 1.4% isoflurane in air (2.5 L/min) was maintained for 30 min. The isoflurane was then discontinued, and only air was delivered continuously. Approximately half an hour after anesthesia recovery, another 5–15-min recording during freely exploring states was performed. Following all recordings, electrical lesions (20 μA for 15 s, DC currents) were applied to verify the recording sites.

The assessment of anesthetic effects on consciousness relied on the righting reflex test (Solt et al., 2011; Pal et al., 2018) and specific EEG signals (Shander et al., 2018) as identified in previous studies. The anesthetic procedure was categorized into five stages based on these approaches. Isoflurane anesthesia was administered in a horizontal Plexiglas® cylinder, as previously described. During isoflurane delivery, the cylinder was manually rotated 180° every 15 s to place the mice on their backs. The time at which mice lost their ability to right themselves within 30 s was defined as the occurrence of the loss of the righting reflex (LORR). The period from the onset of isoflurane exposure to the occurrence of LORR was termed the induction period (284.3 ± 15.7 s in our dataset). Burst suppression (BS) describes an EEG pattern with a continuous alternating appearance of mixed-frequency activity and inactivity periods (Lewis et al., 2013). The period from the occurrence of LORR to the occurrence of BS was defined as the sedation period (107.4 ± 16.4 s in our dataset). After isoflurane shut off, the BS activity persisted and gradually disappeared. The period from the occurrence of BS activity to its disappearance was termed the BS period. Subsequently, mice regained the ability to right themselves as anesthesia gradually wore off. The period from the disappearance of BS activity to the recovery of the righting reflex (RORR) was defined as the emergence period. Lastly, a 5–10-min period from the occurrence of RORR was labeled as post_RORR. Additionally, wake_pre and wake_post represented the awake state before and after anesthesia.

48 h post-electrical lesions, mice underwent intraperitoneal injection of pentobarbital for anesthesia, followed by sequential perfusion with saline and 4% paraformaldehyde (PFA) in PBS. The brain was then extracted and immersed in 15% sucrose PBS for overnight dehydration. Subsequently, coronal sections of 50 μm thickness were obtained from the brains. These sections were stained with DAPI (4′,6-diamidino-2-phenylindole, 1,10,000, D9564, Sigma-Aldrich) and imaged using a wide-field fluorescence microscope (Nikon, ECLIPSE Ni) to visualize the recording sites.

We implanted self-made tetrodes into the CA1 and DG regions to record neuronal activities (Figures 1A,D). Post-recording, tetrode locations were confirmed through histological images (Figures 1B,E). The white traces induced by electrical lesions (red triangles in Figures 1B,E) verified the tetrode tips’ placement in the CA1 pyramidal cell layer and the DG granule cell layer. The positions of all tips of the tetrodes were reconstructed in the mouse brain atlas. The recording sites for CA1 were mainly in the anterior intermediate part of the dorsal CA1 pyramidal layer (Figure 1C). For DG recordings, we had 28 tetrodes in the suprapyramidal region, 16 tetrodes in the infrapyramidal region, and the remaining 4 in the hillus (Figure 1F).

To assess the recordings in hippocampus, we made a comparison of the signals recorded from different tetrodes in freely behaving mice (Figure 2A). The cross correlations and time lags, obtained through lagged cross correlation analysis (Figure 2B), between different tetrodes were used for quantitative comparison. The low-frequency oscillations, including delta, theta, alpha and beta bands, were compared separately. We found that the cross correlation coefficient decreased with the increase of tetrode distance in both CA1(Figure 2C, Kruskal-Wallis test; delta, H = 17.0, p < 0.001; theta, H = 12.2, p = 0.002; alpha, H = 14.5, p < 0.001; beta, H = 15.6, p < 0.001) and DG (Figure 2E, Kruskal-Wallis test; delta, H = 7.2, p = 0.03; theta, H = 12.0, p = 0.002; alpha, H = 19.3, p < 0.001; beta, H = 22.1, p < 0.001). Conversely, the time lag increased with the increase of tetrode distance in theta, alpha and beta bands in both CA1 (Figure 2D, Kruskal-Wallis test; theta, H = 7.5, p = 0.02; alpha, H = 9.8, p = 0.007; beta, H = 9.7, p = 0.008) and DG (Figure 2F, Kruskal-Wallis test; theta, H = 7.8, p = 0.02; alpha, H = 6.5, p = 0.04; beta, H = 9.4, p = 0.009). These results, which were consistent with previous study (Lubenov and Siapas, 2009), and along with our histological results, indicated a reliable recording in CA1 and DG.

Previous studies have shown that general anesthesia can influence memory and cognition (Skvarc et al., 2018; Tasbihgou and Absalom, 2021), while natural sleep, a state of decreased arousal, is essential for memory consolidation (Diekelmann and Born, 2010). To compare the different impacts of natural sleep and anesthesia on CA1 and DG, we initially conducted recordings in both regions during natural sleep. Electrophysiological data were categorized into LFP (low-pass filtered at 250 Hz) and SUA (high-pass filtered at 250 Hz) based on frequency and synchronized with EEG–EMG signals and behavioral videos for analysis (Figure 3A).

We calculated and compared the LFP power spectral densities in CA1 and DG across delta, theta, alpha, and beta bands during NREM sleep, REM sleep, and wakefulness (Figures 3B–D). Distinct differences in LFPs between CA1 and DG were observed throughout the sleep-wakefulness cycle. During NREM sleep, a significant discrepancy was noted in the beta band of normalized LFP power (Figure 3B right; CA1, 0.30 ± 0.005 versus DG, 0.27 ± 0.008; Wilcoxon signed-rank test, p = 0.01). Differences in the theta (CA1, 0.41 ± 0.016 versus DG, 0.36 ± 0.016; unpaired t-test, p = 0.01) and alpha (CA1, 0.08 ± 0.002 versus DG, 0.1 ± 0.004; Wilcoxon signed-rank test, p = 0.01) bands were observed during REM sleep (Figure 3C). Additionally, LFP powers in both CA1 and DG were primarily concentrated in the delta band during NREM sleep, dominated in the theta band during REM sleep and increased in the beta band during wakefulness (Figures 3E,F).

Subsequently, we analyzed SUA in CA1 and DG during sleep-wakefulness cycles. In total, we recorded 189 putative excitatory neurons, 17 putative interneurons in CA1, and 125 putative excitatory neurons, 13 putative interneurons in DG (Figure 3G). The firing rates of DG excitatory neurons were significantly higher than those of CA1 excitatory neurons in both NREM sleep (Figure 3H top; DG, 3.18 ± 0.25 Hz; CA1, 2.52 ± 0.20 Hz; Wilcoxon signed-rank, p = 0.002) and REM sleep (DG, 3.87 ± 0.35 Hz; CA1, 2.90 ± 0.27 Hz; Wilcoxon signed-rank test, p = 0.001), but showed no difference during wakefulness (DG, 2.47 ± 0.26 Hz; CA1, 3.90 ± 0.43 Hz; Wilcoxon signed-rank test, p = 0.97). Moreover, the firing rates of both CA1 and DG excitatory neurons had no significant different across the three sleep-wakefulness states (Friedman test; CA1, χ² = 5.0, p = 0.08; DG, χ² = 5.6, p = 0.06).

However, the firing rates of CA1 interneurons were higher than those of DG interneurons in wakefulness (Figure 3H bottom; CA1, 42.4 ± 9.1 Hz; DG, 12.3 ± 2.3 Hz; Wilcoxon signed-rank, p = 0.001), but not in NREM sleep (CA1, 22.8 ± 3.3 Hz; DG, 16.9 ± 1.5 Hz; Wilcoxon signed-rank, p = 0.54) or REM sleep (CA1, 31.9 ± 3.7 Hz; DG, 27.6 ± 2.4 Hz; Wilcoxon signed-rank, p = 0.62). And the firing rates of CA1 interneuron were lower in NREM sleep (Friedman test, χ² = 8.8, p = 0.01; NREM versus Wake, p = 0.03, NREM versus REM, p = 0.03), while the DG interneurons showed an increase in firing rates in REM sleep (Friedman test, χ² = 17.2, p < 0.001; REM versus Wake, p < 0.001, REM versus NREM, p = 0.005).

To examine the activities of CA1 and DG during anesthesia, we conducted tetrode recordings in mice exposed to isoflurane-induced general anesthesia (Figure 4A). The peak frequency of the theta band shifted to a lower frequency during the induction stage compared to wakefulness (Figures 4B,C; CA1, 6.21 ± 0.08 Hz versus 8.12 ± 0.09 Hz, p < 0.001; DG, 5.51 ± 0.05 Hz versus 7.62 ± 0.03 Hz; paired t-test, p < 0.001, see stage definition in Methods). Furthermore, the LFP theta activities reduced further after the mice lost their righting reflex in subsequent sedation, BS, and emergence stages (Figures 4D–F). During the recovery from anesthesia, the peak frequencies of LFP delta activities in CA1 initially increased after the mice regained their righting reflex (Figures 4F,G; emergence stage, 2.16 ± 0.03 Hz versus post_RORR stage, 2.69 ± 0.06 Hz; paired t-test, p < 0.001). Subsequently, the LFP theta activities gradually recovered when mice resumed locomotion in the wake_post stage (Figure 4H).

We compared the differences in LFP power between CA1 and DG across different anesthesia stages. In the induction stage, the normalized power of the theta band in CA1 was stronger than that in DG (Figure 4C; CA1, 0.26 ± 0.010 versus DG, 0.24 ± 0.008; unpaired t-test, p = 0.01). No significant differences in normalized LFP powers were observed between CA1 and DG in sedation and BS stages (Figures 4D,E). However, significant differences in the delta band (Figure 4F; CA1, 0.24 ± 0.02 versus DG, 0.19 ± 0.01; Wilcoxon signed-rank test, p = 0.04) of the emergence stage and in the alpha band (Figure 4G; CA1, 0.09 ± 0.002 versus DG, 0.10 ± 0.002; unpaired t-test, p = 0.03) of the post_RORR stage were observed. After the mice recovered from anesthesia in the wake_post stage, CA1 LFP exhibited higher theta power (CA1, 0.28 ± 0.01 versus DG, 0.24 ± 0.01; unpaired t-test, p = 0.01) and lower delta power (CA1, 0.17 ± 0.01 Hz versus DG, 0.21 ± 0.01 Hz; unpaired t-test, p = 0.01) than DG LFP.

We then compared the changes in normalized power in different LFP bands throughout the entire isoflurane-induced general anesthesia process (Figures 4I,J). For the delta band in both CA1 and DG LFPs, the power gradually increased and maximized in the BS stage. Subsequently, it started to decrease after isoflurane was turned off and gradually recovered to the pre-anesthesia level (Friedman test; CA1, χ² = 31.6, p < 0.001; DG, χ² = 49.9, p < 0.001). Regarding the theta band, the normalized power began to decrease after isoflurane delivery and remained low in the subsequent sedation, BS, emergence, and post_RORR stages in both CA1 (Friedman test, χ² = 39.4, p < 0.001) and DG (Friedman test, χ² = 35.6, p < 0.001). However, CA1 theta power recovered faster than DG. For the alpha and beta bands, which has been reported to originate from the somatostatin neurons in amygdala (Jackson et al., 2024), their normalized power decreased only in the BS stage in both CA1 and DG.

Previous study has suggested that hippocampal LFP phase was modulated to slow-wave activity during anesthesia (Yang et al., 2021). We then examined the cross-frequency coupling between the phase of slow LFP oscillations and the amplitude of gamma oscillation across different sleep and anesthesia states by doing phase-amplitude coupling (PAC, Figure 5A). During REM sleep, a clear coupling between theta and gamma bands was found, while an additional coupling between delta and gamma bands was observed during NREM sleep (Figure 5B). After isoflurane delivery, the coupling phase shifted from theta to lower frequency gradually (Figure 5B wake_pre, induction and sedation). And after isoflurane shut off, the coupling phase increased from delta band into theta band (Figure 5B emergence, post_RORR and wake_post).

Sharp-wave ripple and dentate spike, which happen in CA1 and DG, respectively, during awake immobility and NREM sleep, are two typical hippocampal LFP activities (Farrell et al., 2024). We additionally checked these two activities in our datasets. We found that both ripples and dentate spikes occurred frequently during NREM sleep, disappeared with isoflurane delivery, and were regenerated after recovery from anesthesia (Figures 6A,B).

We conducted an analysis to characterize the spike firing patterns of CA1 and DG neurons at the single-cell resolution throughout general anesthesia by examining the busty index and firing rate. The spike-time autocorrelogram, ISI histogram and waveform of typical bursty, non-bursty excitatory neuron and interneuron were shown in Figure 7A. A total of 206 CA1 neurons (189 putative excitatory neurons, 17 putative interneurons) and 138 DG neurons (125 putative excitatory neurons, 13 putative interneurons) were recorded. The bursty index of both CA1 (Figure 7B, n = 189, Friedman test, χ² = 317.1, p < 0.001) and DG (Figure 7C, n = 125, Friedman test, χ² = 193.7, p < 0.001) excitatory neurons reduced with the isoflurane delivery, and increased with the recovery from anesthesia. The firing sequences of all recorded neurons across anesthesia stages were depicted in Figure 7D. The average firing rates of putative excitatory neurons in both CA1 and DG gradually reduced with the deepening of anesthesia, followed by a gradual recovery after the cessation of isoflurane (Figures 7D,E, Friedman test; CA1, χ² = 678, p < 0.001; DG, χ² = 415; p < 0.001). Notably, during the sedation stage, the firing rates of DG excitatory neurons were significantly lower than those of CA1 (Figure 7E, Wilcoxon signed-rank test, p < 0.001).

For the interneurons, the firing rates in CA1 and DG also decreased with the application of isoflurane, and increases with the recovery from anesthesia (Figures 7D,F, Friedman test; CA1, χ² = 87, p < 0.001; DG, χ² = 60; p < 0.001). And the firing rates of DG interneurons were consistently lower than that in CA1 except for BS stage (Figure 7F, Wilcoxon signed-rank test; wake_pre, p = 0.02; induction, p < 0.001; sedation, p = 0.009; emergence, p = 0.001; post_RORR, p < 0.001; wake_post, p = 0.002). What’s more, the DG interneurons decreased their firing rates faster in the induction and sedation stages of anesthesia (Figure 7G, Wilcoxon signed-rank test; induction, CA1, 63 ± 9.8%, DG, 14 ± 3.7%, p < 0.001; sedation, CA1, 13 ± 3.9%, DG, 6 ± 3.5%, p = 0.02), and increased slower after the cessation of isoflurane (emergence, CA1, 10 ± 1.9%, DG, 6 ± 4.7%, p = 0.006; post_RORR, CA1, 20 ± 3.6%, DG, 6 ± 3.4%, p < 0.001; wake_post, CA1, 65 ± 5.7%, DG, 43 ± 10.9%, p = 0.02) than CA1 interneurons.

Hippocampal neurons have been reported to be phase-locked to theta oscillation during navigation and memory (Reddy et al., 2021). To test whether the CA1 spikes were phase-locked to LFP oscillations in different sleep and anesthesia stages, we calculated the firing phase of neurons. The firing histograms of an interneuron and two excitatory neurons were shown in Figure 8. We found these three neurons were phase-locked to theta, but not to delta, alpha or beta bands. And this theta phase-locking feature was manifested in REM sleep, NREM sleep, wake_pre, induction and wake_post stages, with the preferred firing phase remaining stable.

While the general trend for CA1 and DG neurons involved a decrease in firing rates during isoflurane delivery and an increase after its removal, individual neurons exhibited distinct firing patterns. We identified six neuronal types based on their firing rates across different anesthesia stages. The normal-type, comprising the largest neuron population, exhibited a gradual decrease in firing rate with isoflurane administration and an increase after its removal (see example in Figure 9A, summary in Figures 9G,H blue line). The induction-type displayed a higher firing rate during the induction stage (see example in Figure 9B, summary in Figures 9G,H green line). Three other types of neurons, showing high firing rates during the sedation, emergence, or post_RORR stage, were categorized as sedation-type (see example in Figure 9C, summary in Figures 9G,H red line), emergence-type (see example in Figure 9D, summary in Figures 9G,H orange line), and post_RORR-type (see example in Figure 9E, summary in Figures 9G,H purple line). The remaining neurons with irregular firing patterns were grouped as other-type (see example in Figure 9F).

We then summarized the proportion of each neuronal type in CA1 and DG, respectively (Figure 9I). Almost all the interneurons belonged to normal-type (CA1, 16 of 17 units, DG, 13 of 13 units). For the excitatory neurons, approximately 64–66% of both CA1 and DG neurons were normal-type. The induction-type shared a similar proportion in both regions. However, the sedation-type (CA1, 15.3%; DG, 11.2%) and post_RORR-type (CA1, 5.3%; DG, 2.4%) occupied a larger proportion in CA1 than in DG. Conversely, the proportion of emergence-type in DG (8.0%) was larger than in CA1 (2.1%). We further compared the firing rates of the five neuronal types for excitatory neurons during sleep-wakefulness cycles. In the CA1 region, the firing rates during NREM sleep were higher than during wakefulness for the sedation-type (Figure 9J; NREM, 1.75 ± 0.24 Hz, wake, 1.48 ± 0.28 Hz, Friedman test, p = 0.002). Additionally, the firing rates of the induction-type during REM sleep were significantly higher than during wakefulness (Figure 9J; REM, 3.82 ± 1.17 Hz, wake, 2.35 ± 0.76 Hz; Friedman test, p = 0.02). Furthermore, a significant difference in firing rate across the sleep-wakefulness cycle was only observed between normal-type and post_RORR-type (Kruskal-Wallis test; NREM, p = 0.002, REM, p = 0.003, wake, p = 0.001). For the DG region, only normal-type neurons showed lower firing rates during wakefulness than during NREM sleep and REM sleep (Figure 9J; NREM, 3.73 ± 0.35 Hz, REM, 4.67 ± 0.47 Hz, wake, 1.65 ± 0.16 Hz; Friedman test; NREM versus wake, p < 0.001, REM versus wake, p < 0.001).