Animals: Nine adult (male, 12–18 weeks, BW: 24g-27g) wild type mice (C57BL/6N, Charles River Laboratories GmbH, Germany) were used in this study. The mice were housed individually under a 12/12-hour light/dark cycle (lights on: 9am, 22°C ± 1°C, humidity: 55% ± 5%) with ad libitum access to food and water. All experimental procedures were approved by the Committee of Animal Health and Care of the State of Upper Bavaria, Germany (ROB-55.2–2532.Vet_02–19–121).
EEG monitor and electrodes: For this study, a BIS™ Complete 2-channel Monitoring System (Covidien Deutschland GmbH, Germany), and subdermal needle electrodes (3mm x 40mm, 0.5” x 27G (Xi’an Friendship Medical Electronics Co., Ltd.) along with a custom-made electrode-monitor interface were used. The custom-made interface was prepared as follows: The BIS™ Quatro Sensors (Covidien Deutschland GmbH, Germany) were horizontally cut at 7cm from the microchip to expose the conducting silver/silver-chloride array. The terminal wire endings of the subdermal needles were soldered to the exposed array using gold wires (diameter: 150µm, standard gold wire, round, which provided a stable but pliable contact) covered with soldering tin. The individual sensor interfaces were electrically insulated using heat shrink insulation tubes. Finally, the subdermal needles were electrically insulated using insulation tubes only to retain a 5mm long conducting tip.
Monitor settings: BIS™ monitor was used with the “filters off” setting, i.e. a bandwidth of 0.25Hz − 100Hz. The additional filters were manually turned off as described in the user manual of the BIS™ monitor. Adequate electrode impedances were automatically measured by the BIS™ device. Small impedance offsets were corrected by delicate adjustments of the electrode placements or electrode angles on the skull. After verifying suitable impedance through all the electrodes, raw EEG signals from two channels were displayed on the monitor. Since the channel(s) to be displayed were not opted manually, the monitor automatically displayed the EEG signals from BIS™ sensor number 1. With a presetting to “export” the signals on the monitor, the displayed EEG signals were saved as R2A files (.r2a-format) to an external USB flash drive attached to the monitor with a sampling rate of 128 Hz. The trend data of the processed EEG indices was also stored to the USB flash drive as SPA (text file: .spa-format) with a resolution of one index per second.
Anesthesia protocol: The mice were briefly anesthetized with isoflurane (CP-Pharma, Germany) using a pre-filled plexiglas box (Induction chamber- 8329001, AgnTho’s AB, Sweden). Then, each mouse was placed in a stereotactic frame (prone position) with an initial maintenance concentration of 1.2% isoflurane (flowrate: 192ml/min) through an anesthesia unit, especially adapted for mice (Univentor 410 Anaesthesia unit, AgnTho’s AB, Sweden). Eye cream (Bepanthen®-Bayer AG, Germany) was applied to the eyes to avoid irritations. An automatic heating pad (Homeothermic Monitoring System, Harvard Apparatus, USA) was placed under the mouse abdomen to maintain a body temperature of 37°C.
Recording scenario: After loss of responsiveness (LOR, paw clip test), four subdermal needle electrodes were inserted into the scalp of the mouse (Fig. 1). Two electrodes (BIS™ sensor numbers 1 and 2) were placed above the left and right frontal lobes and the reference and ground electrodes (BIS™ sensor numbers 3 and 4) were placed above the left and right occipital lobes. The mean duration of anesthesia for the complete preparation before starting the recordings was 10 minutes. To observe concentration-dependent changes in the EEG and the indices provided by the monitor, 3 different concentrations of isoflurane were applied cyclically throughout the process with a flow rate of 192 ml/min. The experiment was started with 1.2% isoflurane. After 7 minutes, the isoflurane concentration was increased to 2.2% for 5 minutes (longer periods lead to very low breathing rates at 2.2%). This anesthesia cycle was repeated one time and then terminated with 1.6% isoflurane for 5min.
Data analysis: Data analysis was performed using native toolboxes and custom scripts in MATLAB-R2019a. The raw EEG signals were converted to vectors stored in the .mat format from r2a file format and were divided into different segments relating to different concentrations. The time series for BIS™ index and suppression ratio were extracted for every concentration with a temporal resolution of 1s.
Since the EEG signals predominantly consisted of bursts (periods of high-voltage electrical activity) and discontinuous patterns of cortical suppression (periods of low or near-isoelectric activity), the analyses were focused on suppression ratios and suppression durations with respect to the changing isoflurane concentrations. A burst-suppression detection algorithm (BSA) was developed to classify bursts and suppressions in the raw EEG signals. The BSA was based on visual and subjective amplitude thresholds and delivered binary series in the form of ONEs (bursts) and ZEROs (suppression). An additional temporal threshold was included in the BSA to classify any isoelectric signal which lasted for more than 0.5 seconds as one bout of suppression [20]. Though BSA was used to show the ease of burst suppression detection at these anesthetic dosages, the algorithm was verified by comparing the resultant binary series with that of a traditional nonlinear energy operator (NLEO) [21] which is widely used to detect bursts and suppressions. Instantaneous energy in the raw EEG signal was assessed to detect spikes [22] and bursts were identified at or above one standard deviation. The resultant binary series of BSA and NLEO across all concentration cycles across all mice were compared using Pearson correlation coefficient.
For the following analyses, only the EEG signals during the last 3 minutes of every concentration were selected to avoid metabolic adaptation artifacts after changing anesthetic concentrations. Using the resultant binary series, suppression durations were calculated for individual concentrations by counting the number of zeros in every suppression segment. To compare suppression durations at different isoflurane concentrations, the time series for suppression durations were calculated with a moving mean (window length of 10 consecutive suppressions) across all the concentrations for all the mice.
Suppression ratios were calculated according to the standard method of the BIS™ monitor, i.e., the percentage of suppressed EEG signals in the previous 63 seconds [20]. Burst ratios were calculated as the percentage of bursts in the previous 63 seconds of the EEG signals. Additionally, across all the concentrations, means of the calculated suppression ratios based on the BSA were compared with the means of the real time suppression ratios provided by the monitor to estimate the difference in burst-suppression detection between the monitor and the BSA.
To estimate the concentration dependent trend of the BIS™ index, the mean BIS™ index for every isoflurane concentration was calculated for each mouse from the real time BIS™ indices provided by the monitor.
Spectral entropy of EEG can be used as a measure of hypnosis during anesthesia [23]. State entropy is a spectral entropy parameter computed from 0.8 to 32 Hz. These entropy indices range from 0 to 91. An index that is equal to 91 means that the subject is fully awake and a state entropy of 0 means the EEG signal is isoelectric with no brain activity [24]. For humans, the recommended range of spectral state entropy index during GA is between 40 and 60 [19]. To estimate state entropy (referred to as spectral entropy in this article) in anesthetized mice, 7 short chunks of data without burst suppression patterns were found with visual inspection in 3 out of 9 mice pooled together. These chunks of EEG signals were looped on the computer to generate 7 sets of EEG signals each of which were 1 minute long. These EEG signals were replayed to the GE ENTROPY™ module (GE Healthcare, Chicago, Illinois, USA) through the computer [25]. The spectral entropy indices for these sets of data were extracted to identify the range of entropy values for anesthetized mice.
Statistical analysis:
All statistical tests and data plots were performed using GraphPad.Prism.9.3.1.471 (GraphPad Software, San Diego, California USA) except for the Durbin & Skillings–Mack test which was performed on XLSTAT v24.3.1342 (Addinsoft, Paris, France). For mouse number 1, the 1.6% and for mouse number 2, the 1.2% isoflurane concentration steps were excluded due to prolonged respiratory depression and signs of movement respectively. Hence, for the comparison of suppression ratios and burst ratios across all the 3 concentrations on 9 mice, the Durbin & Skillings-Mack test was performed, which is a paired nonparametric test for repeated measures with missing values. The same comparison was tested with 7 mice (after removing the 2 mice with missing values) using the Friedman test and a post hoc Dunn’s multiple comparison test. The comparison of suppression ratios in 9 mice calculated by the monitor and the BSA was statistically tested with the Wilcoxon signed-rank test. The comparison of concentration dependent BIS™ index delivered by the monitor was statistically tested with Friedman test and a post hoc Dunn’s multiple comparison test on 8 mice (after removing the mouse with the missing data at 1.6% isoflurane). The summary data across mice for each variable are presented as medians with interquartile range. The statistics are reported at 95% confidence interval (P-value < 0.05). The absolute P-values digits are reported for every statistical test.