Experiments were conducted on adult male B6/129 (hybrid of C57/Bl6X129) F1 mice, weighing 20~35 g, aged 8~12 weeks (n=10). For surgery, animals were maintained in deep anesthesia induced by an intraperitoneal injection of a mixture of ketamine (120 mg/kg) and xylazine (6 mg/kg). A stereotaxic apparatus (Kopf instrument, USA) was used to fix the target area, and the animal's head was restrained by ear bars during surgery. For EEG/EMG electrode implantation, the skin was opened carefully and the surface of the skull was cleaned and dried for implantation of screws. The microscrew electrodes were implanted on the frontal area (anterioposterior (AP) 1.2 mm; mediolateral (ML) 1.5 mm; dorsoventral (DV) 0 mm), parietal area (AP -2 mm; ML -2 mm; DV 0 mm), and both the left and right side of the cerebellum for ground and reference electrodes. For EMG, a tungsten wire (0.003 mm, A-M Systems, USA) electrode was implanted into the neck muscle beyond the skull. All electrode coordinates follow the mouse brain atlas [27]. Next, the electrodes were fixed with dental cement (Vertex Self Curing, Dentimex, Germany). Scalp skin was sutured after cement hardened. All animals had at least one week of recovery period from surgery before their first recording session started.

The recording was performed in a Faraday cage with dim light. Prior to recording, mice were anesthetized in an induction chamber (4% concentration of isoflurane combined with pure oxygen) for 4 to 5 min. After anesthesia on a surgical level was assured by toe and tail pinching, the mouse was placed on a stereotaxic apparatus, and then the nose was connected to a respiration tube. Gel-type eye drops (Liposic ophthalmic gel, Bausch+Lomb, Germany) were spread on the surface of the animal's eye to prevent drying.

Two different paradigms of applying the dose were tested. One paradigm mimicked the induction of anesthesia (referred to as the induction phase), and the dose given to the animal started from 0.5%, with 0.5% step-ups to the highest level. The other paradigm mimicked the emergence from anesthesia (referred to as the induction phase), and the dose given to the animal started from the highest level, with 0.5% step-downs to the lowest level, which was 0.5% or 0.3%. The highest level was determined prior to recording in each mouse, which was set to the maximal dose with no sign of breathing difficulty for several minutes. The highest level was 2.0% in six mice, 1.5% in three mice, and 1.0% in one mouse. During recording, each dose was tested for 15 min. In two mice, the dose was down to 0.3% in the emergence phase, when the animal showed no mydriatic pupil responses at 0.5%, and no response to tail/toe pinch. The induction and emergence phases were tested 16 times in 10 mice.

The pupil was recorded with an infrared LED webcam (DABOA-DA500, DABOA, Korea). In order to avoid the image reflection, the infrared LED was removed from the camera. The camera was fixed on a stereotaxic arm to maintain the focusing distance, which was 0.7~1.3 cm from the mouse's left eye. A custom-built infrared illuminator (850 nm, LED) was fixed above the eyelid. The light from the illuminator was detected by a photodetector and was converted to a sync signal. The video was recorded with 640×480 pixels at 30 fps. To avoid the dark adaptation effect, recording started 15~30 minutes after the elimination of the visible ray.

EEG and EMG were recorded with Intan RHD2000 data acquisition system with RHD2000 Interface software (Intan technologies, USA, setting 1), or SynAmps2 amplifiers with Neuroscan Scan 4.3 software (Compudemics NeuroScan, USA, setting 2), both at a sampling frequency of 2 kHz. EMG signal was re-referenced by subtracting the averaged value of all the electrodes to eliminate the ground and reference.

Pupil dynamics were analyzed frame by frame based on a region-based active contour model, by using a localized_seg.m function presented by Lankton and Tannenbaum [28]. The code is downloadable from http://www.shawnlankton.com/2008/04/active-contour-matlab-code-demo/. The model is based on the Snake algorithm, but finds the segment from the local image statistics rather than global image statistics. Compared to the Hough circle detection algorithm [28,29,30], a widely used shape detection technique, the region-based active contour model finds the mouse pupil more accurately and robustly. Briefly, the algorithm starts segmentation with a rough 'initial contour' drawn manually on the image, then optimizes the region contour by deforming it iteratively while comparing foreground and background image contrast statistics. The contour of the pupil from each frame was used as an initial contour in the subsequent frame. The contour of a full or partially occluded pupil is shown in Fig. 1. Additional post-hoc processing included error correction when necessary. The pixel coordinate was converted to mm. The constriction/dilation responses of the pupil were estimated according to the area of pupil.

After converting the image data into pupil data, the zero time in pupil and EEG/EMG signals were synchronized using the synch signal. As the events of interest, the mydriasis (quickly widening of pupil) in the pupil responses and the burst of EEG were the focus in this study. Typically, mydriasis is referred to as pupil dilation under sympathetic stimulation; here, we refer to it as quickly widening pupillary dilation.

The criteria of mydriasis are as follows: (i) the pupillary aperture starts from a constricted pupil to full dilation, and, (ii) the pupillary dilation occurs within 1 min. The threshold for the constricted pupil was 1/8 in terms of the ratio of the pupil area to eye surface (the area between two eyelids), whereas the threshold for the full dilation was 1/3. The thresholds were determined after a visual inspection of the images. The instantaneous dilation speed was also investigated by calculating the temporal change in the pupil size (mm²/sec).

The burst in EEG was defined by a large and fast fluctuation of EEG for a short period of time, and the burst ratio was calculated according to Land et al. [24]. First, the threshold was determined as a median value of the EEG amplitude during the anesthetic state, acquired for 30 min at a 1.0% concentration dose. The periods of absolute EEG higher than the threshold were counted as the burst period, and the burst ratio was defined by the ratio of the burst period to the observation time period.

Different doses of anesthesia were tested to mimic the induction of anesthesia (induction phase, stepwise increase in anesthetic dose), or emergence of anesthesia (emergence phase, stepwise decrease in anesthetic dose). Changes in the pupillary size were observed in both the induction and recovery modes, including the highest dose with typical patterns of fluctuations, such as slowly widening or shifting motions. The pupil fluctuations occurred irregularly in various patterns. The mydriatic patterns (full dilation of the pupil in a short period of time) were observed in all the recorded mice, but they occurred only in the emergence phase (43 events in total). The maximal dose showing the mydriatic patterns was referred as the critical dose for mydriasis occurrence. The critical dose was 1% in two mice, 0.5% in six mice, and 0.3% in two mice. The occurrence rate of mydriasis in 0.3% was 10±8.4 times per hour; in 0.5% it was 9±1.8 per hour, whereas the value was 1.6±0.54 per hour in 1.0% dose.

Fig. 2 shows the representative traces of pupillary responses and parietal EEG at two different doses (2% versus 0.3%) collected from the emergence phase. The dilation occurred slowly in 2%; hence, it was not counted as a mydriasis event. Although the maximum size of pupil was similar, the EEG traces and spectrum showed a distinctive feature at low and high doses. The parietal EEG showed a slow oscillation under deep anesthesia, whereas the EEG was fast and irregular under light anesthesia.

Fig. 3 shows representative traces of pupil size, frontal and parietal EEG, and EMG. The earliest phase of pupil dilation precedes the responses in EEG and EMG. The typical patterns of EEG under anesthesia are characterized by a slow oscillation with infrequent burst activity (exemplified in Fig. 2B), changing to fast and complex EEG rhythms as the pupil rapidly dilated (exemplified in Fig. 2C). The mydriasis and burst events were logged according to the definition depicted in the Methods. In ten mice, 43 mydriasis events were logged. The burst of EEG occurred intermittently in an episodic way at the critical dose. The episodic EEG burst became continual after mydriasis was monitored (exemplified in Fig. 4). All the mydriasis events were followed by a continual burst of EEG, although the patterns were not unitary (i.e., frequently a occurring burst or a continuous burst). None of the continual burst episodes preceded the mydriasis in the recorded mice. EMG with a high amplitude was also logged, and 47% of the mydriasis were accompanied by EMG responses.

The burst ratio of EEG was calculated every 5 sec and compared to the pupil dynamics. As shown in Fig. 4, the temporal pattern of the burst ratio coincides with the dilation speed rather than the pupil size. The peak latencies of the EEG burst ratio were calculated with respect to the pupil size and the dilation. Within 5 sec of tolerance, no delay was observed with respect to the dilation speed, whereas the latency between the peaks of EEG burst ratio and pupil size was 14±3 sec. Considering the inverse of the burst ratio has been considered as anesthesia depth, pupil dilation indicates the transition of the brain state, which is consistent with previously reported observations by Reimer, et al. [13].

In the present study, mydriasis occurred on the step-wise emergence from anesthetic state, while it did not occur during induction of anesthesia. This result is in line with previous observation that there is a path-dependent feature of brain state during transition of (the) anesthetic state in ketamine-xylazine cocktail induced anesthesia [31], and the observation of Friedman, et al. [32] that there is the path-dependency of behavior response between induction and emergence period of halothane and isoflurane anesthesia. Both observations suppose that this asymmetry between induction and emergence of anesthesia generates from neural inertia, which means the inherent resistance to changes in arousal state. Our observation implies that mydriasis can be a marker for neural inertia.

The simultaneous observation of EEG and EMG shows that the mydriasis implies a mixed state of anesthetized and conscious states. All the mydriasis events were accompanied by the increase of burst activity in the EEG, and half of the events were followed by EMG activity. Considering the recovery of motion requires the transition from anesthetized state to awaken state, the mydriasis preceding the recovery of motion implies a sign of anesthetic level transition. While we claimed mydriasis is a fast dilation compared to other periods during anesthesia, the dilation speed is slower than awaken state shown in the previous studies (Reimer et al, 2014). We observed strong slow rhythm (Fig. 3C) on both frontal and parietal area during mydriasis, which did not occur on awaken animal. Other observation is that some of mydriasis events did not came with EMG signal, a feature of wakefulness. For these reasons, mydriasis is a post-stage of awakening, not the awaken state itself. For these reasons, we suggest that mydriasis is an indicator of transitional state before recovering consciousness.

Monitoring of anesthetized animal is essential on animal research, especially on modern neuroscience such as using fMRI equipment or two-photon microscope. There are studies that dilation and constriction of pupil is correlated with brain state on anesthetized mice and rats [14,15]. Our observation shows that pupil diameter precedes electrophysiological change of brain during transitional state of anesthesia. Observing pupil diameter enables watching out transitional state of anesthesia without recording electrophysiological activation during surgery and experiment of animal research.