Cholinergic Reversal of Isoflurane Anesthesia in Rats as Measured by Cross-approximate Entropy of the Electroencephalogram

The experimental procedures and protocols used in this investigation were reviewed and approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin, Milwaukee, Wisconsin. All procedures conformed to the Guiding Principles in the Care and Use of Animals of the American Physiologic Society and were in accordance with the Guide for the Care and Use of Laboratory Animals  (National Academy Press, Washington, D.C., 1996).

Experiments were performed in 26 adult, male, Sprague-Dawley rats. The animals were anesthetized with 1.5% isoflurane and prepared for bihemispheric electroencephalogram recording. The anesthetics were vaporized into a gas mixture of 30% O², 70% N²that the animals were breathing spontaneously through a gas anesthesia mask (model 51610, Stoelting, Wood Dale, IL). Anesthetic concentration was monitored (POET II monitor; Criticare Systems, Inc., Waukesha, WI) through a sampling line connected to the anesthesia mask. The body temperature was maintained at 37°C with a thermostat-controlled (model 73A, YSI, Yellow Springs, OH), water-circulated (K-mod 100, Baxter Healthcare Corp., Chicago, IL) heating mat. For epidural electroencephalogram recording, two pairs of stainless steel screw electrodes were placed through burr holes in the cranium symmetrically over the left and right frontal cortices (coordinates: 3.5 mm lateral and 3.2 mm rostral or 0.7 mm caudal from bregma). According to the stereotaxic atlas of Paxinos and Watson, 38these coordinates corresponded to primary motor and forelimb somatosensory regions, respectively. For the infusion of cholinergic agents, a 30-gauge needle was implanted into the lateral ventricle (coordinates: 1.2 mm lateral, 0.8 mm caudal, and 3.5 mm deep from bregma). The screws were advanced to the level of the dura. In some animals, the femoral arteries were cannulated for the measurement of blood pressure and for the withdrawal of blood samples. All surgical sites were treated with bupivacaine.

After surgery, the rats were stabilized for 1 h at 1.0% inspired isoflurane, equivalent to 0.7 minimum alveolar concentration (MAC) in the rat. In two separate groups of rats (n = 7 and 8), the cholinesterase inhibitor neostigmine (25 μg) or the muscarinic agonist oxotremorine (25 μg) was infused for 60 min at rates of 5 and 2.5 μl/min, respectively. The doses and infusion rates were established in preliminary experiments to attain maximum electroencephalographic effect and minimum systemic (blood pressure, heart rate) effect, respectively. Each animal received only one type of drug. In a third group of rats (n = 5), the vehicle, artificial cerebrospinal fluid, was infused. In all groups, the electroencephalogram was recorded continuously, starting 10 min before the infusion started. The motor reflex to tail pinch was tested before and after the infusion of agents. Arterial samples were taken at the same time points. Tail pinch was performed with a hemostat, applying a gentle force for a few seconds at approximately mid-tail, with the exact location varied at random. If no response was seen, the pinch was repeated up to three times at slightly different locations.

A fourth group of animals (n = 6) received no drug infusion; these rats were instead allowed to emerge to an awake, sedated state by reducing the isoflurane concentration to 0.4%. Although bupivacaine was applied to all surgical sites, we chose not to remove isoflurane completely so as to minimize stress to the animal. At the beginning of the experiment, the anesthetized animals were placed in a cylindrical, plastic restrainer of 6-cm diameter (Harvard rodent restrainer, model AH-52-0292; Harvard Apparatus, Holliston, MA) that gave them limited movement of their head and limbs but prevented them from crawling out of the cylinder. The assembly was then placed in a transparent Plexiglas anesthesia box. The electroencephalogram was recorded continuously. After a baseline had been established at 1.0% isoflurane, the concentration was lowered to 0.4%. The rats were observed continuously for alertness and the absence of signs of discomfort. An orientation response to gentle knocking on their housing and the presence of spontaneous, calm snouting, sniffing, and whisking was taken as an indication of normal waking state. Potential signs of discomfort would be shaking, tremor, vocalization, or persistent or fierce escape attempt. A final behavioral assessment was done at 1 h.

Bipolar electroencephalographic signals recorded from the two hemispheres were analog-filtered for 1–100 Hz and digitized at 200 Hz using the WinDaq data acquisition software (DATAQ Instruments, Akron, Ohio). We previously tried higher sampling frequencies, such as 400 Hz, but found that C-ApEn worked best with 200-Hz data. This sampling rate makes C-ApEn relatively sensitive to γ-frequencies (30–80 Hz) but excludes high-frequency noise. Although anesthetic agents may influence sub-δ (<1 Hz) rhythms of the electroencephalogram, these waves are often the most contaminated by motion artifacts and therefore were filtered out. We also theorized that the dominant electroencephalographic features that may change with the anesthetic dose near the threshold for loss of consciousness would be the high-frequency rhythms.

For analysis, the electroencephalographic records were divided into epochs of 2-s duration. Epochs containing movement-related artifacts were dropped from the analysis. A computer subroutine designed to detect transient, large-amplitude deflections in the electroencephalogram was used for this purpose. For each valid epoch and hemispheric lead, the SD of the signal was calculated and used for normalization. C-ApEn was calculated as introduced by Pincus et al.  28We present the definition in a simpler, but mathematically equivalent form:

Here, C^im(r) is the relative frequency of finding the same brief signal pattern of length m in both electroencephalogram channels within the epoch. C^{im + 1}(r) is the relative frequency of finding in both electroencephalogram channels the same signal pattern extended to length m + 1. The sum is for i = 1 to n − m + 1, where n is the number of samples used for comparison. The parameter r is the noise threshold; two samples differing by less than r are considered equal. We chose parameter values n = 20, m = 3, and r = 0.25 × SD, where SD is the SD of the signal within each epoch. Using n = 20, the temporal window of C-ApEn calculation was 100 ms. We previously examined 27the dependence of the C-ApEn-anesthetic concentration relationship on window duration between 0.05 and 2 s and found that 100 ms, equivalent to 20 points, produced optimal results in terms of dynamic range and variance. We have also evaluated the effect of choice of m between 1 and 4. The most reproducible results were obtained with m = 3. Another consideration to use m = 3 was to confer a greater sensitivity of C-ApEn to high-frequency components of the electroencephalogram (i.e. , γ). The strong dependence of C-ApEn on electroencephalogram components greater than 20 Hz was demonstrated previously. 27Note that the very-low-frequency electroencephalogram components (less than 1 Hz) were attenuated at the time of recording.

Calculations were performed using a program written in QuickBASIC (Microsoft, Redmond, WA). In essence, recorded electroencephalographic data were read from disk in packets of 400 × 2 (2 s of data times two channels), normalized, and processed to derive one C-ApEn value for each epoch. The resulting data were plotted as a function of time or averaged over 1-minute intervals at selected time points. Statistical analysis was carried out using Analysis of Variance of Microsoft Excel, Office 2000, and NCSS 2001 Statistical Software (NCSS, Kaysville, UT). For post hoc  comparison of C-ApEn group means, the Tukey-Kramer test was applied. The data were tested for normality using the Shapiro-Wilk test, which yielded no reason to reject the normality assumption. Physiologic data means before and after cholinergic activation were compared using a t  test assuming unequal variances.

Animals receiving 0.4% (0.3 MAC) isoflurane produced frequent movements of the snout, such as sniffing, chewing, licking, and occasional gross limb movements, but the animals showed no sign of stress or discomfort. Their eyelid reflex to touch was present, and a motor response to gentle tail pinch could invariably be elicited. In separate, noninstrumented animals, this state of sedation was associated with preserved righting reflex, suggesting a conscious state. 27 

Animals anesthetized with 1.0% (0.7 MAC) isoflurane showed no spontaneous motor signs. Animals subjected to this dose of anesthesia invariably lose their righting reflex, 27suggesting loss of consciousness. However, the motor response to tail pinch was present in all animals, as expected on the basis of the applied MAC fraction.

After the infusion of neostigmine or oxotremorine, animals showed motor signs similar to those under 0.4% isoflurane. In some animals, gross motor behavior was observed (in this case, the isoflurane concentration was immediately raised, and the experiment was terminated). Although the righting reflex was not tested in these animals, their behavioral signs indicated a capability and motivation to right themselves, thus suggesting that they were in the conscious state. In surprising contrast, the motor reflex to tail pinch was absent with no exception, and the response could not be elicited on any repeated trial. Cerebroventricular infusion of artificial cerebrospinal fluid at 1.0% isoflurane anesthesia produced no change in behavior.

Systemic physiologic parameters obtained in three states are displayed in Table 1. The Pco²was moderately elevated, with a corresponding decrease in pH, during 1.0% isoflurane administration before cholinergic drug administration, most likely because of depressed ventilation of the spontaneously breathing animals. This difference was diminished after cholinergic activation, suggesting restored respiratory drive. Similarly, MAP was lower at 1.0% isoflurane (before cholinergic activation) than at 0.4% isoflurane, but it was increased after cholinergic activation. In all three states, Po²and MAP remained within the physiologic limits and within the range of cerebral blood flow autoregulation.

Electroencephalogram signals obtained from one animal are shown in figure 1. The electroencephalogram at 0.4% isoflurane was typical of the awake, sedated rat, essentially desynchronized, with high-frequency activity superimposed on runs of slower θ waves. At 1.0% isoflurane, synchronized δ waves dominated the electroencephalogram. After neostigmine or oxotremorine infusion, the desynchronized electroencephalogram pattern typical to the waking state returned.

With the chosen infusion rate, C-ApEn increased gradually, reaching a maximum within 60 min. The effect time courses for the two agents were similar and are illustrated for neostigmine in figure 2. The maximum effect was an approximately two-fold increase.

Figure 3summarizes the finding that C-ApEn was significantly lower at 1.0% isoflurane than at 0.4% isoflurane, consistent with previous findings. 27Both neostigmine and oxotremorine, administered in the presence of 1.0% isoflurane, increased C-ApEn to approximately the same degree. Postinfusion C-ApEn values were not significantly different from those measured in animals receiving 0.4% isoflurane alone. Cerebroventricular infusion of artificial cerebrospinal fluid in animals anesthetized with 1.0% isoflurane did not change C-ApEn.

Recent years have seen a growing interest in applying novel electroencephalogram-derived indices, particularly various entropies and complexity measures, to assess the depth of anesthesia and loss of consciousness. 21,24–26Although visual inspection of raw electroencephalogram traces can be informative of the anesthetic state, the use of a suitable electroencephalogram-derived index could greatly improve the accuracy, speed, and consistency of the assessment of anesthetic depth. 13–15In addition, an examination of specific properties of the electroencephalogram, such as its complexity 39or interhemispheric synchrony, may help us to better understand the neurophysiologic basis of conscious perception 32,34,35,40and anesthetic action on the CNS. 16 

Most previous studies calculated ApEn from single-channel recordings of the electroencephalogram. A nonlinear, inverse relation between ApEn and the anesthetic dose in human subjects has been established for various anesthetics. 24–26Whereas the single-channel ApEn measures the temporal complexity of the electroencephalogram, the two-channel C-ApEn reflects the spatial and temporal independence of cortical potentials from two remote sites. The temporal relationship between the electroencephalograms of the two hemispheres has traditionally been assessed in terms of cross-correlation or coherence. 31–34Anesthetic-induced changes in interhemispheric electroencephalogram coherence have been demonstrated in surgical patients 16and in healthy adults during drowsiness. 41The application of cross-correlation and coherence techniques requires stationarity and linearity of signals, conditions not typical of the electroencephalogram. Entropy measures, conversely, can be reliably calculated for relatively short runs of data and are less sensitive to nonstationarity.

We recently argued 27that C-ApEn may be interpreted as a statistical measure of the number of neural states independently accessible by the two hemispheres. Thus, a decrease in C-ApEn during anesthesia would reflect a decrease in the number of independent hemispheric states, which may interfere with perceptual information processing. This interpretation is consistent with the strong dependence of C-ApEn on high-frequency electroencephalogram components 27that reflect fast cortical state transitions.

We previously showed that C-ApEn was reduced by isoflurane in rats 27and that this reduction occurred at an agent concentration that abolished the rats’ righting reflex. The righting reflex in rats is lost at a MAC fraction similar to that which produces a loss of response to verbal commands in humans. 42We hypothesized that if C-ApEn correctly reflects the level of sedation, then cholinergic activation that restores wakeful behavioral signs should also reverse the anesthetic effect on C-ApEn. We indeed found that cholinergic agents administered to isoflurane-anesthetized animals increased C-ApEn to a value measured in a lightly sedated, possibly conscious state at 0.4% isoflurane concentration. The level of sedation at 0.4% isoflurane is significantly lighter than that which would ablate the righting reflex. 27Although the return of spontaneous motor signs after cholinergic activation does not guarantee the return of consciousness, the comparably high value of C-ApEN after cholinergic activation suggests that it may have in fact reflected a conscious state.

Cholinergic mechanisms have been known to be important regulators of the state of consciousness 43and to play a key role in conscious information processing. 44The important role of cholinergic systems in general anesthesia has also been recognized. 1Halothane and enflurane have been shown to decrease acetylcholine turnover in the rat cerebral cortex. 45Isoflurane and halothane dose-dependently depress acetylcholine release in cortex and striatum. 46,47Halothane, isoflurane, and enflurane decrease acetylcholine concentration in the pontine reticular formation, 48a source of cortical arousal. In human subjects, isoflurane-induced reductions in regional cerebral metabolism are correlated with muscarinic receptor density, suggesting an involvement of cholinergic antagonism in isoflurane anesthesia. 49 

The behavioral and electroencephalogram-activating effects of cholinergic agents have been studied in freely moving animals. Infusion of the cholinergic agonist carbachol into the pontine reticular formation of rats increased the time spent in rapid eye movement sleep. 7,8Microinjection of neostigmine into the dorsal pontine tegmentum of freely moving cats produced electroencephalogram desynchronization. 4Intracerebroventricular administration of oxotremorine in rats suppressed neocortical sleep spindles. 9The systemic, blood-brain-permeable cholinesterase inhibitor physostigmine produced hemispheric electroencephalogram asymmetry in rats. 50This result is consistent with a greater hemispheric independence, as indicated by the rise in C-ApEn after cholinergic activation in our study.

Cholinergic agents have also been shown to exert an anesthetic sparing or reversal effect. In humans, pretreatment with physostigmine increased the anesthetic requirement to produce unconsciousness in human subjects. 10In a recent study, a close correlation between the return of response to verbal commands, bispectral index, and auditory steady-state response after physostigmine infusion and continued administration of propofol at hypnotic dose was found. 12The antimuscarinic agent scopolamine reduced interhemispheric electroencephalogram coherence in human volunteers. 51In rats, both physostigmine and oxotremorine reduced the duration of ketamine anesthesia. 5,6In dogs, acetylcholinesterase inhibitors were found to antagonize halothane anesthesia and produce an awake-like electroencephalogram. 2,3In rats, injection of carbachol into the pontine reticular formation antagonized sleep spindles during halothane anesthesia. 52 

In our experiments, cholinergic agents produced antinociception concurrently with their arousing effect on the electroencephalogram. Roy and Stullken 3and others have noted this divergence between CNS arousal and nociceptive depression by cholinesterase inhibitors in dogs. In rats, Hartvig et al.  53showed that physostigmine infusion produced analgesia, as indicated by an increase in latency of the tail-flick response to radiant heat. These results are consistent with the differing mechanisms of action for hypnosis and areflexia by general anesthetics. 18,19 

The mechanism of antinociception by neostigmine is not quite clear and may be caused by either a central or a peripheral effect. In the study by Horrigan, 2both neostigmine and physostigmine reduced anesthetic MAC requirement in halothane-anesthetized dogs. Because only physostigmine, not neostigmine, crosses the blood-brain barrier, this suggests a mode of peripheral action. However, the effectiveness of neostigmine administered via  the intracerebroventricular route implicates a central mechanism of antinociception. The antinociceptive and sedative effects of neostigmine injected into the pontine reticular formation support a mechanism of central modulation. 54,55 

The similar nociceptive and electroencephalographic effects of oxotremorine and neostigmine further implicate the involvement of muscarinic receptors. 56Antinociception produced by carbachol injection into the brainstem reticular formation was blocked by the muscarinic antagonist atropine 55or a specific M²antagonist. 57Some investigators 55,58found the MAC sparing of oxotremorine but not of physostigmine in isoflurane-anesthetized rats. 58This difference may be a result of a lower dose or a different route of administration of the cholinesterase inhibitor.

Cholinergic agents may activate the electroencephalogram by reversing an effect of isoflurane on the cholinergic system, or they may work through an independent pathway. General anesthetics depress cholinergic transmission in the CNS 59and have been suggested to affect consciousness through nicotinic receptors. 60A recent study by Flood et al.  61suggests that nicotinic acetylcholine receptors are unlikely to be involved in isoflurane-induced immobility and hypnosis, although they may play a role in amnesia and analgesia. Conversely, the involvement of M¹and M²muscarinic receptors in prefrontal electroencephalogram activation in the rat has been supported. 36,37Thus, it is likely that the electroencephalographic effects of cholinergic activation reflect the reversal of a muscarinic receptor-mediated anesthetic action.

Certain limitations of the present study should be recognized. First, the rats were restrained, which may have contributed to their arousal after cholinergic activation. Second, because of the restraint, we were not able to record an electroencephalogram at zero percent isoflurane. For the same reason, the righting reflex could not be assessed in the same animal in which the electroencephalogram was recorded. Future experiments should be performed in chronically instrumented, freely moving rats to verify and extend the present findings. Also, more specific cholinergic agonists and antagonists could be used and injected at multiple doses at selected cortical and subcortical sites to further delineate the neurophysiologic mechanisms that contribute to the observed changes in the electroencephalogram and C-ApEn. With respect to the electroencephalogram, we derived C-ApEn from frontal leads only. To obtain a fuller characterization of functional connectivity throughout the cortex that may affect consciousness, it would be interesting to examine the electroencephalogram from posterior sites as well and to analyze their intrahemispheric relationship as a function of anesthetic dose. Intracortical electrodes could be used for a better localization of the recorded field potentials. In addition to spontaneous activity, the effect of anesthesia on functional connectivity derived from cortical evoked potentials would be informative for an assessment of cortical sensory processes. Such a study is in progress in our laboratory.

In summary, our present data are consistent with the cortical arousing and antinociceptive effect of cholinesterase inhibitors and cholinergic agonists as found in humans and most mammalian species. They lend further support to the previously recognized mechanistic divergence of anesthetic actions on consciousness and the nociceptive reflex. They suggest that the degree of independence of the frontal hemispheric electroencephalogram, as measured by the interhemispheric C-ApEn, correlates with conscious behavior but not with the nociceptive response. C-ApEn promises to be a suitable indicator of anesthetic hypnosis.