- Open Access
Evidence of a pharmacodynamic EEG profile in rats following clonidine administration using a nonlinear analysis
© Azulay et al; licensee BioMed Central Ltd. 2011
- Received: 11 March 2011
- Accepted: 26 June 2011
- Published: 26 June 2011
Changes caused by clonidine in rodent electroencephalograms (EEG) have been reported with some inconsistency. For this reason, a pre-clinical study was conducted in order to confirm previous findings with both a standard spectral analysis and a sleep stage scoring procedure. In addition, a nonlinear technique for analysing the time-varying signals was implemented to compare its performance against conventional approaches.
The nonlinear method succeeds in quantifying all dose-related responses from the data set relying solely on the EEG trace.
Nonlinear approaches can deliver a suitable alternative to the sleep-stage scoring methods commonly used for drug effect detection.
- Sleep Stage
- NREM Sleep
- NREM Sleep Stage
- Electrode Derivation
The role of the noradrenergic system in sleep physiology has been studied extensively using different pharmacological approaches [1–5]. It is well established that the non-selective alpha-2 agonist clonidine promotes non-rapid eye movement (NREM) sleep in both humans and rats mainly by activating presynaptic inhibitory autoreceptors and thereby inhibiting noradrenergic neurotransmission. However the specific changes caused by clonidine directly in the EEG is less clear both in humans and rats.
In humans, clonidine has been shown to cause changes in the EEG [6–8]. have noticed that power is increased in the 1.0-4.0 Hz range while decreased in the 8.0-12 Hz range. Recently published data suggest that clonidine causes different effects on the EEG depending on level of exposure . In this study the authors showed a significant decrease in power in the 0.5-12 Hz band during rapid eye movement (REM) sleep only.
The findings with clonidine from rodent EEG studies are equally ambiguous . showed that clonidine caused the power to decrease in the 0.1-4 Hz range and increase in the 4.1-8 Hz range compared to vehicle controls specifically in the NREM sleep stage. In contrast,  showed that the changes caused by clonidine were characterized by a significant increase in nearly all the frequency range from 1-30 Hz, with a peak at 13 Hz. Whether these differences in the reported pre-clinical and clinical findings with clonidine are only due to varying exposure levels or whether there are other additional causes (e.g. definitions of bands in the human and rodent studies, varying electrode derivations, changes restricted to specific sleep stages) is unclear.
An EEG recording is the superposition of a number of electrical signals eminating from various regions in the brain. If all these activities generated perfect superpositions of electrical sine wave oscillations then a linear method like a Fourier transform would expose these periodic components . In the case of non-regular time-varying signals, the task is considerably more complex and requires several steps. Nonlinear approaches aim first at detecting if a deterministic structure exists in the waveform before any further calculations. The presence or absence of determinism in a given data set is the corner stone of its numerical analysis since it completely defines which category of algorithms is practicable.
In terms of system dynamics, time series EEGs can be viewed as 2-dimensional windows open to a m-dimensional state space where time points repeat complex cyclic patterns . Since a reliable approximation of m would help understand the underlying mechanisms involved, algorithms designed to estimate this dimension have been applied [14–16]. Although none of the tested algorithms was able to deliver a definitive answer so far, these numerical methods are capable of extracting useful parameters for very specific phases like sleep [17–19], epileptic , schizophrenic  or anaesthesia  stages. These examples suggest that the measurement of the EEG determistic aspect can quantify different episodes of brain activities by using nonlinear methods.
The aim of this study was to identify a sensitive non-linear model of EEG analysis that can extract the pharmacodynamic signal of clonidine from EEG recordings from freely moving animals.
All animal experiments were carried out in accordance with the United Kingdom Animals (Scientific Procedures) Act 1986 and associated guidelines and approved by the local ethics committee. Rats were implanted with radiotelemetry transmitters (Data Sciences International, St Paul, MN, USA) intraperitoneally under isoflurane anaesthesia for the recording of the EEG and electromyogram (EMG). The cortical EEG electrodes (stainless steel screw electrodes) were implanted epidurally over the left parietal cortex (2.0 mm anterior and 2.0 mm lateral to lambda) and over the left frontal cortex (2.0 mm anterior and 2.0 mm lateral to bregma) for a fronto-parietal EEG recording [23–25]. The electrodes and leads attached to the skull were covered with dental acrylic and a second pair of electrodes was attached to the neck muscles to measure general EMG activity needed for the sleep stage analysis only. The rats were allowed to recover from the implantation of the device for at least 2 weeks, and the experiment was started once the animals were certified fit to continue by a veterinary surgeon. Animals were singly housed on a standard 12-12 h light-dark cycle and received standard diet and water ad libitum.
EEG and EMG data were continuously sampled at 250 Hz and the spectral upper limit was set at 40 Hz, with Data Sciences International hardware and software for 12 hours immediately following administration of drug at light onset. Animals were orally dosed with 0.03 (low dose (LD)), 0.1 (medium dose (MD)) or 0.3 (high dose (HD)) mg/kg clonidine or vehicle (VC) (0.5% w/v methylcellulose + 0.1% v/v Tween 80) at light on-set in a four way cross-over design with at least 48 hours between each dose. Recording of EEG and EMG signals began immediately after dosing. In addition to the sampling of these two signals, a general activity measure and core body temperature were collected in parallel. Due to an archiving issue, not all the files were available for the analysis.
Sleep stage scoring principle
EEG traces were cut into 16.384 second epochs (212 points). Since one epoch of EEG data values tend to follow a gaussian distribution the Kolmogorov-Smirnov test was used to filter artefacts : 2% of the epochs were rejected from our data set. The integral local deformation (ILD) time-embedding window algorithm was implemented to evaluate the deterministic structure of the data ; in this work a validation was carried out against known systems .
To allow comparisons with published results, the data analysis relied on a 0-40 Hz power spectrum to identify drug-related signals in the raw EEG data. Efforts were concentrated in distinguishing a clear effect in the EEG down to the lowest administered dose of clonidine.
where the plateau of the graph is lying high around 0.5.
where the plateau of the graph is lying low around 0.2.
From a time-embedding perspective there is no local minimum, that is to say no time delay τ which induces a minimal deviation. The ILD curves systematically reach a plateau after a few iterations. The dimension was arbitrarily fixed at m = 12 since all the plots produced with a dimension estimation m greater than 10 had their plateaux converging.
(null activity, high b) periods are regular throughout the experiment (e.g. time intervals B, M, O, S). They explain most of the main wide peaks in the plot.
(null activity, low b) periods are distributed over the time course (e.g. time intervals A, D, F, H, J, L, N, P, R, T, V). They cause the recurrent flat parts in the plot.
(non-null activity, high b) periods are quite short and randomly localized (e.g. time intervals C, E, G, I, K, Q, U), generating rather sharp peaks.
(non-null activity, low b) periods are quasi inexistent in the experiments.
A common technique to detect drug impacts on EEGs relies on sleep stage scores : time proportions of sleep stage periods  or powers per bands [23, 32] have demonstrated pharmacological effects. Ideally, from a cost and time perspective, the pharmacodynamic effect of a novel drug should be assessed in parallel to standard first in human (FIH) trials. This could potentially be done with a sensitive automatic analysis based on standard wake EEG recordings. However clinical sleep architecture studies tend to be more expensive, mainly due to the time it takes to run these studies in specialist sleep centres. The detection of a pharmacological response directly from the EEG trace would therefore be the preferred solution in an early clinical trial, if achievable.
Nevertheless, links to sleep stages are detailed here for comparison purposes but not retained as part of the final nonlinear analysis workflow. The normalised powers of the different bands per sleep-wake stage during the control condition also confirm the classic distribution of the different frequency bands. For the vehicle dose, a high power in the δ band is associated with the occurrence of NREM sleep stages and a θ band power dominates in REM sleep from a fronto-parietal electrode configuration in rodents. The vehicle plot in Figure 5 shows that NREM epochs are synchronised with a high δ and low b values (bottom right) whereas REM epochs display lower δ but higher b values (top left). One could potentially rely of the localisation of each point in the different regions of the graph to perform an a posteriori sleep stage epoch scoring.
Clonidine has been shown to reduce the firing of noradrenergic neurons in the Locus Coeruleus (LC) via autoreceptors, thus altering the input to the cortex and causing a change in the texture of the recorded EEG signal, which might be reflected in the non-linear parameter b [10, 33].
Averaged normalised powers in spectral bands for the distinct doses and sleep-wake stages during the first hour
NREM powers in %
REM powers in %
WAKE powers in %
Given a new drug, there is always a possibility that its unknown effect might corrupt the usual sleep stage classifiers that are well defined for healthy subjects, but less so for preclinical sleep scoring; as an example of changes in the EEG signal irrespective of sleep stage see .
Assumptions solely formed on sleep scores would then be inconclusive. Continuous characterisations like the nonlinear one presented above are likely to be more robust in these particular cases.
This work fully supports the idea that nonlinear techniques are valuable solutions for analysing EEG. The results presented here are consistent as clonidine is known to affect sleep  and nonlinear approaches are known to discern sleep stages . The combination with the standard δ band commonly associated with the awake stage reinforces the plausibility of having extracted a genuine signal from this data set.
In conclusion, this study describes a generic mathematical framework that can extract a pharmacodynamic profile of clonidine from raw EEG data collected from a fronto-parietal electrode derivation in freely-moving rodents, which could offer an alternative approach to study drug effects in early clinical trials. The model is based on a numerical analysis tool whose field of application is nonlinear dynamic systems. This innovative approach can potentially provide a translatable analysis methodology for assessing central pharmacodynamic effect and bridge the preclinical and clinical EEG observations.
- Aston-Jones G, Chiang C, Alexinsky T: Discharge of noradrenergic locus coeruleus neurons in behaving rats and monkeys suggests a role in vigilance. Progress in brain research. 1991, 88: 501-520.View ArticleGoogle Scholar
- Aston-Jones G, Shipley M, Chouvet G, Ennis M, van Bockstaele E, Pieribone V, Shiekhattar R, Akaoka H, Drolet G, Astier B: Afferent regulation of locus coeruleus neurons: anatomy, physiology and pharmacology. Progress in brain research. 1991, 88: 47-75.View ArticleGoogle Scholar
- Berridge CW: Noradrenergic modulation of arousal. Brain Research Reviews. 2008, 8: 1-17.View ArticleGoogle Scholar
- De Sarro G, Ascioti C, Froio F, Libri V, Nisticò G: Evidence that locus coeruleus is the site where clonidine and drugs acting at alpha 1- and alpha 2-adrenoceptors affect sleep and arousal mechanisms. British Journal of Pharmacology. 1987, 90: 675-685.View ArticleGoogle Scholar
- Nisticó G, De Sarro G, G GB, Mollace V: Altered sensitivity of alpha 2-adrenoceptors in the brain during aging in rats. Annals of the New York Academy of Sciences. 1992, 673: 206-213. 10.1111/j.1749-6632.1992.tb27455.x.View ArticleADSGoogle Scholar
- Itil TM, Itil KZ: Central mechanism of clonidine and propranolol in man. Chest. 1983, 83: 411-416. [Supplement]Google Scholar
- Bischoff P, Scharein E, Schmidt GN, von Knobelsdorff G, Bromm B, am Esch JS: Topography of clonidine-induced electroencephalographic changes evaluated by principal component analysis. Anesthesiology. 2000, 72: 1545-1552.View ArticleGoogle Scholar
- Bischoff P, Schmidt GN, Scharein E, Bromm B, am Esch JS: Clonidine induced sedation and analgesia. Journal of Neurology. 2004, 251: 219-221. 10.1007/s00415-004-0283-9.View ArticleGoogle Scholar
- Miyazaki S, Uchida S, Mukai J, Nishihara K: Clonidine effects on all-night human sleep: Opposite action of low- and medium-dose clonidine on human NREM-REM sleep proportion. Psychiatry and Clinical Neurosciences. 2004, 58: 138-144. 10.1111/j.1440-1819.2003.01207.x.View ArticleGoogle Scholar
- Seidel WF, Maze M, Dement WC, Edgar DM: Alpha-2 adrenergic modulation of sleep: time-of-day-dependent pharmacodynamic profiles of dexmedetomidine and clonidine in the rat. Journal of Pharmacology and Experimental Therapeutics. 1995, 275: 263-273.Google Scholar
- Sebban C, Zhang XQ, Tesolin-Decros B, Millan M, Spedding M: Changes in EEG spectral power in the prefrontal cortex of conscious rats elicited by drugs interacting with dopaminergic and noradrenergic transmission. British Journal of Pharmacology. 1999, 128: 1045-1054. 10.1038/sj.bjp.0702894.View ArticleGoogle Scholar
- Klonowski W: Everything you wanted to ask about EEG but were afraid to get the right answer. Nonlinear Biomedical Physics. 2009, 3: 2-10.1186/1753-4631-3-2.View ArticleGoogle Scholar
- Stam C: Nonlinear dynamical analysis of EEG and MEG: review of an emerging field. Clinical Neurophysiology. 2005, 116: 2266-2301. 10.1016/j.clinph.2005.06.011.View ArticleGoogle Scholar
- Grassberger P, Procaccia I: Measuring the strangeness of strange attractors. Physica. 1983, 189-208.Google Scholar
- Rapp P, Zimmerman I, Albano A, Deguzman G, Greenbaun N: Dynamics of spontaneous neural activity in the simian motor cortex: The dimension of chaotic neurons. Physics Letters A. 1985, 110 (6): 335-338. 10.1016/0375-9601(85)90786-8.View ArticleADSGoogle Scholar
- Mayer-Kress G, Layne SP: Dimension of the Human Electroencephalogram. Annals of the New York Academy of Sciences. 1987, 504: 62-87. 10.1111/j.1749-6632.1987.tb48726.x.View ArticleADSGoogle Scholar
- Pereda E, Gamundi A, Rial R, González J: Non-linear behaviour of human EEG: fractal exponent versus correlation dimension in awake and sleep stages. Neuroscience Letters. 1998, 250: 91-94. 10.1016/S0304-3940(98)00435-2.View ArticleGoogle Scholar
- Kobayashi T, Misaki K, Nakagawa H, Madokoro S, Ota T, Ihara H, Tsuda K, Umezawa Y, Murayama J, Isaki K: Correlation dimension of the human sleep electroencephalogram. Psychiatry and Clinical Neuroscience. 2000, 54: 11-16. 10.1046/j.1440-1819.2000.00629.x.View ArticleGoogle Scholar
- Shen Y, Olbrich E, Achermann P, Meier P: Dimensional complexity and spectral properties of the human sleep EEG. Clinical Neurophysiology. 2003, 114: 199-209. 10.1016/S1388-2457(02)00338-3.View ArticleGoogle Scholar
- Kunhimangalam R, Joseph P, Sujith O: Nonlinear analysis of EEG signals: Surrogate data analysis. Recherche et Ingénierie Biomédicale. 2008, 29: 239-244.Google Scholar
- Lee YJ, Zhu YS, Xu YH, Shen MF, Zhang HX, Thakor N: Detection of non-linearity in the EEG of schizophrenic patient. Clinical Neurophysiology. 2001, 112: 1288-1294. 10.1016/S1388-2457(01)00544-2.View ArticleGoogle Scholar
- Widman G, Schreiber T, Rehberg B, Hoeft A, Elger C: Quantification of depth of anesthesia by nonlinear time series analysis of brain electrical activity. Physical Review E. 2000, 62: 4898-4903. 10.1103/PhysRevE.62.4898.View ArticleADSGoogle Scholar
- Kantor S, Jakus R, R RB, Halasz P, Bagdy G: Acute and long-term effects of the 5-HT2 receptor antagonist ritanserin on EEG power spectra, motor activity, and sleep: changes at the light-dark phase shift. Brain research. 2002, 943: 105-111. 10.1016/S0006-8993(02)02698-7.View ArticleGoogle Scholar
- Ivarsson M, Paterson LM, Hutson PH: Antidepressants and REM in Wistar-Kyoto and Sprague-Dawley rats. European Journal of Pharmacology. 2005, 522: 63-71. 10.1016/j.ejphar.2005.08.050.View ArticleGoogle Scholar
- Paterson LM, Wilson SJ, Nutt DJ, Hutson PH, Ivarsson M: A translational, caffeine-induced model of onset insomnia in rats and healthy volunteers. Psychopharmacology. 2007, 191: 943-950. 10.1007/s00213-006-0672-0.View ArticleGoogle Scholar
- Delorme A, Makeig S: EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. Journal of Neuroscience Methods. 2004, 134: 9-21. 10.1016/j.jneumeth.2003.10.009. [Http://sccn.ucsd.edu/eeglab/]View ArticleGoogle Scholar
- Buzug T, Pfister G: Optimal delay time and embedding dimension for delay-time coordinates by analysis of the global static and local dynamical behavior of strange attractors. Physical review A. 1992, 45 (10): 7073-7088. 10.1103/PhysRevA.45.7073.View ArticleADSGoogle Scholar
- Galka A: Topics in nonlinear time series analysis with implications for EEG analysis. 2000, Advanced series in nonlinear dynamics, World ScientificGoogle Scholar
- Conway EL, Jarrott B: Clonidine distribution in the rat: temporal relationship between tissue levels and blood pressure response. British Journal of Pharmacology. 1980, 71: 473-478.View ArticleGoogle Scholar
- Conway EL, Jarrott B: Tissue Pharmacokinetics and Biopharmaceutics. Journal of Pharmacokinetics and Biopharmaceutics. 1982, 10 (2): 187-200. 10.1007/BF01062335.View ArticleGoogle Scholar
- Duda RO, Hart PE, Stork DG: Pattern classification. 2001, John Wiley and Sons, secondGoogle Scholar
- Bjorvtan B, Fagerland S, Ursin R: EEG power densities (0.5-20 Hz) in different sleep-wake stages in rats. Physiology and Behavior. 1998, 63: 413-417. 10.1016/S0031-9384(97)00460-5.View ArticleGoogle Scholar
- Mallick BN, Singh S, Pal D: Role of alpha and beta adrenoceptors in locus coeruleus stimulation-induced reduction in rapid eye movement sleep in freely moving rats. Behavioural Brain Research. 2005, 158: 9-21. 10.1016/j.bbr.2004.08.004.View ArticleGoogle Scholar
- Winsky-Sommerer R, Vyazovskiy VV, Homanics GE, Tobler I: The EEG effects of THIP (Gaboxadol) on sleep and waking are mediated by the GABAA δ-subunit-containing receptors. European Journal of Neuroscience. 2007, 25: 1893-1899. 10.1111/j.1460-9568.2007.05455.x.View ArticleGoogle Scholar
- Prince JB, Wilens TE, Biederman J, Spencer T, Wozniak JR: Clonidine for sleep disturbances associated with attention-deficit hyperactivity disorder: a systematic chart review of 62 cases. Journal of the American Academy of Child and Adolescent Psychiatry. 1996, 35: 599-605. 10.1097/00004583-199605000-00014.View ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.