Propofol Affects Cortico-Hippocampal Interactions via β3 Subunit-Containing GABAA Receptors

Background: General anesthetics depress neuronal activity. The depression and uncoupling of cortico-hippocampal activity may contribute to anesthetic-induced amnesia. However, the molecular targets involved in this process are not fully characterized. GABAA receptors, especially the type with β3 subunits, represent a main molecular target of propofol. We therefore hypothesized that GABAA receptors with β3 subunits mediate the propofol-induced disturbance of cortico-hippocampal interactions. Methods: We used local field potential (LFP) recordings from chronically implanted cortical and hippocampal electrodes in wild-type and β3(N265M) knock-in mice. In the β3(N265M) mice, the action of propofol via β3subunit containing GABAA receptors is strongly attenuated. The analytical approach contained spectral power, phase locking, and mutual information analyses in the 2–16 Hz range to investigate propofol-induced effects on cortico-hippocampal interactions. Results: Propofol caused a significant increase in spectral power between 14 and 16 Hz in the cortex and hippocampus of wild-type mice. This increase was absent in the β3(N265M) mutant. Propofol strongly decreased phase locking of 6–12 Hz oscillations in wild-type mice. This decrease was attenuated in the β3(N265M) mutant. Finally, propofol reduced the mutual information between 6–16 Hz in wild-type mice, but only between 6 and 8 Hz in the β3(N265M) mutant. Conclusions: GABAA receptors containing β3 subunits contribute to frequency-specific perturbation of cortico-hippocampal interactions. This likely explains some of the amnestic actions of propofol.


Introduction
Commonly used general anesthetics such as propofol influence information processing in the brain. Hypnotic doses of this anesthetic, for instance, impair the transfer of information between the frontal and parietal cortical regions [1][2][3] as well as between the thalamus and the cortex [4,5]. While impairment of thalamocortical activity may relate to the hypnotic component of anesthesia, propofol-induced changes in cortico-hippocampal interactions may contribute to the amnestic component of this drug. Memory-related processes are strongly linked to interactions between these areas [6]. Both regions interact with each other via neural oscillating activity in the mid-frequency (4-12 Hz) range [7]. Volatile anesthetics cause a slowing of the peak oscillation frequency in the hippocampus. This may represent one mechanism of anesthetic-induced amnesia [8,9]. It is difficult to determine which specific molecular pathways are involved in this process, because volatile anesthetics cause many changes via multiple molecular targets in the brain, e.g., potassium channels, HCN channels, acetylcholine receptors, glutamate receptors and GABA A receptors. [10][11][12][13]. Compared to volatile anesthetics, propofol is much more specific for GABA A receptors [11,14]. The anesthetic actions of propofol are mediated almost exclusively via βsubunit containing GABA A receptors [15,16]. Using genetically engineered mice, it was shown that β2subunit-containing GABA A receptors, a frequently expressed subtype of GABA A receptors [17] are involved in the sedative properties of intravenous anesthetics [18,19]. On the other hand, β3 subunit-containing GABA A receptors, being only a minor subtype of GABA A receptors [17], largely contribute to the hypnotic and immobilizing effects of propofol [20]. Furthermore, the electrophysiological effects of propofol in the cerebral cortex critically depend on the presence of β3 subunit-containing GABA A receptors [21]. Therefore, we used wild-type and β3(N265M) knock-in mutant mice to elucidate the impact of β3 subunit-containing GABA A receptors on the effects of propofol on cortico-hippocampal interactions. The β3(N265M) mice are largely resistant to the actions of propofol at this receptor-subtype [22]. We hypothesized that the GABAergic anesthetic propofol would reduce connectivity between the cortex and hippocampus and that this reduction in connectivity is at least in part mediated by β3 subunit-containing GABA A receptors. We tested this hypothesis by (i) analyzing the local field potential (LFP) spectral power in (prefrontal) cortex and hippocampus, (ii) the interaction between these brain areas using phase locking and (iii) the shared information content using the mutual information.

Results
We did not observe significant differences in body weight of the animals from both groups. (WT: 33.4 ± 5.0 g vs. β3(N265M): 36.4 ± 5.9 g). Animals were recorded under awake control conditions for ten minutes, injected i.v. with propofol, and immediately monitored again until they showed recovery from anesthesia in their LFP recordings. Head-fixation impeded with simultaneous test of return of righting reflex (RoRR). Therefore, time to RoRR after propofol injection was measured in separate experiments. We determined RoRR after approximately 10 min in wild-type and 2 min in β3(N265M)-mutant mice.

Effects of Propofol on the Power Spectrum of Cortical and Hippocampal Oscillations in Wild-Type and β3(N265M) Knock-in Mice
Propofol did not change spectral power in the 4-12 Hz range at cortical recording sites in wild-type mice, yet there was a significant increase in power in frequencies above 12 Hz. This increase in higher-frequency power was also observed in data recorded from hippocampal sites. The most striking difference between the hippocampus and cortex, however, was a strong activation of hippocampal but not of cortical oscillations in the 4-6 Hz range. For the β3(N265M) mice, the results are different. Propofol did not have a significant effect on the spectral band power in cortex. The activation of 4-6 Hz oscillatory activity we observed in the wild-type mice in the hippocampus was also absent in the β3(N265M) animals. However, in β3(N265M) mice, we observed a small but significant increase in 8-10 Hz band power. The detailed effects of propofol on spectral band power are presented in Figure 1. PSD error bar plots of the 2 Hz frequency ranges (mean ± sem). (A) LFP recorded from the cortex in the wild-type mice. Propofol has a significant and medium effect on the 12-14 Hz and a significant and strong effect on the 14-16 Hz (Hedges' g: 1.01; CI: (0.38 to 1.96)) frequency range. In both frequency ranges, propofol causes an increase in relative power. (B) There is no significant effect of propofol on the PSD of LFP activity recorded from cortex in the β3(N265M) group. (C) In the hippocampal LFP recordings from wild-types, propofol significantly and strongly increased relative power in 4-6 Hz range and in the 14-16 Hz range (Hedges' g: 0.90; CI: (0.26 to 1.78)). (D) Propofol caused a significant and strong increase in relative power in the 8-10 Hz range of LFP recorded from the hippocampus of the β3(N265M) mice. The ivory areas indicate significance, i.e., a 95% confidence interval of Hedge's g exclusive 0; wild-type (n = 4), β3(N265M) knock-in mice (n = 3); PSD: power spectral density. LFP: local field potential, CI: 95% confidence interval.

Actions of Propofol on Interactions between Cortex and Hippocampus in Wild-Type and β3(N265M) Knock-in Mice
In the next step, we focused on changes in interactions between the cortex and hippocampus using the phase locking value (PLV). In wild-type mice, we found a strong de-synchronization of phase in the 6 to 12 Hz range, indicating that propofol impairs the oscillatory connectivity between the cortex and hippocampus. The β3(N265M) knock-in mice did not show that strong a decrease in PLV. In these animals, PLV showed a significant and strong decrease only in a smaller frequency Figure 1. PSD error bar plots of the 2 Hz frequency ranges (mean ± sem). (A) LFP recorded from the cortex in the wild-type mice. Propofol has a significant and medium effect on the 12-14 Hz and a significant and strong effect on the 14-16 Hz (Hedges' g: 1.01; CI: (0.38 to 1.96)) frequency range. In both frequency ranges, propofol causes an increase in relative power. (B) There is no significant effect of propofol on the PSD of LFP activity recorded from cortex in the β3(N265M) group. (C) In the hippocampal LFP recordings from wild-types, propofol significantly and strongly increased relative power in 4-6 Hz range and in the 14-16 Hz range (Hedges' g: 0.90; CI: (0.26 to 1.78)). (D) Propofol caused a significant and strong increase in relative power in the 8-10 Hz range of LFP recorded from the hippocampus of the β3(N265M) mice. The ivory areas indicate significance, i.e., a 95% confidence interval of Hedge's g exclusive 0; wild-type (n = 4), β3(N265M) knock-in mice (n = 3); PSD: power spectral density. LFP: local field potential, CI: 95% confidence interval.

Actions of Propofol on Interactions between Cortex and Hippocampus in Wild-Type and β3(N265M) Knock-in Mice
In the next step, we focused on changes in interactions between the cortex and hippocampus using the phase locking value (PLV). In wild-type mice, we found a strong de-synchronization of phase in the 6 to 12 Hz range, indicating that propofol impairs the oscillatory connectivity between the cortex and hippocampus. The β3(N265M) knock-in mice did not show that strong a decrease in PLV. In these animals, PLV showed a significant and strong decrease only in a smaller frequency window from 6-8 Hz. Figure 2 displays the PLV for the wild-type and β3(N265M) mice for all frequency ranges. This demonstrates that GABA A receptors containing the β3 subunit might be involved in communicating neuronal information between the cortex and hippocampus in the 8-14 Hz range while receiving propofol sedation/anesthesia. window from 6-8 Hz. Figure 2 displays the PLV for the wild-type and β3(N265M) mice for all frequency ranges. This demonstrates that GABAA receptors containing the β3 subunit might be involved in communicating neuronal information between the cortex and hippocampus in the 8-14 Hz range while receiving propofol sedation/anesthesia.

Effects of Propofol on Shared Information Content between Cortex and Hippocampus in Wild-Type and β3(N265M) Knock-in Mice
Finally, the use of mutual information (MI) served the purpose to estimate propofol-induced changes of shared information between cortex and hippocampus. In wild-type mice, propofol caused a strong reduction in shared information in frequencies from 6-16 Hz. In contrast, MI in β3(N265M) knock-in mice did not show a decrease to that degree, and the effect was limited to a smaller frequency range from 6 to 8 Hz. This is in line with the finding for PLV, where the effects in the β3(N265M) knock-ins were attenuated, i.e., limited to a smaller frequency range as well. Figure 3 shows the effect of propofol on MI in the single frequency ranges for wild-type and β3(N265M) knock-in mice. (B) Propofol only shows a significant (strong: g = 1.06; CI: (0.38, 2.85)) effect on PLV in the 6-8 Hz range in the β3(N265M) mutant group. The ivory areas indicate significance, i.e., a 95% confidence interval of Hedge's g exclusive 0; wild-type (n = 4), β3(N265M) knock-in mice (n = 3); PLV: phase locking value; CI: 95% confidence interval.

Effects of Propofol on Shared Information Content between Cortex and Hippocampus in Wild-Type and β3(N265M) Knock-in Mice
Finally, the use of mutual information (MI) served the purpose to estimate propofol-induced changes of shared information between cortex and hippocampus. In wild-type mice, propofol caused a strong reduction in shared information in frequencies from 6-16 Hz. In contrast, MI in β3(N265M) knock-in mice did not show a decrease to that degree, and the effect was limited to a smaller frequency range from 6 to 8 Hz. This is in line with the finding for PLV, where the effects in the β3(N265M) knock-ins were attenuated, i.e., limited to a smaller frequency range as well. Figure 3 shows the effect of propofol on MI in the single frequency ranges for wild-type and β3(N265M) knock-in mice.  Table 1.   Table 1.

Discussion
We present results regarding different effects of propofol on cortico-hippocampal interactions in two animal models, i.e., in wild type and β3(N265M) knock-in mice. The animals showed an attenuated response to propofol. Investigations of anesthetic-induced modulation of interactions between brain regions have been in the focus of research for some decades. In the mid-1990s, MacIver and colleagues demonstrated that the intravenous anesthetic thiopental affects cortico-hippocampal synchronicity in a dose-dependent manner [23]. More recently, Perouansky and co-workers showed that volatile anesthetics slow hippocampal oscillations in the 4-12 Hz range, thereby inducing amnesia [8]. This knowledge is extended by the present study by showing that the GABAergic anesthetic propofol also affects hippocampal oscillatory activity in this frequency range, weakens cortico-hippocampal interactions and lowers the shared information content between these two brain regions. Based on the observed differences between wild-type and β3(N265M) mutant mice, we conclude that this process critically depends on GABA A receptors containing β3 subunits.
Our experiments served the purpose to investigate propofol-induced effects on interactions between cortex and hippocampus in the range from 4 to 12 Hz that are strongly related to processes that are necessary for memory consolidation [24]. The cortex and the hippocampus play an important role in memory formation. The observed effects of propofol on oscillatory activity in these regions are likely to be associated with the amnestic component of anesthesia. Rodent studies have also demonstrated that parts of the frontal cortex and hippocampus are among the cerebral regions that are most sensitive to anesthetics [11]. The actions of propofol in the CNS are almost exclusively mediated via GABA A receptors containing β subunits [15,25]. There are three different isoforms of β subunits (β1-β3), resulting in at least three different subtypes of GABA A receptors as defined by their β subunits [26,27]. Currently, studies measuring behavioral endpoints of the actions of propofol on β1 subunit-containing GABA A receptors are limited. However, roughly a decade ago, two landmark studies by Reynolds et al. [18] and Jurd et al. [20] provided evidence that sedation by intravenous anesthetics is primarily mediated by β2 subunit-containing GABA A receptors, while hypnosis and immobility involves β3 subunit-containing GABA A receptors. Furthermore, findings using cortical slice cultures from β3(N265M) knock-in mice demonstrate that the typical electrophysiological actions of propofol-the prolongation of inhibitory postsynaptic currents-critically depend on the presence of the β3 subunit-containing GABA A receptor subtype [21]. With this in mind, we evaluated the actions of propofol on cortico-hippocampal interactions in wild-type and β3(N265M) knock-in mice to elucidate the impact of β3 subunit-containing GABA A receptors.
As mentioned above, the cortex is strongly associated with working, i.e., short-term, memory tasks, but also requires interactions with other brain areas like thalamus and hippocampus [28,29]. In order to consolidate memory, the hippocampus is required. Siapas et al. described that activity in the 4-10 Hz frequency range seems important to gate information flow across cortico-hippocampal circuits. The subsequent plastic changes are believed to underlie the storage of information across these networks [30].
We structured our experiments in three steps. The first investigation was to evaluate propofol-induced changes in spectral band-power in the cortex and hippocampus. The analysis of spectral band-power revealed that propofol does not affect oscillations in the mid-frequency range in the cortex. This result complies with findings from other studies showing that propofol, at least at low concentrations, does not affect the (prefrontal) cortex [31]. In our experiments in wild-type mice, propofol caused a strong activation in the 4-6 Hz range, relating to the theta range, in the hippocampus. This observation is in line with previous experimental findings. Perouansky et al. discovered a slowing of oscillations in the 4-12 Hz range in the rat's hippocampus as an effect of subanesthetic doses of inhaled anesthetics. It was hypothesized that this slowing led to impaired learning. The authors concluded that modulating hippocampal θ-oscillations may present a key process of anesthetic-induced amnesia [8]. Pan et al. found a crucial frequency of around 6.5 Hz for hippocampal oscillations. If peak frequency drops below 6.5 Hz, learning impairments occur. This may be due to a slowing of these θ-frequencies or due to a reduction in high frequency θ-oscillations [32]. The activation of low θ-frequencies was absent in the β3(N265M) knock-in mice. This finding may indicate that β3 subunit-containing GABA A receptors, which are heavily expressed in both the cortex and hippocampus [33][34][35][36] are necessary to mediate propofol-induced amnestic actions.
The use of PLV can help to determine effects on oscillation synchrony between cortex and hippocampus and the concept of MI may help to detect changes in the amount of information that is shared by these regions through oscillatory activity in the selected frequency range. The next step consisted of an analysis of cortico-hippocampal interactions using PLV to evaluate phase synchrony and hence the strength of connectivity between the two recording sites. This synchrony may be necessary for interactions between different cerebral structures [37]. Recently, a study using PLV showed that phase coupling in the 3-9 Hz range between the cortex and visual area after a visual stimulus is enhanced during active working memory processes [38]. Furthermore, synchrony between the cortex and hippocampus increases during spatial working memory tasks [39], underlining the importance of synchrony for mnemonic processes. Our analysis was targeted at cortico-hippocampal interactions, possibly reflecting memory consolidation processes rather than working memory activity. Propofol negatively affected cortico-hippocampal PLV in the 6-12 Hz range, i.e., synchrony and hence strength of connection decreased. Propofol also negatively influenced PLV in β3(N265M) knock-in mice, but to a lesser degree as compared to the wild-type, since only PLV in the 6-8 Hz range was significantly reduced.
MI analysis indicated significantly reduced shared information content between the cortex and hippocampus with propofol in the 6-16 Hz oscillation range in wild-type animals. This reduction may be explained by the observed drop in PLV at these very frequencies with propofol. Synchrony and hence possibly the strength of cortico-hippocampal interaction become reduced. This reduction consequently leads to reduced information that is shared between both regions. The shared information is also reduced by propofol in the β3(N265M) mice, but to a lesser degree and in a narrower frequency range as in the wild-types.

Limitations
First of all, we are aware of our small sample size. Nevertheless, we found significant and strong effects that highlight the differences in LFP activity between the strains. The delivery of a single bolus dose of propofol via the tail vein leads to a peak and then decreasing concentrations of propofol in the brain. Hence, we did not investigate the effects at steady state conditions, which is very difficult to do in mice, but rather at selected, artifact-free and non-suppressed episodes that followed the (burst) suppression states after propofol delivery. Our findings highlight differences in LFP activity between the wild type and β3(N265M) mutant mice. LFP effects at distinct propofol concentrations need to be investigated in the future.
We also could not investigate the return of the righting reflex, because the head of the animals was fixated. We also did not perform behavioral experiments to investigate memory performance. This should also be subject to further investigations. Finally, we focused on studying a single point mutation in the β3 subunit of the GABA A receptor. The β3 subunit was selected, as the β3-containing GABA A receptors have been shown to be of major importance for the molecular and behavioral actions of propofol [20,22].

Materials and Methods
Every endeavor was made to minimize both the suffering and the number of mice sacrificed for the study. All experiments and procedures performed were approved by the local animal care committee (1 January 2007-31 December 2010, Eberhard-Karls-University, Tübingen, Germany) and in accordance with the German law on animal experimentation. Wild-type (n = 4) and GABA A receptor β3(N265M) knock-in mice (n = 3) of the same genetic background (87.5% 129 × 1/SvJ, 12.5% 129/Sv) of both sexes were used [20]. The β3(N265M) point mutation renders β3 subunit-containing GABA A receptors insensitive to propofol, but not towards its natural ligand GABA [20,22].

Surgical Procedures
The electrode implantation procedure was described in detail in previous publications [28,[40][41][42][43]. During general anesthesia with isoflurane, a warming apparatus (Fine Science Tools, Heidelberg, Germany) kept body temperature close to 36 • C. The electrode arrays, consisting of four linear electrodes, were implanted after craniotomy using a stereotactic frame, two arrays in each animal on one hemisphere. The cortical array was placed 200 µm below the pia, the other array was placed 900 µm below the pia to target the hippocampus. Electrode arrays were fixed along with their connecting wires to the skull using dental cement (Flowline, Heraeus Kulzer, Hanau, Germany). Finally, the skin incision was sutured and attached to the implant. Post-operatively, the animals were kept warm and analgesia was achieved using carprofen.

Electrophysiology
Custom-built microelectrode arrays were used. A 1 × 4 array of polyimide tubing (HV Technologies, Trenton, GA, USA) served as a frame for electrodes made of platinum and tungsten that was glass coated. Electrodes (Thomas Recording, Giessen, Germany) had a shank diameter of 80 µm, a metal core diameter of 23 µm, and a free tip length of 10 µm and impedance was higher than 1 MΩ. The inter-tip distance in the assembled arrays was around 300 µm. Teflon-insulated silver wires (Science Products, Hofheim, Germany) were soldered to the electrodes on one end and connected to a micro-plug (Bürklin, Munich, Germany). Local field potentials were recorded using a multichannel extracellular amplifier (MultiChannelSystems, Reutlingen, Germany; gain 5000, digital sampling rate 20 kHz) in a head-restraint apparatus.

Recording and Pre-Processing of Local Field Potentials
Adult wild-type and β3(N265M) mice were housed individually for at least one week in a room maintained on a 12-hr light/dark cycle with free access to food and water before electrophysiological recordings. Propofol (10 mg/kg, B.Braun, Melsungen, Germany) was injected intravenously through the tail vein. Electrophysiological recording was started after placing the animal in a head-restraint apparatus. For each experiment, neuronal activity was monitored for ten minutes immediately before propofol treatment (baseline condition) and after propofol administration. The wild-type mice entered a state of isoelectric suppression and required around 50 min to return to baseline LFP activity. A similar state was reached by the mutant mice, already at around 20 min. In the effective period, LFP burst suppression was observed, without ever developing isoelectric LFP.
After the termination of the experiments, the mice were deeply anesthetized and an intracardial perfusion of 0.1 M phosphate puffer prior to 4% paraformaldehyde in phosphate buffer was delivered. The recording sites were marked by electrolytic lesions of the brain tissue, which were identified in the Nissl-stained coronal sections [28].
The initial step of signal preprocessing contained a 100 Hz low pass filtering with a MATLAB (Mathworks, Natick, MA, USA)-based Butterworth filtering routine and a down sampling from f s = 20 kHz to f ds = 250 Hz. We present all used LFP episodes in Figure A1. For analysis in the distinct frequency ranges, the recordings were additionally band pass filtered with the MATLAB routine. We used the filtfilt function, allowing filtering without phase shift. Then, we extracted a simultaneous artifact-free 5 s LFP episode from each of the eight channels from all experiments for analysis. In our analyses, we focused on the 2-16 Hz frequency range of the LFP. In order to evaluate effects in the single oscillation frequencies, we investigated oscillatory activity in non-overlapping 2 Hz frequency bands.

Spectral Power
We used MATLAB to calculate the relative band power for each frequency band represented in the histogram. First, the power spectral density of the 5 s LFP episode over the entire frequency range and for each channel was estimated using the fast Fourier transform. The absolute power spectral density vector was divided through its sum to obtain the relative power. The sum of relative power in the desired frequency range represents the relative band power. Relative band power was calculated for both regions, cortex and hippocampus, to investigate propofol induced effects.

Phase Locking
We used the phase locking value (PLV) to investigate interactions between cortex and hippocampus. Lachaux et al. introduced PLV to detect synchronic behavior in a defined frequency range between two channels [37]. PLV calculation is based on the analytic phase ϕ(t) relationship between both channels. ϕ(t) describes the momentary position of the oscillation for each data point. In contrast to Lachaux, we used the Hilbert transform [44] instead of the wavelet approach to calculate ϕ(t). PLV is based on the difference in analytic phase between two channels, here the difference in the hippocampal and prefrontal cortical analytic phase, ϕ hippocampus (t) and ϕ cortex (t). The details are described in the next section. Figure A2 shows exemplary signals that are either completely phase locked or completely independent.

Description of Hilbert Transform and PLV
The analytic signal X(t) of a recorded and filtered signal x(t), here the envelope of the LFP in the respective frequency range, can be derived with the Hilbert transform [44]. In order to correctly apply this method, the LFP has to be filtered to a narrow frequency range [45], here the described frequency bins of 2 Hz width.
is a series of complex numbers with x RE (t) as the real part and x IM (t) as the imaginary part. The real part x RE (t) complies with the original signal x(t) and x IM (t) complies with x(t) after a 90 • phase shift. Based on X(t), the analytic amplitude A(t) and the analytic phase ϕ(t) can be calculated.
A(t) represents the envelope of x(t) and is derived by A(t) = x RE (t) 2 + x IM (t) 2 . ϕ(t), also called the instantaneous phase, is calculated as φ(t) = arctan(x IM (t)/x RE (t)). The difference Φ(t) is defined as Φ(t) = φ hippocampus (t) − φ cortex (t) PLV can be consequently calculated according to PLV = 1 N exp(iΦ(t)) where N represents the number of data points in the time series.
PLV ranges between 1 and 0 and high values indicate little variance in phase difference, i.e., high synchronization and possibly strong connectivity, whereas a PLV close to 0 indicates that the signals are not connected.
We calculated and averaged the PLV from the extracted 5 s LFP episodes over 5 non-overlapping 1 s (N = 250) segments. Figure A2 presents a graphical description of the PLV calculation.

Mutual Information
Mutual information (MI) evaluates shared information content between two recording sites [46] i.e., the dependence of two channels. The higher the MI, the higher the dependence between two channels. An MI of zero describes two independent channels. MI was calculated over the entire 5 s segments for all hippocampal-cortical channel combinations.
MI calculation is based on the evaluation of probability distributions-in our case, either distribution of LFP amplitudes of a sequence of LFP amplitudes in both selected channels. Therefore, in a first step, the filtered 5 s LFP sequences were normalized, i.e., presented as z-scores . x[t] is a discrete time series, here the filtered and down sampled 5 s LFP episode. In order to generate the amplitude-based probability distributions and the LFP amplitude values of the normalized series x n [t], we used 10 equally spaced bins, as in Figure A3. The width of one bin was the maximum amplitude range, i.e., max(x n [t])-min(x n [t]) divided by the number of bins. For each of the two channels, x n [t] and y n [t], the probability distributions p(x) and p(y) are calculated based on their bin histograms as displayed in Figure A3. Furthermore, the joint probabilities p(x,y) are determined. The mutual information can be estimated by: MI = x∈N y∈N p(x, y) log p(x,y) p(x),p(y) , where N is the number of bins [47].

Statistics
We used the Hedge's g effect size measures for dependent and independent samples. Calculation was performed with the MATLAB based MES toolbox [48]. We also performed a 10,000-fold bootstrapping to generate 95% confidence intervals (CI)-95% CI excluding zero indicates significance at p < 0.05. As a rule of thumb, absolute Hedge's g values > 0.8 indicate a strong effect and values above 0.5 indicate a medium effect [49]. Hence, we discuss the results from the Hedge's g analyses according to this notation, i.e., we describe a strong effect if g > 0.8 and a medium effect if g > 0.5.

Conclusions
Our results suggest that the observed propofol-induced changes in interactions between cortex and hippocampus are largely mediated by β3 subunit-containing GABA A receptors and that this GABA A receptor subtype potentially participates in the process of inducing amnesia by propofol. The decrease in spectral band-power in the 4-6 Hz range was absent in β3(N265M) knock-in mice and the reduction in PLV and MI in our analyses of the θ-frequency range was not as pronounced as in the wild-types.   Figure A2. Example for the calculation of PLV for two signals of "locked phase", i.e., a perfect PLV = one and two signals with "fading phase", i.e., no phase locking leading to a PLV = 0.001 as well as for real prefrontal cortical and hippocampal LFP. (A) "Locked phase" represents the association between the original (black) signal and a signal of same frequency and amplitude shifted by 270 degrees. "Fading phase" represents the association between the original (black) signal and a signal of higher frequency (yellow) leading to a constantly changing phase relationship. (B) The polar representation of the "locked phase" shows that at all times, the phase difference resides in the 270-degree corridor. For the "fading phase", the phase difference is equally distributed over all angles. (C) LFP recordings simultaneously recorded from the prefrontal cortex and the hippocampus that were filtered to the 10-   . Example for the calculation of PLV for two signals of "locked phase", i.e., a perfect PLV = one and two signals with "fading phase", i.e., no phase locking leading to a PLV = 0.001 as well as for real prefrontal cortical and hippocampal LFP. (A) "Locked phase" represents the association between the original (black) signal and a signal of same frequency and amplitude shifted by 270 degrees. "Fading phase" represents the association between the original (black) signal and a signal of higher frequency (yellow) leading to a constantly changing phase relationship. (B) The polar representation of the "locked phase" shows that at all times, the phase difference resides in the 270-degree corridor. For the "fading phase", the phase difference is equally distributed over all angles. (C) LFP recordings simultaneously recorded from the prefrontal cortex and the hippocampus that were filtered to the 10-12 Hz range. (D) The polar representation of the phase differences shows an accumulation of observations in the upper right quadrant. The resulting PLV is 0.465. PLV: phase locking value; LFP: local field potential. Figure A2. Example for the calculation of PLV for two signals of "locked phase", i.e., a perfect PLV = one and two signals with "fading phase", i.e., no phase locking leading to a PLV = 0.001 as well as for real prefrontal cortical and hippocampal LFP. (A) "Locked phase" represents the association between the original (black) signal and a signal of same frequency and amplitude shifted by 270 degrees. "Fading phase" represents the association between the original (black) signal and a signal of higher frequency (yellow) leading to a constantly changing phase relationship. (B) The polar representation of the "locked phase" shows that at all times, the phase difference resides in the 270-degree corridor. For the "fading phase", the phase difference is equally distributed over all angles. (C) LFP recordings simultaneously recorded from the prefrontal cortex and the hippocampus that were filtered to the 10-