This study applied CCG analysis for the first time to investigate the concentration-dependent effects of desflurane anesthesia on putatively classified monosynaptic excitatory and inhibitory functional connections in vivo. We found that anesthesia decreased the number of active units and the absolute number of functional connections they formed. Anesthesia also reduced excitatory connection strength that reflects the efficacy of synaptic transmission. Nevertheless, at a depth of anesthesia that purportedly corresponds to unconsciousness, a significant increase in the ratio of excitatory and inhibitory connection probabilities occurred. The latter change suggests an imbalance of excitatory and inhibitory functional connectivity that may indicate abnormal synaptic communication patterns in the state of suppressed consciousness.
Desflurane was chosen for this study because it is a modern and commonly used anesthetic with favorable pharmacokinetic and pharmacodynamic properties and minimal cardiovascular side effects (Eger and Johnson, 1987
). The anesthetic actions of desflurane are in most respects similar to those of isoflurane (Rehberg et al., 1999
; Murrell et al., 2008
), with the exception of its rapid equilibration, which makes desflurane a preferred choice for experiments to be performed at multiple steady-state anesthetic depths in the same experimental setting.
The effect of anesthesia on spike transmission probabilities in intact cortical neuronal networks, in vivo
, has not been investigated. Previous studies using in vitro
, whole cell or single-channel recordings have established that most anesthetics enhance inhibitory and suppress excitatory synaptic transmission by modulating ligand-gated ion channels (Pearce et al., 1989
; Ries and Puil, 1999
; Pittson et al., 2004
). Our results are consistent with these observations in that inhibitory connection strength was resistant to desflurane, whereas excitatory connection strength was decreased in a concentration-dependent manner. Minor differences with the in vitro
data, such as the lack of enhancement of inhibitory connection strength, are understandable due to the recurrent nature of excitatory and inhibitory interactions in local circuits in vivo
In contrast, the significantly higher excitatory to inhibitory connection probability (E/I balance) at 6% desflurane anesthesia was unexpected. This effect could be due to a change in spike patterns or circuit properties, perhaps to a reduced absolute number (but not strength or efficiency) of inhibitory connections. The difference between the changes in excitatory connection probability and excitatory connection strength is understandable given the different nature of the two parameters. Connection probability measures the frequency of occurrence of functional connections relative to the number of all possible pairs of active units. Connection strength, on the other hand, characterizes the efficacy of spike transmission for each identified functional connection. Thus, it is possible to encounter higher connection probability at lower transmission efficiency.
We also found that the decrease in connection strength could not be directly accounted for by the decrease in spike rate of either the source or the target cells. One explanation for the decreased connection strength may be the change in firing pattern of the presynaptic cell. Short interspike-intervals between pairs of spikes in the presynaptic cells have been shown to more robustly discharge their postsynaptic target (Usrey et al., 1998
; Kara and Reid, 2003
). Anesthetic modulation of spike pattern variability and its consequent effect on connection strength and connection probability may be investigated in the future.
The change in E/I balance observed at 6% desflurane concentration is important. It has been suggested (Shew et al., 2011
) that information transmission and information capacity are maximized at intermediate E/I. An alteration of the E/I balance, in particular elevated E/I, may impair information processing as observed in psychiatric disorders (Yizhar et al., 2011
), suppress memory retrieval and recall (Wang and Zochowski, 2012
) and reduce sensory-motor integration as observed in evoked responses with lower doses of anesthetics (Populin, 2005
). An elevation of E/I from its optimal value may represent insufficient suppression of excitation by inhibition, and was shown to result in excessive correlation, also referred to as hypersynchrony, between neurons (Shew et al., 2011
). In fact, our results revealed an increase in excitatory connection probability at the deepest anesthetic level. As suggested by Buzsaki and colleagues, the lack of inhibition could create an unstable system resulting in an avalanche of excitation (Buzsaki, 2006
; Buzsaki et al., 2007
) that is incompatible with meaningful information processing. Stereotypic hypersynchronous activity is commonly observed in deep anesthesia characterized by electroencephalography (EEG) burst-suppression and is thought to have limited information capacity (Alkire et al., 2008
). In burst-suppression, the cortex displays brief periods of increased activity followed by electrically silent periods. It is accompanied by cortical hyperexcitability through reduced inhibition, therefore causing a shift in the E/I balance (Hudetz and Imas, 2007
; Amzica, 2009
; Ferron et al., 2009
). In our experiments at desflurane concentrations up to 6%, there was no burst-suppression in recorded local field potentials, suggesting that the state of hyperexcitability was not attained. It is possible that neuronal firing may have acquired a bursting pattern—as another form of hyperexcitability, although this was previously observed under deep urethane anesthesia only (Erchova et al., 2002
). This possiblity should be tested in additional studies in the future.
The present results demonstrate that CCG analysis can extract putative monosynaptic connections of functionally interacting neuron pairs at distances up to 1200 μm in the rat visual cortex in both wakefulness and under anesthesia. We also observed a more pronounced deflection in the jittered CCG histogram of inhibitory than excitatory connections. This is consistent with previous in vivo
and in vitro
studies and reflects the slower timecourse of inhibitory postsynaptic potentials (PSPs) relative to excitatory PSPs (Thomson et al., 1996
; Tamas et al., 1997
; Bartho et al., 2004
; Fujisawa et al., 2008
). Similar to previous studies (Bartho et al., 2004
; Buzsaki, 2004
; Fujisawa et al., 2008
), most functional connections were close-range (<200 μm) and found within the same electrode contact. During wakefulness, most excitatory connections projected upward toward more superficial layers consistent with that seen in the auditory cortex of identified pyramidal cells (Crochet and Petersen, 2009
; Sakata and Harris, 2009
). In the anesthetized condition, the spread of excitatory activation was confined to smaller cortical depths suggesting a reduction in information transmission across cortical layers. The reduction in the spatial dispersion of monosynaptic connections may therefore be another indication of reduced cortical communication and integration associated with the anesthetic induced unconsciousness.
We referred to anesthesia at 6% desflurane as a state of unconsciousness. Arguably, consciousness cannot be directly assessed; we can only measure a behavioral surrogate. In rats, the righting reflex is a widely used behavioral index of consciousness because it is abolished at equivalent anesthetic concentrations to those that abolish response to verbal commands in human subjects (Franks, 2008
). The desflurane concentration that suppresses the righting reflex has been previously determined as 4.6 ± 0.45% (Imas et al., 2005b
). The experiments were conducted starting with the anesthetized condition and finishing with the wakeful condition. We chose this order of conditions to the initial threshold selection for spike detection under anesthesia, when signal-to-noise ratio was optimal. Thus, strictly speaking, we investigated the neuronal events associated with regaining as opposed to losing consciousness. During emergence from anesthesia, the threshold for righting reflex may be slightly lower than during induction (Friedman et al., 2010
), indicating hysteresis or “neuronal inertia”. Because our experiments were conducted under steady-state conditions with relatively long equilibration periods before each recording, a hysteresis effect was very unlikely. In preliminary studies with similar equilibration periods, we observed no significant difference in spike rate or interspike intervals between induction and emergence conditions at the same anesthetic concentration. Therefore, 4–6% is a good estimation of range of desflurane concentration at which a reversible transition between consciousness and unconsciousness occurred.
In general, anesthetic drugs target various ligand-gated and voltage-gated ion channels that regulate synaptic transmission (Rudolph and Antkowiak, 2004
; Franks, 2006
; Alkire et al., 2008
) and it is difficult to extrapolate the effect of one agent to that of another. Depression of neuronal excitability has been observed with various anesthetics (Hentschke et al., 2005
; Schumacher et al., 2011
), thought to be primarily caused by enhanced synaptic inhibition at γ-aminobutyric acid A (GABAA
) receptors (Banks and Pearce, 1999
; Bieda and Maciver, 2004
) producing hyperpolarization. A suppression of inhibitory neurotransmitter release (Maclver et al., 1996
) and the anesthetic modulation of sodium and potassium channels (Hemmings et al., 2005
) may also contribute to reduced excitability. Previously, we showed a suppression of baseline and long-latency cortical neuronal responses to stimuli under desflurane in vivo
(Hudetz et al., 2009
). Inhalational anesthetics such as isoflurane and sevoflurane, as well as intravenous anesthetics propofol and midazolam, and perhaps α-chloralose (Garrett and Gan, 1998
), primarily GABAA
potentiators similar to desflurane, may produce comparable results. Generalization to other types of anesthetics, such as ketamine or urethane, with substantially different ionic mechanisms and targets (Harrison and Simmonds, 1985
; Hara and Harris, 2002
; Sceniak and Maciver, 2006
) is not straightforward.
As with all similar studies, a recognized technical limitation is the undersampling of the neuronal population. Although we were able to simultaneously record approximately 70 units during wakefulness in each experiment, this number represents a small percentage of active neurons in the sampled region. Because recorded spike amplitudes are attenuated exponentially with distance, 60–100 neurons could be reliably recorded within a 60 μm radius in the rat hippocampus and medial prefrontal cortex (Buzsaki, 2004
; Fujisawa et al., 2008
). Assuming similar spike amplitude attenuation in the visual cortex, under optimal conditions, we were able to isolate 1–3 units per electrode contact, representing 1–3% of the total possible units. Possible reasons for the relatively low number of recorded cells in our experiments include the potential damage to cells by insertion of the electrode and the possible insulated nature of silicon probe shank that reduces the number of observable neurons (Moffitt and McIntyre, 2005
). Furthermore, the spike rate decreased in a concentration-dependent manner, thus reducing the number of active units (>1 s−1
) used for CCG analysis. The limited number of recorded cells and concentration-dependent change in spike rate may also account for the relatively low percentage of classified monosynaptic connections relative to all possible connected cell pairs as identified by CCG analysis, consistent with previous findings (Csicsvari et al., 1998
; Fujisawa et al., 2008
). Another limitation of the pairwise CCG analysis is that it cannot account for the effects of possible indirect connections (Gerstein and Perkel, 1969
) although the time scale of interactions (~1–5 ms) makes the contribution of multi-synaptic effects unlikely. Recent studies have shown that pairwise analysis may represent the correlated states of a network surprisingly well both in vitro
(Schneidman et al., 2006
; Shlens et al., 2006
; Tang et al., 2008
) and in vivo
(Yu et al., 2008
). Therefore, the pairwise CCG method should represent a reasonable first approximation of population activity.
In summary, our results demonstrate that general anesthesia by desflurane at a concentration that induces unconsciousness alters the excitatory/inhibitory balance of monosynaptic interactions in rat visual cortex neurons in vivo. The elevation of the excitatory-inhibitory balance may result from altered spike firing variability, therefore reducing the efficacy of excitatory transmission among the neurons. Overall, elucidating the effect of general anesthesia on functional communication between cortical neuronal cells should help better understand how changes in spikes modulate population activity as a function of cortical state and awareness.