Human neural correlates of sevoflurane-induced unconsciousness

Abstract

Sevoflurane, a volatile anaesthetic agent well-tolerated for inhalation induction, provides a useful opportunity to elucidate the processes whereby halogenated ethers disrupt consciousness and cognition. Multiple molecular targets of sevoflurane have been identified, complementing imaging and electrophysiologic markers for the mechanistically obscure progression from wakefulness to unconsciousness. Recent investigations have more precisely detailed scalp EEG activity during this transition, with practical clinical implications. The relative timing of scalp potentials in frontal and parietal EEG signals suggests that sevoflurane might perturb the propagation of neural information between underlying cortical regions. Spatially distributed brain activity during general anaesthesia has been further investigated with positron emission tomography (PET) and resting-state functional magnetic resonance imaging (fMRI). Combined EEG and PET investigations have identified changes in cerebral blood flow and metabolic activity in frontal, parietal, and thalamic regions during sevoflurane-induced loss of consciousness. More recent fMRI investigations have revealed that sevoflurane weakens the signal correlations among brain regions that share functionality and specialization during wakefulness. In particular, two such resting-state networks have shown progressive breakdown in intracortical and thalamocortical connectivity with increasing anaesthetic concentrations: the Default Mode Network (introspection and episodic memory) and the Ventral Attention Network (orienting of attention to salient feature of the external world). These data support the hypotheses that perturbations in temporally correlated activity across brain regions contribute to the transition between states of sevoflurane sedation and general anaesthesia.

Introduction

Sevoflurane and other volatile anaesthetics are the principal agents used for maintaining clinical general anaesthesia. Additionally, sevoflurane is well-tolerated for inhalation induction, providing a means of elucidating the neural changes associated with gradual transitions from wakefulness through sedation and beyond the loss of consciousness. The mechanistic cascade from molecular interactions to the suppression of consciousness has yet to be fully characterized. Potential downstream functional targets for sevoflurane include brain networks associated with attention, cognition, and the maintenance of consciousness. Thus, sevoflurane might serve as a paradigm for understanding mechanisms underlying the therapeutic actions of other halogenated ethers.

Framing the problem of anaesthetic-induced unconsciousness

For over two decades, there has been a systematic search for neural correlates and the underpinnings of conscious experience. One approach to this question relates to information synthesis in the brain, a focus motivated by known phenomenological and neurobiological facts. The known phenomenological fact is that our perception of the world is unified—we do not, for example, experience the colour, shape and warmth of the sun as disconnected elements, but rather as a singular whole. The known neurobiological fact is that the brain is subdivided into discrete functional units that independently process sensory modalities (such as vision) and sub-modalities (such as colour). It is therefore critical that the brain has mechanisms to synthesize discrete neural processing in order to generate the unity of experience. If such synthesis is necessary for normal consciousness, it stands to reason that the interruption of this synthesis would be sufficient for unconsciousness.

“Connectivity” is a surrogate for integration in the brain and can be assessed from neuroimaging or neurophysiological data. Four types of brain connectivity are commonly analyzed:1, 2, 3 (1) structural connectivity, which refers to the synaptic connections between brain regions, (2) functional connectivity, evaluated from the covariation of activities within different brain regions over time, (3) directed connectivity, determined as a statistical interdependence of neural activities in one area relative to another region in the past, and (4) effective connectivity, which uses models to infer a causal relationship between the activities of different brain regions. There are advantages and disadvantages to each of these approaches and the various techniques within each of these categories come with assumptions. In this article, we focus primarily on functional connectivity (e.g. as measured through functional magnetic resonance imaging (fMRI) and directed connectivity (e.g. as measured through EEG).

Models for the effects of general anesthesia on consciousness must accommodate the current framework of how signals in the central nervous system are encoded, transmitted, and decoded across multiple scales of space and time. Neural activity is constrained by the current framework of how information is encoded, transmitted, and decoded in the central nervous system. Neurones at a microscale level temporally integrate inputs and transmit either excitatory or inhibitory output in a binary manner. Both types of neurones serve as building blocks of circuits at a mesoscale level. Circuits distributed across the brain can be localized or distributed across brain structures to provide neural markers at a macroscale level. Models for the integration of neural activity across functionally specialized brain regions incorporate these markers to allow inferences based on the phenotype and behaviour at an organismal level.

Direct and indirect actions on cortical neurones

Sevoflurane binds to protein targets on the surface of neurones, but the precise molecular interactions and the neural structures involved in the induction of unconsciousness have yet to be identified. Electrophysiologic effects correspond to sevoflurane interactions with γ-aminobutyric acid (GABA),4, 5, 6, 7 N-methyl-D-aspartate (NMDA),^{8 9} and nicotinic acetylcholine (ACh) receptors;¹⁰ and voltage-gated sodium channels,¹¹ hyperpolarization-activated cyclic nucleotide-gated (HCN) channels¹² and two-pore domain¹³ potassium channels. Whether the relative expression of these proteins on the surface of cortical neurones contributes directly to the unresponsive phenotype remains unknown. In vivo studies have supported the roles of GABA-A receptors,^{14 15} ACh receptors,^{16 17} and HCN channels18, 19, 20 as potential targets for the hypnotic effects of sevoflurane.

Cortical effects of sevoflurane are also likely mediated through arousal centres in the brain. Sevoflurane directly activates adrenergic neurones in the rat locus coeruleus,²¹ potentially contributing to clinical agitation during induction and emergence from general anaesthesia. There are no reports regarding the effects of sevoflurane on dopaminergic centres in the ventral tegmental area, the histaminergic nuclei of the tubermammillary nucleus of the hypothalamus, or the cholinergic centre of the nucleus basalis in the basal forebrain. The attenuation of excitatory output from these subcortical structures by sevoflurane would have downstream effects on cortical neuronal excitability that are measureable from the scalp when synchronized across large populations of neurones.

Visually detectable frontal EEG markers

Wave-like patterns in the frontal EEG provide an estimation of intraoperative “anaesthetic depth” during sevoflurane surgical anaesthesia. The oscillatory waveforms can be described by the dominant frequencies in the EEG, with conventionally monitored frequency bands between 0.5 Hz and 30 Hz. The proportion of delta (0.5–4 Hz), theta (4 to 8 Hz), and alpha (8 to 13 Hz) power can vary across states of general anaesthesia, with lower frequencies dominating at greater sevoflurane concentrations. Alignment of alpha waves in bilateral EEG tracings can also be observed (Fig. 1A). Both slow (<1 Hz) and higher frequency (1 Hz to 4 Hz) delta waves are prominent at sevoflurane concentrations typical of general anaesthesia [e.g. approximately 0.9 age-adjusted minimum alveolar concentration (MAC)]. Delta oscillations do not appear to be aligned across interhemispheric frontal EEG signals (Fig. 1B). Concentrations of sevoflurane associated with general anaesthesia can also produce burst suppression, persistent suppression,²² and epileptiform activity.²³ The sevoflurane concentration at which these phenomena occur may be lower in patients who have increased sensitivity to volatile anaesthetics (e.g. older adults). All these visually recognizable patterns are informative in isolation or for interpreting processed EEG measures.

Fig 1.

Frontal electroencephalographic patterns during sevoflurane general anaesthesia. (A and B) Simultaneous frontal EEG traces acquired from a single patient during 1 MAC sevoflurane general anesthesia. (A) Red dashed vertical lines highlight the interhemispheric phase alignment of oscillatory activity in the alpha frequency band (8–13 Hz) of these two signals, despite a fluctuating baseline. (B) In contrast, the slow oscillations in the delta frequency band (0.5–4 Hz) appear to be out of phase for the two hemispheres (peaks indicated by white arrows). EEG acquired using bipolar montages F7-Fp1 (EEG1) or F8-Fp2 (EEG2) are displayed at 25 mm/sec (A) or 50 mm/sec (B). (C) Population average power spectrogram of EEG acquired during sevoflurane general anaesthesia shows consistent power over time in the alpha, theta (4-8 Hz), and delta frequency bands. (D) Population average power coherogram of EEG recorded during sevoflurane general anaesthesia illustrates greater consistency within the alpha band compared with theta and delta bands. (E) Single patient frontal EEG power spectrogram during the transition from wakefulness, through sedation, sevoflurane-induced unconsciousness, and recovery. Beta band (13–30 Hz) power emerges after drug initiation (first white vertical dashed lines) and during sedation (red arrowhead). With loss of responsiveness (Blue arrow), beta power persists without any emergence in alpha power. EEG were acquired using F7 and a Hjorth reference. These figures were modified from Akeju and colleagues 2014²⁴ (C-D) and Kaskinoro and colleagues 2015³¹ (E). Reproduced with permissions, Wolters Kluwer Health, Inc. and John Wiley and Sons.

Frontal EEG power and coherence

Quantitative measurements of oscillatory EEG waveforms provide useful markers for neuromonitoring during sevoflurane general anaesthesia. The strength of rhythmic activity is described by the amplitude or power for a particular frequency range. The power can be computed across time segments and graphed as a compressed spectral array or spectrogram, with delta, theta, and alpha oscillations predominant during sevoflurane anaesthesia (Fig. 1C).²⁴

The characteristics of frontal EEG oscillations during the maintenance of sevoflurane general anaesthesia appear to evolve as humans age. The amplitude of the EEG increases as infants age,^{25 26} plateaus in early adulthood,²⁶ and appears to fall off with senescence.²⁷ The background EEG varies even between infants²⁵ and pre-term infants.²⁸ Although oscillations in the delta bands have been observed across age groups, those in the theta and alpha bands are observed only in infants four to six months of age;²⁵ and across the population at age >one yr.²⁶ A decline in the strength of alpha oscillations with age²⁷ has implications for brain monitoring geared toward improving perioperative neurological outcomes.

The coupling of EEG oscillations is computed through calculation of coherence or through bispectral analysis, which has been incorporated into proprietary EEG depth of anesthesia monitoring.²⁹ Phase coherence provides an estimate of the extent to which two signals are in phase at a particular frequency. During sevoflurane general anesthesia, interhemispheric coherence is consistently stronger in the alpha band than at lower frequencies. (Fig. 1D).²⁴ While the interhemispheric coherence in the 3.4–10.7 Hz frequency band is greater with sevoflurane than with propofol anaesthesia, the reverse is true in the 11.7–19.5 Hz band. Additional data from children show that the coherence in the alpha band during sevoflurane anaesthesia develops after one yr of age.²⁶ Overall, these data suggest anaesthetic-specific perturbations in intracortical connectivity and the need of addressing age-related variance in future clinical monitoring technology.

EEG markers for transitions between wakefulness and unconsciousness

Recent studies have elucidated the precise temporal order of EEG evolution accompanying the loss of consciousness induced by sevoflurane.^{30 31} Augmentation of frontal beta power occurs during sedation30, 31, 32 and persists despite loss of responsiveness (Fig. 1E).^{30 31} On the other hand, alpha, theta, and delta band power remain unchanged³⁰ and frontal alpha power does not consistently emerge on loss of consciousness.^{30 31} These findings have been clarified in greater detail;³³ a gradual loss of responsiveness is associated with decline in alpha oscillations that originate from posterior brain regions.^{32 33} Decrements in occipital dominant alpha oscillations are accompanied by augmentation in frontal theta and beta activity. Frontal alpha intensifies at greater sevoflurane concentrations, and can consist of sleep spindle-like activity during the maintenance phase of general anaesthesia.^{32 34} Given that alpha activity in the frontal EEG can change over the transition from wakefulness to sevoflurane-induced unconsciousness, EEG dominated by delta waves or exhibiting burst suppression might be more specific for unconsciousness during general anaesthesia.

Phase-amplitude coupling in frontal and parietal EEG

Similar to how relative phases of sound waves can lead to deadening or amplification based on how they interact, disruptions of phase-amplitude relationships of neural activity might relate to loss of consciousness. During wakefulness, 0.1–1 Hz phase and 8–13 Hz amplitude coupling exists in the parietal but not in the frontal EEG.³⁰ Three different patterns have been observed for slow delta phase-alpha amplitude cross-frequency coupling for frontal, parietal, and frontoparietal scalp EEG. For propofol-induced unconsciousness,^{30 35} the peak of slow (0.1–1 Hz) waves coincides with high amplitudes of 8–14 Hz oscillations (peak-max) in the frontal EEG. With ketamine, no significant 0.1–1 Hz-8–14 Hz phase-amplitude coupling is observed in the frontal EEG during the anaesthetized state.³⁶ Blain-Moraes and colleagues³⁰ assessed the effects of sevoflurane on cross-frequency phase-amplitude coupling in frontal and parietal EEG (Fig 2A). Sevoflurane is similar to ketamine in that no significant changes in cross-frequency coupling are observed in the frontal EEG.³⁰ Instead, the phase-amplitude modulation (0.1–1 Hz and 8-13 Hz) present in the parietal EEG during wakefulness is weakened just beyond the loss of consciousness induced by sevoflurane.³⁰ Sevoflurane could undermine the function of the posterior parietal cortical regions that serve as hubs for the integration of oscillations in the alpha band oscillations.³⁷ Corroborating data from other altered conscious states have implicated a posterior parietal region as a “hot spot” for the neural correlates of consciousness.³⁹

Fig 2.

Sevoflurane-induced unconsciousness and altered cross-frequency and cross-regional coupling. (A) Significant parietal cross-frequency phase-amplitude coupling (Modulation Index) is reduced after induction of unresponsiveness with sevoflurane (transition from the period of sedation, S to unconsciousness, U). The phase of slow delta (0.1–1 Hz) modulates the amplitude of alpha (8–13 Hz) in parietal EEG during wakefulness (W), sedation and recovery (R). This relationship is observed in parietal EEG during unconsciousness or in frontal EEG during any epoch. (B) Alpha (8–13 Hz) phase lag between frontal and posterior cortical regions is present at baseline and reversibly attenuated during sevoflurane unconsciousness. Frontal-parietal (F-P), frontal-temporal (F-T), and frontal-occipital (F-O) phase lag index is a measure of the timing phase difference in EEG signals of corresponding cortical lobes. (C) Directed connectivity between frontal and parietal regions is disrupted, following loss of consciousness (LOC) induced by sevoflurane. Normalized symbolic transfer entropy, a surrogate measure of information flow, is preferentially altered for feedback and not feedforward interactions. These figures were modified from Blain-Moraes and colleagues 2015³⁰ (A-B) and Lee et al., 2013⁴⁴ (C). Reproduced with permissions, Wolters Kluwer Health, Inc.

Intracortical connectivity to frontal regions

Disrupted interactions between frontal and parietal regions have also been implicated as markers or potential mechanisms for sevoflurane-induced loss of consciousness. Projections arising from frontal cortical neurones target virtually all arousal centres, key circuits in the basal ganglia and neurones in temporal, parietal, and occipital regions. These feedback projections from frontal cortex are more numerous than feedforward efferents that presumably propagate information forward from sensory cortical regions. Feedback projections are associated with late evoked responses in sensory cortical neurones and serve putative roles in conscious experience through contextual modulation of sensory information, integration of information, or attention. Thus, the preferential weakening of feedback neural activity has been hypothesized as a substrate for sevoflurane-induced loss of consciousness³⁹ with support from electrode recordings in rodents.40, 41, 42

Recent EEG studies lend support to the preferential loss of alpha frequency band feedback connections from frontal to posterior cortical areas during sevoflurane-induced unconsciousness. The relative timing of activity between different regions can be inferred from phase differences in oscillatory components. Blain-Moraes and colleagues³⁰ reported on the phase relationships among 8-13 Hz alpha oscillations of frontal cortical regions relative to those arising from parietal, occipital, and temporal cortical areas. (Fig. 2B). The phase lag index was reduced when volunteers were rendered unconscious with sevoflurane compared to baseline and recovery periods. This suggests that feedback phase interactions in the alpha band were attenuated by sevoflurane. These findings are consistent with surrogates of information transfer from frontal to parietal regions (Fig. 2C), measured by normalized symbolic transfer entropy in the EEG. Although feedback transfer entropy dominated over feedforward measures during wakefulness in the eyes-closed resting-state, it was weakened to the strength of feedforward information after loss of consciousness induced by sevoflurane. This low-resolution EEG study of symbolic transfer entropy has recently been replicated with high-density recordings acquired during sevoflurane-induced unconsciousness.⁴³ Notably, this suppression of frontal-to-parietal transfer entropy was also found after exposure to propofol or ketamine;⁴⁴ findings have also been confirmed with directed phase lag index in the alpha bandwidth for sevoflurane,³⁸ propofol,⁴⁵ or ketamine.³⁶ Neuroimaging studies have attempted to localize these EEG correlates to specific regions of frontal and parietal cortex, and thalamic structures.

Cerebral blood flow and metabolic activity

Combined EEG-PET studies provide additional support for the premise that disruption of connectivity among frontal, parietal, and subcortical areas contribute to sevoflurane-induced unconsciousness. Kaisti and colleagues⁴⁶ assayed Bispectral Index (BIS) and radiolabeled imaging of cerebral blood flow (CBF) during wakefulness and inhalation of sevoflurane held at inspired concentrations of 2%, 3%, and 4%. Sevoflurane reduced CBF across all cortical regions, with 1 MAC robustly reducing CBF in thalamic and parietal cortical regions. Correlation analysis of BIS and CBF identified superficial and deep frontal and parietal regions (Fig. 3A, upper panel).⁴⁶ The same group reported on correlations between CBF and spectral entropy, the extent of disorder in the power spectrum.⁴⁷ CBF in frontal and parietal cortical regions accounted for anaesthetic-induced changes in midfrontal-central spectral entropy (Fig. 3A, lower panel).⁴⁷ Lower sevoflurane concentrations (1.5%) have been used to probe changes in CBF and metabolic consumption of oxygen during the transition to the unconscious state.⁴⁸ Thalamic, frontal, and parietal regions showed robust reductions in CBF but less overlap with metabolic changes for unknown reasons. Across these studies, varied levels of reproducibility were observed for localization of effects to the thalamus^{46 48} and to the posterior cingulate,46, 47, 48 medial prefrontal,⁴⁷ inferior parietal,46, 47, 48 and dorsolateral frontal46, 47, 48 cortical regions. Many of these cortical regions had already been recognized as elements of the Default Mode Network (DMN), identified through notable increases in CBF and metabolic activity during quiet resting.⁴⁹ The clinical significance of reduced CBF and metabolic activity in the DMN remain unclear but could have implications in postoperative delirium and early postoperative cognitive dysfunction. Many of these same regions show functional neuroimaging abnormalities that correlate with dementia severity.^{50 51}

Fig 3.

Distributed alterations in cerebral blood flow and functional connectivity are localized to cortical and thalamic regions during sevoflurane-induced unconsciousness. (A) Upper panel: Cortical and subcortical regions demonstrate correlated changes in cerebral blood flow (CBF) and Bispectral index (BIS) values during positron emission tomography (PET) across the range of 0, 2, 3, and 4% sevoflurane. Superior, lateral, and medial surface registrations show localization mainly to posterior medial and lateral parietal regions. Lower panel: Similarly, another processed EEG measure, spectral entropy, covaries with CBF to lateral parietal and midline frontal regions during sevoflurane general anesthesia. (B) Resting-state functional connectivity magnetic resonance images showing sevoflurane-induced weakening of correlated activity following loss of response to noxious stimulation. At 1.2% sevoflurane, correlations are weaker between medial prefrontal cortex and a midline posterior parietal cortical region of the default mode network (DMN, Upper panel, left), known as the posterior cingulate cortex. Correlations among frontal and posterior regions of the ventral attention network (VAN, Upper panel, right) are also disrupted. Altered thalamocortical connectivity during inhalation of 1.2% sevoflurane is also observed (Thalamus, Lower panel). (C) Reduced thalamocortical resting-state functional connectivity at 2%, 3%, or burst suppression during sevoflurane general anaesthesia. Dark voxels represent weakening of connectivity relative to baseline wakefulness. While mainly frontoparietal (DMN) connectivity is attenuated at 2% sevoflurane, diffuse changes are observed at higher doses. These figures were modified from Kaisti and colleagues 2002⁴⁶ (A, Upper panel), Maksimow et al., 2005⁴⁷ (A, Lower panel), Palanca and colleagues 2015⁵⁶ (B, Upper and lower panels), Ranft and colleagues 2016⁴³ (C, Upper and lower panels). Reproduced with permission, Wolters Kluwer Health, Inc. and John Wiley and Sons.

To date, six independently acquired fMRI data sets have allowed investigation of sevoflurane effects on surrogates of human brain activity and CBF (Table 1).⁴³ 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64 Although many have used task- or stimulation-based paradigms, we will focus on those probing the intrinsic signals that serve as markers of the functional architecture of the brain.

Table 1.

Descriptive summary of human fMRI investigations of sevoflurane sedation and general anaesthesia. Sample size and principal findings are included. The category of Sedation includes moderate sedation (depression of consciousness but intact response to verbal commands) and deep sedation (purposeful responses and arousability in response to repeated or painful stimulation). General denotes general anaesthesia with loss of consciousness and inability to be aroused even by painful stimulation. Burst Suppression indicates general anaesthesia with intermittent periods of low amplitude EEG consistent with suppression of cortical activity. Matched colours represent reports derived from the same data set.

Study	Anaesthetic level	N	Principal findings
Peltier and colleagues 200552	Sedation General	6	Reduction of interhemispheric resting-state connectivity in motor cortex at 1% and 2%.
Kerssens and colleagues 200559	Sedation General	6	Auditory-evoked cortical activity was reduced at 1% and 2% sevoflurane.
Ramani and colleagues 200753	Sedation	16	Sevoflurane at 0.5% reduced task-evoked activation of visual, supplemental motor, thalamic, and hippocampal regions.
Qiu and colleagues 200863	Sedation	22	At 0.5% sevoflurane, regional cerebral blood flow to superficial cortical regions was reduced but was augmented in the insula and anterior cingulate regions.
Qiu and colleagues 200862	Sedation	16	At 0.5% sevoflurane, visual and auditory regions showed a reduction in resting-state BOLD signals, stimulation-evoked BOLD signals, and cerebral blood flow.
Martuzzi and colleagues 201054	General	14	With 1% sevoflurane, functional connectivity was altered in higher order circuits of memory and pain but not in the default mode or sensory networks.
Deshpande and colleagues 201060	General	6	Sevoflurane (1% and 2%) reduced functional connectivity locally within nodes of the default mode network.
Martuzzi and colleagues 201161	General	14	Sevoflurane (1%) augmented or weakened functional connectivity among cortical areas, depending on the region of interest.
Huang and colleagues 201455	General	6	Functional connectivity within the default mode network was stronger during wakefulness than at 2.6% sevoflurane.
Huang and colleagues 201558	General	6	General anaesthesia accompanied reductions in both the temporal variability of the BOLD signal and the synchronization between distant brain regions.
Palanca and colleagues 201556	General	9	Sevoflurane (1.2%) reduced intracortical and thalamocortical functional connectivity of the default mode and ventral attention networks.
Ranft and colleagues 201643	General Burst Suppression	16	Sevoflurane at 2%, 3%, and burst suppression reduced thalamocortical and intracortical functional connectivity, mirroring EEG measures of information.
Golkowski and colleagues 201764	General Burst Suppression	16	At burst suppression, occipital cortical BOLD signals were anti-correlated with EEG bursts, in contrast to signals from other regions.
Kafashan and colleagues 201657	General	9	Sevoflurane (1.2%) disrupted intracortical spatiotemporal motifs of intermediate strength but correlations within resting-state networks were generally maintained.

Default mode and ventral attention resting-state networks

Resting-state networks (RSNs) are defined as regions of gray matter that demonstrate high temporal correlation among fMRI BOLD (Blood-oxygen-level dependent) signalsin the absence of task performance. This phenomenon, referred to as functional connectivity, presumably links these spatially distributed brain regions into specialized groups that underlie specific aspects of sensation, cognition, and behaviour. Resting-state fMRI is used to investigate the strength of functional connectivity within and between RSNs, on the timescale of seconds to minutes. The DMN is an RSN associated with self-referential cognition and episodic memory.⁶⁵ Disruption of DMN functional connectivity has been observed in coma,⁶⁶ non-rapid eye movement sleep,^{67 68} and disorders of consciousness.^{69 70} Orientation of attention to salient features in the external environment is a putative function of the Ventral Attention Network (VAN), which includes anterior insulae, anterior cingulate cortex, and frontal cortical regions.^{71 72}

There is no clear consensus on whether disruption of particular RSNs is critical for anaesthetic-induced unconsciousness. With propofol, loss of DMN functional connectivity between medial prefrontal and posterior cingulate regions has been reported.⁷³ Further analyses of these data demonstrated a lack of specificity for RSN susceptibility, with reduced integration in motor, sensory, and networks of frontoparietal regions.⁷⁴ These findings contrasted with earlier reports on the effects of sevoflurane on functional connectivity. As the first fMRI study of anaesthetized humans, Peltier and colleagues⁵² reported reduced connectivity within the motor system at 1% and 2% sevoflurane. A subsequent study suggested that DMN connectivity did not change between wakefulness and unresponsive states incurred at 1% sevoflurane.⁵⁴ Secondary analyses noted the largest changes in connectivity for the posterior cingulate and insula.⁶¹ Until recently, it remained unclear whether discrepancies in fMRI literature were attributable to drug-specific differences between propofol and sevoflurane, analytical technique, or motion artifact.

Palanca and colleagues⁵⁶ addressed whether perturbations in specific RSNs accounted for sevoflurane loss of consciousness. Ensemble measures of RSN integrity were computed as the average correlation strength between regions independently assigned to different RSNs. The DMN and VAN were the only RSNs with significant differences between wakefulness (0%) and sevoflurane-induced unconsciousness (1.2%). These findings have been reproduced to some extent by a recent investigation.⁴³ At 2% sevoflurane, frontoparietal and anterior DMN connectivity were reduced, but significant changes in the VAN/salience network were not observed. Connectivity measures for the posterior DMN were attenuated at 3% and at burst suppression. Reduction of correlated activity between anterior and posterior regions of the DMN induced by sevoflurane (Fig. 3B, top),⁵⁶ propofol,^{73 75} and ketamine⁷⁶ highlight a possible common correlate of unconsciousness shared by both anaesthetic-induced and pathologic loss of consciousness. The loss of feedback connectivity in EEG studies and posterior-directed connectivity^{30 44} provide potential electrophysiologic analogues to these weakened correlations among low frequency fMRI BOLD signal fluctuations. On the other hand, sensory RSNs appear to be less susceptible to sevoflurane. Correlated activity within the visual network was attenuated at 3%, while disruption in the auditory RSN was only observed during burst suppression.⁴³

Thalamocortical connectivity

Perturbations of thalamic activity and thalamocortical connectivity have been posited as a mechanism underlying anaesthetic-induced unconsciousness.⁷⁷ The thalamus plays several critical roles, making it a potentially critical target for general anaesthetics. First, thalamic neurones relay sensory information to cortical regions. Second, association nuclei of the thalamus, such as at the medial dorsal nucleus, coordinate communication across higher order cortical regions. Third, centromedian and other portions of the intralaminar nuclei project diffusely across cortical regions and are important for arousal. Fourth, the thalamic reticular nucleus partially enshrouds these nuclei and appears to modulate information in distinct subcircuits based on the state of arousal.⁷⁸ The depth, small size, and large number of these nuclei and subnuclei pose significant impediments to their study by fMRI and PET. Nevertheless, thalamic regions had been implicated by prior PET studies involving sevoflurane.^{46 48} Resting-state fMRI data acquired during sevoflurane-induced unconsciousness show increased thalamic connectivity to the insula and supplementary motor areas.⁵⁴ Reduced thalamic connectivity to the caudate and inferior parietal lobule was also observed. Thus, sevoflurane does not appear to simply reduce the strength of thalamocortical connectivity.

A synthesis of human fMRI and rodent studies supports the roles of particular thalamic nuclei as principal sites of action whereby sevoflurane induces unconsciousness. Using the entire thalamus as a single region, Palanca and colleagues⁵⁶ (Fig. 3B, bottom) demonstrated reduced thalamocortical functional connectivity to both the DMN and VAN. This pattern of susceptibility at 1.2% sevoflurane, which is identical to the selectivity for within-RSN intracortical connectivity, suggests that action at certain nuclei of the thalamus relate to the specificity of changes observed in cortical components of RSNs. Additionally, Ranft and colleagues⁴³ demonstrated that thalamocortical connectivity undergoes progressive weakening at 2%, 3%, and burst suppression doses of sevoflurane (Fig. 3C). An anterior midline region of the human thalamus, the medial dorsal thalamus, projects to elements of both the DMN and the VAN (anterior cingulate cortex).⁷⁹ Furthermore, injections of nicotine¹⁷ or voltage-gated Kv1 potassium channel blockers^{19 20} into the central medial thalamus of rodents return elements of responsiveness in sevoflurane-anaesthetized rodents.²⁰ Presumably, activation of nicotinic Ach receptors or suppression of inhibitory K⁺ conductance of neurones in the central thalamus restores a key node for modulating cortical activity and consciousness across propofol, dexmedetomodine, and sleep.⁸⁰ In rodents, action at this thalamic region might also account for the ability of sevoflurane to induce patterns of EEG activity, intracortical connectivity, and acetylcholine activity that mirror slow wave sleep.⁸¹ Sevoflurane also accelerates the return of slow wave homeostasis after sleep deprivation.⁸² Thalamocortical connectivity may differ between states of sleep and high-dose regimes of sevoflurane; total block of sensory input does not occur even at sevoflurane burst suppression.⁸³ Overall, these lines of evidence suggest that connectivity between cortical RSN components and medial anterior thalamic nuclei are critical to the transition to sevoflurane-induced unconsciousness and serve as a conserved locus involved in other altered states of arousal.

Challenges and future directions

Despite progress in elucidating the actions whereby sevoflurane perturbs consciousness, multiple limitations must be noted. Observational end-points can be imprecise or difficult to interpret given the lack of consensus on the definition of consciousness. In humans, responsiveness is commonly used to infer consciousness, yet the two are dissociable in the context of neuromuscular paralysis.⁸⁴ As a corollary, loss of righting reflex is commonly used as a surrogate for loss of consciousness in rodents, with accepted imprecision. Overall, the field is striving for a principled understanding of consciousness instead of imprecise behavioural end-points.⁸⁵

The differentiation of markers and mechanisms also deserves clarification. Neural mechanisms are more easily demonstrated in animal models whereas electrical and imaging markers of consciousness and cognition are more accessible. Markers of anaesthetic-induced perturbation of cognitive function in humans suggest underlying mechanisms that are more readily explored in animal models. Moreover, while such markers of anaesthetic effect on brain activity lack definitive causal importance, they could be more easily translatable into clinical practice throughout the perioperative period. While reproducible measures might reach statistical significance, the manner in which the brain decodes minute changes across multiple regions remains unclear. Admittedly, the findings in humans hinge on small amplitude fluctuations from baseline (∼1% in the case of resting-state BOLD and on the scale of microvolts in EEG). Furthermore, discordance between occipital BOLD fluctuations and EEG bursts argues against a simple relationship of these two imaging modalities.⁶⁴

Aside from these limitations that span both animal and human systems, many unsolved questions remain regarding anaesthetic action on consciousness and cognitive function. It remains unclear whether there are multiple intracortical and thalamocortical pathways invoked by sevoflurane in the process of ablating consciousness. Isolation of thalamic and cortical influences within corticothalamocortical circuitry could prove difficult. Open questions include whether effects in subcortical arousal systems translate directly to effects observed in RSNs and whether specific thalamocortical circuits are important to the loss and return of consciousness. Much of the recent work has addressed resting-state paradigms to assess the effects of sevoflurane on intrinsic or “spontaneous” neural activity. Stimulation across different sensory modalities might clarify the nature of how consciousness is ablated and reconstituted, with clinical implications beyond anaesthesiology.

Conclusions

The synthesis of molecular, electrophysiologic, metabolic, vascular, and network-based imaging studies will ultimately yield a rich model of anaesthetic-induced unconsciousness, spanning interactions from the molecular level to networks of distributed brain regions. Disrupted functional interactions between frontal and parietal cortical regions, in the context of potential interactions with thalamic nuclei, are consistently reported across imaging studies. However, causal interactions and significance remain unknown. The technical limitations in either localizing (EEG) or temporally resolving (PET and fMRI) alterations in spatially distributed neural activity remain. Further investigations employing sevoflurane are warranted to demonstrate the causality of altered temporal relationships among the activities of frontal, parietal, and thalamic regions and the unconscious state.

Authors’ contributions

Study design/planning: B.J.P., G.A.M.

Writing Paper: all authors

Revising paper: all authors

Declaration of interest

None declared.

Funding

This work was supported by the Foundation for Anesthesia Education and Research (BJP); the Washington University Institute of Clinical and Translational Sciences grant UL1TR000448, sub-award KL2TR000450 from the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health (NIH, Bethesda, M.D., USA) (BJP); and the National Institute of General Medical Sciences (Bethesda, MD, USA) R01GM098578 (GAM). The content is solely the responsibility of the authors and does not necessarily represent the official view of the NIH.

Handling editor: Hugh C. Hemmings Jr