Differential Effects of Deep Sedation with Propofol on the Specific and Nonspecific Thalamocortical Systems: A Functional Magnetic Resonance Imaging Study

• Disruption of neuronal information integration, especially involving the thalamocortical system, may be a common mechanism of anesthetic-induced unconsciousness

• Using functional magnetic resonance imaging, the authors demonstrated that propofol conferred differential changes in functional connectivity of the specific and nonspecific thalamocortical systems, particularly in left hemisphere, consistent with the verbal nature of stimuli and task

• The changes in nonspecific thalamocortical connectivity may correlate with the loss and return of consciousness

How general anesthetics suppress consciousness is not fully understood. Recent conceptual approaches focus on the effect of anesthetics on integrated information as the defining property of human consciousness.1,2According to a theory, the brain’s capacity for information integration3,4is determined by the repertoire of its states (information) and the causal interaction of its elements (integration). A reduction of either component, for example, due to anesthesia, could result in a reduction of the level of consciousness. Accordingly, general anesthetics exert a preferential effect on integrative processes of the brain, as opposed to simply a block of sensory transmission or reactivity during reduced consciousness.1As we have proposed, under general anesthesia, information in the brain may be “received but not perceived.”5 

It has also been suggested that the thalamocortical system plays a central role in information integration in the brain.6–8In particular, the two major divisions of the thalamus, the specific relay nuclei and the more diffusely projecting “nonspecific” nuclei, may collaborate to accomplish this task,9–11with the specific system responsible for the transmission and encoding of sensory and motor information and the nonspecific system engaged in the control of cortical arousal and temporal conjunction of information across distributed cortical areas.11These considerations emphasize the importance of the nonspecific thalamocortical system in information integration and raise the possibility that its dysfunction may be a primary, and possibly unitary, mechanism of anesthetic-induced unconsciousness.

Previous studies implied a critical involvement of the nonspecific (intralaminar) thalamic nuclei in the loss and recovery of consciousness in patients in vegetative12or minimally conscious state13and in anesthetized animals.14One of these studies also emphasized the importance of intralaminar thalamocortical functional connectivity in the recovery of consciousness.12However, the role of nonspecific thalamic nuclei under general anesthesia in humans has not been investigated. Therefore, the goal of this work was to examine the effect of propofol sedation on functional connectivity of the specific and nonspecific thalamocortical systems.

We chose to investigate thalamocortical functional connectivity using blood oxygen level-dependent (BOLD) functional magnetic resonance imaging (fMRI) in healthy human volunteers. Functional connectivity has been defined as the temporal correlation of BOLD signals among spatially remote regions of the brain15and is currently a favored approach of neuroimaging. To learn about functional connectivity changes upon loss of consciousness, we targeted an anesthetic depth at which responses to verbal commands and auditory verbal memory were suppressed, but auditory cortical sensory reactivity was preserved. We hypothesized that, under this condition, nonspecific thalamocortical connectivity would be disrupted more than specific thalamocortical connectivity, consistent with a failure of cortical integration but not sensory information transmission. Functional connectivity was evaluated in four conditions (wakeful baseline, light sedation, deep sedation, and recovery) while volunteers listened to word lists that were later used to assess their implicit memory. Seed regions used for functional connectivity analysis were manually defined within the specific (medial dorsal, ventral lateral, ventral posterior, and other) and nonspecific (centromedian and parafascicular) thalamic nuclei. We found that propofol sedation indeed produced distinct changes in the functional connectivity of the two divisions of the thalamocortical system, consistent with their postulated roles in information and integration in specific states of human consciousness.

Eight volunteers of both sexes (four men and four women; aged 24–42 yr; body mass index < 25) provided written informed consent to participate in this study. Experimental protocols were approved by the Institutional Review Board of the Medical College of Wisconsin (Milwaukee, WI). The study participants were native English speakers from Medical College of Wisconsin communities, free of drug administration, and with no history of neurological or psychiatric conditions or structural brain abnormalities.

All participants were made to lie in the scanner and instructed to listen to and try to remember (encode) a distinct set of 40 high-frequency English words (nouns) presented during each of the four experimental sessions in four different states of consciousness: wakeful baseline, light sedation, deep sedation, and recovery (fig. 1A). The rate of word presentation was approximately seven words per minute with a random interstimulus interval during each of the approximately 6-min scanning sessions. BOLD time course data were obtained from the entire duration of word presentation without interruption.16Participants were informed that their recall/recognition performance of words heard during each of the four experimental sessions would be assessed after scanning. The experimental sessions were separated by approximately 15 min for experimental preparations.

The anesthetic agent, propofol, was administered by a bolus followed by a target-controlled continuous infusion (STANPUMP).17We targeted plasma concentration of 1 μg/ml for light sedation and 2 μg/ml for deep sedation. The higher dose for deep sedation was chosen to achieve the desired endpoints of unresponsiveness to verbal commands, intended to induce a global loss of memory.18A behavioral assessment of the level of consciousness was performed right before the start of each fMRI scan. Study participants were asked to respond to questions like “how are doing?,”“take a deep breath,” and “can you squeeze my hand?” by the anesthesiologist through a speaker (the one used in the task or at the bedside of the scanner between recording sessions). In general, during light sedation, participants had lethargic responses to the questions. During deep sedation, participants showed no response to verbal commands. Once a desired sedative state was achieved, the next fMRI scan was initiated. During the scan, the sedation level was maintained by computer-controlled infusion with the preset plasma concentration. Immediately after the scanning of the last sedated state, the administration of propofol was stopped. On the confirmation that participants had recovered responsiveness to verbal commands, the last scanning session defined as “recovery” was initiated. Every participant had two intravenous catheters, one for propofol administration and another for the withdrawal of blood samples for measuring propofol plasma concentration. However, due to a problem with red blood cell lysis, the actually plasma propofol concentrations could not be determined in this study. The order of light and deep sedation was counterbalanced in participants, with four participants receiving the low dose before a high dose and the other four receiving them in a reversed order. Standard American Society of Anesthesiologists monitoring was performed during the experiment, including electrocardiogram, noninvasive blood pressure cuff, pulse oximetry, and end-tidal carbon dioxide gas analysis.

The auditory verbal material was presented using a headphone set (Koss Corporation, Milwaukee, WI) designed to work in the magnetic resonance scanner environment. The word lists were matched for the maximum number of letters, frequency of usage in English, concreteness, and imageability (Paivio, Yuille, and Madigan norms for 925 nouns).19Approximately 20–30 min after the completion of all the experiments (after study participants had been taken out of the scanner), all participants completed a free-recall test followed by a forced-choice recognition test. The time separation between the memory tests and the last scan was intended to neutralize the primacy and recency effects, which occur when the recognition memory test is administered immediately after the stimulus presentations. In the forced-choice recognition test, participants were presented auditorily 320 words, of which 160 words were what they had heard during the experiment and the other 160 words were foils or distractors. Participants were required to press a button if they thought they had already heard the word and another button if the word was new. Participants were instructed to make a decision regarding every presented word as quickly as possible. Each participant’s residual memory to words heard during each of the four experimental sessions was assessed by a few performance indices: the percentage of recalled words, the recognition ratio versus  chance, and the discriminability index (d’ ). Of these, the value of d’ , computed from the hit (correct recognition) rate and false-alarm rate, provides a criterion-independent measure of the internal response of participants (i.e. , regardless of how conservative or liberal participants are in making decisions).20A d’  equal to zero indicates same hit and false-alarm rates, and a d’  value significantly greater than zero indicates a higher hit than a false-alarm rate (a d’  value of three represents minimal overlap between the probability of the occurrence of hit and false alarm).

Imaging acquisition was performed using a 1.5 Tesla GE Signa scanner (General Electric Medical Systems, Milwaukee, WI) with a locally designed gradient and radio frequency coil. Potential head movements were minimized using a chin support system developed at the Medical College of Wisconsin. Functional echo-planar images were obtained using whole-brain imaging in the sagittal plane during each task session (repetition time, 2000 ms; echo time, 40 ms; thickness, 6 mm; in-plane resolution, 3.75 × 3.75 mm; 22 slices; flip angle, 90°; field of view, 24 cm; matrix size, 64 × 64). High-resolution spoiled-gradient–recalled anatomical images were always acquired after the third experimental session for each participant (repetition time, 24 ms; echo time, 5 ms; slice thickness, 1.2 mm; flip angle, 40°; field of view, 24 cm; matrix size, 256 × 128).

The specific and nonspecific thalamic nuclei to be used as seeds for functional connectivity analysis were determined in the coronal plane editorof each individual’s high-resolution spoiled-gradient–recalled images after transformation into the standard Talairach space. Specifically, the nonspecific seed mainly consisted of the intralaminar nuclei, including the centromedian and parafascicular thalamic nuclei located at the ventromedial corners of the left and right thalami (fig. 1B). Anatomical references that could be used to enhance the accuracy of defining the nonspecific thalamic seed included the lateral maximum stretch point of the third ventricle, red nucleus, and the interthalamic adhesion. These referential structures were identifiable in the high-resolution anatomical images of each participant by properly adjusting the brightness and contrast of image pixels. The remaining parts of the thalamus were used as an aggregate for seeding the specific thalamic connections (fig. 1C).

Imaging data analysis was conducted using the software packages Analysis of Functional NeuroImages (AFNI, Bethesda, MD) and Matlab (The MathWorks, Natick, MA). The high-resolution anatomical images were first manually transformed into the standard Talairach space, followed by coregistering the functional data to the Talairach space in 2-mm cubic voxels (adwarp  in AFNI). Subsequent data preprocessing included despiking, detrending (3dDetrend  in AFNI, using the Legendre polynomials with an order of three), and motion correction (3dvolreg  in AFNI, obtaining three translational and three rotational parameters for each image). The first four points of the voxel time series of each section were discarded to reduce the transient effects. To minimize contaminating signals from the white matter and the cerebrospinal fluid, we extracted the average BOLD signal from these brain structures manually identified across each individual’s anatomical images. We then constructed eight regressors using the signals corresponding to the six motion parameters (obtained during the volume registration), white matter, and cerebrospinal fluid signals for the subsequent analysis.

After data preprocessing, the event-related fMRI time series of all task sessions were analyzed by a general linear regression (3dDeconvolve  in AFNI). The regression analysis takes into consideration the eight regressors representing the contribution of noise artifacts from the motion, white matter, and cerebrospinal fluid. The residual signals were considered representative of the task-induced BOLD responses with the potential contamination minimized. In the next step, the averaged voxel time courses of the predefined specific and nonspecific thalamic seed regions were used separately to perform voxel-wise Pearson cross-correlation (3dfim+  in AFNI) of the whole brain. The Fisher transformation, m = 0.5 × ln(1 +r )/(1−r ), was applied to the obtained correlation coefficients (r ) to normalize the output. Spatial smoothing of the m-values was performed using a 3.5-mm full-width half-maximum Gaussian kernel filter to compensate for intersubject variability.

To evaluate the distribution of the specific and nonspecific thalamocortical functional connections across anatomically defined brain regions, we multiplied the obtained functional connectivity maps with a brain mask provided in AFNI (tt_n27_ez_ml.tlrc). The mask provides a reference template that partitions the whole brain into 116 anatomical regions in a voxel-wise manner in the Talairach space.21The corresponding voxel count of each region is taken as a quantitative measure of regional thalamocortical functional connections in each state of consciousness. We defined two indices, I  ^A and I  ^B , to represent the normalized changes in the voxel count of thalamocortical functional connectivity by contrasting the wakeful states (baseline and recovery) to deep sedation (E. 1) and the recovery to wakeful baseline (E. 2), respectively:

where W , S , and R  represent the region-specific voxel count of the specific or nonspecific thalamocortical functional connections in the states of wakeful baseline, deep sedation, and recovery, respectively.

From the group analysis of the hemodynamic activation at the low propofol dose, we could identify a distinct pattern of cerebral hemodynamic responses particularly in the dorsal medial prefrontal cortex.16This was presumably associated with propofol-induced excitation, which is common at low anesthetic dose.22Behavioral observations during the experiments and postscan memory tests corroborated a certain degree of excitation at the low propofol dose. Because of the difficulty in interpreting the state of consciousness and the relatively large interindividual variability at this dose, the imaging data obtained at the low propofol dose were not fully analyzed. For referential information, we provided the thalamocortical connectivity maps obtained at the low propofol dose in Supplemental Digital Content 1, https://links.lww.com/ALN/A889, fig. S1. These results should be interpreted with caution in the aforementioned context.

The group effects of connectivity maps were evaluated by one-sample t  tests followed by transformation to z  scores. Results considered significant at P  value less than 0.05 after correction for multiple comparisons (AlphaSim  in AFNI, a minimum cluster thresholding of 179 voxels of 2-mm cubic in the Talairach space) were reported. The program AlphaSim  considers both voxel probability thresholding and minimum cluster-size thresholding to estimate the probability of false-positive detection from the frequency count of cluster sizes in each image. The underlying principle of this method is that true regions of activation or connection will tend to occur over contiguous voxels, whereas noise has much less of a tendency to form clusters. The minimum cluster-size threshold (179) was obtained based on a combined mask of the specific and nonspecific thalamic functional connections in normal, awake human participants in the resting state, which we reported previously.23To provide a level of inference about the results, we performed a leave-one-out test on the group functional connectivity analysis for each state of consciousness. The test generates a collection of maps corresponding to eight overlapping subsets of participants and allows statistical group comparisons (paired t  test, participant dependent) between the maps of different conditions.

All participants produced purposeful responses to verbal commands during wakeful baseline, light sedation, and recovery. These responses were absent during deep sedation. The postscan memory tests showed that participants could freely recall a small percentage of words heard in the scanner at baseline (9%± 0.3), light sedation (10%± 0.5), and recovery (14%± 0.35), but not during deep sedation (2%± 0.15, not significant). The auditory forced-choice recognition tests performed outside the scanner showed the same trend. Participants were able to distinguish target words from foils heard at baseline (d’ , 0.9 ± 0.25), light sedation (d’ , 0.6 ± 0.24), and recovery (d’ , 1.5 ± 0.42), but not the words heard during deep sedation (d’ , 0.2 ± 0.15). For further details regarding the task and the performance results, we refer the reader to our former publication.16 

In the baseline condition, auditory verbal stimuli induced significant, mainly left-lateralized BOLD activations in multiple temporal and frontal regions and prominent negative BOLD responses in the posterior cingulate cortex and the precuneus. During deep sedation, most of the activations were suppressed, except in a few brain areas in the superior temporal gyrus, centered at the primary auditory cortex. In addition, the negative BOLD effects expanded in the frontal, temporal, and occipital lobes. After the participants regained behavioral responsiveness, brain activation maps were restored to a pattern similar to those observed at baseline.16 

Thalamocortical functional connectivities showed distinct spatial distributions within the two thalamic divisions (fig. 2). In the baseline condition, specific and nonspecific connections were widely distributed, forming large clusters particularly in the frontal lobe (including the prefrontal and cingulate cortices) and the parietal lobe (fig. 2A and 2D). In both thalamocortical divisions, the frontal cluster extended from the ventral medial frontal cortex up to the posterior segment of the dorsal medial frontal cortex. The functional connectivity of the dorsal and medial frontal cortices was more prominent in the nonspecific than in the specific system. In contrast, functional connectivity of the specific thalamus in the bilateral frontal and temporal cortices, especially in the left-lateralized inferior frontal gyrus and superior temporal gyrus, was more prevalent than that associated with the nonspecific thalamus. The specific thalamocortical connections also formed pronounced clusters in the primary visual cortex, whereas the nonspecific connections did not. Functional connections in the areas of the parietal default mode network (i.e ., the precuneus, posterior cingulate cortex, and retrosplenial cortex) were expressed in both thalamocortical divisions. (See also Supplemental Digital Content 1, https://links.lww.com/ALN/A889, fig. S2, which describes the full correlation maps obtained by averaging across the participants.)

The two thalamocortical systems exhibited substantially different changes in functional connectivity in deep sedation as compared with wakeful baseline (fig. 2Band 2E). The extent of connections measured by the number of significantly correlated voxels was reduced in both the specific and the nonspecific systems across the whole brain. However, the nonspecific connections were substantially more suppressed, leaving only a few scattered spots of connectivity in the middle cingulate and premotor areas. Changes in the specific thalamic connectivity were more modest in terms of both the brain regions involved and the overall voxel count. During recovery, the specific and nonspecific connections were restored to a spatial distribution similar to that seen in the baseline condition. Nevertheless, compared with baseline, the extent of connectivity in both systems was generally increased in recovery (fig. 2Cand 2F).

The region-specific changes of functional connectivity of the two thalamic divisions across wakeful baseline, deep sedation, and recovery were quantified by calculating the number of significantly correlated voxels for 20 anatomical regions that covered most of the involved major structures of the brain (fig. 3). The results confirmed that during deep sedation, specific connectivity was moderately reduced (fig. 3A) and nonspecific connectivity was substantially reduced (fig. 3B). The suppressed functional connections were then restored in recovery. Combining the voxel count data from all regions revealed that the overall reduction from wakeful baseline to deep sedation was significantly larger (P < 0.01, group comparison through a leave-one-out test) in the nonspecific system (79 ± 13%) than in the specific system (43 ± 16%) (fig. 4A). Likewise, the overall increases in functional connectivity from deep sedation to recovery and from baseline to recovery were both significantly larger (P < 0.01 and P < 0.05) in the nonspecific (952 ± 251% and 123 ± 42%, respectively) than in the specific system (178 ± 38% and 58 ± 35%, respectively) (fig. 4Band 4C).

We also examined the interhemispheric distribution of thalamic functional connections (fig. 4Dand 4E). The baseline distribution of the specific and nonspecific thalamocortical connections extended over nearly equal number of voxels in the left and right hemispheres. However, in deep sedation, the extent of connectivity was significantly lesser (P < 0.001) in the left hemisphere than in the right hemisphere, with a left versus  right ratio (in voxel count) of 50% in the specific system and 8% in the nonspecific system. During recovery, the interhemispheric balance of connectivity was restored in the nonspecific thalamocortical system, but not in the specific system. These results suggest that the interhemispheric balance of the nonspecific thalamocortical connectivity correlates with the loss and return of consciousness better than that of the specific thalamocortical connectivity.

For further illustration of the region-specific changes of functional connectivity in the two thalamocortical systems, we calculated the two indices, I  ^A (E. 1) and I  ^B (E. 2). For each selected brain region, I  ^A expresses the reduction in the count of connected voxels during deep sedation relative to those of wakeful baseline and recovery. I  ^B expresses the relative difference in connectivity between recovery and wakeful baseline. The values of both indices range from 0 to 1, indicating a magnitude of change from small to large, respectively. Brain regions associated with the specific and nonspecific systems are then identified by the two indices as two dimensions in a scatter plot (fig. 5). The plot shows that brain regions in the specific and nonspecific systems are strongly segregated, expressing a difference in their relative changes in connectivity during sedation and recovery. Specifically, the brain regions involved in the nonspecific thalamic functional connectivity undergo a greater decrease during sedation and a greater increase during recovery than the brain regions involved in the specific connectivity.

Finally, the issue was raised if the relatively large volume of the specific thalamic seed region rendered the specific system less sensitive to propofol than the nonspecific system that was based on a significantly smaller seed volume. To investigate this possibility, additional connectivity analysis was conducted based on a reduced volume of the specific seed. Specifically, we randomly subsampled the original specific seed in a way so that the number of voxels left in the new specific seed was identical to that in the nonspecific seed for each participant. The results confirmed the findings previously obtained with the original specific seed (see Supplemental Digital Content 1, https://links.lww.com/ALN/A889, fig. S3, which describes the functional connectivity maps of the reduced specific seed).

Current investigations into the neurophysiological mechanisms of consciousness and of general anesthesia mutually inform each other,1,24with the neuronal effects of anesthetics considered in the light of the neural correlates of consciousness, and vice versa . In the framework of the information integration theory of human consciousness,3,4we hypothesize that the two necessary properties of consciousness, information and integration, are sustained respectively by the specific and nonspecific divisions of the thalamocortical system10,11and that anesthetics may exert a differential effect on these systems as they suppress consciousness. Our results show that at a level of propofol deep sedation characterized by preserved cortical sensory reactivity,16the specific thalamocortical network is moderately affected, whereas the nonspecific thalamocortical network is severely suppressed and subsequently reactivated after recovery of consciousness.

There is substantial evidence for the role of the thalamocortical system in integrative information processing in the brain.6,8,11,25–27The rich and reciprocal nature of thalamocortical interconnectivity involving differential thalamic divisions establishes oscillatory circuits across multiple cortical layers. Such organizational structures put the thalamocortical networks in a unique and crucial position for implementing integrative functionalities necessary for the development of conscious experience. Consistent with the neuroanatomical and neurofunctional features of the thalamus, converging evidence from brain lesion,28anesthesia,27electrophysiology,11and stimulation13studies suggests that the thalamocortical system is essential in the maintenance and modulation of the state of consciousness.29 

Previous investigations into the mechanism of anesthetic-induced unconsciousness have examined the possible interruption of thalamocortical information transfer within the relay nuclei30,31or the failure of nonspecific thalamocortical functional connections in enabling the conscious state.12,32–34Nonspecific thalamocortical connections that originate from the matrix cells of the thalamus are particularly dense in the intralaminar nuclei and have been implicated in supporting conscious experience.11,32,33,35,36An early publication37that dates back to 1953 indicated that electrical potentials in the central medial thalamus evoked by peripheral afferent stimulation in awake monkeys were blocked in anesthesia, whereas laterally conducted impulses reached the sensory cortex unimpaired. Subsequent studies showed that a selective lesion in the medial/intralaminar thalamus invariably caused loss of consciousness.28,32,35Pharmacological or electrical stimulations to certain intralaminar nuclei helped restore wakeful behavior in anesthetized animals14and, in one instance, in a minimally conscious patient.13An anesthetic effect on the nonspecific thalamus is suggested by a shift from γ-rhythms to frontal α-rhythms in the electroencephalogram by propofol. The 40 Hz γ-band synchrony of thalamocortical circuits has been proposed as a neural correlate of consciousness.11In addition, anesthetics preferentially suppress feedback versus  feedforward signaling in cortical electrophysiological recordings.38,39It has been proposed that feedforward pathways alone are insufficient for conscious perception that also requires reentrant top-down feedback that mediates conscious integration.40–42Whether such selective changes in corticocortical interactions are a manifestation of direct cortical or thalamic anesthetic effects is not yet clear.

A general observation from most previous studies1,5,24is that during anesthetic sedation, the reactivity of the primary sensory cortices (e.g. , the visual43and auditory16,44,45) to external stimuli is preserved, whereas higher-order integrative processing of the stimuli is suppressed. This is consistent with a differential effect of anesthetics on the known functional partitions of the thalamocortical system. That is, the specific thalamic system that encodes and relays sensory information is essentially preserved in anesthesia, whereas the nonspecific thalamic system that facilitates temporal conjunction, binding, and integration of information in the brain11is consistently disrupted during suppressed consciousness.

Previous reports regarding the role of thalamocortical connectivity in anesthetic suppression of consciousness have shown a degree of conflicting results. For example, the studies by White and Alkire26and Boveroux et al.  46suggested a significant suppression of the thalamocortical connectivity, whereas Mhuircheartaigh et al.  47reported relatively preserved thalamocortical connectivity during propofol sedation. The lack of consensus may be related to the just demonstrated differential effects of propofol on the specific and nonspecific thalamic networks. It is possible that a further differentiation of thalamic nuclei may be necessary. In the current study, only two nonspecific (intralaminar) thalamic nuclei were considered as a seed because they admitted a relatively clear identification from the anatomical images obtained at 1.5 Tesla. Also, the specific nuclei were lumped together with no further differentiation into sensory, motor, and other nuclei. It is known that higher-order thalamic relay nuclei may mediate corticocortical communication linking primary and higher-order sensory cortices.48Thus, a more refined differentiation of thalamocortical functional connectivity at higher magnetic field strength may bring further insight.

A notable result of our current study was that during deep sedation, thalamocortical functional connectivity increased in a number of brain regions. These regions included the middle frontal gyrus as part of the specific system (fig. 3A) and the insula, superior temporal gyrus, and angular gyrus as part of the nonspecific system (fig. 3B). The mechanism causing these connectivity increases is currently unclear. It is possible that it reflects a compensatory or stereotypical response to the pharmacological challenge by propofol.

Although we demonstrated significantly differential effects of propofol on the specific and nonspecific thalamocortical functional connectivities, the degrees of reduction in the two systems, which are necessary to diminish consciousness, cannot be determined from the current study. Obtaining this information would require a dose-dependent study with anesthetic depths graded at a finer scale. In addition, other factors affecting the strength of functional connectivity would have to be considered, such as the experimental protocol (e.g. , resting state46,vs . task),47the anesthetic drug (propofol vs . other agents),26the magnetic field strength of the scanner (affecting the signal-to-noise ratio), and others. Nevertheless, the current study highlights the differential effects of propofol on the two thalamic divisions at least at a deep sedative level.

Similar to all other fMRI studies, our results are constrained by the methodological limitation that they indirectly reflect the neuronal mass activity.49In addition, the experimental protocol included a memory task and auditory stimulation that may have modified the connectivity patterns as compared with those observed in the task-free “resting state.” An assessment of functional connectivity during sensory activation is not without precedence.47One such example can be seen in the study by Mhuircheartaigh et al .47Previously, we also showed the connectivity results of the same type of analysis with task effect regressed out of the data.50The patterns and changes of the thalamocortical connectivity across the different states of consciousness showed little difference as compared with the current results derived from data containing task effect.

Two additional observations in our results deserve comments. First, we saw a prominent left-lateralized suppression of thalamic functional connections in deep sedation, especially in the nonspecific system. This is consistent with the left-lateralized distribution of neural structures involved in verbal processing and memory.51,52Second, the dramatic increase in the extent of functional connections in both thalamic divisions during recovery relative to baseline was unexpected. One would have anticipated an incomplete recovery within the timeframe of the study due to a lingering drug effect, seen as emergence delirium. Nevertheless, our results are confirmed by recent studies. For example, the strength of corticocortical connectivity derived from electroencephalogram recordings showed a precipitous increase upon the regaining of consciousness from anesthesia.53Likewise, brain activity in the anterior cingulate cortex revealed a significant anesthetic-independent increase at the emergence of consciousness as characterized by positron emission tomography imaging.54Although the cause for these large increases in connectivity is unclear, they may indicate the presence of a nonlinear process in regaining consciousness.55They may also be related to the phenomenon of hysteresis commonly observed in general anesthesia. Hysteresis describes the existence of differential characteristic paths of anesthetic-modulated transitions toward and away from unconsciousness. It implies a more stringent energy barrier for regaining consciousness from an anesthetized state, as opposed to losing it from wakeful baseline. A recent study shows that hysteresis cannot be fully explained by pharmacokinetics but reflects an inherent property of the central nervous system to resist transitions in the state of arousal, coined as “neural inertia.”56It is therefore conceivable that the restoration of consciousness in recovery requires more extensive neural processing than its maintenance at wakeful baseline. We speculate that awakening after anesthesia is a constructive process that relies on transiently increased neural communication as the brain “reboots” itself to consciousness.

In summary, we demonstrated a differential effect of propofol on the specific and nonspecific thalamocortical functional connectivities. The results are consistent with the presumed roles of the two thalamic divisions in information and integration as necessary conditions for consciousness. They also strengthen the view that the nonspecific thalamocortical system may play an essential role in the neural basis of consciousness.

The authors thank Stephen M. Rao, Ph.D., A.B.P.P.-Cn., Ralph and Luci Schey Chair and Director of the Schey Center for Cognitive Neuroimaging, Cleveland Clinic, Cleveland, Ohio, for his invaluable advice to the original study. They also thank Carrie M. O’Connor, M.A., Editorial Assistant, Department of Biophysics, and Anita Tredeau, B.S., Administrative Assistant, Department of Anesthesiology, Medical College of Wisconsin, Milwaukee, Wisconsin, for editorial assistance.