1. Neuroscience
Download icon

Disruption of thalamic functional connectivity is a neural correlate of dexmedetomidine-induced unconsciousness

  1. Oluwaseun Akeju Is a corresponding author
  2. Marco L Loggia Is a corresponding author
  3. Ciprian Catana
  4. Kara J Pavone
  5. Rafael Vazquez
  6. James Rhee
  7. Violeta Contreras Ramirez
  8. Daniel B Chonde
  9. David Izquierdo-Garcia
  10. Grae Arabasz
  11. Shirley Hsu
  12. Kathleen Habeeb
  13. Jacob M Hooker
  14. Vitaly Napadow
  15. Emery N Brown
  16. Patrick L Purdon Is a corresponding author
  1. Massachusetts General Hospital, Harvard Medical School, United States
  2. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, United States
  3. Brigham and Women's Hospital, Harvard Medical School, United States
  4. Massachusetts General Hospital, United States
  5. Massachusetts Institute of Technology, United States
Research Article
Cited
19
Views
2,320
Comments
0
Cite as: eLife 2014;3:e04499 doi: 10.7554/eLife.04499

Abstract

Understanding the neural basis of consciousness is fundamental to neuroscience research. Disruptions in cortico-cortical connectivity have been suggested as a primary mechanism of unconsciousness. By using a novel combination of positron emission tomography and functional magnetic resonance imaging, we studied anesthesia-induced unconsciousness and recovery using the α2-agonist dexmedetomidine. During unconsciousness, cerebral metabolic rate of glucose and cerebral blood flow were preferentially decreased in the thalamus, the Default Mode Network (DMN), and the bilateral Frontoparietal Networks (FPNs). Cortico-cortical functional connectivity within the DMN and FPNs was preserved. However, DMN thalamo-cortical functional connectivity was disrupted. Recovery from this state was associated with sustained reduction in cerebral blood flow and restored DMN thalamo-cortical functional connectivity. We report that loss of thalamo-cortical functional connectivity is sufficient to produce unconsciousness.

https://doi.org/10.7554/eLife.04499.001

eLife digest

Although we are all familiar with the experience of being conscious, explaining precisely what consciousness is and how it arises from activity in the brain remains extremely challenging. Indeed, explaining consciousness is so challenging that it is sometimes referred to as ‘the hard question’ of neuroscience.

One way to obtain insights into the neural basis of consciousness is to compare patterns of activity in the brains of conscious subjects with patterns of brain activity in the same subjects under anesthesia. The results of some experiments of this kind suggest that loss of consciousness occurs when the communication between specific regions within the outer layer of the brain, the cortex, is disrupted. However, other studies seem to contradict these findings by showing that this communication can sometimes remain intact in unconscious subjects.

Akeju, Loggia et al. have now resolved this issue by using brain imaging to examine the changes that occur as healthy volunteers enter and emerge from a light form of anesthesia roughly equivalent to non-REM sleep. An imaging technique called PET revealed that the loss of consciousness in the subjects was accompanied by reduced activity in a structure deep within the brain called the thalamus. Reduced activity was also seen in areas of cortex at the front and back of the brain.

A technique called fMRI showed in turn that communication between the cortex and the thalamus was disrupted as subjects drifted into unconsciousness, whereas communication between cortical regions was spared. As subjects awakened from the anesthesia, communication between the thalamus and the cortex was restored.

These results suggest that changes within distinct brain regions give rise to different depths of unconsciousness. Loss of communication between the thalamus and the cortex generates the unconsciousness of sleep or light anesthesia, while the additional loss of communication between cortical regions generates the unconsciousness of general anesthesia or coma. In addition to explaining the mixed results seen in previous experiments, this distinction could lead to advances in the diagnosis of patients with disorders of consciousness, and even to the development of therapies that target the thalamus and its connections with cortex.

https://doi.org/10.7554/eLife.04499.002

Introduction

Understanding the neural mechanisms of consciousness is a fundamental challenge of current neuroscience research. We note that precise definitions of the words ‘consciousness’ and ‘unconsciousness’ remain ambiguous. However, loss of voluntary response in a task is an effective and clinically relevant definition of unconsciousness. Over the past decade, important insights have been gained by using functional magnetic resonance imaging (fMRI) to characterize the brain during sleep, and disorders of consciousness (DOC) such as coma, vegetative states, and minimally conscious states (Greicius et al., 2008; Horovitz et al., 2008; Boly et al., 2009; Cauda et al., 2009; Horovitz et al., 2009; Larson-Prior et al., 2009; Vanhaudenhuyse et al., 2010; Samann et al., 2011; Fernandez-Espejo et al., 2012; Picchioni et al., 2014). In addition to being critical for patients to safely and humanely undergo traumatic surgical or diagnostic procedures, anesthesia has long been recognized as a tool for studying the neural mechanisms of loss and recovery of consciousness (Beecher, 1947).

Disruption of cortico-cortical connectivity has been proposed as a central mechanism to explain unconsciousness associated with general anesthesia, sleep, and DOC (Massimini et al., 2005; Mashour, 2006; Alkire et al., 2008; Boly et al., 2009; Cauda et al., 2009; Boveroux et al., 2010; Vanhaudenhuyse et al., 2010; Langsjo et al., 2012; Jordan et al., 2013). However, recent evidence suggests that cortico-cortical functional connectivity may be maintained during unconsciousness (Greicius et al., 2008; Horovitz et al., 2008; Larson-Prior et al., 2009). To develop more precise neuroanatomic and neurophysiological characterizations of this and other putative mechanisms, studies of anesthetic-induced unconsciousness could benefit from the use of multimodal imaging approaches, combined with neurophysiological understanding of the brain state induced by the anesthetic drug being studied.

A challenge in using anesthetics to study unconsciousness is stating precise, possible mechanisms of anesthetic-induced brain states, and using an anesthetic which acts at single rather than multiple brain targets to test those mechanism (Rudolph and Antkowiak, 2004; Franks, 2008; Brown et al., 2011). Among the several anesthetics in current use today, dexmedetomidine is an appealing choice for testing a specific mechanism of how an anesthetic alters the level of consciousness. Dexmedetomidine selectively targets pre-synaptic α2-adrenergic receptors on neurons projecting from the locus ceruleus to the pre-optic area (Correa-Sales et al., 1992; Chiu et al., 1995; Mizobe et al., 1996). This leads to activation of inhibitory outputs to the major arousal centers in the midbrain, pons, and hypothalamus producing a neurophysiological and behavioral state that closely resembles NREM II sleep (Figure 1) (Nelson et al., 2003; Huupponen et al., 2008; Akeju et al., 2014). Dexmedetomidine also acts at the locus ceruleus projections to the intralaminar nucleus of the thalamus, the basal forebrain and the cortex (Espana and Berridge, 2006).

Schematic of dexmedetomidine signaling illustrating similarities to the mechanism proposed for the generation of non-rapid eye movement II sleep.

Dexmedetomidine binds to α2 receptors on neurons emanating from the locus ceruleus to inhibit NE release in the POA. The disinhibited POA reduces arousal by means of GABA- and galanin-mediated inhibition of the midbrain, hypothalamic, and pontine arousal nuclei. Dexmedetomidine also acts at the locus ceruleus projections to the intralaminar nucleus of the hypothalamus, the basal forebrain and the cortex and on post-synaptic α2 receptors. 5HT, serotonin; Ach, acetylcholine; BF, basal forebrain; DA, dopamine; Dex; dexmedetomidine; DR, dorsal raphe; GABA, gamma aminobutyric acid receptor subtype A; Gal, galanin; His, histamine; LC, locus ceruleus; LDT, laterodorsal tegmental area; NE, norepinephrine; PAG, periaqueductal gray; POA, preoptic area; PPT, pedunculopontine tegmental area; TMN, tuberomamillary nucleus.

https://doi.org/10.7554/eLife.04499.003

Therefore, the aim of this study was to use a novel integrated positron emission tomography and magnetic resonance imaging (PET/MR) approach to characterize brain resting-state network activity and metabolism during dexmedetomidine-induced unconsciousness. This strategy offers several appealing features. First, PET scanning using fludeoxyglucose (18F-FDG) provides a sensitive estimate of the brain's metabolic state in terms of the cerebral metabolic rate of glucose (CMRglc). Second, comparison of CMRglc results with regional cerebral blood flow (rCBF) estimates obtained from pulsed Arterial Spin Labeling (pASL) fMRI signals allows a direct assessment of whether, as previously reported, flow-metabolism coupling is maintained during dexmedetomidine-induced unconsciousness (Drummond et al., 2008). Third, we use the CMRglc results to inform the brain functional connectivity analysis derived from the Blood Oxygen Level Dependent (BOLD) fMRI signals.

Results

Decreased CMRglc in the Default Mode Network (DMN), Fronto Parietal Networks (FPNs), and thalamus during dexmedetomidine-induced unconsciousness

In 10 healthy volunteers, we used EEG recordings, and an auditory task to confirm that dexmedetomidine induced a loss of voluntary responsiveness, and exhibited a neurophysiological profile that was similar to non rapid eye movement (NREM) II sleep (Akeju et al., 2014). We then studied 18F-FDG uptake during baseline and dexmedetomidine-induced unconsciousness in these 10 healthy volunteers in a separate experiment where we defined loss of consciousness as the onset of sustained eye closure and lack of response to a verbal request to open the eyes (Figure 2A). During both the baseline and dexmedetomidine-induced unconsciousness 18F-FDG study visits (two per subject), fMRI data were also recorded. Consistent with a global decrease in glucose metabolism, we found that during the unconscious state, there was a broad reduction in 18F-FDG standardized uptake values in the brain (Figure 2—figure supplement 1A). To quantify this result, we calculated CMRglc during the two 18F-FDG visits (Figure 2—figure supplement 1B). We found reduced CMRglc in thalamic, frontal, and parietal brain regions during the unconscious state (Figure 2B, Table 1). The difference in CMRglc exhibited a spatial distribution consistent with the previously described DMN (Raichle et al., 2001; Damoiseaux et al., 2006) and both left and right FPNs (Table 1) (Damoiseaux et al., 2006; Vincent et al., 2008). We found these network-specific CMRglc changes interesting because the DMN has been linked to stimulus-independent thought and self-consciousness (Raichle et al., 2001; Mason et al., 2007; Vincent et al., 2008), while the FPNs have been linked to conscious perception of the external environment and executive control initiation (Boly et al., 2008; Vincent et al., 2008).

Figure 2 with 1 supplement see all
CMRglc and rCBF are decreased in both the default mode and the frontal–parietal network brain regions during unconsciousness.

(A) Study outline of healthy volunteers recruited to undergo both baseline and dexmedetomidine-induced unconsciousness PET brain imaging in a random order. (B) Group wise (n = 10) dexmedetomidine-induced unconsciousness vs baseline changes in CMRglc (PET) displayed on the MNI152 standard volume. Significant CMRglc decreases were localized to brain regions that make up the Default Mode and the Frontal-Parietal Networks. (C) Schematic of fMRI obtained concurrently in combined PET/MR visits (n = 10) and MR/only visits (n = 7). (D) Group wise (n = 17) dexmedetomidine-induced unconsciousness vs awake changes in rCBF (fMRI) displayed on the MNI152 standard volume. Significant rCBF decreases were also localized to brain regions that make up the Default Mode and the Frontal-Parietal Networks. The brain regions that were not included in the rCBF estimation are shaded in the darker hue. ACC, anterior cingulate cortex; CMRglc, cerebral metabolic rate of glucose; Dex, dexmedetomidine; dmPFC, dorsomedial prefrontal cortex; fMRI, functional magnetic resonance; IPL, infraparietal lobule; IPL, infraparietal sulcus; PCC, posterior cingulate cortex; R, right; rCBF, regional cerebral blood flow; RSC, restrosplenial cortex; vmPFC, ventromedial prefrontal cortex.

https://doi.org/10.7554/eLife.04499.004
Table 1

CMRglc, awake vs dex-induced unconsciousness

https://doi.org/10.7554/eLife.04499.006
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > dex-induced unconsciousness
 R middle/inferior frontal gyrus4.3552424179192.95E-12
 R medial orbital gyrus4.222438−16
 L inferior frontal gyrus4.02−523610
 L superior frontal gyrus, medial part4−62040
 L middle frontal gyrus3.94−421838
 L middle frontal gyrus3.86−34866
 R middle frontopolar gyrus3.79−30582
 L lateral orbital gyrus3.58−4628−12
 L medial orbital gyrus3.56−2632−22
 R middle frontal gyrus3.55482044
 R superior frontal gyrus3.4826256
 R inferior frontal gyrus, pars opercularis3.3754164
 L inferior frontal gyrus, pars opercularis3.36−46200
 L superior frontal gyrus, medial part (2)3.33−44222
 inferior rostral gyrus3.15048−10
 R anterior insula2.8240220
 L Posterior Cingulate Gyrus5.56−4−243854432.85E-05
 L precuneus4.27−12−6422
 R precuneus3.8612−6028
 L supramarginal gyrus4.58−40−464228610.0029
 L angular gyrus3.76−48−6834
 R supramarginal gyrus4.5944−464822140.0112
 R angular gyrus434−6244
 R thalamus4.5212−20416100.0444
 L thalamus3.89−14−264
  1. CMRglc, cerebral metabolic rate of glucose; Dex, dexmedetomidine; L, left; MNI, Montreal Neurological Institute; R, right.

Decreased rCBF in DMN, FPN, and thalamus during dexmedetomidine-induced unconsciousness

To further understand the effects of dexmedetomidine-induced unconsciousness on rCBF and brain functional connectivity in these networks, we increased our statistical power for fMRI data analysis by studying an additional seven volunteers using fMRI only (Figure 2C). Analysis of the rCBF difference maps between the unconscious and baseline states showed decreases in rCBF during unconsciousness that overlapped with the same DMN and FPN regions that demonstrated reductions in CMRglc (Figure 2D, Figure 2—figure supplement 1, Table 2). This is consistent with maintenance of blood flow and metabolism coupling during dexmedetomidine-induced unconsciousness.

Table 2

rCBF, awake vs unconscious

https://doi.org/10.7554/eLife.04499.007
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > unconscious
 L thalamus5.36−10−1412108764.20E-11
 L midbrain5.28−32−10
 L posterior cingulate cortex5.14−2−4628
 R thalamus5.038−1210
 R midbrain4.44−6−36−12
 L precuneus4.44−2−7032
 R precuneus4.38−7642
 R intracalcarine cortex4.24−844
 L intracalcarine cortex4.18−2−822
 R hippocampus4.1730−34−4
 L supramarginal gyrus/intraparietal sulcus4.03−34−5044
 L hippocampus3.09−26−22−12
 R frontopolar cortex4.71−2464217730.00463
 L frontopolar cortex4.322062−6
 L middle frontal gyrus3.77−223828
 R inferior frontal gyrus, orbital part2.885040−14
 L dorsal anterior cingulate cortex3.9310382212620.0227
 R dorsal anterior cingulate cortex3.75−103820
 L subgenual anterior cingulate cortex3.3−224−10
 R pregenual anterior cingulate cortex3.21642−6
unconscious > awake
 n.s.
  1. L, left; MNI, Montreal Neurological Institute; n.s, not significant; R, right.

Decreased thalamo-cortical and cortico-cerebellar functional connectivity during dexmedetomidine-induced unconsciousness

Given that we observed reduced flow and metabolism in regions known to correspond to the DMN and FPNs (Raichle et al., 2001; Damoiseaux et al., 2006; Vincent et al., 2008), we next asked whether intrinsic functional connectivity within these networks was altered with dexmedetomidine-induced unconsciousness. First, we confirmed that during both wakefulness and unconsciousness, the DMN and bilateral FPNs were consistently identified in our volunteers (Figure 3, Table 3). Next, we performed a comparison of changes in these networks between the unconscious and awake states. When we compared the DMN between these two states, we found no difference in functional connectivity between the cortical regions of the DMN and FPNs, suggesting that cortico–cortico functional connectivity may be maintained during unconsciousness (Figure 3A–B, Table 3). However, we observed a reduction in functional connectivity between the thalamus and the DMN in the unconscious state (Figure 3A–B, Table 3). This loss of thalamic functional connectivity was observed in a region consistent with intralaminar, midline, mediodorsal, and ventral anterior nuclei. However, our ability to precisely resolve the thalamic nuclei implicated in loss of thalamo-cortical functional connectivity is limited because our methods are not sensitive to small focal differences. We also found that functional connectivity of the left cerebellar representation of the DMN (Buckner et al., 2011) was significantly reduced during unconsciousness (Figure 3B, Table 3). When we compared the FPNs in these two states, we found that the left and right FPNs showed opposite changes in functional connectivity with a cerebellar cluster, likely indicating a switch in hemispheric dominance for cortico-cerebellar functional connectivity to this region during unconsciousness (Figure 3C–F, Table 4, Table 5). Functional connectivity of the cerebellar representations of the FPNs (Buckner et al., 2011) was also significantly reduced (Figure 3D,F, Table 4, Table 5). Since cortico-cerebellar connections are mediated by the thalamus (Guillery and Sherman, 2002), the changes we observed in cortico-cerebellar functional connectivity are likely related to impaired thalamic functioning. Taken together, these results show that unconsciousness is associated with decreased thalamic CMRglc, rCBF, and functional connectivity to the DMN and support the notion that the thalamus plays a critical role in mediating unconsciousness.

Figure 3 with 1 supplement see all
Changes in the default mode and the bilateral frontal parietal networks during unconsciousness.

(A) The DMN extracted from BOLD signals during the awake (n = 16) and unconscious states (n = 16) and displayed on the MNI152 standard volume. (B) A comparison the DMN network during the unconscious vs the awake state identified the putamen and the vlPFC as regions that exhibited increased functional connectivity during dexmedetomidine-induced unconsciousness. This comparison also identified the thalamus and the cerebellar representation of this network as regions that exhibited decreased functional connectivity during dexmedetomidine-induced unconsciousness. Cortico–cortico functional connectivity within this network was maintained. (C) The right FPN extracted from BOLD signals during the awake and unconscious states (n = 16) and displayed on the MNI152 standard volume. (D) A comparison of the right FPN network during the unconscious vs the awake state identified the infraparietal lobule and a right cerebellar cluster as regions that exhibited increased functional connectivity during dexmedetomidine-induced unconsciousness. This comparison also identified the cerebellar representation of this network as a region that exhibited decreased functional connectivity during unconsciousness. Cortico–cortico functional connectivity within this network was maintained. (E) The left FPN extracted from BOLD signals during the awake and unconscious states (n = 16) and displayed on the MNI152 standard volume. (F) A comparison of the left FPN network during the unconscious vs the awake state identified a right cerebellar cluster and the cerebellar representation of the left FPN as regions that exhibited decreased functional connectivity during dexmedetomidine-induced unconsciousness. Cortico–cortico functional connectivity within this network was maintained. BOLD, blood oxygen level dependent; DMN, default mode network; FPN, Frontoparietal Network; R, right; vlPFC, ventrolateral prefrontal cortex.

https://doi.org/10.7554/eLife.04499.008
Table 3

DMN

https://doi.org/10.7554/eLife.04499.010
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > unconscious
 R thalamus4.042−1484380.00193
 L thalamus3.9−4−148
 L cerebellum (Crus II)3.93−44−64−442700.0429
unconscious > awake
 R inferior frontal gyrus, pars opercularis3.835612203990.00381
 R inferior frontal gyrus, pars triangularis3.35543210
 L putamen3.47−246−83820.00517
 L insula3.43−36−10−4
  1. DMN, Default Mode Network; L, left; MNI, Montreal Neurological Institute; R, right.

Table 4

rFCN

https://doi.org/10.7554/eLife.04499.011
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > unconscious
 L cerebellum (Crus II)3.99−36−70−404720.000724
 L cerebellum (VIIb)3.83−34−62−50
 L cerebellum (VI)3.28−28−62−32
unconscious > awake
 R cerebellum (VI)4.0212−68−286931.97E-05
 R cerebellum (Crus II)3.6726−78−42
 R cerebellum (Crus I)3.2420−74−24
 R precuneus2.9926−60185190.000323
 R parietal operculum4.158−12125160.00034
 R insula3.2344−40
 R cerebellum (Crus I)4.2844−56−322920.0212
 R fusiform gyrus3.5930−42−20
 L angular gyrus3.72−38−86202720.0319
 L angular gyrus3.56−48−7830
  1. L, left; MNI, Montreal Neurological Institute; R, right; rFCN, right Frontoparietal Control Network.

Table 5

lFCN

https://doi.org/10.7554/eLife.04499.012
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > unconscious
 R cerebellum (Crus II)4.2438−74−5012235.33E-09
 R cerebellum (VIIb)4.138−56−50
 R cerebellum (VI)4.0828−62−32
 R intracalcarine cortex3.584−86−2
 L middle frontal gyrus3.89−3212544900.000341
 L superior frontal gyrus, medial part3.52−43246
unconscious > awake
 n.s.
  1. L, left; lFCN, left Frontoparietal Control Network; MNI, Montreal Neurological Institute; n.s, not significant; R, right.

Seed-based thalamic connectivity during unconsciousness showed reduced thalamic functional connectivity to posterior cingulate, precuneus, and inferior parietal cortices

In order to further examine the role of the thalamus in unconsciousness, we next performed a seed-based functional connectivity analysis using the thalamic region with decreased DMN functional connectivity as the seed region. This thalamic seed also overlapped with thalamic regions showing reduced rCBF and CMRglc. Our seed-based analysis allowed us to more thoroughly examine changes in thalamic functional connectivity to all brain regions. We found that the thalamic seed was significantly functionally disconnected from posterior regions of the DMN (posterior cingulate cortex, precuneus, and inferior parietal lobules) during unconsciousness (Figure 4A, Table 6). Notably, the cortical regions showing reduced thalamic functional connectivity also overlapped with those demonstrating reduced flow and metabolism (Figure 4B).

Seed based functional connectivity analysis of the brain region with overlapping changes in CMRglc, rCBF, and functional connectivity.

(A) Brain wide representation of regions connected to the thalamic seed during the awake and unconscious states (n = 16). A comparison of thalamic seed functional connectivity during the unconscious vs the awake state identified the PCC, precuneus, thalamus, visual cortex, IPL and IPS as regions exhibiting decreased functional connectivity during unconsciousness. (B) Overlap of the brain regions with changes in CMRglc, rCBF, and functional connectivity. Notably, the posterior thalamus, PCC and the precuneus exhibited overlapping changes. CMRglc, cerebral metabolic rate of glucose; IPL, infraparietal lobule; IPL, infraparietal sulcus; PCC, posterior cingulate cortex; R, right; rCBF, regional cerebral blood flow.

https://doi.org/10.7554/eLife.04499.013
Table 6

Thalamic seed

https://doi.org/10.7554/eLife.04499.014
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
awake > unconscious
 R thalamus5.3312−12834454.05E-16
 L posterior cingulate cortex4.95−6−3630
 R posterior cingulate cortex4.736−2834
 L thalamus4.48−16−1010
 R caudate nucleus3.44121018
 L globus pallidus3.44−122−2
 L precuneus4.03−6−563612301.79E-07
 R precuneus3.7112−6036
 R angular gyrus4.0736−605010949.54E-07
 R supramarginal gyrus3.6748−4836
 R cerebellum (crus II)4.078−76−309068.76E-06
 R cerebellum (crus I)3.8422−88−22
 R intracalcarine cortex3.752−1004
 L cerebellum (crus I)3.14−20−84−28
 L supramarginal gyrus4.6−38−52428312.21E-05
 L angular gyrus3.43−32−7648
unconscious > awake
 L inferior frontal gyrus, pars opercularis4.21−5614814581.80E-08
 L superior temporal gyrus4.2−62−104
 L middle temporal gyrus4.04−50−2−22
 L insula3.55−384−6
 R paracentral lobule3.774−32589634.41E-06
 L paracentral lobule3.64−10−3858
 R superior frontal gyrus, lateral part3.4414−656
 R precentral gyrus3.2314−2066
 L postcentral gyrus3.2−16−4472
 R superior frontal gyrus, medial part2.8610250
 R postcentral gyrus2.7216−3870
 R superior temporal gyrus548−42108681.39E-05
 R middle temporal gyrus4.5344−28−6
 R precentral gyrus3.5154−246
 L postcentral gyrus4.03−62−8364530.00401
 L precentral gyrus3.77−60−436
 L superior temporal gyrus3.97−66−44103400.0244
  1. L, left; MNI, Montreal Neurological Institute; R, right.

Decreases in rCBF in the DMN, FPN, and thalamus do not immediately reverse during recovery from unconsciousness

We next investigated whether the rCBF changes during the unconscious state returned to baseline during the recovery state. We identified 10 volunteers who responded to verbal instructions (opened eyes and gave a thumbs-up signal, though they were still mildly sedated) during at least 6 min of the recovery ASL scan. Surprisingly, when we compared the recovery state to the unconscious state in these volunteers, we found that the rCBF did not increase (cluster corrected, Figure 5A). In fact, when explored at a more liberal threshold (uncorrected, p < 0.05), further decreases in rCBF were observed (Figure 5B). When we compared the recovery state to the awake state, we found that the decrease in rCBF showed a spatial distribution similar to that observed during the unconscious state (Figure 5C). Next, we aligned all subjects to unconscious and recovery time points in order to study the dynamics of rCBF changes within the statistically significant clusters. We confirmed that rCBF decreased during loss of consciousness and that this decrease was maintained during recovery from this state in all regions analyzed (Figure 5—figure supplement 1). These results suggest that cortical rCBF increase was not sufficient for the early stages of recovery from unconsciousness induced by dexmedetomidine.

Figure 5 with 1 supplement see all
Cortical increase in rCBF is not evident during the recovery of consciousness.

(A) Cluster corrected (n = 10) rCBF comparison of the recovery state vs the unconsciousness state displayed on the MNI152 standard volume highlighting no change in rCBF at recovery. (B) Uncorrected (n = 10) rCBF comparison of the recovery state vs the unconsciousness state displayed on the MNI152 standard volume suggesting that a decrease in rCBF occurred during the recovery state. (C) Cluster corrected (n = 10) rCBF comparison of the recovery state vs the awake state displayed on the MNI152 standard volume highlighting a spatial distribution of rCBF decrease similar to that observed during unconscious vs the awake state (n = 17) comparison. IPL, infraparietal lobule; IPL, infraparietal sulcus; PCC, posterior cingulate cortex; rCBF, regional cerebral blood flow; R, right; RSC, retrosplenial cortex. The brain coverage of rCBF estimation is shown in Figure 1D.

https://doi.org/10.7554/eLife.04499.015

Thalamo-cortical and cortical-cerebellar functional connectivity are restored during recovery from dexmedetomidine-induced unconsciousness

We found that thalamic functional connectivity to the DMN was restored during the recovery state (Figure 6A–B, Table 7). We also found that the opposing left and right FPN changes in the cerebellar cluster (Figure 3B–C) were also reversed (Figure 6C–F, Table 7). When we analyzed the functional connectivity of our thalamic seed of interest, we found that thalamic functional connectivity to the posterior cingulate cortex and thalamo-thalamic functional connectivity were restored during the recovery state (Figure 6G–H, Table 7). Taken together, these results provide further evidence that dexmedetomidine-induced unconsciousness is associated with decreased thalamic functional connectivity to the DMN and supports the notion that the thalamus plays a critical role in mediating unconsciousness and recovery from this state.

Functional connectivity changes observed during recovery of consciousness.

(A) The DMN extracted from BOLD signals obtained during the recovery state (n = 15) displayed on the MNI152 standard volume. (B) A comparison the DMN network during the recovery vs the awake state identified the thalamus as the only region with partial recovery of functional connectivity. (C) The right FPN extracted from BOLD signals during the recovery state (n = 15) and displayed on the MNI152 standard volume. (D) A comparison of the right FPN during the recovery vs the awake state showed that right cerebellar cluster from Figure 3B,C now exhibited decreased right FPN functional connectivity. (E) The left FPN extracted from BOLD signals during the recovery state (n = 15) and displayed on the MNI152 standard volume. (F) A comparison of the left FPN during the recovery vs the awake state showed that right cerebellar cluster from Figure 3B,C now exhibited increased left FPN functional connectivity. (G) Brain wide representation of regions connected to the thalamic seed during the recovery state (n = 15) displayed on the MNI152 standard volume. (H) A comparison of thalamic seed functional connectivity during the recovery vs the awake state identified the PCC and thalamus as regions with increased functional connectivity at recovery. BOLD, blood oxygen level dependent; DMN, default mode network; FPN, Frontoparietal Network; PCC, posterior cingulate cortex; R, right.

https://doi.org/10.7554/eLife.04499.017
Table 7

Recovery

https://doi.org/10.7554/eLife.04499.018
MNI coordinate (mm)
LabelZ-statxyzcluster size (# voxels)cluster p-value
DMN, recovery > unconscious
 R thalamus3.32−84780.0345
 L thalamus3−4−48
DMN, unconscious > recovery
 n.s.
rFCN, recovery > unconscious
 n.s.
rFCN, unconscious > recovery
 R cerebellum (VI)3.9214−70−282460.00172
 R cerebellum (Crus I)3.8122−76−28
 R cerebellum (Crus II)3.0624−80−38
lFCN, recovery > unconscious
 R cerebellum (VI)3.5812−66−301180.012
 R cerebellum (Crus I)3.2938−58−32930.0232
 R cerebellum (Crus II)3.0532−64−42
lFCN, unconscious > recovery
 n.s.
Thalamic seed, recovery > unconscious
 posterior cingulate cortex3.530−40223670.00367
 L thalamus3.49−10−22123480.00471
 L putamen2.51−1442
Thalamic seed, unconscious > recovery
 n.s.
  1. DMN, default mode network; L, left; lFCN, left Frontoparietal Control Network; MNI, Montreal Neurological Institute; n.s, not significant; R, right; rFCN, right Frontoparietal Control Network.

Discussion

Resting-state functional brain imaging is increasingly being utilized to probe the mechanisms of consciousness (Fox and Greicius, 2010). By using a very site-specific drug, performing for the first time simultaneous measurements of CMRglc, rCBF, resting state BOLD low frequency oscillations, and analyzing both dexmedetomidine-induced loss of consciousness and recovery, we have been able to identify a minimal disruption of neural networks that is shared between altered consciousness states. Our results show that during dexmedetomidine-induced unconsciousness, cortico-cortical functional connectivity remains intact but thalamo-cortical functional connectivity is disrupted. This disrupted thalamo-cortical functional connectivity appropriately reverses during recovery from dexmedetomidine-induced unconsciousness.

Thalamo-cortical correlates of altered arousal and unconsciousness

The notion that disruption of thalamo-cortical functional connectivity could constitute a neural-network mechanism responsible for unconsciousness that is distinct from disruption of cortico-cortical functional connectivity is supported by a number of findings. During, NREM I and II sleep in humans, DMN cortico-cortical functional connectivity remains preserved (Horovitz et al., 2008; Larson-Prior et al., 2009) while thalamo-cortical functional connectivity is disrupted (Picchioni et al., 2014). However, during NREM III, there is a functional disconnection between frontal and parietal nodes of the DMN (Horovitz et al., 2009; Samann et al., 2011), and disrupted thalamo-cortical functional connectivity (Picchioni et al., 2014). Likewise, in-depth electrode studies in humans show that changes in thalamic activity during sleep precedes changes in cortical activity, suggesting that disrupted thalamo-cortical functional connectivity may underlie changes in consciousness (Magnin et al., 2010). Also, during propofol anesthesia-induced unconsciousness, decreases in thalamo-cortical functional connectivity, alongside disruptions in cortico-cortical functional connectivity in the DMN and FPNs, have been described in healthy volunteers (Boveroux et al., 2010; Ni Mhuircheartaigh et al., 2013) Functional reintegration of the thalamus to key regions of the cortex appears to be necessary for the recovery of consciousness from this state (Langsjo et al., 2012).

A putative functional network for recovery from unconsciousness comprising the locus ceruleus, central thalamus, and posterior cingulate cortex

Norepinephrinergic neurons in the locus ceruleus project to thalamic mediodorsal, midline, intralaminar nuclei, which in turn have neurons that project to the cingulate cortex (Jones and Yang, 1985; Vogt et al., 1987; Buckwalter et al., 2008; Vogt et al., 2008). Our results show that these connections may comprise a functional circuit that is disrupted during dexmedetomidine-induced unconsciousness. Moreover, this functional circuit appears specifically to involve the posterior cingulate cortex, distinct from anterior or mid-cingulate cortex. In our thalamic seed-based functional connectivity analysis, we found that during dexmedetomidine-induced unconsciousness, thalamic regions consistent with intralaminar, midline, mediodorsal, and ventral anterior nuclei were functionally disconnected from posterior DMN structures (precuneus and the posterior cingulate cortex). Perhaps more importantly, recovery from dexmedetomidine-induced unconsciousness was associated with restored thalamo-thalamic and thalamic functional connectivity to the posterior cingulate cortex. Structural connectivity between the thalamus and the posterior cingulate cortex has also recently been correlated with the severity of DOC and used to subcategorize patients in minimally conscious states (Fernandez-Espejo et al., 2012). The above observations lead us to conclude that a noradrenergic circuit involving the thalamus and posterior cingulate cortex may be important for recovery from both dexmedetomidine-induced unconsciousness and DOC states.

Reduced cerebral blood flow during recovery of consciousness

An FDG-PET study has recently described a linear relationship between metabolism in the precuneus and central thalamus in severely brain-injured patients with DOC (Fridman et al., 2014). We note that our observed reduction in blood flow in the precuneus and thalamus during dexmedetomidine-induced unconsciousness parallels this result. However, during recovery from dexmedetomidine-induced unconsciousness, we did not observe an increase in rCBF to any brain regions. Since our current and previously reported (Drummond et al., 2008) results suggest that dexmedetomidine does not decouple rCBF and CMRglc, we surmise that metabolism does not return to baseline during the earliest phases of recovery from dexmedetomidine-induced unconsciousness. Thus, the reduced rCBF we report after recovery from dexmedetomidine-induced unconsciousness could reflect a residual drug effect state in which subjects are responsive to external stimuli but remain in a sedated state. This residual effect may be mediated by the broad and diffuse nor-adrenergic projections from the locus ceruleus to the rest of the brain (Vogt et al., 2008).

This result is interesting because it suggests that recovery from unconsciousness may occur in a graded fashion. For example, performance on more complex cognitive tasks likely differs substantially between the awake and recovery states. This suggests that re-establishment of functional connectivity between vital brain regions precedes increases in rCBF and CMRglc. Return of rCBF and CMRglc to baseline levels may be indicative of the final stages of recovery from dexmedetomidine-induced unconsciousness. Future investigations incorporating cognitive and psychomotor vigilance tasks during recovery from experimental models of unconsciousness would lend further insight on whether this is a drug-specific effect. Nevertheless, this finding is interesting because it suggests that recovery from unconsciousness states may proceed prior to significant increases in rCBF.

Different states of unconsciousness may reflect a hierarchy of disruption in functional circuits

Drawing from previous work on sleep (Horovitz et al., 2008, 2009; Larson-Prior et al., 2009; Magnin et al., 2010; Samann et al., 2011; Picchioni et al., 2014) and anesthesia (White and Alkire, 2003; Alkire et al., 2008; Boveroux et al., 2010; Liu et al., 2013; Ni Mhuircheartaigh et al., 2013), our results suggest a hierarchy of brain network disruptions that can result in different states of altered consciousness. In NREM II (non-slow wave) sleep and dexmedetomidine-induced unconsciousness, patients can be easily aroused to respond to verbal commands by sufficiently strong external stimuli. In these states, cortico-cortical functional connectivity is preserved (Horovitz et al., 2008; Larson-Prior et al., 2009), but thalamo-cortical functional connectivity is disrupted (Picchioni et al., 2014). In propofol anesthesia-induced unconsciousness and some DOC states, where patients cannot be aroused by external stimuli, both cortico-cortical and thalamo-cortical functional connectivity are disrupted (Boveroux et al., 2010).

In the case of sleep, dexmedetomidine-induced unconsciousness, and some minimally conscious states, maintained cortico-cortical functional connectivity likely allows for the cortex to be ‘primed’, ready to recover from the altered consciousness states with restoration of thalamic functional connectivity, which could be triggered by ascending midbrain and brainstem inputs (Solt et al., 2014) or direct therapeutic intervention at the thalamic level (Schiff et al., 2007; Williams et al., 2013). In other altered consciousness states, such as during general anesthesia or comatose states, markedly disrupted cortico-cortical functional connectivity could further impair the ability to recover consciousness. We note that while it might be theoretically possible to attain unconsciousness by disrupting only cortico-cortical functional connectivity while preserving thalamo-cortical functional connectivity, it appears, from our study and others, that this does not happen during sleep and anesthesia-induced unconsciousness.

Compared to the findings supporting a role for disrupted cortico-cortical functional connectivity during unconsciousness, our findings suggest that thalamo-cortical connectivity constitutes a more fundamental disruption required to produce unconsciousness. This insight could lead to a more objective characterization, diagnosis, and prognosis of DOC. Also, the state of cortico-cortical and thalamo-cortical functional connectivity could be used to better select patients who may benefit from therapeutic interventions. For instance, in the event of maintained cortico-cortical functional connectivity but impaired thalamo-cortical functional connectivity with intact thalamo-cortical structural connections (Fernandez-Espejo et al., 2012), therapies aimed at restoring thalamo-cortical functional connectivity might be more likely to help patients recover. These therapies could be pharmacological (Williams et al., 2013), electrical (Schiff et al., 2007), or could involve other forms of stimulation. Since dexmedetomidine appears to disrupt functional network activity in a manner similar to NREM II sleep, our results also suggest that alpha-2-receptor agonists could be further developed as effective sleep therapeutic drugs.

Subjective awareness during sleep and anesthesia

At present, there is a debate pertaining to whether anesthetized and unresponsive patients (without motor impairments) can remain internally conscious to subjective experiences (internally generated or provoked by external stimuli). This debate has been fueled by clinical studies of anesthetized patients that have reported rare intraoperative episodes of awareness (resulting from light anesthetic depth) under anesthesia (Ghoneim et al., 2009) and dreaming under anesthesia (Brandner et al., 1997; Leslie et al., 2005; Errando et al., 2008; Sanders et al., 2012). Noreika et al. studied this phenomenon with dexmedetomidine, sevoflurane, propofol, and xenon (Noreika et al., 2011) and confirmed that subjective experiences of consciousness occur during clinically defined unconsciousness states. However, the experimental paradigm that was used to test for subjective experiences of consciousness involved a gradual increase in drug concentrations. This empiric gradual increase did not directly target anesthetic drug dosing to a neurophysiologically defined brain state nor does it inform us on the amount of time that was spent in lighter levels of anesthesia. Thus, the reported subjective experiences could have occurred when the drug concentration levels were very low. This line of reasoning is supported by evidence confirming that subjective experiences of dreaming under general anesthesia likely occur when patients are emerging from general anesthesia (Leslie et al., 2009).

Subjective experiences of consciousness also occur during REM sleep (Hobson, 2009). To a limited degree, they are also thought to occur during NREM sleep, especially at sleep onset in NREM I and during NREM II sleep periods encountered later at night (Hobson, 2009). However, the current methods for scoring sleep states remain imprecise and highly subjective, with limited state and temporal resolution, and high intra- and inter-scorer variability, raising the possibility that clinically scored NREM I and II sleep may include periods of arousal. In clinical sleep scoring, trained technicians visually score the sleep time-series data in 30-s epochs according to semantically defined sleep stages. These scoring standards grossly simplify sleep electroencephalographic dynamics. For example, consider the characterization of sleep onset. The American Academy of Sleep Medicine defines sleep onset as the first appearance of any 30-s epoch that contains at least 15 s of sleep (Iber et al., 2007). This implies that sleep onset is a binary process. However, recent EEG studies in humans clearly demonstrate that sleep onset, at both a neurophysiological and behavioral level, is a continuous process. As such, methods designed to track moment-to-moment continuous variations in the EEG are significantly better at predicting when subjects lose wakefulness, measured in terms of a behavioral task, than traditional sleep scoring methods (Prerau et al., 2014). In particular, subjects can continue to perform awake behavior despite being scored as asleep (NREM I) using traditional sleep scoring methods. Thus, it is possible that the subjective experiences of consciousness that have been ascribed to NREM sleep could have occurred during subjectively scored sleep stages that included un-scored periods of arousal or wakefulness.

Despite this rapidly evolving understanding of subjective experiences during altered arousal under sleep and anesthesia, we have tried in this study to focus on a particular state of dexmedetomidine-induced unconsciousness. We achieved this state through rapid administration of dexmedetomidine and confirmed that this dose and form of drug administration produces a NREM II-like state (Akeju et al., 2014). We then applied a combination of fMRI and FDG-PET imaging that would characterize the overall time-integrated brain activity during this state. Clinical evidence shows that motor-evoked potentials, and thus motor function from cortex down, remain intact in this state, making it unlikely that the loss of responsiveness we observed is only a result of disrupted motor function (Moore et al., 2006; Garavaglia et al., 2014). Therefore, we feel it is appropriate to refer to the state of altered arousal described in this manuscript as dexmedetomdine-induced unconsciousness.

A long-standing debate in anesthesiology relates to the fundamental systems-level mechanisms for anesthesia-induced unconsciousness, with some investigators proposing a thalamic switch (Alkire et al., 2000) and others proposing disrupted cortico-cortical connections as the central mechanism (Massimini et al., 2005; Mashour, 2006; Alkire et al., 2008). Our results, in combination with previous work, resolve this conflict by showing how both mechanisms could co-exist but at different parts of the continuum of altered arousal or unconsciousness. Moreover, it suggests a mechanistic interpretation for varying levels of anesthesia-induced unconsciousness, where disrupted thalamo-cortical functional connectivity, but intact cortico-cortical functional connectivity reflects lighter states of unconsciousness and disruptions of both cortico-cortical and thalamo-cortical functional connectivity reflect deeper states of unconsciousness. These insights could guide development of system-specific anesthetic drugs that possess minimal side effects and improved monitoring of patients in both the operating room and intensive care settings.

Materials and methods

Subject selection

This study was conducted at the Athinoula A Martinos Center for Biomedical Imaging at the Massachusetts General Hospital. The Human Research Committee and the Radioactive Drug Research Committee at the Massachusetts General Hospital approved the study protocol. After an initial email/phone screen, potential study subjects were invited to participate in a screening visit. At the screening visit, informed consent including the consent to publish was requested after the nature and possible consequences of the study was explained. All subjects provided informed consent and were American Society of Anesthesiology Physical Status I with Mallampati Class I airway anatomy. After providing informed consent, a standard pre-anesthesia assessment was administered and a blood toxicology screen was performed to ensure that subjects were not taking drugs that might confound the interpretation of study results. A complete metabolic panel and a urine pregnancy test were also obtained to confirm that all values were within normal ranges and to confirm non-pregnant status, respectively. A total of 20 subjects participated in screening visits and 18 subjects were deemed eligible to participate in the study protocol. Two subjects were deemed ineligible after medical evaluation and one subject was lost to follow-up after providing informed consent. During all study visits, subjects were required to take nothing by mouth for at least 8 hr. A urine toxicology screen was performed to ensure that subjects had not taken drugs that might confound interpretation of the results. A pregnancy test was also administered (serum for PET visits, urine for fMRI only visits) for each female subject to confirm non-pregnant status.

Imaging visit

We studied awake, unconscious, and recovery states in humans, using an integrated positron PET/MR approach, in healthy volunteers, 18–36 years of age. Brain imaging was performed with the Biograph mMR scanner (Siemens Healthcare, Erlangen, Germany), which allows simultaneous acquisition of whole-body PET and 3 Tesla MRI data. The fully integrated PET detectors use avalanche photoiodide technology and lutetium oxyorthosilicate crystals (8 × 8 arrays of 4 × 4 × 20 mm3 crystals). The PET scanner's transaxial and axial fields of view are 594 mm and 25.8 cm, respectively (Drzezga et al., 2012). Approximately, 5 mCi of FDG was purchased from an outside approved vendor and was administered as a bolus immediately before the initiation of the dexmedetomidine (described below). A PET-compatible 16-channel head and neck array was used to acquire the MR data.

At the beginning of the imaging visit, we acquired structural MRI (MPRAGE volume, TR/TE = 2100/3.24 ms, flip angle = 7°, voxel size = 1 mm isotropic) for the purpose of anatomical localization, spatial normalization, and surface visualization of the imaging data. A 6:08 min pASL scan and a 6:15 min BOLD fMRI scans were performed (‘awake’ pASL and BOLD scans; Figure 1C), in order to assess rCBF and functional connectivity at baseline. The pASL scans were collected using the ‘PICORE-Q2TIPS’ MRI labeling method (Luh et al., 1999) (TR/TE/TI1/TI2 = 3000/13/700/1700 ms, voxel size = 3.5*3.5*5 mm, number of slices = 16). ‘Tag’ images were acquired by labeling a thick inversion slab (110 mm), proximal to the imaging slices (gap = 21.1 mm). ‘Control’ images were acquired interleaved with the tag images, by applying an off-resonance inversion pulse without any spatial encoding gradient. At the beginning of each pASL scan, an M0 scan (i.e., the longitudinal magnetization of fully relaxed tissue) was acquired for rCBF quantification purposes. The pASL imaging volume common to all subjects scanned covered most of the cerebrum, and extended ventrally to the midbrain and dorsally to the vertex (Figure 1D). BOLD fMRI data were collected using a whole brain T2*-weighted gradient echo BOLD echo planar imaging pulse sequence was used (TR/TE = 3000/35 ms, flip angle = 90°, voxel size = 2.3 × 2.3 × 3.8 mm, number of slices = 35).

After the awake scans, dexmedetomidine was administered as a 1 mcg/kg loading bolus over 10 min, followed by a 0.7 mcg/kg/hr infusion to maintain unconsciousness. Another pASL scan was initiated at the onset of the dexmedetomidine infusion, with identical imaging parameters as the awake pASL scan, except for a longer duration (20:08 min). This longer acquisition was performed in order to ensure that perfusion data were collected for a long enough period to capture the transition from the awake to the unconscious state. During the infusion period, the study anesthesiologists monitored cuff blood pressure, capnography, electrocardiogram, and pulse-oximetry. Volunteers were instructed to keep their eyes open during the course of the study; loss of consciousness was defined as the onset of sustained eye closure and lack of response to a verbal request to open the eyes. After the onset of unconsciousness, we then performed a 6:15 min ‘unconscious’ BOLD fMRI scan (to evaluate unconsciousness-related changes in functional connectivity). After acquisition of all images, the dexmedetomidine infusion was discontinued, and a 20:08 min pASL scan was performed to assess changes in rCBF upon recovery. Spontaneous eye opening and a positive response to give a thumbs-up signal were used to determine recovery of consciousness. All 10 subjects who experienced spontaneous eye opening during the final 6:08 min of the recovery of consciousness pASL successfully executed on the thumbs-up signal request. An additional 6:15 min ‘recovery’ BOLD fMRI scan was performed, in order to assess variations in functional connectivity upon recovery.

Data analysis

PET data collected from 10 subjects (two visits each) and stored in list mode format were binned into sinograms. In addition to the static frame corresponding to 40–60 min post FDG administration, dynamic frames of progressively longer duration were generated for quantitative analysis. PET images were reconstructed using an ordered subsets expectation maximization algorithm, with 3 iterations and 21 subsets. Corrections were applied to account for variable detector efficiencies and dead time, photon attenuation and scatter and radioactive decay using the software provided by the manufacturer. We employed a method that is similar to the generation of attenuation maps for computed tomography data to generate our head attenuation maps from the MPRAGE data. This head attenuation map was combined with the hardware (i.e., RF coil, patient table, etc) attenuation map provided by the manufacturer. Spatial smoothing was performed post-reconstruction using a 4 mm full-width-at-half-maximum (FWHM) Gaussian kernel.

Standardized uptake values (SUV; i.e., mean radioactivity/injected dose/weight) were computed voxelwise from the emission data collected 40–60 min post-injection. In order to quantitatively assess metabolic changes, cerebral metabolic rate of glucose (CMRglc) were computed from the dynamic PET frames with the non-invasive method proposed by Wu (2008), using the whole-brain as our reference region. SUV and CMRglc maps were then coregistered with the high resolution MRI scan using BBREGISTER tool (Greve and Fischl, 2009) from the FreeSurfer suite (http://surfer.nmr.mgh.harvard.edu/), normalized to the Montreal Neurological Institute (MNI) space using nonlinear registration (FNIRT, from the FSL suite; FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl/) (Smith et al., 2004) and then smoothed with an 8 mm FWHM kernel. A paired, whole-brain, voxel wise, mixed effects analysis was conducted to compare SUV and CMRglc across visits (n = 10). Statistical parametric maps were thresholded using clusters determined by a voxel-wise threshold (Z > 2.3) and a (corrected) cluster significance threshold of p = 0.05 (Worsley, 2001).

fMRI data were collected from 17 subjects. However, BOLD data (awake, unconscious, and recovery) were successfully collected in only 15 subjects. This is because BOLD data was collected with a different set of parameters in one subject and another subject exited the scanner immediately prior to the recovery BOLD scan. ASL data preprocessing was performed using a combination of analysis packages including FSL, and Freesurfer. The ‘tag’, ‘control’, and M0 scans were first motion-corrected using MCFLIRT (Jenkinson et al., 2002). Then, tag and control scans were surround subtracted (i.e., given each tagX, [(controlX−1 + control X+1)/2 − tagX]) to achieve perfusion-weighted images. Quantification of rCBF was performed using the ASLtbx toolbox (Wang et al., 2008; Wang, 2012) from the following data: the whole 6 min awake scan, the last 6 min of the 20 min induction of unconsciousness scan (i.e., the portion of the scan during which all subjects were unconscious; ‘unconscious’ scan) and the last 6 min of the 20-min recovery of consciousness scan (i.e., the portion of the scan during which 10 subjects had recovered; ‘recovery’ scan).

rCBF maps were coregistered with the high resolution MR scan using Freesurfer's BBREGISTER tool (Greve and Fischl, 2009), normalized to MNI space using nonlinear registration (FNIRT) (Smith et al., 2004) and then smoothed with a 7 mm FWHM kernel. A paired, voxel wise, mixed effects analyses were conducted to compare rCBF between the awake and unconscious states data (n = 17). This contrast was computed only within voxels imaged in all subjects (see Figure 2D for coverage common to all subjects), as assessed through the computation of a conjunction mask from all the MNI-registered M0 scans. Given that rCBF estimation in the white matter presents methodological challenges (particularly given the poor signal-to-noise ratio and longer arterial transit time) (Liu et al., 2010), voxels classified as white matter in (the arbitrary value of) ≥ 70% of subjects by the Harvard–Oxford Subcortical Structural Atlas (Center for Morphometric Analyses, http://www.cma.mgh.harvard.edu/fsl_atlas.html) were excluded from the analyses.

Subsequently, another paired analysis was performed to compare rCBF between the unconscious and recovery states. In this analysis, only the 10 subjects who were awake for the full duration of the last 6 min of the 20-min recovery scans were included. This analysis was focused within search areas determined by the statistically significant clusters in the ‘unconscious vs awake’ contrast. Unless otherwise specified, all rCBF statistical parametric maps were thresholded using clusters determined by a voxel-wise threshold (Z > 2.3) and a (corrected) cluster significance threshold of p = 0.05 (Worsley, 2001).

Resting-state BOLD fMRI data were used to perform functional connectivity analyses. Datasets were slice-timing corrected (SLICETIMER), motion-corrected (MCFLIRT), brain-extracted (BET), registered to MPRAGE (BBREGISTER) and then to the MNI152 atlas (FNIRT), smoothed with a 5 mm FMWH kernel and high-pass filtered (cutoff = 0.008 Hz). As ASL and PET analyses revealed that unconsciousness was associated with a reduction in glucose metabolism and rCBF in regions belonging to distinct resting-state networks (Default Mode and both Frontoparietal Control networks; DMN, FCN), we assessed how intrinsic functional connectivity of these specific networks was affected by unconsciousness, using the dual regression approach (Filippini et al., 2009; Zuo et al., 2010). All the preprocessed datasets were concatenated to create a single 4D dataset. A probabilistic Independent Component Analysis (pICA) (Beckmann et al., 2005) was performed using Multivariate Exploratory Linear Optimized Decomposition into Independent Components (MELODIC) on this concatenated 4D dataset, limiting the number of independent components (ICs) to 25, as in previous publications using the dual regression approach (Filippini et al., 2009; Napadow et al., 2010, 2012; Kim et al., 2013; Loggia et al., 2013). From the pool of the 25 ICs, the DMN, right and left FCNs were clearly identified. Group-level spatial maps for each of these RSNs were used as a set of spatial regressors in a General Linear Model (GLM), in order to identify the individual subjects' time course associated with each group-level map. These time courses were then variance normalized and used as a set of temporal regressors in a GLM, to find subject-specific maps associated with the different group-level independent components. In this GLM, explanatory variables also included six motion parameters and the time courses from ventricles and white matter as covariates of no interest. Subject-specific maps were compared across states (awake, unconscious, recovery) using paired t-tests. As our dual regression ICA approach identified altered DMN functional connectivity to the thalamus, we used this thalamic region as a seed for seed-based functional connectivity analysis. This follow-up analysis was done with the purpose of (1) determining whether specific portions of the DMN (e.g., more posterior regions within the network) may be responsible for driving the effect and (2) assessing changes in functional connectivity between the thalamus and other, non-DMN regions. The extracted fMRI time series, variance normalized, was used as a regressor in a GLM for the awake, unconscious, and recovery data. The same nuisance regressors adopted in the dual regression analyses (see above), with the addition of global signal time course were used in the seed-based analyses. Statistical parametric maps were thresholded using clusters determined by a voxel-wise threshold (Z > 2.3) and a (corrected) cluster significance threshold of p = 0.05 (Worsley, 2001). From representative clusters identified in the various analyses, data were extracted and displayed for illustrative purposes. Human brain atlases were used for anatomical reference of the cerebrum (Mai et al., 2008) and the cerebellum (Diedrichsen et al., 2009).

References

  1. 1
  2. 2
  3. 3
  4. 4
    Investigations into resting-state connectivity using independent component analysis
    1. CF Beckmann
    2. M DeLuca
    3. JT Devlin
    4. SM Smith
    (2005)
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 360:1001–1013.
    https://doi.org/10.1098/rstb.2005.1634
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
    The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specification
    1. CAIS Iber
    2. AL Chesson
    3. SF Quan
    (2007)
    The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specification.
  37. 37
  38. 38
  39. 39
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
    Atlas of the human brain
    1. JK Mai
    2. G Paxinos
    3. T Voss
    (2008)
    New York: Elsevier/Academic Press.
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69
  70. 70
  71. 71
  72. 72
  73. 73
  74. 74
  75. 75
  76. 76
  77. 77
  78. 78
    Functional MRI: An introduction to methods
    1. KJ Worsley
    (2001)
    Functional MRI: An introduction to methods, New York, Oxford University Press.
  79. 79
  80. 80

Decision letter

  1. Jody C Culham
    Reviewing Editor; University of Western Ontario, Canada

eLife posts the editorial decision letter and author response on a selection of the published articles (subject to the approval of the authors). An edited version of the letter sent to the authors after peer review is shown, indicating the substantive concerns or comments; minor concerns are not usually shown. Reviewers have the opportunity to discuss the decision before the letter is sent (see review process). Similarly, the author response typically shows only responses to the major concerns raised by the reviewers.

Thank you for sending your work entitled “Thalamic Functional Connectivity to the Default Mode Network is a Fundamental Neural Correlate of Consciousness” for consideration at eLife. Your article has been favorably evaluated by Eve Marder (Senior editor), Jody Culham (Reviewing editor), and 3 reviewers.

The Reviewing editor and the reviewers discussed their comments before we reached this decision, and the Reviewing editor has assembled the following comments to help you prepare a revised submission.

All three external reviewers made positive comments about the manuscript, noting it “presents an interesting and novel approach to the problem of defining the neural bases of anesthesia-induced unconsciousness”, provide results that are “highly compelling” and “significant”, and overall is a paper that is “quite interesting.”

However, each reviewer brought up points that must be addressed in a revision before the manuscript can be considered suitable for publication in eLife. The following points must be addressed in a revision, which should also take into account the secondary comments (taken directly from the reviewers).

1) Although one reviewer acknowledged that the definition of “loss of consciousness” is a matter of semantics, the authors should rewrite the manuscript to focus on what they actually observed – unresponsiveness – and not perseverate on loss of consciousness.

It is possible that the disconnection of the cortico-thalamic-cerebellar cohesion which the authors claim to be the dexmedetomidine-induced perturbation that produces unresponsiveness may be linked to loss of motor output rather than the conscious experience (Tononi et al., 2004). Without studying the volunteers' subjective experience (after the period of sedation) they would be hard-pressed to describe what was studied as loss of consciousness. Noreika et al. (2011) showed that almost 3 of 4 of their subjects with dexmedetomidine-induced unresponsiveness reported a subjective experience. Absent documentation of presence/absence of subjective experience and therefore presence/absence of consciousness, the possibility remains that most of the subjects retained consciousness and hence cortico–cortico connectivity. A further doubt whether the subjects had lost consciousness is their description of the similarity of the dexmedetomidine-induced behavioral state to EEG-defined stage 2 NREM sleep; this sleep state is not uniformly associated with a lack of subjective experience and therefore unconsciousness.

2) The Introduction should make the main research question and goals clearer. For example, are they using anesthesia as a model to explain pathological loss of consciousness (e.g. disorders of consciousness) or are they interested in the effects of anesthesia in the brain's resting dynamics per se?

3) The revision should discuss/reconcile the decrease in rCBF in DMN, FPN and thalamus with the return of consciousness in the context of the restoration of the functional connectivity in the same areas. One reviewer suggests “perhaps there is some evidence that the alpha2 Dex specific agonism may influence blood vessel responses that could account for both the preservation of the functional connectivity at the cortical level during general anesthesia and failure to normalize?”

4) The spatial resolution of the techniques should be fully considered, particularly in the context of the ability to detect effects in thalamic subnuclei.

5) The authors may wish to take better advantage of the fact that they collected fMRI/PET data simultaneously. All comparisons between both modalities are made on a descriptive basis (or map overlap). A direct comparison of the changes in activation/connectivity/metabolic maps over time would add stronger support to their conclusions.

Reviewer #1:

The authors' reference to the need of combined EEG and fMRI in the Introduction (end of second paragraph) seems out of the scope of their article.

The Methods section describes only 1 imaging visit but Figure 2 describes 2. Please clarify.

Also the number of subjects included in is not clear. Were there 10 subjects for the PET data and 17 or 10 for the fMRI data? Please clarify. Why didn't the authors collect data from all subjects?

How was the onset of the 'unconscious' period determined? In the Methods section the authors define 'sustained eye closure' as a criterion but in the Results section they define an 'EEG response similar to nREM 2 sleep' instead.

The authors may want to consider including a reference to the article below, as it addresses similar research questions:

Altered temporal variance and neural synchronization of spontaneous brain activity in anesthesia.

Huang Z, Wang Z, Zhang J, Dai R, Wu J, Li Y, Liang W, Mao Y, Yang Z, Holland G, Zhang J, Northoff G.

Hum Brain Mapp. 2014 May 28. doi: 10.1002/hbm.22556.

Reviewer #2:

Discussion of the role of noradrenergic projections from locus ceruleus to central thalamus could be amplified and the specific innervation of the central lateral nucleus neurons by noradrenergic afferents noted (Vogt, B.A. et al., 2008. Norepinephrinergic afferent and cytology of the macaque monkey midline, mediodorsal, and intralaminar thalamic nuclei. Brain Struct. Funct. 212: 465-479.) This specificity relates to the findings of the relationship of posterior DMN structures and loss of consciousness identified in the manuscript (see point 3).

The observation that thalamic seed is specifically functionally disconnected from posterior DMN regions (posterior cingulate, precuneus and inferior parietal cortex) is relevant to the specific role of the central lateral nucleus components of the intralaminar nuclei which send broad projections across these cortical regions (Buckwalter et al. J. Comp Neurol. 2008). Recent FDG-PET studies in severely brain-injured subject demonstrate a graded relationship between central thalamus and the precuneus (Fridman et al. PNAS 2014) and provide a nice correlation of these findings which could be compared and discussed.

The correlation of the correlation of the Dex-induced unconscious state with the specific preservation of cortico-cortical functional connectivity suggests an interesting model for minimally conscious state. As the authors have noted this literature in the manuscript it might be worth a brief elaboration of the possible implication of these findings for similar states arising after structural brain injuries.

Reviewer #3:

Introduction:

The authors should quote original articles about the mechanism of action of dexmedetomidine, including the neural loci of action rather than their own review (Brown et al., 2011).

Methods:

The authors mention that principled constraints were placed on the connectivity analysis. These have not been referenced or defined elsewhere. It would be important to clarify.

Results:

Does the Worsley (2001) method have the spatial resolution to be able to detect activity in thalamic subnuclei? If not did you consider using a voxel-level correction (so that the statistical inference is about the peak voxel, and is localised accurately) or cluster level correction with a more stringent voxel level threshold (so the activations are more focal and better localised, albeit still imperfectly)?

Discussion:

The authors have compared their findings using dexmedetomidine with other author's studies on propofol. Is this comparison justified given the author's claim to have used a novel method to identify the changes in connectivity, an analysis that was not used by the other authors? Is it possible that the Akeju et al.'s methods are not as sensitive as other authors work?

There appears to be tautology associated with the authors contention that slow wave sleep is a “deeper level of unconsciousness” than that provided by dexmedetomidine. This therefore assumes that loss of consciousness (unconsciousness) is not binary. Therefore I am confused by their speculation.

The authors do not discuss the finding of reduced CBF on recovery of responsiveness: this is one of the most interesting and perplexing findings that deserves some discussion.

Figure 1 ignores the effect of DEX at postsynaptic alpha2 receptors, which appear to be important for the decrease in anesthetic requirements that is produced by dexmedetomidine (Segal et al., 1988).

https://doi.org/10.7554/eLife.04499.019

Author response

1) Although one reviewer acknowledged that the definition of “loss of consciousness” is a matter of semantics, the authors should rewrite the manuscript to focus on what they actually observed – unresponsiveness – and not perseverate on loss of consciousness.

It is possible that the disconnection of the cortico-thalamic-cerebellar cohesion which the authors claim to be the dexmedetomidine-induced perturbation that produces unresponsiveness may be linked to loss of motor output rather than the conscious experience (Tononi et al., 2004). Without studying the volunteers' subjective experience (after the period of sedation) they would be hard-pressed to describe what was studied as loss of consciousness. Noreika et al. (2011) showed that almost 3 of 4 of their subjects with dexmedetomidine-induced unresponsiveness reported a subjective experience. Absent documentation of presence/absence of subjective experience and therefore presence/absence of consciousness, the possibility remains that most of the subjects retained consciousness and hence cortico–cortico connectivity. A further doubt whether the subjects had lost consciousness is their description of the similarity of the dexmedetomidine-induced behavioral state to EEG-defined stage 2 NREM sleep; this sleep state is not uniformly associated with a lack of subjective experience and therefore unconsciousness.

We thank the reviewers for these comments. We have taken care to rewrite the manuscript without alluding to the notion that we studied consciousness. We agree that the definition of anesthesia-induced loss of consciousness remains a matter of debate in the literature. We have changed the title of this manuscript to reflect that we studied dexmedetomidine-induced unconsciousness:

“Disruption of Thalamic Functional Connectivity is a Neural Correlate of Dexmedetomidine-Induced Unconsciousness”

In our Introduction, we now also explicitly orient the reader as follows:

“We note that precise definitions of the words “consciousness” and “unconsciousness” remain ambiguous. However, loss of voluntary response in a task is an effective and clinically relevant definition of unconsciousness.”

While we agree that alterations in cortico-thalamic-cerebellar networks may in principle reflect loss of motor output, we would like to point out that we did not see evidence of alterations in brain regions associated with motor function. The cerebellum has been traditionally thought to solely contribute to planning and execution of movement, however recent research now demonstrates that the cerebellum is also involved in higher-order neural processing and cognition (Buckner, Neuron 2013 October 30;80(3):807-15). Indeed, the cerebellar regions displaying alterations in connectivity during dexmedetomidine-induced anesthesia in our study did not correspond to the cerebellar representation of the sensorimotor network. Rather they correspond to regions previously described as the cerebellar components of Default Mode Network and the Fronto-parietal networks (Buckner et al., J Neurophysiology 2011 November;106(5):2322-45), which are involved in cognitive processes (including self-referential cognition and attention processes) as opposed to motor processes. At our institution, dexmedetomidine is preferentially used for craniotomies involving cortical motor mapping (masses abutting the motor cortex) due to its favorable “arousability to consciousness” profile. We have not found that dexmedetomidine impairs motor output/motor-evoked potentials (Garavaglia et al., J Neurosurg Anesthesiol. 2014 July;26(3):226-33; Moore et al., Anesth Analg. 2006 May;102(5):1556-8).

The debate on the presence of subjective experiences during anesthesia has been fueled by studies of anesthesia-induced subjective experiences elicited from questionnaires typically administered during the post-operative or post-anesthesia period. This method of investigation does not precisely inform us on the timing of the subjective experience (anesthesia induction vs surgical anesthesia level maintenance vs anesthesia emergence). Nor does it inform us about the timing of that subjective experience relative to the subject’s brain state, which can fluctuate substantially even at a “steady state” of anesthetic administration, and certainly during induction and emergence. In particular, these studies were not performed with an a priori or post hoc analysis of the neurophysiological state(s) elicited by the general anesthetic agent. Rather, the Bispectral Index was typically used as a crude measure of the anesthetized brain. Research and clinical experience from our group and others indicate that the Bispectral Index output varies substantially from clinical assessments of unconsciousness. This is because the output from this device attempts to simplify the complex nature of the electroencephalogram to a single number between 1 and 100. To better inform the readers and to foster more principled approaches to defining and studying unconsciousness/lack of responsiveness in neuroscience, in the revised version of this manuscript, we now describe a few nuances pertaining to anesthesia- and sleep-induced unconsciousness in the context of the subjective experience(s) that have been elicited and positioned to argue against unconsciousness in these states (section titled “Subjective awareness during sleep and anesthesia”).

Therefore, we feel it is appropriate to refer to the state of altered arousal described in this manuscript as dexmedetomdine-induced unconsciousness.

2) The Introduction should make the main research question and goals clearer. For example, are they using anesthesia as a model to explain pathological loss of consciousness (e.g. disorders of consciousness) or are they interested in the effects of anesthesia in the brain's resting dynamics per se?

We agree with this concern, and we have rewritten our Introduction to more accurately reflect that we are studying anesthesia-induced unconsciousness:

“Therefore, the aim of this study was to use a novel integrated positron emission tomography and magnetic resonance imaging (PET/MR) approach characterize brain resting-state network activity and metabolism during dexmedetomidine-induced unconsciousness.”

3) The revision should discuss/reconcile the decrease in rCBF in DMN, FPN and thalamus with the return of consciousness in the context of the restoration of the functional connectivity in the same areas. One reviewer suggests “perhaps there is some evidence that the alpha2 Dex specific agonism may influence blood vessel responses that could account for both the preservation of the functional connectivity at the cortical level during general anesthesia and failure to normalize?”

We agree that this is a point that should be discussed. We have now added more detailed conjecture for the sustained decrease in rCBF (section titled “Reduced cerebral blood flow during recovery of consciousness”).

4) The spatial resolution of the techniques should be fully considered, particularly in the context of the ability to detect effects in thalamic subnuclei.

We agree with the reviewer that the spatial resolution of our techniques should be placed in the appropriate context when reporting results with the thalamic nuclei. We have included the following statement in the Methods section (Decreased thalamo-cortical and cortico-cerebellar functional connectivity during dexmedetomidine-induced unconsciousness):

“However, our ability to precisely resolve the thalamic nuclei implicated in loss of thalamo-cortical functional connectivity is limited because our methods are not sensitive to small focal differences.”

5) The authors may wish to take better advantage of the fact that they collected fMRI/PET data simultaneously. All comparisons between both modalities are made on a descriptive basis (or map overlap). A direct comparison of the changes in activation/connectivity/metabolic maps over time would add stronger support to their conclusions.

We agree with the reviewers that a direct comparison of activation, connectivity and metabolic maps over time would be interesting. However, this is not possible given the technical limitations of these methods. While H2O-PET has been used to characterize group-level dynamics and functional connectivity, the FDG-PET we used in our paper estimates glucose consumption over a time-scale of minutes. The technology lacks the temporal resolution to capture physiological alterations associated with faster brain dynamics such as dexmedetomidine-induced brain changes in glucose utilization. Therefore, changes in glucose utilization induced by experimental paradigms such as ours are typically inferred by comparing two separate scans (baseline vs active study scan). While ASL and BOLD fMRI have a faster temporal resolution at the level of several seconds, the SNR of these methods requires pooling of information across many minutes of imaging (in essence, averaging). Therefore, contemporary analyses of fMRI data are almost always limited to static analyses (brain maps) obtained over a several-minute period. Thus, unfortunately, it is not possible to study detailed time-varying activity with any of these imaging modalities.

Reviewer #1:

The authors' reference to the need of combined EEG and fMRI in the Introduction (end of second paragraph) seems out of the scope of their article.

We agree and have removed this reference in our revision.

The Methods section describes only 1 imaging visit but Figure 2 describes 2. Please clarify.

We apologize for the confusion. We have explicitly added this detail to the Data analysis portion of the Methods section.

“PET data collected from 10 subjects (two visits each) and stored in list mode format were binned into sinograms.”

Also, the caption for Figure 2C has also been amended;

Schematic of fMRI obtained concurrently in combined PET/MR visits (n = 10) and MR/only visits (n = 7)

Also the number of subjects included in is not clear. Were there 10 subjects for the PET data and 17 or 10 for the fMRI data? Please clarify. Why didn't the authors collect data from all subjects?

We regret this. However, in the Results section (Dexmedetomidine-induced unconsciousness decreased CMRglc in the Default Mode Network (DMN), Fronto Parietal Networks (FPNs), and thalamus) we explicitly state the following:

“We then studied 18F-FDG uptake during baseline and dexmedetomidine-induced unconsciousness in these 10 healthy volunteers…”

We have added the following sentence to directly follow this sentence in order to provide more context:

“During both the baseline and dexmedetomidine-induced unconsciousness 18F-FDG study visits (two per subject), fMRI data were also recorded.”

We hope that these revisions will clarify the situation and address the reviewer’s concerns.

Please clarify. Why didn't the authors collect data from all subjects?

FDG-PET studies of propofol and isoflurane induced-unconsciousness, the effect sizes observed were large enough to yield statistically significant effects with only six and five volunteers (Alkire MT. et al., Anesthesiology 1997, 1995; Alkire MT. et al.). Based on discussions with experts in the field, and correspondence with our institutional review board the PET portion of our study was appropriately powered (n = 10, which exceeds that of previous studies without unduly exposing study volunteers to unnecessary radiation risks). In addition, the combined PET-fMRI studies are expensive and very labor intensive. Thus, we felt that a sample size of n = 10 would be more than adequate, exceeding previous similar studies, and would not be overly burdensome in terms of subject risk and cost.

Also, in our Methods section we have added the following to better clarify fMRI data collection:

“fMRI data were collected from 17 subjects. However, BOLD data (awake, unconscious, and recovery) were successfully collected in only 15 subjects. This is because BOLD data was not collected in one subject, and another subject exited the scanner immediately prior to the recovery BOLD scan.”

How was the onset of the 'unconscious' period determined? In the Methods section the authors define 'sustained eye closure' as a criterion but in the Results section they define an 'EEG response similar to nREM 2 sleep' instead.

We have amended the text in our Methods section.

Volunteers were instructed to keep their eyes open during the course of the study; loss of consciousness was defined as the onset of sustained eye closure and lack of response to a verbal request to open the eyes.”

We have also amended the text in our Results section.

“In 10 healthy volunteers, we used EEG recordings, and an auditory task to confirm that dexmedetomidine induced a loss of voluntary responsiveness, and exhibited a neurophysiological profile that was similar to non-rapid eye movement (nREM) II sleep (Akeju et al., 2014). We then studied 18F-FDG uptake during baseline and dexmedetomidine-induced unconsciousness in these 10 healthy volunteers in a separate experiment where we defined loss of consciousness as the onset of sustained eye closure and lack of response to a verbal request to open the eyes (Figure 2A).”

The authors may want to consider including a reference to the article below, as it addresses similar research questions:

Altered temporal variance and neural synchronization of spontaneous brain activity in anesthesia.

Huang Z, Wang Z, Zhang J, Dai R, Wu J, Li Y, Liang W, Mao Y, Yang Z, Holland G, Zhang J, Northoff G.

Hum Brain Mapp. 2014 May 28. doi: 10.1002/hbm.22556.

We thank the reviewer for this suggestion. This paper analyzes patterns of resting-state fMRI activity during propofol and sevoflurane-induced unconsciousness. The authors showed that during anesthesia-induced unconsciousness, functional connectivity and temporal variance within the default-mode network decrease, while temporal variance in lateral cortical networks increase, compared to the conscious state. We don’t know how these variance measures relate to more established measures of functional connectivity, nor how the whole thalamic seed functional connectivity relates to our seed based analysis. This is especially important because non-specific and specific thalamic nulei functional connectivity in both sleep and anesthesia are different (Liu et al., Anesthesiology. 2013 January;118(1):59-69; Picchioni et al., Sleep. 2014 February 1;37(2):387-97). Thus we are hesitant to cite it without a specific understanding of how it relates to our results and previous studies.

Reviewer #2:

Discussion of the role of noradrenergic projections from locus ceruleus to central thalamus could be amplified and the specific innervation of the central lateral nucleus neurons by noradrenergic afferents noted (Vogt, B.A. et al. 2008. Norepinephrinergic afferent and cytology of the macaque monkey midline, mediodorsal, and intralaminar thalamic nuclei. Brain Struct. Funct. 212: 465-479.) This specificity relates to the findings of the relationship of posterior DMN structures and loss of consciousness identified in the manuscript (see point 3).

The observation that thalamic seed is specifically functionally disconnected from posterior DMN regions (posterior cingulate, precuneus and inferior parietal cortex) is relevant to the specific role of the central lateral nucleus components of the intralaminar nuclei which send broad projections across these cortical regions (Buckwalter et al. J. Comp Neurol. 2008).

We thank the reviewer for this suggestion. We have incorporated these insights into the following revised discussion paragraph (“A putative functional network for recovery from unconsciousness comprising the locus ceruleus, central thalamus, and posterior cingulate cortex”).

Recent FDG-PET studies in severely brain injured subject demonstrate a graded relationship between central thalamus and the precuneus (Fridman et al. PNAS 2014) and provide a nice correlation of these findings which could be compared and discussed.

We thank the reviewer for this suggestion. We have modified the text in our Discussion to correlate our findings with the graded relationship between the central thalamus and precuneus that has been recently described in severely brain injured patients (“Reduced cerebral blood flow during recovery of consciousness”).

The correlation of the correlation of the Dex-induced unconscious state with the specific preservation of cortico-cortical functional connectivity suggests an interesting model for minimally conscious state. As the authors have noted this literature in the manuscript it might be worth a brief elaboration of the possible implication of these findings for similar states arising after structural brain injuries.

We thank the reviewer for this statement. In this manuscript we have attempted to carefully balance our anesthesia-induced findings with potential implications for DOC, sleep and the practice of anesthesia. However, we have rewritten our Discussion section to more explicitly elaborate on the potential implications of our findings. We sincerely hope that the modified text in our Discussion section would satisfy the reviewers concerns. Specifically in “Different states of unconsciousness may reflect a hierarchy of disruption in functional circuits” section of the discussion, we now explicitly state the following:

In the case of sleep, dexmedetomidine-induced unconsciousness, and some minimally conscious states, maintained cortico-cortical functional connectivity likely allows for the cortex to be “primed,” ready to recover from the altered consciousness states with restoration of thalamic functional connectivity, which could be triggered by ascending midbrain and brainstem inputs (Solt et al., 2014) or direct therapeutic intervention at the thalamic level (Schiff et al., 2007; Williams et al., 2013).

Reviewer #3:

Introduction:

The authors should quote original articles about the mechanism of action of dexmedetomidine, including the neural loci of action rather than their own review (Brown et al., 2011).

We regret this. We address this concern in this revision.

Methods:

The authors mention that principled constraints were placed on the connectivity analysis. These have not been referenced or defined elsewhere. It would be important to clarify.

By principled constraints, we meant that instead of multiple seed based analysis the FDG-PET and ASL fMRI results helped constrain our analysis to the DMN and FPN. We have rewritten the statement to read as follows:

“Third, we use the CMRglc results to inform the brain functional connectivity analysis derived from the Blood Oxygen Level Dependent (BOLD) fMRI signals.”

Results:

Does the Worsley (2001) method have the spatial resolution to be able to detect activity in thalamic subnuclei? If not did you consider using a voxel-level correction (so that the statistical inference is about the peak voxel, and is localised accurately) or cluster level correction with a more stringent voxel level threshold (so the activations are more focal and better localised, albeit still imperfectly)?

We agree with the reviewer that other methods for statistical thresholding may be more sensitive to focal, rather than distributed effects. However, we feel that our choice of using cluster correction at the standard cluster-forming threshold of z = 2.3 are appropriate for two reasons:

First, and perhaps most importantly, the factor limiting the resolution of our data is the spatial resolution of the scanner itself. It is commonly accepted that an object needs to be 2-3 times the full width at half maximum (FWHM) of the scanner in order to be properly resolved, with limited partial volume effects. The FWHM of the PET camera used in our study -the Siemens Biograph mMR- is 4.3 mm measured at 1 cm from the center of the field of view (Delso et al., J Nucl Med 2011). Therefore, it follows that only objects with a diameter larger than >1 cm would be fully resolved. Since most of the thalamic nuclei have a size that lies below this threshold (Mai, Paxinos and Voss, Atlas of the Human Brain, 2008, Elsevier), we don't feel that we would be able to successfully distinguish different nuclei without a significant amount of partial voluming.

In addition, we expected very widespread changes in our PET and fMRI signals due to the large effect size of our pharmacological-induced unconsciousness experimental model. As a result, we believe that the adoption of our thresholding methods of choice, which is maximally sensitive to widespread signals is appropriate.

Below we present our data from Figure 1D using more stringent cluster forming thresholds and voxel level threshold. Notably, the peak of the voxel-wise correction localized the peaks identified using cluster forming thresholds ( z = 2.3 or 3.1 or 3.3 ).

Author response image 1

Discussion:

The authors have compared their findings using dexmedetomidine with other author's studies on propofol. Is this comparison justified given the author's claim to have used a novel method to identify the changes in connectivity, an analysis that was not used by the other authors? Is it possible that the Akeju et al.'s methods are not as sensitive as other authors work?

We thank the reviewer for this comment. Even though we used a novel PET/MR method to acquire functional and metabolism data in subjects during the same experimental session, the method we used to compute functional connectivity analyses within specific networks (i.e., ‘dual regression’) has been widely used in the literature, and was found to be very sensitive and to have moderate-to-high test–retest reliability. Moreover, the scanning parameters we used generated a slightly higher spatial resolution than most studies (voxel size = 2.3 × 2.3 × 3.8 mm). For these reasons, we feel that the comparison with other connectivity studies in the literature is justified and is not confounded by lower sensitivity in our method.

There appears to be tautology associated with the authors contention that slow wave sleep is a “deeper level of unconsciousness” than that provided by dexmedetomidine. This therefore assumes that loss of consciousness (unconsciousness) is not binary. Therefore I am confused by their speculation.

We have rewritten this sentence to read as follows:

“However, during the more profound sleep state of nREM III, there is a…”

The authors do not discuss the finding of reduced CBF on recovery of responsiveness: this is one of the most interesting and perplexing findings that deserves some discussion.

We thank the reviewer for this comment. We address this concern in the revised version of this manuscript (response to comment 3 of the combined reviews above).

Figure 1 ignores the effect of DEX at postsynaptic alpha2 receptors, which appear to be important for the decrease in anesthetic requirements that is produced by dexmedetomidine (Segal et al 1988).

We regret this omission. In the revised caption, we acknowledge the effect of dexmedetomine at post-synaptic alpha-2-receptors.

https://doi.org/10.7554/eLife.04499.020

Article and author information

Author details

  1. Oluwaseun Akeju

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    Contribution
    OA, Conceived project, Designed experiments, Acquired data, analyzed data, Interpreted data and wrote the manuscript
    Contributed equally with
    Marco L Loggia
    For correspondence
    1. oluwaseun.akeju@mgh.harvard.edu
    Competing interests
    The author declare that no competing interests exist.
  2. Marco L Loggia

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    2. Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, United States
    Contribution
    MLL, Designed experiments, Analyzed data, Interpreted data, Wrote the manuscript
    Contributed equally with
    Oluwaseun Akeju
    For correspondence
    1. marco@nmr.mgh.harvard.edu
    Competing interests
    The author declare that no competing interests exist.
  3. Ciprian Catana

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    CC, Analyzed data, Interpreted data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  4. Kara J Pavone

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    Contribution
    KJP, Analyzed data, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  5. Rafael Vazquez

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    Contribution
    RV, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  6. James Rhee

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    Contribution
    JR, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  7. Violeta Contreras Ramirez

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    Contribution
    VCR, Analyzed data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  8. Daniel B Chonde

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    DBC, Analyzed data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  9. David Izquierdo-Garcia

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    DI-G, Analyzed data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  10. Grae Arabasz

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    GA, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  11. Shirley Hsu

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    SH, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  12. Kathleen Habeeb

    1. Clinical Research Center, Massachusetts General Hospital, Boston, United States
    Contribution
    KH, Acquired data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  13. Jacob M Hooker

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    Contribution
    JMH, Interpreted data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  14. Vitaly Napadow

    1. MGH/MIT/HMS Athinoula A Martinos Center for Biomedical Imaging, Charlestown, United States
    2. Department of Anesthesiology, Perioperative and Pain Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, United States
    Contribution
    VN, Interpreted data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  15. Emery N Brown

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    2. Department of Brain and Cognitive Science and Institute for Medical Engineering and Sciences, Massachusetts Institute of Technology, Cambridge, United States
    3. Harvard-Massachusetts Institute of Technology Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, United States
    Contribution
    ENB, Conceived project, Designed experiments, Interpreted data, Critically revised the manuscript
    Competing interests
    The author declare that no competing interests exist.
  16. Patrick L Purdon

    1. Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Harvard Medical School, Boston, United States
    2. Harvard-Massachusetts Institute of Technology Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, United States
    Contribution
    PLP, Interpreted data, Wrote the manuscript
    For correspondence
    1. patrickp@nmr.mgh.harvard.edu
    Competing interests
    The author declare that no competing interests exist.

Funding

Foundation for Anesthesia Education and Research

  • Oluwaseun Akeju

National Institutes of Health (DP1-OD003646)

  • Emery N Brown

National Institutes of Health (TR01-GM104948)

  • Emery N Brown

National Institutes of Health (DP2-OD006454)

  • Patrick L Purdon

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank P McCarthy, A Omoruan, K Shelton, and Norman Taylor for providing clinical care; J Kim for the acquisition of physiological data. We also thank the Massachusetts General Hospital Clinical Research Center for providing nursing support.

Ethics

Human subjects: The Human Research Committee and the Radioactive Drug Research Committee at the Massachusetts General Hospital approved the study protocol. After an initial email/phone screen, potential study subjects were invited to participate in a screening visit. At the screening visit, informed consent including the consent to publish was requested after the nature and possible consequences of the study was explained. All subjects provided informed consent and were American Society of Anesthesiology Physical Status I with Mallampati Class I airway anatomy.

Reviewing Editor

  1. Jody C Culham, Reviewing Editor, University of Western Ontario, Canada

Publication history

  1. Received: August 25, 2014
  2. Accepted: November 26, 2014
  3. Accepted Manuscript published: November 28, 2014 (version 1)
  4. Version of Record published: January 1, 2015 (version 2)

Copyright

© 2014, Akeju et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 2,320
    Page views
  • 375
    Downloads
  • 19
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Comments

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Developmental Biology and Stem Cells
    Cyrille Ramond et al.
    Research Article