Propofol disrupts the functional core-matrix architecture of the thalamus in humans

Huang, Zirui; Mashour, George A.; Hudetz, Anthony G.

doi:10.1038/s41467-024-51837-1

Download PDF

Article
Open access
Published: 09 September 2024

Propofol disrupts the functional core-matrix architecture of the thalamus in humans

Nature Communications volume 15, Article number: 7496 (2024) Cite this article

6136 Accesses
115 Altmetric
Metrics details

Subjects

Abstract

Research into the role of thalamocortical circuits in anesthesia-induced unconsciousness is difficult due to anatomical and functional complexity. Prior neuroimaging studies have examined either the thalamus as a whole or focused on specific subregions, overlooking the distinct neuronal subtypes like core and matrix cells. We conducted a study of heathy volunteers and functional magnetic resonance imaging during conscious baseline, deep sedation, and recovery. We advanced the functional gradient mapping technique to delineate the functional geometry of thalamocortical circuits, within a framework of the unimodal-transmodal functional axis of the cortex. Here we show a significant shift in this geometry during deep sedation, marked by a transmodal-deficient geometry. This alteration is closely linked to the spatial variations in the matrix cell composition within the thalamus. This research bridges cellular and systems-level understanding, highlighting the crucial role of thalamic core–matrix functional architecture in understanding the neural mechanisms of states of consciousness.

Local orchestration of distributed functional patterns supporting loss and restoration of consciousness in the primate brain

Article Open access 11 March 2024

Loss of consciousness reduces the stability of brain hubs and the heterogeneity of brain dynamics

Article Open access 06 September 2021

Consciousness-specific dynamic interactions of brain integration and functional diversity

Article Open access 10 October 2019

Introduction

Over the past three decades, the role of thalamocortical circuits in consciousness has been a focus of empirical and theoretical research^{1,2,3,4,5,6,7,8,9,10}. However, there is currently no established consensus about the thalamocortical system’s role in consciousness, primarily due to anatomical and functional complexity. The thalamus consists of numerous nuclei, each with distinct anatomical and functional properties^11,12,13. Broadly, thalamic nuclei can be divided into specific and nonspecific types. Specific thalamic nuclei establish direct, targeted connections to defined cortical areas, relaying sensory signals with high precision. Nonspecific nuclei, by contrast, project widely across multiple cortical areas, playing a crucial role in regulating general cortical arousal and supporting complex cognitive functions^4,14,15. Furthermore, the thalamus exhibits a heterogeneous cytoarchitecture with at least two distinct cell classes, known as core and matrix cells, which send differential projections to the cortex^14,15,16. Core cells primarily innervate the granular layers of the cerebral cortex, while matrix cells innervate the supragranular cortex in a relatively diffuse manner. Although specific thalamic nuclei are generally core-rich and nonspecific nuclei are matrix-rich, it is common for individual nuclei to contain both cell types¹⁷, and thus, in this respect, thalamic nuclei form a continuum rather than distinct categories⁷.

Recent investigations suggest that interactions between matrix-rich nonspecific thalamus and thick-tufted cortical layer-5 pyramidal neurons play a crucial role for consciousness^6,18,19,20. Layer-5 pyramidal neurons have distinct dendritic compartments, where the supragranular distal apical dendrites are primarily involved in corticocortical feedback processing, and the infragranular proximal somatic dendrites are primarily involved in corticocortical feedforward processing. Crucially, the matrix-rich nonspecific thalamus targets the oblique dendrites along the apical trunk, modulating the dendrosomatic coupling between these compartments. This coupling enables large-scale cortical integration, binding the more specialized core cells into a distributed coalition²¹. A disruption of the coordination between these apical and somatic inputs²⁰, potentially caused by a suppression of nonspecific thalamocortical connectivity^22,23, may contribute to the loss of consciousness characteristic of general anesthesia.

One class of nonspecific nuclei located within the internal medullary lamina is the intralaminar thalamic nuclei, whose projections innervate multiple cortical and subcortical regions^14,24. Some of the intralaminar nuclei, such as the central lateral and central medial nuclei, are extremely matrix-cell dense. Accordingly, electrical stimulation of these intralaminar nuclei using implanted electrodes awakened monkeys from anesthesia^25,26,27 and facilitated the recovery of consciousness in neuropathological patients^28,29,30. Likewise, microinjection of nicotine or potassium channel-blocking antibody into these intralaminar nuclei restored consciousness in rodents^31,32.

Despite the important roles assigned to nonspecific thalamic nuclei, unconscious states involve a widespread disruption of thalamocortical functional connectivity that affects both specific and nonspecific thalamic nuclei^{33,34,35,36,37,38,39}. To date, these conflicting views have not been reconciled, as most prior neuroimaging studies examined the thalamus either as a whole or focused on subregions rather than identifying specific neuronal subpopulations. Since core and matrix cells are distributed throughout the thalamus, including in both specific and nonspecific thalamic nuclei, there is a need to investigate the spatial distribution of thalamocortical connectivity more systematically, with a focus on core-matrix architecture.

In this work, we combine two techniques to achieve this goal in the human brain. First, we analyze the functional gradients across unimodal and transmodal cortical areas⁴⁰, which have a relationship to thalamic core-matrix architecture⁴¹. Unimodal areas are those that respond primarily to sensory input in a single modality (visual, auditory, or somatosensory), as well as motor areas, collectively underpinning perception and action. Transmodal areas are association cortices whose activity is associated with complex cognitive functions not specific to a single sensory or motor modality^40,42,43. Regions of the thalamus that are abundant in matrix cells exhibit stronger functional connectivity with transmodal cortical areas, whereas regions enriched by core cells display stronger functional connectivity with unimodal cortical areas⁴¹. Second, we advance the functional gradient mapping method⁴⁰ by aligning thalamocortical functional connectivity with the unimodal-transmodal functional axis of the cortex. This technique enables us to map, in a spatially precise manner, the connection between the thalamocortical functional axis and the thalamic core-matrix architecture, as inferred from the mRNA expression levels of two calcium-binding proteins: Calbindin (CALB1) (a marker of matrix cells) and Parvalbumin (PVALB) (a marker of core cells)^41,44. We demonstrate that loss of consciousness due to anesthesia is accompanied by a shift from balanced core–matrix functional geometry to a matrix-deficient one. We use the term “functional geometry” to describe the spatial arrangement of brain areas based on their functional connectivity patterns, as opposed to anatomical geometry, which focuses on their physical spatial configurations.

Results

Unimodal-transmodal functional geometry of thalamocortical circuits

We developed a voxel-based method to analyze the unimodal-transmodal functional geometry of thalamocortical circuits aligned with the unimodal-transmodal functional geometry of the cortex. In contrast to conventional functional connectivity techniques that use discrete areal demarcations, this allowed us to link the thalamocortical functional geometry with the thalamic core-matrix architecture. First, we utilized cortical gradient mapping to transform the functional brain connectome into a non-linear diffusion space⁴⁰. This method breaks down the functional similarity structure of the fMRI data into a series of embedding components (known as gradients), which characterize the major spatial axes along which functional connectivity varies across the entire cortex. Brain regions with similar activity fluctuations over time are grouped at one end of a gradient, and collectively they exhibit less similarity (and greater functional differentiation) than the cluster of regions located at the other end of the gradient. Here we focused on the primary gradient, which corresponds to the unimodal-transmodal functional axis of the cortex⁴⁰. Consistent with our previous findings⁴⁵, we found that the unimodal-transmodal functional gradient was degraded 13.3% during deep sedation (Supplementary Fig. 1).

Second, we examined the thalamic correlates of the unimodal-transmodal cortical gradient (Fig. 1a). Pair-wise correlations were computed between the thalamocortical connectivity values of each thalamic voxel and the cortical gradient values, generating a topographical map of the thalamocortical correlation coefficients of the principal cortical gradient, referred to as the gradient correlation coefficient (GCC). For any thalamic voxel, a positive GCC indicated transmodal-dominant connectivity (stronger functional connectivity with transmodal cortical areas than with unimodal cortical areas), whereas a negative GCC indicated unimodal dominance. Our results revealed that the overarching functional geometry of thalamocortical circuits was transmodal-dominant in the anterior and medial thalamus, and unimodal-dominant in the posterior thalamus (Fig. 1b). The consistency of this result was evaluated across diverse datasets collected from different research sites, utilizing different fMRI scanning parameters, scanning durations, and sample sizes, which included 1009 participants from the Human Connectome Project (HCP)⁴⁶, 116 participants from UCLA⁴⁷, and 27 participants (resting-state conscious baseline) from our current study. We found high spatial similarities across datasets that underscore the robustness of our method (Fig. 1c).

**Fig. 1: Thalamic correlates of the unimodal-transmodal cortical gradient.**

Shifting towards transmodal-deficiency during deep sedation

After validating our method with various datasets, we applied it to our current dataset with conscious baseline, deep sedation, and recovery conditions. Deep sedation was achieved by incrementally increasing the dose of the anesthetic propofol until loss of behavioral responsiveness (see “Methods” for details). We found that the thalamic unimodal-transmodal functional geometry was present during the conscious conditions (baseline and recovery), whereas a transmodal-deficient geometry was present across the entire thalamus during deep sedation (Fig. 2a). To further illustrate the geometric changes of thalamocortical circuits, we extracted the GCC from predefined thalamic areas⁴⁸ (Fig. 2b) and performed statistical comparisons between conscious conditions (baseline and recovery) and deep sedation. During conscious conditions, thalamic areas (bilateral) showed a progression from transmodal-dominant to unimodal-dominant areas as follows: VAs → DAm → VAi → DAl → VPm → DP → VPl. During deep sedation, however, all thalamic areas examined were transmodal-deficient. These findings were consistent irrespective of whether the participant was in a resting state (n = 27) or was listening to music (n = 27), and the results were robust regardless of whether global signal regression (GSR) was applied (Supplementary Fig. 2). Additionally, our findings were reproducible using an independent dataset (n = 26) that we previously published⁴⁹ (Supplementary Fig. 3). Collectively, these results indicate a geometric transformation in thalamocortical functional connectivity during deep sedation, specifically a shift from a unimodal-transmodal geometry to a transmodal-deficient one.

**Fig. 2: Changes of thalamocortical gradient correlation during deep sedation.**

Functional disruption of matrix cells during deep sedation

We investigated whether thalamocortical functional geometry across states was associated with the distribution of core vs. matrix cell compositions. The cell types were inferred from the mRNA expression levels of two calcium-binding proteins, CALB1 and PVALB, provided by the Allen Human Brain Atlas⁴⁴. The relative weighting of the difference between CALB1 and PVALB levels was defined by the CP_T metric⁴¹, where positive values indicated areas with higher CALB1 levels (matrix-rich) and negative values indicated areas with higher PVALB levels (core-rich) (Fig. 3a). We correlated the CP_T values with the group-averaged GCC across all thalamic voxels. During all experimental conditions (conscious baseline, deep sedation, and recovery), we observed statistically significant positive correlations (Fig. 3b), suggesting that matrix-rich thalamic areas were associated with transmodal thalamocortical functional connectivity, whereas core-rich thalamic areas were associated with unimodal thalamocortical functional connectivity. Importantly, this relationship was independent of the state of consciousness, despite the shift from a unimodal-transmodal to transmodal-deficient geometry in deep sedation. Furthermore, correlating CP_T values with changes in GCC (ΔGCC) from conscious baseline to deep sedation (deep sedation vs. conscious) revealed a statistically significant negative correlation (Fig. 3c). This indicates that matrix-rich thalamic areas were associated with a more substantial reduction in GCC, reflecting a greater suppression of transmodal thalamocortical functional connectivity. This finding was mirrored by the observed changes in GCC from deep sedation to recovery. To account for individual variations, we performed confirmatory analyses, with repeated-measures ANOVAs, by dividing thalamic areas into low, medium, and high CP_T groups. The results were consistent with the above correlation analyses using group-averaged GCC values (Supplementary Fig. 4a).

**Fig. 3: Relationship between core-matrix thalamic mRNA expression and gradient correlation coefficient.**

Next, we sought to determine the relative importance of core vs. matrix cell compositions (i.e., PVALB and CALB1 as two independent variables) in predicting the changes of group-averaged GCC (ΔGCC as dependent variable) from conscious to deep sedated states (Fig. 4). Using dominance analysis⁵⁰, we assessed the contribution of each cell composition to the overall model fit (adjusted R²) in a multiple linear regression model. We found that matrix cell composition had a substantially higher percentage of relative importance (CALB1, 84%, incremental R² = 0.395, p < 0.0001) compared to core cell composition (PVALB, 16%, incremental R² = 0.072, p < 0.0001). This indicates that matrix cell composition is a more important predictor of GCC variability during changes in consciousness. This finding was consistent with the observations from deep sedation to recovery (CALB1, 80%, incremental R² = 0.222, p < 0.0001; PVALB, 20%, incremental R² = 0.057, p < 0.0001). Individual-level dominance analysis confirmed these results, with the incremental R² of CALB1 significantly surpassing that of PVALB in both transitions (Supplementary Fig. 4b). To account for possible non-linear effects, we conducted dominance analyses on subsets of data categorized by core-rich (low CP_T), mixed (medium CP_T), or matrix-rich (high CP_T) voxels. Notably, matrix cell composition consistently exhibited a higher percentage of relative importance across all subsets (Supplementary Fig. 5).

**Fig. 4: Relative importance of core vs. matrix cell compositions in predicting the changes of gradient correlation coefficient.**

Taken together, these results suggest that the loss of consciousness during deep sedation is primarily associated with the functional disruption of matrix cells distributed throughout the thalamus.

To evaluate the specificity of core vs. matrix cell compositions in predicting changes in GCC, we extended the dominance analysis by including additional independent variables with neurotransmitter receptors and transporters (Fig. 5). The group-averaged tracer maps were constructed by compiling positron emission tomography (PET) images for 19 different neurotransmitter receptors, transporters, and receptor-binding sites across nine neurotransmitter systems: dopamine, norepinephrine, serotonin, acetylcholine, glutamate, GABA, histamine, cannabinoid, and opioid⁵¹. Our findings reinforced that CALB1 mRNA expression (corresponding to matrix cells) holds high relative importance compared to neurotransmitter receptors and transporters, strengthening our central findings.

**Fig. 5: Relative importance of thalamic cell compositions, neurotransmitter receptors and transporters in predicting the changes of gradient correlation coefficient.**

Additional confirmatory analyses

The following analyses involved conventional functional connectivity methods that are independent of the cortical gradient mapping approach. We aimed to reconcile global vs. specific thalamocortical changes during deep sedation by actively manipulating use of global signal regression, a common practice in fMRI data preprocessing. Based on prior studies, region-specific changes in thalamocortical functional connectivity are expected to occur in the background of an overall reduction in connectivity²². First, to estimate the overall changes in functional connectivity across different conditions, we refrained from performing global signal regression, because the overall level of brain-wide functional connectivity can differ between conscious and anesthetized states^52,53,54. This measure is referred to as absolute (unscaled) thalamocortical functional connectivity. We calculated the pair-wise functional connectivity between each of predefined 400 cortical areas⁵⁵ and each of the predefined 7 (bilateral) thalamic areas⁴⁸. We found an average reduction of 70% in thalamocortical functional connectivity during deep sedation compared to both conscious baseline and recovery. Significant functional connectivity reductions were identified across all thalamic areas we investigated (Supplementary Fig. 6). Second, we examined the relative (scaled) thalamocortical functional connectivity. Here we applied global signal regression, which removed the global effect of the fMRI signal while preserving the topological property of functional connectivity. We computed functional connectivity between each pre-defined thalamic area and the cortical areas associated with both unimodal functions (brain areas in the visual and somatomotor networks) and transmodal functions (brain areas in the frontoparietal and default-mode networks) separately. The results further support our findings regarding a shift from a balanced unimodal-transmodal geometry during consciousness to a deficit of transmodal geometry during deep sedation (Fig. 6). Collectively, these confirmatory analyses indicate both global and specific thalamocortical changes during deep sedation. The former is linked to an overall, absolute reduction in functional connectivity, while the latter is associated with geometric, relative changes characterized by a preferential suppression of transmodal functional connectivity.

**Fig. 6: Relative alterations of thalamocortical functional connectivity during deep sedation.**

Discussion

The goal of this investigation was to better understand the possible role of distributed thalamocortical functional geometry and connectivity across states of consciousness. We found that, alongside the expected decrease in overall thalamocortical functional connectivity, deep sedation was associated with specific alterations in a functional hierarchy within thalamocortical circuits. Specifically, there was a shift from a balanced unimodal-transmodal geometry during consciousness to a deficit of transmodal geometry during deep sedation. This shift in functional geometry was selectively associated with spatial variations in the matrix cell composition within the thalamus. In other words, thalamic regions with a high density of matrix cells exhibited a pronounced reduction in transmodal thalamocortical functional connectivity during deep sedation. Together, our findings suggest that loss of consciousness due to an anesthetic agent may be tied to the preferential disruption of matrix cell connectivity (Fig. 7).

**Fig. 7: Schematic illustration of the primary conclusion.**

Recent research has shifted from traditional methods focused on discrete areas to a continuum-based approach that examines the functional geometry of the cortex⁴⁰ and subcortex⁴⁸. However, the functional geometries of the cortex and thalamus have been examined separately, without an integrated approach that can characterize the organization of thalamocortical circuits as functioning units. Our investigation involved the development of an approach to identify the functional geometry of thalamocortical connectivity. We delineated a unimodal-transmodal gradient of thalamocortical connectivity, along a continuum, corresponding to the unimodal-transmodal functional axis of the cortex.

By applying this approach, we demonstrated that deep sedation was associated with a reorganization of thalamocortical functional geometry, with a shift towards a transmodal-deficient geometry. This result is consistent with but extends our recent observation of the functional degradation of unimodal-transmodal corticocortical gradient during suppressed consciousness⁴⁵. However, our study could not mechanistically determine whether the changes in functional geometry within thalamocortical or corticocortical connections mediated alterations in conscious states. Likewise, it is still unclear whether the anesthetic effects on thalamocortical circuits represent a readout of corticocortical circuits or a direct cause of conscious state transitions^1,56. Conversely, it is conceivable that thalamocortical circuits are just as essential as corticocortical circuits, and the integration of information into a conscious state may necessitate “closing the information loop” through a corticothalamocortical network^13,57.

We found that the distribution of thalamic core and matrix cell compositions closely aligned with the GCC, which reflects the functional geometry of thalamocortical circuits. Specifically, areas of the thalamus rich in matrix cells were linked to transmodal functional connectivity in the anterior and medial regions, while core-rich areas were associated with unimodal functional connectivity in the posterior regions. These findings are supported by the fact that calbindin-stained matrix cells are more abundant in the anterior and medial regions, while parvalbumin-stained core cells are predominantly located in the posterior thalamus^16,41,58,59.

Importantly, we found that the relative contribution of the matrix cell composition in explaining GCC variability for a transition from conscious to deep sedated states was approximately fourfold greater than that of core cells. In other words, the connectivity of matrix cells may have a much stronger influence on the observed changes in the functional geometry of thalamocortical circuits during this transition. Prior theories posited that the core cell-rich specific thalamic circuits responsible for encoding and transmitting sensory information remain largely intact under anesthesia, whereas the matrix cell-rich nonspecific thalamic circuits, which play a greater role in the temporal coordination, binding, and integration of information across the brain, are compromised when consciousness is suppressed⁴. Recent studies confirm that the diffusely projecting matrix cells of the thalamus could play an important role interacting with cortical pyramidal neurons in conscious sensory processing^{6,8,18,19,20,25,26,27,60}, but the current study provides empirical data supporting this hypothesis during conscious state changes in humans.

The shift towards a deficit of transmodal thalamic geometry during deep sedation may be attributed to the differential inputs received by core and matrix thalamic subpopulations, particularly in the ventral tier of the thalamus. Core cells predominantly receive glutamatergic driver inputs from sensory nuclei or the deep nuclei of the cerebellum, while the calbindin-stained matrix cells are under the GABAergic control of the globus pallidus internus, the main inhibitory output from the basal ganglia^7,61. Propofol’s enhancement of GABA-A receptor-mediated inhibition might affect matrix cells more profoundly, leading to the observed shift in thalamic functional geometry. Another possibility could be that matrix cells have more GABA-A receptors in their circuits, leading to a greater inhibitory effect. However, our analysis indicates that this is less likely due to the lower relative importance of GABA-A receptor spatial density compared to other factors. Further investigation is needed to definitively confirm the proposed mechanism and to fully elucidate the causal role of these thalamic subpopulations in the transition between conscious and unconscious states.

In addition, our findings may offer insights into the well-documented EEG phenomena observed during propofol anesthesia. Previous studies have shown a characteristic shift towards high-amplitude slow oscillations and frontal alpha oscillations⁶², with computational modeling highlighting the role of thalamic mechanisms in generating these oscillations⁶³. The disruption we observed in matrix thalamus to transmodal cortical connectivity, including frontal areas, aligns with these EEG findings. The rhythmic patterns of EEG signals likely reflect a breakdown in communication within matrix thalamocortical loops, hindering the ability of higher-order brain regions to exert control over sensory processing and behavior. This aligns with the established role of propofol in enhancing tonic inhibition within the corticothalamic feedback loop and synchronizing inhibitory input from the thalamic reticular nucleus to multiple thalamic nuclei⁶³. However, our data also bring into question the role of GABA-A receptors within the thalamus as mediating the observed oscillatory changes; future computational modeling should include other potential molecular determinants.

Prior research has led to two views on the role of thalamocortical circuits in anesthetic-induced loss of consciousness: (1) a global functional disconnection between the thalamus and cortex^{33,34,35,36,37,38}; (2) preferential disruptions in thalamocortical circuits involving nonspecific thalamic areas^20,22,23,64. Although these views are not mutually exclusive, a consensus has not emerged because most prior investigations either treated the thalamus as a whole or examined a specific subregion rather than exploring it as a functional continuum based on its cellular distribution. Furthermore, the influence of global brain signal changes has not been adequately considered. Our findings reconcile these views by demonstrating that deep sedation induces a shift in thalamocortical functional geometry from a balanced unimodal-transmodal mode towards a transmodal-deficient pattern. Importantly, these changes occur in conjunction with a generalized reduction in functional connectivity.

Our study has several limitations worth noting. First, we focused solely on the unimodal-transmodal functional gradient but at least two other widely studied cortical gradients exist: the visual-to-somatomotor gradient and the task-negative-to-task-positive gradient⁴⁵. It remains to be explored how thalamocortical circuits are associated with these other cortical gradients and whether they are also affected by anesthetics. Second, we used propofol at a dose that induced deep sedation only, whereas the observed effects may be both dose- and agent-dependent. For example, higher doses of propofol at the surgical level might lead to different neural dynamics compared to deep sedation. Additionally, other non-GABAergic anesthetics, such as dexmedetomidine or ketamine, have distinct mechanisms of action and may produce different effects on thalamocortical circuits and cortical gradients. Future studies should investigate a range of doses and various anesthetic agents to better understand their differential impacts on brain function and connectivity. Third, loss of consciousness was inferred from behavioral unresponsiveness to verbal commands, which could indicate a state of disconnectedness but not necessarily complete unconsciousness^65,66. This distinction is crucial because deep sedation, rather than general anesthesia, might allow for internal experiences disconnected from the external environment. If the state in question reflects disconnectedness rather than full unconsciousness, it could imply that matrix thalamocortical interactions are particularly important for maintaining connectedness. Although this cannot be distinguished with the current data, it is important to consider this alternative hypothesis. Fourth, while our study highlights the role of thalamic matrix cells in loss of consciousness, we acknowledge several limitations. Specifically, our analysis was based on the Allen Human Brain Atlas and did not account for potential inter-individual variability in gene expression. Additionally, we inferred the thalamic core-matrix architecture from mRNA expression levels of calcium-binding proteins, rather than directly measuring cell types. Recent work also challenges the traditional core-matrix classification, emphasizing the diverse morphology and connectivity patterns of thalamocortical projections⁶⁷. To further validate and extend our findings, future studies should employ multimodal imaging techniques, such as integrating ultrahigh field strength 7T-fMRI with in vivo cell type classification methods, in larger and more diverse samples. This would enable a more comprehensive understanding of thalamocortical interactions and their role in consciousness, accounting for both inter-individual variability and the complex nature of thalamocortical projections.

We conclude that deep sedation induced by propofol is accompanied by a decrease in overall thalamocortical functional connectivity, with a shift from balanced core–matrix functional geometry to a matrix-deficient one. By synthesizing cellular-level data with systems-level findings, our research illuminates the pivotal role of thalamic matrix cells in understanding the neural mechanisms of states of consciousness.

Methods

Participants

The University of Michigan Institutional Review Board (IRB) approved the experimental protocol. All methods were performed in accordance with the relevant guidelines and regulations. Thirty healthy participants were recruited (male/female: 10/20; age: 18–38 years; right-handed). All participants provided informed consent and received compensation post-experiment. We ensured strict confidentiality at all stages. From their initial involvement, participants were assigned a unique code number, which was used as the sole identifier on all specimen samples, behavioral and physiological records, and magnetic resonance scans.

The sex of participants was determined based on self-report, acknowledging that self-reported sex may not always align with biological sex. Gender identity was not collected as part of this study. Although our study design included both male and female participants, sex-based analyses were not conducted a priori due to the nature of the study design and the limited sample size, which may be insufficient for meaningful post hoc analysis. However, the source data is reported disaggregated by sex for transparency and potential future investigation.

Inclusion and exclusion criterias

Eligible participants were right-handed, healthy, aged 18–40, with a body mass index under 30, and classified as American Society of Anesthesiologists physical status 1. Exclusion criteria included MRI contraindications (potential pregnancy, extreme obesity, metallic body implants, claustrophobia, anxiety, cardiopulmonary issues), history of neurological, cardiovascular, or pulmonary diseases, significant head injury with consciousness loss, learning disabilities, developmental disorders, sleep apnea, severe snoring, sensory/motor impairments affecting study participation, gastroesophageal reflux disease, unwilling to abstain from alcohol 24 h before MRI, drug use history or positive drug screening, tattoos on head/neck, egg allergy, intracranial abnormalities on T1-weighted MRI scans, or discomfort during fMRI scanning.

Anesthetic agents

Propofol served as our reference drug because it is the most utilized agent in human fMRI studies examining anesthetic effects. A favorable property of propofol is that it has minimal impact on cerebral hemodynamics⁶⁸. The primary mechanism through which propofol suppresses neuronal activity involves the enhancement of GABA-A receptor-mediated inhibition, thereby modulating widespread targets throughout the brain¹. Concerning safety in healthy volunteers, a comprehensive multicenter study revealed no adverse effects of propofol-induced anesthesia in the absence of surgery. Cognitive function returned to baseline within 3 h after emerging from prolonged anesthesia, with no indications of disrupted arousal states in the subsequent days⁶⁹.

Anesthetic administration and monitoring

Prior to the study, participants fasted for 8 h. On the day of the experiment, an attending anesthesiologist conducted preoperative assessment and physical examination. Throughout the experiment, two fully trained anesthesiologists were physically present, maintaining continuous monitoring of spontaneous respiration, end-tidal CO₂ levels, heart rate, pulse oximetry, and electrocardiogram. Noninvasive arterial pressure was measured using an MRI-compatible monitor compatible with MRI technology. Following a subcutaneous injection of lidocaine (0.5 ml of 1%) as a local anesthetic, an intravenous cannula was placed. Participants were supplied with supplemental oxygen (administered at 2 l/min via a nasal cannula).

Propofol was administered by an intravenous bolus followed by continuous infusion. The bolus dose, infusion rate, and duration of the infusion were pre-established for each target effect site concentration utilizing the Marsh model incorporated into the STANPUMP software (http://opentci.org/code/stanpump). The infusion rate was manually controlled to achieve the target effect-site concentrations of 1.5, 2.0, 2.5, and 3.0 μg/ml, in a stepwise fashion. Each target concentration was maintained for 4 min. In this way, we were able to titrate the anesthetic level to achieve loss of behavioral responsiveness. To minimize head motion-related artifacts, the effect-site concentration was maintained at one step above the actual concentration for loss of responsiveness for approximately 32 min. For example, if a participant lost responsiveness at 2.0 μg/ml, we maintained 2.5 μg/ml effect-site concentration during the entire period of lost responsiveness. For exceptional cases when participants retained responsiveness at 3.0 μg/ml (occurrence rate of 7%), we increased target effect-site concentrations to 3.5 μg/ml and maintained at maximum of 4.0 μg/ml. The infusion was then terminated and propofol concentration was allowed to gradually decrease. Participants were instructed to perform a behavioral test, rest, or listen to the music as described below.

Experimental design

Eight fMRI scans (16 min per scan) were conducted throughout the experiment. The eight scans included 16 min resting-state and 16-min music-listening during wakefulness baseline, 16 min behavioral test during propofol induction, 16 min resting-state and 16-min music-listening after loss of behavioral responsiveness, 16 min behavioral test during emergence period (after propofol infusion was terminated), 16 min resting-state and 16 min music-listening after recovery of behavioral responsiveness. There was a 1–5 min break between the scans. The imaging protocols and data acquisition were completed within 2.5 h in each subject.

During the rest-state period, participants were asked to lay with eyes closed, try to stay awake without thinking of anything in particular, and try not to move. During the music presentation period, participants were asked to listen to the music keeping their eyes closed, try not to move, and stay awake. We selected well-known music excerpt from four types of music, including Jazz, Rock, Pop and Country, and they were presented in a pseudo-randomized order. Each piece of music was edited into 4 min duration. During the behavioral test period, participants were asked to squeeze an MRI compatible grip dynamometer (a rubber ball) for every 10 s period (96 cycles in total). The beginning of each cycle was cued with the spoken word “squeeze.” The verbal instructions were programmed using E-Prime 3.0 (Psychology Software Tools, Pittsburgh, PA) and delivered via an audiovisual stimulus presentation system designed for an MRI environment. The volume of the headphones was adjusted for subject comfort during wakefulness. The volume was increased to 150% after loss of responsiveness. Behavioral responses were measured in mmHg of air pressure during squeezing the rubber ball, using BIOPAC MP160 system with AcqKnowledge software (V5.0). By comparing the timing of “squeeze” instructions (expected motor response) and the actual motor response, the time points at which a participant lost responsiveness, and regained responsiveness were determined. The onsets of loss and recovery of responsiveness were defined as the time when participants first failed to squeeze, and the first time they were able to squeeze again, respectively.

fMRI data acquisition

Data were acquired using a 3T Philips scanner with a standard 32-channel transmit/receive head coil at University of Michigan Hospital. T1 weighted spoiled gradient recalled echo (SPGR) images were acquired for high spatial resolution of anatomical images with parameters: 170 sagittal slices, 1.0 mm thickness, repetition time (TR) = 8.1 s, echo time (TE) = 3.7 ms, flip angle = 8^o, FOV = 24 cm, image matrix 256 × 256. Functional images over the whole brain were acquired by a gradient-echo echo planar imaging (EPI) pulse sequence: 40 slices, TR/TE = 1400/30 ms by multiband acquisition, MB factor = 4, slice thickness = 2.9 mm, in-plane resolution = 2.75 × 2.75 mm; FOV = 220 mm, flip angle = 76°, image matrix: 80 × 80.

fMRI data preprocessing

Preprocessing steps followed an established pipeline^45,49,70 implemented in AFNI (linux_ubuntu_16_64; http://afni.nimh.nih.gov/) and consisted of the following steps: (1) Slice timing correction, (2) Rigid head motion correction/realignment, (3) Frame-wise scrubbing of head motion, (4) Co-registration with T1 anatomical images, (5) Spatial normalization into MNI stereotactic space resampled to 3 × 3 × 3 mm, (6) Band-pass filtering to 0.01–0.10 Hz with various regressors including linear and nonlinear drift, time series of head motion, white matter and cerebrospinal fluid, (7) Spatial smoothing with 6 mm FWHM isotropic Gaussian kernel, (8) Temporal normalization to zero mean and unit variance. The analysis excluded resting-state data from three participants and music-listening data from two participants due to excessive movement defined as frame-wise displacement greater than 0.8 mm in more than 25% of the data. In one other participant, data from music listening had to be omitted due to MRI technical issues. This resulted in a final dataset of 28 participants: 27 for resting-state data, 27 for music-listening data, with 26 participants overlapping between both types of data. Unless otherwise stated, global signal regression (GSR) procedure was applied to mitigate unwanted confounding factors such as low-frequency respiratory volume and cardiac rate⁷¹ while maintaining the topological characteristics of functional connectivity. Prior research demonstrated that the GSR approach does not impact the outcomes obtained from functional gradient analysis⁴⁵. Nonetheless, to assess the robustness of our findings under various processing methodologies, we conducted additional analyses without utilizing the GSR procedure as a control.

Cortical gradient analysis

Utilizing a common brain parcellation scheme⁷², we extracted fMRI time courses from 400 predetermined cortical regions⁵⁵. For every participant and condition, a 400 × 400 connectivity matrix was generated through Pearson correlation. Cortical gradients were computed using the BrainSpace toolbox (https://brainspace.readthedocs.io/en/latest/)⁷³ as implemented in MATLAB R2023b. Similar to prior studies⁴⁰, we applied a z-transformation and set a threshold for the connectivity matrix at 90% sparsity. This threshold retained the highest 10% of weighted connections per row, after which we computed a normalized cosine angle affinity matrix to quantify the similarity in connectivity profiles among cortical areas. Utilizing a diffusion map embedding algorithm, we identified gradient components as a low-dimensional representation of the connectivity matrix. The algorithm is governed by two key parameters: α, which regulates the influence of the sampling point density on the manifold (α = 0 for maximal influence, α = 1 for no influence), and t, which controls the scale of eigenvalues of the diffusion operator. We held α constant at 0.5 and set t to 0, thereby preserving global relationships between data points in the embedded space. Notably, a value of t = 0 indicates that the diffusion time is determined automatically through a damped regularization process⁴⁰. The selection of parameters including sparsity and α was based on our previous study⁴⁵, which demonstrated that 90% sparsity threshold was optimal for effectively detecting group disparities, and the parameter α had minimal impact on the overall robustness of the results.

Using Procrustes rotation, we aligned gradient solutions to a subset of the HCP dataset consisting of 217 samples available in the Brainspace toolbox⁷³. We aimed to find an orthogonal linear transformation ψ that would superimpose a source representation S onto a target representation T. The goal was to ensure that ψS and T coincide. This transformation was necessary because gradients computed separately from different individuals might not be directly comparable due to variations in eigenvector orderings when eigenvalues are multiple or due to sign ambiguity in the eigenvectors. As a result of this alignment step, the estimation of gradients became more stable, and the solutions became more comparable to those found in existing literature⁷³. To illustrate the organization of cortical gradients at the network level, we extracted gradient eigenvector loading values from seven predefined functional networks⁷². The numerical range of the primary gradient (Gradient-1) was determined by calculating the distance between the minimum and maximum gradient eigenvector values. This measurement provides insight into the segregation, i.e., differences in connectivity profiles, between the extreme points of the gradient.

Thalamic correlates of cortical gradient

A binary thalamic mask consisting of 720 voxels was derived from a previous study⁴⁸. The fMRI time courses were extracted from each thalamic voxel. A thalamocortical connectivity matrix (720 thalamic voxels × 400 cortical areas) was calculated using Pearson correlation with Fisher’s Z transformation applied. Next, we calculated pairwise Pearson correlations between the thalamocortical connectivity values of each thalamic voxel (400 functional connectivity values) and the 400 cortical gradient values. This computation generated a topographical map illustrating the correlation coefficients between the thalamocortical connectivity and the primary cortical gradient (resulting in 720 coefficients). The correlation coefficients, with Fisher’s Z transformation applied, were collectively termed the GCC. In sum, this method quantified the covariation between the functional connectivity profiles of each voxel in the thalamus and the cortical gradients.

Thalamic parcellation

Thalamic parcellation was derived from a Subcortical Atlas⁴⁸ including bilateral superior ventroanterior thalamus (VAs), medial dorsoanterior thalamus (DAm), inferior ventroanterior thalamus (VAi), lateral dorsoanterior thalamus (DAl), medial ventroposterior thalamus (VPm), dorsoposterior thalamus (DP), and lateral ventroposterior thalamus (VPl).

The thalamic matrix and core cell types

The cell types were inferred from the mRNA expression levels of two calcium-binding proteins, CALB1 and PVALB, as provided by the Allen Human Brain Atlas⁴⁴. From the original AHBA analysis, which consisted of 58,692 probes⁴⁴, two probes were selected to estimate PVALB expression (CUST_11451_PI416261804 and A_23_P17844) and three probes were selected for estimating CALB1 expression (CUST_140_PI416408490, CUST_16773_PI416261804, and A_23_P43197). The relative weights of the difference between CALB1 and PVALB levels were determined using the CP_T metric⁴¹, as provided by James M. Shine’s Lab (https://github.com/macshine/corematrix). Positive values indicated regions with higher CALB1 levels (matrix-rich), while negative values indicated regions with higher PVALB levels (core-rich). The binary thalamic mask was constrained by the overlap with the thalamic CP_T map, resulting in 467 voxels. This constrained mask was then used in subsequent analyses to link core-matrix thalamic mRNA expression with GCCs.

Neurotransmitter receptors and transporters

Whole-brain neurotransmitter receptor density maps were provided by Hansen et al.⁵¹. These group-averaged tracer maps were constructed by compiling PET images for 19 different neurotransmitter receptors, transporters, and receptor-binding sites across nine neurotransmitter systems: dopamine, norepinephrine, serotonin, acetylcholine, glutamate, GABA, histamine, cannabinoid, and opioid. Volumetric positron emission tomography images were collected from multiple studies. Detailed information on each study, including the associated receptor/transporter, tracer, number of healthy participants, age, and full methodological details, can be found in ref. ⁵¹. The images provide estimates proportional to receptor densities, referred to as density, based on measured values such as binding potential and tracer distribution volume. Although the dataset includes both cortical and subcortical regions, our analyses were restricted to the thalamus. In total, we present thalamic tracer maps for 19 unique neurotransmitter receptors and transporters. When multiple tracer maps were available for the same receptor/transporter, the map derived from the largest sample size was selected. Finally, each tracer map was z-scored across thalamic voxels.

Statistics and reproducibility

For a given measurement (either GCC value or functional connectivity value), we conducted non-parametric Wilcoxon signed-rank tests to compare conscious baseline versus deep sedation and deep sedation versus recovery for both resting-state and music presentation. Statistics were reported with W values, z values, p values, effect size estimates, and 95% confidence intervals. Using the Benjamini–Hochberg procedure, p values were false discovery rate–corrected (FDR-corrected) for multiple comparisons and thresholded at α = 0.05. Spearman rank correlations were performed between core-matrix thalamic mRNA expression (e.g., CP_T values) and the group-averaged GCC across all thalamic voxels. Statistics were reported with rho values, p values, and 95% confidence intervals. Spin permutation tests, implemented using the ENIGMA Toolbox (https://enigma-toolbox.readthedocs.io), were employed to determine p values while controlling for potential inflation of significance due to spatial autocorrelations⁷⁴. All reported p values are two-sided.

To determine the relative importance of core versus matrix cell compositions (independent variables) in predicting changes in GCC (dependent variable) from conscious to deep sedated states, we performed dominance analysis⁵⁰. Dominance analysis defines “relative importance” based on the additional contribution of a predictor in all subset models. Specifically, the analysis evaluates the contribution of each independent variable to the overall fit (adjusted R²) of a multiple linear regression model. This is achieved by fitting the regression model on every possible combination of input variables (2^p − 1 submodels for a model with p input variables). Total dominance is defined as the average relative increase in R² (incremental R²) when adding a single input variable to a submodel, across all submodels. The sum of the dominance values of all input variables equals the total adjusted R² of the complete model. This provides an intuitive method to partition the total effect size across predictors. Unlike methods that rely on regression coefficients or univariate correlations, dominance analysis accounts for interactions between predictors, making it more interpretable. Statistics were reported with incremental R², percentage relative importance, and p values. All reported p values are two-sided.

To evaluate the specificity of core versus matrix cell compositions in predicting changes in GCC, we extended the dominance analysis by including additional independent variables from 19 receptors and transporters. Since dominance analysis is computationally intensive as it builds all subset models, we selected the top 10 predictors based on F-regression.

Finally, we tested the reproducibility our results in the following three independent datasets.

HCP participants

The study sample was part of the S1200 Release of the WU-Minn HCP database that has been fully described elsewhere⁴⁶. Participants were healthy young adults ranging between 22 and 37 years old. From 1206 healthy participants, participants with fully completed structural MR scans and two sessions (REST1 and REST2) of resting-state fMRI scans were selected, resulting in a total of 1009 participants. The data were acquired using multiband EPI on a customized Siemens 3 T MR scanner (Skyra system), where each session comprised two runs (left-to-right and right-to-left phase encoding) of 14 min and 33 s each (TR = 720 ms, TE = 33.1 ms, voxel dimension: 2-mm isotropic). The two runs were temporally concatenated for each session, yielding 29 min and 6 s of data in each session. Concatenation of the two different phase-encoded data ensured that any potential effect of phase encoding on gradient direction was counterbalanced by the opposing phase encoding. ICA-FIX denoised volumetric data were sourced from the online HCP repository.

UCLA healthy control group

The dataset used in this study was sourced from the OpenNeuro database, which is maintained by the UCLA Consortium for Neuropsychiatric Phenomics⁴⁷. Ethical approval for this research was granted by the IRBs at UCLA and the Los Angeles County Department of Mental Health. All participants provided informed consent and received compensation for their participation following the experiment. The original dataset initially comprised 272 participants aged between 21 and 50 years, including both healthy individuals (n = 130) and individuals with psychiatric disorders. Further information regarding the characteristics of the study population can be found in the reference ⁴⁷. Data were excluded from our analysis if they lacked T-1 images or resting-state data, if overall head motion exceeded a range of 3 mm, or if there were insufficient degrees of freedom after motion scrubbing and band-pass filtering⁷⁰. As a result, our study included 116 healthy individuals.

Michigan propofol deep sedation

The dataset utilized in this study has been previously published with different analyses than those applied here^49,52. The experimental protocol was approved by the IRB at the University of Michigan, and all methods adhered to the relevant guidelines and regulations. Twenty-six healthy participants were recruited, consisting of 13 males and 13 females, with ages ranging from 19 to 34 years. All participants were right-handed and classified as American Society of Anesthesiologists physical status 1. Informed consent was obtained from all participants, who were compensated for their involvement after completing the experiment. Data were acquired using a 3T Philips scanner equipped with a standard 32-channel transmit/receive head coil at Michigan Medicine, University of Michigan, with the following scanning parameters: TR/TE = 800/25 ms, in-plane resolution = 3.4 × 3.4 mm, and slice thickness = 4 mm. Further details on scanning parameters can be found in ref. ⁴⁹. The fMRI scans used in this study included a 15 min conscious baseline, a 30 min period during propofol infusion, a 30 min period after propofol infusion, and another 15 min recovery baseline. Deep sedation was defined by loss of behavioral responsiveness.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Individual thalamic GCC maps are publicly available from figshare repository (https://doi.org/10.6084/m9.figshare.25927939). Source data to plot the figures are provided with this paper. The HCP dataset is available from online repository (https://www.humanconnectome.org/). The UCLA dataset is available at OpenfMRI (https://openfmri.org/dataset/ds000030/). Whole-brain neurotransmitter receptor density maps are available at GitHub (https://github.com/netneurolab/hansen_receptors). Source data are provided with this paper.

Code availability

Publicly available software and toolbox used for analyses include AFNI, BrainSpace (https://brainspace.readthedocs.io/en/latest/), ENIGMA Toolbox (https://enigma-toolbox.readthedocs.io), JASP v0.18.1.0 (https://jasp-stats.org/), and Dominance-Analysis (https://github.com/dominance-analysis/dominance-analysis). Custom code is publicly available from figshare repository (https://doi.org/10.6084/m9.figshare.25927939).

References

Alkire, M. T., Hudetz, A. G. & Tononi, G. Consciousness and anesthesia. Science 322, 876–880 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Laureys, S. et al. Restoration of thalamocortical connectivity after recovery from persistent vegetative state. Lancet 355, 1790–1791 (2000).
Article CAS PubMed Google Scholar
Alkire, M. T., Haier, R. J. & Fallon, J. H. Toward a unified theory of narcosis: brain imaging evidence for a thalamocortical switch as the neurophysiologic basis of anesthetic-induced unconsciousness. Conscious. Cogn. 9, 370–386 (2000).
Article CAS PubMed Google Scholar
Llinás, R., Ribary, U., Contreras, D. & Pedroarena, C. The neuronal basis for consciousness. Philos. Trans. R. Soc. Lond. Ser. B 353, 1841–1849 (1998).
Article Google Scholar
Izhikevich, E. M. & Edelman, G. M. Large-scale model of mammalian thalamocortical systems. Proc. Natl Acad. Sci. 105, 3593–3598 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Aru, J., Suzuki, M. & Larkum, M. E. Cellular mechanisms of conscious processing. Trends Cogn. Sci. 24, 814–825 (2020).
Article PubMed Google Scholar
Shine, J. M. The thalamus integrates the macrosystems of the brain to facilitate complex, adaptive brain network dynamics. Prog. Neurobiol. 199, 101951 (2021).
Article PubMed Google Scholar
Munn, B. R. et al. A thalamocortical substrate for integrated information via critical synchronous bursting. Proc. Natl Acad. Sci. 120, e2308670120 (2023).
Article CAS PubMed PubMed Central Google Scholar
Edelman, G. M. Neural Darwinism: Selection and reentrant signaling in higher brain function. Neuron 10, 115–125 (1993).
Article CAS PubMed Google Scholar
Whyte, C. J., Redinbaugh, M. J., Shine, J. M. & Saalmann, Y. B. Thalamic contributions to the state and contents of consciousness. Neuron 112, 1611–1625 (2024).
Article CAS PubMed Google Scholar
Sherman, S. M. & Guillery, R. W. Functional organization of thalamocortical relays. J. Neurophysiol. 76, 1367–1395 (1996).
Article CAS PubMed Google Scholar
Shine, J. M. et al. The low-dimensional neural architecture of cognitive complexity is related to activity in medial thalamic nuclei. Neuron 104, 849–855.e3 (2019).
Article CAS PubMed Google Scholar
Shine, J. M., Lewis, L. D., Garrett, D. D. & Hwang, K. The impact of the human thalamus on brain-wide information processing. Nat. Rev. Neurosci. 24, 416–430 (2023).
Article CAS PubMed PubMed Central Google Scholar
Jones, E. G. The thalamic matrix and thalamocortical synchrony. Trends Neurosci. 24, 595–601 (2001).
Article CAS PubMed Google Scholar
Jones, E. G. Synchrony in the interconnected circuitry of the thalamus and cerebral cortex. Ann. N.Y. Acad. Sci. 1157, 10–23 (2009).
Article ADS PubMed Google Scholar
Jones, E. G. Viewpoint: the core and matrix of thalamic organization. Neuroscience 85, 331–345 (1998).
Article CAS PubMed Google Scholar
Clascá, F., Rubio‐Garrido, P. & Jabaudon, D. Unveiling the diversity of thalamocortical neuron subtypes. Eur. J. Neurosci. 35, 1524–1532 (2012).
Article PubMed Google Scholar
Aru, J., Suzuki, M., Rutiku, R., Larkum, M. E. & Bachmann, T. Coupling the state and contents of consciousness. Front. Syst. Neurosci. 13, 43 (2019).
Article PubMed PubMed Central Google Scholar
Honjoh, S. et al. Regulation of cortical activity and arousal by the matrix cells of the ventromedial thalamic nucleus. Nat. Commun. 9, 2100 (2018).
Article ADS PubMed PubMed Central Google Scholar
Suzuki, M. & Larkum, M. E. General anesthesia decouples cortical pyramidal neurons. Cell 180, 666–676 (2020).
Article CAS PubMed Google Scholar
Varela, F., Lachaux, J.-P., Rodriguez, E. & Martinerie, J. The brainweb: phase synchronization and large-scale integration. Nat. Rev. Neurosci. 2, 229–239 (2001).
Article CAS PubMed Google Scholar
Liu, X., Lauer, K. K., Ward, B. D., Li, S.-J. & Hudetz, A. G. Differential effects of deep sedation with propofol on the specific and nonspecific thalamocortical systems. Anesthesiology 118, 59–69 (2013).
Article CAS PubMed Google Scholar
Müller, E. J. et al. The non-specific matrix thalamus facilitates the cortical information processing modes relevant for conscious awareness. Cell Rep. 42, 112844 (2023).
Article PubMed Google Scholar
Smith, Y., Raju, D. V., Pare, J.-F. & Sidibe, M. The thalamostriatal system: a highly specific network of the basal ganglia circuitry. Trends Neurosci. 27, 520–527 (2004).
Article CAS PubMed Google Scholar
Bastos, A. M. et al. Neural effects of propofol-induced unconsciousness and its reversal using thalamic stimulation. Elife 10, e60824 (2021).
Article CAS PubMed PubMed Central Google Scholar
Redinbaugh, M. J. et al. Thalamus modulates consciousness via layer-specific control of cortex. Neuron 106, 66–75 (2020).
Article CAS PubMed PubMed Central Google Scholar
Tasserie, J. et al. Deep brain stimulation of the thalamus restores signatures of consciousness in a nonhuman primate model. Sci. Adv. 8, eabl5547 (2022).
Article CAS PubMed PubMed Central Google Scholar
Magrassi, L. et al. Results of a prospective study (CATS) on the effects of thalamic stimulation in minimally conscious and vegetative state patients. J. Neurosurg. 125, 972–981 (2016).
Article PubMed Google Scholar
Schiff, N. D. et al. Behavioural improvements with thalamic stimulation after severe traumatic brain injury. Nature 448, 600–603 (2007).
Article ADS CAS PubMed Google Scholar
Schiff, N. D. et al. Thalamic deep brain stimulation in traumatic brain injury: a phase 1, randomized feasibility study. Nat. Med. 29, 3162–3174 (2023).
Article CAS PubMed PubMed Central Google Scholar
Alkire, M. T., McReynolds, J. R., Hahn, E. L. & Trivedi, A. N. Thalamic microinjection of nicotine reverses sevoflurane-induced loss of righting reflex in the rat. Anesthesiology 107, 264–272 (2007).
Article CAS PubMed Google Scholar
Alkire, M. T., Asher, C. D., Franciscus, A. M. & Hahn, E. L. Thalamic microinfusion of antibody to a voltage-gated potassium channel restores consciousness during anesthesia. Anesthesiology 110, 766–773 (2009).
Article PubMed Google Scholar
White, N. S. & Alkire, M. T. Impaired thalamocortical connectivity in humans during general-anesthetic-induced unconsciousness. Neuroimage 19, 402–411 (2003).
Article PubMed Google Scholar
Boveroux, P. et al. Breakdown of within- and between-network resting state functional magnetic resonance imaging connectivity during propofol-induced loss of consciousness. Anesthesiology 113, 1038–1053 (2010).
Article CAS PubMed Google Scholar
Akeju, O. et al. Disruption of thalamic functional connectivity is a neural correlate of dexmedetomidine-induced unconsciousness. Elife 3, e04499 (2014).
Article PubMed PubMed Central Google Scholar
Guldenmund, P. et al. Thalamus, brainstem and salience network connectivity changes during propofol-induced sedation and unconsciousness. Brain Connect. 3, 273–285 (2013).
Article PubMed Google Scholar
Huang, Z. et al. Altered temporal variance and neural synchronization of spontaneous brain activity in anesthesia. Hum. Brain Mapp. 35, 5368–5378 (2014).
Article PubMed PubMed Central Google Scholar
Malekmohammadi, M., Price, C. M., Hudson, A. E., DiCesare, J. A. T. & Pouratian, N. Propofol-induced loss of consciousness is associated with a decrease in thalamocortical connectivity in humans. Brain 142, 2288–2302 (2019).
Article PubMed PubMed Central Google Scholar
Kantonen, O. et al. Decreased thalamic activity is a correlate for disconnectedness during anesthesia with propofol, dexmedetomidine and sevoflurane but not S-Ketamine. J. Neurosci. 43, 4884–4895 (2023).
Article CAS PubMed PubMed Central Google Scholar
Margulies, D. S. et al. Situating the default-mode network along a principal gradient of macroscale cortical organization. Proc. Natl Acad. Sci. 113, 12574–12579 (2016).
Article ADS CAS PubMed PubMed Central Google Scholar
Müller, E. J. et al. Core and matrix thalamic sub-populations relate to spatio-temporal cortical connectivity gradients. Neuroimage 222, 117224 (2020).
Article PubMed Google Scholar
Huntenburg, J. M., Bazin, P.-L. & Margulies, D. S. Large-scale gradients in human cortical organization. Trends Cogn. Sci. 22, 21–31 (2018).
Article PubMed Google Scholar
Murphy, C. et al. Distant from input: evidence of regions within the default mode network supporting perceptually-decoupled and conceptually-guided cognition. Neuroimage 171, 393–401 (2018).
Article PubMed Google Scholar
Hawrylycz, M. J. et al. An anatomically comprehensive atlas of the adult human brain transcriptome. Nature 489, 391–399 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Huang, Z., Mashour, G. A. & Hudetz, A. G. Functional geometry of the cortex encodes dimensions of consciousness. Nat. Commun. 14, 72 (2023).
Article ADS CAS PubMed PubMed Central Google Scholar
Essen, D. C. V. et al. The WU-Minn human connectome project: an overview. NeuroImage 80, 62–79 (2013).
Article PubMed Google Scholar
Poldrack, R. A. et al. A phenome-wide examination of neural and cognitive function. Sci. Data 3, 160110 (2016).
Article CAS PubMed PubMed Central Google Scholar
Tian, Y., Margulies, D. S., Breakspear, M. & Zalesky, A. Topographic organization of the human subcortex unveiled with functional connectivity gradients. Nat. Neurosci. 23, 1421–1432 (2020).
Article CAS PubMed Google Scholar
Huang, Z. et al. Anterior insula regulates brain network transitions that gate conscious access. Cell Rep. 35, 109081 (2021).
Article CAS PubMed PubMed Central Google Scholar
Azen, R. & Budescu, D. V. The dominance analysis approach for comparing predictors in multiple regression. Psychol. Methods 8, 129–148 (2003).
Article PubMed Google Scholar
Hansen, J. Y. et al. Mapping neurotransmitter systems to the structural and functional organization of the human neocortex. Nat. Neurosci. 25, 1569–1581 (2022).
Article CAS PubMed PubMed Central Google Scholar
Huang, Z. et al. Asymmetric neural dynamics characterize loss and recovery of consciousness. Neuroimage 236, 118042 (2021).
Article PubMed Google Scholar
Huang, Z., Liu, X., Mashour, G. A. & Hudetz, A. G. Timescales of intrinsic BOLD signal dynamics and functional connectivity in pharmacologic and neuropathologic states of unconsciousness. J. Neurosci. 38, 2304–2317 (2018).
Article CAS PubMed PubMed Central Google Scholar
Tanabe, S. et al. Altered global brain signal during physiologic, pharmacologic, and pathologic states of unconsciousness in humans and rats. Anesthesiology 132, 1392–1406 (2020).
Article CAS PubMed Google Scholar
Schaefer, A. et al. Local-global parcellation of the human cerebral cortex from intrinsic functional connectivity MRI. Cereb. Cortex 28, 3095–3114 (2017).
Article PubMed Central Google Scholar
Mashour, G. A. & Alkire, M. T. Consciousness, anesthesia, and the thalamocortical system. Anesthesiology 118, 13–15 (2013).
Article PubMed Google Scholar
Shepherd, G. M. G. & Yamawaki, N. Untangling the cortico-thalamo-cortical loop: cellular pieces of a knotty circuit puzzle. Nat. Rev. Neurosci. 22, 389–406 (2021).
Article CAS PubMed PubMed Central Google Scholar
Münkle, M. C., Waldvogel, H. J. & Faull, R. L. M. Calcium-binding protein immunoreactivity delineates the intralaminar nuclei of the thalamus in the human brain. Neuroscience 90, 485–491 (1999).
Article PubMed Google Scholar
Jones, E. G. & Hendry, S. H. C. Differential calcium binding protein immunoreactivity distinguishes classes of relay neurons in monkey thalamic nuclei. Eur. J. Neurosci. 1, 222–246 (1989).
Article PubMed Google Scholar
Takahashi, N., Oertner, T. G., Hegemann, P. & Larkum, M. E. Active cortical dendrites modulate perception. Science 354, 1587–1590 (2016).
Article ADS CAS PubMed Google Scholar
Kuramoto, E. et al. Two types of thalamocortical projections from the motor thalamic nuclei of the rat: a single neuron-tracing study using viral vectors. Cereb. Cortex 19, 2065–2077 (2009).
Article PubMed Google Scholar
Purdon, P. L. et al. Electroencephalogram signatures of loss and recovery of consciousness from propofol. Proc. Natl Acad. Sci. 110, E1142–E1151 (2013).
Article CAS PubMed PubMed Central Google Scholar
Ching, S., Cimenser, A., Purdon, P. L., Brown, E. N. & Kopell, N. J. Thalamocortical model for a propofol-induced α-rhythm associated with loss of consciousness. Proc. Natl Acad. Sci. 107, 22665–22670 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Weiner, V. S. et al. Propofol disrupts alpha dynamics in functionally distinct thalamocortical networks during loss of consciousness. Proc. Natl Acad. Sci. 120, e2207831120 (2023).
Article CAS PubMed PubMed Central Google Scholar
Hudetz, A. G. & Mashour, G. A. Disconnecting consciousness. Anesth. Analg. 123, 1228–1240 (2016).
Article PubMed PubMed Central Google Scholar
Sanders, R. D., Tononi, G., Laureys, S., Sleigh, J. W. & Warner, D. S. Unresponsiveness ≠ unconsciousness. Anesthesiology 116, 946–959 (2012).
Article PubMed Google Scholar
Sherman, S. M. & Usrey, W. M. A reconsideration of the core and matrix classification of thalamocortical projections. J. Neurosci. 44, e0163242024 (2024).
Article PubMed Google Scholar
Fiset, P. et al. Brain mechanisms of propofol-induced loss of consciousness in humans: a positron emission tomographic study. J. Neurosci. 19, 5506–5513 (1999).
Article CAS PubMed PubMed Central Google Scholar
Mashour, G. A. et al. Recovery of consciousness and cognition after general anesthesia in humans. eLife 10, e59525 (2021).
Article CAS PubMed PubMed Central Google Scholar
Huang, Z., Zhang, J., Wu, J., Mashour, G. A. & Hudetz, A. G. Temporal circuit of macroscale dynamic brain activity supports human consciousness. Sci. Adv. 6, eaaz0087 (2020).
Article ADS PubMed PubMed Central Google Scholar
Murphy, K. & Fox, M. D. Towards a consensus regarding global signal regression for resting state functional connectivity MRI. Neuroimage 154, 169–173 (2017).
Article PubMed Google Scholar
Yeo, B. T. T. et al. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J. Neurophysiol. 106, 1125–1165 (2011).
Article PubMed Google Scholar
Wael et al. BrainSpace: a toolbox for the analysis of macroscale gradients in neuroimaging and connectomics datasets. Commun. Biol. 3, 103 (2020).
Article Google Scholar
Alexander-Bloch, A. F. et al. On testing for spatial correspondence between maps of human brain structure and function. NeuroImage 178, 540–551 (2018).
Article PubMed Google Scholar
Jones, E. G. Thalamic circuitry and thalamocortical synchrony. Philos. Trans. R. Soc. Lond. Ser. B 357, 1659–1673 (2002).
Article Google Scholar

Download references

Acknowledgements

This work was supported by National Institute of General Medical Sciences of the National Institutes of Health (Bethesda, MD) grant R01GM103894 (Co-PIs: A.G.H., and Z.H.). The content is solely the responsibility of the author and does not necessarily represent the official views of the National Institutes of Health. The authors thank Dr. James M. Shine who shared the spatial maps of CALB1 and PVALB.

Author information

Authors and Affiliations

Department of Anesthesiology, University of Michigan Medical School, Ann Arbor, MI, USA
Zirui Huang, George A. Mashour & Anthony G. Hudetz
Center for Consciousness Science, University of Michigan Medical School, Ann Arbor, MI, USA
Zirui Huang, George A. Mashour & Anthony G. Hudetz
Michigan Psychedelic Center, University of Michigan Medical School, Ann Arbor, MI, USA
Zirui Huang, George A. Mashour & Anthony G. Hudetz
Neuroscience Graduate Program, University of Michigan, Ann Arbor, MI, USA
Zirui Huang, George A. Mashour & Anthony G. Hudetz
Department of Pharmacology, University of Michigan Medical School, Ann Arbor, MI, USA
George A. Mashour

Authors

Zirui Huang
View author publications
You can also search for this author in PubMed Google Scholar
George A. Mashour
View author publications
You can also search for this author in PubMed Google Scholar
Anthony G. Hudetz
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

Z.H. conducted the experiment, collected, and analyzed the data, and prepared the figures. Z.H. and A.G.H. designed the study. Z.H., G.A.M., and A.G.H. interpreted the data and wrote the manuscript.

Corresponding author

Correspondence to Zirui Huang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Boris Bernhardt and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Description of Additional Supplementary Files

Supplementary Data 1

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Cite this article

Huang, Z., Mashour, G.A. & Hudetz, A.G. Propofol disrupts the functional core-matrix architecture of the thalamus in humans. Nat Commun 15, 7496 (2024). https://doi.org/10.1038/s41467-024-51837-1

Download citation

Received: 23 January 2024
Accepted: 15 August 2024
Published: 09 September 2024
DOI: https://doi.org/10.1038/s41467-024-51837-1

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.