Cortical connectivity, local dynamics and stability correlates of global conscious states – Communications Biology

What changes in the brain separate wakeful responsiveness from anesthesia-induced unresponsiveness? A new study leverages magnetoencephalography (MEG) and biophysically grounded whole-cortex models to link shifts in local synaptic dynamics, large-scale connectivity, and dynamical stability to global conscious state under xenon sedation. Fifteen healthy participants underwent graded xenon anesthesia up to 1.3 MAC-awake while performing an auditory continuous performance task (ACPT), providing a time-resolved behavioral proxy for responsiveness—and, by assumption, overall consciousness.

How the brain was modeled and measured

Researchers reconstructed source-level MEG signals for 78 cortical regions and fit individual whole-cortex models in short overlapping windows. Each region was represented by a canonical neural mass model (NMM) comprising three interacting populations: excitatory pyramidal cells, excitatory spiny stellate cells, and inhibitory interneurons. Key parameters included excitatory and inhibitory synaptic gains (α terms), membrane potentials (v), and an external cortical drive (μ) summarizing inputs from the rest of the brain. Inter-regional coupling was captured by a connectivity matrix W. An analytic Kalman filter provided time-resolved state and parameter estimates, tracking MEG dynamics with low error while denoising high-frequency activity.

Behavioral anchor: the ACPT

Participants heard 1 or 3 kHz tones and pressed the corresponding button. Accuracy and reaction speed indexed vigilance. Because the task is easy, performance typically stayed at ceiling until sedation altered attention or responsiveness, yielding clear transitions between high and low responsiveness. Although responsiveness and consciousness are not identical, the study treated ACPT accuracy as a monotonic surrogate of global conscious level for analysis.

Connectivity signatures of reduced consciousness

Two consistent patterns emerged during low-responsiveness periods:

• Local weakening: Excitatory synaptic strengths from pyramidal cells to both inhibitory and excitatory targets declined, as did related membrane potentials, indicating a selective reduction in excitatory drive.
• Global uncoupling: Long-range inter-regional coupling dropped and rebounded with recovery, pointing to a transient collapse of large-scale integration.

Network centrality analyses identified hubs whose influence tracked behavior. Centrality in occipital, posterior parietal, and prefrontal cortices decreased with reduced consciousness. Connection-wise analyses highlighted links within and between parietal and occipital regions: stronger parietal→occipital and intra-occipital couplings were associated with higher responsiveness, underscoring a dominant parietal role in modulating conscious state.

Functional networks: DMN and visual systems take center stage

Examining seven canonical networks (including default mode variants, executive control, sensorimotor, auditory, and visual), intra-network connectivity in the visual network (VN) and default mode network (DMN; particularly its ventral and non-anterior components) positively tracked consciousness. Inter-network connections were similarly telling: robust bidirectional VN↔DMN/DMNv links, VN→sensorimotor and VN→auditory, and executive→visual inputs all aligned with higher responsiveness.

To adjudicate the longstanding “front vs. back” debate, the team contrasted posterior parietal and prefrontal networks. Two measures were significant: posterior parietal intra-network strength and posterior parietal→prefrontal coupling, both positively correlated with consciousness. Neither prefrontal intra-network coupling nor prefrontal→posterior parietal projections were significant. A complementary functional metric—weighted symbolic mutual information (wSMI)—converged on the same conclusion: higher functional integration within posterior parietal, between posterior parietal and prefrontal, and across the whole brain tracked higher ACPT accuracy.

Local model parameters that move with responsiveness

“Correlation imaging” of regional NMM parameters revealed that stronger excitatory synaptic gains (pyramidal→excitatory and excitatory→pyramidal) in posterior parietal and occipital cortices were linked to higher responsiveness. External cortical drive (μ) to these regions also increased with consciousness, consistent with their role as key integration hubs. A complementary before–after contrast around xenon equilibration emphasized prefrontal reductions: excitatory gains dropped in frontal and cingulate areas, with broader decreases in external input across frontal, temporal, and posterior cortex—suggesting overlapping but not identical substrates for drug effect and conscious-level modulation. Pyramidal membrane potentials declined most in posterior parietal and occipital regions during sedation.

Stability as a global marker

Within a dynamical-systems framing, the authors computed the Jacobian of the whole-cortex model over time and counted unstable eigenmodes to index cortical stability. Across participants, the cortex stayed near an unstable regime, but the number of unstable modes characteristically fell during loss of responsiveness and rose during recovery—mirroring behavioral state transitions and revealing intermediate, fluctuating phases rather than abrupt switches. Crucially, this stability index tracked consciousness more closely than xenon dose and often lagged dosing changes, reflecting pharmacokinetic/pharmacodynamic delays. Marked individual variability emerged: some subjects showed prolonged unresponsive intervals; others exhibited briefer or incomplete losses.

Why it matters

This study triangulates a coherent picture of how consciousness waxes and wanes under anesthesia:

• Local circuit shifts: selective weakening of excitatory synapses and reduced pyramidal activity.
• Network-level reconfiguration: diminished long-range coupling and fading hub centrality, especially across posterior parietal–occipital–prefrontal circuits.
• Posterior-led integration: posterior parietal intra-network strength and its feedforward projections to prefrontal cortex stand out across both structural-effective coupling and functional integration (wSMI).
• Dynamical regime changes: a model-based stability index follows responsiveness more faithfully than drug concentration, offering a candidate global neural correlate of consciousness.

By uniting source-resolved MEG, mechanistic modeling, and dynamical systems metrics, the work supplies data-constrained biomarkers that cut across scales—from synaptic gains to network integration and stability. The approach could inform patient-specific monitoring in anesthesia, refine assessments in disorders of consciousness, and guide interventions that restore large-scale integration when it falters.