RHO GTPase by L1 GABAergic neurons in frontal cortex and L6 glutamatergic neurons in prefrontal cortex differentiates states of unconsciousness

Abstract

Unconsciousness can be induced by sleep and general anaesthesia alike; however, the extent to which these states differ is not well understood. In this study, we aimed to determine the similarities and differences between the unconscious states caused by sleep and those caused by anaesthesia, as well as investigate gene expression profiles by applying tensor decomposition-based unsupervised feature extraction to the two gene expression profiles. One of the two expression profiles was obtained from mice treated with sevoflurane, a type of inhaled anaesthesia, whereas the other two expression profiles were obtained from sleeping and awake mice. We selected two sets of genes (507 and 1048 genes) that were distinctly expressed between sleep- or anaesthesia-induced unconsciousness and the awake state. Both sets of genes include various distinct genes which are common to the state of unconsciousness during sleep and anaesthesia. The former was enriched in GABAergic synapses and the first layer of the frontal cortex, whereas the latter was enriched in glutamatergic synapses and the sixth layer of the prefrontal cortex. Additionally, both sets of genes were enriched in the RHO GTPase pathway. Based on these results, we hypothesised that L1 GABAergic neurons in the frontal cortex and L6 glutamatergic neurons in the prefrontal cortex use RHO GTPase to differentiate between the states of unconsciousness induced by general anaesthesia and sleep.

Keywords: AI consciousness studies; Anaesthesia; Feature selection; Sleep; Tensor decomposition.

Fig. 1

Schematic representation of the analyses performed in this study. First, two gene expression profiles were multiplied to generate a tensor. Tensor decomposition was applied to the generated tensor and singular value vectors were evaluated. After the selection of the singular value vectors attributed to various experimental conditions (e.g., tissue specificity and sleep vs. awake), the singular value vectors attributed to genes and used to select genes were selected. After optimizing the standard deviation, P-values were attributed to genes, and genes associated with adjusted P-values less than the threshold value () were selected. The selected genes were evaluated using large language model (LLM) and enrichment analysis, based on which we generated the hypothesis.

Fig. 2

Singular value vectors and core tensor G. Top left: , top center: , top right: , bottom left and right: , bottom center: , bottom right: .

Fig. 3

Left: standard deviation of histogram of as a function of for . The vertical red arrow suggests the selected value of . Right: histogram of for the optimal .

Fig. 4

Left: standard deviation of histogram of as a function of for . The vertical red arrow suggests the selected value of . Right: histogram of for the optimal .

Fig. 5

Venn diagram betwenn the 507 genes selected by  ℓ1=3 and the 1,048 genes selected by   ℓ1=4.

Fig. 6

Connectivity graph between biological terms detected by Enrichr (only those with adjusted P-values less than 0.05). Green: (GABAergic), yellow: (glutamatergic). Edge width represents the number of genes shared between adjusted biological terms.