Mapping the neural systems driving breathing at the transition to unconsciousness

After falling asleep, the brain needs to detach from waking activity and reorganize into a functionally distinct state. A functional MRI (fMRI) study has recently revealed that the transition to unconsciousness induced by propofol involves a global decline of brain activity followed by a transient reduction in cortico-subcortical coupling. We have analyzed the relationships between transitional brain activity and breathing changes as one example of a vital function that needs the brain to readapt. Thirty healthy participants were originally examined. The analysis involved the correlation between breathing and fMRI signal upon loss of consciousness. We proposed that a decrease in ventilation would be coupled to the initial decline in fMRI signal in brain areas relevant for modulating breathing in the awake state, and that the subsequent recovery would be coupled to fMRI signal in structures relevant for controlling breathing during the unconscious state. Results showed that a slight reduction in breathing from wakefulness to unconsciousness was distinctively associated with decreased activity in brain systems underlying different aspects of consciousness including the prefrontal cortex, the default mode network and somatosensory areas. Breathing recovery was distinctively coupled to activity in deep brain structures controlling basic behaviors such as the hypothalamus and amygdala. Activity in the brainstem, cerebellum and hippocampus was associated with breathing variations in both states. Therefore, our brain maps illustrate potential drives to breathe, unique to wakefulness, in the form of brain systems underlying cognitive awareness, self-awareness and sensory awareness, and to unconsciousness involving structures controlling instinctive and homeostatic behaviors.


Introduction
During the transition between wakefulness and sleep, the brain needs to adapt functionally to the new context so as to optimally regulate bodily functions. Loss of consciousness may naturally occur in the form of sleep or may be induced by hypnotic agents. Although falling asleep is apparently abrupt in both instances, the phenomenon is complex and requires a large-scale reorganization of brain activity ( Ogilvie, 2001 ;Merica and Fortune, 2004 ;Tanabe et al., 2020 ;Pujol et al., 2021 ). We have recently characterized the transition to unconsciousness induced by a hypnotic agent with a temporal analysis of functional MRI (fMRI) complex while awake, when a greater number of factors may influence respiratory rhythm and amplitude ( Horner, 2009 ;Masaoka et al., 2014 ;Trinder et al., 2014 ). For example, a major source of tonic drive to respiratory motoneurons originates from the brainstem ascending arousal system that promotes wakefulness ( Horner, 2009 ). The breathing pattern, however, is also sensitive to more specific variations in neural activity associated with both the automatic expression of emotions (e.g., laughing) and the volitional control of the motor cortex (e.g., speaking) ( Masaoka et al., 2014 ;Evans, 2010 ). Breathing, therefore, is a common element of the varied behavioral repertory in the conscious state.
Such a waking stimulus to breathe is lost during the transition to unconsciousness, which characteristically presents a transitory reduction in pulmonary ventilation until the unconscious state is reorganized and breathing control fully recovered ( Naifeh and Kamiya, 1981 ;Ogilvie et al., 1989 ;Worsnop et al., 1998 ;Edwards and White, 2011 ;Trinder et al., 2014 ). Reduction of the excitatory inputs combines with the increase in inhibitory -amino butyric acid (GABA) release, and both predispose susceptible individuals to hypoventilation and obstructive sleep apnea ( Horner, 2009 ;Edwards and White, 2011 ). The inhibitory action and the risk of apnea are larger by the effect of hypnotic agents ( Smith and Lee-Chiong, 2008 ).
In the present study, we have analyzed the coupling of breathing to brain activity following a loss of consciousness gently induced by the hypnotic agent propofol. Our aim was to illustrate how the brain reorganizes itself to control a basic behavior when moving between states of consciousness and to further characterize the neural drives to breathing. Our hypothesis was that the transient decrease in ventilation after losing consciousness would coincide with the initial fMRI signal decline observed in the synchronized cerebral cortex prior to the corticosubcortical dissociation phenomenon. At the system level, the transit from stable to transitorily reduced pulmonary ventilation would be coupled to the fMRI signal decline in brain areas relevant for modulating breathing in the awake state, and the subsequent ventilation recovery would be coupled to an increased fMRI signal in structures involved in breathing control during the unconscious state.

Study participants
Thirty healthy right-handed participants were initially selected for the project. The current study was based on 19 subjects showing optimal fMRI and ventilation data. The group included 10 males and 9 females with a mean age of 27.9 years (SD: 6.3 years). The characteristics of the sample and participant selection were fully described in our previous report . Participants were required to meet the criterion of the American Society of Anesthesiologists (ASA) physical status ASA I (healthy, non-smoking, no or minimal alcohol) ( Mayhew et al., 2019 ) and to show normal physical and laboratory exams for eligibility.
The study was conducted in accordance with the guidelines of the Declaration of Helsinki. The protocol was approved by the Spanish Agency of Medicines and Medical Devices, AEMPS (reference 5NFNF6V55C) and by the Ethical Committee of Clinical Research of the Parc de Salut Mar of Barcelona (reference n°2017-7165). Written informed consent was obtained from all participants.

Overview of the experimental design
Functional MRI was continuously acquired for a total of 15 min. In the initial 3-minute period no drug was administered. Subsequently, a continuous intravenous infusion of propofol commenced, which was steeply increased every 2 min until the participant lost consciousness. The dose of propofol was then adjusted (at the effect-site) and maintained constant with the aim of keeping a superficial level of unconsciousness during the rest of the fMRI acquisition with no return to con-sciousness at this stage of the study (participants underwent a second study phase testing other hypotheses).

Induced loss of consciousness with propofol
Our aim was to progressively sedate the participant until consciousness was lost within the fMRI acquisition window using a slow propofol administration regimen. At the attained sedation level, breathing is spontaneous with no need for tracheal intubation and the individual can be awakened with moderately intense stimulation. A behavioral reference indicating "loss of conscious control ", as a surrogate of conventional loss of consciousness, was obtained within the constraints of the MRI environment by recording the moment the participants ceased to squeeze a soft pneumatic balloon with their right hand. Participants were instructed to repeatedly squeeze the balloon at a practiced rate of approximately one movement every 3 s (0.3 Hz) throughout the MRI while they were awake. The actual squeezing rate recorded prior to loss of conscious control showed a mean of 0.30 Hz (SD, 0.08 Hz) .
Loss of conscious control was thus marked by the participants ceasing to move their right hand and served to direct the temporal analyses. We previously identified  that the transitional fMRI signal events (i.e., initial signal fall followed by a transient reduction in corticosubcortical coupling) began just after the participants stopped moving their hand in 9 cases (specifically after a mean of 2.9 s and SD of 3.9 s), and after a temporal lapse (mean, 61.0 s; SD, 25.1 s) in the remaining 10 participants. In no case were the fMRI signal changes evident before the participants ceased to move their hand. In the current evaluation, we analyzed the changes in pulmonary ventilation during the identified fMRI transitional events (see below).

Anesthetic technique
A target-controlled system (Base Primea Orchestra®, Fresenius Kabi, Brézins, France) was used for a continuous intravenous infusion of propofol individually programmed at a target plasma concentration of 3.5 mcg/mL based on the Schnider model ( Schnider et al., 1999 ). Propofol target plasma concentration was steeply increased by 0.5 mcg/mL every 2 min until participants ceased to move their hand for at least 20 s. At this moment (20 s after loss of conscious control), propofol targetcontrolled infusion was targeted at the effect-site (brain) concentration currently estimated by the model. Pulse oximetry (SpO2) and expired capnography (Oral-Trac®, Salter Labs, Arvin CA) were used to continuously monitor respiratory function throughout the experiment for safety purposes. Heart rate, pulse oximetry and carbon dioxide measures were recorded every 10 s. Oxygen (2 L/min) was administered throughout the study .

Continuous breathing recording
Respiration was recorded using a pneumatic sensor (Philips 3T Respiratory Sensor) softly held between the thorax and abdomen, and data logger developed in-house using Labview 8.0 software (National instruments corp. Austin, TX). The measures were registered in a text file with a 25 msec sampling period. As described by Birn et al. (2008) , the amount of air inspired with each breath was estimated by computing the difference between the maximum and minimum sensor positions at the peaks of inspiration and expiration, respectively. This difference was divided by the respiration period (i.e., the time between the peaks of the respiration waveform) to consider changes in both the rate and depth of breathing. Such a respiration volume per time (RVT) quotient allowed us to estimate relative variations in pulmonary ventilation over time ( Fig. 1 ). RVT was expressed as% change from a baseline period (i.e., 30 s preceding the transition to unconsciousness). RVT measures were downsampled by a factor of 80 (i.e., by averaging every 80 samples from the RVT time series) to match the sampling rate of the fMRI data (TR = 2 s), so they could be used as a regressor in the fMRI time series analysis.  . The green square and vertical bar indicate the group average onset of global decline in synchronized brain activity observed after loss of consciousness. The inferior plot for each subject shows the respiration volume per time (RVT) quotient estimated for each image acquisition volume (TR), expressed as% change from a baseline period (i.e., 30 s preceding the transition).

MRI acquisition
A Philips Achieva 3.0 Tesla magnet (Philips Healthcare, Best, The Netherlands), equipped with an eight-channel phased-array head coil and single-shot echoplanar imaging (EPI) software, was used for the MRI assessment. The functional blood oxygen level-dependent (BOLD) sequence consisted of gradient recalled acquisition in the steady state (time of repetition [TR], 2000 ms; time of echo [TE], 35 ms; pulse angle, 70°) within a field of view of 240 × 240 mm, with a 64 × 64-pixel matrix, and a slice thickness of 4 mm (inter-slice gap, 0 mm) and acquisition voxel size of 3.75 × 3.75 × 4 mm. A total of 34 interleaved slices were acquired to cover the whole-brain. Each functional time series consisted of 450 consecutive image sets or volumes obtained over 15 min.

MRI preprocessing
Imaging data were processed using MATLAB version 2016a (The MathWorks Inc, Natick, Mass) and Statistical Parametric Mapping software (SPM12; The Wellcome Department of Imaging Neuroscience, London). Preprocessing involved motion correction, spatial normalization and smoothing by means of a Gaussian filter (full-width half-maximum, 8 mm). Data were normalized to the standard SPM-EPI template and resliced to 3 mm isotropic resolution in Montreal Neurological Institute (MNI) space. Low-frequency fMRI signal drifts were removed using a temporal high-pass filter with a cutoff of 128 s. No low-pass filter was applied.
All image sequences were inspected for potential acquisition and normalization artifacts. At this stage, 9 participants were removed from the initial 30-subject sample as a result of large head displacements preventing adequate image preprocessing or motion artifacts affecting the analyzed image volumes. Two additional participants were excluded due to technical issues in ventilation data registration. Therefore, 19 participants were valid for the analysis after careful selection.

Functional MRI analysis
To analyze the relationships between breathing variations and fMRI signal evolution during the transition to unconsciousness, a total of 80 fMRI volumes (2 min and 40 s) were used in each case that included the entire 100 -s period encompassing the transitional fMRI signal events described in our early report and 60 s preceding the phenomenon. This transitional period was selected on an individual basis using a common point identified in the evolution of the cerebral cortex fMRI signal. Specifically, the common point for valid cases was the point at which the synchronized fMRI signal in the cortex passed from negative to positive values . This benchmark, therefore, served to anchor signal oscillations and average out signal time courses across participants for an optimal group analysis.
SPM maps were generated using individual breathing (RVT) time courses as regressors in the first-level design matrices that also included Group average breathing measures. The superior plot shows the respiration volume per time (RVT) quotient estimated for each image acquisition volume (TR). The middle plot shows the selected "breathing-fall " 100 -s period. The inferior plot shows the selected "breathing-recovery " 100 -s period.
six motion estimates as co-variables and removal of volumes with interframe motion > 0.3 mm ( Power et al., 2014 ). A delay of 4 s was applied to the regressor to adjust for the hemodynamic response latency ( Poldrack et al. 2011 ;Khosla et al. 2021 ). No global MRI signal measures were used as nuisance regressors. This voxel-wise analysis, therefore, estimates the extent to which breathing variations during the transition to unconsciousness are related to variations in regional activity.
Two different fMRI models were estimated. The first model (breathing-fall map) included 30 image volumes (1 min) preceding transition and the subsequent 20 vol (40 s) coinciding with the observed transitory reduction in ventilation (see below in Results). The second model (breathing-recovery map) also included this 20-volume period (40 s) and the subsequent 30 vol (1 min) of breathing recovery ( Fig. 2 ). Therefore, these fMRI models allowed us to differentially identify brain regions coupled to breathing from wakefulness to loss of consciousness and regions coupled to breathing recovery during the unconscious state.
Resulting 1st-level contrast images were carried forward to 2ndlevel analyses in SPM. One-sample t -test maps were estimated for both breathing-fall and breathing-recovery models. Two-sample paired t -test maps were obtained comparing the models. A motion summary measure (mean inter-frame motion ( Pujol et al., 2014 )) for each participant was included as a covariate to further control potential motion effects.
Breathing is often considered a major confounding factor in fMRI, particularly when assessing functional connectivity, and diverse procedures may be applied to remove the artifactual effects ( Murphy et al., 2013 ;Power et al., 2017 ). However, care should be taken to avoid removing the neural signal one is trying to detect. In our analysis of slow ventilation changes, we considered it relevant to assess the potential influence on the results of the delayed vascular effects of breathing variations on the fMRI signal ( Chang and Glover, 2009 ). Therefore, the results were replicated after additionally including the individual RVT time courses with a long delay in the 1st-level models. A lapse of + 18 s was used as the typical time required to achieve the full vascular effects ( Chang and Glover, 2009 ;Birn et al., 2008 ).
Results were considered significant when clusters formed at a threshold of p < 0.005 survived whole-brain family-wise error (FWE) correction ( p < 0.05), calculated using SPM. Fig. 1 shows representative single-subject plots illustrating changes in ventilation associated with propofol-induced loss of consciousness. A transient reduction in ventilation (measured as respiration volume per time, RVT) was identified, with an onset approximately coinciding with the onset of fMRI signal transitional changes (i.e., the initial fMRI signal fall). This transient decrease in ventilation was notably consistent across subjects ( Fig. 2 ). As a group, RVT measures showed values below baseline for 40 s (i.e., 20 image volumes). Interestingly, a degree of breathing synchronization across subjects (i.e., RVT oscillations detectable at the group level) was also observed prior to reduced ventilation and following recovery ( Fig. 2 ).

Ventilation changes at loss of consciousness
Such consistent across-subject variations in ventilation were, however, discreet in terms of their effect on oxygenation and carbon dioxide measures. Oxygen saturation during the transition period showed no Fig. 3. Breathing-coupled brain activity. SPM maps were generated using individual breathing (RVT) time courses as regressors. The "breathing-fall map " (F, superior images) shows cortical and subcortical structures coupled to breathing from wakefulness to loss of consciousness. The "breathing recovery map " (R, middle images) shows deep brain structures coupled to breathing recovery during the unconscious state. The inferior images show the map of the differences (Dif.) between the fall in breathing and breathing recovery, which included cortical areas and the thalamus. The left hemisphere is displayed on the left side of the coronal and axial views.
net change with a ceiling effect consistently at values always of 99% or over in both baseline and transition periods. Note that the participants received a continuous administration of inhaled oxygen at 2 L/min during the experiment. The change in end-tidal carbon dioxide was also not significant, showing a mean ± SD of 34.1 ± 2.4 mmHg prior to the transitional phenomena and 33.9 ± 3.5 mmHg during transition ( t = 0.3; p = 0.773).

Breathing-coupled brain activity at the transition to unconsciousness
Brain elements with fMRI signal significantly coupled to breathing from wakefulness to loss of consciousness included discrete areas in the prefrontal cortex, angular/supramarginal gyri, inferior temporal cortex, cingulate cortex, hippocampus, thalamus, brainstem tegmentum and cerebellum. Structures with fMRI signal coupled to breathing recovery were restricted to the hippocampus, brainstem tegmentum, cerebellum, amygdala and the hypothalamus/subthalamus region ( Fig. 3 and Table 1 ). Differences between both breathing-fall and breathingrecovery maps were observed in the prefrontal cortex, angular gyrus, precuneus, cingulate cortex, supramarginal/postcentral gyri, parietal operculum and thalamus ( Fig. 3 and Table 2 ).
To control for potential vascular effects of breathing variations on fMRI signal, the results of the differences between breathing-fall and breathing-recovery maps were replicated after adjusting the first-level (single-subject) correlation maps by the RVT time courses shifted + 18 s. The analysis with and without the adjustment showed similar results ( Table 3 ).
To further illustrate the coupling of the fMRI signal to breathing, RVT measures were plotted against fMRI signal in representative brain structures ( Fig. 4 ). The angular gyrus showed the best coupling to RVT before and during the breathing fall. In contrast, the hypothalamus region was best coupled to RVT during breathing recovery and the brainstem showed some coupling in both phases.
As mentioned, fMRI signal in the hippocampus was significantly associated with breathing variations in both phases. Nevertheless, the affected hippocampus portion was not identical. In the breathing-fall map, the implicated area extended from the subiculum to the region of the dentate gyrus and CA4, whilst in the breathing recovery map, it extended from the subiculum to the parahippocampal gyrus ( Fig. 5 ). The differences between both maps were not significant. However, at a low threshold (uncorrected p < 0.05), the association between fMRI signal and breathing measures was stronger during unconsciousness in the anterior extent of the subiculum and CA1.

Discussion
We examined the relationships between breathing and fMRI signal changes associated with propofol-induced loss of consciousness. The analysis confirmed the study's hypothesis that a subtle, albeit consistent, reduction in ventilation coincided with the beginning of the fMRI transitional phenomenon, specifically with a general fMRI signal decline in the cerebral cortex. Such a respiratory depression was transitory and recovered during the unconscious state after 40 s in average. The mapped correlation between breathing and fMRI signal variations from wakefulness to loss of consciousness captured a combination of cortical and subcortical structures. In contrast, breathing recovery was coupled to fMRI signal in deep structures only.
We proposed that a transitory decoupling in cortico-subcortical synchrony would allow the brain to detach from waking activity and reorganize during the unconscious state and used breathing as a basic behavior to illustrate such a transition to the new order. The temporal coincidence between the beginning of the ventilation reduction and the initial fMRI signal decline observed in the synchronized cerebral cortex supports that the identified imaging changes captured the transition to unconsciousness gently induced by the hypnotic agent. Interestingly, the relatively long duration of both behavioral and imaging events further indicates that the apparently abrupt loss of consciousness may be only the initial step of a complex transitional phenomenon .
The current analysis showed a notably distinctive set of brain structures coupled to decreased and recovered breathing levels. Only the brainstem tegmentum, cerebellum and hippocampus were identified in both maps. The most basic breathing control takes place in the brainstem and deep brain structures ( Smith and Lee-Chiong, 2008 ; Table 1 Breathing-coupled brain activity at loss of consciousness. All findings correspond to positive associations (i.e., higher RVT with higher fMRI signal). P FWE-corr , P (Family-Wise Error corrected). x y z, coordinates given in Montreal Neurological Institute space.  x y z, coordinates given in Montreal Neurological Institute space. Fig. 4. Group average breathing measures plotted against fMRI signal in representative brain structures. Benarroch, 2019 ;Bosch et al., 2017 ). Therefore, it is likely that breathing variations are coupled to fMRI signal in deep brain systems in both conscious and unconscious states. The breathing-fall map additionally included the prefrontal cortex, inferior temporal cortex, angular/supramarginal gyri, cingulate cortex and thalamus. Moreover, the postcentral gyrus and parietal operculum were identified in the difference map. All these elements may therefore potentially contribute to account for the so-called wakefulness stimulus to breathing. The transient decrease in breathing at the onset of natural or induced sleep is supposed to be, to a notable extent, the result of a sudden loss of the wakefulness drive, defined as brain activity that differentially stimulates breathing while the subject is awake ( Horner, 2009 ;Trinder et al., 2014 ;Orem, 1990 ). Ac-tivity in these structures may express different aspects of consciousness, and they have all been involved in either the voluntary control or conscious perception of breathing ( Evans, 2010 ;Davenport and Vovk, 2009 ). The identified areas in the prefrontal cortex may play an important role in cognitive awareness ( McCaig et al., 2011 ;Brown et al., 2019 ). The angular gyrus and the cingulate cortex are key elements of the default mode network, which has been consistently involved in "self-awareness " ( Moll et al., 2007 ;Leech and Sharp, 2014 ;Pujol et al., 2019 ). As part of the sensory system, the thalamus, postcentral gyrus, parietal operculum and supramarginal gyrus each contribute to somatosensory awareness ( de Haan and Dijkerman, 2020 ;Keller at al., 2020 ). Therefore, the breathing-fall map included elements of brain systems responsible for relevant components of conscious activity with Our results also showed brain structures unique to the breathing recovery map such as the amygdala, hypothalamus and subthalamus. Although we found no significant differences when comparing both maps at this level, the finding is of interest insofar as it suggests a prominent participation of primary systems controlling instinctive and homeostatic behaviors in the modulation of breathing during unconsciousness. Activity in the amygdala and hypothalamus is typically synchronized with a range of autonomic responses including cardio-respiratory coupling, sweating and blood pressure ( Masaoka et al., 2014 ;Fukushi et al., 2019 ). Notable variations in autonomic activity exist during sleep ( Penzel et al., 2016 ;Fukushi et al., 2019 ). Interestingly, the hypothalamus, which has a distinctive role in the general regulation of the wake-sleep cycle ( Saper et al., 2010 ), showed a distinctive fMRI signal dynamics at loss of consciousness  and the closest coupling to breathing recovery in the current analysis ( Fig. 4 ).
As mentioned, the hippocampus was identified in both maps. However, the hippocampal sectors associated with breathing fall and breathing recovery did not completely overlap in our analysis. This interesting observation further emphasizes that the hippocampal neural processes at work during wakefulness and unconsciousness are not identical. Indeed, although vivid experiences with emotional and physiological components may occur both in wakefulness and during natural ( Dresler et al., 2014 ) or propofol-induced unconsciousness ( Sanders et al., 2012 ;Radek et al., 2018 ) in the form of dreams, waking experiences are recorded in memory and dreams are not.
Previous imaging studies have mostly focused on identifying neural systems related to several aspects of conscious breathing based on strategies differing from the one used here, including hypernea, respiratory load, breath-hold, restricted tidal volume and hypercapnia. Although the previous work does not directly compare with our study, it jointly indicates that the major functional brain systems processing sensorimotor, limbic and cognitive activity may be linked to breathing during waking activity (e.g., Davenport and Vovk, 2009 ;Evans, 2010 ;Macey et al., 2016 ;Chan et al., 2018 ). Of special interest may be the studies illustrating how the neural correlates of spontaneous breathing differ during cognitive challenging compared with rest and show a degree of task specificity as to the brain areas implicated ( Evans et al., 2009 ;Zhang et al., 2020 ).
A significant limitation of our study lies in its correlational nature, which allows for alternative interpretations. That is, although the identified brain systems may plausibly contribute to driving breathing, as suggested herein, they could also potentially express bottom-up influences of breathing on brain activity ( Girin et al., 2021 ;Herrero et al., 2018 ;Heck et al., 2017 ), or a combination of both events. An interesting example of bottom-up relationships is the enhancement of our sense of smell during inhalation ( Masaoka et al., 2014 ), via synchronization of brain activity to the respiratory phase ( Ito et al., 2014 ;Nguyen-Chi et al., 2016 ) and enhancement of cortical rhythms specifically during a phase of the respiratory cycle ( Tort et al., 2018 ;Masaoka et al., 2014 ). However, both strong odors and direct stimulation of olfactory-related brain areas produce rapid and important changes in breathing ( Masaoka et al., 2014 ). That is, sensory system activity may drive breathing and, in turn, brain activity rhythms may be phase-locked to the respiratory cycle.
Other interesting examples of observed effects of the respiratory phase on brain activity include enhancement of pain responses ( Iwabe et al., 2014 ) and faster recognition of fear expressions ( Zelano et al., 2016 ) during inspiration.
We would also mention that the transitional breathing changes were pharmacologically induced. Propofol acts substantially through a GABA effect ( Sahinovic et al., 2018 ). The GABA is the second most ubiquitous neurotransmitter present in 15% − 20% of brain neurons ( Buzsáki et al., 2007 ). As other hypnotic agents, when administered at relatively low doses and at low velocity, propofol may potentiate the neural events initiating sleep both by reducing waking cerebral cortex activity and via direct action on the brainstem arousal system ( Horner, 2009 ;Mashour and Hudetz, 2017 ). Interestingly, we have recently illustrated how other hypnotics with robust GABA action, such as benzodiazepines, may alter functional connectivity in the cerebral cortex mostly affecting the primary sensory areas, which suggests that the reduction of sensory processing may be a relevant mediator of the sedative effect . Nevertheless, propofol at higher doses and faster velocity during general anesthesia may virtually affect all neural systems ( Moody et al., 2021 ). Therefore, although we used a slow propofol administration regime, our observations cannot be currently extrapolated to naturally induced sleep.

Conclusions
The study aimed to illustrate how the brain reorganizes itself to control a basic behavior when moving between states of consciousness. The adopted image analysis strategy allowed us to identify brain systems plausibly contributing to breathing modulation during wakefulness and unconsciousness. The fall in breathing from wakefulness to unconsciousness was associated with activity reduction in brain systems underlying distinct aspect of consciousness, such as cognitive awareness, selfawareness and sensory awareness. Breathing recovery during the unconscious state was coupled to deep brain structures controlling instinctive and homeostatic behaviors. All in all, in a broad sense, our results are in line with the proposals that phenomenologically contemplate consciousness as the result of the largest-scale integration of brain activity ( Northoff and Lamme, 2020 ;Tanabe et al., 2020 ;Zhang et al., 2020 ;Pujol et al., 2021 ).

Declaration of Competing Interest
The authors report no financial interests or potential conflicts of interest.