Slow breathing and hypoxic challenge: cardiorespiratory consequences and their central neural substrates

Abstract

Controlled slow breathing (at 6/min, a rate frequently adopted during yoga practice) can benefit cardiovascular function, including responses to hypoxia. We tested the neural substrates of cardiorespiratory control in humans during volitional controlled breathing and hypoxic challenge using functional magnetic resonance imaging (fMRI). Twenty healthy volunteers were scanned during paced (slow and normal rate) breathing and during spontaneous breathing of normoxic and hypoxic (13% inspired O2) air. Cardiovascular and respiratory measures were acquired concurrently, including beat-to-beat blood pressure from a subset of participants (N = 7). Slow breathing was associated with increased tidal ventilatory volume. Induced hypoxia raised heart rate and suppressed heart rate variability. Within the brain, slow breathing activated dorsal pons, periaqueductal grey matter, cerebellum, hypothalamus, thalamus and lateral and anterior insular cortices. Blocks of hypoxia activated mid pons, bilateral amygdalae, anterior insular and occipitotemporal cortices. Interaction between slow breathing and hypoxia was expressed in ventral striatal and frontal polar activity. Across conditions, within brainstem, dorsal medullary and pontine activity correlated with tidal volume and inversely with heart rate. Activity in rostroventral medulla correlated with beat-to-beat blood pressure and heart rate variability. Widespread insula and striatal activity tracked decreases in heart rate, while subregions of insular cortex correlated with momentary increases in tidal volume. Our findings define slow breathing effects on central and cardiovascular responses to hypoxic challenge. They highlight the recruitment of discrete brainstem nuclei to cardiorespiratory control, and the engagement of corticostriatal circuitry in support of physiological responses that accompany breathing regulation during hypoxic challenge.

Fig 1. Physiological data acquisition, illustrating experimental design.

The figure shows data recording (CED Spike software, Cambridge) during the experiment in which participants breathed normoxic and hypoxic air mixtures over periods of approximately 5 minutes at their natural spontaneous rate (unpaced) and at paced rates approximating to a regular rate (paced normal rate: 9.9 breaths per minute) and at a slow rate (paced slow rate 5.5 breaths per minute). These task conditions were counterbalanced within and between participants and for hypoxia periods were preceded by a 1.5 minute ‘wash-in’ period and followed by a 1 minute ‘wash-out’ period. During the experiment, volume changes associated with breathing were recorded directly using pneumotachography alongside measurement of expired CO_2, and through pulse oximetry, arterial oxygen saturation, pulse volume and pulse rate. The experiment entailed repetitions of task conditions in hypoxia and normoxia. For a subset of participants, beat-to-beat blood pressure was also recorded.

Fig 2. Group physiological effects elicited by the experimental manipulations.

Error plots of average physiological response measured in the experimental different task conditions. The data validate the experimental manipulations for respiratory rate and oxygen saturation (Fig 2A and 2B) and show a significant effect of hypoxia on heart rate (Fig 2C). Task effects were also observed on minute ventilation (VE; Fig 2D), tidal volume (VT; Fig 2E) and end tidal CO₂ level (Fig 2F). Results for other parameters are presented within the text.

Fig 3. Physiological data including beat-to-beat blood pressure in subset N = 7.

A) Beat-to-beat blood pressure acquisition during fMRI. The figure presents a close-up snapshot of physiological data recording as in Fig 1, but also including recording of beat-to-beat blood pressure, which was acquired in the experiment from 7 of the 20 participants. B) Beat-to-beat blood pressure effects elicited by experimental manipulations. The figure mean data for the blood pressure changes recorded from across the subset of participants (N = 7). Observed differences were non-significant at standard statistical threshold. C)-H) Experimental evoked physiological responses for the subset of participants). The figure shows the pattern of physiological changes in the subset of participants in whom blood pressure was recorded for comparison with the whole group data depicted in Fig 2. The responses within this subset closely mirrored those seen for the whole group (N = 20).

Fig 4. Brain responses to experimental task manipulations.

Group data is presented on sagittal coronal and horizontal sections of a normalized template brain to illustrate suprathreshold activity differences associated with task conditions. Data is illustrated at a significance of P<0.05 corrected, determined using the combination of using at a voxel-wise threshold significance of P<0.001 uncorrected in combination with a cluster extent threshold > 41 contiguous voxels (computed through Mont Carlo simulation with 1000 iterations [20]). A) Main effect of hypoxia: increased activity when breathing 13%O₂ gas mixture v. normoxic air. Increased activity during hypoxic challenge is observed within regions including occipitotemporal cortex, amygdala and pons. B) Main effect of paced breathing rate: increased activity associated with paced slow breathing v. paced normal rate breathing: Increased activity during slow relative to normal rate breathing is observed within regions including cerebellum, sensorimotor cortices, dorsal pons, midbrain and thalamus. C) Activity reflecting interaction between presence and absence of hypoxia during paced normal rate v. slow breathing. Activity within left lateral frontal pole and ventral striatum reflects the impact of slow breathing on brain response to hypoxia.

Fig 5. Focal brainstem activity correlating with physiological changes.

Selected illustration of activity correlated with physiological changes. Data is illustrated at a voxel-wise threshold significance of P<0.001 uncorrected). A) within medulla, where the same locus reflects blood pressure increases and heart rate variability increases consistent with a neural substrate for the baroreflex mediating adjustments to hypoxic challenge. B) within pons and midbrain, where adjacent nuclei demonstrate relationships with decreases in heart rate and increases in tidal volume, putatively representing human homologues of brainstem centres supporting cardiorespiratory coupling as identified in experimental animals.

Fig 6. Focal activity within insula cortex correlating with physiological changes.

Sagittal and horizontal sections from a standard template brain presented to highlight the location of activity changes within insula cortex associated with task-induced changes in physiological regressors. Of note are the deep insula components of signal change associated with heart rate decrease, merging with a marked engagement of basal ganglia, mirroring striatocortical activation previously observed in relation to expectancy-related heart rate deceleration [22]. The broadly distributed signal change, associated with this particular analysis arguably appears to carry most artefact, beyond what was controlled for by including movement, global and arterial O₂ / end tidalCo₂ regressors as confounding covariates within the analyses. Ventilation (VT) evoked predominantly right hemispheric changes in parietal and insula cortices and basal ganglia. Increases in end tidal CO₂ was associated with enhanced activation within posterior ‘primary interoceptive’ insula, but did not impact global signal at threshold significance. Heart rate variability, blood pressure increases and blood pressure variability were associated with focal activation of distinct subregions of anterior and mid insula consistent with viscerotopography [23]. Data is illustrated at a significance of P<0.05 corrected, determined from the combination of voxel-wise significance and a cluster extent thresholds (see Methods).