From Lung to Brain: Respiration Modulates Neural and Mental Activity

Abstract

Respiration protocols have been developed to manipulate mental states, including their use for therapeutic purposes. In this systematic review, we discuss evidence that respiration may play a fundamental role in coordinating neural activity, behavior, and emotion. The main findings are: (1) respiration affects the neural activity of a wide variety of regions in the brain; (2) respiration modulates different frequency ranges in the brain’s dynamics; (3) different respiration protocols (spontaneous, hyperventilation, slow or resonance respiration) yield different neural and mental effects; and (4) the effects of respiration on the brain are related to concurrent modulation of biochemical (oxygen delivery, pH) and physiological (cerebral blood flow, heart rate variability) variables. We conclude that respiration may be an integral rhythm of the brain’s neural activity. This provides an intimate connection of respiration with neuro-mental features like emotion. A respiratory-neuro-mental connection holds the promise for a brain-based therapeutic usage of respiration in mental disorders.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12264-023-01070-5.

Introduction

Respiration and Brain—Nuisance or Neural?

Respiration, being so closely coupled to heart activity and oxygen supply, is key in maintaining metabolic activity in all organs including the brain. Given the increased efficiency of aerobic over anaerobic conditions in cellular energy dynamics, one might predict that the brain should be optimized to maximize its direct coupling with systems that supply its metabolism [1, 2]. The brain’s metabolic-energetic coupling to respiration is at odds with how neuroscientists methodologically treat respiration. Respiration-related neural activity is typically considered noise, and entire fields are dedicated to stripping it from brain data, particularly in fMRI (e.g., with global signal regression, independent components analysis, and RETROICOR [3–5].

While many researchers do not consider respiration beyond stripping it from their datasets, it is possible that elements of the respiration-coupled neural signal are useful, and can inform us about higher cognitive processing, emotion, and behavior. To test this hypothesis, we have conducted an extensive systematic review of combined brain and respiration studies in both functional magnetic resonance imaging (fMRI) and electroencephalography (EEG). In this analysis, we investigate the topographic and dynamic effects of respiration on the brain’s neural activity. Moreover, to test the strength of the link between neural and respiratory activity, we investigated the impact of different respiration protocols on biochemical, physiological, neural, and mental activity. Our findings should help uncover whether respiration plays an integral role in the brain’s physiology beyond its typical framing as a nuisance variable.

From the Rhythms of Respiration to Brain Rhythms

fMRI studies have shown that respiration frequency and volume are strongly correlated with the blood oxygen level dependent (BOLD) signal [6–10]. In fact, respiration alone has been shown to account for as much as 27% of the variance in the BOLD signal [11]. For this reason, the physiological signals are commonly regressed out of data as noise (for a review see [12]). Emerging evidence, however, suggests that fluctuations in the power of spatially distributed slower neuronal oscillations in the frequency range of respiration (0.01 to 0.3 Hz; [13]) drive the activity of faster frequencies and even mental features like consciousness and self [14–19].

Like fMRI, EEG can be used to assess the respiration-induced modulation of slower neural signals (0.01 to 0.3 Hz [20, 21]. Because of its greater temporal resolution, EEG can be used to investigate the respiration-induced modulation of faster frequencies like theta [22], alpha [23], and gamma [24]. These studies suggest that respiration provides a slow (0.01 to 0.3 Hz) intrinsic rhythm which may be coupled with analogous slower neural rhythms in the same frequency range. Thus, respiration may act as a slow-moving envelope carrying and influencing faster frequencies, which then manifest as a dynamic mixture of respiratory-influenced neural activity.

Respiration and Physiology/Biochemistry

Mammalian respiration is innately connected with other physiology and biochemistry. For example, respiratory sinus arrhythmia (RSA) is a phase relationship between respiration and heart rate. In humans, upon inhalation, the time between heartbeats (measured as R-R intervals or RRIs) tends to decrease in length, which is indicative of heartbeat acceleration. Conversely, upon exhalation, RRIs tend to increase in length, in other words, the heartbeat decelerates [25]. Furthermore, prolonged periods of hyperventilation (>30 s) tend to increase mean heart rate (HR), lower heart rate variability (HRV), reduce the blood gas partial pressure of carbon dioxide (pCO₂), and decrease cerebral blood flow [26–33]. Conversely, slow respiration tends to lower mean heart rate, increase HRV, raise pCO₂, and increase cerebral blood flow [13, 34–37].

How do these physiological and biochemical factors influence the impact of respiration on the brain’s neural activity? Given that the physiological and biochemical variables reviewed above express similar spatiotemporal influences on the brain’s neural activity, one might expect that these more fundamental biochemical processes might modulate and mediate respiration. This would provide a first insight into the physiological mechanisms for the coupling of the brain’s neural activity to respiration.

Respiration and Mental Activity

Neuropsychiatric conditions such as anxiety and panic disorder are often accompanied by irregularities in respiration patterns (for reviews see [26, 38–40]). In fact, hyperventilation is a trait that researchers argue is self-perpetuating and responsible for worsening neuropsychiatric disorders such as anxiety. The hyperventilation theory [41] states that panic patients often present with chronic hyperventilation. By this account, panic attacks are brought on by cognitive misinterpretations of bodily symptoms such as accelerated heart rate, decreased heart rate variability, dizziness, and tingling in the extremities that often accompany hyperventilation.

In contrast to the hyperventilation theory account [42], hyperventilation can be described as a subconscious preventive measure the body uses to maintain the pCO₂ in the blood below the threshold values of peripheral and central chemoreceptors. Klein proposed that over time, chemoreceptors become hypersensitive due to the sustained levels of low pCO₂ brought on by chronic hyperventilation. The authors proposed a positive feedback loop, where hypersensitive chemoreceptors promote hyperventilation to avoid triggering the body’s asphyxiation alarm which, in turn, leads to more panic and anxiety. Slow respiration has also been reported to mitigate problematic mental activity and the accompanying biochemistry/physiology brought on by chronic hyperventilation [43]. In fact, multi-session slow respiration protocols have been shown to increase overall positive affect [44–48].

An analysis by Klimesch and colleagues [49] reported on the potential architecture linking brain and body oscillations. The author argues that large reductions in energy demand drive neural systems to express phase and amplitude coupling with physiological rhythms such as respiration at distinct frequencies. Klimesch claims that, as a result, the respiration signal tends to have a set of “preferred” frequencies centered around 0.08, 0.16, and 0.32 Hz. Klimesch hypothesizes that each frequency carries with it distinct neural and mental correlates. This appears to be directly in line with our hypothesis as we suggest that there are distinct neural and mental features that are coupled to breathing at or close to these frequencies, such as in the case of slow (~0.1 Hz), spontaneous (~0.2 Hz), and fast breathing (~0.5 Hz).

The synchronization between respiration and slow neural activity is likely key to understanding the brain-physiology relationship. Higher degrees of coupling between respiration and brain activity likely manifest as the inverse of symptoms of panic disorder (e.g., relaxation, greater attention, and more measured thoughts). Slow rhythms thus provide a link and shared feature of respiration, neural, and mental activity serving as their “common currency” [18, 19]. We, therefore, include in our review studies using different respiration protocols including slow and fast ones in order to show their impact on both neural and mental activity. By establishing a “common currency”, between respiration, brain, and mind, we hope to provide a novel theoretical framework that serves to inspire future research in this area and provide a springboard for possible therapeutic interventions in neuropsychiatric disorders such as anxiety.

Overview of Steps

We included studies that reported the pCO₂ (via capnography), fMRI, EEG, electrocardiogram (ECG), and subjective questionnaires to piece together the neuronal and mental effects of respiration including their mediation of physiological and biochemical variables. We hypothesized that: (1) respiration affects neural activity in terms of topography and dynamics throughout widespread regions and all frequencies; (2) given the fact that the spontaneous frequency ranges between physiology (heart, pCO₂) and respiration overlap, we also expect their to be overlap in the topography and dynamics of neural signals influenced by these processes; and (3) given the existing evidence of respiratory mechanisms that affect pCO₂ and HRV, we predict that hyperventilation provokes panic or anxiety, and conversely, we predict that slow respiration promotes emotional well-being and relaxation (Fig. 1).

Fig. 1.

The theoretical framework of the coupling from respiration to mental features through biochemical (blood pH, Bohr effect), physiological (heart rate variability/HRV, cerebral blood flow/CBF), and neuronal (topography, dynamic) variables.

Materials and Methods

Article Search

We followed the recommendations of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement [50]; its design and flow are shown in Fig. 2). This review focused on the evaluation of research articles obtained from a survey conducted by the primary author in August 2021. The search included terms such as "respiration", "end-tidal carbon dioxide, "heart rate", and "heart rate variability" in combination with "electrocardiogram", "electroencephalogram", and "functional magnetic resonance imaging". To ensure that this study captured recent literature, the results were restricted to studies published between 2000 and 2021.

Fig. 2.

PRISMA scheme. Representation of the reviewal process for studies included in the survey.

Articles were evaluated for inclusion if they included human subjects, were full-text original research articles, and evaluated respiration’s inherent relationships with heart activity, neuronal activity, CO₂ levels, or mental activity. Articles were excluded if they focused on animals, children (<18 years of age), sleep, or were case studies. Articles were also excluded if they investigated a meditation practice where slow respiration or hyperventilation was not a primary measured outcome. Twenty-nine articles met these exclusion criteria and were not included in the review.

Our initial search captured relatively few studies focusing on the mental implications of the respiration protocols, particularly hyperventilation. As a result, a secondary search was conducted using the online collaborative database OMNI (https://ocul.on.ca/omni/). Identification, review, and article inclusion were conducted by the first author in January 2022. This search intended to investigate the respiratory pathology of anxiety/panic disorders in hopes of linking it to the information compiled in this review. Search algorithms comprised the terms "anxiety" and "panic disorder" used in combination with "hyperventilation". To ensure this search captured the recent literature, studies were limited to the years between 2000 and 2021. This search yielded 8 full-text research articles.

Organization of the Review

As articles were reviewed, they were placed in the following categories: (1) Topography (spatial relationships), (2) Neuronal Dynamics (temporal relationships), (3) Respiration protocols and physiological or neuronal interactions, and (4) Respiration protocols and mental interactions. Further details on references of individual studies as well as which studies were included in categories 1–4 can be found in Supplementary Tables 2–5. Topographical results included fMRI studies investigating regions associated with respiration, heart activity, and pCO₂ fluctuations. ECG, EEG, and fMRI studies that investigated global and regional frequencies associated with respiration, heart activity, and CO₂ fluctuations contributed to the Neuronal Dynamics category. Studies that investigated the effects of respiration protocols fell into two categories. The first included studies that investigated the physiological and neuronal interactions of respiration protocols. The second included studies that investigated the impact of respiration protocols on the mental level, that is, on mood and cognition.

The search yielded 75 unique relevant studies. Some studies reported multiple measures (e.g., EEG, fMRI, ECG, pCO₂, and questionnaires) and were included in multiple categories. Each section yielded studies [9, 26, 27, 34], respectively. Collectively, these studies attempted to investigate the significant physiological, neuronal, and mental variance correlated to fluctuations in respiration, heart activity, and pCO₂.

Results

Topography

Regions Modulated by Respiration

For this review, we compiled 16 studies that collectively attempted to describe the topographical relationships of respiration in the brain under spontaneous conditions. Regions most associated with respiration included frontal [medial and lateral prefrontal cortex (PFC), orbitofrontal PFC, superior frontal gyrus)], temporal [superior temporal gyrus (STG), parietal (somatosensory cortex (SSC), and primary motor cortex (PMC)], occipital [occipital cortex (OC) and supramarginal gyrus SMG)], midline [anterior (ACC), medial (MCC), and posterior cingulate cortex (PCC)/precuneus, and cuneus)], insular [anterior (aINS), medial (mINS), and posterior (pINS)], and cerebellar regions. Subcortical regions (thalamus, caudate, and putamen) also appear to be influenced by respiration. More details on the number of studies and references pertaining to each region are listed in Fig. 3A and Supplementary Table 6, respectively.

Fig. 3.

A–C Regions that have been shown to covary in activity with factors of respiration (A), factors of heart activity (B), and levels of CO₂ (C). D Summary of regions that have been shown to covary with all three respiration, heart activity, and CO₂. Abbreviations: PCC, posterior cingulate cortex; MCC, middle cingulate cortex; ACC, anterior cingulate cortex; STG, superior temporal gyrus; SSC, somatosensory cortex; SMG, supramarginal gyrus; SMA, supplementary motor area; DLPFC, dorsolateral prefrontal cortex; VLPFC, ventrolateral prefrontal cortex; OFPFC, orbitofrontal prefrontal cortex; PMC, primary motor cortex; SFG, superior frontal gyrus; MFG, middle frontal gyrus; DMPFC, dorsomedial prefrontal cortex; SPL, superior parietal lobule; aINS, anterior insula; mINS, middle insula; pINS, posterior insula; Caud, caudate; Thal, thalamus; LPFC, lateral prefrontal cortex; Cereb, cerebellum; amyg, amygdala; hipp, hippocampus; FG, fusiform gyrus; pOP, parietal operculum; AG, angular gyrus.

Regions Modulated by Heart Activity

For this review, we compiled 19 studies that collectively described the topographical relationships of heart activity under spontaneous conditions. Heart activity is most associated with activity in frontal (OFPFC, MPFC, LPFC, and MFG), temporal (STG, MTG, and fusiform gyrus), parietal (PMC, SSC, SPL, pOP, and SMA) occipital (OC and angular gyrus), midline (ACC, MCC, PCC, precuneus), insular (aINS, mINS, pINS), and cerebellar regions. Subcortical (thalamus, putamen, amygdala) and brainstem (pons) regions have also been shown to be influenced by heart activity. More details on the number of studies and references pertaining to each region can be found in Fig. 3B and Supplementary Table 8, respectively.

Regions Modulated by CO₂

For this review, we compiled 10 studies that collectively attempted to describe the topographical relationships of pCO₂ under spontaneous conditions. Regions that are most associated with fluctuations in pCO₂ include frontal (OFPFC, MPFC, LPFC), temporal (STG), parietal (PMC, SSC), occipital (SMG, OC), midline (ACC, MCC, PCC, precuneus), insular (aINS, mINS, pINS), and cerebellar regions. Activity within subcortical regions (thalamus, putamen, caudate) also appears to fluctuate in activity with pCO₂ levels. More details on the number of studies and references pertaining to each region can be found in Fig. 3C and Supplementary Table 7, respectively.

Regions Associated with Physiological Overlap

For this review, we compiled 45 studies (34 unique studies and 11 studies reporting multiple measures) that collectively described the topographical relationships of all the physiology described in this review under spontaneous conditions. Regions associated with overlap of the described physiology include frontal (OFPFC, MPFC, LPFC), temporal (STG), parietal (PMC, SSC), occipital (OC), midline (ACC, MCC, PCC, precuneus), insular (aINS, mINS, pINS), and cerebellar regions. In addition, subcortical regions included the thalamus and the putamen. More details on the number of studies and references pertaining to each region can be found in Fig. 3D and Supplementary Table 9, respectively.

Dynamics

Frequencies Modulated by Respiration

We compiled 11 studies that collectively attempted to describe the dynamic relationship of respiration in the brain under spontaneous conditions. Respiration appears to be coupled to infraslow (IS, 0.01–0.1 Hz), slow cortical potential (SCP, 0.1–1 Hz), delta (1–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (>30 Hz) frequencies. The extent of this coupling varied by region. In the frontal lobe, respiration was shown to be coupled to all frequencies. In the insular lobe, respiration was shown to be coupled to IS, SCP, alpha, and gamma frequencies. In the parietal and occipital lobes, respiration appears to be coupled to IS, SCP, delta, theta, alpha, and beta frequencies. In the temporal lobe, respiration was shown to be coupled to IS, SCP, delta, theta, alpha, and gamma frequencies. In the anterior cingulate cortex, respiration was shown to be coupled to IS, SCP, and alpha frequencies. In the middle and posterior cingulate cortices, respiration was shown to be coupled to IS and SCP frequencies. In the precuneus, respiration was shown to be coupled to IS and alpha frequencies. Respiration was also shown to modulate activity in subcortical regions. In the thalamus, respiration was shown to be coupled with IS and alpha frequencies. Finally, in the amygdala and hippocampus, respiration was shown to be coupled to SCP and gamma frequencies. More details on the number of studies and references pertaining to each region can be found in Fig. 4A and Supplementary Table 10, respectively.

Fig. 4.

A–C Neural dynamics linked to respiration (A), cardiac activity (B) and CO₂ (C). Structures shown in the left lateral view are bilateral. Abbreviations: FL, frontal lobe; PL, parietal lobe; INS, insula; TL, temporal lobe; OL, occipital lobe; ACC, anterior cingulate cortex; MCC, middle cingulate cortex; PCC, posterior cingulate cortex; Precun, precuneus; Thal, thalamus; Amyg, amygdala; hipp, hippocampus; Cereb, cerebellum; MPFC, medial prefrontal cortex; OC, occipital cortex; LFHRV, low frequency heart rate variability; HFHRV, high frequency heart rate variability; IS, infraslow; SCP, slow cortical potentials.

Frequencies Modulated by Heart Activity

In order to investigate the dynamic relationships of heart activity we compiled 11 studies that collectively attempted to describe the relationships between low-frequency heart rate variability (LFHRV, 0.04–0.15 Hz) and high-frequency heart rate variability (HFHRV, 0.15–0.4 Hz) under spontaneous conditions.

(1) LFHRV

LFHRV appears to be coupled to activity within the frontal, parietal, and insular lobes. In addition, it appears to be coupled with activity within midline (MPFC, MCC), subcortical (thalamus, amygdala), and brainstem (pons) regions. More details on the number of studies and references pertaining to each region can be found in Fig. 4B and Supplementary Table 11, respectively.

(2) HFHRV

HFHRV appears to be coupled to activity within the frontal, parietal, occipital, temporal, and insular lobes. In addition, it appears to be coupled to activity within the cerebellum, midline (MPFC, ACC, MCC), subcortical (thalamus, amygdala), and brainstem (pons) regions. More details on the number of studies and references pertaining to each region can be found in Fig. 4B and Supplementary Table 11, respectively.

Frequencies Modulated by CO₂

We compiled 6 studies that collectively attempted to describe the dynamic relationships of fluctuations in the pCO₂ under spontaneous conditions. According to the studies included in this review, pCO₂ appears to be coupled to IS, delta, alpha, beta, and gamma activity. These coupling patterns were consistent throughout the frontal, parietal, temporal, and occipital lobes. More details on the number of studies and references pertaining to each frequency/region can be found in Fig. 4C and Supplementary Table 12, respectively.

Respiration Protocols and Dynamic and Physiological Interactions

Slow Respiration—Resonance Respiration

We compiled 27 studies that collectively attempted to describe the physiological and dynamic changes induced by resonance respiration. Overall, it appears that resonance respiration causes significant increases in LFHRV, slight increases in pCO₂ levels, and an overall shift towards slower neuronal frequencies. According to these data, within the frontal lobe, there appears to be increased IS coherence, increased SCP power, higher levels of inter and intra-hemispheric theta coherence, increased theta power, increased alpha power, increased alpha/high beta ratio, and decreased beta power during resonance respiration compared to spontaneous conditions. Within the temporal lobe, there appears to be increased IS coherence, increased SCP power, increased interhemispheric alpha asymmetry, and increased alpha/high beta ratio during resonance respiration compared to spontaneous conditions. In the occipital lobe, there appears to be increased SCP power, increased inter and intra-hemispheric theta coherence, increased theta power, increased alpha power, and decreased beta power during resonance respiration compared to spontaneous conditions. Within the parietal lobe, there appears to be increased IS coherence, increased SCP power, increased theta power, increased alpha/high beta ratio, increased alpha power, and decreased beta power during resonance respiration compared to spontaneous conditions. Within the cingulate cortex, there appears to be increased activation and increased alpha power during resonance respiration compared to spontaneous conditions. Globally, there appears to be increased IS coherence, increased SCP power, increased delta power, increased alpha power, and increased left brain activation during resonance respiration. Visual representation of the physiological and neuronal changes induced by resonance respiration can be found in Fig. 5B. More details on the main findings of individual references used to construct this figure can be found in Supplementary Table 5.

Fig. 5.

Schematic representation of physiological, neuronal, and mental activity during (A) spontaneous respiration (B) resonance respiration, and (C) hyperventilation.

Fast Respiration—Hyperventilation

We compiled 5 studies that collectively attempted to describe the neuronal dynamic changes induced by hyperventilation. In contrast to resonance respiration, it appears that hyperventilation elicits a decrease in HRV, decreased pCO₂, and increases in higher frequencies. Within the frontal lobe, it appears that there is increased theta and beta power during hyperventilation compared to spontaneous conditions. Within the occipital and parietal lobes, it appears that there is increased theta and alpha power during hyperventilation compared to spontaneous conditions. Globally, there appears to be an increase in theta power compared to spontaneous conditions.

In this study, we captured a relatively small number of papers pertaining to the physiological and dynamic changes induced by hyperventilation. As a consequence, these data should be used as an indication of what might occur, but more research is necessary to understand the full scope of the physiological and dynamic implications of hyperventilation. A visual representation of the changes induced by hyperventilation can be found in Fig. 5C. More details on the main findings from individual references used to construct this figure can be found in Supplementary Table 5.

Respiration Protocols—Impact on mental activity

Our search captured 9 studies that collectively attempted to describe the effects that respiration protocols have on mental activity. Four of these studies investigated the effects of resonance respiration protocols (~0.1 Hz) on perceived anxiety/stress levels [44–46, 48]. A common theme among all these studies was an overall decrease in symptoms of negative affect after their respective intervention periods. Two studies investigated the effects of hyperventilation protocols (~0.5 Hz) on mental activity [51, 52]. Overall, these studies reported increased levels of perceived anxiety and stress during fast respiration, however, participants scored lower on overall perceived anxiety/stress after several weeks of interventions.

Three studies only included spontaneous protocols [53–55]. One study investigated interoceptive awareness of respiration and found that decreasing its capacity was correlated with increased levels of activity within the anterior insula and increased levels of anxiety [53]. Another study described a positive association between HRV and heartbeat detection [54]. This study reported that decreases in heartbeat detection were linked with increased levels of perceived stress and anxiety. The third study looked at patients in remission from major depressive disorder [55]. In this study, patients were found to have increased respiration rates and PCC/parahippocampal gyrus activity compared to healthy controls. In addition, this study found patients to experience increased respiration pause variability compared to healthy controls, and this was correlated to symptom severity. Further information on the main findings of individual studies can be found in Supplementary Table 5.

Secondary Search for Articles on Respiration and Anxiety

Our secondary search yielded a total of 8 studies that investigated the respiratory pathology of anxiety and panic disorder patients. Patients in these studies commonly presented with chronic hyperventilation [26, 27, 30, 32, 33, 39, 40, 56], hypocapnia (low pCO₂; [26, 27, 39, 56]), and lower levels of HF HRV [30] compared to healthy controls. Researchers suggested that patients tend to experience more significant feelings of discomfort, negative affect, and dyspnea at lower pCO₂ threshold values and with smaller increases in pCO₂ compared to healthy controls. As a result, patients tend to hyperventilate as a protective mechanism to keep pCO₂ levels below chemosensory threshold values. Chronic hypocapnia leads to hypersensitive peripheral and central chemoreceptors, which tend to manifest as a self-perpetuating mechanism for hyperventilation, panic, and anxiety [26, 39, 40, 56]. Other theories represented in this literature [41] suggest that the cognitive misinterpretation of hyperventilation and accompanying bodily symptoms themselves (rather than hypersensitive chemoreception) are responsible for the propagation of anxiety and panic. This assumes that patients tend to disproportionately evaluate feelings of dyspnea, elevated heart rate, and low HRV as deadly. This may lead patients to experience increased levels of anxiety and panic [26, 39, 40, 56]. Further information on the main findings of individual studies can be found in Supplementary Table 14.

Discussion

From Lung to Brain I—Respiration and Topography

We first found the involvement of widespread regions in the brain associated with respiration. Rather than implicating specific regions or networks, respiration seems to impact both medial and lateral cortical regions as well as the anterior and posterior cortex including higher- and lower-order regions. The global role of respiration in the brain signal hinges on physiological and biochemical processes. We show significant topographical overlap between respiration, heart activity, and CO₂ fluctuations. Regions associated with this overlap include frontal (OFPFC, MPFC, LPFC), temporal (STG), parietal (PMC, SSC), occipital (OC), midline (ACC, MCC, PCC, precuneus), insular (aINS, mINS, pINS), and cerebellar regions. In addition, subcortical regions included the thalamus and the putamen (for individual references see Supplementary Table 9). Interestingly, many of these regions have been implicated in important networks responsible for the sense of self and cognition such as the default mode and salience networks [57–59]. This, albeit indirectly, provides evidence for the possible connection of respiratory activity with neuro-mental functions like self and consciousness; this is further supported by the differential neuro-mental impact of different slow-fast respiration protocols.

We believe that this global topographic involvement is, at least in part, mediated by the biochemical and physiological dynamics inherently linked to respiration. For example, CO₂ has been shown to fluctuate with the respiratory cycle and has been shown to be a potent vasodilator [6–9, 60, 61]. This dilatory property of CO₂ facilitates modulations in cerebral blood flow. In fact, there have been studies applying transcranial Doppler ultrasound that show as much as a 5% modulation in blood flow per 1 mmHg change in pCO₂ [29]. That considered, one would expect the BOLD signal to be most affected by the respiratory cycle in regions with a high blood supply, which is exactly what we present in this review.

From Lung to Brain Ii—Respiration and Dynamics

It is becoming well established that respiration entrains oscillations at the same frequency as the respiration rate in several regions, at least in the rodent brain [24, 62, 63]. Moreover, these studies particularly implicate the theta and gamma bands in the frontal regions are phase-locked to respiration [24, 62, 63]. Researchers suggest that the respiratory rhythm in rodents, like humans, is propagated and modulated via a central pattern generator (pre-Botzinger complex) buried deep within the brainstem [15, 64, 65].

In mammals, activity within the pre-Botzinger complex is particularly sensitive to changes in blood acidity (pH) levels (for details on the relationship between respiration and pH see the box provided). Central chemoreceptors act in a negative feedback loop with respiratory and cardiac centers to modulate heart and respiration rates accordingly [64, 65]. Given that there is a fundamental connection between respiration and central brain activity, the global frequency involvement of respiration should not come as a surprise.

In line with other studies, our review confirms the involvement of multiple frequencies in the coupling of respiratory and neural activity (for reviews see [43, 66, 67]). Support for the involvement of slower frequencies (0.01 to 0.3 Hz) comes (indirectly) from fMRI where infraslow fluctuations in respiratory volume (0.01 to 0.05 Hz; [6, 12]) or frequency (0.1–0.3 Hz; [10]) modulate neural activity in that frequency range. This is important as the infraslow frequency range of fMRI (0.01 to 0.1 Hz) corresponds to the frequency range of respiration (including its variability; 0.01 to 0.3 Hz).

From Lung to Mind—Respiration and Mental Activity

Although neuroimaging techniques such as fMRI have been used in many studies, about 2% of all examinations had to be aborted due to abnormally high state anxiety related to the scanner [68]. It has been demonstrated that increased state anxiety carries abnormal coupling patterns between breathing and cardiac oscillations. In fact, negative RSA refers to the condition where increased state anxiety induces a complete reversal of the phasic relationship between respiration and cardiac activity where the heart rate slows down during inhalation and speeds up during exhalation [69, 70].

Chronic mental conditions such as anxiety and panic disorder are often accompanied by abnormalities in respiratory patterns [71]. Patients often have a faster respiration rate, higher mean heart rate, diminished HRV amplitude, and a lower level of baseline pCO₂ (for reviews see [26, 38–40, 56]). Over time, diminished levels of pCO₂ lower the threshold of chemoreceptors to trigger the body's intrinsic asphyxiation alarm. With lower threshold values, slight increases in pCO₂ may trigger feelings of dyspnea, discomfort, anxiety, and panic [42]. Interestingly, respiration is an autonomic process that can be somatically modulated. In fact, studies investigating slow respiration techniques have shown it to increase HRV amplitude, raise pCO₂, and promote feelings of positive affect [44–46, 71].

Our findings provide a direct link between respiratory, neural, and mental activity. The impact of different respiration protocols on neural dynamics and topography as well as on mental features like emotions suggest a mechanistic link. We propose the topography and slow-fast dynamics shared by respiratory, physiological, neural, and mental activity act as their “common currency” [18, 19]. This is illustrated in Fig. 5 where we draw connections from respiratory dynamics to neural and mental dynamics mediated by the dynamics of physiological and biochemical variables related to respiration. Beyond opening a novel understanding of the body-brain-mind relationship, this can serve as the basis for the brain-based mechanistic development of novel forms of respiration-based therapeutic intervention in mental disorders like anxiety disorders.

Limitations and Future Directions

A major limitation to consider is that in some instances, sample sizes were small, and methodologies between studies varied significantly. Thus, we acknowledge a certain degree of generalizability may be lacking from this work. In addition, although this manuscript clearly describes clinical implications, at the time of writing, there appears to be a lack of large-scale clinical data on this matter. Although we describe some of the existing clinical research in our analysis, more work is needed to establish a clear neuro-respiratory mechanism for mental activity. The future is bright, however. We are hopeful that recent increases in interest in non-pharmacological interventions for individuals suffering from mental illness and other cognitive disorders may bring much-needed funding into the field so that these data can be acquired.

Future research should focus on refining existing paradigms, such as developing a more precise methodology for collecting and analyzing respiration-entrained neuronal data. Large amounts of high-quality neuro-respiratory data may allow researchers to flesh out the mechanistic links between respiration, neuronal activity, and behavior. In addition, large-scale, longitudinal clinical trials should be conducted to assess the efficacy of deliberate breathing protocols in the treatment of neuro-cognitive disorders in at-risk populations (i.e., advanced aging, anxiety, and depression, …). Further, this “big data” approach should focus on the dynamics of neuro-respiratory covariance and how these interactions impact behavior and mental health.

Conclusions

Respiration is a fundamental activity of life as it provides the necessary metabolic ingredients for all organs of our body including the brain. Does respiration contribute useful information to the neural signal rather than just "noise"? Our review suggests this is indeed the case.

We show that respiration affects a widespread set of regions throughout the brain as well as a range of frequencies ranging from slow (0.01 to 0.3 Hz) to faster (1–80 Hz). Our review furthermore demonstrates that physiological (HRV) and biochemical (CO₂) variables induce similar neural changes in the brain in both its topography and dynamics. These physiological and biochemical variables likely mediate the coupling of respiratory and neural activity thus providing their direct link. Addressing our initial question, these findings strongly suggest that respiration is not a mere nuisance variable but an integral component of the brain’s neuro-mental activity which is mediated by various physiological and biochemical variables.

Our review also shows the differential neural and even mental effects of slow and fast respiration protocols. The respiratory-neural connection seems to have particularly strong effects on emotions: slow respiration protocols like resonance respiration exert a relaxing and calming effect, while faster respiration tends to induce anxiety states. These observations suggest that slow-fast dynamics may be shared by respiratory, physiological, neural, and mental activity, thus providing a mechanistic (or better yet dynamic) link as their “common currency” [18, 19]. We, therefore, propose that an individual's respiratory rhythm serves a fundamental, intrinsic role that modulates the topography and dynamics of the whole brain. Going beyond respiration-brain coupling, this opens the door for the application of respiration as a therapeutic technique in mental disorders like anxiety disorders and others.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Appendix

The following citations were included in the topographic analysis (Supplementary Table 3) but were not referenced in the text [70, 72–95].

The following citations were included in the neural dynamic analysis (Supplementary Table 4) but were not referenced in the text [69, 70, 76, 79, 80, 87, 90–99].

The following citations were included in the effects of breathing protocols on physiological and neuronal dynamics analysis but were not referenced in the text [98, 100–109].

Conflict of interest

The authors declare that there is no conflict of interest regarding the publication of this article.