Recent insights into respiratory modulation of brain activity offer new perspectives on cognition and emotion

Abstract

Over the past six years, a rapidly growing number of studies have shown that respiration exerts a significant influence on sensory, affective, and cognitive processes. At the same time, an increasing amount of experimental evidence indicates that this influence occurs via modulation of neural oscillations and their synchronization between brain areas. In this article, we review the relevant findings and discuss whether they might inform our understanding of a variety of disorders that have been associated with abnormal patterns of respiration. We review literature on the role of respiration in chronic obstructive pulmonary disease (COPD), anxiety (panic attacks), and autism spectrum disorder (ASD), and we conclude that the new insights into respiratory modulation of neuronal activity may help understand the relationship between respiratory abnormalities and cognitive and affective deficits.

The influence of respiration on perception, cognition, and affect is relatively well-documented. For example, it is widely acknowledged that the conscious control of respiration affects heart rate variability and blood pressure, which in turn leads to changes in brain activity (Rau, Pauli, Brody, Elbert, & Birbaumer, 1993), stress reduction, and the regulation of emotional and cognitive processes (Arch & Craske, 2006; Grossman & Christensen, 2011; Paul, Elam, & Verhulst, 2007). These findings suggest that the exchange of oxygen for carbon dioxide, the key function of respiration, is one route of influence of respiration on brain activity, cognition, and affect. However, recent studies in animals and humans have shown that respiration influences rhythmic brain activity in multiple frequency bands and possibly across all brain areas and - at the behavioral level - sensory, affective, and cognitive functions via respiration-locked neuronal activity driven by two major sources. Evidence suggests that this influence is primarily driven by respiration-related sensory inputs and inputs from brain stem pattern generators to subcortical structures. The observable consequence of this influence is a rhythmic modulation of brain activity at the frequency of the breath and a modulation of power of higher frequency oscillations, including gamma band oscillations, which are widely recognized for their association with cognitive processes (Heck et al. 2017; Ito et al. 2014).

Neuronal oscillations occur in a variety of frequencies and show dynamic patterns of phase synchronization, or coherence, between different brain areas. Dynamic patterns of synchronization are implicated in affective and cognitive brain functions whereas atypical oscillation and coherence patterns are associated with affective, neurodevelopmental, and cognitive disorders. Here we review the rapidly growing literature linking respiration to brain activity and function. We discuss whether and how these new findings might inform our understanding of the role of respiration in affective disorders, such as panic attacks, or neurodevelopmental disorders affecting social and cognitive skills such as autism spectrum disorders (ASD). We also review literature on cognitive deficits in chronic obstructive pulmonary disease (COPD), discuss whether those findings support a possible causal link between respiratory and cognitive deficits, and consider the difference between nasal and oral breathing in the context of clarifying the link between respiration and brain activity and function.

We will also consider the possible influence of respiration on the brain’s functional connectome and review relevant literature. The functional connectome is described in terms of temporal correlation of activity between structures, indicating functional connectivity independent of the presence of anatomical connections (Aertsen & Gerstein, 1985; Kim et al. 2013). Respiration-locked activity from the olfactory bulb and likely other sensory inputs provide a rhythmic stimulus that drives neuronal oscillations in many brain areas. Artificial brain stimulation has been shown to reorganize the brains functional connectome (Huang et al. 2019), suggesting that brain stimulation by respiration-locked sensory inputs might influence the functional connectome.

1. Respiration, brain activity and cognition

Since the first report of respiratory modulation of delta and gamma oscillations in a non-olfactory part of the mouse neocortex (Ito et al., 2014), a rapidly growing number of publications confirmed the initial findings in both animals and humans. Studies in humans revealed that, as a functional consequence of respiratory modulation of brain activity, cognitive and affective functions are modulated by the respiratory rhythm (Arshamian, Iravani, Majid & Lundstrom, 2018; Grund et al., 2021; Johannknecht & Kayser, 2021; Nakamura, Fukunaga, & Oku, 2018; Perl et al. 2019; Zelano et al. 2016) (for a recent review see: Heck, Kozma, & Kay, 2019).

In an important study by Zelano and colleagues (2016), participants were exposed to photographs of faces expressing either fear or surprise for a brief duration (100 ms). Participants detected fearful faces more quickly during nasal inspiration than expiration. Moreover, the respiratory phase also appeared to have an effect on retrieval accuracy. Participants were presented with 180 pictures of different everyday objects. In a memory retrieval session, participants were presented pictures from the original set plus pictures they had not seen before. Results showed that retrieval accuracy was significantly higher for images presented during the inspiration phase in the retrieval session. It is important to highlight that nasal respiration and hence the activation of the olfactory bulb played a special role in this study. When participants in the Zelano study were breathing through the mouth, reaction times to both types of faces increased significantly, but there was no longer a difference in detection time between inhalation and exhalation. In the second experiment, respiratory modulation of memory only occurred when participants were breathing through the nose. These results are consistent with the original findings by Ito et al. (2014), showed that respiration, via sensory inputs from the olfactory bulb, modulates neuronal oscillations in the delta and gamma frequency bands in the neocortex of awake mice. The authors carefully dissected the role of the olfactory bulb and highlighted respiration-locked delta oscillations in the whisker barrel cortex (an area of the neocortex not involved in olfactory processing). In the same area, the authors showed that the power of gamma oscillations was modulated in phase with respiration, highlighting that both respiration-locked delta oscillations and respiration-locked power-modulation of gamma oscillations required an intact olfactory bulb (Ito et al., 2014). In turn, when the olfactory bulb was stimulated electrically with a rhythmically applied current, cortical rhythms were entrained to the stimulus rhythm applied to the bulb (Ito et al., 2014).

More recent studies provide additional evidence for respiratory modulation of brain function. Grund et al. (2021) investigated the combined influence of heartbeat and the breathing cycle on conscious sensory perception in humans by asking participants to report perceiving a near-threshold tactile stimulus of the index finger. The detection of near-threshold stimuli was more likely during the diastolic than systolic period of the heartbeat. Analysis of the respiratory cycle revealed that correctly detected stimuli closely clustered in time around inspiration onset, a finding consistent with the suggestion that inspiration onset or possibly the transition from expiration to inspiration is tuning the sensory system to optimize the processing of incoming information (Perl et al., 2019).

Active alignment of respiration to a task context was also observed by Johannknecht and Kayser (2021), who subjected participants to five different sensory detection and discrimination tasks and a short-term memory tasks. The analysis of task performance in relation to the respiratory cycle showed that participants tended to align their respiratory cycle to the timing of stimuli and their own response times. At the same time, the analysis of reaction times and response accuracy showed that reaction times were significantly modulated with the respiratory cycle, while there was no change in response accuracy.

Providing a potential neuronal mechanism underlying the influence of respiration on memory (Heck et al., 2019; Zelano et al., 2016), a study in mice showed that hippocampal sharp wave ripples, a brief high-frequency oscillation characteristic for the hippocampus, are phase locked to respiration (Liu, McAfee, & Heck, 2017). Sharp wave ripples are believed to be crucial for memory consolidation and recall (Buzsaki, 2015), a view that is supported by several studies showing a critical involvement of sharp wave ripples in memory consolidation and retrieval in mice (Malvache, Reichinnek, Villette, Haimerl, & Cossart, 2016; Nicole, Hadzibegovic, Gajda, Bontempi, Bem & Meyrand, 2016; van de Ven, Trouche, McNamara, Allen & Dupret, 2016), rats (Jadhav, Kemere, German, & Frank, 2012; Maingret, Girardeau, Todorova, Goutierre, & Zugaro, 2016; Papale, Zielinski, Frank, Jadhav & Redish, 2016; Pfeiffer & Foster, 2015; Singer, Carr, Karlsson & Frank, 2013), and non-human primates (Leonard & Hoffman, 2017; Logothetis et al. 2012; Ramirez-Villegas, Logothetis, & Besserve, 2015). For an in-depth review of the link between respiration and memory see (Heck et al., 2019).

A recent fMRI study investigated whether the reduction of “default mode network” activity observed during performance of a word memory tasks (Greicius, Supekar, Menon, & Dougherty, 2009) may be linked to stimulus-induced changes in respiratory behavior (Huijbers et al. 2014). Focusing on the relationship between changes in respiration and stimulus presentation, the authors calculated the average stimulus-aligned respiratory signal measured with an MRI compatible air-filled belt around the waist. Analysis of a stimulus-aligned average of the respiratory signal revealed that presentations of word stimuli were followed by a trough at 2 s post-stimulus and a peak at 4 s, indicating a stimulus triggered alignment of respiration patterns. We are using the descriptors “trough” and “peak” here because the authors do not specify whether troughs and peaks correspond to exhale or inhale. A subsequent analysis of respiratory phase aligned to stimulus presentation confirms that respiratory phase is locked to stimulus presentation. In addition, the authors show that the post-stimulus respiratory trough-peak modulation and the respiratory phase locking were significantly higher for correctly remembered words compared to forgotten words. The authors then asked participants to perform the task while holding their breath for 20 s during the presentation of word stimuli. Comparison with the control or “breathing” condition showed that a memory task related silencing of the default mode network activity only occurred during the control condition with participants allowed to breathe. These findings suggest a link between the amplitude and phase of respiratory movements in response to an external stimulus and successful cognitive performance as well as default mode network activity. The authors take their findings to “indicate a region-specific interaction between respiration and the fMRI signal associated with successful task performance” (Huijbers et al., 2014).

Nakamura and colleagues (Nakamura et al., 2018) investigated the link between cognitive functioning and phase transition during a retrieval process, showing that phase transition in the respiratory cycle can modulate cognitive performance. In a delayed visual recognition task, a cue was presented in respiratory phase-locked (phased) or regularly paced (non-phased) presentation paradigms. When the retrieval process encompassed the expiratory-to-inspiratory transition, participants performing the task in the phase-locked presentation condition exhibited increased response time and reduced accuracy. However, when the retrieval process encompassed the inspiratory-to-expiratory transition, there were no significant changes in response time or accuracy.

Oral and nasal breathing do not always display the same patterns of influence on cognitive function. In a recent study by Perl et al. (2019), four different cognitive tests were used to study respiratory modulation of cognitive performance. While three of these turned out to be dependent on nasal breathing, a lexical task was not. Nevertheless, nasal breathing seems to play a special role in human cognitive function. Arshamian and colleagues (2018) conducted a study investigating a possible influence of long-term breathing patterns on cognition function, using olfactory memory as a test. Specifically, the team asked whether long-term breathing (for one hour) either exclusively through the mouth or nose would influence memory consolidation, i.e. the process of transferring memory content from short-term to long-term storage. Healthy adult participants were asked to memorize 12 odors, followed by an hour of resting passively without sleeping (consolidation phase) during which they either breathed through their nose or mouth. Participants were then presented with a sequence of 24 odors, consisting of the 12 familiar and 12 new odors and asked to identify the familiar ones. The results showed that odor recognition memory was significantly increased after nasal respiration during the consolidation phase, compared with oral respiration. These findings provide the first experimental evidence that long-term respiratory behavior, in this case mouth vs. nose breathing, may affect cognitive function. However, since the paradigm used olfactory memory, the memory forming pathway and the pathway conveying respiratory modulation of brain activity were the same. It remains to be seen whether similar results would be obtained if the memory task is based on a non-olfactory sensory modality.

2. Neuronal mechanisms of respiratory modulation of brain activity

The studies discussed thus far provide support for the idea that respiration has a direct influence on perceptual and cognitive processes, most likely via sensory inputs that are phase-locked to breathing. Experimental evidence from animal studies identified respiration-locked activity in the olfactory bulb as a main neuronal driving force for brain-breath coupling (Ito et al., 2014; Liu et al., 2017). In humans the difference between oral and nasal breathing on cognitive processing supports the notion that olfactory bulb activity plays a crucial role in the coupling of brain activity and function to the breath (Arshamian et al., 2018; Perl et al., 2019; Zelano et al., 2016). Provided that the gas-exchange functions of nasal and oral respiration are likely to be similar, the current assumption is that the different effects of nasal and oral breathing on cognitive functions and brain activity are linked to respiration-locked sensory inputs. However, strengthening the case for a causal influence of respiratory activity on cognition would benefit from identifying the underlying mechanism responsible for the influence. This can be accomplished through a closer examination of the neocortical, oscillatory activity involved in cognition.

Local field potential (LFP) oscillations reflect the synaptic input to the area of observation (Buzsaki, Anastassiou, & Koch, 2012). This rhythmically modulated synaptic activity influences the output of the receiving local neurons in the form of the timing and frequency of action potentials (also called “spikes”), which are transient electrical signals that carry information between connected neurons (Buzsaki et al., 2012).

Simultaneous measurements of LFPs and action potentials in the cortex showed that action potentials are synchronized to oscillations in the LFP signal (Eckhorn & Obermueller, 1993; Gray, Kînig, Engel, & Singer, 1989; Jacobs, Kahana, Ekstrom, & Fried, 2007; Murthy & Fetz, 1996; Nase, Singer, Monyer, & Engel, 2003). Spikes typically coincide with the negative phase of LFP oscillations, which correspond to the synaptic influx of positively charged ions into the dendrite of a cell. The negative phase generally reflects increased excitatory synaptic input and thus increased probability of spike activity in the postsynaptic cells (Buzsaki et al., 2012). Oscillations of LFPs temporally organize spike probability or spike synchrony between different parts of the brain, as a means to enhance neuronal communication between task-related parts of the brain for the duration of the task performance (Fries, 2005, 2015; Womelsdorf et al. 2007). Neural oscillations in the LFP signal are thus functionally highly relevant as they directly influence neuronal communication. Investigating the influence of respiration on neuronal oscillations thus helps understand the link between respiratory rhythm and neuronal communication in the brain (Heck et al., 2017; Kluger, Balestrieri, Busch, & Gross, 2021; Rojas-Libano, Wimmer Del Solar, Aguilar-Rivera, Montefusco-Siegmund, & Maldonado, 2018).

The temporal organization of action potentials, especially synchrony in spike activity in response to synaptic inputs has been implicated in cognitive processing, which requires the transmission of information through the synchronous firing of action potentials in large groups of neurons (Diesmann, Gewaltig, & Aertsen, 1999; Uhlhaas, Pipa, Lima, Melloni, Neuenschwander & Nikolic, 2009). There is experimental evidence for a link between cognitive functions and synchronized spike activity. In rats, selectively preventing neurons in a brain area from generating spikes by using locally applied toxin results in the elimination of the function of the relevant brain area (e.g., McLaughlin & See, 2003; Wesierska, Dockery, & Fenton, 2005). In primates, prefrontal cortical neurons synchronize their spike firing during the performance of a task requiring a context-dependent decision to either respond to a signal with a movement (“go” condition) or to ignore the signal (“no-go” condition) (Vaadia, Haalman, Abeles, Bergman, Prut & Slovin, 1995).

Importantly for our purposes, oscillatory neocortical activity in the gamma (30–100 Hz) frequency range is tightly linked to the performance of specific cognitive tasks like decision making (Beshel, Kopell, & Kay, 2007; Siegel, Engel, & Donner, 2011), problem solving (Sheth, Sandkuhler, & Bhattacharya, 2009), memory formation (Osipova, Takashima, Oostenveld, Fernandez, Maris & Jensen, 2006; Sederberg et al. 2007; van Vugt, Schulze-Bonhage, Litt, Brandt & Kahana, 2010) and language processing (Babajani-Feremi, Rezaie, Narayana, Choudhri, Fulton & Boop, 2014; Crone, Hao, Hart, Boatman, Lesser & Irizarry, 2001; Towle, Yoon, Castelle, Edgar, Biassou & Frim, 2008). Correspondingly, deficits in the power or coherence of gamma oscillations have been linked to schizophrenia (e.g., Furth, Mastwal, Wang, Buonanno, & Vullhorst, 2013), ASD (e.g., Brown, Gruber, Boucher, Rippon & Brock, 2005), and Alzheimer’s disease (Basar, Emek-Savas, Guntekin & Yener, 2016; Klein, Donoso, Kempter, Schmitz, & Beed, 2016; Wang et al. 2017).

In mice, respiration directly drives delta/theta rhythms of cortical activity via sensory inputs from the olfactory bulb, but the power of gamma oscillations is also modulated with the respiratory phase (Biskamp, Bartos, & Sauer, 2017; Ito et al., 2014). These findings were confirmed in humans, with intracranial electroencephalogram (iEEG) recordings obtained from pharmaco-resistant epilepsy patients who received implants of subdural grid or intracranial depth electrodes (electro-corticogram) (Herrero, Khuvis, Yeagle, Cerf, & Mehta, 2017; Zelano et al., 2016). Zelano et al. (2016) showed respiration-locked slow oscillations in the piriform cortex, hippocampus, and amygdala and that the power of delta, theta and beta oscillations in all three structures were modulated with the phase of respiration. Importantly, these respiratory modulations of neuronal oscillations were linked to nasal breathing and were not observed during oral breathing. Herrero et al. (2017) recorded iEEG activity in 30 cortical and limbic brain areas, including the three areas recorded by Zelano et al. (2016), but also in several frontal and parietal cortical areas while patients were breathing exclusively through the nose. Respiration-locked iEEG activity was observed in all areas and, using an attention to breath task, the authors could show that the influence of respiration on brain activity increased with attention, reflected in increased breath iEEG coherence. Respiratory modulation of human brain activity can also be captured non-invasively, as recently shown by Kluger and Gross (2020), who used magnetoencephalography (MEG) to show that beta coherence in the sensorimotor cortex is modulated with the respiratory phase. In the same study, the authors also showed that conscious control of respiration reduced the average beta coherence in the sensorimotor cortex but left the respiration-locked modulation of beta power intact.

Overall, experiments in animals and humans show that respiration influences neuronal activity by driving neuronal oscillations that follow the respiratory rhythm and by modulating the power of higher frequency oscillations (including gamma), propelled by respiration-locked sensory input to the cortex from the olfactory system. While more research is needed to confirm this thesis, it is plausible that spike activity and synchrony are a key part of the neuronal mechanism that is responsible for respiratory influence on brain function. Moreover, the full story about the underlying mechanism will likely involve explaining how respiration modulates norepinephrine release.

The respiratory rhythm is generated by brainstem circuits, and brainstem projections carrying respiration-locked rhythmic spike activity have been shown to project to the thalamus. Yackle and colleagues (2017) recently showed that the locus coeruleus (LC), which is the major source of norepinephrine in the mammalian brain (Aston-Jones & Cohen, 2005), receives direct inputs from the mouse preBötzinger complex (preBötC), the primary breathing rhythm generator (Del Negro, Funk, & Feldman, 2018). Importantly, Yackle et al. (2017) showed that the specific subpopulation of neurons in the preBötC that project to the LC regulate the balance between calm and aroused behaviors.

Taken together the above cited studies suggest that respiration rhythmically modulates LC neuronal activity and the excitability of LC neurons. This is of significance for our argument as a recent study by Zerbi and colleagues (Zerbi et al.2019) showed that the stimulation of LC results in norepinephrine release in the entire forebrain and powerfully changes the functional connectome, i.e. the spatiotemporal pattern of synchronous activity between forebrain structures. Neurocomputational work has further demonstrated that baseline modulation of LC output regulates overall neural excitability or gain, dynamically modulating the overall topology of the functional connectome (Aston-Jones & Cohen, 2005; Eldar, Cohen, & Niv, 2013; Warren et al. 2016). The notion that respiration might modulate cortical excitability, possibly via the LC activation received experimental support from a recent study form by Kluger and colleagues (2021), in which the authors employed MEG to measure cortical activity during a near-threshold sensory detection task while also monitoring respiration. This study found that respiration modulated the power of posterior alpha oscillations, which is considered a reliable indicator of cortical excitability. Thus, the projections from the preBötC to the LC as described by Yackle and colleagues (2017) might represent a key neuronal substrate for respiratory modulation of cortical excitability and the functional connectome.

3. Breathing and cognitive function: disorders and possible links to respiration

The direct influence of respiration on brain activity provides potentially new perspectives on somatic, cognitive, and affective disorders that have been associated with breathing. Here we review literature on the role of respiration in COPD, anxiety (panic attacks) and ASD.

We will, however, not discuss links between respiration and motor function which have been demonstrated in both humans (e.g., Kluger & Gross, 2020; Li & Rymer, 2011; Rassler, Ebert, Waurick, & Junghans, 1996) and rodents (e.g., Cao, Roy, Sachdev & Heck, 2012; Welker, 1964), unless the findings are directly relevant to cognitive function.

3.1. Chronic obstructive pulmonary disease

In healthy populations, lung function and cognitive function are closely associated, but offering support for a causal link is difficult because lung function is also a marker of physical activity, which may itself explain the association (Sachdev et al. 2006). The picture is clearer when it comes to the strong correlation between decline in cognitive functioning (reaching from mild impairment to dementia) and COPD (Liao, Ho, Ko, & Li, 2015; Torres-Sanchez, Rodriguez-Alzueta, Cabrera-Martos, Lopez-Torres, Moreno-Ramirez & Valenza, 2015; Villeneuve et al. 2012; von Siemens, Perneczky, Vogelmeier, Behr, Kauffmann-Guerrero & Alter, 2019), a progressive condition characterized by persistent respiratory symptoms and airflow limitation (Pauwels et al., 2001). The exact relationship between cognitive impairment and COPD remains unclear as COPD is correlated with a number of variables (e.g., inflammation, smoking, higher age, reduced physical activity) and comorbidities like diabetes (Arvanitakis, Wilson, Bienias, Evans & Bennett, 2004), sleep apnea (Gosselin, Baril, Osorio, Kaminska, & Carrier, 2019), and depression (Snowden, Atkins, Steinman, Bell, Bryant & Copeland, 2015) which can affect cognitive function and the development of cognitive impairment.

The pathophysiological mechanism most commonly proposed for cognitive dysfunction in COPD is neuronal change or a change in oxygen-dependent enzymes, both mediated through hypoxemia (reduced partial pressure of oxygen in the blood) and hypoxia (reduced level of tissue oxygenation) (Dodd, Getov, & Jones, 2010; Fix, Golden, Daughton, Kass, & Bell, 1982; Grant, Heaton, McSweeny, Adams, & Timms, 1982; Prigatano, Parsons, Wright, Levin, & Hawryluk, 1983). However, these studies do not permit clear conclusions. For example, there are inconsistencies as to whether patients with early disease and mild hypoxemia have significantly impaired cognition, as cognitive impairment has also been found in non-hypoxemic patients (Dodd et al., 2010; Favalli, Miozzo, Cossi, & Marengoni, 2008; Grant et al., 1982; Liesker et al. 2004). Current evidence suggests that hypoxemia alone is not enough to fully account for the cognitive deficits seen in COPD (Dodd et al., 2010).

Interestingly, recent studies using fMRI to explore spontaneous resting-state brain activity in patients with COPD found functional dysconnectivity (Xin et al. 2019) and abnormal synchronization of low-frequency oscillations between brain regions (Yu et al. 2020).

3.2. Panic attacks

There are several lines of evidence linking panic disorder and breathing, most prominently the observation that hyperventilation is a common event during panic attacks (Sinha, Papp, & Gorman, 2000). Studies using panic-inducing manipulations such as lactate injections have shown that induced panic attacks are associated with hyperventilation (Gorman et al. 1986; Liebowitz et al. 1985). Conversely, hypoventilation therapy was shown to reduce panic in patients (Meuret, Ritz, Wilhelm, Roth, & Rosenfield, 2018). While the mechanisms underlying panic disorder are still unknown, two prominent theories focus on PCO₂ levels and PCO₂ sensitivity. Klein (1993) suggests a ‘hypersensitive suffocation alarm system’ detecting rising PCO₂ levels which triggers panic attacks. Alternatively, Ley (1985) points to the panicogenic effects of hyperventilation (hypocapnia) and resultant dyspnea, which in turn leads to more hyperventilation and creates a cycle that leads to a fully developed panic attack. Consequently, the majority of studies have focused on blood oxygen/PCO₂ in the context of breathing (e.g., Goossens et al. 2014; Meuret, Rosenfield, Hofmann, Suvak, & Roth, 2009). To the best of our knowledge, the link between respiration and brain activity in panic disorders have not yet been systematically considered. That said, a study by Meuret and colleagues (2011), however, provides some evidence of a possible link between “resting” breathing patterns and susceptibility to panic attacks, which may be linked to the influence of respiration on brain activity reviewed here.

In their study, Meuret et al. (2011) investigated the triggers of unexpected panic attacks (i.e. panic attacks not thought to be attributed to any situational cue or internal trigger) in people with panic disorders. Experimental sessions involved 24-hour monitoring and the study focused on physiological data during the one-hour time period preceding a panic attack onset until 10 min after. While the minutes leading up to the onset of the panic attack were characterized by a decrease in tidal volume followed by rapid PCO₂ increase, the onset of the attack was characterized by heart rate and tidal volume increases and a decline in PCO₂. The data revealed increased variability in a number of autonomic and respiratory variables that preceded the panic attack onset by as much as 47 min.

Related findings come from a study by Caldirola and colleagues (2004) who compared baseline breathing in panic disorder patients and healthy controls. Participants were asked to breath normally while their breathing activity was recorded for 20 min and to rate their anxiety levels on a 0–100 scale immediately before, halfway through, and immediately after the recording session. Analysis showed that participants with panic disorders had significantly higher approximate entropy indexes than healthy participants for several breathing-related parameters (respiratory rate, tidal volume, minute ventilation, inspiratory drive [tidal volume/time in inspiration], end-tidal PCO₂, and oxygen sensitivity).

These data and results from other studies (e.g., Schwartz, Goetz, Klein, Endicott, & Gorman, 1996) reveal a correlation between panic disorder and increased variability in respiratory activity, which has been suggested as a precursor to rather than a consequence of panic disorders (Coryell, Fyer, Pine, Martinez & Arndt, 2001; Perna, Ieva, Caldirola, Bertani, & Bellodi, 2002). While the source of this greater entropy in panic disorder patients remains unclear, the findings may indicate a link between abnormalities in “baseline” respiration patterns and panic disorders. On the basis of what we now know about the modulation of brain activity through respiration, increased variability in the respiratory rhythm translates to increased variability in rhythmic neuronal activity, including activity in limbic structures (Zelano et al., 2016), and may thus contribute to the development of a panic attack.

EEG measurements of rhythmic neuronal activity in the frontal cortex during exposure to anxiety-triggering stimuli revealed increased power in the beta frequency band in patients compared to healthy control participants (de Carvalho, Velasques, Freire, Cagy, Marques & Teixeira, 2015). De Carvalho and colleagues argue that beta band oscillations are linked to top-down evaluation of stimuli and that the increased beta power is indicative of an imbalance of top-down processing of anxiety related stimuli. Recent findings show that respiration modulates beta band oscillations in the forebrain (Lockmann, Laplagne, & Tort, 2018; Melnychuk, Robertson, Plini, & Dockree, 2021), suggesting that the above-described differences in beta power between patients and controls might be linked to differences in respiration. Changes in oscillation power alone, however, are not necessarily indicative of changes in function, which ultimately requires that neuronal spiking activity is affected. Importantly, Cenier and colleagues (2009) have shown that the phase-locking of neuronal spiking activity to beta oscillations is modulated by respiration, resulting in a respiration-locked gating mechanism that controls neuronal information transmission. Converging evidence thus suggests that the link between panic disorder and respiration may extend beyond PCO2 and include respiratory influence of brain activity.

3.3. Autism spectrum disorders

There are currently no studies that have attempted to directly investigate potential differences in the influence of respiration on brain activity in individuals with ASD compared to neurotypical individuals. We included ASD in this review, however, because several studies of respiratory behavior have identified differences between ASD and neurotypical individuals, such as an atypical nasal cycle (Dane & Balci, 2007) and irregular respiratory patterns known as Biot’s breathing and Cheyne-Stokes Respiration (Ming, Patel, Kang, Chokroverty, & Julu, 2016). While none of the above studies have investigated brain-breath coupling directly, it is interesting to revisit their findings in the light of our new understanding of the influence of respiration on brain activity and function. In this section we review studies of spontaneous respiratory activity in patients with ASD.

Dane and Balci (2007) investigated cerebral lateralization in children with ASD by assessing hand, eye, and nasal dominance in those with ASD compared to neurotypical age-matched peers. They were particularly interested in the nasal cycle, which consists of alternating periods of decongestion in one nostril with increased congestion in the opposite nostril (Hasegawa & Kern, 1977). Nasal cycle dominance typically lasts anywhere between 25 min and eight hours with a peak around 1.5—4 h (Dane & Balci, 2007). The nasal cycle and the ultradian rhythm of cerebral hemisphere dominance, which lasts 1.5–3 h, are tightly coupled and linked to lateralized changes in brain activity (Werntz, Bickford, Bloom, & Shannahoff-Khalsa, 1983). Broadband EEG activity was found to increase in the hemisphere contralateral to the dominant nostril (Shannahoff-Khalsa, 1991; Werntz et al., 1983). This lateralized influence on brain activity matched improvements in two prominently lateralized brain functions, spatial and language abilities (Backon, Matamoros, & Ticho, 1989; Jella & Shannahoff-Khalsa, 1993). Left-sided unilateral forced breathing has been associated with sympathectomy and enhanced spatial abilities while right sided unilateral forced breathing has been associated with vagotomy and heightened verbal skills. Children with ASD had increased rates of left handedness, left eyedness, and left nasal dominance (Dane & Balci, 2007). Furthermore, the nasal cycle for children with ASD was atypical, lasting most of the day; this may suggest that the right hemisphere is continuously exposed to stimulation.

Recent studies by Ming, Julu, Brimacombe, Connor, & Daniels, (2005; Ming et al. 2016) linked differences in respiratory patterns between neurotypical children and children with ASD to autonomic dysfunction, potentially uncontrolled sympathetic activity, triggering hyperarousal. Compared to typically developing age-matched peers, children with ASD have significantly higher heart rate, lower cardiovagal tone, and cardiac sensitivity to baroreceptors (Ming et al., 2016). Moreover, lower cardiovagal tone is associated with significantly higher percentages of apnea, Biot’s breathing, and Cheyne-Stokes Respiration. Biot’s breathing is characterized by abrupt apnea followed by sudden regular or fast breathing, while Cheyne-Stokes Respiration (CSR) is characterized by periodic respiration interrupted by central apnea, where breath gradually increases in depth and decays back to apnea (Wijdicks, 2007). Children with ASD that had higher cardiovagal tone generally had more similar breathing patterns to neurotypical controls (Ming et al., 2016). The findings of this study show a relationship between cardiac arrythmias due to autonomic dysfunction and altered respiratory rhythm, and the authors hypothesize that the observed irregularities are due to brain stem dysfunction.

Cognitive aspects of ASD have also been linked to aberrant LC function, primarily an increase in tonic LC activity with decreased phasic responses to sensory stimuli (Bast, Poustka, & Freitag, 2018). Huang et al. (2021) analyzed the functional connectivity of the LC in boys with ASD using fMRIs. They found decreased connectivity between the LC and brain regions associated with attention networks and areas critical to theory-of-mind. Other studies exploring connectivity show decreased power and phase locking in the alpha and gamma frequency ranges between brain regions involved in sensory processing, a key element of attention (Murphy, Foxe, Peters, & Molholm, 2014; Simon & Wallace, 2016; Sun, Grutzner, Bolte, Wibral, Tozman & Schlitt, 2012).

To the best of our knowledge detailed investigations of brain-breath coupling and of respiratory influence on brain activity in individuals with ASD have not yet been performed. Based on published findings about the respiratory influence on brain activity and function, it seems likely that the observed breathing dysrhythmias in patients with ASD also produce dysrhythmias in oscillatory brain activity, disrupting the normal respiration-locked rhythmic modulation of neuronal oscillations associated with motor and cognitive function.

4. The role of interoception

Respiratory interoception, the ability to perceive, metacognitively monitor, and control breathing-related sensations, is also a key psychological mechanism interlinking breathing behavior to the experience of anxiety, panic, and dyspnea (Benke, Hamm, & Pane-Farre, 2017; Giardino et al. 2010; Nikolova, Harrison, Toohey, Brændholt, Legrand & Correa, 2021). In the cardiac domain, the magnitude of heartbeat-evoked cortical potentials has been shown repeatedly to relate to interoceptive sensitivity (Banellis & Cruse, 2020; Montoya, Schandry, & Muller, 1993). An interesting and thus far untested hypothesis then is that there may be bidirectional modulation between the kinds of broadband neural-respiratory coupling described here, and individual differences in respiratory interoception. In this case, one would expect the intra and inter-individual variance in respiratory modulated brain oscillations, and/or respiratory-evoked cortical potentials, to correlate positively with objective measures of respiratory interoception. As increased sensitivity to respiratory sensations is associated with negative affective outcomes, e.g., in asthma or COPD patients (Dahme, Richter, & Mass, 1996; Giardino et al., 2010), testing this hypothesis may have important implications for treatment of respiratory-related psychopathology, and could guide interventions targeting the remediation of deficits through guided interoceptive feedback or targeted exposure therapy (Barrera, Grubbs, Kunik & Teng, 2014). Further, understanding the link between respiration-brain coupling and respiratory interoception may help to better understand the computational mechanisms interlinking psychological expectations, affective outcomes, and interoception (Allen, 2020; Allen, Varga, & Heck, 2021; Nikolova et al. 2021). Finally, alterations in interoception are linked to a variety of psychiatric outcomes including ASD, anxiety, and depression, implying a third link by which alterations in respiratory-brain coupling may influence mental health (Khalsa et al. 2018; Khalsa & Lapidus, 2016; Owens, Allen, Ondobaka & Friston, 2018).

5. Conclusion

An increasing number of studies have indicated that respiration exerts a significant influence on sensory, affective, and cognitive processes, and that such influence occurs via modulation of neural oscillations and their synchronization between brain areas. The first part of the article reviewed recent findings supporting these theses, while the second part focused on how they could ameliorate our comprehension of certain disorders associated with irregular respiratory patterns. Based on our review of the literature on respiration in COPD, anxiety (panic attacks), and ASD, the insights into respiratory modulation of neuronal activity may provide a better comprehension of the link between respiratory irregularities and cognitive and affective deficits.

Research into breath-brain interaction is still in its infancy, but in light of the evidence reviewed above, it seems safe to maintain that future investigations of brain activity and function could benefit from taking respiration as a modulating factor into account. Furthermore, while published studies from this emerging field have mostly focused on the instantaneous, cycle-by-cycle relationship between respiratory phase and brain activity, respiratory habits could have lasting consequences for brain activity and brain function. The studies on patients with panic disorder and ASD show clear differences in baseline respiratory behavior to healthy controls. So besides factoring in respiration as a modulating influence, it seems promising to explore long term respiratory behavior, both to assess its potential as a biomarker indicating vulnerability but also to examine the potential of interventions that target baseline respiratory patterns.