Harnessing brain rhythms to activate the glymphatic pathway via controlled breathing and intermittent CO₂

Optimal functioning of the glymphatic pathway is essential for maintaining brain health and preventing neurodegenerative diseases.^1,2 This specialized waste clearance system in the brain plays a critical role in removing metabolic byproducts and excess proteins such as α-synuclein, amyloid beta (Aβ), and tau—substances that, when accumulated, are linked to conditions such as Parkinson’s and Alzheimer’s diseases.^1,3 While evidence suggests that the glymphatic pathway is most active during sleep, enhancing or supporting glymphatic activity through lifestyle or therapeutic interventions could offer a powerful strategy for preserving brain function, improving cognition, and reducing the risk of age-related and neurogenerative cognitive decline.

Breath-centered practices such as yoga, tai chi, and qigong have been utilized for centuries to manage stress and promote psychological well-being.^4–6 Emerging research reveals an additional benefit: these practices also stimulate cerebrospinal fluid (CSF) oscillations, potentially activating the glymphatic pathway during wakefulness. CSF oscillations occur through two primary mechanisms when an individual engages in controlled breathing practices: mechanical thoracic pressure changes^7–10 and slow rhythmic CO₂-driven vasomotion (infra-slow rhythmic constriction and dilation of arteries and arterioles). In addition to highlighting the potential for controlled breathing practices to facilitate waste clearance, these findings raise the intriguing possibility that rhythmic, controlled CO₂ exposure may serve as a direct and potentially more effective modulator of glymphatic function. This review synthesizes the current understanding of factors that drive the glymphatic system, highlights how controlled breathing and intermittent CO₂ exposure activates more impactful drivers of the glymphatic pathway, and discusses key considerations for optimizing these interventions to enhance the brain’s housekeeping system.

The brain relies on the flow of CSF through the interstitial space to clear metabolic waste and toxic proteins from the brain (Figure 1).^1,11,12 Perivascular spaces, CSF-filled spaces surrounding the arteries/arterioles and veins/venules that penetrate into the parenchyma, serve as low-resistance conduits, enabling the rapid flow of CSF into deeper brain regions. The transport of CSF from the peri-arterial spaces into the brain interstitial space is referred to as influx, which is driven by oscillatory drivers, such as vasomotion, and mediated by aquaporin 4 (AQP4) water channels on the astrocytic end feet that surround the perivascular space.¹³ The transport of CSF/interstitial fluid out of brain tissue is referred to as clearance or efflux.^14,15 Egress refers to the transport of CSF, extracellular fluid, solutes, and cells from the central nervous system, ultimately into the systemic circulation via multiple pathways.^15,16

Transfer of ISF/CSF/solutes to CSF compartments or directly to egress sites represents efflux—an area of active research with multiple potential pathways. The glymphatic model suggests transport into the perivenous spaces,¹³ though others have suggested alternate pathways such as transport to periarterial spaces (intramural periarterial drainage pathway).¹⁷ Fluid also flows through the spaces between fiber tracts in white matter, subependymal flow beneath the ependymal membrane lining the cerebral ventricles, and perineuronal pathways along the cranial and spinal nerves.^13,18 Egress sites include lymphatic vessels at various locations (e.g. nasal, orbital, dural, parasagittal) and arachnoid projections.^15,16

Compared with wakefulness, the glymphatic pathway is most active during nonrapid eye movement (NREM) sleep, a period during which CSF inflow through the periarterial and parenchymal spaces is greater.¹⁹ This enhanced activity during sleep is linked to a 60% expansion in extracellular volume in the brain, which likely reduces flow resistance and promotes the greater glymphatic influx observed during this state.¹⁹ Given the key role of CSF flow in this system, there are increasing efforts to understand the factors that can drive influx, efflux, and egress to potentially activate this system during wakefulness.

Rhythmic movement of the vascular wall due to vasodilation and vasoconstriction is a common mechanism that drives CSF flow, whether driven by cardiac, respiratory, or vasomotor drivers. The intracranial vault contains the brain, which accounts for 80% of its volume—comprising 66% of the intracellular and 14% of the extracellular compartments. Blood makes up 10% of the space and is distributed among the arterial, capillary, and venous compartments. CSF accounts for 10% of the total volume, with only 20% located within the ventricles and the majority (80%) located in the cranial and spinal subarachnoid spaces.²⁰ Rasmussen et al. described the drivers of CSF flow as a system of extracranial and intracranial coupled oscillators that facilitate rhythmic vasoconstriction and vasodilation through a variety of mechanisms on the basis of initial work by Stefanovska.^15,21 We discuss each of these briefly to provide background information for our subsequent discussion, although we recommend Rasmussen et al.’s¹⁵ review for a more comprehensive discussion.

Extracranial oscillators include cardiac and respiratory cycles.²² Arterial pulsations drive CSF-ISF exchange (influx).^23–26 Mestre et al.²⁷ provided additional quantitative evidence that arterial pulsations drive CSF flow through the perivascular spaces, referred to as “perivascular pumping.” Using particle tracking velocimetry in live mice, CSF flow velocities in perivascular spaces closely match the velocities of the pulsing arterial walls that form the inner boundaries of these spaces. This synchronous motion suggests that the displacement of the arterial wall propels CSF through the perivascular spaces, facilitating the efflux of interstitial solutes. The authors note that CSF flow velocity is driven primarily by the cardiac cycle, with a smaller contribution from the respiratory cycle. Together, this suggests that cardiac pulsations drive CSF flow during spontaneous or normal breathing in wakefulness, but respiration still contributes a measurable effect on CSF flow in the perivascular spaces.¹⁰ Controlled breathing techniques have been shown to amplify respiratory-driven effects on both CSF (primarily measured within the ventricles) and venous circulation, exceeding those associated with cardiac pulsations, potentially more closely aligning this respiratory patterns that occur during sleep.²⁸ We discuss this in detail below.

Intracranial oscillators consist of neural and vascular oscillators. Neural oscillations are widely studied and driven by phasic activity within and between neural circuits. Neural activity can cause cells to swell and shrink, potentially driving extracellular volume changes that could cause flow changes.²⁹ Additionally, neurovascular coupling, the dynamic response of cerebrovascular vessels to increases in energy demands following neural activation (functional hyperemia), results in robust vasoconstriction and vasodilation.^30,31 Unlike extracranial oscillators, functional hyperemia is dependent on external stimuli and is selectively activated only in the brain regions necessary for processing a given stimulus. During sleep, neural oscillations occur due to dynamic interactions across the brain, which vary across sleep stages.³² During NREM sleep, low-frequency neural oscillations (delta frequencies of 0.5–4 Hz) are followed by hemodynamic oscillations, which are coupled to CSF flow in the infraslow frequency range (0.05 Hz).³³ Given the intricate interplay between key regulators of sleep architecture, such as hypothalamic and midbrain structures, including the locus coeruleus, and neural oscillatory dynamics, future research is warranted to elucidate the role of neural oscillators and the influence of phenomena such as microarousals during sleep.³⁴

Infraslow vasomotor oscillations (<0.1 Hz) have emerged as key modulators of glymphatic function.³⁵ Van Veluw et al.³⁶ demonstrated that rhythmic 0.1 Hz vasomotion is present during wakefulness, contributes to a reduction in dextran tracer in the perivascular spaces (suggesting influx into the interstitial spaces), and can be enhanced by visual stimulation in awake head-fixed mice. Specifically, the authors used in vivo two-photon microscopy and reported that rhythmic oscillations in vascular smooth muscle cells at ~0.1 Hz were present in arterioles, but not in venules. They observed a 10%–15% diameter change in the pial arteries and arterioles that was not observed in veins or venules. Bojarskaite et al.³⁷ used two-photon imaging of naturally sleeping male mice to demonstrate sleep cycle-dependent vascular dynamics of pial arteries and penetrating arterioles, whereby slow, large-amplitude oscillations were observed during NREM sleep. The authors modeled this to demonstrate that these perivascular dynamics enhance fluid flow in the perivascular spaces.

Hauglund et al.³⁸ provided further support that vasomotion is important not only for CSF tracer influx, but also for clearance, emphasizing the role of norepinephrine as a key mechanism. Using a combination of optogenetics, fiber photometry, and electroencephalography, the researchers identified synchronized oscillations in norepinephrine, cerebral blood volume (CBV), and CSF as the strongest predictors of glymphatic clearance during NREM sleep. Specifically, norepinephrine is released from the locus coeruleus in an oscillatory fashion at ~0.02 Hz during NREM sleep.³⁹ As norepinephrine is a potent vasoconstrictor, norepinephrine oscillations result in oscillatory vasodilation and vasoconstriction. The authors stimulated the locus coeruleus and reported an increase in norepinephrine and a decrease in blood albumin, suggesting reduced cerebral blood flow (CBF). Miniscope imaging demonstrated that the artery diameter, but not the vein diameter, was inversely correlated with norepinephrine concentration. Following the administration of the tracer into the cisterna magna, the authors demonstrated that the CSF tracer signal during NREM sleep fluctuated inversely with CBF. The increase in the CSF tracer signal was abolished when the mice were treated with medications to facilitate panadrenergic inhibition, demonstrating that norepinephrine mediates these CSF dynamics via its effects on vasomotion. Notably, stimulation of arterial oscillations enhanced CSF inflow, indicating that vasomotion acts as a pump driving CSF into the brain independent of neural activation.

Collectively, these findings highlight that whether driven by cardiac or other oscillators, perivascular pumping is an important underlying mechanism driving influx. Infraslow oscillations, particularly those driven by norepinephrine dynamics, may be the most important drivers of directional flow of CSF and interstitial fluids, promoting the removal of waste. Clearance can occur in the absence of neural oscillations, although neural oscillations are dynamically coupled with vasomotion. Finally, during spontaneous breathing, cardiac pulsations primarily drive CSF flow in perivascular spaces, though respiration still contributes to this process.

Controlled breathing techniques, integral to practices such as yoga,^40,41 tai chi,^42,43 and qigong, have been shown to modulate autonomic nervous system activity^44,45 with the intention of decreasing anxiety and promoting psychological well-being. In addition to these benefits, these techniques may also activate the glymphatic pathway. Since respiration slows and becomes more regular during NREM sleep, it is possible that controlled breathing techniques may mimic this pattern and potentially activate the glymphatic mechanisms that drive waste clearance. Specifically, as individuals transition from wakefulness to NREM sleep, respiratory oscillations progressively slow down and become more regular.^46,47 This transition is accompanied by a shift in the control of respiration from behavioral influences associated with wakefulness to primarily metabolic factors, notably, hypoxic and hypercapnic ventilatory drives.^48,49 During NREM sleep, respiratory-related oscillations in the magnetic resonance encephalography signal within the 0.11–0.44 Hz frequency range are greater than those during wakefulness, suggesting an increase in the influence of respiration on CBF dynamics.⁵⁰

Controlled breathing can amplify respiratory-driven large-scale CSF and venous flow. To use consistent terminology across studies, we use spontaneous breathing to describe normal breathing, deep breathing to describe self-paced reduction in the respiration rate, which is often accompanied by a larger respiratory amplitude, paced breathing to describe timed, directed inspirations and expirations (typically, paradigms include multiple cycles of 2.5–4 s of inspiration followed by 2.5–4 s of expiration, also referred to as forced breathing), and breath-holds (inspiration followed by a period of breath-hold). Dreha-Kulaczewski et al.⁷ evaluated CSF flow responses in the ventricles and aqueduct to (1) paced breathing (eight cycles of 2.5 s inspiration followed by 2.5 s expiration each) and (2) breath-holds for a period of 12 s. Using real-time phase-contrast MRI (RT-PCMRI), they demonstrated increased CSF flow primarily during paced breathing and noted that only a minor flow component could be ascribed to cardiac pulsation. Notably, the design used a 12-s breath-hold and highlighted an attenuated and prolonged CSF response to breath-hold relative to paced breathing. Dreha-Kulaczewski et al.⁸ extended these findings and demonstrated that the CSF flows from the spine toward the head during the initial phase of inspiration measured in the cervical and upper thoracic spine and aqueduct.

Yildiz et al.¹⁰ employed RT-PCMRI to measure CSF velocity and flow in the foramen magnum in healthy subjects under various conditions, including spontaneous breathing, deep breathing, breath-holding (12 s), and coughing. This study further revealed that the magnitude of the CSF velocity was greater during deep abdominal breathing than during natural breathing. A more recent study by Yildiz et al.⁵¹ investigated how different yogic breathing techniques influence CSF flow in healthy adults. Using RT-PCMRI at the foramen magnum, the authors measured CSF velocities during multiple deep breathing techniques, such as slow breathing, deep abdominal breathing (expanding the lower abdomen during inhalation), deep diaphragmatic breathing (the use of the diaphragm to draw air deeply into the lungs), and deep chest breathing (focusing on expanding the chest during inhalation), in 18 participants. All of these breathing techniques led to increases in cranially directed CSF velocities relative to spontaneous breathing, although deep abdominal breathing exhibited the greatest effects.

Kollmeier et al.²⁸ evaluated the coupling between CSF and venous fluid systems, and how these systems differ during spontaneous breathing and paced breathing (eight cycles of 2.5 s forced inspiration and 2.5 s forced expiration). The authors used RT-PCMRI to measure CSF flow in the aqueduct, spinal cervical subarachnoid space, and lumbar subarachnoid space, as well as venous flow in the extracranial epidural vein, internal jugular vein, and inferior vena cava. Paced breathing shifted the dominant spectral components of both CSF and venous flow toward respiratory frequencies and led to a correlation between CSF and venous flow within the major vessels. Laganà et al.⁵² evaluated the extracranial internal jugular veins and the intracranial superior sagittal sinus flow rates and demonstrated an overall reduction in venous flow rate in paced breathing relative to spontaneous breathing overall, but showed an increase in respiratory modulations during forced breathing and paced deep breathing relative to cardiac conditions. Other groups have begun to use RT-PCMRI to measure CSF and cerebral arterial flow.⁹ Liu et al.⁵³ recently demonstrated that changes in CBV driven by respiratory and cardiac activity are strongly inversely coupled with CSF oscillations. Deep breathing (self-paced with no requirements for specific breathing rates or depth) reduced the total cerebral arterial flow rate by 29%. The contributions of cardiac and breathing activities to CBV and CSF volume differed on the basis of breathing pattern.

Thoracic pressure dynamics due to controlled breathing practices are thought to be the primary modulators of CSF oscillations.^7,9,10 Specifically, inspiration lowers intrathoracic venous pressure, enhancing cerebral venous outflow from the brain and reducing intracranial venous volume,^7,9,10 which was empirically demonstrated by Kollmeier et al.²⁸ This transient drop in intracranial pressure generates a cranially directed pressure gradient, propelling CSF upward from the fourth ventricle to the cerebral aqueduct and onward to the third and lateral ventricles (Figure 2).⁸ The opposite occurs in response to expirations, whereby an increase in thoracic pressure increases venous volume in the brain and results in CSF outflow. In line with this mechanistic framework, the reviewed studies demonstrate that deep abdominal breathing—associated with the greatest thoracic pressure fluctuations—elicits the most pronounced changes in CSF flow, though it is unclear whether this translates to greater flow into the perivascular spaces and influx into the interstitial spaces. How CSF oscillations due to thoracic pressure changes influence efflux and egress has yet to be determined.

This highlights several important questions. Does an increase in venous return increase the efflux of interstitial fluid from the brain, effectively having the reverse effect of perivascular pumping? In other words, does a decrease in venous volume increase the perivenous volume that facilitates the flow of fluid from the interstitial spaces into the perivenous spaces? Do the oscillations in venous and CSF flow driven by these mechanisms increase the transport of waste to egress sites? While animal studies can directly measure tracers in the perivascular spaces, human studies primarily rely on measurement of the large-scale fluid flow in the arachnoid spaces, ventricles, and aqueduct due to methodological limitations. While evidence suggests these signals are correlated in NREM sleep,³³ direct evaluation is necessary.

Together, consistent evidence indicates that breathing techniques can increase large-scale CSF flow with varying magnitudes, depending on the technique. While the influence of thoracic pressure changes on venous return explains the immediate CSF response to respiration, it does not account for the delayed increase in CSF flow observed during breath-hold techniques, which is likely driven by CO₂-mediated mechanisms. In the next section, we highlight how these practices result in greater oscillatory CO₂ changes. As CO₂ is a potent driver of vasodilation and vasoconstriction,⁵⁴ we discuss how changes in CO₂ may be an important mechanism by which respiration may impact CSF flow.

Recent investigations that account for the lagged temporal dynamics of CSF responses to various breathing techniques have revealed that the partial pressure of CO₂ in arterial blood (PaCO₂) is a key regulator of cerebrovascular activity (vasomotion) and CSF flow.^55,56 In a pivotal study, Wang et al.⁵⁶ directly compared the effects of cardiac, spontaneous respiration, and deep inspiratory effects on CSF flow in humans. Their findings revealed that deep inspirations displace CSF volumes an order of magnitude greater than those induced by either cardiac or spontaneous respiratory rhythms. The participants were instructed to inhale and exhale over 4-s intervals followed by a period of 50 s until the next deep inspiration, a breathing pattern that produced a marked increase in gray matter signal (reflective of cerebrovascular activity) followed by a pronounced increase in displaced CSF volumes. Critically, the authors observed a delayed peak in CSF flow velocity—~10.4 s following the end of deep inspiration—which is similar to the delays observed in CO₂-induced hypercapnia experiments, suggesting a CO₂-mediated vasoactive mechanism rather than an immediate mechanical effect of thoracic pressure changes.

PaCO₂ is among the most powerful physiological modulators of CBF: for every 1 mmHg increase above resting levels, the CBF increases by ~6%–8%. Conversely, for every 1 mmHg decrease, the CBF decreases by 3%–4%—effects that are consistent until the physiological limits of vasoconstrictive and vasodilatory capacity are reached.⁵⁷ Deep inspirations initially produce a transient decrease in PaCO₂ of ~1–2 mmHg, followed by a rebound increase.⁵⁸ Paced breathing, depending on frequency, can lead to a reduction in PaCO₂ and hypocapnia. In contrast, during breath-holding, PaCO₂ increases rapidly as endogenous CO₂ production continues in the absence of ventilation. Empirical studies have shown that PaCO₂ typically increases by 2–4 mmHg within the first 10 s of a breath-hold, with a continued rise of ~2–3 mmHg every subsequent 10 s, although this rate is modulated by individual differences in metabolic rate, lung volume, and buffering capacity.^59,60 Together, these findings highlight that different controlled breathing techniques lead to different patterns of PaCO₂ concentrations and that the greater the change in CO₂, the greater the change in CBF.

In a 2024 investigation, Nair et al.⁵⁵ conducted a direct comparison of two respiratory paradigms—paced breathing and breath-hold—during MRI BOLD imaging, which facilitated the measurement of CBF and CSF inflow and outflow dynamics. The paced breathing protocol consisted of a 3-s inhalation followed by a 3-s exhalation, which was sustained for durations of 18–48 s and interleaved with 15-s intervals of spontaneous breathing. In contrast, the breath-hold condition involved 20-s breath-holds followed by 15 s of normal respiration. Notably, the breath-hold paradigm elicited a significant increase in biphasic CSF velocities relative to spontaneous breathing, whereas the paced breathing condition produced only modest, statistically nonsignificant enhancements. These findings suggest a more central role of CO₂-mediated vasomotion in modulating CSF dynamics, as breath-holds are associated with more pronounced fluctuations in PaCO₂ levels relative to paced breathing, although this was not objectively assessed. Furthermore, when the authors compared these findings to the CSF changes that occur during NREM sleep from their earlier study,⁶¹ they reported that respiratory challenges induced larger mean amplitude fluctuations in the biphasic CSF velocities, underscoring the physiological impact of respiratory modulation on CSF flow.

A model to explain the coupling between CBF and CSF was proposed by Yang et al.⁶² and emphasizes the Monro–Kellie doctrine.⁶³ Specifically, the model indicates that vasodilation and vasoconstriction will exert force on the walls of ventricles, driving CSF in and out of the ventricles, particularly at the level of the fourth ventricle and central canal. Certainly, a majority of the CSF inflow and outflow at the level of the central canal and fourth ventricle are related to the overall CBV entering and exiting the brain in line with the Monro–Kellie doctrine, which states that volume changes in any intracranial component (blood, brain tissue, CSF) should be counterbalanced by a co-occurring opposite change to maintain intracranial pressure within the fixed volume of the cranium.

Here, we clarify the importance of distinguishing the mechanisms that contribute to CSF flow in the ventricles/arachnoid spaces and the CSF that flows into the perivascular and interstitial spaces. The traditional understanding of CSF physiology suggests that CSF exhibits unidirectional flow.⁶⁴ It is produced primarily by the choroid plexus and is secreted into the lateral ventricles as an electrolyte-rich solution, with the water content being transported mainly through aquaporin-1 water channels. CSF moves caudally from the lateral ventricles to the third and fourth ventricles. CSF then enters the subarachnoid spaces of the brain and spine along with the central canal of the spinal cord. In the traditional model, CSF is then absorbed by the arachnoid villi.

However, it is increasingly appreciated that CSF fluid dynamics are much more complex, in line with our above discussion of the different oscillators that can drive CSF flow and influx. As noted, CSF can reach the parenchyma of the brain by flowing from the cerebral subarachnoid space into the perivascular space, previously referred to as Virchow Robin spaces.^65,66 While some interstitial fluid may be created by the blood-brain barrier,⁶⁷ the evidence discussed above indicates that the activity of cerebral arteries, corresponding to localized vasodilation alternating with vasoconstriction, provides a pressure gradient that acts as a pump for facilitating the flow of CSF along the network of perivascular spaces.²⁷ Therefore, in the absence of oscillatory vasodilation and vasoconstriction, the extent to which CSF is driven into the perivascular and interstitial spaces, which are central to driving the glymphatic pathway, is unclear. Here, we suggest that controlled breathing techniques that lead to fluctuations in CO₂ are critical for eliciting vasomotion, which drives CSF flow into the perivascular and interstitial spaces (Figure 2).

The changes in CBF in response to CO₂ are primarily due to pH-dependent modulation of vascular smooth muscle tone, although complex bidirectional interactions with the autonomic nervous system and noradrenergic signaling further influence cerebrovascular dynamics.^57,68 For example, locus coeruleus neurons drive hypercapnic ventilatory responses and regular normal inspiratory activity.^69,70 Conversely, elevated CO₂ leads the locus coeruleus to release norepinephrine through pH-dependent mechanisms, whereby chemosensitive locus coeruleus neurons increase the firing rate in response to rising levels of CO₂, facilitating the release of norepinephrine from the locus coeruleus.⁷¹ These findings highlight that oscillatory changes in PaCO₂ levels induced by controlled breathing can modulate vasodilation and vasoconstriction both directly—through pH changes and their impact on vascular vessels—and indirectly via activation of the noradrenergic system. Importantly, administering breathing paradigms whereby the cycles occur ~every 50 s (<0.1 Hz oscillations) has the potential to activate infraslow vasomotion, akin to mechanisms observed during NREM sleep.

Considering the central role of vasomotion in driving perivascular CSF influx and clearance,^27,36–38 these findings raise the possibility that respiratory-driven CO₂ oscillations can mimic or augment endogenous vasomotor patterns characteristic of NREM sleep, thereby promoting CSF flow into the perivascular and interstitial spaces.

The above focuses on the role of increased CO₂ in response to controlled breathing techniques. However, directly manipulating CO₂ by exposing individuals to increased concentrations of CO₂ may be a more direct way of activating vasomotion and corresponding CSF flow. Intermittent CO₂ exposure is routinely used to evaluate cerebrovascular reactivity.^72,73 Participants are asked to breathe gas mixtures of elevated concentrations of CO₂ to elicit an increase in CBF in ON-OFF cycles. This approach quantifies the “responsiveness” of the cerebral vasculature, with the intention of serving as an index of cerebrovascular health.

These paradigms increase CBF and CO₂-driven changes in CBF are coupled with CSF inflow, which is consistent with what has been demonstrated in resting-state data collected during wakefulness and sleep.^33,61,62 A recent retrospective analysis by van der Voort et al.⁷⁴ demonstrated that hypercapnia, but not hyperoxic challenges, lead to an increase in CBF, and that CBF is coupled with CSF flow. The design included two hypercapnic blocks (90 s with a +10 mmHg and a 120-s hypercapnic ramp up to a maximum of +12 mmHg) and one hyperoxic block (180-s hyperoxic, targeting an inspired oxygen partial pressure of 680 mmHg). As expected, both hypercapnia blocks resulted in an increase in the CBF. As the study only measured CSF inflow, a large increase in CSF flow was observed following the reduction in CBF.

Given the importance of infraslow oscillatory drivers of perivascular pumping forces,²⁷ we conducted a study to evaluate whether rhythmic exposure to CO₂, which we refer to as intermittent hypercapnia, can elicit vasomotion-induced CSF flow. For the CO₂ exposure, we chose 35 s ON and OFF block lengths because we aimed to mimic infraslow wave oscillations, which result in vasomotion at a frequency of ~0.014 Hz. As anticipated, each CO₂ exposure elicited an infraslow large-amplitude BOLD signal oscillation (reflecting vasomotion) followed by a large-amplitude CSF inflow signal that was highly coupled with the BOLD time course.⁷⁵ This work suggests that inducing oscillatory infraslow vasomotion can increase CSF flow in humans. A limitation of this approach is that CSF is measured at the level of the fourth ventricle, and it is unclear whether vasomotion also drives CSF flow through the perivascular and interstitial space. To evaluate this, we tested whether intermittent hypercapnia can increase the clearance of brain-derived proteins and peptides into the bloodstream. Our initial study conducted in five patients with PD and five HCs who underwent ~30 min of intermittent hypercapnia demonstrated an increase in the plasma levels of proteins and peptides implicated in neurodegeneration. These include total α-synuclein, neurofilament light (NfL), glial fibrillary acidic protein (GFAP), amyloid β_1–42 (Aβ_1–42), amyloid β_1–40 (Aβ_1–40), and phosphorylated tau (pTau217) – proteins and peptides that are implicated in neurodegenerative conditions.⁷⁵ Together, these findings suggest that intermittent CO₂ not only leads to an increase in vasomotion-driven CSF flow, but also enhances the clearance of neurodegeneration-associated proteins and peptides, providing evidence that CO₂-driven physiology may drive perivascular pumping and waste removal from the brain. Notably, prolonged periods of hypercapnia (continuously elevated levels of PaCO₂ without intermittent exposure) impair both glymphatic influx and cervical lymphatic efflux of solutes, highlighting the crucial importance of rhythmic CO₂ exposure to elicit CSF flow and clearance.⁷⁶

Above, we have discussed how infraslow, rhythmic vasoconstriction and vasodilation are the key mechanisms that drive CSF influx and clearance mechanisms. Thus, the goal of future work using controlled breathing techniques or intermittent CO₂ exposure should evaluate approaches on the basis of the current understanding of these mechanisms. First, thoracic pressure changes clearly increase CSF flow largely in the ventricular and arachnoid spaces, but it is possible that these changes may not correspond to CSF flow through the perivascular and interstitial spaces. The same is true for CO₂-mediated CSF changes; however, given that intermittent CO₂ elicits vasomotion, it is more likely to drive CSF into the perivascular and interstitial spaces through perivascular pumping. We hypothesize that breath-hold or intermittent CO₂ paradigms whereby the cycles occur ~every 50 s (~0.02 Hz oscillations) may stimulate infraslow vasomotion, similar to the processes seen during NREM sleep. However, it is important to consider and further evaluate whether the venous return mechanisms driven by thoracic pressure changes may facilitate efflux and egress. Deep inspirations exhibit the greatest impact on large-scale CSF flow, potentially facilitating influx as well as efflux (decreased venous return increases perivenous volume and facilitates the flow of fluid from the interstitial spaces into the perivenous spaces), though this has not been evaluated. As such, future research is necessary to evaluate the extent to which different controlled breathing approaches and intermittent CO₂ exposure lead to increased flow into the perivascular and interstitial spaces (influx). Additionally, evaluation of the degree of efflux and egress in response to controlled breathing and intermittent CO₂ approaches is necessary to more definitively demonstrate the efficacy of these approaches in clearing brain products implicated in neurodegeneration.

Herein, we have suggested that the glymphatic pathway is “most active” during sleep, though this is based primarily on the evaluation of influx. Recent evidence indicates that efflux to lymph nodes is enhanced during wakefulness.⁷⁷ There are likely inherent limitations during wakefulness that may constrain the ability of interventions to activate the glymphatic pathway due to circadian fluctuations. AQP4 channels facilitate the influx of fluid into the interstitial fluid during sleep—AQP4 polarization is thought to be influenced by circadian rhythms, with peak polarization occurring during NREM sleep,⁷⁷ suggesting the efficacy of perivascular pumping during wakefulness may be limited by reduced AQP4 polarization during wakefulness. However, recent work suggests that circadian differences in glymphatic function are mediated by factors other than AQP4.⁷⁸

No study has directly evaluated how controlled breathing or intermittent CO₂ influences AQP4 channels. Increases in AQP4-channel permeability occur in response to lower intracellular pH, with no significant effects of extracellular pH.⁷⁹ CO₂ may pass through AQP4 channels and influence intracellular pH,⁸⁰ however, the pH changes required to impact AQP4 channels⁷⁹ are significantly greater than what is observed even in severe clinical hypercapnia.⁸¹ This suggests that the effect of intermittent hypercapnia as tested is unlikely to induce a significant change in AQP4 permeability. Relatedly, direct measurement of extracellular volume changes due to controlled breathing or intermittent CO₂ paradigms can help determine whether there is an increase in influx of CSF fluid into the interstitial space. Evaluation of this is limited, complicated by the increase in CBF in response to CO₂,⁸² and prior studies have relied on prolonged CO₂ exposure that does not facilitate perivascular pumping.⁷⁶ Since the interstitial volume of the brain increases going from awake to sleep state,⁸³ careful assessments of extracellular space in response to breathing paradigms and intermittent CO₂ and their comparison to sleep remain an important area of future research.

Sympathetic tone exhibits circadian rhythms whereby increases in norepinephrine activity occur during wakefulness.⁸⁴ Prior work demonstrating that inhibition of adrenergic signaling increases interstitial volume fraction suggests that elevated norepinephrine may disrupt influx.¹⁹ Notably, slow breathing activates the parasympathetic nervous system and downregulates sympathetic outflow and norepinephrine activity.⁸⁵ Additionally, we have already highlighted that infraslow oscillatory norepinephrine drives vasomotion, influx, and clearance during sleep, suggesting that rhythmic norepinephrine activity in response to CO₂ may enable vasomotion equivalent to that observed during sleep.³⁸ It is important to determine whether the degree of efficacy of CO₂-induced vasomotion can be equivalent to that observed during sleep, given the potentially higher sympathetic activity during wakefulness.

A separate line of research has begun to evaluate whether functional hyperemia can be used to increase vasomotion, and whether this may enhance glymphatic clearance during wakefulness.^36,86,87 Some studies have demonstrated that functional hyperemia leads to vasomotion that increases CSF flow and influx during wakefulness,^36,87 although others were unable to observe this effect.⁸⁶ Importantly, functional hyperemia selectively activates regions relevant for processing a given stimulus, whereas controlled breathing and intermittent CO₂ paradigms can elicit vasomotion across the brain. Therefore, we posit that the latter approaches may be more efficacious in activating the glymphatic pathway. An additional challenge in the efficacy of such approaches is that neuronal activity reduces extracellular space volume due to cellular swelling and ion shifts, potentially reducing the ability of fluid to effectively flow though the interstitial space. Future work is necessary to compare these approaches to determine which may be most effective in activating the glymphatic pathway during wakefulness.

There are multiple benefits of breathing practices relative to direct manipulation of CO₂. They are convenient and can be readily used by any individual in almost any setting. reath-holds in conjunction with deep diaphragmatic breathing may have combined benefits of CO₂-mediated vasomotion and thoracic pressure changes. While breathing practices are convenient and can be readily used by any individual, there are potential challenges if they need to be applied for regular or extended practices to elicit meaningful effects. Unlike breathing practices, the CO₂ paradigm is relatively passive and thus may be more feasible for use. Breathhold not only induces hypercapnia (elevated CO₂), but can also lead to reduced oxygen (O₂) levels.⁵⁹ While the commonly-used brief breath-holds are unlikely to result in hypoxia, there is greater discomfort associated with decreased O₂. Intermittent CO₂ paradigms typically hold oxygen levels constant (often relying on gas mixtures that include ~5% CO₂ and a mixture of gases that approximate room air). As such, CO₂ exposure with consistent oxygen saturation may facilitate therapeutic effects without the potential negative consequences of reduced oxygen that occurs during breath-holds. Future research is necessary to identify the optimal combination of controlled breathing techniques and intermittent CO₂ exposures.

We want to thank the patients and their families who participated in our research studies.