Dynamic coupling between the central and autonomic nervous systems during sleep: A review

https://doi.org/10.1016/j.neubiorev.2018.03.027Get rights and content

Highlights

  • Central (CNS) and autonomic (ANS) nervous systems are tightly coupled during sleep.

  • Complex regulatory mechanisms (CNS commands and ANS reflexes) drive the coupling.

  • Analytic methods reveal a complex interplay between cortico-cardiac dynamics.

  • Analysis of coupling between CNS events and ANS changes gives physiological insight.

  • Changes in ANS-CNS coupling might be useful as a biomarker of disease.

Abstract

Sleep is characterized by coordinated cortical and cardiac oscillations reflecting communication between the central (CNS) and autonomic (ANS) nervous systems. Here, we review fluctuations in ANS activity in association with CNS-defined sleep stages and cycles, and with phasic cortical events during sleep (e.g., arousals, K-complexes). Recent novel analytic methods reveal a dynamic organization of integrated physiological networks during sleep and indicate how multiple factors (e.g., sleep structure, age, sleep disorders) affect “CNS-ANS coupling”. However, these data are mostly correlational and there is a lack of clarity of the underlying physiology, making it challenging to interpret causality and direction of coupling. Experimental manipulations (e.g., evoking K-complexes or arousals) provide information on the precise temporal sequence of cortical-cardiac activity, and are useful for investigating physiological pathways underlying CNS-ANS coupling. With the emergence of new analytical approaches and a renewed interest in ANS and CNS communication during sleep, future work may reveal novel insights into sleep and cardiovascular interactions during health and disease, in which coupling could be adversely impacted.

Introduction

Sleep is a complex physiological process involving multiple biological systems, fundamental for the health of an individual. It is typically defined as a central nervous system (CNS) phenomenon in which the state of cortical activation determines the individuals’ state of consciousness. Sleep regulatory systems are closely associated, anatomically and physiologically, with the autonomic nervous system (ANS), as described in this review. The ANS regulates the majority of the body’s internal processes (e.g., blood pressure, myocardial function, breathing, body temperature, digestion, urination) via afferent and efferent sympathetic and parasympathetic pathways, allowing adaptive responses to internal and external stressors, guaranteeing the body’s homeostatic milieu.

A growing body of research has investigated ANS function during sleep, mainly focusing on cardiovascular (CV) functioning, using measures derived from heart rate such as heart rate variability (HRV) analysis of the electrocardiography (ECG) signal. Importantly, measures reflecting ANS function fluctuate across a night of sleep, depending on both homeostatic (i.e., related to the duration of sleep and of prior wakefulness) and circadian processes, and in association with phasic events during sleep. These interactions, which reflect the dynamic interplay between the CNS and the ANS, are the focus of the current review.

Over recent years, a number of studies have explored the association between the CNS and the ANS during sleep using various approaches and rationales, while offering promising new perspectives to understand the brain-body interplay during sleep in humans. Here, we review CNS-ANS coupling during sleep, consider the methodologies for measuring sleep and ANS activity (Sections 1.1), identify limitations in assessing ANS function indirectly in humans (Section 1.1), and discuss the underlying physiological basis for the presence of coupling between the two systems (Section 1.2). We then review the body of work investigating ANS-CNS coupling during sleep, ranging from observations of changes in heart rate, HRV, and blood pressure across sleep stages (Section 2) to an analysis of the temporal dynamics between outputs of the two systems (Sections 3–4). Finally, we discuss clinical implications for the interdependency between the CNS and ANS during sleep and conclude with suggestions for future research directions (Sections 5 and 6).

For the purpose of this review, we will consider the relationship between CNS indices and ANS measures as part of the concept of “CNS-ANS coupling”. We conceptualize different levels of evidence and methods (arbitrarily defined for clarity of presentation) supporting CNS-ANS coupling during sleep, which forms the framework for this review:

  • Observations. For example, changes in ANS control of cardiovascular effectors, such as changes in cardiac vagal modulation, according to CNS-defined sleep stages.

  • ANS and/or CNS mediated perturbations. For example, heart rate and blood pressure fluctuations associated with, or in response to, CNS-defined phasic sleep events including spontaneous cortical arousals and slow waves.

  • Analytics of continuous CNS-ANS dynamics, which integrate continuous, simultaneous measures of both ANS and CNS function to explore their interconnections. For example, continuous linear and non-linear relationships between time-series of cortical and cardiac function, considering strength, time-course, and directionality.

Polysomnography (PSG) is the gold-standard method to objectively assess sleep using multiple sources of information. The characterization of sleep-wake states (i.e., wakefulness, the stages N1, N2, and N3 of non-rapid-eye-movement [NREM] sleep, and rapid-eye-movement [REM] sleep) is fully described in the scoring manual of the American Academy of Sleep Medicine (AASM) (see Berry et al., 2017). This characterization follows standardized rules based on visual detection and discrimination of electroencephalography (EEG), electromyography, and electrooculography features within fixed discrete time windows (30 s epochs) across the night. EEG is typically sampled at ≥200 Hz to fully capture high frequency activity.

Sleep can be broadly divided into two main states characterized by specific features and EEG patterns. In humans, REM sleep is characterized by a “wake-like” low-amplitude high frequency EEG activity, irregular and sharp eye movements, and low muscle tone. NREM sleep, which constitutes between 75% and 80% of the total sleep period, is characterized by progressive EEG synchronization. It begins with low amplitude and mixed frequency EEG activity with a predominant theta (4–7 Hz) rhythm in NREM sleep stage N1. This is followed by NREM sleep stage N2, characterized by the appearance of spontaneous K-complexes (the largest event in the human EEG, consisting of a positive-negative-positive waveform ≥0.5 s) and sleep spindles (bursts of distinct 11–16 Hz waves, ≥ 0.5 s). There is then a predominance of high-voltage slow wave oscillations (1–4 Hz) in NREM sleep stage N3. Sleep shows a strong ultradian rhythm such that REM and NREM alternate with a cycle length of ∼90 min across the night. There is predominance of N3 sleep during cycles in the first part of the night with a progressive increase in time spent in REM sleep across the night (Kryger et al., 2015). Importantly, both sleep states (NREM and REM), as well as phasic sleep events (e.g., arousals, K-complexes) are accompanied by distinct patterns of tonic and phasic ANS activation (see Section 1.2).

Conventional sleep staging offers an overall picture of a night’s sleep, and can be clinically useful in assessing sleep quality and composition. However, it is limited by being visually determined, arbitrary, and providing discrete characterization of sleep. The current scoring guidelines (Berry et al., 2017) are essentially similar to those codified 50 years ago by a committee led by Rechtschaffen and Kales (Kales and Rechtschaffen, 1968) for use with analog data recorded on pen chart recorders. These criteria were, in turn, largely based on visual scoring systems developed 80 years ago (Loomis et al., 1937). Importantly, this approach constrains the analysis of all aspects of human physiology (e.g., cardiovascular measures, hormonal variations) within arbitrarily-defined, discrete, and CNS-defined sleep periods.

Automated quantitative EEG analysis has the potential to overcome limitations inherent in visual scoring and provides continuous and detailed information on the state of cortical activation and EEG complexity. Typically, an EEG spectrum is obtained by applying a fast Fourier transform to the signal, after which power spectral density (μV2/Hz) can be calculated for somewhat arbitrarily defined EEG bands: delta (1–4 Hz), theta (4–7 Hz), alpha (8–12 Hz), sigma (12–15 Hz), low- (16–23 Hz) and high- (23–30 Hz) beta, and gamma (>30 Hz) frequencies. Power in each frequency band can then be summarized per epoch, across the night, or within sleep stages and sleep cycles. For example, EEG delta power declines over sequential NREM-REM sleep cycles, a phenomenon interpreted as reflecting a decrease in homeostatic pressure across the night (Dijk, 2009). Several factors including age, sex, menstrual cycle, medications, and presence of insomnia, influence elements of the EEG power spectrum. EEG spectral indices are the principal CNS indices used in studies directly assessing CNS-ANS coupling (see Section 4).

Daytime assessment of ANS function mainly relies on classic ANS tests reflecting the ANS response to standardized maneuvers such as isometric exercises (handgrip), cold pressure test, cognitive load (mental arithmetic), orthostatic testing (including head-up tilt-test), Valsalva maneuver, deep breathing, cold face test of the diving reflex, and more invasive tests such as pharmacological (e.g., intravenous phenylephrine) or mechanical (e.g., neck suction) challenges to stimulate and evaluate baroreflex functioning (see Zygmunt and Stanczyk, 2010). These tests are used in clinical settings to identify ANS abnormalities for the diagnosis and treatment of diseases involving ANS dysfunction.

A number of other techniques, more suitable for research purposes, are available to continuously assess ANS and cardiovascular functioning without active engagement from participants, and can, therefore, be applied during sleep (See Table 1 for a summary of these measures and their limitations). These differ as to the particular property of cardiovascular functioning they assess (e.g., cardiac function such as heart rate, or peripheral sympathetic nerve activity [SNA] using microneurography); whether they measure tonic or acute/transient properties (e.g., average heart rate during NREM sleep, or the heart rate response to a brief arousal from sleep); the ease with which the technique can be applied over prolonged periods of time such as during sleep (e.g., challenges associated with inserted electrodes that need to be kept in position across the night); and the degree of ambiguity in the interpretation of the measure (e.g., the relatively straightforward interpretation of average heart rate, or the more complex interpretation of indices of HRV). Fig. 1 offers an example of co-registration of EEG, ANS and cardiovascular indices during a PSG night of a sample participant at the SRI Human Sleep Research Lab.

The most widely used indicators of ANS activity during sleep are heart rate and HRV. Heart rate may be readily and non-invasively assessed by appropriately placed surface recording electrodes, typically over thoracic regions, producing the ECG. The ECG is an integral component of clinical sleep assessment and, while not being used as a main source of information to determine sleep state, is recommended by the AASM for standard clinical PSG assessments (Berry et al., 2017).

Heart rate (or heart period, which is its reciprocal) has the advantage of enabling assessment of cardiac ANS tone (i.e., average ANS activity) and of transient (beat-to-beat) cardiac modulation, and thus it has been repeatedly applied in studies of CNS-ANS coupling. The available evidence indicates that heart period is linearly related to the frequency of cardiac parasympathetic and sympathetic activity (Berntson et al., 1995). The mean value of heart period is thus a simple and powerful index to assess changes in the balance (i.e., linear sum) of cardiac parasympathetic and sympathetic tone. Since heart rate and heart period are related by a non-linear (hyperbolic) function, it should be noted that they are not equivalent for the purpose of assessing cardiac ANS tone.

HRV quantifies the beat-to-beat variability of heart periods (the time intervals between subsequent R waves on the ECG), generally at frequencies <0.4 Hz, and has been widely used with the purpose of estimating specific components of ANS functioning (Cacioppo et al., 2007). HRV may be analyzed in the time domain or in the frequency domain (in the latter technique by integrating HRV spectral power density, in ms2 Hz−1, over specified frequency bands), within specified time-windows, ideally of 5-min length for the purpose of standardization (Camm et al., 1996), across the sleep period.

HRV spectral power at frequencies below 0.04 Hz (the very low frequency range, VLF) is a measure with dubious interpretation, particularly when assessed from recordings with durations ≤5 min. In contrast, HRV power in the 0.15–0.40 Hz frequency range (high-frequency, HF) reflects respiratory sinus arrhythmia and as a consequence is a useful measure of cardiac vagal modulation (Camm et al., 1996). Cardiac vagal modulation may also be estimated from HRV in the time domain by computing the indices RMSSD, the root mean square of the differences between successive R-R intervals, and pNN50, the percentage of the values of RR-interval that differ from the preceding value by >50 ms (Camm et al., 1996).

HRV spectral power in the low frequency (LF) range (0.04–0.15 Hz) has often been interpreted to reflect sympathetic activity. This interpretation has been particularly controversial because of conceptual shortcomings and because multiple sources contribute to variability in this range (Billman, 2013; Trinder, 2007). Importantly, the sinoatrial node responses to sympathetic and vagal activity indicate that HRV in the LF range may be driven by both sympathetic and vagal modulations (Rossi et al., 2016). In an attempt to provide a solution to this problem, the power ratio LF/HF has been proposed under the assumption that the vagal component contributing to HRV LF power would be at least partially neutralized, such that the ratio would reflect sympathetic modulation or sympatho-vagal balance (Camm et al., 1996). However, the very concept of sympatho-vagal balance has been challenged (Eckberg, 1997) and there is no reason to think that the vagal modulation contributing to LF HRV power is equal to, or from the same source, as that contributing to HF HRV power. Consistent with this, the interpretation of the LF/HF ratio in terms of cardiac sympathetic activity has been recently questioned by direct experimentation on a sheep model of heart failure (Martelli et al., 2014). Nevertheless, even in the absence of a clear-cut physiological interpretation, the LF/HF ratio may be of some value as a descriptor of HRV during sleep.

The LF and HF components of HRV power are often expressed in normalized units (nu), as the % of the total power of HRV. In normalizing LF and HF power, however, it is common practice to remove the VLF component from total power, i.e., to divide LF and HF by the sum (LF + HF) (Camm et al., 1996). As a consequence, the quantities LF/HF, LFnu and HFnu (mathematically identical values) are equivalent carriers of information that reflect the same features of cardiac ANS control. Thus it becomes redundant to report more than one of these variables. This limitation is not substantially lessened if total power of HRV is used as the denominator instead of the sum of LF and HF power, because any difference in information among LF/HF, LFnu, and HFnu will then come from the HRV VLF power, which, as noted above, is hardly interpretable in terms of sympathetic or vagal modulation (Burr, 2007).

If the assumptions of linearity and stationarity of the heart period time series are abandoned, several non-liner methods are available to assess ANS function. For example, the Poincaré plot is a non-linear geometric approach to qualitatively picture and quantitatively assess HRV metrics. It is produced by plotting each RR interval (RRn+1, ms; y axis) against the previous RR interval (RRn, ms; x axis) for a specific time window (e.g., 1 min, 2 min, 5 min). The inter-beat autocorrelation coefficient is a quantitative HRV index derived from the Poincaré plot, and has been frequently used in investigating the dynamics of CNS-ANS coupling (see Section 4). It is calculated as the Pearson product-moment correlation coefficient between RRn+1 and RRn+1. The inter-beat autocorrelation coefficient is highly correlated with the frequency-domain HRV LF/HF index during night-time sleep (Otzenberger et al., 1998), whose physiological interpretation is, as previously discussed, a topic of much debate. Moreover, being a linear measure, the inter-beat autocorrelation index is not sensitive to the nonlinear features displayed by the Poincaré plot (Brennan et al., 2001).

Entropy-derived non-linear indices based on HRV (e.g., detrending fluctuation analysis, approximate entropy, sample entropy, multiscale entropy analyses) have also been used. These and even more sophisticated methods (see Valenza et al., 2017) which measure non-linear HRV features might be better suited to track HRV changes associated with human high-level functioning such as emotional and cognitive processing (see de la Torre-Luque et al., 2016; Young and Benton, 2015). However, their readouts are difficult to interpret in terms of specific anatomic pathways.

Limitations of HRV analysis (see Table 1), are critical when considering the physiological interpretation of the even more complex relationships between the temporal EEG-HRV dynamics, as we will further discuss in Section 4.

Given the lack of a pure sympathetic nervous system measure derived from HRV analysis, sympathetic nerve activity (SNA) has been measured using other techniques. Impedance cardiography allows a noninvasive measure of pre-ejection period, a cardiac ANS index calculated as the time from the electrical systole to the opening of the aortic valve (blood ejection). It is thought to reflect β-adrenergic sympathetic control of the heart (Sherwood et al., 1990). To our knowledge, no study has investigated the temporal dynamics of the relationship between the pre-ejection period and CNS indices (e.g., EEG) during sleep. A different direct measure of (non-cardiac) peripheral SNA is obtained by microneurography, an invasive method measuring the SNA to muscle blood vessels and skin. Despite being used in some sleep studies (see Mano et al., 2006), microneurography cannot be easily adopted in sleep research and is impractical in clinical sleep medicine due to its invasiveness and methodological challenges (e.g., the requirement of immobilization). For example, Tank et al. (2003) recorded muscle SNA from the right peroneal nerve using microneurography together with intra-arterial blood pressure before and after EEG K-complexes. None of the 8 participants studied, were able to reach REM sleep and only 5 were able to reach NREM N3. Skin conductance/resistance can offer a non-invasive continuous measure of electrodermal activity, reflecting cutaneous sudomotor activity and indicating skin sympathetic outflow. The most frequently used method to assess electrodermal activity consist of passing a small current between two surface electrodes usually placed on the palm of the hand. The resulting voltage, when the current is held constant, is proportional to skin resistance (or alternatively, a measure of conductance is obtained using a constant voltage system). Although it is generally accepted that sweat glands have predominant sympathetic cholinergic innervations, neural mechanisms and pathways involved in the central control of electrodermal activity are not fully clarified (Cacioppo et al., 2007). Most of the research on electrodermal activity has been conducted in the context of cognitive and emotional processing, and the application and interpretation of this technique during sleep is questionable. The measurement of circulating catecholamines (plasma epinephrine and norepinephrine concentration) has been also used to evaluate peripheral sympathetic ANS functioning during sleep, although it is invasive, and not suitable to assess CNS-ANS coupling during sleep at high temporal resolution.

In summary, as shown in Table 1, there is no unobtrusive and unconfounded measure of central sympathetic tone which can be used in sleep studies. In light of the methodological constraints of sympathetic assessment and in the face of reliable indices of cardiac vagal modulation, it is surprising that research on CNS-ANS coupling during sleep has over-emphasized the sympathetic over the vagal system.

Other measures reflecting ANS function (however, less frequently used in the context of CNS-ANS coupling) are available and may ultimately further advance the understanding of CNS-ANS dynamics during sleep by providing windows on other aspects of the ANS. These measures include beat-to-beat blood pressure monitoring and others obtainable with photoplethysmographic-type sensors (see Table 1 for details).

The brain is connected to cardiac myocytes and arteriolar smooth muscle cells, which are the CV autonomic effectors, by di-synaptic pathways involving preganglionic and ganglionic autonomic neurons (Silvani et al., 2016). In the afferent direction, changes in blood vessel stretch and skin temperature, which are affected by skin vessel tone, are detected by baroreceptors and thermoreceptors and fed back to the brain (Silvani, 2017; Silvani et al., 2015). Changes in breathing such as those seen when transitioning from wakefulness to sleep (Trinder and Nicholas, 2000), also modulate the ANS both directly, through CNS circuits, and reflexively, for example by changing lung stretch as a function of sleep-related reduction in tidal volume (Colrain et al., 1987). These efferent and afferent pathways constitute the anatomical and functional substrate of the coupling between the CNS and the ANS.

The coexistence of top-down (efferent) and bottom-up (afferent) pathways complicates inferences of causality from observed correlations between CNS and ANS variables (Fig. 2). For example, parasympathetic preganglionic neurons in the medullary nucleus ambiguous are subjected to a respiratory modulation (McAllen and Spyer, 1978) that gates inputs from baroreceptors, chemoreceptors and nasopharyngeal receptors (Gilbey et al., 1984). The cardiac sinoatrial node responds to parasympathetic activity with minimal delay and is responsive to modulations of parasympathetic activity at least up to 0.4 Hz (Berger et al., 1989). As a result, the inspiratory gating of nucleus ambiguous neural activity is one of the mechanisms responsible for respiratory sinus arrhythmia (Eckberg, 2003).

The specific neural pathways that couple CNS to ANS during sleep are still uncharted. It has been suggested (Silvani and Dampney, 2013) that the activities of thalamo-cortical neurons, skeletal muscles, and the ANS, which explain the EEG, electromyographic, and cardiovascular features of the sleep states, respectively, are modulated in parallel by sleep-state control circuits, which are mainly located in the hypothalamus and brainstem (Scammell et al., 2017). This conceptual model predicts that measures of CNS-ANS coupling during sleep, such as correlations between EEG and HRV spectral indexes, do not necessarily imply direct causal links from EEG to HRV. Rather, the CNS-ANS coupling may reflect the common dependence of EEG and HRV variables on the activity of specific hypothalamic and brainstem circuits. The brainstem appears critical in this respect, as key neural elements of autonomic, respiratory, and cardiovascular control are located within the medulla, pons, and midbrain (Silvani and Dampney, 2013), together with key elements of the ascending pathways that control EEG activity through the thalamus and basal forebrain (Scammell et al., 2017).

This conceptual model bears resemblance to that of central pattern generators coordinating skeletal muscle motor neurons, autonomic preganglionic neurons, and neuroendocrine neurons (Swanson, 2011). In the context of physical exercise, the autonomic counterpart of the skeletal muscle activity produced by these central pattern generators is classically termed the central autonomic command (Goodwin et al., 1972). This conceptual model of central autonomic commands in the context of CNS-ANS coupling during sleep provides a framework for future work to examine which sleep-state control circuits modulate cardiovascular activity, how circuits may differ during NREM versus REM sleep as well as during transition periods between NREM and REM sleep, and whether outputs from sleep state circuits to thalamo-cortical and ANS neurons are indeed precisely temporally coupled. In this respect, the coupling between pattern generators should not be viewed as rigid, and may be actually relatively weak, as is thought for the neural oscillators underlying resting brain fluctuations (Deco et al., 2009). Sleep-related central autonomic commands may result from the summation of activity of different central pattern generators (Silvani and Dampney, 2013), either at the same time, as occurs for thermoregulation (Romanovsky, 2007), or at different times, because of the asynchronous development of the sleep process in different neuronal groups (Kim et al., 2017; Pigarev et al., 1997). A summary of the data-driven hypotheses on the similarities and differences in central autonomic control between wakefulness and NREM sleep, and REM sleep is provided in Table 2.

Increases in the activity of the hypothalamic ventrolateral preoptic nucleus may modulate ANS activity during NREM sleep through at least four different pathways (Sherin et al., 1996). The lower values of SNA to the skin during NREM sleep compared to wakefulness (Takeuchi et al., 1994) may result from inhibition of a thermoregulatory pathway for heat generation and retention, which includes the preoptic hypothalamus and the medullary raphe (Morrison, 2016). On the other hand, the occurrence of lower values of SNA to the skeletal muscles (Somers et al., 1993) and kidneys (Miki et al., 2003) during NREM sleep may result from (Silvani, 2017; Silvani and Dampney, 2013): a) inhibition of the hypothalamic paraventricular nucleus (Uschakov et al., 2006), a master ANS controller that integrates neuroendocrine, homeostatic and stress responses during wakefulness (Thompson and Swanson, 2003); b) inhibition of the pedunculopontine nucleus between the caudal midbrain and rostral pons (el Mansari et al., 1989), a region potentially contributing to central autonomic commands associated with locomotion during wakefulness (Garcia-Rill et al., 1987; Padley et al., 2007); and c) potentiation of the baroreceptor reflex, through a pathway that involves the pontine parabrachial nucleus (Saito et al., 1977) and the medullary nucleus of the solitary tract (Eguchi and Satoh, 1980). In particular, increases in parabrachial nucleus activity may inhibit the baroreflex at the level of the first afferent relay, which is the nucleus of the solitary tract (Felder and Mifflin, 1988). Therefore, the tonic decrease in parabrachial nucleus activity during NREM sleep may disinhibit the baroreflex in this state. Conversely, transient increases in parabrachial nucleus activity during autonomic arousals in NREM sleep may contribute to simultaneous increases of blood pressure and heart rate because of transient baroreflex inhibition (Silvani et al., 2015; Silvani and Dampney, 2013). These central neural pathways may also explain other features of the baroreflex during NREM sleep, such as the baroreflex resetting toward lower values of heart rate, blood pressure, and SNA, and the greater baroreflex contribution to cardiac control than in wakefulness (Silvani et al., 2008). Taken together, these changes in baroreflex function may represent an important mechanism of the CNS-ANS coupling during sleep.

Neurons of the nucleus of the solitary tract control both cardiac parasympathetic activity, by means of a direct projection to preganglionic parasympathetic neurons in the nucleus ambiguus of the medulla, and cardiac SNA, by means of indirect projections to the rostral ventrolateral medulla (Silvani et al., 2016). At least in mice, the decrease in heart rate from wakefulness to NREM sleep is caused by strikingly balanced increases and decreases in parasympathetic and sympathetic tone to the heart, respectively (Lo Martire et al., 2018). Sleep-related changes in the neural circuit involving the parabrachial nucleus and the nucleus of the solitary tract may thus contribute to drive the decrease in heart rate during NREM sleep. The hypothalamic paraventricular nucleus also receives synaptic projections from the hypothalamic suprachiasmatic nucleus, which control separate populations of pre-sympathetic and pre-parasympathetic neurons (Buijs et al., 2003). The suprachiasmatic nucleus is the master circadian clock responsible for the circadian rhythms of sleep (Easton et al., 2004) and cardiovascular variables (Janssen et al., 1994). Neurons in the suprachiasmatic nucleus decrease their activity during NREM sleep and increase it during REM sleep, suggesting that the link between suprachiasmatic nucleus activity and sleep is, actually, bidirectional (Deboer et al., 2003). Although much critical evidence is still missing, including on the neurochemical features of the suprachiasmatic neurons projecting to the paraventricular nucleus (Buijs et al., 2003), it is conceivable that this circuit plays a role in the integration of circadian and sleep-dependent central autonomic commands (Silvani et al., 2016).

Episodes of REM sleep can be described as tonic or phasic. Tonic REM sleep refers to relatively long period of stable N1-like EEG and muscle atonia with relatively regular heart rate. Tonic REM sleep is interrupted by phasic burst in which eye movements occur and cardiovascular physiology becomes more unstable.

The cardiovascular changes that occur during episodes of REM sleep are largely distinct from those of NREM sleep, but may also be driven by central autonomic commands capable of biasing the arterial baroreflex (Silvani and Dampney, 2013). The neural circuitry responsible for the occurrence and control of REM sleep involves the glutamatergic sublaterodorsal nucleus of the pons as well as multiple GABAergic cell groups in the brainstem and hypothalamus (Luppi et al., 2017). The tonic periods of REM sleep entail disparate changes in SNA, consisting of decreases in renal and mesenteric SNA and simultaneous increases in skeletal muscle SNA (Yoshimoto et al., 2011). Moreover, recent data on mice indicate that the increase in heart rate on passing from NREM sleep to REM sleep is mediated, at least in part, by an increase in cardiac SNA (Lo Martire et al., 2018). The increases in muscle and cardiac SNA during REM sleep may result from direct excitatory projections from the pontine sublaterodorsal nucleus to the pre-sympathetic neurons in the rostral medulla, which project to the sympathetic preganglionic neurons. Moreover, REM sleep reverses the reductions in the activity of the pedunculopontine nucleus and the parabrachial and solitary tract nuclei, which may underlie the reduction in muscle and cardiac SNA during NREM sleep (Silvani and Dampney, 2013). The disparate changes in SNA to the skeletal muscle, cardiac, mesenteric, and renal vascular beds during REM sleep may reflect modulations of pre-sympathetic neurons by neurons of the medullary raphe obscurus and the ventrolateral and lateral regions of the midbrain periaqueductal gray, which include both sympatho-excitatory and sympatho-inhibitory neurons (Futuro-Neto and Coote, 1982; Silvani and Dampney, 2013). Incidentally, the finding that SNA to different cardiovascular effectors undergoes opposing changes at the same time during REM sleep seriously challenges the concept of a unique central sympathetic tone during sleep. In this light, the existence of a global ANS index/measurement that can closely reflect the central ANS control during sleep appears unlikely, especially if the picture is broadened to include cardiac vagal control, the sympathetic control of thermoregulatory effectors such as brown adipose tissue, and the control of breathing, which is tightly linked to that of the cardiovascular system (Eckberg, 2003).

The occurrence of wide and bizarre fluctuations (‘unpredictable fugues’) of blood pressure and heart rate, which may or may not be associated with bursts of rapid eye movements, ponto-geniculo-occipital waves, and transient changes in the breathing pattern, characterizes the phasic periods of REM sleep (Silvani, 2008; Snyder et al., 1964). The medial and inferior vestibular nuclei in the medulla may play a key role in driving the autonomic and neurophysiological variability during phasic REM sleep (Morrison and Pompeiano, 1970). These nuclei may receive sleep-related synaptic input from the hypothalamic ventrolateral preoptic nucleus, and may modulate the activity of the pedunculopontine tegmental nucleus, the pontine parabrachial nucleus, and the medullary nucleus of the solitary tract (Silvani and Dampney, 2013). The centrally-driven changes in breathing pattern may also contribute to baroreflex modulation and cardiovascular variability during phasic REM sleep, at least in part, by modulating barosensitive neurons in the caudal ventrolateral medulla (Mandel and Schreihofer, 2006; Silvani and Dampney, 2013).

Section snippets

Observations: changes in ANS tone and modulation according to CNS-defined sleep stages

As described above, there are multiple top-down and bottom-up pathways that potentially could underlie the observed dynamic coupling between indices reflecting CNS (EEG) and ANS (cardiovascular variables) modulation during NREM and REM sleep and in relation to phasic sleep events within each sleep state. This section broadly describes how indices reflecting ANS tone and modulation, as measured non-invasively in humans, differ between sleep states. Fig. 3 shows the temporal dynamic of heart rate

ANS and/or CNS mediated perturbations: CNS-ANS coupling across phasic sleep events (grand-average and ERP designs)

A clear manifestation of the CNS-ANS interplay during sleep is evident from the reciprocal cortical and autonomic fluctuations during phasic, transient sleep events. Phasic sleep events are mainly considered manifestations of transient cortical synchronization and/or desynchronization. The majority of studies directly investigating cortico-cardiac coupling during transient sleep events have time-aligned the analysis to the EEG-defined phasic sleep events, and considered the peripheral changes

Analytics of continuous CNS-ANS dynamics

Advanced signal processing techniques promise to offer new insights into the relationships underlying the complex dynamic interplay between CNS and ANS processes during sleep. A number of observational studies have focused on the coupling between spontaneous fluctuations of the EEG and either heart rate or HRV indices. The reason for this emphasis is that sleep is principally defined by the EEG, eye movements (measured with the electrooculogram), and muscle activity (measured with

Conclusion and future directions

A growing body of evidence at anatomical, physiological, and behavioral levels is converging in support of a tight coupling between CNS and ANS measures during sleep, showing that a dynamic interplay across multiple bio-systems occurs. Thus, (1) cyclic oscillations in CNS (cortical synchronizations and desynchronizations) and co-occurring changes in peripheral ANS control characterize the “sleep cycles”; (2) on a shorter time scale, phasic cortical synchronizations and desynchronizations are

Practical points

  • CNS and ANS are tightly coupled during sleep, as evident from different levels of analysis and different time-scales

  • The complexity of the indices used in analyzing CNS-ANS coupling (mainly spectral EEG and HRV indices) makes the physiological interpretation of cortico-cardiac dynamics during sleep challenging. For example, in the evaluation of the cortico-cardiac relationship, the physiology underlying the oscillatory rhythms responsible for the HRV phenomenon is frequently ignored. HRV metrics

Funding sources

This study was supported by National Institutes of Health (NIH) grants, AA021696 (to IMC + FCB), HL103688 (to FCB), AA020565 (to IMC), AA024841 (to IMC + MdZ), and National Health and Medical Research Council (NHMRC) of Australia (grant ID: 1027076, to JT). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies.

Disclosure statement

MdZ, FB, and IMC have received research funding unrelated to this work from Fitbit Inc., Ebb Therapeutics Inc., and International Flavors & Fragrances Inc. JT and AS have nothing to disclose.

References (177)

  • L. Ferini-Strambi et al.

    The impact of cyclic alternating pattern on heart rate variability during sleep in healthy young adults

    Clin. Neurophysiol.

    (2000)
  • H.A. Futuro-Neto et al.

    Desynchronized sleep-like pattern of sympathetic activity elicited by electrical stimulation of sites in the brainstem

    Brain Res.

    (1982)
  • E. Garcia-Rill et al.

    Locomotion-inducing sites in the vicinity of the pedunculopontine nucleus

    Brain Res. Bull.

    (1987)
  • B. Greene et al.

    Combination of EEG and ECG for improved automatic neonatal seizure detection

    Clin. Neurophysiol.

    (2007)
  • R. Horner et al.

    Activation of a distinct arousal state immediately after spontaneous awakening from sleep

    Brain Res. Rev.

    (1997)
  • B. Janssen et al.

    Suprachiasmatic lesions eliminate 24-h blood pressure variability in rats

    Physiol. Behav.

    (1994)
  • F. Jurysta et al.

    The impact of chronic primary insomnia on the heart rate--EEG variability link

    Clin. Neurophysiol.

    (2009)
  • F. Jurysta et al.

    Progressive aging does not alter the interaction between autonomic cardiac activity and delta EEG power

    Clin. Neurophysiol.

    (2005)
  • F. Jurysta et al.

    A study of the dynamic interactions between sleep EEG and heart rate variability in healthy young men

    Clin. Neurophysiol.

    (2003)
  • G. Kerkhof et al.

    Absence of endogenous circadian rhythmicity in blood pressure?

    Am. J. Hypertens.

    (1998)
  • M. Kern et al.

    Heart cycle-related effects on event-related potentials, spectral power changes, and connectivity patterns in the human ECoG

    NeuroImage

    (2013)
  • T. Kuo et al.

    EEG beta power and heart rate variability describe the association between cortical and autonomic arousals across sleep

    Auton. Neurosci.

    (2016)
  • H. Abdullah et al.

    Brain and heart interaction during sleep in the healthy and sleep apnoea, biomedical engineering and sciences (IECBES)

    2010 IEEE EMBS Conference on. IEEE

    (2010)
  • H. Abdullah et al.

    Cross-correlation of EEG frequency bands and heart rate variability for sleep apnoea classification

    Med. Biol. Eng. Comput.

    (2010)
  • H. Abdullah et al.

    Phase synchronization and directional coupling between brain and heart during sleep, biomedical engineering and sciences (IECBES)

    2012 IEEE EMBS Conference on. IEEE

    (2012)
  • M. Ako et al.

    Correlation between electroencephalography and heart rate variability during sleep

    Psychiatry Clin. Neurosci.

    (2003)
  • R.P. Bartsch et al.

    Network physiology: how organ systems dynamically interact

    PLoS One

    (2015)
  • A. Bashan et al.

    Network physiology reveals relations between network topology and physiological function

    Nat. Commun.

    (2012)
  • R.D. Berger et al.

    Transfer function analysis of autonomic regulation. I. Canine atrial rate response

    Am. J. Physiol.

    (1989)
  • G. Berntson et al.

    The metrics of cardiac chronotropism: biometric perspectives

    Psychophysiology

    (1995)
  • R. Berry et al.

    AASM scoring manual updates for 2017 (version 2.4)

    J. Clin. Sleep Med.

    (2017)
  • G.E. Billman

    The LF/HF ratio does not accurately measure cardiac sympatho-vagal balance

    Front. Physiol.

    (2013)
  • H. Bonnemeier et al.

    Circadian profile of cardiac autonomic nervous modulation in healthy subjects: differing effects of aging and gender on heart rate variability

    J. Cardiovasc. Electrophysiol.

    (2003)
  • G. Brandenberger et al.

    Sleep stage 2: an electroencephalographic, autonomic, and hormonal duality

    Sleep

    (2005)
  • G. Brandenberger et al.

    Ultradian oscillations in plasma renin activity: their relationships to meals and sleep stages

    J. Clin. Endocrinol. Metab.

    (1985)
  • G. Brandenberger et al.

    Age-related changes in cardiac autonomic control during sleep

    J. Sleep Res.

    (2003)
  • M. Brennan et al.

    Do existing measures of Poincare plot geometry reflect nonlinear features of heart rate variability?

    IEEE Trans. Biomed. Eng.

    (2001)
  • R. Buijs et al.

    The suprachiasmatic nucleus balances sympathetic and parasympathetic output to peripheral organs through separate preautonomic neurons

    J. Comp. Neurol.

    (2003)
  • H. Burgess et al.

    The relationship between slow-wave activity, body temperature, and cardiac activity during nighttime sleep

    Sleep

    (2001)
  • H. Burgess et al.

    Cardiac activity during sleep onset

    Psychophysiology

    (1999)
  • H. Burgess et al.

    Sleep and circadian influences on cardiac autonomic nervous system activity

    Am. J. Physiol.

    (1997)
  • R.L. Burr

    Interpretation of normalized spectral heart rate variability indices in sleep research: a critical review

    Sleep

    (2007)
  • J. Cacioppo et al.

    Autonomic cardiac control. II. Noninvasive indices and basal response as revealed by autonomic blockades

    Psychophysiology

    (1994)
  • J. Cacioppo et al.

    Handbook of Psychophysiology

    (2007)
  • G. Calandra-Buonaura et al.

    Cardiovascular autonomic dysfunctions and sleep disorders

    Sleep Med. Rev.

    (2015)
  • A.J. Camm et al.

    Heart rate variability: standards of measurement, physiological interpretation and clinical use. Task force of the european society of cardiology and the North American society of pacing and electrophysiology

    Circulation

    (1996)
  • J. Carrier et al.

    The effects of age and gender on sleep EEG power spectral density in the middle years of life (ages 20–60 years old

    Psychophysiology

    (2001)
  • M. Carrington et al.

    Changes in cardiovascular function during the sleep onset period in young adults

    J. Appl. Physiol. (1985)

    (2005)
  • M. Carrington et al.

    Blood pressure and heart rate during continuous experimental sleep fragmentation in healthy adults

    Sleep

    (2008)
  • M. Carrington et al.

    The influence of sleep onset on the diurnal variation in cardiac activity and cardiac control

    J. Sleep Res.

    (2003)
  • Cited by (120)

    • Cardiorespiratory changes associated with micro-arousals during naps

      2023, Neurobiology of Sleep and Circadian Rhythms
    • The complexities of the sleep-pain relationship in adolescents: A critical review

      2023, Sleep Medicine Reviews
      Citation Excerpt :

      PSG has the advantage of providing both sleep macro-structure and micro-structure measures, allowing an in-depth characterization of the physiology of sleep. With PSG and associated measures, it is, therefore, possible to attain an integrated perspective of sleep [186] which can be applied to understand links between sleep disturbances beyond simple metrics, including impaired central and ANS coupling during sleep, elevated inflammatory markers, cortical and ANS hyperarousal, all of which may be implicated in altered pain processing. Given the complexity of the sleep-pain relationship within the context of biobehavioral maturation and social changes (see Section 2), additional experimental and analytic methods might be needed to parse these associations.

    • Performance of consumer wearable sleep technology

      2023, Encyclopedia of Sleep and Circadian Rhythms: Volume 1-6, Second Edition
    View all citing articles on Scopus
    View full text