New Neuroscience Tools That Are Identifying the Sleep–Wake Circuit

Abstract

The complexity of the brain is yielding to technology. In the area of sleep neurobiology, conventional neuroscience tools such as lesions, cell recordings, c-Fos, and axon-tracing methodologies have been instrumental in identifying the complex and intermingled populations of sleep- and arousal-promoting neurons that orchestrate and generate wakefulness, NREM, and REM sleep. In the last decade, new technologies such as optogenetics, chemogenetics, and the CRISPR-Cas system have begun to transform how biologists understand the finer details associated with sleep–wake regulation. These additions to the neuroscience toolkit are helping to identify how discrete populations of brain cells function to trigger and shape the timing and transition into and out of different sleep–wake states, and how glia partner with neurons to regulate sleep. Here, we detail how some of the newest technologies are being applied to understand the neural circuits underlying sleep and wake.

Statement of Significance

Conventional neuroscience tools such as lesions, cell recordings, c-Fos, and tract-tracing methodologies have been instrumental in identifying the complex and intermingled populations of sleep- and arousal-promoting neurons that orchestrate and generate wakefulness, NREM, and REM sleep. The challenge is to manipulate only specific elements of the circuit that control the behavior. This is now feasible because of the development of new genetic technologies such as optogenetics, chemogenetics, and the CRISPR-Cas system. These tools have begun to transform how sleep biologists understand the finer details associated with sleep–wake regulation.

INTRODUCTION

New technologies are key in advancing our understanding of how the nervous system controls sleep and wakefulness. For instance, the invention of the electroencephalograph led to the discovery of brain activity and of REM sleep.¹ The discovery of REM sleep forever laid to rest the belief that sleep is a passive phenomenon resulting from lack of sensory stimulation. The EEG machine is still the central device in a sleep laboratory and the gold standard for identifying sleep. Experimental approaches have also evolved because a century ago Bremer, Jouvet, and other researchers used knife-cuts to transect the brain at various levels to identify the nature of sleep.^2,3 These transection studies were eventually replaced by electrolytic lesions of specific brain areas, which narrowed the brain region and further advanced our understanding of how certain brain areas control different sleep states.² Researchers began making cell-specific lesions,^4–6 monitoring cellular electrophysiology,⁷ and measuring neurotransmitters^8–10 to understand how candidate neurochemicals and the activity of restricted cells groups were associated with sleep–wake behaviors. However, with the introduction of c-FOS as a functional neuroanatomical tool,^11,12 it became apparent that identified neurons were embedded within complex and intermingled populations that participate in other behaviors (such as movement), making it difficult to identify how they directly control specific sleep behaviors. More recent work has shown that astroglia are also important in controlling sleep activity and that they function to regulate the activity of the neurons that control sleep.¹³ Given the growing complexity of sleep circuitry, the challenge faced by a new generation of sleep researchers is how to disentangle the distributed neural networks that regulate different behavioral states.

Fortunately, recent advances in molecular biology have created new genetically engineered tools that are allowing researchers to understand sleep control from organismal, regional, and cellular perspectives. Researchers are now able to use these new tools to identify—with greater precision and rigor—the neural circuits that control sleep–wake behavior.¹⁴ For example, candidate cell circuits can be genetically targeted to determine how they communicate with one another to regulate sleep and wakefulness, and specific genes can be transferred (ie, gene therapy) into these circuits to determine if they can correct abnormal sleep behaviors (eg, narcolepsy). This review summarizes how new technologies are being used to dissect the neural circuitry underlying wake, NREM, and REM sleep, along with strengths and limitations of these tools (Table 1 and Figure 1).

Table 1.

Summary of the effects of optogenetic or chemogenetic activation or inhibition of neurons or astrocytes on wakefulness (W), non-REM (NR), or REM sleep (RS).

	Probability				Percent			Length			Number			Inhibition	References
	W-NR	NR-W	NR-RS	RS-W	Wake	NR	RS	W	NR	RS	W	NR	RS
Wake active
BF – Chat-Cre a		↑	↑	↓	↑	↓	nc	↑							53,55
BF – PV-Crea (GABA)		↑			↑										53
BF – Vgat-Crea (GABA)		↑			↑	↓	↓							DREADD inhibition ↑Sleep	56
BF – Vglut2-Crea		↑			↑										53
BF – Vglut2-Crea					nc	nc	nc								56
Hypocretin		↑		↑										Inhibition	49,50,58
														↓W
														↑NR
Histamine (TMN)
DA (VTA)
NE (LC)		↑			↑	↓	nc	↑	↓					Inhibition	50
														↓W
														↑NR
Serotonin (DRN)
Pons-Ach LDT/PPTa (Chat-Cre)			↑		↑	↓	↑			nc			↑		78
Sleep active
BF – SOMA-Cre (GABA)	↑				↓	↑									53
VLPO
MCH-Cre or MCH- promoter driven	↑		↑		↓	↑	↑	↓	nc	nc	↓	↑	↑	Inhibition	41,48,71,73
														nc
Medulla GABA						↑	↑							Inhibition	80,101
														↓W
														↓RS
Astrocytes
Posterior hypothalamus					↓	↑	↑								86

The cells were stimulated using either optogenetic or chemogenetic methods. Ach = acetylcholine; BF = basal forebrain; DA = dopamine; DRN = dorsal raphe nucleus; GABA = gamma aminobutyric acid; LC = locus ceruleus; LDT/PPT = lateral dorsal tegmentum and pedunculopontine tegmentum; MCH = melanin-concentrating hormone; nc = no change; NE = norepinephrine; PV = parvalbumin; SOMA = somatostatin; TMN = tuberomammillary nucleus; Vglut2 = vesicular glutamine transporter 3; VTA = ventral tegmental area; VLPO = ventral lateral preoptic area.

^aNeuronal activity is high in waking and RS compared to NR.

Figure 1.

A schematic figure summarizing the different approaches that can be used to target, dissect, and map-out the circuits underlying waking, non-REM, and REM sleep. The experimental paradigm consists of manipulating a gene(s) of interest in specific cells using Cre-expressing mice or rats (also see Figures 2 and 3). This figure focuses on cells in the hypothalamus and pons because of their significance to sleep neurobiology. A researcher can manipulate the activity of the cells in these regions using either light, drugs (eg, CNO) or both. The optogenetics approach uses specific wavelengths of light to manipulate (activation or inhibition) the activity of genetically targeted cells by directing light onto the cell soma (A), thereby influencing the post-synaptic targets of manipulated cells. Light can also be delivered directly to a specific downstream pathway to specifically manipulate the presynaptic terminals at that target site (B). A combination of axonal tracers in Cre-expressing animals to stimulate phenotype- and projection-specific neurons (C). The chemogenetic approach (also known as DREADDs) manipulates cell activity by applying an inert drug (called CNO) that remotely activates/inactivates cholinergic receptors expressed on candidate cells (D). A combination of optogenetics and DREADDs can be employed to manipulate multiple cells types in a hypothesized circuit (E). The activity of downstream neurons can also be electrophysiologically recorded before, during, and after optogenetic manipulation of their presynaptic targets (F). The activity patterns of individual neurons (and entire neuronal populations) can be optically monitored (using a head-mounted micro-camera) and then correlated with candidate behaviors (eg, non-REM vs. REM sleep) (G). CLARITY allows visualization of an entire neuronal circuit to understand where and what neurons they communicate with (H).

Although cutting-edge tools have been applied to study rest–activity behaviors in a range of animals (eg, flies and worms),^15–18 we do not summarize this research because our goal is to describe advances in mechanisms of mammalian sleep, and how such discoveries have led to a better understanding of causes and treatments of sleep disorders such as narcolepsy.

THE GENETIC APPROACH TO SLEEP DISORDERS

The genetic approach has been pivotal in understanding at least two sleep disorders—narcolepsy and advanced sleep phase syndrome. With respect to narcolepsy, Emmanuel Mignot used forward genetics to identify the genetic mutation in the hypocretin-2 receptor (orexin-2R) in canine narcolepsy.¹⁹ Masashi Yanagisawa’s group in Texas had discovered orexin (hypocretin)²⁰ and to identify the behavior regulated by the orexin gene they created orexin (hypocretin) knockout mice (reverse genetics). They found that the orexin (hypocretin) knockout mice had narcoleptic symptoms.²¹ Mice that lacked the orexin-2R (hypocretin-2) also had narcoleptic symptoms,²² thereby confirming the data from narcoleptic dogs. Because orexin knockout mice had narcolepsy, it was logical to autopsy the brains of human narcoleptics revealing a massive loss of the orexin (hypocretin) neurons.^23,24 Transgenic mice (orexin-ataxin-3) that mimic the neurodegeneration seen in humans also display narcoleptic symptoms.^25,26 The creation of the transgenic mice models that genetically destroy the orexin neurons is a significant advance over previous work that used saporin to nonspecifically destroy the orexin neurons.⁵ Thus, both forward and reverse genetics established the link between hypocretin (orexin) and narcolepsy.

Advanced sleep phase syndrome is the other disorder where both forward and reverse genetics converged to identify the cause of the sleep behavior. This disorder is characterized by very early sleep onset (9 pm) and offset (5 am).²⁷ Individuals with this disorder also have a phase advance in the melatonin rhythm.²⁸ Forward genetics revealed that individuals with this disorder have a mutation in the hPer2 gene.²⁹ The Per2 component is part of the molecular mechanisms regulating circadian timing.³⁰ Reverse genetics established that mice with this mutation also demonstrated a shorter period of sleep–activity cycle.³¹

REM sleep behavior disorder is an example of a sleep disorder that was first identified in animals,³² which then led to identifying humans with the behavior.³³ Genetic studies have not been done, but this disorder is considered to be an early indicator of synucleinopathies.³⁴ Restless legs syndrome is a promising candidate for genetic linkage but clearly defined patients, and mice models are needed.³⁵

CRE-EXPRESSING MICE AND RATS TO IDENTIFY CELLS UNDERLYING SLEEP CONTROL

Although the loss of orexin (hypocretin) has been linked to narcolepsy, it is not clear how the brain circuits become destabilized to cause the abnormal behavior. To manipulate and map specific brain circuits, researchers can use a new tool to easily insert genes of interest into phenotype-specific cells in a discrete brain region. This new tool is based on the Cre-loxP system.^36,37 The Cre gene encodes a protein of 343 amino acids, cycle recombinase (cre), that recognizes two sites, called LoxP (locus of X-over P1) sites, and the cre enzyme will very efficiently recombine essentially any DNA substrates which contain these sites (Figure 2). Mice and rats that express Cre in specific cells and neurons are now available, and they can be used when the objective is to deliver specific genes of interest (eg, channelrhodopsin-2) in specific cells (eg, hypocretin or MCH; Figure 3).

Figure 2 .

Candidate viral constructs are used to express a specific transgene in a defined Cre-driver line. This construct comprises of inverted terminal repeats (ITR), the promoter (eg, elongation factor-1-alpha), and the transgene of interest (eg, ChR2-EYFP) surrounded by a pair of LoxP sites and Lox2722 sited oriented inward, a post-translational regulatory element (WPRE) and a polyadenylation (polyA) signal. When a virus infects cells with Cre recombinase, the enzyme will cause serial recombination at one of the two sets of recognition sites, thereby allowing the promoter to drive transgene expression.

Figure 3.

A schematic outlining the general principles of optogenetics. A candidate Cre-driver line of mice (eg, Cre-hypocretin) enable a Cre-inducible virus to deliver a candidate effector gene (eg, ChR2) to candidate cells (eg, hypocretin) (A). A Cre-inducible virus is stereotaxically injected into a region of interest (B) so that it can infect Cre-expressing cells (C). A fibre-optic probe is stereotaxically placed over the candidate cells (eg, hypocretin) expressing a candidate effector protein (eg, ChR2) (D). A specific wavelength of light (eg, blue light, 470 nm) is used to activate (or inactivate) the Cre-expressing cells harboring at candidate opsin (E–G). Light triggers cell activation by opening Na⁺ channels to induce action potentials (F); however, cells that do not contain Cre will not express opsins and light will have no effect on these cells (E). The principles of chemogenetics are largely the same as described in A–G. However, instead of inserting opsins, mutated cholinergic receptors (ie, Designer Receptors Exclusively Activated by Designer Drugs; DREADDs) are inserted into Cre-expressing cells, and these receptors are activated by a drug (eg, CNO) instead of light.

The specific genes of interest (eg, ChR2) can be inserted into Cre-expressing cells using ready-made viruses (eg, recombinant adenoassociated virus, AAV; Table 2). There are a number of viruses that are currently available, and each virus has strengths and weaknesses.³⁸ However, recombinant adenoassociated viruses (rAAV) are the tool of choice for transferring genes because of their safety and their ability to express the transgene over many months. There are currently 11 AAV serotypes that are available but serotypes differ in their tropisms for neurons (serotypes 5, 8, DJ are better for neurons).³⁹ Nevertheless, expression of the transgene is also guided by phenotype-specific promoters or presence of Cre. In our studies, we have used both rAAV5⁴⁰ and rAAV8⁴¹ to drive expression of ChR2-eYFP or orexin-EGFP into neurons containing melanin-concentrating hormone under the control of the MCH promoter. There has been significant progress in the safety of virally mediated gene transfer, including lentivirus.⁴² A biosafety level-2 is required for working with viruses.

Table 2.

Examples of viral vector constructs that can be used to manipulate or visualize cell activity in Cre-expressing mice or rats.

Experimental
Excitatory	Inhibitory
AAV-EF1a-DIO-hChR2 (H134R)-eYFP(optogenetics)	AAV-EF1a -DIO -eArch3.0-eYFPAAV-EF1a -DIO -eNpHR3-eYFP(optogenetics)
AAV-hSyn-DIO-hM3D(Gq)-mCherry(chemogenetics)	AAV-hSyn-DIO-hM4D(Gi)-mCherry(chemogenetics)
Control
AAV-EF1a -DIO -eYFP (optogenetics or chemogenetics)
Deep brain-imaging
AAV-EF1a-DIO-GCaMP6s

In Cre-driver lines, a specific viral construct is stereotaxically injected into a specific brain region, and Cre recombinase will enable recombination of a candidate gene(s), thus resulting in its expression in a specific cell type.

In a typical experiment, a small quantity (0.1 µl–1 µl; 1 × 10¹³ viral particles/ml) of the virus is microinjected into the brain region that contains Cre-expressing neurons. The experiment can begin approximately 3 weeks later when peak levels of the transgene (eg, ChR2) will be expressed in Cre-expressing cells (Figure 3E). Most gene constructs also express a reporter gene that drives expression of a fluorescent molecule (eg, eYFP) so that researchers can visually show that effector genes (eg, ChR2) are expressed in candidate cells. Classic immunohistochemical methods are then used to show that effector proteins are expressed in the candidate cells (eg, hypocretin), which allows researchers to verify that they have genetically targeted the correct cells of interest. In-vitro electrophysiology data are necessary to demonstrate that the neurons containing the light-sensitive opsin are responsive to light.

THE ABCs OF OPTOGENETICS

Optogenetics (“opto” for optical stimulation and “genetics” for genetically targeted cell types) uses the light-sensitive proteins channelrhodopsin-2 (ChR2) and halorhodopsin (NpHR) to either activate or inhibit (respectively) the activity of neurons in which they are targeted.⁴³ ChR2 is a cation channel cloned from green algae, Chlamydomonas reinhardtii, that opens in response to blue light (~473 nm) and allows Na⁺ to flow into a cell (Figure 3F). Cells expressing ChR2 are depolarized when exposed to blue light (~473 nm; Figure 3G). Halorhodopsin is a light-gated chloride pump present in Natronomonas pharaonis, a single-celled organism that is abundant in brackish water. Cells expressing NpHR are hyperpolarized when exposed to yellow light (~590 nm). The fast temporal kinetics of light-sensitive opsins makes it possible to reliably drive trains of high-frequency action potentials in ChR2-expressing cells or suppresses action potentials in NpHR-expressing neurons. These two types of light-sensitive proteins can be inserted into cells allowing researchers the ability to manipulate the activity of specific cells with millisecond precision (Figure 3G). Thus, by delivering a specific wavelength of light, it is possible to control specific circuits in a neural network. Since cells controlling specific behaviors are intermingled with cells controlling other behaviors, optogenetics empowers researchers to use light to control the activity of specific neurons in a neural network (Figure 3).

The major strength of optogenetics is that it allows researchers to control the activity of genetically targeted cells. In a typical experimental paradigm, the light-sensitive opsin-containing cells are stimulated by delivery of light pulses, and the effects on behavior are determined. The first question that the researcher faces is whether the rate of the optogenetic cell stimulation should mimic the cell’s normal activity pattern(s). For instance, in normal sleep–wake behavior, the rate of activity of the orexin, MCH, and the noradrenergic locus coeruleus neurons is 1 Hz, but they can also discharge for brief periods at higher rates (20 Hz).^44–47 When we began to stimulate the MCH neurons,⁴¹ we started with the 1 Hz rate, but did not observe an effect (Shiromani, unpublished data). Therefore, we increased the stimulation rate and discovered that sleep was induced when the MCH neurons were stimulated at 5 and 10 Hz in mice and rats.^41,48 Similarly, the orexin and LC neurons induce waking only at higher stimulation rates (5 to 30 Hz).^49,50 It is not clear why higher rates are needed to induce the behavior, but it is possible that there is a threshold that needs to be overcome to shift the network from one state of consciousness to another. In these studies, it is necessary to provide in-vitro data that verifies that the opsin-sensitive neurons can faithfully follow higher stimulation rates without degradation of the action potential. Such data would also rule out depolarization block resulting from hyperexcitation.⁵¹

The optogenetic method can also be combined with electrophysiology to identify the pattern of activity of a specific phenotype of neurons. Indeed, the optogenetic method is replacing the juxtacellular electrophysiology method which provided the data on the orexin and MCH neurons.^44,47 The juxtacellular method is exceedingly labor intensive, can sample only a few neurons, and the method requires the animal to be head-restrained. Instead, the optogenetic method can be used to identify the sleep–wake related activity of a specific phenotype of neurons in freely behaving mice and rats. Moreover, in the same experiment, these neurons can then be stimulated to determine whether the sleep–wake behavior is changed. In a typical experiment, an optrode (an optic probe encircled by several fine wires) is used to deliver light pulses and also record the activity of the surrounding neurons. The light pulse will trigger an action potential with a short latency (<5 ms) only in neurons that contain the light-sensitive opsin. There is the possibility that the observed signal is an artifact of an interaction between the light and the metal electrode, a phenomenon discovered by Alexander-Edmond Becquerel in 1839. However, by varying the intensity and duration of the light pulse, it is easy to distinguish a legitimate action potential from an artifact because an action potential has constant amplitude, profile, and duration.⁵² On the other hand, the amplitude, profile, and duration of an artifact signal will increase with the intensity and duration of the light. Having verified that the action potential triggered by the light pulse is genuine, the activity of the opsin-containing neuron is followed during several cycles of wake-NREM-REM sleep to determine the relationship of the activity to sleep. Subsequently, these neurons can be stimulated to determine whether there is a change in sleep–wake. Thus, the combination of electrophysiology with optogenetic stimulation provides direct evidence linking phenotype-specific neurons to behavior. This methodology was used to separate the different phenotype of neurons in the basal forebrain and the role of each in sleep and wake.⁵³

There are some limitations of the optogenetic approach. For example, heat is produced at the target site in response to light stimulation, which could be problematic if the cells under study are heat sensitive (eg, thermo-sensitive sleep-inducing hypothalamic neurons). However, this potential limitation is overcome by simply applying light pulses to cells that do not express light-sensitive opsins. Another potential limitation of optogenetics is that the light may not reach all of the cells containing the light-sensitive opsins, which makes it difficult to determine the actual percentage of neurons stimulated (or inhibited) by the light stimulus. This limitation is less concerning when candidate cells are tightly packed in a small cluster (eg, locus coeruleus), but is more problematic when cells are diffusely distributed across a large area (eg, hypocretin cells).

OPTOGENETICS AND ITS ROLE IN UNDERSTANDING MECHANISMS OF AROUSAL

Optogenetics is revolutionizing how researchers study sleep–wake mechanisms⁴⁹ because it allows them to determine how phenotypically “identified” neuron populations participate in candidate behaviors such as REM sleep control. For instance, in the BF, there are a number of different phenotypes of neurons, but it has been unclear which phenotype(s) is responsible for waking. The BF-cholinergic neurons were prime suspects based on their activity pattern and release of acetylcholine during wake and REM sleep. However, selective deletion of the BF-cholinergic neurons with a neurotoxin, 192-IgG-saporin, does not change daily levels of waking,⁵⁴ suggesting that there are other neuronal phenotypes exerting a more significant effect on waking compared to the BF-cholinergic neurons. The identity of the mystery phenotype has been resolved with optogenetics which has shown that BF-GABA neurons, rather than BF-cholinergic neurons, regulate overall levels of wakefulness. In a recent study, optogenetics was combined with fine-wire electrophysiology (optrodes) to record the rate of activity of phenotype-specific neurons in the basal forebrain of four separate cre-expressing transgenic mice.⁵³ The authors stimulated opsin-containing neurons, and when a neuron was found to be activated by the light pulses (16 or 33Hz, 5 ms), they followed it across several sleep–wake cycles. They recorded 85 identified neurons and determined that the BF-GABA neurons (parvalbumin-Cre) were selectively active during waking.⁵³ Optogenetic stimulation of these neurons increased wake.⁵³ Optogenetic stimulation of the BF-cholinergic neurons in Chat-IRES-Cre mice during NREM sleep hastens onset to arousal.⁵⁵

Another group confirmed the role of the BF-GABA neurons in sleep by activating these neurons using chemogenetics (this technique is discussed in a later section).⁵⁶ It is useful to note that an advantage of the chemogenetic approach is that if the targeted cells are dispersed over a wide area, then the drug is likely to reach all of them. The limitation of chemogenetics is that it lacks temporal control because it is based on a drug binding to a receptor.

Optogenetics has also been used to understand better how hypocretin neurons participate in the control of arousal^49,50,57 (Table 1). In fact, the first published study that used optogenetic methods to examine behavior in intact mice used photic manipulation of hypocretin cells to identify their role in arousal. This landmark paper used a lentivirus to drive ChR2 expression in hypocretin neurons in mice. Although 1 Hz blue-light pulses did not induce wakefulness, 5–30 Hz stimulation triggered arousal from both NREM and REM sleep,⁴⁹ suggesting that hypocretin cells function to promote arousal. Optogenetic inhibition of hypocretin neurons using archaerhodopsin, another inhibitory opsin, supports the idea that hypocretin cells act to enhance wakefulness. One hour of hypocretin neuron inhibition during the normal active period (ie, dark phase) reduced wakefulness by increasing NREM sleep,⁵⁸ again suggesting that one of their biological functions is to promote behavioral arousal.

Subsequent optogenetic experiments were used to suggest that hypocretin cells might produce their arousal-promoting effects by stimulating noradrenergic cells in the locus coeruleus.⁴² Carter et al. drove expression of either ChR2 or NpHR to either activate or inhibit LC cells.⁵⁰ They found that LC stimulation induced rapid transitions from sleep to wakefulness, but found that LC inhibition reduced the length of wake bouts, suggesting that LC cells may function to promote arousal. However, they also found that high-frequency (~20 Hz) LC stimulation caused “behavioral arrests” that resemble cataplexy in narcoleptic mice.^21,59 They suggest that high-frequency LC stimulation results in depletion of noradrenaline, which results in reduced noradrenergic drive to skeletal motoneurons. However, there is the possibility that the LC neurons may have experienced depolarization block as a result of the high-frequency stimulation. LC neurons are silent during cataplexy, and the depolarization block-induced inactivity of LC neurons would have reduced noradrenaline release onto motoneurons associated with cataplexy.^59,60

OPTOGENETICS AND ITS ROLE IN DISSECTING NEURAL CIRCUITS UNDERLYING NREM SLEEP

Optogenetic methods have also been used to determine how sleep-promoting circuits regulate different sleep states. For example, classic electrophysiological studies identified “sleep-active” neurons in the BF (reviewed in the study of Szymusiak et al.⁶¹), but none of them were able to identify the neurochemical identity of these neurons. Using Cre-specific mouse lines Xu et al.,⁵³ were able to record the discharge profiles of the four major cell types in the BF—ie, cholinergic, glutamate, GABA+somatostatin, and GABA+parvalbumin. In somatostatin-Cre mice, they found that GABA+somatostatin neurons were most active during sleep (relative to waking) and that optogenetic activation of these cells increased the transitions from wakefulness to NREM sleep as well as increasing overall amounts of NREM sleep⁵³ (Table 1).

Optogenetic strategies have also been pivotal in documenting the importance of MCH neurons and their role in sleep regulation. MCH neurons are intermingled with hypocretin neurons and project to many of the same targets.⁶² They begin firing during NREM sleep and become most active during REM sleep,⁴⁷ but are largely silent during wakefulness, and MCH levels are highest at sleep onset,^63,64 suggesting that MCH cells could play a role in the initiation of sleep. Also, ICV injection of MCH increases both NREM and REM sleep,⁶⁵ and MCH injection into NREM sleep-promoting areas (eg, VLPO) increases NREM sleep⁶⁶ while its injection into REM sleep-promoting areas (eg, such as the nucleus pontis oralis) increases REM sleep.^67,68

However, because MCH neurons express a variety of neurochemicals (eg, GABA,⁶² cocaine- and amphetamine-regulated transcript,⁶⁹ and nesfatin⁷⁰), it is difficult to determine which one is sleep inducing. But, recent data show that MCH loss either by killing the neurons themselves⁷¹ or by knocking out the MCH gene⁷² results in marked decreases in NREM sleep, suggesting that MCH itself is indeed sleep-promoting.

Optogenetic methods have provided some of the newest information concerning how MCH cells control sleep. In wildtype mice,⁴¹ and rats,⁴⁸ optogenetic activation of MCH neurons (1 min on- 4 min off over 24 h) increased both NREM and REM sleep amounts as well as increasing delta power—an index of sleep intensity. Importantly, sleep was increased at night when these nocturnal rodents are active, which suggests that MCH neurons promote sleep by inhibiting arousal-promoting circuits. Moreover, because sleep was similarly increased in mice and rats it suggests that activation of MCH neurons drives sleep across mammals. In MCH-cre mice, acute activation of MCH neurons during NREM sleep promotes transition REM sleep, but does not lengthen NREM sleep.⁷³ In that same study, optogenetic inhibition of MCH neurons had no effect on REM sleep bouts. Selective neurodegeneration of the MCH neurons (97% ablation via diphtheria toxin) decreases NREM sleep but not REM sleep,²⁶ and with 30% loss there is little effect on sleep.⁷⁴

OPTOGENETICS AND ITS USE IN DISSECTING REM SLEEP CIRCUITRY

Since the pioneering studies of Professor Jouvet, cholinergic neurons in the pons have been considered key players in REM sleep control.⁷⁵ This idea was founded by the fact that cholinergic neurons in the PPT and LDT are active during REM sleep (and waking),⁷⁶ and that their pharmacological activation can induce REM sleep (reviewed in the study of Shiromani et al.⁷⁷). However, recent optogenetic experiments provide equivocal support for their role in REM sleep generation because their activation (during NREM sleep) only triggered REM sleep 33% of the time, and their activation during wakefulness had negligible effects on REM sleep timing.⁷⁸ Also, it was reported that optogenetic activation of cholinergic cells had a minimal impact on wakefulness despite the fact that these neurons are also active during wakefulness. Therefore, it remains unclear how (or if) pontine cholinergic neurons participate in the generation of either wakefulness or REM sleep.

Brain structures in the medial medulla are important in controlling the muscle paralysis that occurs during REM sleep but is not typically associated with regulating REM sleep itself.^34,79 However, recent work suggests that the ventral medulla could play a role in REM sleep control. Weber et al. found that optogenetically activating GABA cells in the ventral medulla increased the probability of REM sleep episodes and found that chemogenetic inactivation (see below for methodological explanation) of these same cells reduced REM sleep,⁸⁰ suggesting that GABA cells in the ventral medulla could play a role in REM sleep control.

OPTOGENETICS AND UNDERSTANDING GLIAL FUNCTION IN SLEEP

Ramon y Cajal found that there was a close association between astrocytes and brain vasculature, and hypothesized that the astrocytes regulated nutrient flow to the brain by contracting and relaxing (in the study of Garcia-Marin et al.⁸¹). New evidence is now revealing that the astrocytes are a key component of a glymphatic system in the CNS that is responsible for waste disposal and distribution of lipids, glucose, and other nutrients.⁸² Especially intriguing is a study in mice which found that the glymphatic system becomes activated during natural sleep (high delta waves) and anesthesia (ketamine and xylazine), but not wake.⁸³ The glymphatic system also removes brain lactate that has accumulated during waking.⁸⁴ These studies represent the first direct evidence that one function of sleep (in particular NREM sleep) is to clear the brain of waste that has accumulated during waking. The limitation of these studies is that they only examined the superficial layers of the cortex and did not specifically identify REM sleep, a time when brain metabolic activity is the same as during wake.

Cajal also hypothesized that astrocytes regulated sleep and waking by extending and retracting their endfeet into the synaptic cleft (in the study of Garcia-Marin et al.⁸¹). Recent evidence⁸⁵ shows that there are differences in glial morphology in the synaptic cleft during sleep versus waking, but it is opposite to what Cajal had hypothesized. The study used a serial block face scanning electron microscope to obtain ultra high resolution images of dendritic spines in layer II of the prefrontal cortex in mice that were naturally asleep (6–8 h of sleep), acute sleep deprivation (6–8 h; gentle handling and novel object), or chronic sleep restriction (4 days with ~70% sleep loss; mix of novel objects and forced locomotion).⁸⁵ The study found that during sleep astrocytic processes are further away from the synaptic cleft (opposite to Cajal’s hypothesis) and are closer to the synaptic cleft during wake (acute or chronic sleep deprived groups). The authors conclude that when the astrocytic processes are close to the synaptic cleft, it facilitates glutamate clearance. It is unclear whether the changes in glia can be attributed specifically to sleep or whether they reflect a consequence of the sleep deprivation procedure.

On the other hand, with optogenetics, it is possible to link glia to sleep directly. Optogenetic stimulation (10 Hz) of the astrocytes in the posterior hypothalamus significantly increases both NREM and REM sleep.⁸⁶ This is the first time that optogenetic activation of non-neural structures has been shown to influence sleep. It is unclear how astroglial stimulation increased sleep, but astrocytes are a major source of ATP, which is converted to extracellular adenosine that inhibits synaptic transmission via adenosine A1 receptors.⁸⁷ Optogenetic stimulation of astrocytes does affect behavior^88,89 via ATP and adenosine. With respect to sleep, adenosine A1R agonists potently stimulate sleep, whereas antagonists decrease sleep.⁹⁰ Genetically engineered mice with reduced adenosine levels (expression of dominant negative SNARE selectively within astrocytes) do not have a sleep rebound after sleep deprivation.¹³ On the other hand, a conditionally induced knockdown of glial adenosine kinase increases extracellular concentrations of adenosine and potentiates the homeostatic response to sleep deprivation.⁹¹ Microinjection of adenosine in the rat posterior hypothalamus, the site of optogenetic stimulation of astrocytes in the Pelluru et al. paper,⁸⁶ increases sleep⁹² and this has been linked to a reduction in activity of the arousal hypocretin neurons.⁹³ Glial rather than neuronal adenosine may regulate the homeostatic sleep drive response to prior waking.⁹¹ Thus, during waking when metabolic demand is high, the astrocytes are activated and release adenosine, which then influences local neuronal activity.

These recent studies emphasize the importance of glia in CNS function and suggest that current models of sleep–wake regulation need to be re-evaluated. We consider the current models to be neuron-centric since they only consist of interaction between neurons, and that sleep and wake emerge in response to activation of specific neurons. However, these models are not able to explain the waxing and waning of the sleep drive, unihemispheric sleep, or why some animals sleep very little.⁹⁴ The inclusion of glia and the close association between astrocytes, neurons, and blood vessels throughout the brain makes it possible to explain sleep homeostasis and unihemispheric sleep. For instance, a sleep factor, such as adenosine could accumulate locally, and because of the close association between glia, neurons, and vasculature, the homeostatic load could also dissipate at the local level. This is based on the idea of local use-dependent sleep,⁹⁵ and that sleep can be initiated locally by brain activity in small networks of glia and neurons. Thus, in unihemispheric sleep, the glymphatic system may become activated in a brain hemisphere based on the accumulation of waste products, or an animal that sleeps very little may have an efficient glymphatic system.

THE ABCs OF CHEMOGENETICS

Chemogenetics—also called pharmacogenetics—is a complementary technology to optogenetics, but instead of using light to control cell activity, it uses a drug-based approach. The particular branch of chemogenetics we will discuss is DREADDs—Designer Receptors Exclusively Activated by Designer Drugs.⁹⁶ This approach uses two separate types of engineered G protein-coupled receptors to genetically target cells and manipulate their activity. AAVs are used to drive expression of either the hM3Dq muscarinic receptor that acts to excite cells by a Gαq signaling pathway, whereas the hM4Di muscarinic receptor that inhibits cells by a Gαi pathway. Both hM3Dq and hM4Di receptors are stimulated by the inert compound, clozapine-n-oxide (CNO), and are unresponsive to endogenous acetylcholine. Because CNO crosses the blood–brain barrier, it can be delivered by intraperitoneal injection or by putting CNO in an animal’s drinking water.⁹⁶ Unlike optogenetics, which allows rapid, reversible control of neuron activity, manipulation of hM3Dq and hM4Di receptors impacts neuronal activity over a time-course that is influenced by the bioavailability of the CNO (systemic vs. directly to brain). After delivery of the CNO peak effects on sleep are evident for ~60–180 minutes.⁹⁶ One advantage of chemogenetics is that it can activate or suppress the activity of broadly scattered neuron populations (eg, MCH and hypocretin cell fields), making it a more useful approach than optogenetics for manipulating distributed cell groups. The major limitation is that binding of the CNO to the receptor will induce a pattern of activity in the affected neurons that does not mimic the natural activity pattern of those neurons, and once the CNO is delivered it will continuously influence the receptor until it is metabolized (~6 h).⁹⁶

The chemogenetic approach has been used to determine how wake-active cells group participate in behavioral control. For example, Sasaki et al. used CNO-induced activation of hM3Dq receptors on hypocretin neurons to show they increase wakefulness, whereas CNO-induced activation of hM4Di receptors increases sleep for 4 h.⁹⁷ This group has also combined both optogenetics and DREADDs to define the brain region where hypocretin deficiency produces narcoleptic symptoms. They used mice in which both the hypocretin receptors were deleted and then reinserted either hypocretin receptor 1 or 2 in phenotype-specific neurons in the pons.⁹⁸ Cataplexy was suppressed when the hypocretin-2 receptors were restored in the serotonergic neurons of the dorsal raphe, whereas waking behavior, and sleep fragmentation were rescued only when hypocretin-1 receptors were restored in the LC. To confirm their findings, the authors used orexin-ataxin-3 mice, and selectively activated either the dorsal raphe serotoninergic neurons or the LC neurons with DREADDs technology. Chemogenetic-induced activation of the serotonin neurons rescued cataplexy while the LC stimulation increased waking. This finding suggests that cataplexy results from loss of hypocretin-induced activation of raphe neurons and that sleepiness is caused by reduced LC activity, although these findings differ from previous studies, which show that cataplexy results from loss of LC activity.^59,60 A combination of chemogenetics and optogenetics was used to identify GABA neurons in the lateral hypothalamus in regulating waking.^99,100 The chemogenetic method recently determined that activation of GABAergic neurons in the medullary parafacial zone (PZ) rapidly induced NREM sleep regardless of circadian time.¹⁰¹ Anaclet et al. reported that PZ neurons monosynaptically innervate and release GABA onto parabrachial neurons, which in turn project to and release glutamate onto magnocellular basal forebrain neurons. They hypothesized that GABAergic PZ neurons trigger NREM and slow-wave activity via this newly identified medullary pathway. In-vivo electrophysiology studies have yet to verify that the GABA PZ neurons are active during NREM sleep.

Chemogenetics has also been used to identify potential mechanisms of REM sleep control. Weber et al. used hM4Di receptor expression in GABA cells in the ventral medulla to determine if they influence REM sleep.⁸⁰ They found that CNO-induced inactivation of GABA cells caused marked increases in wakefulness and reduced both NREM and REM sleep amounts.⁸⁰

HOW GENE THERAPY IS SHAPING OUR UNDERSTANDING OF SLEEP–WAKE MECHANISMS

Gene therapy is not as new as either optogenetics or chemogenetics, but has nonetheless provided insight into sleep mechanisms, especially in the context of narcolepsy. Gene therapy has been used to deliver specific genes (eg, dopamine) to the CNS of patients with either Parkinson’s or Alzheimer’s disease to improve the motor and cognitive symptoms associated with each disorder.^102,103 A similar approach could be used to restore hypocretin function in human narcolepsy, and studies discussed below support this idea.

Mouse models of narcolepsy provide an ideal tool for testing if hypocretin gene therapy can improve either sleepiness or cataplexy. Several recent studies have shown that hypocretin gene insertion within various neuronal populations can reduce cataplexy and improve waking function.^40,104–106 For example, Liu et al. showed that driving hypocretin gene expression in amygdalar neurons of narcoleptic mice reduces cataplexy attacks associated with fear cues.¹⁰⁷ This same group also showed that cataplexy could be reduced by driving hypocretin gene expression in pontine regions¹⁰⁴ associated with controlling muscle tone.¹⁰⁸ These results suggest that cataplexy can be rescued by restoring hypocretin expression at multiple levels of the neuroaxis.

Gene therapy approaches have also been helpful in consolidating wakefulness in narcolepsy. Kantor et al. found that hypocretin gene expression in the ventrolateral hypothalamus of narcoleptic mice improved wakefulness, but had no effect on cataplexy.¹⁰⁶ These observations broadly support those showing that restoring hypocretin-2 receptor function in tuberomammillary (TMN) neurons of the posterior hypothalamus in hypocretin-2 receptor-deficient mice improved sleepiness by consolidating waking bouts.¹⁰⁹ These two studies suggest that hypocretin sustains wakefulness by activation of histamine cells in the TMN. But in a broader sense, these gene therapy studies illustrate that candidate genes—such as hypocretin—can be transferred to surrogate neurons in target brain regions to control the circuits underlying arousal and motor control, and this strategy could be used to mitigate cataplexy and sleepiness in narcolepsy.

DEEP BRAIN IMAGING OF CELLS DURING SLEEP

One of the newest technologies in neuroscience is deep brain imaging of live cells in the intact brain during behavior.^110,111 Currently, activity in brain regions can be detected using positron emission tomography (PET scan) or functional MRI. However, these methods cannot identify the phenotype of the cells, which hinders our ability to understand how the brain processes complex information to trigger behaviors, such as sleep. On the other hand, voltage-sensitive dyes allow researchers to measure and record the activity patterns of genetically identified cells and their populations from the cell body to terminals.¹¹² The goal of the Brain Research through Advancing Innovative Neurotechnologies (BRAIN) initiative at the National Institutes of Health is to create tools to visualize in a living brain the propagation of activity from single cells to whole populations across the network in real time (https://www.braininitiative.nih.gov/funding/initiatives.htm).

Deep brain imaging (or fiber photometry) measures changes in calcium dynamics in cells using fluorescent probes.¹¹³ New genetically encoded calcium indicators can measure calcium influx in astrocytes¹¹⁴ and neurons.¹¹⁵ The most widely used genetically designed protein-based calcium sensors are GCaMPs, consisting of the calcium-binding protein calmodulin (CaM) fused to green fluorescent protein (GFP), and a linker peptide M13. When calcium attaches to the calcium-binding protein, there is a conformational change in the fluorescent protein, which emits a signal upon excitation with light. At rest, the cells containing the GCaMP will exhibit a basal fluorescent signal. However, when a cell (astrocyte or neuron) is stimulated, the levels of calcium inside the cells will increase, causing an increase in the intensity of the fluorescence. The change in intensity of the fluorescent signal can be empirically determined (df/f) and reflects the activity of the cell.¹¹³

Each successive generation of GCaMPs has been improved to yield better speed, sensitivity, intensity, and durability of the fluorescent signal.^110,111 The current versions of GCaMP6 (6s, 6m, 6f for slow, medium, fast) can detect single action potentials in neurons.¹¹⁵ GCaMP6-expressing transgenic mice are available through commercial sources. These can be crossed with a Cre-expressing line to yield GCaMP6 expression in phenotype-specific cells. GCaMP6 can also be delivered by microinjection of rAAV into a target brain region (see Table 2).

The fluorescence of GCaMP6-expressing cells can be monitored in virtually any brain region. This is achieved by placing a tiny endoscope (0.5–1.0 mm diameter) in the vicinity of targeted cells. The endoscope transmits the fluorescent signal to a miniature fluorescence microscope (2 g) mounted atop the animal’s head.^116,117 This method of using a microscope to visualize the change in fluorescence in individual neurons is referred to as “deep brain imaging.” The significant advantage is that the endoscope-microscope combination can image many individual neurons at the same time, thereby providing unprecedented data on the behavior of individual neurons during a specific behavior.¹¹⁸ Moreover, these neurons can be followed for many days.

An alternative approach is not to use the microscope but instead, gather the signal with a fiber-optic line from the endoscope in the brain. This approach is referred to as “fiber photometry”. The rationale is that since the GCaMP is in phenotype-specific neurons, then it is possible to measure the collective output of this population. The serious limitation of “fiber photometry” is that it cannot isolate the activity of single neurons or locate the source of the fluorescence signal (adjacent vs. distal somata or whether the signal is from a dendritic field). The signal picked up by a single endoscope represents the fluorescence signals from many GCaMP6 cells, both near to the probe as well as distally. Fiber photometry is analogous to multiunit recording since it identifies the activity of a population of neurons.

In theory, the use of GCaMPs is similar to classic single or multiunit recordings in that it allows for the determination of when individual cells or their populations become active in the context of a specific behavior. The limitation of GCaMPs is that it does not indicate whether the neuron fired as single spikes, or in clusters. Thus, it does not provide fine temporal details about the pattern of activity. Researchers should also be aware that GCaMPs bind calcium, and a high concentration of GCaMP in the cell will add to calcium buffering that may, over time, alter the activity of the cell.¹¹⁹ If the concentration of calcium inside the cell is too high, it may also be toxic to the cell.¹¹⁹ Thus, it is necessary to adjust the titer of the rAAV-GCaMP to avoid overloading. Researchers should also be aware that the fluorescent signal will quench with continued excitation. Therefore, it is important to reduce both the intensity of the excitation beam and the duration of exposure.

Deep brain imaging has already proven useful in identifying potential REM sleep generating neurons in the pons. Recently, Cox et al. used GCaMP6s to identify the activity patterns of different cell types in the dorsal pons—a region hypothesized to regulate REM sleep. They found that glutamate cells in this region were most active during REM sleep, whereas GABA cells were most active during wakefulness.¹²⁰ Their results are important because it suggests that glutamate cells could be responsible for controlling REM sleep, supporting the longstanding hypothesis—and multiple lines of evidence—that a glutamatergic mechanism underlies REM sleep regulation.⁷⁹

THE CRISPR-CAS SYSTEM AND ITS POTENTIAL IN DETERMINING SLEEP MECHANISMS

The CRISPR-Cas system is a new gene-editing tool for gene splicing and editing.^121,122 It allows researchers to delete, silence, enhance, or insert specific genes at a precise location within the genome. This new gene manipulation system represents a big step forward in genetic manipulation because it reduces the time, effort, and cost associated with more conventional genetic strategies of gene manipulation. But, perhaps most importantly, it can be applied across all eukaryotes, including humans. There are other approaches to editing genes, such as short interference RNA (siRNA) or micro RNA (miRNA), but the effects are transient, and they could have unintended consequences on other genes as the siRNA or miRNA sequence may be common to other genes (off-target).

CRISPR is an acronym for Clustered Regularly Interspaced Short Palindromic Repeats—segments of DNA containing short repetitions of base sequences with each repeat sequence followed by pieces of spacer DNA. The Cas protein uses CRISPR spacers to recognize and cut these exogenous genetic elements in a manner analogous to RNA interference. By delivering a “Cas9 nuclease” and guide RNAs into a cell, its genome can be manipulated at a specific location so that existing genes can be removed and new ones added. Guide RNA and Cas9 can be inserted into a plasmid, and after packaging it into AAV or lentivirus, it can be injected into a specific brain region containing the cells of interest. To target a specific phenotype of cells, the expression of the guide RNA-Cas9 can be driven by specific promoters or Cre. Multiple guide RNAs can be delivered in a single package to delete multiple genes of interest in the same cell.

The CRISPR-Cas method was used recently to identify how the entire NMDA receptor family contributes to sleep regulation in mice.¹²³ Sunagawa et al. systematically deleted each member (ie, Nr1, Nr2a, Nr2b, Nr2c, Nr2d, Nr3a, and Nr3b) of the NMDA receptor family and then assessed how such manipulations impacted the normal sleep–wake cycle. They found that the triple targeting of a single gene using the CRISPR-cas system achieves almost perfect knockout efficiency (96–100%) of candidate genes. But, most importantly they found that removing Nr3a caused short sleeping mice, which slept 97 minutes less each day than wild-types. This study is important because it shows the utility of CRISPR technology to examine the genes involved in sleep, and it identifies how a single gene (Nr3a) can transform the normal sleep cycle in mice.

NEW HISTOCHEMICAL TISSUE-CLEARING METHODS AND CIRCUIT MAPPING TOOLS

New histological methods are emerging that are likely to replace the familiar cryostat and sliding microtome that we have been using to cut tissue. The new approach is to make the tissue transparent (Figure 4), scan it with a laser from a light-sheet microscope, and then the software compiles a 3D image of the scanned images. These methods are known by various names (3DISCO, CLARITY, CUBIC, SCALE) depending on the process (passive vs. electrochemical) and type of chemicals (hydrogels, organic solvents, and detergents) that are used to stabilize proteins, remove lipids, and render the tissue, such as the brain, transparent.^124–127 While the goal of the various tissue-clearing methods is to produce tissue that is transparent and ready to be scanned with a light-sheet microscope, the chemicals have varying effects on the endogenous proteins and fluorescent markers.¹²⁸ As such, researchers have to empirically determine the tissue-clearing method that is best suited to visualize their fluorophore. Readers should also be aware that multiple high-performance computers (workstations with high-end video graphics capability) and ample data storage (solid-state terabyte drives) are required to process the images.

Figure 4.

Whole mouse brain cleared with the advanced-CLARITY protocol (Konadhode and Shiromani, unpublished data). The top photo depicts a brain that was perfused with 10% formaldehyde but did not undergo further histological clearing. The second brain was processed for 30 days using the advanced-CLARITY protocol.¹²⁷ The CLARITY method uses a hydrogel matrix to crosslink and stabilize proteins, and then the lipids are dissolved with other chemical agents. The final step is to equilibrate the brain in media that produces a suitable refractory index for light to pass through with little or no scattering. The brain appears transparent because light is passing through with very little scattering. The process can be accelerated by using an electric field and pressure (ACT-PRESTO).¹³⁶ Besides CLARITY,¹²⁷ there are other methods to clear tissue.^124–126 Each method has different effects on the proteins, including fluorescent reporter genes. Therefore, researchers have to empirically determine the tissue-clearing method that provides the clearest fluorescent signal.

The enormous appeal of the new histoclearing methods is that they allow visualization of the underlying cytoarchitecture in great detail.^127,129 If the brain contains fluorescent markers in phenotype-specific neurons (eg, orexin or MCH), the 3D image will depict the distribution of their somata along with the neural pathways throughout the brain. This allows for visualization of the relationship between cell types in a way that has not been possible before. For instance, it will be useful in reconstructing visual maps of sleep–wake circuitry to understand how different circuits communicate with one another and where they are anatomically located in the CNS. This method can be used in combination with neuronal activity markers such as c-Fos to determine when particular neuron networks are activated and where they project.¹²⁶

Figure 5 depicts the distribution of MCH neurons in a mouse brain (Shiromani lab, unpublished data) cleared with advanced CLARITY.¹²⁷ Previous studies have described the distribution of MCH neurons, albeit in rat brain, based on the serial reconstruction of brain sections cut on a cryostat.^130,131 However, with the CLARITY method, we gain a new understanding of the spatial distribution of the various clusters and the projections of the neurons, something that is not possible in 2D images from coronal sections cut on a microtome. This allows us to position the probes better for optogenetic stimulation and deep brain imaging of specific clusters (Figure 5 and accompanying Supplementary Video). For instance, if an optogenetic probe were located atop the MCH neurons in the zona incerta then the light will only weakly reach the cluster of MCH neurons located in the perifornical area. The new data also reveal a pattern in the projection pattern of the MCH neuron clusters. For instance, when viewed in the sagittal and horizontal positions, notice that some clusters project toward the anterior of the brain while the MCH neurons located in the perifornical and lateral hypothalamus region project caudally (see Figure 5). Such information can only be gleaned by viewing the same dataset from various perspectives, which is something that is feasible only with the new tissue clearing methods.

Figure 5.

The distribution of MCH neurons (MCH-EYFP) in a mouse brain (unpublished data, Shiromani and Konadhode). A coronal section of the brain (hypothalamus between optic chiasm and midbrain) was processed with the advanced-CLARITY protocol.¹²⁷ After clearing, the tissue was scanned with a Zeiss Z.1 Lightsheet microscope (896 slices in 3 um increments). The images were stitched to yield a single 3D image (shown in video link—place link to supplemental material here). The assembled image was rotated to depict the coronal, horizontal, and sagittal perspectives. This is the first time that a population of neurons implicated in sleep–wake has been reconstructed from a single block of the brain. This new method has revealed for the first time a topographical difference in the projection of MCH neurons. It will accelerate the discovery process.

The power to image the intact brain will be further enhanced by the use of new retrograde tracers. Such tracers have been used before (eg, cholera toxin subunit-b; wheat germ agglutinin-horseradish peroxidase) to identify the connectivity between sleep–wake neuronal populations. However, new genetically assisted circuit mapping tools have been developed that can be used in conjunction with optogenetics in cre-transgenic animals.¹³² These new tracers allow researchers to identify, manipulate, and map the projections of genetically targeted neurons (see Figure 1D). For example, the tracer AAV-EF1alpha–mCherry—IRES-WGA-Cre (WGA is the plant lectin in WGA-HRP used for tract tracing) can be injected into a target site such that Cre is then retrogradely transported to neurons that project to the injection site. It is very likely that multiple phenotypes of neurons project to the target site, all of which will express Cre. The gene of interest, such as ChR2, can be inserted into Cre-expressing neurons at a specific site by injecting AAV-DIO-ChR2-Cre at that site. All neurons containing Cre and that project to the target site will now contain ChR2. Thus, projection specific neurons can be stimulated.^133,134

Cre-transgenic animals can be used to restrict expression of ChR2 only to specific neurons that have a defined projection pathway. Moreover, new tracers, such as canine adenovirus⁴² can be used to identify trans-synaptic connections. The retrograde tracer CAV-Flex-Flp is injected in the target site and is retrogradely transported and expressed only in Cre+ neurons projecting to that site. A second virus containing DIO-ChR2 can then be injected at the site that is to be optogenetically stimulated, thereby allowing targeted activation of a defined projection pathway.^133,135 This methodology can also be refined to kill specific neuron projection pathways by linking candidate “death-genes” (eg, ataxin-3) to the retrograde tracers. The methodology can also be extended to identify the input–output organization of specific circuits. This method, referred to as tracing the relationship between input and output (TRIO), can be combined with cre-transgenic animals (cTRIO).¹³⁵

SUMMARY

This review highlights how some of the newest neuroscience tools can be (and have been) used to study the mechanisms underlying arousal and sleep. We discussed how optogenetics, chemogenetics, genetic-based tract tracing, deep brain imaging, gene therapy, and CRISPR-cas technologies have not only confirmed pre-existing hypotheses about sleep mechanisms but have recently been used to advance our current understanding of such mechanisms. In particular, we summarized how the different strategies have been used to dissect the normal and pathological function of the hypocretin system in sleep–wake control with specific emphasis on narcolepsy. Even though these approaches have made clear advances in how we understand CNS control of sleep–wake behavior, there is no doubt that development of new, and refinement of current technologies, will play a pivotal role in how scientists unravel how and why we sleep. This sentiment is echoed by the celebrated English chemist Sir Humphry Davy who said: “Nothing tends so much to the advancement of knowledge as the application of a new instrument”.

SUPPLEMENTARY MATERIAL

Supplementary data are available at SLEEP online.

FUNDING

PJS was supported in part by the Department of Veterans Affairs, Veterans Health Administration, Office of Research Development (BLR&D) (grant BX000798), and NIH grants NS052287, NS084477, NS098541, and NS079940. JHP was supported by grants from CIHR and NSERC.

DISCLOSURES STATEMENT

None declared.