Tumor Necrosis Factor Alpha in Sleep Regulation

Summary

This review details tumor necrosis factor alpha (TNF) biology and its role in sleep, and describes how TNF medications influence sleep/wake activity. Substantial evidence from healthy young animals indicates acute enhancement or inhibition of endogenous brain TNF respectively promotes and inhibits sleep. In contrast, the role of TNF in sleep in most human studies involves pathological conditions associated with chronic elevations of systemic TNF and disrupted sleep. Normalization of TNF levels in such patients improves sleep. A few studies involving normal healthy humans and their TNF levels and sleep are consistent with the animal studies but are necessarily more limited in scope. TNF can act on established sleep regulatory circuits to promote sleep and on the cortex within small networks, such as cortical columns, to induce sleep-like states. TNF affects multiple synaptic functions, e.g. its role in synaptic scaling is firmly established. The TNF-plasticity actions, like its role in sleep, can be local network events suggesting that sleep and plasticity share biochemical regulatory mechanisms and thus may be inseparable from each other. We conclude that TNF is involved in sleep regulation acting within an extensive tightly orchestrated biochemical network to niche-adapt sleep in health and disease.

Introduction

Sleep remains a fundamental scientific enigma. Although significant progress has been made in elucidating roles for sleep in cognition and brain health, the primary functions of sleep have not been firmly established. Moreover, the regulation of sleep and wake is complex and not fully understood, and newer questions have arisen regarding the role and need for local sleep within a specific brain region versus more generalized sleep states. We do, however, appreciate that sleep manifestations are systemic. Sleep affects almost every physiological function, e.g. body temperature, hormone secretions, and respiratory, cardiac, kidney, and immune functions. Sleep actions on the brain range from altering multiple pathologies to recovery from them, as well as performance, mentation, emotion, learning and memory, etc. At the same time, many physiological functions affect sleep, e.g. body temperature, hunger, sexual drive, development, respiration, etc. and extracellular signals involved in the regulation of these functions affect sleep [1]. This latter point is critical to appreciate, as these physiological effects are not accounted for within the two-process (homeostasis and circadian) regulation of sleep model [2]. Figure 1 illustrates some of the known components of tumor necrosis factor alpha (TNF) regulation and TNF biological activities that are also linked to sleep. Although the biochemical network shown is incomplete, it and the more extensive network (not shown) are likely beyond the capacity of any individual’s ability to fully understand the dynamic nuances of such networks. Further, how the components interact with other TNF-regulated processes to orchestrate sleep niche adaptation further challenges comprehension. Herein we focus on only one sleep regulatory substance, TNF. We do so in recognition that other molecules are also involved in sleep regulation, but focus on TNF to illustrate principles of physiological sleep regulation and function. The reader is referred to other reviews for broader treatments of sleep regulation, both biochemical [3–9], neurobiological [10, 11], and glialogical [12–15].

Figure 1.

Tumor necrosis factor α (TNF) regulation in health and disease and its sleep-linked effectors and actions. Physiological regulation of TNF (upper pathway) includes dampening of TNF production via negative feedback. In contrast, pathogenic stimuli can lead to overwhelming positive feedback (lower pathway). TNF can induce its own and other pro-inflammatory cytokines to amplify host-defense reactions to pathological challenges. Clinically this positive feedback can lead to a cytokine storm with the persistence of intense pathological stimuli. Figure abbreviations: CRH, corticotropin-releasing hormone; GABA, gamma-aminobutyric acid; GHRH, growth-hormone-releasing hormone; Glu, glutamic acid; IL, interleukin; NE, norepinephrine; NFkB, nuclear factor kappa B; NO, nitric oxide; PG, prostaglandins; sTNFR, soluble TNF receptor; TGFβ, transforming growth factor β.

TNF biology

Within the brain, TNF has many functions including mediation of brain damage, e.g. cerebral ischemia [17], cerebral blood flow, neuro-protection, responses to infection, and synaptic scaling [18]. TNF is expressed by microglia, astrocytes, and neurons [19–22]. The actions of TNF depend not only on the receptor type – either 55 kilo-Daltons (kD) or 75 kD – and adaptor proteins, but also on the context of the stimulus and the interaction with substances that modify TNF activity. In addition, TNF influences whole organism functions such as body temperature [23, 24], appetite [25], cognition [26], and brain development [27].

The mechanism of TNF signaling is extensively studied due to its broad influence on physiologic and pathophysiologic processes (Figure 1) [28–36]. The TNF ligand has two forms, a 26 kD trans-membrane protein [31] and a soluble 17 kD protein. Trans-membrane 26 kD TNF predominates in the brain [37] and fat tissue [38]. In contrast the soluble 17 kD form is, for example, more abundent in muscle and liver [ibid]. Soluble 17 kD TNF is first transcribed as the trans-membrane 26 kD form, then cleaved by TNF converting enzyme (Figure 2) to yield the soluble 17 kD TNF form. The regulation of TNF is tissue-specific [39]. Production is driven in part via nuclear factor kappa B and TNF can induce its own expression through this regulatory pathway.

Figure 2.

Potential tumor necrosis factor α (TNF) signaling mechanisms. The upper box illustrates that both TNF ligand and its receptors have transmembrane forms and that the extracellular components of those forms can be cleaved to yield soluble 17 kilo-Daltons (kD) TNF ligand and soluble TNF receptors. The bottom panels illustrate TNF reverse signaling (left), conventional soluble ligand-transmembrane receptor signaling (middle), and potential direct cell-to-cell TNF signaling with intracellular signaling occurring in both cells (right).

Soluble 17 kD TNF can signal via the trans-membrane TNF receptors (Rs), (Figure 2, bottom center) as evidenced by the multiple effects induced by injection of soluble 17 kD TNF into the brain including excess sleep. In addition, trans-membrane 26 kD TNF can complex with the soluble TNFRs to induce intracellular pathways within the cell expressing the transmembrane TNF (Figure 2, left lower panel). The extent of the occurrence of this mechanism in the brain is not known although TNF reverse signaling is described in axonal physiology [40]. Another potential TNF signaling mechanism is direct cell-to-cell contact with the transmembrane TNF binding to the transmembrane R with subsequent intracellular signaling occurring in both cells. Whether this mechanism occurs in brain is unknown (Figure 2, lower right) although TNF-mediated direct cell-to-cell contact appears to be important in embryonic stem cell differentiation [41].

Cellular responses to TNF signaling can vary broadly based on variability in R expression and adaptor proteins present in the target cell. There are two membrane-bound TNF Rs, a 55 kD R and a 75 kD R. Neither R has intrinsic enzymatic activity, instead signal through recruitment of adaptor proteins. The TNF R super family is divided into 2 broad categories, according to the adaptor proteins that are recruited upon binding of a ligand [32–36]. Both the 55 kD and 75 kD TNFRs form trimeric complexes with TNF [36]. The specific R type, adaptor protein, and spatial orientation of R-ligand formations dictate target response, which can range from induction of cell death to protection from cell death [32].

The level of expression of TNF in the brain is activity-dependent. When murine facial whiskers are repeatedly stimulated, neuronal production of the trans-membrane 26 kD TNF increases in the somatosensory cortex [37]. Optogenetic stimulation of neurons in vitro enhances expression of neuronal TNF [42]. The induction of neuronal production of TNF by cell activity suggests that TNF plays a role in neuronal connectivity. TNF has a well-established role in synaptic scaling [18]. TNF potentiates AMPA-induced potentials at the post-synaptic membrane [43], and also acts to increase Ca⁺⁺ conductance via AMPA-induced [44] and voltage-dependent mechanisms [45, 46]. Additionally, TNF modulates glutamatergic transmission [16, 44, 45]. TNF promotes neuronal expression of Homer1a mRNA [47]. Homer1a plays a role in synaptic physiology and is increased after sleep deprivation [48, 49].

TNF in sleep regulation; animal studies

A sleep regulatory substance should be able to satisfy several basic criteria [50–52]. In animal studies, a sleep regulatory substance should enhance a sleep phenotype, such as duration of non-rapid eye movement sleep (NREMS). Inhibition or reduction of the substance should reduce the sleep phenotype. The levels of the purported sleep regulatory substance measured in the brain should correlate with the duration of sleep loss and sleep propensity. Additionally, the sleep regulatory substance should act within known sleep regulatory circuitry and/or induce sleep-like states within local circuits (see below).

The somnogenic actions of TNF in rabbits were first demonstrated in 1987 [53]. Since then, TNF has been shown to promote NREMS duration in every species tested: rats [54], mice [55], and sheep [56]. The effects on NREMS are profound. For instance, when mice are given 3 μg of TNF via intraperitoneal injection, they spend around 90 min longer in NREMS during the first 9 h post-treatment [55]. Additionally, δ power during NREMS also increases after TNF injection (ibid). Δ power amplitude is a sleep phenotype that under normal conditions is indicative of sleep intensity [57]. TNF administration also exerts an effect on rapid eye movement sleep (REMS). While lower doses of TNF are sufficient to promote NREMS, REM sleep is not affected by lower doses of TNF. However, at higher doses of TNF, REMS is inhibited [58]. Interestingly, absence of the 55 kD TNFR in mice is associated with a reduction of REMS duration [55].

While administration of TNF promotes NREMS, inhibiting TNF reduces spontaneous NREMS. NREMS is reduced whether TNF is inhibited by anti-TNF antibodies [59], the soluble TNFR [60], or fragments of the soluble TNFR [61]. The physiologic utility of TNF inhibition is not fully understood. Preliminary data from our lab (unpublished) suggests that treatment of neuronal/glial co-cultures with a soluble TNFR promotes a wake-like state suggesting possible TNF reverse signaling (see Figure 2). Furthermore, if animal subjects are given TNF inhibitors before sleep deprivation, the animals have less sleep rebound than corresponding controls [62]. Substances that modulate TNF production, either directly or indirectly inhibiting its production, inhibit NREMS. For example, IL4, IL10, and IL13 all inhibit NREMS [8]. NREMS increases induced by acute mild increases in environmental temperature are also inhibited when TNF is inhibited [61].

Mice with the 55 kD TNFR knockout sleep less than control mice. Specifically, TNFR knockout mice have less NREMS and REMS than control mice [55], suggesting that TNF has a role in normal sleep physiology, influencing normal sleep times. When these mice are given soluble 17 kD TNF, they do not respond with an increase in NREMS as occurs in control mice (ibid), supporting the concept that TNF as a sleep-promoting substance is acting through the TNFR. Mice lacking both TNFRs also have less spontaneous NREMS than controls [63].

Brain levels of TNF or TNF mRNA correlate with sleep propensity. Within the hypothalamus, concentrations of TNF [64] and TNF mRNA [65–67] fluctuate diurnally. These levels are highest at day break, correlating with the beginning of the rat sleep period. The concentration of TNF protein from trough to peak varies by about a factor of 10. TNF mRNA concentrations double during this same time frame, indicating the presence of post-transcriptional upregulation of TNF underlying this process. TNF levels are also influenced by sleep deprivation. Sleep deprivation increases hypothalamic TNF mRNA concentration. Sleep deprivation also increases expression of 55 kD TNFR mRNA [67]. Expression of TNF in the brain can be influenced by multiple factors, such as inflammation, and these conditions are also affected by sleep loss and REMS deprivation [68–72].

TNF microinjection into brain regions associated with sleep and alertness regulation, including the hypothalamic pre-optic area and the locus coeruleus, enhances sleep [74, 75]. In contrast, inhibition of TNF via injection of soluble TNFRs into the hypothalamus inhibits spontaneous NREMS [74]. When TNF is administered locally into the subarachnoid space inferior to the basal surface of the forebrain via micro infusion, NREMS is enhanced and REMS is inhibited [58]. Furthermore, unilateral application of TNF within the somatosensory cortex induces unilateral changes in state-dependent EEG δ wave power [76, 77]. Conversely, unilateral inhibition of TNF via soluble TNFRs reduces EEG δ wave power measured in NREMS [76, 77]. When a TNF short interfering RNA is applied unilaterally to the surface of the somatosensory cortex, expression of TNF is reduced. Further, this treatment lowers EEG δ wave power during NREMS, but there is little reduction of TNF δ wave power during wake or REMS periods [78].

In summary, research over the past 25 years has comprehensively qualified TNF as a key sleep regulatory molecule, using the widely accepted criteria for sleep regulators [50, 51]. However, TNF animal sleep work is limited because almost all of it was conducted using young healthy mice or rats. In contrast, most human studies of sleep and TNF involve patients with chronic pathologies associated with high levels of TNF. Regardless, the measurement of TNF in human plasma and cerebrospinal fluid has been possible for many years. This has led to many correlative studies linking TNF to sleep, cognition, and other brain functions.

Linking TNF to human sleep

TNF blood levels correlate with EEG δ power during spontaneous sleep in humans [79]. During sleep in healthy young men, serum TNF levels decrease [80]; this finding is consistent with the reduction of brain TNF across the rat daytime sleep period [64]. After sleep deprivation in humans, circulating levels of the 55 kD soluble TNFR, but not the 75 kD soluble TNFR, increase [81, 82]. The soluble 55 kD R is a component of normal cerebrospinal fluid [83]. The TNF system may play a role in normal diurnal temperature regulation as well [82]. In healthy volunteers, intravenous injection of endotoxin increases plasma levels of TNF and the soluble 55 kD TNFR and promotes sleep simultaneously [84]. TNF induces fatigue in patients with metastatic cancer [85]. Further, women with breast cancer experience increases of fatigue and TNF levels during chemotherapy [86]. In pediatric obstructive sleep apnea patients, TNF levels increase, and are tightly related to sleepiness; plasma TNF levels, and sleepiness in these patients decrease after surgery [87] and are also associated with single nucleotide polymorphisms in the TNF gene [88].

The TNF G308A polymorphic variant is associated with metabolic syndrome, insulin resistance, sleep apnea [ibid, 89–92], heart disease [93], tuberculosis [94], Alzheimer’s disease [95], and peri-portal fibrosis regression of schistosomiasis [96]. Furthermore, this variant has relevance to cognitive resilience after sleep deprivation, healthy volunteers with this variant have impaired psychomotor vigilance performance [97].

Many human studies involve the measurement of systemic TNF, or associated molecules. How systemic TNF communicates to the brain in humans remains unknown. However, in animals, systemic TNF can induce its own mRNA production in brain via vagus nerve afferents (Figure 3). Vagotomy blocks this action [73]. These results suggest that TNF-induced action potentials in vagal afferents induce TNF and interleukin-1β mRNA expressions in brain. Thus it seems that systemic TNF influences brain production of TNF and other cytokines via the vagus nerve. TNF can also be directly transported from blood to brain [98]. These studies reinforce the experimental measurement of systemic TNF in humans as a means to assess TNF-sleep interactions. However, the nuances of systemic-brain TNF interactions in health and disease remains understudied.

Figure 3.

Vagotomy attenuates the somnogenic actions of systemic soluble tumor necrosis factor α (TNF). TNF, 1 μg give intraperitoneally to mice, enhances non-rapid eye movement sleep (NREMS) for hours. Vagotomy attenuates the effect of this TNF dose on NREMS (A). Rapid eye movement sleep (REMS) durations after TNF were similar to those after saline injections and did not differ between treatment groups (B). (*) = significant differences for 2 hour time bins between vagotomized and sham-operated mice after TNFα injections compared to saline injections. P< 0.05. White circles = saline injection group; dark circles = TNF injection group. The right graphs show that systemic TNF induces brain and liver TNF mRNA and interleukin-1 β (IL1β) mRNA expressions; TNF enhanced its own expression and IL1β expression. These expressions were attenuated in brain after vagotomy. The results provide reassuring evidence that measures of systemic TNF are related to brain functions including sleep. Abbreviations; S-sham operated controls; Sal-saline; Data are derived from reference [73].

There are multiple pathologies associated with excessive soluble 17 kD TNF plasma levels that are also associated with poor sleep, (e.g. fragmentation, reduced duration, loss of normal sleep architecture, sleepiness). These include sleep apnea, insomnia, excessive daytime sleepiness, chronic fatigue syndrome, cancers, (e.g. gliomas), psoriasis, severe cutaneous lupus, rheumatoid arthritis, severe infectious diseases (e.g. AIDS), post-dialysis fatigue, alcoholism, obesity, Alzheimer disease, ankylosing spondylitis, inflammatory bowel disease, some autoimmune diseases (e.g. osteoarthritis), preeclampsia, and hyperemesis gravidarum. A variety of TNF inhibitors have been used to treat these diseases and in some of these studies sleep measures were taken. TNF inhibitors generally restore sleep to more normal patterns in these clinical conditions while simultaneously alleviating disease symptoms. The effects of these drugs on sleep in normal people are mostly unknown, thus there remains a gap between human and animal TNF-sleep knowledge. For example, Progranulin (PGRN) is an endogenous glycoprotein with roles in neurodevelopment and neurodegeneration [99]. PGRN is a TNFR ligand acting as a competitive TNF antagonist. PGRN is used for disease therapy [100] including in autoimmune diseases, such as osteoarthritis, inflammatory bowel disease, psoriasis, systemic lupus erythematous, systemic sclerosis, multiple sclerosis and Sjogren’s syndrome [100]. Although PGRN treatment can affect mouse sleep-linked behavior [101], its actions on human sleep remain unstudied. Six additional TNF inhibitors used clinically for which sleep-related evidence are discussed herein.

Sleep/wake effects of prescribed TNF-active substances

Etanercept (ETA)

ETA is a recombinant TNFR protein that binds TNF thereby reducing the effects of naturally present TNF [102]. It is not known if ETA can bind to the 26 kD trans-membrane form of TNF to reverse signal (see above), thus interpretation of clinical studies is speculative. Nevertheless, because TNF is a key inflammatory regulator, ETA is used to treat autoimmune diseases [103]. ETA was the first FDA approved medicine for rheumatoid arthritis and polyarticular-course juvenile rheumatoid arthritis [104]. ETA affects sleep. For instance, ETA improves sleep efficiency, total sleep time, and stage 2 sleep duration in rheumatoid arthritis patients [105] and ETA alleviates daytime sleepiness in rheumatoid arthritis patients [106]. In another study, after 36 weeks of ETA-treatment, sleep improved inpatients with moderately active rheumatoid arthritis [107]. Similar results were obtained from psoriasis patients. Most psoriasis patients have disrupted sleep. ETA-treatment for 12 and 24 weeks improved their impaired sleep [108]. ETA also improves the sleep problems associate with ankylosing spondylitis patients [109]. In addition, ETA relieves objective sleepiness in obese patients with obstructive sleep apnea [110]. Finally, ETA reduces the REMS in abstinent alcohol-dependent patients; excessive REMS is associated with alcohol relapse [111].

Adalimumab (ADA)

ADA is an anti-recombinant human TNF monoclonal antibody capable of blocking TNF inflammatory activity [112]. It is approved for treatment of a number of autoimmune diseases [113]. For example, the sleep disturbances associated with rheumatoid arthritis [106] are improved with ADA treatment [114]. In psoriasis patients, ADA treatment improved sleep quality [115]. Fatigue and sleep problems also occur in ankylosing spondylitis patients [116]. ADA improves overall sleep and sleep quality in these patients [117].

Certolizumab pegol (CZP)

CZP is a pegylated fragment of an anti-recombinant human TNF monoclonal antibody [118]. CZP is used for the treatment of rheumatoid arthritis, ankylosing spondylitis and Crohn’s disease etc. [119]. In a clinical trial of axial spondylo arthritis, CZP improved sleep in the CAP-recipient patients but not in those receiving placebo [120].

Golimumab (GLM)

GLM is an anti-human monoclonal TNF antibody that has recently been approved for clinical use [121]. GLM binds to both the soluble and transmembrane forms of TNF [30]. In active ankylosing spondylitis patients, GLM treatment led to significant improvements of sleep as determined by Jenkins Sleep Evaluation Questionnaire [109, 122, 123].

Infliximab (IFX)

IFX is a chimeric anti-human TNF monoclonal antibody [124] use to treat autoimmune diseases [125]. In rheumatoid arthritis patients, IFX lowers circulating TNF levels and improves sleep [126]. IFX also relieves daytime sleepiness in rheumatoid arthritis patients [106]. In ankylosing spondylitis patients, IFX improves sleep [109].

Thalidomide (TH)

TH is an immunomodulatory drug that inhibits TNF expression [127]. The current primary clinical uses of TH are for multiple myeloma and erythema nodosum leprosum [128]. TH treatment of severe cutaneous lupus patients induces drowsiness as a common side effects [129] and it reduces serum TNF levels in systemic lupus erythematosus patients [130]. Using TH as the last option of therapy in advanced glioma patients, the patient’s sleep improved [131]. In a separate study, glioma patients were demonstrated to have high TNF serum levels [132]. In Chron’s disease patients, the cessation of TH treatment caused intractable insomnia [133]. Historically, TH was used to relieve morning sickness in pregnancy [134] but the discovery that TH is teratogenic ended this use [135]. Regardless, TH increased REMS and NREMS (stages 3 and 4) in humans but decreased time in stage 1 NREMS [136].

TNF may also be involved in mental disease. In bipolar disorder patients, serum TNF levels are enhanced during manic and depressive episodes [137] and are positively correlated with inhibitory control [138]. In schizophrenia patients, TNF appears to be a trait marker, as levels remained elevated in acute exacerbations and also during subsequent antipsychotic treatment [139]. There is extensive literature indicating that major depressive disorder is associated with elevated plasma levels of TNF. TNF antagonists (IFX, ADA, and ETA) mitigate depressive symptoms and improve associated cognitive deficits [140]. Although these mental diseases are associated with sleep disturbances, clinical outcomes following TNF inhibitor treatment remains understudied.

TNF sleep biology within the context of brain organization of sleep

Despite millions of human stroke cases and many animal brain lesion studies, if the subject survived, sleep always ensues. Although post-lesion sleep may not have normal structure with oscillations between NREMS and REMS and periodic awaking, sometimes duration of sleep recovers [141–143]. This strongly suggests that no specific circuit is required for sleep and that sleep has self-organizing properties in any viable brain tissue that remains after lesions. Further, several marine mammalian and bird species have unihemispheric sleep [144–146] suggesting that in normal animals half of the brain can be asleep. There are also many disassociated state pathologies suggesting that humans can be awake and asleep simultaneously [147]. For instance, during sleep walking the individual is awake to the extent that they can navigate while walking yet asleep in that there responsiveness to some environmental stimuli is reduced. These findings collectively suggest that sleep can be confined to local areas of the brain. Thus, a bottom-up, local use-dependent sleep regulatory mechanism was proposed [148].

Strong experimental evidence indicates that sleep is a fundamental property of small neuronal/glial networks originates from the findings that individual cortical columns oscillate between sleep-like state and wake-like states [7, 76, 149]. In intact animals, evoked response potentials (ERPs) are greater in magnitude during sleep than during waking. Using an electrode array placed over somatosensory, or auditory cortex, electrical activity of individual cortical columns including ERPs could be isolated after sensory stimulation (either facial whisker twitching or auditor sounds). When the experimental rats were asleep then given a sensory stimuli most, but not all, of the cortical columns were in a sleep-like state, as determined from ERP amplitudes. In contrast, when the rat was awake, ERP magnitude was lower. Further, the longer the column was in a wake-like state the higher the probability it would enter into a sleep like state. In addition, the occurrence of the sleep-like state was activity-dependent in that if facial whiskers were repeatedly stimulated the probability of a column entering the sleep-like state increased. When rats were trained to lick a sweet solution in response to a specific whisker stimulation, if the specific cortical column to which afferent nerve fibers projected to from the stimulated whisker was in a wake-like state, the animal did not make mistakes. In contrast, if the specific cortical column was in the sleep-like state (the whole rat was awake), the rat made errors of commission or omission [7]. Finally, if TNF is applied locally to the cortical column, ERP magnitude is enhanced suggesting that TNF induced a sleep like state in the column [37]. Collectively, these data indicate that individual cortical columns oscillate between sleep- and wake-like states and that TNF drives columns into the sleep-like state.

A logical extension of the local use-dependent sleep hypothesis is that any viable neuronal/glial network, whether in vivo or in vitro should exhibit sleep- and wake-like states, as long as connectivity is intact. Co-cultures of dispersed neurons and glia spontaneously display a sleep-like default state as determined by the burstiness of neuron action potentials, and a reversal to a more wake-like state following treatment with excitatory neurotransmitters or amino acids, or electrical stimulation [42, 150, 151]. The culture’s sleep-like state develops over the course of about 10 days as neurons make connections to each other and emergent network properties become manifest. Further, following prolonged electrical stimulation (30 min), the cultures exhibit a rebound in the sleep-like state the next day suggesting sleep homeostasis. Optogenetic stimulation (30 min) induces TNF neuronal expression in the cultures [42]. Finally, if TNF is added to the cultures they appear to be in a deeper sleep-like state as synchrony of electrical potentials and amplitudes of delta waves (0.5–3.5 Hz) increase [42]. These measurements are also used to characterize NREMS in intact animals including humans. One interpretation of these studies is that reduced small networks in culture not only exhibit sleep states, but mechanistically activity-dependent TNF is involved in eliciting the sleep-like state. As Figure 2 suggests, TNF signaling includes reverse signaling. Our preliminary data hint that addition of the soluble TNFR induces a wake-like state in cultures of neurons and glia derived from mice that lack both TNFRs (unpublished).

There is a rich history linking sleep to brain plasticity (e.g. see Behavioral Brain Research Vol 69:1995 special issue). Brain connectivity theories of sleep function were derived from the learning and memory, computer, and physiological literatures [152–155] and logic considerations [148, 156, 157]. Assuming that simple percepts or memory recalls are an emergent property of dynamic synaptic activation patterns, then logically the brain is confronted with several challenges that may be dealt with in part by sleep. New synapses or enhanced synaptic efficacies are needed to form and consolidate new synaptic activation patterns, thus new memories. There is much evidence that sleep is involved in this process [155]. Further, there is a need to preserve established synaptic networks that have already proved adaptive, as evidenced by the animal being alive, and to preserve synaptic plasticity as well [148, 157]. Finally, because synapses and synaptic efficacies are activity-dependent, and there is always massive brain activity, logically there may be a need for synaptic pruning or synaptic down-scaling [153, 156]. All of these potential plasticity issues require a mechanism to label synapses for preservation, greater efficacy, or pruning; if that mechanism is linked to sleep it will provide strong support for a plasticity function for sleep.

Regardless, brain plasticity has long been recognized as being driven by neuron and glia activity-dependent expression of molecules and thus being initiated within local networks. The recognition of sleep as a local use-dependent process was more recent [7, 148, 158]. As local use-dependent processes, the underlying molecular mechanisms link sleep and plasticity regulations. All well-characterized sleep regulatory substances also affect brain plasticity.

Conclusion

There is ample evidence from animals and humans that TNF is involved in sleep regulation. TNF biology and TNF sleep mechanisms reveal much about sleep regulation, the minimal amount of tissue required for sleep, and sleep function. Local application of TNF to small circuits, such as cortical columns or to co-cultures of neurons and glia induces sleep-like states. The in vivo and in vitro sleep-likes states suggest that sleep is a fundamental property of very small circuits. TNF is expressed in neurons as a consequence of neuronal activity. TNF also has a role in brain plasticity, perhaps best known are its roles in synaptic scaling and glutamate receptor (AMPA) trafficking. Both sleep and plasticity are activity-dependent and thus sleep and plasticity mechanisms are linked, and at least via TNF seem inseparable from each other. This is strong evidence for a connectivity function for sleep.

Practice Points.

1. Patients with pathologies associated with high systemic TNF levels likely also exhibit sleep abnormalities, e.g. fragmented sleep.

2. Systemic TNF induces brain production of sleep-altering cytokines.

3. Normalization of excessive systemic TNF improves sleep.

4. Pro- and anti-inflammatory cytokines influence physiological sleep and sleep responses to pathological insult.

Research Agenda.

1. The human sleep consequences of acute vs chronic changes in TNF need to be defined.

2. How does TNF signaling vary in acute versus chronic exposures?

3. The effects of chronic TNF elevation in animals on their sleep needs to be characterized.

4. What role does TNF play in adaptive brain responses to sleep/wake schedules?

5. Characterization of TNF communication between brain and systemic pools.

6. TNF signaling, e.g. direct cell-to-cell TNF-mediated contact and signaling, requires study.

7. The cytokine orchestra in terms of positive and negative feedback mechanisms in brain and systemic tissues and its role in sleep regulation requires further study.

8. How does TNF NREMS improve peripheral immune responses?

9. Is local sleep a response to reduce local inflammation from column activation?

Abbreviations

ADA

adalimumab

CZP

certolizumab pegol

D

Dalton

EEG

electroencephalogram

ERP

evoked response potential

ETA

etanercept

GLM

golimumab

IFX

infliximab

kD

kilo-Dalton

NREMS

non-rapid eye movement sleep

PGRN

progranulin

R

receptor

REMS

rapid eye movement sleep

TH

thalidomide

TNF

tumor necrosis factor alpha

Glossary of Terms

Evoked response potential

The extracellular localized electrical response of the brain to an afferent input. ERPs can also be obtained from co-cultures of cells grown in vitro in response to a stimulus