Whatever their proper classification, both jet-lag and shift-work insomnia represent important societal problems deserving of public health and medical attention. Barring a world-wide rejection of air-travel, jet lag will continue to afflict tens of thousands of people annually. The effects of jet-lag on human performance, while typically transient, can nonetheless be significant, affecting commerce, government, and even the outcome of professional sports contests (Recht, Lew et al., 1995) . More important, only a global regression to an agrarian economy will eliminate the problem of tens of millions of workers in this country who regularly attempt to work at night and sleep during the day (Gordon, Cleary et al., 1986; Mellor, 1986) . In contrast to jet-lag, shift-work produces chronic sleep disruption lasting for the duration of shift-work exposure. For while individual differences in the ability to adjust to a nocturnal work schedule clearly exist (Sack, Blood et al., 1992) , recent studies suggest that few if any night workers regularly experience restful and restorative day sleep equivalent to that considered normal at night (Akerstedt and Kecklund, 1991) . This chronic sleep limitation is associated with significant increases in a number of consequent problems including sleepiness-related accidents (Smith, Folkard et al., 1994) , social disruption (Colligan and Rosa, 1990) , and psychiatric disturbances (Healy, Minors et al., 1993) . In addition, chronic exposure to shift-work has now been shown to be an independent risk factor for the development of both cardiovascular (Knutsson, 1989) and gastrointestinal (Segawa, Nakazawa et al., 1987) diseases. While these epidemiological studies have not identified the specific aspect of shift-work that is associated with increased risk of these disorders, the chronic limitation and disruption is foremost among plausible factors (Moore-Ede and Richardson, 1985) .

The most important aspect of human circadian physiology that limits adaptation to the extreme schedules inherent in shift-work and jet travel is the primacy of light among entraining signals, or zeitgebers (Czeisler, 1995) . Exposure to sunlight for night shift-workers, or for jet-travelers at their destination, results in maintenance (or resetting) of the clock to environmental time. This response can be prevented or over-ridden with extraordinary avoidance of sunlight or with provision of artificial light of sufficient duration and intensity to negate the sunlight signal, an approach shown to be effective in the treatment of shift-work sleep disruption (Czeisler, Johnson et al., 1990) . Practical issues sharply limit the application of artificial lighting to all shift-work settings, however, and the role for a pharmacological chronobiotic agent capable of accomplishing the same end is potentially very large. The purpose of this review is to summarize available data on the anatomy and physiology of circadian clocks, with a specific focus on potential sites of action for possible chronobiotic agents, and to summarize the status of current drugs and hormones with apparent chronobiotic effects.

Anatomy And Physiology Of The Circadian System

FIGURE 1

The circadian system can be divided into three major components: the central oscillator or pacemaker; the afferent pathways which carry entraining environmental information to the oscillator; and the efferent pathways that communicate the rhythmicity of the oscillator to the physiology and behavior of the organism. As stated above, the suprachiasmatic nuclei constitute the circadian pacemaker in mammals (Moore and Eichler, 1972; Stephan and Zucker, 1972) , including humans (Lydic, Schoene et al., 1980) . This bilateral nucleus in the anterior hypothalamus is located on either side of the third ventricle, just above the optic chiasm. These nuclei are small in volume, with approximately 10,000 cells in each nucleus of the SCN of the rat (van den Pol and Dudek, 1993) . In addition, the cells within the SCN make frequent somato-somatic contacts, a feature that may be important in the stability of the output of the network of oscillators. The SCN project to each other and to numerous hypothalamic, thalamic and cortical regions. Recently, Moore has suggested that the SCN has two anatomically distinct regions, the core, or ventral portion, and the shell, or dorsal portion (Moore, 1997) . These areas differ in the types of projections they receive, in the predominate cell type, and in their efferent pathways (see below). The core contains vasoactive intestinal polypeptide (VIP) positive neurons while the shell contains vasopressin (VP) positive neurons (Figure 1).

The SCN are the biological clock in mammals

A number of lines of evidence support the conclusion that the SCN are the site of the biological clock. Lesions of the SCN eliminate rhythmicity of overt physiological and behavioral events, including sleep and wakefulness (Eastman, Mistlberger et al., 1984; Edgar, Dement et al., 1993) . Transplantation of fetal SCN tissue into lesioned animals restores rhythmicity (Silver, Lehman et al., 1990; Aguilar-Roblero, Drucker-Colin et al., 1992) . SCN cells display rhythmic electrical firing, glucose uptake, and neurotransmitter release, in vivo, and in vitro, either as a slice or as dissociated cells (Welsh, Logothetis et al., 1995) . While these latter findings suggest that individual neurons of the SCN are capable of sustaining circadian oscillations, it appears that the nuclei function as an ensemble oscillator, with network connections providing period and phase stability. The cumulative firing rate of the networked oscillator is higher during the day and lower at night, whether an animal is diurnal or nocturnal (Figure 2).

Afferent projections to the SCN

FIGURE 2

The SCN receives major afferent input via three paths, the retinohypothalamic tract (RHT), the geniculo-hypothalamic tract (GHT), and a serotonergic input from the raphe nuclei. Two other inputs, less well described and/or more species dependent, but for which functional correlates have been identified, are a cholinergic projection from the basal forebrain, and a histaminergic projection from mammillary hypothalamus. Each of these five inputs provides a potential point of entry for pharmacological manipulation of clock function.

First, the SCN receive afferent projections from the retina via a monosynaptic pathway, the retinohypothalamic tract (RHT), which arises from a subset of photoreceptors and retinal ganglion cells that appear to be specialized for sensitivity to luminescence (Foster, Argamaso et al., 1993) . The RHT terminate in the ventral portions, or core, of the SCN, an area rich in vasoactive intestinal polypeptide (VIP) and gastrin releasing peptide (GRP) producing cells. The neurotransmitter of the RHT is glutamate, which acts through both NMDA and non-NMDA receptors in the SCN. Light, or glutamate, stimulates Fos expression in the VIP positive neurons of the SCN (Schwartz, 1997) .

The second major afferent is from the intergeniculate leaflet (IGL). The IGL also receives input directly from the retina via a separate branch of the RHT. The IGL, in turn, projects back to the SCN via the geniculo-hypothalamic tract (GHT), terminating in areas that overlap the direct RHT-SCN projections. The IGL pathway is enkephalin and neuropeptide Y positive. In addition to an ancillary role in photic entrainment, this alternate input appears to play an important role in entrainment mediated by non-photic stimuli such as motor activity. Electrical stimulation of the IGL produces phase shifts similar to those produced by activity (Rusak, Meijer et al., 1989) and lesions of the IGL block activity-induced phase shifts (Janik and Mrosovsky, 1994) .

The third major afferent to the SCN originates in the raphe nuclei. Data suggest that this innervation has two functional components; one direct from the raphe (presumably accounting for 5HT axons that are appreciable within the SCN), and a second indirect projection via the IGL (Figure 1). Several lines of evidence have also implicated the serotonergic input in the effects of activity on the circadian clock, but the effects of pharmacological serotonin manipulation on clock function are complex. A few of the demonstrated interactions include a direct suppressive effect of serotonin on SCN neuronal firing (Meijer and Groos, 1988) , and phase-dependent modulation of the SCN response to photic phase shifts (Mintz, Gillespie et al., 1997).

Pharmacological Chronobiotics: Theoretical Considerations

FIGURE 3

By definition, a chronobiotic is an agent that acts on the circadian clock or its inputs to alter timekeeping function. There are three possible modes of action for putative chronobiotics (Figure 3). First, a chronobiotic could act to modulate input from photic afferents to affect entrainment and phase alterations (A). Second, a chronobiotic could act directly on the clock to alter intrinsic period (B). A change in period would manifest as an alteration in the phase relationship between the circadian rhythm and the light-dark cycle under entrained conditions. Third, a chronobiotic could act upon the oscillator to change the amplitude of clock output (C), an alteration that would have complex effects on different driven rhythms. In the case of the sleep-wakefulness rhythm, current models suggest that an augmentation of circadian clock amplitude would produce improved nocturnal sleep and increased daytime alertness. It should be noted that within each of these broad conceptual categories, current understanding of the anatomic and neurophysiological components of the clock allows much more specific dissection of potential sites of action. This degree of detail is beyond the scope of this review. A more detailed treatment of potential mechanisms of chronobiotic action can be found in (Dawson and Armstrong, 1996) .

In addition, with regards to putative chronobiotic agents and their mechanisms of action, it may be very difficult to distinguish a direct effect on the circadian clock from an effect on a driven circadian process. This is particularly true for sleep and wakefulness where the exact nature of the relationship between the clock and sleep-wake processes remains unclear, and where non-circadian, e.g. homeostatic, processes clearly play a pivotal role in the timing of sleep-wake expression. Thus, an agent that potentiates sleep at non-circadian phases might act directly on the clock to either shift its phase or attenuate its amplitude to facilitate sleep. Alternatively, it might act directly on sleep-wake regulatory centers to potentiate sleep and override the circadian wakefulness drive. Even phase-dependence of such an effect does not definitively implicate the clock as the site of action, given that hypnotic compounds may depend on underlying drive for their effect, and this drive may co-vary with circadian phase in an ineluctable way (Trachsel, Edgar et al., 1992) .

Indeed, only direct effects on the period of the circadian oscillation, as assessed in appropriate constant conditions, definitively identify the circadian clock as the site of action. By this restrictive definition, only a handful of pharmacological agents have been identified (see below). Given the complexity of the circadian clock, it is perhaps surprising that neuroactive drugs do not commonly alter circadian timing. Considered from an evolutionary perspective, however, the resilience of the clock makes more sense. Accurate timekeeping by a biological clock requires careful insulation from physiological variations that might produce irregular changes in clock period. In this context, the remarkable insensitivity of the circadian clock in early experiments is less surprising. As noted by Richter in 1965 (Richter, 1965) :

"How can this clock be speeded up; slowed down; reset; or stopped? Blinded rats were subjected to almost every conceivable kind of metabolic, endocrinological and neurological interference... to no effect"

Insulation Of The Circadian Clock

The period, or frequency, of oscillation of the central circadian pacemaker is relatively insensitive to external and internal environmental influences. The period of the circadian oscillator is stable over long periods of time, despite changes in the physiology of the organism. Early in the study of circadian rhythms it was demonstrated that the period of the oscillator is relatively insensitive to temperature changes. While the rate of most biological processes changes 2-3 fold with a 10o C change in temperature (Q10 of 2-3), the period of overt circadian rhythms shows a Q10 of less than 1. Thus, in hibernating mammals, the oscillator continues to keep relatively accurate time even in the face of a drastic decline in the body temperature of the organism (Ruby and Heller, 1996) The molecular basis of this temperature compensation is not known.

Protection of the period of the oscillator also derives from the fact that it is not a function of sodium-dependent action potentials. Infusion of tetrodotoxin (TTX) into the SCN of the rat prevents sodium dependent action and synaptic potentials. In TTX treated animals, light will not cause phase shifts and circadian rhythmicity of overt rhythms is abolished. However, upon wash out of the TTX, the animals resume rhythmicity at a phase that is predictable from the period of the oscillator prior to TTX treatment (Schwartz, Gross et al., 1987) . A similar experiment has been performed in both SCN slices and in dissociated SCN cells (Welsh, Logothetis et al., 1995) , and the results continue to support the conclusion that action potentials within SCN cells are not a part of the clock time keeping mechanism. The input to the clock and the output from the clock are clearly dependent upon action potentials but the actual clock time keeping mechanism does not.

Putative Chronobiotic Agents

Despite the early difficulty identifying environmental and metabolic manipulations with circadian clock effects, there were a few prominent successes. One early example was deuterium, which when fed to hamsters and mice in drinking water, produced a concentration-dependent lengthening of the circadian period (Figure 4; (Suter and Rawson, 1968; Daan and Pittendrigh, 1976) . Several more recent studies have demonstrated that this agent acts at the SCN itself (Lesauter and Silver, 1993) , but there is still no clear understanding of how deuterium exerts its effects on tau.

FIGURE 4

FIGURE 5

The monoamine oxidase inhibitor, clorgyline, also appears capable of altering the circadian oscillator (Figure 5; (Wehr and Wirz-Justice, 1982) ). Clorgyline lengthens the period of the activity rhythm of the hamster. The entrained phase of overt rhythms were also delayed, the expected consequence of the increase in the oscillator period. Localization of this drug effect to the SCN was corroborated in an experiment in which clorgyline was infused chronically into the third ventricles of freely behaving blinded rats. The period of the free-running rhythm in feeding behavior increased (Wirz-Justice, Groos et al., 1982) . Importantly, there was also a substantial decrease in the amplitude of the rhythm, a change that was indistinguishable from arrhythmicity in some animals.

The tricyclic antidepressant, imipramine, also alters circadian rhythms (Figure 5; (Wehr and Wirz-Justice, 1982) ). When administered to rodents, imipramine lengthens the period and decreases the amplitude of overt rhythms in a manner similar to that of clorgyline. This suggests a common biochemical basis of the effect. While both drugs, particularly imipramine, have complex action on multiple neurotransmitter systems, both clorgyline and imipramine increase presynaptic serotonin levels, an effect that also results from lithium treatment. Serotonin may be the common mediator of antidepressant effects on circadian rhythms.

FIGURE 6

Both the SCN and the IGL receive serotonergic input from the raphe. The median raphe projects to the SCN, but not to the IGL while the dorsal raphe projects to the IGL but not to the SCN (Miller, Morin et al., 1996) . Lesions of the median raphe produce alterations in the period and amplitude of circadian rhythms, while lesions of the dorsal raphe do not alter circadian rhythmicity (Miller, Morin et al., 1996) and specifically, do not mimic the effects of IGL lesions, suggesting that the serotonergic inputs to that structure are not necessary to its timekeeping role.

Within the SCN, a circadian rhythm is serotonin uptake and SCN cell response to serotonin has been documented (Meyer and Quay, 1976) (Mason, 1986) , with maximal activity during the dark phase in the nocturnal rat. Direct application of a serotonin agonist, quipazine, to cultured SCN phase shifts the SCN firing rhythm, with advances occurring during the projected day and delays occurring during the projected night (Figure 6; (Prosser, Miller et al., 1990) ). A similar phase response curve to quipazine has been generated in vivo (Edgar, Dement et al., 1993) . Pharmacological studies suggest that the phase advance effect is mediated by a novel serotonin receptor found in the SCN, 5HT7, a G protein coupled receptor that elevates cAMP levels (Lovenberg, Baron et al., 1993) . The opening of calcium dependent potassium channels may also be necessary for serotonin-induced phase advances (Prosser, Heller et al., 1994) and this may also occur through binding to the 5HT7 receptor.

Several lines of evidence also suggest that serotonergic inputs to the clock are involved in non-photic phase shifting. First, locomotor activity, which shifts oscillator phase in a "dark-pulse" pattern, produces an elevation of serotonin (5HT) in the SCN (Jacobs, 1991) . Second, the phase response curves for wheel running activity and quipazine are very similar (Edgar and Dement, 1991) . Third, lesions of the 5HT pathway to the SCN eliminate the phase shifting effects of activity (Kuroda, Fukushima et al., 1997) , and of triazolam, a chronobiotic whose phase shifting effects are dependent on stimulated motor activity (see below; (Cutrera, Kalsbeek et al., 1994) (Penev, Turek et al., 1993) ) Thus, the feedback effects of locomotor activity on the SCN may be at least partially mediated by the 5HT input from the raphe. Other data are not entirely consistent with singular dependence on serotonergic mechanisms, however. Infusion into the SCN of antibodies to NPY block the phase shifting effects of locomotor activity (Biello, Janik et al., 1994) as well. More important, recent studies utilizing systemic and intra-SCN administration of specific serotonin antagonists do not block activity-induced phase shifts (Antle, Marchant et al., 1997) . While all of the available data are consistent, it appears that activity produces phase shifts of the circadian oscillator either directly, via a projection directly to the SCN, or indirectly via a projection to the IGL and thence via a NPY projection to the SCN, or both.

Sex steroids have been demonstrated to have effects on circadian rhythms of motor activity(Morin, Fitzgerald et al., 1977) sleep/wake state, and SCN firing rates (Kow and Pfaff, 1984) . The SCN have sex steroid receptors (Lauber, Romano et al., 1991) . Estrogen is associated with increased consolidation of motor activity and phase advance of activity onset. In female hamsters, for example, the elevated estrogen levels at pro-estrous are associated with an earlier onset of wheel running than on other days of the estrus cycle producing a characteristic "scalloping" in the activity rhythm (Morin, 1980) Castration/replacement studies have shown a small effect on clock function with shortening of free running period (Morin and Cummings, 1981) . There appear to be important species differences. There is not yet direct evidence to support an important effect in humans. Melatonin

One putative chronobiotic receiving extensive attention recently is melatonin. This naturally occurring hormone is secreted by the pineal gland in a circadian rhythm that is controlled by sympathetic innervation from the SCN (Cassone, Warren et al., 1993) (Figure 7). Melatonin levels rise during the night and decline at dawn in both nocturnal and diurnal species (Reiter, 1986) . In mammals, the majority of melatonin receptors in the brain reside in the SCN (Reppert, Weaver et al., 1994) (Weaver and Reppert, 1996) (Roca, Godson et al., 1996) (Reppert, Weaver et al., 1996) . Melatonin effects on the circadian oscillator have been demonstrated in several species, including humans. This relationship, in which the driven rhythm of pineal melatonin secretion feeds back on the oscillator to influence its function, appears to be a mammalian modification of an evolutionarily older system. In lower vertebrates, including birds and reptiles, the pineal is a functional circadian oscillator (Underwood, Barrett et al., 1990) .Thus, interdependence of these two systems in the mammal may have arisen from an older relationship, when both pineal and SCN exerted clock function.

FIGURE 7

Melatonin modulation of the circadian signal of the SCN has been demonstrated in several studies. Melatonin injections can entrain circadian rhythms in rats (Armstrong and Redman, 1985) and lesions of the SCN eliminate the effect (Cassone, Chesworth et al., 1986) . Melatonin can phase shift circadian rhythms in a number of other species, including sheep (Wood, Claypool et al., 1989) , lizards (Underwood and Harless, 1985) , birds (Heigl and Gwinner, 1995) and humans (Lewy, Ahmed et al., 1992) . Melatonin shifts circadian phase in a "dark-pulse" pattern, i.e., opposite that produced by light (Figure 8). Consistent with this, pinealectomized rats phase shift more rapidly to a change in light cycles than do pineal intact animals (Armstrong and Redman, 1985) . Exogenous melatonin administered to rats late in the circadian day (9-11 hours after activity onset) can accelerate the rate of reentrainment to phase advances but not phase delays of the LD cycle (Redman and Armstrong, 1988) (Armstrong, 1989) .The effects of melatonin are more pronounced under LD or LL conditions than under DD conditions (Chesworth, Cassone et al., 1987) (Redman and Armstrong, 1988) .

FIGURE 8

In vitro studies of direct effects of melatonin on SCN suggest that melatonin's effects on the SCN are complex. Melatonin applied directly to the SCN will both phase shift the neuronal firing rhythm within the cultured nucleus (McArthur, Gillette et al., 1991) and suppress the amplitude of neuronal firing (Mason and Brooks, 1988; Shibata, Cassone et al., 1989) (Stehle, Vanecek et al., 1989) (Liu, Weaver et al., 1997) . It is unclear whether the melatonin inhibition of SCN neuronal firing is fundamental to the phase-shifting effect. It has been suggested that melatonin inhibition of neuronal firing alters SCN sensitivity to other phase shifting stimuli (Liu, Weaver et al., 1997) ; a decrease in the firing rate reflects a neuron in a state of decreased sensitivity to phase shifting stimuli. Thus, the net result of pinealectomy would be an increased sensitivity to phase shifting stimuli. In addition, melatonin effects on 2-deoxy-d-glucose uptake in the SCN demonstrated maximal inhibitory effects of melatonin on SCN metabolism at the circadian phase most sensitive to the phase shifting effects of melatonin (Cassone, Roberts et al., 1988) . Therefore, the nighttime elevation of melatonin serves to protect the oscillator from phase shifting stimuli, providing a degree of inertia to the system. Light, however, can over ride the inhibitory modulation of melatonin on the circadian oscillator, with its two fold action of 1) suppression of melatonin secretion from the pineal and 2) powerful phase shifting effects (Liu, Weaver et al., 1997) .

Recent work characterizing the melatonin receptors suggests that the inhibitory and phase-shifting effects may be more distinct than previously thought. Melatonin binds to high affinity G-protein coupled receptors (Reppert, Weaver et al., 1996) . Melatonin binding has been demonstrated in the SCN of adult mice, rats, Siberian hamsters, and humans (Reppert, Weaver et al., 1988) (Siuciak, Fang et al., 1990) (Weaver, Carlson et al., 1990) (Roca, Godson et al., 1996) (Weaver and Reppert, 1996) . Two types of melatonin receptors have been identified in the SCN (Liu, Weaver et al., 1997) , Mel1a and Mel1b. Expression of the Mel1a receptor may be developmentally regulated, at least in some species, as it has been demonstrated that binding to the receptor decreases with advancing age in hamsters and humans (Duncan and Davis, 1993) (Weaver, Stehle et al., 1993) . The Mel1b receptor subtype is expressed at substantially lower concentrations in the SCN, and more sensitive detection methods, e.g. RT-PCR, have been required to identify mRNA for the subtype in mice (Liu, Weaver et al., 1997) .

FIGURE 9

The Mel1a receptor appears to be necessary for the acute inhibitory action of melatonin on SCN firing, as cultured SCN from knock-out mice with disruption of the Mel1a receptor gene do not decrease firing in response to melatonin (Figure 9; (Liu, Weaver et al., 1997) ). These same mice will continue to phase shift in response to melatonin, suggesting that the phase shifting effect of melatonin occurs through the Mel1b receptor and that the inhibitory effect on neuronal firing is not necessary to the phase shifting effect.

Important species differences exist. Mel1b may be absent in other species, including the Siberian and Syrian hamsters (Weaver, Liu et al., 1996) . Yet in these animals, melatonin will elicit phase shifts, suggesting that the Mel1a receptor mediates phase shifting effects in the hamster, but not the mouse. An alternative explanation is that the Mel1b receptor in the knockout mouse assumed the phase shifting function in the absence of the Mel1a receptor. Additionally, an as yet unidentified receptor may be responsible for phase shifting effects. Resolution of the respective roles of each receptor subtype in the physiology of melatonin effects on the circadian system awaits the development of specific pharmacological probes that selectively activate or inactivate receptor subtypes.

FIGURE 10

If it is established that a specific melatonin receptor subtype is responsible for phase shifting effects of the hormone, then specific agonists of that receptor may yield significantly greater phase responses, potentially large enough to overcome the advancing effects of morning sunlight that are most troublesome to night shiftworkers. Conversely, if the Mel1a subtype produces attenuation of SCN output amplitude, as it seems to do in the mouse, specific agonists of this receptor subtype may prove useful for their daytime hypnotic effects. To date, subtype specific agonists have not yet been developed, but recent progress suggests that these tools will soon be available (Gillen, Li et al., 1997) .

An additional caveat to the application of melatonin agonists to shift-work insomnia is the still incomplete understanding of the vascular effects of the hormone. Specific melatonin receptors mediate vasoactive effects of melatonin directly on arterial smooth muscle, which, at least for coronary arteries, is vasoconstrictive (Viswanathan, Laitinen et al., 1993) . As yet it is not known whether subtype-specific agonists will prove selective at vascular receptors as well; preliminary results have isolated message for both subtypes from cerebral arteries of the rat (Easton, Masana et al., 1997) .

Specific serotonin agonists

An alternate approach to the development of chronobiotic agents has been incompletely explored. Consideration of the phase-shifting effects of serotonin agonists suggest that these effects may potentially be quite large, and the "dark-pulse" orientation indicates that agonists could be used to counteract the effects of uncontrolled sunlight exposure. Indeed, simultaneous administration may render the SCN "blind" to light effects (Rea, Glass et al., 1994) . Characterization of a specific 5HT receptor subtype mediating these effects offers the possibility that specific, short-acting agonists could be developed that would avoid the potential toxicity associated with broad spectrum serotonin stimulation.