Biological Rhythms in Mood Disorders

Anna Wirz-Justice

Evolution has provided us with a day within, an endogenous template that anticipates the demands of the day without—the circadian system. If temporal order is essential for health, as the appropriate timing of psychological, behavioral, physiological, and hormonal rhythms with respect to the external day–night cycle imply, then temporal disorder should have clinical sequelae. Indeed, it is now well established that certain sleep disorders, such as delayed- or advanced-sleep-phase syndrome, arise from inappropriate phasing of the endogenous circadian clock with respect to normal sleep times. These rhythm disturbances are generally not accompanied by any psychiatric illness. Other sleep disorders, such as those related to shift work and transmeridian flight, arise from sudden shifts of the sleep–wake cycle without concomitant synchronization of the endogenous circadian component. Here there appears to be a closer link with mood disorders.

The remarkable episodic nature of some forms of affective illness as a manifestation of abnormal rhythmicity was first described in detail by German psychiatrists at the beginning of the century. The patterns of periodic recurrence, links with time of year or hormonal cycles, characteristic sleep disturbances, or diurnal mood fluctuations, later stimulated very specific hypotheses, beginning with the original studies of Curt Richter in the 1950s (49). The question as to whether the observed disturbances of biological rhythms in mood disorders reflect an underlying disorder of the biological clock is as yet unresolved. The last decade has seen an unprecedented growth into this aspect of depression research. The circadian system provides an integrative framework for concepts of affective illness involving psychopathological changes, sleep regulation, and neuroendocrine and neurotransmitter mechanisms.

The rare clinical observation of precise periodicity in psychopathology has provided certain clues to an underlying aberration of normal physiological rhythms. What mechanisms could explain the extraordinary stability, unaffected by any drug treatment, of the manic–depressive cycles documented for over 35 years in a single patient (Fig. 1A) (40)? Is there a causal link between rapid cycling and the circadian system? How are mood changes in women susceptible to affective illness linked with the menstrual cycle and the postpartum period? (See Mood Disorders Linked to the Reproductive Cycle in Women.) And most remarkably, how can depressive phases be triggered by time of year? A new diagnostic subgroup of depressive disorders with seasonal pattern (seasonal affective disorder, SAD) has been successfully treated with bright light (51). Prospective weekly depression self-ratings in a SAD patient over three consecutive cycles document this replicable seasonality of recurrence in autumn (Fig. 1B). Without treatment, the depressive phase lasted until spring. Light therapy administered toward the end of the depressive phase in January induced remission, light therapy administered at the beginning of the depressive phase in November prevented its evolution. Light therapy can be considered the most successful clinical application of circadian rhythm concepts in psychiatry to date.

The most extensively described, yet still least understood, rhythmic phenomenon in depression is the so-called classic melancholic symptom of diurnal variation of mood (DV) (see ref. 21 for a review). In the last few years, a number of studies have documented that DV is a frequent phenomenon, but not specific for endogenous depression (DV is also found, for example, in reactive depression, or in SAD). It is not even specific for depression, because it occurs in healthy subjects and in other psychiatric illnesses. The presence of DV is not consistent throughout the course of depression: the type of DV (morning low, evening low, indifferent pattern) can change from day to day in a given patient. In short, DV is surprisingly variable when one considers its prominent position in established diagnostic systems.

There is poor agreement between daily mood self-ratings and retrospective judgment of DV. Diurnal variation of mood is not dependent on clinical state (e.g., depth of depression, or season in SAD patients). There is no systematic relationship between the type of DV and melancholic or atypical depressive symptoms.

New long-term prospective documentation of hospitalized patients by the Groningen research group show that every pattern of DV can be documented (2). The underlying potential for manifestation of DV may be present, but DV is not expressed on all days. The patterns in two individual patients, one predominantly with, and one without significant DV during a 90-day period of gradual clinical improvement, are shown in Fig. 2. This list of rather negative observations leads to the moot question as to the relevance of the phenomenon. Clinically, there remain certain consistencies that underline its importance. One of these is that the propensity to produce DV, and in particular, DV with morning low, is a good predictor for sleep deprivation response.

In recent studies of the circadian rhythm of mood during controlled 40-hr sleep deprivation (the constant routine protocol), we have been able to document the kinetics of mood change very precisely (11, 74) (Fig. 3). The subjects are kept sitting in bed without temporal input during this study, and thus mood ratings are not influenced by outside daily events. It can be seen that on the first day, the depressed patients (SAD in winter) show low morning mood that increases slightly in the afternoon, but that during the night, mood declines again. The minimum at 5 to 6 a.m. appears to be the switch point. There is an almost linear improvement after this time. However, the pattern on the second day is more complex, since an afternoon dip is noticeable, with mood improving again before sleep. From these data, it is clear that measurement of mood at two or three time points each day, under normal conditions, may give misleading information as to the presence of DV. Given that the circadian rhythm of mood is individually somewhat different in its timing adds to the potential variance. However, the sleep-deprivation–induced mood improvement is very clear. This pattern can be contrasted with that of control subjects, whose mood is much higher than that of depressed patients throughout the entire experiment, but whose mood rather declines after sleep deprivation. In summary, these studies provide the first data documenting a circadian rhythm in mood (as opposed to diurnal variation). The extent of DV has been suggested to predict later response to conventional antidepressant drugs. As a working hypothesis, it has been proposed that the presence of DV (or greater variability) is a symptom signaling the propensity for change, that is, the potential for improvement, and preliminary findings do support this concept (21).

Sleep disturbances are inextricably linked with depressive illness (3, 4, 65; see also Psychopharmacology of Anorexia Nervosa, Bulimia Nervosa, and Binge Eating). Insomnia with early morning awakening is most characteristic. However, hypersomnia is also found (e.g., in bipolar patients or in atypical depression with or without seasonal pattern). That the interrelation of sleep and mood in major depression is not an epiphenomenon, is most clearly demonstrated by the rapid and dramatic, albeit usually short-lasting, improvement after total sleep deprivation first described by Schulte more than 20 years ago (56), and found in about 60% of depressed patients (3, 4, 32, 34, 53, 65, 67, 70, 75). The equally rapid return of depressive symptoms after subsequent recovery sleep again suggests a crucial role for sleep regulatory mechanisms in the clinical manifestation of depression. The therapeutic response to sleep deprivation is found independent of diagnostic categories, whether endogenous or reactive, psychotic or nonpsychotic, unipolar or bipolar, late luteal phase dysphoric disorder, schizoaffective or seasonal depression. There is evidence that a propensity for diurnal variation of mood with amelioration in the evening predicts sleep deprivation response (summarized in ref. 21).

Other sleep manipulations that have positive effects are partial sleep deprivation in the second half of the night, rapid-eye-movement (REM) sleep deprivation, or phase advance of the sleep–wake cycle. The variety of these sleep manipulations and their temporal course are summarized in Table 1. These experimental results suggest that the depressive process is sleep dependent and requires that sleep coincide with a sleep-sensitive early morning circadian phase (70). The switch out of depression (and into hypomania and mania) often occurs after a spontaneous sleep deprivation. The time when this switch is most likely to occur is also the second half of the night (67). In the rapid cycler above (Fig. 1A), 81% of 64 switches into mania observed over a 10-year period occurred between 11 p.m. and 8 a.m. half of these between 2 a.m. and 8 a.m. (40).

The phenomenon of relapse after recovery sleep has been studied using naps at different times of day, of different duration and sleep architecture. Although a nap in the morning induced relapse more often than an afternoon nap (suggesting a circadian-dependent factor), no clock time has been found where relapse cannot occur, nor is a depressogenic nap predominantly associated with REM- or non-REM sleep (nap studies are summarized in ref. 4).

Preliminary studies suggest that sleep deprivation can potentiate long-term drug treatment and may even predict it. Conversely, there have been attempts to prolong the sleep deprivation response with adjuvant antidepressants, lithium, or thyroid hormones, repeated partial sleep deprivations, or most recently phase advance of the sleep–wake cycle (4).

In order to consider putative mechanisms underlying these clinical observations, the concepts of circadian physiology and sleep regulation need first to be described.

Mammalian circadian rhythms are driven by a central pacemaker located in the suprachiasmatic nuclei (SCN) of the anterior hypothalamus. The SCN is the only pacemaker in mammalian systems for which convincing evidence exists (neural mechanisms of the mammalian circadian system are reviewed in ref. 25). There may be a second, coupled circadian pacemaker in the retina actively gating the transduction of photic information (48).

The endogenous rhythms generated by the SCN (whose periodicity is slightly different from 24 hr) are synchronized (entrained) to the 24-hr day by "zeitgebers" (regular recurring environmental signals). Light is the major, but not unique zeitgeber, entraining the circadian clock to 24 hr as well as providing information about daylength or photoperiod. Two visual pathways mediate entrainment. First, light stimuli reach the SCN through a direct retinohypothalamic tract (RHT) from specific ganglion cells in the retina. Glutamate is a probable neurotransmitter of the RHT. The second visual pathway is the geniculohypothalamic tract. Neurons of the intergeniculate leaflet of the lateral geniculate complex, which contain g-aminobutyric acid (GABA) and neuropeptide Y, also receive light stimuli from the same retinal ganglion cells and project to the zone of the SCN receiving RHT input (Fig. 4).

An important neural pathway leads from the SCN to the pineal gland, via the paraventricular nucleus (PVN). The pineal hormone melatonin is considered to act as a hormonal transducer of the light–dark (LD) cycle to the rest of the organism. The pineal gland is not a circadian pacemaker in mammals; its major output, melatonin, feeds back on to melatonin receptors in the SCN (25) (Fig. 4). Locomotor activity also feeds back on the SCN, although these mechanisms are still being elucidated (42). Serotonin (5-HT) is the putative neurotransmitter of this nonphotic entrainment pathway from the raphé nuclei via the intergeniculate leaflet. The neural substrates of additional zeitgebers, such as food availability and social factors, and the putative central oscillators on which they may act are not yet known (25).

There are two classical techniques to establish whether a stimulus affects the mammalian circadian clock in the SCN. First, the endogenous circadian period of the rest-activity cycle can be documented under constant conditions without time cues (called free running). Second, the susceptibility to stimuli can be mapped as a phase response curve (PRC). When a stimulus is administered, the subsequent direction and amount of phase shift that occurs is dependent on the time at which the stimulus was given. The PRC to light in nocturnal rodents shows phase delays to light pulses given around dusk and early in the night, and phase advances to light pulses given late in the night and around dawn. The PRC to light in humans is similar, although the timing is somewhat different: phase delays occur when light is administered before the circadian temperature minimum near 5 a.m., and advances when light pulses are given after the temperature minimum (11, 39).

In rodents, administration of a dark pulse yields a PRC with nearly inverse pattern to that of a light pulse. Similar PRCs result from locomotor activity pulses induced by a variety of behavioral methods (e.g., novel situations) or drugs (summarized in refs. 25 and 52). There is, as yet, little data for an activity PRC in humans (64). The PRC to melatonin in rodents shows only a narrow sensitive circadian time in the early evening where phase advances can be induced (25). However, a PRC to melatonin application in humans appears to show nearly opposite patterns to that for a light pulse (36).

These stringent techniques, which yield information about pacemaker characteristics, are difficult to apply in humans. Thus, a compromise must be made to at least attempt to measure phase and amplitude (indirect information about pacemaker characteristics) under optimal conditions. A given overt measured rhythm is modified by many internal and external (masking) factors. Sleep, for example, reduces body temperature and suppresses TSH secretion; this is called internal masking. Food or stress stimulate cortisol secretion, and bright light suppresses, whereas supine position diminishes, melatonin: these are examples of external masking. Masking effects, as evoked physiological responses, are thus superimposed upon the expression of the true endogenous circadian rhythm and may give misleading information. Only recently have adequate methods been established to study unmasked circadian rhythms in humans. These important methodological advances provide the techniques for testing circadian hypotheses of affective illness, for example, dim light to measure the precise timing of melatonin onset (35) or the constant routine protocol to measure amplitude and phase of the core temperature rhythm (11).

The above techniques can provide information about circadian rhythm parameters. The circadian pacemaker also regulates aspects of sleep, and, to investigate the etiopathological role of sleep disturbances in depression, the two-process model of sleep regulation has proved a useful framework (5). In this model, the timing, duration, and architecture of sleep are considered to be determined by the interaction of the circadian pacemaker process C with a homeostatic process S dependent solely on prior wakefulness. The level of process S increases during waking and declines exponentially during sleep. Sleep deprivation augments process S even further (Fig. 5). The level of process S is considered to be an indicator of sleep intensity. The time course of the decline in process S is based on electroencephalogram (EEG) slow-wave activity in non-REM sleep. Process C is assumed to modulate thresholds that define the time of sleep onset and termination.

Given these models, we can consider the evidence for disturbances in rhythmic processes underlying mood disorders in very specific terms, whether of the circadian pacemaker (i.e., process C), whether of a sleep regulatory process S, and/or of the sensory systems transducing information from the external environment to the organism. Since there are multiple zeitgebers, the mechanisms of entrainment are quite complex. Not only may a given zeitgeber have multiple entrainment pathways, but such an external stimulus can have direct masking effects. Thus the circadian system is vulnerable to disturbance at many different levels.

In a remarkably prescient review in 1975, Papoušek was the first to integrate rhythm disturbances in mood disorders within a contemporary framework of circadian rhythm regulation. The risk of a depressive episode was considered in terms of a tempora minoris resistentiae, that is, depression could arise out of those temporal constellations that disturb outer and/or inner synchronization in predisposed individuals (44). This concept was later reformulated; in terms of an "internal coincidence" model, depression occurs when sleep (and its concomitant physiological and endocrine changes) occurs at a certain susceptible circadian phase (70), and, in terms of an "external coincidence" model, the susceptible circadian phase is linked with sensitivity to external stimuli such as light (30). More specifically, a phase advance of circadian rhythms was proposed to be pathognomonic for major depression (70). The circadian models proposed in the 1980s were straightforward in their simplicity, and could elegantly explain much of the then available clinical data. The data supporting these hypotheses have been extensively reviewed (1, 10, 20, 53, 67) and this chapter does not reiterate these reviews nor the individual studies but focuses on conceptual approaches. However, at that time there was no information about unmasked human rhythms in depression, nor was the complexity of the circadian system with its multiple entrainment pathways yet recognized. For this reason we are now at a critical stage with respect to circadian hypotheses: the following brief summaries indicate their present status and suggest directions for further research.

Under nychthemeral conditions, the expression of the circadian pacemaker is masked to varying degrees (depending on the "hand of the clock" that is measured) by activity, sleep, meals, light, and ambient temperature. For this reason, the constant routine (CR) protocol (as first designed by Mills and refined by Czeisler in ref. 11) is considered the best method to measure circadian rhythm amplitude and phase in humans. In the constant routine protocol, patients who are studied in a laboratory controlled for temperature, humidity, and light are given regular small isocaloric meals and remain awake in a sitting position for 25 to 40 hours. This method minimizes exogenous effects and unmasks circadian amplitude and phase. In control subjects, certain parameters (e.g., core body temperature) have been validated as markers of the biological clock (11).

The constant routine is only just being adapted for use in depressive patients. One major drawback exists—and it remains an unresolvable paradox. A total sleep deprivation is intrinsic to the constant routine. Sleep deprivation may alleviate depression, so the very method used to analyze a putative circadian disorder changes the clinical state during the course of study. Thus, the uncertainty principle precludes precise attribution of any modification of the circadian system to a neurobiological correlate of depression. A similar problem has been noted during the difficult free-running studies of depressive patients; in some patients, the release from external time cues and social zeitgebers induced a switch out of depression into hypomania, and thus the circadian period measured could not be unequivocally attributed to the depressive state (69).

One important circadian marker is not masked by the sleep–wake cycle. The major influence on the pineal hormone melatonin is its suppression by bright light (35, 37). As an alternative to the constant routine, measurement of melatonin rhythms under dim light conditions does appear to provide a valid estimate of circadian phase in humans (35). However, the new findings that posture masks melatonin secretion (14) may require that previous findings in depressed patients be reevaluated with respect to posture during sample collection. It is not known to what extent immediate or long-term prior light exposure history modifies melatonin production, and how long these aftereffects of light last. Because there is a wide individually determined range in melatonin secretion levels, it is often difficult to find significant (and functionally meaningful) interindividual differences that can be related to the illness being studied.

Sleep and mood are obviously interrelated in major depression (reviewed in refs. 3, 4, and 65; see also Chapter 153). This is particularly evidenced by the rapid antidepressant effect of sleep deprivation and relapse after recovery sleep (3, 4, 32, 34, 53, 65, 67, 70, 75). Process S is hypothesized to be the link between sleep and depression, whereby the buildup of S during waking is impaired, but can be normalized by extending the length of wakefulness, as during sleep deprivation (Fig. 5) (5).

Process S is considered to be represented by the EEG slow-wave activity in non-REM sleep (5). Few studies have used spectral analysis of the EEG to estimate the time course of process S during sleep and after sleep deprivation in major depression, and the hypothesis is in need of stringent testing (reviewed in refs. 3, 4, 5).

If depression is a circadian disorder, this disturbance could occur at one or many levels: the SCN could have an abnormally short endogenous period or there could be abnormalities of zeitgeber input pathways (e.g., the retina, see Fig. 4) or coupling between circadian oscillatory systems. Very precise experimental protocols are required to differentiate potential mechanisms underlying disturbed overt rhythmic behavior.

One of the very first hypotheses considered was that manic–depressive disorder arose as a beat phenomenon between a free-running circadian rhythm and the entrained sleep–wake cycle (18). Very few individuals have been studied over the long-term to detect such free-running rhythmic components in an otherwise entrained rhythm (31, 46). These early findings have not yet been replicated, probably for reasons of inadequate methodology. Such circadian pathology may be specific for extreme cases of rapid cycling (e.g., see ref. 40), but as yet, evidence for this is lacking. Rather, this circadian abnormality appears to be characteristic of certain blind individuals with a periodically recurring sleep disorder: the endogenous pacemaker free runs and cannot be synchronized to normal sleep times (54). There is, however, no evidence in these blind subjects of concurrent periodically recurring mood changes.

The rare investigation of depressive subjects isolated from time cues has not shown any consistent abnormally short period (16, 69). However, these difficult studies have been made even more difficult to interpret by the fact that the clinical state itself changed throughout the time the subject was in temporal isolation.

The phase advance hypothesis of major depression, first proposed by Papoušek (44), was explicitly defined and tested with an experiment of phase advancing the sleep–wake cycle (70). The few replication studies have also found an antidepressant effect (reviewed in ref. 4). A novel experiment has facilitated the phase advance by preceding it with a total sleep deprivation (4). If replicated, this method may be useful to sustain the antidepressant effect of sleep deprivation.

Phase-advanced rhythms in depressed patients have been reviewed for a large number of neurochemical parameters (67), hormones (see The Serotonin Hypothesis of Major Depression), and temperature (53). The most impressive data for a correlation of phase position with clinical state come from the long-term studies in rapid-cycling manic–depressive patients, where, for example the most advanced phase position of temperature and REM sleep is found just before the switch out of mania into depression (67).

One of the earliest and most consistent rhythm abnormalities in depression has been an early nadir and hypersecretion of plasma cortisol. A recent metaanalysis of cortisol rhythms reiterates the validity of this finding (Van Cauter, personal communication). Important, and new in this metaanalysis, is the detailed differentiation of wave-form characteristics. Interpretation is based on twin studies of cortisol rhythms that permit separation of these specific circadian parameters into primarily genetically controlled, or environmentally influenced. The timing of the nocturnal nadir appears to be under genetic control, as is the proportion of overall temporal variability associated with pulsatility (38). The mean level of cortisol secretion and the timing of the morning peak are environmentally modified. The timing of the cortisol nadir is thus a robust marker of circadian phase in humans. This differentiated reevaluation of cortisol abnormalities in major depression lends certain support to the phase-advance hypothesis, albeit in a modified form.

Conversely, attempts have been made to induce depressive symptoms with a phase delay of the sleep–wake cycle in healthy normal subjects. Modest but reliable mood decrements have been induced. Certain individuals, however, became noticeably depressed (e.g., two out of ten in ref. 59). In the extreme case, a healthy subject committed suicide after a phase-shift experiment (50). He was found to have had abnormal and unstable circadian phase relationships during the baseline period. This suggests that sudden circadian phase shifts can induce depressive symptoms in predisposed individuals. For example, subjects with a history of affective illness have become depressed after a westbound flight, or manic after an eastward shift over several time zones (24).

The hypothesis that a phase-delayed circadian system can be depressogenic has only been applied to winter depression (discussed below). Phase-delayed rhythms are often manifested in phase-delayed sleep, from the night-owl behavior in students to the serious disorder of delayed sleep phase syndrome. No augmented incidence of depression has been associated with this sleep disorder.

It has been suggested that the pathophysiology of depression is characterized not by a specific decrease or increase in neurotransmitter function (such as the monoamine hypothesis originally postulated), but by high variance and instability (57). Similarly, in longitudinal studies of depressed patients, the characteristic rhythm disturbance appears to be instability of phase from day to day, or a continuous shift in phase throughout the course of a depressive or manic episode, not an abnormal phase position per se (46, 67). Phase instability could arise through a short endogenous circadian period too near to the 24-hr external LD cycle to entrain properly, low circadian amplitude, or weak coupling between pacemaker and the LD cycle (e.g., low light perception, disturbed retinohypothalamic tract/geniculohypothalamic tract input).

There are few studies addressing this possibility. Two deserve attention for using spectral analysis on longitudinal data. The diurnal rhythm of temperature was measured in a large cohort of major depressed patients who showed higher phase variability and decreased amplitude than controls (63). Both depressive and manic fluctuations reduced circadian fit; instability of the diurnal rhythm was the main feature. Similarly, another large cohort of major depressed patients showed a normal or delayed phase of the temperature rhythm together with a lower amplitude than in controls (13). The 24-hr component was lower in power, with an increase in ultradian components. This suggests a possible weakening of the coupling processes between internal pacemakers and abnormal sensitivity to environmental information.

Diurnal variation of mood has also been used as a variable. A group of patients with major depression carried out hourly mood ratings throughout the day; they showed both greater mood variability and a higher amplitude of ultradian fluctuations than controls (19). In a group of SAD patients, mood rated six times a day during winter depression showed that DV was more unstable than in controls. Light treatment reduced or eliminated all group differences in both mean level and variability (28).

Comparison of multiple circadian rhythms during depression and after recovery have suggested that blunted amplitude is the main chronobiological abnormality (e.g., see refs. 10, 58, 66). However, the majority of studies have been carried out under normal nychthemeral conditions (i.e., masked conditions). Thus, the often-observed diminution of amplitude in the core body temperature rhythm may be a consequence of a disturbed sleep–wake cycle in depression and not of disturbed circadian clock function (58, 66). This can be illustrated by a study of 24-hr heart rate in untreated major depressives; two groups of patients were distinguished, those with a low amplitude rhythm, and patients without any demonstrable 24-hr rhythm at all (61). This dichotomy may be more apparent than real, since the amplitude of a given rhythm probably lies on a continuum, resulting in a nonsignificant circadian frequency component below a given threshold. Additionally, heart rate is highly masked by activity, sleep, and stress. It is unclear how much of the reduced amplitude is an epiphenomenon of behavioral inhibition during the day and poor sleep at night.

Many measures, not reviewed here (but see e.g. refs. 1, 10 and 67) manifest lower amplitude of diurnal variation in major depression than in control subjects. However, to test the hypothesis of circadian amplitude reduction adequately, requires the methodology of the constant routine and validation of the measured variable as an adequate marker of circadian amplitude.

As yet, no studies have been published on circadian rhythms in major depression measured in a constant routine. The need for such studies are illustrated by data from a single severely depressed, hospitalized, nonseasonal patient (J. Anderson et al., unpublished data). The patient was diagnosed with recurrent major depression and was therapy resistant; he was withdrawn from drug treatment 1 week before the study, and carried out two CR protocols prior to and following treatment for 1 month with light therapy alone. During 2 weeks, light was administered for 4 hr/day over midday (to augment amplitude); during the next 2 weeks light was administered in the evening (to delay phase). However, neither treatment regimen induced any amelioration of his depression, whereas during the sleep deprivation of the CR his depressive symptoms transiently improved. In both CRs, his temperature rhythm showed a robust amplitude for a 60-year-old man. A curious response was found on the second day of each CR after sleep deprivation, in that regulation of the temperature set-point appeared to go awry. Temperature increased markedly above normal values, as did heat production. The reproducibility of such temperature increases concomitant with sleep deprivation response remains to be further explored.

Of course, this example cannot define the role of circadian rhythm abnormalities in major depression. Although somewhat heroic for both patient and therapist, further studies using the constant routine in depressive patients may provide data that test these circadian hypotheses. Additionally, clues to the mechanism of the sleep-deprivation–induced mood elevation may also be found, because the timing of the switch can be narrowed very precisely under these controlled conditions and compared with the timing of psychophysiological events.

Most studies of depressed patients document a diminution in melatonin amplitude (e.g., refs. 10 and 58). However, it is not clear whether melatonin can be used as a marker of amplitude in depression, as opposed to being a validated marker of phase (35). This issue is illustrated by melatonin rhythms measured in the CR in two patients (J. Arendt et al., unpublished data). First, the above described major depressive patient had undetectable salivary melatonin secretion when depressed. After 4 weeks of a 4-hr daily midday light treatment, a melatonin rhythm was clearly present. Yet concomitantly, there was no change in depressed state. Second, an euthymic SAD patient in summer had undetectable saliva melatonin secretion. One week of 4-hr daily midday light treatment in summer also augmented melatonin secretion, but again without any change in clinical state. These examples suggest that light has a direct pharmacological effect to increase the level (and thus amplitude) of melatonin secretion. In the same SAD patient, euthymic after 6 weeks of treatment with the selective 5-HT uptake inhibitor citalopram, even higher melatonin levels were found. Thus three different levels of nocturnal melatonin secretion were present during three separate euthymic episodes in one patient, and two different levels of nocturnal melatonin secretion during two separate depressive episodes in the former patient. These data imply that melatonin levels (and thus amplitude) are genetically regulated but also pharmacologically modified. This also means that the amplitude of the melatonin peak has not been validated as a marker for circadian amplitude.

In many species that have seasonal behaviors (e.g., in reproduction, migration, and hibernation), it is the duration of daylength (photoperiod) that initiates these seasonal changes. The circadian system responds with great precision to these changes in photoperiod, for example with expansion and reduction of the locomotor activity: rest ratio (a:r).

Although humans do indeed manifest a number of seasonal rhythms (reviewed in ref. 33), studies have usually been carried out under artificial lighting conditions, which obscure the natural signals of dawn and dusk. Only recently has a winter-time expansion of nocturnal duration of several parameters been documented in humans (decreased a:r from summer to winter) (68). This remarkable study simulated winter and summer photoperiods by controlling the nights: subjects remained in long or short nights of absolute darkness. Such a defined, completely dark phase is never experienced under normal urbanized conditions. Under these controlled photoperiodic conditions, sleep time, melatonin, and prolactin secretion expanded in winter. These data suggest that humans are indeed susceptible to independent shifts in the timing of dawn and dusk markers with season, that is, a change in the a:r ratio.

An analogy with depression was first made by Kripke (29), based on the extensive knowledge of seasonal mechanisms in rodents. The model of separate, but coupled, oscillators for the dawn and dusk signal was applied to circadian concepts of depressive pathophysiology. Instead of simply looking at circadian amplitude and phase (as in the phase-advance or amplitude diminution hypotheses), it becomes necessary to investigate characteristics of the wave form of a rhythm. It has been unambiguously demonstrated that in rodents, the onset and offset of melatonin secretion can shift independently (23). In women with premenstrual depression, the offset of the melatonin rhythm was earlier, reducing the duration of nocturnal secretion (increased a:r) (45). The metaanalysis of cortisol rhythms in depression indicated that a phase advance was specific for the nadir, not the peak, resulting in a longer duration of nocturnal cortisol secretion (decreased a:r ratio) (Van Cauter, personal communication). This permits an hypothesis that disturbances in the a:r ratio may underlie certain mood disorders. Conversely, in the photoperiod study of healthy subjects, one volunteer became severely suicidal after decrease of the a:r ratio (long nights), even though all the others felt remarkably well on this protocol (68).

Zeitgeber strength is a measure of the capacity of an external entraining agent to synchronize circadian rhythms. Zeitgeber strength is relative, being also dependent on the sensitivity of the system to the zeitgeber. Light is the major synchronizing agent. Additionally, the circadian timekeeping system in animals phase shifts in response to periodic locomotor activity (42). Food availability (6) and social cues can also entrain rhythms in certain species.

In the free running experiments in Andechs, where humans were deprived of time cues, social factors were considered the most important synchronizer (71). Later experiments indicated that light was also a major zeitgeber for human circadian rhythms (71, 72). The two zeitgebers are not independent, because social behavior determines whether a person is exposed to the physical zeitgebers of light and temperature, and the timing of going to bed and waking up (which also gates light input via the retina).

Social factors, as social zeitgebers, have been formalized within a circadian hypothesis of mood disorders (17). This conceptual approach has the attractive feature of linking the biological hypotheses with psychosocial research. Social zeitgebers are personal relationships, jobs, social demands, or tasks that serve to entrain biological rhythms. They determine the timing of meals, sleep, and physical exercise. These social factors also have the potential to disrupt circadian rhythms. Some of the particular psychosocial precipitants of depressive disorder, such as life events, chronic stresses, or lack of appropriate social support systems, may be considered to act as precipitants by inducing rhythm disruptions (e.g., acute changes in the sleep–wake cycle, drop in activity level). Conversely, psychotherapy, as well as social interventions, may act to synchronize the circadian system. Perhaps they incidentally enhance exposure to physical zeitgebers such as light. Given the new knowledge about behavioral feedback on the circadian clock, social zeitgebers could be postulated to act through neural pathways related to arousal.

This phenomenon had been described in qualitative terms by Schulte more than 20 years ago (56): the absence of usual habits and sudden lack of duties, as well as changes in the usual social patterns requiring adaptative behavior, were considered potential precipitants of depressive episodes. He pointed out that depressed patients had two major predisposing factors: the chronicity of sleep disturbances, and their vulnerability to changing external conditions. This he interpreted as possibly resulting from insufficient mechanisms for circadian synchronization.

Experimental tests of social zeitgebers have been carried out using scales developed to quantify these daily social rhythms. For example, in the acute bereavement stage of recently widowed subjects, those individuals with highly disrupted social patterns had higher depression scores than those who maintained social routines (17). Overall social activity scores were negatively correlated with depression ratings in hospitalized major depressives and were lower than for a group of control subjects (60); similar correlations were found in the elderly (47). Remitted depressives showed no differences from controls, but enhanced variability (41), which suggests that low zeitgeber strength may be linked with a certain instability.

We now know that there are multiple zeitgebers for the circadian system, of different relative strengths, that act on and interact with different systems (25). It is not known how much social synchronization is actually related to augmented locomotor activity or increased arousal. If further confirmed, this more complex approach also suggests that a variety of zeitgebers could be used singly, or in combination, to achieve improved synchronization. Application has already been made to improve disordered or reversed sleep–wake cycles in demented patients, using both increased social interaction and bright light (43).

Circadian rhythm studies in depression have primarily measured output parameters. For example, the secretion of melatonin provides a measure of phase of the circadian pacemaker, as well as noradrenergic receptor sensitivity at the pinealocyte. Melatonin is suppressed by light given at night. Melatonin suppression has been used as an index of light sensitivity in depressed patients (e.g., ref. 37).

Measuring sensitivity to light via light suppression of melatonin is rather indirect. A novel, and little-used strategy, would be to measure retinal function itself. Although some aspects of retinal function have been investigated in depressive patients, particularly with respect to modification by drugs such as lithium, there is insufficient research on putative disturbances in retinal sensitivity. The retina is a part of the CNS that is directly accessible to measurement. The accumulating evidence that there is a circadian clock in the mammalian eye and that dopamine and melatonin are important neurotransmitters mediating light and dark, respectively (48), makes this approach all the more attractive.

One of the tenets of circadian physiology has been that the biological clock is immutable. This is not so. Not only is circadian timekeeping modulated by changes in internal state (e.g., by reproductive and thyroid hormones), but also its own output can feed back on its function. This is now clear for the pineal hormone melatonin, because melatonin receptors are selective to the SCN (reviewed in ref. 25, Fig 4). More surprisingly, it appears that the rest–activity cycle can, under certain conditions, feed back on the period and phase of the circadian clock. In hamsters, locomotor activity induced at particular circadian phases (by a drug such as triazolam, by a social stimulus such as an estrus female in the adjacent cage, by a novel situation such as changing the cage or a new running wheel) can phase shift the endogenous rhythm (42) and shorten free-running period in mice and rats (52). These findings lend credence to the idea that alterations in behavioral arousal state or activity level can lead to alterations in circadian period and phase.

Psychomotor disturbances are an important hallmark of depression and mania. The causal relationships between agitation and inhibition in depression and the circadian system are still unclear. Preliminary studies in humans suggest that high-level activity can phase shift circadian rhythms, although this is not yet certain (55, 64). Enhanced activity has been used to test the hypothesis that increased zeitgeber strength could treat SAD patients (bright light also increases zeitgeber strength). Two hours of regular exercise in the early morning decreased depressive symptoms, together with a phase advance of the temperature minimum (26).

Exogenous shifts in timing of the sleep–wake cycle in healthy subjects have been shown to induce dysphoric mood, poor performance, fatigue, and anorexia, as well as mimicking sleep patterns found in depressed patients (early morning awakening and short REM latency) (e.g., 59). After transmeridian travel, the incidence of hospitalization for depression was higher after westward than eastward flight, whereas hypomania occurred more often after eastward shifts (24). Some of the neuroendocrine abnormalities in depression may arise from alterations in circadian rhythm organization, as shown by blunting of evening prolactin and by thyrotropin responses to TRH after reversal of the sleep–wake cycle the response was similar to the normal low response to TRH when given in the morning hours (9).

Healy has postulated that mood-related cognitive and attributional disturbances are sequelae of shifting circadian rhythms (22). This clearly shifts the disruption from being innate to the circadian pacemaker itself, to a secondary level, that of subjective interpretation of internal temporal disorder. In a first test of this hypothesis, healthy student nurses undertaking night shift work for the first time were found to manifest enhanced psychosomatic complaints and negative perceptions of altered neurovegetative function, perceived criticism from others, and less sense of purpose and control (22). In this respect, a further point is important: subjects in free running experiments, whose circadian rhythms of temperature and sleep desynchronized, did not notice that this phenomenon had occurred. They showed no decrement in mood or functioning; on the contrary, they felt rather well (71). This switch into positive mood on internal desynchronization has been used as a model for the switch out of depression into hypomania after sleep deprivation (67, 70). Thus the depressive disturbances concomitant with shift work and jet lag in vulnerable subjects must result from external (i.e., with conflicting zeitgebers) and not from internal dyssynchronization.

Models of depression implicating predisposing and precipitating factors, such as genetic vulnerability, sex, age, chronic stress, the change of seasons, can equally be represented in circadian terms. Some of these biological and psychosocial phenomena relevant to circadian hypotheses of affective illness are summarized in Table 2 (see ref. 20 for details). The previous sections have delineated possible disturbances in temporal order according to the formal properties of the circadian system. To focus on practical application of these principles, the following section describes the example of light therapy, a treatment modality that arose out of circadian models of rhythm disorder.

Extensive research in the last decade has focused on a group of patients who indeed appear to fulfill the criteria of suffering from a rhythm disorder. Seasonal affective disorder (SAD) is characterized by recurrent episodes in autumn and winter of depression, hypersomnia, augmented appetite with carbohydrate craving and weight gain (51). Bright light (more than 2500 lux for more than 1 hr per day) reverses these symptoms within 3 to 4 days (51, 62). We are still far from understanding the mechanisms underlying the pathophysiology of SAD and the therapeutic response to light. At last count, eight explicit hypotheses have been proposed, of which the majority are linked with circadian system abnormalities (reviewed in refs. 51 and 62). These hypotheses utilize the concepts presented above and are briefly summarized as follows:

These various hypotheses are not mutually exclusive. Bright light acts directly on the circadian pacemaker and not on sleep-dependent processes. Thus successful treatment of SAD must act on mechanisms within known retinohypothalamic pathways. Conversely, SAD may result from failure of one or several mechanisms at different neuroanatomical foci to respond appropriately to decreasing daylength. In spite of being a relatively homogeneous group of depressed patients in symptomatology, SAD patients are not necessarily homogeneous in the etiology of their illness. Winter depression may arise from a disturbance in of any of the above hypothesized pathways. Thus, although light therapy was initially based on a specific seasonal hypothesis implicating melatonin suppression, this has been disproved.

Two hypotheses have been systematically addressed under the controlled conditions of the constant routine: whether SAD patients are phase-delayed in winter or whether their rhythm amplitude is diminished. In a 25-hr CR protocol, the circadian rhythm of body core temperature and melatonin onset in depressed SAD patients was phase-delayed in winter compared with age-matched controls (12). Morning light treatment phase-advanced these rhythms concomitant with alleviation of symptoms. In a 40-hr CR protocol in SAD patients, a midday light regimen induced antidepressant response (74). The effects of sleep deprivation and light on the circadian rhythm of mood have been detailed in Fig. 3. Depressed SAD patients showed a tendency for core body temperature rhythms in winter to be phase-delayed compared with controls, but there was no delay in melatonin. There was also no amplitude diminution during winter depression, nor did midday light augment it (Wirz-Justice et al., unpublished data).

Since the discovery that light can treat winter depression, a number of further applications have been subjected to experimental testing (summarized in ref. 62). The success of light therapy in SAD has led to investigation of whether seasonal patterns in other psychiatric disorders could predict therapeutic response to light. Indeed, there appears to be wintertime exacerbation of symptoms in certain patients with bulimia, panic attacks, obsessive–compulsive disorder, and improvement with light treatment. In contrast, atypical depressive patients without seasonality did not show amelioration.

Light treatment studies in nonseasonal depression have been mainly negative (summarized in ref. 62). It may be that the intensity and duration of light application (1 to 2 hr of 2500 lux for 1 to 2 weeks) may not be a sufficiently high dose of light for a clinically relevant effect. Antidepressant drug trials in major depression, particularly in hospitalized patients, require at least 3 to 6 weeks of treatment before any statement about nonresponse can be made. Future studies will need such a conservative time course to adequately test the clinical efficacy of light in nonseasonal depression. In an open pilot study of hospitalized, untreated patients with major depression, 4 to 8 hr of 4000 lux light daily over a period of 10 days reduced depressive symptoms by more than 50% (Graw et al., unpublished data), suggesting the range of intensity and duration that should be tested in a longer controlled trial.

Bright light has been employed as a treatment for certain sleep disorders associated with alterations in the circadian timing system (i.e., advanced- and delayed-sleep-phase syndrome, jet lag, and shift work) (15). Light has also a direct activating effect, which may interfere with sleep initiation (7), but it also improves performance (e.g., 74). Another potential treatment group are the elderly. The age-related decrease of circadian rhythm amplitude, and phase advance of its timing has been suggested to cause, in part, the characteristic sleep disturbances associated with aging (8). Light therapy has been employed in nursing home populations (primarily demented patients) to manage the behavioral disorders and often reversed sleep–wake cycle (reviewed in ref. 8).

In contrast to the light PRC, melatonin induces phase advances when given in the late afternoon and phase delays when given in the early morning (36). This phase-shifting property of melatonin has been applied to treat jet lag and to advance the delayed-sleep-phase syndrome in sighted and blind persons (e.g., see ref. 54). Melatonin has not been very successful in treating depressive symptoms, acting on alertness rather than on mood. Thus the future application of melatonin and its analogs appears to lie in certain sleep and not mood disorders (see also Psychopharmacology of Anorexia Nervosa, Bulimia Nervosa, and Binge Eating).

Rhythm disturbances in depression are heterogeneous. Of course, the heterogeneity of the depressive disorders is a basic classificatory problem in depression research. Yet even in the well-defined subgroup of SAD patients, circadian rhythm characteristics (such as melatonin phase) do not appear unitary and circadian phase is neither correlated with depth of depression nor with preferential response to morning or evening light (e.g., see ref. 73). Additionally, chronobiological characteristics may change throughout a depressive episode (see refs. 46 and 67). This renders many of the findings inconclusive, and contributes to the large variance often observed (patients with similar depths of depressive symptoms may be at different stages in the course of their illness). The necessity for individual long-term study is apparent, particularly to distinguish whether any of these rhythm abnormalities are state or trait dependent. It has not yet been possible to attribute a specific pathogenetic role to the circadian system in depression: a patient's prior history has not been known; the use of antidepressant medication at some stage itself impacts on rhythmic phenomena; the behavioral withdrawal during depression from social and physical zeitgebers augments vulnerability to rhythm disruption; and the majority of investigations have not been carried out under the requisite controlled conditions with adequate sampling intervals and validated markers.

Altered rhythmicity could be either a cause or an effect of altered affective state. Both could independently reflect abnormalities in a third system, such as psychomotor activity. The rhythmic phenomena in mood disorders could be largely a consequence of, rather than a cause of, behavioral depression and mania.

The previous summary of potential foci of rhythmic disturbances have been consciously separated for didactic purposes. Yet a single cause is unlikely. Additionally, a single observation of abnormal entrainment phase could arise from different causes: altered underlying period of the circadian pacemaker, diminished amplitude, the pacemaker's response to light, the strength of the zeitgeber, not to mention pacemaker-independent phenomena such as masking. Separating these causes require methods for determining amplitude and phase of good markers of the circadian system under longitudinal conditions (patients as their own state-dependent control). Yet given the multiplicity of zeitgebers, and the multiple CNS entrainment pathways, the intrinsic interactions may prove to be intransigent. A summary of these multiple foci and their putative role in affective illness has been summarized in Table 2. Not only is vulnerability to circadian dysfunction present at the genetic level, but this is modified during ontogeny, by sex and thyroid hormones, by situative factors such as arousal level and stress, and clinically by drugs.

Circadian rhythm and sleep research have provided the rationale for novel, nonpharmacological therapies of depression (sleep deprivation, light, social interaction). The remarkable temporal course of response and relapse after sleep deprivation (hours) and light therapy (days) is a clue to understanding mechanisms of therapeutic response. Sleep deprivation is antidepressant in all diagnostic subgroups. In contrast, light is antidepressant in SAD but appears to have little effect in other forms of depression. This would suggest that sleep deprivation is a more generalized intervention in the depressive process than light. In the two-process model of sleep regulation, sleep deprivation is postulated to act via process S, with little effect on the circadian pacemaker. In contrast, light acts directly on process C and not on sleep dependent processes. Thus further investigation of the neuroendocrinological and neurobiological changes around the switch out of depression (see Fig. 3), and during recovery sleep or depressogenic naps are required.

A recent, excellent review of those animal models that purport to mimic features of depression, as well as involving dysregulation of circadian rhythms and monoaminergic and cholinergic neurotransmitter systems, is summarized here (see ref. 52 for references; also Animal Models of Psychiatric Disorders).

One of the most researched animal models of depression is that of stress-induced behavioral change. In rhythm studies, stress has resulted in failure to entrain to a light–dark cycle, 48-hr days, lengthening of free-running period, and long-term reductions in the 24-hr mean activity level. A second strategy yields findings suggesting that strains selected for specific behavioral or neurochemical properties also manifest altered circadian rhythmicity (e.g., the Flinders sensitive rat strain has both an up-regulated cholinergic system and short circadian period and advanced phase of temperature and REM sleep). A third strategy analyzes circadian rhythm sequelae of specific lesions (e.g., olfactory bulbectomy in rats results in delayed entrainment phase and lengthening of free-running period).

In both depressed patients and experimental animal models, alterations in phase, period, amplitude, 24-hr mean, and coherence of rhythms have been documented, although cause-and-effect relationships have not been established. In depressed patients, both circadian rhythm phase advances and phase delays have been observed and, in experimental animal models, both lengthening and shortening of the free-running period. It is not clear what clinical or neurobiological characteristics distinguish these two groups, for example, psychomotor disturbances or anxiety. The situation becomes even more complex when comparing circadian period and locomotor activity level. Manipulations that diminish or augment activity do not induce predictable and parallel changes in free-running period.

It is often forgotten that even within nocturnal rodents, important species and even strain differences are present: for example, in drug metabolism, sensitivity to stress, or stability of the circadian system. In the course of evolution diverse mechanisms subserve the same phenomenon in different species. Thus any parallels with the human circadian system (and the human is not usually a nocturnal animal) require this caveat. To cite Zucker (76)

The somatic factors of depression (changed appetite and weight, disturbed sleep and psychomotor activity, loss of interest in sex) are intrinsic to the illness. Nuclei in the hypothalamus regulate the physiological functions of sleep, reproduction, weight, food choice, and their daily and seasonal timing. Circadian and seasonal rhythms in sleep, weight, and appetite are intrinsically linked with energy expenditure and energy conservation. These symptoms can be explored and manipulated in both depressive patients and experimental animals in the search for mechanisms. In spite of the temptation to search for a global animal model of depression, it may be more realistic to focus on particular behaviors.

An example of this approach is the model of seasonal hibernation in hamsters, which provided the rationale for initiating light therapy in seasonal depression (51). Although this model does have certain face and predictive validity, it is oversimplified: traits, not individuals, are photoperiodic (76). It is not legitimate to use winter weight gain in hamsters as a model of seasonal weight gain in SAD patients, unless the functions of weight gain are similar in both species (76). Nevertheless, playing with these fruitful analogies is not forbidden! For example, we have focused on the specific SAD symptom of enhanced carbohydrate-rich food intake in autumn, and the gender differences in the incidence of SAD, to postulate a link with classic serotonergic hypotheses of mood disorder (27).

The most rapid rate of decreasing CNS 5-HT turnover occurs in autumn, and the most rapid rate of increasing 5-HT turnover occurs in spring (33). Carbohydrate selection is driven by 5-HT in the PVN and can be reversed by serotonergic agonists (27). In a variety of animal studies (e.g., in ref. 25), it has been shown that 5-HT release occurs in the dark phase; light can modify sensitivity of the SCN to 5-HT, and 5-HT can influence photic responsiveness of the SCN. When a hamster in a summer photoperiod is depleted of 5-HT, behavior is switched into the winter mode with a reduction of the a:r ratio (nocturnal activity onset occurs earlier). Chronic 5-HT depletion can modify entrainment, decrease amplitude, and increase irregularity of the rest–activity cycle. Finally, 5-HT function is modified by gonadal hormones. This brief summary provides hints for serotonergic mechanisms underlying the exaggerated seasonal rhythms in carbohydrate intake in SAD patients, as well as the vulnerability of women to suffer from winter depression.

Input and output pathways from the SCN suggest complex neuroanatomical feedback loops, mediated by a variety of neurotransmitters and neuropeptides (reviewed in ref. 25) (simplified in Fig. 4). Thus, there are multiple loci for circadian rhythm modification by psychopharmacological agents. Psychoactive drugs have become tools to dissect neurochemical substrates of circadian pacemaker function, define neurotransmitters coding for light entrainment, as well as to test hypotheses of circadian rhythm mechanisms mediating antidepressant drug action. The following summary is condensed from the review by Rosenwasser (52).

An emerging neuropsychopharmacology of circadian rhythms indicates that drug effects fall into two main categories of PRC: mimicking either the effects of a light pulse (carbachol, clonidine) or of a dark pulse (glutamate, neuropeptide Y, benzodiazepines, muscimol, serotonergic agonists, protein synthesis inhibitors). A variety of nonphotic and nonpharmacological stimuli also mimic the dark PRC, some mediated by increased activity.

Several antidepressant drugs have been given chronically to hamsters and rats to measure their effects on free-running circadian period and entrained phase position. The early studies on lithium indicated period lengthening and phase delay in most (but not all) species. Rolipram, putative antidepressant, also lengthened period. Rubidium, another alkali metal with antidepressant potential, and valproate, used prophylactically in some bipolar patients, shortened circadian period. Later, clorgyline, an irreversible, selective monoamine oxidase-A (MAO-A) inhibitor, was extensively studied in hamsters, where it also lengthened period, delayed phase, and increased the a:r ratio.

A wide range of antidepressant drugs have been studied in this classical circadian paradigm, but they have failed to reveal a common action on the circadian clock. Imipramine, clomipramine, fluoxetine, citalopram, levoprotiline, and a selective MAO-B inhibitor were without systematic effect. Only desipramine (a selective norepinephrine uptake inhibitor) and moclobemide (a reversible and selective MAO-A inhibitor) shortened circadian period and increased overall locomotor activity and the circadian amplitude of the activity rhythm. This is consistent with the hypothesis that monoaminergic neurotransmitters play an important role in the control of behavioral state and circadian rhythmicity.

Delayed entrainment phase and marked lengthening of free-running period have been induced by chronic treatment with methamphetamine. Monoaminergic agents have also been shown to induce complex dissociations or splitting of free-running rhythms that may reflect weakening of coupling relationships within a network of mutually coupled circadian oscillators.

The effect of psychopharmacological drugs on retinal function requires mention. The retina, itself a part of the CNS, transduces light–dark information to the rest of the brain and may itself contain a circadian pacemaker; dopamine and melatonin are the putative neurotransmitters mediating light and dark signals (48). Thus a given drug can modify retinal sensitivity as well as circadian function: this has been most clearly shown for lithium, clorgyline, and methamphetamine (48).

Finally, a related application of circadian rhythm concepts not reviewed here is that of chronopharmacology. Studies that focus on the optimum time of day for giving the lowest dose, to provide maximum efficacy with minimum side effects are still rare, but may aid in strategies for therapy-resistant depression.

The last decade has seen an unprecedented growth in circadian rhythm neuroscience (reviewed in ref. 25). Important has been the identification of the genetic basis of period. Mutations of period genes in Drosophilia result in long or short circadian period, or arrhythmic behavior. In the mutant hamster, circadian period can be transplanted. It is to be hoped that further knowledge of the molecular biology of the clock may lead to understanding mechanisms underlying putative clock disturbances in humans. For example, the use of immediate early gene expression (such as the protooncogene c-fos) as a very specific functional marker of neuronal activation, has been applied to the circadian timing system. This specificity is most remarkable light pulses induce c-fos expression only in those regions involved in the photic entrainment of circadian rhythms (the retina, SCN, and perhaps the intergeniculate leaflet), and only at those circadian times when light induces a phase advance or phase delay. Thus c-fos localization may be an important tool to localize pathways involved in photic entrainment, as well as phase shifts induced by activity pulses or drugs (25).

The critique of methods has focused on the necessity for a new generation of circadian rhythm studies in affective illness. The importance of dim light and posture for correct interpretation of melatonin rhythm characteristics (as well as documentation of prior lighting history), has been emphasized. Prior light exposure probably also determines the amplitude and phase of other circadian rhythms, although this has not yet been systematically investigated. The majority of markers of the circadian pacemaker are highly vulnerable to masking effects, thus require a constant routine protocol for correct measurement. The blunted rhythm amplitude documented for many parameters in depressive patients cannot be considered more than an epiphenomenon of the activity– rest cycle unless replicated under stringently controlled conditions.

Validated markers now exist for the output of the circadian pacemaker in normal subjects under such controlled conditions (e.g., core body temperature). Different mathematical methods permit estimates of the timing of the temperature minimum and maximum and the rhythm amplitude (e.g., see ref. 11). Melatonin secretion can be measured under dim light conditions, and provide phase estimates of onset, peak, and offset, as well as duration of secretion and amplitude of the fitted peak (35, 68). Cortisol nadir, onset of the nocturnal rise, peak timing, 24-hr mean and relative amplitude, as well as ultradian frequencies of pulsatile secretion, can also be differentiated (38). Thus simple description of a single parameter (e.g., phase advance of the temperature rhythm) is now insufficient to describe the complex changes that occur (e.g., phase advance of the nadir of the cortisol rhythm with increased pulsatility and change in a:r ratio).

We live in a world that is no longer dominated by the day–night cycle or seasonal rhythms. A 24-hr society impacts on circadian rhythm integrity through low exposure to adequate light levels or shift work schedules. A society that can choose its seasons (jet to tropical beaches to avoid the long dark winter nights) is no longer synchronized to the natural ebb and flood of daylength or temperature. Artificial light, a bonus for a dynamic consumer society, simulates a summer day throughout the year, yet this lighting is of insufficient intensity to truly synchronize. Too little is known of the sequelae of such irregular patterns of light and temperature exposure on a vulnerable circadian system, and how light or temperature could trigger or alleviate a depressive phase. Genetic predisposition, hormonal fluctuations, environmental stress, and altered light–dark cycles could all induce rhythm disturbances. Conversely, altered sleep, arousal, mood state could feed back on to the circadian system.

The heuristic value of basic animal research directed at understanding temporal organization in animals has been to provide paradigms for conceptualizing and manipulating human biological rhythms. "The worst that can happen to a scientific hypothesis is that it is ignored" (5). The field of chronobiology and related sleep research has stimulated innovative circadian rhythm models of depression and has provided novel, nonpharmacological therapies.

Postmodern eclecticism has also reached psychiatry. As in other neurotransmitter-based hypotheses, explanations in terms of single causes, linear function, and reductionist simplification have been replaced by interacting, nonlinear dynamical systems, vulnerability thresholds, and efforts toward a more sophisticated bridge linking intrinsic and external phenomena.

Leave of absence at the University of Cambridge is gratefully acknowledged. I thank Alexander Borbély, Hans-Joachim Haug, Alan Rosenwasser, and Michael Terman for comments. The author's studies on SAD patients were supported by the Swiss National Science Foundation.

published 2000