Clinical studies provide clear evidence for a critical role of circadian rhythm and sleep disturbances in the pathophysiology of mood disorders, which are also closely linked to another biological marker of MD, the dysregulation of the HPA axis (for reviews see 
. Applying a selective breeding approach, we developed an animal model that resembles the deviation in sensitivity to stressful encounters 
. The aim of the present study was to investigate this ‘stress reactivity’ mouse model with respect to the clinically relevant endophenotypes of rhythmicity and sleep disturbances.
We found significant differences between HR, IR and LR mice regarding their circadian rhythm of psychomotor activity and GC secretion as well as pronounced alterations in their sleep-EEG profiles. HR mice for instance showed increased wakefulness, locomotor activity and exploratory behaviours towards the end of the resting period. Moreover, the amplitude of the circadian GC rhythm was reduced due to elevated trough levels and the proportion of REM sleep was clearly increased in these animals. NREM sleep and SWA on the other hand were reduced in comparison to the other two lines. No major rhythmicity differences were found between IR and LR mice, except for a significantly higher proportion of slow wave sleep across the day in LR animals.
In the experiments addressing the behavioural activity rhythms of the animals our results revealed significant differences in the diurnal activity patterns of the three mouse lines. In general, as expected for nocturnal rodents, all animals were more active during the dark phase than during the light phase, but compared to the other two lines, HR mice showed a marked increase in activity towards the end of the light phase, i.e. some hours before the light-dark transition. This increased psychomotor activity during the resting period was found in the analysis of locomotion (see ) as well as exploratory behaviours (see ) and can be interpreted as resembling the symptoms of sleep fragmentation and early morning awakenings often seen in melancholically depressed patients 
. This interpretation is also supported by our sleep-EEG data, including a detailed event related analysis (see discussion
below). The fact that LR mice did not differ considerably from IR animals with respect to their behavioural activity rhythm is also in accordance with clinical findings, as MD patients with atypical features are not reported to suffer from sleep continuity disturbances or restlessness 
Concerning the diurnal variation of HPA axis activity, i.e. the circadian rhythm of GC secretion, similar differences between the three mouse lines were found, as observed for the behavioural rhythms. Again, a clear pattern of increased GC concentrations (measured as faecal CM) during the activity period and relatively low levels during the resting period (including a trough at the beginning of the light phase) were observed in all animals (see ). This typical cycle of nadir and peak concentrations is very much in accordance with published data on laboratory rats (plasma samples 
; faecal samples 
) and mice (plasma samples 
; faecal samples 
). However, compared to the other two lines, HR mice showed clearly elevated concentrations of faecal CM during the entire light phase as well as at the end of the dark phase, resulting in a markedly flattened diurnal rhythm (see ). IR and LR mice, on the other hand, did not differ very much, although LR animals tended to have lower CM levels across the 24-h light-dark cycle (see ). These findings further support the close association between HPA axis activity/reactivity and disturbances of neuroendocrine rhythms, as for example very similar alterations, including a reduced amplitude in circadian cortisol secretion patterns, elevated trough cortisol levels and increased 24-h means, have been found in patients suffering from melancholic or psychotic depression, both of which are characterized by a strong increase in HPA axis drive 
. Interestingly, data available for atypical depression suggest no change or a slight decrease in trough cortisol levels 
, indicating similarities with the phenotype observed in the LR mouse line (see also 
). Although our findings match reasonably with these clinical observations, it should be highlighted that in rodents, the entire human syndrome of MD cannot be modelled, but they may share core symptoms of the disease, including the molecular pathways underlying key endophenotypes.
Potential mechanisms that might be involved in bringing about the described alterations in the circadian GC rhythm of our mouse lines include variations in the activity of neural networks (assessable as brain glucose metabolism differences across times of day) as well as abnormal levels or patterns of noradrenalin and melatonin secretion 
. Furthermore, neurodegenerative processes, particularly in structures participating in the regulation of the HPA axis such as the hippocampus, might be an important factor, as similar disturbances in the diurnal variation of GC have been reported in Alzheimer's and Parkinson's disease patients as well as in experimental models of prion disease 
. The deterioration of the circadian rhythm is interestingly often observed before other clinical symptoms are manifested and can be indicative of a relapse in the case of MD. Therefore alterations of the circadian rhythm appear to be closely linked to the body's stress system and might have a significant impact for a number of pathologies, including MD (for reviews see 
Genotyping efforts as well as studies addressing changes in brain neurotransmitter and neuromodulator systems (including CRH, serotonin and noradrenalin) are currently underway, shedding light on the molecular underpinnings of the endophenotypes observed in the HR/IR/LR mouse lines. Potentially, this pre-clinical research will also yield novel insights into the fundamental mechanisms involved in the pathophysiology of human diseases.
As outlined above, sleep abnormalities are very common symptoms of MD patients and have been in the focus of researchers for several decades (for reviews see 
). Sleep is typically divided into NREM sleep and REM sleep episodes; the former can be further subdivided into sleep stages I–IV in humans. Stage I sleep, the transition from wakefulness with its mixed frequency activity and dominant alpha waves (8–12 Hz) to shallow sleep, is marked with dominant EEG frequencies of 4–7 Hz (theta waves). Sleep spindles with frequencies of 12–15 Hz and K-complexes are hallmarks of stage II sleep 
. In sleep stage III, delta waves with a frequency of around 1-3(4) Hz, so-called SWA, are present and become increasingly dominant in stage IV sleep (referred to as slow wave sleep). REM sleep on the other hand is characterised by a desynchronised EEG (similar to wakefulness) and episodic erratic movements of the eyes together with low amplitude electromyogram activity 
In healthy adults, NREM sleep and REM sleep normally alternate periodically through the night starting with around 90 min of NREM sleep, followed by a short REM sleep period of approximately 10 min. This cycle is then repeated four to six times during the night, with decreasing portions of sleep stages III and IV and increasing durations of the successive REM sleep periods towards the end of the night 
. In depressed patients, however, increased stage I sleep, decreased stage III and stage IV sleep, shorter NREM sleep duration, insomnia (involving difficulties falling asleep, sleep fragmentation and early morning awakenings) are often reported 
. In addition, common sleep-EEG alterations include decreased REM sleep latency, increased REM density 
and increased total time spent in REM sleep 
. It has to be noted, however, that these sleep alterations are not uniformly found across all MD patients. In particular, when the different subtypes of melancholic and atypical depression are considered, the emerging picture is different. Melancholic depression, for instance, is characterised by the aforementioned alterations, including poor sleep quality and decreased amounts of sleep, whereas in atypical depression poor sleep quality is rather associated with an overall increased amount of sleep and fatigue-like behaviour during the day 
Interestingly, our findings from the sleep-EEG recordings in HR, IR and LR mice also support this dichotomy of symptom clusters linked with diametral differences in HPA axis reactivity. HR mice were found to have more bouts of wakefulness during the normal resting period of the animals (see and ) and also showed a significant reduction in the amount of NREM sleep at several experimental time points (see ). An extensive event related analysis (applying the ‘event-history-analysis program’ developed by Alexander Yassouridis 
) additionally supports the notion of a shallower and more fragmented sleep in HR mice, as the number of awakenings and stage shifts, particularly from REM sleep to wake, was clearly increased during the light as well as during the dark phase in this mouse line (Fenzl and Touma et al., in preparation). These differences in sleep architecture might be attributed to the increased activation of the HPA axis across the day in the HR mouse line (see discussion
above and ). CRH is known to impair sleep and enhance vigilance, thereby suggesting a causal relationship between shallow sleep and the hyperactivity of the HPA system in melancholic depression 
. Other preclinical studies also support this view. In rats, after intracerebroventricular administration of CRH, waking was enhanced, whereas alpha-helical CRH (a specific CRH receptor antagonist) reduced spontaneous waking 
The most pronounced differences between HR, IR and LR mice, however, were found regarding the amount of REM sleep. At the majority of time points during the animals' normal resting period, HR mice spent much more time in REM sleep than the other two lines (see ). Human sleep data suggest that changes in REM sleep, mediated by the noradrenergic, serotonergic and cholinergic systems, are not only a consequence of depression, but can be seen as true endophenotype of the disease (reviewed in 
). Interestingly, in a transgenic mouse model overexpressing CRH in the entire brain, REM sleep was also significantly enhanced 
, along with a clearly increased responsiveness of the HPA axis to stressors and alterations in emotional behaviour 
, hence largely overlapping with our observations in HR mice (see also 
). Other animal studies as well as clinical findings further support the notion that CRH promotes REM sleep 
, although the effect of CRH on REM sleep seems to be site- and dose-dependent 
. Moreover, our findings are in line with results of sleep investigations performed in different animal models of depression such as exposure to chronic unpredictable stress 
and selection for increased ‘helplessness’ in the tail suspension test 
. These studies revealed very similar alterations in sleep/wake patterns, distribution of sleep stages and facilitation of REM sleep as we saw in the HR mouse line, again underlining the significant impact of stress responsiveness on sleep architecture.
In similarity to REM sleep, significant differences between HR, IR and LR mice were found in the proportion of slow wave sleep. Virtually across the entire light-dark cycle, HR mice showed a dramatically lowered level of SWA, while higher SWA was observed in LR animals (see ). Sleep deprivation studies indicate that SWA reflects sleep intensity, as it was clearly increased as a function of waking 
. In other words, SWA can serve as a distinct marker for homeostatic sleep pressure 
. The regulation of SWA itself was proposed to be a function of the ‘Two Process Model’ 
, depending on the interaction of processes S (sleep dependent) and C (circadian). Sleep propensity, increasingly depending on extended time spent awake, is reflected by process S. In this model, the sleep intensity (process S) is at its maximum at sleep onset, declining during consecutive sleep. It is beyond the scope of this study to reveal whether the decreased amounts of NREM sleep can be attributed to attenuated levels of SWA, but this would implicate that reduced SWA is an intrinsic sleep-physiological feature of the HR mouse line, which might be brought about by a chronic activation of the CRH system 
. Interestingly, clinical studies also report a reduction in SWA in depressed patients 
, although slow wave sleep is not reduced and REM sleep parameters seem to be less consistently altered in patients with atypical depression 
Taken together, our study provides clear evidence for a critical interaction between HPA axis dysregulation and rhythmicity disturbances, including changes in behavioural activity patterns, circadian GC secretion and sleep architecture. In our mouse model, hyper-responsiveness to stressors was associated with psychomotor activity alterations, resembling the restlessness, sleep discontinuity and early awakenings commonly observed in melancholic depression. Furthermore, HR mice also showed neuroendocrine abnormalities such as reduced amplitude of the circadian GC rhythm and elevated trough levels, potentially mimicking similar symptoms in MD patients. The sleep-EEG analyses revealed changes in NREM and REM sleep as well as SWA in HR mice, indicative of reduced sleep efficacy and REM disinhibition, which reasonably overlap with observations in melancholically depressed patients. Thus, by selectively breeding mice for extremes in stress reactivity, clinically relevant endophenotypes of MD can be modelled, presumably including the symptomatology and pathophysiology of specific subtypes of depression.
It should be emphasized, however, that animal models will only be able to mimic certain aspects of the human disease biology rather than the entire clinical syndrome and that not all features of our SR model match with findings in MD patients. Limitations to the clinical relevance of the HR/IR/LR mouse lines for instance include that HPA axis dysregulation is currently not one of the critical diagnostic criteria for MD and that it is a genetic model, i.e. the differences in stress responsiveness are already present early in life, thereby potentially influencing developmental processes that shape the respective endophenotypes. On the other hand, also in humans, the latter mechanisms (driven by both genetic and environmental factors) might represent key variables underlying individual vulnerability to psychiatric disorders 
Therefore, we are convinced that elucidating similar aspects of biological alterations in animal models and human patients can be a major progress and that translational approaches using appropriate animal models can substantially further our understanding of how organisms respond to stress and the nature of inter-individual differences in the stress response. Given the importance of rhythmicity and sleep disturbances as biomarkers of MD, both animal and clinical studies on the interaction of behavioural, neuroendocrine and sleep parameters may reveal molecular pathways that ultimately lead to the discovery of new targets for antidepressant drugs tailored to match specific pathologies within MD.