The hypothesis tested in this study, developed from the findings of Um et al. , was that central activation of AMPK could disrupt the circadian machinery located in the SCN and thus, alter the endogenous timing of the resumption of melatonin secretion by the pineal gland. In our study, such an effect was not observed after i.c.v. treatment with metformin or AICAR, both known activators of AMPK. Surprisingly we found that another component of the endogenous rhythm of melatonin, its amplitude, was altered by these compounds.
There are three main possibilities to account for the rejection of our hypothesis. First, the i.c.v. treatment with either metformin or AICAR was ineffective and did not stimulate AMPK in the hypothalamic structures we examined. Indeed, the similarities of the effects between the i.c.v. injections of both drugs on the amplitude of the rhythm of melatonin secretion suggest that they might be mediated by the activation of AMPK, as a common element along their distinct signalling pathways . However, we do not exclude possible AMPK-independent effects . Second, the treatment effectively activated AMPK within the SCN, but in DD when the SCN is released from entrainment by the prevailing LP produced conflicting signals because of the persistence of entrainment by the preceding LP  and the expected SCN-driven phase shift in the resumption of melatonin secretion. Third, the treatment effectively activated AMPK but not within the SCN or without triggering phase-shifts in the master circadian machinery perhaps because of functional differences  between the master circadian clock and the peripheral-slaves clock that was studied by Um et al. .
The biological relevance of the lowering effects of both metformin and AICAR on the amplitude of the endogenous rhythm of melatonin is difficult to interpret. Indeed, in sheep as in most photoperiodic species, the duration of melatonin secretion is the critical feature of the rhythm of melatonin secretion that transmits photoperiodic information to the reproductive axis ; the role played by the amplitude, the other main feature of this rhythm, remains controversial [27–29]. Some authors have suggested that the amplitude of the melatonin rhythm acts as a potent integrator of environmental factors, other than photoperiod for example, temperature and the food supply (for review see ). In relation to this theory, a complementary study (Menassol et al., unpublished data) has established that feed restriction can alter the seasonal reproductive patterns of the Île-de-France ewe as has been reported in a previous study on sheep , and that this alteration is associated with lower concentrations of plasma melatonin. Therefore we speculate that the central activation of AMPK is involved in the integration of metabolic-related cues that alter the amplitude of the rhythm of melatonin secretion. We speculate further that this effect is mediated by neuropeptides or peptidergic hormones that regulate the amplitude of the rhythm of melatonin secretion such as Vasoactive Intestinal Peptide (VIP), Pituitary Adenylate Cyclase Activating Peptide (PACAP) or Vasopressin (VP) (for review see ) and particularly Neuropeptide Y (NPY) which is activated by AMPK .
The effect of metformin and AICAR on the amplitude of the rhythm of melatonin secretion was not associated with immediate changes in insulin secretion and thus, was not attributable to the interrelationship that exists between these two hormones [11, 12, 34]. Interestingly, some studies have reported a decrease in melatonin concentrations associated with the development of diabetes [35, 36]. Thus, assuming that metformin crosses the blood-brain barrier, one can ask if therapeutic treatment with metformin could not have unreported side-effects on the secretion of melatonin.
The characterization of the expression of AMPK within the hypothalamus, pituitary and pineal glands of the ewes suggests a potentially wide range of action for metformin and AICAR since all subunits were identified in each of the structures that we examined. The differences among structures in the level of expression of the β2 and γ2 subunits may indicate structural specificities in the sensitivity or functionality of the AMPK complex . Indeed, our investigation of the potential central sites of actions of the AMPK complex, point preferentially to the PVN as a structure of particular interest because some of its neurons that express melatonin-regulating peptides project to the pineal gland . In this particular structure, the direct assessment of the phosphorylation rate of AMPK only reveals a trend toward significance (p = 0.1) for a greater activation after i.c.v. injection of AICAR. However the indirect measurement of AMPK activity, trough the assessment of the phosphorylation rate of its downstream target ACC [37, 38], shows that AICAR treatment significantly inhibited ACC activity in both the paraventricular nucleus (p < 0.001) and the pineal gland (p < 0.05). Therefore the PVN is strongly suggested as a key structure that mediates metabolic influences on the rhythm of secretion of melatonin. Moreover, it seems that direct effects of AMPK on the pineal gland activity itself can not be excluded.
In this experiment, the failure to demonstrate significant direct activation of the AMPK complex following treatment with AICAR might be explained by the time elapsed between tissue fixation and injection (1 hr) and the possible action of endogenous phosphatases .