BackgroundMares have an annual reproductive rhythm, with a phase of inactivity in midwinter. The aim of this study was to determine the impact of food restriction on physiological and metabolic hallmarks of this rhythm.MethodsOver three successive years, 3 groups of 10 mares were kept under natural photoperiod. A 'well-fed' group was fed to maintain the mares in good body condition; a 'restricted' group received a diet calculated to keep the mares thin and a 'variable' group was fed during some periods like the 'restricted' group and during some other periods like the 'well-fed' group, with the aim of mimicking the natural seasonal variation of pasture availability, but a few months in advance of this natural rhythm.ResultsWinter ovarian inactivity always occurred and was long in the restricted group. In contrast, in the 'well-fed' group, 40% of mares showed this inactivity, which was shorter than in the other groups. Re-feeding the 'variable' group in autumn and winter did not advance the first ovulation in spring, compared with the 'restricted' group. Measurements of glucose and insulin concentrations in mares from the 'restricted' group during two 24 h periods of blood sampling, revealed no post-prandial peaks. For GH (Growth hormone), IGF-1 and leptin levels, large differences were found between the 'well-fed' group and the other groups. The glucose, insulin, GH and leptin levels but not melatonin level are highly correlated with the duration of ovulatory activity.ConclusionsThe annual rhythm driven by melatonin secretion is only responsible for the timing of the breeding season. The occurrence and length of winter ovarian inactivity is defined by metabolic hormones.