Hugues Dardente, Physiologie de la Reproduction et des Comportements, INRA UMR85, CNRS UMR6175, Université de Tours, Haras nationaux, 37380 Nouzilly, France (e-mail: firstname.lastname@example.org).
Most mammals living at temperate latitudes exhibit marked seasonal variations in reproduction. In long-lived species, it is assumed that timely physiological alternations between a breeding season and a period of sexual rest depend upon the ability of day length (photoperiod) to synchronise an endogenous timing mechanism called the circannual clock. The sheep has been extensively used to characterise the time-measurement mechanisms of seasonal reproduction. Melatonin, secreted only during the night, acts as the endocrine transducer of the photoperiodic message. The present review is concerned with the endocrine mechanisms of seasonal reproduction in sheep and the evidence that long day length and thyroid hormones are mandatory to their proper timing. Recent evidence for a circadian-based molecular mechanism within the pars tuberalis of the pituitary, which ties the short duration melatonin signal reflecting long day length to the hypothalamic increase of triiodothyronine (T3) through a thyroid-stimulating hormone/deiodinase2 paracrine mechanism is presented and evaluated in this context. A parallel is also drawn with the golden hamster, a long-day breeder, aiming to demonstrate that features of seasonality appear to be phylogenetically conserved. Finally, potential mechanisms of T3 action within the hypothalamus/median eminence in relationship to seasonal timing are examined.
Neuroendocrine mechanisms of seasonal reproduction
Subsequent to their initial domestication approximately 12 000 years ago, man has selected sheep for economically valuable phenotypic traits such as an increased production of meat, wool, milk or a higher prolificacy (1–3). As a result, there are over 1000 different breeds of sheep worldwide (http://www.sheep101.info), which display considerable genetic heterogeneity. Sheep are seasonal breeders that exhibit recurrent physiological alternations of reproductive activity and sexual rest under natural conditions. The photoperiod (PP; length of the daily light phase) is the main driver of these profound physiological changes. There are, however, large variations in the degree of photoperiodic responsiveness of various domesticated (e.g. Suffolk, Merino) and feral breeds of sheep (e.g. Soay or mouflon), with a general trend towards a reduction in the amplitude of the seasonal reproductive cycle and an extension of the breeding season with domestication (4–7).
For example, in Suffolk ewes raised at the latitude of Ann Arbor (42°N; MI, USA), the breeding season typically lasts approximately 5.5 months (approximately 165 days), from early September until mid-February (Fig. 1) (8). The anoestrus season spans the remainder of the year (approximately 6.5 months, or 200 days). Because of the timing of its seasonal reproduction, the sheep is considered a ‘short-day breeder’. These swings in reproductive states are mediated by PP through melatonin, a hormone produced mostly by the pineal gland and whose duration of secretion is directly proportional to the length of the night (9–12). Artificially controlled alternations of long days (LD) and short days (SD) are sufficient to entrain cyclical modifications in the endocrine status, similar to those of the seasonal rhythm under natural conditions (6,13,14). It has systematically been found that the transfer from SD to LD quickly prompts the arrest of cyclicity (approximately 30 days), whereas exposure of anoestrus ewes (LD state) to SD elicits a return to cyclicity after a longer duration (often between 60 and 90 days) (9). However, this longer latency has been mostly observed in ovariectomised, oestradiol-implanted ewes (see below), and may not reflect what occurs in intact animals. Indeed, Soay rams transferred from LD to SD show rapid (within weeks) switches in endocrinology (15,16). The photoperiodic system is temporally flexible because submitting animals to alternations of 3 months of SD and LD can yield two breeding seasons per year (13,17).
From the standpoint of endocrinology, ewes (or rams) display high prolactin (PRL) levels under LD (500–800 ng/ml versus 20–30 ng/ml under SD), whereas luteinising hormone (LH) pulse frequency is high under SD and low under LD. The breeding season of the ewe is characterised by regular 17–19 days oestrus cycles with a long luteal phase of approximately 14 days and a short follicular phase of approximately 3–4 days. The demise of the corpus luteum at the end of the luteal phase leads to a decrease in progesterone, which otherwise exerts potent negative-feedback at the central level (18). This allows oestradiol levels to rise and trigger the increase in the frequency of gonadotrophin-releasing hormone (GnRH) pulses, which leads to a parallel increase in the frequency of LH pulses and, ultimately, to the LH surge followed by ovulation some 20 h later (19–22).
There is clear evidence that the difference between the breeding and nonbreeding seasons primarily lies within the GnRH pulse generator because the frequency of GnRH pulses (and consequently LH) is drastically reduced during the anoestrus season compared to the breeding season (23–26). Methodical observation of what occurs during the last oestrus of the breeding season shows that the demise of the corpus luteum is not followed by a surge of GnRH (27). It therefore appears that the inability to mount a GnRH/LH surge is what prevents the sequence from happening again and is characteristic of the anoestrus state. The converse also holds true as the frequency of GnRH pulses increases again during the transition from anoestrus to the breeding season (28).
Classical work using ovariectomised (OVX) ewes with an implant releasing constant levels of oestradiol at mid-luteal phase levels revealed that LH levels were then constantly high during the breeding season (LH > 5–10 ng/ml) and remained at basal levels (LH < 1 ng/ml) during the anoestrus season. This result led to the hypothesis that seasonal reproduction is the result of a marked seasonal shift in the hypothalamic responsiveness of oestradiol, such that it exerts negative-feedback during anoestrus and positive-feedback during the breeding season (23,29,30). This idea that there might be seasonal changes in the effect of sex steroids was initially proposed by Pelletier and Ortavant (31,32) in Ile-de-France rams. These seasonal shifts in reproduction are indeed steroid-dependent because LH is clamped to high levels in OVX ewes, albeit there may also be a minor steroid-independent modulation of the GnRH pulse frequency (24,26,33); but see also (30,34). Progesterone does not appear to play a role in these seasonal transitions and is indeed predictably undetectable during anoestrus (23). It is noteworthy that alternations between reproductive activity and anoestrus occur despite approximately constant levels of oestradiol and over relatively short time-spans. Therefore, the seasonal shift in the hypothalamic responsiveness of oestradiol might be envisioned as the consequence of the interaction between oestradiol, mostly fulfilling a permissive role, with another (endocrine or else) signal whose production varies according to the photoperiod to actively promote switches in reproduction.
The ovine seasonal breeding cycle: fueled by long days
A widely embraced model for seasonal reproduction in sheep and other long-lived mammals is that PP does not generate rhythms but rather acts, through melatonin, as the main synchroniser of an endogenous timing mechanism called the circannual clock. Circannual rhythms have indeed been described not only in mammals, but also in birds and insects and are assumed to be at the basis of several crucial behavioural and physiological phenomenons such as migration, hibernation or moult (35–39). Because of this circannual clock, persistent rhythmicity in reproduction with a period of approximately 1 year (usually approximately 10 months) should be revealed if sheep are maintained under constant conditions. Because it is PP through the action of melatonin that acts as the main synchroniser of circannual clocks, circannual rhythmicity should be evident when melatonin or photoperiod is kept constant. Below, the outcomes of experiments carried out under several constant conditions are reviewed.
Suppression of melatonin
When they are pinealectomised (PX), Suffolk ewes show great variability in individual LH patterns with little evidence for persistent yearly rhythmicity. Some animals do show constant high levels of LH, whereas most exhibit erratic transitions between states of elevated and low LH, the durations of which fluctuate hugely (9,40–43). PX ewes of another breed (Saxon Merino × Border Leicester) also do not exhibit any rhythms in LH/follicle-stimulating hormone (FSH) (44). It is, however, important to stress that significant seasonal variations in all endocrine parameters tested, albeit of lesser amplitude and less precise timing, are conserved if PX (or SCGX; ablation of the superior cervical ganglions that innervate the pineal) animals are maintained outdoors along with non-operated animals (e.g. Soay rams) (45). This illustrates the fact that not only PP, but also other environmental, nutritional, social and olfactory cues play a role in the maintenance of seasonality (4546–48). By contrast, annual rhythms in temperature exert at best a weak influence on seasonal rhythmicity (13,49,50). Nevertheless, when SCGX rams were housed indoors for 5 years, with constant diet and photoperiodic cycles consisting of alternations between 4 months of LD and SD, ‘the persistence of long-term reproductive cycles was less evident (than outdoors) and cycles were poorly defined and irregular’ (45). Another means to clamp melatonin to basal levels is to maintain animals under constant light, a condition under which the suprachiasmatic nuclei suppress melatonin production by the pineal gland. When ewes (Finn × Dorset/Rambouillet breed) are maintained under constant light for a year, PRL levels remain at intermediate levels and ewes appear to end up in a state of prolonged oestrus (51).
Constant release of melatonin
Constant-release melatonin implants in Soay rams blunt PRL and FSH rhythms that were previously sustained by timed alternations between LD (PP 16 : 8, i.e. 16 h of light: 8 h of darkness) and SD (PP 8 : 16), which led to the conclusion that ‘melatonin blocks the effect of LD and in the long term renders rams unresponsive to changes between LD and SD’ (Fig. 2) (52). Melatonin implants also blunt photoperiodic variations in mean LH/FSH concentrations, and, to a lesser extent PRL, in ewes (44). Therefore, either the absence or constant presence of melatonin, leads to at best erratic (i.e. of no obvious consistent period) alternations between breeding and anoestrus.
Constant equatorial photoperiods and shorter photoperiods
Ewes of the Southdown or Merino breeds kept in a constant equatorial PP 12.5 11.5 show very little evidence for seasonal rhythmicity in reproduction, as assessed by oestrus behaviour (49,53). This result has been replicated in further studies using progesterone and PRL as endocrine markers (54). Similarly, Romney rams maintained in PP 12 : 12 do not exhibit rhythmicity in PRL (55). The same result was obtained in Suffolk ewes kept under a constant PP 12 : 12 for 3 years, also using PRL as a marker for rhythmicity (56). Indeed, sheep and goat breeds that inhabit or are transferred to equatorial regions often show irregular oestrus cycles throughout the year or persistent anoestrus (7, 57,58).
Suffolk ewes maintained under a constant winter solstice PP (winter solstice-hold protocol; WSH, which corresponds to PP 10 : 14 at the latitude of Ann Arbor) never resume sexual activity the following year (Fig. 2) (59). Using the same breed (Suffolk) and the same protocol (WSH; PP 10 : 14 at the latitude of Urbana, IL, USA), Lubbers and Jackson (34) confirmed this result because ewes spent over 300 days in anoestrus (as assessed by LH levels). Galway ewes submitted to the same protocol (WSH; PP 8.5 : 15.5 at the latitude of Dublin, Ireland) showed variability as only three of five ewes initiated a breeding season (60); estimated using progesterone and LH levels. In a further study by the same team in the same breed and with the same protocol, 19 of 20 ewes remained in anoestrus the following year (61). Similarly, Soay rams maintained under PP 8 : 16 do not exhibit rhythmicity in either PRL or testosterone (62). Howles et al. (63) confirmed this because they found that ‘there was little evidence for rhythmicity in PRL’ in rams maintained under PP 8 : 16. Langford et al. (64) also made similar observations: ‘Rams exposed to a continuous PP 8 : 16 exhibited no cycles of pituitary-testis activity, individually or as a group’. Finally, Vidal et al. (65) measured LH levels in Black-Belly ewes kept for 3 years under a constant PP 8 : 16 and noted that ‘the existence of an endogenous rhythm is not straightforward’.
The only discordant study is Karsch et al. (66) where Suffolk ewes kept under constant PP 8 : 16 showed perfectly rhythmic LH and PRL profiles for 5 years. In another study carried out later by the same group under the same PP 8 : 16, rhythmicity was much less obvious (67). This result is at odds with data from the WSH protocol mentioned previously and not only obtained in the same breed by the same team (59), but also by others (34). In the study by Karsch et al (66), there were indeed only minute differences in duration, phase or amplitude between endocrine profiles of ewes kept outdoors and under PP 8 : 16, which suggests that rhythmicity may perhaps have been entrained by some unidentified social or olfactory cue (see above).
Constant long photoperiod
Soay rams maintained under PP 16 : 8 show robust variations in testosterone and PRL (62). Rams studied by Howles et al. (63) also showed clear PRL rhythmicity under PP 16 : 8. Suffolk ewes kept in PP 16 : 8 also demonstrated rhythmicity in LH levels (68), as did Romney rams kept in PP 14.5 : 9.5 (55). Both European Mouflon and Spanish Manchega ewes exhibited seasonal variations of progesterone and PRL under constant PP 16 : 8 (69). Similarly, Finn and Galway ewes submitted to a summer solstice-hold (SSH; PP 16 : 8 at the latitude of Dublin) also showed a complete breeding season (i.e. animals underwent reactivation of the gonadotrophic axis, resumed breeding and eventually entered in anoestrus) (60). The exact same result using the same protocol was also reported for Suffolk ewes by Ebling and Foster (70). Note that animals can undergo one complete cycle under constant SD, or constant melatonin release, provided they have previously been exposed to LD for a sufficient duration (52,62,71) (Fig. 2).
Consequence: the critical photoperiod in sheep
Persistent circannual rhythmicity in sheep is not observed in animals with no melatonin, which implies that melatonin is mandatory. However, because circannual rythmicity is also absent in animals with constant-release melatonin implants, it is equally obvious that melatonin should not be present continuously, which supports the notion that there might be a critical duration for melatonin to be effective. Data also provide little support for persistent circannual rhythmicity under PP equal to, or shorter than, 12–12.5 h. It is only when animals are maintained under constant LD (> 12.5 h), when the duration of melatonin secretion remains short, that circannual rhythmicity is sustained. LD might therefore be defined as the ‘driving force’ of the ovine seasonal reproductive cycle. Note that the major role for LD is obvious not only during adulthood, but also for the initiation of puberty in lambs and, indeed, Foster (72,73) wrote that ‘LD set events in motion for normal puberty many weeks later, which can occur in the absence of additional PP information about season’.
The minimal length of light exposure that allows the animal to interpret the PP as being long is called the critical PP. As discussed above, this PP is most likely > 12.5 h. In line with this, Black-Belly ewes maintained under simulated tropical lighting conditions, with only modest variations from PP 11 : 13 to PP 13 : 11 are clearly rhythmic (74). Data reported by Robinson and Karsch (75) concur and indicate that the critical PP is close to, but shorter than, 13 h, in the Suffolk ewe too. Incidentally, when transferred at 19°N (where natural PP variations are larger, from 11.5 : 12.5 to 14 : 10) Suffolk ewes show marked seasonality in reproduction (58). These data indicate that the value for the critical PP probably lies between 12.5 and 13 h. Below, the mechanisms by which LD entrain seasonal rhythmicity are evaluated.
How long days shape the seasonal reproductive cycle
The effects of long days are very well characterised and have been used to develop breeding strategies which are routinely used in sheep and goats to assist farmers and yield out-of-season births (7,76–79). From the literature, three different effects of LD can be disclosed: (i) LD are mandatory to prime the sequence of reactivation and anoestrus that occurs in the following year, (ii) the amount of LD sets the length of the anoestrus period; and (iii) LD exerts a tonic suppressive effect upon the gonadotrophic axis at the end of winter.
LD priming of the next season reproductive sequence
Evidence that LD is the main driving force of seasonal rhythms not only stems from constant PP data, but also from the results of the solstice-hold protocols (Fig. 3). From the summer solstice, animals are submitted to either a constant PP of the solstice duration (SSH) or a natural simulate whereby PP is decreased stepwise to mimic the natural PP, or kept in natural conditions. Strikingly, the timing of onset of the breeding season is then identical between all three groups (early September; 80). Furthermore, when submitted to the mirror protocol (WSH), all ewes also enter in anoestrus during the next months (8). Therefore, it can be concluded that neither decreasing day length after the summer solstice, nor increasing day length after the winter solstice play any role, and this part of the cycle is generated endogenously. Indeed, animals under the SSH can undergo successive sexual reactivation and entry into anoestrus (60,70). The experiments in Soay rams with melatonin implants exemplify this point: when implants are removed, a normal sequence of reactivation/anoestrus only occurs in those rams which experience LD, not SD (52).
Perhaps the best evidence for a major role of LD has been gained in PX animals given melatonin infusions of various durations or mimicking different parts of the natural photoperiodic cycle. Seasonal rhythmicity can be reinstated in PX animals by giving them specifically LD patterns of melatonin (short duration) (81). When ewes are PX during spring or summer, the anoestrus duration is shortened (see below) and the next breeding season does not happen or is greatly perturbed, whereas ewes PX during autumn are similar to control animals and ewes PX during winter end their breeding season normally but fail to reinitiate a breeding season the next year (40,82). Furthermore, daily injection of a long day pattern of melatonin for 70 days (i.e. 70 LD) per year is sufficient to yield synchronised seasonal rhythms in LH (41). The same conclusions arise when only 35 LD are given (42). Definitive evidence for the role of LD was provided in the study by Woodfill et al. (82) where PX animals were given timed infusion patterns of melatonin corresponding to different patterns occurring during the year. A spring melatonin pattern (21 March to 21 June, from a 13 : 11 h photocycle to a 16 : 8 h photocycle) or a summer pattern (21 June to 20 September, from 16 : 8 h photocycle to a 13 : 11 h photocycle) yielded exactly the same entrainment, and the breeding season occurred a month after the infusion was ended. This appears to imply it is the absolute number of days interpreted as LD (identical between both groups as a result of the symmetrical variation pattern) that matters rather than the direction of change. The autumn pattern (20 September to 21 December; from a 13 : 11 h photocycle to a 11 : 13 h photocycle), which includes a smaller number of LD (> 12.5 h of light) also is efficient, albeit synchronisation is not as good. The winter pattern is ineffective. Similar results were reported by Barrell et al. (43).
The amount of LD sets the length of the anoestrus
Interestingly, in a study by Wayne et al. (81), when the LD melatonin infusion starts at the end of the breeding season and is maintained, LH increase occurs after approximately 210 days (as long as the length of anoestrus under natural conditions), whereas, if infusion is stopped after 65 days, then the LH rise occurs earlier, after approximately 145 days. This result was confirmed by Karsch et al. (68); if animals are given 60 days of LD at the end of the breeding season, then the anoestrus lasts approximately 110 days from the start of the melatonin LD treatment. Exposing Suffolk ewes to 7 or 15 days of LD around the winter solstice appears insufficient to yield a breeding season thereafter, whereas some ewes receiving a 30-day block of LD do show short periods of rises in progesterone levels approximately 60–90 days later (83). PX animals which are synchronised by a short 35 LD block of melatonin exhibit rising levels of LH after approximately 95 days (42). Furthermore, when given the 3-month spring or summer melatonin patterns, these PX animals have rising levels of LH within approximately 120 days from the start of the experiment, whereas the rise occurs after only approximately 60 days with the autumn melatonin pattern (82). Ewes receiving daily melatonin supplementation from mid-late June to yield a SD pattern, which diminishes the number of effective LD, have their breeding season advanced by approximately 20–40 days (84,85). When a similar treatment is given earlier, from mid-March, the breeding season is advanced by nearly 4 months (86). These data concur with earlier findings reported by Ducker et al. (87,88) in Clun Forest ewes showing that premature exposure to SD significantly advances the breeding season.
LD exert a tonic suppressive effect which can curtail the breeding season
Exposure of ewes to a short LD message (7–15 days) at the mid-end breeding season is sufficient to precipitate the fall of LH levels and accelerates the entry in anoestrus (9,61,68,83,89,90). When given during the anoestrus season, the LD message invariably advances or delays the next breeding season according to its duration and timing (59,61). Finally, PX animals readily entrain to a melatonin LD signal regardless of whether it coincides with a period of high or low LH, which indicates that LD entrainment is largely independent of the reproductive state of the animal (42,82). The ability of LD to synchronise the PRL and moult rhythms, whatever the initial phase of the seasonal rhythm is, has also been demonstrated in Soay rams (91). However, it has been noted by Sweeney et al. (61) that ‘the reproductive neuroendocrine axis in Galway ewes is insensitive to a LD signal initiated after the autumn equinox and terminated before the winter solstice’. This specific time-window, early in the breeding season, therefore corresponds to a phase of refractoriness to LD.
To summarise, these data show that a minimal duration of exposure to LD (approximately 35 days) is mandatory for the breeding season to happen, that there is a positive correlation between the length of the LD treatment and the length of the anoestrus and that LD can actively suppress breeding. Together, these effects might explain why the period of the circannual rhythm is of only approximately 10 months (39). The simplest model to accommodate these findings is one in which LD actively maintains the animal in a sexually inhibited state during spring/summer (for a maximal of approximately 210 days, after which the animal becomes refractory to LD and resumes breeding), whereas LD are also mandatory for the initiation of a series of events that ultimately culminate in refractoriness and the occurrence of the breeding season. This necessarily implies that LD trigger some melatonin-dependent mechanism that ultimately impinges on GnRH neurones.
No obvious role for short days
As seen above, a minimum duration of exposure to LD is required to yield a breeding season. However, the duration of the breeding season does not appear to be tightly correlated with the duration of LP exposure past this minimum threshold requirement. Indeed, the breeding season is systematically shortened under various indoors conditions as it lasts only approximately 3 months versus approximately 5.5 months under natural conditions (41,59,60,68,81,92). This suggests that SD might somehow be required for the maintenance of the breeding activity. Furthermore, procedures that maintain or even extend the breeding season would bear significant economic value and have therefore been investigated thoroughly.
Unfortunately, attempts to extend SD by various means (e.g. WSH, the use of even shorter PP after the winter solstice) to yield a breeding season of maximal length, are either ineffective or have only marginal effects (59,71,93,94) (Figs 2 and 3). In line with this, PX ewes receiving first a melatonin LD signal of 70 days and then either 70 or 100 days of a melatonin SD message do not exhibit any difference in the timing or duration of their breeding season compared to animals given only the 70 days LD signal (41). Taken together, these data suggest that SD are not necessary to time seasonal rhythmicity, which is also consistent with data of the SSH. Consequently, although the sheep is a ‘short-day breeder’, the proper execution of its seasonal programme depends on LD and not SD. The exact reasons why these procedures do not allow a breeding season of normal duration to be expressed remain unclear.
Thyroid hormones as endocrine mediators of long days
The implication of thyroid hormones in the seasonal programme of reproduction has been established in birds over 70 years ago (95–97). In a series of studies, groups led by Brian Follett and Fred Karsch demonstrated that thyroxine (T4) also plays a crucial role in the sheep and further characterised the temporal and spatial requirements for T4 action (Fig. 4).
When Welsh Mountain ewes are thyroidectomised (THX) at the end of July (anoestrus), sexual activity resumes as expected in autumn (October/November) but the transition to anoestrus, which normally occurs in February, never happens and these ewes remain sexually active (98). Melatonin profile still faithfully tracks PP, which shows that THX does not impair the hormonal photoperiodic read-out (98). Moenter et al. (99) made similar observations: when Suffolk ewes are THX in July, sexual activity resumes at a normal time (mid-September) but animals fail to enter in anoestrus in February, as judged from circulating LH levels. Melatonin profiles and PRL output are not affected by THX and exhibit normal seasonal variations. These results have been confirmed in further studies (melatonin: 100; PRL: 101). When THX was performed in October (101), the PRL rhythm was normal during the following year, which indicates that PRL rhythmicity is not mediated through a T4-dependent mechanism. The above-mentioned role for T4 is not specific to the ewe because similar observations have been made in Soay rams (102,103): THX blocks the decrease in LH/FSH, the involution of the gonads and the drop in testosterone, which all normally occur during the period of sexual rest. Again, PRL rhythms are unaffected by THX.
T4 supplementation in THX ewes reinstates the normal transition to anoestrus (104). The seasonal rhythm in the levels of circulating T4, with higher values during the breeding season (approximately 50 ng/ml from November until May) than during anoestrus (approximately 30–40 ng/ml), appears meaningless for the seasonal reproductive rhythm (105,106). Thyroid hormones therefore appear to play a permissive role (106). In THX animals, the mechanisms of feedback of oestradiol and progesterone are preserved (104). Similarly, the amplitude and frequency of GnRH and LH pulses, the number and overall morphology of GnRH neurones are all similar between THX and control animals (107). The lack of anoestrus in THX ewes appears to be the direct consequence of a permanent activation of the GnRH pulse generator, typical of the breeding season (107).
When and for how long are thyroid hormones required? Thrun et al. (108–110) demonstrated that T4 is only required for a short period of time, approximately 90 days, starting at the end of December that is during the latest portion of the breeding season. Results of a later study, also in Suffolk ewes, showed that T4 is indeed efficient up until June/July that is late during the anoestrus season (101). T4 might therefore exert a potent and active inhibitory action during most of spring and summer that prevents the animal from resuming sexual activity. Such a conclusion is also supported by data obtained in Soay rams showing an almost immediate increase in FSH output and regrowth of the testes when THX in March, at a time when the gonads are completely regressed (102).
Does T4 act in the brain or in the periphery? Viguiéet al. (111) showed that a small dose of T4 injected i.c.v. (approximately 2 ng/ml; similar to the physiological concentration in the CSF) was sufficient to reinstate transition to anoestrus, whereas a similar dose administered peripherally was without effect. It was therefore concluded that the site of action for T4 is central. In an attempt to find this site, T4 implants were placed in various locations within the hypothalamus (112). Only implants in either the ventromedial preoptic area (POA) or the mediobasal hypothalamus (MBH) were efficient. At this stage, both the exact site(s) of action of triiodothyronine (T3), converted from T4 directly within the hypothalamus, and the potential target genes of the liganded thyroid hormone receptors remain to be defined. Although data implicating the dopaminergic A15 nucleus in the physiological transition to anoestrus make it an appealing candidate (113–116), the A15 nucleus is not immediately adjacent to either the ventromedial POA or the MBH (117).
Collectively, these studies demonstrated that T4 plays a permissive role within the POA and/or MBH during a period of time, which extends from the end of the breeding season to late in the anoestrus season, to allow ewes to terminate their breeding season and remain in anoestrus. The striking parallel between the time-windows of action for long days and T4 were suggestive of a model in which long days (through melatonin) trigger a T4 increase within the MBH, which in turn triggers the entry and maintenance in anoestrus. Below, data pertaining to the role of the MBH in the seasonal control of reproduction are reviewed.
Sites of action of melatonin for the seasonal response
Melatonin implants placed in the MBH of animals kept under LD lead to neuroendocrine and reproductive changes typical of those expressed under SD, with an increase in LH/FSH levels and a concomitant decrease in PRL levels (Fig. 5). This pattern occurs in both ewes and rams, albeit with different kinetics. Because these implants release large amounts of melatonin (approximately 5 μg/day) (118,119), which is a highly lipophilic molecule that easily diffuses throughout tissues, it was concluded that ‘melatonin acts within or close to the MBH to mediate the effects of photoperiod’ (118–120). Indeed, the use of radioactive 125I-Mel and 3H-Mel implants in the MBH indicates that melatonin diffuses within a radius of at least 2–6 mm from the implant (118–121).
The pars tuberalis (PT), which is located immediately adjacent to the MBH and expresses the highest levels of melatonin receptors across all mammals, therefore appeared as another plausible target for melatonin coming from these implants (Fig. 5). Binding studies with 125iodo-melatonin (122–125), in situ hybridisation (126) and quantitative reverse transcriptase-polymerase chain reaction (127) all concur with a density of receptors approximately 30- to 100-fold times higher in the PT compared to the MBH. This difference does not imply that the PT would be more physiologically relevant than the MBH, especially because the affinity of the melatonin receptors is quite high (low pM range). More interestingly, melatonin receptors are not found within the MBH of all mammals but, instead, are present in other hypothalamic regions (128–130) or even absent, as in the highly photoperiodic ferret, which led to the conclusion that ‘… melatonin acts through the hypothalamus, without necessarily binding within it’ (131). These data indicate that the role of the MBH as a melatonin target tissue is not phylogenetically conserved.
Using their hypothalamic-pituitary disconnected (HPD) ram model, Lincoln and Clarke provided definitive evidence that melatonin acts directly through the PT (independently of the hypothalamus) to regulate PRL seasonal secretion (Fig. 5) (132,133). HPD animals are hypogonadic, display constant basal level of gonadotrophins and their gonadotrophic axis is completely impervious to PP changes or subcutaneous melatonin implants, although pituitary gonadotrophs are functional as demonstrated by faithful responses to timed-release GnRH pulses (134,135). These characteristics are a result of the fact that the HPD operation permanently destroys the arcuate nucleus and median eminence (hence GnRH projections) and, consequently, all communication between the hypothalamus and the hypophysis (39,136), whereas the PT is preserved (137). This has important implications for seasonal reproduction because the arcuate nucleus is the site where progesterone and oestradiol exert negative-feedback to govern the photoperiodic response of the gonadotrophic axis and the oestrous cycle (138–143). Furthermore, lesions of the arcuate nucleus lead to the same phenotype as the HPD surgery: photoperiodic control of PRL is barely affected, whereas the response of the gonadotrophic axis is lost (136). Therefore, independently of it being a direct target for melatonin or not, the MBH is involved in the transduction of the photoperiodic information towards the gonadotrophic axis.
Melatonin implants aimed at the MBH systematically lead to simultaneous effects on both PRL and gonadotrophins. These effects exhibit a high degree of negative correlation such that implants eliciting the largest increase of LH/FSH secretion are also those providing the largest suppressive effect on PRL levels (118–120,144). When melatonin implants are aimed at the PT (implants placed at the lower point of the infundibular recess in the pituitary stalk) instead of the MBH, the same negative correlation prevails: implants yielding the largest increase of LH/FSH secretion also have the strongest suppressive effect towards PRL (145). When melatonin implants are placed in apposition with the pituitary stalk (trans-nasal surgical approach), close to the PT, they are without effect on either PRL or LH/FSH, whereas s.c. implants or implants in the third ventricle lead to the expected concomitant increase in LH and reduction in PRL secretion (146). Overall, the interpretation of the melatonin implant data is ambiguous because, if melatonin released in the MBH leaks to the PT, the converse might equally be true.
These data are not easily reconciled with the ‘dual sites hypothesis’, which posits that melatonin acts in the PT to control PRL and in the MBH to control the gonadotrophic axis (116,134,147) because a strong prediction from this model (i.e. an asymmetrical response of both endocrine responses according the position of the implant) is not verified. From a functional perspective, having a single site of action for melatonin would arguably be the best way to achieve coordination of the PRL and FSH/LH responses. The only possible site would then be the PT because experiments with the HPD ram unambiguously demonstrated that melatonin acts within the PT to control the seasonality of PRL (132). Although in no case comprising decisive arguments, there are anatomical and functional evidence for a particular microvasculature of the MBH, which would favour the unidirectional flow of molecules from the MBH to the PT (148,149), whereas tanycytes lining the third ventricle also provide a privileged route from the cerebrospinal fluid to the median eminence/PT (150–152).
To conclude, the early HPD experiments made it clear that melatonin mediates seasonal swings in PRL through the PT, whereas experiments using melatonin implants could not define beyond reasonable doubt that the seasonal control over gonadotrophins depends upon melatonin acting within the MBH and/or within the PT (153). The mechanism of the seasonal control of PRL by the PT is independent from T3, as demonstrated by the persistent PRL rhythm in THX animals. This rhythm might therefore be dissociated from the gonadotrophic output. Indeed, the PRL response to both LD and SD is much faster than that of the gonadotroph axis because it occurs within days and is complete after a few weeks (132,154). It has been suggested that the PP impinges on the PT to direct the production of one or several endocrine compounds (collectively termed tuberalins) that are released in the portal system of the anterior pituitary and act directly within the pars distalis to govern seasonal rhythmicity in PRL. Neurokinin A appears to be a strong contender for tuberalin (155). Further information regarding the seasonal control of PRL by the PT is provided in recent reviews (156,157).
There is now strong evidence that the PT indeed acts upon the gonadotrophic axis. By contrast with the anterograde role of tuberalins in the control of PRL, this occurs in a retrograde manner through a relay within the MBH. Below, the findings are described that shed new light on how PT cells decode a LD signal of melatonin and translate it into an endocrine output regulating thyroid hormone metabolism within the MBH.
Molecular decoding of the long day melatonin message within the PT
Our previous work in sheep (158) along with pioneer work in the Japanese quail (159,160) identified a paracrine mechanism that integrates the crucial roles of long days and the action of T3 in the MBH, through the action of melatonin within the PT. Briefly, a short duration melatonin signal (typical of LD) increases thyroid-stimulating hormone (TSH) output by the PT, which then acts retrogradely within the hypothalamus through cognate receptors (TSH-R) expressed by tanycytes lining the third ventricle. This leads to increased cAMP production and transcriptional activation of Dio2 that encodes for an enzyme (type-2 deiodinase; DIO2) which catalyses the deiodination of T4 into T3 and therefore activates thyroid hormone signalling. Interestingly, at least in the quail, simultaneous transcriptional repression is also exerted on the Dio3 (type-3 deiodinase) gene. This coordinated action leads to a tight photoperiodic control of local hypothalamic T3 levels because DIO3 plays a role opposite to that of DIO2 and converts T4 to rT3 and T3 to T2. This dual mechanism explains how T3 is made available in the MBH during the LD of spring and summer. Although there might be species differences in the relative importance of the photoperiodic changes in DIO2 and DIO3 (161–164), the net effect remains the same: T3 signalling is increased upon LD exposure.
The action of melatonin is probably direct because thyrotrophs of the PT bear the MT1 melatonin receptor, at least in rat (165) and European hamster (166). In the ewe, this temporally-gated increase in T3 triggers the transition to anoestrus and the subsequent maintenance in this physiological state until refractoriness to LD develops during autumn. This Mel/TSH/Dio2-Dio3 mechanism has also been uncovered in other species, including the mouse (167), golden hamster (168), European hamster (169) and rat (170). Crucially, the PT behaves as a true ‘TSH-generator’ under LD, independently of the classical negative-feedback exerted by T3 and TRH, because their cognate receptors are not expressed in the PT, at least in sheep (171).
The PT, similar to most tissues, harbours a circadian clock. This one, however, has the unique property of being entrained by melatonin through the induction of the clock gene Cryptochrome1 (Cry1), as demonstrated in sheep (154,172) and rodents (173–175). Therefore, melatonin achieves a daily resetting of the phase of the cellular PT clocks, and several models have been proposed over the years to explain how this clock might be used to measure the duration of the melatonin signal (156,176,177). The mechanism of Cry1 induction in the PT remains incompletely understood, although it might involve one or several transcription factors related to early growth response protein 1 (178). Apart from TSH, it was shown that the transcriptional co-activator Eyes absent homologue 3 (EYA3) is also acutely induced by LD in the PT (155,160). This led us to test the possible implication of EYA3 in the transactivation of the thyrotrophin beta subunit (Tshβ) gene.
This work led to the identification of a molecular mechanism responsible for the LD increase in TSH production within the ovine PT (Fig. 6) (179). Similar data were obtained in the mouse (180). Transactivation at the Tshβ promoter depends on the binding of a transcription factor, Thyrotrophic embyonic factor (TEF), to its specific response element (D-Box) and subsequent recruitment of the co-activators Sine oculis homeobox homolog 1 (SIX1) and EYA3. The system is reliant on: (i) the melatonin synchronisation of the phase of individual cellular circadian oscillators within the PT through the acute induction of Cry1; (ii) the subsequent phase-locking of the expression of Tef and Eya3, which are direct outputs of the circadian clock; and (iii) the inhibitory effect of melatonin on Eya3 expression. Because of these characteristics, the induction of Eya3 invariably occurs approximately 12 h after Cry1 induction, and therefore falls either when melatonin is present (e.g. PP 8 : 16) or absent (e.g. PP 16 : 8), which corresponds to an external coincidence model for photoperiodic time measurement (181–183). This explains why Eya3 expression is acutely induced in LD and how TSH production is consequently increased.
It is remarkable that the gap between Cry1 (melatonin onset) and Eya3 (melatonin offset) peaks be of approximately 12 h, so close to the estimated value for the critical PP. Because of its design principle, this time-measurement mechanism is expected to yield a high TSH output under constant LD conditions (when Eya3 expression can rise) and a small output under constant SP (Eya3 peak blunted by melatonin). Furthermore, because melatonin induces the expression of Cry1 and because negative-feedback by Cry1 is a critical feature of the circadian clock (184–186), PX animals or animals with a constant-release melatonin implant would be unable to exhibit a daily peak of Cry1 and the time-measurement mechanism would be compromised. Although data are required to support these predictions, they are fully compatible with observations made in vivo (i.e. circannual rhythmicity retained under constant LD only).
The seasonal reproductive cycles of the golden hamster and sheep: opposite yet similar
Aside from the sheep, the golden hamster (Mesocricetus auratus) is undoubtedly the species that has been used the most extensively to study the mechanisms of seasonal breeding. As a result of its very short gestational period of 16 days, the reproductive season of the golden hamster occurs during spring/summer, which is why it is categorised as an LD breeder. Although this might appear a radical departure from the sheep, the two models are similar in many respects. Under natural conditions, the breeding season of the hamster starts around mid-March and ends in September, which means that physiological transitions, albeit opposite to those in sheep, happen at the same time of year (187). Here again, melatonin plays the central role because PX animals cannot read photoperiod and are permanently sexually active (188). Interestingly, animals with a constant-release melatonin implant, which blocks the PP response, are also indefinitely sexually active (189).
By the end of the winter, hamsters do show a recrudescence of the gonads that is usually described as ‘spontaneous’ because neither continuous exposure to SD, nor melatonin injections are able to block it (188,190,191). This parallels the results of the SSH and WSH experiments in sheep, in which the termination of the breeding season in spring is generated endogenously. Similar to the curtailing of the breeding season upon LD exposure in sheep, premature gonadal recrudescence at the end of the winter can be induced either by PX (192) or by exposing hamsters with nonfunctional gonads to LD (190,193,194). This effect of LD increases with the duration of SD exposure: LD are ineffective in animals that have just finished gonadal involution and become increasingly efficient after approximately 10 weeks of SD (191). This period would fall around the winter solstice, a time during which LD are also without effect in the sheep.
If sheep are maintained under SD after the winter solstice, they fail to reinitiate breeding during the next fall and remain sexually inactive. What occurs in hamsters is a mirror image because hamsters kept under constant SD remain indefinitely sexually active. Exposure to LD is required in sheep to yield a breeding season and hamsters too need LD exposure to become sensitive to SD again. Here again, melatonin is required for LD to exert its effect (193,195). Much like in the sheep, a minimal duration of exposure to LD (approximately 11 weeks) is required for the hamster to interpret the SD message in autumn (188,191,194,196). The only notable difference between sheep and hamsters appears to be the inability of hamsters to become refractory to LD because hamsters maintained in LD remain sexually active (188), whereas sheep resume their breeding season spontaneously in autumn as demonstrated by the SSH.
It has been hypothesised that sensitivity of target tissues to melatonin is lost upon prolonged exposure to SD and that LD exposure is necessary to resensitise these tissues to the hormone (195). However, the pineal produces melatonin also under LD and there is evidence that the target tissues are indeed quite able to decode the melatonin message (193,195). More importantly, the difference between a constantly active gonadotroph axis and a permanently inactive one is only a matter of minutes, as demonstrated by experiments aimed at the identification of the critical PP, which is of 12.5 h (197–199). This appears to hardly be compatible with the desensitisation hypothesis.
The sequence of events triggered by LD, which culminates in a local increase of T3, has been identified in hamsters as well (168,169). In the adult Siberian hamster (Phodopus sungorus), which displays responses to LD very similar to those described above for the golden hamster (200–203), intrahypothalamic implants of crystalline T3 or daily injections mimic LD because they block the onset of the response to SD (204,205) and also prematurely reactivate the reproductive axis of animals under SD (205,206). Initial mechanisms of the photoperiodic read-out appear very similar in SD and LD breeders and what is responsible for the difference must therefore lie downstream. Indeed, clear similarities exist between seasonal reproductive cycles of phylogenetically distant vertebrates such of birds and mammals, including the role of thyroid hormones, and these have been highlighted in reviews by Nicholls et al. (95,98).
Conclusions and future directions
Although the breeding season of the sheep occurs during the SD of autumn and winter, the seasonal reproduction of this species appears to be governed by the exposure to the LD of spring and summer. LD not only actively suppress reproductive activity from early spring to autumn, but also play an active role that ultimately promotes the occurrence of the next breeding season. As shown by the outcome of experiments with THX sheep, these effects of LD appear to result from the local increase of T3 levels within the MBH. Present evidence favours a model in which melatonin acts through the PT of the anterior pituitary to control the endocrine transitions triggered by varying photoperiods. A molecular mechanism which ties duration of the melatonin signal to the local output of T3, through TSH and DIO2 activation, has been uncovered in the PT and might explain how LD impact physiology. Because of its design, this timing device can precisely adapt its output according to the melatonin duration and could allow the animal to draw a line between SD and LD (i.e. the critical photoperiod). This system does not require any shift in the hypothalamic ability of oestradiol to inhibit or activate the GnRH pulse generator and oestradiol would act as a permanent permissive signal whose action is modified during the anoestrus as a result of T3. Furthermore, despite their opposite breeding seasons, sheep and golden hamster both appear to rely on LD to time their reproductive cycle, which suggests that underlying mechanisms may be fundamentally similar. The current evidence obtained in several species indeed indicates that the PT time-measurement mechanism and role of TSH might be phylogenetically conserved.
There is undeniably strong evidence for an endogenous component in the seasonal programme of the sheep because it ultimately becomes refractory to LD, spontaneously resumes breeding and subsequently becomes sexually quiescent. Because circannual rhythms are clearly expressed only under constant LD conditions, the data might perhaps be more consistent with a complex hourglass mechanism (corresponding to the breeding season) driven by exposure to LD than with a circannual clock. Because LD trigger the sequence of events through T3 production (and other unidentified mechanisms; c.f. PRL rhythmicity), we might advance our understanding of the nature of this endogenous process if we determine the role of T3. Below, several potential research avenues are described.
With the identification of a chain of events and major molecular players (e.g. CRY1, EYA3, TEF, TSH, Dio2/Dio3), it appears legitimate to investigate whether LD-refractoriness might arise at some point in this cascade. We already know that the melatonin pattern always faithfully tracks day length and that the expression of clock genes is therefore unchanged in the PT of refractory animals (207). However, there is evidence from the HPD ram that the mechanism of refractoriness develops within the PT (134). Indeed, putative molecular correlates to LD refractoriness of the reproductive axis are found in the PT: the expression of the α-subunit common to glycoproteins such as TSH returns to LD levels after prolonged exposure to SD, both in sheep (207) and the golden hamster (208). Spontaneous changes in the expression of Dio2 have also been observed in the hamster (163).
We know that thyroid hormones do not passively diffuse through plasma membranes but require active transport (209). In the brain of the sheep, the expression of the most relevant transporters (such MCT8) and of Dio2 is not limited to the tanycytes and also occurs within the median eminence and MBH. These transporters constitute another potential way to regulate T4 access and their distribution implies that LD-induced T3 could play a role in various structures. The data reported by Viguiéet al. (111) indicate that T4 in the third ventricle is required but it may not necessarily act within the MBH because tanycytes provide a route to the median eminence where Dio2 is also expressed (150,151). Because T3 plays major roles in plasticity and neurogenesis, and because the ependymal cells of the basolateral third ventricle comprise a population of stem cells (210,211), the possibility that LD refractoriness and the occurrence of the breeding season arise as a consequence of cell renewal and apoptosis has some strong appeal (39,212,213). A role of neuronal plasticity and neurogenesis in seasonality is indeed supported by several lines of evidence such as dynamic photoperiodic modifications not only at the GnRH cell bodies, but also at their end-feet in the median eminence (97,115,116,214–216).
The kisspeptin (KiSS) neurones of the arcuate nucleus (a nucleus of the MBH) constitute another potential link between T3 and GnRH neurones (217,218). It is noteworthy that KiSS displays opposite melatonin-dependent photoperiodic variations in sheep and hamster (219–221). Transcription of KiSS might be dependent upon a competition between oestrogen receptors and thyroid hormone receptors for DNA binding at its promoter (222) because these nuclear receptors exert transcriptional action towards the same response elements (223,224). It is equally possible that T3 blocks the action of oestradiol without functional cross-talk (e.g. by acting at the level of the median eminence to prevent GnRH release), as suggested in the quail (225).
There are several potential non-mutually exclusive mechanisms that could account for the development of LD refractoriness. Although much has been learned during recent years regarding the molecular menchanisms by which photoperiod synchronises reproduction, we still do not know how animals undergo spontaneous transitions in physiology independently of the photoperiodic message. However, the clarification of the mechanisms leading to the control of T3 by LD now provides a strong lead to allow progress towards an improved understanding on this matter.
I wish to thank the following colleagues for their insightful and constructive criticisms on earlier drafts of this review: Gerald Lincoln (University of Edinburgh, UK); David Hazlerigg (University of Aberdeen, UK); Paul Pévet and Mireille Masson-Pévet (INCI, CNRS UPR 3212, Strasbourg); and Laurence Dufourny and Massimiliano Beltramo (INRA UMR85, Nouzilly).