Persistence of Sleep-Associated Decrease in GnRH Pulse Frequency in the Absence of Gonadal Steroids

Abstract

Context:

There is dramatic slowing of GnRH pulse frequency during sleep in the early follicular phase of the menstrual cycle, but it is unknown whether this represents a primary effect of sleep or is dependent upon the sex steroid environment.

Objectives:

Our objective was to determine 1) whether sleep affects GnRH pulse frequency in postmenopausal women (PMW) in whom gonadal hormones are low and 2) whether this relationship changes with aging.

Design and Setting:

Studies were performed in the Clinical Research Center of an academic medical center.

Subjects:

Subjects included healthy PMW, 45–55 (n = 8) and 70–80 (n = 6) years old.

Interventions:

Subjects were studied during one night of polysomnographic-recorded sleep and one night of monitored wake during which blood was sampled every 5 min for 8 h.

Main Outcome Measures:

Pulsatile secretion of free α-subunit (FAS), a marker of GnRH secretion, was assessed.

Results:

There were no differences in sleep macroarchitecture or sleep efficiency [75 ± 12% (mean ± sd)] between older and younger PMW. The FAS interpulse interval was longer during sleep than nighttime wake in all women (60.5 ± 4.3 vs. 52.0 ± 2.8 min, P = 0.03) with a similar effect in the two groups. FAS pulse amplitude did not differ between sleep and wake periods (474.8 ± 36.7 vs. 478.2 ± 36.5 ng/liter, P = 0.9).

Conclusions:

Sleep is associated with a significant decline in GnRH pulse frequency in both older and younger PMW. Its persistence in PMW reinforces the important connection between sleep and GnRH secretion.

There is a complex relationship between sleep and the central components of the reproductive axis that changes across reproductive development. This interaction is first manifest in early puberty when a sleep-specific rise in LH pulse amplitude is observed (1). In contrast, women in the early follicular phase (EFP) of the menstrual cycle demonstrate a nocturnal slowing of GnRH pulse frequency (2–10), which sleep-reversal studies have confirmed to be specifically tied to sleep rather than to time of day (4, 11). Because FSH synthesis is supported by slow-frequency GnRH pulses (12), we have hypothesized that sleep slowing in the EFP may be critical for maintenance of the FSH levels required for follicular recruitment in the EFP (4). Nighttime slowing has also been demonstrated, although less consistently, in the midluteal phase and late follicular phase (2;8–10;13–17) and was also noted in one study in women with natural menopause (18), although previous studies did not control for time of day and not all studies used polysomnography (PSG) to document sleep.

It is well known that progesterone (P) induces slowing of the GnRH pulse generator, and several lines of evidence further suggest that recent exposure to P may play a role in the sleep-specific slowing of GnRH pulsatility. Women with anovulatory polycystic ovarian syndrome who lack the recent exposure to P that is characteristic of the EFP do not demonstrate wake-sleep differences in LH pulse frequency (19). P administration studies show that, in the presence of midluteal levels of 17β-estradiol (E2), the inhibitory effect of P on LH pulse frequency persists for a full week after its withdrawal (20, 21), suggesting that some effects of P may continue into the EFP of normal cycles. Finally, P negative feedback on the GnRH pulse generator is mediated by opioids (22), and the opioid receptor antagonist naloxone prevents the sleep-associated decrease in LH pulse frequency in EFP women (6).

To determine whether there is an endogenous effect of sleep on the GnRH pulse generator that is seen in the absence of gonadal hormone exposure, we compared GnRH pulse dynamics during wake and electroencephalographically monitored sleep in postmenopausal women (PMW), in whom endogenous gonadal steroids are uniformly low. In the current study, we used free α-subunit (FAS) as a marker of underlying GnRH secretion to demonstrate that GnRH pulse frequency is also decreased during sleep in PMW, suggesting that there is a primary inhibitory effect of sleep on GnRH pulse frequency.

Subjects and Methods

Subjects

Eight younger (age 45–55 yr) and six older (age 70–80 yr) healthy PMW participated. Subjects were more than 12 months from their last menstrual period, nonobese [body mass index (BMI) ≤ 30 kg/m²], euthyroid, and normoprolactinemic. Subjects were screened and excluded for sleep disorders. They were not on any medications, and none had taken hormone replacement therapy within the last 2 months. They kept a regular sleep schedule, confirmed with a diary, and had not undergone transmeridian travel within 3 months of the study. All subjects had serum E2 levels below 20 pg/ml and serum LH and FSH levels above 30 IU/liter, consistent with their postmenopausal status. Subjects took iron supplements (ferrous gluconate 324 mg twice a day) 1 month before until 1 month after the study.

The study protocol was approved by the Partners Human Research Committee, and informed consent was obtained from each subject before her participation.

Experimental protocol

Subjects were required to maintain a constant sleep-wake schedule for 4 wk before the 3-d admission to the Clinical Research Center of the Massachusetts General Hospital. The first night served as an acclimatization night with PSG monitoring [GRASS 8-channel model (GRASS Technologies, Astro-Med, Inc., West Warwick, RI)] during sleep but without blood sampling. The following night, blood sampling was performed every 5 min for 8 h during sleep, monitored with PSG. A long line (422 cm, 6.4 ml residual volume) was connected for blood withdrawal outside of the sleeping room to minimize sleep disruption. Samples were drawn using a discard volume previously determined to result in no dilution of the measured sample and a blood sparing technique in which all discard volume is reinfused (23). On the third night, subjects again underwent 8 h of frequent sampling but remained awake under constant supervision, in dim light (10–20 Lux), in constant temperature, and in a semirecumbent position. FAS was measured in all samples, whereas E2, LH, and FSH were measured in pools from each frequent sampling study. It has previously been shown that a 5-min sampling interval is necessary for optimal pulse detection at the GnRH pulse frequencies that are present in PMW and that FAS is superior to LH as a marker of GnRH at these frequencies (24, 25).

Assays

Serum LH, FSH, FAS, and E2 levels were determined by immunoassay, as previously described (24, 26). The interassay coefficient of variation for FAS is 9.4–20.0% for levels between 163.4 and 595.4 ng/liter. Samples with FAS levels higher than the assay range were diluted and remeasured. The mean intraassay coefficient of variation for FAS was 5.5 ± 1.7%. Gonadotropin values are expressed in international units per liter as equivalents of the Second International Reference Preparation 71/223 of human menopausal gonadotropins (hmg). The following formulas can be used to convert to the pituitary (pit) Second International Standard 80/552: LH (pit) = 0.41 × (LH hmg) − 0.32, and FSH (pit) = 0.57 × (FSH hmg) − 0.25.

Data analysis

Sample size calculation

Relying on our previous work showing a FAS interpulse interval (IPI) of 52.6 min (1 sd = 3.1 min) in PMW studied during the day (24), we found that 14 subjects were needed to detect a 10% difference between wake and sleep IPI at a power of 80% with a two-sided 0.05 significance level.

Polysomnography

The PSG sleep recordings were scored visually by trained sleep technicians in 30-sec epochs according to the criteria of Rechtschaffen and Kales (27). Sleep parameters were compared between younger and older subjects using unpaired, two-sided Student's t tests.

Pulse analysis

Pulsatile secretion of FAS was analyzed using a validated modification of the Santen and Bardin method (25, 28). Mean FAS and pooled E2, LH, and FSH levels were compared between younger and older PMW using unpaired, two-sided Student's t tests (or the Mann Whitney rank sum test if not normally distributed). Two-way ANOVA for repeated measures with post hoc Newman-Keuls testing was performed to examine the independent effects of age and sleep on FAS IPI. Values are expressed as mean ± sem, unless otherwise indicated. P values < 0.05 are considered significant.

Results

Baseline characteristics

The PMW were clearly dichotomized in terms of age and years from menopause but did not differ in BMI. Menopausal status was confirmed in all subjects by low E2 levels and elevated LH, FSH, and FAS levels (Table 1). LH, FSH, and FAS were lower in the older compared with younger PMW, as has been previously reported, whereas E2 was not different between the two age groups (Table 1) or between the sleep and awake nights [23.1 ± 1.7 and 22.8 ± 1.6 pg/ml (84.8 ± 6.2 and 83.6 ± 5.9 pmol/liter), respectively]. Three subjects in the older group had undergone bilateral oophorectomy in the past, whereas the remainder of women in both groups had undergone natural menopause.

Table 1.

Characteristics of younger and older postmenopausal subjects

	Younger PMW	Older PMW
n	8	6
Age (yr) mean ± sd)	51 ± 2	73 ± 2a
yr since menopause	3.3 ± 0.9	28.8 ± 3.2a
BMI (kg/m2)	24.1 ± 1.3	25.4 ± 1.2
E2 (pg/ml)	26.6 ± 3.2	21.6 ± 1.0
LH (IU/liter)	114.7 ± 16.7	56.7 ± 9.6a
FSH (IU/liter)	141.9 ± 14.9	94.6 ± 11.1a
FAS (ng/liter)	1453.8 ± 171.8	491.6 ± 90.0a

Values are expressed as mean ± sem except as indicated. Gonadotropins are expressed in international units per liter as equivalents of the Second International Reference Preparation of human menopausal gonadotropins. Please see text for conversion to international units per liter expressed in relation to pituitary standards for LH and FSH. To convert picograms per milliliter to picomoles per liter, multiply by 3.67.

^aP < 0.05 (younger vs. older).

Sleep characteristics

There were no differences in sleep efficiency or latency between the younger and older PMW and no difference in the relative amounts of non-rapid eye movement (non-REM) or REM sleep (Table 2). Sleep efficiency [75 ± 12% (mean ± sd)] was similar to what has been previously reported in frequent sampling studies during sleep in older women (29). There was also no difference in the total number of awakenings in the PMW compared with what we have previously observed in younger women during the EFP (15.5 ± 1.3 vs. 14.0 ± 1.8, respectively; P = 0.5) (unpublished data from Ref. 4).

Table 2.

There were no differences in sleep parameters between younger and older PMW

	Younger PMW	Older PMW
Stage 1 (% TST)	17.7 ± 2.7	15.0 ± 3.8
Stage 2 (% TST)	37.5 ± 2.2	34.5 ± 7.2
Stage 3 + 4 (% TST)	24.7 ± 2.1	23.6 ± 3.9
Stage REM (% TST)	16.6 ± 3.6	14.1 ± 3.6
Number of awakenings	15.4 ± 1.9	16.0 ± 1.3
Sleep efficiency (TST/TIB %)	76.1 ± 4.8	74.6 ± 4.3
Sleep latency (min)	15.6 ± 6.0	24.5 ± 3.9

Values are expressed as mean ± sem. TIB, Time in bed; TST, total sleep time.

Effect of age and sleep on FAS IPI and pulse amplitude

The FAS IPI was influenced by both age (P < 0.01) and sleep-wake state (P < 0.01). FAS IPI was longer, indicating a slower pulse frequency, in older compared with younger PMW (64.6 ± 3.1 vs. 50.0 ± 3.3 min, P < 0.01), as previously described (24) and was longer during sleep than nighttime wake in all women (60.5 ± 4.3 vs. 52.0 ± 2.8 min, P = 0.03) (Figs. 1 and 2). There was no interaction between age and sleep-wake state, suggesting that the effect of aging on IPI is similar during wake and sleep (Fig. 2). Sleep was associated with a 17.2 ± 11.0 and 17.0 ± 10.0% increase in FAS IPI in younger and older PMW, respectively, which is a smaller sleep-related increment than we have previously reported in women studied during the EFP (68.2 ± 10.6%, P < 0.01) (4) (Fig. 2).

Fig. 1.

Sleep stages plotted for the 8 h during the sleep night, and serum FAS concentrations determined at 5-min intervals for 8 h during sleep and wake nights in representative younger (top) and older (bottom) PMW. The inverted triangles indicate FAS pulses.

Fig. 2.

The FAS IPI increased with aging in PMW and was longer during sleep than during wake in both younger (n = 8) and older (n = 6) PMW. The increase in FAS IPI during sleep in PMW was less than that previously reported in women (n = 11) studied during the EFP (4). White and black lines within bars represent mean − 1 sem. *, P < 0.05, sleep vs. wake for each age group.

The mean FAS pulse amplitude was significantly greater in younger compared with older PMW (625.4 ± 32.6 vs. 188.5 ± 14.2 ng/liter, P < 0.01) as has been previously demonstrated (24), but did not differ between sleep and wake periods (474.8 ± 36.7 vs. 478.2 ± 36.5 ng/liter, P = 0.9).

Discussion

There is now compelling evidence that sleep is an important modulator of GnRH dynamics. Because this effect is most consistently seen during puberty and in the EFP of the normal menstrual cycle in reproductive-aged women, this suggests that the effect of sleep on the GnRH pulse generator may be dependent on the hormonal milieu (2–10). In the current study, we demonstrate that sleep-related slowing of GnRH pulse frequency is maintained in younger and older PMW, although the inhibitory effect of sleep is less than in previous studies of women in the EFP of a normal menstrual cycle. These findings suggest that sleep per se decreases GnRH pulse frequency in reproductive-aged and PMW and that its effect may be amplified by gonadal steroids.

The results of earlier studies on the effect of sleep on LH pulse frequency in PMW have been inconsistent. Studies by Santoro et al. (18) demonstrated sleep-specific slowing, whereas other studies found no difference in LH secretion during waking and sleep in PMW (8, 30). However, all have had limitations that were specifically addressed in the current study by using FAS as a marker of GnRH, by the use of an optimal blood sampling frequency, and by the use of PSG to document sleep. Several previous studies (8, 30) performed blood sampling for LH at 15- or 20-min intervals, which represents an insensitive sampling interval for pulse detection in PMW compared with the 5-min sampling interval used in the present study (24). In addition, due to its shorter half-life, FAS is a more faithful marker of GnRH stimulation of the pituitary than is LH at the fast GnRH pulse frequencies characteristic of PMW (25). This consideration is particularly relevant to the current study in PMW because the half-life of LH is markedly prolonged in the absence of gonadal steroids, whereas that of FAS is unaffected (26). Finally, several studies (8, 18) relied on observer-reported assessments of sleep, raising the possibility that brief awakenings may not have been recognized. Nocturnal awakenings precede and potentially induce LH pulses in younger women (4), and thus, the lack of PSG in these studies may have significantly confounded earlier results.

Our observation that sleep-related slowing of LH pulse frequency is preserved in PMW suggests the presence of an underlying effect of sleep on GnRH pulse frequency. A number of studies have addressed the potential mechanisms underlying sleep-related slowing, but to date, all have been performed in EFP women. There is some evidence from pharmacological studies in humans that serotonin and dopamine modulate the sleep/wake state (31), but studies employing neurotransmitter antagonists have failed to show a role for serotonin or dopamine in mediating this effect (7, 32). Prolactin and melatonin increase during sleep and may be potential mediators of the link between sleep and GnRH pulse frequency. However, studies have not supported a relationship between LH and prolactin in EFP women (5), and an effect of melatonin on GnRH secretion in humans has not been convincingly demonstrated (33). Although the nocturnal decline in LH pulse frequency during the EFP occurs in close association with the nocturnal leptin peak, there are no studies that have dissociated the effects of leptin from sleep in an attempt to address causality (13). The most promising candidates to date for sleep-related modulation of GnRH pulse frequency are endogenous opioids because studies have shown that naloxone, an opioid receptor antagonist, reverses GnRH pulse slowing in women studied during sleep in the EFP (6).

Although significant sleep-related slowing of GnRH pulse frequency was observed in the current studies of PMW, the effect of sleep was considerably less than what we have previously reported in women in the EFP (4). In addition to the possibility that the attenuated effect of sleep in PMW may reflect altered sensitivity of the GnRH pulse generator with aging, there are a number of other possibilities that must be considered. Studies in older women have demonstrated that periodic blood sampling, even when performed remotely through a long iv catheter, causes a significant decrease in sleep efficiency and an increase in the number of awakenings (29). Because awakenings may induce LH pulses (4), our estimate of the slowing of GnRH pulse frequency during sleep in PMW is likely to be a conservative one. Importantly, we did not observe a difference in the number of awakenings between the PMW and younger women previously studied during the EFP, suggesting that differences in sleep consolidation are unlikely to account for the greater effect of sleep on GnRH pulsatility in younger women compared with PMW. Inhibition of the GnRH pulse generator in the EFP is associated with deep sleep, and it is possible that the shift from slow-wave and REM sleep to shallower sleep stages with aging (34) may explain the greater inhibition of GnRH pulse frequency by sleep in reproductive-aged compared with PMW. The effect of aging on sleep parameters, however, plateaus around age 60 (35), which may well account for the lack of a difference in the effect of sleep on GnRH pulse frequency in the 45- to 55- compared with 70- to 80-yr-old PMW in the current study.

An alternative hypothesis for the diminished effect of sleep on GnRH pulse frequency in PMW compared with women studied in the EFP is that gonadal steroids augment the impact of sleep in premenopausal women and, specifically, that recent exposure to P sensitizes the GnRH pulse generator to the inhibitory effects of sleep. In this regard, it is of interest that the inhibitory effect of both sleep and P on the GnRH pulse generator have been linked to endogenous opioids (6, 22). Furthermore, recent studies in the ewe have identified dynorphin as the specific opioid mediating the inhibitory effect of P (36). This finding has been complemented by human postmortem studies demonstrating decreased prodynorphin mRNA expression in PMW, who lack recent P exposure, compared with premenopausal women (37). Although it is dynorphin within the arcuate nucleus that has been tied to P negative feedback (38), dynorphin is also coexpressed by lateral hypothalamic orexin (hypocretin) neurons (39) whose primary function is to maintain wakefulness. At this time, however, it is unclear whether the roles of dynorphin in sleep/wake and reproductive hormone dynamics are linked and, if so, how dynorphin neurons might coordinate these complex physiological systems.

In summary, we have now shown that the sleep-related slowing of the GnRH pulse generator, first documented in EFP women, is present in PMW. The preservation of this sleep-related effect in a low sex steroid environment suggests that it is driven by sleep itself. The relative attenuation of the effect of sleep in PMW compared with those in the EFP further suggests that the inhibitory effect of sleep on the GnRH pulse generator may be augmented by gonadal steroids. Future studies in which gonadal hormone levels are manipulated in PMW will allow us to test this hypothesis directly.