The Ovarian Cycle and Fertility

In one 1984 study, single progesterone samples from normally cycling women in the mid-to-late luteal phase ranged from 2.3 to 40.1 ng per mL. Both ends came from healthy women. The span pools different women and sampling times, but progesterone also swung widely within single 24-hour profiles, so a gland whose output moves that much within a day is not running at a steady rate. This lesson asks how the ovary turns a crowd of growing follicles into one ovulation at a predictable time, and what a single measurement of that system can show.

The first look at the cycle, the uterine cycle and Knobil’s experiment with pulsed and continuous GnRH in monkeys is in animal reproduction; feedback loops and why one sample of a pulsatile hormone is hard to read are in endocrine axes and testing.

Most follicles die; one is chosen

A follicle is an oocyte, arrested in prophase I of meiosis since before birth (chromosomes and meiosis), with the somatic cells around it. The ovary holds a fixed reserve of primordial follicles, each an oocyte with a single flat layer of granulosa cells, and two kinds of recruitment draw on it. Initial recruitment moves primordial follicles into growth continuously, from before puberty on, and does not depend on the cycle’s FSH. Cyclic recruitment acts much later, on follicles that have grown for months into small antral follicles (with a fluid-filled cavity, the antrum). In the human model reviewed by McGee and Hsueh (2000), rising FSH rescues a cohort of small antral follicles from atresia, the degeneration that ends almost every follicle.

The FSH that rescues them rises at the end of the previous cycle: as the corpus luteum regresses, estradiol, progesterone and inhibin A fall, and with that brake lifted FSH rises. The rescued follicles then make estradiol and inhibin B, a granulosa-cell protein that suppresses pituitary FSH secretion. Inhibin B peaks just after the intercycle FSH rise and then falls, so rising estradiol is the main brake through the mid-follicular phase. The cohort lowers the very hormone that saved it: as FSH falls through the mid-follicular phase, usually one follicle keeps growing and becomes dominant, and the rest undergo atresia. FSH responsiveness, the acquisition of granulosa LH receptors and local growth support together decide which one.

Is it the height of FSH that matters, or how long it stays up? Schipper, Hop and Fauser (1998) raised FSH in normally ovulating women in two ways. A brief early rise did not maintain extra follicles; a moderate rise sustained into the mid-to-late follicular phase let several follicles keep growing. That supports an FSH window: selection depends on how long FSH stays above each follicle’s threshold, not on the threshold alone. The arms differed in timing as well as duration, so the study does not define one threshold value.

Baerwald, Adams and Pierson (2003) scanned 50 women by transvaginal ultrasound every day for one interovulatory interval and found two waves of follicle growth in 68 percent and three in 32 percent. A wave is a synchronous round of antral follicle growth, defined here by follicle size; only the last wave of an interval ended in ovulation. The wave model is a proposal from this study and is still debated. Waves are not extra ovulations.

Two cells, two gonadotropins, one estrogen

In the human follicle before ovulation, neither of its two main cell types makes much estradiol from cholesterol on its own; like a two-station kitchen, the layers share the job across a basement membrane. Theca cells, outside the membrane, carry LH receptors and CYP17 (17-alpha-hydroxylase and 17,20-lyase), the enzyme that turns pregnane steroids into androgens. In isolated human theca cells from 4 to 15 mm follicles incubated for two hours, androstenedione was the main steroid measured (Nilsson, Dennefors and Hamberger, 1984). Granulosa cells, inside, carry FSH receptors and aromatase (CYP19A1), the enzyme that converts androgens to estrogens, and FSH raises their aromatase. Androstenedione crosses to the granulosa cells, where aromatase converts it to estrone, and testosterone to estradiol; a 17-beta-hydroxysteroid dehydrogenase interconverts the pairs, and estradiol is the follicle’s main secreted estrogen. This is the two-cell, two-gonadotropin model. Both gonadotropin receptors are G protein-coupled receptors acting largely through cyclic AMP (G protein-coupled receptors, cAMP and PKA).

The model also offers an account of how selection ends: as granulosa cells differentiate under FSH, they acquire functional LH receptors (reviewed by Hillier, Whitelaw and Smyth in 1994; the two-cell idea itself comes from earlier work). The most mature follicle then responds to LH as well as FSH, so it keeps growing and making estradiol while FSH falls. Smaller follicles, whose granulosa cells have not gained the same LH responsiveness, usually cannot.

The inhibins report on the follicles. Groome and colleagues (1996), with assays that measure inhibin A and inhibin B separately, found inhibin B rising rapidly in the early follicular phase and inhibin A reaching its maximum in the midluteal phase. Inhibin B tracks the growing cohort of small antral follicles; inhibin A is highest when the corpus luteum is most active.

From brake to accelerator: the switch to positive feedback

For most of the cycle, estradiol and progesterone are negative feedback signals: they lower GnRH release from the hypothalamus and LH release from the pituitary. Veldhuis and colleagues (1987) tested the estradiol half of this in 12 postmenopausal women, whose ovaries make little estradiol: when estradiol was restored, LH pulse frequency slowed within 24 hours. That study did not manipulate progesterone. Because each LH pulse is thought to follow a GnRH pulse, slower LH pulses point to slower GnRH pulses, an inference: GnRH in portal blood cannot be sampled in women.

Late in the follicular phase the sign reverses: when estradiol from the dominant follicle stays high for long enough, it becomes positive feedback, a signal that raises its own trigger, and LH release climbs steeply in the LH surge (a GnRH surge accompanies it in the rodent models reviewed). A 2020 review by Moenter and colleagues notes that most of the mechanistic work is in laboratory mice and that it is still debated whether women have a GnRH surge. The surge is fast: in five women sampled every 2 hours, Hoff, Quigley and Yen (1983) found LH doubling within 2 hours of onset and a surge lasting about 48 hours on average. Textbooks place surge onset roughly 34 to 36 hours before ovulation, and individual intervals vary widely. No single estradiol level or duration for the switch is given here, because quoted values differ between sources.

Where does estradiol act? In rodents, GnRH neurons carry little or no estrogen receptor alpha (ERalpha), so the signal is thought to pass through other neurons. The leading candidates are kisspeptin neurons, which act on GnRH neurons through the kisspeptin receptor (KISS1R, formerly GPR54), a G protein-coupled receptor; people with loss-of-function KISS1R mutations fail to enter puberty normally (de Roux and colleagues, 2003; Seminara and colleagues, 2003). In gonadectomized mice, arcuate kisspeptin neurons were shown to act as the GnRH pulse generator (Clarkson and colleagues, 2017, told in animal reproduction), while a second population in the anteroventral periventricular nucleus (AVPV), carrying ERalpha, has been proposed to carry positive feedback. Both estrogen receptors and the progesterone receptor are nuclear receptors (nuclear receptors); the portal route to the pituitary is in hypothalamus and autonomic control.

Sharon Dubois and colleagues (2015) asked whether one receptor in one cell class carries both of estradiol’s actions. They bred female mice in which Cre recombinase, made only in kisspeptin-expressing cells, deleted the gene for ERalpha, then ovariectomized adult knockouts and wild-type controls and gave estradiol regimens that produce either negative feedback or a surge. Estradiol still lowered mean LH in the knockouts, but they made no LH surge. In this mouse paradigm, ERalpha in kisspeptin cells is required for the estradiol-induced surge and is not the only route for suppression.

What it does not show matters as much. The deletion was present from development, so developmental and adult signaling requirements are confounded, and both AVPV and arcuate cells lost the receptor, so it cannot say which population is responsible. And it is a mouse result: in women the surge circuitry is less well localized, and the rodent AVPV pathway cannot be stated as the human mechanism. The same brake underlies one way hormonal contraception prevents ovulation: sustained steroid negative feedback blocks the surge.

Ovulation and a corpus luteum that pulses

The surge restarts meiosis in the oocyte, turns the granulosa and theca cells into luteal cells, and leads to rupture of the follicle wall. The oocyte part of this was worked out in mice. Norris and colleagues (2009) imaged living, follicle-enclosed mouse oocytes with FRET sensors, a cyclic-nucleotide-binding domain placed between two fluorescent proteins, so binding changes the energy transfer between them. Before LH, cGMP made in the granulosa cells passes into the oocyte through gap junctions and inhibits PDE3A, the oocyte’s cAMP-degrading phosphodiesterase, so oocyte cAMP stays high and meiosis stays arrested. LH acts on receptors on the somatic cells: it lowered their cGMP from about 2 micromolar to about 80 nanomolar and closed the gap junctions between them. Estimated oocyte cGMP then fell from about 1 micromolar to about 40 nanomolar (about 25-fold), PDE3A activity rose about 5-fold, and cAMP fell from about 700 to about 140 nanomolar (5-fold), and meiosis resumed. The stimulus LH sends goes to the somatic cells; what the oocyte sees is the withdrawal of an inhibitor of an inhibitor. The oocyte then completes meiosis I and arrests again at metaphase II, as animal reproduction describes.

After rupture, luteinized granulosa cells and theca-lutein cells form the corpus luteum, which secretes progesterone and, in the midluteal phase, a second rise of estradiol. Progesterone slows LH pulse frequency. In the cycle-app cohort analysed by Bull and colleagues (told in animal reproduction), the mean estimated luteal phase was 12.4 days. If an embryo implants, its hCG acts through the LH receptor and keeps the corpus luteum working; that rescue belongs to pregnancy, birth and lactation.

The classic experiment: Filicori, Butler and Crowley, 1984

Is the corpus luteum a steady gland once it forms? Marco Filicori, James P. Butler and William F. Crowley Jr. studied 15 normally cycling women, in separate early, middle and late luteal groups, drawing blood every 10 minutes for 24 hours, with LH measured in every sample and progesterone at the same frequency in six of the women. In the early luteal phase it was stable and unrelated to LH, but later it was not. LH pulse frequency fell across the luteal phase, from a mean of 15.2 to 8.4 pulses per 24 hours between the early and late groups. In the mid and late luteal phase, progesterone was episodic, rising after LH pulses, and its measured values ranged from 2.3 to 40.1 ng per mL across the sampled women and times.

Ruled in: the human corpus luteum secretes progesterone in episodes that follow LH pulses, so luteal function remains under pulsatile pituitary control. Ruled out: that one progesterone sample measures the day’s luteal output. Not shown: that each LH pulse causes the episode after it, since the evidence is timing, not blockade; and change within one woman, since each phase was a different group.

Reading the cycle’s hormone curves

Put one cycle on a single time axis, aligned to the day of ovulation rather than to a calendar day 14, and the sequence reads in order. FSH rises at the end of the previous cycle and falls through the mid-follicular phase; estradiol climbs from the dominant follicle to a late-follicular peak; LH surges within a day of that peak; progesterone begins to rise shortly before the surge and climbs after ovulation, with a second, smaller estradiol rise, and both fall at the cycle’s end as FSH rises again. Inhibin B rises early with the cohort, inhibin A in the midluteal phase with the corpus luteum.

A curve is only as fine as the sampling that made it. Using the group means from Filicori and colleagues:

pulse interval = 1,440 min per day / pulses per day

samples per interval = pulse interval / sampling interval

Early luteal, 15.2 pulses per 24 hours: 1,440 / 15.2 = 94.7 min between pulses. Late luteal, 8.4 pulses: 1,440 / 8.4 = 171.4 min. Sampling every 10 minutes gives 94.7 / 10 = 9.5 samples per early-luteal interval and 171.4 / 10 = 17.1 per late-luteal interval, enough to follow the rises and falls in that study. Sampling every 60 minutes gives 1.6 and 2.9. As a rule of thumb, below about two samples per interval pulses cannot be resolved, and even above it peaks can fall between samples. One sample a day gives no pulse information, which is why frequent sampling found luteal progesterone episodes that sparse sampling misses.

Menopause is the same circuit with almost none of its input: as the follicle supply runs out, inhibin B, inhibin A and estradiol fall, ovarian negative feedback is largely lost, and FSH rises, usually by more than LH, with progesterone low for want of a corpus luteum. Ageing of the hypothalamus and pituitary contributes as well, so the rise is not only the loss of feedback.

Design problem

Dubois and colleagues deleted estrogen receptor alpha from every kisspeptin-expressing cell. Design an experiment in mice that asks a narrower question: does estradiol’s positive feedback on the LH surge act through kisspeptin neurons of the anteroventral periventricular region, the arcuate population, or both?