Estradiol action in the female hypothalamo-pituitary-gonadal axis

Abstract

It has now been about a century since a flurry of discoveries identified first the pituitary, then more specifically the anterior pituitary and soon thereafter the central nervous system as components regulating gonadal and downstream reproductive functions. This was an era of ablation/replacement designs using at first rudimentary and then increasingly pure preparations of gonadal and pituitary ‘activities’ or transplanting actual glands, whole or homogenized, among subjects. There was, of course, controversy as is typical of lively and productive scientific debates to this day. The goals of this commentary are to briefly review the history of this work and how the terms referring to interactions among the components of the hypothalamo (as the central neural component was soon associated with)-pituitary-gonadal (HPG) axis evolved, and then to question if the current terms used have kept up with our understanding of the system. The focus in this review will be the actions of estradiol primarily upon the hypothalamus. Important actions of progesterone on the hypothalamus as well as both steroids on the pituitary response to hypothalamic factors are both acknowledged and largely ignored in this document, as are any sex differences as we focus on females.

Early history of the HPG axis

Pflüger wrote in 1877 that “Accommodations of living beings are always teleological, are subject to the law of causality.”¹ Defining the postulated interactions among HPG components in a teleologic manner, that is the purpose they serve vs the cause by which they arise², has remained a consistent theme. Among the first hints of a humoral relationship between the pituitary and the gonads were observations that the former were enlarged upon removal of the latter^3,4. This was due to an increase in size and vacuolization of some basophilic cells, now known to be the source of gonadotropins, after castration⁵. Feeding or intraperitoneal injections of pituitary extracts to rats produced mixed results, with both activation and inhibition of the reproductive system reported in each sex^6–8. The advent of transsphenoidal pituitary ablation in rodents in 1926 by Smith⁹ in combination with pituitary replacement built consensus around an activating effect of the anterior pituitary on the gonads. Ablation of the whole gland, but not just the posterior pituitary, reduced the size of reproductive organs, which could be partially restored by daily intramuscular injections of fresh anterior pituitary. These studies confirmed reports of a lack of effect of posterior pituitary hormones on the gonads in mammals. Together with earlier ablation studies of the various subdivisions of the pituitary of Rana pipiens¹⁰, this work supported the postulate that each part of the pituitary served different physiological functions and that the anterior pituitary was the major influencer of reproductive function in mammals.

In studies of the reproductive tracts, pituitary ablation/replacement along with castration soon demonstrated the requirement of intact gonads for pituitary replacement to have trophic effects upon the reproductive tract in both sexes¹¹. Treating immature female rodents with pituitaries from castrated animals was more effective in advancing vaginal opening and increasing ovarian mass than treating with pituitaries from gonad-intact animals, suggesting functional changes in the pituitary upon castration¹². Activation of immature ovaries upon their transplantation into adults further solidified the relationship between the anterior pituitary and gonads, and led to early speculation that the release of the gonad-activating substance of the pituitary was periodic and perhaps linked to vaginal cycle stages¹¹. As these discoveries progressed, the nature of the relationship appeared to clarify. Work in freemartin cattle by Lillie¹³ and rodents by Steinach¹⁴ led to the hypothesis that the gonadal hormones of the sexes antagonized one another, leading to damage to the gonadal and reproductive systems of the opposite sex. The gonadal antagonism hypothesis had sustained popularity through this period.

A new concept of hormone interactions

Other groups were working to identify the active substances, or hormones as Starling first referred to them¹⁵, that produced these responses. Studies of urine from women in different reproductive states¹⁶ suggested the pituitary likely contained at least two substances with different gonadal functions, soon identified as a luteinizing hormone (LH) and a follicle-stimulating hormone (FSH)¹⁷. The first crystallization of the follicle-derived hormone theelin, or estrone was published in 1930¹⁸, followed by estriol¹⁹ and eventually estradiol²⁰. These biochemical advances helped standardize extract preparation and enabled increased rigor and reproducibility for the physiologic studies.

Pivotal in the evolution of thought on the interrelationships between the pituitary and gonad was the elegant reasoning²¹ and experimental work²² of Moore and Price, who proposed “A new conception of hormone interactions” in their 1932 publication²². These are so simply stated that an attempt to paraphrase would not do them justice, and they are quoted below in full:

• “1. Gonad hormones stimulate homologous reproductive accessories, but are without effect upon heterologous accessories.

• 2. Secretions produced by the hypophysis stimulate the gonads to function both in germ cell production and in hormone secretion.

• 3. Gonad hormones have no direct effect on the gonads of either the same, or the opposite, sex.

• 4. Gonad hormones, of either sex, exert a depressing effect upon the hypophysis which results in a diminished amount of the sex-stimulating factor available to the organism.”

While point 3 is now known to be incorrect for both sexes, these statements explained a great deal of existing data regarding the interactions between the anterior pituitary and gonad, and quickly supplanted the gonadal antagonism hypothesis.

Two additional aspects of this interaction were co-emerging. First, Hohlweg and Junkmann added the central nervous system to the Moore and Price model, proposing that gonadal hormones acted centrally to control pituitary output²³. Although they incorrectly bypassed the pituitary as a recipient of gonadal hormone action, this added the central neural component of the hypothalamo-pituitary-gonadal (HPG) axis. Central involvement in reproduction had been hinted at in prior studies that intentionally damaged the tuber cinereum of the hypothalamus, this left the pituitary intact but nonetheless reduced reproductive organ function^9,11. Second, Hohlweg made what he termed a surprising discovery when larger doses of Progynon, an early estrogen preparation, was observed to induce both pituitary enlargement (with a different underlying histological changes than caused by castration) and luteinization of the follicles²⁴. He speculated the formation of corpora lutea was caused by a central-neural-mediated increase in release of a separate luteinizing hormone from the pituitary. This was contrary to the inhibitory actions on the gonads observed in most studies using lower doses of steroid. Earlier work had suggested a similar dose dependence for what we now know as estrogenic responses with higher concentrations needed to induce estrous behavior than to induce vaginal cornification²⁵. Fevold et al., showed evidence for an initial stimulation and subsequent suppression of ovarian mass in response to high doses of oestrin²⁶; the timing of the observations and use of ovarian mass and structures as the bioassay caused the initial suppression upon estrogen treatment to be missed. The concept of biphasic estrogen action on the gonads and a requirement for the pituitary in these actions was thus emerging and increasingly associated with cyclic changes, including in humans^27–32.

Estradiol and feedback

It is likely that any discussion of the HPG axis amongst scientists will before long mention the phenomena of negative and positive feedback actions of estradiol. Interestingly, none of the early work cited above use the term feedback. Instead, phrases such as: the reaction of the various glands, the reciprocal interaction between the pituitary and gonad, reciprocity, complex interrelationships, interplay, influence, suppression, depression, activation, stimulation are used. Feedback is not only missing with regard to steroids, but entirely from searchable PDFs of the cited work. While not an all-encompassing review, the above is a reasonable sample of studies defining this axis. The lack of use of feedback extends to some classics of the field including Harris³³, Everett and Sawyer³⁴, and Greep and Jones’ report at the Laurentian Conference in 1950³⁵. Even Dorothy Price’s further work did not use the term feedback³⁶. But on page 6 of his iconic 1955 book Neural Control of the Pituitary Gland, Geoffrey Harris used feedback³⁷, a noted change from his 1948 review with the same title³³.

What brought about this change? A likely suspect is the 1952 Ramon Guiteras lecture by Charles Huggins at the annual meeting of the American Urological Association in Atlantic City³⁷. Huggins, who shared the Nobel Prize in Physiology and Medicine in 1966 with Peyton Rous for his work on cancer cell biology, suggested in his address that “Feedback circuits are now extensively utilized by engineers and the feedback concept is useful in visualizing endocrine physiologic control devices of the body.” Huggins melded physiology with emerging concepts from engineering, such as Black’s³⁸ invention of the negative feedback amplifier to produce more stable signals, and Norbert Wiener’s theory of Cybernetics³⁹, “a new field that combined the study of what in a human context is sometimes loosely described as thinking and in engineering is known as control and communication.”⁴⁰ Huggins specifically refers to using the words of “communications experts” when describing feedback. Without examining every publication prior to the Huggins article, it is impossible to know if this was indeed the first use of the term feedback for endocrine systems. The scholarly combination of concepts across disciplines is notable and intentional, and the word feedback increasingly appears in endocrine publications after this date^41–46.

Feedback and the female reproductive cycle-tonic and surge LH release

Coevolving with the adoption of the term feedback was information on the pattern of LH during the female cycle. Early bioassays⁴⁷ revealed LH concentrations were fairly steady in urine and blood samples obtained as often as daily, except for a sharp peak at mid cycle; this peak was preceded by increased estrogens and followed by ovulation and increased progesterone^32,48–53. Involvement of a neural component in this process was supported by the discovery of a 24h rhythm in the LH peak in rodents^34,54. Barraclough and Gorski⁵⁵ used “tonic” to describe the steady, non-peak LH release that caused estrogen synthesis but not ovulation and “surge” to refer to the ovulation-inducing peak. These authors presciently hypothesized that these release modes were controlled by the medial basal and more rostral hypothalamus, respectively, a concept supported by the work of Hálasz⁵⁶

In this tonic-surge understanding of the cycle, the terms negative and positive feedback work fairly well. A feedback loop is formed in a control system when an output is fed into its input, creating a closed loop. In negative feedback, the output has an inhibitory effect on the system, promoting stability of the output within a variable range; this is common in physiology due to the need for stability of the internal environment. While negative feedback is often equated with homeostasis, the term “homeostatic” applied to estradiol negative feedback has likely been misused, including by the authors⁵⁷. Homeostasis strictly requires a regulated variable that is sensed (e.g., blood pressure by baroreceptors) as well as non-regulated variables that are controlled by effectors (e.g., heart rate), but are themselves not directly sensed⁵⁸. An essential element of homeostasis is minimizing error between the set point of a sensed variable and its current value. In the HPG axis, different concentrations of hormones are naturally produced and have varied effects (e.g., tonic vs surge LH pattern), but specific set points have not been described nor have formal sensors that measure hormones vs a desired set point; rather there are receptive cells that change function in response to hormones⁵⁸. In a tonic-surge cycle, negative feedback maintains gonadotropins to promote folliculogenesis and steroidogenesis. In positive feedback, the output has a stimulatory effect on the system, leading to increases in output towards the limit of the system or until something disrupts the positive feedback loop. Physiologic positive feedback systems are less common but are found in contexts that demand deviation from a steady state. In the tonic-surge cycle, positive feedback arises following exposure to sustained high concentrations of estradiol from the dominant follicle (s), resulting in both an increase in output and a marked shift in pattern of gonadotropin release to a surge mode^59–67.

Discoveries in the HPG axis complicate things

As understanding of the hormonal changes of the HPG axis became more detailed, inconsistencies appeared in this straightforward view of estradiol negative and positive feedback. A major shift in understanding was the discovery that tonic LH concentrations were not tonic at all, rather circulating LH concentrations varied in an episodic, or pulsatile, fashion⁶⁸. The hypothalamic factor controlling pituitary gonadotropins, gonadotropin-releasing hormone (GnRH), was sequenced⁶⁹ and LH pulses were shown to be driven by GnRH pulses in pituitary portal blood^70,71. The pituitary was found to need episodic GnRH input to maintain ‘tonic’ gonadotropin concentrations⁷², and circulating gonadotropin concentrations were GnRH-pulse frequency dependent, with higher frequencies favoring LH and lower frequencies favoring FSH⁷³. Natural variations in LH pulse frequency, the best available assay for GnRH release in most mammals⁷⁴, were observed during the primate menstrual cycle^75–80. Specifically, lower LH-pulse frequencies were observed during the early follicular phase when FSH is relatively higher, and higher LH-pulse frequencies during the later follicular phase when the FSH:LH ratio had decreased. Together with the FSH-specific suppression by inhibin⁸¹, changes in GnRH-pulse frequency helped explain how one hypothalamic releasing factor could induce the differential release of LH and FSH needed for proper folliculogenesis. Beyond pulses, estrogens were found to have a similar biphasic effect upon GnRH release mode as they had on LH^82,83. The estrogen receptor needed for these responses, ERα⁸⁴, was not, however, typically detected in GnRH neurons^85,86 indicating a need for estrogen-sensitive GnRH-neuron afferents⁸⁷. The rostral and medial basal hypothalamic regions postulated by Barraclough and Gorski in the early 1960s⁵⁵ do express ERα^86,88,89, and in 2003, kisspeptin-producing, ERα-expressing afferents of GnRH neurons were identified in these two regions in rodents, specifically within the anteroventral periventricular (AVPV) area rostrally and arcuate within the medial basal hypothalamus^90–93. From this rash of discoveries, the complications of the episodic mode of release and multiple sites for estradiol action to control GnRH release are considered below.

Pulses

The discovery of pulses complicated the simple idea that estradiol negative feedback controls LH through the follicular phase. Variables including frequency, amplitude, nadir, total and mean LH concentrations have been reported. The term negative feedback came from the clear reduction in mean/total LH concentrations in response to estrogens^94,95. But in experiments conducted during the follicular phase or with physiologic estradiol concentrations and with a sufficiently high blood sampling frequency, divergent effects were observed on LH-pulse amplitude vs frequency. As mentioned above, LH-pulse frequency increases during the primate follicular phase, in particular when early vs late follicular phases are compared as the peak frequency appears to be achieved within the first week and frequency then plateaus. In sheep, estradiol reduced LH-pulse amplitude⁹⁶, but increased LH-pulse frequency after removal of progesterone⁹⁷ or when compared to ovariectomized animals⁹⁸, indicating a clear stimulation of frequency. Direct measurements of GnRH release in sheep tell a similar story. Increased GnRH-pulse frequency was observed concomitant with increased estradiol concentrations from the early to late follicular phase⁶⁴. Evans, Karsch and co-investigators undertook a series of studies to quantify the effects of estradiol upon GnRH release in the sheep model. Comparing no, basal and peak estradiol during an artificial follicular phase, total GnRH in portal blood was markedly decreased by peak estradiol compared to no estradiol; this was due to reduced GnRH-pulse amplitude but was accompanied by increased GnRH-pulse frequency⁹⁹. Frequent sampling (1-min intervals) of pituitary portal blood during both artificial and natural follicular phases revealed a progressive change from a strictly episodic GnRH release pattern, to pulses imposed on a rising baseline, to continuously elevated GnRH concentrations in which clear pulses were not evident¹⁰⁰; this latter observation confirmed studies in which samples taken as frequently as every 30 seconds during the estradiol-induced surge failed to reveal a pulsatile pattern¹⁰¹. Finally, a careful comparison of individual GnRH-pulse characteristics among ewes with no, basal or peak estradiol treatment confirmed estradiol suppressed total GnRH and GnRH-pulse amplitude and increased GnRH-pulse frequency, and further revealed a progressive decrease in pulse duration and increase in basal GnRH concentrations between pulses with increasing estradiol concentrations¹⁰².

GnRH is the hormonal output from the brain to the pituitary and in most cases LH pulses are a good surrogate⁷⁴. There are also biophysical measurements that have good correspondence with LH pulses and are thought to reflect the operation of the neural component of a GnRH-pulse generator. These have primarily been made in the medial basal hypothalamus and provide an opportunity for long-term monitoring. Volleys of multiunit neural activity (MUA) in the infundibular region of monkeys were first associated with LH pulses in 1984¹⁰³. LH-associated MUA volleys increase in frequency during the early follicular phase in both monkeys and goats^104,105. The increase in goats and most of the increase in monkeys precedes increases in circulating estradiol suggesting removal of progesterone inhibition is the primary cause. Of interest, however, subsequent increases in estradiol do not suppress the frequency of MUA volleys or LH pulses. Also of note, clear resolution of LH pulses can become more difficult when underlying GnRH pulse frequency increases, leading to a potential underestimation of LH pulse frequency⁷⁴. In mice, peaks in intracellular calcium produced by GCaMP6 targeted to arcuate (medial basal hypothalamic) kisspeptin neurons were monitored by fiber photometry over the estrous cycle¹⁰⁶. The nadir in frequency of these events is on estrus following the LH surge. Event frequency is greater on all other cycle days with no statistical differences detected between them. There is, however, a trend for increased frequency from metestrus to diestrus to proestrus. Further, examining the raw data in the example shown for proestrus reveals more “baseline activity” reminiscent of that observed for GnRH release in sheep receiving peak estradiol treatment. An interesting question to contemplate regarding these calcium data is if the biology of the HPG axis is more sensitive than statistical tests as a pulse frequency detector.

A persistent question is whether the episodic mode of release is interrupted/replaced by the surge mode or if it persists throughout. As mentioned, clear pulses have not been detected during the GnRH surge^64,100. The GnRH surge lasts hours longer than the LH surge and well beyond the peak in estradiol that induces it⁶⁴, because the LH surge reduces further estradiol synthesis by the follicle¹⁰⁷. In fact, removal of estradiol near the start of the surge does not block it, although the duration is somewhat reduced (to about 20 h from 30 in persistent E-treated ewes)¹⁰⁸. Similarly, blocking neural activity with general anesthesia is only effective in blocking the LH surge that day if given at the right time before surge onset, after which the surge proceeds and later treatments are ineffective⁵⁴. In this ‘all or none’ nature, the GnRH surge in some respects resembles an action potential: it is induced after a threshold concentration of estradiol is reached and then runs its course well beyond the loss of the induction signal. Unfortunately, the long duration of the GnRH surge and the experimental focus on its onset and generation has resulted is a near absence of data as to what happens to the pattern of release after the surge. There is a single observation in a synchronized ovine follicular phase in which sampling continued through the surge (ewe 1S in⁶⁴); GnRH concentrations in portal blood appear to be at the level of detection for 3–4 hours before small pulses are again observed. In monkeys, hypothalamic MUA is not evident during the estradiol-induced or preovulatory LH surge, suggesting either a cessation of episodic activity or, alternatively, reduced coordination among cells that would preclude detection of bouts of multiunit activity^104,109. MUA frequency declines but does not disappear during the preovulatory LH surge in goats¹⁰⁵. In apparent contrast, the frequency of intracellular calcium events in murine arcuate kisspeptin neurons is maintained for hours into the proestrous LH surge¹⁰⁶. Individual bouts of GnRH release into the median eminence were resolved using electrochemical methods in mouse brain slices during the early estradiol-induced LH surge¹¹⁰; these bouts were higher frequency than during the presurge period but whether this shift reflects the in vivo situation remains to be tested.

So, is estradiol feedback on pulsatile release negative (mean, amplitude) or positive (frequency, increased baseline)? To answer this question, we need to know what effect the LH pattern has at the ovary. Unfortunately, the effects of LH-pulse parameters on follicular maturation have not been determined. Interpretation of work that has been done is tricky because of the two-cell, two-gonadotropin mode of ovarian steroidogenesis¹¹¹, the presence/absence of FSH, which induces expression of aromatase in granulosa cells¹¹², and the use of estradiol, which is dependent upon all of these variables, as an output measure. Nonetheless, in in vitro studies, pulsatile LH administration was moderately more effective at inducing estradiol synthesis and reducing follicle atresia than continuous administration of the same mass of LH^113–115. Reducing LH pulse amplitude with a GnRH receptor antagonist increased the rate of follicular atresia, but this treatment also reduced total LH thus a specific effect of LH-pulse amplitude is difficult to identify¹¹⁶. LH increases steroidogenesis through induction of both steroidogenic acute regulatory protein and cholesterol side-chain cleavage enzyme (CYP11A), which carry out the rate-limiting transport and enzymatic steps of steroidogenesis, respectively¹¹⁷. The effect of LH-pulse parameters on expression of these genes does not appear to have been studied. Two indirect examples provide support for a role of LH-pulse frequency. First, when women with hypothalamic amenorrhea were treated with the same mass of GnRH per day; more robust LH surges were induced in women receiving GnRH at one pulse per hour than one pulse per two hours even though the amplitude in the former was half and GnRH pulse characteristics were likely reflected in LH pulse patterns produced¹¹⁸. Second, an indirect relationship between LH-pulse frequency and steroidogenesis can be inferred from studies in which a stress-mimicking concentrations of cortisol reduced LH-pulse frequency and delayed the estradiol rise in follicular phase ewes¹¹⁹. This suggests lower LH-pulse frequency may be less effective at inducing steroidogenesis, however an inhibitory effect of glucocorticoids on estradiol production by granulosa cells, independent of LH, has also been reported¹²⁰.

If one considers GnRH pulse frequency the primary signal from the hypothalamus, an argument can be made that the substantive effects of estradiol on the hypothalamus to control pulsatile GnRH and LH release are positive or neutral, not negative. LH effects on steroidogenesis are primarily mediated by PKA-dependent pathways, which are induced by relatively low LH concentrations¹²¹ thus higher-frequency low-amplitude LH pulses may effectively produce a sustained increase in estradiol production. In this model, rather than estradiol action switching from negative to positive feedback at surge onset, a second mode of estradiol action is initiated. This mode requires higher concentrations of estradiol and triggers the GnRH and subsequent LH surges in what has classically been referred to as positive feedback. The signal for the second mode is terminated because the LH surge inhibits, rather than stimulates, further production of estradiol because surge LH concentrations engage additional signaling pathways at the ovary and begin the process of luteinization (Figure 1)^107,121.

Figure 1.

Evolution of estradiol action in the HPG axis of females. Left, early HPG axis at the time of Moore and Price and Hohlweg. Middle, refinements to the axis, including the hypothalamus as the main central neural player and the postulated regions of control for tonic and surge gonadotropin release by Barraclough and Gorski. Right, contemporary view with two interacting loops. Loop 1 produces episodic GnRH/LH output and it shows stimulation of pulse frequency by estradiol. Loop 1 cycles for hours, days or week depending on the species to accumulate sufficient estradiol to activate loop 2 to induce the surge. The high concentration of LH during the surge initiate multiple processes at the ovary to trigger ovulation but also terminate the activating estradiol signal for loop 2 by beginning the process of luteinization of the follicle cells. Progesterone from the corpus luteum, not pictured, provides the primary inhibitory input for pulse frequency.

Hypothalamic sites of estradiol action

The dual control of pulse and surge modes of LH release postulated by Barraclough⁵⁵ was advanced by the discovery of two kisspeptin populations in the arcuate and anteroventral periventricular (AVPV) regions. Their roles in the HPG axis have been extensively studied and recently reviewed^{57,93,122–126}. The concept that arcuate kisspeptin neurons mediate negative feedback, whereas AVPV kisspeptin neurons mediate positive feedback arose from and is supported by work in rodents^{124,127–129}. Other species exhibit this dichotomy to different extents but more recent work suggests positive regulation of the more rostral (e.g., AVPV) population by estradiol may be a more common feature among species including humans than previously recognized¹³⁰.

Two points are of note for the consideration of the concepts of estradiol negative and positive feedback as popularly used. First, these hypothalamic neural populations are differentially regulated by estradiol in terms of kisspeptin mRNA expression¹²⁹, and firing activity^128,131,132, with estradiol suppressing or having no effect on these parameters in the arcuate and increasing them in the AVPV. This point calls into question the blanket use of the adjectives negative or positive because both negative (inhibition in the arcuate) and positive (activation in the AVPV) actions can occur simultaneously in the same subject in response to the same estradiol signal. Second, this anatomical division of pulse and surge control of GnRH release violates the requirement that feedback is directed at the source of the original output⁵⁸. Using the arcuate as an example, episodic activity of these neurons controls GnRH pulses and thereby drives the estradiol rise. By strict definition, negative feedback would involve high estradiol regulating arcuate activity to restore baseline estradiol concentration. During the cycle, however, the return towards baseline estradiol concentration is instead brought about by the LH surge, which is initiated by estradiol action on the AVPV to change the mode of GnRH release. Estradiol action thus incorporates feed-forward elements that may be thought of as second ‘feedback’ loops (Figure 1), and the dichotomous framing of the arcuate and AVPV populations as negative vs positive estradiol feedback arms may be oversimplified.

Concluding thoughts

The actions of estradiol within the hypothalamus to regulate reproductive neuroendocrine output are many. These actions, where they occur, the neurobiological processes affected and the direction of change can all vary throughout the female reproductive cycle. In considering the actions of estradiol during the follicular phase as primarily positive, we are giving primacy to the effects of estradiol to increase pulse frequency and induce the surge. This ignores the marked reduction of GnRH- and LH-pulse amplitude by estradiol. While it is intuitive that pulse amplitude is an inverse function of pulse frequency, several points argue that amplitude can be independently regulated. First, electrical stimulation of GnRH neurons in brain slices from mice elicits a greater GnRH secretory response when the slices are from castrated vs intact mice. This suggests increased excitation-secretion efficacy may underlie higher GnRH pulse amplitude observed in castrates in vivo¹³³. Second, estradiol can both reduce and enhance the pituitary response to GnRH^96,134. Third, LH pulse frequency and amplitude are simultaneously elevated in women with polycystic ovary syndrome (PCOS)^135,136. It is possible that the reduced LH-pulse amplitude through the typical follicular phase plays important physiologic roles in selection of the dominant follicle¹¹⁶ and/or guarding against premature ovulation by selectively activating only certain signaling pathways¹²¹. It must be pointed out that while several studies show a lack of estradiol effects to inhibit the frequency of various parameters associated with GnRH output, the best evidence for the ability of estradiol to increase GnRH and LH pulse frequency comes from a single species, the sheep. This could be a specialization of this species. Or it could reflect the ability to measure directly GnRH in a reliable manner, which led to this species dominating research in the area¹³⁷. Regardless, the consideration of the primary effects of estradiol as positive emphasizes the critical nature of progesterone negative feedback to reduce GnRH pulse frequency^64,96 allowing for the early follicular phase FSH rise needed for folliculogenesis. Loss of efficacy of progesterone negative feedback is seen in women with PCOS and the resulting persistent high frequency of GnRH/LH release contributes to disruption of cycles¹³⁸.

It might seem revolutionary to propose we reconsider the terms negative, positive and feedback as used to describe hormonal interactions in the HPG axis. The goal of this review was to stimulate thought about the operation of this axis, not necessarily incite a revolution. We are hopeful that such thought will inspire experiments that address gaps in our understanding of the axis, or that test more firmly established principles in new arenas. For example, direct measurement of GnRH in multiple species would test gonadal-hypothalamic feedback concepts that were established primarily in sheep. These experiments could also reveal if direct measurement of GnRH is indeed clarifying, or if surrogate LH measurements are mostly sufficient for monitoring GnRH release. The applicability of negative/positive feedback terms depends partly on the effects of LH pulse parameters on ovarian estradiol synthesis, and there are few data available on this topic. In vivo studies of the effects of LH frequency and amplitude on ovarian function could either add support to the postulate that frequency is the most critical variable, or force our thinking in new directions. A factor that remains a mystery is why, in common laboratory and domestic species, there is a multi-hour delay from the achievement of peak estradiol levels in the circulation to the switch in estradiol action to induce a surge^67,139–147. Understanding the mechanisms underlying this delay might help explain how the two loops proposed in Figure 1 are connected.

There is a stark difference between the pulse and surge modes of GnRH secretion, and control of these patterns has been mapped primarily in rodents to different hypothalamic kisspeptin neuron populations. The separation of the arcuate and AVPV into purely negative/episodic and positive feedback/surge systems may, however, be premature as it deemphasizes the feed-forward hormonal interplay between these systems as well as possible interactions between these and other central systems.

Adoption of the engineering term of ‘feedback’ to describe hormone interactions in the reproductive and other endocrine systems in the middle of the last century was an appropriate advance in thinking. In light of a better understanding of the elements of the HPG axis, however, a more scrupulous way to refer to estradiol actions in this system may be to revert to the earlier terminology of stimulation and inhibition as appropriate for each individual mechanism being discussed, to better reflect the complex nature of the biology. This could lead to new interpretations of data as “If you change the way you look at things, the things you look at change.” (Wayne Dyer).