9 Kisspeptin
Was this the face that launch’d a thousand ships,
And burnt the topless towers of Ilium?
Sweet Helen, make me immortal with a kiss.
—Christopher Marlowe (1564–1593), “The Face That Launch’d a Thousand Ships”
By the end of the 1990s, reproductive neuroendocrinology was confronted with seemingly intractable puzzles. The most obvious was that, while the ovarian hormones estrogen and progesterone were immensely important for GnRH secretion, we were unsure about how or where they acted.
While estrogen depresses pulsatile LH secretion, high levels of estrogen are an essential predeterminant of the preovulatory LH surge.1 Pulsatile secretion seems to be regulated by the arcuate nucleus, where the axons of GnRH neurons enter the median eminence, but surges seem to be controlled in the rostral hypothalamus where most of the cell bodies of the GnRH neurons are—at least, most of those that control the pituitary.
There are three known estrogen receptors: estrogen receptor alpha (ERα), produced by the gene ESR1; ERβ, produced by the gene ESR2, and a G protein-coupled estrogen receptor. ERα and ERβ are nuclear receptors: their “classic” actions are not directly on neuronal excitability as those of membrane receptors are, but on gene expression. Estrogen is lipid soluble and so it can freely enter cells. If either ERα or ERβ is expressed by a cell then estrogen will bind to it and the bound complex is translocated to the cell nucleus where it interacts with estrogen response elements in the DNA, altering gene expression. Some GnRH cells express ERβ,2 but this receptor is not needed for the effects of estrogen on LH secretion.3 Estrogen inhibits GnRH pulses via ERα, but GnRH neurons do not express it.4
We thus needed to find a population of ERα-expressing neurons to explain the GnRH surge; perhaps they were in the rostral hypothalamus, where many estrogen-receptive cells are found close to the GnRH neurons. If those cells initiate a surge of spiking in the GnRH neurons that leads to the LH surge, then the multiunit recordings of Knobil and others probably did not come from the axons of GnRH neurons, because no multiunit activity is seen in conjunction with the surge.
We also needed to find a population of ERα-expressing neurons to explain how estrogen inhibits GnRH pulses. If Knobil’s LH pulse generator reflects neurons in the arcuate nucleus that regulate GnRH neurons, then perhaps those express ERα. Is it plausible that neurons in the arcuate nucleus generate bursts of spikes that are propagated along axons that end on the cell bodies or dendrites of the GnRH neurons far away in the rostral hypothalamus, which then send axons back through the arcuate nucleus to end in the median eminence? This seemed perverse. It seemed more likely that pulse-generator neurons in the arcuate nucleus would interact directly with the GnRH axons in the arcuate nucleus. But all that synaptic inputs to axons were known to do was to inhibit spike activity—classically, spikes are initiated close to the cell body.
A breakthrough came with the discovery of kisspeptin. Kisspeptin is a peptide encoded by the Kiss1 gene, identified by scientists in Hershey, Pennsylvania, as a gene that could suppress melanoma and breast cancer metastasis.5 They named it after Hershey’s Kisses, the popular American chocolate-teardrop candy—sadly, not for any link with reproductive behavior.
In 2003 it was reported that mutations of the kisspeptin receptor were associated with hypogonadotropic hypogonadism in humans—in this condition, impaired secretion of LH and FSH leads to dysfunction of the testes and ovaries, and puberty is delayed or absent.6,7 Kisspeptin does not affect the secretion of LH and FSH directly, but stimulates GnRH release. It is expressed by two main populations of neurons, one in the arcuate nucleus and a more rostral population. Both express the critical estrogen receptor ERα.
The kisspeptin neurons of the rostral hypothalamus are particularly prominent in females. This is because, in these neurons, Kiss1 expression is upregulated by estrogen. These neurons are essential for the preovulatory LH surge, which can be blocked by infusion of an antibody to kisspeptin into the rostral hypothalamus. By contrast, in the arcuate nucleus, Kiss1 expression is downregulated by estrogen, and these neurons are critical for LH pulses but not for the LH surge. After the menopause, when the ovaries produce no more steroid hormones, pulsatile secretion of LH in women is exaggerated.
Most of the kisspeptin neurons in the arcuate nucleus express two other peptides, neurokinin B and dynorphin, and hence they are sometimes called KNDy neurons (kisspeptin neurons in the rostral hypothalamus contain neither neurokinin B nor dynorphin). In humans, mutations of either TAC3, which encodes neurokinin B, or TAC3R, which encodes its receptor, are associated with gonadotropin deficiency and failure to reach puberty.8 Transgenic mice that lack either dynorphin or the kappa opioid receptors through which dynorphin acts also have deficiencies in LH secretion. Thus neurokinin B and dynorphin are also important for GnRH secretion.
The KNDy neurons talk to each other. They express receptors for neurokinin B and dynorphin, they are surrounded by axons that contain all three peptides, and axons containing neurokinin B connect one arcuate nucleus with the arcuate nucleus of the other side of the brain. Whether kisspeptin itself regulates KNDy neurons is less clear, but it certainly acts on GnRH neurons. In the monkey, kisspeptin axons are densely intermingled with GnRH axons in the median eminence,9 and pulses of kisspeptin release occur in conjunction with pulses of GnRH.10
Transgenic mice that lack kisspeptin receptors are infertile, and restoring kisspeptin receptors in just the GnRH neurons completely rescues their fertility. In Otago, Allan Herbison set about testing whether KNDy neurons could produce pulsatile LH secretion. For this he used optogenetics, which involves making neurons sensitive to light by introducing a gene that encodes a light-sensitive receptor, such as channelrhodopsin2.
Channelrhodopsin2 is normally expressed in green algae; it is an ion channel that guides the movement of the cells in response to light—the algae rise to the surface of a pond by day and sink at night. In 2005, Karl Deisseroth’s lab showed that modified lentiviruses could be used to introduce channelrhodopsin2 into neurons, whose spike activity could then be controlled by exposing them to light of the appropriate wavelength.11 Lentiviruses are retroviruses, which integrate some of their RNA into the DNA of cells that they infect. Now, many transgenic mouse lines are available where many kinds of light-sensitive channels have been introduced into specific neuronal populations, and virus-based approaches have made it possible to easily introduce these channels into specific cell populations in rats.
Allan Herbison and his colleagues introduced channelrhodopsin into the KNDy neurons and placed an optic fiber into the hypothalamus, enabling them to activate the KNDy neurons with flashes of light.12 When they activated the KNDy neurons in male mice or in ovariectomized female mice, making them fire at 10 spikes per second for two minutes was enough to generate an LH pulse. In diestrous female mice, a higher frequency (20 spikes per second) was needed; because there is less kisspeptin in the KNDy neurons at this stage of the ovarian cycle, the neurons must be stimulated harder to achieve the same effect.
However, puzzles remained. Microinjecting kisspeptin either into the rostral hypothalamus or into the arcuate nucleus could stimulate LH secretion, but whereas injecting a kisspeptin antagonist into the arcuate nucleus suppressed LH pulses, rostral injections had no effect on them.13 It thus seemed that, although kisspeptin was an essential driver of the GnRH release that drives LH pulses, it was not acting at the cell bodies of GnRH neurons, but close to the axon terminals. How could an action at that site, far away from the presumed origin of spikes in GnRH neurons, trigger pulses that were presumed to be the result of burst of spikes in GnRH cells?
What had been thought to be the axons of GnRH neurons do not seem to be axons as conventionally understood, but long dendrites from which a cluster of short axons emerges at the level of the median eminence. Herbison called these dendrons to suggest a hybrid of dendritic and axonal properties: the dendron does not merely transmit spikes generated at or close to the cell body—it receives signals along its whole length. Iremonger and Herbison proposed that the GnRH neuron is a hybrid of endocrine cell and neuron and that, at the terminals, kisspeptin controls GnRH secretion not by triggering spikes, but by modulating calcium entry into the distal end of the dendron.14
However, to get enough calcium entry to trigger secretion generally requires calcium channels that are opened only at the voltages reached by spike activity. So might spikes be generated in a dendron? Spikes can be initiated anywhere along an axon, given enough depolarization: all that makes a spike-generating site unique is a high density of voltage-gated sodium channels. In many neurons, dendrites have voltage-gated sodium channels and can propagate spikes; in these neurons, spikes are normally initiated close to the cell body and are propagated along the dendrites as well as along the axons. However, in some neurons, spikes can also be initiated in the dendrites; so perhaps inputs from kisspeptin neurons might trigger bursts of spikes at the end of the dendron. Perhaps the GnRH neurons have two sites of spike initiation: one at the end of the dendron where inputs from KNDy neurons generate the bursts that drive pulsatile secretion, and another, close to the cell bodies, where neurons of the rostral hypothalamus generate a surge of spike activity that produces the preovulatory LH surge15 (figure 9.1).

Kisspeptin and the GnRH neuron. The insets illustrating pulsatile secretion of GnRH and LH on the left and surge secretion on the right are adapted from published work of Sue Moenter, Alain Caraty, Alain Locatelli, and Fred Karsch, showing that “GnRH secretion leading up to ovulation in the ewe is dynamic, beginning with slow pulses during the luteal phase, progressing to higher frequency pulses during the follicular phase and invariably culminating in a robust surge of GnRH.”16
The discovery of kisspeptin has not answered all our questions, but has posed many of them in a new way. It seems likely that bursts are generated by arcuate kisspeptin neurons that signal to the dendrons of GnRH neurons. To understand the generation of those bursts we must begin afresh by studying the arcuate KNDy neurons, which seem to have intrinsic properties compatible with burst firing.17 Bursts can be generated in many different ways, and the kind of properties that we expect of burst-generating neurons are the subject of the next chapter. The answer to the GnRH surge may lie in the rostral kisspeptin neurons, but it has yet to be found: as of now we know little of their properties. The significance of the calcium oscillations in GnRH neurons is still a mystery.
This is what science is like—every discovery raises new questions, often forcing us to reconsider things we have taken for granted. This summary has scarcely begun to encompass the disputes, conundrums, and paradoxes of GnRH research. We respond to new discoveries by trying to incorporate them into our existing body of knowledge. We try to explain as much as possible: we don’t abandon old evidence, though we might look at it a new way. Hypothesis testing as envisaged by Karl Popper is part of this process. Generally, hypotheses are not grand and universal bold hypotheses: they are answers we offer for puzzles posed by observations that we can’t easily dismiss.
I began by emphasizing how few GnRH neurons there are. Without these few GnRH neurons, we cannot reproduce, and our own lines must end; it is a cause for pause to consider how much rests on so few. Yet, there seem to be many more than are strictly needed. Normal mice have about 600 GnRH neurons in the hypothalamus. Their migration is impaired in the mutant mouse strain GNR23, and in homozygous mice only about 70 reach the hypothalamus. Males of this strain have smaller testes than normal, but are fully fertile; females enter puberty normally but are generally infertile. Mice with just one copy of the mutant gene have about 200 GnRH neurons in the hypothalamus, and these females are fully fertile with normal ovarian cycles.
Thus very few GnRH neurons seem to be needed.18 But we cannot be sure. Aging is accompanied by a decline in the production of GnRH. In 2013, a paper in Nature suggested that the decline of GnRH contributes directly to the process of aging and that life span itself can be extended by maintaining the population of GnRH neurons.19 It would be unwise to put too much store on one paper; the scientific literature is full of false leads and dashed hopes, but it is an enticing thought that one day we might indeed be made immortal, and perhaps even with a kiss.
Notes