14 Plasticity
Whether ’tis nobler in the mind to suffer
The slings and arrows of outrageous fortune,
Or to take Arms against a Sea of troubles,
And by opposing end them:
—William Shakespeare, Hamlet
About twenty years ago I was invited to be an instructor at a training school for new researchers in endocrinology. The school was funded by the pharmaceutical company Ferring, and was held on the island of Föhr in the North Sea, at the home of Fredrik Paulsen, the founder of Ferring. The high points of the week, for me, were two after-dinner talks given by Fredrik, one on the history of Ferring, and another on language. Föhr is a small island with just a few thousand inhabitants, but seven languages are spoken there, including one, Fering, which is spoken only on Föhr. Fredrik’s theme was that every language is a unique repository of knowledge and understanding.
Föhr had been a whaling community, and in the early race to develop treatments for the diseases of hormone deficiency, whales played a small but important part. As might be expected, the pituitary of the whale is large and a rich source of hormones very like those of humans. Moreover, the anterior pituitary is separated from the posterior pituitary by a bony plate, making it easy to separate the tissues without cross-contamination, so Fredrik recruited whalers to salvage the pituitaries for research. In Fering, the pituitary is known by an idiom that translates as “the button on the brain.” Whalers had long recognized this conspicuous gland and had noticed that, when a whale had fought particularly long and hard, the pituitary would be engorged with blood. This explains another idiom—“the button on his brain has burst,” applied to someone overcome by stress. Fredrik argued that knowledge of the pituitary’s involvement in stress was embedded in the language, long predating our scientific awareness.
Anyone who attempts to track the historical development of physiological concepts is likely to become frustrated at their imprecision and instability. Terms that appear to define specific biological entities, like genes, neurotransmitters, and hormones, have accrued multiple different and sometimes incompatible meanings that depend on the context in which they are used. Today, in some contexts, a gene is the sequence of DNA that encodes a specific protein; in others, it is the longer unit that spans the exons that encode the protein sequence but also includes the introns that separate the exons and which themselves include sites that regulate the expression of the gene product; in yet others, it is a longer and ill-defined sequence that includes all regulatory sites. Sometimes, such as when we talk of the gene for a certain trait that differs between individuals, we mean a variant version of some site of regulation of a gene. Sometimes we talk loosely about genes as though they were synonymous with specific traits, as in the “genes for hair color.” Sometimes, even more loosely, we talk of genes “for” diseases like cystic fibrosis or Huntington’s disease, as though their purpose is to give us these diseases.
These confusions are as nothing to those entailed by terms that have connotations from their use in normal language. As I write, I have just returned from a meeting with collaborators in a multidisciplinary project to study the determinants of food choice. We spent our last few hours together reflecting on how we use the term hunger, an important reflection if mutual understanding is to be more than an illusion. We variously use “hunger” to refer to: a state where there is not enough available energy to meet current and expected demands; the signals that inform the brain of that state; the effects of those signals on specific brain areas; our conscious awareness of that state; and the motivation to eat regardless of how that motivation arises. We also use diverse “operational definitions” of hunger such as self-reported scales of perceived hunger and measures of the willingness to work to receive a food reward. That these are all different things we must keep in mind whenever we talk. “Hunger” is a ghost in the machine of the brain that assumes different and often intangible forms, and it will not be easily expunged.
“Stress” is even more problematic. Our current use of the term was coined by Hans Selye about eighty years ago. Selye had shown that diverse physical and emotional threats all produced the same three pathological outcomes: enlarged adrenal glands, stomach ulcerations, and lymph tissue shrinkage, and he chose the word stress to describe this triad, defining it as “the non-specific response of the body to any demand for change.” Defining stress in this way forced him to introduce a new term, stressor, for things that produce stress, but keeping the terminology tight and consistent was not easy. Paul Rosch, one of Hans Selye’s collaborators, later reflected: “Even Selye had difficulties, and in helping to prepare the First Annual Report on Stress in 1951, I included the comments of one critic, who, using verbatim citations from Selye’s own writings, concluded that ‘Stress, in addition to being itself, was also the cause of itself, and the result of itself.’”1
Neuroendocrinologists try to avoid such confusions by using operational definitions tied to the secretory activity of the adrenal cortex; thus a stressor is any stimulus that results in increased production of glucocorticoid hormones—mainly cortisol in humans, corticosterone in rodents. These steroid hormones are not stored in the adrenal cortex but must be produced on demand, and their production is regulated mainly by ACTH from the pituitary. Both the production and the secretion of ACTH is regulated by factors released, as we now know, by neurons in the paraventricular nucleus (figure 14.1).
Amplification and feedback in the stress axis. In the rat, CRF (and vasopressin) is produced in about 1,000 neuroendocrine neurons of the paraventricular nucleus. These neurons project to the median eminence, where they release brief pulses of CRF into portal blood vessels that carry the CRF to the anterior pituitary. There, CRF stimulates the corticotrophs to secrete pulses of ACTH, which reaches levels of up to 1 ng/ml in the blood. ACTH acts on about 400,000 cells of the adrenal cortex to stimulate the production of corticosterone, which reaches levels of 80 ng/ml and more in plasma. This acts back on both the pituitary and the brain. The secretion of corticosterone follows a diurnal rhythm—rats are nocturnal, and highest levels occur at about the time they become active. This daily rhythm is governed by light signals, detected by the retina and communicated to the paraventricular nucleus via the suprachiasmatic nucleus. Illustration modified from Spiga F et al.2
The triad of hypothalamus, pituitary, and adrenal cortex is commonly known as the HPA axis. This triad does not capture the full range of physiological responses to stressors: the “fight or flight” responses to acute threats that involve increased heart rate and blood pressure, increased glucose production by the liver, and secretion of adrenaline from the adrenal medulla are regulated by other paraventricular neurons via a pathway that engages the autonomic nervous system. Nor is the HPA axis the only endocrine system engaged: the secretion of many hormones is affected by stressors. However, it is the HPA axis that produces stress as defined by Selye.
Glucocorticoids act on many organ systems. In the liver, fat, and muscle they mobilize energy to meet current or anticipated demand, and they also redirect energy. For example, they suppress inflammation, which is a response to injury that, however useful, might impair the ability to escape from a continuing threat. Glucocorticoids engage long-lasting adaptations, and can do so because they affect the expression of many different genes in different tissues. Glucocorticoids are lipid soluble, and so enter cells freely; inside a cell, if a molecule of glucocorticoid encounters a glucocorticoid receptor the two molecules will bind together, and this complex is ferried into the cell nucleus where it binds to “glucocorticoid response elements” in the DNA. There are such elements in the regulatory regions of many genes.
Glucocorticoids act not only on peripheral organs, but also on the brain and pituitary. They signal through at least two receptor subtypes, the mineralocorticoid receptor and the glucocorticoid receptor: both are expressed in the brain, especially in the hypothalamus but also in the hippocampus, which is involved in learning and memory. This should be no surprise: if an innocuous event is a reliable harbinger of a threat that we can avert by an appropriate behavioral or physiological response, we need to learn this and remember it. A stressor given shortly after learning will generally strengthen memory, especially for stimuli that evoke emotional responses, whereas stressors before learning are distractors. Mineralocorticoid and glucocorticoid receptors are “nuclear” receptors, affecting gene expression, but glucocorticoids can also exert rapid effects on neuronal excitability by effects on membrane receptors.
Glucocorticoids are needed to assure energy availability even in the absence of stress. They are secreted in a daily rhythm, mobilizing energy in anticipation of the normal cycle of energy use. This rhythm is modulated by signals from the periphery that indicate how much energy is available; in rats, sucrose intake will dampen glucocorticoid production, while low levels of free fatty acids, sensed by neurons in the paraventricular nucleus, enhance it. Superimposed on this daily rhythm is a pulsatile pattern of secretion; every hour or so a pulse of ACTH triggers a pulse of glucocorticoid production. Surprisingly, this does not depend on a pulsatile input from the hypothalamus, as revealed in a mathematical model by Stafford Lightman, John Terry, and Jamie Walker.2 Because ACTH triggers the production of glucocorticoids, which takes time, there is a delay between ACTH secretion and the glucocorticoid production that it evokes, and glucocorticoids then act back on the pituitary to suppress the secretion of ACTH. Given a constant hypothalamic drive to ACTH secretion, this cycle will result in a pulsatile pattern of secretion very like that normally observed.
Any novel stimulus is a potential threat, and will evoke a sharp peak of ACTH secretion and glucocorticoid production. We might see this as a signal that puts brain networks “on alert.” The signal goes everywhere, but only recently activated pathways are affected. In the hippocampus, circuits that contain glucocorticoid receptors receive information about diverse sensory stimuli. If a stimulus proves innocuous, then it may have no enduring consequences. However, if it has an injurious sequel, then a large and prolonged rise in glucocorticoid concentration will act on those pathways that have been put on alert to consolidate the memory of the stimulus and encode its linkage with an adverse consequence.
This account just gives the shape of a possible explanation consistent with present knowledge and understanding. If neuroscientists sometimes talk as though they know all the answers, it’s a hubris that is often easily punctured. Our theories are partial and provisional. Paul Rosch, reflecting on Hans Selye’s contributions, wrote, “His real legacy can be summed up by what he often reminded me, namely, that theories do not have to be correct. Only facts do. Some theories are of value because of their heuristic value, in that they encourage others to discover new facts, that then lead to better theories.”1
By looking at the effects of glucocorticoids on the brain, we can see that an indiscriminate hormonal signal, accessing all regions of the brain, can nevertheless have actions that are specific, sophisticated, and enduring. The specificity is only partly because only some neurons have glucocorticoid receptors: the effects of glucocorticoids also depend on context, including the context of recent events, and these effects can have prolonged consequences for the behavior of neurons.
The survival value of associating a stimulus with a threat is obvious, and the association must be learned from a single experience and remembered for a lifetime. In travels to many parts of the world I have eaten and usually enjoyed many different foods. One thwarted me. Lutefisk, long part of Swedish cuisine, is dried cod that has been soaked in a lye solution for several days to rehydrate it before being rinsed with cold water to remove the lye; it is then boiled or baked. My only encounter with it was at a restaurant in Stockholm. Before a spoon touched my mouth I felt my face turn white, broke into a sweat and started to gag. Something had triggered in me the unconscious memory of something encountered, probably in my childhood, of which I have no conscious memory. Memories don’t have to involve any conscious elements or reasoning processes, and it’s often better that they don’t.
If the HPA axis is continually activated, as in many disease conditions that are accompanied by chronic pain or inflammation, then the prolonged production of glucocorticoids can have serious pathological consequences. So, while it is important that we learn that some stimuli are threats, we must also find ways of coping with stressors that we cannot avoid, but which turn out not to be unduly harmful.
Rats, when first shown a dark tube, will readily enter it—they like dark places. But if the tube is then closed so they can’t escape, they become anxious and produce large amounts of corticosterone. If a rat is restrained in this way for an hour each day, the corticosteroid response habituates; each day, it is less than on the previous day. It seems that the rat learns that the experience, though initially unsettling, is not harmful, and you might imagine that this is a reasoned understanding, that the rat learns that nothing untoward ensues and hence does not activate corticosterone production. This is sort of true, but it happens in an interesting way. To understand what happens, we have to understand something of how the paraventricular nucleus regulates ACTH secretion.
After Geoffrey Harris had proposed that secretion from the anterior pituitary was regulated by hypothalamic factors, the first confirmation of their existence came from evidence that extracts of the hypothalamus contained a factor that potently stimulated ACTH secretion. However, this advance was followed by frustration; it seemed impossible to disentangle the effects of any “true” corticotropin-releasing factor (CRF) from those of vasopressin. Vasopressin was always present in tissues that had CRF activity: it could stimulate ACTH secretion, but did not seem potent enough to be the CRF. It did not seem credible that vasopressin could regulate both ACTH secretion and antidiuresis: the levels in plasma seemed too low, and only some stimuli that activated ACTH secretion also activated vasopressin secretion. Studies in the Brattleboro rat, which lacks vasopressin, produced conflicting results: some claimed that the HPA response to stress was normal, but others that it was attenuated. If vasopressin was the CRF, then the Brattleboro rat should have profound deficits in ACTH secretion and corticosterone production—which it does not.
In the late 1970s and early 1980s, several findings emerged in quick succession. It was found that not all of the vasopressin neurons in the paraventricular nucleus projected to the posterior pituitary: some projected to the median eminence, so there was a second vasopressin system, one that might be implicated specifically in ACTH secretion. After removal of the adrenal glands, the vasopressin content of the paraventricular nucleus massively increased, suggesting that its release was regulated by hormonal feedback from the adrenal glands, as would be expected of a hypothalamic system of regulation. So, if vasopressin was not the CRF then surely it had to be a major part of it. There seemed to be three possibilities: that vasopressin was involved in regulating the secretion of a CRF; that a CRF and vasopressin were independently regulated releasing factors for ACTH; and that vasopressin acted as a releasing factor in concert with a CRF.
Phil Lowry and Glenda Gillies had advocated the theory that vasopressin was the main factor regulating ACTH secretion, and when they found that other hypothalamic factors with “weak, labile CRF activity” could potentiate vasopressin-induced ACTH release, they proposed, in a letter to Nature, that “CRF is vasopressin modulated by other hypothalamic factor(s) released into the hypothalamo-hypophyseal portal system.”3
Just two years later, Wylie Vale and his colleagues at the Scripps Institute succeeded in purifying, from extracts of sheep hypothalami, a peptide with extremely high potency for stimulating the secretion of ACTH from culture.4 There was little doubt that they had found the physiological releasing factor, and it was not vasopressin. Soon, the term CRH (corticotropin-releasing hormone) superseded the term CRF as it became generally accepted that the peptide identified by Vale was indeed a hormone that regulated ACTH secretion.
Gillies and Lowry did not let the matter rest. The following year, again in Nature, they published a letter with the title “Corticotropin releasing activity of the new CRF is potentiated several times by vasopressin.”5 They had studied ACTH secretion from isolated pituitary cells; they confirmed that the new CRF was more potent at secreting ACTH than vasopressin, but found a synergistic interaction between this CRF and vasopressin. Synergy is a term that is often used loosely when an effect exerted by two factors in combination is greater than the effect of either alone. Usually such effects are additive, rather than synergistic. True synergy can be seen when two drugs that are both applied at maximally effective concentrations evoke a response that is much greater than the sum of their separate effects. Synergy in this sense cannot be explained by two drugs acting on the same receptors, but implies an interaction between the intracellular signaling pathways that the two drugs engage.
Then, two years later, in 1984, Wylie Vale and colleagues reported that they had found the neurons that make CRH, as expected, in the paraventricular nucleus. Between 1% and 2% of them also contained vasopressin. But when the adrenal glands were removed from rats, about 70% of the CRH neurons contained vasopressin. They concluded that “there is a state-dependent plasticity in the expression of biologically active peptides by individual neuroendocrine neurons.”6
It was well known that the brain is plastic—that experience and physiological state can alter the connectivity of neurons—but the idea that what messengers these neurons use might also change was novel. Here was a neuronal system that used a potent ACTH secretagogue, CRH; now, in addition it seemed to be making another, vasopressin, which alone was less potent than CRH but which in combination with CRH was extremely potent.
In Bristol, Ming Ma and Stafford Lightman began studying how the expression of the genes for vasopressin and CRH in rats changed in response to repeated episodes of restraint stress.7 The first episode evoked a large rise in corticosterone levels, and it stimulated the synthesis of both vasopressin and CRH, as measured by levels of heteronuclear RNA (the immediate copy of the coding region of the DNA before it is processed into mRNA) in the paraventricular nucleus. This is to be expected: whenever a peptide is released, more must be made to replenish the releasable pools. When the stressor was given daily, the corticosterone response declined progressively, and after two weeks it was completely absent. The CRH response declined similarly. So far, the account is as might be expected if the stimulus ceases to become stressful—if the rat learns that it is innocuous.
However, whereas the response of the CRH gene to the stressor decreased with repetition, that of the vasopressin gene to the stressor increased. The neurons in the paraventricular nucleus were still reacting to the stimulus—but differently. If with each episode of a stressor the neurons increase the synthesis of vasopressin but not that of CRH, then progressively their stores of releasable peptide will include more vasopressin and less CRH, and what were CRH-secreting neurons will become vasopressin-secreting neurons.
To see what this might mean we have also to know that although rats exposed to a repeated stressor will habituate to it, they will still react to a new and different stressor—indeed, repeatedly stressed rats show a heightened response to some different stressors.
To make sense of this, we must recognize that different stressors signal to the paraventricular nucleus by different pathways, and do not target exactly the same neurons. So, consider one type of stressor, let us call it stressor A, that activates a subset of CRH neurons—subset A. With repeated exposure to stressor A, these become vasopressin-secreting neurons, and because vasopressin is less potent than CRH, they release less ACTH. A different type of stressor, stressor B, may activate another subset B that are still CRH neurons, so the ACTH response to stressor B is intact although that to stressor A has diminished. The interesting case is stressor C, which activates some neurons from subset A and some from subset B. This will evoke a mixture of vasopressin and CRH release, and so can give a supranormal response.
The phenotypic plasticity of CRH neurons can no longer be regarded as exceptional. Across the hypothalamus, more than a hundred different peptides are expressed in different neuronal subpopulations along with several hundred specific peptide receptors. Many neurons produce several peptides, in a bewildering variety of combinations, and this Kandinsky-like canvas is in constant turmoil, as the expression of each of these genes waxes and wanes with the multifarious rewards and insults of daily life.
One of the most startling changes has been described in the neurons that regulate prolactin secretion. When the pituitary is separated from the hypothalamus, the blood levels of most of its hormones disappear, but prolactin levels increase, because its secretion is governed by a factor that inhibits spontaneous secretion. That factor is dopamine, released from the arcuate nucleus. Dopamine is better known as a neurotransmitter in neuronal circuits within the brain, especially in the “reward circuits” of the brain and in the circuits involved in fine motor control that are affected in Parkinson’s disease; there is no single “dopamine system” but many systems doing very different things. Even in the arcuate nucleus there are at least two populations of dopamine neurons; one regulates prolactin secretion, while another innervates the intermediate pituitary and regulates secretion of α-MSH from melanotrophs into the blood.
Electrophysiological studies of dopamine neurons in the arcuate nucleus indicate that they, rather like vasopressin cells, alternate every 20 seconds or so between an up state during which they fire at about four spikes per second and a quiescent down state. But unlike vasopressin cells they fire in synchrony, releasing dopamine in frequent pulses. Prolactin has a long half-life in blood (about 20 minutes), so the rapidly pulsating dopamine release is clearly not there to drive a slowly pulsing prolactin secretion. However, it is possible to study the activity of the prolactin gene in real time in individual lactotrophs using a “reporter gene” in which the expression of a light-emitting protein (luciferase or green fluorescent protein) is governed by the same DNA sequence that regulates prolactin expression.8,9 When this gene is introduced into a lactotroph, the level of prolactin expression can be inferred from the intensity of light emitted from the cell. These studies have shown that transcription of the prolactin gene is highly pulsatile. In cell cultures, different lactotrophs display very different and asynchronous fluctuations, but within explanted pituitary glands they are less heterogeneous and are bound together by intercellular signals. It is not yet possible to study transcription in real time in lactotrophs in the living animal, but it seems likely that a common rhythmic drive from pulsatile dopamine release coordinates transcription of prolactin among the whole population of lactotrophs.
The functions of prolactin are many and diverse. In fish and amphibians, prolactin is an osmoregulatory hormone, regulating the permeability of the skin to water and salt, and it also regulates reproduction. In some birds, prolactin promotes fluid production in the crop sac to feed fledglings—a function analogous to milk production in mammals. In mammals, prolactin is involved in many aspects of reproduction, and also in metabolic regulation through receptors expressed on adipose tissue, liver, pancreas, and the brain. Prolactin promotes appetite, and this might underlie the weight gain in pregnancy that anticipates the nutritional demands of suckling young. It also causes infertility, by suppressing LH secretion, probably through its actions on GnRH neurons, and is the main reason why women who are exclusively breastfeeding an infant seldom become pregnant.
Prolactin also suppresses libido in both males and females and enhances parental behaviors. When male talapoin monkeys are housed together, they sort themselves into a dominance hierarchy. The most aggressive and sexually dominant males have high levels of testosterone and low levels of prolactin, while the most subservient have high levels of prolactin and low levels of testosterone. These endocrine differences are a consequence of the emergence of a hierarchy, not the cause of it, and they have important consequences. Subordinate individuals withdraw from competition for mates, which might be seen as a way of avoiding injuries that would be incurred by fighting; since dominant males are also those most at risk of injury, we might see subordinate individuals as biding their time.10
However, the best-known role of prolactin is in lactation, when it stimulates the production of milk. Prolactin secretion is stimulated by suckling, and evidence indicates that this must involve an inhibition of dopamine release. Prolactin secretion is regulated in part by a negative-feedback loop—as levels in the blood rise, some is transported into the brain, where it activates the dopamine neurons to inhibit further secretion. This transport mechanism is unchanged in lactation, so if the dopamine neurons are quiescent in lactation despite continued high levels of prolactin, then either something is inhibiting them strongly, or else something is stopping them being activated by prolactin.
Dave Grattan and his colleagues in Otago and Montpellier set about testing these hypotheses, and elegantly refuted both. They found that the electrical properties of the dopamine neurons were unchanged in lactation, and so were their responses to prolactin. However, in one important detail things were different: in lactation, they no longer produced dopamine. Instead, they were making met-enkephalin,11 by which they signal not to the pituitary but to other neurons in the arcuate nucleus: met-enkephalin, among other things, is a potent stimulator of appetite.
To put all this in a broad context, in response to signals from the periphery, including hormonal signals from the adrenal glands and the gonads, some neurons in the hypothalamus don’t merely change their electrical activity, they also change the messengers that they produce and who they talk to. The pattern in which they release their messengers not only affects the electrical and secretory activity of their targets, it can also affect the expression of genes in those targets. We are glimpsing a system that is not like a giant supercomputer executing some vast and sophisticated program, but like an ecology of multiple small computers that are constantly reprogramming themselves or being reprogrammed by external events.
Notes