19 The Sweet Hormone
Yes, oxytocin is in the list of peptides in the brain that affect appetite.1 I’m not going to argue that it is the most important. I don’t know how to distribute importance among the many neuronal contributors to energy balance and don’t imagine that it would be useful or illuminating. I’m giving this chapter to oxytocin, mainly by way of introducing how it is that the same neurons can do several very different things.
In rats, central injections of oxytocin not only affect social and sexual behavior, they also potently inhibit appetite, as do central injections of mesotocin in chicks. As we saw in the previous chapter, the simple fact of an effect on feeding is only a starting point. The questions we ask when we investigate any peptide’s role in energy balance must address how it changes our motivation to eat: whether it affects energy expenditure, pattern of intake, choice of nutrients, intake of fluid; whether it differentially affects females and males; whether it affects taste or the experience of food reward. In the case of oxytocin, the short answer is probably yes, to most if not all of the above.
Some oxytocin neurons in the paraventricular nucleus project to the spinal cord, with effects on thermogenesis, pain processing, and erectile function. Others innervate the dorsal vagal complex in the brainstem, where there is a high density of oxytocin receptors: this region regulates the digestive functions of the stomach and esophagus, and it projects project back to the hypothalamus to stop eating when the stomach is full. Others project to the nucleus accumbens, part of the brain’s “reward circuitry”: there, oxytocin suppresses reward-driven food intake while enhancing the reward associated with social interactions. Transgenic mice that lack oxytocin or its receptor eat about the same amount as wild-type mice, but they show a greater preference for sweet-tasting solutions than wild-type mice, and they are more likely to become overweight.
These different actions of oxytocin are not all hard to explain; some are mediated by different subpopulations of oxytocin neurons that might have properties specialized for those different roles, and which project to different brain sites where the oxytocin they release can act relatively locally. Much more interesting is how the magnocellular oxytocin neurons, all of which are involved in milk ejection and parturition, can also have a role in appetite.
For me, this story came to life when Louise Johnstone in my lab set about using c-fos protein as a marker to determine which neurons in the brain are activated by feeding: c-fos is an “immediate early gene” that neurons begin to express when their activity increases above “normal” levels.2 We had expected that in a hungry rat, the neurons that drive feeding would be active and express c-fos, while neurons that mediate satiety would be active only at the end of a meal. Johnstone allowed rats food for just two hours each day, always at the same time. The rats soon learned to expect the arrival of food, and when it arrived they would eat voraciously for about 90 minutes before grooming themselves and going to sleep. Johnstone looked at the brains of rats killed before or at the expected time of food arrival, and at intervals after food had arrived. We had expected that the NPY neurons would already be active at the expected time of food arrival, stimulated by ghrelin from the empty stomach, and that the lateral hypothalamus, a “hunger center” that contains other appetite-stimulating peptides like orexin and MCH, would be active too. Equally, we had expected that the α-MSH neurons and brain areas involved in satiety would become active toward the end of the meal.
This is not what Johnstone saw. She saw little activation of any neurons before or at the expected time of food arrival, but strong activation in many areas after 30 minutes of feeding. Because there is a delay of about 30 minutes before any stimulus that stimulates c-fos expression can result in the appearance of c-fos protein, these neurons must have been activated as soon as food appeared. The levels of c-fos protein continued to rise, peaking in most areas at about 90 minutes before beginning to slowly decline after the rats stopped eating. It seemed that the neurons driving feeding were activated by the arrival of food, perhaps by its smell and taste. This was perhaps not so surprising: after all, we begin banquets with appetizers to pique our appetite.
But there was no difference in the timing of appearance of c-fos protein between neurons involved in stimulating feeding and those involved in satiety. The populations were activated a simultaneously. The satiety pathways were also engaged as soon as the rats started eating. Perhaps this too should not have surprised us—the signals from the gut that inhibit feeding begin to be released as soon as food arrives there.
Was it really the arrival of food that triggered this response in the brain? Johnstone took other rats and killed them an hour after the time of expected arrival of food but without giving them any food. These showed the same activation of c-fos in areas of the hypothalamus involved in hunger—the neurons were activated not by the taste or smell of food but by the expectation of food arriving. The lateral hypothalamus was even more strongly activated in rats that were not fed. Just as we learn to become hungry as dinnertime approaches, so these rats were intensely hungry when an expected meal did not appear. However, none of the brain regions involved in satiety were activated: the nucleus of the solitary tract was quiet, as were the α-MSH cells.
Of all the brain areas that Johnstone looked at, the strongest activation was in the supraoptic nucleus, where feeding evoked strong expression of c-fos. This activation, like that in the nucleus of the solitary tract and in α-MSH neurons was present only in fed rats, never in rats that were expecting but did not receive food.
The results were, for us, the strongest indication yet that magnocellular oxytocin neurons are involved in appetite. There was, however, already circumstantial evidence that magnocellular neurons might be involved in appetite. These neurons receive an innervation from the nucleus of the solitary tract, and in rats, gastric distension and intravenous injections of cholecystokinin strongly activate them. Cholecystokinin is a hormone secreted from the duodenum in response to food ingestion. It acts on the sensory endings of afferent neurons of the gastric vagus nerve, and this effect is relayed by the nucleus of the solitary tract, by neurons that produce noradrenaline and others that expresses prolactin-releasing peptide (as is often the case with the names of neuropeptides, this name is inappropriate). However, both cholecystokinin and gastric distension were thought to be stressful, and as it was known that other types of stressors would activate magnocellular oxytocin neurons, these findings did not unequivocally suggest a specific role in appetite regulation.
Then, two years after Johnstone published her paper, Giles Yeo and his colleagues in Cambridge showed that leptin strongly affects the oxytocin system.3 They were searching for genes involved in mediating the effects of leptin, and reasoned that such genes would be downregulated in fasted mice when leptin levels were low, and upregulated again when the fasted mice were treated with leptin. They used laser-capture microdissection to cut out just the paraventricular nucleus from brain sections, isolated the RNA from these samples, amplified the RNA, and used a microarray to measure gene expression. In all, 89 genes met stringent criteria to be considered as leptin-regulated genes. The 25 genes most strongly influenced included three secreted peptides: thyrotropin-releasing hormone, oxytocin, and vasopressin. Although the paraventricular nucleus contains both parvocellular and magnocellular oxytocin neurons, the magnocellular neurons are more abundant and express much more oxytocin. To account for the Cambridge results, it seemed that they must be affected by fasting and leptin.
We went back to the supraoptic nucleus to study oxytocin neurons, and found that fasting reduced their spike activity while systemically administered leptin enhanced it. By then, more evidence had accumulated. The oxytocin neurons were responsive to glucose and insulin. They expressed other appetite-inhibiting factors, including cholecystokinin itself and nesfatin. They were activated by oleoylethanolamide (a hypophagic lipid-amide released by the small intestine in response to fat intake), by oral delivery of sweet solutions, and by delivery of food into the stomach by gavage. It was clear that many different appetite-related signals converge on these neurons.
The α-MSH neurons innervate both the paraventricular nucleus and the supraoptic nucleus, which densely express the MC3 and MC4 receptors through which α-MSH acts. In oxytocin neurons, α-MSH induces mobilization of intracellular calcium stores, expression of c-fos, and dendritic oxytocin release, but it also inhibits spiking activity and therefore inhibits secretion into the blood. This inhibition is the consequence of the evoked production of endocannabinoids, which suppress the release of neurotransmitters from afferent endings that express cannabinoid receptors. Interestingly, in pregnancy α-MSH has no effect on oxytocin neurons, and while central injections of α-MSH increase c-fos expression in the paraventricular and supraoptic in nonpregnant rats, this effect of α-MSH is also suppressed in pregnant rats. Perhaps this contributes to the hyperphagia of pregnancy: while mammals generally maintain a very stable body weight, in the last third of pregnancy they gain fat mass through increasing their food intake.
The magnocellular oxytocin neurons are the source of oxytocin in the plasma, but they also release abundant oxytocin from their dendrites, and this acts at two relatively close sites that express abundant oxytocin receptors: the amygdala, which contains only a few oxytocin-containing fibers, and the ventromedial nucleus, which appears to contain none. The ventromedial nucleus is important for glucose homeostasis, and it also controls sexual behavior, feeding, fear behavior, and aggression. These behaviors are not mutually compatible: given the motivation and opportunity to have sex and to eat, animals generally do one or the other, unless they are afraid, in which case they may fight or flee but are unlikely to eat or mate. Oxytocin enhances sexual behavior while suppressing feeding and fear, so, by its actions at the ventromedial nucleus, it is important for fundamental behavioral decisions.
In rodents, oxytocin secretion induced by food intake promotes sodium excretion and influences gastric motility, but these effects are not seen in all mammals. In humans, oxytocin secretion is stimulated by exercise (and this is associated with altered fluid balance), but not by osmotic stimuli as it is in rodents, nor does oxytocin induce natriuresis. Oxytocin secretion in humans is not stimulated by gastric distension or by systemic administration of cholecystokinin; instead, these activate vasopressin secretion. The same is true in ferrets.4 It seems that there may be a difference between species, like ferrets and humans, that can vomit when they overeat and those, like rodents, that can’t.
Nevertheless, oxytocin and its receptor are present throughout the gastrointestinal tract in humans and other mammals. Oxytocin acts as a brake on intestinal motility, promotes the development and survival of enteric neurons and intestinal crypt cells, regulates the permeability of the mucosal lining of the intestine, and protects against inflammation. Oxytocin receptors are also present in the rat pancreas, where oxytocin can stimulate insulin and glucagon secretion. Adipocytes also have oxytocin receptors, and through these, oxytocin induces lipolysis—the breakdown of lipids into glycerol and free fatty acids. Oxytocin also increases the formation of osteoclasts, cells that break down bone and are essential for repair and remodeling of bone, and in mature osteoclasts it inhibits this bone resorption. Mice lacking oxytocin or its receptor develop osteoporosis, a condition in which bones become fragile, and this worsens with age. In rabbits, glucocorticoid-induced osteoporosis can be prevented by giving oxytocin systemically, and in mice, oxytocin can reverse ovariectomy-induced osteopenia and adiposity.
So now, to come to the point of this chapter—how can one hormone, oxytocin, be involved simultaneously in so many different things?
It’s not hard to conceive that different subpopulations of oxytocin neurons do different things, if they project to very different places that are far apart. Nor is it hard to conceive how oxytocin produced at peripheral sites has specific, very local actions on adjacent cells—those cells might be indifferent to the oxytocin concentrations in the plasma but might receive enough oxytocin from adjacent cells to be activated. Harder to understand is how the magnocellular oxytocin neurons can do several apparently very different things simultaneously without conflict. To spell it out, these neurons, in the rat, regulate milk ejection, parturition, food intake, social and sexual behavior, and sodium excretion.
Even milk ejection isn’t simple. In marsupials, milk ejection is regulated by mesotocin, the marsupial homolog of oxytocin; they rear their young in a pouch—a pouch that can contain two young of very different ages. Dennis Lincoln and Marilyn Renfrew studied milk ejection in the wallaby Macropus agilis5 Marsupials give birth to young that are almost embryonic in form; for the first 100 days or so of life, they remain continuously attached to a nipple. By the time it is 200 days old, the infant will have left the pouch, returning intermittently to suckle, by which time a second newborn infant may be attached to a different nipple. So how can mesotocin secreted from the pituitary provide an appropriate supply of milk to each of these two young, whose patterns of suckling and whose needs are so very different? Lincoln and Renfrew showed that the mammary gland that is suckled by the newborn is exquisitely sensitive to mesotocin; this suckling leads to intermittent pulses of mesotocin, which, though small, will trigger milk ejection at this gland, but not at the other gland, suckled by the older sibling, which is less sensitive to mesotocin. The suckling of that older sibling is a stronger stimulus, and leads to occasional large pulses of mesotocin—these will cause milk letdown at both glands—but the gland suckled by the newborn responds similarly to this large pulse as to a small pulse. They therefore concluded that “the wallaby, far from discarding during its evolution the milk-ejection response to oxytocin, has refined the response to permit it to feed simultaneously both newly born pouch young (<1 g) and young at foot (>2,500 g).”5
The response of a tissue to oxytocin (or in this case mesotocin) is defined by the level of expression of oxytocin receptors in that tissue. Turning to how oxytocin can regulate both uterine contractions and milk ejection, the explanation is the same—the uterus is only fully sensitive to oxytocin at the time of parturition, and the mammary gland only during lactation.
But what about natriuresis? Sodium excretion must be regulated throughout life, including during lactation. Here it is the pattern of secretion that is critical. The rodent kidney is extremely sensitive to oxytocin, and at the kidney, natriuresis is regulated by a graded, continual release of oxytocin governed by the continual background activity of oxytocin neurons.6 Natriuresis is a slow and continuing process of sodium excretion: pulses of oxytocin are neither here nor there, their impact on the kidney is too brief and intermittent to count. The mammary glands, by contrast, are relatively insensitive to oxytocin—the graded background release that regulates natriuresis is invisible to them, all that matters is those large pulses. So by regulating their background activity and generating intermittent pulses, magnocellular oxytocin neurons can fulfill both functions without conflict.
These neurons also regulate both feeding behavior and sexual behavior. There is no conflict between these two—oxytocin inhibits one while enhancing the other, but how is a conflict avoided between these and the other functions of oxytocin? Here we must recall that because dendritic secretion of oxytocin does not require spike activity, it can be regulated independently of oxytocin secretion from the posterior pituitary.
The magnocellular oxytocin neurons are supreme multitaskers, with multiple sensory properties and multiple functional roles—characteristics, as we will see, of the neurons of ancient organisms from which they evolved.
Notes