20 The Hybrid Neuron
In the multiplicity and diversity of the physiological effects produced by these various chemical messengers one is apt to lose sight of the fact that we are here investigating one of the fundamental means for the integration of the functions of the body. These are not merely interesting facts which form a pretty story, but they are pregnant of possibilities for our control of the processes of the body.…
—E. H. Starling, The Harveian Oration, delivered before the Royal College of Physicians of London on St. Luke’s Day, 19231
The classical neuron has dendrites, where information is received from other neurons and converted into electrical signals; a cell body, where these signals are converted into spikes; and axons, which carry the spikes to synaptic endings where they trigger the release of neurotransmitters. Neurons differ in their morphology: some have exuberantly bushy dendrites; some have several axons that have many branches; magnocellular vasopressin and oxytocin cells have two or three long but unbranching dendrites and just one long axon. Typically, neurons transmit information at several thousand synapses, and receive several thousand inputs at their cell bodies and dendrites, but there is great heterogeneity—in the rat, about 12,000 dopamine neurons of the substantia nigra project to 2.8 million spiny neurons in the striatum with axons that are on average 40 cm long and which each give rise to between 100,000 and 250,000 synapses.
Membrane properties vary from cell to cell and between the compartments of a cell, and those properties influence how incoming information is processed and how outgoing information is patterned. Some dendrites have extensive signal-processing abilities and can generate spikes; others are passive conductors of synaptic inputs. Classically, synapses are between axon terminals and dendrites, but some connect axon terminals to axon terminals, or dendrites with dendrites. Nevertheless, the classical image remains dominant, conveying the key understanding that the main point of information exchange is the synapse.
The classical synapse has several distinctive features. The presynaptic ending contains about 5,000 small vesicles, each of which contains a few thousand molecules of neurotransmitter, usually glutamate or GABA, sometimes noradrenaline, serotonin, or dopamine, sometimes acetylcholine or histamine. There is continuous remodeling of the synaptic architecture; some synapses proliferate as others are eliminated, guided in part by activity in the presynaptic cell and in part by signals from the postsynaptic cell. This web of interneuronal communication is generally understood to be the basis of information processing in the brain.
But Kyell Fuxe has long argued that some of these “neurotransmitters”—notably serotonin, noradrenaline, and dopamine, act mainly at extrasynaptic receptors. Many of the neurons that release these neurotransmitters have very diffuse projections in the brain and spinal cord, and he proposed that they use a mode of signaling that he called “volume transmission.” Noradrenaline, serotonin, and dopamine are all messengers that act close to their site of release if not purely at synapses, but because they are released simultaneously from diffusely distributed projections, they go indiscriminately to all neurons in a particular brain region. Fuxe suggested that such pathways deliver “emotional” signals to the brain—such as the “reward” signals that act to reinforce experience-dependent plasticity in neuronal circuits.2
The dopamine neurons of the substantia nigra have been studied intensively because it is their death that causes Parkinson’s disease. The substantia nigra is a large structure in the midbrain that is involved in the control of movement and in reward: it gets its name because the abundant dopamine neurons there express a pigment, neuromelanin, giving it a dark color. The substantia nigra is reciprocally connected with the striatum, a large region in the forebrain that also receives major inputs from glutamate neurons of the cortex and thalamus, and its output cells are “medium-sized spiny neurons” that release GABA (spiny neurons are neurons whose dendrites exhibit a profusion of protruding “spines” that increase the surface area of the dendrite, providing more sites at which synapses can form). It was once thought that the dopamine projections from the substantia nigra to the striatum allow an exquisite cell-specific tuning of the responses of the spiny neurons to excitatory inputs from the cortex. The numbers tell another story. In the rat, the axon of each dopamine neuron innervates about 6% of the total area of the striatum.3 These synapses were once thought to target the bases of dendritic spines, where they could specifically modulate the glutamate inputs from cortical projections. But Jonathan Moss and Paul Bolam in Oxford, in a detailed electron microscopic study, carefully quantified the contacts between dopamine terminals and their relationship to the terminals from cortical and thalamic projections. They found that “the organization of the dopaminergic nigrostriatal system is such that striatal neuropil is located within a dense, evenly spaced lattice-work of dopaminergic axons, and the probability of being apposed by a dopaminergic axon is principally dependent on the size of the structure. Thus, all similarly sized striatal structures have an approximately equal probability of being apposed by a dopaminergic axon. The organization is such that apposition by, and proximity to, a dopaminergic axon is random in nature.”4
Thus while dopamine was once thought to have actions confined to the synapse, many dopamine receptors are found outside synapses, and dopamine spills over from synapses to act at these extrasynaptic receptors. Moss and Bolam found that every structure in the striatum was within 1 µm of a dopamine synapse. In short, dopamine in the substantia nigra looks like a diffuse and indiscriminate signal, not one of exquisite temporal and spatial selectivity—more like a hormone than a neurotransmitter.
Neurons are not the only cells in our bodies that process and transmit information. The classic endocrine systems include five major steroid hormones: testosterone, estrogen and progesterone secreted by the gonads, aldosterone and cortisol by the adrenal cortex. These hormones are not stored, but are produced on demand; because they are soluble in lipids, no cell membrane can contain them, and they regulate gene expression in tissues throughout the body, including in the brain—there is no blood-brain barrier to steroids. Prostaglandins are other lipid-soluble hormones that are not stored but are produced on demand. They come in diverse forms, are produced by most tissues in the body, and have diverse actions; they regulate blood flow in many tissues, and regulate uterine contractions in labor. Thyroid hormones are also lipid soluble, but are carried in the blood to their sites of action by binding proteins, and they regulate protein, fat, and carbohydrate metabolism and brain development.
Steroid hormones are soluble in lipids and so generally have no difficulty in accessing the brain, but there are some important exceptions. One of these exceptions is important for understanding the difference between the brain of a male and that of a female.
The mammalian hypothalamus is sexually dimorphic; the hypothalamus differs between males and females in a number of very visible ways. In the human hypothalamus, the sexually dimorphic nucleus is a part of the preoptic area that is twice as large in men as in women, and many other differences are apparent on close examination. In part, the sexual dimorphism arises from the effects of sex steroids released in early postnatal life, effects that last throughout life.
A newborn male mammal has large testes; for a short while after birth, the testes secrete large amounts of testosterone, and its actions “masculinize” the developing hypothalamus: if the testes of a rat are removed soon after birth, the hypothalamus develops as a “female” hypothalamus. If an ovary is transplanted into an adult male rat that was castrated at birth, the rat will show cyclic ovulation, and at the estrous phase of that cycle will respond to a male with a lordosis reflex—the arched-back posture that invites mounting by the male. The same is not true of a male rat castrated in later life: the ovulatory mechanism is potentially present in the male brain at birth, but is prevented from developing by the actions of testosterone at a critical period in neonatal development. Conversely, if a newborn female rat is given an injection of testosterone in her first few days of life, the ovulatory mechanism of the hypothalamus fails to develop and she will be forever sterile.
What does testosterone do that makes such a profound difference to the developing hypothalamus? Strangely, it is not testosterone that masculinizes the male brain, but the female sex hormone, estrogen. If a large amount of estrogen is injected directly into the brain of a newborn female rat, she will develop a male hypothalamus just as if she had been injected systemically with testosterone. The female is normally protected from the relatively low amounts of estrogen that she experiences in neonatal life because the fetus makes a plasma protein called alpha-fetoprotein, which avidly binds estrogen and prevents it from reaching the brain.
In males, by contrast, testosterone is abundant and is less effectively bound in plasma. When testosterone reaches the hypothalamus, some is converted to estrogen by aromatase, an enzyme expressed by neurons in sexually dimorphic regions of the hypothalamus. If this enzyme is blocked, then the brains of male mice develop as female brains and the brains of female mice are resistant to the masculinizing effects of neonatal testosterone.5 The effects of neonatal testosterone exposure are evident in the density of neurons in certain hypothalamic areas, in the connections that neurons in these areas make with other brain regions, and in the phenotypes of neurons in these areas.6 The effects are exerted on neurons that, in early life, express receptors for estrogen; some of these neurons will die, others will change their phenotype and express a different complement of peptides and receptors in later life—and these changes will indirectly affect neurons to which they are connected. One important consequence is that, in adult life, the brain that had developed in the male pattern will respond differently to sex steroids than a brain that has developed in the female pattern.
In the 1980s, Nederlandse Hersenbank (the Netherlands Brain Bank), was set up by Dick Swaab, the head of the Netherlands Institute for Brain Research, to study the effects of aging and other factors on the human brain. Vasopressin was one of the few peptides that could be reliably identified in human postmortem brain tissue, and Swaab and his coworkers set about studying how vasopressin cells differed according to sex, age, and other factors. In the suprachiasmatic nucleus, there seemed to be many more vasopressin cells during the day than at night, and many more in summer than in autumn.7 The cells themselves couldn’t be disappearing and reappearing— there had to be marked changes in how much vasopressin they were producing. The shape of the nucleus also differed between males and females, although the number of vasopressin cells was similar between sexes: the size of the vasopressin cells was changing.
At this time, AIDS was scything down young lives, especially those of homosexual men, and some of the brains of this cruel harvest were donated to the Brain Bank. When Swaab and his team looked at these brains, they found that the suprachiasmatic nucleus contained, on average, more than twice as many vasopressin cells in homosexual men who had died from AIDS as in heterosexual men who had died from AIDS.8 Knowing that male rats exposed prenatally to an aromatase inhibitor show “bisexual” partner preference, Swaab and colleagues went on to look at the brains of these rats—and they, too, had many more vasopressin cells in the suprachiasmatic nucleus than control rats.9
Steroid hormones, thyroid hormones, and prostaglandins all have diverse actions on the brain as well as on peripheral tissues. But most hormones are peptides: the anterior pituitary gland secretes prolactin—which (among many other things) regulates milk production—and five other peptide hormones, and is regulated by at least a dozen peptide hormones secreted by neuroendocrine neurons. From the anterior pituitary, thyroid-stimulating hormone regulates the production of thyroid hormones, and is regulated by thyrotropin-releasing hormone; LH and FSH regulate the gonads and are regulated by GnRH; ACTH regulates the adrenal cortex and is regulated by CRH and vasopressin; and growth hormone, regulated by GHRH and by somatostatin, regulates muscle and bone growth by regulating the secretion of insulin-like growth factor-1 from the liver. α-MSH from the intermediate lobe of the pituitary regulates melanin production. Insulin, amylin, and glucagon from the pancreas control glucose homeostasis. Secretin, the first hormone to be discovered, is secreted from the small intestine and regulates gastric acid secretion and bicarbonate homeostasis. Parathyroid hormone regulates calcium homeostasis. Relaxin from the ovaries controls labor, by softening the cervix. Angiotensin from the kidneys regulates blood pressure. Natriuretic peptides from the heart regulate sodium excretion. Cholecystokinin from the duodenum regulates digestion of fat and protein, and gastrin from the stomach regulates gastric motility. Other hormones from the gastrointestinal tract, including ghrelin, peptide YY, and glucagon-like peptide, signal to the brain to regulate energy balance, as do leptin and other adipokines secreted from adipose tissue.
This is a very partial list, and in many ways it is misleading. Each of these hormones has not one target but many, and it’s often hard to be sure which role is most important, or whether that is even a sensible question. They act at multiple spatial scales, with autocrine effects on the cells that secrete them, paracrine effects on neighboring cells, and hormonal effects on distant targets. These actions are determined by the regulation of production and of secretion, and by the regulation of receptor expression in the target tissues. Their actions include acute activational effects on targets, slower effects on gene expression, and organizational effects on tissue development.
It is probably simplest to say that every cell in the body produces many different molecules that act both as feedback signals to the cells that produce them and as signals to other cells, sometimes only to close neighbors, sometimes to more distant cells. The diversity of signals is enormous, the receptors for them are even more diverse, and the ways that receptors are coupled to cellular function are more diverse again.
The concept of a hormone arose from the discovery by Bayliss and Starling in 1902 of secretin, a factor secreted from the duodenum that regulates pancreatic secretion. Three years later, Starling wrote: “We recognized these so-called internal secretions were merely isolated examples of a great system of correlations of the activities, chemical and otherwise, of different organs, not by the central nervous system but by the intermediation of the blood by the discharge into the blood stream of drug-like substances in minute proportions which evoked an appropriate reaction in distant parts of the body.”10
A hormone, as first understood, was characterized as a substance produced at one site that was secreted in the blood to act at another, and which exerted effects in proportion to its concentration at the target tissue. When defined in this way, hormones are radically different from neurotransmitters. A neurotransmitter carries a message from one cell to another at a specific place—a synapse—and at a specific time—its effect lasts just a few milliseconds. These messages are targeted by the anatomical arrangement of interneuronal connections. By contrast, a hormone carries a message from one population of cells to another; its effects are extended in time, and depend on the ability of target cells to specifically, selectively, and proportionately recognize its presence.
Neurotransmitters are released one vesicle at a time, as discrete signals. When a vesicle is released into one side of the synaptic cleft, the concentration of neurotransmitter in the cleft reaches high levels, and some neurotransmitter will bind to receptors on the other side of the cleft. These receptors are molecules that span the cell membrane; on the extracellular side, the receptor molecule forms a “binding pocket” into which only molecules of a certain shape can fit. Molecules that fit in this way are called ligands for the receptor, and receptors are characterized in terms of their affinity for those ligands. Affinity is a measure of the probability that a molecule of ligand in solution will bind to the receptor. It takes only one molecule to activate a receptor molecule, but receptors with a low affinity need to be exposed to a high concentration of ligand for it to be likely that a molecule will bind to them; those with a high affinity can be sensitive to very low concentrations. Receptors for neurotransmitters do not need a high affinity for their ligands; on the contrary, it is perhaps better that they have a low affinity to ensure that the action is brief and local. The brevity of the actions of neurotransmitters means that they can transmit signals that vary rapidly over time.
Hormones are, by contrast, continuous signals, with half-lives of several minutes in the blood. Their effects are proportional to concentration and are mediated by receptors with very high affinity. They act selectively on populations of cells that express the relevant receptors, and those receptors are specific: they are not activated by any other signals unless present at very high concentrations. These characteristics—specificity, selectivity, and proportionality—are conferred by the density of expression of receptors.
Endocrine cells have neither dendrites nor axons, but many are like neurons in other ways. Some are electrically excitable: when pancreatic beta cells see an increase in extracellular glucose concentration they fire in bursts of spikes that are like the phasic bursts of vasopressin neurons; these bursts lead to calcium entry and trigger insulin secretion. In both neurons and endocrine cells, peptides are packaged in vesicles just as neurotransmitters are. Typically, peptide secretion is the result of the same process as that by which neurotransmitters are released: exocytosis is triggered in both cases by an increase in intracellular calcium. In neurons, this happens when spikes depolarize the neuron, opening voltage-sensitive calcium channels, and the same occurs in spiking endocrine cells.
However, endocrine cells have another trick. The cell bodies of all eukaryotic cells contain rough endoplasmic reticulum, which sequesters free calcium, and activation of receptors for some neurotransmitters or hormones can release calcium from these stores. In many endocrine cells, this “calcium mobilization” can trigger exocytosis of vesicles without any involvement of spikes. There is no rough endoplasmic reticulum in axon terminals, so spikes are necessarily involved in the release of synaptic vesicles.
Although the mechanisms by which vesicles are released are similar for synaptic vesicles and the vesicles that contain peptides, there are differences. Synaptic vesicles are only found at synapses, which contain specific release sites where just a few of the vesicles are “docked” ready to be released. A spike that invades a synapse will seldom release more than one vesicle, and often will release none; this is a stochastic process, and is erratic. The strength of a synapse can be quantified by the probability that a single spike will release any neurotransmitter, and this can change dynamically. The release probability depends on the size of the readily releasable pool—how many vesicles are docked—and declines as the pool is depleted. Conversely, when two spikes occur close together, the release probability can be higher for the second spike than for the first.11 But endocrine cells have no specialized release sites; exocytosis can occur anywhere on the cell membrane. Vesicles are not generally waiting to be released, but, in response to signals, they are delivered to a readily releasable pool from stores deep inside the cell. This “trafficking” is organized by the cytoskeleton, an intracellular scaffold of actin fibers and microtubules that can either impede or facilitate access of vesicles to the plasma membrane.12
Magnocellular oxytocin and vasopressin neurons combine properties of classical neurons and classic endocrine cells. Like classical neurons, they receive inputs at their dendrites, generate spikes at the cell bodies, and propagate these spikes down their axons to trigger release from the nerve endings. However, their dendrites are also packed with vesicles, and these can be released by signals that mobilize intracellular calcium.
Some peptides can trigger secretion from the dendrites while inhibiting electrical activity. α-MSH acts at MC4 receptors on the oxytocin cells to mobilize intracellular calcium; this triggers oxytocin release from the dendrites, but also increases the production of endocannabinoids, which inhibit spiking activity. Thus α-MSH suppresses oxytocin secretion from the pituitary, which depends on spike activity, but stimulates oxytocin release within the brain. By combining properties of neurons and endocrine cells, oxytocin cells can independently regulate what they release centrally and peripherally.
This is exciting but unfortunate. We might wish, as many have wished, that by measuring oxytocin in the blood we might infer what is happening within the brain. We cannot. We might have hoped, as many have hoped, that by recording the electrical activity of oxytocin cells we might know what they are releasing within the brain. We cannot.
These characteristics may be common to the many types of neuron that make peptides in abundance. Vasopressin release from dendrites is even more abundant than oxytocin release, and is similarly governed by intracellular calcium mobilization. Generally, the large vesicles in which all secreted peptides are packaged are not found only or even mainly at synapses. They are distributed throughout the cytoplasm, in dendrites, cell bodies, axons, and terminals, and it appears that for peptides generally there are no specific release sites: exocytosis can occur from any compartment. The textbook image of a neuron is misleading: typically dendrites are shown as minor appendages to the cell, but they normally constitute about 85% of the cell volume. Dendritic release of peptides may not be an exception but the norm.
Endocannabinoids too are widely distributed in the brain; they, like steroids, are not stored in vesicles but are produced on demand. They have many effects, including on appetite, sleep, stress responses, and anxiety-linked behaviors, and the wide distribution of endocannabinoids and their receptors suggests that they are a common retrograde signaling mechanism. There are many others—the gas nitric oxide is one other, produced in some neurons by the enzyme nitric oxide synthase, but also in the endothelial cells of blood vessels by a slightly different version of the same enzyme. Another is adenosine; it is produced by most neurons when they are active and exported by a membrane transport mechanism to the extracellular fluid, where it can act at adenosine receptors on neighboring cells to inhibit them: caffeine has its stimulatory effects on brain function by blocking some of these receptors.
We have classically thought in terms of hierarchies, of higher centers, master controllers and slaves, systems and subsystems, placing the cerebral cortex at the top as the organ of personal responsibility. But all parts of the brain are massively and reciprocally interconnected. The organization is parallel, not serial; information flows between structures equally in two directions, and does so at every level of organization—between neurons and their “inputs,” between neighboring neurons, between cells within networks, and between networks.
We can go further: the brain is massively and reciprocally connected with all parts of the body, the organization is parallel, not serial, and information flows between brain and organs in both directions and between organs and other organs. The brain communicates with organs by both neuronal signals and endocrine signals, and peripheral organs communicate with the brain by both neuronal signals and endocrine signals. The gut has an extensive nervous system all of its own, the enteric nervous system: in humans this contains about 500 million neurons. The brain does far more than we are conscious of, and what we are conscious of is molded by a multitude of things that we are not conscious of.
Notes