neurotransmitters and neuromodulators

The soft warm living substance of the brain and nervous system stands in stark contrast to the rigid metal and plastic hardware of a modern day computer, but at the fundamental level there are clear similarities between these two apparently disparate organizational systems and, of course, one is a product of the other. Not only are the nerve cell units (neurons) self-repairing and self-wiring under the grand design built into our genes, but they can also promote, amplify, block, inhibit, or attenuate the micro-electric signals which are passed to them, and through them. In this way they give rise to signalling patterns of myriad complexity between networks of cerebral neurons, and this provides the physical substrate of mind.

These key processes of signalling by one group, or family, of neurons to another is achieved largely by the secretion of tiny quantities of potent chemical substances by neuronal fibre terminals. These neurotransmitters stimulate selected neighbours, with whom they junction, into producing electrical responses which both qualitatively (i.e. by excitation or inhibition) and quantitatively (i.e. by frequency of neurotransmitter release) reflect the patterns of presynaptic stimulation.

In this way, the nerve impulses are passed on from cell to cell. This continuous alternation between electrical and chemical conveyance of signals on their journeys through the pathways of the brain and nervous system provides a special opportunity for the traffic of electrical impulses to be modulated or blocked as they attempt to jump the gap between one neurone and the next at their junctions, transposed into pulses of chemical substances. This is the point where selected constellations of neurons from the vast array of neuronal populations can effectively interact, one with another, to filter, edit, integrate, and add precise direction to their interplay of communication. Thus, neurotransmitters and their functional partners, the neuromodulators, play a cardinal role in controlling the flow of information through the nervous system.

1. Neurons as information receivers and transmitters
2. How many neurotransmitters are there?
3. How do neurotransmitters work?
4. Neuromodulators
5. Coexistence and co-release of neurotransmitters
6. Postsynaptic and presynaptic neuroreceptors
7. How are neurotransmitters released?
8. Why are there so many neurotransmitters?

1. Neurons as information receivers and transmitters

The extensive web of branching dendrites which characterizes so many neurons in the brain is primarily an adaptation to provide maximal surface area for receiving inputs from other nerve cells with which they make contact. Interaction is very much the principle theme which underlies the shape and cellular anatomy of neurons. Each pyramidal neuron in the vertebrate brain is likely to be receiving up to some 100,000 contacts from the neurons to which they are wired, and the dazzlingly complex and extensive multi-branched dendritic tree of the Purkinje cells of the cerebellum, concerned with learning coordination tasks, probably extends to some 300,000 neuronal contacts. The axons, the single output line of the neuron, can be rather short (e.g. in so-called interneurons) or very long, perhaps 12 metres in cortical pyramidal cells of the giant blue whale, which course from brain to lower spinal cord. At various points along its length, the axon may branch to make contact with local neuronal communities, though most of its contacts are made towards its terminal region. Thus, it is not surprising that the cell-body region of the neuron is estimated to take up only 5 per cent or less of the cellular volume of the brain, the greater part comprising the dense fibrous feltwork of dendrites and axons (Fig. 1).


Fig. 1. Golgi preparation from the visual cortex of a human infant, showing the vertical orientation of many neuronal processes. Dendrites (D) can be identified by the dendrite 'spine' processes which give their surface a rough, granular appearance (see also Figs. 2 and 3). In contrast, axons (A) and neuron cell bodies (CB) are smooth-surfaced.
Sometimes the neuron-to-neuron contacts or synapses, as they are termed, are highly organized junctions allowing an extremely close approach between the two cells so that they are separated only by a very narrow gap or cleft (typically 20 nanometres wide) (Figs. 2 and 3). In this case, neurotransmitters are delivered at a very precise location when their secretion is triggered by the incoming nerve-terminal. In other cases, the axon produces long chains of swellings (so-called varicosities) towards its terminal branching regions, and these provide multiple sites for neurotransmitter release as the nerve-impulse courses through them on its passage towards the distal regions of the neuron (Fig. 4).

These varicosities, while containing synaptic vesicles and granules of the type characteristic of the 'tight-junction' synapses described above, do not exist in highly organized apposition to other neurons, and the highly structured synaptic cleft (gap between neurons) or post-synaptic thickenings typical of 'tight-junction' synapses are rarely present (compare Figs. 3 and 4). The neurotransmitter is probably released from the whole bulbous surface of the varicosity as a miniature cloud which will diffuse away, diminishing in concentration, until it encounters the appropriate neurotransmitter neuroreceptors at which it can bind and act. These varicose modifications of nerve axons were first discovered in the peripheral nervous system, but in the past decade they have become established as a common feature of the synaptic organization of the brain itself, and provide a semi-localized form of neurotransmitter release called paracrine neurosecretion.


Fig. 2. Electron micrograph of a nerve axon ending making a synapse of the 'tight-junction' variety in a rat brain. These synapses are identical in the human brain. The presynaptic nerve axon ending (axon A) makes synaptic contact with a dendrite spine (S) of the dendrite of the next (postsynaptic) neuron. Note the synaptic cleft (C), 20 nm wide, which is the narrow space separating the two neurones, synaptic vesicles (V), and postsynaptic densities (d). Also note the synaptic vesicles emptying their neurotransmitter content into the cleft by exocytosis (compare with Fig. 3)



Fig. 3. Diagrammatic version of an axon terminal forming a synapse on a dendrite spine. The structures shown are only approximately to relative scale. Endocytosis produces coated vesicles consisting of a hexagonal basketwork of fibres (cytonet) which form part of the inner surface of the nerve terminal membrane at regions called 'coated pits'. In this diagram, vesicles are shown which contain monoamines, neuropeptides (in granules), or other neurotransmitters, are shown. A vesicle is shown expelling a neurotransmitter into the synaptic cleft by exocytosis. Compare with Fig. 2.



Fig. 4. Diagram of a neuron with axonal terminal varicosities.

2. How many neurotransmitters are there?

The search for chemical agents which could transmit the activity of peripheral nerves onto their target organs began early in the 20th century. In the 1920s and early 1930s, after a long trail of research acetylcholine was unequivocally demonstrated to be mediating the inhibitory influence of the vagus nerve on the heart, as well as the excitatory action of motor nerve terminals on voluntary muscle. At that point, acetylcholine became the first chemically identified neurotransmitter substance. Adrenaline, too, was an early candidate as neurotransmitter in the peripheral nervous system at ganglionic sites, but after long accumulation of evidence its unmethylated derivative, noradrenaline, was finally shown in the mid-1940s to be the actual agent responsible. Dopamine, a closely related amine (monoamines), followed a similar history, with doubt and then certainty following its progress to acceptance as a neurotransmitter in the 1950s (see catecholamines). In the 1940s another neuroactive monoamine, serotonin, first isolated from blood was accepted as a neurotransmitter in the brain and peripheral nervous system.

Thus, in the 1950s only four compounds, together with a few other unlikely candidates, including a peptide (substance P), were the full armoury of agents known to be acting as neurotransmitters. They were found to be localized to the neurons from which they were released. These substances seemed to be specialized for their task as neurotransmitters and were not involved in other biochemical activities. At that time, it seemed likely that, together, they provided the principal means of chemical neurotransmission throughout the nervous system. However, it was also in the 1950s and later in the 1960s that four amino acids (glutamate and aspartate as excitatory agents, and GABA and glycine as inhibitory agents), also amines (but carrying an acidic group as well), were being considered as new and important contenders as neurotransmitters in the brain and spinal cord. They were most unlikely candidates, being found ubiquitously in all cells and organs in high concentrations, and being involved in a wide range of metabolic pathways and biosyntheses in the general biochemical economy of the cell.

The first members of yet another entirely different biochemical category of neurotransmitters became serious contenders in the 1970s, namely the neuropeptides (2 to 50 residue oligopeptides). Unlike the amino acids, the neuropeptides are mostly present in extremely small quantities in localized regions of the nervous system. The earliest candidates proved to be already operating in the brain as local neurohormones in the hypothalamus and anterior pituitary gland. One example is thyrotropin-releasing hormone (TRH, a tripeptide). During the 1970s and 1980s many or most of the peptides known to be serving an endocrine or neurotransmitter role in the gastrointestinal system were found to be also serving as neurotransmitters in the brain. The significance of the dual existence and bioactivity of these peptides is not clear, but specific neuroreceptors for them exist in both brain and gastrointestinal systems, allowing the possibility of brain–gut interactions at the neurohormone level, and giving rise to the concept of the 'brain–gut axis'.

Many of these neuropeptides seem to evoke rather more complex responses than simple physiological synaptic excitation or inhibition (both of which they also mediate). This includes evocation of behavioural and emotional responses. For example, very small quantities of TRH can induce euphoric states, and it can act as an antidepressant drug for the treatment of affective disorders. Another neuropeptide, β-endorphin, causes muscular rigidity and immobility (catatonia), whilst luteinizing-hormone-releasing hormone (LHRH) is reputed to stimulate the libido, and has been used to cure oligospermy. Cholecystokinin (CCK) and Gastrin promote feelings of appetitive satiety and cause cessation of feeding in animals via the brain–gut axis of communication. Bombesin dramatically lowers body temperatures, controls many aspects of gastric secretion, and stimulates appetite by actions at sites within the brain. The endorphins and encephalins not only produce fairly complex and sophisticated behavioural effects, they also induce analgaesia, behaving like endogenous 'morphine-like compounds'. Unfortunately, they (and their active synthetic derivatives) are also addictive when given in quantity as analgesic drugs. The endorphins seem to serve a neurohumoral role as well as that of neurotransmitter (see psychopharmacology).


Fig. 5. Diagram illustrating the biosynthesis, packaging, and release of neuropeptides, and other neurotransmitters, in neurons. The peptides are generated from very large precursor molecules (pre-proteins) produced in the rough endoplasmic reticulum (RER) of the neuronal body. These are packaged into secretory granules or vesicles in the Golgi membrane stacks. The granules are then transported out of the cell body (by axonal transport) to the terminals. Here they release their contents by exocytosis by incoming nerve impulses. Other neurotransmitters are produced in the cytosol of the cell body, axon, and principally in the nerve terminal. They are then packaged into synaptic vesicles by specific uptake processes.
Peptide-releasing neurons show certain special features in their organization. Thus, cholinergic (i.e. 'worked' by acetylcholine), monoaminergic, and amino acidergic neurons synthesize neurotransmitter principally in their nerve terminals, by simple enzymatic processes. Peptidergic cells, in contrast, synthesize their peptide neurotransmitters as subcomponents of large 'mother' proteins by protein synthesis processes occurring in the cell body. These are then loaded into large granules and transported down the length of the axon to the terminal regions of the neurone. During this journey, the active neuropeptides are 'snipped' off the mother protein by specific enzymes (proteases), and the peptide neurotransmitters are then ready for release on arrival (Fig. 5). Inactivation of the released peptide neurotransmitter is by enzymatic hydrolysis, as for acetylcholine, and unlike most other neurotransmitters, which are rapidly reabsorbed back into the terminals, and into surrounding glial cells.


Table 1. Neurotransmitters, putative neurotransmitters, and neuroactive peptides.
When one surveys the current (2003) list of neurotransmitters (Table 1), it can be seen that apart from the simple amines and amino acids discovered during the first 70 years of the last century, some fifty neuropeptides must be included. The latter serve in hybrid capacities as neurotransmitters-neurohormones-neuromodulators, and in this way trigger off complex patterns of behaviour.In order to be accepted as a neurotransmitter, the substance must satisfy some eight key criteria. Until they satisfy all criteria, they are called putative neurotransmitters. Most of the compounds listed in Table 1 have satisfied these criteria and are fully-fledged neurotransmitters. There are usually several highly qualified putative neurotransmitters awaiting final acceptance or rejection, e.g. currently, agamatine, 3-,4-dihydroxyphenylalanine. Many of the neuropeptides listed (Table 1) are still of putative neurotransmitter/neurohormone status. In recent years, other, atypical, neurotransmitters have been discovered which have been accepted as serving the neurotransmitter function, even though they do not satify all the criteria referred to above. These include the gaseous substances nitric oxide and carbon monoxide. Like monosodium glutamate, viewed prospectively, these gases were most unlikely candidates as neurotransmitters, but now their case is proven (see below). Monosodium glutamate is now a well established excitatory neurotransmitter, which is known to operate far more synapses in the human and animal brain and spinal cord than any other known neurotransmitter!

3. How do neurotransmitters work?

These highly potent substances are released from their storage sites in the close apposition synapses, or in terminal varicosities (or dendrites — see below), and diffuse shorter or longer distances until they encounter neurotransmitter receptors with which they are designed to specifically interact. Once bound to the neurotransmitter in question, the neuroreceptor, which is a large glycoprotein molecule, spanning the membrane thickness (10 nanometres), undergoes conformational or other structural change, and this results in one of two known categories of response. The first of these is called an ionotropic response, and results in the appearance of a 'hole' or 'passage' right through the neuroreceptor protein molecule from outside to inside the membrane, through which only a particular charged ion can pass (Na⁺, K⁺, or CI). The specific ion in question proceeds to move through the neuroreceptor molecule either into or out of the cell interior, driven down its concentration gradient, and attracted or repulsed by the prevailing electric field across the membrane, according to the nature of its own net charge (Figs. 6 and 7). Each neuroreceptor channel may be open for only a very brief period (e.g. 1 microsecond) as the neurotransmitter rapidly dissociates and is inactivated, or may remain open for much longer periods (e.g. 1 sec.) depending on the ion channel concerned. As the postsynaptic membrane is densely packed with these structures, the net effect is a substantial movement of charged ions across the membrane. This movement generates excitatory or inhibitory synaptic potentials and, from this pattern of impingement of electrical signals (information) onto its dendrites and cell body (inhibitory inputs), the target neuron will be triggered to fire its own action potential, or remain quiescent, as appropriate according to the intensity of the excitatory and inhibitory signals received. Individual neuroreceptor channel-opening and -closing events can now be distinguished by so-called 'single channel recordings' by 'patch clamping' electrophysiological recording techniques where the neurotransmitter is perfused onto the postsynaptic membrane and the properties of small patches of membrane are studied.


Fig. 6. Three-dimensional models of the nicotinic acetylcholine receptor from the electric ray fish Torpedo californica depicted as membrane proteins in the postsynaptic membrane of the synapse between the nerve and electric organ. This organ works essentially like a neuromuscular junction, and employs acetylcholine as a neurotransmitter. Note the subunit arrangement around the central channel which conducts ions into, or out of, the electric organ. Small black spots indicate the sites on the two alpha subunits which bind acetylcholine, and other neurotransmitters. This binding leads to the opening of the ion channel. The proposed shape of the central ion channel can be seen in the vertical section. Also shown is the membrane structural protein, of mol. wt. 43,000, often found in association with the receptor in the Torpedo electric organ.



Fig. 7. Electron microscope picture showing acetylcholine-activated neurotransmitter receptors densely packed in the postsynaptic membrane of a cell in the electric organ of Torpedo californica, the electric ray fish. Note the central channel, or hole, through the centre of the neuroreceptor. The scale bar is 100 nanometres, showing that each neurotransmitter receptor is 8.5 nm wide.
The second category of neuroreceptor response is known as metabotropic because enzymes are involved. These enzymes are buried in the lipid membrane close to, and linked permanently or temporarily (depending on the cellular location) to, the neuroreceptor protein molecule. Following neurotransmitter-receptor activation there are conformational changes in the neuroreceptor-enzyme complex, which cause activation of the enzyme. The latter consequently acts upon its substrate, which is situated on the inner surface of the membrane, to produce a soluble product which diffuses into the neuronal cytoplasm and evokes responses which mostly lead on to produce a change in local membrane potential, usually resulting in the generation of a synaptic potential. These soluble enzyme products are called second messengers, as they produce actions which are secondary to the primary action of the neurotransmitter. Examples of such metabotropic, second-messenger, neuroreceptor-enzyme systems would be (1) the cyclic nucleotide system and (2) the phosphoinositide system:

1. The key enzyme here is adenylate cyclase which produces the soluble product cyclic AMP (adenosine monophosphate) inside the postsynaptic neurone. This catalyses a cascade of protein phosphorylations and dephosphorylations resulting in the secondary opening of particular ion channels with accompanying generation of synaptic potentials (Fig. 8). In this system coupling of neuroreceptor to adenylate cyclase is via a so-called G-protein (because it requires guanylate-triphosphate for its activation) which can convey either a stimulatory (N_s) or inhibitory (N_i) influence from the neuroreceptor to adenylate cyclase (Fig. 9). There is a parallel system which employs guanylate cyclase linked to neuroreceptors, which is particularly prominent in the cerebellum.
2. The key enzyme in the phosphoinositide system is a particular molecular subtype of phospholipase C (phosphoinositidase C) which is buried in the lipid membrane in close association with particular neurotransmitter receptors (Fig. 10). When this enzyme is activated indirectly via a G-protein following binding of neurotransmitter to neuroreceptor, its substrate, triphosphoinositide (actually, phosphatidylinositol (4,5) diphosphate), is hydrolysed to release water-soluble inositol triphosphate (IP, Inositol (1,4,5)-triphosphate) (Fig. 10). This IP molecule has the ability to release calcium from intracellular stores, and the calcium (which can be regarded as a third messenger in this case) then initiates a series of other biochemical events, particularly phosphorylations of key proteins in the cell. Such phosphorylations can switch on various ionic movements into the cell through channels (e.g. potassium ions) and thereby cause synaptic potential changes by this delayed and indirect route (Fig. 10).

The other product of phosphoinositidase C hydrolysis is a special neutral fat (diglyceride, containing arachidonic acid, a rather long fatty acid) which, in concert with calcium and a phospholipid, activates another enzyme, protein kinase C. This enzyme initiates a series of protein phosphorylations. These can result in the opening of local ion channels, and can therefore generate membrane or synaptic potentials (Fig. 10) on a par with those initiated by adenylate cyclase activation (see (1) above and Fig. 8). Close control of these two systems is provided by inactivating phosphatase enzymes, which work by rapidly removing the phosphate groups, from the key proteins involved, as fast as they are inserted to activate them (Fig. 10).

These metabotropic responses to neurotransmitter–neuroreceptor activation are necessarily much slower (10 to 30 times) than their ionotropic counterparts, because they involve enzyme activation and subsequent cascades of biochemical responses before inflow/outflow of charged ions. This is an intrinsically slower pathway than the ionotropic response which involves initiation of ion flow through membrane channels following their instant opening.


Fig. 8. The second-messenger concept: the adenylate cyclase system. Neurotransmitter receptor (R), buried in the postsynaptic membrane, interacts with neurotransmitter (T) and initiates cyclic nucleotide formation via the adenylcylase enzyme (AC), which is also in the membrane. The cyclic nucleotide (cyclic AMP) is the second messenger, the first messenger being the neurotransmitter released from the nerve-ending of the presynaptic neuron. The cyclic AMP stimulates phosphorylation of a postsynaptic membrane ion channel protein (IC) via a protein phosphokinase enzyme. This event opens the ion channel, allowing ions to move in or out of the cell, resulting in the generation of electrical signals, finally producing a new nerve impulse. The effect is reversed first by a phosphoprotein phosphatase, which de-phosphorylates the ion channel, and secondly by a phosphodiesterase which inactivates the cyclic AMP. Coupling between R and AC is known to involve guanylnucleotide-binding protein (G-protein). See Fig. 9 for details.



Fig. 9. Scheme showing the coupling inside the postsynaptic membrane between neurotransmitters and adenylate kinase (C) via guanidine nucleotide-binding protein (G-protein). Key: T, neurotransmitter; R, receptor protein; N, guanidine-binding protein; C, catalytic subunit of adenylate cyclase. Subscript 's' to N indicates a stimulatory action; subscript 'i' to N indicates inhibitory action. The coupling between the three membrane components is shown as being reversible. The binding of GTP by N enhances the interaction, which can result in either stimulation or inhibition of adenylate kinase, as shown, depending on the category of N involved. GTP is removed by its hydrolysis. See also Fig. 8.



Fig. 10. The second-messenger concept. The phosphoinositide-inositol system. Neurotransmitter receptor (R), buried in the post-synaptic membrane, interacts with neurotransmitter (T) and initiates formation of inositol (1,4,5) triphosphate (IP₃), the second messenger (the first messenger is the neurotransmitter). This is achieved by activation of the enzyme phosphoinositidase C (PIC, also in the membrane) via G-proteins with the consequent hydrolysis of a membrane phospholipid (PI, phosphatidyl inositol (4,5) diphosphate) which is located close to the enzyme and is indicated by the black heads. The second messenger, IP₃, releases calcium from intracellular stores and thereby raises the level of free calcium in the cytoplasm of that neuron. This second messenger is inactivated by another enzyme, inositol phosphatase. The other product of PI hydrolysis is diglyceride (DG). This activates the enzyme protein kinase C (PKC), which begins to phosphorylate various proteins, including specific ion channels, causing them to open and generate synaptic potentials and eventually to produce, or block, nerve impulses. PKC activity can also influence rates of neurotransmitter release, as well as produce profound effects on cell growth and development by influencing gene expression.

4. Neuromodulators

It is now well established that the synaptic action of a neurotransmitter may be modulated (i.e. made more or less efficient) by a third party, a neuromodulator substance, thereby amplifying or attenuating the action of the neurotransmitter. This seems to be achieved by more than one mechanism. Thus, the neuromodulator has the capacity to enhance or decrease the extent of release of the neurotransmitter following action potential invasion of the nerve terminal. For example, adenosinetriphosphate (ATP), secreted together with the neurotransmitter, will decrease noradrenaline or acetylcholine release from some adrenergic or cholinergic nerves, respectively. In other cases, the neuromodulator will alter the efficiency with which the neurotransmitter interacts with its neuroreceptor so as to allow the inward, or outward, flux of more ions per unit time by: (a) lengthening the opening period, (b) causing a greater frequency of channel opening, or (c) causing a greater activation of neuroreceptor-linked enzymes (e.g. adenyl cyclase).

Another, rather curious, neuromodulator substance is of considerable interest because of its links with benzodiazepine anxiolytic (anxiety-reducing) drugs. In fact, the existence of this naturally occurring neuromodulator has been inferred from the potent facilitatory actions of anxiolytic benzodiazepines, such as diazepam (Valium), on the most widespread inhibitory neurotransmitter system in the nervous system, namely the GABA (γ-aminobutyric acid) system. These benzodiazepine drugs both increase the affinity of GABA for GABA neuroreceptors located on synaptic membranes, and enhance GABA-mediated behavioural responses and synaptic potential generation. Moreover, there are neuroreceptors naturally present in the brain which very specifically bind to the drugs. The endogenous benzodiazepine receptor protein (or binding site) is thought to form part of the GABA receptor complex. When released from its nerve endings (or co-released with GABA from GABAergic nerve endings) it works by increasing the extent, or time period, of opening of GABA-operated chloride ion channels (Fig. 11). There has been a thorough hunt in the brain over the past fifteen years for the indwelling Valium-like endogenous benzodiazepine, but with little firm success. Various candidates have been found, but as yet no endogenous substance has been isolated which completely mirrors the properties of the endogenous neuromodulator, or endozepine, as it has now been termed. The latest front runner is ODN (octadecaneuropeptide), an 18 amino acid neuropeptide which has most of the properties of Valium. A specific blocking agent for ODN has been developed, called Flumazenil, which prevents the actions of both Valium and ODN supporting the case for ODN as the endogenous benzodiazepine.

A precise definition of a neuromodulator is difficult to produce, since the same substance may act as a neurotransmitter in one synapse, and a neuroregulator at another synapse. Therefore, a definition must represent the category of neuroactivity at a particular site, rather than the identity of the substance itself. For instance, ATP, ADP, AMP (adenosine tri-, di-, and mono-phosphates), or adenosine itself, may function as a neurotransmitter or as a neuromodulator according to the nature of the neuroreceptors with which it interacts.


Fig. 11. Neurotransmitter and neuromodulator action. a. Model of GABAergic synapse functioning with a double signal generation, that is GABA and a neuromodulator whose actions are mimicked by benzodiazepine anxiolytic drugs such as Librium and Valium. The coupler for the GABA recognition site and chloride ion (Cl) channel is shown as a postsynaptic membrane protein. In b. the GABA, when released alone, is seen to operate the chloride ion channel at normal levels of opening, allowing standard rates of Cl influx. In c. the released neuromodulator works through the receptor, along with GABA, to increase the influx of Cl through the channel by interacting with the GABA receptor, and changing its protein shape.

5. Coexistence and co-release of neurotransmitters

The long-held dictum, enunciated by H. H. Dale (1935) and later developed by J. C. Eccles in the 1950s, that 'only one and the same neurotransmitter is released by each neuron at all of its nerve terminals, at each of its terminals' (Dale's principle), is no longer universally tenable. We now have many examples where neuroactive peptides coexist with longer-standing ('classical') neurotransmitters, such as acetylcholine, noradrenaline, serotonin, and GABA. The coexistence of other, relatively newly established, neurotransmitters, such as ATP within cholinergic and adrenergic neurones is also now well founded. For the most part, the different neurotransmitters appear to be contained in different categories of storage vesicle or granule within the nerve terminals concerned, and can be released independently of one another. Thus, Vasoactive Intestinal Peptide (VIP) coexists with acetylcholine in parasympathetic nerves supplying the cat salivary gland, each being contained in its own vesicle type. High-frequency stimulation of the nerve releases VIP, whilst low frequencies release acetylcholine. Each neurotransmitter can act as a neuromodulator to influence the extent of release or post-synaptic actions of the other (Fig. 12).


Fig. 12. Co-transmission at a synapse by two neurotransmitters. This is the saliva secreting submaxillary gland in which acetylcholine (ACh), and a neuropeptide neurotransmitter, VIP (vasoactive intestinal peptide), coexist in the parasympathetic nerve terminals supplying this gland. Note that the ACh and VIP are stored in separate vesicles and can therefore be released at different rates at different nerve impulse frequencies to act on the saliva-secreting acinar cells to either increase (+), or decrease (−), saliva secretion. They also act on the blood vessels supplying the gland relaxing (−) the smooth muscles of the blood vessel walls to allow a greater flow of blood carrying precursors to the gland for saliva synthesis. Cooperation between the two neurotransmitters is achieved by selective release of ACh at low nerve impulse frequencies, and of VIP at high nerve impulse frequencies. The two neurotransmitters are also seen acting as neuromodulators mutually influencing the extent of one another's release and postsynaptic action. Key: (+), increased effect; (−), decreased effect.

6. Postsynaptic and presynaptic neuroreceptors

Postsynaptic neuroreceptors represent the longer established category of neuroreceptor, and provide the conventional feed-forward of electrical and trophic influences of neurone on neurone. The past three decades have seen the discovery of presynaptic neuroreceptors which serve a modulatory function in neurotransmission, being primarily concerned with controlling the extent of neurotransmitter release. These neuroreceptors respond to the principal neurotransmitter released by the nerve-ending concerned (so-called autoregulation), or to the actions of co-released, or extraneous, neurotransmitters or neuromodulators of different identity (so-called heteroregulation), by reducing or 'shutting down' the release of that neurotransmitter. This negative-feedback action is the more common consequence of presynaptic neuroreceptor activation (e.g. noradrenaline release from heart, spleen, vas deferens, and also centrally in the brain). A minor category of presynaptic neuroreceptors actually mediate enhancement (positive-feedback) of neurotransmitter release. It seems that both negative-and positive-feedback control can be exercised in the same synapse at a few particular central and peripheral axonal endings, particularly of adrenergic nerves. In this case, integration of their actions is achieved as follows: when low levels of neurotransmitter (e.g. noradrenaline) are present in the synaptic cleft, the facilitatory (usually beta-type) neuroreceptors are activated, leading to increased release of neurotransmitters. When higher concentrations are reached in the cleft, the inhibitory (usually alpha-type) adrenoreceptors come into play, resulting in a reduction in the level of noradrenaline release. Thus, differential sensitivity to neurotransmitter concentration allows a delicate balance to be maintained between the opposing actions of the two categories of pre-synaptic neuroreceptor.

7. How are neurotransmitters released?

Calcium ions are the critical factors in triggering neurotransmitter release from the nerve terminal onto adjoining neurons, muscles, or glands. Normally these calcium ions are at almost undetectable levels in nerve terminal cytoplasm (10M), but rapidly flood into the terminals from the surrounding fluid through special voltage-dependent channels during nerve-terminal depolarization caused by invasion by the action potential. Equally rapidly, the raised cytosolic levels of calcium are reduced to very low, ineffective, concentrations by fast absorption into mitochondria and endoplasmic reticulum within the terminal, and neurotransmitter release therefore ceases.

It has been demonstrated that in neuron-to-neuron communication, neurotransmitters can also be released, in the reverse direction, i.e. from the dendritic tree of the post-synaptic neurone, onto the incoming pre-synaptic axon terminal, as well as from the axon terminals onto this tree, as classically conceived. They can therefore act as retrograde neurotransmitters, providing feedback of information (see below).

It may be that neurotransmitters can be released from other regions of the neuronal cell surface, including the unmyelinated regions of axons, and from the cell bodies themselves, though the greater proportion of neurotransmitter is released from the nerve terminals, where it is mostly to be found.

Whether neurotransmitter is actually released from these various sites by a single process remains unclear. Certainly the traditional view envisages an exocytotic process involving the calcium-triggered fusion of the neurotransmitter-filled synaptic vesicles with the inner side of the nerve terminal wall, resulting in the expulsion of about 1,000 molecules (one quantum) of neurotransmitter into the synaptic cleft from each vesicle. Many multiples of such quanta are released as the nerve impulse invades the nerve terminal, and this involves fusion with the nerve terminal membrane and exocytosis by equivalent multiples of vesicles. However, the synaptic vesicles involved must already be in contact with the terminal membrane, and ready to release their content, as there is less than 200 microseconds between first entry of calcium through its voltage-dependent channel, and first detectable appearance of neurotransmitter in the cleft. The synaptic vesicles which release the neurotransmitters are positioned on the inside face of the membrane close to the point of influx of calcium through its channels, which span the membrane (Fig 13).


Fig. 13. Neurotransmitter release by exocytosis of synaptic vesicles. a. Protein domain structure of synaptotagmin and the other proteins involved in docking the synaptic vesicles at the presynaptic nerve terminal membrane. b. The interrelations of the proteins shown in (a) after docking and before the final exocytosis stage which releases the neurotransmitter into the synaptic cleft. Adapted from Fig. 1 in Littleton and Bellen (1995).
The balance of evidence is in favour of the vesicular release hypothesis, though this may not be exclusive, and vesicular release, and release by controlled outward diffusion through membrane channels, could be occurring, according to the neurotransmitter in question. The structure and function of the protein molecules present in the membrane or the vesicle, and which are involved in the exocytosis of synaptic vesicles, have been well defined over the past decade. At least eight of these have been isolated (Fig 13). First, synaptic vesicles attached, via proteins called neurophysins, to structural cytoskeleton scaffolding (microtubules and microfilaments) to limit their migration, are dissociated from this linkage by the addition of phosphate groups to the neurophysins by a calcium-stimulated phosphorylating enzyme. The vesicles can now move, slowly, but freely. Then, synaptic vesicles, full of neurotransmitter, are recruited, from those in the immediate vicinity of the pre-synaptic membrane, transported to this membrane, and then docked by attachment close to where an N-type voltage-dependent calcium channel protrudes through the membrane. Then the vesicle is 'primed' by addition of a molecule of magnesium-ATP salt complex, to enable it to exocytose to the outside and expel its 1,000 molecules of neurotransmitter into the synaptic cleft. The proteins involved here are known to include syntaxin and snap-25 (in the nerve terminal membrane), and synaptotagmin (a calcium-sensing protein), Rab3a and synaptobrevin (VAMP) (in the synaptic vesicle membrane). The next, and critical step, is the coming together and fusion of the vesicle-associated proteins with the membrane-associate proteins, rather like the fingers of two hands being closely intermeshed (Fig 13). This formation is known as the SNARE complex (i.e. Soluble N-sensitive factor Attachment protein REceptor), which is the last stage before neurotransmitter release. It drives vesicle fusion with the terminal membrane and immediate exocytosis by a process called zippering, in which calcium ions, flooding into the terminal through their membrane channels, bind to synaptotagmin and neurexin to change their conformations. This enables the latter to form the SNARE, and the former to bind it closer to the calcium channel. Other proteins, including alpha-snap, Rop, Rab3a, and NSF (Fig. 13) are necessary for docking of the vesicle at the terminal membrane before triggering of tight zippering of the SNARE proteins. This is followed immediately by exocytosis, with consequent neurotransmitter release. Spontaneous exocytosis in the absence of nerve impulse traffic is limited in rate by the vesicle membrane protein, synaptotagmin, whose block is neutralized when it binds to the calcium influxing during nerve impulse traffic. It is interesting that many of these proteins are also found in to be involved in membrane fusion and exocytosis which occurs in invertebrates and even yeasts, indicating their conservation throughout evolution due to their efficacy in these processes. Most excitatory glutamate receptors are silent most of the time. Glutamate is the most widely used excitatory neurotransmitter in the central nervous system, and operates far more synapses in the brain and spinal cord than any other neurotransmitter. Yet, its most common form, the NMDA ionotropic receptor, is paradoxically silent most of the time and glutamate cannot work as a neurotransmitter via these majority receptors! This is due to its ion channel, which runs across the post-synaptic membrane, being physically blocked by the entry of positively charged magnesium ions which are drawn into the ion channel from the extracellular fluid by the attracting influence of the negative interior of the neurone. This is established by the prevailing negative membrane potential across the neuronal membrane (approx. −70mV). Until this magnesium block is removed by a fall in membrane potential to −40mV or less, the NMDA receptor remains inactive. Membrane potential depolarization, and therefore activation of the NMDA receptor, is achieved by an unusual arrangement in the post-synaptic membrane. This involves the close juxtaposition of the NMDA receptor proteins with other types of glutamate receptors which are not blocked, such as ionotropic AMPA receptors or metabotropic glutamate receptors (mGluRs). Thus, when glutamate is released into the synaptic cleft from the pre-synaptic terminal, it cannot activate the NMDA receptors, but can activate the other, neighbouring, glutamate receptors. As a result, the post-synaptic membrane potential is locally depolarized by excitatory-postsynaptic potentials (EPSPs: usually due to sodium and calcium influx through the AMPA receptor ion channel). This depolarization causes efflux of the magnesium ions blocking the NMDA receptor channel, and this NMDA glutamate receptor can then respond to the synaptically released glutamate neurotransmitter, alongside the AMPA receptors in the same membrane, once sufficient membrane depolarization has occurred (Fig. 14). This functional arrangement means that NMDA receptors will only be recruited during periods of high and persistent neural activity, when glutamate release into the synaptic cleft occurs at a high rate. Once recruited, NMDA receptor activity, which has a large excitatory impact due to the high rate of entry of sodium/calcium ions, will massively boost the effect of the glutamatergic excitatory pathway. This occurs as part of the normal way of providing the upper range of intensity of normal nerve impulse traffic, and it is also involved in abnormal situations such as the initiation and propagation of epileptic fits. Retrograde and gaseous neurotransmitters, and synapses working backwards to establish memory stores. Among the interesting new discoveries in recent years has been the discovery that communication between nerve cells across a synapse can be two-way, i.e. forward directed (anterograde), and also retrograde in direction, back across the synapse. It was long thought that nerve impulses, when they invaded the nerve terminal, released neurotransmitter by causing influx of calcium ions, and the released neurotransmitter then activated specific postsynaptic receptors on the next neurone in the chain, causing post-synaptic electrical impulses which stimulated this receiving neurone. It was not conceived that events could take place inside this receiving (postsynaptic) neuron which could result in a retrograde action on the presynaptic neuron which had released the neurotransmitter. It did not seem possible, as no mechanisms for such retrograde action were known. This has turned out to be wrong; it is possible, and indeed, common. The thinking began to change when two observations were brought together. The first finding, in the 1970s, was that glutamate-operated synapses, in particular, can somehow learn when they receive an invasion, via the incoming axon, of high-frequency (tetanic) nerve impulses — a kind of intensive bout of neural activity, i.e. strong synaptic stimulation. This was shown by making simple electrical recordings from glutamate-operated pathways in isolated slices of the hippocampus, a brain region much involved in memory processes. It has subsequently been demonstrated to occur in synapses located in other brain regions concerned with memory, e.g. the cerebral cortex. Thus, after strong, high-frequency, tetanic stimulation, the post-synaptic response to a subsequent low/normal frequency stimulation was greatly enhanced. This enhancement of postsynaptic response lasted for at least several hours, and has been shown to last weeks. Indeed, the phenomenon was appropriately called Long-Term Potentiation of the postsynaptic response (LTP, Fig. 14). There is also Long-Term Depression (LDP) of the EPSP which can occur in other synapses under similar circumstances. The question was, how could the postsynaptic part of the synapse which was part of the next neurone in the chain apparently influence the presynapse to release more transmitter on all subsequent occasions that it received a low-frequency nerve impulse volley, i.e. how could it learn? After all, there was no known method of communication backwards from the second neuron to the first, i.e. from the postsynapse to the presynapse. The synaptic pathway was thought to be a rectifier, and allow only one-way traffic of the stimulus. It has since been established that there is a parallel increase in the pre-synaptic release of the excitatory neurotransmitter, glutamate, which occurs during the establishment of the LTP. Moreover, LTP was prevented by agents (antagonists) which blocked the synaptic action of glutamate at its post-synaptic receptors, showing that these receptors were somehow involved in LTP.


Fig. 14. Theories of how the retrograde gaseous neurotransmitter nitric oxide could be involved in Long-Term Potentiation (LTP) of a glutamatergic hippocampal synapse. The excitatory neurotransmitter glutamate first activates AMPA and metabotropic glutamate receptors which leads to recruitment of NMDA receptor activity. Calcium ion entry, via NMDA receptor channels, into the postsynaptic site activates nitric oxide synthase (NOS). The nitric oxide gas (NO) produced diffuses to the presynaptic site where it is absorbed by the haem group of an NO-sensitive guanylate kinase enzyme which triggers the production of cyclic GMP which increases neurotransmitter glutamate release. Key: mGluR, metabotropic glutamate receptors; IP, inositol triphosphate second messenger; PDE, phosphodiesterase, which inactivates cyclic GMP. Adapted from Fig 2 in Holscher (1997).
The second observation, in the 1980s and 1990s, which was to throw light on the mechanism of retrograde synaptic action, and the mechanism of LTP, was the finding that at least two gases are working as neurotransmitters, namely nitric oxide and carbon monoxide. The latter is, of course, a very poisonous gas when inhaled, and excess nitric oxide will also kill neurons! These surprisingly atypical neurotransmitters have properties which would allow them to work in a synaptically retrograde fashion, because, as gases, they are able to diffuse in all 3-D spherical directions, once synthesized and released. Unlike most other neurotransmitters, there are no storage pools for these gases. They must be synthesized and released when needed. Nitric oxide synthesis occurs both in the pre-and postsynapse. It readily dissolves in, and passes through, cell membranes, which do not, therefore, provide a barrier to its movement in any direction. It turns out that nitric oxide gas is made in significant amounts for a short time in the postsynaptic dendrite by the enzyme nitric oxide synthase (NOS) following the action of glutamate jointly at its postsynaptic AMPA and NMDA receptors (see 'silent glutamate receptors', above). This is because the ionotropic NMDA receptor channel, once activated following sufficient AMPA-glutamate receptor induced membrane depolarization, allows a large influx of calcium ions into the postsynapse, where it stimulates NOS to produce nitric oxide (Fig. 14). The NOS-enzyme is attached to the postsynaptic density (Figs. 2 and 3), close to the NMDA glutamate receptor. The released nitric oxide gas then diffuses rapidly and reaches the pre-synaptic nerve terminal where it stimulates the formation of the activating agent cyclic-GMP, by the enzyme Guanylate cyclase. The amount of glutamate neurotransmitter release is then increased via the action of this cyclic GMP (Fig. 14). Either the postsynaptic production of nitric oxide (or carbon monoxide; see below) must be constant to maintain the LTP, once established, or another, as yet unidentified mechanism is responsible for maintenance of the synapse in a potentiated state which persists for a lengthy period of time that greatly outlasts the period of application of the inducing stimulus.

Thus, nitric oxide gas is indeed likely to be one of the retrograde chemical signals that stimulate the persistent increased release of glutamate by the presynapse which characterizes LTP. Other candidates include carbon monoxide, which is produced by the enzyme haem oxygenase, which cleaves haem into carbon monoxide and the protein biliverdin. There are many parallels in the range of neuroactivity and the organization of the carbon monoxide and nitric oxide neurotransmitter systems, both in relation to NMDA receptor activation and to calcium influx. The long chain polyunsaturated fatty acid arachidonic acid is another front-running candidate as a retrograde signalling agent initiating LTP. These retrograde messengers, and LTP, are likely to be involved in the early phase of memory, lasting ten or twenty minutes, and involving acquisition and consolidation of the memory. This is where, immediately after acquisition, the memory is transformed from an initially unstable state into a stabilized store.

The calcium-dependent protein kinase enzyme, known as calcium calmodulin-dependent kinase (CaM kinase II) may be also be part of the mechanism which allows persistence and consolidation of the synaptic LTP (i.e. the memory store) triggered by NMDA-receptor activation. Thus, this CaM kinase protein is located postsynaptically, in particularly high concentrations (20–40 per cent) in the postsynaptic densities of glutamatergic synapses (Figs. 2 and 3). It is therefore in a good location to respond immediately to the cytoplasmic calcium ions flooding in through the activated postsynaptic NMDA receptor channels, following high-frequency (tetanic) synaptic stimulation. Furthermore, CaM kinase remains active, phosphorylating its own protein structure (autophosphorylation), as well as that of the juxtaposed NMDA and AMPA glutamate receptors, which further increases the rate of calcium and sodium ion entry through these receptor channels during subsequent nerve impulse invasion and synaptic activation. Metabotropic glutamate receptors and their second messengers such as IP (see Figs. 10 and 14) are also involved in LTP, though this is not so well understood. This activated synaptic state continues long after the calcium levels have dropped back to basal levels. Moreover, the CaM kinase enzyme protein has been reported to increase in amount in the postsynaptic dendrite during the tetanic stimulation which establishes LTP of the hippocampal synapse. Also, more synapses have been reported to develop, i.e. there is remodelling of the post-synaptic dendrite spine. The CaM kinase II enzyme can therefore 'remember' that the juxtaposed NMDA receptors have been activated, and the resulting calcium influx has produce phosphorylated CaM kinase II in the post-synaptic density. The stimulation has also stimulated increased gene-expression of the CaM kinase II protein of the postsynaptic density, dendrite, and cell body of the postsynaptic neurone over the subsequent 30 minutes (Fig. 15). These changes could partly explain why there is persistence of an enhanced postsynaptic response to subsequent pre-synaptic invasion of this potentiated synapse by nerve impulses, i.e. a key part of a the molecular mechanism of establishing memory traces in the brain.


Fig. 15. Theories of the process of induction of Long-Term Potentiation (LTP). From left to right: Normal stimulation of the presynaptic (upper) nerve terminal activates only the postsynaptically positioned glutamate-AMPA receptors, the juxtaposed glutamate-NMDA receptors being sterically blocked by magnesium ion binding to and occupying their ion channels. Strong repetitive (tetanic) stimulation, or high-frequency nerve-impulse invasion, of the presynapse activates first AMPA receptors sufficiently to cause a large depolarization (EPSP) of the postsynaptic membrane, and consequent activation of NMDA receptors by outflow of the blocking magnesium channels ions. As a result, sodium ions enter through the channel of the AMPA receptors, and sodium and calcium ions enter through the NMDA channel, into the cytoplasm of the postsynaptic dendrite, close to the postsynaptic density. This calcium ion influx induces LTP by activating nitric oxide synthase and CaM kinase II (see text). There is evidence that LTP involves, in the long term: increased presynaptic release of glutamate neurotransmitter, permanent increase in the expression and state of activation of CaM kinase II, phosphorylation to activate NMDA and AMPA receptors, as well as the increased synaptic response (e.g. 10-fold) to subsequent normal levels of synaptic stimulation by nerve impulses. Adapted from Sanes and Lichtman (1999), Fig. 1 in Trends in Neuroscience, 2/7.
The dendritic tree and retrograde neurotransmission. It has been demonstrated that neurotransmitters can also be released from the dendritic tree of the neuron as well as from axon terminals as classically conceived. Indeed, the dendrites show localized accumulations of synaptic vesicles which are characteristic of all neurotransmitter release sites (e.g. synapses, terminal varicosities). Both long-established neurotransmitters (e.g. dopamine, ATP), and neuropeptide neurotransmitters, such as vasopressin and oxytocin (normally known only as neurohormones), can be released from the dendrites of neurons at many sites within the CNS, to act as retrograde signals to modulate incoming synaptic transmission, electrical activity, and, in some cases, to modify the morphology of the surrounding presynaptic neurons. Such dopamine release has been unequivocally demonstrated to occur from the dendrites of dopaminergic neurones of the substantia nigra of the brain (see basal ganglia).

The neurotransmitters glutamate and GABA act as retrograde dendritic synaptic signals for the mitral cells and GABAergic axonless granule cells of the olfactory bulb. The dendrites can support an invasion of action potentials spreading, backward and upwards, from the cell body where they are generated. It is these invading action potentials which trigger dendritic neurotransmitter release. Through this mechanism, the neuron cell body can transmit information back, from its dendrites, to its neighbouring presynaptic neurons, through the actions of dendritically released neuroactive substances. The gaseous neurotransmitters nitric oxide and carbon monoxide and the long, polyunsaturated fatty acid arachidonic acid are other membrane-permeable neuroactive compounds that are released from neuronal dendrites onto synaptic inputs where they have a major role in synapse formation, and memory establishment.

8. Why are there so many neurotransmitters?

The answer to this question is far from resolved. Certainly, subgroups of the 50 or so neurotransmitters (Table 1) produce different categories of effect in both qualitative and quantitative respects, and many will coexist in different neuronal populations. First, there is the differential speed of response following receptor activation, with ionotropic being some 10-to 30-fold faster than metabotropic responses. Another dimension is provided in the variety of second-messenger activation of different cascades of biochemical sequelae, with particular metabotropic neurotransmitter actions being mediated by one or other second messenger system. Further possibilities for adding to the qualitative features of the response could come through the range of distances through which the neuroactive compounds exert their influence once released, if they can survive the inactivation processes. This may be 20 nanometres within the synaptic cleft but several millimetres from axonal varicosities or unstructured release points, and of course the concentration of neurotransmitter diminishes rapidly with the distance travelled, and some neuroactive substances will be less efficiently inactivated than others. Actions ranging over a greater volume of tissue would activate specific neuroreceptors sited on a larger variety of neurone types, producing complex sequences of neuronal triggering, and therefore a greater spectrum of overall responses (so-called trophic or paracrine actions). Continuous (tonic) release of neurotransmitter (e.g. from axonal varicosities or dendrites), as opposed to discrete and occasional release, would also provide a basis for variation in the patterns of activation of targeted neurons due to changes in sensitivity, accommodation, and desensitization of neuroreceptors which ensues during continuous neuroreceptor stimulation.

Thus, the large number and the chemical variety of neurotransmitters, together with their tendency to activate anatomically distinct neuronal pathways, often in pairs, and the evidence that they provide many palpable possibilities for variation in response, can be seen to provide a chorus of informational voices, each adding tonal colour or timbre to the final output of the brain and nervous system.

(Published 2004)

— H. F. Bradford

Bibliography
• Bajjalieh, S. (2001). 'SNARES takes the stage: a prime time to trigger neurotransmitter secretion'. Trends in Neurosciences, 24/12.
• Baranano, D. E., Ferris, C. D., and Snyder S. H. (2001). 'Atypical neural messengers'. Trends in Neurosciences, 24/2.
• Bradford, H. F. (1986). Chemical Neurobiology.
• Hampton, L. L. et al. (1998). 'Loss of Bombesin-induced feeding suppression in gastrin-releasing peptide receptor deficient mice'. Proceedings of the National Academy of Sciences of the United States of America, 95/6.
• Holscher, C. (1997). 'Nitric oxide, the enigmatic neuronal messenger: its role in synaptic plasticity'. Trends in Neuroscience, 20/7.
• Kandel, E. R., Schwartz, J. H., and Jessell, T. M. (2000). Principles of Neuroscience (4th edn.).
• Kennedy, M. B. (1998). 'Signal transduction molecules at the glutamatergic postsynaptic membrane'. Brain Research Reviews, 26/2–3.
• Kiss, J. P., and Vizi, E. S. (2001). 'Nitric oxide: a novel link between synaptic and non-synaptic transmission'. Trends in Neuroscience, 24/3.
• Lamacz, M., et al. (1996). 'The endogenous benzodiazepine receptor ligand ODN increases cytosolic calcium in cultured rat astrocytes'. Molecular Brain Research, 37.
• Littleton, J. T., and Bellen, H. J. (1995). 'Synaptotagmin controls and modulates synaptic-vesicle fusion in a calcium-dependent way'. Trends in Neuroscience, 18/4.
• Ludwig, M., and Pittman, Q. P. (2003). 'Talking back: dendritic neurotransmitter release'. Trends in Neurosciences, 26/5.
• Medina, J. H., and Izquierdo, I. (1995). 'Retrograde messengers, long-term potentiation and memory'. Brain Research Reviews, 21/2.
• Nichols, J. G. (2001). From Neuron to Brain.
• Ojima, K., Matsumoto, K., and Watnabe, H. (1997). 'Flumazenil and endogenous benzodizepines'. Brain Research, 747/1–2.
• Siegel, G. J., Agranoff, B., Albers, R. W., Fisher, S. K., and Uhler, M. D. (1999). Basic Neurochemistry: Molecular, Cellular and Medical Aspects (6th edn.).
• Snyder, S. H., Jaffrey, S. R., and Zakhary, R. (1998). 'Nitric oxide and carbon monoxide: parallel roles as neural messengers'. Brain Research Reviews, 26/2–3.

---