The antipsychotic landscape: dopamine and beyond

Abstract

Until recently, the actions of antipsychotic and pro-psychotic drugs have largely been evaluated in the framework of neuronal doctrine – namely, that neurons communicate by releasing transmitters, and that psychiatric disorders are caused by neurotransmitter imbalances. Moreover, the majority of studies have focused on single transmitter systems – neglecting the fact that in the nervous system, different transmitter systems work in concert and impact on not only their immediate receptors but also downstream pathways that shape structural plasticity. In this review, we discuss the history of understanding the antipsychotic and pro-psychotic actions of drugs, recent developments and future perspectives.

Introduction

Psychological health depends upon the health of the neuronal network. A fundamental property of neuronal networks is that they learn from experience. Learning is achieved by adjusting the strength of the connections between neurons. New connections form, and weak connections wither away, essentially a process of re-wiring, as the personality forms and develops.

Of major significance is the dawning realization that, downstream of the initial receptor stage, many psychiatric drugs impact upon structural plasticity in the network. In this review, we focus on the role of antipsychotic drugs on structural plasticity. As we learn more about the mechanisms that control structural plasticity, more precise targets will emerge for the treatment of major psychiatric syndromes.

From the neuronal doctrine to synaptic plasticity

Psychiatric drugs impact upon how neurons in a network communicate with each other. The power of the nervous system does not reside in a single neuron. Nervous tissue is immensely powerful because of the rich connectivity between neurons. A 1 mm voxel of cerebral cortex (a standard fMRI unit) contains ~300 million synaptic connections and ~50,000 neurons. Neuronal networks are the foundation of perception, movement, thinking, memory and the personality. ¹

Until the early 20th century it was unknown how neurons communicate. One school favoured direct electrical transmission, another favoured chemical transmission between neurons. The great Spanish histologist Santiago Ramón y Cajal was an adherent of the latter: the neuronal doctrine. Ramón y Cajal postulated that the small protrusions from dendrites, later termed spines, were connection points between neurons. ² Ramón y Cajal’s genius was to predict that the strength of synaptic connections might be altered by experience and account for memory, a theme we will return to below: synaptic plasticity. ³

In the neuronal doctrine, neurons communicate by releasing transmitter which is recognized by receptors on the target cell. By the 1970s it was possible to quantify neurotransmitter release and to visualize receptors. Twenty years later, receptors would be isolated, cloned, sequenced, catalogued, re-engineered and their 3D structures modelled. ⁴

Drug design in schizophrenia: not all serendipity

It is sometimes said that all the treatments in psychiatry were discovered by chance (or serendipity), rather than by planning. This is not strictly true. In fact, some pharmaceuticals for schizophrenia were discovered by design. The rationale was to start with a molecule that could induce a transient psychosis, even in healthy people. Thereafter the task was to find a drug that could block the effects of the psychosis-inducing compound. Such a drug, it was reasoned, could be an effective treatment for schizophrenia.

The Belgian pharmacologist Paul Janssen pioneered this approach to great effect. He observed the effects of amphetamine in professional cyclists, who were using the drug to combat fatigue. Some of the cyclists developed an acute psychosis that was identical to paranoid schizophrenia. Progress in finding antagonists to amphetamine was rapid and the compound haloperidol was discovered. ⁵ Used in small doses, without interruption, haloperidol is a powerful treatment against hallucinations, delusions and agitation.

The dopamine hypothesis

The antipsychotics were in use for over 10–20 years before it became clear that they blocked dopamine signals in the brain. Given the success of the antipsychotics, some suggested that the cause of schizophrenia was excess dopamine signalling in the first place. ⁶ Dopamine gained entry to the neurotransmitter club in the 1960s. Like much else from that era, the dopamine story is discovered anew by successive generations. That story – that the dramatic symptoms of schizophrenia are caused by too much dopamine – has survived as fashions have come and gone. However, even the most vocal proponents of the dopamine hypothesis agree that the idea of too much transmitter is likely to be an oversimplification. Neuroscience has advanced considerably since the 1960s and we now recognize dynamic, adaptive systems, with a multitude of interacting components in flux. ⁷

There are also some inconsistencies in the dopamine model. For example, it takes repetitive, addictive use of amphetamine/cocaine before psychosis emerges, not a single exposure, as the hypothesis predicts. Initially, stimulants provoke a huge release of dopamine onto target neurons in higher centres, but the initial experiences are in some way the opposite of a paranoid psychosis: feelings of confidence, energy, enhanced focus and wellbeing ⁸ – which is why people take the drug. Of course, stimulant addiction can lead, over time, to a paranoid psychosis, but paradoxically – and this is the important point – by that stage stimulant drugs evoke a relatively meagre dopamine release compared to the initial exhilarating/vitalizing doses.^9,10 For the dopamine hypothesis, this dissociation between the acute versus the chronic pharmacology of stimulants is problematic.

With repetitive patterns of stimulant use, plastic adaptations in the neuronal network occur. The plastic changes encompass not only biochemical, but also structural changes at individual connections in the network. ¹¹ Indeed, it is known that a period of antipsychotic treatment also modifies the structure of connections in the network.^12–15 Put succinctly, dopamine drugs, whether agonists or antagonists, ‘re-wire’ the network.

To understand the mechanisms of a drug it is necessary to look beyond the receptor, at the downstream consequences on protein synthesis and network connectivity. Certainly, short-term fluctuations in dopamine have a significant effect on mental experience, but these experiences recede fairly quickly and are usually inconsequential, whereas re-wiring the network means change in the personality – change in behaviour, attitudes, beliefs, aims in life, relations with others and so forth. Consider how difficult it is for the addicted brain to switch back to its former state. ¹⁶

Serotonin 5HT_2A (but still really D₂)

With the successful development of haloperidol from a blocker of the effects of amphetamine into an all-purpose antipsychotic, attention focused on other pro-psychotics. Researchers utilized LSD as another psychosis-inducing agent. Numerous reports had shown that LSD can transform consciousness in a way that has some similarities to the experience of people with schizophrenia. LSD and other hallucinogens, such as the natural compounds mescaline and psilocybin, completely distort the experience of lived reality. ¹⁷ The effects are pronounced and cannot be overcome by an effort of will or reasoning. Molecules such as LSD demonstrate that higher consciousness – perception, thinking, beliefs and will – have an organic substrate. The effects of psychedelics are attributed to very specific actions on serotonin receptors in nervous tissue. ¹⁸

For drug discovery, what was needed was a compound to block LSD, followed by a trial of the new compound in people with schizophrenia. As had been shown for amphetamine/haloperidol, a drug which is effective against a model psychosis might be effective for endogenous psychosis. Again this approach was successful, yielding the drug risperidone. ¹⁹ But risperidone is not just a serotonin blocker, it also blocks dopamine receptors. With time it has in fact come to be realized that the effect on serotonin signalling – blockade of a subtype of serotonin receptor called 2A – is probably not that important. It still appears that the ability to block dopamine D₂ receptors is the critical effect. Three selective serotonin 2A receptor antagonists were tested in schizophrenia with little success. ²⁰ However, the molecule pimvanserin (a selective 5-HT_2A inverse agonist) gained FDA approval for the treatment of psychosis occurring in Parkinson’s disease. ²¹

NMDA channels

Other drugs are pro-psychotic. The behavioural psychopharmacology of ketamine is similar to a natural molecule from Central Africa called ibogaine, which is used in traditional religious ceremonies. Both molecules can produce bizarre trance-like, mystical states. Ketamine is championed by some as the most convincing drug-model of schizophrenia. ²² Some researchers believe that ketamine produces the full spectrum of schizophrenic symptoms, including negative symptoms.

At the molecular level, ketamine ‘blocks’ an ionotropic receptor for the neurotransmitter glutamate, the NMDA receptor. The NMDA receptor is one of two types of ionotropic receptors that is opened by glutamate. The other receptor is called AMPA. The NMDA receptor is an essential component of learning at synapses. ²³ When there is ‘traffic’ through a synaptic connection, the NMDA channel snaps opens and lets calcium rush into the dendritic spine. Calcium sets off numerous processes which lead to strengthening of the synaptic connection. The more traffic through a synapse, the stronger (and larger) it becomes. The technical name for the process of strengthening of synapses is long-term potentiation (LTP). ²³ Throughout the whole brain, hundreds of billions of synapses have been moulded by the process of LTP. LTP is a mechanism that is in keeping with the initial idea of Ramón y Cajal, that the strength of the connections between neurons in a network can be adjusted during learning to form a permanent memory trace.

One other aspect of the NMDA receptor is notable. There is an additional receptor site on the NMDA protein for another neurotransmitter, called glycine. Both sites (glutamate and glycine) need to be occupied for the channel to open. The practical significance is that many of the drug discovery efforts in schizophrenia have attempted to alter the NMDA receptor by influencing the glycine site, rather than the glutamate site. ²⁴ Partly this is because direct drug actions at the glutamate site can be very toxic.

There was a massive investment in glutamate/glycine drugs, with the hope that these molecules would be effective in schizophrenia, not just for the positive symptoms, but also for the negative symptoms and cognitive impairments. ²⁵ Based on successful animal work – inhibition of the effects of ketamine – two candidates went forward into clinical trials. One strategy was to boost the glycine site on the NMDA channel. The drug bitopertin, a glycine reuptake inhibitor, stops glycine being recycled so that more is available for binding to the glycine site on the NMDA protein. The other strategy involved targeting an auto-receptor on the glutamate terminal. An auto-receptor is a receptor on the same presynaptic terminal as from where transmitter, in this case glutamate, originated. The drug LY2140023 acts at the glutamate auto-receptor. ²⁶

There were initial successes with both bitopertin and LY2140023. The early trials in schizophrenia appeared to work. In the case of bitopertin, a trial in over 300 patients suggested benefits against the negative symptoms of schizophrenia; the first trial of LY2140023 suggested that it was at least as effective as the powerful antipsychotic olanzapine.^27,28 However, the early hopes turned out to be a false dawn. Repeated failure to demonstrate efficacy led to the bitopertin and LY-023 programmes being stopped prematurely by the manufacturers. As a result, some authorities consider that the glutamate hypothesis of schizophrenia is in ‘existential crisis’. ²⁹

Nitric oxide

The glutamate story may yet have a lifeline. Although bitopertin and LY2140023 turned out to be unsuccessful, another approach involves the molecule sodium nitroprusside. Nitroprusside prevents the effects of NMDA channel blockers in animals, which links it to the glutamate story. ³⁰ Nitroprusside is a nitric oxide (or NO) donor. It breaks down into nitric oxide very quickly in the body. Nitroprusside is used in cardiovascular medicine as a treatment for severe hypertension. But nitric oxide is also a natural signalling molecule in the body, including in blood vessels. Its natural role is as a potent vasodilator that relaxes blood vessels, but it was subsequently found that NO is also a neurotransmitter in the brain – a very unusual transmitter, as we shall see below. ³¹

Based on the ability to reverse the effects of ketamine in animals, and with the ketamine model of psychosis in mind, researchers in Brazil looked at whether nitroprusside showed any benefits in schizophrenia. ³² What they found was that administration of nitroprusside led to a marked and very quick reduction in psychotic symptoms, over several hours. The study was small, looking at only 20 patients, but the difference between the active and placebo groups was impressive. ³³ Nitroprusside is delivered by intravenous injection, which is obviously a drawback, but the researchers claimed that a single injection could lead to improvement maintained over the next month. There are now further trials hoping to replicate these findings. One study was carried out in our own institution in 2015–2016. A similar method was used to the Brazilian trial, and again involved 20 patients. Our study did not, unfortunately, replicate the Brazilian trial, and we could not see any clear effect on schizophrenic psychosis. ³⁴ Our patient group had been unwell for a longer period of time, on average 14 years rather than less than 5 years, which may be an important cause of the discrepancy.

In the brain, nitric oxide is produced at glutamate synapses. The signal is calcium rushing through the NMDA channel. The enzyme that synthesizes nitric oxide and the NMDA channel are physically attached to each other, within the dendritic spine. The enzyme senses the Ca²⁺ influx through the channel pore and activates nitric oxide synthesis. ³⁵ The unusual property of nitric oxide is that it is a gas – a gaseous neurotransmitter. As a gas, it can diffuse in all directions after it is produced. Some of the nitric oxide will end up at nearby blood vessels and cause vasodilation, exactly as in the rest of the body. This is an important signal for the so-called BOLD response in functional MRI: active glutamate synapses signal their demand for blood-borne resource by releasing nitric oxide and increasing local blood flow. ³⁶

Nitric oxide also signals to nerve endings in the vicinity of the dendritic spine. This direction of information flow, from the post- to the presynaptic side, is termed retrograde transmission. Conventionally, information flow is from the pre- to the postsynaptic elements. Nitric oxide was the first retrograde signal to be discovered, in the late 1980s. ³¹ Others would be discovered, including endocannabinoids, the brain’s own cannabis system.

Cannabinoids

By the early 1990s the molecular biology underlying the effects of cannabis was elucidated. There are receptors on nerve terminals termed CB₁ cannabinoid receptors. This is where ∆ ⁹ -tetrahydrocannabinol (THC), the main psychoactive ingredient of cannabis, produces its effects. It was clear that if a receptor for a drug like THC exists in the body, there are likely to be natural cannabis molecules as well. These were soon discovered. The brain has a natural inbuilt endocannabinoid system. ³⁷

Since then it has become clear that endocannabinoids are crucially involved in plastic adaptation at synaptic connections. Endocannabinoids are produced in dendritic spines and signal to adjacent nerve terminals (retrograde transmission). They invoke a form of plasticity called long-term depression (LTD), whereby nerve terminals are instructed to downgrade their release of transmitter. ³⁸ (LTD can be thought of as the opposite process to LTP.)

It is common knowledge that cannabis can cause major mental illness. ³⁹ High-potency varieties (termed skunk in the UK) carry particular risk for psychosis and schizophrenia. Community studies in South London have shown that people who use skunk cannabis are especially likely to experience a psychotic illness, and the more they take, the higher the risk. ⁴⁰ Skunk is high in THC content, at the expense of other natural cannabinoids. One of these other molecules in cannabis is cannabidiol (CBD). ⁴¹ Remarkably, CBD appears to have antipsychotic properties. So the natural cannabis plant appears to contain ingredients which are pro-psychotic (THC), and also antipsychotic (CBD). There are concerns that the many Western markets are dominated by high-potency varieties such as skunk, which are unbalanced products devoid of CBD.

Laboratory experiments demonstrated that CBD could ameliorate the effects of THC. We carried out studies in which healthy volunteers came for two sessions. On one occasion they were given both CBD and THC; on another they received THC plus a placebo. Psychosis rating scales, as used in schizophrenia, showed that THC could increase psychosis scores and worsen memory performance. However, the addition of CBD reduced the pro-psychotic and amnestic effects of THC. ⁴² The laboratory findings were in agreement with the community studies, which showed that unbalanced products like skunk were having an adverse effect on mental health. ⁴⁰

If CBD could protect against the effects of THC, then perhaps CBD could be an antipsychotic in its own right, with benefits in endogenous psychotic illness. Early results from pioneering researchers in Brazil hinted that CBD could be effective in schizophrenia. A German group carried out a head-to-head trial in which CBD was compared against amisulpride. By 4 weeks the group randomized to CBD showed as much improvement as those randomized to amisulpride. ⁴³ Although the trial was small, these results were highly encouraging, especially as the side effects of CBD are regarded as minimal. Recently our group in London headed an international study involving a sample of 88 schizophrenic patients which found that 6-week treatment with CBD reduced positive psychotic symptoms (in-press).

It is not entirely clear how CBD acts on neurons. At the molecular level, it appears to inhibit the intracellular second messenger systems that are coupled to CB₁ receptors. But there may be additional mechanisms. ⁴⁴ If CBD turns out to be a useful antipsychotic in clinical practice, there is likely to be a major effort to understand the underlying mechanism of action.

Constellations

Future generations of psychiatric researchers will need to address the complexity of neural processing, rather than focusing on a single transmitter system. ⁴⁵ In nervous tissue, different transmitter systems work in concert, not in isolation (Figure 1). Take, for example, the local circuitry within the striatum. ⁴⁶ The striatum receives input from the whole of the cortical mantle, including parts of the cortex which process emotions and thoughts. The striatum is involved in organizing psychomotor programmes. ⁴⁷ The principal neuron in the striatum is the medium spiny neuron (MSSN). MSSNs are a vast population of over 200 million neurons. Each MSSN receives thousands of cortical inputs. These inputs use glutamate as a transmitter. Glutamate fibres from the cortex make synapses on the mushroom-shaped dendritic spines of MSSNs. The strength of the connection is reflected in the size of the dendritic spine. ⁴⁸

Figure 1.

At the small-world scale, different signalling systems work in concert. In the striatum, cortical fibres synapse on the head of dendritic spines. Dopamine-containing nerve varicosities regulate the strength of the corticostriatal connection. Nitric oxide and endocannabinoids are retrograde transmitters. Receptors for dopamine (D₂), glutamate (AMPA, NMDA) and endocannabinoids (CB₁) are shown.

In real time, the striatal network is believed to judge between competing calls from the cortex, and decide which particular programme will be selected and amplified.^47,49 That is to say, different thoughts, emotions or motor programmes are all competing with each other for attention. The striatum is responsible for the effortless and automatic selection of a particular stream of content. The network is arranged in a winner-takes-all pattern, in which the strongest call wins out. ⁴⁷ In addition, the striatum learns from experience. With time, certain programmes become stronger, and a repertoire of behaviours and thought patterns are ingrained into the circuitry. ⁴⁶

At corticostriatal connections, many of the components discussed above coalesce in the same micro-environment and work in concert (Figure 1). Individual connections can be strengthened (LTP) or weakened (LTD) as the network is ‘wired-up’ by experience. ⁵⁰ We can map the main components in turn: glutamate fibres from the cortex synapse on the head of dendritic spines of MSSNs. The spine makes and releases endocannabinoids – which go back by retrograde transmission – to communicate with the cortical input. And dopamine is released from adjacent varicosities at the neck of the dendritic spine. Dopamine has a vital role in regulating plasticity at individual corticostriatal connections. ⁵⁰

At a higher level of detail, dopamine promotes the weakening (LTD) of corticostriatal connections by acting at D₂ receptors. Correspondingly, a specific D₂ antipsychotic, such as sulpiride, inhibits the weakening (LTD) of corticostriatal connections.^51,52 Studies have shown that repeated treatment with dopamine D₂ antipsychotics increases the size of the striatum.^14,15 Here, we propose that dopamine D₂ antipsychotics invoke striatal enlargement by inhibiting LTD at corticostriatal connections.

The constellation of components – endocannabinoid, dopamine, glutamate and others – function in harmony under physiological conditions, so that ongoing psychomotor behaviour and any new learning are seamlessly integrated. Learning does not only have biochemical foundations, in terms of weakening and strengthening connections, but has structural correlates. New connections can form and enlarge, old connections wither away and the synaptic debris is cleared by the brain’s inbuilt scavenger cells, the microglia.^48,53

Drugs impacting on dopamine, glutamate, etc. impact upon structural plasticity. For example, the effect of antipsychotic drugs on the striatal network is so marked that it can actually be detected by brain MRI imaging. When a person has been taking antipsychotics for some time, the striatum actually enlarges in size. ⁵⁴ The network has expanded because of the antipsychotic drug. The change in volume is attributable to structural plasticity in the fine-grained structure (the neuropil) of the network.

Future landscapes

And yet, the above is still an oversimplification, and our models lag far behind the reality of molecular neuroscience. The next generation of psychiatrists will need to embrace additional complexity at the micro-scale level, far more sophisticated than the idea of too much transmitter or too little receptor, which still characterizes much of present thinking. Essentially this means learning about the multitude of components involved in synaptic health and plasticity.

We can sketch some of the main themes (Figure 2). Network health is vital for mental health. The stabilization of essential connections, the formation of new connections and the controlled elimination of redundant connections involves many components. There are components that span the gap between nerve terminals and dendritic spines to ensure that connections remain tightly bound. There are signalling pathways that control the dynamic, flexible actin scaffold that give terminals and spines their anatomical structure. ⁴⁸ There is ready-to-hand, protein-synthesis machinery for making additional components as learning proceeds. Finally, and most recently explored, there are mechanisms for ‘clearing up’ the debris when connections are no longer required. Such components (microglia, complement proteins) are much more familiar in their role as immune cells and immune signals, but their role extends beyond inflammation – microglia and complement proteins are now recognized as key components in the wiring of the brain as it learns and develops. ⁵³

Figure 2.

Structural plasticity at synapses involves change in the number, shape and size of dendritic spines. Protein scaffolds tether the pre- and postsynaptic elements. Actin fibres give the dendritic spine flexibility. Protein-synthesis machinery is present locally for making new components. Finally, microglia recognize and are activated by ‘eat-me’ signals (complement proteins) to clear redundant connections. Pathology in the various components can cause neuropsychiatric syndromes such as autism, learning disability and neurodegenerative disease.

Where those components involved in the function and structure of synaptic connections are defective, psychiatric illness can result.^55,56 Mutations in the components that bind the nerve terminal and dendritic spine are a cause of autism. The cause of many learning disability cases, hitherto unknown, are mutations in proteins which control the flexible actin scaffold that provides the neuropil with internal structure. The psychiatric manifestations of Fragile X syndrome (intellectual deficits/autistic features/hyperactivity) result from abnormal protein synthesis in dendritic spines and subsequent abnormal local wiring.

The latest components to receive attention, as pertains to psychiatric illness, are the microglia and their signalling pathways, specifically complement proteins. Complement proteins function as a tag, essentially an ‘eat-me’ signal, on synapses destined for elimination. ⁵⁷ The tag is recognized by the phagocytic microglia which engulf and clear the redundant synaptic elements. A major development in Alzheimer’s research has been the recognition of upregulated complement proteins and microglial phagocytosis commensurate with the loss of neuronal connections. The crucial observation is that such changes occur prior to amyloid deposition and tangle formation. ⁵⁸ Alzheimer’s appears to be a disorder of runaway synaptic loss. Schizophrenia, albeit to a far lesser extent than Alzheimer’s, is regarded as a disorder of impoverished connectivity. Variation in the complement C4 gene is strongly associated with schizophrenia and excessive synaptic pruning. ⁵⁹

There have been efforts to target the microglia in schizophrenia. One drug that has showed promise is minocycline. ⁶⁰ Early reports have suggested that minocycline may be effective for schizophrenia, although more convincing data are needed. ⁶¹ But the search will continue. Our current drugs for schizophrenia are reasonably effective, but as our knowledge of the complex dynamics at the neuropil level expands further, there will be new drug targets.

A more basic question goes back to the very roots of modern psychiatry. The question is whether, for some, the neuronal networks are destined to be unwell from the outset (endogenous psychiatric illness), or if, for others, adverse circumstances during development cause the network to wire-up pathologically (exogenous psychiatric illness). Whatever the proximal cause(s), endogenous, exogenous or a combination of both, major psychiatric syndromes appear to stem from abnormal connectivity within neuronal networks. Ultimately, we will become better at ensuring network health, and in steering networks along a trajectory which is commensurate with quality of life and wellbeing.