Beyond dopamine: glutamate, inflammation and other theories

Until 2024, every medicine approved for schizophrenia worked by blocking the dopamine D2 receptor (opens in a new tab).^8,9 The reasoning behind them goes back to 1976, when researchers showed that the doses of antipsychotics (opens in a new tab) used in the clinic closely matched how strongly each drug attaches to dopamine receptors: the stronger the attachment, the lower the dose needed.¹⁰ Drugs that release dopamine pointed the same way. At doses that leave healthy people unaffected, stimulants such as amphetamine and methylphenidate brought on or worsened psychotic symptoms in about 40 in 100 people with schizophrenia.¹¹

Brain scans later located the fault more precisely. Across 44 scanning studies, which compared 618 people with schizophrenia with 606 healthy volunteers, the largest difference lay before the receptors, in the nerve endings of the striatum, a region deep in the brain: these endings make and release more dopamine. The receptors themselves were barely changed. Antipsychotics therefore act one step after the abnormality: they block the receptor that receives the excess signal, not the excess production itself.⁹

In 2009, Oliver Howes and Kapur drew these strands together. Many risk factors for schizophrenia, from genes and complications in pregnancy to stress, trauma and drug use, may converge on excess dopamine in the striatum, which they called the final common pathway to psychosis. If so, current medicines treat the last link of a longer chain, and new ones should aim at the links before it.¹³

What dopamine does not explain

Schizophrenia is more than psychosis. Low motivation, reduced emotional expression and social withdrawal, known as negative symptoms (opens in a new tab), respond poorly to treatment: across 168 placebo-controlled trials, no medicine, brain stimulation or psychological therapy improved them enough to make a clinically meaningful difference.² Problems with memory, attention and planning account for much of the disability the illness brings, and existing treatments do not relieve them.³

Medicines also fail some people outright. Between a quarter and a third of people with schizophrenia do not improve enough on antipsychotics, a condition called treatment-resistant schizophrenia.¹ A brain-scan study in London suggests one reason. In 12 people who had responded to antipsychotics, dopamine production in the striatum was raised, as the theory predicts; in 12 people whose illness had not responded, it was no higher than in 12 healthy volunteers. The authors suggest that the illness of these patients may rest on a different biology, or that treatment affected them differently.¹⁴

A later study found more glutamate in the anterior cingulate cortex, at the front of the brain, in 21 people with treatment-resistant schizophrenia than in 20 who had responded to treatment.¹⁵ Clozapine (opens in a new tab), the one medicine approved for treatment-resistant schizophrenia, blocks D2 only weakly and acts more strongly on several other receptors.¹⁶

Genetic studies reveal a third gap. Few of the genetic variants linked to schizophrenia (opens in a new tab) act directly on dopamine, which suggests that the excess of dopamine arises mainly from other causes.⁶

Glutamate and the NMDA receptor

Glutamate (opens in a new tab) is the brain’s main excitatory messenger: most nerve cells use it to activate the cells they connect to.⁶ The case for its role in schizophrenia began with two anaesthetics, phencyclidine (PCP, sold illegally as “angel dust”) and ketamine, which both block one of its receptors, the NMDA receptor (opens in a new tab). Amphetamine causes mainly hallucinations and paranoia; PCP also brings on emotional withdrawal, slowed movement, disorganised thinking and the difficulties with attention and memory seen in schizophrenia.¹⁷ In 1994, researchers gave 19 healthy volunteers ketamine through a drip, at doses below those used for anaesthesia. It produced behaviour resembling both the positive and the negative symptoms of schizophrenia, altered perception, and impaired attention, verbal fluency and the recall of words after a delay.⁴

How a weakened glutamate receptor could lead to excess dopamine has a proposed answer. NMDA receptors drive fast-firing inhibitory nerve cells, which keep their neighbours in check. When these receptors work poorly, the inhibition weakens and the neighbouring cells become overactive, notably in the hippocampus, a region central to memory; that overactivity can in turn drive the excess of dopamine that produces psychosis.¹⁸ The inhibitory cells use GABA, the brain’s main dampening messenger. In brain tissue from people with schizophrenia, the inhibitory cells that contain a protein called parvalbumin make less GABA in the prefrontal cortex. This may weaken the fast, synchronised rhythm of activity on which working memory depends.¹⁹

Glutamate is harder to measure in the living brain than dopamine. Magnetic resonance spectroscopy, a form of MRI that measures chemicals rather than structures, has been used in 59 studies of 1,686 patients. Together, they found raised levels of glutamate and related compounds in the basal ganglia, the thalamus and the medial temporal lobe, and lowered levels in no region.²⁰

Genetics points in the same direction. In 2014, a study of 36,989 people with schizophrenia linked 108 regions of DNA to the illness; they included the gene for the D2 receptor and several genes involved in glutamate signalling.²¹ The largest study so far, of 76,755 people with schizophrenia and 243,649 without, linked 287 regions. The genes involved are active mainly in nerve cells and govern how synapses (opens in a new tab), the junctions between nerve cells, are organised and pass on signals. Among them is GRIN2A, which carries the instructions for part of the NMDA receptor, and in which rare disruptive mutations are also linked to schizophrenia.⁵

Medicines that strengthen the NMDA receptor have not yet succeeded. The receptor needs glycine, a small amino acid, alongside glutamate to work; bitopertin and iclepertin (opens in a new tab) block the pump that clears glycine away, so that more of it remains near the receptor.⁶ Added to antipsychotics for 24 weeks in two trials of about 600 people each, bitopertin did not improve negative symptoms more than placebo.²² Iclepertin was tested against problems with memory and attention in 1,835 people at 338 centres in 41 countries; after 26 weeks, their scores on thinking tests did not differ from those on placebo.⁷

Antibodies against the NMDA receptor

A rare illness offers the most direct evidence that a fault at the NMDA receptor can cause psychosis. In anti-NMDA receptor encephalitis, the immune system produces antibodies against the receptor. Among the first 100 patients described in detail, mostly young women with a median age of 23, every one began with psychiatric symptoms or memory problems. Most went on to develop seizures, reduced consciousness, involuntary movements, shallow breathing, and an unstable pulse and blood pressure. In the laboratory, the patients’ antibodies reduced the number of NMDA receptors on nerve cells, and the effect reversed when the antibodies were removed. A tumour, most often an ovarian teratoma, was found in 58 of the 98 patients checked. Of the 100, 75 recovered or were left with mild problems; those whose tumour was treated early fared better.²³

Psychosis without any neurological signs is uncommon in this disease: 23 of 571 patients, or 4 in 100, had episodes of purely psychiatric symptoms, at its onset or at relapse, and most recovered with treatment of the immune system.²⁴ In England, 7 of 228 people seen for a first episode of psychosis had antibodies against the NMDA receptor in their blood, against none of 105 healthy volunteers; their symptoms did not differ from those of patients without antibodies.²⁵ Whether such patients form a distinct group, sometimes called autoimmune psychosis, is still debated. In 2020, an international group of psychiatrists and neurologists proposed how to recognise it, which tests to run and how to treat it.²⁶

Inflammation and the immune system

A wider line of research examines the immune system as a whole. Denmark’s national registers, linked over 30 years, showed that a previous autoimmune disease raised the risk of later schizophrenia by 29%, a hospital stay for any infection by 60%, and the two together more than doubled it, to 2.25 times. The risk rose with the number of infections: three or more, together with an autoimmune disease, raised it 3.4 times.²⁷ Infection during pregnancy has been linked to schizophrenia in the child, and in animals, activating the mother’s immune system alters the brain and behaviour of the offspring.²⁸ Because schizophrenia affects fewer than 1 in 100 people, even a doubled risk remains small for any individual.²⁹

In the blood, people with schizophrenia have on average more of the inflammatory signalling proteins interleukin-6 and tumour necrosis factor during an acute episode, and interleukin-6 falls with treatment. The same pattern appears in bipolar disorder and depression, so it is not specific to schizophrenia.³⁰ C-reactive protein (CRP), the marker of inflammation measured in everyday practice, is moderately raised from the first episode onwards, whether or not people take antipsychotics. It is higher in people with a higher body weight and with more severe hallucinations and delusions.³¹

Inside the brain, the evidence is weaker. If the brain’s own immune cells were overactive, scans of a protein they produce should show more of it. Pooled data from five studies, with 75 patients and 77 healthy volunteers, showed less, which may mean that these cells work differently in schizophrenia or are fewer.³²

Treatments aimed at inflammation have so far disappointed. The antibiotic minocycline (opens in a new tab), which also dampens inflammation, was given for a year to 207 people in the UK who had developed a schizophrenia-spectrum illness within the previous five years, in addition to their usual treatment. It did not improve negative symptoms, and the trial found no sign of an inflammatory process for it to treat.³³

Across 70 trials of add-on treatments with possible anti-inflammatory effects, from anti-inflammatory painkillers and minocycline to oestrogens, statins and fatty acids, symptoms improved by an average of 4.6 points more than with placebo on the PANSS (opens in a new tab), a scale that runs from 30 to 210. Medicines designed specifically to suppress inflammation did no better than the rest, and many of the trials were small, which tends to inflate the apparent benefit.³⁴

The Danish researchers read their findings as consistent with an immune cause in a subgroup of patients, not in all.²⁷ The minocycline team reached a similar conclusion from the other side: further trials should wait until an inflammatory process can be shown in the people being treated.³³

Brain development and synapses

In 1987, Daniel Weinberger in the United States, and Robin Murray and Shôn Lewis in London, independently proposed that schizophrenia begins long before its symptoms. In their view, a fault laid down early in brain development stays silent until the brain matures in adolescence and early adulthood, when the illness usually appears.^35,36

On average, children who later develop schizophrenia differ slightly from their peers. In a British cohort of 5,362 people born in one week of March 1946, the 30 who developed schizophrenia as adults had learned to walk about a month later, had more speech problems, scored lower on school tests at the ages of 8, 11 and 15, and at 4 and 6 more often preferred to play alone.²⁹ These are differences between groups: most children who walk late or prefer to play alone never develop schizophrenia.

Adolescence brings a profound reorganisation of the brain: deep sleep and the brain’s energy use decline, and the cortex loses synapses. In 1982, the psychiatrist Irwin Feinberg proposed that a fault in this pruning might underlie the schizophrenia that begins in adolescence.³⁷ Brain tissue supports the idea of fewer connections. In the prefrontal cortex of people with schizophrenia, nerve cells carried 23% fewer dendritic spines, the small protrusions that receive synapses, than in people without mental illness. People with other psychiatric illnesses who had taken antipsychotics showed no such loss, which makes the medicine an unlikely cause.³⁸ Brain scans can now estimate synapses in living people: in 18 people with schizophrenia, a protein found in synapses was reduced in the frontal cortex compared with 18 healthy volunteers, and antipsychotics given to rats did not reduce it.³⁹

Genetics connects pruning to the immune system. The region of DNA most strongly tied to schizophrenia contains immune genes, and in 2016 researchers traced part of that link to complement component 4 (C4), an immune protein that tags material for removal. Versions of the gene that produce more of one form, C4A, in the brain carried more risk, in proportion to how much they produced. C4 was found at synapses in the human brain, and in mice it helped to remove synapses during development. Excessive C4 activity may therefore help explain why the brains of people with schizophrenia have fewer synapses.⁴⁰

Acetylcholine and the first new medicine

Acetylcholine (opens in a new tab) acts in the brain through muscarinic receptors (opens in a new tab), two of which, M1 and M4, influence dopamine signalling indirectly.⁸ In brain tissue from 80 people with schizophrenia, about a quarter (26 in 100) had 74% fewer muscarinic receptors in the prefrontal cortex than people without the illness. The authors proposed that these people form a distinct subgroup, which they called muscarinic receptor-deficit schizophrenia.⁴¹

The medicine that came from this line of research was first tried in another illness. In the 1990s, xanomeline, which stimulates M1 and M4 receptors, was tested in 343 older people with Alzheimer’s disease. The higher the dose, the more it reduced suspiciousness, delusions, agitation and hallucinations; but more than half of those on the highest dose stopped taking it because of side effects, mostly nausea and other digestive complaints, and 13 in 100 fainted.⁴²

What made it usable was a second drug: trospium blocks muscarinic receptors in the rest of the body but barely enters the brain, so it eases those side effects without cancelling xanomeline’s effect. In a 5-week trial of 252 adults in hospital with acute psychosis, the combination improved symptoms by 9.6 points more than placebo on the PANSS, and weight gain and movement side effects were no more common than with placebo.⁸ It was approved in the United States in September 2024, as Cobenfy (opens in a new tab); as of November 2026, it is not approved in the European Union.^43,44

A new target does not guarantee a new medicine. Emraclidine (opens in a new tab), which acts selectively on the M4 receptor, did not improve symptoms more than placebo in two trials reported in November 2024.⁴⁵ Ulotaront (opens in a new tab), which acts on two other receptors, TAAR1 and the serotonin 5-HT1A receptor, failed in two phase 3 trials of 435 and 464 people in 2023; its makers attributed the result partly to an unusually large improvement on placebo.⁴⁶

How the theories connect

The theories connect more than they compete. In the circuit model, a fault at the NMDA receptor weakens inhibition and drives the excess of dopamine.¹⁸ The C4 gene ties the immune system to the loss of synapses.⁴⁰ Genetic and environmental risks appear to disturb both glutamate and dopamine, with dopamine as the final step to psychosis.^6,13

Schizophrenia may also be several illnesses that share a name. Each biological subgroup found so far covers only part of the people with the diagnosis: those whose dopamine production is normal and who do not respond to antipsychotics,¹⁴ the quarter with a marked loss of muscarinic receptors,⁴¹ and the few with antibodies against the NMDA receptor.²⁵ If that is the case, the next medicines may help some people considerably and others not at all, and the task becomes telling the groups apart before treatment begins, for example with scans that measure glutamate.¹⁵

Where each theory stands, as of November 2026:

What we don’t know yet

• Whether schizophrenia is one illness with several mechanisms, or several illnesses with a shared outcome.
• How to tell, before treatment, which mechanism drives a person’s illness.
• Whether medicines aimed at glutamate or inflammation would work if given only to people in whom those systems are disturbed.