Is autoimmune psychosis a thing?

The search for pathological antibodies

Berlin zoo’s polar bear Knut, who died after developing anti-NMDA receptor autoimmune encephalitis

When I first arrived at the Maudsley hospital in 2012, a consultant psychiatrist (now professor) said that in the next few years neurologists would assess each new case of psychosis. Only after they had examined and investigated the patient for reversible causes, would they then pass on management to psychiatry. While the ongoing shortage of UK neurologists precludes this scenario for the foreseeable future, I could see his logic.

The same year, a memoir was released by the journalist Susannah Cahalan about her experience suffering from anti-NMDA receptor autoimmune encephalitis, aptly titled, Brain on Fire: My month of madness. Cahalan initially developed mood swings, anxiety, irritability and paranoia. She struggled with concentration and memory at work. She went on to have auditory hallucinations, persecutory delusions and seizures. First misdiagnosed with a psychiatric disorder or even alcohol withdrawal, testing her spinal fluid revealed the true problem - antibodies against her brain’s NMDA receptors.

Luckily for Cahalan, her astute neurologist Dr Souhel Najjar was familiar with this recently discovered disease and she makes a good recovery with immunotherapy to dampen the inflammation and plasma exchange to remove the harmful antibodies. Cahalan contemplates the alternative, remaining diagnosed with a mental rather neurological illness:

‘…how many people currently are in psychiatric wards and nursing homes denied the relatively simple cure of steroids, plasma exchange, more intense immunotherapy…?’

Another patient, Bex, is featured in BBC Radio 4’s Medical Mysteries. She is a university student who experiences acute anxiety and self-consciousness, along with concentration difficulties and memory lapses. She later develops headaches and is admitted to hospital after having a seizure. In hospital, her condition deteriorates: she loses the ability to speak and hears voices. The medical team initially think she is experiencing an episode of psychosis. Fortunately, working at the hospital was a world-leading expert in autoimmune neurology, Sarosh Irani, who makes the diagnosis of anti-NMDA receptor encephalitis. Bex receives appropriate immunosuppressive treatment and plasma exchange, eventually making a good recovery.

What if patients with this form of encephalitis are not under the care of an eminent neurologist, might they continue to be misdiagnosed with a mental disorder? Sadly, such cases have also been described. Lucy Dawson was detained under the Mental Health Act and treated with ECT, before a diagnosis of autoimmune encephalitis was made. An international rugby player Abi Burton similarly was being treated on a psychiatric ward before a clinician made the connection between her seizures and autoimmune encephalitis.

There are even reports of patients with chronic psychotic disorders like schizophrenia having an autoimmune cause - for example going into spontaneous remission when being treated with immunosuppression, even with no speicific autoantibody identified. Whether there is a type of disorder, autoimmune psychosis, caused by pathological autoantibodies but without neurological symptoms (seizures, cognitive deficits, autonomic dysfunction) is an open question with two possibilities: either we are missing a proportion of reversible psychosis cases or we are talking about a condition that doesn’t actually exist.¹

Autoimmune encephalitis

Remarkably, anti-NMDA receptor encephalitis didn’t exist when I started medical school; the syndrome hadn’t been defined and the antibody hadn’t been discovered. The first description of the phenotype was published as a case series of four women in 2005, by a team led by Josep Dalmau (then of Penn, now in Barcelona). Each case was a young woman with a specific type of ovarian tumour, called a teratoma. They all developed acute behavioural change and psychiatric symptoms (often including hallucinations) followed by neurological deterioration (seizures or reduced consciousness). One patient died, the others made a full recovery.

A subsequent 2007 paper from Dalmau of 12 women confirmed the illness course of psychiatric symptoms→memory problems→seizures/movement disorders→reduced consciousness. Half of the patients were first assessed or admitted under psychiatry before the correct diagnosis was made. All bar one developed seizures and all bar two required mechanical ventilation. All patients had tumours and these were usually ovarian teratomas. This time, they found that all patients had antibodies against the NMDA receptor; a new disorder had been discovered. Before this antibody was identified, were cases which fitted the phenotype simply undiagnosed? It certainly seems that way. In 1997 The Lancet published a case report, An acutely confused 15-year-old girl, about a patient with memory impairment and psychosis, who was found to have an ovarian teratoma. She did not respond to antipsychotics but recovered when the tumour was removed. This was almost certainly an early description of the undiscovered disorder.

Over the past 20 years, this new autoimmune disease has been intensively researched and much of its initial characterisation holds up. It has an incidence of 1.5 per million people per year and is around four times more common in women. It is a disorder of young people, with an average age of onset at 21 years. Around a third of patients have an underlying tumour, which is usually an ovarian teratoma, explaining the marked female:male predominance. Most patients respond to immunotherapy and tumour removal, though recovery can be protracted and the mortality is around 7%.

The initial presentation is usually psychiatric which over a period of a couple of weeks progresses to movement disorders, seizures, reduced respiration and autonomic dysfunction (changes in blood pressure, temperature and sweating). A seminal paper led by Adam Al-Diwani synthesised the psychopathology across cases, shown below. The most common individual symptoms were agitation and hallucinations. Approximately 2/3 had documented psychotic symptoms, almost half had mood changes and 1/3 had catatonic symptoms.

Reproduced under CC-BY from Al-Diwani et al. 2019

Knowing the symptoms and clinical course of this disorder is important. Brain MRI is often normal (another reason not to do these routinely), so clinicians have to go on clinical suspicion to arrange the crucial diagnostic tests of EEG and lumbar puncture.

Most now recognise anti-NMDA receptor encephalitis as a rare but important secondary cause of psychotic symptoms. However, does a forme fruste exist in which antibodies to the NMDA receptor cause psychotic in the absence of typical neurological progression? Before we delve into contentious issue of autoimmune psychosis, it might be worthwhile thinking about the autoimmune pathological process.

Attack of the clones

Autoimmune encephalitis literally means brain inflammation caused by the immune system mistakingly targeting the body’s own proteins, rather than foreign pathogens. How this happens and why it results in psychiatric and neurological symptoms can be thought about within the framework of antigen (NMDA receptor), antibody and phenotype (clinical presentation).

The human NMDA receptor, coloured by subunits

The NMDA receptor sits across the membrane of neurones and is activated by the neurotransmitter glutamate. It is important for memory and learning (i.e. neuroplasticity). The types of drugs that block NMDA receptors (like ketamine and PCP) can induce psychosis-like states and have been used as models for schizophrenia. This led to the glutamate hypothesis which proposes reduced function of the NMDA receptor as a key mechanism, upstream to the most established theory of psychosis dopamine dysregulation. It is not surprising then, that attacking this receptor could produce psychotic symptoms - the next question is why the immune system would target it in error.

Ovarian teratoma - literally ‘monster-tumour’

The immune system’s job is to protect the body from harmful invaders. A major component is the adaptive immune system, adaptive because it learns from experience. After a first encounter with a pathogen, it builds a memory that allows for a faster, stronger response next time. Antibodies are central to this process: when B cells recognise a specific antigen, they undergo clonal expansion, multiplying into an army of antibody-producing cells. On re-exposure, this force can mount a rapid and highly targeted defence.

A feature of the immune system is its ability to distinguish between self and non-self: this is immune tolerance. When this ability goes awry, the immune system mounts a response against the body, leading to autoimmune disease. Generally, the brain is thought as immune privileged meaning that it is less exposed to the parts of the immune system which recognise antigens and mount a response. When the immune system is presented with neural tissue (which can form part of ovarian teratomas), it may recognise it as ‘foreign’. If the tissue expresses NMDA receptors, we can end up with a clone army of B cells producing anti-NMDA receptor antibodies (not good).²

These circulating antibodies then bind to parts of the receptor, causing it to be internalised into the neuron. Essentially, this results in reduced numbers of the NMDA receptors on cell membranes, sometimes called receptor hypofunction. It disrupts the receptors role in memory and learning, causing amnesia as well as psychiatric symptoms. It also disturbs the brains excitation/inhibition balance, resulting in seizures. In this way, the immune response, mounting an antibody attack against its own brain’s NMDA receptors results in the classic encephalitis phenotype.

Autoimmune psychosis

I hope everything I’ve said so far is uncontroversial. Next we’ll wade into murkier waters. Firstly, can cases of confirmed anti-NMDA receptor encephalitis present with an isolated psychiatric syndrome (i.e. without neurological features)? One study from Dalmau’s group examined 571 patients with a confirmed diagnosis. They found 23 patients had episodes with a purely psychiatric phenotype. However, 18 had previous neurological symptoms in an early episode. There were only 5 (<1%) who presented for the first time in the absence of neurological symptoms. All these patients had antibodies to the NMDA receptor in their cerebrospinal fluid, rather than just in their blood.

One percent of a rare disease is a small number. If we go for the quoted incidence of 1.5 per million people per year, we might expected 100 cases of anti-NMDA receptor encephalitis in the UK each year, with perhaps one of them presenting with an isolated psychiatric syndrome. Given the possibility of misclassifying or overlooking symptoms, this number begins to look like a rounding error.

Of course, a reason that the proportion is so low could be because only patients with seizures and gross cognitive impairment get tested for anti-NMDA antibodies. Perhaps testing everyone with psychosis for these antibodies, might identify a number with antibodies who would never have been thought of as having encephalitis?

A team led by Belinda Lennox in Oxford (COI - my head of department) screened 228 patients with first episode psychosis for anti-NMDA receptor and other neuronal autoantibodies. None of the patients had signs of encephalitis. Seven patients (3%) were positive for antibodies to the NMDA receptor while thirteen were positive for other neuronal autoantibodies. While 3% may seem like a small proportion, approximately 10,000-20,000 patients are diagnosed with a psychotic disorder in the UK each year, meaning up to 500 of these could be positive for such antibodies. However, patients positive for NMDA receptor antibodies did not differ from those who were negative in terms of clinical presentations or outcomes. None of the antibody-positive cases developed neurological symptoms.

Crucially, these antibodies were measured from blood samples rather than cerebrospinal fluid so we don’t know whether they could bind to receptors within the brain. Curiously too, a proportion of healthy controls had neuronal autoantibodies, though to other targets than the NMDA receptor. The proportion was similar (4%) in a study of 387 patients with first episode psychosis from Substack’s own Tom Pollak, though there were no differences in symptomatology, treatment response or outcome. Another study from Dalmau’s team recruiting 105 patients with first episode psychosis (who had undergone lumbar puncture) did not identify NMDA receptor antibody in blood or cerebrospinal fluid, except in patients who additionally had the diagnosis of autoimmune encephalitis.

So where do we stand? We can say that a small proportion of patients with first episode psychosis have antibodies to the NMDA receptor from blood samples, at a higher rate than is found in healthy controls. However, lack of cerebrospinal fluid samples and lack of a clinical phenotype specific to antibody-positive patients, calls into question whether these antibodies are pathological. That the antibodies are found in healthy control blood samples suggests they do not necessarily cause a clinical syndrome.

In my mind there are two ways to resolve whether the autoantibodies found in first episode psychosis patients are indeed pathological. The first is to screen large numbers of patients using lumbar puncture to obtain cerebrospinal fluid. This has been advocated by Belinda and Tom but it would require a change in clinical practice and is a logistical challenge. Most patients I see who are acutely psychotic would lack decision-making capacity for a somewhat invasive procedure with uncertain benefits. The second is to do what we always do to resolve clinical equipoise, a randomised trial.

SINAPPS

The Study of Immunology in Antibody Positive Psychosis (SINAPPS) is a randomised placebo-controlled trial of immunotherapy (intravenous immunoglobulin followed by rituximab) as an add-on treatment for patients with acute psychosis who test positive for anti-neuronal antibodies. Importantly the trial isn’t enrolling patients with suspected encephalitis (seizures are an exclusion criterion) but those with an isolated psychiatric syndrome. The inclusion criteria allows for patients who test positive based on blood or cerebrospinal fluid samples, though from the protocol it seems that blood serum will be the main source.

I say this a lot: clinical trials are hard. They are difficult to setup due to ever increasing bureaucracy, which contributes to them being expensive to run. Recruitment, particularly from people experiencing severe episodes of mental illness, is tough. Even as clinical trials go, SINAPPS is more challenging than most. By focussing on the <5% of psychosis patients who are antibody-positive the study needs to enrol a much greater number (thousands) of patients for antibody screening. Only those who test positive will be offered the 50/50 chance of receiving a treatment which, while standard for encephalitis, has some serious side-effects and requires intravenous infusion.

The study (which I should say is partly run out of my department, though I have no inside knowledge) has so far screened almost 6k people with psychosis from more than 40 health services in the UK. Of the n=80 target sample size, 57 antibody-positive participants have been randomised since it launched in 2017 and its projected end date is 2027.

By now I think you all know that I love randomised controlled trials and this one is a good example of why. It is hard, it is long, it is expensive, it is ambitious. It is attempting to definitively answer whether patients with psychosis associated with autoantibodies but without features of encephalitis respond to immunotherapy. If they do, a new treatment paradigm in psychiatry will be unlocked. If the trial is conclusively negative, then we will have disappointing but ultimately important information about the clinical relevance of these antibodies.

So, does autoimmune psychosis exist? You’ll have to wait and see.

1. Any good ideas in this post were stolen from a talk I recently attended by Adam Al-Diwani, anything I got wrong was probably stolen from someone else

2. As a brief aside, it is a bit of a mystery as to why patients without such tumours end up with antibodies targeting their own NMDA receptors. Recent work by Adam suggests this might be through antigens draining from cerebrospinal fluid into cervical lymph nodes