Psychosis: an autoimmune disease?

Summary

Psychotic disorders are common and disabling. Overlaps in clinical course in addition to epidemiological and genetic associations raise the possibility that autoimmune mechanisms may underlie some psychoses, potentially offering novel therapeutic approaches. Several immune loci including the major histocompatibility complex and B‐cell markers CD19 and CD20 achieve genome‐wide significance in schizophrenia. Emerging evidence suggests a potential role via neurodevelopment in addition to classical immune pathways. Additionally, lymphocyte biology is increasingly investigated. Some reports note raised peripheral CD19⁺ and reduced CD3⁺ lymphocyte counts, with altered CD4 : CD8 ratios in acute psychosis. Also, post‐mortem studies have found CD3⁺ and CD20⁺ lymphocyte infiltration in brain regions that are of functional relevance to psychosis. More specifically, the recent paradigm of neuronal surface antibody‐mediated (NSAb) central nervous system disease provides an antigen‐specific model linking adaptive autoimmunity to psychopathology. NSAbs bind extracellular epitopes of signalling molecules that are classically implicated in psychosis such as NMDA and GABA receptors. This interaction may cause circuit dysfunction leading to psychosis among other neurological features in patients with autoimmune encephalitis. The detection of these cases is crucial as autoimmune encephalitis is ameliorated by commonly available immunotherapies. Meanwhile, the prevalence and relevance of these antibodies in people with isolated psychotic disorders is an area of emerging scientific and clinical interest. Collaborative efforts to achieve larger sample sizes, comparison of assay platforms, and placebo‐controlled randomized clinical trials are now needed to establish an autoimmune contribution to psychosis.

Abbreviations

CNS

central nervous system

CSF

cerebrospinal fluid

GABAR

γ‐amino‐butyric acid receptor

GWAS

genome‐wide association screen

IL‐6

interleukin‐6

LGI1

leucine‐rich glioma inactivated 1

NMDAR

N‐methyl‐d‐aspartate receptor

NSAb

neuronal surface antibody

PPP

post‐partum psychosis

VGKC

voltage‐gated potassium channel

Introduction – turning against the self?

Psychotic disorders involve disturbances in thought, perception, cognition and social functioning. They are disabling and costly to the health and wealth of societies worldwide.1 Schizophrenia, the archetypal psychotic illness, is common (1% lifetime prevalence), begins in early adulthood (typical onset in the second to third decades), and may persist throughout life. Relapse and/or progression are familiar and challenging features.2, 3 Early intervention with medical and psycho‐social approaches can be effective, but a better understanding of these heterogeneous conditions is needed to improve treatments.

Mounting evidence underlines the systemic aspects of mental disorders.4 Notably, a greater understanding of immunological and nervous system interactions has challenged the previously held belief of central nervous system immune privilege,5, 6 with autoimmunity emerging as a potential disease mechanism. This tractable and translatable paradigm is beginning to rejuvenate the integrative biology of psychosis.

Here, by focusing on perturbations in adaptive immunity we present a synthesis of evidence for and against psychosis as an autoimmune disease. As in previous reviews of this area,7 we have organized this update around modified Witebsky's postulates of pathogenicity. These criteria – which define an autoimmune disease – include: (i) evidence of disease‐specific adaptive immune response in the affected target tissue, organ or blood; (ii) passive transfer of autoreactive cells or antibodies replicates the disease in experimental animals or in humans via in utero materno‐fetal transfer or transplant procedures; (iii) elimination of autoimmune response modifies disease; (iv) personal or family history of autoimmune disease, and/or MHC associations.8

Overlap between psychosis, autoimmune disorders and autoimmune risk factors

Phenotypic overlap

There is a range of courses seen in psychotic illnesses, ranging from a single episode of illness, through relapsing–remitting courses to progressive accrual of disability. This profile has a degree of overlap with the time–course of some central nervous system (CNS) inflammatory disorders, such as multiple sclerosis.

Psychosis is also recognized, either monosymptomatically or in association with other neuropsychiatric or systemic symptoms, as a consequence of other CNS autoimmune and inflammatory disorders, including systemic lupus erythematosus,9, 10 multiple sclerosis,11, 12 sarcoidosis,13, 14 and autoimmune vasculitides.15, 16, 17 The specific mechanisms underlying psychotic symptoms in these disorders remain largely unknown. An exception may be systemic lupus erythematosus, in which associations between psychosis and autoantibodies such as anti‐ribosomal P and anti‐N‐methyl‐d‐aspartate receptor (NMDAR) NR2 have been described, albeit inconsistently, but these do not have proven clinical value as biomarkers of disease activity in lupus psychosis. 18, 19

Epidemiological and genetic associations

Schizophrenia is at least in part a genetic disorder, with heritability estimated at 80%.20 At an epidemiological level, there is both an increased risk of other autoimmune disorders in those with psychosis and an increased risk of psychotic disorders in people with an autoimmune disorder (relative risk 1·4 for any autoimmune disorder, between 0·7 and 5·4 for specific disorders) and their relatives.21, 22, 23, 24 Interestingly the risk of subsequent diagnosis of psychotic disorder in individuals with autoimmune disorders was increased further still if those individuals had been hospitalized for a serious infection, pointing towards an interplay of autoimmune risk factors and infectious or inflammatory insults.23 The only autoimmune disease that consistently demonstrates a negative association with schizophrenia is rheumatoid arthritis, which has a risk > 29% of that in the general population.25 Reasons for the negative association are not fully understood but may in part reflect a small but robust negative single nucleotide polymorphism‐genetic correlation between the two disorders.26

Psychosis also has shared environmental risk factors with other autoimmune disorders, such as season‐of‐birth and latitude effects, which have been hypothetically related to neonatal vitamin D status.27

Many autoimmune diseases demonstrate associations with alleles in the MHC region located on chromosome 6. HLA association has emerged as the strongest and most consistent genome‐wide association study (GWAS) finding in schizophrenia.28 The paradox remains that despite this, efforts to identify the exact variants have been unsuccessful, largely hampered by the high degree of linkage dysequlibrium within the MHC region.29

One notable exception that has generated interest was the discovery that variation at the complement C4 locus, a gene within the MHC region, conferred schizophrenia risk in a dose‐dependent manner proportional to an allele's tendency to express C4A expression in brain. Furthermore, a separate series of animal studies demonstrated that C4 variation directly shapes microglia‐mediated synaptic pruning processes relevant to schizophrenia pathology.30 That variation within the MHC region on chromosome 6 could in fact contribute to schizophrenia risk through non‐HLA alleles, and through a mechanism ostensibly far‐removed from the classical antigen‐presentation functions attributed to that region, has led some researchers to begin to reconsider the role of the immune system in schizophrenia through a neurodevelopmental lens.

Beyond the HLA association, and statistically independent from it, the largest GWAS to date found that associations were enriched among genes expressed in B‐lymphocyte lineages (CD19 and CD20).28 Pathway analyses of schizophrenia GWAS data have yielded conflicting results: a 2010 study found that three of seven over‐represented pathways related to immune function,31 but in a more recent and larger study pathway analysis of GWAS data identified key immune processes such as T‐cell activation, B‐cell activation and transforming growth factor‐β signalling when schizophrenia, bipolar disorder and major depression were pooled, but no immune pathway associations for schizophrenia alone.32

Nonetheless, as was the case with HLA region candidate genes, immune‐related genome‐wide associations with schizophrenia remain less convincing than those demonstrated for established autoimmune diseases; in one study, no enrichment of immune loci outside the MHC region was found when schizophrenia was compared with five known autoimmune diseases (Crohn's disease, multiple sclerosis, psoriasis, ulcerative colitis and rheumatoid arthritis).33 Further, immune loci that were individually associated with schizophrenia all had alternative, non‐immune roles that could plausibly contribute to schizophrenia aetiopathology. A recent gene‐expression study of post‐mortem schizophrenia brain tissue has led authors to a similarly sceptical conclusion: the authors found surprisingly limited differential expression of immune genes in schizophrenia prenatal cortex, and none in hippocampus, compared with controls; in neither brain region was there differential expression of immune pathways. Further, of nine MHC genes that were differentially expressed in patients, only one had a putative immune function and its expression was decreased in schizophrenia.34

A final consideration regarding shared genetic vulnerability concerns the 22q11 deletion syndrome (di George syndrome, or velocardiofacial syndrome), which among structural variants is the strongest known risk factor for developing schizophrenia‐spectrum disorders in adulthood, with up to 40% of carriers developing such a disorder by their mid‐20s.35, 36 Carriers may also have varying degrees of thymic hypoplasia and T‐cell dysfunction, along with an impaired B‐cell and humoral immune response and a substantially increased rate of occurrence of autoimmune disorders.37 Notably, no authors have assessed psychosis risk in 22q11 deletion syndrome in relation to immunological dysfunction.

Relationship to infection

In the 1980s and 1990s epidemiological evidence that perinatal infections, particularly influenza, were more common in mothers of children who developed schizophrenia helped to steer the neurodevelopmental hypothesis of schizophrenia.38 Today, evidence that infection increases risk of psychotic disorders has been extended to include neonatal, infant and even adult life infections.39

The most frequently implicated organisms in schizophrenia include influenza A and B, toxoplasma, herpes simplex virus‐1, Epstein–Barr virus, cytomegalovirus, rubella and Borna disease virus. Studies have mainly sought to show a higher prevalence of antibodies (either IgG or IgM) to these organisms in patients versus controls, or have found associations between markers of infection and some phenotypic dimension within psychosis, such as cognitive ability40 or psychopathology.41 Relatively few studies have attempted to demonstrate the presence of viral, bacterial or protozoal organisms in the blood, much less the brain, of patients with psychotic disorders. Some authors have reported that schizophrenia‐associated polymorphisms may moderate the relationship between infection and clinical disease expression, possibly by increasing vulnerability to infection. For example, schizophrenia‐associated HLA single nucleotide polymorphisms increase the risk of herpes simplex virus seropositivity.42, 43

Evidence of disease‐specific adaptive immune response in the affected target tissue, organ or blood

Evidence from primary psychotic disorders

Putative disease subgroups, methodological differences, and confounding effects of antipsychotic medication, smoking and substance use are long‐standing challenges in studies of the CNS in psychosis. Nonetheless both meta‐analytic and prospective studies in medication‐free participants using serum, cerebrospinal fluid (CSF), post‐mortem, and neuroimaging data have implicated inflammation per se as a convergent mechanism in at least a proportion of cases. Adaptive immunity more specifically remains an under‐investigated area, but is starting to receive greater attention.

Analyses of CSF have largely focused on non‐disease‐specific cytokine and neurotransmitter metabolite assays, rather than lymphocytes and autoantibodies.44, 45 Nonetheless, efforts such as those by Endres et al., highlight the value of routine and systematic CSF collection in people with psychosis. In a cohort of 180 cases, of which 146 were cases with undifferentiated new‐onset acute psychosis, they found raised protein in 42·2%, raised albumin quotient implying blood–brain barrier disruption in 21·8%, raised white cell counts in 3·4% and oligoclonal bands in 7·2%.46 These findings are complemented by previous work by Nikkilä et al., who found CSF lymphocyte morphology consistent with activation among patients in their first‐episode or relapse receiving short periods or no medication compared with controls.47

Although there is an incomplete picture of lymphocyte biology in CSF, there have been many studies using peripheral blood in psychotic disorders. Elevated B‐cell populations have been described by several reports in acute unmedicated psychosis. For example, using flow cytometry Steiner et al. found higher CD19⁺ lymphocyte counts in an unmedicated group of patients with acute psychosis compared with healthy controls, which decreased after 6 weeks of antipsychotic treatment.48 This was similar to previous findings from Maino et al. who found elevated CD19⁺ and reduced CD3⁺ counts in unmedicated patients that reduced to healthy control levels during antipsychotic treatment.49 Additionally, a meta‐analysis by Miller et al. found reduced CD3⁺ lymphocyte levels and an increased CD4 : CD8 ratio in patients versus controls that reversed following anti‐psychotic medication, although generalization is limited by heterogeneity.50

These observations have been complemented by cytokine studies. There has been meta‐analytic evidence from schizophrenia populations that interleukin‐1 (IL‐1), IL‐6, and transforming growth factor‐β (TGF‐β) increase during acute psychosis and decrease during anti‐psychotic treatment whereas IL‐12, interferon‐γ (IFN‐γ), tumour necrosis factor‐α (TNF‐α) and soluble IL‐2 receptor remain elevated despite clinical status.51, 52 Meta‐analysis focusing specifically on anti‐psychotic‐naive cases of first episode psychosis identified elevations in similar cytokines: IL‐1β, soluble IL‐2 receptor, IL‐6 and TNF‐α. However, the generalizability of these conclusions is limited by important confounders known to influence cytokine levels such as age, sex, body mass and smoking, as well as clinically heterogeneous participant populations.

Post‐mortem brain histology is an established approach in schizophrenia research.53 Whereas microglia and macroglia have been frequently examined,54 data are sparse regarding lymphocytes. A recent meta‐analysis of post‐mortem immune‐related findings in schizophrenia by Van Kesteren et al.55 found that while there are sufficient data to conclude a significant increase in the density of microglia and pro‐inflammatory transcripts and proteins in cases compared with controls, only one study from Busse et al. looked at lymphocytes.56 This study of 17 cases of schizophrenia and 11 controls found CD3⁺ and CD20⁺ lymphocyte infiltration in the posterior hippocampus predominantly in those with a more chronic deficit‐type phenotype, ‘residual schizophrenia’ and higher HLA‐DR⁺ microglia in the more positively psychotic phenotype, ‘paranoid schizophrenia’. Subsequently, using semi‐quantitative methods, Bogerts et al. have added further evidence from CD3 and CD20 staining in brain sections from patients with schizophrenia and mood disorders compared with matched controls.57 Significant differences versus controls were found for both groups for medial temporal lobe CD20⁺ lymphocyte infiltration and for CD3⁺ lymphocytes in additional regions such as cortex and thalamus, however, suicide method and rate of death are potential confounders that could explain these observations rather than being due to an aspect of the illness itself.

Direct in vivo observations of immune‐involvement with spatial and longitudinal resolution before and after illness are highly desirable. Functional neuroimaging techniques such as positron emission tomography in principle offer this opportunity. There have been several attempts to use translocator binding protein‐specific ligands to assay microglial activation. For example, Bloomfield et al. found signal consistent with elevated microglial activity in people at ultra high risk for psychosis that correlated with symptom severity.58 However, this observation has not been replicated and other studies show no change or even a fall in this signal in first episode psychosis.59, 60, 61 Indeed work using animal models implies that translocator binding protein‐signal may be determined by multiple cell types beyond microglia.62 Nonetheless, this work shows, in general terms, the potential of molecular neuroimaging to evaluate immune hypotheses. Application and further development of lymphocyte‐specific tracers and non‐tracer‐dependent magnetic resonance methods should be a focus, for example conjugated antibody and cell positron emission tomography tracers developed to monitor cancer and rheumatology immunotherapeutics may be translatable.63, 64

These findings add some weight to the concept that schizophrenia and other psychotic disorders may be the result of chronic ‘low‐grade’ neuroinflammation, a proposal presciently suggested by Bechter in his ‘mild encephalitis’ hypothesis of schizophrenia.65 The relatively recent description of autoimmune encephalitides caused by antibodies to neuronal surface antigens provides a unique opportunity to explore an antigen‐specific variation of this hypothesis. These neuronal surface antibody (NSAb)‐mediated encephalopathies fulfill Witebsky's postulates more classically and offer plausible direct mechanisms linking humoral autoimmunity to neural circuit dysfunction and psychosis.

Autoimmune encephalitis as a psychosis disease model

The idea that people with schizophrenia and other psychoses could have brain‐reactive autoantibodies of pathological relevance has a long history.66, 67, 68, 69 However, after several promising but unreplicated findings, the development of assays that reliably detect conformationally specific autoantibodies is providing robust experimental data to test previous hypotheses and develop new models.70

The principle that autoantibodies can target specific proteins crucial to neurotransmission, directly cause disease and hence be responsive to autoantibody‐depleting therapies was established with the discovery of antibodies against the acetylcholine receptor and other targets in the neuromuscular junction disease myasthenia gravis.71 Extending this into the CNS has proved more complicated. Although multiple brain‐targeted antibodies have been identified in paraneoplastic encephalitides, these ‘onconeural’ antibodies react with intracellular antigens. Although they are useful biomarkers, neuronal pathology is likely to arise secondary to cytotoxic T‐cell‐driven damage, and the antibodies are not in themselves pathogenic. Correspondingly, the diseases are unresponsive to humorally directed immunotherapies.72

Evidence from these intracellularly directed antibodies and the notion of limited, if any, CNS permeability to antibodies, formerly restricted the clinical focus of pathogenic antibodies to the peripheral nervous system. However, the identification of IgG putatively directed against cell surface voltage‐gated potassium channels (VGKC) in patients with predominantly non‐paraneoplastic limbic encephalitis initiated the current focus on conformationally specific CNS NSAbs.73, 74 The finding that this rapidly progressive encephalopathy was immunotherapy‐responsive transformed the assessment of subacute cognitive impairment. It later emerged that VGKC antibodies did not bind VGKCs themselves but bound targets complexed with the VGKC in the brain tissue used in the radioimmunoassay. These targets were leucine‐rich, glioma‐inactivated 1 (LGI1) and, less so, contactin‐associated protein‐like 2 (CASPR2).75, 76 In addition to these targets, further NSAbs against synaptic membrane receptors with clear relevance to psychosis rapidly emerged including: NMDA and AMPA glutamate receptors, γ‐amino‐butyric acid (GABA_A and GABA_B) receptors, and dopamine D2 receptor. 77, 78, 79, 80, 81 Experimental82, 83, 84 and observational85 clinical data have provided strong evidence to show that these cell‐surface‐binding autoantibodies are directly pathogenic. These illnesses respond to commonly available immunotherapies including corticosteroids, intravenous immunoglobulins and plasma exchange, as well as second‐line agents such as cyclophosphamide, rituximab and bortezomib, consolidating their clinical relevance.86

NMDAR‐antibody encephalitis in particular, presents with a striking polymorphic psychosis and has caught the imagination of psychiatry.87 Early descriptions reported previously well young women, with a rapid‐onset mood disturbance, paranoid thoughts, hallucinations and agitated behaviour, before development of an unusual movement disorder, seizures and a global encephalopathy frequently requiring intensive therapy unit admission and mechanical ventilation.82 It has since been suggested that NMDAR‐antibody encephalitis probably accounts for many cases of previously unexplained neuropsychiatric conditions including: encephalitis lethargica,88 malignant catatonia/neuroleptic malignant syndrome,89, 90 and even cases of supposed ‘demonic possession’.91 Crucially, in the context of increasingly demedicalized mental health systems,92 the majority of these patients were initially admitted to psychiatric hospitals before more clearly ‘neurological’ signs emerged:82 indeed the initial ‘psychiatric’ phase of NMDAR‐antibody encephalitis is often difficult to differentiate from some presentations of first‐episode psychosis and the illness can rarely consist of entirely ‘psychiatric’ symptoms.93 In our experience, despite increasing awareness, delayed recognition is not uncommon. Prompt diagnosis, removal of an ovarian teratoma if present, and timely immunotherapy can promote disease remission, correlating not with major structural abnormalities or focal signs,94 but with reversible disturbance in circuit function due to NMDAR internalization and reduced inhibitory GABAergic neuron signalling. This experiment of nature has been striking for its degree of fit with previous theories of NMDAR hypofunction and NMDAR antagonist (e.g. ketamine/phencyclidine) models of schizophrenia.95 As well as providing simultaneously a neural and immune disease model for psychosis, the intriguing question soon arose: how many patients with hitherto ‘primary’ psychotic disorders had this disease, or at least harboured disease‐relevant NSAbs?96

Defining NSAb prevalence and relevance in psychotic disorders has raised important questions about both disease taxonomy and antibody‐detection methods (Fig. 1).97, 98 Although NMDAR‐antibody encephalitis has an abrupt onset and a distinctive psychiatric phenotype, and usually progresses to frank neurological features, some features of an encephalopathy including catatonia, disorientation and more traditional cognitive deficits are not uncommon in primary psychotic disorders, and are probably under‐recognized.99 Estimates of NSAb prevalence in primary psychotic disorders have varied largely with symptom duration and assay approaches, but generally studies looking at patients early in their illness with live cell‐based assays have detected NMDAR‐IgG antibodies at higher rates than controls,100 whereas those looking at more chronic presentations with commercial fixed cell‐based assays have been negative.101, 102 However, larger sample sizes using multiple approaches to NSAb detection, probably across multiple centres, are still needed to more definitively answer this question.

Figure 1.

Commonly used methods of neuronal surface antibody (NSAb) detection include immunohistochemistry, immunocytochemistry, and cell‐based assays (CBA). These methods complement each other: the first two can detect NSAbs against unknown targets whereas CBAs detect NSAbs against known targets. CBAs involve transfection of a cell line with a cDNA construct encoding the target of interest either tagged or co‐transfected with a fluorescent protein. In all three methods a candidate serum or CSF sample is incubated with the detection substrate and following washing and fixation steps, tagged anti‐human immunoglobulin secondary antibodies are used to visualise binding. Interested readers are directed to Pettingill et al.70 for a more detailed exploration of this area.

Work on the potential contributions of NSAbs to postpartum psychosis (PPP), a severe perinatal mental illness,103 provides an illustrative example. Bergink et al. demonstrated serum IgG reactivity to cultured hippocampal neurons and typical surface staining pattern on rat hippocampal slices in four of 96 cases of previously collected PPP sera, with two of the four positive on NMDAR (NR1 subunit) CBAs and none in 64 control postpartum samples.104 This was interpreted as evidence of autoimmune encephalitis causing a subset of PPP. Indeed all four cases could satisfy emergent criteria based on symptom onset of < 3 months and multiple confirmatory serological tests,105 and the two NMDAR‐antibody‐associated cases in retrospect included features consistent with NMDAR‐antibody encephalitis such as neuroleptic intolerance. However, none of the cases were prospectively distinguishable on clinical grounds from other patients with PPP and had responded to usual psychiatric management such as antipsychotics and lithium, rather than requiring immunotherapy. Nonetheless, this important study has further broadened the phenotype of NSAb‐associated diseases into a distinctive ‘psychiatric’ illness, undermining dualistic approaches to mental disturbance.106 Prospective studies of longitudinal paired serum and CSF, and potentially trials of immunotherapy, will be needed to determine further the clinical relevance of NSAbs in this disorder. PPP is associated with an elevated rate of both postpartum and non‐pregnancy‐associated episodes of predominantly affective psychosis. Evaluating the role of NSAbs as a biomarker or therapeutically relevant target in these sub‐groups in particular may have a useful clinical role and give mechanistic insight.

These clinical–serological correlations illustrate the need to robustly define NSAb positivity throughout the course of these disorders. Importantly, NMDAR IgG antibodies may arise secondary to a variety of disease states, and it is not always clear that they are pathogenically relevant in each case. Examples include Creutzfeld–Jakob disease where they are presumably irrelevant to the inevitable effect of prion‐mediated neurodegeneration.107 By contrast, in post‐herpes simplex virus infection108 they can cause a typical autoimmune encephalitis or modify cognitive performance,109 ^, 110 and generate a partial phenotype such as an isolated movement disorder.111

The cause of neuronal autoantibody production is usually unclear, and is likely to be paraneoplastic in less than half of all autoimmune encephalitis cases. The relationship between ovarian teratoma, NMDAR‐antibody generation, and clinical features is intriguing. Gong et al. did not detect serum NMDAR‐antibodies in 80 women with ovarian teratoma without neuropsychiatric illness and 95 healthy control samples.112 In people with a clinical picture of NMDAR‐antibody encephalitis there are clear mechanistic links between ovarian teratoma and antibody generation,113 as well as therapeutic efficacy of tumour removal. However, this evidence implies that, more generally, in women with ovarian teratoma, NMDAR‐antibody generation may be the exception rather than the norm. This implicates an interaction between other factors such as genetic predisposition and/or infection. LGI1‐antibody disease has an emerging strong HLA association,114, 115 but there has not yet been a robust HLA association with NMDAR‐antibody encephalitis. Nonetheless, a retrospective case series from New Zealand, which found that the incidence of the disease was higher and possibly more severe in those with Pacific Island or Māori ancestry compared with those without these ancestries, supports the search for immune‐related alleles.116

Some authors have suggested that seropositivity for NSAbs is uninformative in itself and that blood–brain barrier permeability/disruption is an essential factor in considering whether NSAbs are pathogenic in a particular disease state.117 Evidence includes NMDAR‐antibody seropositivity associating with stroke lesion outcome118 or psychiatric phenotype in clinical populations being contingent upon APOE genotype, a proxy marker for blood–brain barrier permeability.119, 120 A full discussion of this is beyond the scope of the present review; however, to advance this intriguing hypothesis, in vivo measures of blood–brain barrier integrity from NSAb‐associated clinical phenotypes will be needed. Dynamic contrast‐enhanced magnetic resonance imaging has been advocated, and has given novel insights in other conditions such as migraine,121, 123 but has yet to be validated within autoimmune encephalopathy and seropositive psychosis cohorts.

Ultimately, there must be populations of neuronal surface antigen‐specific plasma cells underlying the production of detectable NSAbs. Understanding the initiation and maintenance of these populations and their distribution between the periphery, the CNS and regional lymph nodes is a priority for establishing the location of antibody‐producing cells. Kreye and colleagues recently showed the potential of this approach by single‐cell cloning of immunoglobulin genes to generate recombinant monoclonal antibodies from CSF B and plasma cells from patients with NMDAR‐antibody encephalitis. Interestingly, NMDAR‐directed populations contained both somatically hypermutated clonal populations, but also non‐mutated cells, implying a contribution from natural immunity. Furthermore, they demonstrated a diverse range of neuronal surface reactivities other than the NMDAR. A similar approach that captures antigen‐specific B‐cell traffic between the CNS and the periphery has also been reported for multiple sclerosis.124 Further development of this in NMDAR‐antibody‐associated psychiatric syndromes may help to understand the extent to which these presentations are immunologically related to autoimmune encephalitis.

In summary, NSAbs, especially NMDAR‐antibodies, provide a fascinating insight into both the neural and immune mechanisms that may contribute to psychotic disorders. NMDAR‐antibody encephalitis is an important differential diagnosis in acute psychiatric illness, meanwhile their phenotypic spectrum exemplifies an ability to contribute to clinical presentations beyond the stereotyped classical progression of autoimmune encephalitis. Moreover, they provide a mechanism through which B‐cell and possibly cytokine abnormalities previously observed in some cases of schizophrenia may be operating (Fig. 2).

Figure 2.

Using assay platforms sensitive to conformationally‐specific antibodies, NMDAR‐antibodies can occasionally be detected in serum from patients with psychosis without other clinical and para‐clinical features typical of autoimmune encephalitis. Interpretation of such a result is challenging. In some patients these antibodies may have a direct relationship with psychopathology as they do in the early stages of NMDAR‐antibody encephalitis. In others they may have a causative role but in a modifying rather than determining relationship, whilst in others they may be irrelevant to mental state or epiphenomenal to the cause(s) of their psychosis. NMDAR‐antibodies are secreted by antigen‐specific B‐cell lineages whose biology is currently poorly understood. They may reside in a bone marrow niche, tumour germinal centre‐like structures, circulate in regional lymph nodes e.g. pelvic nodes if ovarian teratoma present, or cervical nodes receiving CNS antigens, possibly via recently discovered CNS lymphatics, and/or be resident in the CNS. These lineages may have arisen directly due to incomplete tolerance or have been triggered by tumour immunity e.g. neural tissue in ovarian teratoma or viral immunity e.g. immunisation to CNS antigens during HSV encephalitis. HLA and other genetic signals could predispose to the complex bio‐psycho‐social mechanisms involved in psychosis and/or incomplete NMDAR tolerance. These plausible yet speculative mechanisms are a possible model only and require further experimental work to determine their significance.

Demonstration of autoreactive T and B cells and/or autoantibodies that can transfer disease to healthy individuals or animals through adoptive transfer or autoantigen immunization

Notwithstanding conceptual issues regarding whether non‐verbal mammals can be said to be experiencing psychotic symptoms, animal models of psychotic disorders do exist. Behavioural parameters of relevance include selective attention, exploratory and social behaviour, cognitive flexibility, sensorimotor gating and sensitivity to psychotomimetic drugs (such as amphetamine or ketamine).125 Multiple electrophysiological parameters including markers of excitatory/inhibitory potential and evoked potentials and oscillations are also used as endophenotypes.126

Relatively few papers in the modern era have sought to determine whether adaptive immune factors from patients can recapitulate a psychosis phenotype in animals. Work from the group of Diamond has focused on neuropsychiatric systemic lupus erythematosus and putative NMDAR NR2‐reactive antibodies (which are cross‐reactive with dsDNA), which have been suggested to be a potential biomarker for psychosis in patients with systemic lupus erythematosus.127 When human serum containing these antibodies was peripherally administered to rats, along with lipopolysaccharide, animals showed memory deficits and hippocampal IgG deposition;128 when the antibodies were administered along with adrenaline, neuronal damage was observed in the amygdala and animals were impaired on a fear‐conditioning task.129

Within the more recent framework of conformationally relevant NSAbs, Hammer et al. demonstrated that peripheral administration of purified NMDAR NR1 IgG from subjects with schizophrenia altered basal and MK‐801‐induced open‐field activity (both putative psychosis‐relevant phenotypes) in mice, but only in case of congenital blood–brain barrier disruption (ApoE^–/– haplotype) and not in wild‐type animals.119

Two intraventricular transfer animal models of NMDAR‐antibody encephalitis have been reported, demonstrating clearly that NR1 antibodies from patients with NMDAR‐antibody encephalitis can alter behaviour and disrupt neuronal function in vivo.130, 131 However, neither study assessed animals for psychosis‐relevant behavioural or electrophysiological (endo)phenotypes; doing so would be a necessary next step in evaluating whether NSAbs can cause psychosis.

Emerging case reports suggest that passive transfer of antibodies in humans may have the potential to transfer severe neuropsychiatric illness. In one, a 7‐year‐old boy developed autoimmune encephalitis with prominent behavioural features and with measurable VGKC, LGI1 and thyroglobulin antibodies following bone marrow transplantation.132 In another a 57‐year‐old man with no past psychiatric history developed a severe, chronic psychosis following allogenic stem cell transplant from a brother who had a diagnosis of schizophrenia (neuronal autoantibodies were not tested in either donor or recipient).133

The maternal immune activation model of neurodevelopmental disorders has assumed a central position in the literature on animal models of schizophrenia.134 In this model, pregnant animals are administered an immunogen (usually poly (I:C), lipopolysaccharide or specific cytokines such as IL‐6) and offspring exhibit schizophrenia‐related behavioural and cognitive abnormalities during and after late adolescence. Work using the model has focused largely on cytokine production and microglial priming as the chief pathological mechanism, with scant attention paid to adaptive immune system alterations. Given that Toll‐like receptor signalling plays multiple roles in B‐cell development and activation,135 it is possible that the maternal immune activation model could provide useful insights into whether early inflammatory insults in individuals who develop schizophrenia might shape the subsequent course of adaptive immune abnormalities. Further, materno–fetal transfer of specific antibodies may represent an alternative paradigm for maternal immune activation, already showing promise within the autism literature136 and with clear implications for schizophrenia pathogenesis.

Effect of immunomodulatory treatments on schizophrenia

The pharmacological treatment of schizophrenia has remained largely unchanged since the serendipitous discovery of the phenothiazines in 1950s as being calming in patients with schizophrenia. The subsequent analysis of the effects of these drugs, as blocking post‐synaptic dopamine receptors then led to the prevailing theory that schizophrenia is a disorder of raised dopamine levels. Subsequent drug development in the treatment of schizophrenia has been, in almost all cases, through development of other dopamine‐receptor‐blocking drugs.

However, it has also been recognized, since the early study of these antipsychotic drugs that, they are also anti‐inflammatory, suppressing pro‐inflammatory cytokines, IL‐1, IL‐6, TNF‐α and IFN‐γ in association with the improvement of psychotic symptoms in patients during acute episode of psychosis.51

There are a limited number of studies to examine the effect of non‐steroidal anti‐inflammatory drugs as adjunctive treatments in schizophrenia. One meta‐analysis, including six trials of celecoxib (cyclooxygenase‐2 inhibitor) and two of aspirin, found a small overall positive effect on psychotic symptoms.137 There was some indication that these effects were greater when patients were treated earlier in their illness, and that aspirin was more effective than celecoxib, but the total numbers of patients treated are still modest (fewer than 500 in total), and patients were unselected for their inflammatory status, limiting the interpretation of these findings. Furthermore, these agents have minimal, if any, role as disease‐modifying agents in canonical autoimmune disease so alternative agents would be needed to test the hypothesis more clearly.

There have been studies examining the effect of the tetracyclic antibiotic minocycline, which crosses the blood–brain barrier and is therefore considered a plausible anti‐inflammatory agent. There have been widely discrepant findings in the four studies undertaken to date, with no overall effect shown on psychotic symptoms.138

There have been no randomized controlled trials to explore the effects of more significant immune suppression in schizophrenia, using drugs that are used in the treatment of other autoimmune disorders such as corticosteroids, or monoclonal antibodies against immune targets. The often invasive nature and potential adverse effects of the treatments mean that trials need to be careful to select those patients that are most likely to benefit from this approach. However, there are case reports of people with psychosis who have received these immunosuppressive agents, for the treatment of concurrent illness, with improvement of their psychotic symptoms reported.139 One open label study to date, of patients with NSAb‐positive psychosis treated with steroids, plasma exchange and/or intravenous immunoglobulins, demonstrates that immunosuppression can be effective in treating psychosis.140 If replicated and confirmed in a randomized controlled trial setting, this would provide clinically important, albeit indirect, evidence of the relevance of autoimmune or perhaps more non‐specific inflammatory mechanisms contributing to psychosis.

Conclusion

In conclusion, there are several lines of evidence to suggest that a proportion of psychotic illnesses have a plausible autoimmune component. However, defining which component, in whom, and what if anything it means for treatment are ongoing questions for the field. Most of the evidence of the possible autoimmune basis of psychosis is circumstantial – in particular the overlapping clinical phenotype and the shared heritability between schizophrenia with other autoimmune disorders. According to the modified Witebsky postulates there is currently limited direct research evidence that psychosis is an autoimmune disorder, and there is an urgent need for studies to directly address this gap. Further studies should prioritize longitudinal, in vivo modelling of the effects of neuronal surface antibodies in patients with psychosis, as well as clinical trials to evaluate the effects of immunosuppression (see Table 1).

Table 1.

Summary of the current evidence base regarding psychosis as an autoimmune disease

	Evidence for	Evidence against	Evidence needed
Evidence of disease‐specific adaptive immune response in the affected tissue or organ
Presence in target organ	CD3+ and CD20+ lymphocyte hippocampal infiltration in post‐mortem studies.55, 56	Target‐specificity not examined. Potentially confounded by non‐disease‐related factors.	Cerebrospinal fluid (CSF) B‐ and T‐cell studies, e.g. immunoglobulin gene cloning
Presence in circulation	Increased CD19+ lymphocyte counts and raised CD4 : CD8 ratio in acute unmedicated psychosis.47, 48	Target‐specificity not examined.	CSF neuronal surface antibody‐mediated (NSAb) studies in primary psychotic disorders. Evidence of immunoglobulin deposition in brain parenchyma.
	Prevalence studies of NSAbs, e.g. NMDAR.	Not necessarily accessing and modifying brain function. Patients may improve without immune modulation.	Molecular assays of functional antibody activity.
	NMDAR‐specific and NSAbs of unknown target in post‐partum psychosis.103		Whole brain neural approaches, e.g. fMRI/spectroscopy
Transfer of disease via autoreactive cells and/or autoantibodies
Autoreactive cells	Allogeneic stem cell recipient without psychiatric history developed post‐transplant chronic severe psychosis. Fraternal donor had schizophrenia.132	Single case report.	Systematic analysis of psychiatric complications following haematopoietic donation.
		No animal models.
Autoantibodies	NMDAR(NR1) antibody mouse passive transfer model – largely behavioural phenotype.129, 130	Did not assess for psychosis phenotype.	Explore validated psychosis markers in NSAb passive transfer experiments.
	Materno–fetal transfer mouse models of neuronal surface autoantibodies leading to neurodevelopmental disorders.135	Discredited ‘taraxein’ passive transfer experiments on psychosis and EEG in 1960s.	Longitudinal follow up of children born to mothers who were NSAb seropositive.
		Very little experience of developmental trajectory in infants born to mothers with autoimmune encephalitis.	Explore validated psychosis markers in materno–fetal autoantibody transfer models.
Elimination of autoimmune response modifies disease
Anti‐inflammatory medications	Celecoxib, aspirin and minocycline may improve positive symptoms of schizophrenia.136, 137	Variable results.	RCTs including cases stratified by immune markers.
Antibody‐modifying immunotherapy	Immunotherapy of NSAb‐associated psychosis associated with clinical improvement correlating with serum NSAb titres.139	Non‐controlled, open‐label (risk of case selection bias, placebo effect, regression to the mean).	RCTs of specific immunotherapy in cases stratified by NSAb seropositivity not meeting criteria for autoimmune encephalitis.
Clinical features
Personal and/or family history of autoimmune disease	Increased risk of most autoimmune diseases, but risk of rheumatoid arthritis in people with schizophrenia is 29% that of the general population.22, 23, 24, 25	Causal mechanisms under‐explored	Further exploration of mechanistic relevance of differential risks of other autoimmune diseases.
	22q11 deletion syndrome major risk factor for psychosis also causes thymic hypoplasia and both immunodeficiency and autoimmune disease.34, 35, 36
MHC association	Schizophrenia GWAS major association in MHC locus on chromosome 6p.	Specific causal HLA variants have remained elusive or have not been replicated. HLA association may contribute risk via non‐immune role in brain development, e.g. complement C4 and synaptic pruning.	Further exploration of pathogenic relevance of immune loci to clinical syndrome.
	B‐cell markers CD19 and CD20 also strongly associated.27
	LGI1‐antibody encephalitis has a strong specific HLA‐association, occasionally associated with psychotic features.113, 114

Insights into the pathogenic actions of antibodies against neuronal cell surface targets in particular, provide an opportunity for psychiatry to define a proportion of those with a psychotic illness that may be amenable to a novel immunotherapeutic approach. The treatment of autoimmune encephalitis has been transformed in the last decade, from a disorder with a poor outcome and high mortality rate, to one in which patients may recover completely. If a proportion of psychosis is similarly attributable to these same antibodies, then there is the potential for the treatment of psychosis to be similarly transformed, with profound implications for the practice of psychiatry.

Disclosures

Dr Al‐Diwani is the recipient of a Wellcome Trust clinical research training fellowship no. 205126/Z/16/Z and is supported by the Oxford NIHR Biomedical Research Centre. Dr Pollak is supported by a clinical research training fellowship grant from the Wellcome Trust (no 105758/Z/14/Z). Associate Professor Irani is a co‐applicant and receives royalties on patent application WO/2010/046716 titled ‘Neurological autoimmune disorders’. The patent has been licensed to Euroimmun AG for the development of assays for LGI1 and other VGKC‐complex antibodies. He has received speaker fees and research funding from The Wellcome Trust, UCB‐Oxford Alliance, Epilepsy Research UK, MedImmune, and the British Medical Association–Vera Down Grant. Associate Professor Lennox is supported by NIHR CLAHRC Oxford. She has received speaker fees and research funding from Biotest, Lundbeck Foundation, MRC and Stanley Medical Research Centre. The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.