Antibodies that bind to cell surface neuronal, glial, or synaptic targets, collectively known as neural surface antibodies (NSAbs), have attracted significant attention in neurology and psychiatry (2). Their detection in a patient presenting with psychiatric symptoms raises the possibility both of a causal or disease-modifying role and of clinical improvement with immunotherapy (IT). This would represent a major step forward from current largely symptom-targeted psychotropic medications and has been met by clinicians and researchers with enthusiasm.

Although this iteration of autoimmune psychiatry is in its infancy (3), here we argue that there already exists ample evidence to warrant an expanding research program, focusing on robustly establishing the prevalence and relevance of NSAbs in what would otherwise appear to be primary psychiatric disorders. We discuss the controversies in applying knowledge of autoimmune encephalitis (AE) to such psychiatric disorders and suggest experiments by which these controversies may be resolved.

Given the prominence of psychiatric symptoms in many types of AE, and the known importance of receptor targets of NSAbs to psychopathology in psychiatric disorders, there have been extensive efforts to define disease-relevant associations between NSAbs and isolated psychiatric syndromes such as first-episode psychosis (FEP) (4). The development of assays able to detect specific antibodies against central nervous system (CNS) neural surface antigens in combination with careful clinical phenotyping has made this possible. Primary psychiatric disorders, such as schizophrenia, are far more common than AE. If a proportion of these were to have an NSAb-related etiology, then applying the lessons from the rare disease could stand to benefit a far larger group of patients.

Differences in acuity of sample timing, serum testing without paired cerebrospinal fluid (CSF), and variations in assay method have led to a lack of consensus on prevalence of NSAbs in psychiatric illness. Initially, higher rates of autoantibody prevalence were found early in psychotic illnesses compared to the chronic phase (5), but some evidence makes this distinction less clear (6). Live, non-permeabilized cell-based assays (CBAs) largely find higher NSAb prevalence in patients compared to controls (7, 8), but this has not been wholly true of fixed permeabilized CBAs (9–11). The use of CBAs in isolation has been criticized for lacking disease relevance (12); however, some studies supplementing CBA with immunohistochemistry and/or neuronal staining do in fact detect differences between cases and controls in certain psychiatric syndromes such as postpartum psychosis (13).

The rapid development of knowledge on AE, and the increasing and potentially confusing “phenotype spread” associated with NSAbs, have inspired diagnostic criteria for AE (14). These focus on early detection prompting timely IT. Diagnosis is deemed more likely with multiple symptoms, symptoms spanning both neurological and psychiatric domains, and supportive imaging, electroencephalogram (EEG), and CSF changes. Less emphasis is placed on NSAb detection as AEs tend to present stereotypically and can be recognized with readily available investigations, whereas NSAb results may take several weeks to process. However, by definition, psychiatric syndromes with serum NSAb positivity on routinely used CBAs will be incompatible with this system. To illustrate, for NMDAR antibody positivity to be relevant by these criteria, it either needs to be detected in CSF or in serum with multiple assay methods, and “relevance” equates only to a diagnosis of NMDAR-AE. Given that NMDAR antibodies are the most commonly detected NSAb in isolated psychiatric syndromes (9), and that NMDAR-AE occasionally does not progress beyond psychosis, this has implications for much of the field. NSAbs found in isolated psychiatric syndromes, which are currently considered of research interest and potential clinical relevance, would be scientifically neglected or “orphaned.” The reality that CSF is often difficult to obtain in these patients, and multiple assay methods to evaluate serum are rarely available outside specialist centers, compounds this issue.

Furthermore, a negative CSF result should not automatically render NSAb seropositivity irrelevant. First, some NSAbs, such as LGI1, are detected less frequently in CSF, in this case in half or less of LGI1 AE cases (15). Second, if NSAbs are pathogenic in some of the isolated psychiatric presentations in which they are detected, it is likely that they occur at lower titers than in fulminant AE. Furthermore, the brain parenchyma can act as an “immunoprecipitator” of NMDAR antibodies (16), so unless the parenchyma is “saturated” by an extremely high concentration of antibody, it is plausible to expect that NSAbs may not be found in the CSF while still directly disrupting brain function.

We propose that these challenges for the field can be addressed by use of a syndrome-level taxonomy. Patients with isolated psychiatric symptoms and detectable NSAb can be characterized as “synaptic and neuronal autoantibody-associated psychiatric syndromes,” abbreviated to “SNAps” (Figure 1A). SNAps deliberately have less stringent criteria for onset time, antibody class, and assay criteria. This enables a broad category, agnostic to the precise pathogenic role of the NSAb. In real time, additional serum antibody tests, identification of CSF antibody, or further investigations (e.g., MRI, EEG) may reveal abnormalities that meet the diagnostic criteria for AE, and presuming that new symptom domains do not evolve, these patients can then be characterized as having SNAps-AE (see Table 1). Patients with clear features of AE would continue to be classified as such. We suggest that this model, in which an NSAb-associated isolated clinical syndrome and AE partially overlap, can be extended to other syndromes such as epilepsy (17) and cognitive impairment (18) (see Figure 1B). Table 2 demonstrates that many of the signs and symptoms of AE are in fact seen (often to an attenuated degree) in patients with so called “isolated” psychosis, suggesting that the overlap between these conditions may be even greater still. The nascence of the field would suggest this system to be clinically pragmatic and also heuristic, promoting the delineation of biologically defined disease sub-classes and, equally as important, the signals that imply lack of disease relevance.

Most NSAbs appear to share a common mechanism of action in their ability to cause cross-linking and internalization of their target antigen (4). Where the target is a neurotransmitter ion channel receptor, it is likely that the resulting receptor hypofunction results in impaired neurotransmitter signaling, as demonstrated by abolition of currents on some neurophysiological assays and alterations in synaptic plasticity (31, 32). Although not proven, the rationale for this causing the signs and symptoms of NSAb-associated disease is compelling and the “receptor hypofunction” model resonates well with many current theories of psychiatric disorders (33). Importantly, pathogenic potential as indexed by effects on receptor internalization and postsynaptic currents in vitro has been demonstrated for NMDAR antibodies taken from patients with schizophrenia (10, 34) and to a lesser extent bipolar disorder (35), demonstrating that pathogenicity is not restricted to NSAbs found in encephalitic presentations.

Synaptic dysfunction affecting glutamatergic neurotransmission has been proposed as a mechanism in schizophrenia, bipolar disorder, and major depression (36–38). We consider that antibody-mediated glutamatergic synaptic dysfunction, if relevant to psychiatric symptomatology, is likely to have relevance that cuts across traditional diagnostic boundaries. For example, the psychiatric phase of NMDAR-AE can include a diverse range of symptoms including affective and anxiety in addition to psychosis (39) and prevalence estimates detect NMDAR antibodies across traditional diagnostic boundaries (40). If NMDAR antibodies are pathogenic outside of encephalitis, then we could expect the psychiatric manifestations to have similar clinical heterogeneity. Nonetheless, only further clinicopathological correlation of SNAps will help determine an accurate picture. Such study may help validate or suggest new directions for receptor-based models of idiopathic psychiatric syndromes.

Alternatively, NSAbs in SNAps may not be directly pathogenic but still are part of the primary disease process, for example, as part of a broader immune and/or inflammatory syndrome which may be IT responsive. A randomized controlled trial could interrogate this, but to be robust would need to be designed and powered to test multiple forms of IT and detect partial responses.

Another possible role for NSAbs in SNAps is as a prognostic marker, thereby allowing disease stratification. This may include likely response to antipsychotic medication, or illness trajectory. Cohort studies would best assess this hypothesis. Suitable populations might include those at ultra-high risk or in the prodromal phase of a psychiatric disorder, in whom such biomarkers are already sorely needed.

A final potential role of NSAb is as part of a secondary process, following a separate primary disease process. The presence of NSAbs in disorders such as herpes simplex encephalitis, Alzheimer’s disease, Creutzfeldt–Jakob disease, and other dementia types (18, 41–43) strongly suggests that NSAb production sometimes occurs following neuronal destruction. Nonetheless, “secondary” antibodies can still have pathogenic potential: in dementia, NSAbs may confer a higher risk of psychosis (43), and in patients who have had herpes simplex encephalitis, NSAbs associate with greater cognitive impairment (44).

It is plausible that NSAbs associated with psychiatric disease fall into this “phenotype-modifying” category. For example, NMDAR antibodies may be found in psychosis because the primary pathology has rendered NMDARs immunogenic. Other “immunizing” conditions could include late pregnancy and parturition, around which time numerous immunological rebound changes are understood to occur (45). Indeed NSAbs have been detected in the serum of postpartum psychosis cases (13). These women responded to usual psychiatric care and clinically did not present as having AE, but even in these situations, the value of IT remains to be decided by rigorous trials.

The autoimmune encephalitides have largely been described in association with Abs of the IgG isotype, and typically only IgG is screened for clinically. While the majority of autoimmune disorders are indeed IgG-mediated, numerous instances of IgA- and IgM-mediated diseases exist outside of the CNS, such as IgA pemphigus and autoimmune hemolytic anemia. IgA and IgM NSAbs have been consistently reported in psychiatric disorders (9, 46, 47). Importantly, there is in vitro evidence of their pathogenicity (18, 34, 35), and IgA and IgM seropositivity does appear to associate with clinical phenotype in some non-encephalitis conditions (18). We suggest that it is premature to dismiss a priori non-IgG NSAbs as irrelevant to the disease. For the purposes of our categorical model, IgA or IgM seropositivity is included in the definition of SNAps, but not of SNAps-AE, implying possible, but not definite, causal relevance.

Animal models could potentially demonstrate the pathogenicity of NSAbs in SNAps, but animal models of AE have been slow to develop and often recapitulate a limited facet of a complex phenotype (48, 49). The abnormal movements, spontaneous seizures, autonomic instability, or psychosis-like behaviors associated with AE are notably absent. Nonetheless, the animal model of NMDAR encephalitis of Planaguma and colleagues (48) appears to be an unintended but plausible model of NMDAR antibody-mediated depressive behavior. Future work on an animal model of SNAps should attempt to integrate passive NSAb transfer experiments with established psychiatric animal endophenotypes. For example, in a potential model of psychosis, multiple behavioral and neurophysiological indices such as latent inhibition and mismatch negativity could be assessed alongside the neuronal effects of antibodies (50).

Ultimately, in vivo human studies of SNAps will be necessary to elucidate whether NSAbs differentially impact brain function in psychiatric presentations. Functional MRI and magnetic resonance spectroscopy offer equivalent promise in restricted psychiatric presentations as for more fulminant neurological presentations (51, 52). With the development of PET and SPECT ligands for in vivo measurement of microglial activation (53, 54) and the function of individual receptor types [e.g., NMDAR (55) and GABAR (56)] or neurotransmitter synthesis capacity (57), it is likely that this methodology can offer insights into the molecular pathology of SNAps. Correlation of clinical improvement with improvement in disease-relevant biomarkers (for example, a perturbed glutamate MRS signal or ketamine-like functional dysconnectivity) following antibody removal would strengthen an argument for in vivo pathogenicity in SNAps. Conversely, an absence of differences from seronegative patients with psychiatric disorders on such neuroimaging measures would make it less likely that NSAbs directly affect brain function.

The description of AE syndromes caused by NSAbs has profoundly impacted neurological practice and has invigorated neuroimmunology as a basic and clinical science. The extent to which the paradigm—of autoantibodies affecting brain function and behavior—has relevance for clinical psychiatry is a matter of considerable debate.

We have introduced the concept of SNAps as an attempt to clarify some of the issues in this sometimes confusing field. By making the distinction between AE and isolated psychiatric presentations associated with NSAbs, we hope to encourage the latter as an important focus of research in its own right. It is now generally accepted that many psychiatric disorders may be comprised of subgroups that are phenotypically similar but heterogeneous in their etiologies. We believe that the category of SNAps offers a well-defined candidate for one such subgroup, and we have offered suggestions as to how this hypothesis may be tested.

Research on SNAps will of course be informed by the AE literature to date, particularly in terms of pathogenic mechanisms. In this article, we have outlined the increasingly compelling evidence that NSAbs may have pathogenic potential in SNAps: we are clear however that considerable further work needs to be done in this area before generalized statements can be made about pathogenicity. Further, we would suggest that much of this work needs to be led by psychiatrists. The nuances of psychiatric signs and symptoms are often ignored within the neurological literature, and while this remains the case, a “psychiatric phenotype” associated with NSAbs will remain elusive.

The possibility that the SNAps concept may delineate a subgroup of psychiatric patients with a differential treatment response (including a potential IT response) remains an exciting focus of future research. In an area of medicine where novel therapies are relatively rare and where immune therapies are increasingly under the spotlight, SNAps represent a focus for therapeutic studies that could be potentially transformative for the field.

AA-D, TP, AL, and BL conceived the article. AA-D, TP, and AL drafted the text. AA-D and TP prepared the tables and figures. BL revised the text, tables, and figures. All the authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.