Clinical manifestations

Presentation and course

Antibody-associated epilepsy in the setting of encephalitis. Over 2000 patients with antibody-associated encephalitides have been described in the last decade. Most of these patients have seizures as a prominent feature of their illness, sometimes in isolation (02). These conditions have now reached mainstream neurology and are considered frequently by clinicians seeing patients with an acute or subacute onset of amnesia, psychiatric features, disorientation, and/or seizures. Typically, symptoms manifest over days to weeks. Although there are significant clinical differences between patients with the 2 most common antibodies (LGI1 and NMDAR; see Table 1). Overall, in these patients, the seizures often respond relatively poorly to antiepileptic drugs (AEDs) but well to immunotherapies (12; 37; 31; 36; 65; 27; 02). In particular, early immunotherapies appear to offer most substantial benefits (30; 36; 13; 63). One major aim of the immunotherapies is thought to be reduction of antibody levels and accompanying aspects of the underlying inflammatory process. The reduction in antibody levels in particular often correlates modestly well with decreased seizure frequencies in an individual patient, suggesting antibody pathogenicity. With the marked exception of ovarian teratomas in young females with NMDAR-antibodies and some cases of Morvan syndrome with thymomas, it is becoming recognized that the majority of these cases do not have an underlying tumour (30; 35; Vincent et al 2011; 22; 63).

Leucine-rich glioma-inactivated 1 (LGI1) are typically found in patients with limbic encephalitis and faciobrachial dystonic seizures (see Table 1). Males are affected more commonly than females, and the median age of onset is around 63 years with a range from around 20 to 95 years of age. Remarkably few children are described in the literature. They were formally described as a subunit of the VGKC-complex, but the VGKC-complex radioimmunoassay lacks sensitivity and specificity compared to direct testing for LGI1- and CASPR2-antibodies; therefore, it is no longer a useful test and also frequently leads to unnecessary administration of immunotherapies (29; 69; 05).

CASPR2-antibody encephalitis is also frequently associated with seizures (29; 70), and a pure-epilepsy syndrome has also been described in patients with CASPR2-antibodies (67). CASPR2-antibodies are discussed further below in the context of Morvan syndrome.

In addition, other less frequent antibodies directed against the GABAA-, GABAB-, and AMPA-receptors have been detected in a limited number of patients with limbic encephalitis, many of whom have prominent and antiepileptic drug-refractory seizures, and a substantial number of these patients have an underlying tumor (54; 14; 55).

Antibodies against the intracellular protein glutamic acid decarboxylase (GAD) have also been found in a few patients with limbic encephalitis who present with temporal lobe seizures. In 1 study, the frequency of this syndrome was similar to LGI1/CASPR2-antibody limbic encephalitis (47). GAD-antibody encephalitis is usually a nonparaneoplastic condition that often affects young females. The antibodies have a lower likelihood of being pathogenic as they target an intracellular antigen (28), and the levels of the antibody correlate poorly with disease severity with an individual patient. In many of these cases the seizures are often resistant to both antiepileptic drugs and to immunotherapies. These refractory seizures may lead to the unfavorable long-term cognitive outcomes (46). Indeed, 1 study has shown that within a cohort of patients with temporal lobe epilepsy, the subset with GAD-antibodies often have pharmacoresistant seizures, and their disease is associated with depression and memory dysfunction (Malter et al 2009; 18).

Rasmussen encephalitis has been inconsistently associated with autoantibodies. Despite early suggestions that GluR3 antibodies may be present in these patients, these findings have proven difficult to reproduce (72), and preferential responses to T-cell mediated immunotherapies question a humoral immune-mediated mechanism in the pathogenesis of Rasmussen encephalitis. Whether the T-cell mediated mechanism reflects a viral etiology is yet to be determined.

Table 1. Differences among LGI1, CASPR2, NMDAR-, and GAD-antibody associated forms of encephalitis

	LGI1 and CASPR2	NMDAR	GAD
Male:female	2:1 and 8:1	1:3	1:3
Age	Usually > 50 years old, rarely affecting children	Usually < 40 years old, commonly young children	Wide ranging: 17 to 66 years of age in 1 study (47)
Target antigen	Secreted LGI1 or extracellular domain of CASPR2	Extracellular region of NR1 subunit of NMDAR	GAD-65
Tumor association	Likely reflecting background population in LGI1. If so, SCLC and thymoma. In Morvan syndrome and neuromyotonia, CASPR2-antibodies commonly associate with thymoma.	Ovarian teratoma (in 20%), others rarely	Uncommon
Clinical features	Amnesia, disorientation, temporal lobe seizures, faciobrachial dystonic seizures (with LGI1-only), and affective disturbance. CASPR2 antibodies associate with pain, neuromyotonia, and dysautonomia	Psychiatric features, amnesia, disorientation, and seizures in the first stage of the disease; these progress to a second stage with a movement disorder, dysautonomia, and reduction in consciousness (12; 29)	Temporal lobe seizures, amnesia, and affective disturbance
Distinctive clinical features	Faciobrachial dystonic seizures, which most often precede the amnesia and disorientation (32; 34)	Choreoathetoid movement disorder, mutism, usually starting days to weeks after the psychiatric features	None known but may have co-existing cerebellar ataxia, type 1 diabetes, or stiff person syndrome
Blood tests (other than antibody)	Hyponatremia consistent with SIADH (in around 60%)	None known	None known
MRI	Bilateral hippocampal high signal on T2/FLAIR (in around 60%)	Often normal. Occasionally, non-specific high signal or medial temporal lobe high signal	Bilateral hippocampal high signal on T2/FLAIR
CSF	Most commonly normal	Typically early lymphocytic pleocytosis and later oligoclonal bands	Oligoclonal bands and occasional lymphocytic pleocytosis
Immunotherapy regime	Usually good response to 1 or 2 immunotherapies (steroids +/- IVIG/plasma exchange)	Slow response; often requires > 2 immunotherapies, plus utility of tumor removal	Usually poor touches despite several antiepileptic drugs and immunotherapies

Adapted from (37).

Faciobrachial dystonic seizures. This distinctive seizure semiology was first described in 2008 (32). Most recently this was definitively characterized in 103 patients (63). Faciobrachial dystonic seizures describe a semiology with frequent, brief, highly stereotyped dystonic episodes with arm posturing and ipsilateral face grimacing. These patients have LGI1-antibodies, meaning that the clinical semiology is almost pathognomonic of the underlying molecular diagnosis (36; 63).

The main importance of recognizing faciobrachial dystonic seizures lies in their frequently striking response to immunotherapies and their timing within the natural history of LGI1-antibody-associated encephalitis. Ninety percent of faciobrachial dystonic seizures are antiepileptic drug refractory, with the residual 10% responding to sodium channel blockers over levetiracetam (19). Furthermore, antiepileptic drugs are associated with frequent side effects in these patients, with rashes noted at up to 50% rate. However, faciobrachial dystonic seizures are often very sensitive to immunotherapies, particularly steroids. In addition, intravenous immunoglobulins (IVIG) and plasma exchange have reduced seizure frequency in some patients. Their sensitivity is time-dependent, with greater chances of seizure cessation in those treated earlier, and in those with no cognitive impairment (63).

In around 25% of cases with faciobrachial dystonic seizures, seizure onset is after the cognitive impairment that characterizes LGI1-antibody associated limbic encephalitis. In these cases, the observation of this distinctive semiology should prompt confidence in immunotherapy administration before confirmatory serology is available. More importantly, in around 75% of patients with faciobrachial dystonic seizures, the faciobrachial dystonic seizures preceded the onset of cognitive impairment (63). Therefore, it was hypothesized that treatment of faciobrachial dystonic seizures with immunotherapies may prevent limbic encephalitis (34). Indeed, 2 sequential studies confirmed prevention of cognitive impairment in patients whose faciobrachial dystonic seizures were effectively treated (36; 63). Therefore, the clinical recognition of faciobrachial dystonic seizures may aid an early molecular diagnosis, and it appears to support the prompt administration of corticosteroids, which may prevent otherwise ensuing cognitive impairment (36; 63).

Other autoantibody-mediated epilepsies. In addition, several studies have begun to define other characteristics of autoantibody-mediated seizures, which largely follow the paradigms of faciobrachial dystonic seizures. For example, features that may identify patients with a high likelihood of being immunotherapy-responsive include frequent focal seizures, neuropsychiatric features, brain MRI abnormalities, presence of a neural autoantibody, and an inflammatory CSF (58; Iorio et al 2014; 65). Typically, these features are predictive of LGI1- or CASPR2-autoantibodies (58; 71), but also some cases with other neuronal-surface reactivities (58; 01). Applying these features in a scoring system may in fact help select patents with autoantibody mediated epilepsy from unselected general neurology patients (17).

More specifically, the presence of pilomotor dysfunction, paroxysmal thermal sensations, and bradycardia offer some added specificity towards the diagnosis of an antibody-mediated epilepsy (50; 59; 01). Often these patients have LGI1-antibodies. Although, of course, these semiological features are less specific than faciobrachial dystonic seizures, they remain very helpful in clinical identification of antibody-mediated patients.

Unselected epilepsies associated with antibodies. With greater relevance to the more common epilepsies, it was recognized in 2005 that autoantibodies, particularly against the VGKC-complex, were present in around 10% of patients with antiepileptic drug-refractory and cryptogenic epilepsies (48). Since, it has become apparent that these antibodies rarely represented the likely pathogenic LGI1 or CASPR2 reactivities (“double-negative”). In fact, nearly all fail to bind the extracellular aspects of native neuronal proteins in vitro. Furthermore, several of these double-negative antibodies have been shown to bind the intracellular aspects of VGKCs and to dendrotoxin, a nonmammalian protein used in the VGKC-complex assay (33; 70; 42).

However, double-negative VGKC-complex antibodies appear to be found more commonly in patients with focal epilepsy than in those with generalized epilepsy, and in patients whose seizures respond poorly to antiepileptic drugs (06). Maybe in these and other similar patients they represent a biomarker for an underlying inflammatory process that may be part of the etiology.

In addition, some patients have been reported with seizures and NMDAR-antibodies but few other features characteristically associated with NMDAR-antibodies (51; 30; 39; 25).

In these patients it is unlikely that the antibodies – rarely detected in CSF – are directly pathogenic.

In a large series of patients with new onset refractory status epilepticus (NORSE), it has become recognized that autoantibodies are frequently detected (23). This important study requires more detailed serology to understand whether the VGKC-complex antibodies were double-negatives, or targeted LGI1/CASPR2, and whether the NMDAR-antibodies were also present in CSF.

Postical psychosis and autoantibodies. An intriguing article has proposed that after a seizure, antibodies may access the brain via the breach in blood-brain barrier integrity and generate the well-recognized clinical phenomenon of postictal psychosis. This may account for the typical lucid interval seen between seizure and psychosis and for the self-limiting nature of the condition (57). The role of the blood-brain barrier requires more formal investigations in all the autoimmune encephalopathies. Its role in some rodent models has received preliminary attention (09), but the effects in humans of relative contribution of serum-antibody crossing into CSF, versus the intrathecal production of antibody, needs clarification in order to understand the best compartment to target therapeutically.

CNS autoantibody-mediated syndromes without frequent seizures. By contrast to the above findings, 2 central nervous system autoantibody-related syndromes, Morvan syndrome and neuromyelitis optica, appear to be largely nonepileptogenic. In neuromyelitis optica, the autoimmune process appears to be directed against the aquaporin-4 water channel. The often intense inflammatory infiltrates are seen in neuromyelitis optica pathology, however, the burden of disease affects the spinal cord, optic nerves, and subcortical structures, and this may explain the paucity of seizures. Yet, some patients with antibodies to myelin oligodendrocyte glycoprotein do have encephalitis with prominent seizures and some cortical imaging abnormalities.

In Morvan syndrome, the lack of seizures is far harder to understand. Morvan syndrome has been associated with antibodies often target both CASPR2 and LGI1 either protein alone, and sometimes contactin-2 in addition (35; 44). Given that LGI1-antibodies appear to be epileptogenic in faciobrachial dystonic seizures and limbic encephalitis, and CASPR2-antibodies also associate with some cases of limbic encephalitis, it is difficult to understand why only around 35% of patients with Morvan syndrome develop seizures. Furthermore, these seizures are not especially prominent. It may be that future animal models further help inform this question.

Mesial temporal lobe epilepsy and hippocampal sclerosis as a consequence of autoantibodies. It has been shown that limbic encephalitis was a prodromal illness in around half of adult-onset temporal lobe epilepsy with hippocampal sclerosis. It was suggested that the medial temporal lobe inflammation occurring during limbic encephalitis was the substrate for subsequent atrophy and hippocampal sclerosis (03). This observation may indicate that some cases of adult-onset hippocampal sclerosis can be prevented by prompt recognition and treatment of limbic encephalitis. This possibility is even more intriguing given that cases with faciobrachial dystonic seizures, and perhaps patients with other isolated antibody-mediated epilepsies, may benefit from immunotherapy to prevent the onset of the encephalitis (63). It may be that the natural history of these diseases will alter profoundly with immunotherapies.

Clinical vignette

A 92-year-old hypertensive male presented with a 3-month history of events that synchronously affected his right arm and face. The face was pulled upwards and the arm would supinate and posture upwards. Each event lasted for a few seconds. These began at a frequency of 2 per day and now occurred twice per hour; as a result he was dropping objects in his grasp. Some of the episodes were accompanied by speech arrest and loss of awareness. One month into his illness, his general practitioner commenced carbamazepine. Despite this, attack frequency steadily increased. As he developed a rash, his medication was changed to lamotrigine, again without a reduction in seizure frequency. Neuropsychology assessment and brain MRI were normal. On inspection, the events were identical to cases of faciobrachial dystonic seizures, which have been described above. The patient was treated with 1 g of methylprednisolone and on his third dose the events were much less frequent. After 1 month, no further events were observed. His LGI1-antibodies, sent from the initial clinic visit, were positive (measured using a cell-based assay), and he carried HLA-DRB1*07:01. The steroids were tapered over 6 months with a reduction in LGI1-antibody levels, and there was no return of faciobrachial dystonic seizures.

Biological basis

Etiology and pathogenesis

The antibody-mediated pathogenesis. Accumulating clinical and scientific data suggest that these disorders are caused by antibodies directed against the extracellular surface of neuronal proteins such as the NMDAR and LGI1. After binding to their antigenic target, there are 4 potential mechanisms by which antibodies may affect neuronal function. First, as originally shown with AChR antibodies in myasthenia gravis, the antibodies may cross link surface proteins and internalize the target. This has been confirmed for NMDAR-antibodies (12) and shown to be true for LGI1-antibodies after LGI1 is associated with its natural receptor ADAM22 (63). Second, it is possible that bound antibodies may fix complement and deposit membrane attack complex (MAC) on the target membrane. This has been demonstrated in biopsied or postmortem tissue from patients with LGI1- and CASPR2-antibodies (04), and it is now clear that some LGI1-antibodies are of the IgG1 complement-fixing subclass and that the levels of these IgG1-LGI1 antibodies is proportional to the degree of cognitive impairment in 1 study (63). By contrast, complement deposition was not seen in specimens from patients with NMDAR-antibodies. However, this might be explained by different brain regions examined. Third, antibodies may directly modulate the kinetics or function of the target protein, although this has yet to be convincingly demonstrated in antibody-mediated disorders of the central nervous system. Nevertheless, LGI1-antibodies can bind soluble LGI1 and prevent its interaction with the postsynaptic ADAM22 receptor, consequently reducing AMPA receptor clusters in rat hippocampal neurons (52). In addition, genetic downregulation of LGI1 induced a reduction in presynaptic VGKC currents, and perhaps antibodies to LGI1 will mediate similar effects (61). Finally, it is possible that antibodies could cause direct cytotoxic damage (antibody-mediated cytotoxicity), for example, by modulating calcium or potassium entry to neurons. This mechanism is yet to be proven.

More definitive proof of direct pathogenicity requires behavioral and morphological changes consistent with encephalitis in animals exposed to passive transfer of purified IgG. Indeed, a report has shown that a continuous infusion of NMDAR-antibody into the ventricles of mice causes a behavioral deficit that is consistent with hippocampal, and more widespread cortical, dysfunction. This infusion was associated with specific downregulation of NMDARs and the effect resolved after cessation of antibody infusion (56). Indeed, the pathogenicity of these antibodies is supported by their targeting of neuronal protein extracellular domains, modest correlations between serum levels and clinical features, and genetic or pharmacological paradigms, showing that modulation of these proteins can generate often similar clinical features (70).

The underlying immunology. By contrast to the effector mechanisms of autoantibodies, the immunobiology underlying these conditions has received relatively little attention. This knowledge seems key to understand etiology and disease perpetuation. One study has shown that lymphocytes derived from the circulation of patients have the capacity to produce NMDAR-directed autoantibodies (45). These B cells can be stimulated under defined conditions to differentiate into antibody-secreting cells, and produce NR1-autoantibody specificities at levels that correlate with the serum NR1-autoantibody levels. In addition, low to moderate levels of NR1-directed IgM autoantibodies can be observed in patient sera at a variety of time points from disease onset. These novel observations suggest that germinal center reactions may account for a significant proportion of NR1-autoantibody production in vivo. The main germinal centers are likely to be within lymph nodes. In addition, the ovarian teratoma contains T-B cell lymphoid aggregates and the intratumoral lymphocytes can secrete the NR1-autoantibodies. Hence, the teratoma may also be considered a germinal center.

Although small reports suggest NMDAR-antibody encephalitis shows no HLA bias, observations from patients with LGI1 antibodies show a striking MHC class 2 association. Around 90% of 40 patients from differing ethnic backgrounds showed an association with HLA-DRB1*07:01 (40; 68). We have also extended this association to 68 further UK and U.S. patients and demonstrated linkage with several other class 2 and two class 1 loci (70; 05). These observations strongly implicate T cell initiated pathology in patients with LGI1-antibodies and mean that T cell modulating therapies may have important implications for future immunotherapy strategies.

Epidemiology

Although it is difficult to quantify rare diseases, an approximate incidence of 1 to 3 per million new cases per year has been estimated for LGI1 or NMDAR-antibody encephalitis based on ascertainment by our UK reference laboratory (Waters and Irani, unpublished). However, underrecognition remains a clinical problem and these numbers can only increase with time.

Prevention

There is now potential to prevent the progression to limbic encephalitis by prompt detection and treatment of faciobrachial dystonic seizures and to prevent hippocampal sclerosis by recognition and treatment of preceding encephalitis. This has been detailed above alongside the main references (03; 63).

Differential diagnosis

In the differential diagnosis, consider Wernicke-Korsakoff syndrome, infective encephalitis (especially HSV), drug/toxin overdose (for example ketamine for NMDAR-encephalitis), encephalitis lethargica (the hyperkinetic half is now established to be associated with CSF and serum NMDAR-antibodies) (11), Creutzfeldt-Jacob disease—a few patients with LGI1-antibodies have been mistaken for sCJD and showed a good response to immunotherapies (24), and Hashimoto encephalopathy where it is likely that some cases will have antibodies directed against known surface proteins (66). The clinical significance of low levels of autoantibodies, especially NMDAR antibodies, in a small proportion (typically 2% to 5%) of patients with non-immune diseases, including Creutzfeldt-Jacob disease and various neurodegenerative conditions and dementias, and is yet to be clarified (16).

Diagnostic workup

Serum antibody levels are almost always higher than CSF levels in these conditions at disease onset (47; 29). However, testing in both samples offers specificity in addition to sensitivity, and positive results in either compartment have different meanings.

In Oxford, LGI1-, CASPR2- and NMDAR-antibodies are defined as positive when serum or CSF binds to an unpermeabilized antigen-transfected cell line but not to the same cells transfected with a different protein. However, currently available commercial assays used fixed permeabilized cells, and some groups recommend use of brain section staining and cultured neuron staining in addition to the cell-based assay, as confirmatory assays. The different available cell-based assay techniques have been reviewed previously (37; 70).

Low positive values of GAD-antibodies are common in a number of diseases, notably diabetes mellitus. Titers usually need to be higher than 1000 U/mL if relevant to CNS disease, but these high titers can be found in a small proportion of diabetics and in other nonseizure related disorders. As local laboratories may use different detection methods and “cut-offs,” interpretation of results needs to be considered carefully.

Management

The autoantibody studies showing treatment effects and outcomes in epilepsy have been summarized by Bakpa and colleagues (02).

The mainstays of management are symptomatic therapies (eg, antiepileptic drugs and drugs to control dyskinesias), observation for complications characteristically associated with the antibody (eg, NMDAR-antibodies and dysautonomia), reduction of antibody levels with immunotherapies, and tumor removal where relevant.

The tumor removal may act to eliminate lymphocytes within an active germinal center and a source of antigen that stimulates the ongoing immune reactions (45). Although mechanisms of action of corticosteroids and IVIG have been reviewed in great detail, the effect of plasma exchange – removal of circulating antibodies – in CNS conditions has only recently been studied. One elegant study has suggested that the effect may relate to a concomitant reduction in serum autoantibodies, which may correlate with clinical efficacy (15). This intriguing possibility requires further prospective studies to understand the relative timings of antibody reduction and clinical outcomes.

The available data are broadly consistent among studies. First, it is clear that steroids are of benefit in LGI1-antibody encephalitis for seizure control and for reduction of cognitive deficits (30; 36). Although steroids alone may be sufficient for some patients with LGI1-antibodies, IVIG or plasma exchange are reasonable alternatives, and few patients appear refractory to steroids plus 1 of these other 2 options (62). However, long-term outcomes have not been associated with the type of treatment interventions (08).

By contrast, around 50% of patients with NMDAR-antibody encephalitis make a good response to combinations of steroids, IVIG, and plasma exchange. The remaining 50% of patients show an improved outcome with additional therapies, such as cyclophosphamide or rituximab (13; 64). A study suggested a poor long-term cognitive outcome in NMDAR-antibody encephalitis treated with late or limited immunotherapies (20).

Patients with GAD-antibody associated limbic encephalitis tend to have a poor outcome despite often-aggressive immunotherapy (47; 21). Novel therapeutic options are required in this group and may include drugs such as rituximab (43) and bortezomib (60), which eliminate nonplasma cell B cells and plasma cells, respectively.

Special considerations

Pregnancy

Unlike LGI1-antibody encephalitis, NMDAR-antibody encephalitis often affects women of childbearing age. Although it is theoretically possible that IgG-antibodies (particularly the IgG1 subclass) can cross the placenta, 1 study has described 3 pregnant mothers and surprisingly could not find NMDAR-IgG in the serum of the 1 fetus tested (41). However, an article has shown maternal-to-foetal transfer of NMDAR-IgG, and the infant experienced global developmental delay, generalized seizures, and MRI changes consistent with cortical dysplasia (38).

Anesthesia

There are potential risks for anesthesia in all these patients but as the NMDAR-autoantibodies mimic down regulation of the receptor, in a manner akin to NMDAR antagonists such as ketamine, theoretical and practical concerns are being studied (49).