Encephalitis and antibodies to synaptic and neuronal cell surface proteins

Abstract

The identification of encephalitis associated with antibodies against cell surface and synaptic proteins, although recent, has already had a substantial impact in clinical neurology and neuroscience. The target antigens are receptors and proteins that have critical roles in synaptic transmission and plasticity, including the NMDA receptor, the AMPA receptor, the GABA_B receptor, and the glycine receptor. Other autoantigens, such as leucine-rich glioma-inactivated 1 and contactin-associated protein-like 2, form part of trans-synaptic complexes and neuronal cell adhesion molecules involved in fine-tuning synaptic transmission and nerve excitability. Syndromes resulting from these immune responses resemble those of pharmacologic or genetic models in which the antigens are disrupted. For some immune responses, there is evidence that the antibodies alter the structure and function of the antigen, suggesting a direct pathogenic effect. These disorders are important because they can affect children and young adults, are severe and protracted, occur with or without tumor association, and respond to treatment but may relapse. This review provides an update on these syndromes and autoantigens with special emphasis on clinical diagnosis and treatment.

Over the last few years autoantibodies targeting extracellular epitopes of synaptic receptors and components of trans-synaptic protein complexes have been identified in several forms of autoimmune encephalitis or epilepsy (table).^1–5 The discovery of these autoimmune disorders has changed the diagnostic approach to clinical problems as diverse as catatonia, subacute memory disturbance, seizures, abnormal movements, and limbic encephalitis. For example, some patients previously thought to have viral encephalitis or unusual manifestations of schizophrenia will be found to have a treatable autoimmune disease.⁶ Five features characterize these autoimmune responses: 1) the epitopes are extracellular; 2) the antibody binding is visible in cells transfected with the target antigen; 3) for all the disorders studied, the antibodies alter the structure or function of the corresponding neuronal antigen^1,2; 4) the effects of the antibodies are often reversible; and 5) the clinical picture resembles that of pharmacologic or genetic models in which the antigen is disrupted. Failure to fulfill these criteria should bring into question the identity of the antigen.⁴ In this review, we provide an update on these disorders and discuss the clinical implications of the ongoing discovery of neuronal self-antigens.

Table.

Clinical features of encephalitis associated with antibodies to neuronal cell surface antigens

Abbreviations: AChR = acetylcholine receptor; AMPAR = α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor; ANA = antinuclear antibody; Caspr2 = contactin-associated protein-like 2; FLAIR = fluid-attenuated inversion recovery; GABA_B-R = γ-aminobutyric acid-B receptor; GAD₆₅ = glutamic acid decarboxylase 65; LGI1 = leucine-rich glioma-inactivated 1; MuSK = muscle-specific kinase; TPO = thyroid peroxidase; VGCC = voltage-gated calcium channel.

^aClassic limbic encephalitis: subacute confusion, depression, irritability, short-term memory loss, and seizures suggestive of medial temporal lobe dysfunction, usually associated with MRI abnormalities in the medial temporal lobes, without clinical or radiologic evidence of other areas of the brain involved.

^bAbnormal findings refer to lymphocytic pleocytosis that may be accompanied by a moderate increase in protein concentration and CSF-specific oligoclonal bands.

^cRelative frequency is an estimate of the authors based on studies of 700 patients with encephalitis and antibodies against neuronal cell surface antigens. Approximately 9% of patients had antibodies against other cell surface antigens, including glycine receptor or proteins pending to be characterized (unpublished data, 2011).

SYNAPTIC PROTEINS AS AUTOANTIGENS OF CNS DISORDERS

The discovery of antibodies to cell surface or synaptic proteins of the CNS emerged from studies of limbic encephalitis. This disorder was first reported in the 1960s,^e1 and the concept of an immune-mediated pathogenesis gained relevance after anti-Hu and other onconeuronal antibodies against intracellular antigens were identified.^e2-e4 However, cytotoxic T-cell mechanisms are considered to be the main effectors of these immune responses^e5,e6 (reviewed by Tuzun and Dalmau⁷). The implication of cell surface autoantibodies came from the identification of antibodies believed to be directed against the voltage-gated potassium channel (VGKC) in some patients with limbic encephalitis.^8,9 Because this disorder infrequently is associated with cancer, patients with nonparaneoplastic limbic encephalitis were categorized as VGKC antibody–positive or –negative, the later being perceived as a disorder of uncertain etiology and without clear treatment recommendations.^e7 This led Ances et al.¹⁰ to optimize immunohistochemical techniques with rodent brain and cultures of rat hippocampal neurons to visualize cell surface autoantibodies (figure 1). These studies revealed subsets of patients, previously considered “seronegative,” who had antibodies against the neuropil of brain and patterns of immunolabeling resembling those of synaptic proteins.⁷ Immunoprecipitation and sequence analysis of these proteins provided the identity of the antigens, and their expression in human embryonic kidney cells resulted in specific diagnostic tests.^2,3,11

Figure 1. Comparative analysis of antibodies to intracellular and cell surface antigens.

Consecutive sections of rat hippocampus immunostained with CSF of a patient with antibodies against an intracellular antigen (Hu, A) and CSF of a patient with antibodies against a cell surface synaptic antigen (NMDA receptor [NMDAR], B). (C, D) Framed areas shown at higher magnification. Compared with the intracellular antigen, the cell surface synaptic antigen is demonstrated as intense neuropil staining, sparing neuronal cell bodies (nuclei of neurons mildly counterstained with hematoxylin). Using live, nonpermeabilized cultures of dissociated rat hippocampal neurons, the Hu antibody does not show reactivity due to lack of penetration into the neuron (E), whereas the NMDAR antibody shows intense neuronal cell surface immunolabeling (F), indicating that it recognizes an extracellular epitope (nuclei of neurons counterstained in blue with 4′,6-diamidino-2-phenylindole). (A–D) Immunoperoxidase method (A, B 10; C, D 400). (E, F) Immunofluorescence method (800 oil lens). (A–D) Modified, with permission, from Tuzun and Dalmau⁷ (Limbic encephalitis and variants: classification, diagnosis and treatment. Neurologist 2007;13:261–271).

The incidence of these disorders is unknown, but collectively they are at least 5 times more frequent than all encephalitis associated with classic paraneoplastic antibodies, including Hu, CRMP5, Ma2, and amphiphysin. The table shows the relative frequency of each autoimmune synaptic disorder.

ENCEPHALITIS ASSOCIATED WITH ANTIBODIES AGAINST NMDA RECEPTOR

Discovered in 2007,¹ this disorder has become one of the most common types of autoimmune encephalitis associated with antibodies against a neuronal antigen.^11–13 This fact is supported by studies from intensive care,^14,e8,e9 neurology,¹ and pediatric departments.^15,e10 A multicenter, population-based study of causes of encephalitis showed that 4% of the patients had NMDA receptor (NMDAR) antibodies, making this the second most common immune-mediated cause after acute disseminated encephalomyelitis.¹⁶ A retrospective analysis of all patients aged between 18 and 35 years admitted to an intensive care unit for encephalitis of unknown etiology demonstrated that 1% had NMDAR antibodies.^14,e11

The associated syndrome is different from limbic encephalitis and usually develops as a multistage process of illness and recovery.^11–13 Approximately 70% of patients have prodromal headache, fever, or other symptoms that may resemble a viral process. Within a few days, patients develop psychiatric manifestations such as anxiety, insomnia, agitation, bizarre behavior, hallucinations, or delusions. These symptoms are usually followed by orofacial and limb dyskinesias, choreoathetosis, oculogyric crisis, dystonia, rigidity, and opisthotonic postures in the context of catatonia, coma, and autonomic or breathing instability. Seizures can occur at any time during the disease but predominate in early stages. In children, the psychiatric problems are more difficult to identify, and patients often present with hyperactivity, irritability, insomnia, agitation, temper tantrums, seizures, or focal or dystonic deficits.¹⁵ Partial syndromes in which one of the symptoms (e.g., psychosis, movement disorder, or seizures) largely predominates can occur.

A review of 400 patients showed that the likelihood of a tumor varies with patients' age, gender, and ethnicity.¹³ Whereas in women older than 18 years the chance of having unilateral or bilateral ovarian teratomas was approximately 55%, only 15% of women younger than 14 years had teratomas. Approximately 5% of men had testicular germ-cell tumors or other neoplasms. Tumors other than teratoma were identified in only 7 patients (2%). Single case reports of neuroblastoma and Hodgkin lymphoma have been published.^e12,e13 Analysis of 201 consecutive patients showed that African Americans had ovarian teratomas more frequently than did whites.¹³

Prompt immunotherapy and, in patients with teratoma, tumor removal are associated with substantial neurologic recovery (discussed later). The mortality rate is 4%, the main causes being respiratory, cardiac, and infectious problems that usually occur at the stage of intensive care support.¹³

Patients' antibodies target an extracellular epitope region located in the N-terminal domain of the NR1 subunit of the NMDAR.¹¹ By using cultures of rat hippocampal neurons and in vivo infusion of patients' CSF or immunoglobulin G (IgG) into rodent hippocampus, antibodies cause cross-linking and internalization of the target receptors, accompanied by a decrease in synaptic NMDAR-mediated currents.^17,e14 The effects of NMDAR antibodies resemble those of phencyclidine, ketamine, and other NMDAR antagonists,^e15-e17 and the clinical stages of the syndrome probably arise as a consequence of an antibody-mediated progressive decrease of NMDAR clusters and function, followed by a gradual restoration of receptor function during recovery. A model that accounts for most of the clinical manifestations of the disorder has been suggested.¹³ According to this model, a decrease of NMDAR in inhibitory GABAergic neurons (which contain high levels of NMDAR) and glutamatergic synapses causes effects at multiple levels including disinhibition of excitatory pathways and an increase of extracellular glutamate resulting in a frontostriatal syndrome (psychosis, catatonia, mutism, rigidity, and dystonia),^e18,e19 disinhibition of brainstem central pattern generators causing complex elaborate movements and dyskinesias,¹⁸ and disruption of the medullary-pontine respiratory network resulting in breathing dysfunction.^e20 This model is supported by studies showing a dramatic increase of extracellular glutamate after intracerebral injection of patients' antibodies in rodents.¹⁹ Autopsies of patients show infiltrates of B cells and plasma cells, moderate amounts of T cells, deposits of IgG (figure 2), and decreased levels of NMDAR but no evidence of complement deposition, supporting the concept that the symptoms of the disorder arise from functional alteration of the synapses.^1,17,20

Figure 2. Presence of plasma cells and immunoglobulin G (IgG) but not complement in the brain of patients with anti-NMDA receptor (NMDAR) encephalitis.

Demonstration of plasma cells (A), deposits of IgG (B, diffuse brown staining of neuropil), and absence of complement (D) in the brain from the autopsy of a patient with anti-NMDAR encephalitis. (C) Corresponds to the hippocampus of a neurologically normal, NMDAR antibody–negative individual; note the absence of IgG compared with B. The lack of complement in the patient's brain is in contrast with the detection of complement in the associated ovarian teratoma (E). Cells and processes contained in the tumor react with MAP-2, a neuronal dendritic marker (F). Specific markers used include anti-CD138 (specific for plasma cells, A), anti-human IgG (B, C), anti-C9neo (specific for activated complement, D, E), and anti-MAP-2 (specific for dendrites, F). Immunoperoxidase technique, all panels 400.

ENCEPHALITIS ASSOCIATED WITH ANTIBODIES AGAINST AMPA RECEPTOR AND GABA_B RECEPTOR

Antibodies against α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR) and γ-aminobutyric acid-B receptor (GABA_B-R) are associated with a syndrome of limbic encephalitis that overlaps with that of antibodies against the neuronal-secreted protein leucine-rich glioma-inactivated 1 (LGI1; previously considered VGKC).^2–4,21 Patients develop similar MRI and CSF findings, although those with LGI1 antibodies may have milder pleocytosis or normal CSF (figure 3).⁴ However, when LGI1 antibodies are detectable in serum they are always present in the CSF, even when this is considered to be normal (Dalmau, unpublished observations, 2011). Some demographic and clinical features are different among these disorders (table).

Figure 3. MRI of patients with limbic encephalitis and different antibodies to cell surface antigens.

The MRI scans correspond to patients with LGI1 antibodies (A), AMPA-R antibodies (B), and GABA_B-R antibodies (C).

A study examining patients with small-cell lung cancer and limbic encephalitis without onconeuronal antibodies identified GABA_B-R antibodies in 7 of 10 patients (the other 3 had AMPAR antibodies). The conclusion was that GABA_B-R autoimmunity accounts for most of the small-cell lung cancer–associated limbic encephalitis previously considered “seronegative.”²²

Autoimmunity to AMPAR or GABA_B-R may occur in association with antibodies against intracellular antigens, such as thyroid peroxidase, glutamic acid decarboxylase 65 (GAD₆₅), SOX1, and antinuclear antibodies, and also with antibodies against cell surface antigens such as N-type voltage-gated calcium channels.^2,3 These associated antibodies probably reflect a tendency to autoimmunity or an antitumor immune response. The co-occurrence of GABA_B-R antibodies was recently examined in 71 patients with GAD₆₅ antibody–associated nonparaneoplastic disorders and 11 patients with paraneoplastic disorders (5 with limbic encephalitis). GABA_B-R antibodies were only encountered in 3 of the patients with paraneoplastic disorders (2 with limbic encephalitis), suggesting that these antibodies should be suspected in patients with paraneoplastic limbic encephalitis attributed to GAD₆₅ antibodies.²²

AMPAR antibodies target extracellular epitopes of glutamate receptor type 1 or type 2 subunits, cause receptor cross-linking and internalization, and result in a reversible decrease in AMPAR clusters at synapses.² Antibodies to the metabotropic GABA_B-R inhibit receptor function but do not cause receptor internalization (Lancaster, unpublished observations, 2010).

ANTIBODIES TARGETING TRANS-SYNAPTIC AND NEURONAL CELL ADHESION MOLECULES (LGI1, Caspr2)

Over the last 10 years, multiple studies suggested that limbic encephalitis, peripheral nerve hyperexcitability (acquired neuromyotonia or Isaacs syndrome), and Morvan syndrome were all related to VGKC antibodies.^8,9 However, recent experiments could not demonstrate that the target antigens were the VGKCs, and Lai et al.,⁴ using serum of patients with limbic encephalitis, immunoprecipitated LGI1 as the main autoantigen. In the same study, the serum of a patient with encephalitis and peripheral neuropathy precipitated contactin-associated protein-like 2 (Caspr2), an antigen that had been related to neuromyotonia and Morvan syndrome.²³ These findings, confirmed by other investigators,²¹ indicate that at least 2 molecular targets (LGI1 and Caspr2) are the antigens of antibodies previously attributed to VGKCs.

LGI1 antibodies.

LGI1 is a secreted protein that interacts with 2 epilepsy-related proteins (presynaptic ADAM23 and postsynaptic ADAM22), organizing a trans-synaptic protein complex that includes the presynaptic Kv1.1/Kv1.2 potassium channels and postsynaptic AMPAR scaffolds (figure 4).²⁴ Studies using linkage analysis revealed that mutations of LGI1 cause autosomal-dominant partial epilepsy with auditory features,^e21,e22 also known as autosomal-dominant lateral temporal lobe epilepsy (ADLTE).^e23 This is an inherited epileptic syndrome associated with partial seizures and auditory or visual hallucinations. A transgenic mouse expressing a mutant LGI1 identified in human ADLTE showed inhibition of dendritic pruning and increased spine density, resulting in a marked increase of excitatory synaptic transmission.²⁵ In mice with deleted LGI1, this increase in excitability has been attributed to a decrease in AMPAR function in inhibitory neurons,^e24 and to an increase in release of glutamate.²⁶ Knockout models of LGI1, ADAM22,^e24,e25 ADAM23,^e26 or Kv1^e27 result in severe epileptic phenotypes and premature death, suggesting that these proteins are genetically and functionally related.⁴

Figure 4. Interaction of leucine-rich glioma-inactivated 1 (LGI1) with presynaptic and postsynaptic proteins.

LGI1 is a secreted neuronal protein that interacts at the presynapse with ADAM23 and at the postsynapse with ADAM22. The cartoon is based on studies indicating that LGI1 coprecipitates with other proteins including the presynaptic Kv1 potassium channels and a variety of presynaptic and postsynaptic scaffolding proteins.^24,e53 It has been postulated that LGI1 connects presynaptic and postsynaptic protein complexes for finely tuned synaptic transmission. AMPAR = α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor.

Lai et al.⁴ proposed that in patients with LGI1 antibodies, an immune-mediated disruption of LGI1 function causes increased excitability, resulting in seizures and limbic encephalopathy. Most patients with LGI1 antibodies develop the classic syndrome of limbic encephalitis previously attributed to VGKCs.^8,e28 Some patients present with a rapidly progressive dementia that is treatable and is similar to that in previous reports of patients with VGKC antibodies, mimicking Creutzfeldt-Jakob disease.^27,e29

The fact that anti-LGI1–associated symptoms are different from the phenotype associated with mutations of the gene is not surprising, given that some of these mutations alter the postnatal maturation of presynaptic and postsynaptic functions, including glutamatergic circuits.²⁵ However, knowledge of the main autoantigen has clarified the origin of some symptoms and improved treatment. For example, at least 40% of patients with LGI1 antibodies have myoclonic-like movements,⁴ which in some cases may further suggest Creutzfeldt-Jakob disease.²⁷ These movements have been described as “twitches,” “myoclonus,” “stereotyped brief monomorphic movements,” and “seizure-like episodes.”^28,e30 On the other hand, LGI1-null mutant mice show myoclonic seizures that may be preceded by a brief “flagpole” tail dorsiflexion probably representing brief tonic seizures.^26,e31 Using continuous video-EEG recording, these observations led to the demonstration that some of the indicated twitches or myoclonic-like movements in patients with LGI1 antibodies are, in fact, tonic seizures.²⁹ Considering that these seizures may precede other symptoms of limbic dysfunction, their recognition and prompt treatment, including immunotherapy, are important.

Caspr2 antibodies.

Caspr2 is a member of the neurexin superfamily, which mediates cell–cell interactions and has a critical role in concentrating VGKCs in the juxtaparanodal region of myelinated axons.^30,e32 It is also expressed at high concentrations in the hippocampus and cerebellum. Mutations and polymorphisms of CNTNAP2, the gene that codes for Caspr2, have been linked to schizophrenia, psychosis, intractable focal seizures, autism, mental retardation, and cortical dysplasia.^e33-e35 Peripheral nerve hyperexcitability has been observed in some patients with CNTNAP2 mutations.³¹

Patients with Caspr2 antibodies develop encephalitis, peripheral nerve hyperexcitability, or Morvan syndrome.^21,31 The index patient, whose serum was used to precipitate this protein, had limbic encephalitis and thermal allodynia. Among 7 additional patients, 4 had symptoms involving both the CNS (encephalitis and seizures) and peripheral nervous system (hyperexcitability and neuropathy), 2 had encephalitis, and one had neuromyotonia. In 3 patients, the presence of fasciculations and symptoms related to overlapping immune responses, such as bulbar and limb weakness associated with muscle-specific kinase (MuSK) and acetylcholine receptor (AChR) antibodies, initially suggested a motor neuron disorder.³¹ Diagnosis of these patients is important because they recover with immunotherapy.^e36

Although VGKC (now Caspr2) antibodies have been linked to acquired neuromyotonia, many patients do not have these antibodies, suggesting other yet unknown autoimmune mechanisms.^31,32

RARE SYNDROMES OF AUTOIMMUNITY TO SYNAPTIC PROTEINS

In addition to the syndromes described above, other less common disorders of synaptic autoimmunity have been reported. Antibodies to metabotropic glutamate receptor type 1 were described in 3 patients with cerebellar ataxia, 2 of them having a history of Hodgkin lymphoma.^33,e37 Intrathecal injection of antibodies to rodents caused ataxia, suggesting that they are pathogenic.^33,e38 In contrast, antibodies to mGluR5, which is highly homologous to mGluR1 but instead of being predominantly expressed in the cerebellum is expressed in the hippocampus, associate with the Ophelia syndrome or limbic encephalitis related to Hodgkin lymphoma (Lancaster et al., unpublished data, 2011).

Antibodies to the α1 subunit of the glycine receptor (GlyR) occur in some patients with progressive encephalomyelitis with rigidity and myoclonus or acquired hyperekplexia and atypical stiff-person or stiff-limb syndrome without GAD₆₅ antibodies.^5,34 Symptoms associated with GlyR antibodies bear a remarkable resemblance to those resulting from genetic mutations of GlyR subunits and associated proteins (hereditary hyperekplexia; OMIM 149400) or from poisoning with the GlyR antagonist strychnine.^e39 In a report of 3 patients, atypical symptoms included alterations of behavior and sleep, seizures, trismus, and neurogenic pruritus. None of the patients had cancer, and only one had substantial neurologic recovery.³⁴

TREATMENT

Immunotherapy.

There is no standard of care for patients with these disorders; however, most experts agree that detection of antibodies to cell surface or synaptic receptors should prompt the use of immunotherapy while screening for a tumor is conducted.^e40 Many patients are initially treated with first-line immunotherapies, including corticosteroids, IV immunoglobulin (IVIg), plasma exchange, or a combination of these. There are no data supporting one treatment over another. We prefer IVIg over plasma exchange in patients with anti-NMDAR encephalitis. This is due to the young age of many patients and the frequent development of disease-related limitations such as psychosis, extreme agitation, autonomic instability, or nosocomial infections. Between 60% and 80% of patients with AMPAR, GABA_B-R, or LGI1 antibodies respond to these treatments.^2–4,21 For patients with Caspr2 antibodies, the response rate is probably lower, and these patients may need more aggressive immunotherapy.^21,31 For patients with anti-NMDAR encephalitis, the response varies, depending on the presence of a teratoma. A review of 105 patients showed that although there was no difference between the proportions of patients with or without tumor who eventually achieved substantial recovery, patients with tumor responded to first-line treatment (tumor removal and corticosteroids, IVIg, or plasma exchange) more frequently than did patients without tumor who received the same type of immunotherapy. Patients without tumor required second-line immunotherapy (rituximab, cyclophosphamide, or both) more often than those with tumor.¹³

Tumor association.

The likelihood of identifying a tumor varies with the type of antibody and sometimes with the patient's age, gender, and ethnicity, as shown in anti-NMDAR encephalitis.¹³ The main diagnostic tests to demonstrate occult tumors were recently reviewed.³⁵ Fluorodeoxyglucose-PET is useful for detection of many occult malignancies but has limited utility for ovarian teratomas. For this type of tumor, MRI of the abdomen and pelvis is the test of choice, followed by CT and abdominal or transvaginal ultrasound (if age-appropriate).^e41 A common quandary is how aggressive one should be after all test results are negative for the presence of a teratoma. This dilemma is emphasized by the young age of most patients and reports of patients who had occult teratomas after exploratory laparoscopy or oophorectomy^36,e42 and others who did not have a tumor.³⁷ Although the answer depends on prospective and long-term follow-up studies, our approach is to avoid surgery if results of all tumor-screening studies are negative. However, any small cystic and persistent abnormality of the ovary, hemorrhagic or not, should be viewed with a high index of suspicion, and in these patients, we recommend cystectomy or oophorectomy. In patients with anti-NMDAR encephalitis, the teratomas are frequently small and asymptomatic, and the tumor markers (CA125, β-human chorionic gonadotropin, and α-fetoprotein) are often negative. Therefore, the standard criteria to remove ovarian teratomas based on size, local symptoms, and ultrasound and tumor marker findings suggesting malignancy should not be applied to patients with anti-NMDAR encephalitis.

Outcome and relapses.

General concepts about treatment and outcome of classic paraneoplastic syndromes of the CNS do not apply to encephalitides with antibodies to cell surface antigens. For example, whereas classic paraneoplastic syndromes do not respond to immunotherapy unless the tumor is successfully treated, and even then the responses are very limited,^e43 the encephalitides with antibodies to cell surface antigens may respond to immunotherapy before the tumor is identified and treated or improve spontaneously.^1,38 This should not discourage physicians from searching for and treating a tumor if found, because tumor removal accelerates improvement and decreases relapses.^11–13 Relapses occur in 20%–25% of patients with anti-NMDAR encephalitis, predominantly in those without tumor¹¹ or in those who had a rapid immunotherapy taper¹⁵ or prior suboptimal immunotherapy¹² (Graus, personal communication). Relapses are also frequent in syndromes associated with AMPAR² and Caspr2 antibodies³¹ and occurred in 18% of patients in a series with LGI1 antibodies.⁴

Patients with anti-NMDAR encephalitis, mainly if they are older than 18 years,¹³ and patients with anti-AMPAR and GABA_B-R encephalitis should have regular tumor screening for at least 2 years, even if they have recovered neurologically.³⁹ This recommendation is based on the frequent presence of tumors in these disorders, and reports of patients who had a teratoma identified several months or years after recovery from anti-NMDAR encephalitis or at neurologic relapse.^38,e44

Patients with anti-NMDAR encephalitis usually have prolonged hospitalizations (average 3 months), followed by several additional months of physical and cognitive rehabilitation.^11,12,15 In contrast, patients with LGI1, AMPAR, or GABA_B-R antibodies often have shorter hospitalizations and faster responses to treatment, albeit not necessarily better long-term outcomes.^2–4,e45

For all these disorders, neurologic improvement usually correlates with a decrease in serum and CSF antibody titers.^11,12,14 Nevertheless, highly sensitive and specific methods show that after substantial recovery, patients may still have detectable antibodies in serum or CSF.¹³ These patients usually have a decline in intrathecal synthesis of antibodies as opposed to patients who do not improve.^14,40

ALGORITHMIC APPROACH TO DIAGNOSIS AND TREATMENT

On the basis of previous studies and personal experience, an algorithmic approach to diagnosis and treatment of autoimmune encephalitides is proposed in figure 5. Patients with rapidly progressive encephalitis suspected to be autoimmune because of the type of syndrome, MRI and CSF findings, or other associations (e.g., presence of a tumor) should have serum and CSF tested for antibodies to intraneuronal and cell surface antigens. Detection of antibodies to onconeuronal antigens usually carries a poor prognosis because of the frequent association with aggressive tumors, cytotoxic T-cell mechanisms (amphiphysin may be an exception^e46), and irreversible neuronal degeneration.^e43 Less than 10% of these patients have substantial or full recoveries.^e47-e49 In contrast, detection of antibodies to cell surface antigens carries a better prognosis, and this should lead to prompt immunotherapy while screening for an underlying tumor. If, after first-line immunotherapy (corticosteroids and IVIg or plasma exchange), there is no sign of improvement, second-line immunotherapy with rituximab, cyclophosphamide, or both should be considered.¹³ The duration of treatment depends on the clinical response and discretion of the physician. Limited or slow improvement usually is associated with persistently high antibody titers, predominantly in CSF of patients with intrathecal synthesis, and these patients may benefit from continued second-line immunotherapies.⁴⁰ In subphenotypes with a tendency to relapse (anti-NMDAR encephalitis without tumor association or Casp2 or AMPAR antibody–associated encephalitis), chronic immunosuppression with azathioprine or mycophenolate mofetil should be considered.

Figure 5. Algorithmic approach to diagnosis and treatment of encephalitis with antibodies to intracellular and cell surface neuronal antigens.

See discussion in text. *Differential diagnosis and exclusion of other disorders has been examined in several reviews.^7,13,e54 **Unknown antigens refer to antigens whose identity has not been established but are visible using patients' antibodies with brain immunohistochemistry and cultures of neurons (as in figure 1, B, D, and F). †T-cell suppression refers to strategies focused on decreasing T-cell activation (rituximab) and cytotoxic T-cell mechanisms (cyclophosphamide, tacrolimus, or cyclosporine). ††As indicated in text, tumor surveillance and chronic immunosuppression should be considered in some disorders (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor [AMPAR], γ-aminobutyric acid-B receptor [GABA_B-R], and contactin-associated protein-like 2 [Caspr2]) and subgroups of patients with anti-NMDA receptor (NMDAR) encephalitis with higher risk for relapse (e.g., patients without tumor) or to have an underlying tumor (older than 18 years). GlyR = glycine receptor; IVIg = IV immunoglobulin; LGI1 = leucine-rich glioma-inactivated 1.

There are patients with encephalitis suspected to be autoimmune whose serum and CSF react with neuronal-specific cell surface antigens of unknown identity. This finding strongly supports an immune-mediated disorder and the possibility of response to immunotherapy.^10,e50 However, not all antibodies to cell surface antigens predict a good outcome. For example, a patient with fulminant encephalitis, refractory seizures, and brain edema that resulted in death had antibodies against a member of the α-neurexins (Dalmau and Glaser, unpublished data, 2009). These proteins are synaptic cell adhesion molecules that organize presynaptic terminals and connect with postsynaptic neuroligins.^e51 Neurexins are target receptors for α-latrotoxin, a black widow spider venom that induces massive neurotransmitter release.^e52 Thus, given the critical role of neurexins in the assembly of synapses, any antibody-mediated attack should be expected to have dramatic effects, as occurred in our patient.

FUTURE STUDIES

Future studies should focus on better defining the spectrum of symptoms of recently identified disorders and the immune responses of those considered to be seronegative, such as many patients with acquired neuromyotonia.³² The specific antigens and immune mechanisms that underlie poorly defined syndromes should be investigated, including steroid-responsive encephalitis, Hashimoto encephalitis, encephalitis lethargica, or pediatric autoimmune neuropsychiatric disorders associated with streptococci. The frequency of encephalitis associated with NMDAR or LGI1 antibodies is high enough to proceed with trials to determine the best treatment approach and immunologic follow-up. Studies with animal models may help to better understand how antibodies affect memory, behavior, and cognition and cause seizures.

AUTHOR CONTRIBUTIONS

J.D. provided the study concept and designed the study. E.L. and J.D. acquired data. E.L. supervised the study. E.M.-H. analyzed or interpreted data. J.D. contributed vital reagents, tools, and patients. E.L., E.M.-H., and J.D. drafted/revised the manuscript. J.D. obtained funding.

DISCLOSURE

Dr. Lancaster has received research support from Talecris Biotherapeutics (training grant) and the Dana Foundation (Neuro-immunology Award). Dr. Martinez-Hernandez receives research support from Instituto de Salud Carlos III, Fondo de Investigaciones Sanitarias, Spain. Dr. Dalmau is ICREA (Institució Catalana de Recerca i Estudis Avançats) Research Professor at IDIBAPS, Hospital Clinic, University of Barcelona. Dr. Dalmau serves on the editorial board of Neurology® and receives royalties from the editorial board of Up-To-Date; has filed a patent application for the use of LGI1 as a diagnostic test; has received royalties from Memorial Sloan-Kettering Cancer Center for a patent re: Ma2 autoantibody test and has patents pending re: NMDA and GABA_B receptor autoantibody tests (license fee payments received from EUROIMMUN AG); and receives research support from EUROIMMUN AG, the NIH/NCI, and the McKnight Foundation.