Anti-NMDA-receptor encephalitis: case series and analysis of the effects of antibodies

Abstract

Background: A severe form of encephalitis associated with antibodies against NR1-NR2 heteromers of the NMDA receptor was recently identified. We aimed to analyse the clinical and immunological features of patients with the disorder and examine the effects of antibodies against NMDA receptors in neuronal cultures.

Methods: We describe the clinical characteristics of 100 patients with encephalitis and NR1-NR2 antibodies. HEK293 cells ectopically expressing single or assembled NR1-NR2 subunits were used to determine the epitope targeted by the antibodies. Antibody titres were measured with ELISA. The effect of antibodies on neuronal cultures was determined by quantitative analysis of NMDA-receptor clusters.

Findings: Median age of patients was 23 years (range 5-76 years); 91 were women. All patients presented with psychiatric symptoms or memory problems; 76 had seizures, 88 unresponsiveness (decreased consciousness), 86 dyskinesias, 69 autonomic instability, and 66 hypoventilation. 58 (59%) of 98 patients for whom results of oncological assessments were available had tumours, most commonly ovarian teratoma. Patients who received early tumour treatment (usually with immunotherapy) had better outcome (p=0.004) and fewer neurological relapses (p=0.009) than the rest of the patients. 75 patients recovered or had mild deficits and 25 had severe deficits or died. Improvement was associated with a decrease of serum antibody titres. The main epitope targeted by the antibodies is in the extracellular N-terminal domain of the NR1 subunit. Patients' antibodies decreased the numbers of cell-surface NMDA receptors and NMDA-receptor clusters in postsynaptic dendrites, an effect that could be reversed by antibody removal.

Interpretation: A well-defined set of clinical characteristics are associated with anti-NMDA-receptor encephalitis. The pathogenesis of the disorder seems to be mediated by antibodies.

Figure 1. Immunohistochemical criteria for the presence of NR1–NR2B antibodies

Sera and CSF from all patients’ with anti-NMDA-receptor encephalitis showed identical antibody reactivity in three different assays. Coronal section of rat brain incubated with a representative CSF (A) shows intense reactivity predominantly involving the hippocampus. Cultures of non-permeabilised live rat hippocampal neurons (B) incubated with the same CSF show extensive cell-surface immunolabelling. HEK293 cells transfected with NR1 and NR2B (forming NR1–NR2B heteromers of the NMDA receptor) show intense reactivity with patients’ CSF (C); this reactivity co-localises (D) with the reactivity of a monoclonal rabbit antibody against NR1 (E). Immunofluorescence method, nuclei of cells shown with 4′,6-diamidino-2-phenylindole (DAPI). A ×25; B ×800 oil lens; C–E ×400.

Figure 2. Response to treatment

Patients whose tumour was diagnosed and treated within 4 months of neurological symptom development had better outcomes (full recovery or mild deficits) than those whose tumour was treated after 4 months of neurological symptom development or not treated (p=0.03), those without tumour (p=0.006), and these two groups combined (p=0.004). See webtable 1 for more details.

Figure 3. Analysis of NR1 antibody titres

In 53 patients with anti-NMDA-receptor encephalitis, antibody titres were higher in CSF than in serum (A). In 83 patients with anti-NMDA-receptor encephalitis (54 with tumour, 29 without tumour) and 50 controls (B), those with tumours had higher titres than those without (Wilcoxon rank, p<0.0001) and controls (p<0.0001). Six patients (one with tumour, five without tumour) had very low ELISA readings that overlapped with the signal given by negative controls. These six patients had low antibody titres; in contrast, the 50 controls were negative. Solid lines indicate the mean of the titres in each group. The dotted line indicates three SD above the mean value given by background signal of negative controls. Follow-up of serum antibody titres (C) in 14 representative patients who had neurological improvement (black lines) and four who did not (red lines); the second time-point is the sample obtained at the last follow-up (median 5.6 months, range 2–83 months). The dotted line indicates three SD of the mean value given by background signal of 50 negative control sera. Similar results were obtained by ELISA with NR1–NR2 heteromers (data not shown). Values in A, B, and C are given in relative fluorescence units (rfu) from the ELISA reader, and plotted in a logarithmic scale.

Figure 4. Immunolabelling of neuronal NR1 clusters

Left: Hippocampal neurons (14 days in vitro) immunostained with antibody against NR1 (black and white and green) or a patient’s CSF (black and white and red). Right: 91% of NR1 clusters are colabelled with patient’s CSF (yellow puncta in overlay), less than 9% of NR1-positive puncta remain unlabelled (green in overlay). Kruskal-Wallis non-parametric ANOVA followed by Dunn’s pairwise comparison, p<0.01.

Figure 5. Effect of antibodies on the number of NMDA-receptor clusters in live neurons

Representative immunoblot (A) of neuronal surface NR1 detection after incubation with control or patients’ IgG (μg/mL). Protein concentrations were normalised to concentrations of GABA_A receptor subunit a2. Hippocampal neurons (B) cultured with control CSF or patients’ CSF from day 7 to day 14 in vitro (7 day treatment), or with patients’ CSF from day 7 to day 10 in vitro followed by control CSF from day 11 to day 14 in vitro (3 day treatment and 4 day recovery), then immunostained for NR1 (18–36 cells from each of three experiments). Boxed areas are shown below at higher magnification. Fewer NR1-labelled clusters (C) were found in cultures treated with patient CSF for 3 or 7 days compared with those treated with control CSF or cultures treated with patient CSF followed by 4 days recovery (18–36 cells from each of three experiments; Kruskal-Wallis non-parametric ANOVA, Dunn’s pairwise comparison, p<0.01). Incubation with patients’ CSF for 3 days or 7 days did not affect the number of PSD-95 clusters (D; 18–36 cells from each of three experiments).