Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 141 (6), 1691-1702

Ion Channels in EEG: Isolating Channel Dysfunction in NMDA Receptor Antibody Encephalitis


Ion Channels in EEG: Isolating Channel Dysfunction in NMDA Receptor Antibody Encephalitis

Mkael Symmonds et al. Brain.


See Roberts and Breakspear (doi:10.1093/brain/awy136) for a scientific commentary on this article.Neurological and psychiatric practice frequently lack diagnostic probes that can assess mechanisms of neuronal communication non-invasively in humans. In N-methyl-d-aspartate (NMDA) receptor antibody encephalitis, functional molecular assays are particularly important given the presence of NMDA antibodies in healthy populations, the multifarious symptomology and the lack of radiological signs. Recent advances in biophysical modelling techniques suggest that inferring cellular-level properties of neural circuits from macroscopic measures of brain activity is possible. Here, we estimated receptor function from EEG in patients with NMDA receptor antibody encephalitis (n = 29) as well as from encephalopathic and neurological patient controls (n = 36). We show that the autoimmune patients exhibit distinct fronto-parietal network changes from which ion channel estimates can be obtained using a microcircuit model. Specifically, a dynamic causal model of EEG data applied to spontaneous brain responses identifies a selective deficit in signalling at NMDA receptors in patients with NMDA receptor antibody encephalitis but not at other ionotropic receptors. Moreover, though these changes are observed across brain regions, these effects predominate at the NMDA receptors of excitatory neurons rather than at inhibitory interneurons. Given that EEG is a ubiquitously available clinical method, our findings suggest a unique re-purposing of EEG data as an assay of brain network dysfunction at the molecular level.


Figure 1
Figure 1
Spectral characteristics of NMDAR-antibody encephalitis. (A) Left: EEG, scalp data from mid-frontal channel Fz from exemplar patients in each group. Right: Though NMDAR-antibody encephalitis can present with a range of clinical and electrographic features there was no statistical difference in the clinical severity of encephalopathy (West-Haven scale) applied to the two groups. Inset shows merged fluorescence image of cell-based assay of positive NMDAR antibodies in serum. (B) On the far left, scalp maps indicate the 21 sensor locations and power estimates from two exemplar subjects. In the left panel we show the spontaneous delta-theta and high beta-gamma power averaged across the groups. Given previous reports of altered low (delta/theta) and high (beta/agama) frequency power in patients with NMDAR-antibody encephalitis, we tested these ranges and found a significant effect of group [right panel (log-log plot of spectral power)]. For delta responses the spectral differences were driven by enhanced power from 2 to 4 Hz in the control encephalopathic patients (ANOVA: F = 6.0, P = 0.004). For the beta/gamma band it was the NMDAR-antibody encephalitis group that exhibited the greatest power (ANOVA: F = 3.63; P = 0.03). The average response over the band is plotted for each subject (x’s) with inverted triangles indicating the group mean.
Figure 2
Figure 2
The DCM. (A) Sources for the DCM are based on MEG ‘default mode’ studies and comprise two parietal and two frontal sources. These are connected with forward connections (from supragranular layers) from parietal to frontal cortex and with backward connections (from infragranular layers) from frontal to parietal cortex. (B) Each source is populated with a four-population neural mass model comprising supragranular pyramidal cells, inhibitory interneurons, layer IV stellate cells and infragranular pyramidal cell layer. Receptors and intrinsic connections represented in the neural mass models are shown. Our dynamics prescribe changes in postsynaptic membrane potential based on the dynamics of ion channel transmission (Supplementary Fig. 4). These ions are controlled by conductances representing binding to GABAA, NMDA and AMPA receptors. Each of these receptors is imbued with a physiologically plausible time constant, which acts as the inverse of a rate constant controlling the rate of opening and closing. Each channel also has its own reversal potential set at physiological levels (see text). (C) The DCMs produce spectra, which recapitulate the patterns of beta-gamma and delta-theta responses observed in the empirical recordings. Importantly, no group effects of fit were observed. In other words, the model was equally applicable to all data. C = patient controls; E = other encephalopathy; N = NMDA encephalitis.
Figure 3
Figure 3
Receptor fingerprints from DCM parameters. (A) Our DCMs contained parameters relating to three major receptor types, namely NMDA, AMPA and GABAA. Here we show the linear combination of each parameter set that best described the partition amongst the groups. Our plots show each parameter scaled by the canonical vector (Supplementary Fig. 1). Each of our patients are represented by either a red sphere (NMDAR-antibody encephalitis diagnosis) or a grey sphere (for both other encephalopathy and neurological patient controls). Statistically, only the NMDA parameter sets showed any significant multivariate difference between groups. (B) Using parametric empirical Bayes, we tested for a group difference of NMDAR-antibody encephalitis while accounting for the presence or absence of encephalopathy in general using all patient records. We used a model search across group-effects to determine which parameters in our DCM showed significant group differences both in terms of NMDAR-antibody encephalitis status (present in 29 records) and encephalopathy status (present in 47 records). The top bar charts show the Bayesian covariate, which are reduced via model comparison to leave only those significant effects for each group class. The lower bar charts indicate group effect probabilities for each parameter with a significance limit set to >0.95 probability. The insets highlight the set of receptors that exhibit significant group effects. We show that only the NMDA parameters show significant effects for the NMDAR-antibody encephalitis while all three ion channels are significant predictors of encephalopathy. (B) To investigate individual patient classification, we tested NMDAR-antibody encephalitis patients using only those EEG records obtained within the ‘acute phase’ (<3 months since symptom onset, n = 19, see Supplementary Table 1). Fifteen of these 19 patients (red spheres) could be distinguished from the encephalopathic controls (grey spheres). We determined that the misclassifications (striped symbols) could be attributed to a high amplitude occipital alpha rhythm in occipital regions (Supplementary Fig. 5). By accounting for this confound, an accurate class label could be applied to all of our acute patients.

Comment in

Similar articles

See all similar articles

Cited by 6 articles

See all "Cited by" articles


    1. Baker AP, Brookes MJ, Rezek IA, Smith SM, Behrens T, Probert Smith PJ., et al. Fast transient networks in spontaneous human brain activity. Elife 2014; 3: e01867. - PMC - PubMed
    1. Bastos AM, Usrey WM, Adams RA, Mangun GR, Fries P, Friston KJ. Canonical microcircuits for predictive coding. Neuron 2012; 76: 695–711. - PMC - PubMed
    1. Benes FM, Berretta S. GABAergic interneurons: implications for understanding schizophrenia and bipolar disorder. Neuropsychopharmacology 2001; 25: 1–27. - PubMed
    1. Billingslea EN, Tatard-Leitman VM, Anguiano J, Jutzeler CR, Suh J, Saunders JA., et al. Parvalbumin cell ablation of NMDA-R1 causes increased resting network excitability with associated social and self-care deficits. Neuropsychopharmacology 2014; 39: 1603. - PMC - PubMed
    1. Brookes MJ, Woolrich M, Luckhoo H, Price D, Hale JR, Stephenson MC., et al. Investigating the electrophysiological basis of resting state networks using magnetoencephalography. Proc Natl Acad Sci USA 2011; 108: 16783–8. - PMC - PubMed

Publication types

MeSH terms