ARTICLE OPEN ACCESS Placental transfer of NMDAR antibodies causes reversible alterations in mice Anna Garcı́a-Serra, MSc, Marija Radosevic, PhD, Anika Pupak, MSc, Veronica Brito, PhD, José Rı́os, MSc, Correspondence Dr. Dalmau Esther Aguilar, BS, Estibaliz Maudes, MSc, Helena Ariño, MD, PhD, Marianna Spatola, MD, PhD, jdalmau@clinic.cat Francesco Mannara, PhD, Marta Pedreño, PhD, Bastien Joubert, MD, PhD, Silvia Ginés, PhD, Jesús Planagumà, PhD,* and Josep Dalmau, MD, PhD* Neurol Neuroimmunol Neuroinflamm 2021;8:e915. doi:10.1212/NXI.0000000000000915 Abstract Objective To determine whether maternofetal transfer of NMDA receptor (NMDAR) antibodies has pathogenic effects on the fetus and offspring, we developed a model of placental transfer of antibodies. Methods Pregnant C57BL/6J mice were administered via tail vein patients’ or controls’ immunoglobulin G (IgG) on days 14–16 of gestation, when the placenta is able to transport IgG and the immature fetal blood-brain barrier is less restrictive to IgG crossing. Immunohistochemical and DiOlistic (gene gun delivery of fluorescent dye) staining, confocal microscopy, standardized developmental and behavioral tasks, and hippocampal long-term potentiation were used to determine the antibody effects. Results In brains of fetuses, patients’ IgG, but not controls’ IgG, bound to NMDAR, causing a decrease in NMDAR clusters and cortical plate thickness. No increase in neonatal mortality was ob- served, but offspring exposed in utero to patients’ IgG had reduced levels of cell-surface and synaptic NMDAR, increased dendritic arborization, decreased density of mature (mushroom- shaped) spines, microglial activation, and thinning of brain cortical layers II–IV with cellular compaction. These animals also had a delay in innate reflexes and eye opening and during follow-up showed depressive-like behavior, deficits in nest building, poor motor coordination, and impaired social-spatial memory and hippocampal plasticity. Remarkably, all these para- digms progressively improved (becoming similar to those of controls) during follow-up until adulthood. Conclusions In this model, placental transfer of patients’ NMDAR antibodies caused severe but reversible synaptic and neurodevelopmental alterations. Reversible antibody effects may contribute to the infrequent and limited number of complications described in children of patients who develop anti-NMDAR encephalitis during pregnancy. *These authors are joint senior authors. From the Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS) (A.G.-S., M.R., A.P., V.B., E.A., E.M., H.A., M.S., F.M., M.P., B.J., S.G., J.P., J.D.), Hospital Clı́nic, Universitat de Barcelona; Departament de Biomedicina (A.P., V.B., S.G.), Facultat de Medicina, Institut de Neurociències, Universitat de Barcelona; Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas (CIBERNED) (A.P., V.B., S.G.), Madrid; Medical Statistics Core Facility (J.R.), IDIBAPS and Hospital Clı́nic, Barcelona; Biostatistics Unit (J.R.), School of Medicine, Universitat Autònoma de Barcelona; Centro de Investigaciones Biomédicas en Red de Enfermedades Raras (CIBERER) (J.D.), Madrid, Spain; Department of Neurology (J.D.), University of Pennsylvania, Philadelphia; and Institució Catalana de Recerca i Estudis Avançats (ICREA) (J.D.), Barcelona, Spain. Go to Neurology.org/NN for full disclosures. Funding information is provided at the end of the article. The Article Processing Charge was funded by the ISCIII. This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 4.0 (CC BY-NC-ND), which permits downloading and sharing the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal. Copyright © 2020 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Academy of Neurology. 1 Glossary BBB = blood-brain barrier; CBA = cell-based assay; fEPSP = field excitatory postsynaptic potential; GEE = generalized estimated equation; GLM = generalized linear model; IgG = immunoglobulin G; LTP = long-term potentiation; NMDAR = NMDA receptor; NMDAR-ab = NMDAR antibody; NOL = novel object location; PD = postnatal day; PPI = prepulse inhibition; PSD95 = postsynaptic density protein 95. Anti-NMDA receptor (NMDAR) encephalitis is a neuronal encephalitis were women (median age 20 years, range 16–26 antibody-mediated disease that associates with prominent years) with anti-NMDAR encephalitis. The presence of neurologic and psychiatric symptoms.1 Most clinical series NMDAR antibodies (NMDAR-abs) was determined with rat show that 60%–80% of patients are women of childbearing brain immunohistochemistry and a cell-based assay (CBA).1 age2–4 who sometimes develop the encephalitis during To rule out the presence of other antibodies, pooled IgG from pregnancy.5,6 In addition to the obstetric complications that patients was immunoabsorbed with HEK293T cells express- can arise from the severity of the disease,5–7 there is the ing GluN1/2B, and the reactivity determined with brain im- concern that patient’s antibodies can reach the fetal brain and munohistochemistry, CBA of GluN1/2B, and live neurons, cause synaptic and neurodevelopmental alterations. This showing with all 3 techniques abrogation of reactivity.10,18 We concern is supported by the demonstration of the pathoge- confirmed that immunoabsorbed patients’ IgG no longer nicity of the antibodies in cultured neurons8–10 and in a decreased the levels of NMDARs after 12-hour incubation, as mouse model of cerebroventricular transfer of patients’ anti- reported.19 bodies.11 Moreover, peritoneal injection of a human NMDAR monoclonal antibody to pregnant mice resulted in high Animals, infusion of IgG, tissue processing, and neonatal mortality and long-lasting irreversible neuro- determination of antibodies in blood from developmental deficits in the offspring.12 pregnant mice and fetuses Animal care and processing of brain of fetuses and offspring Therefore, considering the severity of patients’ symptoms, were performed as reported.11 Overall, 54 pregnant C57BL/ pathogenicity of the antibodies, and the fact that NMDAR 6J mice, 165 fetuses, and 187 pups were used for behavioral, signaling regulates neuronal maturation,13 migration,14 and electrophysiologic, morphologic, and synaptic brain studies. synaptogenesis,15 one would expect severe abnormalities in the offspring of patients with anti-NMDAR encephalitis. Pooled IgG (800 μg) from patients or controls was injected However, the clinical experience is markedly different; other via tail vein to pregnant mice on days 14, 15, and 16 (E14, than a few exceptional cases,16 most reports indicate that the E15, and E16) of gestation (figure 1A). These experiments offspring of these patients appear to have good outcomes.6 were planned according to the window of time in which the Indeed, among the limited number of problems reported in neonatal receptor (FcRn), which allows IgG transplacental the babies, most are described as transient17 or attributed to transfer, is expressed in placental tissue, and the immature medications.6 These discrepancies led us to investigate in a fetal blood-brain barrier (BBB) does not restrict the crossing model of maternofetal transfer of patients’ immunoglobulin G of IgG (by day E16, the BBB becomes more restrictive).20 (IgG) antibodies, the effects on synaptic NMDAR, dendritic The presence of human NMDAR-abs in blood from pregnant complexity, cortical development, and microglial activation in mice and fetuses was demonstrated with immunolabeling of fetal and postnatal brains. We also determined whether mice live rat hippocampal neurons in culture and conventional exposed in utero to patients’ IgG had impairment of memory, CBA.1 Antibody titers were obtained by serial sample dilution behavior, and hippocampal long-term plasticity and whether using CBA. the antibody effects persisted or resolved from newborn to adulthood stages. Determination of human IgG, NMDAR clusters, and cortical plate thickness in fetal brain tissue To determine whether human NMDAR-abs injected to pregnant mice reached the brain of fetuses, 5-μm-thick fetal Methods brain sections were immunostained for human IgG using Most of the methods and techniques used here have been Alexa Fluor 488 goat anti-human IgG, as above. Quantifica- previously reported11,18 and are described in appendix e-1 tion of fluorescence intensity was performed using the public (links.lww.com/NXI/A341). domain Fiji ImageJ software (fiji.sc/Fiji).8 Confirmation that patients’ IgG was bound to NMDAR was done by Human serum samples, IgG purification, immunoprecipitation.18 and immunoabsorption IgG was isolated by ammonium sulfate precipitation from The effects of patients’ antibodies on total cell-surface and serum of 7 patients with anti-NMDAR encephalitis and 7 synaptic NMDAR clusters and postsynaptic density protein healthy blood donors. All patients with anti-NMDAR 95 (PSD95) were examined using immunohistochemistry 2 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN Figure 1 Patients’ antibodies intravenously administered to pregnant mice reach fetal brain and bind to NMDAR (A) Diagram of the experimental design. IV injection of pooled IgG was performed on gestational days E14, E15, and E16. Brain and blood samples from the fetus or offspring were collected on E16, E17, and E18 and PD 21, PD 70, and PD 155; blood samples of pregnant mice were collected on E16–E18. Brain tissue studies were performed on the same time points and electrophysiology on PD 21 and PD 70. Newborns were assessed daily for neurodevelopmental milestones from birth until PD 21. After breastfeeding withdrawal, mice underwent a battery of behavioral tests at age 1, 2, and 4 months. Gestational period is marked in pink and postnatal period in green. (B) Representative images of HEK293T cells expressing GluN1/2B immunolabeled with human IgG contained in serum of fetuses (E17) exposed to controls’ or patients’ IgG. Scale bar = 10 μm. (C and D) Representative human IgG immunolabeling in E17 fetal brain sections of animals exposed in utero to controls’ (C) or patients’ (D) IgG; the 2 insets (small squares in C and D, developing hippocampus) are shown enlarged on the right of each panel. Scale bar whole brain = 100 μm, insets = 10 μm. (E) Quantification of human IgG immunofluorescence intensity in the developing hippocampus of fetal brains obtained on gestational days E16-E18 exposed to controls’ IgG or patients’ IgG. Controls’ IgG, E16, n = 5; E17, n = 13; E18, n = 8; patients’ IgG, E16, n = 9; E17, n = 16; E18, n = 8. Median intensity of IgG immunofluorescence in brain of mice exposed to controls’ IgG was defined as 100%. Data presented in box plots show the median, 25th, and 75th percentiles; whiskers indicate minimum and maximum. Significance of treatment effect was assessed by the Mann-Whitney U test. ****p < 0.0001. (F) Immunoprecipitated NMDAR-bound IgG from E17 and E18 brains of fetuses exposed in utero to controls’ or patients’ IgG. Lane brain tissue corresponds to homogenate of mouse brain used here as control. The band at ;100 KDa corresponds to the GluN1 subunit of the NMDAR. Each lane corresponds to the immunoprecipitation of the brains from 4 fetuses, per each indicated condition. IgG = immunoglobulin G; NMDAR = NMDA receptor; PD = postnatal day. and confocal microscopy.11 The thickness of the cortical plate stained fetal brain sections and quantified using Fiji ImageJ was measured using 4', 6-diamidino-2-phenylindole (DAPI)- software. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 3 Determination of human IgG, NMDAR clusters, mice underwent a comprehensive battery of memory and synaptic density, cortical layer thickness, and behavioral tests (appendix e-1, links.lww.com/NXI/A341, microglial activation in offspring and figure e-1, links.lww.com/NXI/A342). The presence of human IgG in the brain of offspring and the density of NMDAR were examined as indicated for Statistical analysis fetal brains (time points shown in figure 1A). NMDAR Comparison of human IgG intensity, confocal densities of clusters were quantified from a total of 15 images acquired NMDAR, PSD95, Homer1, Bassoon, Iba-1, CD68, and from CA1, CA3, and dentate gyrus of the hippocampus (5 mushroom-shaped spines between mice exposed to pa- different areas per region). tients’ IgG vs controls’ IgG was performed with the Mann- Whitney U test as non-normally distributed parameters. The density of glutamatergic synapses was assessed with im- Comparison of cortical plate thickness, cortical layers II– munohistochemistry using presynaptic (Bassoon) and post- IV and VI, cortical cell density, and electrophysiologic field synaptic (Homer1) markers and the corresponding secondary excitatory postsynaptic potential (fEPSP) slope changes antibodies (appendix e-1, links.lww.com/NXI/A341). between the 2 experimental groups was performed using independent sample t tests. Sholl analysis data were ana- Cortical layer thickness was measured in sections of brains lyzed with a generalized linear model (GLM) for global immunostained with specific cortical layer markers including a and time point effects. For behavioral paradigms, longi- monoclonal mouse anti-CUTL1 for layers II–IV, and a tudinal analyses were performed by generalized estimated polyclonal sheep anti-FoxP2 for layer VI, followed by the equations (GEEs) using an AR(1) matrix to account for appropriate secondary antibodies. The layers’ thickness was intraindividual correlations. All models include litter size, quantified using Fiji ImageJ software. Cell density in CUTL1- group, and group by time as fixed factors. Time to rooting, labeled layers was measured on brain sections using DAPI ear detachment, auditory startle, and eye opening were staining with spots algorithm in Imaris software. analyzed with the Mann-Whitney U test. All experiments were assessed for outliers with the ROUT method apply- Microglial activation was assessed in brain sections using a ing Q = 1%. In all analyses, we used a 2-sided type I error of monoclonal rat antibody against CD68 (pan-macrophage) 5%. All tests were performed using GraphPad Prism and a polyclonal rabbit antibody against Iba-1 (activated (version 8; GraphPad Inc., San Diego, CA) or SPSS microglia). The CD68- or Iba-1–stained surface density was (version 25; IBM Corp., Armonk, NY) for GLM and GEE quantified using Imaris software. models. Mean values or estimated mean values (body weight, righting reflex, and negative geotaxis for the GEE Dendritic complexity and spine model) were presented with ±SEM. morphology analysis These studies were performed using DiOlistic (gene gun Ethical approval delivery of fluorescent dye) staining and confocal microscopy Written informed consent was obtained from all patients; (appendix e-1, links.lww.com/NXI/A341). Dendritic com- the study was approved by the local institutional review plexity was examined with Sholl analysis in cortical and CA1 board at Hospital Clı́nic, Barcelona (HCB/2018/0192). hippocampal pyramidal neurons.21 Spine morphology was Animal studies were approved by the Ethical Committee of examined in DiI-labeled segments of apical secondary den- the University of Barcelona (2010/63/UE) and Spanish drites of cortical and CA1 pyramidal neurons. Quantification (RD 53/2013) regulations. of mushroom-shaped dendritic spines was performed with the spine classification algorithm (Imaris software). Data availability Data supporting these findings are available on reasonable Field potential recordings, hippocampal long- request. term potentiation, and paired- pulse facilitation Acute sections of the hippocampus of the offspring were used Results to determine long-term plasticity by the classical paradigm of stimulation at the Schaffer collateral pathway and recording Presence of human NMDAR-abs in serum of the field potentials at CA1 synapses.19 pregnant mice and fetuses Serum samples collected on E16, E17, and E18 from pregnant Neurobehavioral assessment in postnatal and mice and fetuses exposed to patients’ IgG, but not those ex- adult stages posed to controls’ IgG, showed human IgG binding to cell From birth to weaning (defined as breastfeeding with- surface of live hippocampal neuronal cultures and drawal on postnatal day [PD] 21), mice were assessed HEK293T cells expressing GluN1/GluN2B (figure 1B). The daily for achievement of developmental milestones using a human NMDAR antibody titer (obtained by serial dilution of modified Fox battery, similar to that previously repor- samples in CBA) was 1:160–1:320 in pregnant mice and 1: ted. 22 From weaning to adulthood (155 days), subsets of 80–1:160 in fetuses (data not shown). 4 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN Patients’ IgG reach fetal brain, causing a hippocampal areas showed a higher dendritic arborization in reduction of synaptic NMDAR and thinning of mice exposed in utero to patients’ IgG (figure 4, C and D, left). the cortical plate No differences were noted on PD 70 (figure 4, C and D, right). On days E16–18, fetuses exposed to patients’ IgG showed increased human IgG fluorescence in the developing hippo- The density of mushroom-shaped (mature) dendritic spines campus compared with those exposed to controls’ IgG (figure was assessed in the cerebral cortex and hippocampus using the 1, C–E). Immunoprecipitation experiments confirmed the Imaris standard parameters for dendritic spine classification presence of NMDAR bound to IgG in the brain precipitate (example shown in figure 4, E–G). On PD 21, the density of (figure 1F). Analysis of the density of NMDAR clusters these spines was significantly decreased in mice exposed in showed that fetuses harboring patients’ IgG had a significant utero to patients’ IgG (cortex, figure 4H, left; hippocampus, decrease in cluster density of total cell-surface and synaptic figure 4I, left). The changes were no longer seen on PD 70 NMDARs (figure 2, A and D). The cortical plate of these (cortex, figure 4H, right; and hippocampus, figure 4I, right). fetuses was also thinner on E17 and E18 compared with that of controls (figure 2, G and J). Intrauterine exposure to patients’ IgG causes a reversible impairment in hippocampal long- Offspring exposed in utero to patients’ IgG term potentiation formation have decreased NMDAR and synaptic density, Assessment of fEPSPs showed that on PD 21, mice that had thinning of cortical layers, and been exposed in utero to patients’ IgG had impairment in microglial activation long-term potentiation (LTP) formation compared with mice On PD 21 and 70, mice exposed in utero to patients’ IgG no exposed in utero to controls’ IgG (figure e-2, A–C, links.lww. longer had significant brain deposits of IgG (not shown). com/NXI/A343). This change in LTP formation was no Despite this, the reduction of total cell-surface NMDAR longer detected on PD 70 (figure e-2, D–F). Paired-pulse clusters persisted until day 70 (figure 2, B and C) with sub- facilitation studies at 50 and 100 milliseconds of interstimuli sequent normalization, so that by day 155, the levels were interval (which assess presynaptic release mechanisms) were similar to those of controls (not shown). In contrast, the levels not different between groups of animals (data not shown). of synaptic NMDAR normalized faster; they were decreased on day 21 (figure 2E), but were normal on days 70 and 155 Neurodevelopmental delay due to in utero (figure 2F, not shown). PSD95 cluster density was not af- exposure to NMDAR-abs fected at any time point (data not shown). Newborn pups exposed in utero to patients’ IgG showed multiple alterations including transient decrease of weight Mice exposed to patients’ IgG, but not controls’ IgG, showed (figure 5A); longer times for proper body righting (figure 5B) thinning of cortical layers II–IV (assessed with CUTL1/FoxP2 and negative geotaxis (figure 5C); delayed eye opening staining) on PD 21 (figure 2, H and K) that was no longer (figure 5D); and longer time in open field locomotion (not present on PD 70 (figure 2, I and L). This transient thinning of shown, p < 0.01). Neonatal mortality occurred in litters of cortical layers II–IV was associated with an increase in cell mice exposed in utero to controls’ IgG (26%) or patients’ IgG density in these layers on PD 21 (p = 0.0065, data not shown). (25%), which is in line with general litter loss in laboratory mice.23 No significant differences were observed in ear de- On PD 21, the density of glutamatergic synapses (assessed by tachment, tactile rooting response, and auditory startle (data colocalization of Homer1/Bassoon clusters) was decreased in not shown). mice exposed in utero to patients’ IgG (figure 3, A and B). These effects were no longer seen on day 70 (not shown). Spontaneous reversal of behavioral alterations from early to mature adulthood On PD 21 and 70, there was an increase of 57% and 53% in At age 1 month, mice exposed in utero to patients’ IgG activated microglia (assessed by expression of Iba-1 in so- showed a significant decrease in the novel object location matosensory cortex) in mice exposed in utero to patients’ IgG (NOL) index indicating impairment in visuospatial memory (figure 3, C and D). This difference in activated microglia was (figure 5E), a higher amount of unused material for nest no longer present on PD 155 (not shown). No significant building (figure 5F), and a longer time of exploration of a difference in the total number of microglial cells (assessed by familiar intruder mouse indicating impairment in social expression of CD68) was noted between groups of mice at memory (figure 5G, trial 4). Assessment of motor co- any time point examined (figure 3E). ordination and balance showed that mice exposed in utero to patients’ IgG had shorter times on the accelerating rotarod Exposure to patients’ IgG causes alterations in (figure 5H). No significant differences were noted in prepulse dendritic complexity, spine density, and spine inhibition (PPI) of acoustic startle reaction, locomotor ac- morphology in cortical and hippocampal tivity, or adult body weight at age 1 month (data not shown). pyramidal neurons Dendritic arborization was assessed by Sholl analysis (figure 4, At age 2 months, the alterations of NOL, nest building, and A and B). On PD 21, the Sholl analysis in cortical and social interaction in mice that had been exposed to patients’ Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 5 Figure 2 Intrauterine exposure to patients’ IgG causes a transient reduction of cell-surface NMDARs and thinning of the cortical plate and cortical layers II–IV (A–F) Quantification of total cell-surface (A–C) and synaptic (D–F) NMDAR cluster density in brains obtained on gestational days E16, E17, and E18 (A and D), PD 21 (B and E), and PD 70 (C and F) from animals exposed in utero to controls’ IgG or patients’ IgG. Median of NMDAR clusters in mice exposed to controls’ IgG was defined as 100%. Controls’ IgG, E16, n = 6; E17, n = 15; E18, n = 7; PD 21, n = 11; PD 70, n = 9; patients’ IgG, E16, n = 13; E17, n = 21; E18, n = 5; PD 21, n = 14; PD 70, n = 10. (G) Representative cortical plate thickness (assessed with DAPI) in E17 and E18 brains of fetuses exposed in utero to controls’ or patients’ IgG. Scale bars = 100 μm. (H–I) Representative sections of the somatosensory cortex (H, PD 21; I, PD 70) of animals exposed in utero to controls’ or patients’ IgG, showing cortical layers II–IV (CUTL1 positive, green fluorescence) and cortical layer VI (FoxP2 positive, red fluorescence); arrows indicate the thickness of the indicated layers. Scale bars = 100 μm. (J) Quantification of cortical plate thickness in E16, E17, and E18 brains of fetuses exposed in utero to controls’ IgG or patients’ IgG. Controls’ IgG, E16, n = 4; E17, n = 11; E18, n = 10; patients’ IgG, E16, n = 8; E17, n = 14; E18, n = 8. (K and L) Quantification of the indicated cortical layers’ thickness (K, PD 21; L, PD 70) in animals exposed to controls’ IgG or patient’s IgG. Controls’ IgG, PD 21, n = 7; PD 70, n = 8; patients’ IgG, PD 21, n = 15; PD 70, n = 10. For J–L, the mean thickness of the cortical plate or the indicated cortical layers in mice exposed in utero to controls’ IgG was defined as 100%. Data presented in box plots show the median, 25th, and 75th percentiles; whiskers indicate minimum and maximum. Significance of treatment effect was assessed by the Mann- Whitney U test in A–F and independent sample t test in J–L. *p < 0.05, **p < 0.01, ****p < 0.0001. IgG = immunoglobulin G; NMDAR = NMDA receptor; PD = postnatal day. IgG were no longer present (figure 5, I and J; social in- time of immobility in the tail suspension test (which assesses teraction not shown). PPI and locomotor activity remained core helplessness) (figure 5K). unaffected at age 2 months (not shown). However, mice ex- posed in utero to patients’ IgG showed a shorter time on the At age 4 months (adulthood), no significant difference was accelerating rotarod (data not shown, p = 0.0292) and longer noted between mice exposed in utero to patients’ IgG and 6 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN Figure 3 Intrauterine exposure to patients’ IgG causes a reversible decrease in synaptic density and increase in activated microglia (A) Representative z-stacks of brain sections obtained on PD 21 from mice exposed in utero to controls’ or patients’ IgG immunostained for Homer1 and Bassoon expression. Colocalization of both markers is defined as a glutamatergic synapse. Scale bar = 2 μm. (B) Quantification of synaptic density (Homer1/ Bassoon colocalization) in brains obtained on PD 21 from animals exposed in utero to controls’ IgG or patients’ IgG. Median density of Homer1/Bassoon synapses in mice exposed in utero to controls’ IgG was defined as 100%. Controls’ IgG, n = 11; patients’ IgG, n = 14. (C) Representative CD68 (macrophage specific) and Iba-1 (activated microglia specific) immunostainings in brains obtained on PD 21 from mice exposed in utero to controls’ or patients’ IgG. Scale bar = 20 μm. (D and E) Quantification of Iba-1 (D) and CD68 (E) immunostaining in brains obtained at PD 21 and PD 70 from mice exposed to controls’ IgG or patients’ IgG. Median surface density of immunostaining in mice exposed in utero to controls’ IgG was defined as 100%. Controls’ IgG, PD 21, n = 11; PD 70, n = 9; patients’ IgG, PD 21, n = 14; PD 70, n = 10. Data presented in box plots show the median, 25th, and 75th percentiles; whiskers indicate the minimum and maximum. Significance of treatment effect was assessed by the Mann-Whitney U test. **p < 0.01, ****p < 0.0001. IgG = immunoglobulin G; PD = postnatal day. those exposed to controls’ IgG in any of the following tests: expected from previous experience with adult mice in which tail suspension (figure 5L), rotarod, locomotor activity, NOL, patients’ IgG was infused in the cerebroventricular system,11,19 PPI, nest building, and social interaction (not shown). we observed a robust antibody-mediated decrease in NMDAR clusters in the brain of the fetuses. These effects occurred quickly (noted 2 days after injecting the antibodies to the mothers), and Discussion the observed reduction of NMDAR clusters and impairment of We show that antibodies from patients with anti-NMDAR en- memory and hippocampal long-term plasticity were detectable cephalitis infused to pregnant mice reached the fetal brain, during the first postnatal month. caused multiple alterations ranging from NMDAR synaptic changes to delayed neurodevelopment and behavioral deficits in In addition, hippocampal and cortical neurons from mice the offspring, and that all these alterations progressively resolved exposed in utero to patients’ IgG showed an increase in after birth (summary in table e-1, links.lww.com/NXI/A341). As dendritic arborization accompanied by a reduction of mature Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 7 Figure 4 Intrauterine exposure to patients’ IgG causes a transient increase in dendritic complexity and decrease in spine density (A and B) Representative confocal image (A) and Sholl analysis (B) of a DiI-stained CA1 hippocampal pyramidal neuron showing in B the 10-μm spaced concentric rings used for quantification of dendritic complexity. Scale bar in A = 50 μm. (C and D) Number of dendrite intersections according to the indicated distance to the neuronal soma in cortical and hippocampal neurons examined at PD 21 and PD 70 of mice exposed in utero to controls’ (Black) or patients’ (blue) IgG. Number of mice exposed in utero to controls’ IgG, PD 21, n = 6; PD 70, n = 4; number of mice exposed in utero to patients’ IgG, PD 21, n = 6; PD 70, n = 4. Per each animal, condition, and brain region, 4 neurons were examined. Data presented as mean ± SEM. Significance of treatment effect was assessed by a generalized lineal model. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. (E–G) Representative images by confocal microscopy of a DiI-stained CA1 hippocampal pyramidal neuron (E) and a secondary apical dendrite (F), with morphologically classified dendritic spines (G, mushroom-shaped spines in green); Scale bar in E = 50 μm; scale bar in F = 5 μm. (H and I) Quantification of mushroom dendritic spine density in cortical and hippocampal neurons examined at PD 21 and PD 70 in mice exposed in utero to controls’ or patients’ IgG. The median density of mushroom spines (number of spines/μm) in mice exposed to controls’ IgG was defined as 100%. Total number of dendrites examined/total number of mice exposed to controls’ IgG: PD 21, cortex n = 122/6, hippocampus n = 120/6; PD 70, cortex n = 81/4, and hippocampus n = 81/4. Total dendrites/mice exposed to patients’ IgG: PD 21, cortex n = 141/6, hippocampus n = 148/6; PD 70, cortex n = 70/4, and hippocampus n = 72/4. Data presented in box plots show the median, 25th, and 75th percentiles; whiskers indicate minimum and maximum. Significance of treatment effect was assessed by the Mann-Whitney U test in H and I. *p < 0.05. IgG = immunoglobulin G; PD = postnatal day. 8 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN Figure 5 Intrauterine exposure to patients’ IgG causes transient developmental delay and behavioral alterations in offspring (A–D) Developmental milestone assessment (from birth to PD 21) in mice that were exposed in utero to controls’ (Black) or patients’ (blue) IgG including (A) body weight, (B) time needed for body righting using the surface righting reflex test (PD 1, 3, 5, 7, 9, and 11), (C) time needed for 180° turning to head up position using the negative geotaxis reflex (PD 2, 4, and 6), and (D) scored eye opening (PD 12–16). Number of mice exposed in utero to controls’ IgG = 36; number of mice exposed to patients’ IgG = 40. Data from A–C are represented as estimated mean ± SEM inferred by the GEE models described in Methods and in D are represented as median. Significance of treatment effect was assessed by a GEE adjusted model (A–C) and Mann-Whitney U test (D). *p < 0.05, ***p < 0.001. (E–L) Battery of behavioral tests performed at the indicated time points including novel object location index (E, 1 month; I, 2 months), nest building test (F, 1 month; J, 2 months), social interaction test (G, 1 month), rotarod test (H, 1 month), and tail suspension test (K, 2 months; L, 4 months). For E–I and K, number of mice exposed to controls’ IgG = 18 and mice exposed to patients’ IgG = 18. For J and L, number of mice exposed to controls’ IgG = 8 and mice exposed to patients’ IgG = 9. Data from panels E, F, and I–L are represented as box plots with median, 25th, and 75th percentiles; whiskers indicate minimum and maximum values; data from G and H are represented as mean ± SEM. For all behavioral tests, significance of treatment effect was assessed by GEE. *p < 0.05, **p < 0.01, ***p < 0.001. GEE = generalized estimated equation; IgG = immunoglobulin G; PD = postnatal day. (mushroom-shaped) dendritic spines and synaptic density. resemble those of the cortex-specific GluN1 knockout, al- These alterations have not been reported in previous models though in our immune model, the alterations were reversible. of immunodepletion of NMDAR and likely contributed to the impairment of hippocampal LTP. The effect of NMDAR We believe that the increased dendritic arborization observed hypofunction on dendritic arborization and the morphology in animals exposed in utero to patients’ IgG was likely caused of the spines are not well understood. In a cortex-specific by the reduction of NMDAR levels along with microglial GluN1 knockout model, the cortical neurons in layer IV activation. Any perinatal immune challenge such as the elaborated exuberant dendritic specializations24; however, in a binding of antibodies to NMDAR in the fetal brain can result study using MK801 (an NMDAR antagonist) during early in microglial activation26 and consequently affect microglia- postnatal development, there was a transient decrease in mediated synaptic pruning. This normal microglial function is dendritic arborization in hippocampal pyramidal neurons that also triggered by neural activity in the developing brain.27,28 reversed when MK801 was discontinued.25 Our findings For example, impaired microglia-neuron interactions in Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 9 CX3CR1-deficient mice showed delayed synaptic functional model, in which serum NMDAR-abs representative of several maturation29 and immature and redundant connectivity, the patients were used, and the model in which a human mono- latter attributed to reduced synaptic pruning.30 These alter- clonal NMDAR-ab was used are unclear. It is likely that the ations spontaneously reverted by PD 30–40,29,30 similar to monoclonal antibody model, lacking the repertoire NMDAR- our findings. abs of patients, is less representative of the disease. Never- theless, the outcome of newborns from pregnant patients with In our model, the brain of fetuses and offspring exposed in anti-NMDAR encephalitis6 is remarkably different from that utero to patients’ IgG had a reduction in the thickness of the obtained with the mouse monoclonal antibody model.12 cortical plate and cortical layers II–IV. In mice, glutamate serves as a chemoattractant for neurons in the developing Our study does not allow generalization of the notion that cortex, signaling cells to migrate into the cortical plate through anti-NMDAR encephalitis during pregnancy does not repre- receptor activation.14 Studies with NMDAR antagonists sent a risk for serious effects in the offspring. For example, our (MK801 or APV) have shown that they block neuronal cell model does not associate with encephalitis of the mothers, migration from the ventricular zone into the cortical plate, which in our experience6 is the most important risk factor for delaying neuronal migration to the upper cortical layers obstetric complications and potential brain damage of the (II–III).14,31 In our immune-mediated model, the transient newborns. However, the low frequency of antibody-mediated decrease in thickness of cortical layers II–IV was associated complications in newborns of patients with anti-NMDAR with an increase in cell density in the upper cortical layers that encephalitis is puzzling and may also be related to the timings likely contributed to the restoration of the cortical layer of IgG placental transfer and BBB restriction properties for thickness after PD 21. These findings are somewhat similar to IgG crossing into the brain. In humans, placental transfer of those obtained in models of genetic disruption of other pro- maternal antibodies starts at 12–13 weeks of pregnancy36–38 teins and receptors where the decrease in cortical thickness is that is approximately the same time the BBB of the fetus starts reversible and associates with an increase in cell density in the to become more restrictive to the passage of albumin (and upper cortical layers.32,33 For example, loss of MeCP2 in the likely Igs) to the brain.39 It is unknown whether this time Rett syndrome model results in a reduction in thickness of window and the efficiency of the fetal BBB are affected by the somatosensory cortical layers II–III and VI, with a total rescue presence of maternal encephalitis or systemic inflammation. of thickness after treatment with mirtazapine during PD Future studies with pregnant patients who develop anti- 28–42.33 In another model based on a constitutive reduction NMDAR encephalitis and close follow-up of the babies are in central serotonin (VMAT2sert-cre mice), the total cortical needed to determine the frequency of deficits (if present, thickness reduction at PD 7 spontaneously recovered by PD mostly overlooked until now) and the timing of improve- 14.32 Like this model, our anti-NMDAR model associates ment. Keeping these limitations in mind, the current model with a temporary delay in the development of the upper suggests that reversibility of some of the antibody-mediated cortical layers along with increased compaction of these alterations can potentially occur. Moreover, the model can be layers. manipulated to assess the effects of systemic inflammation and provides also the possibility to determine whether in- Preceding the indicated memory deficit, the newborn pups hibitors of FcRn have potential utility in preventing the po- exposed in utero to patients’ IgG showed impaired innate tential neonatal abnormalities reported here. reflexes and developmental delay. Similar impairment of na- tive reflexes and developmental delay have been reported in Acknowledgment GluN1 gene-manipulated mice,34 and newborn rats treated The authors thank Dr. Myrna R. Rosenfeld (IDIBAPS, with a competitive NMDAR antagonist.35 University of Barcelona) for critical review of the manuscript and useful suggestions and Mercedes Alba, Eva Caballero, and A previous model of transplacental transfer of a human- Araceli Mellado (IDIBAPS, Hospital Clı́nic, University of derived NMDAR monoclonal antibody showed high neonatal Barcelona) for their technical support. mortality (27%), impairment of native reflexes, and long- lasting neuropathologic alterations without evidence of re- Study funding versibility.12 In contrast, we did not observe increased neo- This study was funded by Plan Nacional de I+D+I and cofi- natal mortality, and the symptoms and developmental nanced by the Instituto de Salud Carlos III (ISCIII)— alterations spontaneously improved. Moreover, although the Subdirección General de Evaluación y Fomento de la Inves- monoclonal antibody model did not assess dendritic arbori- tigación Sanitaria and the Fondo Europeo de Desarrollo Re- zation, spine density, cortical development, and number of gional (ISCIII-FEDER; 17/00234); Project Integrative of cell surface and synaptic NMDAR clusters, we found that Excellence (PIE 16/00014); Centro de Investigación Bio- patients’ IgG antibodies robustly altered all these paradigms, médica en Red de Enfermedades Raras (CIBERER; #CB15/ along with a dramatic decrease in NMDAR clusters and 00010); “La Caixa” Foundation (ID 100010434, under the hippocampal LTP. Thus, the reversibility of these alterations agreement LCF/PR/HR17/52150001); The Safra Founda- and associated symptoms cannot be attributed to a milder tion (J.D.), and Fundació CELLEX (J.D.); Centres de phenotype. The reasons for the discrepancies between our Recerca de Catalunya (CERCA) program by La Generalitat 10 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN de Catalunya (J.R.); Pla Estratègic de Recerca i Innovació en Salut (PERIS, SLT002/16/00346, J.P.); Ministerio de Appendix (continued) Economia y Competitividad (RTI2018-094374-B-I00, S.G.); Name Location Contribution Basque Government Doctoral Fellowship Program (PRE_ 2019_1_0255; E.M.); and Agència de Gestió d’Ajuts Uni- Helena IDIBAPS, Hospital Clı́nic, Performed IV Ariño, MD, Universitat de Barcelona, administrations and versitaris i de Recerca (FI-AGAUR) grant program by La PhD Spain revised the manuscript for Generalitat de Catalunya (2018FI_B_00487 A.P.; 2019FI_ intellectual content B1 00212, A.G.-S.). Marianna IDIBAPS, Hospital Clı́nic, Performed Spatola, MD, Universitat de Barcelona, immunoprecipitation PhD Spain technique and revised the Disclosure manuscript for intellectual A. Garcı́a-Serra, M. Radosevic, A. Pupak, V. Brito, J. Rı́os, E. content Aguilar, E. Maudes, H. Ariño, M. Spatola, F. Mannara, M. Francesco IDIBAPS, Hospital Clı́nic, Revised the manuscript for Pedreño, B. Joubert, S. Ginés, and J. Planagumà report no Mannara, Universitat de Barcelona, intellectual content PhD Spain disclosures. J. Dalmau receives royalties from Athena Diag- nostics for the use of Ma2 as an autoantibody test and from Marta IDIBAPS, Hospital Clı́nic, Revised the manuscript for Euroimmun for the use of NMDA, GABAB receptor, GABAA Pedreño, Universitat de Barcelona, intellectual content PhD Spain receptor, DPPX and IgLON5 as autoantibody tests. Go to Neurology.org/NN for full disclosures. Bastien IDIBAPS, Hospital Clı́nic, Revised the manuscript for Joubert, MD, Universitat de Barcelona, intellectual content PhD Spain Publication history Silvia Ginés, IDIBAPS, Hospital Clı́nic, Revised the manuscript for Received by Neurology: Neuroimmunology & Neuroinflammation PhD and Institut de intellectual content July 13, 2020. Accepted in final form August 26, 2020. Neurociències, Universitat de Barcelona; and CIBERNED, Madrid, Spain Jesús IDIBAPS, Hospital Clı́nic, Major role in the Planagumà, Universitat de Barcelona, acquisition of data; Appendix Authors PhD Spain designed and conceptualized the study; Name Location Contribution and revised the manuscript for intellectual content Anna IDIBAPS, Hospital Clı́nic, Major role in the Garcı́a- Universitat de Barcelona, acquisition of data; Josep IDIBAPS, Hospital Clı́nic, Designed and Serra, MSc Spain analyzed the data; and Dalmau, Universitat de Barcelona, conceptualized the study drafted the manuscript for MD, PhD ICREA; CIBERER, Madrid, and drafted the intellectual content Spain; and University of manuscript for intellectual Pennsylvania, Philadelphia content Marija IDIBAPS, Hospital Clı́nic, Major role in the Radosevic, Universitat de Barcelona, acquisition of data; PhD Spain analyzed the data; and revised the manuscript for References intellectual content 1. Dalmau J, Gleichman AJ, Hughes EG, et al. Anti-NMDA-receptor encephalitis: case series and analysis of the effects of antibodies. Lancet Neurol 2008;7:1091–1098. Anika IDIBAPS, Hospital Clı́nic, Major role in the 2. Titulaer MJ, McCracken L, Gabilondo I, et al. Treatment and prognostic factors for Pupak, MSc and Institut de acquisition of data and long-term outcome in patients with anti-NMDA receptor encephalitis: an observa- Neurociències, Universitat revised the manuscript for tional cohort study. Lancet Neurol 2013;12:157–165. de Barcelona; and intellectual content 3. Irani SR, Bera K, Waters P, et al. N-methyl-D-aspartate antibody encephalitis: tem- CIBERNED, Madrid, Spain poral progression of clinical and paraclinical observations in a predominantly non- paraneoplastic disorder of both sexes. Brain 2010;133:1655–1667. Veronica IDIBAPS, Hospital Clı́nic, Major role in the 4. Viaccoz A, Desestret V, Ducray F, et al. Clinical specificities of adult male patients with Brito, PhD and Institut de acquisition of data and NMDA receptor antibodies encephalitis. Neurology 2014;82:556–563. Neurociències, Universitat revised the manuscript for 5. Kumar MA, Jain A, Dechant VE, et al. Anti-N-methyl-D-aspartate receptor enceph- de Barcelona; and intellectual content alitis during pregnancy. Arch Neurol 2010;67:884–887. CIBERNED, Madrid, Spain 6. Joubert B, Garcia-Serra A, Planaguma J, et al. Pregnancy outcomes in anti-NMDA receptor encephalitis: case series. Neurol Neuroimmunol Neuroinflamm 2020;7: José Rı́os, Medical Statistics Core Analyzed the data and e668. MSc Facility, IDIBAPS and revised the manuscript for 7. Xiao X, Gui S, Bai P, et al. Anti-NMDA-receptor encephalitis during pregnancy: a case Hospital Clı́nic, Barcelona; intellectual content report and literature review. J Obstet Gynaecol Res 2017;43:768–774. and Biostatistics Unit, 8. Hughes EG, Peng X, Gleichman AJ, et al. Cellular and synaptic mechanisms of anti- Faculty of Medicine, NMDA receptor encephalitis. J Neurosci 2010;30:5866–5875. Universitat Autònoma de 9. Mikasova L, De Rossi P, Bouchet D, et al. Disrupted surface cross-talk between Barcelona, Spain NMDA and Ephrin-B2 receptors in anti-NMDA encephalitis. Brain 2012;135: 1606–1621. Esther IDIBAPS, Hospital Clı́nic, Major role in the 10. Moscato EH, Peng X, Jain A, Parsons TD, Dalmau J, Balice-Gordon RJ. Acute Aguilar, BS Universitat de Barcelona, acquisition of data and mechanisms underlying antibody effects in anti-N-methyl-D-aspartate receptor en- Spain revised the manuscript for cephalitis. Ann Neurol 2014;76:108–119. intellectual content 11. Planaguma J, Leypoldt F, Mannara F, et al. Human N-methyl D-aspartate receptor antibodies alter memory and behaviour in mice. Brain 2015;138:94–109. Estibaliz IDIBAPS, Hospital Clı́nic, Major role in the 12. Jurek B, Chayka M, Kreye J, et al. Human gestational N-methyl-d-aspartate receptor Maudes, Universitat de Barcelona, acquisition of data and autoantibodies impair neonatal murine brain function. Ann Neurol 2019;86:656–670. MSc Spain revised the manuscript for 13. Hirasawa T, Wada H, Kohsaka S, Uchino S. Inhibition of NMDA receptors induces intellectual content delayed neuronal maturation and sustained proliferation of progenitor cells during neocortical development. J Neurosci Res 2003;74:676–687. Neurology.org/NN Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 11 14. Behar TN, Scott CA, Greene CL, et al. Glutamate acting at NMDA receptors stim- 27. Schafer DP, Lehrman EK, Kautzman AG, et al. Microglia sculpt postnatal neural ulates embryonic cortical neuronal migration. J Neurosci 1999;19:4449–4461. circuits in an activity and complement-dependent manner. Neuron 2012;74:691–705. 15. Tolias KF, Bikoff JB, Burette A, et al. The Rac1-GEF Tiam1 couples the NMDA 28. Zuo Y, Yang G, Kwon E, Gan WB. Long-term sensory deprivation prevents dendritic receptor to the activity-dependent development of dendritic arbors and spines. spine loss in primary somatosensory cortex. Nature 2005;436:261–265. Neuron 2005;45:525–538. 29. Hoshiko M, Arnoux I, Avignone E, Yamamoto N, Audinat E. Deficiency of the 16. Jagota P, Vincent A, Bhidayasiri R. Transplacental transfer of NMDA receptor anti- microglial receptor CX3CR1 impairs postnatal functional development of thalamo- bodies in an infant with cortical dysplasia. Neurology 2014;82:1662–1663. cortical synapses in the barrel cortex. J Neurosci 2012;32:15106–15111. 17. Hilderink M, Titulaer MJ, Schreurs MW, Keizer K, Bunt JE. Transient anti-NMDAR 30. Paolicelli RC, Bolasco G, Pagani F, et al. Synaptic pruning by microglia is necessary for encephalitis in a newborn infant due to transplacental transmission. Neurol Neuro- normal brain development. Science 2011;333:1456–1458. immunol Neuroinflamm 2015;2:e126. 31. Reiprich P, Kilb W, Luhmann HJ. Neonatal NMDA receptor blockade disturbs 18. Planaguma J, Haselmann H, Mannara F, et al. Ephrin-B2 prevents N-methyl-D- neuronal migration in rat somatosensory cortex in vivo. Cereb Cortex 2005;15: aspartate receptor antibody effects on memory and neuroplasticity. Ann Neurol 2016; 349–358. 80:388–400. 32. Narboux-Neme N, Angenard G, Mosienko V, et al. Postnatal growth defects in mice 19. Mannara F, Radosevic M, Planaguma J, et al. Allosteric modulation of NMDA re- with constitutive depletion of central serotonin. ACS Chem Neurosci 2013;4: ceptors prevents the antibody effects of patients with anti-NMDAR encephalitis. Brain 171–181. 2020;143:2709–2720. 33. Bittolo T, Raminelli CA, Deiana C, et al. Pharmacological treatment with mirtazapine 20. Braniste V, Al-Asmakh M, Kowal C, et al. The gut microbiota influences blood-brain rescues cortical atrophy and respiratory deficits in MeCP2 null mice. Sci Rep 2016;6: barrier permeability in mice. Sci Transl Med 2014;6:263ra158. 19796. 21. Ferreira TA, Blackman AV, Oyrer J, et al. Neuronal morphometry directly from 34. Single FN, Rozov A, Burnashev N, et al. Dysfunctions in mice by NMDA receptor bitmap images. Nat Methods 2014;11:982–984. point mutations NR1(N598Q) and NR1(N598R). J Neurosci 2000;20:2558–2566. 22. Hill JM, Lim MA, Stone MM. Developmental milestones in the newborn mouse. In: 35. Mikulecka A, Mares P. NMDA receptor antagonists impair motor performance in Gozes I, ed. Neuromethods. Totowa: Humana Press; 2008:131–149. immature rats. Psychopharmacology (Berl) 2002;162:364–372. 23. Weber EM, Algers B, Wurbel H, Hultgren J, Olsson IA. Influence of strain and parity 36. Malek A, Sager R, Kuhn P, Nicolaides KH, Schneider H. Evolution of maternofetal on the risk of litter loss in laboratory mice. Reprod Domest Anim 2013;48:292–296. transport of immunoglobulins during human pregnancy. Am J Reprod Immunol 24. Datwani A, Iwasato T, Itohara S, Erzurumlu RS. NMDA receptor-dependent pattern 1996;36:248–255. transfer from afferents to postsynaptic cells and dendritic differentiation in the barrel 37. Simister NE, Story CM, Chen HL, Hunt JS. An IgG-transporting Fc receptor cortex. Mol Cell Neurosci 2002;21:477–492. expressed in the syncytiotrophoblast of human placenta. Eur J Immunol 1996;26: 25. Elhardt M, Martinez L, Tejada-Simon MV. Neurochemical, behavioral and archi- 1527–1531. tectural changes after chronic inactivation of NMDA receptors in mice. Neurosci Lett 38. Dancis J, Lind J, Oratz M, Smolens J, Vara P. Placental transfer of proteins in human 2010;468:166–171. gestation. Am J Obstet Gynecol 1961;82:167–171. 26. Paolicelli RC, Ferretti MT. Function and dysfunction of microglia during brain de- 39. Virgintino D, Robertson D, Benagiano V, et al. Immunogold cytochemistry of the velopment: consequences for synapses and neural circuits. Front Synaptic Neurosci blood-brain barrier glucose transporter GLUT1 and endogenous albumin in the 2017;9:9. developing human brain. Brain Res Dev Brain Res 2000;123:95–101. 12 Neurology: Neuroimmunology & Neuroinflammation | Volume 8, Number 1 | January 2021 Neurology.org/NN