Autoantibodies against NMDA receptor 1 modify rather than cause encephalitis

The etiology and pathogenesis of “anti-N-methyl-D-aspartate-receptor (NMDAR) encephalitis” and the role of autoantibodies (AB) in this condition are still obscure. While NMDAR1-AB exert NMDAR-antagonistic properties by receptor internalization, no firm evidence exists to date that NMDAR1-AB by themselves induce brain inflammation/encephalitis. NMDAR1-AB of all immunoglobulin classes are highly frequent across mammals with multiple possible inducers and boosters. We hypothesized that “NMDAR encephalitis” results from any primary brain inflammation coinciding with the presence of NMDAR1-AB, which may shape the encephalitis phenotype. Thus, we tested whether following immunization with a “cocktail” of 4 NMDAR1 peptides, induction of a spatially and temporally defined sterile encephalitis by diphtheria toxin-mediated ablation of pyramidal neurons (“DTA” mice) would modify/aggravate the ensuing phenotype. In addition, we tried to replicate a recent report claiming that immunizing just against the NMDAR1-N368/G369 region induced brain inflammation. Mice after DTA induction revealed a syndrome comprising hyperactivity, hippocampal learning/memory deficits, prefrontal cortical network dysfunction, lasting blood brain-barrier impairment, brain inflammation, mainly in hippocampal and cortical regions with pyramidal neuronal death, microgliosis, astrogliosis, modest immune cell infiltration, regional atrophy, and relative increases in parvalbumin-positive interneurons. The presence of NMDAR1-AB enhanced the hyperactivity (psychosis-like) phenotype, whereas all other readouts were identical to control-immunized DTA mice. Non-DTA mice with or without NMDAR1-AB were free of any encephalitic signs. Replication of the reported NMDAR1-N368/G369-immunizing protocol in two large independent cohorts of wild-type mice completely failed. To conclude, while NMDAR1-AB can contribute to the behavioral phenotype of an underlying encephalitis, induction of an encephalitis by NMDAR1-AB themselves remains to be proven.


INTRODUCTION
Since, the first report on a novel paraneoplastic autoimmune disease with autoantibodies against the NMDA receptor subunit NR1 (NMDAR1-AB = GluN1-AB) appeared >12 years ago [1,2], this condition attracted considerable attention both by clinicians and basic researchers worldwide, resulting in currently nearly 2000 entries in PubMed. Variably associated with psychosis, cognitive decline, extrapyramidal symptoms, or seizures, this "NMDAR encephalitis" was already early on described to be mediated by lower surface expression of neuronal NMDAR after exposure to NMDAR1-AB of the immunoglobulin G (IgG) class [3]. The resulting NMDAR antagonistic or "ketamine-like" effects, demonstrated also in various rodent models via passive transfer of NMDAR1-AB [4][5][6][7][8][9][10][11][12][13], were interpreted as potential mechanisms of action. Thus far, however, they have failed to explain the inflammatory or encephalitic component of this condition. Interestingly, comparison of NMDAR1-AB-positive and -negative human encephalitis cases did not reveal differences, except for few perhaps NMDAR-antagonistic (ketamine-like) symptoms [14].
Even sophisticated preclinical approaches raised further questions rather than providing answers. For example, immunization of mice with liposome-embedded tetrameric NMDAR was stated to induce NMDAR-AB and fulminant encephalitis. Highly variable behavioral changes in this model included hyperactivity, immobility, circling, seizures, hunched back/lethargy, and increased mortality, including neuroinflammation, and immune cell infiltration [15], all nonspecific features of any severe encephalitis. Adequate "negative" controls like the use of other liposomeembedded holoreceptors or of control proteins of similar size would have been required to link NMDAR-AB causally with the occurrence of an encephalitis. To validate that "disease induction depends on conformationally restricted epitopes" [15], infusion of purified NMDAR-AB should have been used to clarify whether they act encephalitogenic on their own [16].
During performance of the present work, a report claimed that immunizing mice against the NMDAR1-N368/G369 region alone at high dose induced encephalitis-like behavioral and morphological impairments, including blood brain-barrier (BBB) breakdown [17]. We thus decided to follow this initially exciting report's protocol, hoping to answer several important questions related to our own research. However, using two large independent cohorts of wild-type mice, we were unable to reproduce any of the described findings. This replication failure is now also included in the present paper.
To summarize, despite all efforts, etiology and pathogenesis of the "NMDAR encephalitis" are still as uncertain as the role of NMDAR1-AB in this condition. We note that multiple braindirected AB apart from NMDAR1-AB have been reported in serum of healthy humans and of various other mammalian species, likely belonging to the normal pre-existing "autoimmune repertoire." These AB do have potential functionality and also potential pathogenicity [18][19][20][21][22]. In fact, functional serum NMDAR1-AB of all immunoglobulin classes are a highly frequent finding across mammals, with multiple possible inducers or boosters identified thus far, ranging from genetic predisposition, various viral infections (e.g., influenza, herpes), oncological conditions, and brain lesions to chronic life stress [1,6,[23][24][25][26]. However, no solid proof exists to date that NMDAR1-AB by themselves can induce an encephalitic process.
The present study addresses the hypothesis that "NMDAR encephalitis" may result from a primary brain inflammation coinciding with the presence of NMDAR1-AB, which ultimately shape the encephalitis phenotype. Therefore, a spatially and temporally defined" sterile encephalitis was induced in young female NexCreERT2xRosa26-eGFP-DTA ( = "DTA") mice after immunization with a cocktail of 4 NMDAR1 peptides, also including a peptide covering the NMDAR1-N368/G369 region. Females were chosen to account for the~4:1 female/male ratio observed in human "NMDAR encephalitis" [27]. As shown in vivo earlier, this cocktail induces functionally highly active NMDAR1-AB, leading to psychosis-like symptoms in mice with compromised BBB but no brain inflammation [28]. Here, we report that NMDAR1-AB can contribute to the behavioral abnormalities of an underlying gray-matter encephalitis, but that the multifaceted encephalitic phenotype itself, involving pyramidal neurons and their NMDAR, is nearly identical between NMDAR1-AB carriers and noncarriers.

MATERIALS AND METHODS Mice
Mouse experiments (all C57BL/6 background) were approved by the Local Animal Care and Use Committee (LAVES, Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit, Oldenburg, Germany) in accordance with the German animal protection law. Sample sizes were based on previous experience under consideration of the RRR-principle and technical limitations (i.e., maximum of 16 animals per IntelliCage). All experiments were performed by investigators unaware of group assignment (fully-blinded). Mice were separated by genotype and treatment to avoid inclusion effects [29] or aggressive behavior against potentially affected animals. Unless otherwise stated, mice were maintained in temperature-and humidity-controlled environments in 12 h light/dark cycle (light on at 7am) with wood-chip bedding, nesting material (Sizzle Nest, Datesand, Bredbury, UK) and, ad libitum food and water.
Immunization of the replication cohorts followed the protocol of Wagnon et al. [17]. Male C57B/6 J mice were immunized at 8-9 weeks of age with either GluN1 168-187 (control peptide), GluN1 359-378 (active peptide) or for our additional comparison with ovalbumin, each emulsified in an equal volume of CFA (as described above). Antigens (200 µg) were equally distributed over 4 subcutaneous injections, 2 at shoulders and 2 at hind limbs. In addition, mice received 2 intraperitoneal injections of 200 ng of Bordetella pertussis toxin (#180, List Biological Laboratories) in PBS, immediately after and 48 h after immunization.

Behavioral phenotyping
Experiments of the DTA cohort were performed in the following order: LABORAS (baseline-prior to tamoxifen induction), bar test, hurdle test, IntelliCage-based phenotyping including pheromone preference, LABORAS, Morris water maze, hole board, prepulse inhibition (PPI), marble-burying test, and complex wheel running.
Behavioral testing of the replication cohort followed the design of Wagnon et al. [17]. Behavioral analyses comprised open-field activity, Ymaze working memory, elevated plus maze (anxiety), and forced-swim test (depression-like behavior). For the second replication cohort, we rearranged the test schedule to assess anxiety-like behavior in behaviorally naive mice, followed by testing in open field and Y-maze.

LABORAS
To characterize the spontaneous home cage behavior of mice prior to and after a tamoxifen-induced gray-matter inflammation, the LABORAS system (Metris B.V.) was employed [32][33][34]37]. Briefly, mice habituated for 2 nights to the experimental room and single housing in clear polycarbonate cages (Makrolon type I, 22 cm × 16 cm × 14 cm, Tecniplast) with wood-chip bedding, food, and water ad libitum. After two nights, cages were placed on a sensor platform (Carbon Fiber Plate 1000 mm × 700 mm × 700 mm × 30 mm, Metris B.V., Hoofddorp, Netherland) and the resulting electrical signals recorded throughout the dark phase (12 h) and classified into behavioral categories, i.e., eating, drinking, scratching, circling, climbing, immobility, locomotion, and grooming.

IntelliCage-based phenotyping
To assess a variety of cognitive measures with minimal experimenter intervention on a 24/7 basis in a social home cage-based setting, 16 mice per group were placed in standard laboratory rodent cages (55 × 38.5 × 20.5 cm, Tecniplast) equipped with the IntelliCage ® apparatus (TSE-Systems) controlled by NewBehavior software (version 3.1.7), as we described in great detail previously [29]. Water was accessible in four triangular conditioning chambers (15 × 15 × 21 cm), located in cage corners. The conditioning chambers were equipped with temperature sensors and RFID antennas to identify entering mice. If mice entered a correct corner during the allocated time window, doors blocking access to water or sucrose bottles (dependent on test setting) could be opened via nosepokes sensed by light barriers. Lickometers registered licks on bottles.
The experimental setup comprised measurement of place learning (day 1), reversal learning (day 2), sucrose preference (day 3), acquisition and recollection of a place time-reward/episodic-like memory (days 4 and 5), and behavioral extinction of operant responses (days 6 and 7). General activity and day/night pattern were assessed by recording number and timing of visits to operant chambers (corner visits). On days 1-2, each mouse had access to water at a single corner (4 mice/corner), during days 3-5 to a 2% sucrose solution on one corner and water at the opposite site (8 mice/corner). On days 6-7, mice had access to only water at previously rewarded corners. In-between test sessions, corners were re allocated. The efficiency of place and reversal learning was assessed for 24 h each by calculating the percentage of place errors (visits to blocked corner/total corner visits *100%). Sucrose preference means the percentage of licks at sucrose bottle/total licks within 24 h. To evaluate episodic-like memory, access to water and sucrose solution was restricted to the first 2 h after dark-phase onset. Within this 2 h period, preference to sucrose corner was calculated as percentage of visits to sucrose corner/total corner visits.
Two weeks after this IntelliCage testing, mice returned to IntelliCages to evaluate pheromone preference using two social boxes, supplemented with fresh wood-chip bedding, connected to the left and right side of the IntelliCage via two plastic tubes, equipped with two ring RFID antennas to track individual mouse. The time each mouse spent in the IntelliCage, the neutral or the target social box was recorded along with the number of social box visits. After habituation of 1 h, social boxes were replaced by novel cages filled with either fresh wood-chip bedding (neutral site) or bedding from male C3H mice (target site) and mice again had 1 h to explore the social boxes. In addition to time spent on the target site, the time spent exploring either social box was evaluated as the readout for exploratory behavior.

Complex wheel running
To assess locomotor activity and motor-cognitive learning, a complex running-wheel-setup (CRW) was used. Mice were single-housed in type-III cages (42 × 26 × 18 cm, Tecniplast), equipped with CRW (TSE-Systems). CRW are characterized by randomized omitted bars [39][40][41]. Mice were habituated to the experimental room and CRW for 2 h prior to the dark phase. After dark-phase onset, voluntary running was automatically tracked for 4 h via Phenomaster software (TSE-Systems) and the total running distance per mouse calculated.
Measurements assessing blood-brain-barrier integrity BBB integrity was evaluated as previously described [28,42]. Briefly, mice received intravenous injections of Evans blue (50 µg/g body weight, E2129, Sigma) and sodium fluorescein (200 µg/g body weight, F6377, Sigma). After 4 h, mice were anesthetized and transcardially perfused. Brains were collected, frozen on dry ice, and lyophilized. Tracers were extracted from hemispheres with formamide and quantified in triplicates on a fluorescent microscope (Observer Z2, Zeiss). The concentrations of tracers were calculated using a standard curve and normalized to controls.

Flow cytometry
(Beckman Coulter, Krefeld, Germany). Cell numbers were corrected for the number of recorded APC beads. Leukocytes (CD45 high , CD11b low ) and microglia/macrophages (CD45 mid , CD11b high ) were quantified within single-cell gate determined by forward and side scatter. CD4 + and CD8 + T-cells were quantified within leukocyte gate. CD19 + B cells and CD138 + plasma cells were quantified in CD4 -CD8leukocyte gate.

Statistical analysis
Statistical analyses were performed using Prism software (GraphPad Software) with the exception of the mixed ANOVA that was conducted on PPI data with R4.0.3 [44] using the rstatix package [45]. The results are presented as mean ± standard deviations (SD), with few exceptions as indicated in the figure legends. Data normality was assessed using the   Shapiro-Wilk test with an alpha error of 0.05. Dependent on data distribution, two-tailed unpaired Welch's-corrected t-tests or Mann-Whitney U-tests were used to compare groups of 2. Similarly, repeatedmeasure ANOVA, one-way ANOVA or Kruskal-Wallis test were used to compare multiple groups. P values < 0.05 were considered statistically significant.

RESULTS AND DISCUSSION
Diphtheria toxin-mediated ablation of pyramidal neurons to mimic encephalitis affecting primarily gray matter To mimic features of a viral encephalitis in a spatially and temporally defined sterile experimental approach, we employed young female NexCreERT2xRosa26-eGFP-DTA ( = "DTA") mice with an inducible transgene for diphtheria toxin-mediated ablation of pyramidal neurons, generating gray-matter inflammation [30,31,46]. Cell death can be controlled by dosing tamoxifen. Thus, in a series of dose-titrating pilot experiments, we selected a 2-day tamoxifen-injection design (Fig. 1A, B). After 7 days, we histologically confirmed acute neurodegeneration, pyramidal neuron loss, and a distinct inflammatory response comprising micro-and astrogliosis (Fig. 1C, D).
Generation of primary brain inflammation coinciding with the controlled presence of NMDAR1-AB following immunization We next addressed our hypothesis that "NMDAR encephalitis" may result from a primary brain inflammation concurring with the presence of NMDAR1-AB, which ultimately shape the encephalitis phenotype. Therefore, the above-described, defined sterile encephalitis was induced in young female DTA mice after active immunization with a cocktail of four peptides of extracellular NMDAR1/GluN1 domains, including a peptide covering the NMDAR1-N368/G369 region, versus ovalbumin as control immunization. This cocktail induces sufficient titers of highly functional NMDAR1-AB of the IgG class [28]. Active-immunization was performed at postnatal day 30, followed by an experimental scheme as detailed in Fig. 2A, comprising induction of encephalitis, extensive behavioral testing, blood sampling, and finally perfusion.
Immunization of mice against the 4 NMDAR1 peptides led to high circulating levels (even though somewhat variable between subjects) of specific AB, as shown by ELISA and cell-based assay, which persisted throughout the experiment, i.e., over 4 months. Titers did not differ between DTA mice and controls (Fig. 2B-D).
Comparison of multidimensional behavioral readouts among DTA and control groups with or without NMDAR1-AB Testing mice in a multifaceted battery should reveal behavioral domains affected by the induced gray-matter encephalitis and shaped by NMDAR1-AB (Fig. 2E-R; Table 1). Indeed, spontaneous home cage behavior, measured by LABORAS, and corner visits as activity readout in the IntelliCage setup, revealed considerable hyperactivity of the DTA mice, interpretable as psychosis-like behavior or "loss of inhibition" [47]. The hyperactivity measures substantially inter correlated and were more pronounced upon the presence of NMDAR1-AB (Fig. 2E-G). Additional evidence of pathological hyperactivity is presented in Table 1. In fact, this psychosis-like feature seemed to persevere, since also the final test performed before perfusion, complex wheel running, similarly indicated hyperactivity (Fig. 2H).
For hippocampal learning and memory, the classical Morris water maze (MWM) test was employed [48]. Whereas the visible platform days showed comparable ability of task learning among groups (Fig. 2I-K), substantial deficits arose regarding the hidden platform learning curves (Fig. 2L) and the probe trial results (Fig. 2M). Here, DTA mice, independent of the presence of NMDAR1-AB, demonstrated inferior performance consistent with hippocampal damage. This was also seen as a clear tendency in MWM reversal testing as hippocampus-dependent cognitive flexibility measure, again without an appreciable NMDAR1-AB effect (Fig. 2N-O).
Prepulse inhibition of the startle response (PPI) was employed as a surrogate marker for gating defects in the prefrontal cortex-known also as a highly relevant translational test in human patients with psychosis. PPI deficit is among the most reliable objective features of severe neuropsychiatric phenotypes, likely affecting the ability to adequately filter and interpret environmental stimuli [47,[49][50][51][52]. Indeed, DTA mice, again independent of NMDAR1-AB, showed an overall tendency of a PPI deficit, pointing to prefrontal cortical network dysfunction as a consequence of the induced pyramidal neuronal death (Fig. 2P).
To some surprise, our previously designed extensive cognitive, emotional, and social phenotyping of mice in an observerindependent setting, using IntelliCages [29], did not reveal any considerable differences, except for the above-delineated distinct hyperactivity. This is most likely explained by lower sensitivity or ceiling effects of this paradigm (requiring a higher level of damage to show impairment), and the just partial involvement of hippocampal functions that are apparently fully compensated (example of IntelliCage readouts in Fig. 2Q, R). Similarly, all other behavioral tests failed to show appreciable alterations. Few just marginally significant and sometimes rather counter intuitive results proved invalid upon multiple-testing correction (Table 1). In conclusion, behavioral testing uncovered the expected consequences of pyramidal neuronal degeneration, i.e., substantial hippocampal dysfunction (MWM), a strong tendency of gating deficits as prefrontal cortex measure (PPI), and lasting psychosislike hyperactivity, with only the latter augmented by the presence of NMDAR1-AB. Fig. 2 Pathophysiological relevance of NMDAR1-AB (=GluN1-AB) in the context of gray-matter inflammation. A Experimental outline indicating experimental groups, as well as order and time of behavioral tests (DPI = days post induction, i.e., after the last tamoxifen injection). B NMDAR1-AB validation by ELISA. C Cell-based (HEK293T) clinical standard assay for NMDAR1-AB (Euroimmun). D Immunocytochemical colocalization (CBA, Euroimmun) with a commercial rabbit GluN1-AB directed against the C-terminal domain. E Intra individual change of locomotor activity assessed in LABORAS at baseline and after tamoxifen induction. For each mouse, time in locomotion after tamoxifen induction was normalized to time in locomotion prior to tamoxifen application (baseline). F Activity (number of corner visits) over a 7-day IntelliCage session. G Pearson correlation between IntelliCage activity and intra individual changes of LABORAS locomotor activity. H Locomotor activity assessed by 4 h voluntary complex wheel running. I-O Cognitive testing in Morris water maze (MWM). I-K Visible platform task comprising 2 training days, demonstrating the ability for fast escape and simple task learning using within-maze cues. L Training of hidden platform task using extra-maze cues. DTA mice performed significantly worse than control mice (repeated-measures ANOVA, p < 0.0001), whereas no effect of NMDAR1-AB was observed in either DTA (p = 0.3273) or control mice (p = 0.5972). M Evaluation of spatial memory in the probe trial. N Spatial reversal of the hidden platform. O Evaluation of cognitive flexibility and reversal learning in a second probe trial after spatial reversal training. P Prepulse inhibition of acoustic startle. Intra group comparisons performed using repeated-measure one-way ANOVA; inter group comparison between genotypes for 75 and 80 dB prepulses performed using mixed ANOVA. Q IntelliCage-based evaluation of place learning and R reversal learning within 24 h sessions shows similar performance across groups. Dashed lines indicate performance at chance level (75%). Experiments were performed with 14-16 mice/group, except for CRW (H, n = 6 mice/group). Data presented as mean ± SD, except for repeated measure data (L, N, P; mean ± SEM) and AB titers (C, median, range).   BBB integrity and measures of chronic inflammation in DTA and control groups with or without NMDAR1-AB Using fluorescent tracers, as described in detail previously [28,42], we addressed BBB leakiness in our DTA encephalitis model as a prerequisite for substantial NMDAR1-AB transfer to the brain and exertion of measurable effects. Tracer extravasation was enhanced for Evans blue and displayed a strong trend for fluorescein, confirming persistent BBB dysfunction. Brain water-content data showed remarkable scatters but were comparable in DTA and control mice at 3 months after encephalitis induction (Fig. 3A-D).
The chronic microglia and astrocyte response after tamoxifeninduced neuronal death in DTA mice was slightly milder as compared with the acute situation (see Fig. 1C, D), but still very obvious already from overview images (Fig. 3E, F). Histological quantification yielded distinct reductions of the areas of whole hippocampus, CA1 and CA3, whereas the dentate gyrus just revealed a similar tendency (Fig. 4A). Both Iba1 and GFAP quantifications followed the same pattern: strong microgliosis and astrogliosis in the whole hippocampus, CA1 and CA3, but only moderate in dentate gyrus of both DTA groups, with no significant differences in NMDAR1-AB carriers (Fig. 4B, C). NMDAR1immunized control mice showed no signs of inflammation as reported previously [28]. Change of inhibition could be demonstrated by the relative increases in parvalbumin-positive interneurons upon DTA-induced encephalitis, sparing the dentate gyrus. Also regarding these interneurons, NMDAR1-AB presence did not modulate the ensuing picture (Fig. 4D, E). Hippocampal immune cell infiltration, determined by counting CD45 + cells, was modest but clear in DTA mice (Fig. 4F). Collectively, these data document persistent BBB disruption and distinct chronic inflammation in our DTA model but no obvious amplifying influence of NMDAR1-AB on any of these readouts. This additionally argues against any appreciable proinflammatory role of NMDAR1-AB of the IgG class.
Failure to induce any signal of encephalitis by immunizing mice solely against the NMDAR1-N368/G369 region During performance of the present work, a report claimed that immunizing mice against the NMDAR1-N368/G369 region alone induced an encephalopathy with remarkable B-cell response, provoking an autoimmune reaction against NMDAR, and encephalitis-like behavioral deficiencies [17]. We were first excited and wondered whether this model would finally help us to answer several burning questions of our own research, addressing for instance, the cellular-molecular mechanisms of NMDAR1-ABmediated brain inflammation, which we had not been able to observe [28], or late consequences of an autoimmune encephalitis, or reasons for NMDAR1-AB titer fluctuations including the occurence and nature of late boosters [53]. Therefore, we started an extensive experiment, ultimately using two large independent cohorts of wild-type mice, and followed meticulously this report's protocol [17]. However, as detailed in Fig. 5 and Table 2, none of the described findings could be reproduced, although the adopted immunization protocol worked well. We found specific GluN1 359-378 -AB of the IgG class in serum and no immune response using the non immunogenic control peptide GluN1 168-187 (Fig. 5A-D). There was no BBB disturbance (Fig. 5E), all behavioral tests were normal (Fig. 5F-I), histology did not show any abnormalities (Fig. 5J-L), and brain FACS results were physiological and did not reveal any differences between groups (Fig. 5M-R). Therefore, a second mouse cohort was employed to exclude potential by-chance failure of replication but, disappointingly, yielded the same overall negative results (data not shown).
Simultaneous with our anticlimactic replication failure, Ding et al. published a study, using a similar immunization strategy to investigate the pathogenicity of various GluN1 peptides (each also at a dose of 200 µg), including the GluN1 359-377 . Yet, in contrast to the claims of Wagnon et al., the authors did not find GluN1 359-377 - AB in the CSF of immunized mice [54]. In addition, these authors validated the functionality of GluN1 356-385 -specific AB and assessed the in vivo consequences after triple immunization in combination with pertussis toxin, resulting in a behavioral phenotype (impaired social memory and novel object recognition, normal anxiety-, and depressive-like behavior), distinct from the one reported by Wagnon et al. (impaired spatial memory, abnormal anxiety-, and depressive-like behavior) [17,54]. Unfortunately, Ding Fig. 3 Long-lasting inflammatory response at 3 months after tamoxifen induction. A Abbreviated experimental outline (detailed schematic in Fig. 2A). Blood-brain barrier (BBB) permeability, irrespective of NMDAR1-AB status, assessed by Evans blue (B) and fluorescein (C) extravasation, as well as brain-water content (D) in the forebrain and cerebellum of DTA and control mice. Data presented as mean ± SD. E, F Representative images demonstrating persistent inflammatory changes in the hippocampus of DTA mice, including increased microglia density and apparent changes in morphology (E, quantifications in Fig. 4B) and increased expression of glial fibrillary acidic protein, GFAP (F), quantified in Fig. 4C). High-resolution images of CA1, CA3, and dentate gyrus (DG) regions were acquired as 10 µm Z-stacks and displayed as maximum-intensity projections.
on the chronic effects of NMDAR1-AB upon immunization against GluN1 402-421 peptide. Similar to our experience with these immunizations, mice remained healthy and did not develop any encephalitic signs, despite the presence of persistent and functional NMDAR1-AB. The only behavioral alteration reliably observed in GluN1 402-421 -immunized mice was impaired spatial memory and/or novelty recognition, assessed as spontaneous alternation in T-maze [55]. These findings are highly interesting since they underscore the possibility that chronic presence of circulating NMDAR1-AB can mildly modulate behavior, even in the absence of an appreciable BBB dysfunction. This is in line with a previous expert review [18] and our experimental observation that NMDAR1-AB can reach the brain to bind there at low titers even in healthy wild-type mice [26].
We are aware that-for comparison-in experimental autoimmune encephalitis (EAE), fluctuations can occur regarding the severity of the clinical/pathological picture [56][57][58]; however, the absence of all claimed readouts as obtained in our replication approaches, is unheard of and quite surprising. This is more so, since the message of the report by Wagnon et al. will remain, if not questioned, and be taken for granted. This in turn can become an ethical issue, leading to (pre)clinical conclusions that are ultimately damaging for other scientists and unfortunately also for patients, as we observe on a regular basis.
Conclusions from the NMDAR1-immunized DTA model: NMDAR1-AB modify rather than cause an encephalitis Using a well-standardizable, spatially and temporally defined mouse model of viral encephalitis by employing controlled DTA induction in pyramidal neurons, we find a multifaceted encephalitic phenotype, which persists over months. This phenotype involves pyramidal neurons and thus of course also their NMDAR, but is only marginally aggravated in NMDAR1-AB carriers versus non carriers. The aggravation essentially rests on hyperactivity as     psychosis-like behavioral readout but is not reflected in any histological quantification. Similarly, comparison of NMDAR1-ABpositive and -negative human encephalitis cases did not reveal appreciable differences, except for few NMDAR-antagonistic (ketamine-like) symptoms [14]. Most likely, human "NMDAR encephalitis" is simply not a separate condition, but rather marks an encephalitis where the highly prevalent NMDAR1-AB and/or the respective B cells happen to be present in the brain and shape the clinical picture. Therefore, it may be problematic, if the search for encephalitis causes stops after detection of NMDAR1-AB (of the IgG class). Reassuringly, "polypragmatic" treatment of any encephalitis of unknown origin (constituting the majority of cases) should anyhow include antibiotics, antivirals, and eventually corticosteroids/immunosuppressants on top of supporting measures. The N-terminal domain containing the G7 epitope (N368/G369) was first deemed pathognomonic for "NMDAR encephalitis," and believed to be the target region of the pathological NMDAR1-AB of the IgG class seen responsible for this condition [59]. Therefore, the respective immunization model, leading to encephalitis as described by Wagnon and colleagues [17], seemed attractive at first view and worthwhile pursuing. Unfortunately, it was not reproducible in our hands, and is not supported by a similar recent paper [54]. Searching for an explanation by speculating about possible reasons why Wagnon and colleagues found signs of an encephalitis that we did not see, interfering "iatrogenic" issues may be worthwhile considering, like e.g., undetected subclinical infections in their animal facility, leading to "occult" brain inflammation. Also, other factors, e.g., differences in the gut microbiota, which have the potential to modulate disease progression in EAE models [60], cannot be entirely excluded. However, it seems rather unlikely that the overall discrepancy, including BBB dysfunction, can be explained by such physiological factors.
Considering our own previous findings on epitopes recognized by the highly frequent NMDAR1-AB found in human serum, the negative outcome of the replication attempts is actually not too surprising. Epitope mapping using 7 different NMDAR1 constructs revealed recognition by NMDAR1-AB-positive sera of different epitopes, located in the extracellular ligand-binding and the Nterminal domain, as well as the intracellular C-terminal and the extended pore domain. NMDAR1-AB seropositivity was polyclonal/ polyspecific in half of the investigated sera and likely mono-or oligoclonal/oligospecific (mainly IgG) in the other half. Overall, no particular disease-related pattern appeared. NMDAR1 epitopes were comparable across health and disease [61]. Also, the accentuated role of IgG in "NMDAR encephalitis" is still a matter of speculation, but likely related to inflammation-induced class Table 2. Detailed presentation of behavior results obtained in replication study (protocol following Wagnon et al. [17] switch in the brain [62]. As mentioned in the Introduction, NMDAR1-AB are only one of many possible autoantibodies directed against brain epitopes [18][19][20][21][22]. The finally resulting phenotype would then depend on (i) the specific site(s) of brain inflammation, with either resident plasma cells producing AB or a local extent of BBB breach at that site (to allow sufficient AB transfer to the brain), and (ii) the specific circulating brain-reactive autoantibody profile of each individual.
To conclude, while NMDAR1-AB can contribute to the behavioral phenotype of an underlying encephalitis, there is no proof at present for induction of an encephalitis by NMDAR1-AB themselves. Thus, based on the results presented here, the answer to the question of whether or not NMDAR1-AB can, by themselves, induce encephalitis is probably no, with the caveat that perhaps it may be possible in extremely rare patients with an exceptionally high NMDAR1-AB titer. However, this would assume that a very high titer is somehow linked to brain inflammation (as cause or consequence) and/or local increases in BBB permeability, which is neither supported by the data from this nor from other studies.

DATA AVAILABILITY
All data are available upon request.