Neuropathological Variability within a Spectrum of NMDAR‐Encephalitis

To describe the neuropathological features of N‐methyl‐D‐aspartate receptor (NMDAR)‐encephalitis in an archival autopsy cohort.

E ncephalitis with IgG autoantibodies against the Nmethyl-D-aspartate receptor (NMDAR) is associated with a characteristic clinical syndrome presenting with acute psychiatric symptoms, cognitive deficits, epileptic seizures, movement disorders and autonomic dysregulation. 1 In vivo and in vitro experiments demonstrated that the antibodies bind to an extracellular region of the NMDAR causing internalization of the receptor and neuronal dysfunction. [2][3][4] NMDAR-encephalitis is characterized by prominent intrathecal production of pathogenic autoantibodies and the reversibility of neuronal dysfunction is reflected by good response to immunotherapy. 1,5 Due to the frequent recovery of patients, neuropathological studies on NMDAR-encephalitis are rare and mainly based on isolated cases with overlapping pathologies or small biopsy specimens, lacking systematic assessment of immune cell infiltration and anatomic correlations. [6][7][8][9][10][11][12][13][14] This study aims to characterize the spectrum of inflammatory changes in different brain areas in patients who have or have not been treated with immunotherapy. Moreover, we describe the overlapping neuropathology of NMDAR-encephalitis with multiple sclerosis (MS)-type demyelination and post-transplant lymphoproliferative disease (PTLD).

Patient Identification and Inclusion Criteria
Patients were identified at the Neurobiobank of the Division of Neuropathology and Neurochemistry, Department of Neurology, Medical University of Vienna, Austria (3 autopsies) and the Neurological Tissue Bank of the Hospital Clinic -IDIBAPS, Barcelona (1 autopsy). Inclusion criteria were: (1) positive NMDAR-IgG in CSF determined by cell-based and tissue-based assays, and (2) sufficient archival tissue for pathological analysis. A total of 4 autopsies were included in the study. The study was approved by the Institutional Review Board of the Medical University of Vienna (EK 1123/2015), and the Ethics Committee of Hospital Clinic, Barcelona (R091217-12). Patient demographics, clinical data, comorbidities and cause of death are presented in Table 1. Patient 2 was briefly reported in a series of patients with autoimmune encephalitis. 15 Neuropathology and Immunohistochemistry Neuropathological analysis was performed either on double hemispheric brain sections or small sections. Formalinfixed and paraffin-embedded tissue blocks representing frontal, temporal, parietal and occipital cortices and adjacent white matter, thalamus, hippocampus, basal ganglia, amygdala, midbrain, pons, medulla oblongata, and cerebellum were regions of interest. All sections were stained with hematoxylin and eosin (H&E), and a panel of primary antibodies was used for immunohistochemistry techniques. The primary antibodies and type of antigen retrieval used are summarized in Table 2. Sections stained in the absence of primary antibodies and using isotypematched control antibodies served as controls. Image acquisition was performed on a NanoZoomer 2.0-HT digital slide scanner C9600 (Hamamatsu Photonics, Hamamatsu, Japan). Double hemispheric brain sections were scanned using a custom slide scanner as described previously. 16 In situ hybridization for Epstein Barr Virus (EBER) was performed using the EBER pNA detection kit (DAKO Y5200), including an Epstein-Barr virus + (EBV) cerebral lymphoma as a positive control.
For double immunostaining using primary antibodies derived from different species, the same antigen retrieval techniques were applied (Table 2). Visualization was performed by using (1) alkaline phosphatase-conjugated secondary antibodies for subsequent development with fast blue BB salt as well as (2) biotinylated secondary antibodies and peroxidase-conjugated streptavidin for subsequent development with aminoethyl carbazole (AEC). 17,18 NMDAR Antibody Testing CSF NMDAR antibodies were assessed using an in-house tissue-based assay as described elsewhere 1 and a commercial cell-based assay (Euroimmun, Lübeck, Germany) according to the manufacturer's protocol. Positivity was defined as a positive neuropil staining pattern in the tissue-based assay and positive labeling of transfected cells.

Quantitative Analysis
All quantitative data were obtained by one investigator blinded to the clinical data. All sections were overlaid with a microscopic grid and 17 randomly distributed fields in the region of interest were counted. In each ROI, 1 mm 2 was quantified. All values are expressed as cell counts per square millimeter. The number of T cells and B cells were counted separately in the perivascular and parenchymal areas. To assess the anatomic distribution of CD79a positive plasmablasts/plasma cells (cytoplasmic expression of CD79a in plasma cells) in NMDAR-encephalitis overlapping with PTLD, a bi-hemispheric section was stained with anti-CD79a, scanned and each immunolabelled cell was classified as a red dot, using the bioimage analysis platform QuPath. 19

Patients
The median age at disease onset was 45 years, and two patients were female (Table 1). Two patients died in 2004 and 2006 before NMDAR-encephalitis was discovered or systematically tested, and did not receive immunotherapy. NMDAR antibodies were retrospectively tested in archival CSF samples. Two patients were diagnosed with NMDAR-encephalitis during their lifetime and treated with immunotherapy, but developed a co-pathology and died. A summary of the clinical information is shown in Table 1.
Neuropathology of NMDAR-Encephalitis without Immunotherapy Patient 1 was a 33-year-old female patient with unremarkable past medical history or family history and without children, who presented with severe head and neck pain, auditory hallucinations, agitation, fever and generalized seizures. She was admitted to the hospital and was treated for a potential herpes encephalitis as well as bacterial encephalitis with antibiotics and antiviral medication (acyclovir, doxycycline, ceftriaxone). Lumbar puncture revealed 18 WBC/μl, normal protein concentration (37.3 mg/dl) and CSF-specific (serum unmatched) oligoclonal bands. Microbiological studies revealed positive serum IgM antibodies directed against mycoplasma pneumoniae antigen, microbiological cultures and PCR analysis for viruses (including HSV1 and HSV2) resulted negative. Over the ensuing days, she developed central hypoventilation and was transferred to the ICU. Brain MRI revealed bilateral temporal edema. Any attempt to taper the sedation resulted in uncontrollable seizures. Four months after transfer to the ICU, she developed multiple bilateral pulmonary embolisms and died. The patient never received any steroid or immunomodulatory treatment due to the suspicion of an infectious disease. Postmortem examination did not reveal any malignancy or ovarian teratoma. Patient 2 was a 76-year-old male patient with a history of smoking, arterial hypertension, chronic renal insufficiency, and chronic diverticulosis. His daughter noticed behavioral abnormalities, disinhibition, and confusion and brought him to the hospital. Toxicological tests, chest Xray and cranial CT were normal. Over the ensuing days he developed severe anterograde amnesia, confabulations, and preservation of remote memory. CSF showed 20 WBC/μl, normal protein (51 mg/dl) and glucose (61 mg/dl). Microbiological studies and PCR for viruses (including HSV1 and HSV2) were negative. Brain MRI showed small vessel arteriopathy but was otherwise unremarkable. The patient rapidly progressed, became mute without responding to commands, presented frontal release reflexes and an increased muscle tone in all 4 limbs. EEG showed diffuse slowing without epileptic features. The patient developed a syndrome of inappropriate antidiuretic hormone secretion (SIADH) with hyponatremia as well as atrial fibrillation. The patient died 31 days after disease onset due to cardiac arrest. He did not receive any steroid or immunomodulatory treatment. Postmortem examination revealed small cell lung carcinoma with mediastinal and hepatic metastasis. A tumor in the testes was not found. Postmortem brain studies of both patients demonstrated mild to moderate inflammatory infiltrates within the basal ganglia, amygdala, and the surrounding white matter.

Neuropathology of NMDAR-Encephalitis with
Overlapping CNS Comorbidities NMDAR-Encephalitis and PTLD. A 33-year-old man with past medical history of renal transplantation under immunosuppressive therapy (cyclosporine, mycophenolate mofetil, and steroids) presented 13 years after transplantation to the emergency department with behavioral changes, confusion, memory deficits, and hypersexuality over the previous 2 weeks. Brain MRI was unremarkable. The CSF showed pleocytosis (45 WBC/μl), increased protein concentration (54.0 mg/dl), CSF-specific (serum unmatched) oligoclonal bands, and positive EBV-PCR. NMDAR antibodies were positive in serum and CSF. The patient received steroid pulse therapy, plasma exchange, and intravenous immunoglobulins (IVIGs), which lead to partial neurological improvement. Treatment escalation with rituximab was initially avoided due to the risk of EBV-encephalitis. Six months later, he was admitted to the emergency department with altered mental status. Laboratory findings showed hyponatremia, and brain MRI revealed a 1.4 cm mass in the left globus pallidus. CSF showed mild pleocytosis (16 WBC/μl) and continued to be EBV-PCR positive (1.12 Â 10 5 copies/ml) and NMDAR-antibody positive (titer 1:8). Brain biopsy showed a monomorphic PTLD (EBV-associated diffuse large B cell lymphoma, DLBCL). The patient was treated with immuno-and chemotherapy (rituximab, vincristine, methotrexate, and procarbazine) and radiation; however, he died 4 months later due to cardiopulmonary arrest.
Post mortem examination of the brain showed a resection cavity in the left frontal lobe and a necrotic tumor lesion in the left globus pallidus (Fig 2A, B) accompanied by a small residual area of PTLD with cells positive in the in situ hybridization for Epstein-Barr virus latencyassociated RNA (EBER) (Fig 2C).
Anatomic distribution of inflammation in the tumor-affected left hemisphere showed maximal inflammation around the necrotic tumor lesion, with macrophages and activated microglia that strongly expressed pro-inflammatory activation markers HLA-DR, CD68, and IL18 ( Fig 2D) along with numerous EBER-negative plasma cells (Fig 2E). Plasma cells were also abundant in the medial temporal lobe, including the amygdala ( Fig 2F) and hippocampus, as well as inferior and medial temporal gyrus. The right hemisphere did not show any tumor lesions and type of inflammatory infiltrates and distribution was compatible with NMDAR-encephalitis (Fig 2B,  G). In both hemispheres, the inflammation was predominantly composed of CD20 À /CD79a + plasmablasts/plasma cells and only single CD20 + B cells and T cells (Fig 2H--K). Plasma cells showed polytypic staining for kappa and lambda light chain and were EBER-negative. Immunohistochemistry for NMDAR only revealed a mild reduction of immunoreactivity in the hippocampus compared to a healthy control (data not shown). No neuronal loss was observed except around the necrotic tumor lesion.
NMDAR-Encephalitis and Demyelination. A 42-year-old woman with past medical history of borderline personality disorder presented with religious delusions, acoustic hallucinations, mutism, and refusal to eat. Brain MRI was unremarkable; CSF showed normal cell count, positive oligoclonal bands, and NMDAR antibodies, which were also present in serum. Myelin oligodendrocyte glycoprotein (MOG)-and aquaporin-4 (AQP4)-antibodies were negative in serum and CSF. Studies for an infectious process were negative. The patient received initial steroid pulse therapy, followed by IVIGs that lead to fast improvement. CT and MRI of the pelvis were normal. Four years later, she returned because of agitation, hemiparesis and a complex pain syndrome. NMDAR antibodies were again identified in serum (1:20) and CSF (1:8); the CSF revealed 0 WBC/μl, protein concentration (16.4 mg/dl) and CSF-specific (serum unmatched) oligoclonal bands. Brain MRI showed T2 hyperintense white matter abnormalities suggestive for multiple sclerosis. The patient again received steroid pulse therapy and IVIGs, which lead to improvement of the hemiparesis; however, she refused other treatments. Five months later, she was found dead at home; a forensic autopsy suggested that the cause of death was sudden cardiopulmonary arrest.
Analysis of whole brain hemispheric sections showed predominantly periventricular demyelinating plaques with perivenous finger-like extensions into the adjacent white matter (Dawson Fingers) (Fig 3A). In addition, multiple   Two untreated cases, 2 with co-pathology (1 PTLD, 1 demyelination); inflammatory infiltrates most prominent in clinically correlated regions: basal ganglia, amygdala, hippocampus: T cells (between 48-90% perivascular and 84-96% parenchymal) and B cells/plasma cells (CD20 + cells: between 0.4-17% perivascular and 0-3.6% parenchymal; CD79a + cells: between 10-41% perivascular and 0-15.6% parenchymal); co-pathology influences spatial distribution and composition of inflammation; decrease of NMDAR-immunoreactivity in hippocampus; no complement deposition, no obvious neuronal loss; Present manuscript smaller perivenous inflammatory demyelinating lesions throughout the deep white matter were seen (Fig 3A-G). The small perivenous lesions displayed active and demyelinating/post-demyelinating activity, 20 characterized by LFB-and MBP positive myelin degradation products within macrophages (Fig 3F, G). The confluent periventricular lesions showed a rim of activated microglia (TMEM119 + /P2RY12 À ) at the border and an inactive lesion center (Fig 3H, J). The axons were relatively well preserved, with some axonal spheroids at the lesion margin. Few plaques were remyelinated and presented as shadow plaques (Fig 3K). TPPP/p25 positive oligodendrocytes were numerous at the margin of the lesion and periplaque white matter but were almost absent in the inactive lesion centers (Fig 3L, M). Cortical or deep grey matter demyelination was absent. None of the lesions showed deposition of activated complement (C9neo antigen). TUNEL staining revealed some cells with DNA fragmentation in the inactive lesion centers that were negative in the double-staining for MOG and TPPP/p25 (data not shown). Inflammatory reaction was prominent and consisted of perivascular B cells (CD20 + ) and plasma cells (CD79a + IgG + ) next to abundant CD3 + and CD8 + perivascular and parenchymal T cells (ratio CD8:CD4 was 1.38:1) (Fig 3N-Q). Overall, the demyelinating lesions showed the pathological hallmarks of typical MS with inflammatory demyelination and remyelination, without antibody or complement deposition, compatible with a pattern I demyelination. 21 However, in contrast to classic MS, which is dominated by CD8 + T cell inflammation, the inflammatory infiltrates of this patient contained a relatively higher number of CD4 + T cells in addition to abundant infiltrating CD20 + B cells and plasma cells. We did not find features of pattern II demyelination with complement and immunoglobulin deposition, or pattern III demyelination defined by distal oligodendrogliopathy, oligodendrocyte apoptosis, or concentric type of demyelination in this patient. B cells and plasma cells were EBERnegative.
In the normal-appearing white matter, prominent T and B-cell infiltration in the parenchyma and perivascular space was observed. The highest number of parenchymal lymphocytes was identified in the basal ganglia, CA4 sector in the hippocampus, adjacent hippocampal white matter, periaqueductal grey matter and the red nucleus, tegmentum, and the formatio reticularis. Overall, the parenchymal cell distribution was dominated by T-cells with a substantial percentage of plasma cells: 72.2% CD3 + T-cells (72.4% of them were CD8 + ), 26.2% CD79a + plasmablasts/plasma cells and 1.5% CD20 + B cells. In the perivascular compartment the overall cell distribution was 47.3% CD3 + (of these 71.8% CD8 + ), 16.5% CD20 + and 36.2% CD79a + cells. The dura mater with sinus sagittalis was available and showed focal infiltrates of CD79a + /CD138 + plasma cells and T cells. No lymph follicles were found (data not shown). Immunohistochemistry for NMDAR revealed a mild reduction of immunoreactivity in the hippocampus compared to a healthy control.
Compared to the average number of lymphocytes in the two untreated NMDAR-encephalitis cases, we found intra-parenchymally 5.9 times more CD3 + T cells, 8 times more CD8+ T-cells, and 10 times more CD20 + B cells and CD79a + plasmablasts/plasma cells. Perivascularly we observed 3.9 times more CD3 + T cells 5.6 times more CD8 + T cells, 7.2 and 6.3 more CD20/CD79a + B cells, respectively.

Discussion
This is an extensive neuropathological study of NMDARencephalitis with and without overlapping brain pathologies. We observed a different qualitative, quantitative, and topographic distribution of inflammatory reaction, depending on the association with other disorders. In the two untreated NMDAR-encephalitis patients, we found that the areas with maximal inflammation were the amygdala, hippocampus, and basal ganglia, which correlated well with the clinical presentation of abnormal behavior, memory dysfunction, and movement disorders. The numerous plasma cells probably explain the intrathecal NMDAR antibody production reported in patients with this encephalitis. 1,5 In addition, we found numerous CD3 + /CD8 À helper T cells necessary for activating B cells and antibody production, whereas no cytotoxic T cells (CD8 + granzymeB + T cells) were identified. Neurons were well preserved but had a reduced NMDARimmunoreactivity as already shown in previous reports (current and previously reported pathological findings are summarized in Table 3). 2,8,9 The extent of decreased immunoreactivity for NMDAR correlated with disease severity (eg, it was most reduced in the patient with intensive medical treatment), which supports the primary antibody-mediated mechanism in NMDAR-encephalitis. No complement deposition was found. Overall, the findings are clearly different from those associated with cytotoxic T cell mechanisms (with CD8 + granzymeB + T cells and substantial neuronal loss). 7,22,23 One of the patients with kidney transplantation developed NMDAR-encephalitis while he was on immunosuppressive therapy. Together with the NMDARantibodies, he was found to be EBV-positive in CSF. Later, the patient developed an EBV-associated PTLD and was therefore treated with rituximab. Isolated cases of NMDAR-encephalitis have been described in patients treated with immunosuppressants after solid organ transplantation or allogeneic hematopoietic stem cell transplantation. [24][25][26][27][28][29][30] In our patient, it is unclear whether the encephalitis was related to the immunological dysfunction caused by the chronic immunosuppression or the destructive effect of the brain PTLD lesion similar to that seen after herpes simplex type I encephalitis. 31 However, the development of the NMDAR-encephalitis when the brain MRI was normal does not support the latter possibility. Four of the reported 6 cases with NMDAR-encephalitis post-solid organ transplantation had a positive EBV PCR in CSF that initially suggested EBV-related encephalitis. [26][27][28][29] Patients with long-term immunosuppressive therapies are typically at risk of infection or reactivation of EBV. A typical aspect of EBV infection is a polyclonal proliferation of B cells with accompanying antibody formation. It was suggested that in these 4 patients the polyclonal proliferation of B cells was accompanied by the synthesis of antibodies against the NMDAR. In this setting, rituximab is a reasonable option for the treatment of the polyclonal proliferation of B cells as well as for the treatment of NMDARencephalitis. In our patient, neuropathological investigations revealed a residual necrotic tumor in the basal ganglia that was surrounded by numerous plasma cells along with macrophages expressing HLA-DR, CD68, and IL18. Whether this plasma cell accumulation is a result of the proinflammatory microenvironment and/or correspond to a residual infiltrate of the treated PTLD is unclear, but the polytypic staining for kappa and lambda light chains and the prominent neurological symptoms along with the persistent NMDAR antibodies in the CSF make a component of the anti-NMDAR immune response more likely.
The other patient with concomitant pathology showed an overlapping syndrome of NMDAR-encephalitis with demyelination. Approximately 5% of NMDAR-encephalitis patients develop a demyelinating disease before, simultaneously, or after the onset of NMDAR-encephalitis. Some are positive for MOG-or AQP4-antibodies, while others remain seronegative. [32][33][34] Other patients have anti-NMDAR encephalitis that overlaps with MS. [35][36][37][38][39][40][41][42][43] In our patient, brain MRI was initially unremarkable and demyelinating lesions were first found 4 years after the onset of NMDAR-encephalitis and showed imaging features more typical of MS than NMOSD or MOG-antibody-associated disease. The mechanisms that trigger overlapping NMDAR-encephalitis and demyelination are unclear and the pathogenesis may be heterogeneous. In cases with concomitant MOG-or AQP4-antibodies, a genetic susceptibility towards humoral autoimmunity may play a role. Genetic susceptibility variants for NMDAR-encephalitis have only recently been described, with one gene also being under debate in multiple sclerosis. 44,45 In our case, the morphology of the demyelinating lesions was consistent with classic MS with periventricular, plaque-like demyelination with radially expanding lesions, and signs of remyelination with shadow plaques. The absence of complement deposition in the demyelinating lesions may indicate other inflammatory mechanisms, such as microglia-driven tissue injury. Indeed, we found high numbers of activated microglia (TMEM119 + ) in the normal-appearing white and grey matter expressing the pro-inflammatory markers HLA-DR, CD68, and loss of the homeostatic marker P2RY12. 46 Of note was the intense inflammatory infiltrate composed of plasma cells and T cells that was 6 to 10-fold higher than in the two untreated NMDAR-encephalitis cases. The concomitant pathologies may have mutually reinforced the inflammation and microglia activation, which would explain previously published clinical observations that NMDARencephalitis concurrent with demyelination may require more intensive immunotherapy and result in more residual deficits compared with isolated NMDAR-encephalitis. 32 In some patients, white matter changes may be caused by the NMDAR antibodies, although these patients usually do not develop clinical MRI features of demyelination. 47 In fact, in these patients the alteration of myelin is better observed with advanced MRI than in routine MRI studies. 48 Two of our patients underwent a forensic autopsy, one to exclude medical error in the treatment of unclear encephalitis (the patient died before NMDAR-encephalitis was discovered), the other because she was found dead at home. Forensic (neuro-) pathologists are often confronted with challenging situations and should be aware of NMDAR-or other autoimmune encephalitides as a possible cause of unclear encephalitis. Particularly cases of acute psychosis and neuropathological findings of abundant plasma cells should prompt antibody testing in stored serum and CSF samples.
In conclusion, our study reveals several important findings: (1) In untreated patients, the clinical presentation of abnormal behavior, memory dysfunction, and movement disorders correlates well with the topographic distribution of the inflammation with plasma cells in amygdala, hippocampus, and basal ganglia, while disease severity correlates with the decrease of NMDAR-immunoreactivity, (2) overlapping pathologies with NMDARencephalitis may change the distribution and composition of inflammatory infiltrates, and the pro-inflammatory microenvironment may enhance the intensity of inflammation, which in turn may influence the clinical presentation and outcome of patients, and (3) pathogenetic mechanisms of overlapping NMDAR-encephalitis and demyelination may be heterogeneous and the tissue injury may be driven by pro-inflammatory microglia and macrophages in MOG-or AQP4-negative patients.