Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies

We appreciate the authors’ time and effort in responding to our comments, clarifying their cell-based assay method (permebealized and fixed) and tissue fixation (optimized for detection of plasma membrane protein-reactive IgG antibodies, rather than GFAP-IgG).¹ We should also add: the majority of initially reported historical autoimmune GFAP astrocytopathy patients had undergone brain biopsy to determine etiology, prior to autoimmune diagnosis and treatment.² We are thankful invasive surgical procedures are no longer required to make this diagnosis.

To offer the highest quality service in the Mayo Clinic Neuroimmunology Laboratory, we approach test development with the practical, dispassionate countenance of the clinical laboratorian. For our clinical service (as distinct from our research laboratory), we utilize a single tissue fixation protocol, optimized to detect as many clinically pertinent neural IgGs as possible. Testing several hundred specimens each day on multiple differentially-fixed tissue slices would be neither logistically feasible nor cost-effective. Our tissue-based testing is complemented by screening for certain plasma membrane protein-targeted IgGs—such as NMDA-R, aquaporin-4, Lgi1, CASPR2 and MOG—by rigorously validated cell-based methods, read out either by microscopy or by flow cytometry.³ Other analytes again—such as GFAP-IgG—must meet strict immunohistochemical criteria and be confirmed by antigen-specific cell based assay before a positive result can be issued.⁴

As a U.S.-based clinical laboratory, both the College of American Pathologists and New York State Department of Health ensure—on behalf of the Center for Medicare and Medicaid Services [CMS]—our continual compliance with “CLIA’88” regulations through inspections of our laboratory and evaluation of test development results. Before introduction to clinical service, our tests must demonstrate: accuracy, precision, sensitivity and specificity (both analytical and clinical), a validated reference range, and non-interference by irrelevant molecules, such as polyclonal IgGs.⁵ Our experience of evaluating diverse, unselected patients with subacute onset neurologic complaints in the Autoimmune Neurology clinic—some with autoimmune disease, others without—continues to be informative for clinical test validations and interpretative reports.

Disclosures

The authors report no relevant disclosures. Contact [email protected] for full disclosures.

References

1. Martinez-Hernandez E, Guasp M, Garcia-Serra A, et al. Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies. Neurology 2020 Epub Mar 11.
2. Fang B, McKeon A, Hinson SR, et al. Autoimmune Glial Fibrillary Acidic Protein Astrocytopathy: A Novel Meningoencephalomyelitis. JAMA Neurol 2016;73:1297–1307.
3. Fryer JP, Lennon VA, Pittock SJ, et al. AQP4 autoantibody assay performance in clinical laboratory service. Neurol Neuroimmunol Neuroinflamm 2014;1:e11.
4. Flanagan EP, Hinson SR, Lennon VA, et al. Glial fibrillary acidic protein immunoglobulin G as biomarker of autoimmune astrocytopathy: Analysis of 102 patients. Ann Neurol 2017;81:298–309.
5. Knight V, Lebo T. Validation and Quality Control: General Principles and Application to the Clinical Immunology Laboratory. In: Detrick B, Schmitz JL, Hamilton RG., eds. Manual of Molecular and Clinical Laboratory Immunology, 8th ed. Washington, DC: ASM press, 2016: 1180–1192.

Dr. McKeon and colleagues have misread our article.¹ As indicated in our study, GFAP antibodies (GFAP-abs) were only systematically tested in a series of 200 patients with anti-NMDAR encephalitis, and 10 (5%) of them were positive.¹

We appreciate the educational figure of the authors, but it does not clarify disease-specificity and the data provided by McKeon et al. shows otherwise. Specificity means “the probability of being test negative when the disease is absent.” In the note of McKeon et al., there were 11 patients with coexisting NMDAR antibodies (NMDAR-abs) and GFAP-abs with available clinical information. Of these 11 patients, 6 had NMDAR-abs detectable by CBA and tissue staining and the other 5 only by CBA. All 6 (100%) patients with concurrent GFAP and genuine NMDAR-abs had features of anti-NMDAR encephalitis, but only 1 (16%) had accompanying encephalomyelitis. Thus, GFAP-abs had low disease-specificity. A similar problem occurred in previous reports of these authors,² where patients with concurrent NMDAR-abs and GFAP-abs were allocated to the group with autoimmune GFAP astrocytopathy (AGA)—regardless of whether patients had anti-NMDAR encephalitis—while  those with anti-NMDAR encephalitis without GFAP-abs were considered controls, introducing a bias in the assessment of disease-specificity.² For example, among 22 patients with coexisting GFAP-abs and NMDAR-abs in CSF, 20 had encephalitis and 2 meningoencephalitis or meningoencephalomyelitis (none had myelitis, encephalomyelitis, neuropathy, or meningitis), whereas among the 62 cases with only GFAP-abs in CSF, 34 had encephalitis and 28 had one of the other syndromes noted above (Fisher exact test p = 0.0036), confirming the indicated bias. If one includes the cases with concurrent aquaporin-4 (AQP4) antibodies and GFAP-abs that were handled similarly, the problem of disease-specificity becomes even worse.²     

In our experience, the CSF of patients with anti-NMDAR encephalitis always immunoreacts with brain, but in the note of Dr. McKeon et al., 5 (45%) of the 11 patients with CSF NMDAR-abs did not show tissue staining. This suggests 2 potential problems: 1) the CBA or tissue assay was suboptimal, and/or 2) the clinical significance of a positive NMDAR-CBA test (in their lab) is low.  Indeed, only 1 of these 5 patients developed anti-NMDAR encephalitis. A deviation from published methods may account for these problems.

As indicated in the title of our manuscript, our study was focused on determining whether patients with anti-NMDAR encephalitis had concurrent “glial or neuronal surface” antibodies. Therefore, the brain tissue processing was optimized for surface proteins; this does not include GFAP. Yet, due to Mayo Clinic reports indicating the presence of NMDAR-abs and other antibodies (e.g., AQP4) in many patients with GFAP-abs, we systematically re-examined 200 patients for GFAP-abs. The GFAP-ab testing was done with CBA with HEK293T cells expressing human GFAP transcript variant 1 (RG204548 pCMV6-AC-GFAP-α-GFP) examined after tissue fixation and permeabilization (not live cells; misquoted in our manuscript), and incubated with patient’s serum (diluted 1:40) and CSF (1:5), overnight at 4°C. The reactivity was visualized with a standard immunofluorescence technique that also shows co-localization with a commercial GFAP-ab. As indicated, 10 (5%) of 200 patients had GFAP-abs without any clinical or radiological evidence of AGA. 

Therefore, our findings confirm the limited-disease specificity of GFAP-abs shown by the current note and previous reports of McKeon et al.² In our experience transient meningeal enhancement without symptoms of AGA should not lead to GFAP-abs testing.  Rigorous studies are essential for the correct determination of neural antibodies.³ Equally important are the assessment of symptoms and a good understanding of disease-specificity, which in the case of GFAP-abs have not always been achieved.

References

1. Martinez-Hernandez E, Guasp M, García -Serra A, et al. Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies. Neurology 2020 Epub Mar 11.
2. Flanagan EP, Hinson SR, Lennon VA, et al. Glial fibrillary acidic protein immunoglobulin G as biomarker of autoimmune astrocytopathy: Analysis of 102 patients. Ann Neurol 2017;81:298–309.
3. Ebright MJ, Li SH, Reynolds E, et al. Unintended consequences of Mayo paraneoplastic evaluations. Neurology 2018;91:e2057–e2066.

We appreciate Dr. Lancaster’s comments on our article.¹ Indeed, the co-occurrence of MOG- or aquaporin-4- (AQP4) antibodies with anti-NMDAR encephalitis was associated with symptoms of this encephalitis, accompanied by a history or MRI features of a demyelinating disease. However, the co-occurrence of anti-NMDAR encephalitis with other neuronal antibodies was associated with symptoms of this encephalitis and clinical or MRI features of the associated antibody—e.g., medial temporal lobe MRI changes with AMPAR-antibodies or cortical-subcortical FLAIR MRI abnormalities with GABAaR-antibodies. Patients with an additional neuronal surface antibody had a worse prognosis than those with isolated anti-NMDAR encephalitis or with a concurrent glial antibody. Our experience with GFAP-antibodies suggests that the syndrome-specificity of these antibodies is less clear. In our study, GFAP-antibodies did not confer any additional clinical or radiological features to anti-NMDAR encephalitis. Moreover, the meningoencephalomyelitis with linear perivascular enhancement that has been considered characteristic of GFAP-antibodies can also occur without these antibodies.² If a patient with encephalitis and cerebrospinal fluid NMDAR-antibodies develops a syndrome atypical for anti-NMDAR encephalitis, one should be concerned: (1) that a potential laboratory error (e.g., the NMDAR-antibody test was falsely positive), or (2) that the syndrome represents a manifestation of two disorders with concurrent autoantibodies.

Disclosure

The authors report no relevant disclosures. Contact [email protected] for full disclosures.

References

1. Martinez-Hernandez E, Guasp M, Garcia-Serra A, et al. Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies. Neurology 2020 Epub Mar 11.
2. Wickel J, Chung HY, Kirchhof K, et al. Encephalitis with radial perivascular emphasis: Not necessarily associated with GFAP antibodies. Neurol Neuroimmunol Neuroinflamm 2020;7:e670.

Martinez-Hernandez et al.¹ systematically explored the co-occurrence of anti-NMDAR with other cell surface neuronal antibodies or glial antibodies (MOG, AQP4). Anti-NMDAR encephalitis is an autoimmune brain disease characterized by psychosis, memory loss, dystonia, catatonia, seizures, and coma.² Ovarian teratoma triggers about half of cases but other tumors are rare. Anti-NMDAR encephalitis is usually not associated with other types of brain autoimmunity. However, a small group of patients may have co-existing neuromyelitis optica, MOG antibody syndrome, or additional CNS synaptic antibodies.^3,4

In the current study, 7.5% of patients had at least one other immune response when comprehensively tested. MOG and AQP4 antibodies were strongly linked to CNS demyelination. Other CNS synaptic antibodies indicated a more severe disease course, and GABA-A receptor antibodies were strongly associated with the characteristic brain MRI abnormalities of that disorder.⁴ The significance of GFAP antibodies was less clear. Patients with preceding CNS viral infection were particularly predisposed to multiple immune responses.⁵

In summary, clinicians should be especially vigilant for other types of autoimmunity in anti-NMDAR encephalitis patients, especially those with atypical radiological features, signs of CNS demyelination (optic neuritis, myelitis), preceding CNS viral infection, or atypical tumors. 

Disclosure

The author reports no relevant disclosures. Contact [email protected] for full disclosures.

References

1. Martinez-Hernandez E, Guasp M, Garcia-Serra A, et al. Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies. Neurology 2020 Epub Mar 11.
2. 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.
3. Titulaer MJ, Hoftberger R, Iizuka T, et al. Overlapping demyelinating syndromes and anti—N-methyl-D-aspartate receptor encephalitis. Ann Neurol 2014;75:411–428. 
4. Spatola M, Petit-Pedrol M, Simabukuro MM, et al. Investigations in GABAA receptor antibody-associated encephalitis. Neurology 2017;88:1012–1020.
5. Armangue T, Leypoldt F, Málaga I, et al. Herpes simplex virus encephalitis is a trigger of brain autoimmunity. Ann Neurol. 2014;75:317–23.

We welcome the study by Martinez-Hernandez et al.¹ reporting antibody coexistence in anti-n-methyl-D-aspartate receptor (NMDA-R) encephalitis.

The authors confirmed coexistence of aquaporin-4-IgG or myelin oligodendrocyte glycoprotein (MOG)-IgG as predictors of co-occurrence of anti-NMDA-R encephalitis and autoimmune CNS demyelination. In contrast, it was concluded that coexisting GFAP-IgG is “clinicoradiologically non-specific.”¹

The diagnosis of autoimmune GFAP astrocytopathy—AGA, a clinical-radiological-serological entity, characterized by an inflammatory steroid-responsive meningoencephalomyelitis—requires adherence to a 2-step CSF evaluation^{2, 4-6} (Figure): 1) detection–by optimized rodent tissue-based immunohistochemistry–of an astrocytic filamentous-appearing pattern of IgG staining of all of midbrain (meningeal, subpial, and periventricular), cerebellar white—but not grey—matter, hippocampal, and cortical brain regions, and myenteric plexus; 2) positivity by GFAPα isoform-transfected HEK293 CBA (Figure). Serum testing is insensitive and generates non-specific results, as occasionally does CSF if both tissue and cell-based assays steps are not followed.

Two patients with coexisting GFAP-IgG reported by Martinez-Hernandez et al. had meningeal enhancement on MRI, but the authors attribute this to anti-NMDA-R encephalitis alone, though the question of these 2 IgGs coexisting was not addressed systematically.¹ Deviations from published methods may explain the paucity of cases found by Martinez-Hernandez and colleagues—10 of 846 anti-NMDA-R encephalitis patients total, 1.2%.²  Since test “go-live” in June 2019, we have detected GFAP-IgG coexisting in 6% of NMDA-R-IgG positive CSF specimens.

Martinez-Hernandez et al. reported universal non-detection of coexisting GFAP-IgG by immunohistochemistry.² The authors did not state whether fixation and permeabilization were optimized for detection of IgGs reactive with cytoplasmic proteins, such as GFAP. Furthermore, the authors utilized a “live” GFAP-specific cell-based assay (CBA) to identify 10 patients with coexisting GFAP-IgG, yet the intracellular location of GFAP should preclude detection in live cells. Additional details pertaining to the GFAP isoform expressed, and the specimen types tested—CSF or serum—were lacking. All of these factors can impact the sensitivity and specificity of GFAP-IgG testing and frequency of coexistence with NMDA-R-IgG.^2,3

To address the question of neurologic phenotype further, we reviewed data for 248 unique patients in whom CSF GFAP-IgG was detected and confirmed by our 2-step laboratory protocol^2,3 over the course of 4.5 years (January 1^st 2015–June 30^th 2019). Median GFAP-IgG end-point titer was 1:64 (range 4-512; normal, <2). Twenty-three (9%) had coexisting CSF NMDA-R-IgG detected, 11 with clinical information available. Six of 11 patients also had NMDA-R-IgG detected at high titer on tissue-based assay (median, 1:32 [range, 4-128; normal, <2], Figure), confirmed by GluN1-specific CBA. Of those 6, 5 had typical anti-NMDA-R encephalitis clinical phenotype, 4 of whom had anti-NMDA-R encephalitis-typical (normal) MRI brain imaging, and 1had radial periventricular enhancement (AGA-typical). The 6^th patient had an inflammatory encephalomyelitis clinically and radiologically (AGA-typical) but accompanied by orofacial dyskinesias (anti-NMDA-R encephalitis-typical). The remaining 5 patients had NMDA-R-IgG detected (robustly) by GluN1-specific CBA only. Four of 5 patients were AGA-typical radiologically and clinically—with headache and meningism prominent—and 1 was anti-NMDA-R encephalitis-typical. 

Although the pathophysiology of AGA remains to be elucidated, GFAP-IgG is a biomarker of a treatable inflammatory CNS disorder, when appropriately evaluated for. Rigorous CSF evaluation permits clear AGA diagnosis, identification of clinically pertinent coexisting IgGs, avoidance of unnecessary brain biopsies, and facilitation of early treatment.

Disclosure

The authors report no relevant disclosures. Contact [email protected] for full disclosures.

References

1.Martinez-Hernandez E, Guasp M, Garcia-Serra A, et al. Clinical significance of anti-NMDAR concurrent with glial or neuronal surface antibodies. Neurology 2020 Epub Mar 11.

2.Flanagan EP, Hinson SR, Lennon VA, et al. Glial fibrillary acidic protein immunoglobulin G as biomarker of autoimmune astrocytopathy: Analysis of 102 patients. Ann Neurol 2017;81:298–309.

3.Dubey D, Hinson SR, Jolliffe EA, et al. Autoimmune GFAP astrocytopathy: Prospective evaluation of 90 patients in 1 year. J Neuroimmunol 2018;321:157-163.

4.Fang B, McKeon A, Hinson SR, et al. Autoimmune Glial Fibrillary Acidic Protein Astrocytopathy: A Novel Meningoencephalomyelitis. JAMA Neurol 2016;73:1297–1307.

5.Iorio R, Damato V, Evoli A, et al. Clinical and immunological characteristics of the spectrum of GFAP autoimmunity: a case series of 22 patients. J Neurol Neurosurg Psychiatry 2018;89:138–146.

6.Long Y, Liang J, Xu H, et al. Autoimmune glial fibrillary acidic protein astrocytopathy in Chinese patients: a retrospective study. Eur J Neurol 2018;25:477–483.

Figure: Analysis of Patient CSF by Indirect Immunofluorescence Assays

(A, B) Fixed and permeabilized tissue and (C, D) HEK293 cells (GFAPα-transfected), visualized at 20x. Typical filamentous GFAP-IgG staining, identified in the CSF of patients A and B, is evident in the pial and subpial regions, and parenchyma of midbrain (Mb), and hippocampus (Hi), but spares the cerebellar gray matter (Cb). Patient A had no coexisting IgGs detected. Patient B had NMDA-R-IgG coexisting, producing staining of synaptic regions of the cerebellar granular layer and hippocampus, confirmed by GluN1-specific cell-based assay (not shown). C and D, GFP-tagged GFAPα cell expression is shown in green (C.a and D.a). CSF from a patient harbors GFAP-IgG (red, C.b) that co-localizes with GFAP protein (yellow, C.c). D, CSF from a non-autoimmune control subject, produces no staining of cells (D.b), nor co-localization, (D.c).