International multicenter examination of MOG antibody assays

Objective To compare the reproducibility of 11 antibody assays for immunoglobulin (Ig) G and IgM myelin oligodendrocyte glycoprotein antibodies (MOG-IgG and MOG-IgM) from 5 international centers. Methods The following samples were analyzed: MOG-IgG clearly positive sera (n = 39), MOG-IgG low positive sera (n = 39), borderline negative sera (n = 13), clearly negative sera (n = 40), and healthy blood donors (n = 30). As technical controls, 18 replicates (9 MOG-IgG positive and 9 negative) were included. All samples and controls were recoded, aliquoted, and distributed to the 5 testing centers, which performed the following antibody assays: 5 live and 1 fixed immunofluorescence cell-based assays (CBA-IF, 5 MOG-IgG, and 1 MOG-IgM), 3 live flow cytometry cell-based assays (CBA-FACS, all MOG-IgG), and 2 ELISAs (both MOG-IgG). Results We found excellent agreement (96%) between the live CBAs for MOG-IgG for samples previously identified as clearly positive or negative from 4 different national testing centers. The agreement was lower with fixed CBA-IF (90%), and the ELISA showed no concordance with CBAs for detection of human MOG-IgG. All CBAs showed excellent interassay reproducibility. The agreement of MOG-IgG CBAs for borderline negative (77%) and particularly low positive (33%) samples was less good. Finally, most samples from healthy blood donors (97%) were negative for MOG-IgG in all CBAs. Conclusions Live MOG-IgG CBAs showed excellent agreement for high positive and negative samples at 3 international testing centers. Low positive samples were more frequently discordant than in a similar comparison of aquaporin-4 antibody assays. Further research is needed to improve international standardization for clinical care.

Immunoglobulin (Ig) G antibodies to myelin oligodendrocyte glycoprotein (MOG-IgG) are found in adults and children who present with a spectrum of CNS features that include optic neuritis, acute disseminated encephalomyelitis (ADEM), myelitis, seizures, encephalitis, brainstem, and/or cerebellar involvement. In addition, the presence of MOG-IgG can discriminate these disorders from MS. 1 Numerous studies have used different immunoassays for MOG-IgG detection, but it is now clear that native full-length human MOG as an assay substrate is crucial to make this clinical distinction. When measured using first generation assays (ELISA and Western blot), MOG-IgG are prevalent and have been identified in healthy individuals and patients with a wide variety of clinical presentations. Thus, their detection was initially considered to have little clinical utility. However, when measured by live cell-based assays (CBAs), an association between MOG-IgG antibodies and a non-MS demyelinating phenotype has been established. This understanding has driven the establishment of different variants of MOG-IgG assays with native MOG substrates in multiple centers worldwide. There are limited data on assay reproducibility between these centers. In this study, we compared the most frequently used assays for MOG-IgG detection, such as live and fixed immunofluorescence cell-based assays (CBA-IF), [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17] live flow cytometry cell-based assays (CBA-FACS), 4,[18][19][20][21][22][23][24][25][26][27] and ELISA. 28,29 Methods

Patients and controls
The clinical laboratories (Innsbruck, Mayo Clinic, Oxford, and Sydney; centers 1-4) sent the following groups of coded serum samples and clinical information to the Institute for Quality Assurance (IfQ; Lübeck, Germany):

Statistical analysis
Upon completion of the testing, the assay results from each center were entered onto a web-based database. The data were then unblinded and analyzed. Statistical analyses were performed using IBM SPSS software (release 24.0; IBM, Armonk, NY) or GraphPad Prism 8 (GraphPad, San Diego, CA). Correlation of parameters was analyzed with Spearman nonparametric correlation. Kappa statistics were used to assess the concordance between assays. All graphs were created using GraphPad Prism.

Data availability
The data set used and analyzed during the current study is included in the main text and the supplementary files.

Results
CBAs  MOG-IgM antibodies at a titer ≥ 1:160 were a rare finding in samples sent as clear positive (5/39, 13%). One of these 5      samples had a high MOG-IgM titer (1:5,120, figure 2C, assay Nr. 9, large gray dot) and was low positive for MOG-IgG in 4 assays (using IgG (H + L) and IgG1 secondary antibodies), but negative in 4 assays (using IgG(Fc) and IgG1 secondary antibodies). The other 4 samples were positive for MOG-IgG in all CBAs. MOG-IgMs at a titer ≥ 1:160 were absent in all 40 samples sent as negative.
Overall  MOG-associated clinical phenotypes mentioned above, but also from a healthy blood donor and 4 patients with MS. Finally, the 82 samples negative in all live CBAs were from 24 patients with typical MOG-associated clinical phenotypes, MS (9), other neurologic diseases (10), and healthy controls (39).

Discussion
In this study, we compared the reproducibility among the most frequently used assays for serum MOG-IgG detection, such as live and fixed CBA-IF, live CBA-FACS, and ELISA. Our data clearly indicate that strong positive and clearly negative samples are reproducible between centers where live cells expressing native full-length human MOG are used as the assay substrate. In the 4 different national testing centers using different live CBAs, there was 96% concordance for all samples tested. The agreement was less good when a fixed CBA-IF (90%) tested in-house by the company (center 5) was included, which is consistent with recent studies demonstrating that some conformational epitopes of MOG are lost upon fixation of MOG-expressing cells. 4,16,26 Importantly, most of these discordant negative results on the fixed MOG-IgG assay had high MOG-IgG titers in live CBAs and were from typical MOG-IgG-associated demyelinating syndromes. There is utility in the commercial fixed MOG-IgG testing in places where live MOG-IgG CBAs are unavailable, but this assay will miss 10-15% of positive cases. A recent study highlighted an issue with specificity in commercial MOG-IgG testing, particularly in samples that were only positive at low dilutions. 4 Therefore, clinicians should consider retesting unexpected MOG-IgG results at centers offering live CBAs.
Finally, ELISAs did not distinguish between the positive and negative patient samples and showed no concordance with CBAs for detection of human MOG-IgG conclusively demonstrating that ELISAs are not suitable for the detection of human MOG-IgG. Although this has been shown in several studies (summarized in ref 28), some laboratories still use this method for MOG-IgG detection. We hope that our findings inform neurologists that only CBAs should be used for the measurement of human MOG-IgG. Moreover, and in agreement with previous studies, 4,16,26 live CBAs remain the gold standard for the detection of human MOG-IgG.
The agreement of MOG-IgG CBAs for low positive sample was less good (33% concordance), and MOG-IgG assays were particularly discordant at the borderline of positivity. This raises the pertinent question where to draw cutoff values and how they influence the clinical interpretation of diagnostic results. If we examine the clinical phenotype of people with high MOG-IgG levels, which are consistently detected by all CBAs, we identify patients with non-MS demyelinating phenotypes (such as ADEM, NMOSD, optic neuritis, myelitis, and other demyelinating diseases). 1 In contrast, the low positive samples ,which showed a lack of reproducibility between centers, had a wider range of clinical phenotypes that mostly include the same phenotypes (ADEM, NMOSD, optic neuritis, myelitis, and other demyelinating diseases), but also a proportion of every control group (clinically definite MS, other neurologic diseases, and healthy individuals), making their interpretation difficult. It is unlikely that lower levels of pathogenic antibody cause a wider disease presentation, suggesting that some of these phenotypes are not driven by MOG-IgG. Hence, an argument can be made that the presence of low positive MOG-IgG is only meaningful in the correct clinical context such as in patients with ON, myelitis, ADEM, or encephalitis but not in the context of other diseases, particularly MS. 1,19,31 This is a circular but reasonable interpretation of low positive results, but with caveats. There will be a false-positive rate even within the correct clinical context that should be considered, and an estimate of this would be useful for any test. Second, clinical criteria are not perfect. There are individuals who fulfill criteria for MS, but are often atypical; perhaps the MOG-IgG result has utility in this context in ruling out MS and should not be ignored out of hand. Importantly, when extrapolating from experiences on the treatment of NMOSD and a recent larger study on treatment of patients with MOG-IgG from France, disease-modifying treatments for MS may not work in MOG-IgG-positive patients and may even exacerbate disease. 1,[31][32][33] The third interpretation is that these low positive results that are not reproducible between centers are not useful clinically and in fact dilute the utility of a more specific test. Finally, as a general consideration in samples not taken at disease onset, other confounding factors such as preceding steroid use or other immunosuppressive treatments and remission could lower a positive MOG-IgG result. It is important to note that MOG-IgG levels are often non-normally distributed in large patient cohorts, and a skewing toward these lower MOG-IgG titers has been observed in many studies. 1 Samples scored as low positive for MOG antibodies are much less concordant than the clearly positive samples across the 7 live MOG assays (28% vs 92%). Importantly, the MOG-IgG low positives are also less concordant than low positive samples in similar assays for other autoantigens such as AQP4.
In 2016, we published a European multicenter validation experiment comparing 21 AQP4-IgG assays. 34 In this study 5 live CBAs (3 CBA-IF and 2 live CBA-FACS) had sensitivities, specificities, and accuracies greater than 90% similar to the 7 live MOG-IgG CBAs in our current study. These 5 AQP4-IgG assays were compared on 66 AQP4-Ab-positive samples, thereof 52 were high positive (median semiquantitative score in the live CBAs 2.5-4) and 14 low positive (median semiquantitative score in the live CBAs 1-2). 34  Further work is now needed to better define the most useful clinical cutoff and to establish whether there is any added benefit in identifying patients with low positive MOG-IgG. We propose that this should be done in a collaborative effort. We need to better characterize false-positive cases, such as classical MS cases, other neurologic diseases, and healthy individuals, and get more information on the clinical sensitivity and specificity of all assays by using appropriate controls, such as systemic autoimmune diseases, noninflammatory neurologic controls, and healthy controls. It is of great interest to establish how these antibodies relate to different clinical phenotypes and whether they are a mixture of pathogenic and bystander antibodies that all bind MOG in vitro.
To conclude, we have shown that currently used live CBAs to measure MOG-Abs showed excellent agreement for clearly positive and negative samples, but low positive samples were more discordant. Further work is now required to standardize the clinically most useful assay.