Characterization of the human myelin oligodendrocyte glycoprotein antibody response in demyelination

The objective of this study was to characterize the epitope, affinity, and sensitivity to conformational changes of human MOG Ab response in demyelination. Locally and internationally recruited patient sera were screened for antibodies targeting native MOG between 2011 and early 2018. We have identified 287 MOG Ab-positive (MOG Ab+) children (n = 139, < 18 years at disease onset) and adults (n = 148, > 18 years at disease onset). Among the patients for whom we had additional clinical information, disease duration (from onset to first serum tested/baseline sample) was a median 0 year, interquartile range (IQR) 0–0 years and mean 0.47 ± 1.24 years in children (n = 122), and median 0 year, IQR 0–0.1 years and mean 1.46 ± 3.5 years in adults (n = 108). Among these samples, 83% of paediatric and 69% of adult samples were collected at disease onset. One hundred thirty serial sera from 51 MOG Ab+ patients (19 children, 32 adults) and 22 CSF (10 children, 9 adults) were also collected. Clinical phenotypes were retrospectively collected between July 2018 and December 2018, and patients were reported as relapsing if relapses had occurred over the study time frame with a minimum disease duration of 6 months (Table 1). Experimental data in this study was obtained when MOG Ab+ patient sera were combined and retested as one cohort. A flow cytometry live cell-based assay (live flow assay) was used to detect presence of patient serum Ab against conformational native-MOG [1, 13, 49]. The epitope of MOG Ab was assessed using a P42S mutant, consisting of full-length human MOG with the proline at position 42 substituted for serine. Conformational changes in native-MOG were modelled using paraformaldehyde which is known to alter protein structure [37, 59], and binding to fixed MOG was assessed using flow fixed assay and a commercial fixed biochip assay, with samples blinded and independently performed by an accredited external pathology laboratory. Binding to the immobilized extracellular Ig-like domain of MOG (MOG^1–117) enabled detection of high affinity Ab by ELISA. In all flow cytometry and ELISA experiments, samples were reported positive if they were above 99th percentile of the control range in at least two of three quality-controlled experiments.

Patient clinical phenotypes were delineated based on discussion with their treating neurologists and are accurate characterizations of patients’ presentations (Table 1). Detailed clinical and radiological phenotyping was outside of the scope of this study, and we have not conferred neuromyelitis optica spectrum disorder (NMOSD) 2015 diagnosis on these patients [72]. MOG Ab-seropositive patients did not fulfil 2017 revised McDonald criteria [62]. Those classified as ADEM fulfilled Krupp et al. criteria [30]. Seropositivity of 23 children and 12 adults have been previously reported in [47]. Age-matched controls (24 children, 24 adults) were selected on the basis of clinical disease (general medical and non-inflammatory neurological disorders), and not on their negative serostatus. Previously archived MOG Ab seronegative (MOG Ab-) patients (24 children, 24 adults) were also tested. They included patients with monophasic and relapsing disorders not typically associated with MOG Ab, such as multiple sclerosis (MS, fulfilling 2017 revised McDonald criteria), and clinically isolated syndrome (CIS) other than optic neuritis (ON), and longitudinally extensive transverse myelitis (LETM) [23, 26, 49, 52]. All control sera and MOG Ab- sera remained below the positivity threshold, whilst MOG Ab+ patients laid consistently above threshold and had low intra-assay variability (Additional file 1: Table S1). MOG Ab titers were measured in 51 MOG Ab+ patients (19 children, 32 adults) from whom 130 serial samples were tested (54 paediatric samples, median follow-up 17.6 months, IQR 3.6–28.9; 76 adult samples, median follow-up 4.7 months, IQR 1.7–13.) (Additional file 1: Table S2). Intrathecal MOG Ab were detected in undiluted cerebrospinal fluid (CSF) as previously described [13]. We obtained 22 CSF from seropositive-MOG Ab+ patients (12 samples from 10 children, 10 samples from 9 adults). CSF controls (24 non-demyelination controls; 14 demyelination controls) were tested in parallel. Intrathecal production of MOG Ab was assessed by determining the MOG Ab index: MOG Ab index = \( \frac{\mathrm{QMOG}\ \left(\mathrm{CSF}\ \mathrm{MOG}\ \mathrm{Ab}\ \Delta \mathrm{MFI}/\mathrm{Serum}\ \mathrm{MOG}\ \mathrm{Ab}\ \Delta \mathrm{MFI}\right)}{\mathrm{QIgG}\ \left(\mathrm{Total}\ \mathrm{CSF}\ \mathrm{IgG}/\mathrm{Total}\ \mathrm{serum}\ \mathrm{IgG}\right)} \). A MOG Ab index greater than four indicated intrathecal synthesis of as previously described [25, 50, 51]. Total IgG in patient CSF and serum was determined using a human IgG ELISA kit following manufacturer’s instructions (Immunology Consultants Laboratory, Inc.).

Serum IgG binding to full-length native human MOG α1 isoform (native-MOG), a native MOG mutant P42S (native-P42S), and an empty vector control (native-CTL) were assessed. Transduced stable human embryonic kidney (HEK293) cells lines were generated as previously described [13]. Site-directed PCR was used to mutate the Proline42 (P42) for serine of the extracellular monomer MOG domain (1–117aa) (native-MOG^1–117), generating native-P42S^1–117, which was used to confirm the secondary structure of native-P42S. Native-MOG^1–117 and native-P42S^1–117 was subcloned into the pESG-IBA144 Stargate Accepter vector (IBA Lifesciences). HEK293T cells were transfected with Lipofectamine 3000 (ThermoFisher Scientific), secreted monomers were then purified using HisTrap and StrepTrap HP columns (GE Healthcare Lifesciences) following manufacturer’s instructions. Following buffer exchange, protein yield was quantified with Pierce BCA protein assay kit (ThermoFisher Scientific). The secondary structure of native-MOG^1–117 and native-P42S^1–117 were determined using an Aviv 215S circular dichroism (CD) spectrometer (Aviv Biomedical Inc.). Far-UV circular dichroism (CD) spectra were measured over 200-260 nm with a cell path length of 0.1 cm at 0.26–0.43 mg/mL in PBS at 25 °C. The mean of four spectra were obtained and normalized to mean residue weight ellipticity units (θ_MRW).

Flow cytometry cell-based assays were used to detect presence of patient serum Ab (1:50) against conformational native-MOG, native-P42S, and formaldehyde-fixed MOG (fixed-MOG). For the fixed flow assay, HEK293 cells were incubated with fresh 4% formaldehyde for 10 min at RT and washed twice prior to the addition of patient serum (1:50). After extensive washing of patient sera, live or fixed cells were incubated with AlexaFluor 647-conjugated anti-human IgG (H + L) (1:100, A21445) or anti-human IgM (1:100, A21249, ThermoFisher Scientific). Detection of native-MOG IgG1 Ab by live flow assay was similarly performed by which an unconjugated mouse anti-human IgG1 (1:100, MH1013, ThermoFisher Scientific) was added, then washed, and followed by AlexaFluor 647-conjugated anti-mouse IgG (H + L) (1:100, A31571, ThermoFisher Scientific).

Native-MOG and fixed-MOG Ab binding levels were obtained from the delta median fluorescence intensity (ΔMFI): ΔMFI = MOG MFI – CTL MFI. The positive threshold was set at three standard deviations (3SD) above the mean of age-matched controls. Samples were reported positive if they were above the threshold in at least two of three quality-controlled experiments. Patient MOG Ab binding was also assessed using fixed-MOG, P42S (fixed-P42S), and CTL (fixed-CTL) cell lines. The intra-assay variability was higher in the fixed flow assay than the live flow assay (Additional file 1: Table S1). P42S expression was controlled in each experiment using the chimeric 8-18C5 mAb (monoclonal anti-rat MOG with murine variable region and human IgG1 constant region, secondary antibody: AlexaFluor 647-conjugated anti-human IgG (H + L) at 1:100 (A21445 from ThermoFisher Scientific) as described for other mutant antigens [56]. Levels of MOG Ab binding to P42 (either native-P42 or fixed-P42) in patients were determined by the formula: P42 Ab = MOG ΔMFI - P42S ΔMFI. Control sera were similarly analysed and used to establish a control reference range by calculating the 3SD above and below the control P42 Ab mean. Patient MOG Ab were assigned as Proline42 binders (above mean + 3SD), other epitope (between mean-3SD and mean + 3SD), or Serine42 binders (below mean-3SD). Reported samples remained in the same category in at least two of three independent experiments. All flow cytometry data was acquired using the high-throughput system on the LSRII flow cytometer (BD Biosciences). Data was analysed using FlowJo v10 (TreeStar) software and Microsoft Excel. All imaging data was acquired on a TCS SP5 confocal microscope (Leica) with a 63X oil numerical aperture 1.4 objective lens (Westmead Imaging Facility). Data were analysed using ImageJ v1.46 software (NIH).

Controls (10 children, 11 adults), native-MOG Ab- (12 children, 12 adults), and native-MOG Ab+ patients (65 children, 59 adults) were blinded and independently tested using the formaldehyde-fixed indirect immunofluorescence BioChip™ assay (fixed biochip assay) as per manufacturer’s instructions (Euroimmun). Similar numbers of low, mid, and high native-MOG Ab+ patients were randomly selected. The fixed biochip assay was independently performed by an accredited external pathology laboratory at NSW Health Pathology-ICPMR, which is the referral centre for neuronal antibody testing in NSW (average 8000 samples tested per year by Euroimmun assays). All samples were blinded and scored by two independent experienced investigators. Sera were reported negative or scored 1 to 4 if positive. Equivocal patients were re-blinded and independently retested. To avoid bias between assays, and in absence of diagnostic criteria for MOG Ab-associated disorders, sensitivity and specificity were determined using two groups of patients: monophasic and relapsing disorders with reported MOG Ab-association (such as ADEM, ON, BON, LETM) [23, 26, 46, 48, 49, 52, 70], and monophasic and relapsing disorders with not yet reported MOG Ab-association and disorders not associated with MOG Ab (MS, CIS other than ON, and general medical and non-inflammatory neurological disorders) [23, 26, 49, 52, 70].

ELISAs were conducted to detect high affinity MOG Ab binding to the extracellular MOG domain. Purified native-MOG^1–117 and native-P42S^1–117 (10 μg/mL) was coated onto 96-well Nunc MaxiSorp™ plates (ThermoFisher Scientific) in coating buffer (0.1 M Na₂CO₃, 0.1 M NaHCO₃, Milli-Q water, pH 9.6) overnight at 4 °C. Control wells were coated with bovine serum albumin (BSA) (10 μg/mL) to detect background sera binding. Wells were blocked (PBS, 1% BSA) for 3 h at RT, patient sera (1:50) was incubated for 2 h at RT, HRP-conjugated goat anti-human IgG (1:2000) (ThermoFisher Scientific) was added for 1 h at RT, followed by 3,3′,5,5′-tetramethylbenzidine (Sigma-Aldrich, USA) for 15 min at RT, and 1 M HCl was added to stop the reaction. Wells were washed thoroughly with PBS, 0.05% Tween20 following each incubation. Optical density (OD) values were obtained at 450 nm and were wavelength- and blank-corrected. MOG Ab binding to native-MOG^1–117 was determined by the formula: native-MOG^1–117 ΔOD = native-MOG^1–117 OD – BSA OD. The positive threshold was set at 3SD above the control mean ∆OD . Native-MOG^1–117 Ab titers were also independently assessed in 15 serum samples by ELISA, as described above.

Graphs were generated using Prism v7.0a (GraphPad Software) and Adobe Illustrator CC 2015 (Adobe Systems). Statistics were only performed on phenotype groups with eight or more patients (Table 1). The Kruskal-Wallis test was used for continuous data comparisons and a chi-square (χ²) test for categorical comparisons. Correlation analyses and R² values were generated using a linear regression model. Boxplots, bar graphs, age, and disease duration are expressed as median and IQR. One representative out of three independent flow cytometry and ELISA experiments is shown.

Human research ethics approvals (NEAF 12/SCHN/395) were granted by the individual ethics committees for the participating hospitals. Informed consent was obtained from patients or carers in the case of paediatric patients.

IgG antibodies targeting native MOG (native-MOG Ab) were detected in 139 children and 148 adult patients using a live flow assay (Fig. 1a) with low intra-assay variability (Additional file 1: Table S1). Native-MOG Ab-seropositive (native-MOG Ab+) patients exhibited a broad range of median fluorescence intensity (ΔMFI) values and were categorized into three levels of native-MOG Ab, with no significant differences between children and adults (Fig. 1a). Most native-MOG Ab (131/134 children, 98%; 134/147 adults, 91%) were of the IgG1 isotype (Fig. 1a), whereas only a minority was IgM and IgG double-positive (3/133 children, 2%; 10/148 adults, 7%) (Additional file 1: Figure S1A). There was no correlation between IgG and IgM values in all patients, suggesting no cross-reactivity (Additional file 1: Figure S1B). Sensitivity and specificity were high and similar between assays detecting total IgG and IgG1 (Table 2). ΔMFI values were strongly associated with native-MOG Ab titers in children and adults (Fig. 1b and Additional file 1: Figure S2). There was a high concordance of native-MOG Ab-positivity between matched serum and CSF samples, and native-MOG Ab were higher in serum than CSF after normalization to total IgG and serum dilution (8/10 children, 7/9 adults) (Fig. 1c). Four patients (2 children and 2 adults) had increased CSF titers compared to serum, however the two adults had intrathecal synthesis of MOG Ab as their MOG Ab index between intrathecal and extrathecal compartments exceeded four [25, 50] (Additional file 1: Table S3). Among native-MOG Ab+ children, ADEM was the most prevalent phenotype (38%, 53/139), and most exhibited a monophasic course (79%, 42/53) (Table 1, Fig. 1d), whereas in adults, optic neuritis (ON) was most common (61%, 91/148), and most presented with relapsing unilateral ON (UON) (27%, 25/91). Across all phenotypes, the disease course of children was predominantly monophasic (57%, 79/139), whereas a relapsing course was slightly more frequent in adults (41%, 61/148) (Table 1, Fig. 1d). Overall, ON was the major clinical phenotype across the cohort, and mainly occurred in isolation (45%, 129/287), but ON in combination with other demyelinating syndromes (including myelitis and ADEM) was also observed in a small group (9%, 26/287) (Table 1, Fig. 1d). Native-MOG Ab presented in patients between 1 to 80 years of age, predominantly between ages 3–7 in children, and 29–33 and 53–57 in adults (Table 1, Additional file 1: Figure S3A), with no correlation between native-MOG Ab titer and age. ON presentation in isolation were distributed across all ages, whereas ADEM and ADEM/ON was only observed in children, and ON/TM was only observed in adults (Table 1, Fig. 1d and Additional file 1: Figure S3B). Despite a slight female preponderance, native-MOG Ab titers did not differ across sex or paediatric phenotypes (Table 1). Among adults with ON, native-MOG Ab titers were higher in patients with bilateral ON (BON) compared to UON, independently of a monophasic or relapsing course (P = 0.01) (Fig. 1e), and despite disease duration, which was not statistically different between adults with UON (median 0 year, IQR 0–0.25 years; mean 1.32 ± 3.16 years, n = 35) and BON (median 0 year, IQR 0–0.88 years; mean 1.19 ± 2.95 years, n = 35) (P = 0.97). Furthermore, when patients with serum collected at disease onset were independently analyzed (monophasic and relapsing UON, n = 27; monophasic and relapsing BON n = 21), similar results were observed; BON patients exhibited higher MOG Ab titers than UON patients (Fig. 1e) (P = 0.040), suggesting that these results were not influenced by the timing of sample collection.

Antigen conformation is known to influence Ab recognition, however, the effect of conformational changes to MOG on Ab binding remains elusive. Therefore, the sensitivity of human MOG Ab binding to fixed-MOG was assessed in formaldehyde-fixed flow cytometry and commercial biochip assays. Formaldehyde modifies proteins on various structural levels [37, 59], and with five primary amine groups within its extracellular domain, amine cross-linkage may not disrupt MOG secondary structure, but will change the native conformation of MOG. Expression of fixed-MOG was high and comparable to native-MOG (Fig. 2a and b), therefore fixed-MOG was available for binding by patient native-MOG Ab. Only 56% (78/139) of children and 53% (79/148) of adult native-MOG Ab+ sera were able to bind to fixed-MOG (Fig. 2c). The remaining native-MOG Ab+ patients (61/139 children, 44%; 69/148 adults, 47%), all control sera (48/48), and native-MOG Ab- sera (48/48) did not recognize fixed-MOG (Fig. 2c). 59% of patients who failed to recognize fixed-MOG had low native-MOG Ab titers (36/61 children; 41/69 adults) (Fig. 2d), whereas 41% of patients (25/61 children, 41%; 28/69 adults, 41%) did not recognize fixed-MOG despite having mid to high native-MOG Ab titers (Fig. 2d). However, 22% of children (10/46) and 16% of adults (8/49) bound fixed-MOG even with low native-MOG Ab titers (Fig. 2d). Native-MOG Ab titers correlated poorly with fixed-MOG Ab titers (Additional file 1: Figure S4A). These data suggest that failure of native-MOG Ab to bind fixed-MOG was influenced by two key features, which are not mutually exclusive: a low native-MOG Ab titer, and a MOG Ab response with sensitivity to native conformational MOG. Adults with monophasic and relapsing UON exhibited weaker binding to fixed-MOG than BON (P = 0.01, Fig. 2e) and were more likely to be seronegative in the fixed flow assay (P = 0.03, Fig. 2f) (Additional file 1: Table S4). Interestingly, adult females had weaker binding to fixed-MOG and were more likely to become seronegative as assessed by the fixed flow assay than adult males (P = 0.03).

We then examined whether this loss of binding to conformationally altered MOG was echoed in a formaldehyde-fixed biochip assay in a smaller cohort externally tested (65 children, 59 adults). Parallel to the fixed flow assay, a proportion of native-MOG Ab+ patients did not recognize fixed-MOG in the fixed biochip assay (29%, 19/65 children; 41%, 24/59 adults), and all controls (10 children, 11 adults) and all native-MOG Ab- patients (12 children, 12 adults) were negative (Additional file 1: Figure S4B). As in the fixed flow assay, binding to fixed-MOG was also dependent on native-MOG Ab category, as 58% of paediatric and 54% of adult patients who were negative had low native-MOG Ab titers (11/19 children; 13/24 adults) (Fig. 2g). Conformational sensitivity was observed in fixed-MOG Ab- patients with mid to high native-MOG Ab titers (42%, 8/19 children; 46%, 11/24 adults), whereas conformational insensitivity was observed in fixed-MOG Ab+ patients despite low native-MOG Ab titers (34%, 13/24 children; 35%, 7/20 adults) (Fig. 2g).

Between the fixed flow and biochip assays, 65% of children (30/46) and 71% of adults (25/35) were positive in both assays, with less overall seropositivity detected in the fixed flow assay. The score of fixed-MOG Ab, a qualitative measure of fluorescence intensity, in the fixed biochip assay correlated poorly to fixed-MOG Ab titers (Additional file 1: Figure S4C), implying a difference in the fixation effect on MOG between the fixed flow and biochip assays. Across the three assays, 30/66 (45%) children and 25/59 (42%) adults were positive. Overall, although the specificity of detection remained high between all assays, albeit slightly increased in the fixed flow and biochip assays, the sensitivity of the live flow assay was far greater than both fixed MOG assays (Table 2). Importantly, phenotypes which have recently been described as classical for MOG Ab-associated demyelination, such as children presenting with monophasic ADEM, UON, and BON, and adults presenting with relapsing UON and BON, were reported seronegative by fixed assays (Fig. 2f and h, Additional file 1: Table S4). A number of patients in whom detailed clinical information was available, and who tested seronegative by the fixed flow and biochip fixed assays, presented with typical MOG Ab-associated phenotypes and were therefore considered bona fide MOG Ab-seropositive patients by their clinicians rather than false positives (Additional file 1: Supplementary material 1). Furthermore, 19 children and 28 adults, in whom therapeutic data was available, were highly sensitive and responsive to immunotherapy, particularly steroids (Additional file 1: Table S4), once again highlighting that these patients behave in a manner considered typical for MOG Ab-associated demyelination.

To determine whether Proline42 was included in the major antigenic region as reported in children [34], MOG Ab binding was compared between P42S MOG (native-P42S) and native-MOG in our large paediatric and adult cohort. The secondary protein structure of native extracellular P42S MOG (native-P42S^1–117) was not altered by the point mutation, with the circular dichroism (CD) spectrum characteristic of β-sheet folding and similar to extracellular MOG (native-MOG^1–117) (Fig. 3a). Furthermore, surface expression of native-P42S was high and comparable to native-MOG (Figs. 2e and 3b and c), which enabled reliable assessment of binding between native-P42S and native-MOG. Proline42 within the extracellular CC’ loop of native-MOG was crucial to the binding of most human native-MOG Ab+ patients: 85% of children (118/139) and 76% of adults (113/148) (Fig. 3d). The epitope was not identified in a small group who bound similarly to native-P42S and native-MOG (7%, 10/139 children; 9%, 14/148 adults), and a minor group presented with strong immunoreactivity to native-P42S, and therefore bound Serine42 (8%, 11/139 children; 14%, 21/149 adults) (Fig. 3d). Proline42 was an immunodominant epitope, with more than 75% of native-MOG Ab targeting P42 (Additional file 1: Figure S5A). Native-P42 Ab titers and percentage of total MOG Ab response between paediatric and adult sera were similar (Fig. 3d and Additional file 1: Figure S5A). 67% (8/12) of paediatric and 70% (7/10) of adult CSF MOG Ab recognized native-P42. When matched CSF and serum samples were compared, 75% (6/8) of children and 86% (6/7) of adult CSF P42-binders also had serum responses toward Proline42, and the percentages of native-P42 Ab in serum and CSF were correlated (Fig. 3e), suggesting no major change in epitope across intrathecal and peripheral native-MOG Ab responses.

Across all adult patients, reactivity to native-P42 was significantly lower in relapsing phenotypes than monophasic syndromes (P = 0.015). Compared to adults with monophasic ON (UON and BON), relapsing ON patients had lower immunoreactivity to native-P42S (P = 0.014, Additional file 1: Figure S5B), and a lower percentage of native-P42 Abs (P = 0.038, Fig. 3f) with more patients recognizing other epitopes or Serine42, regardless of lesion localization, i.e. bilateral or unilateral. These results show that 34% of relapsing adults (35% of relapsing ON including ON mixed) intrinsically present with a more diverse MOG Ab response characterized by binding of additional epitopes not affected by the P42S mutation (Fig. 3f). Overall, 75% of adult patients with an epitope outside Proline42 presented with a relapsing course (Fig. 3f). No statistical differences were found across paediatric phenotypes.

To assess whether conformational change prevented binding of native-MOG Ab to its epitope, we compared patient recognition of Proline42 between native-P42 and fixed-P42. Of those who bound fixed-MOG, 69/78 children (88%) and 65/79 adults (82%) recognized native-P42. Within these P42-binders, most patients recognized the same Proline42 epitope across the two assays (77%, 53/69 children; 80%, 52/65 adults) (Fig. 3g). However, 16/69 children (23%) and 13/65 adults (20%) who bound native-P42 lost their recognition of Proline42 on fixed-MOG, suggesting that the conformational change abrogated binding to their natively-presented Proline42 epitope (Fig. 3g).

Longitudinal analysis of native-MOG Ab titers was examined in 130 native-MOG Ab+ samples from 51 patients (Additional file 1: Table S2). The majority of patients longitudinally tested had a relapsing course (84%, 16/19 children; 69%, 22/32 adults). Most children presented with monophasic and relapsing ADEM (5/19 in isolation, 5/19 in combination with other syndromes) and ON (3/19 BON, 3/19 UON, 3/19 ON/LETM), whereas most adults presented with ON (13/32 BON, 9/32 ON) (Additional file 1: Table S2). From baseline, represented by the first collected sample, most patients (89%, 17/19 children; 63%, 20/32 adults) had fluctuating native-MOG Ab titers over time, spanning up to 4.7 years in children (median disease duration 17.6 months (interquartile range (IQR) 3.6–28.9) and 9.1 years in adults (median disease duration 4.1 months (IQR 1.7–13.9) (Fig. 4a). The majority of patients, 13/19 children and 14/32 adults, had at least 30% decrease in native-MOG Ab titer from baseline (Fig. 4a), whereas a smaller proportion of 4/19 children and 6/32 adults had a greater than 30% increase of MOG Ab titers. 2/19 children and 12/32 adults showed stable MOG Ab titers over time (Fig. 4a). In this cohort, patient phenotype did not predict an increase or decrease in MOG Ab titer. Persistent levels of native-MOG Ab, reported when patients maintained MOG Ab-seropositivity for at least 3 months, was observed in 15/19 children (79%) and 20/32 adults (63%), with the majority exhibiting a relapsing phenotype (80%, 12/15 children; 70%, 14/20 adults) (Additional file 1: Table S2). Most longitudinal patients harbored native-P42 Ab at baseline (89%, 17/19 children; 84%, 27/32 adults). In all 17 children and most adults (89%, 24/27), the percentage of native-P42 Ab among total native-MOG Ab remained stable across their disease course regardless of their initial native-P42 Ab percentage (Fig. 4b). Longitudinal stability was also observed when native-P42 Ab titer fluctuations paralleled native-MOG Ab titer changes over time (Additional file 1: Figure S5C and D). Two adults had reduced percentage of native-P42 Ab whereas another adult had an increase in percentage of native-P42 Ab (Fig. 4b and Additional file 1: Figure S5E). However, their native-MOG Ab titers fluctuated significantly resulting in the observed large native-P42 Ab percentage change (Additional file 1: Figure S5F).

Native-MOG Ab titers were higher during active disease and decreased during disease remission in all 10 children and 10 adults (children, P = 0.006; adults P = 0.004) (Fig. 4c). Interestingly, although there was stable native-P42 Ab across disease course, the percentage of native-P42 Ab was slightly higher during active disease than at remission in 9/9 children and 6/7 adults (P > 0.05) (Fig. 4d). One adult, who harbored a low percentage of native-P42 Ab during active disease, had a significant increase of native-P42 Ab during remission (Fig. 4d), however this patient had a 10-fold decrease of native-MOG Ab titres from active disease to remission (Fig. 4c and Additional file 1: Figure S4, C and D, Patient A).

Overexpression of antigens in the live flow assay permits high antigen density and oligomerization of cell surface MOG that enables detection of high and low affinity antibodies [33, 58, 69]. A recent study detected high affinity MOG Ab using an immobilized folded extracellular MOG domain [58]. To a similar extent, we immobilized the extracellular domain of MOG spanning amino acids 1–117 (native-MOG^1–117) to detect high affinity MOG Ab by ELISA. 15% of children (20/134) and 18% of adults (26/145) bound to native-MOG^1–117, and therefore harbored high affinity MOG Ab (Fig. 5a). No controls were above the positive threshold, however, among the native-MOG Ab- patients, one adult and one child exhibited low levels of native-MOG^1–117 Ab. The Δ optical density (OD) values were positively correlated with native-MOG^1–117 Ab titer (Additional file 1: Figure S6A), suggesting a higher ΔOD value denoted a higher Ab concentration. The presence of high affinity MOG Ab was observed among a range of native-MOG Ab titers (Fig. 5b). Most native-MOG^1–117 Ab recognized native-P42^1–117 (55%, 11/20 children; 77%, 20/26 adults) and most of these patients (55%, 6/11 children; 70%, 14/20 adults) similarly harbored native-P42 Ab (Additional file 1: Figure S6B). There was no correlation between native-MOG^1–117 and native-MOG Ab titers (Additional file 1: Figure S6C), suggesting these patients have a combination of high and low affinity MOG Ab. Native-MOG^1–117 and fixed-MOG Ab titers were compared to determine whether high affinity MOG Ab allowed binding to fixed-MOG, and no correlation was observed (Additional file 1: Figure S6C), suggesting high binding in the live and fixed flow assays was not dependent on Ab affinity.

Presence of high affinity MOG Ab was sustained for up to 1.6 years in 5/6 children (median 3.4 months, IQR 1.6–5.80) and up to 4.9 years in all adults (5/5, median 5.1 months, IQR 4.2–13.7) (Fig. 5c). High affinity MOG Ab disappeared in one child at 20.8 months (Fig. 5c) which also coincided with an 80% decrease in native-MOG Ab. High affinity MOG Ab did not appear throughout clinical course in any patients with low affinity Ab at baseline, up to 4.7 years in children and 9.1 years in adults. In line with native-MOG Ab responses, all patients who recognized native-P42^1–117 maintained the same epitope recognition throughout their disease course (4 children, 4 adults) (Additional file 1: Figure S6D). Children with high affinity MOG Ab predominantly presented with ADEM (55%, 7/20 monophasic and relapsing ADEM, and 4/20 ADEM/ON), whereas adults with high affinity MOG Ab mainly had BON (30%, 8/26, monophasic and relapsing) followed by UON (24%, 6/26, monophasic and relapsing, Additional file 1: Figure S6E). Monophasic patients were more likely to harbor low affinity MOG Ab than relapsing patients (P = 0.088).

Eight MOG Ab binding patterns were identified based on their epitope, and binding to fixed-MOG and native-MOG^1–117 (Fig. 6). Overall, the human native-MOG Ab response recognized an immunodominant epitope at Proline42, and comprised of conformation-insensitive low affinity Ab, more so in children than in adults (Pattern 1; 42%, 58/139 children; 34%, 51/148 adults) (Fig. 6a). The second-most prevalent MOG Ab response was conformation-sensitive and similarly targeted Proline42 with low affinity (Pattern 3; 30%, 42/139 children; 28%, 41/148 adults). Interestingly, when the response targeted an epitope outside Proline42, although the response was predominantly of low affinity, a sensitivity to conformation was more prevalent (Pattern 7 and 8 combined; 11/139 children, 8%; 20/148 adults, 13%) (Fig. 6a). When clinical phenotypes were analyzed, adults with a monophasic disease course were more likely to have MOG Ab binding Pattern 1, with recognition of Proline42, the ability to bind fixed-MOG, and failure to bind to native-MOG^1–117 (P = 0.028) (Fig. 6b), whereas Patterns 5 and 7, characterized by low affinity Ab and binding to an epitope outside Proline42, were more likely to include adults with a relapsing course, irrespective of conformational sensitivity (P = 0.035) (Fig. 6b). There was no segregation of MOG Ab binding patterns between paediatric phenotypes.