ARTICLE OPEN ACCESS

High-throughput investigation of molecularand cellular biomarkers in NMOSDSoumya S Yandamuri MSE PhD Ruoyi Jiang MS Aditi Sharma MBBS Elizabeth Cotzomi BS

Chrysoula Zografou PhD Anthony K Ma MS Jessica S Alvey MS Lawrence J Cook PhD Terry J Smith MD

Michael R Yeaman PhD and Kevin C OrsquoConnor PhD On behalf of the Guthy-Jackson Charitable Foundation

CIRCLES Study Group

Neurol Neuroimmunol Neuroinflamm 20207e852 doi101212NXI0000000000000852

Correspondence

Dr OrsquoConnor

kevinoconnoryaleedu

AbstractObjectiveTo identify candidate biomarkers associated with neuromyelitis optica spectrum disorder(NMOSD) using high-throughput technologies that broadly assay the concentrations of serumanalytes and frequencies of immune cell subsets

MethodsSera peripheral blood mononuclear cells (PBMCs) and matched clinical data from participantswith NMOSD and healthy controls (HCs) were obtained from the Collaborative InternationalResearch in Clinical and Longitudinal Experience Study NMOSD biorepository Flow cytometrypanels were used to measure the frequencies of 39 T-cell B-cell regulatory T-cell monocytenatural killer (NK) cell and dendritic cell subsets in unstimulated PBMCs In parallel multiplexproteomics assays were used to measure 46 serum cytokines and chemokines in 2 independentNMOSD and HC cohorts Multivariable regression models were used to assess molecular andcellular profiles in NMOSD compared with HC

ResultsNMOSD samples had a lower frequency of CD16+CD56+ NK cells Both serum cohorts andmultivariable logistic regression revealed increased levels of B-cell activating factor associated withNMOSD Interleukin 6 CCL22 and CCL3 were also elevated in 1 NMOSD cohort of the 2analyzed Multivariable linear regression of serum analyte levels revealed a correlation betweenCX3CL1 (fractalkine) levels and the number of days since most recent disease relapse

ConclusionsIntegrative analyses of cytokines chemokines and immune cells in participants with NMOSDand HCs provide congruence with previously identified biomarkers of NMOSD and highlightCD16+CD56+ NK cells and CX3CL1 as potential novel biomarker candidates

From the Department of Neurology (SSY AS EC CZ KCOC) Department of Immunobiology (RJ KCOC) and Department of Pathology (AKM) Yale School of MedicineNew Haven CT University of Utah School of Medicine (JSA LJC) Salt Lake City Departments of Ophthalmology and Visual Sciences and Internal Medicine (TJS) University ofMichigan Medical School Ann Arbor Department of Medicine (MRY) David Geffen School of Medicine at the University of California Los Angeles Divisions of Molecular Medicineamp Infectious Diseases (MRY) Harbor-UCLA Medical Center Torrance and Lundquist Institute for Biomedical Innovation at Harbor-UCLA Medical Center (MRY) Torrance

Go to NeurologyorgNN for full disclosures Funding information is provided at the end of the article

Guthy-Jackson Charitable Foundation CIRCLES Study Group coinvestigators are listed in the appendix 2 at the end of the article

The Article Processing Charge was funded by the Guthy-Jackson Charitable Foundation

This is an open access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License 40 (CC BY-NC-ND) which permits downloadingand sharing the work provided it is properly cited The work cannot be changed in any way or used commercially without permission from the journal

Copyright copy 2020 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the American Academy of Neurology 1

Neuromyelitis optica spectrum disorder (NMOSD) is a rare au-toimmune demyelinating disease of the optic nerve and spinalcord12 Autoantibodies directed at the aquaporin-4 water channel(AQP4) expressed on astrocytes in theCNS have been identifiedas key contributors to NMOSD pathogenesis and are calledAQP4 immunoglobulin G (IgG)3ndash7 Approximately 70 of pa-tients withNMOSDare seropositive for AQP4-IgGwhich appearto contribute to disease by causing internalization of AQP4 andcomplement fixation8ndash10 Twenty to 50 of patients in whomAQP4-IgG cannot be detected are seropositive for myelin oligo-dendrocyte glycoprotein (MOG) autoantibodies (MOG-IgG)11ndash13 The remainder of NMOSD patients appear to be se-ronegative for both AQP4-IgG and MOG-IgG Neither autoan-tibody has been convincingly demonstrated to predict relapseresponse to therapy or prognosis14ndash18 Given the cumulative andpotentially devastating nature of relapses a prognostic biomarkerwould be especially valuable to herald imminent relapses andguide treatment interventionsMoreover specific biomarkersmayreveal disease mechanisms inform disease status provide insightfor development of therapies and helpmonitor treatment efficacy

Biomarker discovery in NMOSD is hampered by disease rarityFurthermore studies often use an a priori single-candidate ap-proach to biomarker discovery limiting the probability of dis-covery of a potential novel biomarker Although imagingtechniques and interrogation of CSF may provide more directinformation on CNS disease the collection of blood samples isless invasive time intensive and costly Therefore we sought toperform an unbiased discovery-based evaluation of candidatebiomarkers that can be accessed from peripheral blood samplesTo this end we acquired peripheral blood mononuclear cells(PBMCs) serum and clinical data from the Collaborative In-ternational Research in Clinical and Longitudinal ExperienceStudy (CIRCLES) NMOSD biorepository19 We applied high-throughput technologies to assess these biospecimens using asimultaneous hypothesis-generating strategy focused on a largeset of immune cell populations cytokines and chemokinesStatistical methods were then used to compare cell subset fre-quencies and serum analyte concentrations in NMOSD vshealthy controls (HC) In addition clinical metadata were in-tegrated into statistical modeling to assess potential relationshipsbetween molecular and cellular profiles and NMOSD relapses

MethodsStudy participants and biospecimensPBMC and serum specimens from participants with NMOSDand HCs were collected and archived through the Guthy-

Jackson Charitable Foundation CIRCLES study19 or through abiorepository established at the Yale University School ofMedicine Department of Neurology Written informed con-sent was obtained from all study participants before samplecollection

Two sets of serum samples were separately obtained andanalyzed independently to avoid batch effect These 2 serumsets were termed cohort 1 and cohort 2

Flow cytometry immunophenotyping ofcell subsetsFive flow cytometry panels previously validated by the HumanImmune Phenotyping Consortium20 were used to examine thefrequencies of 39 immune cell subsets (table e-1 linkslwwcomNXIA292) in unstimulated PBMCs These panels weredeveloped to standardize routine immunophenotyping in hu-mans across various sites All samples analyzed were verified tohave a PBMC viability greater than 80 assessed by7-aminoactinomycin D staining (Thermo Fisher WalthamMA) Briefly samples were thawed and incubated in the darkwith viability dye for 20 minutes Following washing PBMCswere split into 5 times 105 cells for each of the 5 panels Cells wereincubated with an antibody cocktail respective of each panel(table e-1 linkslwwcomNXIA292) at 4degC and then ana-lyzed on a BD Biosciences LSR Fortessa cytometer

Evaluation of serum cytokinesand chemokinesThe concentrations of 46 soluble circulating cytokineschemo-kines were measured using customized multiplex proteomicsassays (RampD Systems Human Magnetic Luminex Assay kitvendor catalog no CUST0I704QT840382 LXSAHM) foreach cohort The assays were conducted according to the man-ufacturerrsquos instructions Briefly diluted serum samples or stan-dardswere added to individual Luminexwells and incubatedwiththemicroparticle cocktail for 2 hours with agitation Sample wellswere washed and incubated in the dark with diluted biotin-antibody cocktail for 1 hour at room temperature with agitationAfter washing streptavidin-PE was added to each well and in-cubated in the dark for 30 minutes with agitation Last magneticmicroparticles were resuspended and read using a Luminex an-alyzer The concentration of each serum analyte was thenquantified using its respective standard curve

In cohort 1 B-cell activating factor (BAFF) which was notavailable in that Luminex kit was analyzed using a commercialELISA kit (RampD Systems Minneapolis MN) following man-ufacturer instructions

GlossaryAQP4 = aquaporin-4 ARR = annual relapse rate AUC = area under the curve BAFF = B-cell activating factor CIRCLES =Collaborative International Research in Clinical and Longitudinal Experience Study HC = healthy control IgG =immunoglobulin G IL = interleukin MOG = myelin oligodendrocyte glycoprotein NK = natural killer NMOSD =neuromyelitis optica spectrum disorder PBMC = peripheral blood mononuclear cell ROC = receiver operating characteristic

2 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 NeurologyorgNN

Statistical analysisFlow cytometry frequencies were calculated using FlowJo v104The percent frequency of each cell subset was determined bycalculating the fraction of the subset population relative to itsparent population Univariate statistical analyses on immuno-phenotyping and serum analyte data were performed using theWelch t test and 95 CIs of the differences of the means (withmean difference defined as meanNMOSDmdashmeanHC) werecalculated in R To correct for multiple hypotheses testing q-value or p-adjusted (p-adj) was calculated while maintaining a5 false discovery rate using the Storey algorithm in R Forunivariate analyses of serum analytes p-adj le 005 was inter-preted as significant whereas for immunophenotyping p le 005was interpreted as significant Further evaluation of univariateanalyses based on rituximab treatment was performed using theWelch t test and p le 005 was interpreted as significant

Multivariate statisticsTo permit comparisons across serum analyte cohorts nor-malization was performed by z-score normalizing assay values(subtracting the mean and dividing by the SD)

Multivariable logistic or linear regression was performed usingthe glm function in base R v342 for the prediction of NMOSDvs HC or for detecting associations with NMOSD clinical datasuch as days since relapse or annual relapse rate (ARR) Allavailable serum analyte or immunophenotyping frequencieswere used as variables in these models and no data weremissing during model training R2 (1 subtracted by the ratio ofnull deviance to residual deviance) was used to evaluate forgoodness of fit of multivariable linear regression models Five-fold cross validation was performed to assess models foroverfitting of multivariable logistic regression models A re-ceiver operating characteristic (ROC) curve was computedand the area under the curve (AUC) was calculated to assessperformance using the pROC v1153 R package The Rpackage boot v13-22 was used to compute CIs for AUCPrincipal component analysis was performed using base RMean and 95CIs of the mean (95CImean) were calculatedfor days since relapse and ARR All codes used in this study areavailable on request A p value le005 was interpreted as sta-tistically significant

Data availabilityAnonymized data will be shared on request from qualifiedinvestigators

ResultsClinical and demographic characteristics ofparticipants with NMOSD and HCsPBMC samples were provided by 12 participants withNMOSD and HCs Serum samples were provided by 27 par-ticipants with NMOSD and 11 HCs for cohort 1 and 29 par-ticipants withNMOSD and 11HCs for cohort 2 Demographicand clinical characteristics of study participants are summarized

in table 1 The mean age of each NMOSD and HC group forthe 3 analyses ranged from 315 to 545 years Themean diseaseduration ranged from 540 to 642 years The majority of par-ticipants were female (between 67 and 84 per group)NMO-IgG serostatus was distributed between seronegativeseropositive and unknown for all NMOSD groups

Primary reported race or ethnicity was heterogeneous betweengroups as was treatment history At least 50 of the partici-pants with NMOSD in each group had been treated with rit-uximab Varying frequencies of participants had been treatedwith corticosteroid IVIg plasma exchange or other immuno-therapies Participants were also heterogeneous in terms ofprimary symptoms and phenotypic presentation includingoptic neuritis or longitudinally extensive transverse myelitis

CD16+CD56+ NK cell frequency is reduced inNMOSD vs HCsWe next sought to determine whether differences in immunecell frequencies between participants with NMOSD and HCscould be identified To that end 5 validated flow cytometrypanels that applied defined markers were used to delineatesubsets of T lymphocytes B lymphocytes monocytes naturalkiller (NK) cells and dendritic cells in unstimulated PBMCsPanel 1 measured central effector and naive memory T cellspanel 2 measured CD4+ Th1 Th2 and Th17 cells as well astheir CD8+ counterparts defined by the samemarkers namelyTh1-like Th2-like and Th17-like CD8+ cells panel 3 mea-sured regulatory T-cell subsets panel 4 measured B-cellsubsets and panel 5 measured monocyte NK cell and den-dritic cell subsets (table e-1 linkslwwcomNXIA292) Atotal of 39 cell subset frequencies were analyzed in the 5panels (table 2)

The frequency of 3 populations was significantly differentbetween participants with NMOSD and HCs CD16+CD56+

NK cell frequency was decreased by 54 in NMOSD vs HC(95 CI minus568 to minus104 p = 00067) A reduction inCD16+CD56+ NK cell frequency was still seen in the partic-ipants with NMOSD who had not been treated by rituximab(95 CI minus697 to minus645 p = 0024) though not in those whohad been treated by rituximab (95 CI minus629 to 472 p =0083) The frequency of Th1-like CD8+ T cells was alsoreduced in NMOSD vs HC by 53 (95 CI minus398 to minus0192p = 0034) Stratification analyses based on rituximab treat-ment revealed that both rituximab-treated (95 CI minus406 tominus0176 p = 0035) and untreated (95CIminus415 to minus0451 p =0019) participants with NMOSDhad reduced Th1-like CD8+

T cells compared with HCs Last the frequency of B cellsoverall was also lower in NMOSD than HC by 54 (95 CIminus633 tominus0702 p = 0017) However further analysis revealedthat rituximab-untreated participants had a similar frequencyof B cells as HCs (95 CI minus472 to 328 p = 0685) and it wasthe rituximab-treated group who had the reduced B-cell fre-quency (95 CI minus796 to minus466 p = 35 times 10minus6) All other cellsubset frequencies were similar in NMOSD comparedwith HC

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 3

Analysis of serum analytes validatesNMOSD biomarkersSerum analyte concentrations were measured using multiplexproteomics assays and ELISA in 2 independent cohorts A total of46 serum cytokines or chemokines weremeasured in each cohort4 of these differed significantly between NMOSD and HC in oneor both cohorts (table 3) BAFFwas elevated inNMOSDsamplesin cohorts 1 (95 CI 671ndash1430 p-adj = 16 times 10minus4) and 2 (95CI 1360ndash3880 p-adj = 00028) As rituximab treatment has beenshown to elevate serum BAFF18 we evaluated whether BAFFserum concentration differed based on previous rituximab treat-ment The groups arranged in ascending order of BAFF con-centration are HC untreated NMOSD all (unstratified)NMOSD and treated NMOSD In both cohorts BAFF con-centration was elevated in rituximab-treated NMOSD (cohort 1

95CI 769ndash1650 p= 13 times 10minus5 cohort 2 95CI 2740ndash6190p = 47 times 10minus4) in comparison to HC In cohort 2 the serumBAFF concentration was elevated in rituximab-treated NMOSDcompared with untreated (95 CI 1740ndash5620 p = 000065) Insummary the rituximab-treated and unstratified NMOSD groupshad significantly elevated serumBAFF in both cohorts The use of2 independent cohorts and 2 different methods (ELISA andmultiplex proteomics assay) validated BAFF as a significant result

The 3 other serum analytes that differed between NMOSD andHCwere interleukin (IL)-6 (95CI 0500ndash197 p-adj = 0013)CCL22 (95 CI 104ndash504 p-adj = 0019) and CCL3 (95 CI849ndash507 p-adj = 0027) which were all elevated in cohort 2only IL-17F S100B FGF-basic IL-15 RANTES and IL-1RAdid not meet the threshold for significance (p-adj le 005) but

Table 1 Clinical and demographic characteristics of NMOSD and HC participants

Immunophenotyping Cohort 1 serum Cohort 2 serum

HC (n = 12) NMOSD (n = 12) HC (n = 11) NMOSD (n = 27) HC (n = 11) NMOSD (n = 29)

Age y mean plusmn SD 486 plusmn 167 443 plusmn 13711 315 plusmn 131 395 plusmn 12924 545 plusmn 175 316 plusmn 168

Disease duration ymean plusmn SD

mdash 555 plusmn 43611 mdash 642 plusmn 67822 mdash 540 plusmn 57426

Sex n ()

Female 8 (67) 10 (83) 8 (73) 2125 (84) 8 (73) 21 (72)

Male 4 (33) 2 (17) 3 (27) 425 (16) 3 (27) 8 (28)

Serostatusa n ()

AQP4-IgG+ mdash 5 (42) mdash 9 (33) mdash 17 (59)

AQP4-IgG2 mdash 3 (25) mdash 7 (26) mdash 7 (24)

AQP4-IgG unknown mdash 4 (33) mdash 11 (41) mdash 5 (17)

Raceethnicity n ()

Caucasian 9 (75) 10 (83) 8 (73) 721 (33) 8 (72) 20 (69)

BlackAfrican American 0 (0) 1 (83) 0 (0) 921 (43) 0 (0) 3 (10)

LatinoHispanic 0 (0) 1 (83) 1 (9) 321 (14) 1 (90) 5 (17)

Asian 3 (25) 0 (0) 2 (18) 321 (14) 2 (18) 0 (0)

Other b 0 (0) 0 (0) 1 (9) 021 (0) 0 (0) 1 (34)

Treatment n ()

Rituximab mdash 6 (50) mdash 20 (74) mdash 16 (55)

Corticosteroids mdash 8 (67) mdash 1323 (57) mdash 8 (28)

Other immunotherapies mdash 6 (50) mdash 111 (91) mdash 8 (28)

IVIg mdash 2 (17) mdash 011 (0) mdash 1 (34)

PLEX mdash 1 (83) mdash 111 (91) mdash 1 (34)

Abbreviations AQP4-IgG = aquaporin-4 immunoglobulin G HC = healthy control IVIg = IV immunoglobulin NMOSD = neuromyelitis optica spectrumdisorder PLEX = plasma exchangeIf demographics or clinical data are unknown for some samples in a group they are depicted as a fraction with the number of known samples indicated in thedenominator For example 111 (91) for PLEX under treatment means that of the 11 samples for which PLEX treatment status is known 1 or 91 hadundergone PLEX treatmenta Serostatus was determined using standard methods approved for NMOSD diagnosis and confirmed by the site investigatorb The ldquoOtherrdquo category includes American Indian First Canadian Alaskan Native Hawaiian Native Pacific Islander or another specified racialethnic identity

4 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 NeurologyorgNN

Table 2 Mean immune cell frequencies in NMOSD and HC PBMC samples

Cell subset HC NMOSD Difference (95 CI) p Value p-adj

CD16+CD56+ NKNK cells 654 319 2568 to 2104 00067 031

B cellsviable CD32 646 294 2633 to 20702 0017 031

Th1-likeCD8+ T cellsa 395 187 2398 to 20192 0034 045

Central memoryCD8+ T cells 508 957 minus103 to 999 010 070

CD162CD56 brightNKNK cells

371 585 minus0479 to 475 010 070

NaiveB cells 617 458 minus373 to 568 013 070

Memory CCR4+Tregs 088 0295 minus139 to 0225 014 070

CXCR5+PD12CD4+ T cells 540 382 minus377 to 0603 015 070

ActivatedCD8+ T cells 122 275 minus0771 to 382 017 070

T follicularhelperCD4+ T cells

131 0541 minus196 to 0414 018 070

Activated CCR4+Tregs 0129 0383 minus0133 to 0641 018 070

CD8+ Tviable CD3+ 262 224 minus104 to 272 024 074

TregCD4+ T cells 196 266 minus0582 to 199 026 074

NK cellsviable CD32192202 315 256 minus167 to 483 026 074

EffectorCD4+ T cells 597 108 minus563 to 152 035 083

Effector memoryCD8+ T cells 424 362 minus198 to 725 035 083

Th2-likeCD8+ T cellsa 0563 0340 minus0711 to 0265 035 083

DCsLin2NK-subset 129 173 minus653 to 154 040 086

PlasmablastsCD19+CD202

viable CD32690 414 minus978 to 424 042 086

IgD2 memoryB cells 951 128 minus584 to 123 045 086

IgD+ memoryB cells 106 163 minus105 to 219 046 086

Plasmacytoid DCsDCs 259 226 minus138 to 727 052 086

CD4+ T cellsviable CD3+ 567 533 minus152 to 851 056 086

MonocytesLin2NK-subset 433 397 minus159 to 878 056 086

Th17-likeCD8+ T cellsa 145 176 minus0899 to 152 060 086

CD45RO+CD4+ T cells 370 332 minus199 to 123 062 086

ActivatedCD4+ T cells 0470 0383 minus0445 to 0278 063 086

NaiveCD4+ T cells 425 379 minus238 to 148 063 086

Th1CD4+ T cells 483 600 minus445 to 679 066 086

NaiveCD8+ T cells 256 287 minus122 to 185 067 086

Th17CD4+ T cells 349 307 minus326 to 243 076 091

Central memoryCD4+

T cells288 272 minus146 to 113 079 091

Naive CCR4+Tregs 174 201 minus193 to 247 080 091

Myeloid DCsDCs 419 434 minus121 to 152 081 091

EffectorCD8+ T cells 270 256 minus165 to 137 085 091

Effector memoryCD4+ T cells 227 242 minus155 to 183 086 091

Continued

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 5

were elevated inNMOSD in 1 cohort (ple 005) All other serumanalyte concentrations in NMOSD were similar to HC

Multivariable logistic regression identifiesBAFF association with NMOSDWe performed multivariable logistic regression analyses tomodel the association of serum analyte concentrations orimmunophenotyping frequencies with NMOSD The abilityto predict NMOSD was assessed using 5-fold cross validationevaluated through quantification of the AUC of the associatedROC curve (figure 1 AndashD table 4) Models constructed onimmunophenotyping and serum analyte concentrations fromboth cohorts had a greater predictive ability (07 lt AUC lt08) than models constructed on serum analyte concentra-tions from cohort 1 or 2 alone (05 lt AUC lt 07) The modelbased on serum analytes from both cohorts had the highestAUC (0734) and standard accuracy value (0700) and wasfurther analyzed to identify significant parameters

In the multivariate serum analyte model based on both co-horts BAFF concentration was found to be associated withNMOSD (p = 0022 figure 1E) consistent with observationsfrom univariate analyses BAFF concentration remained as-sociated with NMOSD in this model when rituximab treat-ment was included as a covariate (p = 0038) A multivariablelogistic regression model constructed on cohort 1 and testedon cohort 2 performed with an AUC of 066 and accuracy of058 suggesting weak generalizability between cohorts (figure1 F and G) The cohort 1ndashtrained model was a relatively poorpredictor (AUC lt07) of NMOSD in cohort 2

A separation of NMOSD and HC was observed when visual-izing components 1 and 2 derived from principal componentanalysis (figure e-1 AndashD linkslwwcomNXIA291) Thesecomponents explained only 45 and 25 respectively of thetotal variance from immunophenotyping and serum analytes(both cohorts)We did not observe distinct clustering based onthese analyses Principal components 1 and 2 explained 73and 78 of the total variance of cohorts 1 and 2 respectivelyAlthough clustering was still not observed in these decompo-sitions someNMOSD samples did separate No differentiationof samples was seen based on institutional source or cohort(figure e-1 EndashJ linkslwwcomNXIA291) In general

principal component analysis of immunophenotyping or serumanalyte data sets could not reliably distinguish NMOSD vsHC

Serum CX3CL1 concentration correlates withdays since relapseWe next sought to investigate whether measured serum analyteconcentrations correlated with temporal measures of clinicaloutcome particularly the number of days since relapse or ARRThe mean number of days since relapse was 590 (95 CImean350ndash820) and the mean ARR was 08 (95 CImean 05ndash12)Multivariable linear regression models were used to identify as-sociations between cohort 1 and 2 serum analyte levels and the 2outcome variables Thesemodelsfit weakly though better for dayssince relapse (R2 = 0602) than ARR (R2 = 0361) Statisticalanalysis of serum analyte parameters (table e-2 linkslwwcomNXIA292) revealed that the chemokine CX3CL1 (fractalkine p= 0031) was associated with days since relapse This associationindicated that CX3CL1 levels in serum decline as a function oftime since previous relapseNoother serumanalytewas significantin our multivariable linear regression analyses for either outcome

DiscussionThe goal of this exploratory study was to assess circulatingimmune cell and cytokinechemokine profiles using an un-biased and hypothesis-generating strategy This approach ach-ieved 3 overarching goals which were to (1) verify the utility ofthe CIRCLES biospecimen repository in supporting molecularand cellular NMOSD research (2) reference known biomarkersas a means of validation using external standards and (3)identify potentially novel candidate NMOSD biomarkers Withthe high-throughput evaluation of a large set of circulating im-mune cell frequencies and serum cytokine and chemokineconcentrations we have validated previously identified bio-markers Beyond this concordance the current studies have alsoidentified CD16+CD56+ NK cells and CX3CL1 as intriguingand novel biomarker candidates to be further evaluated forpossible NMOSD predictive or prognostic applications

A significant decrease in peripheral NK cell frequency has beenreported in patients with NMOSD in comparison to patientswith MS and HCs21 However stratification by clinical status

Table 2 Mean immune cell frequencies in NMOSD and HC PBMC samples (continued)

Cell subset HC NMOSD Difference (95 CI) p Value p-adj

Th2CD4+ T cells 0661 0642 minus0499 to 0460 093 094

CCR4+Tregs 202 210 minus174 to 190 093 094

TransitionalB cells 772 792 minus560 to 600 094 094

Abbreviations DC = dendritic cell HC = healthy control NK cells = natural killer cells NMOSD = neuromyelitis optica spectrum disorder p-adj = p-adjusted (q-value) PBMC = peripheral blood mononuclear cell Tregs = regulatory T cellsStatistical analysis was performed using the Welch t test and Storey adjustment for multiple comparisons and serum analytes with p le 005 are bolded Cellsubsets are in ascending order of p valuea Th1-like Th2-like and Th17-like CD8+ T cells are the CD8+ correlates of CD4+ Th1 Th2 and Th17 subtypes based on marker expression

6 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 NeurologyorgNN

Table 3 Mean NMOSD and HC serum analyte concentrations in 2 cohorts

Analyte

Cohort 1 Cohort 2

HC pgmL NMOSD pgmL Difference (95 CI) p Value (p-adj) HC pgmL NMOSD pgmL Difference (95 CI) p Value (p-adj)

BAFF 527 1580 671 to 1430 44 times 1026 (16 times 1024) 1350 3970 1360 to 3880 18 times 1024 (00028)

IL-6 mdash mdash mdash BL 181 304 0500 to 197 00018 (0013)

CCL22 332 314 minus859 to 489 058 (078) 469 773 104 to 504 00038 (0019)

CCL3 116 874 minus145 to 875 061 (078) 986 128 849 to 507 00071 (0027)

S100B mdash mdash mdash BL 469 901 612 to 804 0024 (0072)

FGF-basic mdash mdash mdash BL 111 157 00262 to 0884 0038 (0089)

IL-15 mdash mdash mdash BL 400 457 000840 to 114 0047 (0089)

IL-1RA 534 571 minus186 to 259 074 (088) 702 1130 665 to 856 0047 (0089)

IL-2 mdash mdash BL 296 479 minus0678 to 374 0058 (0099)

TGF-a mdash mdash BL 500 589 minus0102 to 188 0077 (012)

CCL2 1378 163 minus166 to 679 023 (051) 530 410 minus303 to 641 019 (021)

CX3CL1 553 566 minus143 to 168 087 (089) 9329 1147 minus136 to 564 022 (021)

IP-10 141 155 minus455 to 743 063 (078) 193 252 minus401 to 159 023 (021)

APRIL mdash mdash NA 1830 2200 minus260 to 1010 024 (021)

FLT-3L 351 362 minus642 to 855 077 (088) 732 885 minus108 to 415 024 (021)

IL-1a 171 175 minus00297 to 0125 022 (051) 420 428 minus00543 to 0210 024 (021)

IL-13 mdash mdash mdash BL 419 686 minus261 to 794 031 (021)

IL-21 mdash mdash mdash NA 990 122 minus226 to 676 032 (021)

CCL4 130 115 minus187 to 155 085 (089) 366 208 minus509 to 195 035 (021)

G-CSF mdash mdash BL 144 183 minus489 to 128 037 (021)

IL-8 131 621 minus158 to 114 013 (046) 184 276 minus128 to 312 040 (023)

IL-19 mdash mdash mdash BL 253 249 minus161 to 724 045 (024)

CCL7 mdash mdash mdash BL 910 892 minus707 to 346 048 (024)

Eotaxin 559 712 minus0199 to 308 0053 (032) 120 105 minus599 to 296 049 (024)

Continued

Neurolo

gyorgN

NNeurologyN

euroim

munology

ampNeuroinflam

mation

|Volum

e7N

umber

5|

Septem

ber2020

7

Table 3 Mean NMOSD and HC serum analyte concentrations in 2 cohorts (continued)

Analyte

Cohort 1 Cohort 2

HC pgmL NMOSD pgmL Difference (95 CI) p Value (p-adj) HC pgmL NMOSD pgmL Difference (95 CI) p Value (p-adj)

IL-31 mdash mdash mdash BL 138 165 minus512 to 106 049 (024)

IL-16 mdash mdash mdash NA 159 175 minus437 to 760 059 (026)

IL-7 283 303 minus0280 to 0672 041 (067) 861 964 minus282 to 488 059 (026)

IL-27 mdash mdash mdash BL 164 158 minus398 to 277 072 (030)

Eotaxin-2 535 463 minus276 to 131 047 (073) 1760 1570 minus1480 to 1100 076 (031)

IL-4 263 32 minus0525 to 172 029 (058) 2360 2373 minus0835 to 108 080 (032)

CD40L 2370 2470 minus638 to 834 079 (088) 5120 5200 minus2040 to 2220 093 (036)

IL-33 mdash mdash mdash BL 165 166 minus384 to 408 095 (036)

IL-17F 335 507 384 to 306 0013 (024) mdash mdash mdash NA

RANTES 679 1120 868 to 875 0046 (032) mdash mdash mdash NA

IL-3 565 672 minus0189 to 232 0093 (042) mdash mdash mdash BL

PDGF-BB 113 462 minus421 to 112 036 (062) mdash mdash mdash NA

PDGF-AA 946 109 minus406 to 695 060 (078) mdash mdash mdash NA

Abbreviations BL = concentration below limit for analysis HC = healthy control NA = not available NMOSD = neuromyelitis optica spectrum disorder p-adj = p-adjusted (q-value)Statistical analysis was performed using theWelch t test and Storey adjustment formultiple comparisons and serumanalytes with p le 005 are bolded Serumanalytes are in ascending order of cohort 2 p value and then cohort1p value Ten serumanalyteswere below limit for analysis for both cohorts (IFNγ IL-10 TNFα IL-17A IL-17EIL-25 IL-5 IL-36β GM-CSF IL-22 and IL-12 p70) Three serumanalytesmeasured belowdetection limit in cohort 1 andwere not analyzed in cohort 2 (LT-α IL-12 p40 and IL-11) and 3 serum analytes measured below detection limit in cohort 2 and were not analyzed in cohort 1 (IL-9 TNFβ and IL-23)

8NeurologyN

euroim

muno

logyampNeuroinflam

mation

|Vo

lume7N

umber

5|

Septem

ber

2020NeurologyorgN

N

revealed that only patients with relapsing NMOSD had sig-nificantly less NK cells those in remission did not21 ThereforeNK cell frequencymay be a candidate biomarker with temporal

specificity In this study we did not find a significant change intotal NK cell frequency although we found a decrease inCD16+CD56+ NK cell subset frequency in participants with

Figure 1Multivariable logistic regressions based on immunophenotyping frequencies and serum analyte concentrations

ROC curves for 5-fold cross validation of multi-variable logistic regression models were con-structed for NMOSD predictability based on (A)immunophenotyping (B) cohort 1 (C) cohort 2and (D) cohorts 1 and 2 serum analyte concen-trations Combined serum cohort and immuno-phenotypingmodels have better predictive valuethanmodels constructed on cohorts 1 or 2 alone(E) Coefficients for the combined cohort modelare depicted with SD BAFF concentration wassignificantly associated with NMOSD (p = 0022in orange) Cohort 2 was tested on a multivari-able regressionmodel for cohort 1 for which the(F) ROC curve and (G) confusion matrix areshown They revealed an AUC of 066 and accu-racy of 058 AUC = area under the curve NMOSD= neuromyelitis optica spectrum disorder ROC =receiver operating characteristic p le 005 wasconsidered statistically significant

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 9

NMOSD Of note CD16+ NK cells express perforin and arecytotoxic they are involved in antibody-dependent cellularcytotoxicity which is a mechanism associated with NMOSDpathogenesis2223 Although histopathologic examination ofNMOSD lesions has not identified NK cells as a predominantinfiltrating cell type studies conducted in vitro or with mousemodels have demonstrated NMO-IgGndashdependent astrocytekilling by NK cells highlighting a possible mechanism forNMOSD pathology24ndash26 Other studies elucidate a protectiverole of NK cells in an autoimmune demyelinating diseasemodel and a reduction in NK cell frequency in patients withMS2728 Our findings provide impetus for further inquiry intoCD16+CD56+ NK cells and their functions in association withNMOSD relapses in a larger patient sample

CX3CL1 is also a viable novel biomarker candidate for furtherstudy in NMOSD We found that CX3CL1 was significantlyassociated with the number of days since relapse in a multi-variable linear regression model CX3CL1 also known asfractalkine was originally called neurotactin due to its highexpression in the CNS including by astrocytes2930 CX3CL1 isimportant for chemotaxis of leukocytes to the CNS and iscommonly studied in the context of neurologic disease Forexample CX3CL1 is elevated in the serum of patients withMS31 Its receptor CX3CR1 is expressed on microgliamonocytes dendritic cells and especially NK cells32 NotablyCD16+ NK cells highly express CX3CR133 and variations inCX3CR1 are correlated with age-related macular de-generation34 These collective findings reinforce our currentdata suggesting that CX3CL1 is an intriguing candidate forfurther study in the context of NMOSD relapse prediction

BAFF elevation as a result of rituximab treatment is a recognizedphenomenon and has been implicated in relapses occurringimmediately subsequent to rituximab dosing18 Ostensibly thedepletion of B cells causes a compensatory elevation in BAFFpromoting B-cell activation and disease recrudescence ElevatedBAFF levels have been observed in the CSF and serum ofpatients with NMOSD3536 Autoreactive B cells are especiallydependent on BAFF for survival BAFF selectively rescuesautoreactive B cells from elimination37 Elevated BAFF con-centration has been observed in other autoimmune diseases aswell including systemic lupus erythematosus Sjogren

syndrome and rheumatoid arthritis and it correlates with dis-ease status3839 In the present study BAFF was significantlyelevated in both study cohorts in unstratified NMOSD groupsFurthermore BAFF was significantly associated with NMOSDin a multivariable logistic regression model even with rituximabtreatment used as a covariate Therefore our analyses are con-sistent with previous studies identified BAFF as an NMOSDbiomarker

The current findings were supported by the observation thattreatment history is associated with immunophenotypingfrequencies and serum analyte concentrations For examplewe observed a significant reduction in B-cell frequency inNMOSD samples in relation to rituximab treatment Thisoutcome was anticipated given that rituximab depletesCD20+ B cells from peripheral circulation However ritux-imab also depletes nonndashB-cell CD20+ cells such as a smallpopulation of CD3+ T lymphocytes40 Of interest CD20+

T cells are more likely to be CD8+40 and we observed that thefrequency of Th1-like CD8+ T cells was significantly reducedin participants with NMOSD with a history of rituximabtherapy Thus this study may provide support for the hy-pothesis that rituximab treatment affects cell subsets beyondCD20+ B cells and therefore may contribute to its mecha-nism(s) of action This concept deserves additional in-vestigation Steroids and other preventive treatments used fortheir immunosuppressive properties are also likely to affectimmune cell frequencies and serum analyte concentrationsHere sample sizes for other treatments were not sufficient foranalysis by stratification given that stratification by treatmentrequires sizable cohorts Compounded by the rarity ofNMOSD treatment heterogeneity and limited treatment-naive patients are challenges to biomarker discovery inNMOSD However treated patients should be evaluated forprognostic biomarkers given such biomarkers would havegreater value if they predicted a potential relapse regardless oftreatment

Taken together the present findings provide impetus for fur-ther study of CX3CL1 and CD16+CD56+ NK cells as potentialnovel biomarker candidates specifically associated withNMOSD and its relapse patterns Considering that the eleva-tion of serum CCL22 IL-6 and CCL3 was not validated by asecond cohort it remains to be determined whether signifi-cantly different concentrations would be found in largerNMOSD cohorts Therefore these serum analytes should befurther validated for consideration as biomarkers as shouldTh1-like CD8+ T cells Future studies involving larger cohortsmay reveal that these or other immunologic constituents eitheralone or as components of composite signatures representimportant indicators of disease status In any case the currentstudies affirm the use of high-throughput techniques for bio-marker identification in larger NMOSD cohorts It would beinformative to reevaluate the predictive value of multivariablemodels and principal component analyses using a larger cohortand further stratification based on treatment relapse recencyand other clinical factors Larger sample sizes assessment of

Table 4 AUC and ACC formultivariable logistic regressioncurves

Data AUC mean plusmn SD ACC mean plusmn SD

Immunophenotyping 0700 plusmn 0217 0530 plusmn 0157

Serum analytes cohort 1 0634 plusmn 0239 0464 plusmn 0163

Serum analytes cohort 2 0503 plusmn 0206 0444 plusmn 0193

Serum analytes both 0734 plusmn 0143 0700 plusmn 0142

Abbreviations ACC = standard accuracy value AUC = area under the curveROC = receiver operating characteristic

10 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 NeurologyorgNN

biospecimens from clinical trials and investigation of samplesdrawn immediately before and after relapse events might dis-play improved delineation of NMOSD from HC in multivari-able models We were not able to assess the correlationbetween immunophenotyping and clinical outcomes due tolimited relapse data in these study participants (only 4 partic-ipants had paired relapse data)

Our findings provide validation for the use of high-throughputtechnologies particularly multipanel flow cytometry and multi-plex proteomics assay for identification of candidate NMOSDbiomarkers Furthermore we have identified CX3CL1 andCD16+CD56+ NK cells as intriguing and novel prognostic bio-marker candidates for further study Based on these findings forfuture evaluation of largeNMOSD cohorts with high-throughputtechniques we propose (1) appropriate and standardized col-lection and storage of samples as in the CIRCLES program orclinical trials (2) verification of sample quality such as PBMCviability quantification (3) collection of paired information ontreatment and clinical characteristics such as relapse data and (4)controlling for or stratification analyses based on treatment his-tory current treatment regimen andor clinical characteristics

AcknowledgmentThe authors thank Dr Lesley Devine Scientific Director ofthe Immune Monitoring Core Facility at Yale School ofMedicine for assisting with the FACS and proteomics assayThe authors also appreciate the efforts of the followingindividuals Yanet Babcock J Michael Dean Colleen FarrellHaojun Feng Susan Filomena Toni Ganaway Myka Barnes-Garcia Samuel Glaisher Rivka Green Elizabeth GonzalesElaine Hsu Catherine J Johnson Ilana Katz Sand Ilya KisterMarlene Keymolen Ramirez Allyson Reid Pavle RepovicGloria Rodriguez Jennifer L Sedlak Nancy L Sicotte AngelaStangorone Ben W Thrower Anthony L Traboulsee andGabriella Tosto The CIRCLES project team including sitePIs and coordinators the DCC team and the GJCF and itsadvisors are grateful to study participants caregivers andfamilies who participated in this study The CIRCLES projectis supported by the Guthy-Jackson Charitable Foundation

Study fundingSupported by the Guthy-Jackson Charitable Foundation andits CIRCLES research and biorepository program

DisclosureS S Yandamuri R Jiang A Sharma E Cotzomi C ZografouA K Ma J S Alvey and L J Cook report no disclosures T JSmith is a paid consultant for Horizon Therapeutics andImmunovant has been issued US patents for the diagnosisand treatment of autoimmune diseases with IGF-I receptormonoclonal antibodies and is a medical advisor to theGuthy-Jackson Charitable Foundation Michael R Yeaman isfounder of NovaDigm Therapeutics Inc developing novelvaccines and immunotherapeutics and Metacin Inc de-veloping novel neuroscience pattern recognition tools hasreceived honoraria from Alexion Inc and Genentech-Roche

Inc is a member of the Genentech-Roche Inc ScientificAdvisory Committee and is Chair medical advisor to theGuthy-Jackson Charitable Foundation K C OrsquoConnor hasreceived research support from Ra Pharma is a consultantand equity shareholder of Cabaletta Bio and is the recipientof a sponsored research subaward from the University ofPennsylvania the primary financial sponsor of which is Ca-baletta Bio Go to NeurologyorgNN for full disclosures

Publication historyReceived by Neurology Neuroimmunology amp NeuroinflammationApril 12 2020 Accepted in final form June 19 2020

Appendix 1 Authors

Name Location Contribution

Soumya SYandamuriMSE PhD

Yale UniversityNew Haven

Analyzed data interpreted dataacquired data and draftedmanuscript for intellectual content

Ruoyi JiangMS Yale UniversityNew Haven

Performed multivariate statisticalanalyses and modeling andcontributed to manuscript writing

Aditi SharmaMBBS

Yale UniversityNew Haven

Acquired data analyzed data andcontributed to manuscript writing

ElizabethCotzomi BS

Yale UniversityNew Haven

Acquired data and analyzed data

ChrysoulaZografou PhD

Yale UniversityNew Haven

Analyzed data and revised themanuscript

AnthonyKMaMS

Yale UniversityNew Haven

Performed multivariate statisticalanalyses and modeling

Jessica SAlvey MS

University ofUtah Salt LakeCity

Coordinated biorepositoryparticipant information

Lawrence JCook PhD

University ofUtah Salt LakeCity

Managed biorepository participantinformation

Terry J SmithMD

University ofMichigan AnnArbor

Designed and conceptualized thestudy interpreted data andrevised the manuscript forintellectual content

Michael RYeaman PhD

University ofCalifornia LosAngeles

Designed and conceptualized thestudy interpreted data andrevised the manuscript forintellectual content

Kevin COrsquoConnor PhD

Yale UniversityNew Haven

Designed and conceptualized thestudy supervised data acquisitioninterpreted data and revised themanuscript for intellectual content

Appendix 2 Coinvestigators from the Guthy-JacksonCharitable Foundation CIRCLES Study Group

Name Location Contribution

LilyanaAmezcuaMD MS

Department of Neurology KeckSchool of Medicine Universityof Southern California LosAngeles

Reviewed and revisedthe manuscript forintellectual content

Continued

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 11

Appendix 2 (continued)

Name Location Contribution

ErikaAmundsonBA

Department of Neurology KeckSchool of Medicine Universityof Southern California LosAngeles

Reviewed and revisedthe manuscript forintellectual content

Jeffrey LBennettMD PhD

Departments of Neurology andOphthalmology University ofColorado School of MedicineAurora

Reviewed and revisedthe manuscript forintellectual content

Jacinta MBehne MA

The Guthy-Jackson CharitableFoundation Beverly Hills CA

Reviewed and revisedthe manuscript forintellectual content

MeganKenneallyBehne AS

The Guthy-Jackson CharitableFoundation Beverly Hills CA

Reviewed and revisedthe manuscript forintellectual content

Terrence MBlaschkeMD PhD

Stanford University School ofMedicine Palo Alto CA

Reviewed and revisedthe manuscript forintellectual content

Robert LCarruthersMD

Department of Medicine ampNeurology University of BritishColumbia Vancouver Canada

Reviewed and revisedthe manuscript forintellectual content

TanujaChitnis MD

Department of NeurologyBrigham and WomenrsquosHospital Harvard MedicalSchool Boston MA

Reviewed and revisedthe manuscript forintellectual content

Jeffrey ACohen MD

Mellen Center for MSTreatment and ResearchNeurological InstituteCleveland Clinic OH

Reviewed and revisedthe manuscript forintellectual content

JessicaColeman BA

Department of NeurologyJohns Hopkins UniversitySchool of Medicine BaltimoreMD

Reviewed and revisedthe manuscript forintellectual content

Casey EngelBA

Weill Cornell Medicine NewYork NY

Reviewed andrevised themanuscript forintellectualcontent

LisaEunyoungLee BS MSc

Department of Medicine ampNeurology University ofBritish ColumbiaVancouver Canada

Reviewed and revisedthe manuscript forintellectual content

May H HanMD

Department of Neurology andNeurological Sciences Divisionof Neuroimmunology andMultiple Sclerosis CenterStanford University CA

Reviewed and revisedthe manuscript forintellectual content

RuthJohnson BS

Departments of Neurologyand OphthalmologyUniversity of ColoradoSchool of Medicine Aurora

Reviewed and revisedthe manuscript forintellectual content

DianeIvancicCCRP

Mellen Center for MSTreatment and ResearchNeurological InstituteCleveland Clinic OH

Reviewed and revisedthe manuscript forintellectual content

Eric CKlawiter MD

Department of NeurologyMassachusetts GeneralHospital Harvard MedicalSchool Boston

Reviewed and revisedthe manuscript forintellectual content

AlexandraKocsik BS

Weill Cornell Medicine NewYork NY

Reviewed and revisedthe manuscript forintellectual content

Appendix 2 (continued)

Name Location Contribution

MasonKruse-Hoyer MDMA

Weill Cornell Medicine NewYork NY

Reviewed and revisedthe manuscript forintellectual content

MichaelLevy MDPhD

Department of NeurologyMassachusetts GeneralHospital Harvard MedicalSchool Boston

Reviewed and revisedthe manuscript forintellectual content

Libby LevineRN ANP-BC

Department of NeurologyColumbia University MedicalCenter New York NY

Reviewed and revisedthe manuscript forintellectual content

MelanieMarcille BA

Weill Cornell Medicine NewYork NY

Reviewed and revisedthe manuscript forintellectual content

KatrinaMcMullenPhD

Department of Medicine ampNeurology University of BritishColumbia Vancouver Canada

Reviewed and revisedthe manuscript forintellectual content

Maureen AMealy PhD

Viela Bio Gaithersburg MD(formerly Johns HopkinsUniversity School of MedicineBaltimore MD)

Reviewed and revisedthe manuscript forintellectual content

StephanieMoore MPH

Mellen Center for MSTreatment and ResearchNeurological InstituteCleveland Clinic OH

Reviewed and revisedthe manuscript forintellectual content

Devin SMullin BS

Department of NeurologyBrigham and WomenrsquosHospital Harvard MedicalSchool Boston MA

Reviewed and revisedthe manuscript forintellectual content

Kaho BOnomichiMS

Department of NeurologyColumbia University MedicalCenter New York NY

Reviewed and revisedthe manuscript forintellectual content

Nancy MNealon MD

Weill Cornell Medicine NewYork NY

Reviewed and revisedthe manuscript forintellectual content

Katherine ENelson BA

Department of NeurologyColumbia University MedicalCenter New York NY

Reviewed and revisedthe manuscript forintellectual content

Sarah MPlanchonPhD

Mellen Center for MSTreatment and ResearchNeurological InstituteCleveland Clinic OH

Reviewed and revisedthe manuscript forintellectual content

Ana PruittBS

Department of Neurology KeckSchool of Medicine Universityof Southern California LosAngeles

Reviewed and revisedthe manuscript forintellectual content

Claire SRiley MD

Department of NeurologyColumbia University MedicalCenter New York NY

Reviewed and revisedthe manuscript forintellectual content

Zoe RimlerBS

NYU Langone Health Reviewed and revisedthe manuscript forintellectual content

Andrew WRusso BS

Department of NeurologyMassachusetts GeneralHospital Harvard MedicalSchool Boston

Reviewed and revisedthe manuscript forintellectual content

Julia JSchubert BSBCS

Department of Medicine ampNeurology University ofBritish ColumbiaVancouver Canada

Reviewed and revisedthe manuscript forintellectual content

12 Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 NeurologyorgNN

References1 Wingerchuk DM Lennon VA Lucchinetti CF Pittock SJ Weinshenker BG The

spectrum of neuromyelitis optica Lancet Neurol 20076805ndash8152 Hinson SR Lennon VA Pittock SJ Autoimmune AQP4 channelopathies and neu-

romyelitis optica spectrum disorders Handb Clin Neurol 2016133377ndash4033 Lennon VA Wingerchuk DM Kryzer TJ et al A serum autoantibody marker of neu-

romyelitis optica distinction from multiple sclerosis Lancet 20043642106ndash21124 Lennon VA Kryzer TJ Pittock SJ Verkman AS Hinson SR IgG marker of optic-spinal

multiple sclerosis binds to the aquaporin-4 water channel J ExpMed 2005202473ndash4775 Bennett JL Lam C Kalluri SR et al Intrathecal pathogenic anti-aquaporin-4 anti-

bodies in early neuromyelitis optica Ann Neurol 200966617ndash6296 Asavapanumas N Ratelade J Papadopoulos MC Bennett JL Levin MH Verkman

AS Experimental mouse model of optic neuritis with inflammatory demyelinationproduced by passive transfer of neuromyelitis optica-immunoglobulin GJ Neuroinflammation 20141116

7 Kinoshita M Nakatsuji Y Kimura T et al Neuromyelitis optica passive transfer torats by human immunoglobulin Biochem Biophys Res Commun 2009386623ndash627

8 Weinshenker BG Wingerchuk DM Neuromyelitis spectrum disorders Mayo ClinProc 201792663ndash679

9 Saadoun S Waters P Bell BA Vincent A Verkman AS Papadopoulos MC Intra-cerebral injection of neuromyelitis optica immunoglobulin G and human complementproduces neuromyelitis optica lesions in mice Brain 2010133349ndash361

10 Hinson SR Romero MF Popescu BF et al Molecular outcomes of neuromyelitisoptica (NMO)-IgG binding to aquaporin-4 in astrocytes Proc Natl Acad Sci USA20121091245ndash1250

11 Kitley JWoodhallMWaters P et al Myelin-oligodendrocyte glycoprotein antibodies inadults with a neuromyelitis optica phenotype Neurology 2012791273ndash1277

12 Hamid SHM Whittam D Mutch K et al What proportion of AQP4-IgG-negativeNMO spectrum disorder patients are MOG-IgG positive A cross sectional study of132 patients J Neurol 20172642088ndash2094

13 SatoDKCallegaroD Lana-PeixotoMA et al Distinction betweenMOGantibody-positiveand AQP4 antibody-positive NMO spectrum disorders Neurology 201482474ndash481

14 Wingerchuk DM Banwell B Bennett JL et al International consensus diagnosticcriteria for neuromyelitis optica spectrum disorders Neurology 201585177ndash189

15 Takahashi T Fujihara K Nakashima I et al Anti-aquaporin-4 antibody is involved inthe pathogenesis of NMO a study on antibody titre Brain 20071301235ndash1243

16 Matiello M Lennon VA Jacob A et al NMO-IgG predicts the outcome of recurrentoptic neuritis Neurology 2008702197ndash2200

17 Kang H Chen T Li H Xu Q Cao S Wei S Prognostic factors and disease course inaquaporin-4 antibody-positive Chinese patients with acute optic neuritis J Neurol20172642130ndash2140

18 Pellkofer HL Krumbholz M Berthele A et al Long-term follow-up of patients withneuromyelitis optica after repeated therapy with rituximab Neurology 2011761310ndash1315

19 Cook LJ Rose JW Alvey JS et al Collaborative International Research in Clinical andLongitudinal Experience Study in NMOSD Neurol Neuroimmunol Neuroinflamm20196e583

20 FinakG LangweilerM JaimesM et al Standardizing flow cytometry immunophenotypinganalysis from the Human ImmunoPhenotyping Consortium Sci Rep 2016620686

21 Ding J Zhu DS Hong RH et al The differential expression of natural killer cells inNMOSD and MS J Clin Neurosci 2020719ndash14

22 Ratelade J Asavapanumas N Ritchie AM Wemlinger S Bennett JL Verkman AS In-volvement of antibody-dependent cell-mediated cytotoxicity in inflammatory demyelinationin a mouse model of neuromyelitis optica Acta Neuropathol 2013126699ndash709

23 Lanier LL Le AM Civin CI Loken MR Phillips JH The relationship of CD16 (Leu-11) and Leu-19 (NKH-1) antigen expression on human peripheral blood NK cellsand cytotoxic T lymphocytes J Immunol 19861364480ndash4486

24 Saadoun S Bridges LR Verkman AS Papadopoulos MC Paucity of natural killer andcytotoxic T cells in human neuromyelitis optica lesions Neuroreport 2012231044ndash1047

25 Ratelade J Zhang H Saadoun S Bennett JL Papadopoulos MC Verkman ASNeuromyelitis optica IgG and natural killer cells produce NMO lesions in micewithout myelin loss Acta Neuropathol 2012123861ndash872

26 Vincent T Saikali P Cayrol R et al Functional consequences of neuromyelitis optica-IgG astrocyte interactions on blood-brain barrier permeability and granulocyte re-cruitment J Immunol 20081815730ndash5737

27 Rodriguez-Martin E Picon C Costa-Frossard L et al Natural killer cell subsets incerebrospinal fluid of patients with multiple sclerosis Clin Exp Immunol 2015180243ndash249

28 Zhang B Yamamura T Kondo T Fujiwara M Tabira T Regulation of experimentalautoimmune encephalomyelitis by natural killer (NK) cells J Exp Med 19971861677ndash1687

29 Hatori K Nagai A Heisel R Ryu JK Kim SU Fractalkine and fractalkine receptors inhuman neurons and glial cells J Neurosci Res 200269418ndash426

30 Pan Y Lloyd C Zhou H et al Neurotactin a membrane-anchored chemokineupregulated in brain inflammation Nature 1997387611ndash617

31 Kastenbauer S Koedel U Wick M Kieseier BC Hartung HP Pfister HW CSF andserum levels of soluble fractalkine (CX3CL1) in inflammatory diseases of the nervoussystem J Neuroimmunol 2003137210ndash217

32 Imai T Hieshima K Haskell C et al Identification and molecular characterization offractalkine receptor CX3CR1 which mediates both leukocyte migration and adhe-sion Cell 199791521ndash530

33 Campbell JJ Qin S Unutmaz D et al Unique subpopulations of CD56+ NK and NK-T peripheral blood lymphocytes identified by chemokine receptor expression rep-ertoire J Immunol 20011666477ndash6482

34 Tuo J Smith BC Bojanowski CM et al The involvement of sequence variation andexpression of CX3CR1 in the pathogenesis of age-related macular degenerationFASEB J 2004181297ndash1299

35 Okada K Matsushita T Kira J Tsuji S B-cell activating factor of the TNF family isupregulated in neuromyelitis optica Neurology 201074177ndash178

36 Wang H Wang K Zhong X et al Cerebrospinal fluid BAFF and APRIL levels inneuromyelitis optica and multiple sclerosis patients during relapse J Clin Immunol2012321007ndash1011

37 Lesley R Xu Y Kalled SL et al Reduced competitiveness of autoantigen-engagedB cells due to increased dependence on BAFF Immunity 200420441ndash453

38 Groom J Kalled SL Cutler AH et al Association of BAFFBLyS overexpression andaltered B cell differentiation with Sjogrenrsquos syndrome J Clin Invest 200210959ndash68

39 Pers JO Daridon C Devauchelle V et al BAFF overexpression is associated with auto-antibody production in autoimmune diseases Ann NY Acad Sci 2005105034ndash39

40 Schuh E Berer K Mulazzani M et al Features of human CD3+CD20+ T cellsJ Immunol 20161971111ndash1117

Appendix 2 (continued)

Name Location Contribution

Judy SheardMPH MACCRA

The Guthy-Jackson CharitableFoundation Beverly Hills CA

Reviewed and revisedthe manuscript forintellectual content

Anna JTomczak MS

Department of Neurology andNeurological Sciences Divisionof Neuroimmunology andMultiple Sclerosis CenterStanford University CA

Reviewed and revisedthe manuscript forintellectual content

NeurologyorgNN Neurology Neuroimmunology amp Neuroinflammation | Volume 7 Number 5 | September 2020 13

DOI 101212NXI000000000000085220207 Neurol Neuroimmunol Neuroinflamm

Soumya S Yandamuri Ruoyi Jiang Aditi Sharma et al High-throughput investigation of molecular and cellular biomarkers in NMOSD

This information is current as of August 4 2020

ServicesUpdated Information amp

httpnnneurologyorgcontent75e852fullhtmlincluding high resolution figures can be found at

References httpnnneurologyorgcontent75e852fullhtmlref-list-1

This article cites 40 articles 8 of which you can access for free at

Subspecialty Collections

httpnnneurologyorgcgicollectionautoimmune_diseasesAutoimmune diseasesfollowing collection(s) This article along with others on similar topics appears in the

Permissions amp Licensing

httpnnneurologyorgmiscaboutxhtmlpermissionsits entirety can be found online atInformation about reproducing this article in parts (figurestables) or in

Reprints

httpnnneurologyorgmiscaddirxhtmlreprintsusInformation about ordering reprints can be found online

Academy of Neurology All rights reserved Online ISSN 2332-7812Copyright copy 2020 The Author(s) Published by Wolters Kluwer Health Inc on behalf of the AmericanPublished since April 2014 it is an open-access online-only continuous publication journal Copyright

is an official journal of the American Academy of NeurologyNeurol Neuroimmunol Neuroinflamm