Pediatric cerebrospinal fluid immune profiling distinguishes pediatric-onset multiple sclerosis from other pediatric-onset acute neurological disorders

35 The cerebrospinal fluid (CSF) provides a unique glimpse into the central nervous system (CNS) 36 compartment and offers insights into immune processes associated with both healthy immune 37 surveillance as well as inflammatory disorders of the CNS. The latter include demyelinating 38 disorders, such as multiple sclerosis (MS) and myelin oligodendrocyte glycoprotein antibody -39 associated disease (MOGAD) , that warrant different therapeutic approaches yet are not always 40 straightforward to distinguish on clinical and imaging grounds alone. Here, we establish a 41 comprehensive phenotypic landscape of the pediatric CSF immune compartment across a range of 42 non-inflammatory and inflammatory neurological disorders, with a focus on better elucidating 43 CNS-associated immune mechanisms potentially involved in, and discriminating between, 44 pediatric-onset MS (MS) and other pediatric -onset suspected neuroimmune disorders, including 45 MOGAD. We find that CSF from pediatric patients with non-inflammatory neurological disorders 46 is primarily composed of non-activated CD4+ T cells, with few if any B cells present. CSF from 47 pediatric patients with acquired inflammatory demyelinating disorders is characterized by 48 increased numbers of B cells compared to CSF of both patients with other inflammatory or non -49 inflammatory conditions. Certain features, including particular increased frequencies of antibody-50 secreting cells (ASCs) and decreased frequencies of CD14+ myeloid cells, distinguish MS from 51 MOGAD and other acquired inflammatory demyelinating disorders. 52 53 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint

54 In the non-inflammatory state, the central nervous system (CNS) contains immune cells involved 55 in steady -state immune surveillance 1, while in inflammatory disease states such as multiple 56 sclerosis (MS), the abundance and profile of peripheral immune cells present within the CNS can 57 vary and may include pathogenic immune cells 1,2. Cerebrospinal fluid (CSF), which is produced 58 in the choroid plexus and drains into the subarachnoid granulations, is a fluid that envelops the 59 brain in the subarachnoid space. It thus represents a CNS sub -compartment and may harbor cells 60 that partially mirror disease processes occurring within the CNS tissue. To this end, prior studies 61 have profiled CSF immune cells across neurological diseases including MS 3–6, Alzheimer’s 62 disease7, and brain metastases 8, among others (recently reviewed in 9), to elucidate disease -63 implicated immune mechanisms. However, while CSF cells from adult patients have been profiled 64 extensively 3–6,10–16, there exists limited data on the CSF cellular compartment in the pediatric age 65 group5, whether in health or across neurological disease states. The latter includes pediatric 66 acquired inflammatory demyelinating syndromes (ADS), a collection of disorders in which 67 immune responses are implicated in mediating CNS demyelination. 68 69 Pediatric-onset ADS include pediatric-onset MS (MS) which accounts for approximately 25% of 70 all pediatric ADS diagnoses 17–23; the more recently described MOGAD which accounts for 71 approximately 30% of pediatric ADS diagnoses17; and individuals with ADS who do not meet MS, 72 MOGAD, or neuromyelitis-optica spectrum disorder (NMOSD) diagnostic criteria ( collectively 73 referred to here as ‘other ADS’). MS, and particularly MOGAD, can share considerable overlap in 74 their clinical presentation and disease course (with a relapsing course experienced in some patients 75 with MOGAD). At the time of initial presentation, it may not be straightforward to distinguish the 76 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint two conditions, and results of anti-MOG antibody testing that can be diagnostic for MOGAD may 77 take days to weeks to become available. Like MS, patients with MOGAD can experience relapsing 78 episodes of inflammatory demyelination. However, use of certain MS disease -modifying drugs 79 may not be efficacious in prevention of MOGAD relapses and may even exacerbate its course 24–80 26, pointing to distinct immunopathophysiological mechanisms underlying these two inflammatory 81 demyelinating disorders, and the importance of distinguishing between them. 82 83 Here, we investigate the CSF immune cell compartment across a spectrum of pediatric -onset 84 neurological disorders, including new-onset MS and MOGAD. We first characterize the immune 85 compartment in non-inflammatory CNS conditions across the pediatric age-span. We then identify 86 immune cell features distinguishing ADS from non -inflammatory and other (non-demyelinating) 87 inflammatory disorders and further identify key cell -based immune features distinguishing MS 88 from both MOGAD and other forms of ADS at the time of incident clinical presentation. In 89 addition to establishing a foundational dataset describing the profile and dynamic nature of the 90 cellular composition within non-inflammatory pediatric CSF throughout childhood and 91 adolescence, we provide disease-specific insights into early CNS-associated inflammation in MS, 92 MOGAD, other ADS and other neuroimmune disorders. We expect our findings to represent a 93 useful resource for future studies into the pediatric CNS immune compartment in health and 94 disease. 95 96 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint

97 CSF immune cell profiling in the non-inflammatory pediatric CSF 98 We first established a standardized 16-parameter flow cytometric platform to profile immune cells 99 from fresh pediatric CSF samples obtained at diagnostic lumbar punctures (Fig. 1A, see Methods). 100 This platform enabled us to identify and characterize major cell lineages (e.g. T cell, B cells, 101 myeloid cells), as well as subsets within these lineages, from small volumes of pediatric CSF (0.5-102 7 mL) (Supplementary Fig. 1). 103 104 With this platform established, we next set out to determine the landscape of pediatric CSF across 105 neurological conditions that are not considered primarily inflammatory in nature (i.e. non -106 inflammatory neurological disorders, or NIND, which include disorders such as primary headaches 107 and idiopathic intracranial hypertension, among others (Supplementary Table 1). To this end, we 108 analyzed the CSF flow cytometric profiles of 33 patients with subsequently ascertained diagnoses 109 of NIND (Table 1). NIND CSF cell concentrations ranged from 0 .2-6.0 cells/uL (median 0.63 110 cells/uL, Fig. 1B). The most abundant cell types within the cell fraction of the NIND CSF were 111 CD3+ T cells ( Fig. 1C, median 67.3%, CD45+/CD14 -/CD19-/CD3+) and CD14+ myeloid cells 112 (median 12.9%, CD45+/CD14+). Natural killer (NK cells 4.3%, CD45+/CD14 -/CD3-/CD19-113 /CD56+), dendritic cells (DCs 1.6%, CD45+/CD14-/CD3-/CD19-/CD56-/HLA-DR+/CD11c+), B 114 cells ( 0.6%, CD45+/CD14-/CD3-/CD19+) and polymorphonuclear cells (PMNs 0.3%, 115 CD45+/SSChi, inclusive of neutrophils, eosinophils, and basophils ) were detected at lower 116 frequencies. 117 118 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint Within the CD3+ T-cell compartment of NIND CSF ( Fig. 1D), most cells were CD4+ (median 119 67.2%), whereas CD8+ (median 23.6%) and double-negative (CD4-/CD8-; median 8.4%) T cells 120 were less frequent . Double-positive (CD4+/CD8+) T cells were rarely observed (median 0.7%). 121 Most CD4+ and CD8+ T cells were memory T cells (based on CD27 and CD45RA expression, 122 median 97.8% and 77.0%, respectively, Fig. 1E-F), with larger variability in memory CD8+ T cell 123 frequencies across donors . This variability appeared to be explained, at least in part, by a clear 124 correlation between age and frequency of naive CD8+ T cells, such that youngest children with 125 NIND had high frequencies of naive CD8+ T cells, with gradual increases in the frequency of 126 memory CD8+ T cells observed with increasing age ( Fig. 1G ). Total CD8+ T -cell counts 127 (cells/mL) remained stable over time (Supplementary Fig. 2A), indicating an age-associated shift 128 in the balance of naive and memory CD8+ T cells in the non-inflamed pediatric CSF. We did not 129 observe an age-associated pattern for naive and memory CD4+ T cells (Supplementary Fig. 2B). 130 131 Further assessment of the memory CD4+ and CD8+ T-cell compartments in patients with NIND 132 revealed that memory CD4+ T cells were mostly CD27+/CD45RA -, consistent with a central -133 memory (CM) phenotype (Supplementary Fig. 2C) with the remainder being primarily effector-134 memory (EM) cells (CD27 -/CD45RA-), and few CD4+ TEMRA cells (CD45RA+/CD27 -) 135 detected (Supplementary Fig. 2C). Most memory CD4+ T cells in NIND CSF were Th1-like 136 (CXCR3+/CCR6-, Supplementary Fig. 2D ). Few memory CD4+ T cells were activated 137 (CD38+/HLA-DR+, Supplementary Fig. 2E). Similarly, most memory CD8+ T cells in the NIND 138 patients had a CM phenotype (Supplementary Fig. 2F), and activated memory CD8+ T cells 139 made up a minority of the memory CD8+ T-cell compartment (Supplementary Fig. 2G). 140 141 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint Overall, pediatric NIND CSF was as expected of generally low cellularity, with most leukocytes 142 being CD4+ T cells, most of which were memory T cells with a Th1-like phenotype and non -143 activated. CD8+ T cells were similarly mostly memory T cells and non-activated. We found that 144 younger age correlated with an increased frequency of naive CD8+, but not CD4+, T cells. 145 146 Elevations in B-cell counts highlight the increased cellularity of pediatric ADS CSF 147 Having established the CSF immune cell landscape in pediatric NIND, we next compared CSF 148 cellular profiles of children with ADS (n = 33) to children with both NIND and a range of other 149 inflammatory and non -inflammatory neurological disorders (Table 1, further detail in 150 Supplementary Table 1 ). These disorders included peripheral inflammatory neurological 151 disorders (PIND, n=7, which include inflammatory demyelinating disorders of the peripheral 152 nervous system such as Guillain -Barre syndrome), autoimmune encephalitidies (AIE n = 6, , 153 autoimmune inflammatory disorders of the CNS that are not necessarily demyelinating, such as 154 NMDA-receptor encephalitis), and inherited disorders of white matter (IDWM, n = 4, genetic 155 disorders resulting in abnormal growth or destruction of CNS white matter). 156 157 We first compared the absolute CSF cell counts of all children with ADS to children with NIND, 158 PIND, AIE, and IDWM. We observed a statistically significant increase in the cellularity of ADS 159 CSF compared to NIND, PIND, and AIE CSF (Fig. 2A ). We next examined how individual 160 lineages contributed to the increased cellularity in ADS CSF. We found that counts of total CD3+ 161 T cells, CD14+ myeloid cells, NK cells, DCs, PMNs, and B cells showed statistically significant 162 increases in ADS CSF when compared to NIND CSF (Fig 2B). Similarly, cell counts for these 163 lineages tended to be increased when comparing ADS CSF to PIND, AIE, and IDWM CSF, though 164 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint differences did not reach statistical significance for all comparisons (Fig. 2B). Of all examined 165 lineages, the B-cell lineage exhibited the greatest relative enrichment in ADS CSF, with a greater 166 than tenfold increase of B-cell counts in ADS CSF relative to B-cell counts in NIND, PIND, AIE, 167 or IDWM CSF (Fig. 2C). 168 169 MS CSF is distinguished from MOGAD and other ADS by a high frequency of antibody 170 secreting cells and a decreased frequency of CD14+ myeloid cells 171 Having identified a pronounced B-cell enrichment in the CSF of children with ADS (Fig. 2B-C), 172 we next assessed individual ADS diagnoses ( MS, MOGAD, and other ADS). We observed that 173 CSF B-cell counts were increased in each of MS, MOGAD, and other ADS CSF when 174 independently compared to NIND CSF ( Supplementary Table 2). Among these, B-cell 175 frequencies were particularly elevated in MS CSF in comparison to MOGAD and other ADS CSF 176 (Fig. 3A); B-cell absolute counts also appeared elevated in MS CSF in comparison to MOGAD 177 and other ADS CSF, though this comparison did not reach statistical significance for MS vs. 178 MOGAD, in part likely due to the large range of CSF B-cell and total cell counts observed in 179 MOGAD patients (Fig. 3A, Supplementary Fig. 3A). We next sought to determine whether levels 180 of antibody-secreting cells (ASCs), a B -cell subset that includes plasmablasts and plasma cells , 181 differed between MS, MOGAD, and other ADS CSF. We found that both ASC frequencies and 182 counts were increased in MS CSF when compared to MOGAD and other ADS (Fig. 3B). In fact, 183 ASCs were frequently absent in MOGAD (absent in 5/10 samples) or other ADS CSF (absent in 184 5/10 samples) while almost always detected in MS CSF (Fig. 3B). Finally, frequencies and counts 185 of B cells excluding ASCs (non -ASC B cells) appeared elevated in MS when compared to 186 MOGAD and other ADS (Supplementary Fig. 3B-C). Together these findings point to elevated 187 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint B-cell and particularly ASC levels as a feature distinguishing MS CSF not only from non -188 inflammatory disease controls but also from MOGAD and other ADS (Fig. 3D). 189 190 We further observed that frequencies of CSF CD14+ myeloid cells were lower in CSF of MS 191 compared to both MOGAD and other ADS patients (Fig. 3C-D), which appeared to reflect 192 elevations in CD14+ myeloid cell counts in the CSF of MOGAD and the other ADS patients, 193 relative to NIND CSF (Fig. 3C, Supplementary Table 2). 194 195 CD3+ T-cell frequencies were enriched in MS CSF relative to MOGAD and other ADS CSF 196 (Supplementary Fig. 3B), though no particular subset of T-cells clearly contributed this difference 197 (Supplementary Fig. 5A-B). NK cell and DC frequencies and counts appeared to be similar 198 across the three groups ( Supplementary Fig. 3B-C). PMN frequencies appeared, at times, 199 elevated in MOGAD and other ADS CSF, while largely absent in MS (Supplementary Fig. 3B). 200 To explore this phenomenon further, we set a threshold (defined as >2 times the mean PMN 201 frequency in NIND CSF ) and assessed whether the presence of CSF PMNs above this threshold 202 was associated with MOGAD/other AD S compared to MS (Supplementary Fig. 3D). Indeed, 203 detection of CSF PMNs above this threshold was associated with a diagnosis of other 204 ADS/MOGAD and essentially excluded MS (Supplementary Fig. 3E). 205 206 Potential for the ASC to CD14+ myeloid cell ratio to distinguish MS CSF from MOGAD and 207 other ADS CSF 208 Since an increase in ASC frequencies and a decrease in CD14+ myeloid cell frequencies appeared 209 to best differentiate between CSF of MS versus MOGAD or other ADS patients ( Fig. 3D, 210 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint Supplementary Fig. 4B), we considered whether the ratio of ASC to CD14+ myeloid cell 211 frequencies could be utilized to discriminate MS from MOGAD and other ADS, at the individual 212 patient level. As expected, we found that the CSF ASC to CD14+ myeloid cell ratios (AMRs) were 213 elevated in children with MS, both in comparison to MOGAD and to other ADS patients (p < 214 0.001, Fig. 4A ). A previously reported ratio, termed the neuroinflammatory composite score 6 215 (NCS), similarly incorporated several cellular measures, using both CSF (B-cell, CD14+ myeloid 216 cell, CD56+ CD3+ NKT -cell frequencies and total cell counts) and blood (CD56 dim NK-cell 217 frequencies) to distinguish between neuroinflammatory diagnoses (high composite scores), and 218 non-inflammatory controls (lower composite scores). We utilized a CSF-only version of the NCS 219 (coNCS, see Methods) and similarly found that this coNCS was also elevated in patients with MS, 220 in comparison to MOGAD and other ADS (Fig. 4B). We found similar AMR and coNCS patterns 221 distinguishing between patients who had received systemic glucocorticoid and/or intravenous 222 immunoglobulin (IVIG) treatment prior to sampling versus those who had not (Supplementary 223 Fig. S5A-B). We next assessed the performance of the AMR and the coNCS as classifiers for MS, 224 when compared to either MOGAD, other ADS, or both by constructing receiver-operating 225 characteristic (ROC) curves and found that both the AMR and coNCS performed well at 226 discriminating MS from MOGAD and other ADS, with AUCs > 0.7 (Fig. 4C). The AMR, however, 227 appeared superior to the coNCS in discriminating MS vs MOGAD (AUC=0.95, 0.80, respectively) 228 and MS vs MOGAD/other ADS (AUC=0.94, 0.88, respectively). The AMR also provided 229 excellent performance at distinguishing MS from other ADS (AUC=0.93), though the coNCS 230 outperformed the AMR in this regard. (AUC=0.96). We were able to generate the full NCS in a 231 subset of patients for whom we also had available blood CD56dim NK -cell frequencies and 232 similarly observed that the full NCS was higher in MS relative to MOGAD and other ADS 233 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint (Supplementary Fig. 6A) and overall performed similarly to the AMR (Supplementary Fig. 6B). 234 Finally, we also evaluated the NCS companion “MS score”, which incorporates ASC presence and 235 IgG synthesis rates . We found that it performed relatively poorly in our dataset (correctly 236 classifying only 7 of 15 MS diagnoses ; Supplementary Table 3). Overall, the two -parameter 237 AMR we describe here provided excellent performance at discriminating MS CSF from MOGAD 238 and other ADS CSF and performed at least as well as a previously reported neuro-immunological 239 classifier reliant on additional parameters (NCS). 240 241 Interactive exploration of pediatric CSF immune profiles 242 We next developed a public, web-based R-Shiny application for further analysis and visualization 243 of pediatric CSF immune -profiles across a broader range of disease states 244 (https://diegoalexespi.shinyapps.io/pedscsf/). By extending our CSF profiling to patients with viral 245 encephalitis (VE) and patients with CNS -involved hemophagocytic lymphohistiocytosis (HLH), 246 for example, we are able to observe a breadth of CD8+ T -cell activation states ( Supplementary 247 Fig. S7), such that patients with VE and HLH harbored elevated levels of CD8+ T-cell activation 248 that are rarely observed in other disease states including other ADS , MOGAD, and MS 249 (Supplementary Fig. S7). Such exploratory analyses can provide preliminary insight into further 250 studies of the CSF immune compartment across disease states. 251 252 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint

253 Data on CSF cellular immune profiles in pediatric neurological disorders has been exceptionally 254 limited. Here, we present a flow cytometric characterization of CSF immune cells across a 255 spectrum of pediatric neurological disorders, identifying age -associated profiles in non -256 inflammatory neurological diseases that likely reflect steady-state immune surveillance, as well as 257 disease-specific cellular signatures across a range of inflammatory states. We find that under 258 normal, non-inflammatory conditions, pediatric CSF is not entirely acellular, with small numbers 259 of immune cells predominantly composed of T cells, and with a CD8+ T-cell compartment that 260 evolves with age. B cells are scarce or entirely absent in the CSF of non-inflammatory states while 261 the CSF of children diagnosed with a range of ADS, exhibit notable increases in the number of B 262 cells present. Among ADS presentations, a distinguishing feature of the CSF immune cell profile 263 of children with MS compared to MOGAD and other ADS, is an increased frequency of ASCs and 264 a decreased of frequency of CD14+ myeloid cells. Overall, these data represent the first 265 characterization of pediatric CSF across both noninflammatory states and inflammatory 266 neurological disorders and provides insight into normal immune surveillance as well as putative 267 mechanisms of pediatric CNS disease. 268 269 Our characterization of the CSF profile across the pediatric age-span in non -inflammatory 270 disorders provides us insight into age -associated patterns of CNS immune surveillance in the 271 immunological steady-state. We find that most cells participating in CNS immune surveillance are 272 CD4+ CM T cells across the pediatric age -span. These findings in the pediatric population are 273 overall consistent with prior reports in adults that identified CM CD4+ and CD8+ T cells and 274 CXCR3+ T cells as the primary cellular components of the NIND CSF immune compartment, with 275 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint minimal to no B cells detected 27–29. We extend these observations by finding that, across the 276 pediatric age -span, CD8+ T cells shift from a naive (antigen -inexperienced) to memory 277 (presumably antigen-experienced) phenotype with increasing age . This trend aligns with the 278 observation that adult NIND CSF is largely devoid of naive CD8+ T cells 27. Since the migration 279 of immune cells from the periphery into the CNS is a regulated process and not a simple reflection 280 of the periphery (as both CD4+ and CD8+ T cells are primarily naive throughout childhood30), this 281 observation implies that naive CD8+ T cells earlier in life are endowed with a capacity to migrate 282 into the CNS. One explanation is that these naive CD8+ T cells are “tissue-residency poised31” and 283 mature into CD8+ tissue-resident memory T cells within the CNS that eventually saturate the CNS 284 CD8+ tissue resident T-cell niche. Most parenchymal T cells in human brain (across health and 285 disease states ) appear to be CD8+ rather than CD4+, and express tissue residency markers 32, 286 consistent with the above phenomenon observed in CD8+ but not CD4+ T cells. It is also possible 287 that what we observe as “naive” CD8+ T cells in the CSF, instead reflect a collection of antigen-288 experienced memory CD8+ T cells that share markers with naive CD8+ T cells33,34. In any event, 289 it is evident that there exists an evolving recruitment of CD8+ T-cells of different phenotypes into 290 the CNS across the pediatric age span, likely reflecting in part tandem immunological changes in 291 CD8+ T-cell compartments beyond the CNS35–37. 292 293 We observe a significant elevation in CSF B-cell counts in children with ADS, beyond elevations 294 observed for other lineages, and relative to both non-inflammatory and other inflammatory states. 295 This elevation of B cells in ADS CSF could reflect a trafficking of peripheral B cells into the CNS 296 where they may then participate in mechanisms associated with inflammatory demyelination; B 297 cells are indeed present both in MS and MOGAD brain lesions 38,39. The observation that B-cell 298 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint numbers are elevated in both MS and MOGAD CSF does not necessarily mean they participate in 299 these diseases through the same mechanism. In fact, B-cell depleting therapies, which are highly 300 efficacious in preventing relapse activity in MS40–42, have had mixed results in preventing relapse 301 activity in MOGAD25,26. Thus, it is plausible that these cells play distinct functional roles in 302 different inflammatory demyelinating disease processes. Future studies could further delineate 303 whether and how B cells in MS and MOGAD CSF differ with regards to, for example, their pro- 304 and/or anti-inflammatory functions. 305 306 We found a particular enrichment of ASCs (the antibody -secreting B -cell subset) in MS CSF 307 compared to other ADS and MOGAD. While ASC elevations have previously been described in 308 MS CSF, prior studies have typically compared MS to non-inflammatory controls rather than to 309 other inflammatory states 4,43–45. Here, our direct comparisons with MOGAD and other ADS 310 allowed us to demonstrate that increased presence of ASCs is a relatively distinct feature of MS 311 CSF, rather than a feature of all ADS CSF. While peripheral ASCs may generate circulating 312 antibodies to CNS proteins in certain ADS, such as the generation of circulating antibodies to 313 MOG in MOGAD, the contribution of ASCs to disease in the MS CNS has not been fully 314 elucidated and their function within the CNS may be distinct from that of peripheral ASCs . Of 315 interest is whether and how the enrichment of ASCs in MS CSF may be associated somehow with 316 presence (or possibly generation of) meningeal lymphoid aggregates, a feature also reported in the 317 brains of patients with M S46 but not in MOGAD or other AD S. These aggregates , which are 318 composed of T cells, B cells, plasma cells, and myeloid cells, are thought to contribute to chronic 319 CNS-compartmentalized inflammation and progressive disease47,48, also a feature of MS, but not 320 MOGAD49. If the presence of these ASCs in pediatric MS CSF were indeed associated with the 321 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint presence or development of meningeal lymphoid aggregates, this could imply that the cellular 322 machinery and anatomical structures associated with chronic CNS -compartmentalized 323 inflammation in MS may already be developing very early in the MS disease spectrum. 324 325 We also identify decreased numbers of CD14+ myeloid cells in the CSF (and a concomitant 326 increased ratio of ASCs to CD14+ myeloid cells) as a feature distinguishing MS from MOGAD 327 and other ADS. Further work could dissect the pro/anti-inflammatory states of myeloid cells in the 328 CSF of ADS patients including MOGAD, and assess their potential interactions with B-cell 329 subsets, as such interactions are implicated in the context of CNS inflammatory disease 48. 330 Regardless of the processes that underlie the different CSF ratios of ASC to CD14+ myeloid cells, 331 the ratio in our cohort distinguishes between MS and non-MS ADS including MOGAD and could 332 be further investigated for its potential to provide an MS disease-specific CSF signature at time of 333 initial CSF evaluation. The utility of this ratio may prove complementary to the previously 334 developed CSF and blood “neuroinflammatory composite score (NCS)” that has been used in 335 adults to distinguish between neuroinflammatory syndromes and non-inflammatory syndromes6. 336 337 We note that CSF CD4+ T cell subset profiles did not clearly distinguish between MS and other 338 ADS or MOGAD. While prior reports in adults identified elevated frequencies of CD4+ regulatory 339 T cells (Tregs) in MS CSF when compared to idiopathic intracranial hypertension (IIH) 340 controls4,50, we did not find elevated Treg frequencies in MS CSF when compared to either NIND 341 or to MOGAD or other ADS. Plausible explanations for this discrepancy include the more 342 heterogenous cohort of NINDs included in our study and the differing ages of the patients across 343 studies. While T cells observed in white matter lesions of MOGAD39 have been shown to be 344 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint primarily CD4+ (in contrast to MS lesions, which are CD8+ T-cell dominant38,51), we did not find 345 differences in CSF frequencies of CD4+ T cells or CD8+ T cells between MS and MOGAD 346 patients. 347 348 While we were able to identify a CSF immune profile that distinguishes MS from MOGAD (and 349 other ADS), we did not identify a unique CSF immune profile that clearly distinguished MOGAD 350 patients. In prior studies52,53 PMNs (i.e. neutrophils or eosinophils) have been found to be elevated 351 in MOGAD CSF . Here we identify that elevated frequencies of PMNs (while potentially 352 exclusionary for diagnosis of MS), is a potential feature of both MOGAD and other ADS . The 353 presence of PMNs in the CSF has been historically associated with anti -pathogen responses 354 (typically to bacterial or also viral species) during CNS infection, yet the development of MOGAD 355 or other ADS have not been firmly linked to a specific viral infection (including EBV 54). It is 356 plausible that the presence of PMNs in select MOGAD and other ADS CSF samples could hint at 357 some anti-pathogen (potentially anti -viral) immune processes that, in parallel, also manifest as 358 anti-self, inflammatory CNS demyelination processes. 359 360 Our study has several limitations. While we profile d children as proximal as possible to clinical 361 disease onset, the underlying disease processes may already be well underway prior to clinical 362 manifestations, which may be particularly true in MS where evidence for injury manifesting as 363 elevated serum neurofilament light chain (NfL) levels is documented well prior to clinical disease 364 onset55. Another limitation is absence of CSF profiles from children with neuromyelitis optica 365 spectrum disorder (NMOSD), which is in the differential diagnosis of children with incident ADS 366 presentations. Even though it is relatively uncommon in the pediatric age -range, NMOSD is 367 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint thought to involve pathogenic antibodies, and it will be important to assess in the future whether 368 CSF ASC levels are elevated or not and how the ASC to CD14+ myeloid cell ratio manifests in 369 such children. Given that relapsing disease is a shared feature between NMOSD and MS (and some 370 children with MOGAD), defining these profiles in NMOSD could provide additional insight into 371 the pathophysiology of antibody-mediated ADS and disease chronicity. Finally, our cohort consists 372 of children who underwent diagnostically-necessary lumbar punctures, which may not be wholly 373 representative of all MS, MOGAD, or other ADS diagnoses. The diagnostic criteria for these 374 disorders do not always necessitate CSF analyses, and the set of children who do undergo lumbar 375 punctures may be enriched for those whose presentations elicit diagnostic uncertainty. Clear 376 immune profile patterns nevertheless emerged in our data, such as those clearly separating MS 377 from MOGAD and other ADS. 378 379 In sum, our findings provide insights into CSF immune cell profiles in the context of both normal 380 immune surveillance and CNS non-inflammatory and inflammatory diseases, across the pediatric 381 age-span. Our findings identify CSF cellular features that distinguish pediatric-onset MS from 382 MOGAD and other ADS and may imply that some features of chronic inflammation are present 383 early in the disease course. While no cellular features clearly delineated MOGAD from other ADS, 384 the profiles observed in MOGAD and other ADS, as well as in MS, could reflect biological 385 heterogeneity that may have implications for a patient’s clinical course . For example, a patient’s 386 responsiveness to treatment(s) and risk of future relapse. Future studies could determine such 387 “endophenotypes” by acquiring these CSF measures at the time of presentation and associating 388 them with prospectively ascertained diagnoses, as has been done for peripheral blood profiles in 389 adult MS56. As flow cytometric profiling of CSF samples becomes more widely available (driven 390 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint by newly-developed CSF fixation protocols leveraged for multicenter studies 57,58), CSF immune 391 profiling has the potential to play a n important complementary role in providing improved 392 diagnosis and hence care of patients as part of a larger precision neuroimmunology approaches59. 393 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint

394 Patient eligibility and recruitment 395 Patients were recruited over a period of 3 years from inpatient and outpatient facilities at the 396 Children’s Hospital of Philadelphia (CHOP). Patients were eligible and approached for enrollment 397 (subject to staff availability) if they were undergoing a diagnostic lumbar puncture (LP) for a 398 suspected neurological condition . Written informed consent was obtained from all donors. All 399 protocols and consents were approved by the University of Pennsylvania Institutional Review 400 Board. 401 402 Acquisition, transport, and analysis of CSF samples 403 A total of 0.5-7 mLs of CSF per participant were obtained. CSF samples were collected in 15 mL 404 polypropylene tube s at the time of LP, placed on ice and immediately transported to a single 405 laboratory facility for prompt same-day processing and analysis. CSF from one HLH patient was 406 collected from an external ventricular drain. Standardized operating procedures were consistent 407 with consensus protocols for CSF analysis60. 408 409 Flow cytometry of CSF cells 410 Upon arrival to the laboratory, 20 uL of CSF was placed on the Blood/Hemoglobin region of a 411 Chemstrip (Chemstrip 10 SG, Cat. No. 11895362160) to evaluate for blood and/or hemoglobin 412 contamination in the CSF sample. Samples with Chemstrip readings of ≥50 Erythrocytes/uL using 413 either the blood or hemoglobin reading were not evaluated further. Evaluable samples were 414 immediately spun at 4º C for 10 minutes at 400 RCF. CSF supernatant was then removed from the 415 cell pellet, and the cell pellet was resuspended in 200 uL cold 1X PBS. 10 uL of sample was added 416 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint to 10 uL of Methylene Blue and utilized for cell counting using a hemocytometer. Following cell 417 counting, 800 uL of cold 1X PBS was added to the sample, and the sample was spun again at 4º C 418 for 10 minutes at 400 RCF. Supernatant was discarded and the cell pellet was resuspended in 100 419 uL of cold 1X PBS for flow cytometry staining with a cocktail of 16 antibodies targeting CD45, 420 CD14, CD3, CD19, CD56, HLA-DR, CD4, CD8, CD38, CD27, CD45RA, CD11c, CD127, CD25, 421 CXCR3, and CCR6 ( Supplementary Table 4) for 30 minutes at 4ºC protected from light. 422 Following the 30-minute stain, 900 uL of 1X cold PBS was added to the sample before another 423 spin at 4º C for 10 minutes at 400 RCF. Supernatant was discarded and cell pellet was resuspended 424 in 200-300 uL of 1X cold PBS and placed on ice prior to immediate analysis on an LSR Fortessa. 425 Routinely, paired whole blood was stained and run on the Fortessa side-by-side with the CSF. BD 426 FACSDiva software was utilized to collect flow cytometry data on the LSR Fortessa. Single-stain 427 compensation tubes with beads were run prior to data collection and utilized for compensation on 428 FlowJo software. After gating, frequencies were exported from FlowJo into .csv files for 429 downstream statistical analysis. 430 431 Statistical analysis 432 Statistical analysis was performed utilizing R software. When statistical comparisons were 433 performed between conditions, a Wilcoxon -rank-sum test was utilized. The R packages dplyr, 434 ggplot2, magrittr, tibble, tidyr, ggpubr, pROC, and cowplot were utilized for statistical analysis and 435 figure generation. 436 437 Neuroinflammatory composite score 438 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint A CSF-only neuroinflammatory composite score (coNCS, adapted from 6) was calculated with the 439 following formula and used in Fig. 4B: 440 !"𝑐𝑒𝑙𝑙𝑠 𝑢𝐿 + 1+ ∗ 𝐵 𝑐𝑒𝑙𝑙𝑠 ∗ 100 0 /[𝑀𝑜𝑛𝑜𝑐𝑦𝑡𝑒𝑠 ∗ 𝑁𝐾𝑇 𝑐𝑒𝑙𝑙𝑠] 441 Where cells/uL = CSF cell concentration calculated from hemocytometer manual count, B cells = 442 CSF B cell frequency as % of CSF lymphocytes, Monocytes = CSF CD14+ myeloid cell frequency 443 as % of CSF cells, and NKT cells = CSF CD56+ CD3+ T-cell frequency as % of CSF lymphocytes. 444 If either the denominator or numerator were 0, the coNCS was set to 0. For Supplementary Fig. 445 S6B, the full NCS (with blood CD56dim NK) was utilized when blood CD56dim NK cell 446 frequencies were available: 447 !"𝑐𝑒𝑙𝑙𝑠 𝑢𝐿 + 1+ ∗ 𝐵 𝑐𝑒𝑙𝑙𝑠 ∗ 100 0 /[𝐶𝐷56𝑑𝑖𝑚 𝑁𝐾𝑐𝑒𝑙𝑙𝑠!"##$ ∗ 𝑀𝑜𝑛𝑜𝑐𝑦𝑡𝑒𝑠 ∗ 𝑁𝐾𝑇 𝑐𝑒𝑙𝑙𝑠] 448 Where CD56dim NK cellsblood = blood CD56dim NK cell frequency as % of blood lymphocytes 449 and all other parameters identical to the coNCS. If either the denominator or numerator were 0, 450 the full NCS was set to 0. 451 452 Patient diagnoses and categorical classification 453 Patient diagnoses were ascertained in prospective follow -up by pediatric neurologists with 454 extensive experience in pediatric neuroimmunology, and based on all available clinical and clinical 455 laboratory data (but blinded to the CSF flow cytometry results). When possible, these diagnoses 456 were then binned into five categories: non -inflammatory neurological disorders (NINDs), 457 peripheral inflammatory neurological disorders (PIND), autoimmune encephalitidies (AIE), 458 inherited disorders of white matter (IDWM) . For acquired demyelinating syndromes (ADS) , 459 diagnoses were further characterized by presenting phenotype . The classification of specific 460 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint diagnoses into categories is available within Supplementary Table 1. Within the ADS category, 461 MOGAD was diagnosed utilizing the 2023 International MOGAD Panel proposed criteria 61 and 462 MS was diagnosed utilizing the 2017 McDonald criteria62. For children diagnosed with other ADS 463 or MOGAD, CSF samples were determined to have been procured with a median of 4.5 days or 464 12 days, respectivley, from the onset of most recent episode of clinically active disease 465 (Supplementary Table 5). For children diagnosed with MS, 4 of 15 were found to be in remission 466 at the time of sampling, defined as the absence of clinical symptoms for at least 3 months prior to 467 LP (Supplementary Table 5); for those with active clinical disease, CSF samples were determined 468 to have been procured with a median of 6 days since the onset of most recent episode of clinically 469 active disease (Supplementary Table 5). In Supplementary Fig. S5, other ADS, MOGAD, and 470 CSF samples were further classified as treated if they had received any systemic glucocorticoid 471 and/or IVIG treatment within the last 30 days prior to LP. Children with other ADS, MOGAD, or 472 MS with any history of MS disease-modifying therapies were excluded from all analyses. 473 474 475 476 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint DATA A V AILABILITY AND REPRODUCIBILITY 477 Complete R code (including R markdown notebooks with clear documentation) and final tables 478 used to analyze and visualize our results have been deposited at 479 https://github.com/diegoalexespi/espinozada_pediatricCSF2024. 480 481 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint

482 1. Ousman SS, Kubes P. Immune surveillance in the central nervous system. Nat Neurosci. 483 2012;15(8):1096-1101. doi:10.1038/nn.3161 484 2. Stangel M, Fredrikson S, Meinl E, Petzold A, Stüve O, Tumani H. The utility of cerebrospinal 485 fluid analysis in patients with multiple sclerosis. Nat Rev Neurol. 2013;9(5):267-276. 486 doi:10.1038/nrneurol.2013.41 487 3. Ostkamp P, Deffner M, Schulte-Mecklenbeck A, et al. A single-cell analysis framework 488 allows for characterization of CSF leukocytes and their tissue of origin in multiple sclerosis. 489 Science Translational Medicine. 2022;14(673):eadc9778. doi:10.1126/scitranslmed.adc9778 490 4. Schafflick D, Xu CA, Hartlehnert M, et al. Integrated single cell analysis of blood and 491 cerebrospinal fluid leukocytes in multiple sclerosis. Nature Communications. 492 2020;11(1):247. doi:10.1038/s41467-019-14118-w 493 5. Han S, Lin YC, Wu T, et al. Comprehensive Immunophenotyping of Cerebrospinal Fluid 494 Cells in Patients with Neuroimmunological Diseases. The Journal of Immunology. 495 2014;192(6):2551-2563. doi:10.4049/jimmunol.1302884 496 6. Gross CC, Schulte-Mecklenbeck A, Madireddy L, et al. Classification of neurological 497 diseases using multi-dimensional CSF analysis. Brain. 2021;144(9):2625-2634. 498 doi:10.1093/brain/awab147 499 7. Gate D, Saligrama N, Leventhal O, et al. Clonally expanded CD8 T cells patrol the 500 cerebrospinal fluid in Alzheimer’s disease. Nature. 2020;577(7790):399-404. 501 doi:10.1038/s41586-019-1895-7 502 8. Rubio-Perez C, Planas-Rigol E, Trincado JL, et al. Immune cell profiling of the cerebrospinal 503 fluid enables the characterization of the brain metastasis microenvironment. Nat Commun. 504 2021;12(1):1503. doi:10.1038/s41467-021-21789-x 505 9. Heming M, Börsch AL, Wiendl H, Meyer zu Hörste G. High-dimensional investigation of the 506 cerebrospinal fluid to explore and monitor CNS immune responses. Genome Med. 507 2022;14(1):1-12. doi:10.1186/s13073-022-01097-9 508 10. Pappalardo JL, Zhang L, Pecsok MK, et al. Transcriptomic and clonal characterization of T 509 cells in the human central nervous system. Science Immunology. 2020;5(51):eabb8786. 510 doi:10.1126/sciimmunol.abb8786 511 11. Ramesh A, Schubert RD, Greenfield AL, et al. A pathogenic and clonally expanded B cell 512 transcriptome in active multiple sclerosis. Proceedings of the National Academy of 513 Sciences. 2020;117(37):22932-22943. doi:10.1073/pnas.2008523117 514 12. Jacobs BM, Gasperi C, Kalluri SR, et al. Single-cell analysis of cerebrospinal fluid reveals 515 common features of neuroinflammation. Cell Reports Medicine. Published online December 516 20, 2024:101733. doi:10.1016/j.xcrm.2024.101733 517 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 13. Cantoni C, Smirnov RA, Firulyova M, et al. A single-cell compendium of human 518 cerebrospinal fluid identifies disease-associated immune cell populations. J Clin Invest. 519 2025;135(1). doi:10.1172/JCI177793 520 14. Heming M, Börsch AL, Melnik S, et al. Atlas of Cerebrospinal Fluid Immune Cells Across 521 Neurological Diseases. Ann Neurol. Published online December 12, 2024. 522 doi:10.1002/ana.27157 523 15. Ban M, Bredikhin D, Huang Y, et al. Expression profiling of cerebrospinal fluid identifies 524 dysregulated antiviral mechanisms in multiple sclerosis. Brain. 2024;147(2):554-565. 525 doi:10.1093/brain/awad404 526 16. Lepennetier G, Hracsko Z, Unger M, et al. Cytokine and immune cell profiling in the 527 cerebrospinal fluid of patients with neuro-inflammatory diseases. Journal of 528 Neuroinflammation. 2019;16(1):219. doi:10.1186/s12974-019-1601-6 529 17. Waters P, Fadda G, Woodhall M, et al. Serial Anti–Myelin Oligodendrocyte Glycoprotein 530 Antibody Analyses and Outcomes in Children With Demyelinating Syndromes. JAMA 531 Neurology. 2020;77(1):82-93. doi:10.1001/jamaneurol.2019.2940 532 18. Langer-Gould A, Zhang JL, Chung J, Yeung Y, Waubant E, Yao J. Incidence of acquired 533 CNS demyelinating syndromes in a multiethnic cohort of children. Neurology. 534 2011;77(12):1143-1148. doi:10.1212/WNL.0b013e31822facdd 535 19. Ketelslegers IA, Catsman-Berrevoets CE, Neuteboom RF, et al. Incidence of acquired 536 demyelinating syndromes of the CNS in Dutch children: a nationwide study. J Neurol. 537 2012;259(9):1929-1935. doi:10.1007/s00415-012-6441-6 538 20. Banwell B, Bar-Or A, Arnold DL, et al. Clinical, environmental, and genetic determinants of 539 multiple sclerosis in children with acute demyelination: a prospective national cohort study. 540 The Lancet Neurology. 2011;10(5):436-445. doi:10.1016/S1474-4422(11)70045-X 541 21. Tantsis EM, Prelog K, Brilot F, Dale RC. Risk of multiple sclerosis after a first demyelinating 542 syndrome in an Australian Paediatric cohort: clinical, radiological features and application of 543 the McDonald 2010 MRI criteria. Mult Scler. 2013;19(13):1749-1759. 544 doi:10.1177/1352458513484377 545 22. Mikaeloff Y, Adamsbaum C, Husson B, et al. MRI prognostic factors for relapse after acute 546 CNS inflammatory demyelination in childhood. Brain. 2004;127(9):1942-1947. 547 doi:10.1093/brain/awh218 548 23. Dale RC, Pillai SC. Early relapse risk after a first CNS inflammatory demyelination episode: 549 examining international consensus definitions. Developmental Medicine & Child Neurology. 550 2007;49(12):887-893. doi:10.1111/j.1469-8749.2007.00887.x 551 24. Chen JJ, Flanagan EP, Bhatti MT, et al. Steroid-sparing maintenance immunotherapy for 552 MOG-IgG associated disorder. Neurology. 2020;95(2):e111-e120. 553 doi:10.1212/WNL.0000000000009758 554 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 25. Whittam DH, Cobo-Calvo A, Lopez-Chiriboga AS, et al. Treatment of MOG-IgG-associated 555 disorder with rituximab: An international study of 121 patients. Multiple Sclerosis and 556 Related Disorders. 2020;44:102251. doi:10.1016/j.msard.2020.102251 557 26. Hacohen Y, Wong YY, Lechner C, et al. Disease Course and Treatment Responses in 558 Children With Relapsing Myelin Oligodendrocyte Glycoprotein Antibody–Associated 559 Disease. JAMA Neurology. 2018;75(4):478-487. doi:10.1001/jamaneurol.2017.4601 560 27. Kivisäkk P, Mahad DJ, Callahan MK, et al. Human cerebrospinal fluid central memory CD4+ 561 T cells: Evidence for trafficking through choroid plexus and meninges via P-selectin. 562 Proceedings of the National Academy of Sciences. 2003;100(14):8389-8394. 563 doi:10.1073/pnas.1433000100 564 28. de Graaf MT, Smitt PAES, Luitwieler RL, et al. Central memory CD4+ T cells dominate the 565 normal cerebrospinal fluid. Cytometry Part B: Clinical Cytometry. 2011;80B(1):43-50. 566 doi:10.1002/cyto.b.20542 567 29. KIVISÄKK P, TREBST C, LIU Z, et al. T-cells in the cerebrospinal fluid express a similar 568 repertoire of inflammatory chemokine receptors in the absence or presence of CNS 569 inflammation: implications for CNS trafficking. Clinical and Experimental Immunology. 570 2002;129(3):510-518. doi:10.1046/j.1365-2249.2002.01947.x 571 30. Shearer WT, Rosenblatt HM, Gelman RS, et al. Lymphocyte subsets in healthy children 572 from birth through 18 years of age: The pediatric AIDS clinical trials group P1009 study. 573 Journal of Allergy and Clinical Immunology. 2003;112(5):973-980. 574 doi:10.1016/j.jaci.2003.07.003 575 31. Kok L, Masopust D, Schumacher TN. The precursors of CD8+ tissue resident memory T 576 cells: from lymphoid organs to infected tissues. Nat Rev Immunol. 2022;22(5):283-293. 577 doi:10.1038/s41577-021-00590-3 578 32. Smolders J, Heutinck KM, Fransen NL, et al. Tissue-resident memory T cells populate the 579 human brain. Nat Commun. 2018;9(1):4593. doi:10.1038/s41467-018-07053-9 580 33. van den Broek T, Borghans JAM, van Wijk F. The full spectrum of human naive T cells. Nat 581 Rev Immunol. 2018;18(6):363-373. doi:10.1038/s41577-018-0001-y 582 34. Pulko V, Davies JS, Martinez C, et al. Human memory T cells with a naive phenotype 583 accumulate with aging and respond to persistent viruses. Nat Immunol. 2016;17(8):966-975. 584 doi:10.1038/ni.3483 585 35. Cossarizza A, Ortolani C, Paganelli R, et al. CD45 isoforms expression on CD4+ and CD8+ 586 T cells throughout life, from newborns to centenarians: implications for T cell memory. 587 Mechanisms of Ageing and Development. 1996;86(3):173-195. doi:10.1016/0047-588 6374(95)01691-0 589 36. Farber DL, Yudanin NA, Restifo NP. Human memory T cells: generation, 590 compartmentalization and homeostasis. Nat Rev Immunol. 2014;14(1):24-35. 591 doi:10.1038/nri3567 592 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 37. Thome JJC, Yudanin N, Ohmura Y, et al. Spatial Map of Human T Cell 593 Compartmentalization and Maintenance over Decades of Life. Cell. 2014;159(4):814-828. 594 doi:10.1016/j.cell.2014.10.026 595 38. Machado-Santos J, Saji E, Tröscher AR, et al. The compartmentalized inflammatory 596 response in the multiple sclerosis brain is composed of tissue-resident CD8+ T lymphocytes 597 and B cells. Brain. 2018;141(7):2066-2082. doi:10.1093/brain/awy151 598 39. Höftberger R, Guo Y, Flanagan EP, et al. The pathology of central nervous system 599 inflammatory demyelinating disease accompanying myelin oligodendrocyte glycoprotein 600 autoantibody. Acta Neuropathol. 2020;139(5):875-892. doi:10.1007/s00401-020-02132-y 601 40. Hauser SL, Bar-Or A, Comi G, et al. Ocrelizumab versus Interferon Beta-1a in Relapsing 602 Multiple Sclerosis. N Engl J Med. 2017;376(3):221-234. doi:10.1056/NEJMoa1601277 603 41. Hauser SL, Bar-Or A, Cohen JA, et al. Ofatumumab versus Teriflunomide in Multiple 604 Sclerosis. New England Journal of Medicine. 2020;383(6):546-557. 605 doi:10.1056/NEJMoa1917246 606 42. Hauser SL, Waubant E, Arnold DL, et al. B-Cell Depletion with Rituximab in Relapsing–607 Remitting Multiple Sclerosis. New England Journal of Medicine. 2008;358(7):676-688. 608 doi:10.1056/NEJMoa0706383 609 43. Bogers L, Engelenburg HJ, Janssen M, et al. Selective emergence of antibody-secreting 610 cells in the multiple sclerosis brain. eBioMedicine. 2023;89. 611 doi:10.1016/j.ebiom.2023.104465 612 44. Cepok S, Jacobsen M, Schock S, et al. Patterns of cerebrospinal fluid pathology correlate 613 with disease progression in multiple sclerosis. Brain. 2001;124(11):2169-2176. 614 doi:10.1093/brain/124.11.2169 615 45. Cepok S, Rosche B, Grummel V, et al. Short-lived plasma blasts are the main B cell effector 616 subset during the course of multiple sclerosis. Brain. 2005;128(7):1667-1676. 617 doi:10.1093/brain/awh486 618 46. Magliozzi R, Howell O, Vora A, et al. Meningeal B-cell follicles in secondary progressive 619 multiple sclerosis associate with early onset of disease and severe cortical pathology. Brain. 620 2007;130(4):1089-1104. doi:10.1093/brain/awm038 621 47. Li R, Patterson KR, Bar-Or A. Reassessing B cell contributions in multiple sclerosis. Nat 622 Immunol. 2018;19(7):696-707. doi:10.1038/s41590-018-0135-x 623 48. Touil H, Li R, Zuroff L, et al. Cross-talk between B cells, microglia and macrophages, and 624 implications to central nervous system compartmentalized inflammation and progressive 625 multiple sclerosis. eBioMedicine. 2023;96. doi:10.1016/j.ebiom.2023.104789 626 49. Akaishi T, Misu T, Takahashi T, et al. Progression pattern of neurological disability with 627 respect to clinical attacks in anti-MOG antibody-associated disorders. Journal of 628 Neuroimmunology. 2021;351. doi:10.1016/j.jneuroim.2020.577467 629 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 50. Straeten F, Zhu J, Börsch AL, et al. Integrated single-cell transcriptomics of cerebrospinal 630 fluid cells in treatment-naïve multiple sclerosis. Journal of Neuroinflammation. 631 2022;19(1):306. doi:10.1186/s12974-022-02667-9 632 51. van Nierop GP, van Luijn MM, Michels SS, et al. Phenotypic and functional characterization 633 of T cells in white matter lesions of multiple sclerosis patients. Acta Neuropathol. 634 2017;134(3):383-401. doi:10.1007/s00401-017-1744-4 635 52. Jarius S, Pellkofer H, Siebert N, et al. Cerebrospinal fluid findings in patients with myelin 636 oligodendrocyte glycoprotein (MOG) antibodies. Part 1: Results from 163 lumbar punctures 637 in 100 adult patients. Journal of Neuroinflammation. 2020;17(1):261. doi:10.1186/s12974-638 020-01824-2 639 53. Kornbluh AB, Campano VM, Har C, et al. Cerebrospinal fluid eosinophils in pediatric myelin 640 oligodendrocyte glycoprotein antibody-associated disease. Multiple Sclerosis and Related 641 Disorders. 2024;85. doi:10.1016/j.msard.2024.105526 642 54. Fadda G, Yea C, O’Mahony J, et al. Epstein–Barr Virus Strongly Associates With Pediatric 643 Multiple Sclerosis, But Not Myelin Oligodendrocyte Glycoprotein-Antibody-Associated 644 Disease. Annals of Neurology. 2024;95(4):700-705. doi:10.1002/ana.26890 645 55. Bjornevik K, Munger KL, Cortese M, et al. Serum Neurofilament Light Chain Levels in 646 Patients With Presymptomatic Multiple Sclerosis. JAMA Neurology. 2020;77(1):58-64. 647 doi:10.1001/jamaneurol.2019.3238 648 56. Gross CC, Schulte-Mecklenbeck A, Steinberg OV, et al. Multiple sclerosis endophenotypes 649 identified by high-dimensional blood signatures are associated with distinct disease 650 trajectories. Science Translational Medicine. 2024;16(740):eade8560. 651 doi:10.1126/scitranslmed.ade8560 652 57. Mexhitaj I, Lim N, Fernandez-Velasco JI, et al. Stabilization of leukocytes from cerebrospinal 653 fluid for central immunophenotypic evaluation in multicenter clinical trials. J Immunol 654 Methods. 2022;510:113344. doi:10.1016/j.jim.2022.113344 655 58. Schulte-Mecklenbeck A, Zinnhardt B, Müller-Miny L, et al. Letter to the editor regarding 656 “Stabilization of leukocytes from cerebrospinal fluid for central immunophenotypic evaluation 657 in multicenter clinical trials.” Journal of Immunological Methods. 2023;514:113428. 658 doi:10.1016/j.jim.2023.113428 659 59. Oh J, Bar-Or A. Precision neuroimmunology in multiple sclerosis — the horizon is near. Nat 660 Rev Neurol. Published online July 4, 2024:1-2. doi:10.1038/s41582-024-00992-6 661 60. Teunissen CE, Petzold A, Bennett JL, et al. A consensus protocol for the standardization of 662 cerebrospinal fluid collection and biobanking. Neurology. 2009;73(22):1914-1922. 663 doi:10.1212/WNL.0b013e3181c47cc2 664 61. Banwell B, Bennett JL, Marignier R, et al. Diagnosis of myelin oligodendrocyte glycoprotein 665 antibody-associated disease: International MOGAD Panel proposed criteria. The Lancet 666 Neurology. 2023;22(3):268-282. doi:10.1016/S1474-4422(22)00431-8 667 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 62. Thompson AJ, Banwell BL, Barkhof F, et al. Diagnosis of multiple sclerosis: 2017 revisions 668 of the McDonald criteria. The Lancet Neurology. 2018;17(2):162-173. doi:10.1016/S1474-669 4422(17)30470-2 670 671 ACKNOWLEDGMENTS 672 This study was supported in part through the Melissa and Paul Anderson fund and Rosenblum 673 donation (ABO, BB) with DAE supported in part by NIH Medical Scientist Training Program 674 T32 GM07170, T32 G000046, F31AI167501, the Blavatnik Family Fellowship in Biomedical 675 Research at the University of Pennsylvania, and the Penn Colton Center for Autoimmunity. TZ 676 was supported in part by FWF Schrödinger grant (Nr: J4524). We would also like to thank the 677 Children’s Hospital of Philadelphia Division of Child Neurology for their help in procuring CSF 678 samples. Figure 1A was created with Biorender.com. 679 680 AUTHOR CONTRIBUTIONS 681 DAE performed analysis. DAE, TZ, GB, ST, IM, YK, and ZM performed experiments. DAE, TZ, 682 GB, ST, IM, MB, JL, SF, AnM, and MS assisted in sample transport and patient consenting. AmM 683 ascertained patient diagnoses. DAE and MK designed the web-based R-shiny app for exploration 684 of data. TZ, IM, CD, AS-M, HW, AR, and RL aided with analysis, protocol optimization, and panel 685 development. ATW, SEH, and SN aided with patient recruitment and lumbar punctures. BB and 686 ABO supervised the project. All authors contributed to the writing of the manuscript. 687 688 COMPETING INTERESTS 689 The authors declare no conflict of interest related to this study. 690 691 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint FIGURES AND FIGURE LEGENDS 692 693 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 694 Figure 1: Pediatric NIND CSF is primarily composed of memory CD4+ and CD8+ T cells, 695 with an age -associated increase in CD8+ memory T cell frequency (A) Schematic of CSF 696 sample procurement and flow cytometry data acquisition. (B) Cell count range (cells/mL) of NIND 697 CSF (n=33). (C) Frequencies of CD3+ T cells, CD14+ myeloid cells, NK cells, DCs, PMNs, and 698 B cells in NIND CSF. (D) Frequencies of T-cell subsets, as percent of CD3+ T cells, in NIND CSF. 699 (E) Frequencies of CD4+ naive and memory T cells, as percent of CD4+ T cells, in NIND CSF. 700 (F) Frequencies of CD8+ naive and memory T cells, as percent of CD8+ T cells, in NIND CSF 701 and (G) across the NIND age span. 702 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 703 Figure 2: An increase in CSF B-cell counts best differentiates ADS from both other 704 inflammatory and non-inflammatory neurological disease cohorts (A) Cell counts (cells/mL) 705 in NIND CSF (n=33), PIND CSF (n=7), AIE CSF (n=6), IDWM CSF (n=4), and ADS CSF (n=35). 706 (B) Cell counts (cells/mL) for CD3+ T cells, CD14+ myeloid cells, NK cells, DCs, PMNs, and B 707 cells across the same cohorts as A. (C) log10 of the fold change of median cell counts, comparing 708 the median cell count in ADS to the median cell count in other cohorts for each lineage. For Fig. 709 2A-B, Wilcoxon -rank-sum test used to compare ADS to NIND, PIND, AIE, and IDWM 710 independently (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001 ). NIND = non -711 inflammatory neurological disease, PIND = peripheral inflammatory neurological disease, AIE = 712 autoimmune encephalitidies, IDWM = inherited disorders of white matter, ADS = acquired 713 demyelinating syndromes. 714 715 716 **** * * Cells NINDPINDAIEIDWMADS 103 104 105 106 cells/mL A **** ***** * * * * * ** ****** * * DCs PMNs B cells CD3+ T cells CD14+ myeloid cells NK cells NINDPINDAIEIDWMADS NINDPINDAIEIDWMADS NINDPINDAIEIDWMADS NINDPINDAIEIDWMADS NINDPINDAIEIDWMADS NINDPINDAIEIDWMADS 0 101 102 103 104 0 101 102 103 104 101 102 103 104 0 101 102 103 104 102 103 104 105 0 101 102 103 104 cells/mL B log10(count fold−change) 0.0 0.5 1.0 1.5 2.0 B cells PMNs DCs NK cells CD14+ myeloid cells CD3+ T cells log10 fold−change of medians log10(ADS/comparator) Lineage Comparisons ADS vs. NIND ADS vs. PIND ADS vs. AIE ADS vs. IDWM not significant C .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 717 Figure 3: Pediatric MS CSF exhibits increased ASC frequencies and decreased CD14+ 718 myeloid cell frequencies when compared to MOGAD and other ADS. (A) B-cell, (B) ASC, and 719 (C) CD14+ myeloid cell frequencies and counts (cells/mL) in NIND CSF (n=33), PIND CSF 720 (n=7), AIE CSF (n=6), IDWM CSF (n=4), other ADS CSF (n=10), MOGAD CSF (n=10), and MS 721 CSF (n=15). (D) log10 of the fold change of median frequencies, comparing the median frequency 722 in MS to the median frequency in other ADS and MOGAD for each cell population (lineage or 723 subset). For Fig. 3A-C, Wilcoxon-rank-sum test used to compare MS to MOGAD, MS to other 724 ADS, and MOGAD to other ADS (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p < 0.0001). 725 ASC = antibody secreting cell, NIND = non-inflammatory neurological disease, PIND = peripheral 726 inflammatory neurological disease, AIE = autoimmune encephalitidies, IDWM = inherited 727 disorders of white matter, other ADS = non-MS/non-MOGAD acquired demyelinating syndromes, 728 MOGAD = myelin oligodendrocyte glycoprotein antibody -associated disease, MS = multiple 729 sclerosis. 730 **** * B cells NINDPIND AIEIDWM other ADSMOGAD MS 0 5 10 15 % of cells A ** *** ASCs NINDPIND AIEIDWM other ADSMOGAD MS 0.0 2.5 5.0 7.5% of cells B *** *** CD14+ myeloid cells NINDPIND AIEIDWM other ADSMOGAD MS 0 20 40 60 % of cells C *** B cells NINDPIND AIEIDWM other ADSMOGAD MS 0 101 102 103 cells/mL ** ** ASCs NINDPIND AIEIDWM other ADSMOGAD MS 0 101 102 103 cells/mL ** CD14+ myeloid cells NINDPIND AIEIDWM other ADSMOGAD MS 101 102 103 104 cells/mL log10(frequency fold−change) −2 −1 0 1 2 CD14+ myeloid cells PMNs DCs NK cells CD3+ T cells non−ASC B cells B cells ASCs log10 fold−change of medians log10(MS/comparator) Cell population Comparisons MS vs. MOGAD MS vs. other ADS not significant D .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint 731 Figure 4: Ratio of CSF ASC to CSF CD14+ myeloid cell frequencies differentiates MS from 732 MOGAD and other ADS (A) Ratio of ASC frequencies to CD14+ myeloid cell frequencies 733 (AMR) and ( B) CSF-only neuroinflammatory composite score s (coNCS) across NIND CSF 734 (n=33), PIND CSF (n=7), AIE CSF (n=6), IDWM CSF (n=4), and other ADS CSF (n=10), 735 MOGAD CSF (n=10), and MS CSF (n=15). (C) Receiver operating characteristic (ROC) curves 736 built for AMR or coNCS classifiers for the MS -vs-MOGAD/other ADS comparison, MS -vs-737 MOGAD comparison, and MS -vs-other ADS comparison, along with each corresponding area 738 under the curve (AUC). For Fig. 4A-B, Wilcoxon-rank-sum used to compare MS to MOGAD, MS 739 *** *** ASC/CD14+ Myeloid Ratio NIND PIND AIE IDWM other ADSMOGAD MS −2 −1 0 1 log10(AMR+0.01) A **** * CSF−only Neuroinflammatory Composite Score NIND PIND AIE IDWM other ADSMOGAD MS −2 0 2 4 log10(coNCS+0.01) B AMR AUC = 0.95 coNCS AUC = 0.80 AMR AUC = 0.94 coNCS AUC = 0.88 AMR AUC = 0.93 coNCS AUC = 0.96 MS−vs−MOGAD MS−vs−MOGAD/other ADS MS−vs−other ADS 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.00 0.25 0.50 0.75 1.00 0.0 0.8 False positive rate True positive rate ROC AMR coNCS C .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint to other ADS, and MOGAD to other ADS (* = p < 0.05, ** = p < 0.01, *** = p < 0.001, **** = p 740 < 0.0001). AMR = ASC to CD14+ myeloid cell ratio, coNCS = CSF -only neuroinflammatory 741 composite score, NIND = non -inflammatory neurological disease, PIND = peripheral 742 inflammatory neurological disease, AIE = autoimmune encephalitidies, IDWM = inherited 743 disorders of white matter, other ADS = non-MS/non-MOGAD acquired demyelinating syndromes, 744 MOGAD = myelin oligodendrocyte glycoprotein antibody -associated disease, MS = multiple 745 sclerosis. 746 747 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint Category Median Age Age Range Male Sex Female Sex Days from clinical episode onset to sampling (median for ADS groups only) Days from clinical episode onset to sampling (range for ADS groups only) Number in remission (MS only) NIND 12.2 2.6 - 19.4 21 12 -- -- -- PIND 16.8 3.5 - 19.4 3 4 -- -- -- AIE 14 8.7 - 19 2 4 -- -- -- IDWM 10.1 1.3 - 17.2 4 0 -- -- -- other ADS 9.3 0.7 - 17.9 8 2 4.5 1-40 -- MOGAD 9.3 1.8 - 16.2 5 5 12 1-23 -- MS 16.9 11.6 - 19.1 4 11 6 3-41 4 748 Table 1 Aggregated patient metadata by age and sex. NIND = non-inflammatory neurological 749 disease, PIND = peripheral inflammatory neurological disease, AIE = autoimmune encephalitidies, 750 IDWM = inherited disorders of white matter, other ADS = non -MS/non-MOGAD acquired 751 demyelinating syndromes, MOGAD = myelin oligodendrocyte glycoprotein antibody -associated 752 disease, MS = multiple sclerosis 753 754 755 756 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 1, 2025. ; https://doi.org/10.1101/2025.02.27.637541doi: bioRxiv preprint