Abstract

   Multiple sclerosis (MS) is a chronic inflammatory CNS disease with
   heterogeneous manifestation. Prognostic markers for early
   classification of MS are currently under investigation. Higher
   diagnostic resolution of cerebrospinal fluid (CSF) has the potential to
   contribute significantly to patient stratification, which should be
   especially important for a subgroup of patients with high risk to
   convert to a progressive disease course. This study aimed to determine
   whether spectral flow cytometry of CSF cells could identify pathogenic
   CD4+ T cell subset in MS. Using a two-step approach, we designed a
   marker panel informed by publicly available transcriptomic datasets
   from early human MS and our own single-cell RNA sequencing (scRNA-seq)
   in acute and chronic experimental autoimmune encephalomyelitis (EAE), a
   murine MS model. Notably, chronic (‘phase’) markers such as Il7r and
   Ramp3 (associated with memory T cells), Itgb1 (integrin beta-1) and
   anti-apoptotic genes like Dnaja1, Hsph1 and Jun/AP-1 were enriched in
   EAE. These markers reflect pro-survival signalling and tissue-residency
   characteristics, including CXCR6, CD69 and Bhlhe40, which suggest an
   adaptation of CD4+ T cells towards persistent neuroinflammatory
   responses in chronic EAE. This phase-specific marker profile highlights
   CD4+ T cells as both indicators and contributors to disease progression
   in EAE. Translating these findings to MS datasets, we found an
   enrichment of phase-specific markers in CSF cells. Spectral flow
   cytometry in an independent MS cohort revealed distinct memory and
   effector T cell subsets, indicating unique CSF signatures in MS. This
   study underscores the heterogeneity and dynamic changes of CD4+ T cells
   detectable by spectral flow cytometry, enhancing diagnostic resolution
   of CSF cells and informing more precise therapeutic strategies for MS.

   Keywords: multiple sclerosis, experimental neuroinflammation, chronic
   neuroinflammation, T cell subsets, cerebrospinal fluid
     __________________________________________________________________

   Crowley et al. explored CD4+ T cell heterogeneity in experimental
   autoimmune encephalomyelitis and multiple sclerosis (MS). Using
   single-cell RNA sequencing and spectral flow cytometry, they identified
   distinct phase- and cluster-specific markers. These findings enhance
   diagnostic resolution and highlight potential endophenotypes, paving
   the way for personalized therapeutic strategies in MS.

Graphical Abstract

Graphical Abstract.

   [44]Graphical Abstract
   [45]Open in a new tab

Introduction

   Multiple sclerosis (MS) is the most common chronic inflammatory disease
   of the central nervous system (CNS) leading to disability in young
   adults.^[46]1,[47]2 The disease is thought to be initiated by
   peripheral invading immune cells. T cells can be identified in MS
   lesions early on and throughout the disease course,^[48]3 and are
   thought to be implicated in the complex pathophysiology of MS next to B
   cells.^[49]4,[50]5 CD4+ T cells are the most abundant cell population
   in the cerebral spinal fluid (CSF) and were shown to be enriched in
   MS.^[51]6 Pathogenic T cells are considered one of the earliest to
   breach the blood-brain barrier in nascent demyelinating lesions.^[52]7
   The animal model experimental autoimmune encephalomyelitis (EAE) serves
   to understand the transmigration processes of inflammatory T cells into
   the CNS parenchyma. It has been shown that CNS pathology was initiated
   by CNS-infiltrating myelin-specific CD4+ T cells^[53]8 and carried on
   to the development of clinical EAE by T cell activities dependent on
   the Th17 key transcription factor RORyt.^[54]9 Pathogenic T cells
   isolated from the CNS demonstrated significant changes compared to
   typical Th17 cells, notably losing IL-17A production and shifting to a
   Th1-like (ex-Th17) phenotype.^[55]10 This phenotype is associated with
   gene expression patterns enriched in genetic risk variants linked to
   MS.^[56]11 However, it remains unclear how CD4+ T cells acquire the
   capacity to home to the CNS and why focal inflammation persists over
   extended periods in MS—one correlate of progressive MS. Over the last
   year, numerous single-cell studies have underlined the diagnostic
   potential of in-depth T cell phenotyping not only in MS.^[57]12,[58]13
   These data demonstrate the accumulation of tissue-resident memory T
   cells (TRMs) with characteristic gene expression profiles,^[59]13 which
   were also suggestive as diagnostic marker to identify pre-clinical
   MS.^[60]12 An in-depth analysis using single-cell RNA sequencing of CSF
   T cells and integrative modelling identified T cell subsets with a
   shared TRM signature (CXCR6, ZFP36L2, DUSP1 and ID2) tracking
   CNS-resident immune involvement in MS.^[61]14 Indirect evidence from
   studies on peripheral blood mononuclear cells in MS patients treated
   with Natalizumab (NTZ)—a drug believed to block the transmigration of
   pathogenic T cells into the CNS—suggests that this treatment may lead
   to an accumulation of these pathogenic T cells in the peripheral blood.
   Another study integrated several single-cell studies from CSF and CNS
   tissue from early MS disease stages, non-inflammatory controls and CSF
   from patients efficiently treated with NTZ. They found that CD4+ T
   cells with a distinct transcription factor profile (Runx3 + Eomes +
   T-bet−) and high levels of granzyme K and CCR5, particularly in Th17.1
   cells, exhibit brain-homing abilities. Interestingly, this subset
   decreases in blood and accumulates in cerebrospinal fluid of untreated
   patients, with levels reversing after NTZ treatment.^[62]15

   Our study aimed to set up a multicolour panel for spectral flow
   cytometry, a cost-effective method with the potential to be implemented
   into routine diagnostics with an emphasis on CD4+ T cell markers which
   have been associated with chronic disease progression. Next to
   available transcriptomic datasets of early human MS, we employed
   single-cell RNA sequencing (scRNA-seq) to analyse CD4+ T cell subsets
   in acute and chronic EAE. Findings were then translated to human MS
   transcriptomic datasets, confirming the enrichment of our identified
   cluster markers in CSF cells. Finally, we applied high-dimensional
   spectral flow cytometry to examine the surface expression of both
   established and novel CD4+ T cell markers in CSF from a discovery and
   extended cohort of early MS patients.

Materials and methods

Mice

   All animals were raised and housed under SPF conditions in individually
   ventilated cages. Fluorescent reporter mice (B6.IL-17-EGFP_IFNg-YFP)
   were either derived from the creators^[63]16 or purchased from Jackson
   labs.^[64]17 All animal experiments were conducted according to the
   German Animal Protection Law. Experimental approval (G0009/16) was
   obtained from the State Office for Health and Social Affairs (‘LaGeSo’)
   in Berlin. Both male and female animals were used for EAE experiments.

Experimental autoimmune encephalomyelitis (EAE)

   To induce EAE in mice, 125 µg MOG[35–55] (Charité Universitätsmedizin
   Berlin) emulsified in 400 µg Mycobacterium tuberculosis H-37Ra and
   complete Freund’s adjuvant (Sigma-Aldrich) was injected subcutaneously.
   This was followed by two intraperitoneal injections of 300 ng pertussis
   toxin reconstituted in Phosphate Buffered Saline (PBS, Gibco) at 2 and
   24 h post-immunization. Mice were monitored daily for weight and
   clinical signs of EAE. These were scored as follows, 0.5, tail paresis;
   1.0, tail plegia; 0.5, weak righting reflex; 1.0 hind limb paresis; 2.0
   hind limb paralysis; 0.5 forelimb paresis and 1.0, forelimb paralysis.
   Mice reaching a total score of 3.5 or losing more than 20% of their
   initial weight were euthanized in line with animal welfare regulations.
   To induce remission of EAE, corticosteroid treatment (1.6 mg
   methylprednisolone in 200 µl 0.9% NaCl) was administered i.p. for 3–5
   consecutive days when clinical signs exceeded a score of
   2.^[65]18,[66]19 All animal experiments had been approved by local
   authorities (G0009/16, Landesamt für Gesundheit und Soziales Berlin).

Patient recruitment, ethics approval and consent to participate

   The human subjects’ consent was obtained according to the Declaration
   of Helsinki. Ethical approval was obtained (internal number:
   EA1_023_15) by the ethical committee of Charité Berlin. PwMS (n = 21),
   including diagnoses of clinical or radiological isolated syndrome, and
   non-inflammatory control patients (n = 15) with idiopathic intracranial
   hypertension (IIH) or excluded inflammatory CNS disease were recruited
   during clinical routine diagnostics, where lumbar puncture was part of
   the diagnostic workflow. Ethical approval for this study was obtained
   (EA1_023_15) from the local Ethics Committee of the Charité. CSF
   samples from pwMS were included only if lumbar puncture was performed
   before steroid or other immunomodulatory treatments. Clinical details
   (age, sex) of pwMS and controls are listed in [67]Supplementary Table
   S7.

CNS T cell isolation

   Mice were administered a lethal dose of a ketamine–xylazine mixture
   (ketamine 320 mg/kg; xylazine 5 mg/kg) intraperitoneally, followed by
   transcardial perfusion with 50 ml of ice-cold PBS. This procedure
   resulted in <1%, of contaminating T cells derived from the circulation,
   as verified using intravenously injected anti-CD45 fluorescently
   labelled antibodies in previous experiments. Brainstem and spinal cord
   tissues were removed, placed in ice-cold D-PBS (Ca2+, Mg2+, glucose,
   pyruvate, Gibco®) chopped into smaller pieces and transferred into
   C-tubes® (Miltenyi Biotec, Germany) containing enzyme mix 1 and 2 (for
   further details please refer to Miltenyi's Adult Brain Dissociation Kit
   Protocol). Tissue was homogenized using a gentleMACS® Octo Dissociator
   with Heaters, Miltenyi), programme 37C_ABDK_01. The homogenate was
   briefly centrifuged, resuspended in D-PBS then passed through a 100 µm
   cell strainer and centrifuged at 300×g and 4°C for 10 min. The
   resultant cell pellet was resuspended in 3.1 ml cold D-PBS mixed with
   900 µl Percoll®, transferred to a 15 ml tube, overlaid with 4 ml cold
   D-PBS and centrifuged at 3000 × g for 10 min at 4°C without breaking.
   Following phase separation, the two top phases were aspirated and the
   cell pellet washed in cold D-PBS at 1000 × g for 10 min at 4°C.
   Erythrolysis was performed by resuspending the pellet in 10 ml of 1 ×
   Red Blood Cell Removal Solution®, incubating for 10 min at 4°C and
   halting the reaction with 10 ml of cold D-PBS/BSA buffer (0.5% BSA, pH
   7.2). A final wash at 300 × g for 10 min at 4°C yielded the single-cell
   suspension.

Ex vivo T cell stimulation and immunostaining

   For the cytokine production analysis, single-cell suspensions of CNS
   tissue were resuspended in RPMI-based cell culture medium (Gibco®),
   supplemented with 10% foetal calf serum, 1% penicillin/streptomycin, 1%
   GlutaMAX®, 1% HEPES. Cells were plated on anti-CD3 (BD Pharmingen,
   clone 145-2C11) and anti-CD28 (BD Pharmingen, clone 37.51)-coated
   culture plates and incubated overnight before immunostaining. Brefeldin
   A (BioLegend) was added 2 h after plating. Cells for surface analysis
   and single-cell sequencing were directly immunostained. Fc-blocking (BD
   Pharmingen) was performed for 15 min to reduce non-specific antibody
   bindings. Cells were then incubated for 10 min with the primary
   antibodies (CD11b, clone M1/70, BD Pharmingen; CD45, clone 30F11,
   Miltenyi and CD4, clone GK1.5, BioLegend) diluted in D-PBS/BSA buffer.
   For intracellular staining, cells were fixed in 2% PFA for 20 min
   washed in PBS, permeabilized in a saponin buffer (0.5% saponin, 0.5%
   BSA, pH 7.6), and stained with the antibodies against IFN-γ (clone
   XMG12, Biolegend), GM-CSF (Invitrogen) and IL-17A (clone eBio17B7,
   eBioscience). Stained cells were washed in PBS/BSA buffer before
   analysis.

FACS-sorting

   CNS cell suspensions were immunostained for cell surface antigens (see
   above) and sorted on a BD FACSAria™ II instrument. CD4^+ T cells were
   gated by f size and granularity (FSC/SSC), single cells (FSC-H/FSC-W),
   PI^neg, CD45^hi, CD11b^neg and CD4^+ cells. Sorted cells were collected
   directly into pre-cooled culture medium.

Single-cell transcriptome analysis

   To perform single-cell transcriptome analysis of CNS CD4+ T cells, we
   isolated CD45^hiCD11b^−CD4^+ T cells by FACS-sorting from acute EAE
   (two samples, each two mice) and chronic EAE (two samples, each four
   mice). On average, using FACS, 180 000 cells from the acute phase and
   70 000 cells from the chronic phase were isolated per replicate for
   scRNA-seq. Libraries were prepared according to the manufacturer's
   instructions using the 10 × Genomics platform for the Single Cell
   3-prime Gene Expression v3 workflow. Sequencing was performed on the
   HiSeq 4000 platform and demultiplexed using bcl2fastq v2.20.0.
   Alignment, filtering, UMI counting, barcode-counting and gene-barcode
   matrix generation were done with Cell Ranger (version 3.0.2, 10 ×
   Genomics).

Flow cytometry of human CSF cells

   CSF samples were processed within 1 h after lumbar puncture. Samples
   were centrifuged at 300× g for 10 min. The resulting cell pellet was
   stained with a fluorophore-conjugated antibody mixture for 10 min
   ([68]Table 1). Staining was stopped with 1 ml FACS buffer. After
   washing, cells were fixed with 2% formaldehyde solution at room
   temperature. Following another wash step, cells were resuspended in
   200 µl FACS buffer for analysis. Cytometric measurements were performed
   using a Spectral Flow Cytometer (Cytek Aurora®).

Table 1.

   Monoclonal antibodies utilized for flow cytometry analysis of human CSF
   cells
   Biolegend #Catalog Clone Dilution Marker selection rationale
   (cluster/phase association or lineage)
   CCR5 313711 HEK/1/85a 1:40 C0, C1 phase markers
   CD29 Itgb1 303023 TS2/16 1:40 C3 cluster marker; C0, C2 phase marker
   PD1 329949 EH12.2H7 1:40 C0 phase marker
   CD8 300935 HIT8a 1:40 Lineage marker
   CD45RO 304233 UCHL1 1:40 C1, C2 phase markers
   CD4 317439 OKT4 1:40 Lineage marker
   CD137 309843 4B4 1:40
   CD74 326807 LN2 1:40 C2 cluster/phase marker
   IL7R 351313 A019D5 1:40 C0 phase marker
   CD25 302607 BC96 1:40 C1 phase marker
   CD69 310929 FN50 1:40 C1 phase marker
   GITR 371205 108-17 1:40 C0, C2 phase marker
   Streptavidin 405227 1:100 secondary antibody
   CD103 350220 Ber-ACT8 1:200 C0, C1, C2 phase marker
   KLRG1 138421 2F1/KLRG1 1:100 C2 cluster marker
   CD19 302271 HiB19 1:40 Lineage marker
   BD
    CD45 563791 563791 1:40 Lineage marker
    CD16 612945 612945 1:40 Lineage marker
    CD3 612896 612896 1:40 Lineage marker
    CD14 612764 612764 1:40 Lineage marker
    HLA-DR 612981 612981 1:40 Lineage marker
   [69]Open in a new tab

CSF protein analysis

   Results of CSF protein analysis were obtained from routine diagnostics
   at Labor Berlin—Charité Vivantes GmbH. IgG in the gel were separated
   electrophoretically based on their acidic and basic amino acid
   residues. Each IgG migrates in the electric field until the surrounding
   pH value corresponds to its isoelectric point and can be detected with
   a specific antiserum as a precipitation, which is visible for clonally
   expanded IgG as a band. Albumin, IgG, IgA and IgM were measured by
   nephelometry. Quotient schemes were calculated using the Reiber
   method.^[70]20-22

EAE single-cell transcriptomics analysis and statistics

   Count matrices were analysed in the R package, Seurat.^[71]23,[72]24
   Cell Ranger data were imported using the ‘Read10X’ function. Empty
   droplets and potential doublets were excluded by filtering for cells
   expressing more than 850 genes but fewer than 2800 genes. Contaminating
   macrophages were removed by excluding cells with detectable expression
   of P2ry12, Tmem119, Itgam, H2-Ab1, H2-Aa, Siglech, Ly6d, Cd209a, Cd209d
   or Lyz2. To exclude dead or dying cells, we filtered out those with
   mitochondrial gene expression exceeding 8% of total gene expression.
   Post-filtering, 15 253 cells (Acute 1: 3039, Acute 2: 4240, Chronic 1:
   3445 and Chronic 2: 4529) passed quality control, with an average of
   5000 reads/cell.

   To integrate datasets across replicates and disease conditions,
   log-normalization was applied to each dataset individually, and the top
   3000 variable features were identified. Seurat's IntegrateData function
   was used to combine datasets with default parameters. Clustering was
   performed based on the first 30 dimensions, using the Louvain
   algorithm. Following inspection of dimensionality by elbow plots of the
   first 50 dimensions, principal component analysis, clustering and
   dimensionality reduction via UMAP were performed on the first 20
   dimensions. Cluster markers were identified using the
   FindConservedMarkers function to determine constant markers across
   disease phases. Clustering resolution was adjusted to the point where
   known cell-type-specific markers were identified in single clusters.
   Phase-specific markers were identified using the FindMarkers function
   to compare clusters between disease states. An adjusted p-value cut-off
   of p.adj < 0.05 was applied, and genes meeting this were analysed
   further using Ingenuity Pathway Analysis.

   Publicly available single-cell sequencing data was downloaded from
   Zenodo ([73]https://zenodo.org/records/6910635). Single-cell sequencing
   matrices were imported and Seurat Objects created using the
   CreateSeuratObject function (Seurat v5.0.1). Data were subsetted to
   CD4+ cells based on the provided metadata. To assess the enrichment of
   markers obtained from our mouse dataset, Seurat's AddModuleScore
   function was used to compute an enrichment score for non-mitochondrial
   and ribosomal mouse genes converted to their human homologues.

CSF cells flow cytometry data analysis

   Flow data was analysed using FlowJo® for manual gating or pre-processed
   for unbiased cluster analysis using Seurat. For manual gating, a
   minimum of 100 cells per CSF sample was required for downstream
   analysis. Compensated and FlowJo-normalized data were refined to
   include only lymphocytes. Data were loaded and processed using R
   (v4.3.0). Cells below a mean fluorescence intensity value of 1023 were
   filtered out to ensure data quality. Data were rescaled and centred
   before determining the principal components (PCs) across n–1
   dimensions, where n was the maximum number of dimensions. Given that
   sample analysis spanned multiple days, batch correction was applied
   using Harmony (v0.1.1)^[74]25 to minimize variability in the PC space.
   Each sample was treated as an individual batch, and the integrated PC
   space was used for downstream analysis. A shared nearest neighbour
   graph (k = 20) was constructed as implemented in Seurat (v5). Clusters
   were identified using modularity-based clustering (Louvain
   algorithm^[75]26,[76]27) and visualized with uniform manifold
   approximation and projection (UMAP).^[77]28 For a minimum distance of
   0.3 and 30 neighbouring cells were used to enhance visualization of
   local structures in the dataset. To specifically focus on CD4+ T cells,
   we refined our data to include markers relevant to T cell lineage
   (KLRG1, CD29, CD25, CD103, CD154, IL7R, PD1, CD137, Il1R1, HLA-DR,
   CD69, CCR5, CD45RO and GITR). Clustering and dimensionality reduction
   were rerun on this refined dataset. Plots were generated using ggplot2
   v3.4.2.^[78]29 Statistical analysis of manually gated flow data was
   performed using Graphpad Prism (Version 9.0, GraphPad Software, San
   Diego, CA, USA). Values are expressed as mean ± SEM. Statistical tests
   included Kruskal–Wallis, Dunn's multiple comparison, Mann–Whitney U,
   t-test, or one-way ANOVA, as appropriate (indicated in figure legends;
   P < 0.05 was considered statistically significant).

Results

ScRNA-seq of CNS CD4+ T cells in acute and chronic EAE identifies five
subsets

   In EAE, myelin-specific CD4+ T cells elicit an autoimmune reaction
   within the CNS leading to ascending paralysis that begins 10–14 days
   after subcutaneous immunization with myelin oligodendrocyte
   glycoprotein in conjunction with adjuvants. During the acute phase of
   EAE, clinical signs peak 3–4 days after onset. For single-cell
   transcriptome analysis, mice selected for analysis at the peak of the
   disease had a clinical score ranging from 3.0 to 3.5 ([79]Fig. 1A, left
   axis). Chronic EAE was defined as 30 or more days after the onset of
   clinical signs, with chronic EAE animals showing a score range of
   0.75–2.75. The CNS-isolated CD4+ T cell counts were significantly
   elevated at the peak of the disease (median: 95 565 cells/mouse)
   compared to non-immunized control animals (median: 364 cells/mouse).
   Although CD4+ T cell numbers declined in chronic EAE (median: 15 659
   cells/mouse; [80]Fig. 1A, right axis, statistical analysis in
   [81]Supplementary Fig. S1B), they remained consistently higher than in
   control animals. FACS analysis ([82]Supplementary Fig. S1A) revealed
   that a substantial proportion of CNS effector T cells expressed the
   cytokines IFN-γ, GM-CSF and IL-17A in both acute and chronic EAE. Next,
   we FACS-sorted CD45^hiCD11b^−CD4^+ T cells and used the droplet-based
   10 × Genomics Chromium approach to examine disease states. ScRNA-seq
   was performed on the HiSeq 4000 platform. Using unbiased clustering
   based on the top variable genes present in both conditions, we
   identified six clusters overall (referred to as Cluster 0–5 or C0–C5)
   with C0 being the largest and C5 the smallest ([83]Fig. 1B). No
   clusters were found exclusively in chronic EAE. Still, cluster-specific
   shifts in proportion were observed for clusters C0–C3. C1, identified
   as regulatory T cells, increased in proportion in chronic EAE (from 13%
   in acute EAE to 30.2% in chronic EAE). Conversely, the chronic EAE
   Cluster C0 showed a size reduction (from 25.2% in acute EAE to 18.9% in
   chronic EAE; [84]Fig. 1B). Clusters 3, 4 and 5 were comparable in
   proportion between the two different disease states.

Figure 1.

   [85]Figure 1
   [86]Open in a new tab

   scRNA-seq of CNS CD4+ T cells in acute and chronic EAE reveals five
   subsets. (A) Time course of the mean clinical EAE score (±SEM, left
   y-axis) based on data from at least three independent experiments, with
   individual mouse scores synchronized to the first incidence of a
   clinical score ≥1. Asterisks denote absolute CD4+ T cell counts
   recorded by FACS (right y-axis; CD45hi CD11b− CD4+ T cells) in the CNS,
   which were significantly higher during both the acute and chronic EAE
   phases compared to the naïve control group. Sample sizes: control (n =
   21), acute EAE (n = 10) and chronic EAE (n = 11). Statistical analysis
   was conducted using the Kruskal–Wallis test and Dunn's multiple
   comparison test. (B) UMAP visualization of Seurat-derived cell clusters
   using nonlinear dimensionality reduction, identifying six clusters
   (0–5, color-coded) in acute and chronic EAE (pooled data from two
   independent samples; the number of mice used in the scRNA-seq
   experiment is n = 4 (2 + 2) for acute EAE and n = 8 (4 + 4) for chronic
   EAE, resulting in on average 180 000 cells from the acute phase and 70
   000 analysed cells from the chronic phase sorted and sequenced per
   replicate). (C) Dot plot showing output from IPA's Biomarker tool
   applied to Seurat's FindMarkers results (threshold: Fold Change > 1) to
   identify biologically relevant genes marking each cluster (positive
   expression). Clusters C0–C4 are color-coded as follows: C0: red, C1:
   yellow, C2: green, C3: blue, C4: purple. The fully resolved gene
   abbreviations are provided in [87]Supplementary Table S8.

   Differential lineage marker analysis confirmed the accuracy of our
   positive sorting on CD4+ T cells ([88]Supplementary Fig. S1C). Due to
   very low cell numbers in C5 (acute: 4 cells, or 0.05% in acute EAE), a
   reliable in-depth analysis of this subset was not possible.
   Consequently, all further analysis focused on C0–C4. We used Seurat's
   FindMarkers function and Ingenuity® Pathways Analysis (IPA) Biomarker
   analysis tool, which enabled the cluster differentiation into memory
   (C0), regulatory (Treg) (C1), effector (C2), cytotoxic (C3) and
   lymphoid-tissue-like (C4) phenotypes ([89]Fig. 1C and [90]Supplementary
   Fig. S1D). A complete marker list is available in [91]Supplementary
   Table S1. In Clusters C0–C3, we observed an enrichment of tissue
   residency markers (Cxcr6, Cd69 and Bhlhe40) in C0–C3 ([92]Supplementary
   Fig. S1D). The C0 cluster was memory-like and lacked differentially
   expressed genes, with only an increased expression of innate immune
   receptors Ifitm2 and Ifitm3 could be found. The C1 Treg cluster was
   defined by a large set of differentially upregulated genes involved in
   Treg activation and effector functions that included cytotoxic
   potential. The conventional T effector subpopulation, C2, expressed
   genes involved in pro-inflammatory cytokine production, inhibition of
   proliferation, modulating cytokine/TCR signalling and a diverse set of
   co-stimulatory and exhaustion markers. C3 displayed the characteristic
   gene signature of Tbet-dependent cytotoxic CD4+ T cells, with machinery
   for antiviral responses and cytotoxicity. Cluster C4 exhibited a
   lymphoid-tissue–typical gene expression profile—closely resembling that
   of central memory T cells. Additional information on previously
   reported genes in the context of T cells and neuroinflammation is
   available in [93]Supplementary Table S2.

Cluster-specific adaptations dominate between acute and chronic EAE, and
pro-survival prevails across subsets

   The persistence of T cells within the CNS is one of the main features
   of multiple sclerosis, but the mechanisms behind T cell persistence
   remain unclear. To identify factors involved in these processes, we
   analysed cluster-specific markers, or ‘phase markers’, which define
   chronic versus acute EAE for each of the identified CD4+ T cell
   subpopulations ([94]Fig. 2A). In chronic disease, C0 showed increased
   expression of memory T cell markers such as Il7r and Ramp3 in chronic
   disease. C1 upregulated genes associated with effector or cytotoxic
   Treg in chronic EAE. C2 showed an increase in tissue-resident marker
   expression and demonstrated metabolic changes in chronic EAE. C3 showed
   selective upregulation of integrin beta-1 (Itgb1) and other genes
   upregulated in chronic EAE. For C4, we observed an upregulation of
   typical lymphoid-tissue-related genes. Additional information on
   previously reported genes in the context of T cells and
   neuroinflammation is available in [95]Supplementary Table S3.

Figure 2.

   [96]Figure 2
   [97]Open in a new tab

   Cluster-specific adaptations between acute and chronic EAE highlight
   pro-survival signals across subsets. (A) Cluster-specific gene
   enrichment changes in the chronic phase of EAE. Spot size represents
   the proportion of cells in each cluster expressing the gene, while
   greyscale intensity indicates the average expression level within the
   cluster. Clusters are color-coded as follows: C0: red, C1: yellow, C2:
   green, C3: blue, C4: purple (legend provided for reference). The
   ‘acute’ dataset corresponds to the acute EAE phase, while the ‘chronic’
   dataset represents the chronic EAE phase, as defined in [98]Fig. 1
   (pooled data from two independent samples; the number of mice used in
   the scRNA-seq experiment is n = 4 (2 + 2) for acute EAE and n = 8 (4 +
   4) for chronic EAE, resulting in on average 180 000 cells from the
   acute phase and 70 000 analysed cells from the chronic phase sorted and
   sequenced per replicate). The fully resolved gene abbreviations are
   provided in [99]Supplementary Table S8. (B) Genes enriched across
   multiple clusters (pan-clusters) during the chronic phase of EAE (based
   on scRNA-seq data set as described in A). The fully resolved gene
   abbreviations are provided in [100]Supplementary Table S8. (C)
   Ingenuity Pathway Analysis (IPA) of differentially expressed genes
   between disease phases predicts activation of pathways related to cell
   survival and apoptosis [magenta: chronic phase, grey: acute phase;
   (based on scRNA-seq data set as described in A)]. Z-scores reflect the
   expression direction versus known effects; no statistical testing has
   been applied. (D) IPA of molecular functions and diseases affected
   between chronic and acute EAE based on differentially expressed genes,
   determined by cluster-wise comparisons of disease phases. Adjusted
   P-value threshold = 0.05 (based on scRNA-seq data set as described in
   A). IPA pathway enrichment analysis uses right-tailed Fisher's Exact
   Test and Benjamini-Hochberg (BH) correction. (E) IPA-predicted upstream
   regulators of differentially expressed genes between disease phases
   [magenta: chronic phase, grey: acute phase; (based on scRNA-seq data
   set as described in [101]Fig. 2A)].

   A comparison of disease phases across the entire CD4+ T cell population
   revealed commonalities in metabolism and apoptotic signalling
   ([102]Fig. 2B). When analysing genes enriched in the chronic phase of
   EAE across multiple CD4+ T cell subtypes, a smaller subset of genes
   emerged from the analysis. In particular, genes involved in
   anti-apoptotic signalling—e.g. Dnaja1 (DnaJ homologue subfamily A
   member 1) and heat shock protein-coding genes, e.g. Hsph1/Hspe1/Hspa1b
   were upregulated in chronic EAE and have been previously reported as
   essential for the manifestation of clinical signs in EAE.^[103]30
   Pathway analysis of cluster-specific differential gene expression
   demonstrated phase and cluster-specific pathway activation patterns
   ([104]Fig. 2C). Notably, C1 (Treg) showed chronic phase enrichment in
   programmes associated with effector Treg function ([105]Fig. 2C). For
   C1, IPA-predicted a relative decrease in pathways involved in neuronal
   apoptosis and cell death. C2 (Teff) displayed an increase in leukocyte
   activation signals and inflammatory signalling and diverse Th17-related
   pathways. IPA-based prediction of the molecular functions and regulated
   pathways from our cluster-specific pairwise comparisons showed an
   increase in survival pathways across all subsets in the chronic phase
   of EAE ([106]Fig. 2D). We identified a unifying signature of
   anti-apoptotic signalling involving protein ubiquitination pathways,
   glucocorticoid receptor signalling pathways, heat shock responses and
   unfolded protein responses. Examining potential upstream inducers of
   these programmes, we identified HSF1 as a candidate regulator of
   anti-apoptotic and heat shock responses ([107]Fig. 2E). The relative
   absence of SGPP2 (sphingosine 1 phosphate phosphatase 2) activity
   appears to be a unifying potential regulatory factor for all tissue
   resident clusters (C0–C3). Inflammatory cytokines are known to
   influence chronic EAE; here, we saw the influence of TNF in C1 to C3
   and of IL-1b in C1/C2. In line with these observations, TCR stimulation
   and CD40L activation are potential inducers of the observed chronic EAE
   changes in C1.

   These results indicated that CD4+ T cell subtype changes in persistent
   neuroinflammation include adaptation towards a tissue-resident
   phenotype in C0–C3 and a specialized, antagonistic effector function in
   C1 and C2. C4 exhibits a more pronounced central memory phenotype. A
   common pattern of increased anti-apoptotic and pro-survival programmes
   can be seen in all clusters. Collectively, these findings indicate a
   role for these markers in the CNS persistence of CD4+ cells beyond the
   acute inflammatory period.

Compartment-specific and disease-associated enrichment of EAE CD4+ T cell
cluster and phase markers in MS

   In the next step, we analysed a publicly available single-cell RNA-seq
   dataset integrating CSF and blood samples from early pwMS to assess the
   expression of cluster-specific and differentially upregulated genes
   identified in EAE ([108]Fig. 3A). Using the Seurat workflow, we
   selected conventional CD4+ T cells (Tcon) and regulatory T cells (Treg)
   from the Schafflick et al. dataset and mapped the murine differentially
   expressed genes (DEGs) sets onto the human data ([109]Supplementary
   Fig. S2). The calculated enrichment scores for both the cluster markers
   ([110]Fig. 3B) and phase markers ([111]Fig. 3C) provide a quantitative
   measure of whether a predefined set of DEGs from one dataset is
   statistically over-represented in another, allowing us to assess the
   relevance of EAE-identified signatures in human CD4+ T cell samples.

Figure 3.

   [112]Figure 3
   [113]Open in a new tab

   Mapping EAE-identified cluster and phase markers onto human MS
   scRNA-seq data. (A) Schematic overview of the workflow used to
   integrate murine and human scRNA-seq datasets. A publicly available
   single-cell dataset (reproduced from Schafflick et al.^[114]1) from
   human cerebrospinal fluid (CSF) and blood of early pwMS (pwMS vs
   control; blood (n): 5 versus 5; CSF (n): 4 vs 4) was analysed to assess
   the expression of EAE-identified, cluster-specific markers. After
   selecting CD4+ T cells, murine cluster marker sets (C1–C4, as outlined
   in [115]Supplementary Table S4) were mapped onto human MS and control
   samples to identify enriched genes. (B) Box and whisker plots showing
   the distribution of enrichment scores of cluster markers from clusters
   C0 to C4 in human CD4+ T cells, subdivided into conventional helper T
   cells (Tcon) and regulatory T cells (Treg). C0 markers were enriched in
   both CSF and blood CD4+ conventional T cells, while C1 markers showed a
   notable enrichment in blood-derived Tregs, consistent with their
   characterization as markers of memory and regulatory cells (see also
   [116]Supplementary Fig. S2A). MS = multiple sclerosis, ctrl = control.
   Statistical analyses were performed using the Mann–Whitney U test, with
   Bonferroni correction applied for multiple comparisons (see
   [117]Supplementary Table S5 for full details). (C) Box and whisker
   plots showing the enrichment of chronic phase DEGs from clusters C0 to
   C4 in the human dataset. Phase-defining markers for Clusters C0–C3 were
   enriched in both CSF and blood CD4+ T cells (Tcon and Treg) in pwMS
   compared to controls. In contrast, C4 markers were similarly
   distributed between the two groups, suggesting a lack of
   disease-specific association (see also [118]Supplementary Fig. S2B).
   Statistical analyses were performed using the Mann–Whitney U test, with
   Bonferroni correction applied for multiple comparisons (see
   [119]Supplementary Table S6 for full details).

   The cluster markers ([120]Supplementary Table S4) were mapped
   separately onto conventional (Tcon) and regulatory (Treg) CD4+ T cells
   in the human dataset ([121]Supplementary Fig. S2A), further
   differentiating between MS and control samples. For the cluster markers
   ([122]Fig. 3B; [123]Supplementary Fig. S2B), C1 (Treg) markers showed
   an enrichment in Tregs (using the Tcon as reference) and further
   exhibited disease-specific enrichment in both the CSF and blood of pwMS
   compared to controls. C0 markers were reduced in Tregs across both
   compartments (blood and CSF, referenced to Tcon) but did not show
   strong disease-related differences. C2 (effector) markers increased
   expression in both Tcon and Treg populations in the CSF compared to
   blood Tcon, with an additional enrichment in blood-derived MS Tregs. C3
   markers (cytotoxic) were similarly distributed in Tcon and Treg
   populations in the CSF but were reduced in Tcon cells in the blood. C4
   cluster markers (lymphoid-tissue-like) showed no significant
   differences related to disease status or compartment specificity.
   Overall, Tregs in MS, particularly in CSF, showed strong enrichment for
   the identified cluster markers, suggesting a disease-specific
   regulatory T cell phenotype.

   We then performed a similar analysis for the phase markers ([124]Fig.
   3C, [125]Supplementary Fig. S2C) that distinguished chronic from acute
   EAE in the human dataset ([126]Supplementary Table S4). C4 phase
   markers showed no significant differences between Tcon and Treg
   populations in either the CSF or blood. In contrast, phase markers from
   C0 to C3 were enriched in Tcon and Treg MS samples compared to Tcon
   (blood) across both compartments, with the most pronounced enrichment
   observed in Tregs across all clusters in MS versus control samples.

   Notably, Tregs showed greater enrichment than Tcon in the CNS compared
   to blood, along with clear disease-associated differences within both
   subsets. These findings suggest that EAE-derived phase markers (C0–C3)
   are relevant for the persistence of certain CD4⁺ T cell subpopulations
   in the CNS, reinforcing their involvement in chronic neuroinflammation.
   Consistent and significant enrichment of chronic phase markers in
   MS-derived Tregs and CD4⁺ cells—particularly in the CSF—further
   supports their role in disease progression.

CD4+ T cell subset analysis in early MS distinguishes between a more acute
versus chronic CSF signature

   Building on our findings from the EAE model and publicly available MS
   datasets, we applied spectral flow cytometry to further characterize
   CD4+ T cell subsets in human CSF. Our marker panel was designed based
   on (i) cluster and phase markers identified in EAE ([127]Table 1,
   [128]Supplementary Table S4, [129]Supplementary Fig. S1) and (ii)
   additional markers to distinguish non-CD4+ T cell populations, ensuring
   a comprehensive analysis. Using this approach, we investigated CNS T
   cell subsets in an independent cohort of pwMS and controls to assess
   differences in immune signatures between acute and chronic disease
   states.We recruited pwMS who were referred to our inpatient service due
   to clinical and MRI suspicion of chronic inflammatory disease, along
   with control patients diagnosed with IIH, a non-inflammatory condition,
   during routine diagnostics. We analysed CSF samples from a discovery
   cohort of 25 patients (n = 15 MS, n = 0 IIH).

   Initial analyses were conducted in an unbiased manner with the Seurat
   workflow. Foetal calf serum files were batch-integrated with Harmony,
   ^[130]25 and CD45 + CD3 + CD4+ clusters were selected for a refined
   clustering based on activation markers. UMAP visualization showed
   distinct clustering of CD4+ T cell subpopulations ([131]Fig. 4A) in
   both MS and non-inflammatory control patients. The identified
   subpopulations included resting Tregs (CD25^hiIL7R^low), activated
   Tregs (CD25hiCCR5 + GITR + HLADR+) and conventional T cells, including
   central memory T cells (CD45RO^lowIL7R^hi), effector memory T cells
   (CD45RO^hiPD1 + HLADR + CD137+), effector T cells (CD45RO-CCR5 + PD1 +
   IL7R^low), early effector T cells (CD69 + CD45RO-CCR5+) and a
   cytotoxic/innate-like (KLRG1+) cluster of T cells ([132]Fig. 4A).
   Individual patient analysis revealed differences in CD4+ T cell
   composition, showing either dominant memory/effector memory T cells or
   effector/early effector T cells in pwMS, which were not observed in the
   control cohort ([133]Fig. 4B). Due to variability in CSF cell counts
   and the potential impact of low cell numbers on clustering algorithms,
   we verified these observations by manual gating in FlowJo® for both the
   discovery cohort and an extended cohort (MS = 6, controls = 5,
   totalling MS = 21 and controls = 15) ([134]Supplementary Fig. S3A). We
   found that these differences were primarily driven by CD45RO and CD69
   expression ([135]Supplementary Fig. S3B). Despite considering batch
   effects, the observations remained consistent within the MS cohort,
   across different investigators, and over varying measurement days.
   There were no significant age variations in our cohort; females were
   more frequent in the control group (persons with IIH;
   [136]Supplementary Fig. S3D).

Figure 4.

   [137]Figure 4
   [138]Open in a new tab

   CD4+ T cell subset analysis in early MS reveals distinct acute and
   chronic CSF endophenotypes. As suggested, we have clarified the
   specific statistical test applied to each dataset in the figure. (A)
   UMAP visualization of CD4+ T cell clusters identified through Louvain
   clustering in Seurat. Heatmap displays the markers used for cluster
   annotation, with cluster labels including CD4+ T cells (T), central
   memory T cells (CM T) and regulatory T cells (Treg). Sample size MS =
   15, control = 10. (B) Bar plot showing the distribution of CD4+ T cell
   subsets in individual patient samples, highlighting differences between
   MS and control groups. (C) Ratio of CD4+ T cell subset counts
   normalized to CD14+ monocyte counts, with values expressed as mean ±
   SEM. Graphs compare control and MS groups, as well as MS1 and MS2
   subgroups, across various parameters (y-axis), highlighting differences
   in CD4+ T cell subset composition relative to stable CD14+ monocyte
   counts. Each data point represents one patient sample in groups as
   indicated. Statistical analysis was conducted using the Shapiro–Wilk
   test for normal distribution, followed by either t-test (MS versus
   ctrl) or one-way ANOVA (Ctrl versus MS1 versus MS2). Sample size MS =
   21 (MS1 = 15, MS2 = 6), Control = 15. Significance levels are denoted
   as follows: ns, non-significant, *P < 0.05, **P < 0.01, ***P < 0.001
   and ****P < 0.0001. (D and E) Clinical routine parameters, with values
   expressed as mean ± SEM. Types of oligoclonal bands are defined as
   follows: type 1 = none, type 2 = exclusive to CSF and type 3 =
   identical in both CSF and serum, with additional bands in CSF. CEL ,
   contrast-enhancing lesion. Statistical analysis was conducted using the
   Shapiro–Wilk test for normal distribution, followed by either t-test
   (IgG; Alb ratio) or Mann–Whitney test (IgM; lesion number, CEL). Each
   data point represents one patient sample in groups as indicated. Sample
   size MS = 21 (MS1 = 15, MS2 = 6). Significance levels are denoted as
   follows: ns, non-significant, *P < 0.05, **P < 0.01, ***P < 0.001 and
   ****P < 0.0001.

   Based on these findings, we categorized MS patients into two CSF
   signature groups: MS1, with a CSF signature favouring memory over
   effector T cells (Memory > Effector), and MS2, with a signature
   favouring effector over memory T cells (Effector > Memory). To
   determine whether this difference arose from a shift within CSF CD4+ T
   cell subsets or from additional invading CD4+ T cells, we calculated
   absolute counts of each T cell subpopulation and normalized these to
   CD14+ monocyte counts ([139]Supplementary Fig. S3B) according to our
   previous work.^[140]6 We then analysed ratios of CD4+ T cell subsets to
   CD14+ counts in controls versus MS and MS1 versus MS2 patients,
   confirming the subset distribution observed with our algorithm
   ([141]Fig. 4C).

   We noted a trend towards increased Treg/CD14+ ratios within the MS1
   subtype, while the MS2 subtype exhibited a higher ratio of CCR5+ Tregs
   ([142]Supplementary Fig. S3C). Comparing CSF CD4+ T cell subsets to
   paraclinical CSF data, we found that the MS1 subtype had a trend
   towards higher intrathecal IgG production compared to the MS2 subtype
   ([143]Fig. 4D). The albumin CSF/serum ratio, an indicator of
   blood-brain barrier integrity, showed no significant difference.
   Additionally, we observed no correlation between CSF/serum oligoclonal
   band IgG patterns and T cell signatures ([144]Fig. 4D). PwMS showed no
   differences in disability, age, or sex at the time of CSF analysis
   ([145]Supplementary Table S7), and MRI T2 lesion load, and
   contrast-enhancing lesions on T1 imaging did not differ between MS1 and
   MS2 ([146]Fig. 4E).

Discussion

   Our comprehensive characterization of CD4+ T cells in experimental
   neuroinflammation and MS reveals that phase-specific markers
   distinguishing chronic states of neuroinflammation are detectable in
   the CSF of individuals with early MS. By integrating single-cell RNA
   sequencing and high-dimensional flow cytometry, we demonstrate that
   in-depth profiling of CD4+ T cell subsets can be effectively achieved,
   providing valuable insights beyond conventional diagnostic parameters.
   High-dimensional spectral flow cytometry, implemented as a feasible
   approach for routine diagnostics, identified distinct CD4+ T cell
   signatures in the CSF, potentially enhancing diagnostic resolution and
   advancing our understanding of disease heterogeneity. These findings
   underscore the potential of CSF cellularity analysis to capture unique
   cellular adaptations associated with neuroinflammation stages, which
   could facilitate the development of personalized endophenotype-driven
   therapeutic strategies in MS. While these results are exploratory, they
   pave the way for further studies with larger cohorts to strengthen
   correlations with prospective clinical data and refine endophenotype
   characterization.

   Over recent years, several CSF scRNA-seq studies from pwMS have
   provided critical insights into the cellular landscape of the CSF.
   Notably, Ostkamp and colleagues^[147]14 investigated immune cells in
   CSF and CNS parenchyma of pwMS, comparing early disease stages, relapse
   and stable disease states under α4-integrin blockade via Natalizumab
   (NTZ). Their work revealed a tissue-resident phenotype across multiple
   lineages including T cells with shared expression of CXCR6 and
   differential regulation of CD103, indicating cellular adaptations
   involved in tissue retention. In CD4+ T cells, CCR5^hi Th17.1 and CD4+
   CD45RA+ T effector memory cells showed high tissue retention capacity.
   Our results build on this foundation, identifying phase markers for
   CD4+ T cell subsets with potential roles in chronic inflammation and
   neuroinflammatory processes in MS.

   Several studies emphasize the role of specific CD4+ T cell subsets,
   particularly Th17 cell subsets, in MS. A recent study demonstrated that
   brain-homing Th17.1 cells were resistant to glucocorticoids and were
   elevated in the peripheral blood of NTZ-treated MS patients, while also
   predominating the CSF of early MS.^[148]31 In our study, by mapping
   EAE-derived markers to the work of Schafflick et al.,^[149]32 we
   identified cluster- and phase-specific alterations that suggest the
   utility of phase markers from C1 to C3 in pinpointing potential
   pathogenic T cells in MS. These findings align with our observation
   that specific markers enriched in EAE reflect similar immune
   adaptations in human MS, further supporting the use of EAE as a model
   to explore CD4+ T cell-related mechanisms in the pathophysiology of MS.

   The advantage of single-cell sequencing lies in its ability to analyse
   the heterogeneity of immune cells with pathogenic potential. However,
   limitations such as cost and the need for large cell numbers constrain
   its applicability in routine diagnostics, especially when dealing with
   CSF, where rare cell numbers and interindividual variability make
   analysis challenging. Spectral flow cytometry, a recently upgraded flow
   cytometry technique, allows for the simultaneous analysis of up to 40
   markers, enabling clustering at a resolution comparable to sequencing
   studies.^[150]33,[151]34 Prior work has demonstrated that CSF immune
   cell composition holds prognostic value and could aid in treatment
   decisions.^[152]6,[153]35,[154]36 This technique has proven effective
   in recent studies investigating blood-derived mononuclear cells,
   including those examining pwMS treated with NTZ^[155]37 or dimethyl
   fumarate,^[156]38 to highlight potentially pathogenic immune cell
   subsets in a treatment-dependent manner. Our flow cytometry cohort
   revealed distinct CD4+ T cell signatures, showing the heterogeneity of
   inflammatory states in pwMS. The most notable differences were found in
   CD69+/CD45RO− (effector) and CD69−/CD45RO+ (memory) T cells,
   highlighting an underappreciated phasic aspect of inflammation in MS.
   While our study primarily focused on early active MS, the potential
   impact of relapse versus remission on CD4+ T cell subsets remains to be
   fully elucidated. CSF immune cell composition may change dynamically
   during different disease phases, with a potential shift in T cell
   activation states, tissue retention characteristics, and
   pro-inflammatory versus regulatory balance. Although our findings
   indicate distinct CD4+ T cell signatures in MS, it is currently unclear
   whether these signatures are phase-dependent or reflect a broader
   spectrum of immune responses. Notably, the MS1 subset in our study
   exhibited a CD4+ T cell composition more comparable to non-inflammatory
   controls, raising the possibility that our observed variations may
   correspond to distinct disease states or different immune adaptations
   within the same phase. Future studies with larger, longitudinally
   sampled cohorts will be critical to disentangle these effects and
   determine whether CSF CD4+ T cell profiles could serve as reliable
   markers for disease stage, relapse risk, or progression to more
   advanced MS phenotypes. Additionally, unbiased bioinformatics
   approaches, such as clustering algorithms and machine learning-based
   stratification, could help refine subgroup classification and validate
   the robustness of MS1 and MS2 groupings. However, these methods require
   significantly larger datasets to ensure statistical power and
   generalizability, particularly when distinguishing between disease
   phases and immune response variability. This will be interesting for
   rarer subsets, which may be underappreciated in our study, such as
   activated vs resting Tregs or cytotoxic CD4+ T cells.

   MS1 and MS2 signatures did not correlate with baseline clinical
   presentations typically observed during the initial MS episode.
   However, conventional CSF antibody analysis revealed differences in
   oligoclonal bands, suggesting that CSF CD4+ T cell signatures may
   capture immune processes beyond standard clinical and paraclinical
   markers. Variability in symptom onset, timing of disease presentation,
   and factors not captured by imaging could introduce confounders,
   potentially explaining why conventional parameters did not align with
   our findings.

   Our results highlight the potential of CSF immune profiling to identify
   pathogenic T cell subsets with greater relevance for disease activity
   and progression. While T2 lesion load and oligoclonal bands aid in
   diagnosis, their prognostic value is limited—oligoclonal bands
   primarily distinguish MS from non-MS, and lesion burden correlates only
   moderately with long-term disability. Identifying CSF-resident immune
   signatures may improve patient stratification and treatment selection,
   particularly as many MS patients require therapy adjustments early in
   the disease course due to treatment failures.

   To further validate these findings, experimental models could help
   clarify the role of chronic phase CD4+ T cells in sustaining
   neuroinflammation. Selective depletion assays in EAE and in vitro
   models may reveal mechanisms underlying T cell persistence, tissue
   retention and resistance to apoptosis. Combining CSF immune profiling
   with longitudinal clinical follow-up and experimental studies could
   yield more precise biomarkers for disease course and treatment
   response, ultimately advancing personalized therapy in MS.

   Despite these promising results, our study has some limitations. The
   single-timepoint analysis of the transcriptomic and phenotypic states
   may not fully capture the dynamic nature of CD4+ T cells, particularly
   transcriptomically silent memory T cells. CSF cells also present
   handling challenges due to their poor viability ex vivo, which may lead
   to rapid cell decline. Ongoing research is exploring more robust
   protocols for handling CSF cells.^[157]39 Additionally, the
   methodological transfer from RNA expression patterns into epitope
   staining for flow cytometry requires sufficient protein expression and
   the availability of specific antibodies, which limits the inclusion of
   some cluster-defining markers. Consequently, some markers could not be
   incorporated into our flow cytometry panel due to the absence of
   appropriate antibodies or intracellular localization. A deeper
   integration of cell surface protein expression and transcription in CSF
   T cells of pwMS—linking surface marker expression with transcriptional
   programmes—could refine our understanding of immune dynamics in
   different disease phases. Future studies using CITE-Seq, which combines
   protein and RNA profiling at the single-cell level, may provide a more
   comprehensive characterization of MS1 and MS2 subtypes. Expanding
   spectral flow cytometry cohorts and applying unbiased bioinformatics
   approaches will be essential to uncover potential correlations between
   disease progression, immune signatures and treatment responses. Chronic
   phase markers are consistently and significantly enriched in MS-derived
   Tregs and CD4⁺ cells—especially within the CSF—supporting their role in
   disease persistence and progression.

   Currently, the prediction on disease course is speculative due to small
   patient numbers in our cohort. It also remains to be elucidated, how
   the clinical relapse (as in most of our cases) or a remission
   influences the CSF cellularity. While our panel focuses primarily on
   CD4+ T cells, other CSF immune cells, such as B cells, warrant further
   analysis to gain a more comprehensive understanding of CSF cellularity
   in MS.

Conclusion

   This study demonstrates that in-depth characterization of CD4+ T cells
   via spectral flow cytometry reveals distinct inflammatory disease
   states in CSF of early MS patients. These findings underscore the
   importance of characterizing CSF cellularity across larger patient
   cohorts to uncover potential correlations between cellular markers and
   clinical markers, offering valuable insights for therapeutic
   implications. Establishing robust CSF handling and expanding marker
   panels to include additional immune cell types, such as B cells, will
   be instrumental in advancing our understanding of the cellular
   landscape in MS and other neuroinflammatory diseases.

Supplementary Material

   fcaf231_Supplementary_Data
   [158]fcaf231_supplementary_data.zip^ (3.2MB, zip)

Acknowledgements