Integrin β1 Demarks Precursors of Brain-Residing Antibody-Secreting Cells in Multiple Sclerosis

Abstract

Background and Objectives

Multiple sclerosis (MS) is characterized by CXCR3⁺ memory B cells that infiltrate the CNS to mature into antibody-secreting cells (ASCs). It remains elusive how to benefit from this disease hallmark as a prognostic tool. We aimed to uncover markers that reflect B-cell entry and maturation in the CNS and, on that basis, explore associations with response to high-efficacy MS therapies.

Methods

Ex vivo single-cell technologies were applied to blood samples and postmortem CNS suspensions from MS and control donors, either treated or untreated. In addition, in vitro assays were performed on blood samples from healthy individuals to functionally asses B-cell activation, differentiation, and transmigration.

Results

Single-cell RNA sequencing revealed ITGB1 (integrin β1/CD29) as a major discriminator of CXCR3⁺ memory B cells in blood, which was validated using spectral flow cytometry. CD29⁺CXCR3⁺ memory B cells preferentially crossed a brain microvascular endothelial layer and were highly receptive to T-cell help to become ASCs in vitro. In contrast to blood, CD29 and CXCR3 coexpression was restricted to ex vivo ASCs derived from CSF, meninges, and brain tissue of people with MS, which was supported by enhanced CD29 upregulation on in vitro–differentiated ASCs. While CD29 was downregulated on CXCR3⁺ memory B cells that accumulated in the blood of natalizumab-treated patients, CD29 levels were only reduced on CXCR3⁺ memory B cells that repopulated in cladribine-treated patients without early disease activity.

Discussion

These findings put forward CD29 as a putative marker on CXCR3⁺ precursors of CNS-residing ASCs for predicting treatment responses in MS.

Introduction

Multiple sclerosis (MS) is a chronic inflammatory disease of the CNS, involving demyelination and subsequent lesion formation.¹ B cells are important players in its pathophysiology, as evidenced by the continuous production of unique IgG oligoclonal bands in the CSF² as well as the presence of B-cell clones in the meninges and brain tissue of people with MS (pwMS),^3-5 which associate with worse disease outcomes.⁶ Although their accumulation in the CNS is an established hallmark of MS,⁷ the prognostic value of B cells remains largely unknown, limiting opportunities to refine and optimize the effect of current B cell–targeted therapies.

Previously, memory B cells that infiltrate the CNS of pwMS have been shown to express high levels of the chemokine receptor CXCR3.⁸ Once arrived in the CNS, these types of memory B cells likely mature into antibody-secreting cells (ASCs) that can persist and may trigger inflammation via intrathecal IgG and/or cytokine secretion.^9,10 In mice, a chronically inflamed CNS can provide survival niches for ASCs,¹¹ which also seems to be the case for pwMS.^6,9,12 While the anti-VLA4 therapy natalizumab effectively prevents the migration of circulating memory B cells to the CNS,⁸ the immune reconstitution therapy cladribine has brain-penetrant capacities¹³ and thus the potential to preferentially target local memory B cells and suppress the presence of intrathecal oligoclonal bands in MS.^14,15 Yet, relapses and new lesions reappear years after cladribine treatment, highlighting the need for well-considered selection and timing of B cell–targeted therapies to reach optimal effects in pwMS.

Uncovering features of long-lived ASCs in the CNS, and especially their precursors in the circulation, could yield new targets to improve treatment for pwMS. The aim of this study was to further expand our understanding of CNS-homing CXCR3⁺ memory B cells,⁸ by looking for additional molecules that could explain their ability to enter, reside, and become ASCs within the CNS in pwMS. Using different single-cell technologies, patient cohorts, and in vitro models, we reveal that integrin β1 (CD29) marks both CNS-persistent ASCs and their CXCR3⁺ precursors and can be used to monitor treatment responses in MS.

Methods

Study Design and Participants

Healthy donors and individuals with MS, untreated, treated IV with natalizumab, or treated orally with cladribine tablets, were selected for this study. Cladribine-treated patients were stratified based on the occurrence of early disease activity during repopulation, defined as the detection of new or enhancing lesions on follow-up MRI-scans performed 5–26 months after treatment. In our natalizumab cohort, patients remained stable. Disease activity was recorded in only 2 cases: one during treatment cessation after pregnancy and the other days after treatment initiation. For ex vivo transcriptomic analyses with scRNA-seq, we used thawed blood samples of healthy donors. For ex vivo phenotyping with spectral flow cytometry, we freshly obtained postmortem blood, CSF, meninges and brain tissues from MS brain donors (Netherlands Brain Bank). In addition, we freshly obtained blood or used thawed blood samples of healthy and MS donors. Natalizumab-treated and cladribine-treated MS donors with antidrug antibodies in serum were excluded. For in vitro assays, we either received fresh blood (Sanquin) or used thawed blood samples of healthy and MS donors. Given the higher prevalence of MS in women compared with men,¹⁶ our study included a larger proportion of female participants. Clinical information on all participants was extracted from patient records (eTable 1).

Isolation of Mononuclear Cells From Peripheral Blood and CNS Compartments

Peripheral blood was collected and processed as described previously.⁸ Peripheral blood mononuclear cells (PBMCs) were isolated according to the manufacturer's protocol and either freshly used or frozen in liquid nitrogen for later use. CSF, leptomeninges, and brain tissues were collected, processed, and prepared into single-cell suspensions following previously described methods.⁸ For enzymatic digestion, leptomeninges and brain tissues were incubated with DNAse (Roche, 100 U/mL) and collagenase IV (Worthington, 1.5 mg/mL) for one hour (37°C, 5% CO₂). Single-cell suspensions were used for spectral flow cytometry.

Single-Cell RNA Sequencing

For scRNA-seq analyses, PBMCs from healthy donors (n = 3) were stained with fluorochrome-labeled monoclonal anti-human antibodies to allow for fluorescence-activated cell sorting (FACS) of viable CD19⁺ B cells. (eTable 2). Viable CD19⁺ B cells were sorted in pure fetal calf serum (FCS) using a FACSAria Fusion cell sorter (BD Biosciences). Cells were washed and resuspended in RPMI with 10% FCS, to continue single-cell processing. After cell sorting, cells were screened for viability prior to loading cells for partitioning. Single-cell samples were loaded on a 10xGenomics Next GEM K chip to aim for a final recovery of 10,000 cells. Final 5’ gene expression (5’ GEX) libraries were generated according to the Chromium Next GEM Single Cell 5’ Kit (v2 chemistry, dual index) User Guide (CG000330, 10x Genomics). Libraries were sequenced on a NovaSeq 6000 platform (Illumina), paired-end mode with a 28-10-10-90 cycle setting, with a sequencing depth of around 25,000 reads per cell. 10x Genomics software Cell Ranger (version v7.2.0) was used to preprocess the raw sequencing data and align reads to reference (GRCh38–1.2.0), with setting to include introns, finally resulting in GEX data sets. After processing in Cell Ranger software, the UMI count matrices were further filtered with the Seurat R package (v.5.1.0). The filtering criteria included genes expressed in at least one cell, cells with more than 300 detected features, over 500 UMIs, <25% mitochondrial reads, and >10% ribosomal reads. The data quality was further filtered after data normalization and dimensional reduction steps. First, the RNA data were log-normalized, and the highly variable features were identified using the “FindVariableFeatures” function of the Seurat package (vst method, n = 2,000). Those highly variable features were then used to scale the expression data and perform principal component analysis (PCA) calculations. The samples were split and integrated with the anchor-based reciprocal PCA (RPCA) method. After integration, unbiased shared nearest neighbor analysis was performed using the first 30 PCA dimensions and cluster estimation was conducted with Leiden algorithm at a resolution of 0.8. Each cluster was visualized using cell embedding generated by the uniform manifold approximation and projection method. Possible T cells were filtered out based on the expression of CD3E and CD3D. In addition, differentially expressed features for each cell cluster, compared with all other cells, were identified using the “FindAllMarkers” function, with an adjusted p value threshold of 0.05. One cluster that was discriminated by the monocyte-related and macrophage-related markers CST3 and LYZ was filtered out. We ended up with 24,220 cells in the final data set, with UMI counts within the normal range (1,000–4,000 nCounts RNA) for 98.5% of all cells. The abovementioned steps of data normalization and dimensional reduction were repeated, and final cell embedding was achieved with the first 24 PCA dimensions. Further cell annotations were achieved using the Azimuth R package (v.0.5.0).

To zoom in on cells related to our research question, naive B cells were filtered out based on Azimuth annotation, leaving 12,397 non-naive B cells in the data set. This data set was further split based on CXCR3 expression into 11,378 CXCR3⁻ and 1,019 CXCR3⁺ cells. We performed differential expression analysis to compare CXCR3⁻ and CXCR3⁺ cells using the Seurat default nonparametric Wilcoxon rank-sum test, with genes filtered based on a Bonferroni-adjusted p value cutoff of 0.001 and an absolute log₂(fold change) < −0.5 or >0.5. The log₂ (fold-change) threshold was determined based on its biological relevance within memory B-cell subset comparisons and in accordance with other studies.^17,18 For visualization, all plots were computed using the ggplot2 R package (v.3.5.1) and Seurat R package.

In Vitro Transmigration Assay

CD19⁺ B cells were isolated from buffy coat–derived PBMCs using the B-cell isolation kit I (Miltenyi Biotec) according to the manufacturer's protocol. Subsequently, CD27⁺ memory B cells were sorted by FACS using fluorochrome-labeled monoclonal anti-human antibodies (eTable 2) on a FACSAria Fusion cell sorter (BD Biosciences). As described previously,¹⁹ 5*10⁵ cells were placed on confluent monolayers of human brain endothelial cells (hCMEC/D3) on 5-µm pore size Transwell plates (Corning Life Sciences). RPMI 1640 medium (0.5% BSA) with or without CXCL10 was added to the bottom of the filters. After 5-hour incubation (37°C, 5% CO₂), the migrated and the nonmigrated fraction of memory B cells were analyzed by spectral flow cytometry.

In Vitro Differentiation Assay

Total CD27⁺ memory B cells as well as CD29^−/+CXCR3^−/+ memory B-cell fractions were sorted from MS and/or healthy blood by FACS using fluorochrome-labeled monoclonal anti-human antibodies (eTable 2) on a FACSAria Fusion cell sorter (BD Biosciences) and a FACSAriaIII sorting machine (BD Biosciences). As reported previously,⁸ 1*10³ cells were cocultured with irradiated 3T3-CD40L fibroblasts and stimulated with soluble IL-21 (Thermo Fisher Scientific, 50 ng/mL). After 6-day incubation (37°C, 5% CO₂), cells were analyzed by spectral flow cytometry.

Phosphoflow

A total of 1 × 10⁶ PBMCs were seeded in RPMI+HEPES 1640 medium (10% FCS) in 96-well round-bottom plates. For stimulation, the plate was placed in a water bath at 37°C. The cells were stimulated for 2 minutes with either anti-human Ig-kappa (20 µg/mL, Southern Biotech) and anti-human Ig-lambda (20 µg/mL, Southern Biotech), or soluble recombinant human CD40L (2 µg/mL, R&D Systems) together with soluble recombinant human IL-21 (50 ng/mL, Thermo Fisher), or left unstimulated. Afterward, cold buffer was added and cells were directly placed on ice. The cells were washed and stained extracellularly in FACS buffer for 30 minutes at 4°C in the dark using fluorochrome-labeled monoclonal anti-human antibodies (eTable 3). Then, the cells were washed and stained with the Zombie NIR Fixable Viability Kit (Biolegend) for 10 minutes at 4°C in the dark. Subsequently, the cells were fixed for 30 minutes at 4°C in the dark and permeabilized for 20 minutes on ice using the Transcription Factor Phospho Buffer Set (BD Biosciences). The cells were stained with phospho-S6 ribosomal protein (Ser240/244) (D68F8) XP Rabbit mAb (Cell Signaling Technology) in Perm buffer for 30 minutes at 4°C in the dark, followed by incubation with R-Phycoerythrin AffiniPureTM F(ab’)2 Fragment Donkey Anti-Rabbit IgG (H+L) (Jackson ImmunoResearch) for 15 minutes. Cells were measured on a 5-laser Cytek Aurora flow cytometer (Cytek Biosciences). Data analysis was conducted using OMIQ software from Dotmatic.

Spectral Flow Cytometry

For ex vivo measurements, 2 × 10⁶ cells were seeded in 96-well round-bottom plates. For measurements after in vitro assays, 1 × 10³–5 × 10⁵ cells were seeded in 96-well conically bottom plates. Cells were consequently stained with the Zombie NIR Fixable Viability Kit (Biolegend) for 10 minutes at room temperature in the dark. Subsequently, the cells were washed and stained extracellularly in BSB or FACS buffer for 10 minutes at room temperature in the dark using fluorochrome-labeled monoclonal anti-human antibodies (eTable 3). Afterward, the cells were fixed and permeabilized using the eBioscience Foxp3 Transcription Factor Staining Buffer Set (Invitrogen) according to the manufacturer's protocol. Then, the cells were washed and stained intracellularly in Perm buffer at 4°C in the dark using fluorochrome-labeled monoclonal anti-human antibodies (eTable 3). Cells were measured on a 5-laser Cytek Aurora flow cytometer (Cytek Biosciences). Data analysis was conducted using OMIQ software from Dotmatic or FlowJo software.

Statistical Analysis

GraphPad Prism 9 (Graphpad Software) and Rstudio (version 4.3.2., Posit) were used for statistical analysis. For comparing between 2 groups, nonparametric Wilcoxon signed-rank tests or Mann-Whitney U tests were applied. For comparing more than 2 groups, nonparametric Kruskal-Wallis or Friedman tests were applied in combination with the Dunn post hoc test. Correlations were analyzed using the Spearman correlation coefficient. A p value <0.05 (2-tailed) was considered statistically significant. Statistical significance is represented in all graphs using asterisks, and all data are shown using the median or standard error of the mean (SEM).

Standard Protocol Approvals, Registrations, and Patient Consents

All donors provided written informed consent. Study protocols were approved by the medical ethics committee of the Erasmus Medical Center (2019-0845; 2021-0251; 2021-0946) and the VU University Medical Center (2009-148).

Data Availability

The raw scRNA-seq data and processed matrix files reported in this study are available on the Gene Expression Omnibus under accession number GSE292997. Other experimental data will be made available from the corresponding author on reasonable request.

Code Availability

The code used for scRNA-seq analysis is available on GitHub (github.com/YFWang-YvH/Integrin_b1_demarks_precursors_of_brain_residing_antibody_secreting_cells_in_multiple_sclerosis).

Results

Integrin β1 (CD29) Expression Is Discriminative for CXCR3⁺ Memory B Cells in Human Blood

To explore discriminating features of CXCR3-expressing memory B cells, we conducted single-cell RNA sequencing (scRNA-seq) on purified B cells from healthy donor blood (eFigure 1A). After preprocessing quality controls and standard analyses, naive and memory B-cell clusters were computationally identified using Azimuth annotations (Figure 1A). The cluster identity matched with canonical marker expression of TCL1A, IGHD, and CD27. Within the CD27-expressing B-cell clusters, a population of CXCR3-expressing cells was identified. After naive B-cell exclusion, we performed differentially expressed gene (DEG) analysis on populations with and without CXCR3 transcript, which yielded 89 genes that were differentially expressed (adjusted p value <0.001) and had an absolute log₂ (fold change) < −0.5 or >0.5 (Figure 1B; eTable 4; eFigure 2). Several genes associated with B-cell activation and differentiation were elevated in CXCR3⁺ memory B cells, including the costimulatory marker CD86 (CD86), the CXCR3-inducing transcription factor T-bet (TBX21), and the plasma cell differentiation–associated transcription factor XBP1 (XBP1). A highly differentially expressed gene was ITGB1 (integrin β1; CD29). Integrin β1 is part of the transmembrane glycoprotein signaling receptor family and is known to form heterodimers with other integrins.²⁰ By interacting with CD49d (integrin α4), it forms the very late antigen-4 (VLA-4) complex, which is critical for immune cell migration.^21,22

Figure 1. Integrin β1 (CD29) Is Coexpressed With CXCR3 on Memory B Cells in Human Blood.

Peripheral blood mononuclear cells (PBMCs) were isolated from blood of healthy donors and frozen at −80°C. Ex vivo spectral flow cytometry was performed on total PBMCs. Single-cell RNA sequencing (scRNA-seq) was conducted on CD19⁺ B cells sorted with fluorescence activated cell sorting (FACS). (A) A uniform manifold approximation and projection (UMAP) is presented, embedding 24,220 B cells from healthy donors (n = 3) depicting TCL1A, IGHD, CD27, and CXCR3 expression. Naive B-cell (11,823), intermediate B-cell (8,035), memory B-cell (4,288), and plasmablast (74) identities were annotated using the Azimuth R package (v.0.5.0). (B) A volcano plot illustrates differentially expressed genes (DEGs) for CXCR3-expressing (transcript reads >0) cells after naive B-cell cluster exclusion on Wilcoxon rank-sum analysis. Dashed lines represent significance (p adjusted <0.001) and an absolute log₂ (fold change) < −0.5 or >0.5. (C–D) Ex vivo spectral flow cytometry was performed on total PBMCs from healthy donors (n = 6). CD29 MFI (median) is presented on different B-cell subpopulations. Data were analyzed using Friedman and Dunn post hoc tests (C–D) or Wilcoxon signed-rank tests (D).

To better understand the role of CD29 and its coexpression with CXCR3 on B-cell subsets, we performed spectral flow cytometry on healthy human blood samples (eFigure 1B). Compared with transitional and naive mature subsets, CD29 expression was approximately 2-fold higher on memory B cells (Figure 1C). Within the memory population, we defined IgM⁺, IgG⁺, and IgA⁺ cells and found that CD29 expression was highest in the IgG⁺ population (Figure 1C). As anticipated from the scRNA-seq analysis, CXCR3⁺ memory B cells showed significantly higher CD29 expression compared with CXCR3⁻ memory B cells (Figure 1D). The increase in CD29 expression on CXCR3⁺ vs CXCR3⁻ memory subsets was evident for IgM⁺ and IgG⁺ but not for IgA⁺ memory B cells (Figure 1D).

CD29 and CXCR3 Coexpression Renders Circulating Memory B Cells More Receptive to T-Cell Help

To study whether memory B cells with or without CD29 and CXCR3 expression behave differently, we first assessed the in vitro responsiveness to BCR (anti-Ig) and T-cell costimulatory signals (anti-CD40 and IL-21) using healthy human blood samples (eFigure 3A). We measured phosphorylation of ribosomal protein S6, as this protein is activated by both signaling routes in B cells after in vitro anti-BCR and T cell–mimicking triggers (eFigure 3B).²³ CXCR3⁺ memory B cells showed slightly but significantly increased basal pS6 signaling compared with CXCR3⁻ counterparts, while on stimulation, only T cell–mimicking triggers significantly enhanced S6 phosphorylation (eFigure 3C; Figure 2A). This suggests reduced sensitivity of CXCR3⁺ memory B cells to BCR triggers, in contrast to its receptiveness to T cell–dependent triggers. Evaluating the effect of CD29 and CXCR3 coexpression, we found that CD29⁺CXCR3⁺ memory B cells were more responsive to both anti-BCR and T cell–mimicking triggers (eFigure 3D; Figure 2B). The expression of antigen presentation markers HLA-DR and CD74, as well as stimulatory molecules CD40 and CD86, was increased on CXCR3⁺ compared with CXCR3⁻ memory B cells (Figure 2, C and D). Within the CXCR3⁺ memory B cells, CD29⁺ B cells displayed higher expression levels of these markers compared with CD29⁻ counterparts (Figure 2, C and D). Of note, CD86 (CD86) was one of the discriminative transcripts expressed in CXCR3⁺ vs CXCR3⁻ memory B cells (Figure 1C). We found that CD70 was somewhat downregulated and CD80 upregulated on CXCR3⁺ vs CXCR3⁻ memory B cells, while both showed higher levels in the presence of CD29 (eFigure 3, E and F). T-bet and its associated markers CD19, CD11c, CD85j, and TLR9 were upregulated in CD29⁺CXCR3⁺ memory B cells as well, while CD21 was downregulated (eFigure 4), which is in line with the differential transcript levels of T-bet (TBX21) in CXCR3⁺ vs CXCR3⁻ memory B cells (Figure 1B). Collectively, these findings corroborate that CD29⁺CXCR3⁺ memory B cells display an increased sensitivity for T cell–dependent triggers.

Figure 2. CD29⁺CXCR3⁺ Memory B Cells Are Prone to T-Cell Interaction.

Thawed PBMCs from healthy donors (n = 10) were stimulated with anti-BCR or anti-CD40 and IL-21. Phosphoflow was performed to assess phospho-S6 (pS6) expression in CD29^−/+CXCR3^−/+ memory B-cell populations. (A) Percentages of pS6⁺ within unstimulated, anti-BCR–triggered, and anti-CD40 and IL-21–triggered CXCR3^−/+ memory B cells are presented. (B) Percentages of pS6⁺ within unstimulated, anti-BCR–triggered, and anti-CD40 and IL-21–triggered CD29^−/+ CXCR3^−/+ memory B cells are presented. (C–D) MFI (median) of antigen presentation–related markers HLA-DR and CD74, (C) and costimulatory molecules CD40 and CD80 (D), is presented on CD29^−/+CXCR3^−/+ memory B-cell populations. Data were analyzed using Wilcoxon signed-rank tests.

CD29⁺CXCR3⁺ Memory B Cells Preferentially Cross Brain Endothelial Monolayers and Differentiate Into ASCs In Vitro

After interacting with T helper cells in secondary lymphoid organs, CXCR3⁺ memory B cells likely home into the CNS of pwMS.²⁴ To study how this is associated with the presence of CD29, we compared the ability of memory B cells from healthy donors with and without CD29 and CXCR3 expression to cross a brain microvascular endothelial layer as an in vitro model of the blood-brain barrier (eFigure 5A). To facilitate migration, CXCL10 was used as a chemoattractant in this assay as CXCL10 is shown to be enriched in MS CSF and brain tissue.^25-27 We confirmed that CXCL10 did not affect preferential survival of CD29^+/−CXCR3^+/− memory B-cell subpopulations (data not shown). We observed a selective enrichment of the CD29⁺CXCR3⁺ subset in the CXCL10-migrated compartment, both in comparison with and after correcting for the ex vivo baseline measurement (Figure 3A; eFigure 6).

Figure 3. CD29⁺CXCR3⁺ Memory B Cells Show Enhanced Transmigration and ASC Differentiation In Vitro.

PBMCs were freshly isolated from healthy buffy coats (n = 6). Memory B cells were sorted using automated cell separation and fluorescence-activated cell sorting (FACS). Sorted cells were used for a 5-hour in vitro transmigration assay, with and without addition of CXCL10 in the medium. Spectral flow cytometry was performed to analyze ex vivo memory B cells, migrated fractions, and nonmigrated fractions. (A) Percentages of CD29^−/+CXCR3^−/+ populations are shown within ex vivo memory B cells, migrated fractions, and nonmigrated fractions. (B) PBMCs were isolated from healthy donor blood (n = 8) and frozen at −80°C. CD29^−/+CXCR3^−/+ memory B-cell populations were sorted from thawed PBMCs using fluorescence-activated cell sorting (FACS). Subsequently, these cells were used for a 6-day in vitro coculture with 3T3-CD40L mouse fibroblasts and soluble IL-21. Spectral flow cytometry was performed to analyze formed ASCs. ASC count and percentage are shown within CD19⁺ cells. (C) IgG⁺ percentage is shown within ASCs. IgG secretion in the supernatant are shown. Data are presented as mean ± standard error of the mean (SEM). Data were analyzed using Wilcoxon signed-rank tests (A), as well as Kruskal-Wallis and Dunn post hoc tests (B–C). ASC = antibody-secreting cell; PBMC = peripheral blood mononuclear cell.

On entry into the CNS, CXCR3⁺ memory B cells are likely to encounter T helper cells in the meninges and perivascular spaces, promoting their maturation into antibody producers and triggering inflammation.⁹ This potential seems to be restricted to the expression of both CXCR3 and CD27, which are defining features of CNS-homing B cells in MS.²⁸ As compared with CXCR3-expressing CD27⁺ memory B cells, CXCR3-expressing CD27⁻IgD⁻ (double-negative; DN) B cells show higher levels of T-bet and TLR9, but lower levels of CD40 (eFigure 7, A and B). The latter likely makes DN B cells less responsive to T cell–dependent signals than their CD27⁺ counterparts for differentiating into ASCs (eFigure 7C).

To assess the impact of CXCR3 and CD29 expression by (CD27⁺) memory B cells on ASC outgrowth, CD29⁻CXCR3⁻, CD29⁻CXCR3⁺, CD29⁺CXCR3⁻, and CXCR3⁺CD29⁺ memory subsets were separately isolated from healthy blood and cultured under T cell–dependent conditions for 6 days (eFigure 5B). ASCs were identified as being CD27^highCD38^high and were discriminated based on IgG or IgM isotype expression (eFigure 5B). ASC frequencies were significantly increased in the cocultures with CD29⁺CXCR3⁺ memory B cells, as determined by both absolute counts and percentages (Figure 3B). Moreover, CD29⁺ memory B cell–derived ASCs showed relatively higher IgG expression, which was further enhanced by coexpression with CXCR3 (Figure 3C). This was supported by an increased IgG secretion in the cocultures with CD29⁺CXCR3⁺ memory B cells (Figure 3C), which correlated with the percentage of ASCs induced (eFigure 5C). CD29⁺CXCR3⁺ memory B cells already exhibit higher IgG expression at baseline (Figure 1D), most likely reflecting their IgG production in vitro. Taken together, CD29⁺CXCR3⁺ memory B cells show enhanced in vitro capacity to cross brain microvascular endothelium and become IgG-producing ASCs under T cell–dependent conditions.

CD29 and CXCR3 Are Coexpressed on Ex Vivo Circulating Memory B Cells but Not ASCs in pwMS

Next, we compared the CD29 and CXCR3 expression profile between memory B cells and terminally differentiated ASCs in the blood of progressive, treatment-naïve pwMS. Using fresh blood samples of the same individuals, we found an enrichment of CD29 on CXCR3⁺ memory B cells (Figure 4A; eFigure 8A). By contrast, CD29 and CXCR3 coexpression on ASCs was clearly diminished (Figure 4B; eFigure 8B). A similar trend was observed for healthy donors (eFigure 8, C and D). This indicates that, at least in the periphery, coexpression of CD29 and CXCR3 as observed on the memory B cell is not conserved on ASCs.

Figure 4. CD29 and CXCR3 Are Coexpressed on Circulating Memory B Cells but Not ASCs in MS.

The authors freshly obtained blood from untreated progressive MS donors (n = 19) and performed ex vivo spectral flow cytometry. (A) Percentages of CD29⁻ and CD29⁺ populations are indicated within CXCR3⁺ memory B cells. CD29 MFI (median) is shown on CXCR3⁻ and CXCR3⁺ memory B cells. (B) Percentages of CD29⁻ and CD29⁺ populations are indicated within CXCR3⁺ ASCs. CD29 MFI (median) is shown on CXCR3⁻ and CXCR3⁺ ASCs. (D) CD27⁺ memory B cells were sorted from thawed PBMCs of healthy individuals (n = 6), as well as relapsing MS donors (n = 6) and in vitro cocultured under T cell–dependent conditions (3T3-CD40L mouse fibroblasts and soluble IL-21). MFI (median) of CXCR3 and CD29 are shown on memory B cells at day 0 as well as non-ASCs and ASCs at day 6. Data are presented as mean ± standard error of the mean (SEM). Data were analyzed using Wilcoxon signed-rank tests (A and B) or Mann Whitney U tests (D). ASC = antibody-secreting cell; MS = multiple sclerosis.

Memory B Cells of PwMS Show Stronger CD29 Upregulation on T Cell–Dependent In Vitro Differentiation Into ASCs

To assess intrinsic differences between pwMS and healthy individuals in regulating CD29 and CXCR3 expression on ASCs, we cocultured blood memory B cells from these groups under T cell–dependent conditions (eFigure 4E). First, we evaluated CXCR3 expression, which did not differ both on ex vivo memory B cells and on in vitro–induced CD38^highCD27^high ASCs (Figure 4C). On the contrary, CD29 expression was significantly higher on ASCs from MS donors as induced in vitro, despite no difference being observed on memory B cells ex vivo (Figure 4D). Taken together, although no ex vivo differences were observed between MS and healthy donors, memory B cells from MS donors may be intrinsically more capable of upregulating CD29 on their surface during ASC maturation in response to T cell–dependent signals.

CD29 and CXCR3 Coexpression Is an Exclusive Characteristic of Ex Vivo ASCs in the CNS of pwMS

To investigate whether CD29⁺CXCR3⁺ memory B cells in MS actually migrate to the CNS and locally differentiate into ASCs, we analyzed B cells and ASCs that were freshly isolated from blood, CSF, meninges, normal-appearing white matter (NAWM), and white matter lesions of deceased progressive MS donors (eFigure 9A). In line with previous research,⁹ B-cell frequencies were reduced, while ASC frequencies were increased in CNS compartments compared with blood, most prominently in white matter lesions (eFigure 9, B and C). In contrast to our findings on memory B cells in MS blood, little CD29 and CXCR3 coexpression was observed on B cells in CNS compartments, especially in meninges, NAWM, and white matter lesions (Figure 5, A and B; eFigure 10A). CD29⁺CXCR3⁺ B cells in blood and CNS compartments of MS donors exhibited high expression of T-bet–related markers compared with other subsets (eFigure 9D). It is of interest that while CD29 and CXCR3 were not coexpressed on ASCs in MS blood, coexpression was observed as a dominant trait on ASCs in all CNS compartments (Figure 5, C and D; eFigure 10B). In one MS donor where we performed more detailed spectral flow analysis, we verified that these types of ASCs mainly express IgG (eFigure 11) and show an even higher CXCR3 and CD29 coexpression on CD138⁺ plasma cells (eFigure 12).⁹ These data support the model that after their entry into the CNS, CD29⁺CXCR3⁺ memory B cells have a marked propensity to further differentiate into IgG⁺ ASCs in pwMS. In addition, CD29 and CXCR3 coexpression is a distinct feature of CNS-residing ASCs compared with local B-cell and blood-derived counterparts.

Figure 5. CD29 and CXCR3 Coexpression Is Restricted to ASCs in CNS Compartments of MS Donors.

The authors freshly obtained blood, CSF, meninges, normal-appearing white matter (NAWM), and white matter lesions from progressive MS donors (n = 10). Single cell suspensions were isolated from each compartment, followed by ex vivo spectral flow cytometry. (A–B) Representative gating and percentages of CD29⁺CXCR3⁺ populations within B cells from blood and CNS compartments are shown. (C–D) Representative gating and percentages of CD29^−/+CXCR3^−/+ populations within ASCs from blood and CNS compartments are shown. Data are presented as mean ± standard error of the mean (SEM). Data were analyzed using Kruskal-Wallis and Dunn post hoc tests (B and D). ASC = antibody-secreting cell.

Natalizumab Treatment Reduces CD29 Expression on CXCR3⁺ Memory B Cells That Accumulate in the Blood of pwMS

To investigate how this proposed B-cell mechanism correlates with MS therapies, we evaluated the impact of common MS disease-modifying treatments on CD29 and CXCR3 expression on blood memory B cells. Natalizumab is directed against integrin α4 as part of the VLA-4 complex, thereby hindering lymphocytes from migrating to the CNS,²⁹ resulting in a selective accumulation of circulating CXCR3⁺ memory B cells.^8,30 CD29 is part of the VLA-4 complex and might, therefore, be alternatively regulated on effective VLA-4 targeting. We performed spectral flow cytometric analysis of B cells from paired blood samples of MS donors before and 6 months after natalizumab treatment (Figure 6A). We confirmed the selective increase of circulating CXCR3⁺ memory B cells after treatment with natalizumab (Figure 6B; eFigure 13A). It is of interest that CD29 expression was reduced on the CXCR3⁺ memory B-cell pool compared with the status prior to natalizumab treatment (Figure 6C; eFigure 13B). These data demonstrate that treatment of active MS with natalizumab effectively reduces CD29 expression on B cells, including CXCR3⁺ memory populations that selectively accumulate in the circulation after this treatment.

Figure 6. Natalizumab Therapy Affects CD29 Expression on CXCR3+ Memory B Cells in MS.

(A) Spectral flow cytometry was performed on thawed PBMCs from blood of relapsing MS donors (n = 8) before and 6 months after natalizumab treatment. Created in BioRender. Van luijn, M. (2025) BioRender.com/q8qiv23. (B) Frequencies of total and CXCR3⁺ memory B cells are depicted before and after natalizumab treatment. (C) Quantifications of CD29 MFI (median) on CXCR3⁻ and CXCR3⁺ memory B cells are shown before and after natalizumab treatment. Data are presented as mean ± standard error of the mean (SEM). Data were analyzed using Wilcoxon signed-rank tests (B–C).

CD29 Is Only Downregulated on Repopulating CXCR3⁺ Memory B Cells in the Blood From pwMS Who Respond to Cladribine

In MS, cladribine serves as a highly effective immune reconstitution therapy that significantly affects B-cell frequencies.³¹ In contrast to natalizumab treatment, activity recurrence is seen in people treated with cladribine during prolonged follow-up after immune reconstitution.^32,33 We performed spectral flow cytometry on blood samples from both untreated and cladribine-treated MS patients after B-cell repopulation (Figure 7A). We stratified patients based on the occurrence of early disease activity during repopulation, defined as the detection of new or enhancing lesions on follow-up MRI-scans performed 5–26 months after treatment. Cladribine treatment resulted in lower frequencies of memory B cells compared with untreated patients, while CXCR3⁺ memory B-cell frequencies were not affected (Figure 7B; eFigure 14A). In contrast to repopulating CXCR3⁻ memory B cells, reduced CD29 levels were observed on repopulating CXCR3⁺ memory B cells of cladribine-treated patients without new MRI lesions during follow-up compared with untreated patients and cladribine-treated patients who developed new MRI lesions (Figure 7C; eFigure 14B). These data suggest that phenotypic profiling of repopulating B cells for CD29 and CXCR3 could offer a theragnostic biomarker to evaluate treatment response after cladribine treatment prior to occurrence of tissue damage.

Figure 7. Cladribine Therapy Affects CD29 Expression on CXCR3⁺ Memory B Cells in MS.

(A) Spectral flow cytometry was performed on thawed PBMCs from blood of both untreated (n = 6) and cladribine-treated relapsing MS donors (n = 8). Cladribine-treated patients were labeled as “stable” or “active” based on rebaseline MRI. Created in BioRender. Van luijn, M. (2025) BioRender.com/mt605oi. (B) Frequencies of total and CXCR3⁺ memory B cells are depicted in untreated and stable/active cladribine-treated patients. (C) Quantifications of CD29 MFI (median) on CXCR3⁻ and CXCR3⁺ memory B cells are shown in untreated and stable/active cladribine-treated patients. Data are presented as mean ± standard error of the mean (SEM). Data were analyzed using Kruskal-Wallis and Dunn post hoc tests (B–C).

Discussion

Although CNS-homing CXCR3⁺ B cells have been identified as key players in MS pathophysiology,^8,9 it remains uncertain how such populations locally persist and become ASCs and can be used to make current MS treatment more effective.

Identifying novel, more specific cellular biomarkers that define their pathogenic effector mechanisms is crucial to answer these key questions, as this could not only improve treatment selection but also enhance our ability to predict treatment outcomes. In this study, we highlight CD29 as an important marker on CXCR3⁺ B cells, as coexpression of these markers resulted in enhanced migration capacity and differentiation to ASCs. Moreover, we show indications that natalizumab and cladribine modulate CD29 expression and that CD29 regulation might relate to treatment responsiveness in cladribine-treated pwMS. These findings provide a more precise definition of B-cell pathogenicity in MS, enabling translational studies to further investigate their role in disease progression and explore their potential as a first theragnostic marker for better predicting responses to MS therapies.

Using scRNA-seq and spectral flow cytometry, we found that integrin β1 (CD29) is coexpressed with CXCR3 on memory B cells in healthy and MS blood. CD29 is mostly known for its adhesive function in lymphocyte trafficking as part of the VLA-4 complex together with integrin α4 (CD49d).³⁴ In allergies, CD29 expression has been proposed to explain the efficacy of B-cell homing to inflamed tissues,³⁵ a process that may be similar for the inflamed MS CNS. CXCR3 and CD49d expression correlates with ALCAM expression on circulating B cells,³⁶ which has been reported to mediate B-cell trafficking across CNS barriers,³⁷ suggesting that additional integrins may contribute to B-cell migration into the CNS. Yet, integrins also appear to play a crucial role in determining additional effector functions of immune cells.²² For example, CD29 identifies a distinct subset of T cells with cytotoxic features in different human tissues.^38,39 Furthermore, integrin β1, as part of the integrin α5β1 complex, has a critical role in efficient CD40⁻CD40L signaling,⁴⁰ a signaling pathway that is essential for T cell–dependent B-cell activation.

In MS, CXCR3⁺ memory B cells preferentially infiltrate the CNS,⁸ likely followed by interaction with CNS-resident CD4⁺ T cells to further differentiate into ASCs and produce antibodies.⁹ In this study, we show that CD29 expression improves in vitro migration of CXCR3⁺ memory B cells across a brain microvascular endothelial layer toward CXCL10. This corresponds to the higher CXCL10 levels found in CSF of pwMS.²⁶ In addition, ITGB1 transcripts have been observed in CSF B cells of pwMS,⁴¹ supporting the hypothesis of CD29⁺CXCR3⁺ B-cell migration into the CNS. The heightened responsiveness of these memory B cells to T-cell signals likely promotes their subsequent engagement with local CD4⁺ T cells in so-called tertiary lymphoid structures.⁴² Our finding that CD29 and CXCR3 coexpression was restricted to ASCs in the MS CNS points to circulating CD29⁺CXCR3⁺ memory B cells as the precursors of CNS-residing ASCs. This corresponds to the improved ability of circulating CD29⁺CXCR3⁺ memory B cells to differentiate into ASCs in vitro. This may not only account for the CNS, as ITGB1 is also expressed with CXCR3 in murine B cells that become ASCs in the lung.⁴³ In our scRNA-seq analysis, transcription factor XBP1 (XBP1; downstream of BLIMP1) and IL-6 receptor (IL6R) were found as differentially expressed genes in CXCR3⁺ memory B cells, both of which have been associated with the later stages of ASC development.^44,45 This supports our observed increase in CXCR3 and CD29 in relation to CD138 expression on ASCs in postmortem CSF. CD29 also appears to be physically and functionally linked to CD38, another ASC maturation marker, in B-cell chronic lymphocytic leukemia cells.⁴⁶ Furthermore, CD29 is claimed to be involved in tissue retention,⁴⁷ potentially aiding in the residence of ASCs within survival niches in the chronically inflamed CNS.¹¹

Although we show that CD29 and CXCR3 coexpression defines a potential pathogenic B-cell population in MS, the question remains how these cells arise in pwMS. MS pathogenesis has been linked to both Epstein-Barr virus (EBV) infection and B cell–related genetic risk alleles.^{48-50, e1} EBV likely interacts with such genetic risk factors to modulate B-cell responses, promoting autoreactive B-cell phenotypes,^e2 such as CD29⁺CXCR3⁺ memory B cells.^e3,e4 It is of interest that one MS-related sequence variant has been detected directly in the ITGB1 gene.^e5 The ITGB1 gene even possesses a binding site for EBV proteins^e6 and is potentially upregulated by EBV infection.^e7 This genetic and environmental interplay could possibly explain the altered regulation of CD29 expression found on in vitro–induced ASCs in pwMS.

The specific targeting of CNS-homing memory B cells and CNS-resident ASCs in MS remains challenging, although currently used disease-modifying therapies, including natalizumab and cladribine, may interfere with their disease-related processes. In line with prior studies, while natalizumab treatment resulted in prevention of CNS-directed migration of memory B cells, specifically CXCR3⁺ memory B cells,⁸ cladribine treatment led to reduced memory B-cell frequencies in MS blood.^e8-e10 Building on these findings, we interrogated how CD29 expression is specifically affected by these MS therapies. Extending the observation of lower CD29 surface levels on total B cells in MS blood after natalizumab treatment,^e11 we found CD29 protein expression to be downregulated on the CXCR3⁺ memory population. Because steric hindrance between natalizumab and the CD29 antibody has been excluded,^e11 this reduction in CD29 was not caused by direct competition for the CD29 antibody–binding site. Instead, as all participants displayed active MS prior to the start of natalizumab treatment, the decrease in CD29 staining intensity likely reflects disrupted protein expression, which may be linked to the profound suppression of inflammatory MS activity as inflicted by 6-month treatment with natalizumab.^e12 In contrast to natalizumab, cladribine is uniquely capable of crossing the blood-brain barrier¹³ and thus could exhibit local effects on B cells and ASCs.¹⁵ However, deoxycytidine kinase (DCK), the enzyme responsible for phosphorylating and activating cladribine intracellularly, is significantly downregulated during ASC maturation,¹⁴ which may explain the mixed reports on oligoclonal band depletion in CSF of pwMS on cladribine treatment.^15,31 Cladribine has been previously shown to reduce CD29 expression on B cells, likely because of the decrease of memory B-cell frequencies, while leaving CD29 levels on T cells unaffected.^e13 We demonstrated that cladribine therapy leads to a reduction of CD29 protein expression specifically on repopulating CXCR3⁺ memory B cells in blood of pwMS that showed no early disease activity after treatment. This suggests that CD29 could be a valuable prognostic cellular marker and potential therapeutic target, which might also be relevant in other disease entities.^e14 Notably, we study disease activity during repopulation rather than breakthrough activity after 2 full treatment cycles with cladribine, as not all cladribine-treated patients completed 2 treatment cycles. Despite the need for independent validation studies, this first association can help in assigning peripheral pathogenic B-cell phenotypes to clinical MS features ultimately to assist in treatment strategies.

This work has some limitations. First of all, the sequencing depth of our transcriptomic analysis was limited. Nevertheless, we validated CXCR3 and CD29 coexpression on memory B cells at protein level. Second, we did not compare our findings in pwMS with individuals with other CNS inflammatory diseases, and therefore, cannot confirm whether the observed findings within the CNS are specific to MS. In addition, the small sample size of cladribine-treated patients limited our ability to draw conclusions, highlighting the need for validation in larger study groups. Furthermore, no samples were collected from cladribine-treated patients with MS longitudinally or prior to treatment, so we could not assess long-term effects and only make comparisons with a separate untreated MS cohort. Despite these limitations, our study provides a detailed phenotypical and functional characterization of CNS-residing ASCs and their precursors in MS, along with an accurate evaluation of the impact of commonly used MS disease-modifying therapies. Therefore, this work offers new insights into the B cell–related mechanisms involved in MS pathophysiology.

In conclusion, we identified CD29 as a characteristic and functional marker for CNS-homing CXCR3⁺ memory B cells in MS. CXCR3⁺ memory B cells likely exhibit an enhanced ability to upregulate CD29 after local interaction with T helper cells, enabling them to persist in the CNS as long-lived ASCs to chronically produce antibodies. The levels of CD29 and CXCR3 on memory B cells may be used as a peripheral indicator of this disease hallmark in the brain to more accurately predict the course and refine the treatment of MS.

Glossary

ASC

antibody-secreting cell

FACS

fluorescence-activated cell sorting

FCS

fetal calf serum

MS

multiple sclerosis

PBMC

peripheral blood mononuclear cell

PCA

principal component analysis

pwMS

people with MS

Author Contributions

K.L. Kuiper: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. L. Bogers: drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data. J. Rip: major role in the acquisition of data; analysis or interpretation of data. Y. van Hasselt: analysis or interpretation of data. M.-J. Melief: major role in the acquisition of data. A.F. Wierenga-Wolf: major role in the acquisition of data. R.A.M. Klein Kranenbarg: major role in the acquisition of data. E. Bindels: major role in the acquisition of data. H.J.G. van de Werken: analysis or interpretation of data. J. de Beukelaar: major role in the acquisition of data. I. Smets: major role in the acquisition of data. B. Wokke: major role in the acquisition of data. J. Smolders: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data. M.M. van Luijn: drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data.

Study Funding

The collaboration project is co-funded by PPP Allowance awarded by Health∼Holland, Top Sector Life Sciences & Health, to stimulate public-private partnerships (LSHM21066). In addition, this research is funded by Stichting MS Research (19–1057 MS; 20-490f MS), Nationaal MS Fonds (OZ2018-003, OZ2019-019, and OZ2021-09), Horizon Europe (BEHIND-MS, 101137235), and the healthcare business of Merck (10.13039/100009945). The healthcare business of Merck KGaA reviewed this manuscript for medical accuracy only before journal submission. The authors are fully responsible for the content of this manuscript, and the views and opinions described in the publication reflect solely those of the authors.

Disclosure

J. Smolders reports grants for scientific research from Biogen, Hansa Biopharma, Roche, and Siemens Healthineers and has received speaker and/or consultancy fees from Biogen, Novartis, Roche, Sanofi, and the healthcare business of Merck KGaA. M.M. van Luijn received research support from EMD Serono, Novartis, GSK, Idorsia Pharmaceutical Ltd, and the healthcare business of Merck KGaA. The remaining authors declare no conflicts of interest. Go to Neurology.org/NN for full disclosures.