Patients with relapsing-remitting multiple sclerosis (RRMS), diagnosed according to the 2017 revised McDonald criteria (12), who underwent AHSCT with a cyclophosphamide and anti-thymocyte globulin (ATG) conditioning regimen at Uppsala University Hospital between October 2011 and June 2022, were invited to participate. Blood samples were collected from 22 patients scheduled for AHSCT, at baseline (pre-AHSCT) (n=20) and at various follow-up intervals post-AHSCT. Summary demographics of study subjects ( Table 1 ), as well as detailed demographics, sampling time points and performed analysis are provided in Supplementary Table S1 .

Healthy controls (HC) and newly diagnosed MS patients (ND) served as comparison groups. Pseudonymized HC samples were obtained from blood donors matched by sex and age, whereas ND samples were derived from anonymized biobank material, thus precluding access to detailed clinical and demographic data. Consequently, demographic and clinical characteristics for ND patients are not reported in Table 1 .

Four patients showed potential evidence of disease activity post-AHSCT: three experienced clear clinical relapses, new T2 lesions, and gadolinium-enhancing lesions on MRI, indicating definite inflammatory disease activity. One additional patient had a single new small T2 lesion detected on MRI without associated clinical symptoms or gadolinium enhancement; thus, it is uncertain whether this represents true inflammatory disease activity or an incidental finding. This patient was included in descriptive analyses but is highlighted separately (with a distinct color in figures) to indicate this uncertainty. Due to the small number of patients with disease activity (three definitive, one uncertain), formal statistical comparisons between patients with and without disease activity post-AHSCT were not performed.

Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll-Paque PLUS density gradient centrifugation (Cytiva). Isolated cells were cryopreserved in fetal calf serum supplemented with 10% dimethyl sulfoxide (DMSO) and stored at -170°C.

Blood samples were drawn at the clinic and T cell count were performed as part of the clinical immune reconstitution follow up. For all these timepoints were not phenotype analysis performed. For detailed information about patients and time points for these samples see Supplementary Table S1 . A volume of 50µl blood, acquired in EDTA tubes, were pipetted in to Trucount™ tubes (BD Bioscience) containing a known number of fluorescent beads and stained with antibodies for CD3-FITC, CD45-Per Cp5.5, CD4-PeCy7 and CD8-APCCy7 incubated in the dark for 15 minutes at room temperature (RT). The erythrocytes were lysed using BD FACS™ lysing solution, diluted 1:10, and the sample was incubated for another 15 min in the dark at RT.

Cells were analyzed by flow cytometry on a DxFLEX instrument (Beckman Coulter), CD45⁺ leukocytes were gated on CD3⁺CD4⁺ and CD3⁺CD8⁺ and frequencies of these cells were established. CD4 and CD8 T cells were enumerated in relation to the fixed number of fluorescent beads in each sample.

Cryopreserved PBMCs were analyzed using mass cytometry at the CryoSciLifeLab Cellular Immunomonitoring Facility in Stockholm, Sweden. Analysis included B/Myeloid, T-cell and intracellular panels ( Supplementary Figures 1 , S1.1, S1.2 ). Thawed cells were incubated in Benzonase-supplemented media and allowed to recover for 2 hours at 37°C and 5% CO₂ before staining.

Upon analysis, frozen PBMCs were thawed at 37°C and washed twice in pre-warmed (37°C) RPMI-1640 medium supplemented with 10% heat-inactivated fetal calf serum, 10 mM HEPES, and 2 mM L-glutamine (all from Gibco). Cell viability and counts were assessed, and 1 × 10⁶ cells per sample were resuspended in calcium- and magnesium-free phosphate-buffered saline (PBS) containing 5% normal mouse serum to block nonspecific binding.

Mass cytometry data were normalized and de-barcoded. Files were analyzed using FlowJo (versions 10.6.2 and 10.8.1) for gating and phenotypic assessments. Briefly, cells of all CyTOF FCS files from panel 1 and panel 2 were subjected to initial clean up gating in FlowJo ( Supplementary Figures 1 ; S1.3 ) and FCS files containing CD45⁺ leukocytes together with relevant channels were exported and imported into new FlowJo worksheets. Equal numbers of cells per subset (e.g., 40,000 CD45⁺ cells, 20,000 T cells) were randomly selected (down sampled) and files concatenated for downstream analysis. The resulting mass cytometry data were processed in FlowJo using high-dimensional reduction algorithms (UMAP or t-SNE), followed by cell subset clustering with the graph-based clustering algorithm PhenoGraph clustering tool. The phenotype of each identified cell cluster was further characterized using Cluster Explorer, which enabled the visualization and annotation of clusters based on marker expression profiles. The Myeloid cell population was defined as CD3^-CD19^-CD20^-HLA-DR⁺ and NK cells CD3^-CD19^-CD20^-HLA-DR^neg/low. For gaiting strategy of NK and Myeloid cells and marker inclusions for Myeloid cell subsets see Supplementary Figures 1 ; S1.4, S1.5 . Expression patterns of markers used in each immune cell subset analysis are outlined in Supplementary Figures 1 ; S1.6-S1.8 .

Flow cytometric data was performed in FlowJo (versions 10.6.2 and 10.8.1). Initial gating strategy for phenotype analysis of live PBMCs (panel 1-4) ( Supplementary Figures 1 ; S1.9 ) and for fixated permeabilized PBMCs analyzing nuclear expressed antigens ( Supplementary Figures 1 ; SS1.10 ). Gating strategies for each flow cytometric panel (panel 1-5) ( Supplementary Figures 1 ; S1.11-S1.15 ). Definitions of CD4 T helper subsets are presented in Supplementary Figures 1 , S1.16 .

Data are presented as medians with minimum and maximum values. The Wilcoxon matched-pairs signed-rank test was used for paired baseline and post-AHSCT comparisons. Differences between unpaired groups were assessed using the Mann-Whitney U test. For multiple group comparisons, Kruskal-Wallis tests followed by Dunn’s post hoc tests were applied. Statistical significance was defined as p < 0.05. Analyses were performed using GraphPad Prism (version 10).

The study was approved by the Regional Ethical Review Board in Uppsala (Dnr 2010/450/1 and 2012/080/1). The study was performed in concordance with the Declaration of Helsinki (1964), and all patients provided written informed consent.

Whole blood collected at baseline, 6 months, 1 year, 2 years as well as a few later timepoints were analyzed for absolute counts of CD4 and CD8 T cells by flow cytometry. CD4 T cells demonstrated significantly reduced absolute counts at 6 months, 1 year, and 2 years post-AHSCT compared to baseline, with gradual recovery observed over time. However, for most patients CD4 T cell counts and frequencies remained below baseline throughout the follow-up period ( Figures 1A, B ). In contrast, CD8 T cell counts remained within normal ranges, with no significant changes ( Figure 1D ).

The CD4/CD8 ratio was inverted in all patients at 6 months post AHSCT but showed partial normalization in six patients by two years ( Figures 1G, H ). Patients with EIDA had all regained normal CD4 counts and CD4/CD8 ratios during the observation period. However, this finding was not exclusive for relapsed patients.

Cryopreserved PBMC samples from 12 of the 22 MS patients treated with AHSCT, collected at baseline and two years post-treatment, were selected for immunophenotypic analysis using CyTOF panel 1. Samples from 10 newly diagnosed, untreated MS patients (ND) and 8 healthy controls (HC) were included as reference groups. Two antibody panels were utilized: panel 1 (26 markers), designed to capture an overall leukocyte phenotype with a focus on B cells and myeloid cells ( Supplementary Figures 1 , S1.1 ), and panel 2 (24 markers), tailored for an in-depth analysis of T cell phenotypes ( Supplementary Figures 1 ; S1.2 ).

Mass cytometry data were processed in FlowJo, where high-dimensional reduction algorithms (UMAP or t-SNE) were applied to visualize cellular distributions. Cell subsets were then identified using the PhenoGraph clustering algorithm. The phenotype of each identified cluster was further characterized using Cluster Explorer, which facilitated the visualization and annotation of clusters based on marker expression profiles.

Analysis of CyTOF data from the B-cell/myeloid panel (CyTOF, panel 1) revealed an increase in B-cell frequencies and a concurrent decrease in myeloid cell proportions two years post-AHSCT. In contrast, T- and NK-cell proportions remained stable ( Figures 2A–D , Supplementary Figures 2 , S2.1 ). The ND group exhibited significantly higher T-cell frequencies and lower myeloid cell proportions compared to HC ( Figures 2A, D ).

A more detailed assessment of HLA-DR⁺ myeloid subsets (CD3^-CD20^-CD19^-) showed a reduction in classical monocytes (CD14⁺CD16^-) (p<0.001) and an increase in non-classical monocytes (CD14^-CD16⁺) (p<0.01) following AHSCT (CyTOF, panel 1). Intermediate monocytes (CD14^intCD16^low) remained unchanged ( Figures 3A–C , Supplementary Figures 2 ; S2.2A-C ). Furthermore, the relative frequencies of plasmacytoid dendritic cells (pDCs) and conventional dendritic cells (DCs) were elevated two years post-AHSCT (p<0.01 and p<0.001 respectively) ( Figures 3A–C , Supplementary Figures 2 ; S2.2D, E ). Intriguingly, the composition of the myeloid compartment in post-AHSCT patients resembled the ND group with lower frequencies of classical monocytes and a rise in pDCs and DCs ( Supplementary Figures 2 , S2.2A, D, E ). Two of the three patients with EIDA (one experiencing a clinical relapse and another with a new T2 lesion) exhibited the lowest classical monocyte frequencies and the highest pDC frequencies among transplanted patients ( Supplementary Figures 2 ; S2.2A, D ).

Given the effectiveness of B-cell-depleting therapies in MS treatment, we sought to analyze the B-cell population in greater detail. Manually gated B cells (CD19⁺ CD20^+/- CD3^-CD14^-) from the B-cell/myeloid CyTOF panel 1 were subjected to dimensional reduction analysis using t-SNE, followed by cell clustering with the PhenoGraph algorithm and phenotypic characterization of clusters using Cluster Explorer. Two years post-AHSCT, the B-cell compartment exhibited a reduction in memory B cells and plasma cells, accompanied by a shift toward naïve B cells. In contrast, the proportions of transitional B cells (both TrB1 and TrB2) remained stable, except for one patient with prior rituximab therapy, who displayed elevated TrB2 levels at baseline. Interestingly, patients with EIDA had higher baseline levels of non-switched IgD⁺IgM⁺ memory B cells, suggesting a potential predictive biomarker ( Figures 4A–D , Supplementary Figures 2 ; S2.3A–Q ).

Given the well-established role of T cells in the pathogenesis of multiple sclerosis (MS), we sought to characterize the phenotype of CD4⁺ and CD8⁺ T-cell populations in greater detail. CyTOF data from the T-cell panel (CyTOF, panel 2) were analyzed using a sequential workflow: first, CD4 and CD8 T cells were manually gated, followed by dimensionality reduction using t-SNE, cell subset clustering with PhenoGraph, and phenotypic characterization of clusters using Cluster Explorer. Naïve T cells were defined as CD45RA⁺CD28⁺CD27⁺, memory T cells (primarily central memory with some effector memory) as CD45RA^-CD28⁺CD27⁺, effector memory T cells as CD45RA^-CD28⁺CD27^-, and T_EMRA cells as CD45RA⁺⁺CD28^-CD27^-. Due to poor data quality, two CyTOF runs were excluded from the analysis, affecting samples from five AHSCT patients (pre- and post-treatment), five ND samples, and three HC.

Two years post-AHSCT, naïve CD4 T cells (cluster a, b. d, e), phenotypical separated by HLA-DR, CD38 and CD5 expression, exhibited a modest decline in all patients except for one patient who experienced clinical relapse during the observation period ( Figures 5A–C , Supplementary Figures 2 ; S2.4A, B, D, E ). In contrast, two clusters of naïve CD4 T cells significantly expanded: one enriched for cells expressing CD194/CCR4 (cluster c) and another composed of naïve-like CD4 T cells lacking CD127 (cluster f) (p<0.05 and p<0.01 respectively) ( Figures 5A–C , Supplementary Figures 2 ; S2.4C, F ).

Within the memory CD4 T-cell compartment, a shift from central memory (T_cm) to effector memory (T_em) phenotypes was observed two years post-AHSCT. At this timepoint, the frequency of T_cm CD4 T cells lacking additional defining markers (cluster g) and those expressing CD161, associated with IL-17 production, (cluster h) were consistently reduced across all patients (p<0.01) ( Figures 5D, E ). In contrast, memory CD4 T cells expressing CD194/CCR4 with or without CD161, as well as memory cells with an activated phenotype, remained at baseline levels (clusters i-l) ( Figures 5A–E , Supplementary Figures 2 ; S2.4I–L ). Two distinct clusters of T_em lacking CD27 were identified, indicative of a more antigen-experienced phenotype (cluster m and n). ( Figures 5F, G ). These clusters differed in their expression of CD194/CCR4 and the immune checkpoint receptor PD-1. The relative proportions of these subsets varied more between patients at two years post-AHSCT compared to baseline. Notably, two of the three patients with EIDA exhibited the highest frequencies of CD194/CCR4⁺ T_em (cluster m) but among the lowest frequencies of PD-1⁺ T_em (cluster n) ( Figures 5F, G ). In contrast, all but one patient in remission showed an increased proportion of PD-1⁺ T_em (cluster n) ( Figure 5G ). Additionally, the proportion of effector/T_EMRA CD4 T cells expressing high levels of CD57 (cluster o) was significantly increased two years post-AHSCT, though some inter-patient variability was observed at this timepoint ( Figures 5A–C , Supplementary Figures 2 ; S2.4O ). Regulatory T cells (T_regs; cluster p), characterized by central memory-like properties and high expression of CD39, CD194, and ICOS, maintained frequencies similar to baseline ( Figures 5A–C , Supplementary Figures 2 ; S2.4P ).

CD8 T cells exhibited greater phenotypic diversity than CD4 T cells. Two years post-AHSCT, naïve CD8⁺ T-cell subsets (clusters a and b), (CyTOF panel 2) distinguished by CD31 expression, showed a slight decline in most patients, except for one individual who had a clinical relapse ( Supplementary 2 , Supplementary Figures 2 ; S2.5A ). In contrast, significant reductions were observed in CD38^- naïve CD8 T cells, (cluster c, p<0.05), HLA-DR⁺, (cluster d, p<0.01) and a small population of proliferating naïve-like cells lacking CD127 (cluster f, p<0.05), with this trend consistent across all patients. ( Supplementary 2 , Supplementary Figures 2 ; S2.5C, D, F )

Meanwhile, memory CD8 T cells shifted toward a more antigen-experienced effector memory/exhausted phenotype, characterized by increased frequencies of both CD57⁺ and CD57^- T_EMRA subsets (p<0.05 and p<0.01 respectively). Notably, patients in remission displayed higher T_EMRA cell frequencies compared to those with EIDA ( Figures 6A–C , Supplementary Figures 2 ; S2.5G–P ).

Mucosal-associated invariant T (MAIT) cells, identified by CD161 and CCR5 expression, declined consistently across all patients, whereas TcRγδ T cells (clusters s-u) remained at baseline frequencies ( Figures 6A–C , Supplementary Figures 2 ; S2.5Q–U ).

Building on the CyTOF data, we sought to further characterize circulating T-cell phenotypes before and after AHSCT in a larger group of AHSCT-treated patients, incorporating both earlier and later time points for reference. To this end, we designed three traditional 7-color flow cytometry panels to assess T-cell maturation, functionality (Th-phenotype), and homing properties.

These panels included markers for CD62L, CCR7 (lymph node homing) and CD45RO, absent in the mass cytometric panel, as well as CXCR3, CCR5, CCR6, and CXCR6 (tissue homing and Th-phenotype). For definition of CD4 Th-phenotype see Supplementary Figures 1 ; S1.16 . Additionally, one panel specifically analyzed PD-1 expression in relation to CXCR3 (Th1-associated) and CCR6 (Th17-associated) subsets.

In the CyTOF analysis, we observed a reduction in MAIT cells two years post-AHSCT (identified using CD161 and CCR5 as lineage markers). To further characterize these populations, we employed a seven-color flow cytometry panel incorporating the invariant T-cell receptor TcR Vα7.2, along with CD161, IL-18R, and CXCR6.

Two years post-AHSCT, the overall proportion of TcR Vα7.2⁺ CD8 T cells exhibited a modest but significant decline (p<0.05) ( Figure 10A ). Further stratification into classical MAIT cells (TcR Vα7.2⁺CD161⁺IL-18R⁺) and immature/non-MAIT cells (TcR Vα7.2⁺CD161^-/⁺IL-18R^-/⁺) revealed a pronounced reduction in classical MAIT cells (p<0.0001), whereas the frequencies of immature/non-MAIT cells remained stable across all time points ( Figures 10B, D , Supplementary Figures 2 ; S2.12 ).

Classical MAIT cells exhibited high CXCR6 expression, while immature/non-MAIT cells displayed minimal CXCR6 expression ( Figures 10C, E , Supplementary Figures 2 ; S2.12E ). Within the double-negative (DN) T-cell subset, the proportions of classical and immature MAIT cells closely mirrored those observed in the CD8 T-cell compartment. In contrast, these subsets remained consistently low and unchanged in the CD4 T-cell population at all time points ( Supplementary Figures 2 ; S2.11, S2.12 ).

The CyTOF analysis of CD4 T cells revealed stable frequencies of T_regs at baseline and two years post-AHSCT. To further characterize their phenotype, we performed flow cytometric analyses of FoxP3, Helios, CD62L, CCR5, and HLA-DR.

Consistent with CyTOF data, the overall proportion of FoxP3⁺ CD4 T cells remained unchanged two years post-AHSCT ( Supplementary Figures 2 ; S2.13A–C ). However, within the FoxP3⁺Helios⁺ subset, commonly referred to as natural Tregs (nTregs) or thymic-derived Tregs (tTregs), a slight but significant reduction in relative frequency was observed (p<0.05). In contrast, FoxP3⁺Helios^- Tregs (inducible Tregs, iTregs) maintained frequencies comparable to baseline ( Supplementary Figures 2 ; S2.13D–G ).

Both Treg subsets exhibited an enhanced activation phenotype, as indicated by reduced CD62L expression (p<0.01) and an increase in CCR5⁺ and HLA-DR⁺ cells, particularly within the FoxP3⁺Helios⁺ subset (p<0.01 and p<0.001, respectively) ( Supplementary Figures 2 ; S2.14A–H ).

While AHSCT is known to reset the immune system and reduce disease activity, most previous studies have focused on T-cell reconstitution, leaving the B-cell compartment relatively underexplored. Our findings align with prior studies demonstrating that B cells are efficiently depleted following AHSCT, yet recover rapidly and exceed normal levels within one year post-transplantation (9, 13). The predominance of naïve B cells post-AHSCT, which remains stable for at least two years, mirrors the natural development of memory B cells in childhood, where memory B-cell frequencies gradually stabilize by the age of five (14).

The profound depletion of memory B cells and plasmablasts, coupled with the expansion of naïve B cells, supports the notion that AHSCT effectively resets the B-cell compartment, reducing their capacity for antigen presentation and autoantibody production. Notably, patients with EIDA displayed a higher proportion of non-switched IgD⁺IgM⁺ memory B cells pre-AHSCT, suggesting a role for this subset in MS pathophysiology and highlighting its potential as a predictive biomarker for treatment response, though this needs confirmation in larger studies.

Unswitched IgD⁺IgM⁺ memory B cells, characterized by the expression of CD27 along with IgM and IgD, represent a key subset of the memory B-cell lineage. These cells play a fundamental role in immune surveillance and response, particularly by serving as a first line of defense against pathogens (15). Strong evidence supports the existence of two distinct subsets of IgD⁺IgM⁺CD27⁺ B cells: one generated independently of germinal centers and predominating in early life, and another with key post-germinal center memory B-cell characteristics that dominate in adulthood. These IgD⁺IgM⁺CD27⁺ B cells can differentiate into IgM-secreting plasma cells, providing highly efficient complement fixation, but they also have the unique ability to re-enter germinal center reactions upon re-exposure to the same or a related antigen. This allows for adaptation of their B-cell receptor specificity to altered antigens through new rounds of somatic hypermutation and selection (15).

In general, increased frequencies of memory B cells have been associated with an MRI phenotype with high neurodegeneration, defined by increased numbers of contrast-enhancing lesions and non-enhancing black holes on T1-weighted images, and reduced brain parenchymal fraction (16). Anti-CD20 monoclonal antibodies are very potent B-cell depleting agents and effectively deplete memory B cells (17) and have been shown to prevent new T2 lesions and clinical relapses (18, 19). In contrast, tabalumab, an anti-BAFF monoclonal targeting BAFF, mainly affecting transitional and mature B cells while sparing memory B cells, had no clinical effect in RRMS patients (20) Instead, treatment with tabalumab led to an increase in circulating memory B cells, further underscoring the importance of memory B-cell depletion for therapeutic efficacy (21).

The role of IgD⁺IgM⁺CD27⁺ B cells in the pathophysiology of MS is less clear, but they have increasingly been implicated in MS pathogenesis. Studies have demonstrated that these cells are capable of supporting ectopic lymphoid structures within the CNS. These structures serve as niches for ongoing inflammation and autoantibody production, contributing to chronic disease activity (22, 23). Moreover, unswitched memory B cells play a role in antigen presentation, engaging with autoreactive T cells and amplifying the autoimmune response (23). Their ability to produce pro-inflammatory cytokines, such as IL-6 and GM-CSF, further underscores their contribution to pathogenic inflammation in MS. Conversely, their production of the regulatory cytokine IL-10 appears limited in disease contexts, reflecting a dysregulation of their normal immune-modulating functions (22, 23). Such observations suggest that unswitched memory B cells may preserve autoreactive T cell clones that are resistant to AHSCT. This hypothesis is further supported by the finding that rituximab administration in close proximity to AHSCT is highly protective for relapses post-AHSCT (24).

Myeloid cells, particularly monocytes and macrophages, are known to infiltrate the CNS in MS and contribute to neuroinflammation, demyelination, and lesion formation (25, 26). In our study we observed a significant reduction in classical monocytes post-transplant, while non-classical monocytes increased. This shift could be important, as classical monocytes are major producers of pro-inflammatory cytokines such as TNF-α, IL-6, and IL-1β, which drive MS progression (27). Conversely, non-classical monocytes exhibit a patrolling function and possess regulatory properties, limiting excessive inflammation. This may represent a shift toward immune regulation rather than inflammation, potentially contributing to long-term remission in MS.

We also observed an increase in pDC post-AHSCT. Classically, these have been associated with antiviral responses and type I interferon production (28). This may be related to the shift from a Th17 to a Th1 type response, but could also reflect a response to viral replication post-AHSCT in the immunocompromised host. Future research focusing on functional assessments of these altered myeloid populations will be critical to understanding their roles in maintaining immune tolerance and preventing MS relapse.

T cells play a central role in multiple sclerosis (MS) pathogenesis, with both CD4⁺ and CD8⁺ subsets contributing to neuroinflammation and disease progression. Historically, MS has been considered a predominantly T-cell-driven autoimmune disorder, supported by findings of T-cell infiltration into CNS lesions and the strong genetic association with HLA-DRB1*15:01 (29). While early studies emphasized a Th1-driven immune pathology, more recent evidence suggests that disease progression is also influenced by Th1/Th17-mediated responses (30, 31). Given the critical role of T cells in MS, understanding their reconstitution post-AHSCT is essential for elucidating mechanisms of immune reset and sustained remission.

Our results reaffirm the critical role of immune ablation and subsequent reconstitution in achieving durable remission. The observed reduction in CD4 T-cell counts at six months, one year, and two years post-AHSCT aligns with previous findings (5), underscoring the transient depletion of lymphocyte subsets as an expected outcome of high-dose immunosuppressive conditioning. In contrast, CD8 T-cell counts remained relatively stable. The shifts in the CD4/CD8 ratio, persisting in many patients at least for two years, mirror the long-lasting effect of AHSCT. Importantly, patients exhibiting EIDA post-AHSCT retained normal CD4 counts and ratios, suggesting that the CD4/CD8 ratio is a potential biomarker for incomplete immune reset.

Recovery of naïve CD4 T cells following AHSCT is known to depend on thymic function, with reconstitution typically beginning around one year post-transplant. At the two-year mark, we observed inter-patient variability in naïve CD4 T-cell frequencies, however those with EIDA were among the patients with the highest proportions, suggesting that this is associated with enhanced thymic output. Further analysis of the CD45RA⁺CD45RO^- subset revealed an increase in naïve CD4⁺ T cells lacking CD62L post AHSCT. Although most patients exhibited a slight increase in these atypical naïve cells, the highest frequencies were observed in those who relapsed. Interestingly, newly diagnosed MS patients also displayed a significant elevation of naïve CD4⁺CD62L^- T cells compared to healthy controls, suggesting that this subset may be relevant to early disease processes.

Naïve and T_cm CD4 T cells typically express CD62L and CCR7, which facilitate their recirculation through lymphoid tissues. Upon antigen engagement, CD62L downregulation enables effector differentiation and tissue migration. However, CD62L downregulation can also occur independently of TCR signaling, driven by cytokines, ATP, glucocorticoids, and metalloprotease activity (32). Atypical CD62L^- naïve CD4 T cells have been described in rheumatoid arthritis (RA), where they correlate with disease flares and exhibit IL-1/IL-6/TNF-driven inflammatory signatures (33). A similar subset has been linked to relapses in a rat model of MS (34), suggesting a role in autoimmune pathology. These cells may serve as a pool of pre-activated, bystander-responsive CD4 T cells, capable of amplifying CNS inflammation when recruited to affected tissues.

The increased proportions of CD62L^- naïve CD4 T cells post-AHSCT may reflect an inflammatory milieu favoring cytokine-driven activation. Alternatively, their presence in both relapsed and remission patients suggest a regulatory function in preventing excessive T-cell activation. Future studies should examine their TCR repertoire to determine whether they contribute to disease relapse or immune regulation post-AHSCT.

Our flow cytometric analysis of CD4 T cells post-AHSCT revealed considerable inter-patient variability in immune reconstitution. However, certain shared features emerged two years post-transplant, including a consistent reduction in T_cm CD4 T cells, polyfunctional Th1/Th17 (CCR6⁺CXCR3⁺) cells, and Th17 (CCR6⁺CXCR3^-) cells across all transplanted patients. These changes were accompanied by a variable increase in T_em CD4 T cells and a Th1-skewed phenotype (CXCR3⁺CCR5⁺CCR6^-), findings that align with previous studies (5).

Using CyTOF analysis, where CD161 and CCR4 served as Th17 markers, we could further delineate these shifts. The observed reduction in T_cm CD4 T cells was predominantly restricted to two subsets: one lacking additional defining markers included in the CyTOF panel and another enriched for CD161⁺CCR4^- Th17 cells. In contrast, the frequencies of classical Th17 T_cm CD4 T cells co-expressing CD161 and CCR4 (CD194) or CCR4 alone (Th0/Th2) remained stable. Notably, the Th1/Th17 and Th17 subsets that decreased were predominantly PD-1^-, whereas their PD-1⁺ counterparts were preserved. These findings suggest that the reduction in Th17 cells post-AHSCT is limited to a specific subset of Th17 T_cm cells. However, given that this decline occurred in all patients, regardless of clinical outcome, additional immunological mechanisms likely contribute to long-term disease remission.

In contrast to the more uniform contraction of T_cm subsets, the T_em CD4 T-cell compartment exhibited greater inter-patient variability, particularly in the Th1 subtype and regarding PD-1 expression. Notably, clusters enriched for PD-1⁺ T_em CD4 T cells differed between patients with and without clinical relapse. All but one patient in remission exhibited elevated frequencies of PD-1⁺ T_em, whereas those who relapsed showed frequencies comparable to baseline. A similar, albeit less pronounced, trend was observed in CD4 T-cell clusters containing PD-1⁺CD57⁺ T_EMRA cells. Conversely, two patients with EIDA, including one who experienced a clinical relapse at the two-year time point, exhibited the highest frequencies of Th17-skewed T_em cells (CCR4⁺CD161⁺).

The role of T_regs in the pathology of MS is not completely understood and the interplay between immune regulatory and immune enhancing elements are complex. Genetic variants in CTLA-4 and CD25, along with altered functions and levels of T_regs in the circulation of MS patients have been reported (35, 36). Emerging evidence underscores the importance of brain-resident T_regs and their functional interactions with pathogenic effector T cells and other immune cells, suggesting a potential role in modulating neuroinflammation (37). Previous studies of AHSCT for MS have described an early transient rise in T_reg frequencies followed by return to baseline levels within 1 year post-AHSCT. The origin of these T_regs is presumed to be cells that escaped ablation or were re-infused with the graft (36, 38–40). At later time points, once thymic output is restored, the T_reg compartment would also be expected to reflect ongoing immune renewal (41).

In the present study, the overall proportions of CD4 with a T_reg associated phenotype were similar to baseline at two years post-AHSCT, but a slight decrease in the frequencies of a subset of natural T_regs (FoxP3⁺Helios⁺) was observed two years post-AHSCT. However, the phenotype of these T_regs cells showed an altered phenotype with decreased expression of CD62L, increased CCR5 and HLA-DR which resembles an activated effector memory type with possible CNS-migration properties.

Although much of MS research has focused on CD4 T cells, emerging evidence suggests that CD8 T cells play a more significant role than previously appreciated. Histopathological analyses of post-mortem MS brain tissue have revealed a higher abundance of CD8 T cells compared to CD4 T cells. Additionally, studies in animal models have identified both myelin-reactive CD8 T cells and a potential regulatory function for this subset (42, 43). Given the distinct biological roles of CD4 and CD8 T cells, one plausible interpretation is that CD4 T cells initiate and drive the autoimmune process, whereas cytotoxic CD8 T cells act as effectors, executing tissue damage by targeting glial cells.

In the present study, we confirmed previous findings of a relatively rapid reconstitution of CD8 T cells in circulation post-AHSCT (5). The phenotypic profile of the CD8 compartment largely mirrored that of CD4 T cells, with a predominance of memory T cells at early time points post-transplant and a gradual recovery of naïve CD8 T cells over time.

At two years post-AHSCT, a shift from a naïve/T_cm-dominant to a T_em-skewed CD8 T-cell population remained evident, with significantly reduced proportions of naïve and T_cm CD8 T cells. However, substantial inter-patient variability was observed in chemokine receptor expression, PD-1 receptor expression, and distinct T_em/T_EMRA CD8 T-cell CyTOF clusters.

Previous studies have reported an expansion of atypical terminally differentiated CD28^- CD57⁺ CD8 T cells post-AHSCT, which has been linked to immunosuppression (8, 44). Interestingly, in our cohort, patients with EIDA exhibited high proportions of both CD57⁺ and CD57^- CD8 cells lacking CD28 but co-expressing CD27. In contrast patients in remission showed increased proportions of highly differentiated CD57⁺ CD28^- CD27^- T_EMRA-like cells, mirroring the corresponding antigen-experienced phenotype among the CD4 T-cells. This suggests a possible role for immune senescence in AHSCT-induced long-term tolerance.

MAIT cells represent a unique subset of T cells characterized by the expression of the semi-invariant TCR TRAV1-2 (Vα7.2) paired with a limited set of TCRβ chains, typically TRBV6 or TRBV20. MAIT cells are primarily found in mucosal tissues, liver, and blood, where they respond to microbial vitamin B metabolites presented by MR1, triggering cytokine release and cytotoxic activity (45). Classical MAIT cells co-express CD161 and IL-18R, but a subset of Vα7.2⁺ T cells lack these markers, raising questions about whether these cells represent precursors to MAIT cells, a resting/naïve MAIT state, or an entirely distinct cell type. In MS, MAIT cells exhibit altered functionality and frequency, with their presence in MS lesions suggesting a potential role in MS pathology (46, 47).

Several studies have reported a persistent decline in MAIT cells following AHSCT in MS (38, 48, 49). In the present study, we observed a significant reduction in CD8 and double-negative (DN) MAIT cells at all time points up to two years post-transplant, with a slight recovery at later time points. In contrast, TcR Vα7.2⁺ T cells lacking classical MAIT markers (CD161^-/IL-18R^-) exhibited little variation from baseline. Whether these cells represent precursors to classical MAIT cells remains debated, and a more precise approach to defining their phenotype would involve the use of an MR1 tetramer loaded with an appropriate antigen. Additionally, given the frequent administration of antibiotics post-AHSCT, which alters the gut microbiota, it is possible that the reduction in classical MAIT cells reflects a decreased availability of bacterial-derived antigens necessary for their maintenance.

While our study provides valuable insights into immune reconstitution following AHSCT, several limitations should be acknowledged. The relatively small cohort size and observational design limit the generalizability of our findings, particularly regarding the subgroup experiencing post-AHSCT relapses. Consequently, these results should be considered exploratory and validated in larger cohorts. Moreover, our analyses were limited to circulating immune cell phenotypes, potentially influenced by external pathogens and immune events unrelated to MS.

Due to resource constraints and methodological scope, we did not include functional assays or B-cell receptor sequencing, which could provide deeper mechanistic insights. Specifically, detailed phenotypic and functional characterization of the unswitched IgD⁺IgM⁺ memory B-cell subset, including activation or exhaustion markers, somatic hypermutation status, and antibody secretion, represents an important direction for future research.

Future studies could also leverage single-cell sequencing and proteomics analyses of blood and cerebrospinal fluid (CSF) to further dissect molecular mechanisms driving immune reconstitution and relapse susceptibility.