Specific myeloid signatures in peripheral blood differentiate active and rare clinical phenotypes of multiple sclerosis

Abstract

Current understanding of Multiple Sclerosis (MS) pathophysiology implicates perturbations in adaptive cellular immune responses, predominantly T cells, in Relapsing-Remitting forms (RRMS). Nevertheless, from a clinical perspective MS is a heterogeneous disease reflecting the heterogeneity of involved biological systems. This complexity requires advanced analysis tools at the single-cell level to discover biomarkers for better patient-group stratification. We designed a novel 44-parameter mass cytometry panel to interrogate predominantly the role of effector and regulatory subpopulations of peripheral blood myeloid subsets along with B and T-cells (excluding granulocytes) in MS, assessing three different patient cohorts: RRMS, PPMS (Primary Progressive) and Tumefactive MS patients (TMS) (n=10, 8, 14 respectively). We further subgrouped our cohort into inactive or active disease stages to capture the early underlying events in disease pathophysiology. Peripheral blood analysis showed that TMS cases belonged to the spectrum of RRMS, whereas PPMS cases displayed different features. In particular, TMS patients during a relapse stage were characterized by a specific subset of CD11c+CD14+ CD33+, CD192+, CD172+-myeloid cells with an alternative phenotype of monocyte-derived macrophages (high arginase-1, CD38, HLA-DR-low and endogenous TNF-a production). Moreover, TMS patients in relapse displayed a selective CD4 T-cell lymphopenia of cells with a Th2-like polarised phenotype. PPMS patients did not display substantial differences from healthy controls, apart from a trend toward higher expansion of NK cell subsets. Importantly, we found that myeloid cell populations are reshaped under effective disease-modifying therapy predominantly with glatiramer acetate and to a lesser extent with anti-CD20, suggesting that the identified cell signature represents a specific therapeutic target in TMS. The expanded myeloid signature in TMS patients was also confirmed by flow cytometry. Serum neurofilament light-chain levels confirmed the correlation of this myeloid cell signature with indices of axonal injury. More in-depth analysis of myeloid subsets revealed an increase of a subset of highly cytolytic and terminally differentiated NK cells in PPMS patients with leptomeningeal enhancement (active-PPMS), compared to those without (inactive-PPMS). We have identified previously uncharacterized subsets of circulating myeloid cells and shown them to correlate with distinct disease forms of MS as well as with specific disease states (relapse/remission).

Keywords: B cells; NK cells; T cells; macrophages; mass cytometry; multiple sclerosis; myeloid-signature; tumefactive multiple sclerosis.

Figure 1

Experimental setup and workflow for mass cytometry analysis. (I) Peripheral blood mononuclear cells (PBMCs) from patients with Multiple Sclerosis (n=32; PPMS=8, RRMS=10, TMS=14) and healthy donors (n=10) were isolated and further stratified in more subgroups based on clinical and radiological criteria. RRMS and TMS were subgrouped based on disease activity in those during a clinical relapse (RRMS; n=5, TMS; n=5) and those under remission (RRMS; n=4, TMS; n=5). PPMS were subgrouped based on the presence of leptomeningeal enhancement on MRI (active PPMS; at least one foci of leptomeningeal inflammation, inactive; without evidence of leptomeningeal inflammation based on MRI criteria). (II) PBMCs were labeled with metal-tagged antibodies against the markers shown in boxes and acquired on a Helios mass cytometer. This multiplex analysis allows broad immunophenotyping as well as deeper analysis in B cell and myeloid subpopulations. (III) Acquired data were analyzed using established analysis pipelines for dimensionality reduction and exploratory data analysis, clustering, and differential analysis. PPMS, Primary Progressive Multiple Sclerosis; RRMS, Relapsing Remitting Multiple Sclerosis; HD, Healthy donors; TMS, Tumefactive Multiple Sclerosis; MRI, magnetic resonance imaging; t-SNE, t-distributed stochastic neighbor embedding; FlowSOM, Flow Self-Organizing Map.

Figure 2

Major cell lineage subsets in peripheral blood of patients with MS during different disease stages and healthy donors. (A) Dimensionality reduction with t-SNE CUDA (t Distributed Stochastic Neighborhood Embedding) on all patients and healthy donors tSNE map (left) shows major immune cell subpopulations in the CD45+ compartment (named T4; CD4+ T cells, T8; CD8+ T cells, TCRgd T cells, NK cells, B cells, Myeloid and Dendritic cells). Major PBMC populations were annotated based on the expression of key lineage markers (CD3, CD8, CD19, CD56, CD11c, TCRgd, CD123). (B) t-SNE plots from one representative sample from each subgroup (PPMS, RRMS, TMS and HD). RRMS and TMS were further subdivided in those in remission (RRMS; n=4, TMS; n=5) and relapse (RRMS; n=5, TMS; n=5). (C) Box plots showing frequency (expressed as % of live singlet CD45+ cells) of each major immune cell subset in peripheral blood mononuclear cells of patient subgroup and healthy donors (n=10). Statistically significant changes are shown in the image (p < 0.05 was considered significant, non-parametric Kruskal Wallis test with correction for multiple comparisons (FDR) was applied). PPMS, Primary Progressive Multiple Sclerosis; RRMS, Relapsing Remitting Multiple Sclerosis; Rem, Remission; Rel, Relapse, HD, Healthy donors; TMS, Tumefactive Multiple Sclerosis; T4; CD4+ T cells, T8; CD8+ T cells, M, Myeloid cells; D, Dendritic cells; NK, natural killer cells; t-SNE, t-distributed stochastic neighbor embedding; FlowSOM, Flow Self-Organizing Map; FDR, false discovery rate.

Figure 3

Expansion of a myeloid specific signature in TMS patients during relapse. (A) FlowSOM resulted metaclusters in myeloid cells projected on the tSNE map (a representative example from one healthy donor is shown here). (B) Heatmap for key markers to identify to characterize metaclusters. A tentative biological name was assigned to each metacluster based on the most abundant myeloid markers expressed by each cluster. (C) MS patients stratified in different disease subgroups (PPMS; n=8, RRMS; n=9, TMS; n=10). RRMS and TMS were further subdivided in those in remission (RRMS; n=4, TMS; n=5) and relapse (RRMS; n=4, TMS; n=5). Box plots showing the frequency (expressed as % of CD45+ cells) of each myeloid cell metacluster in peripheral blood mononuclear cells of patient subgroups and healthy donors (n=10). Statistically significant changes are shown in the image (p < 0.05 was considered significant, non-parametric Kruskal Wallis test with correction for multiple comparisons (FDR) was applied). (D, i) Gating strategy for cells belonging to the Myeloid cell compartment (non-T/B/NK) seeded in the CITRUS algorithm. CITRUS results include the features (D, ii), abundance of clusters that differentiate groups in comparison (D, iii) and histograms of the expression of markers to identify the immune subsets that these clusters represent (D, iv). PPMS, Primary Progressive Multiple Sclerosis; RRMS, Relapsing Remitting Multiple Sclerosis; Rem, Remission; Rel, Relapse; TMS, Tumefactive Multiple Sclerosis; HD, Healthy donors; t-SNE, t-distributed stochastic neighbor embedding; FlowSOM, Flow Self-Organizing Map; CCR2 or CD192, C-C chemokine receptor type 2, CD172a/b or SIRPa/b, signal-regulatory protein alpha/beta; PD-L1, Programmed death-ligand 1; TNF-α, tumor necrosis factor alpha; IL-10, Interleukin 10; FDR, false discovery rate.

Figure 4

Selective CD4+ T cell lymphopenia in TMS patients during relapse. (A) FlowSOM resulted metaclusters in CD4+ T helper cells projected on the tSNE map (a representative example from one healthy donor is shown here). (B) Heatmap for key markers to identify to characterize metaclusters. A tentative biological name was assigned to each metacluster based on the most abundant myeloid markers expressed by each cluster. (C) MS patients stratified in different disease subgroups (PPMS; n=8, RRMS; n=9, TMS; n=10). RRMS and TMS were further subdivided in those in remission (RRMS; n=4, TMS; n=5) and relapse (RRMS; n=4, TMS; n=5). Box plots showing the frequency (expressed as % of CD45+ cells) of each CD4+ T cell metacluster in peripheral blood mononuclear cells of patient subgroups and healthy donors (n=10). Each dot represents the value of each sample. Statistically significant changes are shown in the image (p < 0.05 was considered significant, non-parametric Kruskal Wallis test with correction for multiple comparisons (FDR) was applied, p<0.05, p<0.01). PPMS, Primary Progressive Multiple Sclerosis; RRMS, Relapsing-Remitting Multiple Sclerosis; Rem, Remission; Rel, Relapse; TMS, Tumefactive Multiple Sclerosis; HD, Healthy donors; IQR, interquartile range; t-SNE, t-distributed stochastic neighbor embedding; FlowSOM, Flow Self-Organizing Map; Eff., Effector cell population; T4; CD4+ T cell, Fol., Follicular T helper cells, Th1-like, T helper 1-like cells, Th-2 like; T helper 2 -like cells; CCR, CC chemokine receptors; CXCR, CXC chemokine receptor; FDR, false discovery rate.

Figure 5

Reshaping of myeloid signature population under glatiramer acetate and anti-CD20 therapy. (A) Box plots showing the frequency (expressed as % of CD45+ cells) of each major cell-lineage cluster in peripheral blood mononuclear cells of MS patients divided in three groups; those under no anti-DMT therapy (n=13, both TMS and RRMS are included, PPMS excluded), those under anti-CD20 treatments (n=5, both TMS and RRMS are included) and those under glatiramer acetate (n=5, both TMS and RRMS are included). (B) FlowSOM identified metaclusters in myeloid cells and heatmap showing, for each of the metaclusters generated, in the three groups of interest (MS under no DMT, MS under glatiramer acetate, MS under anti-CD20), the average intensity of each myeloid cell related marker (stage or activation marker). Higher average expression of each marker is indicated with a green-yellow color, and lower expression in black. Absence of expression is depicted with blue. (C) Box plots showing the median expression of each marker on total myeloid cells in peripheral blood mononuclear cells of patient subgroups (MS under no DMT, MS under glatiramer acetate, MS under anti-CD20). (D) tSNE plots of one TMS patient during relapse and after anti-CD20 treatment. The expression levels of indicative markers are shown in plots. p < 0.05 was considered significant, non-parametric Kruskal Wallis test with correction for multiple comparisons (FDR) was applied). FDR, false discovery rate; TMS, Tumefactive Multiple Sclerosis; DMT, disease modifying therapy; FlowSOM, Flow Self-Organizing Map; CCR2 or CD192, C-C chemokine receptor type 2TNF-α, tumor necrosis factor alpha, IL-10; Interleukin 10, IL-6; Interleukin 6.

Figure 6

Natural killer (NK) cell populations expanded in PPMS patients characterised by the presence of leptomeningeal enhancement in MRI. (A) FlowSOM identified metaclusters in NK cells (a representative example from one healthy donor is shown here). (B) Box plots showing the frequency (expressed as % of CD45+ cells) of each NK cell metacluster in peripheral blood mononuclear cells of patient subgroups and healthy donors (n=10). MS patients stratified in different disease subgroups (PPMS; n=9, RRMS; n=9, TMS; n=10). RRMS and TMS were further subdivided in those in remission (RRMS; n=4, TMS; n=5) and relapse (RRMS; n=4, TMS; n=5). A tentative biological name was assigned to each metacluster based on marker expression, shown in the heatmap (C). (D) Box plots showing the frequency (expressed as % of CD45+ cells) of each NK cell metacluster in peripheral blood mononuclear cells of PPMS patients subgrouped in those characterised by the presence of leptomeningeal enhancement (PPMSact) and those without (PPMSinact), as assessed by novel MRI techniques (shown in material and methods). p < 0,05 was considered significant, non-parametric Kruskal Wallis test with correction for multiple comparisons (FDR) was applied, * p<0,05). PPMS, Primary Progressive Multiple Sclerosis; act, active; inact, inactive; RRMS, Relapsing Remitting Multiple Sclerosis; Rem, Remission, Rel, Relapse; TMS, Tumefactive Multiple Sclerosis; HD, Healthy donors; IQR, interquartile range; t-SNE, t-distributed stochastic neighbor embedding; FlowSOM, Flow Self-Organizing Map; PD-L1, Programmed death-ligand 1; TNF-α, tumor necrosis factor alpha; IL-10, Interleukin 10; IL-6, Interleukin 6; CCR, CC chemokine receptors; FDR, false discovery rate; i, immature; m, mature.

Figure 7

Hierarchical clustering of normalized peripheral blood immune subset frequencies in HD and MS patients. Hierarchical clustering (A–C) of normalized immune subset frequencies (% of total CD45+ cells) in peripheral mononuclear cells from all patients (n=27; PPMS=8, RRMS=9, TMS=10) and healthy donors (HD; n=10). Gender (male, female) from healthy donors and patients is also indicated in the lower part of the heatmap.

Figure 8

Correlation matrix of immune subsets in relapsing MS (RRMS and TMS) patients. Relationships between immune subsets in RRMS during relapse (matrix A and table B) and TMS patients during relapse (matrix C and table D). In the matrices, (A, C) correlations are color-indicated (red for positive, blue for negative). Selected top significant positive correlations with a Spearman R > 0.9 and negative R < -0.9 are shown in tables, (B, D) *p <0.05, ** p<0.01.