Cell-binding IgM in CSF is distinctive of multiple sclerosis and targets the iron transporter SCARA5

Abstract

Intrathecal IgM production in multiple sclerosis is associated with a worse disease course. To investigate pathogenic relevance of autoreactive IgM in multiple sclerosis, CSF from two independent cohorts, including multiple sclerosis patients and controls, were screened for antibody binding to induced pluripotent stem cell-derived neurons and astrocytes, and a panel of CNS-related cell lines. IgM binding to a primitive neuro-ectodermal tumour cell line discriminated 10% of multiple sclerosis donors from controls. Transcriptomes of single IgM producing CSF B cells from patients with cell-binding IgM were sequenced and used to produce recombinant monoclonal antibodies for characterization and antigen identification. We produced five cell-binding recombinant IgM antibodies, of which one, cloned from an HLA-DR + plasma-like B cell, mediated antigen-dependent complement activation. Immunoprecipitation and mass spectrometry, and biochemical and transcriptome analysis of the target cells identified the iron transport scavenger protein SCARA5 as the antigen target of this antibody. Intrathecal injection of a SCARA5 antibody led to an increased T cell infiltration in an experimental autoimmune encephalomyelitis (EAE) model. CSF IgM might contribute to CNS inflammation in multiple sclerosis by binding to cell surface antigens like SCARA5 and activating complement, or by facilitating immune cell migration into the brain.

Keywords: complement; immunoprecipitation; iron metabolism; plasma cells.

Figure 1

Binding of antibodies from CSF to live cells. (A) Binding of antibodies from CSF to astrocytes or neurons derived from human induced pluripotent stem cells (iPSCs). Each column scatter plot shows results for CSF samples from donors with a diagnosis of clinically isolated syndrome (CIS), clinically definite multiple sclerosis (MS), or controls. Upper row of plots shows binding to neurons and lower plots binding to astrocytes. Plots on left show IgM, plots in middle show IgG and plots on right show IgA. Each dot represents one sample. Vertical axis shows geometric mean fluorescence intensity. Results from CSF samples from donors with CIS are shown in black, MS in red, and controls in blue. (B) Exemplary flow cytometry dot plot comparing IgM binding from CSF against the indicated cell lines. The vertical axis shows the fluorescence intensity of the secondary antibody used to detect the bound antibody, and the horizontal axis shows the intensity of the fluorescent vital dye Cell Trace Violet used to label the various cell lines before mixing, to enable their subsequent separation. Red dot plot shows IgM binding from CSF donated by a patient with CIS. Blue dot plot shows results with secondary antibody only. (C) Relationship between PNET-binding IgM signal (vertical axis, GMFI ratio) and CSF/serum IgM quotient (horizontal axis) (Pearson r = 0.096, P = 0.176). (D) Binding of antibodies from CSF to various cell lines. Vertical axis shows GMFI ratio (antibody + secondary-labelled cells divided by cells labelled with secondary antibody only). The 18 columns represent binding of IgM (upper row) or IgG (lower row) from CSF samples from donors in Basel with CIS (black), MS (red) and controls (blue) to each of six cell lines whose names are shown inside each plot. One-way ANOVA *P = 0.049, ****P < 0.0001. ns = not significant.

Figure 2

Properties of monoclonal IgM isolated from single CSF B cells. (A) Contour plots showing the PNET binding of IgM from the single cell culture supernatants from the wells highlighted in Supplementary Fig. 2C. Names above each plot are the identities of the five wells, derived from their locations in the culture plates, and are used hereafter as identifiers for the monoclonal antibodies derived from these B cells. In each plot, the vertical axis shows IgM binding (fluorescence intensity of anti-human IgM) and the horizontal axis shows the forward scatter. Contours in blue are derived from cells labelled with anti-human IgM only, and contours in red from cells labelled with the single cell supernatants and then with anti-human IgM. (B) Contour plots like those shown in A representing binding of the five monoclonal antibodies cloned from the B cells in each of the five wells. (C) Heat map showing binding of each of the five monoclonal antibodies (rows) to each of the 13 cell lines (columns). Blue intensity shown on bar at right corresponds to the GMFI ratio. (D) Effect of mutational reversion on PNET binding. Binding of the recombinant antibody B3 with the observed mutations as originally cloned, in comparison with partially or fully germline-reverted versions of the same antibody. The red histogram shows binding of the original hypermutated antibody. The orange histogram shows binding of an artificial antibody with the native mutated heavy chain (HC) combined with a germline-reverted light chain (LC). Blue histogram shows binding of antibody with reverted heavy and original mutated light chain. Green shows fully germline reverted version. (E) Transcriptional characteristics of B cell B3. cDNA from each of five singly cultured B cells was subjected to Illumina sequencing and for each gene observed, the log10 number of reads found in B3 was plotted against the log10 mean number of reads found in the other B cells (blue dots). Genes whose expression is >5-fold higher in B3 are marked with red circles and for a subset, their gene symbols printed beside the plotted dots. This information for all the B3 overabundant genes is collated in Supplementary Table 5. Genes regarded as plasma cell markers have their symbols in black, genes associated with antigen presentation in dark blue. (F) Antibody-dependent complement activation mediated by B3. PNET cells were incubated with monoclonal IgM B3, or with M16, in the presence of human serum from a healthy donor as a source of complement. Serum only without additional antibody was used as a control. Antibody-dependent complement activation was detected by bound C3 immunolabelling flow cytometry. Vertical axis shows GMFI ratio. *P = 0.0385, one-way ANOVA. ns = not significant.

Figure 3

Target antigen of IgM B3. (A) Effects of protease treatment or deglycosylation on B3 binding. Histogram on left shows binding to untreated PNET cells in dark blue and binding to cells treated with the non-specific protease mixture Pronase to cleave exposed cell surface proteins in light blue. Histogram on right shows the effect of deglycosylation treatment on B3 binding. Binding to untreated PNET cells is shown in dark blue, binding to cells treated with various deglycosylating enzymes to specifically cleave terminal sugar moieties from cell surface carbohydrates is shown with other coloured histograms, as specified on right of plot. Specificity controls for deglycosylation treatments are shown in Supplementary Fig. 3. (B) Immunoprecipitation and mass spectrometry to identify the unknown protein target of the PNET-binding IgM B3. Live PNET cells were first incubated with B3, then washed and lysed; the lysate was next incubated with anti-IgM coated Dynabeads, washed and magnetically retrieved. Proteins were eluted from the beads, digested with trypsin, and peptides detected by mass spectrometry. Data are pooled from three independent experiments. Vertical axis shows log2 number of peptides precipitated by B3. Horizontal axis shows same parameter from control preparation in which the B3 was omitted. Each dot is one protein. Red dots are proteins whose retrieval was significantly more likely in the B3 samples (two-tailed unpaired t-test, P < 0.1), filtered for only those genes whose gene ontology (GO) terms include ‘plasma membrane’. These proteins are labelled with their gene symbol and the P-value is shown below each symbol. (C) Transcriptional characteristics of the PNET cell line bound by B3. cDNA from each of 13 cell lines was subjected to Illumina sequencing and for each gene observed, the log10 number of reads found in the PNET cell lines was plotted against the log10 maximum number of reads found in the other B cell lines (light blue dots). Genes whose expression is significantly higher in PNET (two-tailed, unpaired t-test, P < 1 × 10⁻⁸), and whose GO terms include ‘plasma membrane’ are marked with red circles and their gene symbols printed adjacent to the plotted. (D) Binding of B3 to cells transfected with either PCDH18 or SCARA5 measured by flow cytometry. B3 non-binding HEK cells were transfected with plasmids mediating expression of either PCDH18 (left) or SCARA5 (right) and co-transfected with fluorescent protein transfection markers (mCherry for PCDH18 and GFP for SCARA5). Vertical axis of each contour plot shows B3 binding and horizontal axis shows transfection marker. (E) Impact of isoform. Analogous experiments to D were conducted using HEK cells transfected with each of the 176-, 400- and 495 amino acid isoforms of SCARA5. Vertical axis shows binding of a commercial mouse anti-human SCARA5 antibody (left plots) or B3 (right plots). Horizontal axis shows forward scatter (no transfection marker used in this experiment). Blue contours are results with the commercial or B3 antibody, red contours are results from cells incubated with the appropriate secondary antibody only. (F) Prevalence of IgM antibodies against PNET cells and against SCARA5 in a cohort of donors with various diagnoses. Vertical axis shows IgM binding. Six sections left to right show results from CSF from donors with different diagnoses (left to right: CIS, PPMS, RRMS, SPMS, healthy controls, inflammatory non-MS, as specified within each plot). Within each of the six plots, the left-most column shows binding to HEK cells transfected with the 495-amino acid isoform, the middle column binding to the 176-amino acid isoform, and the right column binding to PNET cells. (G) Binding of antibodies from serum to SCARA5-expressing cells. Vertical axis shows GMFI ratio (antibody + secondary-labelled cells divided by cells labelled with secondary antibody only obtained from SCARA5-HEK cells, again divided by the same value obtained from untransfected HEK cells). Each separate plot represents binding IgG (on the left), IgA (in the middle) or IgM (on the right) from serum samples from donors with CIS (black), MS (red) and controls (blue). Comparison of antibody binding between different diagnosis was performed by one-way ANOVA. CIS = clinically isolated syndrome; infl. CTRL = inflammatory control; ns = not significant; PPMS = primary progressive multiple sclerosis; RRMS = relapsing remitting multiple sclerosis; SPMS = secondary progressive multiple sclerosis.

Figure 4

Impact of anti-SCARA5 antibodies on immune cell trafficking into brain. (A) Immunofluorescent labelling of SCARA5 expression in mouse brain. Vibratome sections from perfusion-fixed brains of unmanipulated mice were immunolabelled with sheep anti-SCARA5 antibodies and visualized by confocal microscopy. Two upper images show surface (subdural) and penetrating blood vessels of the superficial cortex. The upper left image shows nuclei labelled with DAPI in blue and immunolabelling for SCARA5 in red. The upper right image shows a similar location from a control section, from which the primary antibody was omitted, imaged with the same parameters. In the lower two images, labelling of periventricular blood vessels is shown. The left image is labelled with sheep anti-SCARA5 (green) and the right is a no-primary control. (B) Anti-SCARA5 binding to live blood vessels. A polyclonal sheep anti-SCARA antibody was injected into the lateral ventricle of a mouse and the next day the mouse was sacrificed by perfusion fixation, and the brain labelled with DAPI (blue) and anti-sheep secondary antibody (red). The upper panel is a composite image made by stitching together images captured with a 10 × objective, to show both the ventricles—the ventricle highlighted with a box on the left received the injection. The panel below is a maximum intensity projection of a confocal stack of the area shown in the box, captured with the 40 × objective showing the blood vessels in the wall of the ventricle labelled by the in vivo antibody injection. (C–G) Impact of intraventricular anti-SCARA5 antibodies on encephalitic T cell infiltration. All figures show results pooled from three experiments, each with 3–6 animals per group. (C) Schematic diagram of experimental design (created with BioRender.com). In vitro activated, MOG-specific 2D2 T cells were transferred to recipient animals by intraperitoneal injection, followed by pertussis toxin, and animals were tracked until the first animals showed motor signs, at which point, the remaining asymptomatic animals were randomized into two groups to receive anti-SCARA5 antibody or control. (D) Motor impairment of animals over time. Motor function was subjectively assessed according to a system used for scoring experimental autoimmune encephalomyelitis (EAE), in which 0 is unimpaired, and 4 is severe impairment. Points show the means within the conditions and the bars the standard error of the mean (SEM). Red squares indicate the animals injected with anti-SCARA5 antibody and blue circles control animals injected with PBS or control antibody. P = 0.0936 calculated by two-tailed unpaired t-test between the areas under the curves for each condition. (E) Weights of animals over time of experiment. Weights are expressed as the weight at each time point divided by the weight at the start of the experiment for each animal. Points show the means within the conditions and the bars the SEM. Purple circles represent control animals and blue squares the animals injected with anti-SCARA5. (F) T cell infiltration. Vibratome section of brain from animals sacrificed at 1 week after intraventricular antibody injection in the experiment described above were immunolabelled for laminin (blood vessel walls, red) and CD3-epsilon (T cells, green). (G) Quantification of infiltrating T cells. Images like the ones shown in F were captured for 5–6 sections per mouse by an automated fluorescent microscope and the numbers of T cells per section extracted by an automated script in ImageJ. Points show the median number of T cells per section for each mouse and the whiskers show the interquartile range. **P = 0.0012, two-tailed unpaired t-test. ns = not significant.