Differential contribution of immune effector mechanisms to cortical demyelination in multiple sclerosis

We screened a cohort of 740 archival CNS biopsies of patients diagnosed with inflammatory demyelinating disease between 1999 and 2016 for the presence of cortical demyelination. Biopsies were performed for diagnostic reasons to exclude tumor or infection and targeted MRI-defined white matter lesions (Table S1). The study was approved by the ethical review committee of the University Medical Center Göttingen (#19/09/10). All biopsies were stained with HE and LFB/PAS and Bielschowsky silver impregnated. Immunohistochemistry (IHC) was performed with antibodies against myelin, oligodendrocytes, adaptive and innate immune cells, neurons and axons, astrocytes, including AQP4, JC virus and Ki-67. Biopsies with cortical demyelination were in addition stained with antibodies against CCR2 and double-labelled for CCR2/CD14 and CCR2/KiM1P to identify recently invaded monocytes and tissue macrophages, and GrB/CD3 to identify natural killer (NK) cells. Demyelinating lesion activity was determined according to Brück et al. [10] and Kuhlmann et al. [27]. Lesions without myelin-laden macrophages were staged according to Schirmer et al. [47]. Primary antibodies used for IHC of human MS biopsies are listed in Table S2 and exemplary IHC of tonsillar tissue with antibodies against CD3, CD4, CD8, CD20 and CD138 is provided as Supplemental Fig. 6.

135 biopsies (18%) contained cortical tissue in addition to subcortical white matter (Figure S1). Of those, 26 patients (19%) showed cortical demyelination identified by myelin immunohistochemistry. All but two patients also harbored subcortical white matter for analysis. 20/26 patients showed white matter demyelination in addition to cortical demyelination. Intracortical (type 2) lesions were exceedingly rare in this cohort, and leukocortical (type 1) lesions were not included in the present study.

The median age of the patients at biopsy was 41 (interquartile range, 30–49), the median time to biopsy was 2 months (interquartile range, 0.5–24) and the female to male ratio was 2.3. 10/26 patients had seizures or were diagnosed with symptomatic epilepsy and 8/26 patients presented with moderate to severe neuropsychiatric symptoms (confusion, cognitive impairment, depression; Table S1). The clinical diagnosis at the time of biopsy was CIS in 16, RRMS in 8 and SPMS in 2 cases.

C57BL/6J mice were obtained from Charles River Laboratories (Charles River). 2D2 mice (TCR^MOG) [6], Th/+mice (IgG^MOG, kindly provided by Tobias Litzenburger, Antonio Iglesias and Hartmut Wekerle [31]), OSE double transgenic mice (TCR^MOG × IgG^MOG) [5, 26], RAG1^−/− mice [37], RAG2^−/− γc^−/−, RAG1^−/− γc^-/-, C1q^−/− (kindly provided by Marina Botto [8]) and OT-II [4] mice were bred at the animal facility of the University of Göttingen under SPF conditions. Th/+CCR2^−/−, Th/+C1q^−/− and Th/+CD59a^−/− mice were generated by crossbreeding CCR2^−/− [29] C1q^−/− or CD59a^−/− (kindly provided by B.P. Morgan [20]) mice with Th/Th mice. Knockout animals were backcrossed against C57BL/6J for more than 12 generations and housed at a density of 2–5 animals/cage in a temperature controlled environment with a 12/12 h light/dark cycle and food and water ad libitum. If not otherwise stated, 6–10 week old mice of both sexes were randomized into groups. Adult common marmosets of both sexes (Callithrix jacchus) were provided by Neu Encepharm (Göttingen, Germany).

2D2 animals were immunized subcutaneously (s.c.) with 200 µg MOG₃₅₋₅₅ (purchased from the Institute of Medical Immunology, University Medical Center Charité, Berlin) emulsified in CFA substituted with 5 mg/ml Mycobacterium tuberculosis H37Ra (Difco, 231141). C57BL/6J, Th/+ and OSE mice were immunized s.c. with 100 µg recombinant rat MOG_1–125/CFA. 300 ng/mouse pertussis toxin (PTX) (List Biological Laboratories, #180) were injected i.p. at day 0 and day 2 after immunization. Control animals were naïve or immunized with CFA only. Age- and sex-matched common marmosets were immunized s.c. with 50 µg recombinant rat MOG_1–125 emulsified in IFA supplemented with 0.25 mg/ml Mycobacterium butyricum (Difco, 264010). EAE animals were scored as previously described for mice [39] and marmosets [50].

Monocyte depletion in marmoset monkeys was initiated 14 days after immunization by twice weekly i.v. injections of 5 mg/kg DOC-2 Fr-2 (marmoset IgG1-chimeric humanized mouse anti-human CCR2 antibody). Controls received 5 mg/kg marmoset IgG1-chimeric isotype control antibody. The administration frequency was reduced to once weekly from day 28 to the end of the experiment. In mice, all depletion and blocking experiments started at disease onset. NK cells were depleted in Th/+ mice by daily i.p. injections of 300 µg of the mouse monoclonal anti-NK1.1 antibody (Clone PK136, Bio X Cell, BE0036). Control Th/+ mice received 300 µg i.p. of the isotype control antibody C1.18.4 (Clone C1.18.4, Bio X Cell, BE0085). To block the formation of the membrane attack complex (MAC) 2 µg of the BB5.1 monoclonal antibody against mouse complement component C5 (Hycult biotech, HM1073) [23] or a mouse IgG1 control antibody (BioLegend, Clone MOPC-21) was injected intracerebrally at the time point of stereotactic cytokine injection.

Mice were anaesthetized i.p. by injection of ketamine/xylazine and mounted on a stereotactic device (Stoelting Co, Germany). The scalp was opened to expose the skull and a fine hole was drilled 0.1 mm caudal to the bregma and 0.2 mm lateral to the midline. A finely calibrated glass capillary was inserted into the brain (0.7 mm depth) allowing the intracerebral administration of 2 µl of a mixture composed of 50 ng TNFα (R&D Systems) and 60 ng IFNγ (R&D Systems). Immunized animals were injected on the 2nd day of disease, animals which received cell-depleting antibodies on day 3 after EAE onset. Monastral blue (Sigma-Aldrich) was added to the cytokine mixture to facilitate the identification of the lesions in the tissue.

Spleen cells from 2D2 or OT-II mice were expanded with plate bound anti-CD3 (4 µg/ml, Bio X Cell, Clone 145-2C11, BE0001-1) and soluble anti-CD28 (1 µg/ml, Bio X Cell, Clone PV1, BE0015-5) in the presence of 1 ng/ml rm IL-12 (R&D systems). Cells were restimulated with 15 µg/ml MOG_35–55 or 15 µg/ml chicken ovalbumin 323–339 (OVA) and 30 Gy-irradiated antigen presenting cells for 3 days and 10 million T cell blasts were injected i.p. into RAG1^−/−, RAG1^−/− γc^−/− or RAG2^−/− γc^−/− animals. 12 h after adoptive transfer, all animals were s.c. immunized with 10 µg MOG_35–55 or OVA peptide and received 300 ng PTX i.p. On the 2nd day of EAE 1.5 mg/animal of the MOG-specific antibodies 8-18C5 (IgG1 isotype) or Z2 (IgG2a isotype) was injected i.v., and animals were subjected to stereotactic surgery.

Male Th/+ and C57BL/6J mice were individually equipped with a running wheel with regularly spaced crossbars (conventional wheels) where they could run freely at any time for 2 weeks [30]. After this period they were immunized s.c. with a subclinical dose of 10 µg recombinant rat MOG_1–125/CFA, therefore, not developing clinical disease. The mice were kept on the conventional running wheels for further 11 days and then randomized according to their wheel running performance into two comparable groups which received intracortical stereotactic injections of 2 µl PBS or cytokines. One day after surgery, all animals were put onto wheels with irregularly spaced crossbars (complex wheels). Wheel running performance was recorded continuously by LabVIEW™-based software.

The meninges were removed and the cortex was separated from the white matter 2 days after stereotactic injection, cut and digested for 45 min at 37 °C with 2.5 mg/ml Collagenase D (Roche) and 1 mg/ml DNAse I (Roche). Mononuclear cells were isolated by Percoll gradient centrifugation (37%/70%, GE Healthcare) removed from the interphase, washed and subsequently blocked with αCD16/32 (BioLegend, Clone 93) for 15 min. The following antibodies were used (all from BioLegend or eBioscience): αCD3 (145-2C11), αCD4 (RM4-5), αCD8 (53-6.7), αCD11b (M1/70), αCD11c (N418), αCD19 (eBio1D3), αCD25 (PC61.5), αCD45 (30-F11), αFoxP3 (FJK-16S), αNK1.1 (PK-136), αNkp46 (29A1.4), αLy6C (HK1.4), αLy6G (1A8), αCCR2 (R&D 475301), αγδ TCR (eBioGL3), αB220 (RA3-6B2). For intracellular αFoxP3 staining, cells were fixed after surface staining for 45 min and permeabilized for 30 min using the eBioscience FoxP3 staining kit. All flow cytometry data were acquired on a FACS Canto™ II (BD Bioscience) device and analyzed with the FlowJo software (v. 7.6.1, Tree Star Inc).

Blood samples of marmoset monkeys were analyzed by flow cytometry using the following directly labelled antibodies specific for: CD20 (B-Ly1, DakoCytomation), CD3 (SP34, BD Bioscience), CD14 (M5E2), CD11b (D12). For indirect staining, marmoset or human EDTA blood was first stained with the mouse-anti-human CCR2 antibody DOC-2 (5 µg/ml) followed by PE-labelled rabbit anti-mouse F(ab)2 (DakoCytomation), or with the humanized or marmoset chimeric DOC-2 Fr-2 followed by FITC-labelled goat anti human Fcγ specific F(ab’)2 (Jackson Immunoresearch).

Th/+ mice, healthy or previously immunized with rMOG_1–125, were stereotactically injected with IFNγ and TNFα as described above and perfused 24 h afterwards. One hour before perfusion, the mice received i.v. injections of 100 µg FITC-albumin/g body weight (albumin–fluorescein isothiocyanate conjugate, Sigma, A9771).

To identify mature oligodendrocytes and oligodendrocyte precursor cells (OPC) in the cortical demyelinated areas, double-IHC was performed using a rabbit anti-TPPP/p25 antibody against tubulin polymerization-promoting protein (Abcam, 92305, 1:500) or a rabbit anti-Olig2 antibody (IBL, 18953, 1:300) respectively, on sections previously stained with rat anti-MOG antibody (purified in our laboratory). Double-IHC was also performed for myelin (anti-MOG) and amyloid precursor protein (APP, Chemicon, MAB348, 1:2000) as a marker of axonal damage or NeuN (Millipore, MAB377, 1:200) for neurons. Polymorphonuclear cells were detected by enzyme histochemistry using a chloroacetate esterase assay (Naphtol AS-D Chloroacetate esterase kit; Sigma-Aldrich).

NK cells infiltrating the cortex of stereotactically injected Th/+ mice were detected in cryosections using the primary goat anti-mouse antibody NKp46/NCR1 (R&D Systems, AF2225, 1:50). Briefly, mice where perfused with cold PBS. The brain was sectioned, mounted in OCT medium (Tissue-Tek; Sakura Finetek USA) and frozen in isopentane (Sigma-Aldrich) pre-cooled in liquid nitrogen. Serial sections 6 μm thick were cut in a cryostat at −20 °C and fixed in ethanol for 1 h before applying the NKp46 antibody. Antibody binding was visualized using DAB. To distinguish NK from NKT cells, we performed double-immunofluorescence on cryosections from stereotactically injected Th/+ mice using the goat anti-NKp46 antibody described above and a rabbit anti-CD3 antibody (Dako, A0452, 1:50). For visualizing NKp46, a donkey anti-goat biotinylated secondary antibody (GE Healthcare Ltd, RPN1025, 1:200) was added for 1 h at room temperature in 10% FCS/PBS, followed by 30 min incubation with Cy™3-conjugated streptavidin (Jackson ImmunoResearch Lab, 016160184, 1:100). The AlexaFluor^® 488 goat anti-rabbit antibody (Life Technologies, A11034, 1:200) was used to visualize the CD3 signal. Nuclei were counterstained with 4′,6-diamino-2-phenylindole (DAPI). Sections were examined using a fluorescence microscope (Olympus BX51, Germany).

Spinal cord sections from Th/+ CCR2^+/+ and Th/+ CCR2^−/− were stained with Luxol fast blue (LFB)/periodic acid-Schiff (PAS) and evaluated for the extent of demyelination. Finally, paraffin-embedded brain sections from Th/+ mice injected with FITC-albumin were immunolabelled with a rabbit anti-FITC antibody conjugated to horseradish peroxidase (Dako, P5100, 1:50) to measure the area of FITC extravasation.

The quantification of cortical demyelination was performed on MBP-immunostained sections in mouse brains at the aforementioned time points. Images were acquired using a 40× magnification of a light microscope (Olympus BX51, Germany) equipped with a digital camera (Software CellSens Dimension v.1.7.1, DP71, Olympus, Germany). To measure cortical demyelination in the marmosets, MBP-immunolabelled sections were scanned using the Keyence Bio REVO BZ-X710 microscope (Keyence, Japan). Compound images of the entire coronal brain section analyzed were generated by combining single images using the BZ-II analyzer software (Keyence, Japan). ImageJ (v. 1.46r, NIH, USA) was used to measure the subpial demyelinated area and the extent of perivascular cortical demyelination in the digital composite images. In mice, subpial demyelination was quantified in the ipsilateral (injected) brain hemisphere, whereas perivascular demyelination was measured in both the ipsi- and contralateral cortex. Subpial demyelination (%) and the area of perivascular demyelination (mm²) are reported. Subpial demyelination was calculated as follows:

Since cortical demyelination in mice was most extensive at day 5 post stereotactic injection, we selected this time point (if not stated otherwise) for quantitative analysis. Cell densities were determined using a 10 × 10 ocular morphometric grid (at 400× magnification) (Olympus, Japan), and the results were expressed as cells/mm². The densities of p25+, Olig2+ and NeuN+ cells were assessed in the ipsilateral cortical layers I and II/III of C57BL6/J mice and in the subpial demyelinated areas across the same layers in Th/+ mice. The analysis of APP⁺ axons was restricted to the perivascular demyelinated areas in both hemispheres. For assessment of blood–brain barrier permeability the area of FITC-albumin extravasation was measured using ImageJ. To this end, a color deconvolution plugin for ImageJ was used [45], and the threshold established for quantifying the intensity of the brown (DAB) channel was set at (0; 150). For the analysis of macrophages/activated microglia in mice, the number of intracortical Mac-3+-perivascular cuffings was counted in both hemispheres. For the purpose of quantification, infiltrates restricted to the Virchow–Robin space, without evidence for cellular transmigration into the cortical parenchyma, were not considered bona fide perivascular infiltrates. The area of spinal cord demyelination in mice (Th/+ CCR2^+/+, Th/+ CCR2^−/−) and marmosets was measured on LFB/PAS stained sections using ImageJ. Digital images were acquired from at least eight transversal spinal cord sections from each animal, the demyelinated areas were measured and reported as % of total white matter. The densities of KiM1P+ macrophages/activated microglia and T cells in biopsies from early MS cases were determined using an ocular morphometric grid (Olympus, Japan). Immunostained cells were counted in regions with subpial cortical demyelination, as well as in normal appearing cortical regions (NAGM). At least five separate visual fields were quantified for each cortical region within each sample. Densities were reported as cells/mm². The investigators (NLD, CS) were blinded to the experimental groups.

Th/+ CCR2^+/+ or Th/+ CCR2^−/− mice were perfused with PBS 2 days after stereotactic cytokine injection. Brains were dissected and the ipsilateral cortex near the injection site was separated from the white matter and conserved in QIAzol Lysis Reagent (Qiagen). RNA was isolated from the tissue using the RNeasy Micro Kit (Qiagen, Germany) and subsequently transcribed into cDNA using the High Capacity RNA-to-cDNA™ Kit (Life Technologies). From the cDNA, 12.5 ng were used per PCR reaction. Quantitative PCR was performed on the iQ5™ Real Time PCR Detection System (BIO-Rad) using the qPCR Core Kit obtained from Eurogentec. The following FAM labeled primers/probes (TaqMan^® Gene Expression Assays) were selected to be intron spanning from Life Technologies: Gapdh (Mm99999915 g1), TNF-α (Mm00443258_m1), NOS2/iNOS (Mm00440502_m1) and CCL2 (Mm00441242_m1). The expression levels of TNF-α, NOS2/iNOS and CCL2 are indicated as the percentage of the housekeeping gene Gapdh.

Normality tests (D’Agostino and Pearson omnibus normality test) were applied to all data sets. If the samples followed a normal distribution, the unpaired t test or a paired t test were applied. In the absence of normality, the non-parametric Mann–Whitney U test or the Wilcoxon signed-rank test were used. One-way-ANOVA followed by Dunn’s Multiple Comparison post hoc test was used for comparisons between three or more groups. p < 0.05 was considered to be significant. Unless otherwise specified, all results are shown as mean ± SD.

The sponsors of this study had no role in study design, data collection, data analysis, data interpretation and writing of the manuscript. The corresponding authors had full access to all the data in the study and had final responsibility for the decision to submit the manuscript.

In a next step, we strived to model cortical demyelination in the mouse to obtain an easily manipulable system in which the relative contribution of adaptive and innate immune cells as well as humoral components for cortical demyelination can be assessed. Cortical demyelination is not a regular feature of experimental autoimmune encephalomyelitis (EAE) in MOG_35–55-peptide- or rMOG_1–125-immunized C57BL/6 mice. To mimic the pro-inflammatory cytokine milieu found in the CSF of patients with cortical demyelination, we stereotactically injected TNFα and IFNγ into the right motor cortex of MOG_35–55-peptide immunized 2D2 mice (TCR transgenic for MOG), rMOG_1–125-immunized C57BL/6 mice or rMOG_1–125-immunized Th/+ mice (BCR transgenic for MOG) at the 2nd day of clinical EAE. Subpial and perivascular cortical demyelination were only observed in Th/+ mice (Fig. 3a), which carry the Ig heavy chain knockin of 8-18C5, an established demyelinating MOG-specific antibody. Of note, Th/+ mice developed cortical demyelination subpially and intracortically in both the injected and non-injected hemispheres (Fig. 3a–c). Demyelination in the cortex was most pronounced on day 5 after injection and paralleled by a significant reduction in p25⁺ mature oligodendrocytes (Fig. 3d, h, l). In line with previous studies demonstrating axonal damage in cortical MS lesions, the density of damaged APP⁺ axons was significantly increased in perivascular cortical demyelination of Th/+ mice compared to corresponding cortical regions in C57BL/6J mice (Fig. 3f, j, n), while the density of neurons in the demyelinated cortex was not reduced (Fig. 3g, k, o). A significant decline in perivascular macrophage cuffings and T cell densities could be observed on days 20 and 40 after lesion induction (Fig. 3s, t), and lesions were significantly remyelinated on day 20 perivascularly and on day 40 subpially (Fig. 3p–r). Concordant with the maximum of demyelination observed, cortically demyelinated Th/+ mice were impaired in the motor skill sequence (MOSS) test between days 4 and 6 after lesion induction (Supplemental Fig. 4).

MOG-specific antibodies targeting conformational epitopes are a prerequisite for cortical demyelination in our animal model. They might bind to their cognate antigen on the cell surface of oligodendrocytes and activate C1q, the first component of the classical complement pathway that leads to target cell killing by complement-dependent cytotoxicity (CDC) via membrane attack complexes. In addition, elimination of antibody-opsonized cells might be mediated by NK cells, macrophages, monocytes and neutrophils, which are activated via specific IgG receptors sensing surface bound IgG antibodies. To assess the relative importance of humoral and cellular effector mechanisms for the pathogenesis of myelin damage in subpial and perivascular cortical demyelination, we first strived to detect the membrane attack complex in cortical demyelinated lesions in human MS biopsies. Although we detected subpial IgG leakage in patients with ongoing cortical demyelination (Supplemental Fig. 5 a, c), no evidence of terminal complement activation was found (not depicted). Next, we prevented the generation of the membrane attack complex (MAC) in our animal model by a specific antibody directed against C5 (BB5.1, the mouse analogue of eculizumab) thereby inhibiting its cleavage [51]. Inhibiting terminal complement activation reduced subpial cortical demyelination, whereas the extent of perivascular cortical demyelination was not modulated (Fig. 4a, b). If we compared cortical demyelination in C1q-deficient mice and controls with EAE which received the potent complement-activating and MOG-specific antibody Z2 i.v. prior to stereotactic cytokine injections, the extent of cortical demyelination in C1q-deficient animals was equivalent to controls (Fig. 4c, d). To gain further insight whether CDC is relevant for cortical demyelination in Th/+ mice, we established Th/+ C1q^−/− and Th/+ CD59a^−/− mice, the latter being deficient for a major membrane inhibitor of terminal complement activation in oligodendrocytes. Th/+ C1q^−/− and Th/+ C1q^+/+ mice did neither differ in MOG-specific antibody titres (not depicted) nor in the extent of cortical demyelination (Fig. 4e, f). Furthermore, the extent of cortical demyelination was similar in Th/+ CD59a^−/− and Th/+ CD59a^+/+ mice (Fig. 4g, h). In summary, we provide evidence that CDC activated via the classical complement pathway unlikely contributes to cortical demyelination in Th/+ mice.

We further characterized the cortical inflammatory infiltrate by flow cytometry and identified both adaptive (T and B cells) and innate immune cells (monocytes, granulocytes (PMN) and natural killer (NK) cells) in the demyelinated cortex of Th/+ mice on day 3 after lesion induction (Fig. 4i). To gain insight into the relevance of individual immune cell populations for cortical demyelination, we selectively depleted individual immune cells by specific antibodies or used transgenic mice genetically deficient for adaptive immune cells, NK cells or inflammatory monocytes.

NK cells are the classical effectors of antibody-dependent cellular cytotoxicity (ADCC) against IgG-coated targets and were detected around vessels in both demyelinated cortical lesions of MS biopsies (Fig. 1t) and in cortically demyelinated Th/+ mice (Fig. 5g, h). When we efficiently depleted NK cells in the blood and cortex of Th/+ mice prior to stereotactic surgery (Fig. 5a–c), the extent of subpial cortical demyelination was unaffected by NK cell depletion, but perivascular cortical demyelination was significantly diminished (Fig. 5d–f). Of note, NK cell depletion did not affect the numbers of perivascular or meningeal CD3⁺ T cells (Fig. 5k, l). To further support the relevance of NK cells to perivascular cortical demyelination, we adoptively transferred activated MOG-specific 2D2 T cells into RAG1^−/− (no mature T and B cells) and RAG1^−/− γc^−/− (no mature T and B cells, no NK cells) mice. At disease onset, all animals received 1.5 mg of the demyelinating MOG-specific IgG2a antibody Z2 i.v. and underwent stereotactic cytokine injections into the motor cortex. Whereas RAG1^−/− and RAG1^−/− γc^−/− mice developed similar subpial cortical demyelination, perivascular cortical demyelination was significantly higher in RAG1^−/− compared to RAG1^−/− γc^−/− animals (Fig. 5i, j). Our experiments thus demonstrate that NK cells are relevant for perivascular cortical demyelination in the presence of a pathogenic antibody response, which is in line with their predominant perivascular localization in the inflamed cortex.

Clinical trials in MS have been interpreted in a way that the progressive disease phase might be less dependent on peripherally recruited adaptive immune cells. Since cortical demyelination is prominent in chronic MS, we scrutinized whether T cells are central to cortical lesion formation in our model. Also, we wanted to clarify, whether CNS-antigen specificity and T cell activation influence cortical demyelination. To this end, we combined T- and B cell receptor transgenic mice for MOG (OSE: TCR^MOG × IgG^MOG) and transferred activated MOG-specific T cells (TCR^MOG, named 2D2) or activated, ovalbumin-specific T cells (OT-II, which recognize an antigen not present in mice) into mice deficient for adaptive immune cells (RAG1^−/− mice). Surprisingly, we found that T and B cells were not immediately required for subpial cortical demyelination in the presence of demyelinating antibodies and stereotactically applied cytokines (Fig. 6a, b). However, activated CNS-antigen specific T cells were a prerequisite for perivascular cortical demyelination (Fig. 6c–f). These results were supported by experiments assessing the permeability of the cortical blood–brain barrier via the extravasation of FITC-albumin in healthy, non-immunized Th/+ mice versus diseased Th/+ mice harboring CNS-antigen specific, activated T cells (Fig. 6g). Perivascular cortical extravasation of FITC-albumin was significantly more pronounced in diseased than in healthy Th/+ mice, whereas subpial extravasation of serum proteins was substantial in both experimental paradigms 24 h after lesion induction (Fig. 6h, i). This indicates that intracortically applied cytokines are sufficient to increase the permeability of meningeal and subpial vessels readily while intracortical vessels require the combined action of cytokines and encephalitogenic T cells to become permeable to serum proteins, such as albumin and IgG. In line, IgG leakage was detected subpially in a single patient with ongoing, active cortical demyelination (Supplemental Fig. 5).

Quantitatively, macrophages are by far the dominant inflammatory cell population in MS and correlate with active demyelination. Recent evidence has been provided that blood-derived CCR2^rfp/+ monocytes give rise to highly phagocytic and inflammatory macrophages, which initiate demyelination in EAE [53]. To assess the contribution of CCR2⁺ monocytes for cortical demyelination, Th/+ CCR2^−/− mice and Th/+ CCR2^+/+ controls were immunized with rMOG_1–125 and received stereotactic injections on day 2 after EAE onset. In line with previous reports in CCR2-deficient animals [17], Th/+ CCR2^−/− mice developed EAE at a comparable severity, albeit at later time points than Th/+ CCR2^+/+ controls (Table S2). The recruitment of inflammatory monocytes into the cortex was significantly impaired in Th/+ CCR2^−/− mice compared to Th/+ CCR2^+/+ controls on day 2 after stereotactic injection (Fig. 7a, b). Consistent with the flow cytometry data, Th/+ CCR2^−/− mice had significantly less intracortical Mac-3+ perivascular cuffings in both hemispheres than Th/+ CCR2^+/+ mice (Fig. 7c, d) and reduced mRNA levels of monocyte-related genes such as TNFα and iNOS (Fig. 7e–g). Most importantly, both subpial cortical demyelination and perivascular cortical demyelination were significantly reduced in Th/+ CCR2^−/− compared to Th/+ CCR2^+/+ mice (Fig. 7h, i).

The detection of CD14⁺ CCR2⁺ double-positive monocytes invading the cortex with ongoing, active demyelination (Figs. 1y, 2l) let us hypothesize that targeting CCR2 might represent a valid treatment strategy against cortical demyelination. We, therefore, developed a novel monoclonal mouse anti-human CCR2 antibody (DOC-2) cross-reacting with marmoset CCR2 (Fig. 7j–l). After humanization, DOC-2 antibodies (now named DOC-2 Fr) retained their capability to bind to human and marmoset monocytes (Fig. 7k, l), were well tolerated, and efficiently depleted monocytes in healthy marmosets at a dose of 5 mg/kg given weekly i.v. (Fig. 7m). Next, we specifically depleted CCR2⁺ monocytes (but not T and B cells) in marmosets with EAE using twice weekly i.v. injections of DOC-2 Fr-2 starting 2 weeks after immunization (Fig. 8a–e). Monocyte-depleted animals displayed less subpial type 3 cortical demyelination and significantly less perivascular type 2 cortical demyelination compared to isotype-treated controls (Fig. 8f–h), in line with an ameliorated clinical disease course (Fig. 8i). In summary, these observations provide evidence that inflammatory monocytes are crucial for the initiation of cortical demyelination and thus represent a specific and rational therapeutic target for the reduction of inflammatory cortical pathology.