In this observational case-control study, we included CSF cell samples from 14 symptomatic control subjects, 21 treatment-naïve patients with RRMS, 12 patients treated with natalizumab, and blood cell samples from 21 healthy controls, 21 symptomatic controls, 21 treatment-naïve patients with RRMS, and 13 patients treated with natalizumab. There was no significant difference in sex or age distribution between the groups. Furthermore, cell-free CSF samples from 41 symptomatic controls, 43 treatment-naïve patients with RRMS and 42 natalizumab-treated patients were included. There was no significant difference in sex or age distribution between the groups except a higher age of natalizumab-treated patients compared to untreated patients. Symptomatic control subjects were defined as patients with symptoms warranting a lumbar puncture but with no indication of inflammation or neurological disorder after diagnostic work-up. Symptoms of the symptomatic controls included sensory symptoms, pain, tension headache, migraine, sleep symptoms (hypersomnia), or idiopathic Bell’s palsy (14). All patients were diagnosed with RRMS fulfilled the 2017 McDonald criteria (15) and none had received methylprednisolone treatment within one month prior to sampling. All treatment-naïve patients with RRMS were diagnosed at the time of lumbar puncture and had a disease duration of 10 months (3; 25); median (IQR). Patients treated with natalizumab had been treated for 38 weeks (12; 97) at sampling time, and time since last treatment was 27 days (19; 44), median (IQR). Of the 42 natalizumab-treated patients, 4 switched to another treatment at sampling time due to disease activity. Healthy controls had no neurological, autoimmune, or chronic illness.

All participants gave informed, written consent to participation. The study was approved by the regional scientific ethics committee (protocol number H-17005703 and H-16047666).

Venous blood was collected, and peripheral blood mononuclear cells (PBMC) isolated by density gradient centrifugation using Lymphoprep (Axis-Shield, Oslo, Norway), washed twice in cold PBS/2 mM EDTA and applied directly to ex vivo flow cytometry or in vitro assays. In parallel with blood sampling, 10 ml of CSF was collected in a polypropylene tube on ice and immediately centrifuged for 10 min at 400 g to separate cells from fluid. Cell-free CSF was kept at -80°C until use and CSF cells applied to flow cytometry within one hour of sampling. PBMC were manually counted using a Neubauer chamber, and CSF cell counts were measured during routine assessment with counts below 3 cells/µl reported as 2 cells/µl. CSF cell count of natalizumab-treated patients below 3 cells/µl was reported as 1 cell/µl according to literature (16).

For ex vivo flow cytometry, freshly isolated PBMC and CSF cells were incubated in FcR-blocking reagent (Miltenyi Biotec, Bergisch Gladbach, Germany) to prevent nonspecific Ab binding, and thereafter stained in PBS/2% FBS/0.02% NaN₃ with a combination of fluorochrome-conjugated antibodies against (conjugate; clone): CD14 (BV605; M5E2), CD16 (PerCP; 3G8), CCR2 (BV421; K036C2), CD40 (APC-Cy7; 5C3), CD49d (APC; 9F10), CD86 (PE-Cy7; 2D10), CD206 (PE; 15-2) all from BioLegend (CA, USA). DICAM (unconjugated; 2H2G12A) from MBL International Corporation (USA, IL) was conjugated using the fast AF488-conjugation kit from Abcam (Cambridge, UK), according to manufacturer. Isotype matched controls were used to correct for nonspecific Ab binding and spectral overlap, where appropriate. For in vitro flow cytometry a live-dead stain (Thermo Fisher Scientific, MA, USA) was included in the staining procedure. TruCount staining of whole blood to measure absolute cell count was performed using BD Multitest 6-color TBNK Reagent including a CD14 antibody (BV605; M5E2), according to manufacturer (BD Biosciences, CA, USA). Data were acquired on a FACSymphony or FACS Canto II flow cytometer (BD Biosciences) and data analysis performed using the software FlowJo (TreeStar, OR, USA).

PBMC were purified from healthy donors and monocytes negatively selected using the human monocyte enrichment kit without CD16 depletion from StemCell Technologies (Vancouver, Canada), according to manufacturer. The purity of monocytes was measured by flow cytometry of cells stained with anti-CD14 antibodies (APC-Cy7; M5E2) and live-dead stain (Thermo Fisher Scientific). Hereafter, cells were resuspended in monocyte attachment media (Sigma-Aldrich; MO, USA) at a concentration of 1×10⁶ monocytes/ml and 200.000 monocytes plated in each well of a 96 wells flat bottom plate. After 1 h at 37°C, non-adherent cells were washed off, and medium changed to 100 ng/ml M-CSF (R&D Systems; MN, USA)/macrophage base medium DXF (Sigma-Aldrich), and cells incubated at 37°C, 5% CO₂ for 5 days. The growth medium was then replaced by fresh medium and cells incubated an additional 24 h. Hereafter, cell-free CSF was thawed and centrifuged at 3000g, 10 min to pellet any cell debris. The growth medium was removed from the cells and replaced by fresh growth medium with/without CSF from one of the 42 donors at a concentration of 37.5%. CSF was used from 14 symptomatic controls, 14 treatment-naïve patients with RRMS and 14 patients treated with natalizumab. After 48 h the cells were detached using accutase (Sigma-Aldrich) and analyzed by flow cytometry to measure expression of DICAM. For this, cells were stained with FcR-blocking reagent, DICAM-AF488 and live-dead stain as described in the flow cytometry analysis section above.

Monocytes were purified as described in the in vitro macrophage differentiation section and plated at a concentration of 1×10⁶ monocytes/ml in RPMI with 1–1000 ng/ml LPS (InvivoGen, CA, USA). After 14 h at 37°C, 5% CO₂ the supernatant was applied to ELISA-measurement of sDICAM (human MXRA8-ELISA; MyBioSource, CA, USA), according to manufacturer.

Monocytes were purified from PBMC using the human monocyte enrichment kit without CD16 depletion (StemCell Technologies), and stained with antibodies targeting DICAM (AF488; 2H2G12A) and CD14 (APC-Cy7; M5E2) in PBS/2 mM EDTA/2% FBS and 9000 CD14⁺DICAM^- and 9000 CD14⁺DICAM⁺ cells sorted on a BD Melody FACS at 4°C. Cells were hereafter cultured with 10 ng/ml LPS (InvivoGen) for 6h at 37°C, 5% CO₂ and the supernatant applied to ELISA-measurement of TNFα (DuoSet from R&D Systems). The experiment was repeated using freshly drawn CSF-cells that was likewise incubated with DICAM and CD14 antibodies and 10 ng/ml LPS at 37°C, 5% CO₂.

Biomarker analysis was performed on cell-free CSF samples from 41 symptomatic controls, 43 untreated patients with RRMS and 42 natalizumab-treated patients. For measurement of IL-8, IL-10, IL-12, IL-15, TNFα, and IFNγ electrochemiluminescence assays (Vplex kits) from Meso Scale Diagnostics (MD, USA) was used according to manufacturer. The maximum lower level of detection across plates measured for each analyte was used as a cut-off for detection. The limit of quantification was defined by the lowest standard with a CV below 25%, samples below this limit was assigned a value of 0. Samples were measured in duplicates and excluded if CV exceeded 20%, or 25% if the value was within the two lowest accepted standard. ELISA was used to measure CSF concentrations of sDICAM (MXRA8-ELISA from MyBioSource), NfL (Uman Diagnostics, Umeå, Sweden), and GFAP (SIMOA technology; Quanterix, MA, USA), according to manufacturer. NfL and GFAP were measured in duplicates.

For analysis of sex differences between groups a Chi-square test was performed, and for analysis of age differences between groups a Kruskal-Wallis test was applied. To compare cell frequencies or numbers between healthy controls, symptomatic controls, untreated and natalizumab-treated patients with RRMS, a Kruskal-Wallis and post-hoc Dunn’s test corrected for multiple comparisons was used. For comparison of DICAM⁺ and DICAM^- cell populations a Wilcoxon test was used. For comparison of sDICAM secretion in response to LPS stimulation, paired t-tests between stimulated and unstimulated samples were applied. Correlations were assessed by Spearman rank correlation analysis and corrected for multiple comparisons by FDR. Statistical values < 0.05 was considered statistically significant. Statistical analysis was performed using GraphPad Prism 10.1.

To investigate a possible role of DICAM in neuroinflammation, we collected CSF samples from 14 symptomatic controls, 21 treatment-naïve patients with RRMS, and 12 patients treated with natalizumab, and used flow cytometry to determine the frequency of DICAM⁺ leukocytes ( Figures 1A–C ). This showed that 11% (7.3; 21) of CSF leukocytes expressed DICAM in controls, 4.6% (2.9; 6.2) in untreated patients, and 25% (17; 30) in patients treated with natalizumab; median (IQR). Compared to both symptomatic controls (p=0.002) and natalizumab-treated patients (p<0.0001), untreated patients had a significantly lower frequency of DICAM⁺ cells ( Figure 1D ). Furthermore, the majority of DICAM⁺ leukocytes in both controls (87%) and natalizumab-treated patients (93%) were CD14⁺ mononuclear phagocytes ( Figure 1D ). Of the few DICAM⁺ cells in CSF of treatment-naïve patients, 55% were CD14⁺ mononuclear phagocytes ( Figure 1D ). CD14⁺ mononuclear phagocytes are hereafter referred to as mononuclear phagocytes.

Further analyzing mononuclear phagocytes in the CSF showed that the frequency of mononuclear phagocytes among leukocytes was greatly reduced in untreated patients compared to symptomatic controls (p=0.001) and natalizumab-treated patients (p<0.0001); Figure 2A . Most untreated patients had CSF pleocytosis ( Figure 2B ) and we observed that the absolute number of mononuclear phagocytes in untreated patients was similar to both symptomatic controls and natalizumab-treated patients ( Figure 2C ).

Also, the frequency of mononuclear phagocytes expressing DICAM was lower in untreated patients compared to both symptomatic controls (p=0.016) and natalizumab-treated patients (p=0.016); Figure 2D . Despite an equal number of intrathecal mononuclear phagocytes in the three groups, the absolute number of DICAM⁺ mononuclear phagocytes in untreated patients were lower compared to both symptomatic controls (p=0.031) and natalizumab-treated patients (p=0.028); Figure 2E . This difference was further emphasized by the observation that DICAM⁺ mononuclear phagocytes constituted 24% of all leukocytes in CSF of natalizumab-treated patients and 10% in controls, compared to only 1.5% in untreated patients ( Figure 2F ).

To understand the role of intrathecal DICAM⁺ mononuclear phagocytes in neuroinflammation, we used flow cytometry to characterize the surface expression of various proteins associated with the function of mononuclear phagocytes in the CSF of untreated patients with RRMS. For this, the following gating strategi was applied: the CSF leukocyte population was defined ( Figure 3A ), doublet cells excluded ( Figure 3B ), and CD14⁺DICAM^- and CD14⁺DICAM⁺ mononuclear phagocytes defined ( Figure 3C ). Expression of the surface markers CD16, CCR2, CD40, CD86, and CD206 where hereafter defined ( Figure 3D ) and the gates applied to both the DICAM⁺ and DICAM^- mononuclear phagocyte population. This analysis showed that DICAM⁺ mononuclear phagocytes in CSF predominantly were CD16⁺ and CCR2^- in contrast to DICAM^- mononuclear phagocytes (p<0.0001); Figures 3E, F . Additionally, DICAM⁺ mononuclear phagocytes had an increased expression of CD40 (p<0.0001), CD86 (p=0.0001), and CD206 (p<0.0001) compared to DICAM^- mononuclear phagocytes, Figures 3G–I . Also, like DICAM^- mononuclear phagocytes, 100% of DICAM⁺ mononuclear phagocytes expressed CD49d, which is one of the two subunits of the adhesion molecule Very Late Activation Antigen-4 (VLA-4; data not shown). The same was found in symptomatic controls and natalizumab-treated patients ( Supplementary Figure 1 ).

Comparing the phenotype of DICAM⁺ mononuclear phagocytes of untreated patients with symptomatic controls and natalizumab-treated patients showed a similar frequency of intrathecal DICAM⁺ mononuclear phagocytes expressing CD16, CD40, CD86 and CD49d between the three groups ( Figure 3J , and data not shown). In contrast, the frequency of DICAM⁺ mononuclear phagocytes expressing CCR2 in untreated patients compared to controls and natalizumab-treated patients was increased (p=0.004 and 0.006) and the expression of CD206 decreased (p=0.005 and 0.046) ( Figures 3K, L ).

In the blood, DICAM is expressed on less than 2% of circulating CD14⁺ monocytes, with no difference between healthy controls (n=21), symptomatic controls (n=21), untreated (n=21) or natalizumab-treated (n=13) patients with RRMS ( Figures 4D, E ). Subdividing circulating DICAM⁺ monocytes into classical (CD16^-), intermediate (CD16⁺CCR2⁺) and non-classical (CD16⁺CCR2^-) monocytes ( Figures 4A–C ) showed a higher frequency of DICAM⁺ intermediate monocytes in natalizumab-treated patients compared to both symptomatic controls (p=0.029) and untreated patients (p=0.035), Figure 4F .

To investigate if the enrichment of DICAM⁺ mononuclear phagocytes observed in the CSF was due to selective recruitment of DICAM⁺ monocytes from the blood, differentiation of recruited monocytes to CNS border-patrolling macrophages or an effect of the local intrathecal microenvironment, monocytes from healthy controls were purified and cultured under macrophage differentiating conditions for 6 days. Thereafter, macrophages were grown in medium or CSF from either symptomatic control subjects (n=14), untreated patients (n=14), or natalizumab-treated patients (n=14) for an additional 2 days and DICAM expression was measured ( Figures 4G, H ). Following macrophage differentiation, we observed a strong increase in DICAM⁺ cells to a mean of 26% (dotted line in Figure 4I ). This increase was marginally elevated when cells were cultured in CSF; however, not significantly and all CSF-samples affected the expression of DICAM equally ( Figure 4I ). The experiment was repeated, and the same observations obtained (data not shown).

In addition to a membrane bound form, DICAM exists in a soluble form (sDICAM). To investigate if sDICAM is present in CSF and whether there is a difference between patients and symptomatic controls, we measured sDICAM in CSF from 41 symptomatic controls, 43 treatment-naïve and 42 natalizumab-treated patients by ELISA. sDICAM could be detected in all CSF samples analyzed. The analysis showed higher sDICAM concentrations in natalizumab-treated patients compared to both untreated patients (p=0.0008) and symptomatic controls (p=0.015), Figure 5A . The variation in duration of natalizumab treatment and time since last treatment between treated patients may influence the level of sDICAM in CSF. To analyze this, we performed a correlation analysis between these parameters and the concentration of sDICAM and found no correlation (p=0.084, r_s=0.27 and p=0.124, r_s=-0.24; respectively).

Besides mononuclear phagocytes, a small population of T cells (6), endothelial cells (9), and astrocytes (12) have been shown to express DICAM. The measured sDICAM in CSF is therefore likely of mixed origin; however, we found a positive correlation between sDICAM and the frequency of DICAM⁺ mononuclear phagocytes in CSF of patients with RRMS (p=0.011, r_s=0.40; Figure 5B ). Also, we found that mononuclear phagocytes stimulated in vitro with LPS for 14 h secreted sDICAM ( Figure 5C ); altogether suggesting that these cells are major contributors to CSF sDICAM.

To investigate a clinical implication of DICAM in MS, we analyzed a possible association between sDICAM in CSF of 43 untreated patients with RRMS and the biomarker of neuroaxonal damage neurofilament light chain (NfL) (17) and the astrocyte activity biomarker glial fibrillary acidic protein (GFAP) (18). This showed a weak association between sDICAM and an increased GFAP (q=0.04; r_s=0.35); Figure 5D . Further analyzing an association between sDICAM in CSF and various cytokines showed a negative correlation between sDICAM and TNFα (q=0.028; r_s=-0.39), IFNγ (q=0.0002; r_s=-0.60), IL-12 (q=0.0008; r_s=-0.55), and IL-10 (q=0.001; r_s=0.54); Figure 5D . To improve a link between these observations and the presence of DICAM⁺ mononuclear phagocytes in CSF, we next analyzed correlations between the frequency of DICAM⁺ mononuclear phagocytes in untreated patients with RRMS and the biomarkers that correlated with sDICAM. Despite a considerably lower sample size in this analysis (n=18), we found a negative correlation between DICAM⁺ mononuclear phagocytes and TNFα (q=0.009; r_s=-0.66); Figure 5E . All statistical values shown are FDR-corrected p-values (q-values).

Studies have indicated a role of DICAM in suppressing intracellular signaling pathways (9) and lipopolysaccharide (LPS)-induced production of proinflammatory cytokines including TNFα (10). Based on our data suggesting a relation between DICAM and TNFα in CSF, we investigated DICAM’s direct ability to influence TNFα-production in mononuclear phagocytes. To study this functional property of DICAM, we sorted monocytes into CD14⁺DICAM^- and CD14⁺DICAM⁺ cells using antibodies targeting CD14 and DICAM, at a low temperature to ensure that no signals were induced through the DICAM-receptor by the anti-DICAM antibody during sorting. The two sorted populations were hereafter cultured in 10 ng/ml LPS for 6 h at 37°C allowing anti-DICAM induced signaling through surface expressed DICAM. The culture supernatants were collected, and TNFα-concentration measured. This showed a prompt LPS-induced production of TNFα without triggering of DICAM ( Figure 6A , LPS); a production that was strongly reduced by anti-DICAM antibody triggering of DICAM ( Figure 6A ; LPS+αDICAM, p=0.001). The experiment was repeated three times with a total of 7 donors. To validate this finding, the same cocktail of antibodies and LPS was added to freshly drawn CSF-cells from a patient with RRMS, and TNFα production was measured. Likewise, this showed that a signal through DICAM diminished TNFα production ( Figure 6B ). As a previous study described that DICAM accelerated apoptosis in endothelial cells (9) we measured cell death after 6h LPS-stimulation in the two sorted populations to ensure that the decrease in TNFα-production observed was not simply a result of cell death. This showed that cells in which DICAM was triggered were more sensitive to apoptosis, but not to a large degree ( Figures 6C, D ). Altogether, these data indicate that a functional property of DICAM is to attenuate LPS-induced TNFα production in mononuclear phagocytes.