Immune cell trafficking from the brain maintains CNS immune tolerance (original) (raw)
CD11c+ cells are associated with the rostral migratory stream (RMS). APCs are present in the choroid plexus and the meninges (22); however, their localization to the brain parenchyma in health is disputed (23, 24). Our examination of the normal mouse brain with dual immunolabeling demonstrated an association of CD11c+ cells with the RMS, a pathway along which neuroblasts travel from the lateral ventricle to the olfactory bulb (OB) to replace olfactory interneurons (Figure 1A). CD11c+ cells were found throughout the RMS (Figure 1B), including its terminal field, the granule cell layer of the OB. Additionally, these cells were found in the glomerular layer, adjacent to the cribriform plate. The distribution of CD11c+ cells along the RMS and extending into the OB suggested that this might be an important area for CNS immunoregulation. We thus examined the RMS and surrounding white and gray matter structures for the presence of CD11c+ cells. In the brains of normal, specific pathogen–free mice, CD11c+ cells were preferentially associated with the RMS at significantly higher densities compared with surrounding white or gray matter structures (Figure 1C). To further characterize these putative DCs, we isolated CD11c+ cells from 10 normal mice and examined them by fluorescence-activated cell sorting (FACS). They were predominantly CD11c+CD11b+ DCs, along with typical plasmacytoid DCs (pDCs) and a population of other DC phenotypes (Figure 1D and Supplemental Figure 2; supplemental material available online with this article; doi:10.1172/JCI71544DS1). To address the possibility that the specialized brain macrophage–like cells known as microglia might express CD11c, we performed further confocal examination of brain sections to define the anatomical location of the CD11c+ cells and confirm that they were in fact DCs. This showed the presence of CD11c+CD11b– cells in the choroid plexus and in association with the RMS (Figure 1, E and F); the lack of CD11b on these cells suggested they were true DCs. These cells and a subgroup of CD11c+CD11b+ cells had a similar appearance, distinctly different from that of resident CD11b+CD11c– microglia (Figure 1F). Finally, we found that a substantial portion of these RMS DCs were truly located in the parenchyma (Figure 2).
Immune cells are associated with the RMS. (A) Sagittal sections were labeled for CD11c by IHC, followed by IF staining for DCX (green), showing the course of the RMS from the subventricular zone to the OB. AOV, ventral anterior olfactory nucleus; ACB, nucleus accumbens; CP, caudate putamen. Box indicates the region shown at higher magnification in B. (B) The RMS, showing CD11c cells (red arrowheads) in close association with migrating DCX+ neuroblasts. (C) CD11c+ cells in the RMS were counted in at least 5 sections. The corpus callosum (CC) and frontal cortex (FCx) in the same sections were similarly examined. Data represent mean ± SEM. *P < 0.001 vs. RMS. (D) Phenotypic analysis of CD11c+ cells isolated from 10 mouse forebrains (for gating strategy, see Supplemental Figure 2). CD11c+CD11b+ cells represented 70% of DCs, pDCs represented 10%, and the remainder belonged to other CD11c+CD11b– DC subsets (3 independent experiments). (E) IF confocal images for DCX (green), CD11b (blue), and CD11c (red). CD11c+ (red arrowhead) and CD11b+ (blue arrowhead) cells are identified in the choroid plexus adjacent to the DCX+ subventricular zone in normal mice. (F) CD11c+CD11b+ DCs (blue arrow), CD11c+CD11b– DCs (red arrow), and CD11b+ microglia (white arrows) in close association with the RMS (green; DCX). Scale bars: 250 μm (A); 25 μm (B); 10 μm (E and F).
The majority of RMS CD11c+ cells are not associated with cerebral vasculature. To show that CD11c+ cells in the RMS were intraparenchymal, we performed serial optical sectioning (1-μm steps) through the RMS using confocal microscopy (n = 5). Sections were stained with DCX (RMS; green), CD11c (DCs; red) and CD31 (vascular endothelium; blue). Shown is a representative series demonstrating a typical perivascular CD11c+ cell (blue arrow) and 2 examples of the several CD11c+ cells in the RMS that were not associated with any vasculature (red arrows). Scale bar: 100 μm.
CD11c+ cells in the RMS are DCs derived from the circulation. To establish the origin of DCs present in the RMS, chimeric mice were generated by transplanting BM from syngeneic C57BL/6 mice expressing GFP under the control of the Cx3cr1 promoter, which is active in monocyte lineage cells (25). 12 weeks after BM transplantation, GFP+CD11b– cells were present along the course of the RMS and its terminal field. The parenchymal distribution of these cells was similar to that of normal, nonmanipulated mice, while the coexpression of GFP provided evidence of their peripheral (rather than CNS) origin. Using multilabel confocal microscopy, we identified subsets of GFP+CD11b+ and GFP+CD11b– cells. These cells were again present in the choroid plexus adjacent to the origin of the RMS and throughout the entire RMS and its terminal field (Figure 3, A and B), which suggests the presence of both conventional (CD11c+CD11b+) and nonconventional (CD11c+CD11b–) DCs in this region. For macrophages, choroid plexus recruitment is mediated by VLA4 for attachment and CD73 for migration through the choroid epithelium (26). Indeed, systemic treatment of mice with the CD73 inhibitor methylene ADP led to significant changes in macrophage numbers, but not DC numbers (Supplemental Figure 1), which suggests that DC migration into the CNS is not modulated by CD73. Nonetheless, the presence of CD11c+CD11b– cells with peripheral, circulatory origins represents strong evidence that true DCs are indeed present in the adult CNS.
Blood-derived DCs are recruited to the RMS. _Cx3cr1+/gfp_-to-WT BM chimeric mice were killed 12 weeks after BMT. Brain sections were labeled for DCX (blue), CD11b (blue or red), and/or CD11c (red); GFP fluorescence is shown in green. (A) Confocal images showing the presence of both GFP+CD11c+ cells (red arrow) and GFP+CD11b+ cells (blue arrow; i.e., derived from donor BM) in the choroid plexus. (B) GFP+CD11b– circulation-derived cells, presumably nonmacrophage/microglia and therefore DCs, were again seen in close association with the RMS (red arrow), alongside GFP–CD11b+ (white arrows) and GFP+CD11b+ cells. Scale bars: 10 μm.
DCs migrate from the CNS to the CxLNs. To determine whether leukocytes recruited to the CNS might migrate to the CxLNs, we infused CFSE, a vital fluorochrome that stably labels leukocytes, allowing their migration to be tracked. After 7 days of continuous intracerebroventricular (icv) infusion of CFSE into a lateral ventricle adjacent to the choroid plexus, CFSE-labeled immune cells were detected in the CxLNs (Figure 4A), but not in other secondary peripheral immune organs, including the inguinal LNs (Figure 4B) and spleen (data not shown). These findings indicate that CFSE-labeled cells accumulated in the CxLNs because of lymphatic-like drainage from the CNS.
Leukocytes migrate from the CNS to CxLNs. In icv-cannulated mice, after 7 days of CFSE infusion, CxLN CD45+ cells were gated and examined for CFSE labeling. (A) CFSE-labeled cells were discerned in all mice. The percentage of CFSE+ cells is indicated. (B) The same experiment repeated with fingolimod or vehicle administered i.p. (n = 6; 3 independent experiments), beginning 1 day prior to icv cannula implantation. For gating of CFSE+ CxLN and ILN monocytes, see Supplemental Figure 4. Top: Vehicle-treated mice (large arrow) had substantial populations of CFSE+ presumed macrophages (CD45+CD3–B220–CD11c–CD11b+), which were significantly reduced in CxLNs from fingolimod-treated mice (small arrow) and absent in ILNs of either vehicle- or drug- treated mice. Bottom: CFSE+ DCs (CD45+CD3–CD11c+) were only seen in CxLNs of vehicle-treated mice (arrow), and their accumulation was completely inhibited by systemic fingolimod treatment. (C) Mice that had CFSE infused by icv cannula had their brains removed, and mononuclear cells were extracted for FACS. Top: DC and macrophage/PMN accumulation in the brains of cannulated mice. Bottom: To determine which cell type(s) may have accumulated as a result of injury associated with icv cannulation, we also administered fingolimod (6 μg/d) or vehicle (n = 6; 3 independent experiments) i.p. to noncannulated mice, which demonstrated an increased frequency of DCs in the CNS as assessed by FACS (P = 0.018), with no differences in corresponding macrophage/PMN or T cell numbers. Data represent mean ± SEM. *P < 0.05.
If CFSE-labeled cells did transit from the brain, pharmacological interruption of this process should lead to their retention in the CNS. We therefore used fingolimod, which is known to alter lymphocyte and DC trafficking (19). We repeated the preceding experiment, but this time also gave mice daily i.p. injections of fingolimod or vehicle, starting prior to icv cannulation and continuing through the entire period of CFSE infusion. In vehicle-treated mice, CFSE-labeled leukocytes appeared in the CxLNs of mice sacrificed at the termination of CFSE infusion, and this accumulation was inhibited by systemic fingolimod treatment (Figure 4B). There was no significant accumulation of CD45+ CFSE+ events in the inguinal LNs (ILNs) of either fingolimod- or vehicle-treated mice (Figure 4B). We next characterized the composition of cells that were labeled with CFSE in the CNS and had migrated to the CxLNs. Perhaps not surprisingly, considering the injury induced by cannulation, the majority of cells that migrated to the CxLNs were CD11b+, with forward and side scatter characteristics of macrophages. Consistent with the mediation of macrophage migration by S1P (27), fingolimod almost completely inhibited their transit to the CxLNs (Figure 4B). The next most abundant CFSE-labeled cell type found in the CxLNs was CD3–CD11c+ DCs, whose accumulation was also inhibited by fingolimod (Figure 4B). Virtually no CFSE+CD3+CD4+ T cells were identified in CxLNs of CFSE-infused mice (data not shown). Consistent with this finding, systemic fingolimod treatment led to accumulation of these cell types in the CNS (Figure 4C). The absence of widespread CFSE staining in the CxLNs of mice treated with systemic fingolimod and the confinement of CFSE staining to specific leukocyte subsets also indicated that nonspecific leakage of dye to the CxLNs was not a confounding factor.
Systemic fingolimod treatment leads to DC accumulation in the normal CNS. Having established that fingolimod treatment virtually abolished egress of leukocytes from the brain, we next determined which leukocytes were preferentially retained in healthy brains. These experiments were performed in mice that had not experienced the injury that occurred due to CNS cannulation. Mice received daily systemic (i.p.) treatment of fingolimod or vehicle. Similar to mice undergoing icv cannulation, systemic fingolimod treatment of uncannulated mice led to significantly more DCs accumulating in the CNS (Figure 4C). To exclude the possibility that this was related to increased microglia activation for some unforeseen reason, we further quantified CD3–CD11c+CD11b– DCs, as these could not be classical microglia because of the absence of CD11b. We also found a significant increase in these cells with fingolimod treatment (Figure 4C). These findings indicate that DC entry and exit from the CNS are likely to be regulated by different mechanisms, as their egress (but not accumulation) was at least partly modulated by fingolimod. In the absence of cannulation, fingolimod treatment was not associated with significant differences in the accumulation of the CD45hiCD11c–CD11b+ macrophage/polymorphonuclear cell (macrophage/PMN) subset (Figure 4C).
Systemic fingolimod treatment alters CD11c+ cell distribution along the RMS. Our earlier experiments indicated that the majority of CD11c+ cells, outside the normal CNS support structures, were associated with the RMS (Figure 1). Based on this finding, we reasoned that if egress of these cells was regulated by systemic fingolimod therapy, and if they traveled from the proximal to the distal RMS with migrating neuroblasts, fingolimod treatment should lead to their accumulation in the distal RMS. Furthermore, there should be no significant change in CD11c+ cell numbers in the proximal RMS, as fingolimod therapy did not appear to affect their recruitment. Comparison of fingolimod- and vehicle-treated mice (n = 6 per group) revealed a significant difference in the distribution of CD11c+ cells along the RMS. As predicted, CD11c+ cells were at the same density in proximal parts of the RMS in both groups (Figure 5A). However, the density of CD11c+ cells increased in the more distal parts of the RMS terminal field, peaking at the ventral glomerular layer of the OB, adjacent to the cribriform plate, after fingolimod treatment (Figure 5, A and B). These observations suggest that migration of CD11c+ cells along the RMS was arrested at this point by fingolimod therapy. We next sought to determine whether DCs actually migrate along the RMS.
CD11c+ cells are recruited to the proximal RMS and travel to the OB, and RMS disruptions leads to their accumulation. (A) With i.p. fingolimod treatment (n = 5), the frequency of CD11c+ cells in the proximal RMS (RMSO) remained unchanged. However, the CD11c+ cell density significantly increased in the terminal field of the RMS (P = 0.042) and was further enriched in the ventral part of the OB (VOB; P = 0.014). (B) Representative sections of the ventral OB showing accumulated CD11c+ cells in fingolimod-treated mice compared with vehicle controls. (C) CFSE-labeled BMDCs were given icv (n = 3/group). CFSE+ cells (green) were present in the proximal-mid RMS (DCX, red) at 16 hours (top), and by 24 hours (bottom) had reached the OB. (D) The RMS was ablated with a 7-day icv ARA-c infusion. There were significantly fewer CD11c+ DCs in the proximal region of the ablated RMS (P = 0.04, n = 3/group). (E) However, in similarly treated mice, FACS analysis of CNS mononuclear cells from the forebrain showed significantly increased macrophages and DCs, similar to systemic fingolimod treatment (P = 0.03, n = 6; 2 independent experiments). (F) Representative FACS plots showing CFSE-labeled DCs that had migrated from CNS to CxLNs isolated 48 hours from mice after 3 icv injections of fingolimod-pretreated (100 nM for 24 hours), CFSE-labeled BMDCs (1 × 106 cells/injection) at 12-hour intervals (n = 7/group, 2 independent experiments). (G) Fingolimod treatment of DCs significantly reduced their migration into CxLNs (P = 0.042, n = 7; 2 independent experiments). Scale bars: 100 μm; 10 μm (insets). Data represent mean ± SEM. *P < 0.05.
DCs migrate through the RMS. DCs were generated by culturing murine BM with fms-like tyrosine kinase 3 ligand (FLT3) (28), which yielded a highly enriched CD11c+ cell population (>95% purity) after 9 days in culture. These BM-derived DCs (BMDCs) were then labeled ex vivo with CFSE, after which 106 labeled cells were injected icv. Mice were killed 16 or 24 hours after injection (n = 3 per group). At 16 hours, CFSE+ BMDCs were observed in the proximal and middle RMS, as shown by staining for doublecortin (DCX; Figure 5C and ref. 29). By 24 hours, labeled cells had cleared the proximal and middle RMS and were seen in the terminal field of the RMS within the OB. To confirm that the RMS was unequivocally involved in DC migration from the proximal to the distal RMS, we ablated it with icv infusion of cytosine-β-D-arabinofuranoside (ARA-c) (30). In multiple experiments, this led to retention of DCs and macrophages in the brain (Figure 5E), similar to systemic fingolimod treatment of icv cannulated animals. Furthermore, direct quantification of DCs in the RMS showed that ablation of the subventricular zone and RMS did lead to failure of recruitment of DCs to the RMS (Figure 5D). As CXCL12 normally mediates neuroblast migration through the RMS (31), we treated mice systemically with AMD3100 to inhibit one of its receptors, CXCR4. This treatment resulted in an isolated increase in DCs in the mouse CNS (Supplemental Figure 3), similar to that observed with systemic fingolimod treatment (Figure 4C). Finally, we treated BMDCs ex vivo with fingolimod or vehicle and injected them icv as before, 48 hours after which the CxLNs were isolated. These showed that fingolimod directly acted on adoptively transferred DCs, reducing their egress from the CNS to the CxLNs (Figure 5, F and G). These data confirmed that DCs migrate through the RMS, possibly using chemotactic signals mediated by CXCR4. We next assessed the effect of perturbing this cellular pathway on CNS autoimmune inflammatory disease.
Fingolimod infusion to the RMS increases actively induced EAE severity. Systemic delivery of fingolimod inhibits EAE disease (32). Conversely, we discovered that targeting delivery of fingolimod to the RMS (Figure 6A) led to a dose-related increase in the severity of actively induced EAE with myelin oligodendrocyte glycoprotein35–55 (MOG35–55) peptide (Figure 6B). Infusions of the same dose of fingolimod — either icv into the lateral ventricle, at a CNS parenchymal site distant from the RMS (hippocampus), or systemically — led to relative reductions in clinical EAE (Figure 6, C and D). Together, these data suggested that modulation of immune cell traffic through the RMS leads to altered CNS immunoregulation. As fingolimod treatment of the RMS led to deviation of regulatory immune responses and enhanced anti-CNS immunity, we next investigated whether targeted delivery of this drug could induce spontaneous EAE.
Targeted RMS-fingolimod treatment increases MOG-induced EAE severity. (A) RMS-fingolimod treatment. Cannulae were implanted near the origin of the RMS (top), in the lateral ventricle (LV; bottom), or the hippocampal formation (not shown). Fingolimod was delivered over a 4-week period via an osmotic minipump. Sept, septum. (B) RMS-fingolimod treatment beginning 1 week prior to EAE induction led to a dose-dependent increase in disease severity (n = 3/group), with mice receiving 300 ng/d all dying by day 15. Animals receiving 30 ng/d were less severely affected, but reliably more so than vehicle-infused controls (P < 0.001). (C) Mice that received icv fingolimod (300 ng/d; n = 3) developed significantly less severe EAE than vehicle controls (P = 0.004). (D) Infusion of fingolimod to the hippocampus or systemically also resulted in milder clinical disease compared with RMS-fingolimod treatment (n = 3 per group; P < 0.001). *P < 0.05.
Fingolimod infusion to the RMS increases spontaneous EAE incidence. While the majority of Th cells of Vβ11 T cell receptor transgenic 2D2 C57BL/6 mice respond to MOG, they display a low rate of spontaneous CNS-wide disease. Whereas up to 30% of 2D2 mice develop optic neuritis over a 3-month period, only about 6% go on to develop spontaneous classical (rather than opticospinal) EAE (33, 34). Breaking immune tolerance in 2D2 mice (with pertussis toxin treatment alone) significantly enhances EAE incidence to about 60%, whereas MOG vaccination (without pertussis toxin treatment) does not induce EAE (34). As one of pertussis toxin’s principal actions is to modify immune cell migration (35), we hypothesized that immune cells associated with the RMS may restrain systemic anti-CNS immune responses in 2D2 mice. To test this contention, we administered fingolimod or vehicle directly targeting the RMS (referred to herein as RMS-fingolimod and RMS-vehicle treatment, respectively). Separate cohorts received fingolimod or vehicle via RMS and icv infusion, or directly to the CxLNs via s.c. lymphatics (36), the latter bypassing the CNS. RMS-fingolimod led to a 63% incidence of classical EAE (10 of 16 mice) compared with RMS-vehicle control mice (0 of 9; P < 0.001). Furthermore, there was no case of EAE in mice that had fingolimod or vehicle infused to the lateral ventricle (n = 4) or to the CxLNs (n = 11). Among the 10 mice that developed symptoms of EAE, clinical scores ranged from 1 to 5 (mean ± SEM, 2.0 ± 0.4; Figure 7A). These data confirmed the unique capacity of RMS-fingolimod delivery to induce EAE in this otherwise-resistant mouse line.
RMS-fingolimod treatment induces EAE by disrupting CxLN Treg function. (A) EAE incidence in 2D2 mice increased from 0% with RMS-vehicle treatment (n = 9) to 63% (10 of 16; 4 independent experiments) with RMS-fingolimod treatment. (B) These mice also had more antigen-specific T cells in spinal cords (1.7 ± 0.1 × 105 vs. 0.4 ± 0.2 × 105; P = 0.03; 3 independent experiments), and (C) the Treg/Teff ratio in the CxLNs was negatively related to antigen-specific T cells in spinal cord (SC; n = 6; P = 0.002; 3 independent experiments; see also Supplemental Figures 6 and 7). (D and E) They also had no differences in (D) the number of antigen-specific Tregs between CxLNs and ILNs (CxLN, P = 0.9; ILN, P = 0.5) or in (E) the proportion of antigen-specific Teffs as a percentage of total T cells (CxLN, P = 0.6; ILN, P = 0.3). (F) RMS-fingolimod or direct CxLN fingolimod treatment did not affect the proportion of Teffs (P = 0.6; n = 6; 2 independent experiments). (G) Separate RMS-fingolimod– or RMS-vehicle–treated 2D2 mice (n = 9; 3 independent experiments) had their Treg activity assessed in their CxLNs (see Methods and Supplemental Figures 8 and 9). There was a significant increase in antigen-specific T cell activation when CD25hi Tregs were removed from the CxLN mononuclear population (P = 0.023) of RMS-vehicle–treated, but not RMS-fingolimod–treated, CxLN isolates, in contrast to ILNs (n = 18; 3 independent experiments), indicative of reduced Treg activity in the CxLNs of RMS-fingolimod–treated mice. Data represent mean ± SEM. *P < 0.05.
We also quantified CNS-specific Vβ11+ T cells in the spinal cord of 2D2 mice that received RMS-targeted treatment; this area is affected in classical EAE, but distant from the infusion site. FACS analysis of spinal cord mononuclear cells revealed that the absolute number and proportion of Vβ11+ T cells was significantly higher in fingolimod- versus vehicle-infused mice (Figure 7B and Supplemental Figure 5B), consistent with increased neuroinflammation.
Treg function is compromised in the CxLNs of RMS-fingolimod–treated mice. CxLNs were harvested from untreated 2D2 mice or those subjected to RMS-fingolimod or RMS-vehicle treatment. To identify Tregs, we stained CxLN cells from untreated mice with CD3, CD4, CD25, CD127, and FoxP3. This showed that most, if not all, CD25+CD127lo cells expressed FoxP3 (Supplemental Figure 6), identifying them as bone fide inducible Tregs. We then determined the proportion of antigen-specific T effector cells (Teffs) and Tregs in the CxLNs of treated 2D2 mice. We stained cells with anti-CD45, -CD3, -CD4, -CD44, -CD62L, -CD25, and -CD127 in addition to a Vβ11 antibody to identify the transgenic MOG-specific T cells (Supplemental Figure 7). While the proportion of CxLN CD25hiCD127lo Tregs was not significantly different between mice treated with vehicle or fingolimod (Figure 7D), the Vβ11+ Treg/Teff ratio in the CxLNs was significantly and inversely related to the proportion of CNS-specific cells in the spinal cord (P = 0.002; Figure 7C and Supplemental Figure 7), the latter being a quantitative measure of EAE disease severity. As antigen-specific T cells are activated in the CxLNs prior to migrating to the CNS (14), and Tregs modulate this activation (37), we hypothesized that RMS-fingolimod treatment causes a deficit in Treg function in the CxLNs, thus enhancing anti-CNS immune responses.
RMS-fingolimod administration breaks CNS immune tolerance. To determine whether CxLN Treg function was compromised after RMS-fingolimod infusion of treated 2D2 mice, we directly compared the effect of Treg depletion on in vitro antigen-specific CxLN T cell activation. Treg depletion was accomplished by either leaving or removing CD25hi cells from the mononuclear cell preparation derived from CxLNs (Supplemental Figure 8). This approach was designed to measure the functional capacity of DC-inducible CD25hiCD127lo Tregs (reviewed in ref. 38). While we recognize that this approach also removes CD25+ Teffs, the removal of these cells biases against increased T cell activation. Nonetheless, to exclude any unequal bias, we quantified the antigen-specific Teff frequency in the CxLNs after RMS-fingolimod or RMS-vehicle treatment, as well as after similar treatments administered directly to the CxLNs, and found no significant differences (Figure 7, E and F). RMS-fingolimod treatment resulted in significantly less restraint of Teff cell activation compared with RMS-vehicle, as indicated by a smaller increase in antigen-specific activation with and without CxLN Tregs (P = 0.02; Figure 7G). Given that the CxLNs of the RMS-fingolimod and RMS-vehicle groups showed equal numbers of Tregs (Figure 7D), these data suggest that CxLN Tregs after RMS-fingolimod treatment are less efficient at suppressing T cell activation. In direct contrast, there were no significant differences in the activity of ILN Tregs, regardless of treatment (Figure 7G). As DCs rather than T cells were prevented from traveling to the CxLNs from the CNS, we postulated that DCs mediated the changes in Treg function in the CxLNs.
DCs modify anti-CNS immunity in the CxLNs via regulation of Treg activity. To determine whether impaired DC traffic from the brain, as a result of targeted fingolimod treatment, mediates the altered Treg activity in the CxLNs, we used a model of delayed-type hypersensitivity (DTH). This model is capable of determining whether DCs modify antigen-specific Teff responses in the CxLNs in vivo (39), and whether CxLN Treg activity is similarly modified. In this case, we examined whether RMS-fingolimod treatment altered the capacity of CxLN DCs to modulate Th cell responses to MOG peptide. WT C57BL/6 mice were subjected to a 4-week RMS-fingolimod or RMS-vehicle infusion, after which CxLNs were harvested, mononuclear cells were isolated and surface labeled, and CD45+CD3–CD11+ DCs were purified by FACS to greater than 99% purity (Supplemental Figure 10). The DCs from a single mouse (4–8 × 104 cells) were injected into a pinna of a separate C57BL/6 mouse, which had received a single s.c. vaccination with MOG peptide in complete Freund’s adjuvant 7 days earlier. As a control, the other ear of the same mouse was injected with the acellular vehicle. 10 days later, both ears were challenged with MOG peptide by directly injecting it into the subcutis. To assess T cell–mediated antigen recall responses, ear swelling was measured 24 hours after challenge (39), with increasing ear swelling indicating a greater response. The swelling in the control ear that received the acellular vehicle injection was considered to represent the immune recall response, wholly attributable to the MOG vaccination. The swelling of the ear injected with DCs was directly compared with the control ear of the same mouse and expressed as a percentage of the control ear swelling. In replicate experiments, DCs isolated from the CxLNs of mice with RMS-fingolimod treatment mediated a significant increase in recall responses to the CNS antigen MOG, confirming enhanced anti-CNS immune responses in the CxLNs, whereas RMS-vehicle controls showed no such change (Figure 8A). DCs isolated from the CxLNs of mice receiving icv vehicle or fingolimod showed no capacity to regulate the Teff response to MOG in the CxLNs, as indicated by similar degrees of swelling in acellular vehicle– and DC-injected ears (Figure 8A). These data suggested that impaired egress of DCs from the CNS as a result of RMS-fingolimod treatment was responsible for increased CNS antigen recall responses in the CxLNs. However, a direct effect of fingolimod on CxLN DCs cannot be excluded.
CNS-derived DCs modulate anti-CNS immunity in the CxLNs. (A) CxLN DCs were purified from C57BL/6 mice treated with RMS-fingolimod or -vehicle, or icv fingolimod or vehicle, for 4 weeks. These or acellular vehicle were injected into each pinna of separate C57BL/6 mice prior to MOG DTH assessment (see Methods). Only DCs from fingolimod-treated mice increased MOG DTH responses (RMS-fingolimod, n = 12/group; icv fingolimod, n = 6/group; P = 0.003; 3 independent experiments). (B) Separate C57BL/6 mice were treated with fingolimod or vehicle i.p. for 14 days (n = 10; 3 independent experiments), and CNS DCs were FACS purified and tested in the DTH model, which showed they similarly modified DTH and were not altered by direct fingolimod exposure. (C) 10 mice (5 vehicle; 5 fingolimod) were randomly selected from A, and CxLNs draining the vehicle control– or DC-treated ears were isolated separately for Treg functional assessment. Only the CxLNs draining ears that had RMS-fingolimod DCs delivered had significantly reduced Treg activity (n = 5; P = 0.01). (D) CNS CD45hiCD3–CD11c+CD11b+ and CD45hiCD3–CD11c+CD11b– cells were isolated from 10 normal mice, after which each DC subset (2 × 102 cells) was assessed in the DTH model in separate mice. Only CD11b– CNS DCs significantly reduced DTH to MOG (n = 10; P = 0.011; 3 independent experiments). (E) We repeated the experiment in D, except MOG was replaced by methylated BSA (n = 3/group). There was no significant modulation of methylated BSA response (P > 0.9), which — in contrast to MOG DTH DC–mediated modulation — suggests a CNS-antigen specific mechanism. Data represent mean ± SEM. *P < 0.05.
Fingolimod does not alter the inflammatory activity of CNS DCs. To test whether direct effects of fingolimod on DC inflammatory, rather than migratory, capacity were responsible for our results, we took advantage of the pharmacokinetic properties of fingolimod, which leads to preferential CNS concentration after systemic therapy (32). This ensured that any DCs isolated from the CNS would be exposed to fingolimod in vivo. In this case, DCs from the CxLNs could not be used, as we had shown that systemic fingolimod treatment was likely to alter traffic to the CxLNs and thus bias our results. We treated groups of 10 mice with vehicle or fingolimod (6 μg/d i.p.) for 7 days. CNS DCs were isolated, as above, from the forebrain with the meninges removed (Supplemental Figures 2 and 10). We then compared their capacity to modulate CNS antigen recall responses using the DTH model. There were no significant differences in the recall responses when DCs were isolated from vehicle- or fingolimod-treated mice and subsequently delivered to the CxLNs (Figure 8B), which suggests that the direct effects of fingolimod on DCs did not alter their capacity to modulate inflammation. Having discounted a direct effect of fingolimod on DCs, and considering our observation of altered Treg function in the CxLNs as a result of RMS-fingolimod treatment, we next examined whether adoptively transferred DCs directly modulate Treg function in the CxLNs.
DCs directly modulate Treg function in the CxLNs. To ascertain that transferred DCs modulate anti-CNS immunity by regulating CxLN Treg activity, we examined mice that had CxLN DCs transferred to the ear. After MOG challenge (Figure 8A), the DC- and vehicle-injected ears had their draining CxLNs isolated separately, after which Treg activity was determined, as we had done before (Figure 7G). Treg activity was significantly lower in CxLNs that received DCs from RMS-fingolimod compared with RMS-vehicle–treated mice (Figure 8C). These data thus support the contention that fingolimod restrains the egress of DCs from the brain that would otherwise enhance Treg activity. Furthermore, they indicate that there must be a subpopulation of DCs in the normal CNS that restrain anti-CNS immune responses in the CxLNs.
A subset of forebrain DCs restrains anti-CNS immunity. Our preceding data suggested the presence of a subset of CNS DCs that suppresses anti-CNS immune responses in the CxLNs. We therefore isolated the DCs from the forebrain, where DCs are concentrated along the RMS (Figures 1–3), carefully denuded of meninges to remove any associated DCs. The remaining, predominantly RMS-associated DCs of 10 normal mice were purified by FACS (Supplemental Figure 10). We then tested the ability of sorted CD45hiCD3–CD11c+CD11b+ and CD45hiCD3–CD11c+CD11b– CNS DCs to modulate anti-CNS responses in our DTH model. Isolated DCs were directly injected into one ear, and acellular vehicle control into the other, of MOG-vaccinated mice, and recall responses to MOG were measured. Only CD45hiCD3–CD11c+CD11b– CNS DCs were capable of significantly reducing antigen recall responses to MOG (Figure 8D). This DC subpopulation includes pDCs that are known to augment Treg responses (40). These data confirmed that the normal mouse CNS contains a CD45hiCD3–CD11c+CD11b– cell population capable of suppressing anti-CNS responses in the CxLNs. To show that this effect was likely antigen specific, we performed the same experiment but substituted methylated BSA for MOG for vaccination and challenge, and found no significant modulation of DTH (Figure 8E). This observation suggests that DCs modulate CxLN responses in a CNS antigen–specific manner.