CXCL12 Limits Inflammation by Localizing Mononuclear Infiltrates to the Perivascular Space during Experimental Autoimmune Encephalomyelitis¹

The CNS is considered an immunologically specialized site where leukocyte trafficking is restricted by the blood-brain barrier (BBB),⁴ a complex organization of endothelial cells with tight junctions whose basement membranes are enveloped by glial foot processes (1, 2). Leukocytes that traverse the microvasculature must exit the perivascular space through this glial limitans to gain entry into the CNS parenchyma (3). In the autoimmune disease multiple sclerosis (MS) and its animal model, experimental autoimmune encephalomyelitis (EAE), the development of demyelinating lesions within the CNS is associated with the perivascular accumulation of mononuclear cells (4, 5). Studies suggest that myelin-specific T cells within these infiltrates enter the CNS parenchyma where they initiate inflammation, leading to the subsequent recruitment of effector cells such as macrophages and the activation of microglia (6, 7, 8). These effects ultimately result in demyelination and axonal damage, which produce the clinical deficits observed in MS and EAE. Although several adhesion molecules have been implicated in the interactions between mononuclear cells and the CNS endothelium (9, 10, 11), the chemoattractant molecules responsible for the movement of leukocytes into and out of the perivascular space are unknown. Knowledge of the chemoattractant molecules is of considerable interest for the development of therapies that would limit the development of inflammatory infiltrates in patients with MS.

It is well established that chemokines are essential for directing the movement of leukocytes into tissues (12). Chemokines are a superfamily of >50 structurally homologous chemotactic secreted proteins with their target cell specificity conferred by G_i-coupled, seven-transmembrane glycoprotein chemokine receptors. These proteins may be functionally classified into inflammatory and homeostatic members that direct leukocytes into parenchymal tissues during inflammation and into lymphoid tissues during immune surveillance, respectively (13, 14). Studies in rodents have suggested prominent roles for CCL2 and its receptor, CCR2, in the recruitment of monocytes and for CXCL9–11 and their receptor, CXCR3, in the immune responses of lymphocytes during the induction of EAE (15, 16, 17, 18, 19, 20). Although these and other studies have focused on inflammatory chemokines, the normal CNS has been shown to express several secondary lymphoid T lymphocyte chemoattractants including CCL19, CCL21, and CXCL12 (21, 22, 23). CCL19 and CCL21 promote dendritic and T cell interactions within lymphoid tissues, whereas CXCL12 regulates germinal cell polarity (24, 25). All three of these chemokines are expressed during EAE and can be detected in the cerebrospinal fluid of patients with both inflammatory and noninflammatory neurological diseases (26, 27). Recently, CXCL12 expression by endothelial cells and astrocytes in both active and inactive MS lesions was reported (28). The investigators speculated that endothelial CXCL12 might play a role in the extravasation of leukocytes during the development of inflammatory lesions in patients with MS.

The chemokine CXCL12 and its receptor, CXCR4, have well known roles in the patterning and function of the immune and nervous systems, where they localize various cell types to specific microenvironments (29). CXCL12 was originally identified as a bone marrow stromal cell-derived chemoattractant and proliferative factor for B cell precursors (30, 31). More recently, studies have demonstrated roles for CXCL12 and CXCR4 in multiple aspects of B and T cell development and in effector immune responses (32, 33, 34). CXCL12 is constitutively expressed by a variety of parenchymal tissues, including the CNS where it participates in neuronal proliferation, migration, axonal pathfinding, and myelination during development (35, 36, 37, 38). Although CXCL12 expression on neurons, microvasculature, and glial cells has been described, its role in the functioning of the normal or inflamed adult nervous system has not been extensively investigated. At many tissue sites, CXCL12 expression increases during autoimmune disease, and CXCR4 participates in the localization, proliferation, and activation of effector leukocytes at inflamed tissues sites (39, 40, 41). AMD3100, a bicyclam antagonist of CXCR4 signaling (42, 43), has been used to analyze the role of this receptor in a variety of biological processes as the targeted deletion of either CXCL12 or CXCR4 leads to embryonic lethality due to defects in the development of multiple organ systems (38, 44). Amelioration of disease in a variety of murine models of autoimmunity has been accomplished via chronic treatment with AMD3100 (45, 46), suggesting that CXCR4 activation is required during the expression of certain autoimmune diseases.

Because of the specialized functions of the CNS microvasculature and the high level and wide cellular range of expression of this chemokine in the normal, relatively immunologically quiescent CNS, we thought it unlikely that CXCL12 would function to promote extensive leukocyte trafficking into the normal CNS parenchyma. The endothelial cell expression of CXCL12 in the CNS, where there is little immune surveillance and no intraparenchymal lymphoid tissues, is curious and suggests that this chemokine could act to regulate leukocyte extravasation under normal and inflammatory circumstances. In this study we report that CXCL12 is expressed by normal spinal cord white matter endothelium at the basolateral surface. During induction of EAE, CXCL12 expressed is increased and is associated with perivascular mononuclear cell infiltrates expressing CXCR4. Inhibition of CXCR4 via administration of AMD3100 results in extensive migration of mononuclear cells out of the perivascular space and into the parenchyma, leading to increased inflammatory mediators and worsened clinical disease. Together, these studies suggest that CXCL12 plays a role in limiting the intraparenchymal migration of mononuclear cells during CNS autoimmune disease, identifying a novel anti-inflammatory role for this chemokine within the CNS.

C57BL/6 mice (The Jackson Laboratory) were maintained in pathogen-free conditions (Department of Comparative Medicine, Washington University, St. Louis, MO). Abs used include the following: CXCL12 rabbit polyclonal Ab (PeproTech); actin Abs (Sigma-Aldrich); IgG isotype control Abs (Jackson ImmunoResearch Laboratories); CXCR4 polyclonal Abs (Leinco Technologies); monoclonal rat anti-mouse-CD31 (PECAM), CD11b, mouse anti-human CD3, and rat anti-mouse-CD45 (BD Pharmingen); rat anti-mouse glial fibrillary acidic protein (GFAP) Ab (Zymed Laboratories); monoclonal anti-NeuN Ab (Chemicon International);and fluorescently conjugated Abs against CD3, CD4, CD8, CD19, CD11b, CD45, and Gr-1 (BD Pharmingen).

Active EAE was induced in 8- to 12-wk-old female C57BL/6 (The Jackson Laboratory) mice by s.c. immunization with a murine myelin oligodendroglial glycoprotein (MOG) peptide (MOGp35-55) (Sigma Genosys) in PBS emulsified 1:1 with IFA supplemented with 500 μg/ml inactivated Mycobacterium tuberculosis for a total 50-μg immunizing dose of MOGp35-55. Mice received 300 ng of pertussis toxin (List Laboratories) i.v. at the time of immunization and 48 h later, as previously described (18). For adoptive transfer experiments, MOG-specific T cells were generated as previously described (47). Activated cells were collected and transferred retro-orbitally at 2.5 × 10⁶ cells per mouse. Mice from all experiments were graded for clinical manifestations of EAE by the following criteria: 1, tail weakness; 2, difficulty righting; 3, hind limb paralysis; 4, forelimb weakness or paralysis; 5, moribund or dead.

Subcutaneous osmotic pumps (Alzet) loaded with 2 or 20 mg/ml AMD3100 in sterile PBS or PBS alone were used according to the manufacturer’s instructions. The infusion rate was 0.25 μl/h and consistent administration was achieved for 14 days postimplantation of the pumps. Osmotic pumps were implanted on days 0 and 9 postimmunization with MOGp35-55 or on day 0 after adoptive transfer of MOG-specific Th1 cells. For hemopoietic stem cell mobilization experiments, animals were injected with 5 mg/kg AMD3100 s.c. 1 h before sacrifice via exsanguination.

Animals were deeply anesthetized with a xylazine/ketamine mixture and perfused intracardially with normal saline and then with 4% paraformaldehyde on days 0, 10, and 14 postimmunization. Spinal cords were collected embedded in paraffin or frozen medium, and both transverse and longitudinal sections were generated and stained with H&E. For myelin evaluation, frozen sections were stained with FluoroMyelin Green fluorescent stain according to the manufacturer’s instructions (Molecular Probes).

Fixed, H&E-stained transverse and longitudinal sections of spinal cords from either PBS-treated (n = 3) or AMD3100-treated (n = 3) mice with an EAE grade of 2 were analyzed by a blinded observer under light microscopy. Measurements of white matter lesions were made to the nearest 5 μm via an intra-eyepiece reticle representing 1 mm at ×100 total magnification, with subdivisions representing 10 μm. Inflammatory cell identification was confirmed by morphology at ×400 total magnification. Because the vast majority of lesions approximated triangular shapes, area was calculated by multiplying the maximum measurement in each dimension and dividing by two. Mean lesion size for the respective treatment groups was calculated by summation of all areas calculated and division by the total number of lesions measured in each group (n_PBS = 91; n_AMD3100 = 156).

Cells were isolated from the spinal cords of mice with EAE as previously described (18) and stained with fluorescently conjugated Abs to CD3, CD4, CD8, CD19, CD11b, CD45, and Gr-1 (see above). Data collection and analysis were conducted using a FACScalibur flow cytometer using CellQuest software (BD Biosciences). For fluorescence analysis, a bivariate dot plot of forward vs side scatter was generated and separate regions for lymphocytes and monocytes were delineated. Fluorescent analysis displays then were generated by gating on either lymphocytes or monocytes/ macrophages.

Total RNA was prepared and QPCR was performed as previously described (48) using primers for IL-2, TNF-α, IFN-γ, IL-10, chemokines, and chemokine receptors whose sequences have been previously published (18, 48). Calculated copies were normalized against copies of the housekeeping gene GAPDH.

Paraffin sections underwent deparaffinization, Ag recovery, and permeabilization as previously described (48, 49), and frozen sections were permeabilized, blocked, and stained as previously described (48). For detection of CXCL12 and anti-CD31 (PECAM) or anti-NeuN (neuronal marker), Image-iT FX signal enhancer (Molecular Probes) solution was used according to the manufacturer’s instructions. Primary Abs (see above) were used at the following dilutions: anti-CXCL12 (1/20), anti-mPECAM-1 (1 μg/ml), monoclonal anti-NeuN (1/50), CXCR4 (1 μg/ml), and GFAP, CD11b CD3, or CD45 (1–10 μg/ml). Primary Abs were detected with secondary goat anti-rabbit or mouse IgG conjugated to Alexa Fluor 594 or Alexa Fluor 488 (Molecular Probes) for immunofluorescence staining, and nuclei were counterstained with 4′,6′-diamidino-3-phenylindole (DAPI) or ToPro3. Control sections were incubated with antisera in the presence of a 100 μmol/L excess of peptide or with isotype-matched IgG. Sections were analyzed using a Zeiss LSM 510 laser-scanning confocal microscope and accompanying software (Zeiss). Volocity image analysis software (Improvision) was used to generate and analyze three-dimensional renderings of confocal images. Stained regions were identified by applying a classifier to exclude objects smaller than 0.1 μm³ and pixels of intensity of <45 (scale 0–225).

Female C57BL/6 mice were immunized with MOGp35-55, and pertussis toxin was administered on days 0 and 2 postimmunization as described above. On day 10 postimmunization, AMD3100 (5 mg/kg) was administered to a group of naive female C57BL/6 mice. One hour later, blood was harvested into heparin-containing syringes and PBMC were prepared using Lympholyte-M (Cedarlane Laboratories). Hemopoietic precursor cells (HPCs) were isolated by negative selection by incubating the PBMCs with Mouse Lineage Mixture (Caltag Laboratories), an anti-cell marker Ab mixture, followed by incubation with anti-biotin microbeads (Miltenyi Biotec), and purified hemopoietic stem cells were obtained via magnetic columns (Miltenyi Biotec). At day 10 postimmunization, MOG-immunized mice received 280,000 stem cells in HBSS with 1% FCS and 20 mM HEPES (vehicle) or vehicle alone i.v., which corresponded to the total number of hemopoietic stem cells recovered from one AMD3100-treated animal. Mice were followed on a daily basis and scored for clinical EAE as described above. In some experiments, CD45.1 HPCs were isolated from the peripheral blood of AMD3100-treated mice and i.v. transferred to MOG-immunized CD45.2 mice at day 10 postimmunization, and blood from recipient mice was examined for the presence of transferred CD45.1 cells via flow cytometry at day 14 postimmunization.

All values are expressed as mean ± SEM. The Student t test was used to determine the statistical significance of QPCR, histological, and flow cytometry analyses, whereas mean maximal disease severity significance was determined by the Mann-Whitney nonparametric test with values of p < 0.05 considered statistically significant for all analyses. For analysis of disease severity curves, statistical significance of curves was determined using the Wilcoxon sign rank test.

To define the roles of CXCL12 and CXCR4 in autoimmune disease of the spinal cord, we used the active immunization model of EAE induced by administration of the MOGp35-55 peptide. Expressions of CXCL12 and its receptor, CXCR4, have not been evaluated extensively in the normal or diseased spinal cord. Examination of CXCL12 mRNA expression via QPCR of the spinal cord from unimmunized mice and at 9 (clinical score 0) and 14 days (clinical score 2) postimmunization with MOGp35-55 revealed high levels of expression in the normal tissue that are significantly increased at the peak of clinical disease (Fig. 1,A). Similar evaluation of other secondary lymphoid chemokines, CCL19 and CCL21, revealed that CCL19, but not CCL21, increased during induction of EAE (Fig. 1 A). Previous studies of CCL19 and CCL21 in the brains of mice with adoptive transfer EAE detected increased expression of both of these chemokines (21, 26). The absence of CCL21 up-regulation in the spinal cord of animals with active immunization EAE suggests a differential role for CCL21 in various CNS compartments.

The up-regulation of spinal cord CXCL12 expression during the induction of EAE suggested that this chemokine might be important in leukocyte migration during CNS inflammation. Prior studies of CXCL12 mRNA expression in brain had localized CXCL12 to subpopulations of forebrain neurons and to the microvasculature of both gray and white matter (23). To better evaluate endothelial cell expression of CXCL12 within the spinal cord, double-labeled immunofluorescent confocal microscopy was performed on spinal cord sections from normal mice and from those derived 10 and 14 days postimmunization with the MOGp35-55 peptide. In both normal and inflamed spinal cords, the CXCL12 protein colocalized with the endothelial cell marker PECAM-1 within both gray (not shown) and white matter (Fig. 1,B). Subpopulations of neurons within the gray matter also expressed CXCL12, as demonstrated by double labeling with the neuronal nucleus marker NeuN (Fig. 1,B, upper left panel, inset). In spinal cord sections derived from normal mice and from those obtained 10 days postimmunization with MOGp35-55, quantification of fluorescence intensity during confocal microscopy revealed a polarity in peak endothelial cell CXCL12 expression that localized this chemokine to the basolateral surface of microendothelial cells (parenchymal side), opposing the perivascular space (Fig. 1,B, right panels). Although PECAM staining could be detected easily in vessels of noninflamed spinal cord, the staining intensity was decreased at day 10 postimmunization with MOGp35-55, a time point at which mononuclear cells have begun to enter the CNS in search of Ag. Interestingly, analyses of spinal cord sections obtained from animals at day 14 postimmunization, which is at the peak of disease, revealed that vessels surrounded by numerous mononuclear cells entering the parenchyma have more diffuse PECAM staining and complete loss of CXCL12 polarity (Fig. 1,C, compare top left and right panels). Quantitative confocal microscopy of these highly inflamed vessels demonstrates a more even distribution of CXCL12 and PECAM, with loss of the previous basolateral peak in CXCL12 expression (Fig. 1,C, bottom panel). Prior studies have indicated that activation of CNS microvasculature is associated with more diffuse distribution of PECAM staining (50). Three-dimensional reconstructions of vessels stained with CXCL12 (red) and PECAM (green) better demonstrated this reversal of the detection of expression with noninflamed and day 10 vessels exhibiting red exteriors and green interiors and severely inflamed day 14 vessels exhibiting green exteriors and red interiors (Fig. 1 C). Experiments using control IgG Abs did not demonstrate any specific staining (data not shown). These results suggest that CXCL12 exerts its chemoattraction toward the parenchymal side of vessels and that this polarity of expression is disrupted during endothelial cell activation and increased BBB permeability.

QPCR evaluation of the chemokine receptors for CCL19 and CXCL12, CCR7 and CXCR4, respectively, similarly revealed significantly increased expression of these chemokine receptor mRNAs at the peak of clinical disease, with CXCR4 exhibiting the highest fold increase (Fig. 2,A). Identification of cells expressing CXCR4 via double-labeled immunofluorescent microscopy with anti-CXCR4 Abs revealed them to be CD11b-, GFAP-, and CD45-expressing macrophages/microglia, astrocytes, and leukocytes, respectively (Fig. 2,B). Interestingly, many of the CD11b⁺ and CD45⁺ cells that expressed CXCR4 were vessel associated. The pattern of expression of CXCR4 was quite distinct from GFAP, with CXCR4 more centrally located and GFAP characteristically within astrocyte processes (Fig. 2,B). The use of isotype control Abs did not result in any specific staining (Fig. 2,B). CXCR4 expression was limited to scattered cells within the white matter of spinal cord tissues from unimmunized animals (Fig. 2 B). The location and morphology of these cells suggest that they are oligodendrocytes. Taken together, these results indicate that the increased levels of CXCR4 expression observed during the induction of EAE are the results of both infiltrating CXCR4-expressing mononuclear cells and the expression of CXCR4 by activated astrocytes and microglia within the parenchyma.

Based on the expression patterns of CXCL12 and CXCR4 in the spinal cord microvascular endothelium and perivascular/parenchymal mononuclear cells, we hypothesized that these proteins might play a role in the movement of leukocytes out of the perivascular space during the development of autoimmune inflammation. To test this hypothesis, we administered AMD3100 to animals during the induction of active immunization EAE. AMD3100 is a bicyclam competitive antagonist of CXCL12 binding to CXCR4 (43). This compound has been widely used in basic and clinical studies for evaluating CXCR4 function. AMD3100 induced a dose-dependent, significant worsening of clinical severity when administered to animals during the first 2 wk of the disease (Fig. 3,A). Disease severity in AMD3100-treated animals with EAE displayed a plateau phase without significant recovery, whereas the disease curves of PBS-treated animals were more typical of monophasic EAE induced by immunization with the MOGp35-55 peptide (Fig. 3, A and B). The higher dose of AMD3100 (4 mg/kg/day) has been used to inhibit CXCR4 activation in multiple animal models (45, 46, 51). Thus, this dose was used for the remainder of our experiments. Because CXCL12 is highly expressed in secondary lymphoid tissues and it was possible that AMD3100 affected immunization, we also treated cohorts of animals with AMD3100 beginning at day 9 postimmunization, a time point at which we and others have previously demonstrated that autoreactive T cells are present and have begun trafficking into the CNS (18, 53). Treatment with AMD3100 beginning at day 9 postimmunization also led to significant worsening of disease (Fig. 3,B), suggesting that the development of inflammation at the spinal cord was the target of AMD3100. Finally, administration of AMD3100 to mice during adoptive transfer of MOG-specific T cells also led to significantly worsened disease but with more pronounced recovery (Fig. 3,C), suggesting that CXCR4 antagonism does not lead to worsened EAE via effects at peripheral lymphoid tissues but rather that this worsening is due to direct effects at the spinal cord. Importantly, in all experiments AMD3100 led to a significant increase in overall clinical severity of EAE and a trend toward increased overall incidence but did not appreciably alter the time of onset of disease (Fig. 3 and Table I). In addition, the mean maximal disease scores were significantly higher only in those animals treated with the standard dose of AMD3100 (4 mg/kg/day) (Table I). Thus, CXCR4 is unlikely to play a role in the initiation of disease by MOGp35-55-specific T cells but does play a role in the progression of inflammation.

To examine the basis of our clinical findings, we analyzed spinal cord tissues from MOGp35-55-immunized animals treated with PBS vs AMD3100. Histological examination of spinal cord sections from PBS and AMD3100-treated animals with EAE of the same clinical grade revealed startling differences in the extent of the inflammatory infiltrates within the ventral white matter at day 14 postimmunization (Fig. 4,A). Although the PBS-treated animals developed the typically intense perivascular mononuclear cell infiltrates with moderate parenchymal infiltration, AMD3100-treated animals developed extensive mononuclear parenchymal inflammatory infiltration, essentially obliterating the ventral white matter region. Because AMD3100 mobilizes bone marrow stem cells and thus could theoretically alter the initial neutrophilic component during early EAE induction, we also performed histologic analyses of spinal cords from PBS- and AMD3100-treated animals at 9 and 10 days postimmunization, which did not reveal differences in the neutrophilic infiltrate (data not shown). An analysis of spinal cord sections from PBS- and AMD3100-treated mice with myelin stains at day 14 postimmunization revealed increased demyelination as a result of AMD3100 treatment (Fig. 4,B). Quantitative analyses of lesions revealed a significant increase in the average areas of lesions in AMD3100-treated mice with EAE compared with those that received PBS (Fig. 4 C). Taken together, these results are consistent with the hypothesis that CXCR4 plays a role in adjusting the severity of disease by limiting the extent of the parenchymal infiltration of mononuclear cells.

Because CXCL12 plays an essential role in the localization of lymphocytes to specific zones within lymphoid tissues, we wondered whether its expression patterns were altered during the induction of CNS autoimmunity. QPCR analyses of CXCL12 and CXCR4 revealed that mRNAs for these molecules are significantly decreased during the induction of EAE (Figs. 5, A and B). For control purposes, expressions of CCL19, CCL21, and CCR7 mRNAs were also analyzed, with similar results. These data suggest that migration of mononuclear cells out of lymphoid tissues during the initiation of CNS autoimmune disease may be partly due to release from the localizing effects of secondary lymphoid chemokines via decreased expressions of both ligands and receptors. In addition, they support the notion that AMD3100 is unlikely to exacerbate EAE through effects at peripheral lymphoid tissues.

AMD3100 has been shown to efficiently mobilize HPCs, which use CXCR4 for localization to proliferative microenvironments within the bone marrow (53, 54, 55). To determine whether AMD3100-mediated mobilization of HPCs affected the induction of EAE, we i.v. transferred AMD3100-mobilized HPCs from CD45.1 congenic mice into MOG-immunized C57BL/6 (CD45.2) mice at day 10 postimmunization, a time point when encephalitogenic T cells have begun trafficking into the CNS, and we observed animals for alterations in the clinical severity of the disease. As indicated in Fig. 5,B, there was no difference in the clinical severity of EAE in animals that received vehicle vs those that received vehicle containing HPCs. Flow cytometric analysis of PBMCs obtained from recipient mice at day 14 postimmunization demonstrates that, although transferred HPCs did not affect disease course, they survived and began to express CD45 (Fig. 5 C). Taken together, our data suggest a separate local effect of CXCR4 within the spinal cord during the induction of EAE.

Our histological results demonstrated that AMD3100-treated mice with EAE did not develop the usual intense perivascular infiltrates but instead had widening of intraparenchymal lesions with increased demyelination of white matter tracts. To determine whether loss of CXCR4 activation affected the overall numbers of leukocytes trafficking into the spinal cord, we analyzed leukocyte infiltrates in PBS- and AMD3100-treated mice with EAE of similar clinical grades via flow cytometry (Fig. 6). Leukocytes were isolated from individual spinal cords from PBS- and AMD3100-treated mice with EAE at clinical grade 2 and examined for expression of CD3, CD4, CD8, CD19, CD45, CD11b, and Gr-1. Although the percentages of T and B lymphocytes and granulocytes were similar between the two treatment groups (data not shown), the percentage and total numbers of activated microglia (CD45^highCD11b^low) was significantly increased in the AMD3100-treated groups compared with those that had received PBS vehicle (Fig. 6, A and B). Consistent with this finding, we also observed a significant decrease in the numbers of resting microglia (CD45^lowCD11b^low) in AMD3100-treated animals. Examination of total cell numbers similarly did not reveal any large differences in the numbers of infiltrating leukocytes between the two treatment groups with EAE; however, a significant increase was observed in the penetration of CD45⁺ cells in the parenchyma of AMD3100-treated animals compared with those treated with PBS in all grades of EAE (Fig. 6,C). CD45⁺ cells comprising focal perivascular infiltrates in the PBS-treated mice were instead dispersed throughout the white matter in the AMD3100-treated animals. The increased penetration of CD45⁺ cells in AMD3100-treated animals with the most severe disease (grade 4) was also associated with an increase in more dimly staining CD45⁺ cells, presumed to be microglia, when compared with PBS-treated mice with more severe disease (grade 3) (Fig. 6,C, compare the two right panels). Similar dramatic differences in the positions of CD3- and CD11b-expressing cells were also observed between the two treatment groups (Fig. 6, D and E, respectively). Spinal cord sections from PBS-treated mice had focal perivascular areas containing intensely stained CD3-expressing T cells and CD11b^high-expressing macrophages, whereas the CD3- and CD11b-expressing cells detected in the sections from AMD3100-treated mice were more dispersed and, in the case of CD11b-expressing cells, consisted of both CD11b^high macrophages and CD11b^low microglia (Fig. 6,E, insets). The increased penetration of CD11b-expressing cells and the enhanced microglial activation in the AMD3100-treated animals were also associated with increased demyelination as assessed by double-labeled immunofluorescent labeling of myelin and CD11b expression in spinal cord sections from grade 3 PBS-treated and grade 4 AMD3100-treated mice (Fig. 6 E). These data suggest that CXCR4 determines the extent of mononuclear cell parenchymal penetration and microglial activation during EAE.

Because CXCR4 antagonism led to worsened clinical severity of EAE with a widening of inflammatory infiltrates and increased microglial activation and demyelination, we evaluated the mRNA expression of inflammatory mediators and chemokines within the spinal cords of PBS- and AMD3100-treated mice with grade 2 EAE. QPCR evaluation revealed that AMD3100-treated mice with EAE had significantly increased levels of Th1 cytokines IL-2, TNF-α, and IFN-γ, whereas levels of IL-10 were lower than those in PBS-treated mice, consistent with an anti-inflammatory role for CXCL12/CXCR4 during EAE (Fig. 7,A). Analysis of the regulatory T cell markers CD25, FoxP3, and TGF-β (56) did not reveal any significant differences (data not shown). With the exception of CXCL10, levels of chemokine (Fig. 7B) and chemokine receptor (not shown) mRNAs previously observed to be increased in the spinal cord during EAE induction were similar in the two treatment groups. Because CXCL10 is expressed by activated glia and macrophages that infiltrate the CNS (57, 58), this finding is also consistent with our data suggesting that loss of CXCR4 activity leads to the increased trafficking of mononuclear cells with enhancement of the subsequent parenchymal inflammatory responses.

In the present study we demonstrate that the chemokine CXCL12 is expressed on the basolateral surface of spinal cord microvasculature, where it functions to prevent CXCR4-expressing mononuclear cells from migrating out of the perivascular space and into the parenchymal tissue during CNS autoimmune disease. This distribution pattern is altered during the onset of inflammation, suggesting that disruption of this chemokine gradient could play a role in the extravasation of leukocytes into the CNS parenchyma. Chronic administration of AMD3100, a small molecule inhibitor of CXCR4 signaling, during the induction of EAE led to extensive intraparenchymal migration of mononuclear cells with the resultant activation of microglia, extensive demyelination, increased inflammation, and worsened clinical disease. Normally, EAE and the disease it models, MS, are characterized by intense perivascular cuffs of mononuclear cells with the limited parenchymal migration of leukocytes associated with circumscribed demyelination. In our AMD3100-treated animals, these tight perivascular lesions were replaced by extensive sheets of mononuclear cells infiltrating the ventral white matter, with subsequent damage to the myelin sheaths. Although AMD3100-treated animals with EAE had worsened clinical disease, higher levels of spinal cord Th1 cytokines, and extensive demyelination, the total numbers of mononuclear cells were similar to their PBS-treated counterparts. Interestingly, significant differences were observed in the numbers of CD45^highCD11b^low cells (activated microglia) and CD45^lowCD11b^low cells (resting microglia), which were higher and lower, respectively, in the AMD3100-treated animals with EAE. These data suggest that AMD3100 treatment is associated with an increase in microglial activation and not increased recruitment of peripheral leukocytes. In support of this interpretation, we observed a dramatic difference in the intraparenchymal positions of these mononuclear cells, with cells located throughout the ventral white matter, suggesting that CXCR4 plays a role in preventing mononuclear cell trafficking out of the perivascular space and into the CNS parenchyma, thereby limiting the extent of inflammation within the parenchyma.

Although a variety of studies have analyzed the expression patterns of CXCL12 and CXCR4 in the developing and adult CNS, there has been little work describing a role for these molecules in the pathogenesis of CNS diseases. The detection of CXCL12 in the cerebrospinal fluid of patients without neuroinflammatory diseases indicates that this chemokine does not necessarily promote inflammation within the CNS (28). Indeed, our data would argue that CXCL12 prevents CXCR4-expressing mononuclear cells from penetrating the CNS parenchyma, where they would contribute to microglial activation and demyelination via the actions of inflammatory mediators such as IFN-γ (59, 60). Consistent with this argument, Stumm et al. (23) previously reported that the administration of LPS is associated with decreased expression of CXCL12 by CNS endothelial cells, supporting the notion that CXCL12 does not promote activation of the CNS microvasculature. The observed redistribution of CXCL12 in activated microendothelial cells with perivascular infiltrates argues instead that directed expression of CXCL12 may be altered in response to this activation. Thus, CXCL12 detected in the cerebrospinal fluid of patients may simply be a marker of endothelial cell activation. CXCR4 could also play a role in promoting interactions between endothelial cells, astrocytes, and microglia. Further studies will be necessary to more completely elucidate the role of CXCL12 in the regulation of mononuclear cell trafficking across the BBB.

We also analyzed the kinetics of expression of two other secondary lymphoid chemokines, CCL19 and CCL21, during the induction of EAE via MOG immunization. Both of these chemokines are expressed by normal human brain tissue (61) and elevated in the cerebrospinal fluid of MS patients (27), and cells expressing their receptor, CCR7, accumulate within inflammatory lesions (62). Similar findings have been observed in brain tissue from mice with EAE (21). Although we also observed increased expressions of CCL19 and CCR7 mRNAs in spinal cord tissue from mice with EAE, we did not detect alterations in the expression of CCL21. As CCL19 expression has been limited to large venules (21), it is possible that this chemokine plays a role in the egress of lymphocytes out of the parenchyma during inflammation. Further studies are underway to address the role of CCL19 in lymphocyte movement through the CNS parenchyma during EAE.

Increased CXCR4 expression in the spinal cords of mice with EAE has been previously observed (63). In prior studies it was concluded that endogenous neural cells were the source of this chemokine receptor. In contrast, we observed CXCR4 expression within both mononuclear and parenchymal cells of the spinal cord in mice with EAE. Treatment with AMD3100 altered only the positions of infiltrating macrophages and T cells, leading to the loss of the typical perivascular infiltrates normally observed in EAE. Thus, although CXCR4 activity does not appear to be required for mononuclear cells to cross the endothelial barrier, it does seem to be important for the localization of these cells to the perivascular space. Because we did not see vast differences in the numbers of mononuclear cells obtained from spinal cords of PBS- vs AMD3100-treated animals, it remains unclear whether this localization is required to expose these and the microglial cells to factors produced by activated endothelial cells that would affect their proliferation. Alternatively, the enhanced expression of inflammatory cytokines observed during AMD3100 treatment may indicate that these cells are normally exposed to anti-inflammatory mediators within the perivascular space or that close interactions between T cells and macrophages at this site direct their mutual effector activities. Expressions of the anti-inflammatory cytokines IL-10 and IL-4 have been localized to perivascular macrophages within active MS lesions (64, 65). These and other factors could influence the inflammatory responses of mononuclear cells as they migrate out of the blood and into the CNS parenchyma.

Because CXCL12 is expressed by secondary lymphoid tissue, it is possible that AMD3100 exerts its effects peripherally at the lymph node. However, the administration of AMD3100 9 days postimmunization or during adoptive transfer EAE also led to increased clinical severity as compared with PBS treatment, indicating that CXCR4 antagonism during immunization is not required to observe an increased clinical severity of EAE. In addition, we observed a down-regulation of CXCL12 and CXCR4 expression in splenic tissues during the induction of EAE, suggesting that CXCL12 activity in secondary lymphoid tissues is decreased during EAE induction. This decrease in CXCL12 expression might abrogate CXCR4 activation and, therefore, any activity of AMD3100 treatment in the periphery during the immunization phase of EAE.

AMD3100 was initially developed to prevent lymphocyte entry by CXCR4 using strains of HIV-1 (66, 67). More recently, AMD3100 has been shown to induce the rapid mobilization of hemopoietic stem cell precursors, which require CXCL12/CXCR4 for their localization within the bone marrow (53, 54). In prior studies, another bone marrow stem cell mobilization agent, G-CSF, was shown to ameliorate EAE (65). This effect was associated with a shift toward anti-inflammatory and Th2 cytokines, suggesting that regulatory lymphocytes were involved. In a recent study, G-CSF was observed to down-regulate CXCL12 expression in human bone marrow, leading to the release of CD4⁺CD25⁺ regulatory T cells that were localized there via CXCL12/CXCR4 signaling (68). Although these data suggest that CNS autoimmune disease might be inhibited by the release of bone marrow-derived regulatory T cells, our results with the adoptive transfer of AMD3100-mobilized cells indicate that this mobilization alone does not regulate EAE disease expression. Although it is also possible that CXCR4 activity is required for the migration or action of regulatory T cells within the CNS, AMD3100-mobilized hemopoietic precursor cells did not affect the induction of active immunization EAE when adoptively transferred into MOGp35-55- immunized mice without AMD3100 treatment. In addition, we did not observe any differences in the spinal cord expression of regulatory T cell markers in our PBS- vs AMD3100-treated animals. Thus, CXCL12 appears to have a separate anti-inflammatory role in the CNS that does not involve the specific recruitment of regulatory T cells.

The effect of AMD3100 on autoimmune disease of the CNS is contrary to that observed in studies of autoimmunity at other tissues sites. Prior studies using murine models of arthritis and asthma have demonstrated that the disruption of CXCL12/CXCR4 signaling via the administration of AMD3100 leads to amelioration of disease (45, 69). Because the levels of CXCL12 are relatively low in uninjured murine synovium and tracheal epithelium (70, 71), high levels of its expression during inflammation within these tissues and reduced disease expression after CXCR4 inactivation are consistent with CXCL12 exerting a proinflammatory role in these autoimmune diseases. These studies also observed that CXCR4 antagonism decreased the expression of Th1 cytokines in inflamed tissues, which was not detected in our analyses of spinal cord tissues from PBS- and AMD3100-treated mice. As CXCL12 is expressed in high levels in the normal CNS, it is more logical that it would exert anti-inflammatory effects under normal circumstances. Consistent with this possibility, we found that AMD3100-treated mice had significantly increased levels of IFN-γ, decreased levels of IL-10, and increased levels of cytokines and chemokines associated with monocyte and microglial activation, including TNF-α and CXCL10 (72, 73), indicating that CXCR4 activation is anti-inflammatory in the spinal cord. Others have reported that CXCR4 expression by cells of myeloid lineage is dynamic and regulated by the activation state of the cell (74, 75). The up-regulation of CXCR4 by microglial cells during activation may be important for anti-inflammatory interactions between these cells and CXCL12-expressing endothelial cells at the BBB.

An additional caveat of all pharmacological studies regards the specificity of the administered agent. The specificity of AMD3100 for CXCR4 in human leukocytes was determined to rely on interactions of the drug with two aspartic acid residues located at the interphase between the transmembrane domains and the extracellular regions of human CXCR4 (42). These aspartic acid residues are completely conserved across primate, rodent, and feline species. However, the specificity of AMD3100 for murine chemokine receptors other than CXCR4 has not been as rigorously evaluated as it has been for primate CXCR4. Thus, it is remotely possible that the in vivo use of AMD3100 could lead to nonspecific effects on alternative receptors, including those outside the chemokine receptor family. Studies using transgenic approaches could confirm the in vivo role of CXCR4 during autoimmune diseases.

In summary, CXCR4 antagonism was associated with worsening CNS autoimmune disease due to increased parenchymal penetration of mononuclear cells. The increased clinical severity of disease was associated with increased expression of inflammatory mediators and augmented microglial activation and demyelination. We propose an anti-inflammatory role for CXCL12 and CXCR4 within the CNS and suggest that inhibition of CXCR4 for the prevention of HIV-1 infection of lymphocytes should be approached with caution, as it may disrupt the normal BBB regulation of leukocyte entry and promote the migration of HIV-1-infected macrophages into the CNS.

We thank Julia Sim and Judy Tollett for technical assistance. We also thank Drs. Joshua Rubin, John Russell, Anne Cross, and Michael Diamond for experimental advice and critical comments on the manuscript.

The authors have no financial conflict of interest.