Activation of kinin receptor B1 limits encephalitogenic T lymphocyte recruitment to the central nervous system

Abstract

Previous proteomic and transcriptional analyses of multiple sclerosis lesions^{1, 2, 3} revealed modulation of the renin-angiotensin and the opposing kallikrein-kinin pathways. Here we identify kinin receptor B1 (Bdkrb1) as a specific modulator of immune cell entry into the central nervous system (CNS). We demonstrate that the Bdkrb1 agonist R838 (Sar-[d-Phe]des-Arg⁹-bradykinin) markedly decreases the clinical symptoms of experimental autoimmune encephalomyelitis (EAE) in SJL mice^{4, 5, 6}, whereas the Bdkrb1 antagonist R715 (Ac-Lys-[d-βNal⁷, Ile⁸]des-Arg⁹-bradykinin) resulted in earlier onset and greater severity of the disease. Bdkrb1-deficient (Bdkrb1^−/−) C57BL/6 mice⁷ immunized with a myelin oligodendrocyte glycoprotein fragment, MOG_35–55, showed more severe disease with enhanced CNS-immune cell infiltration. The same held true for mixed bone marrow–chimeric mice reconstituted with Bdkrb1^−/− T lymphocytes, which showed enhanced T helper type 17 (T_H17) cell invasion into the CNS. Pharmacological modulation of Bdkrb1 revealed that in vitro migration of human T_H17 lymphocytes across blood-brain barrier endothelium is regulated by this receptor. Taken together, these results suggest that the kallikrein-kinin system is involved in the regulation of CNS inflammation, limiting encephalitogenic T lymphocyte infiltration into the CNS, and provide evidence that Bdkrb1 could be a new target for the treatment of chronic inflammatory diseases such as multiple sclerosis.

Introduction

Three independent mRNA and protein screens performed on inflammatory lesions of the central nervous system (CNS) (ref. 1,2) and on blood-brain barrier (BBB) membrane microdomains³ revealed unexpected changes in both the renin-angiotensin and the kallikrein-kinin systems (KKS), two pathways that are known mainly for their role in blood pressure regulation and are proposed to counterbalance each other⁸. For example, in a large-scale analysis of mRNA transcripts from multiple sclerosis lesions, prokallikrein KLKB1 was more abundant in samples from all lesion types relative to control brain samples, whereas the kallikrein inhibitor kallistatin was less abundant¹. In a proteomic study, neurolysin, which hydrolyzes bradykinin, was present in chronic active and chronic plaques, whereas lysosomal carboxypeptidase, which cleaves angiotensins as well as des-Arg⁹-bradykinin, was found only in active plaques². Kinins belong to a family of bioactive octa- to decapeptides generated from kininogens in a stepwise cleavage process⁹. Their biological activities are mediated via two pharmacologically distinct G protein–coupled receptors: kinin receptor B1 (Bdkrb1), which under physiological conditions is not found in immune cells, and the ubiquitously expressed kinin receptor B2 (Bdkrb2). In the mouse EAE model of multiple sclerosis, we found upregulation of Bdkrb1, bradykinin, des-Arg⁹-bradykinin, kallikrein-1 and kallikrein-6 as well as low-molecular-weight kininogens (KNGL) in CNS tissue and the cerebrospinal fluid (Supplementary Fig. 1). Because kallikreins cleave the precursor kininogens to bradykinin, which mediates its effect primarily via Bdkrb2, and subsequent proteases generate des-Arg⁹-bradykinin, acting via Bdkrb1, we assumed that the kallikrein-kinin system may contribute to chronic autoimmune neuroinflammation. Bdkrb1 expression has been found not only on brain endothelial cells¹⁰ but also on T lymphocytes from individuals with multiple sclerosis^{11, 12} and, as shown here, on parenchymal CD3⁺ T cells within perivascular lesions (Fig. 1a).

Figure 1. Morphological and functional evidence for the involvement of the kallikrein-kinin system in autoimmune CNS inflammation.

(a) Histopathology for CD3 (green), kinin receptor B1 (Bdkrb1; red) and cell nuclei (Hoechst; blue), including overlay analysis in the CNS from a human with multiple sclerosis (top) or from a mouse with EAE (bottom). Asterisk marks the lumen of a blood vessel. Scale bars, 10 µm. (b) Pharmacological modulation of Bdkrb1 in adoptive transfer EAE in SJL mice. One of two experiments is shown. Mice were distributed into three groups and received intraperitoneal injections of either the Bdkrb1 agonist R838, the Bdkrb1 antagonist R715 or vehicle daily for days 0–10 (n = 6 per group). (c) Therapeutic treatment effect for the Bdkrb1 agonist R838, given after disease onset twice daily, is demonstrated in SJL mice immunized with PLP_139–151 (n = 6 per group). Representative results from two independent experiments are shown. (d) Bdkrb1 deficiency enhances autoimmune neuroinflammation. Clinical scores for wild-type (WT) EAE (n = 9) and Bdkrb1^−/− EAE (n = 9) mice after immunization with MOG_35–55. Representative results from three independent EAE experiments are shown. (e) Combined disruption of Bdkrb1 and Bdkrb2 results in a disease course similar to that seen with Bdkrb1 disruption alone. Bdkrb1^−/− ;Bdkrb2^−/− (n = 5) and Bdkrb1^−/− mice (n = 6) were immunized with MOG_35–55. Representative results from two independent EAE experiments are shown. For all EAE courses, mean disease scores ± s.e.m. are displayed; ^**P < 0.01, ^*P < 0.05, repeated-measures ANOVA.

In light of the known involvement of such T cells in the initiation of a myelin-specific immune attack, we induced EAE via adoptive transfer of activated proteolipid protein (PLP)_139–151–specific lymphocytes into naive SJL/J mice^{5, 6}. We treated recipients with the Bdkrb1 agonist R838 (1 mg kg⁻¹), the Bdkrb1 antagonist R715 (1 mg kg⁻¹), or with vehicle, from day 0 to day 10 after transfer. Activation of Bdkrb1 resulted in a significantly lower maximum clinical disease severity and milder clinical deficits than occurred in vehicle-treated mice (P < 0.01), whereas blocking of Bdkrb1 led to accelerated disease onset (P = 0.01; Fig. 1b). To analyze the potential of Bdkrb1 activation for therapeutic use, we immunized SJL/J mice to produce relapsing-remitting EAE and treated them after disease onset with R838 (1 mg kg⁻¹) or vehicle. Bdkrb1 activation resulted in a significantly attenuated clinical disease course (P < 0.05; Fig. 1c).

To elucidate the role of Bdkrb1 in chronic CNS neuroinflammation, we immunized Bdkrb1-deficient (Bdkrb1^−/−) C57BL/6 mice⁷ or wild-type (WT) C57BL/6 controls with myelin oligodendrocyte glycoprotein (MOG)_35–55. Bdkrb1^−/− mice had both a significantly greater maximum clinical disease severity (P < 0.05) and greater clinical deficits (Fig. 1d). To rule out a possible compensatory role of the constitutively expressed Bdkrb2 in Bdkrb1^−/− mice, we next immunized C57BL/6 mice lacking the genes encoding kinin receptors B1 and B2 (ref. 13) (Bdkrb1^−/−;Bdkrb2^−/− mice) with MOG_35–55. Despite a recently reported potential proinflammatory role of Bdkrb2 (ref. 14), we found no difference between the EAE disease courses in Bdkrb1^−/−;Bdkrb2^−/− and Bdkrb1^−/− mice (P > 0.05), which indicates that Bdkrb1 has a dominant role in influencing the course of the disease (Fig. 1e).

Histological examination at day 23 after immunization (Fig. 1d) revealed large inflammatory infiltrates, including an elevated number of activated microglia and macrophages throughout the brain stem and spinal cord, in Bdkrb1^−/− mice with EAE as compared to control WT mice with EAE (Fig. 2a,b). In parallel, the extent of demyelination and axonal damage, the main pathological feature of multiple sclerosis¹⁵, was greater in the Bdkrb1^−/− mice (Fig. 2c,d). In contrast to the strong effects of Bdkrb1 on disease severity and histological parameters, however, the myelin-specific T cell response as well as the activation status of T cells in the peripheral immune organs of the Bdkrb1^−/− mice were unaltered as compared to those of the controls (Fig. 2e–h). Moreover, the proportion of FoxP3⁺ CD25⁺ regulatory T cells¹⁶ within the CD4⁺ lymphocyte population from draining lymph nodes was similar in the Bdkrb1^−/− mice with EAE (13.1% ± 1.2) and the WT control mice with EAE (11.9% ± 0.8.; P > 0.05).

Figure 2. Bdkrb1 deficiency leads to enhanced EAE pathology.

(a–d) Histopathological analysis of three sections per mouse, comprising the assessment of inflammation by H&E staining (a), microglia and macrophage infiltration by immunohistochemistry for Iba-1 (b), demyelination by luxol fast blue staining (arrows indicate demyelinated areas) (c) and axonal damage by immunohistochemistry for APP (d). Representative images from spinal cord longitudinal sections and corresponding quantifications are shown. Scale bars, 50 µm. (e–h) Deficiency of Bdkrb1 has no impact on myelin-specific inflammatory responses in the periphery. Proliferation in response to MOG_35–55 in cells from draining lymph nodes (e,g,h) or spleens (f) from Bdkrb1^−/− or wild-type (WT) mice with EAE killed after immunization with MOG_35–55 but before the onset of disease. Shown are the results of [³H]thymidine incorporation assays in response to MOG_35–55 (e,f) and flow cytometric assessment of the expression of surface activation markers and cytokines (g,h). Values are means ± s.d.; ^**P < 0.01, ^***P < 0.001, Mann-Whitney U-test.

To further dissect which cell type is influenced by Bdkrb1 signaling during the disease, we generated bone marrow–chimeric mice¹⁷ by injecting C57BL/6-CD45.2 bone marrow into lethally irradiated C57BL/6-CD45.1 recipient mice. In the first group, we reconstituted lethally irradiated congenic C57BL/6-CD45.1 mice with a mixed bone marrow consisting of a 5:1 ratio of T cell receptor (TCR) β-chain (Tcrb^−/−)-deficient and Bdkrb1-deficient bone marrow donors^{18, 19}. The resulting Bdkrb1^−/−Tcrb^−/− → WT mixed-bone-marrow chimera lacked Bdkrb1 on T cells. In a second group, we gave irradiated C57BL/6-CD45.1 mice Bdkrb1^−/− bone marrow only, in which all reconstituted immune cells were devoid of Bdkrb1 (Bdkrb1^−/− → WT). To rule out a bias through irradiation and BM reconstitution, we gave a third group of mice a 5:1 ratio of Tcrb^−/− and C57BL/6 (WT) bone marrow (WT Tcrb^−/− → WT). Two months after bone marrow grafting, we assessed reconstitution by FACS analysis of peripheral blood and PCR analysis of MACS-sorted CD3⁺ and CD11b⁺ cells (Supplementary Fig. 2a). Immunization with MOG_35–55 to produce active EAE revealed an earlier onset and markedly greater clinical deficits in Bdkrb1^−/− → WT bone marrow chimeras as compared to WT Tcrb^−/− → WT EAE controls (P < 0.05; Fig. 3a). This pattern was similar to the EAE disease courses in germline Bdkrb1^−/− mice as compared to WT C57BL/6 controls (Fig. 1d). We found that mean EAE disease scores in Bdkrb1^−/−Tcrb^−/− → WT were virtually indistinguishable from Bdkrb1^−/− → WT EAE, pointing to the importance of T cell–expressed Bdkrb1 in CNS inflammation and to a somewhat accessory role for macrophage or endothelial kinin receptor B1 expression. The results of ex vivo proliferation assays of lymph node–derived T cells in response to stimulation with MOG_35–55 were the same for all three groups (Supplementary Fig. 2b), indicating that antigen-specific immune responses against the encephalitogenic peptide were not compromised by kinin receptor B1 deficiency (Fig. 3a).

Figure 3. Bdkrb1 controls the migratory capacities of T cells targeting the CNS.

(a) Induction of EAE by MOG_35–55 immunization of lethally irradiated C57BL/6-CD45.1 recipients reconstituted with Bdkrb1-deficient (Bdkrb1^−/− → WT; n = 5) bone marrow or a mix of TCR-β-chain– and Bdkrb1-deficient bone marrow (Bdkrb1^−/−Tcrb^−/− → WT; n = 6). To control for the impact of the transplantation procedure, control mice received a mix of TCR-β-chain–deficient and C57BL/6 WT bone marrow (WTTcrb^−/− → WT, n = 7). (b) EAE course in Bdkrb1-deficient mice that received a mix of TCR-β-chain–deficient and C57BL/6-CD45.1 bone marrow (WT Tcrb^−/− → Bdkrb1^−/−, n = 5), as compared to that in Bdkrb1^−/− → WT mice (n = 4) and Bdkrb1^−/− → Bdkrb1^−/− (n = 5) chimeras. Mean clinical disease scores ± s.e.m. are given for all three groups. ^*P < 0.05, repeated-measures ANOVA. Representative results from three (a) and two (b) independent EAE experiments are shown. (c) C57BL/6 recipients were reconstituted with mixed Bdkrb1^−/− (C57BL/6-CD45.2) and WT (C57BL/6-CD45.1) bone marrow (1:1 ratio; n = 6), and CD4⁺ T cells were isolated from the CNS at disease peak. ^**P < 0.01, Mann-Whitney U-test. (d) Migratory capacity, toward a CXCL12 chemokine gradient in a Transwell system, of PLP_139–151-specific T cells (from immunized SJL mice) and MOG_35–55-specific CD4⁺ T cells (from immunized Bdkrb1^−/− mice) that were seeded on a mouse bEnd3 brain–derived endothelial cell monolayer. Bdkrb1 was modulated by incubating T cells with the Bdkrb1 agonist R838 or the Bdkrb1 antagonist R715 before application to the endothelial monolayer. (e) Decreased F-actin polymerization of CD4⁺ T cells upon Bdkrb1 activation. (f) Downregulation of the small GTPase RhoA in CD4⁺ T cells by activation of Bdkrb1. Shown is an immunoblot of a cell extract prepared from T cells after RhoA pulldown. GTPγS-loaded controls were used as a positive control for RhoA pulldown.

Next, we reconstituted Bdkrb1-deficient mice with a mix of TCR-β-chain–deficient and C57BL/6-CD45.1 bone marrow (WT Tcrb^−/− → Bdkrb1^−/−). After immunization with MOG_35–55, EAE course in these mice was compared to that in Bdkrb1^−/− → WT as well as Bdkrb1^−/− → Bdkrb1^−/− mice (Fig. 3b). Disease courses were comparable in mice deficient for Bdkrb1 in the immune system; however, the disease was ameliorated in the mice reconstituted with WT T cells (P < 0.05). These data demonstrate that the protective effect of Bdkrb1 is mediated by its expression on T cells and exclude a possible contribution of Bdkrb1 expression in the CNS.

We therefore investigated whether Bdkrb1 would modulate CNS inflammation by affecting the migration of encephalitogenic T cells across the BBB (Fig. 3). To directly compare and quantify the homing capacity to the CNS in vivo, we reconstituted C57BL/6 mice with mixed bone marrow consisting of a 1:1 ratio of WT (C57BL/6-CD45.1) and Bdkrb1^−/− (C57BL/6-CD45.2) bone marrow. Infiltration of Bdkrb1^−/− T cells into the CNS, normalized to the reconstitution rate, was significantly higher than infiltration of WT T cells at disease peak (Fig. 3c). Moreover, we found more CD4⁺ T cells in the CNS when encephalitogenic Bdkrb1^−/− T cells, as opposed to WT T cells, were transferred to Rag1^−/− recipients (Fig. 4a). In line with this, we observed an EAE incidence of 75% when transferring Bdkrb1^−/− T cells but only of 33% with WT T cells (Supplementary Table 1).

Figure 4. Bdkrb1 activation primarily targets the invasion of T_H17 cells.

(a–c) Proportions of CD4⁺ cells in the CNS (a) and of IL-17⁺ and IFN-γ⁺ cells within the CD4⁺ T cell population (b,c) in C57BL/6 Rag1^−/− mice with adoptive transfer EAE induced by injection of Bdkrb1^−/− or WT T cells. Immune cells were isolated from the CNS of four mice per group at day 13 after cell transfer (which corresponds to the time of disease onset). (d–f) Proportions of IL-17⁺ cells (d), IFN-γ⁺ cells (e) and FoxP3⁺ CD25⁺ cells (f) within the CD4⁺ T cell population from CNS-invading immune cells recovered from bone marrow chimeric mice with T cells lacking Bdkrb1 or from control mice (see Fig. 3a). (g) Bdkrb1 activation primarily targets the migration of human memory CD45RO⁺ T_H17 rather than T_H1 cells across human brain-derived microvascular endothelium; ^*P < 0.05, Mann-Whitney U-test. (h) Increased expression of Bdkrb1 in human T_H17 cells analyzed by FACS. (i,j) Activation of Bdkrb1 decreased the average number of Celltracker Orange–labeled T_H17 cells after application to hippocampal slice cultures for multiphoton microscopy. The average number of T cells per minute and per defined volume (between 60–120 m depth) over time is shown, including quantification (i) (see also Supplementary Movies 1,2,3) and representative overviews (j). Data shown are means ± s.e.m.; ^*P < 0.05, Mann-Whitney U-test.

Using an in vitro Transwell assay, we found that PLP_139–151- and MOG_35–55-specific CD4⁺ T cells isolated from immunized SJL/J and Bdkrb1^−/− mice were attracted to migrate through a monolayer of transformed mouse brain-derived endothelial cells (ref. 20) in a CXCL12 gradient–dependent manner. Although engagement of Bdkrb1 by R838 considerably reduced the number of migrated T cells, the additional application of the Bdkrb1 antagonist R715 restored migration. By contrast, the migration of Bdkrb1^−/− cells was not altered (Fig. 3d). In PLP-specific T cells treated with R838, we did not find any alterations in adhesion molecule expression (LFA-1, VLA-4, ICAM, ALCAM, VCAM-1 and CD6; see Supplementary Table 2 for oligonucleotide primers). In contrast, we observed a markedly decreased polymerization of F-actin upon short-term Bdkrb1 activation (Fig. 3e). Moreover, in CD4⁺ T cells the activity of the small GTPase RhoA—which controls T cell migration²¹—was also downregulated (Fig. 3f). Thus, activation of the G protein–coupled kinin receptor B1 directly regulates signaling events that are important for T lymphocyte motility.

Bdkrb1 seems to affect CNS inflammation and control the migration of proinflammatory T cells across the BBB into the CNS. Indeed, analysis of immune cells recovered from the CNS of mice with EAE revealed that after adoptive transfer of Bdkrb1^−/− into Rag1^−/− mice, the proportion of CD4⁺ T cells within the CNS-infiltrating immune cells was significantly higher than that in mice that had received only WT T cells (Fig. 4a). Notably, within the same CNS-derived CD4⁺ T cell population, the proportion of IL-17⁺ CD4⁺ T lymphocytes—regarded as crucial for the initiation and maintenance of autoimmune neuroinflammation²²—was greater in Bdkrb1^−/− mice with EAE, whereas the proportion of IFN-γ–producing T cells was comparable, relative to those in WT control mice with EAE (Fig. 4b,c).

To investigate whether Bdkrb1 influences the development of IL-17–producing T helper type 17 (T_H17) cells (ref. 23) or the migration of T_H17 lymphocytes to the CNS, we used ovalbumin-transgenic C57BL/6 OT-2 mice and PLP-specific T cells isolated from EAE mice. The proportion of OT-2- and PLP-specific CD4⁺ T_H17 lymphocytes generated with IL-23, IL-6, anti–IL-4, anti–IL-12 and TGF-β (ref. 24) did not differ in the presence or absence of the Bdkrb1 agonist R838 (Supplementary Fig. 3), demonstrating that Bdkrb1 has no impact on the generation of antigen-specific T_H17 lymphocytes. However, in the CNS of Bdkrb1^−/−Tcrb^−/− → WT mice with EAE, in which the T cell population was deficient for Bdkrb1, the proportion of IL-17⁺ subsets was markedly greater than among cells isolated from WT Tcrb^−/− → WT mice with EAE. We did not observe any differences in the proportions of either IFN-γ–producing T cells or FoxP3⁺ CD25⁺ T cells (Fig. 4d–f).

Using in vitro migration assays, we next investigated the effect of R838 and R715 on the migration of mouse and human memory CD45RO⁺ T_H17 and T_H1 lymphocytes (Supplementary Fig. 4)²⁵. Addition of R838 resulted in a significant reduction in the migration of T_H17, but not T_H1, lymphocytes toward a CXCL12 chemokine gradient for mouse T cells and across human brain-derived microvascular endothelial cells for human T cells (Fig. 4g and Supplementary Fig. 5). These data also point to the important contribution of kinin receptor B1 as a modulator of the recruitment of pathogenic lymphocytes to the CNS. To find a possible explanation for the specific effect on T_H17 cells, we analyzed the expression of Bdkrb1 in T_H1 and T_H17 cells generated from different transgenic mouse strains (2d2, OT-2) as well as from human sources. In fact, T_H17 cells showed a markedly higher expression of Bdkrb1 than did T_H1 cells (Fig. 4h and Supplementary Fig. 5).

To finally demonstrate an influence of Bdkrb1 signaling on the migration pattern of T_H17 cells, we treated fluorescence-labeled mouse T_H17 lymphocytes with the Bdkrb1 modulators or vehicle before allowing them to infiltrate into syngeneic hippocampal slice cultures²⁶. Two-photon microscopy analysis revealed a lower mean velocity (Supplementary Fig. 6) and reduced infiltrative behavior upon Bdkrb1 activation (Fig. 4i,j and Supplementary Movies 1, 2,3).

Altogether, our data suggest the existence of a hitherto unknown endogenous control mechanism that limits harmful antigen-specific immune responses targeting the CNS, and they define the kinin receptor B1 as an important regulator for the homing of encephalitogenic T lymphocytes into the CNS. Progress toward the development of new therapies for chronic inflammation is urgently needed^{27, 28}. In principle, activation of the body's innate control mechanisms, such as those identified here, may offer certain advantages over previous strategies aimed at selectively blocking structures involved in inflammatory CNS infiltration²⁹. Newly developed kinin receptor agonists with improved pharmacological properties in regard to half-life and receptor specificity may provide promising new tools for the therapeutic manipulation of the kallikrein-kinin system in immune-mediated diseases. For the related renin-angiotensin system, blockade by an angiotensin-converting enzyme inhibitor suppresses inflammation in the CNS (L.S., personal communication). Thus, modification of major systems known for their cardiovascular roles may open the way toward new therapies for chronic inflammatory diseases such as multiple sclerosis.

Online methods

EAE

We induced active EAE (in C57BL/6 and SJL/J) and adoptive transfer EAE (in SJL/J) as previously described^{4, 5, 6}. Bdkrb1-deficient mice on the SV129 background were backcrossed to C57BL/6 to produce F₁₀ offspring. For pharmacological modulation, we used the Bdkrb1 agonist R838 or antagonist R715 (Biosynthan). For adoptive transfer EAE in C57BL/6 Rag1^−/− mice, we immunized C57BL/6 (WT) and Bdkrb1^−/− mice using 200 µg MOG_35–55 and 200 ng pertussis toxin. Twelve days later, cells from draining lymph nodes and splenocytes were depleted of CD8⁺ T cells (MACS, Miltenyi), re-stimulated for 72 h (15 µg ml⁻¹ MOG_35–55), and intravenously injected into naive C57BL/6 Rag1^−/− recipients (1 × 10⁷ cells per mouse). All animal experiments were conducted according to protocols approved by the local Canadian and German animal welfare committees.

Generation of bone marrow chimeras

We generated conventional and mixed bone marrow chimeras as described previously^{17, 18, 19}. Recipient mice were lethally irradiated with 1,100 cGy (split dose) and reconstituted with 12 × 10⁶ donor bone marrow cells devoid of CD90⁺ T cells (MACS). To track encephalitogenic Bdkrb1^−/− and WT T cells into the CNS, cells were followed after disease induction and the proportion of CNS-derived T cells was analyzed by FACS. We quantified WT and Bdkrb1^−/− T cells recovered from the CNS by using flow cytometry for CD45.1 and CD45.2. We checked reconstitution by analyzing donor-derived peripheral blood mononuclear cells for CD45.2 expression (95% purity). We determined the loss of Bdkrb1 in T cells from mixed bone marrow chimeras (Bdkrb1^−/−Tcrb^−/− → C57BL/6-CD45.1) by PCR (25 cycles) of magnetically sorted CD3⁺ cells (95% purity Dynal CD3 Sort) from spleens. We analyzed Bdkrb1 expression on non-CD3⁺ cells by magnetic enrichment of CD11b⁺ cells (95% purity, Miltenyi).

Generation of T_H17 and T_H1 cells

Mouse cells

CD4⁺CD62L⁺ naive T cells were magnetically sorted from OT-2 mice and stimulated with ovalbumin (0.3 µM OVA_323–339; Pepceuticals) with irradiated antigen-presenting cells (APC) at a 1:5 ratio. T_H17 differentiation was achieved by addition of 3 ng ml⁻¹ TGF-β, 20 ng ml⁻¹ IL-23, 20 ng ml⁻¹ IL-6, 5 µg ml⁻¹ anti–IL-12 (C17.8) and 5 µg ml⁻¹ anti–IL-4 (11B11). For T_H1 cells, 10 ng ml⁻¹ IL-12 and 5 µg ml⁻¹ anti–IL-4 (11B11) were added. Cells were kept in medium supplemented with rhIL-2 (Chiron) and re-stimulated every 7 d. R838 (500 nM) was added to the culture after 7 d and refreshed after medium exchange. After 14 d we checked cytokine production with PMA and ionomycin stimulation. PLP-specific T cells isolated from EAE mice were stimulated with PLP_139–151 (12.5 µg ml⁻¹; Pepceuticals) in the presence of irradiated APCs. T_H17 differentiation was achieved with IL-23 (20 ng ml⁻¹), 10 µg ml⁻¹ anti–IL-12 (C17.8) and 10 µg ml⁻¹ anti–IL-4 (11B11). R838 (500 nM) was added to the culture and refreshed after medium exchange. After 7 d, cytokine production was checked with PMA and ionomycin stimulation.

Human cells

CD4⁺ and CD4⁺CD45RO⁺ T lymphocytes (purity > 97%) were magnetically isolated from the peripheral blood mononuclear cells of healthy human donors, who had given written informed consent (Centre Hospitalier de l'Université de Montréal, ethical approval experimental design number HD04.046) (ref. 25). T cells were cultured with autologous monocytes as APCs in a 2:1 ratio and stimulated with CD3-specific antibody (2.5 µg ml⁻¹; clone OKT3, eBioscience). For T_H1 polarization, we added recombinant hIL-12 (10 ng ml⁻¹) and anti-hIL-4 (5 µg ml⁻¹; clone 3007, R&D), whereas for T_H17 polarization, we cultured T cells with recombinant hIL-23 (10 ng ml⁻¹) and neutralizing antibodies against IFN-γ (5 µg ml⁻¹; clone K3.53, R&D) and IL-4 (R&D). We harvested cells on day 6 for cytokine determination using commercially available ELISA kits for IFN-γ (Becton Dickinson), IL-17 (Biosource) and IL-22 (R&D).

Migration assay

Mouse cells

We performed chemotaxis experiments with PLP_139–151- and MOG_35–55-specific T cells using laminin-coated, 6.5 mm Transwells (Costar) with a confluent monolayer of bEnd3 cells²⁰. Except for in vitro–differentiated T_H1 and T_H17 cells, T cells were pretreated for 24 h with 70 U ml⁻¹ TNF-α and IFN-γ to induce Bdkrb1 expression, and challenged with either R838, R715 or both (500 nM of each) for an additional 3 h. We induced migration by adding 200 ng ml⁻¹ recombinant CXCL12 (R&D) to the lower chamber.

Human cells

BBB endothelial cells were isolated from CNS tissue specimens obtained from temporal lobe resections from young adults undergoing surgery for the treatment of intractable epilepsy, as described previously^{10, 25}. Informed consent and ethical approval were obtained before surgery (Centre Hospitalier de l'Université de Montréal, approval HD04.046). Human T cell migration was assessed using a 24-well-plate modified Boyden chamber^{3, 25}. T_H1 and T_H17 lymphocytes were challenged with R838 or R715 (500 nM each) for 3 h. One million cells per condition were loaded in the upper chamber, and the absolute number of cells that transmigrated to the lower chamber was counted after 18 h.

Statistics

Data were analyzed with SPSS and presented with Prism 4 (GraphPad).