Treatment of ongoing autoimmune encephalomyelitis with activated B-cell progenitors maturing into regulatory B cells

The influence of signals perceived by immature B cells during their development in bone marrow on their subsequent functions as mature cells are poorly defined. Here, we show that bone marrow cells transiently stimulated in vivo or in vitro through the Toll-like receptor 9 generate proB cells (CpG-proBs) that interrupt experimental autoimmune encephalomyelitis (EAE) when transferred at the onset of clinical symptoms. Protection requires differentiation of CpG-proBs into mature B cells that home to reactive lymph nodes, where they trap T cells by releasing the CCR7 ligand, CCL19, and to inflamed central nervous system, where they locally limit immunopathogenesis through interleukin-10 production, thereby cooperatively inhibiting ongoing EAE. These data demonstrate that a transient inflammation at the environment, where proB cells develop, is sufficient to confer regulatory functions onto their mature B-cell progeny. In addition, these properties of CpG-proBs open interesting perspectives for cell therapy of autoimmune diseases.

B lymphocytes exert complex functions in autoimmune diseases. On the one hand they can promote these diseases, as shown by the beneficial effects of B-cell depletion therapies in rheumatoid arthritis or multiple sclerosis (MS) [1][2][3] . On the other hand, their negative regulatory functions can provide protection, as initially shown in models of ulcerative colitis 4 , experimental autoimmune encephalomyelitis (EAE) 5 and collagen-induced arthritis 6 . More precisely, mice with an interleukin (IL)-10 deficiency restricted to B cells developed a severe chronic form of EAE, while those harbouring wild-type (WT) B cells rapidly recovered from disease 5 . The unique capacity of B cells to reduce the severity of autoimmune diseases through provision of IL-10 has kindled enormous interest in the identification of the responsible B-cell sub-populations, and the signals controlling their expression of suppressive functions. Several B-cell subsets can produce IL-10 on stimulation in vitro, and protect recipient mice from autoimmune disease on adoptive transfer. These include B10 cells with a CD5 þ CD1d high phenotype, transitional type 2 (T2)-marginal zone precursors, and CD5 þ B-1 cells (refs 7-9). Moreover, analysis of the cells producing IL-10 in a suppressive manner in vivo identified CD138 hi plasma cells residing either in spleen 10 or LN 11 as major IL-10 producers during EAE. In addition, IL-35 (ref. 10) and PD-L1 (ref. 12) were recently shown to mediate protection against EAE displayed by B regulatory cells.
Toll-like receptor (TLR) agonists are particularly important in this context because of their unique capacity to induce IL-10 expression in mature naive B cells, and the requirement for intrinsic TLR signalling in B cells for recovery from EAE 13 . Similarly, CD5 þ CD1d high B cells depend on activation by TLR-4 or -9 agonists to produce IL-10 in vitro 14 . In addition to intrinsic TLR signalling, signals provided to B cells through the BCR or CD40 are also necessary to achieve a fully protective B-cell-mediated regulation 5,[15][16][17][18][19] .
Human B cells can also produce IL-10, and evidence is accumulating that they can subsequently behave as negative regulators of immunity. Duddy et al. 20 found that B cells from MS patients produce lower amounts of IL-10 on activation than their counterparts from healthy individuals, suggesting that MS is facilitated by a defect in this regulatory circuit. This defective IL-10 production was subsequently confirmed by Correale and Farez 21 who observed that helminth infections could restore normal IL-10 production by B cells in this disease, which correlated with an improvement of the disease course compared with non-infected patients. These data suggest that promoting B-cell-mediated regulation might help to reduce MS progression. Of note, some MS treatments, such as interferon (IFN)-b 22 or glatiramer acetate 23,24 have been shown to enhance IL-10 secretion by B cells.
Although our knowledge of the peripheral B-cell subsets implicated in immune regulation, and the signals controlling their regulatory activity has greatly improved, we still have little data on whether B-cell development in bone marrow (BM) might influence the subsequent capacity of mature B cells to negatively regulate immunity in periphery. We recently gained some evidence of this by showing that proB cells emerging from BM cell cultures transiently stimulated with the TLR-9 agonist CpG prevented type 1 diabetes in NOD mice on adoptive transfer 25 . These cells, which we termed as CpG-proB cells, differentiated into various more mature B-cell subsets in recipient mice. We could demonstrate that they provided protection by reducing pathogenic IL-21 secretion by T cells, and by promoting apoptosis of T-effector cells 25 . However, it was difficult to further analyse how the administered CpG-proBs protected recipient mice from disease in this model due to the lack of mutant mouse strains on the NOD background. We therefore addressed the question, whether a similar subset emerged in the C57BL/6 mouse strain. To this end, we used the myelin oligodendrocyte glycoprotein (MOG)  -induced EAE model in which B-cell-mediated regulation has been extensively studied. So far described regulatory B-cell subsets only acted at the initiation phase of the disease, except PD-L1 hi mature splenic B cells of which 1 million cells injected at day 7 after immunization provided protection against EAE 12 , albeit milder than when injected at day 0. However, we find that a single injection of only 60,000 CpG-proB cells at the onset of EAE clinical signs markedly reduces disease progression in recipient mice. In contrast, adoptive transfer of control non-stimulated proBs does not influence the disease course in recipient mice. This suppression requires the differentiation of the administered CpG-proB cells into mature B cells in the periphery and in the central nervous system (CNS). Collectively, these results shed light on the importance of the BCR-independent signals that B-cell progenitors receive in the BM environment to develop into mature B cells with regulatory rather than inflammatory functions. In addition, these properties of CpG-proBs open interesting perspectives for cell therapy of autoimmune diseases.

Results
ProB cells exposed to CpG protect against ongoing EAE. We previously found that on short-term incubation with CpG-1668 (CpG-B), BM cells from NOD mice gave rise to a population of pro-B cells that protected recipient mice from diabetes on adoptive transfer 25 . To assess whether a similar process took place in a strain of mice which are more amenable to genetic studies, we evaluated the effect of CpG-B on BM cells from C57BL/6J mice. This treatment led to emergence of cells expressing a c-kit þ Sca-1 þ B220 þ IgM À phenotype, and heterogeneous in terms of PDCA-1 expression ( Fig. 1a and Supplementary Fig. 1). A similar BM population was observed in vivo in mice after i.p injection of CpG-B, validating the use of in vitro cultures ( Supplementary Fig. 2). The bright B220 þ cells are gated out since they correspond to the more mature B cells contaminating the c-kit þ magnetically sorted cells. Moreover, since TLR-9 stimulation has been shown to promote deviation of hematopoiesis away from the B-cell lineage towards the PDCA-1 þ plasmacytoid dendritic cell lineage 26 , B-cell precursors were further sorted by excluding the PDCA-1 þ fraction (Fig. 1a). The resulting PDCA-1 À population was closely related to the pro-B cell stage of differentiation, being CD19 þ CD24 þ IgM À CD11b À CD11c À , as well as expressing the IL-7Ra chain (CD127), CD43 and the transcription factor Pax5 ( Fig. 1b and Supplementary  Fig. 3a) characterizing B-cell lineage commitment. They all expressed CD1d, but were negative for CD5 (Fig. 1b). It is noteworthy that this effect was not restricted to TLR-9 agonists, because agonists of TLR-2, -4, -5, -6 and -7 induced development of a similar population, unlike agonists of TLR-1 and -3 (Fig. 1c). As expected, these cells did not appear in BM cell cultures from MyD88-deficient mice after incubation with CpG-B (Fig. 1c). Collectively, these data suggest that TLR agonists induce in vitro and in vivo the formation of a unique population of proB cells in BM from C57BL/6 mice, as previously found in NOD mice 25 .
We next examined whether these cells could protect recipient mice from EAE on adoptive transfer. Remarkably, a single injection of only 60,000 CpG-proBs (Fig. 1d)    CD24 and CD43, from fresh non-stimulated BM, as c-kit þ Sca1 À B220 þ CD24 hi CD43 hi cells ( Fig. 1e and Supplementary  Fig. 3b), nor cells detected within BM culture with PBS over 18 h exhibiting the same phenotype as CpG-proB cells but at a 6-10-fold less frequency, termed PBS-proBs, had any effect in recipient mice (Fig. 1e). Therefore transient in vitro and in vivo TLR-9-dependent activation within the BM confers protective properties onto proB cells against ongoing EAE. A dose-response study ( Supplementary Fig. 4) showed that injection of 60,000 CpG-proBs was optimal for protection against EAE. Neither 25,000 nor 90,000 injected cells provided significant protection, the latter possibly resulting from cell aggregation or, alternatively, from hindrance in access of cells into the hematopoietic niche of non-irradiated recipients.
CpG-proB cells modify immune cell response and distribution.
To characterize the protective effect of the injected progenitors on encephalitogenic T cells, we analysed the expression of proinflammatory cytokines by CD4 þ T cells in treated mice at the peak of the disease (d18-d21). In spinal cord, which is a major target organ in EAE, mice treated with CpG-proB cells had reduced numbers of CD4 þ T cells producing all pathogenic cytokines tested, namely GM-CSF, IL-17 and IFN-g (Fig. 2a,c). In LN, production of GM-CSF by CD4 þ T cells was reduced, whereas expression of IL-17 and IFN-g were not affected in treated mice compared with controls ( Fig. 2b,c). In addition, production of the anti-inflammatory cytokine IL-10 by CD4 þ T and B220 þ B cells was enhanced in treated mice in both tissues ( Fig. 2d-g), which was paralleled by an increased frequency of CD4 þ Foxp3 þ regulatory T cells (Fig. 2h,i), showing a modestly enhanced Foxp3 expression (Fig. 2j).
To further delineate how the transferred CpG-proBs protected from EAE, we analysed immune cell distribution in recipient mice, in comparison with control mice with EAE. Remarkably, mice treated with CpG-proB cells displayed increased cell numbers and particularly CD4 þ T cells in LN, while having markedly fewer total and CD4 þ T cells in spinal cord compared with controls (Fig. 2k).
In sum, these data show that the administered B cells impact on the disease process both at the level of the CNS and LN.
CpG-proB cells migrate and differentiate into mature Bregs. To trace the migration and follow the differentiation of the administered CpG-proB immature cells, we injected 60,000 CpG-proB cells prepared from BM of congenic CD45.1 mice into CD45.2 recipient mice at day 12 after EAE induction (Fig. 3a). On day 9 after transfer, CD45.1 þ cells were mainly found in reactive, cervical, axillary and inguinal, but not mesenteric, lymph nodes (LNs; 113,900 ± 3,500 cells, mean ± s.e.m.), and spinal cord (27,880±3,218 cells), with only few cells present in spleen (5,000 ± 1,098 cells; Fig. 3b). It is noteworthy that the total number of donor-derived cells was about threefold higher in recipients than the number of injected cells, indicating that they expanded in vivo. Using different B-cell and other lineage markers, we assessed that the transferred CpG-proBs differentiated exclusively into the B-cell lineage (Fig. 3c,d). In LN, the recovered CD45.1 þ cells displayed a B220 þ CD19 þ IgM þ IgD þ IgG À CD1d þ CD21 þ CD23 hi phenotype typical of transitional B cells (Fig. 3c). This population was, however, heterogeneous, comprising about 70% of CD11b þ CD11c À , and 20% of CD5 þ cells (Fig. 3c). In spinal cord, donor-derived B cells were comparable to their LN counterparts in their B220 þ CD19 þ IgM þ IgD þ IgG À phenotype ( Fig. 3d) but with lower levels of CD21 and CD23 than in LNs and exclusively CD1d þ CD5 þ CD11b þ , therefore similar to B10 cells. While the CD45.1 þ B-cell progeny of injected progenitors represented only 0.77±0.14% of B cells in the LNs, they represented nearly eightfold more, that is, 5.97 ± 1.32% of B cells, in the spinal cord (Fig. 3e,f). Remarkably, CD45.1 þ B cells produced the anti-inflammatory cytokine IL-10 in spinal cord, but not in LN (Fig. 3g).
These data demonstrate that the transferred CpG-proB cells give rise to more mature B cells that accumulate in reactivecervical and inguinal-LN and inflamed CNS during EAE.
To assess whether such differentiation of CpG-proB cells into more mature B cells was required for protection against EAE, CpG-proB cells from CD45.1 þ Rag2 À / À mice (prepared as in Supplementary Fig. 5) whose differentiation into mature B lymphocytes is blocked, were adoptively transferred. They had no effect on the disease course (Fig. 4a). Injected cells from CD45.1 þ Rag2 À / À donors could not be found in LNs or spinal cord at 3, 5 or 7 days after injection, in contrast to CpG-proB cells from WT donors, suggesting that their reduced lifespan as progenitors hampered them to confer protection against EAE. Therefore, transient treatment of BM cells with TLR-9 agonist generates proB cells that differentiate into more mature B cells with protective function in EAE. Remarkably, CpG-proB cells from Rag2 À / À mice had no effect on cell numbers in LN and CNS (Fig. 4b), suggesting that this alteration of cell distribution directly contributed to the beneficial effect afforded by CpG-proB cells.
Roles of IL-10 production by donor-derived B cells. Since IL-10 is generally a key mediator of B-cell-mediated regulation, we assessed its importance for the suppressive function of CpG-proB cells. IL-10 was necessary for the protective function of CpG-proB cells, because mice treated with IL-10-deficient CpG-proBs (prepared as in Supplementary Fig. 5) were not protected against disease (Fig. 4c), and production of pathogenic cytokines by CD4 þ T cells in spinal cord was not decreased (Fig. 4d,e). However, IL-10-deficient CpG-proBs still promoted CD4 þ T cell accumulation in the reactive LN and reduced their numbers in spinal cord (Fig. 4f) suggesting the involvement of an additional mechanism in this latter effect.
Altogether, these data show that an efficient control of disease by CpG-proBs requires IL-10 production by their mature B-cell progeny. IL-10 produced by these cells is important to reduce inflammation locally in spinal cord, while trapping of immune cells in reactive LN is achieved via an IL-10-independent mechanism.

Role of IFN-c in LN entrapment of T cells by CpG-proBs.
To identify the IL-10-independent mechanism involved in the accumulation of immune cells in the reactive LN of CpG-proB cells-treated mice, we analysed cytokine production by these cells (Fig. 5a,b) in a comprehensive manner, because B cells primarily inhibit immunity through cytokine production 5,10 . Remarkably, freshly prepared CpG-proB cells did not produce IL-10 ( Fig. 5a,b) or IL-35 (Fig. 5b), but produced abundant amounts of IFN-g, with levels about 30 times higher than those detected for the other cytokines, following stimulation for 5 h with PMA plus ionomycin (Fig. 5a). Expression of IFN-g by these cells was also confirmed by flow cytometry (Fig. 5b) and was not significantly different in WT and IL-10-deficient CpG-proBs ( Supplementary  Fig. 6). Moreover, the WT CD45.1 þ B-cell progeny recovered from reactive LN and from the spinal cord still produced IFN-g, although less in spinal cord than in LN, as measured by qRT-PCR analysis (Fig. 5c) and intracellular flow cytometry ( Supplementary Fig. 7). Since IFN-g has protective functions in EAE 27,28 , this prompted us to investigate whether IFN-g played a role in the protective function of CpG-proB cells against EAE.
We tested the role of IFN-g using CpG-proBs prepared from IFN-g-deficient mice. IFN-g-deficient CpG-proB cells did not protect recipient mice against EAE on adoptive transfer (Fig. 5d). Moreover, although the cells derived from IFN-g-deficient CpG-proBs actually reached the spinal cord, where they released as much IL-10 as the WT progeny ( Supplementary Fig. 7), they also produced there fivefold more GM-CSF than the WT progeny. They failed to reduce the encephalitogenic cytokines production by CD4 þ T cells and to enhance IL-10-producing CD4 þ and B220 þ cells in spinal cord ( Supplementary Fig. 8). Similarly, they failed to drive an accumulation of CD4 þ T cells in reactive LN of recipient mice, in contrast to WT CpG-proBs (Fig. 5e). This suggested to us that IFN-g produced by donor-derived B cells contributed to protection from disease not only by limiting encephalitogenic cytokine production within the spinal cord, but also by retaining encephalitogenic T cells in reactive LN, thereby preventing their accumulation in the CNS.
To characterize how IFN-g produced by donor-derived B cells controlled T-cell accumulation in reactive LN, we examined its effect on expression by T cells of the chemokine receptor CCR7 known to control T-cell localization within LN 29 , and to be implicated in neuro-inflammation 30,31 . CCR7 expression was reduced in LN CD4 þ and CD8 þ T cells of the CpG-proBtreated group relative to controls (Fig. 5f,g). Remarkably, CpG-proBs derived from IFN-g-deficient mice, barely affected CCR7 staining on CD4 þ T cells (Fig. 5f,g), contrasting with those derived from IL-10-deficient mice that were as efficient as WT CpG-proBs at retaining CD4 þ T cells within LNs ( Supplementary Fig. 9). When we performed a short acid wash (0.2 M acetic acid, 0.5 mM NaCl), known to eliminate bound ligands from their receptor 32 , before CCR7 staining, the reduction in CCR7 staining in LN T cells from CpG-proB recipients relative to controls was no longer observed (Fig. 5h,i). This suggested that increased local chemokine concentrations in LNs of CpG-proB recipients bound to CCR7 and prevented antibody binding,   Role of CCL19 in T-cell entrapment and disease protection. The protection afforded by B-cell-derived IFN-g identifies a new mechanism of B-cell-mediated regulation in EAE. We sought to gain further insight into this mechanism. Given that IFN-g can modulate chemokine expression 27 , we evaluated whether CpG-proB cells and/or their progeny themselves might be a source of CCL19 and/or CCL21 via a mechanism controlled by IFN-g. To this end, we first quantified Ccl19 and Ccl21 mRNA in WT CpG-proBs, IFN-g-deficient CpG-proBs, and IFN-gdeficient CpG-proBs cultured in presence of IFN-g. We found that WT CpG-proB cells produced both CCL19 and CCL21 (Fig. 6a). The production of these chemokines was reduced in IFN-g-deficient CpG-proBs, and this defect could be erased by exogenous addition of IFN-g (Fig. 6a). Remarkably, in LN, the mature B-cell progeny generated from CpG-proBs in recipients expressed markedly higher levels of Ccl21, and even more Ccl19 than they do in spinal cord and both expressions were higher than in the originally administered CpG-proB cells (Fig. 6b). CCL19 has been shown to induce CCR7 internalization more efficiently than CCL21 (ref. 33). In addition, CD4 þ and CD8 þ LN T cells from mice that received WT but not Ccl19-deficient CpG-proBs displayed reduced cell surface expression of CCR7 (Fig. 6c,d).
Moreover, CpG-proBs sorted from Ccl19-deficient mice neither restrained immune cells in reactive LN (Fig. 6e), nor provided protection against EAE (Fig. 6f), confirming that CCL19 production by CpG-proB cells was essential for T-cell sequestration inside the LN and protection against disease. From these data, we conclude that CCL19 produced by CpG-proB-derived B  cells in response to autocrine IFN-g induces sequestration of T cells via CCR7 in the reactive LN, subsequently limiting their accumulation in the CNS and improving the disease course.

Discussion
In adults, immature B cells develop in BM and then seed secondary lymphoid organs, where they can acquire diverse effector functions depending on their mode of activation. Here, we show that signals perceived by immature B cells in a BCR-independent manner can determine their subsequent effector functions as mature B cells. Indeed, we demonstrate that a transient stimulation of BM cells via TLR-9 is sufficient to divert the differentiation of pro-B precursors towards suppressive B cells. In brief, adoptive transfer of as little as 60,000 pro-B cells recovered from BM cultures shortly stimulated with CpG-B or from the BM of CpG-injected mice markedly reduced the severity of EAE in recipient mice, while unstimulated pro-B cells had no effect even though they encountered the same inflammatory environment in recipient mice. Donor-derived mature B cells mediated this protection because the latter depended on RAG expression.
The protection from disease mediated by the transferred CpG-proB cells involved two complementary mechanisms, one limiting local inflammation in CNS through IL-10 production, and the other restraining encephalitogenic T cells in reactive LN. The latter was achieved through LN retention of T cells, which was mediated by increased CCL19 production by CpG-proBderived B cells under IFN-g autocrine control. None of these effects were observed with control unstimulated pro-B cells.   These data demonstrate that transient changes in the BM environment of pro-B cells can markedly influence their subsequent effector functions as mature B cells in periphery, favouring their anti-inflammatory activities.
The production of IL-10 is a key mechanism, by which B cells can suppress autoimmune diseases [4][5][6]11 . CpG-proBs did not produce IL-10 themselves, but their mature progeny became IL-10-competent strictly in spinal cord of recipient mice. Although CpG-proBs from IL-10-deficient mice normally retained immune cells in reactive LN, this was insufficient to provide protection against EAE since at the onset of clinical signs, pathogenic cells have already reached the CNS and triggered neuro-inflammation. We propose that IL-10 produced by donor-derived B cells is particularly important in the CNS in our model. In addition, the prevalence of IL-10 over GM-CSF production in the spinal cord by the progeny of CpG-proBs may be essential to control neuro-inflammation, as shown in Supplementary Fig. 7. The enhanced production of IL-10 by host CD4 þ T cells and B220 þ B cells in CNS of mice treated with WT CpG-proB might further help to control local inflammation in the target organ. In addition, mice treated with WT CpG-proB cells displayed an increased expansion of Tregs, suggesting that host Tregs are also involved in the protective effect of the transferred B cells, in line with results from other studies 15,19,34,35 .
An important finding of this study is the identification of a novel mechanism of B-cell-mediated immune regulation involving IFN-g production by B cells. Looking for other non IL-10 possible mediators for the protective effects of CpG-proBs against EAE, we found that both CpG-proBs and their LN but not spinal cord progeny displayed a massive production of IFN-g, a cytokine considered pro-inflammatory but that has also been credited with possible regulatory functions. In EAE, both IFNg À / À and IFNgR À / À mice suffer from a more severe disease in both chronic and Remitting-Relapsing-EAE models 27,28 . The capacity to produce IFN-g has been reported for B cells at different levels of maturation and in various activation conditions. Thus BM immature B cells constitutively produce IFN-g at low levels [36][37][38] , whereas mature follicular and marginal zone B cells produce massive amounts of IFN-g on TLR-activation that condition the Th1 response to Salmonella enterica infection 39 . Co-culture with Th1 cells emerging in pathogen-infected animals give rise to so-called Be1 effector B cells producing large amounts of IFN-g that subsequently polarize T cells towards a type 1 immune response 40   in control mice (n ¼ 10) or mice injected either with WT (n ¼ 10) or CCL19-deficient CpG-proBs (n ¼ 10). ***Po0.001, between controls or CCL19 À / À and WT CpG-proB recipients, non significant between controls and CCL19 À / À CpG-proB-treated mice, using the two-way repeated measures ANOVA test.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms12134 ARTICLE infections, produced as much IFN-g as NK cells when re-stimulated ex vivo with anti-CD40. Their provision of IFN-g was required for protection against L. Monocytogenes infection. Conversely, Olalekan et al. 42 reported that IFN-g production by B cells was essential for the development of autoimmune experimental arthritis. Therefore, IFN-g production by CpG-proBs as well as other described B cells have functional, either protective or deleterious, roles and may link innate and adaptive immune responses.
Since PD-L1 hi mature splenic B cells were recently shown to provide protection against EAE 12 and PD-L1 can be induced by IFN-g, we investigated the possible role of PD-L1 in CpG-proBinduced protection. There was no correlation between the level of PD-L1 expression on WT versus IL-10 and IFN-g-deficient CpG-proBs and their relative protective capacity against EAE ( Supplementary Fig. 10a). Furthermore, CpG-proBs isolated from PD-L1-deficient C57BL/6 donor mice showed only a slight (non significant) reduction in their protective effect against EAE ( Supplementary Fig. 10b). Therefore, PD-L1 is not playing a significant role in disease protection against EAE by CpG-proBs.
Our data indicate that the protective effect mediated by B-cell-derived IFN-g operates at least in part by sequestering CD4 þ T cells in the reactive LN. Many presently available treatments of MS aim at preventing T cells from entering the CNS. Several chemokine receptors and their ligands have been shown to take part in the process of neuroinflammation, including CCR7 whose ligands CCL19 and CCL21 are produced by endothelial cells at the inflamed BBB in mice and men 43,44 . CCR7 À / À or plt/plt mice that lack CCL19 and CCL21 are resistant to EAE 45 , demonstrating the requirement of CCL19/CCR7 interactions for the development of EAE. When expressed in peripheral tissues, CCL19 and CCL21 are able to mediate cell recruitment in vivo of both naive and recently activated T cells 46 . CpG-proB recipients showed downregulated staining for CCR7 at the surface of both CD4 þ and CD8 þ LN T cells. We found that IFN-g affected CCR7 expression on T cells indirectly by enhancing the production by CpG-proB cells of both CCR7 ligands, CCL19 and CCL21. Once in the LN, the B-cell progeny produced both chemokines at far higher rates than the progenitors they derived from. The observation that similar levels of cell surface CCR7 were recovered once ligands were eliminated by an acid wash suggested that high-receptor occupancy by CCL19 or CCL21 was taking place in T cells from CpG-proB recipients, anchoring T cells within LN. Britschgi et al. 33 have demonstrated that CCR7 levels and occupancy reflect the amount of CCL19 present in the environment and that high levels of CCL19 reduce CCR7 staining by the same antibody we used. Similarly, increased levels of the CCR7 ligands in inflamed tissues have been reported to reduce CCR7 detection and to induce locally T-cell retention 47 . Thus, high occupancy of CCR7 by its ligands and particularly CCL19 occurs in LN of CpG-proB recipients and impedes leucocyte migration. The particular anchoring role of CCL19 produced by the B-cell progeny was confirmed by the findings that CpG-proB cells isolated from CCL19-deficient mice did not reduce surface CCR7 expression measured without acidic washing procedure and concomitantly lost their capacity to sequester T cells within reactive LN, and to protect against EAE disease. Importantly, the markedly (3,000-fold) enhanced production of CCL19 in LN rather than spinal cord progeny of CpG-proBs strongly supports a major role of this protective mechanism in the reactive LN to keep T cells away from the CNS. Notably, we did not observe significant reduction of S1P1R expression of LN-trapped T cells in CpG-proB recipients relative to control mice with EAE. Therefore, different from FTY720 (ref. 48) which targets the LN egress signal, CpG-proBs promote entrapment of pathogenic T cells inside secondary lymphoid organs by targeting instead the retention signal.
In summary (Fig. 7), the powerful protective effect of CpG-proBs against EAE is due to the capacity of their progeny to anchor T cells within the LNs particularly by releasing CCL19, thereby retaining them away from the spinal cord. This process is indispensable, but not sufficient to confer protection against EAE, at disease onset when T cells have already migrated into the CNS. The B220 þ CD11b þ CD1d þ CD5 þ B-cell progeny of injected CpG-proBs must further release IL-10 within the spinal cord, which is ultimately required for switching the host cytokine profile from inflammatory to immunoregulatory. These molecular mechanisms were obtained using progenitors in which each individual mediator was deficient at a time. BM chimera exhibiting concomitant deficiencies in all the mentioned molecular mediators may be necessary to provide full demonstration of their roles in the protection afforded by CpG-proBs. CpG-proBs are, to our knowledge, the most efficient B-cell subset so far reported to be able, at only 60,000 cells per recipient, to stably inhibit ongoing encephalomyelitis. These properties of CpG-proBs therefore open interesting perspectives for cell therapy of autoimmune diseases. In addition, these data shed light on the importance of the BCR-independent signals that B-cell progenitors receive in the BM environment to develop into mature B cells with regulatory rather than inflammatory functions.

Methods
Mice. WT CD45.2 þ C57BL/6J mice were obtained from Janvier Laboratories (Le Genest Saint Isle, France). Congenic CD45.1 þ , MyD88 À / À , IFN-g À / À , Rag2 À / À C57BL/6J, all backcrossed for at least ten generations, were bred in our animal facility under specific pathogen-free conditions. CCL19 À / À C57BL/6J mice were raised at the Department of Biochemistry at the University of Lausanne, IL-10 À / À C57BL/6J mice at the DRFZ, Berlin and PD-L1 À / À C57BL/6J mice 49 at Trinity Biomedical Sciences Institute, Trinity College Dublin. Knock-out animals and their WT controls from the same origin were systematically used as donor mice for CpG-proBs. Female 12-week-old C57BL/6J mice were used as a model for MOG  -induced EAE and received intravenous progenitor cell transfers at day-12 post-immunization. EAE induction. Active EAE was induced in 12-week-old female mice by s.c. immunization at two sites, upper and lower back, with 200 mg MOG 35-55 peptide emulsified in CFA containing 400 mg heat-killed Mycobacterium tuberculosis H37Ra (Hooke Laboratories, Lawrence, MA, USA), on day 0. In addition, mice received 200 ng pertussis toxin (Hooke Laboratories) i.p. in 0.1 ml per mouse on days 0 and 1. Clinical signs of EAE were assessed daily with a 0-to 5-point scoring system defined as follows: 0, no obvious changes in motor function compared with non-immunized mice; 0.5, tip of tail is limp; 1, limp tail; 1.5, limp tail and hind leg inhibition; 2, limp tail and weakness of hind legs; 2.5, limp tail and dragging of hind legs; 3, limp tail and complete paralysis of hind legs or paralysis of one front and one hind leg; 3.5, limp tail and complete paralysis of hind legs that are together on one side of body; 4, limp tail, complete hind leg and partial front leg paralysis; mouse is still minimally moving and appears feeding; 4.5, complete hind leg and partial front leg paralysis; no movement around the cage, mouse is not alert; 5, death or severe paralysis. Mice with score Z4 for two consecutive days and mice with score 5 were killed.
Isolation of immune cells from the spinal cord. Spinal cords isolated from control and CpG-proB-recipient mice were incubated for 30 min in digestion buffer of DNAse and liberase (27 WU ml À 1 ) in PBS 1 Â at 37°C, mixing every 5 min. EDTA (100 mM, 500 ml) was added for 1 min to end the digestion. The cells were passed through a cell strainer, using a syringe plunger (back side) to smash the tissue. Cells were then resuspended in 3-5 ml of 40% Percoll underlayed with the same volume of 70% Percoll (in PBS) and centrifuged for 35 min at 1,300 g (2,800 r.p.m.) without brakes to form a smooth interface. Cells were collected with a Pasteur pipette and diluted 10 times with complete medium RPMI 10% SVF. After centrifugation cells were resuspended in 2-3 ml of complete medium and further stained with appropriately labelled mAbs.
Cytokine assays. Control and CpG-ProB-recipient mice were killed at the peak of the disease, 18-21 days after immunization. Cytokines were measured in supernatants of recovered LN cells from control or CpG-ProB-recipient mice, re-stimulated with the MOG 35-55 peptide for 3 days followed by 5-h stimulation with PMA þ ionomycin. For determination of spinal cord cytokines, isolation of spinal cord cells was followed by incubation at 37°C/5% CO 2 in the presence of Golgi-stop ( þ PMA/ionomycin) for 3 h. Th1/Th2/Th17 cytokines and GM-CSF were measured by multiplex ELISA using Flow Cytomix analyte detection reagents from eBioscience.
Study approval. All mouse procedures were approved by the Paris Descartes University Animal Experimentation and Ethics Committee. Sample sizes were chosen to ensure reproducibility of the experiments and according to the 3Rs of animal ethics regulation.
Statistics. Non-parametric Mann-Whitney's t-test was used to compare values between the two groups. Disease curves were analysed using two-way repeated measures ANOVA test.
Data availability. All relevant data are available from the authors.