astrocyte-induced regulatory t cells mitigate cns autoimmunity

Astrocyte-Induced Regulatory T CellsMitigate CNS Autoimmunity

VLADIMIR TRAJKOVIC,1 OLIVERA VUCKOVIC,1 STANISLAVA STOSIC-GRUJICIC,2DJORDJE MILJKOVIC,2 DUSAN POPADIC,1 MILOS MARKOVIC,1

VLADIMIR BUMBASIREVIC,3 ALEKSANDAR BACKOVIC,4 IVANA CVETKOVIC,2LJUBICA HARHAJI,2 ZORICA RAMIC,1 AND MARIJA MOSTARICA STOJKOVIC1*

1Institute of Microbiology and Immunology, School of Medicine, University of Belgrade,Belgrade, Serbia and Montenegro

2Institute for Biological Research Sinisa Stankovic, Belgrade, Serbia and Montenegro3Institute of Histology, School of Medicine, University of Belgrade,

Belgrade, Serbia and Montenegro4Institute for Medical Research, Military Medical Academy, Belgrade, Serbia and Montenegro

KEY WORDS experimental autoimmune encephalomyelitis; multiple sclerosis; im-munoregulation

ABSTRACT Although astrocytes presumably participate in maintaining the im-mune privilege of the central nervous system (CNS), the mechanisms behind theirimmunoregulatory properties are still largely undefined. In this study, we describe thedevelopment of regulatory T cells upon contact with astrocytes. Rat T cells pre-incubatedwith astrocytes completely lost the ability to proliferate in response to mitogenic stimuli.The cells were blocked in G0/G1 phase of the cell cycle, expressed less IL-2R, andproduced significantly lower amounts of interferon-� (IFN-�), but not interleukin-2(IL-2), IL-10, or tumor necrosis factor (TNF). These anergic cells completely preventedmitogen-induced growth of normal T lymphocytes, as well as CNS antigen-driven pro-liferation of autoreactive T cells. The suppressive activity resided in both CD4� andCD8� T-cell compartments. Heat-sensitive soluble T-cell factors, not including trans-forming growth factor-� (TGF-�) or IL-10, were solely responsible for the observedsuppression, as well as for the transfer of suppressive activity to normal T cells. Theadministration of astrocyte-induced regulatory T cells markedly alleviated CNS inflam-mation and clinical symptoms of CNS autoimmunity in rats with experimental allergicencephalomyelitis. Finally, the cells with suppressive properties were readily generatedfrom human lymphocytes after contact with astrocytes. Taken together, these dataindicate that astrocyte-induced regulatory T cells might represent an important mech-anism for self-limitation of excessive inflammation in the brain. © 2004 Wiley-Liss, Inc.

INTRODUCTION

Multiple sclerosis (MS) is a demyelinating autoim-mune disease of the central nervous system (CNS),representing the most common cause of neurologicaldisability in young adults (Compston and Coles, 2002).In most cases, MS has a typical relapsing-remittingcourse with exacerbations followed by periods of remis-sion but, over time, most patients develop progressiveneurological deficit. The best-defined experimentalmodel with significant clinical and pathological simi-larities to MS is experimental autoimmune encephalo-myelitis (EAE), which is induced by immunization of

susceptible animals with spinal cord homogenate ordifferent myelin antigens (Zamvil and Steinman,1990). The pathology of EAE is characterized by lym-

V. Trajkovic and O. Vuckovic contributed equally to this work.

Grant sponsor: Ministry of Science, Technology and Development of the Re-public of Serbia; Grant number: 1664; Grant number: 2020.

*Correspondence to: Marija Mostarica Stojkovic, Institute of Microbiology andImmunology, School of Medicine, Dr. Subotica 1, Belgrade 11000, Serbia andMontenegro. E-mail: [email protected]

Received 20 November 2003; Accepted 11 February 2004

DOI 10.1002/glia.20046

Published online 30 April 2004 in Wiley InterScience (www.interscience.wiley.com).

GLIA 47:168–179 (2004)

© 2004 Wiley-Liss, Inc.

phocyte infiltration of the CNS, an increase in blood-brain barrier permeability and demyelination, all ofwhich contribute to the flaccid paralysis as the mostcommonly observed clinical expression of disease.

According to the generally accepted concept of patho-genesis in both EAE and MS, encephalitogenic proteinsor cross-reactive antigens prime peripheral autoreac-tive CD4� T cells that migrate to the CNS and, uponreactivation by target antigen, secrete proinflamma-tory cytokines and produce neurological signs (Comp-ston and Coles, 2002). Eventually, the immune re-sponse wanes and remission follows the exacerbationin MS, as well as recovery follows a clinical disease inEAE. This clearly suggests that some downregulatorymechanism might prevent the perpetuation of the im-mune response within the CNS. Some recent data sug-gest that astrocytes, which are most abundant glial celltype, could be involved in the downmodulation of T-cellautoreactivity in the CNS. Indeed, astrocytes havebeen shown to inhibit microglial production of the es-sential Th1 cytokine IL-12 (Aloisi et al., 1997), as wellas to produce a wide array of immunosuppressive mol-ecules, such as prostaglandin E2 or transforminggrowth factor-� (TGF-�) (Constam et al., 1992; Aloisi etal., 2000). Furthermore, astrocytes inhibit mitogen-in-duced proliferation of rat T cells (Miljkovic et al., 2002)and suppress growth and interferon-� (IFN-�) produc-tion of myelin basic protein (MBP)-specific humanCD4� T-cell lines in vitro (Meinl et al., 1994). Thissuppressive activity was mediated by astrocyte-derivedsoluble factor(s), rather than through direct cellularcontact (Meinl et al., 1994). However, the effect couldnot be ascribed to either anti-inflammatory cytokines(IL-4, IL-10, TGF-�) or nitric oxide (NO), while prosta-glandins were only partially involved (Meinl et al.,1994; Miljkovic et al., 2002). Finally, the contact withastrocytes was found to trigger the apoptotic death of Tcells in vitro and in vivo, probably through a CD95ligand-dependent mechanism (Gold et al., 1996;Pender and Rist, 2001; Bechmann et al., 2002).

We present for the first time evidence for the astro-cyte ability to induce regulatory T cells with immuno-suppressive capacity. The regulatory T cells induced incontact with astrocytes completely prevented mitogen-induced T-cell growth, as well as antigen-driven prolif-eration of CNS-autoreactive T lymphocytes. Moreover,their potential therapeutic use was suggested by mark-edly reduced clinical and pathological signs of EAE, anaccepted animal model of CNS autoimmunity.

MATERIALS AND METHODSCells and Cell Cultures

Astrocytes were isolated from mixed glial cell cul-tures prepared from brains of newborn Dark Agouti(DA) rats as previously described (McCarthy and deVellis, 1980). Cells were maintained in HEPES-buff-ered RPMI-1640 medium (Sigma-Aldrich, Deisenhofen,Germany) with 5% fetal calf serum (FCS), 2 mM L-

glutamine, antibiotics, and sodium pyruvate (culturemedium) supplemented with 4 g/L glucose. After 10days of cultivation, the cultures were shaken mechan-ically to dislodge microglia and oligodendrocytes, whileastrocytes were further purified by repetition oftrypsinization (0.25% trypsin and 0.02% EDTA) andreplating. The cells used in the experiments describedin the present report were obtained after a third tofourth passage, when they were more than 97% posi-tive for glial fibrillary acidic protein (GFAP), an astro-cyte-specific intermediate filament component, andless than 1% positive for microglial surface moleculeCD11b (data not shown). The rat astrocytoma C6 cellline and human astrocytoma cell line U251 were main-tained in culture medium until used for the experi-ments. Resident peritoneal macrophages (Trajkovic etal., 1997), lymph node cells (LNC) (Stosic-Grujicic etal., 2002), spleen fibroblasts and heart vascular endo-thelial cells (Miljkovic et al., 2003) were isolated aspreviously described. Human peripheral blood mono-nuclear cells (PBMC) were obtained from healthy vol-unteers by centrifugation on lymphocyte separationmedium (ICN, Costa Mesa, CA). Following preincuba-tion with astrocytes or in culture medium in a 60-mmPetri dish or 24-mm Transwell system (Corning, Corn-ing, NY), LNC or PBMC were collected, washed exten-sively, and used in experiments as described in thefigure legends. Astrocyte contamination of lymphocytesobtained in a described way was consistently lowerthan 1%, as confirmed by GFAP staining. For someexperiments, lymphocytes were incubated under thesame conditions with C6 or U251 astrocytoma cells,syngeneic macrophages, fibroblasts, or endothelialcells, as described in Figure legends. For proliferation,cytokine production, cell cycle analysis, and determi-nation of CD25 expression, LNC or PBMC were stim-ulated in 96-well plates (Nunc, Roskilde, Denmark)with 2.5 �g/ml of concanavalin A (Con A) or 500 nMPMA � 100 ng/ml ionomycin (all from Sigma) for 48 h.Cycloheximide, emetine dihydrochloride, NG-methyl-L-arginine, indomethacin, �-methyl-DL-tryptophan,L-tryptophan (all from Sigma), neutralizing anti-rat-IL-10 (goat IgG) and pan-specific anti-TGF-� (rabbitIgG) (both from R&D Systems, Minneapolis, MN), orappropriate control antibodies against c-Jun (goat IgG)and c-Fos (rabbit IgG) (both Santa Cruz Biotechnology,Santa Cruz, CA) were used as indicated in the figurelegends. The IL-2-responsive T-cell blasts were pre-pared from Con A-stimulated LNC by density-gradientseparation, as previously described (Simic et al., 1986).LNC apoptosis was induced by a 5-min pulse of ultra-violet (UV) irradiation, delivered by an UV lamp (250nm) at a distance of 10 cm. Purified CD3�, CD4�, andCD8� lymph node cells were positively selected bymagnetic separation, using appropriate mouse anti-ratantibodies (�-CD3 from PharMingen, Heidelberg, Ger-many; �-CD4 and �-CD8, both from Holland Biotech-nology, Leiden, The Netherlands) and CELLection™Pan Mouse IgG Dynabeads� (Dynal A.S., Oslo, Nor-way), according to the manufacturer’s instructions.

169ASTROCYTE-INDUCED REGULATORY T CELLS

The purity of positively selected cell populations wasroutinely higher than 95%, as assessed by flow cytom-etry analysis.

Cell Proliferation and Cytokine Production

Cell proliferation was measured by incorporation of3H-thymidine (Sigma) into DNA. 3H-thymidine (1 �Ci/well) was added to cell cultures during last 16 h of 48 hincubation period, and its incorporation into cellularDNA, expressed as counts per minute (cpm), was de-termined in a scintillation counter. Concentration ofcytokines in cell culture supernatants was measuredby ELISA, using rat IFN-� ELISA kit (Biosource, Fleu-rus, Belgium) and paired anti-rat TNF-�, IL-2, or IL-10antibodies (R&D Systems) according to the manufac-turer’s instructions.

Flow Cytometry Analysis of CD25 Expressionand Cell Cycle

For CD25 analysis, cells were stained with anti-CD25 Ab (PharMingen, San Diego, CA) followed bysecondary FITC-labeled Ab (Sigma), as previously de-scribed (Badovinac et al., 2000). Cell cycle analysis wasperformed as described by determination of the DNAcontent using propidium iodide staining (Jonuleit etal., 2000). Briefly, after Con A stimulation, cells werewashed in phosphate-buffered saline (PBS) and fixed in70% ethanol for 30 min. Afterward, cells (1 � 106) weretreated with PBS containing 1 mg/ml RNase and 0.05mg/ml propidium iodide (both from Sigma) for 30 minat room temperature. Both CD25 expression and DNAcontent of propidium iodide-stained cells were ana-lyzed using Epics flow cytometer (Coulter).

Induction and Clinical Evaluation of EAE

For the induction of EAE, the encephalitogenic emul-sion was prepared by mixing equal amounts of ratspinal cord homogenate (50% w/v in saline) and com-plete Freund adjuvant (CFA) containing 1 mg/ml M.tuberculosis (ICN). Immunization with 100 �g of re-combinant myelin basic protein (MBP; Sigma) in CFAwas used for generation of MBP-specific T cells. DArats were immunized with 0.1 ml of the emulsion, givenas a single intradermal injection in the right footpad.The immunized rats were monitored daily in a blindfashion, using the following method for scoring theclinical signs: 0, no clinical sign; 1, flaccid tail; 2, hind-limb paresis; 3, bilateral hindlimb paralysis; 4, com-plete paralysis of the tail and hindlimbs often associ-ated with incontinence; and 5, moribund state or deathof an animal. Several parameters of disease were ex-amined: mean clinical score, the mean of clinical scoresfor all rats within a group on a given day; incidence, thenumber of rats within a group that developed a clinical

score of �1 in comparison with the starting number ofrats in that group; mean day of onset, the mean daythat affected rats within a group first developed clinicalsigns of disease; and cumulative disease index, the sumof the daily mean clinical scores for a group over a givennumber of days.

Histological Examination

For histological evidence of EAE, rats were sacrificedon day 21 postimmunization. Spinal cords were care-fully removed and fixed by immersion in 4% bufferedformaldehyde. The cervical, thoracic and lumbar seg-ments were routinely processed for embedding in par-affin. Transverse 5-�m-thick sections were cut with aReichert’s microtome and stained with hematoxylinand eosin (H&E). The severity of inflammation wasscored in a blind fashion from 0 to 4 as follows: 0, noinflammatory cells; 1, leptomeningeal and adjacentsubpial cell infiltration; 2, one to three perivascularand parenchymal infiltrates; 3, three to six infiltrates;and 4, more than six infiltrates.

Statistical Analysis

The statistical significance of the results was ana-lyzed by one way analysis of variance (ANOVA) fol-lowed by the Student-Newman-Keuls test, unless indi-cated otherwise.

RESULTSAstrocytes Induce Unresponsiveness

of Rat T Cells

In the first set of experiments, we found that a 24 hpreincubation of rat LNC with syngeneic primary as-trocytes completely abolished their proliferative re-sponse to subsequent stimulation with T-cell mitogenCon A or PMA � ionomicin (Fig. 1A,B). The time-dependence experiments had shown that even 1-h con-tact with astrocytes was sufficient to cause total unre-sponsiveness of T cells to Con A-mediated activation(2,632 � 60 cpm compared with the control prolifera-tion of 101,843 � 9,108 cpm; P � 0.01). A reversibleprotein synthesis inhibitor cycloheximide (5 �g/ml)failed to abolish the observed effect (4% of control pro-liferation) if present during 1 h of lymphocyte/astrocyteco-cultivation, indicating that protein synthesis wasnot required for the astrocyte action. The effect wasMHC-unrestricted, since comparable results were ob-tained with rat C6 astrocytoma cells (Fig. 1A), or whenMHC-mismatched combination of Albino Oxford lym-phocytes and DA astrocytes was used (35,199 � 1,579cpm and 920 � 81 cpm for Con A-stimulated controllymphocytes and lymphocytes preincubated with astro-cytes, respectively). On the other hand, macrophagescompletely failed to mimic astrocyte action, while the

170 TRAJKOVIC ET AL.

slight inhibition of T-cell proliferation (not significantat the level of P � 0.01) was seen after preincubationwith endothelial cells or fibroblasts (Fig. 1A). In accor-dance with the proliferation data, astrocyte-preincu-bated LNC stimulated with Con A showed cell cyclearrest in G0/G1 phase (Fig. 1C). Although the cellviability was similar in control LNC and in those pre-incubated with astrocytes (90%, according to trypanblue staining), an increased number of apoptotic cells(sub-G fraction) was observed in the latter cultures inthe absence of stimulation, and it further increasedfollowing activation with Con A (Fig. 1C). Con A-in-

duced expression of IL-2R� chain (CD25) of trimericIL-2R, which is necessary for optimal binding and ac-tion of T-cell growth factor IL-2, was markedly lower incells that were preincubated with astrocytes (Fig. 1D).However, the contact with astrocytes did not affect ConA-triggered IL-2 production in T cells, thus eliminatingthe possibility that their growth was reduced as a con-sequence of defective IL-2 release (Fig. 1E). Of the twoproinflammatory T-cell cytokines tested - TNF-� andIFN-�, the latter was released at considerably lowerlevel in cultures of astrocyte-preincubated cells, whilethe production of anti-inflammatory cytokine IL-10

Fig. 1. Astrocytes induce T-cellunresponsiveness. A,B: Lymphnode cells (LNC) (1 � 107) of DarkAgouti (DA) rats were preincu-bated for 24 h in culture mediumonly () or in the presence (�) of1 � 106 DA primary astrocytes, C6astrocytoma cells, macrophages,endothelial cells, or fibroblasts. Af-terward, LNC were collected andseeded in 96-well plates (1 � 105/well) with (A) concanavalin A (ConA) or (B) phorbol myristate acetate(PMA) � ionomycin for additional48 h. The data are expressed aspercentage of proliferation ob-tained with control cells (). Con-trol values in A were, from left toright: 114,413, 121,677, 57,697,67,179, and 41,575 cpm. Controlproliferation in B was 91,064 cpm.The incorporation of 3H-thymidinein the absence of stimulation was�1,000 cpm. C: Cell cycle was as-sessed by propidium iodide stain-ing in Con A-stimulated LNC thatwere preincubated for 24 h in theabsence (T control) or presence ofastrocytes (T astro). D: Expressionof CD25 in cells from (C) was de-termined by flow cytometry (meanfluorescence [MF]). In the absenceof Con A stimulation, the propor-tion of CD25� cells was �10% inboth groups. E: Cytokine accumu-lation in the supernatants of thecell cultures from (C) was mea-sured by enzyme-linked immu-nosorbent assay (ELISA). Theamount of cytokines in unstimu-lated cultures was below detectionlimit. F: Responder LNC (2.5 �106) were incubated in the upperchamber of a transwell system,while astrocytes (2.5 � 105) andLNC (2.5 � 106), alone or com-bined, were cultivated in the lowerchamber. After 48 h, responderLNC from upper chamber were col-lected, washed, and stimulated ornot (1 � 105 cells/well) with Con A.Subsequent proliferative responseof LNC (1 � 105/well) to Con A isshown. A–F: Results are presentedas mean �SD of triplicate mea-surements from the representativeof at least three separate experi-ments (A,B,F) or as mean �SDfrom three independent experi-ments (C,E); *P � 0.01.


Figure 2.

was not altered (Fig. 1E). Finally, we assessed the roleof cell-to-cell contact and soluble factors in the induc-tion of T-cell unresponsiveness by astrocytes. Separa-tion of astrocytes and LNC using the Transwell systemcompletely abolished the observed inhibitory effect, ar-guing against the involvement of astrocyte solubleproducts (Fig. 1F). However, the presence of LNC inthe astrocyte compartment of the transwell readily di-minished subsequent Con A-triggered proliferation ofresponder LNC in the upper chamber (Fig. 1F). Thissuggested that the effect was actually mediated bysoluble product(s), the release of which required thecontact between astrocytes and LNC.

Astrocytes Induce Regulatory T Cells WithSuppressive Properties

One possibility for the results obtained in the Trans-well experiments (Fig. 1F) was that contact with astro-cytes can induce in T cells the release of some solublemediator able to suppress T-cell growth. To examinesuch an assumption, LNC were removed after co-culti-vation with astrocytes and tested for their capacity toinhibit growth of Con A-stimulated fresh responders.Indeed, a strong dose-dependent suppression of T-cellproliferation was observed, with complete inhibitionachieved at approximately 1:100 suppressor/responderratio (Fig. 2A). A similar effect was obtained when LNCwere preincubated with C6 astrocytoma cells, whilemacrophages, endothelial cells and fibroblasts werecompletely unable to induce the suppressive activity inlymphocytes (Fig. 2B). While 24-h incubation with as-trocytes did not significantly affect the number of LNCexpressing CD3 (�70%), CD4 (�50%), CD8 (�20%), orCD25 (�5%) (not shown), the LNC that acquired sup-pressive activity were further characterized as CD4�

and CD8� T cells, since CD3�, CD4�, or CD8� cellsreadily blocked Con A-induced proliferation of freshLNC (Fig. 2C). The suppression was contact-indepen-

dent, as previously suggested in transwell experiments(Fig. 1F), and further confirmed by finding that para-formaldehyde-fixed astrocyte-induced regulatory cellswere unable to affect the proliferation of responder Tcells (Fig. 2D). Accordingly, conditioned medium fromastrocyte-preincubated LNC readily exerted the sup-pressive action that was lost after boiling (Fig. 2E), butnot after freezing the supernatant (not shown). Anextremely high effectiveness of astrocyte-induced reg-ulatory cells prompted us to investigate their ability totransfer the suppressive activity to the responder cells.Indeed, fresh LNC exposed to culture supernatants ofastrocyte-induced regulatory cells readily acquired theability to block Con A-induced T-cell proliferation com-pletely (Fig. 2F). It therefore appears that contact withastrocytes induces development of regulatory T cells,which then use soluble mediator(s) for both inhibitionof T-cell growth and for the transfer of the suppressiveactivity to normal T cells.

Mechanisms Responsible for the SuppressiveActivity of Astrocyte-Induced

Regulatory T Cells

As the production of major T-cell growth factor IL-2was not affected by the putative suppressive factor (Fig.1E), we sought to determine if the induction of unrespon-siveness to IL-2 could be a mechanism for the observedT-cell suppressive activity. Therefore, we tested the influ-ence of astrocyte-induced regulatory T cells on the IL-2-induced proliferation of T cells that were previously acti-vated with Con A. The proliferative response of the lattercells to stimulation with exogenous IL-2 was completelyabolished in the presence of astrocyte-induced regulatoryT cells (Fig. 3A), thus confirming that the loss of theresponsiveness to IL-2, rather than the block of its pro-duction, was responsible for the suppression of T-cellgrowth. The LNC undergoing UV light-induced apoptosis(�70% of apoptotic cells 16 h post-irradiation) failed tomimic the suppressive action of astrocyte-induced regu-latory T cells, while the induction of apoptosis markedlyreduced the suppressive capacity of the latter cells (Fig.3B). Therefore, the observed inhibitory effect was notmediated by the immunosuppressive factors releasedfrom apoptotic cells, but apparently required metaboli-cally active cells. The loss of the suppressive activity afterboiling the supernatant of astrocyte-preincubated LNCsuggested that the suppressive factor might be a protein.Indeed, the treatment with irreversible protein synthesisinhibitor emetine almost totally abolished the suppres-sive ability of astrocyte-preincubated LNC (Fig. 3C),without significantly altering their viability (not shown).However, the suppression was probably not mediated bywell-known T-cell-inhibitory cytokines TGF-� and IL-10(Kehrl et al., 1986; Taga et al., 1993). The specific neu-tralizing antibodies at concentration able to neutralize 40ng/ml of IL-10 or 0.25 ng/ml of TGF-�1 (manufacturer’sdata) significantly enhanced Con A-triggered T-cell pro-liferation, but completely failed to affect the suppressive

Fig. 2. Astrocytes induce regulatory T cells with suppressive prop-erties. A–F: Proliferative response of fresh responder lymph node cells(LNC) (1 � 105/well) to concanavalin A (Con A) is shown. The datafrom one of at least three separate experiments with similar resultsare presented as mean � SD of triplicate measurements (*P � 0.01).A: LNC obtained after 24-h incubation with astrocytes (T astro), or inthe medium only (T control), were washed and incubated at variousconcentrations with responder LNC. B: After preincubation in me-dium alone (), or with C6 astrocytoma cells, macrophages, fibro-blasts or endothelial cells (�), LNC were washed and incubated (1 �105/well) with responder LNC. C: Control cells from (a) or purified cellpopulations (�95% purity) obtained after magnetic sorting of LNCpre-incubated with astrocytes were added (2 � 102/well) to responderLNC. D: Various concentrations of conditioned medium from 48 hcultures of astrocyte-preincubated LNC (T astro) or control cells (Tcontrol) were added to responder LNC. Black bar represents theproliferation of responder LNC incubated with astrocyte-preincu-bated lymphocytes fixed with 1% paraformaldehyde. E: Supernatantsfrom (D) or culture medium were boiled for 5 min and added atconcentration of 25% to responder LNC. F: After 24-h incubation in25% supernatants from (D), LNC were washed three times in phos-phate-buffered saline (PBS) and tested for their ability to affect pro-liferation of responder LNC (suppressor/responder 1:10).


activity of astrocyte-induced regulatory T cells (Fig. 3D).We also tested the involvement of other mechanisms re-ported to interfere with T-cell growth, including NO(Liew, 1995) and prostaglandin release (Chouaib et al.,1985), as well as tryptophan consumption (Munn et al.,1999). Specific enzyme inhibitors were used to block NOand prostaglandin production, or tryptophan metabolism(NG-methyl-L-arginine, indomethacin or �-methyl-DL-tryptophan, respectively), while tryptophan was also ex-ogenously added to restore its possible deficit. However,although these treatments restored macrophage-inhib-ited T-cell growth to a certain extent (Fig. 3F), they werecompletely unable to neutralize the inhibitory effect ofastrocyte-induced regulatory T cells (Fig. 3E).

Downregulation of EAE by Astrocyte-InducedRegulatory T Cells

Next, we assessed the ability of astrocyte-induced reg-ulatory T cells to interfere with antigen-specific prolifer-ation of T lymphocytes from rats immunized with thewell-known CNS antigen myelin basic protein (MBP). Aspreviously observed in mitogen-stimulated LNC, thepresence of astrocyte-derived regulatory cells totally abol-ished the proliferative response of encephalitogenic Tlymphocytes to MBP (Fig. 4A). The T-cell response topurified protein derivative (PPD), a mycobacterial com-ponent of CFA, was also completely blocked by regulatorycells (Fig. 4A). The therapeutic efficacy of astrocyte-in-duced regulatory T cells in CNS autoimmunity wastested in EAE induced with rat spinal cord homogenate.Control animals that received 1 � 106 normal LNC de-veloped a characteristic impairment of motor function(Fig. 4B), which was quite similar to that observed uponadministration of spinal cord homogenate alone (data notshown). The severity of clinical signs, however, wasmarkedly reduced in animals that received intravenousinjection of astrocyte-induced suppressor cells, as judgedby significantly lower mean clinical score at the peak ofthe disease (days 11-16) (Fig. 4b), as well as by the cu-mulative disease index (mean �SD: 15.4 � 9.5 vs. 8.4 �4.2 in control and treatment group, respectively; P �0.05). The incidence of the disease was not significantlydifferent between two groups. Although the onset of thedisease was somewhat delayed in the treatment group(13.8 � 2.6 days) compared with control (11.5 � 1.5), thedifference did not reach statistical significance. Allevia-tion of EAE symptoms correlated well with the results ofpathohistological analysis. While massive perivascularinflammatory cell infiltration was observed in the CNS ofcontrol animals, only a few minor infiltrates were foundin the spinal cord sections of rats treated with astrocyte-induced regulatory cells (Fig. 4C,D).

Astrocytes Induce the Suppressor Activity inHuman Lymphocytes

Finally, to determine whether regulatory T cells withsuppressive activity can be induced in contact of hu-

man lymphocytes with astrocytes, PBMCs of healthydonors were incubated with cells of human astrocy-toma cell line U251. In accordance with the resultsobtained in rats, co-incubation of human lymphocyteswith U251 cells caused unresponsiveness to subse-quent Con A stimulation (Fig. 5A). This was probably aconsequence of the induction of T-cell suppressive ac-tivity, as judged by the ability of the cells preincubatedwith U251 astrocytes to inhibit Con A-triggered prolif-eration of fresh autologous PBMC completely (Fig. 5C).Interestingly, both T-cell unresponsiveness, as well asthe induction of T-cell suppressive activity were readilyobserved upon preincubation of human lymphocyteswith rat astrocytoma cell line C6 or rat primary astro-cytes (Fig. 5A–D).

DISCUSSION

In the present study, we propose thus far undisclosedmechanism for astrocyte-mediated immunoregulation,in which astrocytes trigger suppressor activity in Tcells. This astrocyte ability is apparently unique, sincemacrophages, endothelial cells or fibroblasts could notmimic it. Astrocyte-induced regulatory cells are ex-tremely potent inhibitors of T-cell proliferation andhave the ability to alleviate experimentally inducedCNS autoimmunity in rats. The cells with similar sup-pressive activity are readily obtained in humans, indi-cating their potential use in the treatment of autoim-mune disorders.

Astrocytes have been reported to block the prolifer-ation and/or IFN-� production of T cells in both contact-independent and contact-dependent fashion that partlyinvolves prostaglandin action (Meinl et al., 1994) andinduction of T-cell apoptosis (Gold et al., 1996), respec-tively. We have observed that preincubation with as-trocytes promotes apoptosis of T-cell mitogen-stimu-lated LNC, which conforms with the previous resultsabout astrocyte ability to prime rat T lymphocytes forantigen-triggered apoptotic cell death (Gold et al.,1996). The increased rate of apoptosis in astrocyte-preincubated T cells was observed even in the absenceof Con A stimulation, but such relatively modest dif-ference (see Fig. 1C) seems unlikely to account for thecomplete absence of their proliferative capacity. Ourdata, however, clearly demonstrate an additionalmechanism for astrocyte-mediated suppression of theT-cell response, in which the contact with some consti-tutively expressed molecule on astrocyte surface trig-gers in T cells the release of soluble mediator(s) thatcompletely block T-cell growth in an autocrine/para-crine manner. The cell cycle arrest in G0/G1 phase wasprobably related to defective T-cell response to IL-2, asindicated by unaltered release of this T-cell growthfactor and impaired IL-2-driven proliferation of T-cellblasts. Although relatively modest reduction of IL-2R�expression could contribute to, it clearly cannot besolely responsible for the observed growth arrest,


Fig. 3. Mechanisms responsible for the suppressive activity of as-trocyte-derived regulatory T cells. A: Concanavalin A (Con A)-inducedT-cell blasts (2 � 103/well) were stimulated with IL-2 (10 U/ml) in theabsence or presence of lymph node cells (LNC) obtained after 24-hincubation in medium only (T control) or with astrocytes (T astro). B:Fresh responder LNC (1 � 105/well) were stimulated with Con A inthe absence (control) or presence of LNC undergoing UV-inducedapoptosis (T apo; 1 � 105/well), astrocyte-induced regulatory cells (Tastro; 2 � 104/well), or T astro undergoing UV-induced apoptosis (Tastro-apo; 2 � 104/well). C: Con A-stimulated responder LNC (1 �105/well) were incubated alone, or with astrocyte-preincubated LNCthat were treated (T astro � emetine) or not (T astro) with proteinsynthesis inhibitor emetine (1 h at 37°C) after co-cultivation with

astrocytes. D: Con A-stimulated responder LNC (1 � 105/well) wereincubated with or without astrocyte-preincubated LNC (T astro; 2 �104/well), in the presence or absence of anti-TGF-� and anti-IL-10.The dashed line represents Con A-induced proliferation of responderLNC in the presence of both control antibodies. E: Con A-stimulatedresponder LNC (1 � 105/well) were incubated with 2 � 104/well ofnormal (T control) or astrocyte-preincubated LNC (T astro), or (F)macrophages (MF; 5 � 104/well), in the presence or absence of 500 �Mof NG-methyl-L-arginine (L-NMMA), indomethacin, �-methyl-DL-tryp-tophan (�-MDL-tr), or L-tryptophan. A–F: T-cell proliferation fromone of three (A,D,E,F) or two (B,C) independent experiments withsimilar results is presented as mean �SD of triplicate measurements(*P � 0.01).


Fig. 4. Downregulation of experimental autoimmune encephalomy-elitis (EAE) by astrocyte-induced regulatory T cells. A: Popliteallymph node cells (LNC) were collected 7 days after immunization withmyelin basic protein (MBP) in complete Freund adjuvant (CFA). Theresponse of LNC (2 � 105/well) to recombinant MBP (20 �g/ml) orpurified protein derivative (PPD) (10 �g/ml) was tested in the pres-ence of control (preincubated for 24 h in medium only) or astrocyte-preincubated LNC (1 � 105/well). B: EAE was induced and clinicalscore was assessed as described in Materials and Methods. The ex-periment involved 26 animals (two groups of 13) that received either

control or astrocyte-preincubated LNC (1 � 106 cells intravenously) 7days after immunization. The results are presented as mean � SEM(*P � 0.05). Similar results were obtained in another experiment. C:Representative lumbal sections of the spinal cord from the control andtreated animals were photographed at 200 � magnification (the ar-rows indicate cellular infiltrates). D: Higher magnification of an in-flammatory infiltrate from a control animal, containing mainly mono-nuclear cells. E: Extent of the inflammatory infiltration is presentedas the mean percentage of sections (25 sections per animal, 2 animalsper group) in each scoring category (*P � 0.05; Chi-square test).


which probably involved some defects downstream ofIL-2 receptor.

A plausible question is whether the increased apo-ptosis seen upon Con A stimulation was mainly acause, or rather a consequence of the LNC suppressiveactivity that was imparted by pre-incubation with as-trocytes. Our data support the latter assumption, asthe apoptotic T cells did not exert any suppressiveactivity, while that of astrocyte-preincubated T cellswas completely lost upon the induction of apoptosis.Therefore, some actively secreted T-cell product(s) in-duced by contact with astrocytes, rather than the im-munosuppressive factors released from dying cells, ex-erted the observed suppression. While our resultssuggest a protein nature of the putative heat-sensitivesoluble mediator of the T-cell suppressive activity, wecould not demonstrate the involvement of either IL-10or TGF-�, which are among most potent autocrine/paracrine inhibitors of T-cell proliferation. This find-ing, at least when IL-10 is concerned, is consistent withcomparable IL-10 release by astrocyte-derived regula-tory cells and control lymphocytes. With the productionof NO and prostaglandins, as well as tryptophan con-sumption ruled out as possible mediators of the ob-served T-cell unresponsiveness, the search for the mol-ecules responsible for this effect is currently under wayin our laboratory. Our unpublished preliminary results

obtained by using molecular weight sieves indicatethat the putative suppressive molecule is �12 kDa.

Subsets of both CD4� and CD8� T cells have beenimplicated in the suppression of T-cell responses indifferent experimental settings (Athanassakis andVassiliadis, 2002). The suppressive capacity of astro-cyte-induced regulatory T cells apparently was not re-stricted to a specific cell compartment, as it was sharedby both CD4� and CD8� T cells. The possible involve-ment of recently described “professional” regulatoryCD4� T cells that are characterized by co-expression ofCD25 (Shevach, 2002), was excluded by finding thatremoval of CD25� cells did not affect astrocyte induc-tion of suppressor activity in T cells (unpublished ob-servation). Moreover, while CD4�CD25� T cells re-quired TCR activation to exert their regulatoryproperties (Thornton and Shevach, 1998; Dieckmannet al., 2001; Jonuleit et al., 2001; Shevach, 2002), T-cellsuppressor activity in our study was readily triggeredby MHC-unrestricted contact with astrocytes, withoutany need for further activation. Furthermore, the invitro inhibitory action of CD4�CD25� regulatory cellswas mainly contact-dependent (Dieckmann et al.,2001; Jonuleit et al., 2001; Shevach, 2002; Thorntonand Shevach, 1998), unlike that of astrocyte-derivedregulatory cells, which was achieved solely by solublemediator(s). Interestingly, once it was induced, this

Fig. 5. Astrocytes induce sup-pressive activity in human lym-phocytes. Peripheral blood mono-nuclear cells (PBMC) (1 � 107) ofhealthy donors were incubated for24 h with 1 � 106 (A,C) humanU251 astrocytoma cells, rat C6 as-trocytoma cells, or (B,D) rat pri-mary astrocytes. (A,B) PBMCwere collected, washed, and stim-ulated (1 � 105/well) with con-canavalin A (Con A). C: After pre-incubation with U251 or C6 cells,PBMC (5 � 103/well) were addedto Con A-stimulated fresh autolo-gous PBMC (1 � 105/well). D: Var-ious concentrations of normalPBMC (T control) or PBMC pre-incubated with rat primary astro-cytes (T astro) were tested for theability to suppress Con A-inducedproliferation of responder PBMC(1 � 105/well). Results are pre-sented as (A,C) mean � SD valuesobtained with the cells from threedifferent donors, or (B,D) mean �SD from triplicate observations ob-tained with the cells from a singledonor (*P � 0.01).


inhibitory activity did not further require astrocytepresence, since it was apparently propagated among Tcells in a chain reaction-like manner. Such “infectious”suppression has also been attributed to CD4�CD25�

cells, but it was conveyed to conventional T cellsthrough contact-dependent mechanism (Jonuleit et al.,2002). The ability to transfer the suppressor activity tonormal T cells by soluble product(s) was probably re-sponsible for the surprisingly high efficiency of astro-cyte-derived regulatory cells, compared with relativelyweak potency of CD4�CD25� T cells (Dieckmann et al.,2001; Jonuleit et al., 2001). However, it is not clear atpresent whether the same T-cell-secreted molecule me-diated both the suppression and its transfer in ourexperiments.

The activated autoreactive T cells are the first toenter CNS in MS and its animal counterpart EAE.Subsequently, naive T cells are recruited at the site ofinflammation in an antigen-independent fashion andthey rapidly account for the vast majority of infiltratingcells (Cross et al., 1993). It has been proposed thatsome of these secondarily recruited T cells could beactivated in situ upon recognition of self antigens re-leased during the initial tissue damage, which mightresult in epitope spreading and perpetuation of theinflammatory response (Elson et al., 1995). However,because of interference by astrocyte-induced regulatorycells recruited from newly arrived T lymphocytes ofdifferent specificities, these CNS-autoreactive T cellsmight not develop destructive autoimmune capacity.Instead, they could acquire an anergic regulatory phe-notype characterized by the proliferation block and theshift of IFN-�/IL-10 balance toward the release of anti-inflammatory IL-10. Our preliminary observation thatT cells recovered from the spinal cord of animals withEAE can suppress the proliferation of fresh T lympho-cytes (unpublished data) suggest that astrocytes mightindeed induce regulatory T cells in vivo. While possibleinvolvement of astrocyte-induced regulatory cells inpreventing the perpetuation of CNS autoimmunity isintriguing, their therapeutic potential was clearly dem-onstrated in the EAE model. It should be noted that wecannot completely exclude the possibility that immu-nosuppressive properties of apoptotic cells might con-tribute to the beneficial effect of astrocyte-induced reg-ulatory T cells in EAE. However, a relatively smallastrocyte-induced increase in T-cell apoptosis, togetherwith our in vitro observations on the absence of thesuppressive activity of apoptotic cells suggest that sucha mechanism, if operative, was of relatively minor im-portance for the protection seen in our experimentalmodel of autoimmunity.

In conclusion, astrocyte-induced generation of regu-latory T cells might represent an important protectivemechanism partly responsible for self-limiting courseof EAE and remissions seen in MS patients. Impor-tantly, astrocyte-induced regulatory T cells are readilygenerated from human lymphocytes in vitro, and the

molecules involved in their development seem to behighly conserved between species. These findings indi-cate a potential for developing simple therapeuticstrategies for adoptive transfer of ex vivo-generatedregulatory cells in autoimmune disorders. As potentialtherapeutic usefulness of such astrocyte-triggered reg-ulatory T cells in human autoimmunity is obvious, it ison forthcoming studies to further elucidate the mech-anisms and characterize the molecule(s) involved inthis phenomenon.

ACKNOWLEDGMENTS

The authors thank Professor Miodrag Colic (Insti-tute for Medical Research, Military Medical Academy,Belgrade, Serbia and Montenegro) for permission touse the FACS equipment, as well as Drs. Janko Ni-kolich-Zugich (Oregon Health and Science University,Beaverton, Oregon) and Stanislav Vukmanovic (Mi-chael Heidelberger Division of Immunology, Depart-ment of Pathology, Kaplan Cancer Center, NYU Med-ical Center, New York) for helpful discussion.

REFERENCES

Aloisi F, Penna G, Cerase J, Iglesias BM, Adorini L. 1997. IL-12production by central nervous system microglia is inhibited byastrocytes. J Immunol 159:1604–1612.

Aloisi F, Ria F, Adorini, L. 2000. Regulation of T-cell responses byCNS antigen-presenting cells: different roles for microglia and as-trocytes. Immunol Today 21:141–147.

Athanassakis I, Vassiliadis S. 2002. T-regulatory cells: are we re-discovering T suppressors? Immunol Lett 84:179–183.

Badovinac V, Trajkovic V, Mostarica Stojkovic M. 2000. Nitric oxidepromotes growth and major histocompatibility complex-unre-stricted cytotoxicity of interleukin-2-activated rat lymphocytes.Scand J Immunol 52:62–70.

Bechmann I, Steiner B, Gimsa U, Mor G, Wolf S, Beyer M, Nitsch R,Zipp F. 2002. Astrocyte-induced T cell elimination is CD95 liganddependent. J Neuroimmunol 132:60–65.

Chouaib S, Welte K, Mertelsmann R, Dupont B. 1985. ProstaglandinE2 acts at two distinct pathways of T lymphocyte activation: inhi-bition of interleukin 2 production and down-regulation of trans-ferrin receptor expression. J Immunol 135:1172–1179.

Compston A, Coles A. 2002. Multiple sclerosis. Lancet 359:1221–1231.Constam DB, Philipp J, Malipiero UV, ten Dijke P, Schachner M,

Fontana, A. 1992. Differential expression of transforming growthfactor-beta 1, -beta 2, and -beta 3 by glioblastoma cells, astrocytes,and microglia. J Immunol 148:1404–1410.

Cross AH, O’Mara T, Raine CS. 1993. Chronologic localization ofmyelin-reactive cells in the lesions of relapsing EAE: implicationsfor the study of multiple sclerosis. Neurology 43:1028–1033.

Dieckmann D, Plottner H, Berchtold S, Berger S, Schuler G. 2001. Exvivo isolation and characterization of CD4�CD25� T cells withregulatory properties from human blood. J Exp Med 193:1303–1310.

Elson CJ, Barker RN, Thompson SJ, Williams NA. 1995. Immunolog-ically ignorant autoreactive T-cells, epitope spreading and reper-toire limitation. Immunol Today 16:71–76.

Gold R, Schmied M, Tontsch U, Hartung HP, Wekerle H, Toyka KV,Lassmann H. 1996. Antigen presentation by astrocytes primes rat Tlymphocytes for apoptotic cell death. A model for T-cell apoptosis invivo. Brain 119:651–659.

Jonuleit T, van der Kuip H, Miething C, Michels H, Hallek M, DuysterJ, Aulitzky WE. 2000. Bcr-Abl kinase down-regulates cyclin-depen-dent kinase inhibitor p27 in human and murine cell lines. Blood96:1933–1939.

Jonuleit H, Schmitt E, Stassen M, Tuettenberg A, Knop J, Enk AH.2001. Identification and functional characterization of human


CD4�CD25� T cells with regulatory properties isolated from pe-ripheral blood. J Exp Med 193:1285–1294.

Jonuleit H, Schmitt E, Kakirman H, Stassen M, Knop J, Enk AH.2002. Infectious tolerance: human CD25� regulatory T cells conveysuppressor activity to conventional CD4� T helper cells. J Exp Med196:255–260.

Kehrl JH, Wakefield LM, Roberts AB, Jakowlew S, Alvarez-Mon M,Derynck R, Sporn MB, Fauci AS. 1986. Production of transforminggrowth factor-� by human T lymphocytes and its potential role inthe regulation of T cell growth. J Exp Med 163:1037–1050.

Liew FY. 1995. Regulation of lymphocyte functions by nitric oxide.Curr Opin Immunol 7:396–399.

McCarthy KD, de Vellis J. 1980. Preparation of separate astroglialand oligodendroglial cell cultures from rat cerebral tissue. J CellBiol 85:890–902.

Meinl E, Aloisi F, Ertl B, Weber F, de Waal Malefyt R, Wekerle H,Hohlfeld R. 1994. Multiple sclerosis: immunomodulatory effects ofhuman astrocytes on T cells. Brain 117:1323–1332.

Miljkovic D, Samardzic T, Cvetkovic I, Mostarica Stojkovic M, Trajk-ovic V. 2002. Mycophenolic acid downregulates inducible nitric ox-ide synthase induction in astrocytes. Glia 39:247–255.

Miljkovic DJ, Cvetkovic I, Vuckovic O, Stosic-Grujicic S, MostaricaStojkovic M, Trajkovic V. 2003. The role of interleukin-17 in induc-ible nitric oxide synthase-mediated nitric oxide production in endo-thelial cells. Cell Mol Life Sci 60:518–525.

Munn DH, Shafizadeh E, Attwood JT, Bondarev I, Pashine A, MellorAL. 1999. Inhibition of T cell proliferation by macrophage trypto-phan catabolism. J Exp Med 189:1363–1372.

Pender MP, Rist MJ. 2001. Apoptosis of inflammatory cells in immunecontrol of the nervous system: role of glia. Glia 36:137–144.

Shevach EM. 2002. CD4�CD25� suppressor T cells: more questionsthan answers. Nat Rev Immunol 2:389–400.

Simic MM, Ramic Z, Nikolic V. 1986. Selective effects of human cordserum on distinct events of the lectin-induced activation and pro-liferation of thymic T cells. Int J Immunopharmacol 8:763–771.

Stosic-Grujicic S, Savic-Radojevic A, Maksimovic-Ivanic D, MarkovicM, Bumbasirevic V, Ramic Z, Mostarica Stojkovic M. 2002. Down-regulation of experimental allergic encephalomyelitis in DA rats bytiazofurin. J Neuroimmunol 130:66–77.

Taga K, Mostowski H, Tosat G. 1993. Human interleukin-10 candirectly inhibit T-cell growth. Blood 81:2964–2971.

Thornton AM, Shevach EM. 1998. CD4�CD25� immunoregulatory Tcells suppress polyclonal T cell activation in vitro by inhibitinginterleukin 2 production. J Exp Med 188:287–296.

Trajkovic V, Badovinac V, Popadic D, Hadzic O, Stojkovic MM. 1997.Cell-specific effects of pentoxifylline on nitric oxide production andinducible nitric oxide synthase mRNA expression. Immunology 92:402–406.

Zamvil SS, Steinman L. 1990. The T lymphocyte in experimentalallergic encephalomyelitis. Annu Rev Immunol 8:579–621.


astrocyte-induced regulatory t cells mitigate cns autoimmunity

Documents