Proc. Natl. Acad. Sci. USA Vol. 96, pp. 622–627, January 1999

Costimulatory signals are required for induction of transcription factor Nur77 during negative selection of CD4؉CD8؉ thymocytes

DERK AMSEN*, CONCHI REVILLA CALVO*†,BARBARA A. OSBORNE‡, AND ADA M. KRUISBEEK*§

*Division of Immunology, The Cancer Institute, Plesmanlaan 121, 1066 CX Amsterdam, The Netherlands; and ‡Department of Veterinary and Animal Sciences, University of Massachusetts, 304 Paige Lab, Amherst, MA 01003

Edited by Philippa Marrack, National Jewish Medical and Research Center, Denver, CO, and approved November 19, 1998 (received for review September 25, 1998)

ABSTRACT A major question in end-stage devel- presenting cells (APCs) to induce either positive or negative opment is how T cell receptor(TCR) ligation on immature selection may reflect in part their localization in different CD4؉CD8؉ double positive thymocytes is translated into spatial and temporal compartments. In addition, variables such either survival (positive selection) or apoptotic (negative as MHC expression levels and expression of adhesion mole- selection) signals. Because different types of antigen- cules may also contribute to the outcome of the interaction presenting cells (APCs) induce positive or negative selection between DP thymocytes and thymic APCs. Finally, the dif- in the thymus and express different costimulatory molecules, ferential ability of thymic presenting cells to induce positive or involvement of such costimulatory molecules in determining negative selection may reflect cell type-specific expression of cell fate of DP thymocytes is considered here. If TCR- costimulatory molecules. Such molecules may, through en- generated signals are modulated by APCs, this should be gagement of their counter-receptors on DP thymocytes, qual- reflected in the activation of distinct biochemical pathways. itatively influence the signals induced by the TCR, leading to We here demonstrate that costimulatory signals involved in induction of either apoptotic or survival pathways. negative selection also are required for induction of protein In support of a contribution of cell-derived costimulatory expression of Nur77 and its family members. These transcrip- signals, clonal deletion of thymocytes was indeed shown to tion factors are critically involved in negative but not positive occur only if, next to signals through the TCR, additional selection. In contrast, the signals that costimulate negative APC-derived signals are provided (11–14). Whether thymo- selection are not required for induction of several molecular cytes also use the molecular interactions that mature T cells events associated with positive selection. These include acti- exploit to generate costimulatory signals has been controver- vation of the immediate early gene Egr-1, the mitogen- sial. On the one hand, several investigators have reported a activated protein kinase ERK2, and surface expression of the lack of a phenotype with respect to negative selection in CD69 marker. Thus, costimulation for negative selection CD28Ϫ͞Ϫ mice (15–17). On the other hand, apoptosis of selectively provides signals for activation of apoptotic medi- thymocytes induced in suspension can be costimulated by ators. These data provide molecular insights into how TCR- crosslinking the CD28 receptor, either by using (12, engagement by ligands on different thymic APCs can deter- 18, 19) or by its natural ligands CD80 and CD86 (13, 14, 20). mine cell fate. Moreover, these findings have been corroborated in the intact thymus in fetal thymic organ culture and recently also in The random process that creates T cell receptor (TCR) CD28-deficient mice (13, 21), providing formal support for a diversity generates a large pool of thymocytes, of which only a role for CD28 in clonal deletion in vivo. While acknowledging few will migrate to the periphery. Within the population of that other molecules than CD28 can costimulate deletion (refs. ϩ ϩ thymocytes that has arrived at the CD4 CD8 double positive 11–17; present study), these findings document that CD28– (DP) stage of development, positive and negative selection CD80͞86 interactions can serve to provide prototypical co- processes determine which cells are allowed to develop into stimulatory signals for thymocyte deletion. Supportive evi- mature CD4 or CD8 TCRab cells. Both of these selection dence for this notion stems from the observation that thymic processes depend on ligation of the TCR by peptide-loaded medullary expression of CD80 is correlated with preferential MHC molecules present in the thymus but result in diamet- deletion of self-reactive thymocytes in the medulla (22, 23). rically different outcomes—survival and maturation (positive Also, CD86 expression in the medullary region of the thymus selection) or apoptosis (negative selection) (reviewed in ref. 1). is substantially reduced in gp39-deficient mice (24), a pheno- The critical issue in understanding end-stage T cell devel- type correlated with reduced deletion of thymocytes specific opment is how TCR occupancy can be perceived in such for endogenously expressed self-antigens (24). markedly different ways by DP thymocytes. Whether positive Thus far, only two molecular signaling modules have spe- or negative selection is induced appears to depend in part on cifically been linked to either positive or negative selection. the affinity͞avidity of the interaction between thymocyte Activation of the mitogen-activated protein kinase (MAPK) TCRs and major histocompatibility complex (MHC)–peptide cascade is required only for positive selection: transgenically complexes on presenting cells in the thymus (2, 3). However, expressed dominant negative mutants of Ras, Raf1, and a different model of thymocyte selection contends that in fact MEK-1 each reduce positive selection (25–27), whereas neg- the type of thymic cell presenting MHC–peptide ligands to ative selection is unaffected by perturbations in the MAPK developing thymocytes dictates the outcome of selection. cascade. By contrast, a specific role in negative selection has Support for this model has been provided by many different experimental systems (4–10) and is difficult to ignore. This This paper was submitted directly (Track II) to the Proceedings office. differential ability of different types of thymic antigen- Abbreviations: EMSA, electrophoretic mobility-shift assay; APC, antigen-presenting cell; DP, double positive; MHC, major histocom- The publication costs of this article were defrayed in part by page charge patibility complex; GST, glutathione S-transferase; TCR, T cell re- ceptor; ECDI, 1-ethyl-3-(3Ј-dimethylaminopropyl) carbodiimide; payment. This article must therefore be hereby marked ‘‘advertisement’’ in MAPK, mitogen-activated protein kinase. accordance with 18 U.S.C. §1734 solely to indicate this fact. †Present address: Centro de Investigaciones Biolo´gicas, Madrid, Spain. PNAS is available online at www.pnas.org. §To whom reprint requests should be addressed. e-mail: [email protected].

622 Downloaded by guest on September 25, 2021 Immunology: Amsen et al. Proc. Natl. Acad. Sci. USA 96 (1999) 623

been documented for members of the Nur77 family of orphan purified and unfractionated thymocytes and yielded similar steroid receptors. Nur77 is expressed in an immediate early results. fashion in response to apoptotic TCR triggering in thymocytes In Vitro Deletion Assay. Thymocytes were brought into (28, 29), and transgenic expression of Nur77 or its relative single-cell suspension in complete medium. Thymocytes (2– Nor1 in thymocytes leads to massive apoptosis (30). Moreover, 4 ϫ 105) were cultured in 96-well flat-bottom plates overnight mice expressing a transgene that encodes dominant negative with 1 ϫ 105 APCs. After 14–18 hr, cells were harvested and Nur77 lacking its transactivation domain exhibit severely im- prepared for flow cytometry analysis. Fixation of DCEK cells paired ability to induce clonal deletion (31, 32), whereas was carried out as described (47). positive selection proceeds undisturbed. These findings docu- Flow Cytometry. Cells were resuspended in PBS containing ment that transactivation of as-yet-unidentified pro-apoptotic 0.5% BSA and 0.02% sodium azide and incubated with target genes by Nur77 and closely related factors is a prereq- antibodies on ice for 30 min. Incubation with biotin-labeled uisite step in thymocyte clonal deletion but not in positive anti-CD69 (PharMingen) was followed by incubation with selection. The precise identity of these pro-apoptotic target avidin–fluorescein isothiocyanate (FITC) (Sigma). For mea- genes is currently unclear. It has been suggested that Nur77 surement of deletion, thymocytes were (after surface staining ␮ ͞ functions upstream from the CD95 ligand–CD95 interaction. for CD8 expression) exposed to 1 g ml ethidium bromide ϫ 4 ϫ 5 Thus, the massive apoptosis occurring in thymuses from Nur77 (Sigma) as described (48). Cells (1 10 to 2.5 10 per transgenic mice was reported to be reduced by introduction of sample) were analyzed on a Becton Dickinson FACSCAN by the gld mutation (30). However, these results were recently using LYSIS 2 software. APCs were segregated electronically on contradicted by another study (33) in which such rescue was the basis of forward scatter/side scatter pattern. not detected. As expression of dominant negative Nur77 yields Stimulation of Cells and Separation from APCs. APC lines a much more pronounced block in negative selection (31, 32) were grown overnight in the presence of magnetic beads (Biomag, Cambridge, MA). After two rounds of magnetic than mutations in CD95 or its ligand (34), it is likely that the Ј major pathway(s) downstream from Nur77 are independent of selection, APCs were either fixed with 1-ethyl-3-(3 - CD95 and CD95 ligand. dimethylaminopropyl) carbodiimide (ECDI) and subjected to We report here that events associated with positive selec- another round of magnetic selection or left untreated. Subse- tion—such as induction of expression of Egr-1 (35, 36) and quently, APCs were prepulsed for two hr in 15-ml polystyrene tubes with peptide and͞or antibodies. Finally, 5 ϫ 106 thymo- CD69 (14, 37)—as well as activation of the MAPK ERK2 ϫ 6 (25–27) are not affected by costimulation. In sharp contrast, a cytes and 1 10 APCs were mixed in 24-well plates (Costar) strict dependence on costimulation was found for protein in a total volume of 1 ml of prewarmed Iscoves medium containing 10% fetal calf serum. Plates were immediately expression of the orphan steroid receptor Nur77 and for centrifuged for 1 min at 1,200 rpm in a Hettich (Tuttlingen, induction of total DNA-binding activity to the Nur77 binding Germany) Rotanta RP centrifuge and put at 37°C. After element. Given the critical dependence of negative selection incubation, 1 ml of ice-cold PBS was squirted into each well, on transactivation by Nur77 and related factors, these results and vigorous pipetting was used to harvest the cells. Thymo- provide a molecular explanation for the requirement for cytes were separated from the APCs by two rounds of magnetic costimulation during negative selection and further our un- selection at 4°C and contained typically Ͻ1% APC. derstanding of how thymic APCs can determine cell fate Western Blot. Thymocytes were lysed in radioimmunopre- during thymocyte development. cipitation assay (RIPA) buffer containing 10 ␮g͞ml leupeptin and 1 mM phenylmethylsulfonyl fluoride (PMSF). Cell equiv- MATERIALS AND METHODS alents (107 per lane) were loaded onto 7.5% SDS͞PAGE gels. Nitrocellulose membranes (Schleicher and Schuell) were Antibodies and Reagents. The following purified antibodies k blocked with PBS containing 10% nonfat milk powder and were used: 11–4-1 (anti-K ) (38), GL1 (anti-CD86) (39), 0.1% Tween 20 and probed with anti-Nur77 (49) (anti-Nur77 145.2C11 (anti-CD3) (40), 37.51 (anti-CD28) (41), goat anti- ϩ clone 12.14 is available through PharMingen, catalog no. hamster IgG (H L) (Pierce), and rabbit polyclonal IgG 13471A), followed by biotinylated goat anti-mouse IgG anti-ERK-2 (Santa Cruz Biotechnology). Production and pu- (Pierce) and streptavidin–horseradish peroxidase (Amer- rification of glutathione S-transferase (GST)–cJun (1–223) sham). Alternatively, membranes were probed with anti-Egr-1 was done as described (42). Myelin basic protein was pur- (Santa Cruz Biotechnology) followed by biotinylated donkey chased from Sigma. Pigeon cytochrome C peptide 88–104 anti-rabbit Ig (Amersham) and streptavidin–horseradish per- (p17) was synthesized at the peptide synthesis facility at the oxidase. reactivity was detected by using Amer- Netherlands Cancer Institute. sham’s Enhanced Chemiluminescence kit according to man- Cells and Mice. All cells were cultured in Iscove’s modified ufacturer’s instructions. ͞ Dulbecco’s medium (GIBCO BRL) supplemented with 10% Electrophoretic Mobility-Shift Assay (EMSA). Thymocytes ϫ Ϫ5 fetal calf serum (BioWhittaker), 2 10 M 2-mercaptoetha- isolated from coculture with APCs were lysed in a buffer ͞ ␮ ͞ nol (Merck), 100 units ml penicillin, and 100 g ml strepto- containing 150 mM NaCl, 10 mM Tris (pH 8.0), 1% Triton mycin at 37°C in a humidified incubator containing 5% CO2 in X-100, 10 ␮g͞ml leupeptin, and 1 mM PMSF. After clearance, ␣ ϫ ␤ air. Thymocytes from (2B4 2B4 )F1 mice (43) on the 129 protein concentrations were determined by using Bicinchonic MHC II knockout background (44) were used. Mice were acid (BCA) protein assay (Pierce). Subsequently, 20 ␮gof between 4 and 6 weeks of age and were bred under specific- protein was tested by using EMSA as described (50). For pathogen-free conditions. DCEKhi7 is a daughter cell line of competition experiments, 1 ␮l of unlabeled oligonucleotide DAP3.3C1 (45) transfected with I-Ek (J. Miller and R. H. was added to the mixture together with the labeled oligonu- Germain, unpublished data). LEC is a B cell lymphoma cleotide. derived from CBA͞Ht mice (46). ERK2 and JNK Kinase Assays. Purified DP thymocytes Thymocyte Purification. Thymocytes were purified by pan- were rested at 37°C for 5 hr and subsequently incubated in ning on petri dishes coated with 20 mg͞ml 53–6 (American medium with antibodies (10 ␮g͞ml) for 30 min on ice. Cells type Culture Collection) for 45 min at 4°C followed by 12 to 15 were washed twice with cold medium, and goat anti-hamster washes to remove unbound cells. After this, vigorous pipetting antibody was added at 10 ␮g͞ml in warm (37°C) medium. After ϩ ϩ was used to isolate CD4 CD8 thymocytes. Typically, these this, cells were incubated for the indicated times at 37°C. The cells Ͼ96% pure CD4ϩCD8ϩ as measured by using flow reactions were stopped by addition of ice-cold PBS. Cells were cytometry. All experiments have been performed both with lysed in whole-cell extract lysis buffer (51). Subsequently, 50 ␮g Downloaded by guest on September 25, 2021 624 Immunology: Amsen et al. Proc. Natl. Acad. Sci. USA 96 (1999)

of whole-cell lysate per sample was tested for Erk2 and JNK approach that would allow separation of the thymocytes from activities by in vitro kinase assays after immunoprecipitation of the APC. To achieve this, monoclonal transformed APC cell Erk2 or precipitation with GST–c-Jun (1–223) as described lines were first cultured overnight with magnetic beads (53). (51, 52). These beads are ingested by the APCs such that these cells can be magnetically isolated. This procedure has no effect on the RESULTS ability of these cells to induce clonal deletion in an overnight deletion assay and allows efficient separation of thymocytes Anti-CD3-Mediated Expression of Nur77 Is Promoted by from the APC; typically, contamination was Ͻ1% (data not CD28 Engagement. Negative selection of immature DP thy- shown). mocytes requires costimulatory signals in addition to high- B7 Molecules Selectively Cooperate with TCR Triggering affinity TCR engagement. It has been shown that the inter- for Induction of Nur77 but Not Egr-1 Expression. In a previous action between CD80͞CD86 on APCs with CD28 on DP report, we showed that the B cell lymphoma LEC predomi- thymocytes can generate such costimulatory signals both in the nantly uses B7–2 (CD86) for induction of deletion, as an intact thymus (13, 21) and in in vitro suspension cultures (12, antibody to CD86 inhibits its ability to deliver apoptotic signals 20) (Fig. 1A). A critical event in thymocyte negative selection to DP thymocytes by Ͼ80% (13). We therefore tested whether is induction of expression of Nur77 (28, 29). To evaluate the CD86 molecules on this APC function to induce Nur77 whether expression of this protein would be subject to regu- protein expression in DP thymocytes. When prepulsed with lation by costimulation, we triggered the TCR–CD3 complex p17, LEC B lymphoma cells efficiently induce Nur77 protein on purified DP thymocytes in the presence or absence of a expression in 2B4 DP thymocytes (Fig. 2). However, depriving crosslinking anti-CD28 antibody. As shown in Fig. 1, treatment these cells of costimulatory activity by addition of a blocking with a combination of antibodies specific for the TCR–CD3 antibody against CD86 severely perturbs the ability of these complex and the CD28 receptor resulted in expression of cells to induce the expression of Nur77 protein (Fig. 2). Time Nur77 within 2 hr of stimulation. In contrast, engagement of course experiments revealed that anti-CD86 caused a similar only the antigen receptor or CD28 failed to induce significant reduction of Nur77 protein levels after 1, 2, and 4 hr of expression of Nur77 protein. Expression of Nur77 protein is coculture (data not shown). thus promoted by engagement of CD28. To examine the specificity of the regulation of gene expres- Development of an APC-Based System for Studying Thy- sion by CD86-delivered signals, we also tested the effects of mocyte Gene Regulation Associated with Selection. To exclude CD86 blockade on protein expression of Egr-1. This protein the possibility that the stimulatory effect on Nur77 described also is encoded by an immediate early gene, which is rapidly above was an artifact of antibody-mediated engagement of transcribed on TCR ligation (28, 35, 36); expression of this CD28 (20), we examined whether B7 molecules expressed on immediate early gene is associated with positive selection of a APCs would also costimulate expression of Nur77. For this, we DP thymocyte cell line (36). Furthermore, an instrumental made use of thymocytes from 2B4 TCR-transgenic mice, that role of this protein in positive selection is suggested by the react to the 88–104 COOH-terminal peptide from pigeon finding that overexpression of Egr-1 enhances positive selec- cytochrome C (p17) presented on IEk (43).To avoid compli- tion by low-affinity ligands in transgenic mice (54). In contrast cations from the changing requirements of thymocytes at later to Nur77, the dose response for Egr-1 induction was entirely stages of maturity (19), we used only preselection DP thymo- unaffected by antibody-mediated blockade of CD86 (Fig. 2). cytes: 2B4 TCR transgenic mice were backcrossed to 129 MHC These findings demonstrate that costimulatory interactions class II-deficient mice, and thymocytes from such crosses are selectively cooperate with some, but not all TCR-triggered unable to undergo positive selection. Because Nur77 expres- signaling events in DP thymocytes. sion also may appear in the APC used, we developed an B7-Independent Costimulatory Interactions Exist That Regulate Nur77 Expression. The interaction between B7 molecules and CD28 has been shown in various systems (both in vitro and in vivo) to be capable of providing costimulatory signals for clonal deletion. It is clear, however, that receptors other than CD28 also can fulfill this function, as illustrated by the lack of an overt phenotype on negative selection in CD28-deficient mice found in several studies (15, 17). Con- sistent with this observation, we and others identified APC lines whose ability to induce clonal deletion cannot be abro-

FIG. 1. Anti-CD3 induced expression of Nur77 is promoted by engagement of CD28. DP thymocytes were incubated with plate- bound antibodies (10 ␮g͞ml) to CD3, CD28, or control antibody in different combinations. (A) After overnight incubation, cells were FIG. 2. B cell lymphoma LEC uses CD86 molecules to costimulate harvested and stained with ethidium bromide to assess viability. Shown expression of Nur77 but not Egr-1 protein. 2B4 thymocytes were in each histogram is background ethidium bromide staining from cocultured with LEC and different concentrations of p17 (5, 1 or 0 thymocytes incubated without antibody treatment (thin line) or with ␮g/ml). LEC had been preincubated with 5 ␮g͞ml anti-CD86, 10 the treatment indicated in the figure (thick line). (B) Alternatively, ␮g͞ml anti-Kk, or no antibody. After 2 hr of coincubation, thymocytes thymocytes were collected after 2 hr, and whole-cell lysates were were separated from LEC and whole-cell lysates were analyzed by analyzed for Nur77 expression by Western blot analysis. Note that Western blot analysis using antibodies to Nur77 (Upper)ortoEgr-1 Nur77 from thymocytes typically runs at several different sizes, (Lower). Similar results were obtained after 1 and 4 hr of coincubation presumably because of hyperphosphorylation. (data not shown). Downloaded by guest on September 25, 2021 Immunology: Amsen et al. Proc. Natl. Acad. Sci. USA 96 (1999) 625

gated by B7 blocking. For instance, induction of deletion by the sion of a truncated Nur77 protein lacking the transactivation fibroblast cell line DCEKhi7 (DCEK) can only marginally be domain but still binding to the Nur77͞NBRE leads to inhibi- blocked by anti-B7 (13), indicating that this cell line expresses tion of negative selection (31, 32). These findings indicate that other as-yet-unidentified costimulatory molecules. The activ- thymocytes can express factors related to Nur77 that can also ity of such molecules can be destroyed by chemical fixation of transactivate expression of pro-apoptotic genes from Nur77- DCEK cells with ECDI. This treatment leads to an inability of binding elements. Although in mice expressing truncated these APCs to induce clonal deletion, whereas peptide- Nur77 binding to DNA and transactivation by these factors is presenting ability of such fixed cells remains intact (13). blocked, such factors can efficiently compensate for its absence Importantly, fixation with ECDI also destroys the ability of in Nur77-deficient mice. Indeed, thymocytes undergoing neg- DCEK cells to induce Nur77 expression in DP thymocytes ative selection express at least one homologue of Nur77 (Fig. 3). This pronounced difference in induction of Nur77 (termed Nor-1) that binds to the same element as Nur77 and protein expression between thymocytes that recognized p17 shares the capacity to activate transcription from this element with or without costimulation remained at all time points (56). Thus, not Nur77 alone, but most likely Nur77 or closely examined over a 1–4 hr time course. related proteins such as Nor-1, associate with Nur77-binding To evaluate whether the lack of Nur77 protein expression elements to activate transcription of as-yet-unidentified pro- was a reflection of reduced TCR-triggering by fixed DCEK, we apoptotic target genes. also examined induction of Egr-1 protein expression. Like To investigate the costimulation dependency of all factors antibody-mediated blocking of CD86 (see Fig. 2), fixation of that can bind to Nur77-binding elements, we tested thymocyte DCEK failed to significantly alter the induction of expression lysates for NBRE-binding activity in EMSAs. A complex of of the Egr-1 protein (Fig. 3A). These findings demonstrate the proteins binding to the NBRE is clearly detectable after 2 hr capacity of fixed APCs to present peptide to the thymocytes of thymocyte incubation with DCEK and p17, increasing until and the specificity of the effects of destruction of costimulation a