L E T T E R S

1318 VOLUME 9 | NUMBER 10 | OCTOBER 2003 NATURE MEDICINE

Glucocorticoids, acting through the glucocorticoid receptor,potently modulate immune function and are a mainstay oftherapy for treatment of inflammatory conditions, autoimmunediseases, leukemias and lymphomas1. Moreover, removal ofsystemic glucocorticoids, by adrenalectomy in animal modelsor adrenal insufficiency in humans, has shown thatendogenous glucocorticoid production is required forregulation of physiologic immune responses2. These effectshave been attributed to suppression of cytokines, although the crucial cellular and molecular targets remain unknown3. In addition, considerable controversy remains as to whetherglucocorticoids are required for thymocyte development4–7. To assess the role of the glucocorticoid receptor in immunesystem development and function, we generated T-cell-specificglucocorticoid receptor knockout mice. Here we show that theT-cell is a critical cellular target of glucocorticoid receptorsignaling, as immune activation in these mice resulted insignificant mortality. This lethal activation is rescued bycyclooxygenase-2 (COX-2) inhibition but not steroidadministration or cytokine neutralization. These studiesindicate that glucocorticoid receptor suppression of COX-2 iscrucial for curtailing lethal immune activation, and suggestnew therapeutic approaches for regulation of T-cell-mediatedinflammatory diseases.

T-cell glucocorticoid receptor–deficient mice were generated usingLck promoter–driven, Cre recombinase–mediated excision of exon 2of the glucocorticoid receptor gene (Nr3c1; Fig. 1a,b). Althoughglobal inactivation of Nr3c1 results in perinatal lethality8, micehomozygous for the floxed Nr3c1 gene and harboring the Lck-Cretransgene (designated TGRKO) were as healthy as their Lck-Cre–negative homozygous floxed Nr3c1 littermates (used as controls).Nevertheless, we found little glucocorticoid receptor in whole thy-mus, and no glucocorticoid receptor in purified CD4+ thymocytes(includes CD4+CD8+ and CD4+CD8– subpopulations; Fig. 1c) fromTGRKO mice. This was accomplished using either an antibody to theglucocorticoid receptor N terminus, recognizing an epitope in theloxP-flanked exon, or an antibody to an epitope adjacent to the gluco-corticoid receptor DNA-binding domain. We found no glucocorti-

coid receptor in either CD4+ or CD8+ thymocytes from TGRKO miceusing an antibody to the distal C terminus, recognizing an epitope 3′to the deleted exon (Fig. 1d). Thus, the glucocorticoid receptor is effi-ciently and specifically deleted early in thymocyte development inTGRKO mice.

The glucocorticoid receptor maintains homeostasis by mediatingfeedback inhibition of the hypothalamic-pituitary-adrenal (HPA)axis by glucocorticoids. Additionally, the HPA axis is regulated bycytokines, neuropeptides and the sympathetic nervous system9. Todetermine whether deletion of T-cell glucocorticoid receptor altersHPA axis activity, we analyzed plasma corticosterone at circadiannadir (morning) and peak (evening). TGRKO mice showed circadianregulation and corticosterone responses equal to those of controlswhen given a polyclonal T-cell activation stimulus (antibody toCD3ε; Fig. 1e).

Previous studies have shown conflicted roles of the glucocorticoidreceptor in thymocyte development6,7,10. We therefore analyzed thy-mocytes from TGRKO and control mice. We noted no significant dif-ference in thymus cellularity or subset distribution betweengenotypes (85.8 ± 10.5 × 106 total cells for TGRKO (n = 13) versus102.5 ± 11.6 × 106 total cells for control (n = 14); data not shown). Inaddition, T-cell receptor (TCR)-β and CD25 expression in thymo-cytes, and T-cell distribution in spleen and lymph nodes, did not dif-fer between TGRKO and control mice (data not shown). Thus,glucocorticoid receptor seems dispensable for T-cell development.

To evaluate regulation of proinflammatory molecules, we adminis-tered CD3ε-specific antibody to TGRKO and control mice. In thespleen, this polyclonal T-cell activation induced transient transcrip-tion of interleukin (IL)-2, IL-3, IL-4, IL-6, interferon (IFN)-γ and tumor necrosis factor (TNF)-α, leading to measurablebut transient plasma levels of these cytokines11. In mice and humans,this results in a self-limiting syndrome characterized by hypotension,fever, and hypoglycemia, which can be modulated by glucocorticoidadministration12,13. In contrast to the uniform survival in controls,high mortality occurred in TGRKO mice after administering anti-body to CD3ε, which could not be rescued by pretreatment with theglucocorticoid dexamethasone (DEX; Fig. 2a).

To determine whether TGRKO mice were dying from cytokine dys-regulation, we measured plasma cytokines after polyclonal T-cell acti-

1Washington University School of Medicine, Saint Louis, Missouri 63110, USA. 2Pfizer Inc., Chesterfield, Missouri 63017, USA. Correspondence should beaddressed to L.J.M. (Muglia_L@kids.wustl.edu).

Published online 31 August 2003; doi:10.1038/nm895

T-cell glucocorticoid receptor is required to suppressCOX-2-mediated lethal immune activationJudson A Brewer1, Bernard Khor1, Sherri K Vogt1, Lisa M Muglia1, Hideji Fujiwara2, Karen E Haegele1,Barry P Sleckman1 & Louis J Muglia1

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rem

edic

ine

L E T T E R S

NATURE MEDICINE VOLUME 9 | NUMBER 10 | OCTOBER 2003 1319

vation. We noted increases in TNF-α at 2 h, and IFN-γand IL-6 at 8 hafter T-cell activation in TGRKO mice (Fig. 2b). DEX administrationreduced plasma TNF-α in both TGRKO and control mice, suggestinga non-T-cell contribution to regulation of this cytokine. Plasma IFN-γ was not affected by DEX administration in either genotype,suggesting that it is already maximally suppressed by endogenous glu-cocorticoids in control mice (Fig. 2b).

To assess the role of T-cell glucocorticoid receptor in transcrip-tional regulation of these and other inflammatory genes, we com-pared gene expression in spleens of TGRKO and control mice, 8 hafter T-cell activation, by microarray analysis and ribonucleaseprotection assay (RPA). Of the 21 known genes induced 2.5-fold orgreater in microarray analysis of TGRKO compared with controlsplenocytes, 10 have immune function (fold induction indicated inparentheses): T-cell and activation-regulated chemokine (7.0),small inducible cytokine B subfamily member-5 (6.1), IL-6 (4.3),COX-2 (3.5), Src-suppressed C kinase substrate (3.5), matrix met-alloproteinase-1 (3.2), eotaxin precursor (3.2), IFN-γ (2.8), sup-pressor of cytokine signaling-3 (2.6) and protein kinase inhibitor(2.6). Expression of nine out of nine cytokines analyzed by RPA,from these samples or from pooled lymph nodes (data not shown),showed similar induction to that shown by microarray (Fig. 2c).IFN-γ, but not TNF-α or IL-2, was elevated in TGRKO samples (IL-6 was induced in one of two RPA samples). The low levels ofIL-2 and TNF-α mRNA correlated with reduced plasma cytokinemeasurements 8 h after stimulation (data not shown). These datasuggest that in contrast to TNF-α and IL-2, endogenous glucocorticoids are required for transcriptional suppression ofIFN-γ in T cells (possibly by inhibition of Stat-4 phosphoryla-tion14).

To determine whether unchecked IFN-γ production causes mortal-ity in TGRKO mice, we administered neutralizing antibody to IFN-γbefore in vivo T-cell stimulation. We saw no reduction of mortality inTGRKO mice after IFN-γ depletion (three of three mice died at 1.7 ±0.4 d, with plasma IFN-γ below the limit of detection of 624 pg/ml 8 hafter stimulation).

In addition to cytokines, glucocorticoids regulate expression ofother proinflammatory mediators. COX-2 is a glucocorticoid-mod-ulated enzyme induced in monocytes after lipopolysaccharideadministration and in T cells after activation in vitro15. Consistentwith the microarray data above, COX-2 mRNA measured by RT-PCR was increased in purified splenic TGRKO T cells 8 h after treat-ment with antibody to CD3ε; COX-2 mRNA in splenic T cells fromcontrol mice decreased to levels approximating those of PBS-injectedmice (Fig. 2d). Elevated COX-2 mRNA was associated with dysregu-lated prostaglandin (PG) synthesis in the TGRKO mice, as PGD2production was increased in purified T cells from TGRKO mice stim-ulated with antibody to CD3ε in vivo (Fig. 2e). In contrast, PGF2αand PGE2 showed little or no increase in these same samples (datanot shown). This augmented PGD2 production resisted suppressionby DEX (Fig. 2e).

To determine whether COX-2 and prostaglandin dysregulationinduce mortality, we treated TGRKO mice with a selective COX-2inhibitor (SC-236)16,17 and with antibody to CD3ε. TGRKO micetreated with SC-236 were protected from lethality as compared withvehicle-treated mice (Fig. 3a). COX-2 inhibition did not alter plasmaIFN-γ in these mice (19 ± 4 ng/ml with SC-236 versus 16 ± 4 ng/mlwith vehicle). Lethality was similarly prevented by administration ofanother COX-2 inhibitor, NS-398 (Fig. 3a). To extend these observa-tions to the clinical context of global glucocorticoid deficiency or

a b c

d eFigure 1 T-cell glucocorticoid receptor deletion. (a) Exon 2was flanked by loxP sites (�) for targeted deletion of Nr3c1.WT, wild-type. (b) Southern blot analysis of heart (H),thymus (T), kidney (K) and liver (L) DNA from TGRKO mice.Nr3c1 exon 2 was 87% deleted in thymus (representative of three mice). (c) Glucocorticoid receptor (GR) proteinimmunoblot of whole thymus or purified CD4+ thymocytes.Antibodies to the glucocorticoid N terminus (top two panels)and DNA binding domain region (bottom panel) were used.A nonspecific 70-kDa band is present in all samples in thebottom panel. Actin served as loading control. (d) Proteinimmunoblot using antibody to GR C terminus. Shown areChinese hamsters ovary cells transfected with control plasmid (lane 1), GR expression vector (lane 2) or purified CD4+ or CD8+ thymocytes fromcontrol (Con) or TGRKO (KO) mice. Nonspecific immunoreactive bands (∼ 60 and 43 kDa) were present in both transfected cells and thymocytes of both genotypes. (e) Circadian and T-cell activation–induced (by antibody to CD3ε) plasma corticosterone in TGRKO and control mice (n = 4).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rem

edic

ine

L E T T E R S

1320 VOLUME 9 | NUMBER 10 | OCTOBER 2003 NATURE MEDICINE

resistance, we repeated the experiments in control mice pretreatedwith mifepristone (RU-486), a glucocorticoid receptor antagonist.Mice treated with mifepristone and vehicle were significantly moresusceptible to the lethal affects of T-cell activation than control micetreated with vehicle alone (Fig. 3b). As with TGRKO mice treated withSC-236, control mice treated with mifepristone and SC-236 were pro-tected compared with their vehicle-treatedcounterparts (Fig. 3b). In addition, SC-236treatment of adrenalectomized control miceshowed the same degree of rescue, though ona shorter time scale (none of three sham-operated, vehicle-treated mice, two of sixadrenalectomized, SC-236–treated mice, andfour of five adrenalectomized, vehicle-treatedmice died within 8 h of treatment with anti-body to CD3ε).

Bacterial superantigens activate 5–20% ofT cells and induce human diseases such astoxic shock syndrome18. To examine whetherglucocorticoid receptor is important duringsuperantigen-mediated disease, we adminis-tered staphylococcal enterotoxin A (SEA) toTGRKO and control mice. As with adminis-tration of the antibody to CD3ε, this morerestricted stimulus proved lethal for TGRKObut not for control mice. In addition, SEA-induced mortality was rescued by COX-2inhibition (Fig. 3c). These data show that T-cell glucocorticoid receptor modulation of COX-2 expression is required to preventT-cell activation from becoming lethal.

How do activation of glucocorticoid recep-tor–deficient T-cells and dysregulation ofCOX-2 activity cause mortality? Histological

analyses of TGRKO and control mice treated with antibody to CD3ε36–48 h revealed similar vacuolization and apoptosis in liver and kid-ney, and little evidence of damage in brain, heart or lung (data notshown). In contrast, we observed far greater damage to the TGRKOgastrointestinal tract, particularly the cecum (Fig. 3d). TGRKO micehad mucosal and submucosal edema and inflammation, along with

a

b

c d

e

Figure 2 T-cell glucocorticoid receptor is required for prevention of lethality. (a) Kaplan-Meyer plots of mice treated with antibody to CD3ε, with (right) orwithout (left) DEX. �, control (n = 10); �, TGRKO (n = 4); �, control + DEX (n = 10); �, TGRKO + DEX (n = 7). P < 0.01 for TGRKO mice compared withcontrol, in both plots. (b) Plasma cytokines in TGRKO and control mice after injection of antibody to CD3ε, with or without DEX (n = 6–9 for TGRKO and8–13 for control). *, P < 0.05; **, P < 0.01 for TGRKO compared with control. (c) RPA analysis of splenic RNA after injection of antibody to CD3ε. (+), positive control RNA; NC, no change. (d) RT-PCR analysis of COX-2 mRNA in purified T cells from control (Con) and TGRKO (KO) mice after injectionwith PBS or antibody to CD3ε (anti-CD3ε). β2-microglobulin mRNA (β2m) was used as a control for RNA recovery and loading. (e) PGD2 measurement insupernatants of cultured T-cells from mice stimulated in vivo with PBS or antibody to CD3ε.

a b c

d

Figure 3 COX-2 inhibition in TGRKO mice. (a) TGRKO mice treated with antibody to CD3ε and SC-236 (S; n = 7), NS-398 (N; n = 5) or vehicle (V; n = 8). P ≤ 0.05 for NS-398– or SC-236–treatedcompared with vehicle-treated TGRKO mice. (b) Control mice treated with antibody to CD3ε andmifepristone + SC-236 (M/S; n = 10), mifepristone + vehicle (M/V; n = 8) or vehicle + vehicle (V/V; n = 3). P < 0.05 for SC-236–treated compared with vehicle-treated mice. (c) Galactosamine-sensitized, SEA-treated control (n = 9), TGRKO (n = 3) and TGRKO mice given SC-236 (TGRKO + S;n = 5). P < 0.01 for TGRKO compared with control mice. (d) Histological analysis of ceca in micetreated with antibody to CD3ε. Sections are representative of n = 3–5 mice.

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rem

edic

ine

L E T T E R S

NATURE MEDICINE VOLUME 9 | NUMBER 10 | OCTOBER 2003 1321

necrotic regions that were denuded of mucosa. Administration of NS-398 substantially reduced these changes in TGRKO mice (Fig. 3d). Insitu hybridization showed that COX-2 mRNA was induced in theseinflammatory areas of mucosal epithelial cells and submucosa inTGRKO mice, suggesting that COX-2 inhibitors may attenuate mor-tality in TGRKO mice by modulating dysregulated T-cell COX-2activity as well as secondary induction of COX-2 in other cells (datanot shown).

In summary, our data show that glucocorticoid receptor functionin T-cells is essential for survival during polyclonal T-cell activation.T-cell glucocorticoid receptor deficiency results in dysregulation ofseveral cytokines, but redundant mechanisms for downregulation ofsome of these molecules exist. One crucial role for glucocorticoidreceptor in activated T-cells seems to be modulation of COX-2expression. Among T-cells, COX-2 and PGD synthase are coexpressedpredominantly in TH2 cells19. Given our findings of dysregulation ofboth IFN-γ and PGD2, it seems that glucocorticoid receptor modu-lates activation of both TH1 and TH2 cells. Recent reports indicatedthat T-cell activation with antibody to CD3ε causes gastrointestinalmucosal cell damage, as well as functional alterations that are inde-pendent of IFN-γ and only partially ameliorated by inactivation ofTNF-α signaling20,21. Our data indicate that glucocorticoid receptorin T-cells limits the severity of this process by inhibiting prostanoidproduction, particularly the proinflammatory actions of PGD2. Wepropose the use of selective COX-2 inhibition, especially in the settingof glucocorticoid insufficiency or resistance, as therapy for limitingmorbidity and mortality in patients with toxic shock syndrome, graft-versus-host disease and other T-cell-activating processes.

METHODSAnimal handling. Mouse protocols were approved by the Animal Care and UseCommittee of Washington University. Mice were housed on a 12-h light and12-h dark cycle. Blood was collected by retro-orbital phlebotomy intoheparinized capillary tubes, with the time from first handling the animal tocompletion of bleeding not exceeding 30 s. Plasma was stored at –80 °C untilassay. Mice used were 6–10 weeks old and were of a C57BL/6 X 129/Sv back-ground.

Generation of TGRKO mice. The Nr3c1 locus was targeted by inserting a loxPsite into a SacI site upstream of exons 1C and 2 and inserting a PGK neo cas-sette flanked by loxP sites into intron 2. After germline transmission, we matedheterozygous mice harboring the floxed exon 2 allele to Lck-Cre transgenicmice, provided by J. Marth (University of California, San Diego). We matedLck-Cre-harboring Nr3c1-floxed exon 2 homozygotes (Nr3c1flox/floxLckCre,designated TGRKO) and floxed exon 2 homozygotes without the Lck-Cretransgene (Nr3c1flox/flox, used as control) for our experiments. Southern blotanalysis of DNA from individual tissues in TGRKO mice was quantitated on aMolecular Dynamics PhoshorImager. All experiments were done on sex-matched littermates.

Antibody detection of glucocorticoid receptor. We harvested total proteinfrom whole thymus, from CD4+ or CD8+ thymocytes sorted by flow cytome-try (MoFlo; Cytomation) or magnetic beads (Dynal Biotech), or from Chinesehamster ovary cells transfected with a mouse glucocorticoid receptor expres-sion vector (pSV2mGR; kindly provided by J. Bodwell, Dartmouth MedicalSchool) or control plasmid (pBluescript SKII+). We resolved 15–25 µg of pro-tein on 4–12% bis-tris polyacrylamide gels and probed membrane-immobi-lized proteins with antibodies to glucocorticoid receptor (M-20 from SantaCruz Biotechnology, or BuGR2 or PAI-516 from Affinity Bioreagents) or actin(Sigma). Proteins were visualized using enhanced chemiluminescence detec-tion reagents (Amersham).

Cytokine, prostaglandin and corticosterone measurements. We measuredplasma cytokines and corticosterone (PharMingen and ICN, respectively) as

previously described22. Prostaglandins were measured in supernatants of cul-tured splenic T cells. Mice were injected with PBS or 100 µg of antibody toCD3ε (n = 2 per genotype and treatment). Splenocytes were pooled and puri-fied using a SpinSep kit (Stem Cell Technologies) 8 h after injection. Purity ofT cells was >93% by FACS. 5 × 105 purified T cells were cultured for 12 h intriplicate wells containing 200 µl of DMEM with 10% FBS, in the presence orabsence of 10 nM dexamethasone. Supernatants from triplicate wells of eachgroup were pooled for analysis. Quantitation was accomplished by on-line C-18 trapping of prostaglandins (Michrom BioResources and KeystoneScientific) followed by liquid chromatography and tandem mass spectroscopy.An API-4000 mass spectrometer (Applied Sciences) was operated in the nega-tive ion turbospray multiple reaction–monitoring mode using Analyst 1.3software. The amount of PGD2 in the sample was determined by normaliza-tion of integrated PGD2 to a PGE2 tetradeuterated standard chromatographicsignal.

RNA analyses. We isolated splenic or purified T-cell RNA (RNEasy; Qiagen)from control and TGRKO mice (n = 2 per group) 8 h after administration ofPBS or antibody to CD3ε. We did RPA on 2 µg of total RNA from individualmice according to the manufacturer’s instructions (PharMingen). Pooledsplenic RNA from mice treated with antibody to CD3ε was hybridized toAffymetrix murine U74A microarrays. Data was analyzed using MicroarraySuite Version 5.0 software (Affymetrix). RT-PCR analysis was done on RNAisolated from 5 × 105 T cells pooled from 2–3 mice of each genotype. Primersspecific for COX-2 or β2-microglobulin were used to amplify cDNA as previ-ously described23. We evaluated 25–30 cycles of amplification to establish lin-earity (the gel in Figure 2 shows samples amplified for 27 cycles). Sampleslacking reverse transcriptase did not produce any product. RT-PCR experi-ments were repeated on two separate groups of mice with similar results.

Flow cytometry. Thymocytes were dispersed into PBS, washed, counted,stained for cell surface markers (using phycoerythrin-conjugated antibody toCD25, peridinin chlorophyll protein–conjugated antibody to CD8, allophyco-cyanin-conjugated antibody to CD4, FITC-conjugated antibody to CD69 andphycoerythrin-conjugated antibody to TCRβ, all from PharMingen), washed,resuspended in PBS and analyzed on a FACSCaliber (Becton Dickinson).Nonviable cells were excluded based on forward- and side-scatter profiles.

Adrenalectomy. Mice were adrenalectomized and allowed to recover 1 weekbefore being subjected to experimentation as previously described24.

Pharmacological and antibody treatment. Mice were injected intraperi-toneally with 100 µg of antibody to CD3ε (145-2C11) diluted in 250 µl PBS.SEA-treated mice were injected intraperitoneally with 30 mg of galactosamine(Sigma) and, 1.5 h later, 20 µg SEA (Toxin Technologies) diluted in 200 µl PBS.Dexamethasone-treated mice were injected intraperitoneally with 200 µg dex-amethasone phosphate just before and 8 h after antibody challenge.Neutralizing antibody to IFN-γ (H22; 50 µg; gift of R. Schreiber, University ofWashington) was injected intraperitoneally 1 d before administration of anti-body to CD3ε25. Mifepristone (RU486; 0.5 mg in sesame oil; Sigma) was givensubcutaneously the night before and 1 h before inflammatory challenge. Micetreated with COX-2 inhibitor were given 300 µg of either SC-236 (Pharmacia)or NS-398 (Cayman Chemical), or vehicle twice daily for 2 d as described16.

Statistical methods. All results are expressed as mean ± s.e.m. unless otherwisestated. Statistical analysis was done by analysis of variance, with P ≤ 0.05 con-sidered significant.

ACKNOWLEDGMENTSWe thank E. Unanue, R. Schreiber and J. Gitlin for critical review of thismanuscript; A. Cheng for insightful discussions and critical review of thismanuscript; E. Plut, M. Wallace and the Washington University Mouse GeneticsCore for embryonic stem cell injections; K. Hamilton and the Siteman CancerCenter Genechip Facility for microarray experimentation; K. Sheehan and R. Schreiber for providing neutralizing antibodies to IFN-γ; and PharmaciaCorporation for providing SC-236. This work was supported by grants from theNational Institutes of Health and the Pharmacia–Washington UniversityBiomedical Research Program (L.J.M.) and the Medical Scientist Training Program(J.A.B.).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rem

edic

ine

L E T T E R S

1322 VOLUME 9 | NUMBER 10 | OCTOBER 2003 NATURE MEDICINE

COMPETING INTERESTS STATEMENTThe authors declare competing financial interests (see the Nature Medicine websitefor details).

Received 14 March; accepted 2 June 2003Published online at http://www.nature.com/naturemedicine/

1. Webster, J.I., Tonelli, L. & Sternberg, E.M. Neuroendocrine regulation of immunity.Annu. Rev. Immunol. 20, 125–163 (2002).

2. Bertini, R., Bianchi, M. & Ghezzi, P. Adrenalectomy sensitizes mice to the lethaleffects of interleukin 1 and tumor necrosis factor. J. Exp. Med. 167, 1708–1712(1988).

3. Winoto, A. & Littman, D.R. Nuclear hormone receptors in T lymphocytes. Cell 109,S57–S66 (2002).

4. Ashwell, J.D., Vacchio, M.S. & Galon, J. Do glucocorticoids participate in thymocytedevelopment? Immunol. Today 21, 644–646 (2000).

5. Jondal, M., Pazirandeh, A. & Okret, S. A role for glucocorticoids in the thymus?Trends Immunol. 22, 185–186 (2001).

6. King, L.B. et al. A targeted glucocorticoid receptor antisense transgene increases thy-mocyte apoptosis and alters thymocyte development. Immunity 3, 647–656 (1995).

7. Purton, J.F., Boyd, R.L., Cole, T.J. & Godfrey, D.I. Intrathymic T cell development andselection proceeds normally in the absence of glucocorticoid receptor signaling.Immunity 13, 179–186 (2000).

8. Cole, T.J. et al. Targeted disruption of the glucocorticoid receptor gene blocks adren-ergic chromaffin cell development and severely retards lung maturation. Gen. Dev. 9,1608–1621 (1995).

9. Da Silva, J.A. Sex hormones and glucocorticoids: interactions with the immune sys-tem. Ann. N. Y. Acad. Sci. 876, 102–117 (1999).

10. Vacchio, M.S., Lee, J.Y. & Ashwell, J.D. Thymus-derived glucocorticoids set thethresholds for thymocyte selection by inhibiting TCR-mediated thymocyte activation.J. Immunol. 163, 1327–1333 (1999).

11. Scott, D.E., Gause, W.C., Finkelman, F.D. & Steinberg, A.D. Anti-CD3 antibodyinduces rapid expression of cytokine genes in vivo. J. Immunol. 145, 2183–2188(1990).

12. Charpentier, B. et al. Evidence that antihuman tumor necrosis factor monoclonalantibody prevents OKT3-induced acute syndrome. Transplantation 54, 997–1002(1992).

13. Ferran, C. et al. Reduction of morbidity and cytokine release in anti-CD3 MoAb-treated mice by corticosteroids. Transplantation 50, 642–648 (1990).

14. Franchimont, D. et al. Inhibition of Th1 immune response by glucocorticoids: dexam-ethasone selectively inhibits IL-12-induced Stat4 phosphorylation in T lymphocytes.J. Immunol. 164, 1768–1774 (2000).

15. Iniguez, M.A., Punzon, C. & Fresno, M. Induction of cyclooxygenase-2 on activated T lymphocytes: regulation of T cell activation by cyclooxygenase-2 inhibitors. J. Immunol. 163, 111–119 (1999).

16. Gross, G. et al. Inhibition of cyclooxygenase-2 prevents inflammation-mediatedpreterm labor in the mouse. Am. J. Physiol. Regul. Integr. Comp. Physiol. 278,R1415–R1423 (2000).

17. Penning, T.D. et al. Synthesis and biological evaluation of the 1,5-diarylpyrazole classof cyclooxygenase-2 inhibitors: identification of 4-[5-(4-methylphenyl)-3- (trifluo-romethyl)-1H-pyrazol-1-yl]benze nesulfonamide (SC-58635, celecoxib). J. Med.Chem. 40, 1347–1365 (1997).

18. Li, H., Llera, A., Malchiodi, E.L. & Mariuzza, R.A. The structural basis of T cell acti-vation by superantigens. Annu. Rev. Immunol. 17, 435–466 (1999).

19. Tanaka, K. et al. Cutting edge: differential production of prostaglandin D2 by humanhelper T cell subsets. J. Immunol. 164, 2277–2280 (2000).

20. Merger, M. et al. Defining the roles of perforin, Fas/FasL, and tumour necrosis factoralpha in T cell induced mucosal damage in the mouse intestine. Gut 51, 155–163(2002).

21. Musch, M.W. et al. T cell activation causes diarrhea by increasing intestinal perme-ability and inhibiting epithelial Na+/K+-ATPase. J. Clin. Invest. 110, 1739–1747(2002).

22. Bethin, K.E., Vogt, S.K. & Muglia, L.J. Interleukin-6 is an essential, corticotropin-releasing hormone-independent stimulator of the adrenal axis during immune systemactivation. Proc. Natl. Acad. Sci. USA 97, 9317–9322 (2000).

23. Muglia, L.J., Jenkins, N.A., Gilbert, D.J., Copeland, N.G. & Majzoub, J.A. Expressionof the mouse corticotropin-releasing hormone gene in vivo and targeted inactivationin embryonic stem cells. J. Clin. Invest. 93, 2066–2072 (1994).

24. Muglia, L.J. et al. Corticotropin-releasing hormone links pituitary adrenocorticotropingene expression and release during adrenal insufficiency. J. Clin. Invest. 105,1269–1277 (2000).

25. Ferran, C. et al. Cascade modulation by anti-tumor necrosis factor monoclonal anti-body of interferon-gamma, interleukin 3 and interleukin 6 release after triggering ofthe CD3/T cell receptor activation pathway. Eur. J. Immunol. 21, 2349–2353(1991).

©20

03 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://w

ww

.nat

ure

.co

m/n

atu

rem

edic

ine