Cutting Edge: De Novo Glucocorticoid Synthesis by Thymic Epithelial Cells Regulates Antigen-Specific Thymocyte Selection

Generation of a competent but self-tolerant T cell Ag-specific repertoire takes place in the thymus. The fate of CD4⁺CD8⁺ (double positive, DP) thymocytes is determined by recognition of self-peptides presented by MHC molecules (self-pMHC). DP cells with TCRs that do not recognize self-pMHC presented by cortical thymic epithelial cells (TEC) die by neglect. Those that recognize self-pMHC enter the medulla where they encounter migratory dendritic cells, some of which present self-pMHC derived from peripheral tissues, medullary TEC in which the autoimmune regulator (Aire) drives expression of tissue-restricted Ags, and resident dendritic cells bearing peptides transferred from medullary TEC (1). DP cells that have TCRs with strong avidity for self-pMHC die (negative selection), whereas those with intermediate avidity survive (positive selection) and populate the periphery (2, 3).

Glucocorticoids (GC) are steroid hormones that bind the GC receptor (GR), a ligand-dependent transcription factor that translocates to the nucleus and regulates transcription by binding to its response elements or other transcription factors. GC potently downregulate the production of proinflammatory cytokines, chemokines, and PGs, and antagonize NF-κB and AP-1 (4). GC also inhibit transcriptional activity of Nur77 (5), a TCR-induced transcription factor implicated in thymocyte negative selection (6, 7). We previously suggested that by blunting TCR signals at a distal step (i.e., in the nucleus), GC could raise the threshold of avidity for self-pMHC above which negative selection takes place, allowing positive selection of TCRs that would otherwise be negatively selected (8). Evidence for an effect of GC on thymocyte selection was initially obtained from fetal thymic organ cultures in which negative selection of TCR-transgenic thymocytes was increased by pharmacologic inhibition of local GC production (9). This was subsequently supported by in vivo studies in which GR expression was reduced by the expression of an antisense transgene (10–12). The best evidence has been obtained with mice in which the GR was deleted in thymocytes (13). T cells from these mice responded normally to repertoire-independent TCR stimuli, but had diminished responses to immunization with foreign Ag, infection with lymphocytic choriomeningitis virus (LCMV) Armstrong strain, and culture with allogeneic APC, indicating a decrease in the avidity with which the repertoire recognized pMHC (13). Alterations of the TCR repertoire were confirmed by analysis of TCR Vβ CDR3 sequences.

Although circulating GC are primarily produced in the adrenal cortex, the thymus is itself a site of synthesis (14–19). Cultured mouse and chicken TECs express GC-synthetic enzymes and secrete steroid intermediates and GC themselves, production being highest at birth when adrenal production of GC is lowest (14, 17, 20). Direct measurement of thymus GC found corticosterone and its precursor steroid concentrations to be higher than in blood, particularly shortly after birth, confirming thymic GC synthesis in vivo (19). In addition to TEC, it has been proposed that thymocytes themselves are a source of GC, especially later in life (18). The functional contribution of extra-adrenal GC synthesis in the thymus, or any tissue for that matter, is unknown. To address this, we conditionally deleted 11β-hydroxylase (Cyp11b1) (P450 c11b1), the enzyme that catalyzes the conversion of biologically inactive precursors to active GC, in TEC or thymocytes, and characterized the results in thymocytes and T cells.

C57BL/6 (B6) and the congenic strains B10.A and Rag2^−/−, AND TCR-transgenic mice (21), β-actin-FLPe (22), FoxN1-Cre–transgenic mice (23), and β-actin-Cre mice were obtained from the Jackson Laboratory. Lck-Cre–transgenic mice were obtained from Taconic. Nr3c1 (GR) exon 3 conditionally targeted mice were described (13). A conditional Cyp11b1 allele with loxp sites flanking exons 3–5 was generated by recombineering (24) (Supplemental Fig. 1A), and Cyp11b1-floxed mice were generated using C57BL/6 embryonic stem cells. The Neo cassette was removed by crossing floxed mice with β-actin-FLPe transgenic mice. All mice used in this study were backcrossed for at least six generations onto B6. Primer sequences used for genotyping are provided in Supplemental Table I.

Anti-CD3 (145-2C11) and anti-CD28 (37.51) were from BD Pharmingen. For flow cytometry, Abs recognizing CD45.2 (104), CD4 (RM4-5), and PD-1 (J43) were from eBioscience, recognizing Helios (22F6) from BioLegend, and recognizing Bim (C34C5) from Cell Signaling Technology. Abs against EpCAM (G8.8), MHC class II (MHC-II) (M5/114.15.2), CD8α (53-6.7), and TCRβ (H57), as well as Annexin V, were from BD Pharmingen.

Corticosterone was measured by chemiluminescence ELISA (Arbor Assays).

T cells were cultured in RPMI 1640 (Biofluids) supplemented with 10% heat-inactivated calf serum (Sigma), 100 μg/ml gentamicin, 4 mM glutamine, and 50 μM 2-ME. To measure T cell proliferation, 3 × 10⁴ (anti-CD3/CD28) or 1.5 × 10⁵ (alloantigen) purified lymph node T cells were cultured in triplicate in a total volume of 200 μl complete medium with 0.5 μg/ml plate-bound anti-CD28 and the indicated amounts of plate-bound anti-CD3, or with the indicated numbers of irradiated B10.A splenocytes in 96-well plates. After 48 (anti-CD3/anti-CD28) or 72 (alloantigen) h wells were pulsed overnight with ³H-thymidine and harvested. Incorporation of radioactivity was determined using a 1450 Microbeta liquid scintillation counter (Wallac).

TEC were isolated by digestion of minced thymi and enrichment with discontinuous percoll as described (25). Genomic DNA from sorted TECs (Epcam⁺, MHC-II⁺, and CD45.2⁻) and DP thymocytes (TCRβ⁺, CD4⁺CD8α⁺) was purified using a DNeasy kit (Qiagen). Sorts were performed with a FACSAria II or a FACSAria Fusion (BD Biosciences). T cells used in proliferation assays were purified from lymph nodes using Dynabead Untouched Mouse T cell kits (Invitrogen).

Total RNA was isolated with an RNeasy Mini kit (Qiagen) and cDNA generated with Superscript RT (Invitrogen). Real-time PCR was performed with SYBR Green PCR mix (Applied Biosystems) using a QuantStudio 6 (Applied Biosystems). The results are relative to Hprt expression. Primer sequences used for real-time PCR are provided in Supplemental Table I.

LCMV Armstrong was obtained from Dorian McGavern (National Institute of Neurological Disorders and Stroke). Mice were inoculated i.p. with 2 × 10⁵ PFU. Splenocytes were stained with APC-labeled class I tetramers containing LCMV peptides gp33, gp276, and np396 (National Cancer Institute tetramer core facility).

LPS from Escherichia coli (#L2880; Sigma) was solubilized in PBS and injected i.p. at a dose of 3 μg/g body weight. Control mice were injected with PBS alone.

Unless otherwise indicated, statistical analyses were performed using GraphPad Prism software and an unpaired two-tailed Student t test. A p value ≤ 0.05 was considered significant. Averaged results of multiple experiments are presented as the arithmetic mean ± SEM.

Mice with global deletion of exons 3–7 of Cyp11b1 have markedly diminished adrenal corticosterone production (26). To address the role of thymus-derived GC in thymocyte development, we generated mice in which Cyp11b1 could be disrupted in a tissue-specific manner. Mice were generated in which loxp sites flanked Cyp11b1 exons 3–5 (Cyp11b1^fl/fl mice, Supplemental Fig. 1A). Deletion of these exons results in early termination of translation after an open reading frame containing the first 135 of 501 aa followed by five out-of-frame residues (Supplemental Fig. 1B). These mice were crossed with animals expressing actin-Cre to delete Cyp11b1 in the entire germline (Cyp11b1^−/− mice). Cultured wild-type (WT) but not Cyp11b1^−/− adrenal cells synthesized substantial amounts of corticosterone, which was prevented by the Cyp11b1 inhibitor metyrapone (Fig. 1A). Plasma corticosterone levels in Cyp11b1^−/− mice were <50% of WT levels, similar to the reduction reported for Cyp11b1 exon 3–7-targeted mice (26), but were not statistically significantly different from levels in Cyp11b1^foxN1-Cre animals (Fig. 1B). Systemic GC increase in response to an acute stress such as LPS (27). Intraperitoneal injection of LPS increased plasma corticosterone levels in WT but not Cyp11b1^−/− mice (Fig. 1C). Finally, Cyp11b1^−/− mice exhibited adrenal hyperplasia (Fig. 1D), characteristic of impaired GC production (26).

Cyp11b1 was deleted in TEC or thymocytes by crossing Cyp11b1^fl/fl mice with mice expressing Cre knocked into the FoxN1 locus (23) (Cyp11b1^foxN1-Cre mice) or expressed as a transgene driven by the proximal lck promoter (Cyp11b1^lck-Cre mice) (28). The tissue specificity of deletion was shown by PCR of genomic DNA from sort-purified cells, which demonstrated that Cyp11b1 exon 4 was absent in Cyp11b1^foxN1-Cre TEC but not DP thymocytes, with the opposite being true for Cyp11b1^lck-Cre cells (Fig. 2A). Because stress-induced elevations in adrenal-derived GC can cause acute thymic involution (29), it is assumed that the thymus also responds to circulating levels at steady-state. However, the relative contributions of systemic versus paracrine GC have never been experimentally addressed. To do this, the constitutive expression of two GC-responsive genes, Gilz and Lad1 (30), was used as a read-out of GC signaling in freshly isolated thymocytes. In GR-deficient thymocytes (GR^lck-Cre), Gilz and Lad1 mRNA levels were reduced 40–50% compared with WT controls (Fig. 2B). Gilz and Lad1 expression were similarly reduced in Cyp11b1^−/− thymocytes, as expected, and also in Cyp11b1^foxN1-Cre thymocytes, in which only TEC-synthesized GC are absent. In contrast, this reduction was not observed in the thymocytes of Cyp11b1^lck-Cre mice, or in Cyp11b1 heterozygous (Cyp11b1^{fl/+,actin-Cre}) thymocytes. Thymocyte expression of the GC-unresponsive gene Calm3 was similar across genotypes. Differences in Gilz and Lad1 expression were not due to GC hyporesponsiveness, as they were induced by exogenous GC in all GR-sufficient thymocytes (Supplemental Fig. 1C). Together, these data show that biologically active thymus GC are synthesized de novo by TEC, not thymocytes, in situ. Furthermore, under basal conditions, thymocyte GC signaling is primarily driven by TEC- rather than adrenal-derived GC.

Deletion of the GR in immature thymocytes results in the negative selection of cells that otherwise would have been positively selected (13). One example was the reduction in DP and CD4⁺ thymocytes in AND mice whose transgenic TCR normally promotes positive selection of CD4⁺ T cells in H-2^b animals. To prevent rearrangement of endogenous TCRα, recombinase-activating gene 2 (RAG-2) was deleted by crossing AND with RAG2^−/− mice. As observed with AND GR^lck-Cre mice, whose thymocytes cannot respond to GC (13), there was a reduction in the number of DP and CD4⁺ single positive (SP) thymocytes in Cyp11b1^foxN1-Cre AND mice compared with Cyp11b1^fl/fl AND controls (Fig. 3A). Among the molecules upregulated during negative selection are PD-1, Helios, and the proapoptotic Bcl-2 family member Bim (31). Furthermore, Annexin V binds to cells actively undergoing apoptosis (32). We examined preselection thymocytes (DP), TCR-signaled DP thymocytes [TCR^hi (33)], a population of cells enriched for those undergoing negative selection [CD4^low/CD8^low double dull (34–36)], and mature thymoytes (CD4⁺ SP). We found that double dull, and to a lesser extent SP, thymocytes from Cyp11b1^foxN1-Cre AND mice had a larger fraction of Bim^hiPD-1⁺ cells compared with Cyp11b1^fl/fl AND controls (Fig. 3B). Strikingly, the fraction of apoptotic cells (Helios^hiAnnexin V⁺) was increased in Cyp11b1^foxN1-Cre thymocytes, most notably in the double dull subset. Therefore, loss of TEC-derived GC mice resulted in increased negative selection of thymocytes that normally undergo positive selection.

A hallmark of the changed repertoire in GR-deficient T cells is a decreased allogenic response (13). If paracrine production by TEC is the major source of intrathymic GC, Cyp11b1^foxN1-Cre T cells would also be expected to have a reduced response to allogeneic APC. To test this, H-2^b T cells were cultured with irradiated H-2^a splenocytes. Whereas Cyp11b1^foxN1-Cre T cells responded normally to stimulation with plate-bound anti-CD3/anti-CD28, the response of Cyp11b1^foxN1-Cre T cells to allogeneic stimulation was blunted (Fig. 4A). A possible contribution of thymocyte-synthesized GC on the TCR repertoire was addressed by parallel experiments using Cyp11b1^1ck-Cre T cells as responders. The response of WT and Cyp11b1^1ck-Cre T cells was identical to both anti-CD3/anti-CD28 and allogeneic APC (Fig. 4B).

Similar to alloantigen, the altered TCR repertoire in GR^lck-Cre caused T cells to respond less well when mice were infected with LCMV Armstrong strain (13), which elicits a robust CD8⁺ T cell response that peaks at 7 d. WT and Cyp11b1^foxN1-Cre mice were infected with LCMV and 7 d later splenic T cells were characterized. Cyp11b1^foxN1-Cre mice had a 15 and 30% reduction in CD4⁺ and CD8⁺ T cell subsets, respectively, compared with WT (Fig. 5A). There was a decrease in the number of LCMV gp33–specific CD8⁺ T cells in spleens of Cyp11b1^foxN1-Cre mice as measured by peptide–MHC-I tetramer binding. There was also a decrease in LCMV np396–reactive CD8 T cells, although the reduction did not achieve statistical significance. In contrast, targeting Cyp11b1 expression in thymocytes did not affect the response to LCMV (Fig. 5B). Thus, T cells that developed in the absence of TEC- but not thymocyte-provided GC have a reduced ability to respond to pMHC.

It was long assumed that GC acting on the thymus were derived from the circulation. However, the discovery that TEC can produce GC (14, 16) raised the possibility that the locally derived product was biologically active. In fact, blockade of GC production in fetal thymic organ culture resulted in increased TCR-mediated activation and enhanced negative selection (15). To assess the relative contribution of local versus systemic GC in the thymus, we quantified the expression of GC-responsive genes in thymocytes at steady-state. Remarkably, the lack of local synthesis reduced expression of these genes to the same levels as in GR-deficient thymocytes, which cannot respond to GC at all. This implies that the bulk of biologically active GC in the thymus under basal conditions is supplied by TEC in a paracrine manner. This discrepancy with the classical understanding of endocrine GC signaling may be explained by enhanced bioavailability. Interaction between thymocytes and TECs is a prerequisite for the TCR-pMHC interactions underlying selection. Paracrine delivery via cell-cell contact raises the possibility that GC pass directly from TECs to thymocytes without diffusion and dilution into the extracellular space. In addition, hormone delivered directly to thymocytes would bypass carrier proteins. Approximately 80–90% of plasma GC are bound by the corticosteroid-binding globulin and another 5–10% by albumin, leaving only 5% free and available to signal by entering the cell and binding the GR (37). GC that pass directly from TEC to thymocytes would therefore have a much higher effective concentration than those delivered via the blood.

It has been reported that thymocytes express the complete set of steroidogenic enzymes and can produce measurable corticosterone in vitro (18, 38, 39). Thymocytes from adult (14–22 wk of age) mice produced more GC than those from younger mice, leading to the suggestion that thymocytes supplant TECs for GC production later in life (18). We and others were unable to detect thymocyte Cyp11b1 activity in vitro [data not shown and (40)]. We were, however, able test for a role for thymocyte-produced GC genetically by deleting Cyp11b1 in thymocytes and T cells. Expression of GC-sensitive genes in Cyp11b1^lck-Cre thymocytes was normal, indicating that GC levels sensed by thymocytes at the population level were normal. As indicated by the response to allogeneic APC and LCMV, the repertoire of Cyp11b1^lck-Cre T cells was also normal.

The data in this report demonstrate a paracrine mode of action by GC produced in the thymus, where effects of GC have previously been ascribed to hormonal control via adrenal production. Awareness of the impact of locally produced GC could aid development of targeted therapies addressing thymocyte development.

We thank Bei Dong for expert technical assistance and Dorian McGavern for the gift of LCMV.

The authors have no financial conflicts of interest.