NR4A3 Mediates Thymic Negative Selection

This article is featured in Top Reads, p.999

Tcells recognize, via their TCR, a peptide fragment of foreign Ag in association with MHC class I or II molecules. The generation of a repertoire of T cells endowed with the ability to recognize almost all possible foreign Ags is due to TCR gene rearrangement, a process in which juxtaposition of TCR gene segments occurs to create TCR sequence diversity. To select T cells that are able to recognize foreign Ag in the context of self-MHC molecules but do not respond to self-peptide–MHC (pMHC) complexes, an education process must occur (1–3). This education process, termed selection, begins at the CD4⁺CD8⁺ double-positive (DP) stage. At this stage, thymocytes that fail to express a TCR capable of interacting with self-pMHC complexes die by neglect, whereas DP thymocytes expressing a TCR that recognize self-pMHC with low to moderate affinity are positively selected and allowed to further differentiate into single-positive CD4⁺ (SPCD4) or single-positive CD8⁺ (SPCD8) cells. Thymocytes, either at the DP or single-positive (SP) stage, that bear a strongly autoreactive TCR will either die by apoptosis (negative selection) or be diverted into alternative differentiation pathways, leading to the production of, for example, CD8αα intestinal intraepithelial lymphocytes (IEL) or T regulatory cells (Tregs) (2, 4). A substantial portion of negative selection takes place at the DP stage of development in the cortex, where thymocytes bearing TCRs specific for ubiquitously expressed Ags are deleted (5–8). As thymocytes mature, they upregulate CCR7 expression and migrate to the medulla, where developing T cells further test their Ag receptors for reactivity to tissue-restricted Ags (2).

Given that TCR signal strength determines the fate of maturing thymocytes, elements from the TCR signaling cascade, as well as pro- and antiapoptotic molecules, are important in determining cell fate in the thymus. There is also evidence that RNA transcription is required for optimal induction of TCR-induced thymocyte apoptosis (9, 10). Therefore, certain transcription factors may be crucial mediators of negative selection. The transcription factor HELIOS has recently been shown to be a specific marker of clonal deletion in CD4⁺ T cells (11) and is overexpressed in DP thymocytes undergoing negative selection (12).

In particular, NR4A family members NR4A1 (Nur77), NR4A2 (Nurr1), and NR4A3 (Nor-1) are transcription factors of the neuron orphan receptor family that have been linked to the process of negative selection (13). Nr4a1–3 genes have been shown in vivo to be differentially expressed in high-affinity clones and in cells undergoing negative selection in the thymus (12, 14–18). These genes encode for proteins that are largely homologous and have similar expression kinetics upon thymocyte stimulation (14). In addition, they bind to the same DNA motif to induce gene transcription (14, 19–21). Their role in negative selection has been inferred from studies demonstrating massive apoptosis in thymuses from transgenic mice constitutively expressing NR4A1 or NR4A3 (14, 15) and by the inefficient clonal deletion of self-reactive T cells upon expression of a dominant-negative form of NR4A1, which also inhibits NR4A3 (14, 22). Finally, transcriptional analysis of the footprint of negative selection has shown that DP thymocytes that are targeted for clonal deletion increase their expression of Nr4a family mRNAs (12). It was initially proposed that NR4A proteins mediate their role in negative selection by localizing to the mitochondria and converting Bcl-2 to a proapoptotic molecule (16). However, the use of more physiologically relevant models of negative selection failed to confirm an increase in the conversion of Bcl-2 during negative selection (23, 24). In contrast, constructs specifically reducing NR4A1 transcriptional activity reduced its ability to induce negative selection (25).

Early studies have failed to identify a defect of clonal deletion in Nr4a1^−/− mice, and this was thought to result from a functional redundancy with NR4A3 (13, 14). It was through the study of negative selection of MHC class II–restricted TCR transgenic thymocytes that a role for NR4A1 was revealed (26). Other studies have shown that NR4A1 was dispensable for negative selection of MHC class I–restricted TCR transgenic thymocytes to tissue-restricted Ags or ubiquitously expressed Ags (6, 7, 13, 24, 27). However, no study has evaluated the role of NR4A3 during negative selection.

NR4A1 deficiency was also shown to increase the generation of Tregs (26, 27). The latter finding is strengthened by the fact that Nr4a1 is transcribed not only during positive or negative selection but also in cells selected on a strong TCR signal, such as Tregs (17, 28, 29). Another study has observed an increased number of SPCD8 thymocytes in Nr4a1^−/− mice and suggested that NR4A1 inhibited the CD8-determining factor RUNX3, but a similar phenotype was not observed by others (27, 30–32). Thus, the bulk of the current research on the role of NR4A orphan receptors in the process of thymocyte maturation has focused on NR4A1. Whereas it is likely redundant with NR4A1, it is currently not known whether NR4A3 has unique functions in the thymus because studies detailing its specific role in the educational process of T cells are lacking.

We have characterized the pattern of Nr4a3 transcription at different stages of thymocyte maturation and, using NR4A3-deficient mice, determined its importance for proper T cell generation. We demonstrate that NR4A3 has nonredundant functions in the processes involved in sensing strong TCR signals, including negative selection and Treg differentiation.

Nr4a3 knock-in (KI) mice were first described by Ponnio et al. (33) and were a kind gift from Dr. Conneely. Homozygote mice for the Nr4a3 KI locus are deficient in NR4A3 expression and will be referred to as Nr4a3^−/− mice. Nr4a3^+/+, Nr4a3^−/−, Rag1^−/−, OT-I Rag1^−/−, OT-I Rag1^−/−Nr4a3^−/−, B6.SJL, CD45.1.2 (F1 of C57BL/6xB6.SJL), and RIP-mOVA mice (34) were bred at the Maisonneuve-Rosemont Hospital Research Center facility. OT-I mice (35) were crossed to Nr4a3 KI mice to generate OT-I Nr4a3^+/+ and OT-I Nr4a3^−/− mice. Mice were housed in a specific pathogen-free environment and treated in accordance with the Canadian Council on Animal Care guidelines.

Anti-CD8 (53-6.7), -CD4 (GK1.5), -CD3ε (145-2C11) -CD45.2 (104), -CD45.1 (A20), -CD69 (H1.2F3), -CD5 (53-7.3), -CD25 (PC61), -CD62L (MEL-14), -CD44 (IM7), -PD-1 (29F.1A2), -TCRβ (H57-597), -Vα2 (B20.1), -Vβ5 (MR9-4), -CD19 (6D5), -B220 (RA3-6B2), -NK1.1 (PK136), -CD11c (N418), -CD26 (H194-112), -XCR1 (ZET), -SIRPα (P84), CD11b (M1/70), -F4/80 (BM8), and -CD64 (X54-5/7.1) and I-A/I-E (M5/114.15.2) Abs as well as Zombie fixable viability kits were purchased from BioLegend. Anti-CD24 (M1/69), anti-CD124 (mIL4R-M1), and anti–active Caspase-3 (C92-605) were purchased from BD Biosciences. Anti-Foxp3 (FJK-16s), -HELIOS (22F6), and -EOMES (Dan11mag) were purchased from Thermo Fisher Scientific. The CD1d tetramer was obtained through the National Institutes of Health Tetramer Core Facility. Anti-CD16/32 (Fc Block, clone 2.4g2) was purchased from Bio X Cell. Cell surface stainings were performed as previously described (36). For staining of active Caspase-3, cells were fixed following surface stains with BD Cytofix/Cytoperm (BD Biosciences) at 4°C and then washed twice with Perm/Wash (BD Biosciences) prior to intracellular staining for 1 h at room temperature. Otherwise, intracellular stainings were performed using the eBioscience Foxp3 staining buffers (Thermo Fisher Scientific) as per the manufacturer’s protocol. Skin tissue preparations were done as previously described (37). Cells were analyzed on a BD Biosciences FACSCanto I, LSR II, or LSRFortessa X-20 system.

Fluorescein di(β-d-galactopyranoside) (FDG) (Sigma-Aldrich) staining was performed as described (38). Briefly, cells were surface stained and resuspended in PBS. Cells and diluted FDG (7.5 mM) were then incubated for 5 min at 37°C, and 80 μl of warmed FDG was added to cells while gently vortexing. The reaction was stopped by adding 2 ml of ice-cold PBS, and cells were kept on ice for 5 min. After centrifugation, cells were resuspended in PBS 10% horse serum (Thermo Fisher Scientific). Cells were then transferred to a 15°C water bath for 15–20 min to enhance β-galactosidase activity before flow cytometry analysis.

CD45.1.2⁺, RIP-mOVA, or wild-type (WT) recipients were lethally irradiated (12 Gy) and injected with 2 × 10⁶ to 5 × 10⁶ bone marrow cells from Nr4a3^+/+, Nr4a3^−/−, OT-I Nr4a3 WT, or OT-I Nr4a3^−/− donors. In some cases, to delay autoimmunity, a mix of 3% OT-I bone marrow cells with 97% B6.SJL were grafted into RIP-mOVA recipients. In experiments in which OT-I bone marrow cells were grafted, donors were injected i.p. with 100 μg of purified anti-CD8 Ab (clone 2.43; Bio X Cell) on days −2 and −1 before bone marrow harvest to deplete CD8⁺ T cells. Recipients were then injected i.p. on days +1 and +8 with purified anti-Th1.2 Ab (clone 30H12; Bio X Cell) to eliminate any residual mature OT-I cells from the graft. Mice were provided with antibiotic water for 2 wk postinjection.

Total DP^lo (CD4⁺CD8⁺TCRβ⁻CD69^lo), DP^hi (CD4⁺CD8⁺TCRβ⁺CD69^hi), SPCD4 (CD4⁺CD8⁻), and SPCD8 (CD4⁻CD8⁺) thymocytes were sorted directly in TRIzol, and RNA was extracted as instructed by the manufacturer (Life Technologies). RNA was reverse transcribed with SuperScript II with oligonucleotide primers (Life Technologies). SYBR Green–based real-time PCR was performed (Life Technologies), and the Δ cycle threshold (CT) value for each sample was determined by calculating the difference between the CT value of the target and the CT value of the reference gene Hprt. The ΔΔCT value was then determined by subtracting the ΔCT value of the sample from the ΔCT value of a reference sample, as previously described (39). Finally, the relative gene expression of the target was calculated using 2^−ΔΔCT. Primers used in this study were Nr4a1, 5′-GTGCAGTCTGTGGTGACAATGCTT-3′ and 5′-TGTCCACAGGGCAATCCTTGTTTG-3′; Nr4a2, 5′-GCCGAAATCGTTGTCAGTA-3′ and 5′-TTAAACTGTCCGTGCGAA-3′; and Hprt, 5′-CTCCTCAGACCGCTTTTTGC-3′ and 5′-TAACCTGGTTCATCATCGCTAATC-3′.

A total of 1 × 10⁶ sorted SPCD8 (CD4⁻CD8⁺) thymocytes from Nr4a3^+/+ or Nr4a3^−/− mice were adoptively transferred into Rag1^−/− mice. Recipients were euthanized at least 30 d after transfer for analysis.

Statistical analyses for differences between groups were performed using two-tailed Student t test. Welch correction was applied for unequal variances when required. Data are presented as mean ± SEM. A paired t test was used for matched observations. Pearson correlation analysis was used to study the relationship between thymus cellularity or population proportions and time (weeks). Multiple-group comparisons were performed using Kruskal–Wallis ANOVA with Dunn posttest or with repeated measures of ANOVA with Tukey posttest for matched observations in mixed bone marrow chimeras. A value of p < 0.05 was considered statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001).

To gain insight into the thymic expression pattern of NR4A3 (encoded by Nr4a3), we took advantage of the lacZ reported embedded in the Nr4a3 KI locus (33). The activity of the encoded β-galactosidase enzyme can be used to convert the FDG substrate into a fluorogenic compound detectable by flow cytometry and can serve as a surrogate marker for Nr4a3 transcription (Fig. 1A) (33, 38). Using mice heterozygous for the Nr4a3 KI allele, in which no overt phenotype was detected (Supplemental Fig. 1A, 1B), we found that FDG^hi cells at steady state represent ∼2% of the total thymocyte pool (Fig. 1A). These cells were not evenly distributed among the double-negative (DN), DP, SPCD4, and SPCD8 thymocyte subsets. Only 2% of DP cells were FDG^hi, as opposed to 9 and 6% of SPCD4 and SPCD8 thymocytes, respectively (Fig. 1B). It is noteworthy that DN cells that were FDG^hi predominantly expressed cell surface TCRβ, CD5, and PD-1 (Supplemental Fig. 1C), a phenotype associated with diversion from negative selection (40). Therefore, Nr4a3 is only transcribed by a subset of thymocytes.

We further characterized the phenotype of Nr4a3-transcribing cells by comparing FDG^lo cells with FDG^hi DP, SPCD4, and SPCD8 thymocytes. In the DP subset, positively selected thymocytes are defined by increased CD69 and TCRβ-chain expression (41–43). Nr4a3 was preferentially transcribed in postselection DP thymocytes, as ∼50% of FDG^hi DP cells (versus only 6% of FDG^lo cells) displayed elevated expression of CD69 and TCRβ (Fig. 1C). Less than 1% of preselection DP thymocytes (CD69⁻TCRβ⁻) and ∼8% of postselection DP thymocytes were FDG^hi, further supporting the notion that Nr4a3 expression identifies a subset of postselection DP thymocytes (Fig. 1C). In addition, expression of TCR-induced markers CD44, PD-1, and CD5 was also increased in FDG^hi DP thymocytes (Fig. 1D). A similar pattern was observed in the SPCD4 and SPCD8 compartment, where FDG^hi cells showed increased expression of markers typically upregulated following TCR signaling (Fig. 1E). Results were comparable if homozygous KI (Nr4a3^−/−) mice were used to study FDG expression patterns (Supplemental Fig. 1D–F). Interestingly, in the SPCD4 compartment, a subset of FDG⁺ cells expressed CD25 (Fig. 1F), and most SPCD4 CD25⁺ cells, which are likely to become Tregs (44), expressed Nr4a3 (Fig. 1F). This is consistent with published data showing, using an Nr4a3^GFP reporter mouse, that Foxp3 expression is enriched in GFP⁺ thymocytes (45). These results suggest that, from the DP stage onward, Nr4a3 expression in the thymus is associated with TCR signaling. The fact that an increased proportion of SPCD4 cells (and, specifically, CD25⁺ cells; see (Fig. 1B, 1F) over SPCD8 cells were FDG⁺ and that FDG⁺ cells have increased CD5 expression suggests that Nr4a3 expression is particularly associated with a strong TCR signal (17, 28, 29, 46–48).

Because Nr4a3 transcription is associated with a strong TCR signal, it may play a role in thymocyte selection. We therefore studied the phenotype of NR4A3-deficient mice. To do so, we compared the thymuses of WT mice (^+/+) with those of mice homozygous for the KI allele (Nr4a3^−/− mice). We first examined the consequence of NR4A3 deficiency on transcription of the other NR4A family members, Nr4a1 and Nr4a2 (Supplemental Fig. 2A). In sorted thymocyte populations, we observed no statistically different expression levels of Nr4a1, aside from perhaps a trend in DP^hi cells. For Nr4a2, however, there was a slight increase in expression in DP^hi and SPCD8 thymocytes.

Initial examination of the thymus from mice 5–9 wk of age revealed no effect of NR4A3 deficiency. Subset distribution and cell numbers of the DN, DP, SPCD4, and SPCD8 populations were similar in both WT and Nr4a3^−/− mice (Fig. 2A, 2B). However, total thymic cellularity in NR4A3-deficient mice negatively correlated with age for mice 5–15 wk of age (Fig. 2C). Thus, if we included older mice (10–15 wk) in the comparison of cellularity of Nr4a3^+/+ and Nr4a3^−/− thymus, we observed a decrease in total number of DP thymocytes in Nr4a3^−/− mice (Fig. 2D). Although the number of SPCD4 and SPCD8 cells was similar in both groups, we observed, with time, a more-rapid accumulation of an effector-like population (CD44^hiCD62L^lo) in both the SPCD4 and SPCD8 compartment of NR4A3-deficient mice (Fig. 2E).

NR4A1 has been shown to be essential to the development of a CD11b⁻F4/80⁺CD64⁺ macrophage subset in the thymus (49). These macrophages are important for clearance of apoptotic thymocytes, and their absence correlates with accelerated thymic involution in Nr4a1^−/− mice (49). A similar function for NR4A3 in the generation of these macrophages could perhaps provide an explanation for our results in Nr4a3^−/− mice. However, as shown in Supplemental Fig. 2B, 2C, we found no obvious defects in macrophage populations in the thymus of NR4A3-deficient mice. We did, however, note a shift in the balance of thymic conventional dendritic cell 1 (cDC1) and conventional dendritic cell 2 (cDC2) populations (Supplemental Fig. 2D, 2E).

The accumulation of thymic SPCD8 cells with an unconventional activated (innate) phenotype has been observed in mice with mutations compromising TCR signaling, such as Itk^−/−, Rlk^−/−, and Id3^−/− mice (50–53). These innate CD8⁺ thymocytes develop via a cell-extrinsic mechanism that depends on the increased thymic generation of IL-4–secreting NKT cells. Elevated levels of IL-4 then result in increased EOMES expression by SPCD8 thymocytes, which favors an innate T cell phenotype (52–54). Therefore, innate SPCD8 thymocytes are characterized by increased CD44 and EOMES expression as well as sensitivity to IL-4 signals (i.e., increased IL-4R upon exposure to IL-4). In addition, elevated NKT cell numbers should accompany increased numbers of innate CD8⁺ T cells (52–54). However, SPCD8 CD44^hi cells found in Nr4a3^−/− mice did not express EOMES and were not sensitive to IL-4 (Fig. 3A). CD1d-restricted NKT cell numbers, identified using CD1d-tetramer staining, were not increased in NR4A3-deficient mice (Fig. 3B). These results indicate that the activated SPCD8 thymocytes that accumulate in Nr4a3^−/− mice are not similar to the innate thymocytes described by others (50–54). Furthermore, the absence of EOMES expression by Nr4a3^−/− CD44^hiCD62L^lo SPCD8 thymocytes indicate that these cells are not the precursors of self-specific, memory-phenotype CD8 T cells that were recently identified (55). We conclude that the activated CD8⁺ cells that accumulate with time in the thymus of NR4A3-deficient mice are not innate memory or precursor of self-specific, memory-phenotype CD8 T cells.

The capacity of NR4A3 to induce apoptosis upon overexpression and the fact that it is an early effector of the TCR signaling cascade strongly suggest that it has a role in the process of negative selection (14). This is further supported by analysis of the footprint of negative selection, which demonstrates its association with transcription of Nr4a3 (12). We therefore postulated that cells with an activated phenotype seen in Nr4a3^−/− mice resulted from failed negative selection. Phenotypic comparison of SPCD8 thymocytes from WT mice to SPCD8 thymocytes from Nr4a3^−/− mice (either SPCD8 CD44^lo/int cells or SPCD8 CD44^hi cells specifically) revealed that these accumulating activated cells bear a unique signature. They express lower levels of CD24 and TCRβ and increased levels of CD69, CD25, PD-1, and HELIOS when compared with Nr4a3^+/+ SPCD8 and Nr4a3^−/− CD44^lo/int SPCD8 lymphocytes (Fig. 4A). Whereas decreased CD24 levels suggest an advanced maturation stage, the fact that cells with a similar phenotype were not found to be increased in the periphery of Nr4a3^−/− mice suggests that CD44^hiCD62L^lo SPCD8 thymocytes are not likely to be recirculating mature T cells (Supplemental Fig. 2F, 2G). Instead, in the periphery and in the thymus, we found an increase in CD8⁺CD44⁺CD62L⁺ cells (Supplemental Fig. 2H), which could result from the role NR4A3 plays in limiting central memory T cell formation in mature CD8 T cells (56), although we cannot exclude that these are innate memory cells. More likely, although molecular imprinting of negative selection remains somewhat misunderstood, shifts in expression patterns of molecules such as HELIOS, PD-1, CD25, and CD69 may identify thymocytes destined for apoptosis or alternative differentiation because of their strong autoreactivity (11, 12, 17, 28, 29, 42). To further support the idea that SPCD8 CD44^hi cells found in Nr4a3^−/− mice had received a strong TCR signal, we measured levels of CD5 that correlate with signals received by thymocytes during selection (47, 48). As shown in (Fig. 4B, CD5 levels were higher on the CD44^hi SPCD8 thymocytes. Finally, the thymus of lethally irradiated CD45.1.2⁺ recipients reconstituted with a 50:50 mixture of Nr4a3^+/+ and Nr4a3^−/− bone marrow cells demonstrated that the phenotype is cell autonomous, as the increase in CD44^hi SPCD8 thymocytes is only seen in the thymocytes that develop from the Nr4a3^−/− bone marrow graft (Fig. 4C, 4D). Therefore, the impact of NR4A3 deficiency on cDC1 and cDC2 balance noted above could not explain our observations. Note that, as previously published, Nr4a3^−/− cells often have a competitive advantage in mixed bone marrow chimeras (Supplemental Fig. 3A) (37). Importantly, none of these phenotypic differences were observed between Nr4a3^+/+CD45.1⁺ and Nr4a3^+/+CD45.2⁺ control chimeras (Supplemental Fig. 3B). These data support the idea that SPCD8 unconventional thymocytes accumulating in Nr4a3^−/− mice would normally have undergone negative selection but have survived because of the absence of the molecule.

To confirm a defect in negative selection in Nr4a3^−/− mice, we took advantage of a flow cytometry–based assay to measure clonal deletion (6). This assay focuses on thymocytes that have received a TCR signal (TCRβ^hiCD5^hi) and that are undergoing apoptosis (cleaved Caspase-3⁺) to measure clonal deletion, as opposed to cells that are dying by neglect (TCRβ^loCD5^lo active Caspase 3⁺). Indeed, the proportion of cells undergoing clonal deletion in Nr4a3^−/− mice is decreased significantly (Fig. 5A). Decreased clonal deletion was observed in signaled Nr4a3^−/− DP as well as in Nr4a3^−/− SPCD8 thymocytes when compared with controls, indicating that the phenotypic bias in SPCD8 thymocytes could not solely account for the observed differences in the percentage of cleaved Caspase-3⁺ cells (Fig. 5B). Decreased clonal deletion of Nr4a3^−/− thymocytes was also observed in Nr4a3^−/−/Nr4a3^+/+ competitive mixed bone marrow chimeras, but not in WT/WT control chimeras (Fig. 5C, Supplemental Fig. 3C).

To confirm that NR4A3 resulted in escape from negative selection, we generated bone marrow chimeras of lethally irradiated WT or RIP-mOVA recipients reconstituted with Nr4a3^+/+ or Nr4a3^−/− OT-I bone marrow cells. In the presence of the RIP-mOVA transgene, OT-I cells are normally negatively selected (57), but the rapid incidence of diabetes in RIP-mOVA recipients reconstituted with bone marrow from OT-I Nr4a3^−/− mice, but not in other groups, strongly suggests a failure of negative selection when NR4A3 is absent from the hematopoietic system (Fig. 5D). To quantify this, we generated bone marrow chimeras in WT or RIP-mOVA–recipient mice grafted with a mixture of 3% Nr4a3^+/+ or Nr4a3^−/− OT-I (on an RAG1-deficient background) bone marrow cells (CD45.2⁺) mixed with CD45.1⁺ B6.SJL cells. Decreasing the ratio of grafted OT-I cells allowed us to delay appearance of autoimmunity and to phenotype mice 6 wk postirradiation. In RIP-mOVA recipients, we found an increase in OT-I SPCD8 thymocytes when these matured from the Nr4a3^−/− grafted bone marrow cells (Fig. 5E). In addition, whereas all OT-I SPCD8 thymocytes had an activated phenotype in the RIP-mOVA recipients, those deficient in NR4A3 had increased expression of CD69, PD-1, and CD44 (Fig. 5F). The increased frequency of NR4A3-deficient OT-I SPCD8 thymocytes was not due to an increased competitive advantage from OT-I Nr4a3^−/− bone marrow cells, as similar proportions of CD45.2⁺ donor NK1.1⁺ cells were found in all groups (Supplemental Fig. 3D). There was also a lack of CD45.2⁺ donor CD19⁺ cells, as expected, from Rag1^−/− cells (Supplemental Fig. 3D). These results reinforce the fact that NR4A3 deficiency can favor escape from negative selection and imprints an activated phenotype on SPCD8 thymocytes.

To verify whether NR4A3-deficient SPCD8 thymocytes are autoreactive in a polyclonal setting, we sorted these cells from Nr4a3^+/+ and Nr4a3^−/− mice and adoptively transferred them in Rag1^−/− recipients. Cutaneous reactions were observable in 40–50% of recipients of Nr4a3^−/− cells (Supplemental Fig. 3E). When mice were euthanized, at least 30 d posttransfer, there was an increase in CD8⁺ T cells in the skin and skin draining lymph nodes of mice that had received NR4A3-deficient cells (Fig. 5G, 5H, Supplemental Fig. 3F). That there was no such increase in CD8⁺ T cells in the spleen of Rag1^−/− mice (Supplemental Fig. 3G) argues against a general proliferative advantage of Nr4a3^−/− cells and in favor of a localized immune response.

A defect in the molecular mechanism of negative selection may lead to the accumulation of strongly autoreactive cells. These include cells that normally are eliminated by apoptosis as well as cells selected by a strong TCR signal, such as Tregs (2). Given that NR4A1, which shares strong homology with NR4A3, has been shown to regulate Treg differentiation (17, 26, 27, 58), we verified whether the activated cells accumulating in the SPCD4 compartment of Nr4a3^−/− mice were Tregs (normally CD44^hi cells). Nr4a3^−/− mice showed an accumulation of both CD25⁺Foxp3⁻ immature Tregs (44) and CD25⁺Foxp3⁺ Tregs when compared with WT mice (Fig. 6A, 6B). In addition, analysis of CD4⁺Foxp3⁻ cells, to exclude Tregs that express HELIOS, reveals an increase in the expression of HELIOS, a marker of negative selection in SPCD4 thymocytes (Fig. 6C) (11). When we performed mixed bone marrow chimeras, we found that the effect that NR4A3 mediated on Treg differentiation was not T cell intrinsic (Fig. 6D) contrary to what we observed in the SPCD8 compartment. However, we observed a slight increase in Treg frequency in both CD45.1⁺ and CD45.2⁺ SPCD4 thymocytes in the recipient mice reconstituted with a mixture of Nr4a3^+/+ and Nr4a3^−/− hematopoietic cells when compared with control bone marrow chimeras (Fig. 6E). We conclude that hematopoietic-specific NR4A3 deficiency leads to increased survival of SPCD4 thymocytes that should have normally undergone negative selection, as well as increased production of Tregs that are selected following a strong TCR/ligand interaction.

The fact that the NR4A family of orphan nuclear receptors plays a role in negative selection has long been suspected, but the initial difficulties in uncovering a unique phenotype in NR4A1-deficient mice have led to a level of uncertainty concerning the role and importance of these transcription factors in the educational process of thymocytes (13–15). In this study, we show, by studying Nr4a3 transcription patterns and NR4A3-deficient mice, that this molecule has a unique, nonredundant role in the processes of negative selection and Treg differentiation.

Nr4a3 transcription is increased in TCRβ⁺CD69⁺ postselection DP cells and also found in SP thymocytes. Additionally, both in the DP and SP compartment, Nr4a3 gene expression is typically in cells that have increased markers of activation CD44, CD69, and PD-1. Interestingly, PD-1 and CD69 expression at the DP stage have been associated with negative selection (12, 42). This is in accordance with the observation that FDG^hi cells express increased CD5 surface levels, a marker of TCR signal perception (47, 48). Finally, in the SPCD4 cells, Nr4a3 transcription is proportionally higher in CD25⁺ cells, which are likely to be destined to the Treg lineage and therefore were selected on a relatively stronger TCR signal (17, 28, 29). This is consistent with previously published data showing that Foxp3⁺ cells are enriched for NR4A3 expression (45). Using a transgenic mouse expressing GFP from the Nr4a1 locus, Moran et al. (17) have shown that Nr4a1 is expressed during positive and negative selection and its level of expression is proportional to TCR signal strength. They used this system to demonstrate that Treg and invariant NKT cells perceive stronger signals than conventional T cells during development. Similarly, NR4A3 is expressed in thymocytes that have received a TCR signal. However, unlike NR4A1, our Nr4a3 transcriptional reporter is observed only in a subset of postselection DP and SP thymocytes. Combined with higher expression of CD5, this suggests that NR4A3 expression selectively identifies cells that received a strong TCR signal and that should undergo negative selection. The expression pattern of Nr4a3 in DN thymocytes also supports this idea. Pobezinsky et al. (40) have shown that autoreactive DP thymocytes that fail to undergo clonal deletion are diverted into TCRβ⁺CD5⁺PD-1⁺ DN thymocytes. These cells then migrate to the intestine, where they convert to CD8αα⁺ IELs. Given the expression pattern of Nr4a3 in CD4⁻CD8⁻ cells (as measured by FDG expression in Supplemental Fig. 1), it would be interesting to investigate its role in the differentiation of CD8αα⁺ IELs.

Because of previously reported overlapping expression of NR4A1 and NR4A3 and because of the initial difficulties in discerning a unique thymic phenotype in NR4A1-deficient mice, it has been assumed that these homologues have redundant functions. We show that Nr4a3^−/− mice with a polyclonal repertoire are affected by a number of phenotypes in the thymus, something that was not reported for NR4A1-deficient mice. Strikingly, in Nr4a3^−/− mice, we observed a time-dependent accumulation of activated SP thymocytes. In the SPCD8 compartment, activated cells are CD8αβ cells (not shown), excluding that they may be CD8αα⁺ IELs, and do not seem to meet the requirement for innate memory cells. Given their phenotype, there is strong indication that these cells have been activated by a strong TCR signal. It is particularly intriguing that SPCD8 CD44^hi thymocytes express increased levels of HELIOS, which has been shown to be overexpressed in DP thymocytes undergoing negative selection and is a specific marker for negative selection in SPCD4 thymocytes (11, 12). Although the same remains to be demonstrated for SPCD8 thymocytes, it could suggest that these activated cells have received a negatively selecting signal. A failure of negative selection of strongly autoreactive T cells in the absence of NR4A3 is supported by a decreased proportion of TCRβ^hiCD5^hi active Caspase3⁺ cells. Defective negative selection is further supported by accumulation of SPCD8 thymocytes and diabetes induction in the RIP-mOVA OT-I Nr4a3^−/− bone marrow chimeras. The failure of negative selection at the DP and SPCD8 stages may explain the increased Nr4a2 transcription observed in these populations in NR4A3-deficient mice given that the level of expression of NR4As is proportional to TCR signal strength (17, 59).

In the SPCD4 compartment, the accumulating CD44^hi cells are mostly Tregs, suggesting that NR4A3, like NR4A1, is essential to limit this differentiation program (26, 27). Our findings, along with those from Fassett et al. (26) and Hu et al. (27), appear to be in contradiction with those from Sekiya et al. (58), which showed that deficiency of all three members of the nuclear orphan receptor family (NR4A1, NR4A2, and NR4A3) resulted in loss of Tregs. The role of these receptors in Treg differentiation may be complex, playing roles in both limitation of the Treg pool and their generation. The fact that we and Fassett et al. (26) have used germline-deleted mice, whereas Sekiya et al. (58) used a combination of germline and T cell–specific knockout mice, may also account for some of the discrepancies in our results. Another explanation for these contradictory observations is the fact that the increase in Treg differentiation that occurs in the absence of NR4A3 is not T cell autonomous, as observed in our mixed bone marrow chimeras, in which enhanced Treg generation occurred in both Nr4a3^+/+ and Nr4a3^−/− cells. This suggests that NR4A3 deficiency in other hematopoietic cell types influences Treg differentiation. In addition, NR4A1 has been shown to be required for the generation of a subset of thymic CD11b⁻F4/80⁺ macrophages that can clear apoptotic cells, a factor that must be accounted for when studying full NR4A1 knockout mice (49).

Our analysis of SPCD4 and SPCD8 thymocytes from NR4A3-deficient mice supports a nonredundant role for NR4A3 in negative selection, yet some of the data suggest a level of similarity of expression pattern and function of NR4A3 and NR4A1 (13, 14, 16, 25). It would be interesting to verify whether these proteins act in a cascade (one inducing the other) or in tandem, perhaps as a heterodimer (19), which could help better understand how these proteins function. Perhaps more importantly, it is still not clear whether NR4A3 acts to regulate clonal deletion via its proapoptotic function, converting Bcl-2 into a proapoptotic effector (16), or as a transcription factor (25).

One of the striking differences between the role of NR4A1 and role of NR4A3 has been revealed by bone marrow chimera experiments. Indeed, whereas we show that single NR4A3 deficiency is sufficient to induce diabetes in RIP-mOVA mice engrafted with OT-I cells, double NR4A1 and BIM deficiency was required to induce diabetes, which was, in addition, much more delayed when compared with Nr4a3^−/− OT-I bone marrow chimeras (24, 27). One possible explanation is that we have observed defective thymic clonal deletion of SPCD8 thymocytes in the absence of NR4A3, whereas this did not occur in Nr4a1^−/− mice. However, other explanations are possible. We, and others, have shown an important specific role for NR4A3 in the migration of dendritic cells (37, 60). Given the known role these cells play in maintaining immune homeostasis in the periphery, their absence in RIP-mOVA mice grafted with OT-I Nr4a3^−/− bone marrow cells could explain the break in tolerance we observed (61). The fact that BIM is also involved in the biology of DCs might help contextualize the finding that the break in OT-I tolerance induced by NR4A1 and BIM double deficiency is independent from clonal deletion (62, 63).

Perhaps our most intriguing finding is the age-dependent increase of activated (CD44^hiCD62L^lo) SP thymocytes in Nr4a3^−/− mice. As mice age, there is a natural accumulation of recirculating activated mature T cells (64). The fact that the cells observed in Nr4a3^−/− mice are CD24^lo and CD44^hi is consistent with the possibility that there is an accelerated accumulation of recirculating cells rather than a defect in negative selection in these mice. However, several lines of evidence seem to contradict this possibility. First, the observed phenotype in the SPCD8 compartment is T cell autonomous, as only activated Nr4a3^−/− cells are seen in competitive bone marrow chimeras. Second, some of the phenotypes associated with activated SP thymocytes in Nr4a3^−/− mice are observable at the DP stage—particularly, increased CD44 and HELIOS expression (Supplemental Fig. 4). Third, the accumulation of Nr4a3^−/− OT-I SPCD8 thymocytes in the RIP-mOVA bone marrow chimeras is readily observed at 6 wk after grafting. Fourth, activated SPCD8 thymocytes from Nr4a3^−/− mice are enriched for self-reactive cells, as shown following their transfer into Rag1^−/− recipients. Finally, similar cells were not increased in the peripheral lymphoid organs of Nr4a3^−/− mice (Supplemental Fig. 2). Given that this phenotype must be cell autonomous, at least for activated SPCD8 thymocytes, we would propose that cells that have failed negative selection accumulate with time in the thymus. These cells may be retained in the thymus because they continue to express elevated levels of CD69 (65). Alternatively, age-associated changes in selection thresholds may explain the accumulation over time of SPCD8 thymocytes that have failed negative selection. This possibility is reminiscent of recent works showing differences in the selection efficiency in neonates versus adult mice (66–68).

In conclusion, phenotypic analysis of Nr4a3-transcribing cells as well as of NR4A3-deficient mice has revealed a unique function for this receptor in the processes of negative selection and Treg differentiation. It is important to understand the mechanisms by which the T cell autoreactive repertoire is limited to better study certain types of autoimmune diseases.

We acknowledge the members of the laboratory for helpful discussions. We acknowledge Martine Dupuis, flow core manager, the animal care facility for mice husbandry; Dr. O. M. Conneely for NR4A3-deficient mice; and Dr. S. Lesage for providing Rag1^−/− mice. We acknowledge Dr. S. Lesage and Dr. T. Baldwin for critical reading of the manuscript.

The authors have no financial conflicts of interest.